xref: /openbmc/linux/net/ipv6/addrconf.c (revision 4b33b5ff)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 Address [auto]configuration
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  */
10 
11 /*
12  *	Changes:
13  *
14  *	Janos Farkas			:	delete timer on ifdown
15  *	<chexum@bankinf.banki.hu>
16  *	Andi Kleen			:	kill double kfree on module
17  *						unload.
18  *	Maciej W. Rozycki		:	FDDI support
19  *	sekiya@USAGI			:	Don't send too many RS
20  *						packets.
21  *	yoshfuji@USAGI			:       Fixed interval between DAD
22  *						packets.
23  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
24  *						address validation timer.
25  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
26  *						support.
27  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
28  *						address on a same interface.
29  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
30  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
31  *						seq_file.
32  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
33  *						selection; consider scope,
34  *						status etc.
35  */
36 
37 #define pr_fmt(fmt) "IPv6: " fmt
38 
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65 
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
69 
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
88 
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
92 #include <linux/ioam6.h>
93 
94 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
95 
96 #define IPV6_MAX_STRLEN \
97 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98 
99 static inline u32 cstamp_delta(unsigned long cstamp)
100 {
101 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102 }
103 
104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 {
106 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 	u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
108 	do_div(tmp, 1000000);
109 	return (s32)tmp;
110 }
111 
112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 {
114 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 	u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
116 	do_div(tmp, 1000000);
117 	if ((s32)tmp > mrt) {
118 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 		tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
120 		do_div(tmp, 1000000);
121 	}
122 	return (s32)tmp;
123 }
124 
125 #ifdef CONFIG_SYSCTL
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 #else
129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 {
131 	return 0;
132 }
133 
134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135 {
136 }
137 #endif
138 
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140 
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 					u8 dad_count,
145 					const struct inet6_dev *idev);
146 
147 #define IN6_ADDR_HSIZE_SHIFT	8
148 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
149 
150 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
152 
153 static struct workqueue_struct *addrconf_wq;
154 
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157 
158 static void addrconf_type_change(struct net_device *dev,
159 				 unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
161 
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163 						  int plen,
164 						  const struct net_device *dev,
165 						  u32 flags, u32 noflags,
166 						  bool no_gw);
167 
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171 				   bool send_na);
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176 
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 				struct prefix_info *pinfo);
179 
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
181 	.forwarding		= 0,
182 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
183 	.mtu6			= IPV6_MIN_MTU,
184 	.accept_ra		= 1,
185 	.accept_redirects	= 1,
186 	.autoconf		= 1,
187 	.force_mld_version	= 0,
188 	.mldv1_unsolicited_report_interval = 10 * HZ,
189 	.mldv2_unsolicited_report_interval = HZ,
190 	.dad_transmits		= 1,
191 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
192 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
193 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
195 	.use_tempaddr		= 0,
196 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
197 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
198 	.regen_max_retry	= REGEN_MAX_RETRY,
199 	.max_desync_factor	= MAX_DESYNC_FACTOR,
200 	.max_addresses		= IPV6_MAX_ADDRESSES,
201 	.accept_ra_defrtr	= 1,
202 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
203 	.accept_ra_from_local	= 0,
204 	.accept_ra_min_hop_limit= 1,
205 	.accept_ra_pinfo	= 1,
206 #ifdef CONFIG_IPV6_ROUTER_PREF
207 	.accept_ra_rtr_pref	= 1,
208 	.rtr_probe_interval	= 60 * HZ,
209 #ifdef CONFIG_IPV6_ROUTE_INFO
210 	.accept_ra_rt_info_min_plen = 0,
211 	.accept_ra_rt_info_max_plen = 0,
212 #endif
213 #endif
214 	.proxy_ndp		= 0,
215 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
216 	.disable_ipv6		= 0,
217 	.accept_dad		= 0,
218 	.suppress_frag_ndisc	= 1,
219 	.accept_ra_mtu		= 1,
220 	.stable_secret		= {
221 		.initialized = false,
222 	},
223 	.use_oif_addrs_only	= 0,
224 	.ignore_routes_with_linkdown = 0,
225 	.keep_addr_on_down	= 0,
226 	.seg6_enabled		= 0,
227 #ifdef CONFIG_IPV6_SEG6_HMAC
228 	.seg6_require_hmac	= 0,
229 #endif
230 	.enhanced_dad           = 1,
231 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
232 	.disable_policy		= 0,
233 	.rpl_seg_enabled	= 0,
234 	.ioam6_enabled		= 0,
235 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
236 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
237 	.ndisc_evict_nocarrier	= 1,
238 };
239 
240 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
241 	.forwarding		= 0,
242 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
243 	.mtu6			= IPV6_MIN_MTU,
244 	.accept_ra		= 1,
245 	.accept_redirects	= 1,
246 	.autoconf		= 1,
247 	.force_mld_version	= 0,
248 	.mldv1_unsolicited_report_interval = 10 * HZ,
249 	.mldv2_unsolicited_report_interval = HZ,
250 	.dad_transmits		= 1,
251 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
252 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
253 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
254 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
255 	.use_tempaddr		= 0,
256 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
257 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
258 	.regen_max_retry	= REGEN_MAX_RETRY,
259 	.max_desync_factor	= MAX_DESYNC_FACTOR,
260 	.max_addresses		= IPV6_MAX_ADDRESSES,
261 	.accept_ra_defrtr	= 1,
262 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
263 	.accept_ra_from_local	= 0,
264 	.accept_ra_min_hop_limit= 1,
265 	.accept_ra_pinfo	= 1,
266 #ifdef CONFIG_IPV6_ROUTER_PREF
267 	.accept_ra_rtr_pref	= 1,
268 	.rtr_probe_interval	= 60 * HZ,
269 #ifdef CONFIG_IPV6_ROUTE_INFO
270 	.accept_ra_rt_info_min_plen = 0,
271 	.accept_ra_rt_info_max_plen = 0,
272 #endif
273 #endif
274 	.proxy_ndp		= 0,
275 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
276 	.disable_ipv6		= 0,
277 	.accept_dad		= 1,
278 	.suppress_frag_ndisc	= 1,
279 	.accept_ra_mtu		= 1,
280 	.stable_secret		= {
281 		.initialized = false,
282 	},
283 	.use_oif_addrs_only	= 0,
284 	.ignore_routes_with_linkdown = 0,
285 	.keep_addr_on_down	= 0,
286 	.seg6_enabled		= 0,
287 #ifdef CONFIG_IPV6_SEG6_HMAC
288 	.seg6_require_hmac	= 0,
289 #endif
290 	.enhanced_dad           = 1,
291 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
292 	.disable_policy		= 0,
293 	.rpl_seg_enabled	= 0,
294 	.ioam6_enabled		= 0,
295 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
296 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
297 	.ndisc_evict_nocarrier	= 1,
298 };
299 
300 /* Check if link is ready: is it up and is a valid qdisc available */
301 static inline bool addrconf_link_ready(const struct net_device *dev)
302 {
303 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
304 }
305 
306 static void addrconf_del_rs_timer(struct inet6_dev *idev)
307 {
308 	if (del_timer(&idev->rs_timer))
309 		__in6_dev_put(idev);
310 }
311 
312 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
313 {
314 	if (cancel_delayed_work(&ifp->dad_work))
315 		__in6_ifa_put(ifp);
316 }
317 
318 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
319 				  unsigned long when)
320 {
321 	if (!timer_pending(&idev->rs_timer))
322 		in6_dev_hold(idev);
323 	mod_timer(&idev->rs_timer, jiffies + when);
324 }
325 
326 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
327 				   unsigned long delay)
328 {
329 	in6_ifa_hold(ifp);
330 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
331 		in6_ifa_put(ifp);
332 }
333 
334 static int snmp6_alloc_dev(struct inet6_dev *idev)
335 {
336 	int i;
337 
338 	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
339 	if (!idev->stats.ipv6)
340 		goto err_ip;
341 
342 	for_each_possible_cpu(i) {
343 		struct ipstats_mib *addrconf_stats;
344 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
345 		u64_stats_init(&addrconf_stats->syncp);
346 	}
347 
348 
349 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
350 					GFP_KERNEL);
351 	if (!idev->stats.icmpv6dev)
352 		goto err_icmp;
353 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
354 					   GFP_KERNEL);
355 	if (!idev->stats.icmpv6msgdev)
356 		goto err_icmpmsg;
357 
358 	return 0;
359 
360 err_icmpmsg:
361 	kfree(idev->stats.icmpv6dev);
362 err_icmp:
363 	free_percpu(idev->stats.ipv6);
364 err_ip:
365 	return -ENOMEM;
366 }
367 
368 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
369 {
370 	struct inet6_dev *ndev;
371 	int err = -ENOMEM;
372 
373 	ASSERT_RTNL();
374 
375 	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
376 		return ERR_PTR(-EINVAL);
377 
378 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
379 	if (!ndev)
380 		return ERR_PTR(err);
381 
382 	rwlock_init(&ndev->lock);
383 	ndev->dev = dev;
384 	INIT_LIST_HEAD(&ndev->addr_list);
385 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
386 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
387 
388 	if (ndev->cnf.stable_secret.initialized)
389 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
390 
391 	ndev->cnf.mtu6 = dev->mtu;
392 	ndev->ra_mtu = 0;
393 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
394 	if (!ndev->nd_parms) {
395 		kfree(ndev);
396 		return ERR_PTR(err);
397 	}
398 	if (ndev->cnf.forwarding)
399 		dev_disable_lro(dev);
400 	/* We refer to the device */
401 	dev_hold_track(dev, &ndev->dev_tracker, GFP_KERNEL);
402 
403 	if (snmp6_alloc_dev(ndev) < 0) {
404 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
405 			   __func__);
406 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
407 		dev_put_track(dev, &ndev->dev_tracker);
408 		kfree(ndev);
409 		return ERR_PTR(err);
410 	}
411 
412 	if (dev != blackhole_netdev) {
413 		if (snmp6_register_dev(ndev) < 0) {
414 			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
415 				   __func__, dev->name);
416 			goto err_release;
417 		}
418 	}
419 	/* One reference from device. */
420 	refcount_set(&ndev->refcnt, 1);
421 
422 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
423 		ndev->cnf.accept_dad = -1;
424 
425 #if IS_ENABLED(CONFIG_IPV6_SIT)
426 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
427 		pr_info("%s: Disabled Multicast RS\n", dev->name);
428 		ndev->cnf.rtr_solicits = 0;
429 	}
430 #endif
431 
432 	INIT_LIST_HEAD(&ndev->tempaddr_list);
433 	ndev->desync_factor = U32_MAX;
434 	if ((dev->flags&IFF_LOOPBACK) ||
435 	    dev->type == ARPHRD_TUNNEL ||
436 	    dev->type == ARPHRD_TUNNEL6 ||
437 	    dev->type == ARPHRD_SIT ||
438 	    dev->type == ARPHRD_NONE) {
439 		ndev->cnf.use_tempaddr = -1;
440 	}
441 
442 	ndev->token = in6addr_any;
443 
444 	if (netif_running(dev) && addrconf_link_ready(dev))
445 		ndev->if_flags |= IF_READY;
446 
447 	ipv6_mc_init_dev(ndev);
448 	ndev->tstamp = jiffies;
449 	if (dev != blackhole_netdev) {
450 		err = addrconf_sysctl_register(ndev);
451 		if (err) {
452 			ipv6_mc_destroy_dev(ndev);
453 			snmp6_unregister_dev(ndev);
454 			goto err_release;
455 		}
456 	}
457 	/* protected by rtnl_lock */
458 	rcu_assign_pointer(dev->ip6_ptr, ndev);
459 
460 	if (dev != blackhole_netdev) {
461 		/* Join interface-local all-node multicast group */
462 		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
463 
464 		/* Join all-node multicast group */
465 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
466 
467 		/* Join all-router multicast group if forwarding is set */
468 		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
469 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
470 	}
471 	return ndev;
472 
473 err_release:
474 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
475 	ndev->dead = 1;
476 	in6_dev_finish_destroy(ndev);
477 	return ERR_PTR(err);
478 }
479 
480 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
481 {
482 	struct inet6_dev *idev;
483 
484 	ASSERT_RTNL();
485 
486 	idev = __in6_dev_get(dev);
487 	if (!idev) {
488 		idev = ipv6_add_dev(dev);
489 		if (IS_ERR(idev))
490 			return idev;
491 	}
492 
493 	if (dev->flags&IFF_UP)
494 		ipv6_mc_up(idev);
495 	return idev;
496 }
497 
498 static int inet6_netconf_msgsize_devconf(int type)
499 {
500 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
501 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
502 	bool all = false;
503 
504 	if (type == NETCONFA_ALL)
505 		all = true;
506 
507 	if (all || type == NETCONFA_FORWARDING)
508 		size += nla_total_size(4);
509 #ifdef CONFIG_IPV6_MROUTE
510 	if (all || type == NETCONFA_MC_FORWARDING)
511 		size += nla_total_size(4);
512 #endif
513 	if (all || type == NETCONFA_PROXY_NEIGH)
514 		size += nla_total_size(4);
515 
516 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
517 		size += nla_total_size(4);
518 
519 	return size;
520 }
521 
522 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
523 				      struct ipv6_devconf *devconf, u32 portid,
524 				      u32 seq, int event, unsigned int flags,
525 				      int type)
526 {
527 	struct nlmsghdr  *nlh;
528 	struct netconfmsg *ncm;
529 	bool all = false;
530 
531 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
532 			flags);
533 	if (!nlh)
534 		return -EMSGSIZE;
535 
536 	if (type == NETCONFA_ALL)
537 		all = true;
538 
539 	ncm = nlmsg_data(nlh);
540 	ncm->ncm_family = AF_INET6;
541 
542 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
543 		goto nla_put_failure;
544 
545 	if (!devconf)
546 		goto out;
547 
548 	if ((all || type == NETCONFA_FORWARDING) &&
549 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
550 		goto nla_put_failure;
551 #ifdef CONFIG_IPV6_MROUTE
552 	if ((all || type == NETCONFA_MC_FORWARDING) &&
553 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
554 			atomic_read(&devconf->mc_forwarding)) < 0)
555 		goto nla_put_failure;
556 #endif
557 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
558 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
559 		goto nla_put_failure;
560 
561 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
562 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
563 			devconf->ignore_routes_with_linkdown) < 0)
564 		goto nla_put_failure;
565 
566 out:
567 	nlmsg_end(skb, nlh);
568 	return 0;
569 
570 nla_put_failure:
571 	nlmsg_cancel(skb, nlh);
572 	return -EMSGSIZE;
573 }
574 
575 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
576 				  int ifindex, struct ipv6_devconf *devconf)
577 {
578 	struct sk_buff *skb;
579 	int err = -ENOBUFS;
580 
581 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
582 	if (!skb)
583 		goto errout;
584 
585 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
586 					 event, 0, type);
587 	if (err < 0) {
588 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
589 		WARN_ON(err == -EMSGSIZE);
590 		kfree_skb(skb);
591 		goto errout;
592 	}
593 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
594 	return;
595 errout:
596 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
597 }
598 
599 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
600 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
601 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
602 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
603 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
604 };
605 
606 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
607 				       const struct nlmsghdr *nlh,
608 				       struct nlattr **tb,
609 				       struct netlink_ext_ack *extack)
610 {
611 	int i, err;
612 
613 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
614 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
615 		return -EINVAL;
616 	}
617 
618 	if (!netlink_strict_get_check(skb))
619 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
620 					      tb, NETCONFA_MAX,
621 					      devconf_ipv6_policy, extack);
622 
623 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
624 					    tb, NETCONFA_MAX,
625 					    devconf_ipv6_policy, extack);
626 	if (err)
627 		return err;
628 
629 	for (i = 0; i <= NETCONFA_MAX; i++) {
630 		if (!tb[i])
631 			continue;
632 
633 		switch (i) {
634 		case NETCONFA_IFINDEX:
635 			break;
636 		default:
637 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
638 			return -EINVAL;
639 		}
640 	}
641 
642 	return 0;
643 }
644 
645 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
646 				     struct nlmsghdr *nlh,
647 				     struct netlink_ext_ack *extack)
648 {
649 	struct net *net = sock_net(in_skb->sk);
650 	struct nlattr *tb[NETCONFA_MAX+1];
651 	struct inet6_dev *in6_dev = NULL;
652 	struct net_device *dev = NULL;
653 	struct sk_buff *skb;
654 	struct ipv6_devconf *devconf;
655 	int ifindex;
656 	int err;
657 
658 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
659 	if (err < 0)
660 		return err;
661 
662 	if (!tb[NETCONFA_IFINDEX])
663 		return -EINVAL;
664 
665 	err = -EINVAL;
666 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
667 	switch (ifindex) {
668 	case NETCONFA_IFINDEX_ALL:
669 		devconf = net->ipv6.devconf_all;
670 		break;
671 	case NETCONFA_IFINDEX_DEFAULT:
672 		devconf = net->ipv6.devconf_dflt;
673 		break;
674 	default:
675 		dev = dev_get_by_index(net, ifindex);
676 		if (!dev)
677 			return -EINVAL;
678 		in6_dev = in6_dev_get(dev);
679 		if (!in6_dev)
680 			goto errout;
681 		devconf = &in6_dev->cnf;
682 		break;
683 	}
684 
685 	err = -ENOBUFS;
686 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
687 	if (!skb)
688 		goto errout;
689 
690 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
691 					 NETLINK_CB(in_skb).portid,
692 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
693 					 NETCONFA_ALL);
694 	if (err < 0) {
695 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
696 		WARN_ON(err == -EMSGSIZE);
697 		kfree_skb(skb);
698 		goto errout;
699 	}
700 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
701 errout:
702 	if (in6_dev)
703 		in6_dev_put(in6_dev);
704 	dev_put(dev);
705 	return err;
706 }
707 
708 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
709 				      struct netlink_callback *cb)
710 {
711 	const struct nlmsghdr *nlh = cb->nlh;
712 	struct net *net = sock_net(skb->sk);
713 	int h, s_h;
714 	int idx, s_idx;
715 	struct net_device *dev;
716 	struct inet6_dev *idev;
717 	struct hlist_head *head;
718 
719 	if (cb->strict_check) {
720 		struct netlink_ext_ack *extack = cb->extack;
721 		struct netconfmsg *ncm;
722 
723 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
724 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
725 			return -EINVAL;
726 		}
727 
728 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
729 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
730 			return -EINVAL;
731 		}
732 	}
733 
734 	s_h = cb->args[0];
735 	s_idx = idx = cb->args[1];
736 
737 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
738 		idx = 0;
739 		head = &net->dev_index_head[h];
740 		rcu_read_lock();
741 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
742 			  net->dev_base_seq;
743 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
744 			if (idx < s_idx)
745 				goto cont;
746 			idev = __in6_dev_get(dev);
747 			if (!idev)
748 				goto cont;
749 
750 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
751 						       &idev->cnf,
752 						       NETLINK_CB(cb->skb).portid,
753 						       nlh->nlmsg_seq,
754 						       RTM_NEWNETCONF,
755 						       NLM_F_MULTI,
756 						       NETCONFA_ALL) < 0) {
757 				rcu_read_unlock();
758 				goto done;
759 			}
760 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
761 cont:
762 			idx++;
763 		}
764 		rcu_read_unlock();
765 	}
766 	if (h == NETDEV_HASHENTRIES) {
767 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
768 					       net->ipv6.devconf_all,
769 					       NETLINK_CB(cb->skb).portid,
770 					       nlh->nlmsg_seq,
771 					       RTM_NEWNETCONF, NLM_F_MULTI,
772 					       NETCONFA_ALL) < 0)
773 			goto done;
774 		else
775 			h++;
776 	}
777 	if (h == NETDEV_HASHENTRIES + 1) {
778 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
779 					       net->ipv6.devconf_dflt,
780 					       NETLINK_CB(cb->skb).portid,
781 					       nlh->nlmsg_seq,
782 					       RTM_NEWNETCONF, NLM_F_MULTI,
783 					       NETCONFA_ALL) < 0)
784 			goto done;
785 		else
786 			h++;
787 	}
788 done:
789 	cb->args[0] = h;
790 	cb->args[1] = idx;
791 
792 	return skb->len;
793 }
794 
795 #ifdef CONFIG_SYSCTL
796 static void dev_forward_change(struct inet6_dev *idev)
797 {
798 	struct net_device *dev;
799 	struct inet6_ifaddr *ifa;
800 
801 	if (!idev)
802 		return;
803 	dev = idev->dev;
804 	if (idev->cnf.forwarding)
805 		dev_disable_lro(dev);
806 	if (dev->flags & IFF_MULTICAST) {
807 		if (idev->cnf.forwarding) {
808 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
809 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
810 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
811 		} else {
812 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
813 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
814 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
815 		}
816 	}
817 
818 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
819 		if (ifa->flags&IFA_F_TENTATIVE)
820 			continue;
821 		if (idev->cnf.forwarding)
822 			addrconf_join_anycast(ifa);
823 		else
824 			addrconf_leave_anycast(ifa);
825 	}
826 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
827 				     NETCONFA_FORWARDING,
828 				     dev->ifindex, &idev->cnf);
829 }
830 
831 
832 static void addrconf_forward_change(struct net *net, __s32 newf)
833 {
834 	struct net_device *dev;
835 	struct inet6_dev *idev;
836 
837 	for_each_netdev(net, dev) {
838 		idev = __in6_dev_get(dev);
839 		if (idev) {
840 			int changed = (!idev->cnf.forwarding) ^ (!newf);
841 			idev->cnf.forwarding = newf;
842 			if (changed)
843 				dev_forward_change(idev);
844 		}
845 	}
846 }
847 
848 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
849 {
850 	struct net *net;
851 	int old;
852 
853 	if (!rtnl_trylock())
854 		return restart_syscall();
855 
856 	net = (struct net *)table->extra2;
857 	old = *p;
858 	*p = newf;
859 
860 	if (p == &net->ipv6.devconf_dflt->forwarding) {
861 		if ((!newf) ^ (!old))
862 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
863 						     NETCONFA_FORWARDING,
864 						     NETCONFA_IFINDEX_DEFAULT,
865 						     net->ipv6.devconf_dflt);
866 		rtnl_unlock();
867 		return 0;
868 	}
869 
870 	if (p == &net->ipv6.devconf_all->forwarding) {
871 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
872 
873 		net->ipv6.devconf_dflt->forwarding = newf;
874 		if ((!newf) ^ (!old_dflt))
875 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
876 						     NETCONFA_FORWARDING,
877 						     NETCONFA_IFINDEX_DEFAULT,
878 						     net->ipv6.devconf_dflt);
879 
880 		addrconf_forward_change(net, newf);
881 		if ((!newf) ^ (!old))
882 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
883 						     NETCONFA_FORWARDING,
884 						     NETCONFA_IFINDEX_ALL,
885 						     net->ipv6.devconf_all);
886 	} else if ((!newf) ^ (!old))
887 		dev_forward_change((struct inet6_dev *)table->extra1);
888 	rtnl_unlock();
889 
890 	if (newf)
891 		rt6_purge_dflt_routers(net);
892 	return 1;
893 }
894 
895 static void addrconf_linkdown_change(struct net *net, __s32 newf)
896 {
897 	struct net_device *dev;
898 	struct inet6_dev *idev;
899 
900 	for_each_netdev(net, dev) {
901 		idev = __in6_dev_get(dev);
902 		if (idev) {
903 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
904 
905 			idev->cnf.ignore_routes_with_linkdown = newf;
906 			if (changed)
907 				inet6_netconf_notify_devconf(dev_net(dev),
908 							     RTM_NEWNETCONF,
909 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
910 							     dev->ifindex,
911 							     &idev->cnf);
912 		}
913 	}
914 }
915 
916 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
917 {
918 	struct net *net;
919 	int old;
920 
921 	if (!rtnl_trylock())
922 		return restart_syscall();
923 
924 	net = (struct net *)table->extra2;
925 	old = *p;
926 	*p = newf;
927 
928 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
929 		if ((!newf) ^ (!old))
930 			inet6_netconf_notify_devconf(net,
931 						     RTM_NEWNETCONF,
932 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
933 						     NETCONFA_IFINDEX_DEFAULT,
934 						     net->ipv6.devconf_dflt);
935 		rtnl_unlock();
936 		return 0;
937 	}
938 
939 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
940 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
941 		addrconf_linkdown_change(net, newf);
942 		if ((!newf) ^ (!old))
943 			inet6_netconf_notify_devconf(net,
944 						     RTM_NEWNETCONF,
945 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
946 						     NETCONFA_IFINDEX_ALL,
947 						     net->ipv6.devconf_all);
948 	}
949 	rtnl_unlock();
950 
951 	return 1;
952 }
953 
954 #endif
955 
956 /* Nobody refers to this ifaddr, destroy it */
957 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
958 {
959 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
960 
961 #ifdef NET_REFCNT_DEBUG
962 	pr_debug("%s\n", __func__);
963 #endif
964 
965 	in6_dev_put(ifp->idev);
966 
967 	if (cancel_delayed_work(&ifp->dad_work))
968 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
969 			  ifp);
970 
971 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
972 		pr_warn("Freeing alive inet6 address %p\n", ifp);
973 		return;
974 	}
975 
976 	kfree_rcu(ifp, rcu);
977 }
978 
979 static void
980 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
981 {
982 	struct list_head *p;
983 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
984 
985 	/*
986 	 * Each device address list is sorted in order of scope -
987 	 * global before linklocal.
