xref: /linux/net/ipv6/addrconf.c (revision ab59a8605604f71bbbc16077270dc3f39648b7fc)
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/ip_tunnels.h>
67 #include <net/net_namespace.h>
68 #include <net/sock.h>
69 #include <net/snmp.h>
70 
71 #include <net/6lowpan.h>
72 #include <net/firewire.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <net/l3mdev.h>
83 #include <net/netdev_lock.h>
84 #include <linux/if_tunnel.h>
85 #include <linux/rtnetlink.h>
86 #include <linux/netconf.h>
87 #include <linux/random.h>
88 #include <linux/uaccess.h>
89 #include <linux/unaligned.h>
90 
91 #include <linux/proc_fs.h>
92 #include <linux/seq_file.h>
93 #include <linux/export.h>
94 #include <linux/ioam6.h>
95 
96 #define IPV6_MAX_STRLEN \
97 	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98 
cstamp_delta(unsigned long cstamp)99 static inline u32 cstamp_delta(unsigned long cstamp)
100 {
101 	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102 }
103 
rfc3315_s14_backoff_init(s32 irt)104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 {
106 	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 	u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
108 	do_div(tmp, 1000000);
109 	return (s32)tmp;
110 }
111 
rfc3315_s14_backoff_update(s32 rt,s32 mrt)112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 {
114 	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 	u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116 	do_div(tmp, 1000000);
117 	if ((s32)tmp > mrt) {
118 		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 		tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
120 		do_div(tmp, 1000000);
121 	}
122 	return (s32)tmp;
123 }
124 
125 #ifdef CONFIG_SYSCTL
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 #else
addrconf_sysctl_register(struct inet6_dev * idev)129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 {
131 	return 0;
132 }
133 
addrconf_sysctl_unregister(struct inet6_dev * idev)134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135 {
136 }
137 #endif
138 
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140 
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 					u8 dad_count,
145 					const struct inet6_dev *idev);
146 
147 #define IN6_ADDR_HSIZE_SHIFT	8
148 #define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
149 
150 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
152 
153 static struct workqueue_struct *addrconf_wq;
154 
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157 
158 static void addrconf_type_change(struct net_device *dev,
159 				 unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
161 
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163 						  int plen,
164 						  const struct net_device *dev,
165 						  u32 flags, u32 noflags,
166 						  bool no_gw);
167 
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171 				   bool send_na);
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176 
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 				struct prefix_info *pinfo);
179 
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
181 	.forwarding		= 0,
182 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
183 	.mtu6			= IPV6_MIN_MTU,
184 	.accept_ra		= 1,
185 	.accept_redirects	= 1,
186 	.autoconf		= 1,
187 	.force_mld_version	= 0,
188 	.mldv1_unsolicited_report_interval = 10 * HZ,
189 	.mldv2_unsolicited_report_interval = HZ,
190 	.dad_transmits		= 1,
191 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
192 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
193 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
195 	.use_tempaddr		= 0,
196 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
197 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
198 	.regen_min_advance	= REGEN_MIN_ADVANCE,
199 	.regen_max_retry	= REGEN_MAX_RETRY,
200 	.max_desync_factor	= MAX_DESYNC_FACTOR,
201 	.max_addresses		= IPV6_MAX_ADDRESSES,
202 	.accept_ra_defrtr	= 1,
203 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
204 	.accept_ra_from_local	= 0,
205 	.accept_ra_min_hop_limit= 1,
206 	.accept_ra_min_lft	= 0,
207 	.accept_ra_pinfo	= 1,
208 #ifdef CONFIG_IPV6_ROUTER_PREF
209 	.accept_ra_rtr_pref	= 1,
210 	.rtr_probe_interval	= 60 * HZ,
211 #ifdef CONFIG_IPV6_ROUTE_INFO
212 	.accept_ra_rt_info_min_plen = 0,
213 	.accept_ra_rt_info_max_plen = 0,
214 #endif
215 #endif
216 	.proxy_ndp		= 0,
217 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
218 	.disable_ipv6		= 0,
219 	.accept_dad		= 0,
220 	.suppress_frag_ndisc	= 1,
221 	.accept_ra_mtu		= 1,
222 	.stable_secret		= {
223 		.initialized = false,
224 	},
225 	.use_oif_addrs_only	= 0,
226 	.ignore_routes_with_linkdown = 0,
227 	.keep_addr_on_down	= 0,
228 	.seg6_enabled		= 0,
229 #ifdef CONFIG_IPV6_SEG6_HMAC
230 	.seg6_require_hmac	= 0,
231 #endif
232 	.enhanced_dad           = 1,
233 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
234 	.disable_policy		= 0,
235 	.rpl_seg_enabled	= 0,
236 	.ioam6_enabled		= 0,
237 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
238 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
239 	.ndisc_evict_nocarrier	= 1,
240 	.ra_honor_pio_life	= 0,
241 	.ra_honor_pio_pflag	= 0,
242 };
243 
244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
245 	.forwarding		= 0,
246 	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
247 	.mtu6			= IPV6_MIN_MTU,
248 	.accept_ra		= 1,
249 	.accept_redirects	= 1,
250 	.autoconf		= 1,
251 	.force_mld_version	= 0,
252 	.mldv1_unsolicited_report_interval = 10 * HZ,
253 	.mldv2_unsolicited_report_interval = HZ,
254 	.dad_transmits		= 1,
255 	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
256 	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
257 	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
258 	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
259 	.use_tempaddr		= 0,
260 	.temp_valid_lft		= TEMP_VALID_LIFETIME,
261 	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
262 	.regen_min_advance	= REGEN_MIN_ADVANCE,
263 	.regen_max_retry	= REGEN_MAX_RETRY,
264 	.max_desync_factor	= MAX_DESYNC_FACTOR,
265 	.max_addresses		= IPV6_MAX_ADDRESSES,
266 	.accept_ra_defrtr	= 1,
267 	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
268 	.accept_ra_from_local	= 0,
269 	.accept_ra_min_hop_limit= 1,
270 	.accept_ra_min_lft	= 0,
271 	.accept_ra_pinfo	= 1,
272 #ifdef CONFIG_IPV6_ROUTER_PREF
273 	.accept_ra_rtr_pref	= 1,
274 	.rtr_probe_interval	= 60 * HZ,
275 #ifdef CONFIG_IPV6_ROUTE_INFO
276 	.accept_ra_rt_info_min_plen = 0,
277 	.accept_ra_rt_info_max_plen = 0,
278 #endif
279 #endif
280 	.proxy_ndp		= 0,
281 	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
282 	.disable_ipv6		= 0,
283 	.accept_dad		= 1,
284 	.suppress_frag_ndisc	= 1,
285 	.accept_ra_mtu		= 1,
286 	.stable_secret		= {
287 		.initialized = false,
288 	},
289 	.use_oif_addrs_only	= 0,
290 	.ignore_routes_with_linkdown = 0,
291 	.keep_addr_on_down	= 0,
292 	.seg6_enabled		= 0,
293 #ifdef CONFIG_IPV6_SEG6_HMAC
294 	.seg6_require_hmac	= 0,
295 #endif
296 	.enhanced_dad           = 1,
297 	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
298 	.disable_policy		= 0,
299 	.rpl_seg_enabled	= 0,
300 	.ioam6_enabled		= 0,
301 	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
302 	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
303 	.ndisc_evict_nocarrier	= 1,
304 	.ra_honor_pio_life	= 0,
305 	.ra_honor_pio_pflag	= 0,
306 };
307 
308 /* Check if link is ready: is it up and is a valid qdisc available */
addrconf_link_ready(const struct net_device * dev)309 static inline bool addrconf_link_ready(const struct net_device *dev)
310 {
311 	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
312 }
313 
addrconf_del_rs_timer(struct inet6_dev * idev)314 static void addrconf_del_rs_timer(struct inet6_dev *idev)
315 {
316 	if (timer_delete(&idev->rs_timer))
317 		__in6_dev_put(idev);
318 }
319 
addrconf_del_dad_work(struct inet6_ifaddr * ifp)320 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
321 {
322 	if (cancel_delayed_work(&ifp->dad_work))
323 		__in6_ifa_put(ifp);
324 }
325 
addrconf_mod_rs_timer(struct inet6_dev * idev,unsigned long when)326 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
327 				  unsigned long when)
328 {
329 	if (!mod_timer(&idev->rs_timer, jiffies + when))
330 		in6_dev_hold(idev);
331 }
332 
addrconf_mod_dad_work(struct inet6_ifaddr * ifp,unsigned long delay)333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
334 				   unsigned long delay)
335 {
336 	in6_ifa_hold(ifp);
337 	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
338 		in6_ifa_put(ifp);
339 }
340 
snmp6_alloc_dev(struct inet6_dev * idev)341 static int snmp6_alloc_dev(struct inet6_dev *idev)
342 {
343 	int i;
344 
345 	idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
346 	if (!idev->stats.ipv6)
347 		goto err_ip;
348 
349 	for_each_possible_cpu(i) {
350 		struct ipstats_mib *addrconf_stats;
351 		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
352 		u64_stats_init(&addrconf_stats->syncp);
353 	}
354 
355 
356 	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
357 					GFP_KERNEL);
358 	if (!idev->stats.icmpv6dev)
359 		goto err_icmp;
360 	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
361 					   GFP_KERNEL_ACCOUNT);
362 	if (!idev->stats.icmpv6msgdev)
363 		goto err_icmpmsg;
364 
365 	return 0;
366 
367 err_icmpmsg:
368 	kfree(idev->stats.icmpv6dev);
369 err_icmp:
370 	free_percpu(idev->stats.ipv6);
371 err_ip:
372 	return -ENOMEM;
373 }
374 
ipv6_add_dev(struct net_device * dev)375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
376 {
377 	struct inet6_dev *ndev;
378 	int err = -ENOMEM;
379 
380 	ASSERT_RTNL();
381 	netdev_ops_assert_locked(dev);
382 
383 	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
384 		return ERR_PTR(-EINVAL);
385 
386 	ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
387 	if (!ndev)
388 		return ERR_PTR(err);
389 
390 	rwlock_init(&ndev->lock);
391 	ndev->dev = dev;
392 	INIT_LIST_HEAD(&ndev->addr_list);
393 	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
394 	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
395 
396 	if (ndev->cnf.stable_secret.initialized)
397 		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
398 
399 	ndev->cnf.mtu6 = dev->mtu;
400 	ndev->ra_mtu = 0;
401 	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
402 	if (!ndev->nd_parms) {
403 		kfree(ndev);
404 		return ERR_PTR(err);
405 	}
406 	if (ndev->cnf.forwarding)
407 		netif_disable_lro(dev);
408 	/* We refer to the device */
409 	netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
410 
411 	if (snmp6_alloc_dev(ndev) < 0) {
412 		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
413 			   __func__);
414 		neigh_parms_release(&nd_tbl, ndev->nd_parms);
415 		netdev_put(dev, &ndev->dev_tracker);
416 		kfree(ndev);
417 		return ERR_PTR(err);
418 	}
419 
420 	if (dev != blackhole_netdev) {
421 		if (snmp6_register_dev(ndev) < 0) {
422 			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
423 				   __func__, dev->name);
424 			goto err_release;
425 		}
426 	}
427 	/* One reference from device. */
428 	refcount_set(&ndev->refcnt, 1);
429 
430 	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
431 		ndev->cnf.accept_dad = -1;
432 
433 #if IS_ENABLED(CONFIG_IPV6_SIT)
434 	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
435 		pr_info("%s: Disabled Multicast RS\n", dev->name);
436 		ndev->cnf.rtr_solicits = 0;
437 	}
438 #endif
439 
440 	INIT_LIST_HEAD(&ndev->tempaddr_list);
441 	ndev->desync_factor = U32_MAX;
442 	if ((dev->flags&IFF_LOOPBACK) ||
443 	    dev->type == ARPHRD_TUNNEL ||
444 	    dev->type == ARPHRD_TUNNEL6 ||
445 	    dev->type == ARPHRD_SIT ||
446 	    dev->type == ARPHRD_NONE) {
447 		ndev->cnf.use_tempaddr = -1;
448 	}
449 
450 	ndev->token = in6addr_any;
451 
452 	if (netif_running(dev) && addrconf_link_ready(dev))
453 		ndev->if_flags |= IF_READY;
454 
455 	ipv6_mc_init_dev(ndev);
456 	ndev->tstamp = jiffies;
457 	if (dev != blackhole_netdev) {
458 		err = addrconf_sysctl_register(ndev);
459 		if (err) {
460 			ipv6_mc_destroy_dev(ndev);
461 			snmp6_unregister_dev(ndev);
462 			goto err_release;
463 		}
464 	}
465 	/* protected by rtnl_lock */
466 	rcu_assign_pointer(dev->ip6_ptr, ndev);
467 
468 	if (dev != blackhole_netdev) {
469 		/* Join interface-local all-node multicast group */
470 		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
471 
472 		/* Join all-node multicast group */
473 		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
474 
475 		/* Join all-router multicast group if forwarding is set */
476 		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
477 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
478 	}
479 	return ndev;
480 
481 err_release:
482 	neigh_parms_release(&nd_tbl, ndev->nd_parms);
483 	ndev->dead = 1;
484 	in6_dev_finish_destroy(ndev);
485 	return ERR_PTR(err);
486 }
487 
ipv6_find_idev(struct net_device * dev)488 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
489 {
490 	struct inet6_dev *idev;
491 
492 	ASSERT_RTNL();
493 
494 	idev = __in6_dev_get(dev);
495 	if (!idev) {
496 		idev = ipv6_add_dev(dev);
497 		if (IS_ERR(idev))
498 			return idev;
499 	}
500 
501 	if (dev->flags&IFF_UP)
502 		ipv6_mc_up(idev);
503 	return idev;
504 }
505 
inet6_netconf_msgsize_devconf(int type)506 static int inet6_netconf_msgsize_devconf(int type)
507 {
508 	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
509 		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
510 	bool all = false;
511 
512 	if (type == NETCONFA_ALL)
513 		all = true;
514 
515 	if (all || type == NETCONFA_FORWARDING)
516 		size += nla_total_size(4);
517 #ifdef CONFIG_IPV6_MROUTE
518 	if (all || type == NETCONFA_MC_FORWARDING)
519 		size += nla_total_size(4);
520 #endif
521 	if (all || type == NETCONFA_PROXY_NEIGH)
522 		size += nla_total_size(4);
523 
524 	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
525 		size += nla_total_size(4);
526 
527 	return size;
528 }
529 
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)530 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
531 				      struct ipv6_devconf *devconf, u32 portid,
532 				      u32 seq, int event, unsigned int flags,
533 				      int type)
534 {
535 	struct nlmsghdr  *nlh;
536 	struct netconfmsg *ncm;
537 	bool all = false;
538 
539 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
540 			flags);
541 	if (!nlh)
542 		return -EMSGSIZE;
543 
544 	if (type == NETCONFA_ALL)
545 		all = true;
546 
547 	ncm = nlmsg_data(nlh);
548 	ncm->ncm_family = AF_INET6;
549 
550 	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
551 		goto nla_put_failure;
552 
553 	if (!devconf)
554 		goto out;
555 
556 	if ((all || type == NETCONFA_FORWARDING) &&
557 	    nla_put_s32(skb, NETCONFA_FORWARDING,
558 			READ_ONCE(devconf->forwarding)) < 0)
559 		goto nla_put_failure;
560 #ifdef CONFIG_IPV6_MROUTE
561 	if ((all || type == NETCONFA_MC_FORWARDING) &&
562 	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
563 			atomic_read(&devconf->mc_forwarding)) < 0)
564 		goto nla_put_failure;
565 #endif
566 	if ((all || type == NETCONFA_PROXY_NEIGH) &&
567 	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
568 			READ_ONCE(devconf->proxy_ndp)) < 0)
569 		goto nla_put_failure;
570 
571 	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
572 	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
573 			READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
574 		goto nla_put_failure;
575 
576 out:
577 	nlmsg_end(skb, nlh);
578 	return 0;
579 
580 nla_put_failure:
581 	nlmsg_cancel(skb, nlh);
582 	return -EMSGSIZE;
583 }
584 
inet6_netconf_notify_devconf(struct net * net,int event,int type,int ifindex,struct ipv6_devconf * devconf)585 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
586 				  int ifindex, struct ipv6_devconf *devconf)
587 {
588 	struct sk_buff *skb;
589 	int err = -ENOBUFS;
590 
591 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
592 	if (!skb)
593 		goto errout;
594 
595 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
596 					 event, 0, type);
597 	if (err < 0) {
598 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
599 		WARN_ON(err == -EMSGSIZE);
600 		kfree_skb(skb);
601 		goto errout;
602 	}
603 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
604 	return;
605 errout:
606 	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
607 }
608 
609 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
610 	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
611 	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
612 	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
613 	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
614 };
615 
inet6_netconf_valid_get_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)616 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
617 				       const struct nlmsghdr *nlh,
618 				       struct nlattr **tb,
619 				       struct netlink_ext_ack *extack)
620 {
621 	int i, err;
622 
623 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
624 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
625 		return -EINVAL;
626 	}
627 
628 	if (!netlink_strict_get_check(skb))
629 		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
630 					      tb, NETCONFA_MAX,
631 					      devconf_ipv6_policy, extack);
632 
633 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
634 					    tb, NETCONFA_MAX,
635 					    devconf_ipv6_policy, extack);
636 	if (err)
637 		return err;
638 
639 	for (i = 0; i <= NETCONFA_MAX; i++) {
640 		if (!tb[i])
641 			continue;
642 
643 		switch (i) {
644 		case NETCONFA_IFINDEX:
645 			break;
646 		default:
647 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
648 			return -EINVAL;
649 		}
650 	}
651 
652 	return 0;
653 }
654 
inet6_netconf_get_devconf(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)655 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
656 				     struct nlmsghdr *nlh,
657 				     struct netlink_ext_ack *extack)
658 {
659 	struct net *net = sock_net(in_skb->sk);
660 	struct nlattr *tb[NETCONFA_MAX+1];
661 	struct inet6_dev *in6_dev = NULL;
662 	struct net_device *dev = NULL;
663 	struct sk_buff *skb;
664 	struct ipv6_devconf *devconf;
665 	int ifindex;
666 	int err;
667 
668 	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
669 	if (err < 0)
670 		return err;
671 
672 	if (!tb[NETCONFA_IFINDEX])
673 		return -EINVAL;
674 
675 	err = -EINVAL;
676 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
677 	switch (ifindex) {
678 	case NETCONFA_IFINDEX_ALL:
679 		devconf = net->ipv6.devconf_all;
680 		break;
681 	case NETCONFA_IFINDEX_DEFAULT:
682 		devconf = net->ipv6.devconf_dflt;
683 		break;
684 	default:
685 		dev = dev_get_by_index(net, ifindex);
686 		if (!dev)
687 			return -EINVAL;
688 		in6_dev = in6_dev_get(dev);
689 		if (!in6_dev)
690 			goto errout;
691 		devconf = &in6_dev->cnf;
692 		break;
693 	}
694 
695 	err = -ENOBUFS;
696 	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
697 	if (!skb)
698 		goto errout;
699 
700 	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
701 					 NETLINK_CB(in_skb).portid,
702 					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
703 					 NETCONFA_ALL);
704 	if (err < 0) {
705 		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
706 		WARN_ON(err == -EMSGSIZE);
707 		kfree_skb(skb);
708 		goto errout;
709 	}
710 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
711 errout:
712 	if (in6_dev)
713 		in6_dev_put(in6_dev);
714 	dev_put(dev);
715 	return err;
716 }
717 
718 /* Combine dev_addr_genid and dev_base_seq to detect changes.
719  */
inet6_base_seq(const struct net * net)720 static u32 inet6_base_seq(const struct net *net)
721 {
722 	u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
723 		  READ_ONCE(net->dev_base_seq);
724 
725 	/* Must not return 0 (see nl_dump_check_consistent()).
726 	 * Chose a value far away from 0.
727 	 */
728 	if (!res)
729 		res = 0x80000000;
730 	return res;
731 }
732 
inet6_netconf_dump_devconf(struct sk_buff * skb,struct netlink_callback * cb)733 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
734 				      struct netlink_callback *cb)
735 {
736 	const struct nlmsghdr *nlh = cb->nlh;
737 	struct net *net = sock_net(skb->sk);
738 	struct {
739 		unsigned long ifindex;
740 		unsigned int all_default;
741 	} *ctx = (void *)cb->ctx;
742 	struct net_device *dev;
743 	struct inet6_dev *idev;
744 	int err = 0;
745 
746 	if (cb->strict_check) {
747 		struct netlink_ext_ack *extack = cb->extack;
748 		struct netconfmsg *ncm;
749 
750 		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
751 			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
752 			return -EINVAL;
753 		}
754 
755 		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
756 			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
757 			return -EINVAL;
758 		}
759 	}
760 
761 	rcu_read_lock();
762 	for_each_netdev_dump(net, dev, ctx->ifindex) {
763 		idev = __in6_dev_get(dev);
764 		if (!idev)
765 			continue;
766 		err = inet6_netconf_fill_devconf(skb, dev->ifindex,
767 					         &idev->cnf,
768 						 NETLINK_CB(cb->skb).portid,
769 						 nlh->nlmsg_seq,
770 						 RTM_NEWNETCONF,
771 						 NLM_F_MULTI,
772 						 NETCONFA_ALL);
773 		if (err < 0)
774 			goto done;
775 	}
776 	if (ctx->all_default == 0) {
777 		err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
778 						 net->ipv6.devconf_all,
779 						 NETLINK_CB(cb->skb).portid,
780 						 nlh->nlmsg_seq,
781 						 RTM_NEWNETCONF, NLM_F_MULTI,
782 						 NETCONFA_ALL);
783 		if (err < 0)
784 			goto done;
785 		ctx->all_default++;
786 	}
787 	if (ctx->all_default == 1) {
788 		err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
789 						 net->ipv6.devconf_dflt,
790 						 NETLINK_CB(cb->skb).portid,
791 						 nlh->nlmsg_seq,
792 						 RTM_NEWNETCONF, NLM_F_MULTI,
793 						 NETCONFA_ALL);
794 		if (err < 0)
795 			goto done;
796 		ctx->all_default++;
797 	}
798 done:
799 	rcu_read_unlock();
800 	return err;
801 }
802 
803 #ifdef CONFIG_SYSCTL
dev_forward_change(struct inet6_dev * idev)804 static void dev_forward_change(struct inet6_dev *idev)
805 {
806 	struct net_device *dev;
807 	struct inet6_ifaddr *ifa;
808 	LIST_HEAD(tmp_addr_list);
809 
810 	if (!idev)
811 		return;
812 	dev = idev->dev;
813 	if (idev->cnf.forwarding)
814 		dev_disable_lro(dev);
815 	if (dev->flags & IFF_MULTICAST) {
816 		if (idev->cnf.forwarding) {
817 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
818 			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
819 			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
820 		} else {
821 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
822 			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
823 			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
824 		}
825 	}
826 
827 	read_lock_bh(&idev->lock);
828 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
829 		if (ifa->flags&IFA_F_TENTATIVE)
830 			continue;
831 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
832 	}
833 	read_unlock_bh(&idev->lock);
834 
835 	while (!list_empty(&tmp_addr_list)) {
836 		ifa = list_first_entry(&tmp_addr_list,
837 				       struct inet6_ifaddr, if_list_aux);
838 		list_del(&ifa->if_list_aux);
839 		if (idev->cnf.forwarding)
840 			addrconf_join_anycast(ifa);
841 		else
842 			addrconf_leave_anycast(ifa);
843 	}
844 
845 	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
846 				     NETCONFA_FORWARDING,
847 				     dev->ifindex, &idev->cnf);
848 }
849 
850 
addrconf_forward_change(struct net * net,__s32 newf)851 static void addrconf_forward_change(struct net *net, __s32 newf)
852 {
853 	struct net_device *dev;
854 	struct inet6_dev *idev;
855 
856 	for_each_netdev(net, dev) {
857 		idev = __in6_dev_get_rtnl_net(dev);
858 		if (idev) {
859 			int changed = (!idev->cnf.forwarding) ^ (!newf);
860 
861 			WRITE_ONCE(idev->cnf.forwarding, newf);
862 			if (changed)
863 				dev_forward_change(idev);
864 		}
865 	}
866 }
867 
addrconf_fixup_forwarding(const struct ctl_table * table,int * p,int newf)868 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf)
869 {
870 	struct net *net = (struct net *)table->extra2;
871 	int old;
872 
873 	if (!rtnl_net_trylock(net))
874 		return restart_syscall();
875 
876 	old = *p;
877 	WRITE_ONCE(*p, newf);
878 
879 	if (p == &net->ipv6.devconf_dflt->forwarding) {
880 		if ((!newf) ^ (!old))
881 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
882 						     NETCONFA_FORWARDING,
883 						     NETCONFA_IFINDEX_DEFAULT,
884 						     net->ipv6.devconf_dflt);
885 		rtnl_net_unlock(net);
886 		return 0;
887 	}
888 
889 	if (p == &net->ipv6.devconf_all->forwarding) {
890 		int old_dflt = net->ipv6.devconf_dflt->forwarding;
891 
892 		WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
893 		if ((!newf) ^ (!old_dflt))
894 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
895 						     NETCONFA_FORWARDING,
896 						     NETCONFA_IFINDEX_DEFAULT,
897 						     net->ipv6.devconf_dflt);
898 
899 		addrconf_forward_change(net, newf);
900 		if ((!newf) ^ (!old))
901 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
902 						     NETCONFA_FORWARDING,
903 						     NETCONFA_IFINDEX_ALL,
904 						     net->ipv6.devconf_all);
905 	} else if ((!newf) ^ (!old))
906 		dev_forward_change((struct inet6_dev *)table->extra1);
907 	rtnl_net_unlock(net);
908 
909 	if (newf)
910 		rt6_purge_dflt_routers(net);
911 	return 1;
912 }
913 
addrconf_linkdown_change(struct net * net,__s32 newf)914 static void addrconf_linkdown_change(struct net *net, __s32 newf)
915 {
916 	struct net_device *dev;
917 	struct inet6_dev *idev;
918 
919 	for_each_netdev(net, dev) {
920 		idev = __in6_dev_get_rtnl_net(dev);
921 		if (idev) {
922 			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
923 
924 			WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
925 			if (changed)
926 				inet6_netconf_notify_devconf(dev_net(dev),
927 							     RTM_NEWNETCONF,
928 							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
929 							     dev->ifindex,
930 							     &idev->cnf);
931 		}
932 	}
933 }
934 
addrconf_fixup_linkdown(const struct ctl_table * table,int * p,int newf)935 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf)
936 {
937 	struct net *net = (struct net *)table->extra2;
938 	int old;
939 
940 	if (!rtnl_net_trylock(net))
941 		return restart_syscall();
942 
943 	old = *p;
944 	WRITE_ONCE(*p, newf);
945 
946 	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
947 		if ((!newf) ^ (!old))
948 			inet6_netconf_notify_devconf(net,
949 						     RTM_NEWNETCONF,
950 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
951 						     NETCONFA_IFINDEX_DEFAULT,
952 						     net->ipv6.devconf_dflt);
953 		rtnl_net_unlock(net);
954 		return 0;
955 	}
956 
957 	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
958 		WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
959 		addrconf_linkdown_change(net, newf);
960 		if ((!newf) ^ (!old))
961 			inet6_netconf_notify_devconf(net,
962 						     RTM_NEWNETCONF,
963 						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
964 						     NETCONFA_IFINDEX_ALL,
965 						     net->ipv6.devconf_all);
966 	}
967 
968 	rtnl_net_unlock(net);
969 
970 	return 1;
971 }
972 
973 #endif
974 
975 /* Nobody refers to this ifaddr, destroy it */
inet6_ifa_finish_destroy(struct inet6_ifaddr * ifp)976 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
977 {
978 	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
979 
980 #ifdef NET_REFCNT_DEBUG
981 	pr_debug("%s\n", __func__);
982 #endif
983 
984 	in6_dev_put(ifp->idev);
985 
986 	if (cancel_delayed_work(&ifp->dad_work))
987 		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
988 			  ifp);
989 
990 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
991 		pr_warn("Freeing alive inet6 address %p\n", ifp);
992 		return;
993 	}
994 
995 	kfree_rcu(ifp, rcu);
996 }
997 
998 static void
ipv6_link_dev_addr(struct inet6_dev * idev,struct inet6_ifaddr * ifp)999 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1000 {
1001 	struct list_head *p;
1002 	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1003 
1004 	/*
1005 	 * Each device address list is sorted in order of scope -
1006 	 * global before linklocal.
