Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
[linux-2.6-microblaze.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
68
69 #include <net/net_namespace.h>
70 #include <net/sock.h>
71 #include <net/snmp.h>
72
73 #include <net/af_ieee802154.h>
74 #include <net/firewire.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/tcp.h>
81 #include <net/ip.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <linux/if_tunnel.h>
85 #include <linux/rtnetlink.h>
86 #include <linux/netconf.h>
87 #include <linux/random.h>
88 #include <linux/uaccess.h>
89 #include <asm/unaligned.h>
90
91 #include <linux/proc_fs.h>
92 #include <linux/seq_file.h>
93 #include <linux/export.h>
94
95 /* Set to 3 to get tracing... */
96 #define ACONF_DEBUG 2
97
98 #if ACONF_DEBUG >= 3
99 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
100 #else
101 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
102 #endif
103
104 #define INFINITY_LIFE_TIME      0xFFFFFFFF
105
106 #define IPV6_MAX_STRLEN \
107         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
108
109 static inline u32 cstamp_delta(unsigned long cstamp)
110 {
111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
112 }
113
114 #ifdef CONFIG_SYSCTL
115 static int addrconf_sysctl_register(struct inet6_dev *idev);
116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
117 #else
118 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
119 {
120         return 0;
121 }
122
123 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
124 {
125 }
126 #endif
127
128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
130 static void ipv6_regen_rndid(unsigned long data);
131
132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
133 static int ipv6_count_addresses(struct inet6_dev *idev);
134 static int ipv6_generate_stable_address(struct in6_addr *addr,
135                                         u8 dad_count,
136                                         const struct inet6_dev *idev);
137
138 /*
139  *      Configured unicast address hash table
140  */
141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
142 static DEFINE_SPINLOCK(addrconf_hash_lock);
143
144 static void addrconf_verify(void);
145 static void addrconf_verify_rtnl(void);
146 static void addrconf_verify_work(struct work_struct *);
147
148 static struct workqueue_struct *addrconf_wq;
149 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
150
151 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
152 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
153
154 static void addrconf_type_change(struct net_device *dev,
155                                  unsigned long event);
156 static int addrconf_ifdown(struct net_device *dev, int how);
157
158 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
159                                                   int plen,
160                                                   const struct net_device *dev,
161                                                   u32 flags, u32 noflags);
162
163 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
164 static void addrconf_dad_work(struct work_struct *w);
165 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
166 static void addrconf_dad_run(struct inet6_dev *idev);
167 static void addrconf_rs_timer(unsigned long data);
168 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
169 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170
171 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
172                                 struct prefix_info *pinfo);
173 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
174                                struct net_device *dev);
175
176 static struct ipv6_devconf ipv6_devconf __read_mostly = {
177         .forwarding             = 0,
178         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
179         .mtu6                   = IPV6_MIN_MTU,
180         .accept_ra              = 1,
181         .accept_redirects       = 1,
182         .autoconf               = 1,
183         .force_mld_version      = 0,
184         .mldv1_unsolicited_report_interval = 10 * HZ,
185         .mldv2_unsolicited_report_interval = HZ,
186         .dad_transmits          = 1,
187         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
188         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
189         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
190         .use_tempaddr           = 0,
191         .temp_valid_lft         = TEMP_VALID_LIFETIME,
192         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
193         .regen_max_retry        = REGEN_MAX_RETRY,
194         .max_desync_factor      = MAX_DESYNC_FACTOR,
195         .max_addresses          = IPV6_MAX_ADDRESSES,
196         .accept_ra_defrtr       = 1,
197         .accept_ra_from_local   = 0,
198         .accept_ra_min_hop_limit= 1,
199         .accept_ra_pinfo        = 1,
200 #ifdef CONFIG_IPV6_ROUTER_PREF
201         .accept_ra_rtr_pref     = 1,
202         .rtr_probe_interval     = 60 * HZ,
203 #ifdef CONFIG_IPV6_ROUTE_INFO
204         .accept_ra_rt_info_max_plen = 0,
205 #endif
206 #endif
207         .proxy_ndp              = 0,
208         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
209         .disable_ipv6           = 0,
210         .accept_dad             = 1,
211         .suppress_frag_ndisc    = 1,
212         .accept_ra_mtu          = 1,
213         .stable_secret          = {
214                 .initialized = false,
215         },
216         .use_oif_addrs_only     = 0,
217         .ignore_routes_with_linkdown = 0,
218 };
219
220 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
221         .forwarding             = 0,
222         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
223         .mtu6                   = IPV6_MIN_MTU,
224         .accept_ra              = 1,
225         .accept_redirects       = 1,
226         .autoconf               = 1,
227         .force_mld_version      = 0,
228         .mldv1_unsolicited_report_interval = 10 * HZ,
229         .mldv2_unsolicited_report_interval = HZ,
230         .dad_transmits          = 1,
231         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
232         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
233         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
234         .use_tempaddr           = 0,
235         .temp_valid_lft         = TEMP_VALID_LIFETIME,
236         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
237         .regen_max_retry        = REGEN_MAX_RETRY,
238         .max_desync_factor      = MAX_DESYNC_FACTOR,
239         .max_addresses          = IPV6_MAX_ADDRESSES,
240         .accept_ra_defrtr       = 1,
241         .accept_ra_from_local   = 0,
242         .accept_ra_min_hop_limit= 1,
243         .accept_ra_pinfo        = 1,
244 #ifdef CONFIG_IPV6_ROUTER_PREF
245         .accept_ra_rtr_pref     = 1,
246         .rtr_probe_interval     = 60 * HZ,
247 #ifdef CONFIG_IPV6_ROUTE_INFO
248         .accept_ra_rt_info_max_plen = 0,
249 #endif
250 #endif
251         .proxy_ndp              = 0,
252         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
253         .disable_ipv6           = 0,
254         .accept_dad             = 1,
255         .suppress_frag_ndisc    = 1,
256         .accept_ra_mtu          = 1,
257         .stable_secret          = {
258                 .initialized = false,
259         },
260         .use_oif_addrs_only     = 0,
261         .ignore_routes_with_linkdown = 0,
262 };
263
264 /* Check if a valid qdisc is available */
265 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
266 {
267         return !qdisc_tx_is_noop(dev);
268 }
269
270 static void addrconf_del_rs_timer(struct inet6_dev *idev)
271 {
272         if (del_timer(&idev->rs_timer))
273                 __in6_dev_put(idev);
274 }
275
276 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
277 {
278         if (cancel_delayed_work(&ifp->dad_work))
279                 __in6_ifa_put(ifp);
280 }
281
282 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
283                                   unsigned long when)
284 {
285         if (!timer_pending(&idev->rs_timer))
286                 in6_dev_hold(idev);
287         mod_timer(&idev->rs_timer, jiffies + when);
288 }
289
290 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
291                                    unsigned long delay)
292 {
293         if (!delayed_work_pending(&ifp->dad_work))
294                 in6_ifa_hold(ifp);
295         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
296 }
297
298 static int snmp6_alloc_dev(struct inet6_dev *idev)
299 {
300         int i;
301
302         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
303         if (!idev->stats.ipv6)
304                 goto err_ip;
305
306         for_each_possible_cpu(i) {
307                 struct ipstats_mib *addrconf_stats;
308                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
309                 u64_stats_init(&addrconf_stats->syncp);
310         }
311
312
313         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
314                                         GFP_KERNEL);
315         if (!idev->stats.icmpv6dev)
316                 goto err_icmp;
317         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
318                                            GFP_KERNEL);
319         if (!idev->stats.icmpv6msgdev)
320                 goto err_icmpmsg;
321
322         return 0;
323
324 err_icmpmsg:
325         kfree(idev->stats.icmpv6dev);
326 err_icmp:
327         free_percpu(idev->stats.ipv6);
328 err_ip:
329         return -ENOMEM;
330 }
331
332 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
333 {
334         struct inet6_dev *ndev;
335         int err = -ENOMEM;
336
337         ASSERT_RTNL();
338
339         if (dev->mtu < IPV6_MIN_MTU)
340                 return ERR_PTR(-EINVAL);
341
342         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
343         if (!ndev)
344                 return ERR_PTR(err);
345
346         rwlock_init(&ndev->lock);
347         ndev->dev = dev;
348         INIT_LIST_HEAD(&ndev->addr_list);
349         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
350                     (unsigned long)ndev);
351         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
352         ndev->cnf.mtu6 = dev->mtu;
353         ndev->cnf.sysctl = NULL;
354         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
355         if (!ndev->nd_parms) {
356                 kfree(ndev);
357                 return ERR_PTR(err);
358         }
359         if (ndev->cnf.forwarding)
360                 dev_disable_lro(dev);
361         /* We refer to the device */
362         dev_hold(dev);
363
364         if (snmp6_alloc_dev(ndev) < 0) {
365                 ADBG(KERN_WARNING
366                         "%s: cannot allocate memory for statistics; dev=%s.\n",
367                         __func__, dev->name);
368                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
369                 dev_put(dev);
370                 kfree(ndev);
371                 return ERR_PTR(err);
372         }
373
374         if (snmp6_register_dev(ndev) < 0) {
375                 ADBG(KERN_WARNING
376                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
377                         __func__, dev->name);
378                 goto err_release;
379         }
380
381         /* One reference from device.  We must do this before
382          * we invoke __ipv6_regen_rndid().
383          */
384         in6_dev_hold(ndev);
385
386         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
387                 ndev->cnf.accept_dad = -1;
388
389 #if IS_ENABLED(CONFIG_IPV6_SIT)
390         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
391                 pr_info("%s: Disabled Multicast RS\n", dev->name);
392                 ndev->cnf.rtr_solicits = 0;
393         }
394 #endif
395
396         INIT_LIST_HEAD(&ndev->tempaddr_list);
397         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
398         if ((dev->flags&IFF_LOOPBACK) ||
399             dev->type == ARPHRD_TUNNEL ||
400             dev->type == ARPHRD_TUNNEL6 ||
401             dev->type == ARPHRD_SIT ||
402             dev->type == ARPHRD_NONE) {
403                 ndev->cnf.use_tempaddr = -1;
404         } else {
405                 in6_dev_hold(ndev);
406                 ipv6_regen_rndid((unsigned long) ndev);
407         }
408
409         ndev->token = in6addr_any;
410
411         if (netif_running(dev) && addrconf_qdisc_ok(dev))
412                 ndev->if_flags |= IF_READY;
413
414         ipv6_mc_init_dev(ndev);
415         ndev->tstamp = jiffies;
416         err = addrconf_sysctl_register(ndev);
417         if (err) {
418                 ipv6_mc_destroy_dev(ndev);
419                 del_timer(&ndev->regen_timer);
420                 goto err_release;
421         }
422         /* protected by rtnl_lock */
423         rcu_assign_pointer(dev->ip6_ptr, ndev);
424
425         /* Join interface-local all-node multicast group */
426         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
427
428         /* Join all-node multicast group */
429         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
430
431         /* Join all-router multicast group if forwarding is set */
432         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
433                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
434
435         return ndev;
436
437 err_release:
438         neigh_parms_release(&nd_tbl, ndev->nd_parms);
439         ndev->dead = 1;
440         in6_dev_finish_destroy(ndev);
441         return ERR_PTR(err);
442 }
443
444 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
445 {
446         struct inet6_dev *idev;
447
448         ASSERT_RTNL();
449
450         idev = __in6_dev_get(dev);
451         if (!idev) {
452                 idev = ipv6_add_dev(dev);
453                 if (IS_ERR(idev))
454                         return NULL;
455         }
456
457         if (dev->flags&IFF_UP)
458                 ipv6_mc_up(idev);
459         return idev;
460 }
461
462 static int inet6_netconf_msgsize_devconf(int type)
463 {
464         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
465                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
466
467         /* type -1 is used for ALL */
468         if (type == -1 || type == NETCONFA_FORWARDING)
469                 size += nla_total_size(4);
470 #ifdef CONFIG_IPV6_MROUTE
471         if (type == -1 || type == NETCONFA_MC_FORWARDING)
472                 size += nla_total_size(4);
473 #endif
474         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
475                 size += nla_total_size(4);
476
477         if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
478                 size += nla_total_size(4);
479
480         return size;
481 }
482
483 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
484                                       struct ipv6_devconf *devconf, u32 portid,
485                                       u32 seq, int event, unsigned int flags,
486                                       int type)
487 {
488         struct nlmsghdr  *nlh;
489         struct netconfmsg *ncm;
490
491         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
492                         flags);
493         if (!nlh)
494                 return -EMSGSIZE;
495
496         ncm = nlmsg_data(nlh);
497         ncm->ncm_family = AF_INET6;
498
499         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
500                 goto nla_put_failure;
501
502         /* type -1 is used for ALL */
503         if ((type == -1 || type == NETCONFA_FORWARDING) &&
504             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
505                 goto nla_put_failure;
506 #ifdef CONFIG_IPV6_MROUTE
507         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
508             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
509                         devconf->mc_forwarding) < 0)
510                 goto nla_put_failure;
511 #endif
512         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
513             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
514                 goto nla_put_failure;
515
516         if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
517             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
518                         devconf->ignore_routes_with_linkdown) < 0)
519                 goto nla_put_failure;
520
521         nlmsg_end(skb, nlh);
522         return 0;
523
524 nla_put_failure:
525         nlmsg_cancel(skb, nlh);
526         return -EMSGSIZE;
527 }
528
529 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
530                                   struct ipv6_devconf *devconf)
531 {
532         struct sk_buff *skb;
533         int err = -ENOBUFS;
534
535         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
536         if (!skb)
537                 goto errout;
538
539         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
540                                          RTM_NEWNETCONF, 0, type);
541         if (err < 0) {
542                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
543                 WARN_ON(err == -EMSGSIZE);
544                 kfree_skb(skb);
545                 goto errout;
546         }
547         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
548         return;
549 errout:
550         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
551 }
552
553 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
554         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
555         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
556         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
557         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
558 };
559
560 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
561                                      struct nlmsghdr *nlh)
562 {
563         struct net *net = sock_net(in_skb->sk);
564         struct nlattr *tb[NETCONFA_MAX+1];
565         struct netconfmsg *ncm;
566         struct sk_buff *skb;
567         struct ipv6_devconf *devconf;
568         struct inet6_dev *in6_dev;
569         struct net_device *dev;
570         int ifindex;
571         int err;
572
573         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
574                           devconf_ipv6_policy);
575         if (err < 0)
576                 goto errout;
577
578         err = EINVAL;
579         if (!tb[NETCONFA_IFINDEX])
580                 goto errout;
581
582         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
583         switch (ifindex) {
584         case NETCONFA_IFINDEX_ALL:
585                 devconf = net->ipv6.devconf_all;
586                 break;
587         case NETCONFA_IFINDEX_DEFAULT:
588                 devconf = net->ipv6.devconf_dflt;
589                 break;
590         default:
591                 dev = __dev_get_by_index(net, ifindex);
592                 if (!dev)
593                         goto errout;
594                 in6_dev = __in6_dev_get(dev);
595                 if (!in6_dev)
596                         goto errout;
597                 devconf = &in6_dev->cnf;
598                 break;
599         }
600
601         err = -ENOBUFS;
602         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
603         if (!skb)
604                 goto errout;
605
606         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
607                                          NETLINK_CB(in_skb).portid,
608                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
609                                          -1);
610         if (err < 0) {
611                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
612                 WARN_ON(err == -EMSGSIZE);
613                 kfree_skb(skb);
614                 goto errout;
615         }
616         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
617 errout:
618         return err;
619 }
620
621 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
622                                       struct netlink_callback *cb)
623 {
624         struct net *net = sock_net(skb->sk);
625         int h, s_h;
626         int idx, s_idx;
627         struct net_device *dev;
628         struct inet6_dev *idev;
629         struct hlist_head *head;
630
631         s_h = cb->args[0];
632         s_idx = idx = cb->args[1];
633
634         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
635                 idx = 0;
636                 head = &net->dev_index_head[h];
637                 rcu_read_lock();
638                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
639                           net->dev_base_seq;
640                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
641                         if (idx < s_idx)
642                                 goto cont;
643                         idev = __in6_dev_get(dev);
644                         if (!idev)
645                                 goto cont;
646
647                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
648                                                        &idev->cnf,
649                                                        NETLINK_CB(cb->skb).portid,
650                                                        cb->nlh->nlmsg_seq,
651                                                        RTM_NEWNETCONF,
652                                                        NLM_F_MULTI,
653                                                        -1) < 0) {
654                                 rcu_read_unlock();
655                                 goto done;
656                         }
657                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
658 cont:
659                         idx++;
660                 }
661                 rcu_read_unlock();
662         }
663         if (h == NETDEV_HASHENTRIES) {
664                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
665                                                net->ipv6.devconf_all,
666                                                NETLINK_CB(cb->skb).portid,
667                                                cb->nlh->nlmsg_seq,
668                                                RTM_NEWNETCONF, NLM_F_MULTI,
669                                                -1) < 0)
670                         goto done;
671                 else
672                         h++;
673         }
674         if (h == NETDEV_HASHENTRIES + 1) {
675                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
676                                                net->ipv6.devconf_dflt,
677                                                NETLINK_CB(cb->skb).portid,
678                                                cb->nlh->nlmsg_seq,
679                                                RTM_NEWNETCONF, NLM_F_MULTI,
680                                                -1) < 0)
681                         goto done;
682                 else
683                         h++;
684         }
685 done:
686         cb->args[0] = h;
687         cb->args[1] = idx;
688
689         return skb->len;
690 }
691
692 #ifdef CONFIG_SYSCTL
693 static void dev_forward_change(struct inet6_dev *idev)
694 {
695         struct net_device *dev;
696         struct inet6_ifaddr *ifa;
697
698         if (!idev)
699                 return;
700         dev = idev->dev;
701         if (idev->cnf.forwarding)
702                 dev_disable_lro(dev);
703         if (dev->flags & IFF_MULTICAST) {
704                 if (idev->cnf.forwarding) {
705                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
706                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
707                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
708                 } else {
709                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
710                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
711                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
712                 }
713         }
714
715         list_for_each_entry(ifa, &idev->addr_list, if_list) {
716                 if (ifa->flags&IFA_F_TENTATIVE)
717                         continue;
718                 if (idev->cnf.forwarding)
719                         addrconf_join_anycast(ifa);
720                 else
721                         addrconf_leave_anycast(ifa);
722         }
723         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
724                                      dev->ifindex, &idev->cnf);
725 }
726
727
728 static void addrconf_forward_change(struct net *net, __s32 newf)
729 {
730         struct net_device *dev;
731         struct inet6_dev *idev;
732
733         for_each_netdev(net, dev) {
734                 idev = __in6_dev_get(dev);
735                 if (idev) {
736                         int changed = (!idev->cnf.forwarding) ^ (!newf);
737                         idev->cnf.forwarding = newf;
738                         if (changed)
739                                 dev_forward_change(idev);
740                 }
741         }
742 }
743
744 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
745 {
746         struct net *net;
747         int old;
748
749         if (!rtnl_trylock())
750                 return restart_syscall();
751
752         net = (struct net *)table->extra2;
753         old = *p;
754         *p = newf;
755
756         if (p == &net->ipv6.devconf_dflt->forwarding) {
757                 if ((!newf) ^ (!old))
758                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
759                                                      NETCONFA_IFINDEX_DEFAULT,
760                                                      net->ipv6.devconf_dflt);
761                 rtnl_unlock();
762                 return 0;
763         }
764
765         if (p == &net->ipv6.devconf_all->forwarding) {
766                 net->ipv6.devconf_dflt->forwarding = newf;
767                 addrconf_forward_change(net, newf);
768                 if ((!newf) ^ (!old))
769                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
770                                                      NETCONFA_IFINDEX_ALL,
771                                                      net->ipv6.devconf_all);
772         } else if ((!newf) ^ (!old))
773                 dev_forward_change((struct inet6_dev *)table->extra1);
774         rtnl_unlock();
775
776         if (newf)
777                 rt6_purge_dflt_routers(net);
778         return 1;
779 }
780
781 static void addrconf_linkdown_change(struct net *net, __s32 newf)
782 {
783         struct net_device *dev;
784         struct inet6_dev *idev;
785
786         for_each_netdev(net, dev) {
787                 idev = __in6_dev_get(dev);
788                 if (idev) {
789                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
790
791                         idev->cnf.ignore_routes_with_linkdown = newf;
792                         if (changed)
793                                 inet6_netconf_notify_devconf(dev_net(dev),
794                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
795                                                              dev->ifindex,
796                                                              &idev->cnf);
797                 }
798         }
799 }
800
801 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
802 {
803         struct net *net;
804         int old;
805
806         if (!rtnl_trylock())
807                 return restart_syscall();
808
809         net = (struct net *)table->extra2;
810         old = *p;
811         *p = newf;
812
813         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
814                 if ((!newf) ^ (!old))
815                         inet6_netconf_notify_devconf(net,
816                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
817                                                      NETCONFA_IFINDEX_DEFAULT,
818                                                      net->ipv6.devconf_dflt);
819                 rtnl_unlock();
820                 return 0;
821         }
822
823         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
824                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
825                 addrconf_linkdown_change(net, newf);
826                 if ((!newf) ^ (!old))
827                         inet6_netconf_notify_devconf(net,
828                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
829                                                      NETCONFA_IFINDEX_ALL,
830                                                      net->ipv6.devconf_all);
831         }
832         rtnl_unlock();
833
834         return 1;
835 }
836
837 #endif
838
839 /* Nobody refers to this ifaddr, destroy it */
840 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
841 {
842         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
843
844 #ifdef NET_REFCNT_DEBUG
845         pr_debug("%s\n", __func__);
846 #endif
847
848         in6_dev_put(ifp->idev);
849
850         if (cancel_delayed_work(&ifp->dad_work))
851                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
852                           ifp);
853
854         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
855                 pr_warn("Freeing alive inet6 address %p\n", ifp);
856                 return;
857         }
858         ip6_rt_put(ifp->rt);
859
860         kfree_rcu(ifp, rcu);
861 }
862
863 static void
864 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
865 {
866         struct list_head *p;
867         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
868
869         /*
870          * Each device address list is sorted in order of scope -
871          * global before linklocal.
