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_min_lft	= 0,
211 	.accept_ra_pinfo	= 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213 	.accept_ra_rtr_pref	= 1,
214 	.rtr_probe_interval	= 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216 	.accept_ra_rt_info_min_plen = 0,
217 	.accept_ra_rt_info_max_plen = 0,
218 #endif
219 #endif
220 	.proxy_ndp		= 0,
221 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
222 	.disable_ipv6		= 0,
223 	.accept_dad		= 0,
224 	.suppress_frag_ndisc	= 1,
225 	.accept_ra_mtu		= 1,
226 	.stable_secret		= {
227 		.initialized = false,
228 	},
229 	.use_oif_addrs_only	= 0,
230 	.ignore_routes_with_linkdown = 0,
231 	.keep_addr_on_down	= 0,
232 	.seg6_enabled		= 0,
233 #ifdef CONFIG_IPV6_SEG6_HMAC
234 	.seg6_require_hmac	= 0,
235 #endif
236 	.enhanced_dad           = 1,
237 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
238 	.disable_policy		= 0,
239 	.rpl_seg_enabled	= 0,
240 };
241 
242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
243 	.forwarding		= 0,
244 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
245 	.mtu6			= IPV6_MIN_MTU,
246 	.accept_ra		= 1,
247 	.accept_redirects	= 1,
248 	.autoconf		= 1,
249 	.force_mld_version	= 0,
250 	.mldv1_unsolicited_report_interval = 10 * HZ,
251 	.mldv2_unsolicited_report_interval = HZ,
252 	.dad_transmits		= 1,
253 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
254 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
255 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
256 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
257 	.use_tempaddr		= 0,
258 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
259 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
260 	.regen_max_retry	= REGEN_MAX_RETRY,
261 	.max_desync_factor	= MAX_DESYNC_FACTOR,
262 	.max_addresses		= IPV6_MAX_ADDRESSES,
263 	.accept_ra_defrtr	= 1,
264 	.accept_ra_from_local	= 0,
265 	.accept_ra_min_hop_limit= 1,
266 	.accept_ra_min_lft	= 0,
267 	.accept_ra_pinfo	= 1,
268 #ifdef CONFIG_IPV6_ROUTER_PREF
269 	.accept_ra_rtr_pref	= 1,
270 	.rtr_probe_interval	= 60 * HZ,
271 #ifdef CONFIG_IPV6_ROUTE_INFO
272 	.accept_ra_rt_info_min_plen = 0,
273 	.accept_ra_rt_info_max_plen = 0,
274 #endif
275 #endif
276 	.proxy_ndp		= 0,
277 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
278 	.disable_ipv6		= 0,
279 	.accept_dad		= 1,
280 	.suppress_frag_ndisc	= 1,
281 	.accept_ra_mtu		= 1,
282 	.stable_secret		= {
283 		.initialized = false,
284 	},
285 	.use_oif_addrs_only	= 0,
286 	.ignore_routes_with_linkdown = 0,
287 	.keep_addr_on_down	= 0,
288 	.seg6_enabled		= 0,
289 #ifdef CONFIG_IPV6_SEG6_HMAC
290 	.seg6_require_hmac	= 0,
291 #endif
292 	.enhanced_dad           = 1,
293 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
294 	.disable_policy		= 0,
295 	.rpl_seg_enabled	= 0,
296 };
297 
298 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device *dev)299 static inline bool addrconf_link_ready(const struct net_device *dev)
300 {
301 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
302 }
303 
addrconf_del_rs_timer(struct inet6_dev *idev)304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
305 {
306 	if (del_timer(&idev->rs_timer))
307 		__in6_dev_put(idev);
308 }
309 
addrconf_del_dad_work(struct inet6_ifaddr *ifp)310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
311 {
312 	if (cancel_delayed_work(&ifp->dad_work))
313 		__in6_ifa_put(ifp);
314 }
315 
addrconf_mod_rs_timer(struct inet6_dev *idev, unsigned long when)316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
317 				  unsigned long when)
318 {
319 	if (!mod_timer(&idev->rs_timer, jiffies + when))
320 		in6_dev_hold(idev);
321 }
322 
addrconf_mod_dad_work(struct inet6_ifaddr *ifp, unsigned long delay)323 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
324 				   unsigned long delay)
325 {
326 	in6_ifa_hold(ifp);
327 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
328 		in6_ifa_put(ifp);
329 }
330 
snmp6_alloc_dev(struct inet6_dev *idev)331 static int snmp6_alloc_dev(struct inet6_dev *idev)
332 {
333 	int i;
334 
335 	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
336 	if (!idev->stats.ipv6)
337 		goto err_ip;
338 
339 	for_each_possible_cpu(i) {
340 		struct ipstats_mib *addrconf_stats;
341 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
342 		u64_stats_init(&addrconf_stats->syncp);
343 	}
344 
345 
346 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
347 					GFP_KERNEL);
348 	if (!idev->stats.icmpv6dev)
349 		goto err_icmp;
350 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
351 					   GFP_KERNEL);
352 	if (!idev->stats.icmpv6msgdev)
353 		goto err_icmpmsg;
354 
355 	return 0;
356 
357 err_icmpmsg:
358 	kfree(idev->stats.icmpv6dev);
359 err_icmp:
360 	free_percpu(idev->stats.ipv6);
361 err_ip:
362 	return -ENOMEM;
363 }
364 
ipv6_add_dev(struct net_device *dev)365 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
366 {
367 	struct inet6_dev *ndev;
368 	int err = -ENOMEM;
369 
370 	ASSERT_RTNL();
371 
372 	if (dev->mtu < IPV6_MIN_MTU)
373 		return ERR_PTR(-EINVAL);
374 
375 	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
376 	if (!ndev)
377 		return ERR_PTR(err);
378 
379 	rwlock_init(&ndev->lock);
380 	ndev->dev = dev;
381 	INIT_LIST_HEAD(&ndev->addr_list);
382 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
383 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
384 
385 	if (ndev->cnf.stable_secret.initialized)
386 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
387 
388 	ndev->cnf.mtu6 = dev->mtu;
389 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
390 	if (!ndev->nd_parms) {
391 		kfree(ndev);
392 		return ERR_PTR(err);
393 	}
394 	if (ndev->cnf.forwarding)
395 		dev_disable_lro(dev);
396 	/* We refer to the device */
397 	dev_hold(dev);
398 
399 	if (snmp6_alloc_dev(ndev) < 0) {
400 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
401 			   __func__);
402 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
403 		dev_put(dev);
404 		kfree(ndev);
405 		return ERR_PTR(err);
406 	}
407 
408 	if (snmp6_register_dev(ndev) < 0) {
409 		netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
410 			   __func__, dev->name);
411 		goto err_release;
412 	}
413 
414 	/* One reference from device. */
415 	refcount_set(&ndev->refcnt, 1);
416 
417 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
418 		ndev->cnf.accept_dad = -1;
419 
420 #if IS_ENABLED(CONFIG_IPV6_SIT)
421 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
422 		pr_info("%s: Disabled Multicast RS\n", dev->name);
423 		ndev->cnf.rtr_solicits = 0;
424 	}
425 #endif
426 
427 	INIT_LIST_HEAD(&ndev->tempaddr_list);
428 	ndev->desync_factor = U32_MAX;
429 	if ((dev->flags&IFF_LOOPBACK) ||
430 	    dev->type == ARPHRD_TUNNEL ||
431 	    dev->type == ARPHRD_TUNNEL6 ||
432 	    dev->type == ARPHRD_SIT ||
433 	    dev->type == ARPHRD_NONE) {
434 		ndev->cnf.use_tempaddr = -1;
435 	}
436 
437 	ndev->token = in6addr_any;
438 
439 	if (netif_running(dev) && addrconf_link_ready(dev))
440 		ndev->if_flags |= IF_READY;
441 
442 	ipv6_mc_init_dev(ndev);
443 	ndev->tstamp = jiffies;
444 	err = addrconf_sysctl_register(ndev);
445 	if (err) {
446 		ipv6_mc_destroy_dev(ndev);
447 		snmp6_unregister_dev(ndev);
448 		goto err_release;
449 	}
450 	/* protected by rtnl_lock */
451 	rcu_assign_pointer(dev->ip6_ptr, ndev);
452 
453 	/* Join interface-local all-node multicast group */
454 	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
455 
456 	/* Join all-node multicast group */
457 	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
458 
459 	/* Join all-router multicast group if forwarding is set */
460 	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
461 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
462 
463 	return ndev;
464 
465 err_release:
466 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
467 	ndev->dead = 1;
468 	in6_dev_finish_destroy(ndev);
469 	return ERR_PTR(err);
470 }
471 
ipv6_find_idev(struct net_device *dev)472 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
473 {
474 	struct inet6_dev *idev;
475 
476 	ASSERT_RTNL();
477 
478 	idev = __in6_dev_get(dev);
479 	if (!idev) {
480 		idev = ipv6_add_dev(dev);
481 		if (IS_ERR(idev))
482 			return idev;
483 	}
484 
485 	if (dev->flags&IFF_UP)
486 		ipv6_mc_up(idev);
487 	return idev;
488 }
489 
inet6_netconf_msgsize_devconf(int type)490 static int inet6_netconf_msgsize_devconf(int type)
491 {
492 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
493 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
494 	bool all = false;
495 
496 	if (type == NETCONFA_ALL)
497 		all = true;
498 
499 	if (all || type == NETCONFA_FORWARDING)
500 		size += nla_total_size(4);
501 #ifdef CONFIG_IPV6_MROUTE
502 	if (all || type == NETCONFA_MC_FORWARDING)
503 		size += nla_total_size(4);
504 #endif
505 	if (all || type == NETCONFA_PROXY_NEIGH)
506 		size += nla_total_size(4);
507 
508 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
509 		size += nla_total_size(4);
510 
511 	return size;
512 }
513 
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)514 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
515 				      struct ipv6_devconf *devconf, u32 portid,
516 				      u32 seq, int event, unsigned int flags,
517 				      int type)
518 {
519 	struct nlmsghdr  *nlh;
520 	struct netconfmsg *ncm;
521 	bool all = false;
522 
523 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
524 			flags);
525 	if (!nlh)
526 		return -EMSGSIZE;
527 
528 	if (type == NETCONFA_ALL)
529 		all = true;
530 
531 	ncm = nlmsg_data(nlh);
532 	ncm->ncm_family = AF_INET6;
533 
534 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
535 		goto nla_put_failure;
536 
537 	if (!devconf)
538 		goto out;
539 
540 	if ((all || type == NETCONFA_FORWARDING) &&
541 	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
542 		goto nla_put_failure;
543 #ifdef CONFIG_IPV6_MROUTE
544 	if ((all || type == NETCONFA_MC_FORWARDING) &&
545 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
546 			atomic_read(&devconf->mc_forwarding)) < 0)
547 		goto nla_put_failure;
548 #endif
549 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
550 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
551 		goto nla_put_failure;
552 
553 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
554 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
555 			devconf->ignore_routes_with_linkdown) < 0)
556 		goto nla_put_failure;
557 
558 out:
559 	nlmsg_end(skb, nlh);
560 	return 0;
561 
562 nla_put_failure:
563 	nlmsg_cancel(skb, nlh);
564 	return -EMSGSIZE;
565 }
566 
inet6_netconf_notify_devconf(struct net *net, int event, int type, int ifindex, struct ipv6_devconf *devconf)567 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
568 				  int ifindex, struct ipv6_devconf *devconf)
569 {
570 	struct sk_buff *skb;
571 	int err = -ENOBUFS;
572 
573 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
574 	if (!skb)
575 		goto errout;
576 
577 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
578 					 event, 0, type);
579 	if (err < 0) {
580 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
581 		WARN_ON(err == -EMSGSIZE);
582 		kfree_skb(skb);
583 		goto errout;
584 	}
585 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
586 	return;
587 errout:
588 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
589 }
590 
591 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
592 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
593 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
594 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
595 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
596 };
597 
inet6_netconf_valid_get_req(struct sk_buff *skb, const struct nlmsghdr *nlh, struct nlattr **tb, struct netlink_ext_ack *extack)598 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
599 				       const struct nlmsghdr *nlh,
600 				       struct nlattr **tb,
601 				       struct netlink_ext_ack *extack)
602 {
603 	int i, err;
604 
605 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
606 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
607 		return -EINVAL;
608 	}
609 
610 	if (!netlink_strict_get_check(skb))
611 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
612 					      tb, NETCONFA_MAX,
613 					      devconf_ipv6_policy, extack);
614 
615 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
616 					    tb, NETCONFA_MAX,
617 					    devconf_ipv6_policy, extack);
618 	if (err)
619 		return err;
620 
621 	for (i = 0; i <= NETCONFA_MAX; i++) {
622 		if (!tb[i])
623 			continue;
624 
625 		switch (i) {
626 		case NETCONFA_IFINDEX:
627 			break;
628 		default:
629 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
630 			return -EINVAL;
631 		}
632 	}
633 
634 	return 0;
635 }
636 
inet6_netconf_get_devconf(struct sk_buff *in_skb, struct nlmsghdr *nlh, struct netlink_ext_ack *extack)637 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
638 				     struct nlmsghdr *nlh,
639 				     struct netlink_ext_ack *extack)
640 {
641 	struct net *net = sock_net(in_skb->sk);
642 	struct nlattr *tb[NETCONFA_MAX+1];
643 	struct inet6_dev *in6_dev = NULL;
644 	struct net_device *dev = NULL;
645 	struct sk_buff *skb;
646 	struct ipv6_devconf *devconf;
647 	int ifindex;
648 	int err;
649 
650 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
651 	if (err < 0)
652 		return err;
653 
654 	if (!tb[NETCONFA_IFINDEX])
655 		return -EINVAL;
656 
657 	err = -EINVAL;
658 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
659 	switch (ifindex) {
660 	case NETCONFA_IFINDEX_ALL:
661 		devconf = net->ipv6.devconf_all;
662 		break;
663 	case NETCONFA_IFINDEX_DEFAULT:
664 		devconf = net->ipv6.devconf_dflt;
665 		break;
666 	default:
667 		dev = dev_get_by_index(net, ifindex);
668 		if (!dev)
669 			return -EINVAL;
670 		in6_dev = in6_dev_get(dev);
671 		if (!in6_dev)
672 			goto errout;
673 		devconf = &in6_dev->cnf;
674 		break;
675 	}
676 
677 	err = -ENOBUFS;
678 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
679 	if (!skb)
680 		goto errout;
681 
682 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
683 					 NETLINK_CB(in_skb).portid,
684 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
685 					 NETCONFA_ALL);
686 	if (err < 0) {
687 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
688 		WARN_ON(err == -EMSGSIZE);
689 		kfree_skb(skb);
690 		goto errout;
691 	}
692 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
693 errout:
694 	if (in6_dev)
695 		in6_dev_put(in6_dev);
696 	if (dev)
697 		dev_put(dev);
698 	return err;
699 }
700 
inet6_netconf_dump_devconf(struct sk_buff *skb, struct netlink_callback *cb)701 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
702 				      struct netlink_callback *cb)
703 {
704 	const struct nlmsghdr *nlh = cb->nlh;
705 	struct net *net = sock_net(skb->sk);
706 	int h, s_h;
707 	int idx, s_idx;
708 	struct net_device *dev;
709 	struct inet6_dev *idev;
710 	struct hlist_head *head;
711 
712 	if (cb->strict_check) {
713 		struct netlink_ext_ack *extack = cb->extack;
714 		struct netconfmsg *ncm;
715 
716 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
717 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
718 			return -EINVAL;
719 		}
720 
721 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
722 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
723 			return -EINVAL;
724 		}
725 	}
726 
727 	s_h = cb->args[0];
728 	s_idx = idx = cb->args[1];
729 
730 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
731 		idx = 0;
732 		head = &net->dev_index_head[h];
733 		rcu_read_lock();
734 		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
735 			  net->dev_base_seq;
736 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
737 			if (idx < s_idx)
738 				goto cont;
739 			idev = __in6_dev_get(dev);
740 			if (!idev)
741 				goto cont;
742 
743 			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
744 						       &idev->cnf,
745 						       NETLINK_CB(cb->skb).portid,
746 						       nlh->nlmsg_seq,
747 						       RTM_NEWNETCONF,
748 						       NLM_F_MULTI,
749 						       NETCONFA_ALL) < 0) {
750 				rcu_read_unlock();
751 				goto done;
752 			}
753 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
754 cont:
755 			idx++;
756 		}
757 		rcu_read_unlock();
758 	}
759 	if (h == NETDEV_HASHENTRIES) {
760 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
761 					       net->ipv6.devconf_all,
762 					       NETLINK_CB(cb->skb).portid,
763 					       nlh->nlmsg_seq,
764 					       RTM_NEWNETCONF, NLM_F_MULTI,
765 					       NETCONFA_ALL) < 0)
766 			goto done;
767 		else
768 			h++;
769 	}
770 	if (h == NETDEV_HASHENTRIES + 1) {
771 		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
772 					       net->ipv6.devconf_dflt,
773 					       NETLINK_CB(cb->skb).portid,
774 					       nlh->nlmsg_seq,
775 					       RTM_NEWNETCONF, NLM_F_MULTI,
776 					       NETCONFA_ALL) < 0)
777 			goto done;
778 		else
779 			h++;
780 	}
781 done:
782 	cb->args[0] = h;
783 	cb->args[1] = idx;
784 
785 	return skb->len;
786 }
787 
788 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev *idev)789 static void dev_forward_change(struct inet6_dev *idev)
790 {
791 	struct net_device *dev;
792 	struct inet6_ifaddr *ifa;
793 	LIST_HEAD(tmp_addr_list);
794 
795 	if (!idev)
796 		return;
797 	dev = idev->dev;
798 	if (idev->cnf.forwarding)
799 		dev_disable_lro(dev);
800 	if (dev->flags & IFF_MULTICAST) {
801 		if (idev->cnf.forwarding) {
802 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
803 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
804 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
805 		} else {
806 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
807 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
808 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
809 		}
810 	}
811 
812 	read_lock_bh(&idev->lock);
813 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
814 		if (ifa->flags&IFA_F_TENTATIVE)
815 			continue;
816 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
817 	}
818 	read_unlock_bh(&idev->lock);
819 
820 	while (!list_empty(&tmp_addr_list)) {
821 		ifa = list_first_entry(&tmp_addr_list,
822 				       struct inet6_ifaddr, if_list_aux);
823 		list_del(&ifa->if_list_aux);
824 		if (idev->cnf.forwarding)
825 			addrconf_join_anycast(ifa);
826 		else
827 			addrconf_leave_anycast(ifa);
828 	}
829 
830 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
831 				     NETCONFA_FORWARDING,
832 				     dev->ifindex, &idev->cnf);
833 }
834 
835 
addrconf_forward_change(struct net *net, __s32 newf)836 static void addrconf_forward_change(struct net *net, __s32 newf)
837 {
838 	struct net_device *dev;
839 	struct inet6_dev *idev;
840 
841 	for_each_netdev(net, dev) {
842 		idev = __in6_dev_get(dev);
843 		if (idev) {
844 			int changed = (!idev->cnf.forwarding) ^ (!newf);
845 			idev->cnf.forwarding = newf;
846 			if (changed)
847 				dev_forward_change(idev);
848 		}
849 	}
850 }
851 
addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)852 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
853 {
854 	struct net *net;
855 	int old;
856 
857 	if (!rtnl_trylock())
858 		return restart_syscall();
859 
860 	net = (struct net *)table->extra2;
861 	old = *p;
862 	*p = newf;
863 
864 	if (p == &net->ipv6.devconf_dflt->forwarding) {
865 		if ((!newf) ^ (!old))
866 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
867 						     NETCONFA_FORWARDING,
868 						     NETCONFA_IFINDEX_DEFAULT,
869 						     net->ipv6.devconf_dflt);
870 		rtnl_unlock();
871 		return 0;
872 	}
873 
874 	if (p == &net->ipv6.devconf_all->forwarding) {
875 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
876 
877 		net->ipv6.devconf_dflt->forwarding = newf;
878 		if ((!newf) ^ (!old_dflt))
879 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
880 						     NETCONFA_FORWARDING,
881 						     NETCONFA_IFINDEX_DEFAULT,
882 						     net->ipv6.devconf_dflt);
883 
884 		addrconf_forward_change(net, newf);
885 		if ((!newf) ^ (!old))
886 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
887 						     NETCONFA_FORWARDING,
888 						     NETCONFA_IFINDEX_ALL,
889 						     net->ipv6.devconf_all);
890 	} else if ((!newf) ^ (!old))
891 		dev_forward_change((struct inet6_dev *)table->extra1);
892 	rtnl_unlock();
893 
894 	if (newf)
895 		rt6_purge_dflt_routers(net);
896 	return 1;
897 }
898 
addrconf_linkdown_change(struct net *net, __s32 newf)899 static void addrconf_linkdown_change(struct net *net, __s32 newf)
900 {
901 	struct net_device *dev;
902 	struct inet6_dev *idev;
903 
904 	for_each_netdev(net, dev) {
905 		idev = __in6_dev_get(dev);
906 		if (idev) {
907 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
908 
909 			idev->cnf.ignore_routes_with_linkdown = newf;
910 			if (changed)
911 				inet6_netconf_notify_devconf(dev_net(dev),
912 							     RTM_NEWNETCONF,
913 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
914 							     dev->ifindex,
915 							     &idev->cnf);
916 		}
917 	}
918 }
919 
addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)920 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
921 {
922 	struct net *net;
923 	int old;
924 
925 	if (!rtnl_trylock())
926 		return restart_syscall();
927 
928 	net = (struct net *)table->extra2;
929 	old = *p;
930 	*p = newf;
931 
932 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
933 		if ((!newf) ^ (!old))
934 			inet6_netconf_notify_devconf(net,
935 						     RTM_NEWNETCONF,
936 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
937 						     NETCONFA_IFINDEX_DEFAULT,
938 						     net->ipv6.devconf_dflt);
939 		rtnl_unlock();
940 		return 0;
941 	}
942 
943 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
944 		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
945 		addrconf_linkdown_change(net, newf);
946 		if ((!newf) ^ (!old))
947 			inet6_netconf_notify_devconf(net,
948 						     RTM_NEWNETCONF,
949 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
950 						     NETCONFA_IFINDEX_ALL,
951 						     net->ipv6.devconf_all);
952 	}
953 	rtnl_unlock();
954 
955 	return 1;
956 }
957 
958 #endif
959 
960 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)961 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
962 {
963 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
964 
965 #ifdef NET_REFCNT_DEBUG
966 	pr_debug("%s\n", __func__);
967 #endif
968 
969 	in6_dev_put(ifp->idev);
970 
971 	if (cancel_delayed_work(&ifp->dad_work))
972 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
973 			  ifp);
974 
975 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
976 		pr_warn("Freeing alive inet6 address %p\n", ifp);
977 		return;
978 	}
979 
980 	kfree_rcu(ifp, rcu);
981 }
982 
983 static void
ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)984 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
985 {
986 	struct list_head *p;
987 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
988 
989 	/*
990 	 * Each device address list is sorted in order of scope -
991 	 * global before linklocal.
992 	 */
993 	list_for_each(p, &idev->addr_list) {
994 		struct inet6_ifaddr *ifa
995 			= list_entry(p, struct inet6_ifaddr, if_list);
996 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
997 			break;
998 	}
999 
1000 	list_add_tail_rcu(&ifp->if_list, p);
1001 }
1002 
inet6_addr_hash(const struct net *net, const struct in6_addr *addr)1003 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1004 {
1005 	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1006 
1007 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1008 }
1009 
ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr, struct net_device *dev, unsigned int hash)1010 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1011 			       struct net_device *dev, unsigned int hash)
1012 {
1013 	struct inet6_ifaddr *ifp;
1014 
1015 	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1016 		if (!net_eq(dev_net(ifp->idev->dev), net))
1017 			continue;
1018 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1019 			if (!dev || ifp->idev->dev == dev)
1020 				return true;
1021 		}
1022 	}
1023 	return false;
1024 }
1025 
ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)1026 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1027 {
1028 	unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1029 	int err = 0;
1030 
1031 	spin_lock(&addrconf_hash_lock);
1032 
1033 	/* Ignore adding duplicate addresses on an interface */
1034 	if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1035 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1036 		err = -EEXIST;
1037 	} else {
1038 		hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1039 	}
1040 
1041 	spin_unlock(&addrconf_hash_lock);
1042 
1043 	return err;
1044 }
1045 
1046 /* On success it returns ifp with increased reference count */
1047 
1048 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg, bool can_block, struct netlink_ext_ack *extack)1049 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1050 	      bool can_block, struct netlink_ext_ack *extack)
1051 {
1052 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1053 	int addr_type = ipv6_addr_type(cfg->pfx);
1054 	struct net *net = dev_net(idev->dev);
1055 	struct inet6_ifaddr *ifa = NULL;
1056 	struct fib6_info *f6i = NULL;
1057 	int err = 0;
1058 
1059 	if (addr_type == IPV6_ADDR_ANY ||
1060 	    (addr_type & IPV6_ADDR_MULTICAST &&
1061 	     !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1062 	    (!(idev->dev->flags & IFF_LOOPBACK) &&
1063 	     !netif_is_l3_master(idev->dev) &&
1064 	     addr_type & IPV6_ADDR_LOOPBACK))
1065 		return ERR_PTR(-EADDRNOTAVAIL);
1066 
1067 	if (idev->dead) {
1068 		err = -ENODEV;			/*XXX*/
1069 		goto out;
1070 	}
1071 
1072 	if (idev->cnf.disable_ipv6) {
1073 		err = -EACCES;
1074 		goto out;
1075 	}
1076 
1077 	/* validator notifier needs to be blocking;
1078 	 * do not call in atomic context
1079 	 */
1080 	if (can_block) {
1081 		struct in6_validator_info i6vi = {
1082 			.i6vi_addr = *cfg->pfx,
1083 			.i6vi_dev = idev,
1084 			.extack = extack,
1085 		};
1086 
1087 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1088 		err = notifier_to_errno(err);
1089 		if (err < 0)
1090 			goto out;
1091 	}
1092 
1093 	ifa = kzalloc(sizeof(*ifa), gfp_flags);
1094 	if (!ifa) {
1095 		err = -ENOBUFS;
1096 		goto out;
1097 	}
1098 
1099 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1100 	if (IS_ERR(f6i)) {
1101 		err = PTR_ERR(f6i);
1102 		f6i = NULL;
1103 		goto out;
1104 	}
1105 
1106 	neigh_parms_data_state_setall(idev->nd_parms);
1107 
1108 	ifa->addr = *cfg->pfx;
1109 	if (cfg->peer_pfx)
1110 		ifa->peer_addr = *cfg->peer_pfx;
1111 
1112 	spin_lock_init(&ifa->lock);
1113 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1114 	INIT_HLIST_NODE(&ifa->addr_lst);
1115 	ifa->scope = cfg->scope;
1116 	ifa->prefix_len = cfg->plen;
1117 	ifa->rt_priority = cfg->rt_priority;
1118 	ifa->flags = cfg->ifa_flags;
1119 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1120 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1121 		ifa->flags |= IFA_F_TENTATIVE;
1122 	ifa->valid_lft = cfg->valid_lft;
1123 	ifa->prefered_lft = cfg->preferred_lft;
1124 	ifa->cstamp = ifa->tstamp = jiffies;
1125 	ifa->tokenized = false;
1126 
1127 	ifa->rt = f6i;
1128 
1129 	ifa->idev = idev;
1130 	in6_dev_hold(idev);
1131 
1132 	/* For caller */
1133 	refcount_set(&ifa->refcnt, 1);
1134 
1135 	rcu_read_lock_bh();
1136 
1137 	err = ipv6_add_addr_hash(idev->dev, ifa);
1138 	if (err < 0) {
1139 		rcu_read_unlock_bh();
1140 		goto out;
1141 	}
1142 
1143 	write_lock(&idev->lock);
1144 
1145 	/* Add to inet6_dev unicast addr list. */
1146 	ipv6_link_dev_addr(idev, ifa);
1147 
1148 	if (ifa->flags&IFA_F_TEMPORARY) {
1149 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1150 		in6_ifa_hold(ifa);
1151 	}
1152 
1153 	in6_ifa_hold(ifa);
1154 	write_unlock(&idev->lock);
1155 
1156 	rcu_read_unlock_bh();
1157 
1158 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1159 out:
1160 	if (unlikely(err < 0)) {
1161 		fib6_info_release(f6i);
1162 
1163 		if (ifa) {
1164 			if (ifa->idev)
1165 				in6_dev_put(ifa->idev);
1166 			kfree(ifa);
1167 		}
1168 		ifa = ERR_PTR(err);
1169 	}
1170 
1171 	return ifa;
1172 }
1173 
1174 enum cleanup_prefix_rt_t {
1175 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1176 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1177 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1178 };
1179 
1180 /*
1181  * Check, whether the prefix for ifp would still need a prefix route
1182  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1183  * constants.
