netlink: make nlmsg_end() and genlmsg_end() void
[firefly-linux-kernel-4.4.55.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/firewire.h>
74 #include <net/ipv6.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
79 #include <net/tcp.h>
80 #include <net/ip.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
89
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
93
94 /* Set to 3 to get tracing... */
95 #define ACONF_DEBUG 2
96
97 #if ACONF_DEBUG >= 3
98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
99 #else
100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
101 #endif
102
103 #define INFINITY_LIFE_TIME      0xFFFFFFFF
104
105 static inline u32 cstamp_delta(unsigned long cstamp)
106 {
107         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
108 }
109
110 #ifdef CONFIG_SYSCTL
111 static int addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
113 #else
114 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
115 {
116         return 0;
117 }
118
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
120 {
121 }
122 #endif
123
124 static void __ipv6_regen_rndid(struct inet6_dev *idev);
125 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
126 static void ipv6_regen_rndid(unsigned long data);
127
128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
129 static int ipv6_count_addresses(struct inet6_dev *idev);
130
131 /*
132  *      Configured unicast address hash table
133  */
134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
135 static DEFINE_SPINLOCK(addrconf_hash_lock);
136
137 static void addrconf_verify(void);
138 static void addrconf_verify_rtnl(void);
139 static void addrconf_verify_work(struct work_struct *);
140
141 static struct workqueue_struct *addrconf_wq;
142 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
143
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146
147 static void addrconf_type_change(struct net_device *dev,
148                                  unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
150
151 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
152                                                   int plen,
153                                                   const struct net_device *dev,
154                                                   u32 flags, u32 noflags);
155
156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
157 static void addrconf_dad_work(struct work_struct *w);
158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_run(struct inet6_dev *idev);
160 static void addrconf_rs_timer(unsigned long data);
161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163
164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
165                                 struct prefix_info *pinfo);
166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
167                                struct net_device *dev);
168
169 static struct ipv6_devconf ipv6_devconf __read_mostly = {
170         .forwarding             = 0,
171         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
172         .mtu6                   = IPV6_MIN_MTU,
173         .accept_ra              = 1,
174         .accept_redirects       = 1,
175         .autoconf               = 1,
176         .force_mld_version      = 0,
177         .mldv1_unsolicited_report_interval = 10 * HZ,
178         .mldv2_unsolicited_report_interval = HZ,
179         .dad_transmits          = 1,
180         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
181         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
182         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
183         .use_tempaddr           = 0,
184         .temp_valid_lft         = TEMP_VALID_LIFETIME,
185         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
186         .regen_max_retry        = REGEN_MAX_RETRY,
187         .max_desync_factor      = MAX_DESYNC_FACTOR,
188         .max_addresses          = IPV6_MAX_ADDRESSES,
189         .accept_ra_defrtr       = 1,
190         .accept_ra_from_local   = 0,
191         .accept_ra_pinfo        = 1,
192 #ifdef CONFIG_IPV6_ROUTER_PREF
193         .accept_ra_rtr_pref     = 1,
194         .rtr_probe_interval     = 60 * HZ,
195 #ifdef CONFIG_IPV6_ROUTE_INFO
196         .accept_ra_rt_info_max_plen = 0,
197 #endif
198 #endif
199         .proxy_ndp              = 0,
200         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
201         .disable_ipv6           = 0,
202         .accept_dad             = 1,
203         .suppress_frag_ndisc    = 1,
204 };
205
206 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
207         .forwarding             = 0,
208         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
209         .mtu6                   = IPV6_MIN_MTU,
210         .accept_ra              = 1,
211         .accept_redirects       = 1,
212         .autoconf               = 1,
213         .force_mld_version      = 0,
214         .mldv1_unsolicited_report_interval = 10 * HZ,
215         .mldv2_unsolicited_report_interval = HZ,
216         .dad_transmits          = 1,
217         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
218         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
219         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
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         .max_addresses          = IPV6_MAX_ADDRESSES,
226         .accept_ra_defrtr       = 1,
227         .accept_ra_from_local   = 0,
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         .suppress_frag_ndisc    = 1,
241 };
242
243 /* Check if a valid qdisc is available */
244 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
245 {
246         return !qdisc_tx_is_noop(dev);
247 }
248
249 static void addrconf_del_rs_timer(struct inet6_dev *idev)
250 {
251         if (del_timer(&idev->rs_timer))
252                 __in6_dev_put(idev);
253 }
254
255 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
256 {
257         if (cancel_delayed_work(&ifp->dad_work))
258                 __in6_ifa_put(ifp);
259 }
260
261 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
262                                   unsigned long when)
263 {
264         if (!timer_pending(&idev->rs_timer))
265                 in6_dev_hold(idev);
266         mod_timer(&idev->rs_timer, jiffies + when);
267 }
268
269 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
270                                    unsigned long delay)
271 {
272         if (!delayed_work_pending(&ifp->dad_work))
273                 in6_ifa_hold(ifp);
274         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
275 }
276
277 static int snmp6_alloc_dev(struct inet6_dev *idev)
278 {
279         int i;
280
281         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
282         if (!idev->stats.ipv6)
283                 goto err_ip;
284
285         for_each_possible_cpu(i) {
286                 struct ipstats_mib *addrconf_stats;
287                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
288                 u64_stats_init(&addrconf_stats->syncp);
289         }
290
291
292         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
293                                         GFP_KERNEL);
294         if (!idev->stats.icmpv6dev)
295                 goto err_icmp;
296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
297                                            GFP_KERNEL);
298         if (!idev->stats.icmpv6msgdev)
299                 goto err_icmpmsg;
300
301         return 0;
302
303 err_icmpmsg:
304         kfree(idev->stats.icmpv6dev);
305 err_icmp:
306         free_percpu(idev->stats.ipv6);
307 err_ip:
308         return -ENOMEM;
309 }
310
311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
312 {
313         struct inet6_dev *ndev;
314         int err = -ENOMEM;
315
316         ASSERT_RTNL();
317
318         if (dev->mtu < IPV6_MIN_MTU)
319                 return ERR_PTR(-EINVAL);
320
321         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
322         if (ndev == NULL)
323                 return ERR_PTR(err);
324
325         rwlock_init(&ndev->lock);
326         ndev->dev = dev;
327         INIT_LIST_HEAD(&ndev->addr_list);
328         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
329                     (unsigned long)ndev);
330         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
331         ndev->cnf.mtu6 = dev->mtu;
332         ndev->cnf.sysctl = NULL;
333         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
334         if (ndev->nd_parms == NULL) {
335                 kfree(ndev);
336                 return ERR_PTR(err);
337         }
338         if (ndev->cnf.forwarding)
339                 dev_disable_lro(dev);
340         /* We refer to the device */
341         dev_hold(dev);
342
343         if (snmp6_alloc_dev(ndev) < 0) {
344                 ADBG(KERN_WARNING
345                         "%s: cannot allocate memory for statistics; dev=%s.\n",
346                         __func__, dev->name);
347                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
348                 dev_put(dev);
349                 kfree(ndev);
350                 return ERR_PTR(err);
351         }
352
353         if (snmp6_register_dev(ndev) < 0) {
354                 ADBG(KERN_WARNING
355                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
356                         __func__, dev->name);
357                 goto err_release;
358         }
359
360         /* One reference from device.  We must do this before
361          * we invoke __ipv6_regen_rndid().
362          */
363         in6_dev_hold(ndev);
364
365         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
366                 ndev->cnf.accept_dad = -1;
367
368 #if IS_ENABLED(CONFIG_IPV6_SIT)
369         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
370                 pr_info("%s: Disabled Multicast RS\n", dev->name);
371                 ndev->cnf.rtr_solicits = 0;
372         }
373 #endif
374
375         INIT_LIST_HEAD(&ndev->tempaddr_list);
376         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
377         if ((dev->flags&IFF_LOOPBACK) ||
378             dev->type == ARPHRD_TUNNEL ||
379             dev->type == ARPHRD_TUNNEL6 ||
380             dev->type == ARPHRD_SIT ||
381             dev->type == ARPHRD_NONE) {
382                 ndev->cnf.use_tempaddr = -1;
383         } else {
384                 in6_dev_hold(ndev);
385                 ipv6_regen_rndid((unsigned long) ndev);
386         }
387
388         ndev->token = in6addr_any;
389
390         if (netif_running(dev) && addrconf_qdisc_ok(dev))
391                 ndev->if_flags |= IF_READY;
392
393         ipv6_mc_init_dev(ndev);
394         ndev->tstamp = jiffies;
395         err = addrconf_sysctl_register(ndev);
396         if (err) {
397                 ipv6_mc_destroy_dev(ndev);
398                 del_timer(&ndev->regen_timer);
399                 goto err_release;
400         }
401         /* protected by rtnl_lock */
402         rcu_assign_pointer(dev->ip6_ptr, ndev);
403
404         /* Join interface-local all-node multicast group */
405         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
406
407         /* Join all-node multicast group */
408         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
409
410         /* Join all-router multicast group if forwarding is set */
411         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
412                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
413
414         return ndev;
415
416 err_release:
417         neigh_parms_release(&nd_tbl, ndev->nd_parms);
418         ndev->dead = 1;
419         in6_dev_finish_destroy(ndev);
420         return ERR_PTR(err);
421 }
422
423 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
424 {
425         struct inet6_dev *idev;
426
427         ASSERT_RTNL();
428
429         idev = __in6_dev_get(dev);
430         if (!idev) {
431                 idev = ipv6_add_dev(dev);
432                 if (IS_ERR(idev))
433                         return NULL;
434         }
435
436         if (dev->flags&IFF_UP)
437                 ipv6_mc_up(idev);
438         return idev;
439 }
440
441 static int inet6_netconf_msgsize_devconf(int type)
442 {
443         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
444                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
445
446         /* type -1 is used for ALL */
447         if (type == -1 || type == NETCONFA_FORWARDING)
448                 size += nla_total_size(4);
449 #ifdef CONFIG_IPV6_MROUTE
450         if (type == -1 || type == NETCONFA_MC_FORWARDING)
451                 size += nla_total_size(4);
452 #endif
453         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
454                 size += nla_total_size(4);
455
456         return size;
457 }
458
459 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
460                                       struct ipv6_devconf *devconf, u32 portid,
461                                       u32 seq, int event, unsigned int flags,
462                                       int type)
463 {
464         struct nlmsghdr  *nlh;
465         struct netconfmsg *ncm;
466
467         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
468                         flags);
469         if (nlh == NULL)
470                 return -EMSGSIZE;
471
472         ncm = nlmsg_data(nlh);
473         ncm->ncm_family = AF_INET6;
474
475         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
476                 goto nla_put_failure;
477
478         /* type -1 is used for ALL */
479         if ((type == -1 || type == NETCONFA_FORWARDING) &&
480             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
481                 goto nla_put_failure;
482 #ifdef CONFIG_IPV6_MROUTE
483         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
484             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
485                         devconf->mc_forwarding) < 0)
486                 goto nla_put_failure;
487 #endif
488         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
489             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
490                 goto nla_put_failure;
491
492         nlmsg_end(skb, nlh);
493         return 0;
494
495 nla_put_failure:
496         nlmsg_cancel(skb, nlh);
497         return -EMSGSIZE;
498 }
499
500 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
501                                   struct ipv6_devconf *devconf)
502 {
503         struct sk_buff *skb;
504         int err = -ENOBUFS;
505
506         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
507         if (skb == NULL)
508                 goto errout;
509
510         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
511                                          RTM_NEWNETCONF, 0, type);
512         if (err < 0) {
513                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
514                 WARN_ON(err == -EMSGSIZE);
515                 kfree_skb(skb);
516                 goto errout;
517         }
518         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
519         return;
520 errout:
521         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
522 }
523
524 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
525         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
526         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
527         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
528 };
529
530 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
531                                      struct nlmsghdr *nlh)
532 {
533         struct net *net = sock_net(in_skb->sk);
534         struct nlattr *tb[NETCONFA_MAX+1];
535         struct netconfmsg *ncm;
536         struct sk_buff *skb;
537         struct ipv6_devconf *devconf;
538         struct inet6_dev *in6_dev;
539         struct net_device *dev;
540         int ifindex;
541         int err;
542
543         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
544                           devconf_ipv6_policy);
545         if (err < 0)
546                 goto errout;
547
548         err = EINVAL;
549         if (!tb[NETCONFA_IFINDEX])
550                 goto errout;
551
552         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
553         switch (ifindex) {
554         case NETCONFA_IFINDEX_ALL:
555                 devconf = net->ipv6.devconf_all;
556                 break;
557         case NETCONFA_IFINDEX_DEFAULT:
558                 devconf = net->ipv6.devconf_dflt;
559                 break;
560         default:
561                 dev = __dev_get_by_index(net, ifindex);
562                 if (dev == NULL)
563                         goto errout;
564                 in6_dev = __in6_dev_get(dev);
565                 if (in6_dev == NULL)
566                         goto errout;
567                 devconf = &in6_dev->cnf;
568                 break;
569         }
570
571         err = -ENOBUFS;
572         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
573         if (skb == NULL)
574                 goto errout;
575
576         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
577                                          NETLINK_CB(in_skb).portid,
578                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
579                                          -1);
580         if (err < 0) {
581                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
582                 WARN_ON(err == -EMSGSIZE);
583                 kfree_skb(skb);
584                 goto errout;
585         }
586         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
587 errout:
588         return err;
589 }
590
591 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
592                                       struct netlink_callback *cb)
593 {
594         struct net *net = sock_net(skb->sk);
595         int h, s_h;
596         int idx, s_idx;
597         struct net_device *dev;
598         struct inet6_dev *idev;
599         struct hlist_head *head;
600
601         s_h = cb->args[0];
602         s_idx = idx = cb->args[1];
603
604         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
605                 idx = 0;
606                 head = &net->dev_index_head[h];
607                 rcu_read_lock();
608                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
609                           net->dev_base_seq;
610                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
611                         if (idx < s_idx)
612                                 goto cont;
613                         idev = __in6_dev_get(dev);
614                         if (!idev)
615                                 goto cont;
616
617                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
618                                                        &idev->cnf,
619                                                        NETLINK_CB(cb->skb).portid,
620                                                        cb->nlh->nlmsg_seq,
621                                                        RTM_NEWNETCONF,
622                                                        NLM_F_MULTI,
623                                                        -1) < 0) {
624                                 rcu_read_unlock();
625                                 goto done;
626                         }
627                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
628 cont:
629                         idx++;
630                 }
631                 rcu_read_unlock();
632         }
633         if (h == NETDEV_HASHENTRIES) {
634                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
635                                                net->ipv6.devconf_all,
636                                                NETLINK_CB(cb->skb).portid,
637                                                cb->nlh->nlmsg_seq,
638                                                RTM_NEWNETCONF, NLM_F_MULTI,
639                                                -1) < 0)
640                         goto done;
641                 else
642                         h++;
643         }
644         if (h == NETDEV_HASHENTRIES + 1) {
645                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
646                                                net->ipv6.devconf_dflt,
647                                                NETLINK_CB(cb->skb).portid,
648                                                cb->nlh->nlmsg_seq,
649                                                RTM_NEWNETCONF, NLM_F_MULTI,
650                                                -1) < 0)
651                         goto done;
652                 else
653                         h++;
654         }
655 done:
656         cb->args[0] = h;
657         cb->args[1] = idx;
658
659         return skb->len;
660 }
661
662 #ifdef CONFIG_SYSCTL
663 static void dev_forward_change(struct inet6_dev *idev)
664 {
665         struct net_device *dev;
666         struct inet6_ifaddr *ifa;
667
668         if (!idev)
669                 return;
670         dev = idev->dev;
671         if (idev->cnf.forwarding)
672                 dev_disable_lro(dev);
673         if (dev->flags & IFF_MULTICAST) {
674                 if (idev->cnf.forwarding) {
675                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
676                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
677                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
678                 } else {
679                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
680                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
681                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
682                 }
683         }
684
685         list_for_each_entry(ifa, &idev->addr_list, if_list) {
686                 if (ifa->flags&IFA_F_TENTATIVE)
687                         continue;
688                 if (idev->cnf.forwarding)
689                         addrconf_join_anycast(ifa);
690                 else
691                         addrconf_leave_anycast(ifa);
692         }
693         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
694                                      dev->ifindex, &idev->cnf);
695 }
696
697
698 static void addrconf_forward_change(struct net *net, __s32 newf)
699 {
700         struct net_device *dev;
701         struct inet6_dev *idev;
702
703         for_each_netdev(net, dev) {
704                 idev = __in6_dev_get(dev);
705                 if (idev) {
706                         int changed = (!idev->cnf.forwarding) ^ (!newf);
707                         idev->cnf.forwarding = newf;
708                         if (changed)
709                                 dev_forward_change(idev);
710                 }
711         }
712 }
713
714 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
715 {
716         struct net *net;
717         int old;
718
719         if (!rtnl_trylock())
720                 return restart_syscall();
721
722         net = (struct net *)table->extra2;
723         old = *p;
724         *p = newf;
725
726         if (p == &net->ipv6.devconf_dflt->forwarding) {
727                 if ((!newf) ^ (!old))
728                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
729                                                      NETCONFA_IFINDEX_DEFAULT,
730                                                      net->ipv6.devconf_dflt);
731                 rtnl_unlock();
732                 return 0;
733         }
734
735         if (p == &net->ipv6.devconf_all->forwarding) {
736                 net->ipv6.devconf_dflt->forwarding = newf;
737                 addrconf_forward_change(net, newf);
738                 if ((!newf) ^ (!old))
739                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
740                                                      NETCONFA_IFINDEX_ALL,
741                                                      net->ipv6.devconf_all);
742         } else if ((!newf) ^ (!old))
743                 dev_forward_change((struct inet6_dev *)table->extra1);
744         rtnl_unlock();
745
746         if (newf)
747                 rt6_purge_dflt_routers(net);
748         return 1;
749 }
750 #endif
751
752 /* Nobody refers to this ifaddr, destroy it */
753 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
754 {
755         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
756
757 #ifdef NET_REFCNT_DEBUG
758         pr_debug("%s\n", __func__);
759 #endif
760
761         in6_dev_put(ifp->idev);
762
763         if (cancel_delayed_work(&ifp->dad_work))
764                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
765                           ifp);
766
767         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
768                 pr_warn("Freeing alive inet6 address %p\n", ifp);
769                 return;
770         }
771         ip6_rt_put(ifp->rt);
772
773         kfree_rcu(ifp, rcu);
774 }
775
776 static void
777 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
778 {
779         struct list_head *p;
780         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
781
782         /*
783          * Each device address list is sorted in order of scope -
784          * global before linklocal.