988 	 */
989 	list_for_each(p, &idev->addr_list) {
990 		struct inet6_ifaddr *ifa
991 			= list_entry(p, struct inet6_ifaddr, if_list);
992 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
993 			break;
994 	}
995 
996 	list_add_tail_rcu(&ifp->if_list, p);
997 }
998 
999 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1000 {
1001 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1002 
1003 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1004 }
1005 
1006 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1007 			       struct net_device *dev, unsigned int hash)
1008 {
1009 	struct inet6_ifaddr *ifp;
1010 
1011 	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1012 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1013 			if (!dev || ifp->idev->dev == dev)
1014 				return true;
1015 		}
1016 	}
1017 	return false;
1018 }
1019 
1020 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1021 {
1022 	struct net *net = dev_net(dev);
1023 	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1024 	int err = 0;
1025 
1026 	spin_lock(&net->ipv6.addrconf_hash_lock);
1027 
1028 	/* Ignore adding duplicate addresses on an interface */
1029 	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1030 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1031 		err = -EEXIST;
1032 	} else {
1033 		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1034 	}
1035 
1036 	spin_unlock(&net->ipv6.addrconf_hash_lock);
1037 
1038 	return err;
1039 }
1040 
1041 /* On success it returns ifp with increased reference count */
1042 
1043 static struct inet6_ifaddr *
1044 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1045 	      bool can_block, struct netlink_ext_ack *extack)
1046 {
1047 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1048 	int addr_type = ipv6_addr_type(cfg->pfx);
1049 	struct net *net = dev_net(idev->dev);
1050 	struct inet6_ifaddr *ifa = NULL;
1051 	struct fib6_info *f6i = NULL;
1052 	int err = 0;
1053 
1054 	if (addr_type == IPV6_ADDR_ANY ||
1055 	    (addr_type & IPV6_ADDR_MULTICAST &&
1056 	     !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1057 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1058 	     !netif_is_l3_master(idev->dev) &&
1059 	     addr_type & IPV6_ADDR_LOOPBACK))
1060 		return ERR_PTR(-EADDRNOTAVAIL);
1061 
1062 	if (idev->dead) {
1063 		err = -ENODEV;			/*XXX*/
1064 		goto out;
1065 	}
1066 
1067 	if (idev->cnf.disable_ipv6) {
1068 		err = -EACCES;
1069 		goto out;
1070 	}
1071 
1072 	/* validator notifier needs to be blocking;
1073 	 * do not call in atomic context
1074 	 */
1075 	if (can_block) {
1076 		struct in6_validator_info i6vi = {
1077 			.i6vi_addr = *cfg->pfx,
1078 			.i6vi_dev = idev,
1079 			.extack = extack,
1080 		};
1081 
1082 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1083 		err = notifier_to_errno(err);
1084 		if (err < 0)
1085 			goto out;
1086 	}
1087 
1088 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1089 	if (!ifa) {
1090 		err = -ENOBUFS;
1091 		goto out;
1092 	}
1093 
1094 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1095 	if (IS_ERR(f6i)) {
1096 		err = PTR_ERR(f6i);
1097 		f6i = NULL;
1098 		goto out;
1099 	}
1100 
1101 	if (net->ipv6.devconf_all->disable_policy ||
1102 	    idev->cnf.disable_policy)
1103 		f6i->dst_nopolicy = true;
1104 
1105 	neigh_parms_data_state_setall(idev->nd_parms);
1106 
1107 	ifa->addr = *cfg->pfx;
1108 	if (cfg->peer_pfx)
1109 		ifa->peer_addr = *cfg->peer_pfx;
1110 
1111 	spin_lock_init(&ifa->lock);
1112 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1113 	INIT_HLIST_NODE(&ifa->addr_lst);
1114 	ifa->scope = cfg->scope;
1115 	ifa->prefix_len = cfg->plen;
1116 	ifa->rt_priority = cfg->rt_priority;
1117 	ifa->flags = cfg->ifa_flags;
1118 	ifa->ifa_proto = cfg->ifa_proto;
1119 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1120 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1121 		ifa->flags |= IFA_F_TENTATIVE;
1122 	ifa->valid_lft = cfg->valid_lft;
1123 	ifa->prefered_lft = cfg->preferred_lft;
1124 	ifa->cstamp = ifa->tstamp = jiffies;
1125 	ifa->tokenized = false;
1126 
1127 	ifa->rt = f6i;
1128 
1129 	ifa->idev = idev;
1130 	in6_dev_hold(idev);
1131 
1132 	/* For caller */
1133 	refcount_set(&ifa->refcnt, 1);
1134 
1135 	rcu_read_lock_bh();
1136 
1137 	err = ipv6_add_addr_hash(idev->dev, ifa);
1138 	if (err < 0) {
1139 		rcu_read_unlock_bh();
1140 		goto out;
1141 	}
1142 
1143 	write_lock(&idev->lock);
1144 
1145 	/* Add to inet6_dev unicast addr list. */
1146 	ipv6_link_dev_addr(idev, ifa);
1147 
1148 	if (ifa->flags&IFA_F_TEMPORARY) {
1149 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1150 		in6_ifa_hold(ifa);
1151 	}
1152 
1153 	in6_ifa_hold(ifa);
1154 	write_unlock(&idev->lock);
1155 
1156 	rcu_read_unlock_bh();
1157 
1158 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1159 out:
1160 	if (unlikely(err < 0)) {
1161 		fib6_info_release(f6i);
1162 
1163 		if (ifa) {
1164 			if (ifa->idev)
1165 				in6_dev_put(ifa->idev);
1166 			kfree(ifa);
1167 		}
1168 		ifa = ERR_PTR(err);
1169 	}
1170 
1171 	return ifa;
1172 }
1173 
1174 enum cleanup_prefix_rt_t {
1175 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1176 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1177 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1178 };
1179 
1180 /*
1181  * Check, whether the prefix for ifp would still need a prefix route
1182  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1183  * constants.
1184  *
1185  * 1) we don't purge prefix if address was not permanent.
1186  *    prefix is managed by its own lifetime.
1187  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1188  * 3) if there are no addresses, delete prefix.
1189  * 4) if there are still other permanent address(es),
1190  *    corresponding prefix is still permanent.
1191  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1192  *    don't purge the prefix, assume user space is managing it.
1193  * 6) otherwise, update prefix lifetime to the
1194  *    longest valid lifetime among the corresponding
1195  *    addresses on the device.
1196  *    Note: subsequent RA will update lifetime.
1197  **/
1198 static enum cleanup_prefix_rt_t
1199 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1200 {
1201 	struct inet6_ifaddr *ifa;
1202 	struct inet6_dev *idev = ifp->idev;
1203 	unsigned long lifetime;
1204 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1205 
1206 	*expires = jiffies;
1207 
1208 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1209 		if (ifa == ifp)
1210 			continue;
1211 		if (ifa->prefix_len != ifp->prefix_len ||
1212 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1213 				       ifp->prefix_len))
1214 			continue;
1215 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1216 			return CLEANUP_PREFIX_RT_NOP;
1217 
1218 		action = CLEANUP_PREFIX_RT_EXPIRE;
1219 
1220 		spin_lock(&ifa->lock);
1221 
1222 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1223 		/*
1224 		 * Note: Because this address is
1225 		 * not permanent, lifetime <
1226 		 * LONG_MAX / HZ here.
1227 		 */
1228 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1229 			*expires = ifa->tstamp + lifetime * HZ;
1230 		spin_unlock(&ifa->lock);
1231 	}
1232 
1233 	return action;
1234 }
1235 
1236 static void
1237 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1238 		     bool del_rt, bool del_peer)
1239 {
1240 	struct fib6_info *f6i;
1241 
1242 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1243 					ifp->prefix_len,
1244 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1245 	if (f6i) {
1246 		if (del_rt)
1247 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1248 		else {
1249 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1250 				fib6_set_expires(f6i, expires);
1251 			fib6_info_release(f6i);
1252 		}
1253 	}
1254 }
1255 
1256 
1257 /* This function wants to get referenced ifp and releases it before return */
1258 
1259 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1260 {
1261 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1262 	struct net *net = dev_net(ifp->idev->dev);
1263 	unsigned long expires;
1264 	int state;
1265 
1266 	ASSERT_RTNL();
1267 
1268 	spin_lock_bh(&ifp->lock);
1269 	state = ifp->state;
1270 	ifp->state = INET6_IFADDR_STATE_DEAD;
1271 	spin_unlock_bh(&ifp->lock);
1272 
1273 	if (state == INET6_IFADDR_STATE_DEAD)
1274 		goto out;
1275 
1276 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1277 	hlist_del_init_rcu(&ifp->addr_lst);
1278 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1279 
1280 	write_lock_bh(&ifp->idev->lock);
1281 
1282 	if (ifp->flags&IFA_F_TEMPORARY) {
1283 		list_del(&ifp->tmp_list);
1284 		if (ifp->ifpub) {
1285 			in6_ifa_put(ifp->ifpub);
1286 			ifp->ifpub = NULL;
1287 		}
1288 		__in6_ifa_put(ifp);
1289 	}
1290 
1291 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1292 		action = check_cleanup_prefix_route(ifp, &expires);
1293 
1294 	list_del_rcu(&ifp->if_list);
1295 	__in6_ifa_put(ifp);
1296 
1297 	write_unlock_bh(&ifp->idev->lock);
1298 
1299 	addrconf_del_dad_work(ifp);
1300 
1301 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1302 
1303 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1304 
1305 	if (action != CLEANUP_PREFIX_RT_NOP) {
1306 		cleanup_prefix_route(ifp, expires,
1307 			action == CLEANUP_PREFIX_RT_DEL, false);
1308 	}
1309 
1310 	/* clean up prefsrc entries */
1311 	rt6_remove_prefsrc(ifp);
1312 out:
1313 	in6_ifa_put(ifp);
1314 }
1315 
1316 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1317 {
1318 	struct inet6_dev *idev = ifp->idev;
1319 	unsigned long tmp_tstamp, age;
1320 	unsigned long regen_advance;
1321 	unsigned long now = jiffies;
1322 	s32 cnf_temp_preferred_lft;
1323 	struct inet6_ifaddr *ift;
1324 	struct ifa6_config cfg;
1325 	long max_desync_factor;
1326 	struct in6_addr addr;
1327 	int ret = 0;
1328 
1329 	write_lock_bh(&idev->lock);
1330 
1331 retry:
1332 	in6_dev_hold(idev);
1333 	if (idev->cnf.use_tempaddr <= 0) {
1334 		write_unlock_bh(&idev->lock);
1335 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1336 		in6_dev_put(idev);
1337 		ret = -1;
1338 		goto out;
1339 	}
1340 	spin_lock_bh(&ifp->lock);
1341 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1342 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1343 		spin_unlock_bh(&ifp->lock);
1344 		write_unlock_bh(&idev->lock);
1345 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1346 			__func__);
1347 		in6_dev_put(idev);
1348 		ret = -1;
1349 		goto out;
1350 	}
1351 	in6_ifa_hold(ifp);
1352 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1353 	ipv6_gen_rnd_iid(&addr);
1354 
1355 	age = (now - ifp->tstamp) / HZ;
1356 
1357 	regen_advance = idev->cnf.regen_max_retry *
1358 			idev->cnf.dad_transmits *
1359 			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1360 
1361 	/* recalculate max_desync_factor each time and update
1362 	 * idev->desync_factor if it's larger
1363 	 */
1364 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1365 	max_desync_factor = min_t(__u32,
1366 				  idev->cnf.max_desync_factor,
1367 				  cnf_temp_preferred_lft - regen_advance);
1368 
1369 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1370 		if (max_desync_factor > 0) {
1371 			get_random_bytes(&idev->desync_factor,
1372 					 sizeof(idev->desync_factor));
1373 			idev->desync_factor %= max_desync_factor;
1374 		} else {
1375 			idev->desync_factor = 0;
1376 		}
1377 	}
1378 
1379 	memset(&cfg, 0, sizeof(cfg));
1380 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1381 			      idev->cnf.temp_valid_lft + age);
1382 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1383 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1384 
1385 	cfg.plen = ifp->prefix_len;
1386 	tmp_tstamp = ifp->tstamp;
1387 	spin_unlock_bh(&ifp->lock);
1388 
1389 	write_unlock_bh(&idev->lock);
1390 
1391 	/* A temporary address is created only if this calculated Preferred
1392 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1393 	 * an implementation must not create a temporary address with a zero
1394 	 * Preferred Lifetime.
1395 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1396 	 * temporary addresses being generated.
1397 	 */
1398 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1399 	if (cfg.preferred_lft <= regen_advance + age) {
1400 		in6_ifa_put(ifp);
1401 		in6_dev_put(idev);
1402 		ret = -1;
1403 		goto out;
1404 	}
1405 
1406 	cfg.ifa_flags = IFA_F_TEMPORARY;
1407 	/* set in addrconf_prefix_rcv() */
1408 	if (ifp->flags & IFA_F_OPTIMISTIC)
1409 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1410 
1411 	cfg.pfx = &addr;
1412 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1413 
1414 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1415 	if (IS_ERR(ift)) {
1416 		in6_ifa_put(ifp);
1417 		in6_dev_put(idev);
1418 		pr_info("%s: retry temporary address regeneration\n", __func__);
1419 		write_lock_bh(&idev->lock);
1420 		goto retry;
1421 	}
1422 
1423 	spin_lock_bh(&ift->lock);
1424 	ift->ifpub = ifp;
1425 	ift->cstamp = now;
1426 	ift->tstamp = tmp_tstamp;
1427 	spin_unlock_bh(&ift->lock);
1428 
1429 	addrconf_dad_start(ift);
1430 	in6_ifa_put(ift);
1431 	in6_dev_put(idev);
1432 out:
1433 	return ret;
1434 }
1435 
1436 /*
1437  *	Choose an appropriate source address (RFC3484)
1438  */
1439 enum {
1440 	IPV6_SADDR_RULE_INIT = 0,
1441 	IPV6_SADDR_RULE_LOCAL,
1442 	IPV6_SADDR_RULE_SCOPE,
1443 	IPV6_SADDR_RULE_PREFERRED,
1444 #ifdef CONFIG_IPV6_MIP6
1445 	IPV6_SADDR_RULE_HOA,
1446 #endif
1447 	IPV6_SADDR_RULE_OIF,
1448 	IPV6_SADDR_RULE_LABEL,
1449 	IPV6_SADDR_RULE_PRIVACY,
1450 	IPV6_SADDR_RULE_ORCHID,
1451 	IPV6_SADDR_RULE_PREFIX,
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1454 #endif
1455 	IPV6_SADDR_RULE_MAX
1456 };
1457 
1458 struct ipv6_saddr_score {
1459 	int			rule;
1460 	int			addr_type;
1461 	struct inet6_ifaddr	*ifa;
1462 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1463 	int			scopedist;
1464 	int			matchlen;
1465 };
1466 
1467 struct ipv6_saddr_dst {
1468 	const struct in6_addr *addr;
1469 	int ifindex;
1470 	int scope;
1471 	int label;
1472 	unsigned int prefs;
1473 };
1474 
1475 static inline int ipv6_saddr_preferred(int type)
1476 {
1477 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1478 		return 1;
1479 	return 0;
1480 }
1481 
1482 static bool ipv6_use_optimistic_addr(struct net *net,
1483 				     struct inet6_dev *idev)
1484 {
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486 	if (!idev)
1487 		return false;
1488 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1489 		return false;
1490 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1491 		return false;
1492 
1493 	return true;
1494 #else
1495 	return false;
1496 #endif
1497 }
1498 
1499 static bool ipv6_allow_optimistic_dad(struct net *net,
1500 				      struct inet6_dev *idev)
1501 {
1502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503 	if (!idev)
1504 		return false;
1505 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1506 		return false;
1507 
1508 	return true;
1509 #else
1510 	return false;
1511 #endif
1512 }
1513 
1514 static int ipv6_get_saddr_eval(struct net *net,
1515 			       struct ipv6_saddr_score *score,
1516 			       struct ipv6_saddr_dst *dst,
1517 			       int i)
1518 {
1519 	int ret;
1520 
1521 	if (i <= score->rule) {
1522 		switch (i) {
1523 		case IPV6_SADDR_RULE_SCOPE:
1524 			ret = score->scopedist;
1525 			break;
1526 		case IPV6_SADDR_RULE_PREFIX:
1527 			ret = score->matchlen;
1528 			break;
1529 		default:
1530 			ret = !!test_bit(i, score->scorebits);
1531 		}
1532 		goto out;
1533 	}
1534 
1535 	switch (i) {
1536 	case IPV6_SADDR_RULE_INIT:
1537 		/* Rule 0: remember if hiscore is not ready yet */
1538 		ret = !!score->ifa;
1539 		break;
1540 	case IPV6_SADDR_RULE_LOCAL:
1541 		/* Rule 1: Prefer same address */
1542 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1543 		break;
1544 	case IPV6_SADDR_RULE_SCOPE:
1545 		/* Rule 2: Prefer appropriate scope
1546 		 *
1547 		 *      ret
1548 		 *       ^
1549 		 *    -1 |  d 15
1550 		 *    ---+--+-+---> scope
1551 		 *       |
1552 		 *       |             d is scope of the destination.
1553 		 *  B-d  |  \
1554 		 *       |   \      <- smaller scope is better if
1555 		 *  B-15 |    \        if scope is enough for destination.
1556 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1557 		 * d-C-1 | /
1558 		 *       |/         <- greater is better
1559 		 *   -C  /             if scope is not enough for destination.
1560 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1561 		 *
1562 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1563 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1564 		 * Assume B = 0 and we get C > 29.
1565 		 */
1566 		ret = __ipv6_addr_src_scope(score->addr_type);
1567 		if (ret >= dst->scope)
1568 			ret = -ret;
1569 		else
1570 			ret -= 128;	/* 30 is enough */
1571 		score->scopedist = ret;
1572 		break;
1573 	case IPV6_SADDR_RULE_PREFERRED:
1574 	    {
1575 		/* Rule 3: Avoid deprecated and optimistic addresses */
1576 		u8 avoid = IFA_F_DEPRECATED;
1577 
1578 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1579 			avoid |= IFA_F_OPTIMISTIC;
1580 		ret = ipv6_saddr_preferred(score->addr_type) ||
1581 		      !(score->ifa->flags & avoid);
1582 		break;
1583 	    }
1584 #ifdef CONFIG_IPV6_MIP6
1585 	case IPV6_SADDR_RULE_HOA:
1586 	    {
1587 		/* Rule 4: Prefer home address */
1588 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1589 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1590 		break;
1591 	    }
1592 #endif
1593 	case IPV6_SADDR_RULE_OIF:
1594 		/* Rule 5: Prefer outgoing interface */
1595 		ret = (!dst->ifindex ||
1596 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1597 		break;
1598 	case IPV6_SADDR_RULE_LABEL:
1599 		/* Rule 6: Prefer matching label */
1600 		ret = ipv6_addr_label(net,
1601 				      &score->ifa->addr, score->addr_type,
1602 				      score->ifa->idev->dev->ifindex) == dst->label;
1603 		break;
1604 	case IPV6_SADDR_RULE_PRIVACY:
1605 	    {
1606 		/* Rule 7: Prefer public address
1607 		 * Note: prefer temporary address if use_tempaddr >= 2
1608 		 */
1609 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1610 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1611 				score->ifa->idev->cnf.use_tempaddr >= 2;
1612 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613 		break;
1614 	    }
1615 	case IPV6_SADDR_RULE_ORCHID:
1616 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1617 		 *	    non-ORCHID vs non-ORCHID
1618 		 */
1619 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1620 			ipv6_addr_orchid(dst->addr));
1621 		break;
1622 	case IPV6_SADDR_RULE_PREFIX:
1623 		/* Rule 8: Use longest matching prefix */
1624 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1625 		if (ret > score->ifa->prefix_len)
1626 			ret = score->ifa->prefix_len;
1627 		score->matchlen = ret;
1628 		break;
1629 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1630 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1631 		/* Optimistic addresses still have lower precedence than other
1632 		 * preferred addresses.
1633 		 */
1634 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635 		break;
1636 #endif
1637 	default:
1638 		ret = 0;
1639 	}
1640 
1641 	if (ret)
1642 		__set_bit(i, score->scorebits);
1643 	score->rule = i;
1644 out:
1645 	return ret;
1646 }
1647 
1648 static int __ipv6_dev_get_saddr(struct net *net,
1649 				struct ipv6_saddr_dst *dst,
1650 				struct inet6_dev *idev,
1651 				struct ipv6_saddr_score *scores,
1652 				int hiscore_idx)
1653 {
1654 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1655 
1656 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1657 		int i;
1658 
1659 		/*
1660 		 * - Tentative Address (RFC2462 section 5.4)
1661 		 *  - A tentative address is not considered
1662 		 *    "assigned to an interface" in the traditional
1663 		 *    sense, unless it is also flagged as optimistic.
1664 		 * - Candidate Source Address (section 4)
1665 		 *  - In any case, anycast addresses, multicast
1666 		 *    addresses, and the unspecified address MUST
1667 		 *    NOT be included in a candidate set.
1668 		 */
1669 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1670 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671 			continue;
1672 
1673 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1674 
1675 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1676 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1677 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1678 					    idev->dev->name);
1679 			continue;
1680 		}
1681 
1682 		score->rule = -1;
1683 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1684 
1685 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1686 			int minihiscore, miniscore;
1687 
1688 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1689 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1690 
1691 			if (minihiscore > miniscore) {
1692 				if (i == IPV6_SADDR_RULE_SCOPE &&
1693 				    score->scopedist > 0) {
1694 					/*
1695 					 * special case:
1696 					 * each remaining entry
1697 					 * has too small (not enough)
1698 					 * scope, because ifa entries
1699 					 * are sorted by their scope
1700 					 * values.
1701 					 */
1702 					goto out;
1703 				}
1704 				break;
1705 			} else if (minihiscore < miniscore) {
1706 				swap(hiscore, score);
1707 				hiscore_idx = 1 - hiscore_idx;
1708 
1709 				/* restore our iterator */
1710 				score->ifa = hiscore->ifa;
1711 
1712 				break;
1713 			}
1714 		}
1715 	}
1716 out:
1717 	return hiscore_idx;
1718 }
1719 
1720 static int ipv6_get_saddr_master(struct net *net,
1721 				 const struct net_device *dst_dev,
1722 				 const struct net_device *master,
1723 				 struct ipv6_saddr_dst *dst,
1724 				 struct ipv6_saddr_score *scores,
1725 				 int hiscore_idx)
1726 {
1727 	struct inet6_dev *idev;
1728 
1729 	idev = __in6_dev_get(dst_dev);
1730 	if (idev)
1731 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1732 						   scores, hiscore_idx);
1733 
1734 	idev = __in6_dev_get(master);
1735 	if (idev)
1736 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1737 						   scores, hiscore_idx);
1738 
1739 	return hiscore_idx;
1740 }
1741 
1742 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1743 		       const struct in6_addr *daddr, unsigned int prefs,
1744 		       struct in6_addr *saddr)
1745 {
1746 	struct ipv6_saddr_score scores[2], *hiscore;
1747 	struct ipv6_saddr_dst dst;
1748 	struct inet6_dev *idev;
1749 	struct net_device *dev;
1750 	int dst_type;
1751 	bool use_oif_addr = false;
1752 	int hiscore_idx = 0;
1753 	int ret = 0;
1754 
1755 	dst_type = __ipv6_addr_type(daddr);
1756 	dst.addr = daddr;
1757 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1758 	dst.scope = __ipv6_addr_src_scope(dst_type);
1759 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760 	dst.prefs = prefs;
1761 
1762 	scores[hiscore_idx].rule = -1;
1763 	scores[hiscore_idx].ifa = NULL;
1764 
1765 	rcu_read_lock();
1766 
1767 	/* Candidate Source Address (section 4)
1768 	 *  - multicast and link-local destination address,
1769 	 *    the set of candidate source address MUST only
1770 	 *    include addresses assigned to interfaces
1771 	 *    belonging to the same link as the outgoing
1772 	 *    interface.
1773 	 * (- For site-local destination addresses, the
1774 	 *    set of candidate source addresses MUST only
1775 	 *    include addresses assigned to interfaces
1776 	 *    belonging to the same site as the outgoing
1777 	 *    interface.)
1778 	 *  - "It is RECOMMENDED that the candidate source addresses
1779 	 *    be the set of unicast addresses assigned to the
1780 	 *    interface that will be used to send to the destination
1781 	 *    (the 'outgoing' interface)." (RFC 6724)
1782 	 */
1783 	if (dst_dev) {
1784 		idev = __in6_dev_get(dst_dev);
1785 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1786 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1787 		    (idev && idev->cnf.use_oif_addrs_only)) {
1788 			use_oif_addr = true;
1789 		}
1790 	}
1791 
1792 	if (use_oif_addr) {
1793 		if (idev)
1794 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1795 	} else {
1796 		const struct net_device *master;
1797 		int master_idx = 0;
1798 
1799 		/* if dst_dev exists and is enslaved to an L3 device, then
1800 		 * prefer addresses from dst_dev and then the master over
1801 		 * any other enslaved devices in the L3 domain.
1802 		 */
1803 		master = l3mdev_master_dev_rcu(dst_dev);
1804 		if (master) {
1805 			master_idx = master->ifindex;
1806 
1807 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1808 							    master, &dst,
1809 							    scores, hiscore_idx);
1810 
1811 			if (scores[hiscore_idx].ifa)
1812 				goto out;
1813 		}
1814 
1815 		for_each_netdev_rcu(net, dev) {
1816 			/* only consider addresses on devices in the
1817 			 * same L3 domain
1818 			 */
1819 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1820 				continue;
1821 			idev = __in6_dev_get(dev);
1822 			if (!idev)
1823 				continue;
1824 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1825 		}
1826 	}
1827 
1828 out:
1829 	hiscore = &scores[hiscore_idx];
1830 	if (!hiscore->ifa)
1831 		ret = -EADDRNOTAVAIL;
1832 	else
1833 		*saddr = hiscore->ifa->addr;
1834 
1835 	rcu_read_unlock();
1836 	return ret;
1837 }
1838 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1839 
1840 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841 			      u32 banned_flags)
1842 {
1843 	struct inet6_ifaddr *ifp;
1844 	int err = -EADDRNOTAVAIL;
1845 
1846 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1847 		if (ifp->scope > IFA_LINK)
1848 			break;
1849 		if (ifp->scope == IFA_LINK &&
1850 		    !(ifp->flags & banned_flags)) {
1851 			*addr = ifp->addr;
1852 			err = 0;
1853 			break;
1854 		}
1855 	}
1856 	return err;
1857 }
1858 
1859 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860 		    u32 banned_flags)
1861 {
1862 	struct inet6_dev *idev;
1863 	int err = -EADDRNOTAVAIL;
1864 
1865 	rcu_read_lock();
1866 	idev = __in6_dev_get(dev);
1867 	if (idev) {
1868 		read_lock_bh(&idev->lock);
1869 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1870 		read_unlock_bh(&idev->lock);
1871 	}
1872 	rcu_read_unlock();
1873 	return err;
1874 }
1875 
1876 static int ipv6_count_addresses(const struct inet6_dev *idev)
1877 {
1878 	const struct inet6_ifaddr *ifp;
1879 	int cnt = 0;
1880 
1881 	rcu_read_lock();
1882 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1883 		cnt++;
1884 	rcu_read_unlock();
1885 	return cnt;
1886 }
1887 
1888 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1889 		  const struct net_device *dev, int strict)
1890 {
1891 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1892 				       strict, IFA_F_TENTATIVE);
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr);
1895 
1896 /* device argument is used to find the L3 domain of interest. If
1897  * skip_dev_check is set, then the ifp device is not checked against
1898  * the passed in dev argument. So the 2 cases for addresses checks are:
1899  *   1. does the address exist in the L3 domain that dev is part of
1900  *      (skip_dev_check = true), or
1901  *
1902  *   2. does the address exist on the specific device
1903  *      (skip_dev_check = false)
1904  */
1905 static struct net_device *
1906 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1907 			  const struct net_device *dev, bool skip_dev_check,
1908 			  int strict, u32 banned_flags)
1909 {
1910 	unsigned int hash = inet6_addr_hash(net, addr);
1911 	struct net_device *l3mdev, *ndev;
1912 	struct inet6_ifaddr *ifp;
1913 	u32 ifp_flags;
1914 
1915 	rcu_read_lock();
1916 
1917 	l3mdev = l3mdev_master_dev_rcu(dev);
1918 	if (skip_dev_check)
1919 		dev = NULL;
1920 
1921 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1922 		ndev = ifp->idev->dev;
1923 
1924 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1925 			continue;
1926 
1927 		/* Decouple optimistic from tentative for evaluation here.