872          */
873         list_for_each(p, &idev->addr_list) {
874                 struct inet6_ifaddr *ifa
875                         = list_entry(p, struct inet6_ifaddr, if_list);
876                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
877                         break;
878         }
879
880         list_add_tail(&ifp->if_list, p);
881 }
882
883 static u32 inet6_addr_hash(const struct in6_addr *addr)
884 {
885         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
886 }
887
888 /* On success it returns ifp with increased reference count */
889
890 static struct inet6_ifaddr *
891 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
892               const struct in6_addr *peer_addr, int pfxlen,
893               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
894 {
895         struct inet6_ifaddr *ifa = NULL;
896         struct rt6_info *rt;
897         unsigned int hash;
898         int err = 0;
899         int addr_type = ipv6_addr_type(addr);
900
901         if (addr_type == IPV6_ADDR_ANY ||
902             addr_type & IPV6_ADDR_MULTICAST ||
903             (!(idev->dev->flags & IFF_LOOPBACK) &&
904              addr_type & IPV6_ADDR_LOOPBACK))
905                 return ERR_PTR(-EADDRNOTAVAIL);
906
907         rcu_read_lock_bh();
908         if (idev->dead) {
909                 err = -ENODEV;                  /*XXX*/
910                 goto out2;
911         }
912
913         if (idev->cnf.disable_ipv6) {
914                 err = -EACCES;
915                 goto out2;
916         }
917
918         spin_lock(&addrconf_hash_lock);
919
920         /* Ignore adding duplicate addresses on an interface */
921         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
922                 ADBG("ipv6_add_addr: already assigned\n");
923                 err = -EEXIST;
924                 goto out;
925         }
926
927         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
928
929         if (!ifa) {
930                 ADBG("ipv6_add_addr: malloc failed\n");
931                 err = -ENOBUFS;
932                 goto out;
933         }
934
935         rt = addrconf_dst_alloc(idev, addr, false);
936         if (IS_ERR(rt)) {
937                 err = PTR_ERR(rt);
938                 goto out;
939         }
940
941         neigh_parms_data_state_setall(idev->nd_parms);
942
943         ifa->addr = *addr;
944         if (peer_addr)
945                 ifa->peer_addr = *peer_addr;
946
947         spin_lock_init(&ifa->lock);
948         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
949         INIT_HLIST_NODE(&ifa->addr_lst);
950         ifa->scope = scope;
951         ifa->prefix_len = pfxlen;
952         ifa->flags = flags | IFA_F_TENTATIVE;
953         ifa->valid_lft = valid_lft;
954         ifa->prefered_lft = prefered_lft;
955         ifa->cstamp = ifa->tstamp = jiffies;
956         ifa->tokenized = false;
957
958         ifa->rt = rt;
959
960         ifa->idev = idev;
961         in6_dev_hold(idev);
962         /* For caller */
963         in6_ifa_hold(ifa);
964
965         /* Add to big hash table */
966         hash = inet6_addr_hash(addr);
967
968         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
969         spin_unlock(&addrconf_hash_lock);
970
971         write_lock(&idev->lock);
972         /* Add to inet6_dev unicast addr list. */
973         ipv6_link_dev_addr(idev, ifa);
974
975         if (ifa->flags&IFA_F_TEMPORARY) {
976                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
977                 in6_ifa_hold(ifa);
978         }
979
980         in6_ifa_hold(ifa);
981         write_unlock(&idev->lock);
982 out2:
983         rcu_read_unlock_bh();
984
985         if (likely(err == 0))
986                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
987         else {
988                 kfree(ifa);
989                 ifa = ERR_PTR(err);
990         }
991
992         return ifa;
993 out:
994         spin_unlock(&addrconf_hash_lock);
995         goto out2;
996 }
997
998 enum cleanup_prefix_rt_t {
999         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1000         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1001         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1002 };
1003
1004 /*
1005  * Check, whether the prefix for ifp would still need a prefix route
1006  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1007  * constants.
1008  *
1009  * 1) we don't purge prefix if address was not permanent.
1010  *    prefix is managed by its own lifetime.
1011  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1012  * 3) if there are no addresses, delete prefix.
1013  * 4) if there are still other permanent address(es),
1014  *    corresponding prefix is still permanent.
1015  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1016  *    don't purge the prefix, assume user space is managing it.
1017  * 6) otherwise, update prefix lifetime to the
1018  *    longest valid lifetime among the corresponding
1019  *    addresses on the device.
1020  *    Note: subsequent RA will update lifetime.
1021  **/
1022 static enum cleanup_prefix_rt_t
1023 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1024 {
1025         struct inet6_ifaddr *ifa;
1026         struct inet6_dev *idev = ifp->idev;
1027         unsigned long lifetime;
1028         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1029
1030         *expires = jiffies;
1031
1032         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1033                 if (ifa == ifp)
1034                         continue;
1035                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1036                                        ifp->prefix_len))
1037                         continue;
1038                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1039                         return CLEANUP_PREFIX_RT_NOP;
1040
1041                 action = CLEANUP_PREFIX_RT_EXPIRE;
1042
1043                 spin_lock(&ifa->lock);
1044
1045                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1046                 /*
1047                  * Note: Because this address is
1048                  * not permanent, lifetime <
1049                  * LONG_MAX / HZ here.
1050                  */
1051                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1052                         *expires = ifa->tstamp + lifetime * HZ;
1053                 spin_unlock(&ifa->lock);
1054         }
1055
1056         return action;
1057 }
1058
1059 static void
1060 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1061 {
1062         struct rt6_info *rt;
1063
1064         rt = addrconf_get_prefix_route(&ifp->addr,
1065                                        ifp->prefix_len,
1066                                        ifp->idev->dev,
1067                                        0, RTF_GATEWAY | RTF_DEFAULT);
1068         if (rt) {
1069                 if (del_rt)
1070                         ip6_del_rt(rt);
1071                 else {
1072                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1073                                 rt6_set_expires(rt, expires);
1074                         ip6_rt_put(rt);
1075                 }
1076         }
1077 }
1078
1079
1080 /* This function wants to get referenced ifp and releases it before return */
1081
1082 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1083 {
1084         int state;
1085         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1086         unsigned long expires;
1087
1088         ASSERT_RTNL();
1089
1090         spin_lock_bh(&ifp->lock);
1091         state = ifp->state;
1092         ifp->state = INET6_IFADDR_STATE_DEAD;
1093         spin_unlock_bh(&ifp->lock);
1094
1095         if (state == INET6_IFADDR_STATE_DEAD)
1096                 goto out;
1097
1098         spin_lock_bh(&addrconf_hash_lock);
1099         hlist_del_init_rcu(&ifp->addr_lst);
1100         spin_unlock_bh(&addrconf_hash_lock);
1101
1102         write_lock_bh(&ifp->idev->lock);
1103
1104         if (ifp->flags&IFA_F_TEMPORARY) {
1105                 list_del(&ifp->tmp_list);
1106                 if (ifp->ifpub) {
1107                         in6_ifa_put(ifp->ifpub);
1108                         ifp->ifpub = NULL;
1109                 }
1110                 __in6_ifa_put(ifp);
1111         }
1112
1113         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1114                 action = check_cleanup_prefix_route(ifp, &expires);
1115
1116         list_del_init(&ifp->if_list);
1117         __in6_ifa_put(ifp);
1118
1119         write_unlock_bh(&ifp->idev->lock);
1120
1121         addrconf_del_dad_work(ifp);
1122
1123         ipv6_ifa_notify(RTM_DELADDR, ifp);
1124
1125         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1126
1127         if (action != CLEANUP_PREFIX_RT_NOP) {
1128                 cleanup_prefix_route(ifp, expires,
1129                         action == CLEANUP_PREFIX_RT_DEL);
1130         }
1131
1132         /* clean up prefsrc entries */
1133         rt6_remove_prefsrc(ifp);
1134 out:
1135         in6_ifa_put(ifp);
1136 }
1137
1138 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1139 {
1140         struct inet6_dev *idev = ifp->idev;
1141         struct in6_addr addr, *tmpaddr;
1142         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1143         unsigned long regen_advance;
1144         int tmp_plen;
1145         int ret = 0;
1146         u32 addr_flags;
1147         unsigned long now = jiffies;
1148
1149         write_lock_bh(&idev->lock);
1150         if (ift) {
1151                 spin_lock_bh(&ift->lock);
1152                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1153                 spin_unlock_bh(&ift->lock);
1154                 tmpaddr = &addr;
1155         } else {
1156                 tmpaddr = NULL;
1157         }
1158 retry:
1159         in6_dev_hold(idev);
1160         if (idev->cnf.use_tempaddr <= 0) {
1161                 write_unlock_bh(&idev->lock);
1162                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1163                 in6_dev_put(idev);
1164                 ret = -1;
1165                 goto out;
1166         }
1167         spin_lock_bh(&ifp->lock);
1168         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1169                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1170                 spin_unlock_bh(&ifp->lock);
1171                 write_unlock_bh(&idev->lock);
1172                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1173                         __func__);
1174                 in6_dev_put(idev);
1175                 ret = -1;
1176                 goto out;
1177         }
1178         in6_ifa_hold(ifp);
1179         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1180         __ipv6_try_regen_rndid(idev, tmpaddr);
1181         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1182         age = (now - ifp->tstamp) / HZ;
1183         tmp_valid_lft = min_t(__u32,
1184                               ifp->valid_lft,
1185                               idev->cnf.temp_valid_lft + age);
1186         tmp_prefered_lft = min_t(__u32,
1187                                  ifp->prefered_lft,
1188                                  idev->cnf.temp_prefered_lft + age -
1189                                  idev->cnf.max_desync_factor);
1190         tmp_plen = ifp->prefix_len;
1191         tmp_tstamp = ifp->tstamp;
1192         spin_unlock_bh(&ifp->lock);
1193
1194         regen_advance = idev->cnf.regen_max_retry *
1195                         idev->cnf.dad_transmits *
1196                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1197         write_unlock_bh(&idev->lock);
1198
1199         /* A temporary address is created only if this calculated Preferred
1200          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1201          * an implementation must not create a temporary address with a zero
1202          * Preferred Lifetime.
1203          * Use age calculation as in addrconf_verify to avoid unnecessary
1204          * temporary addresses being generated.
1205          */
1206         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1207         if (tmp_prefered_lft <= regen_advance + age) {
1208                 in6_ifa_put(ifp);
1209                 in6_dev_put(idev);
1210                 ret = -1;
1211                 goto out;
1212         }
1213
1214         addr_flags = IFA_F_TEMPORARY;
1215         /* set in addrconf_prefix_rcv() */
1216         if (ifp->flags & IFA_F_OPTIMISTIC)
1217                 addr_flags |= IFA_F_OPTIMISTIC;
1218
1219         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1220                             ipv6_addr_scope(&addr), addr_flags,
1221                             tmp_valid_lft, tmp_prefered_lft);
1222         if (IS_ERR(ift)) {
1223                 in6_ifa_put(ifp);
1224                 in6_dev_put(idev);
1225                 pr_info("%s: retry temporary address regeneration\n", __func__);
1226                 tmpaddr = &addr;
1227                 write_lock_bh(&idev->lock);
1228                 goto retry;
1229         }
1230
1231         spin_lock_bh(&ift->lock);
1232         ift->ifpub = ifp;
1233         ift->cstamp = now;
1234         ift->tstamp = tmp_tstamp;
1235         spin_unlock_bh(&ift->lock);
1236
1237         addrconf_dad_start(ift);
1238         in6_ifa_put(ift);
1239         in6_dev_put(idev);
1240 out:
1241         return ret;
1242 }
1243
1244 /*
1245  *      Choose an appropriate source address (RFC3484)
1246  */
1247 enum {
1248         IPV6_SADDR_RULE_INIT = 0,
1249         IPV6_SADDR_RULE_LOCAL,
1250         IPV6_SADDR_RULE_SCOPE,
1251         IPV6_SADDR_RULE_PREFERRED,
1252 #ifdef CONFIG_IPV6_MIP6
1253         IPV6_SADDR_RULE_HOA,
1254 #endif
1255         IPV6_SADDR_RULE_OIF,
1256         IPV6_SADDR_RULE_LABEL,
1257         IPV6_SADDR_RULE_PRIVACY,
1258         IPV6_SADDR_RULE_ORCHID,
1259         IPV6_SADDR_RULE_PREFIX,
1260 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1261         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1262 #endif
1263         IPV6_SADDR_RULE_MAX
1264 };
1265
1266 struct ipv6_saddr_score {
1267         int                     rule;
1268         int                     addr_type;
1269         struct inet6_ifaddr     *ifa;
1270         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1271         int                     scopedist;
1272         int                     matchlen;
1273 };
1274
1275 struct ipv6_saddr_dst {
1276         const struct in6_addr *addr;
1277         int ifindex;
1278         int scope;
1279         int label;
1280         unsigned int prefs;
1281 };
1282
1283 static inline int ipv6_saddr_preferred(int type)
1284 {
1285         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1286                 return 1;
1287         return 0;
1288 }
1289
1290 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1291 {
1292 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1293         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1294 #else
1295         return false;
1296 #endif
1297 }
1298
1299 static int ipv6_get_saddr_eval(struct net *net,
1300                                struct ipv6_saddr_score *score,
1301                                struct ipv6_saddr_dst *dst,
1302                                int i)
1303 {
1304         int ret;
1305
1306         if (i <= score->rule) {
1307                 switch (i) {
1308                 case IPV6_SADDR_RULE_SCOPE:
1309                         ret = score->scopedist;
1310                         break;
1311                 case IPV6_SADDR_RULE_PREFIX:
1312                         ret = score->matchlen;
1313                         break;
1314                 default:
1315                         ret = !!test_bit(i, score->scorebits);
1316                 }
1317                 goto out;
1318         }
1319
1320         switch (i) {
1321         case IPV6_SADDR_RULE_INIT:
1322                 /* Rule 0: remember if hiscore is not ready yet */
1323                 ret = !!score->ifa;
1324                 break;
1325         case IPV6_SADDR_RULE_LOCAL:
1326                 /* Rule 1: Prefer same address */
1327                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1328                 break;
1329         case IPV6_SADDR_RULE_SCOPE:
1330                 /* Rule 2: Prefer appropriate scope
1331                  *
1332                  *      ret
1333                  *       ^
1334                  *    -1 |  d 15
1335                  *    ---+--+-+---> scope
1336                  *       |
1337                  *       |             d is scope of the destination.
1338                  *  B-d  |  \
1339                  *       |   \      <- smaller scope is better if
1340                  *  B-15 |    \        if scope is enough for destination.
1341                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1342                  * d-C-1 | /
1343                  *       |/         <- greater is better
1344                  *   -C  /             if scope is not enough for destination.
1345                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1346                  *
1347                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1348                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1349                  * Assume B = 0 and we get C > 29.
1350                  */
1351                 ret = __ipv6_addr_src_scope(score->addr_type);
1352                 if (ret >= dst->scope)
1353                         ret = -ret;
1354                 else
1355                         ret -= 128;     /* 30 is enough */
1356                 score->scopedist = ret;
1357                 break;
1358         case IPV6_SADDR_RULE_PREFERRED:
1359             {
1360                 /* Rule 3: Avoid deprecated and optimistic addresses */
1361                 u8 avoid = IFA_F_DEPRECATED;
1362
1363                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1364                         avoid |= IFA_F_OPTIMISTIC;
1365                 ret = ipv6_saddr_preferred(score->addr_type) ||
1366                       !(score->ifa->flags & avoid);
1367                 break;
1368             }
1369 #ifdef CONFIG_IPV6_MIP6
1370         case IPV6_SADDR_RULE_HOA:
1371             {
1372                 /* Rule 4: Prefer home address */
1373                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1374                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1375                 break;
1376             }
1377 #endif
1378         case IPV6_SADDR_RULE_OIF:
1379                 /* Rule 5: Prefer outgoing interface */
1380                 ret = (!dst->ifindex ||
1381                        dst->ifindex == score->ifa->idev->dev->ifindex);
1382                 break;
1383         case IPV6_SADDR_RULE_LABEL:
1384                 /* Rule 6: Prefer matching label */
1385                 ret = ipv6_addr_label(net,
1386                                       &score->ifa->addr, score->addr_type,
1387                                       score->ifa->idev->dev->ifindex) == dst->label;
1388                 break;
1389         case IPV6_SADDR_RULE_PRIVACY:
1390             {
1391                 /* Rule 7: Prefer public address
1392                  * Note: prefer temporary address if use_tempaddr >= 2
1393                  */
1394                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1395                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1396                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1397                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1398                 break;
1399             }
1400         case IPV6_SADDR_RULE_ORCHID:
1401                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1402                  *          non-ORCHID vs non-ORCHID
1403                  */
1404                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1405                         ipv6_addr_orchid(dst->addr));
1406                 break;
1407         case IPV6_SADDR_RULE_PREFIX:
1408                 /* Rule 8: Use longest matching prefix */
1409                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1410                 if (ret > score->ifa->prefix_len)
1411                         ret = score->ifa->prefix_len;
1412                 score->matchlen = ret;
1413                 break;
1414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1415         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1416                 /* Optimistic addresses still have lower precedence than other
1417                  * preferred addresses.