1184  *
1185  * 1) we don't purge prefix if address was not permanent.
1186  *    prefix is managed by its own lifetime.
1187  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1188  * 3) if there are no addresses, delete prefix.
1189  * 4) if there are still other permanent address(es),
1190  *    corresponding prefix is still permanent.
1191  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1192  *    don't purge the prefix, assume user space is managing it.
1193  * 6) otherwise, update prefix lifetime to the
1194  *    longest valid lifetime among the corresponding
1195  *    addresses on the device.
1196  *    Note: subsequent RA will update lifetime.
1197  **/
1198 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)1199 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1200 {
1201 	struct inet6_ifaddr *ifa;
1202 	struct inet6_dev *idev = ifp->idev;
1203 	unsigned long lifetime;
1204 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1205 
1206 	*expires = jiffies;
1207 
1208 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1209 		if (ifa == ifp)
1210 			continue;
1211 		if (ifa->prefix_len != ifp->prefix_len ||
1212 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1213 				       ifp->prefix_len))
1214 			continue;
1215 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1216 			return CLEANUP_PREFIX_RT_NOP;
1217 
1218 		action = CLEANUP_PREFIX_RT_EXPIRE;
1219 
1220 		spin_lock(&ifa->lock);
1221 
1222 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1223 		/*
1224 		 * Note: Because this address is
1225 		 * not permanent, lifetime <
1226 		 * LONG_MAX / HZ here.
1227 		 */
1228 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1229 			*expires = ifa->tstamp + lifetime * HZ;
1230 		spin_unlock(&ifa->lock);
1231 	}
1232 
1233 	return action;
1234 }
1235 
1236 static void
cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt, bool del_peer)1237 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1238 		     bool del_rt, bool del_peer)
1239 {
1240 	struct fib6_info *f6i;
1241 
1242 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1243 					ifp->prefix_len,
1244 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1245 	if (f6i) {
1246 		if (del_rt)
1247 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1248 		else {
1249 			if (!(f6i->fib6_flags & RTF_EXPIRES))
1250 				fib6_set_expires(f6i, expires);
1251 			fib6_info_release(f6i);
1252 		}
1253 	}
1254 }
1255 
1256 
1257 /* This function wants to get referenced ifp and releases it before return */
1258 
ipv6_del_addr(struct inet6_ifaddr *ifp)1259 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1260 {
1261 	int state;
1262 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1263 	unsigned long expires;
1264 
1265 	ASSERT_RTNL();
1266 
1267 	spin_lock_bh(&ifp->lock);
1268 	state = ifp->state;
1269 	ifp->state = INET6_IFADDR_STATE_DEAD;
1270 	spin_unlock_bh(&ifp->lock);
1271 
1272 	if (state == INET6_IFADDR_STATE_DEAD)
1273 		goto out;
1274 
1275 	spin_lock_bh(&addrconf_hash_lock);
1276 	hlist_del_init_rcu(&ifp->addr_lst);
1277 	spin_unlock_bh(&addrconf_hash_lock);
1278 
1279 	write_lock_bh(&ifp->idev->lock);
1280 
1281 	if (ifp->flags&IFA_F_TEMPORARY) {
1282 		list_del(&ifp->tmp_list);
1283 		if (ifp->ifpub) {
1284 			in6_ifa_put(ifp->ifpub);
1285 			ifp->ifpub = NULL;
1286 		}
1287 		__in6_ifa_put(ifp);
1288 	}
1289 
1290 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1291 		action = check_cleanup_prefix_route(ifp, &expires);
1292 
1293 	list_del_rcu(&ifp->if_list);
1294 	__in6_ifa_put(ifp);
1295 
1296 	write_unlock_bh(&ifp->idev->lock);
1297 
1298 	addrconf_del_dad_work(ifp);
1299 
1300 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1301 
1302 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1303 
1304 	if (action != CLEANUP_PREFIX_RT_NOP) {
1305 		cleanup_prefix_route(ifp, expires,
1306 			action == CLEANUP_PREFIX_RT_DEL, false);
1307 	}
1308 
1309 	/* clean up prefsrc entries */
1310 	rt6_remove_prefsrc(ifp);
1311 out:
1312 	in6_ifa_put(ifp);
1313 }
1314 
ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)1315 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1316 {
1317 	struct inet6_dev *idev = ifp->idev;
1318 	unsigned long tmp_tstamp, age;
1319 	unsigned long regen_advance;
1320 	unsigned long now = jiffies;
1321 	s32 cnf_temp_preferred_lft;
1322 	struct inet6_ifaddr *ift;
1323 	struct ifa6_config cfg;
1324 	long max_desync_factor;
1325 	struct in6_addr addr;
1326 	int ret = 0;
1327 
1328 	write_lock_bh(&idev->lock);
1329 
1330 retry:
1331 	in6_dev_hold(idev);
1332 	if (idev->cnf.use_tempaddr <= 0) {
1333 		write_unlock_bh(&idev->lock);
1334 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1335 		in6_dev_put(idev);
1336 		ret = -1;
1337 		goto out;
1338 	}
1339 	spin_lock_bh(&ifp->lock);
1340 	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1341 		idev->cnf.use_tempaddr = -1;	/*XXX*/
1342 		spin_unlock_bh(&ifp->lock);
1343 		write_unlock_bh(&idev->lock);
1344 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1345 			__func__);
1346 		in6_dev_put(idev);
1347 		ret = -1;
1348 		goto out;
1349 	}
1350 	in6_ifa_hold(ifp);
1351 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1352 	ipv6_gen_rnd_iid(&addr);
1353 
1354 	age = (now - ifp->tstamp) / HZ;
1355 
1356 	regen_advance = idev->cnf.regen_max_retry *
1357 			idev->cnf.dad_transmits *
1358 			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1359 
1360 	/* recalculate max_desync_factor each time and update
1361 	 * idev->desync_factor if it's larger
1362 	 */
1363 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1364 	max_desync_factor = min_t(long,
1365 				  idev->cnf.max_desync_factor,
1366 				  cnf_temp_preferred_lft - regen_advance);
1367 
1368 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1369 		if (max_desync_factor > 0) {
1370 			get_random_bytes(&idev->desync_factor,
1371 					 sizeof(idev->desync_factor));
1372 			idev->desync_factor %= max_desync_factor;
1373 		} else {
1374 			idev->desync_factor = 0;
1375 		}
1376 	}
1377 
1378 	memset(&cfg, 0, sizeof(cfg));
1379 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1380 			      idev->cnf.temp_valid_lft + age);
1381 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1382 	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1383 
1384 	cfg.plen = ifp->prefix_len;
1385 	tmp_tstamp = ifp->tstamp;
1386 	spin_unlock_bh(&ifp->lock);
1387 
1388 	write_unlock_bh(&idev->lock);
1389 
1390 	/* A temporary address is created only if this calculated Preferred
1391 	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1392 	 * an implementation must not create a temporary address with a zero
1393 	 * Preferred Lifetime.
1394 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1395 	 * temporary addresses being generated.
1396 	 */
1397 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1398 	if (cfg.preferred_lft <= regen_advance + age) {
1399 		in6_ifa_put(ifp);
1400 		in6_dev_put(idev);
1401 		ret = -1;
1402 		goto out;
1403 	}
1404 
1405 	cfg.ifa_flags = IFA_F_TEMPORARY;
1406 	/* set in addrconf_prefix_rcv() */
1407 	if (ifp->flags & IFA_F_OPTIMISTIC)
1408 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1409 
1410 	cfg.pfx = &addr;
1411 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1412 
1413 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1414 	if (IS_ERR(ift)) {
1415 		in6_ifa_put(ifp);
1416 		in6_dev_put(idev);
1417 		pr_info("%s: retry temporary address regeneration\n", __func__);
1418 		write_lock_bh(&idev->lock);
1419 		goto retry;
1420 	}
1421 
1422 	spin_lock_bh(&ift->lock);
1423 	ift->ifpub = ifp;
1424 	ift->cstamp = now;
1425 	ift->tstamp = tmp_tstamp;
1426 	spin_unlock_bh(&ift->lock);
1427 
1428 	addrconf_dad_start(ift);
1429 	in6_ifa_put(ift);
1430 	in6_dev_put(idev);
1431 out:
1432 	return ret;
1433 }
1434 
1435 /*
1436  *	Choose an appropriate source address (RFC3484)
1437  */
1438 enum {
1439 	IPV6_SADDR_RULE_INIT = 0,
1440 	IPV6_SADDR_RULE_LOCAL,
1441 	IPV6_SADDR_RULE_SCOPE,
1442 	IPV6_SADDR_RULE_PREFERRED,
1443 #ifdef CONFIG_IPV6_MIP6
1444 	IPV6_SADDR_RULE_HOA,
1445 #endif
1446 	IPV6_SADDR_RULE_OIF,
1447 	IPV6_SADDR_RULE_LABEL,
1448 	IPV6_SADDR_RULE_PRIVACY,
1449 	IPV6_SADDR_RULE_ORCHID,
1450 	IPV6_SADDR_RULE_PREFIX,
1451 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1452 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1453 #endif
1454 	IPV6_SADDR_RULE_MAX
1455 };
1456 
1457 struct ipv6_saddr_score {
1458 	int			rule;
1459 	int			addr_type;
1460 	struct inet6_ifaddr	*ifa;
1461 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1462 	int			scopedist;
1463 	int			matchlen;
1464 };
1465 
1466 struct ipv6_saddr_dst {
1467 	const struct in6_addr *addr;
1468 	int ifindex;
1469 	int scope;
1470 	int label;
1471 	unsigned int prefs;
1472 };
1473 
ipv6_saddr_preferred(int type)1474 static inline int ipv6_saddr_preferred(int type)
1475 {
1476 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1477 		return 1;
1478 	return 0;
1479 }
1480 
ipv6_use_optimistic_addr(struct net *net, struct inet6_dev *idev)1481 static bool ipv6_use_optimistic_addr(struct net *net,
1482 				     struct inet6_dev *idev)
1483 {
1484 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1485 	if (!idev)
1486 		return false;
1487 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1488 		return false;
1489 	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1490 		return false;
1491 
1492 	return true;
1493 #else
1494 	return false;
1495 #endif
1496 }
1497 
ipv6_allow_optimistic_dad(struct net *net, struct inet6_dev *idev)1498 static bool ipv6_allow_optimistic_dad(struct net *net,
1499 				      struct inet6_dev *idev)
1500 {
1501 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1502 	if (!idev)
1503 		return false;
1504 	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1505 		return false;
1506 
1507 	return true;
1508 #else
1509 	return false;
1510 #endif
1511 }
1512 
ipv6_get_saddr_eval(struct net *net, struct ipv6_saddr_score *score, struct ipv6_saddr_dst *dst, int i)1513 static int ipv6_get_saddr_eval(struct net *net,
1514 			       struct ipv6_saddr_score *score,
1515 			       struct ipv6_saddr_dst *dst,
1516 			       int i)
1517 {
1518 	int ret;
1519 
1520 	if (i <= score->rule) {
1521 		switch (i) {
1522 		case IPV6_SADDR_RULE_SCOPE:
1523 			ret = score->scopedist;
1524 			break;
1525 		case IPV6_SADDR_RULE_PREFIX:
1526 			ret = score->matchlen;
1527 			break;
1528 		default:
1529 			ret = !!test_bit(i, score->scorebits);
1530 		}
1531 		goto out;
1532 	}
1533 
1534 	switch (i) {
1535 	case IPV6_SADDR_RULE_INIT:
1536 		/* Rule 0: remember if hiscore is not ready yet */
1537 		ret = !!score->ifa;
1538 		break;
1539 	case IPV6_SADDR_RULE_LOCAL:
1540 		/* Rule 1: Prefer same address */
1541 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1542 		break;
1543 	case IPV6_SADDR_RULE_SCOPE:
1544 		/* Rule 2: Prefer appropriate scope
1545 		 *
1546 		 *      ret
1547 		 *       ^
1548 		 *    -1 |  d 15
1549 		 *    ---+--+-+---> scope
1550 		 *       |
1551 		 *       |             d is scope of the destination.
1552 		 *  B-d  |  \
1553 		 *       |   \      <- smaller scope is better if
1554 		 *  B-15 |    \        if scope is enough for destination.
1555 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1556 		 * d-C-1 | /
1557 		 *       |/         <- greater is better
1558 		 *   -C  /             if scope is not enough for destination.
1559 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1560 		 *
1561 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1562 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1563 		 * Assume B = 0 and we get C > 29.
1564 		 */
1565 		ret = __ipv6_addr_src_scope(score->addr_type);
1566 		if (ret >= dst->scope)
1567 			ret = -ret;
1568 		else
1569 			ret -= 128;	/* 30 is enough */
1570 		score->scopedist = ret;
1571 		break;
1572 	case IPV6_SADDR_RULE_PREFERRED:
1573 	    {
1574 		/* Rule 3: Avoid deprecated and optimistic addresses */
1575 		u8 avoid = IFA_F_DEPRECATED;
1576 
1577 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1578 			avoid |= IFA_F_OPTIMISTIC;
1579 		ret = ipv6_saddr_preferred(score->addr_type) ||
1580 		      !(score->ifa->flags & avoid);
1581 		break;
1582 	    }
1583 #ifdef CONFIG_IPV6_MIP6
1584 	case IPV6_SADDR_RULE_HOA:
1585 	    {
1586 		/* Rule 4: Prefer home address */
1587 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1588 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1589 		break;
1590 	    }
1591 #endif
1592 	case IPV6_SADDR_RULE_OIF:
1593 		/* Rule 5: Prefer outgoing interface */
1594 		ret = (!dst->ifindex ||
1595 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1596 		break;
1597 	case IPV6_SADDR_RULE_LABEL:
1598 		/* Rule 6: Prefer matching label */
1599 		ret = ipv6_addr_label(net,
1600 				      &score->ifa->addr, score->addr_type,
1601 				      score->ifa->idev->dev->ifindex) == dst->label;
1602 		break;
1603 	case IPV6_SADDR_RULE_PRIVACY:
1604 	    {
1605 		/* Rule 7: Prefer public address
1606 		 * Note: prefer temporary address if use_tempaddr >= 2
1607 		 */
1608 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1609 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1610 				score->ifa->idev->cnf.use_tempaddr >= 2;
1611 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1612 		break;
1613 	    }
1614 	case IPV6_SADDR_RULE_ORCHID:
1615 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1616 		 *	    non-ORCHID vs non-ORCHID
1617 		 */
1618 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1619 			ipv6_addr_orchid(dst->addr));
1620 		break;
1621 	case IPV6_SADDR_RULE_PREFIX:
1622 		/* Rule 8: Use longest matching prefix */
1623 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1624 		if (ret > score->ifa->prefix_len)
1625 			ret = score->ifa->prefix_len;
1626 		score->matchlen = ret;
1627 		break;
1628 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1629 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1630 		/* Optimistic addresses still have lower precedence than other
1631 		 * preferred addresses.
1632 		 */
1633 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1634 		break;
1635 #endif
1636 	default:
1637 		ret = 0;
1638 	}
1639 
1640 	if (ret)
1641 		__set_bit(i, score->scorebits);
1642 	score->rule = i;
1643 out:
1644 	return ret;
1645 }
1646 
__ipv6_dev_get_saddr(struct net *net, struct ipv6_saddr_dst *dst, struct inet6_dev *idev, struct ipv6_saddr_score *scores, int hiscore_idx)1647 static int __ipv6_dev_get_saddr(struct net *net,
1648 				struct ipv6_saddr_dst *dst,
1649 				struct inet6_dev *idev,
1650 				struct ipv6_saddr_score *scores,
1651 				int hiscore_idx)
1652 {
1653 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1654 
1655 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1656 		int i;
1657 
1658 		/*
1659 		 * - Tentative Address (RFC2462 section 5.4)
1660 		 *  - A tentative address is not considered
1661 		 *    "assigned to an interface" in the traditional
1662 		 *    sense, unless it is also flagged as optimistic.
1663 		 * - Candidate Source Address (section 4)
1664 		 *  - In any case, anycast addresses, multicast
1665 		 *    addresses, and the unspecified address MUST
1666 		 *    NOT be included in a candidate set.
1667 		 */
1668 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1669 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1670 			continue;
1671 
1672 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1673 
1674 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1675 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1676 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1677 					    idev->dev->name);
1678 			continue;
1679 		}
1680 
1681 		score->rule = -1;
1682 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1683 
1684 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1685 			int minihiscore, miniscore;
1686 
1687 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1688 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1689 
1690 			if (minihiscore > miniscore) {
1691 				if (i == IPV6_SADDR_RULE_SCOPE &&
1692 				    score->scopedist > 0) {
1693 					/*
1694 					 * special case:
1695 					 * each remaining entry
1696 					 * has too small (not enough)
1697 					 * scope, because ifa entries
1698 					 * are sorted by their scope
1699 					 * values.
1700 					 */
1701 					goto out;
1702 				}
1703 				break;
1704 			} else if (minihiscore < miniscore) {
1705 				swap(hiscore, score);
1706 				hiscore_idx = 1 - hiscore_idx;
1707 
1708 				/* restore our iterator */
1709 				score->ifa = hiscore->ifa;
1710 
1711 				break;
1712 			}
1713 		}
1714 	}
1715 out:
1716 	return hiscore_idx;
1717 }
1718 
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)1719 static int ipv6_get_saddr_master(struct net *net,
1720 				 const struct net_device *dst_dev,
1721 				 const struct net_device *master,
1722 				 struct ipv6_saddr_dst *dst,
1723 				 struct ipv6_saddr_score *scores,
1724 				 int hiscore_idx)
1725 {
1726 	struct inet6_dev *idev;
1727 
1728 	idev = __in6_dev_get(dst_dev);
1729 	if (idev)
1730 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1731 						   scores, hiscore_idx);
1732 
1733 	idev = __in6_dev_get(master);
1734 	if (idev)
1735 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1736 						   scores, hiscore_idx);
1737 
1738 	return hiscore_idx;
1739 }
1740 
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)1741 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1742 		       const struct in6_addr *daddr, unsigned int prefs,
1743 		       struct in6_addr *saddr)
1744 {
1745 	struct ipv6_saddr_score scores[2], *hiscore;
1746 	struct ipv6_saddr_dst dst;
1747 	struct inet6_dev *idev;
1748 	struct net_device *dev;
1749 	int dst_type;
1750 	bool use_oif_addr = false;
1751 	int hiscore_idx = 0;
1752 	int ret = 0;
1753 
1754 	dst_type = __ipv6_addr_type(daddr);
1755 	dst.addr = daddr;
1756 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1757 	dst.scope = __ipv6_addr_src_scope(dst_type);
1758 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1759 	dst.prefs = prefs;
1760 
1761 	scores[hiscore_idx].rule = -1;
1762 	scores[hiscore_idx].ifa = NULL;
1763 
1764 	rcu_read_lock();
1765 
1766 	/* Candidate Source Address (section 4)
1767 	 *  - multicast and link-local destination address,
1768 	 *    the set of candidate source address MUST only
1769 	 *    include addresses assigned to interfaces
1770 	 *    belonging to the same link as the outgoing
1771 	 *    interface.
1772 	 * (- For site-local destination addresses, the
1773 	 *    set of candidate source addresses MUST only
1774 	 *    include addresses assigned to interfaces
1775 	 *    belonging to the same site as the outgoing
1776 	 *    interface.)
1777 	 *  - "It is RECOMMENDED that the candidate source addresses
1778 	 *    be the set of unicast addresses assigned to the
1779 	 *    interface that will be used to send to the destination
1780 	 *    (the 'outgoing' interface)." (RFC 6724)
1781 	 */
1782 	if (dst_dev) {
1783 		idev = __in6_dev_get(dst_dev);
1784 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1785 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1786 		    (idev && idev->cnf.use_oif_addrs_only)) {
1787 			use_oif_addr = true;
1788 		}
1789 	}
1790 
1791 	if (use_oif_addr) {
1792 		if (idev)
1793 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1794 	} else {
1795 		const struct net_device *master;
1796 		int master_idx = 0;
1797 
1798 		/* if dst_dev exists and is enslaved to an L3 device, then
1799 		 * prefer addresses from dst_dev and then the master over
1800 		 * any other enslaved devices in the L3 domain.
1801 		 */
1802 		master = l3mdev_master_dev_rcu(dst_dev);
1803 		if (master) {
1804 			master_idx = master->ifindex;
1805 
1806 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1807 							    master, &dst,
1808 							    scores, hiscore_idx);
1809 
1810 			if (scores[hiscore_idx].ifa)
1811 				goto out;
1812 		}
1813 
1814 		for_each_netdev_rcu(net, dev) {
1815 			/* only consider addresses on devices in the
1816 			 * same L3 domain
1817 			 */
1818 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1819 				continue;
1820 			idev = __in6_dev_get(dev);
1821 			if (!idev)
1822 				continue;
1823 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1824 		}
1825 	}
1826 
1827 out:
1828 	hiscore = &scores[hiscore_idx];
1829 	if (!hiscore->ifa)
1830 		ret = -EADDRNOTAVAIL;
1831 	else
1832 		*saddr = hiscore->ifa->addr;
1833 
1834 	rcu_read_unlock();
1835 	return ret;
1836 }
1837 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1838 
__ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr, u32 banned_flags)1839 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1840 		      u32 banned_flags)
1841 {
1842 	struct inet6_ifaddr *ifp;
1843 	int err = -EADDRNOTAVAIL;
1844 
1845 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1846 		if (ifp->scope > IFA_LINK)
1847 			break;
1848 		if (ifp->scope == IFA_LINK &&
1849 		    !(ifp->flags & banned_flags)) {
1850 			*addr = ifp->addr;
1851 			err = 0;
1852 			break;
1853 		}
1854 	}
1855 	return err;
1856 }
1857 
ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr, u32 banned_flags)1858 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1859 		    u32 banned_flags)
1860 {
1861 	struct inet6_dev *idev;
1862 	int err = -EADDRNOTAVAIL;
1863 
1864 	rcu_read_lock();
1865 	idev = __in6_dev_get(dev);
1866 	if (idev) {
1867 		read_lock_bh(&idev->lock);
1868 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1869 		read_unlock_bh(&idev->lock);
1870 	}
1871 	rcu_read_unlock();
1872 	return err;
1873 }
1874 
ipv6_count_addresses(const struct inet6_dev *idev)1875 static int ipv6_count_addresses(const struct inet6_dev *idev)
1876 {
1877 	const struct inet6_ifaddr *ifp;
1878 	int cnt = 0;
1879 
1880 	rcu_read_lock();
1881 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1882 		cnt++;
1883 	rcu_read_unlock();
1884 	return cnt;
1885 }
1886 
ipv6_chk_addr(struct net *net, const struct in6_addr *addr, const struct net_device *dev, int strict)1887 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1888 		  const struct net_device *dev, int strict)
1889 {
1890 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1891 				       strict, IFA_F_TENTATIVE);
1892 }
1893 EXPORT_SYMBOL(ipv6_chk_addr);
1894 
1895 /* device argument is used to find the L3 domain of interest. If
1896  * skip_dev_check is set, then the ifp device is not checked against
1897  * the passed in dev argument. So the 2 cases for addresses checks are:
1898  *   1. does the address exist in the L3 domain that dev is part of
1899  *      (skip_dev_check = true), or
1900  *
1901  *   2. does the address exist on the specific device
1902  *      (skip_dev_check = false)
1903  */
1904 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)1905 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1906 			  const struct net_device *dev, bool skip_dev_check,
1907 			  int strict, u32 banned_flags)
1908 {
1909 	unsigned int hash = inet6_addr_hash(net, addr);
1910 	struct net_device *l3mdev, *ndev;
1911 	struct inet6_ifaddr *ifp;
1912 	u32 ifp_flags;
1913 
1914 	rcu_read_lock();
1915 
1916 	l3mdev = l3mdev_master_dev_rcu(dev);
1917 	if (skip_dev_check)
1918 		dev = NULL;
1919 
1920 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1921 		ndev = ifp->idev->dev;
1922 		if (!net_eq(dev_net(ndev), net))
1923 			continue;
1924 
1925 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1926 			continue;
1927 
1928 		/* Decouple optimistic from tentative for evaluation here.
1929 		 * Ban optimistic addresses explicitly, when required.
1930 		 */
1931 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1932 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1933 			    : ifp->flags;
1934 		if (ipv6_addr_equal(&ifp->addr, addr) &&
1935 		    !(ifp_flags&banned_flags) &&
1936 		    (!dev || ndev == dev ||
1937 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1938 			rcu_read_unlock();
1939 			return ndev;
1940 		}
1941 	}
1942 
1943 	rcu_read_unlock();
1944 	return NULL;
1945 }
1946 
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)1947 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1948 			    const struct net_device *dev, bool skip_dev_check,
1949 			    int strict, u32 banned_flags)
1950 {
1951 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1952 					 strict, banned_flags) ? 1 : 0;
1953 }
1954 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1955 
1956 
1957 /* Compares an address/prefix_len with addresses on device @dev.
1958  * If one is found it returns true.
1959  */
ipv6_chk_custom_prefix(const struct in6_addr *addr, const unsigned int prefix_len, struct net_device *dev)1960 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1961 	const unsigned int prefix_len, struct net_device *dev)
1962 {
1963 	const struct inet6_ifaddr *ifa;
1964 	const struct inet6_dev *idev;
1965 	bool ret = false;
1966 
1967 	rcu_read_lock();
1968 	idev = __in6_dev_get(dev);
1969 	if (idev) {
1970 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1971 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1972 			if (ret)
1973 				break;
1974 		}
1975 	}
1976 	rcu_read_unlock();
1977 
1978 	return ret;
1979 }
1980 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1981 
ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)1982 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1983 {
1984 	const struct inet6_ifaddr *ifa;
1985 	const struct inet6_dev *idev;
1986 	int	onlink;
1987 
1988 	onlink = 0;
1989 	rcu_read_lock();
1990 	idev = __in6_dev_get(dev);
1991 	if (idev) {
1992 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1993 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1994 						   ifa->prefix_len);
1995 			if (onlink)
1996 				break;
1997 		}
1998 	}
1999 	rcu_read_unlock();
2000 	return onlink;
2001 }
2002 EXPORT_SYMBOL(ipv6_chk_prefix);
2003 
2004 /**
2005  * ipv6_dev_find - find the first device with a given source address.