785          */
786         list_for_each(p, &idev->addr_list) {
787                 struct inet6_ifaddr *ifa
788                         = list_entry(p, struct inet6_ifaddr, if_list);
789                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
790                         break;
791         }
792
793         list_add_tail(&ifp->if_list, p);
794 }
795
796 static u32 inet6_addr_hash(const struct in6_addr *addr)
797 {
798         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
799 }
800
801 /* On success it returns ifp with increased reference count */
802
803 static struct inet6_ifaddr *
804 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
805               const struct in6_addr *peer_addr, int pfxlen,
806               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
807 {
808         struct inet6_ifaddr *ifa = NULL;
809         struct rt6_info *rt;
810         unsigned int hash;
811         int err = 0;
812         int addr_type = ipv6_addr_type(addr);
813
814         if (addr_type == IPV6_ADDR_ANY ||
815             addr_type & IPV6_ADDR_MULTICAST ||
816             (!(idev->dev->flags & IFF_LOOPBACK) &&
817              addr_type & IPV6_ADDR_LOOPBACK))
818                 return ERR_PTR(-EADDRNOTAVAIL);
819
820         rcu_read_lock_bh();
821         if (idev->dead) {
822                 err = -ENODEV;                  /*XXX*/
823                 goto out2;
824         }
825
826         if (idev->cnf.disable_ipv6) {
827                 err = -EACCES;
828                 goto out2;
829         }
830
831         spin_lock(&addrconf_hash_lock);
832
833         /* Ignore adding duplicate addresses on an interface */
834         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
835                 ADBG("ipv6_add_addr: already assigned\n");
836                 err = -EEXIST;
837                 goto out;
838         }
839
840         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
841
842         if (ifa == NULL) {
843                 ADBG("ipv6_add_addr: malloc failed\n");
844                 err = -ENOBUFS;
845                 goto out;
846         }
847
848         rt = addrconf_dst_alloc(idev, addr, false);
849         if (IS_ERR(rt)) {
850                 err = PTR_ERR(rt);
851                 goto out;
852         }
853
854         neigh_parms_data_state_setall(idev->nd_parms);
855
856         ifa->addr = *addr;
857         if (peer_addr)
858                 ifa->peer_addr = *peer_addr;
859
860         spin_lock_init(&ifa->lock);
861         spin_lock_init(&ifa->state_lock);
862         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
863         INIT_HLIST_NODE(&ifa->addr_lst);
864         ifa->scope = scope;
865         ifa->prefix_len = pfxlen;
866         ifa->flags = flags | IFA_F_TENTATIVE;
867         ifa->valid_lft = valid_lft;
868         ifa->prefered_lft = prefered_lft;
869         ifa->cstamp = ifa->tstamp = jiffies;
870         ifa->tokenized = false;
871
872         ifa->rt = rt;
873
874         ifa->idev = idev;
875         in6_dev_hold(idev);
876         /* For caller */
877         in6_ifa_hold(ifa);
878
879         /* Add to big hash table */
880         hash = inet6_addr_hash(addr);
881
882         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
883         spin_unlock(&addrconf_hash_lock);
884
885         write_lock(&idev->lock);
886         /* Add to inet6_dev unicast addr list. */
887         ipv6_link_dev_addr(idev, ifa);
888
889         if (ifa->flags&IFA_F_TEMPORARY) {
890                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
891                 in6_ifa_hold(ifa);
892         }
893
894         in6_ifa_hold(ifa);
895         write_unlock(&idev->lock);
896 out2:
897         rcu_read_unlock_bh();
898
899         if (likely(err == 0))
900                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
901         else {
902                 kfree(ifa);
903                 ifa = ERR_PTR(err);
904         }
905
906         return ifa;
907 out:
908         spin_unlock(&addrconf_hash_lock);
909         goto out2;
910 }
911
912 enum cleanup_prefix_rt_t {
913         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
914         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
915         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
916 };
917
918 /*
919  * Check, whether the prefix for ifp would still need a prefix route
920  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
921  * constants.
922  *
923  * 1) we don't purge prefix if address was not permanent.
924  *    prefix is managed by its own lifetime.
925  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
926  * 3) if there are no addresses, delete prefix.
927  * 4) if there are still other permanent address(es),
928  *    corresponding prefix is still permanent.
929  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
930  *    don't purge the prefix, assume user space is managing it.
931  * 6) otherwise, update prefix lifetime to the
932  *    longest valid lifetime among the corresponding
933  *    addresses on the device.
934  *    Note: subsequent RA will update lifetime.
935  **/
936 static enum cleanup_prefix_rt_t
937 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
938 {
939         struct inet6_ifaddr *ifa;
940         struct inet6_dev *idev = ifp->idev;
941         unsigned long lifetime;
942         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
943
944         *expires = jiffies;
945
946         list_for_each_entry(ifa, &idev->addr_list, if_list) {
947                 if (ifa == ifp)
948                         continue;
949                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
950                                        ifp->prefix_len))
951                         continue;
952                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
953                         return CLEANUP_PREFIX_RT_NOP;
954
955                 action = CLEANUP_PREFIX_RT_EXPIRE;
956
957                 spin_lock(&ifa->lock);
958
959                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
960                 /*
961                  * Note: Because this address is
962                  * not permanent, lifetime <
963                  * LONG_MAX / HZ here.
964                  */
965                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
966                         *expires = ifa->tstamp + lifetime * HZ;
967                 spin_unlock(&ifa->lock);
968         }
969
970         return action;
971 }
972
973 static void
974 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
975 {
976         struct rt6_info *rt;
977
978         rt = addrconf_get_prefix_route(&ifp->addr,
979                                        ifp->prefix_len,
980                                        ifp->idev->dev,
981                                        0, RTF_GATEWAY | RTF_DEFAULT);
982         if (rt) {
983                 if (del_rt)
984                         ip6_del_rt(rt);
985                 else {
986                         if (!(rt->rt6i_flags & RTF_EXPIRES))
987                                 rt6_set_expires(rt, expires);
988                         ip6_rt_put(rt);
989                 }
990         }
991 }
992
993
994 /* This function wants to get referenced ifp and releases it before return */
995
996 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
997 {
998         int state;
999         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1000         unsigned long expires;
1001
1002         ASSERT_RTNL();
1003
1004         spin_lock_bh(&ifp->state_lock);
1005         state = ifp->state;
1006         ifp->state = INET6_IFADDR_STATE_DEAD;
1007         spin_unlock_bh(&ifp->state_lock);
1008
1009         if (state == INET6_IFADDR_STATE_DEAD)
1010                 goto out;
1011
1012         spin_lock_bh(&addrconf_hash_lock);
1013         hlist_del_init_rcu(&ifp->addr_lst);
1014         spin_unlock_bh(&addrconf_hash_lock);
1015
1016         write_lock_bh(&ifp->idev->lock);
1017
1018         if (ifp->flags&IFA_F_TEMPORARY) {
1019                 list_del(&ifp->tmp_list);
1020                 if (ifp->ifpub) {
1021                         in6_ifa_put(ifp->ifpub);
1022                         ifp->ifpub = NULL;
1023                 }
1024                 __in6_ifa_put(ifp);
1025         }
1026
1027         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1028                 action = check_cleanup_prefix_route(ifp, &expires);
1029
1030         list_del_init(&ifp->if_list);
1031         __in6_ifa_put(ifp);
1032
1033         write_unlock_bh(&ifp->idev->lock);
1034
1035         addrconf_del_dad_work(ifp);
1036
1037         ipv6_ifa_notify(RTM_DELADDR, ifp);
1038
1039         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1040
1041         if (action != CLEANUP_PREFIX_RT_NOP) {
1042                 cleanup_prefix_route(ifp, expires,
1043                         action == CLEANUP_PREFIX_RT_DEL);
1044         }
1045
1046         /* clean up prefsrc entries */
1047         rt6_remove_prefsrc(ifp);
1048 out:
1049         in6_ifa_put(ifp);
1050 }
1051
1052 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1053 {
1054         struct inet6_dev *idev = ifp->idev;
1055         struct in6_addr addr, *tmpaddr;
1056         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1057         unsigned long regen_advance;
1058         int tmp_plen;
1059         int ret = 0;
1060         u32 addr_flags;
1061         unsigned long now = jiffies;
1062
1063         write_lock_bh(&idev->lock);
1064         if (ift) {
1065                 spin_lock_bh(&ift->lock);
1066                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1067                 spin_unlock_bh(&ift->lock);
1068                 tmpaddr = &addr;
1069         } else {
1070                 tmpaddr = NULL;
1071         }
1072 retry:
1073         in6_dev_hold(idev);
1074         if (idev->cnf.use_tempaddr <= 0) {
1075                 write_unlock_bh(&idev->lock);
1076                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1077                 in6_dev_put(idev);
1078                 ret = -1;
1079                 goto out;
1080         }
1081         spin_lock_bh(&ifp->lock);
1082         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1083                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1084                 spin_unlock_bh(&ifp->lock);
1085                 write_unlock_bh(&idev->lock);
1086                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1087                         __func__);
1088                 in6_dev_put(idev);
1089                 ret = -1;
1090                 goto out;
1091         }
1092         in6_ifa_hold(ifp);
1093         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1094         __ipv6_try_regen_rndid(idev, tmpaddr);
1095         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1096         age = (now - ifp->tstamp) / HZ;
1097         tmp_valid_lft = min_t(__u32,
1098                               ifp->valid_lft,
1099                               idev->cnf.temp_valid_lft + age);
1100         tmp_prefered_lft = min_t(__u32,
1101                                  ifp->prefered_lft,
1102                                  idev->cnf.temp_prefered_lft + age -
1103                                  idev->cnf.max_desync_factor);
1104         tmp_plen = ifp->prefix_len;
1105         tmp_tstamp = ifp->tstamp;
1106         spin_unlock_bh(&ifp->lock);
1107
1108         regen_advance = idev->cnf.regen_max_retry *
1109                         idev->cnf.dad_transmits *
1110                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1111         write_unlock_bh(&idev->lock);
1112
1113         /* A temporary address is created only if this calculated Preferred
1114          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1115          * an implementation must not create a temporary address with a zero
1116          * Preferred Lifetime.
1117          * Use age calculation as in addrconf_verify to avoid unnecessary
1118          * temporary addresses being generated.
1119          */
1120         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1121         if (tmp_prefered_lft <= regen_advance + age) {
1122                 in6_ifa_put(ifp);
1123                 in6_dev_put(idev);
1124                 ret = -1;
1125                 goto out;
1126         }
1127
1128         addr_flags = IFA_F_TEMPORARY;
1129         /* set in addrconf_prefix_rcv() */
1130         if (ifp->flags & IFA_F_OPTIMISTIC)
1131                 addr_flags |= IFA_F_OPTIMISTIC;
1132
1133         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1134                             ipv6_addr_scope(&addr), addr_flags,
1135                             tmp_valid_lft, tmp_prefered_lft);
1136         if (IS_ERR(ift)) {
1137                 in6_ifa_put(ifp);
1138                 in6_dev_put(idev);
1139                 pr_info("%s: retry temporary address regeneration\n", __func__);
1140                 tmpaddr = &addr;
1141                 write_lock_bh(&idev->lock);
1142                 goto retry;
1143         }
1144
1145         spin_lock_bh(&ift->lock);
1146         ift->ifpub = ifp;
1147         ift->cstamp = now;
1148         ift->tstamp = tmp_tstamp;
1149         spin_unlock_bh(&ift->lock);
1150
1151         addrconf_dad_start(ift);
1152         in6_ifa_put(ift);
1153         in6_dev_put(idev);
1154 out:
1155         return ret;
1156 }
1157
1158 /*
1159  *      Choose an appropriate source address (RFC3484)
1160  */
1161 enum {
1162         IPV6_SADDR_RULE_INIT = 0,
1163         IPV6_SADDR_RULE_LOCAL,
1164         IPV6_SADDR_RULE_SCOPE,
1165         IPV6_SADDR_RULE_PREFERRED,
1166 #ifdef CONFIG_IPV6_MIP6
1167         IPV6_SADDR_RULE_HOA,
1168 #endif
1169         IPV6_SADDR_RULE_OIF,
1170         IPV6_SADDR_RULE_LABEL,
1171         IPV6_SADDR_RULE_PRIVACY,
1172         IPV6_SADDR_RULE_ORCHID,
1173         IPV6_SADDR_RULE_PREFIX,
1174 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1175         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1176 #endif
1177         IPV6_SADDR_RULE_MAX
1178 };
1179
1180 struct ipv6_saddr_score {
1181         int                     rule;
1182         int                     addr_type;
1183         struct inet6_ifaddr     *ifa;
1184         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1185         int                     scopedist;
1186         int                     matchlen;
1187 };
1188
1189 struct ipv6_saddr_dst {
1190         const struct in6_addr *addr;
1191         int ifindex;
1192         int scope;
1193         int label;
1194         unsigned int prefs;
1195 };
1196
1197 static inline int ipv6_saddr_preferred(int type)
1198 {
1199         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1200                 return 1;
1201         return 0;
1202 }
1203
1204 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1205 {
1206 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1207         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1208 #else
1209         return false;
1210 #endif
1211 }
1212
1213 static int ipv6_get_saddr_eval(struct net *net,
1214                                struct ipv6_saddr_score *score,
1215                                struct ipv6_saddr_dst *dst,
1216                                int i)
1217 {
1218         int ret;
1219
1220         if (i <= score->rule) {
1221                 switch (i) {
1222                 case IPV6_SADDR_RULE_SCOPE:
1223                         ret = score->scopedist;
1224                         break;
1225                 case IPV6_SADDR_RULE_PREFIX:
1226                         ret = score->matchlen;
1227                         break;
1228                 default:
1229                         ret = !!test_bit(i, score->scorebits);
1230                 }
1231                 goto out;
1232         }
1233
1234         switch (i) {
1235         case IPV6_SADDR_RULE_INIT:
1236                 /* Rule 0: remember if hiscore is not ready yet */
1237                 ret = !!score->ifa;
1238                 break;
1239         case IPV6_SADDR_RULE_LOCAL:
1240                 /* Rule 1: Prefer same address */
1241                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1242                 break;
1243         case IPV6_SADDR_RULE_SCOPE:
1244                 /* Rule 2: Prefer appropriate scope
1245                  *
1246                  *      ret
1247                  *       ^
1248                  *    -1 |  d 15
1249                  *    ---+--+-+---> scope
1250                  *       |
1251                  *       |             d is scope of the destination.
1252                  *  B-d  |  \
1253                  *       |   \      <- smaller scope is better if
1254                  *  B-15 |    \        if scope is enough for destination.
1255                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1256                  * d-C-1 | /
1257                  *       |/         <- greater is better
1258                  *   -C  /             if scope is not enough for destination.
1259                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1260                  *
1261                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1262                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1263                  * Assume B = 0 and we get C > 29.
1264                  */
1265                 ret = __ipv6_addr_src_scope(score->addr_type);
1266                 if (ret >= dst->scope)
1267                         ret = -ret;
1268                 else
1269                         ret -= 128;     /* 30 is enough */
1270                 score->scopedist = ret;
1271                 break;
1272         case IPV6_SADDR_RULE_PREFERRED:
1273             {
1274                 /* Rule 3: Avoid deprecated and optimistic addresses */
1275                 u8 avoid = IFA_F_DEPRECATED;
1276
1277                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1278                         avoid |= IFA_F_OPTIMISTIC;
1279                 ret = ipv6_saddr_preferred(score->addr_type) ||
1280                       !(score->ifa->flags & avoid);
1281                 break;
1282             }
1283 #ifdef CONFIG_IPV6_MIP6
1284         case IPV6_SADDR_RULE_HOA:
1285             {
1286                 /* Rule 4: Prefer home address */
1287                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1288                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1289                 break;
1290             }
1291 #endif
1292         case IPV6_SADDR_RULE_OIF:
1293                 /* Rule 5: Prefer outgoing interface */
1294                 ret = (!dst->ifindex ||
1295                        dst->ifindex == score->ifa->idev->dev->ifindex);
1296                 break;
1297         case IPV6_SADDR_RULE_LABEL:
1298                 /* Rule 6: Prefer matching label */
1299                 ret = ipv6_addr_label(net,
1300                                       &score->ifa->addr, score->addr_type,
1301                                       score->ifa->idev->dev->ifindex) == dst->label;
1302                 break;
1303         case IPV6_SADDR_RULE_PRIVACY:
1304             {
1305                 /* Rule 7: Prefer public address
1306                  * Note: prefer temporary address if use_tempaddr >= 2
1307                  */
1308                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1309                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1310                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1311                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1312                 break;
1313             }
1314         case IPV6_SADDR_RULE_ORCHID:
1315                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1316                  *          non-ORCHID vs non-ORCHID
1317                  */
1318                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1319                         ipv6_addr_orchid(dst->addr));
1320                 break;
1321         case IPV6_SADDR_RULE_PREFIX:
1322                 /* Rule 8: Use longest matching prefix */
1323                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1324                 if (ret > score->ifa->prefix_len)
1325                         ret = score->ifa->prefix_len;
1326                 score->matchlen = ret;
1327                 break;
1328 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1329         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1330                 /* Optimistic addresses still have lower precedence than other
1331                  * preferred addresses.
1332                  */
1333                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1334                 break;
1335 #endif
1336         default:
1337                 ret = 0;
1338         }
1339
1340         if (ret)
1341                 __set_bit(i, score->scorebits);
1342         score->rule = i;
1343 out:
1344         return ret;
1345 }
1346
1347 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1348                        const struct in6_addr *daddr, unsigned int prefs,
1349                        struct in6_addr *saddr)
1350 {
1351         struct ipv6_saddr_score scores[2],
1352                                 *score = &scores[0], *hiscore = &scores[1];
1353         struct ipv6_saddr_dst dst;
1354         struct net_device *dev;
1355         int dst_type;
1356
1357         dst_type = __ipv6_addr_type(daddr);
1358         dst.addr = daddr;
1359         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1360         dst.scope = __ipv6_addr_src_scope(dst_type);
1361         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1362         dst.prefs = prefs;
1363
1364         hiscore->rule = -1;
1365         hiscore->ifa = NULL;
1366
1367         rcu_read_lock();
1368
1369         for_each_netdev_rcu(net, dev) {
1370                 struct inet6_dev *idev;
1371
1372                 /* Candidate Source Address (section 4)
1373                  *  - multicast and link-local destination address,
1374                  *    the set of candidate source address MUST only
1375                  *    include addresses assigned to interfaces
1376                  *    belonging to the same link as the outgoing
1377                  *    interface.
1378                  * (- For site-local destination addresses, the
1379                  *    set of candidate source addresses MUST only
1380                  *    include addresses assigned to interfaces
1381                  *    belonging to the same site as the outgoing
1382                  *    interface.)
1383                  */
1384                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1385                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1386                     dst.ifindex && dev->ifindex != dst.ifindex)
1387                         continue;
1388
1389                 idev = __in6_dev_get(dev);
1390                 if (!idev)
1391                         continue;
1392
1393                 read_lock_bh(&idev->lock);
1394                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1395                         int i;
1396
1397                         /*
1398                          * - Tentative Address (RFC2462 section 5.4)
1399                          *  - A tentative address is not considered
1400                          *    "assigned to an interface" in the traditional
1401                          *    sense, unless it is also flagged as optimistic.
1402                          * - Candidate Source Address (section 4)
1403                          *  - In any case, anycast addresses, multicast
1404                          *    addresses, and the unspecified address MUST
1405                          *    NOT be included in a candidate set.
1406                          */
1407                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1408                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1409                                 continue;
1410
1411                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1412
1413                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1414                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1415                                 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1416                                                     dev->name);
1417                                 continue;
1418                         }
1419
1420                         score->rule = -1;
1421                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1422
1423                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1424                                 int minihiscore, miniscore;
1425
1426                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1427                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1428
1429                                 if (minihiscore > miniscore) {
1430                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1431                                             score->scopedist > 0) {
1432                                                 /*
1433                                                  * special case:
1434                                                  * each remaining entry
1435                                                  * has too small (not enough)
1436                                                  * scope, because ifa entries
1437                                                  * are sorted by their scope
1438                                                  * values.
