Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[sfrench/cifs-2.6.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/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
88 #endif
89
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
99
100 #if ACONF_DEBUG >= 3
101 #define ADBG(x) printk x
102 #else
103 #define ADBG(x)
104 #endif
105
106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
107
108 static inline u32 cstamp_delta(unsigned long cstamp)
109 {
110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 }
112
113 #ifdef CONFIG_SYSCTL
114 static void addrconf_sysctl_register(struct inet6_dev *idev);
115 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
116 #else
117 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
118 {
119 }
120
121 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
122 {
123 }
124 #endif
125
126 #ifdef CONFIG_IPV6_PRIVACY
127 static void __ipv6_regen_rndid(struct inet6_dev *idev);
128 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
129 static void ipv6_regen_rndid(unsigned long data);
130 #endif
131
132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
133 static int ipv6_count_addresses(struct inet6_dev *idev);
134
135 /*
136  *      Configured unicast address hash table
137  */
138 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
139 static DEFINE_SPINLOCK(addrconf_hash_lock);
140
141 static void addrconf_verify(unsigned long);
142
143 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
144 static DEFINE_SPINLOCK(addrconf_verify_lock);
145
146 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
148
149 static void addrconf_type_change(struct net_device *dev,
150                                  unsigned long event);
151 static int addrconf_ifdown(struct net_device *dev, int how);
152
153 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
154                                                   int plen,
155                                                   const struct net_device *dev,
156                                                   u32 flags, u32 noflags);
157
158 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_timer(unsigned long data);
160 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
161 static void addrconf_dad_run(struct inet6_dev *idev);
162 static void addrconf_rs_timer(unsigned long data);
163 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
165
166 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
167                                 struct prefix_info *pinfo);
168 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
169                                struct net_device *dev);
170
171 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
172
173 static struct ipv6_devconf ipv6_devconf __read_mostly = {
174         .forwarding             = 0,
175         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
176         .mtu6                   = IPV6_MIN_MTU,
177         .accept_ra              = 1,
178         .accept_redirects       = 1,
179         .autoconf               = 1,
180         .force_mld_version      = 0,
181         .dad_transmits          = 1,
182         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
183         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
184         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
185 #ifdef CONFIG_IPV6_PRIVACY
186         .use_tempaddr           = 0,
187         .temp_valid_lft         = TEMP_VALID_LIFETIME,
188         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
189         .regen_max_retry        = REGEN_MAX_RETRY,
190         .max_desync_factor      = MAX_DESYNC_FACTOR,
191 #endif
192         .max_addresses          = IPV6_MAX_ADDRESSES,
193         .accept_ra_defrtr       = 1,
194         .accept_ra_pinfo        = 1,
195 #ifdef CONFIG_IPV6_ROUTER_PREF
196         .accept_ra_rtr_pref     = 1,
197         .rtr_probe_interval     = 60 * HZ,
198 #ifdef CONFIG_IPV6_ROUTE_INFO
199         .accept_ra_rt_info_max_plen = 0,
200 #endif
201 #endif
202         .proxy_ndp              = 0,
203         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
204         .disable_ipv6           = 0,
205         .accept_dad             = 1,
206 };
207
208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
209         .forwarding             = 0,
210         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
211         .mtu6                   = IPV6_MIN_MTU,
212         .accept_ra              = 1,
213         .accept_redirects       = 1,
214         .autoconf               = 1,
215         .dad_transmits          = 1,
216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
219 #ifdef CONFIG_IPV6_PRIVACY
220         .use_tempaddr           = 0,
221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
223         .regen_max_retry        = REGEN_MAX_RETRY,
224         .max_desync_factor      = MAX_DESYNC_FACTOR,
225 #endif
226         .max_addresses          = IPV6_MAX_ADDRESSES,
227         .accept_ra_defrtr       = 1,
228         .accept_ra_pinfo        = 1,
229 #ifdef CONFIG_IPV6_ROUTER_PREF
230         .accept_ra_rtr_pref     = 1,
231         .rtr_probe_interval     = 60 * HZ,
232 #ifdef CONFIG_IPV6_ROUTE_INFO
233         .accept_ra_rt_info_max_plen = 0,
234 #endif
235 #endif
236         .proxy_ndp              = 0,
237         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
238         .disable_ipv6           = 0,
239         .accept_dad             = 1,
240 };
241
242 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
243 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
244 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
245 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
246 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
247
248 /* Check if a valid qdisc is available */
249 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
250 {
251         return !qdisc_tx_is_noop(dev);
252 }
253
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
255 {
256         if (del_timer(&ifp->timer))
257                 __in6_ifa_put(ifp);
258 }
259
260 enum addrconf_timer_t {
261         AC_NONE,
262         AC_DAD,
263         AC_RS,
264 };
265
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267                                enum addrconf_timer_t what,
268                                unsigned long when)
269 {
270         if (!del_timer(&ifp->timer))
271                 in6_ifa_hold(ifp);
272
273         switch (what) {
274         case AC_DAD:
275                 ifp->timer.function = addrconf_dad_timer;
276                 break;
277         case AC_RS:
278                 ifp->timer.function = addrconf_rs_timer;
279                 break;
280         default:
281                 break;
282         }
283         ifp->timer.expires = jiffies + when;
284         add_timer(&ifp->timer);
285 }
286
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
288 {
289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290                           sizeof(struct ipstats_mib),
291                           __alignof__(struct ipstats_mib)) < 0)
292                 goto err_ip;
293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294                                         GFP_KERNEL);
295         if (!idev->stats.icmpv6dev)
296                 goto err_icmp;
297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298                                            GFP_KERNEL);
299         if (!idev->stats.icmpv6msgdev)
300                 goto err_icmpmsg;
301
302         return 0;
303
304 err_icmpmsg:
305         kfree(idev->stats.icmpv6dev);
306 err_icmp:
307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 err_ip:
309         return -ENOMEM;
310 }
311
312 static void snmp6_free_dev(struct inet6_dev *idev)
313 {
314         kfree(idev->stats.icmpv6msgdev);
315         kfree(idev->stats.icmpv6dev);
316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
317 }
318
319 /* Nobody refers to this device, we may destroy it. */
320
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
322 {
323         struct net_device *dev = idev->dev;
324
325         WARN_ON(!list_empty(&idev->addr_list));
326         WARN_ON(idev->mc_list != NULL);
327
328 #ifdef NET_REFCNT_DEBUG
329         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
330 #endif
331         dev_put(dev);
332         if (!idev->dead) {
333                 pr_warn("Freeing alive inet6 device %p\n", idev);
334                 return;
335         }
336         snmp6_free_dev(idev);
337         kfree_rcu(idev, rcu);
338 }
339 EXPORT_SYMBOL(in6_dev_finish_destroy);
340
341 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
342 {
343         struct inet6_dev *ndev;
344
345         ASSERT_RTNL();
346
347         if (dev->mtu < IPV6_MIN_MTU)
348                 return NULL;
349
350         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
351
352         if (ndev == NULL)
353                 return NULL;
354
355         rwlock_init(&ndev->lock);
356         ndev->dev = dev;
357         INIT_LIST_HEAD(&ndev->addr_list);
358
359         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
360         ndev->cnf.mtu6 = dev->mtu;
361         ndev->cnf.sysctl = NULL;
362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
363         if (ndev->nd_parms == NULL) {
364                 kfree(ndev);
365                 return NULL;
366         }
367         if (ndev->cnf.forwarding)
368                 dev_disable_lro(dev);
369         /* We refer to the device */
370         dev_hold(dev);
371
372         if (snmp6_alloc_dev(ndev) < 0) {
373                 ADBG((KERN_WARNING
374                         "%s: cannot allocate memory for statistics; dev=%s.\n",
375                         __func__, dev->name));
376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
377                 dev_put(dev);
378                 kfree(ndev);
379                 return NULL;
380         }
381
382         if (snmp6_register_dev(ndev) < 0) {
383                 ADBG((KERN_WARNING
384                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
385                         __func__, dev->name));
386                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
387                 ndev->dead = 1;
388                 in6_dev_finish_destroy(ndev);
389                 return NULL;
390         }
391
392         /* One reference from device.  We must do this before
393          * we invoke __ipv6_regen_rndid().
394          */
395         in6_dev_hold(ndev);
396
397         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
398                 ndev->cnf.accept_dad = -1;
399
400 #if IS_ENABLED(CONFIG_IPV6_SIT)
401         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
402                 pr_info("%s: Disabled Multicast RS\n", dev->name);
403                 ndev->cnf.rtr_solicits = 0;
404         }
405 #endif
406
407 #ifdef CONFIG_IPV6_PRIVACY
408         INIT_LIST_HEAD(&ndev->tempaddr_list);
409         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
410         if ((dev->flags&IFF_LOOPBACK) ||
411             dev->type == ARPHRD_TUNNEL ||
412             dev->type == ARPHRD_TUNNEL6 ||
413             dev->type == ARPHRD_SIT ||
414             dev->type == ARPHRD_NONE) {
415                 ndev->cnf.use_tempaddr = -1;
416         } else {
417                 in6_dev_hold(ndev);
418                 ipv6_regen_rndid((unsigned long) ndev);
419         }
420 #endif
421
422         if (netif_running(dev) && addrconf_qdisc_ok(dev))
423                 ndev->if_flags |= IF_READY;
424
425         ipv6_mc_init_dev(ndev);
426         ndev->tstamp = jiffies;
427         addrconf_sysctl_register(ndev);
428         /* protected by rtnl_lock */
429         rcu_assign_pointer(dev->ip6_ptr, ndev);
430
431         /* Join all-node multicast group */
432         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
433
434         /* Join all-router multicast group if forwarding is set */
435         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
436                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
437
438         return ndev;
439 }
440
441 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
442 {
443         struct inet6_dev *idev;
444
445         ASSERT_RTNL();
446
447         idev = __in6_dev_get(dev);
448         if (!idev) {
449                 idev = ipv6_add_dev(dev);
450                 if (!idev)
451                         return NULL;
452         }
453
454         if (dev->flags&IFF_UP)
455                 ipv6_mc_up(idev);
456         return idev;
457 }
458
459 static int inet6_netconf_msgsize_devconf(int type)
460 {
461         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
462                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
463
464         /* type -1 is used for ALL */
465         if (type == -1 || type == NETCONFA_FORWARDING)
466                 size += nla_total_size(4);
467 #ifdef CONFIG_IPV6_MROUTE
468         if (type == -1 || type == NETCONFA_MC_FORWARDING)
469                 size += nla_total_size(4);
470 #endif
471
472         return size;
473 }
474
475 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
476                                       struct ipv6_devconf *devconf, u32 portid,
477                                       u32 seq, int event, unsigned int flags,
478                                       int type)
479 {
480         struct nlmsghdr  *nlh;
481         struct netconfmsg *ncm;
482
483         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
484                         flags);
485         if (nlh == NULL)
486                 return -EMSGSIZE;
487
488         ncm = nlmsg_data(nlh);
489         ncm->ncm_family = AF_INET6;
490
491         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
492                 goto nla_put_failure;
493
494         /* type -1 is used for ALL */
495         if ((type == -1 || type == NETCONFA_FORWARDING) &&
496             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
497                 goto nla_put_failure;
498 #ifdef CONFIG_IPV6_MROUTE
499         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
500             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
501                         devconf->mc_forwarding) < 0)
502                 goto nla_put_failure;
503 #endif
504         return nlmsg_end(skb, nlh);
505
506 nla_put_failure:
507         nlmsg_cancel(skb, nlh);
508         return -EMSGSIZE;
509 }
510
511 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
512                                   struct ipv6_devconf *devconf)
513 {
514         struct sk_buff *skb;
515         int err = -ENOBUFS;
516
517         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
518         if (skb == NULL)
519                 goto errout;
520
521         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
522                                          RTM_NEWNETCONF, 0, type);
523         if (err < 0) {
524                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
525                 WARN_ON(err == -EMSGSIZE);
526                 kfree_skb(skb);
527                 goto errout;
528         }
529         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
530         return;
531 errout:
532         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
533 }
534
535 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
536         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
537         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
538 };
539
540 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
541                                      struct nlmsghdr *nlh,
542                                      void *arg)
543 {
544         struct net *net = sock_net(in_skb->sk);
545         struct nlattr *tb[NETCONFA_MAX+1];
546         struct netconfmsg *ncm;
547         struct sk_buff *skb;
548         struct ipv6_devconf *devconf;
549         struct inet6_dev *in6_dev;
550         struct net_device *dev;
551         int ifindex;
552         int err;
553
554         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
555                           devconf_ipv6_policy);
556         if (err < 0)
557                 goto errout;
558
559         err = EINVAL;
560         if (!tb[NETCONFA_IFINDEX])
561                 goto errout;
562
563         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
564         switch (ifindex) {
565         case NETCONFA_IFINDEX_ALL:
566                 devconf = net->ipv6.devconf_all;
567                 break;
568         case NETCONFA_IFINDEX_DEFAULT:
569                 devconf = net->ipv6.devconf_dflt;
570                 break;
571         default:
572                 dev = __dev_get_by_index(net, ifindex);
573                 if (dev == NULL)
574                         goto errout;
575                 in6_dev = __in6_dev_get(dev);
576                 if (in6_dev == NULL)
577                         goto errout;
578                 devconf = &in6_dev->cnf;
579                 break;
580         }
581
582         err = -ENOBUFS;
583         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
584         if (skb == NULL)
585                 goto errout;
586
587         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
588                                          NETLINK_CB(in_skb).portid,
589                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
590                                          -1);
591         if (err < 0) {
592                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
593                 WARN_ON(err == -EMSGSIZE);
594                 kfree_skb(skb);
595                 goto errout;
596         }
597         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
598 errout:
599         return err;
600 }
601
602 #ifdef CONFIG_SYSCTL
603 static void dev_forward_change(struct inet6_dev *idev)
604 {
605         struct net_device *dev;
606         struct inet6_ifaddr *ifa;
607
608         if (!idev)
609                 return;
610         dev = idev->dev;
611         if (idev->cnf.forwarding)
612                 dev_disable_lro(dev);
613         if (dev->flags & IFF_MULTICAST) {
614                 if (idev->cnf.forwarding)
615                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
616                 else
617                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
618         }
619
620         list_for_each_entry(ifa, &idev->addr_list, if_list) {
621                 if (ifa->flags&IFA_F_TENTATIVE)
622                         continue;
623                 if (idev->cnf.forwarding)
624                         addrconf_join_anycast(ifa);
625                 else
626                         addrconf_leave_anycast(ifa);
627         }
628         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
629                                      dev->ifindex, &idev->cnf);
630 }
631
632
633 static void addrconf_forward_change(struct net *net, __s32 newf)
634 {
635         struct net_device *dev;
636         struct inet6_dev *idev;
637
638         for_each_netdev(net, dev) {
639                 idev = __in6_dev_get(dev);
640                 if (idev) {
641                         int changed = (!idev->cnf.forwarding) ^ (!newf);
642                         idev->cnf.forwarding = newf;
643                         if (changed)
644                                 dev_forward_change(idev);
645                 }
646         }
647 }
648
649 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
650 {
651         struct net *net;
652         int old;
653
654         if (!rtnl_trylock())
655                 return restart_syscall();
656
657         net = (struct net *)table->extra2;
658         old = *p;
659         *p = newf;
660
661         if (p == &net->ipv6.devconf_dflt->forwarding) {
662                 if ((!newf) ^ (!old))
663                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
664                                                      NETCONFA_IFINDEX_DEFAULT,
665                                                      net->ipv6.devconf_dflt);
666                 rtnl_unlock();
667                 return 0;
668         }
669
670         if (p == &net->ipv6.devconf_all->forwarding) {
671                 net->ipv6.devconf_dflt->forwarding = newf;
672                 addrconf_forward_change(net, newf);
673                 if ((!newf) ^ (!old))
674                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
675                                                      NETCONFA_IFINDEX_ALL,
676                                                      net->ipv6.devconf_all);
677         } else if ((!newf) ^ (!old))
678                 dev_forward_change((struct inet6_dev *)table->extra1);
679         rtnl_unlock();
680
681         if (newf)
682                 rt6_purge_dflt_routers(net);
683         return 1;
684 }
685 #endif
686
687 /* Nobody refers to this ifaddr, destroy it */
688 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
689 {
690         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
691
692 #ifdef NET_REFCNT_DEBUG
693         pr_debug("%s\n", __func__);
694 #endif
695
696         in6_dev_put(ifp->idev);
697
698         if (del_timer(&ifp->timer))
699                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
700
701         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
702                 pr_warn("Freeing alive inet6 address %p\n", ifp);
703                 return;
704         }
705         ip6_rt_put(ifp->rt);
706
707         kfree_rcu(ifp, rcu);
708 }
709
710 static void
711 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
712 {
713         struct list_head *p;
714         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
715
716         /*
717          * Each device address list is sorted in order of scope -
718          * global before linklocal.