1007 	 */
1008 	list_for_each(p, &idev->addr_list) {
1009 		struct inet6_ifaddr *ifa
1010 			= list_entry(p, struct inet6_ifaddr, if_list);
1011 		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1012 			break;
1013 	}
1014 
1015 	list_add_tail_rcu(&ifp->if_list, p);
1016 }
1017 
inet6_addr_hash(const struct net * net,const struct in6_addr * addr)1018 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1019 {
1020 	u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net));
1021 
1022 	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1023 }
1024 
ipv6_chk_same_addr(struct net * net,const struct in6_addr * addr,struct net_device * dev,unsigned int hash)1025 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1026 			       struct net_device *dev, unsigned int hash)
1027 {
1028 	struct inet6_ifaddr *ifp;
1029 
1030 	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1031 		if (ipv6_addr_equal(&ifp->addr, addr)) {
1032 			if (!dev || ifp->idev->dev == dev)
1033 				return true;
1034 		}
1035 	}
1036 	return false;
1037 }
1038 
ipv6_add_addr_hash(struct net_device * dev,struct inet6_ifaddr * ifa)1039 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1040 {
1041 	struct net *net = dev_net(dev);
1042 	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1043 	int err = 0;
1044 
1045 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1046 
1047 	/* Ignore adding duplicate addresses on an interface */
1048 	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1049 		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1050 		err = -EEXIST;
1051 	} else {
1052 		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1053 	}
1054 
1055 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1056 
1057 	return err;
1058 }
1059 
1060 /* On success it returns ifp with increased reference count */
1061 
1062 static struct inet6_ifaddr *
ipv6_add_addr(struct inet6_dev * idev,struct ifa6_config * cfg,bool can_block,struct netlink_ext_ack * extack)1063 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1064 	      bool can_block, struct netlink_ext_ack *extack)
1065 {
1066 	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1067 	int addr_type = ipv6_addr_type(cfg->pfx);
1068 	struct net *net = dev_net(idev->dev);
1069 	struct inet6_ifaddr *ifa = NULL;
1070 	struct fib6_info *f6i = NULL;
1071 	int err = 0;
1072 
1073 	if (addr_type == IPV6_ADDR_ANY) {
1074 		NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1075 		return ERR_PTR(-EADDRNOTAVAIL);
1076 	} else if (addr_type & IPV6_ADDR_MULTICAST &&
1077 		   !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1078 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1079 		return ERR_PTR(-EADDRNOTAVAIL);
1080 	} else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1081 		   !netif_is_l3_master(idev->dev) &&
1082 		   addr_type & IPV6_ADDR_LOOPBACK) {
1083 		NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1084 		return ERR_PTR(-EADDRNOTAVAIL);
1085 	}
1086 
1087 	if (idev->dead) {
1088 		NL_SET_ERR_MSG_MOD(extack, "device is going away");
1089 		err = -ENODEV;
1090 		goto out;
1091 	}
1092 
1093 	if (idev->cnf.disable_ipv6) {
1094 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1095 		err = -EACCES;
1096 		goto out;
1097 	}
1098 
1099 	/* validator notifier needs to be blocking;
1100 	 * do not call in atomic context
1101 	 */
1102 	if (can_block) {
1103 		struct in6_validator_info i6vi = {
1104 			.i6vi_addr = *cfg->pfx,
1105 			.i6vi_dev = idev,
1106 			.extack = extack,
1107 		};
1108 
1109 		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1110 		err = notifier_to_errno(err);
1111 		if (err < 0)
1112 			goto out;
1113 	}
1114 
1115 	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1116 	if (!ifa) {
1117 		err = -ENOBUFS;
1118 		goto out;
1119 	}
1120 
1121 	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1122 	if (IS_ERR(f6i)) {
1123 		err = PTR_ERR(f6i);
1124 		f6i = NULL;
1125 		goto out;
1126 	}
1127 
1128 	neigh_parms_data_state_setall(idev->nd_parms);
1129 
1130 	ifa->addr = *cfg->pfx;
1131 	if (cfg->peer_pfx)
1132 		ifa->peer_addr = *cfg->peer_pfx;
1133 
1134 	spin_lock_init(&ifa->lock);
1135 	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1136 	INIT_HLIST_NODE(&ifa->addr_lst);
1137 	ifa->scope = cfg->scope;
1138 	ifa->prefix_len = cfg->plen;
1139 	ifa->rt_priority = cfg->rt_priority;
1140 	ifa->flags = cfg->ifa_flags;
1141 	ifa->ifa_proto = cfg->ifa_proto;
1142 	/* No need to add the TENTATIVE flag for addresses with NODAD */
1143 	if (!(cfg->ifa_flags & IFA_F_NODAD))
1144 		ifa->flags |= IFA_F_TENTATIVE;
1145 	ifa->valid_lft = cfg->valid_lft;
1146 	ifa->prefered_lft = cfg->preferred_lft;
1147 	ifa->cstamp = ifa->tstamp = jiffies;
1148 	ifa->tokenized = false;
1149 
1150 	ifa->rt = f6i;
1151 
1152 	ifa->idev = idev;
1153 	in6_dev_hold(idev);
1154 
1155 	/* For caller */
1156 	refcount_set(&ifa->refcnt, 1);
1157 
1158 	rcu_read_lock();
1159 
1160 	err = ipv6_add_addr_hash(idev->dev, ifa);
1161 	if (err < 0) {
1162 		rcu_read_unlock();
1163 		goto out;
1164 	}
1165 
1166 	write_lock_bh(&idev->lock);
1167 
1168 	/* Add to inet6_dev unicast addr list. */
1169 	ipv6_link_dev_addr(idev, ifa);
1170 
1171 	if (ifa->flags&IFA_F_TEMPORARY) {
1172 		list_add(&ifa->tmp_list, &idev->tempaddr_list);
1173 		in6_ifa_hold(ifa);
1174 	}
1175 
1176 	in6_ifa_hold(ifa);
1177 	write_unlock_bh(&idev->lock);
1178 
1179 	rcu_read_unlock();
1180 
1181 	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1182 out:
1183 	if (unlikely(err < 0)) {
1184 		fib6_info_release(f6i);
1185 
1186 		if (ifa) {
1187 			if (ifa->idev)
1188 				in6_dev_put(ifa->idev);
1189 			kfree(ifa);
1190 		}
1191 		ifa = ERR_PTR(err);
1192 	}
1193 
1194 	return ifa;
1195 }
1196 
1197 enum cleanup_prefix_rt_t {
1198 	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1199 	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1200 	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1201 };
1202 
1203 /*
1204  * Check, whether the prefix for ifp would still need a prefix route
1205  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1206  * constants.
1207  *
1208  * 1) we don't purge prefix if address was not permanent.
1209  *    prefix is managed by its own lifetime.
1210  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1211  * 3) if there are no addresses, delete prefix.
1212  * 4) if there are still other permanent address(es),
1213  *    corresponding prefix is still permanent.
1214  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1215  *    don't purge the prefix, assume user space is managing it.
1216  * 6) otherwise, update prefix lifetime to the
1217  *    longest valid lifetime among the corresponding
1218  *    addresses on the device.
1219  *    Note: subsequent RA will update lifetime.
1220  **/
1221 static enum cleanup_prefix_rt_t
check_cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long * expires)1222 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1223 {
1224 	struct inet6_ifaddr *ifa;
1225 	struct inet6_dev *idev = ifp->idev;
1226 	unsigned long lifetime;
1227 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1228 
1229 	*expires = jiffies;
1230 
1231 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1232 		if (ifa == ifp)
1233 			continue;
1234 		if (ifa->prefix_len != ifp->prefix_len ||
1235 		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1236 				       ifp->prefix_len))
1237 			continue;
1238 		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1239 			return CLEANUP_PREFIX_RT_NOP;
1240 
1241 		action = CLEANUP_PREFIX_RT_EXPIRE;
1242 
1243 		spin_lock(&ifa->lock);
1244 
1245 		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1246 		/*
1247 		 * Note: Because this address is
1248 		 * not permanent, lifetime <
1249 		 * LONG_MAX / HZ here.
1250 		 */
1251 		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1252 			*expires = ifa->tstamp + lifetime * HZ;
1253 		spin_unlock(&ifa->lock);
1254 	}
1255 
1256 	return action;
1257 }
1258 
1259 static void
cleanup_prefix_route(struct inet6_ifaddr * ifp,unsigned long expires,bool del_rt,bool del_peer)1260 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1261 		     bool del_rt, bool del_peer)
1262 {
1263 	struct fib6_table *table;
1264 	struct fib6_info *f6i;
1265 
1266 	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1267 					ifp->prefix_len,
1268 					ifp->idev->dev, 0, RTF_DEFAULT, true);
1269 	if (f6i) {
1270 		if (del_rt)
1271 			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1272 		else {
1273 			if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1274 				table = f6i->fib6_table;
1275 				spin_lock_bh(&table->tb6_lock);
1276 
1277 				fib6_set_expires(f6i, expires);
1278 				fib6_add_gc_list(f6i);
1279 
1280 				spin_unlock_bh(&table->tb6_lock);
1281 			}
1282 			fib6_info_release(f6i);
1283 		}
1284 	}
1285 }
1286 
1287 
1288 /* This function wants to get referenced ifp and releases it before return */
1289 
ipv6_del_addr(struct inet6_ifaddr * ifp)1290 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1291 {
1292 	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1293 	struct net *net = dev_net(ifp->idev->dev);
1294 	unsigned long expires;
1295 	int state;
1296 
1297 	ASSERT_RTNL();
1298 
1299 	spin_lock_bh(&ifp->lock);
1300 	state = ifp->state;
1301 	ifp->state = INET6_IFADDR_STATE_DEAD;
1302 	spin_unlock_bh(&ifp->lock);
1303 
1304 	if (state == INET6_IFADDR_STATE_DEAD)
1305 		goto out;
1306 
1307 	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1308 	hlist_del_init_rcu(&ifp->addr_lst);
1309 	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1310 
1311 	write_lock_bh(&ifp->idev->lock);
1312 
1313 	if (ifp->flags&IFA_F_TEMPORARY) {
1314 		list_del(&ifp->tmp_list);
1315 		if (ifp->ifpub) {
1316 			in6_ifa_put(ifp->ifpub);
1317 			ifp->ifpub = NULL;
1318 		}
1319 		__in6_ifa_put(ifp);
1320 	}
1321 
1322 	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1323 		action = check_cleanup_prefix_route(ifp, &expires);
1324 
1325 	list_del_rcu(&ifp->if_list);
1326 	__in6_ifa_put(ifp);
1327 
1328 	write_unlock_bh(&ifp->idev->lock);
1329 
1330 	addrconf_del_dad_work(ifp);
1331 
1332 	ipv6_ifa_notify(RTM_DELADDR, ifp);
1333 
1334 	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1335 
1336 	if (action != CLEANUP_PREFIX_RT_NOP) {
1337 		cleanup_prefix_route(ifp, expires,
1338 			action == CLEANUP_PREFIX_RT_DEL, false);
1339 	}
1340 
1341 	/* clean up prefsrc entries */
1342 	rt6_remove_prefsrc(ifp);
1343 out:
1344 	in6_ifa_put(ifp);
1345 }
1346 
ipv6_get_regen_advance(const struct inet6_dev * idev)1347 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1348 {
1349 	return READ_ONCE(idev->cnf.regen_min_advance) +
1350 		READ_ONCE(idev->cnf.regen_max_retry) *
1351 		READ_ONCE(idev->cnf.dad_transmits) *
1352 		max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1353 }
1354 
ipv6_create_tempaddr(struct inet6_ifaddr * ifp,bool block)1355 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1356 {
1357 	struct inet6_dev *idev = ifp->idev;
1358 	unsigned long tmp_tstamp, age;
1359 	unsigned long regen_advance;
1360 	unsigned long now = jiffies;
1361 	u32 if_public_preferred_lft;
1362 	s32 cnf_temp_preferred_lft;
1363 	struct inet6_ifaddr *ift;
1364 	struct ifa6_config cfg;
1365 	long max_desync_factor;
1366 	struct in6_addr addr;
1367 	int ret = 0;
1368 
1369 	write_lock_bh(&idev->lock);
1370 
1371 retry:
1372 	in6_dev_hold(idev);
1373 	if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1374 		write_unlock_bh(&idev->lock);
1375 		pr_info("%s: use_tempaddr is disabled\n", __func__);
1376 		in6_dev_put(idev);
1377 		ret = -1;
1378 		goto out;
1379 	}
1380 	spin_lock_bh(&ifp->lock);
1381 	if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1382 		WRITE_ONCE(idev->cnf.use_tempaddr, -1);	/*XXX*/
1383 		spin_unlock_bh(&ifp->lock);
1384 		write_unlock_bh(&idev->lock);
1385 		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1386 			__func__);
1387 		in6_dev_put(idev);
1388 		ret = -1;
1389 		goto out;
1390 	}
1391 	in6_ifa_hold(ifp);
1392 	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1393 	ipv6_gen_rnd_iid(&addr);
1394 
1395 	age = (now - ifp->tstamp) / HZ;
1396 
1397 	regen_advance = ipv6_get_regen_advance(idev);
1398 
1399 	/* recalculate max_desync_factor each time and update
1400 	 * idev->desync_factor if it's larger
1401 	 */
1402 	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1403 	max_desync_factor = min_t(long,
1404 				  READ_ONCE(idev->cnf.max_desync_factor),
1405 				  cnf_temp_preferred_lft - regen_advance);
1406 
1407 	if (unlikely(idev->desync_factor > max_desync_factor)) {
1408 		if (max_desync_factor > 0) {
1409 			get_random_bytes(&idev->desync_factor,
1410 					 sizeof(idev->desync_factor));
1411 			idev->desync_factor %= max_desync_factor;
1412 		} else {
1413 			idev->desync_factor = 0;
1414 		}
1415 	}
1416 
1417 	if_public_preferred_lft = ifp->prefered_lft;
1418 
1419 	memset(&cfg, 0, sizeof(cfg));
1420 	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1421 			      READ_ONCE(idev->cnf.temp_valid_lft) + age);
1422 	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1423 	cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1424 	cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1425 
1426 	cfg.plen = ifp->prefix_len;
1427 	tmp_tstamp = ifp->tstamp;
1428 	spin_unlock_bh(&ifp->lock);
1429 
1430 	write_unlock_bh(&idev->lock);
1431 
1432 	/* From RFC 4941:
1433 	 *
1434 	 *     A temporary address is created only if this calculated Preferred
1435 	 *     Lifetime is greater than REGEN_ADVANCE time units.  In
1436 	 *     particular, an implementation must not create a temporary address
1437 	 *     with a zero Preferred Lifetime.
1438 	 *
1439 	 *     ...
1440 	 *
1441 	 *     When creating a temporary address, the lifetime values MUST be
1442 	 *     derived from the corresponding prefix as follows:
1443 	 *
1444 	 *     ...
1445 	 *
1446 	 *     *  Its Preferred Lifetime is the lower of the Preferred Lifetime
1447 	 *        of the public address or TEMP_PREFERRED_LIFETIME -
1448 	 *        DESYNC_FACTOR.
1449 	 *
1450 	 * To comply with the RFC's requirements, clamp the preferred lifetime
1451 	 * to a minimum of regen_advance, unless that would exceed valid_lft or
1452 	 * ifp->prefered_lft.
1453 	 *
1454 	 * Use age calculation as in addrconf_verify to avoid unnecessary
1455 	 * temporary addresses being generated.
1456 	 */
1457 	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1458 	if (cfg.preferred_lft <= regen_advance + age) {
1459 		cfg.preferred_lft = regen_advance + age + 1;
1460 		if (cfg.preferred_lft > cfg.valid_lft ||
1461 		    cfg.preferred_lft > if_public_preferred_lft) {
1462 			in6_ifa_put(ifp);
1463 			in6_dev_put(idev);
1464 			ret = -1;
1465 			goto out;
1466 		}
1467 	}
1468 
1469 	cfg.ifa_flags = IFA_F_TEMPORARY;
1470 	/* set in addrconf_prefix_rcv() */
1471 	if (ifp->flags & IFA_F_OPTIMISTIC)
1472 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1473 
1474 	cfg.pfx = &addr;
1475 	cfg.scope = ipv6_addr_scope(cfg.pfx);
1476 
1477 	ift = ipv6_add_addr(idev, &cfg, block, NULL);
1478 	if (IS_ERR(ift)) {
1479 		in6_ifa_put(ifp);
1480 		in6_dev_put(idev);
1481 		pr_info("%s: retry temporary address regeneration\n", __func__);
1482 		write_lock_bh(&idev->lock);
1483 		goto retry;
1484 	}
1485 
1486 	spin_lock_bh(&ift->lock);
1487 	ift->ifpub = ifp;
1488 	ift->cstamp = now;
1489 	ift->tstamp = tmp_tstamp;
1490 	spin_unlock_bh(&ift->lock);
1491 
1492 	addrconf_dad_start(ift);
1493 	in6_ifa_put(ift);
1494 	in6_dev_put(idev);
1495 out:
1496 	return ret;
1497 }
1498 
1499 /*
1500  *	Choose an appropriate source address (RFC3484)
1501  */
1502 enum {
1503 	IPV6_SADDR_RULE_INIT = 0,
1504 	IPV6_SADDR_RULE_LOCAL,
1505 	IPV6_SADDR_RULE_SCOPE,
1506 	IPV6_SADDR_RULE_PREFERRED,
1507 #ifdef CONFIG_IPV6_MIP6
1508 	IPV6_SADDR_RULE_HOA,
1509 #endif
1510 	IPV6_SADDR_RULE_OIF,
1511 	IPV6_SADDR_RULE_LABEL,
1512 	IPV6_SADDR_RULE_PRIVACY,
1513 	IPV6_SADDR_RULE_ORCHID,
1514 	IPV6_SADDR_RULE_PREFIX,
1515 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1516 	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1517 #endif
1518 	IPV6_SADDR_RULE_MAX
1519 };
1520 
1521 struct ipv6_saddr_score {
1522 	int			rule;
1523 	int			addr_type;
1524 	struct inet6_ifaddr	*ifa;
1525 	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1526 	int			scopedist;
1527 	int			matchlen;
1528 };
1529 
1530 struct ipv6_saddr_dst {
1531 	const struct in6_addr *addr;
1532 	int ifindex;
1533 	int scope;
1534 	int label;
1535 	unsigned int prefs;
1536 };
1537 
ipv6_saddr_preferred(int type)1538 static inline int ipv6_saddr_preferred(int type)
1539 {
1540 	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1541 		return 1;
1542 	return 0;
1543 }
1544 
ipv6_use_optimistic_addr(const struct net * net,const struct inet6_dev * idev)1545 static bool ipv6_use_optimistic_addr(const struct net *net,
1546 				     const struct inet6_dev *idev)
1547 {
1548 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1549 	if (!idev)
1550 		return false;
1551 	if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1552 	    !READ_ONCE(idev->cnf.optimistic_dad))
1553 		return false;
1554 	if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1555 	    !READ_ONCE(idev->cnf.use_optimistic))
1556 		return false;
1557 
1558 	return true;
1559 #else
1560 	return false;
1561 #endif
1562 }
1563 
ipv6_allow_optimistic_dad(const struct net * net,const struct inet6_dev * idev)1564 static bool ipv6_allow_optimistic_dad(const struct net *net,
1565 				      const struct inet6_dev *idev)
1566 {
1567 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1568 	if (!idev)
1569 		return false;
1570 	if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1571 	    !READ_ONCE(idev->cnf.optimistic_dad))
1572 		return false;
1573 
1574 	return true;
1575 #else
1576 	return false;
1577 #endif
1578 }
1579 
ipv6_get_saddr_eval(struct net * net,struct ipv6_saddr_score * score,struct ipv6_saddr_dst * dst,int i)1580 static int ipv6_get_saddr_eval(struct net *net,
1581 			       struct ipv6_saddr_score *score,
1582 			       struct ipv6_saddr_dst *dst,
1583 			       int i)
1584 {
1585 	int ret;
1586 
1587 	if (i <= score->rule) {
1588 		switch (i) {
1589 		case IPV6_SADDR_RULE_SCOPE:
1590 			ret = score->scopedist;
1591 			break;
1592 		case IPV6_SADDR_RULE_PREFIX:
1593 			ret = score->matchlen;
1594 			break;
1595 		default:
1596 			ret = !!test_bit(i, score->scorebits);
1597 		}
1598 		goto out;
1599 	}
1600 
1601 	switch (i) {
1602 	case IPV6_SADDR_RULE_INIT:
1603 		/* Rule 0: remember if hiscore is not ready yet */
1604 		ret = !!score->ifa;
1605 		break;
1606 	case IPV6_SADDR_RULE_LOCAL:
1607 		/* Rule 1: Prefer same address */
1608 		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1609 		break;
1610 	case IPV6_SADDR_RULE_SCOPE:
1611 		/* Rule 2: Prefer appropriate scope
1612 		 *
1613 		 *      ret
1614 		 *       ^
1615 		 *    -1 |  d 15
1616 		 *    ---+--+-+---> scope
1617 		 *       |
1618 		 *       |             d is scope of the destination.
1619 		 *  B-d  |  \
1620 		 *       |   \      <- smaller scope is better if
1621 		 *  B-15 |    \        if scope is enough for destination.
1622 		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1623 		 * d-C-1 | /
1624 		 *       |/         <- greater is better
1625 		 *   -C  /             if scope is not enough for destination.
1626 		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1627 		 *
1628 		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1629 		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1630 		 * Assume B = 0 and we get C > 29.
1631 		 */
1632 		ret = __ipv6_addr_src_scope(score->addr_type);
1633 		if (ret >= dst->scope)
1634 			ret = -ret;
1635 		else
1636 			ret -= 128;	/* 30 is enough */
1637 		score->scopedist = ret;
1638 		break;
1639 	case IPV6_SADDR_RULE_PREFERRED:
1640 	    {
1641 		/* Rule 3: Avoid deprecated and optimistic addresses */
1642 		u8 avoid = IFA_F_DEPRECATED;
1643 
1644 		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1645 			avoid |= IFA_F_OPTIMISTIC;
1646 		ret = ipv6_saddr_preferred(score->addr_type) ||
1647 		      !(score->ifa->flags & avoid);
1648 		break;
1649 	    }
1650 #ifdef CONFIG_IPV6_MIP6
1651 	case IPV6_SADDR_RULE_HOA:
1652 	    {
1653 		/* Rule 4: Prefer home address */
1654 		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1655 		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1656 		break;
1657 	    }
1658 #endif
1659 	case IPV6_SADDR_RULE_OIF:
1660 		/* Rule 5: Prefer outgoing interface */
1661 		ret = (!dst->ifindex ||
1662 		       dst->ifindex == score->ifa->idev->dev->ifindex);
1663 		break;
1664 	case IPV6_SADDR_RULE_LABEL:
1665 		/* Rule 6: Prefer matching label */
1666 		ret = ipv6_addr_label(net,
1667 				      &score->ifa->addr, score->addr_type,
1668 				      score->ifa->idev->dev->ifindex) == dst->label;
1669 		break;
1670 	case IPV6_SADDR_RULE_PRIVACY:
1671 	    {
1672 		/* Rule 7: Prefer public address
1673 		 * Note: prefer temporary address if use_tempaddr >= 2
1674 		 */
1675 		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1676 				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1677 				READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1678 		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1679 		break;
1680 	    }
1681 	case IPV6_SADDR_RULE_ORCHID:
1682 		/* Rule 8-: Prefer ORCHID vs ORCHID or
1683 		 *	    non-ORCHID vs non-ORCHID
1684 		 */
1685 		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1686 			ipv6_addr_orchid(dst->addr));
1687 		break;
1688 	case IPV6_SADDR_RULE_PREFIX:
1689 		/* Rule 8: Use longest matching prefix */
1690 		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1691 		if (ret > score->ifa->prefix_len)
1692 			ret = score->ifa->prefix_len;
1693 		score->matchlen = ret;
1694 		break;
1695 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1696 	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1697 		/* Optimistic addresses still have lower precedence than other
1698 		 * preferred addresses.
1699 		 */
1700 		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1701 		break;
1702 #endif
1703 	default:
1704 		ret = 0;
1705 	}
1706 
1707 	if (ret)
1708 		__set_bit(i, score->scorebits);
1709 	score->rule = i;
1710 out:
1711 	return ret;
1712 }
1713 
__ipv6_dev_get_saddr(struct net * net,struct ipv6_saddr_dst * dst,struct inet6_dev * idev,struct ipv6_saddr_score * scores,int hiscore_idx)1714 static int __ipv6_dev_get_saddr(struct net *net,
1715 				struct ipv6_saddr_dst *dst,
1716 				struct inet6_dev *idev,
1717 				struct ipv6_saddr_score *scores,
1718 				int hiscore_idx)
1719 {
1720 	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1721 
1722 	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1723 		int i;
1724 
1725 		/*
1726 		 * - Tentative Address (RFC2462 section 5.4)
1727 		 *  - A tentative address is not considered
1728 		 *    "assigned to an interface" in the traditional
1729 		 *    sense, unless it is also flagged as optimistic.
1730 		 * - Candidate Source Address (section 4)
1731 		 *  - In any case, anycast addresses, multicast
1732 		 *    addresses, and the unspecified address MUST
1733 		 *    NOT be included in a candidate set.
1734 		 */
1735 		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1736 		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1737 			continue;
1738 
1739 		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1740 
1741 		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1742 			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1743 			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1744 					    idev->dev->name);
1745 			continue;
1746 		}
1747 
1748 		score->rule = -1;
1749 		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1750 
1751 		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1752 			int minihiscore, miniscore;
1753 
1754 			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1755 			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1756 
1757 			if (minihiscore > miniscore) {
1758 				if (i == IPV6_SADDR_RULE_SCOPE &&
1759 				    score->scopedist > 0) {
1760 					/*
1761 					 * special case:
1762 					 * each remaining entry
1763 					 * has too small (not enough)
1764 					 * scope, because ifa entries
1765 					 * are sorted by their scope
1766 					 * values.
1767 					 */
1768 					goto out;
1769 				}
1770 				break;
1771 			} else if (minihiscore < miniscore) {
1772 				swap(hiscore, score);
1773 				hiscore_idx = 1 - hiscore_idx;
1774 
1775 				/* restore our iterator */
1776 				score->ifa = hiscore->ifa;
1777 
1778 				break;
1779 			}
1780 		}
1781 	}
1782 out:
1783 	return hiscore_idx;
1784 }
1785 
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)1786 static int ipv6_get_saddr_master(struct net *net,
1787 				 const struct net_device *dst_dev,
1788 				 const struct net_device *master,
1789 				 struct ipv6_saddr_dst *dst,
1790 				 struct ipv6_saddr_score *scores,
1791 				 int hiscore_idx)
1792 {
1793 	struct inet6_dev *idev;
1794 
1795 	idev = __in6_dev_get(dst_dev);
1796 	if (idev)
1797 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1798 						   scores, hiscore_idx);
1799 
1800 	idev = __in6_dev_get(master);
1801 	if (idev)
1802 		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1803 						   scores, hiscore_idx);
1804 
1805 	return hiscore_idx;
1806 }
1807 
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)1808 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1809 		       const struct in6_addr *daddr, unsigned int prefs,
1810 		       struct in6_addr *saddr)
1811 {
1812 	struct ipv6_saddr_score scores[2], *hiscore;
1813 	struct ipv6_saddr_dst dst;
1814 	struct inet6_dev *idev;
1815 	struct net_device *dev;
1816 	int dst_type;
1817 	bool use_oif_addr = false;
1818 	int hiscore_idx = 0;
1819 	int ret = 0;
1820 
1821 	dst_type = __ipv6_addr_type(daddr);
1822 	dst.addr = daddr;
1823 	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1824 	dst.scope = __ipv6_addr_src_scope(dst_type);
1825 	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1826 	dst.prefs = prefs;
1827 
1828 	scores[hiscore_idx].rule = -1;
1829 	scores[hiscore_idx].ifa = NULL;
1830 
1831 	rcu_read_lock();
1832 
1833 	/* Candidate Source Address (section 4)
1834 	 *  - multicast and link-local destination address,
1835 	 *    the set of candidate source address MUST only
1836 	 *    include addresses assigned to interfaces
1837 	 *    belonging to the same link as the outgoing
1838 	 *    interface.
1839 	 * (- For site-local destination addresses, the
1840 	 *    set of candidate source addresses MUST only
1841 	 *    include addresses assigned to interfaces
1842 	 *    belonging to the same site as the outgoing
1843 	 *    interface.)
1844 	 *  - "It is RECOMMENDED that the candidate source addresses
1845 	 *    be the set of unicast addresses assigned to the
1846 	 *    interface that will be used to send to the destination
1847 	 *    (the 'outgoing' interface)." (RFC 6724)
1848 	 */
1849 	if (dst_dev) {
1850 		idev = __in6_dev_get(dst_dev);
1851 		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1852 		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1853 		    (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1854 			use_oif_addr = true;
1855 		}
1856 	}
1857 
1858 	if (use_oif_addr) {
1859 		if (idev)
1860 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1861 	} else {
1862 		const struct net_device *master;
1863 		int master_idx = 0;
1864 
1865 		/* if dst_dev exists and is enslaved to an L3 device, then
1866 		 * prefer addresses from dst_dev and then the master over
1867 		 * any other enslaved devices in the L3 domain.
1868 		 */
1869 		master = l3mdev_master_dev_rcu(dst_dev);
1870 		if (master) {
1871 			master_idx = master->ifindex;
1872 
1873 			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1874 							    master, &dst,
1875 							    scores, hiscore_idx);
1876 
1877 			if (scores[hiscore_idx].ifa &&
1878 			    scores[hiscore_idx].scopedist >= 0)
1879 				goto out;
1880 		}
1881 
1882 		for_each_netdev_rcu(net, dev) {
1883 			/* only consider addresses on devices in the
1884 			 * same L3 domain
1885 			 */
1886 			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1887 				continue;
1888 			idev = __in6_dev_get(dev);
1889 			if (!idev)
1890 				continue;
1891 			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1892 		}
1893 	}
1894 
1895 out:
1896 	hiscore = &scores[hiscore_idx];
1897 	if (!hiscore->ifa)
1898 		ret = -EADDRNOTAVAIL;
1899 	else
1900 		*saddr = hiscore->ifa->addr;
1901 
1902 	rcu_read_unlock();
1903 	return ret;
1904 }
1905 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1906 
__ipv6_get_lladdr(struct inet6_dev * idev,struct in6_addr * addr,u32 banned_flags)1907 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1908 			      u32 banned_flags)
1909 {
1910 	struct inet6_ifaddr *ifp;
1911 	int err = -EADDRNOTAVAIL;
1912 
1913 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1914 		if (ifp->scope > IFA_LINK)
1915 			break;
1916 		if (ifp->scope == IFA_LINK &&
1917 		    !(ifp->flags & banned_flags)) {
1918 			*addr = ifp->addr;
1919 			err = 0;
1920 			break;
1921 		}
1922 	}
1923 	return err;
1924 }
1925 
ipv6_get_lladdr(struct net_device * dev,struct in6_addr * addr,u32 banned_flags)1926 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1927 		    u32 banned_flags)
1928 {
1929 	struct inet6_dev *idev;
1930 	int err = -EADDRNOTAVAIL;
1931 
1932 	rcu_read_lock();
1933 	idev = __in6_dev_get(dev);
1934 	if (idev) {
1935 		read_lock_bh(&idev->lock);
1936 		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1937 		read_unlock_bh(&idev->lock);
1938 	}
1939 	rcu_read_unlock();
1940 	return err;
1941 }
1942 
ipv6_count_addresses(const struct inet6_dev * idev)1943 static int ipv6_count_addresses(const struct inet6_dev *idev)
1944 {
1945 	const struct inet6_ifaddr *ifp;
1946 	int cnt = 0;
1947 
1948 	rcu_read_lock();
1949 	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1950 		cnt++;
1951 	rcu_read_unlock();
1952 	return cnt;
1953 }
1954 
ipv6_chk_addr(struct net * net,const struct in6_addr * addr,const struct net_device * dev,int strict)1955 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1956 		  const struct net_device *dev, int strict)
1957 {
1958 	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1959 				       strict, IFA_F_TENTATIVE);
1960 }
1961 EXPORT_SYMBOL(ipv6_chk_addr);
1962 
1963 /* device argument is used to find the L3 domain of interest. If
1964  * skip_dev_check is set, then the ifp device is not checked against
1965  * the passed in dev argument. So the 2 cases for addresses checks are:
1966  *   1. does the address exist in the L3 domain that dev is part of
1967  *      (skip_dev_check = true), or
1968  *
1969  *   2. does the address exist on the specific device
1970  *      (skip_dev_check = false)
1971  */
1972 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)1973 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1974 			  const struct net_device *dev, bool skip_dev_check,
1975 			  int strict, u32 banned_flags)
1976 {
1977 	unsigned int hash = inet6_addr_hash(net, addr);
1978 	struct net_device *l3mdev, *ndev;
1979 	struct inet6_ifaddr *ifp;
1980 	u32 ifp_flags;
1981 
1982 	rcu_read_lock();
1983 
1984 	l3mdev = l3mdev_master_dev_rcu(dev);
1985 	if (skip_dev_check)
1986 		dev = NULL;
1987 
1988 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1989 		ndev = ifp->idev->dev;
1990 
1991 		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1992 			continue;
1993 
1994 		/* Decouple optimistic from tentative for evaluation here.
