Upgrade to 4.4.50-rt62
[kvmfornfv.git] / kernel / 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
354         if (ndev->cnf.stable_secret.initialized)
355                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
356         else
357                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
358
359         ndev->cnf.mtu6 = dev->mtu;
360         ndev->cnf.sysctl = NULL;
361         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
362         if (!ndev->nd_parms) {
363                 kfree(ndev);
364                 return ERR_PTR(err);
365         }
366         if (ndev->cnf.forwarding)
367                 dev_disable_lro(dev);
368         /* We refer to the device */
369         dev_hold(dev);
370
371         if (snmp6_alloc_dev(ndev) < 0) {
372                 ADBG(KERN_WARNING
373                         "%s: cannot allocate memory for statistics; dev=%s.\n",
374                         __func__, dev->name);
375                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
376                 dev_put(dev);
377                 kfree(ndev);
378                 return ERR_PTR(err);
379         }
380
381         if (snmp6_register_dev(ndev) < 0) {
382                 ADBG(KERN_WARNING
383                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
384                         __func__, dev->name);
385                 goto err_release;
386         }
387
388         /* One reference from device.  We must do this before
389          * we invoke __ipv6_regen_rndid().
390          */
391         in6_dev_hold(ndev);
392
393         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
394                 ndev->cnf.accept_dad = -1;
395
396 #if IS_ENABLED(CONFIG_IPV6_SIT)
397         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
398                 pr_info("%s: Disabled Multicast RS\n", dev->name);
399                 ndev->cnf.rtr_solicits = 0;
400         }
401 #endif
402
403         INIT_LIST_HEAD(&ndev->tempaddr_list);
404         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
405         if ((dev->flags&IFF_LOOPBACK) ||
406             dev->type == ARPHRD_TUNNEL ||
407             dev->type == ARPHRD_TUNNEL6 ||
408             dev->type == ARPHRD_SIT ||
409             dev->type == ARPHRD_NONE) {
410                 ndev->cnf.use_tempaddr = -1;
411         } else {
412                 in6_dev_hold(ndev);
413                 ipv6_regen_rndid((unsigned long) ndev);
414         }
415
416         ndev->token = in6addr_any;
417
418         if (netif_running(dev) && addrconf_qdisc_ok(dev))
419                 ndev->if_flags |= IF_READY;
420
421         ipv6_mc_init_dev(ndev);
422         ndev->tstamp = jiffies;
423         err = addrconf_sysctl_register(ndev);
424         if (err) {
425                 ipv6_mc_destroy_dev(ndev);
426                 del_timer(&ndev->regen_timer);
427                 snmp6_unregister_dev(ndev);
428                 goto err_release;
429         }
430         /* protected by rtnl_lock */
431         rcu_assign_pointer(dev->ip6_ptr, ndev);
432
433         /* Join interface-local all-node multicast group */
434         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
435
436         /* Join all-node multicast group */
437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
438
439         /* Join all-router multicast group if forwarding is set */
440         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442
443         return ndev;
444
445 err_release:
446         neigh_parms_release(&nd_tbl, ndev->nd_parms);
447         ndev->dead = 1;
448         in6_dev_finish_destroy(ndev);
449         return ERR_PTR(err);
450 }
451
452 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
453 {
454         struct inet6_dev *idev;
455
456         ASSERT_RTNL();
457
458         idev = __in6_dev_get(dev);
459         if (!idev) {
460                 idev = ipv6_add_dev(dev);
461                 if (IS_ERR(idev))
462                         return NULL;
463         }
464
465         if (dev->flags&IFF_UP)
466                 ipv6_mc_up(idev);
467         return idev;
468 }
469
470 static int inet6_netconf_msgsize_devconf(int type)
471 {
472         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
473                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
474
475         /* type -1 is used for ALL */
476         if (type == -1 || type == NETCONFA_FORWARDING)
477                 size += nla_total_size(4);
478 #ifdef CONFIG_IPV6_MROUTE
479         if (type == -1 || type == NETCONFA_MC_FORWARDING)
480                 size += nla_total_size(4);
481 #endif
482         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
483                 size += nla_total_size(4);
484
485         if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
486                 size += nla_total_size(4);
487
488         return size;
489 }
490
491 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
492                                       struct ipv6_devconf *devconf, u32 portid,
493                                       u32 seq, int event, unsigned int flags,
494                                       int type)
495 {
496         struct nlmsghdr  *nlh;
497         struct netconfmsg *ncm;
498
499         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
500                         flags);
501         if (!nlh)
502                 return -EMSGSIZE;
503
504         ncm = nlmsg_data(nlh);
505         ncm->ncm_family = AF_INET6;
506
507         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
508                 goto nla_put_failure;
509
510         /* type -1 is used for ALL */
511         if ((type == -1 || type == NETCONFA_FORWARDING) &&
512             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
513                 goto nla_put_failure;
514 #ifdef CONFIG_IPV6_MROUTE
515         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
516             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
517                         devconf->mc_forwarding) < 0)
518                 goto nla_put_failure;
519 #endif
520         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
521             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
522                 goto nla_put_failure;
523
524         if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
525             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
526                         devconf->ignore_routes_with_linkdown) < 0)
527                 goto nla_put_failure;
528
529         nlmsg_end(skb, nlh);
530         return 0;
531
532 nla_put_failure:
533         nlmsg_cancel(skb, nlh);
534         return -EMSGSIZE;
535 }
536
537 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
538                                   struct ipv6_devconf *devconf)
539 {
540         struct sk_buff *skb;
541         int err = -ENOBUFS;
542
543         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
544         if (!skb)
545                 goto errout;
546
547         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
548                                          RTM_NEWNETCONF, 0, type);
549         if (err < 0) {
550                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
551                 WARN_ON(err == -EMSGSIZE);
552                 kfree_skb(skb);
553                 goto errout;
554         }
555         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
556         return;
557 errout:
558         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
559 }
560
561 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
562         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
563         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
564         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
565         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
566 };
567
568 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
569                                      struct nlmsghdr *nlh)
570 {
571         struct net *net = sock_net(in_skb->sk);
572         struct nlattr *tb[NETCONFA_MAX+1];
573         struct netconfmsg *ncm;
574         struct sk_buff *skb;
575         struct ipv6_devconf *devconf;
576         struct inet6_dev *in6_dev;
577         struct net_device *dev;
578         int ifindex;
579         int err;
580
581         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
582                           devconf_ipv6_policy);
583         if (err < 0)
584                 goto errout;
585
586         err = -EINVAL;
587         if (!tb[NETCONFA_IFINDEX])
588                 goto errout;
589
590         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
591         switch (ifindex) {
592         case NETCONFA_IFINDEX_ALL:
593                 devconf = net->ipv6.devconf_all;
594                 break;
595         case NETCONFA_IFINDEX_DEFAULT:
596                 devconf = net->ipv6.devconf_dflt;
597                 break;
598         default:
599                 dev = __dev_get_by_index(net, ifindex);
600                 if (!dev)
601                         goto errout;
602                 in6_dev = __in6_dev_get(dev);
603                 if (!in6_dev)
604                         goto errout;
605                 devconf = &in6_dev->cnf;
606                 break;
607         }
608
609         err = -ENOBUFS;
610         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
611         if (!skb)
612                 goto errout;
613
614         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
615                                          NETLINK_CB(in_skb).portid,
616                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
617                                          -1);
618         if (err < 0) {
619                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
620                 WARN_ON(err == -EMSGSIZE);
621                 kfree_skb(skb);
622                 goto errout;
623         }
624         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
625 errout:
626         return err;
627 }
628
629 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
630                                       struct netlink_callback *cb)
631 {
632         struct net *net = sock_net(skb->sk);
633         int h, s_h;
634         int idx, s_idx;
635         struct net_device *dev;
636         struct inet6_dev *idev;
637         struct hlist_head *head;
638
639         s_h = cb->args[0];
640         s_idx = idx = cb->args[1];
641
642         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
643                 idx = 0;
644                 head = &net->dev_index_head[h];
645                 rcu_read_lock();
646                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
647                           net->dev_base_seq;
648                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
649                         if (idx < s_idx)
650                                 goto cont;
651                         idev = __in6_dev_get(dev);
652                         if (!idev)
653                                 goto cont;
654
655                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
656                                                        &idev->cnf,
657                                                        NETLINK_CB(cb->skb).portid,
658                                                        cb->nlh->nlmsg_seq,
659                                                        RTM_NEWNETCONF,
660                                                        NLM_F_MULTI,
661                                                        -1) < 0) {
662                                 rcu_read_unlock();
663                                 goto done;
664                         }
665                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
666 cont:
667                         idx++;
668                 }
669                 rcu_read_unlock();
670         }
671         if (h == NETDEV_HASHENTRIES) {
672                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
673                                                net->ipv6.devconf_all,
674                                                NETLINK_CB(cb->skb).portid,
675                                                cb->nlh->nlmsg_seq,
676                                                RTM_NEWNETCONF, NLM_F_MULTI,
677                                                -1) < 0)
678                         goto done;
679                 else
680                         h++;
681         }
682         if (h == NETDEV_HASHENTRIES + 1) {
683                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
684                                                net->ipv6.devconf_dflt,
685                                                NETLINK_CB(cb->skb).portid,
686                                                cb->nlh->nlmsg_seq,
687                                                RTM_NEWNETCONF, NLM_F_MULTI,
688                                                -1) < 0)
689                         goto done;
690                 else
691                         h++;
692         }
693 done:
694         cb->args[0] = h;
695         cb->args[1] = idx;
696
697         return skb->len;
698 }
699
700 #ifdef CONFIG_SYSCTL
701 static void dev_forward_change(struct inet6_dev *idev)
702 {
703         struct net_device *dev;
704         struct inet6_ifaddr *ifa;
705
706         if (!idev)
707                 return;
708         dev = idev->dev;
709         if (idev->cnf.forwarding)
710                 dev_disable_lro(dev);
711         if (dev->flags & IFF_MULTICAST) {
712                 if (idev->cnf.forwarding) {
713                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
714                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
715                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
716                 } else {
717                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
718                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
719                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
720                 }
721         }
722
723         list_for_each_entry(ifa, &idev->addr_list, if_list) {
724                 if (ifa->flags&IFA_F_TENTATIVE)
725                         continue;
726                 if (idev->cnf.forwarding)
727                         addrconf_join_anycast(ifa);
728                 else
729                         addrconf_leave_anycast(ifa);
730         }
731         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
732                                      dev->ifindex, &idev->cnf);
733 }
734
735
736 static void addrconf_forward_change(struct net *net, __s32 newf)
737 {
738         struct net_device *dev;
739         struct inet6_dev *idev;
740
741         for_each_netdev(net, dev) {
742                 idev = __in6_dev_get(dev);
743                 if (idev) {
744                         int changed = (!idev->cnf.forwarding) ^ (!newf);
745                         idev->cnf.forwarding = newf;
746                         if (changed)
747                                 dev_forward_change(idev);
748                 }
749         }
750 }
751
752 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
753 {
754         struct net *net;
755         int old;
756
757         if (!rtnl_trylock())
758                 return restart_syscall();
759
760         net = (struct net *)table->extra2;
761         old = *p;
762         *p = newf;
763
764         if (p == &net->ipv6.devconf_dflt->forwarding) {
765                 if ((!newf) ^ (!old))
766                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
767                                                      NETCONFA_IFINDEX_DEFAULT,
768                                                      net->ipv6.devconf_dflt);
769                 rtnl_unlock();
770                 return 0;
771         }
772
773         if (p == &net->ipv6.devconf_all->forwarding) {
774                 net->ipv6.devconf_dflt->forwarding = newf;
775                 addrconf_forward_change(net, newf);
776                 if ((!newf) ^ (!old))
777                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
778                                                      NETCONFA_IFINDEX_ALL,
779                                                      net->ipv6.devconf_all);
780         } else if ((!newf) ^ (!old))
781                 dev_forward_change((struct inet6_dev *)table->extra1);
782         rtnl_unlock();
783
784         if (newf)
785                 rt6_purge_dflt_routers(net);
786         return 1;
787 }
788
789 static void addrconf_linkdown_change(struct net *net, __s32 newf)
790 {
791         struct net_device *dev;
792         struct inet6_dev *idev;
793
794         for_each_netdev(net, dev) {
795                 idev = __in6_dev_get(dev);
796                 if (idev) {
797                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
798
799                         idev->cnf.ignore_routes_with_linkdown = newf;
800                         if (changed)
801                                 inet6_netconf_notify_devconf(dev_net(dev),
802                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
803                                                              dev->ifindex,
804                                                              &idev->cnf);
805                 }
806         }
807 }
808
809 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
810 {
811         struct net *net;
812         int old;
813
814         if (!rtnl_trylock())
815                 return restart_syscall();
816
817         net = (struct net *)table->extra2;
818         old = *p;
819         *p = newf;
820
821         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
822                 if ((!newf) ^ (!old))
823                         inet6_netconf_notify_devconf(net,
824                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
825                                                      NETCONFA_IFINDEX_DEFAULT,
826                                                      net->ipv6.devconf_dflt);
827                 rtnl_unlock();
828                 return 0;
829         }
830
831         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
832                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
833                 addrconf_linkdown_change(net, newf);
834                 if ((!newf) ^ (!old))
835                         inet6_netconf_notify_devconf(net,
836                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
837                                                      NETCONFA_IFINDEX_ALL,
838                                                      net->ipv6.devconf_all);
839         }
840         rtnl_unlock();
841
842         return 1;
843 }
844
845 #endif
846
847 /* Nobody refers to this ifaddr, destroy it */
848 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
849 {
850         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
851
852 #ifdef NET_REFCNT_DEBUG
853         pr_debug("%s\n", __func__);
854 #endif
855
856         in6_dev_put(ifp->idev);
857
858         if (cancel_delayed_work(&ifp->dad_work))
859                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
860                           ifp);
861
862         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
863                 pr_warn("Freeing alive inet6 address %p\n", ifp);
864                 return;
865         }
866         ip6_rt_put(ifp->rt);
867
868         kfree_rcu(ifp, rcu);
869 }
870
871 static void
872 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
873 {
874         struct list_head *p;
875         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
876
877         /*
878          * Each device address list is sorted in order of scope -
879          * global before linklocal.
880          */
881         list_for_each(p, &idev->addr_list) {
882                 struct inet6_ifaddr *ifa
883                         = list_entry(p, struct inet6_ifaddr, if_list);
884                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
885                         break;
886         }
887
888         list_add_tail(&ifp->if_list, p);
889 }
890
891 static u32 inet6_addr_hash(const struct in6_addr *addr)
892 {
893         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
894 }
895
896 /* On success it returns ifp with increased reference count */
897
898 static struct inet6_ifaddr *
899 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
900               const struct in6_addr *peer_addr, int pfxlen,
901               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
902 {
903         struct inet6_ifaddr *ifa = NULL;
904         struct rt6_info *rt;
905         unsigned int hash;
906         int err = 0;
907         int addr_type = ipv6_addr_type(addr);
908
909         if (addr_type == IPV6_ADDR_ANY ||
910             addr_type & IPV6_ADDR_MULTICAST ||
911             (!(idev->dev->flags & IFF_LOOPBACK) &&
912              addr_type & IPV6_ADDR_LOOPBACK))
913                 return ERR_PTR(-EADDRNOTAVAIL);
914
915         rcu_read_lock_bh();
916         if (idev->dead) {
917                 err = -ENODEV;                  /*XXX*/
918                 goto out2;
919         }
920
921         if (idev->cnf.disable_ipv6) {
922                 err = -EACCES;
923                 goto out2;
924         }
925
926         spin_lock(&addrconf_hash_lock);
927
928         /* Ignore adding duplicate addresses on an interface */
929         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
930                 ADBG("ipv6_add_addr: already assigned\n");
931                 err = -EEXIST;
932                 goto out;
933         }
934
935         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
936
937         if (!ifa) {
938                 ADBG("ipv6_add_addr: malloc failed\n");
939                 err = -ENOBUFS;
940                 goto out;
941         }
942
943         rt = addrconf_dst_alloc(idev, addr, false);
944         if (IS_ERR(rt)) {
945                 err = PTR_ERR(rt);
946                 goto out;
947         }
948
949         neigh_parms_data_state_setall(idev->nd_parms);
950
951         ifa->addr = *addr;
952         if (peer_addr)
953                 ifa->peer_addr = *peer_addr;
954
955         spin_lock_init(&ifa->lock);
956         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
957         INIT_HLIST_NODE(&ifa->addr_lst);
958         ifa->scope = scope;
959         ifa->prefix_len = pfxlen;
960         ifa->flags = flags | IFA_F_TENTATIVE;
961         ifa->valid_lft = valid_lft;
962         ifa->prefered_lft = prefered_lft;
963         ifa->cstamp = ifa->tstamp = jiffies;
964         ifa->tokenized = false;
965
966         ifa->rt = rt;
967
968         ifa->idev = idev;
969         in6_dev_hold(idev);
970         /* For caller */
971         in6_ifa_hold(ifa);
972
973         /* Add to big hash table */
974         hash = inet6_addr_hash(addr);
975
976         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
977         spin_unlock(&addrconf_hash_lock);
978
979         write_lock(&idev->lock);
980         /* Add to inet6_dev unicast addr list. */
981         ipv6_link_dev_addr(idev, ifa);
982
983         if (ifa->flags&IFA_F_TEMPORARY) {
984                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
985                 in6_ifa_hold(ifa);
986         }
987
988         in6_ifa_hold(ifa);
989         write_unlock(&idev->lock);
990 out2:
991         rcu_read_unlock_bh();
992
993         if (likely(err == 0))
994                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
995         else {
996                 kfree(ifa);
997                 ifa = ERR_PTR(err);
998         }
999
1000         return ifa;
1001 out:
1002         spin_unlock(&addrconf_hash_lock);
1003         goto out2;
1004 }
1005
1006 enum cleanup_prefix_rt_t {
1007         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1008         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1009         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1010 };
1011
1012 /*
1013  * Check, whether the prefix for ifp would still need a prefix route
1014  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1015  * constants.
1016  *
1017  * 1) we don't purge prefix if address was not permanent.
1018  *    prefix is managed by its own lifetime.
1019  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1020  * 3) if there are no addresses, delete prefix.
1021  * 4) if there are still other permanent address(es),
1022  *    corresponding prefix is still permanent.
1023  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1024  *    don't purge the prefix, assume user space is managing it.
1025  * 6) otherwise, update prefix lifetime to the
1026  *    longest valid lifetime among the corresponding
1027  *    addresses on the device.
1028  *    Note: subsequent RA will update lifetime.
1029  **/
1030 static enum cleanup_prefix_rt_t
1031 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1032 {
1033         struct inet6_ifaddr *ifa;
1034         struct inet6_dev *idev = ifp->idev;
1035         unsigned long lifetime;
1036         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1037
1038         *expires = jiffies;
1039
1040         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1041                 if (ifa == ifp)
1042                         continue;
1043                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1044                                        ifp->prefix_len))
1045                         continue;
1046                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1047                         return CLEANUP_PREFIX_RT_NOP;
1048
1049                 action = CLEANUP_PREFIX_RT_EXPIRE;
1050
1051                 spin_lock(&ifa->lock);
1052
1053                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1054                 /*
1055                  * Note: Because this address is
1056                  * not permanent, lifetime <
1057                  * LONG_MAX / HZ here.