2006  * @net: the net namespace
2007  * @addr: the source address
2008  *
2009  * The caller should be protected by RCU, or RTNL.
2010  */
ipv6_dev_find(struct net *net, const struct in6_addr *addr, struct net_device *dev)2011 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2012 				 struct net_device *dev)
2013 {
2014 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2015 					 IFA_F_TENTATIVE);
2016 }
2017 EXPORT_SYMBOL(ipv6_dev_find);
2018 
ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr, struct net_device *dev, int strict)2019 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2020 				     struct net_device *dev, int strict)
2021 {
2022 	unsigned int hash = inet6_addr_hash(net, addr);
2023 	struct inet6_ifaddr *ifp, *result = NULL;
2024 
2025 	rcu_read_lock();
2026 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2027 		if (!net_eq(dev_net(ifp->idev->dev), net))
2028 			continue;
2029 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2030 			if (!dev || ifp->idev->dev == dev ||
2031 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2032 				if (in6_ifa_hold_safe(ifp)) {
2033 					result = ifp;
2034 					break;
2035 				}
2036 			}
2037 		}
2038 	}
2039 	rcu_read_unlock();
2040 
2041 	return result;
2042 }
2043 
2044 /* Gets referenced address, destroys ifaddr */
2045 
addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)2046 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2047 {
2048 	if (dad_failed)
2049 		ifp->flags |= IFA_F_DADFAILED;
2050 
2051 	if (ifp->flags&IFA_F_TEMPORARY) {
2052 		struct inet6_ifaddr *ifpub;
2053 		spin_lock_bh(&ifp->lock);
2054 		ifpub = ifp->ifpub;
2055 		if (ifpub) {
2056 			in6_ifa_hold(ifpub);
2057 			spin_unlock_bh(&ifp->lock);
2058 			ipv6_create_tempaddr(ifpub, true);
2059 			in6_ifa_put(ifpub);
2060 		} else {
2061 			spin_unlock_bh(&ifp->lock);
2062 		}
2063 		ipv6_del_addr(ifp);
2064 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2065 		spin_lock_bh(&ifp->lock);
2066 		addrconf_del_dad_work(ifp);
2067 		ifp->flags |= IFA_F_TENTATIVE;
2068 		if (dad_failed)
2069 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2070 		spin_unlock_bh(&ifp->lock);
2071 		if (dad_failed)
2072 			ipv6_ifa_notify(0, ifp);
2073 		in6_ifa_put(ifp);
2074 	} else {
2075 		ipv6_del_addr(ifp);
2076 	}
2077 }
2078 
addrconf_dad_end(struct inet6_ifaddr *ifp)2079 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2080 {
2081 	int err = -ENOENT;
2082 
2083 	spin_lock_bh(&ifp->lock);
2084 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2085 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2086 		err = 0;
2087 	}
2088 	spin_unlock_bh(&ifp->lock);
2089 
2090 	return err;
2091 }
2092 
addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)2093 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2094 {
2095 	struct inet6_dev *idev = ifp->idev;
2096 	struct net *net = dev_net(ifp->idev->dev);
2097 
2098 	if (addrconf_dad_end(ifp)) {
2099 		in6_ifa_put(ifp);
2100 		return;
2101 	}
2102 
2103 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2104 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2105 
2106 	spin_lock_bh(&ifp->lock);
2107 
2108 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2109 		struct in6_addr new_addr;
2110 		struct inet6_ifaddr *ifp2;
2111 		int retries = ifp->stable_privacy_retry + 1;
2112 		struct ifa6_config cfg = {
2113 			.pfx = &new_addr,
2114 			.plen = ifp->prefix_len,
2115 			.ifa_flags = ifp->flags,
2116 			.valid_lft = ifp->valid_lft,
2117 			.preferred_lft = ifp->prefered_lft,
2118 			.scope = ifp->scope,
2119 		};
2120 
2121 		if (retries > net->ipv6.sysctl.idgen_retries) {
2122 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2123 					     ifp->idev->dev->name);
2124 			goto errdad;
2125 		}
2126 
2127 		new_addr = ifp->addr;
2128 		if (ipv6_generate_stable_address(&new_addr, retries,
2129 						 idev))
2130 			goto errdad;
2131 
2132 		spin_unlock_bh(&ifp->lock);
2133 
2134 		if (idev->cnf.max_addresses &&
2135 		    ipv6_count_addresses(idev) >=
2136 		    idev->cnf.max_addresses)
2137 			goto lock_errdad;
2138 
2139 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2140 				     ifp->idev->dev->name);
2141 
2142 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2143 		if (IS_ERR(ifp2))
2144 			goto lock_errdad;
2145 
2146 		spin_lock_bh(&ifp2->lock);
2147 		ifp2->stable_privacy_retry = retries;
2148 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2149 		spin_unlock_bh(&ifp2->lock);
2150 
2151 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2152 		in6_ifa_put(ifp2);
2153 lock_errdad:
2154 		spin_lock_bh(&ifp->lock);
2155 	}
2156 
2157 errdad:
2158 	/* transition from _POSTDAD to _ERRDAD */
2159 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2160 	spin_unlock_bh(&ifp->lock);
2161 
2162 	addrconf_mod_dad_work(ifp, 0);
2163 	in6_ifa_put(ifp);
2164 }
2165 
2166 /* Join to solicited addr multicast group.
2167  * caller must hold RTNL */
addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)2168 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2169 {
2170 	struct in6_addr maddr;
2171 
2172 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2173 		return;
2174 
2175 	addrconf_addr_solict_mult(addr, &maddr);
2176 	ipv6_dev_mc_inc(dev, &maddr);
2177 }
2178 
2179 /* caller must hold RTNL */
addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)2180 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2181 {
2182 	struct in6_addr maddr;
2183 
2184 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2185 		return;
2186 
2187 	addrconf_addr_solict_mult(addr, &maddr);
2188 	__ipv6_dev_mc_dec(idev, &maddr);
2189 }
2190 
2191 /* caller must hold RTNL */
addrconf_join_anycast(struct inet6_ifaddr *ifp)2192 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2193 {
2194 	struct in6_addr addr;
2195 
2196 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2197 		return;
2198 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2199 	if (ipv6_addr_any(&addr))
2200 		return;
2201 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2202 }
2203 
2204 /* caller must hold RTNL */
addrconf_leave_anycast(struct inet6_ifaddr *ifp)2205 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2206 {
2207 	struct in6_addr addr;
2208 
2209 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2210 		return;
2211 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2212 	if (ipv6_addr_any(&addr))
2213 		return;
2214 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2215 }
2216 
addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)2217 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2218 {
2219 	switch (dev->addr_len) {
2220 	case ETH_ALEN:
2221 		memcpy(eui, dev->dev_addr, 3);
2222 		eui[3] = 0xFF;
2223 		eui[4] = 0xFE;
2224 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2225 		break;
2226 	case EUI64_ADDR_LEN:
2227 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2228 		eui[0] ^= 2;
2229 		break;
2230 	default:
2231 		return -1;
2232 	}
2233 
2234 	return 0;
2235 }
2236 
addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)2237 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2238 {
2239 	union fwnet_hwaddr *ha;
2240 
2241 	if (dev->addr_len != FWNET_ALEN)
2242 		return -1;
2243 
2244 	ha = (union fwnet_hwaddr *)dev->dev_addr;
2245 
2246 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2247 	eui[0] ^= 2;
2248 	return 0;
2249 }
2250 
addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)2251 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2252 {
2253 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2254 	if (dev->addr_len != ARCNET_ALEN)
2255 		return -1;
2256 	memset(eui, 0, 7);
2257 	eui[7] = *(u8 *)dev->dev_addr;
2258 	return 0;
2259 }
2260 
addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)2261 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2262 {
2263 	if (dev->addr_len != INFINIBAND_ALEN)
2264 		return -1;
2265 	memcpy(eui, dev->dev_addr + 12, 8);
2266 	eui[0] |= 2;
2267 	return 0;
2268 }
2269 
__ipv6_isatap_ifid(u8 *eui, __be32 addr)2270 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2271 {
2272 	if (addr == 0)
2273 		return -1;
2274 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2275 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2276 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2277 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2278 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2279 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2280 	eui[1] = 0;
2281 	eui[2] = 0x5E;
2282 	eui[3] = 0xFE;
2283 	memcpy(eui + 4, &addr, 4);
2284 	return 0;
2285 }
2286 
addrconf_ifid_sit(u8 *eui, struct net_device *dev)2287 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2288 {
2289 	if (dev->priv_flags & IFF_ISATAP)
2290 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2291 	return -1;
2292 }
2293 
addrconf_ifid_gre(u8 *eui, struct net_device *dev)2294 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2295 {
2296 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2297 }
2298 
addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)2299 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2300 {
2301 	memcpy(eui, dev->perm_addr, 3);
2302 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2303 	eui[3] = 0xFF;
2304 	eui[4] = 0xFE;
2305 	eui[0] ^= 2;
2306 	return 0;
2307 }
2308 
ipv6_generate_eui64(u8 *eui, struct net_device *dev)2309 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2310 {
2311 	switch (dev->type) {
2312 	case ARPHRD_ETHER:
2313 	case ARPHRD_FDDI:
2314 		return addrconf_ifid_eui48(eui, dev);
2315 	case ARPHRD_ARCNET:
2316 		return addrconf_ifid_arcnet(eui, dev);
2317 	case ARPHRD_INFINIBAND:
2318 		return addrconf_ifid_infiniband(eui, dev);
2319 	case ARPHRD_SIT:
2320 		return addrconf_ifid_sit(eui, dev);
2321 	case ARPHRD_IPGRE:
2322 	case ARPHRD_TUNNEL:
2323 		return addrconf_ifid_gre(eui, dev);
2324 	case ARPHRD_6LOWPAN:
2325 		return addrconf_ifid_6lowpan(eui, dev);
2326 	case ARPHRD_IEEE1394:
2327 		return addrconf_ifid_ieee1394(eui, dev);
2328 	case ARPHRD_TUNNEL6:
2329 	case ARPHRD_IP6GRE:
2330 	case ARPHRD_RAWIP:
2331 		return addrconf_ifid_ip6tnl(eui, dev);
2332 	}
2333 	return -1;
2334 }
2335 
ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)2336 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2337 {
2338 	int err = -1;
2339 	struct inet6_ifaddr *ifp;
2340 
2341 	read_lock_bh(&idev->lock);
2342 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2343 		if (ifp->scope > IFA_LINK)
2344 			break;
2345 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2346 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2347 			err = 0;
2348 			break;
2349 		}
2350 	}
2351 	read_unlock_bh(&idev->lock);
2352 	return err;
2353 }
2354 
2355 /* Generation of a randomized Interface Identifier
2356  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2357  */
2358 
ipv6_gen_rnd_iid(struct in6_addr *addr)2359 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2360 {
2361 regen:
2362 	get_random_bytes(&addr->s6_addr[8], 8);
2363 
2364 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2365 	 * check if generated address is not inappropriate:
2366 	 *
2367 	 * - Reserved IPv6 Interface Identifers
2368 	 * - XXX: already assigned to an address on the device
2369 	 */
2370 
2371 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2372 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2373 		goto regen;
2374 
2375 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2376 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2377 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2378 	 */
2379 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2380 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2381 		goto regen;
2382 
2383 	/* Reserved subnet anycast addresses */
2384 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2385 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2386 		goto regen;
2387 }
2388 
2389 /*
2390  *	Add prefix route.
2391  */
2392 
2393 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)2394 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2395 		      struct net_device *dev, unsigned long expires,
2396 		      u32 flags, gfp_t gfp_flags)
2397 {
2398 	struct fib6_config cfg = {
2399 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2400 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2401 		.fc_ifindex = dev->ifindex,
2402 		.fc_expires = expires,
2403 		.fc_dst_len = plen,
2404 		.fc_flags = RTF_UP | flags,
2405 		.fc_nlinfo.nl_net = dev_net(dev),
2406 		.fc_protocol = RTPROT_KERNEL,
2407 		.fc_type = RTN_UNICAST,
2408 	};
2409 
2410 	cfg.fc_dst = *pfx;
2411 
2412 	/* Prevent useless cloning on PtP SIT.
2413 	   This thing is done here expecting that the whole
2414 	   class of non-broadcast devices need not cloning.
2415 	 */
2416 #if IS_ENABLED(CONFIG_IPV6_SIT)
2417 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2418 		cfg.fc_flags |= RTF_NONEXTHOP;
2419 #endif
2420 
2421 	ip6_route_add(&cfg, gfp_flags, NULL);
2422 }
2423 
2424 
addrconf_get_prefix_route(const struct in6_addr *pfx, int plen, const struct net_device *dev, u32 flags, u32 noflags, bool no_gw)2425 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2426 						  int plen,
2427 						  const struct net_device *dev,
2428 						  u32 flags, u32 noflags,
2429 						  bool no_gw)
2430 {
2431 	struct fib6_node *fn;
2432 	struct fib6_info *rt = NULL;
2433 	struct fib6_table *table;
2434 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2435 
2436 	table = fib6_get_table(dev_net(dev), tb_id);
2437 	if (!table)
2438 		return NULL;
2439 
2440 	rcu_read_lock();
2441 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2442 	if (!fn)
2443 		goto out;
2444 
2445 	for_each_fib6_node_rt_rcu(fn) {
2446 		/* prefix routes only use builtin fib6_nh */
2447 		if (rt->nh)
2448 			continue;
2449 
2450 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2451 			continue;
2452 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2453 			continue;
2454 		if ((rt->fib6_flags & flags) != flags)
2455 			continue;
2456 		if ((rt->fib6_flags & noflags) != 0)
2457 			continue;
2458 		if (!fib6_info_hold_safe(rt))
2459 			continue;
2460 		break;
2461 	}
2462 out:
2463 	rcu_read_unlock();
2464 	return rt;
2465 }
2466 
2467 
2468 /* Create "default" multicast route to the interface */
2469 
addrconf_add_mroute(struct net_device *dev)2470 static void addrconf_add_mroute(struct net_device *dev)
2471 {
2472 	struct fib6_config cfg = {
2473 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2474 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2475 		.fc_ifindex = dev->ifindex,
2476 		.fc_dst_len = 8,
2477 		.fc_flags = RTF_UP,
2478 		.fc_type = RTN_MULTICAST,
2479 		.fc_nlinfo.nl_net = dev_net(dev),
2480 		.fc_protocol = RTPROT_KERNEL,
2481 	};
2482 
2483 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2484 
2485 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2486 }
2487 
addrconf_add_dev(struct net_device *dev)2488 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2489 {
2490 	struct inet6_dev *idev;
2491 
2492 	ASSERT_RTNL();
2493 
2494 	idev = ipv6_find_idev(dev);
2495 	if (IS_ERR(idev))
2496 		return idev;
2497 
2498 	if (idev->cnf.disable_ipv6)
2499 		return ERR_PTR(-EACCES);
2500 
2501 	/* Add default multicast route */
2502 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2503 		addrconf_add_mroute(dev);
2504 
2505 	return idev;
2506 }
2507 
manage_tempaddrs(struct inet6_dev *idev, struct inet6_ifaddr *ifp, __u32 valid_lft, __u32 prefered_lft, bool create, unsigned long now)2508 static void manage_tempaddrs(struct inet6_dev *idev,
2509 			     struct inet6_ifaddr *ifp,
2510 			     __u32 valid_lft, __u32 prefered_lft,
2511 			     bool create, unsigned long now)
2512 {
2513 	u32 flags;
2514 	struct inet6_ifaddr *ift;
2515 
2516 	read_lock_bh(&idev->lock);
2517 	/* update all temporary addresses in the list */
2518 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2519 		int age, max_valid, max_prefered;
2520 
2521 		if (ifp != ift->ifpub)
2522 			continue;
2523 
2524 		/* RFC 4941 section 3.3:
2525 		 * If a received option will extend the lifetime of a public
2526 		 * address, the lifetimes of temporary addresses should
2527 		 * be extended, subject to the overall constraint that no
2528 		 * temporary addresses should ever remain "valid" or "preferred"
2529 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2530 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2531 		 */
2532 		age = (now - ift->cstamp) / HZ;
2533 		max_valid = idev->cnf.temp_valid_lft - age;
2534 		if (max_valid < 0)
2535 			max_valid = 0;
2536 
2537 		max_prefered = idev->cnf.temp_prefered_lft -
2538 			       idev->desync_factor - age;
2539 		if (max_prefered < 0)
2540 			max_prefered = 0;
2541 
2542 		if (valid_lft > max_valid)
2543 			valid_lft = max_valid;
2544 
2545 		if (prefered_lft > max_prefered)
2546 			prefered_lft = max_prefered;
2547 
2548 		spin_lock(&ift->lock);
2549 		flags = ift->flags;
2550 		ift->valid_lft = valid_lft;
2551 		ift->prefered_lft = prefered_lft;
2552 		ift->tstamp = now;
2553 		if (prefered_lft > 0)
2554 			ift->flags &= ~IFA_F_DEPRECATED;
2555 
2556 		spin_unlock(&ift->lock);
2557 		if (!(flags&IFA_F_TENTATIVE))
2558 			ipv6_ifa_notify(0, ift);
2559 	}
2560 
2561 	/* Also create a temporary address if it's enabled but no temporary
2562 	 * address currently exists.
2563 	 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2564 	 * as part of cleanup (ie. deleting the mngtmpaddr).
2565 	 * We don't want that to result in creating a new temporary ip address.
2566 	 */
2567 	if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2568 		create = true;
2569 
2570 	if (create && idev->cnf.use_tempaddr > 0) {
2571 		/* When a new public address is created as described
2572 		 * in [ADDRCONF], also create a new temporary address.
2573 		 */
2574 		read_unlock_bh(&idev->lock);
2575 		ipv6_create_tempaddr(ifp, false);
2576 	} else {
2577 		read_unlock_bh(&idev->lock);
2578 	}
2579 }
2580 
is_addr_mode_generate_stable(struct inet6_dev *idev)2581 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2582 {
2583 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2584 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2585 }
2586 
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)2587 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2588 				 const struct prefix_info *pinfo,
2589 				 struct inet6_dev *in6_dev,
2590 				 const struct in6_addr *addr, int addr_type,
2591 				 u32 addr_flags, bool sllao, bool tokenized,
2592 				 __u32 valid_lft, u32 prefered_lft)
2593 {
2594 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2595 	int create = 0, update_lft = 0;
2596 
2597 	if (!ifp && valid_lft) {
2598 		int max_addresses = in6_dev->cnf.max_addresses;
2599 		struct ifa6_config cfg = {
2600 			.pfx = addr,
2601 			.plen = pinfo->prefix_len,
2602 			.ifa_flags = addr_flags,
2603 			.valid_lft = valid_lft,
2604 			.preferred_lft = prefered_lft,
2605 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2606 		};
2607 
2608 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2609 		if ((net->ipv6.devconf_all->optimistic_dad ||
2610 		     in6_dev->cnf.optimistic_dad) &&
2611 		    !net->ipv6.devconf_all->forwarding && sllao)
2612 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2613 #endif
2614 
2615 		/* Do not allow to create too much of autoconfigured
2616 		 * addresses; this would be too easy way to crash kernel.
2617 		 */
2618 		if (!max_addresses ||
2619 		    ipv6_count_addresses(in6_dev) < max_addresses)
2620 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2621 
2622 		if (IS_ERR_OR_NULL(ifp))
2623 			return -1;
2624 
2625 		create = 1;
2626 		spin_lock_bh(&ifp->lock);
2627 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2628 		ifp->cstamp = jiffies;
2629 		ifp->tokenized = tokenized;
2630 		spin_unlock_bh(&ifp->lock);
2631 		addrconf_dad_start(ifp);
2632 	}
2633 
2634 	if (ifp) {
2635 		u32 flags;
2636 		unsigned long now;
2637 		u32 stored_lft;
2638 
2639 		/* update lifetime (RFC2462 5.5.3 e) */
2640 		spin_lock_bh(&ifp->lock);
2641 		now = jiffies;
2642 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2643 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2644 		else
2645 			stored_lft = 0;
2646 		if (!create && stored_lft) {
2647 			const u32 minimum_lft = min_t(u32,
2648 				stored_lft, MIN_VALID_LIFETIME);
2649 			valid_lft = max(valid_lft, minimum_lft);
2650 
2651 			/* RFC4862 Section 5.5.3e:
2652 			 * "Note that the preferred lifetime of the
2653 			 *  corresponding address is always reset to
2654 			 *  the Preferred Lifetime in the received
2655 			 *  Prefix Information option, regardless of
2656 			 *  whether the valid lifetime is also reset or
2657 			 *  ignored."
2658 			 *
2659 			 * So we should always update prefered_lft here.
2660 			 */
2661 			update_lft = 1;
2662 		}
2663 
2664 		if (update_lft) {
2665 			ifp->valid_lft = valid_lft;
2666 			ifp->prefered_lft = prefered_lft;
2667 			ifp->tstamp = now;
2668 			flags = ifp->flags;
2669 			ifp->flags &= ~IFA_F_DEPRECATED;
2670 			spin_unlock_bh(&ifp->lock);
2671 
2672 			if (!(flags&IFA_F_TENTATIVE))
2673 				ipv6_ifa_notify(0, ifp);
2674 		} else
2675 			spin_unlock_bh(&ifp->lock);
2676 
2677 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2678 				 create, now);
2679 
2680 		in6_ifa_put(ifp);
2681 		addrconf_verify();
2682 	}
2683 
2684 	return 0;
2685 }
2686 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2687 
addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)2688 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2689 {
2690 	struct prefix_info *pinfo;
2691 	__u32 valid_lft;
2692 	__u32 prefered_lft;
2693 	int addr_type, err;
2694 	u32 addr_flags = 0;
2695 	struct inet6_dev *in6_dev;
2696 	struct net *net = dev_net(dev);
2697 
2698 	pinfo = (struct prefix_info *) opt;
2699 
2700 	if (len < sizeof(struct prefix_info)) {
2701 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2702 		return;
2703 	}
2704 
2705 	/*
2706 	 *	Validation checks ([ADDRCONF], page 19)
2707 	 */
2708 
2709 	addr_type = ipv6_addr_type(&pinfo->prefix);
2710 
2711 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2712 		return;
2713 
2714 	valid_lft = ntohl(pinfo->valid);
2715 	prefered_lft = ntohl(pinfo->prefered);
2716 
2717 	if (prefered_lft > valid_lft) {
2718 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2719 		return;
2720 	}
2721 
2722 	in6_dev = in6_dev_get(dev);
2723 
2724 	if (!in6_dev) {
2725 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2726 				    dev->name);
2727 		return;
2728 	}
2729 
2730 	if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2731 		goto put;
2732 
2733 	/*
2734 	 *	Two things going on here:
2735 	 *	1) Add routes for on-link prefixes
2736 	 *	2) Configure prefixes with the auto flag set
2737 	 */
2738 
2739 	if (pinfo->onlink) {
2740 		struct fib6_info *rt;
2741 		unsigned long rt_expires;
2742 
2743 		/* Avoid arithmetic overflow. Really, we could
2744 		 * save rt_expires in seconds, likely valid_lft,
2745 		 * but it would require division in fib gc, that it
2746 		 * not good.
2747 		 */
2748 		if (HZ > USER_HZ)
2749 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2750 		else
2751 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2752 
2753 		if (addrconf_finite_timeout(rt_expires))
2754 			rt_expires *= HZ;
2755 
2756 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2757 					       pinfo->prefix_len,
2758 					       dev,
2759 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2760 					       RTF_DEFAULT, true);
2761 
2762 		if (rt) {
2763 			/* Autoconf prefix route */
2764 			if (valid_lft == 0) {
2765 				ip6_del_rt(net, rt, false);
2766 				rt = NULL;
2767 			} else if (addrconf_finite_timeout(rt_expires)) {
2768 				/* not infinity */
2769 				fib6_set_expires(rt, jiffies + rt_expires);
2770 			} else {
2771 				fib6_clean_expires(rt);
2772 			}
2773 		} else if (valid_lft) {
2774 			clock_t expires = 0;
2775 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2776 			if (addrconf_finite_timeout(rt_expires)) {
2777 				/* not infinity */
2778 				flags |= RTF_EXPIRES;
2779 				expires = jiffies_to_clock_t(rt_expires);
2780 			}
2781 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2782 					      0, dev, expires, flags,
2783 					      GFP_ATOMIC);
2784 		}
2785 		fib6_info_release(rt);
2786 	}
2787 
2788 	/* Try to figure out our local address for this prefix */
2789 
2790 	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2791 		struct in6_addr addr;
2792 		bool tokenized = false, dev_addr_generated = false;
2793 
2794 		if (pinfo->prefix_len == 64) {
2795 			memcpy(&addr, &pinfo->prefix, 8);
2796 
2797 			if (!ipv6_addr_any(&in6_dev->token)) {
2798 				read_lock_bh(&in6_dev->lock);
2799 				memcpy(addr.s6_addr + 8,
2800 				       in6_dev->token.s6_addr + 8, 8);
2801 				read_unlock_bh(&in6_dev->lock);
2802 				tokenized = true;
2803 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2804 				   !ipv6_generate_stable_address(&addr, 0,
2805 								 in6_dev)) {
2806 				addr_flags |= IFA_F_STABLE_PRIVACY;
2807 				goto ok;
2808 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2809 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2810 				goto put;
2811 			} else {
2812 				dev_addr_generated = true;
2813 			}
2814 			goto ok;
2815 		}
2816 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2817 				    pinfo->prefix_len);
2818 		goto put;
2819 
2820 ok:
2821 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2822 						   &addr, addr_type,
2823 						   addr_flags, sllao,
2824 						   tokenized, valid_lft,
2825 						   prefered_lft);
2826 		if (err)
2827 			goto put;
2828 
2829 		/* Ignore error case here because previous prefix add addr was
2830 		 * successful which will be notified.
2831 		 */
2832 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2833 					      addr_type, addr_flags, sllao,
2834 					      tokenized, valid_lft,
2835 					      prefered_lft,
2836 					      dev_addr_generated);
2837 	}
2838 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2839 put:
2840 	in6_dev_put(in6_dev);
2841 }
2842 
addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev, struct in6_ifreq *ireq)2843 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2844 		struct in6_ifreq *ireq)
2845 {
2846 	struct ip_tunnel_parm p = { };
2847 	int err;
2848 
2849 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2850 		return -EADDRNOTAVAIL;
2851 
2852 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2853 	p.iph.version = 4;
2854 	p.iph.ihl = 5;
2855 	p.iph.protocol = IPPROTO_IPV6;
2856 	p.iph.ttl = 64;
2857 
2858 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2859 		return -EOPNOTSUPP;
2860 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2861 	if (err)
2862 		return err;
2863 
2864 	dev = __dev_get_by_name(net, p.name);
2865 	if (!dev)
2866 		return -ENOBUFS;
2867 	return dev_open(dev, NULL);
2868 }
2869 
2870 /*
2871  *	Set destination address.