1439                                                  */
1440                                                 goto try_nextdev;
1441                                         }
1442                                         break;
1443                                 } else if (minihiscore < miniscore) {
1444                                         if (hiscore->ifa)
1445                                                 in6_ifa_put(hiscore->ifa);
1446
1447                                         in6_ifa_hold(score->ifa);
1448
1449                                         swap(hiscore, score);
1450
1451                                         /* restore our iterator */
1452                                         score->ifa = hiscore->ifa;
1453
1454                                         break;
1455                                 }
1456                         }
1457                 }
1458 try_nextdev:
1459                 read_unlock_bh(&idev->lock);
1460         }
1461         rcu_read_unlock();
1462
1463         if (!hiscore->ifa)
1464                 return -EADDRNOTAVAIL;
1465
1466         *saddr = hiscore->ifa->addr;
1467         in6_ifa_put(hiscore->ifa);
1468         return 0;
1469 }
1470 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1471
1472 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1473                       u32 banned_flags)
1474 {
1475         struct inet6_ifaddr *ifp;
1476         int err = -EADDRNOTAVAIL;
1477
1478         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1479                 if (ifp->scope > IFA_LINK)
1480                         break;
1481                 if (ifp->scope == IFA_LINK &&
1482                     !(ifp->flags & banned_flags)) {
1483                         *addr = ifp->addr;
1484                         err = 0;
1485                         break;
1486                 }
1487         }
1488         return err;
1489 }
1490
1491 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1492                     u32 banned_flags)
1493 {
1494         struct inet6_dev *idev;
1495         int err = -EADDRNOTAVAIL;
1496
1497         rcu_read_lock();
1498         idev = __in6_dev_get(dev);
1499         if (idev) {
1500                 read_lock_bh(&idev->lock);
1501                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1502                 read_unlock_bh(&idev->lock);
1503         }
1504         rcu_read_unlock();
1505         return err;
1506 }
1507
1508 static int ipv6_count_addresses(struct inet6_dev *idev)
1509 {
1510         int cnt = 0;
1511         struct inet6_ifaddr *ifp;
1512
1513         read_lock_bh(&idev->lock);
1514         list_for_each_entry(ifp, &idev->addr_list, if_list)
1515                 cnt++;
1516         read_unlock_bh(&idev->lock);
1517         return cnt;
1518 }
1519
1520 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1521                   const struct net_device *dev, int strict)
1522 {
1523         struct inet6_ifaddr *ifp;
1524         unsigned int hash = inet6_addr_hash(addr);
1525
1526         rcu_read_lock_bh();
1527         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1528                 if (!net_eq(dev_net(ifp->idev->dev), net))
1529                         continue;
1530                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1531                     !(ifp->flags&IFA_F_TENTATIVE) &&
1532                     (dev == NULL || ifp->idev->dev == dev ||
1533                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1534                         rcu_read_unlock_bh();
1535                         return 1;
1536                 }
1537         }
1538
1539         rcu_read_unlock_bh();
1540         return 0;
1541 }
1542 EXPORT_SYMBOL(ipv6_chk_addr);
1543
1544 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1545                                struct net_device *dev)
1546 {
1547         unsigned int hash = inet6_addr_hash(addr);
1548         struct inet6_ifaddr *ifp;
1549
1550         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1551                 if (!net_eq(dev_net(ifp->idev->dev), net))
1552                         continue;
1553                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1554                         if (dev == NULL || ifp->idev->dev == dev)
1555                                 return true;
1556                 }
1557         }
1558         return false;
1559 }
1560
1561 /* Compares an address/prefix_len with addresses on device @dev.
1562  * If one is found it returns true.
1563  */
1564 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1565         const unsigned int prefix_len, struct net_device *dev)
1566 {
1567         struct inet6_dev *idev;
1568         struct inet6_ifaddr *ifa;
1569         bool ret = false;
1570
1571         rcu_read_lock();
1572         idev = __in6_dev_get(dev);
1573         if (idev) {
1574                 read_lock_bh(&idev->lock);
1575                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1576                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1577                         if (ret)
1578                                 break;
1579                 }
1580                 read_unlock_bh(&idev->lock);
1581         }
1582         rcu_read_unlock();
1583
1584         return ret;
1585 }
1586 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1587
1588 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1589 {
1590         struct inet6_dev *idev;
1591         struct inet6_ifaddr *ifa;
1592         int     onlink;
1593
1594         onlink = 0;
1595         rcu_read_lock();
1596         idev = __in6_dev_get(dev);
1597         if (idev) {
1598                 read_lock_bh(&idev->lock);
1599                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1600                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1601                                                    ifa->prefix_len);
1602                         if (onlink)
1603                                 break;
1604                 }
1605                 read_unlock_bh(&idev->lock);
1606         }
1607         rcu_read_unlock();
1608         return onlink;
1609 }
1610 EXPORT_SYMBOL(ipv6_chk_prefix);
1611
1612 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1613                                      struct net_device *dev, int strict)
1614 {
1615         struct inet6_ifaddr *ifp, *result = NULL;
1616         unsigned int hash = inet6_addr_hash(addr);
1617
1618         rcu_read_lock_bh();
1619         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1620                 if (!net_eq(dev_net(ifp->idev->dev), net))
1621                         continue;
1622                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1623                         if (dev == NULL || ifp->idev->dev == dev ||
1624                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1625                                 result = ifp;
1626                                 in6_ifa_hold(ifp);
1627                                 break;
1628                         }
1629                 }
1630         }
1631         rcu_read_unlock_bh();
1632
1633         return result;
1634 }
1635
1636 /* Gets referenced address, destroys ifaddr */
1637
1638 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1639 {
1640         if (ifp->flags&IFA_F_PERMANENT) {
1641                 spin_lock_bh(&ifp->lock);
1642                 addrconf_del_dad_work(ifp);
1643                 ifp->flags |= IFA_F_TENTATIVE;
1644                 if (dad_failed)
1645                         ifp->flags |= IFA_F_DADFAILED;
1646                 spin_unlock_bh(&ifp->lock);
1647                 if (dad_failed)
1648                         ipv6_ifa_notify(0, ifp);
1649                 in6_ifa_put(ifp);
1650         } else if (ifp->flags&IFA_F_TEMPORARY) {
1651                 struct inet6_ifaddr *ifpub;
1652                 spin_lock_bh(&ifp->lock);
1653                 ifpub = ifp->ifpub;
1654                 if (ifpub) {
1655                         in6_ifa_hold(ifpub);
1656                         spin_unlock_bh(&ifp->lock);
1657                         ipv6_create_tempaddr(ifpub, ifp);
1658                         in6_ifa_put(ifpub);
1659                 } else {
1660                         spin_unlock_bh(&ifp->lock);
1661                 }
1662                 ipv6_del_addr(ifp);
1663         } else {
1664                 ipv6_del_addr(ifp);
1665         }
1666 }
1667
1668 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1669 {
1670         int err = -ENOENT;
1671
1672         spin_lock_bh(&ifp->state_lock);
1673         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1674                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1675                 err = 0;
1676         }
1677         spin_unlock_bh(&ifp->state_lock);
1678
1679         return err;
1680 }
1681
1682 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1683 {
1684         struct inet6_dev *idev = ifp->idev;
1685
1686         if (addrconf_dad_end(ifp)) {
1687                 in6_ifa_put(ifp);
1688                 return;
1689         }
1690
1691         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1692                              ifp->idev->dev->name, &ifp->addr);
1693
1694         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1695                 struct in6_addr addr;
1696
1697                 addr.s6_addr32[0] = htonl(0xfe800000);
1698                 addr.s6_addr32[1] = 0;
1699
1700                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1701                     ipv6_addr_equal(&ifp->addr, &addr)) {
1702                         /* DAD failed for link-local based on MAC address */
1703                         idev->cnf.disable_ipv6 = 1;
1704
1705                         pr_info("%s: IPv6 being disabled!\n",
1706                                 ifp->idev->dev->name);
1707                 }
1708         }
1709
1710         spin_lock_bh(&ifp->state_lock);
1711         /* transition from _POSTDAD to _ERRDAD */
1712         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1713         spin_unlock_bh(&ifp->state_lock);
1714
1715         addrconf_mod_dad_work(ifp, 0);
1716 }
1717
1718 /* Join to solicited addr multicast group.
1719  * caller must hold RTNL */
1720 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1721 {
1722         struct in6_addr maddr;
1723
1724         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1725                 return;
1726
1727         addrconf_addr_solict_mult(addr, &maddr);
1728         ipv6_dev_mc_inc(dev, &maddr);
1729 }
1730
1731 /* caller must hold RTNL */
1732 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1733 {
1734         struct in6_addr maddr;
1735
1736         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1737                 return;
1738
1739         addrconf_addr_solict_mult(addr, &maddr);
1740         __ipv6_dev_mc_dec(idev, &maddr);
1741 }
1742
1743 /* caller must hold RTNL */
1744 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1745 {
1746         struct in6_addr addr;
1747
1748         if (ifp->prefix_len >= 127) /* RFC 6164 */
1749                 return;
1750         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1751         if (ipv6_addr_any(&addr))
1752                 return;
1753         __ipv6_dev_ac_inc(ifp->idev, &addr);
1754 }
1755
1756 /* caller must hold RTNL */
1757 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1758 {
1759         struct in6_addr addr;
1760
1761         if (ifp->prefix_len >= 127) /* RFC 6164 */
1762                 return;
1763         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1764         if (ipv6_addr_any(&addr))
1765                 return;
1766         __ipv6_dev_ac_dec(ifp->idev, &addr);
1767 }
1768
1769 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1770 {
1771         if (dev->addr_len != ETH_ALEN)
1772                 return -1;
1773         memcpy(eui, dev->dev_addr, 3);
1774         memcpy(eui + 5, dev->dev_addr + 3, 3);
1775
1776         /*
1777          * The zSeries OSA network cards can be shared among various
1778          * OS instances, but the OSA cards have only one MAC address.
1779          * This leads to duplicate address conflicts in conjunction
1780          * with IPv6 if more than one instance uses the same card.
1781          *
1782          * The driver for these cards can deliver a unique 16-bit
1783          * identifier for each instance sharing the same card.  It is
1784          * placed instead of 0xFFFE in the interface identifier.  The
1785          * "u" bit of the interface identifier is not inverted in this
1786          * case.  Hence the resulting interface identifier has local
1787          * scope according to RFC2373.
1788          */
1789         if (dev->dev_id) {
1790                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1791                 eui[4] = dev->dev_id & 0xFF;
1792         } else {
1793                 eui[3] = 0xFF;
1794                 eui[4] = 0xFE;
1795                 eui[0] ^= 2;
1796         }
1797         return 0;
1798 }
1799
1800 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1801 {
1802         if (dev->addr_len != IEEE802154_ADDR_LEN)
1803                 return -1;
1804         memcpy(eui, dev->dev_addr, 8);
1805         eui[0] ^= 2;
1806         return 0;
1807 }
1808
1809 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1810 {
1811         union fwnet_hwaddr *ha;
1812
1813         if (dev->addr_len != FWNET_ALEN)
1814                 return -1;
1815
1816         ha = (union fwnet_hwaddr *)dev->dev_addr;
1817
1818         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1819         eui[0] ^= 2;
1820         return 0;
1821 }
1822
1823 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1824 {
1825         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1826         if (dev->addr_len != ARCNET_ALEN)
1827                 return -1;
1828         memset(eui, 0, 7);
1829         eui[7] = *(u8 *)dev->dev_addr;
1830         return 0;
1831 }
1832
1833 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1834 {
1835         if (dev->addr_len != INFINIBAND_ALEN)
1836                 return -1;
1837         memcpy(eui, dev->dev_addr + 12, 8);
1838         eui[0] |= 2;
1839         return 0;
1840 }
1841
1842 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1843 {
1844         if (addr == 0)
1845                 return -1;
1846         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1847                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1848                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1849                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1850                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1851                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1852         eui[1] = 0;
1853         eui[2] = 0x5E;
1854         eui[3] = 0xFE;
1855         memcpy(eui + 4, &addr, 4);
1856         return 0;
1857 }
1858
1859 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1860 {
1861         if (dev->priv_flags & IFF_ISATAP)
1862                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1863         return -1;
1864 }
1865
1866 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1867 {
1868         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1869 }
1870
1871 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1872 {
1873         memcpy(eui, dev->perm_addr, 3);
1874         memcpy(eui + 5, dev->perm_addr + 3, 3);
1875         eui[3] = 0xFF;
1876         eui[4] = 0xFE;
1877         eui[0] ^= 2;
1878         return 0;
1879 }
1880
1881 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1882 {
1883         switch (dev->type) {
1884         case ARPHRD_ETHER:
1885         case ARPHRD_FDDI:
1886                 return addrconf_ifid_eui48(eui, dev);
1887         case ARPHRD_ARCNET:
1888                 return addrconf_ifid_arcnet(eui, dev);
1889         case ARPHRD_INFINIBAND:
1890                 return addrconf_ifid_infiniband(eui, dev);
1891         case ARPHRD_SIT:
1892                 return addrconf_ifid_sit(eui, dev);
1893         case ARPHRD_IPGRE:
1894                 return addrconf_ifid_gre(eui, dev);
1895         case ARPHRD_6LOWPAN:
1896         case ARPHRD_IEEE802154:
1897                 return addrconf_ifid_eui64(eui, dev);
1898         case ARPHRD_IEEE1394:
1899                 return addrconf_ifid_ieee1394(eui, dev);
1900         case ARPHRD_TUNNEL6:
1901                 return addrconf_ifid_ip6tnl(eui, dev);
1902         }
1903         return -1;
1904 }
1905
1906 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1907 {
1908         int err = -1;
1909         struct inet6_ifaddr *ifp;
1910
1911         read_lock_bh(&idev->lock);
1912         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1913                 if (ifp->scope > IFA_LINK)
1914                         break;
1915                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1916                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1917                         err = 0;
1918                         break;
1919                 }
1920         }
1921         read_unlock_bh(&idev->lock);
1922         return err;
1923 }
1924
1925 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1926 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1927 {
1928 regen:
1929         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1930         idev->rndid[0] &= ~0x02;
1931
1932         /*
1933          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1934          * check if generated address is not inappropriate
1935          *
1936          *  - Reserved subnet anycast (RFC 2526)
1937          *      11111101 11....11 1xxxxxxx
1938          *  - ISATAP (RFC4214) 6.1
1939          *      00-00-5E-FE-xx-xx-xx-xx
1940          *  - value 0
1941          *  - XXX: already assigned to an address on the device
1942          */
1943         if (idev->rndid[0] == 0xfd &&
1944             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1945             (idev->rndid[7]&0x80))
1946                 goto regen;
1947         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1948                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1949                         goto regen;
1950                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1951                         goto regen;
1952         }
1953 }
1954
1955 static void ipv6_regen_rndid(unsigned long data)
1956 {
1957         struct inet6_dev *idev = (struct inet6_dev *) data;
1958         unsigned long expires;
1959
1960         rcu_read_lock_bh();
1961         write_lock_bh(&idev->lock);
1962
1963         if (idev->dead)
1964                 goto out;
1965
1966         __ipv6_regen_rndid(idev);
1967
1968         expires = jiffies +
1969                 idev->cnf.temp_prefered_lft * HZ -
1970                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1971                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1972                 idev->cnf.max_desync_factor * HZ;
1973         if (time_before(expires, jiffies)) {
1974                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1975                         __func__, idev->dev->name);
1976                 goto out;
1977         }
1978
1979         if (!mod_timer(&idev->regen_timer, expires))
1980                 in6_dev_hold(idev);
1981
1982 out:
1983         write_unlock_bh(&idev->lock);
1984         rcu_read_unlock_bh();
1985         in6_dev_put(idev);
1986 }
1987
1988 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1989 {
1990         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1991                 __ipv6_regen_rndid(idev);
1992 }
1993
1994 /*
1995  *      Add prefix route.
1996  */
1997
1998 static void
1999 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2000                       unsigned long expires, u32 flags)
2001 {
2002         struct fib6_config cfg = {
2003                 .fc_table = RT6_TABLE_PREFIX,
2004                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2005                 .fc_ifindex = dev->ifindex,
2006                 .fc_expires = expires,
2007                 .fc_dst_len = plen,
2008                 .fc_flags = RTF_UP | flags,
2009                 .fc_nlinfo.nl_net = dev_net(dev),
2010                 .fc_protocol = RTPROT_KERNEL,
2011         };
2012
2013         cfg.fc_dst = *pfx;
2014
2015         /* Prevent useless cloning on PtP SIT.
2016            This thing is done here expecting that the whole
2017            class of non-broadcast devices need not cloning.
2018          */
2019 #if IS_ENABLED(CONFIG_IPV6_SIT)
2020         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2021                 cfg.fc_flags |= RTF_NONEXTHOP;
2022 #endif
2023
2024         ip6_route_add(&cfg);
2025 }
2026
2027
2028 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2029                                                   int plen,
2030                                                   const struct net_device *dev,
2031                                                   u32 flags, u32 noflags)
2032 {
2033         struct fib6_node *fn;
2034         struct rt6_info *rt = NULL;
2035         struct fib6_table *table;
2036
2037         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2038         if (table == NULL)
2039                 return NULL;
2040
2041         read_lock_bh(&table->tb6_lock);
2042         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2043         if (!fn)
2044                 goto out;
2045         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2046                 if (rt->dst.dev->ifindex != dev->ifindex)
2047                         continue;
2048                 if ((rt->rt6i_flags & flags) != flags)
2049                         continue;
2050                 if ((rt->rt6i_flags & noflags) != 0)
2051                         continue;
2052                 dst_hold(&rt->dst);
2053                 break;
2054         }
2055 out:
2056         read_unlock_bh(&table->tb6_lock);
2057         return rt;
2058 }
2059
2060
2061 /* Create "default" multicast route to the interface */
2062
2063 static void addrconf_add_mroute(struct net_device *dev)
2064 {
2065         struct fib6_config cfg = {
2066                 .fc_table = RT6_TABLE_LOCAL,
2067                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2068                 .fc_ifindex = dev->ifindex,
2069                 .fc_dst_len = 8,
2070                 .fc_flags = RTF_UP,
2071                 .fc_nlinfo.nl_net = dev_net(dev),
2072         };
2073
2074         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2075
2076         ip6_route_add(&cfg);
2077 }
2078
2079 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2080 {
2081         struct inet6_dev *idev;
2082
2083         ASSERT_RTNL();
2084
2085         idev = ipv6_find_idev(dev);
2086         if (!idev)
2087                 return ERR_PTR(-ENOBUFS);
2088
2089         if (idev->cnf.disable_ipv6)
2090                 return ERR_PTR(-EACCES);
2091
2092         /* Add default multicast route */
2093         if (!(dev->flags & IFF_LOOPBACK))
2094                 addrconf_add_mroute(dev);
2095
2096         return idev;
2097 }
2098
2099 static void manage_tempaddrs(struct inet6_dev *idev,
2100                              struct inet6_ifaddr *ifp,
2101                              __u32 valid_lft, __u32 prefered_lft,
2102                              bool create, unsigned long now)
2103 {
2104         u32 flags;
2105         struct inet6_ifaddr *ift;
2106
2107         read_lock_bh(&idev->lock);
2108         /* update all temporary addresses in the list */
2109         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2110                 int age, max_valid, max_prefered;
2111
2112                 if (ifp != ift->ifpub)
2113                         continue;
2114
2115                 /* RFC 4941 section 3.3:
2116                  * If a received option will extend the lifetime of a public
2117                  * address, the lifetimes of temporary addresses should
2118                  * be extended, subject to the overall constraint that no
2119                  * temporary addresses should ever remain "valid" or "preferred"
2120                  * for a time longer than (TEMP_VALID_LIFETIME) or
2121                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2122                  */
2123                 age = (now - ift->cstamp) / HZ;
2124                 max_valid = idev->cnf.temp_valid_lft - age;
2125                 if (max_valid < 0)
2126                         max_valid = 0;
2127
2128                 max_prefered = idev->cnf.temp_prefered_lft -
2129                                idev->cnf.max_desync_factor - age;
2130                 if (max_prefered < 0)
2131                         max_prefered = 0;
2132
2133                 if (valid_lft > max_valid)
2134                         valid_lft = max_valid;
2135
2136                 if (prefered_lft > max_prefered)
2137                         prefered_lft = max_prefered;
2138
2139                 spin_lock(&ift->lock);
2140                 flags = ift->flags;
2141                 ift->valid_lft = valid_lft;
2142                 ift->prefered_lft = prefered_lft;
2143                 ift->tstamp = now;
2144                 if (prefered_lft > 0)
2145                         ift->flags &= ~IFA_F_DEPRECATED;
2146
2147                 spin_unlock(&ift->lock);
2148                 if (!(flags&IFA_F_TENTATIVE))
2149                         ipv6_ifa_notify(0, ift);
2150         }
2151
2152         if ((create || list_empty(&idev->tempaddr_list)) &&
2153             idev->cnf.use_tempaddr > 0) {
2154                 /* When a new public address is created as described
2155                  * in [ADDRCONF], also create a new temporary address.