1418                  */
1419                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1420                 break;
1421 #endif
1422         default:
1423                 ret = 0;
1424         }
1425
1426         if (ret)
1427                 __set_bit(i, score->scorebits);
1428         score->rule = i;
1429 out:
1430         return ret;
1431 }
1432
1433 static int __ipv6_dev_get_saddr(struct net *net,
1434                                 struct ipv6_saddr_dst *dst,
1435                                 struct inet6_dev *idev,
1436                                 struct ipv6_saddr_score *scores,
1437                                 int hiscore_idx)
1438 {
1439         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1440
1441         read_lock_bh(&idev->lock);
1442         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1443                 int i;
1444
1445                 /*
1446                  * - Tentative Address (RFC2462 section 5.4)
1447                  *  - A tentative address is not considered
1448                  *    "assigned to an interface" in the traditional
1449                  *    sense, unless it is also flagged as optimistic.
1450                  * - Candidate Source Address (section 4)
1451                  *  - In any case, anycast addresses, multicast
1452                  *    addresses, and the unspecified address MUST
1453                  *    NOT be included in a candidate set.
1454                  */
1455                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1456                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1457                         continue;
1458
1459                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1460
1461                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1462                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1463                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1464                                             idev->dev->name);
1465                         continue;
1466                 }
1467
1468                 score->rule = -1;
1469                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1470
1471                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1472                         int minihiscore, miniscore;
1473
1474                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1475                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1476
1477                         if (minihiscore > miniscore) {
1478                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1479                                     score->scopedist > 0) {
1480                                         /*
1481                                          * special case:
1482                                          * each remaining entry
1483                                          * has too small (not enough)
1484                                          * scope, because ifa entries
1485                                          * are sorted by their scope
1486                                          * values.
1487                                          */
1488                                         goto out;
1489                                 }
1490                                 break;
1491                         } else if (minihiscore < miniscore) {
1492                                 if (hiscore->ifa)
1493                                         in6_ifa_put(hiscore->ifa);
1494
1495                                 in6_ifa_hold(score->ifa);
1496
1497                                 swap(hiscore, score);
1498                                 hiscore_idx = 1 - hiscore_idx;
1499
1500                                 /* restore our iterator */
1501                                 score->ifa = hiscore->ifa;
1502
1503                                 break;
1504                         }
1505                 }
1506         }
1507 out:
1508         read_unlock_bh(&idev->lock);
1509         return hiscore_idx;
1510 }
1511
1512 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1513                        const struct in6_addr *daddr, unsigned int prefs,
1514                        struct in6_addr *saddr)
1515 {
1516         struct ipv6_saddr_score scores[2], *hiscore;
1517         struct ipv6_saddr_dst dst;
1518         struct inet6_dev *idev;
1519         struct net_device *dev;
1520         int dst_type;
1521         bool use_oif_addr = false;
1522         int hiscore_idx = 0;
1523
1524         dst_type = __ipv6_addr_type(daddr);
1525         dst.addr = daddr;
1526         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1527         dst.scope = __ipv6_addr_src_scope(dst_type);
1528         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1529         dst.prefs = prefs;
1530
1531         scores[hiscore_idx].rule = -1;
1532         scores[hiscore_idx].ifa = NULL;
1533
1534         rcu_read_lock();
1535
1536         /* Candidate Source Address (section 4)
1537          *  - multicast and link-local destination address,
1538          *    the set of candidate source address MUST only
1539          *    include addresses assigned to interfaces
1540          *    belonging to the same link as the outgoing
1541          *    interface.
1542          * (- For site-local destination addresses, the
1543          *    set of candidate source addresses MUST only
1544          *    include addresses assigned to interfaces
1545          *    belonging to the same site as the outgoing
1546          *    interface.)
1547          *  - "It is RECOMMENDED that the candidate source addresses
1548          *    be the set of unicast addresses assigned to the
1549          *    interface that will be used to send to the destination
1550          *    (the 'outgoing' interface)." (RFC 6724)
1551          */
1552         if (dst_dev) {
1553                 idev = __in6_dev_get(dst_dev);
1554                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1555                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1556                     (idev && idev->cnf.use_oif_addrs_only)) {
1557                         use_oif_addr = true;
1558                 }
1559         }
1560
1561         if (use_oif_addr) {
1562                 if (idev)
1563                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1564         } else {
1565                 for_each_netdev_rcu(net, dev) {
1566                         idev = __in6_dev_get(dev);
1567                         if (!idev)
1568                                 continue;
1569                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1570                 }
1571         }
1572         rcu_read_unlock();
1573
1574         hiscore = &scores[hiscore_idx];
1575         if (!hiscore->ifa)
1576                 return -EADDRNOTAVAIL;
1577
1578         *saddr = hiscore->ifa->addr;
1579         in6_ifa_put(hiscore->ifa);
1580         return 0;
1581 }
1582 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1583
1584 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1585                       u32 banned_flags)
1586 {
1587         struct inet6_ifaddr *ifp;
1588         int err = -EADDRNOTAVAIL;
1589
1590         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1591                 if (ifp->scope > IFA_LINK)
1592                         break;
1593                 if (ifp->scope == IFA_LINK &&
1594                     !(ifp->flags & banned_flags)) {
1595                         *addr = ifp->addr;
1596                         err = 0;
1597                         break;
1598                 }
1599         }
1600         return err;
1601 }
1602
1603 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1604                     u32 banned_flags)
1605 {
1606         struct inet6_dev *idev;
1607         int err = -EADDRNOTAVAIL;
1608
1609         rcu_read_lock();
1610         idev = __in6_dev_get(dev);
1611         if (idev) {
1612                 read_lock_bh(&idev->lock);
1613                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1614                 read_unlock_bh(&idev->lock);
1615         }
1616         rcu_read_unlock();
1617         return err;
1618 }
1619
1620 static int ipv6_count_addresses(struct inet6_dev *idev)
1621 {
1622         int cnt = 0;
1623         struct inet6_ifaddr *ifp;
1624
1625         read_lock_bh(&idev->lock);
1626         list_for_each_entry(ifp, &idev->addr_list, if_list)
1627                 cnt++;
1628         read_unlock_bh(&idev->lock);
1629         return cnt;
1630 }
1631
1632 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1633                   const struct net_device *dev, int strict)
1634 {
1635         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1636 }
1637 EXPORT_SYMBOL(ipv6_chk_addr);
1638
1639 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1640                             const struct net_device *dev, int strict,
1641                             u32 banned_flags)
1642 {
1643         struct inet6_ifaddr *ifp;
1644         unsigned int hash = inet6_addr_hash(addr);
1645         u32 ifp_flags;
1646
1647         rcu_read_lock_bh();
1648         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1649                 if (!net_eq(dev_net(ifp->idev->dev), net))
1650                         continue;
1651                 /* Decouple optimistic from tentative for evaluation here.
1652                  * Ban optimistic addresses explicitly, when required.
1653                  */
1654                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1655                             ? (ifp->flags&~IFA_F_TENTATIVE)
1656                             : ifp->flags;
1657                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1658                     !(ifp_flags&banned_flags) &&
1659                     (!dev || ifp->idev->dev == dev ||
1660                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1661                         rcu_read_unlock_bh();
1662                         return 1;
1663                 }
1664         }
1665
1666         rcu_read_unlock_bh();
1667         return 0;
1668 }
1669 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1670
1671 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1672                                struct net_device *dev)
1673 {
1674         unsigned int hash = inet6_addr_hash(addr);
1675         struct inet6_ifaddr *ifp;
1676
1677         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1678                 if (!net_eq(dev_net(ifp->idev->dev), net))
1679                         continue;
1680                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1681                         if (!dev || ifp->idev->dev == dev)
1682                                 return true;
1683                 }
1684         }
1685         return false;
1686 }
1687
1688 /* Compares an address/prefix_len with addresses on device @dev.
1689  * If one is found it returns true.
1690  */
1691 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1692         const unsigned int prefix_len, struct net_device *dev)
1693 {
1694         struct inet6_dev *idev;
1695         struct inet6_ifaddr *ifa;
1696         bool ret = false;
1697
1698         rcu_read_lock();
1699         idev = __in6_dev_get(dev);
1700         if (idev) {
1701                 read_lock_bh(&idev->lock);
1702                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1703                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1704                         if (ret)
1705                                 break;
1706                 }
1707                 read_unlock_bh(&idev->lock);
1708         }
1709         rcu_read_unlock();
1710
1711         return ret;
1712 }
1713 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1714
1715 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1716 {
1717         struct inet6_dev *idev;
1718         struct inet6_ifaddr *ifa;
1719         int     onlink;
1720
1721         onlink = 0;
1722         rcu_read_lock();
1723         idev = __in6_dev_get(dev);
1724         if (idev) {
1725                 read_lock_bh(&idev->lock);
1726                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1727                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1728                                                    ifa->prefix_len);
1729                         if (onlink)
1730                                 break;
1731                 }
1732                 read_unlock_bh(&idev->lock);
1733         }
1734         rcu_read_unlock();
1735         return onlink;
1736 }
1737 EXPORT_SYMBOL(ipv6_chk_prefix);
1738
1739 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1740                                      struct net_device *dev, int strict)
1741 {
1742         struct inet6_ifaddr *ifp, *result = NULL;
1743         unsigned int hash = inet6_addr_hash(addr);
1744
1745         rcu_read_lock_bh();
1746         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1747                 if (!net_eq(dev_net(ifp->idev->dev), net))
1748                         continue;
1749                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1750                         if (!dev || ifp->idev->dev == dev ||
1751                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1752                                 result = ifp;
1753                                 in6_ifa_hold(ifp);
1754                                 break;
1755                         }
1756                 }
1757         }
1758         rcu_read_unlock_bh();
1759
1760         return result;
1761 }
1762
1763 /* Gets referenced address, destroys ifaddr */
1764
1765 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1766 {
1767         if (ifp->flags&IFA_F_PERMANENT) {
1768                 spin_lock_bh(&ifp->lock);
1769                 addrconf_del_dad_work(ifp);
1770                 ifp->flags |= IFA_F_TENTATIVE;
1771                 if (dad_failed)
1772                         ifp->flags |= IFA_F_DADFAILED;
1773                 spin_unlock_bh(&ifp->lock);
1774                 if (dad_failed)
1775                         ipv6_ifa_notify(0, ifp);
1776                 in6_ifa_put(ifp);
1777         } else if (ifp->flags&IFA_F_TEMPORARY) {
1778                 struct inet6_ifaddr *ifpub;
1779                 spin_lock_bh(&ifp->lock);
1780                 ifpub = ifp->ifpub;
1781                 if (ifpub) {
1782                         in6_ifa_hold(ifpub);
1783                         spin_unlock_bh(&ifp->lock);
1784                         ipv6_create_tempaddr(ifpub, ifp);
1785                         in6_ifa_put(ifpub);
1786                 } else {
1787                         spin_unlock_bh(&ifp->lock);
1788                 }
1789                 ipv6_del_addr(ifp);
1790         } else {
1791                 ipv6_del_addr(ifp);
1792         }
1793 }
1794
1795 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1796 {
1797         int err = -ENOENT;
1798
1799         spin_lock_bh(&ifp->lock);
1800         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1801                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1802                 err = 0;
1803         }
1804         spin_unlock_bh(&ifp->lock);
1805
1806         return err;
1807 }
1808
1809 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1810 {
1811         struct in6_addr addr;
1812         struct inet6_dev *idev = ifp->idev;
1813         struct net *net = dev_net(ifp->idev->dev);
1814
1815         if (addrconf_dad_end(ifp)) {
1816                 in6_ifa_put(ifp);
1817                 return;
1818         }
1819
1820         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1821                              ifp->idev->dev->name, &ifp->addr);
1822
1823         spin_lock_bh(&ifp->lock);
1824
1825         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1826                 int scope = ifp->scope;
1827                 u32 flags = ifp->flags;
1828                 struct in6_addr new_addr;
1829                 struct inet6_ifaddr *ifp2;
1830                 u32 valid_lft, preferred_lft;
1831                 int pfxlen = ifp->prefix_len;
1832                 int retries = ifp->stable_privacy_retry + 1;
1833
1834                 if (retries > net->ipv6.sysctl.idgen_retries) {
1835                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1836                                              ifp->idev->dev->name);
1837                         goto errdad;
1838                 }
1839
1840                 new_addr = ifp->addr;
1841                 if (ipv6_generate_stable_address(&new_addr, retries,
1842                                                  idev))
1843                         goto errdad;
1844
1845                 valid_lft = ifp->valid_lft;
1846                 preferred_lft = ifp->prefered_lft;
1847
1848                 spin_unlock_bh(&ifp->lock);
1849
1850                 if (idev->cnf.max_addresses &&
1851                     ipv6_count_addresses(idev) >=
1852                     idev->cnf.max_addresses)
1853                         goto lock_errdad;
1854
1855                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1856                                      ifp->idev->dev->name);
1857
1858                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1859                                      scope, flags, valid_lft,
1860                                      preferred_lft);
1861                 if (IS_ERR(ifp2))
1862                         goto lock_errdad;
1863
1864                 spin_lock_bh(&ifp2->lock);
1865                 ifp2->stable_privacy_retry = retries;
1866                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1867                 spin_unlock_bh(&ifp2->lock);
1868
1869                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1870                 in6_ifa_put(ifp2);
1871 lock_errdad:
1872                 spin_lock_bh(&ifp->lock);
1873         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1874                 addr.s6_addr32[0] = htonl(0xfe800000);
1875                 addr.s6_addr32[1] = 0;
1876
1877                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1878                     ipv6_addr_equal(&ifp->addr, &addr)) {
1879                         /* DAD failed for link-local based on MAC address */
1880                         idev->cnf.disable_ipv6 = 1;
1881
1882                         pr_info("%s: IPv6 being disabled!\n",
1883                                 ifp->idev->dev->name);
1884                 }
1885         }
1886
1887 errdad:
1888         /* transition from _POSTDAD to _ERRDAD */
1889         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1890         spin_unlock_bh(&ifp->lock);
1891
1892         addrconf_mod_dad_work(ifp, 0);
1893 }
1894
1895 /* Join to solicited addr multicast group.
1896  * caller must hold RTNL */
1897 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1898 {
1899         struct in6_addr maddr;
1900
1901         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1902                 return;
1903
1904         addrconf_addr_solict_mult(addr, &maddr);
1905         ipv6_dev_mc_inc(dev, &maddr);
1906 }
1907
1908 /* caller must hold RTNL */
1909 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1910 {
1911         struct in6_addr maddr;
1912
1913         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1914                 return;
1915
1916         addrconf_addr_solict_mult(addr, &maddr);
1917         __ipv6_dev_mc_dec(idev, &maddr);
1918 }
1919
1920 /* caller must hold RTNL */
1921 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1922 {
1923         struct in6_addr addr;
1924
1925         if (ifp->prefix_len >= 127) /* RFC 6164 */
1926                 return;
1927         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1928         if (ipv6_addr_any(&addr))
1929                 return;
1930         __ipv6_dev_ac_inc(ifp->idev, &addr);
1931 }
1932
1933 /* caller must hold RTNL */
1934 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1935 {
1936         struct in6_addr addr;
1937
1938         if (ifp->prefix_len >= 127) /* RFC 6164 */
1939                 return;
1940         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1941         if (ipv6_addr_any(&addr))
1942                 return;
1943         __ipv6_dev_ac_dec(ifp->idev, &addr);
1944 }
1945
1946 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1947 {
1948         if (dev->addr_len != ETH_ALEN)
1949                 return -1;
1950         memcpy(eui, dev->dev_addr, 3);
1951         memcpy(eui + 5, dev->dev_addr + 3, 3);
1952
1953         /*
1954          * The zSeries OSA network cards can be shared among various
1955          * OS instances, but the OSA cards have only one MAC address.
1956          * This leads to duplicate address conflicts in conjunction
1957          * with IPv6 if more than one instance uses the same card.
1958          *
1959          * The driver for these cards can deliver a unique 16-bit
1960          * identifier for each instance sharing the same card.  It is
1961          * placed instead of 0xFFFE in the interface identifier.  The
1962          * "u" bit of the interface identifier is not inverted in this
1963          * case.  Hence the resulting interface identifier has local
1964          * scope according to RFC2373.
1965          */
1966         if (dev->dev_id) {
1967                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1968                 eui[4] = dev->dev_id & 0xFF;
1969         } else {
1970                 eui[3] = 0xFF;
1971                 eui[4] = 0xFE;
1972                 eui[0] ^= 2;
1973         }
1974         return 0;
1975 }
1976
1977 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1978 {
1979         if (dev->addr_len != IEEE802154_ADDR_LEN)
1980                 return -1;
1981         memcpy(eui, dev->dev_addr, 8);
1982         eui[0] ^= 2;
1983         return 0;
1984 }
1985
1986 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1987 {
1988         union fwnet_hwaddr *ha;
1989
1990         if (dev->addr_len != FWNET_ALEN)
1991                 return -1;
1992
1993         ha = (union fwnet_hwaddr *)dev->dev_addr;
1994
1995         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1996         eui[0] ^= 2;
1997         return 0;
1998 }
1999
2000 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2001 {
2002         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2003         if (dev->addr_len != ARCNET_ALEN)
2004                 return -1;
2005         memset(eui, 0, 7);
2006         eui[7] = *(u8 *)dev->dev_addr;
2007         return 0;
2008 }
2009
2010 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2011 {
2012         if (dev->addr_len != INFINIBAND_ALEN)
2013                 return -1;
2014         memcpy(eui, dev->dev_addr + 12, 8);
2015         eui[0] |= 2;
2016         return 0;
2017 }
2018
2019 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2020 {
2021         if (addr == 0)
2022                 return -1;
2023         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2024                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2025                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2026                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2027                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2028                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2029         eui[1] = 0;
2030         eui[2] = 0x5E;
2031         eui[3] = 0xFE;
2032         memcpy(eui + 4, &addr, 4);
2033         return 0;
2034 }
2035
2036 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2037 {
2038         if (dev->priv_flags & IFF_ISATAP)
2039                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2040         return -1;
2041 }
2042
2043 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2044 {
2045         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2046 }
2047
2048 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2049 {
2050         memcpy(eui, dev->perm_addr, 3);
2051         memcpy(eui + 5, dev->perm_addr + 3, 3);
2052         eui[3] = 0xFF;
2053         eui[4] = 0xFE;
2054         eui[0] ^= 2;
2055         return 0;
2056 }
2057
2058 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2059 {
2060         switch (dev->type) {
2061         case ARPHRD_ETHER:
2062         case ARPHRD_FDDI:
2063                 return addrconf_ifid_eui48(eui, dev);
2064         case ARPHRD_ARCNET:
2065                 return addrconf_ifid_arcnet(eui, dev);
2066         case ARPHRD_INFINIBAND:
2067                 return addrconf_ifid_infiniband(eui, dev);
2068         case ARPHRD_SIT:
2069                 return addrconf_ifid_sit(eui, dev);
2070         case ARPHRD_IPGRE:
2071                 return addrconf_ifid_gre(eui, dev);
2072         case ARPHRD_6LOWPAN:
2073         case ARPHRD_IEEE802154:
2074                 return addrconf_ifid_eui64(eui, dev);
2075         case ARPHRD_IEEE1394:
2076                 return addrconf_ifid_ieee1394(eui, dev);
2077         case ARPHRD_TUNNEL6:
2078                 return addrconf_ifid_ip6tnl(eui, dev);
2079         }
2080         return -1;
2081 }
2082
2083 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2084 {
2085         int err = -1;
2086         struct inet6_ifaddr *ifp;
2087
2088         read_lock_bh(&idev->lock);
2089         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2090                 if (ifp->scope > IFA_LINK)
2091                         break;
2092                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2093                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2094                         err = 0;
2095                         break;
2096                 }
2097         }
2098         read_unlock_bh(&idev->lock);
2099         return err;
2100 }
2101
2102 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2103 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2104 {
2105 regen:
2106         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2107         idev->rndid[0] &= ~0x02;
2108
2109         /*
2110          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2111          * check if generated address is not inappropriate
2112          *
2113          *  - Reserved subnet anycast (RFC 2526)
2114          *      11111101 11....11 1xxxxxxx
2115          *  - ISATAP (RFC4214) 6.1
2116          *      00-00-5E-FE-xx-xx-xx-xx
2117          *  - value 0
2118          *  - XXX: already assigned to an address on the device
2119          */
2120         if (idev->rndid[0] == 0xfd &&
2121             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2122             (idev->rndid[7]&0x80))
2123                 goto regen;
2124         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2125                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2126                         goto regen;
2127                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2128                         goto regen;
2129         }
2130 }
2131
2132 static void ipv6_regen_rndid(unsigned long data)
2133 {
2134         struct inet6_dev *idev = (struct inet6_dev *) data;
2135         unsigned long expires;
2136
2137         rcu_read_lock_bh();
2138         write_lock_bh(&idev->lock);
2139
2140         if (idev->dead)
2141                 goto out;
2142
2143         __ipv6_regen_rndid(idev);
2144
2145         expires = jiffies +
2146                 idev->cnf.temp_prefered_lft * HZ -
2147                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2148                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2149                 idev->cnf.max_desync_factor * HZ;
2150         if (time_before(expires, jiffies)) {
2151                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2152                         __func__, idev->dev->name);
2153                 goto out;
2154         }
2155
2156         if (!mod_timer(&idev->regen_timer, expires))
2157                 in6_dev_hold(idev);
2158
2159 out:
2160         write_unlock_bh(&idev->lock);
2161         rcu_read_unlock_bh();
2162         in6_dev_put(idev);
2163 }
2164
2165 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2166 {
2167         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2168                 __ipv6_regen_rndid(idev);
2169 }
2170
2171 /*
2172  *      Add prefix route.