1995 		 * Ban optimistic addresses explicitly, when required.
1996 		 */
1997 		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1998 			    ? (ifp->flags&~IFA_F_TENTATIVE)
1999 			    : ifp->flags;
2000 		if (ipv6_addr_equal(&ifp->addr, addr) &&
2001 		    !(ifp_flags&banned_flags) &&
2002 		    (!dev || ndev == dev ||
2003 		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2004 			rcu_read_unlock();
2005 			return ndev;
2006 		}
2007 	}
2008 
2009 	rcu_read_unlock();
2010 	return NULL;
2011 }
2012 
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)2013 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2014 			    const struct net_device *dev, bool skip_dev_check,
2015 			    int strict, u32 banned_flags)
2016 {
2017 	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2018 					 strict, banned_flags) ? 1 : 0;
2019 }
2020 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2021 
2022 
2023 /* Compares an address/prefix_len with addresses on device @dev.
2024  * If one is found it returns true.
2025  */
ipv6_chk_custom_prefix(const struct in6_addr * addr,const unsigned int prefix_len,struct net_device * dev)2026 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2027 	const unsigned int prefix_len, struct net_device *dev)
2028 {
2029 	const struct inet6_ifaddr *ifa;
2030 	const struct inet6_dev *idev;
2031 	bool ret = false;
2032 
2033 	rcu_read_lock();
2034 	idev = __in6_dev_get(dev);
2035 	if (idev) {
2036 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2037 			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2038 			if (ret)
2039 				break;
2040 		}
2041 	}
2042 	rcu_read_unlock();
2043 
2044 	return ret;
2045 }
2046 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2047 
ipv6_chk_prefix(const struct in6_addr * addr,struct net_device * dev)2048 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2049 {
2050 	const struct inet6_ifaddr *ifa;
2051 	const struct inet6_dev *idev;
2052 	int	onlink;
2053 
2054 	onlink = 0;
2055 	rcu_read_lock();
2056 	idev = __in6_dev_get(dev);
2057 	if (idev) {
2058 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2059 			onlink = ipv6_prefix_equal(addr, &ifa->addr,
2060 						   ifa->prefix_len);
2061 			if (onlink)
2062 				break;
2063 		}
2064 	}
2065 	rcu_read_unlock();
2066 	return onlink;
2067 }
2068 EXPORT_SYMBOL(ipv6_chk_prefix);
2069 
2070 /**
2071  * ipv6_dev_find - find the first device with a given source address.
2072  * @net: the net namespace
2073  * @addr: the source address
2074  * @dev: used to find the L3 domain of interest
2075  *
2076  * The caller should be protected by RCU, or RTNL.
2077  */
ipv6_dev_find(struct net * net,const struct in6_addr * addr,struct net_device * dev)2078 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2079 				 struct net_device *dev)
2080 {
2081 	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2082 					 IFA_F_TENTATIVE);
2083 }
2084 EXPORT_SYMBOL(ipv6_dev_find);
2085 
ipv6_get_ifaddr(struct net * net,const struct in6_addr * addr,struct net_device * dev,int strict)2086 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2087 				     struct net_device *dev, int strict)
2088 {
2089 	unsigned int hash = inet6_addr_hash(net, addr);
2090 	struct inet6_ifaddr *ifp, *result = NULL;
2091 
2092 	rcu_read_lock();
2093 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2094 		if (ipv6_addr_equal(&ifp->addr, addr)) {
2095 			if (!dev || ifp->idev->dev == dev ||
2096 			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2097 				if (in6_ifa_hold_safe(ifp)) {
2098 					result = ifp;
2099 					break;
2100 				}
2101 			}
2102 		}
2103 	}
2104 	rcu_read_unlock();
2105 
2106 	return result;
2107 }
2108 
2109 /* Gets referenced address, destroys ifaddr */
2110 
addrconf_dad_stop(struct inet6_ifaddr * ifp,int dad_failed)2111 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2112 {
2113 	if (dad_failed)
2114 		ifp->flags |= IFA_F_DADFAILED;
2115 
2116 	if (ifp->flags&IFA_F_TEMPORARY) {
2117 		struct inet6_ifaddr *ifpub;
2118 		spin_lock_bh(&ifp->lock);
2119 		ifpub = ifp->ifpub;
2120 		if (ifpub) {
2121 			in6_ifa_hold(ifpub);
2122 			spin_unlock_bh(&ifp->lock);
2123 			ipv6_create_tempaddr(ifpub, true);
2124 			in6_ifa_put(ifpub);
2125 		} else {
2126 			spin_unlock_bh(&ifp->lock);
2127 		}
2128 		ipv6_del_addr(ifp);
2129 	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2130 		spin_lock_bh(&ifp->lock);
2131 		addrconf_del_dad_work(ifp);
2132 		ifp->flags |= IFA_F_TENTATIVE;
2133 		if (dad_failed)
2134 			ifp->flags &= ~IFA_F_OPTIMISTIC;
2135 		spin_unlock_bh(&ifp->lock);
2136 		if (dad_failed)
2137 			ipv6_ifa_notify(0, ifp);
2138 		in6_ifa_put(ifp);
2139 	} else {
2140 		ipv6_del_addr(ifp);
2141 	}
2142 }
2143 
addrconf_dad_end(struct inet6_ifaddr * ifp)2144 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2145 {
2146 	int err = -ENOENT;
2147 
2148 	spin_lock_bh(&ifp->lock);
2149 	if (ifp->state == INET6_IFADDR_STATE_DAD) {
2150 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
2151 		err = 0;
2152 	}
2153 	spin_unlock_bh(&ifp->lock);
2154 
2155 	return err;
2156 }
2157 
addrconf_dad_failure(struct sk_buff * skb,struct inet6_ifaddr * ifp)2158 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2159 {
2160 	struct inet6_dev *idev = ifp->idev;
2161 	struct net *net = dev_net(idev->dev);
2162 	int max_addresses;
2163 
2164 	if (addrconf_dad_end(ifp)) {
2165 		in6_ifa_put(ifp);
2166 		return;
2167 	}
2168 
2169 	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2170 			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2171 
2172 	spin_lock_bh(&ifp->lock);
2173 
2174 	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2175 		struct in6_addr new_addr;
2176 		struct inet6_ifaddr *ifp2;
2177 		int retries = ifp->stable_privacy_retry + 1;
2178 		struct ifa6_config cfg = {
2179 			.pfx = &new_addr,
2180 			.plen = ifp->prefix_len,
2181 			.ifa_flags = ifp->flags,
2182 			.valid_lft = ifp->valid_lft,
2183 			.preferred_lft = ifp->prefered_lft,
2184 			.scope = ifp->scope,
2185 		};
2186 
2187 		if (retries > net->ipv6.sysctl.idgen_retries) {
2188 			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2189 					     ifp->idev->dev->name);
2190 			goto errdad;
2191 		}
2192 
2193 		new_addr = ifp->addr;
2194 		if (ipv6_generate_stable_address(&new_addr, retries,
2195 						 idev))
2196 			goto errdad;
2197 
2198 		spin_unlock_bh(&ifp->lock);
2199 
2200 		max_addresses = READ_ONCE(idev->cnf.max_addresses);
2201 		if (max_addresses &&
2202 		    ipv6_count_addresses(idev) >= max_addresses)
2203 			goto lock_errdad;
2204 
2205 		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2206 				     ifp->idev->dev->name);
2207 
2208 		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2209 		if (IS_ERR(ifp2))
2210 			goto lock_errdad;
2211 
2212 		spin_lock_bh(&ifp2->lock);
2213 		ifp2->stable_privacy_retry = retries;
2214 		ifp2->state = INET6_IFADDR_STATE_PREDAD;
2215 		spin_unlock_bh(&ifp2->lock);
2216 
2217 		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2218 		in6_ifa_put(ifp2);
2219 lock_errdad:
2220 		spin_lock_bh(&ifp->lock);
2221 	}
2222 
2223 errdad:
2224 	/* transition from _POSTDAD to _ERRDAD */
2225 	ifp->state = INET6_IFADDR_STATE_ERRDAD;
2226 	spin_unlock_bh(&ifp->lock);
2227 
2228 	addrconf_mod_dad_work(ifp, 0);
2229 	in6_ifa_put(ifp);
2230 }
2231 
2232 /* Join to solicited addr multicast group.
2233  * caller must hold RTNL */
addrconf_join_solict(struct net_device * dev,const struct in6_addr * addr)2234 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2235 {
2236 	struct in6_addr maddr;
2237 
2238 	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2239 		return;
2240 
2241 	addrconf_addr_solict_mult(addr, &maddr);
2242 	ipv6_dev_mc_inc(dev, &maddr);
2243 }
2244 
2245 /* caller must hold RTNL */
addrconf_leave_solict(struct inet6_dev * idev,const struct in6_addr * addr)2246 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2247 {
2248 	struct in6_addr maddr;
2249 
2250 	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2251 		return;
2252 
2253 	addrconf_addr_solict_mult(addr, &maddr);
2254 	__ipv6_dev_mc_dec(idev, &maddr);
2255 }
2256 
2257 /* caller must hold RTNL */
addrconf_join_anycast(struct inet6_ifaddr * ifp)2258 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2259 {
2260 	struct in6_addr addr;
2261 
2262 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2263 		return;
2264 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2265 	if (ipv6_addr_any(&addr))
2266 		return;
2267 	__ipv6_dev_ac_inc(ifp->idev, &addr);
2268 }
2269 
2270 /* caller must hold RTNL */
addrconf_leave_anycast(struct inet6_ifaddr * ifp)2271 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2272 {
2273 	struct in6_addr addr;
2274 
2275 	if (ifp->prefix_len >= 127) /* RFC 6164 */
2276 		return;
2277 	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2278 	if (ipv6_addr_any(&addr))
2279 		return;
2280 	__ipv6_dev_ac_dec(ifp->idev, &addr);
2281 }
2282 
addrconf_ifid_6lowpan(u8 * eui,struct net_device * dev)2283 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2284 {
2285 	switch (dev->addr_len) {
2286 	case ETH_ALEN:
2287 		memcpy(eui, dev->dev_addr, 3);
2288 		eui[3] = 0xFF;
2289 		eui[4] = 0xFE;
2290 		memcpy(eui + 5, dev->dev_addr + 3, 3);
2291 		break;
2292 	case EUI64_ADDR_LEN:
2293 		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2294 		eui[0] ^= 2;
2295 		break;
2296 	default:
2297 		return -1;
2298 	}
2299 
2300 	return 0;
2301 }
2302 
addrconf_ifid_ieee1394(u8 * eui,struct net_device * dev)2303 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2304 {
2305 	const union fwnet_hwaddr *ha;
2306 
2307 	if (dev->addr_len != FWNET_ALEN)
2308 		return -1;
2309 
2310 	ha = (const union fwnet_hwaddr *)dev->dev_addr;
2311 
2312 	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2313 	eui[0] ^= 2;
2314 	return 0;
2315 }
2316 
addrconf_ifid_arcnet(u8 * eui,struct net_device * dev)2317 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2318 {
2319 	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
2320 	if (dev->addr_len != ARCNET_ALEN)
2321 		return -1;
2322 	memset(eui, 0, 7);
2323 	eui[7] = *(u8 *)dev->dev_addr;
2324 	return 0;
2325 }
2326 
addrconf_ifid_infiniband(u8 * eui,struct net_device * dev)2327 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2328 {
2329 	if (dev->addr_len != INFINIBAND_ALEN)
2330 		return -1;
2331 	memcpy(eui, dev->dev_addr + 12, 8);
2332 	eui[0] |= 2;
2333 	return 0;
2334 }
2335 
__ipv6_isatap_ifid(u8 * eui,__be32 addr)2336 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2337 {
2338 	if (addr == 0)
2339 		return -1;
2340 	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2341 		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2342 		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2343 		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2344 		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2345 		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2346 	eui[1] = 0;
2347 	eui[2] = 0x5E;
2348 	eui[3] = 0xFE;
2349 	memcpy(eui + 4, &addr, 4);
2350 	return 0;
2351 }
2352 
addrconf_ifid_sit(u8 * eui,struct net_device * dev)2353 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2354 {
2355 	if (dev->priv_flags & IFF_ISATAP)
2356 		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2357 	return -1;
2358 }
2359 
addrconf_ifid_gre(u8 * eui,struct net_device * dev)2360 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2361 {
2362 	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2363 }
2364 
addrconf_ifid_ip6tnl(u8 * eui,struct net_device * dev)2365 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2366 {
2367 	memcpy(eui, dev->perm_addr, 3);
2368 	memcpy(eui + 5, dev->perm_addr + 3, 3);
2369 	eui[3] = 0xFF;
2370 	eui[4] = 0xFE;
2371 	eui[0] ^= 2;
2372 	return 0;
2373 }
2374 
ipv6_generate_eui64(u8 * eui,struct net_device * dev)2375 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2376 {
2377 	switch (dev->type) {
2378 	case ARPHRD_ETHER:
2379 	case ARPHRD_FDDI:
2380 		return addrconf_ifid_eui48(eui, dev);
2381 	case ARPHRD_ARCNET:
2382 		return addrconf_ifid_arcnet(eui, dev);
2383 	case ARPHRD_INFINIBAND:
2384 		return addrconf_ifid_infiniband(eui, dev);
2385 	case ARPHRD_SIT:
2386 		return addrconf_ifid_sit(eui, dev);
2387 	case ARPHRD_IPGRE:
2388 	case ARPHRD_TUNNEL:
2389 		return addrconf_ifid_gre(eui, dev);
2390 	case ARPHRD_6LOWPAN:
2391 		return addrconf_ifid_6lowpan(eui, dev);
2392 	case ARPHRD_IEEE1394:
2393 		return addrconf_ifid_ieee1394(eui, dev);
2394 	case ARPHRD_TUNNEL6:
2395 	case ARPHRD_IP6GRE:
2396 	case ARPHRD_RAWIP:
2397 		return addrconf_ifid_ip6tnl(eui, dev);
2398 	}
2399 	return -1;
2400 }
2401 
ipv6_inherit_eui64(u8 * eui,struct inet6_dev * idev)2402 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2403 {
2404 	int err = -1;
2405 	struct inet6_ifaddr *ifp;
2406 
2407 	read_lock_bh(&idev->lock);
2408 	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2409 		if (ifp->scope > IFA_LINK)
2410 			break;
2411 		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2412 			memcpy(eui, ifp->addr.s6_addr+8, 8);
2413 			err = 0;
2414 			break;
2415 		}
2416 	}
2417 	read_unlock_bh(&idev->lock);
2418 	return err;
2419 }
2420 
2421 /* Generation of a randomized Interface Identifier
2422  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2423  */
2424 
ipv6_gen_rnd_iid(struct in6_addr * addr)2425 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2426 {
2427 regen:
2428 	get_random_bytes(&addr->s6_addr[8], 8);
2429 
2430 	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2431 	 * check if generated address is not inappropriate:
2432 	 *
2433 	 * - Reserved IPv6 Interface Identifiers
2434 	 * - XXX: already assigned to an address on the device
2435 	 */
2436 
2437 	/* Subnet-router anycast: 0000:0000:0000:0000 */
2438 	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2439 		goto regen;
2440 
2441 	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2442 	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2443 	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2444 	 */
2445 	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2446 	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2447 		goto regen;
2448 
2449 	/* Reserved subnet anycast addresses */
2450 	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2451 	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2452 		goto regen;
2453 }
2454 
2455 /*
2456  *	Add prefix route.
2457  */
2458 
2459 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)2460 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2461 		      struct net_device *dev, unsigned long expires,
2462 		      u32 flags, gfp_t gfp_flags)
2463 {
2464 	struct fib6_config cfg = {
2465 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2466 		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2467 		.fc_ifindex = dev->ifindex,
2468 		.fc_expires = expires,
2469 		.fc_dst_len = plen,
2470 		.fc_flags = RTF_UP | flags,
2471 		.fc_nlinfo.nl_net = dev_net(dev),
2472 		.fc_protocol = RTPROT_KERNEL,
2473 		.fc_type = RTN_UNICAST,
2474 	};
2475 
2476 	cfg.fc_dst = *pfx;
2477 
2478 	/* Prevent useless cloning on PtP SIT.
2479 	   This thing is done here expecting that the whole
2480 	   class of non-broadcast devices need not cloning.
2481 	 */
2482 #if IS_ENABLED(CONFIG_IPV6_SIT)
2483 	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2484 		cfg.fc_flags |= RTF_NONEXTHOP;
2485 #endif
2486 
2487 	ip6_route_add(&cfg, gfp_flags, NULL);
2488 }
2489 
2490 
addrconf_get_prefix_route(const struct in6_addr * pfx,int plen,const struct net_device * dev,u32 flags,u32 noflags,bool no_gw)2491 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2492 						  int plen,
2493 						  const struct net_device *dev,
2494 						  u32 flags, u32 noflags,
2495 						  bool no_gw)
2496 {
2497 	struct fib6_node *fn;
2498 	struct fib6_info *rt = NULL;
2499 	struct fib6_table *table;
2500 	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2501 
2502 	table = fib6_get_table(dev_net(dev), tb_id);
2503 	if (!table)
2504 		return NULL;
2505 
2506 	rcu_read_lock();
2507 	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2508 	if (!fn)
2509 		goto out;
2510 
2511 	for_each_fib6_node_rt_rcu(fn) {
2512 		/* prefix routes only use builtin fib6_nh */
2513 		if (rt->nh)
2514 			continue;
2515 
2516 		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2517 			continue;
2518 		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2519 			continue;
2520 		if ((rt->fib6_flags & flags) != flags)
2521 			continue;
2522 		if ((rt->fib6_flags & noflags) != 0)
2523 			continue;
2524 		if (!fib6_info_hold_safe(rt))
2525 			continue;
2526 		break;
2527 	}
2528 out:
2529 	rcu_read_unlock();
2530 	return rt;
2531 }
2532 
2533 
2534 /* Create "default" multicast route to the interface */
2535 
addrconf_add_mroute(struct net_device * dev)2536 static void addrconf_add_mroute(struct net_device *dev)
2537 {
2538 	struct fib6_config cfg = {
2539 		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2540 		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2541 		.fc_ifindex = dev->ifindex,
2542 		.fc_dst_len = 8,
2543 		.fc_flags = RTF_UP,
2544 		.fc_type = RTN_MULTICAST,
2545 		.fc_nlinfo.nl_net = dev_net(dev),
2546 		.fc_protocol = RTPROT_KERNEL,
2547 	};
2548 
2549 	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2550 
2551 	ip6_route_add(&cfg, GFP_KERNEL, NULL);
2552 }
2553 
addrconf_add_dev(struct net_device * dev)2554 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2555 {
2556 	struct inet6_dev *idev;
2557 
2558 	ASSERT_RTNL();
2559 
2560 	idev = ipv6_find_idev(dev);
2561 	if (IS_ERR(idev))
2562 		return idev;
2563 
2564 	if (idev->cnf.disable_ipv6)
2565 		return ERR_PTR(-EACCES);
2566 
2567 	/* Add default multicast route */
2568 	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2569 		addrconf_add_mroute(dev);
2570 
2571 	return idev;
2572 }
2573 
delete_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp)2574 static void delete_tempaddrs(struct inet6_dev *idev,
2575 			     struct inet6_ifaddr *ifp)
2576 {
2577 	struct inet6_ifaddr *ift, *tmp;
2578 
2579 	write_lock_bh(&idev->lock);
2580 	list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) {
2581 		if (ift->ifpub != ifp)
2582 			continue;
2583 
2584 		in6_ifa_hold(ift);
2585 		write_unlock_bh(&idev->lock);
2586 		ipv6_del_addr(ift);
2587 		write_lock_bh(&idev->lock);
2588 	}
2589 	write_unlock_bh(&idev->lock);
2590 }
2591 
manage_tempaddrs(struct inet6_dev * idev,struct inet6_ifaddr * ifp,__u32 valid_lft,__u32 prefered_lft,bool create,unsigned long now)2592 static void manage_tempaddrs(struct inet6_dev *idev,
2593 			     struct inet6_ifaddr *ifp,
2594 			     __u32 valid_lft, __u32 prefered_lft,
2595 			     bool create, unsigned long now)
2596 {
2597 	u32 flags;
2598 	struct inet6_ifaddr *ift;
2599 
2600 	read_lock_bh(&idev->lock);
2601 	/* update all temporary addresses in the list */
2602 	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2603 		int age, max_valid, max_prefered;
2604 
2605 		if (ifp != ift->ifpub)
2606 			continue;
2607 
2608 		/* RFC 4941 section 3.3:
2609 		 * If a received option will extend the lifetime of a public
2610 		 * address, the lifetimes of temporary addresses should
2611 		 * be extended, subject to the overall constraint that no
2612 		 * temporary addresses should ever remain "valid" or "preferred"
2613 		 * for a time longer than (TEMP_VALID_LIFETIME) or
2614 		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2615 		 */
2616 		age = (now - ift->cstamp) / HZ;
2617 		max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2618 		if (max_valid < 0)
2619 			max_valid = 0;
2620 
2621 		max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2622 			       idev->desync_factor - age;
2623 		if (max_prefered < 0)
2624 			max_prefered = 0;
2625 
2626 		if (valid_lft > max_valid)
2627 			valid_lft = max_valid;
2628 
2629 		if (prefered_lft > max_prefered)
2630 			prefered_lft = max_prefered;
2631 
2632 		spin_lock(&ift->lock);
2633 		flags = ift->flags;
2634 		ift->valid_lft = valid_lft;
2635 		ift->prefered_lft = prefered_lft;
2636 		ift->tstamp = now;
2637 		if (prefered_lft > 0)
2638 			ift->flags &= ~IFA_F_DEPRECATED;
2639 
2640 		spin_unlock(&ift->lock);
2641 		if (!(flags&IFA_F_TENTATIVE))
2642 			ipv6_ifa_notify(0, ift);
2643 	}
2644 
2645 	/* Also create a temporary address if it's enabled but no temporary
2646 	 * address currently exists.
2647 	 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2648 	 * as part of cleanup (ie. deleting the mngtmpaddr).
2649 	 * We don't want that to result in creating a new temporary ip address.
2650 	 */
2651 	if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2652 		create = true;
2653 
2654 	if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2655 		/* When a new public address is created as described
2656 		 * in [ADDRCONF], also create a new temporary address.
2657 		 */
2658 		read_unlock_bh(&idev->lock);
2659 		ipv6_create_tempaddr(ifp, false);
2660 	} else {
2661 		read_unlock_bh(&idev->lock);
2662 	}
2663 }
2664 
is_addr_mode_generate_stable(struct inet6_dev * idev)2665 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2666 {
2667 	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2668 	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2669 }
2670 
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)2671 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2672 				 const struct prefix_info *pinfo,
2673 				 struct inet6_dev *in6_dev,
2674 				 const struct in6_addr *addr, int addr_type,
2675 				 u32 addr_flags, bool sllao, bool tokenized,
2676 				 __u32 valid_lft, u32 prefered_lft)
2677 {
2678 	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2679 	int create = 0, update_lft = 0;
2680 
2681 	if (!ifp && valid_lft) {
2682 		int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2683 		struct ifa6_config cfg = {
2684 			.pfx = addr,
2685 			.plen = pinfo->prefix_len,
2686 			.ifa_flags = addr_flags,
2687 			.valid_lft = valid_lft,
2688 			.preferred_lft = prefered_lft,
2689 			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2690 			.ifa_proto = IFAPROT_KERNEL_RA
2691 		};
2692 
2693 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2694 		if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2695 		     READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2696 		    !net->ipv6.devconf_all->forwarding && sllao)
2697 			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2698 #endif
2699 
2700 		/* Do not allow to create too much of autoconfigured
2701 		 * addresses; this would be too easy way to crash kernel.
2702 		 */
2703 		if (!max_addresses ||
2704 		    ipv6_count_addresses(in6_dev) < max_addresses)
2705 			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2706 
2707 		if (IS_ERR_OR_NULL(ifp))
2708 			return -1;
2709 
2710 		create = 1;
2711 		spin_lock_bh(&ifp->lock);
2712 		ifp->flags |= IFA_F_MANAGETEMPADDR;
2713 		ifp->cstamp = jiffies;
2714 		ifp->tokenized = tokenized;
2715 		spin_unlock_bh(&ifp->lock);
2716 		addrconf_dad_start(ifp);
2717 	}
2718 
2719 	if (ifp) {
2720 		u32 flags;
2721 		unsigned long now;
2722 		u32 stored_lft;
2723 
2724 		/* update lifetime (RFC2462 5.5.3 e) */
2725 		spin_lock_bh(&ifp->lock);
2726 		now = jiffies;
2727 		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2728 			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2729 		else
2730 			stored_lft = 0;
2731 
2732 		/* RFC4862 Section 5.5.3e:
2733 		 * "Note that the preferred lifetime of the
2734 		 *  corresponding address is always reset to
2735 		 *  the Preferred Lifetime in the received
2736 		 *  Prefix Information option, regardless of
2737 		 *  whether the valid lifetime is also reset or
2738 		 *  ignored."
2739 		 *
2740 		 * So we should always update prefered_lft here.
2741 		 */
2742 		update_lft = !create && stored_lft;
2743 
2744 		if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2745 			const u32 minimum_lft = min_t(u32,
2746 				stored_lft, MIN_VALID_LIFETIME);
2747 			valid_lft = max(valid_lft, minimum_lft);
2748 		}
2749 
2750 		if (update_lft) {
2751 			ifp->valid_lft = valid_lft;
2752 			ifp->prefered_lft = prefered_lft;
2753 			WRITE_ONCE(ifp->tstamp, now);
2754 			flags = ifp->flags;
2755 			ifp->flags &= ~IFA_F_DEPRECATED;
2756 			spin_unlock_bh(&ifp->lock);
2757 
2758 			if (!(flags&IFA_F_TENTATIVE))
2759 				ipv6_ifa_notify(0, ifp);
2760 		} else
2761 			spin_unlock_bh(&ifp->lock);
2762 
2763 		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2764 				 create, now);
2765 
2766 		in6_ifa_put(ifp);
2767 		addrconf_verify(net);
2768 	}
2769 
2770 	return 0;
2771 }
2772 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2773 
addrconf_prefix_rcv(struct net_device * dev,u8 * opt,int len,bool sllao)2774 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2775 {
2776 	struct prefix_info *pinfo;
2777 	struct fib6_table *table;
2778 	__u32 valid_lft;
2779 	__u32 prefered_lft;
2780 	int addr_type, err;
2781 	u32 addr_flags = 0;
2782 	struct inet6_dev *in6_dev;
2783 	struct net *net = dev_net(dev);
2784 	bool ignore_autoconf = false;
2785 
2786 	pinfo = (struct prefix_info *) opt;
2787 
2788 	if (len < sizeof(struct prefix_info)) {
2789 		netdev_dbg(dev, "addrconf: prefix option too short\n");
2790 		return;
2791 	}
2792 
2793 	/*
2794 	 *	Validation checks ([ADDRCONF], page 19)
2795 	 */
2796 
2797 	addr_type = ipv6_addr_type(&pinfo->prefix);
2798 
2799 	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2800 		return;
2801 
2802 	valid_lft = ntohl(pinfo->valid);
2803 	prefered_lft = ntohl(pinfo->prefered);
2804 
2805 	if (prefered_lft > valid_lft) {
2806 		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2807 		return;
2808 	}
2809 
2810 	in6_dev = in6_dev_get(dev);
2811 
2812 	if (!in6_dev) {
2813 		net_dbg_ratelimited("addrconf: device %s not configured\n",
2814 				    dev->name);
2815 		return;
2816 	}
2817 
2818 	if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2819 		goto put;
2820 
2821 	/*
2822 	 *	Two things going on here:
2823 	 *	1) Add routes for on-link prefixes
2824 	 *	2) Configure prefixes with the auto flag set
2825 	 */
2826 
2827 	if (pinfo->onlink) {
2828 		struct fib6_info *rt;
2829 		unsigned long rt_expires;
2830 
2831 		/* Avoid arithmetic overflow. Really, we could
2832 		 * save rt_expires in seconds, likely valid_lft,
2833 		 * but it would require division in fib gc, that it
2834 		 * not good.
2835 		 */
2836 		if (HZ > USER_HZ)
2837 			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2838 		else
2839 			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2840 
2841 		if (addrconf_finite_timeout(rt_expires))
2842 			rt_expires *= HZ;
2843 
2844 		rt = addrconf_get_prefix_route(&pinfo->prefix,
2845 					       pinfo->prefix_len,
2846 					       dev,
2847 					       RTF_ADDRCONF | RTF_PREFIX_RT,
2848 					       RTF_DEFAULT, true);
2849 
2850 		if (rt) {
2851 			/* Autoconf prefix route */
2852 			if (valid_lft == 0) {
2853 				ip6_del_rt(net, rt, false);
2854 				rt = NULL;
2855 			} else {
2856 				table = rt->fib6_table;
2857 				spin_lock_bh(&table->tb6_lock);
2858 
2859 				if (addrconf_finite_timeout(rt_expires)) {
2860 					/* not infinity */
2861 					fib6_set_expires(rt, jiffies + rt_expires);
2862 					fib6_add_gc_list(rt);
2863 				} else {
2864 					fib6_clean_expires(rt);
2865 					fib6_remove_gc_list(rt);
2866 				}
2867 
2868 				spin_unlock_bh(&table->tb6_lock);
2869 			}
2870 		} else if (valid_lft) {
2871 			clock_t expires = 0;
2872 			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2873 			if (addrconf_finite_timeout(rt_expires)) {
2874 				/* not infinity */
2875 				flags |= RTF_EXPIRES;
2876 				expires = jiffies_to_clock_t(rt_expires);
2877 			}
2878 			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2879 					      0, dev, expires, flags,
2880 					      GFP_ATOMIC);
2881 		}
2882 		fib6_info_release(rt);
2883 	}
2884 
2885 	/* Try to figure out our local address for this prefix */
2886 
2887 	ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd;
2888 	if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) {
2889 		struct in6_addr addr;
2890 		bool tokenized = false, dev_addr_generated = false;
2891 
2892 		if (pinfo->prefix_len == 64) {
2893 			memcpy(&addr, &pinfo->prefix, 8);
2894 
2895 			if (!ipv6_addr_any(&in6_dev->token)) {
2896 				read_lock_bh(&in6_dev->lock);
2897 				memcpy(addr.s6_addr + 8,
2898 				       in6_dev->token.s6_addr + 8, 8);
2899 				read_unlock_bh(&in6_dev->lock);
2900 				tokenized = true;
2901 			} else if (is_addr_mode_generate_stable(in6_dev) &&
2902 				   !ipv6_generate_stable_address(&addr, 0,
2903 								 in6_dev)) {
2904 				addr_flags |= IFA_F_STABLE_PRIVACY;
2905 				goto ok;
2906 			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2907 				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2908 				goto put;
2909 			} else {
2910 				dev_addr_generated = true;
2911 			}
2912 			goto ok;
2913 		}
2914 		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2915 				    pinfo->prefix_len);
2916 		goto put;
2917 
2918 ok:
2919 		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2920 						   &addr, addr_type,
2921 						   addr_flags, sllao,
2922 						   tokenized, valid_lft,
2923 						   prefered_lft);
2924 		if (err)
2925 			goto put;
2926 
2927 		/* Ignore error case here because previous prefix add addr was
2928 		 * successful which will be notified.