719          */
720         list_for_each(p, &idev->addr_list) {
721                 struct inet6_ifaddr *ifa
722                         = list_entry(p, struct inet6_ifaddr, if_list);
723                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
724                         break;
725         }
726
727         list_add_tail(&ifp->if_list, p);
728 }
729
730 static u32 inet6_addr_hash(const struct in6_addr *addr)
731 {
732         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
733 }
734
735 /* On success it returns ifp with increased reference count */
736
737 static struct inet6_ifaddr *
738 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
739               int scope, u32 flags)
740 {
741         struct inet6_ifaddr *ifa = NULL;
742         struct rt6_info *rt;
743         unsigned int hash;
744         int err = 0;
745         int addr_type = ipv6_addr_type(addr);
746
747         if (addr_type == IPV6_ADDR_ANY ||
748             addr_type & IPV6_ADDR_MULTICAST ||
749             (!(idev->dev->flags & IFF_LOOPBACK) &&
750              addr_type & IPV6_ADDR_LOOPBACK))
751                 return ERR_PTR(-EADDRNOTAVAIL);
752
753         rcu_read_lock_bh();
754         if (idev->dead) {
755                 err = -ENODEV;                  /*XXX*/
756                 goto out2;
757         }
758
759         if (idev->cnf.disable_ipv6) {
760                 err = -EACCES;
761                 goto out2;
762         }
763
764         spin_lock(&addrconf_hash_lock);
765
766         /* Ignore adding duplicate addresses on an interface */
767         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
768                 ADBG(("ipv6_add_addr: already assigned\n"));
769                 err = -EEXIST;
770                 goto out;
771         }
772
773         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
774
775         if (ifa == NULL) {
776                 ADBG(("ipv6_add_addr: malloc failed\n"));
777                 err = -ENOBUFS;
778                 goto out;
779         }
780
781         rt = addrconf_dst_alloc(idev, addr, false);
782         if (IS_ERR(rt)) {
783                 err = PTR_ERR(rt);
784                 goto out;
785         }
786
787         ifa->addr = *addr;
788
789         spin_lock_init(&ifa->lock);
790         spin_lock_init(&ifa->state_lock);
791         init_timer(&ifa->timer);
792         INIT_HLIST_NODE(&ifa->addr_lst);
793         ifa->timer.data = (unsigned long) ifa;
794         ifa->scope = scope;
795         ifa->prefix_len = pfxlen;
796         ifa->flags = flags | IFA_F_TENTATIVE;
797         ifa->cstamp = ifa->tstamp = jiffies;
798
799         ifa->rt = rt;
800
801         ifa->idev = idev;
802         in6_dev_hold(idev);
803         /* For caller */
804         in6_ifa_hold(ifa);
805
806         /* Add to big hash table */
807         hash = inet6_addr_hash(addr);
808
809         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
810         spin_unlock(&addrconf_hash_lock);
811
812         write_lock(&idev->lock);
813         /* Add to inet6_dev unicast addr list. */
814         ipv6_link_dev_addr(idev, ifa);
815
816 #ifdef CONFIG_IPV6_PRIVACY
817         if (ifa->flags&IFA_F_TEMPORARY) {
818                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
819                 in6_ifa_hold(ifa);
820         }
821 #endif
822
823         in6_ifa_hold(ifa);
824         write_unlock(&idev->lock);
825 out2:
826         rcu_read_unlock_bh();
827
828         if (likely(err == 0))
829                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
830         else {
831                 kfree(ifa);
832                 ifa = ERR_PTR(err);
833         }
834
835         return ifa;
836 out:
837         spin_unlock(&addrconf_hash_lock);
838         goto out2;
839 }
840
841 /* This function wants to get referenced ifp and releases it before return */
842
843 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
844 {
845         struct inet6_ifaddr *ifa, *ifn;
846         struct inet6_dev *idev = ifp->idev;
847         int state;
848         int deleted = 0, onlink = 0;
849         unsigned long expires = jiffies;
850
851         spin_lock_bh(&ifp->state_lock);
852         state = ifp->state;
853         ifp->state = INET6_IFADDR_STATE_DEAD;
854         spin_unlock_bh(&ifp->state_lock);
855
856         if (state == INET6_IFADDR_STATE_DEAD)
857                 goto out;
858
859         spin_lock_bh(&addrconf_hash_lock);
860         hlist_del_init_rcu(&ifp->addr_lst);
861         spin_unlock_bh(&addrconf_hash_lock);
862
863         write_lock_bh(&idev->lock);
864 #ifdef CONFIG_IPV6_PRIVACY
865         if (ifp->flags&IFA_F_TEMPORARY) {
866                 list_del(&ifp->tmp_list);
867                 if (ifp->ifpub) {
868                         in6_ifa_put(ifp->ifpub);
869                         ifp->ifpub = NULL;
870                 }
871                 __in6_ifa_put(ifp);
872         }
873 #endif
874
875         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
876                 if (ifa == ifp) {
877                         list_del_init(&ifp->if_list);
878                         __in6_ifa_put(ifp);
879
880                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
881                                 break;
882                         deleted = 1;
883                         continue;
884                 } else if (ifp->flags & IFA_F_PERMANENT) {
885                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
886                                               ifp->prefix_len)) {
887                                 if (ifa->flags & IFA_F_PERMANENT) {
888                                         onlink = 1;
889                                         if (deleted)
890                                                 break;
891                                 } else {
892                                         unsigned long lifetime;
893
894                                         if (!onlink)
895                                                 onlink = -1;
896
897                                         spin_lock(&ifa->lock);
898
899                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
900                                         /*
901                                          * Note: Because this address is
902                                          * not permanent, lifetime <
903                                          * LONG_MAX / HZ here.
904                                          */
905                                         if (time_before(expires,
906                                                         ifa->tstamp + lifetime * HZ))
907                                                 expires = ifa->tstamp + lifetime * HZ;
908                                         spin_unlock(&ifa->lock);
909                                 }
910                         }
911                 }
912         }
913         write_unlock_bh(&idev->lock);
914
915         addrconf_del_timer(ifp);
916
917         ipv6_ifa_notify(RTM_DELADDR, ifp);
918
919         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
920
921         /*
922          * Purge or update corresponding prefix
923          *
924          * 1) we don't purge prefix here if address was not permanent.
925          *    prefix is managed by its own lifetime.
926          * 2) if there're no addresses, delete prefix.
927          * 3) if there're still other permanent address(es),
928          *    corresponding prefix is still permanent.
929          * 4) otherwise, update prefix lifetime to the
930          *    longest valid lifetime among the corresponding
931          *    addresses on the device.
932          *    Note: subsequent RA will update lifetime.
933          *
934          * --yoshfuji
935          */
936         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
937                 struct in6_addr prefix;
938                 struct rt6_info *rt;
939
940                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
941
942                 rt = addrconf_get_prefix_route(&prefix,
943                                                ifp->prefix_len,
944                                                ifp->idev->dev,
945                                                0, RTF_GATEWAY | RTF_DEFAULT);
946
947                 if (rt) {
948                         if (onlink == 0) {
949                                 ip6_del_rt(rt);
950                                 rt = NULL;
951                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
952                                 rt6_set_expires(rt, expires);
953                         }
954                 }
955                 ip6_rt_put(rt);
956         }
957
958         /* clean up prefsrc entries */
959         rt6_remove_prefsrc(ifp);
960 out:
961         in6_ifa_put(ifp);
962 }
963
964 #ifdef CONFIG_IPV6_PRIVACY
965 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
966 {
967         struct inet6_dev *idev = ifp->idev;
968         struct in6_addr addr, *tmpaddr;
969         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
970         unsigned long regen_advance;
971         int tmp_plen;
972         int ret = 0;
973         int max_addresses;
974         u32 addr_flags;
975         unsigned long now = jiffies;
976
977         write_lock(&idev->lock);
978         if (ift) {
979                 spin_lock_bh(&ift->lock);
980                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
981                 spin_unlock_bh(&ift->lock);
982                 tmpaddr = &addr;
983         } else {
984                 tmpaddr = NULL;
985         }
986 retry:
987         in6_dev_hold(idev);
988         if (idev->cnf.use_tempaddr <= 0) {
989                 write_unlock(&idev->lock);
990                 pr_info("%s: use_tempaddr is disabled\n", __func__);
991                 in6_dev_put(idev);
992                 ret = -1;
993                 goto out;
994         }
995         spin_lock_bh(&ifp->lock);
996         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
997                 idev->cnf.use_tempaddr = -1;    /*XXX*/
998                 spin_unlock_bh(&ifp->lock);
999                 write_unlock(&idev->lock);
1000                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1001                         __func__);
1002                 in6_dev_put(idev);
1003                 ret = -1;
1004                 goto out;
1005         }
1006         in6_ifa_hold(ifp);
1007         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1008         __ipv6_try_regen_rndid(idev, tmpaddr);
1009         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1010         age = (now - ifp->tstamp) / HZ;
1011         tmp_valid_lft = min_t(__u32,
1012                               ifp->valid_lft,
1013                               idev->cnf.temp_valid_lft + age);
1014         tmp_prefered_lft = min_t(__u32,
1015                                  ifp->prefered_lft,
1016                                  idev->cnf.temp_prefered_lft + age -
1017                                  idev->cnf.max_desync_factor);
1018         tmp_plen = ifp->prefix_len;
1019         max_addresses = idev->cnf.max_addresses;
1020         tmp_tstamp = ifp->tstamp;
1021         spin_unlock_bh(&ifp->lock);
1022
1023         regen_advance = idev->cnf.regen_max_retry *
1024                         idev->cnf.dad_transmits *
1025                         idev->nd_parms->retrans_time / HZ;
1026         write_unlock(&idev->lock);
1027
1028         /* A temporary address is created only if this calculated Preferred
1029          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1030          * an implementation must not create a temporary address with a zero
1031          * Preferred Lifetime.
1032          */
1033         if (tmp_prefered_lft <= regen_advance) {
1034                 in6_ifa_put(ifp);
1035                 in6_dev_put(idev);
1036                 ret = -1;
1037                 goto out;
1038         }
1039
1040         addr_flags = IFA_F_TEMPORARY;
1041         /* set in addrconf_prefix_rcv() */
1042         if (ifp->flags & IFA_F_OPTIMISTIC)
1043                 addr_flags |= IFA_F_OPTIMISTIC;
1044
1045         ift = !max_addresses ||
1046               ipv6_count_addresses(idev) < max_addresses ?
1047                 ipv6_add_addr(idev, &addr, tmp_plen,
1048                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1049                               addr_flags) : NULL;
1050         if (IS_ERR_OR_NULL(ift)) {
1051                 in6_ifa_put(ifp);
1052                 in6_dev_put(idev);
1053                 pr_info("%s: retry temporary address regeneration\n", __func__);
1054                 tmpaddr = &addr;
1055                 write_lock(&idev->lock);
1056                 goto retry;
1057         }
1058
1059         spin_lock_bh(&ift->lock);
1060         ift->ifpub = ifp;
1061         ift->valid_lft = tmp_valid_lft;
1062         ift->prefered_lft = tmp_prefered_lft;
1063         ift->cstamp = now;
1064         ift->tstamp = tmp_tstamp;
1065         spin_unlock_bh(&ift->lock);
1066
1067         addrconf_dad_start(ift);
1068         in6_ifa_put(ift);
1069         in6_dev_put(idev);
1070 out:
1071         return ret;
1072 }
1073 #endif
1074
1075 /*
1076  *      Choose an appropriate source address (RFC3484)
1077  */
1078 enum {
1079         IPV6_SADDR_RULE_INIT = 0,
1080         IPV6_SADDR_RULE_LOCAL,
1081         IPV6_SADDR_RULE_SCOPE,
1082         IPV6_SADDR_RULE_PREFERRED,
1083 #ifdef CONFIG_IPV6_MIP6
1084         IPV6_SADDR_RULE_HOA,
1085 #endif
1086         IPV6_SADDR_RULE_OIF,
1087         IPV6_SADDR_RULE_LABEL,
1088 #ifdef CONFIG_IPV6_PRIVACY
1089         IPV6_SADDR_RULE_PRIVACY,
1090 #endif
1091         IPV6_SADDR_RULE_ORCHID,
1092         IPV6_SADDR_RULE_PREFIX,
1093         IPV6_SADDR_RULE_MAX
1094 };
1095
1096 struct ipv6_saddr_score {
1097         int                     rule;
1098         int                     addr_type;
1099         struct inet6_ifaddr     *ifa;
1100         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1101         int                     scopedist;
1102         int                     matchlen;
1103 };
1104
1105 struct ipv6_saddr_dst {
1106         const struct in6_addr *addr;
1107         int ifindex;
1108         int scope;
1109         int label;
1110         unsigned int prefs;
1111 };
1112
1113 static inline int ipv6_saddr_preferred(int type)
1114 {
1115         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1116                 return 1;
1117         return 0;
1118 }
1119
1120 static int ipv6_get_saddr_eval(struct net *net,
1121                                struct ipv6_saddr_score *score,
1122                                struct ipv6_saddr_dst *dst,
1123                                int i)
1124 {
1125         int ret;
1126
1127         if (i <= score->rule) {
1128                 switch (i) {
1129                 case IPV6_SADDR_RULE_SCOPE:
1130                         ret = score->scopedist;
1131                         break;
1132                 case IPV6_SADDR_RULE_PREFIX:
1133                         ret = score->matchlen;
1134                         break;
1135                 default:
1136                         ret = !!test_bit(i, score->scorebits);
1137                 }
1138                 goto out;
1139         }
1140
1141         switch (i) {
1142         case IPV6_SADDR_RULE_INIT:
1143                 /* Rule 0: remember if hiscore is not ready yet */
1144                 ret = !!score->ifa;
1145                 break;
1146         case IPV6_SADDR_RULE_LOCAL:
1147                 /* Rule 1: Prefer same address */
1148                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1149                 break;
1150         case IPV6_SADDR_RULE_SCOPE:
1151                 /* Rule 2: Prefer appropriate scope
1152                  *
1153                  *      ret
1154                  *       ^
1155                  *    -1 |  d 15
1156                  *    ---+--+-+---> scope
1157                  *       |
1158                  *       |             d is scope of the destination.
1159                  *  B-d  |  \
1160                  *       |   \      <- smaller scope is better if
1161                  *  B-15 |    \        if scope is enough for destinaion.
1162                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1163                  * d-C-1 | /
1164                  *       |/         <- greater is better
1165                  *   -C  /             if scope is not enough for destination.
1166                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1167                  *
1168                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1169                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1170                  * Assume B = 0 and we get C > 29.
1171                  */
1172                 ret = __ipv6_addr_src_scope(score->addr_type);
1173                 if (ret >= dst->scope)
1174                         ret = -ret;
1175                 else
1176                         ret -= 128;     /* 30 is enough */
1177                 score->scopedist = ret;
1178                 break;
1179         case IPV6_SADDR_RULE_PREFERRED:
1180                 /* Rule 3: Avoid deprecated and optimistic addresses */
1181                 ret = ipv6_saddr_preferred(score->addr_type) ||
1182                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1183                 break;
1184 #ifdef CONFIG_IPV6_MIP6
1185         case IPV6_SADDR_RULE_HOA:
1186             {
1187                 /* Rule 4: Prefer home address */
1188                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1189                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1190                 break;
1191             }
1192 #endif
1193         case IPV6_SADDR_RULE_OIF:
1194                 /* Rule 5: Prefer outgoing interface */
1195                 ret = (!dst->ifindex ||
1196                        dst->ifindex == score->ifa->idev->dev->ifindex);
1197                 break;
1198         case IPV6_SADDR_RULE_LABEL:
1199                 /* Rule 6: Prefer matching label */
1200                 ret = ipv6_addr_label(net,
1201                                       &score->ifa->addr, score->addr_type,
1202                                       score->ifa->idev->dev->ifindex) == dst->label;
1203                 break;
1204 #ifdef CONFIG_IPV6_PRIVACY
1205         case IPV6_SADDR_RULE_PRIVACY:
1206             {
1207                 /* Rule 7: Prefer public address
1208                  * Note: prefer temporary address if use_tempaddr >= 2
1209                  */
1210                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1211                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1212                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1213                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1214                 break;
1215             }
1216 #endif
1217         case IPV6_SADDR_RULE_ORCHID:
1218                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1219                  *          non-ORCHID vs non-ORCHID
1220                  */
1221                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1222                         ipv6_addr_orchid(dst->addr));
1223                 break;
1224         case IPV6_SADDR_RULE_PREFIX:
1225                 /* Rule 8: Use longest matching prefix */
1226                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1227                 if (ret > score->ifa->prefix_len)
1228                         ret = score->ifa->prefix_len;
1229                 score->matchlen = ret;
1230                 break;
1231         default:
1232                 ret = 0;
1233         }
1234
1235         if (ret)
1236                 __set_bit(i, score->scorebits);
1237         score->rule = i;
1238 out:
1239         return ret;
1240 }
1241
1242 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1243                        const struct in6_addr *daddr, unsigned int prefs,
1244                        struct in6_addr *saddr)
1245 {
1246         struct ipv6_saddr_score scores[2],
1247                                 *score = &scores[0], *hiscore = &scores[1];
1248         struct ipv6_saddr_dst dst;
1249         struct net_device *dev;
1250         int dst_type;
1251
1252         dst_type = __ipv6_addr_type(daddr);
1253         dst.addr = daddr;
1254         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1255         dst.scope = __ipv6_addr_src_scope(dst_type);
1256         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1257         dst.prefs = prefs;
1258
1259         hiscore->rule = -1;
1260         hiscore->ifa = NULL;
1261
1262         rcu_read_lock();
1263
1264         for_each_netdev_rcu(net, dev) {
1265                 struct inet6_dev *idev;
1266
1267                 /* Candidate Source Address (section 4)
1268                  *  - multicast and link-local destination address,
1269                  *    the set of candidate source address MUST only
1270                  *    include addresses assigned to interfaces
1271                  *    belonging to the same link as the outgoing
1272                  *    interface.
1273                  * (- For site-local destination addresses, the
1274                  *    set of candidate source addresses MUST only
1275                  *    include addresses assigned to interfaces
1276                  *    belonging to the same site as the outgoing
1277                  *    interface.)
1278                  */
1279                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1280                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1281                     dst.ifindex && dev->ifindex != dst.ifindex)
1282                         continue;
1283
1284                 idev = __in6_dev_get(dev);
1285                 if (!idev)
1286                         continue;
1287
1288                 read_lock_bh(&idev->lock);
1289                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1290                         int i;
1291
1292                         /*
1293                          * - Tentative Address (RFC2462 section 5.4)
1294                          *  - A tentative address is not considered
1295                          *    "assigned to an interface" in the traditional
1296                          *    sense, unless it is also flagged as optimistic.
1297                          * - Candidate Source Address (section 4)
1298                          *  - In any case, anycast addresses, multicast
1299                          *    addresses, and the unspecified address MUST
1300                          *    NOT be included in a candidate set.