2872  *	Special case for SIT interfaces where we create a new "virtual"
2873  *	device.
2874  */
addrconf_set_dstaddr(struct net *net, void __user *arg)2875 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2876 {
2877 	struct net_device *dev;
2878 	struct in6_ifreq ireq;
2879 	int err = -ENODEV;
2880 
2881 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2882 		return -ENODEV;
2883 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2884 		return -EFAULT;
2885 
2886 	rtnl_lock();
2887 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2888 	if (dev && dev->type == ARPHRD_SIT)
2889 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2890 	rtnl_unlock();
2891 	return err;
2892 }
2893 
ipv6_mc_config(struct sock *sk, bool join, const struct in6_addr *addr, int ifindex)2894 static int ipv6_mc_config(struct sock *sk, bool join,
2895 			  const struct in6_addr *addr, int ifindex)
2896 {
2897 	int ret;
2898 
2899 	ASSERT_RTNL();
2900 
2901 	lock_sock(sk);
2902 	if (join)
2903 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2904 	else
2905 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2906 	release_sock(sk);
2907 
2908 	return ret;
2909 }
2910 
2911 /*
2912  *	Manual configuration of address on an interface
2913  */
inet6_addr_add(struct net *net, int ifindex, struct ifa6_config *cfg, struct netlink_ext_ack *extack)2914 static int inet6_addr_add(struct net *net, int ifindex,
2915 			  struct ifa6_config *cfg,
2916 			  struct netlink_ext_ack *extack)
2917 {
2918 	struct inet6_ifaddr *ifp;
2919 	struct inet6_dev *idev;
2920 	struct net_device *dev;
2921 	unsigned long timeout;
2922 	clock_t expires;
2923 	u32 flags;
2924 
2925 	ASSERT_RTNL();
2926 
2927 	if (cfg->plen > 128)
2928 		return -EINVAL;
2929 
2930 	/* check the lifetime */
2931 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2932 		return -EINVAL;
2933 
2934 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2935 		return -EINVAL;
2936 
2937 	dev = __dev_get_by_index(net, ifindex);
2938 	if (!dev)
2939 		return -ENODEV;
2940 
2941 	idev = addrconf_add_dev(dev);
2942 	if (IS_ERR(idev))
2943 		return PTR_ERR(idev);
2944 
2945 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2946 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2947 					 true, cfg->pfx, ifindex);
2948 
2949 		if (ret < 0)
2950 			return ret;
2951 	}
2952 
2953 	cfg->scope = ipv6_addr_scope(cfg->pfx);
2954 
2955 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2956 	if (addrconf_finite_timeout(timeout)) {
2957 		expires = jiffies_to_clock_t(timeout * HZ);
2958 		cfg->valid_lft = timeout;
2959 		flags = RTF_EXPIRES;
2960 	} else {
2961 		expires = 0;
2962 		flags = 0;
2963 		cfg->ifa_flags |= IFA_F_PERMANENT;
2964 	}
2965 
2966 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2967 	if (addrconf_finite_timeout(timeout)) {
2968 		if (timeout == 0)
2969 			cfg->ifa_flags |= IFA_F_DEPRECATED;
2970 		cfg->preferred_lft = timeout;
2971 	}
2972 
2973 	ifp = ipv6_add_addr(idev, cfg, true, extack);
2974 	if (!IS_ERR(ifp)) {
2975 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2976 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2977 					      ifp->rt_priority, dev, expires,
2978 					      flags, GFP_KERNEL);
2979 		}
2980 
2981 		/* Send a netlink notification if DAD is enabled and
2982 		 * optimistic flag is not set
2983 		 */
2984 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2985 			ipv6_ifa_notify(0, ifp);
2986 		/*
2987 		 * Note that section 3.1 of RFC 4429 indicates
2988 		 * that the Optimistic flag should not be set for
2989 		 * manually configured addresses
2990 		 */
2991 		addrconf_dad_start(ifp);
2992 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2993 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
2994 					 cfg->preferred_lft, true, jiffies);
2995 		in6_ifa_put(ifp);
2996 		addrconf_verify_rtnl();
2997 		return 0;
2998 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2999 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3000 			       cfg->pfx, ifindex);
3001 	}
3002 
3003 	return PTR_ERR(ifp);
3004 }
3005 
inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags, const struct in6_addr *pfx, unsigned int plen)3006 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3007 			  const struct in6_addr *pfx, unsigned int plen)
3008 {
3009 	struct inet6_ifaddr *ifp;
3010 	struct inet6_dev *idev;
3011 	struct net_device *dev;
3012 
3013 	if (plen > 128)
3014 		return -EINVAL;
3015 
3016 	dev = __dev_get_by_index(net, ifindex);
3017 	if (!dev)
3018 		return -ENODEV;
3019 
3020 	idev = __in6_dev_get(dev);
3021 	if (!idev)
3022 		return -ENXIO;
3023 
3024 	read_lock_bh(&idev->lock);
3025 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3026 		if (ifp->prefix_len == plen &&
3027 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3028 			in6_ifa_hold(ifp);
3029 			read_unlock_bh(&idev->lock);
3030 
3031 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3032 			    (ifa_flags & IFA_F_MANAGETEMPADDR))
3033 				manage_tempaddrs(idev, ifp, 0, 0, false,
3034 						 jiffies);
3035 			ipv6_del_addr(ifp);
3036 			addrconf_verify_rtnl();
3037 			if (ipv6_addr_is_multicast(pfx)) {
3038 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3039 					       false, pfx, dev->ifindex);
3040 			}
3041 			return 0;
3042 		}
3043 	}
3044 	read_unlock_bh(&idev->lock);
3045 	return -EADDRNOTAVAIL;
3046 }
3047 
3048 
addrconf_add_ifaddr(struct net *net, void __user *arg)3049 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3050 {
3051 	struct ifa6_config cfg = {
3052 		.ifa_flags = IFA_F_PERMANENT,
3053 		.preferred_lft = INFINITY_LIFE_TIME,
3054 		.valid_lft = INFINITY_LIFE_TIME,
3055 	};
3056 	struct in6_ifreq ireq;
3057 	int err;
3058 
3059 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3060 		return -EPERM;
3061 
3062 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3063 		return -EFAULT;
3064 
3065 	cfg.pfx = &ireq.ifr6_addr;
3066 	cfg.plen = ireq.ifr6_prefixlen;
3067 
3068 	rtnl_lock();
3069 	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3070 	rtnl_unlock();
3071 	return err;
3072 }
3073 
addrconf_del_ifaddr(struct net *net, void __user *arg)3074 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3075 {
3076 	struct in6_ifreq ireq;
3077 	int err;
3078 
3079 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3080 		return -EPERM;
3081 
3082 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3083 		return -EFAULT;
3084 
3085 	rtnl_lock();
3086 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3087 			     ireq.ifr6_prefixlen);
3088 	rtnl_unlock();
3089 	return err;
3090 }
3091 
add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int plen, int scope)3092 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3093 		     int plen, int scope)
3094 {
3095 	struct inet6_ifaddr *ifp;
3096 	struct ifa6_config cfg = {
3097 		.pfx = addr,
3098 		.plen = plen,
3099 		.ifa_flags = IFA_F_PERMANENT,
3100 		.valid_lft = INFINITY_LIFE_TIME,
3101 		.preferred_lft = INFINITY_LIFE_TIME,
3102 		.scope = scope
3103 	};
3104 
3105 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3106 	if (!IS_ERR(ifp)) {
3107 		spin_lock_bh(&ifp->lock);
3108 		ifp->flags &= ~IFA_F_TENTATIVE;
3109 		spin_unlock_bh(&ifp->lock);
3110 		rt_genid_bump_ipv6(dev_net(idev->dev));
3111 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3112 		in6_ifa_put(ifp);
3113 	}
3114 }
3115 
3116 #if IS_ENABLED(CONFIG_IPV6_SIT)
sit_add_v4_addrs(struct inet6_dev *idev)3117 static void sit_add_v4_addrs(struct inet6_dev *idev)
3118 {
3119 	struct in6_addr addr;
3120 	struct net_device *dev;
3121 	struct net *net = dev_net(idev->dev);
3122 	int scope, plen;
3123 	u32 pflags = 0;
3124 
3125 	ASSERT_RTNL();
3126 
3127 	memset(&addr, 0, sizeof(struct in6_addr));
3128 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3129 
3130 	if (idev->dev->flags&IFF_POINTOPOINT) {
3131 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3132 			return;
3133 
3134 		addr.s6_addr32[0] = htonl(0xfe800000);
3135 		scope = IFA_LINK;
3136 		plen = 64;
3137 	} else {
3138 		scope = IPV6_ADDR_COMPATv4;
3139 		plen = 96;
3140 		pflags |= RTF_NONEXTHOP;
3141 	}
3142 
3143 	if (addr.s6_addr32[3]) {
3144 		add_addr(idev, &addr, plen, scope);
3145 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3146 				      GFP_KERNEL);
3147 		return;
3148 	}
3149 
3150 	for_each_netdev(net, dev) {
3151 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3152 		if (in_dev && (dev->flags & IFF_UP)) {
3153 			struct in_ifaddr *ifa;
3154 			int flag = scope;
3155 
3156 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3157 				addr.s6_addr32[3] = ifa->ifa_local;
3158 
3159 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3160 					continue;
3161 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3162 					if (idev->dev->flags&IFF_POINTOPOINT)
3163 						continue;
3164 					flag |= IFA_HOST;
3165 				}
3166 
3167 				add_addr(idev, &addr, plen, flag);
3168 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3169 						      0, pflags, GFP_KERNEL);
3170 			}
3171 		}
3172 	}
3173 }
3174 #endif
3175 
init_loopback(struct net_device *dev)3176 static void init_loopback(struct net_device *dev)
3177 {
3178 	struct inet6_dev  *idev;
3179 
3180 	/* ::1 */
3181 
3182 	ASSERT_RTNL();
3183 
3184 	idev = ipv6_find_idev(dev);
3185 	if (IS_ERR(idev)) {
3186 		pr_debug("%s: add_dev failed\n", __func__);
3187 		return;
3188 	}
3189 
3190 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3191 }
3192 
addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr, u32 flags)3193 void addrconf_add_linklocal(struct inet6_dev *idev,
3194 			    const struct in6_addr *addr, u32 flags)
3195 {
3196 	struct ifa6_config cfg = {
3197 		.pfx = addr,
3198 		.plen = 64,
3199 		.ifa_flags = flags | IFA_F_PERMANENT,
3200 		.valid_lft = INFINITY_LIFE_TIME,
3201 		.preferred_lft = INFINITY_LIFE_TIME,
3202 		.scope = IFA_LINK
3203 	};
3204 	struct inet6_ifaddr *ifp;
3205 
3206 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3207 	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3208 	     idev->cnf.optimistic_dad) &&
3209 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3210 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3211 #endif
3212 
3213 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3214 	if (!IS_ERR(ifp)) {
3215 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3216 				      0, 0, GFP_ATOMIC);
3217 		addrconf_dad_start(ifp);
3218 		in6_ifa_put(ifp);
3219 	}
3220 }
3221 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3222 
ipv6_reserved_interfaceid(struct in6_addr address)3223 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3224 {
3225 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3226 		return true;
3227 
3228 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3229 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3230 		return true;
3231 
3232 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3233 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3234 		return true;
3235 
3236 	return false;
3237 }
3238 
ipv6_generate_stable_address(struct in6_addr *address, u8 dad_count, const struct inet6_dev *idev)3239 static int ipv6_generate_stable_address(struct in6_addr *address,
3240 					u8 dad_count,
3241 					const struct inet6_dev *idev)
3242 {
3243 	static DEFINE_SPINLOCK(lock);
3244 	static __u32 digest[SHA1_DIGEST_WORDS];
3245 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3246 
3247 	static union {
3248 		char __data[SHA1_BLOCK_SIZE];
3249 		struct {
3250 			struct in6_addr secret;
3251 			__be32 prefix[2];
3252 			unsigned char hwaddr[MAX_ADDR_LEN];
3253 			u8 dad_count;
3254 		} __packed;
3255 	} data;
3256 
3257 	struct in6_addr secret;
3258 	struct in6_addr temp;
3259 	struct net *net = dev_net(idev->dev);
3260 
3261 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3262 
3263 	if (idev->cnf.stable_secret.initialized)
3264 		secret = idev->cnf.stable_secret.secret;
3265 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3266 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3267 	else
3268 		return -1;
3269 
3270 retry:
3271 	spin_lock_bh(&lock);
3272 
3273 	sha1_init(digest);
3274 	memset(&data, 0, sizeof(data));
3275 	memset(workspace, 0, sizeof(workspace));
3276 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3277 	data.prefix[0] = address->s6_addr32[0];
3278 	data.prefix[1] = address->s6_addr32[1];
3279 	data.secret = secret;
3280 	data.dad_count = dad_count;
3281 
3282 	sha1_transform(digest, data.__data, workspace);
3283 
3284 	temp = *address;
3285 	temp.s6_addr32[2] = (__force __be32)digest[0];
3286 	temp.s6_addr32[3] = (__force __be32)digest[1];
3287 
3288 	spin_unlock_bh(&lock);
3289 
3290 	if (ipv6_reserved_interfaceid(temp)) {
3291 		dad_count++;
3292 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3293 			return -1;
3294 		goto retry;
3295 	}
3296 
3297 	*address = temp;
3298 	return 0;
3299 }
3300 
ipv6_gen_mode_random_init(struct inet6_dev *idev)3301 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3302 {
3303 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3304 
3305 	if (s->initialized)
3306 		return;
3307 	s = &idev->cnf.stable_secret;
3308 	get_random_bytes(&s->secret, sizeof(s->secret));
3309 	s->initialized = true;
3310 }
3311 
addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)3312 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3313 {
3314 	struct in6_addr addr;
3315 
3316 	/* no link local addresses on L3 master devices */
3317 	if (netif_is_l3_master(idev->dev))
3318 		return;
3319 
3320 	/* no link local addresses on devices flagged as slaves */
3321 	if (idev->dev->flags & IFF_SLAVE)
3322 		return;
3323 
3324 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3325 
3326 	switch (idev->cnf.addr_gen_mode) {
3327 	case IN6_ADDR_GEN_MODE_RANDOM:
3328 		ipv6_gen_mode_random_init(idev);
3329 		fallthrough;
3330 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3331 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3332 			addrconf_add_linklocal(idev, &addr,
3333 					       IFA_F_STABLE_PRIVACY);
3334 		else if (prefix_route)
3335 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3336 					      0, 0, GFP_KERNEL);
3337 		break;
3338 	case IN6_ADDR_GEN_MODE_EUI64:
3339 		/* addrconf_add_linklocal also adds a prefix_route and we
3340 		 * only need to care about prefix routes if ipv6_generate_eui64
3341 		 * couldn't generate one.
3342 		 */
3343 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3344 			addrconf_add_linklocal(idev, &addr, 0);
3345 		else if (prefix_route)
3346 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3347 					      0, 0, GFP_KERNEL);
3348 		break;
3349 	case IN6_ADDR_GEN_MODE_NONE:
3350 	default:
3351 		/* will not add any link local address */
3352 		break;
3353 	}
3354 }
3355 
addrconf_dev_config(struct net_device *dev)3356 static void addrconf_dev_config(struct net_device *dev)
3357 {
3358 	struct inet6_dev *idev;
3359 
3360 	ASSERT_RTNL();
3361 
3362 	if ((dev->type != ARPHRD_ETHER) &&
3363 	    (dev->type != ARPHRD_FDDI) &&
3364 	    (dev->type != ARPHRD_ARCNET) &&
3365 	    (dev->type != ARPHRD_INFINIBAND) &&
3366 	    (dev->type != ARPHRD_IEEE1394) &&
3367 	    (dev->type != ARPHRD_TUNNEL6) &&
3368 	    (dev->type != ARPHRD_6LOWPAN) &&
3369 	    (dev->type != ARPHRD_IP6GRE) &&
3370 	    (dev->type != ARPHRD_IPGRE) &&
3371 	    (dev->type != ARPHRD_TUNNEL) &&
3372 	    (dev->type != ARPHRD_NONE) &&
3373 	    (dev->type != ARPHRD_RAWIP)) {
3374 		/* Alas, we support only Ethernet autoconfiguration. */
3375 		idev = __in6_dev_get(dev);
3376 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3377 		    dev->flags & IFF_MULTICAST)
3378 			ipv6_mc_up(idev);
3379 		return;
3380 	}
3381 
3382 	idev = addrconf_add_dev(dev);
3383 	if (IS_ERR(idev))
3384 		return;
3385 
3386 	/* this device type has no EUI support */
3387 	if (dev->type == ARPHRD_NONE &&
3388 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3389 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3390 
3391 	addrconf_addr_gen(idev, false);
3392 }
3393 
3394 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device *dev)3395 static void addrconf_sit_config(struct net_device *dev)
3396 {
3397 	struct inet6_dev *idev;
3398 
3399 	ASSERT_RTNL();
3400 
3401 	/*
3402 	 * Configure the tunnel with one of our IPv4
3403 	 * addresses... we should configure all of
3404 	 * our v4 addrs in the tunnel
3405 	 */
3406 
3407 	idev = ipv6_find_idev(dev);
3408 	if (IS_ERR(idev)) {
3409 		pr_debug("%s: add_dev failed\n", __func__);
3410 		return;
3411 	}
3412 
3413 	if (dev->priv_flags & IFF_ISATAP) {
3414 		addrconf_addr_gen(idev, false);
3415 		return;
3416 	}
3417 
3418 	sit_add_v4_addrs(idev);
3419 
3420 	if (dev->flags&IFF_POINTOPOINT)
3421 		addrconf_add_mroute(dev);
3422 }
3423 #endif
3424 
3425 #if IS_ENABLED(CONFIG_NET_IPGRE)
addrconf_gre_config(struct net_device *dev)3426 static void addrconf_gre_config(struct net_device *dev)
3427 {
3428 	struct inet6_dev *idev;
3429 
3430 	ASSERT_RTNL();
3431 
3432 	idev = ipv6_find_idev(dev);
3433 	if (IS_ERR(idev)) {
3434 		pr_debug("%s: add_dev failed\n", __func__);
3435 		return;
3436 	}
3437 
3438 	addrconf_addr_gen(idev, true);
3439 	if (dev->flags & IFF_POINTOPOINT)
3440 		addrconf_add_mroute(dev);
3441 }
3442 #endif
3443 
fixup_permanent_addr(struct net *net, struct inet6_dev *idev, struct inet6_ifaddr *ifp)3444 static int fixup_permanent_addr(struct net *net,
3445 				struct inet6_dev *idev,
3446 				struct inet6_ifaddr *ifp)
3447 {
3448 	/* !fib6_node means the host route was removed from the
3449 	 * FIB, for example, if 'lo' device is taken down. In that
3450 	 * case regenerate the host route.
3451 	 */
3452 	if (!ifp->rt || !ifp->rt->fib6_node) {
3453 		struct fib6_info *f6i, *prev;
3454 
3455 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3456 					 GFP_ATOMIC);
3457 		if (IS_ERR(f6i))
3458 			return PTR_ERR(f6i);
3459 
3460 		/* ifp->rt can be accessed outside of rtnl */
3461 		spin_lock(&ifp->lock);
3462 		prev = ifp->rt;
3463 		ifp->rt = f6i;
3464 		spin_unlock(&ifp->lock);
3465 
3466 		fib6_info_release(prev);
3467 	}
3468 
3469 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3470 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3471 				      ifp->rt_priority, idev->dev, 0, 0,
3472 				      GFP_ATOMIC);
3473 	}
3474 
3475 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3476 		addrconf_dad_start(ifp);
3477 
3478 	return 0;
3479 }
3480 
addrconf_permanent_addr(struct net *net, struct net_device *dev)3481 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3482 {
3483 	struct inet6_ifaddr *ifp, *tmp;
3484 	struct inet6_dev *idev;
3485 
3486 	idev = __in6_dev_get(dev);
3487 	if (!idev)
3488 		return;
3489 
3490 	write_lock_bh(&idev->lock);
3491 
3492 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3493 		if ((ifp->flags & IFA_F_PERMANENT) &&
3494 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3495 			write_unlock_bh(&idev->lock);
3496 			in6_ifa_hold(ifp);
3497 			ipv6_del_addr(ifp);
3498 			write_lock_bh(&idev->lock);
3499 
3500 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3501 					     idev->dev->name, &ifp->addr);
3502 		}
3503 	}
3504 
3505 	write_unlock_bh(&idev->lock);
3506 }
3507 
addrconf_notify(struct notifier_block *this, unsigned long event, void *ptr)3508 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3509 			   void *ptr)
3510 {
3511 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3512 	struct netdev_notifier_change_info *change_info;
3513 	struct netdev_notifier_changeupper_info *info;
3514 	struct inet6_dev *idev = __in6_dev_get(dev);
3515 	struct net *net = dev_net(dev);
3516 	int run_pending = 0;
3517 	int err;
3518 
3519 	switch (event) {
3520 	case NETDEV_REGISTER:
3521 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3522 			idev = ipv6_add_dev(dev);
3523 			if (IS_ERR(idev))
3524 				return notifier_from_errno(PTR_ERR(idev));
3525 		}
3526 		break;
3527 
3528 	case NETDEV_CHANGEMTU:
3529 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3530 		if (dev->mtu < IPV6_MIN_MTU) {
3531 			addrconf_ifdown(dev, dev != net->loopback_dev);
3532 			break;
3533 		}
3534 
3535 		if (idev) {
3536 			rt6_mtu_change(dev, dev->mtu);
3537 			idev->cnf.mtu6 = dev->mtu;
3538 			break;
3539 		}
3540 
3541 		/* allocate new idev */
3542 		idev = ipv6_add_dev(dev);
3543 		if (IS_ERR(idev))
3544 			break;
3545 
3546 		/* device is still not ready */
3547 		if (!(idev->if_flags & IF_READY))
3548 			break;
3549 
3550 		run_pending = 1;
3551 		fallthrough;
3552 	case NETDEV_UP:
3553 	case NETDEV_CHANGE:
3554 		if (dev->flags & IFF_SLAVE)
3555 			break;
3556 
3557 		if (idev && idev->cnf.disable_ipv6)
3558 			break;
3559 
3560 		if (event == NETDEV_UP) {
3561 			/* restore routes for permanent addresses */
3562 			addrconf_permanent_addr(net, dev);
3563 
3564 			if (!addrconf_link_ready(dev)) {
3565 				/* device is not ready yet. */
3566 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3567 					 dev->name);
3568 				break;
3569 			}
3570 
3571 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3572 				idev = ipv6_add_dev(dev);
3573 
3574 			if (!IS_ERR_OR_NULL(idev)) {
3575 				idev->if_flags |= IF_READY;
3576 				run_pending = 1;
3577 			}
3578 		} else if (event == NETDEV_CHANGE) {
3579 			if (!addrconf_link_ready(dev)) {
3580 				/* device is still not ready. */
3581 				rt6_sync_down_dev(dev, event);
3582 				break;
3583 			}
3584 
3585 			if (!IS_ERR_OR_NULL(idev)) {
3586 				if (idev->if_flags & IF_READY) {
3587 					/* device is already configured -
3588 					 * but resend MLD reports, we might
3589 					 * have roamed and need to update
3590 					 * multicast snooping switches
3591 					 */
3592 					ipv6_mc_up(idev);
3593 					change_info = ptr;
3594 					if (change_info->flags_changed & IFF_NOARP)
3595 						addrconf_dad_run(idev, true);
3596 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3597 					break;
3598 				}
3599 				idev->if_flags |= IF_READY;
3600 			}
3601 
3602 			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3603 				dev->name);
3604 
3605 			run_pending = 1;
3606 		}
3607 
3608 		switch (dev->type) {
3609 #if IS_ENABLED(CONFIG_IPV6_SIT)
3610 		case ARPHRD_SIT:
3611 			addrconf_sit_config(dev);
3612 			break;
3613 #endif
3614 #if IS_ENABLED(CONFIG_NET_IPGRE)
3615 		case ARPHRD_IPGRE:
3616 			addrconf_gre_config(dev);
3617 			break;
3618 #endif
3619 		case ARPHRD_LOOPBACK:
3620 			init_loopback(dev);
3621 			break;
3622 
3623 		default:
3624 			addrconf_dev_config(dev);
3625 			break;
3626 		}
3627 
3628 		if (!IS_ERR_OR_NULL(idev)) {
3629 			if (run_pending)
3630 				addrconf_dad_run(idev, false);
3631 
3632 			/* Device has an address by now */
3633 			rt6_sync_up(dev, RTNH_F_DEAD);
3634 
3635 			/*
3636 			 * If the MTU changed during the interface down,
3637 			 * when the interface up, the changed MTU must be
3638 			 * reflected in the idev as well as routers.
3639 			 */
3640 			if (idev->cnf.mtu6 != dev->mtu &&
3641 			    dev->mtu >= IPV6_MIN_MTU) {
3642 				rt6_mtu_change(dev, dev->mtu);
3643 				idev->cnf.mtu6 = dev->mtu;
3644 			}
3645 			idev->tstamp = jiffies;
3646 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3647 
3648 			/*
3649 			 * If the changed mtu during down is lower than
3650 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3651 			 */
3652 			if (dev->mtu < IPV6_MIN_MTU)
3653 				addrconf_ifdown(dev, dev != net->loopback_dev);
3654 		}
3655 		break;
3656 
3657 	case NETDEV_DOWN:
3658 	case NETDEV_UNREGISTER:
3659 		/*
3660 		 *	Remove all addresses from this interface.
3661 		 */
3662 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3663 		break;
3664 
3665 	case NETDEV_CHANGENAME:
3666 		if (idev) {
3667 			snmp6_unregister_dev(idev);
3668 			addrconf_sysctl_unregister(idev);
3669 			err = addrconf_sysctl_register(idev);
3670 			if (err)
3671 				return notifier_from_errno(err);
3672 			err = snmp6_register_dev(idev);
3673 			if (err) {
3674 				addrconf_sysctl_unregister(idev);
3675 				return notifier_from_errno(err);
3676 			}
3677 		}
3678 		break;
3679 
3680 	case NETDEV_PRE_TYPE_CHANGE:
3681 	case NETDEV_POST_TYPE_CHANGE:
3682 		if (idev)
3683 			addrconf_type_change(dev, event);
3684 		break;
3685 
3686 	case NETDEV_CHANGEUPPER:
3687 		info = ptr;
3688 
3689 		/* flush all routes if dev is linked to or unlinked from
3690 		 * an L3 master device (e.g., VRF)
3691 		 */
3692 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3693 			addrconf_ifdown(dev, false);
3694 	}
3695 
3696 	return NOTIFY_OK;
3697 }
3698 
3699 /*
3700  *	addrconf module should be notified of a device going up
3701  */
3702 static struct notifier_block ipv6_dev_notf = {
3703 	.notifier_call = addrconf_notify,
3704 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3705 };
3706 
addrconf_type_change(struct net_device *dev, unsigned long event)3707 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3708 {
3709 	struct inet6_dev *idev;
3710 	ASSERT_RTNL();
3711 
3712 	idev = __in6_dev_get(dev);
3713 
3714 	if (event == NETDEV_POST_TYPE_CHANGE)
3715 		ipv6_mc_remap(idev);
3716 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3717 		ipv6_mc_unmap(idev);
3718 }
3719 
addr_is_local(const struct in6_addr *addr)3720 static bool addr_is_local(const struct in6_addr *addr)
3721 {
3722 	return ipv6_addr_type(addr) &
3723 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3724 }
3725 
addrconf_ifdown(struct net_device *dev, bool unregister)3726 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3727 {
3728 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3729 	struct net *net = dev_net(dev);
3730 	struct inet6_dev *idev;
3731 	struct inet6_ifaddr *ifa;
3732 	LIST_HEAD(tmp_addr_list);
3733 	bool keep_addr = false;
3734 	bool was_ready;
3735 	int state, i;
3736 
3737 	ASSERT_RTNL();
3738 
3739 	rt6_disable_ip(dev, event);
3740 
3741 	idev = __in6_dev_get(dev);
3742 	if (!idev)
3743 		return -ENODEV;
3744 
3745 	/*
3746 	 * Step 1: remove reference to ipv6 device from parent device.