2929 		 */
2930 		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2931 					      addr_type, addr_flags, sllao,
2932 					      tokenized, valid_lft,
2933 					      prefered_lft,
2934 					      dev_addr_generated);
2935 	}
2936 	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2937 put:
2938 	in6_dev_put(in6_dev);
2939 }
2940 
addrconf_set_sit_dstaddr(struct net * net,struct net_device * dev,struct in6_ifreq * ireq)2941 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2942 		struct in6_ifreq *ireq)
2943 {
2944 	struct ip_tunnel_parm_kern p = { };
2945 	int err;
2946 
2947 	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2948 		return -EADDRNOTAVAIL;
2949 
2950 	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2951 	p.iph.version = 4;
2952 	p.iph.ihl = 5;
2953 	p.iph.protocol = IPPROTO_IPV6;
2954 	p.iph.ttl = 64;
2955 
2956 	if (!dev->netdev_ops->ndo_tunnel_ctl)
2957 		return -EOPNOTSUPP;
2958 	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2959 	if (err)
2960 		return err;
2961 
2962 	dev = __dev_get_by_name(net, p.name);
2963 	if (!dev)
2964 		return -ENOBUFS;
2965 	return dev_open(dev, NULL);
2966 }
2967 
2968 /*
2969  *	Set destination address.
2970  *	Special case for SIT interfaces where we create a new "virtual"
2971  *	device.
2972  */
addrconf_set_dstaddr(struct net * net,void __user * arg)2973 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2974 {
2975 	struct net_device *dev;
2976 	struct in6_ifreq ireq;
2977 	int err = -ENODEV;
2978 
2979 	if (!IS_ENABLED(CONFIG_IPV6_SIT))
2980 		return -ENODEV;
2981 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2982 		return -EFAULT;
2983 
2984 	rtnl_net_lock(net);
2985 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2986 	if (dev && dev->type == ARPHRD_SIT)
2987 		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2988 	rtnl_net_unlock(net);
2989 	return err;
2990 }
2991 
ipv6_mc_config(struct sock * sk,bool join,const struct in6_addr * addr,int ifindex)2992 static int ipv6_mc_config(struct sock *sk, bool join,
2993 			  const struct in6_addr *addr, int ifindex)
2994 {
2995 	int ret;
2996 
2997 	ASSERT_RTNL();
2998 
2999 	lock_sock(sk);
3000 	if (join)
3001 		ret = ipv6_sock_mc_join(sk, ifindex, addr);
3002 	else
3003 		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
3004 	release_sock(sk);
3005 
3006 	return ret;
3007 }
3008 
3009 /*
3010  *	Manual configuration of address on an interface
3011  */
inet6_addr_add(struct net * net,struct net_device * dev,struct ifa6_config * cfg,clock_t expires,u32 flags,struct netlink_ext_ack * extack)3012 static int inet6_addr_add(struct net *net, struct net_device *dev,
3013 			  struct ifa6_config *cfg, clock_t expires, u32 flags,
3014 			  struct netlink_ext_ack *extack)
3015 {
3016 	struct inet6_ifaddr *ifp;
3017 	struct inet6_dev *idev;
3018 
3019 	ASSERT_RTNL_NET(net);
3020 
3021 	if (cfg->plen > 128) {
3022 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3023 		return -EINVAL;
3024 	}
3025 
3026 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3027 		NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3028 		return -EINVAL;
3029 	}
3030 
3031 	idev = addrconf_add_dev(dev);
3032 	if (IS_ERR(idev)) {
3033 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3034 		return PTR_ERR(idev);
3035 	}
3036 
3037 	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3038 		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3039 					 true, cfg->pfx, dev->ifindex);
3040 
3041 		if (ret < 0) {
3042 			NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3043 			return ret;
3044 		}
3045 	}
3046 
3047 	cfg->scope = ipv6_addr_scope(cfg->pfx);
3048 
3049 	ifp = ipv6_add_addr(idev, cfg, true, extack);
3050 	if (!IS_ERR(ifp)) {
3051 		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3052 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3053 					      ifp->rt_priority, dev, expires,
3054 					      flags, GFP_KERNEL);
3055 		}
3056 
3057 		/* Send a netlink notification if DAD is enabled and
3058 		 * optimistic flag is not set
3059 		 */
3060 		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3061 			ipv6_ifa_notify(0, ifp);
3062 		/*
3063 		 * Note that section 3.1 of RFC 4429 indicates
3064 		 * that the Optimistic flag should not be set for
3065 		 * manually configured addresses
3066 		 */
3067 		addrconf_dad_start(ifp);
3068 		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3069 			manage_tempaddrs(idev, ifp, cfg->valid_lft,
3070 					 cfg->preferred_lft, true, jiffies);
3071 		in6_ifa_put(ifp);
3072 		addrconf_verify_rtnl(net);
3073 		return 0;
3074 	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3075 		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3076 			       cfg->pfx, dev->ifindex);
3077 	}
3078 
3079 	return PTR_ERR(ifp);
3080 }
3081 
inet6_addr_del(struct net * net,int ifindex,u32 ifa_flags,const struct in6_addr * pfx,unsigned int plen,struct netlink_ext_ack * extack)3082 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3083 			  const struct in6_addr *pfx, unsigned int plen,
3084 			  struct netlink_ext_ack *extack)
3085 {
3086 	struct inet6_ifaddr *ifp;
3087 	struct inet6_dev *idev;
3088 	struct net_device *dev;
3089 
3090 	if (plen > 128) {
3091 		NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3092 		return -EINVAL;
3093 	}
3094 
3095 	dev = __dev_get_by_index(net, ifindex);
3096 	if (!dev) {
3097 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3098 		return -ENODEV;
3099 	}
3100 
3101 	idev = __in6_dev_get_rtnl_net(dev);
3102 	if (!idev) {
3103 		NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3104 		return -ENXIO;
3105 	}
3106 
3107 	read_lock_bh(&idev->lock);
3108 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3109 		if (ifp->prefix_len == plen &&
3110 		    ipv6_addr_equal(pfx, &ifp->addr)) {
3111 			in6_ifa_hold(ifp);
3112 			read_unlock_bh(&idev->lock);
3113 
3114 			ipv6_del_addr(ifp);
3115 
3116 			if (!(ifp->flags & IFA_F_TEMPORARY) &&
3117 			    (ifp->flags & IFA_F_MANAGETEMPADDR))
3118 				delete_tempaddrs(idev, ifp);
3119 
3120 			addrconf_verify_rtnl(net);
3121 			if (ipv6_addr_is_multicast(pfx)) {
3122 				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3123 					       false, pfx, dev->ifindex);
3124 			}
3125 			return 0;
3126 		}
3127 	}
3128 	read_unlock_bh(&idev->lock);
3129 
3130 	NL_SET_ERR_MSG_MOD(extack, "address not found");
3131 	return -EADDRNOTAVAIL;
3132 }
3133 
3134 
addrconf_add_ifaddr(struct net * net,void __user * arg)3135 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3136 {
3137 	struct ifa6_config cfg = {
3138 		.ifa_flags = IFA_F_PERMANENT,
3139 		.preferred_lft = INFINITY_LIFE_TIME,
3140 		.valid_lft = INFINITY_LIFE_TIME,
3141 	};
3142 	struct net_device *dev;
3143 	struct in6_ifreq ireq;
3144 	int err;
3145 
3146 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3147 		return -EPERM;
3148 
3149 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3150 		return -EFAULT;
3151 
3152 	cfg.pfx = &ireq.ifr6_addr;
3153 	cfg.plen = ireq.ifr6_prefixlen;
3154 
3155 	rtnl_net_lock(net);
3156 	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
3157 	if (dev) {
3158 		netdev_lock_ops(dev);
3159 		err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL);
3160 		netdev_unlock_ops(dev);
3161 	} else {
3162 		err = -ENODEV;
3163 	}
3164 	rtnl_net_unlock(net);
3165 	return err;
3166 }
3167 
addrconf_del_ifaddr(struct net * net,void __user * arg)3168 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3169 {
3170 	struct in6_ifreq ireq;
3171 	int err;
3172 
3173 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3174 		return -EPERM;
3175 
3176 	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3177 		return -EFAULT;
3178 
3179 	rtnl_net_lock(net);
3180 	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3181 			     ireq.ifr6_prefixlen, NULL);
3182 	rtnl_net_unlock(net);
3183 	return err;
3184 }
3185 
add_addr(struct inet6_dev * idev,const struct in6_addr * addr,int plen,int scope,u8 proto)3186 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3187 		     int plen, int scope, u8 proto)
3188 {
3189 	struct inet6_ifaddr *ifp;
3190 	struct ifa6_config cfg = {
3191 		.pfx = addr,
3192 		.plen = plen,
3193 		.ifa_flags = IFA_F_PERMANENT,
3194 		.valid_lft = INFINITY_LIFE_TIME,
3195 		.preferred_lft = INFINITY_LIFE_TIME,
3196 		.scope = scope,
3197 		.ifa_proto = proto
3198 	};
3199 
3200 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3201 	if (!IS_ERR(ifp)) {
3202 		spin_lock_bh(&ifp->lock);
3203 		ifp->flags &= ~IFA_F_TENTATIVE;
3204 		spin_unlock_bh(&ifp->lock);
3205 		rt_genid_bump_ipv6(dev_net(idev->dev));
3206 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3207 		in6_ifa_put(ifp);
3208 	}
3209 }
3210 
3211 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
add_v4_addrs(struct inet6_dev * idev)3212 static void add_v4_addrs(struct inet6_dev *idev)
3213 {
3214 	struct in6_addr addr;
3215 	struct net_device *dev;
3216 	struct net *net = dev_net(idev->dev);
3217 	int scope, plen, offset = 0;
3218 	u32 pflags = 0;
3219 
3220 	ASSERT_RTNL();
3221 
3222 	memset(&addr, 0, sizeof(struct in6_addr));
3223 	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3224 	if (idev->dev->addr_len == sizeof(struct in6_addr))
3225 		offset = sizeof(struct in6_addr) - 4;
3226 	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3227 
3228 	if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3229 		scope = IPV6_ADDR_COMPATv4;
3230 		plen = 96;
3231 		pflags |= RTF_NONEXTHOP;
3232 	} else {
3233 		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3234 			return;
3235 
3236 		addr.s6_addr32[0] = htonl(0xfe800000);
3237 		scope = IFA_LINK;
3238 		plen = 64;
3239 	}
3240 
3241 	if (addr.s6_addr32[3]) {
3242 		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3243 		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3244 				      GFP_KERNEL);
3245 		return;
3246 	}
3247 
3248 	for_each_netdev(net, dev) {
3249 		struct in_device *in_dev = __in_dev_get_rtnl(dev);
3250 		if (in_dev && (dev->flags & IFF_UP)) {
3251 			struct in_ifaddr *ifa;
3252 			int flag = scope;
3253 
3254 			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3255 				addr.s6_addr32[3] = ifa->ifa_local;
3256 
3257 				if (ifa->ifa_scope == RT_SCOPE_LINK)
3258 					continue;
3259 				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3260 					if (idev->dev->flags&IFF_POINTOPOINT)
3261 						continue;
3262 					flag |= IFA_HOST;
3263 				}
3264 
3265 				add_addr(idev, &addr, plen, flag,
3266 					 IFAPROT_UNSPEC);
3267 				addrconf_prefix_route(&addr, plen, 0, idev->dev,
3268 						      0, pflags, GFP_KERNEL);
3269 			}
3270 		}
3271 	}
3272 }
3273 #endif
3274 
init_loopback(struct net_device * dev)3275 static void init_loopback(struct net_device *dev)
3276 {
3277 	struct inet6_dev  *idev;
3278 
3279 	/* ::1 */
3280 
3281 	ASSERT_RTNL();
3282 
3283 	idev = ipv6_find_idev(dev);
3284 	if (IS_ERR(idev)) {
3285 		pr_debug("%s: add_dev failed\n", __func__);
3286 		return;
3287 	}
3288 
3289 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3290 }
3291 
addrconf_add_linklocal(struct inet6_dev * idev,const struct in6_addr * addr,u32 flags)3292 void addrconf_add_linklocal(struct inet6_dev *idev,
3293 			    const struct in6_addr *addr, u32 flags)
3294 {
3295 	struct ifa6_config cfg = {
3296 		.pfx = addr,
3297 		.plen = 64,
3298 		.ifa_flags = flags | IFA_F_PERMANENT,
3299 		.valid_lft = INFINITY_LIFE_TIME,
3300 		.preferred_lft = INFINITY_LIFE_TIME,
3301 		.scope = IFA_LINK,
3302 		.ifa_proto = IFAPROT_KERNEL_LL
3303 	};
3304 	struct inet6_ifaddr *ifp;
3305 
3306 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3307 	if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3308 	     READ_ONCE(idev->cnf.optimistic_dad)) &&
3309 	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3310 		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3311 #endif
3312 
3313 	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3314 	if (!IS_ERR(ifp)) {
3315 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3316 				      0, 0, GFP_ATOMIC);
3317 		addrconf_dad_start(ifp);
3318 		in6_ifa_put(ifp);
3319 	}
3320 }
3321 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3322 
ipv6_reserved_interfaceid(struct in6_addr address)3323 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3324 {
3325 	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3326 		return true;
3327 
3328 	if (address.s6_addr32[2] == htonl(0x02005eff) &&
3329 	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3330 		return true;
3331 
3332 	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3333 	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3334 		return true;
3335 
3336 	return false;
3337 }
3338 
ipv6_generate_stable_address(struct in6_addr * address,u8 dad_count,const struct inet6_dev * idev)3339 static int ipv6_generate_stable_address(struct in6_addr *address,
3340 					u8 dad_count,
3341 					const struct inet6_dev *idev)
3342 {
3343 	static DEFINE_SPINLOCK(lock);
3344 	static __u32 digest[SHA1_DIGEST_WORDS];
3345 	static __u32 workspace[SHA1_WORKSPACE_WORDS];
3346 
3347 	static union {
3348 		char __data[SHA1_BLOCK_SIZE];
3349 		struct {
3350 			struct in6_addr secret;
3351 			__be32 prefix[2];
3352 			unsigned char hwaddr[MAX_ADDR_LEN];
3353 			u8 dad_count;
3354 		} __packed;
3355 	} data;
3356 
3357 	struct in6_addr secret;
3358 	struct in6_addr temp;
3359 	struct net *net = dev_net(idev->dev);
3360 
3361 	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3362 
3363 	if (idev->cnf.stable_secret.initialized)
3364 		secret = idev->cnf.stable_secret.secret;
3365 	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3366 		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3367 	else
3368 		return -1;
3369 
3370 retry:
3371 	spin_lock_bh(&lock);
3372 
3373 	sha1_init(digest);
3374 	memset(&data, 0, sizeof(data));
3375 	memset(workspace, 0, sizeof(workspace));
3376 	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3377 	data.prefix[0] = address->s6_addr32[0];
3378 	data.prefix[1] = address->s6_addr32[1];
3379 	data.secret = secret;
3380 	data.dad_count = dad_count;
3381 
3382 	sha1_transform(digest, data.__data, workspace);
3383 
3384 	temp = *address;
3385 	temp.s6_addr32[2] = (__force __be32)digest[0];
3386 	temp.s6_addr32[3] = (__force __be32)digest[1];
3387 
3388 	spin_unlock_bh(&lock);
3389 
3390 	if (ipv6_reserved_interfaceid(temp)) {
3391 		dad_count++;
3392 		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3393 			return -1;
3394 		goto retry;
3395 	}
3396 
3397 	*address = temp;
3398 	return 0;
3399 }
3400 
ipv6_gen_mode_random_init(struct inet6_dev * idev)3401 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3402 {
3403 	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3404 
3405 	if (s->initialized)
3406 		return;
3407 	s = &idev->cnf.stable_secret;
3408 	get_random_bytes(&s->secret, sizeof(s->secret));
3409 	s->initialized = true;
3410 }
3411 
addrconf_addr_gen(struct inet6_dev * idev,bool prefix_route)3412 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3413 {
3414 	struct in6_addr addr;
3415 
3416 	/* no link local addresses on L3 master devices */
3417 	if (netif_is_l3_master(idev->dev))
3418 		return;
3419 
3420 	/* no link local addresses on devices flagged as slaves */
3421 	if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3422 		return;
3423 
3424 	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3425 
3426 	switch (idev->cnf.addr_gen_mode) {
3427 	case IN6_ADDR_GEN_MODE_RANDOM:
3428 		ipv6_gen_mode_random_init(idev);
3429 		fallthrough;
3430 	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3431 		if (!ipv6_generate_stable_address(&addr, 0, idev))
3432 			addrconf_add_linklocal(idev, &addr,
3433 					       IFA_F_STABLE_PRIVACY);
3434 		else if (prefix_route)
3435 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3436 					      0, 0, GFP_KERNEL);
3437 		break;
3438 	case IN6_ADDR_GEN_MODE_EUI64:
3439 		/* addrconf_add_linklocal also adds a prefix_route and we
3440 		 * only need to care about prefix routes if ipv6_generate_eui64
3441 		 * couldn't generate one.
3442 		 */
3443 		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3444 			addrconf_add_linklocal(idev, &addr, 0);
3445 		else if (prefix_route)
3446 			addrconf_prefix_route(&addr, 64, 0, idev->dev,
3447 					      0, 0, GFP_KERNEL);
3448 		break;
3449 	case IN6_ADDR_GEN_MODE_NONE:
3450 	default:
3451 		/* will not add any link local address */
3452 		break;
3453 	}
3454 }
3455 
addrconf_dev_config(struct net_device * dev)3456 static void addrconf_dev_config(struct net_device *dev)
3457 {
3458 	struct inet6_dev *idev;
3459 
3460 	ASSERT_RTNL();
3461 
3462 	if ((dev->type != ARPHRD_ETHER) &&
3463 	    (dev->type != ARPHRD_FDDI) &&
3464 	    (dev->type != ARPHRD_ARCNET) &&
3465 	    (dev->type != ARPHRD_INFINIBAND) &&
3466 	    (dev->type != ARPHRD_IEEE1394) &&
3467 	    (dev->type != ARPHRD_TUNNEL6) &&
3468 	    (dev->type != ARPHRD_6LOWPAN) &&
3469 	    (dev->type != ARPHRD_TUNNEL) &&
3470 	    (dev->type != ARPHRD_NONE) &&
3471 	    (dev->type != ARPHRD_RAWIP)) {
3472 		/* Alas, we support only Ethernet autoconfiguration. */
3473 		idev = __in6_dev_get(dev);
3474 		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3475 		    dev->flags & IFF_MULTICAST)
3476 			ipv6_mc_up(idev);
3477 		return;
3478 	}
3479 
3480 	idev = addrconf_add_dev(dev);
3481 	if (IS_ERR(idev))
3482 		return;
3483 
3484 	/* this device type has no EUI support */
3485 	if (dev->type == ARPHRD_NONE &&
3486 	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3487 		WRITE_ONCE(idev->cnf.addr_gen_mode,
3488 			   IN6_ADDR_GEN_MODE_RANDOM);
3489 
3490 	addrconf_addr_gen(idev, false);
3491 }
3492 
3493 #if IS_ENABLED(CONFIG_IPV6_SIT)
addrconf_sit_config(struct net_device * dev)3494 static void addrconf_sit_config(struct net_device *dev)
3495 {
3496 	struct inet6_dev *idev;
3497 
3498 	ASSERT_RTNL();
3499 
3500 	/*
3501 	 * Configure the tunnel with one of our IPv4
3502 	 * addresses... we should configure all of
3503 	 * our v4 addrs in the tunnel
3504 	 */
3505 
3506 	idev = ipv6_find_idev(dev);
3507 	if (IS_ERR(idev)) {
3508 		pr_debug("%s: add_dev failed\n", __func__);
3509 		return;
3510 	}
3511 
3512 	if (dev->priv_flags & IFF_ISATAP) {
3513 		addrconf_addr_gen(idev, false);
3514 		return;
3515 	}
3516 
3517 	add_v4_addrs(idev);
3518 
3519 	if (dev->flags&IFF_POINTOPOINT)
3520 		addrconf_add_mroute(dev);
3521 }
3522 #endif
3523 
3524 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
addrconf_gre_config(struct net_device * dev)3525 static void addrconf_gre_config(struct net_device *dev)
3526 {
3527 	struct inet6_dev *idev;
3528 
3529 	ASSERT_RTNL();
3530 
3531 	idev = ipv6_find_idev(dev);
3532 	if (IS_ERR(idev)) {
3533 		pr_debug("%s: add_dev failed\n", __func__);
3534 		return;
3535 	}
3536 
3537 	if (dev->type == ARPHRD_ETHER) {
3538 		addrconf_addr_gen(idev, true);
3539 		return;
3540 	}
3541 
3542 	add_v4_addrs(idev);
3543 
3544 	if (dev->flags & IFF_POINTOPOINT)
3545 		addrconf_add_mroute(dev);
3546 }
3547 #endif
3548 
addrconf_init_auto_addrs(struct net_device * dev)3549 static void addrconf_init_auto_addrs(struct net_device *dev)
3550 {
3551 	switch (dev->type) {
3552 #if IS_ENABLED(CONFIG_IPV6_SIT)
3553 	case ARPHRD_SIT:
3554 		addrconf_sit_config(dev);
3555 		break;
3556 #endif
3557 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3558 	case ARPHRD_IP6GRE:
3559 	case ARPHRD_IPGRE:
3560 		addrconf_gre_config(dev);
3561 		break;
3562 #endif
3563 	case ARPHRD_LOOPBACK:
3564 		init_loopback(dev);
3565 		break;
3566 
3567 	default:
3568 		addrconf_dev_config(dev);
3569 		break;
3570 	}
3571 }
3572 
fixup_permanent_addr(struct net * net,struct inet6_dev * idev,struct inet6_ifaddr * ifp)3573 static int fixup_permanent_addr(struct net *net,
3574 				struct inet6_dev *idev,
3575 				struct inet6_ifaddr *ifp)
3576 {
3577 	/* !fib6_node means the host route was removed from the
3578 	 * FIB, for example, if 'lo' device is taken down. In that
3579 	 * case regenerate the host route.
3580 	 */
3581 	if (!ifp->rt || !ifp->rt->fib6_node) {
3582 		struct fib6_info *f6i, *prev;
3583 
3584 		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3585 					 GFP_ATOMIC, NULL);
3586 		if (IS_ERR(f6i))
3587 			return PTR_ERR(f6i);
3588 
3589 		/* ifp->rt can be accessed outside of rtnl */
3590 		spin_lock(&ifp->lock);
3591 		prev = ifp->rt;
3592 		ifp->rt = f6i;
3593 		spin_unlock(&ifp->lock);
3594 
3595 		fib6_info_release(prev);
3596 	}
3597 
3598 	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3599 		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3600 				      ifp->rt_priority, idev->dev, 0, 0,
3601 				      GFP_ATOMIC);
3602 	}
3603 
3604 	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3605 		addrconf_dad_start(ifp);
3606 
3607 	return 0;
3608 }
3609 
addrconf_permanent_addr(struct net * net,struct net_device * dev)3610 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3611 {
3612 	struct inet6_ifaddr *ifp, *tmp;
3613 	struct inet6_dev *idev;
3614 
3615 	idev = __in6_dev_get(dev);
3616 	if (!idev)
3617 		return;
3618 
3619 	write_lock_bh(&idev->lock);
3620 
3621 	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3622 		if ((ifp->flags & IFA_F_PERMANENT) &&
3623 		    fixup_permanent_addr(net, idev, ifp) < 0) {
3624 			write_unlock_bh(&idev->lock);
3625 			in6_ifa_hold(ifp);
3626 			ipv6_del_addr(ifp);
3627 			write_lock_bh(&idev->lock);
3628 
3629 			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3630 					     idev->dev->name, &ifp->addr);
3631 		}
3632 	}
3633 
3634 	write_unlock_bh(&idev->lock);
3635 }
3636 
addrconf_notify(struct notifier_block * this,unsigned long event,void * ptr)3637 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3638 			   void *ptr)
3639 {
3640 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3641 	struct netdev_notifier_change_info *change_info;
3642 	struct netdev_notifier_changeupper_info *info;
3643 	struct inet6_dev *idev = __in6_dev_get(dev);
3644 	struct net *net = dev_net(dev);
3645 	int run_pending = 0;
3646 	int err;
3647 
3648 	switch (event) {
3649 	case NETDEV_REGISTER:
3650 		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3651 			idev = ipv6_add_dev(dev);
3652 			if (IS_ERR(idev))
3653 				return notifier_from_errno(PTR_ERR(idev));
3654 		}
3655 		break;
3656 
3657 	case NETDEV_CHANGEMTU:
3658 		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3659 		if (dev->mtu < IPV6_MIN_MTU) {
3660 			addrconf_ifdown(dev, dev != net->loopback_dev);
3661 			break;
3662 		}
3663 
3664 		if (idev) {
3665 			rt6_mtu_change(dev, dev->mtu);
3666 			WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3667 			break;
3668 		}
3669 
3670 		/* allocate new idev */
3671 		idev = ipv6_add_dev(dev);
3672 		if (IS_ERR(idev))
3673 			break;
3674 
3675 		/* device is still not ready */
3676 		if (!(idev->if_flags & IF_READY))
3677 			break;
3678 
3679 		run_pending = 1;
3680 		fallthrough;
3681 	case NETDEV_UP:
3682 	case NETDEV_CHANGE:
3683 		if (idev && idev->cnf.disable_ipv6)
3684 			break;
3685 
3686 		if (dev->priv_flags & IFF_NO_ADDRCONF) {
3687 			if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3688 			    dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3689 				ipv6_mc_up(idev);
3690 			break;
3691 		}
3692 
3693 		if (event == NETDEV_UP) {
3694 			/* restore routes for permanent addresses */
3695 			addrconf_permanent_addr(net, dev);
3696 
3697 			if (!addrconf_link_ready(dev)) {
3698 				/* device is not ready yet. */
3699 				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3700 					 dev->name);
3701 				break;
3702 			}
3703 
3704 			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3705 				idev = ipv6_add_dev(dev);
3706 
3707 			if (!IS_ERR_OR_NULL(idev)) {
3708 				idev->if_flags |= IF_READY;
3709 				run_pending = 1;
3710 			}
3711 		} else if (event == NETDEV_CHANGE) {
3712 			if (!addrconf_link_ready(dev)) {
3713 				/* device is still not ready. */
3714 				rt6_sync_down_dev(dev, event);
3715 				break;
3716 			}
3717 
3718 			if (!IS_ERR_OR_NULL(idev)) {
3719 				if (idev->if_flags & IF_READY) {
3720 					/* device is already configured -
3721 					 * but resend MLD reports, we might
3722 					 * have roamed and need to update
3723 					 * multicast snooping switches
3724 					 */
3725 					ipv6_mc_up(idev);
3726 					change_info = ptr;
3727 					if (change_info->flags_changed & IFF_NOARP)
3728 						addrconf_dad_run(idev, true);
3729 					rt6_sync_up(dev, RTNH_F_LINKDOWN);
3730 					break;
3731 				}
3732 				idev->if_flags |= IF_READY;
3733 			}
3734 
3735 			pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3736 				 dev->name);
3737 
3738 			run_pending = 1;
3739 		}
3740 
3741 		addrconf_init_auto_addrs(dev);
3742 
3743 		if (!IS_ERR_OR_NULL(idev)) {
3744 			if (run_pending)
3745 				addrconf_dad_run(idev, false);
3746 
3747 			/* Device has an address by now */
3748 			rt6_sync_up(dev, RTNH_F_DEAD);
3749 
3750 			/*
3751 			 * If the MTU changed during the interface down,
3752 			 * when the interface up, the changed MTU must be
3753 			 * reflected in the idev as well as routers.
3754 			 */
3755 			if (idev->cnf.mtu6 != dev->mtu &&
3756 			    dev->mtu >= IPV6_MIN_MTU) {
3757 				rt6_mtu_change(dev, dev->mtu);
3758 				WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3759 			}
3760 			WRITE_ONCE(idev->tstamp, jiffies);
3761 			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3762 
3763 			/*
3764 			 * If the changed mtu during down is lower than
3765 			 * IPV6_MIN_MTU stop IPv6 on this interface.
3766 			 */
3767 			if (dev->mtu < IPV6_MIN_MTU)
3768 				addrconf_ifdown(dev, dev != net->loopback_dev);
3769 		}
3770 		break;
3771 
3772 	case NETDEV_DOWN:
3773 	case NETDEV_UNREGISTER:
3774 		/*
3775 		 *	Remove all addresses from this interface.
3776 		 */
3777 		addrconf_ifdown(dev, event != NETDEV_DOWN);
3778 		break;
3779 
3780 	case NETDEV_CHANGENAME:
3781 		if (idev) {
3782 			snmp6_unregister_dev(idev);
3783 			addrconf_sysctl_unregister(idev);
3784 			err = addrconf_sysctl_register(idev);
3785 			if (err)
3786 				return notifier_from_errno(err);
3787 			err = snmp6_register_dev(idev);
3788 			if (err) {
3789 				addrconf_sysctl_unregister(idev);
3790 				return notifier_from_errno(err);
3791 			}
3792 		}
3793 		break;
3794 
3795 	case NETDEV_PRE_TYPE_CHANGE:
3796 	case NETDEV_POST_TYPE_CHANGE:
3797 		if (idev)
3798 			addrconf_type_change(dev, event);
3799 		break;
3800 
3801 	case NETDEV_CHANGEUPPER:
3802 		info = ptr;
3803 
3804 		/* flush all routes if dev is linked to or unlinked from
3805 		 * an L3 master device (e.g., VRF)
3806 		 */
3807 		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3808 			addrconf_ifdown(dev, false);
3809 	}
3810 
3811 	return NOTIFY_OK;
3812 }
3813 
3814 /*
3815  *	addrconf module should be notified of a device going up
3816  */
3817 static struct notifier_block ipv6_dev_notf = {
3818 	.notifier_call = addrconf_notify,
3819 	.priority = ADDRCONF_NOTIFY_PRIORITY,
3820 };
3821 
addrconf_type_change(struct net_device * dev,unsigned long event)3822 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3823 {
3824 	struct inet6_dev *idev;
3825 	ASSERT_RTNL();
3826 
3827 	idev = __in6_dev_get(dev);
3828 
3829 	if (event == NETDEV_POST_TYPE_CHANGE)
3830 		ipv6_mc_remap(idev);
3831 	else if (event == NETDEV_PRE_TYPE_CHANGE)
3832 		ipv6_mc_unmap(idev);
3833 }
3834 
addr_is_local(const struct in6_addr * addr)3835 static bool addr_is_local(const struct in6_addr *addr)
3836 {
3837 	return ipv6_addr_type(addr) &
3838 		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3839 }
3840 
addrconf_ifdown(struct net_device * dev,bool unregister)3841 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3842 {
3843 	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3844 	struct net *net = dev_net(dev);
3845 	struct inet6_dev *idev;
3846 	struct inet6_ifaddr *ifa;
3847 	LIST_HEAD(tmp_addr_list);
3848 	bool keep_addr = false;
3849 	bool was_ready;
3850 	int state, i;
3851 
3852 	ASSERT_RTNL();
3853 
3854 	rt6_disable_ip(dev, event);
3855 
3856 	idev = __in6_dev_get(dev);
3857 	if (!idev)
3858 		return -ENODEV;
3859 
3860 	/*
3861 	 * Step 1: remove reference to ipv6 device from parent device.
