• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 Address [auto]configuration
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  */
10 
11 /*
12  *	Changes:
13  *
14  *	Janos Farkas			:	delete timer on ifdown
15  *	<chexum@bankinf.banki.hu>
16  *	Andi Kleen			:	kill double kfree on module
17  *						unload.
18  *	Maciej W. Rozycki		:	FDDI support
19  *	sekiya@USAGI			:	Don't send too many RS
20  *						packets.
21  *	yoshfuji@USAGI			:       Fixed interval between DAD
22  *						packets.
23  *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
24  *						address validation timer.
25  *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
26  *						support.
27  *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
28  *						address on a same interface.
29  *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
30  *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
31  *						seq_file.
32  *	YOSHIFUJI Hideaki @USAGI	:	improved source address
33  *						selection; consider scope,
34  *						status etc.
35  */
36 
37 #define pr_fmt(fmt) "IPv6: " fmt
38 
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65 
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
69 
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
88 
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
92 
93 #define	INFINITY_LIFE_TIME	0xFFFFFFFF
94 
95 #define IPV6_MAX_STRLEN \
96 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
97 
cstamp_delta(unsigned long cstamp)98 static inline u32 cstamp_delta(unsigned long cstamp)
99 {
100 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
101 }
102 
rfc3315_s14_backoff_init(s32 irt)103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
104 {
105 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
106 	u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
107 	do_div(tmp, 1000000);
108 	return (s32)tmp;
109 }
110 
rfc3315_s14_backoff_update(s32 rt,s32 mrt)111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
112 {
113 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
114 	u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
115 	do_div(tmp, 1000000);
116 	if ((s32)tmp > mrt) {
117 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
118 		tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
119 		do_div(tmp, 1000000);
120 	}
121 	return (s32)tmp;
122 }
123 
124 #ifdef CONFIG_SYSCTL
125 static int addrconf_sysctl_register(struct inet6_dev *idev);
126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
127 #else
addrconf_sysctl_register(struct inet6_dev * idev)128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
129 {
130 	return 0;
131 }
132 
addrconf_sysctl_unregister(struct inet6_dev * idev)133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
134 {
135 }
136 #endif
137 
138 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
139 
140 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
141 static int ipv6_count_addresses(const struct inet6_dev *idev);
142 static int ipv6_generate_stable_address(struct in6_addr *addr,
143 					u8 dad_count,
144 					const struct inet6_dev *idev);
145 
146 #define IN6_ADDR_HSIZE_SHIFT	8
147 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
148 /*
149  *	Configured unicast address hash table
150  */
151 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
152 static DEFINE_SPINLOCK(addrconf_hash_lock);
153 
154 static void addrconf_verify(void);
155 static void addrconf_verify_rtnl(void);
156 static void addrconf_verify_work(struct work_struct *);
157 
158 static struct workqueue_struct *addrconf_wq;
159 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
160 
161 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
162 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
163 
164 static void addrconf_type_change(struct net_device *dev,
165 				 unsigned long event);
166 static int addrconf_ifdown(struct net_device *dev, bool unregister);
167 
168 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
169 						  int plen,
170 						  const struct net_device *dev,
171 						  u32 flags, u32 noflags,
172 						  bool no_gw);
173 
174 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
175 static void addrconf_dad_work(struct work_struct *w);
176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
177 				   bool send_na);
178 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
179 static void addrconf_rs_timer(struct timer_list *t);
180 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
181 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
182 
183 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
184 				struct prefix_info *pinfo);
185 
186 static struct ipv6_devconf ipv6_devconf __read_mostly = {
187 	.forwarding		= 0,
188 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
189 	.mtu6			= IPV6_MIN_MTU,
190 	.accept_ra		= 1,
191 	.accept_redirects	= 1,
192 	.autoconf		= 1,
193 	.force_mld_version	= 0,
194 	.mldv1_unsolicited_report_interval = 10 * HZ,
195 	.mldv2_unsolicited_report_interval = HZ,
196 	.dad_transmits		= 1,
197 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
198 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
199 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
200 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
201 	.use_tempaddr		= 0,
202 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
203 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
204 	.regen_max_retry	= REGEN_MAX_RETRY,
205 	.max_desync_factor	= MAX_DESYNC_FACTOR,
206 	.max_addresses		= IPV6_MAX_ADDRESSES,
207 	.accept_ra_defrtr	= 1,
208 	.accept_ra_from_local	= 0,
209 	.accept_ra_min_hop_limit= 1,
210 	.accept_ra_pinfo	= 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212 	.accept_ra_rtr_pref	= 1,
213 	.rtr_probe_interval	= 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215 	.accept_ra_rt_info_min_plen = 0,
216 	.accept_ra_rt_info_max_plen = 0,
217 #endif
218 #endif
219 	.proxy_ndp		= 0,
220 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
221 	.disable_ipv6		= 0,
222 	.accept_dad		= 0,
223 	.suppress_frag_ndisc	= 1,
224 	.accept_ra_mtu		= 1,
225 	.stable_secret		= {
226 		.initialized = false,
227 	},
228 	.use_oif_addrs_only	= 0,
229 	.ignore_routes_with_linkdown = 0,
230 	.keep_addr_on_down	= 0,
231 	.seg6_enabled		= 0,
232 #ifdef CONFIG_IPV6_SEG6_HMAC
233 	.seg6_require_hmac	= 0,
234 #endif
235 	.enhanced_dad           = 1,
236 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
237 	.disable_policy		= 0,
238 	.rpl_seg_enabled	= 0,
239 };
240 
241 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
242 	.forwarding		= 0,
243 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
244 	.mtu6			= IPV6_MIN_MTU,
245 	.accept_ra		= 1,
246 	.accept_redirects	= 1,
247 	.autoconf		= 1,
248 	.force_mld_version	= 0,
249 	.mldv1_unsolicited_report_interval = 10 * HZ,
250 	.mldv2_unsolicited_report_interval = HZ,
251 	.dad_transmits		= 1,
252 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
253 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
254 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
255 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
256 	.use_tempaddr		= 0,
257 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
258 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
259 	.regen_max_retry	= REGEN_MAX_RETRY,
260 	.max_desync_factor	= MAX_DESYNC_FACTOR,
261 	.max_addresses		= IPV6_MAX_ADDRESSES,
262 	.accept_ra_defrtr	= 1,
263 	.accept_ra_from_local	= 0,
264 	.accept_ra_min_hop_limit= 1,
265 	.accept_ra_pinfo	= 1,
266 #ifdef CONFIG_IPV6_ROUTER_PREF
267 	.accept_ra_rtr_pref	= 1,
268 	.rtr_probe_interval	= 60 * HZ,
269 #ifdef CONFIG_IPV6_ROUTE_INFO
270 	.accept_ra_rt_info_min_plen = 0,
271 	.accept_ra_rt_info_max_plen = 0,
272 #endif
273 #endif
274 	.proxy_ndp		= 0,
275 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
276 	.disable_ipv6		= 0,
277 	.accept_dad		= 1,
278 	.suppress_frag_ndisc	= 1,
279 	.accept_ra_mtu		= 1,
280 	.stable_secret		= {
281 		.initialized = false,
282 	},
283 	.use_oif_addrs_only	= 0,
284 	.ignore_routes_with_linkdown = 0,
285 	.keep_addr_on_down	= 0,
286 	.seg6_enabled		= 0,
287 #ifdef CONFIG_IPV6_SEG6_HMAC
288 	.seg6_require_hmac	= 0,
289 #endif
290 	.enhanced_dad           = 1,
291 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
292 	.disable_policy		= 0,
293 	.rpl_seg_enabled	= 0,
294 };
295 
296 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device * dev)297 static inline bool addrconf_link_ready(const struct net_device *dev)
298 {
299 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
300 }
301 
addrconf_del_rs_timer(struct inet6_dev * idev)302 static void addrconf_del_rs_timer(struct inet6_dev *idev)
303 {
304 	if (del_timer(&idev->rs_timer))
305 		__in6_dev_put(idev);
306 }
307 
addrconf_del_dad_work(struct inet6_ifaddr * ifp)308 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
309 {
310 	if (cancel_delayed_work(&ifp->dad_work))
311 		__in6_ifa_put(ifp);
312 }
313 
addrconf_mod_rs_timer(struct inet6_dev * idev,unsigned long when)314 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
315 				  unsigned long when)
316 {
317 	if (!timer_pending(&idev->rs_timer))
318 		in6_dev_hold(idev);
319 	mod_timer(&idev->rs_timer, jiffies + when);
320 }
321 
addrconf_mod_dad_work(struct inet6_ifaddr * ifp,unsigned long delay)322 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
323 				   unsigned long delay)
324 {
325 	in6_ifa_hold(ifp);
326 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
327 		in6_ifa_put(ifp);
328 }
329 
snmp6_alloc_dev(struct inet6_dev * idev)330 static int snmp6_alloc_dev(struct inet6_dev *idev)
331 {
332 	int i;
333 
334 	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
335 	if (!idev->stats.ipv6)
336 		goto err_ip;
337 
338 	for_each_possible_cpu(i) {
339 		struct ipstats_mib *addrconf_stats;
340 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
341 		u64_stats_init(&addrconf_stats->syncp);
342 	}
343 
344 
345 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
346 					GFP_KERNEL);
347 	if (!idev->stats.icmpv6dev)
348 		goto err_icmp;
349 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
350 					   GFP_KERNEL);
351 	if (!idev->stats.icmpv6msgdev)
352 		goto err_icmpmsg;
353 
354 	return 0;
355 
356 err_icmpmsg:
357 	kfree(idev->stats.icmpv6dev);
358 err_icmp:
359 	free_percpu(idev->stats.ipv6);
360 err_ip:
361 	return -ENOMEM;
362 }
363 
ipv6_add_dev(struct net_device * dev)364 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
365 {
366 	struct inet6_dev *ndev;
367 	int err = -ENOMEM;
368 
369 	ASSERT_RTNL();
370 
371 	if (dev->mtu < IPV6_MIN_MTU)
372 		return ERR_PTR(-EINVAL);
373 
374 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
375 	if (!ndev)
376 		return ERR_PTR(err);
377 
378 	rwlock_init(&ndev->lock);
379 	ndev->dev = dev;
380 	INIT_LIST_HEAD(&ndev->addr_list);
381 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
382 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
383 
384 	if (ndev->cnf.stable_secret.initialized)
385 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
386 
387 	ndev->cnf.mtu6 = dev->mtu;
388 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
389 	if (!ndev->nd_parms) {
390 		kfree(ndev);
391 		return ERR_PTR(err);
392 	}
393 	if (ndev->cnf.forwarding)
394 		dev_disable_lro(dev);
395 	/* We refer to the device */
396 	dev_hold(dev);
397 
398 	if (snmp6_alloc_dev(ndev) < 0) {
399 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
400 			   __func__);
401 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
402 		dev_put(dev);
403 		kfree(ndev);
404 		return ERR_PTR(err);
405 	}
406 
407 	if (snmp6_register_dev(ndev) < 0) {
408 		netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
409 			   __func__, dev->name);
410 		goto err_release;
411 	}
412 
413 	/* One reference from device. */
414 	refcount_set(&ndev->refcnt, 1);
415 
416 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
417 		ndev->cnf.accept_dad = -1;
418 
419 #if IS_ENABLED(CONFIG_IPV6_SIT)
420 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
421 		pr_info("%s: Disabled Multicast RS\n", dev->name);
422 		ndev->cnf.rtr_solicits = 0;
423 	}
424 #endif
425 
426 	INIT_LIST_HEAD(&ndev->tempaddr_list);
427 	ndev->desync_factor = U32_MAX;
428 	if ((dev->flags&IFF_LOOPBACK) ||
429 	    dev->type == ARPHRD_TUNNEL ||
430 	    dev->type == ARPHRD_TUNNEL6 ||
431 	    dev->type == ARPHRD_SIT ||
432 	    dev->type == ARPHRD_NONE) {
433 		ndev->cnf.use_tempaddr = -1;
434 	}
435 
436 	ndev->token = in6addr_any;
437 
438 	if (netif_running(dev) && addrconf_link_ready(dev))
439 		ndev->if_flags |= IF_READY;
440 
441 	ipv6_mc_init_dev(ndev);
442 	ndev->tstamp = jiffies;
443 	err = addrconf_sysctl_register(ndev);
444 	if (err) {
445 		ipv6_mc_destroy_dev(ndev);
446 		snmp6_unregister_dev(ndev);
447 		goto err_release;
448 	}
449 	/* protected by rtnl_lock */
450 	rcu_assign_pointer(dev->ip6_ptr, ndev);
451 
452 	/* Join interface-local all-node multicast group */
453 	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
454 
455 	/* Join all-node multicast group */
456 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
457 
458 	/* Join all-router multicast group if forwarding is set */
459 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
460 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
461 
462 	return ndev;
463 
464 err_release:
465 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
466 	ndev->dead = 1;
467 	in6_dev_finish_destroy(ndev);
468 	return ERR_PTR(err);
469 }
470 
ipv6_find_idev(struct net_device * dev)471 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
472 {
473 	struct inet6_dev *idev;
474 
475 	ASSERT_RTNL();
476 
477 	idev = __in6_dev_get(dev);
478 	if (!idev) {
479 		idev = ipv6_add_dev(dev);
480 		if (IS_ERR(idev))
481 			return idev;
482 	}
483 
484 	if (dev->flags&IFF_UP)
485 		ipv6_mc_up(idev);
486 	return idev;
487 }
488 
inet6_netconf_msgsize_devconf(int type)489 static int inet6_netconf_msgsize_devconf(int type)
490 {
491 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
492 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
493 	bool all = false;
494 
495 	if (type == NETCONFA_ALL)
496 		all = true;
497 
498 	if (all || type == NETCONFA_FORWARDING)
499 		size += nla_total_size(4);
500 #ifdef CONFIG_IPV6_MROUTE
501 	if (all || type == NETCONFA_MC_FORWARDING)
502 		size += nla_total_size(4);
503 #endif
504 	if (all || type == NETCONFA_PROXY_NEIGH)
505 		size += nla_total_size(4);
506 
507 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
508 		size += nla_total_size(4);
509 
510 	return size;
511 }
512 
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)513 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
514 				      struct ipv6_devconf *devconf, u32 portid,
515 				      u32 seq, int event, unsigned int flags,
516 				      int type)
517 {
518 	struct nlmsghdr  *nlh;
519 	struct netconfmsg *ncm;
520 	bool all = false;
521 
522 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
523 			flags);
524 	if (!nlh)
525 		return -EMSGSIZE;
526 
527 	if (type == NETCONFA_ALL)
528 		all = true;
529 
530 	ncm = nlmsg_data(nlh);
531 	ncm->ncm_family = AF_INET6;
532 
533 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
534 		goto nla_put_failure;
535 
536 	if (!devconf)
537 		goto out;
538 
539 	if ((all || type == NETCONFA_FORWARDING) &&
540 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
541 		goto nla_put_failure;
542 #ifdef CONFIG_IPV6_MROUTE
543 	if ((all || type == NETCONFA_MC_FORWARDING) &&
544 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
545 			devconf->mc_forwarding) < 0)
546 		goto nla_put_failure;
547 #endif
548 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
549 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
550 		goto nla_put_failure;
551 
552 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
553 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
554 			devconf->ignore_routes_with_linkdown) < 0)
555 		goto nla_put_failure;
556 
557 out:
558 	nlmsg_end(skb, nlh);
559 	return 0;
560 
561 nla_put_failure:
562 	nlmsg_cancel(skb, nlh);
563 	return -EMSGSIZE;
564 }
565 
inet6_netconf_notify_devconf(struct net * net,int event,int type,int ifindex,struct ipv6_devconf * devconf)566 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
567 				  int ifindex, struct ipv6_devconf *devconf)
568 {
569 	struct sk_buff *skb;
570 	int err = -ENOBUFS;
571 
572 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
573 	if (!skb)
574 		goto errout;
575 
576 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
577 					 event, 0, type);
578 	if (err < 0) {
579 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
580 		WARN_ON(err == -EMSGSIZE);
581 		kfree_skb(skb);
582 		goto errout;
583 	}
584 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
585 	return;
586 errout:
587 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
588 }
589 
590 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
591 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
592 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
593 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
594 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
595 };
596 
inet6_netconf_valid_get_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)597 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
598 				       const struct nlmsghdr *nlh,
599 				       struct nlattr **tb,
600 				       struct netlink_ext_ack *extack)
601 {
602 	int i, err;
603 
604 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
605 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
606 		return -EINVAL;
607 	}
608 
609 	if (!netlink_strict_get_check(skb))
610 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
611 					      tb, NETCONFA_MAX,
612 					      devconf_ipv6_policy, extack);
613 
614 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
615 					    tb, NETCONFA_MAX,
616 					    devconf_ipv6_policy, extack);
617 	if (err)
618 		return err;
619 
620 	for (i = 0; i <= NETCONFA_MAX; i++) {
621 		if (!tb[i])
622 			continue;
623 
624 		switch (i) {
625 		case NETCONFA_IFINDEX:
626 			break;
627 		default:
628 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
629 			return -EINVAL;
630 		}
631 	}
632 
633 	return 0;
634 }
635 
inet6_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)636 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
637 				     struct nlmsghdr *nlh,
638 				     struct netlink_ext_ack *extack)
639 {
640 	struct net *net = sock_net(in_skb->sk);
641 	struct nlattr *tb[NETCONFA_MAX+1];
642 	struct inet6_dev *in6_dev = NULL;
643 	struct net_device *dev = NULL;
644 	struct sk_buff *skb;
645 	struct ipv6_devconf *devconf;
646 	int ifindex;
647 	int err;
648 
649 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
650 	if (err < 0)
651 		return err;
652 
653 	if (!tb[NETCONFA_IFINDEX])
654 		return -EINVAL;
655 
656 	err = -EINVAL;
657 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
658 	switch (ifindex) {
659 	case NETCONFA_IFINDEX_ALL:
660 		devconf = net->ipv6.devconf_all;
661 		break;
662 	case NETCONFA_IFINDEX_DEFAULT:
663 		devconf = net->ipv6.devconf_dflt;
664 		break;
665 	default:
666 		dev = dev_get_by_index(net, ifindex);
667 		if (!dev)
668 			return -EINVAL;
669 		in6_dev = in6_dev_get(dev);
670 		if (!in6_dev)
671 			goto errout;
672 		devconf = &in6_dev->cnf;
673 		break;
674 	}
675 
676 	err = -ENOBUFS;
677 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
678 	if (!skb)
679 		goto errout;
680 
681 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
682 					 NETLINK_CB(in_skb).portid,
683 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
684 					 NETCONFA_ALL);
685 	if (err < 0) {
686 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
687 		WARN_ON(err == -EMSGSIZE);
688 		kfree_skb(skb);
689 		goto errout;
690 	}
691 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
692 errout:
693 	if (in6_dev)
694 		in6_dev_put(in6_dev);
695 	if (dev)
696 		dev_put(dev);
697 	return err;
698 }
699 
inet6_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)700 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
701 				      struct netlink_callback *cb)
702 {
703 	const struct nlmsghdr *nlh = cb->nlh;
704 	struct net *net = sock_net(skb->sk);
705 	int h, s_h;
706 	int idx, s_idx;
707 	struct net_device *dev;
708 	struct inet6_dev *idev;
709 	struct hlist_head *head;
710 
711 	if (cb->strict_check) {
712 		struct netlink_ext_ack *extack = cb->extack;
713 		struct netconfmsg *ncm;
714 
715 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
716 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
717 			return -EINVAL;
718 		}
719 
720 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
721 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
722 			return -EINVAL;
723 		}
724 	}
725 
726 	s_h = cb->args[0];
727 	s_idx = idx = cb->args[1];
728 
729 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
730 		idx = 0;
731 		head = &net->dev_index_head[h];
732 		rcu_read_lock();
733 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
734 			  net->dev_base_seq;
735 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
736 			if (idx < s_idx)
737 				goto cont;
738 			idev = __in6_dev_get(dev);
739 			if (!idev)
740 				goto cont;
741 
742 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
743 						       &idev->cnf,
744 						       NETLINK_CB(cb->skb).portid,
745 						       nlh->nlmsg_seq,
746 						       RTM_NEWNETCONF,
747 						       NLM_F_MULTI,
748 						       NETCONFA_ALL) < 0) {
749 				rcu_read_unlock();
750 				goto done;
751 			}
752 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
753 cont:
754 			idx++;
755 		}
756 		rcu_read_unlock();
757 	}
758 	if (h == NETDEV_HASHENTRIES) {
759 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
760 					       net->ipv6.devconf_all,
761 					       NETLINK_CB(cb->skb).portid,
762 					       nlh->nlmsg_seq,
763 					       RTM_NEWNETCONF, NLM_F_MULTI,
764 					       NETCONFA_ALL) < 0)
765 			goto done;
766 		else
767 			h++;
768 	}
769 	if (h == NETDEV_HASHENTRIES + 1) {
770 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
771 					       net->ipv6.devconf_dflt,
772 					       NETLINK_CB(cb->skb).portid,
773 					       nlh->nlmsg_seq,
774 					       RTM_NEWNETCONF, NLM_F_MULTI,
775 					       NETCONFA_ALL) < 0)
776 			goto done;
777 		else
778 			h++;
779 	}
780 done:
781 	cb->args[0] = h;
782 	cb->args[1] = idx;
783 
784 	return skb->len;
785 }
786 
787 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev * idev)788 static void dev_forward_change(struct inet6_dev *idev)
789 {
790 	struct net_device *dev;
791 	struct inet6_ifaddr *ifa;
792 
793 	if (!idev)
794 		return;
795 	dev = idev->dev;
796 	if (idev->cnf.forwarding)
797 		dev_disable_lro(dev);
798 	if (dev->flags & IFF_MULTICAST) {
799 		if (idev->cnf.forwarding) {
800 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
801 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
802 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
803 		} else {
804 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
805 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
806 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
807 		}
808 	}
809 
810 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
811 		if (ifa->flags&IFA_F_TENTATIVE)
812 			continue;
813 		if (idev->cnf.forwarding)
814 			addrconf_join_anycast(ifa);
815 		else
816 			addrconf_leave_anycast(ifa);
817 	}
818 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
819 				     NETCONFA_FORWARDING,
820 				     dev->ifindex, &idev->cnf);
821 }
822 
823 
addrconf_forward_change(struct net * net,__s32 newf)824 static void addrconf_forward_change(struct net *net, __s32 newf)
825 {
826 	struct net_device *dev;
827 	struct inet6_dev *idev;
828 
829 	for_each_netdev(net, dev) {
830 		idev = __in6_dev_get(dev);
831 		if (idev) {
832 			int changed = (!idev->cnf.forwarding) ^ (!newf);
833 			idev->cnf.forwarding = newf;
834 			if (changed)
835 				dev_forward_change(idev);
836 		}
837 	}
838 }
839 
addrconf_fixup_forwarding(struct ctl_table * table,int * p,int newf)840 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
841 {
842 	struct net *net;
843 	int old;
844 
845 	if (!rtnl_trylock())
846 		return restart_syscall();
847 
848 	net = (struct net *)table->extra2;
849 	old = *p;
850 	*p = newf;
851 
852 	if (p == &net->ipv6.devconf_dflt->forwarding) {
853 		if ((!newf) ^ (!old))
854 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
855 						     NETCONFA_FORWARDING,
856 						     NETCONFA_IFINDEX_DEFAULT,
857 						     net->ipv6.devconf_dflt);
858 		rtnl_unlock();
859 		return 0;
860 	}
861 
862 	if (p == &net->ipv6.devconf_all->forwarding) {
863 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
864 
865 		net->ipv6.devconf_dflt->forwarding = newf;
866 		if ((!newf) ^ (!old_dflt))
867 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
868 						     NETCONFA_FORWARDING,
869 						     NETCONFA_IFINDEX_DEFAULT,
870 						     net->ipv6.devconf_dflt);
871 
872 		addrconf_forward_change(net, newf);
873 		if ((!newf) ^ (!old))
874 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
875 						     NETCONFA_FORWARDING,
876 						     NETCONFA_IFINDEX_ALL,
877 						     net->ipv6.devconf_all);
878 	} else if ((!newf) ^ (!old))
879 		dev_forward_change((struct inet6_dev *)table->extra1);
880 	rtnl_unlock();
881 
882 	if (newf)
883 		rt6_purge_dflt_routers(net);
884 	return 1;
885 }
886 
addrconf_linkdown_change(struct net * net,__s32 newf)887 static void addrconf_linkdown_change(struct net *net, __s32 newf)
888 {
889 	struct net_device *dev;
890 	struct inet6_dev *idev;
891 
892 	for_each_netdev(net, dev) {
893 		idev = __in6_dev_get(dev);
894 		if (idev) {
895 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
896 
897 			idev->cnf.ignore_routes_with_linkdown = newf;
898 			if (changed)
899 				inet6_netconf_notify_devconf(dev_net(dev),
900 							     RTM_NEWNETCONF,
901 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
902 							     dev->ifindex,
903 							     &idev->cnf);
904 		}
905 	}
906 }
907 
addrconf_fixup_linkdown(struct ctl_table * table,int * p,int newf)908 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
909 {
910 	struct net *net;
911 	int old;
912 
913 	if (!rtnl_trylock())
914 		return restart_syscall();
915 
916 	net = (struct net *)table->extra2;
917 	old = *p;
918 	*p = newf;
919 
920 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
921 		if ((!newf) ^ (!old))
922 			inet6_netconf_notify_devconf(net,
923 						     RTM_NEWNETCONF,
924 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
925 						     NETCONFA_IFINDEX_DEFAULT,
926 						     net->ipv6.devconf_dflt);
927 		rtnl_unlock();
928 		return 0;
929 	}
930 
931 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
932 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
933 		addrconf_linkdown_change(net, newf);
934 		if ((!newf) ^ (!old))
935 			inet6_netconf_notify_devconf(net,
936 						     RTM_NEWNETCONF,
937 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
938 						     NETCONFA_IFINDEX_ALL,
939 						     net->ipv6.devconf_all);
940 	}
941 	rtnl_unlock();
942 
943 	return 1;
944 }
945 
946 #endif
947 
948 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)949 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
950 {
951 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
952 
953 #ifdef NET_REFCNT_DEBUG
954 	pr_debug("%s\n", __func__);
955 #endif
956 
957 	in6_dev_put(ifp->idev);
958 
959 	if (cancel_delayed_work(&ifp->dad_work))
960 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
961 			  ifp);
962 
963 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
964 		pr_warn("Freeing alive inet6 address %p\n", ifp);
965 		return;
966 	}
967 
968 	kfree_rcu(ifp, rcu);
969 }
970 
971 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)972 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
973 {
974 	struct list_head *p;
975 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
976 
977 	/*
978 	 * Each device address list is sorted in order of scope -
979 	 * global before linklocal.