2156                  * Also create a temporary address if it's enabled but
2157                  * no temporary address currently exists.
2158                  */
2159                 read_unlock_bh(&idev->lock);
2160                 ipv6_create_tempaddr(ifp, NULL);
2161         } else {
2162                 read_unlock_bh(&idev->lock);
2163         }
2164 }
2165
2166 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2167 {
2168         struct prefix_info *pinfo;
2169         __u32 valid_lft;
2170         __u32 prefered_lft;
2171         int addr_type;
2172         struct inet6_dev *in6_dev;
2173         struct net *net = dev_net(dev);
2174
2175         pinfo = (struct prefix_info *) opt;
2176
2177         if (len < sizeof(struct prefix_info)) {
2178                 ADBG("addrconf: prefix option too short\n");
2179                 return;
2180         }
2181
2182         /*
2183          *      Validation checks ([ADDRCONF], page 19)
2184          */
2185
2186         addr_type = ipv6_addr_type(&pinfo->prefix);
2187
2188         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2189                 return;
2190
2191         valid_lft = ntohl(pinfo->valid);
2192         prefered_lft = ntohl(pinfo->prefered);
2193
2194         if (prefered_lft > valid_lft) {
2195                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2196                 return;
2197         }
2198
2199         in6_dev = in6_dev_get(dev);
2200
2201         if (in6_dev == NULL) {
2202                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2203                                     dev->name);
2204                 return;
2205         }
2206
2207         /*
2208          *      Two things going on here:
2209          *      1) Add routes for on-link prefixes
2210          *      2) Configure prefixes with the auto flag set
2211          */
2212
2213         if (pinfo->onlink) {
2214                 struct rt6_info *rt;
2215                 unsigned long rt_expires;
2216
2217                 /* Avoid arithmetic overflow. Really, we could
2218                  * save rt_expires in seconds, likely valid_lft,
2219                  * but it would require division in fib gc, that it
2220                  * not good.
2221                  */
2222                 if (HZ > USER_HZ)
2223                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2224                 else
2225                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2226
2227                 if (addrconf_finite_timeout(rt_expires))
2228                         rt_expires *= HZ;
2229
2230                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2231                                                pinfo->prefix_len,
2232                                                dev,
2233                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2234                                                RTF_GATEWAY | RTF_DEFAULT);
2235
2236                 if (rt) {
2237                         /* Autoconf prefix route */
2238                         if (valid_lft == 0) {
2239                                 ip6_del_rt(rt);
2240                                 rt = NULL;
2241                         } else if (addrconf_finite_timeout(rt_expires)) {
2242                                 /* not infinity */
2243                                 rt6_set_expires(rt, jiffies + rt_expires);
2244                         } else {
2245                                 rt6_clean_expires(rt);
2246                         }
2247                 } else if (valid_lft) {
2248                         clock_t expires = 0;
2249                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2250                         if (addrconf_finite_timeout(rt_expires)) {
2251                                 /* not infinity */
2252                                 flags |= RTF_EXPIRES;
2253                                 expires = jiffies_to_clock_t(rt_expires);
2254                         }
2255                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2256                                               dev, expires, flags);
2257                 }
2258                 ip6_rt_put(rt);
2259         }
2260
2261         /* Try to figure out our local address for this prefix */
2262
2263         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2264                 struct inet6_ifaddr *ifp;
2265                 struct in6_addr addr;
2266                 int create = 0, update_lft = 0;
2267                 bool tokenized = false;
2268
2269                 if (pinfo->prefix_len == 64) {
2270                         memcpy(&addr, &pinfo->prefix, 8);
2271
2272                         if (!ipv6_addr_any(&in6_dev->token)) {
2273                                 read_lock_bh(&in6_dev->lock);
2274                                 memcpy(addr.s6_addr + 8,
2275                                        in6_dev->token.s6_addr + 8, 8);
2276                                 read_unlock_bh(&in6_dev->lock);
2277                                 tokenized = true;
2278                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2279                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2280                                 in6_dev_put(in6_dev);
2281                                 return;
2282                         }
2283                         goto ok;
2284                 }
2285                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2286                                     pinfo->prefix_len);
2287                 in6_dev_put(in6_dev);
2288                 return;
2289
2290 ok:
2291
2292                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2293
2294                 if (ifp == NULL && valid_lft) {
2295                         int max_addresses = in6_dev->cnf.max_addresses;
2296                         u32 addr_flags = 0;
2297
2298 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2299                         if (in6_dev->cnf.optimistic_dad &&
2300                             !net->ipv6.devconf_all->forwarding && sllao)
2301                                 addr_flags = IFA_F_OPTIMISTIC;
2302 #endif
2303
2304                         /* Do not allow to create too much of autoconfigured
2305                          * addresses; this would be too easy way to crash kernel.
2306                          */
2307                         if (!max_addresses ||
2308                             ipv6_count_addresses(in6_dev) < max_addresses)
2309                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2310                                                     pinfo->prefix_len,
2311                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2312                                                     addr_flags, valid_lft,
2313                                                     prefered_lft);
2314
2315                         if (IS_ERR_OR_NULL(ifp)) {
2316                                 in6_dev_put(in6_dev);
2317                                 return;
2318                         }
2319
2320                         update_lft = 0;
2321                         create = 1;
2322                         spin_lock_bh(&ifp->lock);
2323                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2324                         ifp->cstamp = jiffies;
2325                         ifp->tokenized = tokenized;
2326                         spin_unlock_bh(&ifp->lock);
2327                         addrconf_dad_start(ifp);
2328                 }
2329
2330                 if (ifp) {
2331                         u32 flags;
2332                         unsigned long now;
2333                         u32 stored_lft;
2334
2335                         /* update lifetime (RFC2462 5.5.3 e) */
2336                         spin_lock(&ifp->lock);
2337                         now = jiffies;
2338                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2339                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2340                         else
2341                                 stored_lft = 0;
2342                         if (!update_lft && !create && stored_lft) {
2343                                 const u32 minimum_lft = min_t(u32,
2344                                         stored_lft, MIN_VALID_LIFETIME);
2345                                 valid_lft = max(valid_lft, minimum_lft);
2346
2347                                 /* RFC4862 Section 5.5.3e:
2348                                  * "Note that the preferred lifetime of the
2349                                  *  corresponding address is always reset to
2350                                  *  the Preferred Lifetime in the received
2351                                  *  Prefix Information option, regardless of
2352                                  *  whether the valid lifetime is also reset or
2353                                  *  ignored."
2354                                  *
2355                                  * So we should always update prefered_lft here.
2356                                  */
2357                                 update_lft = 1;
2358                         }
2359
2360                         if (update_lft) {
2361                                 ifp->valid_lft = valid_lft;
2362                                 ifp->prefered_lft = prefered_lft;
2363                                 ifp->tstamp = now;
2364                                 flags = ifp->flags;
2365                                 ifp->flags &= ~IFA_F_DEPRECATED;
2366                                 spin_unlock(&ifp->lock);
2367
2368                                 if (!(flags&IFA_F_TENTATIVE))
2369                                         ipv6_ifa_notify(0, ifp);
2370                         } else
2371                                 spin_unlock(&ifp->lock);
2372
2373                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2374                                          create, now);
2375
2376                         in6_ifa_put(ifp);
2377                         addrconf_verify();
2378                 }
2379         }
2380         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2381         in6_dev_put(in6_dev);
2382 }
2383
2384 /*
2385  *      Set destination address.
2386  *      Special case for SIT interfaces where we create a new "virtual"
2387  *      device.
2388  */
2389 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2390 {
2391         struct in6_ifreq ireq;
2392         struct net_device *dev;
2393         int err = -EINVAL;
2394
2395         rtnl_lock();
2396
2397         err = -EFAULT;
2398         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2399                 goto err_exit;
2400
2401         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2402
2403         err = -ENODEV;
2404         if (dev == NULL)
2405                 goto err_exit;
2406
2407 #if IS_ENABLED(CONFIG_IPV6_SIT)
2408         if (dev->type == ARPHRD_SIT) {
2409                 const struct net_device_ops *ops = dev->netdev_ops;
2410                 struct ifreq ifr;
2411                 struct ip_tunnel_parm p;
2412
2413                 err = -EADDRNOTAVAIL;
2414                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2415                         goto err_exit;
2416
2417                 memset(&p, 0, sizeof(p));
2418                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2419                 p.iph.saddr = 0;
2420                 p.iph.version = 4;
2421                 p.iph.ihl = 5;
2422                 p.iph.protocol = IPPROTO_IPV6;
2423                 p.iph.ttl = 64;
2424                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2425
2426                 if (ops->ndo_do_ioctl) {
2427                         mm_segment_t oldfs = get_fs();
2428
2429                         set_fs(KERNEL_DS);
2430                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2431                         set_fs(oldfs);
2432                 } else
2433                         err = -EOPNOTSUPP;
2434
2435                 if (err == 0) {
2436                         err = -ENOBUFS;
2437                         dev = __dev_get_by_name(net, p.name);
2438                         if (!dev)
2439                                 goto err_exit;
2440                         err = dev_open(dev);
2441                 }
2442         }
2443 #endif
2444
2445 err_exit:
2446         rtnl_unlock();
2447         return err;
2448 }
2449
2450 /*
2451  *      Manual configuration of address on an interface
2452  */
2453 static int inet6_addr_add(struct net *net, int ifindex,
2454                           const struct in6_addr *pfx,
2455                           const struct in6_addr *peer_pfx,
2456                           unsigned int plen, __u32 ifa_flags,
2457                           __u32 prefered_lft, __u32 valid_lft)
2458 {
2459         struct inet6_ifaddr *ifp;
2460         struct inet6_dev *idev;
2461         struct net_device *dev;
2462         int scope;
2463         u32 flags;
2464         clock_t expires;
2465         unsigned long timeout;
2466
2467         ASSERT_RTNL();
2468
2469         if (plen > 128)
2470                 return -EINVAL;
2471
2472         /* check the lifetime */
2473         if (!valid_lft || prefered_lft > valid_lft)
2474                 return -EINVAL;
2475
2476         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2477                 return -EINVAL;
2478
2479         dev = __dev_get_by_index(net, ifindex);
2480         if (!dev)
2481                 return -ENODEV;
2482
2483         idev = addrconf_add_dev(dev);
2484         if (IS_ERR(idev))
2485                 return PTR_ERR(idev);
2486
2487         scope = ipv6_addr_scope(pfx);
2488
2489         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2490         if (addrconf_finite_timeout(timeout)) {
2491                 expires = jiffies_to_clock_t(timeout * HZ);
2492                 valid_lft = timeout;
2493                 flags = RTF_EXPIRES;
2494         } else {
2495                 expires = 0;
2496                 flags = 0;
2497                 ifa_flags |= IFA_F_PERMANENT;
2498         }
2499
2500         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2501         if (addrconf_finite_timeout(timeout)) {
2502                 if (timeout == 0)
2503                         ifa_flags |= IFA_F_DEPRECATED;
2504                 prefered_lft = timeout;
2505         }
2506
2507         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2508                             valid_lft, prefered_lft);
2509
2510         if (!IS_ERR(ifp)) {
2511                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2512                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2513                                               expires, flags);
2514                 }
2515
2516                 /*
2517                  * Note that section 3.1 of RFC 4429 indicates
2518                  * that the Optimistic flag should not be set for
2519                  * manually configured addresses
2520                  */
2521                 addrconf_dad_start(ifp);
2522                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2523                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2524                                          true, jiffies);
2525                 in6_ifa_put(ifp);
2526                 addrconf_verify_rtnl();
2527                 return 0;
2528         }
2529
2530         return PTR_ERR(ifp);
2531 }
2532
2533 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2534                           const struct in6_addr *pfx, unsigned int plen)
2535 {
2536         struct inet6_ifaddr *ifp;
2537         struct inet6_dev *idev;
2538         struct net_device *dev;
2539
2540         if (plen > 128)
2541                 return -EINVAL;
2542
2543         dev = __dev_get_by_index(net, ifindex);
2544         if (!dev)
2545                 return -ENODEV;
2546
2547         idev = __in6_dev_get(dev);
2548         if (idev == NULL)
2549                 return -ENXIO;
2550
2551         read_lock_bh(&idev->lock);
2552         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2553                 if (ifp->prefix_len == plen &&
2554                     ipv6_addr_equal(pfx, &ifp->addr)) {
2555                         in6_ifa_hold(ifp);
2556                         read_unlock_bh(&idev->lock);
2557
2558                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2559                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2560                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2561                                                  jiffies);
2562                         ipv6_del_addr(ifp);
2563                         addrconf_verify_rtnl();
2564                         return 0;
2565                 }
2566         }
2567         read_unlock_bh(&idev->lock);
2568         return -EADDRNOTAVAIL;
2569 }
2570
2571
2572 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2573 {
2574         struct in6_ifreq ireq;
2575         int err;
2576
2577         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2578                 return -EPERM;
2579
2580         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2581                 return -EFAULT;
2582
2583         rtnl_lock();
2584         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2585                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2586                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2587         rtnl_unlock();
2588         return err;
2589 }
2590
2591 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2592 {
2593         struct in6_ifreq ireq;
2594         int err;
2595
2596         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2597                 return -EPERM;
2598
2599         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2600                 return -EFAULT;
2601
2602         rtnl_lock();
2603         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2604                              ireq.ifr6_prefixlen);
2605         rtnl_unlock();
2606         return err;
2607 }
2608
2609 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2610                      int plen, int scope)
2611 {
2612         struct inet6_ifaddr *ifp;
2613
2614         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2615                             scope, IFA_F_PERMANENT,
2616                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2617         if (!IS_ERR(ifp)) {
2618                 spin_lock_bh(&ifp->lock);
2619                 ifp->flags &= ~IFA_F_TENTATIVE;
2620                 spin_unlock_bh(&ifp->lock);
2621                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2622                 in6_ifa_put(ifp);
2623         }
2624 }
2625
2626 #if IS_ENABLED(CONFIG_IPV6_SIT)
2627 static void sit_add_v4_addrs(struct inet6_dev *idev)
2628 {
2629         struct in6_addr addr;
2630         struct net_device *dev;
2631         struct net *net = dev_net(idev->dev);
2632         int scope, plen;
2633         u32 pflags = 0;
2634
2635         ASSERT_RTNL();
2636
2637         memset(&addr, 0, sizeof(struct in6_addr));
2638         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2639
2640         if (idev->dev->flags&IFF_POINTOPOINT) {
2641                 addr.s6_addr32[0] = htonl(0xfe800000);
2642                 scope = IFA_LINK;
2643                 plen = 64;
2644         } else {
2645                 scope = IPV6_ADDR_COMPATv4;
2646                 plen = 96;
2647                 pflags |= RTF_NONEXTHOP;
2648         }
2649
2650         if (addr.s6_addr32[3]) {
2651                 add_addr(idev, &addr, plen, scope);
2652                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2653                 return;
2654         }
2655
2656         for_each_netdev(net, dev) {
2657                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2658                 if (in_dev && (dev->flags & IFF_UP)) {
2659                         struct in_ifaddr *ifa;
2660
2661                         int flag = scope;
2662
2663                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2664
2665                                 addr.s6_addr32[3] = ifa->ifa_local;
2666
2667                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2668                                         continue;
2669                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2670                                         if (idev->dev->flags&IFF_POINTOPOINT)
2671                                                 continue;
2672                                         flag |= IFA_HOST;
2673                                 }
2674
2675                                 add_addr(idev, &addr, plen, flag);
2676                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2677                                                       pflags);
2678                         }
2679                 }
2680         }
2681 }
2682 #endif
2683
2684 static void init_loopback(struct net_device *dev)
2685 {
2686         struct inet6_dev  *idev;
2687         struct net_device *sp_dev;
2688         struct inet6_ifaddr *sp_ifa;
2689         struct rt6_info *sp_rt;
2690
2691         /* ::1 */
2692
2693         ASSERT_RTNL();
2694
2695         idev = ipv6_find_idev(dev);
2696         if (idev == NULL) {
2697                 pr_debug("%s: add_dev failed\n", __func__);
2698                 return;
2699         }
2700
2701         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2702
2703         /* Add routes to other interface's IPv6 addresses */
2704         for_each_netdev(dev_net(dev), sp_dev) {
2705                 if (!strcmp(sp_dev->name, dev->name))
2706                         continue;
2707
2708                 idev = __in6_dev_get(sp_dev);
2709                 if (!idev)
2710                         continue;
2711
2712                 read_lock_bh(&idev->lock);
2713                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2714
2715                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2716                                 continue;
2717
2718                         if (sp_ifa->rt) {
2719                                 /* This dst has been added to garbage list when
2720                                  * lo device down, release this obsolete dst and
2721                                  * reallocate a new router for ifa.
2722                                  */
2723                                 if (sp_ifa->rt->dst.obsolete > 0) {
2724                                         ip6_rt_put(sp_ifa->rt);
2725                                         sp_ifa->rt = NULL;
2726                                 } else {
2727                                         continue;
2728                                 }
2729                         }
2730
2731                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2732
2733                         /* Failure cases are ignored */
2734                         if (!IS_ERR(sp_rt)) {
2735                                 sp_ifa->rt = sp_rt;
2736                                 ip6_ins_rt(sp_rt);
2737                         }
2738                 }
2739                 read_unlock_bh(&idev->lock);
2740         }
2741 }
2742
2743 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2744 {
2745         struct inet6_ifaddr *ifp;
2746         u32 addr_flags = IFA_F_PERMANENT;
2747
2748 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2749         if (idev->cnf.optimistic_dad &&
2750             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2751                 addr_flags |= IFA_F_OPTIMISTIC;
2752 #endif
2753
2754
2755         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2756                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2757         if (!IS_ERR(ifp)) {
2758                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2759                 addrconf_dad_start(ifp);
2760                 in6_ifa_put(ifp);
2761         }
2762 }
2763
2764 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2765 {
2766         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2767                 struct in6_addr addr;
2768
2769                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2770                 /* addrconf_add_linklocal also adds a prefix_route and we
2771                  * only need to care about prefix routes if ipv6_generate_eui64
2772                  * couldn't generate one.
