xref: /linux/net/ipv6/addrconf.c (revision eed4edda910fe34dfae8c6bfbcf57f4593a54295)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 Address [auto]configuration
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  */
10 
11 /*
12  *	Changes:
13  *
14  *	Janos Farkas			:	delete timer on ifdown
15  *	<chexum@bankinf.banki.hu>
16  *	Andi Kleen			:	kill double kfree on module
17  *						unload.
18  *	Maciej W. Rozycki		:	FDDI support
19  *	sekiya@USAGI			:	Don't send too many RS
20  *						packets.
21  *	yoshfuji@USAGI			:       Fixed interval between DAD
22  *						packets.
23  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
24  *						address validation timer.
25  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
26  *						support.
27  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
28  *						address on a same interface.
29  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
30  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
31  *						seq_file.
32  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
33  *						selection; consider scope,
34  *						status etc.
35  */
36 
37 #define pr_fmt(fmt) "IPv6: " fmt
38 
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65 
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
69 
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
88 
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
92 #include <linux/ioam6.h>
93 
94 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
95 
96 #define IPV6_MAX_STRLEN \
97 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98 
99 static inline u32 cstamp_delta(unsigned long cstamp)
100 {
101 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102 }
103 
104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 {
106 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 	u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
108 	do_div(tmp, 1000000);
109 	return (s32)tmp;
110 }
111 
112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 {
114 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 	u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116 	do_div(tmp, 1000000);
117 	if ((s32)tmp > mrt) {
118 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 		tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
120 		do_div(tmp, 1000000);
121 	}
122 	return (s32)tmp;
123 }
124 
125 #ifdef CONFIG_SYSCTL
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 #else
129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 {
131 	return 0;
132 }
133 
134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135 {
136 }
137 #endif
138 
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140 
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 					u8 dad_count,
145 					const struct inet6_dev *idev);
146 
147 #define IN6_ADDR_HSIZE_SHIFT	8
148 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
149 
150 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
152 
153 static struct workqueue_struct *addrconf_wq;
154 
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157 
158 static void addrconf_type_change(struct net_device *dev,
159 				 unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
161 
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163 						  int plen,
164 						  const struct net_device *dev,
165 						  u32 flags, u32 noflags,
166 						  bool no_gw);
167 
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171 				   bool send_na);
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176 
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 				struct prefix_info *pinfo);
179 
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
181 	.forwarding		= 0,
182 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
183 	.mtu6			= IPV6_MIN_MTU,
184 	.accept_ra		= 1,
185 	.accept_redirects	= 1,
186 	.autoconf		= 1,
187 	.force_mld_version	= 0,
188 	.mldv1_unsolicited_report_interval = 10 * HZ,
189 	.mldv2_unsolicited_report_interval = HZ,
190 	.dad_transmits		= 1,
191 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
192 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
193 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
195 	.use_tempaddr		= 0,
196 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
197 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
198 	.regen_max_retry	= REGEN_MAX_RETRY,
199 	.max_desync_factor	= MAX_DESYNC_FACTOR,
200 	.max_addresses		= IPV6_MAX_ADDRESSES,
201 	.accept_ra_defrtr	= 1,
202 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
203 	.accept_ra_from_local	= 0,
204 	.accept_ra_min_hop_limit= 1,
205 	.accept_ra_min_lft	= 0,
206 	.accept_ra_pinfo	= 1,
207 #ifdef CONFIG_IPV6_ROUTER_PREF
208 	.accept_ra_rtr_pref	= 1,
209 	.rtr_probe_interval	= 60 * HZ,
210 #ifdef CONFIG_IPV6_ROUTE_INFO
211 	.accept_ra_rt_info_min_plen = 0,
212 	.accept_ra_rt_info_max_plen = 0,
213 #endif
214 #endif
215 	.proxy_ndp		= 0,
216 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
217 	.disable_ipv6		= 0,
218 	.accept_dad		= 0,
219 	.suppress_frag_ndisc	= 1,
220 	.accept_ra_mtu		= 1,
221 	.stable_secret		= {
222 		.initialized = false,
223 	},
224 	.use_oif_addrs_only	= 0,
225 	.ignore_routes_with_linkdown = 0,
226 	.keep_addr_on_down	= 0,
227 	.seg6_enabled		= 0,
228 #ifdef CONFIG_IPV6_SEG6_HMAC
229 	.seg6_require_hmac	= 0,
230 #endif
231 	.enhanced_dad           = 1,
232 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
233 	.disable_policy		= 0,
234 	.rpl_seg_enabled	= 0,
235 	.ioam6_enabled		= 0,
236 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
237 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
238 	.ndisc_evict_nocarrier	= 1,
239 	.ra_honor_pio_life	= 0,
240 };
241 
242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
243 	.forwarding		= 0,
244 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
245 	.mtu6			= IPV6_MIN_MTU,
246 	.accept_ra		= 1,
247 	.accept_redirects	= 1,
248 	.autoconf		= 1,
249 	.force_mld_version	= 0,
250 	.mldv1_unsolicited_report_interval = 10 * HZ,
251 	.mldv2_unsolicited_report_interval = HZ,
252 	.dad_transmits		= 1,
253 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
254 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
255 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
256 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
257 	.use_tempaddr		= 0,
258 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
259 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
260 	.regen_max_retry	= REGEN_MAX_RETRY,
261 	.max_desync_factor	= MAX_DESYNC_FACTOR,
262 	.max_addresses		= IPV6_MAX_ADDRESSES,
263 	.accept_ra_defrtr	= 1,
264 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
265 	.accept_ra_from_local	= 0,
266 	.accept_ra_min_hop_limit= 1,
267 	.accept_ra_min_lft	= 0,
268 	.accept_ra_pinfo	= 1,
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270 	.accept_ra_rtr_pref	= 1,
271 	.rtr_probe_interval	= 60 * HZ,
272 #ifdef CONFIG_IPV6_ROUTE_INFO
273 	.accept_ra_rt_info_min_plen = 0,
274 	.accept_ra_rt_info_max_plen = 0,
275 #endif
276 #endif
277 	.proxy_ndp		= 0,
278 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
279 	.disable_ipv6		= 0,
280 	.accept_dad		= 1,
281 	.suppress_frag_ndisc	= 1,
282 	.accept_ra_mtu		= 1,
283 	.stable_secret		= {
284 		.initialized = false,
285 	},
286 	.use_oif_addrs_only	= 0,
287 	.ignore_routes_with_linkdown = 0,
288 	.keep_addr_on_down	= 0,
289 	.seg6_enabled		= 0,
290 #ifdef CONFIG_IPV6_SEG6_HMAC
291 	.seg6_require_hmac	= 0,
292 #endif
293 	.enhanced_dad           = 1,
294 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
295 	.disable_policy		= 0,
296 	.rpl_seg_enabled	= 0,
297 	.ioam6_enabled		= 0,
298 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
299 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
300 	.ndisc_evict_nocarrier	= 1,
301 	.ra_honor_pio_life	= 0,
302 };
303 
304 /* Check if link is ready: is it up and is a valid qdisc available */
305 static inline bool addrconf_link_ready(const struct net_device *dev)
306 {
307 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
308 }
309 
310 static void addrconf_del_rs_timer(struct inet6_dev *idev)
311 {
312 	if (del_timer(&idev->rs_timer))
313 		__in6_dev_put(idev);
314 }
315 
316 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
317 {
318 	if (cancel_delayed_work(&ifp->dad_work))
319 		__in6_ifa_put(ifp);
320 }
321 
322 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
323 				  unsigned long when)
324 {
325 	if (!mod_timer(&idev->rs_timer, jiffies + when))
326 		in6_dev_hold(idev);
327 }
328 
329 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
330 				   unsigned long delay)
331 {
332 	in6_ifa_hold(ifp);
333 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
334 		in6_ifa_put(ifp);
335 }
336 
337 static int snmp6_alloc_dev(struct inet6_dev *idev)
338 {
339 	int i;
340 
341 	idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
342 	if (!idev->stats.ipv6)
343 		goto err_ip;
344 
345 	for_each_possible_cpu(i) {
346 		struct ipstats_mib *addrconf_stats;
347 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
348 		u64_stats_init(&addrconf_stats->syncp);
349 	}
350 
351 
352 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
353 					GFP_KERNEL);
354 	if (!idev->stats.icmpv6dev)
355 		goto err_icmp;
356 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
357 					   GFP_KERNEL_ACCOUNT);
358 	if (!idev->stats.icmpv6msgdev)
359 		goto err_icmpmsg;
360 
361 	return 0;
362 
363 err_icmpmsg:
364 	kfree(idev->stats.icmpv6dev);
365 err_icmp:
366 	free_percpu(idev->stats.ipv6);
367 err_ip:
368 	return -ENOMEM;
369 }
370 
371 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
372 {
373 	struct inet6_dev *ndev;
374 	int err = -ENOMEM;
375 
376 	ASSERT_RTNL();
377 
378 	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
379 		return ERR_PTR(-EINVAL);
380 
381 	ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
382 	if (!ndev)
383 		return ERR_PTR(err);
384 
385 	rwlock_init(&ndev->lock);
386 	ndev->dev = dev;
387 	INIT_LIST_HEAD(&ndev->addr_list);
388 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
389 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
390 
391 	if (ndev->cnf.stable_secret.initialized)
392 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
393 
394 	ndev->cnf.mtu6 = dev->mtu;
395 	ndev->ra_mtu = 0;
396 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
397 	if (!ndev->nd_parms) {
398 		kfree(ndev);
399 		return ERR_PTR(err);
400 	}
401 	if (ndev->cnf.forwarding)
402 		dev_disable_lro(dev);
403 	/* We refer to the device */
404 	netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
405 
406 	if (snmp6_alloc_dev(ndev) < 0) {
407 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
408 			   __func__);
409 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
410 		netdev_put(dev, &ndev->dev_tracker);
411 		kfree(ndev);
412 		return ERR_PTR(err);
413 	}
414 
415 	if (dev != blackhole_netdev) {
416 		if (snmp6_register_dev(ndev) < 0) {
417 			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
418 				   __func__, dev->name);
419 			goto err_release;
420 		}
421 	}
422 	/* One reference from device. */
423 	refcount_set(&ndev->refcnt, 1);
424 
425 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
426 		ndev->cnf.accept_dad = -1;
427 
428 #if IS_ENABLED(CONFIG_IPV6_SIT)
429 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
430 		pr_info("%s: Disabled Multicast RS\n", dev->name);
431 		ndev->cnf.rtr_solicits = 0;
432 	}
433 #endif
434 
435 	INIT_LIST_HEAD(&ndev->tempaddr_list);
436 	ndev->desync_factor = U32_MAX;
437 	if ((dev->flags&IFF_LOOPBACK) ||
438 	    dev->type == ARPHRD_TUNNEL ||
439 	    dev->type == ARPHRD_TUNNEL6 ||
440 	    dev->type == ARPHRD_SIT ||
441 	    dev->type == ARPHRD_NONE) {
442 		ndev->cnf.use_tempaddr = -1;
443 	}
444 
445 	ndev->token = in6addr_any;
446 
447 	if (netif_running(dev) && addrconf_link_ready(dev))
448 		ndev->if_flags |= IF_READY;
449 
450 	ipv6_mc_init_dev(ndev);
451 	ndev->tstamp = jiffies;
452 	if (dev != blackhole_netdev) {
453 		err = addrconf_sysctl_register(ndev);
454 		if (err) {
455 			ipv6_mc_destroy_dev(ndev);
456 			snmp6_unregister_dev(ndev);
457 			goto err_release;
458 		}
459 	}
460 	/* protected by rtnl_lock */
461 	rcu_assign_pointer(dev->ip6_ptr, ndev);
462 
463 	if (dev != blackhole_netdev) {
464 		/* Join interface-local all-node multicast group */
465 		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
466 
467 		/* Join all-node multicast group */
468 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
469 
470 		/* Join all-router multicast group if forwarding is set */
471 		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
472 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
473 	}
474 	return ndev;
475 
476 err_release:
477 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
478 	ndev->dead = 1;
479 	in6_dev_finish_destroy(ndev);
480 	return ERR_PTR(err);
481 }
482 
483 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
484 {
485 	struct inet6_dev *idev;
486 
487 	ASSERT_RTNL();
488 
489 	idev = __in6_dev_get(dev);
490 	if (!idev) {
491 		idev = ipv6_add_dev(dev);
492 		if (IS_ERR(idev))
493 			return idev;
494 	}
495 
496 	if (dev->flags&IFF_UP)
497 		ipv6_mc_up(idev);
498 	return idev;
499 }
500 
501 static int inet6_netconf_msgsize_devconf(int type)
502 {
503 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
504 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
505 	bool all = false;
506 
507 	if (type == NETCONFA_ALL)
508 		all = true;
509 
510 	if (all || type == NETCONFA_FORWARDING)
511 		size += nla_total_size(4);
512 #ifdef CONFIG_IPV6_MROUTE
513 	if (all || type == NETCONFA_MC_FORWARDING)
514 		size += nla_total_size(4);
515 #endif
516 	if (all || type == NETCONFA_PROXY_NEIGH)
517 		size += nla_total_size(4);
518 
519 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
520 		size += nla_total_size(4);
521 
522 	return size;
523 }
524 
525 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
526 				      struct ipv6_devconf *devconf, u32 portid,
527 				      u32 seq, int event, unsigned int flags,
528 				      int type)
529 {
530 	struct nlmsghdr  *nlh;
531 	struct netconfmsg *ncm;
532 	bool all = false;
533 
534 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
535 			flags);
536 	if (!nlh)
537 		return -EMSGSIZE;
538 
539 	if (type == NETCONFA_ALL)
540 		all = true;
541 
542 	ncm = nlmsg_data(nlh);
543 	ncm->ncm_family = AF_INET6;
544 
545 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
546 		goto nla_put_failure;
547 
548 	if (!devconf)
549 		goto out;
550 
551 	if ((all || type == NETCONFA_FORWARDING) &&
552 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
553 		goto nla_put_failure;
554 #ifdef CONFIG_IPV6_MROUTE
555 	if ((all || type == NETCONFA_MC_FORWARDING) &&
556 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
557 			atomic_read(&devconf->mc_forwarding)) < 0)
558 		goto nla_put_failure;
559 #endif
560 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
561 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
562 		goto nla_put_failure;
563 
564 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
565 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
566 			devconf->ignore_routes_with_linkdown) < 0)
567 		goto nla_put_failure;
568 
569 out:
570 	nlmsg_end(skb, nlh);
571 	return 0;
572 
573 nla_put_failure:
574 	nlmsg_cancel(skb, nlh);
575 	return -EMSGSIZE;
576 }
577 
578 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
579 				  int ifindex, struct ipv6_devconf *devconf)
580 {
581 	struct sk_buff *skb;
582 	int err = -ENOBUFS;
583 
584 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
585 	if (!skb)
586 		goto errout;
587 
588 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
589 					 event, 0, type);
590 	if (err < 0) {
591 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
592 		WARN_ON(err == -EMSGSIZE);
593 		kfree_skb(skb);
594 		goto errout;
595 	}
596 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
597 	return;
598 errout:
599 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
600 }
601 
602 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
603 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
604 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
605 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
606 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
607 };
608 
609 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
610 				       const struct nlmsghdr *nlh,
611 				       struct nlattr **tb,
612 				       struct netlink_ext_ack *extack)
613 {
614 	int i, err;
615 
616 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
617 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
618 		return -EINVAL;
619 	}
620 
621 	if (!netlink_strict_get_check(skb))
622 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
623 					      tb, NETCONFA_MAX,
624 					      devconf_ipv6_policy, extack);
625 
626 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
627 					    tb, NETCONFA_MAX,
628 					    devconf_ipv6_policy, extack);
629 	if (err)
630 		return err;
631 
632 	for (i = 0; i <= NETCONFA_MAX; i++) {
633 		if (!tb[i])
634 			continue;
635 
636 		switch (i) {
637 		case NETCONFA_IFINDEX:
638 			break;
639 		default:
640 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
641 			return -EINVAL;
642 		}
643 	}
644 
645 	return 0;
646 }
647 
648 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
649 				     struct nlmsghdr *nlh,
650 				     struct netlink_ext_ack *extack)
651 {
652 	struct net *net = sock_net(in_skb->sk);
653 	struct nlattr *tb[NETCONFA_MAX+1];
654 	struct inet6_dev *in6_dev = NULL;
655 	struct net_device *dev = NULL;
656 	struct sk_buff *skb;
657 	struct ipv6_devconf *devconf;
658 	int ifindex;
659 	int err;
660 
661 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
662 	if (err < 0)
663 		return err;
664 
665 	if (!tb[NETCONFA_IFINDEX])
666 		return -EINVAL;
667 
668 	err = -EINVAL;
669 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
670 	switch (ifindex) {
671 	case NETCONFA_IFINDEX_ALL:
672 		devconf = net->ipv6.devconf_all;
673 		break;
674 	case NETCONFA_IFINDEX_DEFAULT:
675 		devconf = net->ipv6.devconf_dflt;
676 		break;
677 	default:
678 		dev = dev_get_by_index(net, ifindex);
679 		if (!dev)
680 			return -EINVAL;
681 		in6_dev = in6_dev_get(dev);
682 		if (!in6_dev)
683 			goto errout;
684 		devconf = &in6_dev->cnf;
685 		break;
686 	}
687 
688 	err = -ENOBUFS;
689 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
690 	if (!skb)
691 		goto errout;
692 
693 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
694 					 NETLINK_CB(in_skb).portid,
695 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
696 					 NETCONFA_ALL);
697 	if (err < 0) {
698 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
699 		WARN_ON(err == -EMSGSIZE);
700 		kfree_skb(skb);
701 		goto errout;
702 	}
703 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
704 errout:
705 	if (in6_dev)
706 		in6_dev_put(in6_dev);
707 	dev_put(dev);
708 	return err;
709 }
710 
711 /* Combine dev_addr_genid and dev_base_seq to detect changes.
712  */
713 static u32 inet6_base_seq(const struct net *net)
714 {
715 	u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
716 		  net->dev_base_seq;
717 
718 	/* Must not return 0 (see nl_dump_check_consistent()).
719 	 * Chose a value far away from 0.
720 	 */
721 	if (!res)
722 		res = 0x80000000;
723 	return res;
724 }
725 
726 
727 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
728 				      struct netlink_callback *cb)
729 {
730 	const struct nlmsghdr *nlh = cb->nlh;
731 	struct net *net = sock_net(skb->sk);
732 	int h, s_h;
733 	int idx, s_idx;
734 	struct net_device *dev;
735 	struct inet6_dev *idev;
736 	struct hlist_head *head;
737 
738 	if (cb->strict_check) {
739 		struct netlink_ext_ack *extack = cb->extack;
740 		struct netconfmsg *ncm;
741 
742 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
743 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
744 			return -EINVAL;
745 		}
746 
747 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
748 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
749 			return -EINVAL;
750 		}
751 	}
752 
753 	s_h = cb->args[0];
754 	s_idx = idx = cb->args[1];
755 
756 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
757 		idx = 0;
758 		head = &net->dev_index_head[h];
759 		rcu_read_lock();
760 		cb->seq = inet6_base_seq(net);
761 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
762 			if (idx < s_idx)
763 				goto cont;
764 			idev = __in6_dev_get(dev);
765 			if (!idev)
766 				goto cont;
767 
768 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
769 						       &idev->cnf,
770 						       NETLINK_CB(cb->skb).portid,
771 						       nlh->nlmsg_seq,
772 						       RTM_NEWNETCONF,
773 						       NLM_F_MULTI,
774 						       NETCONFA_ALL) < 0) {
775 				rcu_read_unlock();
776 				goto done;
777 			}
778 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
779 cont:
780 			idx++;
781 		}
782 		rcu_read_unlock();
783 	}
784 	if (h == NETDEV_HASHENTRIES) {
785 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
786 					       net->ipv6.devconf_all,
787 					       NETLINK_CB(cb->skb).portid,
788 					       nlh->nlmsg_seq,
789 					       RTM_NEWNETCONF, NLM_F_MULTI,
790 					       NETCONFA_ALL) < 0)
791 			goto done;
792 		else
793 			h++;
794 	}
795 	if (h == NETDEV_HASHENTRIES + 1) {
796 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
797 					       net->ipv6.devconf_dflt,
798 					       NETLINK_CB(cb->skb).portid,
799 					       nlh->nlmsg_seq,
800 					       RTM_NEWNETCONF, NLM_F_MULTI,
801 					       NETCONFA_ALL) < 0)
802 			goto done;
803 		else
804 			h++;
805 	}
806 done:
807 	cb->args[0] = h;
808 	cb->args[1] = idx;
809 
810 	return skb->len;
811 }
812 
813 #ifdef CONFIG_SYSCTL
814 static void dev_forward_change(struct inet6_dev *idev)
815 {
816 	struct net_device *dev;
817 	struct inet6_ifaddr *ifa;
818 	LIST_HEAD(tmp_addr_list);
819 
820 	if (!idev)
821 		return;
822 	dev = idev->dev;
823 	if (idev->cnf.forwarding)
824 		dev_disable_lro(dev);
825 	if (dev->flags & IFF_MULTICAST) {
826 		if (idev->cnf.forwarding) {
827 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
828 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
829 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
830 		} else {
831 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
832 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
833 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
834 		}
835 	}
836 
837 	read_lock_bh(&idev->lock);
838 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
839 		if (ifa->flags&IFA_F_TENTATIVE)
840 			continue;
841 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
842 	}
843 	read_unlock_bh(&idev->lock);
844 
845 	while (!list_empty(&tmp_addr_list)) {
846 		ifa = list_first_entry(&tmp_addr_list,
847 				       struct inet6_ifaddr, if_list_aux);
848 		list_del(&ifa->if_list_aux);
849 		if (idev->cnf.forwarding)
850 			addrconf_join_anycast(ifa);
851 		else
852 			addrconf_leave_anycast(ifa);
853 	}
854 
855 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
856 				     NETCONFA_FORWARDING,
857 				     dev->ifindex, &idev->cnf);
858 }
859 
860 
861 static void addrconf_forward_change(struct net *net, __s32 newf)
862 {
863 	struct net_device *dev;
864 	struct inet6_dev *idev;
865 
866 	for_each_netdev(net, dev) {
867 		idev = __in6_dev_get(dev);
868 		if (idev) {
869 			int changed = (!idev->cnf.forwarding) ^ (!newf);
870 			idev->cnf.forwarding = newf;
871 			if (changed)
872 				dev_forward_change(idev);
873 		}
874 	}
875 }
876 
877 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
878 {
879 	struct net *net;
880 	int old;
881 
882 	if (!rtnl_trylock())
883 		return restart_syscall();
884 
885 	net = (struct net *)table->extra2;
886 	old = *p;
887 	*p = newf;
888 
889 	if (p == &net->ipv6.devconf_dflt->forwarding) {
890 		if ((!newf) ^ (!old))
891 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
892 						     NETCONFA_FORWARDING,
893 						     NETCONFA_IFINDEX_DEFAULT,
894 						     net->ipv6.devconf_dflt);
895 		rtnl_unlock();
896 		return 0;
897 	}
898 
899 	if (p == &net->ipv6.devconf_all->forwarding) {
900 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
901 
902 		net->ipv6.devconf_dflt->forwarding = newf;
903 		if ((!newf) ^ (!old_dflt))
904 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
905 						     NETCONFA_FORWARDING,
906 						     NETCONFA_IFINDEX_DEFAULT,
907 						     net->ipv6.devconf_dflt);
908 
909 		addrconf_forward_change(net, newf);
910 		if ((!newf) ^ (!old))
911 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
912 						     NETCONFA_FORWARDING,
913 						     NETCONFA_IFINDEX_ALL,
914 						     net->ipv6.devconf_all);
915 	} else if ((!newf) ^ (!old))
916 		dev_forward_change((struct inet6_dev *)table->extra1);
917 	rtnl_unlock();
918 
919 	if (newf)
920 		rt6_purge_dflt_routers(net);
921 	return 1;
922 }
923 
924 static void addrconf_linkdown_change(struct net *net, __s32 newf)
925 {
926 	struct net_device *dev;
927 	struct inet6_dev *idev;
928 
929 	for_each_netdev(net, dev) {
930 		idev = __in6_dev_get(dev);
931 		if (idev) {
932 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
933 
934 			idev->cnf.ignore_routes_with_linkdown = newf;
935 			if (changed)
936 				inet6_netconf_notify_devconf(dev_net(dev),
937 							     RTM_NEWNETCONF,
938 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
939 							     dev->ifindex,
940 							     &idev->cnf);
941 		}
942 	}
943 }
944 
945 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
946 {
947 	struct net *net;
948 	int old;
949 
950 	if (!rtnl_trylock())
951 		return restart_syscall();
952 
953 	net = (struct net *)table->extra2;
954 	old = *p;
955 	*p = newf;
956 
957 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
958 		if ((!newf) ^ (!old))
959 			inet6_netconf_notify_devconf(net,
960 						     RTM_NEWNETCONF,
961 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
962 						     NETCONFA_IFINDEX_DEFAULT,
963 						     net->ipv6.devconf_dflt);
964 		rtnl_unlock();
965 		return 0;
966 	}
967 
968 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
969 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
970 		addrconf_linkdown_change(net, newf);
971 		if ((!newf) ^ (!old))
972 			inet6_netconf_notify_devconf(net,
973 						     RTM_NEWNETCONF,
974 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
975 						     NETCONFA_IFINDEX_ALL,
976 						     net->ipv6.devconf_all);
977 	}
978 	rtnl_unlock();
979 
980 	return 1;
981 }
982 
983 #endif
984 
985 /* Nobody refers to this ifaddr, destroy it */
986 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
987 {
988 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
989 
990 #ifdef NET_REFCNT_DEBUG
991 	pr_debug("%s\n", __func__);
992 #endif
993 
994 	in6_dev_put(ifp->idev);
995 
996 	if (cancel_delayed_work(&ifp->dad_work))
997 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
998 			  ifp);
999 
1000 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
1001 		pr_warn("Freeing alive inet6 address %p\n", ifp);
1002 		return;
1003 	}
1004 
1005 	kfree_rcu(ifp, rcu);
1006 }
1007 
1008 static void
1009 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1010 {
1011 	struct list_head *p;
1012 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1013 
1014 	/*
1015 	 * Each device address list is sorted in order of scope -
1016 	 * global before linklocal.