1301                          */
1302                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1303                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1304                                 continue;
1305
1306                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1307
1308                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1309                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1310                                 LIMIT_NETDEBUG(KERN_DEBUG
1311                                                "ADDRCONF: unspecified / multicast address "
1312                                                "assigned as unicast address on %s",
1313                                                dev->name);
1314                                 continue;
1315                         }
1316
1317                         score->rule = -1;
1318                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1319
1320                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1321                                 int minihiscore, miniscore;
1322
1323                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1324                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1325
1326                                 if (minihiscore > miniscore) {
1327                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1328                                             score->scopedist > 0) {
1329                                                 /*
1330                                                  * special case:
1331                                                  * each remaining entry
1332                                                  * has too small (not enough)
1333                                                  * scope, because ifa entries
1334                                                  * are sorted by their scope
1335                                                  * values.
1336                                                  */
1337                                                 goto try_nextdev;
1338                                         }
1339                                         break;
1340                                 } else if (minihiscore < miniscore) {
1341                                         if (hiscore->ifa)
1342                                                 in6_ifa_put(hiscore->ifa);
1343
1344                                         in6_ifa_hold(score->ifa);
1345
1346                                         swap(hiscore, score);
1347
1348                                         /* restore our iterator */
1349                                         score->ifa = hiscore->ifa;
1350
1351                                         break;
1352                                 }
1353                         }
1354                 }
1355 try_nextdev:
1356                 read_unlock_bh(&idev->lock);
1357         }
1358         rcu_read_unlock();
1359
1360         if (!hiscore->ifa)
1361                 return -EADDRNOTAVAIL;
1362
1363         *saddr = hiscore->ifa->addr;
1364         in6_ifa_put(hiscore->ifa);
1365         return 0;
1366 }
1367 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1368
1369 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1370                     unsigned char banned_flags)
1371 {
1372         struct inet6_dev *idev;
1373         int err = -EADDRNOTAVAIL;
1374
1375         rcu_read_lock();
1376         idev = __in6_dev_get(dev);
1377         if (idev) {
1378                 struct inet6_ifaddr *ifp;
1379
1380                 read_lock_bh(&idev->lock);
1381                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1382                         if (ifp->scope == IFA_LINK &&
1383                             !(ifp->flags & banned_flags)) {
1384                                 *addr = ifp->addr;
1385                                 err = 0;
1386                                 break;
1387                         }
1388                 }
1389                 read_unlock_bh(&idev->lock);
1390         }
1391         rcu_read_unlock();
1392         return err;
1393 }
1394
1395 static int ipv6_count_addresses(struct inet6_dev *idev)
1396 {
1397         int cnt = 0;
1398         struct inet6_ifaddr *ifp;
1399
1400         read_lock_bh(&idev->lock);
1401         list_for_each_entry(ifp, &idev->addr_list, if_list)
1402                 cnt++;
1403         read_unlock_bh(&idev->lock);
1404         return cnt;
1405 }
1406
1407 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1408                   struct net_device *dev, int strict)
1409 {
1410         struct inet6_ifaddr *ifp;
1411         struct hlist_node *node;
1412         unsigned int hash = inet6_addr_hash(addr);
1413
1414         rcu_read_lock_bh();
1415         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1416                 if (!net_eq(dev_net(ifp->idev->dev), net))
1417                         continue;
1418                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1419                     !(ifp->flags&IFA_F_TENTATIVE) &&
1420                     (dev == NULL || ifp->idev->dev == dev ||
1421                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1422                         rcu_read_unlock_bh();
1423                         return 1;
1424                 }
1425         }
1426
1427         rcu_read_unlock_bh();
1428         return 0;
1429 }
1430 EXPORT_SYMBOL(ipv6_chk_addr);
1431
1432 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1433                                struct net_device *dev)
1434 {
1435         unsigned int hash = inet6_addr_hash(addr);
1436         struct inet6_ifaddr *ifp;
1437         struct hlist_node *node;
1438
1439         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1440                 if (!net_eq(dev_net(ifp->idev->dev), net))
1441                         continue;
1442                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1443                         if (dev == NULL || ifp->idev->dev == dev)
1444                                 return true;
1445                 }
1446         }
1447         return false;
1448 }
1449
1450 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1451 {
1452         struct inet6_dev *idev;
1453         struct inet6_ifaddr *ifa;
1454         int     onlink;
1455
1456         onlink = 0;
1457         rcu_read_lock();
1458         idev = __in6_dev_get(dev);
1459         if (idev) {
1460                 read_lock_bh(&idev->lock);
1461                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1462                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1463                                                    ifa->prefix_len);
1464                         if (onlink)
1465                                 break;
1466                 }
1467                 read_unlock_bh(&idev->lock);
1468         }
1469         rcu_read_unlock();
1470         return onlink;
1471 }
1472 EXPORT_SYMBOL(ipv6_chk_prefix);
1473
1474 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1475                                      struct net_device *dev, int strict)
1476 {
1477         struct inet6_ifaddr *ifp, *result = NULL;
1478         unsigned int hash = inet6_addr_hash(addr);
1479         struct hlist_node *node;
1480
1481         rcu_read_lock_bh();
1482         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1483                 if (!net_eq(dev_net(ifp->idev->dev), net))
1484                         continue;
1485                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1486                         if (dev == NULL || ifp->idev->dev == dev ||
1487                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1488                                 result = ifp;
1489                                 in6_ifa_hold(ifp);
1490                                 break;
1491                         }
1492                 }
1493         }
1494         rcu_read_unlock_bh();
1495
1496         return result;
1497 }
1498
1499 /* Gets referenced address, destroys ifaddr */
1500
1501 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1502 {
1503         if (ifp->flags&IFA_F_PERMANENT) {
1504                 spin_lock_bh(&ifp->lock);
1505                 addrconf_del_timer(ifp);
1506                 ifp->flags |= IFA_F_TENTATIVE;
1507                 if (dad_failed)
1508                         ifp->flags |= IFA_F_DADFAILED;
1509                 spin_unlock_bh(&ifp->lock);
1510                 if (dad_failed)
1511                         ipv6_ifa_notify(0, ifp);
1512                 in6_ifa_put(ifp);
1513 #ifdef CONFIG_IPV6_PRIVACY
1514         } else if (ifp->flags&IFA_F_TEMPORARY) {
1515                 struct inet6_ifaddr *ifpub;
1516                 spin_lock_bh(&ifp->lock);
1517                 ifpub = ifp->ifpub;
1518                 if (ifpub) {
1519                         in6_ifa_hold(ifpub);
1520                         spin_unlock_bh(&ifp->lock);
1521                         ipv6_create_tempaddr(ifpub, ifp);
1522                         in6_ifa_put(ifpub);
1523                 } else {
1524                         spin_unlock_bh(&ifp->lock);
1525                 }
1526                 ipv6_del_addr(ifp);
1527 #endif
1528         } else
1529                 ipv6_del_addr(ifp);
1530 }
1531
1532 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1533 {
1534         int err = -ENOENT;
1535
1536         spin_lock(&ifp->state_lock);
1537         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1538                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1539                 err = 0;
1540         }
1541         spin_unlock(&ifp->state_lock);
1542
1543         return err;
1544 }
1545
1546 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1547 {
1548         struct inet6_dev *idev = ifp->idev;
1549
1550         if (addrconf_dad_end(ifp)) {
1551                 in6_ifa_put(ifp);
1552                 return;
1553         }
1554
1555         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1556                              ifp->idev->dev->name, &ifp->addr);
1557
1558         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1559                 struct in6_addr addr;
1560
1561                 addr.s6_addr32[0] = htonl(0xfe800000);
1562                 addr.s6_addr32[1] = 0;
1563
1564                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1565                     ipv6_addr_equal(&ifp->addr, &addr)) {
1566                         /* DAD failed for link-local based on MAC address */
1567                         idev->cnf.disable_ipv6 = 1;
1568
1569                         pr_info("%s: IPv6 being disabled!\n",
1570                                 ifp->idev->dev->name);
1571                 }
1572         }
1573
1574         addrconf_dad_stop(ifp, 1);
1575 }
1576
1577 /* Join to solicited addr multicast group. */
1578
1579 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1580 {
1581         struct in6_addr maddr;
1582
1583         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1584                 return;
1585
1586         addrconf_addr_solict_mult(addr, &maddr);
1587         ipv6_dev_mc_inc(dev, &maddr);
1588 }
1589
1590 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1591 {
1592         struct in6_addr maddr;
1593
1594         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1595                 return;
1596
1597         addrconf_addr_solict_mult(addr, &maddr);
1598         __ipv6_dev_mc_dec(idev, &maddr);
1599 }
1600
1601 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1602 {
1603         struct in6_addr addr;
1604         if (ifp->prefix_len == 127) /* RFC 6164 */
1605                 return;
1606         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1607         if (ipv6_addr_any(&addr))
1608                 return;
1609         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1610 }
1611
1612 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1613 {
1614         struct in6_addr addr;
1615         if (ifp->prefix_len == 127) /* RFC 6164 */
1616                 return;
1617         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1618         if (ipv6_addr_any(&addr))
1619                 return;
1620         __ipv6_dev_ac_dec(ifp->idev, &addr);
1621 }
1622
1623 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1624 {
1625         if (dev->addr_len != ETH_ALEN)
1626                 return -1;
1627         memcpy(eui, dev->dev_addr, 3);
1628         memcpy(eui + 5, dev->dev_addr + 3, 3);
1629
1630         /*
1631          * The zSeries OSA network cards can be shared among various
1632          * OS instances, but the OSA cards have only one MAC address.
1633          * This leads to duplicate address conflicts in conjunction
1634          * with IPv6 if more than one instance uses the same card.
1635          *
1636          * The driver for these cards can deliver a unique 16-bit
1637          * identifier for each instance sharing the same card.  It is
1638          * placed instead of 0xFFFE in the interface identifier.  The
1639          * "u" bit of the interface identifier is not inverted in this
1640          * case.  Hence the resulting interface identifier has local
1641          * scope according to RFC2373.
1642          */
1643         if (dev->dev_id) {
1644                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1645                 eui[4] = dev->dev_id & 0xFF;
1646         } else {
1647                 eui[3] = 0xFF;
1648                 eui[4] = 0xFE;
1649                 eui[0] ^= 2;
1650         }
1651         return 0;
1652 }
1653
1654 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1655 {
1656         if (dev->addr_len != IEEE802154_ADDR_LEN)
1657                 return -1;
1658         memcpy(eui, dev->dev_addr, 8);
1659         eui[0] ^= 2;
1660         return 0;
1661 }
1662
1663 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1664 {
1665         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1666         if (dev->addr_len != ARCNET_ALEN)
1667                 return -1;
1668         memset(eui, 0, 7);
1669         eui[7] = *(u8 *)dev->dev_addr;
1670         return 0;
1671 }
1672
1673 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1674 {
1675         if (dev->addr_len != INFINIBAND_ALEN)
1676                 return -1;
1677         memcpy(eui, dev->dev_addr + 12, 8);
1678         eui[0] |= 2;
1679         return 0;
1680 }
1681
1682 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1683 {
1684         if (addr == 0)
1685                 return -1;
1686         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1687                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1688                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1689                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1690                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1691                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1692         eui[1] = 0;
1693         eui[2] = 0x5E;
1694         eui[3] = 0xFE;
1695         memcpy(eui + 4, &addr, 4);
1696         return 0;
1697 }
1698
1699 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1700 {
1701         if (dev->priv_flags & IFF_ISATAP)
1702                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1703         return -1;
1704 }
1705
1706 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1707 {
1708         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1709 }
1710
1711 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1712 {
1713         switch (dev->type) {
1714         case ARPHRD_ETHER:
1715         case ARPHRD_FDDI:
1716                 return addrconf_ifid_eui48(eui, dev);
1717         case ARPHRD_ARCNET:
1718                 return addrconf_ifid_arcnet(eui, dev);
1719         case ARPHRD_INFINIBAND:
1720                 return addrconf_ifid_infiniband(eui, dev);
1721         case ARPHRD_SIT:
1722                 return addrconf_ifid_sit(eui, dev);
1723         case ARPHRD_IPGRE:
1724                 return addrconf_ifid_gre(eui, dev);
1725         case ARPHRD_IEEE802154:
1726                 return addrconf_ifid_eui64(eui, dev);
1727         }
1728         return -1;
1729 }
1730
1731 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1732 {
1733         int err = -1;
1734         struct inet6_ifaddr *ifp;
1735
1736         read_lock_bh(&idev->lock);
1737         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1738                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1739                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1740                         err = 0;
1741                         break;
1742                 }
1743         }
1744         read_unlock_bh(&idev->lock);
1745         return err;
1746 }
1747
1748 #ifdef CONFIG_IPV6_PRIVACY
1749 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1750 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1751 {
1752 regen:
1753         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1754         idev->rndid[0] &= ~0x02;
1755
1756         /*
1757          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1758          * check if generated address is not inappropriate
1759          *
1760          *  - Reserved subnet anycast (RFC 2526)
1761          *      11111101 11....11 1xxxxxxx
1762          *  - ISATAP (RFC4214) 6.1
1763          *      00-00-5E-FE-xx-xx-xx-xx
1764          *  - value 0
1765          *  - XXX: already assigned to an address on the device
1766          */
1767         if (idev->rndid[0] == 0xfd &&
1768             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1769             (idev->rndid[7]&0x80))
1770                 goto regen;
1771         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1772                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1773                         goto regen;
1774                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1775                         goto regen;
1776         }
1777 }
1778
1779 static void ipv6_regen_rndid(unsigned long data)
1780 {
1781         struct inet6_dev *idev = (struct inet6_dev *) data;
1782         unsigned long expires;
1783
1784         rcu_read_lock_bh();
1785         write_lock_bh(&idev->lock);
1786
1787         if (idev->dead)
1788                 goto out;
1789
1790         __ipv6_regen_rndid(idev);
1791
1792         expires = jiffies +
1793                 idev->cnf.temp_prefered_lft * HZ -
1794                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1795                 idev->cnf.max_desync_factor * HZ;
1796         if (time_before(expires, jiffies)) {
1797                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1798                         __func__, idev->dev->name);
1799                 goto out;
1800         }
1801
1802         if (!mod_timer(&idev->regen_timer, expires))
1803                 in6_dev_hold(idev);
1804
1805 out:
1806         write_unlock_bh(&idev->lock);
1807         rcu_read_unlock_bh();
1808         in6_dev_put(idev);
1809 }
1810
1811 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1812 {
1813         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1814                 __ipv6_regen_rndid(idev);
1815 }
1816 #endif
1817
1818 /*
1819  *      Add prefix route.
1820  */
1821
1822 static void
1823 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1824                       unsigned long expires, u32 flags)
1825 {
1826         struct fib6_config cfg = {
1827                 .fc_table = RT6_TABLE_PREFIX,
1828                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1829                 .fc_ifindex = dev->ifindex,
1830                 .fc_expires = expires,
1831                 .fc_dst_len = plen,
1832                 .fc_flags = RTF_UP | flags,
1833                 .fc_nlinfo.nl_net = dev_net(dev),
1834                 .fc_protocol = RTPROT_KERNEL,
1835         };
1836
1837         cfg.fc_dst = *pfx;
1838
1839         /* Prevent useless cloning on PtP SIT.
1840            This thing is done here expecting that the whole
1841            class of non-broadcast devices need not cloning.
1842          */
1843 #if IS_ENABLED(CONFIG_IPV6_SIT)
1844         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1845                 cfg.fc_flags |= RTF_NONEXTHOP;
1846 #endif
1847
1848         ip6_route_add(&cfg);
1849 }
1850
1851
1852 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1853                                                   int plen,
1854                                                   const struct net_device *dev,
1855                                                   u32 flags, u32 noflags)
1856 {
1857         struct fib6_node *fn;
1858         struct rt6_info *rt = NULL;
1859         struct fib6_table *table;
1860
1861         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1862         if (table == NULL)
1863                 return NULL;
1864
1865         read_lock_bh(&table->tb6_lock);
1866         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1867         if (!fn)
1868                 goto out;
1869         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1870                 if (rt->dst.dev->ifindex != dev->ifindex)
1871                         continue;
1872                 if ((rt->rt6i_flags & flags) != flags)
1873                         continue;
1874                 if ((rt->rt6i_flags & noflags) != 0)
1875                         continue;
1876                 dst_hold(&rt->dst);
1877                 break;
1878         }
1879 out:
1880         read_unlock_bh(&table->tb6_lock);
1881         return rt;
1882 }
1883
1884
1885 /* Create "default" multicast route to the interface */
1886
1887 static void addrconf_add_mroute(struct net_device *dev)
1888 {
1889         struct fib6_config cfg = {
1890                 .fc_table = RT6_TABLE_LOCAL,
1891                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1892                 .fc_ifindex = dev->ifindex,
1893                 .fc_dst_len = 8,
1894                 .fc_flags = RTF_UP,
1895                 .fc_nlinfo.nl_net = dev_net(dev),
1896         };
1897
1898         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1899
1900         ip6_route_add(&cfg);
1901 }
1902
1903 #if IS_ENABLED(CONFIG_IPV6_SIT)
1904 static void sit_route_add(struct net_device *dev)
1905 {
1906         struct fib6_config cfg = {
1907                 .fc_table = RT6_TABLE_MAIN,
1908                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1909                 .fc_ifindex = dev->ifindex,
1910                 .fc_dst_len = 96,
1911                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1912                 .fc_nlinfo.nl_net = dev_net(dev),
1913         };
1914
1915         /* prefix length - 96 bits "::d.d.d.d" */
1916         ip6_route_add(&cfg);
1917 }
1918 #endif
1919
1920 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1921 {
1922         struct inet6_dev *idev;
1923
1924         ASSERT_RTNL();
1925
1926         idev = ipv6_find_idev(dev);
1927         if (!idev)
1928                 return ERR_PTR(-ENOBUFS);
1929
1930         if (idev->cnf.disable_ipv6)
1931                 return ERR_PTR(-EACCES);
1932
1933         /* Add default multicast route */
1934         if (!(dev->flags & IFF_LOOPBACK))
1935                 addrconf_add_mroute(dev);
1936
1937         return idev;
1938 }
1939
1940 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1941 {
1942         struct prefix_info *pinfo;
1943         __u32 valid_lft;
1944         __u32 prefered_lft;
1945         int addr_type;
1946         struct inet6_dev *in6_dev;
1947         struct net *net = dev_net(dev);
1948
1949         pinfo = (struct prefix_info *) opt;
1950
1951         if (len < sizeof(struct prefix_info)) {
1952                 ADBG(("addrconf: prefix option too short\n"));
1953                 return;
1954         }
1955
1956         /*
1957          *      Validation checks ([ADDRCONF], page 19)
1958          */
1959
1960         addr_type = ipv6_addr_type(&pinfo->prefix);
1961
1962         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1963                 return;
1964
1965         valid_lft = ntohl(pinfo->valid);
1966         prefered_lft = ntohl(pinfo->prefered);
1967
1968         if (prefered_lft > valid_lft) {
1969                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1970                 return;
1971         }
1972
1973         in6_dev = in6_dev_get(dev);
1974
1975         if (in6_dev == NULL) {
1976                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1977                                     dev->name);
1978                 return;
1979         }
1980
1981         /*
1982          *      Two things going on here:
1983          *      1) Add routes for on-link prefixes
1984          *      2) Configure prefixes with the auto flag set
1985          */
1986
1987         if (pinfo->onlink) {
1988                 struct rt6_info *rt;
1989                 unsigned long rt_expires;
1990
1991                 /* Avoid arithmetic overflow. Really, we could
1992                  * save rt_expires in seconds, likely valid_lft,
1993                  * but it would require division in fib gc, that it
1994                  * not good.