980 	 */
981 	list_for_each(p, &idev->addr_list) {
982 		struct inet6_ifaddr *ifa
983 			= list_entry(p, struct inet6_ifaddr, if_list);
984 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
985 			break;
986 	}
987 
988 	list_add_tail_rcu(&ifp->if_list, p);
989 }
990 
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)991 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
992 {
993 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
994 
995 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
996 }
997 
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)998 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
999 			       struct net_device *dev, unsigned int hash)
1000 {
1001 	struct inet6_ifaddr *ifp;
1002 
1003 	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1004 		if (!net_eq(dev_net(ifp->idev->dev), net))
1005 			continue;
1006 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1007 			if (!dev || ifp->idev->dev == dev)
1008 				return true;
1009 		}
1010 	}
1011 	return false;
1012 }
1013 
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)1014 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1015 {
1016 	unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1017 	int err = 0;
1018 
1019 	spin_lock(&addrconf_hash_lock);
1020 
1021 	/* Ignore adding duplicate addresses on an interface */
1022 	if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1023 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1024 		err = -EEXIST;
1025 	} else {
1026 		hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1027 	}
1028 
1029 	spin_unlock(&addrconf_hash_lock);
1030 
1031 	return err;
1032 }
1033 
1034 /* On success it returns ifp with increased reference count */
1035 
1036 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)1037 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1038 	      bool can_block, struct netlink_ext_ack *extack)
1039 {
1040 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1041 	int addr_type = ipv6_addr_type(cfg->pfx);
1042 	struct net *net = dev_net(idev->dev);
1043 	struct inet6_ifaddr *ifa = NULL;
1044 	struct fib6_info *f6i = NULL;
1045 	int err = 0;
1046 
1047 	if (addr_type == IPV6_ADDR_ANY ||
1048 	    (addr_type & IPV6_ADDR_MULTICAST &&
1049 	     !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1050 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1051 	     !netif_is_l3_master(idev->dev) &&
1052 	     addr_type & IPV6_ADDR_LOOPBACK))
1053 		return ERR_PTR(-EADDRNOTAVAIL);
1054 
1055 	if (idev->dead) {
1056 		err = -ENODEV;			/*XXX*/
1057 		goto out;
1058 	}
1059 
1060 	if (idev->cnf.disable_ipv6) {
1061 		err = -EACCES;
1062 		goto out;
1063 	}
1064 
1065 	/* validator notifier needs to be blocking;
1066 	 * do not call in atomic context
1067 	 */
1068 	if (can_block) {
1069 		struct in6_validator_info i6vi = {
1070 			.i6vi_addr = *cfg->pfx,
1071 			.i6vi_dev = idev,
1072 			.extack = extack,
1073 		};
1074 
1075 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1076 		err = notifier_to_errno(err);
1077 		if (err < 0)
1078 			goto out;
1079 	}
1080 
1081 	ifa = kzalloc(sizeof(*ifa), gfp_flags);
1082 	if (!ifa) {
1083 		err = -ENOBUFS;
1084 		goto out;
1085 	}
1086 
1087 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1088 	if (IS_ERR(f6i)) {
1089 		err = PTR_ERR(f6i);
1090 		f6i = NULL;
1091 		goto out;
1092 	}
1093 
1094 	if (net->ipv6.devconf_all->disable_policy ||
1095 	    idev->cnf.disable_policy)
1096 		f6i->dst_nopolicy = true;
1097 
1098 	neigh_parms_data_state_setall(idev->nd_parms);
1099 
1100 	ifa->addr = *cfg->pfx;
1101 	if (cfg->peer_pfx)
1102 		ifa->peer_addr = *cfg->peer_pfx;
1103 
1104 	spin_lock_init(&ifa->lock);
1105 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1106 	INIT_HLIST_NODE(&ifa->addr_lst);
1107 	ifa->scope = cfg->scope;
1108 	ifa->prefix_len = cfg->plen;
1109 	ifa->rt_priority = cfg->rt_priority;
1110 	ifa->flags = cfg->ifa_flags;
1111 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1112 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1113 		ifa->flags |= IFA_F_TENTATIVE;
1114 	ifa->valid_lft = cfg->valid_lft;
1115 	ifa->prefered_lft = cfg->preferred_lft;
1116 	ifa->cstamp = ifa->tstamp = jiffies;
1117 	ifa->tokenized = false;
1118 
1119 	ifa->rt = f6i;
1120 
1121 	ifa->idev = idev;
1122 	in6_dev_hold(idev);
1123 
1124 	/* For caller */
1125 	refcount_set(&ifa->refcnt, 1);
1126 
1127 	rcu_read_lock_bh();
1128 
1129 	err = ipv6_add_addr_hash(idev->dev, ifa);
1130 	if (err < 0) {
1131 		rcu_read_unlock_bh();
1132 		goto out;
1133 	}
1134 
1135 	write_lock(&idev->lock);
1136 
1137 	/* Add to inet6_dev unicast addr list. */
1138 	ipv6_link_dev_addr(idev, ifa);
1139 
1140 	if (ifa->flags&IFA_F_TEMPORARY) {
1141 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1142 		in6_ifa_hold(ifa);
1143 	}
1144 
1145 	in6_ifa_hold(ifa);
1146 	write_unlock(&idev->lock);
1147 
1148 	rcu_read_unlock_bh();
1149 
1150 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1151 out:
1152 	if (unlikely(err < 0)) {
1153 		fib6_info_release(f6i);
1154 
1155 		if (ifa) {
1156 			if (ifa->idev)
1157 				in6_dev_put(ifa->idev);
1158 			kfree(ifa);
1159 		}
1160 		ifa = ERR_PTR(err);
1161 	}
1162 
1163 	return ifa;
1164 }
1165 
1166 enum cleanup_prefix_rt_t {
1167 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1168 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1169 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1170 };
1171 
1172 /*
1173  * Check, whether the prefix for ifp would still need a prefix route
1174  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1175  * constants.
1176  *
1177  * 1) we don't purge prefix if address was not permanent.
1178  *    prefix is managed by its own lifetime.
1179  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1180  * 3) if there are no addresses, delete prefix.
1181  * 4) if there are still other permanent address(es),
1182  *    corresponding prefix is still permanent.
1183  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1184  *    don't purge the prefix, assume user space is managing it.
1185  * 6) otherwise, update prefix lifetime to the
1186  *    longest valid lifetime among the corresponding
1187  *    addresses on the device.
1188  *    Note: subsequent RA will update lifetime.
1189  **/
1190 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1191 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1192 {
1193 	struct inet6_ifaddr *ifa;
1194 	struct inet6_dev *idev = ifp->idev;
1195 	unsigned long lifetime;
1196 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1197 
1198 	*expires = jiffies;
1199 
1200 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1201 		if (ifa == ifp)
1202 			continue;
1203 		if (ifa->prefix_len != ifp->prefix_len ||
1204 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1205 				       ifp->prefix_len))
1206 			continue;
1207 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1208 			return CLEANUP_PREFIX_RT_NOP;
1209 
1210 		action = CLEANUP_PREFIX_RT_EXPIRE;
1211 
1212 		spin_lock(&ifa->lock);
1213 
1214 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1215 		/*
1216 		 * Note: Because this address is
1217 		 * not permanent, lifetime <
1218 		 * LONG_MAX / HZ here.
1219 		 */
1220 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1221 			*expires = ifa->tstamp + lifetime * HZ;
1222 		spin_unlock(&ifa->lock);
1223 	}
1224 
1225 	return action;
1226 }
1227 
1228 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1229 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1230 		     bool del_rt, bool del_peer)
1231 {
1232 	struct fib6_info *f6i;
1233 
1234 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1235 					ifp->prefix_len,
1236 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1237 	if (f6i) {
1238 		if (del_rt)
1239 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1240 		else {
1241 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1242 				fib6_set_expires(f6i, expires);
1243 			fib6_info_release(f6i);
1244 		}
1245 	}
1246 }
1247 
1248 
1249 /* This function wants to get referenced ifp and releases it before return */
1250 
ipv6_del_addr(struct inet6_ifaddr * ifp)1251 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1252 {
1253 	int state;
1254 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1255 	unsigned long expires;
1256 
1257 	ASSERT_RTNL();
1258 
1259 	spin_lock_bh(&ifp->lock);
1260 	state = ifp->state;
1261 	ifp->state = INET6_IFADDR_STATE_DEAD;
1262 	spin_unlock_bh(&ifp->lock);
1263 
1264 	if (state == INET6_IFADDR_STATE_DEAD)
1265 		goto out;
1266 
1267 	spin_lock_bh(&addrconf_hash_lock);
1268 	hlist_del_init_rcu(&ifp->addr_lst);
1269 	spin_unlock_bh(&addrconf_hash_lock);
1270 
1271 	write_lock_bh(&ifp->idev->lock);
1272 
1273 	if (ifp->flags&IFA_F_TEMPORARY) {
1274 		list_del(&ifp->tmp_list);
1275 		if (ifp->ifpub) {
1276 			in6_ifa_put(ifp->ifpub);
1277 			ifp->ifpub = NULL;
1278 		}
1279 		__in6_ifa_put(ifp);
1280 	}
1281 
1282 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1283 		action = check_cleanup_prefix_route(ifp, &expires);
1284 
1285 	list_del_rcu(&ifp->if_list);
1286 	__in6_ifa_put(ifp);
1287 
1288 	write_unlock_bh(&ifp->idev->lock);
1289 
1290 	addrconf_del_dad_work(ifp);
1291 
1292 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1293 
1294 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1295 
1296 	if (action != CLEANUP_PREFIX_RT_NOP) {
1297 		cleanup_prefix_route(ifp, expires,
1298 			action == CLEANUP_PREFIX_RT_DEL, false);
1299 	}
1300 
1301 	/* clean up prefsrc entries */
1302 	rt6_remove_prefsrc(ifp);
1303 out:
1304 	in6_ifa_put(ifp);
1305 }
1306 
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,bool block)1307 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1308 {
1309 	struct inet6_dev *idev = ifp->idev;
1310 	unsigned long tmp_tstamp, age;
1311 	unsigned long regen_advance;
1312 	unsigned long now = jiffies;
1313 	s32 cnf_temp_preferred_lft;
1314 	struct inet6_ifaddr *ift;
1315 	struct ifa6_config cfg;
1316 	long max_desync_factor;
1317 	struct in6_addr addr;
1318 	int ret = 0;
1319 
1320 	write_lock_bh(&idev->lock);
1321 
1322 retry:
1323 	in6_dev_hold(idev);
1324 	if (idev->cnf.use_tempaddr <= 0) {
1325 		write_unlock_bh(&idev->lock);
1326 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1327 		in6_dev_put(idev);
1328 		ret = -1;
1329 		goto out;
1330 	}
1331 	spin_lock_bh(&ifp->lock);
1332 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1333 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1334 		spin_unlock_bh(&ifp->lock);
1335 		write_unlock_bh(&idev->lock);
1336 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1337 			__func__);
1338 		in6_dev_put(idev);
1339 		ret = -1;
1340 		goto out;
1341 	}
1342 	in6_ifa_hold(ifp);
1343 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1344 	ipv6_gen_rnd_iid(&addr);
1345 
1346 	age = (now - ifp->tstamp) / HZ;
1347 
1348 	regen_advance = idev->cnf.regen_max_retry *
1349 			idev->cnf.dad_transmits *
1350 			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1351 
1352 	/* recalculate max_desync_factor each time and update
1353 	 * idev->desync_factor if it's larger
1354 	 */
1355 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1356 	max_desync_factor = min_t(__u32,
1357 				  idev->cnf.max_desync_factor,
1358 				  cnf_temp_preferred_lft - regen_advance);
1359 
1360 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1361 		if (max_desync_factor > 0) {
1362 			get_random_bytes(&idev->desync_factor,
1363 					 sizeof(idev->desync_factor));
1364 			idev->desync_factor %= max_desync_factor;
1365 		} else {
1366 			idev->desync_factor = 0;
1367 		}
1368 	}
1369 
1370 	memset(&cfg, 0, sizeof(cfg));
1371 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1372 			      idev->cnf.temp_valid_lft + age);
1373 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1374 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1375 
1376 	cfg.plen = ifp->prefix_len;
1377 	tmp_tstamp = ifp->tstamp;
1378 	spin_unlock_bh(&ifp->lock);
1379 
1380 	write_unlock_bh(&idev->lock);
1381 
1382 	/* A temporary address is created only if this calculated Preferred
1383 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1384 	 * an implementation must not create a temporary address with a zero
1385 	 * Preferred Lifetime.
1386 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1387 	 * temporary addresses being generated.
1388 	 */
1389 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1390 	if (cfg.preferred_lft <= regen_advance + age) {
1391 		in6_ifa_put(ifp);
1392 		in6_dev_put(idev);
1393 		ret = -1;
1394 		goto out;
1395 	}
1396 
1397 	cfg.ifa_flags = IFA_F_TEMPORARY;
1398 	/* set in addrconf_prefix_rcv() */
1399 	if (ifp->flags & IFA_F_OPTIMISTIC)
1400 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1401 
1402 	cfg.pfx = &addr;
1403 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1404 
1405 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1406 	if (IS_ERR(ift)) {
1407 		in6_ifa_put(ifp);
1408 		in6_dev_put(idev);
1409 		pr_info("%s: retry temporary address regeneration\n", __func__);
1410 		write_lock_bh(&idev->lock);
1411 		goto retry;
1412 	}
1413 
1414 	spin_lock_bh(&ift->lock);
1415 	ift->ifpub = ifp;
1416 	ift->cstamp = now;
1417 	ift->tstamp = tmp_tstamp;
1418 	spin_unlock_bh(&ift->lock);
1419 
1420 	addrconf_dad_start(ift);
1421 	in6_ifa_put(ift);
1422 	in6_dev_put(idev);
1423 out:
1424 	return ret;
1425 }
1426 
1427 /*
1428  *	Choose an appropriate source address (RFC3484)
1429  */
1430 enum {
1431 	IPV6_SADDR_RULE_INIT = 0,
1432 	IPV6_SADDR_RULE_LOCAL,
1433 	IPV6_SADDR_RULE_SCOPE,
1434 	IPV6_SADDR_RULE_PREFERRED,
1435 #ifdef CONFIG_IPV6_MIP6
1436 	IPV6_SADDR_RULE_HOA,
1437 #endif
1438 	IPV6_SADDR_RULE_OIF,
1439 	IPV6_SADDR_RULE_LABEL,
1440 	IPV6_SADDR_RULE_PRIVACY,
1441 	IPV6_SADDR_RULE_ORCHID,
1442 	IPV6_SADDR_RULE_PREFIX,
1443 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1444 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1445 #endif
1446 	IPV6_SADDR_RULE_MAX
1447 };
1448 
1449 struct ipv6_saddr_score {
1450 	int			rule;
1451 	int			addr_type;
1452 	struct inet6_ifaddr	*ifa;
1453 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1454 	int			scopedist;
1455 	int			matchlen;
1456 };
1457 
1458 struct ipv6_saddr_dst {
1459 	const struct in6_addr *addr;
1460 	int ifindex;
1461 	int scope;
1462 	int label;
1463 	unsigned int prefs;
1464 };
1465 
ipv6_saddr_preferred(int type)1466 static inline int ipv6_saddr_preferred(int type)
1467 {
1468 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1469 		return 1;
1470 	return 0;
1471 }
1472 
ipv6_use_optimistic_addr(struct net * net,struct inet6_dev * idev)1473 static bool ipv6_use_optimistic_addr(struct net *net,
1474 				     struct inet6_dev *idev)
1475 {
1476 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1477 	if (!idev)
1478 		return false;
1479 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1480 		return false;
1481 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1482 		return false;
1483 
1484 	return true;
1485 #else
1486 	return false;
1487 #endif
1488 }
1489 
ipv6_allow_optimistic_dad(struct net * net,struct inet6_dev * idev)1490 static bool ipv6_allow_optimistic_dad(struct net *net,
1491 				      struct inet6_dev *idev)
1492 {
1493 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1494 	if (!idev)
1495 		return false;
1496 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1497 		return false;
1498 
1499 	return true;
1500 #else
1501 	return false;
1502 #endif
1503 }
1504 
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1505 static int ipv6_get_saddr_eval(struct net *net,
1506 			       struct ipv6_saddr_score *score,
1507 			       struct ipv6_saddr_dst *dst,
1508 			       int i)
1509 {
1510 	int ret;
1511 
1512 	if (i <= score->rule) {
1513 		switch (i) {
1514 		case IPV6_SADDR_RULE_SCOPE:
1515 			ret = score->scopedist;
1516 			break;
1517 		case IPV6_SADDR_RULE_PREFIX:
1518 			ret = score->matchlen;
1519 			break;
1520 		default:
1521 			ret = !!test_bit(i, score->scorebits);
1522 		}
1523 		goto out;
1524 	}
1525 
1526 	switch (i) {
1527 	case IPV6_SADDR_RULE_INIT:
1528 		/* Rule 0: remember if hiscore is not ready yet */
1529 		ret = !!score->ifa;
1530 		break;
1531 	case IPV6_SADDR_RULE_LOCAL:
1532 		/* Rule 1: Prefer same address */
1533 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1534 		break;
1535 	case IPV6_SADDR_RULE_SCOPE:
1536 		/* Rule 2: Prefer appropriate scope
1537 		 *
1538 		 *      ret
1539 		 *       ^
1540 		 *    -1 |  d 15
1541 		 *    ---+--+-+---> scope
1542 		 *       |
1543 		 *       |             d is scope of the destination.
1544 		 *  B-d  |  \
1545 		 *       |   \      <- smaller scope is better if
1546 		 *  B-15 |    \        if scope is enough for destination.
1547 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1548 		 * d-C-1 | /
1549 		 *       |/         <- greater is better
1550 		 *   -C  /             if scope is not enough for destination.
1551 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1552 		 *
1553 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1554 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1555 		 * Assume B = 0 and we get C > 29.
1556 		 */
1557 		ret = __ipv6_addr_src_scope(score->addr_type);
1558 		if (ret >= dst->scope)
1559 			ret = -ret;
1560 		else
1561 			ret -= 128;	/* 30 is enough */
1562 		score->scopedist = ret;
1563 		break;
1564 	case IPV6_SADDR_RULE_PREFERRED:
1565 	    {
1566 		/* Rule 3: Avoid deprecated and optimistic addresses */
1567 		u8 avoid = IFA_F_DEPRECATED;
1568 
1569 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1570 			avoid |= IFA_F_OPTIMISTIC;
1571 		ret = ipv6_saddr_preferred(score->addr_type) ||
1572 		      !(score->ifa->flags & avoid);
1573 		break;
1574 	    }
1575 #ifdef CONFIG_IPV6_MIP6
1576 	case IPV6_SADDR_RULE_HOA:
1577 	    {
1578 		/* Rule 4: Prefer home address */
1579 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1580 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1581 		break;
1582 	    }
1583 #endif
1584 	case IPV6_SADDR_RULE_OIF:
1585 		/* Rule 5: Prefer outgoing interface */
1586 		ret = (!dst->ifindex ||
1587 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1588 		break;
1589 	case IPV6_SADDR_RULE_LABEL:
1590 		/* Rule 6: Prefer matching label */
1591 		ret = ipv6_addr_label(net,
1592 				      &score->ifa->addr, score->addr_type,
1593 				      score->ifa->idev->dev->ifindex) == dst->label;
1594 		break;
1595 	case IPV6_SADDR_RULE_PRIVACY:
1596 	    {
1597 		/* Rule 7: Prefer public address
1598 		 * Note: prefer temporary address if use_tempaddr >= 2
1599 		 */
1600 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1601 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1602 				score->ifa->idev->cnf.use_tempaddr >= 2;
1603 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1604 		break;
1605 	    }
1606 	case IPV6_SADDR_RULE_ORCHID:
1607 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1608 		 *	    non-ORCHID vs non-ORCHID
1609 		 */
1610 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1611 			ipv6_addr_orchid(dst->addr));
1612 		break;
1613 	case IPV6_SADDR_RULE_PREFIX:
1614 		/* Rule 8: Use longest matching prefix */
1615 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1616 		if (ret > score->ifa->prefix_len)
1617 			ret = score->ifa->prefix_len;
1618 		score->matchlen = ret;
1619 		break;
1620 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1621 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1622 		/* Optimistic addresses still have lower precedence than other
1623 		 * preferred addresses.
1624 		 */
1625 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1626 		break;
1627 #endif
1628 	default:
1629 		ret = 0;
1630 	}
1631 
1632 	if (ret)
1633 		__set_bit(i, score->scorebits);
1634 	score->rule = i;
1635 out:
1636 	return ret;
1637 }
1638 
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1639 static int __ipv6_dev_get_saddr(struct net *net,
1640 				struct ipv6_saddr_dst *dst,
1641 				struct inet6_dev *idev,
1642 				struct ipv6_saddr_score *scores,
1643 				int hiscore_idx)
1644 {
1645 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1646 
1647 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1648 		int i;
1649 
1650 		/*
1651 		 * - Tentative Address (RFC2462 section 5.4)
1652 		 *  - A tentative address is not considered
1653 		 *    "assigned to an interface" in the traditional
1654 		 *    sense, unless it is also flagged as optimistic.
1655 		 * - Candidate Source Address (section 4)
1656 		 *  - In any case, anycast addresses, multicast
1657 		 *    addresses, and the unspecified address MUST
1658 		 *    NOT be included in a candidate set.
1659 		 */
1660 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1661 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1662 			continue;
1663 
1664 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1665 
1666 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1667 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1668 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1669 					    idev->dev->name);
1670 			continue;
1671 		}
1672 
1673 		score->rule = -1;
1674 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1675 
1676 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1677 			int minihiscore, miniscore;
1678 
1679 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1680 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1681 
1682 			if (minihiscore > miniscore) {
1683 				if (i == IPV6_SADDR_RULE_SCOPE &&
1684 				    score->scopedist > 0) {
1685 					/*
1686 					 * special case:
1687 					 * each remaining entry
1688 					 * has too small (not enough)
1689 					 * scope, because ifa entries
1690 					 * are sorted by their scope
1691 					 * values.
1692 					 */
1693 					goto out;
1694 				}
1695 				break;
1696 			} else if (minihiscore < miniscore) {
1697 				swap(hiscore, score);
1698 				hiscore_idx = 1 - hiscore_idx;
1699 
1700 				/* restore our iterator */
1701 				score->ifa = hiscore->ifa;
1702 
1703 				break;
1704 			}
1705 		}
1706 	}
1707 out:
1708 	return hiscore_idx;
1709 }
1710 
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)1711 static int ipv6_get_saddr_master(struct net *net,
1712 				 const struct net_device *dst_dev,
1713 				 const struct net_device *master,
1714 				 struct ipv6_saddr_dst *dst,
1715 				 struct ipv6_saddr_score *scores,
1716 				 int hiscore_idx)
1717 {
1718 	struct inet6_dev *idev;
1719 
1720 	idev = __in6_dev_get(dst_dev);
1721 	if (idev)
1722 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1723 						   scores, hiscore_idx);
1724 
1725 	idev = __in6_dev_get(master);
1726 	if (idev)
1727 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1728 						   scores, hiscore_idx);
1729 
1730 	return hiscore_idx;
1731 }
1732 
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)1733 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1734 		       const struct in6_addr *daddr, unsigned int prefs,
1735 		       struct in6_addr *saddr)
1736 {
1737 	struct ipv6_saddr_score scores[2], *hiscore;
1738 	struct ipv6_saddr_dst dst;
1739 	struct inet6_dev *idev;
1740 	struct net_device *dev;
1741 	int dst_type;
1742 	bool use_oif_addr = false;
1743 	int hiscore_idx = 0;
1744 	int ret = 0;
1745 
1746 	dst_type = __ipv6_addr_type(daddr);
1747 	dst.addr = daddr;
1748 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1749 	dst.scope = __ipv6_addr_src_scope(dst_type);
1750 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1751 	dst.prefs = prefs;
1752 
1753 	scores[hiscore_idx].rule = -1;
1754 	scores[hiscore_idx].ifa = NULL;
1755 
1756 	rcu_read_lock();
1757 
1758 	/* Candidate Source Address (section 4)
1759 	 *  - multicast and link-local destination address,
1760 	 *    the set of candidate source address MUST only
1761 	 *    include addresses assigned to interfaces
1762 	 *    belonging to the same link as the outgoing
1763 	 *    interface.
1764 	 * (- For site-local destination addresses, the
1765 	 *    set of candidate source addresses MUST only
1766 	 *    include addresses assigned to interfaces
1767 	 *    belonging to the same site as the outgoing
1768 	 *    interface.)
1769 	 *  - "It is RECOMMENDED that the candidate source addresses
1770 	 *    be the set of unicast addresses assigned to the
1771 	 *    interface that will be used to send to the destination
1772 	 *    (the 'outgoing' interface)." (RFC 6724)
1773 	 */
1774 	if (dst_dev) {
1775 		idev = __in6_dev_get(dst_dev);
1776 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1777 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1778 		    (idev && idev->cnf.use_oif_addrs_only)) {
1779 			use_oif_addr = true;
1780 		}
1781 	}
1782 
1783 	if (use_oif_addr) {
1784 		if (idev)
1785 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1786 	} else {
1787 		const struct net_device *master;
1788 		int master_idx = 0;
1789 
1790 		/* if dst_dev exists and is enslaved to an L3 device, then
1791 		 * prefer addresses from dst_dev and then the master over
1792 		 * any other enslaved devices in the L3 domain.
1793 		 */
1794 		master = l3mdev_master_dev_rcu(dst_dev);
1795 		if (master) {
1796 			master_idx = master->ifindex;
1797 
1798 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1799 							    master, &dst,
1800 							    scores, hiscore_idx);
1801 
1802 			if (scores[hiscore_idx].ifa)
1803 				goto out;
1804 		}
1805 
1806 		for_each_netdev_rcu(net, dev) {
1807 			/* only consider addresses on devices in the
1808 			 * same L3 domain
1809 			 */
1810 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1811 				continue;
1812 			idev = __in6_dev_get(dev);
1813 			if (!idev)
1814 				continue;
1815 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1816 		}
1817 	}
1818 
1819 out:
1820 	hiscore = &scores[hiscore_idx];
1821 	if (!hiscore->ifa)
1822 		ret = -EADDRNOTAVAIL;
1823 	else
1824 		*saddr = hiscore->ifa->addr;
1825 
1826 	rcu_read_unlock();
1827 	return ret;
1828 }
1829 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1830 
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1831 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1832 		      u32 banned_flags)
1833 {
1834 	struct inet6_ifaddr *ifp;
1835 	int err = -EADDRNOTAVAIL;
1836 
1837 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1838 		if (ifp->scope > IFA_LINK)
1839 			break;
1840 		if (ifp->scope == IFA_LINK &&
1841 		    !(ifp->flags & banned_flags)) {
1842 			*addr = ifp->addr;
1843 			err = 0;
1844 			break;
1845 		}
1846 	}
1847 	return err;
1848 }
1849 
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1850 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1851 		    u32 banned_flags)
1852 {
1853 	struct inet6_dev *idev;
1854 	int err = -EADDRNOTAVAIL;
1855 
1856 	rcu_read_lock();
1857 	idev = __in6_dev_get(dev);
1858 	if (idev) {
1859 		read_lock_bh(&idev->lock);
1860 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1861 		read_unlock_bh(&idev->lock);
1862 	}
1863 	rcu_read_unlock();
1864 	return err;
1865 }
1866 
ipv6_count_addresses(const struct inet6_dev * idev)1867 static int ipv6_count_addresses(const struct inet6_dev *idev)
1868 {
1869 	const struct inet6_ifaddr *ifp;
1870 	int cnt = 0;
1871 
1872 	rcu_read_lock();
1873 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1874 		cnt++;
1875 	rcu_read_unlock();
1876 	return cnt;
1877 }
1878 
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1879 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1880 		  const struct net_device *dev, int strict)
1881 {
1882 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1883 				       strict, IFA_F_TENTATIVE);
1884 }
1885 EXPORT_SYMBOL(ipv6_chk_addr);
1886 
1887 /* device argument is used to find the L3 domain of interest. If
1888  * skip_dev_check is set, then the ifp device is not checked against
1889  * the passed in dev argument. So the 2 cases for addresses checks are:
1890  *   1. does the address exist in the L3 domain that dev is part of
1891  *      (skip_dev_check = true), or
1892  *
1893  *   2. does the address exist on the specific device
1894  *      (skip_dev_check = false)
1895  */
1896 static struct net_device *
__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)1897 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1898 			  const struct net_device *dev, bool skip_dev_check,
1899 			  int strict, u32 banned_flags)
1900 {
1901 	unsigned int hash = inet6_addr_hash(net, addr);
1902 	struct net_device *l3mdev, *ndev;
1903 	struct inet6_ifaddr *ifp;
1904 	u32 ifp_flags;
1905 
1906 	rcu_read_lock();
1907 
1908 	l3mdev = l3mdev_master_dev_rcu(dev);
1909 	if (skip_dev_check)
1910 		dev = NULL;
1911 
1912 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1913 		ndev = ifp->idev->dev;
1914 		if (!net_eq(dev_net(ndev), net))
1915 			continue;
1916 
1917 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1918 			continue;
1919 
1920 		/* Decouple optimistic from tentative for evaluation here.