1928 		 * Ban optimistic addresses explicitly, when required.
1929 		 */
1930 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1931 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1932 			    : ifp->flags;
1933 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1934 		    !(ifp_flags&banned_flags) &&
1935 		    (!dev || ndev == dev ||
1936 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1937 			rcu_read_unlock();
1938 			return ndev;
1939 		}
1940 	}
1941 
1942 	rcu_read_unlock();
1943 	return NULL;
1944 }
1945 
1946 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1947 			    const struct net_device *dev, bool skip_dev_check,
1948 			    int strict, u32 banned_flags)
1949 {
1950 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1951 					 strict, banned_flags) ? 1 : 0;
1952 }
1953 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1954 
1955 
1956 /* Compares an address/prefix_len with addresses on device @dev.
1957  * If one is found it returns true.
1958  */
1959 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1960 	const unsigned int prefix_len, struct net_device *dev)
1961 {
1962 	const struct inet6_ifaddr *ifa;
1963 	const struct inet6_dev *idev;
1964 	bool ret = false;
1965 
1966 	rcu_read_lock();
1967 	idev = __in6_dev_get(dev);
1968 	if (idev) {
1969 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1970 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1971 			if (ret)
1972 				break;
1973 		}
1974 	}
1975 	rcu_read_unlock();
1976 
1977 	return ret;
1978 }
1979 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1980 
1981 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1982 {
1983 	const struct inet6_ifaddr *ifa;
1984 	const struct inet6_dev *idev;
1985 	int	onlink;
1986 
1987 	onlink = 0;
1988 	rcu_read_lock();
1989 	idev = __in6_dev_get(dev);
1990 	if (idev) {
1991 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1992 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1993 						   ifa->prefix_len);
1994 			if (onlink)
1995 				break;
1996 		}
1997 	}
1998 	rcu_read_unlock();
1999 	return onlink;
2000 }
2001 EXPORT_SYMBOL(ipv6_chk_prefix);
2002 
2003 /**
2004  * ipv6_dev_find - find the first device with a given source address.
2005  * @net: the net namespace
2006  * @addr: the source address
2007  * @dev: used to find the L3 domain of interest
2008  *
2009  * The caller should be protected by RCU, or RTNL.
2010  */
2011 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2012 				 struct net_device *dev)
2013 {
2014 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2015 					 IFA_F_TENTATIVE);
2016 }
2017 EXPORT_SYMBOL(ipv6_dev_find);
2018 
2019 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2020 				     struct net_device *dev, int strict)
2021 {
2022 	unsigned int hash = inet6_addr_hash(net, addr);
2023 	struct inet6_ifaddr *ifp, *result = NULL;
2024 
2025 	rcu_read_lock();
2026 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2027 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2028 			if (!dev || ifp->idev->dev == dev ||
2029 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2030 				result = ifp;
2031 				in6_ifa_hold(ifp);
2032 				break;
2033 			}
2034 		}
2035 	}
2036 	rcu_read_unlock();
2037 
2038 	return result;
2039 }
2040 
2041 /* Gets referenced address, destroys ifaddr */
2042 
2043 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2044 {
2045 	if (dad_failed)
2046 		ifp->flags |= IFA_F_DADFAILED;
2047 
2048 	if (ifp->flags&IFA_F_TEMPORARY) {
2049 		struct inet6_ifaddr *ifpub;
2050 		spin_lock_bh(&ifp->lock);
2051 		ifpub = ifp->ifpub;
2052 		if (ifpub) {
2053 			in6_ifa_hold(ifpub);
2054 			spin_unlock_bh(&ifp->lock);
2055 			ipv6_create_tempaddr(ifpub, true);
2056 			in6_ifa_put(ifpub);
2057 		} else {
2058 			spin_unlock_bh(&ifp->lock);
2059 		}
2060 		ipv6_del_addr(ifp);
2061 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2062 		spin_lock_bh(&ifp->lock);
2063 		addrconf_del_dad_work(ifp);
2064 		ifp->flags |= IFA_F_TENTATIVE;
2065 		if (dad_failed)
2066 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2067 		spin_unlock_bh(&ifp->lock);
2068 		if (dad_failed)
2069 			ipv6_ifa_notify(0, ifp);
2070 		in6_ifa_put(ifp);
2071 	} else {
2072 		ipv6_del_addr(ifp);
2073 	}
2074 }
2075 
2076 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2077 {
2078 	int err = -ENOENT;
2079 
2080 	spin_lock_bh(&ifp->lock);
2081 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2082 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2083 		err = 0;
2084 	}
2085 	spin_unlock_bh(&ifp->lock);
2086 
2087 	return err;
2088 }
2089 
2090 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2091 {
2092 	struct inet6_dev *idev = ifp->idev;
2093 	struct net *net = dev_net(idev->dev);
2094 
2095 	if (addrconf_dad_end(ifp)) {
2096 		in6_ifa_put(ifp);
2097 		return;
2098 	}
2099 
2100 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2101 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2102 
2103 	spin_lock_bh(&ifp->lock);
2104 
2105 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2106 		struct in6_addr new_addr;
2107 		struct inet6_ifaddr *ifp2;
2108 		int retries = ifp->stable_privacy_retry + 1;
2109 		struct ifa6_config cfg = {
2110 			.pfx = &new_addr,
2111 			.plen = ifp->prefix_len,
2112 			.ifa_flags = ifp->flags,
2113 			.valid_lft = ifp->valid_lft,
2114 			.preferred_lft = ifp->prefered_lft,
2115 			.scope = ifp->scope,
2116 		};
2117 
2118 		if (retries > net->ipv6.sysctl.idgen_retries) {
2119 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2120 					     ifp->idev->dev->name);
2121 			goto errdad;
2122 		}
2123 
2124 		new_addr = ifp->addr;
2125 		if (ipv6_generate_stable_address(&new_addr, retries,
2126 						 idev))
2127 			goto errdad;
2128 
2129 		spin_unlock_bh(&ifp->lock);
2130 
2131 		if (idev->cnf.max_addresses &&
2132 		    ipv6_count_addresses(idev) >=
2133 		    idev->cnf.max_addresses)
2134 			goto lock_errdad;
2135 
2136 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2137 				     ifp->idev->dev->name);
2138 
2139 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2140 		if (IS_ERR(ifp2))
2141 			goto lock_errdad;
2142 
2143 		spin_lock_bh(&ifp2->lock);
2144 		ifp2->stable_privacy_retry = retries;
2145 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2146 		spin_unlock_bh(&ifp2->lock);
2147 
2148 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2149 		in6_ifa_put(ifp2);
2150 lock_errdad:
2151 		spin_lock_bh(&ifp->lock);
2152 	}
2153 
2154 errdad:
2155 	/* transition from _POSTDAD to _ERRDAD */
2156 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2157 	spin_unlock_bh(&ifp->lock);
2158 
2159 	addrconf_mod_dad_work(ifp, 0);
2160 	in6_ifa_put(ifp);
2161 }
2162 
2163 /* Join to solicited addr multicast group.
2164  * caller must hold RTNL */
2165 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2166 {
2167 	struct in6_addr maddr;
2168 
2169 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2170 		return;
2171 
2172 	addrconf_addr_solict_mult(addr, &maddr);
2173 	ipv6_dev_mc_inc(dev, &maddr);
2174 }
2175 
2176 /* caller must hold RTNL */
2177 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2178 {
2179 	struct in6_addr maddr;
2180 
2181 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2182 		return;
2183 
2184 	addrconf_addr_solict_mult(addr, &maddr);
2185 	__ipv6_dev_mc_dec(idev, &maddr);
2186 }
2187 
2188 /* caller must hold RTNL */
2189 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2190 {
2191 	struct in6_addr addr;
2192 
2193 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2194 		return;
2195 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2196 	if (ipv6_addr_any(&addr))
2197 		return;
2198 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2199 }
2200 
2201 /* caller must hold RTNL */
2202 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2203 {
2204 	struct in6_addr addr;
2205 
2206 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2207 		return;
2208 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2209 	if (ipv6_addr_any(&addr))
2210 		return;
2211 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2212 }
2213 
2214 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2215 {
2216 	switch (dev->addr_len) {
2217 	case ETH_ALEN:
2218 		memcpy(eui, dev->dev_addr, 3);
2219 		eui[3] = 0xFF;
2220 		eui[4] = 0xFE;
2221 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2222 		break;
2223 	case EUI64_ADDR_LEN:
2224 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2225 		eui[0] ^= 2;
2226 		break;
2227 	default:
2228 		return -1;
2229 	}
2230 
2231 	return 0;
2232 }
2233 
2234 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2235 {
2236 	const union fwnet_hwaddr *ha;
2237 
2238 	if (dev->addr_len != FWNET_ALEN)
2239 		return -1;
2240 
2241 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2242 
2243 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2244 	eui[0] ^= 2;
2245 	return 0;
2246 }
2247 
2248 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2249 {
2250 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2251 	if (dev->addr_len != ARCNET_ALEN)
2252 		return -1;
2253 	memset(eui, 0, 7);
2254 	eui[7] = *(u8 *)dev->dev_addr;
2255 	return 0;
2256 }
2257 
2258 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2259 {
2260 	if (dev->addr_len != INFINIBAND_ALEN)
2261 		return -1;
2262 	memcpy(eui, dev->dev_addr + 12, 8);
2263 	eui[0] |= 2;
2264 	return 0;
2265 }
2266 
2267 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2268 {
2269 	if (addr == 0)
2270 		return -1;
2271 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2272 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2273 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2274 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2275 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2276 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2277 	eui[1] = 0;
2278 	eui[2] = 0x5E;
2279 	eui[3] = 0xFE;
2280 	memcpy(eui + 4, &addr, 4);
2281 	return 0;
2282 }
2283 
2284 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2285 {
2286 	if (dev->priv_flags & IFF_ISATAP)
2287 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2288 	return -1;
2289 }
2290 
2291 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2292 {
2293 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2294 }
2295 
2296 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2297 {
2298 	memcpy(eui, dev->perm_addr, 3);
2299 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2300 	eui[3] = 0xFF;
2301 	eui[4] = 0xFE;
2302 	eui[0] ^= 2;
2303 	return 0;
2304 }
2305 
2306 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2307 {
2308 	switch (dev->type) {
2309 	case ARPHRD_ETHER:
2310 	case ARPHRD_FDDI:
2311 		return addrconf_ifid_eui48(eui, dev);
2312 	case ARPHRD_ARCNET:
2313 		return addrconf_ifid_arcnet(eui, dev);
2314 	case ARPHRD_INFINIBAND:
2315 		return addrconf_ifid_infiniband(eui, dev);
2316 	case ARPHRD_SIT:
2317 		return addrconf_ifid_sit(eui, dev);
2318 	case ARPHRD_IPGRE:
2319 	case ARPHRD_TUNNEL:
2320 		return addrconf_ifid_gre(eui, dev);
2321 	case ARPHRD_6LOWPAN:
2322 		return addrconf_ifid_6lowpan(eui, dev);
2323 	case ARPHRD_IEEE1394:
2324 		return addrconf_ifid_ieee1394(eui, dev);
2325 	case ARPHRD_TUNNEL6:
2326 	case ARPHRD_IP6GRE:
2327 	case ARPHRD_RAWIP:
2328 		return addrconf_ifid_ip6tnl(eui, dev);
2329 	}
2330 	return -1;
2331 }
2332 
2333 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2334 {
2335 	int err = -1;
2336 	struct inet6_ifaddr *ifp;
2337 
2338 	read_lock_bh(&idev->lock);
2339 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2340 		if (ifp->scope > IFA_LINK)
2341 			break;
2342 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2343 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2344 			err = 0;
2345 			break;
2346 		}
2347 	}
2348 	read_unlock_bh(&idev->lock);
2349 	return err;
2350 }
2351 
2352 /* Generation of a randomized Interface Identifier
2353  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2354  */
2355 
2356 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2357 {
2358 regen:
2359 	get_random_bytes(&addr->s6_addr[8], 8);
2360 
2361 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2362 	 * check if generated address is not inappropriate:
2363 	 *
2364 	 * - Reserved IPv6 Interface Identifiers
2365 	 * - XXX: already assigned to an address on the device
2366 	 */
2367 
2368 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2369 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2370 		goto regen;
2371 
2372 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2373 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2374 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2375 	 */
2376 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2377 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2378 		goto regen;
2379 
2380 	/* Reserved subnet anycast addresses */
2381 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2382 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2383 		goto regen;
2384 }
2385 
2386 /*
2387  *	Add prefix route.
2388  */
2389 
2390 static void
2391 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2392 		      struct net_device *dev, unsigned long expires,
2393 		      u32 flags, gfp_t gfp_flags)
2394 {
2395 	struct fib6_config cfg = {
2396 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2397 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2398 		.fc_ifindex = dev->ifindex,
2399 		.fc_expires = expires,
2400 		.fc_dst_len = plen,
2401 		.fc_flags = RTF_UP | flags,
2402 		.fc_nlinfo.nl_net = dev_net(dev),
2403 		.fc_protocol = RTPROT_KERNEL,
2404 		.fc_type = RTN_UNICAST,
2405 	};
2406 
2407 	cfg.fc_dst = *pfx;
2408 
2409 	/* Prevent useless cloning on PtP SIT.
2410 	   This thing is done here expecting that the whole
2411 	   class of non-broadcast devices need not cloning.
2412 	 */
2413 #if IS_ENABLED(CONFIG_IPV6_SIT)
2414 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2415 		cfg.fc_flags |= RTF_NONEXTHOP;
2416 #endif
2417 
2418 	ip6_route_add(&cfg, gfp_flags, NULL);
2419 }
2420 
2421 
2422 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2423 						  int plen,
2424 						  const struct net_device *dev,
2425 						  u32 flags, u32 noflags,
2426 						  bool no_gw)
2427 {
2428 	struct fib6_node *fn;
2429 	struct fib6_info *rt = NULL;
2430 	struct fib6_table *table;
2431 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2432 
2433 	table = fib6_get_table(dev_net(dev), tb_id);
2434 	if (!table)
2435 		return NULL;
2436 
2437 	rcu_read_lock();
2438 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2439 	if (!fn)
2440 		goto out;
2441 
2442 	for_each_fib6_node_rt_rcu(fn) {
2443 		/* prefix routes only use builtin fib6_nh */
2444 		if (rt->nh)
2445 			continue;
2446 
2447 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2448 			continue;
2449 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2450 			continue;
2451 		if ((rt->fib6_flags & flags) != flags)
2452 			continue;
2453 		if ((rt->fib6_flags & noflags) != 0)
2454 			continue;
2455 		if (!fib6_info_hold_safe(rt))
2456 			continue;
2457 		break;
2458 	}
2459 out:
2460 	rcu_read_unlock();
2461 	return rt;
2462 }
2463 
2464 
2465 /* Create "default" multicast route to the interface */
2466 
2467 static void addrconf_add_mroute(struct net_device *dev)
2468 {
2469 	struct fib6_config cfg = {
2470 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2471 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2472 		.fc_ifindex = dev->ifindex,
2473 		.fc_dst_len = 8,
2474 		.fc_flags = RTF_UP,
2475 		.fc_type = RTN_MULTICAST,
2476 		.fc_nlinfo.nl_net = dev_net(dev),
2477 		.fc_protocol = RTPROT_KERNEL,
2478 	};
2479 
2480 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2481 
2482 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2483 }
2484 
2485 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2486 {
2487 	struct inet6_dev *idev;
2488 
2489 	ASSERT_RTNL();
2490 
2491 	idev = ipv6_find_idev(dev);
2492 	if (IS_ERR(idev))
2493 		return idev;
2494 
2495 	if (idev->cnf.disable_ipv6)
2496 		return ERR_PTR(-EACCES);
2497 
2498 	/* Add default multicast route */
2499 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2500 		addrconf_add_mroute(dev);
2501 
2502 	return idev;
2503 }
2504 
2505 static void manage_tempaddrs(struct inet6_dev *idev,
2506 			     struct inet6_ifaddr *ifp,
2507 			     __u32 valid_lft, __u32 prefered_lft,
2508 			     bool create, unsigned long now)
2509 {
2510 	u32 flags;
2511 	struct inet6_ifaddr *ift;
2512 
2513 	read_lock_bh(&idev->lock);
2514 	/* update all temporary addresses in the list */
2515 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2516 		int age, max_valid, max_prefered;
2517 
2518 		if (ifp != ift->ifpub)
2519 			continue;
2520 
2521 		/* RFC 4941 section 3.3:
2522 		 * If a received option will extend the lifetime of a public
2523 		 * address, the lifetimes of temporary addresses should
2524 		 * be extended, subject to the overall constraint that no
2525 		 * temporary addresses should ever remain "valid" or "preferred"
2526 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2527 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2528 		 */
2529 		age = (now - ift->cstamp) / HZ;
2530 		max_valid = idev->cnf.temp_valid_lft - age;
2531 		if (max_valid < 0)
2532 			max_valid = 0;
2533 
2534 		max_prefered = idev->cnf.temp_prefered_lft -
2535 			       idev->desync_factor - age;
2536 		if (max_prefered < 0)
2537 			max_prefered = 0;
2538 
2539 		if (valid_lft > max_valid)
2540 			valid_lft = max_valid;
2541 
2542 		if (prefered_lft > max_prefered)
2543 			prefered_lft = max_prefered;
2544 
2545 		spin_lock(&ift->lock);
2546 		flags = ift->flags;
2547 		ift->valid_lft = valid_lft;
2548 		ift->prefered_lft = prefered_lft;
2549 		ift->tstamp = now;
2550 		if (prefered_lft > 0)
2551 			ift->flags &= ~IFA_F_DEPRECATED;
2552 
2553 		spin_unlock(&ift->lock);
2554 		if (!(flags&IFA_F_TENTATIVE))
2555 			ipv6_ifa_notify(0, ift);
2556 	}
2557 
2558 	if ((create || list_empty(&idev->tempaddr_list)) &&
2559 	    idev->cnf.use_tempaddr > 0) {
2560 		/* When a new public address is created as described
2561 		 * in [ADDRCONF], also create a new temporary address.
2562 		 * Also create a temporary address if it's enabled but
2563 		 * no temporary address currently exists.
2564 		 */
2565 		read_unlock_bh(&idev->lock);
2566 		ipv6_create_tempaddr(ifp, false);
2567 	} else {
2568 		read_unlock_bh(&idev->lock);
2569 	}
2570 }
2571 
2572 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2573 {
2574 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2575 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2576 }
2577 
2578 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2579 				 const struct prefix_info *pinfo,
2580 				 struct inet6_dev *in6_dev,
2581 				 const struct in6_addr *addr, int addr_type,
2582 				 u32 addr_flags, bool sllao, bool tokenized,
2583 				 __u32 valid_lft, u32 prefered_lft)
2584 {
2585 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2586 	int create = 0, update_lft = 0;
2587 
2588 	if (!ifp && valid_lft) {
2589 		int max_addresses = in6_dev->cnf.max_addresses;
2590 		struct ifa6_config cfg = {
2591 			.pfx = addr,
2592 			.plen = pinfo->prefix_len,
2593 			.ifa_flags = addr_flags,
2594 			.valid_lft = valid_lft,
2595 			.preferred_lft = prefered_lft,
2596 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2597 			.ifa_proto = IFAPROT_KERNEL_RA
2598 		};
2599 
2600 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2601 		if ((net->ipv6.devconf_all->optimistic_dad ||
2602 		     in6_dev->cnf.optimistic_dad) &&
2603 		    !net->ipv6.devconf_all->forwarding && sllao)
2604 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2605 #endif
2606 
2607 		/* Do not allow to create too much of autoconfigured
2608 		 * addresses; this would be too easy way to crash kernel.
2609 		 */
2610 		if (!max_addresses ||
2611 		    ipv6_count_addresses(in6_dev) < max_addresses)
2612 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2613 
2614 		if (IS_ERR_OR_NULL(ifp))
2615 			return -1;
2616 
2617 		create = 1;
2618 		spin_lock_bh(&ifp->lock);
2619 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2620 		ifp->cstamp = jiffies;
2621 		ifp->tokenized = tokenized;
2622 		spin_unlock_bh(&ifp->lock);
2623 		addrconf_dad_start(ifp);
2624 	}
2625 
2626 	if (ifp) {
2627 		u32 flags;
2628 		unsigned long now;
2629 		u32 stored_lft;
2630 
2631 		/* update lifetime (RFC2462 5.5.3 e) */
2632 		spin_lock_bh(&ifp->lock);
2633 		now = jiffies;
2634 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2635 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2636 		else
2637 			stored_lft = 0;
2638 		if (!create && stored_lft) {
2639 			const u32 minimum_lft = min_t(u32,
2640 				stored_lft, MIN_VALID_LIFETIME);
2641 			valid_lft = max(valid_lft, minimum_lft);
2642 
2643 			/* RFC4862 Section 5.5.3e:
2644 			 * "Note that the preferred lifetime of the
2645 			 *  corresponding address is always reset to
2646 			 *  the Preferred Lifetime in the received
2647 			 *  Prefix Information option, regardless of
2648 			 *  whether the valid lifetime is also reset or
2649 			 *  ignored."
2650 			 *
2651 			 * So we should always update prefered_lft here.
2652 			 */
2653 			update_lft = 1;
2654 		}
2655 
2656 		if (update_lft) {
2657 			ifp->valid_lft = valid_lft;
2658 			ifp->prefered_lft = prefered_lft;
2659 			ifp->tstamp = now;
2660 			flags = ifp->flags;
2661 			ifp->flags &= ~IFA_F_DEPRECATED;
2662 			spin_unlock_bh(&ifp->lock);
2663 
2664 			if (!(flags&IFA_F_TENTATIVE))
2665 				ipv6_ifa_notify(0, ifp);
2666 		} else
2667 			spin_unlock_bh(&ifp->lock);
2668 
2669 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2670 				 create, now);
2671 
2672 		in6_ifa_put(ifp);
2673 		addrconf_verify(net);
2674 	}
2675 
2676 	return 0;
2677 }
2678 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2679 
2680 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2681 {
2682 	struct prefix_info *pinfo;
2683 	__u32 valid_lft;
2684 	__u32 prefered_lft;
2685 	int addr_type, err;
2686 	u32 addr_flags = 0;
2687 	struct inet6_dev *in6_dev;
2688 	struct net *net = dev_net(dev);
2689 
2690 	pinfo = (struct prefix_info *) opt;
2691 
2692 	if (len < sizeof(struct prefix_info)) {
2693 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2694 		return;
2695 	}
2696 
2697 	/*
2698 	 *	Validation checks ([ADDRCONF], page 19)
2699 	 */
2700 
2701 	addr_type = ipv6_addr_type(&pinfo->prefix);
2702 
2703 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2704 		return;
2705 
2706 	valid_lft = ntohl(pinfo->valid);
2707 	prefered_lft = ntohl(pinfo->prefered);
2708 
2709 	if (prefered_lft > valid_lft) {
2710 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2711 		return;
2712 	}
2713 
2714 	in6_dev = in6_dev_get(dev);
2715 
2716 	if (!in6_dev) {
2717 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2718 				    dev->name);
2719 		return;
2720 	}
2721 
2722 	/*
2723 	 *	Two things going on here:
2724 	 *	1) Add routes for on-link prefixes
2725 	 *	2) Configure prefixes with the auto flag set
2726 	 */
2727 
2728 	if (pinfo->onlink) {
2729 		struct fib6_info *rt;
2730 		unsigned long rt_expires;
2731 
2732 		/* Avoid arithmetic overflow. Really, we could
2733 		 * save rt_expires in seconds, likely valid_lft,
2734 		 * but it would require division in fib gc, that it
2735 		 * not good.
2736 		 */
2737 		if (HZ > USER_HZ)
2738 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2739 		else
2740 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2741 
2742 		if (addrconf_finite_timeout(rt_expires))
2743 			rt_expires *= HZ;
2744 
2745 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2746 					       pinfo->prefix_len,
2747 					       dev,
2748 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2749 					       RTF_DEFAULT, true);
2750 
2751 		if (rt) {
2752 			/* Autoconf prefix route */
2753 			if (valid_lft == 0) {
2754 				ip6_del_rt(net, rt, false);
2755 				rt = NULL;
2756 			} else if (addrconf_finite_timeout(rt_expires)) {
2757 				/* not infinity */
2758 				fib6_set_expires(rt, jiffies + rt_expires);
2759 			} else {
2760 				fib6_clean_expires(rt);
2761 			}
2762 		} else if (valid_lft) {
2763 			clock_t expires = 0;
2764 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2765 			if (addrconf_finite_timeout(rt_expires)) {
2766 				/* not infinity */
2767 				flags |= RTF_EXPIRES;
2768 				expires = jiffies_to_clock_t(rt_expires);
2769 			}
2770 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2771 					      0, dev, expires, flags,
2772 					      GFP_ATOMIC);
2773 		}
2774 		fib6_info_release(rt);
2775 	}
2776 
2777 	/* Try to figure out our local address for this prefix */
2778 
2779 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2780 		struct in6_addr addr;
2781 		bool tokenized = false, dev_addr_generated = false;
2782 
2783 		if (pinfo->prefix_len == 64) {
2784 			memcpy(&addr, &pinfo->prefix, 8);
2785 
2786 			if (!ipv6_addr_any(&in6_dev->token)) {
2787 				read_lock_bh(&in6_dev->lock);
2788 				memcpy(addr.s6_addr + 8,
2789 				       in6_dev->token.s6_addr + 8, 8);
2790 				read_unlock_bh(&in6_dev->lock);
2791 				tokenized = true;
2792 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2793 				   !ipv6_generate_stable_address(&addr, 0,
2794 								 in6_dev)) {
2795 				addr_flags |= IFA_F_STABLE_PRIVACY;
2796 				goto ok;
2797 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2798 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2799 				goto put;
2800 			} else {
2801 				dev_addr_generated = true;
2802 			}
2803 			goto ok;
2804 		}
2805 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2806 				    pinfo->prefix_len);
2807 		goto put;
2808 
2809 ok:
2810 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2811 						   &addr, addr_type,
2812 						   addr_flags, sllao,
2813 						   tokenized, valid_lft,
2814 						   prefered_lft);
2815 		if (err)
2816 			goto put;
2817 
2818 		/* Ignore error case here because previous prefix add addr was
2819 		 * successful which will be notified.