1058                  */
1059                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1060                         *expires = ifa->tstamp + lifetime * HZ;
1061                 spin_unlock(&ifa->lock);
1062         }
1063
1064         return action;
1065 }
1066
1067 static void
1068 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1069 {
1070         struct rt6_info *rt;
1071
1072         rt = addrconf_get_prefix_route(&ifp->addr,
1073                                        ifp->prefix_len,
1074                                        ifp->idev->dev,
1075                                        0, RTF_GATEWAY | RTF_DEFAULT);
1076         if (rt) {
1077                 if (del_rt)
1078                         ip6_del_rt(rt);
1079                 else {
1080                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1081                                 rt6_set_expires(rt, expires);
1082                         ip6_rt_put(rt);
1083                 }
1084         }
1085 }
1086
1087
1088 /* This function wants to get referenced ifp and releases it before return */
1089
1090 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1091 {
1092         int state;
1093         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1094         unsigned long expires;
1095
1096         ASSERT_RTNL();
1097
1098         spin_lock_bh(&ifp->lock);
1099         state = ifp->state;
1100         ifp->state = INET6_IFADDR_STATE_DEAD;
1101         spin_unlock_bh(&ifp->lock);
1102
1103         if (state == INET6_IFADDR_STATE_DEAD)
1104                 goto out;
1105
1106         spin_lock_bh(&addrconf_hash_lock);
1107         hlist_del_init_rcu(&ifp->addr_lst);
1108         spin_unlock_bh(&addrconf_hash_lock);
1109
1110         write_lock_bh(&ifp->idev->lock);
1111
1112         if (ifp->flags&IFA_F_TEMPORARY) {
1113                 list_del(&ifp->tmp_list);
1114                 if (ifp->ifpub) {
1115                         in6_ifa_put(ifp->ifpub);
1116                         ifp->ifpub = NULL;
1117                 }
1118                 __in6_ifa_put(ifp);
1119         }
1120
1121         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1122                 action = check_cleanup_prefix_route(ifp, &expires);
1123
1124         list_del_init(&ifp->if_list);
1125         __in6_ifa_put(ifp);
1126
1127         write_unlock_bh(&ifp->idev->lock);
1128
1129         addrconf_del_dad_work(ifp);
1130
1131         ipv6_ifa_notify(RTM_DELADDR, ifp);
1132
1133         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1134
1135         if (action != CLEANUP_PREFIX_RT_NOP) {
1136                 cleanup_prefix_route(ifp, expires,
1137                         action == CLEANUP_PREFIX_RT_DEL);
1138         }
1139
1140         /* clean up prefsrc entries */
1141         rt6_remove_prefsrc(ifp);
1142 out:
1143         in6_ifa_put(ifp);
1144 }
1145
1146 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1147 {
1148         struct inet6_dev *idev = ifp->idev;
1149         struct in6_addr addr, *tmpaddr;
1150         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1151         unsigned long regen_advance;
1152         int tmp_plen;
1153         int ret = 0;
1154         u32 addr_flags;
1155         unsigned long now = jiffies;
1156
1157         write_lock_bh(&idev->lock);
1158         if (ift) {
1159                 spin_lock_bh(&ift->lock);
1160                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1161                 spin_unlock_bh(&ift->lock);
1162                 tmpaddr = &addr;
1163         } else {
1164                 tmpaddr = NULL;
1165         }
1166 retry:
1167         in6_dev_hold(idev);
1168         if (idev->cnf.use_tempaddr <= 0) {
1169                 write_unlock_bh(&idev->lock);
1170                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1171                 in6_dev_put(idev);
1172                 ret = -1;
1173                 goto out;
1174         }
1175         spin_lock_bh(&ifp->lock);
1176         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1177                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1178                 spin_unlock_bh(&ifp->lock);
1179                 write_unlock_bh(&idev->lock);
1180                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1181                         __func__);
1182                 in6_dev_put(idev);
1183                 ret = -1;
1184                 goto out;
1185         }
1186         in6_ifa_hold(ifp);
1187         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1188         __ipv6_try_regen_rndid(idev, tmpaddr);
1189         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1190         age = (now - ifp->tstamp) / HZ;
1191         tmp_valid_lft = min_t(__u32,
1192                               ifp->valid_lft,
1193                               idev->cnf.temp_valid_lft + age);
1194         tmp_prefered_lft = min_t(__u32,
1195                                  ifp->prefered_lft,
1196                                  idev->cnf.temp_prefered_lft + age -
1197                                  idev->cnf.max_desync_factor);
1198         tmp_plen = ifp->prefix_len;
1199         tmp_tstamp = ifp->tstamp;
1200         spin_unlock_bh(&ifp->lock);
1201
1202         regen_advance = idev->cnf.regen_max_retry *
1203                         idev->cnf.dad_transmits *
1204                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1205         write_unlock_bh(&idev->lock);
1206
1207         /* A temporary address is created only if this calculated Preferred
1208          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1209          * an implementation must not create a temporary address with a zero
1210          * Preferred Lifetime.
1211          * Use age calculation as in addrconf_verify to avoid unnecessary
1212          * temporary addresses being generated.
1213          */
1214         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1215         if (tmp_prefered_lft <= regen_advance + age) {
1216                 in6_ifa_put(ifp);
1217                 in6_dev_put(idev);
1218                 ret = -1;
1219                 goto out;
1220         }
1221
1222         addr_flags = IFA_F_TEMPORARY;
1223         /* set in addrconf_prefix_rcv() */
1224         if (ifp->flags & IFA_F_OPTIMISTIC)
1225                 addr_flags |= IFA_F_OPTIMISTIC;
1226
1227         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1228                             ipv6_addr_scope(&addr), addr_flags,
1229                             tmp_valid_lft, tmp_prefered_lft);
1230         if (IS_ERR(ift)) {
1231                 in6_ifa_put(ifp);
1232                 in6_dev_put(idev);
1233                 pr_info("%s: retry temporary address regeneration\n", __func__);
1234                 tmpaddr = &addr;
1235                 write_lock_bh(&idev->lock);
1236                 goto retry;
1237         }
1238
1239         spin_lock_bh(&ift->lock);
1240         ift->ifpub = ifp;
1241         ift->cstamp = now;
1242         ift->tstamp = tmp_tstamp;
1243         spin_unlock_bh(&ift->lock);
1244
1245         addrconf_dad_start(ift);
1246         in6_ifa_put(ift);
1247         in6_dev_put(idev);
1248 out:
1249         return ret;
1250 }
1251
1252 /*
1253  *      Choose an appropriate source address (RFC3484)
1254  */
1255 enum {
1256         IPV6_SADDR_RULE_INIT = 0,
1257         IPV6_SADDR_RULE_LOCAL,
1258         IPV6_SADDR_RULE_SCOPE,
1259         IPV6_SADDR_RULE_PREFERRED,
1260 #ifdef CONFIG_IPV6_MIP6
1261         IPV6_SADDR_RULE_HOA,
1262 #endif
1263         IPV6_SADDR_RULE_OIF,
1264         IPV6_SADDR_RULE_LABEL,
1265         IPV6_SADDR_RULE_PRIVACY,
1266         IPV6_SADDR_RULE_ORCHID,
1267         IPV6_SADDR_RULE_PREFIX,
1268 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1269         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1270 #endif
1271         IPV6_SADDR_RULE_MAX
1272 };
1273
1274 struct ipv6_saddr_score {
1275         int                     rule;
1276         int                     addr_type;
1277         struct inet6_ifaddr     *ifa;
1278         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1279         int                     scopedist;
1280         int                     matchlen;
1281 };
1282
1283 struct ipv6_saddr_dst {
1284         const struct in6_addr *addr;
1285         int ifindex;
1286         int scope;
1287         int label;
1288         unsigned int prefs;
1289 };
1290
1291 static inline int ipv6_saddr_preferred(int type)
1292 {
1293         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1294                 return 1;
1295         return 0;
1296 }
1297
1298 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1299 {
1300 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1301         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1302 #else
1303         return false;
1304 #endif
1305 }
1306
1307 static int ipv6_get_saddr_eval(struct net *net,
1308                                struct ipv6_saddr_score *score,
1309                                struct ipv6_saddr_dst *dst,
1310                                int i)
1311 {
1312         int ret;
1313
1314         if (i <= score->rule) {
1315                 switch (i) {
1316                 case IPV6_SADDR_RULE_SCOPE:
1317                         ret = score->scopedist;
1318                         break;
1319                 case IPV6_SADDR_RULE_PREFIX:
1320                         ret = score->matchlen;
1321                         break;
1322                 default:
1323                         ret = !!test_bit(i, score->scorebits);
1324                 }
1325                 goto out;
1326         }
1327
1328         switch (i) {
1329         case IPV6_SADDR_RULE_INIT:
1330                 /* Rule 0: remember if hiscore is not ready yet */
1331                 ret = !!score->ifa;
1332                 break;
1333         case IPV6_SADDR_RULE_LOCAL:
1334                 /* Rule 1: Prefer same address */
1335                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1336                 break;
1337         case IPV6_SADDR_RULE_SCOPE:
1338                 /* Rule 2: Prefer appropriate scope
1339                  *
1340                  *      ret
1341                  *       ^
1342                  *    -1 |  d 15
1343                  *    ---+--+-+---> scope
1344                  *       |
1345                  *       |             d is scope of the destination.
1346                  *  B-d  |  \
1347                  *       |   \      <- smaller scope is better if
1348                  *  B-15 |    \        if scope is enough for destination.
1349                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1350                  * d-C-1 | /
1351                  *       |/         <- greater is better
1352                  *   -C  /             if scope is not enough for destination.
1353                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1354                  *
1355                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1356                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1357                  * Assume B = 0 and we get C > 29.
1358                  */
1359                 ret = __ipv6_addr_src_scope(score->addr_type);
1360                 if (ret >= dst->scope)
1361                         ret = -ret;
1362                 else
1363                         ret -= 128;     /* 30 is enough */
1364                 score->scopedist = ret;
1365                 break;
1366         case IPV6_SADDR_RULE_PREFERRED:
1367             {
1368                 /* Rule 3: Avoid deprecated and optimistic addresses */
1369                 u8 avoid = IFA_F_DEPRECATED;
1370
1371                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1372                         avoid |= IFA_F_OPTIMISTIC;
1373                 ret = ipv6_saddr_preferred(score->addr_type) ||
1374                       !(score->ifa->flags & avoid);
1375                 break;
1376             }
1377 #ifdef CONFIG_IPV6_MIP6
1378         case IPV6_SADDR_RULE_HOA:
1379             {
1380                 /* Rule 4: Prefer home address */
1381                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1382                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1383                 break;
1384             }
1385 #endif
1386         case IPV6_SADDR_RULE_OIF:
1387                 /* Rule 5: Prefer outgoing interface */
1388                 ret = (!dst->ifindex ||
1389                        dst->ifindex == score->ifa->idev->dev->ifindex);
1390                 break;
1391         case IPV6_SADDR_RULE_LABEL:
1392                 /* Rule 6: Prefer matching label */
1393                 ret = ipv6_addr_label(net,
1394                                       &score->ifa->addr, score->addr_type,
1395                                       score->ifa->idev->dev->ifindex) == dst->label;
1396                 break;
1397         case IPV6_SADDR_RULE_PRIVACY:
1398             {
1399                 /* Rule 7: Prefer public address
1400                  * Note: prefer temporary address if use_tempaddr >= 2
1401                  */
1402                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1403                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1404                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1405                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1406                 break;
1407             }
1408         case IPV6_SADDR_RULE_ORCHID:
1409                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1410                  *          non-ORCHID vs non-ORCHID
1411                  */
1412                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1413                         ipv6_addr_orchid(dst->addr));
1414                 break;
1415         case IPV6_SADDR_RULE_PREFIX:
1416                 /* Rule 8: Use longest matching prefix */
1417                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1418                 if (ret > score->ifa->prefix_len)
1419                         ret = score->ifa->prefix_len;
1420                 score->matchlen = ret;
1421                 break;
1422 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1423         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1424                 /* Optimistic addresses still have lower precedence than other
1425                  * preferred addresses.
1426                  */
1427                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1428                 break;
1429 #endif
1430         default:
1431                 ret = 0;
1432         }
1433
1434         if (ret)
1435                 __set_bit(i, score->scorebits);
1436         score->rule = i;
1437 out:
1438         return ret;
1439 }
1440
1441 static int __ipv6_dev_get_saddr(struct net *net,
1442                                 struct ipv6_saddr_dst *dst,
1443                                 struct inet6_dev *idev,
1444                                 struct ipv6_saddr_score *scores,
1445                                 int hiscore_idx)
1446 {
1447         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1448
1449         read_lock_bh(&idev->lock);
1450         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1451                 int i;
1452
1453                 /*
1454                  * - Tentative Address (RFC2462 section 5.4)
1455                  *  - A tentative address is not considered
1456                  *    "assigned to an interface" in the traditional
1457                  *    sense, unless it is also flagged as optimistic.
1458                  * - Candidate Source Address (section 4)
1459                  *  - In any case, anycast addresses, multicast
1460                  *    addresses, and the unspecified address MUST
1461                  *    NOT be included in a candidate set.
1462                  */
1463                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1464                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1465                         continue;
1466
1467                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1468
1469                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1470                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1471                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1472                                             idev->dev->name);
1473                         continue;
1474                 }
1475
1476                 score->rule = -1;
1477                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1478
1479                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1480                         int minihiscore, miniscore;
1481
1482                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1483                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1484
1485                         if (minihiscore > miniscore) {
1486                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1487                                     score->scopedist > 0) {
1488                                         /*
1489                                          * special case:
1490                                          * each remaining entry
1491                                          * has too small (not enough)
1492                                          * scope, because ifa entries
1493                                          * are sorted by their scope
1494                                          * values.
1495                                          */
1496                                         goto out;
1497                                 }
1498                                 break;
1499                         } else if (minihiscore < miniscore) {
1500                                 if (hiscore->ifa)
1501                                         in6_ifa_put(hiscore->ifa);
1502
1503                                 in6_ifa_hold(score->ifa);
1504
1505                                 swap(hiscore, score);
1506                                 hiscore_idx = 1 - hiscore_idx;
1507
1508                                 /* restore our iterator */
1509                                 score->ifa = hiscore->ifa;
1510
1511                                 break;
1512                         }
1513                 }
1514         }
1515 out:
1516         read_unlock_bh(&idev->lock);
1517         return hiscore_idx;
1518 }
1519
1520 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1521                        const struct in6_addr *daddr, unsigned int prefs,
1522                        struct in6_addr *saddr)
1523 {
1524         struct ipv6_saddr_score scores[2], *hiscore;
1525         struct ipv6_saddr_dst dst;
1526         struct inet6_dev *idev;
1527         struct net_device *dev;
1528         int dst_type;
1529         bool use_oif_addr = false;
1530         int hiscore_idx = 0;
1531
1532         dst_type = __ipv6_addr_type(daddr);
1533         dst.addr = daddr;
1534         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1535         dst.scope = __ipv6_addr_src_scope(dst_type);
1536         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1537         dst.prefs = prefs;
1538
1539         scores[hiscore_idx].rule = -1;
1540         scores[hiscore_idx].ifa = NULL;
1541
1542         rcu_read_lock();
1543
1544         /* Candidate Source Address (section 4)
1545          *  - multicast and link-local destination address,
1546          *    the set of candidate source address MUST only
1547          *    include addresses assigned to interfaces
1548          *    belonging to the same link as the outgoing
1549          *    interface.
1550          * (- For site-local destination addresses, the
1551          *    set of candidate source addresses MUST only
1552          *    include addresses assigned to interfaces
1553          *    belonging to the same site as the outgoing
1554          *    interface.)
1555          *  - "It is RECOMMENDED that the candidate source addresses
1556          *    be the set of unicast addresses assigned to the
1557          *    interface that will be used to send to the destination
1558          *    (the 'outgoing' interface)." (RFC 6724)
1559          */
1560         if (dst_dev) {
1561                 idev = __in6_dev_get(dst_dev);
1562                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1563                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1564                     (idev && idev->cnf.use_oif_addrs_only)) {
1565                         use_oif_addr = true;
1566                 }
1567         }
1568
1569         if (use_oif_addr) {
1570                 if (idev)
1571                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1572         } else {
1573                 for_each_netdev_rcu(net, dev) {
1574                         idev = __in6_dev_get(dev);
1575                         if (!idev)
1576                                 continue;
1577                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1578                 }
1579         }
1580         rcu_read_unlock();
1581
1582         hiscore = &scores[hiscore_idx];
1583         if (!hiscore->ifa)
1584                 return -EADDRNOTAVAIL;
1585
1586         *saddr = hiscore->ifa->addr;
1587         in6_ifa_put(hiscore->ifa);
1588         return 0;
1589 }
1590 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1591
1592 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1593                       u32 banned_flags)
1594 {
1595         struct inet6_ifaddr *ifp;
1596         int err = -EADDRNOTAVAIL;
1597
1598         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1599                 if (ifp->scope > IFA_LINK)
1600                         break;
1601                 if (ifp->scope == IFA_LINK &&
1602                     !(ifp->flags & banned_flags)) {
1603                         *addr = ifp->addr;
1604                         err = 0;
1605                         break;
1606                 }
1607         }
1608         return err;
1609 }
1610
1611 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1612                     u32 banned_flags)
1613 {
1614         struct inet6_dev *idev;
1615         int err = -EADDRNOTAVAIL;
1616
1617         rcu_read_lock();
1618         idev = __in6_dev_get(dev);
1619         if (idev) {
1620                 read_lock_bh(&idev->lock);
1621                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1622                 read_unlock_bh(&idev->lock);
1623         }
1624         rcu_read_unlock();
1625         return err;
1626 }
1627
1628 static int ipv6_count_addresses(struct inet6_dev *idev)
1629 {
1630         int cnt = 0;
1631         struct inet6_ifaddr *ifp;
1632
1633         read_lock_bh(&idev->lock);
1634         list_for_each_entry(ifp, &idev->addr_list, if_list)
1635                 cnt++;
1636         read_unlock_bh(&idev->lock);
1637         return cnt;
1638 }
1639
1640 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1641                   const struct net_device *dev, int strict)
1642 {
1643         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1644 }
1645 EXPORT_SYMBOL(ipv6_chk_addr);
1646
1647 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1648                             const struct net_device *dev, int strict,
1649                             u32 banned_flags)
1650 {
1651         struct inet6_ifaddr *ifp;
1652         unsigned int hash = inet6_addr_hash(addr);
1653         u32 ifp_flags;
1654
1655         rcu_read_lock_bh();
1656         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1657                 if (!net_eq(dev_net(ifp->idev->dev), net))
1658                         continue;
1659                 /* Decouple optimistic from tentative for evaluation here.
1660                  * Ban optimistic addresses explicitly, when required.
1661                  */
1662                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1663                             ? (ifp->flags&~IFA_F_TENTATIVE)
1664                             : ifp->flags;
1665                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1666                     !(ifp_flags&banned_flags) &&
1667                     (!dev || ifp->idev->dev == dev ||
1668                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1669                         rcu_read_unlock_bh();
1670                         return 1;
1671                 }
1672         }
1673
1674         rcu_read_unlock_bh();
1675         return 0;
1676 }
1677 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1678
1679 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1680                                struct net_device *dev)
1681 {
1682         unsigned int hash = inet6_addr_hash(addr);
1683         struct inet6_ifaddr *ifp;
1684
1685         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1686                 if (!net_eq(dev_net(ifp->idev->dev), net))
1687                         continue;
1688                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1689                         if (!dev || ifp->idev->dev == dev)
1690                                 return true;
1691                 }
1692         }
1693         return false;
1694 }
1695
1696 /* Compares an address/prefix_len with addresses on device @dev.
1697  * If one is found it returns true.
