xref: /openbmc/linux/net/ipv6/addrconf.c (revision dba52960)
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  *	Configured unicast address hash table
151  */
152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
153 static DEFINE_SPINLOCK(addrconf_hash_lock);
154 
155 static void addrconf_verify(void);
156 static void addrconf_verify_rtnl(void);
157 static void addrconf_verify_work(struct work_struct *);
158 
159 static struct workqueue_struct *addrconf_wq;
160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
161 
162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
164 
165 static void addrconf_type_change(struct net_device *dev,
166 				 unsigned long event);
167 static int addrconf_ifdown(struct net_device *dev, bool unregister);
168 
169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
170 						  int plen,
171 						  const struct net_device *dev,
172 						  u32 flags, u32 noflags,
173 						  bool no_gw);
174 
175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
176 static void addrconf_dad_work(struct work_struct *w);
177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
178 				   bool send_na);
179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
180 static void addrconf_rs_timer(struct timer_list *t);
181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
183 
184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
185 				struct prefix_info *pinfo);
186 
187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
188 	.forwarding		= 0,
189 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
190 	.mtu6			= IPV6_MIN_MTU,
191 	.accept_ra		= 1,
192 	.accept_redirects	= 1,
193 	.autoconf		= 1,
194 	.force_mld_version	= 0,
195 	.mldv1_unsolicited_report_interval = 10 * HZ,
196 	.mldv2_unsolicited_report_interval = HZ,
197 	.dad_transmits		= 1,
198 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
199 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
200 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
201 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
202 	.use_tempaddr		= 0,
203 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
204 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
205 	.regen_max_retry	= REGEN_MAX_RETRY,
206 	.max_desync_factor	= MAX_DESYNC_FACTOR,
207 	.max_addresses		= IPV6_MAX_ADDRESSES,
208 	.accept_ra_defrtr	= 1,
209 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
210 	.accept_ra_from_local	= 0,
211 	.accept_ra_min_hop_limit= 1,
212 	.accept_ra_pinfo	= 1,
213 #ifdef CONFIG_IPV6_ROUTER_PREF
214 	.accept_ra_rtr_pref	= 1,
215 	.rtr_probe_interval	= 60 * HZ,
216 #ifdef CONFIG_IPV6_ROUTE_INFO
217 	.accept_ra_rt_info_min_plen = 0,
218 	.accept_ra_rt_info_max_plen = 0,
219 #endif
220 #endif
221 	.proxy_ndp		= 0,
222 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
223 	.disable_ipv6		= 0,
224 	.accept_dad		= 0,
225 	.suppress_frag_ndisc	= 1,
226 	.accept_ra_mtu		= 1,
227 	.stable_secret		= {
228 		.initialized = false,
229 	},
230 	.use_oif_addrs_only	= 0,
231 	.ignore_routes_with_linkdown = 0,
232 	.keep_addr_on_down	= 0,
233 	.seg6_enabled		= 0,
234 #ifdef CONFIG_IPV6_SEG6_HMAC
235 	.seg6_require_hmac	= 0,
236 #endif
237 	.enhanced_dad           = 1,
238 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
239 	.disable_policy		= 0,
240 	.rpl_seg_enabled	= 0,
241 	.ioam6_enabled		= 0,
242 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
243 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
244 	.ndisc_evict_nocarrier	= 1,
245 };
246 
247 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
248 	.forwarding		= 0,
249 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
250 	.mtu6			= IPV6_MIN_MTU,
251 	.accept_ra		= 1,
252 	.accept_redirects	= 1,
253 	.autoconf		= 1,
254 	.force_mld_version	= 0,
255 	.mldv1_unsolicited_report_interval = 10 * HZ,
256 	.mldv2_unsolicited_report_interval = HZ,
257 	.dad_transmits		= 1,
258 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
259 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
260 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
261 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
262 	.use_tempaddr		= 0,
263 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
264 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
265 	.regen_max_retry	= REGEN_MAX_RETRY,
266 	.max_desync_factor	= MAX_DESYNC_FACTOR,
267 	.max_addresses		= IPV6_MAX_ADDRESSES,
268 	.accept_ra_defrtr	= 1,
269 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
270 	.accept_ra_from_local	= 0,
271 	.accept_ra_min_hop_limit= 1,
272 	.accept_ra_pinfo	= 1,
273 #ifdef CONFIG_IPV6_ROUTER_PREF
274 	.accept_ra_rtr_pref	= 1,
275 	.rtr_probe_interval	= 60 * HZ,
276 #ifdef CONFIG_IPV6_ROUTE_INFO
277 	.accept_ra_rt_info_min_plen = 0,
278 	.accept_ra_rt_info_max_plen = 0,
279 #endif
280 #endif
281 	.proxy_ndp		= 0,
282 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
283 	.disable_ipv6		= 0,
284 	.accept_dad		= 1,
285 	.suppress_frag_ndisc	= 1,
286 	.accept_ra_mtu		= 1,
287 	.stable_secret		= {
288 		.initialized = false,
289 	},
290 	.use_oif_addrs_only	= 0,
291 	.ignore_routes_with_linkdown = 0,
292 	.keep_addr_on_down	= 0,
293 	.seg6_enabled		= 0,
294 #ifdef CONFIG_IPV6_SEG6_HMAC
295 	.seg6_require_hmac	= 0,
296 #endif
297 	.enhanced_dad           = 1,
298 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
299 	.disable_policy		= 0,
300 	.rpl_seg_enabled	= 0,
301 	.ioam6_enabled		= 0,
302 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
303 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
304 	.ndisc_evict_nocarrier	= 1,
305 };
306 
307 /* Check if link is ready: is it up and is a valid qdisc available */
308 static inline bool addrconf_link_ready(const struct net_device *dev)
309 {
310 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
311 }
312 
313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
314 {
315 	if (del_timer(&idev->rs_timer))
316 		__in6_dev_put(idev);
317 }
318 
319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
320 {
321 	if (cancel_delayed_work(&ifp->dad_work))
322 		__in6_ifa_put(ifp);
323 }
324 
325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
326 				  unsigned long when)
327 {
328 	if (!timer_pending(&idev->rs_timer))
329 		in6_dev_hold(idev);
330 	mod_timer(&idev->rs_timer, jiffies + when);
331 }
332 
333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
334 				   unsigned long delay)
335 {
336 	in6_ifa_hold(ifp);
337 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
338 		in6_ifa_put(ifp);
339 }
340 
341 static int snmp6_alloc_dev(struct inet6_dev *idev)
342 {
343 	int i;
344 
345 	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
346 	if (!idev->stats.ipv6)
347 		goto err_ip;
348 
349 	for_each_possible_cpu(i) {
350 		struct ipstats_mib *addrconf_stats;
351 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
352 		u64_stats_init(&addrconf_stats->syncp);
353 	}
354 
355 
356 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
357 					GFP_KERNEL);
358 	if (!idev->stats.icmpv6dev)
359 		goto err_icmp;
360 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
361 					   GFP_KERNEL);
362 	if (!idev->stats.icmpv6msgdev)
363 		goto err_icmpmsg;
364 
365 	return 0;
366 
367 err_icmpmsg:
368 	kfree(idev->stats.icmpv6dev);
369 err_icmp:
370 	free_percpu(idev->stats.ipv6);
371 err_ip:
372 	return -ENOMEM;
373 }
374 
375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
376 {
377 	struct inet6_dev *ndev;
378 	int err = -ENOMEM;
379 
380 	ASSERT_RTNL();
381 
382 	if (dev->mtu < IPV6_MIN_MTU)
383 		return ERR_PTR(-EINVAL);
384 
385 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
386 	if (!ndev)
387 		return ERR_PTR(err);
388 
389 	rwlock_init(&ndev->lock);
390 	ndev->dev = dev;
391 	INIT_LIST_HEAD(&ndev->addr_list);
392 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
393 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
394 
395 	if (ndev->cnf.stable_secret.initialized)
396 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
397 
398 	ndev->cnf.mtu6 = dev->mtu;
399 	ndev->ra_mtu = 0;
400 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
401 	if (!ndev->nd_parms) {
402 		kfree(ndev);
403 		return ERR_PTR(err);
404 	}
405 	if (ndev->cnf.forwarding)
406 		dev_disable_lro(dev);
407 	/* We refer to the device */
408 	dev_hold_track(dev, &ndev->dev_tracker, GFP_KERNEL);
409 
410 	if (snmp6_alloc_dev(ndev) < 0) {
411 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
412 			   __func__);
413 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
414 		dev_put_track(dev, &ndev->dev_tracker);
415 		kfree(ndev);
416 		return ERR_PTR(err);
417 	}
418 
419 	if (snmp6_register_dev(ndev) < 0) {
420 		netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
421 			   __func__, dev->name);
422 		goto err_release;
423 	}
424 
425 	/* One reference from device. */
426 	refcount_set(&ndev->refcnt, 1);
427 
428 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
429 		ndev->cnf.accept_dad = -1;
430 
431 #if IS_ENABLED(CONFIG_IPV6_SIT)
432 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
433 		pr_info("%s: Disabled Multicast RS\n", dev->name);
434 		ndev->cnf.rtr_solicits = 0;
435 	}
436 #endif
437 
438 	INIT_LIST_HEAD(&ndev->tempaddr_list);
439 	ndev->desync_factor = U32_MAX;
440 	if ((dev->flags&IFF_LOOPBACK) ||
441 	    dev->type == ARPHRD_TUNNEL ||
442 	    dev->type == ARPHRD_TUNNEL6 ||
443 	    dev->type == ARPHRD_SIT ||
444 	    dev->type == ARPHRD_NONE) {
445 		ndev->cnf.use_tempaddr = -1;
446 	}
447 
448 	ndev->token = in6addr_any;
449 
450 	if (netif_running(dev) && addrconf_link_ready(dev))
451 		ndev->if_flags |= IF_READY;
452 
453 	ipv6_mc_init_dev(ndev);
454 	ndev->tstamp = jiffies;
455 	err = addrconf_sysctl_register(ndev);
456 	if (err) {
457 		ipv6_mc_destroy_dev(ndev);
458 		snmp6_unregister_dev(ndev);
459 		goto err_release;
460 	}
461 	/* protected by rtnl_lock */
462 	rcu_assign_pointer(dev->ip6_ptr, ndev);
463 
464 	/* Join interface-local all-node multicast group */
465 	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
466 
467 	/* Join all-node multicast group */
468 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
469 
470 	/* Join all-router multicast group if forwarding is set */
471 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
472 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
473 
474 	return ndev;
475 
476 err_release:
477 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
478 	ndev->dead = 1;
479 	in6_dev_finish_destroy(ndev);
480 	return ERR_PTR(err);
481 }
482 
483 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
484 {
485 	struct inet6_dev *idev;
486 
487 	ASSERT_RTNL();
488 
489 	idev = __in6_dev_get(dev);
490 	if (!idev) {
491 		idev = ipv6_add_dev(dev);
492 		if (IS_ERR(idev))
493 			return idev;
494 	}
495 
496 	if (dev->flags&IFF_UP)
497 		ipv6_mc_up(idev);
498 	return idev;
499 }
500 
501 static int inet6_netconf_msgsize_devconf(int type)
502 {
503 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
504 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
505 	bool all = false;
506 
507 	if (type == NETCONFA_ALL)
508 		all = true;
509 
510 	if (all || type == NETCONFA_FORWARDING)
511 		size += nla_total_size(4);
512 #ifdef CONFIG_IPV6_MROUTE
513 	if (all || type == NETCONFA_MC_FORWARDING)
514 		size += nla_total_size(4);
515 #endif
516 	if (all || type == NETCONFA_PROXY_NEIGH)
517 		size += nla_total_size(4);
518 
519 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
520 		size += nla_total_size(4);
521 
522 	return size;
523 }
524 
525 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
526 				      struct ipv6_devconf *devconf, u32 portid,
527 				      u32 seq, int event, unsigned int flags,
528 				      int type)
529 {
530 	struct nlmsghdr  *nlh;
531 	struct netconfmsg *ncm;
532 	bool all = false;
533 
534 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
535 			flags);
536 	if (!nlh)
537 		return -EMSGSIZE;
538 
539 	if (type == NETCONFA_ALL)
540 		all = true;
541 
542 	ncm = nlmsg_data(nlh);
543 	ncm->ncm_family = AF_INET6;
544 
545 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
546 		goto nla_put_failure;
547 
548 	if (!devconf)
549 		goto out;
550 
551 	if ((all || type == NETCONFA_FORWARDING) &&
552 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
553 		goto nla_put_failure;
554 #ifdef CONFIG_IPV6_MROUTE
555 	if ((all || type == NETCONFA_MC_FORWARDING) &&
556 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
557 			devconf->mc_forwarding) < 0)
558 		goto nla_put_failure;
559 #endif
560 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
561 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
562 		goto nla_put_failure;
563 
564 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
565 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
566 			devconf->ignore_routes_with_linkdown) < 0)
567 		goto nla_put_failure;
568 
569 out:
570 	nlmsg_end(skb, nlh);
571 	return 0;
572 
573 nla_put_failure:
574 	nlmsg_cancel(skb, nlh);
575 	return -EMSGSIZE;
576 }
577 
578 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
579 				  int ifindex, struct ipv6_devconf *devconf)
580 {
581 	struct sk_buff *skb;
582 	int err = -ENOBUFS;
583 
584 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
585 	if (!skb)
586 		goto errout;
587 
588 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
589 					 event, 0, type);
590 	if (err < 0) {
591 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
592 		WARN_ON(err == -EMSGSIZE);
593 		kfree_skb(skb);
594 		goto errout;
595 	}
596 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
597 	return;
598 errout:
599 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
600 }
601 
602 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
603 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
604 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
605 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
606 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
607 };
608 
609 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
610 				       const struct nlmsghdr *nlh,
611 				       struct nlattr **tb,
612 				       struct netlink_ext_ack *extack)
613 {
614 	int i, err;
615 
616 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
617 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
618 		return -EINVAL;
619 	}
620 
621 	if (!netlink_strict_get_check(skb))
622 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
623 					      tb, NETCONFA_MAX,
624 					      devconf_ipv6_policy, extack);
625 
626 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
627 					    tb, NETCONFA_MAX,
628 					    devconf_ipv6_policy, extack);
629 	if (err)
630 		return err;
631 
632 	for (i = 0; i <= NETCONFA_MAX; i++) {
633 		if (!tb[i])
634 			continue;
635 
636 		switch (i) {
637 		case NETCONFA_IFINDEX:
638 			break;
639 		default:
640 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
641 			return -EINVAL;
642 		}
643 	}
644 
645 	return 0;
646 }
647 
648 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
649 				     struct nlmsghdr *nlh,
650 				     struct netlink_ext_ack *extack)
651 {
652 	struct net *net = sock_net(in_skb->sk);
653 	struct nlattr *tb[NETCONFA_MAX+1];
654 	struct inet6_dev *in6_dev = NULL;
655 	struct net_device *dev = NULL;
656 	struct sk_buff *skb;
657 	struct ipv6_devconf *devconf;
658 	int ifindex;
659 	int err;
660 
661 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
662 	if (err < 0)
663 		return err;
664 
665 	if (!tb[NETCONFA_IFINDEX])
666 		return -EINVAL;
667 
668 	err = -EINVAL;
669 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
670 	switch (ifindex) {
671 	case NETCONFA_IFINDEX_ALL:
672 		devconf = net->ipv6.devconf_all;
673 		break;
674 	case NETCONFA_IFINDEX_DEFAULT:
675 		devconf = net->ipv6.devconf_dflt;
676 		break;
677 	default:
678 		dev = dev_get_by_index(net, ifindex);
679 		if (!dev)
680 			return -EINVAL;
681 		in6_dev = in6_dev_get(dev);
682 		if (!in6_dev)
683 			goto errout;
684 		devconf = &in6_dev->cnf;
685 		break;
686 	}
687 
688 	err = -ENOBUFS;
689 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
690 	if (!skb)
691 		goto errout;
692 
693 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
694 					 NETLINK_CB(in_skb).portid,
695 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
696 					 NETCONFA_ALL);
697 	if (err < 0) {
698 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
699 		WARN_ON(err == -EMSGSIZE);
700 		kfree_skb(skb);
701 		goto errout;
702 	}
703 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
704 errout:
705 	if (in6_dev)
706 		in6_dev_put(in6_dev);
707 	dev_put(dev);
708 	return err;
709 }
710 
711 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
712 				      struct netlink_callback *cb)
713 {
714 	const struct nlmsghdr *nlh = cb->nlh;
715 	struct net *net = sock_net(skb->sk);
716 	int h, s_h;
717 	int idx, s_idx;
718 	struct net_device *dev;
719 	struct inet6_dev *idev;
720 	struct hlist_head *head;
721 
722 	if (cb->strict_check) {
723 		struct netlink_ext_ack *extack = cb->extack;
724 		struct netconfmsg *ncm;
725 
726 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
727 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
728 			return -EINVAL;
729 		}
730 
731 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
732 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
733 			return -EINVAL;
734 		}
735 	}
736 
737 	s_h = cb->args[0];
738 	s_idx = idx = cb->args[1];
739 
740 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
741 		idx = 0;
742 		head = &net->dev_index_head[h];
743 		rcu_read_lock();
744 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
745 			  net->dev_base_seq;
746 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
747 			if (idx < s_idx)
748 				goto cont;
749 			idev = __in6_dev_get(dev);
750 			if (!idev)
751 				goto cont;
752 
753 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
754 						       &idev->cnf,
755 						       NETLINK_CB(cb->skb).portid,
756 						       nlh->nlmsg_seq,
757 						       RTM_NEWNETCONF,
758 						       NLM_F_MULTI,
759 						       NETCONFA_ALL) < 0) {
760 				rcu_read_unlock();
761 				goto done;
762 			}
763 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
764 cont:
765 			idx++;
766 		}
767 		rcu_read_unlock();
768 	}
769 	if (h == NETDEV_HASHENTRIES) {
770 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
771 					       net->ipv6.devconf_all,
772 					       NETLINK_CB(cb->skb).portid,
773 					       nlh->nlmsg_seq,
774 					       RTM_NEWNETCONF, NLM_F_MULTI,
775 					       NETCONFA_ALL) < 0)
776 			goto done;
777 		else
778 			h++;
779 	}
780 	if (h == NETDEV_HASHENTRIES + 1) {
781 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
782 					       net->ipv6.devconf_dflt,
783 					       NETLINK_CB(cb->skb).portid,
784 					       nlh->nlmsg_seq,
785 					       RTM_NEWNETCONF, NLM_F_MULTI,
786 					       NETCONFA_ALL) < 0)
787 			goto done;
788 		else
789 			h++;
790 	}
791 done:
792 	cb->args[0] = h;
793 	cb->args[1] = idx;
794 
795 	return skb->len;
796 }
797 
798 #ifdef CONFIG_SYSCTL
799 static void dev_forward_change(struct inet6_dev *idev)
800 {
801 	struct net_device *dev;
802 	struct inet6_ifaddr *ifa;
803 
804 	if (!idev)
805 		return;
806 	dev = idev->dev;
807 	if (idev->cnf.forwarding)
808 		dev_disable_lro(dev);
809 	if (dev->flags & IFF_MULTICAST) {
810 		if (idev->cnf.forwarding) {
811 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
812 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
813 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
814 		} else {
815 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
816 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
817 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
818 		}
819 	}
820 
821 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
822 		if (ifa->flags&IFA_F_TENTATIVE)
823 			continue;
824 		if (idev->cnf.forwarding)
825 			addrconf_join_anycast(ifa);
826 		else
827 			addrconf_leave_anycast(ifa);
828 	}
829 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
830 				     NETCONFA_FORWARDING,
831 				     dev->ifindex, &idev->cnf);
832 }
833 
834 
835 static void addrconf_forward_change(struct net *net, __s32 newf)
836 {
837 	struct net_device *dev;
838 	struct inet6_dev *idev;
839 
840 	for_each_netdev(net, dev) {
841 		idev = __in6_dev_get(dev);
842 		if (idev) {
843 			int changed = (!idev->cnf.forwarding) ^ (!newf);
844 			idev->cnf.forwarding = newf;
845 			if (changed)
846 				dev_forward_change(idev);
847 		}
848 	}
849 }
850 
851 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
852 {
853 	struct net *net;
854 	int old;
855 
856 	if (!rtnl_trylock())
857 		return restart_syscall();
858 
859 	net = (struct net *)table->extra2;
860 	old = *p;
861 	*p = newf;
862 
863 	if (p == &net->ipv6.devconf_dflt->forwarding) {
864 		if ((!newf) ^ (!old))
865 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
866 						     NETCONFA_FORWARDING,
867 						     NETCONFA_IFINDEX_DEFAULT,
868 						     net->ipv6.devconf_dflt);
869 		rtnl_unlock();
870 		return 0;
871 	}
872 
873 	if (p == &net->ipv6.devconf_all->forwarding) {
874 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
875 
876 		net->ipv6.devconf_dflt->forwarding = newf;
877 		if ((!newf) ^ (!old_dflt))
878 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
879 						     NETCONFA_FORWARDING,
880 						     NETCONFA_IFINDEX_DEFAULT,
881 						     net->ipv6.devconf_dflt);
882 
883 		addrconf_forward_change(net, newf);
884 		if ((!newf) ^ (!old))
885 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
886 						     NETCONFA_FORWARDING,
887 						     NETCONFA_IFINDEX_ALL,
888 						     net->ipv6.devconf_all);
889 	} else if ((!newf) ^ (!old))
890 		dev_forward_change((struct inet6_dev *)table->extra1);
891 	rtnl_unlock();
892 
893 	if (newf)
894 		rt6_purge_dflt_routers(net);
895 	return 1;
896 }
897 
898 static void addrconf_linkdown_change(struct net *net, __s32 newf)
899 {
900 	struct net_device *dev;
901 	struct inet6_dev *idev;
902 
903 	for_each_netdev(net, dev) {
904 		idev = __in6_dev_get(dev);
905 		if (idev) {
906 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
907 
908 			idev->cnf.ignore_routes_with_linkdown = newf;
909 			if (changed)
910 				inet6_netconf_notify_devconf(dev_net(dev),
911 							     RTM_NEWNETCONF,
912 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
913 							     dev->ifindex,
914 							     &idev->cnf);
915 		}
916 	}
917 }
918 
919 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
920 {
921 	struct net *net;
922 	int old;
923 
924 	if (!rtnl_trylock())
925 		return restart_syscall();
926 
927 	net = (struct net *)table->extra2;
928 	old = *p;
929 	*p = newf;
930 
931 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
932 		if ((!newf) ^ (!old))
933 			inet6_netconf_notify_devconf(net,
934 						     RTM_NEWNETCONF,
935 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
936 						     NETCONFA_IFINDEX_DEFAULT,
937 						     net->ipv6.devconf_dflt);
938 		rtnl_unlock();
939 		return 0;
940 	}
941 
942 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
943 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
944 		addrconf_linkdown_change(net, newf);
945 		if ((!newf) ^ (!old))
946 			inet6_netconf_notify_devconf(net,
947 						     RTM_NEWNETCONF,
948 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
949 						     NETCONFA_IFINDEX_ALL,
950 						     net->ipv6.devconf_all);
951 	}
952 	rtnl_unlock();
953 
954 	return 1;
955 }
956 
957 #endif
958 
959 /* Nobody refers to this ifaddr, destroy it */
960 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
961 {
962 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
963 
964 #ifdef NET_REFCNT_DEBUG
965 	pr_debug("%s\n", __func__);
966 #endif
967 
968 	in6_dev_put(ifp->idev);
969 
970 	if (cancel_delayed_work(&ifp->dad_work))
971 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
972 			  ifp);
973 
974 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
975 		pr_warn("Freeing alive inet6 address %p\n", ifp);
976 		return;
977 	}
978 
979 	kfree_rcu(ifp, rcu);
980 }
981 
982 static void
983 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
984 {
985 	struct list_head *p;
986 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
987 
988 	/*
989 	 * Each device address list is sorted in order of scope -
990 	 * global before linklocal.