1995                  */
1996                 if (HZ > USER_HZ)
1997                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1998                 else
1999                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2000
2001                 if (addrconf_finite_timeout(rt_expires))
2002                         rt_expires *= HZ;
2003
2004                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2005                                                pinfo->prefix_len,
2006                                                dev,
2007                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2008                                                RTF_GATEWAY | RTF_DEFAULT);
2009
2010                 if (rt) {
2011                         /* Autoconf prefix route */
2012                         if (valid_lft == 0) {
2013                                 ip6_del_rt(rt);
2014                                 rt = NULL;
2015                         } else if (addrconf_finite_timeout(rt_expires)) {
2016                                 /* not infinity */
2017                                 rt6_set_expires(rt, jiffies + rt_expires);
2018                         } else {
2019                                 rt6_clean_expires(rt);
2020                         }
2021                 } else if (valid_lft) {
2022                         clock_t expires = 0;
2023                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2024                         if (addrconf_finite_timeout(rt_expires)) {
2025                                 /* not infinity */
2026                                 flags |= RTF_EXPIRES;
2027                                 expires = jiffies_to_clock_t(rt_expires);
2028                         }
2029                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2030                                               dev, expires, flags);
2031                 }
2032                 ip6_rt_put(rt);
2033         }
2034
2035         /* Try to figure out our local address for this prefix */
2036
2037         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2038                 struct inet6_ifaddr *ifp;
2039                 struct in6_addr addr;
2040                 int create = 0, update_lft = 0;
2041
2042                 if (pinfo->prefix_len == 64) {
2043                         memcpy(&addr, &pinfo->prefix, 8);
2044                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2045                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2046                                 in6_dev_put(in6_dev);
2047                                 return;
2048                         }
2049                         goto ok;
2050                 }
2051                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2052                                     pinfo->prefix_len);
2053                 in6_dev_put(in6_dev);
2054                 return;
2055
2056 ok:
2057
2058                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2059
2060                 if (ifp == NULL && valid_lft) {
2061                         int max_addresses = in6_dev->cnf.max_addresses;
2062                         u32 addr_flags = 0;
2063
2064 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2065                         if (in6_dev->cnf.optimistic_dad &&
2066                             !net->ipv6.devconf_all->forwarding && sllao)
2067                                 addr_flags = IFA_F_OPTIMISTIC;
2068 #endif
2069
2070                         /* Do not allow to create too much of autoconfigured
2071                          * addresses; this would be too easy way to crash kernel.
2072                          */
2073                         if (!max_addresses ||
2074                             ipv6_count_addresses(in6_dev) < max_addresses)
2075                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2076                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2077                                                     addr_flags);
2078
2079                         if (IS_ERR_OR_NULL(ifp)) {
2080                                 in6_dev_put(in6_dev);
2081                                 return;
2082                         }
2083
2084                         update_lft = create = 1;
2085                         ifp->cstamp = jiffies;
2086                         addrconf_dad_start(ifp);
2087                 }
2088
2089                 if (ifp) {
2090                         int flags;
2091                         unsigned long now;
2092 #ifdef CONFIG_IPV6_PRIVACY
2093                         struct inet6_ifaddr *ift;
2094 #endif
2095                         u32 stored_lft;
2096
2097                         /* update lifetime (RFC2462 5.5.3 e) */
2098                         spin_lock(&ifp->lock);
2099                         now = jiffies;
2100                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2101                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2102                         else
2103                                 stored_lft = 0;
2104                         if (!update_lft && stored_lft) {
2105                                 if (valid_lft > MIN_VALID_LIFETIME ||
2106                                     valid_lft > stored_lft)
2107                                         update_lft = 1;
2108                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2109                                         /* valid_lft <= stored_lft is always true */
2110                                         /*
2111                                          * RFC 4862 Section 5.5.3e:
2112                                          * "Note that the preferred lifetime of
2113                                          *  the corresponding address is always
2114                                          *  reset to the Preferred Lifetime in
2115                                          *  the received Prefix Information
2116                                          *  option, regardless of whether the
2117                                          *  valid lifetime is also reset or
2118                                          *  ignored."
2119                                          *
2120                                          *  So if the preferred lifetime in
2121                                          *  this advertisement is different
2122                                          *  than what we have stored, but the
2123                                          *  valid lifetime is invalid, just
2124                                          *  reset prefered_lft.
2125                                          *
2126                                          *  We must set the valid lifetime
2127                                          *  to the stored lifetime since we'll
2128                                          *  be updating the timestamp below,
2129                                          *  else we'll set it back to the
2130                                          *  minimum.
2131                                          */
2132                                         if (prefered_lft != ifp->prefered_lft) {
2133                                                 valid_lft = stored_lft;
2134                                                 update_lft = 1;
2135                                         }
2136                                 } else {
2137                                         valid_lft = MIN_VALID_LIFETIME;
2138                                         if (valid_lft < prefered_lft)
2139                                                 prefered_lft = valid_lft;
2140                                         update_lft = 1;
2141                                 }
2142                         }
2143
2144                         if (update_lft) {
2145                                 ifp->valid_lft = valid_lft;
2146                                 ifp->prefered_lft = prefered_lft;
2147                                 ifp->tstamp = now;
2148                                 flags = ifp->flags;
2149                                 ifp->flags &= ~IFA_F_DEPRECATED;
2150                                 spin_unlock(&ifp->lock);
2151
2152                                 if (!(flags&IFA_F_TENTATIVE))
2153                                         ipv6_ifa_notify(0, ifp);
2154                         } else
2155                                 spin_unlock(&ifp->lock);
2156
2157 #ifdef CONFIG_IPV6_PRIVACY
2158                         read_lock_bh(&in6_dev->lock);
2159                         /* update all temporary addresses in the list */
2160                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2161                                             tmp_list) {
2162                                 int age, max_valid, max_prefered;
2163
2164                                 if (ifp != ift->ifpub)
2165                                         continue;
2166
2167                                 /*
2168                                  * RFC 4941 section 3.3:
2169                                  * If a received option will extend the lifetime
2170                                  * of a public address, the lifetimes of
2171                                  * temporary addresses should be extended,
2172                                  * subject to the overall constraint that no
2173                                  * temporary addresses should ever remain
2174                                  * "valid" or "preferred" for a time longer than
2175                                  * (TEMP_VALID_LIFETIME) or
2176                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2177                                  * respectively.
2178                                  */
2179                                 age = (now - ift->cstamp) / HZ;
2180                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2181                                 if (max_valid < 0)
2182                                         max_valid = 0;
2183
2184                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2185                                                in6_dev->cnf.max_desync_factor -
2186                                                age;
2187                                 if (max_prefered < 0)
2188                                         max_prefered = 0;
2189
2190                                 if (valid_lft > max_valid)
2191                                         valid_lft = max_valid;
2192
2193                                 if (prefered_lft > max_prefered)
2194                                         prefered_lft = max_prefered;
2195
2196                                 spin_lock(&ift->lock);
2197                                 flags = ift->flags;
2198                                 ift->valid_lft = valid_lft;
2199                                 ift->prefered_lft = prefered_lft;
2200                                 ift->tstamp = now;
2201                                 if (prefered_lft > 0)
2202                                         ift->flags &= ~IFA_F_DEPRECATED;
2203
2204                                 spin_unlock(&ift->lock);
2205                                 if (!(flags&IFA_F_TENTATIVE))
2206                                         ipv6_ifa_notify(0, ift);
2207                         }
2208
2209                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2210                                 /*
2211                                  * When a new public address is created as
2212                                  * described in [ADDRCONF], also create a new
2213                                  * temporary address. Also create a temporary
2214                                  * address if it's enabled but no temporary
2215                                  * address currently exists.
2216                                  */
2217                                 read_unlock_bh(&in6_dev->lock);
2218                                 ipv6_create_tempaddr(ifp, NULL);
2219                         } else {
2220                                 read_unlock_bh(&in6_dev->lock);
2221                         }
2222 #endif
2223                         in6_ifa_put(ifp);
2224                         addrconf_verify(0);
2225                 }
2226         }
2227         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2228         in6_dev_put(in6_dev);
2229 }
2230
2231 /*
2232  *      Set destination address.
2233  *      Special case for SIT interfaces where we create a new "virtual"
2234  *      device.
2235  */
2236 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2237 {
2238         struct in6_ifreq ireq;
2239         struct net_device *dev;
2240         int err = -EINVAL;
2241
2242         rtnl_lock();
2243
2244         err = -EFAULT;
2245         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2246                 goto err_exit;
2247
2248         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2249
2250         err = -ENODEV;
2251         if (dev == NULL)
2252                 goto err_exit;
2253
2254 #if IS_ENABLED(CONFIG_IPV6_SIT)
2255         if (dev->type == ARPHRD_SIT) {
2256                 const struct net_device_ops *ops = dev->netdev_ops;
2257                 struct ifreq ifr;
2258                 struct ip_tunnel_parm p;
2259
2260                 err = -EADDRNOTAVAIL;
2261                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2262                         goto err_exit;
2263
2264                 memset(&p, 0, sizeof(p));
2265                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2266                 p.iph.saddr = 0;
2267                 p.iph.version = 4;
2268                 p.iph.ihl = 5;
2269                 p.iph.protocol = IPPROTO_IPV6;
2270                 p.iph.ttl = 64;
2271                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2272
2273                 if (ops->ndo_do_ioctl) {
2274                         mm_segment_t oldfs = get_fs();
2275
2276                         set_fs(KERNEL_DS);
2277                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2278                         set_fs(oldfs);
2279                 } else
2280                         err = -EOPNOTSUPP;
2281
2282                 if (err == 0) {
2283                         err = -ENOBUFS;
2284                         dev = __dev_get_by_name(net, p.name);
2285                         if (!dev)
2286                                 goto err_exit;
2287                         err = dev_open(dev);
2288                 }
2289         }
2290 #endif
2291
2292 err_exit:
2293         rtnl_unlock();
2294         return err;
2295 }
2296
2297 /*
2298  *      Manual configuration of address on an interface
2299  */
2300 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2301                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2302                           __u32 valid_lft)
2303 {
2304         struct inet6_ifaddr *ifp;
2305         struct inet6_dev *idev;
2306         struct net_device *dev;
2307         int scope;
2308         u32 flags;
2309         clock_t expires;
2310         unsigned long timeout;
2311
2312         ASSERT_RTNL();
2313
2314         if (plen > 128)
2315                 return -EINVAL;
2316
2317         /* check the lifetime */
2318         if (!valid_lft || prefered_lft > valid_lft)
2319                 return -EINVAL;
2320
2321         dev = __dev_get_by_index(net, ifindex);
2322         if (!dev)
2323                 return -ENODEV;
2324
2325         idev = addrconf_add_dev(dev);
2326         if (IS_ERR(idev))
2327                 return PTR_ERR(idev);
2328
2329         scope = ipv6_addr_scope(pfx);
2330
2331         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2332         if (addrconf_finite_timeout(timeout)) {
2333                 expires = jiffies_to_clock_t(timeout * HZ);
2334                 valid_lft = timeout;
2335                 flags = RTF_EXPIRES;
2336         } else {
2337                 expires = 0;
2338                 flags = 0;
2339                 ifa_flags |= IFA_F_PERMANENT;
2340         }
2341
2342         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2343         if (addrconf_finite_timeout(timeout)) {
2344                 if (timeout == 0)
2345                         ifa_flags |= IFA_F_DEPRECATED;
2346                 prefered_lft = timeout;
2347         }
2348
2349         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2350
2351         if (!IS_ERR(ifp)) {
2352                 spin_lock_bh(&ifp->lock);
2353                 ifp->valid_lft = valid_lft;
2354                 ifp->prefered_lft = prefered_lft;
2355                 ifp->tstamp = jiffies;
2356                 spin_unlock_bh(&ifp->lock);
2357
2358                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2359                                       expires, flags);
2360                 /*
2361                  * Note that section 3.1 of RFC 4429 indicates
2362                  * that the Optimistic flag should not be set for
2363                  * manually configured addresses
2364                  */
2365                 addrconf_dad_start(ifp);
2366                 in6_ifa_put(ifp);
2367                 addrconf_verify(0);
2368                 return 0;
2369         }
2370
2371         return PTR_ERR(ifp);
2372 }
2373
2374 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2375                           unsigned int plen)
2376 {
2377         struct inet6_ifaddr *ifp;
2378         struct inet6_dev *idev;
2379         struct net_device *dev;
2380
2381         if (plen > 128)
2382                 return -EINVAL;
2383
2384         dev = __dev_get_by_index(net, ifindex);
2385         if (!dev)
2386                 return -ENODEV;
2387
2388         if ((idev = __in6_dev_get(dev)) == NULL)
2389                 return -ENXIO;
2390
2391         read_lock_bh(&idev->lock);
2392         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2393                 if (ifp->prefix_len == plen &&
2394                     ipv6_addr_equal(pfx, &ifp->addr)) {
2395                         in6_ifa_hold(ifp);
2396                         read_unlock_bh(&idev->lock);
2397
2398                         ipv6_del_addr(ifp);
2399
2400                         /* If the last address is deleted administratively,
2401                            disable IPv6 on this interface.
