1/*
2 *	IPv6 Address [auto]configuration
3 *	Linux INET6 implementation
4 *
5 *	Authors:
6 *	Pedro Roque		<roque@di.fc.ul.pt>
7 *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8 *
9 *	This program is free software; you can redistribute it and/or
10 *      modify it under the terms of the GNU General Public License
11 *      as published by the Free Software Foundation; either version
12 *      2 of the License, or (at your option) any later version.
13 */
14
15/*
16 *	Changes:
17 *
18 *	Janos Farkas			:	delete timer on ifdown
19 *	<chexum@bankinf.banki.hu>
20 *	Andi Kleen			:	kill double kfree on module
21 *						unload.
22 *	Maciej W. Rozycki		:	FDDI support
23 *	sekiya@USAGI			:	Don't send too many RS
24 *						packets.
25 *	yoshfuji@USAGI			:       Fixed interval between DAD
26 *						packets.
27 *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
28 *						address validation timer.
29 *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
30 *						support.
31 *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
32 *						address on a same interface.
33 *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
34 *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
35 *						seq_file.
36 *	YOSHIFUJI Hideaki @USAGI	:	improved source address
37 *						selection; consider scope,
38 *						status etc.
39 */
40
41#define pr_fmt(fmt) "IPv6: " fmt
42
43#include <linux/errno.h>
44#include <linux/types.h>
45#include <linux/kernel.h>
46#include <linux/socket.h>
47#include <linux/sockios.h>
48#include <linux/net.h>
49#include <linux/inet.h>
50#include <linux/in6.h>
51#include <linux/netdevice.h>
52#include <linux/if_addr.h>
53#include <linux/if_arp.h>
54#include <linux/if_arcnet.h>
55#include <linux/if_infiniband.h>
56#include <linux/route.h>
57#include <linux/inetdevice.h>
58#include <linux/init.h>
59#include <linux/slab.h>
60#ifdef CONFIG_SYSCTL
61#include <linux/sysctl.h>
62#endif
63#include <linux/capability.h>
64#include <linux/delay.h>
65#include <linux/notifier.h>
66#include <linux/string.h>
67#include <linux/hash.h>
68
69#include <net/net_namespace.h>
70#include <net/sock.h>
71#include <net/snmp.h>
72
73#include <net/af_ieee802154.h>
74#include <net/firewire.h>
75#include <net/ipv6.h>
76#include <net/protocol.h>
77#include <net/ndisc.h>
78#include <net/ip6_route.h>
79#include <net/addrconf.h>
80#include <net/tcp.h>
81#include <net/ip.h>
82#include <net/netlink.h>
83#include <net/pkt_sched.h>
84#include <net/l3mdev.h>
85#include <linux/if_tunnel.h>
86#include <linux/rtnetlink.h>
87#include <linux/netconf.h>
88#include <linux/random.h>
89#include <linux/uaccess.h>
90#include <asm/unaligned.h>
91
92#include <linux/proc_fs.h>
93#include <linux/seq_file.h>
94#include <linux/export.h>
95
96/* Set to 3 to get tracing... */
97#define ACONF_DEBUG 2
98
99#if ACONF_DEBUG >= 3
100#define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101#else
102#define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103#endif
104
105#define	INFINITY_LIFE_TIME	0xFFFFFFFF
106
107#define IPV6_MAX_STRLEN \
108	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
109
110static inline u32 cstamp_delta(unsigned long cstamp)
111{
112	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
113}
114
115#ifdef CONFIG_SYSCTL
116static int addrconf_sysctl_register(struct inet6_dev *idev);
117static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118#else
119static inline int addrconf_sysctl_register(struct inet6_dev *idev)
120{
121	return 0;
122}
123
124static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
125{
126}
127#endif
128
129static void __ipv6_regen_rndid(struct inet6_dev *idev);
130static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131static void ipv6_regen_rndid(unsigned long data);
132
133static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134static int ipv6_count_addresses(struct inet6_dev *idev);
135static int ipv6_generate_stable_address(struct in6_addr *addr,
136					u8 dad_count,
137					const struct inet6_dev *idev);
138
139/*
140 *	Configured unicast address hash table
141 */
142static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143static DEFINE_SPINLOCK(addrconf_hash_lock);
144
145static void addrconf_verify(void);
146static void addrconf_verify_rtnl(void);
147static void addrconf_verify_work(struct work_struct *);
148
149static struct workqueue_struct *addrconf_wq;
150static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
151
152static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
154
155static void addrconf_type_change(struct net_device *dev,
156				 unsigned long event);
157static int addrconf_ifdown(struct net_device *dev, int how);
158
159static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
160						  int plen,
161						  const struct net_device *dev,
162						  u32 flags, u32 noflags);
163
164static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165static void addrconf_dad_work(struct work_struct *w);
166static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167static void addrconf_dad_run(struct inet6_dev *idev);
168static void addrconf_rs_timer(unsigned long data);
169static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
171
172static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173				struct prefix_info *pinfo);
174static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175			       struct net_device *dev);
176
177static struct ipv6_devconf ipv6_devconf __read_mostly = {
178	.forwarding		= 0,
179	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
180	.mtu6			= IPV6_MIN_MTU,
181	.accept_ra		= 1,
182	.accept_redirects	= 1,
183	.autoconf		= 1,
184	.force_mld_version	= 0,
185	.mldv1_unsolicited_report_interval = 10 * HZ,
186	.mldv2_unsolicited_report_interval = HZ,
187	.dad_transmits		= 1,
188	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
189	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
190	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
191	.use_tempaddr		= 0,
192	.temp_valid_lft		= TEMP_VALID_LIFETIME,
193	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
194	.regen_max_retry	= REGEN_MAX_RETRY,
195	.max_desync_factor	= MAX_DESYNC_FACTOR,
196	.max_addresses		= IPV6_MAX_ADDRESSES,
197	.accept_ra_defrtr	= 1,
198	.accept_ra_from_local	= 0,
199	.accept_ra_min_hop_limit= 1,
200	.accept_ra_pinfo	= 1,
201#ifdef CONFIG_IPV6_ROUTER_PREF
202	.accept_ra_rtr_pref	= 1,
203	.rtr_probe_interval	= 60 * HZ,
204#ifdef CONFIG_IPV6_ROUTE_INFO
205	.accept_ra_rt_info_max_plen = 0,
206#endif
207#endif
208	.proxy_ndp		= 0,
209	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
210	.disable_ipv6		= 0,
211	.accept_dad		= 1,
212	.suppress_frag_ndisc	= 1,
213	.accept_ra_mtu		= 1,
214	.stable_secret		= {
215		.initialized = false,
216	},
217	.use_oif_addrs_only	= 0,
218	.ignore_routes_with_linkdown = 0,
219};
220
221static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
222	.forwarding		= 0,
223	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
224	.mtu6			= IPV6_MIN_MTU,
225	.accept_ra		= 1,
226	.accept_redirects	= 1,
227	.autoconf		= 1,
228	.force_mld_version	= 0,
229	.mldv1_unsolicited_report_interval = 10 * HZ,
230	.mldv2_unsolicited_report_interval = HZ,
231	.dad_transmits		= 1,
232	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
233	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
234	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
235	.use_tempaddr		= 0,
236	.temp_valid_lft		= TEMP_VALID_LIFETIME,
237	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
238	.regen_max_retry	= REGEN_MAX_RETRY,
239	.max_desync_factor	= MAX_DESYNC_FACTOR,
240	.max_addresses		= IPV6_MAX_ADDRESSES,
241	.accept_ra_defrtr	= 1,
242	.accept_ra_from_local	= 0,
243	.accept_ra_min_hop_limit= 1,
244	.accept_ra_pinfo	= 1,
245#ifdef CONFIG_IPV6_ROUTER_PREF
246	.accept_ra_rtr_pref	= 1,
247	.rtr_probe_interval	= 60 * HZ,
248#ifdef CONFIG_IPV6_ROUTE_INFO
249	.accept_ra_rt_info_max_plen = 0,
250#endif
251#endif
252	.proxy_ndp		= 0,
253	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
254	.disable_ipv6		= 0,
255	.accept_dad		= 1,
256	.suppress_frag_ndisc	= 1,
257	.accept_ra_mtu		= 1,
258	.stable_secret		= {
259		.initialized = false,
260	},
261	.use_oif_addrs_only	= 0,
262	.ignore_routes_with_linkdown = 0,
263};
264
265/* Check if a valid qdisc is available */
266static inline bool addrconf_qdisc_ok(const struct net_device *dev)
267{
268	return !qdisc_tx_is_noop(dev);
269}
270
271static void addrconf_del_rs_timer(struct inet6_dev *idev)
272{
273	if (del_timer(&idev->rs_timer))
274		__in6_dev_put(idev);
275}
276
277static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
278{
279	if (cancel_delayed_work(&ifp->dad_work))
280		__in6_ifa_put(ifp);
281}
282
283static void addrconf_mod_rs_timer(struct inet6_dev *idev,
284				  unsigned long when)
285{
286	if (!timer_pending(&idev->rs_timer))
287		in6_dev_hold(idev);
288	mod_timer(&idev->rs_timer, jiffies + when);
289}
290
291static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
292				   unsigned long delay)
293{
294	if (!delayed_work_pending(&ifp->dad_work))
295		in6_ifa_hold(ifp);
296	mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
297}
298
299static int snmp6_alloc_dev(struct inet6_dev *idev)
300{
301	int i;
302
303	idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
304	if (!idev->stats.ipv6)
305		goto err_ip;
306
307	for_each_possible_cpu(i) {
308		struct ipstats_mib *addrconf_stats;
309		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
310		u64_stats_init(&addrconf_stats->syncp);
311	}
312
313
314	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
315					GFP_KERNEL);
316	if (!idev->stats.icmpv6dev)
317		goto err_icmp;
318	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
319					   GFP_KERNEL);
320	if (!idev->stats.icmpv6msgdev)
321		goto err_icmpmsg;
322
323	return 0;
324
325err_icmpmsg:
326	kfree(idev->stats.icmpv6dev);
327err_icmp:
328	free_percpu(idev->stats.ipv6);
329err_ip:
330	return -ENOMEM;
331}
332
333static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
334{
335	struct inet6_dev *ndev;
336	int err = -ENOMEM;
337
338	ASSERT_RTNL();
339
340	if (dev->mtu < IPV6_MIN_MTU)
341		return ERR_PTR(-EINVAL);
342
343	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
344	if (!ndev)
345		return ERR_PTR(err);
346
347	rwlock_init(&ndev->lock);
348	ndev->dev = dev;
349	INIT_LIST_HEAD(&ndev->addr_list);
350	setup_timer(&ndev->rs_timer, addrconf_rs_timer,
351		    (unsigned long)ndev);
352	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
353
354	if (ndev->cnf.stable_secret.initialized)
355		ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
356	else
357		ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
358
359	ndev->cnf.mtu6 = dev->mtu;
360	ndev->cnf.sysctl = NULL;
361	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
362	if (!ndev->nd_parms) {
363		kfree(ndev);
364		return ERR_PTR(err);
365	}
366	if (ndev->cnf.forwarding)
367		dev_disable_lro(dev);
368	/* We refer to the device */
369	dev_hold(dev);
370
371	if (snmp6_alloc_dev(ndev) < 0) {
372		ADBG(KERN_WARNING
373			"%s: cannot allocate memory for statistics; dev=%s.\n",
374			__func__, dev->name);
375		neigh_parms_release(&nd_tbl, ndev->nd_parms);
376		dev_put(dev);
377		kfree(ndev);
378		return ERR_PTR(err);
379	}
380
381	if (snmp6_register_dev(ndev) < 0) {
382		ADBG(KERN_WARNING
383			"%s: cannot create /proc/net/dev_snmp6/%s\n",
384			__func__, dev->name);
385		goto err_release;
386	}
387
388	/* One reference from device.  We must do this before
389	 * we invoke __ipv6_regen_rndid().
390	 */
391	in6_dev_hold(ndev);
392
393	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
394		ndev->cnf.accept_dad = -1;
395
396#if IS_ENABLED(CONFIG_IPV6_SIT)
397	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
398		pr_info("%s: Disabled Multicast RS\n", dev->name);
399		ndev->cnf.rtr_solicits = 0;
400	}
401#endif
402
403	INIT_LIST_HEAD(&ndev->tempaddr_list);
404	setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
405	if ((dev->flags&IFF_LOOPBACK) ||
406	    dev->type == ARPHRD_TUNNEL ||
407	    dev->type == ARPHRD_TUNNEL6 ||
408	    dev->type == ARPHRD_SIT ||
409	    dev->type == ARPHRD_NONE) {
410		ndev->cnf.use_tempaddr = -1;
411	} else {
412		in6_dev_hold(ndev);
413		ipv6_regen_rndid((unsigned long) ndev);
414	}
415
416	ndev->token = in6addr_any;
417
418	if (netif_running(dev) && addrconf_qdisc_ok(dev))
419		ndev->if_flags |= IF_READY;
420
421	ipv6_mc_init_dev(ndev);
422	ndev->tstamp = jiffies;
423	err = addrconf_sysctl_register(ndev);
424	if (err) {
425		ipv6_mc_destroy_dev(ndev);
426		del_timer(&ndev->regen_timer);
427		snmp6_unregister_dev(ndev);
428		goto err_release;
429	}
430	/* protected by rtnl_lock */
431	rcu_assign_pointer(dev->ip6_ptr, ndev);
432
433	/* Join interface-local all-node multicast group */
434	ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
435
436	/* Join all-node multicast group */
437	ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
438
439	/* Join all-router multicast group if forwarding is set */
440	if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442
443	return ndev;
444
445err_release:
446	neigh_parms_release(&nd_tbl, ndev->nd_parms);
447	ndev->dead = 1;
448	in6_dev_finish_destroy(ndev);
449	return ERR_PTR(err);
450}
451
452static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
453{
454	struct inet6_dev *idev;
455
456	ASSERT_RTNL();
457
458	idev = __in6_dev_get(dev);
459	if (!idev) {
460		idev = ipv6_add_dev(dev);
461		if (IS_ERR(idev))
462			return NULL;
463	}
464
465	if (dev->flags&IFF_UP)
466		ipv6_mc_up(idev);
467	return idev;
468}
469
470static int inet6_netconf_msgsize_devconf(int type)
471{
472	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
473		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
474
475	/* type -1 is used for ALL */
476	if (type == -1 || type == NETCONFA_FORWARDING)
477		size += nla_total_size(4);
478#ifdef CONFIG_IPV6_MROUTE
479	if (type == -1 || type == NETCONFA_MC_FORWARDING)
480		size += nla_total_size(4);
481#endif
482	if (type == -1 || type == NETCONFA_PROXY_NEIGH)
483		size += nla_total_size(4);
484
485	if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
486		size += nla_total_size(4);
487
488	return size;
489}
490
491static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
492				      struct ipv6_devconf *devconf, u32 portid,
493				      u32 seq, int event, unsigned int flags,
494				      int type)
495{
496	struct nlmsghdr  *nlh;
497	struct netconfmsg *ncm;
498
499	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
500			flags);
501	if (!nlh)
502		return -EMSGSIZE;
503
504	ncm = nlmsg_data(nlh);
505	ncm->ncm_family = AF_INET6;
506
507	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
508		goto nla_put_failure;
509
510	/* type -1 is used for ALL */
511	if ((type == -1 || type == NETCONFA_FORWARDING) &&
512	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
513		goto nla_put_failure;
514#ifdef CONFIG_IPV6_MROUTE
515	if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
516	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
517			devconf->mc_forwarding) < 0)
518		goto nla_put_failure;
519#endif
520	if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
521	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
522		goto nla_put_failure;
523
524	if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
525	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
526			devconf->ignore_routes_with_linkdown) < 0)
527		goto nla_put_failure;
528
529	nlmsg_end(skb, nlh);
530	return 0;
531
532nla_put_failure:
533	nlmsg_cancel(skb, nlh);
534	return -EMSGSIZE;
535}
536
537void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
538				  struct ipv6_devconf *devconf)
539{
540	struct sk_buff *skb;
541	int err = -ENOBUFS;
542
543	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
544	if (!skb)
545		goto errout;
546
547	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
548					 RTM_NEWNETCONF, 0, type);
549	if (err < 0) {
550		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
551		WARN_ON(err == -EMSGSIZE);
552		kfree_skb(skb);
553		goto errout;
554	}
555	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
556	return;
557errout:
558	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
559}
560
561static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
562	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
563	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
564	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
565	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
566};
567
568static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
569				     struct nlmsghdr *nlh)
570{
571	struct net *net = sock_net(in_skb->sk);
572	struct nlattr *tb[NETCONFA_MAX+1];
573	struct netconfmsg *ncm;
574	struct sk_buff *skb;
575	struct ipv6_devconf *devconf;
576	struct inet6_dev *in6_dev;
577	struct net_device *dev;
578	int ifindex;
579	int err;
580
581	err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
582			  devconf_ipv6_policy);
583	if (err < 0)
584		goto errout;
585
586	err = -EINVAL;
587	if (!tb[NETCONFA_IFINDEX])
588		goto errout;
589
590	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
591	switch (ifindex) {
592	case NETCONFA_IFINDEX_ALL:
593		devconf = net->ipv6.devconf_all;
594		break;
595	case NETCONFA_IFINDEX_DEFAULT:
596		devconf = net->ipv6.devconf_dflt;
597		break;
598	default:
599		dev = __dev_get_by_index(net, ifindex);
600		if (!dev)
601			goto errout;
602		in6_dev = __in6_dev_get(dev);
603		if (!in6_dev)
604			goto errout;
605		devconf = &in6_dev->cnf;
606		break;
607	}
608
609	err = -ENOBUFS;
610	skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
611	if (!skb)
612		goto errout;
613
614	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
615					 NETLINK_CB(in_skb).portid,
616					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
617					 -1);
618	if (err < 0) {
619		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
620		WARN_ON(err == -EMSGSIZE);
621		kfree_skb(skb);
622		goto errout;
623	}
624	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
625errout:
626	return err;
627}
628
629static int inet6_netconf_dump_devconf(struct sk_buff *skb,
630				      struct netlink_callback *cb)
631{
632	struct net *net = sock_net(skb->sk);
633	int h, s_h;
634	int idx, s_idx;
635	struct net_device *dev;
636	struct inet6_dev *idev;
637	struct hlist_head *head;
638
639	s_h = cb->args[0];
640	s_idx = idx = cb->args[1];
641
642	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
643		idx = 0;
644		head = &net->dev_index_head[h];
645		rcu_read_lock();
646		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
647			  net->dev_base_seq;
648		hlist_for_each_entry_rcu(dev, head, index_hlist) {
649			if (idx < s_idx)
650				goto cont;
651			idev = __in6_dev_get(dev);
652			if (!idev)
653				goto cont;
654
655			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
656						       &idev->cnf,
657						       NETLINK_CB(cb->skb).portid,
658						       cb->nlh->nlmsg_seq,
659						       RTM_NEWNETCONF,
660						       NLM_F_MULTI,
661						       -1) < 0) {
662				rcu_read_unlock();
663				goto done;
664			}
665			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
666cont:
667			idx++;
668		}
669		rcu_read_unlock();
670	}
671	if (h == NETDEV_HASHENTRIES) {
672		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
673					       net->ipv6.devconf_all,
674					       NETLINK_CB(cb->skb).portid,
675					       cb->nlh->nlmsg_seq,
676					       RTM_NEWNETCONF, NLM_F_MULTI,
677					       -1) < 0)
678			goto done;
679		else
680			h++;
681	}
682	if (h == NETDEV_HASHENTRIES + 1) {
683		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
684					       net->ipv6.devconf_dflt,
685					       NETLINK_CB(cb->skb).portid,
686					       cb->nlh->nlmsg_seq,
687					       RTM_NEWNETCONF, NLM_F_MULTI,
688					       -1) < 0)
689			goto done;
690		else
691			h++;
692	}
693done:
694	cb->args[0] = h;
695	cb->args[1] = idx;
696
697	return skb->len;
698}
699
700#ifdef CONFIG_SYSCTL
701static void dev_forward_change(struct inet6_dev *idev)
702{
703	struct net_device *dev;
704	struct inet6_ifaddr *ifa;
705
706	if (!idev)
707		return;
708	dev = idev->dev;
709	if (idev->cnf.forwarding)
710		dev_disable_lro(dev);
711	if (dev->flags & IFF_MULTICAST) {
712		if (idev->cnf.forwarding) {
713			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
714			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
715			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
716		} else {
717			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
718			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
719			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
720		}
721	}
722
723	list_for_each_entry(ifa, &idev->addr_list, if_list) {
724		if (ifa->flags&IFA_F_TENTATIVE)
725			continue;
726		if (idev->cnf.forwarding)
727			addrconf_join_anycast(ifa);
728		else
729			addrconf_leave_anycast(ifa);
730	}
731	inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
732				     dev->ifindex, &idev->cnf);
733}
734
735
736static void addrconf_forward_change(struct net *net, __s32 newf)
737{
738	struct net_device *dev;
739	struct inet6_dev *idev;
740
741	for_each_netdev(net, dev) {
742		idev = __in6_dev_get(dev);
743		if (idev) {
744			int changed = (!idev->cnf.forwarding) ^ (!newf);
745			idev->cnf.forwarding = newf;
746			if (changed)
747				dev_forward_change(idev);
748		}
749	}
750}
751
752static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
753{
754	struct net *net;
755	int old;
756
757	if (!rtnl_trylock())
758		return restart_syscall();
759
760	net = (struct net *)table->extra2;
761	old = *p;
762	*p = newf;
763
764	if (p == &net->ipv6.devconf_dflt->forwarding) {
765		if ((!newf) ^ (!old))
766			inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
767						     NETCONFA_IFINDEX_DEFAULT,
768						     net->ipv6.devconf_dflt);
769		rtnl_unlock();
770		return 0;
771	}
772
773	if (p == &net->ipv6.devconf_all->forwarding) {
774		net->ipv6.devconf_dflt->forwarding = newf;
775		addrconf_forward_change(net, newf);
776		if ((!newf) ^ (!old))
777			inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
778						     NETCONFA_IFINDEX_ALL,
779						     net->ipv6.devconf_all);
780	} else if ((!newf) ^ (!old))
781		dev_forward_change((struct inet6_dev *)table->extra1);
782	rtnl_unlock();
783
784	if (newf)
785		rt6_purge_dflt_routers(net);
786	return 1;
787}
788
789static void addrconf_linkdown_change(struct net *net, __s32 newf)
790{
791	struct net_device *dev;
792	struct inet6_dev *idev;
793
794	for_each_netdev(net, dev) {
795		idev = __in6_dev_get(dev);
796		if (idev) {
797			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
798
799			idev->cnf.ignore_routes_with_linkdown = newf;
800			if (changed)
801				inet6_netconf_notify_devconf(dev_net(dev),
802							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
803							     dev->ifindex,
804							     &idev->cnf);
805		}
806	}
807}
808
809static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
810{
811	struct net *net;
812	int old;
813
814	if (!rtnl_trylock())
815		return restart_syscall();
816
817	net = (struct net *)table->extra2;
818	old = *p;
819	*p = newf;
820
821	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
822		if ((!newf) ^ (!old))
823			inet6_netconf_notify_devconf(net,
824						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
825						     NETCONFA_IFINDEX_DEFAULT,
826						     net->ipv6.devconf_dflt);
827		rtnl_unlock();
828		return 0;
829	}
830
831	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
832		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
833		addrconf_linkdown_change(net, newf);
834		if ((!newf) ^ (!old))
835			inet6_netconf_notify_devconf(net,
836						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
837						     NETCONFA_IFINDEX_ALL,
838						     net->ipv6.devconf_all);
839	}
840	rtnl_unlock();
841
842	return 1;
843}
844
845#endif
846
847/* Nobody refers to this ifaddr, destroy it */
848void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
849{
850	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
851
852#ifdef NET_REFCNT_DEBUG
853	pr_debug("%s\n", __func__);
854#endif
855
856	in6_dev_put(ifp->idev);
857
858	if (cancel_delayed_work(&ifp->dad_work))
859		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
860			  ifp);
861
862	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
863		pr_warn("Freeing alive inet6 address %p\n", ifp);
864		return;
865	}
866	ip6_rt_put(ifp->rt);
867
868	kfree_rcu(ifp, rcu);
869}
870
871static void
872ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
873{
874	struct list_head *p;
875	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
876
877	/*
878	 * Each device address list is sorted in order of scope -
879	 * global before linklocal.