1698  */
1699 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1700         const unsigned int prefix_len, struct net_device *dev)
1701 {
1702         struct inet6_dev *idev;
1703         struct inet6_ifaddr *ifa;
1704         bool ret = false;
1705
1706         rcu_read_lock();
1707         idev = __in6_dev_get(dev);
1708         if (idev) {
1709                 read_lock_bh(&idev->lock);
1710                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1711                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1712                         if (ret)
1713                                 break;
1714                 }
1715                 read_unlock_bh(&idev->lock);
1716         }
1717         rcu_read_unlock();
1718
1719         return ret;
1720 }
1721 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1722
1723 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1724 {
1725         struct inet6_dev *idev;
1726         struct inet6_ifaddr *ifa;
1727         int     onlink;
1728
1729         onlink = 0;
1730         rcu_read_lock();
1731         idev = __in6_dev_get(dev);
1732         if (idev) {
1733                 read_lock_bh(&idev->lock);
1734                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1735                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1736                                                    ifa->prefix_len);
1737                         if (onlink)
1738                                 break;
1739                 }
1740                 read_unlock_bh(&idev->lock);
1741         }
1742         rcu_read_unlock();
1743         return onlink;
1744 }
1745 EXPORT_SYMBOL(ipv6_chk_prefix);
1746
1747 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1748                                      struct net_device *dev, int strict)
1749 {
1750         struct inet6_ifaddr *ifp, *result = NULL;
1751         unsigned int hash = inet6_addr_hash(addr);
1752
1753         rcu_read_lock_bh();
1754         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1755                 if (!net_eq(dev_net(ifp->idev->dev), net))
1756                         continue;
1757                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1758                         if (!dev || ifp->idev->dev == dev ||
1759                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1760                                 result = ifp;
1761                                 in6_ifa_hold(ifp);
1762                                 break;
1763                         }
1764                 }
1765         }
1766         rcu_read_unlock_bh();
1767
1768         return result;
1769 }
1770
1771 /* Gets referenced address, destroys ifaddr */
1772
1773 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1774 {
1775         if (ifp->flags&IFA_F_PERMANENT) {
1776                 spin_lock_bh(&ifp->lock);
1777                 addrconf_del_dad_work(ifp);
1778                 ifp->flags |= IFA_F_TENTATIVE;
1779                 if (dad_failed)
1780                         ifp->flags |= IFA_F_DADFAILED;
1781                 spin_unlock_bh(&ifp->lock);
1782                 if (dad_failed)
1783                         ipv6_ifa_notify(0, ifp);
1784                 in6_ifa_put(ifp);
1785         } else if (ifp->flags&IFA_F_TEMPORARY) {
1786                 struct inet6_ifaddr *ifpub;
1787                 spin_lock_bh(&ifp->lock);
1788                 ifpub = ifp->ifpub;
1789                 if (ifpub) {
1790                         in6_ifa_hold(ifpub);
1791                         spin_unlock_bh(&ifp->lock);
1792                         ipv6_create_tempaddr(ifpub, ifp);
1793                         in6_ifa_put(ifpub);
1794                 } else {
1795                         spin_unlock_bh(&ifp->lock);
1796                 }
1797                 ipv6_del_addr(ifp);
1798         } else {
1799                 ipv6_del_addr(ifp);
1800         }
1801 }
1802
1803 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1804 {
1805         int err = -ENOENT;
1806
1807         spin_lock_bh(&ifp->lock);
1808         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1809                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1810                 err = 0;
1811         }
1812         spin_unlock_bh(&ifp->lock);
1813
1814         return err;
1815 }
1816
1817 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1818 {
1819         struct in6_addr addr;
1820         struct inet6_dev *idev = ifp->idev;
1821         struct net *net = dev_net(ifp->idev->dev);
1822
1823         if (addrconf_dad_end(ifp)) {
1824                 in6_ifa_put(ifp);
1825                 return;
1826         }
1827
1828         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1829                              ifp->idev->dev->name, &ifp->addr);
1830
1831         spin_lock_bh(&ifp->lock);
1832
1833         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1834                 int scope = ifp->scope;
1835                 u32 flags = ifp->flags;
1836                 struct in6_addr new_addr;
1837                 struct inet6_ifaddr *ifp2;
1838                 u32 valid_lft, preferred_lft;
1839                 int pfxlen = ifp->prefix_len;
1840                 int retries = ifp->stable_privacy_retry + 1;
1841
1842                 if (retries > net->ipv6.sysctl.idgen_retries) {
1843                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1844                                              ifp->idev->dev->name);
1845                         goto errdad;
1846                 }
1847
1848                 new_addr = ifp->addr;
1849                 if (ipv6_generate_stable_address(&new_addr, retries,
1850                                                  idev))
1851                         goto errdad;
1852
1853                 valid_lft = ifp->valid_lft;
1854                 preferred_lft = ifp->prefered_lft;
1855
1856                 spin_unlock_bh(&ifp->lock);
1857
1858                 if (idev->cnf.max_addresses &&
1859                     ipv6_count_addresses(idev) >=
1860                     idev->cnf.max_addresses)
1861                         goto lock_errdad;
1862
1863                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1864                                      ifp->idev->dev->name);
1865
1866                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1867                                      scope, flags, valid_lft,
1868                                      preferred_lft);
1869                 if (IS_ERR(ifp2))
1870                         goto lock_errdad;
1871
1872                 spin_lock_bh(&ifp2->lock);
1873                 ifp2->stable_privacy_retry = retries;
1874                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1875                 spin_unlock_bh(&ifp2->lock);
1876
1877                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1878                 in6_ifa_put(ifp2);
1879 lock_errdad:
1880                 spin_lock_bh(&ifp->lock);
1881         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1882                 addr.s6_addr32[0] = htonl(0xfe800000);
1883                 addr.s6_addr32[1] = 0;
1884
1885                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1886                     ipv6_addr_equal(&ifp->addr, &addr)) {
1887                         /* DAD failed for link-local based on MAC address */
1888                         idev->cnf.disable_ipv6 = 1;
1889
1890                         pr_info("%s: IPv6 being disabled!\n",
1891                                 ifp->idev->dev->name);
1892                 }
1893         }
1894
1895 errdad:
1896         /* transition from _POSTDAD to _ERRDAD */
1897         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1898         spin_unlock_bh(&ifp->lock);
1899
1900         addrconf_mod_dad_work(ifp, 0);
1901         in6_ifa_put(ifp);
1902 }
1903
1904 /* Join to solicited addr multicast group.
1905  * caller must hold RTNL */
1906 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1907 {
1908         struct in6_addr maddr;
1909
1910         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1911                 return;
1912
1913         addrconf_addr_solict_mult(addr, &maddr);
1914         ipv6_dev_mc_inc(dev, &maddr);
1915 }
1916
1917 /* caller must hold RTNL */
1918 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1919 {
1920         struct in6_addr maddr;
1921
1922         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1923                 return;
1924
1925         addrconf_addr_solict_mult(addr, &maddr);
1926         __ipv6_dev_mc_dec(idev, &maddr);
1927 }
1928
1929 /* caller must hold RTNL */
1930 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1931 {
1932         struct in6_addr addr;
1933
1934         if (ifp->prefix_len >= 127) /* RFC 6164 */
1935                 return;
1936         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1937         if (ipv6_addr_any(&addr))
1938                 return;
1939         __ipv6_dev_ac_inc(ifp->idev, &addr);
1940 }
1941
1942 /* caller must hold RTNL */
1943 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1944 {
1945         struct in6_addr addr;
1946
1947         if (ifp->prefix_len >= 127) /* RFC 6164 */
1948                 return;
1949         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1950         if (ipv6_addr_any(&addr))
1951                 return;
1952         __ipv6_dev_ac_dec(ifp->idev, &addr);
1953 }
1954
1955 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1956 {
1957         if (dev->addr_len != IEEE802154_ADDR_LEN)
1958                 return -1;
1959         memcpy(eui, dev->dev_addr, 8);
1960         eui[0] ^= 2;
1961         return 0;
1962 }
1963
1964 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1965 {
1966         union fwnet_hwaddr *ha;
1967
1968         if (dev->addr_len != FWNET_ALEN)
1969                 return -1;
1970
1971         ha = (union fwnet_hwaddr *)dev->dev_addr;
1972
1973         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1974         eui[0] ^= 2;
1975         return 0;
1976 }
1977
1978 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1979 {
1980         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1981         if (dev->addr_len != ARCNET_ALEN)
1982                 return -1;
1983         memset(eui, 0, 7);
1984         eui[7] = *(u8 *)dev->dev_addr;
1985         return 0;
1986 }
1987
1988 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1989 {
1990         if (dev->addr_len != INFINIBAND_ALEN)
1991                 return -1;
1992         memcpy(eui, dev->dev_addr + 12, 8);
1993         eui[0] |= 2;
1994         return 0;
1995 }
1996
1997 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1998 {
1999         if (addr == 0)
2000                 return -1;
2001         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2002                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2003                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2004                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2005                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2006                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2007         eui[1] = 0;
2008         eui[2] = 0x5E;
2009         eui[3] = 0xFE;
2010         memcpy(eui + 4, &addr, 4);
2011         return 0;
2012 }
2013
2014 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2015 {
2016         if (dev->priv_flags & IFF_ISATAP)
2017                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2018         return -1;
2019 }
2020
2021 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2022 {
2023         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2024 }
2025
2026 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2027 {
2028         memcpy(eui, dev->perm_addr, 3);
2029         memcpy(eui + 5, dev->perm_addr + 3, 3);
2030         eui[3] = 0xFF;
2031         eui[4] = 0xFE;
2032         eui[0] ^= 2;
2033         return 0;
2034 }
2035
2036 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2037 {
2038         switch (dev->type) {
2039         case ARPHRD_ETHER:
2040         case ARPHRD_FDDI:
2041                 return addrconf_ifid_eui48(eui, dev);
2042         case ARPHRD_ARCNET:
2043                 return addrconf_ifid_arcnet(eui, dev);
2044         case ARPHRD_INFINIBAND:
2045                 return addrconf_ifid_infiniband(eui, dev);
2046         case ARPHRD_SIT:
2047                 return addrconf_ifid_sit(eui, dev);
2048         case ARPHRD_IPGRE:
2049                 return addrconf_ifid_gre(eui, dev);
2050         case ARPHRD_6LOWPAN:
2051         case ARPHRD_IEEE802154:
2052                 return addrconf_ifid_eui64(eui, dev);
2053         case ARPHRD_IEEE1394:
2054                 return addrconf_ifid_ieee1394(eui, dev);
2055         case ARPHRD_TUNNEL6:
2056                 return addrconf_ifid_ip6tnl(eui, dev);
2057         }
2058         return -1;
2059 }
2060
2061 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2062 {
2063         int err = -1;
2064         struct inet6_ifaddr *ifp;
2065
2066         read_lock_bh(&idev->lock);
2067         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2068                 if (ifp->scope > IFA_LINK)
2069                         break;
2070                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2071                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2072                         err = 0;
2073                         break;
2074                 }
2075         }
2076         read_unlock_bh(&idev->lock);
2077         return err;
2078 }
2079
2080 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2081 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2082 {
2083 regen:
2084         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2085         idev->rndid[0] &= ~0x02;
2086
2087         /*
2088          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2089          * check if generated address is not inappropriate
2090          *
2091          *  - Reserved subnet anycast (RFC 2526)
2092          *      11111101 11....11 1xxxxxxx
2093          *  - ISATAP (RFC4214) 6.1
2094          *      00-00-5E-FE-xx-xx-xx-xx
2095          *  - value 0
2096          *  - XXX: already assigned to an address on the device
2097          */
2098         if (idev->rndid[0] == 0xfd &&
2099             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2100             (idev->rndid[7]&0x80))
2101                 goto regen;
2102         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2103                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2104                         goto regen;
2105                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2106                         goto regen;
2107         }
2108 }
2109
2110 static void ipv6_regen_rndid(unsigned long data)
2111 {
2112         struct inet6_dev *idev = (struct inet6_dev *) data;
2113         unsigned long expires;
2114
2115         rcu_read_lock_bh();
2116         write_lock_bh(&idev->lock);
2117
2118         if (idev->dead)
2119                 goto out;
2120
2121         __ipv6_regen_rndid(idev);
2122
2123         expires = jiffies +
2124                 idev->cnf.temp_prefered_lft * HZ -
2125                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2126                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2127                 idev->cnf.max_desync_factor * HZ;
2128         if (time_before(expires, jiffies)) {
2129                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2130                         __func__, idev->dev->name);
2131                 goto out;
2132         }
2133
2134         if (!mod_timer(&idev->regen_timer, expires))
2135                 in6_dev_hold(idev);
2136
2137 out:
2138         write_unlock_bh(&idev->lock);
2139         rcu_read_unlock_bh();
2140         in6_dev_put(idev);
2141 }
2142
2143 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2144 {
2145         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2146                 __ipv6_regen_rndid(idev);
2147 }
2148
2149 /*
2150  *      Add prefix route.
2151  */
2152
2153 static void
2154 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2155                       unsigned long expires, u32 flags)
2156 {
2157         struct fib6_config cfg = {
2158                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2159                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2160                 .fc_ifindex = dev->ifindex,
2161                 .fc_expires = expires,
2162                 .fc_dst_len = plen,
2163                 .fc_flags = RTF_UP | flags,
2164                 .fc_nlinfo.nl_net = dev_net(dev),
2165                 .fc_protocol = RTPROT_KERNEL,
2166         };
2167
2168         cfg.fc_dst = *pfx;
2169
2170         /* Prevent useless cloning on PtP SIT.
2171            This thing is done here expecting that the whole
2172            class of non-broadcast devices need not cloning.
2173          */
2174 #if IS_ENABLED(CONFIG_IPV6_SIT)
2175         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2176                 cfg.fc_flags |= RTF_NONEXTHOP;
2177 #endif
2178
2179         ip6_route_add(&cfg);
2180 }
2181
2182
2183 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2184                                                   int plen,
2185                                                   const struct net_device *dev,
2186                                                   u32 flags, u32 noflags)
2187 {
2188         struct fib6_node *fn;
2189         struct rt6_info *rt = NULL;
2190         struct fib6_table *table;
2191         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2192
2193         table = fib6_get_table(dev_net(dev), tb_id);
2194         if (!table)
2195                 return NULL;
2196
2197         read_lock_bh(&table->tb6_lock);
2198         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2199         if (!fn)
2200                 goto out;
2201
2202         noflags |= RTF_CACHE;
2203         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2204                 if (rt->dst.dev->ifindex != dev->ifindex)
2205                         continue;
2206                 if ((rt->rt6i_flags & flags) != flags)
2207                         continue;
2208                 if ((rt->rt6i_flags & noflags) != 0)
2209                         continue;
2210                 dst_hold(&rt->dst);
2211                 break;
2212         }
2213 out:
2214         read_unlock_bh(&table->tb6_lock);
2215         return rt;
2216 }
2217
2218
2219 /* Create "default" multicast route to the interface */
2220
2221 static void addrconf_add_mroute(struct net_device *dev)
2222 {
2223         struct fib6_config cfg = {
2224                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2225                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2226                 .fc_ifindex = dev->ifindex,
2227                 .fc_dst_len = 8,
2228                 .fc_flags = RTF_UP,
2229                 .fc_nlinfo.nl_net = dev_net(dev),
2230         };
2231
2232         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2233
2234         ip6_route_add(&cfg);
2235 }
2236
2237 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2238 {
2239         struct inet6_dev *idev;
2240
2241         ASSERT_RTNL();
2242
2243         idev = ipv6_find_idev(dev);
2244         if (!idev)
2245                 return ERR_PTR(-ENOBUFS);
2246
2247         if (idev->cnf.disable_ipv6)
2248                 return ERR_PTR(-EACCES);
2249
2250         /* Add default multicast route */
2251         if (!(dev->flags & IFF_LOOPBACK))
2252                 addrconf_add_mroute(dev);
2253
2254         return idev;
2255 }
2256
2257 static void manage_tempaddrs(struct inet6_dev *idev,
2258                              struct inet6_ifaddr *ifp,
2259                              __u32 valid_lft, __u32 prefered_lft,
2260                              bool create, unsigned long now)
2261 {
2262         u32 flags;
2263         struct inet6_ifaddr *ift;
2264
2265         read_lock_bh(&idev->lock);
2266         /* update all temporary addresses in the list */
2267         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2268                 int age, max_valid, max_prefered;
2269
2270                 if (ifp != ift->ifpub)
2271                         continue;
2272
2273                 /* RFC 4941 section 3.3:
2274                  * If a received option will extend the lifetime of a public
2275                  * address, the lifetimes of temporary addresses should
2276                  * be extended, subject to the overall constraint that no
2277                  * temporary addresses should ever remain "valid" or "preferred"
2278                  * for a time longer than (TEMP_VALID_LIFETIME) or
2279                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2280                  */
2281                 age = (now - ift->cstamp) / HZ;
2282                 max_valid = idev->cnf.temp_valid_lft - age;
2283                 if (max_valid < 0)
2284                         max_valid = 0;
2285
2286                 max_prefered = idev->cnf.temp_prefered_lft -
2287                                idev->cnf.max_desync_factor - age;
2288                 if (max_prefered < 0)
2289                         max_prefered = 0;
2290
2291                 if (valid_lft > max_valid)
2292                         valid_lft = max_valid;
2293
2294                 if (prefered_lft > max_prefered)
2295                         prefered_lft = max_prefered;
2296
2297                 spin_lock(&ift->lock);
2298                 flags = ift->flags;
2299                 ift->valid_lft = valid_lft;
2300                 ift->prefered_lft = prefered_lft;
2301                 ift->tstamp = now;
2302                 if (prefered_lft > 0)
2303                         ift->flags &= ~IFA_F_DEPRECATED;
2304
2305                 spin_unlock(&ift->lock);
2306                 if (!(flags&IFA_F_TENTATIVE))
2307                         ipv6_ifa_notify(0, ift);
2308         }
2309
2310         if ((create || list_empty(&idev->tempaddr_list)) &&
2311             idev->cnf.use_tempaddr > 0) {
2312                 /* When a new public address is created as described
2313                  * in [ADDRCONF], also create a new temporary address.
2314                  * Also create a temporary address if it's enabled but
2315                  * no temporary address currently exists.
2316                  */
2317                 read_unlock_bh(&idev->lock);
2318                 ipv6_create_tempaddr(ifp, NULL);
2319         } else {
2320                 read_unlock_bh(&idev->lock);
2321         }
2322 }
2323
2324 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2325 {
2326         struct prefix_info *pinfo;
2327         __u32 valid_lft;
2328         __u32 prefered_lft;
2329         int addr_type;
2330         u32 addr_flags = 0;
2331         struct inet6_dev *in6_dev;
2332         struct net *net = dev_net(dev);
2333
2334         pinfo = (struct prefix_info *) opt;
2335
2336         if (len < sizeof(struct prefix_info)) {
2337                 ADBG("addrconf: prefix option too short\n");
2338                 return;
2339         }
2340
2341         /*
2342          *      Validation checks ([ADDRCONF], page 19)
2343          */
2344
2345         addr_type = ipv6_addr_type(&pinfo->prefix);
2346
2347         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2348                 return;
2349
2350         valid_lft = ntohl(pinfo->valid);
2351         prefered_lft = ntohl(pinfo->prefered);
2352
2353         if (prefered_lft > valid_lft) {
2354                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2355                 return;
2356         }
2357
2358         in6_dev = in6_dev_get(dev);
2359
2360         if (!in6_dev) {
2361                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2362                                     dev->name);
2363                 return;
2364         }
2365
2366         /*
2367          *      Two things going on here:
2368          *      1) Add routes for on-link prefixes
2369          *      2) Configure prefixes with the auto flag set
2370          */
2371
2372         if (pinfo->onlink) {
2373                 struct rt6_info *rt;
2374                 unsigned long rt_expires;
2375
2376                 /* Avoid arithmetic overflow. Really, we could
2377                  * save rt_expires in seconds, likely valid_lft,
2378                  * but it would require division in fib gc, that it
2379                  * not good.
2380                  */
2381                 if (HZ > USER_HZ)
2382                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2383                 else
2384                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2385
2386                 if (addrconf_finite_timeout(rt_expires))
2387                         rt_expires *= HZ;
2388
2389                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2390                                                pinfo->prefix_len,
2391                                                dev,
2392                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2393                                                RTF_GATEWAY | RTF_DEFAULT);
2394
2395                 if (rt) {
2396                         /* Autoconf prefix route */
2397                         if (valid_lft == 0) {
2398                                 ip6_del_rt(rt);
2399                                 rt = NULL;
2400                         } else if (addrconf_finite_timeout(rt_expires)) {
2401                                 /* not infinity */
2402                                 rt6_set_expires(rt, jiffies + rt_expires);
2403                         } else {
2404                                 rt6_clean_expires(rt);
2405                         }
2406                 } else if (valid_lft) {
2407                         clock_t expires = 0;
2408                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2409                         if (addrconf_finite_timeout(rt_expires)) {
2410                                 /* not infinity */
2411                                 flags |= RTF_EXPIRES;
2412                                 expires = jiffies_to_clock_t(rt_expires);
2413                         }
2414                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2415                                               dev, expires, flags);
2416                 }
2417                 ip6_rt_put(rt);
2418         }
2419
2420         /* Try to figure out our local address for this prefix */
2421
2422         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2423                 struct inet6_ifaddr *ifp;
2424                 struct in6_addr addr;
2425                 int create = 0, update_lft = 0;
2426                 bool tokenized = false;
2427
2428                 if (pinfo->prefix_len == 64) {
2429                         memcpy(&addr, &pinfo->prefix, 8);
2430
2431                         if (!ipv6_addr_any(&in6_dev->token)) {
2432                                 read_lock_bh(&in6_dev->lock);
2433                                 memcpy(addr.s6_addr + 8,
2434                                        in6_dev->token.s6_addr + 8, 8);
2435                                 read_unlock_bh(&in6_dev->lock);
2436                                 tokenized = true;
2437                         } else if (in6_dev->addr_gen_mode ==
2438                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2439                                    !ipv6_generate_stable_address(&addr, 0,
2440                                                                  in6_dev)) {
2441                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2442                                 goto ok;
2443                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2444                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2445                                 in6_dev_put(in6_dev);
2446                                 return;
2447                         }
2448                         goto ok;
2449                 }
2450                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2451                                     pinfo->prefix_len);
2452                 in6_dev_put(in6_dev);
2453                 return;
2454
2455 ok:
2456
2457                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2458
2459                 if (!ifp && valid_lft) {
2460                         int max_addresses = in6_dev->cnf.max_addresses;
2461
2462 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2463                         if (in6_dev->cnf.optimistic_dad &&
2464                             !net->ipv6.devconf_all->forwarding && sllao)
2465                                 addr_flags |= IFA_F_OPTIMISTIC;
2466 #endif
2467
2468                         /* Do not allow to create too much of autoconfigured
2469                          * addresses; this would be too easy way to crash kernel.