3862 	 *	   Do not dev_put!
3863 	 */
3864 	if (unregister) {
3865 		idev->dead = 1;
3866 
3867 		/* protected by rtnl_lock */
3868 		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3869 
3870 		/* Step 1.5: remove snmp6 entry */
3871 		snmp6_unregister_dev(idev);
3872 
3873 	}
3874 
3875 	/* combine the user config with event to determine if permanent
3876 	 * addresses are to be removed from address hash table
3877 	 */
3878 	if (!unregister && !idev->cnf.disable_ipv6) {
3879 		/* aggregate the system setting and interface setting */
3880 		int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3881 
3882 		if (!_keep_addr)
3883 			_keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3884 
3885 		keep_addr = (_keep_addr > 0);
3886 	}
3887 
3888 	/* Step 2: clear hash table */
3889 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3890 		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3891 
3892 		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3893 restart:
3894 		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3895 			if (ifa->idev == idev) {
3896 				addrconf_del_dad_work(ifa);
3897 				/* combined flag + permanent flag decide if
3898 				 * address is retained on a down event
3899 				 */
3900 				if (!keep_addr ||
3901 				    !(ifa->flags & IFA_F_PERMANENT) ||
3902 				    addr_is_local(&ifa->addr)) {
3903 					hlist_del_init_rcu(&ifa->addr_lst);
3904 					goto restart;
3905 				}
3906 			}
3907 		}
3908 		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3909 	}
3910 
3911 	write_lock_bh(&idev->lock);
3912 
3913 	addrconf_del_rs_timer(idev);
3914 
3915 	/* Step 2: clear flags for stateless addrconf, repeated down
3916 	 *         detection
3917 	 */
3918 	was_ready = idev->if_flags & IF_READY;
3919 	if (!unregister)
3920 		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3921 
3922 	/* Step 3: clear tempaddr list */
3923 	while (!list_empty(&idev->tempaddr_list)) {
3924 		ifa = list_first_entry(&idev->tempaddr_list,
3925 				       struct inet6_ifaddr, tmp_list);
3926 		list_del(&ifa->tmp_list);
3927 		write_unlock_bh(&idev->lock);
3928 		spin_lock_bh(&ifa->lock);
3929 
3930 		if (ifa->ifpub) {
3931 			in6_ifa_put(ifa->ifpub);
3932 			ifa->ifpub = NULL;
3933 		}
3934 		spin_unlock_bh(&ifa->lock);
3935 		in6_ifa_put(ifa);
3936 		write_lock_bh(&idev->lock);
3937 	}
3938 
3939 	list_for_each_entry(ifa, &idev->addr_list, if_list)
3940 		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3941 	write_unlock_bh(&idev->lock);
3942 
3943 	while (!list_empty(&tmp_addr_list)) {
3944 		struct fib6_info *rt = NULL;
3945 		bool keep;
3946 
3947 		ifa = list_first_entry(&tmp_addr_list,
3948 				       struct inet6_ifaddr, if_list_aux);
3949 		list_del(&ifa->if_list_aux);
3950 
3951 		addrconf_del_dad_work(ifa);
3952 
3953 		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3954 			!addr_is_local(&ifa->addr);
3955 
3956 		spin_lock_bh(&ifa->lock);
3957 
3958 		if (keep) {
3959 			/* set state to skip the notifier below */
3960 			state = INET6_IFADDR_STATE_DEAD;
3961 			ifa->state = INET6_IFADDR_STATE_PREDAD;
3962 			if (!(ifa->flags & IFA_F_NODAD))
3963 				ifa->flags |= IFA_F_TENTATIVE;
3964 
3965 			rt = ifa->rt;
3966 			ifa->rt = NULL;
3967 		} else {
3968 			state = ifa->state;
3969 			ifa->state = INET6_IFADDR_STATE_DEAD;
3970 		}
3971 
3972 		spin_unlock_bh(&ifa->lock);
3973 
3974 		if (rt)
3975 			ip6_del_rt(net, rt, false);
3976 
3977 		if (state != INET6_IFADDR_STATE_DEAD) {
3978 			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3979 			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3980 		} else {
3981 			if (idev->cnf.forwarding)
3982 				addrconf_leave_anycast(ifa);
3983 			addrconf_leave_solict(ifa->idev, &ifa->addr);
3984 		}
3985 
3986 		if (!keep) {
3987 			write_lock_bh(&idev->lock);
3988 			list_del_rcu(&ifa->if_list);
3989 			write_unlock_bh(&idev->lock);
3990 			in6_ifa_put(ifa);
3991 		}
3992 	}
3993 
3994 	/* Step 5: Discard anycast and multicast list */
3995 	if (unregister) {
3996 		ipv6_ac_destroy_dev(idev);
3997 		ipv6_mc_destroy_dev(idev);
3998 	} else if (was_ready) {
3999 		ipv6_mc_down(idev);
4000 	}
4001 
4002 	WRITE_ONCE(idev->tstamp, jiffies);
4003 	idev->ra_mtu = 0;
4004 
4005 	/* Last: Shot the device (if unregistered) */
4006 	if (unregister) {
4007 		addrconf_sysctl_unregister(idev);
4008 		neigh_parms_release(&nd_tbl, idev->nd_parms);
4009 		neigh_ifdown(&nd_tbl, dev);
4010 		in6_dev_put(idev);
4011 	}
4012 	return 0;
4013 }
4014 
addrconf_rs_timer(struct timer_list * t)4015 static void addrconf_rs_timer(struct timer_list *t)
4016 {
4017 	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
4018 	struct net_device *dev = idev->dev;
4019 	struct in6_addr lladdr;
4020 	int rtr_solicits;
4021 
4022 	write_lock(&idev->lock);
4023 	if (idev->dead || !(idev->if_flags & IF_READY))
4024 		goto out;
4025 
4026 	if (!ipv6_accept_ra(idev))
4027 		goto out;
4028 
4029 	/* Announcement received after solicitation was sent */
4030 	if (idev->if_flags & IF_RA_RCVD)
4031 		goto out;
4032 
4033 	rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4034 
4035 	if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4036 		write_unlock(&idev->lock);
4037 		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4038 			ndisc_send_rs(dev, &lladdr,
4039 				      &in6addr_linklocal_allrouters);
4040 		else
4041 			goto put;
4042 
4043 		write_lock(&idev->lock);
4044 		idev->rs_interval = rfc3315_s14_backoff_update(
4045 				idev->rs_interval,
4046 				READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4047 		/* The wait after the last probe can be shorter */
4048 		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4049 					     READ_ONCE(idev->cnf.rtr_solicits)) ?
4050 				      READ_ONCE(idev->cnf.rtr_solicit_delay) :
4051 				      idev->rs_interval);
4052 	} else {
4053 		/*
4054 		 * Note: we do not support deprecated "all on-link"
4055 		 * assumption any longer.
4056 		 */
4057 		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4058 	}
4059 
4060 out:
4061 	write_unlock(&idev->lock);
4062 put:
4063 	in6_dev_put(idev);
4064 }
4065 
4066 /*
4067  *	Duplicate Address Detection
4068  */
addrconf_dad_kick(struct inet6_ifaddr * ifp)4069 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4070 {
4071 	struct inet6_dev *idev = ifp->idev;
4072 	unsigned long rand_num;
4073 	u64 nonce;
4074 
4075 	if (ifp->flags & IFA_F_OPTIMISTIC)
4076 		rand_num = 0;
4077 	else
4078 		rand_num = get_random_u32_below(
4079 				READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4080 
4081 	nonce = 0;
4082 	if (READ_ONCE(idev->cnf.enhanced_dad) ||
4083 	    READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4084 		do
4085 			get_random_bytes(&nonce, 6);
4086 		while (nonce == 0);
4087 	}
4088 	ifp->dad_nonce = nonce;
4089 	ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4090 	addrconf_mod_dad_work(ifp, rand_num);
4091 }
4092 
addrconf_dad_begin(struct inet6_ifaddr * ifp)4093 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4094 {
4095 	struct inet6_dev *idev = ifp->idev;
4096 	struct net_device *dev = idev->dev;
4097 	bool bump_id, notify = false;
4098 	struct net *net;
4099 
4100 	addrconf_join_solict(dev, &ifp->addr);
4101 
4102 	read_lock_bh(&idev->lock);
4103 	spin_lock(&ifp->lock);
4104 	if (ifp->state == INET6_IFADDR_STATE_DEAD)
4105 		goto out;
4106 
4107 	net = dev_net(dev);
4108 	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4109 	    (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4110 	     READ_ONCE(idev->cnf.accept_dad) < 1) ||
4111 	    !(ifp->flags&IFA_F_TENTATIVE) ||
4112 	    ifp->flags & IFA_F_NODAD) {
4113 		bool send_na = false;
4114 
4115 		if (ifp->flags & IFA_F_TENTATIVE &&
4116 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4117 			send_na = true;
4118 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4119 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4120 		spin_unlock(&ifp->lock);
4121 		read_unlock_bh(&idev->lock);
4122 
4123 		addrconf_dad_completed(ifp, bump_id, send_na);
4124 		return;
4125 	}
4126 
4127 	if (!(idev->if_flags & IF_READY)) {
4128 		spin_unlock(&ifp->lock);
4129 		read_unlock_bh(&idev->lock);
4130 		/*
4131 		 * If the device is not ready:
4132 		 * - keep it tentative if it is a permanent address.
4133 		 * - otherwise, kill it.
4134 		 */
4135 		in6_ifa_hold(ifp);
4136 		addrconf_dad_stop(ifp, 0);
4137 		return;
4138 	}
4139 
4140 	/*
4141 	 * Optimistic nodes can start receiving
4142 	 * Frames right away
4143 	 */
4144 	if (ifp->flags & IFA_F_OPTIMISTIC) {
4145 		ip6_ins_rt(net, ifp->rt);
4146 		if (ipv6_use_optimistic_addr(net, idev)) {
4147 			/* Because optimistic nodes can use this address,
4148 			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4149 			 */
4150 			notify = true;
4151 		}
4152 	}
4153 
4154 	addrconf_dad_kick(ifp);
4155 out:
4156 	spin_unlock(&ifp->lock);
4157 	read_unlock_bh(&idev->lock);
4158 	if (notify)
4159 		ipv6_ifa_notify(RTM_NEWADDR, ifp);
4160 }
4161 
addrconf_dad_start(struct inet6_ifaddr * ifp)4162 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4163 {
4164 	bool begin_dad = false;
4165 
4166 	spin_lock_bh(&ifp->lock);
4167 	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4168 		ifp->state = INET6_IFADDR_STATE_PREDAD;
4169 		begin_dad = true;
4170 	}
4171 	spin_unlock_bh(&ifp->lock);
4172 
4173 	if (begin_dad)
4174 		addrconf_mod_dad_work(ifp, 0);
4175 }
4176 
addrconf_dad_work(struct work_struct * w)4177 static void addrconf_dad_work(struct work_struct *w)
4178 {
4179 	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4180 						struct inet6_ifaddr,
4181 						dad_work);
4182 	struct inet6_dev *idev = ifp->idev;
4183 	bool bump_id, disable_ipv6 = false;
4184 	struct in6_addr mcaddr;
4185 	struct net *net;
4186 
4187 	enum {
4188 		DAD_PROCESS,
4189 		DAD_BEGIN,
4190 		DAD_ABORT,
4191 	} action = DAD_PROCESS;
4192 
4193 	net = dev_net(idev->dev);
4194 
4195 	rtnl_net_lock(net);
4196 
4197 	spin_lock_bh(&ifp->lock);
4198 	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4199 		action = DAD_BEGIN;
4200 		ifp->state = INET6_IFADDR_STATE_DAD;
4201 	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4202 		action = DAD_ABORT;
4203 		ifp->state = INET6_IFADDR_STATE_POSTDAD;
4204 
4205 		if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 ||
4206 		     READ_ONCE(idev->cnf.accept_dad) > 1) &&
4207 		    !idev->cnf.disable_ipv6 &&
4208 		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4209 			struct in6_addr addr;
4210 
4211 			addr.s6_addr32[0] = htonl(0xfe800000);
4212 			addr.s6_addr32[1] = 0;
4213 
4214 			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4215 			    ipv6_addr_equal(&ifp->addr, &addr)) {
4216 				/* DAD failed for link-local based on MAC */
4217 				WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4218 
4219 				pr_info("%s: IPv6 being disabled!\n",
4220 					ifp->idev->dev->name);
4221 				disable_ipv6 = true;
4222 			}
4223 		}
4224 	}
4225 	spin_unlock_bh(&ifp->lock);
4226 
4227 	if (action == DAD_BEGIN) {
4228 		addrconf_dad_begin(ifp);
4229 		goto out;
4230 	} else if (action == DAD_ABORT) {
4231 		in6_ifa_hold(ifp);
4232 		addrconf_dad_stop(ifp, 1);
4233 		if (disable_ipv6)
4234 			addrconf_ifdown(idev->dev, false);
4235 		goto out;
4236 	}
4237 
4238 	if (!ifp->dad_probes && addrconf_dad_end(ifp))
4239 		goto out;
4240 
4241 	write_lock_bh(&idev->lock);
4242 	if (idev->dead || !(idev->if_flags & IF_READY)) {
4243 		write_unlock_bh(&idev->lock);
4244 		goto out;
4245 	}
4246 
4247 	spin_lock(&ifp->lock);
4248 	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4249 		spin_unlock(&ifp->lock);
4250 		write_unlock_bh(&idev->lock);
4251 		goto out;
4252 	}
4253 
4254 	if (ifp->dad_probes == 0) {
4255 		bool send_na = false;
4256 
4257 		/*
4258 		 * DAD was successful
4259 		 */
4260 
4261 		if (ifp->flags & IFA_F_TENTATIVE &&
4262 		    !(ifp->flags & IFA_F_OPTIMISTIC))
4263 			send_na = true;
4264 		bump_id = ifp->flags & IFA_F_TENTATIVE;
4265 		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4266 		spin_unlock(&ifp->lock);
4267 		write_unlock_bh(&idev->lock);
4268 
4269 		addrconf_dad_completed(ifp, bump_id, send_na);
4270 
4271 		goto out;
4272 	}
4273 
4274 	ifp->dad_probes--;
4275 	addrconf_mod_dad_work(ifp,
4276 			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4277 				  HZ/100));
4278 	spin_unlock(&ifp->lock);
4279 	write_unlock_bh(&idev->lock);
4280 
4281 	/* send a neighbour solicitation for our addr */
4282 	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4283 	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4284 		      ifp->dad_nonce);
4285 out:
4286 	in6_ifa_put(ifp);
4287 	rtnl_net_unlock(net);
4288 }
4289 
4290 /* ifp->idev must be at least read locked */
ipv6_lonely_lladdr(struct inet6_ifaddr * ifp)4291 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4292 {
4293 	struct inet6_ifaddr *ifpiter;
4294 	struct inet6_dev *idev = ifp->idev;
4295 
4296 	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4297 		if (ifpiter->scope > IFA_LINK)
4298 			break;
4299 		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4300 		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4301 				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4302 		    IFA_F_PERMANENT)
4303 			return false;
4304 	}
4305 	return true;
4306 }
4307 
addrconf_dad_completed(struct inet6_ifaddr * ifp,bool bump_id,bool send_na)4308 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4309 				   bool send_na)
4310 {
4311 	struct net_device *dev = ifp->idev->dev;
4312 	struct in6_addr lladdr;
4313 	bool send_rs, send_mld;
4314 
4315 	addrconf_del_dad_work(ifp);
4316 
4317 	/*
4318 	 *	Configure the address for reception. Now it is valid.
4319 	 */
4320 
4321 	ipv6_ifa_notify(RTM_NEWADDR, ifp);
4322 
4323 	/* If added prefix is link local and we are prepared to process
4324 	   router advertisements, start sending router solicitations.
4325 	 */
4326 
4327 	read_lock_bh(&ifp->idev->lock);
4328 	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4329 	send_rs = send_mld &&
4330 		  ipv6_accept_ra(ifp->idev) &&
4331 		  READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4332 		  (dev->flags & IFF_LOOPBACK) == 0 &&
4333 		  (dev->type != ARPHRD_TUNNEL) &&
4334 		  !netif_is_team_port(dev);
4335 	read_unlock_bh(&ifp->idev->lock);
4336 
4337 	/* While dad is in progress mld report's source address is in6_addrany.
4338 	 * Resend with proper ll now.
4339 	 */
4340 	if (send_mld)
4341 		ipv6_mc_dad_complete(ifp->idev);
4342 
4343 	/* send unsolicited NA if enabled */
4344 	if (send_na &&
4345 	    (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4346 	     READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4347 		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4348 			      /*router=*/ !!ifp->idev->cnf.forwarding,
4349 			      /*solicited=*/ false, /*override=*/ true,
4350 			      /*inc_opt=*/ true);
4351 	}
4352 
4353 	if (send_rs) {
4354 		/*
4355 		 *	If a host as already performed a random delay
4356 		 *	[...] as part of DAD [...] there is no need
4357 		 *	to delay again before sending the first RS
4358 		 */
4359 		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4360 			return;
4361 		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4362 
4363 		write_lock_bh(&ifp->idev->lock);
4364 		spin_lock(&ifp->lock);
4365 		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4366 			READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4367 		ifp->idev->rs_probes = 1;
4368 		ifp->idev->if_flags |= IF_RS_SENT;
4369 		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4370 		spin_unlock(&ifp->lock);
4371 		write_unlock_bh(&ifp->idev->lock);
4372 	}
4373 
4374 	if (bump_id)
4375 		rt_genid_bump_ipv6(dev_net(dev));
4376 
4377 	/* Make sure that a new temporary address will be created
4378 	 * before this temporary address becomes deprecated.
4379 	 */
4380 	if (ifp->flags & IFA_F_TEMPORARY)
4381 		addrconf_verify_rtnl(dev_net(dev));
4382 }
4383 
addrconf_dad_run(struct inet6_dev * idev,bool restart)4384 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4385 {
4386 	struct inet6_ifaddr *ifp;
4387 
4388 	read_lock_bh(&idev->lock);
4389 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4390 		spin_lock(&ifp->lock);
4391 		if ((ifp->flags & IFA_F_TENTATIVE &&
4392 		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4393 			if (restart)
4394 				ifp->state = INET6_IFADDR_STATE_PREDAD;
4395 			addrconf_dad_kick(ifp);
4396 		}
4397 		spin_unlock(&ifp->lock);
4398 	}
4399 	read_unlock_bh(&idev->lock);
4400 }
4401 
4402 #ifdef CONFIG_PROC_FS
4403 struct if6_iter_state {
4404 	struct seq_net_private p;
4405 	int bucket;
4406 	int offset;
4407 };
4408 
if6_get_first(struct seq_file * seq,loff_t pos)4409 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4410 {
4411 	struct if6_iter_state *state = seq->private;
4412 	struct net *net = seq_file_net(seq);
4413 	struct inet6_ifaddr *ifa = NULL;
4414 	int p = 0;
4415 
4416 	/* initial bucket if pos is 0 */
4417 	if (pos == 0) {
4418 		state->bucket = 0;
4419 		state->offset = 0;
4420 	}
4421 
4422 	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4423 		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4424 					 addr_lst) {
4425 			/* sync with offset */
4426 			if (p < state->offset) {
4427 				p++;
4428 				continue;
4429 			}
4430 			return ifa;
4431 		}
4432 
4433 		/* prepare for next bucket */
4434 		state->offset = 0;
4435 		p = 0;
4436 	}
4437 	return NULL;
4438 }
4439 
if6_get_next(struct seq_file * seq,struct inet6_ifaddr * ifa)4440 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4441 					 struct inet6_ifaddr *ifa)
4442 {
4443 	struct if6_iter_state *state = seq->private;
4444 	struct net *net = seq_file_net(seq);
4445 
4446 	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4447 		state->offset++;
4448 		return ifa;
4449 	}
4450 
4451 	state->offset = 0;
4452 	while (++state->bucket < IN6_ADDR_HSIZE) {
4453 		hlist_for_each_entry_rcu(ifa,
4454 				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4455 			return ifa;
4456 		}
4457 	}
4458 
4459 	return NULL;
4460 }
4461 
if6_seq_start(struct seq_file * seq,loff_t * pos)4462 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4463 	__acquires(rcu)
4464 {
4465 	rcu_read_lock();
4466 	return if6_get_first(seq, *pos);
4467 }
4468 
if6_seq_next(struct seq_file * seq,void * v,loff_t * pos)4469 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4470 {
4471 	struct inet6_ifaddr *ifa;
4472 
4473 	ifa = if6_get_next(seq, v);
4474 	++*pos;
4475 	return ifa;
4476 }
4477 
if6_seq_stop(struct seq_file * seq,void * v)4478 static void if6_seq_stop(struct seq_file *seq, void *v)
4479 	__releases(rcu)
4480 {
4481 	rcu_read_unlock();
4482 }
4483 
if6_seq_show(struct seq_file * seq,void * v)4484 static int if6_seq_show(struct seq_file *seq, void *v)
4485 {
4486 	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4487 	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4488 		   &ifp->addr,
4489 		   ifp->idev->dev->ifindex,
4490 		   ifp->prefix_len,
4491 		   ifp->scope,
4492 		   (u8) ifp->flags,
4493 		   ifp->idev->dev->name);
4494 	return 0;
4495 }
4496 
4497 static const struct seq_operations if6_seq_ops = {
4498 	.start	= if6_seq_start,
4499 	.next	= if6_seq_next,
4500 	.show	= if6_seq_show,
4501 	.stop	= if6_seq_stop,
4502 };
4503 
if6_proc_net_init(struct net * net)4504 static int __net_init if6_proc_net_init(struct net *net)
4505 {
4506 	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4507 			sizeof(struct if6_iter_state)))
4508 		return -ENOMEM;
4509 	return 0;
4510 }
4511 
if6_proc_net_exit(struct net * net)4512 static void __net_exit if6_proc_net_exit(struct net *net)
4513 {
4514 	remove_proc_entry("if_inet6", net->proc_net);
4515 }
4516 
4517 static struct pernet_operations if6_proc_net_ops = {
4518 	.init = if6_proc_net_init,
4519 	.exit = if6_proc_net_exit,
4520 };
4521 
if6_proc_init(void)4522 int __init if6_proc_init(void)
4523 {
4524 	return register_pernet_subsys(&if6_proc_net_ops);
4525 }
4526 
if6_proc_exit(void)4527 void if6_proc_exit(void)
4528 {
4529 	unregister_pernet_subsys(&if6_proc_net_ops);
4530 }
4531 #endif	/* CONFIG_PROC_FS */
4532 
4533 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4534 /* Check if address is a home address configured on any interface. */
ipv6_chk_home_addr(struct net * net,const struct in6_addr * addr)4535 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4536 {
4537 	unsigned int hash = inet6_addr_hash(net, addr);
4538 	struct inet6_ifaddr *ifp = NULL;
4539 	int ret = 0;
4540 
4541 	rcu_read_lock();
4542 	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4543 		if (ipv6_addr_equal(&ifp->addr, addr) &&
4544 		    (ifp->flags & IFA_F_HOMEADDRESS)) {
4545 			ret = 1;
4546 			break;
4547 		}
4548 	}
4549 	rcu_read_unlock();
4550 	return ret;
4551 }
4552 #endif
4553 
4554 /* RFC6554 has some algorithm to avoid loops in segment routing by
4555  * checking if the segments contains any of a local interface address.
4556  *
4557  * Quote:
4558  *
4559  * To detect loops in the SRH, a router MUST determine if the SRH
4560  * includes multiple addresses assigned to any interface on that router.
4561  * If such addresses appear more than once and are separated by at least
4562  * one address not assigned to that router.
4563  */
ipv6_chk_rpl_srh_loop(struct net * net,const struct in6_addr * segs,unsigned char nsegs)4564 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4565 			  unsigned char nsegs)
4566 {
4567 	const struct in6_addr *addr;
4568 	int i, ret = 0, found = 0;
4569 	struct inet6_ifaddr *ifp;
4570 	bool separated = false;
4571 	unsigned int hash;
4572 	bool hash_found;
4573 
4574 	rcu_read_lock();
4575 	for (i = 0; i < nsegs; i++) {
4576 		addr = &segs[i];
4577 		hash = inet6_addr_hash(net, addr);
4578 
4579 		hash_found = false;
4580 		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4581 
4582 			if (ipv6_addr_equal(&ifp->addr, addr)) {
4583 				hash_found = true;
4584 				break;
4585 			}
4586 		}
4587 
4588 		if (hash_found) {
4589 			if (found > 1 && separated) {
4590 				ret = 1;
4591 				break;
4592 			}
4593 
4594 			separated = false;
4595 			found++;
4596 		} else {
4597 			separated = true;
4598 		}
4599 	}
4600 	rcu_read_unlock();
4601 
4602 	return ret;
4603 }
4604 
4605 /*
4606  *	Periodic address status verification
4607  */
4608 
addrconf_verify_rtnl(struct net * net)4609 static void addrconf_verify_rtnl(struct net *net)
4610 {
4611 	unsigned long now, next, next_sec, next_sched;
4612 	struct inet6_ifaddr *ifp;
4613 	int i;
4614 
4615 	ASSERT_RTNL();
4616 
4617 	rcu_read_lock_bh();
4618 	now = jiffies;
4619 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4620 
4621 	cancel_delayed_work(&net->ipv6.addr_chk_work);
4622 
4623 	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4624 restart:
4625 		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4626 			unsigned long age;
4627 
4628 			/* When setting preferred_lft to a value not zero or
4629 			 * infinity, while valid_lft is infinity
4630 			 * IFA_F_PERMANENT has a non-infinity life time.