880	 */
881	list_for_each(p, &idev->addr_list) {
882		struct inet6_ifaddr *ifa
883			= list_entry(p, struct inet6_ifaddr, if_list);
884		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
885			break;
886	}
887
888	list_add_tail(&ifp->if_list, p);
889}
890
891static u32 inet6_addr_hash(const struct in6_addr *addr)
892{
893	return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
894}
895
896/* On success it returns ifp with increased reference count */
897
898static struct inet6_ifaddr *
899ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
900	      const struct in6_addr *peer_addr, int pfxlen,
901	      int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
902{
903	struct inet6_ifaddr *ifa = NULL;
904	struct rt6_info *rt;
905	unsigned int hash;
906	int err = 0;
907	int addr_type = ipv6_addr_type(addr);
908
909	if (addr_type == IPV6_ADDR_ANY ||
910	    addr_type & IPV6_ADDR_MULTICAST ||
911	    (!(idev->dev->flags & IFF_LOOPBACK) &&
912	     addr_type & IPV6_ADDR_LOOPBACK))
913		return ERR_PTR(-EADDRNOTAVAIL);
914
915	rcu_read_lock_bh();
916	if (idev->dead) {
917		err = -ENODEV;			/*XXX*/
918		goto out2;
919	}
920
921	if (idev->cnf.disable_ipv6) {
922		err = -EACCES;
923		goto out2;
924	}
925
926	spin_lock(&addrconf_hash_lock);
927
928	/* Ignore adding duplicate addresses on an interface */
929	if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
930		ADBG("ipv6_add_addr: already assigned\n");
931		err = -EEXIST;
932		goto out;
933	}
934
935	ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
936
937	if (!ifa) {
938		ADBG("ipv6_add_addr: malloc failed\n");
939		err = -ENOBUFS;
940		goto out;
941	}
942
943	rt = addrconf_dst_alloc(idev, addr, false);
944	if (IS_ERR(rt)) {
945		err = PTR_ERR(rt);
946		goto out;
947	}
948
949	neigh_parms_data_state_setall(idev->nd_parms);
950
951	ifa->addr = *addr;
952	if (peer_addr)
953		ifa->peer_addr = *peer_addr;
954
955	spin_lock_init(&ifa->lock);
956	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
957	INIT_HLIST_NODE(&ifa->addr_lst);
958	ifa->scope = scope;
959	ifa->prefix_len = pfxlen;
960	ifa->flags = flags | IFA_F_TENTATIVE;
961	ifa->valid_lft = valid_lft;
962	ifa->prefered_lft = prefered_lft;
963	ifa->cstamp = ifa->tstamp = jiffies;
964	ifa->tokenized = false;
965
966	ifa->rt = rt;
967
968	ifa->idev = idev;
969	in6_dev_hold(idev);
970	/* For caller */
971	in6_ifa_hold(ifa);
972
973	/* Add to big hash table */
974	hash = inet6_addr_hash(addr);
975
976	hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
977	spin_unlock(&addrconf_hash_lock);
978
979	write_lock(&idev->lock);
980	/* Add to inet6_dev unicast addr list. */
981	ipv6_link_dev_addr(idev, ifa);
982
983	if (ifa->flags&IFA_F_TEMPORARY) {
984		list_add(&ifa->tmp_list, &idev->tempaddr_list);
985		in6_ifa_hold(ifa);
986	}
987
988	in6_ifa_hold(ifa);
989	write_unlock(&idev->lock);
990out2:
991	rcu_read_unlock_bh();
992
993	if (likely(err == 0))
994		inet6addr_notifier_call_chain(NETDEV_UP, ifa);
995	else {
996		kfree(ifa);
997		ifa = ERR_PTR(err);
998	}
999
1000	return ifa;
1001out:
1002	spin_unlock(&addrconf_hash_lock);
1003	goto out2;
1004}
1005
1006enum cleanup_prefix_rt_t {
1007	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1008	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1009	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1010};
1011
1012/*
1013 * Check, whether the prefix for ifp would still need a prefix route
1014 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1015 * constants.
1016 *
1017 * 1) we don't purge prefix if address was not permanent.
1018 *    prefix is managed by its own lifetime.
1019 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1020 * 3) if there are no addresses, delete prefix.
1021 * 4) if there are still other permanent address(es),
1022 *    corresponding prefix is still permanent.
1023 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1024 *    don't purge the prefix, assume user space is managing it.
1025 * 6) otherwise, update prefix lifetime to the
1026 *    longest valid lifetime among the corresponding
1027 *    addresses on the device.
1028 *    Note: subsequent RA will update lifetime.
1029 **/
1030static enum cleanup_prefix_rt_t
1031check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1032{
1033	struct inet6_ifaddr *ifa;
1034	struct inet6_dev *idev = ifp->idev;
1035	unsigned long lifetime;
1036	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1037
1038	*expires = jiffies;
1039
1040	list_for_each_entry(ifa, &idev->addr_list, if_list) {
1041		if (ifa == ifp)
1042			continue;
1043		if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1044				       ifp->prefix_len))
1045			continue;
1046		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1047			return CLEANUP_PREFIX_RT_NOP;
1048
1049		action = CLEANUP_PREFIX_RT_EXPIRE;
1050
1051		spin_lock(&ifa->lock);
1052
1053		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1054		/*
1055		 * Note: Because this address is
1056		 * not permanent, lifetime <
1057		 * LONG_MAX / HZ here.
1058		 */
1059		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1060			*expires = ifa->tstamp + lifetime * HZ;
1061		spin_unlock(&ifa->lock);
1062	}
1063
1064	return action;
1065}
1066
1067static void
1068cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1069{
1070	struct rt6_info *rt;
1071
1072	rt = addrconf_get_prefix_route(&ifp->addr,
1073				       ifp->prefix_len,
1074				       ifp->idev->dev,
1075				       0, RTF_GATEWAY | RTF_DEFAULT);
1076	if (rt) {
1077		if (del_rt)
1078			ip6_del_rt(rt);
1079		else {
1080			if (!(rt->rt6i_flags & RTF_EXPIRES))
1081				rt6_set_expires(rt, expires);
1082			ip6_rt_put(rt);
1083		}
1084	}
1085}
1086
1087
1088/* This function wants to get referenced ifp and releases it before return */
1089
1090static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1091{
1092	int state;
1093	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1094	unsigned long expires;
1095
1096	ASSERT_RTNL();
1097
1098	spin_lock_bh(&ifp->lock);
1099	state = ifp->state;
1100	ifp->state = INET6_IFADDR_STATE_DEAD;
1101	spin_unlock_bh(&ifp->lock);
1102
1103	if (state == INET6_IFADDR_STATE_DEAD)
1104		goto out;
1105
1106	spin_lock_bh(&addrconf_hash_lock);
1107	hlist_del_init_rcu(&ifp->addr_lst);
1108	spin_unlock_bh(&addrconf_hash_lock);
1109
1110	write_lock_bh(&ifp->idev->lock);
1111
1112	if (ifp->flags&IFA_F_TEMPORARY) {
1113		list_del(&ifp->tmp_list);
1114		if (ifp->ifpub) {
1115			in6_ifa_put(ifp->ifpub);
1116			ifp->ifpub = NULL;
1117		}
1118		__in6_ifa_put(ifp);
1119	}
1120
1121	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1122		action = check_cleanup_prefix_route(ifp, &expires);
1123
1124	list_del_init(&ifp->if_list);
1125	__in6_ifa_put(ifp);
1126
1127	write_unlock_bh(&ifp->idev->lock);
1128
1129	addrconf_del_dad_work(ifp);
1130
1131	ipv6_ifa_notify(RTM_DELADDR, ifp);
1132
1133	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1134
1135	if (action != CLEANUP_PREFIX_RT_NOP) {
1136		cleanup_prefix_route(ifp, expires,
1137			action == CLEANUP_PREFIX_RT_DEL);
1138	}
1139
1140	/* clean up prefsrc entries */
1141	rt6_remove_prefsrc(ifp);
1142out:
1143	in6_ifa_put(ifp);
1144}
1145
1146static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1147{
1148	struct inet6_dev *idev = ifp->idev;
1149	struct in6_addr addr, *tmpaddr;
1150	unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1151	unsigned long regen_advance;
1152	int tmp_plen;
1153	int ret = 0;
1154	u32 addr_flags;
1155	unsigned long now = jiffies;
1156
1157	write_lock_bh(&idev->lock);
1158	if (ift) {
1159		spin_lock_bh(&ift->lock);
1160		memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1161		spin_unlock_bh(&ift->lock);
1162		tmpaddr = &addr;
1163	} else {
1164		tmpaddr = NULL;
1165	}
1166retry:
1167	in6_dev_hold(idev);
1168	if (idev->cnf.use_tempaddr <= 0) {
1169		write_unlock_bh(&idev->lock);
1170		pr_info("%s: use_tempaddr is disabled\n", __func__);
1171		in6_dev_put(idev);
1172		ret = -1;
1173		goto out;
1174	}
1175	spin_lock_bh(&ifp->lock);
1176	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1177		idev->cnf.use_tempaddr = -1;	/*XXX*/
1178		spin_unlock_bh(&ifp->lock);
1179		write_unlock_bh(&idev->lock);
1180		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1181			__func__);
1182		in6_dev_put(idev);
1183		ret = -1;
1184		goto out;
1185	}
1186	in6_ifa_hold(ifp);
1187	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1188	__ipv6_try_regen_rndid(idev, tmpaddr);
1189	memcpy(&addr.s6_addr[8], idev->rndid, 8);
1190	age = (now - ifp->tstamp) / HZ;
1191	tmp_valid_lft = min_t(__u32,
1192			      ifp->valid_lft,
1193			      idev->cnf.temp_valid_lft + age);
1194	tmp_prefered_lft = min_t(__u32,
1195				 ifp->prefered_lft,
1196				 idev->cnf.temp_prefered_lft + age -
1197				 idev->cnf.max_desync_factor);
1198	tmp_plen = ifp->prefix_len;
1199	tmp_tstamp = ifp->tstamp;
1200	spin_unlock_bh(&ifp->lock);
1201
1202	regen_advance = idev->cnf.regen_max_retry *
1203			idev->cnf.dad_transmits *
1204			NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1205	write_unlock_bh(&idev->lock);
1206
1207	/* A temporary address is created only if this calculated Preferred
1208	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1209	 * an implementation must not create a temporary address with a zero
1210	 * Preferred Lifetime.
1211	 * Use age calculation as in addrconf_verify to avoid unnecessary
1212	 * temporary addresses being generated.
1213	 */
1214	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1215	if (tmp_prefered_lft <= regen_advance + age) {
1216		in6_ifa_put(ifp);
1217		in6_dev_put(idev);
1218		ret = -1;
1219		goto out;
1220	}
1221
1222	addr_flags = IFA_F_TEMPORARY;
1223	/* set in addrconf_prefix_rcv() */
1224	if (ifp->flags & IFA_F_OPTIMISTIC)
1225		addr_flags |= IFA_F_OPTIMISTIC;
1226
1227	ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1228			    ipv6_addr_scope(&addr), addr_flags,
1229			    tmp_valid_lft, tmp_prefered_lft);
1230	if (IS_ERR(ift)) {
1231		in6_ifa_put(ifp);
1232		in6_dev_put(idev);
1233		pr_info("%s: retry temporary address regeneration\n", __func__);
1234		tmpaddr = &addr;
1235		write_lock_bh(&idev->lock);
1236		goto retry;
1237	}
1238
1239	spin_lock_bh(&ift->lock);
1240	ift->ifpub = ifp;
1241	ift->cstamp = now;
1242	ift->tstamp = tmp_tstamp;
1243	spin_unlock_bh(&ift->lock);
1244
1245	addrconf_dad_start(ift);
1246	in6_ifa_put(ift);
1247	in6_dev_put(idev);
1248out:
1249	return ret;
1250}
1251
1252/*
1253 *	Choose an appropriate source address (RFC3484)
1254 */
1255enum {
1256	IPV6_SADDR_RULE_INIT = 0,
1257	IPV6_SADDR_RULE_LOCAL,
1258	IPV6_SADDR_RULE_SCOPE,
1259	IPV6_SADDR_RULE_PREFERRED,
1260#ifdef CONFIG_IPV6_MIP6
1261	IPV6_SADDR_RULE_HOA,
1262#endif
1263	IPV6_SADDR_RULE_OIF,
1264	IPV6_SADDR_RULE_LABEL,
1265	IPV6_SADDR_RULE_PRIVACY,
1266	IPV6_SADDR_RULE_ORCHID,
1267	IPV6_SADDR_RULE_PREFIX,
1268#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1269	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1270#endif
1271	IPV6_SADDR_RULE_MAX
1272};
1273
1274struct ipv6_saddr_score {
1275	int			rule;
1276	int			addr_type;
1277	struct inet6_ifaddr	*ifa;
1278	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1279	int			scopedist;
1280	int			matchlen;
1281};
1282
1283struct ipv6_saddr_dst {
1284	const struct in6_addr *addr;
1285	int ifindex;
1286	int scope;
1287	int label;
1288	unsigned int prefs;
1289};
1290
1291static inline int ipv6_saddr_preferred(int type)
1292{
1293	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1294		return 1;
1295	return 0;
1296}
1297
1298static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1299{
1300#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1301	return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1302#else
1303	return false;
1304#endif
1305}
1306
1307static int ipv6_get_saddr_eval(struct net *net,
1308			       struct ipv6_saddr_score *score,
1309			       struct ipv6_saddr_dst *dst,
1310			       int i)
1311{
1312	int ret;
1313
1314	if (i <= score->rule) {
1315		switch (i) {
1316		case IPV6_SADDR_RULE_SCOPE:
1317			ret = score->scopedist;
1318			break;
1319		case IPV6_SADDR_RULE_PREFIX:
1320			ret = score->matchlen;
1321			break;
1322		default:
1323			ret = !!test_bit(i, score->scorebits);
1324		}
1325		goto out;
1326	}
1327
1328	switch (i) {
1329	case IPV6_SADDR_RULE_INIT:
1330		/* Rule 0: remember if hiscore is not ready yet */
1331		ret = !!score->ifa;
1332		break;
1333	case IPV6_SADDR_RULE_LOCAL:
1334		/* Rule 1: Prefer same address */
1335		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1336		break;
1337	case IPV6_SADDR_RULE_SCOPE:
1338		/* Rule 2: Prefer appropriate scope
1339		 *
1340		 *      ret
1341		 *       ^
1342		 *    -1 |  d 15
1343		 *    ---+--+-+---> scope
1344		 *       |
1345		 *       |             d is scope of the destination.
1346		 *  B-d  |  \
1347		 *       |   \      <- smaller scope is better if
1348		 *  B-15 |    \        if scope is enough for destination.
1349		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
1350		 * d-C-1 | /
1351		 *       |/         <- greater is better
1352		 *   -C  /             if scope is not enough for destination.
1353		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
1354		 *
1355		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1356		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1357		 * Assume B = 0 and we get C > 29.
1358		 */
1359		ret = __ipv6_addr_src_scope(score->addr_type);
1360		if (ret >= dst->scope)
1361			ret = -ret;
1362		else
1363			ret -= 128;	/* 30 is enough */
1364		score->scopedist = ret;
1365		break;
1366	case IPV6_SADDR_RULE_PREFERRED:
1367	    {
1368		/* Rule 3: Avoid deprecated and optimistic addresses */
1369		u8 avoid = IFA_F_DEPRECATED;
1370
1371		if (!ipv6_use_optimistic_addr(score->ifa->idev))
1372			avoid |= IFA_F_OPTIMISTIC;
1373		ret = ipv6_saddr_preferred(score->addr_type) ||
1374		      !(score->ifa->flags & avoid);
1375		break;
1376	    }
1377#ifdef CONFIG_IPV6_MIP6
1378	case IPV6_SADDR_RULE_HOA:
1379	    {
1380		/* Rule 4: Prefer home address */
1381		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1382		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1383		break;
1384	    }
1385#endif
1386	case IPV6_SADDR_RULE_OIF:
1387		/* Rule 5: Prefer outgoing interface */
1388		ret = (!dst->ifindex ||
1389		       dst->ifindex == score->ifa->idev->dev->ifindex);
1390		break;
1391	case IPV6_SADDR_RULE_LABEL:
1392		/* Rule 6: Prefer matching label */
1393		ret = ipv6_addr_label(net,
1394				      &score->ifa->addr, score->addr_type,
1395				      score->ifa->idev->dev->ifindex) == dst->label;
1396		break;
1397	case IPV6_SADDR_RULE_PRIVACY:
1398	    {
1399		/* Rule 7: Prefer public address
1400		 * Note: prefer temporary address if use_tempaddr >= 2
1401		 */
1402		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1403				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1404				score->ifa->idev->cnf.use_tempaddr >= 2;
1405		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1406		break;
1407	    }
1408	case IPV6_SADDR_RULE_ORCHID:
1409		/* Rule 8-: Prefer ORCHID vs ORCHID or
1410		 *	    non-ORCHID vs non-ORCHID
1411		 */
1412		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1413			ipv6_addr_orchid(dst->addr));
1414		break;
1415	case IPV6_SADDR_RULE_PREFIX:
1416		/* Rule 8: Use longest matching prefix */
1417		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1418		if (ret > score->ifa->prefix_len)
1419			ret = score->ifa->prefix_len;
1420		score->matchlen = ret;
1421		break;
1422#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1423	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1424		/* Optimistic addresses still have lower precedence than other
1425		 * preferred addresses.
1426		 */
1427		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1428		break;
1429#endif
1430	default:
1431		ret = 0;
1432	}
1433
1434	if (ret)
1435		__set_bit(i, score->scorebits);
1436	score->rule = i;
1437out:
1438	return ret;
1439}
1440
1441static int __ipv6_dev_get_saddr(struct net *net,
1442				struct ipv6_saddr_dst *dst,
1443				struct inet6_dev *idev,
1444				struct ipv6_saddr_score *scores,
1445				int hiscore_idx)
1446{
1447	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1448
1449	read_lock_bh(&idev->lock);
1450	list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1451		int i;
1452
1453		/*
1454		 * - Tentative Address (RFC2462 section 5.4)
1455		 *  - A tentative address is not considered
1456		 *    "assigned to an interface" in the traditional
1457		 *    sense, unless it is also flagged as optimistic.
1458		 * - Candidate Source Address (section 4)
1459		 *  - In any case, anycast addresses, multicast
1460		 *    addresses, and the unspecified address MUST
1461		 *    NOT be included in a candidate set.
1462		 */
1463		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1464		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1465			continue;
1466
1467		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1468
1469		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1470			     score->addr_type & IPV6_ADDR_MULTICAST)) {
1471			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1472					    idev->dev->name);
1473			continue;
1474		}
1475
1476		score->rule = -1;
1477		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1478
1479		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1480			int minihiscore, miniscore;
1481
1482			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1483			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1484
1485			if (minihiscore > miniscore) {
1486				if (i == IPV6_SADDR_RULE_SCOPE &&
1487				    score->scopedist > 0) {
1488					/*
1489					 * special case:
1490					 * each remaining entry
1491					 * has too small (not enough)
1492					 * scope, because ifa entries
1493					 * are sorted by their scope
1494					 * values.