991 	 */
992 	list_for_each(p, &idev->addr_list) {
993 		struct inet6_ifaddr *ifa
994 			= list_entry(p, struct inet6_ifaddr, if_list);
995 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
996 			break;
997 	}
998 
999 	list_add_tail_rcu(&ifp->if_list, p);
1000 }
1001 
1002 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1003 {
1004 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1005 
1006 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1007 }
1008 
1009 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1010 			       struct net_device *dev, unsigned int hash)
1011 {
1012 	struct inet6_ifaddr *ifp;
1013 
1014 	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1015 		if (!net_eq(dev_net(ifp->idev->dev), net))
1016 			continue;
1017 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1018 			if (!dev || ifp->idev->dev == dev)
1019 				return true;
1020 		}
1021 	}
1022 	return false;
1023 }
1024 
1025 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1026 {
1027 	unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1028 	int err = 0;
1029 
1030 	spin_lock(&addrconf_hash_lock);
1031 
1032 	/* Ignore adding duplicate addresses on an interface */
1033 	if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1034 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1035 		err = -EEXIST;
1036 	} else {
1037 		hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1038 	}
1039 
1040 	spin_unlock(&addrconf_hash_lock);
1041 
1042 	return err;
1043 }
1044 
1045 /* On success it returns ifp with increased reference count */
1046 
1047 static struct inet6_ifaddr *
1048 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1049 	      bool can_block, struct netlink_ext_ack *extack)
1050 {
1051 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1052 	int addr_type = ipv6_addr_type(cfg->pfx);
1053 	struct net *net = dev_net(idev->dev);
1054 	struct inet6_ifaddr *ifa = NULL;
1055 	struct fib6_info *f6i = NULL;
1056 	int err = 0;
1057 
1058 	if (addr_type == IPV6_ADDR_ANY ||
1059 	    (addr_type & IPV6_ADDR_MULTICAST &&
1060 	     !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1061 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1062 	     !netif_is_l3_master(idev->dev) &&
1063 	     addr_type & IPV6_ADDR_LOOPBACK))
1064 		return ERR_PTR(-EADDRNOTAVAIL);
1065 
1066 	if (idev->dead) {
1067 		err = -ENODEV;			/*XXX*/
1068 		goto out;
1069 	}
1070 
1071 	if (idev->cnf.disable_ipv6) {
1072 		err = -EACCES;
1073 		goto out;
1074 	}
1075 
1076 	/* validator notifier needs to be blocking;
1077 	 * do not call in atomic context
1078 	 */
1079 	if (can_block) {
1080 		struct in6_validator_info i6vi = {
1081 			.i6vi_addr = *cfg->pfx,
1082 			.i6vi_dev = idev,
1083 			.extack = extack,
1084 		};
1085 
1086 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1087 		err = notifier_to_errno(err);
1088 		if (err < 0)
1089 			goto out;
1090 	}
1091 
1092 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1093 	if (!ifa) {
1094 		err = -ENOBUFS;
1095 		goto out;
1096 	}
1097 
1098 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1099 	if (IS_ERR(f6i)) {
1100 		err = PTR_ERR(f6i);
1101 		f6i = NULL;
1102 		goto out;
1103 	}
1104 
1105 	if (net->ipv6.devconf_all->disable_policy ||
1106 	    idev->cnf.disable_policy)
1107 		f6i->dst_nopolicy = true;
1108 
1109 	neigh_parms_data_state_setall(idev->nd_parms);
1110 
1111 	ifa->addr = *cfg->pfx;
1112 	if (cfg->peer_pfx)
1113 		ifa->peer_addr = *cfg->peer_pfx;
1114 
1115 	spin_lock_init(&ifa->lock);
1116 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1117 	INIT_HLIST_NODE(&ifa->addr_lst);
1118 	ifa->scope = cfg->scope;
1119 	ifa->prefix_len = cfg->plen;
1120 	ifa->rt_priority = cfg->rt_priority;
1121 	ifa->flags = cfg->ifa_flags;
1122 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1123 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1124 		ifa->flags |= IFA_F_TENTATIVE;
1125 	ifa->valid_lft = cfg->valid_lft;
1126 	ifa->prefered_lft = cfg->preferred_lft;
1127 	ifa->cstamp = ifa->tstamp = jiffies;
1128 	ifa->tokenized = false;
1129 
1130 	ifa->rt = f6i;
1131 
1132 	ifa->idev = idev;
1133 	in6_dev_hold(idev);
1134 
1135 	/* For caller */
1136 	refcount_set(&ifa->refcnt, 1);
1137 
1138 	rcu_read_lock_bh();
1139 
1140 	err = ipv6_add_addr_hash(idev->dev, ifa);
1141 	if (err < 0) {
1142 		rcu_read_unlock_bh();
1143 		goto out;
1144 	}
1145 
1146 	write_lock(&idev->lock);
1147 
1148 	/* Add to inet6_dev unicast addr list. */
1149 	ipv6_link_dev_addr(idev, ifa);
1150 
1151 	if (ifa->flags&IFA_F_TEMPORARY) {
1152 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1153 		in6_ifa_hold(ifa);
1154 	}
1155 
1156 	in6_ifa_hold(ifa);
1157 	write_unlock(&idev->lock);
1158 
1159 	rcu_read_unlock_bh();
1160 
1161 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1162 out:
1163 	if (unlikely(err < 0)) {
1164 		fib6_info_release(f6i);
1165 
1166 		if (ifa) {
1167 			if (ifa->idev)
1168 				in6_dev_put(ifa->idev);
1169 			kfree(ifa);
1170 		}
1171 		ifa = ERR_PTR(err);
1172 	}
1173 
1174 	return ifa;
1175 }
1176 
1177 enum cleanup_prefix_rt_t {
1178 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1179 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1180 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1181 };
1182 
1183 /*
1184  * Check, whether the prefix for ifp would still need a prefix route
1185  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1186  * constants.
1187  *
1188  * 1) we don't purge prefix if address was not permanent.
1189  *    prefix is managed by its own lifetime.
1190  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1191  * 3) if there are no addresses, delete prefix.
1192  * 4) if there are still other permanent address(es),
1193  *    corresponding prefix is still permanent.
1194  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1195  *    don't purge the prefix, assume user space is managing it.
1196  * 6) otherwise, update prefix lifetime to the
1197  *    longest valid lifetime among the corresponding
1198  *    addresses on the device.
1199  *    Note: subsequent RA will update lifetime.
1200  **/
1201 static enum cleanup_prefix_rt_t
1202 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1203 {
1204 	struct inet6_ifaddr *ifa;
1205 	struct inet6_dev *idev = ifp->idev;
1206 	unsigned long lifetime;
1207 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1208 
1209 	*expires = jiffies;
1210 
1211 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1212 		if (ifa == ifp)
1213 			continue;
1214 		if (ifa->prefix_len != ifp->prefix_len ||
1215 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1216 				       ifp->prefix_len))
1217 			continue;
1218 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1219 			return CLEANUP_PREFIX_RT_NOP;
1220 
1221 		action = CLEANUP_PREFIX_RT_EXPIRE;
1222 
1223 		spin_lock(&ifa->lock);
1224 
1225 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1226 		/*
1227 		 * Note: Because this address is
1228 		 * not permanent, lifetime <
1229 		 * LONG_MAX / HZ here.
1230 		 */
1231 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1232 			*expires = ifa->tstamp + lifetime * HZ;
1233 		spin_unlock(&ifa->lock);
1234 	}
1235 
1236 	return action;
1237 }
1238 
1239 static void
1240 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1241 		     bool del_rt, bool del_peer)
1242 {
1243 	struct fib6_info *f6i;
1244 
1245 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1246 					ifp->prefix_len,
1247 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1248 	if (f6i) {
1249 		if (del_rt)
1250 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1251 		else {
1252 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1253 				fib6_set_expires(f6i, expires);
1254 			fib6_info_release(f6i);
1255 		}
1256 	}
1257 }
1258 
1259 
1260 /* This function wants to get referenced ifp and releases it before return */
1261 
1262 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1263 {
1264 	int state;
1265 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1266 	unsigned long expires;
1267 
1268 	ASSERT_RTNL();
1269 
1270 	spin_lock_bh(&ifp->lock);
1271 	state = ifp->state;
1272 	ifp->state = INET6_IFADDR_STATE_DEAD;
1273 	spin_unlock_bh(&ifp->lock);
1274 
1275 	if (state == INET6_IFADDR_STATE_DEAD)
1276 		goto out;
1277 
1278 	spin_lock_bh(&addrconf_hash_lock);
1279 	hlist_del_init_rcu(&ifp->addr_lst);
1280 	spin_unlock_bh(&addrconf_hash_lock);
1281 
1282 	write_lock_bh(&ifp->idev->lock);
1283 
1284 	if (ifp->flags&IFA_F_TEMPORARY) {
1285 		list_del(&ifp->tmp_list);
1286 		if (ifp->ifpub) {
1287 			in6_ifa_put(ifp->ifpub);
1288 			ifp->ifpub = NULL;
1289 		}
1290 		__in6_ifa_put(ifp);
1291 	}
1292 
1293 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1294 		action = check_cleanup_prefix_route(ifp, &expires);
1295 
1296 	list_del_rcu(&ifp->if_list);
1297 	__in6_ifa_put(ifp);
1298 
1299 	write_unlock_bh(&ifp->idev->lock);
1300 
1301 	addrconf_del_dad_work(ifp);
1302 
1303 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1304 
1305 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1306 
1307 	if (action != CLEANUP_PREFIX_RT_NOP) {
1308 		cleanup_prefix_route(ifp, expires,
1309 			action == CLEANUP_PREFIX_RT_DEL, false);
1310 	}
1311 
1312 	/* clean up prefsrc entries */
1313 	rt6_remove_prefsrc(ifp);
1314 out:
1315 	in6_ifa_put(ifp);
1316 }
1317 
1318 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1319 {
1320 	struct inet6_dev *idev = ifp->idev;
1321 	unsigned long tmp_tstamp, age;
1322 	unsigned long regen_advance;
1323 	unsigned long now = jiffies;
1324 	s32 cnf_temp_preferred_lft;
1325 	struct inet6_ifaddr *ift;
1326 	struct ifa6_config cfg;
1327 	long max_desync_factor;
1328 	struct in6_addr addr;
1329 	int ret = 0;
1330 
1331 	write_lock_bh(&idev->lock);
1332 
1333 retry:
1334 	in6_dev_hold(idev);
1335 	if (idev->cnf.use_tempaddr <= 0) {
1336 		write_unlock_bh(&idev->lock);
1337 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1338 		in6_dev_put(idev);
1339 		ret = -1;
1340 		goto out;
1341 	}
1342 	spin_lock_bh(&ifp->lock);
1343 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1344 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1345 		spin_unlock_bh(&ifp->lock);
1346 		write_unlock_bh(&idev->lock);
1347 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1348 			__func__);
1349 		in6_dev_put(idev);
1350 		ret = -1;
1351 		goto out;
1352 	}
1353 	in6_ifa_hold(ifp);
1354 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1355 	ipv6_gen_rnd_iid(&addr);
1356 
1357 	age = (now - ifp->tstamp) / HZ;
1358 
1359 	regen_advance = idev->cnf.regen_max_retry *
1360 			idev->cnf.dad_transmits *
1361 			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1362 
1363 	/* recalculate max_desync_factor each time and update
1364 	 * idev->desync_factor if it's larger
1365 	 */
1366 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1367 	max_desync_factor = min_t(__u32,
1368 				  idev->cnf.max_desync_factor,
1369 				  cnf_temp_preferred_lft - regen_advance);
1370 
1371 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1372 		if (max_desync_factor > 0) {
1373 			get_random_bytes(&idev->desync_factor,
1374 					 sizeof(idev->desync_factor));
1375 			idev->desync_factor %= max_desync_factor;
1376 		} else {
1377 			idev->desync_factor = 0;
1378 		}
1379 	}
1380 
1381 	memset(&cfg, 0, sizeof(cfg));
1382 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1383 			      idev->cnf.temp_valid_lft + age);
1384 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1385 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1386 
1387 	cfg.plen = ifp->prefix_len;
1388 	tmp_tstamp = ifp->tstamp;
1389 	spin_unlock_bh(&ifp->lock);
1390 
1391 	write_unlock_bh(&idev->lock);
1392 
1393 	/* A temporary address is created only if this calculated Preferred
1394 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1395 	 * an implementation must not create a temporary address with a zero
1396 	 * Preferred Lifetime.
1397 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1398 	 * temporary addresses being generated.
1399 	 */
1400 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1401 	if (cfg.preferred_lft <= regen_advance + age) {
1402 		in6_ifa_put(ifp);
1403 		in6_dev_put(idev);
1404 		ret = -1;
1405 		goto out;
1406 	}
1407 
1408 	cfg.ifa_flags = IFA_F_TEMPORARY;
1409 	/* set in addrconf_prefix_rcv() */
1410 	if (ifp->flags & IFA_F_OPTIMISTIC)
1411 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1412 
1413 	cfg.pfx = &addr;
1414 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1415 
1416 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1417 	if (IS_ERR(ift)) {
1418 		in6_ifa_put(ifp);
1419 		in6_dev_put(idev);
1420 		pr_info("%s: retry temporary address regeneration\n", __func__);
1421 		write_lock_bh(&idev->lock);
1422 		goto retry;
1423 	}
1424 
1425 	spin_lock_bh(&ift->lock);
1426 	ift->ifpub = ifp;
1427 	ift->cstamp = now;
1428 	ift->tstamp = tmp_tstamp;
1429 	spin_unlock_bh(&ift->lock);
1430 
1431 	addrconf_dad_start(ift);
1432 	in6_ifa_put(ift);
1433 	in6_dev_put(idev);
1434 out:
1435 	return ret;
1436 }
1437 
1438 /*
1439  *	Choose an appropriate source address (RFC3484)
1440  */
1441 enum {
1442 	IPV6_SADDR_RULE_INIT = 0,
1443 	IPV6_SADDR_RULE_LOCAL,
1444 	IPV6_SADDR_RULE_SCOPE,
1445 	IPV6_SADDR_RULE_PREFERRED,
1446 #ifdef CONFIG_IPV6_MIP6
1447 	IPV6_SADDR_RULE_HOA,
1448 #endif
1449 	IPV6_SADDR_RULE_OIF,
1450 	IPV6_SADDR_RULE_LABEL,
1451 	IPV6_SADDR_RULE_PRIVACY,
1452 	IPV6_SADDR_RULE_ORCHID,
1453 	IPV6_SADDR_RULE_PREFIX,
1454 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1455 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1456 #endif
1457 	IPV6_SADDR_RULE_MAX
1458 };
1459 
1460 struct ipv6_saddr_score {
1461 	int			rule;
1462 	int			addr_type;
1463 	struct inet6_ifaddr	*ifa;
1464 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1465 	int			scopedist;
1466 	int			matchlen;
1467 };
1468 
1469 struct ipv6_saddr_dst {
1470 	const struct in6_addr *addr;
1471 	int ifindex;
1472 	int scope;
1473 	int label;
1474 	unsigned int prefs;
1475 };
1476 
1477 static inline int ipv6_saddr_preferred(int type)
1478 {
1479 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1480 		return 1;
1481 	return 0;
1482 }
1483 
1484 static bool ipv6_use_optimistic_addr(struct net *net,
1485 				     struct inet6_dev *idev)
1486 {
1487 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1488 	if (!idev)
1489 		return false;
1490 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1491 		return false;
1492 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1493 		return false;
1494 
1495 	return true;
1496 #else
1497 	return false;
1498 #endif
1499 }
1500 
1501 static bool ipv6_allow_optimistic_dad(struct net *net,
1502 				      struct inet6_dev *idev)
1503 {
1504 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1505 	if (!idev)
1506 		return false;
1507 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1508 		return false;
1509 
1510 	return true;
1511 #else
1512 	return false;
1513 #endif
1514 }
1515 
1516 static int ipv6_get_saddr_eval(struct net *net,
1517 			       struct ipv6_saddr_score *score,
1518 			       struct ipv6_saddr_dst *dst,
1519 			       int i)
1520 {
1521 	int ret;
1522 
1523 	if (i <= score->rule) {
1524 		switch (i) {
1525 		case IPV6_SADDR_RULE_SCOPE:
1526 			ret = score->scopedist;
1527 			break;
1528 		case IPV6_SADDR_RULE_PREFIX:
1529 			ret = score->matchlen;
1530 			break;
1531 		default:
1532 			ret = !!test_bit(i, score->scorebits);
1533 		}
1534 		goto out;
1535 	}
1536 
1537 	switch (i) {
1538 	case IPV6_SADDR_RULE_INIT:
1539 		/* Rule 0: remember if hiscore is not ready yet */
1540 		ret = !!score->ifa;
1541 		break;
1542 	case IPV6_SADDR_RULE_LOCAL:
1543 		/* Rule 1: Prefer same address */
1544 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1545 		break;
1546 	case IPV6_SADDR_RULE_SCOPE:
1547 		/* Rule 2: Prefer appropriate scope
1548 		 *
1549 		 *      ret
1550 		 *       ^
1551 		 *    -1 |  d 15
1552 		 *    ---+--+-+---> scope
1553 		 *       |
1554 		 *       |             d is scope of the destination.
1555 		 *  B-d  |  \
1556 		 *       |   \      <- smaller scope is better if
1557 		 *  B-15 |    \        if scope is enough for destination.
1558 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1559 		 * d-C-1 | /
1560 		 *       |/         <- greater is better
1561 		 *   -C  /             if scope is not enough for destination.
1562 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1563 		 *
1564 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1565 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1566 		 * Assume B = 0 and we get C > 29.
1567 		 */
1568 		ret = __ipv6_addr_src_scope(score->addr_type);
1569 		if (ret >= dst->scope)
1570 			ret = -ret;
1571 		else
1572 			ret -= 128;	/* 30 is enough */
1573 		score->scopedist = ret;
1574 		break;
1575 	case IPV6_SADDR_RULE_PREFERRED:
1576 	    {
1577 		/* Rule 3: Avoid deprecated and optimistic addresses */
1578 		u8 avoid = IFA_F_DEPRECATED;
1579 
1580 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1581 			avoid |= IFA_F_OPTIMISTIC;
1582 		ret = ipv6_saddr_preferred(score->addr_type) ||
1583 		      !(score->ifa->flags & avoid);
1584 		break;
1585 	    }
1586 #ifdef CONFIG_IPV6_MIP6
1587 	case IPV6_SADDR_RULE_HOA:
1588 	    {
1589 		/* Rule 4: Prefer home address */
1590 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1591 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1592 		break;
1593 	    }
1594 #endif
1595 	case IPV6_SADDR_RULE_OIF:
1596 		/* Rule 5: Prefer outgoing interface */
1597 		ret = (!dst->ifindex ||
1598 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1599 		break;
1600 	case IPV6_SADDR_RULE_LABEL:
1601 		/* Rule 6: Prefer matching label */
1602 		ret = ipv6_addr_label(net,
1603 				      &score->ifa->addr, score->addr_type,
1604 				      score->ifa->idev->dev->ifindex) == dst->label;
1605 		break;
1606 	case IPV6_SADDR_RULE_PRIVACY:
1607 	    {
1608 		/* Rule 7: Prefer public address
1609 		 * Note: prefer temporary address if use_tempaddr >= 2
1610 		 */
1611 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1612 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1613 				score->ifa->idev->cnf.use_tempaddr >= 2;
1614 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1615 		break;
1616 	    }
1617 	case IPV6_SADDR_RULE_ORCHID:
1618 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1619 		 *	    non-ORCHID vs non-ORCHID
1620 		 */
1621 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1622 			ipv6_addr_orchid(dst->addr));
1623 		break;
1624 	case IPV6_SADDR_RULE_PREFIX:
1625 		/* Rule 8: Use longest matching prefix */
1626 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1627 		if (ret > score->ifa->prefix_len)
1628 			ret = score->ifa->prefix_len;
1629 		score->matchlen = ret;
1630 		break;
1631 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1632 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1633 		/* Optimistic addresses still have lower precedence than other
1634 		 * preferred addresses.
1635 		 */
1636 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1637 		break;
1638 #endif
1639 	default:
1640 		ret = 0;
1641 	}
1642 
1643 	if (ret)
1644 		__set_bit(i, score->scorebits);
1645 	score->rule = i;
1646 out:
1647 	return ret;
1648 }
1649 
1650 static int __ipv6_dev_get_saddr(struct net *net,
1651 				struct ipv6_saddr_dst *dst,
1652 				struct inet6_dev *idev,
1653 				struct ipv6_saddr_score *scores,
1654 				int hiscore_idx)
1655 {
1656 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1657 
1658 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1659 		int i;
1660 
1661 		/*
1662 		 * - Tentative Address (RFC2462 section 5.4)
1663 		 *  - A tentative address is not considered
1664 		 *    "assigned to an interface" in the traditional
1665 		 *    sense, unless it is also flagged as optimistic.
1666 		 * - Candidate Source Address (section 4)
1667 		 *  - In any case, anycast addresses, multicast
1668 		 *    addresses, and the unspecified address MUST
1669 		 *    NOT be included in a candidate set.
1670 		 */
1671 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1672 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1673 			continue;
1674 
1675 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1676 
1677 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1678 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1679 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1680 					    idev->dev->name);
1681 			continue;
1682 		}
1683 
1684 		score->rule = -1;
1685 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1686 
1687 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1688 			int minihiscore, miniscore;
1689 
1690 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1691 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1692 
1693 			if (minihiscore > miniscore) {
1694 				if (i == IPV6_SADDR_RULE_SCOPE &&
1695 				    score->scopedist > 0) {
1696 					/*
1697 					 * special case:
1698 					 * each remaining entry
1699 					 * has too small (not enough)
1700 					 * scope, because ifa entries
1701 					 * are sorted by their scope
1702 					 * values.
1703 					 */
1704 					goto out;
1705 				}
1706 				break;
1707 			} else if (minihiscore < miniscore) {
1708 				swap(hiscore, score);
1709 				hiscore_idx = 1 - hiscore_idx;
1710 
1711 				/* restore our iterator */
1712 				score->ifa = hiscore->ifa;
1713 
1714 				break;
1715 			}
1716 		}
1717 	}
1718 out:
1719 	return hiscore_idx;
1720 }
1721 
1722 static int ipv6_get_saddr_master(struct net *net,
1723 				 const struct net_device *dst_dev,
1724 				 const struct net_device *master,
1725 				 struct ipv6_saddr_dst *dst,
1726 				 struct ipv6_saddr_score *scores,
1727 				 int hiscore_idx)
1728 {
1729 	struct inet6_dev *idev;
1730 
1731 	idev = __in6_dev_get(dst_dev);
1732 	if (idev)
1733 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1734 						   scores, hiscore_idx);
1735 
1736 	idev = __in6_dev_get(master);
1737 	if (idev)
1738 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1739 						   scores, hiscore_idx);
1740 
1741 	return hiscore_idx;
1742 }
1743 
1744 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1745 		       const struct in6_addr *daddr, unsigned int prefs,
1746 		       struct in6_addr *saddr)
1747 {
1748 	struct ipv6_saddr_score scores[2], *hiscore;
1749 	struct ipv6_saddr_dst dst;
1750 	struct inet6_dev *idev;
1751 	struct net_device *dev;
1752 	int dst_type;
1753 	bool use_oif_addr = false;
1754 	int hiscore_idx = 0;
1755 	int ret = 0;
1756 
1757 	dst_type = __ipv6_addr_type(daddr);
1758 	dst.addr = daddr;
1759 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1760 	dst.scope = __ipv6_addr_src_scope(dst_type);
1761 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1762 	dst.prefs = prefs;
1763 
1764 	scores[hiscore_idx].rule = -1;
1765 	scores[hiscore_idx].ifa = NULL;
1766 
1767 	rcu_read_lock();
1768 
1769 	/* Candidate Source Address (section 4)
1770 	 *  - multicast and link-local destination address,
1771 	 *    the set of candidate source address MUST only
1772 	 *    include addresses assigned to interfaces
1773 	 *    belonging to the same link as the outgoing
1774 	 *    interface.
1775 	 * (- For site-local destination addresses, the
1776 	 *    set of candidate source addresses MUST only
1777 	 *    include addresses assigned to interfaces
1778 	 *    belonging to the same site as the outgoing
1779 	 *    interface.)
1780 	 *  - "It is RECOMMENDED that the candidate source addresses
1781 	 *    be the set of unicast addresses assigned to the
1782 	 *    interface that will be used to send to the destination
1783 	 *    (the 'outgoing' interface)." (RFC 6724)
1784 	 */
1785 	if (dst_dev) {
1786 		idev = __in6_dev_get(dst_dev);
1787 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1788 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1789 		    (idev && idev->cnf.use_oif_addrs_only)) {
1790 			use_oif_addr = true;
1791 		}
1792 	}
1793 
1794 	if (use_oif_addr) {
1795 		if (idev)
1796 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1797 	} else {
1798 		const struct net_device *master;
1799 		int master_idx = 0;
1800 
1801 		/* if dst_dev exists and is enslaved to an L3 device, then
1802 		 * prefer addresses from dst_dev and then the master over
1803 		 * any other enslaved devices in the L3 domain.
1804 		 */
1805 		master = l3mdev_master_dev_rcu(dst_dev);
1806 		if (master) {
1807 			master_idx = master->ifindex;
1808 
1809 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1810 							    master, &dst,
1811 							    scores, hiscore_idx);
1812 
1813 			if (scores[hiscore_idx].ifa)
1814 				goto out;
1815 		}
1816 
1817 		for_each_netdev_rcu(net, dev) {
1818 			/* only consider addresses on devices in the
1819 			 * same L3 domain
1820 			 */
1821 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1822 				continue;
1823 			idev = __in6_dev_get(dev);
1824 			if (!idev)
1825 				continue;
1826 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1827 		}
1828 	}
1829 
1830 out:
1831 	hiscore = &scores[hiscore_idx];
1832 	if (!hiscore->ifa)
1833 		ret = -EADDRNOTAVAIL;
1834 	else
1835 		*saddr = hiscore->ifa->addr;
1836 
1837 	rcu_read_unlock();
1838 	return ret;
1839 }
1840 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1841 
1842 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1843 		      u32 banned_flags)
1844 {
1845 	struct inet6_ifaddr *ifp;
1846 	int err = -EADDRNOTAVAIL;
1847 
1848 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1849 		if (ifp->scope > IFA_LINK)
1850 			break;
1851 		if (ifp->scope == IFA_LINK &&
1852 		    !(ifp->flags & banned_flags)) {
1853 			*addr = ifp->addr;
1854 			err = 0;
1855 			break;
1856 		}
1857 	}
1858 	return err;
1859 }
1860 
1861 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1862 		    u32 banned_flags)
1863 {
1864 	struct inet6_dev *idev;
1865 	int err = -EADDRNOTAVAIL;
1866 
1867 	rcu_read_lock();
1868 	idev = __in6_dev_get(dev);
1869 	if (idev) {
1870 		read_lock_bh(&idev->lock);
1871 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1872 		read_unlock_bh(&idev->lock);
1873 	}
1874 	rcu_read_unlock();
1875 	return err;
1876 }
1877 
1878 static int ipv6_count_addresses(const struct inet6_dev *idev)
1879 {
1880 	const struct inet6_ifaddr *ifp;
1881 	int cnt = 0;
1882 
1883 	rcu_read_lock();
1884 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1885 		cnt++;
1886 	rcu_read_unlock();
1887 	return cnt;
1888 }
1889 
1890 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1891 		  const struct net_device *dev, int strict)
1892 {
1893 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1894 				       strict, IFA_F_TENTATIVE);
1895 }
1896 EXPORT_SYMBOL(ipv6_chk_addr);
1897 
1898 /* device argument is used to find the L3 domain of interest. If
1899  * skip_dev_check is set, then the ifp device is not checked against
1900  * the passed in dev argument. So the 2 cases for addresses checks are:
1901  *   1. does the address exist in the L3 domain that dev is part of
1902  *      (skip_dev_check = true), or
1903  *
1904  *   2. does the address exist on the specific device
1905  *      (skip_dev_check = false)
1906  */
1907 static struct net_device *
1908 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1909 			  const struct net_device *dev, bool skip_dev_check,
1910 			  int strict, u32 banned_flags)
1911 {
1912 	unsigned int hash = inet6_addr_hash(net, addr);
1913 	struct net_device *l3mdev, *ndev;
1914 	struct inet6_ifaddr *ifp;
1915 	u32 ifp_flags;
1916 
1917 	rcu_read_lock();
1918 
1919 	l3mdev = l3mdev_master_dev_rcu(dev);
1920 	if (skip_dev_check)
1921 		dev = NULL;
1922 
1923 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1924 		ndev = ifp->idev->dev;
1925 		if (!net_eq(dev_net(ndev), net))
1926 			continue;
1927 
1928 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1929 			continue;
1930 
1931 		/* Decouple optimistic from tentative for evaluation here.
1932 		 * Ban optimistic addresses explicitly, when required.
