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