1495					 */
1496					goto out;
1497				}
1498				break;
1499			} else if (minihiscore < miniscore) {
1500				if (hiscore->ifa)
1501					in6_ifa_put(hiscore->ifa);
1502
1503				in6_ifa_hold(score->ifa);
1504
1505				swap(hiscore, score);
1506				hiscore_idx = 1 - hiscore_idx;
1507
1508				/* restore our iterator */
1509				score->ifa = hiscore->ifa;
1510
1511				break;
1512			}
1513		}
1514	}
1515out:
1516	read_unlock_bh(&idev->lock);
1517	return hiscore_idx;
1518}
1519
1520int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1521		       const struct in6_addr *daddr, unsigned int prefs,
1522		       struct in6_addr *saddr)
1523{
1524	struct ipv6_saddr_score scores[2], *hiscore;
1525	struct ipv6_saddr_dst dst;
1526	struct inet6_dev *idev;
1527	struct net_device *dev;
1528	int dst_type;
1529	bool use_oif_addr = false;
1530	int hiscore_idx = 0;
1531
1532	dst_type = __ipv6_addr_type(daddr);
1533	dst.addr = daddr;
1534	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1535	dst.scope = __ipv6_addr_src_scope(dst_type);
1536	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1537	dst.prefs = prefs;
1538
1539	scores[hiscore_idx].rule = -1;
1540	scores[hiscore_idx].ifa = NULL;
1541
1542	rcu_read_lock();
1543
1544	/* Candidate Source Address (section 4)
1545	 *  - multicast and link-local destination address,
1546	 *    the set of candidate source address MUST only
1547	 *    include addresses assigned to interfaces
1548	 *    belonging to the same link as the outgoing
1549	 *    interface.
1550	 * (- For site-local destination addresses, the
1551	 *    set of candidate source addresses MUST only
1552	 *    include addresses assigned to interfaces
1553	 *    belonging to the same site as the outgoing
1554	 *    interface.)
1555	 *  - "It is RECOMMENDED that the candidate source addresses
1556	 *    be the set of unicast addresses assigned to the
1557	 *    interface that will be used to send to the destination
1558	 *    (the 'outgoing' interface)." (RFC 6724)
1559	 */
1560	if (dst_dev) {
1561		idev = __in6_dev_get(dst_dev);
1562		if ((dst_type & IPV6_ADDR_MULTICAST) ||
1563		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1564		    (idev && idev->cnf.use_oif_addrs_only)) {
1565			use_oif_addr = true;
1566		}
1567	}
1568
1569	if (use_oif_addr) {
1570		if (idev)
1571			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1572	} else {
1573		for_each_netdev_rcu(net, dev) {
1574			idev = __in6_dev_get(dev);
1575			if (!idev)
1576				continue;
1577			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1578		}
1579	}
1580	rcu_read_unlock();
1581
1582	hiscore = &scores[hiscore_idx];
1583	if (!hiscore->ifa)
1584		return -EADDRNOTAVAIL;
1585
1586	*saddr = hiscore->ifa->addr;
1587	in6_ifa_put(hiscore->ifa);
1588	return 0;
1589}
1590EXPORT_SYMBOL(ipv6_dev_get_saddr);
1591
1592int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1593		      u32 banned_flags)
1594{
1595	struct inet6_ifaddr *ifp;
1596	int err = -EADDRNOTAVAIL;
1597
1598	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1599		if (ifp->scope > IFA_LINK)
1600			break;
1601		if (ifp->scope == IFA_LINK &&
1602		    !(ifp->flags & banned_flags)) {
1603			*addr = ifp->addr;
1604			err = 0;
1605			break;
1606		}
1607	}
1608	return err;
1609}
1610
1611int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1612		    u32 banned_flags)
1613{
1614	struct inet6_dev *idev;
1615	int err = -EADDRNOTAVAIL;
1616
1617	rcu_read_lock();
1618	idev = __in6_dev_get(dev);
1619	if (idev) {
1620		read_lock_bh(&idev->lock);
1621		err = __ipv6_get_lladdr(idev, addr, banned_flags);
1622		read_unlock_bh(&idev->lock);
1623	}
1624	rcu_read_unlock();
1625	return err;
1626}
1627
1628static int ipv6_count_addresses(struct inet6_dev *idev)
1629{
1630	int cnt = 0;
1631	struct inet6_ifaddr *ifp;
1632
1633	read_lock_bh(&idev->lock);
1634	list_for_each_entry(ifp, &idev->addr_list, if_list)
1635		cnt++;
1636	read_unlock_bh(&idev->lock);
1637	return cnt;
1638}
1639
1640int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1641		  const struct net_device *dev, int strict)
1642{
1643	return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1644}
1645EXPORT_SYMBOL(ipv6_chk_addr);
1646
1647int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1648			    const struct net_device *dev, int strict,
1649			    u32 banned_flags)
1650{
1651	struct inet6_ifaddr *ifp;
1652	unsigned int hash = inet6_addr_hash(addr);
1653	u32 ifp_flags;
1654
1655	rcu_read_lock_bh();
1656	hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1657		if (!net_eq(dev_net(ifp->idev->dev), net))
1658			continue;
1659		/* Decouple optimistic from tentative for evaluation here.
1660		 * Ban optimistic addresses explicitly, when required.
1661		 */
1662		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1663			    ? (ifp->flags&~IFA_F_TENTATIVE)
1664			    : ifp->flags;
1665		if (ipv6_addr_equal(&ifp->addr, addr) &&
1666		    !(ifp_flags&banned_flags) &&
1667		    (!dev || ifp->idev->dev == dev ||
1668		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1669			rcu_read_unlock_bh();
1670			return 1;
1671		}
1672	}
1673
1674	rcu_read_unlock_bh();
1675	return 0;
1676}
1677EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1678
1679static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1680			       struct net_device *dev)
1681{
1682	unsigned int hash = inet6_addr_hash(addr);
1683	struct inet6_ifaddr *ifp;
1684
1685	hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1686		if (!net_eq(dev_net(ifp->idev->dev), net))
1687			continue;
1688		if (ipv6_addr_equal(&ifp->addr, addr)) {
1689			if (!dev || ifp->idev->dev == dev)
1690				return true;
1691		}
1692	}
1693	return false;
1694}
1695
1696/* Compares an address/prefix_len with addresses on device @dev.
1697 * If one is found it returns true.
1698 */
1699bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1700	const unsigned int prefix_len, struct net_device *dev)
1701{
1702	struct inet6_dev *idev;
1703	struct inet6_ifaddr *ifa;
1704	bool ret = false;
1705
1706	rcu_read_lock();
1707	idev = __in6_dev_get(dev);
1708	if (idev) {
1709		read_lock_bh(&idev->lock);
1710		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1711			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1712			if (ret)
1713				break;
1714		}
1715		read_unlock_bh(&idev->lock);
1716	}
1717	rcu_read_unlock();
1718
1719	return ret;
1720}
1721EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1722
1723int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1724{
1725	struct inet6_dev *idev;
1726	struct inet6_ifaddr *ifa;
1727	int	onlink;
1728
1729	onlink = 0;
1730	rcu_read_lock();
1731	idev = __in6_dev_get(dev);
1732	if (idev) {
1733		read_lock_bh(&idev->lock);
1734		list_for_each_entry(ifa, &idev->addr_list, if_list) {
1735			onlink = ipv6_prefix_equal(addr, &ifa->addr,
1736						   ifa->prefix_len);
1737			if (onlink)
1738				break;
1739		}
1740		read_unlock_bh(&idev->lock);
1741	}
1742	rcu_read_unlock();
1743	return onlink;
1744}
1745EXPORT_SYMBOL(ipv6_chk_prefix);
1746
1747struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1748				     struct net_device *dev, int strict)
1749{
1750	struct inet6_ifaddr *ifp, *result = NULL;
1751	unsigned int hash = inet6_addr_hash(addr);
1752
1753	rcu_read_lock_bh();
1754	hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1755		if (!net_eq(dev_net(ifp->idev->dev), net))
1756			continue;
1757		if (ipv6_addr_equal(&ifp->addr, addr)) {
1758			if (!dev || ifp->idev->dev == dev ||
1759			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1760				result = ifp;
1761				in6_ifa_hold(ifp);
1762				break;
1763			}
1764		}
1765	}
1766	rcu_read_unlock_bh();
1767
1768	return result;
1769}
1770
1771/* Gets referenced address, destroys ifaddr */
1772
1773static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1774{
1775	if (ifp->flags&IFA_F_PERMANENT) {
1776		spin_lock_bh(&ifp->lock);
1777		addrconf_del_dad_work(ifp);
1778		ifp->flags |= IFA_F_TENTATIVE;
1779		if (dad_failed)
1780			ifp->flags |= IFA_F_DADFAILED;
1781		spin_unlock_bh(&ifp->lock);
1782		if (dad_failed)
1783			ipv6_ifa_notify(0, ifp);
1784		in6_ifa_put(ifp);
1785	} else if (ifp->flags&IFA_F_TEMPORARY) {
1786		struct inet6_ifaddr *ifpub;
1787		spin_lock_bh(&ifp->lock);
1788		ifpub = ifp->ifpub;
1789		if (ifpub) {
1790			in6_ifa_hold(ifpub);
1791			spin_unlock_bh(&ifp->lock);
1792			ipv6_create_tempaddr(ifpub, ifp);
1793			in6_ifa_put(ifpub);
1794		} else {
1795			spin_unlock_bh(&ifp->lock);
1796		}
1797		ipv6_del_addr(ifp);
1798	} else {
1799		ipv6_del_addr(ifp);
1800	}
1801}
1802
1803static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1804{
1805	int err = -ENOENT;
1806
1807	spin_lock_bh(&ifp->lock);
1808	if (ifp->state == INET6_IFADDR_STATE_DAD) {
1809		ifp->state = INET6_IFADDR_STATE_POSTDAD;
1810		err = 0;
1811	}
1812	spin_unlock_bh(&ifp->lock);
1813
1814	return err;
1815}
1816
1817void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1818{
1819	struct in6_addr addr;
1820	struct inet6_dev *idev = ifp->idev;
1821	struct net *net = dev_net(ifp->idev->dev);
1822
1823	if (addrconf_dad_end(ifp)) {
1824		in6_ifa_put(ifp);
1825		return;
1826	}
1827
1828	net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1829			     ifp->idev->dev->name, &ifp->addr);
1830
1831	spin_lock_bh(&ifp->lock);
1832
1833	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1834		int scope = ifp->scope;
1835		u32 flags = ifp->flags;
1836		struct in6_addr new_addr;
1837		struct inet6_ifaddr *ifp2;
1838		u32 valid_lft, preferred_lft;
1839		int pfxlen = ifp->prefix_len;
1840		int retries = ifp->stable_privacy_retry + 1;
1841
1842		if (retries > net->ipv6.sysctl.idgen_retries) {
1843			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1844					     ifp->idev->dev->name);
1845			goto errdad;
1846		}
1847
1848		new_addr = ifp->addr;
1849		if (ipv6_generate_stable_address(&new_addr, retries,
1850						 idev))
1851			goto errdad;
1852
1853		valid_lft = ifp->valid_lft;
1854		preferred_lft = ifp->prefered_lft;
1855
1856		spin_unlock_bh(&ifp->lock);
1857
1858		if (idev->cnf.max_addresses &&
1859		    ipv6_count_addresses(idev) >=
1860		    idev->cnf.max_addresses)
1861			goto lock_errdad;
1862
1863		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1864				     ifp->idev->dev->name);
1865
1866		ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1867				     scope, flags, valid_lft,
1868				     preferred_lft);
1869		if (IS_ERR(ifp2))
1870			goto lock_errdad;
1871
1872		spin_lock_bh(&ifp2->lock);
1873		ifp2->stable_privacy_retry = retries;
1874		ifp2->state = INET6_IFADDR_STATE_PREDAD;
1875		spin_unlock_bh(&ifp2->lock);
1876
1877		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1878		in6_ifa_put(ifp2);
1879lock_errdad:
1880		spin_lock_bh(&ifp->lock);
1881	} else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1882		addr.s6_addr32[0] = htonl(0xfe800000);
1883		addr.s6_addr32[1] = 0;
1884
1885		if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1886		    ipv6_addr_equal(&ifp->addr, &addr)) {
1887			/* DAD failed for link-local based on MAC address */
1888			idev->cnf.disable_ipv6 = 1;
1889
1890			pr_info("%s: IPv6 being disabled!\n",
1891				ifp->idev->dev->name);
1892		}
1893	}
1894
1895errdad:
1896	/* transition from _POSTDAD to _ERRDAD */
1897	ifp->state = INET6_IFADDR_STATE_ERRDAD;
1898	spin_unlock_bh(&ifp->lock);
1899
1900	addrconf_mod_dad_work(ifp, 0);
1901}
1902
1903/* Join to solicited addr multicast group.
1904 * caller must hold RTNL */
1905void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1906{
1907	struct in6_addr maddr;
1908
1909	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1910		return;
1911
1912	addrconf_addr_solict_mult(addr, &maddr);
1913	ipv6_dev_mc_inc(dev, &maddr);
1914}
1915
1916/* caller must hold RTNL */
1917void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1918{
1919	struct in6_addr maddr;
1920
1921	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1922		return;
1923
1924	addrconf_addr_solict_mult(addr, &maddr);
1925	__ipv6_dev_mc_dec(idev, &maddr);
1926}
1927
1928/* caller must hold RTNL */
1929static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1930{
1931	struct in6_addr addr;
1932
1933	if (ifp->prefix_len >= 127) /* RFC 6164 */
1934		return;
1935	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1936	if (ipv6_addr_any(&addr))
1937		return;
1938	__ipv6_dev_ac_inc(ifp->idev, &addr);
1939}
1940
1941/* caller must hold RTNL */
1942static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1943{
1944	struct in6_addr addr;
1945
1946	if (ifp->prefix_len >= 127) /* RFC 6164 */
1947		return;
1948	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1949	if (ipv6_addr_any(&addr))
1950		return;
1951	__ipv6_dev_ac_dec(ifp->idev, &addr);
1952}
1953
1954static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1955{
1956	if (dev->addr_len != IEEE802154_ADDR_LEN)
1957		return -1;
1958	memcpy(eui, dev->dev_addr, 8);
1959	eui[0] ^= 2;
1960	return 0;
1961}
1962
1963static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1964{
1965	union fwnet_hwaddr *ha;
1966
1967	if (dev->addr_len != FWNET_ALEN)
1968		return -1;
1969
1970	ha = (union fwnet_hwaddr *)dev->dev_addr;
1971
1972	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1973	eui[0] ^= 2;
1974	return 0;
1975}
1976
1977static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1978{
1979	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
1980	if (dev->addr_len != ARCNET_ALEN)
1981		return -1;
1982	memset(eui, 0, 7);
1983	eui[7] = *(u8 *)dev->dev_addr;
1984	return 0;
1985}
1986
1987static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1988{
1989	if (dev->addr_len != INFINIBAND_ALEN)
1990		return -1;
1991	memcpy(eui, dev->dev_addr + 12, 8);
1992	eui[0] |= 2;
1993	return 0;
1994}
1995
1996static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1997{
1998	if (addr == 0)
1999		return -1;
2000	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2001		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2002		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2003		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2004		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2005		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2006	eui[1] = 0;
2007	eui[2] = 0x5E;
2008	eui[3] = 0xFE;
2009	memcpy(eui + 4, &addr, 4);
2010	return 0;
2011}
2012
2013static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2014{
2015	if (dev->priv_flags & IFF_ISATAP)
2016		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2017	return -1;
2018}
2019
2020static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2021{
2022	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2023}
2024
2025static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2026{
2027	memcpy(eui, dev->perm_addr, 3);
2028	memcpy(eui + 5, dev->perm_addr + 3, 3);
2029	eui[3] = 0xFF;
2030	eui[4] = 0xFE;
2031	eui[0] ^= 2;
2032	return 0;
2033}
2034
2035static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2036{
2037	switch (dev->type) {
2038	case ARPHRD_ETHER:
2039	case ARPHRD_FDDI:
2040		return addrconf_ifid_eui48(eui, dev);
2041	case ARPHRD_ARCNET:
2042		return addrconf_ifid_arcnet(eui, dev);
2043	case ARPHRD_INFINIBAND:
2044		return addrconf_ifid_infiniband(eui, dev);
2045	case ARPHRD_SIT:
2046		return addrconf_ifid_sit(eui, dev);
2047	case ARPHRD_IPGRE:
2048		return addrconf_ifid_gre(eui, dev);
2049	case ARPHRD_6LOWPAN:
2050	case ARPHRD_IEEE802154:
2051		return addrconf_ifid_eui64(eui, dev);
2052	case ARPHRD_IEEE1394:
2053		return addrconf_ifid_ieee1394(eui, dev);
2054	case ARPHRD_TUNNEL6:
2055		return addrconf_ifid_ip6tnl(eui, dev);
2056	}
2057	return -1;
2058}
2059
2060static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2061{
2062	int err = -1;
2063	struct inet6_ifaddr *ifp;
2064
2065	read_lock_bh(&idev->lock);
2066	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2067		if (ifp->scope > IFA_LINK)
2068			break;
2069		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2070			memcpy(eui, ifp->addr.s6_addr+8, 8);
2071			err = 0;
2072			break;
2073		}
2074	}
2075	read_unlock_bh(&idev->lock);
2076	return err;
2077}
2078
2079/* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2080static void __ipv6_regen_rndid(struct inet6_dev *idev)
2081{
2082regen:
2083	get_random_bytes(idev->rndid, sizeof(idev->rndid));
2084	idev->rndid[0] &= ~0x02;
2085
2086	/*
2087	 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2088	 * check if generated address is not inappropriate
2089	 *
2090	 *  - Reserved subnet anycast (RFC 2526)
2091	 *	11111101 11....11 1xxxxxxx
2092	 *  - ISATAP (RFC4214) 6.1
2093	 *	00-00-5E-FE-xx-xx-xx-xx
2094	 *  - value 0
2095	 *  - XXX: already assigned to an address on the device
2096	 */
2097	if (idev->rndid[0] == 0xfd &&
2098	    (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2099	    (idev->rndid[7]&0x80))
2100		goto regen;
2101	if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2102		if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2103			goto regen;
2104		if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2105			goto regen;
2106	}
2107}
2108
2109static void ipv6_regen_rndid(unsigned long data)
2110{
2111	struct inet6_dev *idev = (struct inet6_dev *) data;
2112	unsigned long expires;
2113
2114	rcu_read_lock_bh();
2115	write_lock_bh(&idev->lock);
2116
2117	if (idev->dead)
2118		goto out;
2119
2120	__ipv6_regen_rndid(idev);
2121
2122	expires = jiffies +
2123		idev->cnf.temp_prefered_lft * HZ -
2124		idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2125		NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2126		idev->cnf.max_desync_factor * HZ;
2127	if (time_before(expires, jiffies)) {
2128		pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2129			__func__, idev->dev->name);
2130		goto out;
2131	}
2132
2133	if (!mod_timer(&idev->regen_timer, expires))
2134		in6_dev_hold(idev);
2135
2136out:
2137	write_unlock_bh(&idev->lock);
2138	rcu_read_unlock_bh();
2139	in6_dev_put(idev);
2140}
2141
2142static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2143{
2144	if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2145		__ipv6_regen_rndid(idev);
2146}
2147
2148/*
2149 *	Add prefix route.
2150 */
2151
2152static void
2153addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2154		      unsigned long expires, u32 flags)
2155{
2156	struct fib6_config cfg = {
2157		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2158		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2159		.fc_ifindex = dev->ifindex,
2160		.fc_expires = expires,
2161		.fc_dst_len = plen,
2162		.fc_flags = RTF_UP | flags,
2163		.fc_nlinfo.nl_net = dev_net(dev),
2164		.fc_protocol = RTPROT_KERNEL,
2165	};
2166
2167	cfg.fc_dst = *pfx;
2168
2169	/* Prevent useless cloning on PtP SIT.
2170	   This thing is done here expecting that the whole
2171	   class of non-broadcast devices need not cloning.
2172	 */
2173#if IS_ENABLED(CONFIG_IPV6_SIT)
2174	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2175		cfg.fc_flags |= RTF_NONEXTHOP;
2176#endif
2177
2178	ip6_route_add(&cfg);
2179}
2180
2181
2182static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2183						  int plen,
2184						  const struct net_device *dev,
2185						  u32 flags, u32 noflags)
2186{
2187	struct fib6_node *fn;
2188	struct rt6_info *rt = NULL;
2189	struct fib6_table *table;
2190	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2191
2192	table = fib6_get_table(dev_net(dev), tb_id);
2193	if (!table)
2194		return NULL;
2195
2196	read_lock_bh(&table->tb6_lock);
2197	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2198	if (!fn)
2199		goto out;
2200
2201	noflags |= RTF_CACHE;
2202	for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2203		if (rt->dst.dev->ifindex != dev->ifindex)
2204			continue;
2205		if ((rt->rt6i_flags & flags) != flags)
2206			continue;
2207		if ((rt->rt6i_flags & noflags) != 0)
2208			continue;
2209		dst_hold(&rt->dst);
2210		break;
2211	}
2212out:
2213	read_unlock_bh(&table->tb6_lock);
2214	return rt;
2215}
2216
2217
2218/* Create "default" multicast route to the interface */
2219
2220static void addrconf_add_mroute(struct net_device *dev)
2221{
2222	struct fib6_config cfg = {
2223		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2224		.fc_metric = IP6_RT_PRIO_ADDRCONF,
2225		.fc_ifindex = dev->ifindex,
2226		.fc_dst_len = 8,
2227		.fc_flags = RTF_UP,
2228		.fc_nlinfo.nl_net = dev_net(dev),
2229	};
2230
2231	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2232
2233	ip6_route_add(&cfg);
2234}
2235
2236static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2237{
2238	struct inet6_dev *idev;
2239
2240	ASSERT_RTNL();
2241
2242	idev = ipv6_find_idev(dev);
2243	if (!idev)
2244		return ERR_PTR(-ENOBUFS);
2245
2246	if (idev->cnf.disable_ipv6)
2247		return ERR_PTR(-EACCES);
2248
2249	/* Add default multicast route */
2250	if (!(dev->flags & IFF_LOOPBACK))
2251		addrconf_add_mroute(dev);
2252
2253	return idev;
2254}
2255
2256static void manage_tempaddrs(struct inet6_dev *idev,
2257			     struct inet6_ifaddr *ifp,
2258			     __u32 valid_lft, __u32 prefered_lft,
2259			     bool create, unsigned long now)
2260{
2261	u32 flags;
2262	struct inet6_ifaddr *ift;
2263
2264	read_lock_bh(&idev->lock);
2265	/* update all temporary addresses in the list */
2266	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2267		int age, max_valid, max_prefered;
2268
2269		if (ifp != ift->ifpub)
2270			continue;
2271
2272		/* RFC 4941 section 3.3:
2273		 * If a received option will extend the lifetime of a public
2274		 * address, the lifetimes of temporary addresses should
2275		 * be extended, subject to the overall constraint that no
2276		 * temporary addresses should ever remain "valid" or "preferred"
2277		 * for a time longer than (TEMP_VALID_LIFETIME) or
2278		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2279		 */
2280		age = (now - ift->cstamp) / HZ;
2281		max_valid = idev->cnf.temp_valid_lft - age;
2282		if (max_valid < 0)
2283			max_valid = 0;
2284
2285		max_prefered = idev->cnf.temp_prefered_lft -
2286			       idev->cnf.max_desync_factor - age;
2287		if (max_prefered < 0)
2288			max_prefered = 0;
2289
2290		if (valid_lft > max_valid)
2291			valid_lft = max_valid;
2292
2293		if (prefered_lft > max_prefered)
2294			prefered_lft = max_prefered;
2295
2296		spin_lock(&ift->lock);
2297		flags = ift->flags;
2298		ift->valid_lft = valid_lft;
2299		ift->prefered_lft = prefered_lft;
2300		ift->tstamp = now;
2301		if (prefered_lft > 0)
2302			ift->flags &= ~IFA_F_DEPRECATED;
2303
2304		spin_unlock(&ift->lock);
2305		if (!(flags&IFA_F_TENTATIVE))
2306			ipv6_ifa_notify(0, ift);
2307	}
2308
2309	if ((create || list_empty(&idev->tempaddr_list)) &&
2310	    idev->cnf.use_tempaddr > 0) {
2311		/* When a new public address is created as described
2312		 * in [ADDRCONF], also create a new temporary address.