1017 	 */
1018 	list_for_each(p, &idev->addr_list) {
1019 		struct inet6_ifaddr *ifa
1020 			= list_entry(p, struct inet6_ifaddr, if_list);
1021 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1022 			break;
1023 	}
1024 
1025 	list_add_tail_rcu(&ifp->if_list, p);
1026 }
1027 
1028 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1029 {
1030 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1031 
1032 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1033 }
1034 
1035 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1036 			       struct net_device *dev, unsigned int hash)
1037 {
1038 	struct inet6_ifaddr *ifp;
1039 
1040 	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1041 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1042 			if (!dev || ifp->idev->dev == dev)
1043 				return true;
1044 		}
1045 	}
1046 	return false;
1047 }
1048 
1049 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1050 {
1051 	struct net *net = dev_net(dev);
1052 	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1053 	int err = 0;
1054 
1055 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1056 
1057 	/* Ignore adding duplicate addresses on an interface */
1058 	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1059 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1060 		err = -EEXIST;
1061 	} else {
1062 		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1063 	}
1064 
1065 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1066 
1067 	return err;
1068 }
1069 
1070 /* On success it returns ifp with increased reference count */
1071 
1072 static struct inet6_ifaddr *
1073 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1074 	      bool can_block, struct netlink_ext_ack *extack)
1075 {
1076 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1077 	int addr_type = ipv6_addr_type(cfg->pfx);
1078 	struct net *net = dev_net(idev->dev);
1079 	struct inet6_ifaddr *ifa = NULL;
1080 	struct fib6_info *f6i = NULL;
1081 	int err = 0;
1082 
1083 	if (addr_type == IPV6_ADDR_ANY) {
1084 		NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1085 		return ERR_PTR(-EADDRNOTAVAIL);
1086 	} else if (addr_type & IPV6_ADDR_MULTICAST &&
1087 		   !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1088 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1089 		return ERR_PTR(-EADDRNOTAVAIL);
1090 	} else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1091 		   !netif_is_l3_master(idev->dev) &&
1092 		   addr_type & IPV6_ADDR_LOOPBACK) {
1093 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1094 		return ERR_PTR(-EADDRNOTAVAIL);
1095 	}
1096 
1097 	if (idev->dead) {
1098 		NL_SET_ERR_MSG_MOD(extack, "device is going away");
1099 		err = -ENODEV;
1100 		goto out;
1101 	}
1102 
1103 	if (idev->cnf.disable_ipv6) {
1104 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1105 		err = -EACCES;
1106 		goto out;
1107 	}
1108 
1109 	/* validator notifier needs to be blocking;
1110 	 * do not call in atomic context
1111 	 */
1112 	if (can_block) {
1113 		struct in6_validator_info i6vi = {
1114 			.i6vi_addr = *cfg->pfx,
1115 			.i6vi_dev = idev,
1116 			.extack = extack,
1117 		};
1118 
1119 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1120 		err = notifier_to_errno(err);
1121 		if (err < 0)
1122 			goto out;
1123 	}
1124 
1125 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1126 	if (!ifa) {
1127 		err = -ENOBUFS;
1128 		goto out;
1129 	}
1130 
1131 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1132 	if (IS_ERR(f6i)) {
1133 		err = PTR_ERR(f6i);
1134 		f6i = NULL;
1135 		goto out;
1136 	}
1137 
1138 	neigh_parms_data_state_setall(idev->nd_parms);
1139 
1140 	ifa->addr = *cfg->pfx;
1141 	if (cfg->peer_pfx)
1142 		ifa->peer_addr = *cfg->peer_pfx;
1143 
1144 	spin_lock_init(&ifa->lock);
1145 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1146 	INIT_HLIST_NODE(&ifa->addr_lst);
1147 	ifa->scope = cfg->scope;
1148 	ifa->prefix_len = cfg->plen;
1149 	ifa->rt_priority = cfg->rt_priority;
1150 	ifa->flags = cfg->ifa_flags;
1151 	ifa->ifa_proto = cfg->ifa_proto;
1152 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1153 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1154 		ifa->flags |= IFA_F_TENTATIVE;
1155 	ifa->valid_lft = cfg->valid_lft;
1156 	ifa->prefered_lft = cfg->preferred_lft;
1157 	ifa->cstamp = ifa->tstamp = jiffies;
1158 	ifa->tokenized = false;
1159 
1160 	ifa->rt = f6i;
1161 
1162 	ifa->idev = idev;
1163 	in6_dev_hold(idev);
1164 
1165 	/* For caller */
1166 	refcount_set(&ifa->refcnt, 1);
1167 
1168 	rcu_read_lock();
1169 
1170 	err = ipv6_add_addr_hash(idev->dev, ifa);
1171 	if (err < 0) {
1172 		rcu_read_unlock();
1173 		goto out;
1174 	}
1175 
1176 	write_lock_bh(&idev->lock);
1177 
1178 	/* Add to inet6_dev unicast addr list. */
1179 	ipv6_link_dev_addr(idev, ifa);
1180 
1181 	if (ifa->flags&IFA_F_TEMPORARY) {
1182 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1183 		in6_ifa_hold(ifa);
1184 	}
1185 
1186 	in6_ifa_hold(ifa);
1187 	write_unlock_bh(&idev->lock);
1188 
1189 	rcu_read_unlock();
1190 
1191 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1192 out:
1193 	if (unlikely(err < 0)) {
1194 		fib6_info_release(f6i);
1195 
1196 		if (ifa) {
1197 			if (ifa->idev)
1198 				in6_dev_put(ifa->idev);
1199 			kfree(ifa);
1200 		}
1201 		ifa = ERR_PTR(err);
1202 	}
1203 
1204 	return ifa;
1205 }
1206 
1207 enum cleanup_prefix_rt_t {
1208 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1209 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1210 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1211 };
1212 
1213 /*
1214  * Check, whether the prefix for ifp would still need a prefix route
1215  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1216  * constants.
1217  *
1218  * 1) we don't purge prefix if address was not permanent.
1219  *    prefix is managed by its own lifetime.
1220  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1221  * 3) if there are no addresses, delete prefix.
1222  * 4) if there are still other permanent address(es),
1223  *    corresponding prefix is still permanent.
1224  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1225  *    don't purge the prefix, assume user space is managing it.
1226  * 6) otherwise, update prefix lifetime to the
1227  *    longest valid lifetime among the corresponding
1228  *    addresses on the device.
1229  *    Note: subsequent RA will update lifetime.
1230  **/
1231 static enum cleanup_prefix_rt_t
1232 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1233 {
1234 	struct inet6_ifaddr *ifa;
1235 	struct inet6_dev *idev = ifp->idev;
1236 	unsigned long lifetime;
1237 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1238 
1239 	*expires = jiffies;
1240 
1241 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1242 		if (ifa == ifp)
1243 			continue;
1244 		if (ifa->prefix_len != ifp->prefix_len ||
1245 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1246 				       ifp->prefix_len))
1247 			continue;
1248 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1249 			return CLEANUP_PREFIX_RT_NOP;
1250 
1251 		action = CLEANUP_PREFIX_RT_EXPIRE;
1252 
1253 		spin_lock(&ifa->lock);
1254 
1255 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1256 		/*
1257 		 * Note: Because this address is
1258 		 * not permanent, lifetime <
1259 		 * LONG_MAX / HZ here.
1260 		 */
1261 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1262 			*expires = ifa->tstamp + lifetime * HZ;
1263 		spin_unlock(&ifa->lock);
1264 	}
1265 
1266 	return action;
1267 }
1268 
1269 static void
1270 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1271 		     bool del_rt, bool del_peer)
1272 {
1273 	struct fib6_info *f6i;
1274 
1275 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1276 					ifp->prefix_len,
1277 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1278 	if (f6i) {
1279 		if (del_rt)
1280 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1281 		else {
1282 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1283 				fib6_set_expires(f6i, expires);
1284 			fib6_info_release(f6i);
1285 		}
1286 	}
1287 }
1288 
1289 
1290 /* This function wants to get referenced ifp and releases it before return */
1291 
1292 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1293 {
1294 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1295 	struct net *net = dev_net(ifp->idev->dev);
1296 	unsigned long expires;
1297 	int state;
1298 
1299 	ASSERT_RTNL();
1300 
1301 	spin_lock_bh(&ifp->lock);
1302 	state = ifp->state;
1303 	ifp->state = INET6_IFADDR_STATE_DEAD;
1304 	spin_unlock_bh(&ifp->lock);
1305 
1306 	if (state == INET6_IFADDR_STATE_DEAD)
1307 		goto out;
1308 
1309 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1310 	hlist_del_init_rcu(&ifp->addr_lst);
1311 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1312 
1313 	write_lock_bh(&ifp->idev->lock);
1314 
1315 	if (ifp->flags&IFA_F_TEMPORARY) {
1316 		list_del(&ifp->tmp_list);
1317 		if (ifp->ifpub) {
1318 			in6_ifa_put(ifp->ifpub);
1319 			ifp->ifpub = NULL;
1320 		}
1321 		__in6_ifa_put(ifp);
1322 	}
1323 
1324 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1325 		action = check_cleanup_prefix_route(ifp, &expires);
1326 
1327 	list_del_rcu(&ifp->if_list);
1328 	__in6_ifa_put(ifp);
1329 
1330 	write_unlock_bh(&ifp->idev->lock);
1331 
1332 	addrconf_del_dad_work(ifp);
1333 
1334 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1335 
1336 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1337 
1338 	if (action != CLEANUP_PREFIX_RT_NOP) {
1339 		cleanup_prefix_route(ifp, expires,
1340 			action == CLEANUP_PREFIX_RT_DEL, false);
1341 	}
1342 
1343 	/* clean up prefsrc entries */
1344 	rt6_remove_prefsrc(ifp);
1345 out:
1346 	in6_ifa_put(ifp);
1347 }
1348 
1349 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1350 {
1351 	struct inet6_dev *idev = ifp->idev;
1352 	unsigned long tmp_tstamp, age;
1353 	unsigned long regen_advance;
1354 	unsigned long now = jiffies;
1355 	s32 cnf_temp_preferred_lft;
1356 	struct inet6_ifaddr *ift;
1357 	struct ifa6_config cfg;
1358 	long max_desync_factor;
1359 	struct in6_addr addr;
1360 	int ret = 0;
1361 
1362 	write_lock_bh(&idev->lock);
1363 
1364 retry:
1365 	in6_dev_hold(idev);
1366 	if (idev->cnf.use_tempaddr <= 0) {
1367 		write_unlock_bh(&idev->lock);
1368 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1369 		in6_dev_put(idev);
1370 		ret = -1;
1371 		goto out;
1372 	}
1373 	spin_lock_bh(&ifp->lock);
1374 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1375 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1376 		spin_unlock_bh(&ifp->lock);
1377 		write_unlock_bh(&idev->lock);
1378 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1379 			__func__);
1380 		in6_dev_put(idev);
1381 		ret = -1;
1382 		goto out;
1383 	}
1384 	in6_ifa_hold(ifp);
1385 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1386 	ipv6_gen_rnd_iid(&addr);
1387 
1388 	age = (now - ifp->tstamp) / HZ;
1389 
1390 	regen_advance = idev->cnf.regen_max_retry *
1391 			idev->cnf.dad_transmits *
1392 			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1393 
1394 	/* recalculate max_desync_factor each time and update
1395 	 * idev->desync_factor if it's larger
1396 	 */
1397 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1398 	max_desync_factor = min_t(long,
1399 				  idev->cnf.max_desync_factor,
1400 				  cnf_temp_preferred_lft - regen_advance);
1401 
1402 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1403 		if (max_desync_factor > 0) {
1404 			get_random_bytes(&idev->desync_factor,
1405 					 sizeof(idev->desync_factor));
1406 			idev->desync_factor %= max_desync_factor;
1407 		} else {
1408 			idev->desync_factor = 0;
1409 		}
1410 	}
1411 
1412 	memset(&cfg, 0, sizeof(cfg));
1413 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1414 			      idev->cnf.temp_valid_lft + age);
1415 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1416 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1417 	cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1418 
1419 	cfg.plen = ifp->prefix_len;
1420 	tmp_tstamp = ifp->tstamp;
1421 	spin_unlock_bh(&ifp->lock);
1422 
1423 	write_unlock_bh(&idev->lock);
1424 
1425 	/* A temporary address is created only if this calculated Preferred
1426 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1427 	 * an implementation must not create a temporary address with a zero
1428 	 * Preferred Lifetime.
1429 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1430 	 * temporary addresses being generated.
1431 	 */
1432 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1433 	if (cfg.preferred_lft <= regen_advance + age) {
1434 		in6_ifa_put(ifp);
1435 		in6_dev_put(idev);
1436 		ret = -1;
1437 		goto out;
1438 	}
1439 
1440 	cfg.ifa_flags = IFA_F_TEMPORARY;
1441 	/* set in addrconf_prefix_rcv() */
1442 	if (ifp->flags & IFA_F_OPTIMISTIC)
1443 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1444 
1445 	cfg.pfx = &addr;
1446 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1447 
1448 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1449 	if (IS_ERR(ift)) {
1450 		in6_ifa_put(ifp);
1451 		in6_dev_put(idev);
1452 		pr_info("%s: retry temporary address regeneration\n", __func__);
1453 		write_lock_bh(&idev->lock);
1454 		goto retry;
1455 	}
1456 
1457 	spin_lock_bh(&ift->lock);
1458 	ift->ifpub = ifp;
1459 	ift->cstamp = now;
1460 	ift->tstamp = tmp_tstamp;
1461 	spin_unlock_bh(&ift->lock);
1462 
1463 	addrconf_dad_start(ift);
1464 	in6_ifa_put(ift);
1465 	in6_dev_put(idev);
1466 out:
1467 	return ret;
1468 }
1469 
1470 /*
1471  *	Choose an appropriate source address (RFC3484)
1472  */
1473 enum {
1474 	IPV6_SADDR_RULE_INIT = 0,
1475 	IPV6_SADDR_RULE_LOCAL,
1476 	IPV6_SADDR_RULE_SCOPE,
1477 	IPV6_SADDR_RULE_PREFERRED,
1478 #ifdef CONFIG_IPV6_MIP6
1479 	IPV6_SADDR_RULE_HOA,
1480 #endif
1481 	IPV6_SADDR_RULE_OIF,
1482 	IPV6_SADDR_RULE_LABEL,
1483 	IPV6_SADDR_RULE_PRIVACY,
1484 	IPV6_SADDR_RULE_ORCHID,
1485 	IPV6_SADDR_RULE_PREFIX,
1486 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1487 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1488 #endif
1489 	IPV6_SADDR_RULE_MAX
1490 };
1491 
1492 struct ipv6_saddr_score {
1493 	int			rule;
1494 	int			addr_type;
1495 	struct inet6_ifaddr	*ifa;
1496 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1497 	int			scopedist;
1498 	int			matchlen;
1499 };
1500 
1501 struct ipv6_saddr_dst {
1502 	const struct in6_addr *addr;
1503 	int ifindex;
1504 	int scope;
1505 	int label;
1506 	unsigned int prefs;
1507 };
1508 
1509 static inline int ipv6_saddr_preferred(int type)
1510 {
1511 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1512 		return 1;
1513 	return 0;
1514 }
1515 
1516 static bool ipv6_use_optimistic_addr(struct net *net,
1517 				     struct inet6_dev *idev)
1518 {
1519 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1520 	if (!idev)
1521 		return false;
1522 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1523 		return false;
1524 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1525 		return false;
1526 
1527 	return true;
1528 #else
1529 	return false;
1530 #endif
1531 }
1532 
1533 static bool ipv6_allow_optimistic_dad(struct net *net,
1534 				      struct inet6_dev *idev)
1535 {
1536 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1537 	if (!idev)
1538 		return false;
1539 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1540 		return false;
1541 
1542 	return true;
1543 #else
1544 	return false;
1545 #endif
1546 }
1547 
1548 static int ipv6_get_saddr_eval(struct net *net,
1549 			       struct ipv6_saddr_score *score,
1550 			       struct ipv6_saddr_dst *dst,
1551 			       int i)
1552 {
1553 	int ret;
1554 
1555 	if (i <= score->rule) {
1556 		switch (i) {
1557 		case IPV6_SADDR_RULE_SCOPE:
1558 			ret = score->scopedist;
1559 			break;
1560 		case IPV6_SADDR_RULE_PREFIX:
1561 			ret = score->matchlen;
1562 			break;
1563 		default:
1564 			ret = !!test_bit(i, score->scorebits);
1565 		}
1566 		goto out;
1567 	}
1568 
1569 	switch (i) {
1570 	case IPV6_SADDR_RULE_INIT:
1571 		/* Rule 0: remember if hiscore is not ready yet */
1572 		ret = !!score->ifa;
1573 		break;
1574 	case IPV6_SADDR_RULE_LOCAL:
1575 		/* Rule 1: Prefer same address */
1576 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1577 		break;
1578 	case IPV6_SADDR_RULE_SCOPE:
1579 		/* Rule 2: Prefer appropriate scope
1580 		 *
1581 		 *      ret
1582 		 *       ^
1583 		 *    -1 |  d 15
1584 		 *    ---+--+-+---> scope
1585 		 *       |
1586 		 *       |             d is scope of the destination.
1587 		 *  B-d  |  \
1588 		 *       |   \      <- smaller scope is better if
1589 		 *  B-15 |    \        if scope is enough for destination.
1590 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1591 		 * d-C-1 | /
1592 		 *       |/         <- greater is better
1593 		 *   -C  /             if scope is not enough for destination.
1594 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1595 		 *
1596 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1597 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1598 		 * Assume B = 0 and we get C > 29.
1599 		 */
1600 		ret = __ipv6_addr_src_scope(score->addr_type);
1601 		if (ret >= dst->scope)
1602 			ret = -ret;
1603 		else
1604 			ret -= 128;	/* 30 is enough */
1605 		score->scopedist = ret;
1606 		break;
1607 	case IPV6_SADDR_RULE_PREFERRED:
1608 	    {
1609 		/* Rule 3: Avoid deprecated and optimistic addresses */
1610 		u8 avoid = IFA_F_DEPRECATED;
1611 
1612 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1613 			avoid |= IFA_F_OPTIMISTIC;
1614 		ret = ipv6_saddr_preferred(score->addr_type) ||
1615 		      !(score->ifa->flags & avoid);
1616 		break;
1617 	    }
1618 #ifdef CONFIG_IPV6_MIP6
1619 	case IPV6_SADDR_RULE_HOA:
1620 	    {
1621 		/* Rule 4: Prefer home address */
1622 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1623 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1624 		break;
1625 	    }
1626 #endif
1627 	case IPV6_SADDR_RULE_OIF:
1628 		/* Rule 5: Prefer outgoing interface */
1629 		ret = (!dst->ifindex ||
1630 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1631 		break;
1632 	case IPV6_SADDR_RULE_LABEL:
1633 		/* Rule 6: Prefer matching label */
1634 		ret = ipv6_addr_label(net,
1635 				      &score->ifa->addr, score->addr_type,
1636 				      score->ifa->idev->dev->ifindex) == dst->label;
1637 		break;
1638 	case IPV6_SADDR_RULE_PRIVACY:
1639 	    {
1640 		/* Rule 7: Prefer public address
1641 		 * Note: prefer temporary address if use_tempaddr >= 2
1642 		 */
1643 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1644 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1645 				score->ifa->idev->cnf.use_tempaddr >= 2;
1646 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1647 		break;
1648 	    }
1649 	case IPV6_SADDR_RULE_ORCHID:
1650 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1651 		 *	    non-ORCHID vs non-ORCHID
1652 		 */
1653 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1654 			ipv6_addr_orchid(dst->addr));
1655 		break;
1656 	case IPV6_SADDR_RULE_PREFIX:
1657 		/* Rule 8: Use longest matching prefix */
1658 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1659 		if (ret > score->ifa->prefix_len)
1660 			ret = score->ifa->prefix_len;
1661 		score->matchlen = ret;
1662 		break;
1663 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1664 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1665 		/* Optimistic addresses still have lower precedence than other
1666 		 * preferred addresses.
1667 		 */
1668 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1669 		break;
1670 #endif
1671 	default:
1672 		ret = 0;
1673 	}
1674 
1675 	if (ret)
1676 		__set_bit(i, score->scorebits);
1677 	score->rule = i;
1678 out:
1679 	return ret;
1680 }
1681 
1682 static int __ipv6_dev_get_saddr(struct net *net,
1683 				struct ipv6_saddr_dst *dst,
1684 				struct inet6_dev *idev,
1685 				struct ipv6_saddr_score *scores,
1686 				int hiscore_idx)
1687 {
1688 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1689 
1690 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1691 		int i;
1692 
1693 		/*
1694 		 * - Tentative Address (RFC2462 section 5.4)
1695 		 *  - A tentative address is not considered
1696 		 *    "assigned to an interface" in the traditional
1697 		 *    sense, unless it is also flagged as optimistic.
1698 		 * - Candidate Source Address (section 4)
1699 		 *  - In any case, anycast addresses, multicast
1700 		 *    addresses, and the unspecified address MUST
1701 		 *    NOT be included in a candidate set.
1702 		 */
1703 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1704 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1705 			continue;
1706 
1707 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1708 
1709 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1710 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1711 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1712 					    idev->dev->name);
1713 			continue;
1714 		}
1715 
1716 		score->rule = -1;
1717 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1718 
1719 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1720 			int minihiscore, miniscore;
1721 
1722 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1723 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1724 
1725 			if (minihiscore > miniscore) {
1726 				if (i == IPV6_SADDR_RULE_SCOPE &&
1727 				    score->scopedist > 0) {
1728 					/*
1729 					 * special case:
1730 					 * each remaining entry
1731 					 * has too small (not enough)
1732 					 * scope, because ifa entries
1733 					 * are sorted by their scope
1734 					 * values.
1735 					 */
1736 					goto out;
1737 				}
1738 				break;
1739 			} else if (minihiscore < miniscore) {
1740 				swap(hiscore, score);
1741 				hiscore_idx = 1 - hiscore_idx;
1742 
1743 				/* restore our iterator */
1744 				score->ifa = hiscore->ifa;
1745 
1746 				break;
1747 			}
1748 		}
1749 	}
1750 out:
1751 	return hiscore_idx;
1752 }
1753 
1754 static int ipv6_get_saddr_master(struct net *net,
1755 				 const struct net_device *dst_dev,
1756 				 const struct net_device *master,
1757 				 struct ipv6_saddr_dst *dst,
1758 				 struct ipv6_saddr_score *scores,
1759 				 int hiscore_idx)
1760 {
1761 	struct inet6_dev *idev;
1762 
1763 	idev = __in6_dev_get(dst_dev);
1764 	if (idev)
1765 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1766 						   scores, hiscore_idx);
1767 
1768 	idev = __in6_dev_get(master);
1769 	if (idev)
1770 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1771 						   scores, hiscore_idx);
1772 
1773 	return hiscore_idx;
1774 }
1775 
1776 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1777 		       const struct in6_addr *daddr, unsigned int prefs,
1778 		       struct in6_addr *saddr)
1779 {
1780 	struct ipv6_saddr_score scores[2], *hiscore;
1781 	struct ipv6_saddr_dst dst;
1782 	struct inet6_dev *idev;
1783 	struct net_device *dev;
1784 	int dst_type;
1785 	bool use_oif_addr = false;
1786 	int hiscore_idx = 0;
1787 	int ret = 0;
1788 
1789 	dst_type = __ipv6_addr_type(daddr);
1790 	dst.addr = daddr;
1791 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1792 	dst.scope = __ipv6_addr_src_scope(dst_type);
1793 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1794 	dst.prefs = prefs;
1795 
1796 	scores[hiscore_idx].rule = -1;
1797 	scores[hiscore_idx].ifa = NULL;
1798 
1799 	rcu_read_lock();
1800 
1801 	/* Candidate Source Address (section 4)
1802 	 *  - multicast and link-local destination address,
1803 	 *    the set of candidate source address MUST only
1804 	 *    include addresses assigned to interfaces
1805 	 *    belonging to the same link as the outgoing
1806 	 *    interface.
1807 	 * (- For site-local destination addresses, the
1808 	 *    set of candidate source addresses MUST only
1809 	 *    include addresses assigned to interfaces
1810 	 *    belonging to the same site as the outgoing
1811 	 *    interface.)
1812 	 *  - "It is RECOMMENDED that the candidate source addresses
1813 	 *    be the set of unicast addresses assigned to the
1814 	 *    interface that will be used to send to the destination
1815 	 *    (the 'outgoing' interface)." (RFC 6724)
1816 	 */
1817 	if (dst_dev) {
1818 		idev = __in6_dev_get(dst_dev);
1819 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1820 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1821 		    (idev && idev->cnf.use_oif_addrs_only)) {
1822 			use_oif_addr = true;
1823 		}
1824 	}
1825 
1826 	if (use_oif_addr) {
1827 		if (idev)
1828 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1829 	} else {
1830 		const struct net_device *master;
1831 		int master_idx = 0;
1832 
1833 		/* if dst_dev exists and is enslaved to an L3 device, then
1834 		 * prefer addresses from dst_dev and then the master over
1835 		 * any other enslaved devices in the L3 domain.
1836 		 */
1837 		master = l3mdev_master_dev_rcu(dst_dev);
1838 		if (master) {
1839 			master_idx = master->ifindex;
1840 
1841 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1842 							    master, &dst,
1843 							    scores, hiscore_idx);
1844 
1845 			if (scores[hiscore_idx].ifa)
1846 				goto out;
1847 		}
1848 
1849 		for_each_netdev_rcu(net, dev) {
1850 			/* only consider addresses on devices in the
1851 			 * same L3 domain
1852 			 */
1853 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1854 				continue;
1855 			idev = __in6_dev_get(dev);
1856 			if (!idev)
1857 				continue;
1858 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1859 		}
1860 	}
1861 
1862 out:
1863 	hiscore = &scores[hiscore_idx];
1864 	if (!hiscore->ifa)
1865 		ret = -EADDRNOTAVAIL;
1866 	else
1867 		*saddr = hiscore->ifa->addr;
1868 
1869 	rcu_read_unlock();
1870 	return ret;
1871 }
1872 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1873 
1874 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1875 			      u32 banned_flags)
1876 {
1877 	struct inet6_ifaddr *ifp;
1878 	int err = -EADDRNOTAVAIL;
1879 
1880 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1881 		if (ifp->scope > IFA_LINK)
1882 			break;
1883 		if (ifp->scope == IFA_LINK &&
1884 		    !(ifp->flags & banned_flags)) {
1885 			*addr = ifp->addr;
1886 			err = 0;
1887 			break;
1888 		}
1889 	}
1890 	return err;
1891 }
1892 
1893 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1894 		    u32 banned_flags)
1895 {
1896 	struct inet6_dev *idev;
1897 	int err = -EADDRNOTAVAIL;
1898 
1899 	rcu_read_lock();
1900 	idev = __in6_dev_get(dev);
1901 	if (idev) {
1902 		read_lock_bh(&idev->lock);
1903 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1904 		read_unlock_bh(&idev->lock);
1905 	}
1906 	rcu_read_unlock();
1907 	return err;
1908 }
1909 
1910 static int ipv6_count_addresses(const struct inet6_dev *idev)
1911 {
1912 	const struct inet6_ifaddr *ifp;
1913 	int cnt = 0;
1914 
1915 	rcu_read_lock();
1916 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1917 		cnt++;
1918 	rcu_read_unlock();
1919 	return cnt;
1920 }
1921 
1922 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1923 		  const struct net_device *dev, int strict)
1924 {
1925 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1926 				       strict, IFA_F_TENTATIVE);
1927 }
1928 EXPORT_SYMBOL(ipv6_chk_addr);
1929 
1930 /* device argument is used to find the L3 domain of interest. If
1931  * skip_dev_check is set, then the ifp device is not checked against
1932  * the passed in dev argument. So the 2 cases for addresses checks are:
1933  *   1. does the address exist in the L3 domain that dev is part of
1934  *      (skip_dev_check = true), or
1935  *
1936  *   2. does the address exist on the specific device
1937  *      (skip_dev_check = false)
1938  */
1939 static struct net_device *
1940 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1941 			  const struct net_device *dev, bool skip_dev_check,
1942 			  int strict, u32 banned_flags)
1943 {
1944 	unsigned int hash = inet6_addr_hash(net, addr);
1945 	struct net_device *l3mdev, *ndev;
1946 	struct inet6_ifaddr *ifp;
1947 	u32 ifp_flags;
1948 
1949 	rcu_read_lock();
1950 
1951 	l3mdev = l3mdev_master_dev_rcu(dev);
1952 	if (skip_dev_check)
1953 		dev = NULL;
1954 
1955 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1956 		ndev = ifp->idev->dev;
1957 
1958 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1959 			continue;
1960 
1961 		/* Decouple optimistic from tentative for evaluation here.
