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