2773                  */
2774                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2775                         addrconf_add_linklocal(idev, &addr);
2776                 else if (prefix_route)
2777                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2778         }
2779 }
2780
2781 static void addrconf_dev_config(struct net_device *dev)
2782 {
2783         struct inet6_dev *idev;
2784
2785         ASSERT_RTNL();
2786
2787         if ((dev->type != ARPHRD_ETHER) &&
2788             (dev->type != ARPHRD_FDDI) &&
2789             (dev->type != ARPHRD_ARCNET) &&
2790             (dev->type != ARPHRD_INFINIBAND) &&
2791             (dev->type != ARPHRD_IEEE802154) &&
2792             (dev->type != ARPHRD_IEEE1394) &&
2793             (dev->type != ARPHRD_TUNNEL6) &&
2794             (dev->type != ARPHRD_6LOWPAN)) {
2795                 /* Alas, we support only Ethernet autoconfiguration. */
2796                 return;
2797         }
2798
2799         idev = addrconf_add_dev(dev);
2800         if (IS_ERR(idev))
2801                 return;
2802
2803         addrconf_addr_gen(idev, false);
2804 }
2805
2806 #if IS_ENABLED(CONFIG_IPV6_SIT)
2807 static void addrconf_sit_config(struct net_device *dev)
2808 {
2809         struct inet6_dev *idev;
2810
2811         ASSERT_RTNL();
2812
2813         /*
2814          * Configure the tunnel with one of our IPv4
2815          * addresses... we should configure all of
2816          * our v4 addrs in the tunnel
2817          */
2818
2819         idev = ipv6_find_idev(dev);
2820         if (idev == NULL) {
2821                 pr_debug("%s: add_dev failed\n", __func__);
2822                 return;
2823         }
2824
2825         if (dev->priv_flags & IFF_ISATAP) {
2826                 addrconf_addr_gen(idev, false);
2827                 return;
2828         }
2829
2830         sit_add_v4_addrs(idev);
2831
2832         if (dev->flags&IFF_POINTOPOINT)
2833                 addrconf_add_mroute(dev);
2834 }
2835 #endif
2836
2837 #if IS_ENABLED(CONFIG_NET_IPGRE)
2838 static void addrconf_gre_config(struct net_device *dev)
2839 {
2840         struct inet6_dev *idev;
2841
2842         ASSERT_RTNL();
2843
2844         idev = ipv6_find_idev(dev);
2845         if (idev == NULL) {
2846                 pr_debug("%s: add_dev failed\n", __func__);
2847                 return;
2848         }
2849
2850         addrconf_addr_gen(idev, true);
2851 }
2852 #endif
2853
2854 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2855                            void *ptr)
2856 {
2857         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2858         struct inet6_dev *idev = __in6_dev_get(dev);
2859         int run_pending = 0;
2860         int err;
2861
2862         switch (event) {
2863         case NETDEV_REGISTER:
2864                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2865                         idev = ipv6_add_dev(dev);
2866                         if (IS_ERR(idev))
2867                                 return notifier_from_errno(PTR_ERR(idev));
2868                 }
2869                 break;
2870
2871         case NETDEV_UP:
2872         case NETDEV_CHANGE:
2873                 if (dev->flags & IFF_SLAVE)
2874                         break;
2875
2876                 if (idev && idev->cnf.disable_ipv6)
2877                         break;
2878
2879                 if (event == NETDEV_UP) {
2880                         if (!addrconf_qdisc_ok(dev)) {
2881                                 /* device is not ready yet. */
2882                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2883                                         dev->name);
2884                                 break;
2885                         }
2886
2887                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2888                                 idev = ipv6_add_dev(dev);
2889
2890                         if (!IS_ERR_OR_NULL(idev)) {
2891                                 idev->if_flags |= IF_READY;
2892                                 run_pending = 1;
2893                         }
2894                 } else {
2895                         if (!addrconf_qdisc_ok(dev)) {
2896                                 /* device is still not ready. */
2897                                 break;
2898                         }
2899
2900                         if (idev) {
2901                                 if (idev->if_flags & IF_READY)
2902                                         /* device is already configured. */
2903                                         break;
2904                                 idev->if_flags |= IF_READY;
2905                         }
2906
2907                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2908                                 dev->name);
2909
2910                         run_pending = 1;
2911                 }
2912
2913                 switch (dev->type) {
2914 #if IS_ENABLED(CONFIG_IPV6_SIT)
2915                 case ARPHRD_SIT:
2916                         addrconf_sit_config(dev);
2917                         break;
2918 #endif
2919 #if IS_ENABLED(CONFIG_NET_IPGRE)
2920                 case ARPHRD_IPGRE:
2921                         addrconf_gre_config(dev);
2922                         break;
2923 #endif
2924                 case ARPHRD_LOOPBACK:
2925                         init_loopback(dev);
2926                         break;
2927
2928                 default:
2929                         addrconf_dev_config(dev);
2930                         break;
2931                 }
2932
2933                 if (!IS_ERR_OR_NULL(idev)) {
2934                         if (run_pending)
2935                                 addrconf_dad_run(idev);
2936
2937                         /*
2938                          * If the MTU changed during the interface down,
2939                          * when the interface up, the changed MTU must be
2940                          * reflected in the idev as well as routers.
2941                          */
2942                         if (idev->cnf.mtu6 != dev->mtu &&
2943                             dev->mtu >= IPV6_MIN_MTU) {
2944                                 rt6_mtu_change(dev, dev->mtu);
2945                                 idev->cnf.mtu6 = dev->mtu;
2946                         }
2947                         idev->tstamp = jiffies;
2948                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2949
2950                         /*
2951                          * If the changed mtu during down is lower than
2952                          * IPV6_MIN_MTU stop IPv6 on this interface.
2953                          */
2954                         if (dev->mtu < IPV6_MIN_MTU)
2955                                 addrconf_ifdown(dev, 1);
2956                 }
2957                 break;
2958
2959         case NETDEV_CHANGEMTU:
2960                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2961                         rt6_mtu_change(dev, dev->mtu);
2962                         idev->cnf.mtu6 = dev->mtu;
2963                         break;
2964                 }
2965
2966                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2967                         idev = ipv6_add_dev(dev);
2968                         if (!IS_ERR(idev))
2969                                 break;
2970                 }
2971
2972                 /*
2973                  * if MTU under IPV6_MIN_MTU.
2974                  * Stop IPv6 on this interface.
2975                  */
2976
2977         case NETDEV_DOWN:
2978         case NETDEV_UNREGISTER:
2979                 /*
2980                  *      Remove all addresses from this interface.
2981                  */
2982                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2983                 break;
2984
2985         case NETDEV_CHANGENAME:
2986                 if (idev) {
2987                         snmp6_unregister_dev(idev);
2988                         addrconf_sysctl_unregister(idev);
2989                         err = addrconf_sysctl_register(idev);
2990                         if (err)
2991                                 return notifier_from_errno(err);
2992                         err = snmp6_register_dev(idev);
2993                         if (err) {
2994                                 addrconf_sysctl_unregister(idev);
2995                                 return notifier_from_errno(err);
2996                         }
2997                 }
2998                 break;
2999
3000         case NETDEV_PRE_TYPE_CHANGE:
3001         case NETDEV_POST_TYPE_CHANGE:
3002                 addrconf_type_change(dev, event);
3003                 break;
3004         }
3005
3006         return NOTIFY_OK;
3007 }
3008
3009 /*
3010  *      addrconf module should be notified of a device going up
3011  */
3012 static struct notifier_block ipv6_dev_notf = {
3013         .notifier_call = addrconf_notify,
3014 };
3015
3016 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3017 {
3018         struct inet6_dev *idev;
3019         ASSERT_RTNL();
3020
3021         idev = __in6_dev_get(dev);
3022
3023         if (event == NETDEV_POST_TYPE_CHANGE)
3024                 ipv6_mc_remap(idev);
3025         else if (event == NETDEV_PRE_TYPE_CHANGE)
3026                 ipv6_mc_unmap(idev);
3027 }
3028
3029 static int addrconf_ifdown(struct net_device *dev, int how)
3030 {
3031         struct net *net = dev_net(dev);
3032         struct inet6_dev *idev;
3033         struct inet6_ifaddr *ifa;
3034         int state, i;
3035
3036         ASSERT_RTNL();
3037
3038         rt6_ifdown(net, dev);
3039         neigh_ifdown(&nd_tbl, dev);
3040
3041         idev = __in6_dev_get(dev);
3042         if (idev == NULL)
3043                 return -ENODEV;
3044
3045         /*
3046          * Step 1: remove reference to ipv6 device from parent device.
3047          *         Do not dev_put!
3048          */
3049         if (how) {
3050                 idev->dead = 1;
3051
3052                 /* protected by rtnl_lock */
3053                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3054
3055                 /* Step 1.5: remove snmp6 entry */
3056                 snmp6_unregister_dev(idev);
3057
3058         }
3059
3060         /* Step 2: clear hash table */
3061         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3062                 struct hlist_head *h = &inet6_addr_lst[i];
3063
3064                 spin_lock_bh(&addrconf_hash_lock);
3065 restart:
3066                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3067                         if (ifa->idev == idev) {
3068                                 hlist_del_init_rcu(&ifa->addr_lst);
3069                                 addrconf_del_dad_work(ifa);
3070                                 goto restart;
3071                         }
3072                 }
3073                 spin_unlock_bh(&addrconf_hash_lock);
3074         }
3075
3076         write_lock_bh(&idev->lock);
3077
3078         addrconf_del_rs_timer(idev);
3079
3080         /* Step 2: clear flags for stateless addrconf */
3081         if (!how)
3082                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3083
3084         if (how && del_timer(&idev->regen_timer))
3085                 in6_dev_put(idev);
3086
3087         /* Step 3: clear tempaddr list */
3088         while (!list_empty(&idev->tempaddr_list)) {
3089                 ifa = list_first_entry(&idev->tempaddr_list,
3090                                        struct inet6_ifaddr, tmp_list);
3091                 list_del(&ifa->tmp_list);
3092                 write_unlock_bh(&idev->lock);
3093                 spin_lock_bh(&ifa->lock);
3094
3095                 if (ifa->ifpub) {
3096                         in6_ifa_put(ifa->ifpub);
3097                         ifa->ifpub = NULL;
3098                 }
3099                 spin_unlock_bh(&ifa->lock);
3100                 in6_ifa_put(ifa);
3101                 write_lock_bh(&idev->lock);
3102         }
3103
3104         while (!list_empty(&idev->addr_list)) {
3105                 ifa = list_first_entry(&idev->addr_list,
3106                                        struct inet6_ifaddr, if_list);
3107                 addrconf_del_dad_work(ifa);
3108
3109                 list_del(&ifa->if_list);
3110
3111                 write_unlock_bh(&idev->lock);
3112
3113                 spin_lock_bh(&ifa->state_lock);
3114                 state = ifa->state;
3115                 ifa->state = INET6_IFADDR_STATE_DEAD;
3116                 spin_unlock_bh(&ifa->state_lock);
3117
3118                 if (state != INET6_IFADDR_STATE_DEAD) {
3119                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3120                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3121                 }
3122                 in6_ifa_put(ifa);
3123
3124                 write_lock_bh(&idev->lock);
3125         }
3126
3127         write_unlock_bh(&idev->lock);
3128
3129         /* Step 5: Discard anycast and multicast list */
3130         if (how) {
3131                 ipv6_ac_destroy_dev(idev);
3132                 ipv6_mc_destroy_dev(idev);
3133         } else {
3134                 ipv6_mc_down(idev);
3135         }
3136
3137         idev->tstamp = jiffies;
3138
3139         /* Last: Shot the device (if unregistered) */
3140         if (how) {
3141                 addrconf_sysctl_unregister(idev);
3142                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3143                 neigh_ifdown(&nd_tbl, dev);
3144                 in6_dev_put(idev);
3145         }
3146         return 0;
3147 }
3148
3149 static void addrconf_rs_timer(unsigned long data)
3150 {
3151         struct inet6_dev *idev = (struct inet6_dev *)data;
3152         struct net_device *dev = idev->dev;
3153         struct in6_addr lladdr;
3154
3155         write_lock(&idev->lock);
3156         if (idev->dead || !(idev->if_flags & IF_READY))
3157                 goto out;
3158
3159         if (!ipv6_accept_ra(idev))
3160                 goto out;
3161
3162         /* Announcement received after solicitation was sent */
3163         if (idev->if_flags & IF_RA_RCVD)
3164                 goto out;
3165
3166         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3167                 write_unlock(&idev->lock);
3168                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3169                         ndisc_send_rs(dev, &lladdr,
3170                                       &in6addr_linklocal_allrouters);
3171                 else
3172                         goto put;
3173
3174                 write_lock(&idev->lock);
3175                 /* The wait after the last probe can be shorter */
3176                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3177                                              idev->cnf.rtr_solicits) ?
3178                                       idev->cnf.rtr_solicit_delay :
3179                                       idev->cnf.rtr_solicit_interval);
3180         } else {
3181                 /*
3182                  * Note: we do not support deprecated "all on-link"
3183                  * assumption any longer.
3184                  */
3185                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3186         }
3187
3188 out:
3189         write_unlock(&idev->lock);
3190 put:
3191         in6_dev_put(idev);
3192 }
3193
3194 /*
3195  *      Duplicate Address Detection
3196  */
3197 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3198 {
3199         unsigned long rand_num;
3200         struct inet6_dev *idev = ifp->idev;
3201
3202         if (ifp->flags & IFA_F_OPTIMISTIC)
3203                 rand_num = 0;
3204         else
3205                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3206
3207         ifp->dad_probes = idev->cnf.dad_transmits;
3208         addrconf_mod_dad_work(ifp, rand_num);
3209 }
3210
3211 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3212 {
3213         struct inet6_dev *idev = ifp->idev;
3214         struct net_device *dev = idev->dev;
3215
3216         addrconf_join_solict(dev, &ifp->addr);
3217
3218         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3219
3220         read_lock_bh(&idev->lock);
3221         spin_lock(&ifp->lock);
3222         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3223                 goto out;
3224
3225         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3226             idev->cnf.accept_dad < 1 ||
3227             !(ifp->flags&IFA_F_TENTATIVE) ||
3228             ifp->flags & IFA_F_NODAD) {
3229                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3230                 spin_unlock(&ifp->lock);
3231                 read_unlock_bh(&idev->lock);
3232
3233                 addrconf_dad_completed(ifp);
3234                 return;
3235         }
3236
3237         if (!(idev->if_flags & IF_READY)) {
3238                 spin_unlock(&ifp->lock);
3239                 read_unlock_bh(&idev->lock);
3240                 /*
3241                  * If the device is not ready:
3242                  * - keep it tentative if it is a permanent address.
3243                  * - otherwise, kill it.
3244                  */
3245                 in6_ifa_hold(ifp);
3246                 addrconf_dad_stop(ifp, 0);
3247                 return;
3248         }
3249
3250         /*
3251          * Optimistic nodes can start receiving
3252          * Frames right away
3253          */
3254         if (ifp->flags & IFA_F_OPTIMISTIC) {
3255                 ip6_ins_rt(ifp->rt);
3256                 if (ipv6_use_optimistic_addr(idev)) {
3257                         /* Because optimistic nodes can use this address,
3258                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3259                          */
3260                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3261                 }
3262         }
3263
3264         addrconf_dad_kick(ifp);
3265 out:
3266         spin_unlock(&ifp->lock);
3267         read_unlock_bh(&idev->lock);
3268 }
3269
3270 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3271 {
3272         bool begin_dad = false;
3273
3274         spin_lock_bh(&ifp->state_lock);
3275         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3276                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3277                 begin_dad = true;
3278         }
3279         spin_unlock_bh(&ifp->state_lock);
3280
3281         if (begin_dad)
3282                 addrconf_mod_dad_work(ifp, 0);
3283 }
3284
3285 static void addrconf_dad_work(struct work_struct *w)
3286 {
3287         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3288                                                 struct inet6_ifaddr,
3289                                                 dad_work);
3290         struct inet6_dev *idev = ifp->idev;
3291         struct in6_addr mcaddr;
3292
3293         enum {
3294                 DAD_PROCESS,
3295                 DAD_BEGIN,
3296                 DAD_ABORT,
3297         } action = DAD_PROCESS;
3298
3299         rtnl_lock();
3300
3301         spin_lock_bh(&ifp->state_lock);
3302         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3303                 action = DAD_BEGIN;
3304                 ifp->state = INET6_IFADDR_STATE_DAD;
3305         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3306                 action = DAD_ABORT;
3307                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3308         }
3309         spin_unlock_bh(&ifp->state_lock);
3310
3311         if (action == DAD_BEGIN) {
3312                 addrconf_dad_begin(ifp);
3313                 goto out;
3314         } else if (action == DAD_ABORT) {
3315                 addrconf_dad_stop(ifp, 1);
3316                 goto out;
3317         }
3318
3319         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3320                 goto out;
3321
3322         write_lock_bh(&idev->lock);
3323         if (idev->dead || !(idev->if_flags & IF_READY)) {
3324                 write_unlock_bh(&idev->lock);
3325                 goto out;
3326         }
3327
3328         spin_lock(&ifp->lock);
3329         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3330                 spin_unlock(&ifp->lock);
3331                 write_unlock_bh(&idev->lock);
3332                 goto out;
3333         }
3334
3335         if (ifp->dad_probes == 0) {
3336                 /*
3337                  * DAD was successful
3338                  */
3339
3340                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3341                 spin_unlock(&ifp->lock);
3342                 write_unlock_bh(&idev->lock);
3343
3344                 addrconf_dad_completed(ifp);
3345
3346                 goto out;
3347         }
3348
3349         ifp->dad_probes--;
3350         addrconf_mod_dad_work(ifp,
3351                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3352         spin_unlock(&ifp->lock);
3353         write_unlock_bh(&idev->lock);
3354
3355         /* send a neighbour solicitation for our addr */
3356         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3357         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3358 out:
3359         in6_ifa_put(ifp);
3360         rtnl_unlock();
3361 }
3362
3363 /* ifp->idev must be at least read locked */
3364 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3365 {
3366         struct inet6_ifaddr *ifpiter;
3367         struct inet6_dev *idev = ifp->idev;
3368
3369         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3370                 if (ifpiter->scope > IFA_LINK)
3371                         break;
3372                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3373                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3374                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3375                     IFA_F_PERMANENT)
3376                         return false;
3377         }
3378         return true;
3379 }
3380
3381 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3382 {
3383         struct net_device *dev = ifp->idev->dev;
3384         struct in6_addr lladdr;
3385         bool send_rs, send_mld;
3386
3387         addrconf_del_dad_work(ifp);
3388
3389         /*
3390          *      Configure the address for reception. Now it is valid.
3391          */
3392
3393         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3394
3395         /* If added prefix is link local and we are prepared to process
3396            router advertisements, start sending router solicitations.
3397          */
3398
3399         read_lock_bh(&ifp->idev->lock);
3400         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3401         send_rs = send_mld &&
3402                   ipv6_accept_ra(ifp->idev) &&
3403                   ifp->idev->cnf.rtr_solicits > 0 &&
3404                   (dev->flags&IFF_LOOPBACK) == 0;
3405         read_unlock_bh(&ifp->idev->lock);
3406
3407         /* While dad is in progress mld report's source address is in6_addrany.
3408          * Resend with proper ll now.