2173  */
2174
2175 static void
2176 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2177                       unsigned long expires, u32 flags)
2178 {
2179         struct fib6_config cfg = {
2180                 .fc_table = RT6_TABLE_PREFIX,
2181                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2182                 .fc_ifindex = dev->ifindex,
2183                 .fc_expires = expires,
2184                 .fc_dst_len = plen,
2185                 .fc_flags = RTF_UP | flags,
2186                 .fc_nlinfo.nl_net = dev_net(dev),
2187                 .fc_protocol = RTPROT_KERNEL,
2188         };
2189
2190         cfg.fc_dst = *pfx;
2191
2192         /* Prevent useless cloning on PtP SIT.
2193            This thing is done here expecting that the whole
2194            class of non-broadcast devices need not cloning.
2195          */
2196 #if IS_ENABLED(CONFIG_IPV6_SIT)
2197         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2198                 cfg.fc_flags |= RTF_NONEXTHOP;
2199 #endif
2200
2201         ip6_route_add(&cfg);
2202 }
2203
2204
2205 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2206                                                   int plen,
2207                                                   const struct net_device *dev,
2208                                                   u32 flags, u32 noflags)
2209 {
2210         struct fib6_node *fn;
2211         struct rt6_info *rt = NULL;
2212         struct fib6_table *table;
2213
2214         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2215         if (!table)
2216                 return NULL;
2217
2218         read_lock_bh(&table->tb6_lock);
2219         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2220         if (!fn)
2221                 goto out;
2222
2223         noflags |= RTF_CACHE;
2224         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2225                 if (rt->dst.dev->ifindex != dev->ifindex)
2226                         continue;
2227                 if ((rt->rt6i_flags & flags) != flags)
2228                         continue;
2229                 if ((rt->rt6i_flags & noflags) != 0)
2230                         continue;
2231                 dst_hold(&rt->dst);
2232                 break;
2233         }
2234 out:
2235         read_unlock_bh(&table->tb6_lock);
2236         return rt;
2237 }
2238
2239
2240 /* Create "default" multicast route to the interface */
2241
2242 static void addrconf_add_mroute(struct net_device *dev)
2243 {
2244         struct fib6_config cfg = {
2245                 .fc_table = RT6_TABLE_LOCAL,
2246                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2247                 .fc_ifindex = dev->ifindex,
2248                 .fc_dst_len = 8,
2249                 .fc_flags = RTF_UP,
2250                 .fc_nlinfo.nl_net = dev_net(dev),
2251         };
2252
2253         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2254
2255         ip6_route_add(&cfg);
2256 }
2257
2258 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2259 {
2260         struct inet6_dev *idev;
2261
2262         ASSERT_RTNL();
2263
2264         idev = ipv6_find_idev(dev);
2265         if (!idev)
2266                 return ERR_PTR(-ENOBUFS);
2267
2268         if (idev->cnf.disable_ipv6)
2269                 return ERR_PTR(-EACCES);
2270
2271         /* Add default multicast route */
2272         if (!(dev->flags & IFF_LOOPBACK))
2273                 addrconf_add_mroute(dev);
2274
2275         return idev;
2276 }
2277
2278 static void manage_tempaddrs(struct inet6_dev *idev,
2279                              struct inet6_ifaddr *ifp,
2280                              __u32 valid_lft, __u32 prefered_lft,
2281                              bool create, unsigned long now)
2282 {
2283         u32 flags;
2284         struct inet6_ifaddr *ift;
2285
2286         read_lock_bh(&idev->lock);
2287         /* update all temporary addresses in the list */
2288         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2289                 int age, max_valid, max_prefered;
2290
2291                 if (ifp != ift->ifpub)
2292                         continue;
2293
2294                 /* RFC 4941 section 3.3:
2295                  * If a received option will extend the lifetime of a public
2296                  * address, the lifetimes of temporary addresses should
2297                  * be extended, subject to the overall constraint that no
2298                  * temporary addresses should ever remain "valid" or "preferred"
2299                  * for a time longer than (TEMP_VALID_LIFETIME) or
2300                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2301                  */
2302                 age = (now - ift->cstamp) / HZ;
2303                 max_valid = idev->cnf.temp_valid_lft - age;
2304                 if (max_valid < 0)
2305                         max_valid = 0;
2306
2307                 max_prefered = idev->cnf.temp_prefered_lft -
2308                                idev->cnf.max_desync_factor - age;
2309                 if (max_prefered < 0)
2310                         max_prefered = 0;
2311
2312                 if (valid_lft > max_valid)
2313                         valid_lft = max_valid;
2314
2315                 if (prefered_lft > max_prefered)
2316                         prefered_lft = max_prefered;
2317
2318                 spin_lock(&ift->lock);
2319                 flags = ift->flags;
2320                 ift->valid_lft = valid_lft;
2321                 ift->prefered_lft = prefered_lft;
2322                 ift->tstamp = now;
2323                 if (prefered_lft > 0)
2324                         ift->flags &= ~IFA_F_DEPRECATED;
2325
2326                 spin_unlock(&ift->lock);
2327                 if (!(flags&IFA_F_TENTATIVE))
2328                         ipv6_ifa_notify(0, ift);
2329         }
2330
2331         if ((create || list_empty(&idev->tempaddr_list)) &&
2332             idev->cnf.use_tempaddr > 0) {
2333                 /* When a new public address is created as described
2334                  * in [ADDRCONF], also create a new temporary address.
2335                  * Also create a temporary address if it's enabled but
2336                  * no temporary address currently exists.
2337                  */
2338                 read_unlock_bh(&idev->lock);
2339                 ipv6_create_tempaddr(ifp, NULL);
2340         } else {
2341                 read_unlock_bh(&idev->lock);
2342         }
2343 }
2344
2345 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2346 {
2347         struct prefix_info *pinfo;
2348         __u32 valid_lft;
2349         __u32 prefered_lft;
2350         int addr_type;
2351         u32 addr_flags = 0;
2352         struct inet6_dev *in6_dev;
2353         struct net *net = dev_net(dev);
2354
2355         pinfo = (struct prefix_info *) opt;
2356
2357         if (len < sizeof(struct prefix_info)) {
2358                 ADBG("addrconf: prefix option too short\n");
2359                 return;
2360         }
2361
2362         /*
2363          *      Validation checks ([ADDRCONF], page 19)
2364          */
2365
2366         addr_type = ipv6_addr_type(&pinfo->prefix);
2367
2368         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2369                 return;
2370
2371         valid_lft = ntohl(pinfo->valid);
2372         prefered_lft = ntohl(pinfo->prefered);
2373
2374         if (prefered_lft > valid_lft) {
2375                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2376                 return;
2377         }
2378
2379         in6_dev = in6_dev_get(dev);
2380
2381         if (!in6_dev) {
2382                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2383                                     dev->name);
2384                 return;
2385         }
2386
2387         /*
2388          *      Two things going on here:
2389          *      1) Add routes for on-link prefixes
2390          *      2) Configure prefixes with the auto flag set
2391          */
2392
2393         if (pinfo->onlink) {
2394                 struct rt6_info *rt;
2395                 unsigned long rt_expires;
2396
2397                 /* Avoid arithmetic overflow. Really, we could
2398                  * save rt_expires in seconds, likely valid_lft,
2399                  * but it would require division in fib gc, that it
2400                  * not good.
2401                  */
2402                 if (HZ > USER_HZ)
2403                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2404                 else
2405                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2406
2407                 if (addrconf_finite_timeout(rt_expires))
2408                         rt_expires *= HZ;
2409
2410                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2411                                                pinfo->prefix_len,
2412                                                dev,
2413                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2414                                                RTF_GATEWAY | RTF_DEFAULT);
2415
2416                 if (rt) {
2417                         /* Autoconf prefix route */
2418                         if (valid_lft == 0) {
2419                                 ip6_del_rt(rt);
2420                                 rt = NULL;
2421                         } else if (addrconf_finite_timeout(rt_expires)) {
2422                                 /* not infinity */
2423                                 rt6_set_expires(rt, jiffies + rt_expires);
2424                         } else {
2425                                 rt6_clean_expires(rt);
2426                         }
2427                 } else if (valid_lft) {
2428                         clock_t expires = 0;
2429                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2430                         if (addrconf_finite_timeout(rt_expires)) {
2431                                 /* not infinity */
2432                                 flags |= RTF_EXPIRES;
2433                                 expires = jiffies_to_clock_t(rt_expires);
2434                         }
2435                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2436                                               dev, expires, flags);
2437                 }
2438                 ip6_rt_put(rt);
2439         }
2440
2441         /* Try to figure out our local address for this prefix */
2442
2443         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2444                 struct inet6_ifaddr *ifp;
2445                 struct in6_addr addr;
2446                 int create = 0, update_lft = 0;
2447                 bool tokenized = false;
2448
2449                 if (pinfo->prefix_len == 64) {
2450                         memcpy(&addr, &pinfo->prefix, 8);
2451
2452                         if (!ipv6_addr_any(&in6_dev->token)) {
2453                                 read_lock_bh(&in6_dev->lock);
2454                                 memcpy(addr.s6_addr + 8,
2455                                        in6_dev->token.s6_addr + 8, 8);
2456                                 read_unlock_bh(&in6_dev->lock);
2457                                 tokenized = true;
2458                         } else if (in6_dev->addr_gen_mode ==
2459                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2460                                    !ipv6_generate_stable_address(&addr, 0,
2461                                                                  in6_dev)) {
2462                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2463                                 goto ok;
2464                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2465                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2466                                 in6_dev_put(in6_dev);
2467                                 return;
2468                         }
2469                         goto ok;
2470                 }
2471                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2472                                     pinfo->prefix_len);
2473                 in6_dev_put(in6_dev);
2474                 return;
2475
2476 ok:
2477
2478                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2479
2480                 if (!ifp && valid_lft) {
2481                         int max_addresses = in6_dev->cnf.max_addresses;
2482
2483 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2484                         if (in6_dev->cnf.optimistic_dad &&
2485                             !net->ipv6.devconf_all->forwarding && sllao)
2486                                 addr_flags = IFA_F_OPTIMISTIC;
2487 #endif
2488
2489                         /* Do not allow to create too much of autoconfigured
2490                          * addresses; this would be too easy way to crash kernel.
2491                          */
2492                         if (!max_addresses ||
2493                             ipv6_count_addresses(in6_dev) < max_addresses)
2494                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2495                                                     pinfo->prefix_len,
2496                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2497                                                     addr_flags, valid_lft,
2498                                                     prefered_lft);
2499
2500                         if (IS_ERR_OR_NULL(ifp)) {
2501                                 in6_dev_put(in6_dev);
2502                                 return;
2503                         }
2504
2505                         update_lft = 0;
2506                         create = 1;
2507                         spin_lock_bh(&ifp->lock);
2508                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2509                         ifp->cstamp = jiffies;
2510                         ifp->tokenized = tokenized;
2511                         spin_unlock_bh(&ifp->lock);
2512                         addrconf_dad_start(ifp);
2513                 }
2514
2515                 if (ifp) {
2516                         u32 flags;
2517                         unsigned long now;
2518                         u32 stored_lft;
2519
2520                         /* update lifetime (RFC2462 5.5.3 e) */
2521                         spin_lock_bh(&ifp->lock);
2522                         now = jiffies;
2523                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2524                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2525                         else
2526                                 stored_lft = 0;
2527                         if (!update_lft && !create && stored_lft) {
2528                                 const u32 minimum_lft = min_t(u32,
2529                                         stored_lft, MIN_VALID_LIFETIME);
2530                                 valid_lft = max(valid_lft, minimum_lft);
2531
2532                                 /* RFC4862 Section 5.5.3e:
2533                                  * "Note that the preferred lifetime of the
2534                                  *  corresponding address is always reset to
2535                                  *  the Preferred Lifetime in the received
2536                                  *  Prefix Information option, regardless of
2537                                  *  whether the valid lifetime is also reset or
2538                                  *  ignored."
2539                                  *
2540                                  * So we should always update prefered_lft here.
2541                                  */
2542                                 update_lft = 1;
2543                         }
2544
2545                         if (update_lft) {
2546                                 ifp->valid_lft = valid_lft;
2547                                 ifp->prefered_lft = prefered_lft;
2548                                 ifp->tstamp = now;
2549                                 flags = ifp->flags;
2550                                 ifp->flags &= ~IFA_F_DEPRECATED;
2551                                 spin_unlock_bh(&ifp->lock);
2552
2553                                 if (!(flags&IFA_F_TENTATIVE))
2554                                         ipv6_ifa_notify(0, ifp);
2555                         } else
2556                                 spin_unlock_bh(&ifp->lock);
2557
2558                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2559                                          create, now);
2560
2561                         in6_ifa_put(ifp);
2562                         addrconf_verify();
2563                 }
2564         }
2565         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2566         in6_dev_put(in6_dev);
2567 }
2568
2569 /*
2570  *      Set destination address.
2571  *      Special case for SIT interfaces where we create a new "virtual"
2572  *      device.