2820 		 */
2821 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2822 					      addr_type, addr_flags, sllao,
2823 					      tokenized, valid_lft,
2824 					      prefered_lft,
2825 					      dev_addr_generated);
2826 	}
2827 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2828 put:
2829 	in6_dev_put(in6_dev);
2830 }
2831 
2832 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2833 		struct in6_ifreq *ireq)
2834 {
2835 	struct ip_tunnel_parm p = { };
2836 	int err;
2837 
2838 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2839 		return -EADDRNOTAVAIL;
2840 
2841 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2842 	p.iph.version = 4;
2843 	p.iph.ihl = 5;
2844 	p.iph.protocol = IPPROTO_IPV6;
2845 	p.iph.ttl = 64;
2846 
2847 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2848 		return -EOPNOTSUPP;
2849 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2850 	if (err)
2851 		return err;
2852 
2853 	dev = __dev_get_by_name(net, p.name);
2854 	if (!dev)
2855 		return -ENOBUFS;
2856 	return dev_open(dev, NULL);
2857 }
2858 
2859 /*
2860  *	Set destination address.
2861  *	Special case for SIT interfaces where we create a new "virtual"
2862  *	device.
2863  */
2864 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2865 {
2866 	struct net_device *dev;
2867 	struct in6_ifreq ireq;
2868 	int err = -ENODEV;
2869 
2870 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2871 		return -ENODEV;
2872 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2873 		return -EFAULT;
2874 
2875 	rtnl_lock();
2876 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2877 	if (dev && dev->type == ARPHRD_SIT)
2878 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2879 	rtnl_unlock();
2880 	return err;
2881 }
2882 
2883 static int ipv6_mc_config(struct sock *sk, bool join,
2884 			  const struct in6_addr *addr, int ifindex)
2885 {
2886 	int ret;
2887 
2888 	ASSERT_RTNL();
2889 
2890 	lock_sock(sk);
2891 	if (join)
2892 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2893 	else
2894 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2895 	release_sock(sk);
2896 
2897 	return ret;
2898 }
2899 
2900 /*
2901  *	Manual configuration of address on an interface
2902  */
2903 static int inet6_addr_add(struct net *net, int ifindex,
2904 			  struct ifa6_config *cfg,
2905 			  struct netlink_ext_ack *extack)
2906 {
2907 	struct inet6_ifaddr *ifp;
2908 	struct inet6_dev *idev;
2909 	struct net_device *dev;
2910 	unsigned long timeout;
2911 	clock_t expires;
2912 	u32 flags;
2913 
2914 	ASSERT_RTNL();
2915 
2916 	if (cfg->plen > 128)
2917 		return -EINVAL;
2918 
2919 	/* check the lifetime */
2920 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2921 		return -EINVAL;
2922 
2923 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2924 		return -EINVAL;
2925 
2926 	dev = __dev_get_by_index(net, ifindex);
2927 	if (!dev)
2928 		return -ENODEV;
2929 
2930 	idev = addrconf_add_dev(dev);
2931 	if (IS_ERR(idev))
2932 		return PTR_ERR(idev);
2933 
2934 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2935 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2936 					 true, cfg->pfx, ifindex);
2937 
2938 		if (ret < 0)
2939 			return ret;
2940 	}
2941 
2942 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2943 
2944 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2945 	if (addrconf_finite_timeout(timeout)) {
2946 		expires = jiffies_to_clock_t(timeout * HZ);
2947 		cfg->valid_lft = timeout;
2948 		flags = RTF_EXPIRES;
2949 	} else {
2950 		expires = 0;
2951 		flags = 0;
2952 		cfg->ifa_flags |= IFA_F_PERMANENT;
2953 	}
2954 
2955 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2956 	if (addrconf_finite_timeout(timeout)) {
2957 		if (timeout == 0)
2958 			cfg->ifa_flags |= IFA_F_DEPRECATED;
2959 		cfg->preferred_lft = timeout;
2960 	}
2961 
2962 	ifp = ipv6_add_addr(idev, cfg, true, extack);
2963 	if (!IS_ERR(ifp)) {
2964 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2965 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2966 					      ifp->rt_priority, dev, expires,
2967 					      flags, GFP_KERNEL);
2968 		}
2969 
2970 		/* Send a netlink notification if DAD is enabled and
2971 		 * optimistic flag is not set
2972 		 */
2973 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2974 			ipv6_ifa_notify(0, ifp);
2975 		/*
2976 		 * Note that section 3.1 of RFC 4429 indicates
2977 		 * that the Optimistic flag should not be set for
2978 		 * manually configured addresses
2979 		 */
2980 		addrconf_dad_start(ifp);
2981 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2982 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
2983 					 cfg->preferred_lft, true, jiffies);
2984 		in6_ifa_put(ifp);
2985 		addrconf_verify_rtnl(net);
2986 		return 0;
2987 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2988 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2989 			       cfg->pfx, ifindex);
2990 	}
2991 
2992 	return PTR_ERR(ifp);
2993 }
2994 
2995 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2996 			  const struct in6_addr *pfx, unsigned int plen)
2997 {
2998 	struct inet6_ifaddr *ifp;
2999 	struct inet6_dev *idev;
3000 	struct net_device *dev;
3001 
3002 	if (plen > 128)
3003 		return -EINVAL;
3004 
3005 	dev = __dev_get_by_index(net, ifindex);
3006 	if (!dev)
3007 		return -ENODEV;
3008 
3009 	idev = __in6_dev_get(dev);
3010 	if (!idev)
3011 		return -ENXIO;
3012 
3013 	read_lock_bh(&idev->lock);
3014 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3015 		if (ifp->prefix_len == plen &&
3016 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3017 			in6_ifa_hold(ifp);
3018 			read_unlock_bh(&idev->lock);
3019 
3020 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3021 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
3022 				manage_tempaddrs(idev, ifp, 0, 0, false,
3023 						 jiffies);
3024 			ipv6_del_addr(ifp);
3025 			addrconf_verify_rtnl(net);
3026 			if (ipv6_addr_is_multicast(pfx)) {
3027 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3028 					       false, pfx, dev->ifindex);
3029 			}
3030 			return 0;
3031 		}
3032 	}
3033 	read_unlock_bh(&idev->lock);
3034 	return -EADDRNOTAVAIL;
3035 }
3036 
3037 
3038 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3039 {
3040 	struct ifa6_config cfg = {
3041 		.ifa_flags = IFA_F_PERMANENT,
3042 		.preferred_lft = INFINITY_LIFE_TIME,
3043 		.valid_lft = INFINITY_LIFE_TIME,
3044 	};
3045 	struct in6_ifreq ireq;
3046 	int err;
3047 
3048 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3049 		return -EPERM;
3050 
3051 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3052 		return -EFAULT;
3053 
3054 	cfg.pfx = &ireq.ifr6_addr;
3055 	cfg.plen = ireq.ifr6_prefixlen;
3056 
3057 	rtnl_lock();
3058 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3059 	rtnl_unlock();
3060 	return err;
3061 }
3062 
3063 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3064 {
3065 	struct in6_ifreq ireq;
3066 	int err;
3067 
3068 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3069 		return -EPERM;
3070 
3071 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3072 		return -EFAULT;
3073 
3074 	rtnl_lock();
3075 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3076 			     ireq.ifr6_prefixlen);
3077 	rtnl_unlock();
3078 	return err;
3079 }
3080 
3081 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3082 		     int plen, int scope, u8 proto)
3083 {
3084 	struct inet6_ifaddr *ifp;
3085 	struct ifa6_config cfg = {
3086 		.pfx = addr,
3087 		.plen = plen,
3088 		.ifa_flags = IFA_F_PERMANENT,
3089 		.valid_lft = INFINITY_LIFE_TIME,
3090 		.preferred_lft = INFINITY_LIFE_TIME,
3091 		.scope = scope,
3092 		.ifa_proto = proto
3093 	};
3094 
3095 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3096 	if (!IS_ERR(ifp)) {
3097 		spin_lock_bh(&ifp->lock);
3098 		ifp->flags &= ~IFA_F_TENTATIVE;
3099 		spin_unlock_bh(&ifp->lock);
3100 		rt_genid_bump_ipv6(dev_net(idev->dev));
3101 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3102 		in6_ifa_put(ifp);
3103 	}
3104 }
3105 
3106 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3107 static void add_v4_addrs(struct inet6_dev *idev)
3108 {
3109 	struct in6_addr addr;
3110 	struct net_device *dev;
3111 	struct net *net = dev_net(idev->dev);
3112 	int scope, plen, offset = 0;
3113 	u32 pflags = 0;
3114 
3115 	ASSERT_RTNL();
3116 
3117 	memset(&addr, 0, sizeof(struct in6_addr));
3118 	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3119 	if (idev->dev->addr_len == sizeof(struct in6_addr))
3120 		offset = sizeof(struct in6_addr) - 4;
3121 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3122 
3123 	if (idev->dev->flags&IFF_POINTOPOINT) {
3124 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3125 			return;
3126 
3127 		addr.s6_addr32[0] = htonl(0xfe800000);
3128 		scope = IFA_LINK;
3129 		plen = 64;
3130 	} else {
3131 		scope = IPV6_ADDR_COMPATv4;
3132 		plen = 96;
3133 		pflags |= RTF_NONEXTHOP;
3134 	}
3135 
3136 	if (addr.s6_addr32[3]) {
3137 		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3138 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3139 				      GFP_KERNEL);
3140 		return;
3141 	}
3142 
3143 	for_each_netdev(net, dev) {
3144 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3145 		if (in_dev && (dev->flags & IFF_UP)) {
3146 			struct in_ifaddr *ifa;
3147 			int flag = scope;
3148 
3149 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3150 				addr.s6_addr32[3] = ifa->ifa_local;
3151 
3152 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3153 					continue;
3154 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3155 					if (idev->dev->flags&IFF_POINTOPOINT)
3156 						continue;
3157 					flag |= IFA_HOST;
3158 				}
3159 
3160 				add_addr(idev, &addr, plen, flag,
3161 					 IFAPROT_UNSPEC);
3162 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3163 						      0, pflags, GFP_KERNEL);
3164 			}
3165 		}
3166 	}
3167 }
3168 #endif
3169 
3170 static void init_loopback(struct net_device *dev)
3171 {
3172 	struct inet6_dev  *idev;
3173 
3174 	/* ::1 */
3175 
3176 	ASSERT_RTNL();
3177 
3178 	idev = ipv6_find_idev(dev);
3179 	if (IS_ERR(idev)) {
3180 		pr_debug("%s: add_dev failed\n", __func__);
3181 		return;
3182 	}
3183 
3184 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3185 }
3186 
3187 void addrconf_add_linklocal(struct inet6_dev *idev,
3188 			    const struct in6_addr *addr, u32 flags)
3189 {
3190 	struct ifa6_config cfg = {
3191 		.pfx = addr,
3192 		.plen = 64,
3193 		.ifa_flags = flags | IFA_F_PERMANENT,
3194 		.valid_lft = INFINITY_LIFE_TIME,
3195 		.preferred_lft = INFINITY_LIFE_TIME,
3196 		.scope = IFA_LINK,
3197 		.ifa_proto = IFAPROT_KERNEL_LL
3198 	};
3199 	struct inet6_ifaddr *ifp;
3200 
3201 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3202 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3203 	     idev->cnf.optimistic_dad) &&
3204 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3205 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3206 #endif
3207 
3208 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3209 	if (!IS_ERR(ifp)) {
3210 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3211 				      0, 0, GFP_ATOMIC);
3212 		addrconf_dad_start(ifp);
3213 		in6_ifa_put(ifp);
3214 	}
3215 }
3216 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3217 
3218 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3219 {
3220 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3221 		return true;
3222 
3223 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3224 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3225 		return true;
3226 
3227 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3228 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3229 		return true;
3230 
3231 	return false;
3232 }
3233 
3234 static int ipv6_generate_stable_address(struct in6_addr *address,
3235 					u8 dad_count,
3236 					const struct inet6_dev *idev)
3237 {
3238 	static DEFINE_SPINLOCK(lock);
3239 	static __u32 digest[SHA1_DIGEST_WORDS];
3240 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3241 
3242 	static union {
3243 		char __data[SHA1_BLOCK_SIZE];
3244 		struct {
3245 			struct in6_addr secret;
3246 			__be32 prefix[2];
3247 			unsigned char hwaddr[MAX_ADDR_LEN];
3248 			u8 dad_count;
3249 		} __packed;
3250 	} data;
3251 
3252 	struct in6_addr secret;
3253 	struct in6_addr temp;
3254 	struct net *net = dev_net(idev->dev);
3255 
3256 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3257 
3258 	if (idev->cnf.stable_secret.initialized)
3259 		secret = idev->cnf.stable_secret.secret;
3260 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3261 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3262 	else
3263 		return -1;
3264 
3265 retry:
3266 	spin_lock_bh(&lock);
3267 
3268 	sha1_init(digest);
3269 	memset(&data, 0, sizeof(data));
3270 	memset(workspace, 0, sizeof(workspace));
3271 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3272 	data.prefix[0] = address->s6_addr32[0];
3273 	data.prefix[1] = address->s6_addr32[1];
3274 	data.secret = secret;
3275 	data.dad_count = dad_count;
3276 
3277 	sha1_transform(digest, data.__data, workspace);
3278 
3279 	temp = *address;
3280 	temp.s6_addr32[2] = (__force __be32)digest[0];
3281 	temp.s6_addr32[3] = (__force __be32)digest[1];
3282 
3283 	spin_unlock_bh(&lock);
3284 
3285 	if (ipv6_reserved_interfaceid(temp)) {
3286 		dad_count++;
3287 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3288 			return -1;
3289 		goto retry;
3290 	}
3291 
3292 	*address = temp;
3293 	return 0;
3294 }
3295 
3296 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3297 {
3298 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3299 
3300 	if (s->initialized)
3301 		return;
3302 	s = &idev->cnf.stable_secret;
3303 	get_random_bytes(&s->secret, sizeof(s->secret));
3304 	s->initialized = true;
3305 }
3306 
3307 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3308 {
3309 	struct in6_addr addr;
3310 
3311 	/* no link local addresses on L3 master devices */
3312 	if (netif_is_l3_master(idev->dev))
3313 		return;
3314 
3315 	/* no link local addresses on devices flagged as slaves */
3316 	if (idev->dev->flags & IFF_SLAVE)
3317 		return;
3318 
3319 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3320 
3321 	switch (idev->cnf.addr_gen_mode) {
3322 	case IN6_ADDR_GEN_MODE_RANDOM:
3323 		ipv6_gen_mode_random_init(idev);
3324 		fallthrough;
3325 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3326 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3327 			addrconf_add_linklocal(idev, &addr,
3328 					       IFA_F_STABLE_PRIVACY);
3329 		else if (prefix_route)
3330 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3331 					      0, 0, GFP_KERNEL);
3332 		break;
3333 	case IN6_ADDR_GEN_MODE_EUI64:
3334 		/* addrconf_add_linklocal also adds a prefix_route and we
3335 		 * only need to care about prefix routes if ipv6_generate_eui64
3336 		 * couldn't generate one.
3337 		 */
3338 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3339 			addrconf_add_linklocal(idev, &addr, 0);
3340 		else if (prefix_route)
3341 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3342 					      0, 0, GFP_KERNEL);
3343 		break;
3344 	case IN6_ADDR_GEN_MODE_NONE:
3345 	default:
3346 		/* will not add any link local address */
3347 		break;
3348 	}
3349 }
3350 
3351 static void addrconf_dev_config(struct net_device *dev)
3352 {
3353 	struct inet6_dev *idev;
3354 
3355 	ASSERT_RTNL();
3356 
3357 	if ((dev->type != ARPHRD_ETHER) &&
3358 	    (dev->type != ARPHRD_FDDI) &&
3359 	    (dev->type != ARPHRD_ARCNET) &&
3360 	    (dev->type != ARPHRD_INFINIBAND) &&
3361 	    (dev->type != ARPHRD_IEEE1394) &&
3362 	    (dev->type != ARPHRD_TUNNEL6) &&
3363 	    (dev->type != ARPHRD_6LOWPAN) &&
3364 	    (dev->type != ARPHRD_TUNNEL) &&
3365 	    (dev->type != ARPHRD_NONE) &&
3366 	    (dev->type != ARPHRD_RAWIP)) {
3367 		/* Alas, we support only Ethernet autoconfiguration. */
3368 		idev = __in6_dev_get(dev);
3369 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3370 		    dev->flags & IFF_MULTICAST)
3371 			ipv6_mc_up(idev);
3372 		return;
3373 	}
3374 
3375 	idev = addrconf_add_dev(dev);
3376 	if (IS_ERR(idev))
3377 		return;
3378 
3379 	/* this device type has no EUI support */
3380 	if (dev->type == ARPHRD_NONE &&
3381 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3382 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3383 
3384 	addrconf_addr_gen(idev, false);
3385 }
3386 
3387 #if IS_ENABLED(CONFIG_IPV6_SIT)
3388 static void addrconf_sit_config(struct net_device *dev)
3389 {
3390 	struct inet6_dev *idev;
3391 
3392 	ASSERT_RTNL();
3393 
3394 	/*
3395 	 * Configure the tunnel with one of our IPv4
3396 	 * addresses... we should configure all of
3397 	 * our v4 addrs in the tunnel
3398 	 */
3399 
3400 	idev = ipv6_find_idev(dev);
3401 	if (IS_ERR(idev)) {
3402 		pr_debug("%s: add_dev failed\n", __func__);
3403 		return;
3404 	}
3405 
3406 	if (dev->priv_flags & IFF_ISATAP) {
3407 		addrconf_addr_gen(idev, false);
3408 		return;
3409 	}
3410 
3411 	add_v4_addrs(idev);
3412 
3413 	if (dev->flags&IFF_POINTOPOINT)
3414 		addrconf_add_mroute(dev);
3415 }
3416 #endif
3417 
3418 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3419 static void addrconf_gre_config(struct net_device *dev)
3420 {
3421 	struct inet6_dev *idev;
3422 
3423 	ASSERT_RTNL();
3424 
3425 	idev = ipv6_find_idev(dev);
3426 	if (IS_ERR(idev)) {
3427 		pr_debug("%s: add_dev failed\n", __func__);
3428 		return;
3429 	}
3430 
3431 	if (dev->type == ARPHRD_ETHER) {
3432 		addrconf_addr_gen(idev, true);
3433 		return;
3434 	}
3435 
3436 	add_v4_addrs(idev);
3437 
3438 	if (dev->flags & IFF_POINTOPOINT)
3439 		addrconf_add_mroute(dev);
3440 }
3441 #endif
3442 
3443 static int fixup_permanent_addr(struct net *net,
3444 				struct inet6_dev *idev,
3445 				struct inet6_ifaddr *ifp)
3446 {
3447 	/* !fib6_node means the host route was removed from the
3448 	 * FIB, for example, if 'lo' device is taken down. In that
3449 	 * case regenerate the host route.
3450 	 */
3451 	if (!ifp->rt || !ifp->rt->fib6_node) {
3452 		struct fib6_info *f6i, *prev;
3453 
3454 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3455 					 GFP_ATOMIC);
3456 		if (IS_ERR(f6i))
3457 			return PTR_ERR(f6i);
3458 
3459 		/* ifp->rt can be accessed outside of rtnl */
3460 		spin_lock(&ifp->lock);
3461 		prev = ifp->rt;
3462 		ifp->rt = f6i;
3463 		spin_unlock(&ifp->lock);
3464 
3465 		fib6_info_release(prev);
3466 	}
3467 
3468 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3469 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3470 				      ifp->rt_priority, idev->dev, 0, 0,
3471 				      GFP_ATOMIC);
3472 	}
3473 
3474 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3475 		addrconf_dad_start(ifp);
3476 
3477 	return 0;
3478 }
3479 
3480 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3481 {
3482 	struct inet6_ifaddr *ifp, *tmp;
3483 	struct inet6_dev *idev;
3484 
3485 	idev = __in6_dev_get(dev);
3486 	if (!idev)
3487 		return;
3488 
3489 	write_lock_bh(&idev->lock);
3490 
3491 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3492 		if ((ifp->flags & IFA_F_PERMANENT) &&
3493 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3494 			write_unlock_bh(&idev->lock);
3495 			in6_ifa_hold(ifp);
3496 			ipv6_del_addr(ifp);
3497 			write_lock_bh(&idev->lock);
3498 
3499 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3500 					     idev->dev->name, &ifp->addr);
3501 		}
3502 	}
3503 
3504 	write_unlock_bh(&idev->lock);
3505 }
3506 
3507 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3508 			   void *ptr)
3509 {
3510 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3511 	struct netdev_notifier_change_info *change_info;
3512 	struct netdev_notifier_changeupper_info *info;
3513 	struct inet6_dev *idev = __in6_dev_get(dev);
3514 	struct net *net = dev_net(dev);
3515 	int run_pending = 0;
3516 	int err;
3517 
3518 	switch (event) {
3519 	case NETDEV_REGISTER:
3520 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3521 			idev = ipv6_add_dev(dev);
3522 			if (IS_ERR(idev))
3523 				return notifier_from_errno(PTR_ERR(idev));
3524 		}
3525 		break;
3526 
3527 	case NETDEV_CHANGEMTU:
3528 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3529 		if (dev->mtu < IPV6_MIN_MTU) {
3530 			addrconf_ifdown(dev, dev != net->loopback_dev);
3531 			break;
3532 		}
3533 
3534 		if (idev) {
3535 			rt6_mtu_change(dev, dev->mtu);
3536 			idev->cnf.mtu6 = dev->mtu;
3537 			break;
3538 		}
3539 
3540 		/* allocate new idev */
3541 		idev = ipv6_add_dev(dev);
3542 		if (IS_ERR(idev))
3543 			break;
3544 
3545 		/* device is still not ready */
3546 		if (!(idev->if_flags & IF_READY))
3547 			break;
3548 
3549 		run_pending = 1;
3550 		fallthrough;
3551 	case NETDEV_UP:
3552 	case NETDEV_CHANGE:
3553 		if (dev->flags & IFF_SLAVE)
3554 			break;
3555 
3556 		if (idev && idev->cnf.disable_ipv6)
3557 			break;
3558 
3559 		if (event == NETDEV_UP) {
3560 			/* restore routes for permanent addresses */
3561 			addrconf_permanent_addr(net, dev);
3562 
3563 			if (!addrconf_link_ready(dev)) {
3564 				/* device is not ready yet. */
3565 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3566 					 dev->name);
3567 				break;
3568 			}
3569 
3570 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3571 				idev = ipv6_add_dev(dev);
3572 
3573 			if (!IS_ERR_OR_NULL(idev)) {
3574 				idev->if_flags |= IF_READY;
3575 				run_pending = 1;
3576 			}
3577 		} else if (event == NETDEV_CHANGE) {
3578 			if (!addrconf_link_ready(dev)) {
3579 				/* device is still not ready. */
3580 				rt6_sync_down_dev(dev, event);
3581 				break;
3582 			}
3583 
3584 			if (!IS_ERR_OR_NULL(idev)) {
3585 				if (idev->if_flags & IF_READY) {
3586 					/* device is already configured -
3587 					 * but resend MLD reports, we might
3588 					 * have roamed and need to update
3589 					 * multicast snooping switches
3590 					 */
3591 					ipv6_mc_up(idev);
3592 					change_info = ptr;
3593 					if (change_info->flags_changed & IFF_NOARP)
3594 						addrconf_dad_run(idev, true);
3595 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3596 					break;
3597 				}
3598 				idev->if_flags |= IF_READY;
3599 			}
3600 
3601 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3602 				dev->name);
3603 
3604 			run_pending = 1;
3605 		}
3606 
3607 		switch (dev->type) {
3608 #if IS_ENABLED(CONFIG_IPV6_SIT)
3609 		case ARPHRD_SIT:
3610 			addrconf_sit_config(dev);
3611 			break;
3612 #endif
3613 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3614 		case ARPHRD_IP6GRE:
3615 		case ARPHRD_IPGRE:
3616 			addrconf_gre_config(dev);
3617 			break;
3618 #endif
3619 		case ARPHRD_LOOPBACK:
3620 			init_loopback(dev);
3621 			break;
3622 
3623 		default:
3624 			addrconf_dev_config(dev);
3625 			break;
3626 		}
3627 
3628 		if (!IS_ERR_OR_NULL(idev)) {
3629 			if (run_pending)
3630 				addrconf_dad_run(idev, false);
3631 
3632 			/* Device has an address by now */
3633 			rt6_sync_up(dev, RTNH_F_DEAD);
3634 
3635 			/*
3636 			 * If the MTU changed during the interface down,
3637 			 * when the interface up, the changed MTU must be
3638 			 * reflected in the idev as well as routers.
3639 			 */
3640 			if (idev->cnf.mtu6 != dev->mtu &&
3641 			    dev->mtu >= IPV6_MIN_MTU) {
3642 				rt6_mtu_change(dev, dev->mtu);
3643 				idev->cnf.mtu6 = dev->mtu;
3644 			}
3645 			idev->tstamp = jiffies;
3646 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3647 
3648 			/*
3649 			 * If the changed mtu during down is lower than
3650 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3651 			 */
3652 			if (dev->mtu < IPV6_MIN_MTU)
3653 				addrconf_ifdown(dev, dev != net->loopback_dev);
3654 		}
3655 		break;
3656 
3657 	case NETDEV_DOWN:
3658 	case NETDEV_UNREGISTER:
3659 		/*
3660 		 *	Remove all addresses from this interface.