4631 			 */
4632 			if ((ifp->flags & IFA_F_PERMANENT) &&
4633 			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
4634 				continue;
4635 
4636 			spin_lock(&ifp->lock);
4637 			/* We try to batch several events at once. */
4638 			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4639 
4640 			if ((ifp->flags&IFA_F_TEMPORARY) &&
4641 			    !(ifp->flags&IFA_F_TENTATIVE) &&
4642 			    ifp->prefered_lft != INFINITY_LIFE_TIME &&
4643 			    !ifp->regen_count && ifp->ifpub) {
4644 				/* This is a non-regenerated temporary addr. */
4645 
4646 				unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4647 
4648 				if (age + regen_advance >= ifp->prefered_lft) {
4649 					struct inet6_ifaddr *ifpub = ifp->ifpub;
4650 					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4651 						next = ifp->tstamp + ifp->prefered_lft * HZ;
4652 
4653 					ifp->regen_count++;
4654 					in6_ifa_hold(ifp);
4655 					in6_ifa_hold(ifpub);
4656 					spin_unlock(&ifp->lock);
4657 
4658 					spin_lock(&ifpub->lock);
4659 					ifpub->regen_count = 0;
4660 					spin_unlock(&ifpub->lock);
4661 					rcu_read_unlock_bh();
4662 					ipv6_create_tempaddr(ifpub, true);
4663 					in6_ifa_put(ifpub);
4664 					in6_ifa_put(ifp);
4665 					rcu_read_lock_bh();
4666 					goto restart;
4667 				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4668 					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4669 			}
4670 
4671 			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4672 			    age >= ifp->valid_lft) {
4673 				spin_unlock(&ifp->lock);
4674 				in6_ifa_hold(ifp);
4675 				rcu_read_unlock_bh();
4676 				ipv6_del_addr(ifp);
4677 				rcu_read_lock_bh();
4678 				goto restart;
4679 			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4680 				spin_unlock(&ifp->lock);
4681 				continue;
4682 			} else if (age >= ifp->prefered_lft) {
4683 				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4684 				int deprecate = 0;
4685 
4686 				if (!(ifp->flags&IFA_F_DEPRECATED)) {
4687 					deprecate = 1;
4688 					ifp->flags |= IFA_F_DEPRECATED;
4689 				}
4690 
4691 				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4692 				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4693 					next = ifp->tstamp + ifp->valid_lft * HZ;
4694 
4695 				spin_unlock(&ifp->lock);
4696 
4697 				if (deprecate) {
4698 					in6_ifa_hold(ifp);
4699 
4700 					ipv6_ifa_notify(0, ifp);
4701 					in6_ifa_put(ifp);
4702 					goto restart;
4703 				}
4704 			} else {
4705 				/* ifp->prefered_lft <= ifp->valid_lft */
4706 				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4707 					next = ifp->tstamp + ifp->prefered_lft * HZ;
4708 				spin_unlock(&ifp->lock);
4709 			}
4710 		}
4711 	}
4712 
4713 	next_sec = round_jiffies_up(next);
4714 	next_sched = next;
4715 
4716 	/* If rounded timeout is accurate enough, accept it. */
4717 	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4718 		next_sched = next_sec;
4719 
4720 	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4721 	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4722 		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4723 
4724 	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4725 		 now, next, next_sec, next_sched);
4726 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4727 	rcu_read_unlock_bh();
4728 }
4729 
addrconf_verify_work(struct work_struct * w)4730 static void addrconf_verify_work(struct work_struct *w)
4731 {
4732 	struct net *net = container_of(to_delayed_work(w), struct net,
4733 				       ipv6.addr_chk_work);
4734 
4735 	rtnl_net_lock(net);
4736 	addrconf_verify_rtnl(net);
4737 	rtnl_net_unlock(net);
4738 }
4739 
addrconf_verify(struct net * net)4740 static void addrconf_verify(struct net *net)
4741 {
4742 	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4743 }
4744 
extract_addr(struct nlattr * addr,struct nlattr * local,struct in6_addr ** peer_pfx)4745 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4746 				     struct in6_addr **peer_pfx)
4747 {
4748 	struct in6_addr *pfx = NULL;
4749 
4750 	*peer_pfx = NULL;
4751 
4752 	if (addr)
4753 		pfx = nla_data(addr);
4754 
4755 	if (local) {
4756 		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4757 			*peer_pfx = pfx;
4758 		pfx = nla_data(local);
4759 	}
4760 
4761 	return pfx;
4762 }
4763 
4764 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4765 	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4766 	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4767 	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4768 	[IFA_FLAGS]		= { .len = sizeof(u32) },
4769 	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
4770 	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
4771 	[IFA_PROTO]		= { .type = NLA_U8 },
4772 };
4773 
4774 static int
inet6_rtm_deladdr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4775 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4776 		  struct netlink_ext_ack *extack)
4777 {
4778 	struct net *net = sock_net(skb->sk);
4779 	struct ifaddrmsg *ifm;
4780 	struct nlattr *tb[IFA_MAX+1];
4781 	struct in6_addr *pfx, *peer_pfx;
4782 	u32 ifa_flags;
4783 	int err;
4784 
4785 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4786 				     ifa_ipv6_policy, extack);
4787 	if (err < 0)
4788 		return err;
4789 
4790 	ifm = nlmsg_data(nlh);
4791 	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4792 	if (!pfx)
4793 		return -EINVAL;
4794 
4795 	ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4796 
4797 	/* We ignore other flags so far. */
4798 	ifa_flags &= IFA_F_MANAGETEMPADDR;
4799 
4800 	rtnl_net_lock(net);
4801 	err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4802 			     ifm->ifa_prefixlen, extack);
4803 	rtnl_net_unlock(net);
4804 
4805 	return err;
4806 }
4807 
modify_prefix_route(struct net * net,struct inet6_ifaddr * ifp,unsigned long expires,u32 flags,bool modify_peer)4808 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp,
4809 			       unsigned long expires, u32 flags,
4810 			       bool modify_peer)
4811 {
4812 	struct fib6_table *table;
4813 	struct fib6_info *f6i;
4814 	u32 prio;
4815 
4816 	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4817 					ifp->prefix_len,
4818 					ifp->idev->dev, 0, RTF_DEFAULT, true);
4819 	if (!f6i)
4820 		return -ENOENT;
4821 
4822 	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4823 	if (f6i->fib6_metric != prio) {
4824 		/* delete old one */
4825 		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4826 
4827 		/* add new one */
4828 		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4829 				      ifp->prefix_len,
4830 				      ifp->rt_priority, ifp->idev->dev,
4831 				      expires, flags, GFP_KERNEL);
4832 		return 0;
4833 	}
4834 	if (f6i != net->ipv6.fib6_null_entry) {
4835 		table = f6i->fib6_table;
4836 		spin_lock_bh(&table->tb6_lock);
4837 
4838 		if (!(flags & RTF_EXPIRES)) {
4839 			fib6_clean_expires(f6i);
4840 			fib6_remove_gc_list(f6i);
4841 		} else {
4842 			fib6_set_expires(f6i, expires);
4843 			fib6_add_gc_list(f6i);
4844 		}
4845 
4846 		spin_unlock_bh(&table->tb6_lock);
4847 	}
4848 	fib6_info_release(f6i);
4849 
4850 	return 0;
4851 }
4852 
inet6_addr_modify(struct net * net,struct inet6_ifaddr * ifp,struct ifa6_config * cfg,clock_t expires,u32 flags)4853 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4854 			     struct ifa6_config *cfg, clock_t expires,
4855 			     u32 flags)
4856 {
4857 	bool was_managetempaddr;
4858 	bool new_peer = false;
4859 	bool had_prefixroute;
4860 
4861 	ASSERT_RTNL_NET(net);
4862 
4863 	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4864 	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4865 		return -EINVAL;
4866 
4867 	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4868 		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4869 
4870 	if (cfg->peer_pfx &&
4871 	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4872 		if (!ipv6_addr_any(&ifp->peer_addr))
4873 			cleanup_prefix_route(ifp, expires, true, true);
4874 		new_peer = true;
4875 	}
4876 
4877 	spin_lock_bh(&ifp->lock);
4878 	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4879 	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4880 			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4881 	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4882 			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4883 			IFA_F_NOPREFIXROUTE);
4884 	ifp->flags |= cfg->ifa_flags;
4885 	WRITE_ONCE(ifp->tstamp, jiffies);
4886 	WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4887 	WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4888 	WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4889 
4890 	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4891 		WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4892 
4893 	if (new_peer)
4894 		ifp->peer_addr = *cfg->peer_pfx;
4895 
4896 	spin_unlock_bh(&ifp->lock);
4897 	if (!(ifp->flags&IFA_F_TENTATIVE))
4898 		ipv6_ifa_notify(0, ifp);
4899 
4900 	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4901 		int rc = -ENOENT;
4902 
4903 		if (had_prefixroute)
4904 			rc = modify_prefix_route(net, ifp, expires, flags, false);
4905 
4906 		/* prefix route could have been deleted; if so restore it */
4907 		if (rc == -ENOENT) {
4908 			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4909 					      ifp->rt_priority, ifp->idev->dev,
4910 					      expires, flags, GFP_KERNEL);
4911 		}
4912 
4913 		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4914 			rc = modify_prefix_route(net, ifp, expires, flags, true);
4915 
4916 		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4917 			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4918 					      ifp->rt_priority, ifp->idev->dev,
4919 					      expires, flags, GFP_KERNEL);
4920 		}
4921 	} else if (had_prefixroute) {
4922 		enum cleanup_prefix_rt_t action;
4923 		unsigned long rt_expires;
4924 
4925 		write_lock_bh(&ifp->idev->lock);
4926 		action = check_cleanup_prefix_route(ifp, &rt_expires);
4927 		write_unlock_bh(&ifp->idev->lock);
4928 
4929 		if (action != CLEANUP_PREFIX_RT_NOP) {
4930 			cleanup_prefix_route(ifp, rt_expires,
4931 				action == CLEANUP_PREFIX_RT_DEL, false);
4932 		}
4933 	}
4934 
4935 	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4936 		if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4937 			delete_tempaddrs(ifp->idev, ifp);
4938 		else
4939 			manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4940 					 cfg->preferred_lft, !was_managetempaddr,
4941 					 jiffies);
4942 	}
4943 
4944 	addrconf_verify_rtnl(net);
4945 
4946 	return 0;
4947 }
4948 
4949 static int
inet6_rtm_newaddr(struct sk_buff * skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)4950 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4951 		  struct netlink_ext_ack *extack)
4952 {
4953 	struct net *net = sock_net(skb->sk);
4954 	struct nlattr *tb[IFA_MAX+1];
4955 	struct in6_addr *peer_pfx;
4956 	struct inet6_ifaddr *ifa;
4957 	struct net_device *dev;
4958 	struct inet6_dev *idev;
4959 	struct ifa6_config cfg;
4960 	struct ifaddrmsg *ifm;
4961 	unsigned long timeout;
4962 	clock_t expires;
4963 	u32 flags;
4964 	int err;
4965 
4966 	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4967 				     ifa_ipv6_policy, extack);
4968 	if (err < 0)
4969 		return err;
4970 
4971 	memset(&cfg, 0, sizeof(cfg));
4972 
4973 	ifm = nlmsg_data(nlh);
4974 	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4975 	if (!cfg.pfx)
4976 		return -EINVAL;
4977 
4978 	cfg.peer_pfx = peer_pfx;
4979 	cfg.plen = ifm->ifa_prefixlen;
4980 	if (tb[IFA_RT_PRIORITY])
4981 		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4982 
4983 	if (tb[IFA_PROTO])
4984 		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4985 
4986 	cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
4987 
4988 	/* We ignore other flags so far. */
4989 	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4990 			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4991 			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4992 
4993 	cfg.ifa_flags |= IFA_F_PERMANENT;
4994 	cfg.valid_lft = INFINITY_LIFE_TIME;
4995 	cfg.preferred_lft = INFINITY_LIFE_TIME;
4996 	expires = 0;
4997 	flags = 0;
4998 
4999 	if (tb[IFA_CACHEINFO]) {
5000 		struct ifa_cacheinfo *ci;
5001 
5002 		ci = nla_data(tb[IFA_CACHEINFO]);
5003 		cfg.valid_lft = ci->ifa_valid;
5004 		cfg.preferred_lft = ci->ifa_prefered;
5005 
5006 		if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) {
5007 			NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
5008 			return -EINVAL;
5009 		}
5010 
5011 		timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ);
5012 		if (addrconf_finite_timeout(timeout)) {
5013 			cfg.ifa_flags &= ~IFA_F_PERMANENT;
5014 			cfg.valid_lft = timeout;
5015 			expires = jiffies_to_clock_t(timeout * HZ);
5016 			flags = RTF_EXPIRES;
5017 		}
5018 
5019 		timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ);
5020 		if (addrconf_finite_timeout(timeout)) {
5021 			if (timeout == 0)
5022 				cfg.ifa_flags |= IFA_F_DEPRECATED;
5023 
5024 			cfg.preferred_lft = timeout;
5025 		}
5026 	}
5027 
5028 	rtnl_net_lock(net);
5029 
5030 	dev =  __dev_get_by_index(net, ifm->ifa_index);
5031 	if (!dev) {
5032 		NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5033 		err = -ENODEV;
5034 		goto unlock_rtnl;
5035 	}
5036 
5037 	netdev_lock_ops(dev);
5038 	idev = ipv6_find_idev(dev);
5039 	if (IS_ERR(idev)) {
5040 		err = PTR_ERR(idev);
5041 		goto unlock;
5042 	}
5043 
5044 	if (!ipv6_allow_optimistic_dad(net, idev))
5045 		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5046 
5047 	if (cfg.ifa_flags & IFA_F_NODAD &&
5048 	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5049 		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5050 		err = -EINVAL;
5051 		goto unlock;
5052 	}
5053 
5054 	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5055 	if (!ifa) {
5056 		/*
5057 		 * It would be best to check for !NLM_F_CREATE here but
5058 		 * userspace already relies on not having to provide this.
5059 		 */
5060 		err = inet6_addr_add(net, dev, &cfg, expires, flags, extack);
5061 		goto unlock;
5062 	}
5063 
5064 	if (nlh->nlmsg_flags & NLM_F_EXCL ||
5065 	    !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5066 		NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5067 		err = -EEXIST;
5068 	} else {
5069 		err = inet6_addr_modify(net, ifa, &cfg, expires, flags);
5070 	}
5071 
5072 	in6_ifa_put(ifa);
5073 unlock:
5074 	netdev_unlock_ops(dev);
5075 unlock_rtnl:
5076 	rtnl_net_unlock(net);
5077 
5078 	return err;
5079 }
5080 
put_ifaddrmsg(struct nlmsghdr * nlh,u8 prefixlen,u32 flags,u8 scope,int ifindex)5081 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5082 			  u8 scope, int ifindex)
5083 {
5084 	struct ifaddrmsg *ifm;
5085 
5086 	ifm = nlmsg_data(nlh);
5087 	ifm->ifa_family = AF_INET6;
5088 	ifm->ifa_prefixlen = prefixlen;
5089 	ifm->ifa_flags = flags;
5090 	ifm->ifa_scope = scope;
5091 	ifm->ifa_index = ifindex;
5092 }
5093 
put_cacheinfo(struct sk_buff * skb,unsigned long cstamp,unsigned long tstamp,u32 preferred,u32 valid)5094 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5095 			 unsigned long tstamp, u32 preferred, u32 valid)
5096 {
5097 	struct ifa_cacheinfo ci;
5098 
5099 	ci.cstamp = cstamp_delta(cstamp);
5100 	ci.tstamp = cstamp_delta(tstamp);
5101 	ci.ifa_prefered = preferred;
5102 	ci.ifa_valid = valid;
5103 
5104 	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5105 }
5106 
rt_scope(int ifa_scope)5107 static inline int rt_scope(int ifa_scope)
5108 {
5109 	if (ifa_scope & IFA_HOST)
5110 		return RT_SCOPE_HOST;
5111 	else if (ifa_scope & IFA_LINK)
5112 		return RT_SCOPE_LINK;
5113 	else if (ifa_scope & IFA_SITE)
5114 		return RT_SCOPE_SITE;
5115 	else
5116 		return RT_SCOPE_UNIVERSE;
5117 }
5118 
inet6_ifaddr_msgsize(void)5119 static inline int inet6_ifaddr_msgsize(void)
5120 {
5121 	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5122 	       + nla_total_size(16) /* IFA_LOCAL */
5123 	       + nla_total_size(16) /* IFA_ADDRESS */
5124 	       + nla_total_size(sizeof(struct ifa_cacheinfo))
5125 	       + nla_total_size(4)  /* IFA_FLAGS */
5126 	       + nla_total_size(1)  /* IFA_PROTO */
5127 	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5128 }
5129 
inet6_fill_ifaddr(struct sk_buff * skb,const struct inet6_ifaddr * ifa,struct inet6_fill_args * args)5130 static int inet6_fill_ifaddr(struct sk_buff *skb,
5131 			     const struct inet6_ifaddr *ifa,
5132 			     struct inet6_fill_args *args)
5133 {
5134 	struct nlmsghdr *nlh;
5135 	u32 preferred, valid;
5136 	u32 flags, priority;
5137 	u8 proto;
5138 
5139 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5140 			sizeof(struct ifaddrmsg), args->flags);
5141 	if (!nlh)
5142 		return -EMSGSIZE;
5143 
5144 	flags = READ_ONCE(ifa->flags);
5145 	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5146 		      ifa->idev->dev->ifindex);
5147 
5148 	if (args->netnsid >= 0 &&
5149 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5150 		goto error;
5151 
5152 	preferred = READ_ONCE(ifa->prefered_lft);
5153 	valid = READ_ONCE(ifa->valid_lft);
5154 
5155 	if (!((flags & IFA_F_PERMANENT) &&
5156 	      (preferred == INFINITY_LIFE_TIME))) {
5157 		if (preferred != INFINITY_LIFE_TIME) {
5158 			long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5159 
5160 			if (preferred > tval)
5161 				preferred -= tval;
5162 			else
5163 				preferred = 0;
5164 			if (valid != INFINITY_LIFE_TIME) {
5165 				if (valid > tval)
5166 					valid -= tval;
5167 				else
5168 					valid = 0;
5169 			}
5170 		}
5171 	} else {
5172 		preferred = INFINITY_LIFE_TIME;
5173 		valid = INFINITY_LIFE_TIME;
5174 	}
5175 
5176 	if (!ipv6_addr_any(&ifa->peer_addr)) {
5177 		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5178 		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5179 			goto error;
5180 	} else {
5181 		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5182 			goto error;
5183 	}
5184 
5185 	priority = READ_ONCE(ifa->rt_priority);
5186 	if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5187 		goto error;
5188 
5189 	if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5190 			  preferred, valid) < 0)
5191 		goto error;
5192 
5193 	if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5194 		goto error;
5195 
5196 	proto = READ_ONCE(ifa->ifa_proto);
5197 	if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5198 		goto error;
5199 
5200 	nlmsg_end(skb, nlh);
5201 	return 0;
5202 
5203 error:
5204 	nlmsg_cancel(skb, nlh);
5205 	return -EMSGSIZE;
5206 }
5207 
inet6_fill_ifmcaddr(struct sk_buff * skb,const struct ifmcaddr6 * ifmca,struct inet6_fill_args * args)5208 int inet6_fill_ifmcaddr(struct sk_buff *skb,
5209 			const struct ifmcaddr6 *ifmca,
5210 			struct inet6_fill_args *args)
5211 {
5212 	int ifindex = ifmca->idev->dev->ifindex;
5213 	u8 scope = RT_SCOPE_UNIVERSE;
5214 	struct nlmsghdr *nlh;
5215 
5216 	if (!args->force_rt_scope_universe &&
5217 	    ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5218 		scope = RT_SCOPE_SITE;
5219 
5220 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5221 			sizeof(struct ifaddrmsg), args->flags);
5222 	if (!nlh)
5223 		return -EMSGSIZE;
5224 
5225 	if (args->netnsid >= 0 &&
5226 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5227 		nlmsg_cancel(skb, nlh);
5228 		return -EMSGSIZE;
5229 	}
5230 
5231 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5232 	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5233 	    put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5234 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5235 		nlmsg_cancel(skb, nlh);
5236 		return -EMSGSIZE;
5237 	}
5238 
5239 	nlmsg_end(skb, nlh);
5240 	return 0;
5241 }
5242 
inet6_fill_ifacaddr(struct sk_buff * skb,const struct ifacaddr6 * ifaca,struct inet6_fill_args * args)5243 int inet6_fill_ifacaddr(struct sk_buff *skb,
5244 			const struct ifacaddr6 *ifaca,
5245 			struct inet6_fill_args *args)
5246 {
5247 	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5248 	int ifindex = dev ? dev->ifindex : 1;
5249 	u8 scope = RT_SCOPE_UNIVERSE;
5250 	struct nlmsghdr *nlh;
5251 
5252 	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5253 		scope = RT_SCOPE_SITE;
5254 
5255 	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5256 			sizeof(struct ifaddrmsg), args->flags);
5257 	if (!nlh)
5258 		return -EMSGSIZE;
5259 
5260 	if (args->netnsid >= 0 &&
5261 	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5262 		nlmsg_cancel(skb, nlh);
5263 		return -EMSGSIZE;
5264 	}
5265 
5266 	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5267 	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5268 	    put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5269 			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5270 		nlmsg_cancel(skb, nlh);
5271 		return -EMSGSIZE;
5272 	}
5273 
5274 	nlmsg_end(skb, nlh);
5275 	return 0;
5276 }
5277 
5278 /* called with rcu_read_lock() */
in6_dump_addrs(const struct inet6_dev * idev,struct sk_buff * skb,struct netlink_callback * cb,int * s_ip_idx,struct inet6_fill_args * fillargs)5279 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5280 			  struct netlink_callback *cb, int *s_ip_idx,
5281 			  struct inet6_fill_args *fillargs)
5282 {
5283 	const struct ifmcaddr6 *ifmca;
5284 	const struct ifacaddr6 *ifaca;
5285 	int ip_idx = 0;
5286 	int err = 0;
5287 
5288 	switch (fillargs->type) {
5289 	case UNICAST_ADDR: {
5290 		const struct inet6_ifaddr *ifa;
5291 		fillargs->event = RTM_NEWADDR;
5292 
5293 		/* unicast address incl. temp addr */
5294 		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5295 			if (ip_idx < *s_ip_idx)
5296 				goto next;
5297 			err = inet6_fill_ifaddr(skb, ifa, fillargs);
5298 			if (err < 0)
5299 				break;
5300 			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5301 next:
5302 			ip_idx++;
5303 		}
5304 		break;
5305 	}
5306 	case MULTICAST_ADDR:
5307 		fillargs->event = RTM_GETMULTICAST;
5308 
5309 		/* multicast address */
5310 		for (ifmca = rcu_dereference(idev->mc_list);
5311 		     ifmca;
5312 		     ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5313 			if (ip_idx < *s_ip_idx)
5314 				continue;
5315 			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5316 			if (err < 0)
5317 				break;
5318 		}
5319 		break;
5320 	case ANYCAST_ADDR:
5321 		fillargs->event = RTM_GETANYCAST;
5322 		/* anycast address */
5323 		for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5324 		     ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5325 			if (ip_idx < *s_ip_idx)
5326 				continue;
5327 			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5328 			if (err < 0)
5329 				break;
5330 		}
5331 		break;
5332 	default:
5333 		break;
5334 	}
5335 	*s_ip_idx = err ? ip_idx : 0;
5336 	return err;
5337 }
5338 
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)5339 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5340 				       struct inet6_fill_args *fillargs,
5341 				       struct net **tgt_net, struct sock *sk,
5342 				       struct netlink_callback *cb)
5343 {
5344 	struct netlink_ext_ack *extack = cb->extack;
5345 	struct nlattr *tb[IFA_MAX+1];
5346 	struct ifaddrmsg *ifm;
5347 	int err, i;
5348 
5349 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5350 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5351 		return -EINVAL;
5352 	}
5353 
5354 	ifm = nlmsg_data(nlh);
5355 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5356 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5357 		return -EINVAL;
5358 	}
5359 
5360 	fillargs->ifindex = ifm->ifa_index;
5361 	if (fillargs->ifindex) {
5362 		cb->answer_flags |= NLM_F_DUMP_FILTERED;
5363 		fillargs->flags |= NLM_F_DUMP_FILTERED;
5364 	}
5365 
5366 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5367 					    ifa_ipv6_policy, extack);
5368 	if (err < 0)
5369 		return err;
5370 
5371 	for (i = 0; i <= IFA_MAX; ++i) {
5372 		if (!tb[i])
5373 			continue;
5374 
5375 		if (i == IFA_TARGET_NETNSID) {
5376 			struct net *net;
5377 
5378 			fillargs->netnsid = nla_get_s32(tb[i]);
5379 			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5380 			if (IS_ERR(net)) {
5381 				fillargs->netnsid = -1;
5382 				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5383 				return PTR_ERR(net);
5384 			}
5385 			*tgt_net = net;
5386 		} else {
5387 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5388 			return -EINVAL;
5389 		}
5390 	}
5391 
5392 	return 0;
5393 }
5394 
inet6_dump_addr(struct sk_buff * skb,struct netlink_callback * cb,enum addr_type_t type)5395 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5396 			   enum addr_type_t type)
5397 {
5398 	struct net *tgt_net = sock_net(skb->sk);
5399 	const struct nlmsghdr *nlh = cb->nlh;
5400 	struct inet6_fill_args fillargs = {
5401 		.portid = NETLINK_CB(cb->skb).portid,
5402 		.seq = cb->nlh->nlmsg_seq,
5403 		.flags = NLM_F_MULTI,
5404 		.netnsid = -1,
5405 		.type = type,
5406 		.force_rt_scope_universe = false,
5407 	};
5408 	struct {
5409 		unsigned long ifindex;
5410 		int ip_idx;
5411 	} *ctx = (void *)cb->ctx;
5412 	struct net_device *dev;
5413 	struct inet6_dev *idev;
5414 	int err = 0;
5415 
5416 	rcu_read_lock();
5417 	if (cb->strict_check) {
5418 		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5419 						  skb->sk, cb);
5420 		if (err < 0)
5421 			goto done;
5422 
5423 		err = 0;
5424 		if (fillargs.ifindex) {
5425 			dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5426 			if (!dev) {
5427 				err = -ENODEV;
5428 				goto done;
5429 			}
5430 			idev = __in6_dev_get(dev);
5431 			if (idev)
5432 				err = in6_dump_addrs(idev, skb, cb,
5433 						     &ctx->ip_idx,
5434 						     &fillargs);
5435 			goto done;
5436 		}
5437 	}
5438 
5439 	cb->seq = inet6_base_seq(tgt_net);
5440 	for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5441 		idev = __in6_dev_get(dev);
5442 		if (!idev)
5443 			continue;
5444 		err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5445 				     &fillargs);
5446 		if (err < 0)
5447 			goto done;
5448 	}
5449 done:
5450 	rcu_read_unlock();
5451 	if (fillargs.netnsid >= 0)
5452 		put_net(tgt_net);
5453 
5454 	return err;
5455 }
5456 
inet6_dump_ifaddr(struct sk_buff * skb,struct netlink_callback * cb)5457 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5458 {
5459 	enum addr_type_t type = UNICAST_ADDR;
5460 
5461 	return inet6_dump_addr(skb, cb, type);
5462 }
5463 
inet6_dump_ifmcaddr(struct sk_buff * skb,struct netlink_callback * cb)5464 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5465 {
5466 	enum addr_type_t type = MULTICAST_ADDR;
5467 
5468 	return inet6_dump_addr(skb, cb, type);
5469 }
5470 
5471 
inet6_dump_ifacaddr(struct sk_buff * skb,struct netlink_callback * cb)5472 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5473 {
5474 	enum addr_type_t type = ANYCAST_ADDR;
5475 
5476 	return inet6_dump_addr(skb, cb, type);
5477 }
5478 
inet6_rtm_valid_getaddr_req(struct sk_buff * skb,const struct nlmsghdr * nlh,struct nlattr ** tb,struct netlink_ext_ack * extack)5479 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5480 				       const struct nlmsghdr *nlh,
5481 				       struct nlattr **tb,
5482 				       struct netlink_ext_ack *extack)
5483 {
5484 	struct ifaddrmsg *ifm;
5485 	int i, err;
5486 
5487 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5488 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5489 		return -EINVAL;
5490 	}
5491 
5492 	if (!netlink_strict_get_check(skb))
5493 		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5494 					      ifa_ipv6_policy, extack);
5495 
5496 	ifm = nlmsg_data(nlh);
5497 	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5498 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5499 		return -EINVAL;
5500 	}
5501 
5502 	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5503 					    ifa_ipv6_policy, extack);
5504 	if (err)
5505 		return err;
5506 
5507 	for (i = 0; i <= IFA_MAX; i++) {
5508 		if (!tb[i])
5509 			continue;
5510 
5511 		switch (i) {
5512 		case IFA_TARGET_NETNSID:
5513 		case IFA_ADDRESS:
5514 		case IFA_LOCAL:
5515 			break;
5516 		default:
5517 			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5518 			return -EINVAL;
5519 		}
5520 	}
5521 
5522 	return 0;
5523 }
5524 
inet6_rtm_getaddr(struct sk_buff * in_skb,struct nlmsghdr * nlh,struct netlink_ext_ack * extack)5525 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5526 			     struct netlink_ext_ack *extack)
5527 {
5528 	struct net *tgt_net = sock_net(in_skb->sk);
5529 	struct inet6_fill_args fillargs = {
5530 		.portid = NETLINK_CB(in_skb).portid,
5531 		.seq = nlh->nlmsg_seq,
5532 		.event = RTM_NEWADDR,
5533 		.flags = 0,
5534 		.netnsid = -1,
5535 		.force_rt_scope_universe = false,
5536 	};
5537 	struct ifaddrmsg *ifm;
5538 	struct nlattr *tb[IFA_MAX+1];
5539 	struct in6_addr *addr = NULL, *peer;
5540 	struct net_device *dev = NULL;
5541 	struct inet6_ifaddr *ifa;
5542 	struct sk_buff *skb;
5543 	int err;
5544 
5545 	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5546 	if (err < 0)
5547 		return err;
5548 
5549 	if (tb[IFA_TARGET_NETNSID]) {
5550 		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5551 
5552 		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5553 						  fillargs.netnsid);
5554 		if (IS_ERR(tgt_net))
5555 			return PTR_ERR(tgt_net);
5556 	}
5557 
5558 	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5559 	if (!addr) {
5560 		err = -EINVAL;
5561 		goto errout;
5562 	}
5563 	ifm = nlmsg_data(nlh);
5564 	if (ifm->ifa_index)
5565 		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5566 
5567 	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5568 	if (!ifa) {
5569 		err = -EADDRNOTAVAIL;
5570 		goto errout;
5571 	}
5572 
5573 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5574 	if (!skb) {
5575 		err = -ENOBUFS;
5576 		goto errout_ifa;
5577 	}
5578 
5579 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5580 	if (err < 0) {
5581 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5582 		WARN_ON(err == -EMSGSIZE);
5583 		kfree_skb(skb);
5584 		goto errout_ifa;
5585 	}
5586 	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5587 errout_ifa:
5588 	in6_ifa_put(ifa);
5589 errout:
5590 	dev_put(dev);
5591 	if (fillargs.netnsid >= 0)
5592 		put_net(tgt_net);
5593 
5594 	return err;
5595 }
5596 
inet6_ifa_notify(int event,struct inet6_ifaddr * ifa)5597 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5598 {
5599 	struct sk_buff *skb;
5600 	struct net *net = dev_net(ifa->idev->dev);
5601 	struct inet6_fill_args fillargs = {
5602 		.portid = 0,
5603 		.seq = 0,
5604 		.event = event,
5605 		.flags = 0,
5606 		.netnsid = -1,
5607 		.force_rt_scope_universe = false,
5608 	};
5609 	int err = -ENOBUFS;
5610 
5611 	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5612 	if (!skb)
5613 		goto errout;
5614 
5615 	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5616 	if (err < 0) {
5617 		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5618 		WARN_ON(err == -EMSGSIZE);
5619 		kfree_skb(skb);
5620 		goto errout;
5621 	}
5622 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5623 	return;
5624 errout:
5625 	rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5626 }
5627 
ipv6_store_devconf(const struct ipv6_devconf * cnf,__s32 * array,int bytes)5628 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5629 			       __s32 *array, int bytes)
5630 {
5631 	BUG_ON(bytes < (DEVCONF_MAX * 4));
5632 
5633 	memset(array, 0, bytes);
5634 	array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5635 	array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5636 	array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5637 	array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5638 	array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5639 	array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5640 	array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5641 	array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5642 	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5643 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5644 	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5645 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5646 	array[DEVCONF_RTR_SOLICIT_DELAY] =
5647 		jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5648 	array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5649 	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5650 		jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5651 	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5652 		jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5653 	array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5654 	array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5655 	array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5656 	array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5657 	array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5658 	array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5659 	array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5660 	array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5661 	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5662 		READ_ONCE(cnf->accept_ra_min_hop_limit);
5663 	array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5664 #ifdef CONFIG_IPV6_ROUTER_PREF
5665 	array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5666 	array[DEVCONF_RTR_PROBE_INTERVAL] =
5667 		jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5668 #ifdef CONFIG_IPV6_ROUTE_INFO
5669 	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5670 		READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5671 	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5672 		READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5673 #endif
5674 #endif
5675 	array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5676 	array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5677 		READ_ONCE(cnf->accept_source_route);
5678 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5679 	array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5680 	array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5681 #endif
5682 #ifdef CONFIG_IPV6_MROUTE
5683 	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5684 #endif
5685 	array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5686 	array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5687 	array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5688 	array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5689 	array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5690 		READ_ONCE(cnf->suppress_frag_ndisc);
5691 	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5692 		READ_ONCE(cnf->accept_ra_from_local);
5693 	array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5694 	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5695 		READ_ONCE(cnf->ignore_routes_with_linkdown);
5696 	/* we omit DEVCONF_STABLE_SECRET for now */
5697 	array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5698 	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5699 		READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5700 	array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5701 	array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5702 	array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5703 #ifdef CONFIG_IPV6_SEG6_HMAC
5704 	array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5705 #endif
5706 	array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5707 	array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5708 	array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5709 	array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5710 	array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5711 	array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5712 	array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5713 	array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5714 	array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5715 		READ_ONCE(cnf->ndisc_evict_nocarrier);
5716 	array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5717 		READ_ONCE(cnf->accept_untracked_na);
5718 	array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5719 }
5720 
inet6_ifla6_size(void)5721 static inline size_t inet6_ifla6_size(void)
5722 {
5723 	return nla_total_size(4) /* IFLA_INET6_FLAGS */
5724 	     + nla_total_size(sizeof(struct ifla_cacheinfo))
5725 	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5726 	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5727 	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5728 	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5729 	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5730 	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5731 	     + 0;
5732 }
5733 
inet6_if_nlmsg_size(void)5734 static inline size_t inet6_if_nlmsg_size(void)
5735 {
5736 	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5737 	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5738 	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5739 	       + nla_total_size(4) /* IFLA_MTU */
5740 	       + nla_total_size(4) /* IFLA_LINK */
5741 	       + nla_total_size(1) /* IFLA_OPERSTATE */
5742 	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5743 }
5744 
__snmp6_fill_statsdev(u64 * stats,atomic_long_t * mib,int bytes)5745 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5746 					int bytes)
5747 {
5748 	int i;
5749 	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5750 	BUG_ON(pad < 0);
5751 
5752 	/* Use put_unaligned() because stats may not be aligned for u64. */
5753 	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5754 	for (i = 1; i < ICMP6_MIB_MAX; i++)
5755 		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5756 
5757 	memset(&stats[ICMP6_MIB_MAX], 0, pad);
5758 }
5759 
__snmp6_fill_stats64(u64 * stats,void __percpu * mib,int bytes,size_t syncpoff)5760 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5761 					int bytes, size_t syncpoff)
5762 {
5763 	int i, c;
5764 	u64 buff[IPSTATS_MIB_MAX];
5765 	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5766 
5767 	BUG_ON(pad < 0);
5768 
5769 	memset(buff, 0, sizeof(buff));
5770 	buff[0] = IPSTATS_MIB_MAX;
5771 
5772 	for_each_possible_cpu(c) {
5773 		for (i = 1; i < IPSTATS_MIB_MAX; i++)
5774 			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5775 	}
5776 
5777 	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5778 	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5779 }
5780 
snmp6_fill_stats(u64 * stats,struct inet6_dev * idev,int attrtype,int bytes)5781 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5782 			     int bytes)
5783 {
5784 	switch (attrtype) {
5785 	case IFLA_INET6_STATS:
5786 		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5787 				     offsetof(struct ipstats_mib, syncp));
5788 		break;
5789 	case IFLA_INET6_ICMP6STATS:
5790 		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5791 		break;
5792 	}
5793 }
5794 
inet6_fill_ifla6_stats_attrs(struct sk_buff * skb,struct inet6_dev * idev)5795 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb,
5796 					struct inet6_dev *idev)
5797 {
5798 	struct nlattr *nla;
5799 
5800 	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5801 	if (!nla)
5802 		goto nla_put_failure;
5803 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5804 
5805 	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5806 	if (!nla)
5807 		goto nla_put_failure;
5808 	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5809 
5810 	return 0;
5811 
5812 nla_put_failure:
5813 	return -EMSGSIZE;
5814 }
5815 
inet6_fill_ifla6_attrs(struct sk_buff * skb,struct inet6_dev * idev,u32 ext_filter_mask)5816 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5817 				  u32 ext_filter_mask)
5818 {
5819 	struct ifla_cacheinfo ci;
5820 	struct nlattr *nla;
5821 	u32 ra_mtu;
5822 
5823 	if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5824 		goto nla_put_failure;
5825 	ci.max_reasm_len = IPV6_MAXPLEN;
5826 	ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5827 	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5828 	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5829 	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5830 		goto nla_put_failure;
5831 	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5832 	if (!nla)
5833 		goto nla_put_failure;
5834 	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5835 
5836 	/* XXX - MC not implemented */
5837 
5838 	if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) {
5839 		if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0)
5840 			goto nla_put_failure;
5841 	}
5842 
5843 	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5844 	if (!nla)
5845 		goto nla_put_failure;
5846 	read_lock_bh(&idev->lock);
5847 	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5848 	read_unlock_bh(&idev->lock);
5849 
5850 	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5851 		       READ_ONCE(idev->cnf.addr_gen_mode)))
5852 		goto nla_put_failure;
5853 
5854 	ra_mtu = READ_ONCE(idev->ra_mtu);
5855 	if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5856 		goto nla_put_failure;
5857 
5858 	return 0;
5859 
5860 nla_put_failure:
5861 	return -EMSGSIZE;
5862 }
5863 
inet6_get_link_af_size(const struct net_device * dev,u32 ext_filter_mask)5864 static size_t inet6_get_link_af_size(const struct net_device *dev,
5865 				     u32 ext_filter_mask)
5866 {
5867 	if (!__in6_dev_get(dev))
5868 		return 0;
5869 
5870 	return inet6_ifla6_size();
5871 }
5872 
inet6_fill_link_af(struct sk_buff * skb,const struct net_device * dev,u32 ext_filter_mask)5873 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5874 			      u32 ext_filter_mask)
5875 {
5876 	struct inet6_dev *idev = __in6_dev_get(dev);
5877 
5878 	if (!idev)
5879 		return -ENODATA;
5880 
5881 	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5882 		return -EMSGSIZE;
5883 
5884 	return 0;
5885 }
5886 
inet6_set_iftoken(struct inet6_dev * idev,struct in6_addr * token,struct netlink_ext_ack * extack)5887 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5888 			     struct netlink_ext_ack *extack)
5889 {
5890 	struct inet6_ifaddr *ifp;
5891 	struct net_device *dev = idev->dev;
5892 	bool clear_token, update_rs = false;
5893 	struct in6_addr ll_addr;
5894 
5895 	ASSERT_RTNL();
5896 
5897 	if (!token)
5898 		return -EINVAL;
5899 
5900 	if (dev->flags & IFF_LOOPBACK) {
5901 		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5902 		return -EINVAL;
5903 	}
5904 
5905 	if (dev->flags & IFF_NOARP) {
5906 		NL_SET_ERR_MSG_MOD(extack,
5907 				   "Device does not do neighbour discovery");
5908 		return -EINVAL;
5909 	}
5910 
5911 	if (!ipv6_accept_ra(idev)) {
5912 		NL_SET_ERR_MSG_MOD(extack,
5913 				   "Router advertisement is disabled on device");
5914 		return -EINVAL;
5915 	}
5916 
5917 	if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5918 		NL_SET_ERR_MSG(extack,
5919 			       "Router solicitation is disabled on device");
5920 		return -EINVAL;
5921 	}
5922 
5923 	write_lock_bh(&idev->lock);
5924 
5925 	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5926 	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5927 
5928 	write_unlock_bh(&idev->lock);
5929 
5930 	clear_token = ipv6_addr_any(token);
5931 	if (clear_token)
5932 		goto update_lft;
5933 
5934 	if (!idev->dead && (idev->if_flags & IF_READY) &&
5935 	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5936 			     IFA_F_OPTIMISTIC)) {
5937 		/* If we're not ready, then normal ifup will take care
5938 		 * of this. Otherwise, we need to request our rs here.