1921 		 * Ban optimistic addresses explicitly, when required.
1922 		 */
1923 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1924 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1925 			    : ifp->flags;
1926 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1927 		    !(ifp_flags&banned_flags) &&
1928 		    (!dev || ndev == dev ||
1929 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1930 			rcu_read_unlock();
1931 			return ndev;
1932 		}
1933 	}
1934 
1935 	rcu_read_unlock();
1936 	return NULL;
1937 }
1938 
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)1939 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1940 			    const struct net_device *dev, bool skip_dev_check,
1941 			    int strict, u32 banned_flags)
1942 {
1943 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1944 					 strict, banned_flags) ? 1 : 0;
1945 }
1946 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1947 
1948 
1949 /* Compares an address/prefix_len with addresses on device @dev.
1950  * If one is found it returns true.
1951  */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)1952 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1953 	const unsigned int prefix_len, struct net_device *dev)
1954 {
1955 	const struct inet6_ifaddr *ifa;
1956 	const struct inet6_dev *idev;
1957 	bool ret = false;
1958 
1959 	rcu_read_lock();
1960 	idev = __in6_dev_get(dev);
1961 	if (idev) {
1962 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1963 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1964 			if (ret)
1965 				break;
1966 		}
1967 	}
1968 	rcu_read_unlock();
1969 
1970 	return ret;
1971 }
1972 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1973 
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)1974 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1975 {
1976 	const struct inet6_ifaddr *ifa;
1977 	const struct inet6_dev *idev;
1978 	int	onlink;
1979 
1980 	onlink = 0;
1981 	rcu_read_lock();
1982 	idev = __in6_dev_get(dev);
1983 	if (idev) {
1984 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1985 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1986 						   ifa->prefix_len);
1987 			if (onlink)
1988 				break;
1989 		}
1990 	}
1991 	rcu_read_unlock();
1992 	return onlink;
1993 }
1994 EXPORT_SYMBOL(ipv6_chk_prefix);
1995 
1996 /**
1997  * ipv6_dev_find - find the first device with a given source address.
1998  * @net: the net namespace
1999  * @addr: the source address
2000  *
2001  * The caller should be protected by RCU, or RTNL.
2002  */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2003 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2004 				 struct net_device *dev)
2005 {
2006 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2007 					 IFA_F_TENTATIVE);
2008 }
2009 EXPORT_SYMBOL(ipv6_dev_find);
2010 
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2011 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2012 				     struct net_device *dev, int strict)
2013 {
2014 	unsigned int hash = inet6_addr_hash(net, addr);
2015 	struct inet6_ifaddr *ifp, *result = NULL;
2016 
2017 	rcu_read_lock();
2018 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2019 		if (!net_eq(dev_net(ifp->idev->dev), net))
2020 			continue;
2021 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2022 			if (!dev || ifp->idev->dev == dev ||
2023 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2024 				result = ifp;
2025 				in6_ifa_hold(ifp);
2026 				break;
2027 			}
2028 		}
2029 	}
2030 	rcu_read_unlock();
2031 
2032 	return result;
2033 }
2034 
2035 /* Gets referenced address, destroys ifaddr */
2036 
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2037 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2038 {
2039 	if (dad_failed)
2040 		ifp->flags |= IFA_F_DADFAILED;
2041 
2042 	if (ifp->flags&IFA_F_TEMPORARY) {
2043 		struct inet6_ifaddr *ifpub;
2044 		spin_lock_bh(&ifp->lock);
2045 		ifpub = ifp->ifpub;
2046 		if (ifpub) {
2047 			in6_ifa_hold(ifpub);
2048 			spin_unlock_bh(&ifp->lock);
2049 			ipv6_create_tempaddr(ifpub, true);
2050 			in6_ifa_put(ifpub);
2051 		} else {
2052 			spin_unlock_bh(&ifp->lock);
2053 		}
2054 		ipv6_del_addr(ifp);
2055 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2056 		spin_lock_bh(&ifp->lock);
2057 		addrconf_del_dad_work(ifp);
2058 		ifp->flags |= IFA_F_TENTATIVE;
2059 		if (dad_failed)
2060 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2061 		spin_unlock_bh(&ifp->lock);
2062 		if (dad_failed)
2063 			ipv6_ifa_notify(0, ifp);
2064 		in6_ifa_put(ifp);
2065 	} else {
2066 		ipv6_del_addr(ifp);
2067 	}
2068 }
2069 
addrconf_dad_end(struct inet6_ifaddr * ifp)2070 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2071 {
2072 	int err = -ENOENT;
2073 
2074 	spin_lock_bh(&ifp->lock);
2075 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2076 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2077 		err = 0;
2078 	}
2079 	spin_unlock_bh(&ifp->lock);
2080 
2081 	return err;
2082 }
2083 
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2084 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2085 {
2086 	struct inet6_dev *idev = ifp->idev;
2087 	struct net *net = dev_net(ifp->idev->dev);
2088 
2089 	if (addrconf_dad_end(ifp)) {
2090 		in6_ifa_put(ifp);
2091 		return;
2092 	}
2093 
2094 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2095 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2096 
2097 	spin_lock_bh(&ifp->lock);
2098 
2099 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2100 		struct in6_addr new_addr;
2101 		struct inet6_ifaddr *ifp2;
2102 		int retries = ifp->stable_privacy_retry + 1;
2103 		struct ifa6_config cfg = {
2104 			.pfx = &new_addr,
2105 			.plen = ifp->prefix_len,
2106 			.ifa_flags = ifp->flags,
2107 			.valid_lft = ifp->valid_lft,
2108 			.preferred_lft = ifp->prefered_lft,
2109 			.scope = ifp->scope,
2110 		};
2111 
2112 		if (retries > net->ipv6.sysctl.idgen_retries) {
2113 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2114 					     ifp->idev->dev->name);
2115 			goto errdad;
2116 		}
2117 
2118 		new_addr = ifp->addr;
2119 		if (ipv6_generate_stable_address(&new_addr, retries,
2120 						 idev))
2121 			goto errdad;
2122 
2123 		spin_unlock_bh(&ifp->lock);
2124 
2125 		if (idev->cnf.max_addresses &&
2126 		    ipv6_count_addresses(idev) >=
2127 		    idev->cnf.max_addresses)
2128 			goto lock_errdad;
2129 
2130 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2131 				     ifp->idev->dev->name);
2132 
2133 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2134 		if (IS_ERR(ifp2))
2135 			goto lock_errdad;
2136 
2137 		spin_lock_bh(&ifp2->lock);
2138 		ifp2->stable_privacy_retry = retries;
2139 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2140 		spin_unlock_bh(&ifp2->lock);
2141 
2142 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2143 		in6_ifa_put(ifp2);
2144 lock_errdad:
2145 		spin_lock_bh(&ifp->lock);
2146 	}
2147 
2148 errdad:
2149 	/* transition from _POSTDAD to _ERRDAD */
2150 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2151 	spin_unlock_bh(&ifp->lock);
2152 
2153 	addrconf_mod_dad_work(ifp, 0);
2154 	in6_ifa_put(ifp);
2155 }
2156 
2157 /* Join to solicited addr multicast group.
2158  * caller must hold RTNL */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2159 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2160 {
2161 	struct in6_addr maddr;
2162 
2163 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2164 		return;
2165 
2166 	addrconf_addr_solict_mult(addr, &maddr);
2167 	ipv6_dev_mc_inc(dev, &maddr);
2168 }
2169 
2170 /* caller must hold RTNL */
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2171 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2172 {
2173 	struct in6_addr maddr;
2174 
2175 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2176 		return;
2177 
2178 	addrconf_addr_solict_mult(addr, &maddr);
2179 	__ipv6_dev_mc_dec(idev, &maddr);
2180 }
2181 
2182 /* caller must hold RTNL */
addrconf_join_anycast(struct inet6_ifaddr * ifp)2183 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2184 {
2185 	struct in6_addr addr;
2186 
2187 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2188 		return;
2189 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2190 	if (ipv6_addr_any(&addr))
2191 		return;
2192 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2193 }
2194 
2195 /* caller must hold RTNL */
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2196 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2197 {
2198 	struct in6_addr addr;
2199 
2200 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2201 		return;
2202 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2203 	if (ipv6_addr_any(&addr))
2204 		return;
2205 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2206 }
2207 
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2208 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2209 {
2210 	switch (dev->addr_len) {
2211 	case ETH_ALEN:
2212 		memcpy(eui, dev->dev_addr, 3);
2213 		eui[3] = 0xFF;
2214 		eui[4] = 0xFE;
2215 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2216 		break;
2217 	case EUI64_ADDR_LEN:
2218 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2219 		eui[0] ^= 2;
2220 		break;
2221 	default:
2222 		return -1;
2223 	}
2224 
2225 	return 0;
2226 }
2227 
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2228 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2229 {
2230 	union fwnet_hwaddr *ha;
2231 
2232 	if (dev->addr_len != FWNET_ALEN)
2233 		return -1;
2234 
2235 	ha = (union fwnet_hwaddr *)dev->dev_addr;
2236 
2237 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2238 	eui[0] ^= 2;
2239 	return 0;
2240 }
2241 
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2242 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2243 {
2244 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2245 	if (dev->addr_len != ARCNET_ALEN)
2246 		return -1;
2247 	memset(eui, 0, 7);
2248 	eui[7] = *(u8 *)dev->dev_addr;
2249 	return 0;
2250 }
2251 
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2252 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2253 {
2254 	if (dev->addr_len != INFINIBAND_ALEN)
2255 		return -1;
2256 	memcpy(eui, dev->dev_addr + 12, 8);
2257 	eui[0] |= 2;
2258 	return 0;
2259 }
2260 
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2261 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2262 {
2263 	if (addr == 0)
2264 		return -1;
2265 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2266 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2267 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2268 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2269 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2270 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2271 	eui[1] = 0;
2272 	eui[2] = 0x5E;
2273 	eui[3] = 0xFE;
2274 	memcpy(eui + 4, &addr, 4);
2275 	return 0;
2276 }
2277 
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2278 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2279 {
2280 	if (dev->priv_flags & IFF_ISATAP)
2281 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2282 	return -1;
2283 }
2284 
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2285 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2286 {
2287 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2288 }
2289 
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2290 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2291 {
2292 	memcpy(eui, dev->perm_addr, 3);
2293 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2294 	eui[3] = 0xFF;
2295 	eui[4] = 0xFE;
2296 	eui[0] ^= 2;
2297 	return 0;
2298 }
2299 
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2300 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2301 {
2302 	switch (dev->type) {
2303 	case ARPHRD_ETHER:
2304 	case ARPHRD_FDDI:
2305 		return addrconf_ifid_eui48(eui, dev);
2306 	case ARPHRD_ARCNET:
2307 		return addrconf_ifid_arcnet(eui, dev);
2308 	case ARPHRD_INFINIBAND:
2309 		return addrconf_ifid_infiniband(eui, dev);
2310 	case ARPHRD_SIT:
2311 		return addrconf_ifid_sit(eui, dev);
2312 	case ARPHRD_IPGRE:
2313 	case ARPHRD_TUNNEL:
2314 		return addrconf_ifid_gre(eui, dev);
2315 	case ARPHRD_6LOWPAN:
2316 		return addrconf_ifid_6lowpan(eui, dev);
2317 	case ARPHRD_IEEE1394:
2318 		return addrconf_ifid_ieee1394(eui, dev);
2319 	case ARPHRD_TUNNEL6:
2320 	case ARPHRD_IP6GRE:
2321 	case ARPHRD_RAWIP:
2322 		return addrconf_ifid_ip6tnl(eui, dev);
2323 	}
2324 	return -1;
2325 }
2326 
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2327 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2328 {
2329 	int err = -1;
2330 	struct inet6_ifaddr *ifp;
2331 
2332 	read_lock_bh(&idev->lock);
2333 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2334 		if (ifp->scope > IFA_LINK)
2335 			break;
2336 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2337 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2338 			err = 0;
2339 			break;
2340 		}
2341 	}
2342 	read_unlock_bh(&idev->lock);
2343 	return err;
2344 }
2345 
2346 /* Generation of a randomized Interface Identifier
2347  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2348  */
2349 
ipv6_gen_rnd_iid(struct in6_addr * addr)2350 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2351 {
2352 regen:
2353 	get_random_bytes(&addr->s6_addr[8], 8);
2354 
2355 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2356 	 * check if generated address is not inappropriate:
2357 	 *
2358 	 * - Reserved IPv6 Interface Identifers
2359 	 * - XXX: already assigned to an address on the device
2360 	 */
2361 
2362 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2363 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2364 		goto regen;
2365 
2366 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2367 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2368 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2369 	 */
2370 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2371 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2372 		goto regen;
2373 
2374 	/* Reserved subnet anycast addresses */
2375 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2376 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2377 		goto regen;
2378 }
2379 
2380 /*
2381  *	Add prefix route.
2382  */
2383 
2384 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)2385 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2386 		      struct net_device *dev, unsigned long expires,
2387 		      u32 flags, gfp_t gfp_flags)
2388 {
2389 	struct fib6_config cfg = {
2390 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2391 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2392 		.fc_ifindex = dev->ifindex,
2393 		.fc_expires = expires,
2394 		.fc_dst_len = plen,
2395 		.fc_flags = RTF_UP | flags,
2396 		.fc_nlinfo.nl_net = dev_net(dev),
2397 		.fc_protocol = RTPROT_KERNEL,
2398 		.fc_type = RTN_UNICAST,
2399 	};
2400 
2401 	cfg.fc_dst = *pfx;
2402 
2403 	/* Prevent useless cloning on PtP SIT.
2404 	   This thing is done here expecting that the whole
2405 	   class of non-broadcast devices need not cloning.
2406 	 */
2407 #if IS_ENABLED(CONFIG_IPV6_SIT)
2408 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2409 		cfg.fc_flags |= RTF_NONEXTHOP;
2410 #endif
2411 
2412 	ip6_route_add(&cfg, gfp_flags, NULL);
2413 }
2414 
2415 
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2416 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2417 						  int plen,
2418 						  const struct net_device *dev,
2419 						  u32 flags, u32 noflags,
2420 						  bool no_gw)
2421 {
2422 	struct fib6_node *fn;
2423 	struct fib6_info *rt = NULL;
2424 	struct fib6_table *table;
2425 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2426 
2427 	table = fib6_get_table(dev_net(dev), tb_id);
2428 	if (!table)
2429 		return NULL;
2430 
2431 	rcu_read_lock();
2432 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2433 	if (!fn)
2434 		goto out;
2435 
2436 	for_each_fib6_node_rt_rcu(fn) {
2437 		/* prefix routes only use builtin fib6_nh */
2438 		if (rt->nh)
2439 			continue;
2440 
2441 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2442 			continue;
2443 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2444 			continue;
2445 		if ((rt->fib6_flags & flags) != flags)
2446 			continue;
2447 		if ((rt->fib6_flags & noflags) != 0)
2448 			continue;
2449 		if (!fib6_info_hold_safe(rt))
2450 			continue;
2451 		break;
2452 	}
2453 out:
2454 	rcu_read_unlock();
2455 	return rt;
2456 }
2457 
2458 
2459 /* Create "default" multicast route to the interface */
2460 
addrconf_add_mroute(struct net_device * dev)2461 static void addrconf_add_mroute(struct net_device *dev)
2462 {
2463 	struct fib6_config cfg = {
2464 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2465 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2466 		.fc_ifindex = dev->ifindex,
2467 		.fc_dst_len = 8,
2468 		.fc_flags = RTF_UP,
2469 		.fc_type = RTN_MULTICAST,
2470 		.fc_nlinfo.nl_net = dev_net(dev),
2471 		.fc_protocol = RTPROT_KERNEL,
2472 	};
2473 
2474 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2475 
2476 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2477 }
2478 
addrconf_add_dev(struct net_device * dev)2479 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2480 {
2481 	struct inet6_dev *idev;
2482 
2483 	ASSERT_RTNL();
2484 
2485 	idev = ipv6_find_idev(dev);
2486 	if (IS_ERR(idev))
2487 		return idev;
2488 
2489 	if (idev->cnf.disable_ipv6)
2490 		return ERR_PTR(-EACCES);
2491 
2492 	/* Add default multicast route */
2493 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2494 		addrconf_add_mroute(dev);
2495 
2496 	return idev;
2497 }
2498 
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2499 static void manage_tempaddrs(struct inet6_dev *idev,
2500 			     struct inet6_ifaddr *ifp,
2501 			     __u32 valid_lft, __u32 prefered_lft,
2502 			     bool create, unsigned long now)
2503 {
2504 	u32 flags;
2505 	struct inet6_ifaddr *ift;
2506 
2507 	read_lock_bh(&idev->lock);
2508 	/* update all temporary addresses in the list */
2509 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2510 		int age, max_valid, max_prefered;
2511 
2512 		if (ifp != ift->ifpub)
2513 			continue;
2514 
2515 		/* RFC 4941 section 3.3:
2516 		 * If a received option will extend the lifetime of a public
2517 		 * address, the lifetimes of temporary addresses should
2518 		 * be extended, subject to the overall constraint that no
2519 		 * temporary addresses should ever remain "valid" or "preferred"
2520 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2521 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2522 		 */
2523 		age = (now - ift->cstamp) / HZ;
2524 		max_valid = idev->cnf.temp_valid_lft - age;
2525 		if (max_valid < 0)
2526 			max_valid = 0;
2527 
2528 		max_prefered = idev->cnf.temp_prefered_lft -
2529 			       idev->desync_factor - age;
2530 		if (max_prefered < 0)
2531 			max_prefered = 0;
2532 
2533 		if (valid_lft > max_valid)
2534 			valid_lft = max_valid;
2535 
2536 		if (prefered_lft > max_prefered)
2537 			prefered_lft = max_prefered;
2538 
2539 		spin_lock(&ift->lock);
2540 		flags = ift->flags;
2541 		ift->valid_lft = valid_lft;
2542 		ift->prefered_lft = prefered_lft;
2543 		ift->tstamp = now;
2544 		if (prefered_lft > 0)
2545 			ift->flags &= ~IFA_F_DEPRECATED;
2546 
2547 		spin_unlock(&ift->lock);
2548 		if (!(flags&IFA_F_TENTATIVE))
2549 			ipv6_ifa_notify(0, ift);
2550 	}
2551 
2552 	if ((create || list_empty(&idev->tempaddr_list)) &&
2553 	    idev->cnf.use_tempaddr > 0) {
2554 		/* When a new public address is created as described
2555 		 * in [ADDRCONF], also create a new temporary address.
2556 		 * Also create a temporary address if it's enabled but
2557 		 * no temporary address currently exists.
2558 		 */
2559 		read_unlock_bh(&idev->lock);
2560 		ipv6_create_tempaddr(ifp, false);
2561 	} else {
2562 		read_unlock_bh(&idev->lock);
2563 	}
2564 }
2565 
is_addr_mode_generate_stable(struct inet6_dev * idev)2566 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2567 {
2568 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2569 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2570 }
2571 
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)2572 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2573 				 const struct prefix_info *pinfo,
2574 				 struct inet6_dev *in6_dev,
2575 				 const struct in6_addr *addr, int addr_type,
2576 				 u32 addr_flags, bool sllao, bool tokenized,
2577 				 __u32 valid_lft, u32 prefered_lft)
2578 {
2579 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2580 	int create = 0, update_lft = 0;
2581 
2582 	if (!ifp && valid_lft) {
2583 		int max_addresses = in6_dev->cnf.max_addresses;
2584 		struct ifa6_config cfg = {
2585 			.pfx = addr,
2586 			.plen = pinfo->prefix_len,
2587 			.ifa_flags = addr_flags,
2588 			.valid_lft = valid_lft,
2589 			.preferred_lft = prefered_lft,
2590 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2591 		};
2592 
2593 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2594 		if ((net->ipv6.devconf_all->optimistic_dad ||
2595 		     in6_dev->cnf.optimistic_dad) &&
2596 		    !net->ipv6.devconf_all->forwarding && sllao)
2597 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2598 #endif
2599 
2600 		/* Do not allow to create too much of autoconfigured
2601 		 * addresses; this would be too easy way to crash kernel.
2602 		 */
2603 		if (!max_addresses ||
2604 		    ipv6_count_addresses(in6_dev) < max_addresses)
2605 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2606 
2607 		if (IS_ERR_OR_NULL(ifp))
2608 			return -1;
2609 
2610 		create = 1;
2611 		spin_lock_bh(&ifp->lock);
2612 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2613 		ifp->cstamp = jiffies;
2614 		ifp->tokenized = tokenized;
2615 		spin_unlock_bh(&ifp->lock);
2616 		addrconf_dad_start(ifp);
2617 	}
2618 
2619 	if (ifp) {
2620 		u32 flags;
2621 		unsigned long now;
2622 		u32 stored_lft;
2623 
2624 		/* update lifetime (RFC2462 5.5.3 e) */
2625 		spin_lock_bh(&ifp->lock);
2626 		now = jiffies;
2627 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2628 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2629 		else
2630 			stored_lft = 0;
2631 		if (!create && stored_lft) {
2632 			const u32 minimum_lft = min_t(u32,
2633 				stored_lft, MIN_VALID_LIFETIME);
2634 			valid_lft = max(valid_lft, minimum_lft);
2635 
2636 			/* RFC4862 Section 5.5.3e:
2637 			 * "Note that the preferred lifetime of the
2638 			 *  corresponding address is always reset to
2639 			 *  the Preferred Lifetime in the received
2640 			 *  Prefix Information option, regardless of
2641 			 *  whether the valid lifetime is also reset or
2642 			 *  ignored."
2643 			 *
2644 			 * So we should always update prefered_lft here.
2645 			 */
2646 			update_lft = 1;
2647 		}
2648 
2649 		if (update_lft) {
2650 			ifp->valid_lft = valid_lft;
2651 			ifp->prefered_lft = prefered_lft;
2652 			ifp->tstamp = now;
2653 			flags = ifp->flags;
2654 			ifp->flags &= ~IFA_F_DEPRECATED;
2655 			spin_unlock_bh(&ifp->lock);
2656 
2657 			if (!(flags&IFA_F_TENTATIVE))
2658 				ipv6_ifa_notify(0, ifp);
2659 		} else
2660 			spin_unlock_bh(&ifp->lock);
2661 
2662 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2663 				 create, now);
2664 
2665 		in6_ifa_put(ifp);
2666 		addrconf_verify();
2667 	}
2668 
2669 	return 0;
2670 }
2671 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2672 
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2673 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2674 {
2675 	struct prefix_info *pinfo;
2676 	__u32 valid_lft;
2677 	__u32 prefered_lft;
2678 	int addr_type, err;
2679 	u32 addr_flags = 0;
2680 	struct inet6_dev *in6_dev;
2681 	struct net *net = dev_net(dev);
2682 
2683 	pinfo = (struct prefix_info *) opt;
2684 
2685 	if (len < sizeof(struct prefix_info)) {
2686 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2687 		return;
2688 	}
2689 
2690 	/*
2691 	 *	Validation checks ([ADDRCONF], page 19)
2692 	 */
2693 
2694 	addr_type = ipv6_addr_type(&pinfo->prefix);
2695 
2696 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2697 		return;
2698 
2699 	valid_lft = ntohl(pinfo->valid);
2700 	prefered_lft = ntohl(pinfo->prefered);
2701 
2702 	if (prefered_lft > valid_lft) {
2703 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2704 		return;
2705 	}
2706 
2707 	in6_dev = in6_dev_get(dev);
2708 
2709 	if (!in6_dev) {
2710 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2711 				    dev->name);
2712 		return;
2713 	}
2714 
2715 	/*
2716 	 *	Two things going on here:
2717 	 *	1) Add routes for on-link prefixes
2718 	 *	2) Configure prefixes with the auto flag set
2719 	 */
2720 
2721 	if (pinfo->onlink) {
2722 		struct fib6_info *rt;
2723 		unsigned long rt_expires;
2724 
2725 		/* Avoid arithmetic overflow. Really, we could
2726 		 * save rt_expires in seconds, likely valid_lft,
2727 		 * but it would require division in fib gc, that it
2728 		 * not good.
2729 		 */
2730 		if (HZ > USER_HZ)
2731 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2732 		else
2733 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2734 
2735 		if (addrconf_finite_timeout(rt_expires))
2736 			rt_expires *= HZ;
2737 
2738 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2739 					       pinfo->prefix_len,
2740 					       dev,
2741 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2742 					       RTF_DEFAULT, true);
2743 
2744 		if (rt) {
2745 			/* Autoconf prefix route */
2746 			if (valid_lft == 0) {
2747 				ip6_del_rt(net, rt, false);
2748 				rt = NULL;
2749 			} else if (addrconf_finite_timeout(rt_expires)) {
2750 				/* not infinity */
2751 				fib6_set_expires(rt, jiffies + rt_expires);
2752 			} else {
2753 				fib6_clean_expires(rt);
2754 			}
2755 		} else if (valid_lft) {
2756 			clock_t expires = 0;
2757 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2758 			if (addrconf_finite_timeout(rt_expires)) {
2759 				/* not infinity */
2760 				flags |= RTF_EXPIRES;
2761 				expires = jiffies_to_clock_t(rt_expires);
2762 			}
2763 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2764 					      0, dev, expires, flags,
2765 					      GFP_ATOMIC);
2766 		}
2767 		fib6_info_release(rt);
2768 	}
2769 
2770 	/* Try to figure out our local address for this prefix */
2771 
2772 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2773 		struct in6_addr addr;
2774 		bool tokenized = false, dev_addr_generated = false;
2775 
2776 		if (pinfo->prefix_len == 64) {
2777 			memcpy(&addr, &pinfo->prefix, 8);
2778 
2779 			if (!ipv6_addr_any(&in6_dev->token)) {
2780 				read_lock_bh(&in6_dev->lock);
2781 				memcpy(addr.s6_addr + 8,
2782 				       in6_dev->token.s6_addr + 8, 8);
2783 				read_unlock_bh(&in6_dev->lock);
2784 				tokenized = true;
2785 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2786 				   !ipv6_generate_stable_address(&addr, 0,
2787 								 in6_dev)) {
2788 				addr_flags |= IFA_F_STABLE_PRIVACY;
2789 				goto ok;
2790 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2791 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2792 				goto put;
2793 			} else {
2794 				dev_addr_generated = true;
2795 			}
2796 			goto ok;
2797 		}
2798 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2799 				    pinfo->prefix_len);
2800 		goto put;
2801 
2802 ok:
2803 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2804 						   &addr, addr_type,
2805 						   addr_flags, sllao,
2806 						   tokenized, valid_lft,
2807 						   prefered_lft);
2808 		if (err)
2809 			goto put;
2810 
2811 		/* Ignore error case here because previous prefix add addr was
2812 		 * successful which will be notified.