3409          */
3410         if (send_mld)
3411                 ipv6_mc_dad_complete(ifp->idev);
3412
3413         if (send_rs) {
3414                 /*
3415                  *      If a host as already performed a random delay
3416                  *      [...] as part of DAD [...] there is no need
3417                  *      to delay again before sending the first RS
3418                  */
3419                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3420                         return;
3421                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3422
3423                 write_lock_bh(&ifp->idev->lock);
3424                 spin_lock(&ifp->lock);
3425                 ifp->idev->rs_probes = 1;
3426                 ifp->idev->if_flags |= IF_RS_SENT;
3427                 addrconf_mod_rs_timer(ifp->idev,
3428                                       ifp->idev->cnf.rtr_solicit_interval);
3429                 spin_unlock(&ifp->lock);
3430                 write_unlock_bh(&ifp->idev->lock);
3431         }
3432 }
3433
3434 static void addrconf_dad_run(struct inet6_dev *idev)
3435 {
3436         struct inet6_ifaddr *ifp;
3437
3438         read_lock_bh(&idev->lock);
3439         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3440                 spin_lock(&ifp->lock);
3441                 if (ifp->flags & IFA_F_TENTATIVE &&
3442                     ifp->state == INET6_IFADDR_STATE_DAD)
3443                         addrconf_dad_kick(ifp);
3444                 spin_unlock(&ifp->lock);
3445         }
3446         read_unlock_bh(&idev->lock);
3447 }
3448
3449 #ifdef CONFIG_PROC_FS
3450 struct if6_iter_state {
3451         struct seq_net_private p;
3452         int bucket;
3453         int offset;
3454 };
3455
3456 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3457 {
3458         struct inet6_ifaddr *ifa = NULL;
3459         struct if6_iter_state *state = seq->private;
3460         struct net *net = seq_file_net(seq);
3461         int p = 0;
3462
3463         /* initial bucket if pos is 0 */
3464         if (pos == 0) {
3465                 state->bucket = 0;
3466                 state->offset = 0;
3467         }
3468
3469         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3470                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3471                                          addr_lst) {
3472                         if (!net_eq(dev_net(ifa->idev->dev), net))
3473                                 continue;
3474                         /* sync with offset */
3475                         if (p < state->offset) {
3476                                 p++;
3477                                 continue;
3478                         }
3479                         state->offset++;
3480                         return ifa;
3481                 }
3482
3483                 /* prepare for next bucket */
3484                 state->offset = 0;
3485                 p = 0;
3486         }
3487         return NULL;
3488 }
3489
3490 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3491                                          struct inet6_ifaddr *ifa)
3492 {
3493         struct if6_iter_state *state = seq->private;
3494         struct net *net = seq_file_net(seq);
3495
3496         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3497                 if (!net_eq(dev_net(ifa->idev->dev), net))
3498                         continue;
3499                 state->offset++;
3500                 return ifa;
3501         }
3502
3503         while (++state->bucket < IN6_ADDR_HSIZE) {
3504                 state->offset = 0;
3505                 hlist_for_each_entry_rcu_bh(ifa,
3506                                      &inet6_addr_lst[state->bucket], addr_lst) {
3507                         if (!net_eq(dev_net(ifa->idev->dev), net))
3508                                 continue;
3509                         state->offset++;
3510                         return ifa;
3511                 }
3512         }
3513
3514         return NULL;
3515 }
3516
3517 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3518         __acquires(rcu_bh)
3519 {
3520         rcu_read_lock_bh();
3521         return if6_get_first(seq, *pos);
3522 }
3523
3524 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3525 {
3526         struct inet6_ifaddr *ifa;
3527
3528         ifa = if6_get_next(seq, v);
3529         ++*pos;
3530         return ifa;
3531 }
3532
3533 static void if6_seq_stop(struct seq_file *seq, void *v)
3534         __releases(rcu_bh)
3535 {
3536         rcu_read_unlock_bh();
3537 }
3538
3539 static int if6_seq_show(struct seq_file *seq, void *v)
3540 {
3541         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3542         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3543                    &ifp->addr,
3544                    ifp->idev->dev->ifindex,
3545                    ifp->prefix_len,
3546                    ifp->scope,
3547                    (u8) ifp->flags,
3548                    ifp->idev->dev->name);
3549         return 0;
3550 }
3551
3552 static const struct seq_operations if6_seq_ops = {
3553         .start  = if6_seq_start,
3554         .next   = if6_seq_next,
3555         .show   = if6_seq_show,
3556         .stop   = if6_seq_stop,
3557 };
3558
3559 static int if6_seq_open(struct inode *inode, struct file *file)
3560 {
3561         return seq_open_net(inode, file, &if6_seq_ops,
3562                             sizeof(struct if6_iter_state));
3563 }
3564
3565 static const struct file_operations if6_fops = {
3566         .owner          = THIS_MODULE,
3567         .open           = if6_seq_open,
3568         .read           = seq_read,
3569         .llseek         = seq_lseek,
3570         .release        = seq_release_net,
3571 };
3572
3573 static int __net_init if6_proc_net_init(struct net *net)
3574 {
3575         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3576                 return -ENOMEM;
3577         return 0;
3578 }
3579
3580 static void __net_exit if6_proc_net_exit(struct net *net)
3581 {
3582         remove_proc_entry("if_inet6", net->proc_net);
3583 }
3584
3585 static struct pernet_operations if6_proc_net_ops = {
3586         .init = if6_proc_net_init,
3587         .exit = if6_proc_net_exit,
3588 };
3589
3590 int __init if6_proc_init(void)
3591 {
3592         return register_pernet_subsys(&if6_proc_net_ops);
3593 }
3594
3595 void if6_proc_exit(void)
3596 {
3597         unregister_pernet_subsys(&if6_proc_net_ops);
3598 }
3599 #endif  /* CONFIG_PROC_FS */
3600
3601 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3602 /* Check if address is a home address configured on any interface. */
3603 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3604 {
3605         int ret = 0;
3606         struct inet6_ifaddr *ifp = NULL;
3607         unsigned int hash = inet6_addr_hash(addr);
3608
3609         rcu_read_lock_bh();
3610         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3611                 if (!net_eq(dev_net(ifp->idev->dev), net))
3612                         continue;
3613                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3614                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3615                         ret = 1;
3616                         break;
3617                 }
3618         }
3619         rcu_read_unlock_bh();
3620         return ret;
3621 }
3622 #endif
3623
3624 /*
3625  *      Periodic address status verification
3626  */
3627
3628 static void addrconf_verify_rtnl(void)
3629 {
3630         unsigned long now, next, next_sec, next_sched;
3631         struct inet6_ifaddr *ifp;
3632         int i;
3633
3634         ASSERT_RTNL();
3635
3636         rcu_read_lock_bh();
3637         now = jiffies;
3638         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3639
3640         cancel_delayed_work(&addr_chk_work);
3641
3642         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3643 restart:
3644                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3645                         unsigned long age;
3646
3647                         /* When setting preferred_lft to a value not zero or
3648                          * infinity, while valid_lft is infinity
3649                          * IFA_F_PERMANENT has a non-infinity life time.
3650                          */
3651                         if ((ifp->flags & IFA_F_PERMANENT) &&
3652                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3653                                 continue;
3654
3655                         spin_lock(&ifp->lock);
3656                         /* We try to batch several events at once. */
3657                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3658
3659                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3660                             age >= ifp->valid_lft) {
3661                                 spin_unlock(&ifp->lock);
3662                                 in6_ifa_hold(ifp);
3663                                 ipv6_del_addr(ifp);
3664                                 goto restart;
3665                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3666                                 spin_unlock(&ifp->lock);
3667                                 continue;
3668                         } else if (age >= ifp->prefered_lft) {
3669                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3670                                 int deprecate = 0;
3671
3672                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3673                                         deprecate = 1;
3674                                         ifp->flags |= IFA_F_DEPRECATED;
3675                                 }
3676
3677                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3678                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3679                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3680
3681                                 spin_unlock(&ifp->lock);
3682
3683                                 if (deprecate) {
3684                                         in6_ifa_hold(ifp);
3685
3686                                         ipv6_ifa_notify(0, ifp);
3687                                         in6_ifa_put(ifp);
3688                                         goto restart;
3689                                 }
3690                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3691                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3692                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3693                                         ifp->idev->cnf.dad_transmits *
3694                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3695
3696                                 if (age >= ifp->prefered_lft - regen_advance) {
3697                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3698                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3699                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3700                                         if (!ifp->regen_count && ifpub) {
3701                                                 ifp->regen_count++;
3702                                                 in6_ifa_hold(ifp);
3703                                                 in6_ifa_hold(ifpub);
3704                                                 spin_unlock(&ifp->lock);
3705
3706                                                 spin_lock(&ifpub->lock);
3707                                                 ifpub->regen_count = 0;
3708                                                 spin_unlock(&ifpub->lock);
3709                                                 ipv6_create_tempaddr(ifpub, ifp);
3710                                                 in6_ifa_put(ifpub);
3711                                                 in6_ifa_put(ifp);
3712                                                 goto restart;
3713                                         }
3714                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3715                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3716                                 spin_unlock(&ifp->lock);
3717                         } else {
3718                                 /* ifp->prefered_lft <= ifp->valid_lft */
3719                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3720                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3721                                 spin_unlock(&ifp->lock);
3722                         }
3723                 }
3724         }
3725
3726         next_sec = round_jiffies_up(next);
3727         next_sched = next;
3728
3729         /* If rounded timeout is accurate enough, accept it. */
3730         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3731                 next_sched = next_sec;
3732
3733         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3734         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3735                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3736
3737         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3738               now, next, next_sec, next_sched);
3739         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3740         rcu_read_unlock_bh();
3741 }
3742
3743 static void addrconf_verify_work(struct work_struct *w)
3744 {
3745         rtnl_lock();
3746         addrconf_verify_rtnl();
3747         rtnl_unlock();
3748 }
3749
3750 static void addrconf_verify(void)
3751 {
3752         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3753 }
3754
3755 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3756                                      struct in6_addr **peer_pfx)
3757 {
3758         struct in6_addr *pfx = NULL;
3759
3760         *peer_pfx = NULL;
3761
3762         if (addr)
3763                 pfx = nla_data(addr);
3764
3765         if (local) {
3766                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3767                         *peer_pfx = pfx;
3768                 pfx = nla_data(local);
3769         }
3770
3771         return pfx;
3772 }
3773
3774 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3775         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3776         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3777         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3778         [IFA_FLAGS]             = { .len = sizeof(u32) },
3779 };
3780
3781 static int
3782 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3783 {
3784         struct net *net = sock_net(skb->sk);
3785         struct ifaddrmsg *ifm;
3786         struct nlattr *tb[IFA_MAX+1];
3787         struct in6_addr *pfx, *peer_pfx;
3788         u32 ifa_flags;
3789         int err;
3790
3791         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3792         if (err < 0)
3793                 return err;
3794
3795         ifm = nlmsg_data(nlh);
3796         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3797         if (pfx == NULL)
3798                 return -EINVAL;
3799
3800         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3801
3802         /* We ignore other flags so far. */
3803         ifa_flags &= IFA_F_MANAGETEMPADDR;
3804
3805         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3806                               ifm->ifa_prefixlen);
3807 }
3808
3809 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3810                              u32 prefered_lft, u32 valid_lft)
3811 {
3812         u32 flags;
3813         clock_t expires;
3814         unsigned long timeout;
3815         bool was_managetempaddr;
3816         bool had_prefixroute;
3817
3818         ASSERT_RTNL();
3819
3820         if (!valid_lft || (prefered_lft > valid_lft))
3821                 return -EINVAL;
3822
3823         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3824             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3825                 return -EINVAL;
3826
3827         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3828         if (addrconf_finite_timeout(timeout)) {
3829                 expires = jiffies_to_clock_t(timeout * HZ);
3830                 valid_lft = timeout;
3831                 flags = RTF_EXPIRES;
3832         } else {
3833                 expires = 0;
3834                 flags = 0;
3835                 ifa_flags |= IFA_F_PERMANENT;
3836         }
3837
3838         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3839         if (addrconf_finite_timeout(timeout)) {
3840                 if (timeout == 0)
3841                         ifa_flags |= IFA_F_DEPRECATED;
3842                 prefered_lft = timeout;
3843         }
3844
3845         spin_lock_bh(&ifp->lock);
3846         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3847         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3848                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
3849         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3850                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3851                         IFA_F_NOPREFIXROUTE);
3852         ifp->flags |= ifa_flags;
3853         ifp->tstamp = jiffies;
3854         ifp->valid_lft = valid_lft;
3855         ifp->prefered_lft = prefered_lft;
3856
3857         spin_unlock_bh(&ifp->lock);
3858         if (!(ifp->flags&IFA_F_TENTATIVE))
3859                 ipv6_ifa_notify(0, ifp);
3860
3861         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3862                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3863                                       expires, flags);
3864         } else if (had_prefixroute) {
3865                 enum cleanup_prefix_rt_t action;
3866                 unsigned long rt_expires;
3867
3868                 write_lock_bh(&ifp->idev->lock);
3869                 action = check_cleanup_prefix_route(ifp, &rt_expires);
3870                 write_unlock_bh(&ifp->idev->lock);
3871
3872                 if (action != CLEANUP_PREFIX_RT_NOP) {
3873                         cleanup_prefix_route(ifp, rt_expires,
3874                                 action == CLEANUP_PREFIX_RT_DEL);
3875                 }
3876         }
3877
3878         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3879                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3880                         valid_lft = prefered_lft = 0;
3881                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3882                                  !was_managetempaddr, jiffies);
3883         }
3884
3885         addrconf_verify_rtnl();
3886
3887         return 0;
3888 }
3889
3890 static int
3891 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3892 {
3893         struct net *net = sock_net(skb->sk);
3894         struct ifaddrmsg *ifm;
3895         struct nlattr *tb[IFA_MAX+1];
3896         struct in6_addr *pfx, *peer_pfx;
3897         struct inet6_ifaddr *ifa;
3898         struct net_device *dev;
3899         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3900         u32 ifa_flags;
3901         int err;
3902
3903         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3904         if (err < 0)
3905                 return err;
3906
3907         ifm = nlmsg_data(nlh);
3908         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3909         if (pfx == NULL)
3910                 return -EINVAL;
3911
3912         if (tb[IFA_CACHEINFO]) {
3913                 struct ifa_cacheinfo *ci;
3914
3915                 ci = nla_data(tb[IFA_CACHEINFO]);
3916                 valid_lft = ci->ifa_valid;
3917                 preferred_lft = ci->ifa_prefered;
3918         } else {
3919                 preferred_lft = INFINITY_LIFE_TIME;
3920                 valid_lft = INFINITY_LIFE_TIME;
3921         }
3922
3923         dev =  __dev_get_by_index(net, ifm->ifa_index);
3924         if (dev == NULL)
3925                 return -ENODEV;
3926
3927         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3928
3929         /* We ignore other flags so far. */
3930         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3931                      IFA_F_NOPREFIXROUTE;
3932
3933         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3934         if (ifa == NULL) {
3935                 /*
3936                  * It would be best to check for !NLM_F_CREATE here but
3937                  * userspace already relies on not having to provide this.
3938                  */
3939                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3940                                       ifm->ifa_prefixlen, ifa_flags,
3941                                       preferred_lft, valid_lft);
3942         }
3943
3944         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3945             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3946                 err = -EEXIST;
3947         else
3948                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3949
3950         in6_ifa_put(ifa);
3951
3952         return err;
3953 }
3954
3955 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3956                           u8 scope, int ifindex)
3957 {
3958         struct ifaddrmsg *ifm;
3959
3960         ifm = nlmsg_data(nlh);
3961         ifm->ifa_family = AF_INET6;
3962         ifm->ifa_prefixlen = prefixlen;
3963         ifm->ifa_flags = flags;
3964         ifm->ifa_scope = scope;
3965         ifm->ifa_index = ifindex;
3966 }
3967
3968 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3969                          unsigned long tstamp, u32 preferred, u32 valid)
3970 {
3971         struct ifa_cacheinfo ci;
3972
3973         ci.cstamp = cstamp_delta(cstamp);
3974         ci.tstamp = cstamp_delta(tstamp);
3975         ci.ifa_prefered = preferred;
3976         ci.ifa_valid = valid;
3977
3978         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3979 }
3980
3981 static inline int rt_scope(int ifa_scope)
3982 {
3983         if (ifa_scope & IFA_HOST)
3984                 return RT_SCOPE_HOST;
3985         else if (ifa_scope & IFA_LINK)
3986                 return RT_SCOPE_LINK;
3987         else if (ifa_scope & IFA_SITE)
3988                 return RT_SCOPE_SITE;
3989         else
3990                 return RT_SCOPE_UNIVERSE;
3991 }
3992
3993 static inline int inet6_ifaddr_msgsize(void)
3994 {
3995         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3996                + nla_total_size(16) /* IFA_LOCAL */
3997                + nla_total_size(16) /* IFA_ADDRESS */
3998                + nla_total_size(sizeof(struct ifa_cacheinfo))
3999                + nla_total_size(4)  /* IFA_FLAGS */;
4000 }
4001
4002 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4003                              u32 portid, u32 seq, int event, unsigned int flags)
4004 {
4005         struct nlmsghdr  *nlh;
4006         u32 preferred, valid;
4007
4008         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4009         if (nlh == NULL)
4010                 return -EMSGSIZE;
4011
4012         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4013                       ifa->idev->dev->ifindex);
4014
4015         if (!((ifa->flags&IFA_F_PERMANENT) &&
4016               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4017                 preferred = ifa->prefered_lft;
4018                 valid = ifa->valid_lft;
4019                 if (preferred != INFINITY_LIFE_TIME) {
4020                         long tval = (jiffies - ifa->tstamp)/HZ;
4021                         if (preferred > tval)
4022                                 preferred -= tval;
4023                         else
4024                                 preferred = 0;
4025                         if (valid != INFINITY_LIFE_TIME) {
4026                                 if (valid > tval)
4027                                         valid -= tval;
4028                                 else
4029                                         valid = 0;
4030                         }
4031                 }
4032         } else {
4033                 preferred = INFINITY_LIFE_TIME;
4034                 valid = INFINITY_LIFE_TIME;
4035         }
4036
4037         if (!ipv6_addr_any(&ifa->peer_addr)) {
4038                 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
4039                     nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4040                         goto error;
4041         } else
4042                 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4043                         goto error;
4044
4045         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4046                 goto error;
4047
4048         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4049                 goto error;
4050
4051         nlmsg_end(skb, nlh);
4052         return 0;
4053
4054 error:
4055         nlmsg_cancel(skb, nlh);
4056         return -EMSGSIZE;
4057 }
4058
4059 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4060                                 u32 portid, u32 seq, int event, u16 flags)
4061 {
4062         struct nlmsghdr  *nlh;
4063         u8 scope = RT_SCOPE_UNIVERSE;
4064         int ifindex = ifmca->idev->dev->ifindex;
4065
4066         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4067                 scope = RT_SCOPE_SITE;
4068
4069         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4070         if (nlh == NULL)
4071                 return -EMSGSIZE;
4072
4073         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4074         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4075             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4076                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4077                 nlmsg_cancel(skb, nlh);
4078                 return -EMSGSIZE;
4079         }
4080
4081         nlmsg_end(skb, nlh);
4082         return 0;
4083 }
4084
4085 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4086                                 u32 portid, u32 seq, int event, unsigned int flags)
4087 {
4088         struct nlmsghdr  *nlh;
4089         u8 scope = RT_SCOPE_UNIVERSE;
4090         int ifindex = ifaca->aca_idev->dev->ifindex;
4091
4092         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4093                 scope = RT_SCOPE_SITE;
4094
4095         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4096         if (nlh == NULL)
4097                 return -EMSGSIZE;
4098
4099         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4100         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4101             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4102                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4103                 nlmsg_cancel(skb, nlh);
4104                 return -EMSGSIZE;
4105         }
4106
4107         nlmsg_end(skb, nlh);
4108         return 0;
4109 }
4110
4111 enum addr_type_t {
4112         UNICAST_ADDR,
4113         MULTICAST_ADDR,
4114         ANYCAST_ADDR,
4115 };
4116
4117 /* called with rcu_read_lock() */
4118 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4119                           struct netlink_callback *cb, enum addr_type_t type,
4120                           int s_ip_idx, int *p_ip_idx)
4121 {
4122         struct ifmcaddr6 *ifmca;
4123         struct ifacaddr6 *ifaca;
4124         int err = 1;
4125         int ip_idx = *p_ip_idx;
4126
4127         read_lock_bh(&idev->lock);
4128         switch (type) {
4129         case UNICAST_ADDR: {
4130                 struct inet6_ifaddr *ifa;
4131
4132                 /* unicast address incl. temp addr */
4133                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4134                         if (++ip_idx < s_ip_idx)
4135                                 continue;
4136                         err = inet6_fill_ifaddr(skb, ifa,
4137                                                 NETLINK_CB(cb->skb).portid,
4138                                                 cb->nlh->nlmsg_seq,
4139                                                 RTM_NEWADDR,
4140                                                 NLM_F_MULTI);
4141                         if (err < 0)
4142                                 break;
4143                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4144                 }
4145                 break;
4146         }
4147         case MULTICAST_ADDR:
4148                 /* multicast address */
4149                 for (ifmca = idev->mc_list; ifmca;
4150                      ifmca = ifmca->next, ip_idx++) {
4151                         if (ip_idx < s_ip_idx)
4152                                 continue;
4153                         err = inet6_fill_ifmcaddr(skb, ifmca,
4154                                                   NETLINK_CB(cb->skb).portid,
4155                                                   cb->nlh->nlmsg_seq,
4156                                                   RTM_GETMULTICAST,
4157                                                   NLM_F_MULTI);
4158                         if (err < 0)