5939 		 */
5940 		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5941 		update_rs = true;
5942 	}
5943 
5944 update_lft:
5945 	write_lock_bh(&idev->lock);
5946 
5947 	if (update_rs) {
5948 		idev->if_flags |= IF_RS_SENT;
5949 		idev->rs_interval = rfc3315_s14_backoff_init(
5950 			READ_ONCE(idev->cnf.rtr_solicit_interval));
5951 		idev->rs_probes = 1;
5952 		addrconf_mod_rs_timer(idev, idev->rs_interval);
5953 	}
5954 
5955 	/* Well, that's kinda nasty ... */
5956 	list_for_each_entry(ifp, &idev->addr_list, if_list) {
5957 		spin_lock(&ifp->lock);
5958 		if (ifp->tokenized) {
5959 			ifp->valid_lft = 0;
5960 			ifp->prefered_lft = 0;
5961 		}
5962 		spin_unlock(&ifp->lock);
5963 	}
5964 
5965 	write_unlock_bh(&idev->lock);
5966 	inet6_ifinfo_notify(RTM_NEWLINK, idev);
5967 	addrconf_verify_rtnl(dev_net(dev));
5968 	return 0;
5969 }
5970 
5971 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5972 	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
5973 	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
5974 	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
5975 					    .reject_message =
5976 						"IFLA_INET6_RA_MTU can not be set" },
5977 };
5978 
check_addr_gen_mode(int mode)5979 static int check_addr_gen_mode(int mode)
5980 {
5981 	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5982 	    mode != IN6_ADDR_GEN_MODE_NONE &&
5983 	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5984 	    mode != IN6_ADDR_GEN_MODE_RANDOM)
5985 		return -EINVAL;
5986 	return 1;
5987 }
5988 
check_stable_privacy(struct inet6_dev * idev,struct net * net,int mode)5989 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5990 				int mode)
5991 {
5992 	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5993 	    !idev->cnf.stable_secret.initialized &&
5994 	    !net->ipv6.devconf_dflt->stable_secret.initialized)
5995 		return -EINVAL;
5996 	return 1;
5997 }
5998 
inet6_validate_link_af(const struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)5999 static int inet6_validate_link_af(const struct net_device *dev,
6000 				  const struct nlattr *nla,
6001 				  struct netlink_ext_ack *extack)
6002 {
6003 	struct nlattr *tb[IFLA_INET6_MAX + 1];
6004 	struct inet6_dev *idev = NULL;
6005 	int err;
6006 
6007 	if (dev) {
6008 		idev = __in6_dev_get(dev);
6009 		if (!idev)
6010 			return -EAFNOSUPPORT;
6011 	}
6012 
6013 	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
6014 					  inet6_af_policy, extack);
6015 	if (err)
6016 		return err;
6017 
6018 	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6019 		return -EINVAL;
6020 
6021 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6022 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6023 
6024 		if (check_addr_gen_mode(mode) < 0)
6025 			return -EINVAL;
6026 		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6027 			return -EINVAL;
6028 	}
6029 
6030 	return 0;
6031 }
6032 
inet6_set_link_af(struct net_device * dev,const struct nlattr * nla,struct netlink_ext_ack * extack)6033 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6034 			     struct netlink_ext_ack *extack)
6035 {
6036 	struct inet6_dev *idev = __in6_dev_get(dev);
6037 	struct nlattr *tb[IFLA_INET6_MAX + 1];
6038 	int err;
6039 
6040 	if (!idev)
6041 		return -EAFNOSUPPORT;
6042 
6043 	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6044 		return -EINVAL;
6045 
6046 	if (tb[IFLA_INET6_TOKEN]) {
6047 		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6048 					extack);
6049 		if (err)
6050 			return err;
6051 	}
6052 
6053 	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6054 		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6055 
6056 		WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6057 	}
6058 
6059 	return 0;
6060 }
6061 
inet6_fill_ifinfo(struct sk_buff * skb,struct inet6_dev * idev,u32 portid,u32 seq,int event,unsigned int flags)6062 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6063 			     u32 portid, u32 seq, int event, unsigned int flags)
6064 {
6065 	struct net_device *dev = idev->dev;
6066 	struct ifinfomsg *hdr;
6067 	struct nlmsghdr *nlh;
6068 	int ifindex, iflink;
6069 	void *protoinfo;
6070 
6071 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6072 	if (!nlh)
6073 		return -EMSGSIZE;
6074 
6075 	hdr = nlmsg_data(nlh);
6076 	hdr->ifi_family = AF_INET6;
6077 	hdr->__ifi_pad = 0;
6078 	hdr->ifi_type = dev->type;
6079 	ifindex = READ_ONCE(dev->ifindex);
6080 	hdr->ifi_index = ifindex;
6081 	hdr->ifi_flags = dev_get_flags(dev);
6082 	hdr->ifi_change = 0;
6083 
6084 	iflink = dev_get_iflink(dev);
6085 	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6086 	    (dev->addr_len &&
6087 	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6088 	    nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6089 	    (ifindex != iflink &&
6090 	     nla_put_u32(skb, IFLA_LINK, iflink)) ||
6091 	    nla_put_u8(skb, IFLA_OPERSTATE,
6092 		       netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6093 		goto nla_put_failure;
6094 	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6095 	if (!protoinfo)
6096 		goto nla_put_failure;
6097 
6098 	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6099 		goto nla_put_failure;
6100 
6101 	nla_nest_end(skb, protoinfo);
6102 	nlmsg_end(skb, nlh);
6103 	return 0;
6104 
6105 nla_put_failure:
6106 	nlmsg_cancel(skb, nlh);
6107 	return -EMSGSIZE;
6108 }
6109 
inet6_valid_dump_ifinfo(const struct nlmsghdr * nlh,struct netlink_ext_ack * extack)6110 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6111 				   struct netlink_ext_ack *extack)
6112 {
6113 	struct ifinfomsg *ifm;
6114 
6115 	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6116 		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6117 		return -EINVAL;
6118 	}
6119 
6120 	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6121 		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6122 		return -EINVAL;
6123 	}
6124 
6125 	ifm = nlmsg_data(nlh);
6126 	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6127 	    ifm->ifi_change || ifm->ifi_index) {
6128 		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6129 		return -EINVAL;
6130 	}
6131 
6132 	return 0;
6133 }
6134 
inet6_dump_ifinfo(struct sk_buff * skb,struct netlink_callback * cb)6135 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6136 {
6137 	struct net *net = sock_net(skb->sk);
6138 	struct {
6139 		unsigned long ifindex;
6140 	} *ctx = (void *)cb->ctx;
6141 	struct net_device *dev;
6142 	struct inet6_dev *idev;
6143 	int err;
6144 
6145 	/* only requests using strict checking can pass data to
6146 	 * influence the dump
6147 	 */
6148 	if (cb->strict_check) {
6149 		err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6150 
6151 		if (err < 0)
6152 			return err;
6153 	}
6154 
6155 	err = 0;
6156 	rcu_read_lock();
6157 	for_each_netdev_dump(net, dev, ctx->ifindex) {
6158 		idev = __in6_dev_get(dev);
6159 		if (!idev)
6160 			continue;
6161 		err = inet6_fill_ifinfo(skb, idev,
6162 					NETLINK_CB(cb->skb).portid,
6163 					cb->nlh->nlmsg_seq,
6164 					RTM_NEWLINK, NLM_F_MULTI);
6165 		if (err < 0)
6166 			break;
6167 	}
6168 	rcu_read_unlock();
6169 
6170 	return err;
6171 }
6172 
inet6_ifinfo_notify(int event,struct inet6_dev * idev)6173 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6174 {
6175 	struct sk_buff *skb;
6176 	struct net *net = dev_net(idev->dev);
6177 	int err = -ENOBUFS;
6178 
6179 	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6180 	if (!skb)
6181 		goto errout;
6182 
6183 	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6184 	if (err < 0) {
6185 		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6186 		WARN_ON(err == -EMSGSIZE);
6187 		kfree_skb(skb);
6188 		goto errout;
6189 	}
6190 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6191 	return;
6192 errout:
6193 	rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6194 }
6195 
inet6_prefix_nlmsg_size(void)6196 static inline size_t inet6_prefix_nlmsg_size(void)
6197 {
6198 	return NLMSG_ALIGN(sizeof(struct prefixmsg))
6199 	       + nla_total_size(sizeof(struct in6_addr))
6200 	       + nla_total_size(sizeof(struct prefix_cacheinfo));
6201 }
6202 
inet6_fill_prefix(struct sk_buff * skb,struct inet6_dev * idev,struct prefix_info * pinfo,u32 portid,u32 seq,int event,unsigned int flags)6203 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6204 			     struct prefix_info *pinfo, u32 portid, u32 seq,
6205 			     int event, unsigned int flags)
6206 {
6207 	struct prefixmsg *pmsg;
6208 	struct nlmsghdr *nlh;
6209 	struct prefix_cacheinfo	ci;
6210 
6211 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6212 	if (!nlh)
6213 		return -EMSGSIZE;
6214 
6215 	pmsg = nlmsg_data(nlh);
6216 	pmsg->prefix_family = AF_INET6;
6217 	pmsg->prefix_pad1 = 0;
6218 	pmsg->prefix_pad2 = 0;
6219 	pmsg->prefix_ifindex = idev->dev->ifindex;
6220 	pmsg->prefix_len = pinfo->prefix_len;
6221 	pmsg->prefix_type = pinfo->type;
6222 	pmsg->prefix_pad3 = 0;
6223 	pmsg->prefix_flags = pinfo->flags;
6224 
6225 	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6226 		goto nla_put_failure;
6227 	ci.preferred_time = ntohl(pinfo->prefered);
6228 	ci.valid_time = ntohl(pinfo->valid);
6229 	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6230 		goto nla_put_failure;
6231 	nlmsg_end(skb, nlh);
6232 	return 0;
6233 
6234 nla_put_failure:
6235 	nlmsg_cancel(skb, nlh);
6236 	return -EMSGSIZE;
6237 }
6238 
inet6_prefix_notify(int event,struct inet6_dev * idev,struct prefix_info * pinfo)6239 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6240 			 struct prefix_info *pinfo)
6241 {
6242 	struct sk_buff *skb;
6243 	struct net *net = dev_net(idev->dev);
6244 	int err = -ENOBUFS;
6245 
6246 	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6247 	if (!skb)
6248 		goto errout;
6249 
6250 	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6251 	if (err < 0) {
6252 		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6253 		WARN_ON(err == -EMSGSIZE);
6254 		kfree_skb(skb);
6255 		goto errout;
6256 	}
6257 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6258 	return;
6259 errout:
6260 	rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6261 }
6262 
__ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6263 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6264 {
6265 	struct net *net = dev_net(ifp->idev->dev);
6266 
6267 	if (event)
6268 		ASSERT_RTNL();
6269 
6270 	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6271 
6272 	switch (event) {
6273 	case RTM_NEWADDR:
6274 		/*
6275 		 * If the address was optimistic we inserted the route at the
6276 		 * start of our DAD process, so we don't need to do it again.
6277 		 * If the device was taken down in the middle of the DAD
6278 		 * cycle there is a race where we could get here without a
6279 		 * host route, so nothing to insert. That will be fixed when
6280 		 * the device is brought up.
6281 		 */
6282 		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6283 			ip6_ins_rt(net, ifp->rt);
6284 		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6285 			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6286 				&ifp->addr, ifp->idev->dev->name);
6287 		}
6288 
6289 		if (ifp->idev->cnf.forwarding)
6290 			addrconf_join_anycast(ifp);
6291 		if (!ipv6_addr_any(&ifp->peer_addr))
6292 			addrconf_prefix_route(&ifp->peer_addr, 128,
6293 					      ifp->rt_priority, ifp->idev->dev,
6294 					      0, 0, GFP_ATOMIC);
6295 		break;
6296 	case RTM_DELADDR:
6297 		if (ifp->idev->cnf.forwarding)
6298 			addrconf_leave_anycast(ifp);
6299 		addrconf_leave_solict(ifp->idev, &ifp->addr);
6300 		if (!ipv6_addr_any(&ifp->peer_addr)) {
6301 			struct fib6_info *rt;
6302 
6303 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6304 						       ifp->idev->dev, 0, 0,
6305 						       false);
6306 			if (rt)
6307 				ip6_del_rt(net, rt, false);
6308 		}
6309 		if (ifp->rt) {
6310 			ip6_del_rt(net, ifp->rt, false);
6311 			ifp->rt = NULL;
6312 		}
6313 		rt_genid_bump_ipv6(net);
6314 		break;
6315 	}
6316 	atomic_inc(&net->ipv6.dev_addr_genid);
6317 }
6318 
ipv6_ifa_notify(int event,struct inet6_ifaddr * ifp)6319 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6320 {
6321 	if (likely(ifp->idev->dead == 0))
6322 		__ipv6_ifa_notify(event, ifp);
6323 }
6324 
6325 #ifdef CONFIG_SYSCTL
6326 
addrconf_sysctl_forward(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6327 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write,
6328 		void *buffer, size_t *lenp, loff_t *ppos)
6329 {
6330 	int *valp = ctl->data;
6331 	int val = *valp;
6332 	loff_t pos = *ppos;
6333 	struct ctl_table lctl;
6334 	int ret;
6335 
6336 	/*
6337 	 * ctl->data points to idev->cnf.forwarding, we should
6338 	 * not modify it until we get the rtnl lock.
6339 	 */
6340 	lctl = *ctl;
6341 	lctl.data = &val;
6342 
6343 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6344 
6345 	if (write)
6346 		ret = addrconf_fixup_forwarding(ctl, valp, val);
6347 	if (ret)
6348 		*ppos = pos;
6349 	return ret;
6350 }
6351 
addrconf_sysctl_mtu(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6352 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write,
6353 		void *buffer, size_t *lenp, loff_t *ppos)
6354 {
6355 	struct inet6_dev *idev = ctl->extra1;
6356 	int min_mtu = IPV6_MIN_MTU;
6357 	struct ctl_table lctl;
6358 
6359 	lctl = *ctl;
6360 	lctl.extra1 = &min_mtu;
6361 	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6362 
6363 	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6364 }
6365 
dev_disable_change(struct inet6_dev * idev)6366 static void dev_disable_change(struct inet6_dev *idev)
6367 {
6368 	struct netdev_notifier_info info;
6369 
6370 	if (!idev || !idev->dev)
6371 		return;
6372 
6373 	netdev_notifier_info_init(&info, idev->dev);
6374 	if (idev->cnf.disable_ipv6)
6375 		addrconf_notify(NULL, NETDEV_DOWN, &info);
6376 	else
6377 		addrconf_notify(NULL, NETDEV_UP, &info);
6378 }
6379 
addrconf_disable_change(struct net * net,__s32 newf)6380 static void addrconf_disable_change(struct net *net, __s32 newf)
6381 {
6382 	struct net_device *dev;
6383 	struct inet6_dev *idev;
6384 
6385 	for_each_netdev(net, dev) {
6386 		idev = __in6_dev_get_rtnl_net(dev);
6387 		if (idev) {
6388 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6389 
6390 			WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6391 			if (changed)
6392 				dev_disable_change(idev);
6393 		}
6394 	}
6395 }
6396 
addrconf_disable_ipv6(const struct ctl_table * table,int * p,int newf)6397 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf)
6398 {
6399 	struct net *net = (struct net *)table->extra2;
6400 	int old;
6401 
6402 	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6403 		WRITE_ONCE(*p, newf);
6404 		return 0;
6405 	}
6406 
6407 	if (!rtnl_net_trylock(net))
6408 		return restart_syscall();
6409 
6410 	old = *p;
6411 	WRITE_ONCE(*p, newf);
6412 
6413 	if (p == &net->ipv6.devconf_all->disable_ipv6) {
6414 		WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6415 		addrconf_disable_change(net, newf);
6416 	} else if ((!newf) ^ (!old)) {
6417 		dev_disable_change((struct inet6_dev *)table->extra1);
6418 	}
6419 
6420 	rtnl_net_unlock(net);
6421 	return 0;
6422 }
6423 
addrconf_sysctl_disable(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6424 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write,
6425 		void *buffer, size_t *lenp, loff_t *ppos)
6426 {
6427 	int *valp = ctl->data;
6428 	int val = *valp;
6429 	loff_t pos = *ppos;
6430 	struct ctl_table lctl;
6431 	int ret;
6432 
6433 	/*
6434 	 * ctl->data points to idev->cnf.disable_ipv6, we should
6435 	 * not modify it until we get the rtnl lock.
6436 	 */
6437 	lctl = *ctl;
6438 	lctl.data = &val;
6439 
6440 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6441 
6442 	if (write)
6443 		ret = addrconf_disable_ipv6(ctl, valp, val);
6444 	if (ret)
6445 		*ppos = pos;
6446 	return ret;
6447 }
6448 
addrconf_sysctl_proxy_ndp(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6449 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write,
6450 		void *buffer, size_t *lenp, loff_t *ppos)
6451 {
6452 	int *valp = ctl->data;
6453 	int ret;
6454 	int old, new;
6455 
6456 	old = *valp;
6457 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6458 	new = *valp;
6459 
6460 	if (write && old != new) {
6461 		struct net *net = ctl->extra2;
6462 
6463 		if (!rtnl_net_trylock(net))
6464 			return restart_syscall();
6465 
6466 		if (valp == &net->ipv6.devconf_dflt->proxy_ndp) {
6467 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6468 						     NETCONFA_PROXY_NEIGH,
6469 						     NETCONFA_IFINDEX_DEFAULT,
6470 						     net->ipv6.devconf_dflt);
6471 		} else if (valp == &net->ipv6.devconf_all->proxy_ndp) {
6472 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6473 						     NETCONFA_PROXY_NEIGH,
6474 						     NETCONFA_IFINDEX_ALL,
6475 						     net->ipv6.devconf_all);
6476 		} else {
6477 			struct inet6_dev *idev = ctl->extra1;
6478 
6479 			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6480 						     NETCONFA_PROXY_NEIGH,
6481 						     idev->dev->ifindex,
6482 						     &idev->cnf);
6483 		}
6484 		rtnl_net_unlock(net);
6485 	}
6486 
6487 	return ret;
6488 }
6489 
addrconf_sysctl_addr_gen_mode(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6490 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write,
6491 					 void *buffer, size_t *lenp,
6492 					 loff_t *ppos)
6493 {
6494 	int ret = 0;
6495 	u32 new_val;
6496 	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6497 	struct net *net = (struct net *)ctl->extra2;
6498 	struct ctl_table tmp = {
6499 		.data = &new_val,
6500 		.maxlen = sizeof(new_val),
6501 		.mode = ctl->mode,
6502 	};
6503 
6504 	if (!rtnl_net_trylock(net))
6505 		return restart_syscall();
6506 
6507 	new_val = *((u32 *)ctl->data);
6508 
6509 	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6510 	if (ret != 0)
6511 		goto out;
6512 
6513 	if (write) {
6514 		if (check_addr_gen_mode(new_val) < 0) {
6515 			ret = -EINVAL;
6516 			goto out;
6517 		}
6518 
6519 		if (idev) {
6520 			if (check_stable_privacy(idev, net, new_val) < 0) {
6521 				ret = -EINVAL;
6522 				goto out;
6523 			}
6524 
6525 			if (idev->cnf.addr_gen_mode != new_val) {
6526 				WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6527 				netdev_lock_ops(idev->dev);
6528 				addrconf_init_auto_addrs(idev->dev);
6529 				netdev_unlock_ops(idev->dev);
6530 			}
6531 		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6532 			struct net_device *dev;
6533 
6534 			WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6535 			for_each_netdev(net, dev) {
6536 				idev = __in6_dev_get_rtnl_net(dev);
6537 				if (idev &&
6538 				    idev->cnf.addr_gen_mode != new_val) {
6539 					WRITE_ONCE(idev->cnf.addr_gen_mode,
6540 						  new_val);
6541 					netdev_lock_ops(idev->dev);
6542 					addrconf_init_auto_addrs(idev->dev);
6543 					netdev_unlock_ops(idev->dev);
6544 				}
6545 			}
6546 		}
6547 
6548 		WRITE_ONCE(*((u32 *)ctl->data), new_val);
6549 	}
6550 
6551 out:
6552 	rtnl_net_unlock(net);
6553 
6554 	return ret;
6555 }
6556 
addrconf_sysctl_stable_secret(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6557 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write,
6558 					 void *buffer, size_t *lenp,
6559 					 loff_t *ppos)
6560 {
6561 	int err;
6562 	struct in6_addr addr;
6563 	char str[IPV6_MAX_STRLEN];
6564 	struct ctl_table lctl = *ctl;
6565 	struct net *net = ctl->extra2;
6566 	struct ipv6_stable_secret *secret = ctl->data;
6567 
6568 	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6569 		return -EIO;
6570 
6571 	lctl.maxlen = IPV6_MAX_STRLEN;
6572 	lctl.data = str;
6573 
6574 	if (!rtnl_net_trylock(net))
6575 		return restart_syscall();
6576 
6577 	if (!write && !secret->initialized) {
6578 		err = -EIO;
6579 		goto out;
6580 	}
6581 
6582 	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6583 	if (err >= sizeof(str)) {
6584 		err = -EIO;
6585 		goto out;
6586 	}
6587 
6588 	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6589 	if (err || !write)
6590 		goto out;
6591 
6592 	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6593 		err = -EIO;
6594 		goto out;
6595 	}
6596 
6597 	secret->initialized = true;
6598 	secret->secret = addr;
6599 
6600 	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6601 		struct net_device *dev;
6602 
6603 		for_each_netdev(net, dev) {
6604 			struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev);
6605 
6606 			if (idev) {
6607 				WRITE_ONCE(idev->cnf.addr_gen_mode,
6608 					   IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6609 			}
6610 		}
6611 	} else {
6612 		struct inet6_dev *idev = ctl->extra1;
6613 
6614 		WRITE_ONCE(idev->cnf.addr_gen_mode,
6615 			   IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6616 	}
6617 
6618 out:
6619 	rtnl_net_unlock(net);
6620 
6621 	return err;
6622 }
6623 
6624 static
addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6625 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl,
6626 						int write, void *buffer,
6627 						size_t *lenp,
6628 						loff_t *ppos)
6629 {
6630 	int *valp = ctl->data;
6631 	int val = *valp;
6632 	loff_t pos = *ppos;
6633 	struct ctl_table lctl;
6634 	int ret;
6635 
6636 	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6637 	 * we should not modify it until we get the rtnl lock.