2813 		 */
2814 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2815 					      addr_type, addr_flags, sllao,
2816 					      tokenized, valid_lft,
2817 					      prefered_lft,
2818 					      dev_addr_generated);
2819 	}
2820 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2821 put:
2822 	in6_dev_put(in6_dev);
2823 }
2824 
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2825 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2826 		struct in6_ifreq *ireq)
2827 {
2828 	struct ip_tunnel_parm p = { };
2829 	int err;
2830 
2831 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2832 		return -EADDRNOTAVAIL;
2833 
2834 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2835 	p.iph.version = 4;
2836 	p.iph.ihl = 5;
2837 	p.iph.protocol = IPPROTO_IPV6;
2838 	p.iph.ttl = 64;
2839 
2840 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2841 		return -EOPNOTSUPP;
2842 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2843 	if (err)
2844 		return err;
2845 
2846 	dev = __dev_get_by_name(net, p.name);
2847 	if (!dev)
2848 		return -ENOBUFS;
2849 	return dev_open(dev, NULL);
2850 }
2851 
2852 /*
2853  *	Set destination address.
2854  *	Special case for SIT interfaces where we create a new "virtual"
2855  *	device.
2856  */
addrconf_set_dstaddr(struct net * net,void __user * arg)2857 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2858 {
2859 	struct net_device *dev;
2860 	struct in6_ifreq ireq;
2861 	int err = -ENODEV;
2862 
2863 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2864 		return -ENODEV;
2865 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2866 		return -EFAULT;
2867 
2868 	rtnl_lock();
2869 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2870 	if (dev && dev->type == ARPHRD_SIT)
2871 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2872 	rtnl_unlock();
2873 	return err;
2874 }
2875 
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2876 static int ipv6_mc_config(struct sock *sk, bool join,
2877 			  const struct in6_addr *addr, int ifindex)
2878 {
2879 	int ret;
2880 
2881 	ASSERT_RTNL();
2882 
2883 	lock_sock(sk);
2884 	if (join)
2885 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2886 	else
2887 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2888 	release_sock(sk);
2889 
2890 	return ret;
2891 }
2892 
2893 /*
2894  *	Manual configuration of address on an interface
2895  */
inet6_addr_add(struct net * net,int ifindex,struct ifa6_config * cfg,struct netlink_ext_ack * extack)2896 static int inet6_addr_add(struct net *net, int ifindex,
2897 			  struct ifa6_config *cfg,
2898 			  struct netlink_ext_ack *extack)
2899 {
2900 	struct inet6_ifaddr *ifp;
2901 	struct inet6_dev *idev;
2902 	struct net_device *dev;
2903 	unsigned long timeout;
2904 	clock_t expires;
2905 	u32 flags;
2906 
2907 	ASSERT_RTNL();
2908 
2909 	if (cfg->plen > 128)
2910 		return -EINVAL;
2911 
2912 	/* check the lifetime */
2913 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2914 		return -EINVAL;
2915 
2916 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2917 		return -EINVAL;
2918 
2919 	dev = __dev_get_by_index(net, ifindex);
2920 	if (!dev)
2921 		return -ENODEV;
2922 
2923 	idev = addrconf_add_dev(dev);
2924 	if (IS_ERR(idev))
2925 		return PTR_ERR(idev);
2926 
2927 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2928 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2929 					 true, cfg->pfx, ifindex);
2930 
2931 		if (ret < 0)
2932 			return ret;
2933 	}
2934 
2935 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2936 
2937 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2938 	if (addrconf_finite_timeout(timeout)) {
2939 		expires = jiffies_to_clock_t(timeout * HZ);
2940 		cfg->valid_lft = timeout;
2941 		flags = RTF_EXPIRES;
2942 	} else {
2943 		expires = 0;
2944 		flags = 0;
2945 		cfg->ifa_flags |= IFA_F_PERMANENT;
2946 	}
2947 
2948 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2949 	if (addrconf_finite_timeout(timeout)) {
2950 		if (timeout == 0)
2951 			cfg->ifa_flags |= IFA_F_DEPRECATED;
2952 		cfg->preferred_lft = timeout;
2953 	}
2954 
2955 	ifp = ipv6_add_addr(idev, cfg, true, extack);
2956 	if (!IS_ERR(ifp)) {
2957 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2958 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2959 					      ifp->rt_priority, dev, expires,
2960 					      flags, GFP_KERNEL);
2961 		}
2962 
2963 		/* Send a netlink notification if DAD is enabled and
2964 		 * optimistic flag is not set
2965 		 */
2966 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2967 			ipv6_ifa_notify(0, ifp);
2968 		/*
2969 		 * Note that section 3.1 of RFC 4429 indicates
2970 		 * that the Optimistic flag should not be set for
2971 		 * manually configured addresses
2972 		 */
2973 		addrconf_dad_start(ifp);
2974 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2975 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
2976 					 cfg->preferred_lft, true, jiffies);
2977 		in6_ifa_put(ifp);
2978 		addrconf_verify_rtnl();
2979 		return 0;
2980 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2981 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2982 			       cfg->pfx, ifindex);
2983 	}
2984 
2985 	return PTR_ERR(ifp);
2986 }
2987 
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen)2988 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2989 			  const struct in6_addr *pfx, unsigned int plen)
2990 {
2991 	struct inet6_ifaddr *ifp;
2992 	struct inet6_dev *idev;
2993 	struct net_device *dev;
2994 
2995 	if (plen > 128)
2996 		return -EINVAL;
2997 
2998 	dev = __dev_get_by_index(net, ifindex);
2999 	if (!dev)
3000 		return -ENODEV;
3001 
3002 	idev = __in6_dev_get(dev);
3003 	if (!idev)
3004 		return -ENXIO;
3005 
3006 	read_lock_bh(&idev->lock);
3007 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3008 		if (ifp->prefix_len == plen &&
3009 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3010 			in6_ifa_hold(ifp);
3011 			read_unlock_bh(&idev->lock);
3012 
3013 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3014 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
3015 				manage_tempaddrs(idev, ifp, 0, 0, false,
3016 						 jiffies);
3017 			ipv6_del_addr(ifp);
3018 			addrconf_verify_rtnl();
3019 			if (ipv6_addr_is_multicast(pfx)) {
3020 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3021 					       false, pfx, dev->ifindex);
3022 			}
3023 			return 0;
3024 		}
3025 	}
3026 	read_unlock_bh(&idev->lock);
3027 	return -EADDRNOTAVAIL;
3028 }
3029 
3030 
addrconf_add_ifaddr(struct net * net,void __user * arg)3031 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3032 {
3033 	struct ifa6_config cfg = {
3034 		.ifa_flags = IFA_F_PERMANENT,
3035 		.preferred_lft = INFINITY_LIFE_TIME,
3036 		.valid_lft = INFINITY_LIFE_TIME,
3037 	};
3038 	struct in6_ifreq ireq;
3039 	int err;
3040 
3041 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3042 		return -EPERM;
3043 
3044 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3045 		return -EFAULT;
3046 
3047 	cfg.pfx = &ireq.ifr6_addr;
3048 	cfg.plen = ireq.ifr6_prefixlen;
3049 
3050 	rtnl_lock();
3051 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3052 	rtnl_unlock();
3053 	return err;
3054 }
3055 
addrconf_del_ifaddr(struct net * net,void __user * arg)3056 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3057 {
3058 	struct in6_ifreq ireq;
3059 	int err;
3060 
3061 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3062 		return -EPERM;
3063 
3064 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3065 		return -EFAULT;
3066 
3067 	rtnl_lock();
3068 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3069 			     ireq.ifr6_prefixlen);
3070 	rtnl_unlock();
3071 	return err;
3072 }
3073 
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope)3074 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3075 		     int plen, int scope)
3076 {
3077 	struct inet6_ifaddr *ifp;
3078 	struct ifa6_config cfg = {
3079 		.pfx = addr,
3080 		.plen = plen,
3081 		.ifa_flags = IFA_F_PERMANENT,
3082 		.valid_lft = INFINITY_LIFE_TIME,
3083 		.preferred_lft = INFINITY_LIFE_TIME,
3084 		.scope = scope
3085 	};
3086 
3087 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3088 	if (!IS_ERR(ifp)) {
3089 		spin_lock_bh(&ifp->lock);
3090 		ifp->flags &= ~IFA_F_TENTATIVE;
3091 		spin_unlock_bh(&ifp->lock);
3092 		rt_genid_bump_ipv6(dev_net(idev->dev));
3093 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3094 		in6_ifa_put(ifp);
3095 	}
3096 }
3097 
3098 #if IS_ENABLED(CONFIG_IPV6_SIT)
sit_add_v4_addrs(struct inet6_dev * idev)3099 static void sit_add_v4_addrs(struct inet6_dev *idev)
3100 {
3101 	struct in6_addr addr;
3102 	struct net_device *dev;
3103 	struct net *net = dev_net(idev->dev);
3104 	int scope, plen;
3105 	u32 pflags = 0;
3106 
3107 	ASSERT_RTNL();
3108 
3109 	memset(&addr, 0, sizeof(struct in6_addr));
3110 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3111 
3112 	if (idev->dev->flags&IFF_POINTOPOINT) {
3113 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3114 			return;
3115 
3116 		addr.s6_addr32[0] = htonl(0xfe800000);
3117 		scope = IFA_LINK;
3118 		plen = 64;
3119 	} else {
3120 		scope = IPV6_ADDR_COMPATv4;
3121 		plen = 96;
3122 		pflags |= RTF_NONEXTHOP;
3123 	}
3124 
3125 	if (addr.s6_addr32[3]) {
3126 		add_addr(idev, &addr, plen, scope);
3127 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3128 				      GFP_KERNEL);
3129 		return;
3130 	}
3131 
3132 	for_each_netdev(net, dev) {
3133 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3134 		if (in_dev && (dev->flags & IFF_UP)) {
3135 			struct in_ifaddr *ifa;
3136 			int flag = scope;
3137 
3138 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3139 				addr.s6_addr32[3] = ifa->ifa_local;
3140 
3141 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3142 					continue;
3143 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3144 					if (idev->dev->flags&IFF_POINTOPOINT)
3145 						continue;
3146 					flag |= IFA_HOST;
3147 				}
3148 
3149 				add_addr(idev, &addr, plen, flag);
3150 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3151 						      0, pflags, GFP_KERNEL);
3152 			}
3153 		}
3154 	}
3155 }
3156 #endif
3157 
init_loopback(struct net_device * dev)3158 static void init_loopback(struct net_device *dev)
3159 {
3160 	struct inet6_dev  *idev;
3161 
3162 	/* ::1 */
3163 
3164 	ASSERT_RTNL();
3165 
3166 	idev = ipv6_find_idev(dev);
3167 	if (IS_ERR(idev)) {
3168 		pr_debug("%s: add_dev failed\n", __func__);
3169 		return;
3170 	}
3171 
3172 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3173 }
3174 
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3175 void addrconf_add_linklocal(struct inet6_dev *idev,
3176 			    const struct in6_addr *addr, u32 flags)
3177 {
3178 	struct ifa6_config cfg = {
3179 		.pfx = addr,
3180 		.plen = 64,
3181 		.ifa_flags = flags | IFA_F_PERMANENT,
3182 		.valid_lft = INFINITY_LIFE_TIME,
3183 		.preferred_lft = INFINITY_LIFE_TIME,
3184 		.scope = IFA_LINK
3185 	};
3186 	struct inet6_ifaddr *ifp;
3187 
3188 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3189 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3190 	     idev->cnf.optimistic_dad) &&
3191 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3192 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3193 #endif
3194 
3195 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3196 	if (!IS_ERR(ifp)) {
3197 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3198 				      0, 0, GFP_ATOMIC);
3199 		addrconf_dad_start(ifp);
3200 		in6_ifa_put(ifp);
3201 	}
3202 }
3203 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3204 
ipv6_reserved_interfaceid(struct in6_addr address)3205 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3206 {
3207 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3208 		return true;
3209 
3210 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3211 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3212 		return true;
3213 
3214 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3215 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3216 		return true;
3217 
3218 	return false;
3219 }
3220 
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3221 static int ipv6_generate_stable_address(struct in6_addr *address,
3222 					u8 dad_count,
3223 					const struct inet6_dev *idev)
3224 {
3225 	static DEFINE_SPINLOCK(lock);
3226 	static __u32 digest[SHA1_DIGEST_WORDS];
3227 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3228 
3229 	static union {
3230 		char __data[SHA1_BLOCK_SIZE];
3231 		struct {
3232 			struct in6_addr secret;
3233 			__be32 prefix[2];
3234 			unsigned char hwaddr[MAX_ADDR_LEN];
3235 			u8 dad_count;
3236 		} __packed;
3237 	} data;
3238 
3239 	struct in6_addr secret;
3240 	struct in6_addr temp;
3241 	struct net *net = dev_net(idev->dev);
3242 
3243 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3244 
3245 	if (idev->cnf.stable_secret.initialized)
3246 		secret = idev->cnf.stable_secret.secret;
3247 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3248 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3249 	else
3250 		return -1;
3251 
3252 retry:
3253 	spin_lock_bh(&lock);
3254 
3255 	sha1_init(digest);
3256 	memset(&data, 0, sizeof(data));
3257 	memset(workspace, 0, sizeof(workspace));
3258 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3259 	data.prefix[0] = address->s6_addr32[0];
3260 	data.prefix[1] = address->s6_addr32[1];
3261 	data.secret = secret;
3262 	data.dad_count = dad_count;
3263 
3264 	sha1_transform(digest, data.__data, workspace);
3265 
3266 	temp = *address;
3267 	temp.s6_addr32[2] = (__force __be32)digest[0];
3268 	temp.s6_addr32[3] = (__force __be32)digest[1];
3269 
3270 	spin_unlock_bh(&lock);
3271 
3272 	if (ipv6_reserved_interfaceid(temp)) {
3273 		dad_count++;
3274 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3275 			return -1;
3276 		goto retry;
3277 	}
3278 
3279 	*address = temp;
3280 	return 0;
3281 }
3282 
ipv6_gen_mode_random_init(struct inet6_dev * idev)3283 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3284 {
3285 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3286 
3287 	if (s->initialized)
3288 		return;
3289 	s = &idev->cnf.stable_secret;
3290 	get_random_bytes(&s->secret, sizeof(s->secret));
3291 	s->initialized = true;
3292 }
3293 
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3294 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3295 {
3296 	struct in6_addr addr;
3297 
3298 	/* no link local addresses on L3 master devices */
3299 	if (netif_is_l3_master(idev->dev))
3300 		return;
3301 
3302 	/* no link local addresses on devices flagged as slaves */
3303 	if (idev->dev->flags & IFF_SLAVE)
3304 		return;
3305 
3306 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3307 
3308 	switch (idev->cnf.addr_gen_mode) {
3309 	case IN6_ADDR_GEN_MODE_RANDOM:
3310 		ipv6_gen_mode_random_init(idev);
3311 		fallthrough;
3312 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3313 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3314 			addrconf_add_linklocal(idev, &addr,
3315 					       IFA_F_STABLE_PRIVACY);
3316 		else if (prefix_route)
3317 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3318 					      0, 0, GFP_KERNEL);
3319 		break;
3320 	case IN6_ADDR_GEN_MODE_EUI64:
3321 		/* addrconf_add_linklocal also adds a prefix_route and we
3322 		 * only need to care about prefix routes if ipv6_generate_eui64
3323 		 * couldn't generate one.
3324 		 */
3325 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3326 			addrconf_add_linklocal(idev, &addr, 0);
3327 		else if (prefix_route)
3328 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3329 					      0, 0, GFP_KERNEL);
3330 		break;
3331 	case IN6_ADDR_GEN_MODE_NONE:
3332 	default:
3333 		/* will not add any link local address */
3334 		break;
3335 	}
3336 }
3337 
addrconf_dev_config(struct net_device * dev)3338 static void addrconf_dev_config(struct net_device *dev)
3339 {
3340 	struct inet6_dev *idev;
3341 
3342 	ASSERT_RTNL();
3343 
3344 	if ((dev->type != ARPHRD_ETHER) &&
3345 	    (dev->type != ARPHRD_FDDI) &&
3346 	    (dev->type != ARPHRD_ARCNET) &&
3347 	    (dev->type != ARPHRD_INFINIBAND) &&
3348 	    (dev->type != ARPHRD_IEEE1394) &&
3349 	    (dev->type != ARPHRD_TUNNEL6) &&
3350 	    (dev->type != ARPHRD_6LOWPAN) &&
3351 	    (dev->type != ARPHRD_IP6GRE) &&
3352 	    (dev->type != ARPHRD_IPGRE) &&
3353 	    (dev->type != ARPHRD_TUNNEL) &&
3354 	    (dev->type != ARPHRD_NONE) &&
3355 	    (dev->type != ARPHRD_RAWIP)) {
3356 		/* Alas, we support only Ethernet autoconfiguration. */
3357 		idev = __in6_dev_get(dev);
3358 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3359 		    dev->flags & IFF_MULTICAST)
3360 			ipv6_mc_up(idev);
3361 		return;
3362 	}
3363 
3364 	idev = addrconf_add_dev(dev);
3365 	if (IS_ERR(idev))
3366 		return;
3367 
3368 	/* this device type has no EUI support */
3369 	if (dev->type == ARPHRD_NONE &&
3370 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3371 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3372 
3373 	addrconf_addr_gen(idev, false);
3374 }
3375 
3376 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3377 static void addrconf_sit_config(struct net_device *dev)
3378 {
3379 	struct inet6_dev *idev;
3380 
3381 	ASSERT_RTNL();
3382 
3383 	/*
3384 	 * Configure the tunnel with one of our IPv4
3385 	 * addresses... we should configure all of
3386 	 * our v4 addrs in the tunnel
3387 	 */
3388 
3389 	idev = ipv6_find_idev(dev);
3390 	if (IS_ERR(idev)) {
3391 		pr_debug("%s: add_dev failed\n", __func__);
3392 		return;
3393 	}
3394 
3395 	if (dev->priv_flags & IFF_ISATAP) {
3396 		addrconf_addr_gen(idev, false);
3397 		return;
3398 	}
3399 
3400 	sit_add_v4_addrs(idev);
3401 
3402 	if (dev->flags&IFF_POINTOPOINT)
3403 		addrconf_add_mroute(dev);
3404 }
3405 #endif
3406 
3407 #if IS_ENABLED(CONFIG_NET_IPGRE)
addrconf_gre_config(struct net_device * dev)3408 static void addrconf_gre_config(struct net_device *dev)
3409 {
3410 	struct inet6_dev *idev;
3411 
3412 	ASSERT_RTNL();
3413 
3414 	idev = ipv6_find_idev(dev);
3415 	if (IS_ERR(idev)) {
3416 		pr_debug("%s: add_dev failed\n", __func__);
3417 		return;
3418 	}
3419 
3420 	addrconf_addr_gen(idev, true);
3421 	if (dev->flags & IFF_POINTOPOINT)
3422 		addrconf_add_mroute(dev);
3423 }
3424 #endif
3425 
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3426 static int fixup_permanent_addr(struct net *net,
3427 				struct inet6_dev *idev,
3428 				struct inet6_ifaddr *ifp)
3429 {
3430 	/* !fib6_node means the host route was removed from the
3431 	 * FIB, for example, if 'lo' device is taken down. In that
3432 	 * case regenerate the host route.
3433 	 */
3434 	if (!ifp->rt || !ifp->rt->fib6_node) {
3435 		struct fib6_info *f6i, *prev;
3436 
3437 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3438 					 GFP_ATOMIC);
3439 		if (IS_ERR(f6i))
3440 			return PTR_ERR(f6i);
3441 
3442 		/* ifp->rt can be accessed outside of rtnl */
3443 		spin_lock(&ifp->lock);
3444 		prev = ifp->rt;
3445 		ifp->rt = f6i;
3446 		spin_unlock(&ifp->lock);
3447 
3448 		fib6_info_release(prev);
3449 	}
3450 
3451 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3452 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3453 				      ifp->rt_priority, idev->dev, 0, 0,
3454 				      GFP_ATOMIC);
3455 	}
3456 
3457 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3458 		addrconf_dad_start(ifp);
3459 
3460 	return 0;
3461 }
3462 
addrconf_permanent_addr(struct net * net,struct net_device * dev)3463 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3464 {
3465 	struct inet6_ifaddr *ifp, *tmp;
3466 	struct inet6_dev *idev;
3467 
3468 	idev = __in6_dev_get(dev);
3469 	if (!idev)
3470 		return;
3471 
3472 	write_lock_bh(&idev->lock);
3473 
3474 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3475 		if ((ifp->flags & IFA_F_PERMANENT) &&
3476 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3477 			write_unlock_bh(&idev->lock);
3478 			in6_ifa_hold(ifp);
3479 			ipv6_del_addr(ifp);
3480 			write_lock_bh(&idev->lock);
3481 
3482 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3483 					     idev->dev->name, &ifp->addr);
3484 		}
3485 	}
3486 
3487 	write_unlock_bh(&idev->lock);
3488 }
3489 
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3490 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3491 			   void *ptr)
3492 {
3493 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3494 	struct netdev_notifier_change_info *change_info;
3495 	struct netdev_notifier_changeupper_info *info;
3496 	struct inet6_dev *idev = __in6_dev_get(dev);
3497 	struct net *net = dev_net(dev);
3498 	int run_pending = 0;
3499 	int err;
3500 
3501 	switch (event) {
3502 	case NETDEV_REGISTER:
3503 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3504 			idev = ipv6_add_dev(dev);
3505 			if (IS_ERR(idev))
3506 				return notifier_from_errno(PTR_ERR(idev));
3507 		}
3508 		break;
3509 
3510 	case NETDEV_CHANGEMTU:
3511 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3512 		if (dev->mtu < IPV6_MIN_MTU) {
3513 			addrconf_ifdown(dev, dev != net->loopback_dev);
3514 			break;
3515 		}
3516 
3517 		if (idev) {
3518 			rt6_mtu_change(dev, dev->mtu);
3519 			idev->cnf.mtu6 = dev->mtu;
3520 			break;
3521 		}
3522 
3523 		/* allocate new idev */
3524 		idev = ipv6_add_dev(dev);
3525 		if (IS_ERR(idev))
3526 			break;
3527 
3528 		/* device is still not ready */
3529 		if (!(idev->if_flags & IF_READY))
3530 			break;
3531 
3532 		run_pending = 1;
3533 		fallthrough;
3534 	case NETDEV_UP:
3535 	case NETDEV_CHANGE:
3536 		if (dev->flags & IFF_SLAVE)
3537 			break;
3538 
3539 		if (idev && idev->cnf.disable_ipv6)
3540 			break;
3541 
3542 		if (event == NETDEV_UP) {
3543 			/* restore routes for permanent addresses */
3544 			addrconf_permanent_addr(net, dev);
3545 
3546 			if (!addrconf_link_ready(dev)) {
3547 				/* device is not ready yet. */
3548 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3549 					 dev->name);
3550 				break;
3551 			}
3552 
3553 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3554 				idev = ipv6_add_dev(dev);
3555 
3556 			if (!IS_ERR_OR_NULL(idev)) {
3557 				idev->if_flags |= IF_READY;
3558 				run_pending = 1;
3559 			}
3560 		} else if (event == NETDEV_CHANGE) {
3561 			if (!addrconf_link_ready(dev)) {
3562 				/* device is still not ready. */
3563 				rt6_sync_down_dev(dev, event);
3564 				break;
3565 			}
3566 
3567 			if (!IS_ERR_OR_NULL(idev)) {
3568 				if (idev->if_flags & IF_READY) {
3569 					/* device is already configured -
3570 					 * but resend MLD reports, we might
3571 					 * have roamed and need to update
3572 					 * multicast snooping switches
3573 					 */
3574 					ipv6_mc_up(idev);
3575 					change_info = ptr;
3576 					if (change_info->flags_changed & IFF_NOARP)
3577 						addrconf_dad_run(idev, true);
3578 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3579 					break;
3580 				}
3581 				idev->if_flags |= IF_READY;
3582 			}
3583 
3584 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3585 				dev->name);
3586 
3587 			run_pending = 1;
3588 		}
3589 
3590 		switch (dev->type) {
3591 #if IS_ENABLED(CONFIG_IPV6_SIT)
3592 		case ARPHRD_SIT:
3593 			addrconf_sit_config(dev);
3594 			break;
3595 #endif
3596 #if IS_ENABLED(CONFIG_NET_IPGRE)
3597 		case ARPHRD_IPGRE:
3598 			addrconf_gre_config(dev);
3599 			break;
3600 #endif
3601 		case ARPHRD_LOOPBACK:
3602 			init_loopback(dev);
3603 			break;
3604 
3605 		default:
3606 			addrconf_dev_config(dev);
3607 			break;
3608 		}
3609 
3610 		if (!IS_ERR_OR_NULL(idev)) {
3611 			if (run_pending)
3612 				addrconf_dad_run(idev, false);
3613 
3614 			/* Device has an address by now */
3615 			rt6_sync_up(dev, RTNH_F_DEAD);
3616 
3617 			/*
3618 			 * If the MTU changed during the interface down,
3619 			 * when the interface up, the changed MTU must be
3620 			 * reflected in the idev as well as routers.
3621 			 */
3622 			if (idev->cnf.mtu6 != dev->mtu &&
3623 			    dev->mtu >= IPV6_MIN_MTU) {
3624 				rt6_mtu_change(dev, dev->mtu);
3625 				idev->cnf.mtu6 = dev->mtu;
3626 			}
3627 			idev->tstamp = jiffies;
3628 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3629 
3630 			/*
3631 			 * If the changed mtu during down is lower than
3632 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3633 			 */
3634 			if (dev->mtu < IPV6_MIN_MTU)
3635 				addrconf_ifdown(dev, dev != net->loopback_dev);
3636 		}
3637 		break;
3638 
3639 	case NETDEV_DOWN:
3640 	case NETDEV_UNREGISTER:
3641 		/*
3642 		 *	Remove all addresses from this interface.