2313		 * Also create a temporary address if it's enabled but
2314		 * no temporary address currently exists.
2315		 */
2316		read_unlock_bh(&idev->lock);
2317		ipv6_create_tempaddr(ifp, NULL);
2318	} else {
2319		read_unlock_bh(&idev->lock);
2320	}
2321}
2322
2323void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2324{
2325	struct prefix_info *pinfo;
2326	__u32 valid_lft;
2327	__u32 prefered_lft;
2328	int addr_type;
2329	u32 addr_flags = 0;
2330	struct inet6_dev *in6_dev;
2331	struct net *net = dev_net(dev);
2332
2333	pinfo = (struct prefix_info *) opt;
2334
2335	if (len < sizeof(struct prefix_info)) {
2336		ADBG("addrconf: prefix option too short\n");
2337		return;
2338	}
2339
2340	/*
2341	 *	Validation checks ([ADDRCONF], page 19)
2342	 */
2343
2344	addr_type = ipv6_addr_type(&pinfo->prefix);
2345
2346	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2347		return;
2348
2349	valid_lft = ntohl(pinfo->valid);
2350	prefered_lft = ntohl(pinfo->prefered);
2351
2352	if (prefered_lft > valid_lft) {
2353		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2354		return;
2355	}
2356
2357	in6_dev = in6_dev_get(dev);
2358
2359	if (!in6_dev) {
2360		net_dbg_ratelimited("addrconf: device %s not configured\n",
2361				    dev->name);
2362		return;
2363	}
2364
2365	/*
2366	 *	Two things going on here:
2367	 *	1) Add routes for on-link prefixes
2368	 *	2) Configure prefixes with the auto flag set
2369	 */
2370
2371	if (pinfo->onlink) {
2372		struct rt6_info *rt;
2373		unsigned long rt_expires;
2374
2375		/* Avoid arithmetic overflow. Really, we could
2376		 * save rt_expires in seconds, likely valid_lft,
2377		 * but it would require division in fib gc, that it
2378		 * not good.
2379		 */
2380		if (HZ > USER_HZ)
2381			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2382		else
2383			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2384
2385		if (addrconf_finite_timeout(rt_expires))
2386			rt_expires *= HZ;
2387
2388		rt = addrconf_get_prefix_route(&pinfo->prefix,
2389					       pinfo->prefix_len,
2390					       dev,
2391					       RTF_ADDRCONF | RTF_PREFIX_RT,
2392					       RTF_GATEWAY | RTF_DEFAULT);
2393
2394		if (rt) {
2395			/* Autoconf prefix route */
2396			if (valid_lft == 0) {
2397				ip6_del_rt(rt);
2398				rt = NULL;
2399			} else if (addrconf_finite_timeout(rt_expires)) {
2400				/* not infinity */
2401				rt6_set_expires(rt, jiffies + rt_expires);
2402			} else {
2403				rt6_clean_expires(rt);
2404			}
2405		} else if (valid_lft) {
2406			clock_t expires = 0;
2407			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2408			if (addrconf_finite_timeout(rt_expires)) {
2409				/* not infinity */
2410				flags |= RTF_EXPIRES;
2411				expires = jiffies_to_clock_t(rt_expires);
2412			}
2413			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2414					      dev, expires, flags);
2415		}
2416		ip6_rt_put(rt);
2417	}
2418
2419	/* Try to figure out our local address for this prefix */
2420
2421	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2422		struct inet6_ifaddr *ifp;
2423		struct in6_addr addr;
2424		int create = 0, update_lft = 0;
2425		bool tokenized = false;
2426
2427		if (pinfo->prefix_len == 64) {
2428			memcpy(&addr, &pinfo->prefix, 8);
2429
2430			if (!ipv6_addr_any(&in6_dev->token)) {
2431				read_lock_bh(&in6_dev->lock);
2432				memcpy(addr.s6_addr + 8,
2433				       in6_dev->token.s6_addr + 8, 8);
2434				read_unlock_bh(&in6_dev->lock);
2435				tokenized = true;
2436			} else if (in6_dev->addr_gen_mode ==
2437				   IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2438				   !ipv6_generate_stable_address(&addr, 0,
2439								 in6_dev)) {
2440				addr_flags |= IFA_F_STABLE_PRIVACY;
2441				goto ok;
2442			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2443				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2444				in6_dev_put(in6_dev);
2445				return;
2446			}
2447			goto ok;
2448		}
2449		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2450				    pinfo->prefix_len);
2451		in6_dev_put(in6_dev);
2452		return;
2453
2454ok:
2455
2456		ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2457
2458		if (!ifp && valid_lft) {
2459			int max_addresses = in6_dev->cnf.max_addresses;
2460
2461#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2462			if (in6_dev->cnf.optimistic_dad &&
2463			    !net->ipv6.devconf_all->forwarding && sllao)
2464				addr_flags |= IFA_F_OPTIMISTIC;
2465#endif
2466
2467			/* Do not allow to create too much of autoconfigured
2468			 * addresses; this would be too easy way to crash kernel.
2469			 */
2470			if (!max_addresses ||
2471			    ipv6_count_addresses(in6_dev) < max_addresses)
2472				ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2473						    pinfo->prefix_len,
2474						    addr_type&IPV6_ADDR_SCOPE_MASK,
2475						    addr_flags, valid_lft,
2476						    prefered_lft);
2477
2478			if (IS_ERR_OR_NULL(ifp)) {
2479				in6_dev_put(in6_dev);
2480				return;
2481			}
2482
2483			update_lft = 0;
2484			create = 1;
2485			spin_lock_bh(&ifp->lock);
2486			ifp->flags |= IFA_F_MANAGETEMPADDR;
2487			ifp->cstamp = jiffies;
2488			ifp->tokenized = tokenized;
2489			spin_unlock_bh(&ifp->lock);
2490			addrconf_dad_start(ifp);
2491		}
2492
2493		if (ifp) {
2494			u32 flags;
2495			unsigned long now;
2496			u32 stored_lft;
2497
2498			/* update lifetime (RFC2462 5.5.3 e) */
2499			spin_lock_bh(&ifp->lock);
2500			now = jiffies;
2501			if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2502				stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2503			else
2504				stored_lft = 0;
2505			if (!update_lft && !create && stored_lft) {
2506				const u32 minimum_lft = min_t(u32,
2507					stored_lft, MIN_VALID_LIFETIME);
2508				valid_lft = max(valid_lft, minimum_lft);
2509
2510				/* RFC4862 Section 5.5.3e:
2511				 * "Note that the preferred lifetime of the
2512				 *  corresponding address is always reset to
2513				 *  the Preferred Lifetime in the received
2514				 *  Prefix Information option, regardless of
2515				 *  whether the valid lifetime is also reset or
2516				 *  ignored."
2517				 *
2518				 * So we should always update prefered_lft here.
2519				 */
2520				update_lft = 1;
2521			}
2522
2523			if (update_lft) {
2524				ifp->valid_lft = valid_lft;
2525				ifp->prefered_lft = prefered_lft;
2526				ifp->tstamp = now;
2527				flags = ifp->flags;
2528				ifp->flags &= ~IFA_F_DEPRECATED;
2529				spin_unlock_bh(&ifp->lock);
2530
2531				if (!(flags&IFA_F_TENTATIVE))
2532					ipv6_ifa_notify(0, ifp);
2533			} else
2534				spin_unlock_bh(&ifp->lock);
2535
2536			manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2537					 create, now);
2538
2539			in6_ifa_put(ifp);
2540			addrconf_verify();
2541		}
2542	}
2543	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2544	in6_dev_put(in6_dev);
2545}
2546
2547/*
2548 *	Set destination address.
2549 *	Special case for SIT interfaces where we create a new "virtual"
2550 *	device.
2551 */
2552int addrconf_set_dstaddr(struct net *net, void __user *arg)
2553{
2554	struct in6_ifreq ireq;
2555	struct net_device *dev;
2556	int err = -EINVAL;
2557
2558	rtnl_lock();
2559
2560	err = -EFAULT;
2561	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2562		goto err_exit;
2563
2564	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2565
2566	err = -ENODEV;
2567	if (!dev)
2568		goto err_exit;
2569
2570#if IS_ENABLED(CONFIG_IPV6_SIT)
2571	if (dev->type == ARPHRD_SIT) {
2572		const struct net_device_ops *ops = dev->netdev_ops;
2573		struct ifreq ifr;
2574		struct ip_tunnel_parm p;
2575
2576		err = -EADDRNOTAVAIL;
2577		if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2578			goto err_exit;
2579
2580		memset(&p, 0, sizeof(p));
2581		p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2582		p.iph.saddr = 0;
2583		p.iph.version = 4;
2584		p.iph.ihl = 5;
2585		p.iph.protocol = IPPROTO_IPV6;
2586		p.iph.ttl = 64;
2587		ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2588
2589		if (ops->ndo_do_ioctl) {
2590			mm_segment_t oldfs = get_fs();
2591
2592			set_fs(KERNEL_DS);
2593			err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2594			set_fs(oldfs);
2595		} else
2596			err = -EOPNOTSUPP;
2597
2598		if (err == 0) {
2599			err = -ENOBUFS;
2600			dev = __dev_get_by_name(net, p.name);
2601			if (!dev)
2602				goto err_exit;
2603			err = dev_open(dev);
2604		}
2605	}
2606#endif
2607
2608err_exit:
2609	rtnl_unlock();
2610	return err;
2611}
2612
2613static int ipv6_mc_config(struct sock *sk, bool join,
2614			  const struct in6_addr *addr, int ifindex)
2615{
2616	int ret;
2617
2618	ASSERT_RTNL();
2619
2620	lock_sock(sk);
2621	if (join)
2622		ret = ipv6_sock_mc_join(sk, ifindex, addr);
2623	else
2624		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2625	release_sock(sk);
2626
2627	return ret;
2628}
2629
2630/*
2631 *	Manual configuration of address on an interface
2632 */
2633static int inet6_addr_add(struct net *net, int ifindex,
2634			  const struct in6_addr *pfx,
2635			  const struct in6_addr *peer_pfx,
2636			  unsigned int plen, __u32 ifa_flags,
2637			  __u32 prefered_lft, __u32 valid_lft)
2638{
2639	struct inet6_ifaddr *ifp;
2640	struct inet6_dev *idev;
2641	struct net_device *dev;
2642	unsigned long timeout;
2643	clock_t expires;
2644	int scope;
2645	u32 flags;
2646
2647	ASSERT_RTNL();
2648
2649	if (plen > 128)
2650		return -EINVAL;
2651
2652	/* check the lifetime */
2653	if (!valid_lft || prefered_lft > valid_lft)
2654		return -EINVAL;
2655
2656	if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2657		return -EINVAL;
2658
2659	dev = __dev_get_by_index(net, ifindex);
2660	if (!dev)
2661		return -ENODEV;
2662
2663	idev = addrconf_add_dev(dev);
2664	if (IS_ERR(idev))
2665		return PTR_ERR(idev);
2666
2667	if (ifa_flags & IFA_F_MCAUTOJOIN) {
2668		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2669					 true, pfx, ifindex);
2670
2671		if (ret < 0)
2672			return ret;
2673	}
2674
2675	scope = ipv6_addr_scope(pfx);
2676
2677	timeout = addrconf_timeout_fixup(valid_lft, HZ);
2678	if (addrconf_finite_timeout(timeout)) {
2679		expires = jiffies_to_clock_t(timeout * HZ);
2680		valid_lft = timeout;
2681		flags = RTF_EXPIRES;
2682	} else {
2683		expires = 0;
2684		flags = 0;
2685		ifa_flags |= IFA_F_PERMANENT;
2686	}
2687
2688	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2689	if (addrconf_finite_timeout(timeout)) {
2690		if (timeout == 0)
2691			ifa_flags |= IFA_F_DEPRECATED;
2692		prefered_lft = timeout;
2693	}
2694
2695	ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2696			    valid_lft, prefered_lft);
2697
2698	if (!IS_ERR(ifp)) {
2699		if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2700			addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2701					      expires, flags);
2702		}
2703
2704		/*
2705		 * Note that section 3.1 of RFC 4429 indicates
2706		 * that the Optimistic flag should not be set for
2707		 * manually configured addresses
2708		 */
2709		addrconf_dad_start(ifp);
2710		if (ifa_flags & IFA_F_MANAGETEMPADDR)
2711			manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2712					 true, jiffies);
2713		in6_ifa_put(ifp);
2714		addrconf_verify_rtnl();
2715		return 0;
2716	} else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2717		ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2718			       false, pfx, ifindex);
2719	}
2720
2721	return PTR_ERR(ifp);
2722}
2723
2724static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2725			  const struct in6_addr *pfx, unsigned int plen)
2726{
2727	struct inet6_ifaddr *ifp;
2728	struct inet6_dev *idev;
2729	struct net_device *dev;
2730
2731	if (plen > 128)
2732		return -EINVAL;
2733
2734	dev = __dev_get_by_index(net, ifindex);
2735	if (!dev)
2736		return -ENODEV;
2737
2738	idev = __in6_dev_get(dev);
2739	if (!idev)
2740		return -ENXIO;
2741
2742	read_lock_bh(&idev->lock);
2743	list_for_each_entry(ifp, &idev->addr_list, if_list) {
2744		if (ifp->prefix_len == plen &&
2745		    ipv6_addr_equal(pfx, &ifp->addr)) {
2746			in6_ifa_hold(ifp);
2747			read_unlock_bh(&idev->lock);
2748
2749			if (!(ifp->flags & IFA_F_TEMPORARY) &&
2750			    (ifa_flags & IFA_F_MANAGETEMPADDR))
2751				manage_tempaddrs(idev, ifp, 0, 0, false,
2752						 jiffies);
2753			ipv6_del_addr(ifp);
2754			addrconf_verify_rtnl();
2755			if (ipv6_addr_is_multicast(pfx)) {
2756				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2757					       false, pfx, dev->ifindex);
2758			}
2759			return 0;
2760		}
2761	}
2762	read_unlock_bh(&idev->lock);
2763	return -EADDRNOTAVAIL;
2764}
2765
2766
2767int addrconf_add_ifaddr(struct net *net, void __user *arg)
2768{
2769	struct in6_ifreq ireq;
2770	int err;
2771
2772	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2773		return -EPERM;
2774
2775	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2776		return -EFAULT;
2777
2778	rtnl_lock();
2779	err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2780			     ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2781			     INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2782	rtnl_unlock();
2783	return err;
2784}
2785
2786int addrconf_del_ifaddr(struct net *net, void __user *arg)
2787{
2788	struct in6_ifreq ireq;
2789	int err;
2790
2791	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2792		return -EPERM;
2793
2794	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2795		return -EFAULT;
2796
2797	rtnl_lock();
2798	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2799			     ireq.ifr6_prefixlen);
2800	rtnl_unlock();
2801	return err;
2802}
2803
2804static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2805		     int plen, int scope)
2806{
2807	struct inet6_ifaddr *ifp;
2808
2809	ifp = ipv6_add_addr(idev, addr, NULL, plen,
2810			    scope, IFA_F_PERMANENT,
2811			    INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2812	if (!IS_ERR(ifp)) {
2813		spin_lock_bh(&ifp->lock);
2814		ifp->flags &= ~IFA_F_TENTATIVE;
2815		spin_unlock_bh(&ifp->lock);
2816		ipv6_ifa_notify(RTM_NEWADDR, ifp);
2817		in6_ifa_put(ifp);
2818	}
2819}
2820
2821#if IS_ENABLED(CONFIG_IPV6_SIT)
2822static void sit_add_v4_addrs(struct inet6_dev *idev)
2823{
2824	struct in6_addr addr;
2825	struct net_device *dev;
2826	struct net *net = dev_net(idev->dev);
2827	int scope, plen;
2828	u32 pflags = 0;
2829
2830	ASSERT_RTNL();
2831
2832	memset(&addr, 0, sizeof(struct in6_addr));
2833	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2834
2835	if (idev->dev->flags&IFF_POINTOPOINT) {
2836		addr.s6_addr32[0] = htonl(0xfe800000);
2837		scope = IFA_LINK;
2838		plen = 64;
2839	} else {
2840		scope = IPV6_ADDR_COMPATv4;
2841		plen = 96;
2842		pflags |= RTF_NONEXTHOP;
2843	}
2844
2845	if (addr.s6_addr32[3]) {
2846		add_addr(idev, &addr, plen, scope);
2847		addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2848		return;
2849	}
2850
2851	for_each_netdev(net, dev) {
2852		struct in_device *in_dev = __in_dev_get_rtnl(dev);
2853		if (in_dev && (dev->flags & IFF_UP)) {
2854			struct in_ifaddr *ifa;
2855
2856			int flag = scope;
2857
2858			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2859
2860				addr.s6_addr32[3] = ifa->ifa_local;
2861
2862				if (ifa->ifa_scope == RT_SCOPE_LINK)
2863					continue;
2864				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2865					if (idev->dev->flags&IFF_POINTOPOINT)
2866						continue;
2867					flag |= IFA_HOST;
2868				}
2869
2870				add_addr(idev, &addr, plen, flag);
2871				addrconf_prefix_route(&addr, plen, idev->dev, 0,
2872						      pflags);
2873			}
2874		}
2875	}
2876}
2877#endif
2878
2879static void init_loopback(struct net_device *dev)
2880{
2881	struct inet6_dev  *idev;
2882	struct net_device *sp_dev;
2883	struct inet6_ifaddr *sp_ifa;
2884	struct rt6_info *sp_rt;
2885
2886	/* ::1 */
2887
2888	ASSERT_RTNL();
2889
2890	idev = ipv6_find_idev(dev);
2891	if (!idev) {
2892		pr_debug("%s: add_dev failed\n", __func__);
2893		return;
2894	}
2895
2896	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2897
2898	/* Add routes to other interface's IPv6 addresses */
2899	for_each_netdev(dev_net(dev), sp_dev) {
2900		if (!strcmp(sp_dev->name, dev->name))
2901			continue;
2902
2903		idev = __in6_dev_get(sp_dev);
2904		if (!idev)
2905			continue;
2906
2907		read_lock_bh(&idev->lock);
2908		list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2909
2910			if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2911				continue;
2912
2913			if (sp_ifa->rt) {
2914				/* This dst has been added to garbage list when
2915				 * lo device down, release this obsolete dst and
2916				 * reallocate a new router for ifa.
2917				 */
2918				if (sp_ifa->rt->dst.obsolete > 0) {
2919					ip6_rt_put(sp_ifa->rt);
2920					sp_ifa->rt = NULL;
2921				} else {
2922					continue;
2923				}
2924			}
2925
2926			sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2927
2928			/* Failure cases are ignored */
2929			if (!IS_ERR(sp_rt)) {
2930				sp_ifa->rt = sp_rt;
2931				ip6_ins_rt(sp_rt);
2932			}
2933		}
2934		read_unlock_bh(&idev->lock);
2935	}
2936}
2937
2938static void addrconf_add_linklocal(struct inet6_dev *idev,
2939				   const struct in6_addr *addr, u32 flags)
2940{
2941	struct inet6_ifaddr *ifp;
2942	u32 addr_flags = flags | IFA_F_PERMANENT;
2943
2944#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2945	if (idev->cnf.optimistic_dad &&
2946	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2947		addr_flags |= IFA_F_OPTIMISTIC;
2948#endif
2949
2950	ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2951			    INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2952	if (!IS_ERR(ifp)) {
2953		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2954		addrconf_dad_start(ifp);
2955		in6_ifa_put(ifp);
2956	}
2957}
2958
2959static bool ipv6_reserved_interfaceid(struct in6_addr address)
2960{
2961	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2962		return true;
2963
2964	if (address.s6_addr32[2] == htonl(0x02005eff) &&
2965	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2966		return true;
2967
2968	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2969	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2970		return true;
2971
2972	return false;
2973}
2974
2975static int ipv6_generate_stable_address(struct in6_addr *address,
2976					u8 dad_count,
2977					const struct inet6_dev *idev)
2978{
2979	static DEFINE_SPINLOCK(lock);
2980	static __u32 digest[SHA_DIGEST_WORDS];
2981	static __u32 workspace[SHA_WORKSPACE_WORDS];
2982
2983	static union {
2984		char __data[SHA_MESSAGE_BYTES];
2985		struct {
2986			struct in6_addr secret;
2987			__be32 prefix[2];
2988			unsigned char hwaddr[MAX_ADDR_LEN];
2989			u8 dad_count;
2990		} __packed;
2991	} data;
2992
2993	struct in6_addr secret;
2994	struct in6_addr temp;
2995	struct net *net = dev_net(idev->dev);
2996
2997	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2998
2999	if (idev->cnf.stable_secret.initialized)
3000		secret = idev->cnf.stable_secret.secret;
3001	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3002		secret = net->ipv6.devconf_dflt->stable_secret.secret;
3003	else
3004		return -1;
3005
3006retry:
3007	spin_lock_bh(&lock);
3008
3009	sha_init(digest);
3010	memset(&data, 0, sizeof(data));
3011	memset(workspace, 0, sizeof(workspace));
3012	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3013	data.prefix[0] = address->s6_addr32[0];
3014	data.prefix[1] = address->s6_addr32[1];
3015	data.secret = secret;
3016	data.dad_count = dad_count;
3017
3018	sha_transform(digest, data.__data, workspace);
3019
3020	temp = *address;
3021	temp.s6_addr32[2] = (__force __be32)digest[0];
3022	temp.s6_addr32[3] = (__force __be32)digest[1];
3023
3024	spin_unlock_bh(&lock);
3025
3026	if (ipv6_reserved_interfaceid(temp)) {
3027		dad_count++;
3028		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3029			return -1;
3030		goto retry;
3031	}
3032
3033	*address = temp;
3034	return 0;
3035}
3036
3037static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3038{
3039	struct in6_addr addr;
3040
3041	/* no link local addresses on L3 master devices */
3042	if (netif_is_l3_master(idev->dev))
3043		return;
3044
3045	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3046
3047	if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3048		if (!ipv6_generate_stable_address(&addr, 0, idev))
3049			addrconf_add_linklocal(idev, &addr,
3050					       IFA_F_STABLE_PRIVACY);
3051		else if (prefix_route)
3052			addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3053	} else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3054		/* addrconf_add_linklocal also adds a prefix_route and we
3055		 * only need to care about prefix routes if ipv6_generate_eui64
3056		 * couldn't generate one.