3747 	 *	   Do not dev_put!
3748 	 */
3749 	if (unregister) {
3750 		idev->dead = 1;
3751 
3752 		/* protected by rtnl_lock */
3753 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3754 
3755 		/* Step 1.5: remove snmp6 entry */
3756 		snmp6_unregister_dev(idev);
3757 
3758 	}
3759 
3760 	/* combine the user config with event to determine if permanent
3761 	 * addresses are to be removed from address hash table
3762 	 */
3763 	if (!unregister && !idev->cnf.disable_ipv6) {
3764 		/* aggregate the system setting and interface setting */
3765 		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3766 
3767 		if (!_keep_addr)
3768 			_keep_addr = idev->cnf.keep_addr_on_down;
3769 
3770 		keep_addr = (_keep_addr > 0);
3771 	}
3772 
3773 	/* Step 2: clear hash table */
3774 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3775 		struct hlist_head *h = &inet6_addr_lst[i];
3776 
3777 		spin_lock_bh(&addrconf_hash_lock);
3778 restart:
3779 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3780 			if (ifa->idev == idev) {
3781 				addrconf_del_dad_work(ifa);
3782 				/* combined flag + permanent flag decide if
3783 				 * address is retained on a down event
3784 				 */
3785 				if (!keep_addr ||
3786 				    !(ifa->flags & IFA_F_PERMANENT) ||
3787 				    addr_is_local(&ifa->addr)) {
3788 					hlist_del_init_rcu(&ifa->addr_lst);
3789 					goto restart;
3790 				}
3791 			}
3792 		}
3793 		spin_unlock_bh(&addrconf_hash_lock);
3794 	}
3795 
3796 	write_lock_bh(&idev->lock);
3797 
3798 	addrconf_del_rs_timer(idev);
3799 
3800 	/* Step 2: clear flags for stateless addrconf, repeated down
3801 	 *         detection
3802 	 */
3803 	was_ready = idev->if_flags & IF_READY;
3804 	if (!unregister)
3805 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3806 
3807 	/* Step 3: clear tempaddr list */
3808 	while (!list_empty(&idev->tempaddr_list)) {
3809 		ifa = list_first_entry(&idev->tempaddr_list,
3810 				       struct inet6_ifaddr, tmp_list);
3811 		list_del(&ifa->tmp_list);
3812 		write_unlock_bh(&idev->lock);
3813 		spin_lock_bh(&ifa->lock);
3814 
3815 		if (ifa->ifpub) {
3816 			in6_ifa_put(ifa->ifpub);
3817 			ifa->ifpub = NULL;
3818 		}
3819 		spin_unlock_bh(&ifa->lock);
3820 		in6_ifa_put(ifa);
3821 		write_lock_bh(&idev->lock);
3822 	}
3823 
3824 	list_for_each_entry(ifa, &idev->addr_list, if_list)
3825 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3826 	write_unlock_bh(&idev->lock);
3827 
3828 	while (!list_empty(&tmp_addr_list)) {
3829 		struct fib6_info *rt = NULL;
3830 		bool keep;
3831 
3832 		ifa = list_first_entry(&tmp_addr_list,
3833 				       struct inet6_ifaddr, if_list_aux);
3834 		list_del(&ifa->if_list_aux);
3835 
3836 		addrconf_del_dad_work(ifa);
3837 
3838 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3839 			!addr_is_local(&ifa->addr);
3840 
3841 		spin_lock_bh(&ifa->lock);
3842 
3843 		if (keep) {
3844 			/* set state to skip the notifier below */
3845 			state = INET6_IFADDR_STATE_DEAD;
3846 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3847 			if (!(ifa->flags & IFA_F_NODAD))
3848 				ifa->flags |= IFA_F_TENTATIVE;
3849 
3850 			rt = ifa->rt;
3851 			ifa->rt = NULL;
3852 		} else {
3853 			state = ifa->state;
3854 			ifa->state = INET6_IFADDR_STATE_DEAD;
3855 		}
3856 
3857 		spin_unlock_bh(&ifa->lock);
3858 
3859 		if (rt)
3860 			ip6_del_rt(net, rt, false);
3861 
3862 		if (state != INET6_IFADDR_STATE_DEAD) {
3863 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3864 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3865 		} else {
3866 			if (idev->cnf.forwarding)
3867 				addrconf_leave_anycast(ifa);
3868 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3869 		}
3870 
3871 		if (!keep) {
3872 			write_lock_bh(&idev->lock);
3873 			list_del_rcu(&ifa->if_list);
3874 			write_unlock_bh(&idev->lock);
3875 			in6_ifa_put(ifa);
3876 		}
3877 	}
3878 
3879 	/* Step 5: Discard anycast and multicast list */
3880 	if (unregister) {
3881 		ipv6_ac_destroy_dev(idev);
3882 		ipv6_mc_destroy_dev(idev);
3883 	} else if (was_ready) {
3884 		ipv6_mc_down(idev);
3885 	}
3886 
3887 	idev->tstamp = jiffies;
3888 
3889 	/* Last: Shot the device (if unregistered) */
3890 	if (unregister) {
3891 		addrconf_sysctl_unregister(idev);
3892 		neigh_parms_release(&nd_tbl, idev->nd_parms);
3893 		neigh_ifdown(&nd_tbl, dev);
3894 		in6_dev_put(idev);
3895 	}
3896 	return 0;
3897 }
3898 
addrconf_rs_timer(struct timer_list *t)3899 static void addrconf_rs_timer(struct timer_list *t)
3900 {
3901 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3902 	struct net_device *dev = idev->dev;
3903 	struct in6_addr lladdr;
3904 
3905 	write_lock(&idev->lock);
3906 	if (idev->dead || !(idev->if_flags & IF_READY))
3907 		goto out;
3908 
3909 	if (!ipv6_accept_ra(idev))
3910 		goto out;
3911 
3912 	/* Announcement received after solicitation was sent */
3913 	if (idev->if_flags & IF_RA_RCVD)
3914 		goto out;
3915 
3916 	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3917 		write_unlock(&idev->lock);
3918 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3919 			ndisc_send_rs(dev, &lladdr,
3920 				      &in6addr_linklocal_allrouters);
3921 		else
3922 			goto put;
3923 
3924 		write_lock(&idev->lock);
3925 		idev->rs_interval = rfc3315_s14_backoff_update(
3926 			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3927 		/* The wait after the last probe can be shorter */
3928 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3929 					     idev->cnf.rtr_solicits) ?
3930 				      idev->cnf.rtr_solicit_delay :
3931 				      idev->rs_interval);
3932 	} else {
3933 		/*
3934 		 * Note: we do not support deprecated "all on-link"
3935 		 * assumption any longer.
3936 		 */
3937 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3938 	}
3939 
3940 out:
3941 	write_unlock(&idev->lock);
3942 put:
3943 	in6_dev_put(idev);
3944 }
3945 
3946 /*
3947  *	Duplicate Address Detection
3948  */
addrconf_dad_kick(struct inet6_ifaddr *ifp)3949 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3950 {
3951 	unsigned long rand_num;
3952 	struct inet6_dev *idev = ifp->idev;
3953 	u64 nonce;
3954 
3955 	if (ifp->flags & IFA_F_OPTIMISTIC)
3956 		rand_num = 0;
3957 	else
3958 		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3959 
3960 	nonce = 0;
3961 	if (idev->cnf.enhanced_dad ||
3962 	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3963 		do
3964 			get_random_bytes(&nonce, 6);
3965 		while (nonce == 0);
3966 	}
3967 	ifp->dad_nonce = nonce;
3968 	ifp->dad_probes = idev->cnf.dad_transmits;
3969 	addrconf_mod_dad_work(ifp, rand_num);
3970 }
3971 
addrconf_dad_begin(struct inet6_ifaddr *ifp)3972 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3973 {
3974 	struct inet6_dev *idev = ifp->idev;
3975 	struct net_device *dev = idev->dev;
3976 	bool bump_id, notify = false;
3977 	struct net *net;
3978 
3979 	addrconf_join_solict(dev, &ifp->addr);
3980 
3981 	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3982 
3983 	read_lock_bh(&idev->lock);
3984 	spin_lock(&ifp->lock);
3985 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3986 		goto out;
3987 
3988 	net = dev_net(dev);
3989 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3990 	    (net->ipv6.devconf_all->accept_dad < 1 &&
3991 	     idev->cnf.accept_dad < 1) ||
3992 	    !(ifp->flags&IFA_F_TENTATIVE) ||
3993 	    ifp->flags & IFA_F_NODAD) {
3994 		bool send_na = false;
3995 
3996 		if (ifp->flags & IFA_F_TENTATIVE &&
3997 		    !(ifp->flags & IFA_F_OPTIMISTIC))
3998 			send_na = true;
3999 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4000 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4001 		spin_unlock(&ifp->lock);
4002 		read_unlock_bh(&idev->lock);
4003 
4004 		addrconf_dad_completed(ifp, bump_id, send_na);
4005 		return;
4006 	}
4007 
4008 	if (!(idev->if_flags & IF_READY)) {
4009 		spin_unlock(&ifp->lock);
4010 		read_unlock_bh(&idev->lock);
4011 		/*
4012 		 * If the device is not ready:
4013 		 * - keep it tentative if it is a permanent address.
4014 		 * - otherwise, kill it.
4015 		 */
4016 		in6_ifa_hold(ifp);
4017 		addrconf_dad_stop(ifp, 0);
4018 		return;
4019 	}
4020 
4021 	/*
4022 	 * Optimistic nodes can start receiving
4023 	 * Frames right away
4024 	 */
4025 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4026 		ip6_ins_rt(net, ifp->rt);
4027 		if (ipv6_use_optimistic_addr(net, idev)) {
4028 			/* Because optimistic nodes can use this address,
4029 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4030 			 */
4031 			notify = true;
4032 		}
4033 	}
4034 
4035 	addrconf_dad_kick(ifp);
4036 out:
4037 	spin_unlock(&ifp->lock);
4038 	read_unlock_bh(&idev->lock);
4039 	if (notify)
4040 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4041 }
4042 
addrconf_dad_start(struct inet6_ifaddr *ifp)4043 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4044 {
4045 	bool begin_dad = false;
4046 
4047 	spin_lock_bh(&ifp->lock);
4048 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4049 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4050 		begin_dad = true;
4051 	}
4052 	spin_unlock_bh(&ifp->lock);
4053 
4054 	if (begin_dad)
4055 		addrconf_mod_dad_work(ifp, 0);
4056 }
4057 
addrconf_dad_work(struct work_struct *w)4058 static void addrconf_dad_work(struct work_struct *w)
4059 {
4060 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4061 						struct inet6_ifaddr,
4062 						dad_work);
4063 	struct inet6_dev *idev = ifp->idev;
4064 	bool bump_id, disable_ipv6 = false;
4065 	struct in6_addr mcaddr;
4066 
4067 	enum {
4068 		DAD_PROCESS,
4069 		DAD_BEGIN,
4070 		DAD_ABORT,
4071 	} action = DAD_PROCESS;
4072 
4073 	rtnl_lock();
4074 
4075 	spin_lock_bh(&ifp->lock);
4076 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4077 		action = DAD_BEGIN;
4078 		ifp->state = INET6_IFADDR_STATE_DAD;
4079 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4080 		action = DAD_ABORT;
4081 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4082 
4083 		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4084 		     idev->cnf.accept_dad > 1) &&
4085 		    !idev->cnf.disable_ipv6 &&
4086 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4087 			struct in6_addr addr;
4088 
4089 			addr.s6_addr32[0] = htonl(0xfe800000);
4090 			addr.s6_addr32[1] = 0;
4091 
4092 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4093 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4094 				/* DAD failed for link-local based on MAC */
4095 				idev->cnf.disable_ipv6 = 1;
4096 
4097 				pr_info("%s: IPv6 being disabled!\n",
4098 					ifp->idev->dev->name);
4099 				disable_ipv6 = true;
4100 			}
4101 		}
4102 	}
4103 	spin_unlock_bh(&ifp->lock);
4104 
4105 	if (action == DAD_BEGIN) {
4106 		addrconf_dad_begin(ifp);
4107 		goto out;
4108 	} else if (action == DAD_ABORT) {
4109 		in6_ifa_hold(ifp);
4110 		addrconf_dad_stop(ifp, 1);
4111 		if (disable_ipv6)
4112 			addrconf_ifdown(idev->dev, false);
4113 		goto out;
4114 	}
4115 
4116 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4117 		goto out;
4118 
4119 	write_lock_bh(&idev->lock);
4120 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4121 		write_unlock_bh(&idev->lock);
4122 		goto out;
4123 	}
4124 
4125 	spin_lock(&ifp->lock);
4126 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4127 		spin_unlock(&ifp->lock);
4128 		write_unlock_bh(&idev->lock);
4129 		goto out;
4130 	}
4131 
4132 	if (ifp->dad_probes == 0) {
4133 		bool send_na = false;
4134 
4135 		/*
4136 		 * DAD was successful
4137 		 */
4138 
4139 		if (ifp->flags & IFA_F_TENTATIVE &&
4140 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4141 			send_na = true;
4142 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4143 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4144 		spin_unlock(&ifp->lock);
4145 		write_unlock_bh(&idev->lock);
4146 
4147 		addrconf_dad_completed(ifp, bump_id, send_na);
4148 
4149 		goto out;
4150 	}
4151 
4152 	ifp->dad_probes--;
4153 	addrconf_mod_dad_work(ifp,
4154 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4155 				  HZ/100));
4156 	spin_unlock(&ifp->lock);
4157 	write_unlock_bh(&idev->lock);
4158 
4159 	/* send a neighbour solicitation for our addr */
4160 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4161 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4162 		      ifp->dad_nonce);
4163 out:
4164 	in6_ifa_put(ifp);
4165 	rtnl_unlock();
4166 }
4167 
4168 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)4169 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4170 {
4171 	struct inet6_ifaddr *ifpiter;
4172 	struct inet6_dev *idev = ifp->idev;
4173 
4174 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4175 		if (ifpiter->scope > IFA_LINK)
4176 			break;
4177 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4178 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4179 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4180 		    IFA_F_PERMANENT)
4181 			return false;
4182 	}
4183 	return true;
4184 }
4185 
addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, bool send_na)4186 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4187 				   bool send_na)
4188 {
4189 	struct net_device *dev = ifp->idev->dev;
4190 	struct in6_addr lladdr;
4191 	bool send_rs, send_mld;
4192 
4193 	addrconf_del_dad_work(ifp);
4194 
4195 	/*
4196 	 *	Configure the address for reception. Now it is valid.
4197 	 */
4198 
4199 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4200 
4201 	/* If added prefix is link local and we are prepared to process
4202 	   router advertisements, start sending router solicitations.
4203 	 */
4204 
4205 	read_lock_bh(&ifp->idev->lock);
4206 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4207 	send_rs = send_mld &&
4208 		  ipv6_accept_ra(ifp->idev) &&
4209 		  ifp->idev->cnf.rtr_solicits != 0 &&
4210 		  (dev->flags & IFF_LOOPBACK) == 0 &&
4211 		  (dev->type != ARPHRD_TUNNEL);
4212 	read_unlock_bh(&ifp->idev->lock);
4213 
4214 	/* While dad is in progress mld report's source address is in6_addrany.
4215 	 * Resend with proper ll now.
4216 	 */
4217 	if (send_mld)
4218 		ipv6_mc_dad_complete(ifp->idev);
4219 
4220 	/* send unsolicited NA if enabled */
4221 	if (send_na &&
4222 	    (ifp->idev->cnf.ndisc_notify ||
4223 	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4224 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4225 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4226 			      /*solicited=*/ false, /*override=*/ true,
4227 			      /*inc_opt=*/ true);
4228 	}
4229 
4230 	if (send_rs) {
4231 		/*
4232 		 *	If a host as already performed a random delay
4233 		 *	[...] as part of DAD [...] there is no need
4234 		 *	to delay again before sending the first RS
4235 		 */
4236 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4237 			return;
4238 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4239 
4240 		write_lock_bh(&ifp->idev->lock);
4241 		spin_lock(&ifp->lock);
4242 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4243 			ifp->idev->cnf.rtr_solicit_interval);
4244 		ifp->idev->rs_probes = 1;
4245 		ifp->idev->if_flags |= IF_RS_SENT;
4246 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4247 		spin_unlock(&ifp->lock);
4248 		write_unlock_bh(&ifp->idev->lock);
4249 	}
4250 
4251 	if (bump_id)
4252 		rt_genid_bump_ipv6(dev_net(dev));
4253 
4254 	/* Make sure that a new temporary address will be created
4255 	 * before this temporary address becomes deprecated.
4256 	 */
4257 	if (ifp->flags & IFA_F_TEMPORARY)
4258 		addrconf_verify_rtnl();
4259 }
4260 
addrconf_dad_run(struct inet6_dev *idev, bool restart)4261 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4262 {
4263 	struct inet6_ifaddr *ifp;
4264 
4265 	read_lock_bh(&idev->lock);
4266 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4267 		spin_lock(&ifp->lock);
4268 		if ((ifp->flags & IFA_F_TENTATIVE &&
4269 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4270 			if (restart)
4271 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4272 			addrconf_dad_kick(ifp);
4273 		}
4274 		spin_unlock(&ifp->lock);
4275 	}
4276 	read_unlock_bh(&idev->lock);
4277 }
4278 
4279 #ifdef CONFIG_PROC_FS
4280 struct if6_iter_state {
4281 	struct seq_net_private p;
4282 	int bucket;
4283 	int offset;
4284 };
4285 
if6_get_first(struct seq_file *seq, loff_t pos)4286 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4287 {
4288 	struct if6_iter_state *state = seq->private;
4289 	struct net *net = seq_file_net(seq);
4290 	struct inet6_ifaddr *ifa = NULL;
4291 	int p = 0;
4292 
4293 	/* initial bucket if pos is 0 */
4294 	if (pos == 0) {
4295 		state->bucket = 0;
4296 		state->offset = 0;
4297 	}
4298 
4299 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4300 		hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4301 					 addr_lst) {
4302 			if (!net_eq(dev_net(ifa->idev->dev), net))
4303 				continue;
4304 			/* sync with offset */
4305 			if (p < state->offset) {
4306 				p++;
4307 				continue;
4308 			}
4309 			return ifa;
4310 		}
4311 
4312 		/* prepare for next bucket */
4313 		state->offset = 0;
4314 		p = 0;
4315 	}
4316 	return NULL;
4317 }
4318 
if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)4319 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4320 					 struct inet6_ifaddr *ifa)
4321 {
4322 	struct if6_iter_state *state = seq->private;
4323 	struct net *net = seq_file_net(seq);
4324 
4325 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4326 		if (!net_eq(dev_net(ifa->idev->dev), net))
4327 			continue;
4328 		state->offset++;
4329 		return ifa;
4330 	}
4331 
4332 	state->offset = 0;
4333 	while (++state->bucket < IN6_ADDR_HSIZE) {
4334 		hlist_for_each_entry_rcu(ifa,
4335 				     &inet6_addr_lst[state->bucket], addr_lst) {
4336 			if (!net_eq(dev_net(ifa->idev->dev), net))
4337 				continue;
4338 			return ifa;
4339 		}
4340 	}
4341 
4342 	return NULL;
4343 }
4344 
4345 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
__acquiresnull4346 	__acquires(rcu)
4347 {
4348 	rcu_read_lock();
4349 	return if6_get_first(seq, *pos);
4350 }
4351 
if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)4352 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4353 {
4354 	struct inet6_ifaddr *ifa;
4355 
4356 	ifa = if6_get_next(seq, v);
4357 	++*pos;
4358 	return ifa;
4359 }
4360 
4361 static void if6_seq_stop(struct seq_file *seq, void *v)
__releasesnull4362 	__releases(rcu)
4363 {
4364 	rcu_read_unlock();
4365 }
4366 
if6_seq_show(struct seq_file *seq, void *v)4367 static int if6_seq_show(struct seq_file *seq, void *v)
4368 {
4369 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4370 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4371 		   &ifp->addr,
4372 		   ifp->idev->dev->ifindex,
4373 		   ifp->prefix_len,
4374 		   ifp->scope,
4375 		   (u8) ifp->flags,
4376 		   ifp->idev->dev->name);
4377 	return 0;
4378 }
4379 
4380 static const struct seq_operations if6_seq_ops = {
4381 	.start	= if6_seq_start,
4382 	.next	= if6_seq_next,
4383 	.show	= if6_seq_show,
4384 	.stop	= if6_seq_stop,
4385 };
4386 
if6_proc_net_init(struct net *net)4387 static int __net_init if6_proc_net_init(struct net *net)
4388 {
4389 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4390 			sizeof(struct if6_iter_state)))
4391 		return -ENOMEM;
4392 	return 0;
4393 }
4394 
if6_proc_net_exit(struct net *net)4395 static void __net_exit if6_proc_net_exit(struct net *net)
4396 {
4397 	remove_proc_entry("if_inet6", net->proc_net);
4398 }
4399 
4400 static struct pernet_operations if6_proc_net_ops = {
4401 	.init = if6_proc_net_init,
4402 	.exit = if6_proc_net_exit,
4403 };
4404 
if6_proc_init(void)4405 int __init if6_proc_init(void)
4406 {
4407 	return register_pernet_subsys(&if6_proc_net_ops);
4408 }
4409 
if6_proc_exit(void)4410 void if6_proc_exit(void)
4411 {
4412 	unregister_pernet_subsys(&if6_proc_net_ops);
4413 }
4414 #endif	/* CONFIG_PROC_FS */
4415 
4416 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4417 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)4418 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4419 {
4420 	unsigned int hash = inet6_addr_hash(net, addr);
4421 	struct inet6_ifaddr *ifp = NULL;
4422 	int ret = 0;
4423 
4424 	rcu_read_lock();
4425 	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4426 		if (!net_eq(dev_net(ifp->idev->dev), net))
4427 			continue;
4428 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4429 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4430 			ret = 1;
4431 			break;
4432 		}
4433 	}
4434 	rcu_read_unlock();
4435 	return ret;
4436 }
4437 #endif
4438 
4439 /* RFC6554 has some algorithm to avoid loops in segment routing by
4440  * checking if the segments contains any of a local interface address.
4441  *
4442  * Quote:
4443  *
4444  * To detect loops in the SRH, a router MUST determine if the SRH
4445  * includes multiple addresses assigned to any interface on that router.
4446  * If such addresses appear more than once and are separated by at least
4447  * one address not assigned to that router.
4448  */
ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs, unsigned char nsegs)4449 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4450 			  unsigned char nsegs)
4451 {
4452 	const struct in6_addr *addr;
4453 	int i, ret = 0, found = 0;
4454 	struct inet6_ifaddr *ifp;
4455 	bool separated = false;
4456 	unsigned int hash;
4457 	bool hash_found;
4458 
4459 	rcu_read_lock();
4460 	for (i = 0; i < nsegs; i++) {
4461 		addr = &segs[i];
4462 		hash = inet6_addr_hash(net, addr);
4463 
4464 		hash_found = false;
4465 		hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4466 			if (!net_eq(dev_net(ifp->idev->dev), net))
4467 				continue;
4468 
4469 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4470 				hash_found = true;
4471 				break;
4472 			}
4473 		}
4474 
4475 		if (hash_found) {
4476 			if (found > 1 && separated) {
4477 				ret = 1;
4478 				break;
4479 			}
4480 
4481 			separated = false;
4482 			found++;
4483 		} else {
4484 			separated = true;
4485 		}
4486 	}
4487 	rcu_read_unlock();
4488 
4489 	return ret;
4490 }
4491 
4492 /*
4493  *	Periodic address status verification
4494  */
4495 
addrconf_verify_rtnl(void)4496 static void addrconf_verify_rtnl(void)
4497 {
4498 	unsigned long now, next, next_sec, next_sched;
4499 	struct inet6_ifaddr *ifp;
4500 	int i;
4501 
4502 	ASSERT_RTNL();
4503 
4504 	rcu_read_lock_bh();
4505 	now = jiffies;
4506 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4507 
4508 	cancel_delayed_work(&addr_chk_work);
4509 
4510 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4511 restart:
4512 		hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4513 			unsigned long age;
4514 
4515 			/* When setting preferred_lft to a value not zero or
4516 			 * infinity, while valid_lft is infinity
4517 			 * IFA_F_PERMANENT has a non-infinity life time.