1962 		 * Ban optimistic addresses explicitly, when required.
1963 		 */
1964 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1965 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1966 			    : ifp->flags;
1967 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1968 		    !(ifp_flags&banned_flags) &&
1969 		    (!dev || ndev == dev ||
1970 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1971 			rcu_read_unlock();
1972 			return ndev;
1973 		}
1974 	}
1975 
1976 	rcu_read_unlock();
1977 	return NULL;
1978 }
1979 
1980 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1981 			    const struct net_device *dev, bool skip_dev_check,
1982 			    int strict, u32 banned_flags)
1983 {
1984 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1985 					 strict, banned_flags) ? 1 : 0;
1986 }
1987 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1988 
1989 
1990 /* Compares an address/prefix_len with addresses on device @dev.
1991  * If one is found it returns true.
1992  */
1993 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1994 	const unsigned int prefix_len, struct net_device *dev)
1995 {
1996 	const struct inet6_ifaddr *ifa;
1997 	const struct inet6_dev *idev;
1998 	bool ret = false;
1999 
2000 	rcu_read_lock();
2001 	idev = __in6_dev_get(dev);
2002 	if (idev) {
2003 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2004 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2005 			if (ret)
2006 				break;
2007 		}
2008 	}
2009 	rcu_read_unlock();
2010 
2011 	return ret;
2012 }
2013 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2014 
2015 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2016 {
2017 	const struct inet6_ifaddr *ifa;
2018 	const struct inet6_dev *idev;
2019 	int	onlink;
2020 
2021 	onlink = 0;
2022 	rcu_read_lock();
2023 	idev = __in6_dev_get(dev);
2024 	if (idev) {
2025 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2026 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
2027 						   ifa->prefix_len);
2028 			if (onlink)
2029 				break;
2030 		}
2031 	}
2032 	rcu_read_unlock();
2033 	return onlink;
2034 }
2035 EXPORT_SYMBOL(ipv6_chk_prefix);
2036 
2037 /**
2038  * ipv6_dev_find - find the first device with a given source address.
2039  * @net: the net namespace
2040  * @addr: the source address
2041  * @dev: used to find the L3 domain of interest
2042  *
2043  * The caller should be protected by RCU, or RTNL.
2044  */
2045 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2046 				 struct net_device *dev)
2047 {
2048 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2049 					 IFA_F_TENTATIVE);
2050 }
2051 EXPORT_SYMBOL(ipv6_dev_find);
2052 
2053 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2054 				     struct net_device *dev, int strict)
2055 {
2056 	unsigned int hash = inet6_addr_hash(net, addr);
2057 	struct inet6_ifaddr *ifp, *result = NULL;
2058 
2059 	rcu_read_lock();
2060 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2061 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2062 			if (!dev || ifp->idev->dev == dev ||
2063 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2064 				result = ifp;
2065 				in6_ifa_hold(ifp);
2066 				break;
2067 			}
2068 		}
2069 	}
2070 	rcu_read_unlock();
2071 
2072 	return result;
2073 }
2074 
2075 /* Gets referenced address, destroys ifaddr */
2076 
2077 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2078 {
2079 	if (dad_failed)
2080 		ifp->flags |= IFA_F_DADFAILED;
2081 
2082 	if (ifp->flags&IFA_F_TEMPORARY) {
2083 		struct inet6_ifaddr *ifpub;
2084 		spin_lock_bh(&ifp->lock);
2085 		ifpub = ifp->ifpub;
2086 		if (ifpub) {
2087 			in6_ifa_hold(ifpub);
2088 			spin_unlock_bh(&ifp->lock);
2089 			ipv6_create_tempaddr(ifpub, true);
2090 			in6_ifa_put(ifpub);
2091 		} else {
2092 			spin_unlock_bh(&ifp->lock);
2093 		}
2094 		ipv6_del_addr(ifp);
2095 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2096 		spin_lock_bh(&ifp->lock);
2097 		addrconf_del_dad_work(ifp);
2098 		ifp->flags |= IFA_F_TENTATIVE;
2099 		if (dad_failed)
2100 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2101 		spin_unlock_bh(&ifp->lock);
2102 		if (dad_failed)
2103 			ipv6_ifa_notify(0, ifp);
2104 		in6_ifa_put(ifp);
2105 	} else {
2106 		ipv6_del_addr(ifp);
2107 	}
2108 }
2109 
2110 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2111 {
2112 	int err = -ENOENT;
2113 
2114 	spin_lock_bh(&ifp->lock);
2115 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2116 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2117 		err = 0;
2118 	}
2119 	spin_unlock_bh(&ifp->lock);
2120 
2121 	return err;
2122 }
2123 
2124 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2125 {
2126 	struct inet6_dev *idev = ifp->idev;
2127 	struct net *net = dev_net(idev->dev);
2128 
2129 	if (addrconf_dad_end(ifp)) {
2130 		in6_ifa_put(ifp);
2131 		return;
2132 	}
2133 
2134 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2135 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2136 
2137 	spin_lock_bh(&ifp->lock);
2138 
2139 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2140 		struct in6_addr new_addr;
2141 		struct inet6_ifaddr *ifp2;
2142 		int retries = ifp->stable_privacy_retry + 1;
2143 		struct ifa6_config cfg = {
2144 			.pfx = &new_addr,
2145 			.plen = ifp->prefix_len,
2146 			.ifa_flags = ifp->flags,
2147 			.valid_lft = ifp->valid_lft,
2148 			.preferred_lft = ifp->prefered_lft,
2149 			.scope = ifp->scope,
2150 		};
2151 
2152 		if (retries > net->ipv6.sysctl.idgen_retries) {
2153 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2154 					     ifp->idev->dev->name);
2155 			goto errdad;
2156 		}
2157 
2158 		new_addr = ifp->addr;
2159 		if (ipv6_generate_stable_address(&new_addr, retries,
2160 						 idev))
2161 			goto errdad;
2162 
2163 		spin_unlock_bh(&ifp->lock);
2164 
2165 		if (idev->cnf.max_addresses &&
2166 		    ipv6_count_addresses(idev) >=
2167 		    idev->cnf.max_addresses)
2168 			goto lock_errdad;
2169 
2170 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2171 				     ifp->idev->dev->name);
2172 
2173 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2174 		if (IS_ERR(ifp2))
2175 			goto lock_errdad;
2176 
2177 		spin_lock_bh(&ifp2->lock);
2178 		ifp2->stable_privacy_retry = retries;
2179 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2180 		spin_unlock_bh(&ifp2->lock);
2181 
2182 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2183 		in6_ifa_put(ifp2);
2184 lock_errdad:
2185 		spin_lock_bh(&ifp->lock);
2186 	}
2187 
2188 errdad:
2189 	/* transition from _POSTDAD to _ERRDAD */
2190 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2191 	spin_unlock_bh(&ifp->lock);
2192 
2193 	addrconf_mod_dad_work(ifp, 0);
2194 	in6_ifa_put(ifp);
2195 }
2196 
2197 /* Join to solicited addr multicast group.
2198  * caller must hold RTNL */
2199 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2200 {
2201 	struct in6_addr maddr;
2202 
2203 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2204 		return;
2205 
2206 	addrconf_addr_solict_mult(addr, &maddr);
2207 	ipv6_dev_mc_inc(dev, &maddr);
2208 }
2209 
2210 /* caller must hold RTNL */
2211 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2212 {
2213 	struct in6_addr maddr;
2214 
2215 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2216 		return;
2217 
2218 	addrconf_addr_solict_mult(addr, &maddr);
2219 	__ipv6_dev_mc_dec(idev, &maddr);
2220 }
2221 
2222 /* caller must hold RTNL */
2223 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2224 {
2225 	struct in6_addr addr;
2226 
2227 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2228 		return;
2229 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2230 	if (ipv6_addr_any(&addr))
2231 		return;
2232 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2233 }
2234 
2235 /* caller must hold RTNL */
2236 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2237 {
2238 	struct in6_addr addr;
2239 
2240 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2241 		return;
2242 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2243 	if (ipv6_addr_any(&addr))
2244 		return;
2245 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2246 }
2247 
2248 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2249 {
2250 	switch (dev->addr_len) {
2251 	case ETH_ALEN:
2252 		memcpy(eui, dev->dev_addr, 3);
2253 		eui[3] = 0xFF;
2254 		eui[4] = 0xFE;
2255 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2256 		break;
2257 	case EUI64_ADDR_LEN:
2258 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2259 		eui[0] ^= 2;
2260 		break;
2261 	default:
2262 		return -1;
2263 	}
2264 
2265 	return 0;
2266 }
2267 
2268 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2269 {
2270 	const union fwnet_hwaddr *ha;
2271 
2272 	if (dev->addr_len != FWNET_ALEN)
2273 		return -1;
2274 
2275 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2276 
2277 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2278 	eui[0] ^= 2;
2279 	return 0;
2280 }
2281 
2282 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2283 {
2284 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2285 	if (dev->addr_len != ARCNET_ALEN)
2286 		return -1;
2287 	memset(eui, 0, 7);
2288 	eui[7] = *(u8 *)dev->dev_addr;
2289 	return 0;
2290 }
2291 
2292 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2293 {
2294 	if (dev->addr_len != INFINIBAND_ALEN)
2295 		return -1;
2296 	memcpy(eui, dev->dev_addr + 12, 8);
2297 	eui[0] |= 2;
2298 	return 0;
2299 }
2300 
2301 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2302 {
2303 	if (addr == 0)
2304 		return -1;
2305 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2306 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2307 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2308 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2309 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2310 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2311 	eui[1] = 0;
2312 	eui[2] = 0x5E;
2313 	eui[3] = 0xFE;
2314 	memcpy(eui + 4, &addr, 4);
2315 	return 0;
2316 }
2317 
2318 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2319 {
2320 	if (dev->priv_flags & IFF_ISATAP)
2321 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2322 	return -1;
2323 }
2324 
2325 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2326 {
2327 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2328 }
2329 
2330 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2331 {
2332 	memcpy(eui, dev->perm_addr, 3);
2333 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2334 	eui[3] = 0xFF;
2335 	eui[4] = 0xFE;
2336 	eui[0] ^= 2;
2337 	return 0;
2338 }
2339 
2340 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2341 {
2342 	switch (dev->type) {
2343 	case ARPHRD_ETHER:
2344 	case ARPHRD_FDDI:
2345 		return addrconf_ifid_eui48(eui, dev);
2346 	case ARPHRD_ARCNET:
2347 		return addrconf_ifid_arcnet(eui, dev);
2348 	case ARPHRD_INFINIBAND:
2349 		return addrconf_ifid_infiniband(eui, dev);
2350 	case ARPHRD_SIT:
2351 		return addrconf_ifid_sit(eui, dev);
2352 	case ARPHRD_IPGRE:
2353 	case ARPHRD_TUNNEL:
2354 		return addrconf_ifid_gre(eui, dev);
2355 	case ARPHRD_6LOWPAN:
2356 		return addrconf_ifid_6lowpan(eui, dev);
2357 	case ARPHRD_IEEE1394:
2358 		return addrconf_ifid_ieee1394(eui, dev);
2359 	case ARPHRD_TUNNEL6:
2360 	case ARPHRD_IP6GRE:
2361 	case ARPHRD_RAWIP:
2362 		return addrconf_ifid_ip6tnl(eui, dev);
2363 	}
2364 	return -1;
2365 }
2366 
2367 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2368 {
2369 	int err = -1;
2370 	struct inet6_ifaddr *ifp;
2371 
2372 	read_lock_bh(&idev->lock);
2373 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2374 		if (ifp->scope > IFA_LINK)
2375 			break;
2376 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2377 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2378 			err = 0;
2379 			break;
2380 		}
2381 	}
2382 	read_unlock_bh(&idev->lock);
2383 	return err;
2384 }
2385 
2386 /* Generation of a randomized Interface Identifier
2387  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2388  */
2389 
2390 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2391 {
2392 regen:
2393 	get_random_bytes(&addr->s6_addr[8], 8);
2394 
2395 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2396 	 * check if generated address is not inappropriate:
2397 	 *
2398 	 * - Reserved IPv6 Interface Identifiers
2399 	 * - XXX: already assigned to an address on the device
2400 	 */
2401 
2402 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2403 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2404 		goto regen;
2405 
2406 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2407 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2408 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2409 	 */
2410 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2411 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2412 		goto regen;
2413 
2414 	/* Reserved subnet anycast addresses */
2415 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2416 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2417 		goto regen;
2418 }
2419 
2420 /*
2421  *	Add prefix route.
2422  */
2423 
2424 static void
2425 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2426 		      struct net_device *dev, unsigned long expires,
2427 		      u32 flags, gfp_t gfp_flags)
2428 {
2429 	struct fib6_config cfg = {
2430 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2431 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2432 		.fc_ifindex = dev->ifindex,
2433 		.fc_expires = expires,
2434 		.fc_dst_len = plen,
2435 		.fc_flags = RTF_UP | flags,
2436 		.fc_nlinfo.nl_net = dev_net(dev),
2437 		.fc_protocol = RTPROT_KERNEL,
2438 		.fc_type = RTN_UNICAST,
2439 	};
2440 
2441 	cfg.fc_dst = *pfx;
2442 
2443 	/* Prevent useless cloning on PtP SIT.
2444 	   This thing is done here expecting that the whole
2445 	   class of non-broadcast devices need not cloning.
2446 	 */
2447 #if IS_ENABLED(CONFIG_IPV6_SIT)
2448 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2449 		cfg.fc_flags |= RTF_NONEXTHOP;
2450 #endif
2451 
2452 	ip6_route_add(&cfg, gfp_flags, NULL);
2453 }
2454 
2455 
2456 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2457 						  int plen,
2458 						  const struct net_device *dev,
2459 						  u32 flags, u32 noflags,
2460 						  bool no_gw)
2461 {
2462 	struct fib6_node *fn;
2463 	struct fib6_info *rt = NULL;
2464 	struct fib6_table *table;
2465 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2466 
2467 	table = fib6_get_table(dev_net(dev), tb_id);
2468 	if (!table)
2469 		return NULL;
2470 
2471 	rcu_read_lock();
2472 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2473 	if (!fn)
2474 		goto out;
2475 
2476 	for_each_fib6_node_rt_rcu(fn) {
2477 		/* prefix routes only use builtin fib6_nh */
2478 		if (rt->nh)
2479 			continue;
2480 
2481 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2482 			continue;
2483 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2484 			continue;
2485 		if ((rt->fib6_flags & flags) != flags)
2486 			continue;
2487 		if ((rt->fib6_flags & noflags) != 0)
2488 			continue;
2489 		if (!fib6_info_hold_safe(rt))
2490 			continue;
2491 		break;
2492 	}
2493 out:
2494 	rcu_read_unlock();
2495 	return rt;
2496 }
2497 
2498 
2499 /* Create "default" multicast route to the interface */
2500 
2501 static void addrconf_add_mroute(struct net_device *dev)
2502 {
2503 	struct fib6_config cfg = {
2504 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2505 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2506 		.fc_ifindex = dev->ifindex,
2507 		.fc_dst_len = 8,
2508 		.fc_flags = RTF_UP,
2509 		.fc_type = RTN_MULTICAST,
2510 		.fc_nlinfo.nl_net = dev_net(dev),
2511 		.fc_protocol = RTPROT_KERNEL,
2512 	};
2513 
2514 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2515 
2516 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2517 }
2518 
2519 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2520 {
2521 	struct inet6_dev *idev;
2522 
2523 	ASSERT_RTNL();
2524 
2525 	idev = ipv6_find_idev(dev);
2526 	if (IS_ERR(idev))
2527 		return idev;
2528 
2529 	if (idev->cnf.disable_ipv6)
2530 		return ERR_PTR(-EACCES);
2531 
2532 	/* Add default multicast route */
2533 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2534 		addrconf_add_mroute(dev);
2535 
2536 	return idev;
2537 }
2538 
2539 static void manage_tempaddrs(struct inet6_dev *idev,
2540 			     struct inet6_ifaddr *ifp,
2541 			     __u32 valid_lft, __u32 prefered_lft,
2542 			     bool create, unsigned long now)
2543 {
2544 	u32 flags;
2545 	struct inet6_ifaddr *ift;
2546 
2547 	read_lock_bh(&idev->lock);
2548 	/* update all temporary addresses in the list */
2549 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2550 		int age, max_valid, max_prefered;
2551 
2552 		if (ifp != ift->ifpub)
2553 			continue;
2554 
2555 		/* RFC 4941 section 3.3:
2556 		 * If a received option will extend the lifetime of a public
2557 		 * address, the lifetimes of temporary addresses should
2558 		 * be extended, subject to the overall constraint that no
2559 		 * temporary addresses should ever remain "valid" or "preferred"
2560 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2561 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2562 		 */
2563 		age = (now - ift->cstamp) / HZ;
2564 		max_valid = idev->cnf.temp_valid_lft - age;
2565 		if (max_valid < 0)
2566 			max_valid = 0;
2567 
2568 		max_prefered = idev->cnf.temp_prefered_lft -
2569 			       idev->desync_factor - age;
2570 		if (max_prefered < 0)
2571 			max_prefered = 0;
2572 
2573 		if (valid_lft > max_valid)
2574 			valid_lft = max_valid;
2575 
2576 		if (prefered_lft > max_prefered)
2577 			prefered_lft = max_prefered;
2578 
2579 		spin_lock(&ift->lock);
2580 		flags = ift->flags;
2581 		ift->valid_lft = valid_lft;
2582 		ift->prefered_lft = prefered_lft;
2583 		ift->tstamp = now;
2584 		if (prefered_lft > 0)
2585 			ift->flags &= ~IFA_F_DEPRECATED;
2586 
2587 		spin_unlock(&ift->lock);
2588 		if (!(flags&IFA_F_TENTATIVE))
2589 			ipv6_ifa_notify(0, ift);
2590 	}
2591 
2592 	/* Also create a temporary address if it's enabled but no temporary
2593 	 * address currently exists.
2594 	 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2595 	 * as part of cleanup (ie. deleting the mngtmpaddr).
2596 	 * We don't want that to result in creating a new temporary ip address.
2597 	 */
2598 	if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2599 		create = true;
2600 
2601 	if (create && idev->cnf.use_tempaddr > 0) {
2602 		/* When a new public address is created as described
2603 		 * in [ADDRCONF], also create a new temporary address.
2604 		 */
2605 		read_unlock_bh(&idev->lock);
2606 		ipv6_create_tempaddr(ifp, false);
2607 	} else {
2608 		read_unlock_bh(&idev->lock);
2609 	}
2610 }
2611 
2612 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2613 {
2614 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2615 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2616 }
2617 
2618 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2619 				 const struct prefix_info *pinfo,
2620 				 struct inet6_dev *in6_dev,
2621 				 const struct in6_addr *addr, int addr_type,
2622 				 u32 addr_flags, bool sllao, bool tokenized,
2623 				 __u32 valid_lft, u32 prefered_lft)
2624 {
2625 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2626 	int create = 0, update_lft = 0;
2627 
2628 	if (!ifp && valid_lft) {
2629 		int max_addresses = in6_dev->cnf.max_addresses;
2630 		struct ifa6_config cfg = {
2631 			.pfx = addr,
2632 			.plen = pinfo->prefix_len,
2633 			.ifa_flags = addr_flags,
2634 			.valid_lft = valid_lft,
2635 			.preferred_lft = prefered_lft,
2636 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2637 			.ifa_proto = IFAPROT_KERNEL_RA
2638 		};
2639 
2640 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2641 		if ((net->ipv6.devconf_all->optimistic_dad ||
2642 		     in6_dev->cnf.optimistic_dad) &&
2643 		    !net->ipv6.devconf_all->forwarding && sllao)
2644 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2645 #endif
2646 
2647 		/* Do not allow to create too much of autoconfigured
2648 		 * addresses; this would be too easy way to crash kernel.
2649 		 */
2650 		if (!max_addresses ||
2651 		    ipv6_count_addresses(in6_dev) < max_addresses)
2652 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2653 
2654 		if (IS_ERR_OR_NULL(ifp))
2655 			return -1;
2656 
2657 		create = 1;
2658 		spin_lock_bh(&ifp->lock);
2659 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2660 		ifp->cstamp = jiffies;
2661 		ifp->tokenized = tokenized;
2662 		spin_unlock_bh(&ifp->lock);
2663 		addrconf_dad_start(ifp);
2664 	}
2665 
2666 	if (ifp) {
2667 		u32 flags;
2668 		unsigned long now;
2669 		u32 stored_lft;
2670 
2671 		/* update lifetime (RFC2462 5.5.3 e) */
2672 		spin_lock_bh(&ifp->lock);
2673 		now = jiffies;
2674 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2675 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2676 		else
2677 			stored_lft = 0;
2678 
2679 		/* RFC4862 Section 5.5.3e:
2680 		 * "Note that the preferred lifetime of the
2681 		 *  corresponding address is always reset to
2682 		 *  the Preferred Lifetime in the received
2683 		 *  Prefix Information option, regardless of
2684 		 *  whether the valid lifetime is also reset or
2685 		 *  ignored."
2686 		 *
2687 		 * So we should always update prefered_lft here.
2688 		 */
2689 		update_lft = !create && stored_lft;
2690 
2691 		if (update_lft && !in6_dev->cnf.ra_honor_pio_life) {
2692 			const u32 minimum_lft = min_t(u32,
2693 				stored_lft, MIN_VALID_LIFETIME);
2694 			valid_lft = max(valid_lft, minimum_lft);
2695 		}
2696 
2697 		if (update_lft) {
2698 			ifp->valid_lft = valid_lft;
2699 			ifp->prefered_lft = prefered_lft;
2700 			ifp->tstamp = now;
2701 			flags = ifp->flags;
2702 			ifp->flags &= ~IFA_F_DEPRECATED;
2703 			spin_unlock_bh(&ifp->lock);
2704 
2705 			if (!(flags&IFA_F_TENTATIVE))
2706 				ipv6_ifa_notify(0, ifp);
2707 		} else
2708 			spin_unlock_bh(&ifp->lock);
2709 
2710 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2711 				 create, now);
2712 
2713 		in6_ifa_put(ifp);
2714 		addrconf_verify(net);
2715 	}
2716 
2717 	return 0;
2718 }
2719 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2720 
2721 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2722 {
2723 	struct prefix_info *pinfo;
2724 	__u32 valid_lft;
2725 	__u32 prefered_lft;
2726 	int addr_type, err;
2727 	u32 addr_flags = 0;
2728 	struct inet6_dev *in6_dev;
2729 	struct net *net = dev_net(dev);
2730 
2731 	pinfo = (struct prefix_info *) opt;
2732 
2733 	if (len < sizeof(struct prefix_info)) {
2734 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2735 		return;
2736 	}
2737 
2738 	/*
2739 	 *	Validation checks ([ADDRCONF], page 19)
2740 	 */
2741 
2742 	addr_type = ipv6_addr_type(&pinfo->prefix);
2743 
2744 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2745 		return;
2746 
2747 	valid_lft = ntohl(pinfo->valid);
2748 	prefered_lft = ntohl(pinfo->prefered);
2749 
2750 	if (prefered_lft > valid_lft) {
2751 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2752 		return;
2753 	}
2754 
2755 	in6_dev = in6_dev_get(dev);
2756 
2757 	if (!in6_dev) {
2758 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2759 				    dev->name);
2760 		return;
2761 	}
2762 
2763 	if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2764 		goto put;
2765 
2766 	/*
2767 	 *	Two things going on here:
2768 	 *	1) Add routes for on-link prefixes
2769 	 *	2) Configure prefixes with the auto flag set
2770 	 */
2771 
2772 	if (pinfo->onlink) {
2773 		struct fib6_info *rt;
2774 		unsigned long rt_expires;
2775 
2776 		/* Avoid arithmetic overflow. Really, we could
2777 		 * save rt_expires in seconds, likely valid_lft,
2778 		 * but it would require division in fib gc, that it
2779 		 * not good.
2780 		 */
2781 		if (HZ > USER_HZ)
2782 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2783 		else
2784 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2785 
2786 		if (addrconf_finite_timeout(rt_expires))
2787 			rt_expires *= HZ;
2788 
2789 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2790 					       pinfo->prefix_len,
2791 					       dev,
2792 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2793 					       RTF_DEFAULT, true);
2794 
2795 		if (rt) {
2796 			/* Autoconf prefix route */
2797 			if (valid_lft == 0) {
2798 				ip6_del_rt(net, rt, false);
2799 				rt = NULL;
2800 			} else if (addrconf_finite_timeout(rt_expires)) {
2801 				/* not infinity */
2802 				fib6_set_expires(rt, jiffies + rt_expires);
2803 			} else {
2804 				fib6_clean_expires(rt);
2805 			}
2806 		} else if (valid_lft) {
2807 			clock_t expires = 0;
2808 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2809 			if (addrconf_finite_timeout(rt_expires)) {
2810 				/* not infinity */
2811 				flags |= RTF_EXPIRES;
2812 				expires = jiffies_to_clock_t(rt_expires);
2813 			}
2814 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2815 					      0, dev, expires, flags,
2816 					      GFP_ATOMIC);
2817 		}
2818 		fib6_info_release(rt);
2819 	}
2820 
2821 	/* Try to figure out our local address for this prefix */
2822 
2823 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2824 		struct in6_addr addr;
2825 		bool tokenized = false, dev_addr_generated = false;
2826 
2827 		if (pinfo->prefix_len == 64) {
2828 			memcpy(&addr, &pinfo->prefix, 8);
2829 
2830 			if (!ipv6_addr_any(&in6_dev->token)) {
2831 				read_lock_bh(&in6_dev->lock);
2832 				memcpy(addr.s6_addr + 8,
2833 				       in6_dev->token.s6_addr + 8, 8);
2834 				read_unlock_bh(&in6_dev->lock);
2835 				tokenized = true;
2836 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2837 				   !ipv6_generate_stable_address(&addr, 0,
2838 								 in6_dev)) {
2839 				addr_flags |= IFA_F_STABLE_PRIVACY;
2840 				goto ok;
2841 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2842 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2843 				goto put;
2844 			} else {
2845 				dev_addr_generated = true;
2846 			}
2847 			goto ok;
2848 		}
2849 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2850 				    pinfo->prefix_len);
2851 		goto put;
2852 
2853 ok:
2854 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2855 						   &addr, addr_type,
2856 						   addr_flags, sllao,
2857 						   tokenized, valid_lft,
2858 						   prefered_lft);
2859 		if (err)
2860 			goto put;
2861 
2862 		/* Ignore error case here because previous prefix add addr was
2863 		 * successful which will be notified.
