• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  *	IPv6 Address [auto]configuration
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14 
15 /*
16  *	Changes:
17  *
18  *	Janos Farkas			:	delete timer on ifdown
19  *	<chexum@bankinf.banki.hu>
20  *	Andi Kleen			:	kill double kfree on module
21  *						unload.
22  *	Maciej W. Rozycki		:	FDDI support
23  *	sekiya@USAGI			:	Don't send too many RS
24  *						packets.
25  *	yoshfuji@USAGI			:       Fixed interval between DAD
26  *						packets.
27  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
28  *						address validation timer.
29  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
30  *						support.
31  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
32  *						address on a same interface.
33  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
34  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
35  *						seq_file.
36  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
37  *						selection; consider scope,
38  *						status etc.
39  */
40 
41 #define pr_fmt(fmt) "IPv6: " fmt
42 
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
61 #ifdef CONFIG_SYSCTL
62 #include <linux/sysctl.h>
63 #endif
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
69 
70 #include <net/net_namespace.h>
71 #include <net/sock.h>
72 #include <net/snmp.h>
73 
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
76 #include <net/ipv6.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
81 #include <net/tcp.h>
82 #include <net/ip.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92 
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96 
97 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
98 
99 #define IPV6_MAX_STRLEN \
100 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
101 
cstamp_delta(unsigned long cstamp)102 static inline u32 cstamp_delta(unsigned long cstamp)
103 {
104 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
105 }
106 
rfc3315_s14_backoff_init(s32 irt)107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
108 {
109 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
110 	u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
111 	do_div(tmp, 1000000);
112 	return (s32)tmp;
113 }
114 
rfc3315_s14_backoff_update(s32 rt,s32 mrt)115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
116 {
117 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
118 	u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
119 	do_div(tmp, 1000000);
120 	if ((s32)tmp > mrt) {
121 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
122 		tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
123 		do_div(tmp, 1000000);
124 	}
125 	return (s32)tmp;
126 }
127 
128 #ifdef CONFIG_SYSCTL
129 static int addrconf_sysctl_register(struct inet6_dev *idev);
130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
131 #else
addrconf_sysctl_register(struct inet6_dev * idev)132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
133 {
134 	return 0;
135 }
136 
addrconf_sysctl_unregister(struct inet6_dev * idev)137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
138 {
139 }
140 #endif
141 
142 static void ipv6_regen_rndid(struct inet6_dev *idev);
143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
144 
145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
146 static int ipv6_count_addresses(const struct inet6_dev *idev);
147 static int ipv6_generate_stable_address(struct in6_addr *addr,
148 					u8 dad_count,
149 					const struct inet6_dev *idev);
150 
151 #define IN6_ADDR_HSIZE_SHIFT	8
152 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
153 /*
154  *	Configured unicast address hash table
155  */
156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
157 static DEFINE_SPINLOCK(addrconf_hash_lock);
158 
159 static void addrconf_verify(void);
160 static void addrconf_verify_rtnl(void);
161 static void addrconf_verify_work(struct work_struct *);
162 
163 static struct workqueue_struct *addrconf_wq;
164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
165 
166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
168 
169 static void addrconf_type_change(struct net_device *dev,
170 				 unsigned long event);
171 static int addrconf_ifdown(struct net_device *dev, int how);
172 
173 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
174 						  int plen,
175 						  const struct net_device *dev,
176 						  u32 flags, u32 noflags);
177 
178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
179 static void addrconf_dad_work(struct work_struct *w);
180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
181 				   bool send_na);
182 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
183 static void addrconf_rs_timer(struct timer_list *t);
184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
186 
187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
188 				struct prefix_info *pinfo);
189 
190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
191 	.forwarding		= 0,
192 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
193 	.mtu6			= IPV6_MIN_MTU,
194 	.accept_ra		= 1,
195 	.accept_redirects	= 1,
196 	.autoconf		= 1,
197 	.force_mld_version	= 0,
198 	.mldv1_unsolicited_report_interval = 10 * HZ,
199 	.mldv2_unsolicited_report_interval = HZ,
200 	.dad_transmits		= 1,
201 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
202 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
203 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
204 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
205 	.use_tempaddr		= 0,
206 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
207 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
208 	.regen_max_retry	= REGEN_MAX_RETRY,
209 	.max_desync_factor	= MAX_DESYNC_FACTOR,
210 	.max_addresses		= IPV6_MAX_ADDRESSES,
211 	.accept_ra_defrtr	= 1,
212 	.accept_ra_from_local	= 0,
213 	.accept_ra_min_hop_limit= 1,
214 	.accept_ra_pinfo	= 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216 	.accept_ra_rtr_pref	= 1,
217 	.rtr_probe_interval	= 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219 	.accept_ra_rt_info_min_plen = 0,
220 	.accept_ra_rt_info_max_plen = 0,
221 #endif
222 #endif
223 	.proxy_ndp		= 0,
224 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
225 	.disable_ipv6		= 0,
226 	.accept_dad		= 0,
227 	.suppress_frag_ndisc	= 1,
228 	.accept_ra_mtu		= 1,
229 	.stable_secret		= {
230 		.initialized = false,
231 	},
232 	.use_oif_addrs_only	= 0,
233 	.ignore_routes_with_linkdown = 0,
234 	.keep_addr_on_down	= 0,
235 	.seg6_enabled		= 0,
236 #ifdef CONFIG_IPV6_SEG6_HMAC
237 	.seg6_require_hmac	= 0,
238 #endif
239 	.enhanced_dad           = 1,
240 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
241 	.disable_policy		= 0,
242 };
243 
244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
245 	.forwarding		= 0,
246 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
247 	.mtu6			= IPV6_MIN_MTU,
248 	.accept_ra		= 1,
249 	.accept_redirects	= 1,
250 	.autoconf		= 1,
251 	.force_mld_version	= 0,
252 	.mldv1_unsolicited_report_interval = 10 * HZ,
253 	.mldv2_unsolicited_report_interval = HZ,
254 	.dad_transmits		= 1,
255 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
256 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
257 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
258 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
259 	.use_tempaddr		= 0,
260 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
261 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
262 	.regen_max_retry	= REGEN_MAX_RETRY,
263 	.max_desync_factor	= MAX_DESYNC_FACTOR,
264 	.max_addresses		= IPV6_MAX_ADDRESSES,
265 	.accept_ra_defrtr	= 1,
266 	.accept_ra_from_local	= 0,
267 	.accept_ra_min_hop_limit= 1,
268 	.accept_ra_pinfo	= 1,
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270 	.accept_ra_rtr_pref	= 1,
271 	.rtr_probe_interval	= 60 * HZ,
272 #ifdef CONFIG_IPV6_ROUTE_INFO
273 	.accept_ra_rt_info_min_plen = 0,
274 	.accept_ra_rt_info_max_plen = 0,
275 #endif
276 #endif
277 	.proxy_ndp		= 0,
278 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
279 	.disable_ipv6		= 0,
280 	.accept_dad		= 1,
281 	.suppress_frag_ndisc	= 1,
282 	.accept_ra_mtu		= 1,
283 	.stable_secret		= {
284 		.initialized = false,
285 	},
286 	.use_oif_addrs_only	= 0,
287 	.ignore_routes_with_linkdown = 0,
288 	.keep_addr_on_down	= 0,
289 	.seg6_enabled		= 0,
290 #ifdef CONFIG_IPV6_SEG6_HMAC
291 	.seg6_require_hmac	= 0,
292 #endif
293 	.enhanced_dad           = 1,
294 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
295 	.disable_policy		= 0,
296 };
297 
298 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device * dev)299 static inline bool addrconf_link_ready(const struct net_device *dev)
300 {
301 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
302 }
303 
addrconf_del_rs_timer(struct inet6_dev * idev)304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
305 {
306 	if (del_timer(&idev->rs_timer))
307 		__in6_dev_put(idev);
308 }
309 
addrconf_del_dad_work(struct inet6_ifaddr * ifp)310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
311 {
312 	if (cancel_delayed_work(&ifp->dad_work))
313 		__in6_ifa_put(ifp);
314 }
315 
addrconf_mod_rs_timer(struct inet6_dev * idev,unsigned long when)316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
317 				  unsigned long when)
318 {
319 	if (!timer_pending(&idev->rs_timer))
320 		in6_dev_hold(idev);
321 	mod_timer(&idev->rs_timer, jiffies + when);
322 }
323 
addrconf_mod_dad_work(struct inet6_ifaddr * ifp,unsigned long delay)324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
325 				   unsigned long delay)
326 {
327 	in6_ifa_hold(ifp);
328 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
329 		in6_ifa_put(ifp);
330 }
331 
snmp6_alloc_dev(struct inet6_dev * idev)332 static int snmp6_alloc_dev(struct inet6_dev *idev)
333 {
334 	int i;
335 
336 	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
337 	if (!idev->stats.ipv6)
338 		goto err_ip;
339 
340 	for_each_possible_cpu(i) {
341 		struct ipstats_mib *addrconf_stats;
342 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
343 		u64_stats_init(&addrconf_stats->syncp);
344 	}
345 
346 
347 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
348 					GFP_KERNEL);
349 	if (!idev->stats.icmpv6dev)
350 		goto err_icmp;
351 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
352 					   GFP_KERNEL);
353 	if (!idev->stats.icmpv6msgdev)
354 		goto err_icmpmsg;
355 
356 	return 0;
357 
358 err_icmpmsg:
359 	kfree(idev->stats.icmpv6dev);
360 err_icmp:
361 	free_percpu(idev->stats.ipv6);
362 err_ip:
363 	return -ENOMEM;
364 }
365 
ipv6_add_dev(struct net_device * dev)366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
367 {
368 	struct inet6_dev *ndev;
369 	int err = -ENOMEM;
370 
371 	ASSERT_RTNL();
372 
373 	if (dev->mtu < IPV6_MIN_MTU)
374 		return ERR_PTR(-EINVAL);
375 
376 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
377 	if (!ndev)
378 		return ERR_PTR(err);
379 
380 	rwlock_init(&ndev->lock);
381 	ndev->dev = dev;
382 	INIT_LIST_HEAD(&ndev->addr_list);
383 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
384 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
385 
386 	if (ndev->cnf.stable_secret.initialized)
387 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
388 
389 	ndev->cnf.mtu6 = dev->mtu;
390 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
391 	if (!ndev->nd_parms) {
392 		kfree(ndev);
393 		return ERR_PTR(err);
394 	}
395 	if (ndev->cnf.forwarding)
396 		dev_disable_lro(dev);
397 	/* We refer to the device */
398 	dev_hold(dev);
399 
400 	if (snmp6_alloc_dev(ndev) < 0) {
401 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
402 			   __func__);
403 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
404 		dev_put(dev);
405 		kfree(ndev);
406 		return ERR_PTR(err);
407 	}
408 
409 	if (snmp6_register_dev(ndev) < 0) {
410 		netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
411 			   __func__, dev->name);
412 		goto err_release;
413 	}
414 
415 	/* One reference from device. */
416 	refcount_set(&ndev->refcnt, 1);
417 
418 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
419 		ndev->cnf.accept_dad = -1;
420 
421 #if IS_ENABLED(CONFIG_IPV6_SIT)
422 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
423 		pr_info("%s: Disabled Multicast RS\n", dev->name);
424 		ndev->cnf.rtr_solicits = 0;
425 	}
426 #endif
427 
428 	INIT_LIST_HEAD(&ndev->tempaddr_list);
429 	ndev->desync_factor = U32_MAX;
430 	if ((dev->flags&IFF_LOOPBACK) ||
431 	    dev->type == ARPHRD_TUNNEL ||
432 	    dev->type == ARPHRD_TUNNEL6 ||
433 	    dev->type == ARPHRD_SIT ||
434 	    dev->type == ARPHRD_NONE) {
435 		ndev->cnf.use_tempaddr = -1;
436 	} else
437 		ipv6_regen_rndid(ndev);
438 
439 	ndev->token = in6addr_any;
440 
441 	if (netif_running(dev) && addrconf_link_ready(dev))
442 		ndev->if_flags |= IF_READY;
443 
444 	ipv6_mc_init_dev(ndev);
445 	ndev->tstamp = jiffies;
446 	err = addrconf_sysctl_register(ndev);
447 	if (err) {
448 		ipv6_mc_destroy_dev(ndev);
449 		snmp6_unregister_dev(ndev);
450 		goto err_release;
451 	}
452 	/* protected by rtnl_lock */
453 	rcu_assign_pointer(dev->ip6_ptr, ndev);
454 
455 	/* Join interface-local all-node multicast group */
456 	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
457 
458 	/* Join all-node multicast group */
459 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
460 
461 	/* Join all-router multicast group if forwarding is set */
462 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
463 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
464 
465 	return ndev;
466 
467 err_release:
468 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
469 	ndev->dead = 1;
470 	in6_dev_finish_destroy(ndev);
471 	return ERR_PTR(err);
472 }
473 
ipv6_find_idev(struct net_device * dev)474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
475 {
476 	struct inet6_dev *idev;
477 
478 	ASSERT_RTNL();
479 
480 	idev = __in6_dev_get(dev);
481 	if (!idev) {
482 		idev = ipv6_add_dev(dev);
483 		if (IS_ERR(idev))
484 			return NULL;
485 	}
486 
487 	if (dev->flags&IFF_UP)
488 		ipv6_mc_up(idev);
489 	return idev;
490 }
491 
inet6_netconf_msgsize_devconf(int type)492 static int inet6_netconf_msgsize_devconf(int type)
493 {
494 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
495 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
496 	bool all = false;
497 
498 	if (type == NETCONFA_ALL)
499 		all = true;
500 
501 	if (all || type == NETCONFA_FORWARDING)
502 		size += nla_total_size(4);
503 #ifdef CONFIG_IPV6_MROUTE
504 	if (all || type == NETCONFA_MC_FORWARDING)
505 		size += nla_total_size(4);
506 #endif
507 	if (all || type == NETCONFA_PROXY_NEIGH)
508 		size += nla_total_size(4);
509 
510 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
511 		size += nla_total_size(4);
512 
513 	return size;
514 }
515 
inet6_netconf_fill_devconf(struct sk_buff * skb,int ifindex,struct ipv6_devconf * devconf,u32 portid,u32 seq,int event,unsigned int flags,int type)516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
517 				      struct ipv6_devconf *devconf, u32 portid,
518 				      u32 seq, int event, unsigned int flags,
519 				      int type)
520 {
521 	struct nlmsghdr  *nlh;
522 	struct netconfmsg *ncm;
523 	bool all = false;
524 
525 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
526 			flags);
527 	if (!nlh)
528 		return -EMSGSIZE;
529 
530 	if (type == NETCONFA_ALL)
531 		all = true;
532 
533 	ncm = nlmsg_data(nlh);
534 	ncm->ncm_family = AF_INET6;
535 
536 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
537 		goto nla_put_failure;
538 
539 	if (!devconf)
540 		goto out;
541 
542 	if ((all || type == NETCONFA_FORWARDING) &&
543 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
544 		goto nla_put_failure;
545 #ifdef CONFIG_IPV6_MROUTE
546 	if ((all || type == NETCONFA_MC_FORWARDING) &&
547 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
548 			devconf->mc_forwarding) < 0)
549 		goto nla_put_failure;
550 #endif
551 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
552 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
553 		goto nla_put_failure;
554 
555 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
556 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
557 			devconf->ignore_routes_with_linkdown) < 0)
558 		goto nla_put_failure;
559 
560 out:
561 	nlmsg_end(skb, nlh);
562 	return 0;
563 
564 nla_put_failure:
565 	nlmsg_cancel(skb, nlh);
566 	return -EMSGSIZE;
567 }
568 
inet6_netconf_notify_devconf(struct net * net,int event,int type,int ifindex,struct ipv6_devconf * devconf)569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
570 				  int ifindex, struct ipv6_devconf *devconf)
571 {
572 	struct sk_buff *skb;
573 	int err = -ENOBUFS;
574 
575 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
576 	if (!skb)
577 		goto errout;
578 
579 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
580 					 event, 0, type);
581 	if (err < 0) {
582 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
583 		WARN_ON(err == -EMSGSIZE);
584 		kfree_skb(skb);
585 		goto errout;
586 	}
587 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
588 	return;
589 errout:
590 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
591 }
592 
593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
594 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
595 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
596 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
597 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
598 };
599 
inet6_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)600 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
601 				     struct nlmsghdr *nlh,
602 				     struct netlink_ext_ack *extack)
603 {
604 	struct net *net = sock_net(in_skb->sk);
605 	struct nlattr *tb[NETCONFA_MAX+1];
606 	struct inet6_dev *in6_dev = NULL;
607 	struct net_device *dev = NULL;
608 	struct netconfmsg *ncm;
609 	struct sk_buff *skb;
610 	struct ipv6_devconf *devconf;
611 	int ifindex;
612 	int err;
613 
614 	err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
615 			  devconf_ipv6_policy, extack);
616 	if (err < 0)
617 		return err;
618 
619 	if (!tb[NETCONFA_IFINDEX])
620 		return -EINVAL;
621 
622 	err = -EINVAL;
623 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
624 	switch (ifindex) {
625 	case NETCONFA_IFINDEX_ALL:
626 		devconf = net->ipv6.devconf_all;
627 		break;
628 	case NETCONFA_IFINDEX_DEFAULT:
629 		devconf = net->ipv6.devconf_dflt;
630 		break;
631 	default:
632 		dev = dev_get_by_index(net, ifindex);
633 		if (!dev)
634 			return -EINVAL;
635 		in6_dev = in6_dev_get(dev);
636 		if (!in6_dev)
637 			goto errout;
638 		devconf = &in6_dev->cnf;
639 		break;
640 	}
641 
642 	err = -ENOBUFS;
643 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
644 	if (!skb)
645 		goto errout;
646 
647 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
648 					 NETLINK_CB(in_skb).portid,
649 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
650 					 NETCONFA_ALL);
651 	if (err < 0) {
652 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
653 		WARN_ON(err == -EMSGSIZE);
654 		kfree_skb(skb);
655 		goto errout;
656 	}
657 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
658 errout:
659 	if (in6_dev)
660 		in6_dev_put(in6_dev);
661 	if (dev)
662 		dev_put(dev);
663 	return err;
664 }
665 
inet6_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
667 				      struct netlink_callback *cb)
668 {
669 	struct net *net = sock_net(skb->sk);
670 	int h, s_h;
671 	int idx, s_idx;
672 	struct net_device *dev;
673 	struct inet6_dev *idev;
674 	struct hlist_head *head;
675 
676 	s_h = cb->args[0];
677 	s_idx = idx = cb->args[1];
678 
679 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
680 		idx = 0;
681 		head = &net->dev_index_head[h];
682 		rcu_read_lock();
683 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
684 			  net->dev_base_seq;
685 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
686 			if (idx < s_idx)
687 				goto cont;
688 			idev = __in6_dev_get(dev);
689 			if (!idev)
690 				goto cont;
691 
692 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
693 						       &idev->cnf,
694 						       NETLINK_CB(cb->skb).portid,
695 						       cb->nlh->nlmsg_seq,
696 						       RTM_NEWNETCONF,
697 						       NLM_F_MULTI,
698 						       NETCONFA_ALL) < 0) {
699 				rcu_read_unlock();
700 				goto done;
701 			}
702 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
703 cont:
704 			idx++;
705 		}
706 		rcu_read_unlock();
707 	}
708 	if (h == NETDEV_HASHENTRIES) {
709 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
710 					       net->ipv6.devconf_all,
711 					       NETLINK_CB(cb->skb).portid,
712 					       cb->nlh->nlmsg_seq,
713 					       RTM_NEWNETCONF, NLM_F_MULTI,
714 					       NETCONFA_ALL) < 0)
715 			goto done;
716 		else
717 			h++;
718 	}
719 	if (h == NETDEV_HASHENTRIES + 1) {
720 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
721 					       net->ipv6.devconf_dflt,
722 					       NETLINK_CB(cb->skb).portid,
723 					       cb->nlh->nlmsg_seq,
724 					       RTM_NEWNETCONF, NLM_F_MULTI,
725 					       NETCONFA_ALL) < 0)
726 			goto done;
727 		else
728 			h++;
729 	}
730 done:
731 	cb->args[0] = h;
732 	cb->args[1] = idx;
733 
734 	return skb->len;
735 }
736 
737 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev * idev)738 static void dev_forward_change(struct inet6_dev *idev)
739 {
740 	struct net_device *dev;
741 	struct inet6_ifaddr *ifa;
742 
743 	if (!idev)
744 		return;
745 	dev = idev->dev;
746 	if (idev->cnf.forwarding)
747 		dev_disable_lro(dev);
748 	if (dev->flags & IFF_MULTICAST) {
749 		if (idev->cnf.forwarding) {
750 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
751 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
752 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
753 		} else {
754 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
755 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
756 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
757 		}
758 	}
759 
760 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
761 		if (ifa->flags&IFA_F_TENTATIVE)
762 			continue;
763 		if (idev->cnf.forwarding)
764 			addrconf_join_anycast(ifa);
765 		else
766 			addrconf_leave_anycast(ifa);
767 	}
768 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
769 				     NETCONFA_FORWARDING,
770 				     dev->ifindex, &idev->cnf);
771 }
772 
773 
addrconf_forward_change(struct net * net,__s32 newf)774 static void addrconf_forward_change(struct net *net, __s32 newf)
775 {
776 	struct net_device *dev;
777 	struct inet6_dev *idev;
778 
779 	for_each_netdev(net, dev) {
780 		idev = __in6_dev_get(dev);
781 		if (idev) {
782 			int changed = (!idev->cnf.forwarding) ^ (!newf);
783 			idev->cnf.forwarding = newf;
784 			if (changed)
785 				dev_forward_change(idev);
786 		}
787 	}
788 }
789 
addrconf_fixup_forwarding(struct ctl_table * table,int * p,int newf)790 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
791 {
792 	struct net *net;
793 	int old;
794 
795 	if (!rtnl_trylock())
796 		return restart_syscall();
797 
798 	net = (struct net *)table->extra2;
799 	old = *p;
800 	*p = newf;
801 
802 	if (p == &net->ipv6.devconf_dflt->forwarding) {
803 		if ((!newf) ^ (!old))
804 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
805 						     NETCONFA_FORWARDING,
806 						     NETCONFA_IFINDEX_DEFAULT,
807 						     net->ipv6.devconf_dflt);
808 		rtnl_unlock();
809 		return 0;
810 	}
811 
812 	if (p == &net->ipv6.devconf_all->forwarding) {
813 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
814 
815 		net->ipv6.devconf_dflt->forwarding = newf;
816 		if ((!newf) ^ (!old_dflt))
817 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
818 						     NETCONFA_FORWARDING,
819 						     NETCONFA_IFINDEX_DEFAULT,
820 						     net->ipv6.devconf_dflt);
821 
822 		addrconf_forward_change(net, newf);
823 		if ((!newf) ^ (!old))
824 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
825 						     NETCONFA_FORWARDING,
826 						     NETCONFA_IFINDEX_ALL,
827 						     net->ipv6.devconf_all);
828 	} else if ((!newf) ^ (!old))
829 		dev_forward_change((struct inet6_dev *)table->extra1);
830 	rtnl_unlock();
831 
832 	if (newf)
833 		rt6_purge_dflt_routers(net);
834 	return 1;
835 }
836 
addrconf_linkdown_change(struct net * net,__s32 newf)837 static void addrconf_linkdown_change(struct net *net, __s32 newf)
838 {
839 	struct net_device *dev;
840 	struct inet6_dev *idev;
841 
842 	for_each_netdev(net, dev) {
843 		idev = __in6_dev_get(dev);
844 		if (idev) {
845 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
846 
847 			idev->cnf.ignore_routes_with_linkdown = newf;
848 			if (changed)
849 				inet6_netconf_notify_devconf(dev_net(dev),
850 							     RTM_NEWNETCONF,
851 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
852 							     dev->ifindex,
853 							     &idev->cnf);
854 		}
855 	}
856 }
857 
addrconf_fixup_linkdown(struct ctl_table * table,int * p,int newf)858 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
859 {
860 	struct net *net;
861 	int old;
862 
863 	if (!rtnl_trylock())
864 		return restart_syscall();
865 
866 	net = (struct net *)table->extra2;
867 	old = *p;
868 	*p = newf;
869 
870 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
871 		if ((!newf) ^ (!old))
872 			inet6_netconf_notify_devconf(net,
873 						     RTM_NEWNETCONF,
874 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
875 						     NETCONFA_IFINDEX_DEFAULT,
876 						     net->ipv6.devconf_dflt);
877 		rtnl_unlock();
878 		return 0;
879 	}
880 
881 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
882 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
883 		addrconf_linkdown_change(net, newf);
884 		if ((!newf) ^ (!old))
885 			inet6_netconf_notify_devconf(net,
886 						     RTM_NEWNETCONF,
887 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
888 						     NETCONFA_IFINDEX_ALL,
889 						     net->ipv6.devconf_all);
890 	}
891 	rtnl_unlock();
892 
893 	return 1;
894 }
895 
896 #endif
897 
898 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)899 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
900 {
901 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
902 
903 #ifdef NET_REFCNT_DEBUG
904 	pr_debug("%s\n", __func__);
905 #endif
906 
907 	in6_dev_put(ifp->idev);
908 
909 	if (cancel_delayed_work(&ifp->dad_work))
910 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
911 			  ifp);
912 
913 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
914 		pr_warn("Freeing alive inet6 address %p\n", ifp);
915 		return;
916 	}
917 
918 	kfree_rcu(ifp, rcu);
919 }
920 
921 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)922 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
923 {
924 	struct list_head *p;
925 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
926 
927 	/*
928 	 * Each device address list is sorted in order of scope -
929 	 * global before linklocal.