2573  */
2574 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2575 {
2576         struct in6_ifreq ireq;
2577         struct net_device *dev;
2578         int err = -EINVAL;
2579
2580         rtnl_lock();
2581
2582         err = -EFAULT;
2583         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2584                 goto err_exit;
2585
2586         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2587
2588         err = -ENODEV;
2589         if (!dev)
2590                 goto err_exit;
2591
2592 #if IS_ENABLED(CONFIG_IPV6_SIT)
2593         if (dev->type == ARPHRD_SIT) {
2594                 const struct net_device_ops *ops = dev->netdev_ops;
2595                 struct ifreq ifr;
2596                 struct ip_tunnel_parm p;
2597
2598                 err = -EADDRNOTAVAIL;
2599                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2600                         goto err_exit;
2601
2602                 memset(&p, 0, sizeof(p));
2603                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2604                 p.iph.saddr = 0;
2605                 p.iph.version = 4;
2606                 p.iph.ihl = 5;
2607                 p.iph.protocol = IPPROTO_IPV6;
2608                 p.iph.ttl = 64;
2609                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2610
2611                 if (ops->ndo_do_ioctl) {
2612                         mm_segment_t oldfs = get_fs();
2613
2614                         set_fs(KERNEL_DS);
2615                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2616                         set_fs(oldfs);
2617                 } else
2618                         err = -EOPNOTSUPP;
2619
2620                 if (err == 0) {
2621                         err = -ENOBUFS;
2622                         dev = __dev_get_by_name(net, p.name);
2623                         if (!dev)
2624                                 goto err_exit;
2625                         err = dev_open(dev);
2626                 }
2627         }
2628 #endif
2629
2630 err_exit:
2631         rtnl_unlock();
2632         return err;
2633 }
2634
2635 static int ipv6_mc_config(struct sock *sk, bool join,
2636                           const struct in6_addr *addr, int ifindex)
2637 {
2638         int ret;
2639
2640         ASSERT_RTNL();
2641
2642         lock_sock(sk);
2643         if (join)
2644                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2645         else
2646                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2647         release_sock(sk);
2648
2649         return ret;
2650 }
2651
2652 /*
2653  *      Manual configuration of address on an interface
2654  */
2655 static int inet6_addr_add(struct net *net, int ifindex,
2656                           const struct in6_addr *pfx,
2657                           const struct in6_addr *peer_pfx,
2658                           unsigned int plen, __u32 ifa_flags,
2659                           __u32 prefered_lft, __u32 valid_lft)
2660 {
2661         struct inet6_ifaddr *ifp;
2662         struct inet6_dev *idev;
2663         struct net_device *dev;
2664         unsigned long timeout;
2665         clock_t expires;
2666         int scope;
2667         u32 flags;
2668
2669         ASSERT_RTNL();
2670
2671         if (plen > 128)
2672                 return -EINVAL;
2673
2674         /* check the lifetime */
2675         if (!valid_lft || prefered_lft > valid_lft)
2676                 return -EINVAL;
2677
2678         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2679                 return -EINVAL;
2680
2681         dev = __dev_get_by_index(net, ifindex);
2682         if (!dev)
2683                 return -ENODEV;
2684
2685         idev = addrconf_add_dev(dev);
2686         if (IS_ERR(idev))
2687                 return PTR_ERR(idev);
2688
2689         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2690                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2691                                          true, pfx, ifindex);
2692
2693                 if (ret < 0)
2694                         return ret;
2695         }
2696
2697         scope = ipv6_addr_scope(pfx);
2698
2699         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2700         if (addrconf_finite_timeout(timeout)) {
2701                 expires = jiffies_to_clock_t(timeout * HZ);
2702                 valid_lft = timeout;
2703                 flags = RTF_EXPIRES;
2704         } else {
2705                 expires = 0;
2706                 flags = 0;
2707                 ifa_flags |= IFA_F_PERMANENT;
2708         }
2709
2710         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2711         if (addrconf_finite_timeout(timeout)) {
2712                 if (timeout == 0)
2713                         ifa_flags |= IFA_F_DEPRECATED;
2714                 prefered_lft = timeout;
2715         }
2716
2717         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2718                             valid_lft, prefered_lft);
2719
2720         if (!IS_ERR(ifp)) {
2721                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2722                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2723                                               expires, flags);
2724                 }
2725
2726                 /*
2727                  * Note that section 3.1 of RFC 4429 indicates
2728                  * that the Optimistic flag should not be set for
2729                  * manually configured addresses
2730                  */
2731                 addrconf_dad_start(ifp);
2732                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2733                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2734                                          true, jiffies);
2735                 in6_ifa_put(ifp);
2736                 addrconf_verify_rtnl();
2737                 return 0;
2738         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2739                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2740                                false, pfx, ifindex);
2741         }
2742
2743         return PTR_ERR(ifp);
2744 }
2745
2746 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2747                           const struct in6_addr *pfx, unsigned int plen)
2748 {
2749         struct inet6_ifaddr *ifp;
2750         struct inet6_dev *idev;
2751         struct net_device *dev;
2752
2753         if (plen > 128)
2754                 return -EINVAL;
2755
2756         dev = __dev_get_by_index(net, ifindex);
2757         if (!dev)
2758                 return -ENODEV;
2759
2760         idev = __in6_dev_get(dev);
2761         if (!idev)
2762                 return -ENXIO;
2763
2764         read_lock_bh(&idev->lock);
2765         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2766                 if (ifp->prefix_len == plen &&
2767                     ipv6_addr_equal(pfx, &ifp->addr)) {
2768                         in6_ifa_hold(ifp);
2769                         read_unlock_bh(&idev->lock);
2770
2771                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2772                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2773                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2774                                                  jiffies);
2775                         ipv6_del_addr(ifp);
2776                         addrconf_verify_rtnl();
2777                         if (ipv6_addr_is_multicast(pfx)) {
2778                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2779                                                false, pfx, dev->ifindex);
2780                         }
2781                         return 0;
2782                 }
2783         }
2784         read_unlock_bh(&idev->lock);
2785         return -EADDRNOTAVAIL;
2786 }
2787
2788
2789 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2790 {
2791         struct in6_ifreq ireq;
2792         int err;
2793
2794         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2795                 return -EPERM;
2796
2797         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2798                 return -EFAULT;
2799
2800         rtnl_lock();
2801         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2802                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2803                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2804         rtnl_unlock();
2805         return err;
2806 }
2807
2808 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2809 {
2810         struct in6_ifreq ireq;
2811         int err;
2812
2813         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2814                 return -EPERM;
2815
2816         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2817                 return -EFAULT;
2818
2819         rtnl_lock();
2820         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2821                              ireq.ifr6_prefixlen);
2822         rtnl_unlock();
2823         return err;
2824 }
2825
2826 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2827                      int plen, int scope)
2828 {
2829         struct inet6_ifaddr *ifp;
2830
2831         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2832                             scope, IFA_F_PERMANENT,
2833                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2834         if (!IS_ERR(ifp)) {
2835                 spin_lock_bh(&ifp->lock);
2836                 ifp->flags &= ~IFA_F_TENTATIVE;
2837                 spin_unlock_bh(&ifp->lock);
2838                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2839                 in6_ifa_put(ifp);
2840         }
2841 }
2842
2843 #if IS_ENABLED(CONFIG_IPV6_SIT)
2844 static void sit_add_v4_addrs(struct inet6_dev *idev)
2845 {
2846         struct in6_addr addr;
2847         struct net_device *dev;
2848         struct net *net = dev_net(idev->dev);
2849         int scope, plen;
2850         u32 pflags = 0;
2851
2852         ASSERT_RTNL();
2853
2854         memset(&addr, 0, sizeof(struct in6_addr));
2855         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2856
2857         if (idev->dev->flags&IFF_POINTOPOINT) {
2858                 addr.s6_addr32[0] = htonl(0xfe800000);
2859                 scope = IFA_LINK;
2860                 plen = 64;
2861         } else {
2862                 scope = IPV6_ADDR_COMPATv4;
2863                 plen = 96;
2864                 pflags |= RTF_NONEXTHOP;
2865         }
2866
2867         if (addr.s6_addr32[3]) {
2868                 add_addr(idev, &addr, plen, scope);
2869                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2870                 return;
2871         }
2872
2873         for_each_netdev(net, dev) {
2874                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2875                 if (in_dev && (dev->flags & IFF_UP)) {
2876                         struct in_ifaddr *ifa;
2877
2878                         int flag = scope;
2879
2880                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2881
2882                                 addr.s6_addr32[3] = ifa->ifa_local;
2883
2884                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2885                                         continue;
2886                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2887                                         if (idev->dev->flags&IFF_POINTOPOINT)
2888                                                 continue;
2889                                         flag |= IFA_HOST;
2890                                 }
2891
2892                                 add_addr(idev, &addr, plen, flag);
2893                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2894                                                       pflags);
2895                         }
2896                 }
2897         }
2898 }
2899 #endif
2900
2901 static void init_loopback(struct net_device *dev)
2902 {
2903         struct inet6_dev  *idev;
2904         struct net_device *sp_dev;
2905         struct inet6_ifaddr *sp_ifa;
2906         struct rt6_info *sp_rt;
2907
2908         /* ::1 */
2909
2910         ASSERT_RTNL();
2911
2912         idev = ipv6_find_idev(dev);
2913         if (!idev) {
2914                 pr_debug("%s: add_dev failed\n", __func__);
2915                 return;
2916         }
2917
2918         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2919
2920         /* Add routes to other interface's IPv6 addresses */
2921         for_each_netdev(dev_net(dev), sp_dev) {
2922                 if (!strcmp(sp_dev->name, dev->name))
2923                         continue;
2924
2925                 idev = __in6_dev_get(sp_dev);
2926                 if (!idev)
2927                         continue;
2928
2929                 read_lock_bh(&idev->lock);
2930                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2931
2932                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2933                                 continue;
2934
2935                         if (sp_ifa->rt) {
2936                                 /* This dst has been added to garbage list when
2937                                  * lo device down, release this obsolete dst and
2938                                  * reallocate a new router for ifa.
2939                                  */
2940                                 if (sp_ifa->rt->dst.obsolete > 0) {
2941                                         ip6_rt_put(sp_ifa->rt);
2942                                         sp_ifa->rt = NULL;
2943                                 } else {
2944                                         continue;
2945                                 }
2946                         }
2947
2948                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2949
2950                         /* Failure cases are ignored */
2951                         if (!IS_ERR(sp_rt)) {
2952                                 sp_ifa->rt = sp_rt;
2953                                 ip6_ins_rt(sp_rt);
2954                         }
2955                 }
2956                 read_unlock_bh(&idev->lock);
2957         }
2958 }
2959
2960 static void addrconf_add_linklocal(struct inet6_dev *idev,
2961                                    const struct in6_addr *addr, u32 flags)
2962 {
2963         struct inet6_ifaddr *ifp;
2964         u32 addr_flags = flags | IFA_F_PERMANENT;
2965
2966 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2967         if (idev->cnf.optimistic_dad &&
2968             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2969                 addr_flags |= IFA_F_OPTIMISTIC;
2970 #endif
2971
2972         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2973                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2974         if (!IS_ERR(ifp)) {
2975                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2976                 addrconf_dad_start(ifp);
2977                 in6_ifa_put(ifp);
2978         }
2979 }
2980
2981 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2982 {
2983         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2984                 return true;
2985
2986         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2987             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2988                 return true;
2989
2990         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2991             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2992                 return true;
2993
2994         return false;
2995 }
2996
2997 static int ipv6_generate_stable_address(struct in6_addr *address,
2998                                         u8 dad_count,
2999                                         const struct inet6_dev *idev)
3000 {
3001         static DEFINE_SPINLOCK(lock);
3002         static __u32 digest[SHA_DIGEST_WORDS];
3003         static __u32 workspace[SHA_WORKSPACE_WORDS];
3004
3005         static union {
3006                 char __data[SHA_MESSAGE_BYTES];
3007                 struct {
3008                         struct in6_addr secret;
3009                         __be32 prefix[2];
3010                         unsigned char hwaddr[MAX_ADDR_LEN];
3011                         u8 dad_count;
3012                 } __packed;
3013         } data;
3014
3015         struct in6_addr secret;
3016         struct in6_addr temp;
3017         struct net *net = dev_net(idev->dev);
3018
3019         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3020
3021         if (idev->cnf.stable_secret.initialized)
3022                 secret = idev->cnf.stable_secret.secret;
3023         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3024                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3025         else
3026                 return -1;
3027
3028 retry:
3029         spin_lock_bh(&lock);
3030
3031         sha_init(digest);
3032         memset(&data, 0, sizeof(data));
3033         memset(workspace, 0, sizeof(workspace));
3034         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3035         data.prefix[0] = address->s6_addr32[0];
3036         data.prefix[1] = address->s6_addr32[1];
3037         data.secret = secret;
3038         data.dad_count = dad_count;
3039
3040         sha_transform(digest, data.__data, workspace);
3041
3042         temp = *address;
3043         temp.s6_addr32[2] = (__force __be32)digest[0];
3044         temp.s6_addr32[3] = (__force __be32)digest[1];
3045
3046         spin_unlock_bh(&lock);
3047
3048         if (ipv6_reserved_interfaceid(temp)) {
3049                 dad_count++;
3050                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3051                         return -1;
3052                 goto retry;
3053         }
3054
3055         *address = temp;
3056         return 0;
3057 }
3058
3059 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3060 {
3061         struct in6_addr addr;
3062
3063         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3064
3065         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3066                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3067                         addrconf_add_linklocal(idev, &addr,
3068                                                IFA_F_STABLE_PRIVACY);
3069                 else if (prefix_route)
3070                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3071         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3072                 /* addrconf_add_linklocal also adds a prefix_route and we
3073                  * only need to care about prefix routes if ipv6_generate_eui64
3074                  * couldn't generate one.
3075                  */
3076                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3077                         addrconf_add_linklocal(idev, &addr, 0);
3078                 else if (prefix_route)
3079                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3080         }
3081 }
3082
3083 static void addrconf_dev_config(struct net_device *dev)
3084 {
3085         struct inet6_dev *idev;
3086
3087         ASSERT_RTNL();
3088
3089         if ((dev->type != ARPHRD_ETHER) &&
3090             (dev->type != ARPHRD_FDDI) &&
3091             (dev->type != ARPHRD_ARCNET) &&
3092             (dev->type != ARPHRD_INFINIBAND) &&
3093             (dev->type != ARPHRD_IEEE802154) &&
3094             (dev->type != ARPHRD_IEEE1394) &&
3095             (dev->type != ARPHRD_TUNNEL6) &&
3096             (dev->type != ARPHRD_6LOWPAN)) {
3097                 /* Alas, we support only Ethernet autoconfiguration. */
3098                 return;
3099         }
3100
3101         idev = addrconf_add_dev(dev);
3102         if (IS_ERR(idev))
3103                 return;
3104
3105         addrconf_addr_gen(idev, false);
3106 }
3107
3108 #if IS_ENABLED(CONFIG_IPV6_SIT)
3109 static void addrconf_sit_config(struct net_device *dev)
3110 {
3111         struct inet6_dev *idev;
3112
3113         ASSERT_RTNL();
3114
3115         /*
3116          * Configure the tunnel with one of our IPv4
3117          * addresses... we should configure all of
3118          * our v4 addrs in the tunnel
3119          */
3120
3121         idev = ipv6_find_idev(dev);
3122         if (!idev) {
3123                 pr_debug("%s: add_dev failed\n", __func__);
3124                 return;
3125         }
3126
3127         if (dev->priv_flags & IFF_ISATAP) {
3128                 addrconf_addr_gen(idev, false);
3129                 return;
3130         }
3131
3132         sit_add_v4_addrs(idev);
3133
3134         if (dev->flags&IFF_POINTOPOINT)
3135                 addrconf_add_mroute(dev);
3136 }
3137 #endif
3138
3139 #if IS_ENABLED(CONFIG_NET_IPGRE)
3140 static void addrconf_gre_config(struct net_device *dev)
3141 {
3142         struct inet6_dev *idev;
3143
3144         ASSERT_RTNL();
3145
3146         idev = ipv6_find_idev(dev);
3147         if (!idev) {
3148                 pr_debug("%s: add_dev failed\n", __func__);
3149                 return;
3150         }
3151
3152         addrconf_addr_gen(idev, true);
3153 }
3154 #endif
3155
3156 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3157                            void *ptr)
3158 {
3159         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3160         struct inet6_dev *idev = __in6_dev_get(dev);
3161         int run_pending = 0;
3162         int err;
3163
3164         switch (event) {
3165         case NETDEV_REGISTER:
3166                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3167                         idev = ipv6_add_dev(dev);
3168                         if (IS_ERR(idev))
3169                                 return notifier_from_errno(PTR_ERR(idev));
3170                 }
3171                 break;
3172
3173         case NETDEV_UP:
3174         case NETDEV_CHANGE:
3175                 if (dev->flags & IFF_SLAVE)
3176                         break;
3177
3178                 if (idev && idev->cnf.disable_ipv6)
3179                         break;
3180
3181                 if (event == NETDEV_UP) {
3182                         if (!addrconf_qdisc_ok(dev)) {
3183                                 /* device is not ready yet. */
3184                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3185                                         dev->name);
3186                                 break;
3187                         }
3188
3189                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3190                                 idev = ipv6_add_dev(dev);
3191
3192                         if (!IS_ERR_OR_NULL(idev)) {
3193                                 idev->if_flags |= IF_READY;
3194                                 run_pending = 1;
3195                         }
3196                 } else {
3197                         if (!addrconf_qdisc_ok(dev)) {
3198                                 /* device is still not ready. */
3199                                 break;
3200                         }
3201
3202                         if (idev) {
3203                                 if (idev->if_flags & IF_READY)
3204                                         /* device is already configured. */
3205                                         break;
3206                                 idev->if_flags |= IF_READY;
3207                         }
3208
3209                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3210                                 dev->name);
3211
3212                         run_pending = 1;
3213                 }
3214
3215                 switch (dev->type) {
3216 #if IS_ENABLED(CONFIG_IPV6_SIT)
3217                 case ARPHRD_SIT:
3218                         addrconf_sit_config(dev);
3219                         break;
3220 #endif
3221 #if IS_ENABLED(CONFIG_NET_IPGRE)
3222                 case ARPHRD_IPGRE:
3223                         addrconf_gre_config(dev);
3224                         break;
3225 #endif
3226                 case ARPHRD_LOOPBACK:
3227                         init_loopback(dev);
3228                         break;
3229
3230                 default:
3231                         addrconf_dev_config(dev);
3232                         break;
3233                 }
3234
3235                 if (!IS_ERR_OR_NULL(idev)) {
3236                         if (run_pending)
3237                                 addrconf_dad_run(idev);
3238
3239                         /*
3240                          * If the MTU changed during the interface down,
3241                          * when the interface up, the changed MTU must be
3242                          * reflected in the idev as well as routers.
3243                          */
3244                         if (idev->cnf.mtu6 != dev->mtu &&
3245                             dev->mtu >= IPV6_MIN_MTU) {
3246                                 rt6_mtu_change(dev, dev->mtu);
3247                                 idev->cnf.mtu6 = dev->mtu;
3248                         }
3249                         idev->tstamp = jiffies;
3250                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3251
3252                         /*
3253                          * If the changed mtu during down is lower than
3254                          * IPV6_MIN_MTU stop IPv6 on this interface.
3255                          */
3256                         if (dev->mtu < IPV6_MIN_MTU)
3257                                 addrconf_ifdown(dev, 1);
3258                 }
3259                 break;
3260
3261         case NETDEV_CHANGEMTU:
3262                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3263                         rt6_mtu_change(dev, dev->mtu);
3264                         idev->cnf.mtu6 = dev->mtu;
3265                         break;
3266                 }
3267
3268                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3269                         idev = ipv6_add_dev(dev);
3270                         if (!IS_ERR(idev))
3271                                 break;
3272                 }
3273
3274                 /*
3275                  * if MTU under IPV6_MIN_MTU.
3276                  * Stop IPv6 on this interface.
3277                  */
3278
3279         case NETDEV_DOWN:
3280         case NETDEV_UNREGISTER:
3281                 /*
3282                  *      Remove all addresses from this interface.
3283                  */
3284                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3285                 break;
3286
3287         case NETDEV_CHANGENAME:
3288                 if (idev) {
3289                         snmp6_unregister_dev(idev);
3290                         addrconf_sysctl_unregister(idev);
3291                         err = addrconf_sysctl_register(idev);
3292                         if (err)
3293                                 return notifier_from_errno(err);
3294                         err = snmp6_register_dev(idev);
3295                         if (err) {
3296                                 addrconf_sysctl_unregister(idev);
3297                                 return notifier_from_errno(err);
3298                         }
3299                 }
3300                 break;
3301
3302         case NETDEV_PRE_TYPE_CHANGE:
3303         case NETDEV_POST_TYPE_CHANGE:
3304                 addrconf_type_change(dev, event);
3305                 break;
3306         }
3307
3308         return NOTIFY_OK;
3309 }
3310
3311 /*
3312  *      addrconf module should be notified of a device going up
3313  */
3314 static struct notifier_block ipv6_dev_notf = {
3315         .notifier_call = addrconf_notify,
3316 };
3317
3318 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3319 {
3320         struct inet6_dev *idev;
3321         ASSERT_RTNL();
3322
3323         idev = __in6_dev_get(dev);
3324
3325         if (event == NETDEV_POST_TYPE_CHANGE)
3326                 ipv6_mc_remap(idev);
3327         else if (event == NETDEV_PRE_TYPE_CHANGE)
3328                 ipv6_mc_unmap(idev);
3329 }
3330
3331 static int addrconf_ifdown(struct net_device *dev, int how)
3332 {
3333         struct net *net = dev_net(dev);
3334         struct inet6_dev *idev;
3335         struct inet6_ifaddr *ifa;
3336         int state, i;
3337
3338         ASSERT_RTNL();
3339
3340         rt6_ifdown(net, dev);
3341         neigh_ifdown(&nd_tbl, dev);
3342
3343         idev = __in6_dev_get(dev);
3344         if (!idev)
3345                 return -ENODEV;
3346
3347         /*
3348          * Step 1: remove reference to ipv6 device from parent device.
3349          *         Do not dev_put!