4518 			 */
4519 			if ((ifp->flags & IFA_F_PERMANENT) &&
4520 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4521 				continue;
4522 
4523 			spin_lock(&ifp->lock);
4524 			/* We try to batch several events at once. */
4525 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4526 
4527 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4528 			    age >= ifp->valid_lft) {
4529 				spin_unlock(&ifp->lock);
4530 				in6_ifa_hold(ifp);
4531 				ipv6_del_addr(ifp);
4532 				goto restart;
4533 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4534 				spin_unlock(&ifp->lock);
4535 				continue;
4536 			} else if (age >= ifp->prefered_lft) {
4537 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4538 				int deprecate = 0;
4539 
4540 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4541 					deprecate = 1;
4542 					ifp->flags |= IFA_F_DEPRECATED;
4543 				}
4544 
4545 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4546 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4547 					next = ifp->tstamp + ifp->valid_lft * HZ;
4548 
4549 				spin_unlock(&ifp->lock);
4550 
4551 				if (deprecate) {
4552 					in6_ifa_hold(ifp);
4553 
4554 					ipv6_ifa_notify(0, ifp);
4555 					in6_ifa_put(ifp);
4556 					goto restart;
4557 				}
4558 			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
4559 				   !(ifp->flags&IFA_F_TENTATIVE)) {
4560 				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4561 					ifp->idev->cnf.dad_transmits *
4562 					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4563 
4564 				if (age >= ifp->prefered_lft - regen_advance) {
4565 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4566 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4567 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4568 					if (!ifp->regen_count && ifpub) {
4569 						ifp->regen_count++;
4570 						in6_ifa_hold(ifp);
4571 						in6_ifa_hold(ifpub);
4572 						spin_unlock(&ifp->lock);
4573 
4574 						spin_lock(&ifpub->lock);
4575 						ifpub->regen_count = 0;
4576 						spin_unlock(&ifpub->lock);
4577 						rcu_read_unlock_bh();
4578 						ipv6_create_tempaddr(ifpub, true);
4579 						in6_ifa_put(ifpub);
4580 						in6_ifa_put(ifp);
4581 						rcu_read_lock_bh();
4582 						goto restart;
4583 					}
4584 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4585 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4586 				spin_unlock(&ifp->lock);
4587 			} else {
4588 				/* ifp->prefered_lft <= ifp->valid_lft */
4589 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4590 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4591 				spin_unlock(&ifp->lock);
4592 			}
4593 		}
4594 	}
4595 
4596 	next_sec = round_jiffies_up(next);
4597 	next_sched = next;
4598 
4599 	/* If rounded timeout is accurate enough, accept it. */
4600 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4601 		next_sched = next_sec;
4602 
4603 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4604 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4605 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4606 
4607 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4608 		 now, next, next_sec, next_sched);
4609 	mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4610 	rcu_read_unlock_bh();
4611 }
4612 
addrconf_verify_work(struct work_struct *w)4613 static void addrconf_verify_work(struct work_struct *w)
4614 {
4615 	rtnl_lock();
4616 	addrconf_verify_rtnl();
4617 	rtnl_unlock();
4618 }
4619 
addrconf_verify(void)4620 static void addrconf_verify(void)
4621 {
4622 	mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4623 }
4624 
extract_addr(struct nlattr *addr, struct nlattr *local, struct in6_addr **peer_pfx)4625 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4626 				     struct in6_addr **peer_pfx)
4627 {
4628 	struct in6_addr *pfx = NULL;
4629 
4630 	*peer_pfx = NULL;
4631 
4632 	if (addr)
4633 		pfx = nla_data(addr);
4634 
4635 	if (local) {
4636 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4637 			*peer_pfx = pfx;
4638 		pfx = nla_data(local);
4639 	}
4640 
4641 	return pfx;
4642 }
4643 
4644 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4645 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4646 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4647 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4648 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4649 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4650 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4651 };
4652 
4653 static int
inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, struct netlink_ext_ack *extack)4654 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4655 		  struct netlink_ext_ack *extack)
4656 {
4657 	struct net *net = sock_net(skb->sk);
4658 	struct ifaddrmsg *ifm;
4659 	struct nlattr *tb[IFA_MAX+1];
4660 	struct in6_addr *pfx, *peer_pfx;
4661 	u32 ifa_flags;
4662 	int err;
4663 
4664 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4665 				     ifa_ipv6_policy, extack);
4666 	if (err < 0)
4667 		return err;
4668 
4669 	ifm = nlmsg_data(nlh);
4670 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4671 	if (!pfx)
4672 		return -EINVAL;
4673 
4674 	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4675 
4676 	/* We ignore other flags so far. */
4677 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4678 
4679 	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4680 			      ifm->ifa_prefixlen);
4681 }
4682 
modify_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, u32 flags, bool modify_peer)4683 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4684 			       unsigned long expires, u32 flags,
4685 			       bool modify_peer)
4686 {
4687 	struct fib6_info *f6i;
4688 	u32 prio;
4689 
4690 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4691 					ifp->prefix_len,
4692 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4693 	if (!f6i)
4694 		return -ENOENT;
4695 
4696 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4697 	if (f6i->fib6_metric != prio) {
4698 		/* delete old one */
4699 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4700 
4701 		/* add new one */
4702 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4703 				      ifp->prefix_len,
4704 				      ifp->rt_priority, ifp->idev->dev,
4705 				      expires, flags, GFP_KERNEL);
4706 	} else {
4707 		if (!expires)
4708 			fib6_clean_expires(f6i);
4709 		else
4710 			fib6_set_expires(f6i, expires);
4711 
4712 		fib6_info_release(f6i);
4713 	}
4714 
4715 	return 0;
4716 }
4717 
inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)4718 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4719 {
4720 	u32 flags;
4721 	clock_t expires;
4722 	unsigned long timeout;
4723 	bool was_managetempaddr;
4724 	bool had_prefixroute;
4725 	bool new_peer = false;
4726 
4727 	ASSERT_RTNL();
4728 
4729 	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4730 		return -EINVAL;
4731 
4732 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4733 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4734 		return -EINVAL;
4735 
4736 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4737 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4738 
4739 	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4740 	if (addrconf_finite_timeout(timeout)) {
4741 		expires = jiffies_to_clock_t(timeout * HZ);
4742 		cfg->valid_lft = timeout;
4743 		flags = RTF_EXPIRES;
4744 	} else {
4745 		expires = 0;
4746 		flags = 0;
4747 		cfg->ifa_flags |= IFA_F_PERMANENT;
4748 	}
4749 
4750 	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4751 	if (addrconf_finite_timeout(timeout)) {
4752 		if (timeout == 0)
4753 			cfg->ifa_flags |= IFA_F_DEPRECATED;
4754 		cfg->preferred_lft = timeout;
4755 	}
4756 
4757 	if (cfg->peer_pfx &&
4758 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4759 		if (!ipv6_addr_any(&ifp->peer_addr))
4760 			cleanup_prefix_route(ifp, expires, true, true);
4761 		new_peer = true;
4762 	}
4763 
4764 	spin_lock_bh(&ifp->lock);
4765 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4766 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4767 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4768 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4769 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4770 			IFA_F_NOPREFIXROUTE);
4771 	ifp->flags |= cfg->ifa_flags;
4772 	ifp->tstamp = jiffies;
4773 	ifp->valid_lft = cfg->valid_lft;
4774 	ifp->prefered_lft = cfg->preferred_lft;
4775 
4776 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4777 		ifp->rt_priority = cfg->rt_priority;
4778 
4779 	if (new_peer)
4780 		ifp->peer_addr = *cfg->peer_pfx;
4781 
4782 	spin_unlock_bh(&ifp->lock);
4783 	if (!(ifp->flags&IFA_F_TENTATIVE))
4784 		ipv6_ifa_notify(0, ifp);
4785 
4786 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4787 		int rc = -ENOENT;
4788 
4789 		if (had_prefixroute)
4790 			rc = modify_prefix_route(ifp, expires, flags, false);
4791 
4792 		/* prefix route could have been deleted; if so restore it */
4793 		if (rc == -ENOENT) {
4794 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4795 					      ifp->rt_priority, ifp->idev->dev,
4796 					      expires, flags, GFP_KERNEL);
4797 		}
4798 
4799 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4800 			rc = modify_prefix_route(ifp, expires, flags, true);
4801 
4802 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4803 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4804 					      ifp->rt_priority, ifp->idev->dev,
4805 					      expires, flags, GFP_KERNEL);
4806 		}
4807 	} else if (had_prefixroute) {
4808 		enum cleanup_prefix_rt_t action;
4809 		unsigned long rt_expires;
4810 
4811 		write_lock_bh(&ifp->idev->lock);
4812 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4813 		write_unlock_bh(&ifp->idev->lock);
4814 
4815 		if (action != CLEANUP_PREFIX_RT_NOP) {
4816 			cleanup_prefix_route(ifp, rt_expires,
4817 				action == CLEANUP_PREFIX_RT_DEL, false);
4818 		}
4819 	}
4820 
4821 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4822 		if (was_managetempaddr &&
4823 		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4824 			cfg->valid_lft = 0;
4825 			cfg->preferred_lft = 0;
4826 		}
4827 		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4828 				 cfg->preferred_lft, !was_managetempaddr,
4829 				 jiffies);
4830 	}
4831 
4832 	addrconf_verify_rtnl();
4833 
4834 	return 0;
4835 }
4836 
4837 static int
inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, struct netlink_ext_ack *extack)4838 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4839 		  struct netlink_ext_ack *extack)
4840 {
4841 	struct net *net = sock_net(skb->sk);
4842 	struct ifaddrmsg *ifm;
4843 	struct nlattr *tb[IFA_MAX+1];
4844 	struct in6_addr *peer_pfx;
4845 	struct inet6_ifaddr *ifa;
4846 	struct net_device *dev;
4847 	struct inet6_dev *idev;
4848 	struct ifa6_config cfg;
4849 	int err;
4850 
4851 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4852 				     ifa_ipv6_policy, extack);
4853 	if (err < 0)
4854 		return err;
4855 
4856 	memset(&cfg, 0, sizeof(cfg));
4857 
4858 	ifm = nlmsg_data(nlh);
4859 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4860 	if (!cfg.pfx)
4861 		return -EINVAL;
4862 
4863 	cfg.peer_pfx = peer_pfx;
4864 	cfg.plen = ifm->ifa_prefixlen;
4865 	if (tb[IFA_RT_PRIORITY])
4866 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4867 
4868 	cfg.valid_lft = INFINITY_LIFE_TIME;
4869 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4870 
4871 	if (tb[IFA_CACHEINFO]) {
4872 		struct ifa_cacheinfo *ci;
4873 
4874 		ci = nla_data(tb[IFA_CACHEINFO]);
4875 		cfg.valid_lft = ci->ifa_valid;
4876 		cfg.preferred_lft = ci->ifa_prefered;
4877 	}
4878 
4879 	dev =  __dev_get_by_index(net, ifm->ifa_index);
4880 	if (!dev)
4881 		return -ENODEV;
4882 
4883 	if (tb[IFA_FLAGS])
4884 		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4885 	else
4886 		cfg.ifa_flags = ifm->ifa_flags;
4887 
4888 	/* We ignore other flags so far. */
4889 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4890 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4891 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4892 
4893 	idev = ipv6_find_idev(dev);
4894 	if (IS_ERR(idev))
4895 		return PTR_ERR(idev);
4896 
4897 	if (!ipv6_allow_optimistic_dad(net, idev))
4898 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4899 
4900 	if (cfg.ifa_flags & IFA_F_NODAD &&
4901 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4902 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4903 		return -EINVAL;
4904 	}
4905 
4906 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4907 	if (!ifa) {
4908 		/*
4909 		 * It would be best to check for !NLM_F_CREATE here but
4910 		 * userspace already relies on not having to provide this.
4911 		 */
4912 		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4913 	}
4914 
4915 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4916 	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4917 		err = -EEXIST;
4918 	else
4919 		err = inet6_addr_modify(ifa, &cfg);
4920 
4921 	in6_ifa_put(ifa);
4922 
4923 	return err;
4924 }
4925 
put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags, u8 scope, int ifindex)4926 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4927 			  u8 scope, int ifindex)
4928 {
4929 	struct ifaddrmsg *ifm;
4930 
4931 	ifm = nlmsg_data(nlh);
4932 	ifm->ifa_family = AF_INET6;
4933 	ifm->ifa_prefixlen = prefixlen;
4934 	ifm->ifa_flags = flags;
4935 	ifm->ifa_scope = scope;
4936 	ifm->ifa_index = ifindex;
4937 }
4938 
put_cacheinfo(struct sk_buff *skb, unsigned long cstamp, unsigned long tstamp, u32 preferred, u32 valid)4939 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4940 			 unsigned long tstamp, u32 preferred, u32 valid)
4941 {
4942 	struct ifa_cacheinfo ci;
4943 
4944 	ci.cstamp = cstamp_delta(cstamp);
4945 	ci.tstamp = cstamp_delta(tstamp);
4946 	ci.ifa_prefered = preferred;
4947 	ci.ifa_valid = valid;
4948 
4949 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4950 }
4951 
rt_scope(int ifa_scope)4952 static inline int rt_scope(int ifa_scope)
4953 {
4954 	if (ifa_scope & IFA_HOST)
4955 		return RT_SCOPE_HOST;
4956 	else if (ifa_scope & IFA_LINK)
4957 		return RT_SCOPE_LINK;
4958 	else if (ifa_scope & IFA_SITE)
4959 		return RT_SCOPE_SITE;
4960 	else
4961 		return RT_SCOPE_UNIVERSE;
4962 }
4963 
inet6_ifaddr_msgsize(void)4964 static inline int inet6_ifaddr_msgsize(void)
4965 {
4966 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4967 	       + nla_total_size(16) /* IFA_LOCAL */
4968 	       + nla_total_size(16) /* IFA_ADDRESS */
4969 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4970 	       + nla_total_size(4)  /* IFA_FLAGS */
4971 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4972 }
4973 
4974 enum addr_type_t {
4975 	UNICAST_ADDR,
4976 	MULTICAST_ADDR,
4977 	ANYCAST_ADDR,
4978 };
4979 
4980 struct inet6_fill_args {
4981 	u32 portid;
4982 	u32 seq;
4983 	int event;
4984 	unsigned int flags;
4985 	int netnsid;
4986 	int ifindex;
4987 	enum addr_type_t type;
4988 };
4989 
inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa, struct inet6_fill_args *args)4990 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4991 			     struct inet6_fill_args *args)
4992 {
4993 	struct nlmsghdr  *nlh;
4994 	u32 preferred, valid;
4995 
4996 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4997 			sizeof(struct ifaddrmsg), args->flags);
4998 	if (!nlh)
4999 		return -EMSGSIZE;
5000 
5001 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5002 		      ifa->idev->dev->ifindex);
5003 
5004 	if (args->netnsid >= 0 &&
5005 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5006 		goto error;
5007 
5008 	spin_lock_bh(&ifa->lock);
5009 	if (!((ifa->flags&IFA_F_PERMANENT) &&
5010 	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
5011 		preferred = ifa->prefered_lft;
5012 		valid = ifa->valid_lft;
5013 		if (preferred != INFINITY_LIFE_TIME) {
5014 			long tval = (jiffies - ifa->tstamp)/HZ;
5015 			if (preferred > tval)
5016 				preferred -= tval;
5017 			else
5018 				preferred = 0;
5019 			if (valid != INFINITY_LIFE_TIME) {
5020 				if (valid > tval)
5021 					valid -= tval;
5022 				else
5023 					valid = 0;
5024 			}
5025 		}
5026 	} else {
5027 		preferred = INFINITY_LIFE_TIME;
5028 		valid = INFINITY_LIFE_TIME;
5029 	}
5030 	spin_unlock_bh(&ifa->lock);
5031 
5032 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5033 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5034 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5035 			goto error;
5036 	} else
5037 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5038 			goto error;
5039 
5040 	if (ifa->rt_priority &&
5041 	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5042 		goto error;
5043 
5044 	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5045 		goto error;
5046 
5047 	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5048 		goto error;
5049 
5050 	nlmsg_end(skb, nlh);
5051 	return 0;
5052 
5053 error:
5054 	nlmsg_cancel(skb, nlh);
5055 	return -EMSGSIZE;
5056 }
5057 
inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca, struct inet6_fill_args *args)5058 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5059 			       struct inet6_fill_args *args)
5060 {
5061 	struct nlmsghdr  *nlh;
5062 	u8 scope = RT_SCOPE_UNIVERSE;
5063 	int ifindex = ifmca->idev->dev->ifindex;
5064 
5065 	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5066 		scope = RT_SCOPE_SITE;
5067 
5068 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5069 			sizeof(struct ifaddrmsg), args->flags);
5070 	if (!nlh)
5071 		return -EMSGSIZE;
5072 
5073 	if (args->netnsid >= 0 &&
5074 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5075 		nlmsg_cancel(skb, nlh);
5076 		return -EMSGSIZE;
5077 	}
5078 
5079 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5080 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5081 	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5082 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5083 		nlmsg_cancel(skb, nlh);
5084 		return -EMSGSIZE;
5085 	}
5086 
5087 	nlmsg_end(skb, nlh);
5088 	return 0;
5089 }
5090 
inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca, struct inet6_fill_args *args)5091 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5092 			       struct inet6_fill_args *args)
5093 {
5094 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5095 	int ifindex = dev ? dev->ifindex : 1;
5096 	struct nlmsghdr  *nlh;
5097 	u8 scope = RT_SCOPE_UNIVERSE;
5098 
5099 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5100 		scope = RT_SCOPE_SITE;
5101 
5102 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5103 			sizeof(struct ifaddrmsg), args->flags);
5104 	if (!nlh)
5105 		return -EMSGSIZE;
5106 
5107 	if (args->netnsid >= 0 &&
5108 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5109 		nlmsg_cancel(skb, nlh);
5110 		return -EMSGSIZE;
5111 	}
5112 
5113 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5114 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5115 	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5116 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5117 		nlmsg_cancel(skb, nlh);
5118 		return -EMSGSIZE;
5119 	}
5120 
5121 	nlmsg_end(skb, nlh);
5122 	return 0;
5123 }
5124 
5125 /* 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)5126 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5127 			  struct netlink_callback *cb, int s_ip_idx,
5128 			  struct inet6_fill_args *fillargs)
5129 {
5130 	struct ifmcaddr6 *ifmca;
5131 	struct ifacaddr6 *ifaca;
5132 	int ip_idx = 0;
5133 	int err = 1;
5134 
5135 	read_lock_bh(&idev->lock);
5136 	switch (fillargs->type) {
5137 	case UNICAST_ADDR: {
5138 		struct inet6_ifaddr *ifa;
5139 		fillargs->event = RTM_NEWADDR;
5140 
5141 		/* unicast address incl. temp addr */
5142 		list_for_each_entry(ifa, &idev->addr_list, if_list) {
5143 			if (ip_idx < s_ip_idx)
5144 				goto next;
5145 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5146 			if (err < 0)
5147 				break;
5148 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5149 next:
5150 			ip_idx++;
5151 		}
5152 		break;
5153 	}
5154 	case MULTICAST_ADDR:
5155 		fillargs->event = RTM_GETMULTICAST;
5156 
5157 		/* multicast address */
5158 		for (ifmca = idev->mc_list; ifmca;
5159 		     ifmca = ifmca->next, ip_idx++) {
5160 			if (ip_idx < s_ip_idx)
5161 				continue;
5162 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5163 			if (err < 0)
5164 				break;
5165 		}
5166 		break;
5167 	case ANYCAST_ADDR:
5168 		fillargs->event = RTM_GETANYCAST;
5169 		/* anycast address */
5170 		for (ifaca = idev->ac_list; ifaca;
5171 		     ifaca = ifaca->aca_next, ip_idx++) {
5172 			if (ip_idx < s_ip_idx)
5173 				continue;
5174 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5175 			if (err < 0)
5176 				break;
5177 		}
5178 		break;
5179 	default:
5180 		break;
5181 	}
5182 	read_unlock_bh(&idev->lock);
5183 	cb->args[2] = ip_idx;
5184 	return err;
5185 }
5186 
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)5187 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5188 				       struct inet6_fill_args *fillargs,
5189 				       struct net **tgt_net, struct sock *sk,
5190 				       struct netlink_callback *cb)
5191 {
5192 	struct netlink_ext_ack *extack = cb->extack;
5193 	struct nlattr *tb[IFA_MAX+1];
5194 	struct ifaddrmsg *ifm;
5195 	int err, i;
5196 
5197 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5198 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5199 		return -EINVAL;
5200 	}
5201 
5202 	ifm = nlmsg_data(nlh);
5203 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5204 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5205 		return -EINVAL;
5206 	}
5207 
5208 	fillargs->ifindex = ifm->ifa_index;
5209 	if (fillargs->ifindex) {
5210 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5211 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5212 	}
5213 
5214 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5215 					    ifa_ipv6_policy, extack);
5216 	if (err < 0)
5217 		return err;
5218 
5219 	for (i = 0; i <= IFA_MAX; ++i) {
5220 		if (!tb[i])
5221 			continue;
5222 
5223 		if (i == IFA_TARGET_NETNSID) {
5224 			struct net *net;
5225 
5226 			fillargs->netnsid = nla_get_s32(tb[i]);
5227 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5228 			if (IS_ERR(net)) {
5229 				fillargs->netnsid = -1;
5230 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5231 				return PTR_ERR(net);
5232 			}
5233 			*tgt_net = net;
5234 		} else {
5235 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5236 			return -EINVAL;
5237 		}
5238 	}
5239 
5240 	return 0;
5241 }
5242 
inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb, enum addr_type_t type)5243 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5244 			   enum addr_type_t type)
5245 {
5246 	const struct nlmsghdr *nlh = cb->nlh;
5247 	struct inet6_fill_args fillargs = {
5248 		.portid = NETLINK_CB(cb->skb).portid,
5249 		.seq = cb->nlh->nlmsg_seq,
5250 		.flags = NLM_F_MULTI,
5251 		.netnsid = -1,
5252 		.type = type,
5253 	};
5254 	struct net *net = sock_net(skb->sk);
5255 	struct net *tgt_net = net;
5256 	int idx, s_idx, s_ip_idx;
5257 	int h, s_h;
5258 	struct net_device *dev;
5259 	struct inet6_dev *idev;
5260 	struct hlist_head *head;
5261 	int err = 0;
5262 
5263 	s_h = cb->args[0];
5264 	s_idx = idx = cb->args[1];
5265 	s_ip_idx = cb->args[2];
5266 
5267 	if (cb->strict_check) {
5268 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5269 						  skb->sk, cb);
5270 		if (err < 0)
5271 			goto put_tgt_net;
5272 
5273 		err = 0;
5274 		if (fillargs.ifindex) {
5275 			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5276 			if (!dev) {
5277 				err = -ENODEV;
5278 				goto put_tgt_net;
5279 			}
5280 			idev = __in6_dev_get(dev);
5281 			if (idev) {
5282 				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5283 						     &fillargs);
5284 				if (err > 0)
5285 					err = 0;
5286 			}
5287 			goto put_tgt_net;
5288 		}
5289 	}
5290 
5291 	rcu_read_lock();
5292 	cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5293 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5294 		idx = 0;
5295 		head = &tgt_net->dev_index_head[h];
5296 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5297 			if (idx < s_idx)
5298 				goto cont;
5299 			if (h > s_h || idx > s_idx)
5300 				s_ip_idx = 0;
5301 			idev = __in6_dev_get(dev);
5302 			if (!idev)
5303 				goto cont;
5304 
5305 			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5306 					   &fillargs) < 0)
5307 				goto done;
5308 cont:
5309 			idx++;
5310 		}
5311 	}
5312 done:
5313 	rcu_read_unlock();
5314 	cb->args[0] = h;
5315 	cb->args[1] = idx;
5316 put_tgt_net:
5317 	if (fillargs.netnsid >= 0)
5318 		put_net(tgt_net);
5319 
5320 	return skb->len ? : err;
5321 }
5322 
inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)5323 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5324 {
5325 	enum addr_type_t type = UNICAST_ADDR;
5326 
5327 	return inet6_dump_addr(skb, cb, type);
5328 }
5329 
inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)5330 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5331 {
5332 	enum addr_type_t type = MULTICAST_ADDR;
5333 
5334 	return inet6_dump_addr(skb, cb, type);
5335 }
5336 
5337 
inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)5338 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5339 {
5340 	enum addr_type_t type = ANYCAST_ADDR;
5341 
5342 	return inet6_dump_addr(skb, cb, type);
5343 }
5344 
inet6_rtm_valid_getaddr_req(struct sk_buff *skb, const struct nlmsghdr *nlh, struct nlattr **tb, struct netlink_ext_ack *extack)5345 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5346 				       const struct nlmsghdr *nlh,
5347 				       struct nlattr **tb,
5348 				       struct netlink_ext_ack *extack)
5349 {
5350 	struct ifaddrmsg *ifm;
5351 	int i, err;
5352 
5353 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5354 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5355 		return -EINVAL;
5356 	}
5357 
5358 	if (!netlink_strict_get_check(skb))
5359 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5360 					      ifa_ipv6_policy, extack);
5361 
5362 	ifm = nlmsg_data(nlh);
5363 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5364 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5365 		return -EINVAL;
5366 	}
5367 
5368 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5369 					    ifa_ipv6_policy, extack);
5370 	if (err)
5371 		return err;
5372 
5373 	for (i = 0; i <= IFA_MAX; i++) {
5374 		if (!tb[i])
5375 			continue;
5376 
5377 		switch (i) {
5378 		case IFA_TARGET_NETNSID:
5379 		case IFA_ADDRESS:
5380 		case IFA_LOCAL:
5381 			break;
5382 		default:
5383 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5384 			return -EINVAL;
5385 		}
5386 	}
5387 
5388 	return 0;
5389 }
5390 
inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh, struct netlink_ext_ack *extack)5391 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5392 			     struct netlink_ext_ack *extack)
5393 {
5394 	struct net *net = sock_net(in_skb->sk);
5395 	struct inet6_fill_args fillargs = {
5396 		.portid = NETLINK_CB(in_skb).portid,
5397 		.seq = nlh->nlmsg_seq,
5398 		.event = RTM_NEWADDR,
5399 		.flags = 0,
5400 		.netnsid = -1,
5401 	};
5402 	struct net *tgt_net = net;
5403 	struct ifaddrmsg *ifm;
5404 	struct nlattr *tb[IFA_MAX+1];
5405 	struct in6_addr *addr = NULL, *peer;
5406 	struct net_device *dev = NULL;
5407 	struct inet6_ifaddr *ifa;
5408 	struct sk_buff *skb;
5409 	int err;
5410 
5411 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5412 	if (err < 0)
5413 		return err;
5414 
5415 	if (tb[IFA_TARGET_NETNSID]) {
5416 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5417 
5418 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5419 						  fillargs.netnsid);
5420 		if (IS_ERR(tgt_net))
5421 			return PTR_ERR(tgt_net);
5422 	}
5423 
5424 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5425 	if (!addr) {
5426 		err = -EINVAL;
5427 		goto errout;
5428 	}
5429 	ifm = nlmsg_data(nlh);
5430 	if (ifm->ifa_index)
5431 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5432 
5433 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5434 	if (!ifa) {
5435 		err = -EADDRNOTAVAIL;
5436 		goto errout;
5437 	}
5438 
5439 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5440 	if (!skb) {
5441 		err = -ENOBUFS;
5442 		goto errout_ifa;
5443 	}
5444 
5445 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5446 	if (err < 0) {
5447 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5448 		WARN_ON(err == -EMSGSIZE);
5449 		kfree_skb(skb);
5450 		goto errout_ifa;
5451 	}
5452 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5453 errout_ifa:
5454 	in6_ifa_put(ifa);
5455 errout:
5456 	if (dev)
5457 		dev_put(dev);
5458 	if (fillargs.netnsid >= 0)
5459 		put_net(tgt_net);
5460 
5461 	return err;
5462 }
5463 
inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)5464 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5465 {
5466 	struct sk_buff *skb;
5467 	struct net *net = dev_net(ifa->idev->dev);
5468 	struct inet6_fill_args fillargs = {
5469 		.portid = 0,
5470 		.seq = 0,
5471 		.event = event,
5472 		.flags = 0,
5473 		.netnsid = -1,
5474 	};
5475 	int err = -ENOBUFS;
5476 
5477 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5478 	if (!skb)
5479 		goto errout;
5480 
5481 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5482 	if (err < 0) {
5483 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5484 		WARN_ON(err == -EMSGSIZE);
5485 		kfree_skb(skb);
5486 		goto errout;
5487 	}
5488 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5489 	return;
5490 errout:
5491 	if (err < 0)
5492 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5493 }
5494 
ipv6_store_devconf(struct ipv6_devconf *cnf, __s32 *array, int bytes)5495 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5496 				__s32 *array, int bytes)
5497 {
5498 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5499 
5500 	memset(array, 0, bytes);
5501 	array[DEVCONF_FORWARDING] = cnf->forwarding;
5502 	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5503 	array[DEVCONF_MTU6] = cnf->mtu6;
5504 	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5505 	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5506 	array[DEVCONF_AUTOCONF] = cnf->autoconf;
5507 	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5508 	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5509 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5510 		jiffies_to_msecs(cnf->rtr_solicit_interval);
5511 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5512 		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5513 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5514 		jiffies_to_msecs(cnf->rtr_solicit_delay);
5515 	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5516 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5517 		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5518 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5519 		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5520 	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5521 	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5522 	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5523 	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5524 	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5525 	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5526 	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5527 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5528 	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5529 #ifdef CONFIG_IPV6_ROUTER_PREF
5530 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5531 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5532 		jiffies_to_msecs(cnf->rtr_probe_interval);
5533 #ifdef CONFIG_IPV6_ROUTE_INFO
5534 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5535 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5536 #endif
5537 #endif
5538 	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5539 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5540 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5541 	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5542 	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5543 #endif
5544 #ifdef CONFIG_IPV6_MROUTE
5545 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5546 #endif
5547 	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5548 	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5549 	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5550 	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5551 	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5552 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5553 	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5554 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5555 	/* we omit DEVCONF_STABLE_SECRET for now */
5556 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5557 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5558 	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5559 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5560 	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5561 #ifdef CONFIG_IPV6_SEG6_HMAC
5562 	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5563 #endif
5564 	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5565 	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5566 	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5567 	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5568 	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5569 	array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft;
5570 }
5571 
inet6_ifla6_size(void)5572 static inline size_t inet6_ifla6_size(void)
5573 {
5574 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5575 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5576 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5577 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5578 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5579 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5580 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5581 	     + 0;
5582 }
5583 
inet6_if_nlmsg_size(void)5584 static inline size_t inet6_if_nlmsg_size(void)
5585 {
5586 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5587 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5588 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5589 	       + nla_total_size(4) /* IFLA_MTU */
5590 	       + nla_total_size(4) /* IFLA_LINK */
5591 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5592 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5593 }
5594 
__snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib, int bytes)5595 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5596 					int bytes)
5597 {
5598 	int i;
5599 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5600 	BUG_ON(pad < 0);
5601 
5602 	/* Use put_unaligned() because stats may not be aligned for u64. */
5603 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5604 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5605 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5606 
5607 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5608 }
5609 
__snmp6_fill_stats64(u64 *stats, void __percpu *mib, int bytes, size_t syncpoff)5610 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5611 					int bytes, size_t syncpoff)
5612 {
5613 	int i, c;
5614 	u64 buff[IPSTATS_MIB_MAX];
5615 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5616 
5617 	BUG_ON(pad < 0);
5618 
5619 	memset(buff, 0, sizeof(buff));
5620 	buff[0] = IPSTATS_MIB_MAX;
5621 
5622 	for_each_possible_cpu(c) {
5623 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5624 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5625 	}
5626 
5627 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5628 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5629 }
5630 
snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype, int bytes)5631 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5632 			     int bytes)
5633 {
5634 	switch (attrtype) {
5635 	case IFLA_INET6_STATS:
5636 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5637 				     offsetof(struct ipstats_mib, syncp));
5638 		break;
5639 	case IFLA_INET6_ICMP6STATS:
5640 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5641 		break;
5642 	}
5643 }
5644 
inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev, u32 ext_filter_mask)5645 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5646 				  u32 ext_filter_mask)
5647 {
5648 	struct nlattr *nla;
5649 	struct ifla_cacheinfo ci;
5650 
5651 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5652 		goto nla_put_failure;
5653 	ci.max_reasm_len = IPV6_MAXPLEN;
5654 	ci.tstamp = cstamp_delta(idev->tstamp);
5655 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5656 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5657 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5658 		goto nla_put_failure;
5659 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5660 	if (!nla)
5661 		goto nla_put_failure;
5662 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5663 
5664 	/* XXX - MC not implemented */
5665 
5666 	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5667 		return 0;
5668 
5669 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5670 	if (!nla)
5671 		goto nla_put_failure;
5672 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5673 
5674 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5675 	if (!nla)
5676 		goto nla_put_failure;
5677 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5678 
5679 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5680 	if (!nla)
5681 		goto nla_put_failure;
5682 	read_lock_bh(&idev->lock);
5683 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5684 	read_unlock_bh(&idev->lock);
5685 
5686 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5687 		goto nla_put_failure;
5688 
5689 	return 0;
5690 
5691 nla_put_failure:
5692 	return -EMSGSIZE;
5693 }
5694 
inet6_get_link_af_size(const struct net_device *dev, u32 ext_filter_mask)5695 static size_t inet6_get_link_af_size(const struct net_device *dev,
5696 				     u32 ext_filter_mask)
5697 {
5698 	if (!__in6_dev_get(dev))
5699 		return 0;
5700 
5701 	return inet6_ifla6_size();
5702 }
5703 
inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev, u32 ext_filter_mask)5704 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5705 			      u32 ext_filter_mask)
5706 {
5707 	struct inet6_dev *idev = __in6_dev_get(dev);
5708 
5709 	if (!idev)
5710 		return -ENODATA;
5711 
5712 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5713 		return -EMSGSIZE;
5714 
5715 	return 0;
5716 }
5717 
inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)5718 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5719 {
5720 	struct inet6_ifaddr *ifp;
5721 	struct net_device *dev = idev->dev;
5722 	bool clear_token, update_rs = false;
5723 	struct in6_addr ll_addr;
5724 
5725 	ASSERT_RTNL();
5726 
5727 	if (!token)
5728 		return -EINVAL;
5729 	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5730 		return -EINVAL;
5731 	if (!ipv6_accept_ra(idev))
5732 		return -EINVAL;
5733 	if (idev->cnf.rtr_solicits == 0)
5734 		return -EINVAL;
5735 
5736 	write_lock_bh(&idev->lock);
5737 
5738 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5739 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5740 
5741 	write_unlock_bh(&idev->lock);
5742 
5743 	clear_token = ipv6_addr_any(token);
5744 	if (clear_token)
5745 		goto update_lft;
5746 
5747 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5748 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5749 			     IFA_F_OPTIMISTIC)) {
5750 		/* If we're not ready, then normal ifup will take care
5751 		 * of this. Otherwise, we need to request our rs here.