2864 		 */
2865 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2866 					      addr_type, addr_flags, sllao,
2867 					      tokenized, valid_lft,
2868 					      prefered_lft,
2869 					      dev_addr_generated);
2870 	}
2871 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2872 put:
2873 	in6_dev_put(in6_dev);
2874 }
2875 
2876 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2877 		struct in6_ifreq *ireq)
2878 {
2879 	struct ip_tunnel_parm p = { };
2880 	int err;
2881 
2882 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2883 		return -EADDRNOTAVAIL;
2884 
2885 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2886 	p.iph.version = 4;
2887 	p.iph.ihl = 5;
2888 	p.iph.protocol = IPPROTO_IPV6;
2889 	p.iph.ttl = 64;
2890 
2891 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2892 		return -EOPNOTSUPP;
2893 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2894 	if (err)
2895 		return err;
2896 
2897 	dev = __dev_get_by_name(net, p.name);
2898 	if (!dev)
2899 		return -ENOBUFS;
2900 	return dev_open(dev, NULL);
2901 }
2902 
2903 /*
2904  *	Set destination address.
2905  *	Special case for SIT interfaces where we create a new "virtual"
2906  *	device.
2907  */
2908 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2909 {
2910 	struct net_device *dev;
2911 	struct in6_ifreq ireq;
2912 	int err = -ENODEV;
2913 
2914 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2915 		return -ENODEV;
2916 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2917 		return -EFAULT;
2918 
2919 	rtnl_lock();
2920 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2921 	if (dev && dev->type == ARPHRD_SIT)
2922 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2923 	rtnl_unlock();
2924 	return err;
2925 }
2926 
2927 static int ipv6_mc_config(struct sock *sk, bool join,
2928 			  const struct in6_addr *addr, int ifindex)
2929 {
2930 	int ret;
2931 
2932 	ASSERT_RTNL();
2933 
2934 	lock_sock(sk);
2935 	if (join)
2936 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2937 	else
2938 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2939 	release_sock(sk);
2940 
2941 	return ret;
2942 }
2943 
2944 /*
2945  *	Manual configuration of address on an interface
2946  */
2947 static int inet6_addr_add(struct net *net, int ifindex,
2948 			  struct ifa6_config *cfg,
2949 			  struct netlink_ext_ack *extack)
2950 {
2951 	struct inet6_ifaddr *ifp;
2952 	struct inet6_dev *idev;
2953 	struct net_device *dev;
2954 	unsigned long timeout;
2955 	clock_t expires;
2956 	u32 flags;
2957 
2958 	ASSERT_RTNL();
2959 
2960 	if (cfg->plen > 128) {
2961 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
2962 		return -EINVAL;
2963 	}
2964 
2965 	/* check the lifetime */
2966 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
2967 		NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
2968 		return -EINVAL;
2969 	}
2970 
2971 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
2972 		NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
2973 		return -EINVAL;
2974 	}
2975 
2976 	dev = __dev_get_by_index(net, ifindex);
2977 	if (!dev)
2978 		return -ENODEV;
2979 
2980 	idev = addrconf_add_dev(dev);
2981 	if (IS_ERR(idev)) {
2982 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
2983 		return PTR_ERR(idev);
2984 	}
2985 
2986 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2987 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2988 					 true, cfg->pfx, ifindex);
2989 
2990 		if (ret < 0) {
2991 			NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
2992 			return ret;
2993 		}
2994 	}
2995 
2996 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2997 
2998 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2999 	if (addrconf_finite_timeout(timeout)) {
3000 		expires = jiffies_to_clock_t(timeout * HZ);
3001 		cfg->valid_lft = timeout;
3002 		flags = RTF_EXPIRES;
3003 	} else {
3004 		expires = 0;
3005 		flags = 0;
3006 		cfg->ifa_flags |= IFA_F_PERMANENT;
3007 	}
3008 
3009 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3010 	if (addrconf_finite_timeout(timeout)) {
3011 		if (timeout == 0)
3012 			cfg->ifa_flags |= IFA_F_DEPRECATED;
3013 		cfg->preferred_lft = timeout;
3014 	}
3015 
3016 	ifp = ipv6_add_addr(idev, cfg, true, extack);
3017 	if (!IS_ERR(ifp)) {
3018 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3019 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3020 					      ifp->rt_priority, dev, expires,
3021 					      flags, GFP_KERNEL);
3022 		}
3023 
3024 		/* Send a netlink notification if DAD is enabled and
3025 		 * optimistic flag is not set
3026 		 */
3027 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3028 			ipv6_ifa_notify(0, ifp);
3029 		/*
3030 		 * Note that section 3.1 of RFC 4429 indicates
3031 		 * that the Optimistic flag should not be set for
3032 		 * manually configured addresses
3033 		 */
3034 		addrconf_dad_start(ifp);
3035 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3036 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
3037 					 cfg->preferred_lft, true, jiffies);
3038 		in6_ifa_put(ifp);
3039 		addrconf_verify_rtnl(net);
3040 		return 0;
3041 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3042 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3043 			       cfg->pfx, ifindex);
3044 	}
3045 
3046 	return PTR_ERR(ifp);
3047 }
3048 
3049 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3050 			  const struct in6_addr *pfx, unsigned int plen,
3051 			  struct netlink_ext_ack *extack)
3052 {
3053 	struct inet6_ifaddr *ifp;
3054 	struct inet6_dev *idev;
3055 	struct net_device *dev;
3056 
3057 	if (plen > 128) {
3058 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3059 		return -EINVAL;
3060 	}
3061 
3062 	dev = __dev_get_by_index(net, ifindex);
3063 	if (!dev) {
3064 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3065 		return -ENODEV;
3066 	}
3067 
3068 	idev = __in6_dev_get(dev);
3069 	if (!idev) {
3070 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3071 		return -ENXIO;
3072 	}
3073 
3074 	read_lock_bh(&idev->lock);
3075 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3076 		if (ifp->prefix_len == plen &&
3077 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3078 			in6_ifa_hold(ifp);
3079 			read_unlock_bh(&idev->lock);
3080 
3081 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3082 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
3083 				manage_tempaddrs(idev, ifp, 0, 0, false,
3084 						 jiffies);
3085 			ipv6_del_addr(ifp);
3086 			addrconf_verify_rtnl(net);
3087 			if (ipv6_addr_is_multicast(pfx)) {
3088 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3089 					       false, pfx, dev->ifindex);
3090 			}
3091 			return 0;
3092 		}
3093 	}
3094 	read_unlock_bh(&idev->lock);
3095 
3096 	NL_SET_ERR_MSG_MOD(extack, "address not found");
3097 	return -EADDRNOTAVAIL;
3098 }
3099 
3100 
3101 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3102 {
3103 	struct ifa6_config cfg = {
3104 		.ifa_flags = IFA_F_PERMANENT,
3105 		.preferred_lft = INFINITY_LIFE_TIME,
3106 		.valid_lft = INFINITY_LIFE_TIME,
3107 	};
3108 	struct in6_ifreq ireq;
3109 	int err;
3110 
3111 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3112 		return -EPERM;
3113 
3114 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3115 		return -EFAULT;
3116 
3117 	cfg.pfx = &ireq.ifr6_addr;
3118 	cfg.plen = ireq.ifr6_prefixlen;
3119 
3120 	rtnl_lock();
3121 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3122 	rtnl_unlock();
3123 	return err;
3124 }
3125 
3126 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3127 {
3128 	struct in6_ifreq ireq;
3129 	int err;
3130 
3131 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3132 		return -EPERM;
3133 
3134 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3135 		return -EFAULT;
3136 
3137 	rtnl_lock();
3138 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3139 			     ireq.ifr6_prefixlen, NULL);
3140 	rtnl_unlock();
3141 	return err;
3142 }
3143 
3144 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3145 		     int plen, int scope, u8 proto)
3146 {
3147 	struct inet6_ifaddr *ifp;
3148 	struct ifa6_config cfg = {
3149 		.pfx = addr,
3150 		.plen = plen,
3151 		.ifa_flags = IFA_F_PERMANENT,
3152 		.valid_lft = INFINITY_LIFE_TIME,
3153 		.preferred_lft = INFINITY_LIFE_TIME,
3154 		.scope = scope,
3155 		.ifa_proto = proto
3156 	};
3157 
3158 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3159 	if (!IS_ERR(ifp)) {
3160 		spin_lock_bh(&ifp->lock);
3161 		ifp->flags &= ~IFA_F_TENTATIVE;
3162 		spin_unlock_bh(&ifp->lock);
3163 		rt_genid_bump_ipv6(dev_net(idev->dev));
3164 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3165 		in6_ifa_put(ifp);
3166 	}
3167 }
3168 
3169 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3170 static void add_v4_addrs(struct inet6_dev *idev)
3171 {
3172 	struct in6_addr addr;
3173 	struct net_device *dev;
3174 	struct net *net = dev_net(idev->dev);
3175 	int scope, plen, offset = 0;
3176 	u32 pflags = 0;
3177 
3178 	ASSERT_RTNL();
3179 
3180 	memset(&addr, 0, sizeof(struct in6_addr));
3181 	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3182 	if (idev->dev->addr_len == sizeof(struct in6_addr))
3183 		offset = sizeof(struct in6_addr) - 4;
3184 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3185 
3186 	if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3187 		scope = IPV6_ADDR_COMPATv4;
3188 		plen = 96;
3189 		pflags |= RTF_NONEXTHOP;
3190 	} else {
3191 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3192 			return;
3193 
3194 		addr.s6_addr32[0] = htonl(0xfe800000);
3195 		scope = IFA_LINK;
3196 		plen = 64;
3197 	}
3198 
3199 	if (addr.s6_addr32[3]) {
3200 		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3201 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3202 				      GFP_KERNEL);
3203 		return;
3204 	}
3205 
3206 	for_each_netdev(net, dev) {
3207 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3208 		if (in_dev && (dev->flags & IFF_UP)) {
3209 			struct in_ifaddr *ifa;
3210 			int flag = scope;
3211 
3212 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3213 				addr.s6_addr32[3] = ifa->ifa_local;
3214 
3215 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3216 					continue;
3217 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3218 					if (idev->dev->flags&IFF_POINTOPOINT)
3219 						continue;
3220 					flag |= IFA_HOST;
3221 				}
3222 
3223 				add_addr(idev, &addr, plen, flag,
3224 					 IFAPROT_UNSPEC);
3225 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3226 						      0, pflags, GFP_KERNEL);
3227 			}
3228 		}
3229 	}
3230 }
3231 #endif
3232 
3233 static void init_loopback(struct net_device *dev)
3234 {
3235 	struct inet6_dev  *idev;
3236 
3237 	/* ::1 */
3238 
3239 	ASSERT_RTNL();
3240 
3241 	idev = ipv6_find_idev(dev);
3242 	if (IS_ERR(idev)) {
3243 		pr_debug("%s: add_dev failed\n", __func__);
3244 		return;
3245 	}
3246 
3247 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3248 }
3249 
3250 void addrconf_add_linklocal(struct inet6_dev *idev,
3251 			    const struct in6_addr *addr, u32 flags)
3252 {
3253 	struct ifa6_config cfg = {
3254 		.pfx = addr,
3255 		.plen = 64,
3256 		.ifa_flags = flags | IFA_F_PERMANENT,
3257 		.valid_lft = INFINITY_LIFE_TIME,
3258 		.preferred_lft = INFINITY_LIFE_TIME,
3259 		.scope = IFA_LINK,
3260 		.ifa_proto = IFAPROT_KERNEL_LL
3261 	};
3262 	struct inet6_ifaddr *ifp;
3263 
3264 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3265 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3266 	     idev->cnf.optimistic_dad) &&
3267 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3268 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3269 #endif
3270 
3271 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3272 	if (!IS_ERR(ifp)) {
3273 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3274 				      0, 0, GFP_ATOMIC);
3275 		addrconf_dad_start(ifp);
3276 		in6_ifa_put(ifp);
3277 	}
3278 }
3279 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3280 
3281 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3282 {
3283 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3284 		return true;
3285 
3286 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3287 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3288 		return true;
3289 
3290 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3291 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3292 		return true;
3293 
3294 	return false;
3295 }
3296 
3297 static int ipv6_generate_stable_address(struct in6_addr *address,
3298 					u8 dad_count,
3299 					const struct inet6_dev *idev)
3300 {
3301 	static DEFINE_SPINLOCK(lock);
3302 	static __u32 digest[SHA1_DIGEST_WORDS];
3303 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3304 
3305 	static union {
3306 		char __data[SHA1_BLOCK_SIZE];
3307 		struct {
3308 			struct in6_addr secret;
3309 			__be32 prefix[2];
3310 			unsigned char hwaddr[MAX_ADDR_LEN];
3311 			u8 dad_count;
3312 		} __packed;
3313 	} data;
3314 
3315 	struct in6_addr secret;
3316 	struct in6_addr temp;
3317 	struct net *net = dev_net(idev->dev);
3318 
3319 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3320 
3321 	if (idev->cnf.stable_secret.initialized)
3322 		secret = idev->cnf.stable_secret.secret;
3323 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3324 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3325 	else
3326 		return -1;
3327 
3328 retry:
3329 	spin_lock_bh(&lock);
3330 
3331 	sha1_init(digest);
3332 	memset(&data, 0, sizeof(data));
3333 	memset(workspace, 0, sizeof(workspace));
3334 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3335 	data.prefix[0] = address->s6_addr32[0];
3336 	data.prefix[1] = address->s6_addr32[1];
3337 	data.secret = secret;
3338 	data.dad_count = dad_count;
3339 
3340 	sha1_transform(digest, data.__data, workspace);
3341 
3342 	temp = *address;
3343 	temp.s6_addr32[2] = (__force __be32)digest[0];
3344 	temp.s6_addr32[3] = (__force __be32)digest[1];
3345 
3346 	spin_unlock_bh(&lock);
3347 
3348 	if (ipv6_reserved_interfaceid(temp)) {
3349 		dad_count++;
3350 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3351 			return -1;
3352 		goto retry;
3353 	}
3354 
3355 	*address = temp;
3356 	return 0;
3357 }
3358 
3359 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3360 {
3361 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3362 
3363 	if (s->initialized)
3364 		return;
3365 	s = &idev->cnf.stable_secret;
3366 	get_random_bytes(&s->secret, sizeof(s->secret));
3367 	s->initialized = true;
3368 }
3369 
3370 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3371 {
3372 	struct in6_addr addr;
3373 
3374 	/* no link local addresses on L3 master devices */
3375 	if (netif_is_l3_master(idev->dev))
3376 		return;
3377 
3378 	/* no link local addresses on devices flagged as slaves */
3379 	if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3380 		return;
3381 
3382 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3383 
3384 	switch (idev->cnf.addr_gen_mode) {
3385 	case IN6_ADDR_GEN_MODE_RANDOM:
3386 		ipv6_gen_mode_random_init(idev);
3387 		fallthrough;
3388 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3389 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3390 			addrconf_add_linklocal(idev, &addr,
3391 					       IFA_F_STABLE_PRIVACY);
3392 		else if (prefix_route)
3393 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3394 					      0, 0, GFP_KERNEL);
3395 		break;
3396 	case IN6_ADDR_GEN_MODE_EUI64:
3397 		/* addrconf_add_linklocal also adds a prefix_route and we
3398 		 * only need to care about prefix routes if ipv6_generate_eui64
3399 		 * couldn't generate one.
3400 		 */
3401 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3402 			addrconf_add_linklocal(idev, &addr, 0);
3403 		else if (prefix_route)
3404 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3405 					      0, 0, GFP_KERNEL);
3406 		break;
3407 	case IN6_ADDR_GEN_MODE_NONE:
3408 	default:
3409 		/* will not add any link local address */
3410 		break;
3411 	}
3412 }
3413 
3414 static void addrconf_dev_config(struct net_device *dev)
3415 {
3416 	struct inet6_dev *idev;
3417 
3418 	ASSERT_RTNL();
3419 
3420 	if ((dev->type != ARPHRD_ETHER) &&
3421 	    (dev->type != ARPHRD_FDDI) &&
3422 	    (dev->type != ARPHRD_ARCNET) &&
3423 	    (dev->type != ARPHRD_INFINIBAND) &&
3424 	    (dev->type != ARPHRD_IEEE1394) &&
3425 	    (dev->type != ARPHRD_TUNNEL6) &&
3426 	    (dev->type != ARPHRD_6LOWPAN) &&
3427 	    (dev->type != ARPHRD_TUNNEL) &&
3428 	    (dev->type != ARPHRD_NONE) &&
3429 	    (dev->type != ARPHRD_RAWIP)) {
3430 		/* Alas, we support only Ethernet autoconfiguration. */
3431 		idev = __in6_dev_get(dev);
3432 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3433 		    dev->flags & IFF_MULTICAST)
3434 			ipv6_mc_up(idev);
3435 		return;
3436 	}
3437 
3438 	idev = addrconf_add_dev(dev);
3439 	if (IS_ERR(idev))
3440 		return;
3441 
3442 	/* this device type has no EUI support */
3443 	if (dev->type == ARPHRD_NONE &&
3444 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3445 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3446 
3447 	addrconf_addr_gen(idev, false);
3448 }
3449 
3450 #if IS_ENABLED(CONFIG_IPV6_SIT)
3451 static void addrconf_sit_config(struct net_device *dev)
3452 {
3453 	struct inet6_dev *idev;
3454 
3455 	ASSERT_RTNL();
3456 
3457 	/*
3458 	 * Configure the tunnel with one of our IPv4
3459 	 * addresses... we should configure all of
3460 	 * our v4 addrs in the tunnel
3461 	 */
3462 
3463 	idev = ipv6_find_idev(dev);
3464 	if (IS_ERR(idev)) {
3465 		pr_debug("%s: add_dev failed\n", __func__);
3466 		return;
3467 	}
3468 
3469 	if (dev->priv_flags & IFF_ISATAP) {
3470 		addrconf_addr_gen(idev, false);
3471 		return;
3472 	}
3473 
3474 	add_v4_addrs(idev);
3475 
3476 	if (dev->flags&IFF_POINTOPOINT)
3477 		addrconf_add_mroute(dev);
3478 }
3479 #endif
3480 
3481 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3482 static void addrconf_gre_config(struct net_device *dev)
3483 {
3484 	struct inet6_dev *idev;
3485 
3486 	ASSERT_RTNL();
3487 
3488 	idev = ipv6_find_idev(dev);
3489 	if (IS_ERR(idev)) {
3490 		pr_debug("%s: add_dev failed\n", __func__);
3491 		return;
3492 	}
3493 
3494 	if (dev->type == ARPHRD_ETHER) {
3495 		addrconf_addr_gen(idev, true);
3496 		return;
3497 	}
3498 
3499 	add_v4_addrs(idev);
3500 
3501 	if (dev->flags & IFF_POINTOPOINT)
3502 		addrconf_add_mroute(dev);
3503 }
3504 #endif
3505 
3506 static void addrconf_init_auto_addrs(struct net_device *dev)
3507 {
3508 	switch (dev->type) {
3509 #if IS_ENABLED(CONFIG_IPV6_SIT)
3510 	case ARPHRD_SIT:
3511 		addrconf_sit_config(dev);
3512 		break;
3513 #endif
3514 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3515 	case ARPHRD_IP6GRE:
3516 	case ARPHRD_IPGRE:
3517 		addrconf_gre_config(dev);
3518 		break;
3519 #endif
3520 	case ARPHRD_LOOPBACK:
3521 		init_loopback(dev);
3522 		break;
3523 
3524 	default:
3525 		addrconf_dev_config(dev);
3526 		break;
3527 	}
3528 }
3529 
3530 static int fixup_permanent_addr(struct net *net,
3531 				struct inet6_dev *idev,
3532 				struct inet6_ifaddr *ifp)
3533 {
3534 	/* !fib6_node means the host route was removed from the
3535 	 * FIB, for example, if 'lo' device is taken down. In that
3536 	 * case regenerate the host route.
3537 	 */
3538 	if (!ifp->rt || !ifp->rt->fib6_node) {
3539 		struct fib6_info *f6i, *prev;
3540 
3541 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3542 					 GFP_ATOMIC, NULL);
3543 		if (IS_ERR(f6i))
3544 			return PTR_ERR(f6i);
3545 
3546 		/* ifp->rt can be accessed outside of rtnl */
3547 		spin_lock(&ifp->lock);
3548 		prev = ifp->rt;
3549 		ifp->rt = f6i;
3550 		spin_unlock(&ifp->lock);
3551 
3552 		fib6_info_release(prev);
3553 	}
3554 
3555 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3556 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3557 				      ifp->rt_priority, idev->dev, 0, 0,
3558 				      GFP_ATOMIC);
3559 	}
3560 
3561 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3562 		addrconf_dad_start(ifp);
3563 
3564 	return 0;
3565 }
3566 
3567 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3568 {
3569 	struct inet6_ifaddr *ifp, *tmp;
3570 	struct inet6_dev *idev;
3571 
3572 	idev = __in6_dev_get(dev);
3573 	if (!idev)
3574 		return;
3575 
3576 	write_lock_bh(&idev->lock);
3577 
3578 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3579 		if ((ifp->flags & IFA_F_PERMANENT) &&
3580 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3581 			write_unlock_bh(&idev->lock);
3582 			in6_ifa_hold(ifp);
3583 			ipv6_del_addr(ifp);
3584 			write_lock_bh(&idev->lock);
3585 
3586 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3587 					     idev->dev->name, &ifp->addr);
3588 		}
3589 	}
3590 
3591 	write_unlock_bh(&idev->lock);
3592 }
3593 
3594 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3595 			   void *ptr)
3596 {
3597 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3598 	struct netdev_notifier_change_info *change_info;
3599 	struct netdev_notifier_changeupper_info *info;
3600 	struct inet6_dev *idev = __in6_dev_get(dev);
3601 	struct net *net = dev_net(dev);
3602 	int run_pending = 0;
3603 	int err;
3604 
3605 	switch (event) {
3606 	case NETDEV_REGISTER:
3607 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3608 			idev = ipv6_add_dev(dev);
3609 			if (IS_ERR(idev))
3610 				return notifier_from_errno(PTR_ERR(idev));
3611 		}
3612 		break;
3613 
3614 	case NETDEV_CHANGEMTU:
3615 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3616 		if (dev->mtu < IPV6_MIN_MTU) {
3617 			addrconf_ifdown(dev, dev != net->loopback_dev);
3618 			break;
3619 		}
3620 
3621 		if (idev) {
3622 			rt6_mtu_change(dev, dev->mtu);
3623 			idev->cnf.mtu6 = dev->mtu;
3624 			break;
3625 		}
3626 
3627 		/* allocate new idev */
3628 		idev = ipv6_add_dev(dev);
3629 		if (IS_ERR(idev))
3630 			break;
3631 
3632 		/* device is still not ready */
3633 		if (!(idev->if_flags & IF_READY))
3634 			break;
3635 
3636 		run_pending = 1;
3637 		fallthrough;
3638 	case NETDEV_UP:
3639 	case NETDEV_CHANGE:
3640 		if (idev && idev->cnf.disable_ipv6)
3641 			break;
3642 
3643 		if (dev->priv_flags & IFF_NO_ADDRCONF) {
3644 			if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3645 			    dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3646 				ipv6_mc_up(idev);
3647 			break;
3648 		}
3649 
3650 		if (event == NETDEV_UP) {
3651 			/* restore routes for permanent addresses */
3652 			addrconf_permanent_addr(net, dev);
3653 
3654 			if (!addrconf_link_ready(dev)) {
3655 				/* device is not ready yet. */
3656 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3657 					 dev->name);
3658 				break;
3659 			}
3660 
3661 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3662 				idev = ipv6_add_dev(dev);
3663 
3664 			if (!IS_ERR_OR_NULL(idev)) {
3665 				idev->if_flags |= IF_READY;
3666 				run_pending = 1;
3667 			}
3668 		} else if (event == NETDEV_CHANGE) {
3669 			if (!addrconf_link_ready(dev)) {
3670 				/* device is still not ready. */
3671 				rt6_sync_down_dev(dev, event);
3672 				break;
3673 			}
3674 
3675 			if (!IS_ERR_OR_NULL(idev)) {
3676 				if (idev->if_flags & IF_READY) {
3677 					/* device is already configured -
3678 					 * but resend MLD reports, we might
3679 					 * have roamed and need to update
3680 					 * multicast snooping switches
3681 					 */
3682 					ipv6_mc_up(idev);
3683 					change_info = ptr;
3684 					if (change_info->flags_changed & IFF_NOARP)
3685 						addrconf_dad_run(idev, true);
3686 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3687 					break;
3688 				}
3689 				idev->if_flags |= IF_READY;
3690 			}
3691 
3692 			pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3693 				 dev->name);
3694 
3695 			run_pending = 1;
3696 		}
3697 
3698 		addrconf_init_auto_addrs(dev);
3699 
3700 		if (!IS_ERR_OR_NULL(idev)) {
3701 			if (run_pending)
3702 				addrconf_dad_run(idev, false);
3703 
3704 			/* Device has an address by now */
3705 			rt6_sync_up(dev, RTNH_F_DEAD);
3706 
3707 			/*
3708 			 * If the MTU changed during the interface down,
3709 			 * when the interface up, the changed MTU must be
3710 			 * reflected in the idev as well as routers.
3711 			 */
3712 			if (idev->cnf.mtu6 != dev->mtu &&
3713 			    dev->mtu >= IPV6_MIN_MTU) {
3714 				rt6_mtu_change(dev, dev->mtu);
3715 				idev->cnf.mtu6 = dev->mtu;
3716 			}
3717 			idev->tstamp = jiffies;
3718 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3719 
3720 			/*
3721 			 * If the changed mtu during down is lower than
3722 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3723 			 */
3724 			if (dev->mtu < IPV6_MIN_MTU)
3725 				addrconf_ifdown(dev, dev != net->loopback_dev);
3726 		}
3727 		break;
3728 
3729 	case NETDEV_DOWN:
3730 	case NETDEV_UNREGISTER:
3731 		/*
3732 		 *	Remove all addresses from this interface.
3733 		 */
3734 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3735 		break;
3736 
3737 	case NETDEV_CHANGENAME:
3738 		if (idev) {
3739 			snmp6_unregister_dev(idev);
3740 			addrconf_sysctl_unregister(idev);
3741 			err = addrconf_sysctl_register(idev);
3742 			if (err)
3743 				return notifier_from_errno(err);
3744 			err = snmp6_register_dev(idev);
3745 			if (err) {
3746 				addrconf_sysctl_unregister(idev);
3747 				return notifier_from_errno(err);
3748 			}
3749 		}
3750 		break;
3751 
3752 	case NETDEV_PRE_TYPE_CHANGE:
3753 	case NETDEV_POST_TYPE_CHANGE:
3754 		if (idev)
3755 			addrconf_type_change(dev, event);
3756 		break;
3757 
3758 	case NETDEV_CHANGEUPPER:
3759 		info = ptr;
3760 
3761 		/* flush all routes if dev is linked to or unlinked from
3762 		 * an L3 master device (e.g., VRF)
3763 		 */
3764 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3765 			addrconf_ifdown(dev, false);
3766 	}
3767 
3768 	return NOTIFY_OK;
3769 }
3770 
3771 /*
3772  *	addrconf module should be notified of a device going up
3773  */
3774 static struct notifier_block ipv6_dev_notf = {
3775 	.notifier_call = addrconf_notify,
3776 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3777 };
3778 
3779 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3780 {
3781 	struct inet6_dev *idev;
3782 	ASSERT_RTNL();
3783 
3784 	idev = __in6_dev_get(dev);
3785 
3786 	if (event == NETDEV_POST_TYPE_CHANGE)
3787 		ipv6_mc_remap(idev);
3788 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3789 		ipv6_mc_unmap(idev);
3790 }
3791 
3792 static bool addr_is_local(const struct in6_addr *addr)
3793 {
3794 	return ipv6_addr_type(addr) &
3795 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3796 }
3797 
3798 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3799 {
3800 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3801 	struct net *net = dev_net(dev);
3802 	struct inet6_dev *idev;
3803 	struct inet6_ifaddr *ifa;
3804 	LIST_HEAD(tmp_addr_list);
3805 	bool keep_addr = false;
3806 	bool was_ready;
3807 	int state, i;
3808 
3809 	ASSERT_RTNL();
3810 
3811 	rt6_disable_ip(dev, event);
3812 
3813 	idev = __in6_dev_get(dev);
3814 	if (!idev)
3815 		return -ENODEV;
3816 
3817 	/*
3818 	 * Step 1: remove reference to ipv6 device from parent device.