6638 	 */
6639 	lctl = *ctl;
6640 	lctl.data = &val;
6641 
6642 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6643 
6644 	if (write)
6645 		ret = addrconf_fixup_linkdown(ctl, valp, val);
6646 	if (ret)
6647 		*ppos = pos;
6648 	return ret;
6649 }
6650 
6651 static
addrconf_set_nopolicy(struct rt6_info * rt,int action)6652 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6653 {
6654 	if (rt) {
6655 		if (action)
6656 			rt->dst.flags |= DST_NOPOLICY;
6657 		else
6658 			rt->dst.flags &= ~DST_NOPOLICY;
6659 	}
6660 }
6661 
6662 static
addrconf_disable_policy_idev(struct inet6_dev * idev,int val)6663 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6664 {
6665 	struct inet6_ifaddr *ifa;
6666 
6667 	read_lock_bh(&idev->lock);
6668 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
6669 		spin_lock(&ifa->lock);
6670 		if (ifa->rt) {
6671 			/* host routes only use builtin fib6_nh */
6672 			struct fib6_nh *nh = ifa->rt->fib6_nh;
6673 			int cpu;
6674 
6675 			rcu_read_lock();
6676 			ifa->rt->dst_nopolicy = val ? true : false;
6677 			if (nh->rt6i_pcpu) {
6678 				for_each_possible_cpu(cpu) {
6679 					struct rt6_info **rtp;
6680 
6681 					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6682 					addrconf_set_nopolicy(*rtp, val);
6683 				}
6684 			}
6685 			rcu_read_unlock();
6686 		}
6687 		spin_unlock(&ifa->lock);
6688 	}
6689 	read_unlock_bh(&idev->lock);
6690 }
6691 
6692 static
addrconf_disable_policy(const struct ctl_table * ctl,int * valp,int val)6693 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val)
6694 {
6695 	struct net *net = (struct net *)ctl->extra2;
6696 	struct inet6_dev *idev;
6697 
6698 	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6699 		WRITE_ONCE(*valp, val);
6700 		return 0;
6701 	}
6702 
6703 	if (!rtnl_net_trylock(net))
6704 		return restart_syscall();
6705 
6706 	WRITE_ONCE(*valp, val);
6707 
6708 	if (valp == &net->ipv6.devconf_all->disable_policy)  {
6709 		struct net_device *dev;
6710 
6711 		for_each_netdev(net, dev) {
6712 			idev = __in6_dev_get_rtnl_net(dev);
6713 			if (idev)
6714 				addrconf_disable_policy_idev(idev, val);
6715 		}
6716 	} else {
6717 		idev = (struct inet6_dev *)ctl->extra1;
6718 		addrconf_disable_policy_idev(idev, val);
6719 	}
6720 
6721 	rtnl_net_unlock(net);
6722 	return 0;
6723 }
6724 
addrconf_sysctl_disable_policy(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)6725 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write,
6726 				   void *buffer, size_t *lenp, loff_t *ppos)
6727 {
6728 	int *valp = ctl->data;
6729 	int val = *valp;
6730 	loff_t pos = *ppos;
6731 	struct ctl_table lctl;
6732 	int ret;
6733 
6734 	lctl = *ctl;
6735 	lctl.data = &val;
6736 	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6737 
6738 	if (write && (*valp != val))
6739 		ret = addrconf_disable_policy(ctl, valp, val);
6740 
6741 	if (ret)
6742 		*ppos = pos;
6743 
6744 	return ret;
6745 }
6746 
6747 static int minus_one = -1;
6748 static const int two_five_five = 255;
6749 static u32 ioam6_if_id_max = U16_MAX;
6750 
6751 static const struct ctl_table addrconf_sysctl[] = {
6752 	{
6753 		.procname	= "forwarding",
6754 		.data		= &ipv6_devconf.forwarding,
6755 		.maxlen		= sizeof(int),
6756 		.mode		= 0644,
6757 		.proc_handler	= addrconf_sysctl_forward,
6758 	},
6759 	{
6760 		.procname	= "hop_limit",
6761 		.data		= &ipv6_devconf.hop_limit,
6762 		.maxlen		= sizeof(int),
6763 		.mode		= 0644,
6764 		.proc_handler	= proc_dointvec_minmax,
6765 		.extra1		= (void *)SYSCTL_ONE,
6766 		.extra2		= (void *)&two_five_five,
6767 	},
6768 	{
6769 		.procname	= "mtu",
6770 		.data		= &ipv6_devconf.mtu6,
6771 		.maxlen		= sizeof(int),
6772 		.mode		= 0644,
6773 		.proc_handler	= addrconf_sysctl_mtu,
6774 	},
6775 	{
6776 		.procname	= "accept_ra",
6777 		.data		= &ipv6_devconf.accept_ra,
6778 		.maxlen		= sizeof(int),
6779 		.mode		= 0644,
6780 		.proc_handler	= proc_dointvec,
6781 	},
6782 	{
6783 		.procname	= "accept_redirects",
6784 		.data		= &ipv6_devconf.accept_redirects,
6785 		.maxlen		= sizeof(int),
6786 		.mode		= 0644,
6787 		.proc_handler	= proc_dointvec,
6788 	},
6789 	{
6790 		.procname	= "autoconf",
6791 		.data		= &ipv6_devconf.autoconf,
6792 		.maxlen		= sizeof(int),
6793 		.mode		= 0644,
6794 		.proc_handler	= proc_dointvec,
6795 	},
6796 	{
6797 		.procname	= "dad_transmits",
6798 		.data		= &ipv6_devconf.dad_transmits,
6799 		.maxlen		= sizeof(int),
6800 		.mode		= 0644,
6801 		.proc_handler	= proc_dointvec,
6802 	},
6803 	{
6804 		.procname	= "router_solicitations",
6805 		.data		= &ipv6_devconf.rtr_solicits,
6806 		.maxlen		= sizeof(int),
6807 		.mode		= 0644,
6808 		.proc_handler	= proc_dointvec_minmax,
6809 		.extra1		= &minus_one,
6810 	},
6811 	{
6812 		.procname	= "router_solicitation_interval",
6813 		.data		= &ipv6_devconf.rtr_solicit_interval,
6814 		.maxlen		= sizeof(int),
6815 		.mode		= 0644,
6816 		.proc_handler	= proc_dointvec_jiffies,
6817 	},
6818 	{
6819 		.procname	= "router_solicitation_max_interval",
6820 		.data		= &ipv6_devconf.rtr_solicit_max_interval,
6821 		.maxlen		= sizeof(int),
6822 		.mode		= 0644,
6823 		.proc_handler	= proc_dointvec_jiffies,
6824 	},
6825 	{
6826 		.procname	= "router_solicitation_delay",
6827 		.data		= &ipv6_devconf.rtr_solicit_delay,
6828 		.maxlen		= sizeof(int),
6829 		.mode		= 0644,
6830 		.proc_handler	= proc_dointvec_jiffies,
6831 	},
6832 	{
6833 		.procname	= "force_mld_version",
6834 		.data		= &ipv6_devconf.force_mld_version,
6835 		.maxlen		= sizeof(int),
6836 		.mode		= 0644,
6837 		.proc_handler	= proc_dointvec,
6838 	},
6839 	{
6840 		.procname	= "mldv1_unsolicited_report_interval",
6841 		.data		=
6842 			&ipv6_devconf.mldv1_unsolicited_report_interval,
6843 		.maxlen		= sizeof(int),
6844 		.mode		= 0644,
6845 		.proc_handler	= proc_dointvec_ms_jiffies,
6846 	},
6847 	{
6848 		.procname	= "mldv2_unsolicited_report_interval",
6849 		.data		=
6850 			&ipv6_devconf.mldv2_unsolicited_report_interval,
6851 		.maxlen		= sizeof(int),
6852 		.mode		= 0644,
6853 		.proc_handler	= proc_dointvec_ms_jiffies,
6854 	},
6855 	{
6856 		.procname	= "use_tempaddr",
6857 		.data		= &ipv6_devconf.use_tempaddr,
6858 		.maxlen		= sizeof(int),
6859 		.mode		= 0644,
6860 		.proc_handler	= proc_dointvec,
6861 	},
6862 	{
6863 		.procname	= "temp_valid_lft",
6864 		.data		= &ipv6_devconf.temp_valid_lft,
6865 		.maxlen		= sizeof(int),
6866 		.mode		= 0644,
6867 		.proc_handler	= proc_dointvec,
6868 	},
6869 	{
6870 		.procname	= "temp_prefered_lft",
6871 		.data		= &ipv6_devconf.temp_prefered_lft,
6872 		.maxlen		= sizeof(int),
6873 		.mode		= 0644,
6874 		.proc_handler	= proc_dointvec,
6875 	},
6876 	{
6877 		.procname       = "regen_min_advance",
6878 		.data           = &ipv6_devconf.regen_min_advance,
6879 		.maxlen         = sizeof(int),
6880 		.mode           = 0644,
6881 		.proc_handler   = proc_dointvec,
6882 	},
6883 	{
6884 		.procname	= "regen_max_retry",
6885 		.data		= &ipv6_devconf.regen_max_retry,
6886 		.maxlen		= sizeof(int),
6887 		.mode		= 0644,
6888 		.proc_handler	= proc_dointvec,
6889 	},
6890 	{
6891 		.procname	= "max_desync_factor",
6892 		.data		= &ipv6_devconf.max_desync_factor,
6893 		.maxlen		= sizeof(int),
6894 		.mode		= 0644,
6895 		.proc_handler	= proc_dointvec,
6896 	},
6897 	{
6898 		.procname	= "max_addresses",
6899 		.data		= &ipv6_devconf.max_addresses,
6900 		.maxlen		= sizeof(int),
6901 		.mode		= 0644,
6902 		.proc_handler	= proc_dointvec,
6903 	},
6904 	{
6905 		.procname	= "accept_ra_defrtr",
6906 		.data		= &ipv6_devconf.accept_ra_defrtr,
6907 		.maxlen		= sizeof(int),
6908 		.mode		= 0644,
6909 		.proc_handler	= proc_dointvec,
6910 	},
6911 	{
6912 		.procname	= "ra_defrtr_metric",
6913 		.data		= &ipv6_devconf.ra_defrtr_metric,
6914 		.maxlen		= sizeof(u32),
6915 		.mode		= 0644,
6916 		.proc_handler	= proc_douintvec_minmax,
6917 		.extra1		= (void *)SYSCTL_ONE,
6918 	},
6919 	{
6920 		.procname	= "accept_ra_min_hop_limit",
6921 		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
6922 		.maxlen		= sizeof(int),
6923 		.mode		= 0644,
6924 		.proc_handler	= proc_dointvec,
6925 	},
6926 	{
6927 		.procname	= "accept_ra_min_lft",
6928 		.data		= &ipv6_devconf.accept_ra_min_lft,
6929 		.maxlen		= sizeof(int),
6930 		.mode		= 0644,
6931 		.proc_handler	= proc_dointvec,
6932 	},
6933 	{
6934 		.procname	= "accept_ra_pinfo",
6935 		.data		= &ipv6_devconf.accept_ra_pinfo,
6936 		.maxlen		= sizeof(int),
6937 		.mode		= 0644,
6938 		.proc_handler	= proc_dointvec,
6939 	},
6940 	{
6941 		.procname	= "ra_honor_pio_life",
6942 		.data		= &ipv6_devconf.ra_honor_pio_life,
6943 		.maxlen		= sizeof(u8),
6944 		.mode		= 0644,
6945 		.proc_handler	= proc_dou8vec_minmax,
6946 		.extra1		= SYSCTL_ZERO,
6947 		.extra2		= SYSCTL_ONE,
6948 	},
6949 	{
6950 		.procname	= "ra_honor_pio_pflag",
6951 		.data		= &ipv6_devconf.ra_honor_pio_pflag,
6952 		.maxlen		= sizeof(u8),
6953 		.mode		= 0644,
6954 		.proc_handler	= proc_dou8vec_minmax,
6955 		.extra1		= SYSCTL_ZERO,
6956 		.extra2		= SYSCTL_ONE,
6957 	},
6958 #ifdef CONFIG_IPV6_ROUTER_PREF
6959 	{
6960 		.procname	= "accept_ra_rtr_pref",
6961 		.data		= &ipv6_devconf.accept_ra_rtr_pref,
6962 		.maxlen		= sizeof(int),
6963 		.mode		= 0644,
6964 		.proc_handler	= proc_dointvec,
6965 	},
6966 	{
6967 		.procname	= "router_probe_interval",
6968 		.data		= &ipv6_devconf.rtr_probe_interval,
6969 		.maxlen		= sizeof(int),
6970 		.mode		= 0644,
6971 		.proc_handler	= proc_dointvec_jiffies,
6972 	},
6973 #ifdef CONFIG_IPV6_ROUTE_INFO
6974 	{
6975 		.procname	= "accept_ra_rt_info_min_plen",
6976 		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
6977 		.maxlen		= sizeof(int),
6978 		.mode		= 0644,
6979 		.proc_handler	= proc_dointvec,
6980 	},
6981 	{
6982 		.procname	= "accept_ra_rt_info_max_plen",
6983 		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
6984 		.maxlen		= sizeof(int),
6985 		.mode		= 0644,
6986 		.proc_handler	= proc_dointvec,
6987 	},
6988 #endif
6989 #endif
6990 	{
6991 		.procname	= "proxy_ndp",
6992 		.data		= &ipv6_devconf.proxy_ndp,
6993 		.maxlen		= sizeof(int),
6994 		.mode		= 0644,
6995 		.proc_handler	= addrconf_sysctl_proxy_ndp,
6996 	},
6997 	{
6998 		.procname	= "accept_source_route",
6999 		.data		= &ipv6_devconf.accept_source_route,
7000 		.maxlen		= sizeof(int),
7001 		.mode		= 0644,
7002 		.proc_handler	= proc_dointvec,
7003 	},
7004 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
7005 	{
7006 		.procname	= "optimistic_dad",
7007 		.data		= &ipv6_devconf.optimistic_dad,
7008 		.maxlen		= sizeof(int),
7009 		.mode		= 0644,
7010 		.proc_handler   = proc_dointvec,
7011 	},
7012 	{
7013 		.procname	= "use_optimistic",
7014 		.data		= &ipv6_devconf.use_optimistic,
7015 		.maxlen		= sizeof(int),
7016 		.mode		= 0644,
7017 		.proc_handler	= proc_dointvec,
7018 	},
7019 #endif
7020 #ifdef CONFIG_IPV6_MROUTE
7021 	{
7022 		.procname	= "mc_forwarding",
7023 		.data		= &ipv6_devconf.mc_forwarding,
7024 		.maxlen		= sizeof(int),
7025 		.mode		= 0444,
7026 		.proc_handler	= proc_dointvec,
7027 	},
7028 #endif
7029 	{
7030 		.procname	= "disable_ipv6",
7031 		.data		= &ipv6_devconf.disable_ipv6,
7032 		.maxlen		= sizeof(int),
7033 		.mode		= 0644,
7034 		.proc_handler	= addrconf_sysctl_disable,
7035 	},
7036 	{
7037 		.procname	= "accept_dad",
7038 		.data		= &ipv6_devconf.accept_dad,
7039 		.maxlen		= sizeof(int),
7040 		.mode		= 0644,
7041 		.proc_handler	= proc_dointvec,
7042 	},
7043 	{
7044 		.procname	= "force_tllao",
7045 		.data		= &ipv6_devconf.force_tllao,
7046 		.maxlen		= sizeof(int),
7047 		.mode		= 0644,
7048 		.proc_handler	= proc_dointvec
7049 	},
7050 	{
7051 		.procname	= "ndisc_notify",
7052 		.data		= &ipv6_devconf.ndisc_notify,
7053 		.maxlen		= sizeof(int),
7054 		.mode		= 0644,
7055 		.proc_handler	= proc_dointvec
7056 	},
7057 	{
7058 		.procname	= "suppress_frag_ndisc",
7059 		.data		= &ipv6_devconf.suppress_frag_ndisc,
7060 		.maxlen		= sizeof(int),
7061 		.mode		= 0644,
7062 		.proc_handler	= proc_dointvec
7063 	},
7064 	{
7065 		.procname	= "accept_ra_from_local",
7066 		.data		= &ipv6_devconf.accept_ra_from_local,
7067 		.maxlen		= sizeof(int),
7068 		.mode		= 0644,
7069 		.proc_handler	= proc_dointvec,
7070 	},
7071 	{
7072 		.procname	= "accept_ra_mtu",
7073 		.data		= &ipv6_devconf.accept_ra_mtu,
7074 		.maxlen		= sizeof(int),
7075 		.mode		= 0644,
7076 		.proc_handler	= proc_dointvec,
7077 	},
7078 	{
7079 		.procname	= "stable_secret",
7080 		.data		= &ipv6_devconf.stable_secret,
7081 		.maxlen		= IPV6_MAX_STRLEN,
7082 		.mode		= 0600,
7083 		.proc_handler	= addrconf_sysctl_stable_secret,
7084 	},
7085 	{
7086 		.procname	= "use_oif_addrs_only",
7087 		.data		= &ipv6_devconf.use_oif_addrs_only,
7088 		.maxlen		= sizeof(int),
7089 		.mode		= 0644,
7090 		.proc_handler	= proc_dointvec,
7091 	},
7092 	{
7093 		.procname	= "ignore_routes_with_linkdown",
7094 		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
7095 		.maxlen		= sizeof(int),
7096 		.mode		= 0644,
7097 		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
7098 	},
7099 	{
7100 		.procname	= "drop_unicast_in_l2_multicast",
7101 		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
7102 		.maxlen		= sizeof(int),
7103 		.mode		= 0644,
7104 		.proc_handler	= proc_dointvec,
7105 	},
7106 	{
7107 		.procname	= "drop_unsolicited_na",
7108 		.data		= &ipv6_devconf.drop_unsolicited_na,
7109 		.maxlen		= sizeof(int),
7110 		.mode		= 0644,
7111 		.proc_handler	= proc_dointvec,
7112 	},
7113 	{
7114 		.procname	= "keep_addr_on_down",
7115 		.data		= &ipv6_devconf.keep_addr_on_down,
7116 		.maxlen		= sizeof(int),
7117 		.mode		= 0644,
7118 		.proc_handler	= proc_dointvec,
7119 
7120 	},
7121 	{
7122 		.procname	= "seg6_enabled",
7123 		.data		= &ipv6_devconf.seg6_enabled,
7124 		.maxlen		= sizeof(int),
7125 		.mode		= 0644,
7126 		.proc_handler	= proc_dointvec,
7127 	},
7128 #ifdef CONFIG_IPV6_SEG6_HMAC
7129 	{
7130 		.procname	= "seg6_require_hmac",
7131 		.data		= &ipv6_devconf.seg6_require_hmac,
7132 		.maxlen		= sizeof(int),
7133 		.mode		= 0644,
7134 		.proc_handler	= proc_dointvec,
7135 	},
7136 #endif
7137 	{
7138 		.procname       = "enhanced_dad",
7139 		.data           = &ipv6_devconf.enhanced_dad,
7140 		.maxlen         = sizeof(int),
7141 		.mode           = 0644,
7142 		.proc_handler   = proc_dointvec,
7143 	},
7144 	{
7145 		.procname	= "addr_gen_mode",
7146 		.data		= &ipv6_devconf.addr_gen_mode,
7147 		.maxlen		= sizeof(int),
7148 		.mode		= 0644,
7149 		.proc_handler	= addrconf_sysctl_addr_gen_mode,
7150 	},
7151 	{
7152 		.procname       = "disable_policy",
7153 		.data           = &ipv6_devconf.disable_policy,
7154 		.maxlen         = sizeof(int),
7155 		.mode           = 0644,
7156 		.proc_handler   = addrconf_sysctl_disable_policy,
7157 	},
7158 	{
7159 		.procname	= "ndisc_tclass",
7160 		.data		= &ipv6_devconf.ndisc_tclass,
7161 		.maxlen		= sizeof(int),
7162 		.mode		= 0644,
7163 		.proc_handler	= proc_dointvec_minmax,
7164 		.extra1		= (void *)SYSCTL_ZERO,
7165 		.extra2		= (void *)&two_five_five,
7166 	},
7167 	{
7168 		.procname	= "rpl_seg_enabled",
7169 		.data		= &ipv6_devconf.rpl_seg_enabled,
7170 		.maxlen		= sizeof(int),
7171 		.mode		= 0644,
7172 		.proc_handler	= proc_dointvec,
7173 	},
7174 	{
7175 		.procname	= "ioam6_enabled",
7176 		.data		= &ipv6_devconf.ioam6_enabled,
7177 		.maxlen		= sizeof(u8),
7178 		.mode		= 0644,
7179 		.proc_handler	= proc_dou8vec_minmax,
7180 		.extra1		= (void *)SYSCTL_ZERO,
7181 		.extra2		= (void *)SYSCTL_ONE,
7182 	},
7183 	{
7184 		.procname	= "ioam6_id",
7185 		.data		= &ipv6_devconf.ioam6_id,
7186 		.maxlen		= sizeof(u32),
7187 		.mode		= 0644,
7188 		.proc_handler	= proc_douintvec_minmax,
7189 		.extra1		= (void *)SYSCTL_ZERO,
7190 		.extra2		= (void *)&ioam6_if_id_max,
7191 	},
7192 	{
7193 		.procname	= "ioam6_id_wide",
7194 		.data		= &ipv6_devconf.ioam6_id_wide,
7195 		.maxlen		= sizeof(u32),
7196 		.mode		= 0644,
7197 		.proc_handler	= proc_douintvec,
7198 	},
7199 	{
7200 		.procname	= "ndisc_evict_nocarrier",
7201 		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
7202 		.maxlen		= sizeof(u8),
7203 		.mode		= 0644,
7204 		.proc_handler	= proc_dou8vec_minmax,
7205 		.extra1		= (void *)SYSCTL_ZERO,
7206 		.extra2		= (void *)SYSCTL_ONE,
7207 	},
7208 	{
7209 		.procname	= "accept_untracked_na",
7210 		.data		= &ipv6_devconf.accept_untracked_na,
7211 		.maxlen		= sizeof(int),
7212 		.mode		= 0644,
7213 		.proc_handler	= proc_dointvec_minmax,
7214 		.extra1		= SYSCTL_ZERO,
7215 		.extra2		= SYSCTL_TWO,
7216 	},
7217 };
7218 
__addrconf_sysctl_register(struct net * net,char * dev_name,struct inet6_dev * idev,struct ipv6_devconf * p)7219 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7220 		struct inet6_dev *idev, struct ipv6_devconf *p)
7221 {
7222 	size_t table_size = ARRAY_SIZE(addrconf_sysctl);
7223 	int i, ifindex;
7224 	struct ctl_table *table;
7225 	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7226 
7227 	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7228 	if (!table)
7229 		goto out;
7230 
7231 	for (i = 0; i < table_size; i++) {
7232 		table[i].data += (char *)p - (char *)&ipv6_devconf;
7233 		/* If one of these is already set, then it is not safe to
7234 		 * overwrite either of them: this makes proc_dointvec_minmax
7235 		 * usable.
7236 		 */
7237 		if (!table[i].extra1 && !table[i].extra2) {
7238 			table[i].extra1 = idev; /* embedded; no ref */
7239 			table[i].extra2 = net;
7240 		}
7241 	}
7242 
7243 	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7244 
7245 	p->sysctl_header = register_net_sysctl_sz(net, path, table,
7246 						  table_size);
7247 	if (!p->sysctl_header)
7248 		goto free;
7249 
7250 	if (!strcmp(dev_name, "all"))
7251 		ifindex = NETCONFA_IFINDEX_ALL;
7252 	else if (!strcmp(dev_name, "default"))
7253 		ifindex = NETCONFA_IFINDEX_DEFAULT;
7254 	else
7255 		ifindex = idev->dev->ifindex;
7256 	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7257 				     ifindex, p);
7258 	return 0;
7259 
7260 free:
7261 	kfree(table);
7262 out:
7263 	return -ENOBUFS;
7264 }
7265 
__addrconf_sysctl_unregister(struct net * net,struct ipv6_devconf * p,int ifindex)7266 static void __addrconf_sysctl_unregister(struct net *net,
7267 					 struct ipv6_devconf *p, int ifindex)
7268 {
7269 	const struct ctl_table *table;
7270 
7271 	if (!p->sysctl_header)
7272 		return;
7273 
7274 	table = p->sysctl_header->ctl_table_arg;
7275 	unregister_net_sysctl_table(p->sysctl_header);
7276 	p->sysctl_header = NULL;
7277 	kfree(table);
7278 
7279 	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7280 }
7281 
addrconf_sysctl_register(struct inet6_dev * idev)7282 static int addrconf_sysctl_register(struct inet6_dev *idev)
7283 {
7284 	int err;
7285 
7286 	if (!sysctl_dev_name_is_allowed(idev->dev->name))
7287 		return -EINVAL;
7288 
7289 	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7290 				    &ndisc_ifinfo_sysctl_change);
7291 	if (err)
7292 		return err;
7293 	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7294 					 idev, &idev->cnf);
7295 	if (err)
7296 		neigh_sysctl_unregister(idev->nd_parms);
7297 
7298 	return err;
7299 }
7300 
addrconf_sysctl_unregister(struct inet6_dev * idev)7301 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7302 {
7303 	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7304 				     idev->dev->ifindex);
7305 	neigh_sysctl_unregister(idev->nd_parms);
7306 }
7307 
7308 
7309 #endif
7310 
addrconf_init_net(struct net * net)7311 static int __net_init addrconf_init_net(struct net *net)
7312 {
7313 	int err = -ENOMEM;
7314 	struct ipv6_devconf *all, *dflt;
7315 
7316 	spin_lock_init(&net->ipv6.addrconf_hash_lock);
7317 	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7318 	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7319 					   sizeof(struct hlist_head),
7320 					   GFP_KERNEL);
7321 	if (!net->ipv6.inet6_addr_lst)
7322 		goto err_alloc_addr;
7323 
7324 	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7325 	if (!all)
7326 		goto err_alloc_all;
7327 
7328 	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7329 	if (!dflt)
7330 		goto err_alloc_dflt;
7331 
7332 	if (!net_eq(net, &init_net)) {
7333 		switch (net_inherit_devconf()) {
7334 		case 1:  /* copy from init_net */
7335 			memcpy(all, init_net.ipv6.devconf_all,
7336 			       sizeof(ipv6_devconf));
7337 			memcpy(dflt, init_net.ipv6.devconf_dflt,
7338 			       sizeof(ipv6_devconf_dflt));
7339 			break;
7340 		case 3: /* copy from the current netns */
7341 			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7342 			       sizeof(ipv6_devconf));
7343 			memcpy(dflt,
7344 			       current->nsproxy->net_ns->ipv6.devconf_dflt,
7345 			       sizeof(ipv6_devconf_dflt));
7346 			break;
7347 		case 0:
7348 		case 2:
7349 			/* use compiled values */
7350 			break;
7351 		}
7352 	}
7353 
7354 	/* these will be inherited by all namespaces */
7355 	dflt->autoconf = ipv6_defaults.autoconf;
7356 	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7357 
7358 	dflt->stable_secret.initialized = false;
7359 	all->stable_secret.initialized = false;
7360 
7361 	net->ipv6.devconf_all = all;
7362 	net->ipv6.devconf_dflt = dflt;
7363 
7364 #ifdef CONFIG_SYSCTL
7365 	err = __addrconf_sysctl_register(net, "all", NULL, all);
7366 	if (err < 0)
7367 		goto err_reg_all;
7368 
7369 	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7370 	if (err < 0)
7371 		goto err_reg_dflt;
7372 #endif
7373 	return 0;
7374 
7375 #ifdef CONFIG_SYSCTL
7376 err_reg_dflt:
7377 	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7378 err_reg_all:
7379 	kfree(dflt);
7380 	net->ipv6.devconf_dflt = NULL;
7381 #endif
7382 err_alloc_dflt:
7383 	kfree(all);
7384 	net->ipv6.devconf_all = NULL;
7385 err_alloc_all:
7386 	kfree(net->ipv6.inet6_addr_lst);
7387 err_alloc_addr:
7388 	return err;
7389 }
7390 
addrconf_exit_net(struct net * net)7391 static void __net_exit addrconf_exit_net(struct net *net)
7392 {
7393 	int i;
7394 
7395 #ifdef CONFIG_SYSCTL
7396 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7397 				     NETCONFA_IFINDEX_DEFAULT);
7398 	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7399 				     NETCONFA_IFINDEX_ALL);
7400 #endif
7401 	kfree(net->ipv6.devconf_dflt);
7402 	net->ipv6.devconf_dflt = NULL;
7403 	kfree(net->ipv6.devconf_all);
7404 	net->ipv6.devconf_all = NULL;
7405 
7406 	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7407 	/*
7408 	 *	Check hash table, then free it.
7409 	 */
7410 	for (i = 0; i < IN6_ADDR_HSIZE; i++)
7411 		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7412 
7413 	kfree(net->ipv6.inet6_addr_lst);
7414 	net->ipv6.inet6_addr_lst = NULL;
7415 }
7416 
7417 static struct pernet_operations addrconf_ops = {
7418 	.init = addrconf_init_net,
7419 	.exit = addrconf_exit_net,
7420 };
7421 
7422 static struct rtnl_af_ops inet6_ops __read_mostly = {
7423 	.family		  = AF_INET6,
7424 	.fill_link_af	  = inet6_fill_link_af,
7425 	.get_link_af_size = inet6_get_link_af_size,
7426 	.validate_link_af = inet6_validate_link_af,
7427 	.set_link_af	  = inet6_set_link_af,
7428 };
7429 
7430 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = {
7431 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK,
7432 	 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED},
7433 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR,
7434 	 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET},
7435 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR,
7436 	 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET},
7437 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR,
7438 	 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr,
7439 	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7440 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST,
7441 	 .dumpit = inet6_dump_ifmcaddr,
7442 	 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7443 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST,
7444 	 .dumpit = inet6_dump_ifacaddr,
7445 	 .flags = RTNL_FLAG_DUMP_UNLOCKED},
7446 	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF,
7447 	 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf,
7448 	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
7449 };
7450 
7451 /*
7452  *	Init / cleanup code
7453  */
7454 
addrconf_init(void)7455 int __init addrconf_init(void)
7456 {
7457 	struct inet6_dev *idev;
7458 	int err;
7459 
7460 	err = ipv6_addr_label_init();
7461 	if (err < 0) {
7462 		pr_crit("%s: cannot initialize default policy table: %d\n",
7463 			__func__, err);
7464 		goto out;
7465 	}
7466 
7467 	err = register_pernet_subsys(&addrconf_ops);
7468 	if (err < 0)
7469 		goto out_addrlabel;
7470 
7471 	/* All works using addrconf_wq need to lock rtnl. */
7472 	addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7473 	if (!addrconf_wq) {
7474 		err = -ENOMEM;
7475 		goto out_nowq;
7476 	}
7477 
7478 	rtnl_net_lock(&init_net);
7479 	idev = ipv6_add_dev(blackhole_netdev);
7480 	rtnl_net_unlock(&init_net);
7481 	if (IS_ERR(idev)) {
7482 		err = PTR_ERR(idev);
7483 		goto errlo;
7484 	}
7485 
7486 	ip6_route_init_special_entries();
7487 
7488 	register_netdevice_notifier(&ipv6_dev_notf);
7489 
7490 	addrconf_verify(&init_net);
7491 
7492 	err = rtnl_af_register(&inet6_ops);
7493 	if (err)
7494 		goto erraf;
7495 
7496 	err = rtnl_register_many(addrconf_rtnl_msg_handlers);
7497 	if (err)
7498 		goto errout;
7499 
7500 	err = ipv6_addr_label_rtnl_register();
7501 	if (err < 0)
7502 		goto errout;
7503 
7504 	return 0;
7505 errout:
7506 	rtnl_unregister_all(PF_INET6);
7507 	rtnl_af_unregister(&inet6_ops);
7508 erraf:
7509 	unregister_netdevice_notifier(&ipv6_dev_notf);
7510 errlo:
7511 	destroy_workqueue(addrconf_wq);
7512 out_nowq:
7513 	unregister_pernet_subsys(&addrconf_ops);
7514 out_addrlabel:
7515 	ipv6_addr_label_cleanup();
7516 out:
7517 	return err;
7518 }
7519 
addrconf_cleanup(void)7520 void addrconf_cleanup(void)
7521 {
7522 	struct net_device *dev;
7523 
7524 	unregister_netdevice_notifier(&ipv6_dev_notf);
7525 	unregister_pernet_subsys(&addrconf_ops);
7526 	ipv6_addr_label_cleanup();
7527 
7528 	rtnl_af_unregister(&inet6_ops);
7529 
7530 	rtnl_net_lock(&init_net);
7531 
7532 	/* clean dev list */
7533 	for_each_netdev(&init_net, dev) {
7534 		if (!__in6_dev_get_rtnl_net(dev))
7535 			continue;
7536 		addrconf_ifdown(dev, true);
7537 	}
7538 	addrconf_ifdown(init_net.loopback_dev, true);
7539 
7540 	rtnl_net_unlock(&init_net);
7541 
7542 	destroy_workqueue(addrconf_wq);
7543 }
7544