930 	 */
931 	list_for_each(p, &idev->addr_list) {
932 		struct inet6_ifaddr *ifa
933 			= list_entry(p, struct inet6_ifaddr, if_list);
934 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
935 			break;
936 	}
937 
938 	list_add_tail_rcu(&ifp->if_list, p);
939 }
940 
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)941 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
942 {
943 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
944 
945 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
946 }
947 
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)948 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
949 			       struct net_device *dev, unsigned int hash)
950 {
951 	struct inet6_ifaddr *ifp;
952 
953 	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
954 		if (!net_eq(dev_net(ifp->idev->dev), net))
955 			continue;
956 		if (ipv6_addr_equal(&ifp->addr, addr)) {
957 			if (!dev || ifp->idev->dev == dev)
958 				return true;
959 		}
960 	}
961 	return false;
962 }
963 
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)964 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
965 {
966 	unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
967 	int err = 0;
968 
969 	spin_lock(&addrconf_hash_lock);
970 
971 	/* Ignore adding duplicate addresses on an interface */
972 	if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
973 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
974 		err = -EEXIST;
975 	} else {
976 		hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
977 	}
978 
979 	spin_unlock(&addrconf_hash_lock);
980 
981 	return err;
982 }
983 
984 /* On success it returns ifp with increased reference count */
985 
986 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)987 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
988 	      bool can_block, struct netlink_ext_ack *extack)
989 {
990 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
991 	int addr_type = ipv6_addr_type(cfg->pfx);
992 	struct net *net = dev_net(idev->dev);
993 	struct inet6_ifaddr *ifa = NULL;
994 	struct fib6_info *f6i = NULL;
995 	int err = 0;
996 
997 	if (addr_type == IPV6_ADDR_ANY ||
998 	    (addr_type & IPV6_ADDR_MULTICAST &&
999 	     !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1000 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1001 	     addr_type & IPV6_ADDR_LOOPBACK))
1002 		return ERR_PTR(-EADDRNOTAVAIL);
1003 
1004 	if (idev->dead) {
1005 		err = -ENODEV;			/*XXX*/
1006 		goto out;
1007 	}
1008 
1009 	if (idev->cnf.disable_ipv6) {
1010 		err = -EACCES;
1011 		goto out;
1012 	}
1013 
1014 	/* validator notifier needs to be blocking;
1015 	 * do not call in atomic context
1016 	 */
1017 	if (can_block) {
1018 		struct in6_validator_info i6vi = {
1019 			.i6vi_addr = *cfg->pfx,
1020 			.i6vi_dev = idev,
1021 			.extack = extack,
1022 		};
1023 
1024 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1025 		err = notifier_to_errno(err);
1026 		if (err < 0)
1027 			goto out;
1028 	}
1029 
1030 	ifa = kzalloc(sizeof(*ifa), gfp_flags);
1031 	if (!ifa) {
1032 		err = -ENOBUFS;
1033 		goto out;
1034 	}
1035 
1036 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1037 	if (IS_ERR(f6i)) {
1038 		err = PTR_ERR(f6i);
1039 		f6i = NULL;
1040 		goto out;
1041 	}
1042 
1043 	if (net->ipv6.devconf_all->disable_policy ||
1044 	    idev->cnf.disable_policy)
1045 		f6i->dst_nopolicy = true;
1046 
1047 	neigh_parms_data_state_setall(idev->nd_parms);
1048 
1049 	ifa->addr = *cfg->pfx;
1050 	if (cfg->peer_pfx)
1051 		ifa->peer_addr = *cfg->peer_pfx;
1052 
1053 	spin_lock_init(&ifa->lock);
1054 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1055 	INIT_HLIST_NODE(&ifa->addr_lst);
1056 	ifa->scope = cfg->scope;
1057 	ifa->prefix_len = cfg->plen;
1058 	ifa->rt_priority = cfg->rt_priority;
1059 	ifa->flags = cfg->ifa_flags;
1060 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1061 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1062 		ifa->flags |= IFA_F_TENTATIVE;
1063 	ifa->valid_lft = cfg->valid_lft;
1064 	ifa->prefered_lft = cfg->preferred_lft;
1065 	ifa->cstamp = ifa->tstamp = jiffies;
1066 	ifa->tokenized = false;
1067 
1068 	ifa->rt = f6i;
1069 
1070 	ifa->idev = idev;
1071 	in6_dev_hold(idev);
1072 
1073 	/* For caller */
1074 	refcount_set(&ifa->refcnt, 1);
1075 
1076 	rcu_read_lock_bh();
1077 
1078 	err = ipv6_add_addr_hash(idev->dev, ifa);
1079 	if (err < 0) {
1080 		rcu_read_unlock_bh();
1081 		goto out;
1082 	}
1083 
1084 	write_lock(&idev->lock);
1085 
1086 	/* Add to inet6_dev unicast addr list. */
1087 	ipv6_link_dev_addr(idev, ifa);
1088 
1089 	if (ifa->flags&IFA_F_TEMPORARY) {
1090 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1091 		in6_ifa_hold(ifa);
1092 	}
1093 
1094 	in6_ifa_hold(ifa);
1095 	write_unlock(&idev->lock);
1096 
1097 	rcu_read_unlock_bh();
1098 
1099 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1100 out:
1101 	if (unlikely(err < 0)) {
1102 		fib6_info_release(f6i);
1103 
1104 		if (ifa) {
1105 			if (ifa->idev)
1106 				in6_dev_put(ifa->idev);
1107 			kfree(ifa);
1108 		}
1109 		ifa = ERR_PTR(err);
1110 	}
1111 
1112 	return ifa;
1113 }
1114 
1115 enum cleanup_prefix_rt_t {
1116 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1117 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1118 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1119 };
1120 
1121 /*
1122  * Check, whether the prefix for ifp would still need a prefix route
1123  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1124  * constants.
1125  *
1126  * 1) we don't purge prefix if address was not permanent.
1127  *    prefix is managed by its own lifetime.
1128  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1129  * 3) if there are no addresses, delete prefix.
1130  * 4) if there are still other permanent address(es),
1131  *    corresponding prefix is still permanent.
1132  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1133  *    don't purge the prefix, assume user space is managing it.
1134  * 6) otherwise, update prefix lifetime to the
1135  *    longest valid lifetime among the corresponding
1136  *    addresses on the device.
1137  *    Note: subsequent RA will update lifetime.
1138  **/
1139 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1140 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1141 {
1142 	struct inet6_ifaddr *ifa;
1143 	struct inet6_dev *idev = ifp->idev;
1144 	unsigned long lifetime;
1145 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1146 
1147 	*expires = jiffies;
1148 
1149 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1150 		if (ifa == ifp)
1151 			continue;
1152 		if (ifa->prefix_len != ifp->prefix_len ||
1153 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1154 				       ifp->prefix_len))
1155 			continue;
1156 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1157 			return CLEANUP_PREFIX_RT_NOP;
1158 
1159 		action = CLEANUP_PREFIX_RT_EXPIRE;
1160 
1161 		spin_lock(&ifa->lock);
1162 
1163 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1164 		/*
1165 		 * Note: Because this address is
1166 		 * not permanent, lifetime <
1167 		 * LONG_MAX / HZ here.
1168 		 */
1169 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1170 			*expires = ifa->tstamp + lifetime * HZ;
1171 		spin_unlock(&ifa->lock);
1172 	}
1173 
1174 	return action;
1175 }
1176 
1177 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1178 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1179 		     bool del_rt, bool del_peer)
1180 {
1181 	struct fib6_info *f6i;
1182 
1183 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1184 				       ifp->prefix_len,
1185 				       ifp->idev->dev,
1186 				       0, RTF_GATEWAY | RTF_DEFAULT);
1187 	if (f6i) {
1188 		if (del_rt)
1189 			ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1190 		else {
1191 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1192 				fib6_set_expires(f6i, expires);
1193 			fib6_info_release(f6i);
1194 		}
1195 	}
1196 }
1197 
1198 
1199 /* This function wants to get referenced ifp and releases it before return */
1200 
ipv6_del_addr(struct inet6_ifaddr * ifp)1201 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1202 {
1203 	int state;
1204 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1205 	unsigned long expires;
1206 
1207 	ASSERT_RTNL();
1208 
1209 	spin_lock_bh(&ifp->lock);
1210 	state = ifp->state;
1211 	ifp->state = INET6_IFADDR_STATE_DEAD;
1212 	spin_unlock_bh(&ifp->lock);
1213 
1214 	if (state == INET6_IFADDR_STATE_DEAD)
1215 		goto out;
1216 
1217 	spin_lock_bh(&addrconf_hash_lock);
1218 	hlist_del_init_rcu(&ifp->addr_lst);
1219 	spin_unlock_bh(&addrconf_hash_lock);
1220 
1221 	write_lock_bh(&ifp->idev->lock);
1222 
1223 	if (ifp->flags&IFA_F_TEMPORARY) {
1224 		list_del(&ifp->tmp_list);
1225 		if (ifp->ifpub) {
1226 			in6_ifa_put(ifp->ifpub);
1227 			ifp->ifpub = NULL;
1228 		}
1229 		__in6_ifa_put(ifp);
1230 	}
1231 
1232 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1233 		action = check_cleanup_prefix_route(ifp, &expires);
1234 
1235 	list_del_rcu(&ifp->if_list);
1236 	__in6_ifa_put(ifp);
1237 
1238 	write_unlock_bh(&ifp->idev->lock);
1239 
1240 	addrconf_del_dad_work(ifp);
1241 
1242 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1243 
1244 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1245 
1246 	if (action != CLEANUP_PREFIX_RT_NOP) {
1247 		cleanup_prefix_route(ifp, expires,
1248 			action == CLEANUP_PREFIX_RT_DEL, false);
1249 	}
1250 
1251 	/* clean up prefsrc entries */
1252 	rt6_remove_prefsrc(ifp);
1253 out:
1254 	in6_ifa_put(ifp);
1255 }
1256 
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,struct inet6_ifaddr * ift,bool block)1257 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1258 				struct inet6_ifaddr *ift,
1259 				bool block)
1260 {
1261 	struct inet6_dev *idev = ifp->idev;
1262 	struct in6_addr addr, *tmpaddr;
1263 	unsigned long tmp_tstamp, age;
1264 	unsigned long regen_advance;
1265 	struct ifa6_config cfg;
1266 	int ret = 0;
1267 	unsigned long now = jiffies;
1268 	long max_desync_factor;
1269 	s32 cnf_temp_preferred_lft;
1270 
1271 	write_lock_bh(&idev->lock);
1272 	if (ift) {
1273 		spin_lock_bh(&ift->lock);
1274 		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1275 		spin_unlock_bh(&ift->lock);
1276 		tmpaddr = &addr;
1277 	} else {
1278 		tmpaddr = NULL;
1279 	}
1280 retry:
1281 	in6_dev_hold(idev);
1282 	if (idev->cnf.use_tempaddr <= 0) {
1283 		write_unlock_bh(&idev->lock);
1284 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1285 		in6_dev_put(idev);
1286 		ret = -1;
1287 		goto out;
1288 	}
1289 	spin_lock_bh(&ifp->lock);
1290 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1291 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1292 		spin_unlock_bh(&ifp->lock);
1293 		write_unlock_bh(&idev->lock);
1294 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1295 			__func__);
1296 		in6_dev_put(idev);
1297 		ret = -1;
1298 		goto out;
1299 	}
1300 	in6_ifa_hold(ifp);
1301 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1302 	ipv6_try_regen_rndid(idev, tmpaddr);
1303 	memcpy(&addr.s6_addr[8], idev->rndid, 8);
1304 	age = (now - ifp->tstamp) / HZ;
1305 
1306 	regen_advance = idev->cnf.regen_max_retry *
1307 			idev->cnf.dad_transmits *
1308 			NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1309 
1310 	/* recalculate max_desync_factor each time and update
1311 	 * idev->desync_factor if it's larger
1312 	 */
1313 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1314 	max_desync_factor = min_t(__u32,
1315 				  idev->cnf.max_desync_factor,
1316 				  cnf_temp_preferred_lft - regen_advance);
1317 
1318 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1319 		if (max_desync_factor > 0) {
1320 			get_random_bytes(&idev->desync_factor,
1321 					 sizeof(idev->desync_factor));
1322 			idev->desync_factor %= max_desync_factor;
1323 		} else {
1324 			idev->desync_factor = 0;
1325 		}
1326 	}
1327 
1328 	memset(&cfg, 0, sizeof(cfg));
1329 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1330 			      idev->cnf.temp_valid_lft + age);
1331 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1332 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1333 
1334 	cfg.plen = ifp->prefix_len;
1335 	tmp_tstamp = ifp->tstamp;
1336 	spin_unlock_bh(&ifp->lock);
1337 
1338 	write_unlock_bh(&idev->lock);
1339 
1340 	/* A temporary address is created only if this calculated Preferred
1341 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1342 	 * an implementation must not create a temporary address with a zero
1343 	 * Preferred Lifetime.
1344 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1345 	 * temporary addresses being generated.
1346 	 */
1347 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1348 	if (cfg.preferred_lft <= regen_advance + age) {
1349 		in6_ifa_put(ifp);
1350 		in6_dev_put(idev);
1351 		ret = -1;
1352 		goto out;
1353 	}
1354 
1355 	cfg.ifa_flags = IFA_F_TEMPORARY;
1356 	/* set in addrconf_prefix_rcv() */
1357 	if (ifp->flags & IFA_F_OPTIMISTIC)
1358 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1359 
1360 	cfg.pfx = &addr;
1361 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1362 
1363 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1364 	if (IS_ERR(ift)) {
1365 		in6_ifa_put(ifp);
1366 		in6_dev_put(idev);
1367 		pr_info("%s: retry temporary address regeneration\n", __func__);
1368 		tmpaddr = &addr;
1369 		write_lock_bh(&idev->lock);
1370 		goto retry;
1371 	}
1372 
1373 	spin_lock_bh(&ift->lock);
1374 	ift->ifpub = ifp;
1375 	ift->cstamp = now;
1376 	ift->tstamp = tmp_tstamp;
1377 	spin_unlock_bh(&ift->lock);
1378 
1379 	addrconf_dad_start(ift);
1380 	in6_ifa_put(ift);
1381 	in6_dev_put(idev);
1382 out:
1383 	return ret;
1384 }
1385 
1386 /*
1387  *	Choose an appropriate source address (RFC3484)
1388  */
1389 enum {
1390 	IPV6_SADDR_RULE_INIT = 0,
1391 	IPV6_SADDR_RULE_LOCAL,
1392 	IPV6_SADDR_RULE_SCOPE,
1393 	IPV6_SADDR_RULE_PREFERRED,
1394 #ifdef CONFIG_IPV6_MIP6
1395 	IPV6_SADDR_RULE_HOA,
1396 #endif
1397 	IPV6_SADDR_RULE_OIF,
1398 	IPV6_SADDR_RULE_LABEL,
1399 	IPV6_SADDR_RULE_PRIVACY,
1400 	IPV6_SADDR_RULE_ORCHID,
1401 	IPV6_SADDR_RULE_PREFIX,
1402 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1403 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1404 #endif
1405 	IPV6_SADDR_RULE_MAX
1406 };
1407 
1408 struct ipv6_saddr_score {
1409 	int			rule;
1410 	int			addr_type;
1411 	struct inet6_ifaddr	*ifa;
1412 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1413 	int			scopedist;
1414 	int			matchlen;
1415 };
1416 
1417 struct ipv6_saddr_dst {
1418 	const struct in6_addr *addr;
1419 	int ifindex;
1420 	int scope;
1421 	int label;
1422 	unsigned int prefs;
1423 };
1424 
ipv6_saddr_preferred(int type)1425 static inline int ipv6_saddr_preferred(int type)
1426 {
1427 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1428 		return 1;
1429 	return 0;
1430 }
1431 
ipv6_use_optimistic_addr(struct net * net,struct inet6_dev * idev)1432 static bool ipv6_use_optimistic_addr(struct net *net,
1433 				     struct inet6_dev *idev)
1434 {
1435 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1436 	if (!idev)
1437 		return false;
1438 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1439 		return false;
1440 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1441 		return false;
1442 
1443 	return true;
1444 #else
1445 	return false;
1446 #endif
1447 }
1448 
ipv6_allow_optimistic_dad(struct net * net,struct inet6_dev * idev)1449 static bool ipv6_allow_optimistic_dad(struct net *net,
1450 				      struct inet6_dev *idev)
1451 {
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453 	if (!idev)
1454 		return false;
1455 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1456 		return false;
1457 
1458 	return true;
1459 #else
1460 	return false;
1461 #endif
1462 }
1463 
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1464 static int ipv6_get_saddr_eval(struct net *net,
1465 			       struct ipv6_saddr_score *score,
1466 			       struct ipv6_saddr_dst *dst,
1467 			       int i)
1468 {
1469 	int ret;
1470 
1471 	if (i <= score->rule) {
1472 		switch (i) {
1473 		case IPV6_SADDR_RULE_SCOPE:
1474 			ret = score->scopedist;
1475 			break;
1476 		case IPV6_SADDR_RULE_PREFIX:
1477 			ret = score->matchlen;
1478 			break;
1479 		default:
1480 			ret = !!test_bit(i, score->scorebits);
1481 		}
1482 		goto out;
1483 	}
1484 
1485 	switch (i) {
1486 	case IPV6_SADDR_RULE_INIT:
1487 		/* Rule 0: remember if hiscore is not ready yet */
1488 		ret = !!score->ifa;
1489 		break;
1490 	case IPV6_SADDR_RULE_LOCAL:
1491 		/* Rule 1: Prefer same address */
1492 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1493 		break;
1494 	case IPV6_SADDR_RULE_SCOPE:
1495 		/* Rule 2: Prefer appropriate scope
1496 		 *
1497 		 *      ret
1498 		 *       ^
1499 		 *    -1 |  d 15
1500 		 *    ---+--+-+---> scope
1501 		 *       |
1502 		 *       |             d is scope of the destination.
1503 		 *  B-d  |  \
1504 		 *       |   \      <- smaller scope is better if
1505 		 *  B-15 |    \        if scope is enough for destination.
1506 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1507 		 * d-C-1 | /
1508 		 *       |/         <- greater is better
1509 		 *   -C  /             if scope is not enough for destination.
1510 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1511 		 *
1512 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1513 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1514 		 * Assume B = 0 and we get C > 29.
1515 		 */
1516 		ret = __ipv6_addr_src_scope(score->addr_type);
1517 		if (ret >= dst->scope)
1518 			ret = -ret;
1519 		else
1520 			ret -= 128;	/* 30 is enough */
1521 		score->scopedist = ret;
1522 		break;
1523 	case IPV6_SADDR_RULE_PREFERRED:
1524 	    {
1525 		/* Rule 3: Avoid deprecated and optimistic addresses */
1526 		u8 avoid = IFA_F_DEPRECATED;
1527 
1528 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1529 			avoid |= IFA_F_OPTIMISTIC;
1530 		ret = ipv6_saddr_preferred(score->addr_type) ||
1531 		      !(score->ifa->flags & avoid);
1532 		break;
1533 	    }
1534 #ifdef CONFIG_IPV6_MIP6
1535 	case IPV6_SADDR_RULE_HOA:
1536 	    {
1537 		/* Rule 4: Prefer home address */
1538 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1539 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1540 		break;
1541 	    }
1542 #endif
1543 	case IPV6_SADDR_RULE_OIF:
1544 		/* Rule 5: Prefer outgoing interface */
1545 		ret = (!dst->ifindex ||
1546 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1547 		break;
1548 	case IPV6_SADDR_RULE_LABEL:
1549 		/* Rule 6: Prefer matching label */
1550 		ret = ipv6_addr_label(net,
1551 				      &score->ifa->addr, score->addr_type,
1552 				      score->ifa->idev->dev->ifindex) == dst->label;
1553 		break;
1554 	case IPV6_SADDR_RULE_PRIVACY:
1555 	    {
1556 		/* Rule 7: Prefer public address
1557 		 * Note: prefer temporary address if use_tempaddr >= 2
1558 		 */
1559 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1560 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1561 				score->ifa->idev->cnf.use_tempaddr >= 2;
1562 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1563 		break;
1564 	    }
1565 	case IPV6_SADDR_RULE_ORCHID:
1566 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1567 		 *	    non-ORCHID vs non-ORCHID
1568 		 */
1569 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1570 			ipv6_addr_orchid(dst->addr));
1571 		break;
1572 	case IPV6_SADDR_RULE_PREFIX:
1573 		/* Rule 8: Use longest matching prefix */
1574 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1575 		if (ret > score->ifa->prefix_len)
1576 			ret = score->ifa->prefix_len;
1577 		score->matchlen = ret;
1578 		break;
1579 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1580 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1581 		/* Optimistic addresses still have lower precedence than other
1582 		 * preferred addresses.
1583 		 */
1584 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1585 		break;
1586 #endif
1587 	default:
1588 		ret = 0;
1589 	}
1590 
1591 	if (ret)
1592 		__set_bit(i, score->scorebits);
1593 	score->rule = i;
1594 out:
1595 	return ret;
1596 }
1597 
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1598 static int __ipv6_dev_get_saddr(struct net *net,
1599 				struct ipv6_saddr_dst *dst,
1600 				struct inet6_dev *idev,
1601 				struct ipv6_saddr_score *scores,
1602 				int hiscore_idx)
1603 {
1604 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1605 
1606 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1607 		int i;
1608 
1609 		/*
1610 		 * - Tentative Address (RFC2462 section 5.4)
1611 		 *  - A tentative address is not considered
1612 		 *    "assigned to an interface" in the traditional
1613 		 *    sense, unless it is also flagged as optimistic.
1614 		 * - Candidate Source Address (section 4)
1615 		 *  - In any case, anycast addresses, multicast
1616 		 *    addresses, and the unspecified address MUST
1617 		 *    NOT be included in a candidate set.
1618 		 */
1619 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1620 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1621 			continue;
1622 
1623 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1624 
1625 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1626 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1627 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1628 					    idev->dev->name);
1629 			continue;
1630 		}
1631 
1632 		score->rule = -1;
1633 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1634 
1635 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1636 			int minihiscore, miniscore;
1637 
1638 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1639 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1640 
1641 			if (minihiscore > miniscore) {
1642 				if (i == IPV6_SADDR_RULE_SCOPE &&
1643 				    score->scopedist > 0) {
1644 					/*
1645 					 * special case:
1646 					 * each remaining entry
1647 					 * has too small (not enough)
1648 					 * scope, because ifa entries
1649 					 * are sorted by their scope
1650 					 * values.
1651 					 */
1652 					goto out;
1653 				}
1654 				break;
1655 			} else if (minihiscore < miniscore) {
1656 				swap(hiscore, score);
1657 				hiscore_idx = 1 - hiscore_idx;
1658 
1659 				/* restore our iterator */
1660 				score->ifa = hiscore->ifa;
1661 
1662 				break;
1663 			}
1664 		}
1665 	}
1666 out:
1667 	return hiscore_idx;
1668 }
1669 
ipv6_get_saddr_master(struct net * net,const struct net_device * dst_dev,const struct net_device * master,struct ipv6_saddr_dst * dst,struct ipv6_saddr_score * scores,int hiscore_idx)1670 static int ipv6_get_saddr_master(struct net *net,
1671 				 const struct net_device *dst_dev,
1672 				 const struct net_device *master,
1673 				 struct ipv6_saddr_dst *dst,
1674 				 struct ipv6_saddr_score *scores,
1675 				 int hiscore_idx)
1676 {
1677 	struct inet6_dev *idev;
1678 
1679 	idev = __in6_dev_get(dst_dev);
1680 	if (idev)
1681 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1682 						   scores, hiscore_idx);
1683 
1684 	idev = __in6_dev_get(master);
1685 	if (idev)
1686 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1687 						   scores, hiscore_idx);
1688 
1689 	return hiscore_idx;
1690 }
1691 
ipv6_dev_get_saddr(struct net * net,const struct net_device * dst_dev,const struct in6_addr * daddr,unsigned int prefs,struct in6_addr * saddr)1692 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1693 		       const struct in6_addr *daddr, unsigned int prefs,
1694 		       struct in6_addr *saddr)
1695 {
1696 	struct ipv6_saddr_score scores[2], *hiscore;
1697 	struct ipv6_saddr_dst dst;
1698 	struct inet6_dev *idev;
1699 	struct net_device *dev;
1700 	int dst_type;
1701 	bool use_oif_addr = false;
1702 	int hiscore_idx = 0;
1703 	int ret = 0;
1704 
1705 	dst_type = __ipv6_addr_type(daddr);
1706 	dst.addr = daddr;
1707 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1708 	dst.scope = __ipv6_addr_src_scope(dst_type);
1709 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1710 	dst.prefs = prefs;
1711 
1712 	scores[hiscore_idx].rule = -1;
1713 	scores[hiscore_idx].ifa = NULL;
1714 
1715 	rcu_read_lock();
1716 
1717 	/* Candidate Source Address (section 4)
1718 	 *  - multicast and link-local destination address,
1719 	 *    the set of candidate source address MUST only
1720 	 *    include addresses assigned to interfaces
1721 	 *    belonging to the same link as the outgoing
1722 	 *    interface.
1723 	 * (- For site-local destination addresses, the
1724 	 *    set of candidate source addresses MUST only
1725 	 *    include addresses assigned to interfaces
1726 	 *    belonging to the same site as the outgoing
1727 	 *    interface.)
1728 	 *  - "It is RECOMMENDED that the candidate source addresses
1729 	 *    be the set of unicast addresses assigned to the
1730 	 *    interface that will be used to send to the destination
1731 	 *    (the 'outgoing' interface)." (RFC 6724)
1732 	 */
1733 	if (dst_dev) {
1734 		idev = __in6_dev_get(dst_dev);
1735 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1736 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1737 		    (idev && idev->cnf.use_oif_addrs_only)) {
1738 			use_oif_addr = true;
1739 		}
1740 	}
1741 
1742 	if (use_oif_addr) {
1743 		if (idev)
1744 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1745 	} else {
1746 		const struct net_device *master;
1747 		int master_idx = 0;
1748 
1749 		/* if dst_dev exists and is enslaved to an L3 device, then
1750 		 * prefer addresses from dst_dev and then the master over
1751 		 * any other enslaved devices in the L3 domain.