3661 		 */
3662 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3663 		break;
3664 
3665 	case NETDEV_CHANGENAME:
3666 		if (idev) {
3667 			snmp6_unregister_dev(idev);
3668 			addrconf_sysctl_unregister(idev);
3669 			err = addrconf_sysctl_register(idev);
3670 			if (err)
3671 				return notifier_from_errno(err);
3672 			err = snmp6_register_dev(idev);
3673 			if (err) {
3674 				addrconf_sysctl_unregister(idev);
3675 				return notifier_from_errno(err);
3676 			}
3677 		}
3678 		break;
3679 
3680 	case NETDEV_PRE_TYPE_CHANGE:
3681 	case NETDEV_POST_TYPE_CHANGE:
3682 		if (idev)
3683 			addrconf_type_change(dev, event);
3684 		break;
3685 
3686 	case NETDEV_CHANGEUPPER:
3687 		info = ptr;
3688 
3689 		/* flush all routes if dev is linked to or unlinked from
3690 		 * an L3 master device (e.g., VRF)
3691 		 */
3692 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3693 			addrconf_ifdown(dev, false);
3694 	}
3695 
3696 	return NOTIFY_OK;
3697 }
3698 
3699 /*
3700  *	addrconf module should be notified of a device going up
3701  */
3702 static struct notifier_block ipv6_dev_notf = {
3703 	.notifier_call = addrconf_notify,
3704 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3705 };
3706 
3707 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3708 {
3709 	struct inet6_dev *idev;
3710 	ASSERT_RTNL();
3711 
3712 	idev = __in6_dev_get(dev);
3713 
3714 	if (event == NETDEV_POST_TYPE_CHANGE)
3715 		ipv6_mc_remap(idev);
3716 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3717 		ipv6_mc_unmap(idev);
3718 }
3719 
3720 static bool addr_is_local(const struct in6_addr *addr)
3721 {
3722 	return ipv6_addr_type(addr) &
3723 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3724 }
3725 
3726 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3727 {
3728 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3729 	struct net *net = dev_net(dev);
3730 	struct inet6_dev *idev;
3731 	struct inet6_ifaddr *ifa, *tmp;
3732 	bool keep_addr = false;
3733 	bool was_ready;
3734 	int state, i;
3735 
3736 	ASSERT_RTNL();
3737 
3738 	rt6_disable_ip(dev, event);
3739 
3740 	idev = __in6_dev_get(dev);
3741 	if (!idev)
3742 		return -ENODEV;
3743 
3744 	/*
3745 	 * Step 1: remove reference to ipv6 device from parent device.
3746 	 *	   Do not dev_put!
3747 	 */
3748 	if (unregister) {
3749 		idev->dead = 1;
3750 
3751 		/* protected by rtnl_lock */
3752 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3753 
3754 		/* Step 1.5: remove snmp6 entry */
3755 		snmp6_unregister_dev(idev);
3756 
3757 	}
3758 
3759 	/* combine the user config with event to determine if permanent
3760 	 * addresses are to be removed from address hash table
3761 	 */
3762 	if (!unregister && !idev->cnf.disable_ipv6) {
3763 		/* aggregate the system setting and interface setting */
3764 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3765 
3766 		if (!_keep_addr)
3767 			_keep_addr = idev->cnf.keep_addr_on_down;
3768 
3769 		keep_addr = (_keep_addr > 0);
3770 	}
3771 
3772 	/* Step 2: clear hash table */
3773 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3774 		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3775 
3776 		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3777 restart:
3778 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3779 			if (ifa->idev == idev) {
3780 				addrconf_del_dad_work(ifa);
3781 				/* combined flag + permanent flag decide if
3782 				 * address is retained on a down event
3783 				 */
3784 				if (!keep_addr ||
3785 				    !(ifa->flags & IFA_F_PERMANENT) ||
3786 				    addr_is_local(&ifa->addr)) {
3787 					hlist_del_init_rcu(&ifa->addr_lst);
3788 					goto restart;
3789 				}
3790 			}
3791 		}
3792 		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3793 	}
3794 
3795 	write_lock_bh(&idev->lock);
3796 
3797 	addrconf_del_rs_timer(idev);
3798 
3799 	/* Step 2: clear flags for stateless addrconf, repeated down
3800 	 *         detection
3801 	 */
3802 	was_ready = idev->if_flags & IF_READY;
3803 	if (!unregister)
3804 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3805 
3806 	/* Step 3: clear tempaddr list */
3807 	while (!list_empty(&idev->tempaddr_list)) {
3808 		ifa = list_first_entry(&idev->tempaddr_list,
3809 				       struct inet6_ifaddr, tmp_list);
3810 		list_del(&ifa->tmp_list);
3811 		write_unlock_bh(&idev->lock);
3812 		spin_lock_bh(&ifa->lock);
3813 
3814 		if (ifa->ifpub) {
3815 			in6_ifa_put(ifa->ifpub);
3816 			ifa->ifpub = NULL;
3817 		}
3818 		spin_unlock_bh(&ifa->lock);
3819 		in6_ifa_put(ifa);
3820 		write_lock_bh(&idev->lock);
3821 	}
3822 
3823 	list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3824 		struct fib6_info *rt = NULL;
3825 		bool keep;
3826 
3827 		addrconf_del_dad_work(ifa);
3828 
3829 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3830 			!addr_is_local(&ifa->addr);
3831 
3832 		write_unlock_bh(&idev->lock);
3833 		spin_lock_bh(&ifa->lock);
3834 
3835 		if (keep) {
3836 			/* set state to skip the notifier below */
3837 			state = INET6_IFADDR_STATE_DEAD;
3838 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3839 			if (!(ifa->flags & IFA_F_NODAD))
3840 				ifa->flags |= IFA_F_TENTATIVE;
3841 
3842 			rt = ifa->rt;
3843 			ifa->rt = NULL;
3844 		} else {
3845 			state = ifa->state;
3846 			ifa->state = INET6_IFADDR_STATE_DEAD;
3847 		}
3848 
3849 		spin_unlock_bh(&ifa->lock);
3850 
3851 		if (rt)
3852 			ip6_del_rt(net, rt, false);
3853 
3854 		if (state != INET6_IFADDR_STATE_DEAD) {
3855 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3856 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3857 		} else {
3858 			if (idev->cnf.forwarding)
3859 				addrconf_leave_anycast(ifa);
3860 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3861 		}
3862 
3863 		write_lock_bh(&idev->lock);
3864 		if (!keep) {
3865 			list_del_rcu(&ifa->if_list);
3866 			in6_ifa_put(ifa);
3867 		}
3868 	}
3869 
3870 	write_unlock_bh(&idev->lock);
3871 
3872 	/* Step 5: Discard anycast and multicast list */
3873 	if (unregister) {
3874 		ipv6_ac_destroy_dev(idev);
3875 		ipv6_mc_destroy_dev(idev);
3876 	} else if (was_ready) {
3877 		ipv6_mc_down(idev);
3878 	}
3879 
3880 	idev->tstamp = jiffies;
3881 	idev->ra_mtu = 0;
3882 
3883 	/* Last: Shot the device (if unregistered) */
3884 	if (unregister) {
3885 		addrconf_sysctl_unregister(idev);
3886 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3887 		neigh_ifdown(&nd_tbl, dev);
3888 		in6_dev_put(idev);
3889 	}
3890 	return 0;
3891 }
3892 
3893 static void addrconf_rs_timer(struct timer_list *t)
3894 {
3895 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3896 	struct net_device *dev = idev->dev;
3897 	struct in6_addr lladdr;
3898 
3899 	write_lock(&idev->lock);
3900 	if (idev->dead || !(idev->if_flags & IF_READY))
3901 		goto out;
3902 
3903 	if (!ipv6_accept_ra(idev))
3904 		goto out;
3905 
3906 	/* Announcement received after solicitation was sent */
3907 	if (idev->if_flags & IF_RA_RCVD)
3908 		goto out;
3909 
3910 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3911 		write_unlock(&idev->lock);
3912 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3913 			ndisc_send_rs(dev, &lladdr,
3914 				      &in6addr_linklocal_allrouters);
3915 		else
3916 			goto put;
3917 
3918 		write_lock(&idev->lock);
3919 		idev->rs_interval = rfc3315_s14_backoff_update(
3920 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3921 		/* The wait after the last probe can be shorter */
3922 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3923 					     idev->cnf.rtr_solicits) ?
3924 				      idev->cnf.rtr_solicit_delay :
3925 				      idev->rs_interval);
3926 	} else {
3927 		/*
3928 		 * Note: we do not support deprecated "all on-link"
3929 		 * assumption any longer.
3930 		 */
3931 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3932 	}
3933 
3934 out:
3935 	write_unlock(&idev->lock);
3936 put:
3937 	in6_dev_put(idev);
3938 }
3939 
3940 /*
3941  *	Duplicate Address Detection
3942  */
3943 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3944 {
3945 	unsigned long rand_num;
3946 	struct inet6_dev *idev = ifp->idev;
3947 	u64 nonce;
3948 
3949 	if (ifp->flags & IFA_F_OPTIMISTIC)
3950 		rand_num = 0;
3951 	else
3952 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3953 
3954 	nonce = 0;
3955 	if (idev->cnf.enhanced_dad ||
3956 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3957 		do
3958 			get_random_bytes(&nonce, 6);
3959 		while (nonce == 0);
3960 	}
3961 	ifp->dad_nonce = nonce;
3962 	ifp->dad_probes = idev->cnf.dad_transmits;
3963 	addrconf_mod_dad_work(ifp, rand_num);
3964 }
3965 
3966 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3967 {
3968 	struct inet6_dev *idev = ifp->idev;
3969 	struct net_device *dev = idev->dev;
3970 	bool bump_id, notify = false;
3971 	struct net *net;
3972 
3973 	addrconf_join_solict(dev, &ifp->addr);
3974 
3975 	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3976 
3977 	read_lock_bh(&idev->lock);
3978 	spin_lock(&ifp->lock);
3979 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3980 		goto out;
3981 
3982 	net = dev_net(dev);
3983 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3984 	    (net->ipv6.devconf_all->accept_dad < 1 &&
3985 	     idev->cnf.accept_dad < 1) ||
3986 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3987 	    ifp->flags & IFA_F_NODAD) {
3988 		bool send_na = false;
3989 
3990 		if (ifp->flags & IFA_F_TENTATIVE &&
3991 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3992 			send_na = true;
3993 		bump_id = ifp->flags & IFA_F_TENTATIVE;
3994 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3995 		spin_unlock(&ifp->lock);
3996 		read_unlock_bh(&idev->lock);
3997 
3998 		addrconf_dad_completed(ifp, bump_id, send_na);
3999 		return;
4000 	}
4001 
4002 	if (!(idev->if_flags & IF_READY)) {
4003 		spin_unlock(&ifp->lock);
4004 		read_unlock_bh(&idev->lock);
4005 		/*
4006 		 * If the device is not ready:
4007 		 * - keep it tentative if it is a permanent address.
4008 		 * - otherwise, kill it.
4009 		 */
4010 		in6_ifa_hold(ifp);
4011 		addrconf_dad_stop(ifp, 0);
4012 		return;
4013 	}
4014 
4015 	/*
4016 	 * Optimistic nodes can start receiving
4017 	 * Frames right away
4018 	 */
4019 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4020 		ip6_ins_rt(net, ifp->rt);
4021 		if (ipv6_use_optimistic_addr(net, idev)) {
4022 			/* Because optimistic nodes can use this address,
4023 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4024 			 */
4025 			notify = true;
4026 		}
4027 	}
4028 
4029 	addrconf_dad_kick(ifp);
4030 out:
4031 	spin_unlock(&ifp->lock);
4032 	read_unlock_bh(&idev->lock);
4033 	if (notify)
4034 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4035 }
4036 
4037 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4038 {
4039 	bool begin_dad = false;
4040 
4041 	spin_lock_bh(&ifp->lock);
4042 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4043 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4044 		begin_dad = true;
4045 	}
4046 	spin_unlock_bh(&ifp->lock);
4047 
4048 	if (begin_dad)
4049 		addrconf_mod_dad_work(ifp, 0);
4050 }
4051 
4052 static void addrconf_dad_work(struct work_struct *w)
4053 {
4054 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4055 						struct inet6_ifaddr,
4056 						dad_work);
4057 	struct inet6_dev *idev = ifp->idev;
4058 	bool bump_id, disable_ipv6 = false;
4059 	struct in6_addr mcaddr;
4060 
4061 	enum {
4062 		DAD_PROCESS,
4063 		DAD_BEGIN,
4064 		DAD_ABORT,
4065 	} action = DAD_PROCESS;
4066 
4067 	rtnl_lock();
4068 
4069 	spin_lock_bh(&ifp->lock);
4070 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4071 		action = DAD_BEGIN;
4072 		ifp->state = INET6_IFADDR_STATE_DAD;
4073 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4074 		action = DAD_ABORT;
4075 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4076 
4077 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4078 		     idev->cnf.accept_dad > 1) &&
4079 		    !idev->cnf.disable_ipv6 &&
4080 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4081 			struct in6_addr addr;
4082 
4083 			addr.s6_addr32[0] = htonl(0xfe800000);
4084 			addr.s6_addr32[1] = 0;
4085 
4086 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4087 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4088 				/* DAD failed for link-local based on MAC */
4089 				idev->cnf.disable_ipv6 = 1;
4090 
4091 				pr_info("%s: IPv6 being disabled!\n",
4092 					ifp->idev->dev->name);
4093 				disable_ipv6 = true;
4094 			}
4095 		}
4096 	}
4097 	spin_unlock_bh(&ifp->lock);
4098 
4099 	if (action == DAD_BEGIN) {
4100 		addrconf_dad_begin(ifp);
4101 		goto out;
4102 	} else if (action == DAD_ABORT) {
4103 		in6_ifa_hold(ifp);
4104 		addrconf_dad_stop(ifp, 1);
4105 		if (disable_ipv6)
4106 			addrconf_ifdown(idev->dev, false);
4107 		goto out;
4108 	}
4109 
4110 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4111 		goto out;
4112 
4113 	write_lock_bh(&idev->lock);
4114 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4115 		write_unlock_bh(&idev->lock);
4116 		goto out;
4117 	}
4118 
4119 	spin_lock(&ifp->lock);
4120 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4121 		spin_unlock(&ifp->lock);
4122 		write_unlock_bh(&idev->lock);
4123 		goto out;
4124 	}
4125 
4126 	if (ifp->dad_probes == 0) {
4127 		bool send_na = false;
4128 
4129 		/*
4130 		 * DAD was successful
4131 		 */
4132 
4133 		if (ifp->flags & IFA_F_TENTATIVE &&
4134 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4135 			send_na = true;
4136 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4137 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4138 		spin_unlock(&ifp->lock);
4139 		write_unlock_bh(&idev->lock);
4140 
4141 		addrconf_dad_completed(ifp, bump_id, send_na);
4142 
4143 		goto out;
4144 	}
4145 
4146 	ifp->dad_probes--;
4147 	addrconf_mod_dad_work(ifp,
4148 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4149 				  HZ/100));
4150 	spin_unlock(&ifp->lock);
4151 	write_unlock_bh(&idev->lock);
4152 
4153 	/* send a neighbour solicitation for our addr */
4154 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4155 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4156 		      ifp->dad_nonce);
4157 out:
4158 	in6_ifa_put(ifp);
4159 	rtnl_unlock();
4160 }
4161 
4162 /* ifp->idev must be at least read locked */
4163 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4164 {
4165 	struct inet6_ifaddr *ifpiter;
4166 	struct inet6_dev *idev = ifp->idev;
4167 
4168 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4169 		if (ifpiter->scope > IFA_LINK)
4170 			break;
4171 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4172 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4173 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4174 		    IFA_F_PERMANENT)
4175 			return false;
4176 	}
4177 	return true;
4178 }
4179 
4180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4181 				   bool send_na)
4182 {
4183 	struct net_device *dev = ifp->idev->dev;
4184 	struct in6_addr lladdr;
4185 	bool send_rs, send_mld;
4186 
4187 	addrconf_del_dad_work(ifp);
4188 
4189 	/*
4190 	 *	Configure the address for reception. Now it is valid.
4191 	 */
4192 
4193 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4194 
4195 	/* If added prefix is link local and we are prepared to process
4196 	   router advertisements, start sending router solicitations.
4197 	 */
4198 
4199 	read_lock_bh(&ifp->idev->lock);
4200 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4201 	send_rs = send_mld &&
4202 		  ipv6_accept_ra(ifp->idev) &&
4203 		  ifp->idev->cnf.rtr_solicits != 0 &&
4204 		  (dev->flags&IFF_LOOPBACK) == 0;
4205 	read_unlock_bh(&ifp->idev->lock);
4206 
4207 	/* While dad is in progress mld report's source address is in6_addrany.
4208 	 * Resend with proper ll now.
4209 	 */
4210 	if (send_mld)
4211 		ipv6_mc_dad_complete(ifp->idev);
4212 
4213 	/* send unsolicited NA if enabled */
4214 	if (send_na &&
4215 	    (ifp->idev->cnf.ndisc_notify ||
4216 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4217 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4218 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4219 			      /*solicited=*/ false, /*override=*/ true,
4220 			      /*inc_opt=*/ true);
4221 	}
4222 
4223 	if (send_rs) {
4224 		/*
4225 		 *	If a host as already performed a random delay
4226 		 *	[...] as part of DAD [...] there is no need
4227 		 *	to delay again before sending the first RS
4228 		 */
4229 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4230 			return;
4231 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4232 
4233 		write_lock_bh(&ifp->idev->lock);
4234 		spin_lock(&ifp->lock);
4235 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4236 			ifp->idev->cnf.rtr_solicit_interval);
4237 		ifp->idev->rs_probes = 1;
4238 		ifp->idev->if_flags |= IF_RS_SENT;
4239 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4240 		spin_unlock(&ifp->lock);
4241 		write_unlock_bh(&ifp->idev->lock);
4242 	}
4243 
4244 	if (bump_id)
4245 		rt_genid_bump_ipv6(dev_net(dev));
4246 
4247 	/* Make sure that a new temporary address will be created
4248 	 * before this temporary address becomes deprecated.
4249 	 */
4250 	if (ifp->flags & IFA_F_TEMPORARY)
4251 		addrconf_verify_rtnl(dev_net(dev));
4252 }
4253 
4254 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4255 {
4256 	struct inet6_ifaddr *ifp;
4257 
4258 	read_lock_bh(&idev->lock);
4259 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4260 		spin_lock(&ifp->lock);
4261 		if ((ifp->flags & IFA_F_TENTATIVE &&
4262 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4263 			if (restart)
4264 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4265 			addrconf_dad_kick(ifp);
4266 		}
4267 		spin_unlock(&ifp->lock);
4268 	}
4269 	read_unlock_bh(&idev->lock);
4270 }
4271 
4272 #ifdef CONFIG_PROC_FS
4273 struct if6_iter_state {
4274 	struct seq_net_private p;
4275 	int bucket;
4276 	int offset;
4277 };
4278 
4279 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4280 {
4281 	struct if6_iter_state *state = seq->private;
4282 	struct net *net = seq_file_net(seq);
4283 	struct inet6_ifaddr *ifa = NULL;
4284 	int p = 0;
4285 
4286 	/* initial bucket if pos is 0 */
4287 	if (pos == 0) {
4288 		state->bucket = 0;
4289 		state->offset = 0;
4290 	}
4291 
4292 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4293 		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4294 					 addr_lst) {
4295 			/* sync with offset */
4296 			if (p < state->offset) {
4297 				p++;
4298 				continue;
4299 			}
4300 			return ifa;
4301 		}
4302 
4303 		/* prepare for next bucket */
4304 		state->offset = 0;
4305 		p = 0;
4306 	}
4307 	return NULL;
4308 }
4309 
4310 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4311 					 struct inet6_ifaddr *ifa)
4312 {
4313 	struct if6_iter_state *state = seq->private;
4314 	struct net *net = seq_file_net(seq);
4315 
4316 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4317 		state->offset++;
4318 		return ifa;
4319 	}
4320 
4321 	state->offset = 0;
4322 	while (++state->bucket < IN6_ADDR_HSIZE) {
4323 		hlist_for_each_entry_rcu(ifa,
4324 				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4325 			return ifa;
4326 		}
4327 	}
4328 
4329 	return NULL;
4330 }
4331 
4332 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4333 	__acquires(rcu)
4334 {
4335 	rcu_read_lock();
4336 	return if6_get_first(seq, *pos);
4337 }
4338 
4339 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4340 {
4341 	struct inet6_ifaddr *ifa;
4342 
4343 	ifa = if6_get_next(seq, v);
4344 	++*pos;
4345 	return ifa;
4346 }
4347 
4348 static void if6_seq_stop(struct seq_file *seq, void *v)
4349 	__releases(rcu)
4350 {
4351 	rcu_read_unlock();
4352 }
4353 
4354 static int if6_seq_show(struct seq_file *seq, void *v)
4355 {
4356 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4357 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4358 		   &ifp->addr,
4359 		   ifp->idev->dev->ifindex,
4360 		   ifp->prefix_len,
4361 		   ifp->scope,
4362 		   (u8) ifp->flags,
4363 		   ifp->idev->dev->name);
4364 	return 0;
4365 }
4366 
4367 static const struct seq_operations if6_seq_ops = {
4368 	.start	= if6_seq_start,
4369 	.next	= if6_seq_next,
4370 	.show	= if6_seq_show,
4371 	.stop	= if6_seq_stop,
4372 };
4373 
4374 static int __net_init if6_proc_net_init(struct net *net)
4375 {
4376 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4377 			sizeof(struct if6_iter_state)))
4378 		return -ENOMEM;
4379 	return 0;
4380 }
4381 
4382 static void __net_exit if6_proc_net_exit(struct net *net)
4383 {
4384 	remove_proc_entry("if_inet6", net->proc_net);
4385 }
4386 
4387 static struct pernet_operations if6_proc_net_ops = {
4388 	.init = if6_proc_net_init,
4389 	.exit = if6_proc_net_exit,
4390 };
4391 
4392 int __init if6_proc_init(void)
4393 {
4394 	return register_pernet_subsys(&if6_proc_net_ops);
4395 }
4396 
4397 void if6_proc_exit(void)
4398 {
4399 	unregister_pernet_subsys(&if6_proc_net_ops);
4400 }
4401 #endif	/* CONFIG_PROC_FS */
4402 
4403 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4404 /* Check if address is a home address configured on any interface. */
4405 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4406 {
4407 	unsigned int hash = inet6_addr_hash(net, addr);
4408 	struct inet6_ifaddr *ifp = NULL;
4409 	int ret = 0;
4410 
4411 	rcu_read_lock();
4412 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4413 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4414 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4415 			ret = 1;
4416 			break;
4417 		}
4418 	}
4419 	rcu_read_unlock();
4420 	return ret;
4421 }
4422 #endif
4423 
4424 /* RFC6554 has some algorithm to avoid loops in segment routing by
4425  * checking if the segments contains any of a local interface address.
4426  *
4427  * Quote:
4428  *
4429  * To detect loops in the SRH, a router MUST determine if the SRH
4430  * includes multiple addresses assigned to any interface on that router.
4431  * If such addresses appear more than once and are separated by at least
4432  * one address not assigned to that router.
4433  */
4434 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4435 			  unsigned char nsegs)
4436 {
4437 	const struct in6_addr *addr;
4438 	int i, ret = 0, found = 0;
4439 	struct inet6_ifaddr *ifp;
4440 	bool separated = false;
4441 	unsigned int hash;
4442 	bool hash_found;
4443 
4444 	rcu_read_lock();
4445 	for (i = 0; i < nsegs; i++) {
4446 		addr = &segs[i];
4447 		hash = inet6_addr_hash(net, addr);
4448 
4449 		hash_found = false;
4450 		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4451 
4452 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4453 				hash_found = true;
4454 				break;
4455 			}
4456 		}
4457 
4458 		if (hash_found) {
4459 			if (found > 1 && separated) {
4460 				ret = 1;
4461 				break;
4462 			}
4463 
4464 			separated = false;
4465 			found++;
4466 		} else {
4467 			separated = true;
4468 		}
4469 	}
4470 	rcu_read_unlock();
4471 
4472 	return ret;
4473 }
4474 
4475 /*
4476  *	Periodic address status verification
4477  */
4478 
4479 static void addrconf_verify_rtnl(struct net *net)
4480 {
4481 	unsigned long now, next, next_sec, next_sched;
4482 	struct inet6_ifaddr *ifp;
4483 	int i;
4484 
4485 	ASSERT_RTNL();
4486 
4487 	rcu_read_lock_bh();
4488 	now = jiffies;
4489 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4490 
4491 	cancel_delayed_work(&net->ipv6.addr_chk_work);
4492 
4493 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4494 restart:
4495 		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4496 			unsigned long age;
4497 
4498 			/* When setting preferred_lft to a value not zero or
4499 			 * infinity, while valid_lft is infinity
4500 			 * IFA_F_PERMANENT has a non-infinity life time.