4159                                 break;
4160                 }
4161                 break;
4162         case ANYCAST_ADDR:
4163                 /* anycast address */
4164                 for (ifaca = idev->ac_list; ifaca;
4165                      ifaca = ifaca->aca_next, ip_idx++) {
4166                         if (ip_idx < s_ip_idx)
4167                                 continue;
4168                         err = inet6_fill_ifacaddr(skb, ifaca,
4169                                                   NETLINK_CB(cb->skb).portid,
4170                                                   cb->nlh->nlmsg_seq,
4171                                                   RTM_GETANYCAST,
4172                                                   NLM_F_MULTI);
4173                         if (err < 0)
4174                                 break;
4175                 }
4176                 break;
4177         default:
4178                 break;
4179         }
4180         read_unlock_bh(&idev->lock);
4181         *p_ip_idx = ip_idx;
4182         return err;
4183 }
4184
4185 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4186                            enum addr_type_t type)
4187 {
4188         struct net *net = sock_net(skb->sk);
4189         int h, s_h;
4190         int idx, ip_idx;
4191         int s_idx, s_ip_idx;
4192         struct net_device *dev;
4193         struct inet6_dev *idev;
4194         struct hlist_head *head;
4195
4196         s_h = cb->args[0];
4197         s_idx = idx = cb->args[1];
4198         s_ip_idx = ip_idx = cb->args[2];
4199
4200         rcu_read_lock();
4201         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4202         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4203                 idx = 0;
4204                 head = &net->dev_index_head[h];
4205                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4206                         if (idx < s_idx)
4207                                 goto cont;
4208                         if (h > s_h || idx > s_idx)
4209                                 s_ip_idx = 0;
4210                         ip_idx = 0;
4211                         idev = __in6_dev_get(dev);
4212                         if (!idev)
4213                                 goto cont;
4214
4215                         if (in6_dump_addrs(idev, skb, cb, type,
4216                                            s_ip_idx, &ip_idx) <= 0)
4217                                 goto done;
4218 cont:
4219                         idx++;
4220                 }
4221         }
4222 done:
4223         rcu_read_unlock();
4224         cb->args[0] = h;
4225         cb->args[1] = idx;
4226         cb->args[2] = ip_idx;
4227
4228         return skb->len;
4229 }
4230
4231 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4232 {
4233         enum addr_type_t type = UNICAST_ADDR;
4234
4235         return inet6_dump_addr(skb, cb, type);
4236 }
4237
4238 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4239 {
4240         enum addr_type_t type = MULTICAST_ADDR;
4241
4242         return inet6_dump_addr(skb, cb, type);
4243 }
4244
4245
4246 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4247 {
4248         enum addr_type_t type = ANYCAST_ADDR;
4249
4250         return inet6_dump_addr(skb, cb, type);
4251 }
4252
4253 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4254 {
4255         struct net *net = sock_net(in_skb->sk);
4256         struct ifaddrmsg *ifm;
4257         struct nlattr *tb[IFA_MAX+1];
4258         struct in6_addr *addr = NULL, *peer;
4259         struct net_device *dev = NULL;
4260         struct inet6_ifaddr *ifa;
4261         struct sk_buff *skb;
4262         int err;
4263
4264         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4265         if (err < 0)
4266                 goto errout;
4267
4268         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4269         if (addr == NULL) {
4270                 err = -EINVAL;
4271                 goto errout;
4272         }
4273
4274         ifm = nlmsg_data(nlh);
4275         if (ifm->ifa_index)
4276                 dev = __dev_get_by_index(net, ifm->ifa_index);
4277
4278         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4279         if (!ifa) {
4280                 err = -EADDRNOTAVAIL;
4281                 goto errout;
4282         }
4283
4284         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4285         if (!skb) {
4286                 err = -ENOBUFS;
4287                 goto errout_ifa;
4288         }
4289
4290         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4291                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4292         if (err < 0) {
4293                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4294                 WARN_ON(err == -EMSGSIZE);
4295                 kfree_skb(skb);
4296                 goto errout_ifa;
4297         }
4298         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4299 errout_ifa:
4300         in6_ifa_put(ifa);
4301 errout:
4302         return err;
4303 }
4304
4305 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4306 {
4307         struct sk_buff *skb;
4308         struct net *net = dev_net(ifa->idev->dev);
4309         int err = -ENOBUFS;
4310
4311         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4312         if (skb == NULL)
4313                 goto errout;
4314
4315         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4316         if (err < 0) {
4317                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4318                 WARN_ON(err == -EMSGSIZE);
4319                 kfree_skb(skb);
4320                 goto errout;
4321         }
4322         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4323         return;
4324 errout:
4325         if (err < 0)
4326                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4327 }
4328
4329 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4330                                 __s32 *array, int bytes)
4331 {
4332         BUG_ON(bytes < (DEVCONF_MAX * 4));
4333
4334         memset(array, 0, bytes);
4335         array[DEVCONF_FORWARDING] = cnf->forwarding;
4336         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4337         array[DEVCONF_MTU6] = cnf->mtu6;
4338         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4339         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4340         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4341         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4342         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4343         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4344                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4345         array[DEVCONF_RTR_SOLICIT_DELAY] =
4346                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4347         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4348         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4349                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4350         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4351                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4352         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4353         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4354         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4355         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4356         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4357         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4358         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4359         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4360 #ifdef CONFIG_IPV6_ROUTER_PREF
4361         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4362         array[DEVCONF_RTR_PROBE_INTERVAL] =
4363                 jiffies_to_msecs(cnf->rtr_probe_interval);
4364 #ifdef CONFIG_IPV6_ROUTE_INFO
4365         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4366 #endif
4367 #endif
4368         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4369         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4370 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4371         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4372         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4373 #endif
4374 #ifdef CONFIG_IPV6_MROUTE
4375         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4376 #endif
4377         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4378         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4379         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4380         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4381         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4382         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4383 }
4384
4385 static inline size_t inet6_ifla6_size(void)
4386 {
4387         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4388              + nla_total_size(sizeof(struct ifla_cacheinfo))
4389              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4390              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4391              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4392              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4393 }
4394
4395 static inline size_t inet6_if_nlmsg_size(void)
4396 {
4397         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4398                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4399                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4400                + nla_total_size(4) /* IFLA_MTU */
4401                + nla_total_size(4) /* IFLA_LINK */
4402                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4403 }
4404
4405 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4406                                       int items, int bytes)
4407 {
4408         int i;
4409         int pad = bytes - sizeof(u64) * items;
4410         BUG_ON(pad < 0);
4411
4412         /* Use put_unaligned() because stats may not be aligned for u64. */
4413         put_unaligned(items, &stats[0]);
4414         for (i = 1; i < items; i++)
4415                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4416
4417         memset(&stats[items], 0, pad);
4418 }
4419
4420 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4421                                       int items, int bytes, size_t syncpoff)
4422 {
4423         int i;
4424         int pad = bytes - sizeof(u64) * items;
4425         BUG_ON(pad < 0);
4426
4427         /* Use put_unaligned() because stats may not be aligned for u64. */
4428         put_unaligned(items, &stats[0]);
4429         for (i = 1; i < items; i++)
4430                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4431
4432         memset(&stats[items], 0, pad);
4433 }
4434
4435 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4436                              int bytes)
4437 {
4438         switch (attrtype) {
4439         case IFLA_INET6_STATS:
4440                 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4441                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4442                 break;
4443         case IFLA_INET6_ICMP6STATS:
4444                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4445                 break;
4446         }
4447 }
4448
4449 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4450 {
4451         struct nlattr *nla;
4452         struct ifla_cacheinfo ci;
4453
4454         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4455                 goto nla_put_failure;
4456         ci.max_reasm_len = IPV6_MAXPLEN;
4457         ci.tstamp = cstamp_delta(idev->tstamp);
4458         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4459         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4460         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4461                 goto nla_put_failure;
4462         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4463         if (nla == NULL)
4464                 goto nla_put_failure;
4465         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4466
4467         /* XXX - MC not implemented */
4468
4469         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4470         if (nla == NULL)
4471                 goto nla_put_failure;
4472         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4473
4474         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4475         if (nla == NULL)
4476                 goto nla_put_failure;
4477         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4478
4479         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4480         if (nla == NULL)
4481                 goto nla_put_failure;
4482
4483         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4484                 goto nla_put_failure;
4485
4486         read_lock_bh(&idev->lock);
4487         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4488         read_unlock_bh(&idev->lock);
4489
4490         return 0;
4491
4492 nla_put_failure:
4493         return -EMSGSIZE;
4494 }
4495
4496 static size_t inet6_get_link_af_size(const struct net_device *dev)
4497 {
4498         if (!__in6_dev_get(dev))
4499                 return 0;
4500
4501         return inet6_ifla6_size();
4502 }
4503
4504 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4505 {
4506         struct inet6_dev *idev = __in6_dev_get(dev);
4507
4508         if (!idev)
4509                 return -ENODATA;
4510
4511         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4512                 return -EMSGSIZE;
4513
4514         return 0;
4515 }
4516
4517 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4518 {
4519         struct inet6_ifaddr *ifp;
4520         struct net_device *dev = idev->dev;
4521         bool update_rs = false;
4522         struct in6_addr ll_addr;
4523
4524         ASSERT_RTNL();
4525
4526         if (token == NULL)
4527                 return -EINVAL;
4528         if (ipv6_addr_any(token))
4529                 return -EINVAL;
4530         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4531                 return -EINVAL;
4532         if (!ipv6_accept_ra(idev))
4533                 return -EINVAL;
4534         if (idev->cnf.rtr_solicits <= 0)
4535                 return -EINVAL;
4536
4537         write_lock_bh(&idev->lock);
4538
4539         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4540         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4541
4542         write_unlock_bh(&idev->lock);
4543
4544         if (!idev->dead && (idev->if_flags & IF_READY) &&
4545             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4546                              IFA_F_OPTIMISTIC)) {
4547
4548                 /* If we're not ready, then normal ifup will take care
4549                  * of this. Otherwise, we need to request our rs here.
4550                  */
4551                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4552                 update_rs = true;
4553         }
4554
4555         write_lock_bh(&idev->lock);
4556
4557         if (update_rs) {
4558                 idev->if_flags |= IF_RS_SENT;
4559                 idev->rs_probes = 1;
4560                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4561         }
4562
4563         /* Well, that's kinda nasty ... */
4564         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4565                 spin_lock(&ifp->lock);
4566                 if (ifp->tokenized) {
4567                         ifp->valid_lft = 0;
4568                         ifp->prefered_lft = 0;
4569                 }
4570                 spin_unlock(&ifp->lock);
4571         }
4572
4573         write_unlock_bh(&idev->lock);
4574         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4575         addrconf_verify_rtnl();
4576         return 0;
4577 }
4578
4579 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4580 {
4581         int err = -EINVAL;
4582         struct inet6_dev *idev = __in6_dev_get(dev);
4583         struct nlattr *tb[IFLA_INET6_MAX + 1];
4584
4585         if (!idev)
4586                 return -EAFNOSUPPORT;
4587
4588         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4589                 BUG();
4590
4591         if (tb[IFLA_INET6_TOKEN]) {
4592                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4593                 if (err)
4594                         return err;
4595         }
4596
4597         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4598                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4599
4600                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4601                     mode != IN6_ADDR_GEN_MODE_NONE)
4602                         return -EINVAL;
4603                 idev->addr_gen_mode = mode;
4604                 err = 0;
4605         }
4606
4607         return err;
4608 }
4609
4610 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4611                              u32 portid, u32 seq, int event, unsigned int flags)
4612 {
4613         struct net_device *dev = idev->dev;
4614         struct ifinfomsg *hdr;
4615         struct nlmsghdr *nlh;
4616         void *protoinfo;
4617
4618         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4619         if (nlh == NULL)
4620                 return -EMSGSIZE;
4621
4622         hdr = nlmsg_data(nlh);
4623         hdr->ifi_family = AF_INET6;
4624         hdr->__ifi_pad = 0;
4625         hdr->ifi_type = dev->type;
4626         hdr->ifi_index = dev->ifindex;
4627         hdr->ifi_flags = dev_get_flags(dev);
4628         hdr->ifi_change = 0;
4629
4630         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4631             (dev->addr_len &&
4632              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4633             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4634             (dev->ifindex != dev->iflink &&
4635              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4636                 goto nla_put_failure;
4637         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4638         if (protoinfo == NULL)
4639                 goto nla_put_failure;
4640
4641         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4642                 goto nla_put_failure;
4643
4644         nla_nest_end(skb, protoinfo);
4645         nlmsg_end(skb, nlh);
4646         return 0;
4647
4648 nla_put_failure:
4649         nlmsg_cancel(skb, nlh);
4650         return -EMSGSIZE;
4651 }
4652
4653 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4654 {
4655         struct net *net = sock_net(skb->sk);
4656         int h, s_h;
4657         int idx = 0, s_idx;
4658         struct net_device *dev;
4659         struct inet6_dev *idev;
4660         struct hlist_head *head;
4661
4662         s_h = cb->args[0];
4663         s_idx = cb->args[1];
4664
4665         rcu_read_lock();
4666         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4667                 idx = 0;
4668                 head = &net->dev_index_head[h];
4669                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4670                         if (idx < s_idx)
4671                                 goto cont;
4672                         idev = __in6_dev_get(dev);
4673                         if (!idev)
4674                                 goto cont;
4675                         if (inet6_fill_ifinfo(skb, idev,
4676                                               NETLINK_CB(cb->skb).portid,
4677                                               cb->nlh->nlmsg_seq,
4678                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
4679                                 goto out;
4680 cont:
4681                         idx++;
4682                 }
4683         }
4684 out:
4685         rcu_read_unlock();
4686         cb->args[1] = idx;
4687         cb->args[0] = h;
4688
4689         return skb->len;
4690 }
4691
4692 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4693 {
4694         struct sk_buff *skb;
4695         struct net *net = dev_net(idev->dev);
4696         int err = -ENOBUFS;
4697
4698         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4699         if (skb == NULL)
4700                 goto errout;
4701
4702         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4703         if (err < 0) {
4704                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4705                 WARN_ON(err == -EMSGSIZE);
4706                 kfree_skb(skb);
4707                 goto errout;
4708         }
4709         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4710         return;
4711 errout:
4712         if (err < 0)
4713                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4714 }
4715
4716 static inline size_t inet6_prefix_nlmsg_size(void)
4717 {
4718         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4719                + nla_total_size(sizeof(struct in6_addr))
4720                + nla_total_size(sizeof(struct prefix_cacheinfo));
4721 }
4722
4723 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4724                              struct prefix_info *pinfo, u32 portid, u32 seq,
4725                              int event, unsigned int flags)
4726 {
4727         struct prefixmsg *pmsg;
4728         struct nlmsghdr *nlh;
4729         struct prefix_cacheinfo ci;
4730
4731         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4732         if (nlh == NULL)
4733                 return -EMSGSIZE;
4734
4735         pmsg = nlmsg_data(nlh);
4736         pmsg->prefix_family = AF_INET6;
4737         pmsg->prefix_pad1 = 0;
4738         pmsg->prefix_pad2 = 0;
4739         pmsg->prefix_ifindex = idev->dev->ifindex;
4740         pmsg->prefix_len = pinfo->prefix_len;
4741         pmsg->prefix_type = pinfo->type;
4742         pmsg->prefix_pad3 = 0;
4743         pmsg->prefix_flags = 0;
4744         if (pinfo->onlink)
4745                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4746         if (pinfo->autoconf)
4747                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4748
4749         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4750                 goto nla_put_failure;
4751         ci.preferred_time = ntohl(pinfo->prefered);
4752         ci.valid_time = ntohl(pinfo->valid);
4753         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4754                 goto nla_put_failure;
4755         nlmsg_end(skb, nlh);
4756         return 0;
4757
4758 nla_put_failure:
4759         nlmsg_cancel(skb, nlh);
4760         return -EMSGSIZE;
4761 }
4762
4763 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4764                          struct prefix_info *pinfo)
4765 {
4766         struct sk_buff *skb;
4767         struct net *net = dev_net(idev->dev);
4768         int err = -ENOBUFS;
4769
4770         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4771         if (skb == NULL)
4772                 goto errout;
4773
4774         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4775         if (err < 0) {
4776                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4777                 WARN_ON(err == -EMSGSIZE);
4778                 kfree_skb(skb);
4779                 goto errout;
4780         }
4781         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4782         return;
4783 errout:
4784         if (err < 0)
4785                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4786 }
4787
4788 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4789 {
4790         struct net *net = dev_net(ifp->idev->dev);
4791
4792         if (event)
4793                 ASSERT_RTNL();
4794
4795         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4796
4797         switch (event) {
4798         case RTM_NEWADDR:
4799                 /*
4800                  * If the address was optimistic
4801                  * we inserted the route at the start of
4802                  * our DAD process, so we don't need
4803                  * to do it again
4804                  */
4805                 if (!(ifp->rt->rt6i_node))
4806                         ip6_ins_rt(ifp->rt);
4807                 if (ifp->idev->cnf.forwarding)
4808                         addrconf_join_anycast(ifp);
4809                 if (!ipv6_addr_any(&ifp->peer_addr))
4810                         addrconf_prefix_route(&ifp->peer_addr, 128,
4811                                               ifp->idev->dev, 0, 0);
4812                 break;
4813         case RTM_DELADDR:
4814                 if (ifp->idev->cnf.forwarding)
4815                         addrconf_leave_anycast(ifp);
4816                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4817                 if (!ipv6_addr_any(&ifp->peer_addr)) {
4818                         struct rt6_info *rt;
4819
4820                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
4821                                                        ifp->idev->dev, 0, 0);
4822                         if (rt && ip6_del_rt(rt))
4823                                 dst_free(&rt->dst);
4824                 }
4825                 dst_hold(&ifp->rt->dst);
4826
4827                 if (ip6_del_rt(ifp->rt))
4828                         dst_free(&ifp->rt->dst);
4829
4830                 rt_genid_bump_ipv6(net);
4831                 break;
4832         }
4833         atomic_inc(&net->ipv6.dev_addr_genid);
4834 }
4835
4836 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4837 {
4838         rcu_read_lock_bh();
4839         if (likely(ifp->idev->dead == 0))
4840                 __ipv6_ifa_notify(event, ifp);
4841         rcu_read_unlock_bh();
4842 }
4843
4844 #ifdef CONFIG_SYSCTL
4845
4846 static
4847 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4848                            void __user *buffer, size_t *lenp, loff_t *ppos)
4849 {
4850         int *valp = ctl->data;
4851         int val = *valp;
4852         loff_t pos = *ppos;
4853         struct ctl_table lctl;
4854         int ret;
4855
4856         /*
4857          * ctl->data points to idev->cnf.forwarding, we should
4858          * not modify it until we get the rtnl lock.
4859          */
4860         lctl = *ctl;
4861         lctl.data = &val;
4862
4863         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4864
4865         if (write)
4866                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4867         if (ret)
4868                 *ppos = pos;
4869         return ret;
4870 }
4871
4872 static void dev_disable_change(struct inet6_dev *idev)
4873 {
4874         struct netdev_notifier_info info;
4875
4876         if (!idev || !idev->dev)
4877                 return;
4878
4879         netdev_notifier_info_init(&info, idev->dev);
4880         if (idev->cnf.disable_ipv6)
4881                 addrconf_notify(NULL, NETDEV_DOWN, &info);
4882         else
4883                 addrconf_notify(NULL, NETDEV_UP, &info);
4884 }
4885
4886 static void addrconf_disable_change(struct net *net, __s32 newf)
4887 {
4888         struct net_device *dev;
4889         struct inet6_dev *idev;
4890
4891         rcu_read_lock();
4892         for_each_netdev_rcu(net, dev) {
4893                 idev = __in6_dev_get(dev);
4894                 if (idev) {
4895                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4896                         idev->cnf.disable_ipv6 = newf;
4897                         if (changed)
4898                                 dev_disable_change(idev);
4899                 }
4900         }
4901         rcu_read_unlock();
4902 }
4903
4904 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4905 {
4906         struct net *net;
4907         int old;
4908
4909         if (!rtnl_trylock())
4910                 return restart_syscall();
4911
4912         net = (struct net *)table->extra2;
4913         old = *p;
4914         *p = newf;
4915
4916         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4917                 rtnl_unlock();
4918                 return 0;
4919         }
4920
4921         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4922                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4923                 addrconf_disable_change(net, newf);
4924         } else if ((!newf) ^ (!old))
4925                 dev_disable_change((struct inet6_dev *)table->extra1);
4926
4927         rtnl_unlock();
4928         return 0;
4929 }
4930
4931 static
4932 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4933                             void __user *buffer, size_t *lenp, loff_t *ppos)
4934 {
4935         int *valp = ctl->data;
4936         int val = *valp;
4937         loff_t pos = *ppos;
4938         struct ctl_table lctl;
4939         int ret;
4940
4941         /*
4942          * ctl->data points to idev->cnf.disable_ipv6, we should
4943          * not modify it until we get the rtnl lock.