2470                          */
2471                         if (!max_addresses ||
2472                             ipv6_count_addresses(in6_dev) < max_addresses)
2473                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2474                                                     pinfo->prefix_len,
2475                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2476                                                     addr_flags, valid_lft,
2477                                                     prefered_lft);
2478
2479                         if (IS_ERR_OR_NULL(ifp)) {
2480                                 in6_dev_put(in6_dev);
2481                                 return;
2482                         }
2483
2484                         update_lft = 0;
2485                         create = 1;
2486                         spin_lock_bh(&ifp->lock);
2487                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2488                         ifp->cstamp = jiffies;
2489                         ifp->tokenized = tokenized;
2490                         spin_unlock_bh(&ifp->lock);
2491                         addrconf_dad_start(ifp);
2492                 }
2493
2494                 if (ifp) {
2495                         u32 flags;
2496                         unsigned long now;
2497                         u32 stored_lft;
2498
2499                         /* update lifetime (RFC2462 5.5.3 e) */
2500                         spin_lock_bh(&ifp->lock);
2501                         now = jiffies;
2502                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2503                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2504                         else
2505                                 stored_lft = 0;
2506                         if (!update_lft && !create && stored_lft) {
2507                                 const u32 minimum_lft = min_t(u32,
2508                                         stored_lft, MIN_VALID_LIFETIME);
2509                                 valid_lft = max(valid_lft, minimum_lft);
2510
2511                                 /* RFC4862 Section 5.5.3e:
2512                                  * "Note that the preferred lifetime of the
2513                                  *  corresponding address is always reset to
2514                                  *  the Preferred Lifetime in the received
2515                                  *  Prefix Information option, regardless of
2516                                  *  whether the valid lifetime is also reset or
2517                                  *  ignored."
2518                                  *
2519                                  * So we should always update prefered_lft here.
2520                                  */
2521                                 update_lft = 1;
2522                         }
2523
2524                         if (update_lft) {
2525                                 ifp->valid_lft = valid_lft;
2526                                 ifp->prefered_lft = prefered_lft;
2527                                 ifp->tstamp = now;
2528                                 flags = ifp->flags;
2529                                 ifp->flags &= ~IFA_F_DEPRECATED;
2530                                 spin_unlock_bh(&ifp->lock);
2531
2532                                 if (!(flags&IFA_F_TENTATIVE))
2533                                         ipv6_ifa_notify(0, ifp);
2534                         } else
2535                                 spin_unlock_bh(&ifp->lock);
2536
2537                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2538                                          create, now);
2539
2540                         in6_ifa_put(ifp);
2541                         addrconf_verify();
2542                 }
2543         }
2544         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2545         in6_dev_put(in6_dev);
2546 }
2547
2548 /*
2549  *      Set destination address.
2550  *      Special case for SIT interfaces where we create a new "virtual"
2551  *      device.
2552  */
2553 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2554 {
2555         struct in6_ifreq ireq;
2556         struct net_device *dev;
2557         int err = -EINVAL;
2558
2559         rtnl_lock();
2560
2561         err = -EFAULT;
2562         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2563                 goto err_exit;
2564
2565         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2566
2567         err = -ENODEV;
2568         if (!dev)
2569                 goto err_exit;
2570
2571 #if IS_ENABLED(CONFIG_IPV6_SIT)
2572         if (dev->type == ARPHRD_SIT) {
2573                 const struct net_device_ops *ops = dev->netdev_ops;
2574                 struct ifreq ifr;
2575                 struct ip_tunnel_parm p;
2576
2577                 err = -EADDRNOTAVAIL;
2578                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2579                         goto err_exit;
2580
2581                 memset(&p, 0, sizeof(p));
2582                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2583                 p.iph.saddr = 0;
2584                 p.iph.version = 4;
2585                 p.iph.ihl = 5;
2586                 p.iph.protocol = IPPROTO_IPV6;
2587                 p.iph.ttl = 64;
2588                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2589
2590                 if (ops->ndo_do_ioctl) {
2591                         mm_segment_t oldfs = get_fs();
2592
2593                         set_fs(KERNEL_DS);
2594                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2595                         set_fs(oldfs);
2596                 } else
2597                         err = -EOPNOTSUPP;
2598
2599                 if (err == 0) {
2600                         err = -ENOBUFS;
2601                         dev = __dev_get_by_name(net, p.name);
2602                         if (!dev)
2603                                 goto err_exit;
2604                         err = dev_open(dev);
2605                 }
2606         }
2607 #endif
2608
2609 err_exit:
2610         rtnl_unlock();
2611         return err;
2612 }
2613
2614 static int ipv6_mc_config(struct sock *sk, bool join,
2615                           const struct in6_addr *addr, int ifindex)
2616 {
2617         int ret;
2618
2619         ASSERT_RTNL();
2620
2621         lock_sock(sk);
2622         if (join)
2623                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2624         else
2625                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2626         release_sock(sk);
2627
2628         return ret;
2629 }
2630
2631 /*
2632  *      Manual configuration of address on an interface
2633  */
2634 static int inet6_addr_add(struct net *net, int ifindex,
2635                           const struct in6_addr *pfx,
2636                           const struct in6_addr *peer_pfx,
2637                           unsigned int plen, __u32 ifa_flags,
2638                           __u32 prefered_lft, __u32 valid_lft)
2639 {
2640         struct inet6_ifaddr *ifp;
2641         struct inet6_dev *idev;
2642         struct net_device *dev;
2643         unsigned long timeout;
2644         clock_t expires;
2645         int scope;
2646         u32 flags;
2647
2648         ASSERT_RTNL();
2649
2650         if (plen > 128)
2651                 return -EINVAL;
2652
2653         /* check the lifetime */
2654         if (!valid_lft || prefered_lft > valid_lft)
2655                 return -EINVAL;
2656
2657         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2658                 return -EINVAL;
2659
2660         dev = __dev_get_by_index(net, ifindex);
2661         if (!dev)
2662                 return -ENODEV;
2663
2664         idev = addrconf_add_dev(dev);
2665         if (IS_ERR(idev))
2666                 return PTR_ERR(idev);
2667
2668         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2669                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2670                                          true, pfx, ifindex);
2671
2672                 if (ret < 0)
2673                         return ret;
2674         }
2675
2676         scope = ipv6_addr_scope(pfx);
2677
2678         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2679         if (addrconf_finite_timeout(timeout)) {
2680                 expires = jiffies_to_clock_t(timeout * HZ);
2681                 valid_lft = timeout;
2682                 flags = RTF_EXPIRES;
2683         } else {
2684                 expires = 0;
2685                 flags = 0;
2686                 ifa_flags |= IFA_F_PERMANENT;
2687         }
2688
2689         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2690         if (addrconf_finite_timeout(timeout)) {
2691                 if (timeout == 0)
2692                         ifa_flags |= IFA_F_DEPRECATED;
2693                 prefered_lft = timeout;
2694         }
2695
2696         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2697                             valid_lft, prefered_lft);
2698
2699         if (!IS_ERR(ifp)) {
2700                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2701                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2702                                               expires, flags);
2703                 }
2704
2705                 /*
2706                  * Note that section 3.1 of RFC 4429 indicates
2707                  * that the Optimistic flag should not be set for
2708                  * manually configured addresses
2709                  */
2710                 addrconf_dad_start(ifp);
2711                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2712                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2713                                          true, jiffies);
2714                 in6_ifa_put(ifp);
2715                 addrconf_verify_rtnl();
2716                 return 0;
2717         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2718                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2719                                false, pfx, ifindex);
2720         }
2721
2722         return PTR_ERR(ifp);
2723 }
2724
2725 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2726                           const struct in6_addr *pfx, unsigned int plen)
2727 {
2728         struct inet6_ifaddr *ifp;
2729         struct inet6_dev *idev;
2730         struct net_device *dev;
2731
2732         if (plen > 128)
2733                 return -EINVAL;
2734
2735         dev = __dev_get_by_index(net, ifindex);
2736         if (!dev)
2737                 return -ENODEV;
2738
2739         idev = __in6_dev_get(dev);
2740         if (!idev)
2741                 return -ENXIO;
2742
2743         read_lock_bh(&idev->lock);
2744         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2745                 if (ifp->prefix_len == plen &&
2746                     ipv6_addr_equal(pfx, &ifp->addr)) {
2747                         in6_ifa_hold(ifp);
2748                         read_unlock_bh(&idev->lock);
2749
2750                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2751                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2752                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2753                                                  jiffies);
2754                         ipv6_del_addr(ifp);
2755                         addrconf_verify_rtnl();
2756                         if (ipv6_addr_is_multicast(pfx)) {
2757                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2758                                                false, pfx, dev->ifindex);
2759                         }
2760                         return 0;
2761                 }
2762         }
2763         read_unlock_bh(&idev->lock);
2764         return -EADDRNOTAVAIL;
2765 }
2766
2767
2768 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2769 {
2770         struct in6_ifreq ireq;
2771         int err;
2772
2773         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2774                 return -EPERM;
2775
2776         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2777                 return -EFAULT;
2778
2779         rtnl_lock();
2780         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2781                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2782                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2783         rtnl_unlock();
2784         return err;
2785 }
2786
2787 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2788 {
2789         struct in6_ifreq ireq;
2790         int err;
2791
2792         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2793                 return -EPERM;
2794
2795         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2796                 return -EFAULT;
2797
2798         rtnl_lock();
2799         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2800                              ireq.ifr6_prefixlen);
2801         rtnl_unlock();
2802         return err;
2803 }
2804
2805 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2806                      int plen, int scope)
2807 {
2808         struct inet6_ifaddr *ifp;
2809
2810         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2811                             scope, IFA_F_PERMANENT,
2812                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2813         if (!IS_ERR(ifp)) {
2814                 spin_lock_bh(&ifp->lock);
2815                 ifp->flags &= ~IFA_F_TENTATIVE;
2816                 spin_unlock_bh(&ifp->lock);
2817                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2818                 in6_ifa_put(ifp);
2819         }
2820 }
2821
2822 #if IS_ENABLED(CONFIG_IPV6_SIT)
2823 static void sit_add_v4_addrs(struct inet6_dev *idev)
2824 {
2825         struct in6_addr addr;
2826         struct net_device *dev;
2827         struct net *net = dev_net(idev->dev);
2828         int scope, plen;
2829         u32 pflags = 0;
2830
2831         ASSERT_RTNL();
2832
2833         memset(&addr, 0, sizeof(struct in6_addr));
2834         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2835
2836         if (idev->dev->flags&IFF_POINTOPOINT) {
2837                 addr.s6_addr32[0] = htonl(0xfe800000);
2838                 scope = IFA_LINK;
2839                 plen = 64;
2840         } else {
2841                 scope = IPV6_ADDR_COMPATv4;
2842                 plen = 96;
2843                 pflags |= RTF_NONEXTHOP;
2844         }
2845
2846         if (addr.s6_addr32[3]) {
2847                 add_addr(idev, &addr, plen, scope);
2848                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2849                 return;
2850         }
2851
2852         for_each_netdev(net, dev) {
2853                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2854                 if (in_dev && (dev->flags & IFF_UP)) {
2855                         struct in_ifaddr *ifa;
2856
2857                         int flag = scope;
2858
2859                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2860
2861                                 addr.s6_addr32[3] = ifa->ifa_local;
2862
2863                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2864                                         continue;
2865                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2866                                         if (idev->dev->flags&IFF_POINTOPOINT)
2867                                                 continue;
2868                                         flag |= IFA_HOST;
2869                                 }
2870
2871                                 add_addr(idev, &addr, plen, flag);
2872                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2873                                                       pflags);
2874                         }
2875                 }
2876         }
2877 }
2878 #endif
2879
2880 static void init_loopback(struct net_device *dev)
2881 {
2882         struct inet6_dev  *idev;
2883         struct net_device *sp_dev;
2884         struct inet6_ifaddr *sp_ifa;
2885         struct rt6_info *sp_rt;
2886
2887         /* ::1 */
2888
2889         ASSERT_RTNL();
2890
2891         idev = ipv6_find_idev(dev);
2892         if (!idev) {
2893                 pr_debug("%s: add_dev failed\n", __func__);
2894                 return;
2895         }
2896
2897         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2898
2899         /* Add routes to other interface's IPv6 addresses */
2900         for_each_netdev(dev_net(dev), sp_dev) {
2901                 if (!strcmp(sp_dev->name, dev->name))
2902                         continue;
2903
2904                 idev = __in6_dev_get(sp_dev);
2905                 if (!idev)
2906                         continue;
2907
2908                 read_lock_bh(&idev->lock);
2909                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2910
2911                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2912                                 continue;
2913
2914                         if (sp_ifa->rt) {
2915                                 /* This dst has been added to garbage list when
2916                                  * lo device down, release this obsolete dst and
2917                                  * reallocate a new router for ifa.
2918                                  */
2919                                 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
2920                                         ip6_rt_put(sp_ifa->rt);
2921                                         sp_ifa->rt = NULL;
2922                                 } else {
2923                                         continue;
2924                                 }
2925                         }
2926
2927                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2928
2929                         /* Failure cases are ignored */
2930                         if (!IS_ERR(sp_rt)) {
2931                                 sp_ifa->rt = sp_rt;
2932                                 ip6_ins_rt(sp_rt);
2933                         }
2934                 }
2935                 read_unlock_bh(&idev->lock);
2936         }
2937 }
2938
2939 static void addrconf_add_linklocal(struct inet6_dev *idev,
2940                                    const struct in6_addr *addr, u32 flags)
2941 {
2942         struct inet6_ifaddr *ifp;
2943         u32 addr_flags = flags | IFA_F_PERMANENT;
2944
2945 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2946         if (idev->cnf.optimistic_dad &&
2947             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2948                 addr_flags |= IFA_F_OPTIMISTIC;
2949 #endif
2950
2951         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2952                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2953         if (!IS_ERR(ifp)) {
2954                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2955                 addrconf_dad_start(ifp);
2956                 in6_ifa_put(ifp);
2957         }
2958 }
2959
2960 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2961 {
2962         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2963                 return true;
2964
2965         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2966             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2967                 return true;
2968
2969         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2970             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2971                 return true;
2972
2973         return false;
2974 }
2975
2976 static int ipv6_generate_stable_address(struct in6_addr *address,
2977                                         u8 dad_count,
2978                                         const struct inet6_dev *idev)
2979 {
2980         static DEFINE_SPINLOCK(lock);
2981         static __u32 digest[SHA_DIGEST_WORDS];
2982         static __u32 workspace[SHA_WORKSPACE_WORDS];
2983
2984         static union {
2985                 char __data[SHA_MESSAGE_BYTES];
2986                 struct {
2987                         struct in6_addr secret;
2988                         __be32 prefix[2];
2989                         unsigned char hwaddr[MAX_ADDR_LEN];
2990                         u8 dad_count;
2991                 } __packed;
2992         } data;
2993
2994         struct in6_addr secret;
2995         struct in6_addr temp;
2996         struct net *net = dev_net(idev->dev);
2997
2998         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2999
3000         if (idev->cnf.stable_secret.initialized)
3001                 secret = idev->cnf.stable_secret.secret;
3002         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3003                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3004         else
3005                 return -1;
3006
3007 retry:
3008         spin_lock_bh(&lock);
3009
3010         sha_init(digest);
3011         memset(&data, 0, sizeof(data));
3012         memset(workspace, 0, sizeof(workspace));
3013         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3014         data.prefix[0] = address->s6_addr32[0];
3015         data.prefix[1] = address->s6_addr32[1];
3016         data.secret = secret;
3017         data.dad_count = dad_count;
3018
3019         sha_transform(digest, data.__data, workspace);
3020
3021         temp = *address;
3022         temp.s6_addr32[2] = (__force __be32)digest[0];
3023         temp.s6_addr32[3] = (__force __be32)digest[1];
3024
3025         spin_unlock_bh(&lock);
3026
3027         if (ipv6_reserved_interfaceid(temp)) {
3028                 dad_count++;
3029                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3030                         return -1;
3031                 goto retry;
3032         }
3033
3034         *address = temp;
3035         return 0;
3036 }
3037
3038 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3039 {
3040         struct in6_addr addr;
3041
3042         /* no link local addresses on L3 master devices */
3043         if (netif_is_l3_master(idev->dev))
3044                 return;
3045
3046         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3047
3048         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3049                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3050                         addrconf_add_linklocal(idev, &addr,
3051                                                IFA_F_STABLE_PRIVACY);
3052                 else if (prefix_route)
3053                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3054         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3055                 /* addrconf_add_linklocal also adds a prefix_route and we
3056                  * only need to care about prefix routes if ipv6_generate_eui64
3057                  * couldn't generate one.
3058                  */
3059                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3060                         addrconf_add_linklocal(idev, &addr, 0);
3061                 else if (prefix_route)
3062                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3063         }
3064 }
3065
3066 static void addrconf_dev_config(struct net_device *dev)
3067 {
3068         struct inet6_dev *idev;
3069
3070         ASSERT_RTNL();
3071
3072         if ((dev->type != ARPHRD_ETHER) &&
3073             (dev->type != ARPHRD_FDDI) &&
3074             (dev->type != ARPHRD_ARCNET) &&
3075             (dev->type != ARPHRD_INFINIBAND) &&
3076             (dev->type != ARPHRD_IEEE802154) &&
3077             (dev->type != ARPHRD_IEEE1394) &&
3078             (dev->type != ARPHRD_TUNNEL6) &&
3079             (dev->type != ARPHRD_6LOWPAN)) {
3080                 /* Alas, we support only Ethernet autoconfiguration. */
3081                 return;
3082         }
3083
3084         idev = addrconf_add_dev(dev);
3085         if (IS_ERR(idev))
3086                 return;
3087
3088         addrconf_addr_gen(idev, false);
3089 }
3090
3091 #if IS_ENABLED(CONFIG_IPV6_SIT)
3092 static void addrconf_sit_config(struct net_device *dev)
3093 {
3094         struct inet6_dev *idev;
3095
3096         ASSERT_RTNL();
3097
3098         /*
3099          * Configure the tunnel with one of our IPv4
3100          * addresses... we should configure all of
3101          * our v4 addrs in the tunnel
3102          */
3103
3104         idev = ipv6_find_idev(dev);
3105         if (!idev) {
3106                 pr_debug("%s: add_dev failed\n", __func__);
3107                 return;
3108         }
3109
3110         if (dev->priv_flags & IFF_ISATAP) {
3111                 addrconf_addr_gen(idev, false);
3112                 return;
3113         }
3114
3115         sit_add_v4_addrs(idev);
3116
3117         if (dev->flags&IFF_POINTOPOINT)
3118                 addrconf_add_mroute(dev);
3119 }
3120 #endif
3121
3122 #if IS_ENABLED(CONFIG_NET_IPGRE)
3123 static void addrconf_gre_config(struct net_device *dev)
3124 {
3125         struct inet6_dev *idev;
3126
3127         ASSERT_RTNL();
3128
3129         idev = ipv6_find_idev(dev);
3130         if (!idev) {
3131                 pr_debug("%s: add_dev failed\n", __func__);
3132                 return;
3133         }
3134
3135         addrconf_addr_gen(idev, true);
3136         if (dev->flags & IFF_POINTOPOINT)
3137                 addrconf_add_mroute(dev);
3138 }
3139 #endif
3140
3141 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3142                            void *ptr)
3143 {
3144         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3145         struct inet6_dev *idev = __in6_dev_get(dev);
3146         int run_pending = 0;
3147         int err;
3148
3149         switch (event) {
3150         case NETDEV_REGISTER:
3151                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3152                         idev = ipv6_add_dev(dev);
3153                         if (IS_ERR(idev))
3154                                 return notifier_from_errno(PTR_ERR(idev));
3155                 }
3156                 break;
3157
3158         case NETDEV_CHANGEMTU:
3159                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3160                 if (dev->mtu < IPV6_MIN_MTU) {
3161                         addrconf_ifdown(dev, 1);
3162                         break;
3163                 }
3164
3165                 if (idev) {
3166                         rt6_mtu_change(dev, dev->mtu);
3167                         idev->cnf.mtu6 = dev->mtu;
3168                         break;
3169                 }
3170
3171                 /* allocate new idev */
3172                 idev = ipv6_add_dev(dev);
3173                 if (IS_ERR(idev))
3174                         break;
3175
3176                 /* device is still not ready */
3177                 if (!(idev->if_flags & IF_READY))
3178                         break;
3179
3180                 run_pending = 1;
3181
3182                 /* fall through */
3183
3184         case NETDEV_UP:
3185         case NETDEV_CHANGE:
3186                 if (dev->flags & IFF_SLAVE)
3187                         break;
3188
3189                 if (idev && idev->cnf.disable_ipv6)
3190                         break;
3191
3192                 if (event == NETDEV_UP) {
3193                         if (!addrconf_qdisc_ok(dev)) {
3194                                 /* device is not ready yet. */
3195                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3196                                         dev->name);
3197                                 break;
3198                         }
3199
3200                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3201                                 idev = ipv6_add_dev(dev);
3202
3203                         if (!IS_ERR_OR_NULL(idev)) {
3204                                 idev->if_flags |= IF_READY;
3205                                 run_pending = 1;
3206                         }
3207                 } else if (event == NETDEV_CHANGE) {
3208                         if (!addrconf_qdisc_ok(dev)) {
3209                                 /* device is still not ready. */
3210                                 break;
3211                         }
3212
3213                         if (idev) {
3214                                 if (idev->if_flags & IF_READY)
3215                                         /* device is already configured. */
3216                                         break;
3217                                 idev->if_flags |= IF_READY;
3218                         }
3219
3220                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3221                                 dev->name);
3222
3223                         run_pending = 1;
3224                 }
3225
3226                 switch (dev->type) {
3227 #if IS_ENABLED(CONFIG_IPV6_SIT)
3228                 case ARPHRD_SIT:
3229                         addrconf_sit_config(dev);
3230                         break;
3231 #endif
3232 #if IS_ENABLED(CONFIG_NET_IPGRE)
3233                 case ARPHRD_IPGRE:
3234                         addrconf_gre_config(dev);
3235                         break;
3236 #endif
3237                 case ARPHRD_LOOPBACK:
3238                         init_loopback(dev);
3239                         break;
3240
3241                 default:
3242                         addrconf_dev_config(dev);
3243                         break;
3244                 }
3245
3246                 if (!IS_ERR_OR_NULL(idev)) {
3247                         if (run_pending)
3248                                 addrconf_dad_run(idev);
3249
3250                         /*
3251                          * If the MTU changed during the interface down,
3252                          * when the interface up, the changed MTU must be
3253                          * reflected in the idev as well as routers.