5752 		 */
5753 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5754 		update_rs = true;
5755 	}
5756 
5757 update_lft:
5758 	write_lock_bh(&idev->lock);
5759 
5760 	if (update_rs) {
5761 		idev->if_flags |= IF_RS_SENT;
5762 		idev->rs_interval = rfc3315_s14_backoff_init(
5763 			idev->cnf.rtr_solicit_interval);
5764 		idev->rs_probes = 1;
5765 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5766 	}
5767 
5768 	/* Well, that's kinda nasty ... */
5769 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5770 		spin_lock(&ifp->lock);
5771 		if (ifp->tokenized) {
5772 			ifp->valid_lft = 0;
5773 			ifp->prefered_lft = 0;
5774 		}
5775 		spin_unlock(&ifp->lock);
5776 	}
5777 
5778 	write_unlock_bh(&idev->lock);
5779 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5780 	addrconf_verify_rtnl();
5781 	return 0;
5782 }
5783 
5784 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5785 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5786 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5787 };
5788 
check_addr_gen_mode(int mode)5789 static int check_addr_gen_mode(int mode)
5790 {
5791 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5792 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5793 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5794 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5795 		return -EINVAL;
5796 	return 1;
5797 }
5798 
check_stable_privacy(struct inet6_dev *idev, struct net *net, int mode)5799 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5800 				int mode)
5801 {
5802 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5803 	    !idev->cnf.stable_secret.initialized &&
5804 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5805 		return -EINVAL;
5806 	return 1;
5807 }
5808 
inet6_validate_link_af(const struct net_device *dev, const struct nlattr *nla)5809 static int inet6_validate_link_af(const struct net_device *dev,
5810 				  const struct nlattr *nla)
5811 {
5812 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5813 	struct inet6_dev *idev = NULL;
5814 	int err;
5815 
5816 	if (dev) {
5817 		idev = __in6_dev_get(dev);
5818 		if (!idev)
5819 			return -EAFNOSUPPORT;
5820 	}
5821 
5822 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5823 					  inet6_af_policy, NULL);
5824 	if (err)
5825 		return err;
5826 
5827 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5828 		return -EINVAL;
5829 
5830 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5831 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5832 
5833 		if (check_addr_gen_mode(mode) < 0)
5834 			return -EINVAL;
5835 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5836 			return -EINVAL;
5837 	}
5838 
5839 	return 0;
5840 }
5841 
inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)5842 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5843 {
5844 	struct inet6_dev *idev = __in6_dev_get(dev);
5845 	struct nlattr *tb[IFLA_INET6_MAX + 1];
5846 	int err;
5847 
5848 	if (!idev)
5849 		return -EAFNOSUPPORT;
5850 
5851 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5852 		return -EINVAL;
5853 
5854 	if (tb[IFLA_INET6_TOKEN]) {
5855 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5856 		if (err)
5857 			return err;
5858 	}
5859 
5860 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5861 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5862 
5863 		idev->cnf.addr_gen_mode = mode;
5864 	}
5865 
5866 	return 0;
5867 }
5868 
inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, u32 portid, u32 seq, int event, unsigned int flags)5869 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5870 			     u32 portid, u32 seq, int event, unsigned int flags)
5871 {
5872 	struct net_device *dev = idev->dev;
5873 	struct ifinfomsg *hdr;
5874 	struct nlmsghdr *nlh;
5875 	void *protoinfo;
5876 
5877 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5878 	if (!nlh)
5879 		return -EMSGSIZE;
5880 
5881 	hdr = nlmsg_data(nlh);
5882 	hdr->ifi_family = AF_INET6;
5883 	hdr->__ifi_pad = 0;
5884 	hdr->ifi_type = dev->type;
5885 	hdr->ifi_index = dev->ifindex;
5886 	hdr->ifi_flags = dev_get_flags(dev);
5887 	hdr->ifi_change = 0;
5888 
5889 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5890 	    (dev->addr_len &&
5891 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5892 	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5893 	    (dev->ifindex != dev_get_iflink(dev) &&
5894 	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5895 	    nla_put_u8(skb, IFLA_OPERSTATE,
5896 		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5897 		goto nla_put_failure;
5898 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5899 	if (!protoinfo)
5900 		goto nla_put_failure;
5901 
5902 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5903 		goto nla_put_failure;
5904 
5905 	nla_nest_end(skb, protoinfo);
5906 	nlmsg_end(skb, nlh);
5907 	return 0;
5908 
5909 nla_put_failure:
5910 	nlmsg_cancel(skb, nlh);
5911 	return -EMSGSIZE;
5912 }
5913 
inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh, struct netlink_ext_ack *extack)5914 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5915 				   struct netlink_ext_ack *extack)
5916 {
5917 	struct ifinfomsg *ifm;
5918 
5919 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5920 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5921 		return -EINVAL;
5922 	}
5923 
5924 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5925 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5926 		return -EINVAL;
5927 	}
5928 
5929 	ifm = nlmsg_data(nlh);
5930 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5931 	    ifm->ifi_change || ifm->ifi_index) {
5932 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5933 		return -EINVAL;
5934 	}
5935 
5936 	return 0;
5937 }
5938 
inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)5939 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5940 {
5941 	struct net *net = sock_net(skb->sk);
5942 	int h, s_h;
5943 	int idx = 0, s_idx;
5944 	struct net_device *dev;
5945 	struct inet6_dev *idev;
5946 	struct hlist_head *head;
5947 
5948 	/* only requests using strict checking can pass data to
5949 	 * influence the dump
5950 	 */
5951 	if (cb->strict_check) {
5952 		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5953 
5954 		if (err < 0)
5955 			return err;
5956 	}
5957 
5958 	s_h = cb->args[0];
5959 	s_idx = cb->args[1];
5960 
5961 	rcu_read_lock();
5962 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5963 		idx = 0;
5964 		head = &net->dev_index_head[h];
5965 		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5966 			if (idx < s_idx)
5967 				goto cont;
5968 			idev = __in6_dev_get(dev);
5969 			if (!idev)
5970 				goto cont;
5971 			if (inet6_fill_ifinfo(skb, idev,
5972 					      NETLINK_CB(cb->skb).portid,
5973 					      cb->nlh->nlmsg_seq,
5974 					      RTM_NEWLINK, NLM_F_MULTI) < 0)
5975 				goto out;
5976 cont:
5977 			idx++;
5978 		}
5979 	}
5980 out:
5981 	rcu_read_unlock();
5982 	cb->args[1] = idx;
5983 	cb->args[0] = h;
5984 
5985 	return skb->len;
5986 }
5987 
inet6_ifinfo_notify(int event, struct inet6_dev *idev)5988 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5989 {
5990 	struct sk_buff *skb;
5991 	struct net *net = dev_net(idev->dev);
5992 	int err = -ENOBUFS;
5993 
5994 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5995 	if (!skb)
5996 		goto errout;
5997 
5998 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5999 	if (err < 0) {
6000 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6001 		WARN_ON(err == -EMSGSIZE);
6002 		kfree_skb(skb);
6003 		goto errout;
6004 	}
6005 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6006 	return;
6007 errout:
6008 	if (err < 0)
6009 		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6010 }
6011 
inet6_prefix_nlmsg_size(void)6012 static inline size_t inet6_prefix_nlmsg_size(void)
6013 {
6014 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6015 	       + nla_total_size(sizeof(struct in6_addr))
6016 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6017 }
6018 
inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev, struct prefix_info *pinfo, u32 portid, u32 seq, int event, unsigned int flags)6019 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6020 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6021 			     int event, unsigned int flags)
6022 {
6023 	struct prefixmsg *pmsg;
6024 	struct nlmsghdr *nlh;
6025 	struct prefix_cacheinfo	ci;
6026 
6027 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6028 	if (!nlh)
6029 		return -EMSGSIZE;
6030 
6031 	pmsg = nlmsg_data(nlh);
6032 	pmsg->prefix_family = AF_INET6;
6033 	pmsg->prefix_pad1 = 0;
6034 	pmsg->prefix_pad2 = 0;
6035 	pmsg->prefix_ifindex = idev->dev->ifindex;
6036 	pmsg->prefix_len = pinfo->prefix_len;
6037 	pmsg->prefix_type = pinfo->type;
6038 	pmsg->prefix_pad3 = 0;
6039 	pmsg->prefix_flags = pinfo->flags;
6040 
6041 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6042 		goto nla_put_failure;
6043 	ci.preferred_time = ntohl(pinfo->prefered);
6044 	ci.valid_time = ntohl(pinfo->valid);
6045 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6046 		goto nla_put_failure;
6047 	nlmsg_end(skb, nlh);
6048 	return 0;
6049 
6050 nla_put_failure:
6051 	nlmsg_cancel(skb, nlh);
6052 	return -EMSGSIZE;
6053 }
6054 
inet6_prefix_notify(int event, struct inet6_dev *idev, struct prefix_info *pinfo)6055 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6056 			 struct prefix_info *pinfo)
6057 {
6058 	struct sk_buff *skb;
6059 	struct net *net = dev_net(idev->dev);
6060 	int err = -ENOBUFS;
6061 
6062 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6063 	if (!skb)
6064 		goto errout;
6065 
6066 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6067 	if (err < 0) {
6068 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6069 		WARN_ON(err == -EMSGSIZE);
6070 		kfree_skb(skb);
6071 		goto errout;
6072 	}
6073 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6074 	return;
6075 errout:
6076 	if (err < 0)
6077 		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6078 }
6079 
__ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)6080 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6081 {
6082 	struct net *net = dev_net(ifp->idev->dev);
6083 
6084 	if (event)
6085 		ASSERT_RTNL();
6086 
6087 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6088 
6089 	switch (event) {
6090 	case RTM_NEWADDR:
6091 		/*
6092 		 * If the address was optimistic we inserted the route at the
6093 		 * start of our DAD process, so we don't need to do it again.
6094 		 * If the device was taken down in the middle of the DAD
6095 		 * cycle there is a race where we could get here without a
6096 		 * host route, so nothing to insert. That will be fixed when
6097 		 * the device is brought up.
6098 		 */
6099 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6100 			ip6_ins_rt(net, ifp->rt);
6101 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6102 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6103 				&ifp->addr, ifp->idev->dev->name);
6104 		}
6105 
6106 		if (ifp->idev->cnf.forwarding)
6107 			addrconf_join_anycast(ifp);
6108 		if (!ipv6_addr_any(&ifp->peer_addr))
6109 			addrconf_prefix_route(&ifp->peer_addr, 128,
6110 					      ifp->rt_priority, ifp->idev->dev,
6111 					      0, 0, GFP_ATOMIC);
6112 		break;
6113 	case RTM_DELADDR:
6114 		if (ifp->idev->cnf.forwarding)
6115 			addrconf_leave_anycast(ifp);
6116 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6117 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6118 			struct fib6_info *rt;
6119 
6120 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6121 						       ifp->idev->dev, 0, 0,
6122 						       false);
6123 			if (rt)
6124 				ip6_del_rt(net, rt, false);
6125 		}
6126 		if (ifp->rt) {
6127 			ip6_del_rt(net, ifp->rt, false);
6128 			ifp->rt = NULL;
6129 		}
6130 		rt_genid_bump_ipv6(net);
6131 		break;
6132 	}
6133 	atomic_inc(&net->ipv6.dev_addr_genid);
6134 }
6135 
ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)6136 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6137 {
6138 	rcu_read_lock_bh();
6139 	if (likely(ifp->idev->dead == 0))
6140 		__ipv6_ifa_notify(event, ifp);
6141 	rcu_read_unlock_bh();
6142 }
6143 
6144 #ifdef CONFIG_SYSCTL
6145 
addrconf_sysctl_forward(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6146 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6147 		void *buffer, size_t *lenp, loff_t *ppos)
6148 {
6149 	int *valp = ctl->data;
6150 	int val = *valp;
6151 	loff_t pos = *ppos;
6152 	struct ctl_table lctl;
6153 	int ret;
6154 
6155 	/*
6156 	 * ctl->data points to idev->cnf.forwarding, we should
6157 	 * not modify it until we get the rtnl lock.
6158 	 */
6159 	lctl = *ctl;
6160 	lctl.data = &val;
6161 
6162 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6163 
6164 	if (write)
6165 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6166 	if (ret)
6167 		*ppos = pos;
6168 	return ret;
6169 }
6170 
addrconf_sysctl_mtu(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6171 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6172 		void *buffer, size_t *lenp, loff_t *ppos)
6173 {
6174 	struct inet6_dev *idev = ctl->extra1;
6175 	int min_mtu = IPV6_MIN_MTU;
6176 	struct ctl_table lctl;
6177 
6178 	lctl = *ctl;
6179 	lctl.extra1 = &min_mtu;
6180 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6181 
6182 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6183 }
6184 
dev_disable_change(struct inet6_dev *idev)6185 static void dev_disable_change(struct inet6_dev *idev)
6186 {
6187 	struct netdev_notifier_info info;
6188 
6189 	if (!idev || !idev->dev)
6190 		return;
6191 
6192 	netdev_notifier_info_init(&info, idev->dev);
6193 	if (idev->cnf.disable_ipv6)
6194 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6195 	else
6196 		addrconf_notify(NULL, NETDEV_UP, &info);
6197 }
6198 
addrconf_disable_change(struct net *net, __s32 newf)6199 static void addrconf_disable_change(struct net *net, __s32 newf)
6200 {
6201 	struct net_device *dev;
6202 	struct inet6_dev *idev;
6203 
6204 	for_each_netdev(net, dev) {
6205 		idev = __in6_dev_get(dev);
6206 		if (idev) {
6207 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6208 			idev->cnf.disable_ipv6 = newf;
6209 			if (changed)
6210 				dev_disable_change(idev);
6211 		}
6212 	}
6213 }
6214 
addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)6215 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6216 {
6217 	struct net *net;
6218 	int old;
6219 
6220 	if (!rtnl_trylock())
6221 		return restart_syscall();
6222 
6223 	net = (struct net *)table->extra2;
6224 	old = *p;
6225 	*p = newf;
6226 
6227 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6228 		rtnl_unlock();
6229 		return 0;
6230 	}
6231 
6232 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6233 		net->ipv6.devconf_dflt->disable_ipv6 = newf;
6234 		addrconf_disable_change(net, newf);
6235 	} else if ((!newf) ^ (!old))
6236 		dev_disable_change((struct inet6_dev *)table->extra1);
6237 
6238 	rtnl_unlock();
6239 	return 0;
6240 }
6241 
addrconf_sysctl_disable(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6242 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6243 		void *buffer, size_t *lenp, loff_t *ppos)
6244 {
6245 	int *valp = ctl->data;
6246 	int val = *valp;
6247 	loff_t pos = *ppos;
6248 	struct ctl_table lctl;
6249 	int ret;
6250 
6251 	/*
6252 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6253 	 * not modify it until we get the rtnl lock.
6254 	 */
6255 	lctl = *ctl;
6256 	lctl.data = &val;
6257 
6258 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6259 
6260 	if (write)
6261 		ret = addrconf_disable_ipv6(ctl, valp, val);
6262 	if (ret)
6263 		*ppos = pos;
6264 	return ret;
6265 }
6266 
addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6267 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6268 		void *buffer, size_t *lenp, loff_t *ppos)
6269 {
6270 	int *valp = ctl->data;
6271 	int ret;
6272 	int old, new;
6273 
6274 	old = *valp;
6275 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6276 	new = *valp;
6277 
6278 	if (write && old != new) {
6279 		struct net *net = ctl->extra2;
6280 
6281 		if (!rtnl_trylock())
6282 			return restart_syscall();
6283 
6284 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6285 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6286 						     NETCONFA_PROXY_NEIGH,
6287 						     NETCONFA_IFINDEX_DEFAULT,
6288 						     net->ipv6.devconf_dflt);
6289 		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6290 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6291 						     NETCONFA_PROXY_NEIGH,
6292 						     NETCONFA_IFINDEX_ALL,
6293 						     net->ipv6.devconf_all);
6294 		else {
6295 			struct inet6_dev *idev = ctl->extra1;
6296 
6297 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6298 						     NETCONFA_PROXY_NEIGH,
6299 						     idev->dev->ifindex,
6300 						     &idev->cnf);
6301 		}
6302 		rtnl_unlock();
6303 	}
6304 
6305 	return ret;
6306 }
6307 
addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6308 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6309 					 void *buffer, size_t *lenp,
6310 					 loff_t *ppos)
6311 {
6312 	int ret = 0;
6313 	u32 new_val;
6314 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6315 	struct net *net = (struct net *)ctl->extra2;
6316 	struct ctl_table tmp = {
6317 		.data = &new_val,
6318 		.maxlen = sizeof(new_val),
6319 		.mode = ctl->mode,
6320 	};
6321 
6322 	if (!rtnl_trylock())
6323 		return restart_syscall();
6324 
6325 	new_val = *((u32 *)ctl->data);
6326 
6327 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6328 	if (ret != 0)
6329 		goto out;
6330 
6331 	if (write) {
6332 		if (check_addr_gen_mode(new_val) < 0) {
6333 			ret = -EINVAL;
6334 			goto out;
6335 		}
6336 
6337 		if (idev) {
6338 			if (check_stable_privacy(idev, net, new_val) < 0) {
6339 				ret = -EINVAL;
6340 				goto out;
6341 			}
6342 
6343 			if (idev->cnf.addr_gen_mode != new_val) {
6344 				idev->cnf.addr_gen_mode = new_val;
6345 				addrconf_dev_config(idev->dev);
6346 			}
6347 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6348 			struct net_device *dev;
6349 
6350 			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6351 			for_each_netdev(net, dev) {
6352 				idev = __in6_dev_get(dev);
6353 				if (idev &&
6354 				    idev->cnf.addr_gen_mode != new_val) {
6355 					idev->cnf.addr_gen_mode = new_val;
6356 					addrconf_dev_config(idev->dev);
6357 				}
6358 			}
6359 		}
6360 
6361 		*((u32 *)ctl->data) = new_val;
6362 	}
6363 
6364 out:
6365 	rtnl_unlock();
6366 
6367 	return ret;
6368 }
6369 
addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6370 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6371 					 void *buffer, size_t *lenp,
6372 					 loff_t *ppos)
6373 {
6374 	int err;
6375 	struct in6_addr addr;
6376 	char str[IPV6_MAX_STRLEN];
6377 	struct ctl_table lctl = *ctl;
6378 	struct net *net = ctl->extra2;
6379 	struct ipv6_stable_secret *secret = ctl->data;
6380 
6381 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6382 		return -EIO;
6383 
6384 	lctl.maxlen = IPV6_MAX_STRLEN;
6385 	lctl.data = str;
6386 
6387 	if (!rtnl_trylock())
6388 		return restart_syscall();
6389 
6390 	if (!write && !secret->initialized) {
6391 		err = -EIO;
6392 		goto out;
6393 	}
6394 
6395 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6396 	if (err >= sizeof(str)) {
6397 		err = -EIO;
6398 		goto out;
6399 	}
6400 
6401 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6402 	if (err || !write)
6403 		goto out;
6404 
6405 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6406 		err = -EIO;
6407 		goto out;
6408 	}
6409 
6410 	secret->initialized = true;
6411 	secret->secret = addr;
6412 
6413 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6414 		struct net_device *dev;
6415 
6416 		for_each_netdev(net, dev) {
6417 			struct inet6_dev *idev = __in6_dev_get(dev);
6418 
6419 			if (idev) {
6420 				idev->cnf.addr_gen_mode =
6421 					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6422 			}
6423 		}
6424 	} else {
6425 		struct inet6_dev *idev = ctl->extra1;
6426 
6427 		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6428 	}
6429 
6430 out:
6431 	rtnl_unlock();
6432 
6433 	return err;
6434 }
6435 
6436 static
addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6437 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6438 						int write, void *buffer,
6439 						size_t *lenp,
6440 						loff_t *ppos)
6441 {
6442 	int *valp = ctl->data;
6443 	int val = *valp;
6444 	loff_t pos = *ppos;
6445 	struct ctl_table lctl;
6446 	int ret;
6447 
6448 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6449 	 * we should not modify it until we get the rtnl lock.