4944          */
4945         lctl = *ctl;
4946         lctl.data = &val;
4947
4948         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4949
4950         if (write)
4951                 ret = addrconf_disable_ipv6(ctl, valp, val);
4952         if (ret)
4953                 *ppos = pos;
4954         return ret;
4955 }
4956
4957 static
4958 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4959                               void __user *buffer, size_t *lenp, loff_t *ppos)
4960 {
4961         int *valp = ctl->data;
4962         int ret;
4963         int old, new;
4964
4965         old = *valp;
4966         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4967         new = *valp;
4968
4969         if (write && old != new) {
4970                 struct net *net = ctl->extra2;
4971
4972                 if (!rtnl_trylock())
4973                         return restart_syscall();
4974
4975                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4976                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4977                                                      NETCONFA_IFINDEX_DEFAULT,
4978                                                      net->ipv6.devconf_dflt);
4979                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4980                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4981                                                      NETCONFA_IFINDEX_ALL,
4982                                                      net->ipv6.devconf_all);
4983                 else {
4984                         struct inet6_dev *idev = ctl->extra1;
4985
4986                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4987                                                      idev->dev->ifindex,
4988                                                      &idev->cnf);
4989                 }
4990                 rtnl_unlock();
4991         }
4992
4993         return ret;
4994 }
4995
4996
4997 static struct addrconf_sysctl_table
4998 {
4999         struct ctl_table_header *sysctl_header;
5000         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5001 } addrconf_sysctl __read_mostly = {
5002         .sysctl_header = NULL,
5003         .addrconf_vars = {
5004                 {
5005                         .procname       = "forwarding",
5006                         .data           = &ipv6_devconf.forwarding,
5007                         .maxlen         = sizeof(int),
5008                         .mode           = 0644,
5009                         .proc_handler   = addrconf_sysctl_forward,
5010                 },
5011                 {
5012                         .procname       = "hop_limit",
5013                         .data           = &ipv6_devconf.hop_limit,
5014                         .maxlen         = sizeof(int),
5015                         .mode           = 0644,
5016                         .proc_handler   = proc_dointvec,
5017                 },
5018                 {
5019                         .procname       = "mtu",
5020                         .data           = &ipv6_devconf.mtu6,
5021                         .maxlen         = sizeof(int),
5022                         .mode           = 0644,
5023                         .proc_handler   = proc_dointvec,
5024                 },
5025                 {
5026                         .procname       = "accept_ra",
5027                         .data           = &ipv6_devconf.accept_ra,
5028                         .maxlen         = sizeof(int),
5029                         .mode           = 0644,
5030                         .proc_handler   = proc_dointvec,
5031                 },
5032                 {
5033                         .procname       = "accept_redirects",
5034                         .data           = &ipv6_devconf.accept_redirects,
5035                         .maxlen         = sizeof(int),
5036                         .mode           = 0644,
5037                         .proc_handler   = proc_dointvec,
5038                 },
5039                 {
5040                         .procname       = "autoconf",
5041                         .data           = &ipv6_devconf.autoconf,
5042                         .maxlen         = sizeof(int),
5043                         .mode           = 0644,
5044                         .proc_handler   = proc_dointvec,
5045                 },
5046                 {
5047                         .procname       = "dad_transmits",
5048                         .data           = &ipv6_devconf.dad_transmits,
5049                         .maxlen         = sizeof(int),
5050                         .mode           = 0644,
5051                         .proc_handler   = proc_dointvec,
5052                 },
5053                 {
5054                         .procname       = "router_solicitations",
5055                         .data           = &ipv6_devconf.rtr_solicits,
5056                         .maxlen         = sizeof(int),
5057                         .mode           = 0644,
5058                         .proc_handler   = proc_dointvec,
5059                 },
5060                 {
5061                         .procname       = "router_solicitation_interval",
5062                         .data           = &ipv6_devconf.rtr_solicit_interval,
5063                         .maxlen         = sizeof(int),
5064                         .mode           = 0644,
5065                         .proc_handler   = proc_dointvec_jiffies,
5066                 },
5067                 {
5068                         .procname       = "router_solicitation_delay",
5069                         .data           = &ipv6_devconf.rtr_solicit_delay,
5070                         .maxlen         = sizeof(int),
5071                         .mode           = 0644,
5072                         .proc_handler   = proc_dointvec_jiffies,
5073                 },
5074                 {
5075                         .procname       = "force_mld_version",
5076                         .data           = &ipv6_devconf.force_mld_version,
5077                         .maxlen         = sizeof(int),
5078                         .mode           = 0644,
5079                         .proc_handler   = proc_dointvec,
5080                 },
5081                 {
5082                         .procname       = "mldv1_unsolicited_report_interval",
5083                         .data           =
5084                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5085                         .maxlen         = sizeof(int),
5086                         .mode           = 0644,
5087                         .proc_handler   = proc_dointvec_ms_jiffies,
5088                 },
5089                 {
5090                         .procname       = "mldv2_unsolicited_report_interval",
5091                         .data           =
5092                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5093                         .maxlen         = sizeof(int),
5094                         .mode           = 0644,
5095                         .proc_handler   = proc_dointvec_ms_jiffies,
5096                 },
5097                 {
5098                         .procname       = "use_tempaddr",
5099                         .data           = &ipv6_devconf.use_tempaddr,
5100                         .maxlen         = sizeof(int),
5101                         .mode           = 0644,
5102                         .proc_handler   = proc_dointvec,
5103                 },
5104                 {
5105                         .procname       = "temp_valid_lft",
5106                         .data           = &ipv6_devconf.temp_valid_lft,
5107                         .maxlen         = sizeof(int),
5108                         .mode           = 0644,
5109                         .proc_handler   = proc_dointvec,
5110                 },
5111                 {
5112                         .procname       = "temp_prefered_lft",
5113                         .data           = &ipv6_devconf.temp_prefered_lft,
5114                         .maxlen         = sizeof(int),
5115                         .mode           = 0644,
5116                         .proc_handler   = proc_dointvec,
5117                 },
5118                 {
5119                         .procname       = "regen_max_retry",
5120                         .data           = &ipv6_devconf.regen_max_retry,
5121                         .maxlen         = sizeof(int),
5122                         .mode           = 0644,
5123                         .proc_handler   = proc_dointvec,
5124                 },
5125                 {
5126                         .procname       = "max_desync_factor",
5127                         .data           = &ipv6_devconf.max_desync_factor,
5128                         .maxlen         = sizeof(int),
5129                         .mode           = 0644,
5130                         .proc_handler   = proc_dointvec,
5131                 },
5132                 {
5133                         .procname       = "max_addresses",
5134                         .data           = &ipv6_devconf.max_addresses,
5135                         .maxlen         = sizeof(int),
5136                         .mode           = 0644,
5137                         .proc_handler   = proc_dointvec,
5138                 },
5139                 {
5140                         .procname       = "accept_ra_defrtr",
5141                         .data           = &ipv6_devconf.accept_ra_defrtr,
5142                         .maxlen         = sizeof(int),
5143                         .mode           = 0644,
5144                         .proc_handler   = proc_dointvec,
5145                 },
5146                 {
5147                         .procname       = "accept_ra_pinfo",
5148                         .data           = &ipv6_devconf.accept_ra_pinfo,
5149                         .maxlen         = sizeof(int),
5150                         .mode           = 0644,
5151                         .proc_handler   = proc_dointvec,
5152                 },
5153 #ifdef CONFIG_IPV6_ROUTER_PREF
5154                 {
5155                         .procname       = "accept_ra_rtr_pref",
5156                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5157                         .maxlen         = sizeof(int),
5158                         .mode           = 0644,
5159                         .proc_handler   = proc_dointvec,
5160                 },
5161                 {
5162                         .procname       = "router_probe_interval",
5163                         .data           = &ipv6_devconf.rtr_probe_interval,
5164                         .maxlen         = sizeof(int),
5165                         .mode           = 0644,
5166                         .proc_handler   = proc_dointvec_jiffies,
5167                 },
5168 #ifdef CONFIG_IPV6_ROUTE_INFO
5169                 {
5170                         .procname       = "accept_ra_rt_info_max_plen",
5171                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5172                         .maxlen         = sizeof(int),
5173                         .mode           = 0644,
5174                         .proc_handler   = proc_dointvec,
5175                 },
5176 #endif
5177 #endif
5178                 {
5179                         .procname       = "proxy_ndp",
5180                         .data           = &ipv6_devconf.proxy_ndp,
5181                         .maxlen         = sizeof(int),
5182                         .mode           = 0644,
5183                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5184                 },
5185                 {
5186                         .procname       = "accept_source_route",
5187                         .data           = &ipv6_devconf.accept_source_route,
5188                         .maxlen         = sizeof(int),
5189                         .mode           = 0644,
5190                         .proc_handler   = proc_dointvec,
5191                 },
5192 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5193                 {
5194                         .procname       = "optimistic_dad",
5195                         .data           = &ipv6_devconf.optimistic_dad,
5196                         .maxlen         = sizeof(int),
5197                         .mode           = 0644,
5198                         .proc_handler   = proc_dointvec,
5199
5200                 },
5201                 {
5202                         .procname       = "use_optimistic",
5203                         .data           = &ipv6_devconf.use_optimistic,
5204                         .maxlen         = sizeof(int),
5205                         .mode           = 0644,
5206                         .proc_handler   = proc_dointvec,
5207
5208                 },
5209 #endif
5210 #ifdef CONFIG_IPV6_MROUTE
5211                 {
5212                         .procname       = "mc_forwarding",
5213                         .data           = &ipv6_devconf.mc_forwarding,
5214                         .maxlen         = sizeof(int),
5215                         .mode           = 0444,
5216                         .proc_handler   = proc_dointvec,
5217                 },
5218 #endif
5219                 {
5220                         .procname       = "disable_ipv6",
5221                         .data           = &ipv6_devconf.disable_ipv6,
5222                         .maxlen         = sizeof(int),
5223                         .mode           = 0644,
5224                         .proc_handler   = addrconf_sysctl_disable,
5225                 },
5226                 {
5227                         .procname       = "accept_dad",
5228                         .data           = &ipv6_devconf.accept_dad,
5229                         .maxlen         = sizeof(int),
5230                         .mode           = 0644,
5231                         .proc_handler   = proc_dointvec,
5232                 },
5233                 {
5234                         .procname       = "force_tllao",
5235                         .data           = &ipv6_devconf.force_tllao,
5236                         .maxlen         = sizeof(int),
5237                         .mode           = 0644,
5238                         .proc_handler   = proc_dointvec
5239                 },
5240                 {
5241                         .procname       = "ndisc_notify",
5242                         .data           = &ipv6_devconf.ndisc_notify,
5243                         .maxlen         = sizeof(int),
5244                         .mode           = 0644,
5245                         .proc_handler   = proc_dointvec
5246                 },
5247                 {
5248                         .procname       = "suppress_frag_ndisc",
5249                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5250                         .maxlen         = sizeof(int),
5251                         .mode           = 0644,
5252                         .proc_handler   = proc_dointvec
5253                 },
5254                 {
5255                         .procname       = "accept_ra_from_local",
5256                         .data           = &ipv6_devconf.accept_ra_from_local,
5257                         .maxlen         = sizeof(int),
5258                         .mode           = 0644,
5259                         .proc_handler   = proc_dointvec,
5260                 },
5261                 {
5262                         /* sentinel */
5263                 }
5264         },
5265 };
5266
5267 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5268                 struct inet6_dev *idev, struct ipv6_devconf *p)
5269 {
5270         int i;
5271         struct addrconf_sysctl_table *t;
5272         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5273
5274         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5275         if (t == NULL)
5276                 goto out;
5277
5278         for (i = 0; t->addrconf_vars[i].data; i++) {
5279                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5280                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5281                 t->addrconf_vars[i].extra2 = net;
5282         }
5283
5284         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5285
5286         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5287         if (t->sysctl_header == NULL)
5288                 goto free;
5289
5290         p->sysctl = t;
5291         return 0;
5292
5293 free:
5294         kfree(t);
5295 out:
5296         return -ENOBUFS;
5297 }
5298
5299 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5300 {
5301         struct addrconf_sysctl_table *t;
5302
5303         if (p->sysctl == NULL)
5304                 return;
5305
5306         t = p->sysctl;
5307         p->sysctl = NULL;
5308         unregister_net_sysctl_table(t->sysctl_header);
5309         kfree(t);
5310 }
5311
5312 static int addrconf_sysctl_register(struct inet6_dev *idev)
5313 {
5314         int err;
5315
5316         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5317                 return -EINVAL;
5318
5319         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5320                                     &ndisc_ifinfo_sysctl_change);
5321         if (err)
5322                 return err;
5323         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5324                                          idev, &idev->cnf);
5325         if (err)
5326                 neigh_sysctl_unregister(idev->nd_parms);
5327
5328         return err;
5329 }
5330
5331 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5332 {
5333         __addrconf_sysctl_unregister(&idev->cnf);
5334         neigh_sysctl_unregister(idev->nd_parms);
5335 }
5336
5337
5338 #endif
5339
5340 static int __net_init addrconf_init_net(struct net *net)
5341 {
5342         int err = -ENOMEM;
5343         struct ipv6_devconf *all, *dflt;
5344
5345         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5346         if (all == NULL)
5347                 goto err_alloc_all;
5348
5349         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5350         if (dflt == NULL)
5351                 goto err_alloc_dflt;
5352
5353         /* these will be inherited by all namespaces */
5354         dflt->autoconf = ipv6_defaults.autoconf;
5355         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5356
5357         net->ipv6.devconf_all = all;
5358         net->ipv6.devconf_dflt = dflt;
5359
5360 #ifdef CONFIG_SYSCTL
5361         err = __addrconf_sysctl_register(net, "all", NULL, all);
5362         if (err < 0)
5363                 goto err_reg_all;
5364
5365         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5366         if (err < 0)
5367                 goto err_reg_dflt;
5368 #endif
5369         return 0;
5370
5371 #ifdef CONFIG_SYSCTL
5372 err_reg_dflt:
5373         __addrconf_sysctl_unregister(all);
5374 err_reg_all:
5375         kfree(dflt);
5376 #endif
5377 err_alloc_dflt:
5378         kfree(all);
5379 err_alloc_all:
5380         return err;
5381 }
5382
5383 static void __net_exit addrconf_exit_net(struct net *net)
5384 {
5385 #ifdef CONFIG_SYSCTL
5386         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5387         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5388 #endif
5389         kfree(net->ipv6.devconf_dflt);
5390         kfree(net->ipv6.devconf_all);
5391 }
5392
5393 static struct pernet_operations addrconf_ops = {
5394         .init = addrconf_init_net,
5395         .exit = addrconf_exit_net,
5396 };
5397
5398 static struct rtnl_af_ops inet6_ops = {
5399         .family           = AF_INET6,
5400         .fill_link_af     = inet6_fill_link_af,
5401         .get_link_af_size = inet6_get_link_af_size,
5402         .set_link_af      = inet6_set_link_af,
5403 };
5404
5405 /*
5406  *      Init / cleanup code
5407  */
5408
5409 int __init addrconf_init(void)
5410 {
5411         struct inet6_dev *idev;
5412         int i, err;
5413
5414         err = ipv6_addr_label_init();
5415         if (err < 0) {
5416                 pr_crit("%s: cannot initialize default policy table: %d\n",
5417                         __func__, err);
5418                 goto out;
5419         }
5420
5421         err = register_pernet_subsys(&addrconf_ops);
5422         if (err < 0)
5423                 goto out_addrlabel;
5424
5425         addrconf_wq = create_workqueue("ipv6_addrconf");
5426         if (!addrconf_wq) {
5427                 err = -ENOMEM;
5428                 goto out_nowq;
5429         }
5430
5431         /* The addrconf netdev notifier requires that loopback_dev
5432          * has it's ipv6 private information allocated and setup
5433          * before it can bring up and give link-local addresses
5434          * to other devices which are up.
5435          *
5436          * Unfortunately, loopback_dev is not necessarily the first
5437          * entry in the global dev_base list of net devices.  In fact,
5438          * it is likely to be the very last entry on that list.
5439          * So this causes the notifier registry below to try and
5440          * give link-local addresses to all devices besides loopback_dev
5441          * first, then loopback_dev, which cases all the non-loopback_dev
5442          * devices to fail to get a link-local address.
5443          *
5444          * So, as a temporary fix, allocate the ipv6 structure for
5445          * loopback_dev first by hand.
5446          * Longer term, all of the dependencies ipv6 has upon the loopback
5447          * device and it being up should be removed.
5448          */
5449         rtnl_lock();
5450         idev = ipv6_add_dev(init_net.loopback_dev);
5451         rtnl_unlock();
5452         if (IS_ERR(idev)) {
5453                 err = PTR_ERR(idev);
5454                 goto errlo;
5455         }
5456
5457         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5458                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5459
5460         register_netdevice_notifier(&ipv6_dev_notf);
5461
5462         addrconf_verify();
5463
5464         rtnl_af_register(&inet6_ops);
5465
5466         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5467                               NULL);
5468         if (err < 0)
5469                 goto errout;
5470
5471         /* Only the first call to __rtnl_register can fail */
5472         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5473         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5474         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5475                         inet6_dump_ifaddr, NULL);
5476         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5477                         inet6_dump_ifmcaddr, NULL);
5478         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5479                         inet6_dump_ifacaddr, NULL);
5480         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5481                         inet6_netconf_dump_devconf, NULL);
5482
5483         ipv6_addr_label_rtnl_register();
5484
5485         return 0;
5486 errout:
5487         rtnl_af_unregister(&inet6_ops);
5488         unregister_netdevice_notifier(&ipv6_dev_notf);
5489 errlo:
5490         destroy_workqueue(addrconf_wq);
5491 out_nowq:
5492         unregister_pernet_subsys(&addrconf_ops);
5493 out_addrlabel:
5494         ipv6_addr_label_cleanup();
5495 out:
5496         return err;
5497 }
5498
5499 void addrconf_cleanup(void)
5500 {
5501         struct net_device *dev;
5502         int i;
5503
5504         unregister_netdevice_notifier(&ipv6_dev_notf);
5505         unregister_pernet_subsys(&addrconf_ops);
5506         ipv6_addr_label_cleanup();
5507
5508         rtnl_lock();
5509
5510         __rtnl_af_unregister(&inet6_ops);
5511
5512         /* clean dev list */
5513         for_each_netdev(&init_net, dev) {
5514                 if (__in6_dev_get(dev) == NULL)
5515                         continue;
5516                 addrconf_ifdown(dev, 1);
5517         }
5518         addrconf_ifdown(init_net.loopback_dev, 2);
5519
5520         /*
5521          *      Check hash table.
5522          */
5523         spin_lock_bh(&addrconf_hash_lock);
5524         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5525                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5526         spin_unlock_bh(&addrconf_hash_lock);
5527         cancel_delayed_work(&addr_chk_work);
5528         rtnl_unlock();
5529
5530         destroy_workqueue(addrconf_wq);
5531 }