3819 	 *	   Do not dev_put!
3820 	 */
3821 	if (unregister) {
3822 		idev->dead = 1;
3823 
3824 		/* protected by rtnl_lock */
3825 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3826 
3827 		/* Step 1.5: remove snmp6 entry */
3828 		snmp6_unregister_dev(idev);
3829 
3830 	}
3831 
3832 	/* combine the user config with event to determine if permanent
3833 	 * addresses are to be removed from address hash table
3834 	 */
3835 	if (!unregister && !idev->cnf.disable_ipv6) {
3836 		/* aggregate the system setting and interface setting */
3837 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3838 
3839 		if (!_keep_addr)
3840 			_keep_addr = idev->cnf.keep_addr_on_down;
3841 
3842 		keep_addr = (_keep_addr > 0);
3843 	}
3844 
3845 	/* Step 2: clear hash table */
3846 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3847 		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3848 
3849 		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3850 restart:
3851 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3852 			if (ifa->idev == idev) {
3853 				addrconf_del_dad_work(ifa);
3854 				/* combined flag + permanent flag decide if
3855 				 * address is retained on a down event
3856 				 */
3857 				if (!keep_addr ||
3858 				    !(ifa->flags & IFA_F_PERMANENT) ||
3859 				    addr_is_local(&ifa->addr)) {
3860 					hlist_del_init_rcu(&ifa->addr_lst);
3861 					goto restart;
3862 				}
3863 			}
3864 		}
3865 		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3866 	}
3867 
3868 	write_lock_bh(&idev->lock);
3869 
3870 	addrconf_del_rs_timer(idev);
3871 
3872 	/* Step 2: clear flags for stateless addrconf, repeated down
3873 	 *         detection
3874 	 */
3875 	was_ready = idev->if_flags & IF_READY;
3876 	if (!unregister)
3877 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3878 
3879 	/* Step 3: clear tempaddr list */
3880 	while (!list_empty(&idev->tempaddr_list)) {
3881 		ifa = list_first_entry(&idev->tempaddr_list,
3882 				       struct inet6_ifaddr, tmp_list);
3883 		list_del(&ifa->tmp_list);
3884 		write_unlock_bh(&idev->lock);
3885 		spin_lock_bh(&ifa->lock);
3886 
3887 		if (ifa->ifpub) {
3888 			in6_ifa_put(ifa->ifpub);
3889 			ifa->ifpub = NULL;
3890 		}
3891 		spin_unlock_bh(&ifa->lock);
3892 		in6_ifa_put(ifa);
3893 		write_lock_bh(&idev->lock);
3894 	}
3895 
3896 	list_for_each_entry(ifa, &idev->addr_list, if_list)
3897 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3898 	write_unlock_bh(&idev->lock);
3899 
3900 	while (!list_empty(&tmp_addr_list)) {
3901 		struct fib6_info *rt = NULL;
3902 		bool keep;
3903 
3904 		ifa = list_first_entry(&tmp_addr_list,
3905 				       struct inet6_ifaddr, if_list_aux);
3906 		list_del(&ifa->if_list_aux);
3907 
3908 		addrconf_del_dad_work(ifa);
3909 
3910 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3911 			!addr_is_local(&ifa->addr);
3912 
3913 		spin_lock_bh(&ifa->lock);
3914 
3915 		if (keep) {
3916 			/* set state to skip the notifier below */
3917 			state = INET6_IFADDR_STATE_DEAD;
3918 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3919 			if (!(ifa->flags & IFA_F_NODAD))
3920 				ifa->flags |= IFA_F_TENTATIVE;
3921 
3922 			rt = ifa->rt;
3923 			ifa->rt = NULL;
3924 		} else {
3925 			state = ifa->state;
3926 			ifa->state = INET6_IFADDR_STATE_DEAD;
3927 		}
3928 
3929 		spin_unlock_bh(&ifa->lock);
3930 
3931 		if (rt)
3932 			ip6_del_rt(net, rt, false);
3933 
3934 		if (state != INET6_IFADDR_STATE_DEAD) {
3935 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3936 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3937 		} else {
3938 			if (idev->cnf.forwarding)
3939 				addrconf_leave_anycast(ifa);
3940 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3941 		}
3942 
3943 		if (!keep) {
3944 			write_lock_bh(&idev->lock);
3945 			list_del_rcu(&ifa->if_list);
3946 			write_unlock_bh(&idev->lock);
3947 			in6_ifa_put(ifa);
3948 		}
3949 	}
3950 
3951 	/* Step 5: Discard anycast and multicast list */
3952 	if (unregister) {
3953 		ipv6_ac_destroy_dev(idev);
3954 		ipv6_mc_destroy_dev(idev);
3955 	} else if (was_ready) {
3956 		ipv6_mc_down(idev);
3957 	}
3958 
3959 	idev->tstamp = jiffies;
3960 	idev->ra_mtu = 0;
3961 
3962 	/* Last: Shot the device (if unregistered) */
3963 	if (unregister) {
3964 		addrconf_sysctl_unregister(idev);
3965 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3966 		neigh_ifdown(&nd_tbl, dev);
3967 		in6_dev_put(idev);
3968 	}
3969 	return 0;
3970 }
3971 
3972 static void addrconf_rs_timer(struct timer_list *t)
3973 {
3974 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3975 	struct net_device *dev = idev->dev;
3976 	struct in6_addr lladdr;
3977 
3978 	write_lock(&idev->lock);
3979 	if (idev->dead || !(idev->if_flags & IF_READY))
3980 		goto out;
3981 
3982 	if (!ipv6_accept_ra(idev))
3983 		goto out;
3984 
3985 	/* Announcement received after solicitation was sent */
3986 	if (idev->if_flags & IF_RA_RCVD)
3987 		goto out;
3988 
3989 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3990 		write_unlock(&idev->lock);
3991 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3992 			ndisc_send_rs(dev, &lladdr,
3993 				      &in6addr_linklocal_allrouters);
3994 		else
3995 			goto put;
3996 
3997 		write_lock(&idev->lock);
3998 		idev->rs_interval = rfc3315_s14_backoff_update(
3999 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
4000 		/* The wait after the last probe can be shorter */
4001 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4002 					     idev->cnf.rtr_solicits) ?
4003 				      idev->cnf.rtr_solicit_delay :
4004 				      idev->rs_interval);
4005 	} else {
4006 		/*
4007 		 * Note: we do not support deprecated "all on-link"
4008 		 * assumption any longer.
4009 		 */
4010 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4011 	}
4012 
4013 out:
4014 	write_unlock(&idev->lock);
4015 put:
4016 	in6_dev_put(idev);
4017 }
4018 
4019 /*
4020  *	Duplicate Address Detection
4021  */
4022 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4023 {
4024 	unsigned long rand_num;
4025 	struct inet6_dev *idev = ifp->idev;
4026 	u64 nonce;
4027 
4028 	if (ifp->flags & IFA_F_OPTIMISTIC)
4029 		rand_num = 0;
4030 	else
4031 		rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1);
4032 
4033 	nonce = 0;
4034 	if (idev->cnf.enhanced_dad ||
4035 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
4036 		do
4037 			get_random_bytes(&nonce, 6);
4038 		while (nonce == 0);
4039 	}
4040 	ifp->dad_nonce = nonce;
4041 	ifp->dad_probes = idev->cnf.dad_transmits;
4042 	addrconf_mod_dad_work(ifp, rand_num);
4043 }
4044 
4045 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4046 {
4047 	struct inet6_dev *idev = ifp->idev;
4048 	struct net_device *dev = idev->dev;
4049 	bool bump_id, notify = false;
4050 	struct net *net;
4051 
4052 	addrconf_join_solict(dev, &ifp->addr);
4053 
4054 	read_lock_bh(&idev->lock);
4055 	spin_lock(&ifp->lock);
4056 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
4057 		goto out;
4058 
4059 	net = dev_net(dev);
4060 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4061 	    (net->ipv6.devconf_all->accept_dad < 1 &&
4062 	     idev->cnf.accept_dad < 1) ||
4063 	    !(ifp->flags&IFA_F_TENTATIVE) ||
4064 	    ifp->flags & IFA_F_NODAD) {
4065 		bool send_na = false;
4066 
4067 		if (ifp->flags & IFA_F_TENTATIVE &&
4068 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4069 			send_na = true;
4070 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4071 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4072 		spin_unlock(&ifp->lock);
4073 		read_unlock_bh(&idev->lock);
4074 
4075 		addrconf_dad_completed(ifp, bump_id, send_na);
4076 		return;
4077 	}
4078 
4079 	if (!(idev->if_flags & IF_READY)) {
4080 		spin_unlock(&ifp->lock);
4081 		read_unlock_bh(&idev->lock);
4082 		/*
4083 		 * If the device is not ready:
4084 		 * - keep it tentative if it is a permanent address.
4085 		 * - otherwise, kill it.
4086 		 */
4087 		in6_ifa_hold(ifp);
4088 		addrconf_dad_stop(ifp, 0);
4089 		return;
4090 	}
4091 
4092 	/*
4093 	 * Optimistic nodes can start receiving
4094 	 * Frames right away
4095 	 */
4096 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4097 		ip6_ins_rt(net, ifp->rt);
4098 		if (ipv6_use_optimistic_addr(net, idev)) {
4099 			/* Because optimistic nodes can use this address,
4100 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4101 			 */
4102 			notify = true;
4103 		}
4104 	}
4105 
4106 	addrconf_dad_kick(ifp);
4107 out:
4108 	spin_unlock(&ifp->lock);
4109 	read_unlock_bh(&idev->lock);
4110 	if (notify)
4111 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4112 }
4113 
4114 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4115 {
4116 	bool begin_dad = false;
4117 
4118 	spin_lock_bh(&ifp->lock);
4119 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4120 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4121 		begin_dad = true;
4122 	}
4123 	spin_unlock_bh(&ifp->lock);
4124 
4125 	if (begin_dad)
4126 		addrconf_mod_dad_work(ifp, 0);
4127 }
4128 
4129 static void addrconf_dad_work(struct work_struct *w)
4130 {
4131 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4132 						struct inet6_ifaddr,
4133 						dad_work);
4134 	struct inet6_dev *idev = ifp->idev;
4135 	bool bump_id, disable_ipv6 = false;
4136 	struct in6_addr mcaddr;
4137 
4138 	enum {
4139 		DAD_PROCESS,
4140 		DAD_BEGIN,
4141 		DAD_ABORT,
4142 	} action = DAD_PROCESS;
4143 
4144 	rtnl_lock();
4145 
4146 	spin_lock_bh(&ifp->lock);
4147 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4148 		action = DAD_BEGIN;
4149 		ifp->state = INET6_IFADDR_STATE_DAD;
4150 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4151 		action = DAD_ABORT;
4152 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4153 
4154 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4155 		     idev->cnf.accept_dad > 1) &&
4156 		    !idev->cnf.disable_ipv6 &&
4157 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4158 			struct in6_addr addr;
4159 
4160 			addr.s6_addr32[0] = htonl(0xfe800000);
4161 			addr.s6_addr32[1] = 0;
4162 
4163 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4164 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4165 				/* DAD failed for link-local based on MAC */
4166 				idev->cnf.disable_ipv6 = 1;
4167 
4168 				pr_info("%s: IPv6 being disabled!\n",
4169 					ifp->idev->dev->name);
4170 				disable_ipv6 = true;
4171 			}
4172 		}
4173 	}
4174 	spin_unlock_bh(&ifp->lock);
4175 
4176 	if (action == DAD_BEGIN) {
4177 		addrconf_dad_begin(ifp);
4178 		goto out;
4179 	} else if (action == DAD_ABORT) {
4180 		in6_ifa_hold(ifp);
4181 		addrconf_dad_stop(ifp, 1);
4182 		if (disable_ipv6)
4183 			addrconf_ifdown(idev->dev, false);
4184 		goto out;
4185 	}
4186 
4187 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4188 		goto out;
4189 
4190 	write_lock_bh(&idev->lock);
4191 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4192 		write_unlock_bh(&idev->lock);
4193 		goto out;
4194 	}
4195 
4196 	spin_lock(&ifp->lock);
4197 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4198 		spin_unlock(&ifp->lock);
4199 		write_unlock_bh(&idev->lock);
4200 		goto out;
4201 	}
4202 
4203 	if (ifp->dad_probes == 0) {
4204 		bool send_na = false;
4205 
4206 		/*
4207 		 * DAD was successful
4208 		 */
4209 
4210 		if (ifp->flags & IFA_F_TENTATIVE &&
4211 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4212 			send_na = true;
4213 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4214 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4215 		spin_unlock(&ifp->lock);
4216 		write_unlock_bh(&idev->lock);
4217 
4218 		addrconf_dad_completed(ifp, bump_id, send_na);
4219 
4220 		goto out;
4221 	}
4222 
4223 	ifp->dad_probes--;
4224 	addrconf_mod_dad_work(ifp,
4225 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4226 				  HZ/100));
4227 	spin_unlock(&ifp->lock);
4228 	write_unlock_bh(&idev->lock);
4229 
4230 	/* send a neighbour solicitation for our addr */
4231 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4232 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4233 		      ifp->dad_nonce);
4234 out:
4235 	in6_ifa_put(ifp);
4236 	rtnl_unlock();
4237 }
4238 
4239 /* ifp->idev must be at least read locked */
4240 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4241 {
4242 	struct inet6_ifaddr *ifpiter;
4243 	struct inet6_dev *idev = ifp->idev;
4244 
4245 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4246 		if (ifpiter->scope > IFA_LINK)
4247 			break;
4248 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4249 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4250 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4251 		    IFA_F_PERMANENT)
4252 			return false;
4253 	}
4254 	return true;
4255 }
4256 
4257 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4258 				   bool send_na)
4259 {
4260 	struct net_device *dev = ifp->idev->dev;
4261 	struct in6_addr lladdr;
4262 	bool send_rs, send_mld;
4263 
4264 	addrconf_del_dad_work(ifp);
4265 
4266 	/*
4267 	 *	Configure the address for reception. Now it is valid.
4268 	 */
4269 
4270 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4271 
4272 	/* If added prefix is link local and we are prepared to process
4273 	   router advertisements, start sending router solicitations.
4274 	 */
4275 
4276 	read_lock_bh(&ifp->idev->lock);
4277 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4278 	send_rs = send_mld &&
4279 		  ipv6_accept_ra(ifp->idev) &&
4280 		  ifp->idev->cnf.rtr_solicits != 0 &&
4281 		  (dev->flags & IFF_LOOPBACK) == 0 &&
4282 		  (dev->type != ARPHRD_TUNNEL) &&
4283 		  !netif_is_team_port(dev);
4284 	read_unlock_bh(&ifp->idev->lock);
4285 
4286 	/* While dad is in progress mld report's source address is in6_addrany.
4287 	 * Resend with proper ll now.
4288 	 */
4289 	if (send_mld)
4290 		ipv6_mc_dad_complete(ifp->idev);
4291 
4292 	/* send unsolicited NA if enabled */
4293 	if (send_na &&
4294 	    (ifp->idev->cnf.ndisc_notify ||
4295 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4296 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4297 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4298 			      /*solicited=*/ false, /*override=*/ true,
4299 			      /*inc_opt=*/ true);
4300 	}
4301 
4302 	if (send_rs) {
4303 		/*
4304 		 *	If a host as already performed a random delay
4305 		 *	[...] as part of DAD [...] there is no need
4306 		 *	to delay again before sending the first RS
4307 		 */
4308 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4309 			return;
4310 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4311 
4312 		write_lock_bh(&ifp->idev->lock);
4313 		spin_lock(&ifp->lock);
4314 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4315 			ifp->idev->cnf.rtr_solicit_interval);
4316 		ifp->idev->rs_probes = 1;
4317 		ifp->idev->if_flags |= IF_RS_SENT;
4318 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4319 		spin_unlock(&ifp->lock);
4320 		write_unlock_bh(&ifp->idev->lock);
4321 	}
4322 
4323 	if (bump_id)
4324 		rt_genid_bump_ipv6(dev_net(dev));
4325 
4326 	/* Make sure that a new temporary address will be created
4327 	 * before this temporary address becomes deprecated.
4328 	 */
4329 	if (ifp->flags & IFA_F_TEMPORARY)
4330 		addrconf_verify_rtnl(dev_net(dev));
4331 }
4332 
4333 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4334 {
4335 	struct inet6_ifaddr *ifp;
4336 
4337 	read_lock_bh(&idev->lock);
4338 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4339 		spin_lock(&ifp->lock);
4340 		if ((ifp->flags & IFA_F_TENTATIVE &&
4341 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4342 			if (restart)
4343 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4344 			addrconf_dad_kick(ifp);
4345 		}
4346 		spin_unlock(&ifp->lock);
4347 	}
4348 	read_unlock_bh(&idev->lock);
4349 }
4350 
4351 #ifdef CONFIG_PROC_FS
4352 struct if6_iter_state {
4353 	struct seq_net_private p;
4354 	int bucket;
4355 	int offset;
4356 };
4357 
4358 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4359 {
4360 	struct if6_iter_state *state = seq->private;
4361 	struct net *net = seq_file_net(seq);
4362 	struct inet6_ifaddr *ifa = NULL;
4363 	int p = 0;
4364 
4365 	/* initial bucket if pos is 0 */
4366 	if (pos == 0) {
4367 		state->bucket = 0;
4368 		state->offset = 0;
4369 	}
4370 
4371 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4372 		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4373 					 addr_lst) {
4374 			/* sync with offset */
4375 			if (p < state->offset) {
4376 				p++;
4377 				continue;
4378 			}
4379 			return ifa;
4380 		}
4381 
4382 		/* prepare for next bucket */
4383 		state->offset = 0;
4384 		p = 0;
4385 	}
4386 	return NULL;
4387 }
4388 
4389 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4390 					 struct inet6_ifaddr *ifa)
4391 {
4392 	struct if6_iter_state *state = seq->private;
4393 	struct net *net = seq_file_net(seq);
4394 
4395 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4396 		state->offset++;
4397 		return ifa;
4398 	}
4399 
4400 	state->offset = 0;
4401 	while (++state->bucket < IN6_ADDR_HSIZE) {
4402 		hlist_for_each_entry_rcu(ifa,
4403 				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4404 			return ifa;
4405 		}
4406 	}
4407 
4408 	return NULL;
4409 }
4410 
4411 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4412 	__acquires(rcu)
4413 {
4414 	rcu_read_lock();
4415 	return if6_get_first(seq, *pos);
4416 }
4417 
4418 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4419 {
4420 	struct inet6_ifaddr *ifa;
4421 
4422 	ifa = if6_get_next(seq, v);
4423 	++*pos;
4424 	return ifa;
4425 }
4426 
4427 static void if6_seq_stop(struct seq_file *seq, void *v)
4428 	__releases(rcu)
4429 {
4430 	rcu_read_unlock();
4431 }
4432 
4433 static int if6_seq_show(struct seq_file *seq, void *v)
4434 {
4435 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4436 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4437 		   &ifp->addr,
4438 		   ifp->idev->dev->ifindex,
4439 		   ifp->prefix_len,
4440 		   ifp->scope,
4441 		   (u8) ifp->flags,
4442 		   ifp->idev->dev->name);
4443 	return 0;
4444 }
4445 
4446 static const struct seq_operations if6_seq_ops = {
4447 	.start	= if6_seq_start,
4448 	.next	= if6_seq_next,
4449 	.show	= if6_seq_show,
4450 	.stop	= if6_seq_stop,
4451 };
4452 
4453 static int __net_init if6_proc_net_init(struct net *net)
4454 {
4455 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4456 			sizeof(struct if6_iter_state)))
4457 		return -ENOMEM;
4458 	return 0;
4459 }
4460 
4461 static void __net_exit if6_proc_net_exit(struct net *net)
4462 {
4463 	remove_proc_entry("if_inet6", net->proc_net);
4464 }
4465 
4466 static struct pernet_operations if6_proc_net_ops = {
4467 	.init = if6_proc_net_init,
4468 	.exit = if6_proc_net_exit,
4469 };
4470 
4471 int __init if6_proc_init(void)
4472 {
4473 	return register_pernet_subsys(&if6_proc_net_ops);
4474 }
4475 
4476 void if6_proc_exit(void)
4477 {
4478 	unregister_pernet_subsys(&if6_proc_net_ops);
4479 }
4480 #endif	/* CONFIG_PROC_FS */
4481 
4482 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4483 /* Check if address is a home address configured on any interface. */
4484 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4485 {
4486 	unsigned int hash = inet6_addr_hash(net, addr);
4487 	struct inet6_ifaddr *ifp = NULL;
4488 	int ret = 0;
4489 
4490 	rcu_read_lock();
4491 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4492 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4493 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4494 			ret = 1;
4495 			break;
4496 		}
4497 	}
4498 	rcu_read_unlock();
4499 	return ret;
4500 }
4501 #endif
4502 
4503 /* RFC6554 has some algorithm to avoid loops in segment routing by
4504  * checking if the segments contains any of a local interface address.
4505  *
4506  * Quote:
4507  *
4508  * To detect loops in the SRH, a router MUST determine if the SRH
4509  * includes multiple addresses assigned to any interface on that router.
4510  * If such addresses appear more than once and are separated by at least
4511  * one address not assigned to that router.
4512  */
4513 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4514 			  unsigned char nsegs)
4515 {
4516 	const struct in6_addr *addr;
4517 	int i, ret = 0, found = 0;
4518 	struct inet6_ifaddr *ifp;
4519 	bool separated = false;
4520 	unsigned int hash;
4521 	bool hash_found;
4522 
4523 	rcu_read_lock();
4524 	for (i = 0; i < nsegs; i++) {
4525 		addr = &segs[i];
4526 		hash = inet6_addr_hash(net, addr);
4527 
4528 		hash_found = false;
4529 		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4530 
4531 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4532 				hash_found = true;
4533 				break;
4534 			}
4535 		}
4536 
4537 		if (hash_found) {
4538 			if (found > 1 && separated) {
4539 				ret = 1;
4540 				break;
4541 			}
4542 
4543 			separated = false;
4544 			found++;
4545 		} else {
4546 			separated = true;
4547 		}
4548 	}
4549 	rcu_read_unlock();
4550 
4551 	return ret;
4552 }
4553 
4554 /*
4555  *	Periodic address status verification
4556  */
4557 
4558 static void addrconf_verify_rtnl(struct net *net)
4559 {
4560 	unsigned long now, next, next_sec, next_sched;
4561 	struct inet6_ifaddr *ifp;
4562 	int i;
4563 
4564 	ASSERT_RTNL();
4565 
4566 	rcu_read_lock_bh();
4567 	now = jiffies;
4568 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4569 
4570 	cancel_delayed_work(&net->ipv6.addr_chk_work);
4571 
4572 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4573 restart:
4574 		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4575 			unsigned long age;
4576 
4577 			/* When setting preferred_lft to a value not zero or
4578 			 * infinity, while valid_lft is infinity
4579 			 * IFA_F_PERMANENT has a non-infinity life time.