1933 		 */
1934 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1935 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1936 			    : ifp->flags;
1937 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1938 		    !(ifp_flags&banned_flags) &&
1939 		    (!dev || ndev == dev ||
1940 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1941 			rcu_read_unlock();
1942 			return ndev;
1943 		}
1944 	}
1945 
1946 	rcu_read_unlock();
1947 	return NULL;
1948 }
1949 
1950 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1951 			    const struct net_device *dev, bool skip_dev_check,
1952 			    int strict, u32 banned_flags)
1953 {
1954 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1955 					 strict, banned_flags) ? 1 : 0;
1956 }
1957 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1958 
1959 
1960 /* Compares an address/prefix_len with addresses on device @dev.
1961  * If one is found it returns true.
1962  */
1963 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1964 	const unsigned int prefix_len, struct net_device *dev)
1965 {
1966 	const struct inet6_ifaddr *ifa;
1967 	const struct inet6_dev *idev;
1968 	bool ret = false;
1969 
1970 	rcu_read_lock();
1971 	idev = __in6_dev_get(dev);
1972 	if (idev) {
1973 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1974 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1975 			if (ret)
1976 				break;
1977 		}
1978 	}
1979 	rcu_read_unlock();
1980 
1981 	return ret;
1982 }
1983 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1984 
1985 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1986 {
1987 	const struct inet6_ifaddr *ifa;
1988 	const struct inet6_dev *idev;
1989 	int	onlink;
1990 
1991 	onlink = 0;
1992 	rcu_read_lock();
1993 	idev = __in6_dev_get(dev);
1994 	if (idev) {
1995 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1996 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1997 						   ifa->prefix_len);
1998 			if (onlink)
1999 				break;
2000 		}
2001 	}
2002 	rcu_read_unlock();
2003 	return onlink;
2004 }
2005 EXPORT_SYMBOL(ipv6_chk_prefix);
2006 
2007 /**
2008  * ipv6_dev_find - find the first device with a given source address.
2009  * @net: the net namespace
2010  * @addr: the source address
2011  * @dev: used to find the L3 domain of interest
2012  *
2013  * The caller should be protected by RCU, or RTNL.
2014  */
2015 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2016 				 struct net_device *dev)
2017 {
2018 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2019 					 IFA_F_TENTATIVE);
2020 }
2021 EXPORT_SYMBOL(ipv6_dev_find);
2022 
2023 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2024 				     struct net_device *dev, int strict)
2025 {
2026 	unsigned int hash = inet6_addr_hash(net, addr);
2027 	struct inet6_ifaddr *ifp, *result = NULL;
2028 
2029 	rcu_read_lock();
2030 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2031 		if (!net_eq(dev_net(ifp->idev->dev), net))
2032 			continue;
2033 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2034 			if (!dev || ifp->idev->dev == dev ||
2035 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2036 				result = ifp;
2037 				in6_ifa_hold(ifp);
2038 				break;
2039 			}
2040 		}
2041 	}
2042 	rcu_read_unlock();
2043 
2044 	return result;
2045 }
2046 
2047 /* Gets referenced address, destroys ifaddr */
2048 
2049 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2050 {
2051 	if (dad_failed)
2052 		ifp->flags |= IFA_F_DADFAILED;
2053 
2054 	if (ifp->flags&IFA_F_TEMPORARY) {
2055 		struct inet6_ifaddr *ifpub;
2056 		spin_lock_bh(&ifp->lock);
2057 		ifpub = ifp->ifpub;
2058 		if (ifpub) {
2059 			in6_ifa_hold(ifpub);
2060 			spin_unlock_bh(&ifp->lock);
2061 			ipv6_create_tempaddr(ifpub, true);
2062 			in6_ifa_put(ifpub);
2063 		} else {
2064 			spin_unlock_bh(&ifp->lock);
2065 		}
2066 		ipv6_del_addr(ifp);
2067 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2068 		spin_lock_bh(&ifp->lock);
2069 		addrconf_del_dad_work(ifp);
2070 		ifp->flags |= IFA_F_TENTATIVE;
2071 		if (dad_failed)
2072 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2073 		spin_unlock_bh(&ifp->lock);
2074 		if (dad_failed)
2075 			ipv6_ifa_notify(0, ifp);
2076 		in6_ifa_put(ifp);
2077 	} else {
2078 		ipv6_del_addr(ifp);
2079 	}
2080 }
2081 
2082 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2083 {
2084 	int err = -ENOENT;
2085 
2086 	spin_lock_bh(&ifp->lock);
2087 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2088 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2089 		err = 0;
2090 	}
2091 	spin_unlock_bh(&ifp->lock);
2092 
2093 	return err;
2094 }
2095 
2096 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2097 {
2098 	struct inet6_dev *idev = ifp->idev;
2099 	struct net *net = dev_net(ifp->idev->dev);
2100 
2101 	if (addrconf_dad_end(ifp)) {
2102 		in6_ifa_put(ifp);
2103 		return;
2104 	}
2105 
2106 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2107 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2108 
2109 	spin_lock_bh(&ifp->lock);
2110 
2111 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2112 		struct in6_addr new_addr;
2113 		struct inet6_ifaddr *ifp2;
2114 		int retries = ifp->stable_privacy_retry + 1;
2115 		struct ifa6_config cfg = {
2116 			.pfx = &new_addr,
2117 			.plen = ifp->prefix_len,
2118 			.ifa_flags = ifp->flags,
2119 			.valid_lft = ifp->valid_lft,
2120 			.preferred_lft = ifp->prefered_lft,
2121 			.scope = ifp->scope,
2122 		};
2123 
2124 		if (retries > net->ipv6.sysctl.idgen_retries) {
2125 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2126 					     ifp->idev->dev->name);
2127 			goto errdad;
2128 		}
2129 
2130 		new_addr = ifp->addr;
2131 		if (ipv6_generate_stable_address(&new_addr, retries,
2132 						 idev))
2133 			goto errdad;
2134 
2135 		spin_unlock_bh(&ifp->lock);
2136 
2137 		if (idev->cnf.max_addresses &&
2138 		    ipv6_count_addresses(idev) >=
2139 		    idev->cnf.max_addresses)
2140 			goto lock_errdad;
2141 
2142 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2143 				     ifp->idev->dev->name);
2144 
2145 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2146 		if (IS_ERR(ifp2))
2147 			goto lock_errdad;
2148 
2149 		spin_lock_bh(&ifp2->lock);
2150 		ifp2->stable_privacy_retry = retries;
2151 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2152 		spin_unlock_bh(&ifp2->lock);
2153 
2154 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2155 		in6_ifa_put(ifp2);
2156 lock_errdad:
2157 		spin_lock_bh(&ifp->lock);
2158 	}
2159 
2160 errdad:
2161 	/* transition from _POSTDAD to _ERRDAD */
2162 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2163 	spin_unlock_bh(&ifp->lock);
2164 
2165 	addrconf_mod_dad_work(ifp, 0);
2166 	in6_ifa_put(ifp);
2167 }
2168 
2169 /* Join to solicited addr multicast group.
2170  * caller must hold RTNL */
2171 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2172 {
2173 	struct in6_addr maddr;
2174 
2175 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2176 		return;
2177 
2178 	addrconf_addr_solict_mult(addr, &maddr);
2179 	ipv6_dev_mc_inc(dev, &maddr);
2180 }
2181 
2182 /* caller must hold RTNL */
2183 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2184 {
2185 	struct in6_addr maddr;
2186 
2187 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2188 		return;
2189 
2190 	addrconf_addr_solict_mult(addr, &maddr);
2191 	__ipv6_dev_mc_dec(idev, &maddr);
2192 }
2193 
2194 /* caller must hold RTNL */
2195 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2196 {
2197 	struct in6_addr addr;
2198 
2199 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2200 		return;
2201 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2202 	if (ipv6_addr_any(&addr))
2203 		return;
2204 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2205 }
2206 
2207 /* caller must hold RTNL */
2208 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2209 {
2210 	struct in6_addr addr;
2211 
2212 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2213 		return;
2214 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2215 	if (ipv6_addr_any(&addr))
2216 		return;
2217 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2218 }
2219 
2220 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2221 {
2222 	switch (dev->addr_len) {
2223 	case ETH_ALEN:
2224 		memcpy(eui, dev->dev_addr, 3);
2225 		eui[3] = 0xFF;
2226 		eui[4] = 0xFE;
2227 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2228 		break;
2229 	case EUI64_ADDR_LEN:
2230 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2231 		eui[0] ^= 2;
2232 		break;
2233 	default:
2234 		return -1;
2235 	}
2236 
2237 	return 0;
2238 }
2239 
2240 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2241 {
2242 	const union fwnet_hwaddr *ha;
2243 
2244 	if (dev->addr_len != FWNET_ALEN)
2245 		return -1;
2246 
2247 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2248 
2249 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2250 	eui[0] ^= 2;
2251 	return 0;
2252 }
2253 
2254 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2255 {
2256 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2257 	if (dev->addr_len != ARCNET_ALEN)
2258 		return -1;
2259 	memset(eui, 0, 7);
2260 	eui[7] = *(u8 *)dev->dev_addr;
2261 	return 0;
2262 }
2263 
2264 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2265 {
2266 	if (dev->addr_len != INFINIBAND_ALEN)
2267 		return -1;
2268 	memcpy(eui, dev->dev_addr + 12, 8);
2269 	eui[0] |= 2;
2270 	return 0;
2271 }
2272 
2273 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2274 {
2275 	if (addr == 0)
2276 		return -1;
2277 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2278 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2279 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2280 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2281 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2282 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2283 	eui[1] = 0;
2284 	eui[2] = 0x5E;
2285 	eui[3] = 0xFE;
2286 	memcpy(eui + 4, &addr, 4);
2287 	return 0;
2288 }
2289 
2290 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2291 {
2292 	if (dev->priv_flags & IFF_ISATAP)
2293 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2294 	return -1;
2295 }
2296 
2297 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2298 {
2299 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2300 }
2301 
2302 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2303 {
2304 	memcpy(eui, dev->perm_addr, 3);
2305 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2306 	eui[3] = 0xFF;
2307 	eui[4] = 0xFE;
2308 	eui[0] ^= 2;
2309 	return 0;
2310 }
2311 
2312 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2313 {
2314 	switch (dev->type) {
2315 	case ARPHRD_ETHER:
2316 	case ARPHRD_FDDI:
2317 		return addrconf_ifid_eui48(eui, dev);
2318 	case ARPHRD_ARCNET:
2319 		return addrconf_ifid_arcnet(eui, dev);
2320 	case ARPHRD_INFINIBAND:
2321 		return addrconf_ifid_infiniband(eui, dev);
2322 	case ARPHRD_SIT:
2323 		return addrconf_ifid_sit(eui, dev);
2324 	case ARPHRD_IPGRE:
2325 	case ARPHRD_TUNNEL:
2326 		return addrconf_ifid_gre(eui, dev);
2327 	case ARPHRD_6LOWPAN:
2328 		return addrconf_ifid_6lowpan(eui, dev);
2329 	case ARPHRD_IEEE1394:
2330 		return addrconf_ifid_ieee1394(eui, dev);
2331 	case ARPHRD_TUNNEL6:
2332 	case ARPHRD_IP6GRE:
2333 	case ARPHRD_RAWIP:
2334 		return addrconf_ifid_ip6tnl(eui, dev);
2335 	}
2336 	return -1;
2337 }
2338 
2339 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2340 {
2341 	int err = -1;
2342 	struct inet6_ifaddr *ifp;
2343 
2344 	read_lock_bh(&idev->lock);
2345 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2346 		if (ifp->scope > IFA_LINK)
2347 			break;
2348 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2349 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2350 			err = 0;
2351 			break;
2352 		}
2353 	}
2354 	read_unlock_bh(&idev->lock);
2355 	return err;
2356 }
2357 
2358 /* Generation of a randomized Interface Identifier
2359  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2360  */
2361 
2362 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2363 {
2364 regen:
2365 	get_random_bytes(&addr->s6_addr[8], 8);
2366 
2367 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2368 	 * check if generated address is not inappropriate:
2369 	 *
2370 	 * - Reserved IPv6 Interface Identifiers
2371 	 * - XXX: already assigned to an address on the device
2372 	 */
2373 
2374 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2375 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2376 		goto regen;
2377 
2378 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2379 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2380 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2381 	 */
2382 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2383 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2384 		goto regen;
2385 
2386 	/* Reserved subnet anycast addresses */
2387 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2388 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2389 		goto regen;
2390 }
2391 
2392 /*
2393  *	Add prefix route.
2394  */
2395 
2396 static void
2397 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2398 		      struct net_device *dev, unsigned long expires,
2399 		      u32 flags, gfp_t gfp_flags)
2400 {
2401 	struct fib6_config cfg = {
2402 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2403 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2404 		.fc_ifindex = dev->ifindex,
2405 		.fc_expires = expires,
2406 		.fc_dst_len = plen,
2407 		.fc_flags = RTF_UP | flags,
2408 		.fc_nlinfo.nl_net = dev_net(dev),
2409 		.fc_protocol = RTPROT_KERNEL,
2410 		.fc_type = RTN_UNICAST,
2411 	};
2412 
2413 	cfg.fc_dst = *pfx;
2414 
2415 	/* Prevent useless cloning on PtP SIT.
2416 	   This thing is done here expecting that the whole
2417 	   class of non-broadcast devices need not cloning.
2418 	 */
2419 #if IS_ENABLED(CONFIG_IPV6_SIT)
2420 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2421 		cfg.fc_flags |= RTF_NONEXTHOP;
2422 #endif
2423 
2424 	ip6_route_add(&cfg, gfp_flags, NULL);
2425 }
2426 
2427 
2428 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2429 						  int plen,
2430 						  const struct net_device *dev,
2431 						  u32 flags, u32 noflags,
2432 						  bool no_gw)
2433 {
2434 	struct fib6_node *fn;
2435 	struct fib6_info *rt = NULL;
2436 	struct fib6_table *table;
2437 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2438 
2439 	table = fib6_get_table(dev_net(dev), tb_id);
2440 	if (!table)
2441 		return NULL;
2442 
2443 	rcu_read_lock();
2444 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2445 	if (!fn)
2446 		goto out;
2447 
2448 	for_each_fib6_node_rt_rcu(fn) {
2449 		/* prefix routes only use builtin fib6_nh */
2450 		if (rt->nh)
2451 			continue;
2452 
2453 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2454 			continue;
2455 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2456 			continue;
2457 		if ((rt->fib6_flags & flags) != flags)
2458 			continue;
2459 		if ((rt->fib6_flags & noflags) != 0)
2460 			continue;
2461 		if (!fib6_info_hold_safe(rt))
2462 			continue;
2463 		break;
2464 	}
2465 out:
2466 	rcu_read_unlock();
2467 	return rt;
2468 }
2469 
2470 
2471 /* Create "default" multicast route to the interface */
2472 
2473 static void addrconf_add_mroute(struct net_device *dev)
2474 {
2475 	struct fib6_config cfg = {
2476 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2477 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2478 		.fc_ifindex = dev->ifindex,
2479 		.fc_dst_len = 8,
2480 		.fc_flags = RTF_UP,
2481 		.fc_type = RTN_MULTICAST,
2482 		.fc_nlinfo.nl_net = dev_net(dev),
2483 		.fc_protocol = RTPROT_KERNEL,
2484 	};
2485 
2486 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2487 
2488 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2489 }
2490 
2491 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2492 {
2493 	struct inet6_dev *idev;
2494 
2495 	ASSERT_RTNL();
2496 
2497 	idev = ipv6_find_idev(dev);
2498 	if (IS_ERR(idev))
2499 		return idev;
2500 
2501 	if (idev->cnf.disable_ipv6)
2502 		return ERR_PTR(-EACCES);
2503 
2504 	/* Add default multicast route */
2505 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2506 		addrconf_add_mroute(dev);
2507 
2508 	return idev;
2509 }
2510 
2511 static void manage_tempaddrs(struct inet6_dev *idev,
2512 			     struct inet6_ifaddr *ifp,
2513 			     __u32 valid_lft, __u32 prefered_lft,
2514 			     bool create, unsigned long now)
2515 {
2516 	u32 flags;
2517 	struct inet6_ifaddr *ift;
2518 
2519 	read_lock_bh(&idev->lock);
2520 	/* update all temporary addresses in the list */
2521 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2522 		int age, max_valid, max_prefered;
2523 
2524 		if (ifp != ift->ifpub)
2525 			continue;
2526 
2527 		/* RFC 4941 section 3.3:
2528 		 * If a received option will extend the lifetime of a public
2529 		 * address, the lifetimes of temporary addresses should
2530 		 * be extended, subject to the overall constraint that no
2531 		 * temporary addresses should ever remain "valid" or "preferred"
2532 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2533 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2534 		 */
2535 		age = (now - ift->cstamp) / HZ;
2536 		max_valid = idev->cnf.temp_valid_lft - age;
2537 		if (max_valid < 0)
2538 			max_valid = 0;
2539 
2540 		max_prefered = idev->cnf.temp_prefered_lft -
2541 			       idev->desync_factor - age;
2542 		if (max_prefered < 0)
2543 			max_prefered = 0;
2544 
2545 		if (valid_lft > max_valid)
2546 			valid_lft = max_valid;
2547 
2548 		if (prefered_lft > max_prefered)
2549 			prefered_lft = max_prefered;
2550 
2551 		spin_lock(&ift->lock);
2552 		flags = ift->flags;
2553 		ift->valid_lft = valid_lft;
2554 		ift->prefered_lft = prefered_lft;
2555 		ift->tstamp = now;
2556 		if (prefered_lft > 0)
2557 			ift->flags &= ~IFA_F_DEPRECATED;
2558 
2559 		spin_unlock(&ift->lock);
2560 		if (!(flags&IFA_F_TENTATIVE))
2561 			ipv6_ifa_notify(0, ift);
2562 	}
2563 
2564 	if ((create || list_empty(&idev->tempaddr_list)) &&
2565 	    idev->cnf.use_tempaddr > 0) {
2566 		/* When a new public address is created as described
2567 		 * in [ADDRCONF], also create a new temporary address.
2568 		 * Also create a temporary address if it's enabled but
2569 		 * no temporary address currently exists.
2570 		 */
2571 		read_unlock_bh(&idev->lock);
2572 		ipv6_create_tempaddr(ifp, false);
2573 	} else {
2574 		read_unlock_bh(&idev->lock);
2575 	}
2576 }
2577 
2578 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2579 {
2580 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2581 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2582 }
2583 
2584 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2585 				 const struct prefix_info *pinfo,
2586 				 struct inet6_dev *in6_dev,
2587 				 const struct in6_addr *addr, int addr_type,
2588 				 u32 addr_flags, bool sllao, bool tokenized,
2589 				 __u32 valid_lft, u32 prefered_lft)
2590 {
2591 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2592 	int create = 0;
2593 
2594 	if (!ifp && valid_lft) {
2595 		int max_addresses = in6_dev->cnf.max_addresses;
2596 		struct ifa6_config cfg = {
2597 			.pfx = addr,
2598 			.plen = pinfo->prefix_len,
2599 			.ifa_flags = addr_flags,
2600 			.valid_lft = valid_lft,
2601 			.preferred_lft = prefered_lft,
2602 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2603 		};
2604 
2605 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2606 		if ((net->ipv6.devconf_all->optimistic_dad ||
2607 		     in6_dev->cnf.optimistic_dad) &&
2608 		    !net->ipv6.devconf_all->forwarding && sllao)
2609 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2610 #endif
2611 
2612 		/* Do not allow to create too much of autoconfigured
2613 		 * addresses; this would be too easy way to crash kernel.
2614 		 */
2615 		if (!max_addresses ||
2616 		    ipv6_count_addresses(in6_dev) < max_addresses)
2617 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2618 
2619 		if (IS_ERR_OR_NULL(ifp))
2620 			return -1;
2621 
2622 		create = 1;
2623 		spin_lock_bh(&ifp->lock);
2624 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2625 		ifp->cstamp = jiffies;
2626 		ifp->tokenized = tokenized;
2627 		spin_unlock_bh(&ifp->lock);
2628 		addrconf_dad_start(ifp);
2629 	}
2630 
2631 	if (ifp) {
2632 		u32 flags;
2633 		unsigned long now;
2634 		u32 stored_lft;
2635 
2636 		/* Update lifetime (RFC4862 5.5.3 e)
2637 		 * We deviate from RFC4862 by honoring all Valid Lifetimes to
2638 		 * improve the reaction of SLAAC to renumbering events
2639 		 * (draft-gont-6man-slaac-renum-06, Section 4.2)
2640 		 */
2641 		spin_lock_bh(&ifp->lock);
2642 		now = jiffies;
2643 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2644 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2645 		else
2646 			stored_lft = 0;
2647 
2648 		if (!create && stored_lft) {
2649 			ifp->valid_lft = valid_lft;
2650 			ifp->prefered_lft = prefered_lft;
2651 			ifp->tstamp = now;
2652 			flags = ifp->flags;
2653 			ifp->flags &= ~IFA_F_DEPRECATED;
2654 			spin_unlock_bh(&ifp->lock);
2655 
2656 			if (!(flags&IFA_F_TENTATIVE))
2657 				ipv6_ifa_notify(0, ifp);
2658 		} else
2659 			spin_unlock_bh(&ifp->lock);
2660 
2661 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2662 				 create, now);
2663 
2664 		in6_ifa_put(ifp);
2665 		addrconf_verify();
2666 	}
2667 
2668 	return 0;
2669 }
2670 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2671 
2672 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2673 {
2674 	struct prefix_info *pinfo;
2675 	__u32 valid_lft;
2676 	__u32 prefered_lft;
2677 	int addr_type, err;
2678 	u32 addr_flags = 0;
2679 	struct inet6_dev *in6_dev;
2680 	struct net *net = dev_net(dev);
2681 
2682 	pinfo = (struct prefix_info *) opt;
2683 
2684 	if (len < sizeof(struct prefix_info)) {
2685 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2686 		return;
2687 	}
2688 
2689 	/*
2690 	 *	Validation checks ([ADDRCONF], page 19)
2691 	 */
2692 
2693 	addr_type = ipv6_addr_type(&pinfo->prefix);
2694 
2695 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2696 		return;
2697 
2698 	valid_lft = ntohl(pinfo->valid);
2699 	prefered_lft = ntohl(pinfo->prefered);
2700 
2701 	if (prefered_lft > valid_lft) {
2702 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2703 		return;
2704 	}
2705 
2706 	in6_dev = in6_dev_get(dev);
2707 
2708 	if (!in6_dev) {
2709 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2710 				    dev->name);
2711 		return;
2712 	}
2713 
2714 	/*
2715 	 *	Two things going on here:
2716 	 *	1) Add routes for on-link prefixes
2717 	 *	2) Configure prefixes with the auto flag set
2718 	 */
2719 
2720 	if (pinfo->onlink) {
2721 		struct fib6_info *rt;
2722 		unsigned long rt_expires;
2723 
2724 		/* Avoid arithmetic overflow. Really, we could
2725 		 * save rt_expires in seconds, likely valid_lft,
2726 		 * but it would require division in fib gc, that it
2727 		 * not good.
2728 		 */
2729 		if (HZ > USER_HZ)
2730 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2731 		else
2732 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2733 
2734 		if (addrconf_finite_timeout(rt_expires))
2735 			rt_expires *= HZ;
2736 
2737 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2738 					       pinfo->prefix_len,
2739 					       dev,
2740 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2741 					       RTF_DEFAULT, true);
2742 
2743 		if (rt) {
2744 			/* Autoconf prefix route */
2745 			if (valid_lft == 0) {
2746 				ip6_del_rt(net, rt, false);
2747 				rt = NULL;
2748 			} else if (addrconf_finite_timeout(rt_expires)) {
2749 				/* not infinity */
2750 				fib6_set_expires(rt, jiffies + rt_expires);
2751 			} else {
2752 				fib6_clean_expires(rt);
2753 			}
2754 		} else if (valid_lft) {
2755 			clock_t expires = 0;
2756 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2757 			if (addrconf_finite_timeout(rt_expires)) {
2758 				/* not infinity */
2759 				flags |= RTF_EXPIRES;
2760 				expires = jiffies_to_clock_t(rt_expires);
2761 			}
2762 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2763 					      0, dev, expires, flags,
2764 					      GFP_ATOMIC);
2765 		}
2766 		fib6_info_release(rt);
2767 	}
2768 
2769 	/* Try to figure out our local address for this prefix */
2770 
2771 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2772 		struct in6_addr addr;
2773 		bool tokenized = false, dev_addr_generated = false;
2774 
2775 		if (pinfo->prefix_len == 64) {
2776 			memcpy(&addr, &pinfo->prefix, 8);
2777 
2778 			if (!ipv6_addr_any(&in6_dev->token)) {
2779 				read_lock_bh(&in6_dev->lock);
2780 				memcpy(addr.s6_addr + 8,
2781 				       in6_dev->token.s6_addr + 8, 8);
2782 				read_unlock_bh(&in6_dev->lock);
2783 				tokenized = true;
2784 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2785 				   !ipv6_generate_stable_address(&addr, 0,
2786 								 in6_dev)) {
2787 				addr_flags |= IFA_F_STABLE_PRIVACY;
2788 				goto ok;
2789 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2790 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2791 				goto put;
2792 			} else {
2793 				dev_addr_generated = true;
2794 			}
2795 			goto ok;
2796 		}
2797 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2798 				    pinfo->prefix_len);
2799 		goto put;
2800 
2801 ok:
2802 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2803 						   &addr, addr_type,
2804 						   addr_flags, sllao,
2805 						   tokenized, valid_lft,
2806 						   prefered_lft);
2807 		if (err)
2808 			goto put;
2809 
2810 		/* Ignore error case here because previous prefix add addr was
2811 		 * successful which will be notified.
2812 		 */
2813 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2814 					      addr_type, addr_flags, sllao,
2815 					      tokenized, valid_lft,
2816 					      prefered_lft,
2817 					      dev_addr_generated);
2818 	}
2819 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2820 put:
2821 	in6_dev_put(in6_dev);
2822 }
2823 
2824 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2825 		struct in6_ifreq *ireq)
2826 {
2827 	struct ip_tunnel_parm p = { };
2828 	int err;
2829 
2830 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2831 		return -EADDRNOTAVAIL;
2832 
2833 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2834 	p.iph.version = 4;
2835 	p.iph.ihl = 5;
2836 	p.iph.protocol = IPPROTO_IPV6;
2837 	p.iph.ttl = 64;
2838 
2839 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2840 		return -EOPNOTSUPP;
2841 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2842 	if (err)
2843 		return err;
2844 
2845 	dev = __dev_get_by_name(net, p.name);
2846 	if (!dev)
2847 		return -ENOBUFS;
2848 	return dev_open(dev, NULL);
2849 }
2850 
2851 /*
2852  *	Set destination address.