3643 		 */
3644 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3645 		break;
3646 
3647 	case NETDEV_CHANGENAME:
3648 		if (idev) {
3649 			snmp6_unregister_dev(idev);
3650 			addrconf_sysctl_unregister(idev);
3651 			err = addrconf_sysctl_register(idev);
3652 			if (err)
3653 				return notifier_from_errno(err);
3654 			err = snmp6_register_dev(idev);
3655 			if (err) {
3656 				addrconf_sysctl_unregister(idev);
3657 				return notifier_from_errno(err);
3658 			}
3659 		}
3660 		break;
3661 
3662 	case NETDEV_PRE_TYPE_CHANGE:
3663 	case NETDEV_POST_TYPE_CHANGE:
3664 		if (idev)
3665 			addrconf_type_change(dev, event);
3666 		break;
3667 
3668 	case NETDEV_CHANGEUPPER:
3669 		info = ptr;
3670 
3671 		/* flush all routes if dev is linked to or unlinked from
3672 		 * an L3 master device (e.g., VRF)
3673 		 */
3674 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3675 			addrconf_ifdown(dev, false);
3676 	}
3677 
3678 	return NOTIFY_OK;
3679 }
3680 
3681 /*
3682  *	addrconf module should be notified of a device going up
3683  */
3684 static struct notifier_block ipv6_dev_notf = {
3685 	.notifier_call = addrconf_notify,
3686 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3687 };
3688 
addrconf_type_change(struct net_device * dev,unsigned long event)3689 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3690 {
3691 	struct inet6_dev *idev;
3692 	ASSERT_RTNL();
3693 
3694 	idev = __in6_dev_get(dev);
3695 
3696 	if (event == NETDEV_POST_TYPE_CHANGE)
3697 		ipv6_mc_remap(idev);
3698 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3699 		ipv6_mc_unmap(idev);
3700 }
3701 
addr_is_local(const struct in6_addr * addr)3702 static bool addr_is_local(const struct in6_addr *addr)
3703 {
3704 	return ipv6_addr_type(addr) &
3705 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3706 }
3707 
addrconf_ifdown(struct net_device * dev,bool unregister)3708 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3709 {
3710 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3711 	struct net *net = dev_net(dev);
3712 	struct inet6_dev *idev;
3713 	struct inet6_ifaddr *ifa, *tmp;
3714 	bool keep_addr = false;
3715 	int state, i;
3716 
3717 	ASSERT_RTNL();
3718 
3719 	rt6_disable_ip(dev, event);
3720 
3721 	idev = __in6_dev_get(dev);
3722 	if (!idev)
3723 		return -ENODEV;
3724 
3725 	/*
3726 	 * Step 1: remove reference to ipv6 device from parent device.
3727 	 *	   Do not dev_put!
3728 	 */
3729 	if (unregister) {
3730 		idev->dead = 1;
3731 
3732 		/* protected by rtnl_lock */
3733 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3734 
3735 		/* Step 1.5: remove snmp6 entry */
3736 		snmp6_unregister_dev(idev);
3737 
3738 	}
3739 
3740 	/* combine the user config with event to determine if permanent
3741 	 * addresses are to be removed from address hash table
3742 	 */
3743 	if (!unregister && !idev->cnf.disable_ipv6) {
3744 		/* aggregate the system setting and interface setting */
3745 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3746 
3747 		if (!_keep_addr)
3748 			_keep_addr = idev->cnf.keep_addr_on_down;
3749 
3750 		keep_addr = (_keep_addr > 0);
3751 	}
3752 
3753 	/* Step 2: clear hash table */
3754 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3755 		struct hlist_head *h = &inet6_addr_lst[i];
3756 
3757 		spin_lock_bh(&addrconf_hash_lock);
3758 restart:
3759 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3760 			if (ifa->idev == idev) {
3761 				addrconf_del_dad_work(ifa);
3762 				/* combined flag + permanent flag decide if
3763 				 * address is retained on a down event
3764 				 */
3765 				if (!keep_addr ||
3766 				    !(ifa->flags & IFA_F_PERMANENT) ||
3767 				    addr_is_local(&ifa->addr)) {
3768 					hlist_del_init_rcu(&ifa->addr_lst);
3769 					goto restart;
3770 				}
3771 			}
3772 		}
3773 		spin_unlock_bh(&addrconf_hash_lock);
3774 	}
3775 
3776 	write_lock_bh(&idev->lock);
3777 
3778 	addrconf_del_rs_timer(idev);
3779 
3780 	/* Step 2: clear flags for stateless addrconf */
3781 	if (!unregister)
3782 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3783 
3784 	/* Step 3: clear tempaddr list */
3785 	while (!list_empty(&idev->tempaddr_list)) {
3786 		ifa = list_first_entry(&idev->tempaddr_list,
3787 				       struct inet6_ifaddr, tmp_list);
3788 		list_del(&ifa->tmp_list);
3789 		write_unlock_bh(&idev->lock);
3790 		spin_lock_bh(&ifa->lock);
3791 
3792 		if (ifa->ifpub) {
3793 			in6_ifa_put(ifa->ifpub);
3794 			ifa->ifpub = NULL;
3795 		}
3796 		spin_unlock_bh(&ifa->lock);
3797 		in6_ifa_put(ifa);
3798 		write_lock_bh(&idev->lock);
3799 	}
3800 
3801 	list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3802 		struct fib6_info *rt = NULL;
3803 		bool keep;
3804 
3805 		addrconf_del_dad_work(ifa);
3806 
3807 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3808 			!addr_is_local(&ifa->addr);
3809 
3810 		write_unlock_bh(&idev->lock);
3811 		spin_lock_bh(&ifa->lock);
3812 
3813 		if (keep) {
3814 			/* set state to skip the notifier below */
3815 			state = INET6_IFADDR_STATE_DEAD;
3816 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3817 			if (!(ifa->flags & IFA_F_NODAD))
3818 				ifa->flags |= IFA_F_TENTATIVE;
3819 
3820 			rt = ifa->rt;
3821 			ifa->rt = NULL;
3822 		} else {
3823 			state = ifa->state;
3824 			ifa->state = INET6_IFADDR_STATE_DEAD;
3825 		}
3826 
3827 		spin_unlock_bh(&ifa->lock);
3828 
3829 		if (rt)
3830 			ip6_del_rt(net, rt, false);
3831 
3832 		if (state != INET6_IFADDR_STATE_DEAD) {
3833 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3834 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3835 		} else {
3836 			if (idev->cnf.forwarding)
3837 				addrconf_leave_anycast(ifa);
3838 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3839 		}
3840 
3841 		write_lock_bh(&idev->lock);
3842 		if (!keep) {
3843 			list_del_rcu(&ifa->if_list);
3844 			in6_ifa_put(ifa);
3845 		}
3846 	}
3847 
3848 	write_unlock_bh(&idev->lock);
3849 
3850 	/* Step 5: Discard anycast and multicast list */
3851 	if (unregister) {
3852 		ipv6_ac_destroy_dev(idev);
3853 		ipv6_mc_destroy_dev(idev);
3854 	} else {
3855 		ipv6_mc_down(idev);
3856 	}
3857 
3858 	idev->tstamp = jiffies;
3859 
3860 	/* Last: Shot the device (if unregistered) */
3861 	if (unregister) {
3862 		addrconf_sysctl_unregister(idev);
3863 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3864 		neigh_ifdown(&nd_tbl, dev);
3865 		in6_dev_put(idev);
3866 	}
3867 	return 0;
3868 }
3869 
addrconf_rs_timer(struct timer_list * t)3870 static void addrconf_rs_timer(struct timer_list *t)
3871 {
3872 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3873 	struct net_device *dev = idev->dev;
3874 	struct in6_addr lladdr;
3875 
3876 	write_lock(&idev->lock);
3877 	if (idev->dead || !(idev->if_flags & IF_READY))
3878 		goto out;
3879 
3880 	if (!ipv6_accept_ra(idev))
3881 		goto out;
3882 
3883 	/* Announcement received after solicitation was sent */
3884 	if (idev->if_flags & IF_RA_RCVD)
3885 		goto out;
3886 
3887 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3888 		write_unlock(&idev->lock);
3889 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3890 			ndisc_send_rs(dev, &lladdr,
3891 				      &in6addr_linklocal_allrouters);
3892 		else
3893 			goto put;
3894 
3895 		write_lock(&idev->lock);
3896 		idev->rs_interval = rfc3315_s14_backoff_update(
3897 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3898 		/* The wait after the last probe can be shorter */
3899 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3900 					     idev->cnf.rtr_solicits) ?
3901 				      idev->cnf.rtr_solicit_delay :
3902 				      idev->rs_interval);
3903 	} else {
3904 		/*
3905 		 * Note: we do not support deprecated "all on-link"
3906 		 * assumption any longer.
3907 		 */
3908 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3909 	}
3910 
3911 out:
3912 	write_unlock(&idev->lock);
3913 put:
3914 	in6_dev_put(idev);
3915 }
3916 
3917 /*
3918  *	Duplicate Address Detection
3919  */
addrconf_dad_kick(struct inet6_ifaddr * ifp)3920 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3921 {
3922 	unsigned long rand_num;
3923 	struct inet6_dev *idev = ifp->idev;
3924 	u64 nonce;
3925 
3926 	if (ifp->flags & IFA_F_OPTIMISTIC)
3927 		rand_num = 0;
3928 	else
3929 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3930 
3931 	nonce = 0;
3932 	if (idev->cnf.enhanced_dad ||
3933 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3934 		do
3935 			get_random_bytes(&nonce, 6);
3936 		while (nonce == 0);
3937 	}
3938 	ifp->dad_nonce = nonce;
3939 	ifp->dad_probes = idev->cnf.dad_transmits;
3940 	addrconf_mod_dad_work(ifp, rand_num);
3941 }
3942 
addrconf_dad_begin(struct inet6_ifaddr * ifp)3943 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3944 {
3945 	struct inet6_dev *idev = ifp->idev;
3946 	struct net_device *dev = idev->dev;
3947 	bool bump_id, notify = false;
3948 	struct net *net;
3949 
3950 	addrconf_join_solict(dev, &ifp->addr);
3951 
3952 	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3953 
3954 	read_lock_bh(&idev->lock);
3955 	spin_lock(&ifp->lock);
3956 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3957 		goto out;
3958 
3959 	net = dev_net(dev);
3960 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3961 	    (net->ipv6.devconf_all->accept_dad < 1 &&
3962 	     idev->cnf.accept_dad < 1) ||
3963 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3964 	    ifp->flags & IFA_F_NODAD) {
3965 		bool send_na = false;
3966 
3967 		if (ifp->flags & IFA_F_TENTATIVE &&
3968 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3969 			send_na = true;
3970 		bump_id = ifp->flags & IFA_F_TENTATIVE;
3971 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3972 		spin_unlock(&ifp->lock);
3973 		read_unlock_bh(&idev->lock);
3974 
3975 		addrconf_dad_completed(ifp, bump_id, send_na);
3976 		return;
3977 	}
3978 
3979 	if (!(idev->if_flags & IF_READY)) {
3980 		spin_unlock(&ifp->lock);
3981 		read_unlock_bh(&idev->lock);
3982 		/*
3983 		 * If the device is not ready:
3984 		 * - keep it tentative if it is a permanent address.
3985 		 * - otherwise, kill it.
3986 		 */
3987 		in6_ifa_hold(ifp);
3988 		addrconf_dad_stop(ifp, 0);
3989 		return;
3990 	}
3991 
3992 	/*
3993 	 * Optimistic nodes can start receiving
3994 	 * Frames right away
3995 	 */
3996 	if (ifp->flags & IFA_F_OPTIMISTIC) {
3997 		ip6_ins_rt(net, ifp->rt);
3998 		if (ipv6_use_optimistic_addr(net, idev)) {
3999 			/* Because optimistic nodes can use this address,
4000 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4001 			 */
4002 			notify = true;
4003 		}
4004 	}
4005 
4006 	addrconf_dad_kick(ifp);
4007 out:
4008 	spin_unlock(&ifp->lock);
4009 	read_unlock_bh(&idev->lock);
4010 	if (notify)
4011 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4012 }
4013 
addrconf_dad_start(struct inet6_ifaddr * ifp)4014 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4015 {
4016 	bool begin_dad = false;
4017 
4018 	spin_lock_bh(&ifp->lock);
4019 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4020 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4021 		begin_dad = true;
4022 	}
4023 	spin_unlock_bh(&ifp->lock);
4024 
4025 	if (begin_dad)
4026 		addrconf_mod_dad_work(ifp, 0);
4027 }
4028 
addrconf_dad_work(struct work_struct * w)4029 static void addrconf_dad_work(struct work_struct *w)
4030 {
4031 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4032 						struct inet6_ifaddr,
4033 						dad_work);
4034 	struct inet6_dev *idev = ifp->idev;
4035 	bool bump_id, disable_ipv6 = false;
4036 	struct in6_addr mcaddr;
4037 
4038 	enum {
4039 		DAD_PROCESS,
4040 		DAD_BEGIN,
4041 		DAD_ABORT,
4042 	} action = DAD_PROCESS;
4043 
4044 	rtnl_lock();
4045 
4046 	spin_lock_bh(&ifp->lock);
4047 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4048 		action = DAD_BEGIN;
4049 		ifp->state = INET6_IFADDR_STATE_DAD;
4050 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4051 		action = DAD_ABORT;
4052 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4053 
4054 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4055 		     idev->cnf.accept_dad > 1) &&
4056 		    !idev->cnf.disable_ipv6 &&
4057 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4058 			struct in6_addr addr;
4059 
4060 			addr.s6_addr32[0] = htonl(0xfe800000);
4061 			addr.s6_addr32[1] = 0;
4062 
4063 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4064 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4065 				/* DAD failed for link-local based on MAC */
4066 				idev->cnf.disable_ipv6 = 1;
4067 
4068 				pr_info("%s: IPv6 being disabled!\n",
4069 					ifp->idev->dev->name);
4070 				disable_ipv6 = true;
4071 			}
4072 		}
4073 	}
4074 	spin_unlock_bh(&ifp->lock);
4075 
4076 	if (action == DAD_BEGIN) {
4077 		addrconf_dad_begin(ifp);
4078 		goto out;
4079 	} else if (action == DAD_ABORT) {
4080 		in6_ifa_hold(ifp);
4081 		addrconf_dad_stop(ifp, 1);
4082 		if (disable_ipv6)
4083 			addrconf_ifdown(idev->dev, false);
4084 		goto out;
4085 	}
4086 
4087 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4088 		goto out;
4089 
4090 	write_lock_bh(&idev->lock);
4091 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4092 		write_unlock_bh(&idev->lock);
4093 		goto out;
4094 	}
4095 
4096 	spin_lock(&ifp->lock);
4097 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4098 		spin_unlock(&ifp->lock);
4099 		write_unlock_bh(&idev->lock);
4100 		goto out;
4101 	}
4102 
4103 	if (ifp->dad_probes == 0) {
4104 		bool send_na = false;
4105 
4106 		/*
4107 		 * DAD was successful
4108 		 */
4109 
4110 		if (ifp->flags & IFA_F_TENTATIVE &&
4111 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4112 			send_na = true;
4113 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4114 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4115 		spin_unlock(&ifp->lock);
4116 		write_unlock_bh(&idev->lock);
4117 
4118 		addrconf_dad_completed(ifp, bump_id, send_na);
4119 
4120 		goto out;
4121 	}
4122 
4123 	ifp->dad_probes--;
4124 	addrconf_mod_dad_work(ifp,
4125 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4126 				  HZ/100));
4127 	spin_unlock(&ifp->lock);
4128 	write_unlock_bh(&idev->lock);
4129 
4130 	/* send a neighbour solicitation for our addr */
4131 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4132 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4133 		      ifp->dad_nonce);
4134 out:
4135 	in6_ifa_put(ifp);
4136 	rtnl_unlock();
4137 }
4138 
4139 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4140 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4141 {
4142 	struct inet6_ifaddr *ifpiter;
4143 	struct inet6_dev *idev = ifp->idev;
4144 
4145 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4146 		if (ifpiter->scope > IFA_LINK)
4147 			break;
4148 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4149 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4150 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4151 		    IFA_F_PERMANENT)
4152 			return false;
4153 	}
4154 	return true;
4155 }
4156 
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4157 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4158 				   bool send_na)
4159 {
4160 	struct net_device *dev = ifp->idev->dev;
4161 	struct in6_addr lladdr;
4162 	bool send_rs, send_mld;
4163 
4164 	addrconf_del_dad_work(ifp);
4165 
4166 	/*
4167 	 *	Configure the address for reception. Now it is valid.
4168 	 */
4169 
4170 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4171 
4172 	/* If added prefix is link local and we are prepared to process
4173 	   router advertisements, start sending router solicitations.
4174 	 */
4175 
4176 	read_lock_bh(&ifp->idev->lock);
4177 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4178 	send_rs = send_mld &&
4179 		  ipv6_accept_ra(ifp->idev) &&
4180 		  ifp->idev->cnf.rtr_solicits != 0 &&
4181 		  (dev->flags&IFF_LOOPBACK) == 0;
4182 	read_unlock_bh(&ifp->idev->lock);
4183 
4184 	/* While dad is in progress mld report's source address is in6_addrany.
4185 	 * Resend with proper ll now.
4186 	 */
4187 	if (send_mld)
4188 		ipv6_mc_dad_complete(ifp->idev);
4189 
4190 	/* send unsolicited NA if enabled */
4191 	if (send_na &&
4192 	    (ifp->idev->cnf.ndisc_notify ||
4193 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4194 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4195 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4196 			      /*solicited=*/ false, /*override=*/ true,
4197 			      /*inc_opt=*/ true);
4198 	}
4199 
4200 	if (send_rs) {
4201 		/*
4202 		 *	If a host as already performed a random delay
4203 		 *	[...] as part of DAD [...] there is no need
4204 		 *	to delay again before sending the first RS
4205 		 */
4206 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4207 			return;
4208 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4209 
4210 		write_lock_bh(&ifp->idev->lock);
4211 		spin_lock(&ifp->lock);
4212 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4213 			ifp->idev->cnf.rtr_solicit_interval);
4214 		ifp->idev->rs_probes = 1;
4215 		ifp->idev->if_flags |= IF_RS_SENT;
4216 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4217 		spin_unlock(&ifp->lock);
4218 		write_unlock_bh(&ifp->idev->lock);
4219 	}
4220 
4221 	if (bump_id)
4222 		rt_genid_bump_ipv6(dev_net(dev));
4223 
4224 	/* Make sure that a new temporary address will be created
4225 	 * before this temporary address becomes deprecated.
4226 	 */
4227 	if (ifp->flags & IFA_F_TEMPORARY)
4228 		addrconf_verify_rtnl();
4229 }
4230 
addrconf_dad_run(struct inet6_dev * idev,bool restart)4231 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4232 {
4233 	struct inet6_ifaddr *ifp;
4234 
4235 	read_lock_bh(&idev->lock);
4236 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4237 		spin_lock(&ifp->lock);
4238 		if ((ifp->flags & IFA_F_TENTATIVE &&
4239 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4240 			if (restart)
4241 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4242 			addrconf_dad_kick(ifp);
4243 		}
4244 		spin_unlock(&ifp->lock);
4245 	}
4246 	read_unlock_bh(&idev->lock);
4247 }
4248 
4249 #ifdef CONFIG_PROC_FS
4250 struct if6_iter_state {
4251 	struct seq_net_private p;
4252 	int bucket;
4253 	int offset;
4254 };
4255 
if6_get_first(struct seq_file * seq,loff_t pos)4256 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4257 {
4258 	struct if6_iter_state *state = seq->private;
4259 	struct net *net = seq_file_net(seq);
4260 	struct inet6_ifaddr *ifa = NULL;
4261 	int p = 0;
4262 
4263 	/* initial bucket if pos is 0 */
4264 	if (pos == 0) {
4265 		state->bucket = 0;
4266 		state->offset = 0;
4267 	}
4268 
4269 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4270 		hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4271 					 addr_lst) {
4272 			if (!net_eq(dev_net(ifa->idev->dev), net))
4273 				continue;
4274 			/* sync with offset */
4275 			if (p < state->offset) {
4276 				p++;
4277 				continue;
4278 			}
4279 			return ifa;
4280 		}
4281 
4282 		/* prepare for next bucket */
4283 		state->offset = 0;
4284 		p = 0;
4285 	}
4286 	return NULL;
4287 }
4288 
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4289 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4290 					 struct inet6_ifaddr *ifa)
4291 {
4292 	struct if6_iter_state *state = seq->private;
4293 	struct net *net = seq_file_net(seq);
4294 
4295 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4296 		if (!net_eq(dev_net(ifa->idev->dev), net))
4297 			continue;
4298 		state->offset++;
4299 		return ifa;
4300 	}
4301 
4302 	state->offset = 0;
4303 	while (++state->bucket < IN6_ADDR_HSIZE) {
4304 		hlist_for_each_entry_rcu(ifa,
4305 				     &inet6_addr_lst[state->bucket], addr_lst) {
4306 			if (!net_eq(dev_net(ifa->idev->dev), net))
4307 				continue;
4308 			return ifa;
4309 		}
4310 	}
4311 
4312 	return NULL;
4313 }
4314 
if6_seq_start(struct seq_file * seq,loff_t * pos)4315 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4316 	__acquires(rcu)
4317 {
4318 	rcu_read_lock();
4319 	return if6_get_first(seq, *pos);
4320 }
4321 
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4322 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4323 {
4324 	struct inet6_ifaddr *ifa;
4325 
4326 	ifa = if6_get_next(seq, v);
4327 	++*pos;
4328 	return ifa;
4329 }
4330 
if6_seq_stop(struct seq_file * seq,void * v)4331 static void if6_seq_stop(struct seq_file *seq, void *v)
4332 	__releases(rcu)
4333 {
4334 	rcu_read_unlock();
4335 }
4336 
if6_seq_show(struct seq_file * seq,void * v)4337 static int if6_seq_show(struct seq_file *seq, void *v)
4338 {
4339 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4340 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4341 		   &ifp->addr,
4342 		   ifp->idev->dev->ifindex,
4343 		   ifp->prefix_len,
4344 		   ifp->scope,
4345 		   (u8) ifp->flags,
4346 		   ifp->idev->dev->name);
4347 	return 0;
4348 }
4349 
4350 static const struct seq_operations if6_seq_ops = {
4351 	.start	= if6_seq_start,
4352 	.next	= if6_seq_next,
4353 	.show	= if6_seq_show,
4354 	.stop	= if6_seq_stop,
4355 };
4356 
if6_proc_net_init(struct net * net)4357 static int __net_init if6_proc_net_init(struct net *net)
4358 {
4359 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4360 			sizeof(struct if6_iter_state)))
4361 		return -ENOMEM;
4362 	return 0;
4363 }
4364 
if6_proc_net_exit(struct net * net)4365 static void __net_exit if6_proc_net_exit(struct net *net)
4366 {
4367 	remove_proc_entry("if_inet6", net->proc_net);
4368 }
4369 
4370 static struct pernet_operations if6_proc_net_ops = {
4371 	.init = if6_proc_net_init,
4372 	.exit = if6_proc_net_exit,
4373 };
4374 
if6_proc_init(void)4375 int __init if6_proc_init(void)
4376 {
4377 	return register_pernet_subsys(&if6_proc_net_ops);
4378 }
4379 
if6_proc_exit(void)4380 void if6_proc_exit(void)
4381 {
4382 	unregister_pernet_subsys(&if6_proc_net_ops);
4383 }
4384 #endif	/* CONFIG_PROC_FS */
4385 
4386 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4387 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4388 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4389 {
4390 	unsigned int hash = inet6_addr_hash(net, addr);
4391 	struct inet6_ifaddr *ifp = NULL;
4392 	int ret = 0;
4393 
4394 	rcu_read_lock();
4395 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4396 		if (!net_eq(dev_net(ifp->idev->dev), net))
4397 			continue;
4398 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4399 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4400 			ret = 1;
4401 			break;
4402 		}
4403 	}
4404 	rcu_read_unlock();
4405 	return ret;
4406 }
4407 #endif
4408 
4409 /* RFC6554 has some algorithm to avoid loops in segment routing by
4410  * checking if the segments contains any of a local interface address.
4411  *
4412  * Quote:
4413  *
4414  * To detect loops in the SRH, a router MUST determine if the SRH
4415  * includes multiple addresses assigned to any interface on that router.
4416  * If such addresses appear more than once and are separated by at least
4417  * one address not assigned to that router.
4418  */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4419 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4420 			  unsigned char nsegs)
4421 {
4422 	const struct in6_addr *addr;
4423 	int i, ret = 0, found = 0;
4424 	struct inet6_ifaddr *ifp;
4425 	bool separated = false;
4426 	unsigned int hash;
4427 	bool hash_found;
4428 
4429 	rcu_read_lock();
4430 	for (i = 0; i < nsegs; i++) {
4431 		addr = &segs[i];
4432 		hash = inet6_addr_hash(net, addr);
4433 
4434 		hash_found = false;
4435 		hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4436 			if (!net_eq(dev_net(ifp->idev->dev), net))
4437 				continue;
4438 
4439 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4440 				hash_found = true;
4441 				break;
4442 			}
4443 		}
4444 
4445 		if (hash_found) {
4446 			if (found > 1 && separated) {
4447 				ret = 1;
4448 				break;
4449 			}
4450 
4451 			separated = false;
4452 			found++;
4453 		} else {
4454 			separated = true;
4455 		}
4456 	}
4457 	rcu_read_unlock();
4458 
4459 	return ret;
4460 }
4461 
4462 /*
4463  *	Periodic address status verification
4464  */
4465 
addrconf_verify_rtnl(void)4466 static void addrconf_verify_rtnl(void)
4467 {
4468 	unsigned long now, next, next_sec, next_sched;
4469 	struct inet6_ifaddr *ifp;
4470 	int i;
4471 
4472 	ASSERT_RTNL();
4473 
4474 	rcu_read_lock_bh();
4475 	now = jiffies;
4476 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4477 
4478 	cancel_delayed_work(&addr_chk_work);
4479 
4480 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4481 restart:
4482 		hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4483 			unsigned long age;
4484 
4485 			/* When setting preferred_lft to a value not zero or
4486 			 * infinity, while valid_lft is infinity
4487 			 * IFA_F_PERMANENT has a non-infinity life time.