2402                          */
2403                         if (list_empty(&idev->addr_list))
2404                                 addrconf_ifdown(idev->dev, 1);
2405                         return 0;
2406                 }
2407         }
2408         read_unlock_bh(&idev->lock);
2409         return -EADDRNOTAVAIL;
2410 }
2411
2412
2413 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2414 {
2415         struct in6_ifreq ireq;
2416         int err;
2417
2418         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2419                 return -EPERM;
2420
2421         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2422                 return -EFAULT;
2423
2424         rtnl_lock();
2425         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2426                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2427                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2428         rtnl_unlock();
2429         return err;
2430 }
2431
2432 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2433 {
2434         struct in6_ifreq ireq;
2435         int err;
2436
2437         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2438                 return -EPERM;
2439
2440         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2441                 return -EFAULT;
2442
2443         rtnl_lock();
2444         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2445                              ireq.ifr6_prefixlen);
2446         rtnl_unlock();
2447         return err;
2448 }
2449
2450 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2451                      int plen, int scope)
2452 {
2453         struct inet6_ifaddr *ifp;
2454
2455         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2456         if (!IS_ERR(ifp)) {
2457                 spin_lock_bh(&ifp->lock);
2458                 ifp->flags &= ~IFA_F_TENTATIVE;
2459                 spin_unlock_bh(&ifp->lock);
2460                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2461                 in6_ifa_put(ifp);
2462         }
2463 }
2464
2465 #if IS_ENABLED(CONFIG_IPV6_SIT)
2466 static void sit_add_v4_addrs(struct inet6_dev *idev)
2467 {
2468         struct in6_addr addr;
2469         struct net_device *dev;
2470         struct net *net = dev_net(idev->dev);
2471         int scope;
2472
2473         ASSERT_RTNL();
2474
2475         memset(&addr, 0, sizeof(struct in6_addr));
2476         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2477
2478         if (idev->dev->flags&IFF_POINTOPOINT) {
2479                 addr.s6_addr32[0] = htonl(0xfe800000);
2480                 scope = IFA_LINK;
2481         } else {
2482                 scope = IPV6_ADDR_COMPATv4;
2483         }
2484
2485         if (addr.s6_addr32[3]) {
2486                 add_addr(idev, &addr, 128, scope);
2487                 return;
2488         }
2489
2490         for_each_netdev(net, dev) {
2491                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2492                 if (in_dev && (dev->flags & IFF_UP)) {
2493                         struct in_ifaddr *ifa;
2494
2495                         int flag = scope;
2496
2497                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2498                                 int plen;
2499
2500                                 addr.s6_addr32[3] = ifa->ifa_local;
2501
2502                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2503                                         continue;
2504                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2505                                         if (idev->dev->flags&IFF_POINTOPOINT)
2506                                                 continue;
2507                                         flag |= IFA_HOST;
2508                                 }
2509                                 if (idev->dev->flags&IFF_POINTOPOINT)
2510                                         plen = 64;
2511                                 else
2512                                         plen = 96;
2513
2514                                 add_addr(idev, &addr, plen, flag);
2515                         }
2516                 }
2517         }
2518 }
2519 #endif
2520
2521 static void init_loopback(struct net_device *dev)
2522 {
2523         struct inet6_dev  *idev;
2524
2525         /* ::1 */
2526
2527         ASSERT_RTNL();
2528
2529         if ((idev = ipv6_find_idev(dev)) == NULL) {
2530                 pr_debug("%s: add_dev failed\n", __func__);
2531                 return;
2532         }
2533
2534         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2535 }
2536
2537 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2538 {
2539         struct inet6_ifaddr *ifp;
2540         u32 addr_flags = IFA_F_PERMANENT;
2541
2542 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2543         if (idev->cnf.optimistic_dad &&
2544             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2545                 addr_flags |= IFA_F_OPTIMISTIC;
2546 #endif
2547
2548
2549         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2550         if (!IS_ERR(ifp)) {
2551                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2552                 addrconf_dad_start(ifp);
2553                 in6_ifa_put(ifp);
2554         }
2555 }
2556
2557 static void addrconf_dev_config(struct net_device *dev)
2558 {
2559         struct in6_addr addr;
2560         struct inet6_dev *idev;
2561
2562         ASSERT_RTNL();
2563
2564         if ((dev->type != ARPHRD_ETHER) &&
2565             (dev->type != ARPHRD_FDDI) &&
2566             (dev->type != ARPHRD_ARCNET) &&
2567             (dev->type != ARPHRD_INFINIBAND) &&
2568             (dev->type != ARPHRD_IEEE802154)) {
2569                 /* Alas, we support only Ethernet autoconfiguration. */
2570                 return;
2571         }
2572
2573         idev = addrconf_add_dev(dev);
2574         if (IS_ERR(idev))
2575                 return;
2576
2577         memset(&addr, 0, sizeof(struct in6_addr));
2578         addr.s6_addr32[0] = htonl(0xFE800000);
2579
2580         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2581                 addrconf_add_linklocal(idev, &addr);
2582 }
2583
2584 #if IS_ENABLED(CONFIG_IPV6_SIT)
2585 static void addrconf_sit_config(struct net_device *dev)
2586 {
2587         struct inet6_dev *idev;
2588
2589         ASSERT_RTNL();
2590
2591         /*
2592          * Configure the tunnel with one of our IPv4
2593          * addresses... we should configure all of
2594          * our v4 addrs in the tunnel
2595          */
2596
2597         if ((idev = ipv6_find_idev(dev)) == NULL) {
2598                 pr_debug("%s: add_dev failed\n", __func__);
2599                 return;
2600         }
2601
2602         if (dev->priv_flags & IFF_ISATAP) {
2603                 struct in6_addr addr;
2604
2605                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2606                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2607                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2608                         addrconf_add_linklocal(idev, &addr);
2609                 return;
2610         }
2611
2612         sit_add_v4_addrs(idev);
2613
2614         if (dev->flags&IFF_POINTOPOINT)
2615                 addrconf_add_mroute(dev);
2616         else
2617                 sit_route_add(dev);
2618 }
2619 #endif
2620
2621 #if IS_ENABLED(CONFIG_NET_IPGRE)
2622 static void addrconf_gre_config(struct net_device *dev)
2623 {
2624         struct inet6_dev *idev;
2625         struct in6_addr addr;
2626
2627         pr_info("%s(%s)\n", __func__, dev->name);
2628
2629         ASSERT_RTNL();
2630
2631         if ((idev = ipv6_find_idev(dev)) == NULL) {
2632                 pr_debug("%s: add_dev failed\n", __func__);
2633                 return;
2634         }
2635
2636         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2637         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2638
2639         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2640                 addrconf_add_linklocal(idev, &addr);
2641 }
2642 #endif
2643
2644 static inline int
2645 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2646 {
2647         struct in6_addr lladdr;
2648
2649         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2650                 addrconf_add_linklocal(idev, &lladdr);
2651                 return 0;
2652         }
2653         return -1;
2654 }
2655
2656 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2657 {
2658         struct net_device *link_dev;
2659         struct net *net = dev_net(idev->dev);
2660
2661         /* first try to inherit the link-local address from the link device */
2662         if (idev->dev->iflink &&
2663             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2664                 if (!ipv6_inherit_linklocal(idev, link_dev))
2665                         return;
2666         }
2667         /* then try to inherit it from any device */
2668         for_each_netdev(net, link_dev) {
2669                 if (!ipv6_inherit_linklocal(idev, link_dev))
2670                         return;
2671         }
2672         pr_debug("init ip6-ip6: add_linklocal failed\n");
2673 }
2674
2675 /*
2676  * Autoconfigure tunnel with a link-local address so routing protocols,
2677  * DHCPv6, MLD etc. can be run over the virtual link
2678  */
2679
2680 static void addrconf_ip6_tnl_config(struct net_device *dev)
2681 {
2682         struct inet6_dev *idev;
2683
2684         ASSERT_RTNL();
2685
2686         idev = addrconf_add_dev(dev);
2687         if (IS_ERR(idev)) {
2688                 pr_debug("init ip6-ip6: add_dev failed\n");
2689                 return;
2690         }
2691         ip6_tnl_add_linklocal(idev);
2692 }
2693
2694 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2695                            void *data)
2696 {
2697         struct net_device *dev = (struct net_device *) data;
2698         struct inet6_dev *idev = __in6_dev_get(dev);
2699         int run_pending = 0;
2700         int err;
2701
2702         switch (event) {
2703         case NETDEV_REGISTER:
2704                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2705                         idev = ipv6_add_dev(dev);
2706                         if (!idev)
2707                                 return notifier_from_errno(-ENOMEM);
2708                 }
2709                 break;
2710
2711         case NETDEV_UP:
2712         case NETDEV_CHANGE:
2713                 if (dev->flags & IFF_SLAVE)
2714                         break;
2715
2716                 if (event == NETDEV_UP) {
2717                         if (!addrconf_qdisc_ok(dev)) {
2718                                 /* device is not ready yet. */
2719                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2720                                         dev->name);
2721                                 break;
2722                         }
2723
2724                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2725                                 idev = ipv6_add_dev(dev);
2726
2727                         if (idev) {
2728                                 idev->if_flags |= IF_READY;
2729                                 run_pending = 1;
2730                         }
2731                 } else {
2732                         if (!addrconf_qdisc_ok(dev)) {
2733                                 /* device is still not ready. */
2734                                 break;
2735                         }
2736
2737                         if (idev) {
2738                                 if (idev->if_flags & IF_READY)
2739                                         /* device is already configured. */
2740                                         break;
2741                                 idev->if_flags |= IF_READY;
2742                         }
2743
2744                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2745                                 dev->name);
2746
2747                         run_pending = 1;
2748                 }
2749
2750                 switch (dev->type) {
2751 #if IS_ENABLED(CONFIG_IPV6_SIT)
2752                 case ARPHRD_SIT:
2753                         addrconf_sit_config(dev);
2754                         break;
2755 #endif
2756 #if IS_ENABLED(CONFIG_NET_IPGRE)
2757                 case ARPHRD_IPGRE:
2758                         addrconf_gre_config(dev);
2759                         break;
2760 #endif
2761                 case ARPHRD_TUNNEL6:
2762                         addrconf_ip6_tnl_config(dev);
2763                         break;
2764                 case ARPHRD_LOOPBACK:
2765                         init_loopback(dev);
2766                         break;
2767
2768                 default:
2769                         addrconf_dev_config(dev);
2770                         break;
2771                 }
2772
2773                 if (idev) {
2774                         if (run_pending)
2775                                 addrconf_dad_run(idev);
2776
2777                         /*
2778                          * If the MTU changed during the interface down,
2779                          * when the interface up, the changed MTU must be
2780                          * reflected in the idev as well as routers.
2781                          */
2782                         if (idev->cnf.mtu6 != dev->mtu &&
2783                             dev->mtu >= IPV6_MIN_MTU) {
2784                                 rt6_mtu_change(dev, dev->mtu);
2785                                 idev->cnf.mtu6 = dev->mtu;
2786                         }
2787                         idev->tstamp = jiffies;
2788                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2789
2790                         /*
2791                          * If the changed mtu during down is lower than
2792                          * IPV6_MIN_MTU stop IPv6 on this interface.
2793                          */
2794                         if (dev->mtu < IPV6_MIN_MTU)
2795                                 addrconf_ifdown(dev, 1);
2796                 }
2797                 break;
2798
2799         case NETDEV_CHANGEMTU:
2800                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2801                         rt6_mtu_change(dev, dev->mtu);
2802                         idev->cnf.mtu6 = dev->mtu;
2803                         break;
2804                 }
2805
2806                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2807                         idev = ipv6_add_dev(dev);
2808                         if (idev)
2809                                 break;
2810                 }
2811
2812                 /*
2813                  * MTU falled under IPV6_MIN_MTU.
2814                  * Stop IPv6 on this interface.
2815                  */
2816
2817         case NETDEV_DOWN:
2818         case NETDEV_UNREGISTER:
2819                 /*
2820                  *      Remove all addresses from this interface.
2821                  */
2822                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2823                 break;
2824
2825         case NETDEV_CHANGENAME:
2826                 if (idev) {
2827                         snmp6_unregister_dev(idev);
2828                         addrconf_sysctl_unregister(idev);
2829                         addrconf_sysctl_register(idev);
2830                         err = snmp6_register_dev(idev);
2831                         if (err)
2832                                 return notifier_from_errno(err);
2833                 }
2834                 break;
2835
2836         case NETDEV_PRE_TYPE_CHANGE:
2837         case NETDEV_POST_TYPE_CHANGE:
2838                 addrconf_type_change(dev, event);
2839                 break;
2840         }
2841
2842         return NOTIFY_OK;
2843 }
2844
2845 /*
2846  *      addrconf module should be notified of a device going up
2847  */
2848 static struct notifier_block ipv6_dev_notf = {
2849         .notifier_call = addrconf_notify,
2850 };
2851
2852 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2853 {
2854         struct inet6_dev *idev;
2855         ASSERT_RTNL();
2856
2857         idev = __in6_dev_get(dev);
2858
2859         if (event == NETDEV_POST_TYPE_CHANGE)
2860                 ipv6_mc_remap(idev);
2861         else if (event == NETDEV_PRE_TYPE_CHANGE)
2862                 ipv6_mc_unmap(idev);
2863 }
2864
2865 static int addrconf_ifdown(struct net_device *dev, int how)
2866 {
2867         struct net *net = dev_net(dev);
2868         struct inet6_dev *idev;
2869         struct inet6_ifaddr *ifa;
2870         int state, i;
2871
2872         ASSERT_RTNL();
2873
2874         rt6_ifdown(net, dev);
2875         neigh_ifdown(&nd_tbl, dev);
2876
2877         idev = __in6_dev_get(dev);
2878         if (idev == NULL)
2879                 return -ENODEV;
2880
2881         /*
2882          * Step 1: remove reference to ipv6 device from parent device.
2883          *         Do not dev_put!
2884          */
2885         if (how) {
2886                 idev->dead = 1;
2887
2888                 /* protected by rtnl_lock */
2889                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2890
2891                 /* Step 1.5: remove snmp6 entry */
2892                 snmp6_unregister_dev(idev);
2893
2894         }
2895
2896         /* Step 2: clear hash table */
2897         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2898                 struct hlist_head *h = &inet6_addr_lst[i];
2899                 struct hlist_node *n;
2900
2901                 spin_lock_bh(&addrconf_hash_lock);
2902         restart:
2903                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2904                         if (ifa->idev == idev) {
2905                                 hlist_del_init_rcu(&ifa->addr_lst);
2906                                 addrconf_del_timer(ifa);
2907                                 goto restart;
2908                         }
2909                 }
2910                 spin_unlock_bh(&addrconf_hash_lock);
2911         }
2912
2913         write_lock_bh(&idev->lock);
2914
2915         /* Step 2: clear flags for stateless addrconf */
2916         if (!how)
2917                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2918
2919 #ifdef CONFIG_IPV6_PRIVACY
2920         if (how && del_timer(&idev->regen_timer))
2921                 in6_dev_put(idev);
2922
2923         /* Step 3: clear tempaddr list */
2924         while (!list_empty(&idev->tempaddr_list)) {
2925                 ifa = list_first_entry(&idev->tempaddr_list,
2926                                        struct inet6_ifaddr, tmp_list);
2927                 list_del(&ifa->tmp_list);
2928                 write_unlock_bh(&idev->lock);
2929                 spin_lock_bh(&ifa->lock);
2930
2931                 if (ifa->ifpub) {
2932                         in6_ifa_put(ifa->ifpub);
2933                         ifa->ifpub = NULL;
2934                 }
2935                 spin_unlock_bh(&ifa->lock);
2936                 in6_ifa_put(ifa);
2937                 write_lock_bh(&idev->lock);
2938         }
2939 #endif
2940
2941         while (!list_empty(&idev->addr_list)) {
2942                 ifa = list_first_entry(&idev->addr_list,
2943                                        struct inet6_ifaddr, if_list);
2944                 addrconf_del_timer(ifa);
2945
2946                 list_del(&ifa->if_list);
2947
2948                 write_unlock_bh(&idev->lock);
2949
2950                 spin_lock_bh(&ifa->state_lock);
2951                 state = ifa->state;
2952                 ifa->state = INET6_IFADDR_STATE_DEAD;
2953                 spin_unlock_bh(&ifa->state_lock);
2954
2955                 if (state != INET6_IFADDR_STATE_DEAD) {
2956                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2957                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2958                 }
2959                 in6_ifa_put(ifa);
2960
2961                 write_lock_bh(&idev->lock);
2962         }
2963
2964         write_unlock_bh(&idev->lock);
2965
2966         /* Step 5: Discard multicast list */
2967         if (how)
2968                 ipv6_mc_destroy_dev(idev);
2969         else
2970                 ipv6_mc_down(idev);
2971
2972         idev->tstamp = jiffies;
2973
2974         /* Last: Shot the device (if unregistered) */
2975         if (how) {
2976                 addrconf_sysctl_unregister(idev);
2977                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2978                 neigh_ifdown(&nd_tbl, dev);
2979                 in6_dev_put(idev);
2980         }
2981         return 0;
2982 }
2983
2984 static void addrconf_rs_timer(unsigned long data)
2985 {
2986         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2987         struct inet6_dev *idev = ifp->idev;
2988
2989         read_lock(&idev->lock);
2990         if (idev->dead || !(idev->if_flags & IF_READY))
2991                 goto out;
2992
2993         if (!ipv6_accept_ra(idev))
2994                 goto out;
2995
2996         /* Announcement received after solicitation was sent */
2997         if (idev->if_flags & IF_RA_RCVD)
2998                 goto out;
2999
3000         spin_lock(&ifp->lock);
3001         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3002                 /* The wait after the last probe can be shorter */
3003                 addrconf_mod_timer(ifp, AC_RS,
3004                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3005                                    idev->cnf.rtr_solicit_delay :
3006                                    idev->cnf.rtr_solicit_interval);
3007                 spin_unlock(&ifp->lock);
3008
3009                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3010         } else {
3011                 spin_unlock(&ifp->lock);
3012                 /*
3013                  * Note: we do not support deprecated "all on-link"
3014                  * assumption any longer.
3015                  */
3016                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3017         }
3018
3019 out:
3020         read_unlock(&idev->lock);
3021         in6_ifa_put(ifp);
3022 }
3023
3024 /*
3025  *      Duplicate Address Detection
3026  */
3027 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3028 {
3029         unsigned long rand_num;
3030         struct inet6_dev *idev = ifp->idev;
3031
3032         if (ifp->flags & IFA_F_OPTIMISTIC)
3033                 rand_num = 0;
3034         else
3035                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3036
3037         ifp->probes = idev->cnf.dad_transmits;
3038         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3039 }
3040
3041 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3042 {
3043         struct inet6_dev *idev = ifp->idev;
3044         struct net_device *dev = idev->dev;
3045
3046         addrconf_join_solict(dev, &ifp->addr);
3047
3048         net_srandom(ifp->addr.s6_addr32[3]);
3049
3050         read_lock_bh(&idev->lock);
3051         spin_lock(&ifp->lock);
3052         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3053                 goto out;
3054
3055         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3056             idev->cnf.accept_dad < 1 ||
3057             !(ifp->flags&IFA_F_TENTATIVE) ||
3058             ifp->flags & IFA_F_NODAD) {
3059                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3060                 spin_unlock(&ifp->lock);
3061                 read_unlock_bh(&idev->lock);
3062
3063                 addrconf_dad_completed(ifp);
3064                 return;
3065         }
3066
3067         if (!(idev->if_flags & IF_READY)) {
3068                 spin_unlock(&ifp->lock);
3069                 read_unlock_bh(&idev->lock);
3070                 /*
3071                  * If the device is not ready:
3072                  * - keep it tentative if it is a permanent address.
3073                  * - otherwise, kill it.
3074                  */
3075                 in6_ifa_hold(ifp);
3076                 addrconf_dad_stop(ifp, 0);
3077                 return;
3078         }
3079
3080         /*
3081          * Optimistic nodes can start receiving
3082          * Frames right away
3083          */
3084         if (ifp->flags & IFA_F_OPTIMISTIC)
3085                 ip6_ins_rt(ifp->rt);
3086
3087         addrconf_dad_kick(ifp);
3088 out:
3089         spin_unlock(&ifp->lock);
3090         read_unlock_bh(&idev->lock);
3091 }
3092
3093 static void addrconf_dad_timer(unsigned long data)
3094 {
3095         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3096         struct inet6_dev *idev = ifp->idev;
3097         struct in6_addr mcaddr;
3098
3099         if (!ifp->probes && addrconf_dad_end(ifp))
3100                 goto out;
3101
3102         read_lock(&idev->lock);
3103         if (idev->dead || !(idev->if_flags & IF_READY)) {
3104                 read_unlock(&idev->lock);
3105                 goto out;
3106         }
3107
3108         spin_lock(&ifp->lock);
3109         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3110                 spin_unlock(&ifp->lock);
3111                 read_unlock(&idev->lock);
3112                 goto out;
3113         }
3114
3115         if (ifp->probes == 0) {
3116                 /*
3117                  * DAD was successful
3118                  */
3119
3120                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3121                 spin_unlock(&ifp->lock);
3122                 read_unlock(&idev->lock);
3123
3124                 addrconf_dad_completed(ifp);
3125
3126                 goto out;
3127         }
3128
3129         ifp->probes--;
3130         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3131         spin_unlock(&ifp->lock);
3132         read_unlock(&idev->lock);
3133
3134         /* send a neighbour solicitation for our addr */
3135         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3136         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3137 out:
3138         in6_ifa_put(ifp);
3139 }
3140
3141 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3142 {
3143         struct net_device *dev = ifp->idev->dev;
3144
3145         /*
3146          *      Configure the address for reception. Now it is valid.