2853  *	Special case for SIT interfaces where we create a new "virtual"
2854  *	device.
2855  */
2856 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2857 {
2858 	struct net_device *dev;
2859 	struct in6_ifreq ireq;
2860 	int err = -ENODEV;
2861 
2862 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2863 		return -ENODEV;
2864 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2865 		return -EFAULT;
2866 
2867 	rtnl_lock();
2868 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2869 	if (dev && dev->type == ARPHRD_SIT)
2870 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2871 	rtnl_unlock();
2872 	return err;
2873 }
2874 
2875 static int ipv6_mc_config(struct sock *sk, bool join,
2876 			  const struct in6_addr *addr, int ifindex)
2877 {
2878 	int ret;
2879 
2880 	ASSERT_RTNL();
2881 
2882 	lock_sock(sk);
2883 	if (join)
2884 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2885 	else
2886 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2887 	release_sock(sk);
2888 
2889 	return ret;
2890 }
2891 
2892 /*
2893  *	Manual configuration of address on an interface
2894  */
2895 static int inet6_addr_add(struct net *net, int ifindex,
2896 			  struct ifa6_config *cfg,
2897 			  struct netlink_ext_ack *extack)
2898 {
2899 	struct inet6_ifaddr *ifp;
2900 	struct inet6_dev *idev;
2901 	struct net_device *dev;
2902 	unsigned long timeout;
2903 	clock_t expires;
2904 	u32 flags;
2905 
2906 	ASSERT_RTNL();
2907 
2908 	if (cfg->plen > 128)
2909 		return -EINVAL;
2910 
2911 	/* check the lifetime */
2912 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2913 		return -EINVAL;
2914 
2915 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2916 		return -EINVAL;
2917 
2918 	dev = __dev_get_by_index(net, ifindex);
2919 	if (!dev)
2920 		return -ENODEV;
2921 
2922 	idev = addrconf_add_dev(dev);
2923 	if (IS_ERR(idev))
2924 		return PTR_ERR(idev);
2925 
2926 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2927 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2928 					 true, cfg->pfx, ifindex);
2929 
2930 		if (ret < 0)
2931 			return ret;
2932 	}
2933 
2934 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2935 
2936 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2937 	if (addrconf_finite_timeout(timeout)) {
2938 		expires = jiffies_to_clock_t(timeout * HZ);
2939 		cfg->valid_lft = timeout;
2940 		flags = RTF_EXPIRES;
2941 	} else {
2942 		expires = 0;
2943 		flags = 0;
2944 		cfg->ifa_flags |= IFA_F_PERMANENT;
2945 	}
2946 
2947 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2948 	if (addrconf_finite_timeout(timeout)) {
2949 		if (timeout == 0)
2950 			cfg->ifa_flags |= IFA_F_DEPRECATED;
2951 		cfg->preferred_lft = timeout;
2952 	}
2953 
2954 	ifp = ipv6_add_addr(idev, cfg, true, extack);
2955 	if (!IS_ERR(ifp)) {
2956 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2957 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2958 					      ifp->rt_priority, dev, expires,
2959 					      flags, GFP_KERNEL);
2960 		}
2961 
2962 		/* Send a netlink notification if DAD is enabled and
2963 		 * optimistic flag is not set
2964 		 */
2965 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2966 			ipv6_ifa_notify(0, ifp);
2967 		/*
2968 		 * Note that section 3.1 of RFC 4429 indicates
2969 		 * that the Optimistic flag should not be set for
2970 		 * manually configured addresses
2971 		 */
2972 		addrconf_dad_start(ifp);
2973 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2974 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
2975 					 cfg->preferred_lft, true, jiffies);
2976 		in6_ifa_put(ifp);
2977 		addrconf_verify_rtnl();
2978 		return 0;
2979 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2980 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2981 			       cfg->pfx, ifindex);
2982 	}
2983 
2984 	return PTR_ERR(ifp);
2985 }
2986 
2987 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2988 			  const struct in6_addr *pfx, unsigned int plen)
2989 {
2990 	struct inet6_ifaddr *ifp;
2991 	struct inet6_dev *idev;
2992 	struct net_device *dev;
2993 
2994 	if (plen > 128)
2995 		return -EINVAL;
2996 
2997 	dev = __dev_get_by_index(net, ifindex);
2998 	if (!dev)
2999 		return -ENODEV;
3000 
3001 	idev = __in6_dev_get(dev);
3002 	if (!idev)
3003 		return -ENXIO;
3004 
3005 	read_lock_bh(&idev->lock);
3006 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3007 		if (ifp->prefix_len == plen &&
3008 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3009 			in6_ifa_hold(ifp);
3010 			read_unlock_bh(&idev->lock);
3011 
3012 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3013 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
3014 				manage_tempaddrs(idev, ifp, 0, 0, false,
3015 						 jiffies);
3016 			ipv6_del_addr(ifp);
3017 			addrconf_verify_rtnl();
3018 			if (ipv6_addr_is_multicast(pfx)) {
3019 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3020 					       false, pfx, dev->ifindex);
3021 			}
3022 			return 0;
3023 		}
3024 	}
3025 	read_unlock_bh(&idev->lock);
3026 	return -EADDRNOTAVAIL;
3027 }
3028 
3029 
3030 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3031 {
3032 	struct ifa6_config cfg = {
3033 		.ifa_flags = IFA_F_PERMANENT,
3034 		.preferred_lft = INFINITY_LIFE_TIME,
3035 		.valid_lft = INFINITY_LIFE_TIME,
3036 	};
3037 	struct in6_ifreq ireq;
3038 	int err;
3039 
3040 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3041 		return -EPERM;
3042 
3043 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3044 		return -EFAULT;
3045 
3046 	cfg.pfx = &ireq.ifr6_addr;
3047 	cfg.plen = ireq.ifr6_prefixlen;
3048 
3049 	rtnl_lock();
3050 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3051 	rtnl_unlock();
3052 	return err;
3053 }
3054 
3055 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3056 {
3057 	struct in6_ifreq ireq;
3058 	int err;
3059 
3060 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3061 		return -EPERM;
3062 
3063 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3064 		return -EFAULT;
3065 
3066 	rtnl_lock();
3067 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3068 			     ireq.ifr6_prefixlen);
3069 	rtnl_unlock();
3070 	return err;
3071 }
3072 
3073 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3074 		     int plen, int scope)
3075 {
3076 	struct inet6_ifaddr *ifp;
3077 	struct ifa6_config cfg = {
3078 		.pfx = addr,
3079 		.plen = plen,
3080 		.ifa_flags = IFA_F_PERMANENT,
3081 		.valid_lft = INFINITY_LIFE_TIME,
3082 		.preferred_lft = INFINITY_LIFE_TIME,
3083 		.scope = scope
3084 	};
3085 
3086 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3087 	if (!IS_ERR(ifp)) {
3088 		spin_lock_bh(&ifp->lock);
3089 		ifp->flags &= ~IFA_F_TENTATIVE;
3090 		spin_unlock_bh(&ifp->lock);
3091 		rt_genid_bump_ipv6(dev_net(idev->dev));
3092 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3093 		in6_ifa_put(ifp);
3094 	}
3095 }
3096 
3097 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3098 static void add_v4_addrs(struct inet6_dev *idev)
3099 {
3100 	struct in6_addr addr;
3101 	struct net_device *dev;
3102 	struct net *net = dev_net(idev->dev);
3103 	int scope, plen, offset = 0;
3104 	u32 pflags = 0;
3105 
3106 	ASSERT_RTNL();
3107 
3108 	memset(&addr, 0, sizeof(struct in6_addr));
3109 	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3110 	if (idev->dev->addr_len == sizeof(struct in6_addr))
3111 		offset = sizeof(struct in6_addr) - 4;
3112 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3113 
3114 	if (idev->dev->flags&IFF_POINTOPOINT) {
3115 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3116 			return;
3117 
3118 		addr.s6_addr32[0] = htonl(0xfe800000);
3119 		scope = IFA_LINK;
3120 		plen = 64;
3121 	} else {
3122 		scope = IPV6_ADDR_COMPATv4;
3123 		plen = 96;
3124 		pflags |= RTF_NONEXTHOP;
3125 	}
3126 
3127 	if (addr.s6_addr32[3]) {
3128 		add_addr(idev, &addr, plen, scope);
3129 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3130 				      GFP_KERNEL);
3131 		return;
3132 	}
3133 
3134 	for_each_netdev(net, dev) {
3135 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3136 		if (in_dev && (dev->flags & IFF_UP)) {
3137 			struct in_ifaddr *ifa;
3138 			int flag = scope;
3139 
3140 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3141 				addr.s6_addr32[3] = ifa->ifa_local;
3142 
3143 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3144 					continue;
3145 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3146 					if (idev->dev->flags&IFF_POINTOPOINT)
3147 						continue;
3148 					flag |= IFA_HOST;
3149 				}
3150 
3151 				add_addr(idev, &addr, plen, flag);
3152 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3153 						      0, pflags, GFP_KERNEL);
3154 			}
3155 		}
3156 	}
3157 }
3158 #endif
3159 
3160 static void init_loopback(struct net_device *dev)
3161 {
3162 	struct inet6_dev  *idev;
3163 
3164 	/* ::1 */
3165 
3166 	ASSERT_RTNL();
3167 
3168 	idev = ipv6_find_idev(dev);
3169 	if (IS_ERR(idev)) {
3170 		pr_debug("%s: add_dev failed\n", __func__);
3171 		return;
3172 	}
3173 
3174 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3175 }
3176 
3177 void addrconf_add_linklocal(struct inet6_dev *idev,
3178 			    const struct in6_addr *addr, u32 flags)
3179 {
3180 	struct ifa6_config cfg = {
3181 		.pfx = addr,
3182 		.plen = 64,
3183 		.ifa_flags = flags | IFA_F_PERMANENT,
3184 		.valid_lft = INFINITY_LIFE_TIME,
3185 		.preferred_lft = INFINITY_LIFE_TIME,
3186 		.scope = IFA_LINK
3187 	};
3188 	struct inet6_ifaddr *ifp;
3189 
3190 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3191 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3192 	     idev->cnf.optimistic_dad) &&
3193 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3194 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3195 #endif
3196 
3197 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3198 	if (!IS_ERR(ifp)) {
3199 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3200 				      0, 0, GFP_ATOMIC);
3201 		addrconf_dad_start(ifp);
3202 		in6_ifa_put(ifp);
3203 	}
3204 }
3205 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3206 
3207 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3208 {
3209 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3210 		return true;
3211 
3212 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3213 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3214 		return true;
3215 
3216 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3217 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3218 		return true;
3219 
3220 	return false;
3221 }
3222 
3223 static int ipv6_generate_stable_address(struct in6_addr *address,
3224 					u8 dad_count,
3225 					const struct inet6_dev *idev)
3226 {
3227 	static DEFINE_SPINLOCK(lock);
3228 	static __u32 digest[SHA1_DIGEST_WORDS];
3229 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3230 
3231 	static union {
3232 		char __data[SHA1_BLOCK_SIZE];
3233 		struct {
3234 			struct in6_addr secret;
3235 			__be32 prefix[2];
3236 			unsigned char hwaddr[MAX_ADDR_LEN];
3237 			u8 dad_count;
3238 		} __packed;
3239 	} data;
3240 
3241 	struct in6_addr secret;
3242 	struct in6_addr temp;
3243 	struct net *net = dev_net(idev->dev);
3244 
3245 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3246 
3247 	if (idev->cnf.stable_secret.initialized)
3248 		secret = idev->cnf.stable_secret.secret;
3249 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3250 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3251 	else
3252 		return -1;
3253 
3254 retry:
3255 	spin_lock_bh(&lock);
3256 
3257 	sha1_init(digest);
3258 	memset(&data, 0, sizeof(data));
3259 	memset(workspace, 0, sizeof(workspace));
3260 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3261 	data.prefix[0] = address->s6_addr32[0];
3262 	data.prefix[1] = address->s6_addr32[1];
3263 	data.secret = secret;
3264 	data.dad_count = dad_count;
3265 
3266 	sha1_transform(digest, data.__data, workspace);
3267 
3268 	temp = *address;
3269 	temp.s6_addr32[2] = (__force __be32)digest[0];
3270 	temp.s6_addr32[3] = (__force __be32)digest[1];
3271 
3272 	spin_unlock_bh(&lock);
3273 
3274 	if (ipv6_reserved_interfaceid(temp)) {
3275 		dad_count++;
3276 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3277 			return -1;
3278 		goto retry;
3279 	}
3280 
3281 	*address = temp;
3282 	return 0;
3283 }
3284 
3285 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3286 {
3287 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3288 
3289 	if (s->initialized)
3290 		return;
3291 	s = &idev->cnf.stable_secret;
3292 	get_random_bytes(&s->secret, sizeof(s->secret));
3293 	s->initialized = true;
3294 }
3295 
3296 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3297 {
3298 	struct in6_addr addr;
3299 
3300 	/* no link local addresses on L3 master devices */
3301 	if (netif_is_l3_master(idev->dev))
3302 		return;
3303 
3304 	/* no link local addresses on devices flagged as slaves */
3305 	if (idev->dev->flags & IFF_SLAVE)
3306 		return;
3307 
3308 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3309 
3310 	switch (idev->cnf.addr_gen_mode) {
3311 	case IN6_ADDR_GEN_MODE_RANDOM:
3312 		ipv6_gen_mode_random_init(idev);
3313 		fallthrough;
3314 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3315 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3316 			addrconf_add_linklocal(idev, &addr,
3317 					       IFA_F_STABLE_PRIVACY);
3318 		else if (prefix_route)
3319 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3320 					      0, 0, GFP_KERNEL);
3321 		break;
3322 	case IN6_ADDR_GEN_MODE_EUI64:
3323 		/* addrconf_add_linklocal also adds a prefix_route and we
3324 		 * only need to care about prefix routes if ipv6_generate_eui64
3325 		 * couldn't generate one.
3326 		 */
3327 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3328 			addrconf_add_linklocal(idev, &addr, 0);
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_NONE:
3334 	default:
3335 		/* will not add any link local address */
3336 		break;
3337 	}
3338 }
3339 
3340 static void addrconf_dev_config(struct net_device *dev)
3341 {
3342 	struct inet6_dev *idev;
3343 
3344 	ASSERT_RTNL();
3345 
3346 	if ((dev->type != ARPHRD_ETHER) &&
3347 	    (dev->type != ARPHRD_FDDI) &&
3348 	    (dev->type != ARPHRD_ARCNET) &&
3349 	    (dev->type != ARPHRD_INFINIBAND) &&
3350 	    (dev->type != ARPHRD_IEEE1394) &&
3351 	    (dev->type != ARPHRD_TUNNEL6) &&
3352 	    (dev->type != ARPHRD_6LOWPAN) &&
3353 	    (dev->type != ARPHRD_TUNNEL) &&
3354 	    (dev->type != ARPHRD_NONE) &&
3355 	    (dev->type != ARPHRD_RAWIP)) {
3356 		/* Alas, we support only Ethernet autoconfiguration. */
3357 		idev = __in6_dev_get(dev);
3358 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3359 		    dev->flags & IFF_MULTICAST)
3360 			ipv6_mc_up(idev);
3361 		return;
3362 	}
3363 
3364 	idev = addrconf_add_dev(dev);
3365 	if (IS_ERR(idev))
3366 		return;
3367 
3368 	/* this device type has no EUI support */
3369 	if (dev->type == ARPHRD_NONE &&
3370 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3371 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3372 
3373 	addrconf_addr_gen(idev, false);
3374 }
3375 
3376 #if IS_ENABLED(CONFIG_IPV6_SIT)
3377 static void addrconf_sit_config(struct net_device *dev)
3378 {
3379 	struct inet6_dev *idev;
3380 
3381 	ASSERT_RTNL();
3382 
3383 	/*
3384 	 * Configure the tunnel with one of our IPv4
3385 	 * addresses... we should configure all of
3386 	 * our v4 addrs in the tunnel
3387 	 */
3388 
3389 	idev = ipv6_find_idev(dev);
3390 	if (IS_ERR(idev)) {
3391 		pr_debug("%s: add_dev failed\n", __func__);
3392 		return;
3393 	}
3394 
3395 	if (dev->priv_flags & IFF_ISATAP) {
3396 		addrconf_addr_gen(idev, false);
3397 		return;
3398 	}
3399 
3400 	add_v4_addrs(idev);
3401 
3402 	if (dev->flags&IFF_POINTOPOINT)
3403 		addrconf_add_mroute(dev);
3404 }
3405 #endif
3406 
3407 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3408 static void addrconf_gre_config(struct net_device *dev)
3409 {
3410 	struct inet6_dev *idev;
3411 
3412 	ASSERT_RTNL();
3413 
3414 	idev = ipv6_find_idev(dev);
3415 	if (IS_ERR(idev)) {
3416 		pr_debug("%s: add_dev failed\n", __func__);
3417 		return;
3418 	}
3419 
3420 	if (dev->type == ARPHRD_ETHER) {
3421 		addrconf_addr_gen(idev, true);
3422 		return;
3423 	}
3424 
3425 	add_v4_addrs(idev);
3426 
3427 	if (dev->flags & IFF_POINTOPOINT)
3428 		addrconf_add_mroute(dev);
3429 }
3430 #endif
3431 
3432 static int fixup_permanent_addr(struct net *net,
3433 				struct inet6_dev *idev,
3434 				struct inet6_ifaddr *ifp)
3435 {
3436 	/* !fib6_node means the host route was removed from the
3437 	 * FIB, for example, if 'lo' device is taken down. In that
3438 	 * case regenerate the host route.
3439 	 */
3440 	if (!ifp->rt || !ifp->rt->fib6_node) {
3441 		struct fib6_info *f6i, *prev;
3442 
3443 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3444 					 GFP_ATOMIC);
3445 		if (IS_ERR(f6i))
3446 			return PTR_ERR(f6i);
3447 
3448 		/* ifp->rt can be accessed outside of rtnl */
3449 		spin_lock(&ifp->lock);
3450 		prev = ifp->rt;
3451 		ifp->rt = f6i;
3452 		spin_unlock(&ifp->lock);
3453 
3454 		fib6_info_release(prev);
3455 	}
3456 
3457 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3458 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3459 				      ifp->rt_priority, idev->dev, 0, 0,
3460 				      GFP_ATOMIC);
3461 	}
3462 
3463 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3464 		addrconf_dad_start(ifp);
3465 
3466 	return 0;
3467 }
3468 
3469 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3470 {
3471 	struct inet6_ifaddr *ifp, *tmp;
3472 	struct inet6_dev *idev;
3473 
3474 	idev = __in6_dev_get(dev);
3475 	if (!idev)
3476 		return;
3477 
3478 	write_lock_bh(&idev->lock);
3479 
3480 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3481 		if ((ifp->flags & IFA_F_PERMANENT) &&
3482 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3483 			write_unlock_bh(&idev->lock);
3484 			in6_ifa_hold(ifp);
3485 			ipv6_del_addr(ifp);
3486 			write_lock_bh(&idev->lock);
3487 
3488 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3489 					     idev->dev->name, &ifp->addr);
3490 		}
3491 	}
3492 
3493 	write_unlock_bh(&idev->lock);
3494 }
3495 
3496 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3497 			   void *ptr)
3498 {
3499 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3500 	struct netdev_notifier_change_info *change_info;
3501 	struct netdev_notifier_changeupper_info *info;
3502 	struct inet6_dev *idev = __in6_dev_get(dev);
3503 	struct net *net = dev_net(dev);
3504 	int run_pending = 0;
3505 	int err;
3506 
3507 	switch (event) {
3508 	case NETDEV_REGISTER:
3509 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3510 			idev = ipv6_add_dev(dev);
3511 			if (IS_ERR(idev))
3512 				return notifier_from_errno(PTR_ERR(idev));
3513 		}
3514 		break;
3515 
3516 	case NETDEV_CHANGEMTU:
3517 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3518 		if (dev->mtu < IPV6_MIN_MTU) {
3519 			addrconf_ifdown(dev, dev != net->loopback_dev);
3520 			break;
3521 		}
3522 
3523 		if (idev) {
3524 			rt6_mtu_change(dev, dev->mtu);
3525 			idev->cnf.mtu6 = dev->mtu;
3526 			break;
3527 		}
3528 
3529 		/* allocate new idev */
3530 		idev = ipv6_add_dev(dev);
3531 		if (IS_ERR(idev))
3532 			break;
3533 
3534 		/* device is still not ready */
3535 		if (!(idev->if_flags & IF_READY))
3536 			break;
3537 
3538 		run_pending = 1;
3539 		fallthrough;
3540 	case NETDEV_UP:
3541 	case NETDEV_CHANGE:
3542 		if (dev->flags & IFF_SLAVE)
3543 			break;
3544 
3545 		if (idev && idev->cnf.disable_ipv6)
3546 			break;
3547 
3548 		if (event == NETDEV_UP) {
3549 			/* restore routes for permanent addresses */
3550 			addrconf_permanent_addr(net, dev);
3551 
3552 			if (!addrconf_link_ready(dev)) {
3553 				/* device is not ready yet. */
3554 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3555 					 dev->name);
3556 				break;
3557 			}
3558 
3559 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3560 				idev = ipv6_add_dev(dev);
3561 
3562 			if (!IS_ERR_OR_NULL(idev)) {
3563 				idev->if_flags |= IF_READY;
3564 				run_pending = 1;
3565 			}
3566 		} else if (event == NETDEV_CHANGE) {
3567 			if (!addrconf_link_ready(dev)) {
3568 				/* device is still not ready. */
3569 				rt6_sync_down_dev(dev, event);
3570 				break;
3571 			}
3572 
3573 			if (!IS_ERR_OR_NULL(idev)) {
3574 				if (idev->if_flags & IF_READY) {
3575 					/* device is already configured -
3576 					 * but resend MLD reports, we might
3577 					 * have roamed and need to update
3578 					 * multicast snooping switches
3579 					 */
3580 					ipv6_mc_up(idev);
3581 					change_info = ptr;
3582 					if (change_info->flags_changed & IFF_NOARP)
3583 						addrconf_dad_run(idev, true);
3584 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3585 					break;
3586 				}
3587 				idev->if_flags |= IF_READY;
3588 			}
3589 
3590 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3591 				dev->name);
3592 
3593 			run_pending = 1;
3594 		}
3595 
3596 		switch (dev->type) {
3597 #if IS_ENABLED(CONFIG_IPV6_SIT)
3598 		case ARPHRD_SIT:
3599 			addrconf_sit_config(dev);
3600 			break;
3601 #endif
3602 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3603 		case ARPHRD_IP6GRE:
3604 		case ARPHRD_IPGRE:
3605 			addrconf_gre_config(dev);
3606 			break;
3607 #endif
3608 		case ARPHRD_LOOPBACK:
3609 			init_loopback(dev);
3610 			break;
3611 
3612 		default:
3613 			addrconf_dev_config(dev);
3614 			break;
3615 		}
3616 
3617 		if (!IS_ERR_OR_NULL(idev)) {
3618 			if (run_pending)
3619 				addrconf_dad_run(idev, false);
3620 
3621 			/* Device has an address by now */
3622 			rt6_sync_up(dev, RTNH_F_DEAD);
3623 
3624 			/*
3625 			 * If the MTU changed during the interface down,
3626 			 * when the interface up, the changed MTU must be
3627 			 * reflected in the idev as well as routers.
3628 			 */
3629 			if (idev->cnf.mtu6 != dev->mtu &&
3630 			    dev->mtu >= IPV6_MIN_MTU) {
3631 				rt6_mtu_change(dev, dev->mtu);
3632 				idev->cnf.mtu6 = dev->mtu;
3633 			}
3634 			idev->tstamp = jiffies;
3635 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3636 
3637 			/*
3638 			 * If the changed mtu during down is lower than
3639 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3640 			 */
3641 			if (dev->mtu < IPV6_MIN_MTU)
3642 				addrconf_ifdown(dev, dev != net->loopback_dev);
3643 		}
3644 		break;
3645 
3646 	case NETDEV_DOWN:
3647 	case NETDEV_UNREGISTER:
3648 		/*
3649 		 *	Remove all addresses from this interface.
3650 		 */
3651 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3652 		break;
3653 
3654 	case NETDEV_CHANGENAME:
3655 		if (idev) {
3656 			snmp6_unregister_dev(idev);
3657 			addrconf_sysctl_unregister(idev);
3658 			err = addrconf_sysctl_register(idev);
3659 			if (err)
3660 				return notifier_from_errno(err);
3661 			err = snmp6_register_dev(idev);
3662 			if (err) {
3663 				addrconf_sysctl_unregister(idev);
3664 				return notifier_from_errno(err);
3665 			}
3666 		}
3667 		break;
3668 
3669 	case NETDEV_PRE_TYPE_CHANGE:
3670 	case NETDEV_POST_TYPE_CHANGE:
3671 		if (idev)
3672 			addrconf_type_change(dev, event);
3673 		break;
3674 
3675 	case NETDEV_CHANGEUPPER:
3676 		info = ptr;
3677 
3678 		/* flush all routes if dev is linked to or unlinked from
3679 		 * an L3 master device (e.g., VRF)
3680 		 */
3681 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3682 			addrconf_ifdown(dev, false);
3683 	}
3684 
3685 	return NOTIFY_OK;
3686 }
3687 
3688 /*
3689  *	addrconf module should be notified of a device going up
3690  */
3691 static struct notifier_block ipv6_dev_notf = {
3692 	.notifier_call = addrconf_notify,
3693 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3694 };
3695 
3696 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3697 {
3698 	struct inet6_dev *idev;
3699 	ASSERT_RTNL();
3700 
3701 	idev = __in6_dev_get(dev);
3702 
3703 	if (event == NETDEV_POST_TYPE_CHANGE)
3704 		ipv6_mc_remap(idev);
3705 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3706 		ipv6_mc_unmap(idev);
3707 }
3708 
3709 static bool addr_is_local(const struct in6_addr *addr)
3710 {
3711 	return ipv6_addr_type(addr) &
3712 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3713 }
3714 
3715 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3716 {
3717 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3718 	struct net *net = dev_net(dev);
3719 	struct inet6_dev *idev;
3720 	struct inet6_ifaddr *ifa, *tmp;
3721 	bool keep_addr = false;
3722 	int state, i;
3723 
3724 	ASSERT_RTNL();
3725 
3726 	rt6_disable_ip(dev, event);
3727 
3728 	idev = __in6_dev_get(dev);
3729 	if (!idev)
3730 		return -ENODEV;
3731 
3732 	/*
3733 	 * Step 1: remove reference to ipv6 device from parent device.