3057		 */
3058		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3059			addrconf_add_linklocal(idev, &addr, 0);
3060		else if (prefix_route)
3061			addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3062	}
3063}
3064
3065static void addrconf_dev_config(struct net_device *dev)
3066{
3067	struct inet6_dev *idev;
3068
3069	ASSERT_RTNL();
3070
3071	if ((dev->type != ARPHRD_ETHER) &&
3072	    (dev->type != ARPHRD_FDDI) &&
3073	    (dev->type != ARPHRD_ARCNET) &&
3074	    (dev->type != ARPHRD_INFINIBAND) &&
3075	    (dev->type != ARPHRD_IEEE802154) &&
3076	    (dev->type != ARPHRD_IEEE1394) &&
3077	    (dev->type != ARPHRD_TUNNEL6) &&
3078	    (dev->type != ARPHRD_6LOWPAN)) {
3079		/* Alas, we support only Ethernet autoconfiguration. */
3080		return;
3081	}
3082
3083	idev = addrconf_add_dev(dev);
3084	if (IS_ERR(idev))
3085		return;
3086
3087	addrconf_addr_gen(idev, false);
3088}
3089
3090#if IS_ENABLED(CONFIG_IPV6_SIT)
3091static void addrconf_sit_config(struct net_device *dev)
3092{
3093	struct inet6_dev *idev;
3094
3095	ASSERT_RTNL();
3096
3097	/*
3098	 * Configure the tunnel with one of our IPv4
3099	 * addresses... we should configure all of
3100	 * our v4 addrs in the tunnel
3101	 */
3102
3103	idev = ipv6_find_idev(dev);
3104	if (!idev) {
3105		pr_debug("%s: add_dev failed\n", __func__);
3106		return;
3107	}
3108
3109	if (dev->priv_flags & IFF_ISATAP) {
3110		addrconf_addr_gen(idev, false);
3111		return;
3112	}
3113
3114	sit_add_v4_addrs(idev);
3115
3116	if (dev->flags&IFF_POINTOPOINT)
3117		addrconf_add_mroute(dev);
3118}
3119#endif
3120
3121#if IS_ENABLED(CONFIG_NET_IPGRE)
3122static void addrconf_gre_config(struct net_device *dev)
3123{
3124	struct inet6_dev *idev;
3125
3126	ASSERT_RTNL();
3127
3128	idev = ipv6_find_idev(dev);
3129	if (!idev) {
3130		pr_debug("%s: add_dev failed\n", __func__);
3131		return;
3132	}
3133
3134	addrconf_addr_gen(idev, true);
3135	if (dev->flags & IFF_POINTOPOINT)
3136		addrconf_add_mroute(dev);
3137}
3138#endif
3139
3140static int addrconf_notify(struct notifier_block *this, unsigned long event,
3141			   void *ptr)
3142{
3143	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3144	struct inet6_dev *idev = __in6_dev_get(dev);
3145	int run_pending = 0;
3146	int err;
3147
3148	switch (event) {
3149	case NETDEV_REGISTER:
3150		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3151			idev = ipv6_add_dev(dev);
3152			if (IS_ERR(idev))
3153				return notifier_from_errno(PTR_ERR(idev));
3154		}
3155		break;
3156
3157	case NETDEV_CHANGEMTU:
3158		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3159		if (dev->mtu < IPV6_MIN_MTU) {
3160			addrconf_ifdown(dev, 1);
3161			break;
3162		}
3163
3164		if (idev) {
3165			rt6_mtu_change(dev, dev->mtu);
3166			idev->cnf.mtu6 = dev->mtu;
3167			break;
3168		}
3169
3170		/* allocate new idev */
3171		idev = ipv6_add_dev(dev);
3172		if (IS_ERR(idev))
3173			break;
3174
3175		/* device is still not ready */
3176		if (!(idev->if_flags & IF_READY))
3177			break;
3178
3179		run_pending = 1;
3180
3181		/* fall through */
3182
3183	case NETDEV_UP:
3184	case NETDEV_CHANGE:
3185		if (dev->flags & IFF_SLAVE)
3186			break;
3187
3188		if (idev && idev->cnf.disable_ipv6)
3189			break;
3190
3191		if (event == NETDEV_UP) {
3192			if (!addrconf_qdisc_ok(dev)) {
3193				/* device is not ready yet. */
3194				pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3195					dev->name);
3196				break;
3197			}
3198
3199			if (!idev && dev->mtu >= IPV6_MIN_MTU)
3200				idev = ipv6_add_dev(dev);
3201
3202			if (!IS_ERR_OR_NULL(idev)) {
3203				idev->if_flags |= IF_READY;
3204				run_pending = 1;
3205			}
3206		} else if (event == NETDEV_CHANGE) {
3207			if (!addrconf_qdisc_ok(dev)) {
3208				/* device is still not ready. */
3209				break;
3210			}
3211
3212			if (idev) {
3213				if (idev->if_flags & IF_READY)
3214					/* device is already configured. */
3215					break;
3216				idev->if_flags |= IF_READY;
3217			}
3218
3219			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3220				dev->name);
3221
3222			run_pending = 1;
3223		}
3224
3225		switch (dev->type) {
3226#if IS_ENABLED(CONFIG_IPV6_SIT)
3227		case ARPHRD_SIT:
3228			addrconf_sit_config(dev);
3229			break;
3230#endif
3231#if IS_ENABLED(CONFIG_NET_IPGRE)
3232		case ARPHRD_IPGRE:
3233			addrconf_gre_config(dev);
3234			break;
3235#endif
3236		case ARPHRD_LOOPBACK:
3237			init_loopback(dev);
3238			break;
3239
3240		default:
3241			addrconf_dev_config(dev);
3242			break;
3243		}
3244
3245		if (!IS_ERR_OR_NULL(idev)) {
3246			if (run_pending)
3247				addrconf_dad_run(idev);
3248
3249			/*
3250			 * If the MTU changed during the interface down,
3251			 * when the interface up, the changed MTU must be
3252			 * reflected in the idev as well as routers.
3253			 */
3254			if (idev->cnf.mtu6 != dev->mtu &&
3255			    dev->mtu >= IPV6_MIN_MTU) {
3256				rt6_mtu_change(dev, dev->mtu);
3257				idev->cnf.mtu6 = dev->mtu;
3258			}
3259			idev->tstamp = jiffies;
3260			inet6_ifinfo_notify(RTM_NEWLINK, idev);
3261
3262			/*
3263			 * If the changed mtu during down is lower than
3264			 * IPV6_MIN_MTU stop IPv6 on this interface.
3265			 */
3266			if (dev->mtu < IPV6_MIN_MTU)
3267				addrconf_ifdown(dev, 1);
3268		}
3269		break;
3270
3271	case NETDEV_DOWN:
3272	case NETDEV_UNREGISTER:
3273		/*
3274		 *	Remove all addresses from this interface.
3275		 */
3276		addrconf_ifdown(dev, event != NETDEV_DOWN);
3277		break;
3278
3279	case NETDEV_CHANGENAME:
3280		if (idev) {
3281			snmp6_unregister_dev(idev);
3282			addrconf_sysctl_unregister(idev);
3283			err = addrconf_sysctl_register(idev);
3284			if (err)
3285				return notifier_from_errno(err);
3286			err = snmp6_register_dev(idev);
3287			if (err) {
3288				addrconf_sysctl_unregister(idev);
3289				return notifier_from_errno(err);
3290			}
3291		}
3292		break;
3293
3294	case NETDEV_PRE_TYPE_CHANGE:
3295	case NETDEV_POST_TYPE_CHANGE:
3296		addrconf_type_change(dev, event);
3297		break;
3298	}
3299
3300	return NOTIFY_OK;
3301}
3302
3303/*
3304 *	addrconf module should be notified of a device going up
3305 */
3306static struct notifier_block ipv6_dev_notf = {
3307	.notifier_call = addrconf_notify,
3308};
3309
3310static void addrconf_type_change(struct net_device *dev, unsigned long event)
3311{
3312	struct inet6_dev *idev;
3313	ASSERT_RTNL();
3314
3315	idev = __in6_dev_get(dev);
3316
3317	if (event == NETDEV_POST_TYPE_CHANGE)
3318		ipv6_mc_remap(idev);
3319	else if (event == NETDEV_PRE_TYPE_CHANGE)
3320		ipv6_mc_unmap(idev);
3321}
3322
3323static int addrconf_ifdown(struct net_device *dev, int how)
3324{
3325	struct net *net = dev_net(dev);
3326	struct inet6_dev *idev;
3327	struct inet6_ifaddr *ifa;
3328	int state, i;
3329
3330	ASSERT_RTNL();
3331
3332	rt6_ifdown(net, dev);
3333	neigh_ifdown(&nd_tbl, dev);
3334
3335	idev = __in6_dev_get(dev);
3336	if (!idev)
3337		return -ENODEV;
3338
3339	/*
3340	 * Step 1: remove reference to ipv6 device from parent device.
3341	 *	   Do not dev_put!
3342	 */
3343	if (how) {
3344		idev->dead = 1;
3345
3346		/* protected by rtnl_lock */
3347		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3348
3349		/* Step 1.5: remove snmp6 entry */
3350		snmp6_unregister_dev(idev);
3351
3352	}
3353
3354	/* Step 2: clear hash table */
3355	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3356		struct hlist_head *h = &inet6_addr_lst[i];
3357
3358		spin_lock_bh(&addrconf_hash_lock);
3359restart:
3360		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3361			if (ifa->idev == idev) {
3362				hlist_del_init_rcu(&ifa->addr_lst);
3363				addrconf_del_dad_work(ifa);
3364				goto restart;
3365			}
3366		}
3367		spin_unlock_bh(&addrconf_hash_lock);
3368	}
3369
3370	write_lock_bh(&idev->lock);
3371
3372	addrconf_del_rs_timer(idev);
3373
3374	/* Step 2: clear flags for stateless addrconf */
3375	if (!how)
3376		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3377
3378	if (how && del_timer(&idev->regen_timer))
3379		in6_dev_put(idev);
3380
3381	/* Step 3: clear tempaddr list */
3382	while (!list_empty(&idev->tempaddr_list)) {
3383		ifa = list_first_entry(&idev->tempaddr_list,
3384				       struct inet6_ifaddr, tmp_list);
3385		list_del(&ifa->tmp_list);
3386		write_unlock_bh(&idev->lock);
3387		spin_lock_bh(&ifa->lock);
3388
3389		if (ifa->ifpub) {
3390			in6_ifa_put(ifa->ifpub);
3391			ifa->ifpub = NULL;
3392		}
3393		spin_unlock_bh(&ifa->lock);
3394		in6_ifa_put(ifa);
3395		write_lock_bh(&idev->lock);
3396	}
3397
3398	while (!list_empty(&idev->addr_list)) {
3399		ifa = list_first_entry(&idev->addr_list,
3400				       struct inet6_ifaddr, if_list);
3401		addrconf_del_dad_work(ifa);
3402
3403		list_del(&ifa->if_list);
3404
3405		write_unlock_bh(&idev->lock);
3406
3407		spin_lock_bh(&ifa->lock);
3408		state = ifa->state;
3409		ifa->state = INET6_IFADDR_STATE_DEAD;
3410		spin_unlock_bh(&ifa->lock);
3411
3412		if (state != INET6_IFADDR_STATE_DEAD) {
3413			__ipv6_ifa_notify(RTM_DELADDR, ifa);
3414			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3415		}
3416		in6_ifa_put(ifa);
3417
3418		write_lock_bh(&idev->lock);
3419	}
3420
3421	write_unlock_bh(&idev->lock);
3422
3423	/* Step 5: Discard anycast and multicast list */
3424	if (how) {
3425		ipv6_ac_destroy_dev(idev);
3426		ipv6_mc_destroy_dev(idev);
3427	} else {
3428		ipv6_mc_down(idev);
3429	}
3430
3431	idev->tstamp = jiffies;
3432
3433	/* Last: Shot the device (if unregistered) */
3434	if (how) {
3435		addrconf_sysctl_unregister(idev);
3436		neigh_parms_release(&nd_tbl, idev->nd_parms);
3437		neigh_ifdown(&nd_tbl, dev);
3438		in6_dev_put(idev);
3439	}
3440	return 0;
3441}
3442
3443static void addrconf_rs_timer(unsigned long data)
3444{
3445	struct inet6_dev *idev = (struct inet6_dev *)data;
3446	struct net_device *dev = idev->dev;
3447	struct in6_addr lladdr;
3448
3449	write_lock(&idev->lock);
3450	if (idev->dead || !(idev->if_flags & IF_READY))
3451		goto out;
3452
3453	if (!ipv6_accept_ra(idev))
3454		goto out;
3455
3456	/* Announcement received after solicitation was sent */
3457	if (idev->if_flags & IF_RA_RCVD)
3458		goto out;
3459
3460	if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3461		write_unlock(&idev->lock);
3462		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3463			ndisc_send_rs(dev, &lladdr,
3464				      &in6addr_linklocal_allrouters);
3465		else
3466			goto put;
3467
3468		write_lock(&idev->lock);
3469		/* The wait after the last probe can be shorter */
3470		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3471					     idev->cnf.rtr_solicits) ?
3472				      idev->cnf.rtr_solicit_delay :
3473				      idev->cnf.rtr_solicit_interval);
3474	} else {
3475		/*
3476		 * Note: we do not support deprecated "all on-link"
3477		 * assumption any longer.
3478		 */
3479		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3480	}
3481
3482out:
3483	write_unlock(&idev->lock);
3484put:
3485	in6_dev_put(idev);
3486}
3487
3488/*
3489 *	Duplicate Address Detection
3490 */
3491static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3492{
3493	unsigned long rand_num;
3494	struct inet6_dev *idev = ifp->idev;
3495
3496	if (ifp->flags & IFA_F_OPTIMISTIC)
3497		rand_num = 0;
3498	else
3499		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3500
3501	ifp->dad_probes = idev->cnf.dad_transmits;
3502	addrconf_mod_dad_work(ifp, rand_num);
3503}
3504
3505static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3506{
3507	struct inet6_dev *idev = ifp->idev;
3508	struct net_device *dev = idev->dev;
3509	bool notify = false;
3510
3511	addrconf_join_solict(dev, &ifp->addr);
3512
3513	prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3514
3515	read_lock_bh(&idev->lock);
3516	spin_lock(&ifp->lock);
3517	if (ifp->state == INET6_IFADDR_STATE_DEAD)
3518		goto out;
3519
3520	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3521	    idev->cnf.accept_dad < 1 ||
3522	    !(ifp->flags&IFA_F_TENTATIVE) ||
3523	    ifp->flags & IFA_F_NODAD) {
3524		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3525		spin_unlock(&ifp->lock);
3526		read_unlock_bh(&idev->lock);
3527
3528		addrconf_dad_completed(ifp);
3529		return;
3530	}
3531
3532	if (!(idev->if_flags & IF_READY)) {
3533		spin_unlock(&ifp->lock);
3534		read_unlock_bh(&idev->lock);
3535		/*
3536		 * If the device is not ready:
3537		 * - keep it tentative if it is a permanent address.
3538		 * - otherwise, kill it.
3539		 */
3540		in6_ifa_hold(ifp);
3541		addrconf_dad_stop(ifp, 0);
3542		return;
3543	}
3544
3545	/*
3546	 * Optimistic nodes can start receiving
3547	 * Frames right away
3548	 */
3549	if (ifp->flags & IFA_F_OPTIMISTIC) {
3550		ip6_ins_rt(ifp->rt);
3551		if (ipv6_use_optimistic_addr(idev)) {
3552			/* Because optimistic nodes can use this address,
3553			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3554			 */
3555			notify = true;
3556		}
3557	}
3558
3559	addrconf_dad_kick(ifp);
3560out:
3561	spin_unlock(&ifp->lock);
3562	read_unlock_bh(&idev->lock);
3563	if (notify)
3564		ipv6_ifa_notify(RTM_NEWADDR, ifp);
3565}
3566
3567static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3568{
3569	bool begin_dad = false;
3570
3571	spin_lock_bh(&ifp->lock);
3572	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3573		ifp->state = INET6_IFADDR_STATE_PREDAD;
3574		begin_dad = true;
3575	}
3576	spin_unlock_bh(&ifp->lock);
3577
3578	if (begin_dad)
3579		addrconf_mod_dad_work(ifp, 0);
3580}
3581
3582static void addrconf_dad_work(struct work_struct *w)
3583{
3584	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3585						struct inet6_ifaddr,
3586						dad_work);
3587	struct inet6_dev *idev = ifp->idev;
3588	struct in6_addr mcaddr;
3589
3590	enum {
3591		DAD_PROCESS,
3592		DAD_BEGIN,
3593		DAD_ABORT,
3594	} action = DAD_PROCESS;
3595
3596	rtnl_lock();
3597
3598	spin_lock_bh(&ifp->lock);
3599	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3600		action = DAD_BEGIN;
3601		ifp->state = INET6_IFADDR_STATE_DAD;
3602	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3603		action = DAD_ABORT;
3604		ifp->state = INET6_IFADDR_STATE_POSTDAD;
3605	}
3606	spin_unlock_bh(&ifp->lock);
3607
3608	if (action == DAD_BEGIN) {
3609		addrconf_dad_begin(ifp);
3610		goto out;
3611	} else if (action == DAD_ABORT) {
3612		addrconf_dad_stop(ifp, 1);
3613		goto out;
3614	}
3615
3616	if (!ifp->dad_probes && addrconf_dad_end(ifp))
3617		goto out;
3618
3619	write_lock_bh(&idev->lock);
3620	if (idev->dead || !(idev->if_flags & IF_READY)) {
3621		write_unlock_bh(&idev->lock);
3622		goto out;
3623	}
3624
3625	spin_lock(&ifp->lock);
3626	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3627		spin_unlock(&ifp->lock);
3628		write_unlock_bh(&idev->lock);
3629		goto out;
3630	}
3631
3632	if (ifp->dad_probes == 0) {
3633		/*
3634		 * DAD was successful
3635		 */
3636
3637		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3638		spin_unlock(&ifp->lock);
3639		write_unlock_bh(&idev->lock);
3640
3641		addrconf_dad_completed(ifp);
3642
3643		goto out;
3644	}
3645
3646	ifp->dad_probes--;
3647	addrconf_mod_dad_work(ifp,
3648			      NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3649	spin_unlock(&ifp->lock);
3650	write_unlock_bh(&idev->lock);
3651
3652	/* send a neighbour solicitation for our addr */
3653	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3654	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3655out:
3656	in6_ifa_put(ifp);
3657	rtnl_unlock();
3658}
3659
3660/* ifp->idev must be at least read locked */
3661static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3662{
3663	struct inet6_ifaddr *ifpiter;
3664	struct inet6_dev *idev = ifp->idev;
3665
3666	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3667		if (ifpiter->scope > IFA_LINK)
3668			break;
3669		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3670		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3671				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3672		    IFA_F_PERMANENT)
3673			return false;
3674	}
3675	return true;
3676}
3677
3678static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3679{
3680	struct net_device *dev = ifp->idev->dev;
3681	struct in6_addr lladdr;
3682	bool send_rs, send_mld;
3683
3684	addrconf_del_dad_work(ifp);
3685
3686	/*
3687	 *	Configure the address for reception. Now it is valid.
3688	 */
3689
3690	ipv6_ifa_notify(RTM_NEWADDR, ifp);
3691
3692	/* If added prefix is link local and we are prepared to process
3693	   router advertisements, start sending router solicitations.
3694	 */
3695
3696	read_lock_bh(&ifp->idev->lock);
3697	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3698	send_rs = send_mld &&
3699		  ipv6_accept_ra(ifp->idev) &&
3700		  ifp->idev->cnf.rtr_solicits > 0 &&
3701		  (dev->flags&IFF_LOOPBACK) == 0;
3702	read_unlock_bh(&ifp->idev->lock);
3703
3704	/* While dad is in progress mld report's source address is in6_addrany.
3705	 * Resend with proper ll now.