4580 			 */
4581 			if ((ifp->flags & IFA_F_PERMANENT) &&
4582 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4583 				continue;
4584 
4585 			spin_lock(&ifp->lock);
4586 			/* We try to batch several events at once. */
4587 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4588 
4589 			if ((ifp->flags&IFA_F_TEMPORARY) &&
4590 			    !(ifp->flags&IFA_F_TENTATIVE) &&
4591 			    ifp->prefered_lft != INFINITY_LIFE_TIME &&
4592 			    !ifp->regen_count && ifp->ifpub) {
4593 				/* This is a non-regenerated temporary addr. */
4594 
4595 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4596 					ifp->idev->cnf.dad_transmits *
4597 					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4598 
4599 				if (age + regen_advance >= ifp->prefered_lft) {
4600 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4601 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4602 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4603 
4604 					ifp->regen_count++;
4605 					in6_ifa_hold(ifp);
4606 					in6_ifa_hold(ifpub);
4607 					spin_unlock(&ifp->lock);
4608 
4609 					spin_lock(&ifpub->lock);
4610 					ifpub->regen_count = 0;
4611 					spin_unlock(&ifpub->lock);
4612 					rcu_read_unlock_bh();
4613 					ipv6_create_tempaddr(ifpub, true);
4614 					in6_ifa_put(ifpub);
4615 					in6_ifa_put(ifp);
4616 					rcu_read_lock_bh();
4617 					goto restart;
4618 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4619 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4620 			}
4621 
4622 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4623 			    age >= ifp->valid_lft) {
4624 				spin_unlock(&ifp->lock);
4625 				in6_ifa_hold(ifp);
4626 				rcu_read_unlock_bh();
4627 				ipv6_del_addr(ifp);
4628 				rcu_read_lock_bh();
4629 				goto restart;
4630 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4631 				spin_unlock(&ifp->lock);
4632 				continue;
4633 			} else if (age >= ifp->prefered_lft) {
4634 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4635 				int deprecate = 0;
4636 
4637 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4638 					deprecate = 1;
4639 					ifp->flags |= IFA_F_DEPRECATED;
4640 				}
4641 
4642 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4643 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4644 					next = ifp->tstamp + ifp->valid_lft * HZ;
4645 
4646 				spin_unlock(&ifp->lock);
4647 
4648 				if (deprecate) {
4649 					in6_ifa_hold(ifp);
4650 
4651 					ipv6_ifa_notify(0, ifp);
4652 					in6_ifa_put(ifp);
4653 					goto restart;
4654 				}
4655 			} else {
4656 				/* ifp->prefered_lft <= ifp->valid_lft */
4657 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4658 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4659 				spin_unlock(&ifp->lock);
4660 			}
4661 		}
4662 	}
4663 
4664 	next_sec = round_jiffies_up(next);
4665 	next_sched = next;
4666 
4667 	/* If rounded timeout is accurate enough, accept it. */
4668 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4669 		next_sched = next_sec;
4670 
4671 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4672 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4673 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4674 
4675 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4676 		 now, next, next_sec, next_sched);
4677 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4678 	rcu_read_unlock_bh();
4679 }
4680 
4681 static void addrconf_verify_work(struct work_struct *w)
4682 {
4683 	struct net *net = container_of(to_delayed_work(w), struct net,
4684 				       ipv6.addr_chk_work);
4685 
4686 	rtnl_lock();
4687 	addrconf_verify_rtnl(net);
4688 	rtnl_unlock();
4689 }
4690 
4691 static void addrconf_verify(struct net *net)
4692 {
4693 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4694 }
4695 
4696 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4697 				     struct in6_addr **peer_pfx)
4698 {
4699 	struct in6_addr *pfx = NULL;
4700 
4701 	*peer_pfx = NULL;
4702 
4703 	if (addr)
4704 		pfx = nla_data(addr);
4705 
4706 	if (local) {
4707 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4708 			*peer_pfx = pfx;
4709 		pfx = nla_data(local);
4710 	}
4711 
4712 	return pfx;
4713 }
4714 
4715 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4716 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4717 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4718 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4719 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4720 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4721 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4722 	[IFA_PROTO]		= { .type = NLA_U8 },
4723 };
4724 
4725 static int
4726 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4727 		  struct netlink_ext_ack *extack)
4728 {
4729 	struct net *net = sock_net(skb->sk);
4730 	struct ifaddrmsg *ifm;
4731 	struct nlattr *tb[IFA_MAX+1];
4732 	struct in6_addr *pfx, *peer_pfx;
4733 	u32 ifa_flags;
4734 	int err;
4735 
4736 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4737 				     ifa_ipv6_policy, extack);
4738 	if (err < 0)
4739 		return err;
4740 
4741 	ifm = nlmsg_data(nlh);
4742 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4743 	if (!pfx)
4744 		return -EINVAL;
4745 
4746 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4747 
4748 	/* We ignore other flags so far. */
4749 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4750 
4751 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4752 			      ifm->ifa_prefixlen, extack);
4753 }
4754 
4755 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4756 			       unsigned long expires, u32 flags,
4757 			       bool modify_peer)
4758 {
4759 	struct fib6_info *f6i;
4760 	u32 prio;
4761 
4762 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4763 					ifp->prefix_len,
4764 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4765 	if (!f6i)
4766 		return -ENOENT;
4767 
4768 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4769 	if (f6i->fib6_metric != prio) {
4770 		/* delete old one */
4771 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4772 
4773 		/* add new one */
4774 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4775 				      ifp->prefix_len,
4776 				      ifp->rt_priority, ifp->idev->dev,
4777 				      expires, flags, GFP_KERNEL);
4778 	} else {
4779 		if (!expires)
4780 			fib6_clean_expires(f6i);
4781 		else
4782 			fib6_set_expires(f6i, expires);
4783 
4784 		fib6_info_release(f6i);
4785 	}
4786 
4787 	return 0;
4788 }
4789 
4790 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4791 			     struct ifa6_config *cfg)
4792 {
4793 	u32 flags;
4794 	clock_t expires;
4795 	unsigned long timeout;
4796 	bool was_managetempaddr;
4797 	bool had_prefixroute;
4798 	bool new_peer = false;
4799 
4800 	ASSERT_RTNL();
4801 
4802 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4803 		return -EINVAL;
4804 
4805 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4806 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4807 		return -EINVAL;
4808 
4809 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4810 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4811 
4812 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4813 	if (addrconf_finite_timeout(timeout)) {
4814 		expires = jiffies_to_clock_t(timeout * HZ);
4815 		cfg->valid_lft = timeout;
4816 		flags = RTF_EXPIRES;
4817 	} else {
4818 		expires = 0;
4819 		flags = 0;
4820 		cfg->ifa_flags |= IFA_F_PERMANENT;
4821 	}
4822 
4823 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4824 	if (addrconf_finite_timeout(timeout)) {
4825 		if (timeout == 0)
4826 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4827 		cfg->preferred_lft = timeout;
4828 	}
4829 
4830 	if (cfg->peer_pfx &&
4831 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4832 		if (!ipv6_addr_any(&ifp->peer_addr))
4833 			cleanup_prefix_route(ifp, expires, true, true);
4834 		new_peer = true;
4835 	}
4836 
4837 	spin_lock_bh(&ifp->lock);
4838 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4839 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4840 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4841 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4842 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4843 			IFA_F_NOPREFIXROUTE);
4844 	ifp->flags |= cfg->ifa_flags;
4845 	ifp->tstamp = jiffies;
4846 	ifp->valid_lft = cfg->valid_lft;
4847 	ifp->prefered_lft = cfg->preferred_lft;
4848 	ifp->ifa_proto = cfg->ifa_proto;
4849 
4850 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4851 		ifp->rt_priority = cfg->rt_priority;
4852 
4853 	if (new_peer)
4854 		ifp->peer_addr = *cfg->peer_pfx;
4855 
4856 	spin_unlock_bh(&ifp->lock);
4857 	if (!(ifp->flags&IFA_F_TENTATIVE))
4858 		ipv6_ifa_notify(0, ifp);
4859 
4860 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4861 		int rc = -ENOENT;
4862 
4863 		if (had_prefixroute)
4864 			rc = modify_prefix_route(ifp, expires, flags, false);
4865 
4866 		/* prefix route could have been deleted; if so restore it */
4867 		if (rc == -ENOENT) {
4868 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4869 					      ifp->rt_priority, ifp->idev->dev,
4870 					      expires, flags, GFP_KERNEL);
4871 		}
4872 
4873 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4874 			rc = modify_prefix_route(ifp, expires, flags, true);
4875 
4876 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4877 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4878 					      ifp->rt_priority, ifp->idev->dev,
4879 					      expires, flags, GFP_KERNEL);
4880 		}
4881 	} else if (had_prefixroute) {
4882 		enum cleanup_prefix_rt_t action;
4883 		unsigned long rt_expires;
4884 
4885 		write_lock_bh(&ifp->idev->lock);
4886 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4887 		write_unlock_bh(&ifp->idev->lock);
4888 
4889 		if (action != CLEANUP_PREFIX_RT_NOP) {
4890 			cleanup_prefix_route(ifp, rt_expires,
4891 				action == CLEANUP_PREFIX_RT_DEL, false);
4892 		}
4893 	}
4894 
4895 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4896 		if (was_managetempaddr &&
4897 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4898 			cfg->valid_lft = 0;
4899 			cfg->preferred_lft = 0;
4900 		}
4901 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4902 				 cfg->preferred_lft, !was_managetempaddr,
4903 				 jiffies);
4904 	}
4905 
4906 	addrconf_verify_rtnl(net);
4907 
4908 	return 0;
4909 }
4910 
4911 static int
4912 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4913 		  struct netlink_ext_ack *extack)
4914 {
4915 	struct net *net = sock_net(skb->sk);
4916 	struct ifaddrmsg *ifm;
4917 	struct nlattr *tb[IFA_MAX+1];
4918 	struct in6_addr *peer_pfx;
4919 	struct inet6_ifaddr *ifa;
4920 	struct net_device *dev;
4921 	struct inet6_dev *idev;
4922 	struct ifa6_config cfg;
4923 	int err;
4924 
4925 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4926 				     ifa_ipv6_policy, extack);
4927 	if (err < 0)
4928 		return err;
4929 
4930 	memset(&cfg, 0, sizeof(cfg));
4931 
4932 	ifm = nlmsg_data(nlh);
4933 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4934 	if (!cfg.pfx)
4935 		return -EINVAL;
4936 
4937 	cfg.peer_pfx = peer_pfx;
4938 	cfg.plen = ifm->ifa_prefixlen;
4939 	if (tb[IFA_RT_PRIORITY])
4940 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4941 
4942 	if (tb[IFA_PROTO])
4943 		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4944 
4945 	cfg.valid_lft = INFINITY_LIFE_TIME;
4946 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4947 
4948 	if (tb[IFA_CACHEINFO]) {
4949 		struct ifa_cacheinfo *ci;
4950 
4951 		ci = nla_data(tb[IFA_CACHEINFO]);
4952 		cfg.valid_lft = ci->ifa_valid;
4953 		cfg.preferred_lft = ci->ifa_prefered;
4954 	}
4955 
4956 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4957 	if (!dev) {
4958 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
4959 		return -ENODEV;
4960 	}
4961 
4962 	if (tb[IFA_FLAGS])
4963 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4964 	else
4965 		cfg.ifa_flags = ifm->ifa_flags;
4966 
4967 	/* We ignore other flags so far. */
4968 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4969 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4970 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4971 
4972 	idev = ipv6_find_idev(dev);
4973 	if (IS_ERR(idev))
4974 		return PTR_ERR(idev);
4975 
4976 	if (!ipv6_allow_optimistic_dad(net, idev))
4977 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4978 
4979 	if (cfg.ifa_flags & IFA_F_NODAD &&
4980 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4981 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4982 		return -EINVAL;
4983 	}
4984 
4985 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4986 	if (!ifa) {
4987 		/*
4988 		 * It would be best to check for !NLM_F_CREATE here but
4989 		 * userspace already relies on not having to provide this.
4990 		 */
4991 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4992 	}
4993 
4994 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4995 	    !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
4996 		NL_SET_ERR_MSG_MOD(extack, "address already assigned");
4997 		err = -EEXIST;
4998 	} else {
4999 		err = inet6_addr_modify(net, ifa, &cfg);
5000 	}
5001 
5002 	in6_ifa_put(ifa);
5003 
5004 	return err;
5005 }
5006 
5007 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5008 			  u8 scope, int ifindex)
5009 {
5010 	struct ifaddrmsg *ifm;
5011 
5012 	ifm = nlmsg_data(nlh);
5013 	ifm->ifa_family = AF_INET6;
5014 	ifm->ifa_prefixlen = prefixlen;
5015 	ifm->ifa_flags = flags;
5016 	ifm->ifa_scope = scope;
5017 	ifm->ifa_index = ifindex;
5018 }
5019 
5020 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5021 			 unsigned long tstamp, u32 preferred, u32 valid)
5022 {
5023 	struct ifa_cacheinfo ci;
5024 
5025 	ci.cstamp = cstamp_delta(cstamp);
5026 	ci.tstamp = cstamp_delta(tstamp);
5027 	ci.ifa_prefered = preferred;
5028 	ci.ifa_valid = valid;
5029 
5030 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5031 }
5032 
5033 static inline int rt_scope(int ifa_scope)
5034 {
5035 	if (ifa_scope & IFA_HOST)
5036 		return RT_SCOPE_HOST;
5037 	else if (ifa_scope & IFA_LINK)
5038 		return RT_SCOPE_LINK;
5039 	else if (ifa_scope & IFA_SITE)
5040 		return RT_SCOPE_SITE;
5041 	else
5042 		return RT_SCOPE_UNIVERSE;
5043 }
5044 
5045 static inline int inet6_ifaddr_msgsize(void)
5046 {
5047 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5048 	       + nla_total_size(16) /* IFA_LOCAL */
5049 	       + nla_total_size(16) /* IFA_ADDRESS */
5050 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
5051 	       + nla_total_size(4)  /* IFA_FLAGS */
5052 	       + nla_total_size(1)  /* IFA_PROTO */
5053 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5054 }
5055 
5056 enum addr_type_t {
5057 	UNICAST_ADDR,
5058 	MULTICAST_ADDR,
5059 	ANYCAST_ADDR,
5060 };
5061 
5062 struct inet6_fill_args {
5063 	u32 portid;
5064 	u32 seq;
5065 	int event;
5066 	unsigned int flags;
5067 	int netnsid;
5068 	int ifindex;
5069 	enum addr_type_t type;
5070 };
5071 
5072 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
5073 			     struct inet6_fill_args *args)
5074 {
5075 	struct nlmsghdr  *nlh;
5076 	u32 preferred, valid;
5077 
5078 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5079 			sizeof(struct ifaddrmsg), args->flags);
5080 	if (!nlh)
5081 		return -EMSGSIZE;
5082 
5083 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5084 		      ifa->idev->dev->ifindex);
5085 
5086 	if (args->netnsid >= 0 &&
5087 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5088 		goto error;
5089 
5090 	spin_lock_bh(&ifa->lock);
5091 	if (!((ifa->flags&IFA_F_PERMANENT) &&
5092 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5093 		preferred = ifa->prefered_lft;
5094 		valid = ifa->valid_lft;
5095 		if (preferred != INFINITY_LIFE_TIME) {
5096 			long tval = (jiffies - ifa->tstamp)/HZ;
5097 			if (preferred > tval)
5098 				preferred -= tval;
5099 			else
5100 				preferred = 0;
5101 			if (valid != INFINITY_LIFE_TIME) {
5102 				if (valid > tval)
5103 					valid -= tval;
5104 				else
5105 					valid = 0;
5106 			}
5107 		}
5108 	} else {
5109 		preferred = INFINITY_LIFE_TIME;
5110 		valid = INFINITY_LIFE_TIME;
5111 	}
5112 	spin_unlock_bh(&ifa->lock);
5113 
5114 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5115 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5116 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5117 			goto error;
5118 	} else
5119 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5120 			goto error;
5121 
5122 	if (ifa->rt_priority &&
5123 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5124 		goto error;
5125 
5126 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5127 		goto error;
5128 
5129 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5130 		goto error;
5131 
5132 	if (ifa->ifa_proto &&
5133 	    nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5134 		goto error;
5135 
5136 	nlmsg_end(skb, nlh);
5137 	return 0;
5138 
5139 error:
5140 	nlmsg_cancel(skb, nlh);
5141 	return -EMSGSIZE;
5142 }
5143 
5144 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5145 			       struct inet6_fill_args *args)
5146 {
5147 	struct nlmsghdr  *nlh;
5148 	u8 scope = RT_SCOPE_UNIVERSE;
5149 	int ifindex = ifmca->idev->dev->ifindex;
5150 
5151 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5152 		scope = RT_SCOPE_SITE;
5153 
5154 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5155 			sizeof(struct ifaddrmsg), args->flags);
5156 	if (!nlh)
5157 		return -EMSGSIZE;
5158 
5159 	if (args->netnsid >= 0 &&
5160 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5161 		nlmsg_cancel(skb, nlh);
5162 		return -EMSGSIZE;
5163 	}
5164 
5165 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5166 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5167 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5168 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5169 		nlmsg_cancel(skb, nlh);
5170 		return -EMSGSIZE;
5171 	}
5172 
5173 	nlmsg_end(skb, nlh);
5174 	return 0;
5175 }
5176 
5177 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5178 			       struct inet6_fill_args *args)
5179 {
5180 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5181 	int ifindex = dev ? dev->ifindex : 1;
5182 	struct nlmsghdr  *nlh;
5183 	u8 scope = RT_SCOPE_UNIVERSE;
5184 
5185 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5186 		scope = RT_SCOPE_SITE;
5187 
5188 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5189 			sizeof(struct ifaddrmsg), args->flags);
5190 	if (!nlh)
5191 		return -EMSGSIZE;
5192 
5193 	if (args->netnsid >= 0 &&
5194 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5195 		nlmsg_cancel(skb, nlh);
5196 		return -EMSGSIZE;
5197 	}
5198 
5199 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5200 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5201 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5202 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5203 		nlmsg_cancel(skb, nlh);
5204 		return -EMSGSIZE;
5205 	}
5206 
5207 	nlmsg_end(skb, nlh);
5208 	return 0;
5209 }
5210 
5211 /* called with rcu_read_lock() */
5212 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5213 			  struct netlink_callback *cb, int s_ip_idx,
5214 			  struct inet6_fill_args *fillargs)
5215 {
5216 	struct ifmcaddr6 *ifmca;
5217 	struct ifacaddr6 *ifaca;
5218 	int ip_idx = 0;
5219 	int err = 1;
5220 
5221 	read_lock_bh(&idev->lock);
5222 	switch (fillargs->type) {
5223 	case UNICAST_ADDR: {
5224 		struct inet6_ifaddr *ifa;
5225 		fillargs->event = RTM_NEWADDR;
5226 
5227 		/* unicast address incl. temp addr */
5228 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
5229 			if (ip_idx < s_ip_idx)
5230 				goto next;
5231 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5232 			if (err < 0)
5233 				break;
5234 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5235 next:
5236 			ip_idx++;
5237 		}
5238 		break;
5239 	}
5240 	case MULTICAST_ADDR:
5241 		read_unlock_bh(&idev->lock);
5242 		fillargs->event = RTM_GETMULTICAST;
5243 
5244 		/* multicast address */
5245 		for (ifmca = rtnl_dereference(idev->mc_list);
5246 		     ifmca;
5247 		     ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5248 			if (ip_idx < s_ip_idx)
5249 				continue;
5250 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5251 			if (err < 0)
5252 				break;
5253 		}
5254 		read_lock_bh(&idev->lock);
5255 		break;
5256 	case ANYCAST_ADDR:
5257 		fillargs->event = RTM_GETANYCAST;
5258 		/* anycast address */
5259 		for (ifaca = idev->ac_list; ifaca;
5260 		     ifaca = ifaca->aca_next, ip_idx++) {
5261 			if (ip_idx < s_ip_idx)
5262 				continue;
5263 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5264 			if (err < 0)
5265 				break;
5266 		}
5267 		break;
5268 	default:
5269 		break;
5270 	}
5271 	read_unlock_bh(&idev->lock);
5272 	cb->args[2] = ip_idx;
5273 	return err;
5274 }
5275 
5276 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5277 				       struct inet6_fill_args *fillargs,
5278 				       struct net **tgt_net, struct sock *sk,
5279 				       struct netlink_callback *cb)
5280 {
5281 	struct netlink_ext_ack *extack = cb->extack;
5282 	struct nlattr *tb[IFA_MAX+1];
5283 	struct ifaddrmsg *ifm;
5284 	int err, i;
5285 
5286 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5287 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5288 		return -EINVAL;
5289 	}
5290 
5291 	ifm = nlmsg_data(nlh);
5292 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5293 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5294 		return -EINVAL;
5295 	}
5296 
5297 	fillargs->ifindex = ifm->ifa_index;
5298 	if (fillargs->ifindex) {
5299 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5300 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5301 	}
5302 
5303 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5304 					    ifa_ipv6_policy, extack);
5305 	if (err < 0)
5306 		return err;
5307 
5308 	for (i = 0; i <= IFA_MAX; ++i) {
5309 		if (!tb[i])
5310 			continue;
5311 
5312 		if (i == IFA_TARGET_NETNSID) {
5313 			struct net *net;
5314 
5315 			fillargs->netnsid = nla_get_s32(tb[i]);
5316 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5317 			if (IS_ERR(net)) {
5318 				fillargs->netnsid = -1;
5319 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5320 				return PTR_ERR(net);
5321 			}
5322 			*tgt_net = net;
5323 		} else {
5324 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5325 			return -EINVAL;
5326 		}
5327 	}
5328 
5329 	return 0;
5330 }
5331 
5332 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5333 			   enum addr_type_t type)
5334 {
5335 	const struct nlmsghdr *nlh = cb->nlh;
5336 	struct inet6_fill_args fillargs = {
5337 		.portid = NETLINK_CB(cb->skb).portid,
5338 		.seq = cb->nlh->nlmsg_seq,
5339 		.flags = NLM_F_MULTI,
5340 		.netnsid = -1,
5341 		.type = type,
5342 	};
5343 	struct net *tgt_net = sock_net(skb->sk);
5344 	int idx, s_idx, s_ip_idx;
5345 	int h, s_h;
5346 	struct net_device *dev;
5347 	struct inet6_dev *idev;
5348 	struct hlist_head *head;
5349 	int err = 0;
5350 
5351 	s_h = cb->args[0];
5352 	s_idx = idx = cb->args[1];
5353 	s_ip_idx = cb->args[2];
5354 
5355 	if (cb->strict_check) {
5356 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5357 						  skb->sk, cb);
5358 		if (err < 0)
5359 			goto put_tgt_net;
5360 
5361 		err = 0;
5362 		if (fillargs.ifindex) {
5363 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5364 			if (!dev) {
5365 				err = -ENODEV;
5366 				goto put_tgt_net;
5367 			}
5368 			idev = __in6_dev_get(dev);
5369 			if (idev) {
5370 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5371 						     &fillargs);
5372 				if (err > 0)
5373 					err = 0;
5374 			}
5375 			goto put_tgt_net;
5376 		}
5377 	}
5378 
5379 	rcu_read_lock();
5380 	cb->seq = inet6_base_seq(tgt_net);
5381 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5382 		idx = 0;
5383 		head = &tgt_net->dev_index_head[h];
5384 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5385 			if (idx < s_idx)
5386 				goto cont;
5387 			if (h > s_h || idx > s_idx)
5388 				s_ip_idx = 0;
5389 			idev = __in6_dev_get(dev);
5390 			if (!idev)
5391 				goto cont;
5392 
5393 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5394 					   &fillargs) < 0)
5395 				goto done;
5396 cont:
5397 			idx++;
5398 		}
5399 	}
5400 done:
5401 	rcu_read_unlock();
5402 	cb->args[0] = h;
5403 	cb->args[1] = idx;
5404 put_tgt_net:
5405 	if (fillargs.netnsid >= 0)
5406 		put_net(tgt_net);
5407 
5408 	return skb->len ? : err;
5409 }
5410 
5411 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5412 {
5413 	enum addr_type_t type = UNICAST_ADDR;
5414 
5415 	return inet6_dump_addr(skb, cb, type);
5416 }
5417 
5418 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5419 {
5420 	enum addr_type_t type = MULTICAST_ADDR;
5421 
5422 	return inet6_dump_addr(skb, cb, type);
5423 }
5424 
5425 
5426 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5427 {
5428 	enum addr_type_t type = ANYCAST_ADDR;
5429 
5430 	return inet6_dump_addr(skb, cb, type);
5431 }
5432 
5433 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5434 				       const struct nlmsghdr *nlh,
5435 				       struct nlattr **tb,
5436 				       struct netlink_ext_ack *extack)
5437 {
5438 	struct ifaddrmsg *ifm;
5439 	int i, err;
5440 
5441 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5442 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5443 		return -EINVAL;
5444 	}
5445 
5446 	if (!netlink_strict_get_check(skb))
5447 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5448 					      ifa_ipv6_policy, extack);
5449 
5450 	ifm = nlmsg_data(nlh);
5451 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5452 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5453 		return -EINVAL;
5454 	}
5455 
5456 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5457 					    ifa_ipv6_policy, extack);
5458 	if (err)
5459 		return err;
5460 
5461 	for (i = 0; i <= IFA_MAX; i++) {
5462 		if (!tb[i])
5463 			continue;
5464 
5465 		switch (i) {
5466 		case IFA_TARGET_NETNSID:
5467 		case IFA_ADDRESS:
5468 		case IFA_LOCAL:
5469 			break;
5470 		default:
5471 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5472 			return -EINVAL;
5473 		}
5474 	}
5475 
5476 	return 0;
5477 }
5478 
5479 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5480 			     struct netlink_ext_ack *extack)
5481 {
5482 	struct net *tgt_net = sock_net(in_skb->sk);
5483 	struct inet6_fill_args fillargs = {
5484 		.portid = NETLINK_CB(in_skb).portid,
5485 		.seq = nlh->nlmsg_seq,
5486 		.event = RTM_NEWADDR,
5487 		.flags = 0,
5488 		.netnsid = -1,
5489 	};
5490 	struct ifaddrmsg *ifm;
5491 	struct nlattr *tb[IFA_MAX+1];
5492 	struct in6_addr *addr = NULL, *peer;
5493 	struct net_device *dev = NULL;
5494 	struct inet6_ifaddr *ifa;
5495 	struct sk_buff *skb;
5496 	int err;
5497 
5498 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5499 	if (err < 0)
5500 		return err;
5501 
5502 	if (tb[IFA_TARGET_NETNSID]) {
5503 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5504 
5505 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5506 						  fillargs.netnsid);
5507 		if (IS_ERR(tgt_net))
5508 			return PTR_ERR(tgt_net);
5509 	}
5510 
5511 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5512 	if (!addr)
5513 		return -EINVAL;
5514 
5515 	ifm = nlmsg_data(nlh);
5516 	if (ifm->ifa_index)
5517 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5518 
5519 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5520 	if (!ifa) {
5521 		err = -EADDRNOTAVAIL;
5522 		goto errout;
5523 	}
5524 
5525 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5526 	if (!skb) {
5527 		err = -ENOBUFS;
5528 		goto errout_ifa;
5529 	}
5530 
5531 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5532 	if (err < 0) {
5533 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5534 		WARN_ON(err == -EMSGSIZE);
5535 		kfree_skb(skb);
5536 		goto errout_ifa;
5537 	}
5538 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5539 errout_ifa:
5540 	in6_ifa_put(ifa);
5541 errout:
5542 	dev_put(dev);
5543 	if (fillargs.netnsid >= 0)
5544 		put_net(tgt_net);
5545 
5546 	return err;
5547 }
5548 
5549 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5550 {
5551 	struct sk_buff *skb;
5552 	struct net *net = dev_net(ifa->idev->dev);
5553 	struct inet6_fill_args fillargs = {
5554 		.portid = 0,
5555 		.seq = 0,
5556 		.event = event,
5557 		.flags = 0,
5558 		.netnsid = -1,
5559 	};
5560 	int err = -ENOBUFS;
5561 
5562 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5563 	if (!skb)
5564 		goto errout;
5565 
5566 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5567 	if (err < 0) {
5568 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5569 		WARN_ON(err == -EMSGSIZE);
5570 		kfree_skb(skb);
5571 		goto errout;
5572 	}
5573 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5574 	return;
5575 errout:
5576 	if (err < 0)
5577 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5578 }
5579 
5580 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5581 				__s32 *array, int bytes)
5582 {
5583 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5584 
5585 	memset(array, 0, bytes);
5586 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5587 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5588 	array[DEVCONF_MTU6] = cnf->mtu6;
5589 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5590 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5591 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5592 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5593 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5594 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5595 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5596 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5597 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5598 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5599 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5600 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5601 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5602 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5603 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5604 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5605 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5606 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5607 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5608 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5609 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5610 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5611 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5612 	array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5613 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5614 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5615 #ifdef CONFIG_IPV6_ROUTER_PREF
5616 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5617 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5618 		jiffies_to_msecs(cnf->rtr_probe_interval);
5619 #ifdef CONFIG_IPV6_ROUTE_INFO
5620 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5621 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5622 #endif
5623 #endif
5624 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5625 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5626 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5627 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5628 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5629 #endif
5630 #ifdef CONFIG_IPV6_MROUTE
5631 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5632 #endif
5633 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5634 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5635 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5636 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5637 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5638 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5639 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5640 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5641 	/* we omit DEVCONF_STABLE_SECRET for now */
5642 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5643 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5644 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5645 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5646 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5647 #ifdef CONFIG_IPV6_SEG6_HMAC
5648 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5649 #endif
5650 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5651 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5652 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5653 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5654 	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5655 	array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5656 	array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5657 	array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5658 	array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5659 	array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
5660 	array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5661 }
5662 
5663 static inline size_t inet6_ifla6_size(void)
5664 {
5665 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5666 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5667 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5668 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5669 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5670 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5671 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5672 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5673 	     + 0;
5674 }
5675 
5676 static inline size_t inet6_if_nlmsg_size(void)
5677 {
5678 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5679 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5680 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5681 	       + nla_total_size(4) /* IFLA_MTU */
5682 	       + nla_total_size(4) /* IFLA_LINK */
5683 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5684 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5685 }
5686 
5687 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5688 					int bytes)
5689 {
5690 	int i;
5691 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5692 	BUG_ON(pad < 0);
5693 
5694 	/* Use put_unaligned() because stats may not be aligned for u64. */
5695 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5696 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5697 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5698 
5699 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5700 }
5701 
5702 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5703 					int bytes, size_t syncpoff)
5704 {
5705 	int i, c;
5706 	u64 buff[IPSTATS_MIB_MAX];
5707 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5708 
5709 	BUG_ON(pad < 0);
5710 
5711 	memset(buff, 0, sizeof(buff));
5712 	buff[0] = IPSTATS_MIB_MAX;
5713 
5714 	for_each_possible_cpu(c) {
5715 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5716 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5717 	}
5718 
5719 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5720 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5721 }
5722 
5723 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5724 			     int bytes)
5725 {
5726 	switch (attrtype) {
5727 	case IFLA_INET6_STATS:
5728 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5729 				     offsetof(struct ipstats_mib, syncp));
5730 		break;
5731 	case IFLA_INET6_ICMP6STATS:
5732 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5733 		break;
5734 	}
5735 }
5736 
5737 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5738 				  u32 ext_filter_mask)
5739 {
5740 	struct nlattr *nla;
5741 	struct ifla_cacheinfo ci;
5742 
5743 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5744 		goto nla_put_failure;
5745 	ci.max_reasm_len = IPV6_MAXPLEN;
5746 	ci.tstamp = cstamp_delta(idev->tstamp);
5747 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5748 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5749 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5750 		goto nla_put_failure;
5751 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5752 	if (!nla)
5753 		goto nla_put_failure;
5754 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5755 
5756 	/* XXX - MC not implemented */
5757 
5758 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5759 		return 0;
5760 
5761 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5762 	if (!nla)
5763 		goto nla_put_failure;
5764 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5765 
5766 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5767 	if (!nla)
5768 		goto nla_put_failure;
5769 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5770 
5771 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5772 	if (!nla)
5773 		goto nla_put_failure;
5774 	read_lock_bh(&idev->lock);
5775 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5776 	read_unlock_bh(&idev->lock);
5777 
5778 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5779 		goto nla_put_failure;
5780 
5781 	if (idev->ra_mtu &&
5782 	    nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5783 		goto nla_put_failure;
5784 
5785 	return 0;
5786 
5787 nla_put_failure:
5788 	return -EMSGSIZE;
5789 }
5790 
5791 static size_t inet6_get_link_af_size(const struct net_device *dev,
5792 				     u32 ext_filter_mask)
5793 {
5794 	if (!__in6_dev_get(dev))
5795 		return 0;
5796 
5797 	return inet6_ifla6_size();
5798 }
5799 
5800 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5801 			      u32 ext_filter_mask)
5802 {
5803 	struct inet6_dev *idev = __in6_dev_get(dev);
5804 
5805 	if (!idev)
5806 		return -ENODATA;
5807 
5808 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5809 		return -EMSGSIZE;
5810 
5811 	return 0;
5812 }
5813 
5814 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5815 			     struct netlink_ext_ack *extack)
5816 {
5817 	struct inet6_ifaddr *ifp;
5818 	struct net_device *dev = idev->dev;
5819 	bool clear_token, update_rs = false;
5820 	struct in6_addr ll_addr;
5821 
5822 	ASSERT_RTNL();
5823 
5824 	if (!token)
5825 		return -EINVAL;
5826 
5827 	if (dev->flags & IFF_LOOPBACK) {
5828 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5829 		return -EINVAL;
5830 	}
5831 
5832 	if (dev->flags & IFF_NOARP) {
5833 		NL_SET_ERR_MSG_MOD(extack,
5834 				   "Device does not do neighbour discovery");
5835 		return -EINVAL;
5836 	}
5837 
5838 	if (!ipv6_accept_ra(idev)) {
5839 		NL_SET_ERR_MSG_MOD(extack,
5840 				   "Router advertisement is disabled on device");
5841 		return -EINVAL;
5842 	}
5843 
5844 	if (idev->cnf.rtr_solicits == 0) {
5845 		NL_SET_ERR_MSG(extack,
5846 			       "Router solicitation is disabled on device");
5847 		return -EINVAL;
5848 	}
5849 
5850 	write_lock_bh(&idev->lock);
5851 
5852 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5853 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5854 
5855 	write_unlock_bh(&idev->lock);
5856 
5857 	clear_token = ipv6_addr_any(token);
5858 	if (clear_token)
5859 		goto update_lft;
5860 
5861 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5862 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5863 			     IFA_F_OPTIMISTIC)) {
5864 		/* If we're not ready, then normal ifup will take care
5865 		 * of this. Otherwise, we need to request our rs here.