4488 			 */
4489 			if ((ifp->flags & IFA_F_PERMANENT) &&
4490 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4491 				continue;
4492 
4493 			spin_lock(&ifp->lock);
4494 			/* We try to batch several events at once. */
4495 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4496 
4497 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4498 			    age >= ifp->valid_lft) {
4499 				spin_unlock(&ifp->lock);
4500 				in6_ifa_hold(ifp);
4501 				ipv6_del_addr(ifp);
4502 				goto restart;
4503 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4504 				spin_unlock(&ifp->lock);
4505 				continue;
4506 			} else if (age >= ifp->prefered_lft) {
4507 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4508 				int deprecate = 0;
4509 
4510 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4511 					deprecate = 1;
4512 					ifp->flags |= IFA_F_DEPRECATED;
4513 				}
4514 
4515 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4516 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4517 					next = ifp->tstamp + ifp->valid_lft * HZ;
4518 
4519 				spin_unlock(&ifp->lock);
4520 
4521 				if (deprecate) {
4522 					in6_ifa_hold(ifp);
4523 
4524 					ipv6_ifa_notify(0, ifp);
4525 					in6_ifa_put(ifp);
4526 					goto restart;
4527 				}
4528 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4529 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4530 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4531 					ifp->idev->cnf.dad_transmits *
4532 					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4533 
4534 				if (age >= ifp->prefered_lft - regen_advance) {
4535 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4536 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4537 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4538 					if (!ifp->regen_count && ifpub) {
4539 						ifp->regen_count++;
4540 						in6_ifa_hold(ifp);
4541 						in6_ifa_hold(ifpub);
4542 						spin_unlock(&ifp->lock);
4543 
4544 						spin_lock(&ifpub->lock);
4545 						ifpub->regen_count = 0;
4546 						spin_unlock(&ifpub->lock);
4547 						rcu_read_unlock_bh();
4548 						ipv6_create_tempaddr(ifpub, true);
4549 						in6_ifa_put(ifpub);
4550 						in6_ifa_put(ifp);
4551 						rcu_read_lock_bh();
4552 						goto restart;
4553 					}
4554 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4555 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4556 				spin_unlock(&ifp->lock);
4557 			} else {
4558 				/* ifp->prefered_lft <= ifp->valid_lft */
4559 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4560 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4561 				spin_unlock(&ifp->lock);
4562 			}
4563 		}
4564 	}
4565 
4566 	next_sec = round_jiffies_up(next);
4567 	next_sched = next;
4568 
4569 	/* If rounded timeout is accurate enough, accept it. */
4570 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4571 		next_sched = next_sec;
4572 
4573 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4574 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4575 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4576 
4577 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4578 		 now, next, next_sec, next_sched);
4579 	mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4580 	rcu_read_unlock_bh();
4581 }
4582 
addrconf_verify_work(struct work_struct * w)4583 static void addrconf_verify_work(struct work_struct *w)
4584 {
4585 	rtnl_lock();
4586 	addrconf_verify_rtnl();
4587 	rtnl_unlock();
4588 }
4589 
addrconf_verify(void)4590 static void addrconf_verify(void)
4591 {
4592 	mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4593 }
4594 
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4595 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4596 				     struct in6_addr **peer_pfx)
4597 {
4598 	struct in6_addr *pfx = NULL;
4599 
4600 	*peer_pfx = NULL;
4601 
4602 	if (addr)
4603 		pfx = nla_data(addr);
4604 
4605 	if (local) {
4606 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4607 			*peer_pfx = pfx;
4608 		pfx = nla_data(local);
4609 	}
4610 
4611 	return pfx;
4612 }
4613 
4614 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4615 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4616 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4617 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4618 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4619 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4620 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4621 };
4622 
4623 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4624 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4625 		  struct netlink_ext_ack *extack)
4626 {
4627 	struct net *net = sock_net(skb->sk);
4628 	struct ifaddrmsg *ifm;
4629 	struct nlattr *tb[IFA_MAX+1];
4630 	struct in6_addr *pfx, *peer_pfx;
4631 	u32 ifa_flags;
4632 	int err;
4633 
4634 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4635 				     ifa_ipv6_policy, extack);
4636 	if (err < 0)
4637 		return err;
4638 
4639 	ifm = nlmsg_data(nlh);
4640 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4641 	if (!pfx)
4642 		return -EINVAL;
4643 
4644 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4645 
4646 	/* We ignore other flags so far. */
4647 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4648 
4649 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4650 			      ifm->ifa_prefixlen);
4651 }
4652 
modify_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4653 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4654 			       unsigned long expires, u32 flags,
4655 			       bool modify_peer)
4656 {
4657 	struct fib6_info *f6i;
4658 	u32 prio;
4659 
4660 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4661 					ifp->prefix_len,
4662 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4663 	if (!f6i)
4664 		return -ENOENT;
4665 
4666 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4667 	if (f6i->fib6_metric != prio) {
4668 		/* delete old one */
4669 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4670 
4671 		/* add new one */
4672 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4673 				      ifp->prefix_len,
4674 				      ifp->rt_priority, ifp->idev->dev,
4675 				      expires, flags, GFP_KERNEL);
4676 	} else {
4677 		if (!expires)
4678 			fib6_clean_expires(f6i);
4679 		else
4680 			fib6_set_expires(f6i, expires);
4681 
4682 		fib6_info_release(f6i);
4683 	}
4684 
4685 	return 0;
4686 }
4687 
inet6_addr_modify(struct inet6_ifaddr * ifp,struct ifa6_config * cfg)4688 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4689 {
4690 	u32 flags;
4691 	clock_t expires;
4692 	unsigned long timeout;
4693 	bool was_managetempaddr;
4694 	bool had_prefixroute;
4695 	bool new_peer = false;
4696 
4697 	ASSERT_RTNL();
4698 
4699 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4700 		return -EINVAL;
4701 
4702 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4703 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4704 		return -EINVAL;
4705 
4706 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4707 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4708 
4709 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4710 	if (addrconf_finite_timeout(timeout)) {
4711 		expires = jiffies_to_clock_t(timeout * HZ);
4712 		cfg->valid_lft = timeout;
4713 		flags = RTF_EXPIRES;
4714 	} else {
4715 		expires = 0;
4716 		flags = 0;
4717 		cfg->ifa_flags |= IFA_F_PERMANENT;
4718 	}
4719 
4720 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4721 	if (addrconf_finite_timeout(timeout)) {
4722 		if (timeout == 0)
4723 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4724 		cfg->preferred_lft = timeout;
4725 	}
4726 
4727 	if (cfg->peer_pfx &&
4728 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4729 		if (!ipv6_addr_any(&ifp->peer_addr))
4730 			cleanup_prefix_route(ifp, expires, true, true);
4731 		new_peer = true;
4732 	}
4733 
4734 	spin_lock_bh(&ifp->lock);
4735 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4736 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4737 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4738 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4739 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4740 			IFA_F_NOPREFIXROUTE);
4741 	ifp->flags |= cfg->ifa_flags;
4742 	ifp->tstamp = jiffies;
4743 	ifp->valid_lft = cfg->valid_lft;
4744 	ifp->prefered_lft = cfg->preferred_lft;
4745 
4746 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4747 		ifp->rt_priority = cfg->rt_priority;
4748 
4749 	if (new_peer)
4750 		ifp->peer_addr = *cfg->peer_pfx;
4751 
4752 	spin_unlock_bh(&ifp->lock);
4753 	if (!(ifp->flags&IFA_F_TENTATIVE))
4754 		ipv6_ifa_notify(0, ifp);
4755 
4756 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4757 		int rc = -ENOENT;
4758 
4759 		if (had_prefixroute)
4760 			rc = modify_prefix_route(ifp, expires, flags, false);
4761 
4762 		/* prefix route could have been deleted; if so restore it */
4763 		if (rc == -ENOENT) {
4764 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4765 					      ifp->rt_priority, ifp->idev->dev,
4766 					      expires, flags, GFP_KERNEL);
4767 		}
4768 
4769 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4770 			rc = modify_prefix_route(ifp, expires, flags, true);
4771 
4772 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4773 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4774 					      ifp->rt_priority, ifp->idev->dev,
4775 					      expires, flags, GFP_KERNEL);
4776 		}
4777 	} else if (had_prefixroute) {
4778 		enum cleanup_prefix_rt_t action;
4779 		unsigned long rt_expires;
4780 
4781 		write_lock_bh(&ifp->idev->lock);
4782 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4783 		write_unlock_bh(&ifp->idev->lock);
4784 
4785 		if (action != CLEANUP_PREFIX_RT_NOP) {
4786 			cleanup_prefix_route(ifp, rt_expires,
4787 				action == CLEANUP_PREFIX_RT_DEL, false);
4788 		}
4789 	}
4790 
4791 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4792 		if (was_managetempaddr &&
4793 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4794 			cfg->valid_lft = 0;
4795 			cfg->preferred_lft = 0;
4796 		}
4797 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4798 				 cfg->preferred_lft, !was_managetempaddr,
4799 				 jiffies);
4800 	}
4801 
4802 	addrconf_verify_rtnl();
4803 
4804 	return 0;
4805 }
4806 
4807 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4808 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4809 		  struct netlink_ext_ack *extack)
4810 {
4811 	struct net *net = sock_net(skb->sk);
4812 	struct ifaddrmsg *ifm;
4813 	struct nlattr *tb[IFA_MAX+1];
4814 	struct in6_addr *peer_pfx;
4815 	struct inet6_ifaddr *ifa;
4816 	struct net_device *dev;
4817 	struct inet6_dev *idev;
4818 	struct ifa6_config cfg;
4819 	int err;
4820 
4821 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4822 				     ifa_ipv6_policy, extack);
4823 	if (err < 0)
4824 		return err;
4825 
4826 	memset(&cfg, 0, sizeof(cfg));
4827 
4828 	ifm = nlmsg_data(nlh);
4829 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4830 	if (!cfg.pfx)
4831 		return -EINVAL;
4832 
4833 	cfg.peer_pfx = peer_pfx;
4834 	cfg.plen = ifm->ifa_prefixlen;
4835 	if (tb[IFA_RT_PRIORITY])
4836 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4837 
4838 	cfg.valid_lft = INFINITY_LIFE_TIME;
4839 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4840 
4841 	if (tb[IFA_CACHEINFO]) {
4842 		struct ifa_cacheinfo *ci;
4843 
4844 		ci = nla_data(tb[IFA_CACHEINFO]);
4845 		cfg.valid_lft = ci->ifa_valid;
4846 		cfg.preferred_lft = ci->ifa_prefered;
4847 	}
4848 
4849 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4850 	if (!dev)
4851 		return -ENODEV;
4852 
4853 	if (tb[IFA_FLAGS])
4854 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4855 	else
4856 		cfg.ifa_flags = ifm->ifa_flags;
4857 
4858 	/* We ignore other flags so far. */
4859 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4860 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4861 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4862 
4863 	idev = ipv6_find_idev(dev);
4864 	if (IS_ERR(idev))
4865 		return PTR_ERR(idev);
4866 
4867 	if (!ipv6_allow_optimistic_dad(net, idev))
4868 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4869 
4870 	if (cfg.ifa_flags & IFA_F_NODAD &&
4871 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4872 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4873 		return -EINVAL;
4874 	}
4875 
4876 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4877 	if (!ifa) {
4878 		/*
4879 		 * It would be best to check for !NLM_F_CREATE here but
4880 		 * userspace already relies on not having to provide this.
4881 		 */
4882 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4883 	}
4884 
4885 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4886 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4887 		err = -EEXIST;
4888 	else
4889 		err = inet6_addr_modify(ifa, &cfg);
4890 
4891 	in6_ifa_put(ifa);
4892 
4893 	return err;
4894 }
4895 
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)4896 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4897 			  u8 scope, int ifindex)
4898 {
4899 	struct ifaddrmsg *ifm;
4900 
4901 	ifm = nlmsg_data(nlh);
4902 	ifm->ifa_family = AF_INET6;
4903 	ifm->ifa_prefixlen = prefixlen;
4904 	ifm->ifa_flags = flags;
4905 	ifm->ifa_scope = scope;
4906 	ifm->ifa_index = ifindex;
4907 }
4908 
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)4909 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4910 			 unsigned long tstamp, u32 preferred, u32 valid)
4911 {
4912 	struct ifa_cacheinfo ci;
4913 
4914 	ci.cstamp = cstamp_delta(cstamp);
4915 	ci.tstamp = cstamp_delta(tstamp);
4916 	ci.ifa_prefered = preferred;
4917 	ci.ifa_valid = valid;
4918 
4919 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4920 }
4921 
rt_scope(int ifa_scope)4922 static inline int rt_scope(int ifa_scope)
4923 {
4924 	if (ifa_scope & IFA_HOST)
4925 		return RT_SCOPE_HOST;
4926 	else if (ifa_scope & IFA_LINK)
4927 		return RT_SCOPE_LINK;
4928 	else if (ifa_scope & IFA_SITE)
4929 		return RT_SCOPE_SITE;
4930 	else
4931 		return RT_SCOPE_UNIVERSE;
4932 }
4933 
inet6_ifaddr_msgsize(void)4934 static inline int inet6_ifaddr_msgsize(void)
4935 {
4936 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4937 	       + nla_total_size(16) /* IFA_LOCAL */
4938 	       + nla_total_size(16) /* IFA_ADDRESS */
4939 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4940 	       + nla_total_size(4)  /* IFA_FLAGS */
4941 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4942 }
4943 
4944 enum addr_type_t {
4945 	UNICAST_ADDR,
4946 	MULTICAST_ADDR,
4947 	ANYCAST_ADDR,
4948 };
4949 
4950 struct inet6_fill_args {
4951 	u32 portid;
4952 	u32 seq;
4953 	int event;
4954 	unsigned int flags;
4955 	int netnsid;
4956 	int ifindex;
4957 	enum addr_type_t type;
4958 };
4959 
inet6_fill_ifaddr(struct sk_buff * skb,struct inet6_ifaddr * ifa,struct inet6_fill_args * args)4960 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4961 			     struct inet6_fill_args *args)
4962 {
4963 	struct nlmsghdr  *nlh;
4964 	u32 preferred, valid;
4965 
4966 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4967 			sizeof(struct ifaddrmsg), args->flags);
4968 	if (!nlh)
4969 		return -EMSGSIZE;
4970 
4971 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4972 		      ifa->idev->dev->ifindex);
4973 
4974 	if (args->netnsid >= 0 &&
4975 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4976 		goto error;
4977 
4978 	if (!((ifa->flags&IFA_F_PERMANENT) &&
4979 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4980 		preferred = ifa->prefered_lft;
4981 		valid = ifa->valid_lft;
4982 		if (preferred != INFINITY_LIFE_TIME) {
4983 			long tval = (jiffies - ifa->tstamp)/HZ;
4984 			if (preferred > tval)
4985 				preferred -= tval;
4986 			else
4987 				preferred = 0;
4988 			if (valid != INFINITY_LIFE_TIME) {
4989 				if (valid > tval)
4990 					valid -= tval;
4991 				else
4992 					valid = 0;
4993 			}
4994 		}
4995 	} else {
4996 		preferred = INFINITY_LIFE_TIME;
4997 		valid = INFINITY_LIFE_TIME;
4998 	}
4999 
5000 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5001 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5002 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5003 			goto error;
5004 	} else
5005 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5006 			goto error;
5007 
5008 	if (ifa->rt_priority &&
5009 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5010 		goto error;
5011 
5012 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5013 		goto error;
5014 
5015 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5016 		goto error;
5017 
5018 	nlmsg_end(skb, nlh);
5019 	return 0;
5020 
5021 error:
5022 	nlmsg_cancel(skb, nlh);
5023 	return -EMSGSIZE;
5024 }
5025 
inet6_fill_ifmcaddr(struct sk_buff * skb,struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5026 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5027 			       struct inet6_fill_args *args)
5028 {
5029 	struct nlmsghdr  *nlh;
5030 	u8 scope = RT_SCOPE_UNIVERSE;
5031 	int ifindex = ifmca->idev->dev->ifindex;
5032 
5033 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5034 		scope = RT_SCOPE_SITE;
5035 
5036 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5037 			sizeof(struct ifaddrmsg), args->flags);
5038 	if (!nlh)
5039 		return -EMSGSIZE;
5040 
5041 	if (args->netnsid >= 0 &&
5042 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5043 		nlmsg_cancel(skb, nlh);
5044 		return -EMSGSIZE;
5045 	}
5046 
5047 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5048 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5049 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5050 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5051 		nlmsg_cancel(skb, nlh);
5052 		return -EMSGSIZE;
5053 	}
5054 
5055 	nlmsg_end(skb, nlh);
5056 	return 0;
5057 }
5058 
inet6_fill_ifacaddr(struct sk_buff * skb,struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5059 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5060 			       struct inet6_fill_args *args)
5061 {
5062 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5063 	int ifindex = dev ? dev->ifindex : 1;
5064 	struct nlmsghdr  *nlh;
5065 	u8 scope = RT_SCOPE_UNIVERSE;
5066 
5067 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5068 		scope = RT_SCOPE_SITE;
5069 
5070 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5071 			sizeof(struct ifaddrmsg), args->flags);
5072 	if (!nlh)
5073 		return -EMSGSIZE;
5074 
5075 	if (args->netnsid >= 0 &&
5076 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5077 		nlmsg_cancel(skb, nlh);
5078 		return -EMSGSIZE;
5079 	}
5080 
5081 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5082 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5083 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5084 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5085 		nlmsg_cancel(skb, nlh);
5086 		return -EMSGSIZE;
5087 	}
5088 
5089 	nlmsg_end(skb, nlh);
5090 	return 0;
5091 }
5092 
5093 /* called with rcu_read_lock() */
in6_dump_addrs(struct inet6_dev * idev,struct sk_buff * skb,struct netlink_callback * cb,int s_ip_idx,struct inet6_fill_args * fillargs)5094 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5095 			  struct netlink_callback *cb, int s_ip_idx,
5096 			  struct inet6_fill_args *fillargs)
5097 {
5098 	struct ifmcaddr6 *ifmca;
5099 	struct ifacaddr6 *ifaca;
5100 	int ip_idx = 0;
5101 	int err = 1;
5102 
5103 	read_lock_bh(&idev->lock);
5104 	switch (fillargs->type) {
5105 	case UNICAST_ADDR: {
5106 		struct inet6_ifaddr *ifa;
5107 		fillargs->event = RTM_NEWADDR;
5108 
5109 		/* unicast address incl. temp addr */
5110 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
5111 			if (ip_idx < s_ip_idx)
5112 				goto next;
5113 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5114 			if (err < 0)
5115 				break;
5116 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5117 next:
5118 			ip_idx++;
5119 		}
5120 		break;
5121 	}
5122 	case MULTICAST_ADDR:
5123 		fillargs->event = RTM_GETMULTICAST;
5124 
5125 		/* multicast address */
5126 		for (ifmca = idev->mc_list; ifmca;
5127 		     ifmca = ifmca->next, ip_idx++) {
5128 			if (ip_idx < s_ip_idx)
5129 				continue;
5130 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5131 			if (err < 0)
5132 				break;
5133 		}
5134 		break;
5135 	case ANYCAST_ADDR:
5136 		fillargs->event = RTM_GETANYCAST;
5137 		/* anycast address */
5138 		for (ifaca = idev->ac_list; ifaca;
5139 		     ifaca = ifaca->aca_next, ip_idx++) {
5140 			if (ip_idx < s_ip_idx)
5141 				continue;
5142 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5143 			if (err < 0)
5144 				break;
5145 		}
5146 		break;
5147 	default:
5148 		break;
5149 	}
5150 	read_unlock_bh(&idev->lock);
5151 	cb->args[2] = ip_idx;
5152 	return err;
5153 }
5154 
inet6_valid_dump_ifaddr_req(const struct nlmsghdr * nlh,struct inet6_fill_args * fillargs,struct net ** tgt_net,struct sock * sk,struct netlink_callback * cb)5155 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5156 				       struct inet6_fill_args *fillargs,
5157 				       struct net **tgt_net, struct sock *sk,
5158 				       struct netlink_callback *cb)
5159 {
5160 	struct netlink_ext_ack *extack = cb->extack;
5161 	struct nlattr *tb[IFA_MAX+1];
5162 	struct ifaddrmsg *ifm;
5163 	int err, i;
5164 
5165 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5166 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5167 		return -EINVAL;
5168 	}
5169 
5170 	ifm = nlmsg_data(nlh);
5171 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5172 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5173 		return -EINVAL;
5174 	}
5175 
5176 	fillargs->ifindex = ifm->ifa_index;
5177 	if (fillargs->ifindex) {
5178 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5179 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5180 	}
5181 
5182 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5183 					    ifa_ipv6_policy, extack);
5184 	if (err < 0)
5185 		return err;
5186 
5187 	for (i = 0; i <= IFA_MAX; ++i) {
5188 		if (!tb[i])
5189 			continue;
5190 
5191 		if (i == IFA_TARGET_NETNSID) {
5192 			struct net *net;
5193 
5194 			fillargs->netnsid = nla_get_s32(tb[i]);
5195 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5196 			if (IS_ERR(net)) {
5197 				fillargs->netnsid = -1;
5198 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5199 				return PTR_ERR(net);
5200 			}
5201 			*tgt_net = net;
5202 		} else {
5203 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5204 			return -EINVAL;
5205 		}
5206 	}
5207 
5208 	return 0;
5209 }
5210 
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5211 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5212 			   enum addr_type_t type)
5213 {
5214 	const struct nlmsghdr *nlh = cb->nlh;
5215 	struct inet6_fill_args fillargs = {
5216 		.portid = NETLINK_CB(cb->skb).portid,
5217 		.seq = cb->nlh->nlmsg_seq,
5218 		.flags = NLM_F_MULTI,
5219 		.netnsid = -1,
5220 		.type = type,
5221 	};
5222 	struct net *net = sock_net(skb->sk);
5223 	struct net *tgt_net = net;
5224 	int idx, s_idx, s_ip_idx;
5225 	int h, s_h;
5226 	struct net_device *dev;
5227 	struct inet6_dev *idev;
5228 	struct hlist_head *head;
5229 	int err = 0;
5230 
5231 	s_h = cb->args[0];
5232 	s_idx = idx = cb->args[1];
5233 	s_ip_idx = cb->args[2];
5234 
5235 	if (cb->strict_check) {
5236 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5237 						  skb->sk, cb);
5238 		if (err < 0)
5239 			goto put_tgt_net;
5240 
5241 		err = 0;
5242 		if (fillargs.ifindex) {
5243 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5244 			if (!dev) {
5245 				err = -ENODEV;
5246 				goto put_tgt_net;
5247 			}
5248 			idev = __in6_dev_get(dev);
5249 			if (idev) {
5250 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5251 						     &fillargs);
5252 				if (err > 0)
5253 					err = 0;
5254 			}
5255 			goto put_tgt_net;
5256 		}
5257 	}
5258 
5259 	rcu_read_lock();
5260 	cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5261 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5262 		idx = 0;
5263 		head = &tgt_net->dev_index_head[h];
5264 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5265 			if (idx < s_idx)
5266 				goto cont;
5267 			if (h > s_h || idx > s_idx)
5268 				s_ip_idx = 0;
5269 			idev = __in6_dev_get(dev);
5270 			if (!idev)
5271 				goto cont;
5272 
5273 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5274 					   &fillargs) < 0)
5275 				goto done;
5276 cont:
5277 			idx++;
5278 		}
5279 	}
5280 done:
5281 	rcu_read_unlock();
5282 	cb->args[0] = h;
5283 	cb->args[1] = idx;
5284 put_tgt_net:
5285 	if (fillargs.netnsid >= 0)
5286 		put_net(tgt_net);
5287 
5288 	return skb->len ? : err;
5289 }
5290 
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5291 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5292 {
5293 	enum addr_type_t type = UNICAST_ADDR;
5294 
5295 	return inet6_dump_addr(skb, cb, type);
5296 }
5297 
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5298 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5299 {
5300 	enum addr_type_t type = MULTICAST_ADDR;
5301 
5302 	return inet6_dump_addr(skb, cb, type);
5303 }
5304 
5305 
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5306 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5307 {
5308 	enum addr_type_t type = ANYCAST_ADDR;
5309 
5310 	return inet6_dump_addr(skb, cb, type);
5311 }
5312 
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5313 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5314 				       const struct nlmsghdr *nlh,
5315 				       struct nlattr **tb,
5316 				       struct netlink_ext_ack *extack)
5317 {
5318 	struct ifaddrmsg *ifm;
5319 	int i, err;
5320 
5321 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5322 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5323 		return -EINVAL;
5324 	}
5325 
5326 	if (!netlink_strict_get_check(skb))
5327 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5328 					      ifa_ipv6_policy, extack);
5329 
5330 	ifm = nlmsg_data(nlh);
5331 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5332 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5333 		return -EINVAL;
5334 	}
5335 
5336 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5337 					    ifa_ipv6_policy, extack);
5338 	if (err)
5339 		return err;
5340 
5341 	for (i = 0; i <= IFA_MAX; i++) {
5342 		if (!tb[i])
5343 			continue;
5344 
5345 		switch (i) {
5346 		case IFA_TARGET_NETNSID:
5347 		case IFA_ADDRESS:
5348 		case IFA_LOCAL:
5349 			break;
5350 		default:
5351 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5352 			return -EINVAL;
5353 		}
5354 	}
5355 
5356 	return 0;
5357 }
5358 
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5359 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5360 			     struct netlink_ext_ack *extack)
5361 {
5362 	struct net *net = sock_net(in_skb->sk);
5363 	struct inet6_fill_args fillargs = {
5364 		.portid = NETLINK_CB(in_skb).portid,
5365 		.seq = nlh->nlmsg_seq,
5366 		.event = RTM_NEWADDR,
5367 		.flags = 0,
5368 		.netnsid = -1,
5369 	};
5370 	struct net *tgt_net = net;
5371 	struct ifaddrmsg *ifm;
5372 	struct nlattr *tb[IFA_MAX+1];
5373 	struct in6_addr *addr = NULL, *peer;
5374 	struct net_device *dev = NULL;
5375 	struct inet6_ifaddr *ifa;
5376 	struct sk_buff *skb;
5377 	int err;
5378 
5379 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5380 	if (err < 0)
5381 		return err;
5382 
5383 	if (tb[IFA_TARGET_NETNSID]) {
5384 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5385 
5386 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5387 						  fillargs.netnsid);
5388 		if (IS_ERR(tgt_net))
5389 			return PTR_ERR(tgt_net);
5390 	}
5391 
5392 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5393 	if (!addr)
5394 		return -EINVAL;
5395 
5396 	ifm = nlmsg_data(nlh);
5397 	if (ifm->ifa_index)
5398 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5399 
5400 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5401 	if (!ifa) {
5402 		err = -EADDRNOTAVAIL;
5403 		goto errout;
5404 	}
5405 
5406 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5407 	if (!skb) {
5408 		err = -ENOBUFS;
5409 		goto errout_ifa;
5410 	}
5411 
5412 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5413 	if (err < 0) {
5414 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5415 		WARN_ON(err == -EMSGSIZE);
5416 		kfree_skb(skb);
5417 		goto errout_ifa;
5418 	}
5419 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5420 errout_ifa:
5421 	in6_ifa_put(ifa);
5422 errout:
5423 	if (dev)
5424 		dev_put(dev);
5425 	if (fillargs.netnsid >= 0)
5426 		put_net(tgt_net);
5427 
5428 	return err;
5429 }
5430 
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5431 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5432 {
5433 	struct sk_buff *skb;
5434 	struct net *net = dev_net(ifa->idev->dev);
5435 	struct inet6_fill_args fillargs = {
5436 		.portid = 0,
5437 		.seq = 0,
5438 		.event = event,
5439 		.flags = 0,
5440 		.netnsid = -1,
5441 	};
5442 	int err = -ENOBUFS;
5443 
5444 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5445 	if (!skb)
5446 		goto errout;
5447 
5448 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5449 	if (err < 0) {
5450 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5451 		WARN_ON(err == -EMSGSIZE);
5452 		kfree_skb(skb);
5453 		goto errout;
5454 	}
5455 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5456 	return;
5457 errout:
5458 	if (err < 0)
5459 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5460 }
5461 
ipv6_store_devconf(struct ipv6_devconf * cnf,__s32 * array,int bytes)5462 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5463 				__s32 *array, int bytes)
5464 {
5465 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5466 
5467 	memset(array, 0, bytes);
5468 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5469 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5470 	array[DEVCONF_MTU6] = cnf->mtu6;
5471 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5472 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5473 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5474 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5475 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5476 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5477 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5478 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5479 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5480 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5481 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5482 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5483 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5484 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5485 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5486 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5487 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5488 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5489 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5490 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5491 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5492 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5493 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5494 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5495 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5496 #ifdef CONFIG_IPV6_ROUTER_PREF
5497 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5498 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5499 		jiffies_to_msecs(cnf->rtr_probe_interval);
5500 #ifdef CONFIG_IPV6_ROUTE_INFO
5501 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5502 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5503 #endif
5504 #endif
5505 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5506 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5507 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5508 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5509 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5510 #endif
5511 #ifdef CONFIG_IPV6_MROUTE
5512 	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5513 #endif
5514 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5515 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5516 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5517 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5518 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5519 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5520 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5521 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5522 	/* we omit DEVCONF_STABLE_SECRET for now */
5523 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5524 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5525 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5526 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5527 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5528 #ifdef CONFIG_IPV6_SEG6_HMAC
5529 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5530 #endif
5531 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5532 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5533 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5534 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5535 	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5536 }
5537 
inet6_ifla6_size(void)5538 static inline size_t inet6_ifla6_size(void)
5539 {
5540 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5541 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5542 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5543 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5544 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5545 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5546 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5547 	     + 0;
5548 }
5549 
inet6_if_nlmsg_size(void)5550 static inline size_t inet6_if_nlmsg_size(void)
5551 {
5552 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5553 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5554 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5555 	       + nla_total_size(4) /* IFLA_MTU */
5556 	       + nla_total_size(4) /* IFLA_LINK */
5557 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5558 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5559 }
5560 
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5561 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5562 					int bytes)
5563 {
5564 	int i;
5565 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5566 	BUG_ON(pad < 0);
5567 
5568 	/* Use put_unaligned() because stats may not be aligned for u64. */
5569 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5570 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5571 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5572 
5573 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5574 }
5575 
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5576 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5577 					int bytes, size_t syncpoff)
5578 {
5579 	int i, c;
5580 	u64 buff[IPSTATS_MIB_MAX];
5581 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5582 
5583 	BUG_ON(pad < 0);
5584 
5585 	memset(buff, 0, sizeof(buff));
5586 	buff[0] = IPSTATS_MIB_MAX;
5587 
5588 	for_each_possible_cpu(c) {
5589 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5590 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5591 	}
5592 
5593 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5594 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5595 }
5596 
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5597 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5598 			     int bytes)
5599 {
5600 	switch (attrtype) {
5601 	case IFLA_INET6_STATS:
5602 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5603 				     offsetof(struct ipstats_mib, syncp));
5604 		break;
5605 	case IFLA_INET6_ICMP6STATS:
5606 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5607 		break;
5608 	}
5609 }
5610 
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5611 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5612 				  u32 ext_filter_mask)
5613 {
5614 	struct nlattr *nla;
5615 	struct ifla_cacheinfo ci;
5616 
5617 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5618 		goto nla_put_failure;
5619 	ci.max_reasm_len = IPV6_MAXPLEN;
5620 	ci.tstamp = cstamp_delta(idev->tstamp);
5621 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5622 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5623 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5624 		goto nla_put_failure;
5625 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5626 	if (!nla)
5627 		goto nla_put_failure;
5628 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5629 
5630 	/* XXX - MC not implemented */
5631 
5632 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5633 		return 0;
5634 
5635 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5636 	if (!nla)
5637 		goto nla_put_failure;
5638 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5639 
5640 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5641 	if (!nla)
5642 		goto nla_put_failure;
5643 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5644 
5645 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5646 	if (!nla)
5647 		goto nla_put_failure;
5648 	read_lock_bh(&idev->lock);
5649 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5650 	read_unlock_bh(&idev->lock);
5651 
5652 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5653 		goto nla_put_failure;
5654 
5655 	return 0;
5656 
5657 nla_put_failure:
5658 	return -EMSGSIZE;
5659 }
5660 
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5661 static size_t inet6_get_link_af_size(const struct net_device *dev,
5662 				     u32 ext_filter_mask)
5663 {
5664 	if (!__in6_dev_get(dev))
5665 		return 0;
5666 
5667 	return inet6_ifla6_size();
5668 }
5669 
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5670 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5671 			      u32 ext_filter_mask)
5672 {
5673 	struct inet6_dev *idev = __in6_dev_get(dev);
5674 
5675 	if (!idev)
5676 		return -ENODATA;
5677 
5678 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5679 		return -EMSGSIZE;
5680 
5681 	return 0;
5682 }
5683 
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token)5684 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5685 {
5686 	struct inet6_ifaddr *ifp;
5687 	struct net_device *dev = idev->dev;
5688 	bool clear_token, update_rs = false;
5689 	struct in6_addr ll_addr;
5690 
5691 	ASSERT_RTNL();
5692 
5693 	if (!token)
5694 		return -EINVAL;
5695 	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5696 		return -EINVAL;
5697 	if (!ipv6_accept_ra(idev))
5698 		return -EINVAL;
5699 	if (idev->cnf.rtr_solicits == 0)
5700 		return -EINVAL;
5701 
5702 	write_lock_bh(&idev->lock);
5703 
5704 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5705 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5706 
5707 	write_unlock_bh(&idev->lock);
5708 
5709 	clear_token = ipv6_addr_any(token);
5710 	if (clear_token)
5711 		goto update_lft;
5712 
5713 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5714 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5715 			     IFA_F_OPTIMISTIC)) {
5716 		/* If we're not ready, then normal ifup will take care
5717 		 * of this. Otherwise, we need to request our rs here.