3734 	 *	   Do not dev_put!
3735 	 */
3736 	if (unregister) {
3737 		idev->dead = 1;
3738 
3739 		/* protected by rtnl_lock */
3740 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3741 
3742 		/* Step 1.5: remove snmp6 entry */
3743 		snmp6_unregister_dev(idev);
3744 
3745 	}
3746 
3747 	/* combine the user config with event to determine if permanent
3748 	 * addresses are to be removed from address hash table
3749 	 */
3750 	if (!unregister && !idev->cnf.disable_ipv6) {
3751 		/* aggregate the system setting and interface setting */
3752 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3753 
3754 		if (!_keep_addr)
3755 			_keep_addr = idev->cnf.keep_addr_on_down;
3756 
3757 		keep_addr = (_keep_addr > 0);
3758 	}
3759 
3760 	/* Step 2: clear hash table */
3761 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3762 		struct hlist_head *h = &inet6_addr_lst[i];
3763 
3764 		spin_lock_bh(&addrconf_hash_lock);
3765 restart:
3766 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3767 			if (ifa->idev == idev) {
3768 				addrconf_del_dad_work(ifa);
3769 				/* combined flag + permanent flag decide if
3770 				 * address is retained on a down event
3771 				 */
3772 				if (!keep_addr ||
3773 				    !(ifa->flags & IFA_F_PERMANENT) ||
3774 				    addr_is_local(&ifa->addr)) {
3775 					hlist_del_init_rcu(&ifa->addr_lst);
3776 					goto restart;
3777 				}
3778 			}
3779 		}
3780 		spin_unlock_bh(&addrconf_hash_lock);
3781 	}
3782 
3783 	write_lock_bh(&idev->lock);
3784 
3785 	addrconf_del_rs_timer(idev);
3786 
3787 	/* Step 2: clear flags for stateless addrconf */
3788 	if (!unregister)
3789 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3790 
3791 	/* Step 3: clear tempaddr list */
3792 	while (!list_empty(&idev->tempaddr_list)) {
3793 		ifa = list_first_entry(&idev->tempaddr_list,
3794 				       struct inet6_ifaddr, tmp_list);
3795 		list_del(&ifa->tmp_list);
3796 		write_unlock_bh(&idev->lock);
3797 		spin_lock_bh(&ifa->lock);
3798 
3799 		if (ifa->ifpub) {
3800 			in6_ifa_put(ifa->ifpub);
3801 			ifa->ifpub = NULL;
3802 		}
3803 		spin_unlock_bh(&ifa->lock);
3804 		in6_ifa_put(ifa);
3805 		write_lock_bh(&idev->lock);
3806 	}
3807 
3808 	list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3809 		struct fib6_info *rt = NULL;
3810 		bool keep;
3811 
3812 		addrconf_del_dad_work(ifa);
3813 
3814 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3815 			!addr_is_local(&ifa->addr);
3816 
3817 		write_unlock_bh(&idev->lock);
3818 		spin_lock_bh(&ifa->lock);
3819 
3820 		if (keep) {
3821 			/* set state to skip the notifier below */
3822 			state = INET6_IFADDR_STATE_DEAD;
3823 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3824 			if (!(ifa->flags & IFA_F_NODAD))
3825 				ifa->flags |= IFA_F_TENTATIVE;
3826 
3827 			rt = ifa->rt;
3828 			ifa->rt = NULL;
3829 		} else {
3830 			state = ifa->state;
3831 			ifa->state = INET6_IFADDR_STATE_DEAD;
3832 		}
3833 
3834 		spin_unlock_bh(&ifa->lock);
3835 
3836 		if (rt)
3837 			ip6_del_rt(net, rt, false);
3838 
3839 		if (state != INET6_IFADDR_STATE_DEAD) {
3840 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3841 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3842 		} else {
3843 			if (idev->cnf.forwarding)
3844 				addrconf_leave_anycast(ifa);
3845 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3846 		}
3847 
3848 		write_lock_bh(&idev->lock);
3849 		if (!keep) {
3850 			list_del_rcu(&ifa->if_list);
3851 			in6_ifa_put(ifa);
3852 		}
3853 	}
3854 
3855 	write_unlock_bh(&idev->lock);
3856 
3857 	/* Step 5: Discard anycast and multicast list */
3858 	if (unregister) {
3859 		ipv6_ac_destroy_dev(idev);
3860 		ipv6_mc_destroy_dev(idev);
3861 	} else {
3862 		ipv6_mc_down(idev);
3863 	}
3864 
3865 	idev->tstamp = jiffies;
3866 	idev->ra_mtu = 0;
3867 
3868 	/* Last: Shot the device (if unregistered) */
3869 	if (unregister) {
3870 		addrconf_sysctl_unregister(idev);
3871 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3872 		neigh_ifdown(&nd_tbl, dev);
3873 		in6_dev_put(idev);
3874 	}
3875 	return 0;
3876 }
3877 
3878 static void addrconf_rs_timer(struct timer_list *t)
3879 {
3880 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3881 	struct net_device *dev = idev->dev;
3882 	struct in6_addr lladdr;
3883 
3884 	write_lock(&idev->lock);
3885 	if (idev->dead || !(idev->if_flags & IF_READY))
3886 		goto out;
3887 
3888 	if (!ipv6_accept_ra(idev))
3889 		goto out;
3890 
3891 	/* Announcement received after solicitation was sent */
3892 	if (idev->if_flags & IF_RA_RCVD)
3893 		goto out;
3894 
3895 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3896 		write_unlock(&idev->lock);
3897 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3898 			ndisc_send_rs(dev, &lladdr,
3899 				      &in6addr_linklocal_allrouters);
3900 		else
3901 			goto put;
3902 
3903 		write_lock(&idev->lock);
3904 		idev->rs_interval = rfc3315_s14_backoff_update(
3905 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3906 		/* The wait after the last probe can be shorter */
3907 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3908 					     idev->cnf.rtr_solicits) ?
3909 				      idev->cnf.rtr_solicit_delay :
3910 				      idev->rs_interval);
3911 	} else {
3912 		/*
3913 		 * Note: we do not support deprecated "all on-link"
3914 		 * assumption any longer.
3915 		 */
3916 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3917 	}
3918 
3919 out:
3920 	write_unlock(&idev->lock);
3921 put:
3922 	in6_dev_put(idev);
3923 }
3924 
3925 /*
3926  *	Duplicate Address Detection
3927  */
3928 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3929 {
3930 	unsigned long rand_num;
3931 	struct inet6_dev *idev = ifp->idev;
3932 	u64 nonce;
3933 
3934 	if (ifp->flags & IFA_F_OPTIMISTIC)
3935 		rand_num = 0;
3936 	else
3937 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3938 
3939 	nonce = 0;
3940 	if (idev->cnf.enhanced_dad ||
3941 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3942 		do
3943 			get_random_bytes(&nonce, 6);
3944 		while (nonce == 0);
3945 	}
3946 	ifp->dad_nonce = nonce;
3947 	ifp->dad_probes = idev->cnf.dad_transmits;
3948 	addrconf_mod_dad_work(ifp, rand_num);
3949 }
3950 
3951 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3952 {
3953 	struct inet6_dev *idev = ifp->idev;
3954 	struct net_device *dev = idev->dev;
3955 	bool bump_id, notify = false;
3956 	struct net *net;
3957 
3958 	addrconf_join_solict(dev, &ifp->addr);
3959 
3960 	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3961 
3962 	read_lock_bh(&idev->lock);
3963 	spin_lock(&ifp->lock);
3964 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3965 		goto out;
3966 
3967 	net = dev_net(dev);
3968 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3969 	    (net->ipv6.devconf_all->accept_dad < 1 &&
3970 	     idev->cnf.accept_dad < 1) ||
3971 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3972 	    ifp->flags & IFA_F_NODAD) {
3973 		bool send_na = false;
3974 
3975 		if (ifp->flags & IFA_F_TENTATIVE &&
3976 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3977 			send_na = true;
3978 		bump_id = ifp->flags & IFA_F_TENTATIVE;
3979 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3980 		spin_unlock(&ifp->lock);
3981 		read_unlock_bh(&idev->lock);
3982 
3983 		addrconf_dad_completed(ifp, bump_id, send_na);
3984 		return;
3985 	}
3986 
3987 	if (!(idev->if_flags & IF_READY)) {
3988 		spin_unlock(&ifp->lock);
3989 		read_unlock_bh(&idev->lock);
3990 		/*
3991 		 * If the device is not ready:
3992 		 * - keep it tentative if it is a permanent address.
3993 		 * - otherwise, kill it.
3994 		 */
3995 		in6_ifa_hold(ifp);
3996 		addrconf_dad_stop(ifp, 0);
3997 		return;
3998 	}
3999 
4000 	/*
4001 	 * Optimistic nodes can start receiving
4002 	 * Frames right away
4003 	 */
4004 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4005 		ip6_ins_rt(net, ifp->rt);
4006 		if (ipv6_use_optimistic_addr(net, idev)) {
4007 			/* Because optimistic nodes can use this address,
4008 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4009 			 */
4010 			notify = true;
4011 		}
4012 	}
4013 
4014 	addrconf_dad_kick(ifp);
4015 out:
4016 	spin_unlock(&ifp->lock);
4017 	read_unlock_bh(&idev->lock);
4018 	if (notify)
4019 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4020 }
4021 
4022 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4023 {
4024 	bool begin_dad = false;
4025 
4026 	spin_lock_bh(&ifp->lock);
4027 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4028 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4029 		begin_dad = true;
4030 	}
4031 	spin_unlock_bh(&ifp->lock);
4032 
4033 	if (begin_dad)
4034 		addrconf_mod_dad_work(ifp, 0);
4035 }
4036 
4037 static void addrconf_dad_work(struct work_struct *w)
4038 {
4039 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4040 						struct inet6_ifaddr,
4041 						dad_work);
4042 	struct inet6_dev *idev = ifp->idev;
4043 	bool bump_id, disable_ipv6 = false;
4044 	struct in6_addr mcaddr;
4045 
4046 	enum {
4047 		DAD_PROCESS,
4048 		DAD_BEGIN,
4049 		DAD_ABORT,
4050 	} action = DAD_PROCESS;
4051 
4052 	rtnl_lock();
4053 
4054 	spin_lock_bh(&ifp->lock);
4055 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4056 		action = DAD_BEGIN;
4057 		ifp->state = INET6_IFADDR_STATE_DAD;
4058 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4059 		action = DAD_ABORT;
4060 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4061 
4062 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4063 		     idev->cnf.accept_dad > 1) &&
4064 		    !idev->cnf.disable_ipv6 &&
4065 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4066 			struct in6_addr addr;
4067 
4068 			addr.s6_addr32[0] = htonl(0xfe800000);
4069 			addr.s6_addr32[1] = 0;
4070 
4071 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4072 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4073 				/* DAD failed for link-local based on MAC */
4074 				idev->cnf.disable_ipv6 = 1;
4075 
4076 				pr_info("%s: IPv6 being disabled!\n",
4077 					ifp->idev->dev->name);
4078 				disable_ipv6 = true;
4079 			}
4080 		}
4081 	}
4082 	spin_unlock_bh(&ifp->lock);
4083 
4084 	if (action == DAD_BEGIN) {
4085 		addrconf_dad_begin(ifp);
4086 		goto out;
4087 	} else if (action == DAD_ABORT) {
4088 		in6_ifa_hold(ifp);
4089 		addrconf_dad_stop(ifp, 1);
4090 		if (disable_ipv6)
4091 			addrconf_ifdown(idev->dev, false);
4092 		goto out;
4093 	}
4094 
4095 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4096 		goto out;
4097 
4098 	write_lock_bh(&idev->lock);
4099 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4100 		write_unlock_bh(&idev->lock);
4101 		goto out;
4102 	}
4103 
4104 	spin_lock(&ifp->lock);
4105 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4106 		spin_unlock(&ifp->lock);
4107 		write_unlock_bh(&idev->lock);
4108 		goto out;
4109 	}
4110 
4111 	if (ifp->dad_probes == 0) {
4112 		bool send_na = false;
4113 
4114 		/*
4115 		 * DAD was successful
4116 		 */
4117 
4118 		if (ifp->flags & IFA_F_TENTATIVE &&
4119 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4120 			send_na = true;
4121 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4122 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4123 		spin_unlock(&ifp->lock);
4124 		write_unlock_bh(&idev->lock);
4125 
4126 		addrconf_dad_completed(ifp, bump_id, send_na);
4127 
4128 		goto out;
4129 	}
4130 
4131 	ifp->dad_probes--;
4132 	addrconf_mod_dad_work(ifp,
4133 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4134 				  HZ/100));
4135 	spin_unlock(&ifp->lock);
4136 	write_unlock_bh(&idev->lock);
4137 
4138 	/* send a neighbour solicitation for our addr */
4139 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4140 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4141 		      ifp->dad_nonce);
4142 out:
4143 	in6_ifa_put(ifp);
4144 	rtnl_unlock();
4145 }
4146 
4147 /* ifp->idev must be at least read locked */
4148 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4149 {
4150 	struct inet6_ifaddr *ifpiter;
4151 	struct inet6_dev *idev = ifp->idev;
4152 
4153 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4154 		if (ifpiter->scope > IFA_LINK)
4155 			break;
4156 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4157 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4158 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4159 		    IFA_F_PERMANENT)
4160 			return false;
4161 	}
4162 	return true;
4163 }
4164 
4165 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4166 				   bool send_na)
4167 {
4168 	struct net_device *dev = ifp->idev->dev;
4169 	struct in6_addr lladdr;
4170 	bool send_rs, send_mld;
4171 
4172 	addrconf_del_dad_work(ifp);
4173 
4174 	/*
4175 	 *	Configure the address for reception. Now it is valid.
4176 	 */
4177 
4178 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4179 
4180 	/* If added prefix is link local and we are prepared to process
4181 	   router advertisements, start sending router solicitations.
4182 	 */
4183 
4184 	read_lock_bh(&ifp->idev->lock);
4185 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4186 	send_rs = send_mld &&
4187 		  ipv6_accept_ra(ifp->idev) &&
4188 		  ifp->idev->cnf.rtr_solicits != 0 &&
4189 		  (dev->flags&IFF_LOOPBACK) == 0;
4190 	read_unlock_bh(&ifp->idev->lock);
4191 
4192 	/* While dad is in progress mld report's source address is in6_addrany.
4193 	 * Resend with proper ll now.
4194 	 */
4195 	if (send_mld)
4196 		ipv6_mc_dad_complete(ifp->idev);
4197 
4198 	/* send unsolicited NA if enabled */
4199 	if (send_na &&
4200 	    (ifp->idev->cnf.ndisc_notify ||
4201 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4202 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4203 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4204 			      /*solicited=*/ false, /*override=*/ true,
4205 			      /*inc_opt=*/ true);
4206 	}
4207 
4208 	if (send_rs) {
4209 		/*
4210 		 *	If a host as already performed a random delay
4211 		 *	[...] as part of DAD [...] there is no need
4212 		 *	to delay again before sending the first RS
4213 		 */
4214 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4215 			return;
4216 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4217 
4218 		write_lock_bh(&ifp->idev->lock);
4219 		spin_lock(&ifp->lock);
4220 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4221 			ifp->idev->cnf.rtr_solicit_interval);
4222 		ifp->idev->rs_probes = 1;
4223 		ifp->idev->if_flags |= IF_RS_SENT;
4224 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4225 		spin_unlock(&ifp->lock);
4226 		write_unlock_bh(&ifp->idev->lock);
4227 	}
4228 
4229 	if (bump_id)
4230 		rt_genid_bump_ipv6(dev_net(dev));
4231 
4232 	/* Make sure that a new temporary address will be created
4233 	 * before this temporary address becomes deprecated.
4234 	 */
4235 	if (ifp->flags & IFA_F_TEMPORARY)
4236 		addrconf_verify_rtnl();
4237 }
4238 
4239 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4240 {
4241 	struct inet6_ifaddr *ifp;
4242 
4243 	read_lock_bh(&idev->lock);
4244 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4245 		spin_lock(&ifp->lock);
4246 		if ((ifp->flags & IFA_F_TENTATIVE &&
4247 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4248 			if (restart)
4249 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4250 			addrconf_dad_kick(ifp);
4251 		}
4252 		spin_unlock(&ifp->lock);
4253 	}
4254 	read_unlock_bh(&idev->lock);
4255 }
4256 
4257 #ifdef CONFIG_PROC_FS
4258 struct if6_iter_state {
4259 	struct seq_net_private p;
4260 	int bucket;
4261 	int offset;
4262 };
4263 
4264 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4265 {
4266 	struct if6_iter_state *state = seq->private;
4267 	struct net *net = seq_file_net(seq);
4268 	struct inet6_ifaddr *ifa = NULL;
4269 	int p = 0;
4270 
4271 	/* initial bucket if pos is 0 */
4272 	if (pos == 0) {
4273 		state->bucket = 0;
4274 		state->offset = 0;
4275 	}
4276 
4277 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4278 		hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4279 					 addr_lst) {
4280 			if (!net_eq(dev_net(ifa->idev->dev), net))
4281 				continue;
4282 			/* sync with offset */
4283 			if (p < state->offset) {
4284 				p++;
4285 				continue;
4286 			}
4287 			return ifa;
4288 		}
4289 
4290 		/* prepare for next bucket */
4291 		state->offset = 0;
4292 		p = 0;
4293 	}
4294 	return NULL;
4295 }
4296 
4297 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4298 					 struct inet6_ifaddr *ifa)
4299 {
4300 	struct if6_iter_state *state = seq->private;
4301 	struct net *net = seq_file_net(seq);
4302 
4303 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4304 		if (!net_eq(dev_net(ifa->idev->dev), net))
4305 			continue;
4306 		state->offset++;
4307 		return ifa;
4308 	}
4309 
4310 	state->offset = 0;
4311 	while (++state->bucket < IN6_ADDR_HSIZE) {
4312 		hlist_for_each_entry_rcu(ifa,
4313 				     &inet6_addr_lst[state->bucket], addr_lst) {
4314 			if (!net_eq(dev_net(ifa->idev->dev), net))
4315 				continue;
4316 			return ifa;
4317 		}
4318 	}
4319 
4320 	return NULL;
4321 }
4322 
4323 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4324 	__acquires(rcu)
4325 {
4326 	rcu_read_lock();
4327 	return if6_get_first(seq, *pos);
4328 }
4329 
4330 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4331 {
4332 	struct inet6_ifaddr *ifa;
4333 
4334 	ifa = if6_get_next(seq, v);
4335 	++*pos;
4336 	return ifa;
4337 }
4338 
4339 static void if6_seq_stop(struct seq_file *seq, void *v)
4340 	__releases(rcu)
4341 {
4342 	rcu_read_unlock();
4343 }
4344 
4345 static int if6_seq_show(struct seq_file *seq, void *v)
4346 {
4347 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4348 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4349 		   &ifp->addr,
4350 		   ifp->idev->dev->ifindex,
4351 		   ifp->prefix_len,
4352 		   ifp->scope,
4353 		   (u8) ifp->flags,
4354 		   ifp->idev->dev->name);
4355 	return 0;
4356 }
4357 
4358 static const struct seq_operations if6_seq_ops = {
4359 	.start	= if6_seq_start,
4360 	.next	= if6_seq_next,
4361 	.show	= if6_seq_show,
4362 	.stop	= if6_seq_stop,
4363 };
4364 
4365 static int __net_init if6_proc_net_init(struct net *net)
4366 {
4367 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4368 			sizeof(struct if6_iter_state)))
4369 		return -ENOMEM;
4370 	return 0;
4371 }
4372 
4373 static void __net_exit if6_proc_net_exit(struct net *net)
4374 {
4375 	remove_proc_entry("if_inet6", net->proc_net);
4376 }
4377 
4378 static struct pernet_operations if6_proc_net_ops = {
4379 	.init = if6_proc_net_init,
4380 	.exit = if6_proc_net_exit,
4381 };
4382 
4383 int __init if6_proc_init(void)
4384 {
4385 	return register_pernet_subsys(&if6_proc_net_ops);
4386 }
4387 
4388 void if6_proc_exit(void)
4389 {
4390 	unregister_pernet_subsys(&if6_proc_net_ops);
4391 }
4392 #endif	/* CONFIG_PROC_FS */
4393 
4394 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4395 /* Check if address is a home address configured on any interface. */
4396 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4397 {
4398 	unsigned int hash = inet6_addr_hash(net, addr);
4399 	struct inet6_ifaddr *ifp = NULL;
4400 	int ret = 0;
4401 
4402 	rcu_read_lock();
4403 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4404 		if (!net_eq(dev_net(ifp->idev->dev), net))
4405 			continue;
4406 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4407 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4408 			ret = 1;
4409 			break;
4410 		}
4411 	}
4412 	rcu_read_unlock();
4413 	return ret;
4414 }
4415 #endif
4416 
4417 /* RFC6554 has some algorithm to avoid loops in segment routing by
4418  * checking if the segments contains any of a local interface address.
4419  *
4420  * Quote:
4421  *
4422  * To detect loops in the SRH, a router MUST determine if the SRH
4423  * includes multiple addresses assigned to any interface on that router.
4424  * If such addresses appear more than once and are separated by at least
4425  * one address not assigned to that router.
4426  */
4427 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4428 			  unsigned char nsegs)
4429 {
4430 	const struct in6_addr *addr;
4431 	int i, ret = 0, found = 0;
4432 	struct inet6_ifaddr *ifp;
4433 	bool separated = false;
4434 	unsigned int hash;
4435 	bool hash_found;
4436 
4437 	rcu_read_lock();
4438 	for (i = 0; i < nsegs; i++) {
4439 		addr = &segs[i];
4440 		hash = inet6_addr_hash(net, addr);
4441 
4442 		hash_found = false;
4443 		hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4444 			if (!net_eq(dev_net(ifp->idev->dev), net))
4445 				continue;
4446 
4447 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4448 				hash_found = true;
4449 				break;
4450 			}
4451 		}
4452 
4453 		if (hash_found) {
4454 			if (found > 1 && separated) {
4455 				ret = 1;
4456 				break;
4457 			}
4458 
4459 			separated = false;
4460 			found++;
4461 		} else {
4462 			separated = true;
4463 		}
4464 	}
4465 	rcu_read_unlock();
4466 
4467 	return ret;
4468 }
4469 
4470 /*
4471  *	Periodic address status verification
4472  */
4473 
4474 static void addrconf_verify_rtnl(void)
4475 {
4476 	unsigned long now, next, next_sec, next_sched;
4477 	struct inet6_ifaddr *ifp;
4478 	int i;
4479 
4480 	ASSERT_RTNL();
4481 
4482 	rcu_read_lock_bh();
4483 	now = jiffies;
4484 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4485 
4486 	cancel_delayed_work(&addr_chk_work);
4487 
4488 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4489 restart:
4490 		hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4491 			unsigned long age;
4492 
4493 			/* When setting preferred_lft to a value not zero or
4494 			 * infinity, while valid_lft is infinity
4495 			 * IFA_F_PERMANENT has a non-infinity life time.