4501 			 */
4502 			if ((ifp->flags & IFA_F_PERMANENT) &&
4503 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4504 				continue;
4505 
4506 			spin_lock(&ifp->lock);
4507 			/* We try to batch several events at once. */
4508 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4509 
4510 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4511 			    age >= ifp->valid_lft) {
4512 				spin_unlock(&ifp->lock);
4513 				in6_ifa_hold(ifp);
4514 				rcu_read_unlock_bh();
4515 				ipv6_del_addr(ifp);
4516 				rcu_read_lock_bh();
4517 				goto restart;
4518 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4519 				spin_unlock(&ifp->lock);
4520 				continue;
4521 			} else if (age >= ifp->prefered_lft) {
4522 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4523 				int deprecate = 0;
4524 
4525 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4526 					deprecate = 1;
4527 					ifp->flags |= IFA_F_DEPRECATED;
4528 				}
4529 
4530 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4531 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4532 					next = ifp->tstamp + ifp->valid_lft * HZ;
4533 
4534 				spin_unlock(&ifp->lock);
4535 
4536 				if (deprecate) {
4537 					in6_ifa_hold(ifp);
4538 
4539 					ipv6_ifa_notify(0, ifp);
4540 					in6_ifa_put(ifp);
4541 					goto restart;
4542 				}
4543 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4544 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4545 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4546 					ifp->idev->cnf.dad_transmits *
4547 					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4548 
4549 				if (age >= ifp->prefered_lft - regen_advance) {
4550 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4551 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4552 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4553 					if (!ifp->regen_count && ifpub) {
4554 						ifp->regen_count++;
4555 						in6_ifa_hold(ifp);
4556 						in6_ifa_hold(ifpub);
4557 						spin_unlock(&ifp->lock);
4558 
4559 						spin_lock(&ifpub->lock);
4560 						ifpub->regen_count = 0;
4561 						spin_unlock(&ifpub->lock);
4562 						rcu_read_unlock_bh();
4563 						ipv6_create_tempaddr(ifpub, true);
4564 						in6_ifa_put(ifpub);
4565 						in6_ifa_put(ifp);
4566 						rcu_read_lock_bh();
4567 						goto restart;
4568 					}
4569 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4570 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4571 				spin_unlock(&ifp->lock);
4572 			} else {
4573 				/* ifp->prefered_lft <= ifp->valid_lft */
4574 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4575 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4576 				spin_unlock(&ifp->lock);
4577 			}
4578 		}
4579 	}
4580 
4581 	next_sec = round_jiffies_up(next);
4582 	next_sched = next;
4583 
4584 	/* If rounded timeout is accurate enough, accept it. */
4585 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4586 		next_sched = next_sec;
4587 
4588 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4589 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4590 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4591 
4592 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4593 		 now, next, next_sec, next_sched);
4594 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4595 	rcu_read_unlock_bh();
4596 }
4597 
4598 static void addrconf_verify_work(struct work_struct *w)
4599 {
4600 	struct net *net = container_of(to_delayed_work(w), struct net,
4601 				       ipv6.addr_chk_work);
4602 
4603 	rtnl_lock();
4604 	addrconf_verify_rtnl(net);
4605 	rtnl_unlock();
4606 }
4607 
4608 static void addrconf_verify(struct net *net)
4609 {
4610 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4611 }
4612 
4613 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4614 				     struct in6_addr **peer_pfx)
4615 {
4616 	struct in6_addr *pfx = NULL;
4617 
4618 	*peer_pfx = NULL;
4619 
4620 	if (addr)
4621 		pfx = nla_data(addr);
4622 
4623 	if (local) {
4624 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4625 			*peer_pfx = pfx;
4626 		pfx = nla_data(local);
4627 	}
4628 
4629 	return pfx;
4630 }
4631 
4632 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4633 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4634 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4635 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4636 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4637 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4638 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4639 	[IFA_PROTO]		= { .type = NLA_U8 },
4640 };
4641 
4642 static int
4643 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4644 		  struct netlink_ext_ack *extack)
4645 {
4646 	struct net *net = sock_net(skb->sk);
4647 	struct ifaddrmsg *ifm;
4648 	struct nlattr *tb[IFA_MAX+1];
4649 	struct in6_addr *pfx, *peer_pfx;
4650 	u32 ifa_flags;
4651 	int err;
4652 
4653 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4654 				     ifa_ipv6_policy, extack);
4655 	if (err < 0)
4656 		return err;
4657 
4658 	ifm = nlmsg_data(nlh);
4659 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4660 	if (!pfx)
4661 		return -EINVAL;
4662 
4663 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4664 
4665 	/* We ignore other flags so far. */
4666 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4667 
4668 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4669 			      ifm->ifa_prefixlen);
4670 }
4671 
4672 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4673 			       unsigned long expires, u32 flags,
4674 			       bool modify_peer)
4675 {
4676 	struct fib6_info *f6i;
4677 	u32 prio;
4678 
4679 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4680 					ifp->prefix_len,
4681 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4682 	if (!f6i)
4683 		return -ENOENT;
4684 
4685 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4686 	if (f6i->fib6_metric != prio) {
4687 		/* delete old one */
4688 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4689 
4690 		/* add new one */
4691 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4692 				      ifp->prefix_len,
4693 				      ifp->rt_priority, ifp->idev->dev,
4694 				      expires, flags, GFP_KERNEL);
4695 	} else {
4696 		if (!expires)
4697 			fib6_clean_expires(f6i);
4698 		else
4699 			fib6_set_expires(f6i, expires);
4700 
4701 		fib6_info_release(f6i);
4702 	}
4703 
4704 	return 0;
4705 }
4706 
4707 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4708 			     struct ifa6_config *cfg)
4709 {
4710 	u32 flags;
4711 	clock_t expires;
4712 	unsigned long timeout;
4713 	bool was_managetempaddr;
4714 	bool had_prefixroute;
4715 	bool new_peer = false;
4716 
4717 	ASSERT_RTNL();
4718 
4719 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4720 		return -EINVAL;
4721 
4722 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4723 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4724 		return -EINVAL;
4725 
4726 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4727 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4728 
4729 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4730 	if (addrconf_finite_timeout(timeout)) {
4731 		expires = jiffies_to_clock_t(timeout * HZ);
4732 		cfg->valid_lft = timeout;
4733 		flags = RTF_EXPIRES;
4734 	} else {
4735 		expires = 0;
4736 		flags = 0;
4737 		cfg->ifa_flags |= IFA_F_PERMANENT;
4738 	}
4739 
4740 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4741 	if (addrconf_finite_timeout(timeout)) {
4742 		if (timeout == 0)
4743 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4744 		cfg->preferred_lft = timeout;
4745 	}
4746 
4747 	if (cfg->peer_pfx &&
4748 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4749 		if (!ipv6_addr_any(&ifp->peer_addr))
4750 			cleanup_prefix_route(ifp, expires, true, true);
4751 		new_peer = true;
4752 	}
4753 
4754 	spin_lock_bh(&ifp->lock);
4755 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4756 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4757 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4758 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4759 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4760 			IFA_F_NOPREFIXROUTE);
4761 	ifp->flags |= cfg->ifa_flags;
4762 	ifp->tstamp = jiffies;
4763 	ifp->valid_lft = cfg->valid_lft;
4764 	ifp->prefered_lft = cfg->preferred_lft;
4765 	ifp->ifa_proto = cfg->ifa_proto;
4766 
4767 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4768 		ifp->rt_priority = cfg->rt_priority;
4769 
4770 	if (new_peer)
4771 		ifp->peer_addr = *cfg->peer_pfx;
4772 
4773 	spin_unlock_bh(&ifp->lock);
4774 	if (!(ifp->flags&IFA_F_TENTATIVE))
4775 		ipv6_ifa_notify(0, ifp);
4776 
4777 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4778 		int rc = -ENOENT;
4779 
4780 		if (had_prefixroute)
4781 			rc = modify_prefix_route(ifp, expires, flags, false);
4782 
4783 		/* prefix route could have been deleted; if so restore it */
4784 		if (rc == -ENOENT) {
4785 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4786 					      ifp->rt_priority, ifp->idev->dev,
4787 					      expires, flags, GFP_KERNEL);
4788 		}
4789 
4790 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4791 			rc = modify_prefix_route(ifp, expires, flags, true);
4792 
4793 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4794 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4795 					      ifp->rt_priority, ifp->idev->dev,
4796 					      expires, flags, GFP_KERNEL);
4797 		}
4798 	} else if (had_prefixroute) {
4799 		enum cleanup_prefix_rt_t action;
4800 		unsigned long rt_expires;
4801 
4802 		write_lock_bh(&ifp->idev->lock);
4803 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4804 		write_unlock_bh(&ifp->idev->lock);
4805 
4806 		if (action != CLEANUP_PREFIX_RT_NOP) {
4807 			cleanup_prefix_route(ifp, rt_expires,
4808 				action == CLEANUP_PREFIX_RT_DEL, false);
4809 		}
4810 	}
4811 
4812 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4813 		if (was_managetempaddr &&
4814 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4815 			cfg->valid_lft = 0;
4816 			cfg->preferred_lft = 0;
4817 		}
4818 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4819 				 cfg->preferred_lft, !was_managetempaddr,
4820 				 jiffies);
4821 	}
4822 
4823 	addrconf_verify_rtnl(net);
4824 
4825 	return 0;
4826 }
4827 
4828 static int
4829 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4830 		  struct netlink_ext_ack *extack)
4831 {
4832 	struct net *net = sock_net(skb->sk);
4833 	struct ifaddrmsg *ifm;
4834 	struct nlattr *tb[IFA_MAX+1];
4835 	struct in6_addr *peer_pfx;
4836 	struct inet6_ifaddr *ifa;
4837 	struct net_device *dev;
4838 	struct inet6_dev *idev;
4839 	struct ifa6_config cfg;
4840 	int err;
4841 
4842 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4843 				     ifa_ipv6_policy, extack);
4844 	if (err < 0)
4845 		return err;
4846 
4847 	memset(&cfg, 0, sizeof(cfg));
4848 
4849 	ifm = nlmsg_data(nlh);
4850 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4851 	if (!cfg.pfx)
4852 		return -EINVAL;
4853 
4854 	cfg.peer_pfx = peer_pfx;
4855 	cfg.plen = ifm->ifa_prefixlen;
4856 	if (tb[IFA_RT_PRIORITY])
4857 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4858 
4859 	if (tb[IFA_PROTO])
4860 		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4861 
4862 	cfg.valid_lft = INFINITY_LIFE_TIME;
4863 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4864 
4865 	if (tb[IFA_CACHEINFO]) {
4866 		struct ifa_cacheinfo *ci;
4867 
4868 		ci = nla_data(tb[IFA_CACHEINFO]);
4869 		cfg.valid_lft = ci->ifa_valid;
4870 		cfg.preferred_lft = ci->ifa_prefered;
4871 	}
4872 
4873 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4874 	if (!dev)
4875 		return -ENODEV;
4876 
4877 	if (tb[IFA_FLAGS])
4878 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4879 	else
4880 		cfg.ifa_flags = ifm->ifa_flags;
4881 
4882 	/* We ignore other flags so far. */
4883 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4884 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4885 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4886 
4887 	idev = ipv6_find_idev(dev);
4888 	if (IS_ERR(idev))
4889 		return PTR_ERR(idev);
4890 
4891 	if (!ipv6_allow_optimistic_dad(net, idev))
4892 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4893 
4894 	if (cfg.ifa_flags & IFA_F_NODAD &&
4895 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4896 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4897 		return -EINVAL;
4898 	}
4899 
4900 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4901 	if (!ifa) {
4902 		/*
4903 		 * It would be best to check for !NLM_F_CREATE here but
4904 		 * userspace already relies on not having to provide this.
4905 		 */
4906 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4907 	}
4908 
4909 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4910 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4911 		err = -EEXIST;
4912 	else
4913 		err = inet6_addr_modify(net, ifa, &cfg);
4914 
4915 	in6_ifa_put(ifa);
4916 
4917 	return err;
4918 }
4919 
4920 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4921 			  u8 scope, int ifindex)
4922 {
4923 	struct ifaddrmsg *ifm;
4924 
4925 	ifm = nlmsg_data(nlh);
4926 	ifm->ifa_family = AF_INET6;
4927 	ifm->ifa_prefixlen = prefixlen;
4928 	ifm->ifa_flags = flags;
4929 	ifm->ifa_scope = scope;
4930 	ifm->ifa_index = ifindex;
4931 }
4932 
4933 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4934 			 unsigned long tstamp, u32 preferred, u32 valid)
4935 {
4936 	struct ifa_cacheinfo ci;
4937 
4938 	ci.cstamp = cstamp_delta(cstamp);
4939 	ci.tstamp = cstamp_delta(tstamp);
4940 	ci.ifa_prefered = preferred;
4941 	ci.ifa_valid = valid;
4942 
4943 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4944 }
4945 
4946 static inline int rt_scope(int ifa_scope)
4947 {
4948 	if (ifa_scope & IFA_HOST)
4949 		return RT_SCOPE_HOST;
4950 	else if (ifa_scope & IFA_LINK)
4951 		return RT_SCOPE_LINK;
4952 	else if (ifa_scope & IFA_SITE)
4953 		return RT_SCOPE_SITE;
4954 	else
4955 		return RT_SCOPE_UNIVERSE;
4956 }
4957 
4958 static inline int inet6_ifaddr_msgsize(void)
4959 {
4960 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4961 	       + nla_total_size(16) /* IFA_LOCAL */
4962 	       + nla_total_size(16) /* IFA_ADDRESS */
4963 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4964 	       + nla_total_size(4)  /* IFA_FLAGS */
4965 	       + nla_total_size(1)  /* IFA_PROTO */
4966 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4967 }
4968 
4969 enum addr_type_t {
4970 	UNICAST_ADDR,
4971 	MULTICAST_ADDR,
4972 	ANYCAST_ADDR,
4973 };
4974 
4975 struct inet6_fill_args {
4976 	u32 portid;
4977 	u32 seq;
4978 	int event;
4979 	unsigned int flags;
4980 	int netnsid;
4981 	int ifindex;
4982 	enum addr_type_t type;
4983 };
4984 
4985 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4986 			     struct inet6_fill_args *args)
4987 {
4988 	struct nlmsghdr  *nlh;
4989 	u32 preferred, valid;
4990 
4991 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4992 			sizeof(struct ifaddrmsg), args->flags);
4993 	if (!nlh)
4994 		return -EMSGSIZE;
4995 
4996 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4997 		      ifa->idev->dev->ifindex);
4998 
4999 	if (args->netnsid >= 0 &&
5000 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5001 		goto error;
5002 
5003 	spin_lock_bh(&ifa->lock);
5004 	if (!((ifa->flags&IFA_F_PERMANENT) &&
5005 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5006 		preferred = ifa->prefered_lft;
5007 		valid = ifa->valid_lft;
5008 		if (preferred != INFINITY_LIFE_TIME) {
5009 			long tval = (jiffies - ifa->tstamp)/HZ;
5010 			if (preferred > tval)
5011 				preferred -= tval;
5012 			else
5013 				preferred = 0;
5014 			if (valid != INFINITY_LIFE_TIME) {
5015 				if (valid > tval)
5016 					valid -= tval;
5017 				else
5018 					valid = 0;
5019 			}
5020 		}
5021 	} else {
5022 		preferred = INFINITY_LIFE_TIME;
5023 		valid = INFINITY_LIFE_TIME;
5024 	}
5025 	spin_unlock_bh(&ifa->lock);
5026 
5027 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5028 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5029 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5030 			goto error;
5031 	} else
5032 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5033 			goto error;
5034 
5035 	if (ifa->rt_priority &&
5036 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5037 		goto error;
5038 
5039 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5040 		goto error;
5041 
5042 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5043 		goto error;
5044 
5045 	if (ifa->ifa_proto &&
5046 	    nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
5047 		goto error;
5048 
5049 	nlmsg_end(skb, nlh);
5050 	return 0;
5051 
5052 error:
5053 	nlmsg_cancel(skb, nlh);
5054 	return -EMSGSIZE;
5055 }
5056 
5057 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5058 			       struct inet6_fill_args *args)
5059 {
5060 	struct nlmsghdr  *nlh;
5061 	u8 scope = RT_SCOPE_UNIVERSE;
5062 	int ifindex = ifmca->idev->dev->ifindex;
5063 
5064 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5065 		scope = RT_SCOPE_SITE;
5066 
5067 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5068 			sizeof(struct ifaddrmsg), args->flags);
5069 	if (!nlh)
5070 		return -EMSGSIZE;
5071 
5072 	if (args->netnsid >= 0 &&
5073 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5074 		nlmsg_cancel(skb, nlh);
5075 		return -EMSGSIZE;
5076 	}
5077 
5078 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5079 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5080 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5081 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5082 		nlmsg_cancel(skb, nlh);
5083 		return -EMSGSIZE;
5084 	}
5085 
5086 	nlmsg_end(skb, nlh);
5087 	return 0;
5088 }
5089 
5090 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5091 			       struct inet6_fill_args *args)
5092 {
5093 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5094 	int ifindex = dev ? dev->ifindex : 1;
5095 	struct nlmsghdr  *nlh;
5096 	u8 scope = RT_SCOPE_UNIVERSE;
5097 
5098 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5099 		scope = RT_SCOPE_SITE;
5100 
5101 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5102 			sizeof(struct ifaddrmsg), args->flags);
5103 	if (!nlh)
5104 		return -EMSGSIZE;
5105 
5106 	if (args->netnsid >= 0 &&
5107 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5108 		nlmsg_cancel(skb, nlh);
5109 		return -EMSGSIZE;
5110 	}
5111 
5112 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5113 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5114 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5115 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5116 		nlmsg_cancel(skb, nlh);
5117 		return -EMSGSIZE;
5118 	}
5119 
5120 	nlmsg_end(skb, nlh);
5121 	return 0;
5122 }
5123 
5124 /* called with rcu_read_lock() */
5125 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5126 			  struct netlink_callback *cb, int s_ip_idx,
5127 			  struct inet6_fill_args *fillargs)
5128 {
5129 	struct ifmcaddr6 *ifmca;
5130 	struct ifacaddr6 *ifaca;
5131 	int ip_idx = 0;
5132 	int err = 1;
5133 
5134 	read_lock_bh(&idev->lock);
5135 	switch (fillargs->type) {
5136 	case UNICAST_ADDR: {
5137 		struct inet6_ifaddr *ifa;
5138 		fillargs->event = RTM_NEWADDR;
5139 
5140 		/* unicast address incl. temp addr */
5141 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
5142 			if (ip_idx < s_ip_idx)
5143 				goto next;
5144 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5145 			if (err < 0)
5146 				break;
5147 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5148 next:
5149 			ip_idx++;
5150 		}
5151 		break;
5152 	}
5153 	case MULTICAST_ADDR:
5154 		read_unlock_bh(&idev->lock);
5155 		fillargs->event = RTM_GETMULTICAST;
5156 
5157 		/* multicast address */
5158 		for (ifmca = rcu_dereference(idev->mc_list);
5159 		     ifmca;
5160 		     ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5161 			if (ip_idx < s_ip_idx)
5162 				continue;
5163 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5164 			if (err < 0)
5165 				break;
5166 		}
5167 		read_lock_bh(&idev->lock);
5168 		break;
5169 	case ANYCAST_ADDR:
5170 		fillargs->event = RTM_GETANYCAST;
5171 		/* anycast address */
5172 		for (ifaca = idev->ac_list; ifaca;
5173 		     ifaca = ifaca->aca_next, ip_idx++) {
5174 			if (ip_idx < s_ip_idx)
5175 				continue;
5176 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5177 			if (err < 0)
5178 				break;
5179 		}
5180 		break;
5181 	default:
5182 		break;
5183 	}
5184 	read_unlock_bh(&idev->lock);
5185 	cb->args[2] = ip_idx;
5186 	return err;
5187 }
5188 
5189 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5190 				       struct inet6_fill_args *fillargs,
5191 				       struct net **tgt_net, struct sock *sk,
5192 				       struct netlink_callback *cb)
5193 {
5194 	struct netlink_ext_ack *extack = cb->extack;
5195 	struct nlattr *tb[IFA_MAX+1];
5196 	struct ifaddrmsg *ifm;
5197 	int err, i;
5198 
5199 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5200 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5201 		return -EINVAL;
5202 	}
5203 
5204 	ifm = nlmsg_data(nlh);
5205 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5206 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5207 		return -EINVAL;
5208 	}
5209 
5210 	fillargs->ifindex = ifm->ifa_index;
5211 	if (fillargs->ifindex) {
5212 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5213 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5214 	}
5215 
5216 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5217 					    ifa_ipv6_policy, extack);
5218 	if (err < 0)
5219 		return err;
5220 
5221 	for (i = 0; i <= IFA_MAX; ++i) {
5222 		if (!tb[i])
5223 			continue;
5224 
5225 		if (i == IFA_TARGET_NETNSID) {
5226 			struct net *net;
5227 
5228 			fillargs->netnsid = nla_get_s32(tb[i]);
5229 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5230 			if (IS_ERR(net)) {
5231 				fillargs->netnsid = -1;
5232 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5233 				return PTR_ERR(net);
5234 			}
5235 			*tgt_net = net;
5236 		} else {
5237 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5238 			return -EINVAL;
5239 		}
5240 	}
5241 
5242 	return 0;
5243 }
5244 
5245 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5246 			   enum addr_type_t type)
5247 {
5248 	const struct nlmsghdr *nlh = cb->nlh;
5249 	struct inet6_fill_args fillargs = {
5250 		.portid = NETLINK_CB(cb->skb).portid,
5251 		.seq = cb->nlh->nlmsg_seq,
5252 		.flags = NLM_F_MULTI,
5253 		.netnsid = -1,
5254 		.type = type,
5255 	};
5256 	struct net *tgt_net = sock_net(skb->sk);
5257 	int idx, s_idx, s_ip_idx;
5258 	int h, s_h;
5259 	struct net_device *dev;
5260 	struct inet6_dev *idev;
5261 	struct hlist_head *head;
5262 	int err = 0;
5263 
5264 	s_h = cb->args[0];
5265 	s_idx = idx = cb->args[1];
5266 	s_ip_idx = cb->args[2];
5267 
5268 	if (cb->strict_check) {
5269 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5270 						  skb->sk, cb);
5271 		if (err < 0)
5272 			goto put_tgt_net;
5273 
5274 		err = 0;
5275 		if (fillargs.ifindex) {
5276 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5277 			if (!dev) {
5278 				err = -ENODEV;
5279 				goto put_tgt_net;
5280 			}
5281 			idev = __in6_dev_get(dev);
5282 			if (idev) {
5283 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5284 						     &fillargs);
5285 				if (err > 0)
5286 					err = 0;
5287 			}
5288 			goto put_tgt_net;
5289 		}
5290 	}
5291 
5292 	rcu_read_lock();
5293 	cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5294 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5295 		idx = 0;
5296 		head = &tgt_net->dev_index_head[h];
5297 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5298 			if (idx < s_idx)
5299 				goto cont;
5300 			if (h > s_h || idx > s_idx)
5301 				s_ip_idx = 0;
5302 			idev = __in6_dev_get(dev);
5303 			if (!idev)
5304 				goto cont;
5305 
5306 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5307 					   &fillargs) < 0)
5308 				goto done;
5309 cont:
5310 			idx++;
5311 		}
5312 	}
5313 done:
5314 	rcu_read_unlock();
5315 	cb->args[0] = h;
5316 	cb->args[1] = idx;
5317 put_tgt_net:
5318 	if (fillargs.netnsid >= 0)
5319 		put_net(tgt_net);
5320 
5321 	return skb->len ? : err;
5322 }
5323 
5324 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5325 {
5326 	enum addr_type_t type = UNICAST_ADDR;
5327 
5328 	return inet6_dump_addr(skb, cb, type);
5329 }
5330 
5331 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5332 {
5333 	enum addr_type_t type = MULTICAST_ADDR;
5334 
5335 	return inet6_dump_addr(skb, cb, type);
5336 }
5337 
5338 
5339 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5340 {
5341 	enum addr_type_t type = ANYCAST_ADDR;
5342 
5343 	return inet6_dump_addr(skb, cb, type);
5344 }
5345 
5346 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5347 				       const struct nlmsghdr *nlh,
5348 				       struct nlattr **tb,
5349 				       struct netlink_ext_ack *extack)
5350 {
5351 	struct ifaddrmsg *ifm;
5352 	int i, err;
5353 
5354 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5355 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5356 		return -EINVAL;
5357 	}
5358 
5359 	if (!netlink_strict_get_check(skb))
5360 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5361 					      ifa_ipv6_policy, extack);
5362 
5363 	ifm = nlmsg_data(nlh);
5364 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5365 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5366 		return -EINVAL;
5367 	}
5368 
5369 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5370 					    ifa_ipv6_policy, extack);
5371 	if (err)
5372 		return err;
5373 
5374 	for (i = 0; i <= IFA_MAX; i++) {
5375 		if (!tb[i])
5376 			continue;
5377 
5378 		switch (i) {
5379 		case IFA_TARGET_NETNSID:
5380 		case IFA_ADDRESS:
5381 		case IFA_LOCAL:
5382 			break;
5383 		default:
5384 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5385 			return -EINVAL;
5386 		}
5387 	}
5388 
5389 	return 0;
5390 }
5391 
5392 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5393 			     struct netlink_ext_ack *extack)
5394 {
5395 	struct net *tgt_net = sock_net(in_skb->sk);
5396 	struct inet6_fill_args fillargs = {
5397 		.portid = NETLINK_CB(in_skb).portid,
5398 		.seq = nlh->nlmsg_seq,
5399 		.event = RTM_NEWADDR,
5400 		.flags = 0,
5401 		.netnsid = -1,
5402 	};
5403 	struct ifaddrmsg *ifm;
5404 	struct nlattr *tb[IFA_MAX+1];
5405 	struct in6_addr *addr = NULL, *peer;
5406 	struct net_device *dev = NULL;
5407 	struct inet6_ifaddr *ifa;
5408 	struct sk_buff *skb;
5409 	int err;
5410 
5411 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5412 	if (err < 0)
5413 		return err;
5414 
5415 	if (tb[IFA_TARGET_NETNSID]) {
5416 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5417 
5418 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5419 						  fillargs.netnsid);
5420 		if (IS_ERR(tgt_net))
5421 			return PTR_ERR(tgt_net);
5422 	}
5423 
5424 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5425 	if (!addr)
5426 		return -EINVAL;
5427 
5428 	ifm = nlmsg_data(nlh);
5429 	if (ifm->ifa_index)
5430 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5431 
5432 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5433 	if (!ifa) {
5434 		err = -EADDRNOTAVAIL;
5435 		goto errout;
5436 	}
5437 
5438 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5439 	if (!skb) {
5440 		err = -ENOBUFS;
5441 		goto errout_ifa;
5442 	}
5443 
5444 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5445 	if (err < 0) {
5446 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5447 		WARN_ON(err == -EMSGSIZE);
5448 		kfree_skb(skb);
5449 		goto errout_ifa;
5450 	}
5451 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5452 errout_ifa:
5453 	in6_ifa_put(ifa);
5454 errout:
5455 	dev_put(dev);
5456 	if (fillargs.netnsid >= 0)
5457 		put_net(tgt_net);
5458 
5459 	return err;
5460 }
5461 
5462 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5463 {
5464 	struct sk_buff *skb;
5465 	struct net *net = dev_net(ifa->idev->dev);
5466 	struct inet6_fill_args fillargs = {
5467 		.portid = 0,
5468 		.seq = 0,
5469 		.event = event,
5470 		.flags = 0,
5471 		.netnsid = -1,
5472 	};
5473 	int err = -ENOBUFS;
5474 
5475 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5476 	if (!skb)
5477 		goto errout;
5478 
5479 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5480 	if (err < 0) {
5481 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5482 		WARN_ON(err == -EMSGSIZE);
5483 		kfree_skb(skb);
5484 		goto errout;
5485 	}
5486 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5487 	return;
5488 errout:
5489 	if (err < 0)
5490 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5491 }
5492 
5493 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5494 				__s32 *array, int bytes)
5495 {
5496 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5497 
5498 	memset(array, 0, bytes);
5499 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5500 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5501 	array[DEVCONF_MTU6] = cnf->mtu6;
5502 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5503 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5504 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5505 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5506 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5507 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5508 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5509 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5510 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5511 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5512 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5513 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5514 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5515 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5516 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5517 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5518 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5519 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5520 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5521 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5522 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5523 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5524 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5525 	array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5526 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5527 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5528 #ifdef CONFIG_IPV6_ROUTER_PREF
5529 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5530 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5531 		jiffies_to_msecs(cnf->rtr_probe_interval);
5532 #ifdef CONFIG_IPV6_ROUTE_INFO
5533 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5534 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5535 #endif
5536 #endif
5537 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5538 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5539 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5540 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5541 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5542 #endif
5543 #ifdef CONFIG_IPV6_MROUTE
5544 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5545 #endif
5546 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5547 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5548 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5549 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5550 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5551 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5552 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5553 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5554 	/* we omit DEVCONF_STABLE_SECRET for now */
5555 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5556 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5557 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5558 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5559 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5560 #ifdef CONFIG_IPV6_SEG6_HMAC
5561 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5562 #endif
5563 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5564 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5565 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5566 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5567 	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5568 	array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5569 	array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5570 	array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5571 	array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5572 }
5573 
5574 static inline size_t inet6_ifla6_size(void)
5575 {
5576 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5577 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5578 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5579 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5580 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5581 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5582 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5583 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5584 	     + 0;
5585 }
5586 
5587 static inline size_t inet6_if_nlmsg_size(void)
5588 {
5589 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5590 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5591 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5592 	       + nla_total_size(4) /* IFLA_MTU */
5593 	       + nla_total_size(4) /* IFLA_LINK */
5594 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5595 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5596 }
5597 
5598 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5599 					int bytes)
5600 {
5601 	int i;
5602 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5603 	BUG_ON(pad < 0);
5604 
5605 	/* Use put_unaligned() because stats may not be aligned for u64. */
5606 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5607 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5608 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5609 
5610 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5611 }
5612 
5613 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5614 					int bytes, size_t syncpoff)
5615 {
5616 	int i, c;
5617 	u64 buff[IPSTATS_MIB_MAX];
5618 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5619 
5620 	BUG_ON(pad < 0);
5621 
5622 	memset(buff, 0, sizeof(buff));
5623 	buff[0] = IPSTATS_MIB_MAX;
5624 
5625 	for_each_possible_cpu(c) {
5626 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5627 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5628 	}
5629 
5630 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5631 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5632 }
5633 
5634 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5635 			     int bytes)
5636 {
5637 	switch (attrtype) {
5638 	case IFLA_INET6_STATS:
5639 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5640 				     offsetof(struct ipstats_mib, syncp));
5641 		break;
5642 	case IFLA_INET6_ICMP6STATS:
5643 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5644 		break;
5645 	}
5646 }
5647 
5648 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5649 				  u32 ext_filter_mask)
5650 {
5651 	struct nlattr *nla;
5652 	struct ifla_cacheinfo ci;
5653 
5654 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5655 		goto nla_put_failure;
5656 	ci.max_reasm_len = IPV6_MAXPLEN;
5657 	ci.tstamp = cstamp_delta(idev->tstamp);
5658 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5659 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5660 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5661 		goto nla_put_failure;
5662 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5663 	if (!nla)
5664 		goto nla_put_failure;
5665 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5666 
5667 	/* XXX - MC not implemented */
5668 
5669 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5670 		return 0;
5671 
5672 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5673 	if (!nla)
5674 		goto nla_put_failure;
5675 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5676 
5677 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5678 	if (!nla)
5679 		goto nla_put_failure;
5680 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5681 
5682 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5683 	if (!nla)
5684 		goto nla_put_failure;
5685 	read_lock_bh(&idev->lock);
5686 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5687 	read_unlock_bh(&idev->lock);
5688 
5689 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5690 		goto nla_put_failure;
5691 
5692 	if (idev->ra_mtu &&
5693 	    nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5694 		goto nla_put_failure;
5695 
5696 	return 0;
5697 
5698 nla_put_failure:
5699 	return -EMSGSIZE;
5700 }
5701 
5702 static size_t inet6_get_link_af_size(const struct net_device *dev,
5703 				     u32 ext_filter_mask)
5704 {
5705 	if (!__in6_dev_get(dev))
5706 		return 0;
5707 
5708 	return inet6_ifla6_size();
5709 }
5710 
5711 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5712 			      u32 ext_filter_mask)
5713 {
5714 	struct inet6_dev *idev = __in6_dev_get(dev);
5715 
5716 	if (!idev)
5717 		return -ENODATA;
5718 
5719 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5720 		return -EMSGSIZE;
5721 
5722 	return 0;
5723 }
5724 
5725 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5726 			     struct netlink_ext_ack *extack)
5727 {
5728 	struct inet6_ifaddr *ifp;
5729 	struct net_device *dev = idev->dev;
5730 	bool clear_token, update_rs = false;
5731 	struct in6_addr ll_addr;
5732 
5733 	ASSERT_RTNL();
5734 
5735 	if (!token)
5736 		return -EINVAL;
5737 
5738 	if (dev->flags & IFF_LOOPBACK) {
5739 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5740 		return -EINVAL;
5741 	}
5742 
5743 	if (dev->flags & IFF_NOARP) {
5744 		NL_SET_ERR_MSG_MOD(extack,
5745 				   "Device does not do neighbour discovery");
5746 		return -EINVAL;
5747 	}
5748 
5749 	if (!ipv6_accept_ra(idev)) {
5750 		NL_SET_ERR_MSG_MOD(extack,
5751 				   "Router advertisement is disabled on device");
5752 		return -EINVAL;
5753 	}
5754 
5755 	if (idev->cnf.rtr_solicits == 0) {
5756 		NL_SET_ERR_MSG(extack,
5757 			       "Router solicitation is disabled on device");
5758 		return -EINVAL;
5759 	}
5760 
5761 	write_lock_bh(&idev->lock);
5762 
5763 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5764 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5765 
5766 	write_unlock_bh(&idev->lock);
5767 
5768 	clear_token = ipv6_addr_any(token);
5769 	if (clear_token)
5770 		goto update_lft;
5771 
5772 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5773 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5774 			     IFA_F_OPTIMISTIC)) {
5775 		/* If we're not ready, then normal ifup will take care
5776 		 * of this. Otherwise, we need to request our rs here.