1752 		 */
1753 		master = l3mdev_master_dev_rcu(dst_dev);
1754 		if (master) {
1755 			master_idx = master->ifindex;
1756 
1757 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1758 							    master, &dst,
1759 							    scores, hiscore_idx);
1760 
1761 			if (scores[hiscore_idx].ifa)
1762 				goto out;
1763 		}
1764 
1765 		for_each_netdev_rcu(net, dev) {
1766 			/* only consider addresses on devices in the
1767 			 * same L3 domain
1768 			 */
1769 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1770 				continue;
1771 			idev = __in6_dev_get(dev);
1772 			if (!idev)
1773 				continue;
1774 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1775 		}
1776 	}
1777 
1778 out:
1779 	hiscore = &scores[hiscore_idx];
1780 	if (!hiscore->ifa)
1781 		ret = -EADDRNOTAVAIL;
1782 	else
1783 		*saddr = hiscore->ifa->addr;
1784 
1785 	rcu_read_unlock();
1786 	return ret;
1787 }
1788 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1789 
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1790 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1791 		      u32 banned_flags)
1792 {
1793 	struct inet6_ifaddr *ifp;
1794 	int err = -EADDRNOTAVAIL;
1795 
1796 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1797 		if (ifp->scope > IFA_LINK)
1798 			break;
1799 		if (ifp->scope == IFA_LINK &&
1800 		    !(ifp->flags & banned_flags)) {
1801 			*addr = ifp->addr;
1802 			err = 0;
1803 			break;
1804 		}
1805 	}
1806 	return err;
1807 }
1808 
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1809 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1810 		    u32 banned_flags)
1811 {
1812 	struct inet6_dev *idev;
1813 	int err = -EADDRNOTAVAIL;
1814 
1815 	rcu_read_lock();
1816 	idev = __in6_dev_get(dev);
1817 	if (idev) {
1818 		read_lock_bh(&idev->lock);
1819 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1820 		read_unlock_bh(&idev->lock);
1821 	}
1822 	rcu_read_unlock();
1823 	return err;
1824 }
1825 
ipv6_count_addresses(const struct inet6_dev * idev)1826 static int ipv6_count_addresses(const struct inet6_dev *idev)
1827 {
1828 	const struct inet6_ifaddr *ifp;
1829 	int cnt = 0;
1830 
1831 	rcu_read_lock();
1832 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1833 		cnt++;
1834 	rcu_read_unlock();
1835 	return cnt;
1836 }
1837 
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1838 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1839 		  const struct net_device *dev, int strict)
1840 {
1841 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1842 				       strict, IFA_F_TENTATIVE);
1843 }
1844 EXPORT_SYMBOL(ipv6_chk_addr);
1845 
1846 /* device argument is used to find the L3 domain of interest. If
1847  * skip_dev_check is set, then the ifp device is not checked against
1848  * the passed in dev argument. So the 2 cases for addresses checks are:
1849  *   1. does the address exist in the L3 domain that dev is part of
1850  *      (skip_dev_check = true), or
1851  *
1852  *   2. does the address exist on the specific device
1853  *      (skip_dev_check = false)
1854  */
ipv6_chk_addr_and_flags(struct net * net,const struct in6_addr * addr,const struct net_device * dev,bool skip_dev_check,int strict,u32 banned_flags)1855 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1856 			    const struct net_device *dev, bool skip_dev_check,
1857 			    int strict, u32 banned_flags)
1858 {
1859 	unsigned int hash = inet6_addr_hash(net, addr);
1860 	const struct net_device *l3mdev;
1861 	struct inet6_ifaddr *ifp;
1862 	u32 ifp_flags;
1863 
1864 	rcu_read_lock();
1865 
1866 	l3mdev = l3mdev_master_dev_rcu(dev);
1867 	if (skip_dev_check)
1868 		dev = NULL;
1869 
1870 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1871 		if (!net_eq(dev_net(ifp->idev->dev), net))
1872 			continue;
1873 
1874 		if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1875 			continue;
1876 
1877 		/* Decouple optimistic from tentative for evaluation here.
1878 		 * Ban optimistic addresses explicitly, when required.
1879 		 */
1880 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1881 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1882 			    : ifp->flags;
1883 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1884 		    !(ifp_flags&banned_flags) &&
1885 		    (!dev || ifp->idev->dev == dev ||
1886 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1887 			rcu_read_unlock();
1888 			return 1;
1889 		}
1890 	}
1891 
1892 	rcu_read_unlock();
1893 	return 0;
1894 }
1895 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1896 
1897 
1898 /* Compares an address/prefix_len with addresses on device @dev.
1899  * If one is found it returns true.
1900  */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)1901 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1902 	const unsigned int prefix_len, struct net_device *dev)
1903 {
1904 	const struct inet6_ifaddr *ifa;
1905 	const struct inet6_dev *idev;
1906 	bool ret = false;
1907 
1908 	rcu_read_lock();
1909 	idev = __in6_dev_get(dev);
1910 	if (idev) {
1911 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1912 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1913 			if (ret)
1914 				break;
1915 		}
1916 	}
1917 	rcu_read_unlock();
1918 
1919 	return ret;
1920 }
1921 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1922 
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)1923 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1924 {
1925 	const struct inet6_ifaddr *ifa;
1926 	const struct inet6_dev *idev;
1927 	int	onlink;
1928 
1929 	onlink = 0;
1930 	rcu_read_lock();
1931 	idev = __in6_dev_get(dev);
1932 	if (idev) {
1933 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1934 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1935 						   ifa->prefix_len);
1936 			if (onlink)
1937 				break;
1938 		}
1939 	}
1940 	rcu_read_unlock();
1941 	return onlink;
1942 }
1943 EXPORT_SYMBOL(ipv6_chk_prefix);
1944 
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)1945 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1946 				     struct net_device *dev, int strict)
1947 {
1948 	unsigned int hash = inet6_addr_hash(net, addr);
1949 	struct inet6_ifaddr *ifp, *result = NULL;
1950 
1951 	rcu_read_lock();
1952 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1953 		if (!net_eq(dev_net(ifp->idev->dev), net))
1954 			continue;
1955 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1956 			if (!dev || ifp->idev->dev == dev ||
1957 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1958 				result = ifp;
1959 				in6_ifa_hold(ifp);
1960 				break;
1961 			}
1962 		}
1963 	}
1964 	rcu_read_unlock();
1965 
1966 	return result;
1967 }
1968 
1969 /* Gets referenced address, destroys ifaddr */
1970 
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)1971 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1972 {
1973 	if (dad_failed)
1974 		ifp->flags |= IFA_F_DADFAILED;
1975 
1976 	if (ifp->flags&IFA_F_TEMPORARY) {
1977 		struct inet6_ifaddr *ifpub;
1978 		spin_lock_bh(&ifp->lock);
1979 		ifpub = ifp->ifpub;
1980 		if (ifpub) {
1981 			in6_ifa_hold(ifpub);
1982 			spin_unlock_bh(&ifp->lock);
1983 			ipv6_create_tempaddr(ifpub, ifp, true);
1984 			in6_ifa_put(ifpub);
1985 		} else {
1986 			spin_unlock_bh(&ifp->lock);
1987 		}
1988 		ipv6_del_addr(ifp);
1989 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1990 		spin_lock_bh(&ifp->lock);
1991 		addrconf_del_dad_work(ifp);
1992 		ifp->flags |= IFA_F_TENTATIVE;
1993 		if (dad_failed)
1994 			ifp->flags &= ~IFA_F_OPTIMISTIC;
1995 		spin_unlock_bh(&ifp->lock);
1996 		if (dad_failed)
1997 			ipv6_ifa_notify(0, ifp);
1998 		in6_ifa_put(ifp);
1999 	} else {
2000 		ipv6_del_addr(ifp);
2001 	}
2002 }
2003 
addrconf_dad_end(struct inet6_ifaddr * ifp)2004 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2005 {
2006 	int err = -ENOENT;
2007 
2008 	spin_lock_bh(&ifp->lock);
2009 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2010 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2011 		err = 0;
2012 	}
2013 	spin_unlock_bh(&ifp->lock);
2014 
2015 	return err;
2016 }
2017 
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2018 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2019 {
2020 	struct inet6_dev *idev = ifp->idev;
2021 	struct net *net = dev_net(ifp->idev->dev);
2022 
2023 	if (addrconf_dad_end(ifp)) {
2024 		in6_ifa_put(ifp);
2025 		return;
2026 	}
2027 
2028 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2029 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2030 
2031 	spin_lock_bh(&ifp->lock);
2032 
2033 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2034 		struct in6_addr new_addr;
2035 		struct inet6_ifaddr *ifp2;
2036 		int retries = ifp->stable_privacy_retry + 1;
2037 		struct ifa6_config cfg = {
2038 			.pfx = &new_addr,
2039 			.plen = ifp->prefix_len,
2040 			.ifa_flags = ifp->flags,
2041 			.valid_lft = ifp->valid_lft,
2042 			.preferred_lft = ifp->prefered_lft,
2043 			.scope = ifp->scope,
2044 		};
2045 
2046 		if (retries > net->ipv6.sysctl.idgen_retries) {
2047 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2048 					     ifp->idev->dev->name);
2049 			goto errdad;
2050 		}
2051 
2052 		new_addr = ifp->addr;
2053 		if (ipv6_generate_stable_address(&new_addr, retries,
2054 						 idev))
2055 			goto errdad;
2056 
2057 		spin_unlock_bh(&ifp->lock);
2058 
2059 		if (idev->cnf.max_addresses &&
2060 		    ipv6_count_addresses(idev) >=
2061 		    idev->cnf.max_addresses)
2062 			goto lock_errdad;
2063 
2064 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2065 				     ifp->idev->dev->name);
2066 
2067 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2068 		if (IS_ERR(ifp2))
2069 			goto lock_errdad;
2070 
2071 		spin_lock_bh(&ifp2->lock);
2072 		ifp2->stable_privacy_retry = retries;
2073 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2074 		spin_unlock_bh(&ifp2->lock);
2075 
2076 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2077 		in6_ifa_put(ifp2);
2078 lock_errdad:
2079 		spin_lock_bh(&ifp->lock);
2080 	}
2081 
2082 errdad:
2083 	/* transition from _POSTDAD to _ERRDAD */
2084 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2085 	spin_unlock_bh(&ifp->lock);
2086 
2087 	addrconf_mod_dad_work(ifp, 0);
2088 	in6_ifa_put(ifp);
2089 }
2090 
2091 /* Join to solicited addr multicast group.
2092  * caller must hold RTNL */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2093 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2094 {
2095 	struct in6_addr maddr;
2096 
2097 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2098 		return;
2099 
2100 	addrconf_addr_solict_mult(addr, &maddr);
2101 	ipv6_dev_mc_inc(dev, &maddr);
2102 }
2103 
2104 /* caller must hold RTNL */
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2105 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2106 {
2107 	struct in6_addr maddr;
2108 
2109 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2110 		return;
2111 
2112 	addrconf_addr_solict_mult(addr, &maddr);
2113 	__ipv6_dev_mc_dec(idev, &maddr);
2114 }
2115 
2116 /* caller must hold RTNL */
addrconf_join_anycast(struct inet6_ifaddr * ifp)2117 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2118 {
2119 	struct in6_addr addr;
2120 
2121 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2122 		return;
2123 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2124 	if (ipv6_addr_any(&addr))
2125 		return;
2126 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2127 }
2128 
2129 /* caller must hold RTNL */
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2130 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2131 {
2132 	struct in6_addr addr;
2133 
2134 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2135 		return;
2136 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2137 	if (ipv6_addr_any(&addr))
2138 		return;
2139 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2140 }
2141 
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2142 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2143 {
2144 	switch (dev->addr_len) {
2145 	case ETH_ALEN:
2146 		memcpy(eui, dev->dev_addr, 3);
2147 		eui[3] = 0xFF;
2148 		eui[4] = 0xFE;
2149 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2150 		break;
2151 	case EUI64_ADDR_LEN:
2152 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2153 		eui[0] ^= 2;
2154 		break;
2155 	default:
2156 		return -1;
2157 	}
2158 
2159 	return 0;
2160 }
2161 
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2162 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2163 {
2164 	union fwnet_hwaddr *ha;
2165 
2166 	if (dev->addr_len != FWNET_ALEN)
2167 		return -1;
2168 
2169 	ha = (union fwnet_hwaddr *)dev->dev_addr;
2170 
2171 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2172 	eui[0] ^= 2;
2173 	return 0;
2174 }
2175 
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2176 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2177 {
2178 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2179 	if (dev->addr_len != ARCNET_ALEN)
2180 		return -1;
2181 	memset(eui, 0, 7);
2182 	eui[7] = *(u8 *)dev->dev_addr;
2183 	return 0;
2184 }
2185 
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2186 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2187 {
2188 	if (dev->addr_len != INFINIBAND_ALEN)
2189 		return -1;
2190 	memcpy(eui, dev->dev_addr + 12, 8);
2191 	eui[0] |= 2;
2192 	return 0;
2193 }
2194 
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2195 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2196 {
2197 	if (addr == 0)
2198 		return -1;
2199 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2200 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2201 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2202 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2203 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2204 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2205 	eui[1] = 0;
2206 	eui[2] = 0x5E;
2207 	eui[3] = 0xFE;
2208 	memcpy(eui + 4, &addr, 4);
2209 	return 0;
2210 }
2211 
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2212 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2213 {
2214 	if (dev->priv_flags & IFF_ISATAP)
2215 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2216 	return -1;
2217 }
2218 
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2219 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2220 {
2221 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2222 }
2223 
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2224 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2225 {
2226 	memcpy(eui, dev->perm_addr, 3);
2227 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2228 	eui[3] = 0xFF;
2229 	eui[4] = 0xFE;
2230 	eui[0] ^= 2;
2231 	return 0;
2232 }
2233 
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2234 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2235 {
2236 	switch (dev->type) {
2237 	case ARPHRD_ETHER:
2238 	case ARPHRD_FDDI:
2239 		return addrconf_ifid_eui48(eui, dev);
2240 	case ARPHRD_ARCNET:
2241 		return addrconf_ifid_arcnet(eui, dev);
2242 	case ARPHRD_INFINIBAND:
2243 		return addrconf_ifid_infiniband(eui, dev);
2244 	case ARPHRD_SIT:
2245 		return addrconf_ifid_sit(eui, dev);
2246 	case ARPHRD_IPGRE:
2247 	case ARPHRD_TUNNEL:
2248 		return addrconf_ifid_gre(eui, dev);
2249 	case ARPHRD_6LOWPAN:
2250 		return addrconf_ifid_6lowpan(eui, dev);
2251 	case ARPHRD_IEEE1394:
2252 		return addrconf_ifid_ieee1394(eui, dev);
2253 	case ARPHRD_TUNNEL6:
2254 	case ARPHRD_IP6GRE:
2255 	case ARPHRD_RAWIP:
2256 		return addrconf_ifid_ip6tnl(eui, dev);
2257 	}
2258 	return -1;
2259 }
2260 
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2261 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2262 {
2263 	int err = -1;
2264 	struct inet6_ifaddr *ifp;
2265 
2266 	read_lock_bh(&idev->lock);
2267 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2268 		if (ifp->scope > IFA_LINK)
2269 			break;
2270 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2271 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2272 			err = 0;
2273 			break;
2274 		}
2275 	}
2276 	read_unlock_bh(&idev->lock);
2277 	return err;
2278 }
2279 
2280 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
ipv6_regen_rndid(struct inet6_dev * idev)2281 static void ipv6_regen_rndid(struct inet6_dev *idev)
2282 {
2283 regen:
2284 	get_random_bytes(idev->rndid, sizeof(idev->rndid));
2285 	idev->rndid[0] &= ~0x02;
2286 
2287 	/*
2288 	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2289 	 * check if generated address is not inappropriate
2290 	 *
2291 	 *  - Reserved subnet anycast (RFC 2526)
2292 	 *	11111101 11....11 1xxxxxxx
2293 	 *  - ISATAP (RFC4214) 6.1
2294 	 *	00-00-5E-FE-xx-xx-xx-xx
2295 	 *  - value 0
2296 	 *  - XXX: already assigned to an address on the device
2297 	 */
2298 	if (idev->rndid[0] == 0xfd &&
2299 	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2300 	    (idev->rndid[7]&0x80))
2301 		goto regen;
2302 	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2303 		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2304 			goto regen;
2305 		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2306 			goto regen;
2307 	}
2308 }
2309 
ipv6_try_regen_rndid(struct inet6_dev * idev,struct in6_addr * tmpaddr)2310 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2311 {
2312 	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2313 		ipv6_regen_rndid(idev);
2314 }
2315 
2316 /*
2317  *	Add prefix route.
2318  */
2319 
2320 static void
addrconf_prefix_route(struct in6_addr * pfx,int plen,u32 metric,struct net_device * dev,unsigned long expires,u32 flags,gfp_t gfp_flags)2321 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2322 		      struct net_device *dev, unsigned long expires,
2323 		      u32 flags, gfp_t gfp_flags)
2324 {
2325 	struct fib6_config cfg = {
2326 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2327 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2328 		.fc_ifindex = dev->ifindex,
2329 		.fc_expires = expires,
2330 		.fc_dst_len = plen,
2331 		.fc_flags = RTF_UP | flags,
2332 		.fc_nlinfo.nl_net = dev_net(dev),
2333 		.fc_protocol = RTPROT_KERNEL,
2334 		.fc_type = RTN_UNICAST,
2335 	};
2336 
2337 	cfg.fc_dst = *pfx;
2338 
2339 	/* Prevent useless cloning on PtP SIT.
2340 	   This thing is done here expecting that the whole
2341 	   class of non-broadcast devices need not cloning.
2342 	 */
2343 #if IS_ENABLED(CONFIG_IPV6_SIT)
2344 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2345 		cfg.fc_flags |= RTF_NONEXTHOP;
2346 #endif
2347 
2348 	ip6_route_add(&cfg, gfp_flags, NULL);
2349 }
2350 
2351 
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags)2352 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2353 						  int plen,
2354 						  const struct net_device *dev,
2355 						  u32 flags, u32 noflags)
2356 {
2357 	struct fib6_node *fn;
2358 	struct fib6_info *rt = NULL;
2359 	struct fib6_table *table;
2360 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2361 
2362 	table = fib6_get_table(dev_net(dev), tb_id);
2363 	if (!table)
2364 		return NULL;
2365 
2366 	rcu_read_lock();
2367 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2368 	if (!fn)
2369 		goto out;
2370 
2371 	for_each_fib6_node_rt_rcu(fn) {
2372 		if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2373 			continue;
2374 		if ((rt->fib6_flags & flags) != flags)
2375 			continue;
2376 		if ((rt->fib6_flags & noflags) != 0)
2377 			continue;
2378 		if (!fib6_info_hold_safe(rt))
2379 			continue;
2380 		break;
2381 	}
2382 out:
2383 	rcu_read_unlock();
2384 	return rt;
2385 }
2386 
2387 
2388 /* Create "default" multicast route to the interface */
2389 
addrconf_add_mroute(struct net_device * dev)2390 static void addrconf_add_mroute(struct net_device *dev)
2391 {
2392 	struct fib6_config cfg = {
2393 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2394 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2395 		.fc_ifindex = dev->ifindex,
2396 		.fc_dst_len = 8,
2397 		.fc_flags = RTF_UP,
2398 		.fc_type = RTN_UNICAST,
2399 		.fc_nlinfo.nl_net = dev_net(dev),
2400 	};
2401 
2402 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2403 
2404 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2405 }
2406 
addrconf_add_dev(struct net_device * dev)2407 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2408 {
2409 	struct inet6_dev *idev;
2410 
2411 	ASSERT_RTNL();
2412 
2413 	idev = ipv6_find_idev(dev);
2414 	if (!idev)
2415 		return ERR_PTR(-ENOBUFS);
2416 
2417 	if (idev->cnf.disable_ipv6)
2418 		return ERR_PTR(-EACCES);
2419 
2420 	/* Add default multicast route */
2421 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2422 		addrconf_add_mroute(dev);
2423 
2424 	return idev;
2425 }
2426 
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2427 static void manage_tempaddrs(struct inet6_dev *idev,
2428 			     struct inet6_ifaddr *ifp,
2429 			     __u32 valid_lft, __u32 prefered_lft,
2430 			     bool create, unsigned long now)
2431 {
2432 	u32 flags;
2433 	struct inet6_ifaddr *ift;
2434 
2435 	read_lock_bh(&idev->lock);
2436 	/* update all temporary addresses in the list */
2437 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2438 		int age, max_valid, max_prefered;
2439 
2440 		if (ifp != ift->ifpub)
2441 			continue;
2442 
2443 		/* RFC 4941 section 3.3:
2444 		 * If a received option will extend the lifetime of a public
2445 		 * address, the lifetimes of temporary addresses should
2446 		 * be extended, subject to the overall constraint that no
2447 		 * temporary addresses should ever remain "valid" or "preferred"
2448 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2449 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2450 		 */
2451 		age = (now - ift->cstamp) / HZ;
2452 		max_valid = idev->cnf.temp_valid_lft - age;
2453 		if (max_valid < 0)
2454 			max_valid = 0;
2455 
2456 		max_prefered = idev->cnf.temp_prefered_lft -
2457 			       idev->desync_factor - age;
2458 		if (max_prefered < 0)
2459 			max_prefered = 0;
2460 
2461 		if (valid_lft > max_valid)
2462 			valid_lft = max_valid;
2463 
2464 		if (prefered_lft > max_prefered)
2465 			prefered_lft = max_prefered;
2466 
2467 		spin_lock(&ift->lock);
2468 		flags = ift->flags;
2469 		ift->valid_lft = valid_lft;
2470 		ift->prefered_lft = prefered_lft;
2471 		ift->tstamp = now;
2472 		if (prefered_lft > 0)
2473 			ift->flags &= ~IFA_F_DEPRECATED;
2474 
2475 		spin_unlock(&ift->lock);
2476 		if (!(flags&IFA_F_TENTATIVE))
2477 			ipv6_ifa_notify(0, ift);
2478 	}
2479 
2480 	if ((create || list_empty(&idev->tempaddr_list)) &&
2481 	    idev->cnf.use_tempaddr > 0) {
2482 		/* When a new public address is created as described
2483 		 * in [ADDRCONF], also create a new temporary address.
2484 		 * Also create a temporary address if it's enabled but
2485 		 * no temporary address currently exists.
2486 		 */
2487 		read_unlock_bh(&idev->lock);
2488 		ipv6_create_tempaddr(ifp, NULL, false);
2489 	} else {
2490 		read_unlock_bh(&idev->lock);
2491 	}
2492 }
2493 
is_addr_mode_generate_stable(struct inet6_dev * idev)2494 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2495 {
2496 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2497 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2498 }
2499 
addrconf_prefix_rcv_add_addr(struct net * net,struct net_device * dev,const struct prefix_info * pinfo,struct inet6_dev * in6_dev,const struct in6_addr * addr,int addr_type,u32 addr_flags,bool sllao,bool tokenized,__u32 valid_lft,u32 prefered_lft)2500 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2501 				 const struct prefix_info *pinfo,
2502 				 struct inet6_dev *in6_dev,
2503 				 const struct in6_addr *addr, int addr_type,
2504 				 u32 addr_flags, bool sllao, bool tokenized,
2505 				 __u32 valid_lft, u32 prefered_lft)
2506 {
2507 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2508 	int create = 0, update_lft = 0;
2509 
2510 	if (!ifp && valid_lft) {
2511 		int max_addresses = in6_dev->cnf.max_addresses;
2512 		struct ifa6_config cfg = {
2513 			.pfx = addr,
2514 			.plen = pinfo->prefix_len,
2515 			.ifa_flags = addr_flags,
2516 			.valid_lft = valid_lft,
2517 			.preferred_lft = prefered_lft,
2518 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2519 		};
2520 
2521 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2522 		if ((net->ipv6.devconf_all->optimistic_dad ||
2523 		     in6_dev->cnf.optimistic_dad) &&
2524 		    !net->ipv6.devconf_all->forwarding && sllao)
2525 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2526 #endif
2527 
2528 		/* Do not allow to create too much of autoconfigured
2529 		 * addresses; this would be too easy way to crash kernel.
2530 		 */
2531 		if (!max_addresses ||
2532 		    ipv6_count_addresses(in6_dev) < max_addresses)
2533 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2534 
2535 		if (IS_ERR_OR_NULL(ifp))
2536 			return -1;
2537 
2538 		create = 1;
2539 		spin_lock_bh(&ifp->lock);
2540 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2541 		ifp->cstamp = jiffies;
2542 		ifp->tokenized = tokenized;
2543 		spin_unlock_bh(&ifp->lock);
2544 		addrconf_dad_start(ifp);
2545 	}
2546 
2547 	if (ifp) {
2548 		u32 flags;
2549 		unsigned long now;
2550 		u32 stored_lft;
2551 
2552 		/* update lifetime (RFC2462 5.5.3 e) */
2553 		spin_lock_bh(&ifp->lock);
2554 		now = jiffies;
2555 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2556 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2557 		else
2558 			stored_lft = 0;
2559 		if (!create && stored_lft) {
2560 			const u32 minimum_lft = min_t(u32,
2561 				stored_lft, MIN_VALID_LIFETIME);
2562 			valid_lft = max(valid_lft, minimum_lft);
2563 
2564 			/* RFC4862 Section 5.5.3e:
2565 			 * "Note that the preferred lifetime of the
2566 			 *  corresponding address is always reset to
2567 			 *  the Preferred Lifetime in the received
2568 			 *  Prefix Information option, regardless of
2569 			 *  whether the valid lifetime is also reset or
2570 			 *  ignored."
2571 			 *
2572 			 * So we should always update prefered_lft here.
2573 			 */
2574 			update_lft = 1;
2575 		}
2576 
2577 		if (update_lft) {
2578 			ifp->valid_lft = valid_lft;
2579 			ifp->prefered_lft = prefered_lft;
2580 			ifp->tstamp = now;
2581 			flags = ifp->flags;
2582 			ifp->flags &= ~IFA_F_DEPRECATED;
2583 			spin_unlock_bh(&ifp->lock);
2584 
2585 			if (!(flags&IFA_F_TENTATIVE))
2586 				ipv6_ifa_notify(0, ifp);
2587 		} else
2588 			spin_unlock_bh(&ifp->lock);
2589 
2590 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2591 				 create, now);
2592 
2593 		in6_ifa_put(ifp);
2594 		addrconf_verify();
2595 	}
2596 
2597 	return 0;
2598 }
2599 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2600 
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2601 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2602 {
2603 	struct prefix_info *pinfo;
2604 	__u32 valid_lft;
2605 	__u32 prefered_lft;
2606 	int addr_type, err;
2607 	u32 addr_flags = 0;
2608 	struct inet6_dev *in6_dev;
2609 	struct net *net = dev_net(dev);
2610 
2611 	pinfo = (struct prefix_info *) opt;
2612 
2613 	if (len < sizeof(struct prefix_info)) {
2614 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2615 		return;
2616 	}
2617 
2618 	/*
2619 	 *	Validation checks ([ADDRCONF], page 19)
2620 	 */
2621 
2622 	addr_type = ipv6_addr_type(&pinfo->prefix);
2623 
2624 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2625 		return;
2626 
2627 	valid_lft = ntohl(pinfo->valid);
2628 	prefered_lft = ntohl(pinfo->prefered);
2629 
2630 	if (prefered_lft > valid_lft) {
2631 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2632 		return;
2633 	}
2634 
2635 	in6_dev = in6_dev_get(dev);
2636 
2637 	if (!in6_dev) {
2638 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2639 				    dev->name);
2640 		return;
2641 	}
2642 
2643 	/*
2644 	 *	Two things going on here:
2645 	 *	1) Add routes for on-link prefixes
2646 	 *	2) Configure prefixes with the auto flag set
2647 	 */
2648 
2649 	if (pinfo->onlink) {
2650 		struct fib6_info *rt;
2651 		unsigned long rt_expires;
2652 
2653 		/* Avoid arithmetic overflow. Really, we could
2654 		 * save rt_expires in seconds, likely valid_lft,
2655 		 * but it would require division in fib gc, that it
2656 		 * not good.