3254                          */
3255                         if (idev->cnf.mtu6 != dev->mtu &&
3256                             dev->mtu >= IPV6_MIN_MTU) {
3257                                 rt6_mtu_change(dev, dev->mtu);
3258                                 idev->cnf.mtu6 = dev->mtu;
3259                         }
3260                         idev->tstamp = jiffies;
3261                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3262
3263                         /*
3264                          * If the changed mtu during down is lower than
3265                          * IPV6_MIN_MTU stop IPv6 on this interface.
3266                          */
3267                         if (dev->mtu < IPV6_MIN_MTU)
3268                                 addrconf_ifdown(dev, 1);
3269                 }
3270                 break;
3271
3272         case NETDEV_DOWN:
3273         case NETDEV_UNREGISTER:
3274                 /*
3275                  *      Remove all addresses from this interface.
3276                  */
3277                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3278                 break;
3279
3280         case NETDEV_CHANGENAME:
3281                 if (idev) {
3282                         snmp6_unregister_dev(idev);
3283                         addrconf_sysctl_unregister(idev);
3284                         err = addrconf_sysctl_register(idev);
3285                         if (err)
3286                                 return notifier_from_errno(err);
3287                         err = snmp6_register_dev(idev);
3288                         if (err) {
3289                                 addrconf_sysctl_unregister(idev);
3290                                 return notifier_from_errno(err);
3291                         }
3292                 }
3293                 break;
3294
3295         case NETDEV_PRE_TYPE_CHANGE:
3296         case NETDEV_POST_TYPE_CHANGE:
3297                 addrconf_type_change(dev, event);
3298                 break;
3299         }
3300
3301         return NOTIFY_OK;
3302 }
3303
3304 /*
3305  *      addrconf module should be notified of a device going up
3306  */
3307 static struct notifier_block ipv6_dev_notf = {
3308         .notifier_call = addrconf_notify,
3309 };
3310
3311 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3312 {
3313         struct inet6_dev *idev;
3314         ASSERT_RTNL();
3315
3316         idev = __in6_dev_get(dev);
3317
3318         if (event == NETDEV_POST_TYPE_CHANGE)
3319                 ipv6_mc_remap(idev);
3320         else if (event == NETDEV_PRE_TYPE_CHANGE)
3321                 ipv6_mc_unmap(idev);
3322 }
3323
3324 static int addrconf_ifdown(struct net_device *dev, int how)
3325 {
3326         struct net *net = dev_net(dev);
3327         struct inet6_dev *idev;
3328         struct inet6_ifaddr *ifa;
3329         int state, i;
3330
3331         ASSERT_RTNL();
3332
3333         rt6_ifdown(net, dev);
3334         neigh_ifdown(&nd_tbl, dev);
3335
3336         idev = __in6_dev_get(dev);
3337         if (!idev)
3338                 return -ENODEV;
3339
3340         /*
3341          * Step 1: remove reference to ipv6 device from parent device.
3342          *         Do not dev_put!
3343          */
3344         if (how) {
3345                 idev->dead = 1;
3346
3347                 /* protected by rtnl_lock */
3348                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3349
3350                 /* Step 1.5: remove snmp6 entry */
3351                 snmp6_unregister_dev(idev);
3352
3353         }
3354
3355         /* Step 2: clear hash table */
3356         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3357                 struct hlist_head *h = &inet6_addr_lst[i];
3358
3359                 spin_lock_bh(&addrconf_hash_lock);
3360 restart:
3361                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3362                         if (ifa->idev == idev) {
3363                                 hlist_del_init_rcu(&ifa->addr_lst);
3364                                 addrconf_del_dad_work(ifa);
3365                                 goto restart;
3366                         }
3367                 }
3368                 spin_unlock_bh(&addrconf_hash_lock);
3369         }
3370
3371         write_lock_bh(&idev->lock);
3372
3373         addrconf_del_rs_timer(idev);
3374
3375         /* Step 2: clear flags for stateless addrconf */
3376         if (!how)
3377                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3378
3379         if (how && del_timer(&idev->regen_timer))
3380                 in6_dev_put(idev);
3381
3382         /* Step 3: clear tempaddr list */
3383         while (!list_empty(&idev->tempaddr_list)) {
3384                 ifa = list_first_entry(&idev->tempaddr_list,
3385                                        struct inet6_ifaddr, tmp_list);
3386                 list_del(&ifa->tmp_list);
3387                 write_unlock_bh(&idev->lock);
3388                 spin_lock_bh(&ifa->lock);
3389
3390                 if (ifa->ifpub) {
3391                         in6_ifa_put(ifa->ifpub);
3392                         ifa->ifpub = NULL;
3393                 }
3394                 spin_unlock_bh(&ifa->lock);
3395                 in6_ifa_put(ifa);
3396                 write_lock_bh(&idev->lock);
3397         }
3398
3399         while (!list_empty(&idev->addr_list)) {
3400                 ifa = list_first_entry(&idev->addr_list,
3401                                        struct inet6_ifaddr, if_list);
3402                 addrconf_del_dad_work(ifa);
3403
3404                 list_del(&ifa->if_list);
3405
3406                 write_unlock_bh(&idev->lock);
3407
3408                 spin_lock_bh(&ifa->lock);
3409                 state = ifa->state;
3410                 ifa->state = INET6_IFADDR_STATE_DEAD;
3411                 spin_unlock_bh(&ifa->lock);
3412
3413                 if (state != INET6_IFADDR_STATE_DEAD) {
3414                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3415                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3416                 }
3417                 in6_ifa_put(ifa);
3418
3419                 write_lock_bh(&idev->lock);
3420         }
3421
3422         write_unlock_bh(&idev->lock);
3423
3424         /* Step 5: Discard anycast and multicast list */
3425         if (how) {
3426                 ipv6_ac_destroy_dev(idev);
3427                 ipv6_mc_destroy_dev(idev);
3428         } else {
3429                 ipv6_mc_down(idev);
3430         }
3431
3432         idev->tstamp = jiffies;
3433
3434         /* Last: Shot the device (if unregistered) */
3435         if (how) {
3436                 addrconf_sysctl_unregister(idev);
3437                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3438                 neigh_ifdown(&nd_tbl, dev);
3439                 in6_dev_put(idev);
3440         }
3441         return 0;
3442 }
3443
3444 static void addrconf_rs_timer(unsigned long data)
3445 {
3446         struct inet6_dev *idev = (struct inet6_dev *)data;
3447         struct net_device *dev = idev->dev;
3448         struct in6_addr lladdr;
3449
3450         write_lock(&idev->lock);
3451         if (idev->dead || !(idev->if_flags & IF_READY))
3452                 goto out;
3453
3454         if (!ipv6_accept_ra(idev))
3455                 goto out;
3456
3457         /* Announcement received after solicitation was sent */
3458         if (idev->if_flags & IF_RA_RCVD)
3459                 goto out;
3460
3461         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3462                 write_unlock(&idev->lock);
3463                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3464                         ndisc_send_rs(dev, &lladdr,
3465                                       &in6addr_linklocal_allrouters);
3466                 else
3467                         goto put;
3468
3469                 write_lock(&idev->lock);
3470                 /* The wait after the last probe can be shorter */
3471                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3472                                              idev->cnf.rtr_solicits) ?
3473                                       idev->cnf.rtr_solicit_delay :
3474                                       idev->cnf.rtr_solicit_interval);
3475         } else {
3476                 /*
3477                  * Note: we do not support deprecated "all on-link"
3478                  * assumption any longer.
3479                  */
3480                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3481         }
3482
3483 out:
3484         write_unlock(&idev->lock);
3485 put:
3486         in6_dev_put(idev);
3487 }
3488
3489 /*
3490  *      Duplicate Address Detection
3491  */
3492 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3493 {
3494         unsigned long rand_num;
3495         struct inet6_dev *idev = ifp->idev;
3496
3497         if (ifp->flags & IFA_F_OPTIMISTIC)
3498                 rand_num = 0;
3499         else
3500                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3501
3502         ifp->dad_probes = idev->cnf.dad_transmits;
3503         addrconf_mod_dad_work(ifp, rand_num);
3504 }
3505
3506 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3507 {
3508         struct inet6_dev *idev = ifp->idev;
3509         struct net_device *dev = idev->dev;
3510         bool notify = false;
3511
3512         addrconf_join_solict(dev, &ifp->addr);
3513
3514         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3515
3516         read_lock_bh(&idev->lock);
3517         spin_lock(&ifp->lock);
3518         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3519                 goto out;
3520
3521         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3522             idev->cnf.accept_dad < 1 ||
3523             !(ifp->flags&IFA_F_TENTATIVE) ||
3524             ifp->flags & IFA_F_NODAD) {
3525                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3526                 spin_unlock(&ifp->lock);
3527                 read_unlock_bh(&idev->lock);
3528
3529                 addrconf_dad_completed(ifp);
3530                 return;
3531         }
3532
3533         if (!(idev->if_flags & IF_READY)) {
3534                 spin_unlock(&ifp->lock);
3535                 read_unlock_bh(&idev->lock);
3536                 /*
3537                  * If the device is not ready:
3538                  * - keep it tentative if it is a permanent address.
3539                  * - otherwise, kill it.
3540                  */
3541                 in6_ifa_hold(ifp);
3542                 addrconf_dad_stop(ifp, 0);
3543                 return;
3544         }
3545
3546         /*
3547          * Optimistic nodes can start receiving
3548          * Frames right away
3549          */
3550         if (ifp->flags & IFA_F_OPTIMISTIC) {
3551                 ip6_ins_rt(ifp->rt);
3552                 if (ipv6_use_optimistic_addr(idev)) {
3553                         /* Because optimistic nodes can use this address,
3554                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3555                          */
3556                         notify = true;
3557                 }
3558         }
3559
3560         addrconf_dad_kick(ifp);
3561 out:
3562         spin_unlock(&ifp->lock);
3563         read_unlock_bh(&idev->lock);
3564         if (notify)
3565                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3566 }
3567
3568 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3569 {
3570         bool begin_dad = false;
3571
3572         spin_lock_bh(&ifp->lock);
3573         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3574                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3575                 begin_dad = true;
3576         }
3577         spin_unlock_bh(&ifp->lock);
3578
3579         if (begin_dad)
3580                 addrconf_mod_dad_work(ifp, 0);
3581 }
3582
3583 static void addrconf_dad_work(struct work_struct *w)
3584 {
3585         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3586                                                 struct inet6_ifaddr,
3587                                                 dad_work);
3588         struct inet6_dev *idev = ifp->idev;
3589         struct in6_addr mcaddr;
3590
3591         enum {
3592                 DAD_PROCESS,
3593                 DAD_BEGIN,
3594                 DAD_ABORT,
3595         } action = DAD_PROCESS;
3596
3597         rtnl_lock();
3598
3599         spin_lock_bh(&ifp->lock);
3600         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3601                 action = DAD_BEGIN;
3602                 ifp->state = INET6_IFADDR_STATE_DAD;
3603         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3604                 action = DAD_ABORT;
3605                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3606         }
3607         spin_unlock_bh(&ifp->lock);
3608
3609         if (action == DAD_BEGIN) {
3610                 addrconf_dad_begin(ifp);
3611                 goto out;
3612         } else if (action == DAD_ABORT) {
3613                 in6_ifa_hold(ifp);
3614                 addrconf_dad_stop(ifp, 1);
3615                 goto out;
3616         }
3617
3618         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3619                 goto out;
3620
3621         write_lock_bh(&idev->lock);
3622         if (idev->dead || !(idev->if_flags & IF_READY)) {
3623                 write_unlock_bh(&idev->lock);
3624                 goto out;
3625         }
3626
3627         spin_lock(&ifp->lock);
3628         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3629                 spin_unlock(&ifp->lock);
3630                 write_unlock_bh(&idev->lock);
3631                 goto out;
3632         }
3633
3634         if (ifp->dad_probes == 0) {
3635                 /*
3636                  * DAD was successful
3637                  */
3638
3639                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3640                 spin_unlock(&ifp->lock);
3641                 write_unlock_bh(&idev->lock);
3642
3643                 addrconf_dad_completed(ifp);
3644
3645                 goto out;
3646         }
3647
3648         ifp->dad_probes--;
3649         addrconf_mod_dad_work(ifp,
3650                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3651         spin_unlock(&ifp->lock);
3652         write_unlock_bh(&idev->lock);
3653
3654         /* send a neighbour solicitation for our addr */
3655         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3656         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3657 out:
3658         in6_ifa_put(ifp);
3659         rtnl_unlock();
3660 }
3661
3662 /* ifp->idev must be at least read locked */
3663 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3664 {
3665         struct inet6_ifaddr *ifpiter;
3666         struct inet6_dev *idev = ifp->idev;
3667
3668         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3669                 if (ifpiter->scope > IFA_LINK)
3670                         break;
3671                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3672                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3673                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3674                     IFA_F_PERMANENT)
3675                         return false;
3676         }
3677         return true;
3678 }
3679
3680 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3681 {
3682         struct net_device *dev = ifp->idev->dev;
3683         struct in6_addr lladdr;
3684         bool send_rs, send_mld;
3685
3686         addrconf_del_dad_work(ifp);
3687
3688         /*
3689          *      Configure the address for reception. Now it is valid.
3690          */
3691
3692         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3693
3694         /* If added prefix is link local and we are prepared to process
3695            router advertisements, start sending router solicitations.
3696          */
3697
3698         read_lock_bh(&ifp->idev->lock);
3699         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3700         send_rs = send_mld &&
3701                   ipv6_accept_ra(ifp->idev) &&
3702                   ifp->idev->cnf.rtr_solicits > 0 &&
3703                   (dev->flags&IFF_LOOPBACK) == 0;
3704         read_unlock_bh(&ifp->idev->lock);
3705
3706         /* While dad is in progress mld report's source address is in6_addrany.
3707          * Resend with proper ll now.
3708          */
3709         if (send_mld)
3710                 ipv6_mc_dad_complete(ifp->idev);
3711
3712         if (send_rs) {
3713                 /*
3714                  *      If a host as already performed a random delay
3715                  *      [...] as part of DAD [...] there is no need
3716                  *      to delay again before sending the first RS
3717                  */
3718                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3719                         return;
3720                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3721
3722                 write_lock_bh(&ifp->idev->lock);
3723                 spin_lock(&ifp->lock);
3724                 ifp->idev->rs_probes = 1;
3725                 ifp->idev->if_flags |= IF_RS_SENT;
3726                 addrconf_mod_rs_timer(ifp->idev,
3727                                       ifp->idev->cnf.rtr_solicit_interval);
3728                 spin_unlock(&ifp->lock);
3729                 write_unlock_bh(&ifp->idev->lock);
3730         }
3731 }
3732
3733 static void addrconf_dad_run(struct inet6_dev *idev)
3734 {
3735         struct inet6_ifaddr *ifp;
3736
3737         read_lock_bh(&idev->lock);
3738         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3739                 spin_lock(&ifp->lock);
3740                 if (ifp->flags & IFA_F_TENTATIVE &&
3741                     ifp->state == INET6_IFADDR_STATE_DAD)
3742                         addrconf_dad_kick(ifp);
3743                 spin_unlock(&ifp->lock);
3744         }
3745         read_unlock_bh(&idev->lock);
3746 }
3747
3748 #ifdef CONFIG_PROC_FS
3749 struct if6_iter_state {
3750         struct seq_net_private p;
3751         int bucket;
3752         int offset;
3753 };
3754
3755 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3756 {
3757         struct inet6_ifaddr *ifa = NULL;
3758         struct if6_iter_state *state = seq->private;
3759         struct net *net = seq_file_net(seq);
3760         int p = 0;
3761
3762         /* initial bucket if pos is 0 */
3763         if (pos == 0) {
3764                 state->bucket = 0;
3765                 state->offset = 0;
3766         }
3767
3768         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3769                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3770                                          addr_lst) {
3771                         if (!net_eq(dev_net(ifa->idev->dev), net))
3772                                 continue;
3773                         /* sync with offset */
3774                         if (p < state->offset) {
3775                                 p++;
3776                                 continue;
3777                         }
3778                         state->offset++;
3779                         return ifa;
3780                 }
3781
3782                 /* prepare for next bucket */
3783                 state->offset = 0;
3784                 p = 0;
3785         }
3786         return NULL;
3787 }
3788
3789 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3790                                          struct inet6_ifaddr *ifa)
3791 {
3792         struct if6_iter_state *state = seq->private;
3793         struct net *net = seq_file_net(seq);
3794
3795         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3796                 if (!net_eq(dev_net(ifa->idev->dev), net))
3797                         continue;
3798                 state->offset++;
3799                 return ifa;
3800         }
3801
3802         while (++state->bucket < IN6_ADDR_HSIZE) {
3803                 state->offset = 0;
3804                 hlist_for_each_entry_rcu_bh(ifa,
3805                                      &inet6_addr_lst[state->bucket], addr_lst) {
3806                         if (!net_eq(dev_net(ifa->idev->dev), net))
3807                                 continue;
3808                         state->offset++;
3809                         return ifa;
3810                 }
3811         }
3812
3813         return NULL;
3814 }
3815
3816 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3817         __acquires(rcu_bh)
3818 {
3819         rcu_read_lock_bh();
3820         return if6_get_first(seq, *pos);
3821 }
3822
3823 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3824 {
3825         struct inet6_ifaddr *ifa;
3826
3827         ifa = if6_get_next(seq, v);
3828         ++*pos;
3829         return ifa;
3830 }
3831
3832 static void if6_seq_stop(struct seq_file *seq, void *v)
3833         __releases(rcu_bh)
3834 {
3835         rcu_read_unlock_bh();
3836 }
3837
3838 static int if6_seq_show(struct seq_file *seq, void *v)
3839 {
3840         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3841         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3842                    &ifp->addr,
3843                    ifp->idev->dev->ifindex,
3844                    ifp->prefix_len,
3845                    ifp->scope,
3846                    (u8) ifp->flags,
3847                    ifp->idev->dev->name);
3848         return 0;
3849 }
3850
3851 static const struct seq_operations if6_seq_ops = {
3852         .start  = if6_seq_start,
3853         .next   = if6_seq_next,
3854         .show   = if6_seq_show,
3855         .stop   = if6_seq_stop,
3856 };
3857
3858 static int if6_seq_open(struct inode *inode, struct file *file)
3859 {
3860         return seq_open_net(inode, file, &if6_seq_ops,
3861                             sizeof(struct if6_iter_state));
3862 }
3863
3864 static const struct file_operations if6_fops = {
3865         .owner          = THIS_MODULE,
3866         .open           = if6_seq_open,
3867         .read           = seq_read,
3868         .llseek         = seq_lseek,
3869         .release        = seq_release_net,
3870 };
3871
3872 static int __net_init if6_proc_net_init(struct net *net)
3873 {
3874         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3875                 return -ENOMEM;
3876         return 0;
3877 }
3878
3879 static void __net_exit if6_proc_net_exit(struct net *net)
3880 {
3881         remove_proc_entry("if_inet6", net->proc_net);
3882 }
3883
3884 static struct pernet_operations if6_proc_net_ops = {
3885         .init = if6_proc_net_init,
3886         .exit = if6_proc_net_exit,
3887 };
3888
3889 int __init if6_proc_init(void)
3890 {
3891         return register_pernet_subsys(&if6_proc_net_ops);
3892 }
3893
3894 void if6_proc_exit(void)
3895 {
3896         unregister_pernet_subsys(&if6_proc_net_ops);
3897 }
3898 #endif  /* CONFIG_PROC_FS */
3899
3900 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3901 /* Check if address is a home address configured on any interface. */
3902 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3903 {
3904         int ret = 0;
3905         struct inet6_ifaddr *ifp = NULL;
3906         unsigned int hash = inet6_addr_hash(addr);
3907
3908         rcu_read_lock_bh();
3909         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3910                 if (!net_eq(dev_net(ifp->idev->dev), net))
3911                         continue;
3912                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3913                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3914                         ret = 1;
3915                         break;
3916                 }
3917         }
3918         rcu_read_unlock_bh();
3919         return ret;
3920 }
3921 #endif
3922
3923 /*
3924  *      Periodic address status verification
3925  */
3926
3927 static void addrconf_verify_rtnl(void)
3928 {
3929         unsigned long now, next, next_sec, next_sched;
3930         struct inet6_ifaddr *ifp;
3931         int i;
3932
3933         ASSERT_RTNL();
3934
3935         rcu_read_lock_bh();
3936         now = jiffies;
3937         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3938
3939         cancel_delayed_work(&addr_chk_work);
3940
3941         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3942 restart:
3943                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3944                         unsigned long age;
3945
3946                         /* When setting preferred_lft to a value not zero or
3947                          * infinity, while valid_lft is infinity
3948                          * IFA_F_PERMANENT has a non-infinity life time.