5866 		 */
5867 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5868 		update_rs = true;
5869 	}
5870 
5871 update_lft:
5872 	write_lock_bh(&idev->lock);
5873 
5874 	if (update_rs) {
5875 		idev->if_flags |= IF_RS_SENT;
5876 		idev->rs_interval = rfc3315_s14_backoff_init(
5877 			idev->cnf.rtr_solicit_interval);
5878 		idev->rs_probes = 1;
5879 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5880 	}
5881 
5882 	/* Well, that's kinda nasty ... */
5883 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5884 		spin_lock(&ifp->lock);
5885 		if (ifp->tokenized) {
5886 			ifp->valid_lft = 0;
5887 			ifp->prefered_lft = 0;
5888 		}
5889 		spin_unlock(&ifp->lock);
5890 	}
5891 
5892 	write_unlock_bh(&idev->lock);
5893 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5894 	addrconf_verify_rtnl(dev_net(dev));
5895 	return 0;
5896 }
5897 
5898 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5899 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5900 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5901 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5902 					    .reject_message =
5903 						"IFLA_INET6_RA_MTU can not be set" },
5904 };
5905 
5906 static int check_addr_gen_mode(int mode)
5907 {
5908 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5909 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5910 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5911 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5912 		return -EINVAL;
5913 	return 1;
5914 }
5915 
5916 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5917 				int mode)
5918 {
5919 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5920 	    !idev->cnf.stable_secret.initialized &&
5921 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5922 		return -EINVAL;
5923 	return 1;
5924 }
5925 
5926 static int inet6_validate_link_af(const struct net_device *dev,
5927 				  const struct nlattr *nla,
5928 				  struct netlink_ext_ack *extack)
5929 {
5930 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5931 	struct inet6_dev *idev = NULL;
5932 	int err;
5933 
5934 	if (dev) {
5935 		idev = __in6_dev_get(dev);
5936 		if (!idev)
5937 			return -EAFNOSUPPORT;
5938 	}
5939 
5940 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5941 					  inet6_af_policy, extack);
5942 	if (err)
5943 		return err;
5944 
5945 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5946 		return -EINVAL;
5947 
5948 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5949 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5950 
5951 		if (check_addr_gen_mode(mode) < 0)
5952 			return -EINVAL;
5953 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5954 			return -EINVAL;
5955 	}
5956 
5957 	return 0;
5958 }
5959 
5960 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5961 			     struct netlink_ext_ack *extack)
5962 {
5963 	struct inet6_dev *idev = __in6_dev_get(dev);
5964 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5965 	int err;
5966 
5967 	if (!idev)
5968 		return -EAFNOSUPPORT;
5969 
5970 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5971 		return -EINVAL;
5972 
5973 	if (tb[IFLA_INET6_TOKEN]) {
5974 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5975 					extack);
5976 		if (err)
5977 			return err;
5978 	}
5979 
5980 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5981 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5982 
5983 		idev->cnf.addr_gen_mode = mode;
5984 	}
5985 
5986 	return 0;
5987 }
5988 
5989 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5990 			     u32 portid, u32 seq, int event, unsigned int flags)
5991 {
5992 	struct net_device *dev = idev->dev;
5993 	struct ifinfomsg *hdr;
5994 	struct nlmsghdr *nlh;
5995 	void *protoinfo;
5996 
5997 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5998 	if (!nlh)
5999 		return -EMSGSIZE;
6000 
6001 	hdr = nlmsg_data(nlh);
6002 	hdr->ifi_family = AF_INET6;
6003 	hdr->__ifi_pad = 0;
6004 	hdr->ifi_type = dev->type;
6005 	hdr->ifi_index = dev->ifindex;
6006 	hdr->ifi_flags = dev_get_flags(dev);
6007 	hdr->ifi_change = 0;
6008 
6009 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6010 	    (dev->addr_len &&
6011 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6012 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
6013 	    (dev->ifindex != dev_get_iflink(dev) &&
6014 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
6015 	    nla_put_u8(skb, IFLA_OPERSTATE,
6016 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
6017 		goto nla_put_failure;
6018 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6019 	if (!protoinfo)
6020 		goto nla_put_failure;
6021 
6022 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6023 		goto nla_put_failure;
6024 
6025 	nla_nest_end(skb, protoinfo);
6026 	nlmsg_end(skb, nlh);
6027 	return 0;
6028 
6029 nla_put_failure:
6030 	nlmsg_cancel(skb, nlh);
6031 	return -EMSGSIZE;
6032 }
6033 
6034 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6035 				   struct netlink_ext_ack *extack)
6036 {
6037 	struct ifinfomsg *ifm;
6038 
6039 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6040 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6041 		return -EINVAL;
6042 	}
6043 
6044 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6045 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6046 		return -EINVAL;
6047 	}
6048 
6049 	ifm = nlmsg_data(nlh);
6050 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6051 	    ifm->ifi_change || ifm->ifi_index) {
6052 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6053 		return -EINVAL;
6054 	}
6055 
6056 	return 0;
6057 }
6058 
6059 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6060 {
6061 	struct net *net = sock_net(skb->sk);
6062 	int h, s_h;
6063 	int idx = 0, s_idx;
6064 	struct net_device *dev;
6065 	struct inet6_dev *idev;
6066 	struct hlist_head *head;
6067 
6068 	/* only requests using strict checking can pass data to
6069 	 * influence the dump
6070 	 */
6071 	if (cb->strict_check) {
6072 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6073 
6074 		if (err < 0)
6075 			return err;
6076 	}
6077 
6078 	s_h = cb->args[0];
6079 	s_idx = cb->args[1];
6080 
6081 	rcu_read_lock();
6082 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6083 		idx = 0;
6084 		head = &net->dev_index_head[h];
6085 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
6086 			if (idx < s_idx)
6087 				goto cont;
6088 			idev = __in6_dev_get(dev);
6089 			if (!idev)
6090 				goto cont;
6091 			if (inet6_fill_ifinfo(skb, idev,
6092 					      NETLINK_CB(cb->skb).portid,
6093 					      cb->nlh->nlmsg_seq,
6094 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
6095 				goto out;
6096 cont:
6097 			idx++;
6098 		}
6099 	}
6100 out:
6101 	rcu_read_unlock();
6102 	cb->args[1] = idx;
6103 	cb->args[0] = h;
6104 
6105 	return skb->len;
6106 }
6107 
6108 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6109 {
6110 	struct sk_buff *skb;
6111 	struct net *net = dev_net(idev->dev);
6112 	int err = -ENOBUFS;
6113 
6114 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6115 	if (!skb)
6116 		goto errout;
6117 
6118 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6119 	if (err < 0) {
6120 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6121 		WARN_ON(err == -EMSGSIZE);
6122 		kfree_skb(skb);
6123 		goto errout;
6124 	}
6125 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6126 	return;
6127 errout:
6128 	if (err < 0)
6129 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6130 }
6131 
6132 static inline size_t inet6_prefix_nlmsg_size(void)
6133 {
6134 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6135 	       + nla_total_size(sizeof(struct in6_addr))
6136 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6137 }
6138 
6139 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6140 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6141 			     int event, unsigned int flags)
6142 {
6143 	struct prefixmsg *pmsg;
6144 	struct nlmsghdr *nlh;
6145 	struct prefix_cacheinfo	ci;
6146 
6147 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6148 	if (!nlh)
6149 		return -EMSGSIZE;
6150 
6151 	pmsg = nlmsg_data(nlh);
6152 	pmsg->prefix_family = AF_INET6;
6153 	pmsg->prefix_pad1 = 0;
6154 	pmsg->prefix_pad2 = 0;
6155 	pmsg->prefix_ifindex = idev->dev->ifindex;
6156 	pmsg->prefix_len = pinfo->prefix_len;
6157 	pmsg->prefix_type = pinfo->type;
6158 	pmsg->prefix_pad3 = 0;
6159 	pmsg->prefix_flags = pinfo->flags;
6160 
6161 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6162 		goto nla_put_failure;
6163 	ci.preferred_time = ntohl(pinfo->prefered);
6164 	ci.valid_time = ntohl(pinfo->valid);
6165 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6166 		goto nla_put_failure;
6167 	nlmsg_end(skb, nlh);
6168 	return 0;
6169 
6170 nla_put_failure:
6171 	nlmsg_cancel(skb, nlh);
6172 	return -EMSGSIZE;
6173 }
6174 
6175 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6176 			 struct prefix_info *pinfo)
6177 {
6178 	struct sk_buff *skb;
6179 	struct net *net = dev_net(idev->dev);
6180 	int err = -ENOBUFS;
6181 
6182 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6183 	if (!skb)
6184 		goto errout;
6185 
6186 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6187 	if (err < 0) {
6188 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6189 		WARN_ON(err == -EMSGSIZE);
6190 		kfree_skb(skb);
6191 		goto errout;
6192 	}
6193 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6194 	return;
6195 errout:
6196 	if (err < 0)
6197 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6198 }
6199 
6200 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6201 {
6202 	struct net *net = dev_net(ifp->idev->dev);
6203 
6204 	if (event)
6205 		ASSERT_RTNL();
6206 
6207 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6208 
6209 	switch (event) {
6210 	case RTM_NEWADDR:
6211 		/*
6212 		 * If the address was optimistic we inserted the route at the
6213 		 * start of our DAD process, so we don't need to do it again.
6214 		 * If the device was taken down in the middle of the DAD
6215 		 * cycle there is a race where we could get here without a
6216 		 * host route, so nothing to insert. That will be fixed when
6217 		 * the device is brought up.
6218 		 */
6219 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6220 			ip6_ins_rt(net, ifp->rt);
6221 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6222 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6223 				&ifp->addr, ifp->idev->dev->name);
6224 		}
6225 
6226 		if (ifp->idev->cnf.forwarding)
6227 			addrconf_join_anycast(ifp);
6228 		if (!ipv6_addr_any(&ifp->peer_addr))
6229 			addrconf_prefix_route(&ifp->peer_addr, 128,
6230 					      ifp->rt_priority, ifp->idev->dev,
6231 					      0, 0, GFP_ATOMIC);
6232 		break;
6233 	case RTM_DELADDR:
6234 		if (ifp->idev->cnf.forwarding)
6235 			addrconf_leave_anycast(ifp);
6236 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6237 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6238 			struct fib6_info *rt;
6239 
6240 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6241 						       ifp->idev->dev, 0, 0,
6242 						       false);
6243 			if (rt)
6244 				ip6_del_rt(net, rt, false);
6245 		}
6246 		if (ifp->rt) {
6247 			ip6_del_rt(net, ifp->rt, false);
6248 			ifp->rt = NULL;
6249 		}
6250 		rt_genid_bump_ipv6(net);
6251 		break;
6252 	}
6253 	atomic_inc(&net->ipv6.dev_addr_genid);
6254 }
6255 
6256 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6257 {
6258 	if (likely(ifp->idev->dead == 0))
6259 		__ipv6_ifa_notify(event, ifp);
6260 }
6261 
6262 #ifdef CONFIG_SYSCTL
6263 
6264 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6265 		void *buffer, size_t *lenp, loff_t *ppos)
6266 {
6267 	int *valp = ctl->data;
6268 	int val = *valp;
6269 	loff_t pos = *ppos;
6270 	struct ctl_table lctl;
6271 	int ret;
6272 
6273 	/*
6274 	 * ctl->data points to idev->cnf.forwarding, we should
6275 	 * not modify it until we get the rtnl lock.
6276 	 */
6277 	lctl = *ctl;
6278 	lctl.data = &val;
6279 
6280 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6281 
6282 	if (write)
6283 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6284 	if (ret)
6285 		*ppos = pos;
6286 	return ret;
6287 }
6288 
6289 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6290 		void *buffer, size_t *lenp, loff_t *ppos)
6291 {
6292 	struct inet6_dev *idev = ctl->extra1;
6293 	int min_mtu = IPV6_MIN_MTU;
6294 	struct ctl_table lctl;
6295 
6296 	lctl = *ctl;
6297 	lctl.extra1 = &min_mtu;
6298 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6299 
6300 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6301 }
6302 
6303 static void dev_disable_change(struct inet6_dev *idev)
6304 {
6305 	struct netdev_notifier_info info;
6306 
6307 	if (!idev || !idev->dev)
6308 		return;
6309 
6310 	netdev_notifier_info_init(&info, idev->dev);
6311 	if (idev->cnf.disable_ipv6)
6312 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6313 	else
6314 		addrconf_notify(NULL, NETDEV_UP, &info);
6315 }
6316 
6317 static void addrconf_disable_change(struct net *net, __s32 newf)
6318 {
6319 	struct net_device *dev;
6320 	struct inet6_dev *idev;
6321 
6322 	for_each_netdev(net, dev) {
6323 		idev = __in6_dev_get(dev);
6324 		if (idev) {
6325 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6326 			idev->cnf.disable_ipv6 = newf;
6327 			if (changed)
6328 				dev_disable_change(idev);
6329 		}
6330 	}
6331 }
6332 
6333 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6334 {
6335 	struct net *net;
6336 	int old;
6337 
6338 	if (!rtnl_trylock())
6339 		return restart_syscall();
6340 
6341 	net = (struct net *)table->extra2;
6342 	old = *p;
6343 	*p = newf;
6344 
6345 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6346 		rtnl_unlock();
6347 		return 0;
6348 	}
6349 
6350 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6351 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
6352 		addrconf_disable_change(net, newf);
6353 	} else if ((!newf) ^ (!old))
6354 		dev_disable_change((struct inet6_dev *)table->extra1);
6355 
6356 	rtnl_unlock();
6357 	return 0;
6358 }
6359 
6360 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6361 		void *buffer, size_t *lenp, loff_t *ppos)
6362 {
6363 	int *valp = ctl->data;
6364 	int val = *valp;
6365 	loff_t pos = *ppos;
6366 	struct ctl_table lctl;
6367 	int ret;
6368 
6369 	/*
6370 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6371 	 * not modify it until we get the rtnl lock.
6372 	 */
6373 	lctl = *ctl;
6374 	lctl.data = &val;
6375 
6376 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6377 
6378 	if (write)
6379 		ret = addrconf_disable_ipv6(ctl, valp, val);
6380 	if (ret)
6381 		*ppos = pos;
6382 	return ret;
6383 }
6384 
6385 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6386 		void *buffer, size_t *lenp, loff_t *ppos)
6387 {
6388 	int *valp = ctl->data;
6389 	int ret;
6390 	int old, new;
6391 
6392 	old = *valp;
6393 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6394 	new = *valp;
6395 
6396 	if (write && old != new) {
6397 		struct net *net = ctl->extra2;
6398 
6399 		if (!rtnl_trylock())
6400 			return restart_syscall();
6401 
6402 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6403 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6404 						     NETCONFA_PROXY_NEIGH,
6405 						     NETCONFA_IFINDEX_DEFAULT,
6406 						     net->ipv6.devconf_dflt);
6407 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6408 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6409 						     NETCONFA_PROXY_NEIGH,
6410 						     NETCONFA_IFINDEX_ALL,
6411 						     net->ipv6.devconf_all);
6412 		else {
6413 			struct inet6_dev *idev = ctl->extra1;
6414 
6415 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6416 						     NETCONFA_PROXY_NEIGH,
6417 						     idev->dev->ifindex,
6418 						     &idev->cnf);
6419 		}
6420 		rtnl_unlock();
6421 	}
6422 
6423 	return ret;
6424 }
6425 
6426 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6427 					 void *buffer, size_t *lenp,
6428 					 loff_t *ppos)
6429 {
6430 	int ret = 0;
6431 	u32 new_val;
6432 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6433 	struct net *net = (struct net *)ctl->extra2;
6434 	struct ctl_table tmp = {
6435 		.data = &new_val,
6436 		.maxlen = sizeof(new_val),
6437 		.mode = ctl->mode,
6438 	};
6439 
6440 	if (!rtnl_trylock())
6441 		return restart_syscall();
6442 
6443 	new_val = *((u32 *)ctl->data);
6444 
6445 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6446 	if (ret != 0)
6447 		goto out;
6448 
6449 	if (write) {
6450 		if (check_addr_gen_mode(new_val) < 0) {
6451 			ret = -EINVAL;
6452 			goto out;
6453 		}
6454 
6455 		if (idev) {
6456 			if (check_stable_privacy(idev, net, new_val) < 0) {
6457 				ret = -EINVAL;
6458 				goto out;
6459 			}
6460 
6461 			if (idev->cnf.addr_gen_mode != new_val) {
6462 				idev->cnf.addr_gen_mode = new_val;
6463 				addrconf_init_auto_addrs(idev->dev);
6464 			}
6465 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6466 			struct net_device *dev;
6467 
6468 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6469 			for_each_netdev(net, dev) {
6470 				idev = __in6_dev_get(dev);
6471 				if (idev &&
6472 				    idev->cnf.addr_gen_mode != new_val) {
6473 					idev->cnf.addr_gen_mode = new_val;
6474 					addrconf_init_auto_addrs(idev->dev);
6475 				}
6476 			}
6477 		}
6478 
6479 		*((u32 *)ctl->data) = new_val;
6480 	}
6481 
6482 out:
6483 	rtnl_unlock();
6484 
6485 	return ret;
6486 }
6487 
6488 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6489 					 void *buffer, size_t *lenp,
6490 					 loff_t *ppos)
6491 {
6492 	int err;
6493 	struct in6_addr addr;
6494 	char str[IPV6_MAX_STRLEN];
6495 	struct ctl_table lctl = *ctl;
6496 	struct net *net = ctl->extra2;
6497 	struct ipv6_stable_secret *secret = ctl->data;
6498 
6499 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6500 		return -EIO;
6501 
6502 	lctl.maxlen = IPV6_MAX_STRLEN;
6503 	lctl.data = str;
6504 
6505 	if (!rtnl_trylock())
6506 		return restart_syscall();
6507 
6508 	if (!write && !secret->initialized) {
6509 		err = -EIO;
6510 		goto out;
6511 	}
6512 
6513 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6514 	if (err >= sizeof(str)) {
6515 		err = -EIO;
6516 		goto out;
6517 	}
6518 
6519 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6520 	if (err || !write)
6521 		goto out;
6522 
6523 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6524 		err = -EIO;
6525 		goto out;
6526 	}
6527 
6528 	secret->initialized = true;
6529 	secret->secret = addr;
6530 
6531 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6532 		struct net_device *dev;
6533 
6534 		for_each_netdev(net, dev) {
6535 			struct inet6_dev *idev = __in6_dev_get(dev);
6536 
6537 			if (idev) {
6538 				idev->cnf.addr_gen_mode =
6539 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6540 			}
6541 		}
6542 	} else {
6543 		struct inet6_dev *idev = ctl->extra1;
6544 
6545 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6546 	}
6547 
6548 out:
6549 	rtnl_unlock();
6550 
6551 	return err;
6552 }
6553 
6554 static
6555 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6556 						int write, void *buffer,
6557 						size_t *lenp,
6558 						loff_t *ppos)
6559 {
6560 	int *valp = ctl->data;
6561 	int val = *valp;
6562 	loff_t pos = *ppos;
6563 	struct ctl_table lctl;
6564 	int ret;
6565 
6566 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6567 	 * we should not modify it until we get the rtnl lock.