2657 		 */
2658 		if (HZ > USER_HZ)
2659 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2660 		else
2661 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2662 
2663 		if (addrconf_finite_timeout(rt_expires))
2664 			rt_expires *= HZ;
2665 
2666 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2667 					       pinfo->prefix_len,
2668 					       dev,
2669 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2670 					       RTF_GATEWAY | RTF_DEFAULT);
2671 
2672 		if (rt) {
2673 			/* Autoconf prefix route */
2674 			if (valid_lft == 0) {
2675 				ip6_del_rt(net, rt);
2676 				rt = NULL;
2677 			} else if (addrconf_finite_timeout(rt_expires)) {
2678 				/* not infinity */
2679 				fib6_set_expires(rt, jiffies + rt_expires);
2680 			} else {
2681 				fib6_clean_expires(rt);
2682 			}
2683 		} else if (valid_lft) {
2684 			clock_t expires = 0;
2685 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2686 			if (addrconf_finite_timeout(rt_expires)) {
2687 				/* not infinity */
2688 				flags |= RTF_EXPIRES;
2689 				expires = jiffies_to_clock_t(rt_expires);
2690 			}
2691 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2692 					      0, dev, expires, flags,
2693 					      GFP_ATOMIC);
2694 		}
2695 		fib6_info_release(rt);
2696 	}
2697 
2698 	/* Try to figure out our local address for this prefix */
2699 
2700 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2701 		struct in6_addr addr;
2702 		bool tokenized = false, dev_addr_generated = false;
2703 
2704 		if (pinfo->prefix_len == 64) {
2705 			memcpy(&addr, &pinfo->prefix, 8);
2706 
2707 			if (!ipv6_addr_any(&in6_dev->token)) {
2708 				read_lock_bh(&in6_dev->lock);
2709 				memcpy(addr.s6_addr + 8,
2710 				       in6_dev->token.s6_addr + 8, 8);
2711 				read_unlock_bh(&in6_dev->lock);
2712 				tokenized = true;
2713 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2714 				   !ipv6_generate_stable_address(&addr, 0,
2715 								 in6_dev)) {
2716 				addr_flags |= IFA_F_STABLE_PRIVACY;
2717 				goto ok;
2718 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2719 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2720 				goto put;
2721 			} else {
2722 				dev_addr_generated = true;
2723 			}
2724 			goto ok;
2725 		}
2726 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2727 				    pinfo->prefix_len);
2728 		goto put;
2729 
2730 ok:
2731 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2732 						   &addr, addr_type,
2733 						   addr_flags, sllao,
2734 						   tokenized, valid_lft,
2735 						   prefered_lft);
2736 		if (err)
2737 			goto put;
2738 
2739 		/* Ignore error case here because previous prefix add addr was
2740 		 * successful which will be notified.
2741 		 */
2742 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2743 					      addr_type, addr_flags, sllao,
2744 					      tokenized, valid_lft,
2745 					      prefered_lft,
2746 					      dev_addr_generated);
2747 	}
2748 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2749 put:
2750 	in6_dev_put(in6_dev);
2751 }
2752 
2753 /*
2754  *	Set destination address.
2755  *	Special case for SIT interfaces where we create a new "virtual"
2756  *	device.
2757  */
addrconf_set_dstaddr(struct net * net,void __user * arg)2758 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2759 {
2760 	struct in6_ifreq ireq;
2761 	struct net_device *dev;
2762 	int err = -EINVAL;
2763 
2764 	rtnl_lock();
2765 
2766 	err = -EFAULT;
2767 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2768 		goto err_exit;
2769 
2770 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2771 
2772 	err = -ENODEV;
2773 	if (!dev)
2774 		goto err_exit;
2775 
2776 #if IS_ENABLED(CONFIG_IPV6_SIT)
2777 	if (dev->type == ARPHRD_SIT) {
2778 		const struct net_device_ops *ops = dev->netdev_ops;
2779 		struct ifreq ifr;
2780 		struct ip_tunnel_parm p;
2781 
2782 		err = -EADDRNOTAVAIL;
2783 		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2784 			goto err_exit;
2785 
2786 		memset(&p, 0, sizeof(p));
2787 		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2788 		p.iph.saddr = 0;
2789 		p.iph.version = 4;
2790 		p.iph.ihl = 5;
2791 		p.iph.protocol = IPPROTO_IPV6;
2792 		p.iph.ttl = 64;
2793 		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2794 
2795 		if (ops->ndo_do_ioctl) {
2796 			mm_segment_t oldfs = get_fs();
2797 
2798 			set_fs(KERNEL_DS);
2799 			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2800 			set_fs(oldfs);
2801 		} else
2802 			err = -EOPNOTSUPP;
2803 
2804 		if (err == 0) {
2805 			err = -ENOBUFS;
2806 			dev = __dev_get_by_name(net, p.name);
2807 			if (!dev)
2808 				goto err_exit;
2809 			err = dev_open(dev);
2810 		}
2811 	}
2812 #endif
2813 
2814 err_exit:
2815 	rtnl_unlock();
2816 	return err;
2817 }
2818 
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2819 static int ipv6_mc_config(struct sock *sk, bool join,
2820 			  const struct in6_addr *addr, int ifindex)
2821 {
2822 	int ret;
2823 
2824 	ASSERT_RTNL();
2825 
2826 	lock_sock(sk);
2827 	if (join)
2828 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2829 	else
2830 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2831 	release_sock(sk);
2832 
2833 	return ret;
2834 }
2835 
2836 /*
2837  *	Manual configuration of address on an interface
2838  */
inet6_addr_add(struct net * net,int ifindex,struct ifa6_config * cfg,struct netlink_ext_ack * extack)2839 static int inet6_addr_add(struct net *net, int ifindex,
2840 			  struct ifa6_config *cfg,
2841 			  struct netlink_ext_ack *extack)
2842 {
2843 	struct inet6_ifaddr *ifp;
2844 	struct inet6_dev *idev;
2845 	struct net_device *dev;
2846 	unsigned long timeout;
2847 	clock_t expires;
2848 	u32 flags;
2849 
2850 	ASSERT_RTNL();
2851 
2852 	if (cfg->plen > 128)
2853 		return -EINVAL;
2854 
2855 	/* check the lifetime */
2856 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2857 		return -EINVAL;
2858 
2859 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2860 		return -EINVAL;
2861 
2862 	dev = __dev_get_by_index(net, ifindex);
2863 	if (!dev)
2864 		return -ENODEV;
2865 
2866 	idev = addrconf_add_dev(dev);
2867 	if (IS_ERR(idev))
2868 		return PTR_ERR(idev);
2869 
2870 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2871 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2872 					 true, cfg->pfx, ifindex);
2873 
2874 		if (ret < 0)
2875 			return ret;
2876 	}
2877 
2878 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2879 
2880 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2881 	if (addrconf_finite_timeout(timeout)) {
2882 		expires = jiffies_to_clock_t(timeout * HZ);
2883 		cfg->valid_lft = timeout;
2884 		flags = RTF_EXPIRES;
2885 	} else {
2886 		expires = 0;
2887 		flags = 0;
2888 		cfg->ifa_flags |= IFA_F_PERMANENT;
2889 	}
2890 
2891 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2892 	if (addrconf_finite_timeout(timeout)) {
2893 		if (timeout == 0)
2894 			cfg->ifa_flags |= IFA_F_DEPRECATED;
2895 		cfg->preferred_lft = timeout;
2896 	}
2897 
2898 	ifp = ipv6_add_addr(idev, cfg, true, extack);
2899 	if (!IS_ERR(ifp)) {
2900 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2901 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2902 					      ifp->rt_priority, dev, expires,
2903 					      flags, GFP_KERNEL);
2904 		}
2905 
2906 		/* Send a netlink notification if DAD is enabled and
2907 		 * optimistic flag is not set
2908 		 */
2909 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2910 			ipv6_ifa_notify(0, ifp);
2911 		/*
2912 		 * Note that section 3.1 of RFC 4429 indicates
2913 		 * that the Optimistic flag should not be set for
2914 		 * manually configured addresses
2915 		 */
2916 		addrconf_dad_start(ifp);
2917 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2918 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
2919 					 cfg->preferred_lft, true, jiffies);
2920 		in6_ifa_put(ifp);
2921 		addrconf_verify_rtnl();
2922 		return 0;
2923 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2924 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2925 			       cfg->pfx, ifindex);
2926 	}
2927 
2928 	return PTR_ERR(ifp);
2929 }
2930 
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen)2931 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2932 			  const struct in6_addr *pfx, unsigned int plen)
2933 {
2934 	struct inet6_ifaddr *ifp;
2935 	struct inet6_dev *idev;
2936 	struct net_device *dev;
2937 
2938 	if (plen > 128)
2939 		return -EINVAL;
2940 
2941 	dev = __dev_get_by_index(net, ifindex);
2942 	if (!dev)
2943 		return -ENODEV;
2944 
2945 	idev = __in6_dev_get(dev);
2946 	if (!idev)
2947 		return -ENXIO;
2948 
2949 	read_lock_bh(&idev->lock);
2950 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2951 		if (ifp->prefix_len == plen &&
2952 		    ipv6_addr_equal(pfx, &ifp->addr)) {
2953 			in6_ifa_hold(ifp);
2954 			read_unlock_bh(&idev->lock);
2955 
2956 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
2957 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
2958 				manage_tempaddrs(idev, ifp, 0, 0, false,
2959 						 jiffies);
2960 			ipv6_del_addr(ifp);
2961 			addrconf_verify_rtnl();
2962 			if (ipv6_addr_is_multicast(pfx)) {
2963 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2964 					       false, pfx, dev->ifindex);
2965 			}
2966 			return 0;
2967 		}
2968 	}
2969 	read_unlock_bh(&idev->lock);
2970 	return -EADDRNOTAVAIL;
2971 }
2972 
2973 
addrconf_add_ifaddr(struct net * net,void __user * arg)2974 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2975 {
2976 	struct ifa6_config cfg = {
2977 		.ifa_flags = IFA_F_PERMANENT,
2978 		.preferred_lft = INFINITY_LIFE_TIME,
2979 		.valid_lft = INFINITY_LIFE_TIME,
2980 	};
2981 	struct in6_ifreq ireq;
2982 	int err;
2983 
2984 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2985 		return -EPERM;
2986 
2987 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2988 		return -EFAULT;
2989 
2990 	cfg.pfx = &ireq.ifr6_addr;
2991 	cfg.plen = ireq.ifr6_prefixlen;
2992 
2993 	rtnl_lock();
2994 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
2995 	rtnl_unlock();
2996 	return err;
2997 }
2998 
addrconf_del_ifaddr(struct net * net,void __user * arg)2999 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3000 {
3001 	struct in6_ifreq ireq;
3002 	int err;
3003 
3004 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3005 		return -EPERM;
3006 
3007 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3008 		return -EFAULT;
3009 
3010 	rtnl_lock();
3011 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3012 			     ireq.ifr6_prefixlen);
3013 	rtnl_unlock();
3014 	return err;
3015 }
3016 
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope)3017 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3018 		     int plen, int scope)
3019 {
3020 	struct inet6_ifaddr *ifp;
3021 	struct ifa6_config cfg = {
3022 		.pfx = addr,
3023 		.plen = plen,
3024 		.ifa_flags = IFA_F_PERMANENT,
3025 		.valid_lft = INFINITY_LIFE_TIME,
3026 		.preferred_lft = INFINITY_LIFE_TIME,
3027 		.scope = scope
3028 	};
3029 
3030 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3031 	if (!IS_ERR(ifp)) {
3032 		spin_lock_bh(&ifp->lock);
3033 		ifp->flags &= ~IFA_F_TENTATIVE;
3034 		spin_unlock_bh(&ifp->lock);
3035 		rt_genid_bump_ipv6(dev_net(idev->dev));
3036 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3037 		in6_ifa_put(ifp);
3038 	}
3039 }
3040 
3041 #if IS_ENABLED(CONFIG_IPV6_SIT)
sit_add_v4_addrs(struct inet6_dev * idev)3042 static void sit_add_v4_addrs(struct inet6_dev *idev)
3043 {
3044 	struct in6_addr addr;
3045 	struct net_device *dev;
3046 	struct net *net = dev_net(idev->dev);
3047 	int scope, plen;
3048 	u32 pflags = 0;
3049 
3050 	ASSERT_RTNL();
3051 
3052 	memset(&addr, 0, sizeof(struct in6_addr));
3053 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3054 
3055 	if (idev->dev->flags&IFF_POINTOPOINT) {
3056 		addr.s6_addr32[0] = htonl(0xfe800000);
3057 		scope = IFA_LINK;
3058 		plen = 64;
3059 	} else {
3060 		scope = IPV6_ADDR_COMPATv4;
3061 		plen = 96;
3062 		pflags |= RTF_NONEXTHOP;
3063 	}
3064 
3065 	if (addr.s6_addr32[3]) {
3066 		add_addr(idev, &addr, plen, scope);
3067 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3068 				      GFP_KERNEL);
3069 		return;
3070 	}
3071 
3072 	for_each_netdev(net, dev) {
3073 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3074 		if (in_dev && (dev->flags & IFF_UP)) {
3075 			struct in_ifaddr *ifa;
3076 
3077 			int flag = scope;
3078 
3079 			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3080 
3081 				addr.s6_addr32[3] = ifa->ifa_local;
3082 
3083 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3084 					continue;
3085 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3086 					if (idev->dev->flags&IFF_POINTOPOINT)
3087 						continue;
3088 					flag |= IFA_HOST;
3089 				}
3090 
3091 				add_addr(idev, &addr, plen, flag);
3092 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3093 						      0, pflags, GFP_KERNEL);
3094 			}
3095 		}
3096 	}
3097 }
3098 #endif
3099 
init_loopback(struct net_device * dev)3100 static void init_loopback(struct net_device *dev)
3101 {
3102 	struct inet6_dev  *idev;
3103 
3104 	/* ::1 */
3105 
3106 	ASSERT_RTNL();
3107 
3108 	idev = ipv6_find_idev(dev);
3109 	if (!idev) {
3110 		pr_debug("%s: add_dev failed\n", __func__);
3111 		return;
3112 	}
3113 
3114 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3115 }
3116 
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3117 void addrconf_add_linklocal(struct inet6_dev *idev,
3118 			    const struct in6_addr *addr, u32 flags)
3119 {
3120 	struct ifa6_config cfg = {
3121 		.pfx = addr,
3122 		.plen = 64,
3123 		.ifa_flags = flags | IFA_F_PERMANENT,
3124 		.valid_lft = INFINITY_LIFE_TIME,
3125 		.preferred_lft = INFINITY_LIFE_TIME,
3126 		.scope = IFA_LINK
3127 	};
3128 	struct inet6_ifaddr *ifp;
3129 
3130 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3131 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3132 	     idev->cnf.optimistic_dad) &&
3133 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3134 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3135 #endif
3136 
3137 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3138 	if (!IS_ERR(ifp)) {
3139 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3140 				      0, 0, GFP_ATOMIC);
3141 		addrconf_dad_start(ifp);
3142 		in6_ifa_put(ifp);
3143 	}
3144 }
3145 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3146 
ipv6_reserved_interfaceid(struct in6_addr address)3147 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3148 {
3149 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3150 		return true;
3151 
3152 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3153 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3154 		return true;
3155 
3156 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3157 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3158 		return true;
3159 
3160 	return false;
3161 }
3162 
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3163 static int ipv6_generate_stable_address(struct in6_addr *address,
3164 					u8 dad_count,
3165 					const struct inet6_dev *idev)
3166 {
3167 	static DEFINE_SPINLOCK(lock);
3168 	static __u32 digest[SHA_DIGEST_WORDS];
3169 	static __u32 workspace[SHA_WORKSPACE_WORDS];
3170 
3171 	static union {
3172 		char __data[SHA_MESSAGE_BYTES];
3173 		struct {
3174 			struct in6_addr secret;
3175 			__be32 prefix[2];
3176 			unsigned char hwaddr[MAX_ADDR_LEN];
3177 			u8 dad_count;
3178 		} __packed;
3179 	} data;
3180 
3181 	struct in6_addr secret;
3182 	struct in6_addr temp;
3183 	struct net *net = dev_net(idev->dev);
3184 
3185 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3186 
3187 	if (idev->cnf.stable_secret.initialized)
3188 		secret = idev->cnf.stable_secret.secret;
3189 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3190 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3191 	else
3192 		return -1;
3193 
3194 retry:
3195 	spin_lock_bh(&lock);
3196 
3197 	sha_init(digest);
3198 	memset(&data, 0, sizeof(data));
3199 	memset(workspace, 0, sizeof(workspace));
3200 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3201 	data.prefix[0] = address->s6_addr32[0];
3202 	data.prefix[1] = address->s6_addr32[1];
3203 	data.secret = secret;
3204 	data.dad_count = dad_count;
3205 
3206 	sha_transform(digest, data.__data, workspace);
3207 
3208 	temp = *address;
3209 	temp.s6_addr32[2] = (__force __be32)digest[0];
3210 	temp.s6_addr32[3] = (__force __be32)digest[1];
3211 
3212 	spin_unlock_bh(&lock);
3213 
3214 	if (ipv6_reserved_interfaceid(temp)) {
3215 		dad_count++;
3216 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3217 			return -1;
3218 		goto retry;
3219 	}
3220 
3221 	*address = temp;
3222 	return 0;
3223 }
3224 
ipv6_gen_mode_random_init(struct inet6_dev * idev)3225 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3226 {
3227 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3228 
3229 	if (s->initialized)
3230 		return;
3231 	s = &idev->cnf.stable_secret;
3232 	get_random_bytes(&s->secret, sizeof(s->secret));
3233 	s->initialized = true;
3234 }
3235 
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3236 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3237 {
3238 	struct in6_addr addr;
3239 
3240 	/* no link local addresses on L3 master devices */
3241 	if (netif_is_l3_master(idev->dev))
3242 		return;
3243 
3244 	/* no link local addresses on devices flagged as slaves */
3245 	if (idev->dev->flags & IFF_SLAVE)
3246 		return;
3247 
3248 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3249 
3250 	switch (idev->cnf.addr_gen_mode) {
3251 	case IN6_ADDR_GEN_MODE_RANDOM:
3252 		ipv6_gen_mode_random_init(idev);
3253 		/* fallthrough */
3254 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3255 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3256 			addrconf_add_linklocal(idev, &addr,
3257 					       IFA_F_STABLE_PRIVACY);
3258 		else if (prefix_route)
3259 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3260 					      0, 0, GFP_KERNEL);
3261 		break;
3262 	case IN6_ADDR_GEN_MODE_EUI64:
3263 		/* addrconf_add_linklocal also adds a prefix_route and we
3264 		 * only need to care about prefix routes if ipv6_generate_eui64
3265 		 * couldn't generate one.
3266 		 */
3267 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3268 			addrconf_add_linklocal(idev, &addr, 0);
3269 		else if (prefix_route)
3270 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3271 					      0, 0, GFP_KERNEL);
3272 		break;
3273 	case IN6_ADDR_GEN_MODE_NONE:
3274 	default:
3275 		/* will not add any link local address */
3276 		break;
3277 	}
3278 }
3279 
addrconf_dev_config(struct net_device * dev)3280 static void addrconf_dev_config(struct net_device *dev)
3281 {
3282 	struct inet6_dev *idev;
3283 
3284 	ASSERT_RTNL();
3285 
3286 	if ((dev->type != ARPHRD_ETHER) &&
3287 	    (dev->type != ARPHRD_FDDI) &&
3288 	    (dev->type != ARPHRD_ARCNET) &&
3289 	    (dev->type != ARPHRD_INFINIBAND) &&
3290 	    (dev->type != ARPHRD_IEEE1394) &&
3291 	    (dev->type != ARPHRD_TUNNEL6) &&
3292 	    (dev->type != ARPHRD_6LOWPAN) &&
3293 	    (dev->type != ARPHRD_IP6GRE) &&
3294 	    (dev->type != ARPHRD_IPGRE) &&
3295 	    (dev->type != ARPHRD_TUNNEL) &&
3296 	    (dev->type != ARPHRD_NONE) &&
3297 	    (dev->type != ARPHRD_RAWIP)) {
3298 		/* Alas, we support only Ethernet autoconfiguration. */
3299 		idev = __in6_dev_get(dev);
3300 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3301 		    dev->flags & IFF_MULTICAST)
3302 			ipv6_mc_up(idev);
3303 		return;
3304 	}
3305 
3306 	idev = addrconf_add_dev(dev);
3307 	if (IS_ERR(idev))
3308 		return;
3309 
3310 	/* this device type has no EUI support */
3311 	if (dev->type == ARPHRD_NONE &&
3312 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3313 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3314 
3315 	addrconf_addr_gen(idev, false);
3316 }
3317 
3318 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3319 static void addrconf_sit_config(struct net_device *dev)
3320 {
3321 	struct inet6_dev *idev;
3322 
3323 	ASSERT_RTNL();
3324 
3325 	/*
3326 	 * Configure the tunnel with one of our IPv4
3327 	 * addresses... we should configure all of
3328 	 * our v4 addrs in the tunnel
3329 	 */
3330 
3331 	idev = ipv6_find_idev(dev);
3332 	if (!idev) {
3333 		pr_debug("%s: add_dev failed\n", __func__);
3334 		return;
3335 	}
3336 
3337 	if (dev->priv_flags & IFF_ISATAP) {
3338 		addrconf_addr_gen(idev, false);
3339 		return;
3340 	}
3341 
3342 	sit_add_v4_addrs(idev);
3343 
3344 	if (dev->flags&IFF_POINTOPOINT)
3345 		addrconf_add_mroute(dev);
3346 }
3347 #endif
3348 
3349 #if IS_ENABLED(CONFIG_NET_IPGRE)
addrconf_gre_config(struct net_device * dev)3350 static void addrconf_gre_config(struct net_device *dev)
3351 {
3352 	struct inet6_dev *idev;
3353 
3354 	ASSERT_RTNL();
3355 
3356 	idev = ipv6_find_idev(dev);
3357 	if (!idev) {
3358 		pr_debug("%s: add_dev failed\n", __func__);
3359 		return;
3360 	}
3361 
3362 	addrconf_addr_gen(idev, true);
3363 	if (dev->flags & IFF_POINTOPOINT)
3364 		addrconf_add_mroute(dev);
3365 }
3366 #endif
3367 
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3368 static int fixup_permanent_addr(struct net *net,
3369 				struct inet6_dev *idev,
3370 				struct inet6_ifaddr *ifp)
3371 {
3372 	/* !fib6_node means the host route was removed from the
3373 	 * FIB, for example, if 'lo' device is taken down. In that
3374 	 * case regenerate the host route.