3949                          */
3950                         if ((ifp->flags & IFA_F_PERMANENT) &&
3951                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3952                                 continue;
3953
3954                         spin_lock(&ifp->lock);
3955                         /* We try to batch several events at once. */
3956                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3957
3958                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3959                             age >= ifp->valid_lft) {
3960                                 spin_unlock(&ifp->lock);
3961                                 in6_ifa_hold(ifp);
3962                                 ipv6_del_addr(ifp);
3963                                 goto restart;
3964                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3965                                 spin_unlock(&ifp->lock);
3966                                 continue;
3967                         } else if (age >= ifp->prefered_lft) {
3968                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3969                                 int deprecate = 0;
3970
3971                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3972                                         deprecate = 1;
3973                                         ifp->flags |= IFA_F_DEPRECATED;
3974                                 }
3975
3976                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3977                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3978                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3979
3980                                 spin_unlock(&ifp->lock);
3981
3982                                 if (deprecate) {
3983                                         in6_ifa_hold(ifp);
3984
3985                                         ipv6_ifa_notify(0, ifp);
3986                                         in6_ifa_put(ifp);
3987                                         goto restart;
3988                                 }
3989                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3990                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3991                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3992                                         ifp->idev->cnf.dad_transmits *
3993                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3994
3995                                 if (age >= ifp->prefered_lft - regen_advance) {
3996                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3997                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3998                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3999                                         if (!ifp->regen_count && ifpub) {
4000                                                 ifp->regen_count++;
4001                                                 in6_ifa_hold(ifp);
4002                                                 in6_ifa_hold(ifpub);
4003                                                 spin_unlock(&ifp->lock);
4004
4005                                                 spin_lock(&ifpub->lock);
4006                                                 ifpub->regen_count = 0;
4007                                                 spin_unlock(&ifpub->lock);
4008                                                 ipv6_create_tempaddr(ifpub, ifp);
4009                                                 in6_ifa_put(ifpub);
4010                                                 in6_ifa_put(ifp);
4011                                                 goto restart;
4012                                         }
4013                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4014                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4015                                 spin_unlock(&ifp->lock);
4016                         } else {
4017                                 /* ifp->prefered_lft <= ifp->valid_lft */
4018                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4019                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4020                                 spin_unlock(&ifp->lock);
4021                         }
4022                 }
4023         }
4024
4025         next_sec = round_jiffies_up(next);
4026         next_sched = next;
4027
4028         /* If rounded timeout is accurate enough, accept it. */
4029         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4030                 next_sched = next_sec;
4031
4032         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4033         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4034                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4035
4036         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4037               now, next, next_sec, next_sched);
4038         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4039         rcu_read_unlock_bh();
4040 }
4041
4042 static void addrconf_verify_work(struct work_struct *w)
4043 {
4044         rtnl_lock();
4045         addrconf_verify_rtnl();
4046         rtnl_unlock();
4047 }
4048
4049 static void addrconf_verify(void)
4050 {
4051         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4052 }
4053
4054 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4055                                      struct in6_addr **peer_pfx)
4056 {
4057         struct in6_addr *pfx = NULL;
4058
4059         *peer_pfx = NULL;
4060
4061         if (addr)
4062                 pfx = nla_data(addr);
4063
4064         if (local) {
4065                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4066                         *peer_pfx = pfx;
4067                 pfx = nla_data(local);
4068         }
4069
4070         return pfx;
4071 }
4072
4073 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4074         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4075         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4076         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4077         [IFA_FLAGS]             = { .len = sizeof(u32) },
4078 };
4079
4080 static int
4081 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4082 {
4083         struct net *net = sock_net(skb->sk);
4084         struct ifaddrmsg *ifm;
4085         struct nlattr *tb[IFA_MAX+1];
4086         struct in6_addr *pfx, *peer_pfx;
4087         u32 ifa_flags;
4088         int err;
4089
4090         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4091         if (err < 0)
4092                 return err;
4093
4094         ifm = nlmsg_data(nlh);
4095         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4096         if (!pfx)
4097                 return -EINVAL;
4098
4099         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4100
4101         /* We ignore other flags so far. */
4102         ifa_flags &= IFA_F_MANAGETEMPADDR;
4103
4104         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4105                               ifm->ifa_prefixlen);
4106 }
4107
4108 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4109                              u32 prefered_lft, u32 valid_lft)
4110 {
4111         u32 flags;
4112         clock_t expires;
4113         unsigned long timeout;
4114         bool was_managetempaddr;
4115         bool had_prefixroute;
4116
4117         ASSERT_RTNL();
4118
4119         if (!valid_lft || (prefered_lft > valid_lft))
4120                 return -EINVAL;
4121
4122         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4123             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4124                 return -EINVAL;
4125
4126         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4127         if (addrconf_finite_timeout(timeout)) {
4128                 expires = jiffies_to_clock_t(timeout * HZ);
4129                 valid_lft = timeout;
4130                 flags = RTF_EXPIRES;
4131         } else {
4132                 expires = 0;
4133                 flags = 0;
4134                 ifa_flags |= IFA_F_PERMANENT;
4135         }
4136
4137         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4138         if (addrconf_finite_timeout(timeout)) {
4139                 if (timeout == 0)
4140                         ifa_flags |= IFA_F_DEPRECATED;
4141                 prefered_lft = timeout;
4142         }
4143
4144         spin_lock_bh(&ifp->lock);
4145         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4146         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4147                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4148         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4149                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4150                         IFA_F_NOPREFIXROUTE);
4151         ifp->flags |= ifa_flags;
4152         ifp->tstamp = jiffies;
4153         ifp->valid_lft = valid_lft;
4154         ifp->prefered_lft = prefered_lft;
4155
4156         spin_unlock_bh(&ifp->lock);
4157         if (!(ifp->flags&IFA_F_TENTATIVE))
4158                 ipv6_ifa_notify(0, ifp);
4159
4160         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4161                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4162                                       expires, flags);
4163         } else if (had_prefixroute) {
4164                 enum cleanup_prefix_rt_t action;
4165                 unsigned long rt_expires;
4166
4167                 write_lock_bh(&ifp->idev->lock);
4168                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4169                 write_unlock_bh(&ifp->idev->lock);
4170
4171                 if (action != CLEANUP_PREFIX_RT_NOP) {
4172                         cleanup_prefix_route(ifp, rt_expires,
4173                                 action == CLEANUP_PREFIX_RT_DEL);
4174                 }
4175         }
4176
4177         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4178                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4179                         valid_lft = prefered_lft = 0;
4180                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4181                                  !was_managetempaddr, jiffies);
4182         }
4183
4184         addrconf_verify_rtnl();
4185
4186         return 0;
4187 }
4188
4189 static int
4190 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4191 {
4192         struct net *net = sock_net(skb->sk);
4193         struct ifaddrmsg *ifm;
4194         struct nlattr *tb[IFA_MAX+1];
4195         struct in6_addr *pfx, *peer_pfx;
4196         struct inet6_ifaddr *ifa;
4197         struct net_device *dev;
4198         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4199         u32 ifa_flags;
4200         int err;
4201
4202         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4203         if (err < 0)
4204                 return err;
4205
4206         ifm = nlmsg_data(nlh);
4207         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4208         if (!pfx)
4209                 return -EINVAL;
4210
4211         if (tb[IFA_CACHEINFO]) {
4212                 struct ifa_cacheinfo *ci;
4213
4214                 ci = nla_data(tb[IFA_CACHEINFO]);
4215                 valid_lft = ci->ifa_valid;
4216                 preferred_lft = ci->ifa_prefered;
4217         } else {
4218                 preferred_lft = INFINITY_LIFE_TIME;
4219                 valid_lft = INFINITY_LIFE_TIME;
4220         }
4221
4222         dev =  __dev_get_by_index(net, ifm->ifa_index);
4223         if (!dev)
4224                 return -ENODEV;
4225
4226         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4227
4228         /* We ignore other flags so far. */
4229         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4230                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4231
4232         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4233         if (!ifa) {
4234                 /*
4235                  * It would be best to check for !NLM_F_CREATE here but
4236                  * userspace already relies on not having to provide this.
4237                  */
4238                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4239                                       ifm->ifa_prefixlen, ifa_flags,
4240                                       preferred_lft, valid_lft);
4241         }
4242
4243         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4244             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4245                 err = -EEXIST;
4246         else
4247                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4248
4249         in6_ifa_put(ifa);
4250
4251         return err;
4252 }
4253
4254 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4255                           u8 scope, int ifindex)
4256 {
4257         struct ifaddrmsg *ifm;
4258
4259         ifm = nlmsg_data(nlh);
4260         ifm->ifa_family = AF_INET6;
4261         ifm->ifa_prefixlen = prefixlen;
4262         ifm->ifa_flags = flags;
4263         ifm->ifa_scope = scope;
4264         ifm->ifa_index = ifindex;
4265 }
4266
4267 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4268                          unsigned long tstamp, u32 preferred, u32 valid)
4269 {
4270         struct ifa_cacheinfo ci;
4271
4272         ci.cstamp = cstamp_delta(cstamp);
4273         ci.tstamp = cstamp_delta(tstamp);
4274         ci.ifa_prefered = preferred;
4275         ci.ifa_valid = valid;
4276
4277         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4278 }
4279
4280 static inline int rt_scope(int ifa_scope)
4281 {
4282         if (ifa_scope & IFA_HOST)
4283                 return RT_SCOPE_HOST;
4284         else if (ifa_scope & IFA_LINK)
4285                 return RT_SCOPE_LINK;
4286         else if (ifa_scope & IFA_SITE)
4287                 return RT_SCOPE_SITE;
4288         else
4289                 return RT_SCOPE_UNIVERSE;
4290 }
4291
4292 static inline int inet6_ifaddr_msgsize(void)
4293 {
4294         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4295                + nla_total_size(16) /* IFA_LOCAL */
4296                + nla_total_size(16) /* IFA_ADDRESS */
4297                + nla_total_size(sizeof(struct ifa_cacheinfo))
4298                + nla_total_size(4)  /* IFA_FLAGS */;
4299 }
4300
4301 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4302                              u32 portid, u32 seq, int event, unsigned int flags)
4303 {
4304         struct nlmsghdr  *nlh;
4305         u32 preferred, valid;
4306
4307         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4308         if (!nlh)
4309                 return -EMSGSIZE;
4310
4311         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4312                       ifa->idev->dev->ifindex);
4313
4314         if (!((ifa->flags&IFA_F_PERMANENT) &&
4315               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4316                 preferred = ifa->prefered_lft;
4317                 valid = ifa->valid_lft;
4318                 if (preferred != INFINITY_LIFE_TIME) {
4319                         long tval = (jiffies - ifa->tstamp)/HZ;
4320                         if (preferred > tval)
4321                                 preferred -= tval;
4322                         else
4323                                 preferred = 0;
4324                         if (valid != INFINITY_LIFE_TIME) {
4325                                 if (valid > tval)
4326                                         valid -= tval;
4327                                 else
4328                                         valid = 0;
4329                         }
4330                 }
4331         } else {
4332                 preferred = INFINITY_LIFE_TIME;
4333                 valid = INFINITY_LIFE_TIME;
4334         }
4335
4336         if (!ipv6_addr_any(&ifa->peer_addr)) {
4337                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4338                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4339                         goto error;
4340         } else
4341                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4342                         goto error;
4343
4344         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4345                 goto error;
4346
4347         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4348                 goto error;
4349
4350         nlmsg_end(skb, nlh);
4351         return 0;
4352
4353 error:
4354         nlmsg_cancel(skb, nlh);
4355         return -EMSGSIZE;
4356 }
4357
4358 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4359                                 u32 portid, u32 seq, int event, u16 flags)
4360 {
4361         struct nlmsghdr  *nlh;
4362         u8 scope = RT_SCOPE_UNIVERSE;
4363         int ifindex = ifmca->idev->dev->ifindex;
4364
4365         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4366                 scope = RT_SCOPE_SITE;
4367
4368         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4369         if (!nlh)
4370                 return -EMSGSIZE;
4371
4372         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4373         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4374             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4375                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4376                 nlmsg_cancel(skb, nlh);
4377                 return -EMSGSIZE;
4378         }
4379
4380         nlmsg_end(skb, nlh);
4381         return 0;
4382 }
4383
4384 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4385                                 u32 portid, u32 seq, int event, unsigned int flags)
4386 {
4387         struct nlmsghdr  *nlh;
4388         u8 scope = RT_SCOPE_UNIVERSE;
4389         int ifindex = ifaca->aca_idev->dev->ifindex;
4390
4391         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4392                 scope = RT_SCOPE_SITE;
4393
4394         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4395         if (!nlh)
4396                 return -EMSGSIZE;
4397
4398         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4399         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4400             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4401                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4402                 nlmsg_cancel(skb, nlh);
4403                 return -EMSGSIZE;
4404         }
4405
4406         nlmsg_end(skb, nlh);
4407         return 0;
4408 }
4409
4410 enum addr_type_t {
4411         UNICAST_ADDR,
4412         MULTICAST_ADDR,
4413         ANYCAST_ADDR,
4414 };
4415
4416 /* called with rcu_read_lock() */
4417 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4418                           struct netlink_callback *cb, enum addr_type_t type,
4419                           int s_ip_idx, int *p_ip_idx)
4420 {
4421         struct ifmcaddr6 *ifmca;
4422         struct ifacaddr6 *ifaca;
4423         int err = 1;
4424         int ip_idx = *p_ip_idx;
4425
4426         read_lock_bh(&idev->lock);
4427         switch (type) {
4428         case UNICAST_ADDR: {
4429                 struct inet6_ifaddr *ifa;
4430
4431                 /* unicast address incl. temp addr */
4432                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4433                         if (++ip_idx < s_ip_idx)
4434                                 continue;
4435                         err = inet6_fill_ifaddr(skb, ifa,
4436                                                 NETLINK_CB(cb->skb).portid,
4437                                                 cb->nlh->nlmsg_seq,
4438                                                 RTM_NEWADDR,
4439                                                 NLM_F_MULTI);
4440                         if (err < 0)
4441                                 break;
4442                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4443                 }
4444                 break;
4445         }
4446         case MULTICAST_ADDR:
4447                 /* multicast address */
4448                 for (ifmca = idev->mc_list; ifmca;
4449                      ifmca = ifmca->next, ip_idx++) {
4450                         if (ip_idx < s_ip_idx)
4451                                 continue;
4452                         err = inet6_fill_ifmcaddr(skb, ifmca,
4453                                                   NETLINK_CB(cb->skb).portid,
4454                                                   cb->nlh->nlmsg_seq,
4455                                                   RTM_GETMULTICAST,
4456                                                   NLM_F_MULTI);
4457                         if (err < 0)
4458                                 break;
4459                 }
4460                 break;
4461         case ANYCAST_ADDR:
4462                 /* anycast address */
4463                 for (ifaca = idev->ac_list; ifaca;
4464                      ifaca = ifaca->aca_next, ip_idx++) {
4465                         if (ip_idx < s_ip_idx)
4466                                 continue;
4467                         err = inet6_fill_ifacaddr(skb, ifaca,
4468                                                   NETLINK_CB(cb->skb).portid,
4469                                                   cb->nlh->nlmsg_seq,
4470                                                   RTM_GETANYCAST,
4471                                                   NLM_F_MULTI);
4472                         if (err < 0)
4473                                 break;
4474                 }
4475                 break;
4476         default:
4477                 break;
4478         }
4479         read_unlock_bh(&idev->lock);
4480         *p_ip_idx = ip_idx;
4481         return err;
4482 }
4483
4484 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4485                            enum addr_type_t type)
4486 {
4487         struct net *net = sock_net(skb->sk);
4488         int h, s_h;
4489         int idx, ip_idx;
4490         int s_idx, s_ip_idx;
4491         struct net_device *dev;
4492         struct inet6_dev *idev;
4493         struct hlist_head *head;
4494
4495         s_h = cb->args[0];
4496         s_idx = idx = cb->args[1];
4497         s_ip_idx = ip_idx = cb->args[2];
4498
4499         rcu_read_lock();
4500         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4501         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4502                 idx = 0;
4503                 head = &net->dev_index_head[h];
4504                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4505                         if (idx < s_idx)
4506                                 goto cont;
4507                         if (h > s_h || idx > s_idx)
4508                                 s_ip_idx = 0;
4509                         ip_idx = 0;
4510                         idev = __in6_dev_get(dev);
4511                         if (!idev)
4512                                 goto cont;
4513
4514                         if (in6_dump_addrs(idev, skb, cb, type,
4515                                            s_ip_idx, &ip_idx) < 0)
4516                                 goto done;
4517 cont:
4518                         idx++;
4519                 }
4520         }
4521 done:
4522         rcu_read_unlock();
4523         cb->args[0] = h;
4524         cb->args[1] = idx;
4525         cb->args[2] = ip_idx;
4526
4527         return skb->len;
4528 }
4529
4530 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4531 {
4532         enum addr_type_t type = UNICAST_ADDR;
4533
4534         return inet6_dump_addr(skb, cb, type);
4535 }
4536
4537 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4538 {
4539         enum addr_type_t type = MULTICAST_ADDR;
4540
4541         return inet6_dump_addr(skb, cb, type);
4542 }
4543
4544
4545 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4546 {
4547         enum addr_type_t type = ANYCAST_ADDR;
4548
4549         return inet6_dump_addr(skb, cb, type);
4550 }
4551
4552 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4553 {
4554         struct net *net = sock_net(in_skb->sk);
4555         struct ifaddrmsg *ifm;
4556         struct nlattr *tb[IFA_MAX+1];
4557         struct in6_addr *addr = NULL, *peer;
4558         struct net_device *dev = NULL;
4559         struct inet6_ifaddr *ifa;
4560         struct sk_buff *skb;
4561         int err;
4562
4563         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4564         if (err < 0)
4565                 goto errout;
4566
4567         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4568         if (!addr) {
4569                 err = -EINVAL;
4570                 goto errout;
4571         }
4572
4573         ifm = nlmsg_data(nlh);
4574         if (ifm->ifa_index)
4575                 dev = __dev_get_by_index(net, ifm->ifa_index);
4576
4577         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4578         if (!ifa) {
4579                 err = -EADDRNOTAVAIL;
4580                 goto errout;
4581         }
4582
4583         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4584         if (!skb) {
4585                 err = -ENOBUFS;
4586                 goto errout_ifa;
4587         }
4588
4589         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4590                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4591         if (err < 0) {
4592                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4593                 WARN_ON(err == -EMSGSIZE);
4594                 kfree_skb(skb);
4595                 goto errout_ifa;
4596         }
4597         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4598 errout_ifa:
4599         in6_ifa_put(ifa);
4600 errout:
4601         return err;
4602 }
4603
4604 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4605 {
4606         struct sk_buff *skb;
4607         struct net *net = dev_net(ifa->idev->dev);
4608         int err = -ENOBUFS;
4609
4610         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4611         if (!skb)
4612                 goto errout;
4613
4614         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4615         if (err < 0) {
4616                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4617                 WARN_ON(err == -EMSGSIZE);
4618                 kfree_skb(skb);
4619                 goto errout;
4620         }
4621         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4622         return;
4623 errout:
4624         if (err < 0)
4625                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4626 }
4627
4628 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4629                                 __s32 *array, int bytes)
4630 {
4631         BUG_ON(bytes < (DEVCONF_MAX * 4));
4632
4633         memset(array, 0, bytes);
4634         array[DEVCONF_FORWARDING] = cnf->forwarding;
4635         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4636         array[DEVCONF_MTU6] = cnf->mtu6;
4637         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4638         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4639         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4640         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4641         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4642         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4643                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4644         array[DEVCONF_RTR_SOLICIT_DELAY] =
4645                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4646         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4647         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4648                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4649         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4650                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4651         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4652         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4653         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4654         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4655         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4656         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4657         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4658         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4659         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4660 #ifdef CONFIG_IPV6_ROUTER_PREF