3706	 */
3707	if (send_mld)
3708		ipv6_mc_dad_complete(ifp->idev);
3709
3710	if (send_rs) {
3711		/*
3712		 *	If a host as already performed a random delay
3713		 *	[...] as part of DAD [...] there is no need
3714		 *	to delay again before sending the first RS
3715		 */
3716		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3717			return;
3718		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3719
3720		write_lock_bh(&ifp->idev->lock);
3721		spin_lock(&ifp->lock);
3722		ifp->idev->rs_probes = 1;
3723		ifp->idev->if_flags |= IF_RS_SENT;
3724		addrconf_mod_rs_timer(ifp->idev,
3725				      ifp->idev->cnf.rtr_solicit_interval);
3726		spin_unlock(&ifp->lock);
3727		write_unlock_bh(&ifp->idev->lock);
3728	}
3729}
3730
3731static void addrconf_dad_run(struct inet6_dev *idev)
3732{
3733	struct inet6_ifaddr *ifp;
3734
3735	read_lock_bh(&idev->lock);
3736	list_for_each_entry(ifp, &idev->addr_list, if_list) {
3737		spin_lock(&ifp->lock);
3738		if (ifp->flags & IFA_F_TENTATIVE &&
3739		    ifp->state == INET6_IFADDR_STATE_DAD)
3740			addrconf_dad_kick(ifp);
3741		spin_unlock(&ifp->lock);
3742	}
3743	read_unlock_bh(&idev->lock);
3744}
3745
3746#ifdef CONFIG_PROC_FS
3747struct if6_iter_state {
3748	struct seq_net_private p;
3749	int bucket;
3750	int offset;
3751};
3752
3753static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3754{
3755	struct inet6_ifaddr *ifa = NULL;
3756	struct if6_iter_state *state = seq->private;
3757	struct net *net = seq_file_net(seq);
3758	int p = 0;
3759
3760	/* initial bucket if pos is 0 */
3761	if (pos == 0) {
3762		state->bucket = 0;
3763		state->offset = 0;
3764	}
3765
3766	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3767		hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3768					 addr_lst) {
3769			if (!net_eq(dev_net(ifa->idev->dev), net))
3770				continue;
3771			/* sync with offset */
3772			if (p < state->offset) {
3773				p++;
3774				continue;
3775			}
3776			state->offset++;
3777			return ifa;
3778		}
3779
3780		/* prepare for next bucket */
3781		state->offset = 0;
3782		p = 0;
3783	}
3784	return NULL;
3785}
3786
3787static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3788					 struct inet6_ifaddr *ifa)
3789{
3790	struct if6_iter_state *state = seq->private;
3791	struct net *net = seq_file_net(seq);
3792
3793	hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3794		if (!net_eq(dev_net(ifa->idev->dev), net))
3795			continue;
3796		state->offset++;
3797		return ifa;
3798	}
3799
3800	while (++state->bucket < IN6_ADDR_HSIZE) {
3801		state->offset = 0;
3802		hlist_for_each_entry_rcu_bh(ifa,
3803				     &inet6_addr_lst[state->bucket], addr_lst) {
3804			if (!net_eq(dev_net(ifa->idev->dev), net))
3805				continue;
3806			state->offset++;
3807			return ifa;
3808		}
3809	}
3810
3811	return NULL;
3812}
3813
3814static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3815	__acquires(rcu_bh)
3816{
3817	rcu_read_lock_bh();
3818	return if6_get_first(seq, *pos);
3819}
3820
3821static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3822{
3823	struct inet6_ifaddr *ifa;
3824
3825	ifa = if6_get_next(seq, v);
3826	++*pos;
3827	return ifa;
3828}
3829
3830static void if6_seq_stop(struct seq_file *seq, void *v)
3831	__releases(rcu_bh)
3832{
3833	rcu_read_unlock_bh();
3834}
3835
3836static int if6_seq_show(struct seq_file *seq, void *v)
3837{
3838	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3839	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3840		   &ifp->addr,
3841		   ifp->idev->dev->ifindex,
3842		   ifp->prefix_len,
3843		   ifp->scope,
3844		   (u8) ifp->flags,
3845		   ifp->idev->dev->name);
3846	return 0;
3847}
3848
3849static const struct seq_operations if6_seq_ops = {
3850	.start	= if6_seq_start,
3851	.next	= if6_seq_next,
3852	.show	= if6_seq_show,
3853	.stop	= if6_seq_stop,
3854};
3855
3856static int if6_seq_open(struct inode *inode, struct file *file)
3857{
3858	return seq_open_net(inode, file, &if6_seq_ops,
3859			    sizeof(struct if6_iter_state));
3860}
3861
3862static const struct file_operations if6_fops = {
3863	.owner		= THIS_MODULE,
3864	.open		= if6_seq_open,
3865	.read		= seq_read,
3866	.llseek		= seq_lseek,
3867	.release	= seq_release_net,
3868};
3869
3870static int __net_init if6_proc_net_init(struct net *net)
3871{
3872	if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3873		return -ENOMEM;
3874	return 0;
3875}
3876
3877static void __net_exit if6_proc_net_exit(struct net *net)
3878{
3879	remove_proc_entry("if_inet6", net->proc_net);
3880}
3881
3882static struct pernet_operations if6_proc_net_ops = {
3883	.init = if6_proc_net_init,
3884	.exit = if6_proc_net_exit,
3885};
3886
3887int __init if6_proc_init(void)
3888{
3889	return register_pernet_subsys(&if6_proc_net_ops);
3890}
3891
3892void if6_proc_exit(void)
3893{
3894	unregister_pernet_subsys(&if6_proc_net_ops);
3895}
3896#endif	/* CONFIG_PROC_FS */
3897
3898#if IS_ENABLED(CONFIG_IPV6_MIP6)
3899/* Check if address is a home address configured on any interface. */
3900int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3901{
3902	int ret = 0;
3903	struct inet6_ifaddr *ifp = NULL;
3904	unsigned int hash = inet6_addr_hash(addr);
3905
3906	rcu_read_lock_bh();
3907	hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3908		if (!net_eq(dev_net(ifp->idev->dev), net))
3909			continue;
3910		if (ipv6_addr_equal(&ifp->addr, addr) &&
3911		    (ifp->flags & IFA_F_HOMEADDRESS)) {
3912			ret = 1;
3913			break;
3914		}
3915	}
3916	rcu_read_unlock_bh();
3917	return ret;
3918}
3919#endif
3920
3921/*
3922 *	Periodic address status verification
3923 */
3924
3925static void addrconf_verify_rtnl(void)
3926{
3927	unsigned long now, next, next_sec, next_sched;
3928	struct inet6_ifaddr *ifp;
3929	int i;
3930
3931	ASSERT_RTNL();
3932
3933	rcu_read_lock_bh();
3934	now = jiffies;
3935	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3936
3937	cancel_delayed_work(&addr_chk_work);
3938
3939	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3940restart:
3941		hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3942			unsigned long age;
3943
3944			/* When setting preferred_lft to a value not zero or
3945			 * infinity, while valid_lft is infinity
3946			 * IFA_F_PERMANENT has a non-infinity life time.
3947			 */
3948			if ((ifp->flags & IFA_F_PERMANENT) &&
3949			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
3950				continue;
3951
3952			spin_lock(&ifp->lock);
3953			/* We try to batch several events at once. */
3954			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3955
3956			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3957			    age >= ifp->valid_lft) {
3958				spin_unlock(&ifp->lock);
3959				in6_ifa_hold(ifp);
3960				ipv6_del_addr(ifp);
3961				goto restart;
3962			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3963				spin_unlock(&ifp->lock);
3964				continue;
3965			} else if (age >= ifp->prefered_lft) {
3966				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3967				int deprecate = 0;
3968
3969				if (!(ifp->flags&IFA_F_DEPRECATED)) {
3970					deprecate = 1;
3971					ifp->flags |= IFA_F_DEPRECATED;
3972				}
3973
3974				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3975				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3976					next = ifp->tstamp + ifp->valid_lft * HZ;
3977
3978				spin_unlock(&ifp->lock);
3979
3980				if (deprecate) {
3981					in6_ifa_hold(ifp);
3982
3983					ipv6_ifa_notify(0, ifp);
3984					in6_ifa_put(ifp);
3985					goto restart;
3986				}
3987			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
3988				   !(ifp->flags&IFA_F_TENTATIVE)) {
3989				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3990					ifp->idev->cnf.dad_transmits *
3991					NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3992
3993				if (age >= ifp->prefered_lft - regen_advance) {
3994					struct inet6_ifaddr *ifpub = ifp->ifpub;
3995					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3996						next = ifp->tstamp + ifp->prefered_lft * HZ;
3997					if (!ifp->regen_count && ifpub) {
3998						ifp->regen_count++;
3999						in6_ifa_hold(ifp);
4000						in6_ifa_hold(ifpub);
4001						spin_unlock(&ifp->lock);
4002
4003						spin_lock(&ifpub->lock);
4004						ifpub->regen_count = 0;
4005						spin_unlock(&ifpub->lock);
4006						ipv6_create_tempaddr(ifpub, ifp);
4007						in6_ifa_put(ifpub);
4008						in6_ifa_put(ifp);
4009						goto restart;
4010					}
4011				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4012					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4013				spin_unlock(&ifp->lock);
4014			} else {
4015				/* ifp->prefered_lft <= ifp->valid_lft */
4016				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4017					next = ifp->tstamp + ifp->prefered_lft * HZ;
4018				spin_unlock(&ifp->lock);
4019			}
4020		}
4021	}
4022
4023	next_sec = round_jiffies_up(next);
4024	next_sched = next;
4025
4026	/* If rounded timeout is accurate enough, accept it. */
4027	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4028		next_sched = next_sec;
4029
4030	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4031	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4032		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4033
4034	ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4035	      now, next, next_sec, next_sched);
4036	mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4037	rcu_read_unlock_bh();
4038}
4039
4040static void addrconf_verify_work(struct work_struct *w)
4041{
4042	rtnl_lock();
4043	addrconf_verify_rtnl();
4044	rtnl_unlock();
4045}
4046
4047static void addrconf_verify(void)
4048{
4049	mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4050}
4051
4052static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4053				     struct in6_addr **peer_pfx)
4054{
4055	struct in6_addr *pfx = NULL;
4056
4057	*peer_pfx = NULL;
4058
4059	if (addr)
4060		pfx = nla_data(addr);
4061
4062	if (local) {
4063		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4064			*peer_pfx = pfx;
4065		pfx = nla_data(local);
4066	}
4067
4068	return pfx;
4069}
4070
4071static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4072	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
4073	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
4074	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
4075	[IFA_FLAGS]		= { .len = sizeof(u32) },
4076};
4077
4078static int
4079inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4080{
4081	struct net *net = sock_net(skb->sk);
4082	struct ifaddrmsg *ifm;
4083	struct nlattr *tb[IFA_MAX+1];
4084	struct in6_addr *pfx, *peer_pfx;
4085	u32 ifa_flags;
4086	int err;
4087
4088	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4089	if (err < 0)
4090		return err;
4091
4092	ifm = nlmsg_data(nlh);
4093	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4094	if (!pfx)
4095		return -EINVAL;
4096
4097	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4098
4099	/* We ignore other flags so far. */
4100	ifa_flags &= IFA_F_MANAGETEMPADDR;
4101
4102	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4103			      ifm->ifa_prefixlen);
4104}
4105
4106static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4107			     u32 prefered_lft, u32 valid_lft)
4108{
4109	u32 flags;
4110	clock_t expires;
4111	unsigned long timeout;
4112	bool was_managetempaddr;
4113	bool had_prefixroute;
4114
4115	ASSERT_RTNL();
4116
4117	if (!valid_lft || (prefered_lft > valid_lft))
4118		return -EINVAL;
4119
4120	if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4121	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4122		return -EINVAL;
4123
4124	timeout = addrconf_timeout_fixup(valid_lft, HZ);
4125	if (addrconf_finite_timeout(timeout)) {
4126		expires = jiffies_to_clock_t(timeout * HZ);
4127		valid_lft = timeout;
4128		flags = RTF_EXPIRES;
4129	} else {
4130		expires = 0;
4131		flags = 0;
4132		ifa_flags |= IFA_F_PERMANENT;
4133	}
4134
4135	timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4136	if (addrconf_finite_timeout(timeout)) {
4137		if (timeout == 0)
4138			ifa_flags |= IFA_F_DEPRECATED;
4139		prefered_lft = timeout;
4140	}
4141
4142	spin_lock_bh(&ifp->lock);
4143	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4144	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4145			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
4146	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4147			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4148			IFA_F_NOPREFIXROUTE);
4149	ifp->flags |= ifa_flags;
4150	ifp->tstamp = jiffies;
4151	ifp->valid_lft = valid_lft;
4152	ifp->prefered_lft = prefered_lft;
4153
4154	spin_unlock_bh(&ifp->lock);
4155	if (!(ifp->flags&IFA_F_TENTATIVE))
4156		ipv6_ifa_notify(0, ifp);
4157
4158	if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4159		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4160				      expires, flags);
4161	} else if (had_prefixroute) {
4162		enum cleanup_prefix_rt_t action;
4163		unsigned long rt_expires;
4164
4165		write_lock_bh(&ifp->idev->lock);
4166		action = check_cleanup_prefix_route(ifp, &rt_expires);
4167		write_unlock_bh(&ifp->idev->lock);
4168
4169		if (action != CLEANUP_PREFIX_RT_NOP) {
4170			cleanup_prefix_route(ifp, rt_expires,
4171				action == CLEANUP_PREFIX_RT_DEL);
4172		}
4173	}
4174
4175	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4176		if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4177			valid_lft = prefered_lft = 0;
4178		manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4179				 !was_managetempaddr, jiffies);
4180	}
4181
4182	addrconf_verify_rtnl();
4183
4184	return 0;
4185}
4186
4187static int
4188inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4189{
4190	struct net *net = sock_net(skb->sk);
4191	struct ifaddrmsg *ifm;
4192	struct nlattr *tb[IFA_MAX+1];
4193	struct in6_addr *pfx, *peer_pfx;
4194	struct inet6_ifaddr *ifa;
4195	struct net_device *dev;
4196	u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4197	u32 ifa_flags;
4198	int err;
4199
4200	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4201	if (err < 0)
4202		return err;
4203
4204	ifm = nlmsg_data(nlh);
4205	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4206	if (!pfx)
4207		return -EINVAL;
4208
4209	if (tb[IFA_CACHEINFO]) {
4210		struct ifa_cacheinfo *ci;
4211
4212		ci = nla_data(tb[IFA_CACHEINFO]);
4213		valid_lft = ci->ifa_valid;
4214		preferred_lft = ci->ifa_prefered;
4215	} else {
4216		preferred_lft = INFINITY_LIFE_TIME;
4217		valid_lft = INFINITY_LIFE_TIME;
4218	}
4219
4220	dev =  __dev_get_by_index(net, ifm->ifa_index);
4221	if (!dev)
4222		return -ENODEV;
4223
4224	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4225
4226	/* We ignore other flags so far. */
4227	ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4228		     IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4229
4230	ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4231	if (!ifa) {
4232		/*
4233		 * It would be best to check for !NLM_F_CREATE here but
4234		 * userspace already relies on not having to provide this.
4235		 */
4236		return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4237				      ifm->ifa_prefixlen, ifa_flags,
4238				      preferred_lft, valid_lft);
4239	}
4240
4241	if (nlh->nlmsg_flags & NLM_F_EXCL ||
4242	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
4243		err = -EEXIST;
4244	else
4245		err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4246
4247	in6_ifa_put(ifa);
4248
4249	return err;
4250}
4251
4252static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4253			  u8 scope, int ifindex)
4254{
4255	struct ifaddrmsg *ifm;
4256
4257	ifm = nlmsg_data(nlh);
4258	ifm->ifa_family = AF_INET6;
4259	ifm->ifa_prefixlen = prefixlen;
4260	ifm->ifa_flags = flags;
4261	ifm->ifa_scope = scope;
4262	ifm->ifa_index = ifindex;
4263}
4264
4265static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4266			 unsigned long tstamp, u32 preferred, u32 valid)
4267{
4268	struct ifa_cacheinfo ci;
4269
4270	ci.cstamp = cstamp_delta(cstamp);
4271	ci.tstamp = cstamp_delta(tstamp);
4272	ci.ifa_prefered = preferred;
4273	ci.ifa_valid = valid;
4274
4275	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4276}
4277
4278static inline int rt_scope(int ifa_scope)
4279{
4280	if (ifa_scope & IFA_HOST)
4281		return RT_SCOPE_HOST;
4282	else if (ifa_scope & IFA_LINK)
4283		return RT_SCOPE_LINK;
4284	else if (ifa_scope & IFA_SITE)
4285		return RT_SCOPE_SITE;
4286	else
4287		return RT_SCOPE_UNIVERSE;
4288}
4289
4290static inline int inet6_ifaddr_msgsize(void)
4291{
4292	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4293	       + nla_total_size(16) /* IFA_LOCAL */
4294	       + nla_total_size(16) /* IFA_ADDRESS */
4295	       + nla_total_size(sizeof(struct ifa_cacheinfo))
4296	       + nla_total_size(4)  /* IFA_FLAGS */;
4297}
4298
4299static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4300			     u32 portid, u32 seq, int event, unsigned int flags)
4301{
4302	struct nlmsghdr  *nlh;
4303	u32 preferred, valid;
4304
4305	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4306	if (!nlh)
4307		return -EMSGSIZE;
4308
4309	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4310		      ifa->idev->dev->ifindex);
4311
4312	if (!((ifa->flags&IFA_F_PERMANENT) &&
4313	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4314		preferred = ifa->prefered_lft;
4315		valid = ifa->valid_lft;
4316		if (preferred != INFINITY_LIFE_TIME) {
4317			long tval = (jiffies - ifa->tstamp)/HZ;
4318			if (preferred > tval)
4319				preferred -= tval;
4320			else
4321				preferred = 0;
4322			if (valid != INFINITY_LIFE_TIME) {
4323				if (valid > tval)
4324					valid -= tval;
4325				else
4326					valid = 0;
4327			}
4328		}
4329	} else {
4330		preferred = INFINITY_LIFE_TIME;
4331		valid = INFINITY_LIFE_TIME;
4332	}
4333
4334	if (!ipv6_addr_any(&ifa->peer_addr)) {
4335		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4336		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4337			goto error;
4338	} else
4339		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4340			goto error;
4341
4342	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4343		goto error;
4344
4345	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4346		goto error;
4347
4348	nlmsg_end(skb, nlh);
4349	return 0;
4350
4351error:
4352	nlmsg_cancel(skb, nlh);
4353	return -EMSGSIZE;
4354}
4355
4356static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4357				u32 portid, u32 seq, int event, u16 flags)
4358{
4359	struct nlmsghdr  *nlh;
4360	u8 scope = RT_SCOPE_UNIVERSE;
4361	int ifindex = ifmca->idev->dev->ifindex;
4362
4363	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4364		scope = RT_SCOPE_SITE;
4365
4366	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4367	if (!nlh)
4368		return -EMSGSIZE;
4369
4370	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4371	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4372	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4373			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4374		nlmsg_cancel(skb, nlh);
4375		return -EMSGSIZE;
4376	}
4377
4378	nlmsg_end(skb, nlh);
4379	return 0;
4380}
4381
4382static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4383				u32 portid, u32 seq, int event, unsigned int flags)
4384{
4385	struct nlmsghdr  *nlh;
4386	u8 scope = RT_SCOPE_UNIVERSE;
4387	int ifindex = ifaca->aca_idev->dev->ifindex;
4388
4389	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4390		scope = RT_SCOPE_SITE;
4391
4392	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4393	if (!nlh)
4394		return -EMSGSIZE;
4395
4396	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4397	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4398	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4399			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4400		nlmsg_cancel(skb, nlh);
4401		return -EMSGSIZE;
4402	}
4403
4404	nlmsg_end(skb, nlh);
4405	return 0;
4406}
4407
4408enum addr_type_t {
4409	UNICAST_ADDR,
4410	MULTICAST_ADDR,
4411	ANYCAST_ADDR,
4412};
4413
4414/* called with rcu_read_lock() */
4415static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4416			  struct netlink_callback *cb, enum addr_type_t type,
4417			  int s_ip_idx, int *p_ip_idx)
4418{
4419	struct ifmcaddr6 *ifmca;
4420	struct ifacaddr6 *ifaca;
4421	int err = 1;
4422	int ip_idx = *p_ip_idx;
4423
4424	read_lock_bh(&idev->lock);
4425	switch (type) {
4426	case UNICAST_ADDR: {
4427		struct inet6_ifaddr *ifa;
4428
4429		/* unicast address incl. temp addr */
4430		list_for_each_entry(ifa, &idev->addr_list, if_list) {
4431			if (++ip_idx < s_ip_idx)
4432				continue;
4433			err = inet6_fill_ifaddr(skb, ifa,
4434						NETLINK_CB(cb->skb).portid,
4435						cb->nlh->nlmsg_seq,
4436						RTM_NEWADDR,
4437						NLM_F_MULTI);
4438			if (err < 0)
4439				break;
4440			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4441		}
4442		break;
4443	}
4444	case MULTICAST_ADDR:
4445		/* multicast address */
4446		for (ifmca = idev->mc_list; ifmca;
4447		     ifmca = ifmca->next, ip_idx++) {
4448			if (ip_idx < s_ip_idx)
4449				continue;
4450			err = inet6_fill_ifmcaddr(skb, ifmca,
4451						  NETLINK_CB(cb->skb).portid,
4452						  cb->nlh->nlmsg_seq,
4453						  RTM_GETMULTICAST,
4454						  NLM_F_MULTI);
4455			if (err < 0)
4456				break;
4457		}
4458		break;
4459	case ANYCAST_ADDR:
4460		/* anycast address */
4461		for (ifaca = idev->ac_list; ifaca;
4462		     ifaca = ifaca->aca_next, ip_idx++) {
4463			if (ip_idx < s_ip_idx)
4464				continue;
4465			err = inet6_fill_ifacaddr(skb, ifaca,
4466						  NETLINK_CB(cb->skb).portid,
4467						  cb->nlh->nlmsg_seq,
4468						  RTM_GETANYCAST,
4469						  NLM_F_MULTI);
4470			if (err < 0)
4471				break;
4472		}
4473		break;
4474	default:
4475		break;
4476	}
4477	read_unlock_bh(&idev->lock);
4478	*p_ip_idx = ip_idx;
4479	return err;
4480}
4481
4482static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4483			   enum addr_type_t type)
4484{
4485	struct net *net = sock_net(skb->sk);
4486	int h, s_h;
4487	int idx, ip_idx;
4488	int s_idx, s_ip_idx;
4489	struct net_device *dev;
4490	struct inet6_dev *idev;
4491	struct hlist_head *head;
4492
4493	s_h = cb->args[0];
4494	s_idx = idx = cb->args[1];
4495	s_ip_idx = ip_idx = cb->args[2];
4496
4497	rcu_read_lock();
4498	cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4499	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4500		idx = 0;
4501		head = &net->dev_index_head[h];
4502		hlist_for_each_entry_rcu(dev, head, index_hlist) {
4503			if (idx < s_idx)
4504				goto cont;
4505			if (h > s_h || idx > s_idx)
4506				s_ip_idx = 0;
4507			ip_idx = 0;
4508			idev = __in6_dev_get(dev);
4509			if (!idev)
4510				goto cont;
4511
4512			if (in6_dump_addrs(idev, skb, cb, type,
4513					   s_ip_idx, &ip_idx) < 0)
4514				goto done;
4515cont:
4516			idx++;
4517		}
4518	}
4519done:
4520	rcu_read_unlock();
4521	cb->args[0] = h;
4522	cb->args[1] = idx;
4523	cb->args[2] = ip_idx;
4524
4525	return skb->len;
4526}
4527
4528static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4529{
4530	enum addr_type_t type = UNICAST_ADDR;
4531
4532	return inet6_dump_addr(skb, cb, type);
4533}
4534
4535static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4536{
4537	enum addr_type_t type = MULTICAST_ADDR;
4538
4539	return inet6_dump_addr(skb, cb, type);
4540}
4541
4542
4543static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4544{
4545	enum addr_type_t type = ANYCAST_ADDR;
4546
4547	return inet6_dump_addr(skb, cb, type);
4548}
4549
4550static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4551{
4552	struct net *net = sock_net(in_skb->sk);
4553	struct ifaddrmsg *ifm;
4554	struct nlattr *tb[IFA_MAX+1];
4555	struct in6_addr *addr = NULL, *peer;
4556	struct net_device *dev = NULL;
4557	struct inet6_ifaddr *ifa;
4558	struct sk_buff *skb;
4559	int err;
4560
4561	err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4562	if (err < 0)
4563		goto errout;
4564
4565	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4566	if (!addr) {
4567		err = -EINVAL;
4568		goto errout;
4569	}
4570
4571	ifm = nlmsg_data(nlh);
4572	if (ifm->ifa_index)
4573		dev = __dev_get_by_index(net, ifm->ifa_index);
4574
4575	ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4576	if (!ifa) {
4577		err = -EADDRNOTAVAIL;
4578		goto errout;
4579	}
4580
4581	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4582	if (!skb) {
4583		err = -ENOBUFS;
4584		goto errout_ifa;
4585	}
4586
4587	err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4588				nlh->nlmsg_seq, RTM_NEWADDR, 0);
4589	if (err < 0) {
4590		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4591		WARN_ON(err == -EMSGSIZE);
4592		kfree_skb(skb);
4593		goto errout_ifa;
4594	}
4595	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4596errout_ifa:
4597	in6_ifa_put(ifa);
4598errout:
4599	return err;
4600}
4601
4602static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4603{
4604	struct sk_buff *skb;
4605	struct net *net = dev_net(ifa->idev->dev);
4606	int err = -ENOBUFS;
4607
4608	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4609	if (!skb)
4610		goto errout;
4611
4612	err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4613	if (err < 0) {
4614		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4615		WARN_ON(err == -EMSGSIZE);
4616		kfree_skb(skb);
4617		goto errout;
4618	}
4619	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4620	return;
4621errout:
4622	if (err < 0)
4623		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4624}
4625
4626static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4627				__s32 *array, int bytes)
4628{
4629	BUG_ON(bytes < (DEVCONF_MAX * 4));
4630
4631	memset(array, 0, bytes);
4632	array[DEVCONF_FORWARDING] = cnf->forwarding;
4633	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4634	array[DEVCONF_MTU6] = cnf->mtu6;
4635	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4636	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4637	array[DEVCONF_AUTOCONF] = cnf->autoconf;
4638	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4639	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4640	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4641		jiffies_to_msecs(cnf->rtr_solicit_interval);
4642	array[DEVCONF_RTR_SOLICIT_DELAY] =
4643		jiffies_to_msecs(cnf->rtr_solicit_delay);
4644	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4645	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4646		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4647	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4648		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4649	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4650	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4651	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4652	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4653	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4654	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4655	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4656	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4657	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4658#ifdef CONFIG_IPV6_ROUTER_PREF
4659	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4660	array[DEVCONF_RTR_PROBE_INTERVAL] =
4661		jiffies_to_msecs(cnf->rtr_probe_interval);
4662#ifdef CONFIG_IPV6_ROUTE_INFO
4663	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4664#endif
4665#endif
4666	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4667	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4668#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4669	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4670	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4671#endif
4672#ifdef CONFIG_IPV6_MROUTE
4673	array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4674#endif
4675	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4676	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4677	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4678	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4679	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4680	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4681	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4682	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4683	/* we omit DEVCONF_STABLE_SECRET for now */
4684	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4685}
4686
4687static inline size_t inet6_ifla6_size(void)
4688{
4689	return nla_total_size(4) /* IFLA_INET6_FLAGS */
4690	     + nla_total_size(sizeof(struct ifla_cacheinfo))
4691	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4692	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4693	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4694	     + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4695}
4696
4697static inline size_t inet6_if_nlmsg_size(void)
4698{
4699	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4700	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4701	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4702	       + nla_total_size(4) /* IFLA_MTU */
4703	       + nla_total_size(4) /* IFLA_LINK */
4704	       + nla_total_size(1) /* IFLA_OPERSTATE */
4705	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4706}
4707
4708static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4709				      int items, int bytes)
4710{
4711	int i;
4712	int pad = bytes - sizeof(u64) * items;
4713	BUG_ON(pad < 0);
4714
4715	/* Use put_unaligned() because stats may not be aligned for u64. */
4716	put_unaligned(items, &stats[0]);
4717	for (i = 1; i < items; i++)
4718		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4719
4720	memset(&stats[items], 0, pad);
4721}
4722
4723static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4724					int bytes, size_t syncpoff)
4725{
4726	int i, c;
4727	u64 buff[IPSTATS_MIB_MAX];
4728	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4729
4730	BUG_ON(pad < 0);
4731
4732	memset(buff, 0, sizeof(buff));
4733	buff[0] = IPSTATS_MIB_MAX;
4734
4735	for_each_possible_cpu(c) {
4736		for (i = 1; i < IPSTATS_MIB_MAX; i++)
4737			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4738	}
4739
4740	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4741	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4742}
4743
4744static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4745			     int bytes)
4746{
4747	switch (attrtype) {
4748	case IFLA_INET6_STATS:
4749		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4750				     offsetof(struct ipstats_mib, syncp));
4751		break;
4752	case IFLA_INET6_ICMP6STATS:
4753		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4754		break;
4755	}
4756}
4757
4758static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4759				  u32 ext_filter_mask)
4760{
4761	struct nlattr *nla;
4762	struct ifla_cacheinfo ci;
4763
4764	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4765		goto nla_put_failure;
4766	ci.max_reasm_len = IPV6_MAXPLEN;
4767	ci.tstamp = cstamp_delta(idev->tstamp);
4768	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4769	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4770	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4771		goto nla_put_failure;
4772	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4773	if (!nla)
4774		goto nla_put_failure;
4775	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4776
4777	/* XXX - MC not implemented */
4778
4779	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4780		return 0;
4781
4782	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4783	if (!nla)
4784		goto nla_put_failure;
4785	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4786
4787	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4788	if (!nla)
4789		goto nla_put_failure;
4790	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4791
4792	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4793	if (!nla)
4794		goto nla_put_failure;
4795
4796	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4797		goto nla_put_failure;
4798
4799	read_lock_bh(&idev->lock);
4800	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4801	read_unlock_bh(&idev->lock);
4802
4803	return 0;
4804
4805nla_put_failure:
4806	return -EMSGSIZE;
4807}
4808
4809static size_t inet6_get_link_af_size(const struct net_device *dev,
4810				     u32 ext_filter_mask)
4811{
4812	if (!__in6_dev_get(dev))
4813		return 0;
4814
4815	return inet6_ifla6_size();
4816}
4817
4818static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4819			      u32 ext_filter_mask)
4820{
4821	struct inet6_dev *idev = __in6_dev_get(dev);
4822
4823	if (!idev)
4824		return -ENODATA;
4825
4826	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4827		return -EMSGSIZE;
4828
4829	return 0;
4830}
4831
4832static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4833{
4834	struct inet6_ifaddr *ifp;
4835	struct net_device *dev = idev->dev;
4836	bool update_rs = false;
4837	struct in6_addr ll_addr;
4838
4839	ASSERT_RTNL();
4840
4841	if (!token)
4842		return -EINVAL;
4843	if (ipv6_addr_any(token))
4844		return -EINVAL;
4845	if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4846		return -EINVAL;
4847	if (!ipv6_accept_ra(idev))
4848		return -EINVAL;
4849	if (idev->cnf.rtr_solicits <= 0)
4850		return -EINVAL;
4851
4852	write_lock_bh(&idev->lock);
4853
4854	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4855	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4856
4857	write_unlock_bh(&idev->lock);
4858
4859	if (!idev->dead && (idev->if_flags & IF_READY) &&
4860	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4861			     IFA_F_OPTIMISTIC)) {
4862
4863		/* If we're not ready, then normal ifup will take care
4864		 * of this. Otherwise, we need to request our rs here.
4865		 */
4866		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4867		update_rs = true;
4868	}
4869
4870	write_lock_bh(&idev->lock);
4871
4872	if (update_rs) {
4873		idev->if_flags |= IF_RS_SENT;
4874		idev->rs_probes = 1;
4875		addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4876	}
4877
4878	/* Well, that's kinda nasty ... */
4879	list_for_each_entry(ifp, &idev->addr_list, if_list) {
4880		spin_lock(&ifp->lock);
4881		if (ifp->tokenized) {
4882			ifp->valid_lft = 0;
4883			ifp->prefered_lft = 0;
4884		}
4885		spin_unlock(&ifp->lock);
4886	}
4887
4888	write_unlock_bh(&idev->lock);
4889	inet6_ifinfo_notify(RTM_NEWLINK, idev);
4890	addrconf_verify_rtnl();
4891	return 0;
4892}
4893
4894static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4895	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
4896	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
4897};
4898
4899static int inet6_validate_link_af(const struct net_device *dev,
4900				  const struct nlattr *nla)
4901{
4902	struct nlattr *tb[IFLA_INET6_MAX + 1];
4903
4904	if (dev && !__in6_dev_get(dev))
4905		return -EAFNOSUPPORT;
4906
4907	return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4908}
4909
4910static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4911{
4912	int err = -EINVAL;
4913	struct inet6_dev *idev = __in6_dev_get(dev);
4914	struct nlattr *tb[IFLA_INET6_MAX + 1];
4915
4916	if (!idev)
4917		return -EAFNOSUPPORT;
4918
4919	if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4920		BUG();
4921
4922	if (tb[IFLA_INET6_TOKEN]) {
4923		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4924		if (err)
4925			return err;
4926	}
4927
4928	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4929		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4930
4931		if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4932		    mode != IN6_ADDR_GEN_MODE_NONE &&
4933		    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4934			return -EINVAL;
4935
4936		if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4937		    !idev->cnf.stable_secret.initialized &&
4938		    !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4939			return -EINVAL;
4940
4941		idev->addr_gen_mode = mode;
4942		err = 0;
4943	}
4944
4945	return err;
4946}
4947
4948static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4949			     u32 portid, u32 seq, int event, unsigned int flags)
4950{
4951	struct net_device *dev = idev->dev;
4952	struct ifinfomsg *hdr;
4953	struct nlmsghdr *nlh;
4954	void *protoinfo;
4955
4956	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4957	if (!nlh)
4958		return -EMSGSIZE;
4959
4960	hdr = nlmsg_data(nlh);
4961	hdr->ifi_family = AF_INET6;
4962	hdr->__ifi_pad = 0;
4963	hdr->ifi_type = dev->type;
4964	hdr->ifi_index = dev->ifindex;
4965	hdr->ifi_flags = dev_get_flags(dev);
4966	hdr->ifi_change = 0;
4967
4968	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4969	    (dev->addr_len &&
4970	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4971	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4972	    (dev->ifindex != dev_get_iflink(dev) &&
4973	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
4974	    nla_put_u8(skb, IFLA_OPERSTATE,
4975		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
4976		goto nla_put_failure;
4977	protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4978	if (!protoinfo)
4979		goto nla_put_failure;
4980
4981	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
4982		goto nla_put_failure;
4983
4984	nla_nest_end(skb, protoinfo);
4985	nlmsg_end(skb, nlh);
4986	return 0;
4987
4988nla_put_failure:
4989	nlmsg_cancel(skb, nlh);
4990	return -EMSGSIZE;
4991}
4992
4993static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4994{
4995	struct net *net = sock_net(skb->sk);
4996	int h, s_h;
4997	int idx = 0, s_idx;
4998	struct net_device *dev;
4999	struct inet6_dev *idev;
5000	struct hlist_head *head;
5001
5002	s_h = cb->args[0];
5003	s_idx = cb->args[1];
5004
5005	rcu_read_lock();
5006	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5007		idx = 0;
5008		head = &net->dev_index_head[h];
5009		hlist_for_each_entry_rcu(dev, head, index_hlist) {
5010			if (idx < s_idx)
5011				goto cont;
5012			idev = __in6_dev_get(dev);
5013			if (!idev)
5014				goto cont;
5015			if (inet6_fill_ifinfo(skb, idev,
5016					      NETLINK_CB(cb->skb).portid,
5017					      cb->nlh->nlmsg_seq,
5018					      RTM_NEWLINK, NLM_F_MULTI) < 0)
5019				goto out;
5020cont:
5021			idx++;
5022		}
5023	}
5024out:
5025	rcu_read_unlock();
5026	cb->args[1] = idx;
5027	cb->args[0] = h;
5028
5029	return skb->len;
5030}
5031
5032void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5033{
5034	struct sk_buff *skb;
5035	struct net *net = dev_net(idev->dev);
5036	int err = -ENOBUFS;
5037
5038	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5039	if (!skb)
5040		goto errout;
5041
5042	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5043	if (err < 0) {
5044		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5045		WARN_ON(err == -EMSGSIZE);
5046		kfree_skb(skb);
5047		goto errout;
5048	}
5049	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5050	return;
5051errout:
5052	if (err < 0)
5053		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5054}
5055
5056static inline size_t inet6_prefix_nlmsg_size(void)
5057{
5058	return NLMSG_ALIGN(sizeof(struct prefixmsg))
5059	       + nla_total_size(sizeof(struct in6_addr))
5060	       + nla_total_size(sizeof(struct prefix_cacheinfo));
5061}
5062
5063static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5064			     struct prefix_info *pinfo, u32 portid, u32 seq,
5065			     int event, unsigned int flags)
5066{
5067	struct prefixmsg *pmsg;
5068	struct nlmsghdr *nlh;
5069	struct prefix_cacheinfo	ci;
5070
5071	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5072	if (!nlh)
5073		return -EMSGSIZE;
5074
5075	pmsg = nlmsg_data(nlh);
5076	pmsg->prefix_family = AF_INET6;
5077	pmsg->prefix_pad1 = 0;
5078	pmsg->prefix_pad2 = 0;
5079	pmsg->prefix_ifindex = idev->dev->ifindex;
5080	pmsg->prefix_len = pinfo->prefix_len;
5081	pmsg->prefix_type = pinfo->type;
5082	pmsg->prefix_pad3 = 0;
5083	pmsg->prefix_flags = 0;
5084	if (pinfo->onlink)
5085		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5086	if (pinfo->autoconf)
5087		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5088
5089	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5090		goto nla_put_failure;
5091	ci.preferred_time = ntohl(pinfo->prefered);
5092	ci.valid_time = ntohl(pinfo->valid);
5093	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5094		goto nla_put_failure;
5095	nlmsg_end(skb, nlh);
5096	return 0;
5097
5098nla_put_failure:
5099	nlmsg_cancel(skb, nlh);
5100	return -EMSGSIZE;
5101}
5102
5103static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5104			 struct prefix_info *pinfo)
5105{
5106	struct sk_buff *skb;
5107	struct net *net = dev_net(idev->dev);
5108	int err = -ENOBUFS;
5109
5110	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5111	if (!skb)
5112		goto errout;
5113
5114	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5115	if (err < 0) {
5116		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5117		WARN_ON(err == -EMSGSIZE);
5118		kfree_skb(skb);
5119		goto errout;
5120	}
5121	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5122	return;
5123errout:
5124	if (err < 0)
5125		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5126}
5127
5128static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5129{
5130	struct net *net = dev_net(ifp->idev->dev);
5131
5132	if (event)
5133		ASSERT_RTNL();
5134
5135	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5136
5137	switch (event) {
5138	case RTM_NEWADDR:
5139		/*
5140		 * If the address was optimistic
5141		 * we inserted the route at the start of
5142		 * our DAD process, so we don't need
5143		 * to do it again
5144		 */
5145		if (!(ifp->rt->rt6i_node))
5146			ip6_ins_rt(ifp->rt);
5147		if (ifp->idev->cnf.forwarding)
5148			addrconf_join_anycast(ifp);
5149		if (!ipv6_addr_any(&ifp->peer_addr))
5150			addrconf_prefix_route(&ifp->peer_addr, 128,
5151					      ifp->idev->dev, 0, 0);
5152		break;
5153	case RTM_DELADDR:
5154		if (ifp->idev->cnf.forwarding)
5155			addrconf_leave_anycast(ifp);
5156		addrconf_leave_solict(ifp->idev, &ifp->addr);
5157		if (!ipv6_addr_any(&ifp->peer_addr)) {
5158			struct rt6_info *rt;
5159
5160			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5161						       ifp->idev->dev, 0, 0);
5162			if (rt)
5163				ip6_del_rt(rt);
5164		}
5165		dst_hold(&ifp->rt->dst);
5166
5167		ip6_del_rt(ifp->rt);
5168
5169		rt_genid_bump_ipv6(net);
5170		break;
5171	}
5172	atomic_inc(&net->ipv6.dev_addr_genid);
5173}
5174
5175static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5176{
5177	rcu_read_lock_bh();
5178	if (likely(ifp->idev->dead == 0))
5179		__ipv6_ifa_notify(event, ifp);
5180	rcu_read_unlock_bh();
5181}
5182
5183#ifdef CONFIG_SYSCTL
5184
5185static
5186int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5187			   void __user *buffer, size_t *lenp, loff_t *ppos)
5188{
5189	int *valp = ctl->data;
5190	int val = *valp;
5191	loff_t pos = *ppos;
5192	struct ctl_table lctl;
5193	int ret;
5194
5195	/*
5196	 * ctl->data points to idev->cnf.forwarding, we should
5197	 * not modify it until we get the rtnl lock.
5198	 */
5199	lctl = *ctl;
5200	lctl.data = &val;
5201
5202	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5203
5204	if (write)
5205		ret = addrconf_fixup_forwarding(ctl, valp, val);
5206	if (ret)
5207		*ppos = pos;
5208	return ret;
5209}
5210
5211static
5212int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5213			void __user *buffer, size_t *lenp, loff_t *ppos)
5214{
5215	struct inet6_dev *idev = ctl->extra1;
5216	int min_mtu = IPV6_MIN_MTU;
5217	struct ctl_table lctl;
5218
5219	lctl = *ctl;
5220	lctl.extra1 = &min_mtu;
5221	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5222
5223	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5224}
5225
5226static void dev_disable_change(struct inet6_dev *idev)
5227{
5228	struct netdev_notifier_info info;
5229
5230	if (!idev || !idev->dev)
5231		return;
5232
5233	netdev_notifier_info_init(&info, idev->dev);
5234	if (idev->cnf.disable_ipv6)
5235		addrconf_notify(NULL, NETDEV_DOWN, &info);
5236	else
5237		addrconf_notify(NULL, NETDEV_UP, &info);
5238}
5239
5240static void addrconf_disable_change(struct net *net, __s32 newf)
5241{
5242	struct net_device *dev;
5243	struct inet6_dev *idev;
5244
5245	rcu_read_lock();
5246	for_each_netdev_rcu(net, dev) {
5247		idev = __in6_dev_get(dev);
5248		if (idev) {
5249			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5250			idev->cnf.disable_ipv6 = newf;
5251			if (changed)
5252				dev_disable_change(idev);
5253		}
5254	}
5255	rcu_read_unlock();
5256}
5257
5258static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5259{
5260	struct net *net;
5261	int old;
5262
5263	if (!rtnl_trylock())
5264		return restart_syscall();
5265
5266	net = (struct net *)table->extra2;
5267	old = *p;
5268	*p = newf;
5269
5270	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5271		rtnl_unlock();
5272		return 0;
5273	}
5274
5275	if (p == &net->ipv6.devconf_all->disable_ipv6) {
5276		net->ipv6.devconf_dflt->disable_ipv6 = newf;
5277		addrconf_disable_change(net, newf);
5278	} else if ((!newf) ^ (!old))
5279		dev_disable_change((struct inet6_dev *)table->extra1);
5280
5281	rtnl_unlock();
5282	return 0;
5283}
5284
5285static
5286int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5287			    void __user *buffer, size_t *lenp, loff_t *ppos)
5288{
5289	int *valp = ctl->data;
5290	int val = *valp;
5291	loff_t pos = *ppos;
5292	struct ctl_table lctl;
5293	int ret;
5294
5295	/*
5296	 * ctl->data points to idev->cnf.disable_ipv6, we should
5297	 * not modify it until we get the rtnl lock.