3375 	 */
3376 	if (!ifp->rt || !ifp->rt->fib6_node) {
3377 		struct fib6_info *f6i, *prev;
3378 
3379 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3380 					 GFP_ATOMIC);
3381 		if (IS_ERR(f6i))
3382 			return PTR_ERR(f6i);
3383 
3384 		/* ifp->rt can be accessed outside of rtnl */
3385 		spin_lock(&ifp->lock);
3386 		prev = ifp->rt;
3387 		ifp->rt = f6i;
3388 		spin_unlock(&ifp->lock);
3389 
3390 		fib6_info_release(prev);
3391 	}
3392 
3393 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3394 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3395 				      ifp->rt_priority, idev->dev, 0, 0,
3396 				      GFP_ATOMIC);
3397 	}
3398 
3399 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3400 		addrconf_dad_start(ifp);
3401 
3402 	return 0;
3403 }
3404 
addrconf_permanent_addr(struct net * net,struct net_device * dev)3405 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3406 {
3407 	struct inet6_ifaddr *ifp, *tmp;
3408 	struct inet6_dev *idev;
3409 
3410 	idev = __in6_dev_get(dev);
3411 	if (!idev)
3412 		return;
3413 
3414 	write_lock_bh(&idev->lock);
3415 
3416 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3417 		if ((ifp->flags & IFA_F_PERMANENT) &&
3418 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3419 			write_unlock_bh(&idev->lock);
3420 			in6_ifa_hold(ifp);
3421 			ipv6_del_addr(ifp);
3422 			write_lock_bh(&idev->lock);
3423 
3424 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3425 					     idev->dev->name, &ifp->addr);
3426 		}
3427 	}
3428 
3429 	write_unlock_bh(&idev->lock);
3430 }
3431 
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3432 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3433 			   void *ptr)
3434 {
3435 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3436 	struct netdev_notifier_change_info *change_info;
3437 	struct netdev_notifier_changeupper_info *info;
3438 	struct inet6_dev *idev = __in6_dev_get(dev);
3439 	struct net *net = dev_net(dev);
3440 	int run_pending = 0;
3441 	int err;
3442 
3443 	switch (event) {
3444 	case NETDEV_REGISTER:
3445 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3446 			idev = ipv6_add_dev(dev);
3447 			if (IS_ERR(idev))
3448 				return notifier_from_errno(PTR_ERR(idev));
3449 		}
3450 		break;
3451 
3452 	case NETDEV_CHANGEMTU:
3453 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3454 		if (dev->mtu < IPV6_MIN_MTU) {
3455 			addrconf_ifdown(dev, dev != net->loopback_dev);
3456 			break;
3457 		}
3458 
3459 		if (idev) {
3460 			rt6_mtu_change(dev, dev->mtu);
3461 			idev->cnf.mtu6 = dev->mtu;
3462 			break;
3463 		}
3464 
3465 		/* allocate new idev */
3466 		idev = ipv6_add_dev(dev);
3467 		if (IS_ERR(idev))
3468 			break;
3469 
3470 		/* device is still not ready */
3471 		if (!(idev->if_flags & IF_READY))
3472 			break;
3473 
3474 		run_pending = 1;
3475 
3476 		/* fall through */
3477 
3478 	case NETDEV_UP:
3479 	case NETDEV_CHANGE:
3480 		if (dev->flags & IFF_SLAVE)
3481 			break;
3482 
3483 		if (idev && idev->cnf.disable_ipv6)
3484 			break;
3485 
3486 		if (event == NETDEV_UP) {
3487 			/* restore routes for permanent addresses */
3488 			addrconf_permanent_addr(net, dev);
3489 
3490 			if (!addrconf_link_ready(dev)) {
3491 				/* device is not ready yet. */
3492 				pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3493 					dev->name);
3494 				break;
3495 			}
3496 
3497 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3498 				idev = ipv6_add_dev(dev);
3499 
3500 			if (!IS_ERR_OR_NULL(idev)) {
3501 				idev->if_flags |= IF_READY;
3502 				run_pending = 1;
3503 			}
3504 		} else if (event == NETDEV_CHANGE) {
3505 			if (!addrconf_link_ready(dev)) {
3506 				/* device is still not ready. */
3507 				rt6_sync_down_dev(dev, event);
3508 				break;
3509 			}
3510 
3511 			if (!IS_ERR_OR_NULL(idev)) {
3512 				if (idev->if_flags & IF_READY) {
3513 					/* device is already configured -
3514 					 * but resend MLD reports, we might
3515 					 * have roamed and need to update
3516 					 * multicast snooping switches
3517 					 */
3518 					ipv6_mc_up(idev);
3519 					change_info = ptr;
3520 					if (change_info->flags_changed & IFF_NOARP)
3521 						addrconf_dad_run(idev, true);
3522 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3523 					break;
3524 				}
3525 				idev->if_flags |= IF_READY;
3526 			}
3527 
3528 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3529 				dev->name);
3530 
3531 			run_pending = 1;
3532 		}
3533 
3534 		switch (dev->type) {
3535 #if IS_ENABLED(CONFIG_IPV6_SIT)
3536 		case ARPHRD_SIT:
3537 			addrconf_sit_config(dev);
3538 			break;
3539 #endif
3540 #if IS_ENABLED(CONFIG_NET_IPGRE)
3541 		case ARPHRD_IPGRE:
3542 			addrconf_gre_config(dev);
3543 			break;
3544 #endif
3545 		case ARPHRD_LOOPBACK:
3546 			init_loopback(dev);
3547 			break;
3548 
3549 		default:
3550 			addrconf_dev_config(dev);
3551 			break;
3552 		}
3553 
3554 		if (!IS_ERR_OR_NULL(idev)) {
3555 			if (run_pending)
3556 				addrconf_dad_run(idev, false);
3557 
3558 			/* Device has an address by now */
3559 			rt6_sync_up(dev, RTNH_F_DEAD);
3560 
3561 			/*
3562 			 * If the MTU changed during the interface down,
3563 			 * when the interface up, the changed MTU must be
3564 			 * reflected in the idev as well as routers.
3565 			 */
3566 			if (idev->cnf.mtu6 != dev->mtu &&
3567 			    dev->mtu >= IPV6_MIN_MTU) {
3568 				rt6_mtu_change(dev, dev->mtu);
3569 				idev->cnf.mtu6 = dev->mtu;
3570 			}
3571 			idev->tstamp = jiffies;
3572 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3573 
3574 			/*
3575 			 * If the changed mtu during down is lower than
3576 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3577 			 */
3578 			if (dev->mtu < IPV6_MIN_MTU)
3579 				addrconf_ifdown(dev, dev != net->loopback_dev);
3580 		}
3581 		break;
3582 
3583 	case NETDEV_DOWN:
3584 	case NETDEV_UNREGISTER:
3585 		/*
3586 		 *	Remove all addresses from this interface.
3587 		 */
3588 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3589 		break;
3590 
3591 	case NETDEV_CHANGENAME:
3592 		if (idev) {
3593 			snmp6_unregister_dev(idev);
3594 			addrconf_sysctl_unregister(idev);
3595 			err = addrconf_sysctl_register(idev);
3596 			if (err)
3597 				return notifier_from_errno(err);
3598 			err = snmp6_register_dev(idev);
3599 			if (err) {
3600 				addrconf_sysctl_unregister(idev);
3601 				return notifier_from_errno(err);
3602 			}
3603 		}
3604 		break;
3605 
3606 	case NETDEV_PRE_TYPE_CHANGE:
3607 	case NETDEV_POST_TYPE_CHANGE:
3608 		if (idev)
3609 			addrconf_type_change(dev, event);
3610 		break;
3611 
3612 	case NETDEV_CHANGEUPPER:
3613 		info = ptr;
3614 
3615 		/* flush all routes if dev is linked to or unlinked from
3616 		 * an L3 master device (e.g., VRF)
3617 		 */
3618 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3619 			addrconf_ifdown(dev, 0);
3620 	}
3621 
3622 	return NOTIFY_OK;
3623 }
3624 
3625 /*
3626  *	addrconf module should be notified of a device going up
3627  */
3628 static struct notifier_block ipv6_dev_notf = {
3629 	.notifier_call = addrconf_notify,
3630 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3631 };
3632 
addrconf_type_change(struct net_device * dev,unsigned long event)3633 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3634 {
3635 	struct inet6_dev *idev;
3636 	ASSERT_RTNL();
3637 
3638 	idev = __in6_dev_get(dev);
3639 
3640 	if (event == NETDEV_POST_TYPE_CHANGE)
3641 		ipv6_mc_remap(idev);
3642 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3643 		ipv6_mc_unmap(idev);
3644 }
3645 
addr_is_local(const struct in6_addr * addr)3646 static bool addr_is_local(const struct in6_addr *addr)
3647 {
3648 	return ipv6_addr_type(addr) &
3649 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3650 }
3651 
addrconf_ifdown(struct net_device * dev,int how)3652 static int addrconf_ifdown(struct net_device *dev, int how)
3653 {
3654 	unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3655 	struct net *net = dev_net(dev);
3656 	struct inet6_dev *idev;
3657 	struct inet6_ifaddr *ifa, *tmp;
3658 	bool keep_addr = false;
3659 	int state, i;
3660 
3661 	ASSERT_RTNL();
3662 
3663 	rt6_disable_ip(dev, event);
3664 
3665 	idev = __in6_dev_get(dev);
3666 	if (!idev)
3667 		return -ENODEV;
3668 
3669 	/*
3670 	 * Step 1: remove reference to ipv6 device from parent device.
3671 	 *	   Do not dev_put!
3672 	 */
3673 	if (how) {
3674 		idev->dead = 1;
3675 
3676 		/* protected by rtnl_lock */
3677 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3678 
3679 		/* Step 1.5: remove snmp6 entry */
3680 		snmp6_unregister_dev(idev);
3681 
3682 	}
3683 
3684 	/* combine the user config with event to determine if permanent
3685 	 * addresses are to be removed from address hash table
3686 	 */
3687 	if (!how && !idev->cnf.disable_ipv6) {
3688 		/* aggregate the system setting and interface setting */
3689 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3690 
3691 		if (!_keep_addr)
3692 			_keep_addr = idev->cnf.keep_addr_on_down;
3693 
3694 		keep_addr = (_keep_addr > 0);
3695 	}
3696 
3697 	/* Step 2: clear hash table */
3698 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3699 		struct hlist_head *h = &inet6_addr_lst[i];
3700 
3701 		spin_lock_bh(&addrconf_hash_lock);
3702 restart:
3703 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3704 			if (ifa->idev == idev) {
3705 				addrconf_del_dad_work(ifa);
3706 				/* combined flag + permanent flag decide if
3707 				 * address is retained on a down event
3708 				 */
3709 				if (!keep_addr ||
3710 				    !(ifa->flags & IFA_F_PERMANENT) ||
3711 				    addr_is_local(&ifa->addr)) {
3712 					hlist_del_init_rcu(&ifa->addr_lst);
3713 					goto restart;
3714 				}
3715 			}
3716 		}
3717 		spin_unlock_bh(&addrconf_hash_lock);
3718 	}
3719 
3720 	write_lock_bh(&idev->lock);
3721 
3722 	addrconf_del_rs_timer(idev);
3723 
3724 	/* Step 2: clear flags for stateless addrconf */
3725 	if (!how)
3726 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3727 
3728 	/* Step 3: clear tempaddr list */
3729 	while (!list_empty(&idev->tempaddr_list)) {
3730 		ifa = list_first_entry(&idev->tempaddr_list,
3731 				       struct inet6_ifaddr, tmp_list);
3732 		list_del(&ifa->tmp_list);
3733 		write_unlock_bh(&idev->lock);
3734 		spin_lock_bh(&ifa->lock);
3735 
3736 		if (ifa->ifpub) {
3737 			in6_ifa_put(ifa->ifpub);
3738 			ifa->ifpub = NULL;
3739 		}
3740 		spin_unlock_bh(&ifa->lock);
3741 		in6_ifa_put(ifa);
3742 		write_lock_bh(&idev->lock);
3743 	}
3744 
3745 	list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3746 		struct fib6_info *rt = NULL;
3747 		bool keep;
3748 
3749 		addrconf_del_dad_work(ifa);
3750 
3751 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3752 			!addr_is_local(&ifa->addr);
3753 
3754 		write_unlock_bh(&idev->lock);
3755 		spin_lock_bh(&ifa->lock);
3756 
3757 		if (keep) {
3758 			/* set state to skip the notifier below */
3759 			state = INET6_IFADDR_STATE_DEAD;
3760 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3761 			if (!(ifa->flags & IFA_F_NODAD))
3762 				ifa->flags |= IFA_F_TENTATIVE;
3763 
3764 			rt = ifa->rt;
3765 			ifa->rt = NULL;
3766 		} else {
3767 			state = ifa->state;
3768 			ifa->state = INET6_IFADDR_STATE_DEAD;
3769 		}
3770 
3771 		spin_unlock_bh(&ifa->lock);
3772 
3773 		if (rt)
3774 			ip6_del_rt(net, rt);
3775 
3776 		if (state != INET6_IFADDR_STATE_DEAD) {
3777 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3778 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3779 		} else {
3780 			if (idev->cnf.forwarding)
3781 				addrconf_leave_anycast(ifa);
3782 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3783 		}
3784 
3785 		write_lock_bh(&idev->lock);
3786 		if (!keep) {
3787 			list_del_rcu(&ifa->if_list);
3788 			in6_ifa_put(ifa);
3789 		}
3790 	}
3791 
3792 	write_unlock_bh(&idev->lock);
3793 
3794 	/* Step 5: Discard anycast and multicast list */
3795 	if (how) {
3796 		ipv6_ac_destroy_dev(idev);
3797 		ipv6_mc_destroy_dev(idev);
3798 	} else {
3799 		ipv6_mc_down(idev);
3800 	}
3801 
3802 	idev->tstamp = jiffies;
3803 
3804 	/* Last: Shot the device (if unregistered) */
3805 	if (how) {
3806 		addrconf_sysctl_unregister(idev);
3807 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3808 		neigh_ifdown(&nd_tbl, dev);
3809 		in6_dev_put(idev);
3810 	}
3811 	return 0;
3812 }
3813 
addrconf_rs_timer(struct timer_list * t)3814 static void addrconf_rs_timer(struct timer_list *t)
3815 {
3816 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3817 	struct net_device *dev = idev->dev;
3818 	struct in6_addr lladdr;
3819 
3820 	write_lock(&idev->lock);
3821 	if (idev->dead || !(idev->if_flags & IF_READY))
3822 		goto out;
3823 
3824 	if (!ipv6_accept_ra(idev))
3825 		goto out;
3826 
3827 	/* Announcement received after solicitation was sent */
3828 	if (idev->if_flags & IF_RA_RCVD)
3829 		goto out;
3830 
3831 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3832 		write_unlock(&idev->lock);
3833 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3834 			ndisc_send_rs(dev, &lladdr,
3835 				      &in6addr_linklocal_allrouters);
3836 		else
3837 			goto put;
3838 
3839 		write_lock(&idev->lock);
3840 		idev->rs_interval = rfc3315_s14_backoff_update(
3841 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3842 		/* The wait after the last probe can be shorter */
3843 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3844 					     idev->cnf.rtr_solicits) ?
3845 				      idev->cnf.rtr_solicit_delay :
3846 				      idev->rs_interval);
3847 	} else {
3848 		/*
3849 		 * Note: we do not support deprecated "all on-link"
3850 		 * assumption any longer.
3851 		 */
3852 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3853 	}
3854 
3855 out:
3856 	write_unlock(&idev->lock);
3857 put:
3858 	in6_dev_put(idev);
3859 }
3860 
3861 /*
3862  *	Duplicate Address Detection
3863  */
addrconf_dad_kick(struct inet6_ifaddr * ifp)3864 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3865 {
3866 	unsigned long rand_num;
3867 	struct inet6_dev *idev = ifp->idev;
3868 	u64 nonce;
3869 
3870 	if (ifp->flags & IFA_F_OPTIMISTIC)
3871 		rand_num = 0;
3872 	else
3873 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3874 
3875 	nonce = 0;
3876 	if (idev->cnf.enhanced_dad ||
3877 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3878 		do
3879 			get_random_bytes(&nonce, 6);
3880 		while (nonce == 0);
3881 	}
3882 	ifp->dad_nonce = nonce;
3883 	ifp->dad_probes = idev->cnf.dad_transmits;
3884 	addrconf_mod_dad_work(ifp, rand_num);
3885 }
3886 
addrconf_dad_begin(struct inet6_ifaddr * ifp)3887 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3888 {
3889 	struct inet6_dev *idev = ifp->idev;
3890 	struct net_device *dev = idev->dev;
3891 	bool bump_id, notify = false;
3892 	struct net *net;
3893 
3894 	addrconf_join_solict(dev, &ifp->addr);
3895 
3896 	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3897 
3898 	read_lock_bh(&idev->lock);
3899 	spin_lock(&ifp->lock);
3900 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3901 		goto out;
3902 
3903 	net = dev_net(dev);
3904 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3905 	    (net->ipv6.devconf_all->accept_dad < 1 &&
3906 	     idev->cnf.accept_dad < 1) ||
3907 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3908 	    ifp->flags & IFA_F_NODAD) {
3909 		bool send_na = false;
3910 
3911 		if (ifp->flags & IFA_F_TENTATIVE &&
3912 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3913 			send_na = true;
3914 		bump_id = ifp->flags & IFA_F_TENTATIVE;
3915 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3916 		spin_unlock(&ifp->lock);
3917 		read_unlock_bh(&idev->lock);
3918 
3919 		addrconf_dad_completed(ifp, bump_id, send_na);
3920 		return;
3921 	}
3922 
3923 	if (!(idev->if_flags & IF_READY)) {
3924 		spin_unlock(&ifp->lock);
3925 		read_unlock_bh(&idev->lock);
3926 		/*
3927 		 * If the device is not ready:
3928 		 * - keep it tentative if it is a permanent address.
3929 		 * - otherwise, kill it.
3930 		 */
3931 		in6_ifa_hold(ifp);
3932 		addrconf_dad_stop(ifp, 0);
3933 		return;
3934 	}
3935 
3936 	/*
3937 	 * Optimistic nodes can start receiving
3938 	 * Frames right away
3939 	 */
3940 	if (ifp->flags & IFA_F_OPTIMISTIC) {
3941 		ip6_ins_rt(net, ifp->rt);
3942 		if (ipv6_use_optimistic_addr(net, idev)) {
3943 			/* Because optimistic nodes can use this address,
3944 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3945 			 */
3946 			notify = true;
3947 		}
3948 	}
3949 
3950 	addrconf_dad_kick(ifp);
3951 out:
3952 	spin_unlock(&ifp->lock);
3953 	read_unlock_bh(&idev->lock);
3954 	if (notify)
3955 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3956 }
3957 
addrconf_dad_start(struct inet6_ifaddr * ifp)3958 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3959 {
3960 	bool begin_dad = false;
3961 
3962 	spin_lock_bh(&ifp->lock);
3963 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3964 		ifp->state = INET6_IFADDR_STATE_PREDAD;
3965 		begin_dad = true;
3966 	}
3967 	spin_unlock_bh(&ifp->lock);
3968 
3969 	if (begin_dad)
3970 		addrconf_mod_dad_work(ifp, 0);
3971 }
3972 
addrconf_dad_work(struct work_struct * w)3973 static void addrconf_dad_work(struct work_struct *w)
3974 {
3975 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3976 						struct inet6_ifaddr,
3977 						dad_work);
3978 	struct inet6_dev *idev = ifp->idev;
3979 	bool bump_id, disable_ipv6 = false;
3980 	struct in6_addr mcaddr;
3981 
3982 	enum {
3983 		DAD_PROCESS,
3984 		DAD_BEGIN,
3985 		DAD_ABORT,
3986 	} action = DAD_PROCESS;
3987 
3988 	rtnl_lock();
3989 
3990 	spin_lock_bh(&ifp->lock);
3991 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3992 		action = DAD_BEGIN;
3993 		ifp->state = INET6_IFADDR_STATE_DAD;
3994 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3995 		action = DAD_ABORT;
3996 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
3997 
3998 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3999 		     idev->cnf.accept_dad > 1) &&
4000 		    !idev->cnf.disable_ipv6 &&
4001 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4002 			struct in6_addr addr;
4003 
4004 			addr.s6_addr32[0] = htonl(0xfe800000);
4005 			addr.s6_addr32[1] = 0;
4006 
4007 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4008 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4009 				/* DAD failed for link-local based on MAC */
4010 				idev->cnf.disable_ipv6 = 1;
4011 
4012 				pr_info("%s: IPv6 being disabled!\n",
4013 					ifp->idev->dev->name);
4014 				disable_ipv6 = true;
4015 			}
4016 		}
4017 	}
4018 	spin_unlock_bh(&ifp->lock);
4019 
4020 	if (action == DAD_BEGIN) {
4021 		addrconf_dad_begin(ifp);
4022 		goto out;
4023 	} else if (action == DAD_ABORT) {
4024 		in6_ifa_hold(ifp);
4025 		addrconf_dad_stop(ifp, 1);
4026 		if (disable_ipv6)
4027 			addrconf_ifdown(idev->dev, 0);
4028 		goto out;
4029 	}
4030 
4031 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4032 		goto out;
4033 
4034 	write_lock_bh(&idev->lock);
4035 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4036 		write_unlock_bh(&idev->lock);
4037 		goto out;
4038 	}
4039 
4040 	spin_lock(&ifp->lock);
4041 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4042 		spin_unlock(&ifp->lock);
4043 		write_unlock_bh(&idev->lock);
4044 		goto out;
4045 	}
4046 
4047 	if (ifp->dad_probes == 0) {
4048 		bool send_na = false;
4049 
4050 		/*
4051 		 * DAD was successful
4052 		 */
4053 
4054 		if (ifp->flags & IFA_F_TENTATIVE &&
4055 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4056 			send_na = true;
4057 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4058 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4059 		spin_unlock(&ifp->lock);
4060 		write_unlock_bh(&idev->lock);
4061 
4062 		addrconf_dad_completed(ifp, bump_id, send_na);
4063 
4064 		goto out;
4065 	}
4066 
4067 	ifp->dad_probes--;
4068 	addrconf_mod_dad_work(ifp,
4069 			      NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4070 	spin_unlock(&ifp->lock);
4071 	write_unlock_bh(&idev->lock);
4072 
4073 	/* send a neighbour solicitation for our addr */
4074 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4075 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4076 		      ifp->dad_nonce);
4077 out:
4078 	in6_ifa_put(ifp);
4079 	rtnl_unlock();
4080 }
4081 
4082 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4083 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4084 {
4085 	struct inet6_ifaddr *ifpiter;
4086 	struct inet6_dev *idev = ifp->idev;
4087 
4088 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4089 		if (ifpiter->scope > IFA_LINK)
4090 			break;
4091 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4092 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4093 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4094 		    IFA_F_PERMANENT)
4095 			return false;
4096 	}
4097 	return true;
4098 }
4099 
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4100 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4101 				   bool send_na)
4102 {
4103 	struct net_device *dev = ifp->idev->dev;
4104 	struct in6_addr lladdr;
4105 	bool send_rs, send_mld;
4106 
4107 	addrconf_del_dad_work(ifp);
4108 
4109 	/*
4110 	 *	Configure the address for reception. Now it is valid.
4111 	 */
4112 
4113 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4114 
4115 	/* If added prefix is link local and we are prepared to process
4116 	   router advertisements, start sending router solicitations.
4117 	 */
4118 
4119 	read_lock_bh(&ifp->idev->lock);
4120 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4121 	send_rs = send_mld &&
4122 		  ipv6_accept_ra(ifp->idev) &&
4123 		  ifp->idev->cnf.rtr_solicits != 0 &&
4124 		  (dev->flags&IFF_LOOPBACK) == 0;
4125 	read_unlock_bh(&ifp->idev->lock);
4126 
4127 	/* While dad is in progress mld report's source address is in6_addrany.
4128 	 * Resend with proper ll now.
4129 	 */
4130 	if (send_mld)
4131 		ipv6_mc_dad_complete(ifp->idev);
4132 
4133 	/* send unsolicited NA if enabled */
4134 	if (send_na &&
4135 	    (ifp->idev->cnf.ndisc_notify ||
4136 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4137 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4138 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4139 			      /*solicited=*/ false, /*override=*/ true,
4140 			      /*inc_opt=*/ true);
4141 	}
4142 
4143 	if (send_rs) {
4144 		/*
4145 		 *	If a host as already performed a random delay
4146 		 *	[...] as part of DAD [...] there is no need
4147 		 *	to delay again before sending the first RS
4148 		 */
4149 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4150 			return;
4151 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4152 
4153 		write_lock_bh(&ifp->idev->lock);
4154 		spin_lock(&ifp->lock);
4155 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4156 			ifp->idev->cnf.rtr_solicit_interval);
4157 		ifp->idev->rs_probes = 1;
4158 		ifp->idev->if_flags |= IF_RS_SENT;
4159 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4160 		spin_unlock(&ifp->lock);
4161 		write_unlock_bh(&ifp->idev->lock);
4162 	}
4163 
4164 	if (bump_id)
4165 		rt_genid_bump_ipv6(dev_net(dev));
4166 
4167 	/* Make sure that a new temporary address will be created
4168 	 * before this temporary address becomes deprecated.
4169 	 */
4170 	if (ifp->flags & IFA_F_TEMPORARY)
4171 		addrconf_verify_rtnl();
4172 }
4173 
addrconf_dad_run(struct inet6_dev * idev,bool restart)4174 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4175 {
4176 	struct inet6_ifaddr *ifp;
4177 
4178 	read_lock_bh(&idev->lock);
4179 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4180 		spin_lock(&ifp->lock);
4181 		if ((ifp->flags & IFA_F_TENTATIVE &&
4182 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4183 			if (restart)
4184 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4185 			addrconf_dad_kick(ifp);
4186 		}
4187 		spin_unlock(&ifp->lock);
4188 	}
4189 	read_unlock_bh(&idev->lock);
4190 }
4191 
4192 #ifdef CONFIG_PROC_FS
4193 struct if6_iter_state {
4194 	struct seq_net_private p;
4195 	int bucket;
4196 	int offset;
4197 };
4198 
if6_get_first(struct seq_file * seq,loff_t pos)4199 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4200 {
4201 	struct if6_iter_state *state = seq->private;
4202 	struct net *net = seq_file_net(seq);
4203 	struct inet6_ifaddr *ifa = NULL;
4204 	int p = 0;
4205 
4206 	/* initial bucket if pos is 0 */
4207 	if (pos == 0) {
4208 		state->bucket = 0;
4209 		state->offset = 0;
4210 	}
4211 
4212 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4213 		hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4214 					 addr_lst) {
4215 			if (!net_eq(dev_net(ifa->idev->dev), net))
4216 				continue;
4217 			/* sync with offset */
4218 			if (p < state->offset) {
4219 				p++;
4220 				continue;
4221 			}
4222 			return ifa;
4223 		}
4224 
4225 		/* prepare for next bucket */
4226 		state->offset = 0;
4227 		p = 0;
4228 	}
4229 	return NULL;
4230 }
4231 
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4232 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4233 					 struct inet6_ifaddr *ifa)
4234 {
4235 	struct if6_iter_state *state = seq->private;
4236 	struct net *net = seq_file_net(seq);
4237 
4238 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4239 		if (!net_eq(dev_net(ifa->idev->dev), net))
4240 			continue;
4241 		state->offset++;
4242 		return ifa;
4243 	}
4244 
4245 	state->offset = 0;
4246 	while (++state->bucket < IN6_ADDR_HSIZE) {
4247 		hlist_for_each_entry_rcu(ifa,
4248 				     &inet6_addr_lst[state->bucket], addr_lst) {
4249 			if (!net_eq(dev_net(ifa->idev->dev), net))
4250 				continue;
4251 			return ifa;
4252 		}
4253 	}
4254 
4255 	return NULL;
4256 }
4257 
if6_seq_start(struct seq_file * seq,loff_t * pos)4258 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4259 	__acquires(rcu)
4260 {
4261 	rcu_read_lock();
4262 	return if6_get_first(seq, *pos);
4263 }
4264 
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4265 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4266 {
4267 	struct inet6_ifaddr *ifa;
4268 
4269 	ifa = if6_get_next(seq, v);
4270 	++*pos;
4271 	return ifa;
4272 }
4273 
if6_seq_stop(struct seq_file * seq,void * v)4274 static void if6_seq_stop(struct seq_file *seq, void *v)
4275 	__releases(rcu)
4276 {
4277 	rcu_read_unlock();
4278 }
4279 
if6_seq_show(struct seq_file * seq,void * v)4280 static int if6_seq_show(struct seq_file *seq, void *v)
4281 {
4282 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4283 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4284 		   &ifp->addr,
4285 		   ifp->idev->dev->ifindex,
4286 		   ifp->prefix_len,
4287 		   ifp->scope,
4288 		   (u8) ifp->flags,
4289 		   ifp->idev->dev->name);
4290 	return 0;
4291 }
4292 
4293 static const struct seq_operations if6_seq_ops = {
4294 	.start	= if6_seq_start,
4295 	.next	= if6_seq_next,
4296 	.show	= if6_seq_show,
4297 	.stop	= if6_seq_stop,
4298 };
4299 
if6_proc_net_init(struct net * net)4300 static int __net_init if6_proc_net_init(struct net *net)
4301 {
4302 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4303 			sizeof(struct if6_iter_state)))
4304 		return -ENOMEM;
4305 	return 0;
4306 }
4307 
if6_proc_net_exit(struct net * net)4308 static void __net_exit if6_proc_net_exit(struct net *net)
4309 {
4310 	remove_proc_entry("if_inet6", net->proc_net);
4311 }
4312 
4313 static struct pernet_operations if6_proc_net_ops = {
4314 	.init = if6_proc_net_init,
4315 	.exit = if6_proc_net_exit,
4316 };
4317 
if6_proc_init(void)4318 int __init if6_proc_init(void)
4319 {
4320 	return register_pernet_subsys(&if6_proc_net_ops);
4321 }
4322 
if6_proc_exit(void)4323 void if6_proc_exit(void)
4324 {
4325 	unregister_pernet_subsys(&if6_proc_net_ops);
4326 }
4327 #endif	/* CONFIG_PROC_FS */
4328 
4329 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4330 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4331 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4332 {
4333 	unsigned int hash = inet6_addr_hash(net, addr);
4334 	struct inet6_ifaddr *ifp = NULL;
4335 	int ret = 0;
4336 
4337 	rcu_read_lock();
4338 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4339 		if (!net_eq(dev_net(ifp->idev->dev), net))
4340 			continue;
4341 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4342 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4343 			ret = 1;
4344 			break;
4345 		}
4346 	}
4347 	rcu_read_unlock();
4348 	return ret;
4349 }
4350 #endif
4351 
4352 /*
4353  *	Periodic address status verification
4354  */
4355 
addrconf_verify_rtnl(void)4356 static void addrconf_verify_rtnl(void)
4357 {
4358 	unsigned long now, next, next_sec, next_sched;
4359 	struct inet6_ifaddr *ifp;
4360 	int i;
4361 
4362 	ASSERT_RTNL();
4363 
4364 	rcu_read_lock_bh();
4365 	now = jiffies;
4366 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4367 
4368 	cancel_delayed_work(&addr_chk_work);
4369 
4370 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4371 restart:
4372 		hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4373 			unsigned long age;
4374 
4375 			/* When setting preferred_lft to a value not zero or
4376 			 * infinity, while valid_lft is infinity
4377 			 * IFA_F_PERMANENT has a non-infinity life time.