5777 		 */
5778 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5779 		update_rs = true;
5780 	}
5781 
5782 update_lft:
5783 	write_lock_bh(&idev->lock);
5784 
5785 	if (update_rs) {
5786 		idev->if_flags |= IF_RS_SENT;
5787 		idev->rs_interval = rfc3315_s14_backoff_init(
5788 			idev->cnf.rtr_solicit_interval);
5789 		idev->rs_probes = 1;
5790 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5791 	}
5792 
5793 	/* Well, that's kinda nasty ... */
5794 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5795 		spin_lock(&ifp->lock);
5796 		if (ifp->tokenized) {
5797 			ifp->valid_lft = 0;
5798 			ifp->prefered_lft = 0;
5799 		}
5800 		spin_unlock(&ifp->lock);
5801 	}
5802 
5803 	write_unlock_bh(&idev->lock);
5804 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5805 	addrconf_verify_rtnl(dev_net(dev));
5806 	return 0;
5807 }
5808 
5809 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5810 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5811 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5812 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5813 					    .reject_message =
5814 						"IFLA_INET6_RA_MTU can not be set" },
5815 };
5816 
5817 static int check_addr_gen_mode(int mode)
5818 {
5819 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5820 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5821 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5822 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5823 		return -EINVAL;
5824 	return 1;
5825 }
5826 
5827 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5828 				int mode)
5829 {
5830 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5831 	    !idev->cnf.stable_secret.initialized &&
5832 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5833 		return -EINVAL;
5834 	return 1;
5835 }
5836 
5837 static int inet6_validate_link_af(const struct net_device *dev,
5838 				  const struct nlattr *nla,
5839 				  struct netlink_ext_ack *extack)
5840 {
5841 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5842 	struct inet6_dev *idev = NULL;
5843 	int err;
5844 
5845 	if (dev) {
5846 		idev = __in6_dev_get(dev);
5847 		if (!idev)
5848 			return -EAFNOSUPPORT;
5849 	}
5850 
5851 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5852 					  inet6_af_policy, extack);
5853 	if (err)
5854 		return err;
5855 
5856 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5857 		return -EINVAL;
5858 
5859 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5860 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5861 
5862 		if (check_addr_gen_mode(mode) < 0)
5863 			return -EINVAL;
5864 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5865 			return -EINVAL;
5866 	}
5867 
5868 	return 0;
5869 }
5870 
5871 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5872 			     struct netlink_ext_ack *extack)
5873 {
5874 	struct inet6_dev *idev = __in6_dev_get(dev);
5875 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5876 	int err;
5877 
5878 	if (!idev)
5879 		return -EAFNOSUPPORT;
5880 
5881 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5882 		return -EINVAL;
5883 
5884 	if (tb[IFLA_INET6_TOKEN]) {
5885 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5886 					extack);
5887 		if (err)
5888 			return err;
5889 	}
5890 
5891 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5892 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5893 
5894 		idev->cnf.addr_gen_mode = mode;
5895 	}
5896 
5897 	return 0;
5898 }
5899 
5900 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5901 			     u32 portid, u32 seq, int event, unsigned int flags)
5902 {
5903 	struct net_device *dev = idev->dev;
5904 	struct ifinfomsg *hdr;
5905 	struct nlmsghdr *nlh;
5906 	void *protoinfo;
5907 
5908 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5909 	if (!nlh)
5910 		return -EMSGSIZE;
5911 
5912 	hdr = nlmsg_data(nlh);
5913 	hdr->ifi_family = AF_INET6;
5914 	hdr->__ifi_pad = 0;
5915 	hdr->ifi_type = dev->type;
5916 	hdr->ifi_index = dev->ifindex;
5917 	hdr->ifi_flags = dev_get_flags(dev);
5918 	hdr->ifi_change = 0;
5919 
5920 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5921 	    (dev->addr_len &&
5922 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5923 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5924 	    (dev->ifindex != dev_get_iflink(dev) &&
5925 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5926 	    nla_put_u8(skb, IFLA_OPERSTATE,
5927 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5928 		goto nla_put_failure;
5929 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5930 	if (!protoinfo)
5931 		goto nla_put_failure;
5932 
5933 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5934 		goto nla_put_failure;
5935 
5936 	nla_nest_end(skb, protoinfo);
5937 	nlmsg_end(skb, nlh);
5938 	return 0;
5939 
5940 nla_put_failure:
5941 	nlmsg_cancel(skb, nlh);
5942 	return -EMSGSIZE;
5943 }
5944 
5945 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5946 				   struct netlink_ext_ack *extack)
5947 {
5948 	struct ifinfomsg *ifm;
5949 
5950 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5951 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5952 		return -EINVAL;
5953 	}
5954 
5955 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5956 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5957 		return -EINVAL;
5958 	}
5959 
5960 	ifm = nlmsg_data(nlh);
5961 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5962 	    ifm->ifi_change || ifm->ifi_index) {
5963 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5964 		return -EINVAL;
5965 	}
5966 
5967 	return 0;
5968 }
5969 
5970 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5971 {
5972 	struct net *net = sock_net(skb->sk);
5973 	int h, s_h;
5974 	int idx = 0, s_idx;
5975 	struct net_device *dev;
5976 	struct inet6_dev *idev;
5977 	struct hlist_head *head;
5978 
5979 	/* only requests using strict checking can pass data to
5980 	 * influence the dump
5981 	 */
5982 	if (cb->strict_check) {
5983 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5984 
5985 		if (err < 0)
5986 			return err;
5987 	}
5988 
5989 	s_h = cb->args[0];
5990 	s_idx = cb->args[1];
5991 
5992 	rcu_read_lock();
5993 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5994 		idx = 0;
5995 		head = &net->dev_index_head[h];
5996 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5997 			if (idx < s_idx)
5998 				goto cont;
5999 			idev = __in6_dev_get(dev);
6000 			if (!idev)
6001 				goto cont;
6002 			if (inet6_fill_ifinfo(skb, idev,
6003 					      NETLINK_CB(cb->skb).portid,
6004 					      cb->nlh->nlmsg_seq,
6005 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
6006 				goto out;
6007 cont:
6008 			idx++;
6009 		}
6010 	}
6011 out:
6012 	rcu_read_unlock();
6013 	cb->args[1] = idx;
6014 	cb->args[0] = h;
6015 
6016 	return skb->len;
6017 }
6018 
6019 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6020 {
6021 	struct sk_buff *skb;
6022 	struct net *net = dev_net(idev->dev);
6023 	int err = -ENOBUFS;
6024 
6025 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6026 	if (!skb)
6027 		goto errout;
6028 
6029 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6030 	if (err < 0) {
6031 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6032 		WARN_ON(err == -EMSGSIZE);
6033 		kfree_skb(skb);
6034 		goto errout;
6035 	}
6036 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6037 	return;
6038 errout:
6039 	if (err < 0)
6040 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6041 }
6042 
6043 static inline size_t inet6_prefix_nlmsg_size(void)
6044 {
6045 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6046 	       + nla_total_size(sizeof(struct in6_addr))
6047 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6048 }
6049 
6050 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6051 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6052 			     int event, unsigned int flags)
6053 {
6054 	struct prefixmsg *pmsg;
6055 	struct nlmsghdr *nlh;
6056 	struct prefix_cacheinfo	ci;
6057 
6058 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6059 	if (!nlh)
6060 		return -EMSGSIZE;
6061 
6062 	pmsg = nlmsg_data(nlh);
6063 	pmsg->prefix_family = AF_INET6;
6064 	pmsg->prefix_pad1 = 0;
6065 	pmsg->prefix_pad2 = 0;
6066 	pmsg->prefix_ifindex = idev->dev->ifindex;
6067 	pmsg->prefix_len = pinfo->prefix_len;
6068 	pmsg->prefix_type = pinfo->type;
6069 	pmsg->prefix_pad3 = 0;
6070 	pmsg->prefix_flags = 0;
6071 	if (pinfo->onlink)
6072 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6073 	if (pinfo->autoconf)
6074 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6075 
6076 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6077 		goto nla_put_failure;
6078 	ci.preferred_time = ntohl(pinfo->prefered);
6079 	ci.valid_time = ntohl(pinfo->valid);
6080 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6081 		goto nla_put_failure;
6082 	nlmsg_end(skb, nlh);
6083 	return 0;
6084 
6085 nla_put_failure:
6086 	nlmsg_cancel(skb, nlh);
6087 	return -EMSGSIZE;
6088 }
6089 
6090 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6091 			 struct prefix_info *pinfo)
6092 {
6093 	struct sk_buff *skb;
6094 	struct net *net = dev_net(idev->dev);
6095 	int err = -ENOBUFS;
6096 
6097 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6098 	if (!skb)
6099 		goto errout;
6100 
6101 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6102 	if (err < 0) {
6103 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6104 		WARN_ON(err == -EMSGSIZE);
6105 		kfree_skb(skb);
6106 		goto errout;
6107 	}
6108 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6109 	return;
6110 errout:
6111 	if (err < 0)
6112 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6113 }
6114 
6115 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6116 {
6117 	struct net *net = dev_net(ifp->idev->dev);
6118 
6119 	if (event)
6120 		ASSERT_RTNL();
6121 
6122 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6123 
6124 	switch (event) {
6125 	case RTM_NEWADDR:
6126 		/*
6127 		 * If the address was optimistic we inserted the route at the
6128 		 * start of our DAD process, so we don't need to do it again.
6129 		 * If the device was taken down in the middle of the DAD
6130 		 * cycle there is a race where we could get here without a
6131 		 * host route, so nothing to insert. That will be fixed when
6132 		 * the device is brought up.
6133 		 */
6134 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6135 			ip6_ins_rt(net, ifp->rt);
6136 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6137 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6138 				&ifp->addr, ifp->idev->dev->name);
6139 		}
6140 
6141 		if (ifp->idev->cnf.forwarding)
6142 			addrconf_join_anycast(ifp);
6143 		if (!ipv6_addr_any(&ifp->peer_addr))
6144 			addrconf_prefix_route(&ifp->peer_addr, 128,
6145 					      ifp->rt_priority, ifp->idev->dev,
6146 					      0, 0, GFP_ATOMIC);
6147 		break;
6148 	case RTM_DELADDR:
6149 		if (ifp->idev->cnf.forwarding)
6150 			addrconf_leave_anycast(ifp);
6151 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6152 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6153 			struct fib6_info *rt;
6154 
6155 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6156 						       ifp->idev->dev, 0, 0,
6157 						       false);
6158 			if (rt)
6159 				ip6_del_rt(net, rt, false);
6160 		}
6161 		if (ifp->rt) {
6162 			ip6_del_rt(net, ifp->rt, false);
6163 			ifp->rt = NULL;
6164 		}
6165 		rt_genid_bump_ipv6(net);
6166 		break;
6167 	}
6168 	atomic_inc(&net->ipv6.dev_addr_genid);
6169 }
6170 
6171 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6172 {
6173 	if (likely(ifp->idev->dead == 0))
6174 		__ipv6_ifa_notify(event, ifp);
6175 }
6176 
6177 #ifdef CONFIG_SYSCTL
6178 
6179 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6180 		void *buffer, size_t *lenp, loff_t *ppos)
6181 {
6182 	int *valp = ctl->data;
6183 	int val = *valp;
6184 	loff_t pos = *ppos;
6185 	struct ctl_table lctl;
6186 	int ret;
6187 
6188 	/*
6189 	 * ctl->data points to idev->cnf.forwarding, we should
6190 	 * not modify it until we get the rtnl lock.
6191 	 */
6192 	lctl = *ctl;
6193 	lctl.data = &val;
6194 
6195 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6196 
6197 	if (write)
6198 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6199 	if (ret)
6200 		*ppos = pos;
6201 	return ret;
6202 }
6203 
6204 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6205 		void *buffer, size_t *lenp, loff_t *ppos)
6206 {
6207 	struct inet6_dev *idev = ctl->extra1;
6208 	int min_mtu = IPV6_MIN_MTU;
6209 	struct ctl_table lctl;
6210 
6211 	lctl = *ctl;
6212 	lctl.extra1 = &min_mtu;
6213 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6214 
6215 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6216 }
6217 
6218 static void dev_disable_change(struct inet6_dev *idev)
6219 {
6220 	struct netdev_notifier_info info;
6221 
6222 	if (!idev || !idev->dev)
6223 		return;
6224 
6225 	netdev_notifier_info_init(&info, idev->dev);
6226 	if (idev->cnf.disable_ipv6)
6227 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6228 	else
6229 		addrconf_notify(NULL, NETDEV_UP, &info);
6230 }
6231 
6232 static void addrconf_disable_change(struct net *net, __s32 newf)
6233 {
6234 	struct net_device *dev;
6235 	struct inet6_dev *idev;
6236 
6237 	for_each_netdev(net, dev) {
6238 		idev = __in6_dev_get(dev);
6239 		if (idev) {
6240 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6241 			idev->cnf.disable_ipv6 = newf;
6242 			if (changed)
6243 				dev_disable_change(idev);
6244 		}
6245 	}
6246 }
6247 
6248 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6249 {
6250 	struct net *net;
6251 	int old;
6252 
6253 	if (!rtnl_trylock())
6254 		return restart_syscall();
6255 
6256 	net = (struct net *)table->extra2;
6257 	old = *p;
6258 	*p = newf;
6259 
6260 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6261 		rtnl_unlock();
6262 		return 0;
6263 	}
6264 
6265 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6266 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
6267 		addrconf_disable_change(net, newf);
6268 	} else if ((!newf) ^ (!old))
6269 		dev_disable_change((struct inet6_dev *)table->extra1);
6270 
6271 	rtnl_unlock();
6272 	return 0;
6273 }
6274 
6275 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6276 		void *buffer, size_t *lenp, loff_t *ppos)
6277 {
6278 	int *valp = ctl->data;
6279 	int val = *valp;
6280 	loff_t pos = *ppos;
6281 	struct ctl_table lctl;
6282 	int ret;
6283 
6284 	/*
6285 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6286 	 * not modify it until we get the rtnl lock.
6287 	 */
6288 	lctl = *ctl;
6289 	lctl.data = &val;
6290 
6291 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6292 
6293 	if (write)
6294 		ret = addrconf_disable_ipv6(ctl, valp, val);
6295 	if (ret)
6296 		*ppos = pos;
6297 	return ret;
6298 }
6299 
6300 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6301 		void *buffer, size_t *lenp, loff_t *ppos)
6302 {
6303 	int *valp = ctl->data;
6304 	int ret;
6305 	int old, new;
6306 
6307 	old = *valp;
6308 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6309 	new = *valp;
6310 
6311 	if (write && old != new) {
6312 		struct net *net = ctl->extra2;
6313 
6314 		if (!rtnl_trylock())
6315 			return restart_syscall();
6316 
6317 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6318 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6319 						     NETCONFA_PROXY_NEIGH,
6320 						     NETCONFA_IFINDEX_DEFAULT,
6321 						     net->ipv6.devconf_dflt);
6322 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6323 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6324 						     NETCONFA_PROXY_NEIGH,
6325 						     NETCONFA_IFINDEX_ALL,
6326 						     net->ipv6.devconf_all);
6327 		else {
6328 			struct inet6_dev *idev = ctl->extra1;
6329 
6330 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6331 						     NETCONFA_PROXY_NEIGH,
6332 						     idev->dev->ifindex,
6333 						     &idev->cnf);
6334 		}
6335 		rtnl_unlock();
6336 	}
6337 
6338 	return ret;
6339 }
6340 
6341 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6342 					 void *buffer, size_t *lenp,
6343 					 loff_t *ppos)
6344 {
6345 	int ret = 0;
6346 	u32 new_val;
6347 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6348 	struct net *net = (struct net *)ctl->extra2;
6349 	struct ctl_table tmp = {
6350 		.data = &new_val,
6351 		.maxlen = sizeof(new_val),
6352 		.mode = ctl->mode,
6353 	};
6354 
6355 	if (!rtnl_trylock())
6356 		return restart_syscall();
6357 
6358 	new_val = *((u32 *)ctl->data);
6359 
6360 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6361 	if (ret != 0)
6362 		goto out;
6363 
6364 	if (write) {
6365 		if (check_addr_gen_mode(new_val) < 0) {
6366 			ret = -EINVAL;
6367 			goto out;
6368 		}
6369 
6370 		if (idev) {
6371 			if (check_stable_privacy(idev, net, new_val) < 0) {
6372 				ret = -EINVAL;
6373 				goto out;
6374 			}
6375 
6376 			if (idev->cnf.addr_gen_mode != new_val) {
6377 				idev->cnf.addr_gen_mode = new_val;
6378 				addrconf_dev_config(idev->dev);
6379 			}
6380 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6381 			struct net_device *dev;
6382 
6383 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6384 			for_each_netdev(net, dev) {
6385 				idev = __in6_dev_get(dev);
6386 				if (idev &&
6387 				    idev->cnf.addr_gen_mode != new_val) {
6388 					idev->cnf.addr_gen_mode = new_val;
6389 					addrconf_dev_config(idev->dev);
6390 				}
6391 			}
6392 		}
6393 
6394 		*((u32 *)ctl->data) = new_val;
6395 	}
6396 
6397 out:
6398 	rtnl_unlock();
6399 
6400 	return ret;
6401 }
6402 
6403 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6404 					 void *buffer, size_t *lenp,
6405 					 loff_t *ppos)
6406 {
6407 	int err;
6408 	struct in6_addr addr;
6409 	char str[IPV6_MAX_STRLEN];
6410 	struct ctl_table lctl = *ctl;
6411 	struct net *net = ctl->extra2;
6412 	struct ipv6_stable_secret *secret = ctl->data;
6413 
6414 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6415 		return -EIO;
6416 
6417 	lctl.maxlen = IPV6_MAX_STRLEN;
6418 	lctl.data = str;
6419 
6420 	if (!rtnl_trylock())
6421 		return restart_syscall();
6422 
6423 	if (!write && !secret->initialized) {
6424 		err = -EIO;
6425 		goto out;
6426 	}
6427 
6428 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6429 	if (err >= sizeof(str)) {
6430 		err = -EIO;
6431 		goto out;
6432 	}
6433 
6434 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6435 	if (err || !write)
6436 		goto out;
6437 
6438 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6439 		err = -EIO;
6440 		goto out;
6441 	}
6442 
6443 	secret->initialized = true;
6444 	secret->secret = addr;
6445 
6446 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6447 		struct net_device *dev;
6448 
6449 		for_each_netdev(net, dev) {
6450 			struct inet6_dev *idev = __in6_dev_get(dev);
6451 
6452 			if (idev) {
6453 				idev->cnf.addr_gen_mode =
6454 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6455 			}
6456 		}
6457 	} else {
6458 		struct inet6_dev *idev = ctl->extra1;
6459 
6460 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6461 	}
6462 
6463 out:
6464 	rtnl_unlock();
6465 
6466 	return err;
6467 }
6468 
6469 static
6470 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6471 						int write, void *buffer,
6472 						size_t *lenp,
6473 						loff_t *ppos)
6474 {
6475 	int *valp = ctl->data;
6476 	int val = *valp;
6477 	loff_t pos = *ppos;
6478 	struct ctl_table lctl;
6479 	int ret;
6480 
6481 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6482 	 * we should not modify it until we get the rtnl lock.