4496 			 */
4497 			if ((ifp->flags & IFA_F_PERMANENT) &&
4498 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4499 				continue;
4500 
4501 			spin_lock(&ifp->lock);
4502 			/* We try to batch several events at once. */
4503 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4504 
4505 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4506 			    age >= ifp->valid_lft) {
4507 				spin_unlock(&ifp->lock);
4508 				in6_ifa_hold(ifp);
4509 				rcu_read_unlock_bh();
4510 				ipv6_del_addr(ifp);
4511 				rcu_read_lock_bh();
4512 				goto restart;
4513 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4514 				spin_unlock(&ifp->lock);
4515 				continue;
4516 			} else if (age >= ifp->prefered_lft) {
4517 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4518 				int deprecate = 0;
4519 
4520 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4521 					deprecate = 1;
4522 					ifp->flags |= IFA_F_DEPRECATED;
4523 				}
4524 
4525 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4526 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4527 					next = ifp->tstamp + ifp->valid_lft * HZ;
4528 
4529 				spin_unlock(&ifp->lock);
4530 
4531 				if (deprecate) {
4532 					in6_ifa_hold(ifp);
4533 
4534 					ipv6_ifa_notify(0, ifp);
4535 					in6_ifa_put(ifp);
4536 					goto restart;
4537 				}
4538 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4539 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4540 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4541 					ifp->idev->cnf.dad_transmits *
4542 					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4543 
4544 				if (age >= ifp->prefered_lft - regen_advance) {
4545 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4546 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4547 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4548 					if (!ifp->regen_count && ifpub) {
4549 						ifp->regen_count++;
4550 						in6_ifa_hold(ifp);
4551 						in6_ifa_hold(ifpub);
4552 						spin_unlock(&ifp->lock);
4553 
4554 						spin_lock(&ifpub->lock);
4555 						ifpub->regen_count = 0;
4556 						spin_unlock(&ifpub->lock);
4557 						rcu_read_unlock_bh();
4558 						ipv6_create_tempaddr(ifpub, true);
4559 						in6_ifa_put(ifpub);
4560 						in6_ifa_put(ifp);
4561 						rcu_read_lock_bh();
4562 						goto restart;
4563 					}
4564 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4565 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4566 				spin_unlock(&ifp->lock);
4567 			} else {
4568 				/* ifp->prefered_lft <= ifp->valid_lft */
4569 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4570 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4571 				spin_unlock(&ifp->lock);
4572 			}
4573 		}
4574 	}
4575 
4576 	next_sec = round_jiffies_up(next);
4577 	next_sched = next;
4578 
4579 	/* If rounded timeout is accurate enough, accept it. */
4580 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4581 		next_sched = next_sec;
4582 
4583 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4584 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4585 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4586 
4587 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4588 		 now, next, next_sec, next_sched);
4589 	mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4590 	rcu_read_unlock_bh();
4591 }
4592 
4593 static void addrconf_verify_work(struct work_struct *w)
4594 {
4595 	rtnl_lock();
4596 	addrconf_verify_rtnl();
4597 	rtnl_unlock();
4598 }
4599 
4600 static void addrconf_verify(void)
4601 {
4602 	mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4603 }
4604 
4605 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4606 				     struct in6_addr **peer_pfx)
4607 {
4608 	struct in6_addr *pfx = NULL;
4609 
4610 	*peer_pfx = NULL;
4611 
4612 	if (addr)
4613 		pfx = nla_data(addr);
4614 
4615 	if (local) {
4616 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4617 			*peer_pfx = pfx;
4618 		pfx = nla_data(local);
4619 	}
4620 
4621 	return pfx;
4622 }
4623 
4624 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4625 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4626 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4627 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4628 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4629 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4630 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4631 };
4632 
4633 static int
4634 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4635 		  struct netlink_ext_ack *extack)
4636 {
4637 	struct net *net = sock_net(skb->sk);
4638 	struct ifaddrmsg *ifm;
4639 	struct nlattr *tb[IFA_MAX+1];
4640 	struct in6_addr *pfx, *peer_pfx;
4641 	u32 ifa_flags;
4642 	int err;
4643 
4644 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4645 				     ifa_ipv6_policy, extack);
4646 	if (err < 0)
4647 		return err;
4648 
4649 	ifm = nlmsg_data(nlh);
4650 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4651 	if (!pfx)
4652 		return -EINVAL;
4653 
4654 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4655 
4656 	/* We ignore other flags so far. */
4657 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4658 
4659 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4660 			      ifm->ifa_prefixlen);
4661 }
4662 
4663 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4664 			       unsigned long expires, u32 flags,
4665 			       bool modify_peer)
4666 {
4667 	struct fib6_info *f6i;
4668 	u32 prio;
4669 
4670 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4671 					ifp->prefix_len,
4672 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4673 	if (!f6i)
4674 		return -ENOENT;
4675 
4676 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4677 	if (f6i->fib6_metric != prio) {
4678 		/* delete old one */
4679 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4680 
4681 		/* add new one */
4682 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4683 				      ifp->prefix_len,
4684 				      ifp->rt_priority, ifp->idev->dev,
4685 				      expires, flags, GFP_KERNEL);
4686 	} else {
4687 		if (!expires)
4688 			fib6_clean_expires(f6i);
4689 		else
4690 			fib6_set_expires(f6i, expires);
4691 
4692 		fib6_info_release(f6i);
4693 	}
4694 
4695 	return 0;
4696 }
4697 
4698 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4699 {
4700 	u32 flags;
4701 	clock_t expires;
4702 	unsigned long timeout;
4703 	bool was_managetempaddr;
4704 	bool had_prefixroute;
4705 	bool new_peer = false;
4706 
4707 	ASSERT_RTNL();
4708 
4709 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4710 		return -EINVAL;
4711 
4712 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4713 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4714 		return -EINVAL;
4715 
4716 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4717 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4718 
4719 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4720 	if (addrconf_finite_timeout(timeout)) {
4721 		expires = jiffies_to_clock_t(timeout * HZ);
4722 		cfg->valid_lft = timeout;
4723 		flags = RTF_EXPIRES;
4724 	} else {
4725 		expires = 0;
4726 		flags = 0;
4727 		cfg->ifa_flags |= IFA_F_PERMANENT;
4728 	}
4729 
4730 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4731 	if (addrconf_finite_timeout(timeout)) {
4732 		if (timeout == 0)
4733 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4734 		cfg->preferred_lft = timeout;
4735 	}
4736 
4737 	if (cfg->peer_pfx &&
4738 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4739 		if (!ipv6_addr_any(&ifp->peer_addr))
4740 			cleanup_prefix_route(ifp, expires, true, true);
4741 		new_peer = true;
4742 	}
4743 
4744 	spin_lock_bh(&ifp->lock);
4745 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4746 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4747 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4748 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4749 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4750 			IFA_F_NOPREFIXROUTE);
4751 	ifp->flags |= cfg->ifa_flags;
4752 	ifp->tstamp = jiffies;
4753 	ifp->valid_lft = cfg->valid_lft;
4754 	ifp->prefered_lft = cfg->preferred_lft;
4755 
4756 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4757 		ifp->rt_priority = cfg->rt_priority;
4758 
4759 	if (new_peer)
4760 		ifp->peer_addr = *cfg->peer_pfx;
4761 
4762 	spin_unlock_bh(&ifp->lock);
4763 	if (!(ifp->flags&IFA_F_TENTATIVE))
4764 		ipv6_ifa_notify(0, ifp);
4765 
4766 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4767 		int rc = -ENOENT;
4768 
4769 		if (had_prefixroute)
4770 			rc = modify_prefix_route(ifp, expires, flags, false);
4771 
4772 		/* prefix route could have been deleted; if so restore it */
4773 		if (rc == -ENOENT) {
4774 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4775 					      ifp->rt_priority, ifp->idev->dev,
4776 					      expires, flags, GFP_KERNEL);
4777 		}
4778 
4779 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4780 			rc = modify_prefix_route(ifp, expires, flags, true);
4781 
4782 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4783 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4784 					      ifp->rt_priority, ifp->idev->dev,
4785 					      expires, flags, GFP_KERNEL);
4786 		}
4787 	} else if (had_prefixroute) {
4788 		enum cleanup_prefix_rt_t action;
4789 		unsigned long rt_expires;
4790 
4791 		write_lock_bh(&ifp->idev->lock);
4792 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4793 		write_unlock_bh(&ifp->idev->lock);
4794 
4795 		if (action != CLEANUP_PREFIX_RT_NOP) {
4796 			cleanup_prefix_route(ifp, rt_expires,
4797 				action == CLEANUP_PREFIX_RT_DEL, false);
4798 		}
4799 	}
4800 
4801 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4802 		if (was_managetempaddr &&
4803 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4804 			cfg->valid_lft = 0;
4805 			cfg->preferred_lft = 0;
4806 		}
4807 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4808 				 cfg->preferred_lft, !was_managetempaddr,
4809 				 jiffies);
4810 	}
4811 
4812 	addrconf_verify_rtnl();
4813 
4814 	return 0;
4815 }
4816 
4817 static int
4818 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4819 		  struct netlink_ext_ack *extack)
4820 {
4821 	struct net *net = sock_net(skb->sk);
4822 	struct ifaddrmsg *ifm;
4823 	struct nlattr *tb[IFA_MAX+1];
4824 	struct in6_addr *peer_pfx;
4825 	struct inet6_ifaddr *ifa;
4826 	struct net_device *dev;
4827 	struct inet6_dev *idev;
4828 	struct ifa6_config cfg;
4829 	int err;
4830 
4831 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4832 				     ifa_ipv6_policy, extack);
4833 	if (err < 0)
4834 		return err;
4835 
4836 	memset(&cfg, 0, sizeof(cfg));
4837 
4838 	ifm = nlmsg_data(nlh);
4839 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4840 	if (!cfg.pfx)
4841 		return -EINVAL;
4842 
4843 	cfg.peer_pfx = peer_pfx;
4844 	cfg.plen = ifm->ifa_prefixlen;
4845 	if (tb[IFA_RT_PRIORITY])
4846 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4847 
4848 	cfg.valid_lft = INFINITY_LIFE_TIME;
4849 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4850 
4851 	if (tb[IFA_CACHEINFO]) {
4852 		struct ifa_cacheinfo *ci;
4853 
4854 		ci = nla_data(tb[IFA_CACHEINFO]);
4855 		cfg.valid_lft = ci->ifa_valid;
4856 		cfg.preferred_lft = ci->ifa_prefered;
4857 	}
4858 
4859 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4860 	if (!dev)
4861 		return -ENODEV;
4862 
4863 	if (tb[IFA_FLAGS])
4864 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4865 	else
4866 		cfg.ifa_flags = ifm->ifa_flags;
4867 
4868 	/* We ignore other flags so far. */
4869 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4870 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4871 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4872 
4873 	idev = ipv6_find_idev(dev);
4874 	if (IS_ERR(idev))
4875 		return PTR_ERR(idev);
4876 
4877 	if (!ipv6_allow_optimistic_dad(net, idev))
4878 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4879 
4880 	if (cfg.ifa_flags & IFA_F_NODAD &&
4881 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4882 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4883 		return -EINVAL;
4884 	}
4885 
4886 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4887 	if (!ifa) {
4888 		/*
4889 		 * It would be best to check for !NLM_F_CREATE here but
4890 		 * userspace already relies on not having to provide this.
4891 		 */
4892 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4893 	}
4894 
4895 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4896 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4897 		err = -EEXIST;
4898 	else
4899 		err = inet6_addr_modify(ifa, &cfg);
4900 
4901 	in6_ifa_put(ifa);
4902 
4903 	return err;
4904 }
4905 
4906 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4907 			  u8 scope, int ifindex)
4908 {
4909 	struct ifaddrmsg *ifm;
4910 
4911 	ifm = nlmsg_data(nlh);
4912 	ifm->ifa_family = AF_INET6;
4913 	ifm->ifa_prefixlen = prefixlen;
4914 	ifm->ifa_flags = flags;
4915 	ifm->ifa_scope = scope;
4916 	ifm->ifa_index = ifindex;
4917 }
4918 
4919 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4920 			 unsigned long tstamp, u32 preferred, u32 valid)
4921 {
4922 	struct ifa_cacheinfo ci;
4923 
4924 	ci.cstamp = cstamp_delta(cstamp);
4925 	ci.tstamp = cstamp_delta(tstamp);
4926 	ci.ifa_prefered = preferred;
4927 	ci.ifa_valid = valid;
4928 
4929 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4930 }
4931 
4932 static inline int rt_scope(int ifa_scope)
4933 {
4934 	if (ifa_scope & IFA_HOST)
4935 		return RT_SCOPE_HOST;
4936 	else if (ifa_scope & IFA_LINK)
4937 		return RT_SCOPE_LINK;
4938 	else if (ifa_scope & IFA_SITE)
4939 		return RT_SCOPE_SITE;
4940 	else
4941 		return RT_SCOPE_UNIVERSE;
4942 }
4943 
4944 static inline int inet6_ifaddr_msgsize(void)
4945 {
4946 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4947 	       + nla_total_size(16) /* IFA_LOCAL */
4948 	       + nla_total_size(16) /* IFA_ADDRESS */
4949 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4950 	       + nla_total_size(4)  /* IFA_FLAGS */
4951 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4952 }
4953 
4954 enum addr_type_t {
4955 	UNICAST_ADDR,
4956 	MULTICAST_ADDR,
4957 	ANYCAST_ADDR,
4958 };
4959 
4960 struct inet6_fill_args {
4961 	u32 portid;
4962 	u32 seq;
4963 	int event;
4964 	unsigned int flags;
4965 	int netnsid;
4966 	int ifindex;
4967 	enum addr_type_t type;
4968 };
4969 
4970 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4971 			     struct inet6_fill_args *args)
4972 {
4973 	struct nlmsghdr  *nlh;
4974 	u32 preferred, valid;
4975 
4976 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4977 			sizeof(struct ifaddrmsg), args->flags);
4978 	if (!nlh)
4979 		return -EMSGSIZE;
4980 
4981 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4982 		      ifa->idev->dev->ifindex);
4983 
4984 	if (args->netnsid >= 0 &&
4985 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4986 		goto error;
4987 
4988 	if (!((ifa->flags&IFA_F_PERMANENT) &&
4989 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4990 		preferred = ifa->prefered_lft;
4991 		valid = ifa->valid_lft;
4992 		if (preferred != INFINITY_LIFE_TIME) {
4993 			long tval = (jiffies - ifa->tstamp)/HZ;
4994 			if (preferred > tval)
4995 				preferred -= tval;
4996 			else
4997 				preferred = 0;
4998 			if (valid != INFINITY_LIFE_TIME) {
4999 				if (valid > tval)
5000 					valid -= tval;
5001 				else
5002 					valid = 0;
5003 			}
5004 		}
5005 	} else {
5006 		preferred = INFINITY_LIFE_TIME;
5007 		valid = INFINITY_LIFE_TIME;
5008 	}
5009 
5010 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5011 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5012 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5013 			goto error;
5014 	} else
5015 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5016 			goto error;
5017 
5018 	if (ifa->rt_priority &&
5019 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5020 		goto error;
5021 
5022 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5023 		goto error;
5024 
5025 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5026 		goto error;
5027 
5028 	nlmsg_end(skb, nlh);
5029 	return 0;
5030 
5031 error:
5032 	nlmsg_cancel(skb, nlh);
5033 	return -EMSGSIZE;
5034 }
5035 
5036 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5037 			       struct inet6_fill_args *args)
5038 {
5039 	struct nlmsghdr  *nlh;
5040 	u8 scope = RT_SCOPE_UNIVERSE;
5041 	int ifindex = ifmca->idev->dev->ifindex;
5042 
5043 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5044 		scope = RT_SCOPE_SITE;
5045 
5046 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5047 			sizeof(struct ifaddrmsg), args->flags);
5048 	if (!nlh)
5049 		return -EMSGSIZE;
5050 
5051 	if (args->netnsid >= 0 &&
5052 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5053 		nlmsg_cancel(skb, nlh);
5054 		return -EMSGSIZE;
5055 	}
5056 
5057 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5058 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5059 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5060 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5061 		nlmsg_cancel(skb, nlh);
5062 		return -EMSGSIZE;
5063 	}
5064 
5065 	nlmsg_end(skb, nlh);
5066 	return 0;
5067 }
5068 
5069 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5070 			       struct inet6_fill_args *args)
5071 {
5072 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5073 	int ifindex = dev ? dev->ifindex : 1;
5074 	struct nlmsghdr  *nlh;
5075 	u8 scope = RT_SCOPE_UNIVERSE;
5076 
5077 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5078 		scope = RT_SCOPE_SITE;
5079 
5080 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5081 			sizeof(struct ifaddrmsg), args->flags);
5082 	if (!nlh)
5083 		return -EMSGSIZE;
5084 
5085 	if (args->netnsid >= 0 &&
5086 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5087 		nlmsg_cancel(skb, nlh);
5088 		return -EMSGSIZE;
5089 	}
5090 
5091 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5092 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5093 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5094 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5095 		nlmsg_cancel(skb, nlh);
5096 		return -EMSGSIZE;
5097 	}
5098 
5099 	nlmsg_end(skb, nlh);
5100 	return 0;
5101 }
5102 
5103 /* called with rcu_read_lock() */
5104 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5105 			  struct netlink_callback *cb, int s_ip_idx,
5106 			  struct inet6_fill_args *fillargs)
5107 {
5108 	struct ifmcaddr6 *ifmca;
5109 	struct ifacaddr6 *ifaca;
5110 	int ip_idx = 0;
5111 	int err = 1;
5112 
5113 	read_lock_bh(&idev->lock);
5114 	switch (fillargs->type) {
5115 	case UNICAST_ADDR: {
5116 		struct inet6_ifaddr *ifa;
5117 		fillargs->event = RTM_NEWADDR;
5118 
5119 		/* unicast address incl. temp addr */
5120 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
5121 			if (ip_idx < s_ip_idx)
5122 				goto next;
5123 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5124 			if (err < 0)
5125 				break;
5126 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5127 next:
5128 			ip_idx++;
5129 		}
5130 		break;
5131 	}
5132 	case MULTICAST_ADDR:
5133 		read_unlock_bh(&idev->lock);
5134 		fillargs->event = RTM_GETMULTICAST;
5135 
5136 		/* multicast address */
5137 		for (ifmca = rcu_dereference(idev->mc_list);
5138 		     ifmca;
5139 		     ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5140 			if (ip_idx < s_ip_idx)
5141 				continue;
5142 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5143 			if (err < 0)
5144 				break;
5145 		}
5146 		read_lock_bh(&idev->lock);
5147 		break;
5148 	case ANYCAST_ADDR:
5149 		fillargs->event = RTM_GETANYCAST;
5150 		/* anycast address */
5151 		for (ifaca = idev->ac_list; ifaca;
5152 		     ifaca = ifaca->aca_next, ip_idx++) {
5153 			if (ip_idx < s_ip_idx)
5154 				continue;
5155 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5156 			if (err < 0)
5157 				break;
5158 		}
5159 		break;
5160 	default:
5161 		break;
5162 	}
5163 	read_unlock_bh(&idev->lock);
5164 	cb->args[2] = ip_idx;
5165 	return err;
5166 }
5167 
5168 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5169 				       struct inet6_fill_args *fillargs,
5170 				       struct net **tgt_net, struct sock *sk,
5171 				       struct netlink_callback *cb)
5172 {
5173 	struct netlink_ext_ack *extack = cb->extack;
5174 	struct nlattr *tb[IFA_MAX+1];
5175 	struct ifaddrmsg *ifm;
5176 	int err, i;
5177 
5178 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5179 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5180 		return -EINVAL;
5181 	}
5182 
5183 	ifm = nlmsg_data(nlh);
5184 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5185 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5186 		return -EINVAL;
5187 	}
5188 
5189 	fillargs->ifindex = ifm->ifa_index;
5190 	if (fillargs->ifindex) {
5191 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5192 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5193 	}
5194 
5195 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5196 					    ifa_ipv6_policy, extack);
5197 	if (err < 0)
5198 		return err;
5199 
5200 	for (i = 0; i <= IFA_MAX; ++i) {
5201 		if (!tb[i])
5202 			continue;
5203 
5204 		if (i == IFA_TARGET_NETNSID) {
5205 			struct net *net;
5206 
5207 			fillargs->netnsid = nla_get_s32(tb[i]);
5208 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5209 			if (IS_ERR(net)) {
5210 				fillargs->netnsid = -1;
5211 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5212 				return PTR_ERR(net);
5213 			}
5214 			*tgt_net = net;
5215 		} else {
5216 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5217 			return -EINVAL;
5218 		}
5219 	}
5220 
5221 	return 0;
5222 }
5223 
5224 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5225 			   enum addr_type_t type)
5226 {
5227 	const struct nlmsghdr *nlh = cb->nlh;
5228 	struct inet6_fill_args fillargs = {
5229 		.portid = NETLINK_CB(cb->skb).portid,
5230 		.seq = cb->nlh->nlmsg_seq,
5231 		.flags = NLM_F_MULTI,
5232 		.netnsid = -1,
5233 		.type = type,
5234 	};
5235 	struct net *tgt_net = sock_net(skb->sk);
5236 	int idx, s_idx, s_ip_idx;
5237 	int h, s_h;
5238 	struct net_device *dev;
5239 	struct inet6_dev *idev;
5240 	struct hlist_head *head;
5241 	int err = 0;
5242 
5243 	s_h = cb->args[0];
5244 	s_idx = idx = cb->args[1];
5245 	s_ip_idx = cb->args[2];
5246 
5247 	if (cb->strict_check) {
5248 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5249 						  skb->sk, cb);
5250 		if (err < 0)
5251 			goto put_tgt_net;
5252 
5253 		err = 0;
5254 		if (fillargs.ifindex) {
5255 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5256 			if (!dev) {
5257 				err = -ENODEV;
5258 				goto put_tgt_net;
5259 			}
5260 			idev = __in6_dev_get(dev);
5261 			if (idev) {
5262 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5263 						     &fillargs);
5264 				if (err > 0)
5265 					err = 0;
5266 			}
5267 			goto put_tgt_net;
5268 		}
5269 	}
5270 
5271 	rcu_read_lock();
5272 	cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5273 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5274 		idx = 0;
5275 		head = &tgt_net->dev_index_head[h];
5276 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5277 			if (idx < s_idx)
5278 				goto cont;
5279 			if (h > s_h || idx > s_idx)
5280 				s_ip_idx = 0;
5281 			idev = __in6_dev_get(dev);
5282 			if (!idev)
5283 				goto cont;
5284 
5285 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5286 					   &fillargs) < 0)
5287 				goto done;
5288 cont:
5289 			idx++;
5290 		}
5291 	}
5292 done:
5293 	rcu_read_unlock();
5294 	cb->args[0] = h;
5295 	cb->args[1] = idx;
5296 put_tgt_net:
5297 	if (fillargs.netnsid >= 0)
5298 		put_net(tgt_net);
5299 
5300 	return skb->len ? : err;
5301 }
5302 
5303 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5304 {
5305 	enum addr_type_t type = UNICAST_ADDR;
5306 
5307 	return inet6_dump_addr(skb, cb, type);
5308 }
5309 
5310 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5311 {
5312 	enum addr_type_t type = MULTICAST_ADDR;
5313 
5314 	return inet6_dump_addr(skb, cb, type);
5315 }
5316 
5317 
5318 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5319 {
5320 	enum addr_type_t type = ANYCAST_ADDR;
5321 
5322 	return inet6_dump_addr(skb, cb, type);
5323 }
5324 
5325 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5326 				       const struct nlmsghdr *nlh,
5327 				       struct nlattr **tb,
5328 				       struct netlink_ext_ack *extack)
5329 {
5330 	struct ifaddrmsg *ifm;
5331 	int i, err;
5332 
5333 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5334 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5335 		return -EINVAL;
5336 	}
5337 
5338 	if (!netlink_strict_get_check(skb))
5339 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5340 					      ifa_ipv6_policy, extack);
5341 
5342 	ifm = nlmsg_data(nlh);
5343 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5344 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5345 		return -EINVAL;
5346 	}
5347 
5348 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5349 					    ifa_ipv6_policy, extack);
5350 	if (err)
5351 		return err;
5352 
5353 	for (i = 0; i <= IFA_MAX; i++) {
5354 		if (!tb[i])
5355 			continue;
5356 
5357 		switch (i) {
5358 		case IFA_TARGET_NETNSID:
5359 		case IFA_ADDRESS:
5360 		case IFA_LOCAL:
5361 			break;
5362 		default:
5363 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5364 			return -EINVAL;
5365 		}
5366 	}
5367 
5368 	return 0;
5369 }
5370 
5371 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5372 			     struct netlink_ext_ack *extack)
5373 {
5374 	struct net *tgt_net = sock_net(in_skb->sk);
5375 	struct inet6_fill_args fillargs = {
5376 		.portid = NETLINK_CB(in_skb).portid,
5377 		.seq = nlh->nlmsg_seq,
5378 		.event = RTM_NEWADDR,
5379 		.flags = 0,
5380 		.netnsid = -1,
5381 	};
5382 	struct ifaddrmsg *ifm;
5383 	struct nlattr *tb[IFA_MAX+1];
5384 	struct in6_addr *addr = NULL, *peer;
5385 	struct net_device *dev = NULL;
5386 	struct inet6_ifaddr *ifa;
5387 	struct sk_buff *skb;
5388 	int err;
5389 
5390 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5391 	if (err < 0)
5392 		return err;
5393 
5394 	if (tb[IFA_TARGET_NETNSID]) {
5395 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5396 
5397 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5398 						  fillargs.netnsid);
5399 		if (IS_ERR(tgt_net))
5400 			return PTR_ERR(tgt_net);
5401 	}
5402 
5403 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5404 	if (!addr)
5405 		return -EINVAL;
5406 
5407 	ifm = nlmsg_data(nlh);
5408 	if (ifm->ifa_index)
5409 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5410 
5411 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5412 	if (!ifa) {
5413 		err = -EADDRNOTAVAIL;
5414 		goto errout;
5415 	}
5416 
5417 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5418 	if (!skb) {
5419 		err = -ENOBUFS;
5420 		goto errout_ifa;
5421 	}
5422 
5423 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5424 	if (err < 0) {
5425 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5426 		WARN_ON(err == -EMSGSIZE);
5427 		kfree_skb(skb);
5428 		goto errout_ifa;
5429 	}
5430 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5431 errout_ifa:
5432 	in6_ifa_put(ifa);
5433 errout:
5434 	dev_put(dev);
5435 	if (fillargs.netnsid >= 0)
5436 		put_net(tgt_net);
5437 
5438 	return err;
5439 }
5440 
5441 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5442 {
5443 	struct sk_buff *skb;
5444 	struct net *net = dev_net(ifa->idev->dev);
5445 	struct inet6_fill_args fillargs = {
5446 		.portid = 0,
5447 		.seq = 0,
5448 		.event = event,
5449 		.flags = 0,
5450 		.netnsid = -1,
5451 	};
5452 	int err = -ENOBUFS;
5453 
5454 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5455 	if (!skb)
5456 		goto errout;
5457 
5458 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5459 	if (err < 0) {
5460 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5461 		WARN_ON(err == -EMSGSIZE);
5462 		kfree_skb(skb);
5463 		goto errout;
5464 	}
5465 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5466 	return;
5467 errout:
5468 	if (err < 0)
5469 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5470 }
5471 
5472 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5473 				__s32 *array, int bytes)
5474 {
5475 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5476 
5477 	memset(array, 0, bytes);
5478 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5479 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5480 	array[DEVCONF_MTU6] = cnf->mtu6;
5481 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5482 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5483 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5484 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5485 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5486 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5487 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5488 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5489 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5490 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5491 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5492 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5493 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5494 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5495 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5496 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5497 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5498 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5499 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5500 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5501 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5502 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5503 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5504 	array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5505 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5506 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5507 #ifdef CONFIG_IPV6_ROUTER_PREF
5508 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5509 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5510 		jiffies_to_msecs(cnf->rtr_probe_interval);
5511 #ifdef CONFIG_IPV6_ROUTE_INFO
5512 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5513 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5514 #endif
5515 #endif
5516 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5517 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5518 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5519 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5520 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5521 #endif
5522 #ifdef CONFIG_IPV6_MROUTE
5523 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5524 #endif
5525 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5526 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5527 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5528 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5529 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5530 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5531 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5532 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5533 	/* we omit DEVCONF_STABLE_SECRET for now */
5534 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5535 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5536 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5537 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5538 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5539 #ifdef CONFIG_IPV6_SEG6_HMAC
5540 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5541 #endif
5542 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5543 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5544 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5545 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5546 	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5547 	array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5548 	array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5549 	array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5550 	array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
5551 }
5552 
5553 static inline size_t inet6_ifla6_size(void)
5554 {
5555 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5556 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5557 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5558 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5559 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5560 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5561 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5562 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5563 	     + 0;
5564 }
5565 
5566 static inline size_t inet6_if_nlmsg_size(void)
5567 {
5568 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5569 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5570 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5571 	       + nla_total_size(4) /* IFLA_MTU */
5572 	       + nla_total_size(4) /* IFLA_LINK */
5573 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5574 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5575 }
5576 
5577 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5578 					int bytes)
5579 {
5580 	int i;
5581 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5582 	BUG_ON(pad < 0);
5583 
5584 	/* Use put_unaligned() because stats may not be aligned for u64. */
5585 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5586 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5587 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5588 
5589 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5590 }
5591 
5592 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5593 					int bytes, size_t syncpoff)
5594 {
5595 	int i, c;
5596 	u64 buff[IPSTATS_MIB_MAX];
5597 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5598 
5599 	BUG_ON(pad < 0);
5600 
5601 	memset(buff, 0, sizeof(buff));
5602 	buff[0] = IPSTATS_MIB_MAX;
5603 
5604 	for_each_possible_cpu(c) {
5605 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5606 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5607 	}
5608 
5609 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5610 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5611 }
5612 
5613 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5614 			     int bytes)
5615 {
5616 	switch (attrtype) {
5617 	case IFLA_INET6_STATS:
5618 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5619 				     offsetof(struct ipstats_mib, syncp));
5620 		break;
5621 	case IFLA_INET6_ICMP6STATS:
5622 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5623 		break;
5624 	}
5625 }
5626 
5627 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5628 				  u32 ext_filter_mask)
5629 {
5630 	struct nlattr *nla;
5631 	struct ifla_cacheinfo ci;
5632 
5633 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5634 		goto nla_put_failure;
5635 	ci.max_reasm_len = IPV6_MAXPLEN;
5636 	ci.tstamp = cstamp_delta(idev->tstamp);
5637 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5638 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5639 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5640 		goto nla_put_failure;
5641 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5642 	if (!nla)
5643 		goto nla_put_failure;
5644 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5645 
5646 	/* XXX - MC not implemented */
5647 
5648 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5649 		return 0;
5650 
5651 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5652 	if (!nla)
5653 		goto nla_put_failure;
5654 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5655 
5656 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5657 	if (!nla)
5658 		goto nla_put_failure;
5659 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5660 
5661 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5662 	if (!nla)
5663 		goto nla_put_failure;
5664 	read_lock_bh(&idev->lock);
5665 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5666 	read_unlock_bh(&idev->lock);
5667 
5668 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5669 		goto nla_put_failure;
5670 
5671 	if (idev->ra_mtu &&
5672 	    nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5673 		goto nla_put_failure;
5674 
5675 	return 0;
5676 
5677 nla_put_failure:
5678 	return -EMSGSIZE;
5679 }
5680 
5681 static size_t inet6_get_link_af_size(const struct net_device *dev,
5682 				     u32 ext_filter_mask)
5683 {
5684 	if (!__in6_dev_get(dev))
5685 		return 0;
5686 
5687 	return inet6_ifla6_size();
5688 }
5689 
5690 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5691 			      u32 ext_filter_mask)
5692 {
5693 	struct inet6_dev *idev = __in6_dev_get(dev);
5694 
5695 	if (!idev)
5696 		return -ENODATA;
5697 
5698 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5699 		return -EMSGSIZE;
5700 
5701 	return 0;
5702 }
5703 
5704 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5705 			     struct netlink_ext_ack *extack)
5706 {
5707 	struct inet6_ifaddr *ifp;
5708 	struct net_device *dev = idev->dev;
5709 	bool clear_token, update_rs = false;
5710 	struct in6_addr ll_addr;
5711 
5712 	ASSERT_RTNL();
5713 
5714 	if (!token)
5715 		return -EINVAL;
5716 
5717 	if (dev->flags & IFF_LOOPBACK) {
5718 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5719 		return -EINVAL;
5720 	}
5721 
5722 	if (dev->flags & IFF_NOARP) {
5723 		NL_SET_ERR_MSG_MOD(extack,
5724 				   "Device does not do neighbour discovery");
5725 		return -EINVAL;
5726 	}
5727 
5728 	if (!ipv6_accept_ra(idev)) {
5729 		NL_SET_ERR_MSG_MOD(extack,
5730 				   "Router advertisement is disabled on device");
5731 		return -EINVAL;
5732 	}
5733 
5734 	if (idev->cnf.rtr_solicits == 0) {
5735 		NL_SET_ERR_MSG(extack,
5736 			       "Router solicitation is disabled on device");
5737 		return -EINVAL;
5738 	}
5739 
5740 	write_lock_bh(&idev->lock);
5741 
5742 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5743 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5744 
5745 	write_unlock_bh(&idev->lock);
5746 
5747 	clear_token = ipv6_addr_any(token);
5748 	if (clear_token)
5749 		goto update_lft;
5750 
5751 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5752 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5753 			     IFA_F_OPTIMISTIC)) {
5754 		/* If we're not ready, then normal ifup will take care
5755 		 * of this. Otherwise, we need to request our rs here.