4378 			 */
4379 			if ((ifp->flags & IFA_F_PERMANENT) &&
4380 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4381 				continue;
4382 
4383 			spin_lock(&ifp->lock);
4384 			/* We try to batch several events at once. */
4385 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4386 
4387 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4388 			    age >= ifp->valid_lft) {
4389 				spin_unlock(&ifp->lock);
4390 				in6_ifa_hold(ifp);
4391 				ipv6_del_addr(ifp);
4392 				goto restart;
4393 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4394 				spin_unlock(&ifp->lock);
4395 				continue;
4396 			} else if (age >= ifp->prefered_lft) {
4397 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4398 				int deprecate = 0;
4399 
4400 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4401 					deprecate = 1;
4402 					ifp->flags |= IFA_F_DEPRECATED;
4403 				}
4404 
4405 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4406 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4407 					next = ifp->tstamp + ifp->valid_lft * HZ;
4408 
4409 				spin_unlock(&ifp->lock);
4410 
4411 				if (deprecate) {
4412 					in6_ifa_hold(ifp);
4413 
4414 					ipv6_ifa_notify(0, ifp);
4415 					in6_ifa_put(ifp);
4416 					goto restart;
4417 				}
4418 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4419 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4420 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4421 					ifp->idev->cnf.dad_transmits *
4422 					NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4423 
4424 				if (age >= ifp->prefered_lft - regen_advance) {
4425 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4426 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4427 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4428 					if (!ifp->regen_count && ifpub) {
4429 						ifp->regen_count++;
4430 						in6_ifa_hold(ifp);
4431 						in6_ifa_hold(ifpub);
4432 						spin_unlock(&ifp->lock);
4433 
4434 						spin_lock(&ifpub->lock);
4435 						ifpub->regen_count = 0;
4436 						spin_unlock(&ifpub->lock);
4437 						rcu_read_unlock_bh();
4438 						ipv6_create_tempaddr(ifpub, ifp, true);
4439 						in6_ifa_put(ifpub);
4440 						in6_ifa_put(ifp);
4441 						rcu_read_lock_bh();
4442 						goto restart;
4443 					}
4444 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4445 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4446 				spin_unlock(&ifp->lock);
4447 			} else {
4448 				/* ifp->prefered_lft <= ifp->valid_lft */
4449 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4450 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4451 				spin_unlock(&ifp->lock);
4452 			}
4453 		}
4454 	}
4455 
4456 	next_sec = round_jiffies_up(next);
4457 	next_sched = next;
4458 
4459 	/* If rounded timeout is accurate enough, accept it. */
4460 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4461 		next_sched = next_sec;
4462 
4463 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4464 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4465 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4466 
4467 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4468 		 now, next, next_sec, next_sched);
4469 	mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4470 	rcu_read_unlock_bh();
4471 }
4472 
addrconf_verify_work(struct work_struct * w)4473 static void addrconf_verify_work(struct work_struct *w)
4474 {
4475 	rtnl_lock();
4476 	addrconf_verify_rtnl();
4477 	rtnl_unlock();
4478 }
4479 
addrconf_verify(void)4480 static void addrconf_verify(void)
4481 {
4482 	mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4483 }
4484 
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4485 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4486 				     struct in6_addr **peer_pfx)
4487 {
4488 	struct in6_addr *pfx = NULL;
4489 
4490 	*peer_pfx = NULL;
4491 
4492 	if (addr)
4493 		pfx = nla_data(addr);
4494 
4495 	if (local) {
4496 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4497 			*peer_pfx = pfx;
4498 		pfx = nla_data(local);
4499 	}
4500 
4501 	return pfx;
4502 }
4503 
4504 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4505 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4506 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4507 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4508 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4509 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4510 };
4511 
4512 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4513 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4514 		  struct netlink_ext_ack *extack)
4515 {
4516 	struct net *net = sock_net(skb->sk);
4517 	struct ifaddrmsg *ifm;
4518 	struct nlattr *tb[IFA_MAX+1];
4519 	struct in6_addr *pfx, *peer_pfx;
4520 	u32 ifa_flags;
4521 	int err;
4522 
4523 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4524 			  extack);
4525 	if (err < 0)
4526 		return err;
4527 
4528 	ifm = nlmsg_data(nlh);
4529 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4530 	if (!pfx)
4531 		return -EINVAL;
4532 
4533 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4534 
4535 	/* We ignore other flags so far. */
4536 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4537 
4538 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4539 			      ifm->ifa_prefixlen);
4540 }
4541 
modify_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4542 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4543 			       unsigned long expires, u32 flags,
4544 			       bool modify_peer)
4545 {
4546 	struct fib6_info *f6i;
4547 	u32 prio;
4548 
4549 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4550 					ifp->prefix_len,
4551 					ifp->idev->dev,
4552 					0, RTF_GATEWAY | RTF_DEFAULT);
4553 	if (!f6i)
4554 		return -ENOENT;
4555 
4556 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4557 	if (f6i->fib6_metric != prio) {
4558 		/* delete old one */
4559 		ip6_del_rt(dev_net(ifp->idev->dev), f6i);
4560 
4561 		/* add new one */
4562 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4563 				      ifp->prefix_len,
4564 				      ifp->rt_priority, ifp->idev->dev,
4565 				      expires, flags, GFP_KERNEL);
4566 	} else {
4567 		if (!expires)
4568 			fib6_clean_expires(f6i);
4569 		else
4570 			fib6_set_expires(f6i, expires);
4571 
4572 		fib6_info_release(f6i);
4573 	}
4574 
4575 	return 0;
4576 }
4577 
inet6_addr_modify(struct inet6_ifaddr * ifp,struct ifa6_config * cfg)4578 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4579 {
4580 	u32 flags;
4581 	clock_t expires;
4582 	unsigned long timeout;
4583 	bool was_managetempaddr;
4584 	bool had_prefixroute;
4585 	bool new_peer = false;
4586 
4587 	ASSERT_RTNL();
4588 
4589 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4590 		return -EINVAL;
4591 
4592 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4593 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4594 		return -EINVAL;
4595 
4596 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4597 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4598 
4599 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4600 	if (addrconf_finite_timeout(timeout)) {
4601 		expires = jiffies_to_clock_t(timeout * HZ);
4602 		cfg->valid_lft = timeout;
4603 		flags = RTF_EXPIRES;
4604 	} else {
4605 		expires = 0;
4606 		flags = 0;
4607 		cfg->ifa_flags |= IFA_F_PERMANENT;
4608 	}
4609 
4610 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4611 	if (addrconf_finite_timeout(timeout)) {
4612 		if (timeout == 0)
4613 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4614 		cfg->preferred_lft = timeout;
4615 	}
4616 
4617 	if (cfg->peer_pfx &&
4618 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4619 		if (!ipv6_addr_any(&ifp->peer_addr))
4620 			cleanup_prefix_route(ifp, expires, true, true);
4621 		new_peer = true;
4622 	}
4623 
4624 	spin_lock_bh(&ifp->lock);
4625 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4626 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4627 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4628 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4629 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4630 			IFA_F_NOPREFIXROUTE);
4631 	ifp->flags |= cfg->ifa_flags;
4632 	ifp->tstamp = jiffies;
4633 	ifp->valid_lft = cfg->valid_lft;
4634 	ifp->prefered_lft = cfg->preferred_lft;
4635 
4636 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4637 		ifp->rt_priority = cfg->rt_priority;
4638 
4639 	if (new_peer)
4640 		ifp->peer_addr = *cfg->peer_pfx;
4641 
4642 	spin_unlock_bh(&ifp->lock);
4643 	if (!(ifp->flags&IFA_F_TENTATIVE))
4644 		ipv6_ifa_notify(0, ifp);
4645 
4646 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4647 		int rc = -ENOENT;
4648 
4649 		if (had_prefixroute)
4650 			rc = modify_prefix_route(ifp, expires, flags, false);
4651 
4652 		/* prefix route could have been deleted; if so restore it */
4653 		if (rc == -ENOENT) {
4654 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4655 					      ifp->rt_priority, ifp->idev->dev,
4656 					      expires, flags, GFP_KERNEL);
4657 		}
4658 
4659 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4660 			rc = modify_prefix_route(ifp, expires, flags, true);
4661 
4662 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4663 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4664 					      ifp->rt_priority, ifp->idev->dev,
4665 					      expires, flags, GFP_KERNEL);
4666 		}
4667 	} else if (had_prefixroute) {
4668 		enum cleanup_prefix_rt_t action;
4669 		unsigned long rt_expires;
4670 
4671 		write_lock_bh(&ifp->idev->lock);
4672 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4673 		write_unlock_bh(&ifp->idev->lock);
4674 
4675 		if (action != CLEANUP_PREFIX_RT_NOP) {
4676 			cleanup_prefix_route(ifp, rt_expires,
4677 				action == CLEANUP_PREFIX_RT_DEL, false);
4678 		}
4679 	}
4680 
4681 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4682 		if (was_managetempaddr &&
4683 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4684 			cfg->valid_lft = 0;
4685 			cfg->preferred_lft = 0;
4686 		}
4687 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4688 				 cfg->preferred_lft, !was_managetempaddr,
4689 				 jiffies);
4690 	}
4691 
4692 	addrconf_verify_rtnl();
4693 
4694 	return 0;
4695 }
4696 
4697 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4698 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4699 		  struct netlink_ext_ack *extack)
4700 {
4701 	struct net *net = sock_net(skb->sk);
4702 	struct ifaddrmsg *ifm;
4703 	struct nlattr *tb[IFA_MAX+1];
4704 	struct in6_addr *peer_pfx;
4705 	struct inet6_ifaddr *ifa;
4706 	struct net_device *dev;
4707 	struct inet6_dev *idev;
4708 	struct ifa6_config cfg;
4709 	int err;
4710 
4711 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4712 			  extack);
4713 	if (err < 0)
4714 		return err;
4715 
4716 	memset(&cfg, 0, sizeof(cfg));
4717 
4718 	ifm = nlmsg_data(nlh);
4719 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4720 	if (!cfg.pfx)
4721 		return -EINVAL;
4722 
4723 	cfg.peer_pfx = peer_pfx;
4724 	cfg.plen = ifm->ifa_prefixlen;
4725 	if (tb[IFA_RT_PRIORITY])
4726 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4727 
4728 	cfg.valid_lft = INFINITY_LIFE_TIME;
4729 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4730 
4731 	if (tb[IFA_CACHEINFO]) {
4732 		struct ifa_cacheinfo *ci;
4733 
4734 		ci = nla_data(tb[IFA_CACHEINFO]);
4735 		cfg.valid_lft = ci->ifa_valid;
4736 		cfg.preferred_lft = ci->ifa_prefered;
4737 	}
4738 
4739 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4740 	if (!dev)
4741 		return -ENODEV;
4742 
4743 	if (tb[IFA_FLAGS])
4744 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4745 	else
4746 		cfg.ifa_flags = ifm->ifa_flags;
4747 
4748 	/* We ignore other flags so far. */
4749 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4750 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4751 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4752 
4753 	idev = ipv6_find_idev(dev);
4754 	if (!idev)
4755 		return -ENOBUFS;
4756 
4757 	if (!ipv6_allow_optimistic_dad(net, idev))
4758 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4759 
4760 	if (cfg.ifa_flags & IFA_F_NODAD &&
4761 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4762 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4763 		return -EINVAL;
4764 	}
4765 
4766 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4767 	if (!ifa) {
4768 		/*
4769 		 * It would be best to check for !NLM_F_CREATE here but
4770 		 * userspace already relies on not having to provide this.
4771 		 */
4772 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4773 	}
4774 
4775 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4776 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4777 		err = -EEXIST;
4778 	else
4779 		err = inet6_addr_modify(ifa, &cfg);
4780 
4781 	in6_ifa_put(ifa);
4782 
4783 	return err;
4784 }
4785 
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)4786 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4787 			  u8 scope, int ifindex)
4788 {
4789 	struct ifaddrmsg *ifm;
4790 
4791 	ifm = nlmsg_data(nlh);
4792 	ifm->ifa_family = AF_INET6;
4793 	ifm->ifa_prefixlen = prefixlen;
4794 	ifm->ifa_flags = flags;
4795 	ifm->ifa_scope = scope;
4796 	ifm->ifa_index = ifindex;
4797 }
4798 
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)4799 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4800 			 unsigned long tstamp, u32 preferred, u32 valid)
4801 {
4802 	struct ifa_cacheinfo ci;
4803 
4804 	ci.cstamp = cstamp_delta(cstamp);
4805 	ci.tstamp = cstamp_delta(tstamp);
4806 	ci.ifa_prefered = preferred;
4807 	ci.ifa_valid = valid;
4808 
4809 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4810 }
4811 
rt_scope(int ifa_scope)4812 static inline int rt_scope(int ifa_scope)
4813 {
4814 	if (ifa_scope & IFA_HOST)
4815 		return RT_SCOPE_HOST;
4816 	else if (ifa_scope & IFA_LINK)
4817 		return RT_SCOPE_LINK;
4818 	else if (ifa_scope & IFA_SITE)
4819 		return RT_SCOPE_SITE;
4820 	else
4821 		return RT_SCOPE_UNIVERSE;
4822 }
4823 
inet6_ifaddr_msgsize(void)4824 static inline int inet6_ifaddr_msgsize(void)
4825 {
4826 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4827 	       + nla_total_size(16) /* IFA_LOCAL */
4828 	       + nla_total_size(16) /* IFA_ADDRESS */
4829 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4830 	       + nla_total_size(4)  /* IFA_FLAGS */
4831 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4832 }
4833 
inet6_fill_ifaddr(struct sk_buff * skb,struct inet6_ifaddr * ifa,u32 portid,u32 seq,int event,unsigned int flags)4834 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4835 			     u32 portid, u32 seq, int event, unsigned int flags)
4836 {
4837 	struct nlmsghdr  *nlh;
4838 	u32 preferred, valid;
4839 
4840 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4841 	if (!nlh)
4842 		return -EMSGSIZE;
4843 
4844 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4845 		      ifa->idev->dev->ifindex);
4846 
4847 	if (!((ifa->flags&IFA_F_PERMANENT) &&
4848 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4849 		preferred = ifa->prefered_lft;
4850 		valid = ifa->valid_lft;
4851 		if (preferred != INFINITY_LIFE_TIME) {
4852 			long tval = (jiffies - ifa->tstamp)/HZ;
4853 			if (preferred > tval)
4854 				preferred -= tval;
4855 			else
4856 				preferred = 0;
4857 			if (valid != INFINITY_LIFE_TIME) {
4858 				if (valid > tval)
4859 					valid -= tval;
4860 				else
4861 					valid = 0;
4862 			}
4863 		}
4864 	} else {
4865 		preferred = INFINITY_LIFE_TIME;
4866 		valid = INFINITY_LIFE_TIME;
4867 	}
4868 
4869 	if (!ipv6_addr_any(&ifa->peer_addr)) {
4870 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4871 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4872 			goto error;
4873 	} else
4874 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4875 			goto error;
4876 
4877 	if (ifa->rt_priority &&
4878 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4879 		goto error;
4880 
4881 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4882 		goto error;
4883 
4884 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4885 		goto error;
4886 
4887 	nlmsg_end(skb, nlh);
4888 	return 0;
4889 
4890 error:
4891 	nlmsg_cancel(skb, nlh);
4892 	return -EMSGSIZE;
4893 }
4894 
inet6_fill_ifmcaddr(struct sk_buff * skb,struct ifmcaddr6 * ifmca,u32 portid,u32 seq,int event,u16 flags)4895 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4896 				u32 portid, u32 seq, int event, u16 flags)
4897 {
4898 	struct nlmsghdr  *nlh;
4899 	u8 scope = RT_SCOPE_UNIVERSE;
4900 	int ifindex = ifmca->idev->dev->ifindex;
4901 
4902 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4903 		scope = RT_SCOPE_SITE;
4904 
4905 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4906 	if (!nlh)
4907 		return -EMSGSIZE;
4908 
4909 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4910 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4911 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4912 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4913 		nlmsg_cancel(skb, nlh);
4914 		return -EMSGSIZE;
4915 	}
4916 
4917 	nlmsg_end(skb, nlh);
4918 	return 0;
4919 }
4920 
inet6_fill_ifacaddr(struct sk_buff * skb,struct ifacaddr6 * ifaca,u32 portid,u32 seq,int event,unsigned int flags)4921 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4922 				u32 portid, u32 seq, int event, unsigned int flags)
4923 {
4924 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
4925 	int ifindex = dev ? dev->ifindex : 1;
4926 	struct nlmsghdr  *nlh;
4927 	u8 scope = RT_SCOPE_UNIVERSE;
4928 
4929 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4930 		scope = RT_SCOPE_SITE;
4931 
4932 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4933 	if (!nlh)
4934 		return -EMSGSIZE;
4935 
4936 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4937 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4938 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4939 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4940 		nlmsg_cancel(skb, nlh);
4941 		return -EMSGSIZE;
4942 	}
4943 
4944 	nlmsg_end(skb, nlh);
4945 	return 0;
4946 }
4947 
4948 enum addr_type_t {
4949 	UNICAST_ADDR,
4950 	MULTICAST_ADDR,
4951 	ANYCAST_ADDR,
4952 };
4953 
4954 /* called with rcu_read_lock() */
in6_dump_addrs(struct inet6_dev * idev,struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type,int s_ip_idx,int * p_ip_idx)4955 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4956 			  struct netlink_callback *cb, enum addr_type_t type,
4957 			  int s_ip_idx, int *p_ip_idx)
4958 {
4959 	struct ifmcaddr6 *ifmca;
4960 	struct ifacaddr6 *ifaca;
4961 	int err = 1;
4962 	int ip_idx = *p_ip_idx;
4963 
4964 	read_lock_bh(&idev->lock);
4965 	switch (type) {
4966 	case UNICAST_ADDR: {
4967 		struct inet6_ifaddr *ifa;
4968 
4969 		/* unicast address incl. temp addr */
4970 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
4971 			if (ip_idx < s_ip_idx)
4972 				goto next;
4973 			err = inet6_fill_ifaddr(skb, ifa,
4974 						NETLINK_CB(cb->skb).portid,
4975 						cb->nlh->nlmsg_seq,
4976 						RTM_NEWADDR,
4977 						NLM_F_MULTI);
4978 			if (err < 0)
4979 				break;
4980 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4981 next:
4982 			ip_idx++;
4983 		}
4984 		break;
4985 	}
4986 	case MULTICAST_ADDR:
4987 		/* multicast address */
4988 		for (ifmca = idev->mc_list; ifmca;
4989 		     ifmca = ifmca->next, ip_idx++) {
4990 			if (ip_idx < s_ip_idx)
4991 				continue;
4992 			err = inet6_fill_ifmcaddr(skb, ifmca,
4993 						  NETLINK_CB(cb->skb).portid,
4994 						  cb->nlh->nlmsg_seq,
4995 						  RTM_GETMULTICAST,
4996 						  NLM_F_MULTI);
4997 			if (err < 0)
4998 				break;
4999 		}
5000 		break;
5001 	case ANYCAST_ADDR:
5002 		/* anycast address */
5003 		for (ifaca = idev->ac_list; ifaca;
5004 		     ifaca = ifaca->aca_next, ip_idx++) {
5005 			if (ip_idx < s_ip_idx)
5006 				continue;
5007 			err = inet6_fill_ifacaddr(skb, ifaca,
5008 						  NETLINK_CB(cb->skb).portid,
5009 						  cb->nlh->nlmsg_seq,
5010 						  RTM_GETANYCAST,
5011 						  NLM_F_MULTI);
5012 			if (err < 0)
5013 				break;
5014 		}
5015 		break;
5016 	default:
5017 		break;
5018 	}
5019 	read_unlock_bh(&idev->lock);
5020 	*p_ip_idx = ip_idx;
5021 	return err;
5022 }
5023 
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5024 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5025 			   enum addr_type_t type)
5026 {
5027 	struct net *net = sock_net(skb->sk);
5028 	int h, s_h;
5029 	int idx, ip_idx;
5030 	int s_idx, s_ip_idx;
5031 	struct net_device *dev;
5032 	struct inet6_dev *idev;
5033 	struct hlist_head *head;
5034 
5035 	s_h = cb->args[0];
5036 	s_idx = idx = cb->args[1];
5037 	s_ip_idx = ip_idx = cb->args[2];
5038 
5039 	rcu_read_lock();
5040 	cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
5041 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5042 		idx = 0;
5043 		head = &net->dev_index_head[h];
5044 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5045 			if (idx < s_idx)
5046 				goto cont;
5047 			if (h > s_h || idx > s_idx)
5048 				s_ip_idx = 0;
5049 			ip_idx = 0;
5050 			idev = __in6_dev_get(dev);
5051 			if (!idev)
5052 				goto cont;
5053 
5054 			if (in6_dump_addrs(idev, skb, cb, type,
5055 					   s_ip_idx, &ip_idx) < 0)
5056 				goto done;
5057 cont:
5058 			idx++;
5059 		}
5060 	}
5061 done:
5062 	rcu_read_unlock();
5063 	cb->args[0] = h;
5064 	cb->args[1] = idx;
5065 	cb->args[2] = ip_idx;
5066 
5067 	return skb->len;
5068 }
5069 
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5070 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5071 {
5072 	enum addr_type_t type = UNICAST_ADDR;
5073 
5074 	return inet6_dump_addr(skb, cb, type);
5075 }
5076 
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5077 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5078 {
5079 	enum addr_type_t type = MULTICAST_ADDR;
5080 
5081 	return inet6_dump_addr(skb, cb, type);
5082 }
5083 
5084 
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5085 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5086 {
5087 	enum addr_type_t type = ANYCAST_ADDR;
5088 
5089 	return inet6_dump_addr(skb, cb, type);
5090 }
5091 
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5092 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5093 			     struct netlink_ext_ack *extack)
5094 {
5095 	struct net *net = sock_net(in_skb->sk);
5096 	struct ifaddrmsg *ifm;
5097 	struct nlattr *tb[IFA_MAX+1];
5098 	struct in6_addr *addr = NULL, *peer;
5099 	struct net_device *dev = NULL;
5100 	struct inet6_ifaddr *ifa;
5101 	struct sk_buff *skb;
5102 	int err;
5103 
5104 	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
5105 			  extack);
5106 	if (err < 0)
5107 		return err;
5108 
5109 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5110 	if (!addr)
5111 		return -EINVAL;
5112 
5113 	ifm = nlmsg_data(nlh);
5114 	if (ifm->ifa_index)
5115 		dev = dev_get_by_index(net, ifm->ifa_index);
5116 
5117 	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
5118 	if (!ifa) {
5119 		err = -EADDRNOTAVAIL;
5120 		goto errout;
5121 	}
5122 
5123 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5124 	if (!skb) {
5125 		err = -ENOBUFS;
5126 		goto errout_ifa;
5127 	}
5128 
5129 	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5130 				nlh->nlmsg_seq, RTM_NEWADDR, 0);
5131 	if (err < 0) {
5132 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5133 		WARN_ON(err == -EMSGSIZE);
5134 		kfree_skb(skb);
5135 		goto errout_ifa;
5136 	}
5137 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5138 errout_ifa:
5139 	in6_ifa_put(ifa);
5140 errout:
5141 	if (dev)
5142 		dev_put(dev);
5143 	return err;
5144 }
5145 
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5146 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5147 {
5148 	struct sk_buff *skb;
5149 	struct net *net = dev_net(ifa->idev->dev);
5150 	int err = -ENOBUFS;
5151 
5152 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5153 	if (!skb)
5154 		goto errout;
5155 
5156 	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5157 	if (err < 0) {
5158 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5159 		WARN_ON(err == -EMSGSIZE);
5160 		kfree_skb(skb);
5161 		goto errout;
5162 	}
5163 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5164 	return;
5165 errout:
5166 	if (err < 0)
5167 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5168 }
5169 
ipv6_store_devconf(struct ipv6_devconf * cnf,__s32 * array,int bytes)5170 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5171 				__s32 *array, int bytes)
5172 {
5173 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5174 
5175 	memset(array, 0, bytes);
5176 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5177 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5178 	array[DEVCONF_MTU6] = cnf->mtu6;
5179 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5180 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5181 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5182 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5183 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5184 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5185 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5186 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5187 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5188 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5189 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5190 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5191 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5192 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5193 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5194 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5195 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5196 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5197 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5198 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5199 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5200 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5201 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5202 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5203 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5204 #ifdef CONFIG_IPV6_ROUTER_PREF
5205 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5206 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5207 		jiffies_to_msecs(cnf->rtr_probe_interval);
5208 #ifdef CONFIG_IPV6_ROUTE_INFO
5209 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5210 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5211 #endif
5212 #endif
5213 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5214 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5215 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5216 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5217 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5218 #endif
5219 #ifdef CONFIG_IPV6_MROUTE
5220 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5221 #endif
5222 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5223 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5224 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5225 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5226 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5227 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5228 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5229 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5230 	/* we omit DEVCONF_STABLE_SECRET for now */
5231 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5232 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5233 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5234 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5235 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5236 #ifdef CONFIG_IPV6_SEG6_HMAC
5237 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5238 #endif
5239 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5240 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5241 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5242 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5243 }
5244 
inet6_ifla6_size(void)5245 static inline size_t inet6_ifla6_size(void)
5246 {
5247 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5248 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5249 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5250 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5251 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5252 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5253 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5254 	     + 0;
5255 }
5256 
inet6_if_nlmsg_size(void)5257 static inline size_t inet6_if_nlmsg_size(void)
5258 {
5259 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5260 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5261 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5262 	       + nla_total_size(4) /* IFLA_MTU */
5263 	       + nla_total_size(4) /* IFLA_LINK */
5264 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5265 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5266 }
5267 
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5268 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5269 					int bytes)
5270 {
5271 	int i;
5272 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5273 	BUG_ON(pad < 0);
5274 
5275 	/* Use put_unaligned() because stats may not be aligned for u64. */
5276 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5277 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5278 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5279 
5280 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5281 }
5282 
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5283 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5284 					int bytes, size_t syncpoff)
5285 {
5286 	int i, c;
5287 	u64 buff[IPSTATS_MIB_MAX];
5288 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5289 
5290 	BUG_ON(pad < 0);
5291 
5292 	memset(buff, 0, sizeof(buff));
5293 	buff[0] = IPSTATS_MIB_MAX;
5294 
5295 	for_each_possible_cpu(c) {
5296 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5297 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5298 	}
5299 
5300 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5301 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5302 }
5303 
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5304 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5305 			     int bytes)
5306 {
5307 	switch (attrtype) {
5308 	case IFLA_INET6_STATS:
5309 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5310 				     offsetof(struct ipstats_mib, syncp));
5311 		break;
5312 	case IFLA_INET6_ICMP6STATS:
5313 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5314 		break;
5315 	}
5316 }
5317 
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5318 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5319 				  u32 ext_filter_mask)
5320 {
5321 	struct nlattr *nla;
5322 	struct ifla_cacheinfo ci;
5323 
5324 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5325 		goto nla_put_failure;
5326 	ci.max_reasm_len = IPV6_MAXPLEN;
5327 	ci.tstamp = cstamp_delta(idev->tstamp);
5328 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5329 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5330 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5331 		goto nla_put_failure;
5332 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5333 	if (!nla)
5334 		goto nla_put_failure;
5335 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5336 
5337 	/* XXX - MC not implemented */
5338 
5339 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5340 		return 0;
5341 
5342 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5343 	if (!nla)
5344 		goto nla_put_failure;
5345 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5346 
5347 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5348 	if (!nla)
5349 		goto nla_put_failure;
5350 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5351 
5352 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5353 	if (!nla)
5354 		goto nla_put_failure;
5355 
5356 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5357 		goto nla_put_failure;
5358 
5359 	read_lock_bh(&idev->lock);
5360 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5361 	read_unlock_bh(&idev->lock);
5362 
5363 	return 0;
5364 
5365 nla_put_failure:
5366 	return -EMSGSIZE;
5367 }
5368 
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5369 static size_t inet6_get_link_af_size(const struct net_device *dev,
5370 				     u32 ext_filter_mask)
5371 {
5372 	if (!__in6_dev_get(dev))
5373 		return 0;
5374 
5375 	return inet6_ifla6_size();
5376 }
5377 
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5378 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5379 			      u32 ext_filter_mask)
5380 {
5381 	struct inet6_dev *idev = __in6_dev_get(dev);
5382 
5383 	if (!idev)
5384 		return -ENODATA;
5385 
5386 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5387 		return -EMSGSIZE;
5388 
5389 	return 0;
5390 }
5391 
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token)5392 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5393 {
5394 	struct inet6_ifaddr *ifp;
5395 	struct net_device *dev = idev->dev;
5396 	bool clear_token, update_rs = false;
5397 	struct in6_addr ll_addr;
5398 
5399 	ASSERT_RTNL();
5400 
5401 	if (!token)
5402 		return -EINVAL;
5403 	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5404 		return -EINVAL;
5405 	if (!ipv6_accept_ra(idev))
5406 		return -EINVAL;
5407 	if (idev->cnf.rtr_solicits == 0)
5408 		return -EINVAL;
5409 
5410 	write_lock_bh(&idev->lock);
5411 
5412 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5413 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5414 
5415 	write_unlock_bh(&idev->lock);
5416 
5417 	clear_token = ipv6_addr_any(token);
5418 	if (clear_token)
5419 		goto update_lft;
5420 
5421 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5422 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5423 			     IFA_F_OPTIMISTIC)) {
5424 		/* If we're not ready, then normal ifup will take care
5425 		 * of this. Otherwise, we need to request our rs here.