3147          */
3148
3149         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3150
3151         /* If added prefix is link local and we are prepared to process
3152            router advertisements, start sending router solicitations.
3153          */
3154
3155         if (ipv6_accept_ra(ifp->idev) &&
3156             ifp->idev->cnf.rtr_solicits > 0 &&
3157             (dev->flags&IFF_LOOPBACK) == 0 &&
3158             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3159                 /*
3160                  *      If a host as already performed a random delay
3161                  *      [...] as part of DAD [...] there is no need
3162                  *      to delay again before sending the first RS
3163                  */
3164                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3165
3166                 spin_lock_bh(&ifp->lock);
3167                 ifp->probes = 1;
3168                 ifp->idev->if_flags |= IF_RS_SENT;
3169                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3170                 spin_unlock_bh(&ifp->lock);
3171         }
3172 }
3173
3174 static void addrconf_dad_run(struct inet6_dev *idev)
3175 {
3176         struct inet6_ifaddr *ifp;
3177
3178         read_lock_bh(&idev->lock);
3179         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3180                 spin_lock(&ifp->lock);
3181                 if (ifp->flags & IFA_F_TENTATIVE &&
3182                     ifp->state == INET6_IFADDR_STATE_DAD)
3183                         addrconf_dad_kick(ifp);
3184                 spin_unlock(&ifp->lock);
3185         }
3186         read_unlock_bh(&idev->lock);
3187 }
3188
3189 #ifdef CONFIG_PROC_FS
3190 struct if6_iter_state {
3191         struct seq_net_private p;
3192         int bucket;
3193         int offset;
3194 };
3195
3196 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3197 {
3198         struct inet6_ifaddr *ifa = NULL;
3199         struct if6_iter_state *state = seq->private;
3200         struct net *net = seq_file_net(seq);
3201         int p = 0;
3202
3203         /* initial bucket if pos is 0 */
3204         if (pos == 0) {
3205                 state->bucket = 0;
3206                 state->offset = 0;
3207         }
3208
3209         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3210                 struct hlist_node *n;
3211                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3212                                          addr_lst) {
3213                         if (!net_eq(dev_net(ifa->idev->dev), net))
3214                                 continue;
3215                         /* sync with offset */
3216                         if (p < state->offset) {
3217                                 p++;
3218                                 continue;
3219                         }
3220                         state->offset++;
3221                         return ifa;
3222                 }
3223
3224                 /* prepare for next bucket */
3225                 state->offset = 0;
3226                 p = 0;
3227         }
3228         return NULL;
3229 }
3230
3231 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3232                                          struct inet6_ifaddr *ifa)
3233 {
3234         struct if6_iter_state *state = seq->private;
3235         struct net *net = seq_file_net(seq);
3236         struct hlist_node *n = &ifa->addr_lst;
3237
3238         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3239                 if (!net_eq(dev_net(ifa->idev->dev), net))
3240                         continue;
3241                 state->offset++;
3242                 return ifa;
3243         }
3244
3245         while (++state->bucket < IN6_ADDR_HSIZE) {
3246                 state->offset = 0;
3247                 hlist_for_each_entry_rcu_bh(ifa, n,
3248                                      &inet6_addr_lst[state->bucket], addr_lst) {
3249                         if (!net_eq(dev_net(ifa->idev->dev), net))
3250                                 continue;
3251                         state->offset++;
3252                         return ifa;
3253                 }
3254         }
3255
3256         return NULL;
3257 }
3258
3259 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3260         __acquires(rcu_bh)
3261 {
3262         rcu_read_lock_bh();
3263         return if6_get_first(seq, *pos);
3264 }
3265
3266 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3267 {
3268         struct inet6_ifaddr *ifa;
3269
3270         ifa = if6_get_next(seq, v);
3271         ++*pos;
3272         return ifa;
3273 }
3274
3275 static void if6_seq_stop(struct seq_file *seq, void *v)
3276         __releases(rcu_bh)
3277 {
3278         rcu_read_unlock_bh();
3279 }
3280
3281 static int if6_seq_show(struct seq_file *seq, void *v)
3282 {
3283         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3284         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3285                    &ifp->addr,
3286                    ifp->idev->dev->ifindex,
3287                    ifp->prefix_len,
3288                    ifp->scope,
3289                    ifp->flags,
3290                    ifp->idev->dev->name);
3291         return 0;
3292 }
3293
3294 static const struct seq_operations if6_seq_ops = {
3295         .start  = if6_seq_start,
3296         .next   = if6_seq_next,
3297         .show   = if6_seq_show,
3298         .stop   = if6_seq_stop,
3299 };
3300
3301 static int if6_seq_open(struct inode *inode, struct file *file)
3302 {
3303         return seq_open_net(inode, file, &if6_seq_ops,
3304                             sizeof(struct if6_iter_state));
3305 }
3306
3307 static const struct file_operations if6_fops = {
3308         .owner          = THIS_MODULE,
3309         .open           = if6_seq_open,
3310         .read           = seq_read,
3311         .llseek         = seq_lseek,
3312         .release        = seq_release_net,
3313 };
3314
3315 static int __net_init if6_proc_net_init(struct net *net)
3316 {
3317         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3318                 return -ENOMEM;
3319         return 0;
3320 }
3321
3322 static void __net_exit if6_proc_net_exit(struct net *net)
3323 {
3324        proc_net_remove(net, "if_inet6");
3325 }
3326
3327 static struct pernet_operations if6_proc_net_ops = {
3328        .init = if6_proc_net_init,
3329        .exit = if6_proc_net_exit,
3330 };
3331
3332 int __init if6_proc_init(void)
3333 {
3334         return register_pernet_subsys(&if6_proc_net_ops);
3335 }
3336
3337 void if6_proc_exit(void)
3338 {
3339         unregister_pernet_subsys(&if6_proc_net_ops);
3340 }
3341 #endif  /* CONFIG_PROC_FS */
3342
3343 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3344 /* Check if address is a home address configured on any interface. */
3345 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3346 {
3347         int ret = 0;
3348         struct inet6_ifaddr *ifp = NULL;
3349         struct hlist_node *n;
3350         unsigned int hash = inet6_addr_hash(addr);
3351
3352         rcu_read_lock_bh();
3353         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3354                 if (!net_eq(dev_net(ifp->idev->dev), net))
3355                         continue;
3356                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3357                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3358                         ret = 1;
3359                         break;
3360                 }
3361         }
3362         rcu_read_unlock_bh();
3363         return ret;
3364 }
3365 #endif
3366
3367 /*
3368  *      Periodic address status verification
3369  */
3370
3371 static void addrconf_verify(unsigned long foo)
3372 {
3373         unsigned long now, next, next_sec, next_sched;
3374         struct inet6_ifaddr *ifp;
3375         struct hlist_node *node;
3376         int i;
3377
3378         rcu_read_lock_bh();
3379         spin_lock(&addrconf_verify_lock);
3380         now = jiffies;
3381         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3382
3383         del_timer(&addr_chk_timer);
3384
3385         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3386 restart:
3387                 hlist_for_each_entry_rcu_bh(ifp, node,
3388                                          &inet6_addr_lst[i], addr_lst) {
3389                         unsigned long age;
3390
3391                         if (ifp->flags & IFA_F_PERMANENT)
3392                                 continue;
3393
3394                         spin_lock(&ifp->lock);
3395                         /* We try to batch several events at once. */
3396                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3397
3398                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3399                             age >= ifp->valid_lft) {
3400                                 spin_unlock(&ifp->lock);
3401                                 in6_ifa_hold(ifp);
3402                                 ipv6_del_addr(ifp);
3403                                 goto restart;
3404                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3405                                 spin_unlock(&ifp->lock);
3406                                 continue;
3407                         } else if (age >= ifp->prefered_lft) {
3408                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3409                                 int deprecate = 0;
3410
3411                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3412                                         deprecate = 1;
3413                                         ifp->flags |= IFA_F_DEPRECATED;
3414                                 }
3415
3416                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3417                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3418
3419                                 spin_unlock(&ifp->lock);
3420
3421                                 if (deprecate) {
3422                                         in6_ifa_hold(ifp);
3423
3424                                         ipv6_ifa_notify(0, ifp);
3425                                         in6_ifa_put(ifp);
3426                                         goto restart;
3427                                 }
3428 #ifdef CONFIG_IPV6_PRIVACY
3429                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3430                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3431                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3432                                         ifp->idev->cnf.dad_transmits *
3433                                         ifp->idev->nd_parms->retrans_time / HZ;
3434
3435                                 if (age >= ifp->prefered_lft - regen_advance) {
3436                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3437                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3438                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3439                                         if (!ifp->regen_count && ifpub) {
3440                                                 ifp->regen_count++;
3441                                                 in6_ifa_hold(ifp);
3442                                                 in6_ifa_hold(ifpub);
3443                                                 spin_unlock(&ifp->lock);
3444
3445                                                 spin_lock(&ifpub->lock);
3446                                                 ifpub->regen_count = 0;
3447                                                 spin_unlock(&ifpub->lock);
3448                                                 ipv6_create_tempaddr(ifpub, ifp);
3449                                                 in6_ifa_put(ifpub);
3450                                                 in6_ifa_put(ifp);
3451                                                 goto restart;
3452                                         }
3453                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3454                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3455                                 spin_unlock(&ifp->lock);
3456 #endif
3457                         } else {
3458                                 /* ifp->prefered_lft <= ifp->valid_lft */
3459                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3460                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3461                                 spin_unlock(&ifp->lock);
3462                         }
3463                 }
3464         }
3465
3466         next_sec = round_jiffies_up(next);
3467         next_sched = next;
3468
3469         /* If rounded timeout is accurate enough, accept it. */
3470         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3471                 next_sched = next_sec;
3472
3473         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3474         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3475                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3476
3477         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3478               now, next, next_sec, next_sched));
3479
3480         addr_chk_timer.expires = next_sched;
3481         add_timer(&addr_chk_timer);
3482         spin_unlock(&addrconf_verify_lock);
3483         rcu_read_unlock_bh();
3484 }
3485
3486 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3487 {
3488         struct in6_addr *pfx = NULL;
3489
3490         if (addr)
3491                 pfx = nla_data(addr);
3492
3493         if (local) {
3494                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3495                         pfx = NULL;
3496                 else
3497                         pfx = nla_data(local);
3498         }
3499
3500         return pfx;
3501 }
3502
3503 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3504         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3505         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3506         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3507 };
3508
3509 static int
3510 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3511 {
3512         struct net *net = sock_net(skb->sk);
3513         struct ifaddrmsg *ifm;
3514         struct nlattr *tb[IFA_MAX+1];
3515         struct in6_addr *pfx;
3516         int err;
3517
3518         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3519         if (err < 0)
3520                 return err;
3521
3522         ifm = nlmsg_data(nlh);
3523         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3524         if (pfx == NULL)
3525                 return -EINVAL;
3526
3527         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3528 }
3529
3530 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3531                              u32 prefered_lft, u32 valid_lft)
3532 {
3533         u32 flags;
3534         clock_t expires;
3535         unsigned long timeout;
3536
3537         if (!valid_lft || (prefered_lft > valid_lft))
3538                 return -EINVAL;
3539
3540         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3541         if (addrconf_finite_timeout(timeout)) {
3542                 expires = jiffies_to_clock_t(timeout * HZ);
3543                 valid_lft = timeout;
3544                 flags = RTF_EXPIRES;
3545         } else {
3546                 expires = 0;
3547                 flags = 0;
3548                 ifa_flags |= IFA_F_PERMANENT;
3549         }
3550
3551         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3552         if (addrconf_finite_timeout(timeout)) {
3553                 if (timeout == 0)
3554                         ifa_flags |= IFA_F_DEPRECATED;
3555                 prefered_lft = timeout;
3556         }
3557
3558         spin_lock_bh(&ifp->lock);
3559         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3560         ifp->tstamp = jiffies;
3561         ifp->valid_lft = valid_lft;
3562         ifp->prefered_lft = prefered_lft;
3563
3564         spin_unlock_bh(&ifp->lock);
3565         if (!(ifp->flags&IFA_F_TENTATIVE))
3566                 ipv6_ifa_notify(0, ifp);
3567
3568         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3569                               expires, flags);
3570         addrconf_verify(0);
3571
3572         return 0;
3573 }
3574
3575 static int
3576 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3577 {
3578         struct net *net = sock_net(skb->sk);
3579         struct ifaddrmsg *ifm;
3580         struct nlattr *tb[IFA_MAX+1];
3581         struct in6_addr *pfx;
3582         struct inet6_ifaddr *ifa;
3583         struct net_device *dev;
3584         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3585         u8 ifa_flags;
3586         int err;
3587
3588         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3589         if (err < 0)
3590                 return err;
3591
3592         ifm = nlmsg_data(nlh);
3593         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3594         if (pfx == NULL)
3595                 return -EINVAL;
3596
3597         if (tb[IFA_CACHEINFO]) {
3598                 struct ifa_cacheinfo *ci;
3599
3600                 ci = nla_data(tb[IFA_CACHEINFO]);
3601                 valid_lft = ci->ifa_valid;
3602                 preferred_lft = ci->ifa_prefered;
3603         } else {
3604                 preferred_lft = INFINITY_LIFE_TIME;
3605                 valid_lft = INFINITY_LIFE_TIME;
3606         }
3607
3608         dev =  __dev_get_by_index(net, ifm->ifa_index);
3609         if (dev == NULL)
3610                 return -ENODEV;
3611
3612         /* We ignore other flags so far. */
3613         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3614
3615         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3616         if (ifa == NULL) {
3617                 /*
3618                  * It would be best to check for !NLM_F_CREATE here but
3619                  * userspace alreay relies on not having to provide this.