3350          */
3351         if (how) {
3352                 idev->dead = 1;
3353
3354                 /* protected by rtnl_lock */
3355                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3356
3357                 /* Step 1.5: remove snmp6 entry */
3358                 snmp6_unregister_dev(idev);
3359
3360         }
3361
3362         /* Step 2: clear hash table */
3363         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3364                 struct hlist_head *h = &inet6_addr_lst[i];
3365
3366                 spin_lock_bh(&addrconf_hash_lock);
3367 restart:
3368                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3369                         if (ifa->idev == idev) {
3370                                 hlist_del_init_rcu(&ifa->addr_lst);
3371                                 addrconf_del_dad_work(ifa);
3372                                 goto restart;
3373                         }
3374                 }
3375                 spin_unlock_bh(&addrconf_hash_lock);
3376         }
3377
3378         write_lock_bh(&idev->lock);
3379
3380         addrconf_del_rs_timer(idev);
3381
3382         /* Step 2: clear flags for stateless addrconf */
3383         if (!how)
3384                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3385
3386         if (how && del_timer(&idev->regen_timer))
3387                 in6_dev_put(idev);
3388
3389         /* Step 3: clear tempaddr list */
3390         while (!list_empty(&idev->tempaddr_list)) {
3391                 ifa = list_first_entry(&idev->tempaddr_list,
3392                                        struct inet6_ifaddr, tmp_list);
3393                 list_del(&ifa->tmp_list);
3394                 write_unlock_bh(&idev->lock);
3395                 spin_lock_bh(&ifa->lock);
3396
3397                 if (ifa->ifpub) {
3398                         in6_ifa_put(ifa->ifpub);
3399                         ifa->ifpub = NULL;
3400                 }
3401                 spin_unlock_bh(&ifa->lock);
3402                 in6_ifa_put(ifa);
3403                 write_lock_bh(&idev->lock);
3404         }
3405
3406         while (!list_empty(&idev->addr_list)) {
3407                 ifa = list_first_entry(&idev->addr_list,
3408                                        struct inet6_ifaddr, if_list);
3409                 addrconf_del_dad_work(ifa);
3410
3411                 list_del(&ifa->if_list);
3412
3413                 write_unlock_bh(&idev->lock);
3414
3415                 spin_lock_bh(&ifa->lock);
3416                 state = ifa->state;
3417                 ifa->state = INET6_IFADDR_STATE_DEAD;
3418                 spin_unlock_bh(&ifa->lock);
3419
3420                 if (state != INET6_IFADDR_STATE_DEAD) {
3421                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3422                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3423                 }
3424                 in6_ifa_put(ifa);
3425
3426                 write_lock_bh(&idev->lock);
3427         }
3428
3429         write_unlock_bh(&idev->lock);
3430
3431         /* Step 5: Discard anycast and multicast list */
3432         if (how) {
3433                 ipv6_ac_destroy_dev(idev);
3434                 ipv6_mc_destroy_dev(idev);
3435         } else {
3436                 ipv6_mc_down(idev);
3437         }
3438
3439         idev->tstamp = jiffies;
3440
3441         /* Last: Shot the device (if unregistered) */
3442         if (how) {
3443                 addrconf_sysctl_unregister(idev);
3444                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3445                 neigh_ifdown(&nd_tbl, dev);
3446                 in6_dev_put(idev);
3447         }
3448         return 0;
3449 }
3450
3451 static void addrconf_rs_timer(unsigned long data)
3452 {
3453         struct inet6_dev *idev = (struct inet6_dev *)data;
3454         struct net_device *dev = idev->dev;
3455         struct in6_addr lladdr;
3456
3457         write_lock(&idev->lock);
3458         if (idev->dead || !(idev->if_flags & IF_READY))
3459                 goto out;
3460
3461         if (!ipv6_accept_ra(idev))
3462                 goto out;
3463
3464         /* Announcement received after solicitation was sent */
3465         if (idev->if_flags & IF_RA_RCVD)
3466                 goto out;
3467
3468         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3469                 write_unlock(&idev->lock);
3470                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3471                         ndisc_send_rs(dev, &lladdr,
3472                                       &in6addr_linklocal_allrouters);
3473                 else
3474                         goto put;
3475
3476                 write_lock(&idev->lock);
3477                 /* The wait after the last probe can be shorter */
3478                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3479                                              idev->cnf.rtr_solicits) ?
3480                                       idev->cnf.rtr_solicit_delay :
3481                                       idev->cnf.rtr_solicit_interval);
3482         } else {
3483                 /*
3484                  * Note: we do not support deprecated "all on-link"
3485                  * assumption any longer.
3486                  */
3487                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3488         }
3489
3490 out:
3491         write_unlock(&idev->lock);
3492 put:
3493         in6_dev_put(idev);
3494 }
3495
3496 /*
3497  *      Duplicate Address Detection
3498  */
3499 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3500 {
3501         unsigned long rand_num;
3502         struct inet6_dev *idev = ifp->idev;
3503
3504         if (ifp->flags & IFA_F_OPTIMISTIC)
3505                 rand_num = 0;
3506         else
3507                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3508
3509         ifp->dad_probes = idev->cnf.dad_transmits;
3510         addrconf_mod_dad_work(ifp, rand_num);
3511 }
3512
3513 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3514 {
3515         struct inet6_dev *idev = ifp->idev;
3516         struct net_device *dev = idev->dev;
3517
3518         addrconf_join_solict(dev, &ifp->addr);
3519
3520         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3521
3522         read_lock_bh(&idev->lock);
3523         spin_lock(&ifp->lock);
3524         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3525                 goto out;
3526
3527         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3528             idev->cnf.accept_dad < 1 ||
3529             !(ifp->flags&IFA_F_TENTATIVE) ||
3530             ifp->flags & IFA_F_NODAD) {
3531                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3532                 spin_unlock(&ifp->lock);
3533                 read_unlock_bh(&idev->lock);
3534
3535                 addrconf_dad_completed(ifp);
3536                 return;
3537         }
3538
3539         if (!(idev->if_flags & IF_READY)) {
3540                 spin_unlock(&ifp->lock);
3541                 read_unlock_bh(&idev->lock);
3542                 /*
3543                  * If the device is not ready:
3544                  * - keep it tentative if it is a permanent address.
3545                  * - otherwise, kill it.
3546                  */
3547                 in6_ifa_hold(ifp);
3548                 addrconf_dad_stop(ifp, 0);
3549                 return;
3550         }
3551
3552         /*
3553          * Optimistic nodes can start receiving
3554          * Frames right away
3555          */
3556         if (ifp->flags & IFA_F_OPTIMISTIC) {
3557                 ip6_ins_rt(ifp->rt);
3558                 if (ipv6_use_optimistic_addr(idev)) {
3559                         /* Because optimistic nodes can use this address,
3560                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3561                          */
3562                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3563                 }
3564         }
3565
3566         addrconf_dad_kick(ifp);
3567 out:
3568         spin_unlock(&ifp->lock);
3569         read_unlock_bh(&idev->lock);
3570 }
3571
3572 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3573 {
3574         bool begin_dad = false;
3575
3576         spin_lock_bh(&ifp->lock);
3577         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3578                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3579                 begin_dad = true;
3580         }
3581         spin_unlock_bh(&ifp->lock);
3582
3583         if (begin_dad)
3584                 addrconf_mod_dad_work(ifp, 0);
3585 }
3586
3587 static void addrconf_dad_work(struct work_struct *w)
3588 {
3589         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3590                                                 struct inet6_ifaddr,
3591                                                 dad_work);
3592         struct inet6_dev *idev = ifp->idev;
3593         struct in6_addr mcaddr;
3594
3595         enum {
3596                 DAD_PROCESS,
3597                 DAD_BEGIN,
3598                 DAD_ABORT,
3599         } action = DAD_PROCESS;
3600
3601         rtnl_lock();
3602
3603         spin_lock_bh(&ifp->lock);
3604         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3605                 action = DAD_BEGIN;
3606                 ifp->state = INET6_IFADDR_STATE_DAD;
3607         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3608                 action = DAD_ABORT;
3609                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3610         }
3611         spin_unlock_bh(&ifp->lock);
3612
3613         if (action == DAD_BEGIN) {
3614                 addrconf_dad_begin(ifp);
3615                 goto out;
3616         } else if (action == DAD_ABORT) {
3617                 addrconf_dad_stop(ifp, 1);
3618                 goto out;
3619         }
3620
3621         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3622                 goto out;
3623
3624         write_lock_bh(&idev->lock);
3625         if (idev->dead || !(idev->if_flags & IF_READY)) {
3626                 write_unlock_bh(&idev->lock);
3627                 goto out;
3628         }
3629
3630         spin_lock(&ifp->lock);
3631         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3632                 spin_unlock(&ifp->lock);
3633                 write_unlock_bh(&idev->lock);
3634                 goto out;
3635         }
3636
3637         if (ifp->dad_probes == 0) {
3638                 /*
3639                  * DAD was successful
3640                  */
3641
3642                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3643                 spin_unlock(&ifp->lock);
3644                 write_unlock_bh(&idev->lock);
3645
3646                 addrconf_dad_completed(ifp);
3647
3648                 goto out;
3649         }
3650
3651         ifp->dad_probes--;
3652         addrconf_mod_dad_work(ifp,
3653                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3654         spin_unlock(&ifp->lock);
3655         write_unlock_bh(&idev->lock);
3656
3657         /* send a neighbour solicitation for our addr */
3658         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3659         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any, NULL);
3660 out:
3661         in6_ifa_put(ifp);
3662         rtnl_unlock();
3663 }
3664
3665 /* ifp->idev must be at least read locked */
3666 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3667 {
3668         struct inet6_ifaddr *ifpiter;
3669         struct inet6_dev *idev = ifp->idev;
3670
3671         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3672                 if (ifpiter->scope > IFA_LINK)
3673                         break;
3674                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3675                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3676                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3677                     IFA_F_PERMANENT)
3678                         return false;
3679         }
3680         return true;
3681 }
3682
3683 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3684 {
3685         struct net_device *dev = ifp->idev->dev;
3686         struct in6_addr lladdr;
3687         bool send_rs, send_mld;
3688
3689         addrconf_del_dad_work(ifp);
3690
3691         /*
3692          *      Configure the address for reception. Now it is valid.
3693          */
3694
3695         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3696
3697         /* If added prefix is link local and we are prepared to process
3698            router advertisements, start sending router solicitations.
3699          */
3700
3701         read_lock_bh(&ifp->idev->lock);
3702         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3703         send_rs = send_mld &&
3704                   ipv6_accept_ra(ifp->idev) &&
3705                   ifp->idev->cnf.rtr_solicits > 0 &&
3706                   (dev->flags&IFF_LOOPBACK) == 0;
3707         read_unlock_bh(&ifp->idev->lock);
3708
3709         /* While dad is in progress mld report's source address is in6_addrany.
3710          * Resend with proper ll now.
3711          */
3712         if (send_mld)
3713                 ipv6_mc_dad_complete(ifp->idev);
3714
3715         if (send_rs) {
3716                 /*
3717                  *      If a host as already performed a random delay
3718                  *      [...] as part of DAD [...] there is no need
3719                  *      to delay again before sending the first RS
3720                  */
3721                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3722                         return;
3723                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3724
3725                 write_lock_bh(&ifp->idev->lock);
3726                 spin_lock(&ifp->lock);
3727                 ifp->idev->rs_probes = 1;
3728                 ifp->idev->if_flags |= IF_RS_SENT;
3729                 addrconf_mod_rs_timer(ifp->idev,
3730                                       ifp->idev->cnf.rtr_solicit_interval);
3731                 spin_unlock(&ifp->lock);
3732                 write_unlock_bh(&ifp->idev->lock);
3733         }
3734 }
3735
3736 static void addrconf_dad_run(struct inet6_dev *idev)
3737 {
3738         struct inet6_ifaddr *ifp;
3739
3740         read_lock_bh(&idev->lock);
3741         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3742                 spin_lock(&ifp->lock);
3743                 if (ifp->flags & IFA_F_TENTATIVE &&
3744                     ifp->state == INET6_IFADDR_STATE_DAD)
3745                         addrconf_dad_kick(ifp);
3746                 spin_unlock(&ifp->lock);
3747         }
3748         read_unlock_bh(&idev->lock);
3749 }
3750
3751 #ifdef CONFIG_PROC_FS
3752 struct if6_iter_state {
3753         struct seq_net_private p;
3754         int bucket;
3755         int offset;
3756 };
3757
3758 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3759 {
3760         struct inet6_ifaddr *ifa = NULL;
3761         struct if6_iter_state *state = seq->private;
3762         struct net *net = seq_file_net(seq);
3763         int p = 0;
3764
3765         /* initial bucket if pos is 0 */
3766         if (pos == 0) {
3767                 state->bucket = 0;
3768                 state->offset = 0;
3769         }
3770
3771         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3772                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3773                                          addr_lst) {
3774                         if (!net_eq(dev_net(ifa->idev->dev), net))
3775                                 continue;
3776                         /* sync with offset */
3777                         if (p < state->offset) {
3778                                 p++;
3779                                 continue;
3780                         }
3781                         state->offset++;
3782                         return ifa;
3783                 }
3784
3785                 /* prepare for next bucket */
3786                 state->offset = 0;
3787                 p = 0;
3788         }
3789         return NULL;
3790 }
3791
3792 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3793                                          struct inet6_ifaddr *ifa)
3794 {
3795         struct if6_iter_state *state = seq->private;
3796         struct net *net = seq_file_net(seq);
3797
3798         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3799                 if (!net_eq(dev_net(ifa->idev->dev), net))
3800                         continue;
3801                 state->offset++;
3802                 return ifa;
3803         }
3804
3805         while (++state->bucket < IN6_ADDR_HSIZE) {
3806                 state->offset = 0;
3807                 hlist_for_each_entry_rcu_bh(ifa,
3808                                      &inet6_addr_lst[state->bucket], addr_lst) {
3809                         if (!net_eq(dev_net(ifa->idev->dev), net))
3810                                 continue;
3811                         state->offset++;
3812                         return ifa;
3813                 }
3814         }
3815
3816         return NULL;
3817 }
3818
3819 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3820         __acquires(rcu_bh)
3821 {
3822         rcu_read_lock_bh();
3823         return if6_get_first(seq, *pos);
3824 }
3825
3826 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3827 {
3828         struct inet6_ifaddr *ifa;
3829
3830         ifa = if6_get_next(seq, v);
3831         ++*pos;
3832         return ifa;
3833 }
3834
3835 static void if6_seq_stop(struct seq_file *seq, void *v)
3836         __releases(rcu_bh)
3837 {
3838         rcu_read_unlock_bh();
3839 }
3840
3841 static int if6_seq_show(struct seq_file *seq, void *v)
3842 {
3843         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3844         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3845                    &ifp->addr,
3846                    ifp->idev->dev->ifindex,
3847                    ifp->prefix_len,
3848                    ifp->scope,
3849                    (u8) ifp->flags,
3850                    ifp->idev->dev->name);
3851         return 0;
3852 }
3853
3854 static const struct seq_operations if6_seq_ops = {
3855         .start  = if6_seq_start,
3856         .next   = if6_seq_next,
3857         .show   = if6_seq_show,
3858         .stop   = if6_seq_stop,
3859 };
3860
3861 static int if6_seq_open(struct inode *inode, struct file *file)
3862 {
3863         return seq_open_net(inode, file, &if6_seq_ops,
3864                             sizeof(struct if6_iter_state));
3865 }
3866
3867 static const struct file_operations if6_fops = {
3868         .owner          = THIS_MODULE,
3869         .open           = if6_seq_open,
3870         .read           = seq_read,
3871         .llseek         = seq_lseek,
3872         .release        = seq_release_net,
3873 };
3874
3875 static int __net_init if6_proc_net_init(struct net *net)
3876 {
3877         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3878                 return -ENOMEM;
3879         return 0;
3880 }
3881
3882 static void __net_exit if6_proc_net_exit(struct net *net)
3883 {
3884         remove_proc_entry("if_inet6", net->proc_net);
3885 }
3886
3887 static struct pernet_operations if6_proc_net_ops = {
3888         .init = if6_proc_net_init,
3889         .exit = if6_proc_net_exit,
3890 };
3891
3892 int __init if6_proc_init(void)
3893 {
3894         return register_pernet_subsys(&if6_proc_net_ops);
3895 }
3896
3897 void if6_proc_exit(void)
3898 {
3899         unregister_pernet_subsys(&if6_proc_net_ops);
3900 }
3901 #endif  /* CONFIG_PROC_FS */
3902
3903 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3904 /* Check if address is a home address configured on any interface. */
3905 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3906 {
3907         int ret = 0;
3908         struct inet6_ifaddr *ifp = NULL;
3909         unsigned int hash = inet6_addr_hash(addr);
3910
3911         rcu_read_lock_bh();
3912         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3913                 if (!net_eq(dev_net(ifp->idev->dev), net))
3914                         continue;
3915                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3916                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3917                         ret = 1;
3918                         break;
3919                 }
3920         }
3921         rcu_read_unlock_bh();
3922         return ret;
3923 }
3924 #endif
3925
3926 /*
3927  *      Periodic address status verification
3928  */
3929
3930 static void addrconf_verify_rtnl(void)
3931 {
3932         unsigned long now, next, next_sec, next_sched;
3933         struct inet6_ifaddr *ifp;
3934         int i;
3935
3936         ASSERT_RTNL();
3937
3938         rcu_read_lock_bh();
3939         now = jiffies;
3940         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3941
3942         cancel_delayed_work(&addr_chk_work);
3943
3944         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3945 restart:
3946                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3947                         unsigned long age;
3948
3949                         /* When setting preferred_lft to a value not zero or
3950                          * infinity, while valid_lft is infinity
3951                          * IFA_F_PERMANENT has a non-infinity life time.