5298	 */
5299	lctl = *ctl;
5300	lctl.data = &val;
5301
5302	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5303
5304	if (write)
5305		ret = addrconf_disable_ipv6(ctl, valp, val);
5306	if (ret)
5307		*ppos = pos;
5308	return ret;
5309}
5310
5311static
5312int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5313			      void __user *buffer, size_t *lenp, loff_t *ppos)
5314{
5315	int *valp = ctl->data;
5316	int ret;
5317	int old, new;
5318
5319	old = *valp;
5320	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5321	new = *valp;
5322
5323	if (write && old != new) {
5324		struct net *net = ctl->extra2;
5325
5326		if (!rtnl_trylock())
5327			return restart_syscall();
5328
5329		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5330			inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5331						     NETCONFA_IFINDEX_DEFAULT,
5332						     net->ipv6.devconf_dflt);
5333		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5334			inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5335						     NETCONFA_IFINDEX_ALL,
5336						     net->ipv6.devconf_all);
5337		else {
5338			struct inet6_dev *idev = ctl->extra1;
5339
5340			inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5341						     idev->dev->ifindex,
5342						     &idev->cnf);
5343		}
5344		rtnl_unlock();
5345	}
5346
5347	return ret;
5348}
5349
5350static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5351					 void __user *buffer, size_t *lenp,
5352					 loff_t *ppos)
5353{
5354	int err;
5355	struct in6_addr addr;
5356	char str[IPV6_MAX_STRLEN];
5357	struct ctl_table lctl = *ctl;
5358	struct net *net = ctl->extra2;
5359	struct ipv6_stable_secret *secret = ctl->data;
5360
5361	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5362		return -EIO;
5363
5364	lctl.maxlen = IPV6_MAX_STRLEN;
5365	lctl.data = str;
5366
5367	if (!rtnl_trylock())
5368		return restart_syscall();
5369
5370	if (!write && !secret->initialized) {
5371		err = -EIO;
5372		goto out;
5373	}
5374
5375	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5376	if (err >= sizeof(str)) {
5377		err = -EIO;
5378		goto out;
5379	}
5380
5381	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5382	if (err || !write)
5383		goto out;
5384
5385	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5386		err = -EIO;
5387		goto out;
5388	}
5389
5390	secret->initialized = true;
5391	secret->secret = addr;
5392
5393	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5394		struct net_device *dev;
5395
5396		for_each_netdev(net, dev) {
5397			struct inet6_dev *idev = __in6_dev_get(dev);
5398
5399			if (idev) {
5400				idev->addr_gen_mode =
5401					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5402			}
5403		}
5404	} else {
5405		struct inet6_dev *idev = ctl->extra1;
5406
5407		idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5408	}
5409
5410out:
5411	rtnl_unlock();
5412
5413	return err;
5414}
5415
5416static
5417int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5418						int write,
5419						void __user *buffer,
5420						size_t *lenp,
5421						loff_t *ppos)
5422{
5423	int *valp = ctl->data;
5424	int val = *valp;
5425	loff_t pos = *ppos;
5426	struct ctl_table lctl;
5427	int ret;
5428
5429	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5430	 * we should not modify it until we get the rtnl lock.
5431	 */
5432	lctl = *ctl;
5433	lctl.data = &val;
5434
5435	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5436
5437	if (write)
5438		ret = addrconf_fixup_linkdown(ctl, valp, val);
5439	if (ret)
5440		*ppos = pos;
5441	return ret;
5442}
5443
5444static struct addrconf_sysctl_table
5445{
5446	struct ctl_table_header *sysctl_header;
5447	struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5448} addrconf_sysctl __read_mostly = {
5449	.sysctl_header = NULL,
5450	.addrconf_vars = {
5451		{
5452			.procname	= "forwarding",
5453			.data		= &ipv6_devconf.forwarding,
5454			.maxlen		= sizeof(int),
5455			.mode		= 0644,
5456			.proc_handler	= addrconf_sysctl_forward,
5457		},
5458		{
5459			.procname	= "hop_limit",
5460			.data		= &ipv6_devconf.hop_limit,
5461			.maxlen		= sizeof(int),
5462			.mode		= 0644,
5463			.proc_handler	= proc_dointvec,
5464		},
5465		{
5466			.procname	= "mtu",
5467			.data		= &ipv6_devconf.mtu6,
5468			.maxlen		= sizeof(int),
5469			.mode		= 0644,
5470			.proc_handler	= addrconf_sysctl_mtu,
5471		},
5472		{
5473			.procname	= "accept_ra",
5474			.data		= &ipv6_devconf.accept_ra,
5475			.maxlen		= sizeof(int),
5476			.mode		= 0644,
5477			.proc_handler	= proc_dointvec,
5478		},
5479		{
5480			.procname	= "accept_redirects",
5481			.data		= &ipv6_devconf.accept_redirects,
5482			.maxlen		= sizeof(int),
5483			.mode		= 0644,
5484			.proc_handler	= proc_dointvec,
5485		},
5486		{
5487			.procname	= "autoconf",
5488			.data		= &ipv6_devconf.autoconf,
5489			.maxlen		= sizeof(int),
5490			.mode		= 0644,
5491			.proc_handler	= proc_dointvec,
5492		},
5493		{
5494			.procname	= "dad_transmits",
5495			.data		= &ipv6_devconf.dad_transmits,
5496			.maxlen		= sizeof(int),
5497			.mode		= 0644,
5498			.proc_handler	= proc_dointvec,
5499		},
5500		{
5501			.procname	= "router_solicitations",
5502			.data		= &ipv6_devconf.rtr_solicits,
5503			.maxlen		= sizeof(int),
5504			.mode		= 0644,
5505			.proc_handler	= proc_dointvec,
5506		},
5507		{
5508			.procname	= "router_solicitation_interval",
5509			.data		= &ipv6_devconf.rtr_solicit_interval,
5510			.maxlen		= sizeof(int),
5511			.mode		= 0644,
5512			.proc_handler	= proc_dointvec_jiffies,
5513		},
5514		{
5515			.procname	= "router_solicitation_delay",
5516			.data		= &ipv6_devconf.rtr_solicit_delay,
5517			.maxlen		= sizeof(int),
5518			.mode		= 0644,
5519			.proc_handler	= proc_dointvec_jiffies,
5520		},
5521		{
5522			.procname	= "force_mld_version",
5523			.data		= &ipv6_devconf.force_mld_version,
5524			.maxlen		= sizeof(int),
5525			.mode		= 0644,
5526			.proc_handler	= proc_dointvec,
5527		},
5528		{
5529			.procname	= "mldv1_unsolicited_report_interval",
5530			.data		=
5531				&ipv6_devconf.mldv1_unsolicited_report_interval,
5532			.maxlen		= sizeof(int),
5533			.mode		= 0644,
5534			.proc_handler	= proc_dointvec_ms_jiffies,
5535		},
5536		{
5537			.procname	= "mldv2_unsolicited_report_interval",
5538			.data		=
5539				&ipv6_devconf.mldv2_unsolicited_report_interval,
5540			.maxlen		= sizeof(int),
5541			.mode		= 0644,
5542			.proc_handler	= proc_dointvec_ms_jiffies,
5543		},
5544		{
5545			.procname	= "use_tempaddr",
5546			.data		= &ipv6_devconf.use_tempaddr,
5547			.maxlen		= sizeof(int),
5548			.mode		= 0644,
5549			.proc_handler	= proc_dointvec,
5550		},
5551		{
5552			.procname	= "temp_valid_lft",
5553			.data		= &ipv6_devconf.temp_valid_lft,
5554			.maxlen		= sizeof(int),
5555			.mode		= 0644,
5556			.proc_handler	= proc_dointvec,
5557		},
5558		{
5559			.procname	= "temp_prefered_lft",
5560			.data		= &ipv6_devconf.temp_prefered_lft,
5561			.maxlen		= sizeof(int),
5562			.mode		= 0644,
5563			.proc_handler	= proc_dointvec,
5564		},
5565		{
5566			.procname	= "regen_max_retry",
5567			.data		= &ipv6_devconf.regen_max_retry,
5568			.maxlen		= sizeof(int),
5569			.mode		= 0644,
5570			.proc_handler	= proc_dointvec,
5571		},
5572		{
5573			.procname	= "max_desync_factor",
5574			.data		= &ipv6_devconf.max_desync_factor,
5575			.maxlen		= sizeof(int),
5576			.mode		= 0644,
5577			.proc_handler	= proc_dointvec,
5578		},
5579		{
5580			.procname	= "max_addresses",
5581			.data		= &ipv6_devconf.max_addresses,
5582			.maxlen		= sizeof(int),
5583			.mode		= 0644,
5584			.proc_handler	= proc_dointvec,
5585		},
5586		{
5587			.procname	= "accept_ra_defrtr",
5588			.data		= &ipv6_devconf.accept_ra_defrtr,
5589			.maxlen		= sizeof(int),
5590			.mode		= 0644,
5591			.proc_handler	= proc_dointvec,
5592		},
5593		{
5594			.procname	= "accept_ra_min_hop_limit",
5595			.data		= &ipv6_devconf.accept_ra_min_hop_limit,
5596			.maxlen		= sizeof(int),
5597			.mode		= 0644,
5598			.proc_handler	= proc_dointvec,
5599		},
5600		{
5601			.procname	= "accept_ra_pinfo",
5602			.data		= &ipv6_devconf.accept_ra_pinfo,
5603			.maxlen		= sizeof(int),
5604			.mode		= 0644,
5605			.proc_handler	= proc_dointvec,
5606		},
5607#ifdef CONFIG_IPV6_ROUTER_PREF
5608		{
5609			.procname	= "accept_ra_rtr_pref",
5610			.data		= &ipv6_devconf.accept_ra_rtr_pref,
5611			.maxlen		= sizeof(int),
5612			.mode		= 0644,
5613			.proc_handler	= proc_dointvec,
5614		},
5615		{
5616			.procname	= "router_probe_interval",
5617			.data		= &ipv6_devconf.rtr_probe_interval,
5618			.maxlen		= sizeof(int),
5619			.mode		= 0644,
5620			.proc_handler	= proc_dointvec_jiffies,
5621		},
5622#ifdef CONFIG_IPV6_ROUTE_INFO
5623		{
5624			.procname	= "accept_ra_rt_info_max_plen",
5625			.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
5626			.maxlen		= sizeof(int),
5627			.mode		= 0644,
5628			.proc_handler	= proc_dointvec,
5629		},
5630#endif
5631#endif
5632		{
5633			.procname	= "proxy_ndp",
5634			.data		= &ipv6_devconf.proxy_ndp,
5635			.maxlen		= sizeof(int),
5636			.mode		= 0644,
5637			.proc_handler	= addrconf_sysctl_proxy_ndp,
5638		},
5639		{
5640			.procname	= "accept_source_route",
5641			.data		= &ipv6_devconf.accept_source_route,
5642			.maxlen		= sizeof(int),
5643			.mode		= 0644,
5644			.proc_handler	= proc_dointvec,
5645		},
5646#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5647		{
5648			.procname       = "optimistic_dad",
5649			.data           = &ipv6_devconf.optimistic_dad,
5650			.maxlen         = sizeof(int),
5651			.mode           = 0644,
5652			.proc_handler   = proc_dointvec,
5653
5654		},
5655		{
5656			.procname       = "use_optimistic",
5657			.data           = &ipv6_devconf.use_optimistic,
5658			.maxlen         = sizeof(int),
5659			.mode           = 0644,
5660			.proc_handler   = proc_dointvec,
5661
5662		},
5663#endif
5664#ifdef CONFIG_IPV6_MROUTE
5665		{
5666			.procname	= "mc_forwarding",
5667			.data		= &ipv6_devconf.mc_forwarding,
5668			.maxlen		= sizeof(int),
5669			.mode		= 0444,
5670			.proc_handler	= proc_dointvec,
5671		},
5672#endif
5673		{
5674			.procname	= "disable_ipv6",
5675			.data		= &ipv6_devconf.disable_ipv6,
5676			.maxlen		= sizeof(int),
5677			.mode		= 0644,
5678			.proc_handler	= addrconf_sysctl_disable,
5679		},
5680		{
5681			.procname	= "accept_dad",
5682			.data		= &ipv6_devconf.accept_dad,
5683			.maxlen		= sizeof(int),
5684			.mode		= 0644,
5685			.proc_handler	= proc_dointvec,
5686		},
5687		{
5688			.procname       = "force_tllao",
5689			.data           = &ipv6_devconf.force_tllao,
5690			.maxlen         = sizeof(int),
5691			.mode           = 0644,
5692			.proc_handler   = proc_dointvec
5693		},
5694		{
5695			.procname       = "ndisc_notify",
5696			.data           = &ipv6_devconf.ndisc_notify,
5697			.maxlen         = sizeof(int),
5698			.mode           = 0644,
5699			.proc_handler   = proc_dointvec
5700		},
5701		{
5702			.procname	= "suppress_frag_ndisc",
5703			.data		= &ipv6_devconf.suppress_frag_ndisc,
5704			.maxlen		= sizeof(int),
5705			.mode		= 0644,
5706			.proc_handler	= proc_dointvec
5707		},
5708		{
5709			.procname	= "accept_ra_from_local",
5710			.data		= &ipv6_devconf.accept_ra_from_local,
5711			.maxlen		= sizeof(int),
5712			.mode		= 0644,
5713			.proc_handler	= proc_dointvec,
5714		},
5715		{
5716			.procname	= "accept_ra_mtu",
5717			.data		= &ipv6_devconf.accept_ra_mtu,
5718			.maxlen		= sizeof(int),
5719			.mode		= 0644,
5720			.proc_handler	= proc_dointvec,
5721		},
5722		{
5723			.procname	= "stable_secret",
5724			.data		= &ipv6_devconf.stable_secret,
5725			.maxlen		= IPV6_MAX_STRLEN,
5726			.mode		= 0600,
5727			.proc_handler	= addrconf_sysctl_stable_secret,
5728		},
5729		{
5730			.procname       = "use_oif_addrs_only",
5731			.data           = &ipv6_devconf.use_oif_addrs_only,
5732			.maxlen         = sizeof(int),
5733			.mode           = 0644,
5734			.proc_handler   = proc_dointvec,
5735		},
5736		{
5737			.procname	= "ignore_routes_with_linkdown",
5738			.data		= &ipv6_devconf.ignore_routes_with_linkdown,
5739			.maxlen		= sizeof(int),
5740			.mode		= 0644,
5741			.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
5742		},
5743		{
5744			/* sentinel */
5745		}
5746	},
5747};
5748
5749static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5750		struct inet6_dev *idev, struct ipv6_devconf *p)
5751{
5752	int i;
5753	struct addrconf_sysctl_table *t;
5754	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5755
5756	t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5757	if (!t)
5758		goto out;
5759
5760	for (i = 0; t->addrconf_vars[i].data; i++) {
5761		t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5762		t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5763		t->addrconf_vars[i].extra2 = net;
5764	}
5765
5766	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5767
5768	t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5769	if (!t->sysctl_header)
5770		goto free;
5771
5772	p->sysctl = t;
5773	return 0;
5774
5775free:
5776	kfree(t);
5777out:
5778	return -ENOBUFS;
5779}
5780
5781static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5782{
5783	struct addrconf_sysctl_table *t;
5784
5785	if (!p->sysctl)
5786		return;
5787
5788	t = p->sysctl;
5789	p->sysctl = NULL;
5790	unregister_net_sysctl_table(t->sysctl_header);
5791	kfree(t);
5792}
5793
5794static int addrconf_sysctl_register(struct inet6_dev *idev)
5795{
5796	int err;
5797
5798	if (!sysctl_dev_name_is_allowed(idev->dev->name))
5799		return -EINVAL;
5800
5801	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5802				    &ndisc_ifinfo_sysctl_change);
5803	if (err)
5804		return err;
5805	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5806					 idev, &idev->cnf);
5807	if (err)
5808		neigh_sysctl_unregister(idev->nd_parms);
5809
5810	return err;
5811}
5812
5813static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5814{
5815	__addrconf_sysctl_unregister(&idev->cnf);
5816	neigh_sysctl_unregister(idev->nd_parms);
5817}
5818
5819
5820#endif
5821
5822static int __net_init addrconf_init_net(struct net *net)
5823{
5824	int err = -ENOMEM;
5825	struct ipv6_devconf *all, *dflt;
5826
5827	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5828	if (!all)
5829		goto err_alloc_all;
5830
5831	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5832	if (!dflt)
5833		goto err_alloc_dflt;
5834
5835	/* these will be inherited by all namespaces */
5836	dflt->autoconf = ipv6_defaults.autoconf;
5837	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5838
5839	dflt->stable_secret.initialized = false;
5840	all->stable_secret.initialized = false;
5841
5842	net->ipv6.devconf_all = all;
5843	net->ipv6.devconf_dflt = dflt;
5844
5845#ifdef CONFIG_SYSCTL
5846	err = __addrconf_sysctl_register(net, "all", NULL, all);
5847	if (err < 0)
5848		goto err_reg_all;
5849
5850	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5851	if (err < 0)
5852		goto err_reg_dflt;
5853#endif
5854	return 0;
5855
5856#ifdef CONFIG_SYSCTL
5857err_reg_dflt:
5858	__addrconf_sysctl_unregister(all);
5859err_reg_all:
5860	kfree(dflt);
5861#endif
5862err_alloc_dflt:
5863	kfree(all);
5864err_alloc_all:
5865	return err;
5866}
5867
5868static void __net_exit addrconf_exit_net(struct net *net)
5869{
5870#ifdef CONFIG_SYSCTL
5871	__addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5872	__addrconf_sysctl_unregister(net->ipv6.devconf_all);
5873#endif
5874	kfree(net->ipv6.devconf_dflt);
5875	kfree(net->ipv6.devconf_all);
5876}
5877
5878static struct pernet_operations addrconf_ops = {
5879	.init = addrconf_init_net,
5880	.exit = addrconf_exit_net,
5881};
5882
5883static struct rtnl_af_ops inet6_ops __read_mostly = {
5884	.family		  = AF_INET6,
5885	.fill_link_af	  = inet6_fill_link_af,
5886	.get_link_af_size = inet6_get_link_af_size,
5887	.validate_link_af = inet6_validate_link_af,
5888	.set_link_af	  = inet6_set_link_af,
5889};
5890
5891/*
5892 *	Init / cleanup code
5893 */
5894
5895int __init addrconf_init(void)
5896{
5897	struct inet6_dev *idev;
5898	int i, err;
5899
5900	err = ipv6_addr_label_init();
5901	if (err < 0) {
5902		pr_crit("%s: cannot initialize default policy table: %d\n",
5903			__func__, err);
5904		goto out;
5905	}
5906
5907	err = register_pernet_subsys(&addrconf_ops);
5908	if (err < 0)
5909		goto out_addrlabel;
5910
5911	addrconf_wq = create_workqueue("ipv6_addrconf");
5912	if (!addrconf_wq) {
5913		err = -ENOMEM;
5914		goto out_nowq;
5915	}
5916
5917	/* The addrconf netdev notifier requires that loopback_dev
5918	 * has it's ipv6 private information allocated and setup
5919	 * before it can bring up and give link-local addresses
5920	 * to other devices which are up.
5921	 *
5922	 * Unfortunately, loopback_dev is not necessarily the first
5923	 * entry in the global dev_base list of net devices.  In fact,
5924	 * it is likely to be the very last entry on that list.
5925	 * So this causes the notifier registry below to try and
5926	 * give link-local addresses to all devices besides loopback_dev
5927	 * first, then loopback_dev, which cases all the non-loopback_dev
5928	 * devices to fail to get a link-local address.
5929	 *
5930	 * So, as a temporary fix, allocate the ipv6 structure for
5931	 * loopback_dev first by hand.
5932	 * Longer term, all of the dependencies ipv6 has upon the loopback
5933	 * device and it being up should be removed.
5934	 */
5935	rtnl_lock();
5936	idev = ipv6_add_dev(init_net.loopback_dev);
5937	rtnl_unlock();
5938	if (IS_ERR(idev)) {
5939		err = PTR_ERR(idev);
5940		goto errlo;
5941	}
5942
5943	for (i = 0; i < IN6_ADDR_HSIZE; i++)
5944		INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5945
5946	register_netdevice_notifier(&ipv6_dev_notf);
5947
5948	addrconf_verify();
5949
5950	rtnl_af_register(&inet6_ops);
5951
5952	err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5953			      NULL);
5954	if (err < 0)
5955		goto errout;
5956
5957	/* Only the first call to __rtnl_register can fail */
5958	__rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5959	__rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5960	__rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5961			inet6_dump_ifaddr, NULL);
5962	__rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5963			inet6_dump_ifmcaddr, NULL);
5964	__rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5965			inet6_dump_ifacaddr, NULL);
5966	__rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5967			inet6_netconf_dump_devconf, NULL);
5968
5969	ipv6_addr_label_rtnl_register();
5970
5971	return 0;
5972errout:
5973	rtnl_af_unregister(&inet6_ops);
5974	unregister_netdevice_notifier(&ipv6_dev_notf);
5975errlo:
5976	destroy_workqueue(addrconf_wq);
5977out_nowq:
5978	unregister_pernet_subsys(&addrconf_ops);
5979out_addrlabel:
5980	ipv6_addr_label_cleanup();
5981out:
5982	return err;
5983}
5984
5985void addrconf_cleanup(void)
5986{
5987	struct net_device *dev;
5988	int i;
5989
5990	unregister_netdevice_notifier(&ipv6_dev_notf);
5991	unregister_pernet_subsys(&addrconf_ops);
5992	ipv6_addr_label_cleanup();
5993
5994	rtnl_lock();
5995
5996	__rtnl_af_unregister(&inet6_ops);
5997
5998	/* clean dev list */
5999	for_each_netdev(&init_net, dev) {
6000		if (__in6_dev_get(dev) == NULL)
6001			continue;
6002		addrconf_ifdown(dev, 1);
6003	}
6004	addrconf_ifdown(init_net.loopback_dev, 2);
6005
6006	/*
6007	 *	Check hash table.
6008	 */
6009	spin_lock_bh(&addrconf_hash_lock);
6010	for (i = 0; i < IN6_ADDR_HSIZE; i++)
6011		WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6012	spin_unlock_bh(&addrconf_hash_lock);
6013	cancel_delayed_work(&addr_chk_work);
6014	rtnl_unlock();
6015
6016	destroy_workqueue(addrconf_wq);
6017}
6018