3620                  */
3621                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3622                                       ifm->ifa_prefixlen, ifa_flags,
3623                                       preferred_lft, valid_lft);
3624         }
3625
3626         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3627             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3628                 err = -EEXIST;
3629         else
3630                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3631
3632         in6_ifa_put(ifa);
3633
3634         return err;
3635 }
3636
3637 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3638                           u8 scope, int ifindex)
3639 {
3640         struct ifaddrmsg *ifm;
3641
3642         ifm = nlmsg_data(nlh);
3643         ifm->ifa_family = AF_INET6;
3644         ifm->ifa_prefixlen = prefixlen;
3645         ifm->ifa_flags = flags;
3646         ifm->ifa_scope = scope;
3647         ifm->ifa_index = ifindex;
3648 }
3649
3650 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3651                          unsigned long tstamp, u32 preferred, u32 valid)
3652 {
3653         struct ifa_cacheinfo ci;
3654
3655         ci.cstamp = cstamp_delta(cstamp);
3656         ci.tstamp = cstamp_delta(tstamp);
3657         ci.ifa_prefered = preferred;
3658         ci.ifa_valid = valid;
3659
3660         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3661 }
3662
3663 static inline int rt_scope(int ifa_scope)
3664 {
3665         if (ifa_scope & IFA_HOST)
3666                 return RT_SCOPE_HOST;
3667         else if (ifa_scope & IFA_LINK)
3668                 return RT_SCOPE_LINK;
3669         else if (ifa_scope & IFA_SITE)
3670                 return RT_SCOPE_SITE;
3671         else
3672                 return RT_SCOPE_UNIVERSE;
3673 }
3674
3675 static inline int inet6_ifaddr_msgsize(void)
3676 {
3677         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3678                + nla_total_size(16) /* IFA_ADDRESS */
3679                + nla_total_size(sizeof(struct ifa_cacheinfo));
3680 }
3681
3682 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3683                              u32 portid, u32 seq, int event, unsigned int flags)
3684 {
3685         struct nlmsghdr  *nlh;
3686         u32 preferred, valid;
3687
3688         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3689         if (nlh == NULL)
3690                 return -EMSGSIZE;
3691
3692         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3693                       ifa->idev->dev->ifindex);
3694
3695         if (!(ifa->flags&IFA_F_PERMANENT)) {
3696                 preferred = ifa->prefered_lft;
3697                 valid = ifa->valid_lft;
3698                 if (preferred != INFINITY_LIFE_TIME) {
3699                         long tval = (jiffies - ifa->tstamp)/HZ;
3700                         if (preferred > tval)
3701                                 preferred -= tval;
3702                         else
3703                                 preferred = 0;
3704                         if (valid != INFINITY_LIFE_TIME) {
3705                                 if (valid > tval)
3706                                         valid -= tval;
3707                                 else
3708                                         valid = 0;
3709                         }
3710                 }
3711         } else {
3712                 preferred = INFINITY_LIFE_TIME;
3713                 valid = INFINITY_LIFE_TIME;
3714         }
3715
3716         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3717             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3718                 nlmsg_cancel(skb, nlh);
3719                 return -EMSGSIZE;
3720         }
3721
3722         return nlmsg_end(skb, nlh);
3723 }
3724
3725 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3726                                 u32 portid, u32 seq, int event, u16 flags)
3727 {
3728         struct nlmsghdr  *nlh;
3729         u8 scope = RT_SCOPE_UNIVERSE;
3730         int ifindex = ifmca->idev->dev->ifindex;
3731
3732         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3733                 scope = RT_SCOPE_SITE;
3734
3735         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3736         if (nlh == NULL)
3737                 return -EMSGSIZE;
3738
3739         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3740         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3741             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3742                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3743                 nlmsg_cancel(skb, nlh);
3744                 return -EMSGSIZE;
3745         }
3746
3747         return nlmsg_end(skb, nlh);
3748 }
3749
3750 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3751                                 u32 portid, u32 seq, int event, unsigned int flags)
3752 {
3753         struct nlmsghdr  *nlh;
3754         u8 scope = RT_SCOPE_UNIVERSE;
3755         int ifindex = ifaca->aca_idev->dev->ifindex;
3756
3757         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3758                 scope = RT_SCOPE_SITE;
3759
3760         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3761         if (nlh == NULL)
3762                 return -EMSGSIZE;
3763
3764         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3765         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3766             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3767                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3768                 nlmsg_cancel(skb, nlh);
3769                 return -EMSGSIZE;
3770         }
3771
3772         return nlmsg_end(skb, nlh);
3773 }
3774
3775 enum addr_type_t {
3776         UNICAST_ADDR,
3777         MULTICAST_ADDR,
3778         ANYCAST_ADDR,
3779 };
3780
3781 /* called with rcu_read_lock() */
3782 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3783                           struct netlink_callback *cb, enum addr_type_t type,
3784                           int s_ip_idx, int *p_ip_idx)
3785 {
3786         struct ifmcaddr6 *ifmca;
3787         struct ifacaddr6 *ifaca;
3788         int err = 1;
3789         int ip_idx = *p_ip_idx;
3790
3791         read_lock_bh(&idev->lock);
3792         switch (type) {
3793         case UNICAST_ADDR: {
3794                 struct inet6_ifaddr *ifa;
3795
3796                 /* unicast address incl. temp addr */
3797                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3798                         if (++ip_idx < s_ip_idx)
3799                                 continue;
3800                         err = inet6_fill_ifaddr(skb, ifa,
3801                                                 NETLINK_CB(cb->skb).portid,
3802                                                 cb->nlh->nlmsg_seq,
3803                                                 RTM_NEWADDR,
3804                                                 NLM_F_MULTI);
3805                         if (err <= 0)
3806                                 break;
3807                 }
3808                 break;
3809         }
3810         case MULTICAST_ADDR:
3811                 /* multicast address */
3812                 for (ifmca = idev->mc_list; ifmca;
3813                      ifmca = ifmca->next, ip_idx++) {
3814                         if (ip_idx < s_ip_idx)
3815                                 continue;
3816                         err = inet6_fill_ifmcaddr(skb, ifmca,
3817                                                   NETLINK_CB(cb->skb).portid,
3818                                                   cb->nlh->nlmsg_seq,
3819                                                   RTM_GETMULTICAST,
3820                                                   NLM_F_MULTI);
3821                         if (err <= 0)
3822                                 break;
3823                 }
3824                 break;
3825         case ANYCAST_ADDR:
3826                 /* anycast address */
3827                 for (ifaca = idev->ac_list; ifaca;
3828                      ifaca = ifaca->aca_next, ip_idx++) {
3829                         if (ip_idx < s_ip_idx)
3830                                 continue;
3831                         err = inet6_fill_ifacaddr(skb, ifaca,
3832                                                   NETLINK_CB(cb->skb).portid,
3833                                                   cb->nlh->nlmsg_seq,
3834                                                   RTM_GETANYCAST,
3835                                                   NLM_F_MULTI);
3836                         if (err <= 0)
3837                                 break;
3838                 }
3839                 break;
3840         default:
3841                 break;
3842         }
3843         read_unlock_bh(&idev->lock);
3844         *p_ip_idx = ip_idx;
3845         return err;
3846 }
3847
3848 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3849                            enum addr_type_t type)
3850 {
3851         struct net *net = sock_net(skb->sk);
3852         int h, s_h;
3853         int idx, ip_idx;
3854         int s_idx, s_ip_idx;
3855         struct net_device *dev;
3856         struct inet6_dev *idev;
3857         struct hlist_head *head;
3858         struct hlist_node *node;
3859
3860         s_h = cb->args[0];
3861         s_idx = idx = cb->args[1];
3862         s_ip_idx = ip_idx = cb->args[2];
3863
3864         rcu_read_lock();
3865         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3866                 idx = 0;
3867                 head = &net->dev_index_head[h];
3868                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3869                         if (idx < s_idx)
3870                                 goto cont;
3871                         if (h > s_h || idx > s_idx)
3872                                 s_ip_idx = 0;
3873                         ip_idx = 0;
3874                         idev = __in6_dev_get(dev);
3875                         if (!idev)
3876                                 goto cont;
3877
3878                         if (in6_dump_addrs(idev, skb, cb, type,
3879                                            s_ip_idx, &ip_idx) <= 0)
3880                                 goto done;
3881 cont:
3882                         idx++;
3883                 }
3884         }
3885 done:
3886         rcu_read_unlock();
3887         cb->args[0] = h;
3888         cb->args[1] = idx;
3889         cb->args[2] = ip_idx;
3890
3891         return skb->len;
3892 }
3893
3894 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3895 {
3896         enum addr_type_t type = UNICAST_ADDR;
3897
3898         return inet6_dump_addr(skb, cb, type);
3899 }
3900
3901 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3902 {
3903         enum addr_type_t type = MULTICAST_ADDR;
3904
3905         return inet6_dump_addr(skb, cb, type);
3906 }
3907
3908
3909 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3910 {
3911         enum addr_type_t type = ANYCAST_ADDR;
3912
3913         return inet6_dump_addr(skb, cb, type);
3914 }
3915
3916 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3917                              void *arg)
3918 {
3919         struct net *net = sock_net(in_skb->sk);
3920         struct ifaddrmsg *ifm;
3921         struct nlattr *tb[IFA_MAX+1];
3922         struct in6_addr *addr = NULL;
3923         struct net_device *dev = NULL;
3924         struct inet6_ifaddr *ifa;
3925         struct sk_buff *skb;
3926         int err;
3927
3928         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3929         if (err < 0)
3930                 goto errout;
3931
3932         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3933         if (addr == NULL) {
3934                 err = -EINVAL;
3935                 goto errout;
3936         }
3937
3938         ifm = nlmsg_data(nlh);
3939         if (ifm->ifa_index)
3940                 dev = __dev_get_by_index(net, ifm->ifa_index);
3941
3942         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3943         if (!ifa) {
3944                 err = -EADDRNOTAVAIL;
3945                 goto errout;
3946         }
3947
3948         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3949         if (!skb) {
3950                 err = -ENOBUFS;
3951                 goto errout_ifa;
3952         }
3953
3954         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3955                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3956         if (err < 0) {
3957                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3958                 WARN_ON(err == -EMSGSIZE);
3959                 kfree_skb(skb);
3960                 goto errout_ifa;
3961         }
3962         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3963 errout_ifa:
3964         in6_ifa_put(ifa);
3965 errout:
3966         return err;
3967 }
3968
3969 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3970 {
3971         struct sk_buff *skb;
3972         struct net *net = dev_net(ifa->idev->dev);
3973         int err = -ENOBUFS;
3974
3975         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3976         if (skb == NULL)
3977                 goto errout;
3978
3979         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3980         if (err < 0) {
3981                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3982                 WARN_ON(err == -EMSGSIZE);
3983                 kfree_skb(skb);
3984                 goto errout;
3985         }
3986         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3987         return;
3988 errout:
3989         if (err < 0)
3990                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3991 }
3992
3993 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3994                                 __s32 *array, int bytes)
3995 {
3996         BUG_ON(bytes < (DEVCONF_MAX * 4));
3997
3998         memset(array, 0, bytes);
3999         array[DEVCONF_FORWARDING] = cnf->forwarding;
4000         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4001         array[DEVCONF_MTU6] = cnf->mtu6;
4002         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4003         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4004         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4005         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4006         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4007         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4008                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4009         array[DEVCONF_RTR_SOLICIT_DELAY] =
4010                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4011         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4012 #ifdef CONFIG_IPV6_PRIVACY
4013         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4014         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4015         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4016         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4017         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4018 #endif
4019         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4020         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4021         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4022 #ifdef CONFIG_IPV6_ROUTER_PREF
4023         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4024         array[DEVCONF_RTR_PROBE_INTERVAL] =
4025                 jiffies_to_msecs(cnf->rtr_probe_interval);
4026 #ifdef CONFIG_IPV6_ROUTE_INFO
4027         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4028 #endif
4029 #endif
4030         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4031         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4032 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4033         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4034 #endif
4035 #ifdef CONFIG_IPV6_MROUTE
4036         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4037 #endif
4038         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4039         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4040         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4041         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4042 }
4043
4044 static inline size_t inet6_ifla6_size(void)
4045 {
4046         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4047              + nla_total_size(sizeof(struct ifla_cacheinfo))
4048              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4049              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4050              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4051 }
4052
4053 static inline size_t inet6_if_nlmsg_size(void)
4054 {
4055         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4056                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4057                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4058                + nla_total_size(4) /* IFLA_MTU */
4059                + nla_total_size(4) /* IFLA_LINK */
4060                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4061 }
4062
4063 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4064                                       int items, int bytes)
4065 {
4066         int i;
4067         int pad = bytes - sizeof(u64) * items;
4068         BUG_ON(pad < 0);
4069
4070         /* Use put_unaligned() because stats may not be aligned for u64. */
4071         put_unaligned(items, &stats[0]);
4072         for (i = 1; i < items; i++)
4073                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4074
4075         memset(&stats[items], 0, pad);
4076 }
4077
4078 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4079                                       int items, int bytes, size_t syncpoff)
4080 {
4081         int i;
4082         int pad = bytes - sizeof(u64) * items;
4083         BUG_ON(pad < 0);
4084
4085         /* Use put_unaligned() because stats may not be aligned for u64. */
4086         put_unaligned(items, &stats[0]);
4087         for (i = 1; i < items; i++)
4088                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4089
4090         memset(&stats[items], 0, pad);
4091 }
4092
4093 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4094                              int bytes)
4095 {
4096         switch (attrtype) {
4097         case IFLA_INET6_STATS:
4098                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4099                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4100                 break;
4101         case IFLA_INET6_ICMP6STATS:
4102                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4103                 break;
4104         }
4105 }
4106
4107 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4108 {
4109         struct nlattr *nla;
4110         struct ifla_cacheinfo ci;
4111
4112         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4113                 goto nla_put_failure;
4114         ci.max_reasm_len = IPV6_MAXPLEN;
4115         ci.tstamp = cstamp_delta(idev->tstamp);
4116         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4117         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4118         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4119                 goto nla_put_failure;
4120         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4121         if (nla == NULL)
4122                 goto nla_put_failure;
4123         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4124
4125         /* XXX - MC not implemented */
4126
4127         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4128         if (nla == NULL)
4129                 goto nla_put_failure;
4130         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4131
4132         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4133         if (nla == NULL)
4134                 goto nla_put_failure;
4135         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4136
4137         return 0;
4138
4139 nla_put_failure:
4140         return -EMSGSIZE;
4141 }
4142
4143 static size_t inet6_get_link_af_size(const struct net_device *dev)
4144 {
4145         if (!__in6_dev_get(dev))
4146                 return 0;
4147
4148         return inet6_ifla6_size();
4149 }
4150
4151 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4152 {
4153         struct inet6_dev *idev = __in6_dev_get(dev);
4154
4155         if (!idev)
4156                 return -ENODATA;
4157
4158         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4159                 return -EMSGSIZE;
4160
4161         return 0;
4162 }
4163
4164 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4165                              u32 portid, u32 seq, int event, unsigned int flags)
4166 {
4167         struct net_device *dev = idev->dev;
4168         struct ifinfomsg *hdr;
4169         struct nlmsghdr *nlh;
4170         void *protoinfo;
4171
4172         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4173         if (nlh == NULL)
4174                 return -EMSGSIZE;
4175
4176         hdr = nlmsg_data(nlh);
4177         hdr->ifi_family = AF_INET6;
4178         hdr->__ifi_pad = 0;
4179         hdr->ifi_type = dev->type;
4180         hdr->ifi_index = dev->ifindex;
4181         hdr->ifi_flags = dev_get_flags(dev);
4182         hdr->ifi_change = 0;
4183
4184         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4185             (dev->addr_len &&
4186              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4187             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4188             (dev->ifindex != dev->iflink &&
4189              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4190                 goto nla_put_failure;
4191         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4192         if (protoinfo == NULL)
4193                 goto nla_put_failure;
4194
4195         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4196                 goto nla_put_failure;
4197
4198         nla_nest_end(skb, protoinfo);
4199         return nlmsg_end(skb, nlh);
4200
4201 nla_put_failure:
4202         nlmsg_cancel(skb, nlh);
4203         return -EMSGSIZE;
4204 }
4205
4206 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4207 {
4208         struct net *net = sock_net(skb->sk);
4209         int h, s_h;
4210         int idx = 0, s_idx;
4211         struct net_device *dev;
4212         struct inet6_dev *idev;
4213         struct hlist_head *head;
4214         struct hlist_node *node;
4215
4216         s_h = cb->args[0];
4217         s_idx = cb->args[1];
4218
4219         rcu_read_lock();
4220         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4221                 idx = 0;
4222                 head = &net->dev_index_head[h];
4223                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4224                         if (idx < s_idx)
4225                                 goto cont;
4226                         idev = __in6_dev_get(dev);
4227                         if (!idev)
4228                                 goto cont;
4229                         if (inet6_fill_ifinfo(skb, idev,
4230                                               NETLINK_CB(cb->skb).portid,
4231                                               cb->nlh->nlmsg_seq,
4232                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4233                                 goto out;
4234 cont:
4235                         idx++;
4236                 }
4237         }
4238 out:
4239         rcu_read_unlock();
4240         cb->args[1] = idx;
4241         cb->args[0] = h;
4242
4243         return skb->len;
4244 }
4245
4246 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4247 {
4248         struct sk_buff *skb;
4249         struct net *net = dev_net(idev->dev);
4250         int err = -ENOBUFS;
4251
4252         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4253         if (skb == NULL)
4254                 goto errout;
4255
4256         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4257         if (err < 0) {
4258                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4259                 WARN_ON(err == -EMSGSIZE);
4260                 kfree_skb(skb);
4261                 goto errout;
4262         }
4263         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4264         return;
4265 errout:
4266         if (err < 0)
4267                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4268 }
4269
4270 static inline size_t inet6_prefix_nlmsg_size(void)
4271 {
4272         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4273                + nla_total_size(sizeof(struct in6_addr))
4274                + nla_total_size(sizeof(struct prefix_cacheinfo));
4275 }
4276
4277 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4278                              struct prefix_info *pinfo, u32 portid, u32 seq,
4279                              int event, unsigned int flags)
4280 {
4281         struct prefixmsg *pmsg;
4282         struct nlmsghdr *nlh;
4283         struct prefix_cacheinfo ci;
4284
4285         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4286         if (nlh == NULL)
4287                 return -EMSGSIZE;
4288
4289         pmsg = nlmsg_data(nlh);
4290         pmsg->prefix_family = AF_INET6;
4291         pmsg->prefix_pad1 = 0;
4292         pmsg->prefix_pad2 = 0;
4293         pmsg->prefix_ifindex = idev->dev->ifindex;
4294         pmsg->prefix_len = pinfo->prefix_len;
4295         pmsg->prefix_type = pinfo->type;
4296         pmsg->prefix_pad3 = 0;
4297         pmsg->prefix_flags = 0;
4298         if (pinfo->onlink)
4299                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4300         if (pinfo->autoconf)
4301                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4302
4303         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4304                 goto nla_put_failure;
4305         ci.preferred_time = ntohl(pinfo->prefered);
4306         ci.valid_time = ntohl(pinfo->valid);
4307         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4308                 goto nla_put_failure;
4309         return nlmsg_end(skb, nlh);
4310
4311 nla_put_failure:
4312         nlmsg_cancel(skb, nlh);
4313         return -EMSGSIZE;
4314 }
4315
4316 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4317                          struct prefix_info *pinfo)
4318 {
4319         struct sk_buff *skb;
4320         struct net *net = dev_net(idev->dev);
4321         int err = -ENOBUFS;
4322
4323         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4324         if (skb == NULL)
4325                 goto errout;
4326
4327         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4328         if (err < 0) {
4329                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4330                 WARN_ON(err == -EMSGSIZE);
4331                 kfree_skb(skb);
4332                 goto errout;
4333         }
4334         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4335         return;
4336 errout:
4337         if (err < 0)
4338                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4339 }
4340
4341 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4342 {
4343         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4344
4345         switch (event) {
4346         case RTM_NEWADDR:
4347                 /*
4348                  * If the address was optimistic
4349                  * we inserted the route at the start of
4350                  * our DAD process, so we don't need
4351                  * to do it again
4352                  */
4353                 if (!(ifp->rt->rt6i_node))
4354                         ip6_ins_rt(ifp->rt);
4355                 if (ifp->idev->cnf.forwarding)
4356                         addrconf_join_anycast(ifp);
4357                 break;
4358         case RTM_DELADDR:
4359                 if (ifp->idev->cnf.forwarding)
4360                         addrconf_leave_anycast(ifp);
4361                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4362                 dst_hold(&ifp->rt->dst);
4363
4364                 if (ip6_del_rt(ifp->rt))
4365                         dst_free(&ifp->rt->dst);
4366                 break;
4367         }
4368 }
4369
4370 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4371 {
4372         rcu_read_lock_bh();
4373         if (likely(ifp->idev->dead == 0))
4374                 __ipv6_ifa_notify(event, ifp);
4375         rcu_read_unlock_bh();
4376 }
4377
4378 #ifdef CONFIG_SYSCTL
4379
4380 static
4381 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4382                            void __user *buffer, size_t *lenp, loff_t *ppos)
4383 {
4384         int *valp = ctl->data;
4385         int val = *valp;
4386         loff_t pos = *ppos;
4387         ctl_table lctl;
4388         int ret;
4389
4390         /*
4391          * ctl->data points to idev->cnf.forwarding, we should
4392          * not modify it until we get the rtnl lock.