3952                          */
3953                         if ((ifp->flags & IFA_F_PERMANENT) &&
3954                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3955                                 continue;
3956
3957                         spin_lock(&ifp->lock);
3958                         /* We try to batch several events at once. */
3959                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3960
3961                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3962                             age >= ifp->valid_lft) {
3963                                 spin_unlock(&ifp->lock);
3964                                 in6_ifa_hold(ifp);
3965                                 ipv6_del_addr(ifp);
3966                                 goto restart;
3967                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3968                                 spin_unlock(&ifp->lock);
3969                                 continue;
3970                         } else if (age >= ifp->prefered_lft) {
3971                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3972                                 int deprecate = 0;
3973
3974                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3975                                         deprecate = 1;
3976                                         ifp->flags |= IFA_F_DEPRECATED;
3977                                 }
3978
3979                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3980                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3981                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3982
3983                                 spin_unlock(&ifp->lock);
3984
3985                                 if (deprecate) {
3986                                         in6_ifa_hold(ifp);
3987
3988                                         ipv6_ifa_notify(0, ifp);
3989                                         in6_ifa_put(ifp);
3990                                         goto restart;
3991                                 }
3992                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3993                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3994                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3995                                         ifp->idev->cnf.dad_transmits *
3996                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3997
3998                                 if (age >= ifp->prefered_lft - regen_advance) {
3999                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4000                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4001                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4002                                         if (!ifp->regen_count && ifpub) {
4003                                                 ifp->regen_count++;
4004                                                 in6_ifa_hold(ifp);
4005                                                 in6_ifa_hold(ifpub);
4006                                                 spin_unlock(&ifp->lock);
4007
4008                                                 spin_lock(&ifpub->lock);
4009                                                 ifpub->regen_count = 0;
4010                                                 spin_unlock(&ifpub->lock);
4011                                                 ipv6_create_tempaddr(ifpub, ifp);
4012                                                 in6_ifa_put(ifpub);
4013                                                 in6_ifa_put(ifp);
4014                                                 goto restart;
4015                                         }
4016                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4017                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4018                                 spin_unlock(&ifp->lock);
4019                         } else {
4020                                 /* ifp->prefered_lft <= ifp->valid_lft */
4021                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4022                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4023                                 spin_unlock(&ifp->lock);
4024                         }
4025                 }
4026         }
4027
4028         next_sec = round_jiffies_up(next);
4029         next_sched = next;
4030
4031         /* If rounded timeout is accurate enough, accept it. */
4032         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4033                 next_sched = next_sec;
4034
4035         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4036         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4037                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4038
4039         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4040               now, next, next_sec, next_sched);
4041         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4042         rcu_read_unlock_bh();
4043 }
4044
4045 static void addrconf_verify_work(struct work_struct *w)
4046 {
4047         rtnl_lock();
4048         addrconf_verify_rtnl();
4049         rtnl_unlock();
4050 }
4051
4052 static void addrconf_verify(void)
4053 {
4054         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4055 }
4056
4057 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4058                                      struct in6_addr **peer_pfx)
4059 {
4060         struct in6_addr *pfx = NULL;
4061
4062         *peer_pfx = NULL;
4063
4064         if (addr)
4065                 pfx = nla_data(addr);
4066
4067         if (local) {
4068                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4069                         *peer_pfx = pfx;
4070                 pfx = nla_data(local);
4071         }
4072
4073         return pfx;
4074 }
4075
4076 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4077         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4078         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4079         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4080         [IFA_FLAGS]             = { .len = sizeof(u32) },
4081 };
4082
4083 static int
4084 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4085 {
4086         struct net *net = sock_net(skb->sk);
4087         struct ifaddrmsg *ifm;
4088         struct nlattr *tb[IFA_MAX+1];
4089         struct in6_addr *pfx, *peer_pfx;
4090         u32 ifa_flags;
4091         int err;
4092
4093         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4094         if (err < 0)
4095                 return err;
4096
4097         ifm = nlmsg_data(nlh);
4098         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4099         if (!pfx)
4100                 return -EINVAL;
4101
4102         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4103
4104         /* We ignore other flags so far. */
4105         ifa_flags &= IFA_F_MANAGETEMPADDR;
4106
4107         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4108                               ifm->ifa_prefixlen);
4109 }
4110
4111 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4112                              u32 prefered_lft, u32 valid_lft)
4113 {
4114         u32 flags;
4115         clock_t expires;
4116         unsigned long timeout;
4117         bool was_managetempaddr;
4118         bool had_prefixroute;
4119
4120         ASSERT_RTNL();
4121
4122         if (!valid_lft || (prefered_lft > valid_lft))
4123                 return -EINVAL;
4124
4125         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4126             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4127                 return -EINVAL;
4128
4129         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4130         if (addrconf_finite_timeout(timeout)) {
4131                 expires = jiffies_to_clock_t(timeout * HZ);
4132                 valid_lft = timeout;
4133                 flags = RTF_EXPIRES;
4134         } else {
4135                 expires = 0;
4136                 flags = 0;
4137                 ifa_flags |= IFA_F_PERMANENT;
4138         }
4139
4140         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4141         if (addrconf_finite_timeout(timeout)) {
4142                 if (timeout == 0)
4143                         ifa_flags |= IFA_F_DEPRECATED;
4144                 prefered_lft = timeout;
4145         }
4146
4147         spin_lock_bh(&ifp->lock);
4148         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4149         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4150                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4151         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4152                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4153                         IFA_F_NOPREFIXROUTE);
4154         ifp->flags |= ifa_flags;
4155         ifp->tstamp = jiffies;
4156         ifp->valid_lft = valid_lft;
4157         ifp->prefered_lft = prefered_lft;
4158
4159         spin_unlock_bh(&ifp->lock);
4160         if (!(ifp->flags&IFA_F_TENTATIVE))
4161                 ipv6_ifa_notify(0, ifp);
4162
4163         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4164                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4165                                       expires, flags);
4166         } else if (had_prefixroute) {
4167                 enum cleanup_prefix_rt_t action;
4168                 unsigned long rt_expires;
4169
4170                 write_lock_bh(&ifp->idev->lock);
4171                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4172                 write_unlock_bh(&ifp->idev->lock);
4173
4174                 if (action != CLEANUP_PREFIX_RT_NOP) {
4175                         cleanup_prefix_route(ifp, rt_expires,
4176                                 action == CLEANUP_PREFIX_RT_DEL);
4177                 }
4178         }
4179
4180         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4181                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4182                         valid_lft = prefered_lft = 0;
4183                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4184                                  !was_managetempaddr, jiffies);
4185         }
4186
4187         addrconf_verify_rtnl();
4188
4189         return 0;
4190 }
4191
4192 static int
4193 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4194 {
4195         struct net *net = sock_net(skb->sk);
4196         struct ifaddrmsg *ifm;
4197         struct nlattr *tb[IFA_MAX+1];
4198         struct in6_addr *pfx, *peer_pfx;
4199         struct inet6_ifaddr *ifa;
4200         struct net_device *dev;
4201         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4202         u32 ifa_flags;
4203         int err;
4204
4205         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4206         if (err < 0)
4207                 return err;
4208
4209         ifm = nlmsg_data(nlh);
4210         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4211         if (!pfx)
4212                 return -EINVAL;
4213
4214         if (tb[IFA_CACHEINFO]) {
4215                 struct ifa_cacheinfo *ci;
4216
4217                 ci = nla_data(tb[IFA_CACHEINFO]);
4218                 valid_lft = ci->ifa_valid;
4219                 preferred_lft = ci->ifa_prefered;
4220         } else {
4221                 preferred_lft = INFINITY_LIFE_TIME;
4222                 valid_lft = INFINITY_LIFE_TIME;
4223         }
4224
4225         dev =  __dev_get_by_index(net, ifm->ifa_index);
4226         if (!dev)
4227                 return -ENODEV;
4228
4229         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4230
4231         /* We ignore other flags so far. */
4232         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4233                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4234
4235         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4236         if (!ifa) {
4237                 /*
4238                  * It would be best to check for !NLM_F_CREATE here but
4239                  * userspace already relies on not having to provide this.
4240                  */
4241                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4242                                       ifm->ifa_prefixlen, ifa_flags,
4243                                       preferred_lft, valid_lft);
4244         }
4245
4246         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4247             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4248                 err = -EEXIST;
4249         else
4250                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4251
4252         in6_ifa_put(ifa);
4253
4254         return err;
4255 }
4256
4257 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4258                           u8 scope, int ifindex)
4259 {
4260         struct ifaddrmsg *ifm;
4261
4262         ifm = nlmsg_data(nlh);
4263         ifm->ifa_family = AF_INET6;
4264         ifm->ifa_prefixlen = prefixlen;
4265         ifm->ifa_flags = flags;
4266         ifm->ifa_scope = scope;
4267         ifm->ifa_index = ifindex;
4268 }
4269
4270 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4271                          unsigned long tstamp, u32 preferred, u32 valid)
4272 {
4273         struct ifa_cacheinfo ci;
4274
4275         ci.cstamp = cstamp_delta(cstamp);
4276         ci.tstamp = cstamp_delta(tstamp);
4277         ci.ifa_prefered = preferred;
4278         ci.ifa_valid = valid;
4279
4280         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4281 }
4282
4283 static inline int rt_scope(int ifa_scope)
4284 {
4285         if (ifa_scope & IFA_HOST)
4286                 return RT_SCOPE_HOST;
4287         else if (ifa_scope & IFA_LINK)
4288                 return RT_SCOPE_LINK;
4289         else if (ifa_scope & IFA_SITE)
4290                 return RT_SCOPE_SITE;
4291         else
4292                 return RT_SCOPE_UNIVERSE;
4293 }
4294
4295 static inline int inet6_ifaddr_msgsize(void)
4296 {
4297         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4298                + nla_total_size(16) /* IFA_LOCAL */
4299                + nla_total_size(16) /* IFA_ADDRESS */
4300                + nla_total_size(sizeof(struct ifa_cacheinfo))
4301                + nla_total_size(4)  /* IFA_FLAGS */;
4302 }
4303
4304 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4305                              u32 portid, u32 seq, int event, unsigned int flags)
4306 {
4307         struct nlmsghdr  *nlh;
4308         u32 preferred, valid;
4309
4310         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4311         if (!nlh)
4312                 return -EMSGSIZE;
4313
4314         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4315                       ifa->idev->dev->ifindex);
4316
4317         if (!((ifa->flags&IFA_F_PERMANENT) &&
4318               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4319                 preferred = ifa->prefered_lft;
4320                 valid = ifa->valid_lft;
4321                 if (preferred != INFINITY_LIFE_TIME) {
4322                         long tval = (jiffies - ifa->tstamp)/HZ;
4323                         if (preferred > tval)
4324                                 preferred -= tval;
4325                         else
4326                                 preferred = 0;
4327                         if (valid != INFINITY_LIFE_TIME) {
4328                                 if (valid > tval)
4329                                         valid -= tval;
4330                                 else
4331                                         valid = 0;
4332                         }
4333                 }
4334         } else {
4335                 preferred = INFINITY_LIFE_TIME;
4336                 valid = INFINITY_LIFE_TIME;
4337         }
4338
4339         if (!ipv6_addr_any(&ifa->peer_addr)) {
4340                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4341                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4342                         goto error;
4343         } else
4344                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4345                         goto error;
4346
4347         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4348                 goto error;
4349
4350         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4351                 goto error;
4352
4353         nlmsg_end(skb, nlh);
4354         return 0;
4355
4356 error:
4357         nlmsg_cancel(skb, nlh);
4358         return -EMSGSIZE;
4359 }
4360
4361 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4362                                 u32 portid, u32 seq, int event, u16 flags)
4363 {
4364         struct nlmsghdr  *nlh;
4365         u8 scope = RT_SCOPE_UNIVERSE;
4366         int ifindex = ifmca->idev->dev->ifindex;
4367
4368         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4369                 scope = RT_SCOPE_SITE;
4370
4371         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4372         if (!nlh)
4373                 return -EMSGSIZE;
4374
4375         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4376         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4377             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4378                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4379                 nlmsg_cancel(skb, nlh);
4380                 return -EMSGSIZE;
4381         }
4382
4383         nlmsg_end(skb, nlh);
4384         return 0;
4385 }
4386
4387 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4388                                 u32 portid, u32 seq, int event, unsigned int flags)
4389 {
4390         struct nlmsghdr  *nlh;
4391         u8 scope = RT_SCOPE_UNIVERSE;
4392         int ifindex = ifaca->aca_idev->dev->ifindex;
4393
4394         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4395                 scope = RT_SCOPE_SITE;
4396
4397         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4398         if (!nlh)
4399                 return -EMSGSIZE;
4400
4401         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4402         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4403             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4404                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4405                 nlmsg_cancel(skb, nlh);
4406                 return -EMSGSIZE;
4407         }
4408
4409         nlmsg_end(skb, nlh);
4410         return 0;
4411 }
4412
4413 enum addr_type_t {
4414         UNICAST_ADDR,
4415         MULTICAST_ADDR,
4416         ANYCAST_ADDR,
4417 };
4418
4419 /* called with rcu_read_lock() */
4420 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4421                           struct netlink_callback *cb, enum addr_type_t type,
4422                           int s_ip_idx, int *p_ip_idx)
4423 {
4424         struct ifmcaddr6 *ifmca;
4425         struct ifacaddr6 *ifaca;
4426         int err = 1;
4427         int ip_idx = *p_ip_idx;
4428
4429         read_lock_bh(&idev->lock);
4430         switch (type) {
4431         case UNICAST_ADDR: {
4432                 struct inet6_ifaddr *ifa;
4433
4434                 /* unicast address incl. temp addr */
4435                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4436                         if (++ip_idx < s_ip_idx)
4437                                 continue;
4438                         err = inet6_fill_ifaddr(skb, ifa,
4439                                                 NETLINK_CB(cb->skb).portid,
4440                                                 cb->nlh->nlmsg_seq,
4441                                                 RTM_NEWADDR,
4442                                                 NLM_F_MULTI);
4443                         if (err < 0)
4444                                 break;
4445                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4446                 }
4447                 break;
4448         }
4449         case MULTICAST_ADDR:
4450                 /* multicast address */
4451                 for (ifmca = idev->mc_list; ifmca;
4452                      ifmca = ifmca->next, ip_idx++) {
4453                         if (ip_idx < s_ip_idx)
4454                                 continue;
4455                         err = inet6_fill_ifmcaddr(skb, ifmca,
4456                                                   NETLINK_CB(cb->skb).portid,
4457                                                   cb->nlh->nlmsg_seq,
4458                                                   RTM_GETMULTICAST,
4459                                                   NLM_F_MULTI);
4460                         if (err < 0)
4461                                 break;
4462                 }
4463                 break;
4464         case ANYCAST_ADDR:
4465                 /* anycast address */
4466                 for (ifaca = idev->ac_list; ifaca;
4467                      ifaca = ifaca->aca_next, ip_idx++) {
4468                         if (ip_idx < s_ip_idx)
4469                                 continue;
4470                         err = inet6_fill_ifacaddr(skb, ifaca,
4471                                                   NETLINK_CB(cb->skb).portid,
4472                                                   cb->nlh->nlmsg_seq,
4473                                                   RTM_GETANYCAST,
4474                                                   NLM_F_MULTI);
4475                         if (err < 0)
4476                                 break;
4477                 }
4478                 break;
4479         default:
4480                 break;
4481         }
4482         read_unlock_bh(&idev->lock);
4483         *p_ip_idx = ip_idx;
4484         return err;
4485 }
4486
4487 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4488                            enum addr_type_t type)
4489 {
4490         struct net *net = sock_net(skb->sk);
4491         int h, s_h;
4492         int idx, ip_idx;
4493         int s_idx, s_ip_idx;
4494         struct net_device *dev;
4495         struct inet6_dev *idev;
4496         struct hlist_head *head;
4497
4498         s_h = cb->args[0];
4499         s_idx = idx = cb->args[1];
4500         s_ip_idx = ip_idx = cb->args[2];
4501
4502         rcu_read_lock();
4503         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4504         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4505                 idx = 0;
4506                 head = &net->dev_index_head[h];
4507                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4508                         if (idx < s_idx)
4509                                 goto cont;
4510                         if (h > s_h || idx > s_idx)
4511                                 s_ip_idx = 0;
4512                         ip_idx = 0;
4513                         idev = __in6_dev_get(dev);
4514                         if (!idev)
4515                                 goto cont;
4516
4517                         if (in6_dump_addrs(idev, skb, cb, type,
4518                                            s_ip_idx, &ip_idx) < 0)
4519                                 goto done;
4520 cont:
4521                         idx++;
4522                 }
4523         }
4524 done:
4525         rcu_read_unlock();
4526         cb->args[0] = h;
4527         cb->args[1] = idx;
4528         cb->args[2] = ip_idx;
4529
4530         return skb->len;
4531 }
4532
4533 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4534 {
4535         enum addr_type_t type = UNICAST_ADDR;
4536
4537         return inet6_dump_addr(skb, cb, type);
4538 }
4539
4540 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4541 {
4542         enum addr_type_t type = MULTICAST_ADDR;
4543
4544         return inet6_dump_addr(skb, cb, type);
4545 }
4546
4547
4548 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4549 {
4550         enum addr_type_t type = ANYCAST_ADDR;
4551
4552         return inet6_dump_addr(skb, cb, type);
4553 }
4554
4555 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4556 {
4557         struct net *net = sock_net(in_skb->sk);
4558         struct ifaddrmsg *ifm;
4559         struct nlattr *tb[IFA_MAX+1];
4560         struct in6_addr *addr = NULL, *peer;
4561         struct net_device *dev = NULL;
4562         struct inet6_ifaddr *ifa;
4563         struct sk_buff *skb;
4564         int err;
4565
4566         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4567         if (err < 0)
4568                 goto errout;
4569
4570         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4571         if (!addr) {
4572                 err = -EINVAL;
4573                 goto errout;
4574         }
4575
4576         ifm = nlmsg_data(nlh);
4577         if (ifm->ifa_index)
4578                 dev = __dev_get_by_index(net, ifm->ifa_index);
4579
4580         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4581         if (!ifa) {
4582                 err = -EADDRNOTAVAIL;
4583                 goto errout;
4584         }
4585
4586         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4587         if (!skb) {
4588                 err = -ENOBUFS;
4589                 goto errout_ifa;
4590         }
4591
4592         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4593                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4594         if (err < 0) {
4595                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4596                 WARN_ON(err == -EMSGSIZE);
4597                 kfree_skb(skb);
4598                 goto errout_ifa;
4599         }
4600         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4601 errout_ifa:
4602         in6_ifa_put(ifa);
4603 errout:
4604         return err;
4605 }
4606
4607 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4608 {
4609         struct sk_buff *skb;
4610         struct net *net = dev_net(ifa->idev->dev);
4611         int err = -ENOBUFS;
4612
4613         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4614         if (!skb)
4615                 goto errout;
4616
4617         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4618         if (err < 0) {
4619                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4620                 WARN_ON(err == -EMSGSIZE);
4621                 kfree_skb(skb);
4622                 goto errout;
4623         }
4624         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4625         return;
4626 errout:
4627         if (err < 0)
4628                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4629 }
4630
4631 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4632                                 __s32 *array, int bytes)
4633 {
4634         BUG_ON(bytes < (DEVCONF_MAX * 4));
4635
4636         memset(array, 0, bytes);
4637         array[DEVCONF_FORWARDING] = cnf->forwarding;
4638         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4639         array[DEVCONF_MTU6] = cnf->mtu6;
4640         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4641         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4642         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4643         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4644         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4645         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4646                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4647         array[DEVCONF_RTR_SOLICIT_DELAY] =
4648                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4649         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4650         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4651                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4652         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4653                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4654         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4655         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4656         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4657         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4658         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4659         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4660         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4661         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4662         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4663 #ifdef CONFIG_IPV6_ROUTER_PREF
4664         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4665         array[DEVCONF_RTR_PROBE_INTERVAL] =
4666                 jiffies_to_msecs(cnf->rtr_probe_interval);
4667 #ifdef CONFIG_IPV6_ROUTE_INFO
4668         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4669 #endif
4670 #endif
4671         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4672         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4673 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4674         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4675         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4676 #endif
4677 #ifdef CONFIG_IPV6_MROUTE
4678         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4679 #endif
4680         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4681         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4682         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4683         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4684         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4685         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4686         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4687         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4688         /* we omit DEVCONF_STABLE_SECRET for now */
4689         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4690 }
4691
4692 static inline size_t inet6_ifla6_size(void)
4693 {
4694         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4695              + nla_total_size(sizeof(struct ifla_cacheinfo))
4696              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4697              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4698              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4699              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4700 }
4701
4702 static inline size_t inet6_if_nlmsg_size(void)
4703 {
4704         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4705                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4706                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4707                + nla_total_size(4) /* IFLA_MTU */
4708                + nla_total_size(4) /* IFLA_LINK */
4709                + nla_total_size(1) /* IFLA_OPERSTATE */
4710                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4711 }
4712
4713 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4714                                       int items, int bytes)
4715 {
4716         int i;
4717         int pad = bytes - sizeof(u64) * items;
4718         BUG_ON(pad < 0);
4719
4720         /* Use put_unaligned() because stats may not be aligned for u64. */
4721         put_unaligned(items, &stats[0]);
4722         for (i = 1; i < items; i++)
4723                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4724
4725         memset(&stats[items], 0, pad);
4726 }
4727
4728 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4729                                         int bytes, size_t syncpoff)
4730 {
4731         int i, c;
4732         u64 buff[IPSTATS_MIB_MAX];
4733         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4734
4735         BUG_ON(pad < 0);
4736
4737         memset(buff, 0, sizeof(buff));
4738         buff[0] = IPSTATS_MIB_MAX;
4739
4740         for_each_possible_cpu(c) {
4741                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4742                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4743         }
4744
4745         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4746         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4747 }
4748
4749 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4750                              int bytes)
4751 {
4752         switch (attrtype) {
4753         case IFLA_INET6_STATS:
4754                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4755                                      offsetof(struct ipstats_mib, syncp));
4756                 break;
4757         case IFLA_INET6_ICMP6STATS:
4758                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4759                 break;
4760         }
4761 }
4762
4763 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4764 {
4765         struct nlattr *nla;
4766         struct ifla_cacheinfo ci;
4767
4768         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4769                 goto nla_put_failure;
4770         ci.max_reasm_len = IPV6_MAXPLEN;
4771         ci.tstamp = cstamp_delta(idev->tstamp);
4772         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4773         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4774         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4775                 goto nla_put_failure;
4776         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4777         if (!nla)
4778                 goto nla_put_failure;
4779         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4780
4781         /* XXX - MC not implemented */
4782
4783         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4784         if (!nla)
4785                 goto nla_put_failure;
4786         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4787
4788         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4789         if (!nla)
4790                 goto nla_put_failure;
4791         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4792
4793         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4794         if (!nla)
4795                 goto nla_put_failure;
4796
4797         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4798                 goto nla_put_failure;
4799
4800         read_lock_bh(&idev->lock);
4801         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4802         read_unlock_bh(&idev->lock);
4803
4804         return 0;
4805
4806 nla_put_failure:
4807         return -EMSGSIZE;
4808 }
4809
4810 static size_t inet6_get_link_af_size(const struct net_device *dev)
4811 {
4812         if (!__in6_dev_get(dev))
4813                 return 0;
4814
4815         return inet6_ifla6_size();
4816 }
4817
4818 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4819 {
4820         struct inet6_dev *idev = __in6_dev_get(dev);
4821
4822         if (!idev)
4823                 return -ENODATA;
4824
4825         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4826                 return -EMSGSIZE;
4827
4828         return 0;
4829 }
4830
4831 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4832 {
4833         struct inet6_ifaddr *ifp;
4834         struct net_device *dev = idev->dev;
4835         bool update_rs = false;
4836         struct in6_addr ll_addr;
4837
4838         ASSERT_RTNL();
4839
4840         if (!token)
4841                 return -EINVAL;
4842         if (ipv6_addr_any(token))
4843                 return -EINVAL;
4844         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4845                 return -EINVAL;
4846         if (!ipv6_accept_ra(idev))
4847                 return -EINVAL;
4848         if (idev->cnf.rtr_solicits <= 0)
4849                 return -EINVAL;
4850
4851         write_lock_bh(&idev->lock);
4852
4853         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4854         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4855
4856         write_unlock_bh(&idev->lock);
4857
4858         if (!idev->dead && (idev->if_flags & IF_READY) &&
4859             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4860                              IFA_F_OPTIMISTIC)) {
4861
4862                 /* If we're not ready, then normal ifup will take care
4863                  * of this. Otherwise, we need to request our rs here.