6450 	 */
6451 	lctl = *ctl;
6452 	lctl.data = &val;
6453 
6454 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6455 
6456 	if (write)
6457 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6458 	if (ret)
6459 		*ppos = pos;
6460 	return ret;
6461 }
6462 
6463 static
addrconf_set_nopolicy(struct rt6_info *rt, int action)6464 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6465 {
6466 	if (rt) {
6467 		if (action)
6468 			rt->dst.flags |= DST_NOPOLICY;
6469 		else
6470 			rt->dst.flags &= ~DST_NOPOLICY;
6471 	}
6472 }
6473 
6474 static
addrconf_disable_policy_idev(struct inet6_dev *idev, int val)6475 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6476 {
6477 	struct inet6_ifaddr *ifa;
6478 
6479 	read_lock_bh(&idev->lock);
6480 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6481 		spin_lock(&ifa->lock);
6482 		if (ifa->rt) {
6483 			/* host routes only use builtin fib6_nh */
6484 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6485 			int cpu;
6486 
6487 			rcu_read_lock();
6488 			ifa->rt->dst_nopolicy = val ? true : false;
6489 			if (nh->rt6i_pcpu) {
6490 				for_each_possible_cpu(cpu) {
6491 					struct rt6_info **rtp;
6492 
6493 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6494 					addrconf_set_nopolicy(*rtp, val);
6495 				}
6496 			}
6497 			rcu_read_unlock();
6498 		}
6499 		spin_unlock(&ifa->lock);
6500 	}
6501 	read_unlock_bh(&idev->lock);
6502 }
6503 
6504 static
addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)6505 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6506 {
6507 	struct inet6_dev *idev;
6508 	struct net *net;
6509 
6510 	if (!rtnl_trylock())
6511 		return restart_syscall();
6512 
6513 	*valp = val;
6514 
6515 	net = (struct net *)ctl->extra2;
6516 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6517 		rtnl_unlock();
6518 		return 0;
6519 	}
6520 
6521 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6522 		struct net_device *dev;
6523 
6524 		for_each_netdev(net, dev) {
6525 			idev = __in6_dev_get(dev);
6526 			if (idev)
6527 				addrconf_disable_policy_idev(idev, val);
6528 		}
6529 	} else {
6530 		idev = (struct inet6_dev *)ctl->extra1;
6531 		addrconf_disable_policy_idev(idev, val);
6532 	}
6533 
6534 	rtnl_unlock();
6535 	return 0;
6536 }
6537 
addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write, void *buffer, size_t *lenp, loff_t *ppos)6538 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6539 				   void *buffer, size_t *lenp, loff_t *ppos)
6540 {
6541 	int *valp = ctl->data;
6542 	int val = *valp;
6543 	loff_t pos = *ppos;
6544 	struct ctl_table lctl;
6545 	int ret;
6546 
6547 	lctl = *ctl;
6548 	lctl.data = &val;
6549 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6550 
6551 	if (write && (*valp != val))
6552 		ret = addrconf_disable_policy(ctl, valp, val);
6553 
6554 	if (ret)
6555 		*ppos = pos;
6556 
6557 	return ret;
6558 }
6559 
6560 static int minus_one = -1;
6561 static const int two_five_five = 255;
6562 
6563 static const struct ctl_table addrconf_sysctl[] = {
6564 	{
6565 		.procname	= "forwarding",
6566 		.data		= &ipv6_devconf.forwarding,
6567 		.maxlen		= sizeof(int),
6568 		.mode		= 0644,
6569 		.proc_handler	= addrconf_sysctl_forward,
6570 	},
6571 	{
6572 		.procname	= "hop_limit",
6573 		.data		= &ipv6_devconf.hop_limit,
6574 		.maxlen		= sizeof(int),
6575 		.mode		= 0644,
6576 		.proc_handler	= proc_dointvec_minmax,
6577 		.extra1		= (void *)SYSCTL_ONE,
6578 		.extra2		= (void *)&two_five_five,
6579 	},
6580 	{
6581 		.procname	= "mtu",
6582 		.data		= &ipv6_devconf.mtu6,
6583 		.maxlen		= sizeof(int),
6584 		.mode		= 0644,
6585 		.proc_handler	= addrconf_sysctl_mtu,
6586 	},
6587 	{
6588 		.procname	= "accept_ra",
6589 		.data		= &ipv6_devconf.accept_ra,
6590 		.maxlen		= sizeof(int),
6591 		.mode		= 0644,
6592 		.proc_handler	= proc_dointvec,
6593 	},
6594 	{
6595 		.procname	= "accept_redirects",
6596 		.data		= &ipv6_devconf.accept_redirects,
6597 		.maxlen		= sizeof(int),
6598 		.mode		= 0644,
6599 		.proc_handler	= proc_dointvec,
6600 	},
6601 	{
6602 		.procname	= "autoconf",
6603 		.data		= &ipv6_devconf.autoconf,
6604 		.maxlen		= sizeof(int),
6605 		.mode		= 0644,
6606 		.proc_handler	= proc_dointvec,
6607 	},
6608 	{
6609 		.procname	= "dad_transmits",
6610 		.data		= &ipv6_devconf.dad_transmits,
6611 		.maxlen		= sizeof(int),
6612 		.mode		= 0644,
6613 		.proc_handler	= proc_dointvec,
6614 	},
6615 	{
6616 		.procname	= "router_solicitations",
6617 		.data		= &ipv6_devconf.rtr_solicits,
6618 		.maxlen		= sizeof(int),
6619 		.mode		= 0644,
6620 		.proc_handler	= proc_dointvec_minmax,
6621 		.extra1		= &minus_one,
6622 	},
6623 	{
6624 		.procname	= "router_solicitation_interval",
6625 		.data		= &ipv6_devconf.rtr_solicit_interval,
6626 		.maxlen		= sizeof(int),
6627 		.mode		= 0644,
6628 		.proc_handler	= proc_dointvec_jiffies,
6629 	},
6630 	{
6631 		.procname	= "router_solicitation_max_interval",
6632 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6633 		.maxlen		= sizeof(int),
6634 		.mode		= 0644,
6635 		.proc_handler	= proc_dointvec_jiffies,
6636 	},
6637 	{
6638 		.procname	= "router_solicitation_delay",
6639 		.data		= &ipv6_devconf.rtr_solicit_delay,
6640 		.maxlen		= sizeof(int),
6641 		.mode		= 0644,
6642 		.proc_handler	= proc_dointvec_jiffies,
6643 	},
6644 	{
6645 		.procname	= "force_mld_version",
6646 		.data		= &ipv6_devconf.force_mld_version,
6647 		.maxlen		= sizeof(int),
6648 		.mode		= 0644,
6649 		.proc_handler	= proc_dointvec,
6650 	},
6651 	{
6652 		.procname	= "mldv1_unsolicited_report_interval",
6653 		.data		=
6654 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6655 		.maxlen		= sizeof(int),
6656 		.mode		= 0644,
6657 		.proc_handler	= proc_dointvec_ms_jiffies,
6658 	},
6659 	{
6660 		.procname	= "mldv2_unsolicited_report_interval",
6661 		.data		=
6662 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6663 		.maxlen		= sizeof(int),
6664 		.mode		= 0644,
6665 		.proc_handler	= proc_dointvec_ms_jiffies,
6666 	},
6667 	{
6668 		.procname	= "use_tempaddr",
6669 		.data		= &ipv6_devconf.use_tempaddr,
6670 		.maxlen		= sizeof(int),
6671 		.mode		= 0644,
6672 		.proc_handler	= proc_dointvec,
6673 	},
6674 	{
6675 		.procname	= "temp_valid_lft",
6676 		.data		= &ipv6_devconf.temp_valid_lft,
6677 		.maxlen		= sizeof(int),
6678 		.mode		= 0644,
6679 		.proc_handler	= proc_dointvec,
6680 	},
6681 	{
6682 		.procname	= "temp_prefered_lft",
6683 		.data		= &ipv6_devconf.temp_prefered_lft,
6684 		.maxlen		= sizeof(int),
6685 		.mode		= 0644,
6686 		.proc_handler	= proc_dointvec,
6687 	},
6688 	{
6689 		.procname	= "regen_max_retry",
6690 		.data		= &ipv6_devconf.regen_max_retry,
6691 		.maxlen		= sizeof(int),
6692 		.mode		= 0644,
6693 		.proc_handler	= proc_dointvec,
6694 	},
6695 	{
6696 		.procname	= "max_desync_factor",
6697 		.data		= &ipv6_devconf.max_desync_factor,
6698 		.maxlen		= sizeof(int),
6699 		.mode		= 0644,
6700 		.proc_handler	= proc_dointvec,
6701 	},
6702 	{
6703 		.procname	= "max_addresses",
6704 		.data		= &ipv6_devconf.max_addresses,
6705 		.maxlen		= sizeof(int),
6706 		.mode		= 0644,
6707 		.proc_handler	= proc_dointvec,
6708 	},
6709 	{
6710 		.procname	= "accept_ra_defrtr",
6711 		.data		= &ipv6_devconf.accept_ra_defrtr,
6712 		.maxlen		= sizeof(int),
6713 		.mode		= 0644,
6714 		.proc_handler	= proc_dointvec,
6715 	},
6716 	{
6717 		.procname	= "accept_ra_min_hop_limit",
6718 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6719 		.maxlen		= sizeof(int),
6720 		.mode		= 0644,
6721 		.proc_handler	= proc_dointvec,
6722 	},
6723 	{
6724 		.procname	= "accept_ra_min_lft",
6725 		.data		= &ipv6_devconf.accept_ra_min_lft,
6726 		.maxlen		= sizeof(int),
6727 		.mode		= 0644,
6728 		.proc_handler	= proc_dointvec,
6729 	},
6730 	{
6731 		.procname	= "accept_ra_pinfo",
6732 		.data		= &ipv6_devconf.accept_ra_pinfo,
6733 		.maxlen		= sizeof(int),
6734 		.mode		= 0644,
6735 		.proc_handler	= proc_dointvec,
6736 	},
6737 #ifdef CONFIG_IPV6_ROUTER_PREF
6738 	{
6739 		.procname	= "accept_ra_rtr_pref",
6740 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6741 		.maxlen		= sizeof(int),
6742 		.mode		= 0644,
6743 		.proc_handler	= proc_dointvec,
6744 	},
6745 	{
6746 		.procname	= "router_probe_interval",
6747 		.data		= &ipv6_devconf.rtr_probe_interval,
6748 		.maxlen		= sizeof(int),
6749 		.mode		= 0644,
6750 		.proc_handler	= proc_dointvec_jiffies,
6751 	},
6752 #ifdef CONFIG_IPV6_ROUTE_INFO
6753 	{
6754 		.procname	= "accept_ra_rt_info_min_plen",
6755 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6756 		.maxlen		= sizeof(int),
6757 		.mode		= 0644,
6758 		.proc_handler	= proc_dointvec,
6759 	},
6760 	{
6761 		.procname	= "accept_ra_rt_info_max_plen",
6762 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6763 		.maxlen		= sizeof(int),
6764 		.mode		= 0644,
6765 		.proc_handler	= proc_dointvec,
6766 	},
6767 #endif
6768 #endif
6769 	{
6770 		.procname	= "proxy_ndp",
6771 		.data		= &ipv6_devconf.proxy_ndp,
6772 		.maxlen		= sizeof(int),
6773 		.mode		= 0644,
6774 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6775 	},
6776 	{
6777 		.procname	= "accept_source_route",
6778 		.data		= &ipv6_devconf.accept_source_route,
6779 		.maxlen		= sizeof(int),
6780 		.mode		= 0644,
6781 		.proc_handler	= proc_dointvec,
6782 	},
6783 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6784 	{
6785 		.procname	= "optimistic_dad",
6786 		.data		= &ipv6_devconf.optimistic_dad,
6787 		.maxlen		= sizeof(int),
6788 		.mode		= 0644,
6789 		.proc_handler   = proc_dointvec,
6790 	},
6791 	{
6792 		.procname	= "use_optimistic",
6793 		.data		= &ipv6_devconf.use_optimistic,
6794 		.maxlen		= sizeof(int),
6795 		.mode		= 0644,
6796 		.proc_handler	= proc_dointvec,
6797 	},
6798 #endif
6799 #ifdef CONFIG_IPV6_MROUTE
6800 	{
6801 		.procname	= "mc_forwarding",
6802 		.data		= &ipv6_devconf.mc_forwarding,
6803 		.maxlen		= sizeof(int),
6804 		.mode		= 0444,
6805 		.proc_handler	= proc_dointvec,
6806 	},
6807 #endif
6808 	{
6809 		.procname	= "disable_ipv6",
6810 		.data		= &ipv6_devconf.disable_ipv6,
6811 		.maxlen		= sizeof(int),
6812 		.mode		= 0644,
6813 		.proc_handler	= addrconf_sysctl_disable,
6814 	},
6815 	{
6816 		.procname	= "accept_dad",
6817 		.data		= &ipv6_devconf.accept_dad,
6818 		.maxlen		= sizeof(int),
6819 		.mode		= 0644,
6820 		.proc_handler	= proc_dointvec,
6821 	},
6822 	{
6823 		.procname	= "force_tllao",
6824 		.data		= &ipv6_devconf.force_tllao,
6825 		.maxlen		= sizeof(int),
6826 		.mode		= 0644,
6827 		.proc_handler	= proc_dointvec
6828 	},
6829 	{
6830 		.procname	= "ndisc_notify",
6831 		.data		= &ipv6_devconf.ndisc_notify,
6832 		.maxlen		= sizeof(int),
6833 		.mode		= 0644,
6834 		.proc_handler	= proc_dointvec
6835 	},
6836 	{
6837 		.procname	= "suppress_frag_ndisc",
6838 		.data		= &ipv6_devconf.suppress_frag_ndisc,
6839 		.maxlen		= sizeof(int),
6840 		.mode		= 0644,
6841 		.proc_handler	= proc_dointvec
6842 	},
6843 	{
6844 		.procname	= "accept_ra_from_local",
6845 		.data		= &ipv6_devconf.accept_ra_from_local,
6846 		.maxlen		= sizeof(int),
6847 		.mode		= 0644,
6848 		.proc_handler	= proc_dointvec,
6849 	},
6850 	{
6851 		.procname	= "accept_ra_mtu",
6852 		.data		= &ipv6_devconf.accept_ra_mtu,
6853 		.maxlen		= sizeof(int),
6854 		.mode		= 0644,
6855 		.proc_handler	= proc_dointvec,
6856 	},
6857 	{
6858 		.procname	= "stable_secret",
6859 		.data		= &ipv6_devconf.stable_secret,
6860 		.maxlen		= IPV6_MAX_STRLEN,
6861 		.mode		= 0600,
6862 		.proc_handler	= addrconf_sysctl_stable_secret,
6863 	},
6864 	{
6865 		.procname	= "use_oif_addrs_only",
6866 		.data		= &ipv6_devconf.use_oif_addrs_only,
6867 		.maxlen		= sizeof(int),
6868 		.mode		= 0644,
6869 		.proc_handler	= proc_dointvec,
6870 	},
6871 	{
6872 		.procname	= "ignore_routes_with_linkdown",
6873 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
6874 		.maxlen		= sizeof(int),
6875 		.mode		= 0644,
6876 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
6877 	},
6878 	{
6879 		.procname	= "drop_unicast_in_l2_multicast",
6880 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
6881 		.maxlen		= sizeof(int),
6882 		.mode		= 0644,
6883 		.proc_handler	= proc_dointvec,
6884 	},
6885 	{
6886 		.procname	= "drop_unsolicited_na",
6887 		.data		= &ipv6_devconf.drop_unsolicited_na,
6888 		.maxlen		= sizeof(int),
6889 		.mode		= 0644,
6890 		.proc_handler	= proc_dointvec,
6891 	},
6892 	{
6893 		.procname	= "keep_addr_on_down",
6894 		.data		= &ipv6_devconf.keep_addr_on_down,
6895 		.maxlen		= sizeof(int),
6896 		.mode		= 0644,
6897 		.proc_handler	= proc_dointvec,
6898 
6899 	},
6900 	{
6901 		.procname	= "seg6_enabled",
6902 		.data		= &ipv6_devconf.seg6_enabled,
6903 		.maxlen		= sizeof(int),
6904 		.mode		= 0644,
6905 		.proc_handler	= proc_dointvec,
6906 	},
6907 #ifdef CONFIG_IPV6_SEG6_HMAC
6908 	{
6909 		.procname	= "seg6_require_hmac",
6910 		.data		= &ipv6_devconf.seg6_require_hmac,
6911 		.maxlen		= sizeof(int),
6912 		.mode		= 0644,
6913 		.proc_handler	= proc_dointvec,
6914 	},
6915 #endif
6916 	{
6917 		.procname       = "enhanced_dad",
6918 		.data           = &ipv6_devconf.enhanced_dad,
6919 		.maxlen         = sizeof(int),
6920 		.mode           = 0644,
6921 		.proc_handler   = proc_dointvec,
6922 	},
6923 	{
6924 		.procname		= "addr_gen_mode",
6925 		.data			= &ipv6_devconf.addr_gen_mode,
6926 		.maxlen			= sizeof(int),
6927 		.mode			= 0644,
6928 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
6929 	},
6930 	{
6931 		.procname       = "disable_policy",
6932 		.data           = &ipv6_devconf.disable_policy,
6933 		.maxlen         = sizeof(int),
6934 		.mode           = 0644,
6935 		.proc_handler   = addrconf_sysctl_disable_policy,
6936 	},
6937 	{
6938 		.procname	= "ndisc_tclass",
6939 		.data		= &ipv6_devconf.ndisc_tclass,
6940 		.maxlen		= sizeof(int),
6941 		.mode		= 0644,
6942 		.proc_handler	= proc_dointvec_minmax,
6943 		.extra1		= (void *)SYSCTL_ZERO,
6944 		.extra2		= (void *)&two_five_five,
6945 	},
6946 	{
6947 		.procname	= "rpl_seg_enabled",
6948 		.data		= &ipv6_devconf.rpl_seg_enabled,
6949 		.maxlen		= sizeof(int),
6950 		.mode		= 0644,
6951 		.proc_handler	= proc_dointvec,
6952 	},
6953 	{
6954 		/* sentinel */
6955 	}
6956 };
6957 
__addrconf_sysctl_register(struct net *net, char *dev_name, struct inet6_dev *idev, struct ipv6_devconf *p)6958 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6959 		struct inet6_dev *idev, struct ipv6_devconf *p)
6960 {
6961 	int i, ifindex;
6962 	struct ctl_table *table;
6963 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6964 
6965 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6966 	if (!table)
6967 		goto out;
6968 
6969 	for (i = 0; table[i].data; i++) {
6970 		table[i].data += (char *)p - (char *)&ipv6_devconf;
6971 		/* If one of these is already set, then it is not safe to
6972 		 * overwrite either of them: this makes proc_dointvec_minmax
6973 		 * usable.
6974 		 */
6975 		if (!table[i].extra1 && !table[i].extra2) {
6976 			table[i].extra1 = idev; /* embedded; no ref */
6977 			table[i].extra2 = net;
6978 		}
6979 	}
6980 
6981 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6982 
6983 	p->sysctl_header = register_net_sysctl(net, path, table);
6984 	if (!p->sysctl_header)
6985 		goto free;
6986 
6987 	if (!strcmp(dev_name, "all"))
6988 		ifindex = NETCONFA_IFINDEX_ALL;
6989 	else if (!strcmp(dev_name, "default"))
6990 		ifindex = NETCONFA_IFINDEX_DEFAULT;
6991 	else
6992 		ifindex = idev->dev->ifindex;
6993 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6994 				     ifindex, p);
6995 	return 0;
6996 
6997 free:
6998 	kfree(table);
6999 out:
7000 	return -ENOBUFS;
7001 }
7002 
__addrconf_sysctl_unregister(struct net *net, struct ipv6_devconf *p, int ifindex)7003 static void __addrconf_sysctl_unregister(struct net *net,
7004 					 struct ipv6_devconf *p, int ifindex)
7005 {
7006 	struct ctl_table *table;
7007 
7008 	if (!p->sysctl_header)
7009 		return;
7010 
7011 	table = p->sysctl_header->ctl_table_arg;
7012 	unregister_net_sysctl_table(p->sysctl_header);
7013 	p->sysctl_header = NULL;
7014 	kfree(table);
7015 
7016 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7017 }
7018 
addrconf_sysctl_register(struct inet6_dev *idev)7019 static int addrconf_sysctl_register(struct inet6_dev *idev)
7020 {
7021 	int err;
7022 
7023 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7024 		return -EINVAL;
7025 
7026 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7027 				    &ndisc_ifinfo_sysctl_change);
7028 	if (err)
7029 		return err;
7030 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7031 					 idev, &idev->cnf);
7032 	if (err)
7033 		neigh_sysctl_unregister(idev->nd_parms);
7034 
7035 	return err;
7036 }
7037 
addrconf_sysctl_unregister(struct inet6_dev *idev)7038 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7039 {
7040 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7041 				     idev->dev->ifindex);
7042 	neigh_sysctl_unregister(idev->nd_parms);
7043 }
7044 
7045 
7046 #endif
7047 
addrconf_init_net(struct net *net)7048 static int __net_init addrconf_init_net(struct net *net)
7049 {
7050 	int err = -ENOMEM;
7051 	struct ipv6_devconf *all, *dflt;
7052 
7053 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7054 	if (!all)
7055 		goto err_alloc_all;
7056 
7057 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7058 	if (!dflt)
7059 		goto err_alloc_dflt;
7060 
7061 	if (!net_eq(net, &init_net)) {
7062 		switch (net_inherit_devconf()) {
7063 		case 1:  /* copy from init_net */
7064 			memcpy(all, init_net.ipv6.devconf_all,
7065 			       sizeof(ipv6_devconf));
7066 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7067 			       sizeof(ipv6_devconf_dflt));
7068 			break;
7069 		case 3: /* copy from the current netns */
7070 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7071 			       sizeof(ipv6_devconf));
7072 			memcpy(dflt,
7073 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7074 			       sizeof(ipv6_devconf_dflt));
7075 			break;
7076 		case 0:
7077 		case 2:
7078 			/* use compiled values */
7079 			break;
7080 		}
7081 	}
7082 
7083 	/* these will be inherited by all namespaces */
7084 	dflt->autoconf = ipv6_defaults.autoconf;
7085 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7086 
7087 	dflt->stable_secret.initialized = false;
7088 	all->stable_secret.initialized = false;
7089 
7090 	net->ipv6.devconf_all = all;
7091 	net->ipv6.devconf_dflt = dflt;
7092 
7093 #ifdef CONFIG_SYSCTL
7094 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7095 	if (err < 0)
7096 		goto err_reg_all;
7097 
7098 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7099 	if (err < 0)
7100 		goto err_reg_dflt;
7101 #endif
7102 	return 0;
7103 
7104 #ifdef CONFIG_SYSCTL
7105 err_reg_dflt:
7106 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7107 err_reg_all:
7108 	kfree(dflt);
7109 #endif
7110 err_alloc_dflt:
7111 	kfree(all);
7112 err_alloc_all:
7113 	return err;
7114 }
7115 
addrconf_exit_net(struct net *net)7116 static void __net_exit addrconf_exit_net(struct net *net)
7117 {
7118 #ifdef CONFIG_SYSCTL
7119 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7120 				     NETCONFA_IFINDEX_DEFAULT);
7121 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7122 				     NETCONFA_IFINDEX_ALL);
7123 #endif
7124 	kfree(net->ipv6.devconf_dflt);
7125 	kfree(net->ipv6.devconf_all);
7126 }
7127 
7128 static struct pernet_operations addrconf_ops = {
7129 	.init = addrconf_init_net,
7130 	.exit = addrconf_exit_net,
7131 };
7132 
7133 static struct rtnl_af_ops inet6_ops __read_mostly = {
7134 	.family		  = AF_INET6,
7135 	.fill_link_af	  = inet6_fill_link_af,
7136 	.get_link_af_size = inet6_get_link_af_size,
7137 	.validate_link_af = inet6_validate_link_af,
7138 	.set_link_af	  = inet6_set_link_af,
7139 };
7140 
7141 /*
7142  *	Init / cleanup code
7143  */
7144 
addrconf_init(void)7145 int __init addrconf_init(void)
7146 {
7147 	struct inet6_dev *idev;
7148 	int i, err;
7149 
7150 	err = ipv6_addr_label_init();
7151 	if (err < 0) {
7152 		pr_crit("%s: cannot initialize default policy table: %d\n",
7153 			__func__, err);
7154 		goto out;
7155 	}
7156 
7157 	err = register_pernet_subsys(&addrconf_ops);
7158 	if (err < 0)
7159 		goto out_addrlabel;
7160 
7161 	addrconf_wq = create_workqueue("ipv6_addrconf");
7162 	if (!addrconf_wq) {
7163 		err = -ENOMEM;
7164 		goto out_nowq;
7165 	}
7166 
7167 	/* The addrconf netdev notifier requires that loopback_dev
7168 	 * has it's ipv6 private information allocated and setup
7169 	 * before it can bring up and give link-local addresses
7170 	 * to other devices which are up.
7171 	 *
7172 	 * Unfortunately, loopback_dev is not necessarily the first
7173 	 * entry in the global dev_base list of net devices.  In fact,
7174 	 * it is likely to be the very last entry on that list.
7175 	 * So this causes the notifier registry below to try and
7176 	 * give link-local addresses to all devices besides loopback_dev
7177 	 * first, then loopback_dev, which cases all the non-loopback_dev
7178 	 * devices to fail to get a link-local address.
7179 	 *
7180 	 * So, as a temporary fix, allocate the ipv6 structure for
7181 	 * loopback_dev first by hand.
7182 	 * Longer term, all of the dependencies ipv6 has upon the loopback
7183 	 * device and it being up should be removed.
7184 	 */
7185 	rtnl_lock();
7186 	idev = ipv6_add_dev(init_net.loopback_dev);
7187 	rtnl_unlock();
7188 	if (IS_ERR(idev)) {
7189 		err = PTR_ERR(idev);
7190 		goto errlo;
7191 	}
7192 
7193 	ip6_route_init_special_entries();
7194 
7195 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7196 		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7197 
7198 	register_netdevice_notifier(&ipv6_dev_notf);
7199 
7200 	addrconf_verify();
7201 
7202 	rtnl_af_register(&inet6_ops);
7203 
7204 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7205 				   NULL, inet6_dump_ifinfo, 0);
7206 	if (err < 0)
7207 		goto errout;
7208 
7209 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7210 				   inet6_rtm_newaddr, NULL, 0);
7211 	if (err < 0)
7212 		goto errout;
7213 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7214 				   inet6_rtm_deladdr, NULL, 0);
7215 	if (err < 0)
7216 		goto errout;
7217 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7218 				   inet6_rtm_getaddr, inet6_dump_ifaddr,
7219 				   RTNL_FLAG_DOIT_UNLOCKED);
7220 	if (err < 0)
7221 		goto errout;
7222 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7223 				   NULL, inet6_dump_ifmcaddr, 0);
7224 	if (err < 0)
7225 		goto errout;
7226 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7227 				   NULL, inet6_dump_ifacaddr, 0);
7228 	if (err < 0)
7229 		goto errout;
7230 	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7231 				   inet6_netconf_get_devconf,
7232 				   inet6_netconf_dump_devconf,
7233 				   RTNL_FLAG_DOIT_UNLOCKED);
7234 	if (err < 0)
7235 		goto errout;
7236 	err = ipv6_addr_label_rtnl_register();
7237 	if (err < 0)
7238 		goto errout;
7239 
7240 	return 0;
7241 errout:
7242 	rtnl_unregister_all(PF_INET6);
7243 	rtnl_af_unregister(&inet6_ops);
7244 	unregister_netdevice_notifier(&ipv6_dev_notf);
7245 errlo:
7246 	destroy_workqueue(addrconf_wq);
7247 out_nowq:
7248 	unregister_pernet_subsys(&addrconf_ops);
7249 out_addrlabel:
7250 	ipv6_addr_label_cleanup();
7251 out:
7252 	return err;
7253 }
7254 
addrconf_cleanup(void)7255 void addrconf_cleanup(void)
7256 {
7257 	struct net_device *dev;
7258 	int i;
7259 
7260 	unregister_netdevice_notifier(&ipv6_dev_notf);
7261 	unregister_pernet_subsys(&addrconf_ops);
7262 	ipv6_addr_label_cleanup();
7263 
7264 	rtnl_af_unregister(&inet6_ops);
7265 
7266 	rtnl_lock();
7267 
7268 	/* clean dev list */
7269 	for_each_netdev(&init_net, dev) {
7270 		if (__in6_dev_get(dev) == NULL)
7271 			continue;
7272 		addrconf_ifdown(dev, true);
7273 	}
7274 	addrconf_ifdown(init_net.loopback_dev, true);
7275 
7276 	/*
7277 	 *	Check hash table.
7278 	 */
7279 	spin_lock_bh(&addrconf_hash_lock);
7280 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7281 		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7282 	spin_unlock_bh(&addrconf_hash_lock);
7283 	cancel_delayed_work(&addr_chk_work);
7284 	rtnl_unlock();
7285 
7286 	destroy_workqueue(addrconf_wq);
7287 }
7288