5426 		 */
5427 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5428 		update_rs = true;
5429 	}
5430 
5431 update_lft:
5432 	write_lock_bh(&idev->lock);
5433 
5434 	if (update_rs) {
5435 		idev->if_flags |= IF_RS_SENT;
5436 		idev->rs_interval = rfc3315_s14_backoff_init(
5437 			idev->cnf.rtr_solicit_interval);
5438 		idev->rs_probes = 1;
5439 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5440 	}
5441 
5442 	/* Well, that's kinda nasty ... */
5443 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5444 		spin_lock(&ifp->lock);
5445 		if (ifp->tokenized) {
5446 			ifp->valid_lft = 0;
5447 			ifp->prefered_lft = 0;
5448 		}
5449 		spin_unlock(&ifp->lock);
5450 	}
5451 
5452 	write_unlock_bh(&idev->lock);
5453 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5454 	addrconf_verify_rtnl();
5455 	return 0;
5456 }
5457 
5458 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5459 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5460 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5461 };
5462 
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla)5463 static int inet6_validate_link_af(const struct net_device *dev,
5464 				  const struct nlattr *nla)
5465 {
5466 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5467 
5468 	if (dev && !__in6_dev_get(dev))
5469 		return -EAFNOSUPPORT;
5470 
5471 	return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5472 				NULL);
5473 }
5474 
check_addr_gen_mode(int mode)5475 static int check_addr_gen_mode(int mode)
5476 {
5477 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5478 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5479 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5480 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5481 		return -EINVAL;
5482 	return 1;
5483 }
5484 
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)5485 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5486 				int mode)
5487 {
5488 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5489 	    !idev->cnf.stable_secret.initialized &&
5490 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5491 		return -EINVAL;
5492 	return 1;
5493 }
5494 
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla)5495 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5496 {
5497 	int err = -EINVAL;
5498 	struct inet6_dev *idev = __in6_dev_get(dev);
5499 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5500 
5501 	if (!idev)
5502 		return -EAFNOSUPPORT;
5503 
5504 	if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5505 		BUG();
5506 
5507 	if (tb[IFLA_INET6_TOKEN]) {
5508 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5509 		if (err)
5510 			return err;
5511 	}
5512 
5513 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5514 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5515 
5516 		if (check_addr_gen_mode(mode) < 0 ||
5517 		    check_stable_privacy(idev, dev_net(dev), mode) < 0)
5518 			return -EINVAL;
5519 
5520 		idev->cnf.addr_gen_mode = mode;
5521 		err = 0;
5522 	}
5523 
5524 	return err;
5525 }
5526 
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)5527 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5528 			     u32 portid, u32 seq, int event, unsigned int flags)
5529 {
5530 	struct net_device *dev = idev->dev;
5531 	struct ifinfomsg *hdr;
5532 	struct nlmsghdr *nlh;
5533 	void *protoinfo;
5534 
5535 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5536 	if (!nlh)
5537 		return -EMSGSIZE;
5538 
5539 	hdr = nlmsg_data(nlh);
5540 	hdr->ifi_family = AF_INET6;
5541 	hdr->__ifi_pad = 0;
5542 	hdr->ifi_type = dev->type;
5543 	hdr->ifi_index = dev->ifindex;
5544 	hdr->ifi_flags = dev_get_flags(dev);
5545 	hdr->ifi_change = 0;
5546 
5547 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5548 	    (dev->addr_len &&
5549 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5550 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5551 	    (dev->ifindex != dev_get_iflink(dev) &&
5552 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5553 	    nla_put_u8(skb, IFLA_OPERSTATE,
5554 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5555 		goto nla_put_failure;
5556 	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5557 	if (!protoinfo)
5558 		goto nla_put_failure;
5559 
5560 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5561 		goto nla_put_failure;
5562 
5563 	nla_nest_end(skb, protoinfo);
5564 	nlmsg_end(skb, nlh);
5565 	return 0;
5566 
5567 nla_put_failure:
5568 	nlmsg_cancel(skb, nlh);
5569 	return -EMSGSIZE;
5570 }
5571 
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)5572 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5573 {
5574 	struct net *net = sock_net(skb->sk);
5575 	int h, s_h;
5576 	int idx = 0, s_idx;
5577 	struct net_device *dev;
5578 	struct inet6_dev *idev;
5579 	struct hlist_head *head;
5580 
5581 	s_h = cb->args[0];
5582 	s_idx = cb->args[1];
5583 
5584 	rcu_read_lock();
5585 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5586 		idx = 0;
5587 		head = &net->dev_index_head[h];
5588 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5589 			if (idx < s_idx)
5590 				goto cont;
5591 			idev = __in6_dev_get(dev);
5592 			if (!idev)
5593 				goto cont;
5594 			if (inet6_fill_ifinfo(skb, idev,
5595 					      NETLINK_CB(cb->skb).portid,
5596 					      cb->nlh->nlmsg_seq,
5597 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
5598 				goto out;
5599 cont:
5600 			idx++;
5601 		}
5602 	}
5603 out:
5604 	rcu_read_unlock();
5605 	cb->args[1] = idx;
5606 	cb->args[0] = h;
5607 
5608 	return skb->len;
5609 }
5610 
inet6_ifinfo_notify(int event,struct inet6_dev * idev)5611 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5612 {
5613 	struct sk_buff *skb;
5614 	struct net *net = dev_net(idev->dev);
5615 	int err = -ENOBUFS;
5616 
5617 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5618 	if (!skb)
5619 		goto errout;
5620 
5621 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5622 	if (err < 0) {
5623 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5624 		WARN_ON(err == -EMSGSIZE);
5625 		kfree_skb(skb);
5626 		goto errout;
5627 	}
5628 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5629 	return;
5630 errout:
5631 	if (err < 0)
5632 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5633 }
5634 
inet6_prefix_nlmsg_size(void)5635 static inline size_t inet6_prefix_nlmsg_size(void)
5636 {
5637 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
5638 	       + nla_total_size(sizeof(struct in6_addr))
5639 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
5640 }
5641 
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)5642 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5643 			     struct prefix_info *pinfo, u32 portid, u32 seq,
5644 			     int event, unsigned int flags)
5645 {
5646 	struct prefixmsg *pmsg;
5647 	struct nlmsghdr *nlh;
5648 	struct prefix_cacheinfo	ci;
5649 
5650 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5651 	if (!nlh)
5652 		return -EMSGSIZE;
5653 
5654 	pmsg = nlmsg_data(nlh);
5655 	pmsg->prefix_family = AF_INET6;
5656 	pmsg->prefix_pad1 = 0;
5657 	pmsg->prefix_pad2 = 0;
5658 	pmsg->prefix_ifindex = idev->dev->ifindex;
5659 	pmsg->prefix_len = pinfo->prefix_len;
5660 	pmsg->prefix_type = pinfo->type;
5661 	pmsg->prefix_pad3 = 0;
5662 	pmsg->prefix_flags = 0;
5663 	if (pinfo->onlink)
5664 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5665 	if (pinfo->autoconf)
5666 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5667 
5668 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5669 		goto nla_put_failure;
5670 	ci.preferred_time = ntohl(pinfo->prefered);
5671 	ci.valid_time = ntohl(pinfo->valid);
5672 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5673 		goto nla_put_failure;
5674 	nlmsg_end(skb, nlh);
5675 	return 0;
5676 
5677 nla_put_failure:
5678 	nlmsg_cancel(skb, nlh);
5679 	return -EMSGSIZE;
5680 }
5681 
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)5682 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5683 			 struct prefix_info *pinfo)
5684 {
5685 	struct sk_buff *skb;
5686 	struct net *net = dev_net(idev->dev);
5687 	int err = -ENOBUFS;
5688 
5689 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5690 	if (!skb)
5691 		goto errout;
5692 
5693 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5694 	if (err < 0) {
5695 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5696 		WARN_ON(err == -EMSGSIZE);
5697 		kfree_skb(skb);
5698 		goto errout;
5699 	}
5700 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5701 	return;
5702 errout:
5703 	if (err < 0)
5704 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5705 }
5706 
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)5707 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5708 {
5709 	struct net *net = dev_net(ifp->idev->dev);
5710 
5711 	if (event)
5712 		ASSERT_RTNL();
5713 
5714 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5715 
5716 	switch (event) {
5717 	case RTM_NEWADDR:
5718 		/*
5719 		 * If the address was optimistic we inserted the route at the
5720 		 * start of our DAD process, so we don't need to do it again.
5721 		 * If the device was taken down in the middle of the DAD
5722 		 * cycle there is a race where we could get here without a
5723 		 * host route, so nothing to insert. That will be fixed when
5724 		 * the device is brought up.
5725 		 */
5726 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
5727 			ip6_ins_rt(net, ifp->rt);
5728 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
5729 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
5730 				&ifp->addr, ifp->idev->dev->name);
5731 		}
5732 
5733 		if (ifp->idev->cnf.forwarding)
5734 			addrconf_join_anycast(ifp);
5735 		if (!ipv6_addr_any(&ifp->peer_addr))
5736 			addrconf_prefix_route(&ifp->peer_addr, 128,
5737 					      ifp->rt_priority, ifp->idev->dev,
5738 					      0, 0, GFP_ATOMIC);
5739 		break;
5740 	case RTM_DELADDR:
5741 		if (ifp->idev->cnf.forwarding)
5742 			addrconf_leave_anycast(ifp);
5743 		addrconf_leave_solict(ifp->idev, &ifp->addr);
5744 		if (!ipv6_addr_any(&ifp->peer_addr)) {
5745 			struct fib6_info *rt;
5746 
5747 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5748 						       ifp->idev->dev, 0, 0);
5749 			if (rt)
5750 				ip6_del_rt(net, rt);
5751 		}
5752 		if (ifp->rt) {
5753 			ip6_del_rt(net, ifp->rt);
5754 			ifp->rt = NULL;
5755 		}
5756 		rt_genid_bump_ipv6(net);
5757 		break;
5758 	}
5759 	atomic_inc(&net->ipv6.dev_addr_genid);
5760 }
5761 
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)5762 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5763 {
5764 	rcu_read_lock_bh();
5765 	if (likely(ifp->idev->dead == 0))
5766 		__ipv6_ifa_notify(event, ifp);
5767 	rcu_read_unlock_bh();
5768 }
5769 
5770 #ifdef CONFIG_SYSCTL
5771 
5772 static
addrconf_sysctl_forward(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)5773 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5774 			   void __user *buffer, size_t *lenp, loff_t *ppos)
5775 {
5776 	int *valp = ctl->data;
5777 	int val = *valp;
5778 	loff_t pos = *ppos;
5779 	struct ctl_table lctl;
5780 	int ret;
5781 
5782 	/*
5783 	 * ctl->data points to idev->cnf.forwarding, we should
5784 	 * not modify it until we get the rtnl lock.
5785 	 */
5786 	lctl = *ctl;
5787 	lctl.data = &val;
5788 
5789 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5790 
5791 	if (write)
5792 		ret = addrconf_fixup_forwarding(ctl, valp, val);
5793 	if (ret)
5794 		*ppos = pos;
5795 	return ret;
5796 }
5797 
5798 static
addrconf_sysctl_mtu(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)5799 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5800 			void __user *buffer, size_t *lenp, loff_t *ppos)
5801 {
5802 	struct inet6_dev *idev = ctl->extra1;
5803 	int min_mtu = IPV6_MIN_MTU;
5804 	struct ctl_table lctl;
5805 
5806 	lctl = *ctl;
5807 	lctl.extra1 = &min_mtu;
5808 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5809 
5810 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5811 }
5812 
dev_disable_change(struct inet6_dev * idev)5813 static void dev_disable_change(struct inet6_dev *idev)
5814 {
5815 	struct netdev_notifier_info info;
5816 
5817 	if (!idev || !idev->dev)
5818 		return;
5819 
5820 	netdev_notifier_info_init(&info, idev->dev);
5821 	if (idev->cnf.disable_ipv6)
5822 		addrconf_notify(NULL, NETDEV_DOWN, &info);
5823 	else
5824 		addrconf_notify(NULL, NETDEV_UP, &info);
5825 }
5826 
addrconf_disable_change(struct net * net,__s32 newf)5827 static void addrconf_disable_change(struct net *net, __s32 newf)
5828 {
5829 	struct net_device *dev;
5830 	struct inet6_dev *idev;
5831 
5832 	for_each_netdev(net, dev) {
5833 		idev = __in6_dev_get(dev);
5834 		if (idev) {
5835 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5836 			idev->cnf.disable_ipv6 = newf;
5837 			if (changed)
5838 				dev_disable_change(idev);
5839 		}
5840 	}
5841 }
5842 
addrconf_disable_ipv6(struct ctl_table * table,int * p,int newf)5843 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5844 {
5845 	struct net *net;
5846 	int old;
5847 
5848 	if (!rtnl_trylock())
5849 		return restart_syscall();
5850 
5851 	net = (struct net *)table->extra2;
5852 	old = *p;
5853 	*p = newf;
5854 
5855 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5856 		rtnl_unlock();
5857 		return 0;
5858 	}
5859 
5860 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
5861 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
5862 		addrconf_disable_change(net, newf);
5863 	} else if ((!newf) ^ (!old))
5864 		dev_disable_change((struct inet6_dev *)table->extra1);
5865 
5866 	rtnl_unlock();
5867 	return 0;
5868 }
5869 
5870 static
addrconf_sysctl_disable(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)5871 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5872 			    void __user *buffer, size_t *lenp, loff_t *ppos)
5873 {
5874 	int *valp = ctl->data;
5875 	int val = *valp;
5876 	loff_t pos = *ppos;
5877 	struct ctl_table lctl;
5878 	int ret;
5879 
5880 	/*
5881 	 * ctl->data points to idev->cnf.disable_ipv6, we should
5882 	 * not modify it until we get the rtnl lock.
5883 	 */
5884 	lctl = *ctl;
5885 	lctl.data = &val;
5886 
5887 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5888 
5889 	if (write)
5890 		ret = addrconf_disable_ipv6(ctl, valp, val);
5891 	if (ret)
5892 		*ppos = pos;
5893 	return ret;
5894 }
5895 
5896 static
addrconf_sysctl_proxy_ndp(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)5897 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5898 			      void __user *buffer, size_t *lenp, loff_t *ppos)
5899 {
5900 	int *valp = ctl->data;
5901 	int ret;
5902 	int old, new;
5903 
5904 	old = *valp;
5905 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5906 	new = *valp;
5907 
5908 	if (write && old != new) {
5909 		struct net *net = ctl->extra2;
5910 
5911 		if (!rtnl_trylock())
5912 			return restart_syscall();
5913 
5914 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5915 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5916 						     NETCONFA_PROXY_NEIGH,
5917 						     NETCONFA_IFINDEX_DEFAULT,
5918 						     net->ipv6.devconf_dflt);
5919 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5920 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5921 						     NETCONFA_PROXY_NEIGH,
5922 						     NETCONFA_IFINDEX_ALL,
5923 						     net->ipv6.devconf_all);
5924 		else {
5925 			struct inet6_dev *idev = ctl->extra1;
5926 
5927 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5928 						     NETCONFA_PROXY_NEIGH,
5929 						     idev->dev->ifindex,
5930 						     &idev->cnf);
5931 		}
5932 		rtnl_unlock();
5933 	}
5934 
5935 	return ret;
5936 }
5937 
addrconf_sysctl_addr_gen_mode(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)5938 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5939 					 void __user *buffer, size_t *lenp,
5940 					 loff_t *ppos)
5941 {
5942 	int ret = 0;
5943 	u32 new_val;
5944 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5945 	struct net *net = (struct net *)ctl->extra2;
5946 	struct ctl_table tmp = {
5947 		.data = &new_val,
5948 		.maxlen = sizeof(new_val),
5949 		.mode = ctl->mode,
5950 	};
5951 
5952 	if (!rtnl_trylock())
5953 		return restart_syscall();
5954 
5955 	new_val = *((u32 *)ctl->data);
5956 
5957 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
5958 	if (ret != 0)
5959 		goto out;
5960 
5961 	if (write) {
5962 		if (check_addr_gen_mode(new_val) < 0) {
5963 			ret = -EINVAL;
5964 			goto out;
5965 		}
5966 
5967 		if (idev) {
5968 			if (check_stable_privacy(idev, net, new_val) < 0) {
5969 				ret = -EINVAL;
5970 				goto out;
5971 			}
5972 
5973 			if (idev->cnf.addr_gen_mode != new_val) {
5974 				idev->cnf.addr_gen_mode = new_val;
5975 				addrconf_dev_config(idev->dev);
5976 			}
5977 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
5978 			struct net_device *dev;
5979 
5980 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
5981 			for_each_netdev(net, dev) {
5982 				idev = __in6_dev_get(dev);
5983 				if (idev &&
5984 				    idev->cnf.addr_gen_mode != new_val) {
5985 					idev->cnf.addr_gen_mode = new_val;
5986 					addrconf_dev_config(idev->dev);
5987 				}
5988 			}
5989 		}
5990 
5991 		*((u32 *)ctl->data) = new_val;
5992 	}
5993 
5994 out:
5995 	rtnl_unlock();
5996 
5997 	return ret;
5998 }
5999 
addrconf_sysctl_stable_secret(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)6000 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6001 					 void __user *buffer, size_t *lenp,
6002 					 loff_t *ppos)
6003 {
6004 	int err;
6005 	struct in6_addr addr;
6006 	char str[IPV6_MAX_STRLEN];
6007 	struct ctl_table lctl = *ctl;
6008 	struct net *net = ctl->extra2;
6009 	struct ipv6_stable_secret *secret = ctl->data;
6010 
6011 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6012 		return -EIO;
6013 
6014 	lctl.maxlen = IPV6_MAX_STRLEN;
6015 	lctl.data = str;
6016 
6017 	if (!rtnl_trylock())
6018 		return restart_syscall();
6019 
6020 	if (!write && !secret->initialized) {
6021 		err = -EIO;
6022 		goto out;
6023 	}
6024 
6025 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6026 	if (err >= sizeof(str)) {
6027 		err = -EIO;
6028 		goto out;
6029 	}
6030 
6031 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6032 	if (err || !write)
6033 		goto out;
6034 
6035 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6036 		err = -EIO;
6037 		goto out;
6038 	}
6039 
6040 	secret->initialized = true;
6041 	secret->secret = addr;
6042 
6043 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6044 		struct net_device *dev;
6045 
6046 		for_each_netdev(net, dev) {
6047 			struct inet6_dev *idev = __in6_dev_get(dev);
6048 
6049 			if (idev) {
6050 				idev->cnf.addr_gen_mode =
6051 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6052 			}
6053 		}
6054 	} else {
6055 		struct inet6_dev *idev = ctl->extra1;
6056 
6057 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6058 	}
6059 
6060 out:
6061 	rtnl_unlock();
6062 
6063 	return err;
6064 }
6065 
6066 static
addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)6067 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6068 						int write,
6069 						void __user *buffer,
6070 						size_t *lenp,
6071 						loff_t *ppos)
6072 {
6073 	int *valp = ctl->data;
6074 	int val = *valp;
6075 	loff_t pos = *ppos;
6076 	struct ctl_table lctl;
6077 	int ret;
6078 
6079 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6080 	 * we should not modify it until we get the rtnl lock.