5756 		 */
5757 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5758 		update_rs = true;
5759 	}
5760 
5761 update_lft:
5762 	write_lock_bh(&idev->lock);
5763 
5764 	if (update_rs) {
5765 		idev->if_flags |= IF_RS_SENT;
5766 		idev->rs_interval = rfc3315_s14_backoff_init(
5767 			idev->cnf.rtr_solicit_interval);
5768 		idev->rs_probes = 1;
5769 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5770 	}
5771 
5772 	/* Well, that's kinda nasty ... */
5773 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5774 		spin_lock(&ifp->lock);
5775 		if (ifp->tokenized) {
5776 			ifp->valid_lft = 0;
5777 			ifp->prefered_lft = 0;
5778 		}
5779 		spin_unlock(&ifp->lock);
5780 	}
5781 
5782 	write_unlock_bh(&idev->lock);
5783 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5784 	addrconf_verify_rtnl();
5785 	return 0;
5786 }
5787 
5788 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5789 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5790 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5791 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5792 					    .reject_message =
5793 						"IFLA_INET6_RA_MTU can not be set" },
5794 };
5795 
5796 static int check_addr_gen_mode(int mode)
5797 {
5798 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5799 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5800 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5801 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5802 		return -EINVAL;
5803 	return 1;
5804 }
5805 
5806 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5807 				int mode)
5808 {
5809 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5810 	    !idev->cnf.stable_secret.initialized &&
5811 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5812 		return -EINVAL;
5813 	return 1;
5814 }
5815 
5816 static int inet6_validate_link_af(const struct net_device *dev,
5817 				  const struct nlattr *nla,
5818 				  struct netlink_ext_ack *extack)
5819 {
5820 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5821 	struct inet6_dev *idev = NULL;
5822 	int err;
5823 
5824 	if (dev) {
5825 		idev = __in6_dev_get(dev);
5826 		if (!idev)
5827 			return -EAFNOSUPPORT;
5828 	}
5829 
5830 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5831 					  inet6_af_policy, extack);
5832 	if (err)
5833 		return err;
5834 
5835 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5836 		return -EINVAL;
5837 
5838 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5839 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5840 
5841 		if (check_addr_gen_mode(mode) < 0)
5842 			return -EINVAL;
5843 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5844 			return -EINVAL;
5845 	}
5846 
5847 	return 0;
5848 }
5849 
5850 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5851 			     struct netlink_ext_ack *extack)
5852 {
5853 	struct inet6_dev *idev = __in6_dev_get(dev);
5854 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5855 	int err;
5856 
5857 	if (!idev)
5858 		return -EAFNOSUPPORT;
5859 
5860 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5861 		return -EINVAL;
5862 
5863 	if (tb[IFLA_INET6_TOKEN]) {
5864 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5865 					extack);
5866 		if (err)
5867 			return err;
5868 	}
5869 
5870 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5871 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5872 
5873 		idev->cnf.addr_gen_mode = mode;
5874 	}
5875 
5876 	return 0;
5877 }
5878 
5879 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5880 			     u32 portid, u32 seq, int event, unsigned int flags)
5881 {
5882 	struct net_device *dev = idev->dev;
5883 	struct ifinfomsg *hdr;
5884 	struct nlmsghdr *nlh;
5885 	void *protoinfo;
5886 
5887 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5888 	if (!nlh)
5889 		return -EMSGSIZE;
5890 
5891 	hdr = nlmsg_data(nlh);
5892 	hdr->ifi_family = AF_INET6;
5893 	hdr->__ifi_pad = 0;
5894 	hdr->ifi_type = dev->type;
5895 	hdr->ifi_index = dev->ifindex;
5896 	hdr->ifi_flags = dev_get_flags(dev);
5897 	hdr->ifi_change = 0;
5898 
5899 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5900 	    (dev->addr_len &&
5901 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5902 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5903 	    (dev->ifindex != dev_get_iflink(dev) &&
5904 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5905 	    nla_put_u8(skb, IFLA_OPERSTATE,
5906 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5907 		goto nla_put_failure;
5908 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5909 	if (!protoinfo)
5910 		goto nla_put_failure;
5911 
5912 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5913 		goto nla_put_failure;
5914 
5915 	nla_nest_end(skb, protoinfo);
5916 	nlmsg_end(skb, nlh);
5917 	return 0;
5918 
5919 nla_put_failure:
5920 	nlmsg_cancel(skb, nlh);
5921 	return -EMSGSIZE;
5922 }
5923 
5924 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5925 				   struct netlink_ext_ack *extack)
5926 {
5927 	struct ifinfomsg *ifm;
5928 
5929 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5930 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5931 		return -EINVAL;
5932 	}
5933 
5934 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5935 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5936 		return -EINVAL;
5937 	}
5938 
5939 	ifm = nlmsg_data(nlh);
5940 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5941 	    ifm->ifi_change || ifm->ifi_index) {
5942 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5943 		return -EINVAL;
5944 	}
5945 
5946 	return 0;
5947 }
5948 
5949 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5950 {
5951 	struct net *net = sock_net(skb->sk);
5952 	int h, s_h;
5953 	int idx = 0, s_idx;
5954 	struct net_device *dev;
5955 	struct inet6_dev *idev;
5956 	struct hlist_head *head;
5957 
5958 	/* only requests using strict checking can pass data to
5959 	 * influence the dump
5960 	 */
5961 	if (cb->strict_check) {
5962 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5963 
5964 		if (err < 0)
5965 			return err;
5966 	}
5967 
5968 	s_h = cb->args[0];
5969 	s_idx = cb->args[1];
5970 
5971 	rcu_read_lock();
5972 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5973 		idx = 0;
5974 		head = &net->dev_index_head[h];
5975 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5976 			if (idx < s_idx)
5977 				goto cont;
5978 			idev = __in6_dev_get(dev);
5979 			if (!idev)
5980 				goto cont;
5981 			if (inet6_fill_ifinfo(skb, idev,
5982 					      NETLINK_CB(cb->skb).portid,
5983 					      cb->nlh->nlmsg_seq,
5984 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
5985 				goto out;
5986 cont:
5987 			idx++;
5988 		}
5989 	}
5990 out:
5991 	rcu_read_unlock();
5992 	cb->args[1] = idx;
5993 	cb->args[0] = h;
5994 
5995 	return skb->len;
5996 }
5997 
5998 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5999 {
6000 	struct sk_buff *skb;
6001 	struct net *net = dev_net(idev->dev);
6002 	int err = -ENOBUFS;
6003 
6004 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6005 	if (!skb)
6006 		goto errout;
6007 
6008 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6009 	if (err < 0) {
6010 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6011 		WARN_ON(err == -EMSGSIZE);
6012 		kfree_skb(skb);
6013 		goto errout;
6014 	}
6015 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6016 	return;
6017 errout:
6018 	if (err < 0)
6019 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6020 }
6021 
6022 static inline size_t inet6_prefix_nlmsg_size(void)
6023 {
6024 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6025 	       + nla_total_size(sizeof(struct in6_addr))
6026 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6027 }
6028 
6029 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6030 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6031 			     int event, unsigned int flags)
6032 {
6033 	struct prefixmsg *pmsg;
6034 	struct nlmsghdr *nlh;
6035 	struct prefix_cacheinfo	ci;
6036 
6037 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6038 	if (!nlh)
6039 		return -EMSGSIZE;
6040 
6041 	pmsg = nlmsg_data(nlh);
6042 	pmsg->prefix_family = AF_INET6;
6043 	pmsg->prefix_pad1 = 0;
6044 	pmsg->prefix_pad2 = 0;
6045 	pmsg->prefix_ifindex = idev->dev->ifindex;
6046 	pmsg->prefix_len = pinfo->prefix_len;
6047 	pmsg->prefix_type = pinfo->type;
6048 	pmsg->prefix_pad3 = 0;
6049 	pmsg->prefix_flags = 0;
6050 	if (pinfo->onlink)
6051 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6052 	if (pinfo->autoconf)
6053 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6054 
6055 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6056 		goto nla_put_failure;
6057 	ci.preferred_time = ntohl(pinfo->prefered);
6058 	ci.valid_time = ntohl(pinfo->valid);
6059 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6060 		goto nla_put_failure;
6061 	nlmsg_end(skb, nlh);
6062 	return 0;
6063 
6064 nla_put_failure:
6065 	nlmsg_cancel(skb, nlh);
6066 	return -EMSGSIZE;
6067 }
6068 
6069 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6070 			 struct prefix_info *pinfo)
6071 {
6072 	struct sk_buff *skb;
6073 	struct net *net = dev_net(idev->dev);
6074 	int err = -ENOBUFS;
6075 
6076 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6077 	if (!skb)
6078 		goto errout;
6079 
6080 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6081 	if (err < 0) {
6082 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6083 		WARN_ON(err == -EMSGSIZE);
6084 		kfree_skb(skb);
6085 		goto errout;
6086 	}
6087 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6088 	return;
6089 errout:
6090 	if (err < 0)
6091 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6092 }
6093 
6094 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6095 {
6096 	struct net *net = dev_net(ifp->idev->dev);
6097 
6098 	if (event)
6099 		ASSERT_RTNL();
6100 
6101 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6102 
6103 	switch (event) {
6104 	case RTM_NEWADDR:
6105 		/*
6106 		 * If the address was optimistic we inserted the route at the
6107 		 * start of our DAD process, so we don't need to do it again.
6108 		 * If the device was taken down in the middle of the DAD
6109 		 * cycle there is a race where we could get here without a
6110 		 * host route, so nothing to insert. That will be fixed when
6111 		 * the device is brought up.
6112 		 */
6113 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6114 			ip6_ins_rt(net, ifp->rt);
6115 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6116 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6117 				&ifp->addr, ifp->idev->dev->name);
6118 		}
6119 
6120 		if (ifp->idev->cnf.forwarding)
6121 			addrconf_join_anycast(ifp);
6122 		if (!ipv6_addr_any(&ifp->peer_addr))
6123 			addrconf_prefix_route(&ifp->peer_addr, 128,
6124 					      ifp->rt_priority, ifp->idev->dev,
6125 					      0, 0, GFP_ATOMIC);
6126 		break;
6127 	case RTM_DELADDR:
6128 		if (ifp->idev->cnf.forwarding)
6129 			addrconf_leave_anycast(ifp);
6130 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6131 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6132 			struct fib6_info *rt;
6133 
6134 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6135 						       ifp->idev->dev, 0, 0,
6136 						       false);
6137 			if (rt)
6138 				ip6_del_rt(net, rt, false);
6139 		}
6140 		if (ifp->rt) {
6141 			ip6_del_rt(net, ifp->rt, false);
6142 			ifp->rt = NULL;
6143 		}
6144 		rt_genid_bump_ipv6(net);
6145 		break;
6146 	}
6147 	atomic_inc(&net->ipv6.dev_addr_genid);
6148 }
6149 
6150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6151 {
6152 	if (likely(ifp->idev->dead == 0))
6153 		__ipv6_ifa_notify(event, ifp);
6154 }
6155 
6156 #ifdef CONFIG_SYSCTL
6157 
6158 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6159 		void *buffer, size_t *lenp, loff_t *ppos)
6160 {
6161 	int *valp = ctl->data;
6162 	int val = *valp;
6163 	loff_t pos = *ppos;
6164 	struct ctl_table lctl;
6165 	int ret;
6166 
6167 	/*
6168 	 * ctl->data points to idev->cnf.forwarding, we should
6169 	 * not modify it until we get the rtnl lock.
6170 	 */
6171 	lctl = *ctl;
6172 	lctl.data = &val;
6173 
6174 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6175 
6176 	if (write)
6177 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6178 	if (ret)
6179 		*ppos = pos;
6180 	return ret;
6181 }
6182 
6183 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6184 		void *buffer, size_t *lenp, loff_t *ppos)
6185 {
6186 	struct inet6_dev *idev = ctl->extra1;
6187 	int min_mtu = IPV6_MIN_MTU;
6188 	struct ctl_table lctl;
6189 
6190 	lctl = *ctl;
6191 	lctl.extra1 = &min_mtu;
6192 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6193 
6194 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6195 }
6196 
6197 static void dev_disable_change(struct inet6_dev *idev)
6198 {
6199 	struct netdev_notifier_info info;
6200 
6201 	if (!idev || !idev->dev)
6202 		return;
6203 
6204 	netdev_notifier_info_init(&info, idev->dev);
6205 	if (idev->cnf.disable_ipv6)
6206 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6207 	else
6208 		addrconf_notify(NULL, NETDEV_UP, &info);
6209 }
6210 
6211 static void addrconf_disable_change(struct net *net, __s32 newf)
6212 {
6213 	struct net_device *dev;
6214 	struct inet6_dev *idev;
6215 
6216 	for_each_netdev(net, dev) {
6217 		idev = __in6_dev_get(dev);
6218 		if (idev) {
6219 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6220 			idev->cnf.disable_ipv6 = newf;
6221 			if (changed)
6222 				dev_disable_change(idev);
6223 		}
6224 	}
6225 }
6226 
6227 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6228 {
6229 	struct net *net;
6230 	int old;
6231 
6232 	if (!rtnl_trylock())
6233 		return restart_syscall();
6234 
6235 	net = (struct net *)table->extra2;
6236 	old = *p;
6237 	*p = newf;
6238 
6239 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6240 		rtnl_unlock();
6241 		return 0;
6242 	}
6243 
6244 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6245 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
6246 		addrconf_disable_change(net, newf);
6247 	} else if ((!newf) ^ (!old))
6248 		dev_disable_change((struct inet6_dev *)table->extra1);
6249 
6250 	rtnl_unlock();
6251 	return 0;
6252 }
6253 
6254 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6255 		void *buffer, size_t *lenp, loff_t *ppos)
6256 {
6257 	int *valp = ctl->data;
6258 	int val = *valp;
6259 	loff_t pos = *ppos;
6260 	struct ctl_table lctl;
6261 	int ret;
6262 
6263 	/*
6264 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6265 	 * not modify it until we get the rtnl lock.
6266 	 */
6267 	lctl = *ctl;
6268 	lctl.data = &val;
6269 
6270 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6271 
6272 	if (write)
6273 		ret = addrconf_disable_ipv6(ctl, valp, val);
6274 	if (ret)
6275 		*ppos = pos;
6276 	return ret;
6277 }
6278 
6279 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6280 		void *buffer, size_t *lenp, loff_t *ppos)
6281 {
6282 	int *valp = ctl->data;
6283 	int ret;
6284 	int old, new;
6285 
6286 	old = *valp;
6287 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6288 	new = *valp;
6289 
6290 	if (write && old != new) {
6291 		struct net *net = ctl->extra2;
6292 
6293 		if (!rtnl_trylock())
6294 			return restart_syscall();
6295 
6296 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6297 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6298 						     NETCONFA_PROXY_NEIGH,
6299 						     NETCONFA_IFINDEX_DEFAULT,
6300 						     net->ipv6.devconf_dflt);
6301 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6302 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6303 						     NETCONFA_PROXY_NEIGH,
6304 						     NETCONFA_IFINDEX_ALL,
6305 						     net->ipv6.devconf_all);
6306 		else {
6307 			struct inet6_dev *idev = ctl->extra1;
6308 
6309 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6310 						     NETCONFA_PROXY_NEIGH,
6311 						     idev->dev->ifindex,
6312 						     &idev->cnf);
6313 		}
6314 		rtnl_unlock();
6315 	}
6316 
6317 	return ret;
6318 }
6319 
6320 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6321 					 void *buffer, size_t *lenp,
6322 					 loff_t *ppos)
6323 {
6324 	int ret = 0;
6325 	u32 new_val;
6326 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6327 	struct net *net = (struct net *)ctl->extra2;
6328 	struct ctl_table tmp = {
6329 		.data = &new_val,
6330 		.maxlen = sizeof(new_val),
6331 		.mode = ctl->mode,
6332 	};
6333 
6334 	if (!rtnl_trylock())
6335 		return restart_syscall();
6336 
6337 	new_val = *((u32 *)ctl->data);
6338 
6339 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6340 	if (ret != 0)
6341 		goto out;
6342 
6343 	if (write) {
6344 		if (check_addr_gen_mode(new_val) < 0) {
6345 			ret = -EINVAL;
6346 			goto out;
6347 		}
6348 
6349 		if (idev) {
6350 			if (check_stable_privacy(idev, net, new_val) < 0) {
6351 				ret = -EINVAL;
6352 				goto out;
6353 			}
6354 
6355 			if (idev->cnf.addr_gen_mode != new_val) {
6356 				idev->cnf.addr_gen_mode = new_val;
6357 				addrconf_dev_config(idev->dev);
6358 			}
6359 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6360 			struct net_device *dev;
6361 
6362 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6363 			for_each_netdev(net, dev) {
6364 				idev = __in6_dev_get(dev);
6365 				if (idev &&
6366 				    idev->cnf.addr_gen_mode != new_val) {
6367 					idev->cnf.addr_gen_mode = new_val;
6368 					addrconf_dev_config(idev->dev);
6369 				}
6370 			}
6371 		}
6372 
6373 		*((u32 *)ctl->data) = new_val;
6374 	}
6375 
6376 out:
6377 	rtnl_unlock();
6378 
6379 	return ret;
6380 }
6381 
6382 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6383 					 void *buffer, size_t *lenp,
6384 					 loff_t *ppos)
6385 {
6386 	int err;
6387 	struct in6_addr addr;
6388 	char str[IPV6_MAX_STRLEN];
6389 	struct ctl_table lctl = *ctl;
6390 	struct net *net = ctl->extra2;
6391 	struct ipv6_stable_secret *secret = ctl->data;
6392 
6393 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6394 		return -EIO;
6395 
6396 	lctl.maxlen = IPV6_MAX_STRLEN;
6397 	lctl.data = str;
6398 
6399 	if (!rtnl_trylock())
6400 		return restart_syscall();
6401 
6402 	if (!write && !secret->initialized) {
6403 		err = -EIO;
6404 		goto out;
6405 	}
6406 
6407 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6408 	if (err >= sizeof(str)) {
6409 		err = -EIO;
6410 		goto out;
6411 	}
6412 
6413 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6414 	if (err || !write)
6415 		goto out;
6416 
6417 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6418 		err = -EIO;
6419 		goto out;
6420 	}
6421 
6422 	secret->initialized = true;
6423 	secret->secret = addr;
6424 
6425 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6426 		struct net_device *dev;
6427 
6428 		for_each_netdev(net, dev) {
6429 			struct inet6_dev *idev = __in6_dev_get(dev);
6430 
6431 			if (idev) {
6432 				idev->cnf.addr_gen_mode =
6433 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6434 			}
6435 		}
6436 	} else {
6437 		struct inet6_dev *idev = ctl->extra1;
6438 
6439 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6440 	}
6441 
6442 out:
6443 	rtnl_unlock();
6444 
6445 	return err;
6446 }
6447 
6448 static
6449 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6450 						int write, void *buffer,
6451 						size_t *lenp,
6452 						loff_t *ppos)
6453 {
6454 	int *valp = ctl->data;
6455 	int val = *valp;
6456 	loff_t pos = *ppos;
6457 	struct ctl_table lctl;
6458 	int ret;
6459 
6460 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6461 	 * we should not modify it until we get the rtnl lock.