5718 		 */
5719 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5720 		update_rs = true;
5721 	}
5722 
5723 update_lft:
5724 	write_lock_bh(&idev->lock);
5725 
5726 	if (update_rs) {
5727 		idev->if_flags |= IF_RS_SENT;
5728 		idev->rs_interval = rfc3315_s14_backoff_init(
5729 			idev->cnf.rtr_solicit_interval);
5730 		idev->rs_probes = 1;
5731 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5732 	}
5733 
5734 	/* Well, that's kinda nasty ... */
5735 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5736 		spin_lock(&ifp->lock);
5737 		if (ifp->tokenized) {
5738 			ifp->valid_lft = 0;
5739 			ifp->prefered_lft = 0;
5740 		}
5741 		spin_unlock(&ifp->lock);
5742 	}
5743 
5744 	write_unlock_bh(&idev->lock);
5745 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5746 	addrconf_verify_rtnl();
5747 	return 0;
5748 }
5749 
5750 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5751 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5752 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5753 };
5754 
check_addr_gen_mode(int mode)5755 static int check_addr_gen_mode(int mode)
5756 {
5757 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5758 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5759 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5760 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5761 		return -EINVAL;
5762 	return 1;
5763 }
5764 
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)5765 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5766 				int mode)
5767 {
5768 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5769 	    !idev->cnf.stable_secret.initialized &&
5770 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5771 		return -EINVAL;
5772 	return 1;
5773 }
5774 
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla)5775 static int inet6_validate_link_af(const struct net_device *dev,
5776 				  const struct nlattr *nla)
5777 {
5778 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5779 	struct inet6_dev *idev = NULL;
5780 	int err;
5781 
5782 	if (dev) {
5783 		idev = __in6_dev_get(dev);
5784 		if (!idev)
5785 			return -EAFNOSUPPORT;
5786 	}
5787 
5788 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5789 					  inet6_af_policy, NULL);
5790 	if (err)
5791 		return err;
5792 
5793 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5794 		return -EINVAL;
5795 
5796 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5797 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5798 
5799 		if (check_addr_gen_mode(mode) < 0)
5800 			return -EINVAL;
5801 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5802 			return -EINVAL;
5803 	}
5804 
5805 	return 0;
5806 }
5807 
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla)5808 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5809 {
5810 	struct inet6_dev *idev = __in6_dev_get(dev);
5811 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5812 	int err;
5813 
5814 	if (!idev)
5815 		return -EAFNOSUPPORT;
5816 
5817 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5818 		return -EINVAL;
5819 
5820 	if (tb[IFLA_INET6_TOKEN]) {
5821 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5822 		if (err)
5823 			return err;
5824 	}
5825 
5826 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5827 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5828 
5829 		idev->cnf.addr_gen_mode = mode;
5830 	}
5831 
5832 	return 0;
5833 }
5834 
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)5835 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5836 			     u32 portid, u32 seq, int event, unsigned int flags)
5837 {
5838 	struct net_device *dev = idev->dev;
5839 	struct ifinfomsg *hdr;
5840 	struct nlmsghdr *nlh;
5841 	void *protoinfo;
5842 
5843 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5844 	if (!nlh)
5845 		return -EMSGSIZE;
5846 
5847 	hdr = nlmsg_data(nlh);
5848 	hdr->ifi_family = AF_INET6;
5849 	hdr->__ifi_pad = 0;
5850 	hdr->ifi_type = dev->type;
5851 	hdr->ifi_index = dev->ifindex;
5852 	hdr->ifi_flags = dev_get_flags(dev);
5853 	hdr->ifi_change = 0;
5854 
5855 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5856 	    (dev->addr_len &&
5857 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5858 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5859 	    (dev->ifindex != dev_get_iflink(dev) &&
5860 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5861 	    nla_put_u8(skb, IFLA_OPERSTATE,
5862 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5863 		goto nla_put_failure;
5864 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5865 	if (!protoinfo)
5866 		goto nla_put_failure;
5867 
5868 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5869 		goto nla_put_failure;
5870 
5871 	nla_nest_end(skb, protoinfo);
5872 	nlmsg_end(skb, nlh);
5873 	return 0;
5874 
5875 nla_put_failure:
5876 	nlmsg_cancel(skb, nlh);
5877 	return -EMSGSIZE;
5878 }
5879 
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5880 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5881 				   struct netlink_ext_ack *extack)
5882 {
5883 	struct ifinfomsg *ifm;
5884 
5885 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5886 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5887 		return -EINVAL;
5888 	}
5889 
5890 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5891 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5892 		return -EINVAL;
5893 	}
5894 
5895 	ifm = nlmsg_data(nlh);
5896 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5897 	    ifm->ifi_change || ifm->ifi_index) {
5898 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5899 		return -EINVAL;
5900 	}
5901 
5902 	return 0;
5903 }
5904 
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)5905 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5906 {
5907 	struct net *net = sock_net(skb->sk);
5908 	int h, s_h;
5909 	int idx = 0, s_idx;
5910 	struct net_device *dev;
5911 	struct inet6_dev *idev;
5912 	struct hlist_head *head;
5913 
5914 	/* only requests using strict checking can pass data to
5915 	 * influence the dump
5916 	 */
5917 	if (cb->strict_check) {
5918 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5919 
5920 		if (err < 0)
5921 			return err;
5922 	}
5923 
5924 	s_h = cb->args[0];
5925 	s_idx = cb->args[1];
5926 
5927 	rcu_read_lock();
5928 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5929 		idx = 0;
5930 		head = &net->dev_index_head[h];
5931 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5932 			if (idx < s_idx)
5933 				goto cont;
5934 			idev = __in6_dev_get(dev);
5935 			if (!idev)
5936 				goto cont;
5937 			if (inet6_fill_ifinfo(skb, idev,
5938 					      NETLINK_CB(cb->skb).portid,
5939 					      cb->nlh->nlmsg_seq,
5940 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
5941 				goto out;
5942 cont:
5943 			idx++;
5944 		}
5945 	}
5946 out:
5947 	rcu_read_unlock();
5948 	cb->args[1] = idx;
5949 	cb->args[0] = h;
5950 
5951 	return skb->len;
5952 }
5953 
inet6_ifinfo_notify(int event,struct inet6_dev * idev)5954 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5955 {
5956 	struct sk_buff *skb;
5957 	struct net *net = dev_net(idev->dev);
5958 	int err = -ENOBUFS;
5959 
5960 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5961 	if (!skb)
5962 		goto errout;
5963 
5964 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5965 	if (err < 0) {
5966 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5967 		WARN_ON(err == -EMSGSIZE);
5968 		kfree_skb(skb);
5969 		goto errout;
5970 	}
5971 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5972 	return;
5973 errout:
5974 	if (err < 0)
5975 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5976 }
5977 
inet6_prefix_nlmsg_size(void)5978 static inline size_t inet6_prefix_nlmsg_size(void)
5979 {
5980 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
5981 	       + nla_total_size(sizeof(struct in6_addr))
5982 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
5983 }
5984 
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)5985 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5986 			     struct prefix_info *pinfo, u32 portid, u32 seq,
5987 			     int event, unsigned int flags)
5988 {
5989 	struct prefixmsg *pmsg;
5990 	struct nlmsghdr *nlh;
5991 	struct prefix_cacheinfo	ci;
5992 
5993 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5994 	if (!nlh)
5995 		return -EMSGSIZE;
5996 
5997 	pmsg = nlmsg_data(nlh);
5998 	pmsg->prefix_family = AF_INET6;
5999 	pmsg->prefix_pad1 = 0;
6000 	pmsg->prefix_pad2 = 0;
6001 	pmsg->prefix_ifindex = idev->dev->ifindex;
6002 	pmsg->prefix_len = pinfo->prefix_len;
6003 	pmsg->prefix_type = pinfo->type;
6004 	pmsg->prefix_pad3 = 0;
6005 	pmsg->prefix_flags = 0;
6006 	if (pinfo->onlink)
6007 		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6008 	if (pinfo->autoconf)
6009 		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6010 
6011 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6012 		goto nla_put_failure;
6013 	ci.preferred_time = ntohl(pinfo->prefered);
6014 	ci.valid_time = ntohl(pinfo->valid);
6015 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6016 		goto nla_put_failure;
6017 	nlmsg_end(skb, nlh);
6018 	return 0;
6019 
6020 nla_put_failure:
6021 	nlmsg_cancel(skb, nlh);
6022 	return -EMSGSIZE;
6023 }
6024 
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6025 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6026 			 struct prefix_info *pinfo)
6027 {
6028 	struct sk_buff *skb;
6029 	struct net *net = dev_net(idev->dev);
6030 	int err = -ENOBUFS;
6031 
6032 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6033 	if (!skb)
6034 		goto errout;
6035 
6036 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6037 	if (err < 0) {
6038 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6039 		WARN_ON(err == -EMSGSIZE);
6040 		kfree_skb(skb);
6041 		goto errout;
6042 	}
6043 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6044 	return;
6045 errout:
6046 	if (err < 0)
6047 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6048 }
6049 
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6050 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6051 {
6052 	struct net *net = dev_net(ifp->idev->dev);
6053 
6054 	if (event)
6055 		ASSERT_RTNL();
6056 
6057 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6058 
6059 	switch (event) {
6060 	case RTM_NEWADDR:
6061 		/*
6062 		 * If the address was optimistic we inserted the route at the
6063 		 * start of our DAD process, so we don't need to do it again.
6064 		 * If the device was taken down in the middle of the DAD
6065 		 * cycle there is a race where we could get here without a
6066 		 * host route, so nothing to insert. That will be fixed when
6067 		 * the device is brought up.
6068 		 */
6069 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6070 			ip6_ins_rt(net, ifp->rt);
6071 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6072 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6073 				&ifp->addr, ifp->idev->dev->name);
6074 		}
6075 
6076 		if (ifp->idev->cnf.forwarding)
6077 			addrconf_join_anycast(ifp);
6078 		if (!ipv6_addr_any(&ifp->peer_addr))
6079 			addrconf_prefix_route(&ifp->peer_addr, 128,
6080 					      ifp->rt_priority, ifp->idev->dev,
6081 					      0, 0, GFP_ATOMIC);
6082 		break;
6083 	case RTM_DELADDR:
6084 		if (ifp->idev->cnf.forwarding)
6085 			addrconf_leave_anycast(ifp);
6086 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6087 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6088 			struct fib6_info *rt;
6089 
6090 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6091 						       ifp->idev->dev, 0, 0,
6092 						       false);
6093 			if (rt)
6094 				ip6_del_rt(net, rt, false);
6095 		}
6096 		if (ifp->rt) {
6097 			ip6_del_rt(net, ifp->rt, false);
6098 			ifp->rt = NULL;
6099 		}
6100 		rt_genid_bump_ipv6(net);
6101 		break;
6102 	}
6103 	atomic_inc(&net->ipv6.dev_addr_genid);
6104 }
6105 
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6106 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6107 {
6108 	rcu_read_lock_bh();
6109 	if (likely(ifp->idev->dead == 0))
6110 		__ipv6_ifa_notify(event, ifp);
6111 	rcu_read_unlock_bh();
6112 }
6113 
6114 #ifdef CONFIG_SYSCTL
6115 
addrconf_sysctl_forward(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6116 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6117 		void *buffer, size_t *lenp, loff_t *ppos)
6118 {
6119 	int *valp = ctl->data;
6120 	int val = *valp;
6121 	loff_t pos = *ppos;
6122 	struct ctl_table lctl;
6123 	int ret;
6124 
6125 	/*
6126 	 * ctl->data points to idev->cnf.forwarding, we should
6127 	 * not modify it until we get the rtnl lock.
6128 	 */
6129 	lctl = *ctl;
6130 	lctl.data = &val;
6131 
6132 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6133 
6134 	if (write)
6135 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6136 	if (ret)
6137 		*ppos = pos;
6138 	return ret;
6139 }
6140 
addrconf_sysctl_mtu(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6141 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6142 		void *buffer, size_t *lenp, loff_t *ppos)
6143 {
6144 	struct inet6_dev *idev = ctl->extra1;
6145 	int min_mtu = IPV6_MIN_MTU;
6146 	struct ctl_table lctl;
6147 
6148 	lctl = *ctl;
6149 	lctl.extra1 = &min_mtu;
6150 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6151 
6152 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6153 }
6154 
dev_disable_change(struct inet6_dev * idev)6155 static void dev_disable_change(struct inet6_dev *idev)
6156 {
6157 	struct netdev_notifier_info info;
6158 
6159 	if (!idev || !idev->dev)
6160 		return;
6161 
6162 	netdev_notifier_info_init(&info, idev->dev);
6163 	if (idev->cnf.disable_ipv6)
6164 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6165 	else
6166 		addrconf_notify(NULL, NETDEV_UP, &info);
6167 }
6168 
addrconf_disable_change(struct net * net,__s32 newf)6169 static void addrconf_disable_change(struct net *net, __s32 newf)
6170 {
6171 	struct net_device *dev;
6172 	struct inet6_dev *idev;
6173 
6174 	for_each_netdev(net, dev) {
6175 		idev = __in6_dev_get(dev);
6176 		if (idev) {
6177 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6178 			idev->cnf.disable_ipv6 = newf;
6179 			if (changed)
6180 				dev_disable_change(idev);
6181 		}
6182 	}
6183 }
6184 
addrconf_disable_ipv6(struct ctl_table * table,int * p,int newf)6185 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6186 {
6187 	struct net *net;
6188 	int old;
6189 
6190 	if (!rtnl_trylock())
6191 		return restart_syscall();
6192 
6193 	net = (struct net *)table->extra2;
6194 	old = *p;
6195 	*p = newf;
6196 
6197 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6198 		rtnl_unlock();
6199 		return 0;
6200 	}
6201 
6202 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6203 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
6204 		addrconf_disable_change(net, newf);
6205 	} else if ((!newf) ^ (!old))
6206 		dev_disable_change((struct inet6_dev *)table->extra1);
6207 
6208 	rtnl_unlock();
6209 	return 0;
6210 }
6211 
addrconf_sysctl_disable(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6212 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6213 		void *buffer, size_t *lenp, loff_t *ppos)
6214 {
6215 	int *valp = ctl->data;
6216 	int val = *valp;
6217 	loff_t pos = *ppos;
6218 	struct ctl_table lctl;
6219 	int ret;
6220 
6221 	/*
6222 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6223 	 * not modify it until we get the rtnl lock.
6224 	 */
6225 	lctl = *ctl;
6226 	lctl.data = &val;
6227 
6228 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6229 
6230 	if (write)
6231 		ret = addrconf_disable_ipv6(ctl, valp, val);
6232 	if (ret)
6233 		*ppos = pos;
6234 	return ret;
6235 }
6236 
addrconf_sysctl_proxy_ndp(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6237 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6238 		void *buffer, size_t *lenp, loff_t *ppos)
6239 {
6240 	int *valp = ctl->data;
6241 	int ret;
6242 	int old, new;
6243 
6244 	old = *valp;
6245 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6246 	new = *valp;
6247 
6248 	if (write && old != new) {
6249 		struct net *net = ctl->extra2;
6250 
6251 		if (!rtnl_trylock())
6252 			return restart_syscall();
6253 
6254 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6255 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6256 						     NETCONFA_PROXY_NEIGH,
6257 						     NETCONFA_IFINDEX_DEFAULT,
6258 						     net->ipv6.devconf_dflt);
6259 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6260 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6261 						     NETCONFA_PROXY_NEIGH,
6262 						     NETCONFA_IFINDEX_ALL,
6263 						     net->ipv6.devconf_all);
6264 		else {
6265 			struct inet6_dev *idev = ctl->extra1;
6266 
6267 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6268 						     NETCONFA_PROXY_NEIGH,
6269 						     idev->dev->ifindex,
6270 						     &idev->cnf);
6271 		}
6272 		rtnl_unlock();
6273 	}
6274 
6275 	return ret;
6276 }
6277 
addrconf_sysctl_addr_gen_mode(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6278 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6279 					 void *buffer, size_t *lenp,
6280 					 loff_t *ppos)
6281 {
6282 	int ret = 0;
6283 	u32 new_val;
6284 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6285 	struct net *net = (struct net *)ctl->extra2;
6286 	struct ctl_table tmp = {
6287 		.data = &new_val,
6288 		.maxlen = sizeof(new_val),
6289 		.mode = ctl->mode,
6290 	};
6291 
6292 	if (!rtnl_trylock())
6293 		return restart_syscall();
6294 
6295 	new_val = *((u32 *)ctl->data);
6296 
6297 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6298 	if (ret != 0)
6299 		goto out;
6300 
6301 	if (write) {
6302 		if (check_addr_gen_mode(new_val) < 0) {
6303 			ret = -EINVAL;
6304 			goto out;
6305 		}
6306 
6307 		if (idev) {
6308 			if (check_stable_privacy(idev, net, new_val) < 0) {
6309 				ret = -EINVAL;
6310 				goto out;
6311 			}
6312 
6313 			if (idev->cnf.addr_gen_mode != new_val) {
6314 				idev->cnf.addr_gen_mode = new_val;
6315 				addrconf_dev_config(idev->dev);
6316 			}
6317 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6318 			struct net_device *dev;
6319 
6320 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6321 			for_each_netdev(net, dev) {
6322 				idev = __in6_dev_get(dev);
6323 				if (idev &&
6324 				    idev->cnf.addr_gen_mode != new_val) {
6325 					idev->cnf.addr_gen_mode = new_val;
6326 					addrconf_dev_config(idev->dev);
6327 				}
6328 			}
6329 		}
6330 
6331 		*((u32 *)ctl->data) = new_val;
6332 	}
6333 
6334 out:
6335 	rtnl_unlock();
6336 
6337 	return ret;
6338 }
6339 
addrconf_sysctl_stable_secret(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6340 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6341 					 void *buffer, size_t *lenp,
6342 					 loff_t *ppos)
6343 {
6344 	int err;
6345 	struct in6_addr addr;
6346 	char str[IPV6_MAX_STRLEN];
6347 	struct ctl_table lctl = *ctl;
6348 	struct net *net = ctl->extra2;
6349 	struct ipv6_stable_secret *secret = ctl->data;
6350 
6351 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6352 		return -EIO;
6353 
6354 	lctl.maxlen = IPV6_MAX_STRLEN;
6355 	lctl.data = str;
6356 
6357 	if (!rtnl_trylock())
6358 		return restart_syscall();
6359 
6360 	if (!write && !secret->initialized) {
6361 		err = -EIO;
6362 		goto out;
6363 	}
6364 
6365 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6366 	if (err >= sizeof(str)) {
6367 		err = -EIO;
6368 		goto out;
6369 	}
6370 
6371 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6372 	if (err || !write)
6373 		goto out;
6374 
6375 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6376 		err = -EIO;
6377 		goto out;
6378 	}
6379 
6380 	secret->initialized = true;
6381 	secret->secret = addr;
6382 
6383 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6384 		struct net_device *dev;
6385 
6386 		for_each_netdev(net, dev) {
6387 			struct inet6_dev *idev = __in6_dev_get(dev);
6388 
6389 			if (idev) {
6390 				idev->cnf.addr_gen_mode =
6391 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6392 			}
6393 		}
6394 	} else {
6395 		struct inet6_dev *idev = ctl->extra1;
6396 
6397 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6398 	}
6399 
6400 out:
6401 	rtnl_unlock();
6402 
6403 	return err;
6404 }
6405 
6406 static
addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6407 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6408 						int write, void *buffer,
6409 						size_t *lenp,
6410 						loff_t *ppos)
6411 {
6412 	int *valp = ctl->data;
6413 	int val = *valp;
6414 	loff_t pos = *ppos;
6415 	struct ctl_table lctl;
6416 	int ret;
6417 
6418 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6419 	 * we should not modify it until we get the rtnl lock.