6081 	 */
6082 	lctl = *ctl;
6083 	lctl.data = &val;
6084 
6085 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6086 
6087 	if (write)
6088 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6089 	if (ret)
6090 		*ppos = pos;
6091 	return ret;
6092 }
6093 
6094 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6095 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6096 {
6097 	if (rt) {
6098 		if (action)
6099 			rt->dst.flags |= DST_NOPOLICY;
6100 		else
6101 			rt->dst.flags &= ~DST_NOPOLICY;
6102 	}
6103 }
6104 
6105 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6106 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6107 {
6108 	struct inet6_ifaddr *ifa;
6109 
6110 	read_lock_bh(&idev->lock);
6111 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6112 		spin_lock(&ifa->lock);
6113 		if (ifa->rt) {
6114 			struct fib6_info *rt = ifa->rt;
6115 			int cpu;
6116 
6117 			rcu_read_lock();
6118 			ifa->rt->dst_nopolicy = val ? true : false;
6119 			if (rt->rt6i_pcpu) {
6120 				for_each_possible_cpu(cpu) {
6121 					struct rt6_info **rtp;
6122 
6123 					rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
6124 					addrconf_set_nopolicy(*rtp, val);
6125 				}
6126 			}
6127 			rcu_read_unlock();
6128 		}
6129 		spin_unlock(&ifa->lock);
6130 	}
6131 	read_unlock_bh(&idev->lock);
6132 }
6133 
6134 static
addrconf_disable_policy(struct ctl_table * ctl,int * valp,int val)6135 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6136 {
6137 	struct inet6_dev *idev;
6138 	struct net *net;
6139 
6140 	if (!rtnl_trylock())
6141 		return restart_syscall();
6142 
6143 	*valp = val;
6144 
6145 	net = (struct net *)ctl->extra2;
6146 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6147 		rtnl_unlock();
6148 		return 0;
6149 	}
6150 
6151 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6152 		struct net_device *dev;
6153 
6154 		for_each_netdev(net, dev) {
6155 			idev = __in6_dev_get(dev);
6156 			if (idev)
6157 				addrconf_disable_policy_idev(idev, val);
6158 		}
6159 	} else {
6160 		idev = (struct inet6_dev *)ctl->extra1;
6161 		addrconf_disable_policy_idev(idev, val);
6162 	}
6163 
6164 	rtnl_unlock();
6165 	return 0;
6166 }
6167 
6168 static
addrconf_sysctl_disable_policy(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)6169 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6170 				   void __user *buffer, size_t *lenp,
6171 				   loff_t *ppos)
6172 {
6173 	int *valp = ctl->data;
6174 	int val = *valp;
6175 	loff_t pos = *ppos;
6176 	struct ctl_table lctl;
6177 	int ret;
6178 
6179 	lctl = *ctl;
6180 	lctl.data = &val;
6181 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6182 
6183 	if (write && (*valp != val))
6184 		ret = addrconf_disable_policy(ctl, valp, val);
6185 
6186 	if (ret)
6187 		*ppos = pos;
6188 
6189 	return ret;
6190 }
6191 
6192 static int minus_one = -1;
6193 static const int zero = 0;
6194 static const int one = 1;
6195 static const int two_five_five = 255;
6196 
6197 static const struct ctl_table addrconf_sysctl[] = {
6198 	{
6199 		.procname	= "forwarding",
6200 		.data		= &ipv6_devconf.forwarding,
6201 		.maxlen		= sizeof(int),
6202 		.mode		= 0644,
6203 		.proc_handler	= addrconf_sysctl_forward,
6204 	},
6205 	{
6206 		.procname	= "hop_limit",
6207 		.data		= &ipv6_devconf.hop_limit,
6208 		.maxlen		= sizeof(int),
6209 		.mode		= 0644,
6210 		.proc_handler	= proc_dointvec_minmax,
6211 		.extra1		= (void *)&one,
6212 		.extra2		= (void *)&two_five_five,
6213 	},
6214 	{
6215 		.procname	= "mtu",
6216 		.data		= &ipv6_devconf.mtu6,
6217 		.maxlen		= sizeof(int),
6218 		.mode		= 0644,
6219 		.proc_handler	= addrconf_sysctl_mtu,
6220 	},
6221 	{
6222 		.procname	= "accept_ra",
6223 		.data		= &ipv6_devconf.accept_ra,
6224 		.maxlen		= sizeof(int),
6225 		.mode		= 0644,
6226 		.proc_handler	= proc_dointvec,
6227 	},
6228 	{
6229 		.procname	= "accept_redirects",
6230 		.data		= &ipv6_devconf.accept_redirects,
6231 		.maxlen		= sizeof(int),
6232 		.mode		= 0644,
6233 		.proc_handler	= proc_dointvec,
6234 	},
6235 	{
6236 		.procname	= "autoconf",
6237 		.data		= &ipv6_devconf.autoconf,
6238 		.maxlen		= sizeof(int),
6239 		.mode		= 0644,
6240 		.proc_handler	= proc_dointvec,
6241 	},
6242 	{
6243 		.procname	= "dad_transmits",
6244 		.data		= &ipv6_devconf.dad_transmits,
6245 		.maxlen		= sizeof(int),
6246 		.mode		= 0644,
6247 		.proc_handler	= proc_dointvec,
6248 	},
6249 	{
6250 		.procname	= "router_solicitations",
6251 		.data		= &ipv6_devconf.rtr_solicits,
6252 		.maxlen		= sizeof(int),
6253 		.mode		= 0644,
6254 		.proc_handler	= proc_dointvec_minmax,
6255 		.extra1		= &minus_one,
6256 	},
6257 	{
6258 		.procname	= "router_solicitation_interval",
6259 		.data		= &ipv6_devconf.rtr_solicit_interval,
6260 		.maxlen		= sizeof(int),
6261 		.mode		= 0644,
6262 		.proc_handler	= proc_dointvec_jiffies,
6263 	},
6264 	{
6265 		.procname	= "router_solicitation_max_interval",
6266 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6267 		.maxlen		= sizeof(int),
6268 		.mode		= 0644,
6269 		.proc_handler	= proc_dointvec_jiffies,
6270 	},
6271 	{
6272 		.procname	= "router_solicitation_delay",
6273 		.data		= &ipv6_devconf.rtr_solicit_delay,
6274 		.maxlen		= sizeof(int),
6275 		.mode		= 0644,
6276 		.proc_handler	= proc_dointvec_jiffies,
6277 	},
6278 	{
6279 		.procname	= "force_mld_version",
6280 		.data		= &ipv6_devconf.force_mld_version,
6281 		.maxlen		= sizeof(int),
6282 		.mode		= 0644,
6283 		.proc_handler	= proc_dointvec,
6284 	},
6285 	{
6286 		.procname	= "mldv1_unsolicited_report_interval",
6287 		.data		=
6288 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6289 		.maxlen		= sizeof(int),
6290 		.mode		= 0644,
6291 		.proc_handler	= proc_dointvec_ms_jiffies,
6292 	},
6293 	{
6294 		.procname	= "mldv2_unsolicited_report_interval",
6295 		.data		=
6296 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6297 		.maxlen		= sizeof(int),
6298 		.mode		= 0644,
6299 		.proc_handler	= proc_dointvec_ms_jiffies,
6300 	},
6301 	{
6302 		.procname	= "use_tempaddr",
6303 		.data		= &ipv6_devconf.use_tempaddr,
6304 		.maxlen		= sizeof(int),
6305 		.mode		= 0644,
6306 		.proc_handler	= proc_dointvec,
6307 	},
6308 	{
6309 		.procname	= "temp_valid_lft",
6310 		.data		= &ipv6_devconf.temp_valid_lft,
6311 		.maxlen		= sizeof(int),
6312 		.mode		= 0644,
6313 		.proc_handler	= proc_dointvec,
6314 	},
6315 	{
6316 		.procname	= "temp_prefered_lft",
6317 		.data		= &ipv6_devconf.temp_prefered_lft,
6318 		.maxlen		= sizeof(int),
6319 		.mode		= 0644,
6320 		.proc_handler	= proc_dointvec,
6321 	},
6322 	{
6323 		.procname	= "regen_max_retry",
6324 		.data		= &ipv6_devconf.regen_max_retry,
6325 		.maxlen		= sizeof(int),
6326 		.mode		= 0644,
6327 		.proc_handler	= proc_dointvec,
6328 	},
6329 	{
6330 		.procname	= "max_desync_factor",
6331 		.data		= &ipv6_devconf.max_desync_factor,
6332 		.maxlen		= sizeof(int),
6333 		.mode		= 0644,
6334 		.proc_handler	= proc_dointvec,
6335 	},
6336 	{
6337 		.procname	= "max_addresses",
6338 		.data		= &ipv6_devconf.max_addresses,
6339 		.maxlen		= sizeof(int),
6340 		.mode		= 0644,
6341 		.proc_handler	= proc_dointvec,
6342 	},
6343 	{
6344 		.procname	= "accept_ra_defrtr",
6345 		.data		= &ipv6_devconf.accept_ra_defrtr,
6346 		.maxlen		= sizeof(int),
6347 		.mode		= 0644,
6348 		.proc_handler	= proc_dointvec,
6349 	},
6350 	{
6351 		.procname	= "accept_ra_min_hop_limit",
6352 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6353 		.maxlen		= sizeof(int),
6354 		.mode		= 0644,
6355 		.proc_handler	= proc_dointvec,
6356 	},
6357 	{
6358 		.procname	= "accept_ra_pinfo",
6359 		.data		= &ipv6_devconf.accept_ra_pinfo,
6360 		.maxlen		= sizeof(int),
6361 		.mode		= 0644,
6362 		.proc_handler	= proc_dointvec,
6363 	},
6364 #ifdef CONFIG_IPV6_ROUTER_PREF
6365 	{
6366 		.procname	= "accept_ra_rtr_pref",
6367 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6368 		.maxlen		= sizeof(int),
6369 		.mode		= 0644,
6370 		.proc_handler	= proc_dointvec,
6371 	},
6372 	{
6373 		.procname	= "router_probe_interval",
6374 		.data		= &ipv6_devconf.rtr_probe_interval,
6375 		.maxlen		= sizeof(int),
6376 		.mode		= 0644,
6377 		.proc_handler	= proc_dointvec_jiffies,
6378 	},
6379 #ifdef CONFIG_IPV6_ROUTE_INFO
6380 	{
6381 		.procname	= "accept_ra_rt_info_min_plen",
6382 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6383 		.maxlen		= sizeof(int),
6384 		.mode		= 0644,
6385 		.proc_handler	= proc_dointvec,
6386 	},
6387 	{
6388 		.procname	= "accept_ra_rt_info_max_plen",
6389 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6390 		.maxlen		= sizeof(int),
6391 		.mode		= 0644,
6392 		.proc_handler	= proc_dointvec,
6393 	},
6394 #endif
6395 #endif
6396 	{
6397 		.procname	= "proxy_ndp",
6398 		.data		= &ipv6_devconf.proxy_ndp,
6399 		.maxlen		= sizeof(int),
6400 		.mode		= 0644,
6401 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6402 	},
6403 	{
6404 		.procname	= "accept_source_route",
6405 		.data		= &ipv6_devconf.accept_source_route,
6406 		.maxlen		= sizeof(int),
6407 		.mode		= 0644,
6408 		.proc_handler	= proc_dointvec,
6409 	},
6410 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6411 	{
6412 		.procname	= "optimistic_dad",
6413 		.data		= &ipv6_devconf.optimistic_dad,
6414 		.maxlen		= sizeof(int),
6415 		.mode		= 0644,
6416 		.proc_handler   = proc_dointvec,
6417 	},
6418 	{
6419 		.procname	= "use_optimistic",
6420 		.data		= &ipv6_devconf.use_optimistic,
6421 		.maxlen		= sizeof(int),
6422 		.mode		= 0644,
6423 		.proc_handler	= proc_dointvec,
6424 	},
6425 #endif
6426 #ifdef CONFIG_IPV6_MROUTE
6427 	{
6428 		.procname	= "mc_forwarding",
6429 		.data		= &ipv6_devconf.mc_forwarding,
6430 		.maxlen		= sizeof(int),
6431 		.mode		= 0444,
6432 		.proc_handler	= proc_dointvec,
6433 	},
6434 #endif
6435 	{
6436 		.procname	= "disable_ipv6",
6437 		.data		= &ipv6_devconf.disable_ipv6,
6438 		.maxlen		= sizeof(int),
6439 		.mode		= 0644,
6440 		.proc_handler	= addrconf_sysctl_disable,
6441 	},
6442 	{
6443 		.procname	= "accept_dad",
6444 		.data		= &ipv6_devconf.accept_dad,
6445 		.maxlen		= sizeof(int),
6446 		.mode		= 0644,
6447 		.proc_handler	= proc_dointvec,
6448 	},
6449 	{
6450 		.procname	= "force_tllao",
6451 		.data		= &ipv6_devconf.force_tllao,
6452 		.maxlen		= sizeof(int),
6453 		.mode		= 0644,
6454 		.proc_handler	= proc_dointvec
6455 	},
6456 	{
6457 		.procname	= "ndisc_notify",
6458 		.data		= &ipv6_devconf.ndisc_notify,
6459 		.maxlen		= sizeof(int),
6460 		.mode		= 0644,
6461 		.proc_handler	= proc_dointvec
6462 	},
6463 	{
6464 		.procname	= "suppress_frag_ndisc",
6465 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6466 		.maxlen		= sizeof(int),
6467 		.mode		= 0644,
6468 		.proc_handler	= proc_dointvec
6469 	},
6470 	{
6471 		.procname	= "accept_ra_from_local",
6472 		.data		= &ipv6_devconf.accept_ra_from_local,
6473 		.maxlen		= sizeof(int),
6474 		.mode		= 0644,
6475 		.proc_handler	= proc_dointvec,
6476 	},
6477 	{
6478 		.procname	= "accept_ra_mtu",
6479 		.data		= &ipv6_devconf.accept_ra_mtu,
6480 		.maxlen		= sizeof(int),
6481 		.mode		= 0644,
6482 		.proc_handler	= proc_dointvec,
6483 	},
6484 	{
6485 		.procname	= "stable_secret",
6486 		.data		= &ipv6_devconf.stable_secret,
6487 		.maxlen		= IPV6_MAX_STRLEN,
6488 		.mode		= 0600,
6489 		.proc_handler	= addrconf_sysctl_stable_secret,
6490 	},
6491 	{
6492 		.procname	= "use_oif_addrs_only",
6493 		.data		= &ipv6_devconf.use_oif_addrs_only,
6494 		.maxlen		= sizeof(int),
6495 		.mode		= 0644,
6496 		.proc_handler	= proc_dointvec,
6497 	},
6498 	{
6499 		.procname	= "ignore_routes_with_linkdown",
6500 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6501 		.maxlen		= sizeof(int),
6502 		.mode		= 0644,
6503 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6504 	},
6505 	{
6506 		.procname	= "drop_unicast_in_l2_multicast",
6507 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6508 		.maxlen		= sizeof(int),
6509 		.mode		= 0644,
6510 		.proc_handler	= proc_dointvec,
6511 	},
6512 	{
6513 		.procname	= "drop_unsolicited_na",
6514 		.data		= &ipv6_devconf.drop_unsolicited_na,
6515 		.maxlen		= sizeof(int),
6516 		.mode		= 0644,
6517 		.proc_handler	= proc_dointvec,
6518 	},
6519 	{
6520 		.procname	= "keep_addr_on_down",
6521 		.data		= &ipv6_devconf.keep_addr_on_down,
6522 		.maxlen		= sizeof(int),
6523 		.mode		= 0644,
6524 		.proc_handler	= proc_dointvec,
6525 
6526 	},
6527 	{
6528 		.procname	= "seg6_enabled",
6529 		.data		= &ipv6_devconf.seg6_enabled,
6530 		.maxlen		= sizeof(int),
6531 		.mode		= 0644,
6532 		.proc_handler	= proc_dointvec,
6533 	},
6534 #ifdef CONFIG_IPV6_SEG6_HMAC
6535 	{
6536 		.procname	= "seg6_require_hmac",
6537 		.data		= &ipv6_devconf.seg6_require_hmac,
6538 		.maxlen		= sizeof(int),
6539 		.mode		= 0644,
6540 		.proc_handler	= proc_dointvec,
6541 	},
6542 #endif
6543 	{
6544 		.procname       = "enhanced_dad",
6545 		.data           = &ipv6_devconf.enhanced_dad,
6546 		.maxlen         = sizeof(int),
6547 		.mode           = 0644,
6548 		.proc_handler   = proc_dointvec,
6549 	},
6550 	{
6551 		.procname		= "addr_gen_mode",
6552 		.data			= &ipv6_devconf.addr_gen_mode,
6553 		.maxlen			= sizeof(int),
6554 		.mode			= 0644,
6555 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6556 	},
6557 	{
6558 		.procname       = "disable_policy",
6559 		.data           = &ipv6_devconf.disable_policy,
6560 		.maxlen         = sizeof(int),
6561 		.mode           = 0644,
6562 		.proc_handler   = addrconf_sysctl_disable_policy,
6563 	},
6564 	{
6565 		.procname	= "ndisc_tclass",
6566 		.data		= &ipv6_devconf.ndisc_tclass,
6567 		.maxlen		= sizeof(int),
6568 		.mode		= 0644,
6569 		.proc_handler	= proc_dointvec_minmax,
6570 		.extra1		= (void *)&zero,
6571 		.extra2		= (void *)&two_five_five,
6572 	},
6573 	{
6574 		/* sentinel */
6575 	}
6576 };
6577 
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)6578 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6579 		struct inet6_dev *idev, struct ipv6_devconf *p)
6580 {
6581 	int i, ifindex;
6582 	struct ctl_table *table;
6583 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6584 
6585 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6586 	if (!table)
6587 		goto out;
6588 
6589 	for (i = 0; table[i].data; i++) {
6590 		table[i].data += (char *)p - (char *)&ipv6_devconf;
6591 		/* If one of these is already set, then it is not safe to
6592 		 * overwrite either of them: this makes proc_dointvec_minmax
6593 		 * usable.
6594 		 */
6595 		if (!table[i].extra1 && !table[i].extra2) {
6596 			table[i].extra1 = idev; /* embedded; no ref */
6597 			table[i].extra2 = net;
6598 		}
6599 	}
6600 
6601 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6602 
6603 	p->sysctl_header = register_net_sysctl(net, path, table);
6604 	if (!p->sysctl_header)
6605 		goto free;
6606 
6607 	if (!strcmp(dev_name, "all"))
6608 		ifindex = NETCONFA_IFINDEX_ALL;
6609 	else if (!strcmp(dev_name, "default"))
6610 		ifindex = NETCONFA_IFINDEX_DEFAULT;
6611 	else
6612 		ifindex = idev->dev->ifindex;
6613 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6614 				     ifindex, p);
6615 	return 0;
6616 
6617 free:
6618 	kfree(table);
6619 out:
6620 	return -ENOBUFS;
6621 }
6622 
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)6623 static void __addrconf_sysctl_unregister(struct net *net,
6624 					 struct ipv6_devconf *p, int ifindex)
6625 {
6626 	struct ctl_table *table;
6627 
6628 	if (!p->sysctl_header)
6629 		return;
6630 
6631 	table = p->sysctl_header->ctl_table_arg;
6632 	unregister_net_sysctl_table(p->sysctl_header);
6633 	p->sysctl_header = NULL;
6634 	kfree(table);
6635 
6636 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6637 }
6638 
addrconf_sysctl_register(struct inet6_dev * idev)6639 static int addrconf_sysctl_register(struct inet6_dev *idev)
6640 {
6641 	int err;
6642 
6643 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
6644 		return -EINVAL;
6645 
6646 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6647 				    &ndisc_ifinfo_sysctl_change);
6648 	if (err)
6649 		return err;
6650 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6651 					 idev, &idev->cnf);
6652 	if (err)
6653 		neigh_sysctl_unregister(idev->nd_parms);
6654 
6655 	return err;
6656 }
6657 
addrconf_sysctl_unregister(struct inet6_dev * idev)6658 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6659 {
6660 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6661 				     idev->dev->ifindex);
6662 	neigh_sysctl_unregister(idev->nd_parms);
6663 }
6664 
6665 
6666 #endif
6667 
addrconf_init_net(struct net * net)6668 static int __net_init addrconf_init_net(struct net *net)
6669 {
6670 	int err = -ENOMEM;
6671 	struct ipv6_devconf *all, *dflt;
6672 
6673 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6674 	if (!all)
6675 		goto err_alloc_all;
6676 
6677 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6678 	if (!dflt)
6679 		goto err_alloc_dflt;
6680 
6681 	/* these will be inherited by all namespaces */
6682 	dflt->autoconf = ipv6_defaults.autoconf;
6683 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6684 
6685 	dflt->stable_secret.initialized = false;
6686 	all->stable_secret.initialized = false;
6687 
6688 	net->ipv6.devconf_all = all;
6689 	net->ipv6.devconf_dflt = dflt;
6690 
6691 #ifdef CONFIG_SYSCTL
6692 	err = __addrconf_sysctl_register(net, "all", NULL, all);
6693 	if (err < 0)
6694 		goto err_reg_all;
6695 
6696 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6697 	if (err < 0)
6698 		goto err_reg_dflt;
6699 #endif
6700 	return 0;
6701 
6702 #ifdef CONFIG_SYSCTL
6703 err_reg_dflt:
6704 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6705 err_reg_all:
6706 	kfree(dflt);
6707 #endif
6708 err_alloc_dflt:
6709 	kfree(all);
6710 err_alloc_all:
6711 	return err;
6712 }
6713 
addrconf_exit_net(struct net * net)6714 static void __net_exit addrconf_exit_net(struct net *net)
6715 {
6716 #ifdef CONFIG_SYSCTL
6717 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6718 				     NETCONFA_IFINDEX_DEFAULT);
6719 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6720 				     NETCONFA_IFINDEX_ALL);
6721 #endif
6722 	kfree(net->ipv6.devconf_dflt);
6723 	kfree(net->ipv6.devconf_all);
6724 }
6725 
6726 static struct pernet_operations addrconf_ops = {
6727 	.init = addrconf_init_net,
6728 	.exit = addrconf_exit_net,
6729 };
6730 
6731 static struct rtnl_af_ops inet6_ops __read_mostly = {
6732 	.family		  = AF_INET6,
6733 	.fill_link_af	  = inet6_fill_link_af,
6734 	.get_link_af_size = inet6_get_link_af_size,
6735 	.validate_link_af = inet6_validate_link_af,
6736 	.set_link_af	  = inet6_set_link_af,
6737 };
6738 
6739 /*
6740  *	Init / cleanup code
6741  */
6742 
addrconf_init(void)6743 int __init addrconf_init(void)
6744 {
6745 	struct inet6_dev *idev;
6746 	int i, err;
6747 
6748 	err = ipv6_addr_label_init();
6749 	if (err < 0) {
6750 		pr_crit("%s: cannot initialize default policy table: %d\n",
6751 			__func__, err);
6752 		goto out;
6753 	}
6754 
6755 	err = register_pernet_subsys(&addrconf_ops);
6756 	if (err < 0)
6757 		goto out_addrlabel;
6758 
6759 	addrconf_wq = create_workqueue("ipv6_addrconf");
6760 	if (!addrconf_wq) {
6761 		err = -ENOMEM;
6762 		goto out_nowq;
6763 	}
6764 
6765 	/* The addrconf netdev notifier requires that loopback_dev
6766 	 * has it's ipv6 private information allocated and setup
6767 	 * before it can bring up and give link-local addresses
6768 	 * to other devices which are up.
6769 	 *
6770 	 * Unfortunately, loopback_dev is not necessarily the first
6771 	 * entry in the global dev_base list of net devices.  In fact,
6772 	 * it is likely to be the very last entry on that list.
6773 	 * So this causes the notifier registry below to try and
6774 	 * give link-local addresses to all devices besides loopback_dev
6775 	 * first, then loopback_dev, which cases all the non-loopback_dev
6776 	 * devices to fail to get a link-local address.
6777 	 *
6778 	 * So, as a temporary fix, allocate the ipv6 structure for
6779 	 * loopback_dev first by hand.
6780 	 * Longer term, all of the dependencies ipv6 has upon the loopback
6781 	 * device and it being up should be removed.
6782 	 */
6783 	rtnl_lock();
6784 	idev = ipv6_add_dev(init_net.loopback_dev);
6785 	rtnl_unlock();
6786 	if (IS_ERR(idev)) {
6787 		err = PTR_ERR(idev);
6788 		goto errlo;
6789 	}
6790 
6791 	ip6_route_init_special_entries();
6792 
6793 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
6794 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6795 
6796 	register_netdevice_notifier(&ipv6_dev_notf);
6797 
6798 	addrconf_verify();
6799 
6800 	rtnl_af_register(&inet6_ops);
6801 
6802 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6803 				   NULL, inet6_dump_ifinfo, 0);
6804 	if (err < 0)
6805 		goto errout;
6806 
6807 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6808 				   inet6_rtm_newaddr, NULL, 0);
6809 	if (err < 0)
6810 		goto errout;
6811 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6812 				   inet6_rtm_deladdr, NULL, 0);
6813 	if (err < 0)
6814 		goto errout;
6815 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6816 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
6817 				   RTNL_FLAG_DOIT_UNLOCKED);
6818 	if (err < 0)
6819 		goto errout;
6820 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6821 				   NULL, inet6_dump_ifmcaddr, 0);
6822 	if (err < 0)
6823 		goto errout;
6824 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6825 				   NULL, inet6_dump_ifacaddr, 0);
6826 	if (err < 0)
6827 		goto errout;
6828 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6829 				   inet6_netconf_get_devconf,
6830 				   inet6_netconf_dump_devconf,
6831 				   RTNL_FLAG_DOIT_UNLOCKED);
6832 	if (err < 0)
6833 		goto errout;
6834 	err = ipv6_addr_label_rtnl_register();
6835 	if (err < 0)
6836 		goto errout;
6837 
6838 	return 0;
6839 errout:
6840 	rtnl_unregister_all(PF_INET6);
6841 	rtnl_af_unregister(&inet6_ops);
6842 	unregister_netdevice_notifier(&ipv6_dev_notf);
6843 errlo:
6844 	destroy_workqueue(addrconf_wq);
6845 out_nowq:
6846 	unregister_pernet_subsys(&addrconf_ops);
6847 out_addrlabel:
6848 	ipv6_addr_label_cleanup();
6849 out:
6850 	return err;
6851 }
6852 
addrconf_cleanup(void)6853 void addrconf_cleanup(void)
6854 {
6855 	struct net_device *dev;
6856 	int i;
6857 
6858 	unregister_netdevice_notifier(&ipv6_dev_notf);
6859 	unregister_pernet_subsys(&addrconf_ops);
6860 	ipv6_addr_label_cleanup();
6861 
6862 	rtnl_af_unregister(&inet6_ops);
6863 
6864 	rtnl_lock();
6865 
6866 	/* clean dev list */
6867 	for_each_netdev(&init_net, dev) {
6868 		if (__in6_dev_get(dev) == NULL)
6869 			continue;
6870 		addrconf_ifdown(dev, 1);
6871 	}
6872 	addrconf_ifdown(init_net.loopback_dev, 2);
6873 
6874 	/*
6875 	 *	Check hash table.
6876 	 */
6877 	spin_lock_bh(&addrconf_hash_lock);
6878 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
6879 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6880 	spin_unlock_bh(&addrconf_hash_lock);
6881 	cancel_delayed_work(&addr_chk_work);
6882 	rtnl_unlock();
6883 
6884 	destroy_workqueue(addrconf_wq);
6885 }
6886