6483 	 */
6484 	lctl = *ctl;
6485 	lctl.data = &val;
6486 
6487 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6488 
6489 	if (write)
6490 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6491 	if (ret)
6492 		*ppos = pos;
6493 	return ret;
6494 }
6495 
6496 static
6497 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6498 {
6499 	if (rt) {
6500 		if (action)
6501 			rt->dst.flags |= DST_NOPOLICY;
6502 		else
6503 			rt->dst.flags &= ~DST_NOPOLICY;
6504 	}
6505 }
6506 
6507 static
6508 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6509 {
6510 	struct inet6_ifaddr *ifa;
6511 
6512 	read_lock_bh(&idev->lock);
6513 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6514 		spin_lock(&ifa->lock);
6515 		if (ifa->rt) {
6516 			/* host routes only use builtin fib6_nh */
6517 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6518 			int cpu;
6519 
6520 			rcu_read_lock();
6521 			ifa->rt->dst_nopolicy = val ? true : false;
6522 			if (nh->rt6i_pcpu) {
6523 				for_each_possible_cpu(cpu) {
6524 					struct rt6_info **rtp;
6525 
6526 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6527 					addrconf_set_nopolicy(*rtp, val);
6528 				}
6529 			}
6530 			rcu_read_unlock();
6531 		}
6532 		spin_unlock(&ifa->lock);
6533 	}
6534 	read_unlock_bh(&idev->lock);
6535 }
6536 
6537 static
6538 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6539 {
6540 	struct inet6_dev *idev;
6541 	struct net *net;
6542 
6543 	if (!rtnl_trylock())
6544 		return restart_syscall();
6545 
6546 	*valp = val;
6547 
6548 	net = (struct net *)ctl->extra2;
6549 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6550 		rtnl_unlock();
6551 		return 0;
6552 	}
6553 
6554 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6555 		struct net_device *dev;
6556 
6557 		for_each_netdev(net, dev) {
6558 			idev = __in6_dev_get(dev);
6559 			if (idev)
6560 				addrconf_disable_policy_idev(idev, val);
6561 		}
6562 	} else {
6563 		idev = (struct inet6_dev *)ctl->extra1;
6564 		addrconf_disable_policy_idev(idev, val);
6565 	}
6566 
6567 	rtnl_unlock();
6568 	return 0;
6569 }
6570 
6571 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6572 				   void *buffer, size_t *lenp, loff_t *ppos)
6573 {
6574 	int *valp = ctl->data;
6575 	int val = *valp;
6576 	loff_t pos = *ppos;
6577 	struct ctl_table lctl;
6578 	int ret;
6579 
6580 	lctl = *ctl;
6581 	lctl.data = &val;
6582 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6583 
6584 	if (write && (*valp != val))
6585 		ret = addrconf_disable_policy(ctl, valp, val);
6586 
6587 	if (ret)
6588 		*ppos = pos;
6589 
6590 	return ret;
6591 }
6592 
6593 static int minus_one = -1;
6594 static const int two_five_five = 255;
6595 static u32 ioam6_if_id_max = U16_MAX;
6596 
6597 static const struct ctl_table addrconf_sysctl[] = {
6598 	{
6599 		.procname	= "forwarding",
6600 		.data		= &ipv6_devconf.forwarding,
6601 		.maxlen		= sizeof(int),
6602 		.mode		= 0644,
6603 		.proc_handler	= addrconf_sysctl_forward,
6604 	},
6605 	{
6606 		.procname	= "hop_limit",
6607 		.data		= &ipv6_devconf.hop_limit,
6608 		.maxlen		= sizeof(int),
6609 		.mode		= 0644,
6610 		.proc_handler	= proc_dointvec_minmax,
6611 		.extra1		= (void *)SYSCTL_ONE,
6612 		.extra2		= (void *)&two_five_five,
6613 	},
6614 	{
6615 		.procname	= "mtu",
6616 		.data		= &ipv6_devconf.mtu6,
6617 		.maxlen		= sizeof(int),
6618 		.mode		= 0644,
6619 		.proc_handler	= addrconf_sysctl_mtu,
6620 	},
6621 	{
6622 		.procname	= "accept_ra",
6623 		.data		= &ipv6_devconf.accept_ra,
6624 		.maxlen		= sizeof(int),
6625 		.mode		= 0644,
6626 		.proc_handler	= proc_dointvec,
6627 	},
6628 	{
6629 		.procname	= "accept_redirects",
6630 		.data		= &ipv6_devconf.accept_redirects,
6631 		.maxlen		= sizeof(int),
6632 		.mode		= 0644,
6633 		.proc_handler	= proc_dointvec,
6634 	},
6635 	{
6636 		.procname	= "autoconf",
6637 		.data		= &ipv6_devconf.autoconf,
6638 		.maxlen		= sizeof(int),
6639 		.mode		= 0644,
6640 		.proc_handler	= proc_dointvec,
6641 	},
6642 	{
6643 		.procname	= "dad_transmits",
6644 		.data		= &ipv6_devconf.dad_transmits,
6645 		.maxlen		= sizeof(int),
6646 		.mode		= 0644,
6647 		.proc_handler	= proc_dointvec,
6648 	},
6649 	{
6650 		.procname	= "router_solicitations",
6651 		.data		= &ipv6_devconf.rtr_solicits,
6652 		.maxlen		= sizeof(int),
6653 		.mode		= 0644,
6654 		.proc_handler	= proc_dointvec_minmax,
6655 		.extra1		= &minus_one,
6656 	},
6657 	{
6658 		.procname	= "router_solicitation_interval",
6659 		.data		= &ipv6_devconf.rtr_solicit_interval,
6660 		.maxlen		= sizeof(int),
6661 		.mode		= 0644,
6662 		.proc_handler	= proc_dointvec_jiffies,
6663 	},
6664 	{
6665 		.procname	= "router_solicitation_max_interval",
6666 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6667 		.maxlen		= sizeof(int),
6668 		.mode		= 0644,
6669 		.proc_handler	= proc_dointvec_jiffies,
6670 	},
6671 	{
6672 		.procname	= "router_solicitation_delay",
6673 		.data		= &ipv6_devconf.rtr_solicit_delay,
6674 		.maxlen		= sizeof(int),
6675 		.mode		= 0644,
6676 		.proc_handler	= proc_dointvec_jiffies,
6677 	},
6678 	{
6679 		.procname	= "force_mld_version",
6680 		.data		= &ipv6_devconf.force_mld_version,
6681 		.maxlen		= sizeof(int),
6682 		.mode		= 0644,
6683 		.proc_handler	= proc_dointvec,
6684 	},
6685 	{
6686 		.procname	= "mldv1_unsolicited_report_interval",
6687 		.data		=
6688 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6689 		.maxlen		= sizeof(int),
6690 		.mode		= 0644,
6691 		.proc_handler	= proc_dointvec_ms_jiffies,
6692 	},
6693 	{
6694 		.procname	= "mldv2_unsolicited_report_interval",
6695 		.data		=
6696 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6697 		.maxlen		= sizeof(int),
6698 		.mode		= 0644,
6699 		.proc_handler	= proc_dointvec_ms_jiffies,
6700 	},
6701 	{
6702 		.procname	= "use_tempaddr",
6703 		.data		= &ipv6_devconf.use_tempaddr,
6704 		.maxlen		= sizeof(int),
6705 		.mode		= 0644,
6706 		.proc_handler	= proc_dointvec,
6707 	},
6708 	{
6709 		.procname	= "temp_valid_lft",
6710 		.data		= &ipv6_devconf.temp_valid_lft,
6711 		.maxlen		= sizeof(int),
6712 		.mode		= 0644,
6713 		.proc_handler	= proc_dointvec,
6714 	},
6715 	{
6716 		.procname	= "temp_prefered_lft",
6717 		.data		= &ipv6_devconf.temp_prefered_lft,
6718 		.maxlen		= sizeof(int),
6719 		.mode		= 0644,
6720 		.proc_handler	= proc_dointvec,
6721 	},
6722 	{
6723 		.procname	= "regen_max_retry",
6724 		.data		= &ipv6_devconf.regen_max_retry,
6725 		.maxlen		= sizeof(int),
6726 		.mode		= 0644,
6727 		.proc_handler	= proc_dointvec,
6728 	},
6729 	{
6730 		.procname	= "max_desync_factor",
6731 		.data		= &ipv6_devconf.max_desync_factor,
6732 		.maxlen		= sizeof(int),
6733 		.mode		= 0644,
6734 		.proc_handler	= proc_dointvec,
6735 	},
6736 	{
6737 		.procname	= "max_addresses",
6738 		.data		= &ipv6_devconf.max_addresses,
6739 		.maxlen		= sizeof(int),
6740 		.mode		= 0644,
6741 		.proc_handler	= proc_dointvec,
6742 	},
6743 	{
6744 		.procname	= "accept_ra_defrtr",
6745 		.data		= &ipv6_devconf.accept_ra_defrtr,
6746 		.maxlen		= sizeof(int),
6747 		.mode		= 0644,
6748 		.proc_handler	= proc_dointvec,
6749 	},
6750 	{
6751 		.procname	= "ra_defrtr_metric",
6752 		.data		= &ipv6_devconf.ra_defrtr_metric,
6753 		.maxlen		= sizeof(u32),
6754 		.mode		= 0644,
6755 		.proc_handler	= proc_douintvec_minmax,
6756 		.extra1		= (void *)SYSCTL_ONE,
6757 	},
6758 	{
6759 		.procname	= "accept_ra_min_hop_limit",
6760 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6761 		.maxlen		= sizeof(int),
6762 		.mode		= 0644,
6763 		.proc_handler	= proc_dointvec,
6764 	},
6765 	{
6766 		.procname	= "accept_ra_pinfo",
6767 		.data		= &ipv6_devconf.accept_ra_pinfo,
6768 		.maxlen		= sizeof(int),
6769 		.mode		= 0644,
6770 		.proc_handler	= proc_dointvec,
6771 	},
6772 #ifdef CONFIG_IPV6_ROUTER_PREF
6773 	{
6774 		.procname	= "accept_ra_rtr_pref",
6775 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6776 		.maxlen		= sizeof(int),
6777 		.mode		= 0644,
6778 		.proc_handler	= proc_dointvec,
6779 	},
6780 	{
6781 		.procname	= "router_probe_interval",
6782 		.data		= &ipv6_devconf.rtr_probe_interval,
6783 		.maxlen		= sizeof(int),
6784 		.mode		= 0644,
6785 		.proc_handler	= proc_dointvec_jiffies,
6786 	},
6787 #ifdef CONFIG_IPV6_ROUTE_INFO
6788 	{
6789 		.procname	= "accept_ra_rt_info_min_plen",
6790 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6791 		.maxlen		= sizeof(int),
6792 		.mode		= 0644,
6793 		.proc_handler	= proc_dointvec,
6794 	},
6795 	{
6796 		.procname	= "accept_ra_rt_info_max_plen",
6797 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6798 		.maxlen		= sizeof(int),
6799 		.mode		= 0644,
6800 		.proc_handler	= proc_dointvec,
6801 	},
6802 #endif
6803 #endif
6804 	{
6805 		.procname	= "proxy_ndp",
6806 		.data		= &ipv6_devconf.proxy_ndp,
6807 		.maxlen		= sizeof(int),
6808 		.mode		= 0644,
6809 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6810 	},
6811 	{
6812 		.procname	= "accept_source_route",
6813 		.data		= &ipv6_devconf.accept_source_route,
6814 		.maxlen		= sizeof(int),
6815 		.mode		= 0644,
6816 		.proc_handler	= proc_dointvec,
6817 	},
6818 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6819 	{
6820 		.procname	= "optimistic_dad",
6821 		.data		= &ipv6_devconf.optimistic_dad,
6822 		.maxlen		= sizeof(int),
6823 		.mode		= 0644,
6824 		.proc_handler   = proc_dointvec,
6825 	},
6826 	{
6827 		.procname	= "use_optimistic",
6828 		.data		= &ipv6_devconf.use_optimistic,
6829 		.maxlen		= sizeof(int),
6830 		.mode		= 0644,
6831 		.proc_handler	= proc_dointvec,
6832 	},
6833 #endif
6834 #ifdef CONFIG_IPV6_MROUTE
6835 	{
6836 		.procname	= "mc_forwarding",
6837 		.data		= &ipv6_devconf.mc_forwarding,
6838 		.maxlen		= sizeof(int),
6839 		.mode		= 0444,
6840 		.proc_handler	= proc_dointvec,
6841 	},
6842 #endif
6843 	{
6844 		.procname	= "disable_ipv6",
6845 		.data		= &ipv6_devconf.disable_ipv6,
6846 		.maxlen		= sizeof(int),
6847 		.mode		= 0644,
6848 		.proc_handler	= addrconf_sysctl_disable,
6849 	},
6850 	{
6851 		.procname	= "accept_dad",
6852 		.data		= &ipv6_devconf.accept_dad,
6853 		.maxlen		= sizeof(int),
6854 		.mode		= 0644,
6855 		.proc_handler	= proc_dointvec,
6856 	},
6857 	{
6858 		.procname	= "force_tllao",
6859 		.data		= &ipv6_devconf.force_tllao,
6860 		.maxlen		= sizeof(int),
6861 		.mode		= 0644,
6862 		.proc_handler	= proc_dointvec
6863 	},
6864 	{
6865 		.procname	= "ndisc_notify",
6866 		.data		= &ipv6_devconf.ndisc_notify,
6867 		.maxlen		= sizeof(int),
6868 		.mode		= 0644,
6869 		.proc_handler	= proc_dointvec
6870 	},
6871 	{
6872 		.procname	= "suppress_frag_ndisc",
6873 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6874 		.maxlen		= sizeof(int),
6875 		.mode		= 0644,
6876 		.proc_handler	= proc_dointvec
6877 	},
6878 	{
6879 		.procname	= "accept_ra_from_local",
6880 		.data		= &ipv6_devconf.accept_ra_from_local,
6881 		.maxlen		= sizeof(int),
6882 		.mode		= 0644,
6883 		.proc_handler	= proc_dointvec,
6884 	},
6885 	{
6886 		.procname	= "accept_ra_mtu",
6887 		.data		= &ipv6_devconf.accept_ra_mtu,
6888 		.maxlen		= sizeof(int),
6889 		.mode		= 0644,
6890 		.proc_handler	= proc_dointvec,
6891 	},
6892 	{
6893 		.procname	= "stable_secret",
6894 		.data		= &ipv6_devconf.stable_secret,
6895 		.maxlen		= IPV6_MAX_STRLEN,
6896 		.mode		= 0600,
6897 		.proc_handler	= addrconf_sysctl_stable_secret,
6898 	},
6899 	{
6900 		.procname	= "use_oif_addrs_only",
6901 		.data		= &ipv6_devconf.use_oif_addrs_only,
6902 		.maxlen		= sizeof(int),
6903 		.mode		= 0644,
6904 		.proc_handler	= proc_dointvec,
6905 	},
6906 	{
6907 		.procname	= "ignore_routes_with_linkdown",
6908 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6909 		.maxlen		= sizeof(int),
6910 		.mode		= 0644,
6911 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6912 	},
6913 	{
6914 		.procname	= "drop_unicast_in_l2_multicast",
6915 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6916 		.maxlen		= sizeof(int),
6917 		.mode		= 0644,
6918 		.proc_handler	= proc_dointvec,
6919 	},
6920 	{
6921 		.procname	= "drop_unsolicited_na",
6922 		.data		= &ipv6_devconf.drop_unsolicited_na,
6923 		.maxlen		= sizeof(int),
6924 		.mode		= 0644,
6925 		.proc_handler	= proc_dointvec,
6926 	},
6927 	{
6928 		.procname	= "keep_addr_on_down",
6929 		.data		= &ipv6_devconf.keep_addr_on_down,
6930 		.maxlen		= sizeof(int),
6931 		.mode		= 0644,
6932 		.proc_handler	= proc_dointvec,
6933 
6934 	},
6935 	{
6936 		.procname	= "seg6_enabled",
6937 		.data		= &ipv6_devconf.seg6_enabled,
6938 		.maxlen		= sizeof(int),
6939 		.mode		= 0644,
6940 		.proc_handler	= proc_dointvec,
6941 	},
6942 #ifdef CONFIG_IPV6_SEG6_HMAC
6943 	{
6944 		.procname	= "seg6_require_hmac",
6945 		.data		= &ipv6_devconf.seg6_require_hmac,
6946 		.maxlen		= sizeof(int),
6947 		.mode		= 0644,
6948 		.proc_handler	= proc_dointvec,
6949 	},
6950 #endif
6951 	{
6952 		.procname       = "enhanced_dad",
6953 		.data           = &ipv6_devconf.enhanced_dad,
6954 		.maxlen         = sizeof(int),
6955 		.mode           = 0644,
6956 		.proc_handler   = proc_dointvec,
6957 	},
6958 	{
6959 		.procname	= "addr_gen_mode",
6960 		.data		= &ipv6_devconf.addr_gen_mode,
6961 		.maxlen		= sizeof(int),
6962 		.mode		= 0644,
6963 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6964 	},
6965 	{
6966 		.procname       = "disable_policy",
6967 		.data           = &ipv6_devconf.disable_policy,
6968 		.maxlen         = sizeof(int),
6969 		.mode           = 0644,
6970 		.proc_handler   = addrconf_sysctl_disable_policy,
6971 	},
6972 	{
6973 		.procname	= "ndisc_tclass",
6974 		.data		= &ipv6_devconf.ndisc_tclass,
6975 		.maxlen		= sizeof(int),
6976 		.mode		= 0644,
6977 		.proc_handler	= proc_dointvec_minmax,
6978 		.extra1		= (void *)SYSCTL_ZERO,
6979 		.extra2		= (void *)&two_five_five,
6980 	},
6981 	{
6982 		.procname	= "rpl_seg_enabled",
6983 		.data		= &ipv6_devconf.rpl_seg_enabled,
6984 		.maxlen		= sizeof(int),
6985 		.mode		= 0644,
6986 		.proc_handler	= proc_dointvec,
6987 	},
6988 	{
6989 		.procname	= "ioam6_enabled",
6990 		.data		= &ipv6_devconf.ioam6_enabled,
6991 		.maxlen		= sizeof(u8),
6992 		.mode		= 0644,
6993 		.proc_handler	= proc_dou8vec_minmax,
6994 		.extra1		= (void *)SYSCTL_ZERO,
6995 		.extra2		= (void *)SYSCTL_ONE,
6996 	},
6997 	{
6998 		.procname	= "ioam6_id",
6999 		.data		= &ipv6_devconf.ioam6_id,
7000 		.maxlen		= sizeof(u32),
7001 		.mode		= 0644,
7002 		.proc_handler	= proc_douintvec_minmax,
7003 		.extra1		= (void *)SYSCTL_ZERO,
7004 		.extra2		= (void *)&ioam6_if_id_max,
7005 	},
7006 	{
7007 		.procname	= "ioam6_id_wide",
7008 		.data		= &ipv6_devconf.ioam6_id_wide,
7009 		.maxlen		= sizeof(u32),
7010 		.mode		= 0644,
7011 		.proc_handler	= proc_douintvec,
7012 	},
7013 	{
7014 		.procname	= "ndisc_evict_nocarrier",
7015 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
7016 		.maxlen		= sizeof(u8),
7017 		.mode		= 0644,
7018 		.proc_handler	= proc_dou8vec_minmax,
7019 		.extra1		= (void *)SYSCTL_ZERO,
7020 		.extra2		= (void *)SYSCTL_ONE,
7021 	},
7022 	{
7023 		/* sentinel */
7024 	}
7025 };
7026 
7027 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7028 		struct inet6_dev *idev, struct ipv6_devconf *p)
7029 {
7030 	int i, ifindex;
7031 	struct ctl_table *table;
7032 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7033 
7034 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
7035 	if (!table)
7036 		goto out;
7037 
7038 	for (i = 0; table[i].data; i++) {
7039 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7040 		/* If one of these is already set, then it is not safe to
7041 		 * overwrite either of them: this makes proc_dointvec_minmax
7042 		 * usable.
7043 		 */
7044 		if (!table[i].extra1 && !table[i].extra2) {
7045 			table[i].extra1 = idev; /* embedded; no ref */
7046 			table[i].extra2 = net;
7047 		}
7048 	}
7049 
7050 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7051 
7052 	p->sysctl_header = register_net_sysctl(net, path, table);
7053 	if (!p->sysctl_header)
7054 		goto free;
7055 
7056 	if (!strcmp(dev_name, "all"))
7057 		ifindex = NETCONFA_IFINDEX_ALL;
7058 	else if (!strcmp(dev_name, "default"))
7059 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7060 	else
7061 		ifindex = idev->dev->ifindex;
7062 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7063 				     ifindex, p);
7064 	return 0;
7065 
7066 free:
7067 	kfree(table);
7068 out:
7069 	return -ENOBUFS;
7070 }
7071 
7072 static void __addrconf_sysctl_unregister(struct net *net,
7073 					 struct ipv6_devconf *p, int ifindex)
7074 {
7075 	struct ctl_table *table;
7076 
7077 	if (!p->sysctl_header)
7078 		return;
7079 
7080 	table = p->sysctl_header->ctl_table_arg;
7081 	unregister_net_sysctl_table(p->sysctl_header);
7082 	p->sysctl_header = NULL;
7083 	kfree(table);
7084 
7085 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7086 }
7087 
7088 static int addrconf_sysctl_register(struct inet6_dev *idev)
7089 {
7090 	int err;
7091 
7092 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7093 		return -EINVAL;
7094 
7095 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7096 				    &ndisc_ifinfo_sysctl_change);
7097 	if (err)
7098 		return err;
7099 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7100 					 idev, &idev->cnf);
7101 	if (err)
7102 		neigh_sysctl_unregister(idev->nd_parms);
7103 
7104 	return err;
7105 }
7106 
7107 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7108 {
7109 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7110 				     idev->dev->ifindex);
7111 	neigh_sysctl_unregister(idev->nd_parms);
7112 }
7113 
7114 
7115 #endif
7116 
7117 static int __net_init addrconf_init_net(struct net *net)
7118 {
7119 	int err = -ENOMEM;
7120 	struct ipv6_devconf *all, *dflt;
7121 
7122 	spin_lock_init(&net->ipv6.addrconf_hash_lock);
7123 	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7124 	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7125 					   sizeof(struct hlist_head),
7126 					   GFP_KERNEL);
7127 	if (!net->ipv6.inet6_addr_lst)
7128 		goto err_alloc_addr;
7129 
7130 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7131 	if (!all)
7132 		goto err_alloc_all;
7133 
7134 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7135 	if (!dflt)
7136 		goto err_alloc_dflt;
7137 
7138 	if (IS_ENABLED(CONFIG_SYSCTL) &&
7139 	    !net_eq(net, &init_net)) {
7140 		switch (sysctl_devconf_inherit_init_net) {
7141 		case 1:  /* copy from init_net */
7142 			memcpy(all, init_net.ipv6.devconf_all,
7143 			       sizeof(ipv6_devconf));
7144 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7145 			       sizeof(ipv6_devconf_dflt));
7146 			break;
7147 		case 3: /* copy from the current netns */
7148 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7149 			       sizeof(ipv6_devconf));
7150 			memcpy(dflt,
7151 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7152 			       sizeof(ipv6_devconf_dflt));
7153 			break;
7154 		case 0:
7155 		case 2:
7156 			/* use compiled values */
7157 			break;
7158 		}
7159 	}
7160 
7161 	/* these will be inherited by all namespaces */
7162 	dflt->autoconf = ipv6_defaults.autoconf;
7163 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7164 
7165 	dflt->stable_secret.initialized = false;
7166 	all->stable_secret.initialized = false;
7167 
7168 	net->ipv6.devconf_all = all;
7169 	net->ipv6.devconf_dflt = dflt;
7170 
7171 #ifdef CONFIG_SYSCTL
7172 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7173 	if (err < 0)
7174 		goto err_reg_all;
7175 
7176 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7177 	if (err < 0)
7178 		goto err_reg_dflt;
7179 #endif
7180 	return 0;
7181 
7182 #ifdef CONFIG_SYSCTL
7183 err_reg_dflt:
7184 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7185 err_reg_all:
7186 	kfree(dflt);
7187 #endif
7188 err_alloc_dflt:
7189 	kfree(all);
7190 err_alloc_all:
7191 	kfree(net->ipv6.inet6_addr_lst);
7192 err_alloc_addr:
7193 	return err;
7194 }
7195 
7196 static void __net_exit addrconf_exit_net(struct net *net)
7197 {
7198 	int i;
7199 
7200 #ifdef CONFIG_SYSCTL
7201 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7202 				     NETCONFA_IFINDEX_DEFAULT);
7203 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7204 				     NETCONFA_IFINDEX_ALL);
7205 #endif
7206 	kfree(net->ipv6.devconf_dflt);
7207 	net->ipv6.devconf_dflt = NULL;
7208 	kfree(net->ipv6.devconf_all);
7209 	net->ipv6.devconf_all = NULL;
7210 
7211 	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7212 	/*
7213 	 *	Check hash table, then free it.
7214 	 */
7215 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7216 		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7217 
7218 	kfree(net->ipv6.inet6_addr_lst);
7219 	net->ipv6.inet6_addr_lst = NULL;
7220 }
7221 
7222 static struct pernet_operations addrconf_ops = {
7223 	.init = addrconf_init_net,
7224 	.exit = addrconf_exit_net,
7225 };
7226 
7227 static struct rtnl_af_ops inet6_ops __read_mostly = {
7228 	.family		  = AF_INET6,
7229 	.fill_link_af	  = inet6_fill_link_af,
7230 	.get_link_af_size = inet6_get_link_af_size,
7231 	.validate_link_af = inet6_validate_link_af,
7232 	.set_link_af	  = inet6_set_link_af,
7233 };
7234 
7235 /*
7236  *	Init / cleanup code
7237  */
7238 
7239 int __init addrconf_init(void)
7240 {
7241 	struct inet6_dev *idev;
7242 	int err;
7243 
7244 	err = ipv6_addr_label_init();
7245 	if (err < 0) {
7246 		pr_crit("%s: cannot initialize default policy table: %d\n",
7247 			__func__, err);
7248 		goto out;
7249 	}
7250 
7251 	err = register_pernet_subsys(&addrconf_ops);
7252 	if (err < 0)
7253 		goto out_addrlabel;
7254 
7255 	addrconf_wq = create_workqueue("ipv6_addrconf");
7256 	if (!addrconf_wq) {
7257 		err = -ENOMEM;
7258 		goto out_nowq;
7259 	}
7260 
7261 	rtnl_lock();
7262 	idev = ipv6_add_dev(blackhole_netdev);
7263 	rtnl_unlock();
7264 	if (IS_ERR(idev)) {
7265 		err = PTR_ERR(idev);
7266 		goto errlo;
7267 	}
7268 
7269 	ip6_route_init_special_entries();
7270 
7271 	register_netdevice_notifier(&ipv6_dev_notf);
7272 
7273 	addrconf_verify(&init_net);
7274 
7275 	rtnl_af_register(&inet6_ops);
7276 
7277 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7278 				   NULL, inet6_dump_ifinfo, 0);
7279 	if (err < 0)
7280 		goto errout;
7281 
7282 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7283 				   inet6_rtm_newaddr, NULL, 0);
7284 	if (err < 0)
7285 		goto errout;
7286 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7287 				   inet6_rtm_deladdr, NULL, 0);
7288 	if (err < 0)
7289 		goto errout;
7290 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7291 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7292 				   RTNL_FLAG_DOIT_UNLOCKED);
7293 	if (err < 0)
7294 		goto errout;
7295 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7296 				   NULL, inet6_dump_ifmcaddr, 0);
7297 	if (err < 0)
7298 		goto errout;
7299 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7300 				   NULL, inet6_dump_ifacaddr, 0);
7301 	if (err < 0)
7302 		goto errout;
7303 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7304 				   inet6_netconf_get_devconf,
7305 				   inet6_netconf_dump_devconf,
7306 				   RTNL_FLAG_DOIT_UNLOCKED);
7307 	if (err < 0)
7308 		goto errout;
7309 	err = ipv6_addr_label_rtnl_register();
7310 	if (err < 0)
7311 		goto errout;
7312 
7313 	return 0;
7314 errout:
7315 	rtnl_unregister_all(PF_INET6);
7316 	rtnl_af_unregister(&inet6_ops);
7317 	unregister_netdevice_notifier(&ipv6_dev_notf);
7318 errlo:
7319 	destroy_workqueue(addrconf_wq);
7320 out_nowq:
7321 	unregister_pernet_subsys(&addrconf_ops);
7322 out_addrlabel:
7323 	ipv6_addr_label_cleanup();
7324 out:
7325 	return err;
7326 }
7327 
7328 void addrconf_cleanup(void)
7329 {
7330 	struct net_device *dev;
7331 
7332 	unregister_netdevice_notifier(&ipv6_dev_notf);
7333 	unregister_pernet_subsys(&addrconf_ops);
7334 	ipv6_addr_label_cleanup();
7335 
7336 	rtnl_af_unregister(&inet6_ops);
7337 
7338 	rtnl_lock();
7339 
7340 	/* clean dev list */
7341 	for_each_netdev(&init_net, dev) {
7342 		if (__in6_dev_get(dev) == NULL)
7343 			continue;
7344 		addrconf_ifdown(dev, true);
7345 	}
7346 	addrconf_ifdown(init_net.loopback_dev, true);
7347 
7348 	rtnl_unlock();
7349 
7350 	destroy_workqueue(addrconf_wq);
7351 }
7352