6420 	 */
6421 	lctl = *ctl;
6422 	lctl.data = &val;
6423 
6424 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6425 
6426 	if (write)
6427 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6428 	if (ret)
6429 		*ppos = pos;
6430 	return ret;
6431 }
6432 
6433 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6434 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6435 {
6436 	if (rt) {
6437 		if (action)
6438 			rt->dst.flags |= DST_NOPOLICY;
6439 		else
6440 			rt->dst.flags &= ~DST_NOPOLICY;
6441 	}
6442 }
6443 
6444 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6445 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6446 {
6447 	struct inet6_ifaddr *ifa;
6448 
6449 	read_lock_bh(&idev->lock);
6450 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6451 		spin_lock(&ifa->lock);
6452 		if (ifa->rt) {
6453 			/* host routes only use builtin fib6_nh */
6454 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6455 			int cpu;
6456 
6457 			rcu_read_lock();
6458 			ifa->rt->dst_nopolicy = val ? true : false;
6459 			if (nh->rt6i_pcpu) {
6460 				for_each_possible_cpu(cpu) {
6461 					struct rt6_info **rtp;
6462 
6463 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6464 					addrconf_set_nopolicy(*rtp, val);
6465 				}
6466 			}
6467 			rcu_read_unlock();
6468 		}
6469 		spin_unlock(&ifa->lock);
6470 	}
6471 	read_unlock_bh(&idev->lock);
6472 }
6473 
6474 static
addrconf_disable_policy(struct ctl_table * ctl,int * valp,int val)6475 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6476 {
6477 	struct inet6_dev *idev;
6478 	struct net *net;
6479 
6480 	if (!rtnl_trylock())
6481 		return restart_syscall();
6482 
6483 	*valp = val;
6484 
6485 	net = (struct net *)ctl->extra2;
6486 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6487 		rtnl_unlock();
6488 		return 0;
6489 	}
6490 
6491 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6492 		struct net_device *dev;
6493 
6494 		for_each_netdev(net, dev) {
6495 			idev = __in6_dev_get(dev);
6496 			if (idev)
6497 				addrconf_disable_policy_idev(idev, val);
6498 		}
6499 	} else {
6500 		idev = (struct inet6_dev *)ctl->extra1;
6501 		addrconf_disable_policy_idev(idev, val);
6502 	}
6503 
6504 	rtnl_unlock();
6505 	return 0;
6506 }
6507 
addrconf_sysctl_disable_policy(struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6508 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6509 				   void *buffer, size_t *lenp, loff_t *ppos)
6510 {
6511 	int *valp = ctl->data;
6512 	int val = *valp;
6513 	loff_t pos = *ppos;
6514 	struct ctl_table lctl;
6515 	int ret;
6516 
6517 	lctl = *ctl;
6518 	lctl.data = &val;
6519 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6520 
6521 	if (write && (*valp != val))
6522 		ret = addrconf_disable_policy(ctl, valp, val);
6523 
6524 	if (ret)
6525 		*ppos = pos;
6526 
6527 	return ret;
6528 }
6529 
6530 static int minus_one = -1;
6531 static const int two_five_five = 255;
6532 
6533 static const struct ctl_table addrconf_sysctl[] = {
6534 	{
6535 		.procname	= "forwarding",
6536 		.data		= &ipv6_devconf.forwarding,
6537 		.maxlen		= sizeof(int),
6538 		.mode		= 0644,
6539 		.proc_handler	= addrconf_sysctl_forward,
6540 	},
6541 	{
6542 		.procname	= "hop_limit",
6543 		.data		= &ipv6_devconf.hop_limit,
6544 		.maxlen		= sizeof(int),
6545 		.mode		= 0644,
6546 		.proc_handler	= proc_dointvec_minmax,
6547 		.extra1		= (void *)SYSCTL_ONE,
6548 		.extra2		= (void *)&two_five_five,
6549 	},
6550 	{
6551 		.procname	= "mtu",
6552 		.data		= &ipv6_devconf.mtu6,
6553 		.maxlen		= sizeof(int),
6554 		.mode		= 0644,
6555 		.proc_handler	= addrconf_sysctl_mtu,
6556 	},
6557 	{
6558 		.procname	= "accept_ra",
6559 		.data		= &ipv6_devconf.accept_ra,
6560 		.maxlen		= sizeof(int),
6561 		.mode		= 0644,
6562 		.proc_handler	= proc_dointvec,
6563 	},
6564 	{
6565 		.procname	= "accept_redirects",
6566 		.data		= &ipv6_devconf.accept_redirects,
6567 		.maxlen		= sizeof(int),
6568 		.mode		= 0644,
6569 		.proc_handler	= proc_dointvec,
6570 	},
6571 	{
6572 		.procname	= "autoconf",
6573 		.data		= &ipv6_devconf.autoconf,
6574 		.maxlen		= sizeof(int),
6575 		.mode		= 0644,
6576 		.proc_handler	= proc_dointvec,
6577 	},
6578 	{
6579 		.procname	= "dad_transmits",
6580 		.data		= &ipv6_devconf.dad_transmits,
6581 		.maxlen		= sizeof(int),
6582 		.mode		= 0644,
6583 		.proc_handler	= proc_dointvec,
6584 	},
6585 	{
6586 		.procname	= "router_solicitations",
6587 		.data		= &ipv6_devconf.rtr_solicits,
6588 		.maxlen		= sizeof(int),
6589 		.mode		= 0644,
6590 		.proc_handler	= proc_dointvec_minmax,
6591 		.extra1		= &minus_one,
6592 	},
6593 	{
6594 		.procname	= "router_solicitation_interval",
6595 		.data		= &ipv6_devconf.rtr_solicit_interval,
6596 		.maxlen		= sizeof(int),
6597 		.mode		= 0644,
6598 		.proc_handler	= proc_dointvec_jiffies,
6599 	},
6600 	{
6601 		.procname	= "router_solicitation_max_interval",
6602 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6603 		.maxlen		= sizeof(int),
6604 		.mode		= 0644,
6605 		.proc_handler	= proc_dointvec_jiffies,
6606 	},
6607 	{
6608 		.procname	= "router_solicitation_delay",
6609 		.data		= &ipv6_devconf.rtr_solicit_delay,
6610 		.maxlen		= sizeof(int),
6611 		.mode		= 0644,
6612 		.proc_handler	= proc_dointvec_jiffies,
6613 	},
6614 	{
6615 		.procname	= "force_mld_version",
6616 		.data		= &ipv6_devconf.force_mld_version,
6617 		.maxlen		= sizeof(int),
6618 		.mode		= 0644,
6619 		.proc_handler	= proc_dointvec,
6620 	},
6621 	{
6622 		.procname	= "mldv1_unsolicited_report_interval",
6623 		.data		=
6624 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6625 		.maxlen		= sizeof(int),
6626 		.mode		= 0644,
6627 		.proc_handler	= proc_dointvec_ms_jiffies,
6628 	},
6629 	{
6630 		.procname	= "mldv2_unsolicited_report_interval",
6631 		.data		=
6632 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6633 		.maxlen		= sizeof(int),
6634 		.mode		= 0644,
6635 		.proc_handler	= proc_dointvec_ms_jiffies,
6636 	},
6637 	{
6638 		.procname	= "use_tempaddr",
6639 		.data		= &ipv6_devconf.use_tempaddr,
6640 		.maxlen		= sizeof(int),
6641 		.mode		= 0644,
6642 		.proc_handler	= proc_dointvec,
6643 	},
6644 	{
6645 		.procname	= "temp_valid_lft",
6646 		.data		= &ipv6_devconf.temp_valid_lft,
6647 		.maxlen		= sizeof(int),
6648 		.mode		= 0644,
6649 		.proc_handler	= proc_dointvec,
6650 	},
6651 	{
6652 		.procname	= "temp_prefered_lft",
6653 		.data		= &ipv6_devconf.temp_prefered_lft,
6654 		.maxlen		= sizeof(int),
6655 		.mode		= 0644,
6656 		.proc_handler	= proc_dointvec,
6657 	},
6658 	{
6659 		.procname	= "regen_max_retry",
6660 		.data		= &ipv6_devconf.regen_max_retry,
6661 		.maxlen		= sizeof(int),
6662 		.mode		= 0644,
6663 		.proc_handler	= proc_dointvec,
6664 	},
6665 	{
6666 		.procname	= "max_desync_factor",
6667 		.data		= &ipv6_devconf.max_desync_factor,
6668 		.maxlen		= sizeof(int),
6669 		.mode		= 0644,
6670 		.proc_handler	= proc_dointvec,
6671 	},
6672 	{
6673 		.procname	= "max_addresses",
6674 		.data		= &ipv6_devconf.max_addresses,
6675 		.maxlen		= sizeof(int),
6676 		.mode		= 0644,
6677 		.proc_handler	= proc_dointvec,
6678 	},
6679 	{
6680 		.procname	= "accept_ra_defrtr",
6681 		.data		= &ipv6_devconf.accept_ra_defrtr,
6682 		.maxlen		= sizeof(int),
6683 		.mode		= 0644,
6684 		.proc_handler	= proc_dointvec,
6685 	},
6686 	{
6687 		.procname	= "accept_ra_min_hop_limit",
6688 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6689 		.maxlen		= sizeof(int),
6690 		.mode		= 0644,
6691 		.proc_handler	= proc_dointvec,
6692 	},
6693 	{
6694 		.procname	= "accept_ra_pinfo",
6695 		.data		= &ipv6_devconf.accept_ra_pinfo,
6696 		.maxlen		= sizeof(int),
6697 		.mode		= 0644,
6698 		.proc_handler	= proc_dointvec,
6699 	},
6700 #ifdef CONFIG_IPV6_ROUTER_PREF
6701 	{
6702 		.procname	= "accept_ra_rtr_pref",
6703 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6704 		.maxlen		= sizeof(int),
6705 		.mode		= 0644,
6706 		.proc_handler	= proc_dointvec,
6707 	},
6708 	{
6709 		.procname	= "router_probe_interval",
6710 		.data		= &ipv6_devconf.rtr_probe_interval,
6711 		.maxlen		= sizeof(int),
6712 		.mode		= 0644,
6713 		.proc_handler	= proc_dointvec_jiffies,
6714 	},
6715 #ifdef CONFIG_IPV6_ROUTE_INFO
6716 	{
6717 		.procname	= "accept_ra_rt_info_min_plen",
6718 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6719 		.maxlen		= sizeof(int),
6720 		.mode		= 0644,
6721 		.proc_handler	= proc_dointvec,
6722 	},
6723 	{
6724 		.procname	= "accept_ra_rt_info_max_plen",
6725 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6726 		.maxlen		= sizeof(int),
6727 		.mode		= 0644,
6728 		.proc_handler	= proc_dointvec,
6729 	},
6730 #endif
6731 #endif
6732 	{
6733 		.procname	= "proxy_ndp",
6734 		.data		= &ipv6_devconf.proxy_ndp,
6735 		.maxlen		= sizeof(int),
6736 		.mode		= 0644,
6737 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6738 	},
6739 	{
6740 		.procname	= "accept_source_route",
6741 		.data		= &ipv6_devconf.accept_source_route,
6742 		.maxlen		= sizeof(int),
6743 		.mode		= 0644,
6744 		.proc_handler	= proc_dointvec,
6745 	},
6746 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6747 	{
6748 		.procname	= "optimistic_dad",
6749 		.data		= &ipv6_devconf.optimistic_dad,
6750 		.maxlen		= sizeof(int),
6751 		.mode		= 0644,
6752 		.proc_handler   = proc_dointvec,
6753 	},
6754 	{
6755 		.procname	= "use_optimistic",
6756 		.data		= &ipv6_devconf.use_optimistic,
6757 		.maxlen		= sizeof(int),
6758 		.mode		= 0644,
6759 		.proc_handler	= proc_dointvec,
6760 	},
6761 #endif
6762 #ifdef CONFIG_IPV6_MROUTE
6763 	{
6764 		.procname	= "mc_forwarding",
6765 		.data		= &ipv6_devconf.mc_forwarding,
6766 		.maxlen		= sizeof(int),
6767 		.mode		= 0444,
6768 		.proc_handler	= proc_dointvec,
6769 	},
6770 #endif
6771 	{
6772 		.procname	= "disable_ipv6",
6773 		.data		= &ipv6_devconf.disable_ipv6,
6774 		.maxlen		= sizeof(int),
6775 		.mode		= 0644,
6776 		.proc_handler	= addrconf_sysctl_disable,
6777 	},
6778 	{
6779 		.procname	= "accept_dad",
6780 		.data		= &ipv6_devconf.accept_dad,
6781 		.maxlen		= sizeof(int),
6782 		.mode		= 0644,
6783 		.proc_handler	= proc_dointvec,
6784 	},
6785 	{
6786 		.procname	= "force_tllao",
6787 		.data		= &ipv6_devconf.force_tllao,
6788 		.maxlen		= sizeof(int),
6789 		.mode		= 0644,
6790 		.proc_handler	= proc_dointvec
6791 	},
6792 	{
6793 		.procname	= "ndisc_notify",
6794 		.data		= &ipv6_devconf.ndisc_notify,
6795 		.maxlen		= sizeof(int),
6796 		.mode		= 0644,
6797 		.proc_handler	= proc_dointvec
6798 	},
6799 	{
6800 		.procname	= "suppress_frag_ndisc",
6801 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6802 		.maxlen		= sizeof(int),
6803 		.mode		= 0644,
6804 		.proc_handler	= proc_dointvec
6805 	},
6806 	{
6807 		.procname	= "accept_ra_from_local",
6808 		.data		= &ipv6_devconf.accept_ra_from_local,
6809 		.maxlen		= sizeof(int),
6810 		.mode		= 0644,
6811 		.proc_handler	= proc_dointvec,
6812 	},
6813 	{
6814 		.procname	= "accept_ra_mtu",
6815 		.data		= &ipv6_devconf.accept_ra_mtu,
6816 		.maxlen		= sizeof(int),
6817 		.mode		= 0644,
6818 		.proc_handler	= proc_dointvec,
6819 	},
6820 	{
6821 		.procname	= "stable_secret",
6822 		.data		= &ipv6_devconf.stable_secret,
6823 		.maxlen		= IPV6_MAX_STRLEN,
6824 		.mode		= 0600,
6825 		.proc_handler	= addrconf_sysctl_stable_secret,
6826 	},
6827 	{
6828 		.procname	= "use_oif_addrs_only",
6829 		.data		= &ipv6_devconf.use_oif_addrs_only,
6830 		.maxlen		= sizeof(int),
6831 		.mode		= 0644,
6832 		.proc_handler	= proc_dointvec,
6833 	},
6834 	{
6835 		.procname	= "ignore_routes_with_linkdown",
6836 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6837 		.maxlen		= sizeof(int),
6838 		.mode		= 0644,
6839 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6840 	},
6841 	{
6842 		.procname	= "drop_unicast_in_l2_multicast",
6843 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6844 		.maxlen		= sizeof(int),
6845 		.mode		= 0644,
6846 		.proc_handler	= proc_dointvec,
6847 	},
6848 	{
6849 		.procname	= "drop_unsolicited_na",
6850 		.data		= &ipv6_devconf.drop_unsolicited_na,
6851 		.maxlen		= sizeof(int),
6852 		.mode		= 0644,
6853 		.proc_handler	= proc_dointvec,
6854 	},
6855 	{
6856 		.procname	= "keep_addr_on_down",
6857 		.data		= &ipv6_devconf.keep_addr_on_down,
6858 		.maxlen		= sizeof(int),
6859 		.mode		= 0644,
6860 		.proc_handler	= proc_dointvec,
6861 
6862 	},
6863 	{
6864 		.procname	= "seg6_enabled",
6865 		.data		= &ipv6_devconf.seg6_enabled,
6866 		.maxlen		= sizeof(int),
6867 		.mode		= 0644,
6868 		.proc_handler	= proc_dointvec,
6869 	},
6870 #ifdef CONFIG_IPV6_SEG6_HMAC
6871 	{
6872 		.procname	= "seg6_require_hmac",
6873 		.data		= &ipv6_devconf.seg6_require_hmac,
6874 		.maxlen		= sizeof(int),
6875 		.mode		= 0644,
6876 		.proc_handler	= proc_dointvec,
6877 	},
6878 #endif
6879 	{
6880 		.procname       = "enhanced_dad",
6881 		.data           = &ipv6_devconf.enhanced_dad,
6882 		.maxlen         = sizeof(int),
6883 		.mode           = 0644,
6884 		.proc_handler   = proc_dointvec,
6885 	},
6886 	{
6887 		.procname		= "addr_gen_mode",
6888 		.data			= &ipv6_devconf.addr_gen_mode,
6889 		.maxlen			= sizeof(int),
6890 		.mode			= 0644,
6891 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6892 	},
6893 	{
6894 		.procname       = "disable_policy",
6895 		.data           = &ipv6_devconf.disable_policy,
6896 		.maxlen         = sizeof(int),
6897 		.mode           = 0644,
6898 		.proc_handler   = addrconf_sysctl_disable_policy,
6899 	},
6900 	{
6901 		.procname	= "ndisc_tclass",
6902 		.data		= &ipv6_devconf.ndisc_tclass,
6903 		.maxlen		= sizeof(int),
6904 		.mode		= 0644,
6905 		.proc_handler	= proc_dointvec_minmax,
6906 		.extra1		= (void *)SYSCTL_ZERO,
6907 		.extra2		= (void *)&two_five_five,
6908 	},
6909 	{
6910 		.procname	= "rpl_seg_enabled",
6911 		.data		= &ipv6_devconf.rpl_seg_enabled,
6912 		.maxlen		= sizeof(int),
6913 		.mode		= 0644,
6914 		.proc_handler	= proc_dointvec,
6915 	},
6916 	{
6917 		/* sentinel */
6918 	}
6919 };
6920 
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)6921 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6922 		struct inet6_dev *idev, struct ipv6_devconf *p)
6923 {
6924 	int i, ifindex;
6925 	struct ctl_table *table;
6926 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6927 
6928 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6929 	if (!table)
6930 		goto out;
6931 
6932 	for (i = 0; table[i].data; i++) {
6933 		table[i].data += (char *)p - (char *)&ipv6_devconf;
6934 		/* If one of these is already set, then it is not safe to
6935 		 * overwrite either of them: this makes proc_dointvec_minmax
6936 		 * usable.
6937 		 */
6938 		if (!table[i].extra1 && !table[i].extra2) {
6939 			table[i].extra1 = idev; /* embedded; no ref */
6940 			table[i].extra2 = net;
6941 		}
6942 	}
6943 
6944 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6945 
6946 	p->sysctl_header = register_net_sysctl(net, path, table);
6947 	if (!p->sysctl_header)
6948 		goto free;
6949 
6950 	if (!strcmp(dev_name, "all"))
6951 		ifindex = NETCONFA_IFINDEX_ALL;
6952 	else if (!strcmp(dev_name, "default"))
6953 		ifindex = NETCONFA_IFINDEX_DEFAULT;
6954 	else
6955 		ifindex = idev->dev->ifindex;
6956 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6957 				     ifindex, p);
6958 	return 0;
6959 
6960 free:
6961 	kfree(table);
6962 out:
6963 	return -ENOBUFS;
6964 }
6965 
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)6966 static void __addrconf_sysctl_unregister(struct net *net,
6967 					 struct ipv6_devconf *p, int ifindex)
6968 {
6969 	struct ctl_table *table;
6970 
6971 	if (!p->sysctl_header)
6972 		return;
6973 
6974 	table = p->sysctl_header->ctl_table_arg;
6975 	unregister_net_sysctl_table(p->sysctl_header);
6976 	p->sysctl_header = NULL;
6977 	kfree(table);
6978 
6979 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6980 }
6981 
addrconf_sysctl_register(struct inet6_dev * idev)6982 static int addrconf_sysctl_register(struct inet6_dev *idev)
6983 {
6984 	int err;
6985 
6986 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
6987 		return -EINVAL;
6988 
6989 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6990 				    &ndisc_ifinfo_sysctl_change);
6991 	if (err)
6992 		return err;
6993 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6994 					 idev, &idev->cnf);
6995 	if (err)
6996 		neigh_sysctl_unregister(idev->nd_parms);
6997 
6998 	return err;
6999 }
7000 
addrconf_sysctl_unregister(struct inet6_dev * idev)7001 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7002 {
7003 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7004 				     idev->dev->ifindex);
7005 	neigh_sysctl_unregister(idev->nd_parms);
7006 }
7007 
7008 
7009 #endif
7010 
addrconf_init_net(struct net * net)7011 static int __net_init addrconf_init_net(struct net *net)
7012 {
7013 	int err = -ENOMEM;
7014 	struct ipv6_devconf *all, *dflt;
7015 
7016 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7017 	if (!all)
7018 		goto err_alloc_all;
7019 
7020 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7021 	if (!dflt)
7022 		goto err_alloc_dflt;
7023 
7024 	if (IS_ENABLED(CONFIG_SYSCTL) &&
7025 	    !net_eq(net, &init_net)) {
7026 		switch (sysctl_devconf_inherit_init_net) {
7027 		case 1:  /* copy from init_net */
7028 			memcpy(all, init_net.ipv6.devconf_all,
7029 			       sizeof(ipv6_devconf));
7030 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7031 			       sizeof(ipv6_devconf_dflt));
7032 			break;
7033 		case 3: /* copy from the current netns */
7034 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7035 			       sizeof(ipv6_devconf));
7036 			memcpy(dflt,
7037 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7038 			       sizeof(ipv6_devconf_dflt));
7039 			break;
7040 		case 0:
7041 		case 2:
7042 			/* use compiled values */
7043 			break;
7044 		}
7045 	}
7046 
7047 	/* these will be inherited by all namespaces */
7048 	dflt->autoconf = ipv6_defaults.autoconf;
7049 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7050 
7051 	dflt->stable_secret.initialized = false;
7052 	all->stable_secret.initialized = false;
7053 
7054 	net->ipv6.devconf_all = all;
7055 	net->ipv6.devconf_dflt = dflt;
7056 
7057 #ifdef CONFIG_SYSCTL
7058 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7059 	if (err < 0)
7060 		goto err_reg_all;
7061 
7062 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7063 	if (err < 0)
7064 		goto err_reg_dflt;
7065 #endif
7066 	return 0;
7067 
7068 #ifdef CONFIG_SYSCTL
7069 err_reg_dflt:
7070 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7071 err_reg_all:
7072 	kfree(dflt);
7073 #endif
7074 err_alloc_dflt:
7075 	kfree(all);
7076 err_alloc_all:
7077 	return err;
7078 }
7079 
addrconf_exit_net(struct net * net)7080 static void __net_exit addrconf_exit_net(struct net *net)
7081 {
7082 #ifdef CONFIG_SYSCTL
7083 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7084 				     NETCONFA_IFINDEX_DEFAULT);
7085 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7086 				     NETCONFA_IFINDEX_ALL);
7087 #endif
7088 	kfree(net->ipv6.devconf_dflt);
7089 	kfree(net->ipv6.devconf_all);
7090 }
7091 
7092 static struct pernet_operations addrconf_ops = {
7093 	.init = addrconf_init_net,
7094 	.exit = addrconf_exit_net,
7095 };
7096 
7097 static struct rtnl_af_ops inet6_ops __read_mostly = {
7098 	.family		  = AF_INET6,
7099 	.fill_link_af	  = inet6_fill_link_af,
7100 	.get_link_af_size = inet6_get_link_af_size,
7101 	.validate_link_af = inet6_validate_link_af,
7102 	.set_link_af	  = inet6_set_link_af,
7103 };
7104 
7105 /*
7106  *	Init / cleanup code
7107  */
7108 
addrconf_init(void)7109 int __init addrconf_init(void)
7110 {
7111 	struct inet6_dev *idev;
7112 	int i, err;
7113 
7114 	err = ipv6_addr_label_init();
7115 	if (err < 0) {
7116 		pr_crit("%s: cannot initialize default policy table: %d\n",
7117 			__func__, err);
7118 		goto out;
7119 	}
7120 
7121 	err = register_pernet_subsys(&addrconf_ops);
7122 	if (err < 0)
7123 		goto out_addrlabel;
7124 
7125 	addrconf_wq = create_workqueue("ipv6_addrconf");
7126 	if (!addrconf_wq) {
7127 		err = -ENOMEM;
7128 		goto out_nowq;
7129 	}
7130 
7131 	/* The addrconf netdev notifier requires that loopback_dev
7132 	 * has it's ipv6 private information allocated and setup
7133 	 * before it can bring up and give link-local addresses
7134 	 * to other devices which are up.
7135 	 *
7136 	 * Unfortunately, loopback_dev is not necessarily the first
7137 	 * entry in the global dev_base list of net devices.  In fact,
7138 	 * it is likely to be the very last entry on that list.
7139 	 * So this causes the notifier registry below to try and
7140 	 * give link-local addresses to all devices besides loopback_dev
7141 	 * first, then loopback_dev, which cases all the non-loopback_dev
7142 	 * devices to fail to get a link-local address.
7143 	 *
7144 	 * So, as a temporary fix, allocate the ipv6 structure for
7145 	 * loopback_dev first by hand.
7146 	 * Longer term, all of the dependencies ipv6 has upon the loopback
7147 	 * device and it being up should be removed.
7148 	 */
7149 	rtnl_lock();
7150 	idev = ipv6_add_dev(init_net.loopback_dev);
7151 	rtnl_unlock();
7152 	if (IS_ERR(idev)) {
7153 		err = PTR_ERR(idev);
7154 		goto errlo;
7155 	}
7156 
7157 	ip6_route_init_special_entries();
7158 
7159 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7160 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7161 
7162 	register_netdevice_notifier(&ipv6_dev_notf);
7163 
7164 	addrconf_verify();
7165 
7166 	rtnl_af_register(&inet6_ops);
7167 
7168 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7169 				   NULL, inet6_dump_ifinfo, 0);
7170 	if (err < 0)
7171 		goto errout;
7172 
7173 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7174 				   inet6_rtm_newaddr, NULL, 0);
7175 	if (err < 0)
7176 		goto errout;
7177 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7178 				   inet6_rtm_deladdr, NULL, 0);
7179 	if (err < 0)
7180 		goto errout;
7181 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7182 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7183 				   RTNL_FLAG_DOIT_UNLOCKED);
7184 	if (err < 0)
7185 		goto errout;
7186 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7187 				   NULL, inet6_dump_ifmcaddr, 0);
7188 	if (err < 0)
7189 		goto errout;
7190 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7191 				   NULL, inet6_dump_ifacaddr, 0);
7192 	if (err < 0)
7193 		goto errout;
7194 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7195 				   inet6_netconf_get_devconf,
7196 				   inet6_netconf_dump_devconf,
7197 				   RTNL_FLAG_DOIT_UNLOCKED);
7198 	if (err < 0)
7199 		goto errout;
7200 	err = ipv6_addr_label_rtnl_register();
7201 	if (err < 0)
7202 		goto errout;
7203 
7204 	return 0;
7205 errout:
7206 	rtnl_unregister_all(PF_INET6);
7207 	rtnl_af_unregister(&inet6_ops);
7208 	unregister_netdevice_notifier(&ipv6_dev_notf);
7209 errlo:
7210 	destroy_workqueue(addrconf_wq);
7211 out_nowq:
7212 	unregister_pernet_subsys(&addrconf_ops);
7213 out_addrlabel:
7214 	ipv6_addr_label_cleanup();
7215 out:
7216 	return err;
7217 }
7218 
addrconf_cleanup(void)7219 void addrconf_cleanup(void)
7220 {
7221 	struct net_device *dev;
7222 	int i;
7223 
7224 	unregister_netdevice_notifier(&ipv6_dev_notf);
7225 	unregister_pernet_subsys(&addrconf_ops);
7226 	ipv6_addr_label_cleanup();
7227 
7228 	rtnl_af_unregister(&inet6_ops);
7229 
7230 	rtnl_lock();
7231 
7232 	/* clean dev list */
7233 	for_each_netdev(&init_net, dev) {
7234 		if (__in6_dev_get(dev) == NULL)
7235 			continue;
7236 		addrconf_ifdown(dev, true);
7237 	}
7238 	addrconf_ifdown(init_net.loopback_dev, true);
7239 
7240 	/*
7241 	 *	Check hash table.
7242 	 */
7243 	spin_lock_bh(&addrconf_hash_lock);
7244 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7245 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7246 	spin_unlock_bh(&addrconf_hash_lock);
7247 	cancel_delayed_work(&addr_chk_work);
7248 	rtnl_unlock();
7249 
7250 	destroy_workqueue(addrconf_wq);
7251 }
7252