6568 	 */
6569 	lctl = *ctl;
6570 	lctl.data = &val;
6571 
6572 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6573 
6574 	if (write)
6575 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6576 	if (ret)
6577 		*ppos = pos;
6578 	return ret;
6579 }
6580 
6581 static
6582 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6583 {
6584 	if (rt) {
6585 		if (action)
6586 			rt->dst.flags |= DST_NOPOLICY;
6587 		else
6588 			rt->dst.flags &= ~DST_NOPOLICY;
6589 	}
6590 }
6591 
6592 static
6593 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6594 {
6595 	struct inet6_ifaddr *ifa;
6596 
6597 	read_lock_bh(&idev->lock);
6598 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6599 		spin_lock(&ifa->lock);
6600 		if (ifa->rt) {
6601 			/* host routes only use builtin fib6_nh */
6602 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6603 			int cpu;
6604 
6605 			rcu_read_lock();
6606 			ifa->rt->dst_nopolicy = val ? true : false;
6607 			if (nh->rt6i_pcpu) {
6608 				for_each_possible_cpu(cpu) {
6609 					struct rt6_info **rtp;
6610 
6611 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6612 					addrconf_set_nopolicy(*rtp, val);
6613 				}
6614 			}
6615 			rcu_read_unlock();
6616 		}
6617 		spin_unlock(&ifa->lock);
6618 	}
6619 	read_unlock_bh(&idev->lock);
6620 }
6621 
6622 static
6623 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6624 {
6625 	struct inet6_dev *idev;
6626 	struct net *net;
6627 
6628 	if (!rtnl_trylock())
6629 		return restart_syscall();
6630 
6631 	*valp = val;
6632 
6633 	net = (struct net *)ctl->extra2;
6634 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6635 		rtnl_unlock();
6636 		return 0;
6637 	}
6638 
6639 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6640 		struct net_device *dev;
6641 
6642 		for_each_netdev(net, dev) {
6643 			idev = __in6_dev_get(dev);
6644 			if (idev)
6645 				addrconf_disable_policy_idev(idev, val);
6646 		}
6647 	} else {
6648 		idev = (struct inet6_dev *)ctl->extra1;
6649 		addrconf_disable_policy_idev(idev, val);
6650 	}
6651 
6652 	rtnl_unlock();
6653 	return 0;
6654 }
6655 
6656 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6657 				   void *buffer, size_t *lenp, loff_t *ppos)
6658 {
6659 	int *valp = ctl->data;
6660 	int val = *valp;
6661 	loff_t pos = *ppos;
6662 	struct ctl_table lctl;
6663 	int ret;
6664 
6665 	lctl = *ctl;
6666 	lctl.data = &val;
6667 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6668 
6669 	if (write && (*valp != val))
6670 		ret = addrconf_disable_policy(ctl, valp, val);
6671 
6672 	if (ret)
6673 		*ppos = pos;
6674 
6675 	return ret;
6676 }
6677 
6678 static int minus_one = -1;
6679 static const int two_five_five = 255;
6680 static u32 ioam6_if_id_max = U16_MAX;
6681 
6682 static const struct ctl_table addrconf_sysctl[] = {
6683 	{
6684 		.procname	= "forwarding",
6685 		.data		= &ipv6_devconf.forwarding,
6686 		.maxlen		= sizeof(int),
6687 		.mode		= 0644,
6688 		.proc_handler	= addrconf_sysctl_forward,
6689 	},
6690 	{
6691 		.procname	= "hop_limit",
6692 		.data		= &ipv6_devconf.hop_limit,
6693 		.maxlen		= sizeof(int),
6694 		.mode		= 0644,
6695 		.proc_handler	= proc_dointvec_minmax,
6696 		.extra1		= (void *)SYSCTL_ONE,
6697 		.extra2		= (void *)&two_five_five,
6698 	},
6699 	{
6700 		.procname	= "mtu",
6701 		.data		= &ipv6_devconf.mtu6,
6702 		.maxlen		= sizeof(int),
6703 		.mode		= 0644,
6704 		.proc_handler	= addrconf_sysctl_mtu,
6705 	},
6706 	{
6707 		.procname	= "accept_ra",
6708 		.data		= &ipv6_devconf.accept_ra,
6709 		.maxlen		= sizeof(int),
6710 		.mode		= 0644,
6711 		.proc_handler	= proc_dointvec,
6712 	},
6713 	{
6714 		.procname	= "accept_redirects",
6715 		.data		= &ipv6_devconf.accept_redirects,
6716 		.maxlen		= sizeof(int),
6717 		.mode		= 0644,
6718 		.proc_handler	= proc_dointvec,
6719 	},
6720 	{
6721 		.procname	= "autoconf",
6722 		.data		= &ipv6_devconf.autoconf,
6723 		.maxlen		= sizeof(int),
6724 		.mode		= 0644,
6725 		.proc_handler	= proc_dointvec,
6726 	},
6727 	{
6728 		.procname	= "dad_transmits",
6729 		.data		= &ipv6_devconf.dad_transmits,
6730 		.maxlen		= sizeof(int),
6731 		.mode		= 0644,
6732 		.proc_handler	= proc_dointvec,
6733 	},
6734 	{
6735 		.procname	= "router_solicitations",
6736 		.data		= &ipv6_devconf.rtr_solicits,
6737 		.maxlen		= sizeof(int),
6738 		.mode		= 0644,
6739 		.proc_handler	= proc_dointvec_minmax,
6740 		.extra1		= &minus_one,
6741 	},
6742 	{
6743 		.procname	= "router_solicitation_interval",
6744 		.data		= &ipv6_devconf.rtr_solicit_interval,
6745 		.maxlen		= sizeof(int),
6746 		.mode		= 0644,
6747 		.proc_handler	= proc_dointvec_jiffies,
6748 	},
6749 	{
6750 		.procname	= "router_solicitation_max_interval",
6751 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6752 		.maxlen		= sizeof(int),
6753 		.mode		= 0644,
6754 		.proc_handler	= proc_dointvec_jiffies,
6755 	},
6756 	{
6757 		.procname	= "router_solicitation_delay",
6758 		.data		= &ipv6_devconf.rtr_solicit_delay,
6759 		.maxlen		= sizeof(int),
6760 		.mode		= 0644,
6761 		.proc_handler	= proc_dointvec_jiffies,
6762 	},
6763 	{
6764 		.procname	= "force_mld_version",
6765 		.data		= &ipv6_devconf.force_mld_version,
6766 		.maxlen		= sizeof(int),
6767 		.mode		= 0644,
6768 		.proc_handler	= proc_dointvec,
6769 	},
6770 	{
6771 		.procname	= "mldv1_unsolicited_report_interval",
6772 		.data		=
6773 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6774 		.maxlen		= sizeof(int),
6775 		.mode		= 0644,
6776 		.proc_handler	= proc_dointvec_ms_jiffies,
6777 	},
6778 	{
6779 		.procname	= "mldv2_unsolicited_report_interval",
6780 		.data		=
6781 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6782 		.maxlen		= sizeof(int),
6783 		.mode		= 0644,
6784 		.proc_handler	= proc_dointvec_ms_jiffies,
6785 	},
6786 	{
6787 		.procname	= "use_tempaddr",
6788 		.data		= &ipv6_devconf.use_tempaddr,
6789 		.maxlen		= sizeof(int),
6790 		.mode		= 0644,
6791 		.proc_handler	= proc_dointvec,
6792 	},
6793 	{
6794 		.procname	= "temp_valid_lft",
6795 		.data		= &ipv6_devconf.temp_valid_lft,
6796 		.maxlen		= sizeof(int),
6797 		.mode		= 0644,
6798 		.proc_handler	= proc_dointvec,
6799 	},
6800 	{
6801 		.procname	= "temp_prefered_lft",
6802 		.data		= &ipv6_devconf.temp_prefered_lft,
6803 		.maxlen		= sizeof(int),
6804 		.mode		= 0644,
6805 		.proc_handler	= proc_dointvec,
6806 	},
6807 	{
6808 		.procname	= "regen_max_retry",
6809 		.data		= &ipv6_devconf.regen_max_retry,
6810 		.maxlen		= sizeof(int),
6811 		.mode		= 0644,
6812 		.proc_handler	= proc_dointvec,
6813 	},
6814 	{
6815 		.procname	= "max_desync_factor",
6816 		.data		= &ipv6_devconf.max_desync_factor,
6817 		.maxlen		= sizeof(int),
6818 		.mode		= 0644,
6819 		.proc_handler	= proc_dointvec,
6820 	},
6821 	{
6822 		.procname	= "max_addresses",
6823 		.data		= &ipv6_devconf.max_addresses,
6824 		.maxlen		= sizeof(int),
6825 		.mode		= 0644,
6826 		.proc_handler	= proc_dointvec,
6827 	},
6828 	{
6829 		.procname	= "accept_ra_defrtr",
6830 		.data		= &ipv6_devconf.accept_ra_defrtr,
6831 		.maxlen		= sizeof(int),
6832 		.mode		= 0644,
6833 		.proc_handler	= proc_dointvec,
6834 	},
6835 	{
6836 		.procname	= "ra_defrtr_metric",
6837 		.data		= &ipv6_devconf.ra_defrtr_metric,
6838 		.maxlen		= sizeof(u32),
6839 		.mode		= 0644,
6840 		.proc_handler	= proc_douintvec_minmax,
6841 		.extra1		= (void *)SYSCTL_ONE,
6842 	},
6843 	{
6844 		.procname	= "accept_ra_min_hop_limit",
6845 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6846 		.maxlen		= sizeof(int),
6847 		.mode		= 0644,
6848 		.proc_handler	= proc_dointvec,
6849 	},
6850 	{
6851 		.procname	= "accept_ra_min_lft",
6852 		.data		= &ipv6_devconf.accept_ra_min_lft,
6853 		.maxlen		= sizeof(int),
6854 		.mode		= 0644,
6855 		.proc_handler	= proc_dointvec,
6856 	},
6857 	{
6858 		.procname	= "accept_ra_pinfo",
6859 		.data		= &ipv6_devconf.accept_ra_pinfo,
6860 		.maxlen		= sizeof(int),
6861 		.mode		= 0644,
6862 		.proc_handler	= proc_dointvec,
6863 	},
6864 	{
6865 		.procname	= "ra_honor_pio_life",
6866 		.data		= &ipv6_devconf.ra_honor_pio_life,
6867 		.maxlen		= sizeof(u8),
6868 		.mode		= 0644,
6869 		.proc_handler	= proc_dou8vec_minmax,
6870 		.extra1		= SYSCTL_ZERO,
6871 		.extra2		= SYSCTL_ONE,
6872 	},
6873 #ifdef CONFIG_IPV6_ROUTER_PREF
6874 	{
6875 		.procname	= "accept_ra_rtr_pref",
6876 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6877 		.maxlen		= sizeof(int),
6878 		.mode		= 0644,
6879 		.proc_handler	= proc_dointvec,
6880 	},
6881 	{
6882 		.procname	= "router_probe_interval",
6883 		.data		= &ipv6_devconf.rtr_probe_interval,
6884 		.maxlen		= sizeof(int),
6885 		.mode		= 0644,
6886 		.proc_handler	= proc_dointvec_jiffies,
6887 	},
6888 #ifdef CONFIG_IPV6_ROUTE_INFO
6889 	{
6890 		.procname	= "accept_ra_rt_info_min_plen",
6891 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6892 		.maxlen		= sizeof(int),
6893 		.mode		= 0644,
6894 		.proc_handler	= proc_dointvec,
6895 	},
6896 	{
6897 		.procname	= "accept_ra_rt_info_max_plen",
6898 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6899 		.maxlen		= sizeof(int),
6900 		.mode		= 0644,
6901 		.proc_handler	= proc_dointvec,
6902 	},
6903 #endif
6904 #endif
6905 	{
6906 		.procname	= "proxy_ndp",
6907 		.data		= &ipv6_devconf.proxy_ndp,
6908 		.maxlen		= sizeof(int),
6909 		.mode		= 0644,
6910 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6911 	},
6912 	{
6913 		.procname	= "accept_source_route",
6914 		.data		= &ipv6_devconf.accept_source_route,
6915 		.maxlen		= sizeof(int),
6916 		.mode		= 0644,
6917 		.proc_handler	= proc_dointvec,
6918 	},
6919 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6920 	{
6921 		.procname	= "optimistic_dad",
6922 		.data		= &ipv6_devconf.optimistic_dad,
6923 		.maxlen		= sizeof(int),
6924 		.mode		= 0644,
6925 		.proc_handler   = proc_dointvec,
6926 	},
6927 	{
6928 		.procname	= "use_optimistic",
6929 		.data		= &ipv6_devconf.use_optimistic,
6930 		.maxlen		= sizeof(int),
6931 		.mode		= 0644,
6932 		.proc_handler	= proc_dointvec,
6933 	},
6934 #endif
6935 #ifdef CONFIG_IPV6_MROUTE
6936 	{
6937 		.procname	= "mc_forwarding",
6938 		.data		= &ipv6_devconf.mc_forwarding,
6939 		.maxlen		= sizeof(int),
6940 		.mode		= 0444,
6941 		.proc_handler	= proc_dointvec,
6942 	},
6943 #endif
6944 	{
6945 		.procname	= "disable_ipv6",
6946 		.data		= &ipv6_devconf.disable_ipv6,
6947 		.maxlen		= sizeof(int),
6948 		.mode		= 0644,
6949 		.proc_handler	= addrconf_sysctl_disable,
6950 	},
6951 	{
6952 		.procname	= "accept_dad",
6953 		.data		= &ipv6_devconf.accept_dad,
6954 		.maxlen		= sizeof(int),
6955 		.mode		= 0644,
6956 		.proc_handler	= proc_dointvec,
6957 	},
6958 	{
6959 		.procname	= "force_tllao",
6960 		.data		= &ipv6_devconf.force_tllao,
6961 		.maxlen		= sizeof(int),
6962 		.mode		= 0644,
6963 		.proc_handler	= proc_dointvec
6964 	},
6965 	{
6966 		.procname	= "ndisc_notify",
6967 		.data		= &ipv6_devconf.ndisc_notify,
6968 		.maxlen		= sizeof(int),
6969 		.mode		= 0644,
6970 		.proc_handler	= proc_dointvec
6971 	},
6972 	{
6973 		.procname	= "suppress_frag_ndisc",
6974 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6975 		.maxlen		= sizeof(int),
6976 		.mode		= 0644,
6977 		.proc_handler	= proc_dointvec
6978 	},
6979 	{
6980 		.procname	= "accept_ra_from_local",
6981 		.data		= &ipv6_devconf.accept_ra_from_local,
6982 		.maxlen		= sizeof(int),
6983 		.mode		= 0644,
6984 		.proc_handler	= proc_dointvec,
6985 	},
6986 	{
6987 		.procname	= "accept_ra_mtu",
6988 		.data		= &ipv6_devconf.accept_ra_mtu,
6989 		.maxlen		= sizeof(int),
6990 		.mode		= 0644,
6991 		.proc_handler	= proc_dointvec,
6992 	},
6993 	{
6994 		.procname	= "stable_secret",
6995 		.data		= &ipv6_devconf.stable_secret,
6996 		.maxlen		= IPV6_MAX_STRLEN,
6997 		.mode		= 0600,
6998 		.proc_handler	= addrconf_sysctl_stable_secret,
6999 	},
7000 	{
7001 		.procname	= "use_oif_addrs_only",
7002 		.data		= &ipv6_devconf.use_oif_addrs_only,
7003 		.maxlen		= sizeof(int),
7004 		.mode		= 0644,
7005 		.proc_handler	= proc_dointvec,
7006 	},
7007 	{
7008 		.procname	= "ignore_routes_with_linkdown",
7009 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
7010 		.maxlen		= sizeof(int),
7011 		.mode		= 0644,
7012 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
7013 	},
7014 	{
7015 		.procname	= "drop_unicast_in_l2_multicast",
7016 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
7017 		.maxlen		= sizeof(int),
7018 		.mode		= 0644,
7019 		.proc_handler	= proc_dointvec,
7020 	},
7021 	{
7022 		.procname	= "drop_unsolicited_na",
7023 		.data		= &ipv6_devconf.drop_unsolicited_na,
7024 		.maxlen		= sizeof(int),
7025 		.mode		= 0644,
7026 		.proc_handler	= proc_dointvec,
7027 	},
7028 	{
7029 		.procname	= "keep_addr_on_down",
7030 		.data		= &ipv6_devconf.keep_addr_on_down,
7031 		.maxlen		= sizeof(int),
7032 		.mode		= 0644,
7033 		.proc_handler	= proc_dointvec,
7034 
7035 	},
7036 	{
7037 		.procname	= "seg6_enabled",
7038 		.data		= &ipv6_devconf.seg6_enabled,
7039 		.maxlen		= sizeof(int),
7040 		.mode		= 0644,
7041 		.proc_handler	= proc_dointvec,
7042 	},
7043 #ifdef CONFIG_IPV6_SEG6_HMAC
7044 	{
7045 		.procname	= "seg6_require_hmac",
7046 		.data		= &ipv6_devconf.seg6_require_hmac,
7047 		.maxlen		= sizeof(int),
7048 		.mode		= 0644,
7049 		.proc_handler	= proc_dointvec,
7050 	},
7051 #endif
7052 	{
7053 		.procname       = "enhanced_dad",
7054 		.data           = &ipv6_devconf.enhanced_dad,
7055 		.maxlen         = sizeof(int),
7056 		.mode           = 0644,
7057 		.proc_handler   = proc_dointvec,
7058 	},
7059 	{
7060 		.procname	= "addr_gen_mode",
7061 		.data		= &ipv6_devconf.addr_gen_mode,
7062 		.maxlen		= sizeof(int),
7063 		.mode		= 0644,
7064 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
7065 	},
7066 	{
7067 		.procname       = "disable_policy",
7068 		.data           = &ipv6_devconf.disable_policy,
7069 		.maxlen         = sizeof(int),
7070 		.mode           = 0644,
7071 		.proc_handler   = addrconf_sysctl_disable_policy,
7072 	},
7073 	{
7074 		.procname	= "ndisc_tclass",
7075 		.data		= &ipv6_devconf.ndisc_tclass,
7076 		.maxlen		= sizeof(int),
7077 		.mode		= 0644,
7078 		.proc_handler	= proc_dointvec_minmax,
7079 		.extra1		= (void *)SYSCTL_ZERO,
7080 		.extra2		= (void *)&two_five_five,
7081 	},
7082 	{
7083 		.procname	= "rpl_seg_enabled",
7084 		.data		= &ipv6_devconf.rpl_seg_enabled,
7085 		.maxlen		= sizeof(int),
7086 		.mode		= 0644,
7087 		.proc_handler	= proc_dointvec,
7088 	},
7089 	{
7090 		.procname	= "ioam6_enabled",
7091 		.data		= &ipv6_devconf.ioam6_enabled,
7092 		.maxlen		= sizeof(u8),
7093 		.mode		= 0644,
7094 		.proc_handler	= proc_dou8vec_minmax,
7095 		.extra1		= (void *)SYSCTL_ZERO,
7096 		.extra2		= (void *)SYSCTL_ONE,
7097 	},
7098 	{
7099 		.procname	= "ioam6_id",
7100 		.data		= &ipv6_devconf.ioam6_id,
7101 		.maxlen		= sizeof(u32),
7102 		.mode		= 0644,
7103 		.proc_handler	= proc_douintvec_minmax,
7104 		.extra1		= (void *)SYSCTL_ZERO,
7105 		.extra2		= (void *)&ioam6_if_id_max,
7106 	},
7107 	{
7108 		.procname	= "ioam6_id_wide",
7109 		.data		= &ipv6_devconf.ioam6_id_wide,
7110 		.maxlen		= sizeof(u32),
7111 		.mode		= 0644,
7112 		.proc_handler	= proc_douintvec,
7113 	},
7114 	{
7115 		.procname	= "ndisc_evict_nocarrier",
7116 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
7117 		.maxlen		= sizeof(u8),
7118 		.mode		= 0644,
7119 		.proc_handler	= proc_dou8vec_minmax,
7120 		.extra1		= (void *)SYSCTL_ZERO,
7121 		.extra2		= (void *)SYSCTL_ONE,
7122 	},
7123 	{
7124 		.procname	= "accept_untracked_na",
7125 		.data		= &ipv6_devconf.accept_untracked_na,
7126 		.maxlen		= sizeof(int),
7127 		.mode		= 0644,
7128 		.proc_handler	= proc_dointvec_minmax,
7129 		.extra1		= SYSCTL_ZERO,
7130 		.extra2		= SYSCTL_TWO,
7131 	},
7132 	{
7133 		/* sentinel */
7134 	}
7135 };
7136 
7137 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7138 		struct inet6_dev *idev, struct ipv6_devconf *p)
7139 {
7140 	int i, ifindex;
7141 	struct ctl_table *table;
7142 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7143 
7144 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7145 	if (!table)
7146 		goto out;
7147 
7148 	for (i = 0; table[i].data; i++) {
7149 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7150 		/* If one of these is already set, then it is not safe to
7151 		 * overwrite either of them: this makes proc_dointvec_minmax
7152 		 * usable.
7153 		 */
7154 		if (!table[i].extra1 && !table[i].extra2) {
7155 			table[i].extra1 = idev; /* embedded; no ref */
7156 			table[i].extra2 = net;
7157 		}
7158 	}
7159 
7160 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7161 
7162 	p->sysctl_header = register_net_sysctl_sz(net, path, table,
7163 						  ARRAY_SIZE(addrconf_sysctl));
7164 	if (!p->sysctl_header)
7165 		goto free;
7166 
7167 	if (!strcmp(dev_name, "all"))
7168 		ifindex = NETCONFA_IFINDEX_ALL;
7169 	else if (!strcmp(dev_name, "default"))
7170 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7171 	else
7172 		ifindex = idev->dev->ifindex;
7173 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7174 				     ifindex, p);
7175 	return 0;
7176 
7177 free:
7178 	kfree(table);
7179 out:
7180 	return -ENOBUFS;
7181 }
7182 
7183 static void __addrconf_sysctl_unregister(struct net *net,
7184 					 struct ipv6_devconf *p, int ifindex)
7185 {
7186 	struct ctl_table *table;
7187 
7188 	if (!p->sysctl_header)
7189 		return;
7190 
7191 	table = p->sysctl_header->ctl_table_arg;
7192 	unregister_net_sysctl_table(p->sysctl_header);
7193 	p->sysctl_header = NULL;
7194 	kfree(table);
7195 
7196 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7197 }
7198 
7199 static int addrconf_sysctl_register(struct inet6_dev *idev)
7200 {
7201 	int err;
7202 
7203 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7204 		return -EINVAL;
7205 
7206 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7207 				    &ndisc_ifinfo_sysctl_change);
7208 	if (err)
7209 		return err;
7210 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7211 					 idev, &idev->cnf);
7212 	if (err)
7213 		neigh_sysctl_unregister(idev->nd_parms);
7214 
7215 	return err;
7216 }
7217 
7218 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7219 {
7220 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7221 				     idev->dev->ifindex);
7222 	neigh_sysctl_unregister(idev->nd_parms);
7223 }
7224 
7225 
7226 #endif
7227 
7228 static int __net_init addrconf_init_net(struct net *net)
7229 {
7230 	int err = -ENOMEM;
7231 	struct ipv6_devconf *all, *dflt;
7232 
7233 	spin_lock_init(&net->ipv6.addrconf_hash_lock);
7234 	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7235 	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7236 					   sizeof(struct hlist_head),
7237 					   GFP_KERNEL);
7238 	if (!net->ipv6.inet6_addr_lst)
7239 		goto err_alloc_addr;
7240 
7241 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7242 	if (!all)
7243 		goto err_alloc_all;
7244 
7245 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7246 	if (!dflt)
7247 		goto err_alloc_dflt;
7248 
7249 	if (!net_eq(net, &init_net)) {
7250 		switch (net_inherit_devconf()) {
7251 		case 1:  /* copy from init_net */
7252 			memcpy(all, init_net.ipv6.devconf_all,
7253 			       sizeof(ipv6_devconf));
7254 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7255 			       sizeof(ipv6_devconf_dflt));
7256 			break;
7257 		case 3: /* copy from the current netns */
7258 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7259 			       sizeof(ipv6_devconf));
7260 			memcpy(dflt,
7261 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7262 			       sizeof(ipv6_devconf_dflt));
7263 			break;
7264 		case 0:
7265 		case 2:
7266 			/* use compiled values */
7267 			break;
7268 		}
7269 	}
7270 
7271 	/* these will be inherited by all namespaces */
7272 	dflt->autoconf = ipv6_defaults.autoconf;
7273 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7274 
7275 	dflt->stable_secret.initialized = false;
7276 	all->stable_secret.initialized = false;
7277 
7278 	net->ipv6.devconf_all = all;
7279 	net->ipv6.devconf_dflt = dflt;
7280 
7281 #ifdef CONFIG_SYSCTL
7282 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7283 	if (err < 0)
7284 		goto err_reg_all;
7285 
7286 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7287 	if (err < 0)
7288 		goto err_reg_dflt;
7289 #endif
7290 	return 0;
7291 
7292 #ifdef CONFIG_SYSCTL
7293 err_reg_dflt:
7294 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7295 err_reg_all:
7296 	kfree(dflt);
7297 	net->ipv6.devconf_dflt = NULL;
7298 #endif
7299 err_alloc_dflt:
7300 	kfree(all);
7301 	net->ipv6.devconf_all = NULL;
7302 err_alloc_all:
7303 	kfree(net->ipv6.inet6_addr_lst);
7304 err_alloc_addr:
7305 	return err;
7306 }
7307 
7308 static void __net_exit addrconf_exit_net(struct net *net)
7309 {
7310 	int i;
7311 
7312 #ifdef CONFIG_SYSCTL
7313 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7314 				     NETCONFA_IFINDEX_DEFAULT);
7315 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7316 				     NETCONFA_IFINDEX_ALL);
7317 #endif
7318 	kfree(net->ipv6.devconf_dflt);
7319 	net->ipv6.devconf_dflt = NULL;
7320 	kfree(net->ipv6.devconf_all);
7321 	net->ipv6.devconf_all = NULL;
7322 
7323 	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7324 	/*
7325 	 *	Check hash table, then free it.
7326 	 */
7327 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7328 		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7329 
7330 	kfree(net->ipv6.inet6_addr_lst);
7331 	net->ipv6.inet6_addr_lst = NULL;
7332 }
7333 
7334 static struct pernet_operations addrconf_ops = {
7335 	.init = addrconf_init_net,
7336 	.exit = addrconf_exit_net,
7337 };
7338 
7339 static struct rtnl_af_ops inet6_ops __read_mostly = {
7340 	.family		  = AF_INET6,
7341 	.fill_link_af	  = inet6_fill_link_af,
7342 	.get_link_af_size = inet6_get_link_af_size,
7343 	.validate_link_af = inet6_validate_link_af,
7344 	.set_link_af	  = inet6_set_link_af,
7345 };
7346 
7347 /*
7348  *	Init / cleanup code
7349  */
7350 
7351 int __init addrconf_init(void)
7352 {
7353 	struct inet6_dev *idev;
7354 	int err;
7355 
7356 	err = ipv6_addr_label_init();
7357 	if (err < 0) {
7358 		pr_crit("%s: cannot initialize default policy table: %d\n",
7359 			__func__, err);
7360 		goto out;
7361 	}
7362 
7363 	err = register_pernet_subsys(&addrconf_ops);
7364 	if (err < 0)
7365 		goto out_addrlabel;
7366 
7367 	addrconf_wq = create_workqueue("ipv6_addrconf");
7368 	if (!addrconf_wq) {
7369 		err = -ENOMEM;
7370 		goto out_nowq;
7371 	}
7372 
7373 	rtnl_lock();
7374 	idev = ipv6_add_dev(blackhole_netdev);
7375 	rtnl_unlock();
7376 	if (IS_ERR(idev)) {
7377 		err = PTR_ERR(idev);
7378 		goto errlo;
7379 	}
7380 
7381 	ip6_route_init_special_entries();
7382 
7383 	register_netdevice_notifier(&ipv6_dev_notf);
7384 
7385 	addrconf_verify(&init_net);
7386 
7387 	rtnl_af_register(&inet6_ops);
7388 
7389 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7390 				   NULL, inet6_dump_ifinfo, 0);
7391 	if (err < 0)
7392 		goto errout;
7393 
7394 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7395 				   inet6_rtm_newaddr, NULL, 0);
7396 	if (err < 0)
7397 		goto errout;
7398 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7399 				   inet6_rtm_deladdr, NULL, 0);
7400 	if (err < 0)
7401 		goto errout;
7402 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7403 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7404 				   RTNL_FLAG_DOIT_UNLOCKED);
7405 	if (err < 0)
7406 		goto errout;
7407 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7408 				   NULL, inet6_dump_ifmcaddr, 0);
7409 	if (err < 0)
7410 		goto errout;
7411 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7412 				   NULL, inet6_dump_ifacaddr, 0);
7413 	if (err < 0)
7414 		goto errout;
7415 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7416 				   inet6_netconf_get_devconf,
7417 				   inet6_netconf_dump_devconf,
7418 				   RTNL_FLAG_DOIT_UNLOCKED);
7419 	if (err < 0)
7420 		goto errout;
7421 	err = ipv6_addr_label_rtnl_register();
7422 	if (err < 0)
7423 		goto errout;
7424 
7425 	return 0;
7426 errout:
7427 	rtnl_unregister_all(PF_INET6);
7428 	rtnl_af_unregister(&inet6_ops);
7429 	unregister_netdevice_notifier(&ipv6_dev_notf);
7430 errlo:
7431 	destroy_workqueue(addrconf_wq);
7432 out_nowq:
7433 	unregister_pernet_subsys(&addrconf_ops);
7434 out_addrlabel:
7435 	ipv6_addr_label_cleanup();
7436 out:
7437 	return err;
7438 }
7439 
7440 void addrconf_cleanup(void)
7441 {
7442 	struct net_device *dev;
7443 
7444 	unregister_netdevice_notifier(&ipv6_dev_notf);
7445 	unregister_pernet_subsys(&addrconf_ops);
7446 	ipv6_addr_label_cleanup();
7447 
7448 	rtnl_af_unregister(&inet6_ops);
7449 
7450 	rtnl_lock();
7451 
7452 	/* clean dev list */
7453 	for_each_netdev(&init_net, dev) {
7454 		if (__in6_dev_get(dev) == NULL)
7455 			continue;
7456 		addrconf_ifdown(dev, true);
7457 	}
7458 	addrconf_ifdown(init_net.loopback_dev, true);
7459 
7460 	rtnl_unlock();
7461 
7462 	destroy_workqueue(addrconf_wq);
7463 }
7464