4661         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4662         array[DEVCONF_RTR_PROBE_INTERVAL] =
4663                 jiffies_to_msecs(cnf->rtr_probe_interval);
4664 #ifdef CONFIG_IPV6_ROUTE_INFO
4665         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4666 #endif
4667 #endif
4668         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4669         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4670 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4671         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4672         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4673 #endif
4674 #ifdef CONFIG_IPV6_MROUTE
4675         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4676 #endif
4677         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4678         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4679         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4680         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4681         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4682         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4683         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4684         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4685         /* we omit DEVCONF_STABLE_SECRET for now */
4686         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4687 }
4688
4689 static inline size_t inet6_ifla6_size(void)
4690 {
4691         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4692              + nla_total_size(sizeof(struct ifla_cacheinfo))
4693              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4694              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4695              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4696              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4697 }
4698
4699 static inline size_t inet6_if_nlmsg_size(void)
4700 {
4701         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4702                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4703                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4704                + nla_total_size(4) /* IFLA_MTU */
4705                + nla_total_size(4) /* IFLA_LINK */
4706                + nla_total_size(1) /* IFLA_OPERSTATE */
4707                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4708 }
4709
4710 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4711                                       int items, int bytes)
4712 {
4713         int i;
4714         int pad = bytes - sizeof(u64) * items;
4715         BUG_ON(pad < 0);
4716
4717         /* Use put_unaligned() because stats may not be aligned for u64. */
4718         put_unaligned(items, &stats[0]);
4719         for (i = 1; i < items; i++)
4720                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4721
4722         memset(&stats[items], 0, pad);
4723 }
4724
4725 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4726                                         int bytes, size_t syncpoff)
4727 {
4728         int i, c;
4729         u64 buff[IPSTATS_MIB_MAX];
4730         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4731
4732         BUG_ON(pad < 0);
4733
4734         memset(buff, 0, sizeof(buff));
4735         buff[0] = IPSTATS_MIB_MAX;
4736
4737         for_each_possible_cpu(c) {
4738                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4739                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4740         }
4741
4742         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4743         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4744 }
4745
4746 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4747                              int bytes)
4748 {
4749         switch (attrtype) {
4750         case IFLA_INET6_STATS:
4751                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4752                                      offsetof(struct ipstats_mib, syncp));
4753                 break;
4754         case IFLA_INET6_ICMP6STATS:
4755                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4756                 break;
4757         }
4758 }
4759
4760 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4761                                   u32 ext_filter_mask)
4762 {
4763         struct nlattr *nla;
4764         struct ifla_cacheinfo ci;
4765
4766         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4767                 goto nla_put_failure;
4768         ci.max_reasm_len = IPV6_MAXPLEN;
4769         ci.tstamp = cstamp_delta(idev->tstamp);
4770         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4771         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4772         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4773                 goto nla_put_failure;
4774         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4775         if (!nla)
4776                 goto nla_put_failure;
4777         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4778
4779         /* XXX - MC not implemented */
4780
4781         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4782                 return 0;
4783
4784         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4785         if (!nla)
4786                 goto nla_put_failure;
4787         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4788
4789         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4790         if (!nla)
4791                 goto nla_put_failure;
4792         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4793
4794         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4795         if (!nla)
4796                 goto nla_put_failure;
4797
4798         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4799                 goto nla_put_failure;
4800
4801         read_lock_bh(&idev->lock);
4802         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4803         read_unlock_bh(&idev->lock);
4804
4805         return 0;
4806
4807 nla_put_failure:
4808         return -EMSGSIZE;
4809 }
4810
4811 static size_t inet6_get_link_af_size(const struct net_device *dev,
4812                                      u32 ext_filter_mask)
4813 {
4814         if (!__in6_dev_get(dev))
4815                 return 0;
4816
4817         return inet6_ifla6_size();
4818 }
4819
4820 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4821                               u32 ext_filter_mask)
4822 {
4823         struct inet6_dev *idev = __in6_dev_get(dev);
4824
4825         if (!idev)
4826                 return -ENODATA;
4827
4828         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4829                 return -EMSGSIZE;
4830
4831         return 0;
4832 }
4833
4834 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4835 {
4836         struct inet6_ifaddr *ifp;
4837         struct net_device *dev = idev->dev;
4838         bool update_rs = false;
4839         struct in6_addr ll_addr;
4840
4841         ASSERT_RTNL();
4842
4843         if (!token)
4844                 return -EINVAL;
4845         if (ipv6_addr_any(token))
4846                 return -EINVAL;
4847         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4848                 return -EINVAL;
4849         if (!ipv6_accept_ra(idev))
4850                 return -EINVAL;
4851         if (idev->cnf.rtr_solicits <= 0)
4852                 return -EINVAL;
4853
4854         write_lock_bh(&idev->lock);
4855
4856         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4857         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4858
4859         write_unlock_bh(&idev->lock);
4860
4861         if (!idev->dead && (idev->if_flags & IF_READY) &&
4862             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4863                              IFA_F_OPTIMISTIC)) {
4864
4865                 /* If we're not ready, then normal ifup will take care
4866                  * of this. Otherwise, we need to request our rs here.
4867                  */
4868                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4869                 update_rs = true;
4870         }
4871
4872         write_lock_bh(&idev->lock);
4873
4874         if (update_rs) {
4875                 idev->if_flags |= IF_RS_SENT;
4876                 idev->rs_probes = 1;
4877                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4878         }
4879
4880         /* Well, that's kinda nasty ... */
4881         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4882                 spin_lock(&ifp->lock);
4883                 if (ifp->tokenized) {
4884                         ifp->valid_lft = 0;
4885                         ifp->prefered_lft = 0;
4886                 }
4887                 spin_unlock(&ifp->lock);
4888         }
4889
4890         write_unlock_bh(&idev->lock);
4891         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4892         addrconf_verify_rtnl();
4893         return 0;
4894 }
4895
4896 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4897         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4898         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4899 };
4900
4901 static int inet6_validate_link_af(const struct net_device *dev,
4902                                   const struct nlattr *nla)
4903 {
4904         struct nlattr *tb[IFLA_INET6_MAX + 1];
4905
4906         if (dev && !__in6_dev_get(dev))
4907                 return -EAFNOSUPPORT;
4908
4909         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4910 }
4911
4912 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4913 {
4914         int err = -EINVAL;
4915         struct inet6_dev *idev = __in6_dev_get(dev);
4916         struct nlattr *tb[IFLA_INET6_MAX + 1];
4917
4918         if (!idev)
4919                 return -EAFNOSUPPORT;
4920
4921         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4922                 BUG();
4923
4924         if (tb[IFLA_INET6_TOKEN]) {
4925                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4926                 if (err)
4927                         return err;
4928         }
4929
4930         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4931                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4932
4933                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4934                     mode != IN6_ADDR_GEN_MODE_NONE &&
4935                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4936                         return -EINVAL;
4937
4938                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4939                     !idev->cnf.stable_secret.initialized &&
4940                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4941                         return -EINVAL;
4942
4943                 idev->addr_gen_mode = mode;
4944                 err = 0;
4945         }
4946
4947         return err;
4948 }
4949
4950 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4951                              u32 portid, u32 seq, int event, unsigned int flags)
4952 {
4953         struct net_device *dev = idev->dev;
4954         struct ifinfomsg *hdr;
4955         struct nlmsghdr *nlh;
4956         void *protoinfo;
4957
4958         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4959         if (!nlh)
4960                 return -EMSGSIZE;
4961
4962         hdr = nlmsg_data(nlh);
4963         hdr->ifi_family = AF_INET6;
4964         hdr->__ifi_pad = 0;
4965         hdr->ifi_type = dev->type;
4966         hdr->ifi_index = dev->ifindex;
4967         hdr->ifi_flags = dev_get_flags(dev);
4968         hdr->ifi_change = 0;
4969
4970         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4971             (dev->addr_len &&
4972              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4973             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4974             (dev->ifindex != dev_get_iflink(dev) &&
4975              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
4976             nla_put_u8(skb, IFLA_OPERSTATE,
4977                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
4978                 goto nla_put_failure;
4979         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4980         if (!protoinfo)
4981                 goto nla_put_failure;
4982
4983         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
4984                 goto nla_put_failure;
4985
4986         nla_nest_end(skb, protoinfo);
4987         nlmsg_end(skb, nlh);
4988         return 0;
4989
4990 nla_put_failure:
4991         nlmsg_cancel(skb, nlh);
4992         return -EMSGSIZE;
4993 }
4994
4995 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4996 {
4997         struct net *net = sock_net(skb->sk);
4998         int h, s_h;
4999         int idx = 0, s_idx;
5000         struct net_device *dev;
5001         struct inet6_dev *idev;
5002         struct hlist_head *head;
5003
5004         s_h = cb->args[0];
5005         s_idx = cb->args[1];
5006
5007         rcu_read_lock();
5008         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5009                 idx = 0;
5010                 head = &net->dev_index_head[h];
5011                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5012                         if (idx < s_idx)
5013                                 goto cont;
5014                         idev = __in6_dev_get(dev);
5015                         if (!idev)
5016                                 goto cont;
5017                         if (inet6_fill_ifinfo(skb, idev,
5018                                               NETLINK_CB(cb->skb).portid,
5019                                               cb->nlh->nlmsg_seq,
5020                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5021                                 goto out;
5022 cont:
5023                         idx++;
5024                 }
5025         }
5026 out:
5027         rcu_read_unlock();
5028         cb->args[1] = idx;
5029         cb->args[0] = h;
5030
5031         return skb->len;
5032 }
5033
5034 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5035 {
5036         struct sk_buff *skb;
5037         struct net *net = dev_net(idev->dev);
5038         int err = -ENOBUFS;
5039
5040         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5041         if (!skb)
5042                 goto errout;
5043
5044         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5045         if (err < 0) {
5046                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5047                 WARN_ON(err == -EMSGSIZE);
5048                 kfree_skb(skb);
5049                 goto errout;
5050         }
5051         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5052         return;
5053 errout:
5054         if (err < 0)
5055                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5056 }
5057
5058 static inline size_t inet6_prefix_nlmsg_size(void)
5059 {
5060         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5061                + nla_total_size(sizeof(struct in6_addr))
5062                + nla_total_size(sizeof(struct prefix_cacheinfo));
5063 }
5064
5065 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5066                              struct prefix_info *pinfo, u32 portid, u32 seq,
5067                              int event, unsigned int flags)
5068 {
5069         struct prefixmsg *pmsg;
5070         struct nlmsghdr *nlh;
5071         struct prefix_cacheinfo ci;
5072
5073         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5074         if (!nlh)
5075                 return -EMSGSIZE;
5076
5077         pmsg = nlmsg_data(nlh);
5078         pmsg->prefix_family = AF_INET6;
5079         pmsg->prefix_pad1 = 0;
5080         pmsg->prefix_pad2 = 0;
5081         pmsg->prefix_ifindex = idev->dev->ifindex;
5082         pmsg->prefix_len = pinfo->prefix_len;
5083         pmsg->prefix_type = pinfo->type;
5084         pmsg->prefix_pad3 = 0;
5085         pmsg->prefix_flags = 0;
5086         if (pinfo->onlink)
5087                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5088         if (pinfo->autoconf)
5089                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5090
5091         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5092                 goto nla_put_failure;
5093         ci.preferred_time = ntohl(pinfo->prefered);
5094         ci.valid_time = ntohl(pinfo->valid);
5095         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5096                 goto nla_put_failure;
5097         nlmsg_end(skb, nlh);
5098         return 0;
5099
5100 nla_put_failure:
5101         nlmsg_cancel(skb, nlh);
5102         return -EMSGSIZE;
5103 }
5104
5105 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5106                          struct prefix_info *pinfo)
5107 {
5108         struct sk_buff *skb;
5109         struct net *net = dev_net(idev->dev);
5110         int err = -ENOBUFS;
5111
5112         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5113         if (!skb)
5114                 goto errout;
5115
5116         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5117         if (err < 0) {
5118                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5119                 WARN_ON(err == -EMSGSIZE);
5120                 kfree_skb(skb);
5121                 goto errout;
5122         }
5123         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5124         return;
5125 errout:
5126         if (err < 0)
5127                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5128 }
5129
5130 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5131 {
5132         struct net *net = dev_net(ifp->idev->dev);
5133
5134         if (event)
5135                 ASSERT_RTNL();
5136
5137         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5138
5139         switch (event) {
5140         case RTM_NEWADDR:
5141                 /*
5142                  * If the address was optimistic
5143                  * we inserted the route at the start of
5144                  * our DAD process, so we don't need
5145                  * to do it again
5146                  */
5147                 if (!(ifp->rt->rt6i_node))
5148                         ip6_ins_rt(ifp->rt);
5149                 if (ifp->idev->cnf.forwarding)
5150                         addrconf_join_anycast(ifp);
5151                 if (!ipv6_addr_any(&ifp->peer_addr))
5152                         addrconf_prefix_route(&ifp->peer_addr, 128,
5153                                               ifp->idev->dev, 0, 0);
5154                 break;
5155         case RTM_DELADDR:
5156                 if (ifp->idev->cnf.forwarding)
5157                         addrconf_leave_anycast(ifp);
5158                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5159                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5160                         struct rt6_info *rt;
5161
5162                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5163                                                        ifp->idev->dev, 0, 0);
5164                         if (rt)
5165                                 ip6_del_rt(rt);
5166                 }
5167                 dst_hold(&ifp->rt->dst);
5168
5169                 ip6_del_rt(ifp->rt);
5170
5171                 rt_genid_bump_ipv6(net);
5172                 break;
5173         }
5174         atomic_inc(&net->ipv6.dev_addr_genid);
5175 }
5176
5177 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5178 {
5179         rcu_read_lock_bh();
5180         if (likely(ifp->idev->dead == 0))
5181                 __ipv6_ifa_notify(event, ifp);
5182         rcu_read_unlock_bh();
5183 }
5184
5185 #ifdef CONFIG_SYSCTL
5186
5187 static
5188 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5189                            void __user *buffer, size_t *lenp, loff_t *ppos)
5190 {
5191         int *valp = ctl->data;
5192         int val = *valp;
5193         loff_t pos = *ppos;
5194         struct ctl_table lctl;
5195         int ret;
5196
5197         /*
5198          * ctl->data points to idev->cnf.forwarding, we should
5199          * not modify it until we get the rtnl lock.
5200          */
5201         lctl = *ctl;
5202         lctl.data = &val;
5203
5204         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5205
5206         if (write)
5207                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5208         if (ret)
5209                 *ppos = pos;
5210         return ret;
5211 }
5212
5213 static
5214 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5215                         void __user *buffer, size_t *lenp, loff_t *ppos)
5216 {
5217         struct inet6_dev *idev = ctl->extra1;
5218         int min_mtu = IPV6_MIN_MTU;
5219         struct ctl_table lctl;
5220
5221         lctl = *ctl;
5222         lctl.extra1 = &min_mtu;
5223         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5224
5225         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5226 }
5227
5228 static void dev_disable_change(struct inet6_dev *idev)
5229 {
5230         struct netdev_notifier_info info;
5231
5232         if (!idev || !idev->dev)
5233                 return;
5234
5235         netdev_notifier_info_init(&info, idev->dev);
5236         if (idev->cnf.disable_ipv6)
5237                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5238         else
5239                 addrconf_notify(NULL, NETDEV_UP, &info);
5240 }
5241
5242 static void addrconf_disable_change(struct net *net, __s32 newf)
5243 {
5244         struct net_device *dev;
5245         struct inet6_dev *idev;
5246
5247         for_each_netdev(net, dev) {
5248                 idev = __in6_dev_get(dev);
5249                 if (idev) {
5250                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5251                         idev->cnf.disable_ipv6 = newf;
5252                         if (changed)
5253                                 dev_disable_change(idev);
5254                 }
5255         }
5256 }
5257
5258 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5259 {
5260         struct net *net;
5261         int old;
5262
5263         if (!rtnl_trylock())
5264                 return restart_syscall();
5265
5266         net = (struct net *)table->extra2;
5267         old = *p;
5268         *p = newf;
5269
5270         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5271                 rtnl_unlock();
5272                 return 0;
5273         }
5274
5275         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5276                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5277                 addrconf_disable_change(net, newf);
5278         } else if ((!newf) ^ (!old))
5279                 dev_disable_change((struct inet6_dev *)table->extra1);
5280
5281         rtnl_unlock();
5282         return 0;
5283 }
5284
5285 static
5286 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5287                             void __user *buffer, size_t *lenp, loff_t *ppos)
5288 {
5289         int *valp = ctl->data;
5290         int val = *valp;
5291         loff_t pos = *ppos;
5292         struct ctl_table lctl;
5293         int ret;
5294
5295         /*
5296          * ctl->data points to idev->cnf.disable_ipv6, we should
5297          * not modify it until we get the rtnl lock.
5298          */
5299         lctl = *ctl;
5300         lctl.data = &val;
5301
5302         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5303
5304         if (write)
5305                 ret = addrconf_disable_ipv6(ctl, valp, val);
5306         if (ret)
5307                 *ppos = pos;
5308         return ret;
5309 }
5310
5311 static
5312 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5313                               void __user *buffer, size_t *lenp, loff_t *ppos)
5314 {
5315         int *valp = ctl->data;
5316         int ret;
5317         int old, new;
5318
5319         old = *valp;
5320         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5321         new = *valp;
5322
5323         if (write && old != new) {
5324                 struct net *net = ctl->extra2;
5325
5326                 if (!rtnl_trylock())
5327                         return restart_syscall();
5328
5329                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5330                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5331                                                      NETCONFA_IFINDEX_DEFAULT,
5332                                                      net->ipv6.devconf_dflt);
5333                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5334                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5335                                                      NETCONFA_IFINDEX_ALL,
5336                                                      net->ipv6.devconf_all);
5337                 else {
5338                         struct inet6_dev *idev = ctl->extra1;
5339
5340                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5341                                                      idev->dev->ifindex,
5342                                                      &idev->cnf);
5343                 }
5344                 rtnl_unlock();
5345         }
5346
5347         return ret;
5348 }
5349
5350 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5351                                          void __user *buffer, size_t *lenp,
5352                                          loff_t *ppos)
5353 {
5354         int err;
5355         struct in6_addr addr;
5356         char str[IPV6_MAX_STRLEN];
5357         struct ctl_table lctl = *ctl;
5358         struct net *net = ctl->extra2;
5359         struct ipv6_stable_secret *secret = ctl->data;
5360
5361         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5362                 return -EIO;
5363
5364         lctl.maxlen = IPV6_MAX_STRLEN;
5365         lctl.data = str;
5366
5367         if (!rtnl_trylock())
5368                 return restart_syscall();
5369
5370         if (!write && !secret->initialized) {
5371                 err = -EIO;
5372                 goto out;
5373         }
5374
5375         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5376         if (err >= sizeof(str)) {
5377                 err = -EIO;
5378                 goto out;
5379         }
5380
5381         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5382         if (err || !write)
5383                 goto out;
5384
5385         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5386                 err = -EIO;
5387                 goto out;
5388         }
5389
5390         secret->initialized = true;
5391         secret->secret = addr;
5392
5393         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5394                 struct net_device *dev;
5395
5396                 for_each_netdev(net, dev) {
5397                         struct inet6_dev *idev = __in6_dev_get(dev);
5398
5399                         if (idev) {
5400                                 idev->addr_gen_mode =
5401                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5402                         }
5403                 }
5404         } else {
5405                 struct inet6_dev *idev = ctl->extra1;
5406
5407                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5408         }
5409
5410 out:
5411         rtnl_unlock();
5412
5413         return err;
5414 }
5415
5416 static
5417 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5418                                                 int write,
5419                                                 void __user *buffer,
5420                                                 size_t *lenp,
5421                                                 loff_t *ppos)
5422 {
5423         int *valp = ctl->data;
5424         int val = *valp;
5425         loff_t pos = *ppos;
5426         struct ctl_table lctl;
5427         int ret;
5428
5429         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5430          * we should not modify it until we get the rtnl lock.