6462 	 */
6463 	lctl = *ctl;
6464 	lctl.data = &val;
6465 
6466 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6467 
6468 	if (write)
6469 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6470 	if (ret)
6471 		*ppos = pos;
6472 	return ret;
6473 }
6474 
6475 static
6476 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6477 {
6478 	if (rt) {
6479 		if (action)
6480 			rt->dst.flags |= DST_NOPOLICY;
6481 		else
6482 			rt->dst.flags &= ~DST_NOPOLICY;
6483 	}
6484 }
6485 
6486 static
6487 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6488 {
6489 	struct inet6_ifaddr *ifa;
6490 
6491 	read_lock_bh(&idev->lock);
6492 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6493 		spin_lock(&ifa->lock);
6494 		if (ifa->rt) {
6495 			/* host routes only use builtin fib6_nh */
6496 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6497 			int cpu;
6498 
6499 			rcu_read_lock();
6500 			ifa->rt->dst_nopolicy = val ? true : false;
6501 			if (nh->rt6i_pcpu) {
6502 				for_each_possible_cpu(cpu) {
6503 					struct rt6_info **rtp;
6504 
6505 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6506 					addrconf_set_nopolicy(*rtp, val);
6507 				}
6508 			}
6509 			rcu_read_unlock();
6510 		}
6511 		spin_unlock(&ifa->lock);
6512 	}
6513 	read_unlock_bh(&idev->lock);
6514 }
6515 
6516 static
6517 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6518 {
6519 	struct inet6_dev *idev;
6520 	struct net *net;
6521 
6522 	if (!rtnl_trylock())
6523 		return restart_syscall();
6524 
6525 	*valp = val;
6526 
6527 	net = (struct net *)ctl->extra2;
6528 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6529 		rtnl_unlock();
6530 		return 0;
6531 	}
6532 
6533 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6534 		struct net_device *dev;
6535 
6536 		for_each_netdev(net, dev) {
6537 			idev = __in6_dev_get(dev);
6538 			if (idev)
6539 				addrconf_disable_policy_idev(idev, val);
6540 		}
6541 	} else {
6542 		idev = (struct inet6_dev *)ctl->extra1;
6543 		addrconf_disable_policy_idev(idev, val);
6544 	}
6545 
6546 	rtnl_unlock();
6547 	return 0;
6548 }
6549 
6550 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6551 				   void *buffer, size_t *lenp, loff_t *ppos)
6552 {
6553 	int *valp = ctl->data;
6554 	int val = *valp;
6555 	loff_t pos = *ppos;
6556 	struct ctl_table lctl;
6557 	int ret;
6558 
6559 	lctl = *ctl;
6560 	lctl.data = &val;
6561 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6562 
6563 	if (write && (*valp != val))
6564 		ret = addrconf_disable_policy(ctl, valp, val);
6565 
6566 	if (ret)
6567 		*ppos = pos;
6568 
6569 	return ret;
6570 }
6571 
6572 static int minus_one = -1;
6573 static const int two_five_five = 255;
6574 static u32 ioam6_if_id_max = U16_MAX;
6575 
6576 static const struct ctl_table addrconf_sysctl[] = {
6577 	{
6578 		.procname	= "forwarding",
6579 		.data		= &ipv6_devconf.forwarding,
6580 		.maxlen		= sizeof(int),
6581 		.mode		= 0644,
6582 		.proc_handler	= addrconf_sysctl_forward,
6583 	},
6584 	{
6585 		.procname	= "hop_limit",
6586 		.data		= &ipv6_devconf.hop_limit,
6587 		.maxlen		= sizeof(int),
6588 		.mode		= 0644,
6589 		.proc_handler	= proc_dointvec_minmax,
6590 		.extra1		= (void *)SYSCTL_ONE,
6591 		.extra2		= (void *)&two_five_five,
6592 	},
6593 	{
6594 		.procname	= "mtu",
6595 		.data		= &ipv6_devconf.mtu6,
6596 		.maxlen		= sizeof(int),
6597 		.mode		= 0644,
6598 		.proc_handler	= addrconf_sysctl_mtu,
6599 	},
6600 	{
6601 		.procname	= "accept_ra",
6602 		.data		= &ipv6_devconf.accept_ra,
6603 		.maxlen		= sizeof(int),
6604 		.mode		= 0644,
6605 		.proc_handler	= proc_dointvec,
6606 	},
6607 	{
6608 		.procname	= "accept_redirects",
6609 		.data		= &ipv6_devconf.accept_redirects,
6610 		.maxlen		= sizeof(int),
6611 		.mode		= 0644,
6612 		.proc_handler	= proc_dointvec,
6613 	},
6614 	{
6615 		.procname	= "autoconf",
6616 		.data		= &ipv6_devconf.autoconf,
6617 		.maxlen		= sizeof(int),
6618 		.mode		= 0644,
6619 		.proc_handler	= proc_dointvec,
6620 	},
6621 	{
6622 		.procname	= "dad_transmits",
6623 		.data		= &ipv6_devconf.dad_transmits,
6624 		.maxlen		= sizeof(int),
6625 		.mode		= 0644,
6626 		.proc_handler	= proc_dointvec,
6627 	},
6628 	{
6629 		.procname	= "router_solicitations",
6630 		.data		= &ipv6_devconf.rtr_solicits,
6631 		.maxlen		= sizeof(int),
6632 		.mode		= 0644,
6633 		.proc_handler	= proc_dointvec_minmax,
6634 		.extra1		= &minus_one,
6635 	},
6636 	{
6637 		.procname	= "router_solicitation_interval",
6638 		.data		= &ipv6_devconf.rtr_solicit_interval,
6639 		.maxlen		= sizeof(int),
6640 		.mode		= 0644,
6641 		.proc_handler	= proc_dointvec_jiffies,
6642 	},
6643 	{
6644 		.procname	= "router_solicitation_max_interval",
6645 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6646 		.maxlen		= sizeof(int),
6647 		.mode		= 0644,
6648 		.proc_handler	= proc_dointvec_jiffies,
6649 	},
6650 	{
6651 		.procname	= "router_solicitation_delay",
6652 		.data		= &ipv6_devconf.rtr_solicit_delay,
6653 		.maxlen		= sizeof(int),
6654 		.mode		= 0644,
6655 		.proc_handler	= proc_dointvec_jiffies,
6656 	},
6657 	{
6658 		.procname	= "force_mld_version",
6659 		.data		= &ipv6_devconf.force_mld_version,
6660 		.maxlen		= sizeof(int),
6661 		.mode		= 0644,
6662 		.proc_handler	= proc_dointvec,
6663 	},
6664 	{
6665 		.procname	= "mldv1_unsolicited_report_interval",
6666 		.data		=
6667 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6668 		.maxlen		= sizeof(int),
6669 		.mode		= 0644,
6670 		.proc_handler	= proc_dointvec_ms_jiffies,
6671 	},
6672 	{
6673 		.procname	= "mldv2_unsolicited_report_interval",
6674 		.data		=
6675 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6676 		.maxlen		= sizeof(int),
6677 		.mode		= 0644,
6678 		.proc_handler	= proc_dointvec_ms_jiffies,
6679 	},
6680 	{
6681 		.procname	= "use_tempaddr",
6682 		.data		= &ipv6_devconf.use_tempaddr,
6683 		.maxlen		= sizeof(int),
6684 		.mode		= 0644,
6685 		.proc_handler	= proc_dointvec,
6686 	},
6687 	{
6688 		.procname	= "temp_valid_lft",
6689 		.data		= &ipv6_devconf.temp_valid_lft,
6690 		.maxlen		= sizeof(int),
6691 		.mode		= 0644,
6692 		.proc_handler	= proc_dointvec,
6693 	},
6694 	{
6695 		.procname	= "temp_prefered_lft",
6696 		.data		= &ipv6_devconf.temp_prefered_lft,
6697 		.maxlen		= sizeof(int),
6698 		.mode		= 0644,
6699 		.proc_handler	= proc_dointvec,
6700 	},
6701 	{
6702 		.procname	= "regen_max_retry",
6703 		.data		= &ipv6_devconf.regen_max_retry,
6704 		.maxlen		= sizeof(int),
6705 		.mode		= 0644,
6706 		.proc_handler	= proc_dointvec,
6707 	},
6708 	{
6709 		.procname	= "max_desync_factor",
6710 		.data		= &ipv6_devconf.max_desync_factor,
6711 		.maxlen		= sizeof(int),
6712 		.mode		= 0644,
6713 		.proc_handler	= proc_dointvec,
6714 	},
6715 	{
6716 		.procname	= "max_addresses",
6717 		.data		= &ipv6_devconf.max_addresses,
6718 		.maxlen		= sizeof(int),
6719 		.mode		= 0644,
6720 		.proc_handler	= proc_dointvec,
6721 	},
6722 	{
6723 		.procname	= "accept_ra_defrtr",
6724 		.data		= &ipv6_devconf.accept_ra_defrtr,
6725 		.maxlen		= sizeof(int),
6726 		.mode		= 0644,
6727 		.proc_handler	= proc_dointvec,
6728 	},
6729 	{
6730 		.procname	= "ra_defrtr_metric",
6731 		.data		= &ipv6_devconf.ra_defrtr_metric,
6732 		.maxlen		= sizeof(u32),
6733 		.mode		= 0644,
6734 		.proc_handler	= proc_douintvec_minmax,
6735 		.extra1		= (void *)SYSCTL_ONE,
6736 	},
6737 	{
6738 		.procname	= "accept_ra_min_hop_limit",
6739 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6740 		.maxlen		= sizeof(int),
6741 		.mode		= 0644,
6742 		.proc_handler	= proc_dointvec,
6743 	},
6744 	{
6745 		.procname	= "accept_ra_pinfo",
6746 		.data		= &ipv6_devconf.accept_ra_pinfo,
6747 		.maxlen		= sizeof(int),
6748 		.mode		= 0644,
6749 		.proc_handler	= proc_dointvec,
6750 	},
6751 #ifdef CONFIG_IPV6_ROUTER_PREF
6752 	{
6753 		.procname	= "accept_ra_rtr_pref",
6754 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6755 		.maxlen		= sizeof(int),
6756 		.mode		= 0644,
6757 		.proc_handler	= proc_dointvec,
6758 	},
6759 	{
6760 		.procname	= "router_probe_interval",
6761 		.data		= &ipv6_devconf.rtr_probe_interval,
6762 		.maxlen		= sizeof(int),
6763 		.mode		= 0644,
6764 		.proc_handler	= proc_dointvec_jiffies,
6765 	},
6766 #ifdef CONFIG_IPV6_ROUTE_INFO
6767 	{
6768 		.procname	= "accept_ra_rt_info_min_plen",
6769 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6770 		.maxlen		= sizeof(int),
6771 		.mode		= 0644,
6772 		.proc_handler	= proc_dointvec,
6773 	},
6774 	{
6775 		.procname	= "accept_ra_rt_info_max_plen",
6776 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6777 		.maxlen		= sizeof(int),
6778 		.mode		= 0644,
6779 		.proc_handler	= proc_dointvec,
6780 	},
6781 #endif
6782 #endif
6783 	{
6784 		.procname	= "proxy_ndp",
6785 		.data		= &ipv6_devconf.proxy_ndp,
6786 		.maxlen		= sizeof(int),
6787 		.mode		= 0644,
6788 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6789 	},
6790 	{
6791 		.procname	= "accept_source_route",
6792 		.data		= &ipv6_devconf.accept_source_route,
6793 		.maxlen		= sizeof(int),
6794 		.mode		= 0644,
6795 		.proc_handler	= proc_dointvec,
6796 	},
6797 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6798 	{
6799 		.procname	= "optimistic_dad",
6800 		.data		= &ipv6_devconf.optimistic_dad,
6801 		.maxlen		= sizeof(int),
6802 		.mode		= 0644,
6803 		.proc_handler   = proc_dointvec,
6804 	},
6805 	{
6806 		.procname	= "use_optimistic",
6807 		.data		= &ipv6_devconf.use_optimistic,
6808 		.maxlen		= sizeof(int),
6809 		.mode		= 0644,
6810 		.proc_handler	= proc_dointvec,
6811 	},
6812 #endif
6813 #ifdef CONFIG_IPV6_MROUTE
6814 	{
6815 		.procname	= "mc_forwarding",
6816 		.data		= &ipv6_devconf.mc_forwarding,
6817 		.maxlen		= sizeof(int),
6818 		.mode		= 0444,
6819 		.proc_handler	= proc_dointvec,
6820 	},
6821 #endif
6822 	{
6823 		.procname	= "disable_ipv6",
6824 		.data		= &ipv6_devconf.disable_ipv6,
6825 		.maxlen		= sizeof(int),
6826 		.mode		= 0644,
6827 		.proc_handler	= addrconf_sysctl_disable,
6828 	},
6829 	{
6830 		.procname	= "accept_dad",
6831 		.data		= &ipv6_devconf.accept_dad,
6832 		.maxlen		= sizeof(int),
6833 		.mode		= 0644,
6834 		.proc_handler	= proc_dointvec,
6835 	},
6836 	{
6837 		.procname	= "force_tllao",
6838 		.data		= &ipv6_devconf.force_tllao,
6839 		.maxlen		= sizeof(int),
6840 		.mode		= 0644,
6841 		.proc_handler	= proc_dointvec
6842 	},
6843 	{
6844 		.procname	= "ndisc_notify",
6845 		.data		= &ipv6_devconf.ndisc_notify,
6846 		.maxlen		= sizeof(int),
6847 		.mode		= 0644,
6848 		.proc_handler	= proc_dointvec
6849 	},
6850 	{
6851 		.procname	= "suppress_frag_ndisc",
6852 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6853 		.maxlen		= sizeof(int),
6854 		.mode		= 0644,
6855 		.proc_handler	= proc_dointvec
6856 	},
6857 	{
6858 		.procname	= "accept_ra_from_local",
6859 		.data		= &ipv6_devconf.accept_ra_from_local,
6860 		.maxlen		= sizeof(int),
6861 		.mode		= 0644,
6862 		.proc_handler	= proc_dointvec,
6863 	},
6864 	{
6865 		.procname	= "accept_ra_mtu",
6866 		.data		= &ipv6_devconf.accept_ra_mtu,
6867 		.maxlen		= sizeof(int),
6868 		.mode		= 0644,
6869 		.proc_handler	= proc_dointvec,
6870 	},
6871 	{
6872 		.procname	= "stable_secret",
6873 		.data		= &ipv6_devconf.stable_secret,
6874 		.maxlen		= IPV6_MAX_STRLEN,
6875 		.mode		= 0600,
6876 		.proc_handler	= addrconf_sysctl_stable_secret,
6877 	},
6878 	{
6879 		.procname	= "use_oif_addrs_only",
6880 		.data		= &ipv6_devconf.use_oif_addrs_only,
6881 		.maxlen		= sizeof(int),
6882 		.mode		= 0644,
6883 		.proc_handler	= proc_dointvec,
6884 	},
6885 	{
6886 		.procname	= "ignore_routes_with_linkdown",
6887 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6888 		.maxlen		= sizeof(int),
6889 		.mode		= 0644,
6890 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6891 	},
6892 	{
6893 		.procname	= "drop_unicast_in_l2_multicast",
6894 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6895 		.maxlen		= sizeof(int),
6896 		.mode		= 0644,
6897 		.proc_handler	= proc_dointvec,
6898 	},
6899 	{
6900 		.procname	= "drop_unsolicited_na",
6901 		.data		= &ipv6_devconf.drop_unsolicited_na,
6902 		.maxlen		= sizeof(int),
6903 		.mode		= 0644,
6904 		.proc_handler	= proc_dointvec,
6905 	},
6906 	{
6907 		.procname	= "keep_addr_on_down",
6908 		.data		= &ipv6_devconf.keep_addr_on_down,
6909 		.maxlen		= sizeof(int),
6910 		.mode		= 0644,
6911 		.proc_handler	= proc_dointvec,
6912 
6913 	},
6914 	{
6915 		.procname	= "seg6_enabled",
6916 		.data		= &ipv6_devconf.seg6_enabled,
6917 		.maxlen		= sizeof(int),
6918 		.mode		= 0644,
6919 		.proc_handler	= proc_dointvec,
6920 	},
6921 #ifdef CONFIG_IPV6_SEG6_HMAC
6922 	{
6923 		.procname	= "seg6_require_hmac",
6924 		.data		= &ipv6_devconf.seg6_require_hmac,
6925 		.maxlen		= sizeof(int),
6926 		.mode		= 0644,
6927 		.proc_handler	= proc_dointvec,
6928 	},
6929 #endif
6930 	{
6931 		.procname       = "enhanced_dad",
6932 		.data           = &ipv6_devconf.enhanced_dad,
6933 		.maxlen         = sizeof(int),
6934 		.mode           = 0644,
6935 		.proc_handler   = proc_dointvec,
6936 	},
6937 	{
6938 		.procname	= "addr_gen_mode",
6939 		.data		= &ipv6_devconf.addr_gen_mode,
6940 		.maxlen		= sizeof(int),
6941 		.mode		= 0644,
6942 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6943 	},
6944 	{
6945 		.procname       = "disable_policy",
6946 		.data           = &ipv6_devconf.disable_policy,
6947 		.maxlen         = sizeof(int),
6948 		.mode           = 0644,
6949 		.proc_handler   = addrconf_sysctl_disable_policy,
6950 	},
6951 	{
6952 		.procname	= "ndisc_tclass",
6953 		.data		= &ipv6_devconf.ndisc_tclass,
6954 		.maxlen		= sizeof(int),
6955 		.mode		= 0644,
6956 		.proc_handler	= proc_dointvec_minmax,
6957 		.extra1		= (void *)SYSCTL_ZERO,
6958 		.extra2		= (void *)&two_five_five,
6959 	},
6960 	{
6961 		.procname	= "rpl_seg_enabled",
6962 		.data		= &ipv6_devconf.rpl_seg_enabled,
6963 		.maxlen		= sizeof(int),
6964 		.mode		= 0644,
6965 		.proc_handler	= proc_dointvec,
6966 	},
6967 	{
6968 		.procname	= "ioam6_enabled",
6969 		.data		= &ipv6_devconf.ioam6_enabled,
6970 		.maxlen		= sizeof(u8),
6971 		.mode		= 0644,
6972 		.proc_handler	= proc_dou8vec_minmax,
6973 		.extra1		= (void *)SYSCTL_ZERO,
6974 		.extra2		= (void *)SYSCTL_ONE,
6975 	},
6976 	{
6977 		.procname	= "ioam6_id",
6978 		.data		= &ipv6_devconf.ioam6_id,
6979 		.maxlen		= sizeof(u32),
6980 		.mode		= 0644,
6981 		.proc_handler	= proc_douintvec_minmax,
6982 		.extra1		= (void *)SYSCTL_ZERO,
6983 		.extra2		= (void *)&ioam6_if_id_max,
6984 	},
6985 	{
6986 		.procname	= "ioam6_id_wide",
6987 		.data		= &ipv6_devconf.ioam6_id_wide,
6988 		.maxlen		= sizeof(u32),
6989 		.mode		= 0644,
6990 		.proc_handler	= proc_douintvec,
6991 	},
6992 	{
6993 		.procname	= "ndisc_evict_nocarrier",
6994 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
6995 		.maxlen		= sizeof(u8),
6996 		.mode		= 0644,
6997 		.proc_handler	= proc_dou8vec_minmax,
6998 		.extra1		= (void *)SYSCTL_ZERO,
6999 		.extra2		= (void *)SYSCTL_ONE,
7000 	},
7001 	{
7002 		/* sentinel */
7003 	}
7004 };
7005 
7006 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7007 		struct inet6_dev *idev, struct ipv6_devconf *p)
7008 {
7009 	int i, ifindex;
7010 	struct ctl_table *table;
7011 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7012 
7013 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
7014 	if (!table)
7015 		goto out;
7016 
7017 	for (i = 0; table[i].data; i++) {
7018 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7019 		/* If one of these is already set, then it is not safe to
7020 		 * overwrite either of them: this makes proc_dointvec_minmax
7021 		 * usable.
7022 		 */
7023 		if (!table[i].extra1 && !table[i].extra2) {
7024 			table[i].extra1 = idev; /* embedded; no ref */
7025 			table[i].extra2 = net;
7026 		}
7027 	}
7028 
7029 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7030 
7031 	p->sysctl_header = register_net_sysctl(net, path, table);
7032 	if (!p->sysctl_header)
7033 		goto free;
7034 
7035 	if (!strcmp(dev_name, "all"))
7036 		ifindex = NETCONFA_IFINDEX_ALL;
7037 	else if (!strcmp(dev_name, "default"))
7038 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7039 	else
7040 		ifindex = idev->dev->ifindex;
7041 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7042 				     ifindex, p);
7043 	return 0;
7044 
7045 free:
7046 	kfree(table);
7047 out:
7048 	return -ENOBUFS;
7049 }
7050 
7051 static void __addrconf_sysctl_unregister(struct net *net,
7052 					 struct ipv6_devconf *p, int ifindex)
7053 {
7054 	struct ctl_table *table;
7055 
7056 	if (!p->sysctl_header)
7057 		return;
7058 
7059 	table = p->sysctl_header->ctl_table_arg;
7060 	unregister_net_sysctl_table(p->sysctl_header);
7061 	p->sysctl_header = NULL;
7062 	kfree(table);
7063 
7064 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7065 }
7066 
7067 static int addrconf_sysctl_register(struct inet6_dev *idev)
7068 {
7069 	int err;
7070 
7071 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7072 		return -EINVAL;
7073 
7074 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7075 				    &ndisc_ifinfo_sysctl_change);
7076 	if (err)
7077 		return err;
7078 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7079 					 idev, &idev->cnf);
7080 	if (err)
7081 		neigh_sysctl_unregister(idev->nd_parms);
7082 
7083 	return err;
7084 }
7085 
7086 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7087 {
7088 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7089 				     idev->dev->ifindex);
7090 	neigh_sysctl_unregister(idev->nd_parms);
7091 }
7092 
7093 
7094 #endif
7095 
7096 static int __net_init addrconf_init_net(struct net *net)
7097 {
7098 	int err = -ENOMEM;
7099 	struct ipv6_devconf *all, *dflt;
7100 
7101 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7102 	if (!all)
7103 		goto err_alloc_all;
7104 
7105 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7106 	if (!dflt)
7107 		goto err_alloc_dflt;
7108 
7109 	if (IS_ENABLED(CONFIG_SYSCTL) &&
7110 	    !net_eq(net, &init_net)) {
7111 		switch (sysctl_devconf_inherit_init_net) {
7112 		case 1:  /* copy from init_net */
7113 			memcpy(all, init_net.ipv6.devconf_all,
7114 			       sizeof(ipv6_devconf));
7115 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7116 			       sizeof(ipv6_devconf_dflt));
7117 			break;
7118 		case 3: /* copy from the current netns */
7119 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7120 			       sizeof(ipv6_devconf));
7121 			memcpy(dflt,
7122 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7123 			       sizeof(ipv6_devconf_dflt));
7124 			break;
7125 		case 0:
7126 		case 2:
7127 			/* use compiled values */
7128 			break;
7129 		}
7130 	}
7131 
7132 	/* these will be inherited by all namespaces */
7133 	dflt->autoconf = ipv6_defaults.autoconf;
7134 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7135 
7136 	dflt->stable_secret.initialized = false;
7137 	all->stable_secret.initialized = false;
7138 
7139 	net->ipv6.devconf_all = all;
7140 	net->ipv6.devconf_dflt = dflt;
7141 
7142 #ifdef CONFIG_SYSCTL
7143 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7144 	if (err < 0)
7145 		goto err_reg_all;
7146 
7147 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7148 	if (err < 0)
7149 		goto err_reg_dflt;
7150 #endif
7151 	return 0;
7152 
7153 #ifdef CONFIG_SYSCTL
7154 err_reg_dflt:
7155 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7156 err_reg_all:
7157 	kfree(dflt);
7158 #endif
7159 err_alloc_dflt:
7160 	kfree(all);
7161 err_alloc_all:
7162 	return err;
7163 }
7164 
7165 static void __net_exit addrconf_exit_net(struct net *net)
7166 {
7167 #ifdef CONFIG_SYSCTL
7168 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7169 				     NETCONFA_IFINDEX_DEFAULT);
7170 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7171 				     NETCONFA_IFINDEX_ALL);
7172 #endif
7173 	kfree(net->ipv6.devconf_dflt);
7174 	kfree(net->ipv6.devconf_all);
7175 }
7176 
7177 static struct pernet_operations addrconf_ops = {
7178 	.init = addrconf_init_net,
7179 	.exit = addrconf_exit_net,
7180 };
7181 
7182 static struct rtnl_af_ops inet6_ops __read_mostly = {
7183 	.family		  = AF_INET6,
7184 	.fill_link_af	  = inet6_fill_link_af,
7185 	.get_link_af_size = inet6_get_link_af_size,
7186 	.validate_link_af = inet6_validate_link_af,
7187 	.set_link_af	  = inet6_set_link_af,
7188 };
7189 
7190 /*
7191  *	Init / cleanup code
7192  */
7193 
7194 int __init addrconf_init(void)
7195 {
7196 	struct inet6_dev *idev;
7197 	int i, err;
7198 
7199 	err = ipv6_addr_label_init();
7200 	if (err < 0) {
7201 		pr_crit("%s: cannot initialize default policy table: %d\n",
7202 			__func__, err);
7203 		goto out;
7204 	}
7205 
7206 	err = register_pernet_subsys(&addrconf_ops);
7207 	if (err < 0)
7208 		goto out_addrlabel;
7209 
7210 	addrconf_wq = create_workqueue("ipv6_addrconf");
7211 	if (!addrconf_wq) {
7212 		err = -ENOMEM;
7213 		goto out_nowq;
7214 	}
7215 
7216 	/* The addrconf netdev notifier requires that loopback_dev
7217 	 * has it's ipv6 private information allocated and setup
7218 	 * before it can bring up and give link-local addresses
7219 	 * to other devices which are up.
7220 	 *
7221 	 * Unfortunately, loopback_dev is not necessarily the first
7222 	 * entry in the global dev_base list of net devices.  In fact,
7223 	 * it is likely to be the very last entry on that list.
7224 	 * So this causes the notifier registry below to try and
7225 	 * give link-local addresses to all devices besides loopback_dev
7226 	 * first, then loopback_dev, which cases all the non-loopback_dev
7227 	 * devices to fail to get a link-local address.
7228 	 *
7229 	 * So, as a temporary fix, allocate the ipv6 structure for
7230 	 * loopback_dev first by hand.
7231 	 * Longer term, all of the dependencies ipv6 has upon the loopback
7232 	 * device and it being up should be removed.
7233 	 */
7234 	rtnl_lock();
7235 	idev = ipv6_add_dev(init_net.loopback_dev);
7236 	rtnl_unlock();
7237 	if (IS_ERR(idev)) {
7238 		err = PTR_ERR(idev);
7239 		goto errlo;
7240 	}
7241 
7242 	ip6_route_init_special_entries();
7243 
7244 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7245 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7246 
7247 	register_netdevice_notifier(&ipv6_dev_notf);
7248 
7249 	addrconf_verify();
7250 
7251 	rtnl_af_register(&inet6_ops);
7252 
7253 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7254 				   NULL, inet6_dump_ifinfo, 0);
7255 	if (err < 0)
7256 		goto errout;
7257 
7258 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7259 				   inet6_rtm_newaddr, NULL, 0);
7260 	if (err < 0)
7261 		goto errout;
7262 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7263 				   inet6_rtm_deladdr, NULL, 0);
7264 	if (err < 0)
7265 		goto errout;
7266 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7267 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7268 				   RTNL_FLAG_DOIT_UNLOCKED);
7269 	if (err < 0)
7270 		goto errout;
7271 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7272 				   NULL, inet6_dump_ifmcaddr, 0);
7273 	if (err < 0)
7274 		goto errout;
7275 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7276 				   NULL, inet6_dump_ifacaddr, 0);
7277 	if (err < 0)
7278 		goto errout;
7279 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7280 				   inet6_netconf_get_devconf,
7281 				   inet6_netconf_dump_devconf,
7282 				   RTNL_FLAG_DOIT_UNLOCKED);
7283 	if (err < 0)
7284 		goto errout;
7285 	err = ipv6_addr_label_rtnl_register();
7286 	if (err < 0)
7287 		goto errout;
7288 
7289 	return 0;
7290 errout:
7291 	rtnl_unregister_all(PF_INET6);
7292 	rtnl_af_unregister(&inet6_ops);
7293 	unregister_netdevice_notifier(&ipv6_dev_notf);
7294 errlo:
7295 	destroy_workqueue(addrconf_wq);
7296 out_nowq:
7297 	unregister_pernet_subsys(&addrconf_ops);
7298 out_addrlabel:
7299 	ipv6_addr_label_cleanup();
7300 out:
7301 	return err;
7302 }
7303 
7304 void addrconf_cleanup(void)
7305 {
7306 	struct net_device *dev;
7307 	int i;
7308 
7309 	unregister_netdevice_notifier(&ipv6_dev_notf);
7310 	unregister_pernet_subsys(&addrconf_ops);
7311 	ipv6_addr_label_cleanup();
7312 
7313 	rtnl_af_unregister(&inet6_ops);
7314 
7315 	rtnl_lock();
7316 
7317 	/* clean dev list */
7318 	for_each_netdev(&init_net, dev) {
7319 		if (__in6_dev_get(dev) == NULL)
7320 			continue;
7321 		addrconf_ifdown(dev, true);
7322 	}
7323 	addrconf_ifdown(init_net.loopback_dev, true);
7324 
7325 	/*
7326 	 *	Check hash table.
7327 	 */
7328 	spin_lock_bh(&addrconf_hash_lock);
7329 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7330 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7331 	spin_unlock_bh(&addrconf_hash_lock);
7332 	cancel_delayed_work(&addr_chk_work);
7333 	rtnl_unlock();
7334 
7335 	destroy_workqueue(addrconf_wq);
7336 }
7337