4393          */
4394         lctl = *ctl;
4395         lctl.data = &val;
4396
4397         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4398
4399         if (write)
4400                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4401         if (ret)
4402                 *ppos = pos;
4403         return ret;
4404 }
4405
4406 static void dev_disable_change(struct inet6_dev *idev)
4407 {
4408         if (!idev || !idev->dev)
4409                 return;
4410
4411         if (idev->cnf.disable_ipv6)
4412                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4413         else
4414                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4415 }
4416
4417 static void addrconf_disable_change(struct net *net, __s32 newf)
4418 {
4419         struct net_device *dev;
4420         struct inet6_dev *idev;
4421
4422         rcu_read_lock();
4423         for_each_netdev_rcu(net, dev) {
4424                 idev = __in6_dev_get(dev);
4425                 if (idev) {
4426                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4427                         idev->cnf.disable_ipv6 = newf;
4428                         if (changed)
4429                                 dev_disable_change(idev);
4430                 }
4431         }
4432         rcu_read_unlock();
4433 }
4434
4435 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4436 {
4437         struct net *net;
4438         int old;
4439
4440         if (!rtnl_trylock())
4441                 return restart_syscall();
4442
4443         net = (struct net *)table->extra2;
4444         old = *p;
4445         *p = newf;
4446
4447         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4448                 rtnl_unlock();
4449                 return 0;
4450         }
4451
4452         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4453                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4454                 addrconf_disable_change(net, newf);
4455         } else if ((!newf) ^ (!old))
4456                 dev_disable_change((struct inet6_dev *)table->extra1);
4457
4458         rtnl_unlock();
4459         return 0;
4460 }
4461
4462 static
4463 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4464                             void __user *buffer, size_t *lenp, loff_t *ppos)
4465 {
4466         int *valp = ctl->data;
4467         int val = *valp;
4468         loff_t pos = *ppos;
4469         ctl_table lctl;
4470         int ret;
4471
4472         /*
4473          * ctl->data points to idev->cnf.disable_ipv6, we should
4474          * not modify it until we get the rtnl lock.
4475          */
4476         lctl = *ctl;
4477         lctl.data = &val;
4478
4479         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4480
4481         if (write)
4482                 ret = addrconf_disable_ipv6(ctl, valp, val);
4483         if (ret)
4484                 *ppos = pos;
4485         return ret;
4486 }
4487
4488 static struct addrconf_sysctl_table
4489 {
4490         struct ctl_table_header *sysctl_header;
4491         ctl_table addrconf_vars[DEVCONF_MAX+1];
4492 } addrconf_sysctl __read_mostly = {
4493         .sysctl_header = NULL,
4494         .addrconf_vars = {
4495                 {
4496                         .procname       = "forwarding",
4497                         .data           = &ipv6_devconf.forwarding,
4498                         .maxlen         = sizeof(int),
4499                         .mode           = 0644,
4500                         .proc_handler   = addrconf_sysctl_forward,
4501                 },
4502                 {
4503                         .procname       = "hop_limit",
4504                         .data           = &ipv6_devconf.hop_limit,
4505                         .maxlen         = sizeof(int),
4506                         .mode           = 0644,
4507                         .proc_handler   = proc_dointvec,
4508                 },
4509                 {
4510                         .procname       = "mtu",
4511                         .data           = &ipv6_devconf.mtu6,
4512                         .maxlen         = sizeof(int),
4513                         .mode           = 0644,
4514                         .proc_handler   = proc_dointvec,
4515                 },
4516                 {
4517                         .procname       = "accept_ra",
4518                         .data           = &ipv6_devconf.accept_ra,
4519                         .maxlen         = sizeof(int),
4520                         .mode           = 0644,
4521                         .proc_handler   = proc_dointvec,
4522                 },
4523                 {
4524                         .procname       = "accept_redirects",
4525                         .data           = &ipv6_devconf.accept_redirects,
4526                         .maxlen         = sizeof(int),
4527                         .mode           = 0644,
4528                         .proc_handler   = proc_dointvec,
4529                 },
4530                 {
4531                         .procname       = "autoconf",
4532                         .data           = &ipv6_devconf.autoconf,
4533                         .maxlen         = sizeof(int),
4534                         .mode           = 0644,
4535                         .proc_handler   = proc_dointvec,
4536                 },
4537                 {
4538                         .procname       = "dad_transmits",
4539                         .data           = &ipv6_devconf.dad_transmits,
4540                         .maxlen         = sizeof(int),
4541                         .mode           = 0644,
4542                         .proc_handler   = proc_dointvec,
4543                 },
4544                 {
4545                         .procname       = "router_solicitations",
4546                         .data           = &ipv6_devconf.rtr_solicits,
4547                         .maxlen         = sizeof(int),
4548                         .mode           = 0644,
4549                         .proc_handler   = proc_dointvec,
4550                 },
4551                 {
4552                         .procname       = "router_solicitation_interval",
4553                         .data           = &ipv6_devconf.rtr_solicit_interval,
4554                         .maxlen         = sizeof(int),
4555                         .mode           = 0644,
4556                         .proc_handler   = proc_dointvec_jiffies,
4557                 },
4558                 {
4559                         .procname       = "router_solicitation_delay",
4560                         .data           = &ipv6_devconf.rtr_solicit_delay,
4561                         .maxlen         = sizeof(int),
4562                         .mode           = 0644,
4563                         .proc_handler   = proc_dointvec_jiffies,
4564                 },
4565                 {
4566                         .procname       = "force_mld_version",
4567                         .data           = &ipv6_devconf.force_mld_version,
4568                         .maxlen         = sizeof(int),
4569                         .mode           = 0644,
4570                         .proc_handler   = proc_dointvec,
4571                 },
4572 #ifdef CONFIG_IPV6_PRIVACY
4573                 {
4574                         .procname       = "use_tempaddr",
4575                         .data           = &ipv6_devconf.use_tempaddr,
4576                         .maxlen         = sizeof(int),
4577                         .mode           = 0644,
4578                         .proc_handler   = proc_dointvec,
4579                 },
4580                 {
4581                         .procname       = "temp_valid_lft",
4582                         .data           = &ipv6_devconf.temp_valid_lft,
4583                         .maxlen         = sizeof(int),
4584                         .mode           = 0644,
4585                         .proc_handler   = proc_dointvec,
4586                 },
4587                 {
4588                         .procname       = "temp_prefered_lft",
4589                         .data           = &ipv6_devconf.temp_prefered_lft,
4590                         .maxlen         = sizeof(int),
4591                         .mode           = 0644,
4592                         .proc_handler   = proc_dointvec,
4593                 },
4594                 {
4595                         .procname       = "regen_max_retry",
4596                         .data           = &ipv6_devconf.regen_max_retry,
4597                         .maxlen         = sizeof(int),
4598                         .mode           = 0644,
4599                         .proc_handler   = proc_dointvec,
4600                 },
4601                 {
4602                         .procname       = "max_desync_factor",
4603                         .data           = &ipv6_devconf.max_desync_factor,
4604                         .maxlen         = sizeof(int),
4605                         .mode           = 0644,
4606                         .proc_handler   = proc_dointvec,
4607                 },
4608 #endif
4609                 {
4610                         .procname       = "max_addresses",
4611                         .data           = &ipv6_devconf.max_addresses,
4612                         .maxlen         = sizeof(int),
4613                         .mode           = 0644,
4614                         .proc_handler   = proc_dointvec,
4615                 },
4616                 {
4617                         .procname       = "accept_ra_defrtr",
4618                         .data           = &ipv6_devconf.accept_ra_defrtr,
4619                         .maxlen         = sizeof(int),
4620                         .mode           = 0644,
4621                         .proc_handler   = proc_dointvec,
4622                 },
4623                 {
4624                         .procname       = "accept_ra_pinfo",
4625                         .data           = &ipv6_devconf.accept_ra_pinfo,
4626                         .maxlen         = sizeof(int),
4627                         .mode           = 0644,
4628                         .proc_handler   = proc_dointvec,
4629                 },
4630 #ifdef CONFIG_IPV6_ROUTER_PREF
4631                 {
4632                         .procname       = "accept_ra_rtr_pref",
4633                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4634                         .maxlen         = sizeof(int),
4635                         .mode           = 0644,
4636                         .proc_handler   = proc_dointvec,
4637                 },
4638                 {
4639                         .procname       = "router_probe_interval",
4640                         .data           = &ipv6_devconf.rtr_probe_interval,
4641                         .maxlen         = sizeof(int),
4642                         .mode           = 0644,
4643                         .proc_handler   = proc_dointvec_jiffies,
4644                 },
4645 #ifdef CONFIG_IPV6_ROUTE_INFO
4646                 {
4647                         .procname       = "accept_ra_rt_info_max_plen",
4648                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4649                         .maxlen         = sizeof(int),
4650                         .mode           = 0644,
4651                         .proc_handler   = proc_dointvec,
4652                 },
4653 #endif
4654 #endif
4655                 {
4656                         .procname       = "proxy_ndp",
4657                         .data           = &ipv6_devconf.proxy_ndp,
4658                         .maxlen         = sizeof(int),
4659                         .mode           = 0644,
4660                         .proc_handler   = proc_dointvec,
4661                 },
4662                 {
4663                         .procname       = "accept_source_route",
4664                         .data           = &ipv6_devconf.accept_source_route,
4665                         .maxlen         = sizeof(int),
4666                         .mode           = 0644,
4667                         .proc_handler   = proc_dointvec,
4668                 },
4669 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4670                 {
4671                         .procname       = "optimistic_dad",
4672                         .data           = &ipv6_devconf.optimistic_dad,
4673                         .maxlen         = sizeof(int),
4674                         .mode           = 0644,
4675                         .proc_handler   = proc_dointvec,
4676
4677                 },
4678 #endif
4679 #ifdef CONFIG_IPV6_MROUTE
4680                 {
4681                         .procname       = "mc_forwarding",
4682                         .data           = &ipv6_devconf.mc_forwarding,
4683                         .maxlen         = sizeof(int),
4684                         .mode           = 0444,
4685                         .proc_handler   = proc_dointvec,
4686                 },
4687 #endif
4688                 {
4689                         .procname       = "disable_ipv6",
4690                         .data           = &ipv6_devconf.disable_ipv6,
4691                         .maxlen         = sizeof(int),
4692                         .mode           = 0644,
4693                         .proc_handler   = addrconf_sysctl_disable,
4694                 },
4695                 {
4696                         .procname       = "accept_dad",
4697                         .data           = &ipv6_devconf.accept_dad,
4698                         .maxlen         = sizeof(int),
4699                         .mode           = 0644,
4700                         .proc_handler   = proc_dointvec,
4701                 },
4702                 {
4703                         .procname       = "force_tllao",
4704                         .data           = &ipv6_devconf.force_tllao,
4705                         .maxlen         = sizeof(int),
4706                         .mode           = 0644,
4707                         .proc_handler   = proc_dointvec
4708                 },
4709                 {
4710                         .procname       = "ndisc_notify",
4711                         .data           = &ipv6_devconf.ndisc_notify,
4712                         .maxlen         = sizeof(int),
4713                         .mode           = 0644,
4714                         .proc_handler   = proc_dointvec
4715                 },
4716                 {
4717                         /* sentinel */
4718                 }
4719         },
4720 };
4721
4722 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4723                 struct inet6_dev *idev, struct ipv6_devconf *p)
4724 {
4725         int i;
4726         struct addrconf_sysctl_table *t;
4727         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4728
4729         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4730         if (t == NULL)
4731                 goto out;
4732
4733         for (i = 0; t->addrconf_vars[i].data; i++) {
4734                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4735                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4736                 t->addrconf_vars[i].extra2 = net;
4737         }
4738
4739         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4740
4741         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4742         if (t->sysctl_header == NULL)
4743                 goto free;
4744
4745         p->sysctl = t;
4746         return 0;
4747
4748 free:
4749         kfree(t);
4750 out:
4751         return -ENOBUFS;
4752 }
4753
4754 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4755 {
4756         struct addrconf_sysctl_table *t;
4757
4758         if (p->sysctl == NULL)
4759                 return;
4760
4761         t = p->sysctl;
4762         p->sysctl = NULL;
4763         unregister_net_sysctl_table(t->sysctl_header);
4764         kfree(t);
4765 }
4766
4767 static void addrconf_sysctl_register(struct inet6_dev *idev)
4768 {
4769         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4770                               &ndisc_ifinfo_sysctl_change);
4771         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4772                                         idev, &idev->cnf);
4773 }
4774
4775 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4776 {
4777         __addrconf_sysctl_unregister(&idev->cnf);
4778         neigh_sysctl_unregister(idev->nd_parms);
4779 }
4780
4781
4782 #endif
4783
4784 static int __net_init addrconf_init_net(struct net *net)
4785 {
4786         int err;
4787         struct ipv6_devconf *all, *dflt;
4788
4789         err = -ENOMEM;
4790         all = &ipv6_devconf;
4791         dflt = &ipv6_devconf_dflt;
4792
4793         if (!net_eq(net, &init_net)) {
4794                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4795                 if (all == NULL)
4796                         goto err_alloc_all;
4797
4798                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4799                 if (dflt == NULL)
4800                         goto err_alloc_dflt;
4801         } else {
4802                 /* these will be inherited by all namespaces */
4803                 dflt->autoconf = ipv6_defaults.autoconf;
4804                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4805         }
4806
4807         net->ipv6.devconf_all = all;
4808         net->ipv6.devconf_dflt = dflt;
4809
4810 #ifdef CONFIG_SYSCTL
4811         err = __addrconf_sysctl_register(net, "all", NULL, all);
4812         if (err < 0)
4813                 goto err_reg_all;
4814
4815         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4816         if (err < 0)
4817                 goto err_reg_dflt;
4818 #endif
4819         return 0;
4820
4821 #ifdef CONFIG_SYSCTL
4822 err_reg_dflt:
4823         __addrconf_sysctl_unregister(all);
4824 err_reg_all:
4825         kfree(dflt);
4826 #endif
4827 err_alloc_dflt:
4828         kfree(all);
4829 err_alloc_all:
4830         return err;
4831 }
4832
4833 static void __net_exit addrconf_exit_net(struct net *net)
4834 {
4835 #ifdef CONFIG_SYSCTL
4836         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4837         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4838 #endif
4839         if (!net_eq(net, &init_net)) {
4840                 kfree(net->ipv6.devconf_dflt);
4841                 kfree(net->ipv6.devconf_all);
4842         }
4843 }
4844
4845 static struct pernet_operations addrconf_ops = {
4846         .init = addrconf_init_net,
4847         .exit = addrconf_exit_net,
4848 };
4849
4850 /*
4851  *      Device notifier
4852  */
4853
4854 int register_inet6addr_notifier(struct notifier_block *nb)
4855 {
4856         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4857 }
4858 EXPORT_SYMBOL(register_inet6addr_notifier);
4859
4860 int unregister_inet6addr_notifier(struct notifier_block *nb)
4861 {
4862         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4863 }
4864 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4865
4866 static struct rtnl_af_ops inet6_ops = {
4867         .family           = AF_INET6,
4868         .fill_link_af     = inet6_fill_link_af,
4869         .get_link_af_size = inet6_get_link_af_size,
4870 };
4871
4872 /*
4873  *      Init / cleanup code
4874  */
4875
4876 int __init addrconf_init(void)
4877 {
4878         int i, err;
4879
4880         err = ipv6_addr_label_init();
4881         if (err < 0) {
4882                 pr_crit("%s: cannot initialize default policy table: %d\n",
4883                         __func__, err);
4884                 goto out;
4885         }
4886
4887         err = register_pernet_subsys(&addrconf_ops);
4888         if (err < 0)
4889                 goto out_addrlabel;
4890
4891         /* The addrconf netdev notifier requires that loopback_dev
4892          * has it's ipv6 private information allocated and setup
4893          * before it can bring up and give link-local addresses
4894          * to other devices which are up.
4895          *
4896          * Unfortunately, loopback_dev is not necessarily the first
4897          * entry in the global dev_base list of net devices.  In fact,
4898          * it is likely to be the very last entry on that list.
4899          * So this causes the notifier registry below to try and
4900          * give link-local addresses to all devices besides loopback_dev
4901          * first, then loopback_dev, which cases all the non-loopback_dev
4902          * devices to fail to get a link-local address.
4903          *
4904          * So, as a temporary fix, allocate the ipv6 structure for
4905          * loopback_dev first by hand.
4906          * Longer term, all of the dependencies ipv6 has upon the loopback
4907          * device and it being up should be removed.
4908          */
4909         rtnl_lock();
4910         if (!ipv6_add_dev(init_net.loopback_dev))
4911                 err = -ENOMEM;
4912         rtnl_unlock();
4913         if (err)
4914                 goto errlo;
4915
4916         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4917                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4918
4919         register_netdevice_notifier(&ipv6_dev_notf);
4920
4921         addrconf_verify(0);
4922
4923         err = rtnl_af_register(&inet6_ops);
4924         if (err < 0)
4925                 goto errout_af;
4926
4927         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4928                               NULL);
4929         if (err < 0)
4930                 goto errout;
4931
4932         /* Only the first call to __rtnl_register can fail */
4933         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4934         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4935         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4936                         inet6_dump_ifaddr, NULL);
4937         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4938                         inet6_dump_ifmcaddr, NULL);
4939         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4940                         inet6_dump_ifacaddr, NULL);
4941         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4942                         NULL, NULL);
4943
4944         ipv6_addr_label_rtnl_register();
4945
4946         return 0;
4947 errout:
4948         rtnl_af_unregister(&inet6_ops);
4949 errout_af:
4950         unregister_netdevice_notifier(&ipv6_dev_notf);
4951 errlo:
4952         unregister_pernet_subsys(&addrconf_ops);
4953 out_addrlabel:
4954         ipv6_addr_label_cleanup();
4955 out:
4956         return err;
4957 }
4958
4959 void addrconf_cleanup(void)
4960 {
4961         struct net_device *dev;
4962         int i;
4963
4964         unregister_netdevice_notifier(&ipv6_dev_notf);
4965         unregister_pernet_subsys(&addrconf_ops);
4966         ipv6_addr_label_cleanup();
4967
4968         rtnl_lock();
4969
4970         __rtnl_af_unregister(&inet6_ops);
4971
4972         /* clean dev list */
4973         for_each_netdev(&init_net, dev) {
4974                 if (__in6_dev_get(dev) == NULL)
4975                         continue;
4976                 addrconf_ifdown(dev, 1);
4977         }
4978         addrconf_ifdown(init_net.loopback_dev, 2);
4979
4980         /*
4981          *      Check hash table.
4982          */
4983         spin_lock_bh(&addrconf_hash_lock);
4984         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4985                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4986         spin_unlock_bh(&addrconf_hash_lock);
4987
4988         del_timer(&addr_chk_timer);
4989         rtnl_unlock();
4990 }