5431          */
5432         lctl = *ctl;
5433         lctl.data = &val;
5434
5435         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5436
5437         if (write)
5438                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5439         if (ret)
5440                 *ppos = pos;
5441         return ret;
5442 }
5443
5444 static struct addrconf_sysctl_table
5445 {
5446         struct ctl_table_header *sysctl_header;
5447         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5448 } addrconf_sysctl __read_mostly = {
5449         .sysctl_header = NULL,
5450         .addrconf_vars = {
5451                 {
5452                         .procname       = "forwarding",
5453                         .data           = &ipv6_devconf.forwarding,
5454                         .maxlen         = sizeof(int),
5455                         .mode           = 0644,
5456                         .proc_handler   = addrconf_sysctl_forward,
5457                 },
5458                 {
5459                         .procname       = "hop_limit",
5460                         .data           = &ipv6_devconf.hop_limit,
5461                         .maxlen         = sizeof(int),
5462                         .mode           = 0644,
5463                         .proc_handler   = proc_dointvec,
5464                 },
5465                 {
5466                         .procname       = "mtu",
5467                         .data           = &ipv6_devconf.mtu6,
5468                         .maxlen         = sizeof(int),
5469                         .mode           = 0644,
5470                         .proc_handler   = addrconf_sysctl_mtu,
5471                 },
5472                 {
5473                         .procname       = "accept_ra",
5474                         .data           = &ipv6_devconf.accept_ra,
5475                         .maxlen         = sizeof(int),
5476                         .mode           = 0644,
5477                         .proc_handler   = proc_dointvec,
5478                 },
5479                 {
5480                         .procname       = "accept_redirects",
5481                         .data           = &ipv6_devconf.accept_redirects,
5482                         .maxlen         = sizeof(int),
5483                         .mode           = 0644,
5484                         .proc_handler   = proc_dointvec,
5485                 },
5486                 {
5487                         .procname       = "autoconf",
5488                         .data           = &ipv6_devconf.autoconf,
5489                         .maxlen         = sizeof(int),
5490                         .mode           = 0644,
5491                         .proc_handler   = proc_dointvec,
5492                 },
5493                 {
5494                         .procname       = "dad_transmits",
5495                         .data           = &ipv6_devconf.dad_transmits,
5496                         .maxlen         = sizeof(int),
5497                         .mode           = 0644,
5498                         .proc_handler   = proc_dointvec,
5499                 },
5500                 {
5501                         .procname       = "router_solicitations",
5502                         .data           = &ipv6_devconf.rtr_solicits,
5503                         .maxlen         = sizeof(int),
5504                         .mode           = 0644,
5505                         .proc_handler   = proc_dointvec,
5506                 },
5507                 {
5508                         .procname       = "router_solicitation_interval",
5509                         .data           = &ipv6_devconf.rtr_solicit_interval,
5510                         .maxlen         = sizeof(int),
5511                         .mode           = 0644,
5512                         .proc_handler   = proc_dointvec_jiffies,
5513                 },
5514                 {
5515                         .procname       = "router_solicitation_delay",
5516                         .data           = &ipv6_devconf.rtr_solicit_delay,
5517                         .maxlen         = sizeof(int),
5518                         .mode           = 0644,
5519                         .proc_handler   = proc_dointvec_jiffies,
5520                 },
5521                 {
5522                         .procname       = "force_mld_version",
5523                         .data           = &ipv6_devconf.force_mld_version,
5524                         .maxlen         = sizeof(int),
5525                         .mode           = 0644,
5526                         .proc_handler   = proc_dointvec,
5527                 },
5528                 {
5529                         .procname       = "mldv1_unsolicited_report_interval",
5530                         .data           =
5531                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5532                         .maxlen         = sizeof(int),
5533                         .mode           = 0644,
5534                         .proc_handler   = proc_dointvec_ms_jiffies,
5535                 },
5536                 {
5537                         .procname       = "mldv2_unsolicited_report_interval",
5538                         .data           =
5539                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5540                         .maxlen         = sizeof(int),
5541                         .mode           = 0644,
5542                         .proc_handler   = proc_dointvec_ms_jiffies,
5543                 },
5544                 {
5545                         .procname       = "use_tempaddr",
5546                         .data           = &ipv6_devconf.use_tempaddr,
5547                         .maxlen         = sizeof(int),
5548                         .mode           = 0644,
5549                         .proc_handler   = proc_dointvec,
5550                 },
5551                 {
5552                         .procname       = "temp_valid_lft",
5553                         .data           = &ipv6_devconf.temp_valid_lft,
5554                         .maxlen         = sizeof(int),
5555                         .mode           = 0644,
5556                         .proc_handler   = proc_dointvec,
5557                 },
5558                 {
5559                         .procname       = "temp_prefered_lft",
5560                         .data           = &ipv6_devconf.temp_prefered_lft,
5561                         .maxlen         = sizeof(int),
5562                         .mode           = 0644,
5563                         .proc_handler   = proc_dointvec,
5564                 },
5565                 {
5566                         .procname       = "regen_max_retry",
5567                         .data           = &ipv6_devconf.regen_max_retry,
5568                         .maxlen         = sizeof(int),
5569                         .mode           = 0644,
5570                         .proc_handler   = proc_dointvec,
5571                 },
5572                 {
5573                         .procname       = "max_desync_factor",
5574                         .data           = &ipv6_devconf.max_desync_factor,
5575                         .maxlen         = sizeof(int),
5576                         .mode           = 0644,
5577                         .proc_handler   = proc_dointvec,
5578                 },
5579                 {
5580                         .procname       = "max_addresses",
5581                         .data           = &ipv6_devconf.max_addresses,
5582                         .maxlen         = sizeof(int),
5583                         .mode           = 0644,
5584                         .proc_handler   = proc_dointvec,
5585                 },
5586                 {
5587                         .procname       = "accept_ra_defrtr",
5588                         .data           = &ipv6_devconf.accept_ra_defrtr,
5589                         .maxlen         = sizeof(int),
5590                         .mode           = 0644,
5591                         .proc_handler   = proc_dointvec,
5592                 },
5593                 {
5594                         .procname       = "accept_ra_min_hop_limit",
5595                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5596                         .maxlen         = sizeof(int),
5597                         .mode           = 0644,
5598                         .proc_handler   = proc_dointvec,
5599                 },
5600                 {
5601                         .procname       = "accept_ra_pinfo",
5602                         .data           = &ipv6_devconf.accept_ra_pinfo,
5603                         .maxlen         = sizeof(int),
5604                         .mode           = 0644,
5605                         .proc_handler   = proc_dointvec,
5606                 },
5607 #ifdef CONFIG_IPV6_ROUTER_PREF
5608                 {
5609                         .procname       = "accept_ra_rtr_pref",
5610                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5611                         .maxlen         = sizeof(int),
5612                         .mode           = 0644,
5613                         .proc_handler   = proc_dointvec,
5614                 },
5615                 {
5616                         .procname       = "router_probe_interval",
5617                         .data           = &ipv6_devconf.rtr_probe_interval,
5618                         .maxlen         = sizeof(int),
5619                         .mode           = 0644,
5620                         .proc_handler   = proc_dointvec_jiffies,
5621                 },
5622 #ifdef CONFIG_IPV6_ROUTE_INFO
5623                 {
5624                         .procname       = "accept_ra_rt_info_max_plen",
5625                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5626                         .maxlen         = sizeof(int),
5627                         .mode           = 0644,
5628                         .proc_handler   = proc_dointvec,
5629                 },
5630 #endif
5631 #endif
5632                 {
5633                         .procname       = "proxy_ndp",
5634                         .data           = &ipv6_devconf.proxy_ndp,
5635                         .maxlen         = sizeof(int),
5636                         .mode           = 0644,
5637                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5638                 },
5639                 {
5640                         .procname       = "accept_source_route",
5641                         .data           = &ipv6_devconf.accept_source_route,
5642                         .maxlen         = sizeof(int),
5643                         .mode           = 0644,
5644                         .proc_handler   = proc_dointvec,
5645                 },
5646 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5647                 {
5648                         .procname       = "optimistic_dad",
5649                         .data           = &ipv6_devconf.optimistic_dad,
5650                         .maxlen         = sizeof(int),
5651                         .mode           = 0644,
5652                         .proc_handler   = proc_dointvec,
5653
5654                 },
5655                 {
5656                         .procname       = "use_optimistic",
5657                         .data           = &ipv6_devconf.use_optimistic,
5658                         .maxlen         = sizeof(int),
5659                         .mode           = 0644,
5660                         .proc_handler   = proc_dointvec,
5661
5662                 },
5663 #endif
5664 #ifdef CONFIG_IPV6_MROUTE
5665                 {
5666                         .procname       = "mc_forwarding",
5667                         .data           = &ipv6_devconf.mc_forwarding,
5668                         .maxlen         = sizeof(int),
5669                         .mode           = 0444,
5670                         .proc_handler   = proc_dointvec,
5671                 },
5672 #endif
5673                 {
5674                         .procname       = "disable_ipv6",
5675                         .data           = &ipv6_devconf.disable_ipv6,
5676                         .maxlen         = sizeof(int),
5677                         .mode           = 0644,
5678                         .proc_handler   = addrconf_sysctl_disable,
5679                 },
5680                 {
5681                         .procname       = "accept_dad",
5682                         .data           = &ipv6_devconf.accept_dad,
5683                         .maxlen         = sizeof(int),
5684                         .mode           = 0644,
5685                         .proc_handler   = proc_dointvec,
5686                 },
5687                 {
5688                         .procname       = "force_tllao",
5689                         .data           = &ipv6_devconf.force_tllao,
5690                         .maxlen         = sizeof(int),
5691                         .mode           = 0644,
5692                         .proc_handler   = proc_dointvec
5693                 },
5694                 {
5695                         .procname       = "ndisc_notify",
5696                         .data           = &ipv6_devconf.ndisc_notify,
5697                         .maxlen         = sizeof(int),
5698                         .mode           = 0644,
5699                         .proc_handler   = proc_dointvec
5700                 },
5701                 {
5702                         .procname       = "suppress_frag_ndisc",
5703                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5704                         .maxlen         = sizeof(int),
5705                         .mode           = 0644,
5706                         .proc_handler   = proc_dointvec
5707                 },
5708                 {
5709                         .procname       = "accept_ra_from_local",
5710                         .data           = &ipv6_devconf.accept_ra_from_local,
5711                         .maxlen         = sizeof(int),
5712                         .mode           = 0644,
5713                         .proc_handler   = proc_dointvec,
5714                 },
5715                 {
5716                         .procname       = "accept_ra_mtu",
5717                         .data           = &ipv6_devconf.accept_ra_mtu,
5718                         .maxlen         = sizeof(int),
5719                         .mode           = 0644,
5720                         .proc_handler   = proc_dointvec,
5721                 },
5722                 {
5723                         .procname       = "stable_secret",
5724                         .data           = &ipv6_devconf.stable_secret,
5725                         .maxlen         = IPV6_MAX_STRLEN,
5726                         .mode           = 0600,
5727                         .proc_handler   = addrconf_sysctl_stable_secret,
5728                 },
5729                 {
5730                         .procname       = "use_oif_addrs_only",
5731                         .data           = &ipv6_devconf.use_oif_addrs_only,
5732                         .maxlen         = sizeof(int),
5733                         .mode           = 0644,
5734                         .proc_handler   = proc_dointvec,
5735                 },
5736                 {
5737                         .procname       = "ignore_routes_with_linkdown",
5738                         .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5739                         .maxlen         = sizeof(int),
5740                         .mode           = 0644,
5741                         .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5742                 },
5743                 {
5744                         /* sentinel */
5745                 }
5746         },
5747 };
5748
5749 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5750                 struct inet6_dev *idev, struct ipv6_devconf *p)
5751 {
5752         int i;
5753         struct addrconf_sysctl_table *t;
5754         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5755
5756         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5757         if (!t)
5758                 goto out;
5759
5760         for (i = 0; t->addrconf_vars[i].data; i++) {
5761                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5762                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5763                 t->addrconf_vars[i].extra2 = net;
5764         }
5765
5766         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5767
5768         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5769         if (!t->sysctl_header)
5770                 goto free;
5771
5772         p->sysctl = t;
5773         return 0;
5774
5775 free:
5776         kfree(t);
5777 out:
5778         return -ENOBUFS;
5779 }
5780
5781 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5782 {
5783         struct addrconf_sysctl_table *t;
5784
5785         if (!p->sysctl)
5786                 return;
5787
5788         t = p->sysctl;
5789         p->sysctl = NULL;
5790         unregister_net_sysctl_table(t->sysctl_header);
5791         kfree(t);
5792 }
5793
5794 static int addrconf_sysctl_register(struct inet6_dev *idev)
5795 {
5796         int err;
5797
5798         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5799                 return -EINVAL;
5800
5801         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5802                                     &ndisc_ifinfo_sysctl_change);
5803         if (err)
5804                 return err;
5805         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5806                                          idev, &idev->cnf);
5807         if (err)
5808                 neigh_sysctl_unregister(idev->nd_parms);
5809
5810         return err;
5811 }
5812
5813 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5814 {
5815         __addrconf_sysctl_unregister(&idev->cnf);
5816         neigh_sysctl_unregister(idev->nd_parms);
5817 }
5818
5819
5820 #endif
5821
5822 static int __net_init addrconf_init_net(struct net *net)
5823 {
5824         int err = -ENOMEM;
5825         struct ipv6_devconf *all, *dflt;
5826
5827         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5828         if (!all)
5829                 goto err_alloc_all;
5830
5831         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5832         if (!dflt)
5833                 goto err_alloc_dflt;
5834
5835         /* these will be inherited by all namespaces */
5836         dflt->autoconf = ipv6_defaults.autoconf;
5837         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5838
5839         dflt->stable_secret.initialized = false;
5840         all->stable_secret.initialized = false;
5841
5842         net->ipv6.devconf_all = all;
5843         net->ipv6.devconf_dflt = dflt;
5844
5845 #ifdef CONFIG_SYSCTL
5846         err = __addrconf_sysctl_register(net, "all", NULL, all);
5847         if (err < 0)
5848                 goto err_reg_all;
5849
5850         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5851         if (err < 0)
5852                 goto err_reg_dflt;
5853 #endif
5854         return 0;
5855
5856 #ifdef CONFIG_SYSCTL
5857 err_reg_dflt:
5858         __addrconf_sysctl_unregister(all);
5859 err_reg_all:
5860         kfree(dflt);
5861 #endif
5862 err_alloc_dflt:
5863         kfree(all);
5864 err_alloc_all:
5865         return err;
5866 }
5867
5868 static void __net_exit addrconf_exit_net(struct net *net)
5869 {
5870 #ifdef CONFIG_SYSCTL
5871         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5872         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5873 #endif
5874         kfree(net->ipv6.devconf_dflt);
5875         kfree(net->ipv6.devconf_all);
5876 }
5877
5878 static struct pernet_operations addrconf_ops = {
5879         .init = addrconf_init_net,
5880         .exit = addrconf_exit_net,
5881 };
5882
5883 static struct rtnl_af_ops inet6_ops __read_mostly = {
5884         .family           = AF_INET6,
5885         .fill_link_af     = inet6_fill_link_af,
5886         .get_link_af_size = inet6_get_link_af_size,
5887         .validate_link_af = inet6_validate_link_af,
5888         .set_link_af      = inet6_set_link_af,
5889 };
5890
5891 /*
5892  *      Init / cleanup code
5893  */
5894
5895 int __init addrconf_init(void)
5896 {
5897         struct inet6_dev *idev;
5898         int i, err;
5899
5900         err = ipv6_addr_label_init();
5901         if (err < 0) {
5902                 pr_crit("%s: cannot initialize default policy table: %d\n",
5903                         __func__, err);
5904                 goto out;
5905         }
5906
5907         err = register_pernet_subsys(&addrconf_ops);
5908         if (err < 0)
5909                 goto out_addrlabel;
5910
5911         addrconf_wq = create_workqueue("ipv6_addrconf");
5912         if (!addrconf_wq) {
5913                 err = -ENOMEM;
5914                 goto out_nowq;
5915         }
5916
5917         /* The addrconf netdev notifier requires that loopback_dev
5918          * has it's ipv6 private information allocated and setup
5919          * before it can bring up and give link-local addresses
5920          * to other devices which are up.
5921          *
5922          * Unfortunately, loopback_dev is not necessarily the first
5923          * entry in the global dev_base list of net devices.  In fact,
5924          * it is likely to be the very last entry on that list.
5925          * So this causes the notifier registry below to try and
5926          * give link-local addresses to all devices besides loopback_dev
5927          * first, then loopback_dev, which cases all the non-loopback_dev
5928          * devices to fail to get a link-local address.
5929          *
5930          * So, as a temporary fix, allocate the ipv6 structure for
5931          * loopback_dev first by hand.
5932          * Longer term, all of the dependencies ipv6 has upon the loopback
5933          * device and it being up should be removed.
5934          */
5935         rtnl_lock();
5936         idev = ipv6_add_dev(init_net.loopback_dev);
5937         rtnl_unlock();
5938         if (IS_ERR(idev)) {
5939                 err = PTR_ERR(idev);
5940                 goto errlo;
5941         }
5942
5943         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5944                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5945
5946         register_netdevice_notifier(&ipv6_dev_notf);
5947
5948         addrconf_verify();
5949
5950         rtnl_af_register(&inet6_ops);
5951
5952         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5953                               NULL);
5954         if (err < 0)
5955                 goto errout;
5956
5957         /* Only the first call to __rtnl_register can fail */
5958         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5959         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5960         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5961                         inet6_dump_ifaddr, NULL);
5962         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5963                         inet6_dump_ifmcaddr, NULL);
5964         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5965                         inet6_dump_ifacaddr, NULL);
5966         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5967                         inet6_netconf_dump_devconf, NULL);
5968
5969         ipv6_addr_label_rtnl_register();
5970
5971         return 0;
5972 errout:
5973         rtnl_af_unregister(&inet6_ops);
5974         unregister_netdevice_notifier(&ipv6_dev_notf);
5975 errlo:
5976         destroy_workqueue(addrconf_wq);
5977 out_nowq:
5978         unregister_pernet_subsys(&addrconf_ops);
5979 out_addrlabel:
5980         ipv6_addr_label_cleanup();
5981 out:
5982         return err;
5983 }
5984
5985 void addrconf_cleanup(void)
5986 {
5987         struct net_device *dev;
5988         int i;
5989
5990         unregister_netdevice_notifier(&ipv6_dev_notf);
5991         unregister_pernet_subsys(&addrconf_ops);
5992         ipv6_addr_label_cleanup();
5993
5994         rtnl_lock();
5995
5996         __rtnl_af_unregister(&inet6_ops);
5997
5998         /* clean dev list */
5999         for_each_netdev(&init_net, dev) {
6000                 if (__in6_dev_get(dev) == NULL)
6001                         continue;
6002                 addrconf_ifdown(dev, 1);
6003         }
6004         addrconf_ifdown(init_net.loopback_dev, 2);
6005
6006         /*
6007          *      Check hash table.
6008          */
6009         spin_lock_bh(&addrconf_hash_lock);
6010         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6011                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6012         spin_unlock_bh(&addrconf_hash_lock);
6013         cancel_delayed_work(&addr_chk_work);
6014         rtnl_unlock();
6015
6016         destroy_workqueue(addrconf_wq);
6017 }