4864                  */
4865                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4866                 update_rs = true;
4867         }
4868
4869         write_lock_bh(&idev->lock);
4870
4871         if (update_rs) {
4872                 idev->if_flags |= IF_RS_SENT;
4873                 idev->rs_probes = 1;
4874                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4875         }
4876
4877         /* Well, that's kinda nasty ... */
4878         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4879                 spin_lock(&ifp->lock);
4880                 if (ifp->tokenized) {
4881                         ifp->valid_lft = 0;
4882                         ifp->prefered_lft = 0;
4883                 }
4884                 spin_unlock(&ifp->lock);
4885         }
4886
4887         write_unlock_bh(&idev->lock);
4888         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4889         addrconf_verify_rtnl();
4890         return 0;
4891 }
4892
4893 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4894         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4895         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4896 };
4897
4898 static int inet6_validate_link_af(const struct net_device *dev,
4899                                   const struct nlattr *nla)
4900 {
4901         struct nlattr *tb[IFLA_INET6_MAX + 1];
4902
4903         if (dev && !__in6_dev_get(dev))
4904                 return -EAFNOSUPPORT;
4905
4906         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4907 }
4908
4909 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4910 {
4911         int err = -EINVAL;
4912         struct inet6_dev *idev = __in6_dev_get(dev);
4913         struct nlattr *tb[IFLA_INET6_MAX + 1];
4914
4915         if (!idev)
4916                 return -EAFNOSUPPORT;
4917
4918         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4919                 BUG();
4920
4921         if (tb[IFLA_INET6_TOKEN]) {
4922                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4923                 if (err)
4924                         return err;
4925         }
4926
4927         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4928                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4929
4930                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4931                     mode != IN6_ADDR_GEN_MODE_NONE &&
4932                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4933                         return -EINVAL;
4934
4935                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4936                     !idev->cnf.stable_secret.initialized &&
4937                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4938                         return -EINVAL;
4939
4940                 idev->addr_gen_mode = mode;
4941                 err = 0;
4942         }
4943
4944         return err;
4945 }
4946
4947 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4948                              u32 portid, u32 seq, int event, unsigned int flags)
4949 {
4950         struct net_device *dev = idev->dev;
4951         struct ifinfomsg *hdr;
4952         struct nlmsghdr *nlh;
4953         void *protoinfo;
4954
4955         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4956         if (!nlh)
4957                 return -EMSGSIZE;
4958
4959         hdr = nlmsg_data(nlh);
4960         hdr->ifi_family = AF_INET6;
4961         hdr->__ifi_pad = 0;
4962         hdr->ifi_type = dev->type;
4963         hdr->ifi_index = dev->ifindex;
4964         hdr->ifi_flags = dev_get_flags(dev);
4965         hdr->ifi_change = 0;
4966
4967         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4968             (dev->addr_len &&
4969              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4970             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4971             (dev->ifindex != dev_get_iflink(dev) &&
4972              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
4973             nla_put_u8(skb, IFLA_OPERSTATE,
4974                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
4975                 goto nla_put_failure;
4976         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4977         if (!protoinfo)
4978                 goto nla_put_failure;
4979
4980         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4981                 goto nla_put_failure;
4982
4983         nla_nest_end(skb, protoinfo);
4984         nlmsg_end(skb, nlh);
4985         return 0;
4986
4987 nla_put_failure:
4988         nlmsg_cancel(skb, nlh);
4989         return -EMSGSIZE;
4990 }
4991
4992 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4993 {
4994         struct net *net = sock_net(skb->sk);
4995         int h, s_h;
4996         int idx = 0, s_idx;
4997         struct net_device *dev;
4998         struct inet6_dev *idev;
4999         struct hlist_head *head;
5000
5001         s_h = cb->args[0];
5002         s_idx = cb->args[1];
5003
5004         rcu_read_lock();
5005         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5006                 idx = 0;
5007                 head = &net->dev_index_head[h];
5008                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5009                         if (idx < s_idx)
5010                                 goto cont;
5011                         idev = __in6_dev_get(dev);
5012                         if (!idev)
5013                                 goto cont;
5014                         if (inet6_fill_ifinfo(skb, idev,
5015                                               NETLINK_CB(cb->skb).portid,
5016                                               cb->nlh->nlmsg_seq,
5017                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5018                                 goto out;
5019 cont:
5020                         idx++;
5021                 }
5022         }
5023 out:
5024         rcu_read_unlock();
5025         cb->args[1] = idx;
5026         cb->args[0] = h;
5027
5028         return skb->len;
5029 }
5030
5031 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5032 {
5033         struct sk_buff *skb;
5034         struct net *net = dev_net(idev->dev);
5035         int err = -ENOBUFS;
5036
5037         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5038         if (!skb)
5039                 goto errout;
5040
5041         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5042         if (err < 0) {
5043                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5044                 WARN_ON(err == -EMSGSIZE);
5045                 kfree_skb(skb);
5046                 goto errout;
5047         }
5048         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5049         return;
5050 errout:
5051         if (err < 0)
5052                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5053 }
5054
5055 static inline size_t inet6_prefix_nlmsg_size(void)
5056 {
5057         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5058                + nla_total_size(sizeof(struct in6_addr))
5059                + nla_total_size(sizeof(struct prefix_cacheinfo));
5060 }
5061
5062 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5063                              struct prefix_info *pinfo, u32 portid, u32 seq,
5064                              int event, unsigned int flags)
5065 {
5066         struct prefixmsg *pmsg;
5067         struct nlmsghdr *nlh;
5068         struct prefix_cacheinfo ci;
5069
5070         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5071         if (!nlh)
5072                 return -EMSGSIZE;
5073
5074         pmsg = nlmsg_data(nlh);
5075         pmsg->prefix_family = AF_INET6;
5076         pmsg->prefix_pad1 = 0;
5077         pmsg->prefix_pad2 = 0;
5078         pmsg->prefix_ifindex = idev->dev->ifindex;
5079         pmsg->prefix_len = pinfo->prefix_len;
5080         pmsg->prefix_type = pinfo->type;
5081         pmsg->prefix_pad3 = 0;
5082         pmsg->prefix_flags = 0;
5083         if (pinfo->onlink)
5084                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5085         if (pinfo->autoconf)
5086                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5087
5088         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5089                 goto nla_put_failure;
5090         ci.preferred_time = ntohl(pinfo->prefered);
5091         ci.valid_time = ntohl(pinfo->valid);
5092         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5093                 goto nla_put_failure;
5094         nlmsg_end(skb, nlh);
5095         return 0;
5096
5097 nla_put_failure:
5098         nlmsg_cancel(skb, nlh);
5099         return -EMSGSIZE;
5100 }
5101
5102 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5103                          struct prefix_info *pinfo)
5104 {
5105         struct sk_buff *skb;
5106         struct net *net = dev_net(idev->dev);
5107         int err = -ENOBUFS;
5108
5109         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5110         if (!skb)
5111                 goto errout;
5112
5113         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5114         if (err < 0) {
5115                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5116                 WARN_ON(err == -EMSGSIZE);
5117                 kfree_skb(skb);
5118                 goto errout;
5119         }
5120         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5121         return;
5122 errout:
5123         if (err < 0)
5124                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5125 }
5126
5127 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5128 {
5129         struct net *net = dev_net(ifp->idev->dev);
5130
5131         if (event)
5132                 ASSERT_RTNL();
5133
5134         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5135
5136         switch (event) {
5137         case RTM_NEWADDR:
5138                 /*
5139                  * If the address was optimistic
5140                  * we inserted the route at the start of
5141                  * our DAD process, so we don't need
5142                  * to do it again
5143                  */
5144                 if (!(ifp->rt->rt6i_node))
5145                         ip6_ins_rt(ifp->rt);
5146                 if (ifp->idev->cnf.forwarding)
5147                         addrconf_join_anycast(ifp);
5148                 if (!ipv6_addr_any(&ifp->peer_addr))
5149                         addrconf_prefix_route(&ifp->peer_addr, 128,
5150                                               ifp->idev->dev, 0, 0);
5151                 break;
5152         case RTM_DELADDR:
5153                 if (ifp->idev->cnf.forwarding)
5154                         addrconf_leave_anycast(ifp);
5155                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5156                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5157                         struct rt6_info *rt;
5158
5159                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5160                                                        ifp->idev->dev, 0, 0);
5161                         if (rt && ip6_del_rt(rt))
5162                                 dst_free(&rt->dst);
5163                 }
5164                 dst_hold(&ifp->rt->dst);
5165
5166                 if (ip6_del_rt(ifp->rt))
5167                         dst_free(&ifp->rt->dst);
5168
5169                 rt_genid_bump_ipv6(net);
5170                 break;
5171         }
5172         atomic_inc(&net->ipv6.dev_addr_genid);
5173 }
5174
5175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5176 {
5177         rcu_read_lock_bh();
5178         if (likely(ifp->idev->dead == 0))
5179                 __ipv6_ifa_notify(event, ifp);
5180         rcu_read_unlock_bh();
5181 }
5182
5183 #ifdef CONFIG_SYSCTL
5184
5185 static
5186 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5187                            void __user *buffer, size_t *lenp, loff_t *ppos)
5188 {
5189         int *valp = ctl->data;
5190         int val = *valp;
5191         loff_t pos = *ppos;
5192         struct ctl_table lctl;
5193         int ret;
5194
5195         /*
5196          * ctl->data points to idev->cnf.forwarding, we should
5197          * not modify it until we get the rtnl lock.
5198          */
5199         lctl = *ctl;
5200         lctl.data = &val;
5201
5202         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5203
5204         if (write)
5205                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5206         if (ret)
5207                 *ppos = pos;
5208         return ret;
5209 }
5210
5211 static
5212 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5213                         void __user *buffer, size_t *lenp, loff_t *ppos)
5214 {
5215         struct inet6_dev *idev = ctl->extra1;
5216         int min_mtu = IPV6_MIN_MTU;
5217         struct ctl_table lctl;
5218
5219         lctl = *ctl;
5220         lctl.extra1 = &min_mtu;
5221         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5222
5223         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5224 }
5225
5226 static void dev_disable_change(struct inet6_dev *idev)
5227 {
5228         struct netdev_notifier_info info;
5229
5230         if (!idev || !idev->dev)
5231                 return;
5232
5233         netdev_notifier_info_init(&info, idev->dev);
5234         if (idev->cnf.disable_ipv6)
5235                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5236         else
5237                 addrconf_notify(NULL, NETDEV_UP, &info);
5238 }
5239
5240 static void addrconf_disable_change(struct net *net, __s32 newf)
5241 {
5242         struct net_device *dev;
5243         struct inet6_dev *idev;
5244
5245         rcu_read_lock();
5246         for_each_netdev_rcu(net, dev) {
5247                 idev = __in6_dev_get(dev);
5248                 if (idev) {
5249                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5250                         idev->cnf.disable_ipv6 = newf;
5251                         if (changed)
5252                                 dev_disable_change(idev);
5253                 }
5254         }
5255         rcu_read_unlock();
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         if (!write) {
5376                 err = snprintf(str, sizeof(str), "%pI6",
5377                                &secret->secret);
5378                 if (err >= sizeof(str)) {
5379                         err = -EIO;
5380                         goto out;
5381                 }
5382         }
5383
5384         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5385         if (err || !write)
5386                 goto out;
5387
5388         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5389                 err = -EIO;
5390                 goto out;
5391         }
5392
5393         secret->initialized = true;
5394         secret->secret = addr;
5395
5396         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5397                 struct net_device *dev;
5398
5399                 for_each_netdev(net, dev) {
5400                         struct inet6_dev *idev = __in6_dev_get(dev);
5401
5402                         if (idev) {
5403                                 idev->addr_gen_mode =
5404                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5405                         }
5406                 }
5407         } else {
5408                 struct inet6_dev *idev = ctl->extra1;
5409
5410                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5411         }
5412
5413 out:
5414         rtnl_unlock();
5415
5416         return err;
5417 }
5418
5419 static
5420 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5421                                                 int write,
5422                                                 void __user *buffer,
5423                                                 size_t *lenp,
5424                                                 loff_t *ppos)
5425 {
5426         int *valp = ctl->data;
5427         int val = *valp;
5428         loff_t pos = *ppos;
5429         struct ctl_table lctl;
5430         int ret;
5431
5432         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5433          * we should not modify it until we get the rtnl lock.
5434          */
5435         lctl = *ctl;
5436         lctl.data = &val;
5437
5438         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5439
5440         if (write)
5441                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5442         if (ret)
5443                 *ppos = pos;
5444         return ret;
5445 }
5446
5447 static struct addrconf_sysctl_table
5448 {
5449         struct ctl_table_header *sysctl_header;
5450         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5451 } addrconf_sysctl __read_mostly = {
5452         .sysctl_header = NULL,
5453         .addrconf_vars = {
5454                 {
5455                         .procname       = "forwarding",
5456                         .data           = &ipv6_devconf.forwarding,
5457                         .maxlen         = sizeof(int),
5458                         .mode           = 0644,
5459                         .proc_handler   = addrconf_sysctl_forward,
5460                 },
5461                 {
5462                         .procname       = "hop_limit",
5463                         .data           = &ipv6_devconf.hop_limit,
5464                         .maxlen         = sizeof(int),
5465                         .mode           = 0644,
5466                         .proc_handler   = proc_dointvec,
5467                 },
5468                 {
5469                         .procname       = "mtu",
5470                         .data           = &ipv6_devconf.mtu6,
5471                         .maxlen         = sizeof(int),
5472                         .mode           = 0644,
5473                         .proc_handler   = addrconf_sysctl_mtu,
5474                 },
5475                 {
5476                         .procname       = "accept_ra",
5477                         .data           = &ipv6_devconf.accept_ra,
5478                         .maxlen         = sizeof(int),
5479                         .mode           = 0644,
5480                         .proc_handler   = proc_dointvec,
5481                 },
5482                 {
5483                         .procname       = "accept_redirects",
5484                         .data           = &ipv6_devconf.accept_redirects,
5485                         .maxlen         = sizeof(int),
5486                         .mode           = 0644,
5487                         .proc_handler   = proc_dointvec,
5488                 },
5489                 {
5490                         .procname       = "autoconf",
5491                         .data           = &ipv6_devconf.autoconf,
5492                         .maxlen         = sizeof(int),
5493                         .mode           = 0644,
5494                         .proc_handler   = proc_dointvec,
5495                 },
5496                 {
5497                         .procname       = "dad_transmits",
5498                         .data           = &ipv6_devconf.dad_transmits,
5499                         .maxlen         = sizeof(int),
5500                         .mode           = 0644,
5501                         .proc_handler   = proc_dointvec,
5502                 },
5503                 {
5504                         .procname       = "router_solicitations",
5505                         .data           = &ipv6_devconf.rtr_solicits,
5506                         .maxlen         = sizeof(int),
5507                         .mode           = 0644,
5508                         .proc_handler   = proc_dointvec,
5509                 },
5510                 {
5511                         .procname       = "router_solicitation_interval",
5512                         .data           = &ipv6_devconf.rtr_solicit_interval,
5513                         .maxlen         = sizeof(int),
5514                         .mode           = 0644,
5515                         .proc_handler   = proc_dointvec_jiffies,
5516                 },
5517                 {
5518                         .procname       = "router_solicitation_delay",
5519                         .data           = &ipv6_devconf.rtr_solicit_delay,
5520                         .maxlen         = sizeof(int),
5521                         .mode           = 0644,
5522                         .proc_handler   = proc_dointvec_jiffies,
5523                 },
5524                 {
5525                         .procname       = "force_mld_version",
5526                         .data           = &ipv6_devconf.force_mld_version,
5527                         .maxlen         = sizeof(int),
5528                         .mode           = 0644,
5529                         .proc_handler   = proc_dointvec,
5530                 },
5531                 {
5532                         .procname       = "mldv1_unsolicited_report_interval",
5533                         .data           =
5534                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5535                         .maxlen         = sizeof(int),
5536                         .mode           = 0644,
5537                         .proc_handler   = proc_dointvec_ms_jiffies,
5538                 },
5539                 {
5540                         .procname       = "mldv2_unsolicited_report_interval",
5541                         .data           =
5542                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5543                         .maxlen         = sizeof(int),
5544                         .mode           = 0644,
5545                         .proc_handler   = proc_dointvec_ms_jiffies,
5546                 },
5547                 {
5548                         .procname       = "use_tempaddr",
5549                         .data           = &ipv6_devconf.use_tempaddr,
5550                         .maxlen         = sizeof(int),
5551                         .mode           = 0644,
5552                         .proc_handler   = proc_dointvec,
5553                 },
5554                 {
5555                         .procname       = "temp_valid_lft",
5556                         .data           = &ipv6_devconf.temp_valid_lft,
5557                         .maxlen         = sizeof(int),
5558                         .mode           = 0644,
5559                         .proc_handler   = proc_dointvec,
5560                 },
5561                 {
5562                         .procname       = "temp_prefered_lft",
5563                         .data           = &ipv6_devconf.temp_prefered_lft,
5564                         .maxlen         = sizeof(int),
5565                         .mode           = 0644,
5566                         .proc_handler   = proc_dointvec,
5567                 },
5568                 {
5569                         .procname       = "regen_max_retry",
5570                         .data           = &ipv6_devconf.regen_max_retry,
5571                         .maxlen         = sizeof(int),
5572                         .mode           = 0644,
5573                         .proc_handler   = proc_dointvec,
5574                 },
5575                 {
5576                         .procname       = "max_desync_factor",
5577                         .data           = &ipv6_devconf.max_desync_factor,
5578                         .maxlen         = sizeof(int),
5579                         .mode           = 0644,
5580                         .proc_handler   = proc_dointvec,
5581                 },
5582                 {
5583                         .procname       = "max_addresses",
5584                         .data           = &ipv6_devconf.max_addresses,
5585                         .maxlen         = sizeof(int),
5586                         .mode           = 0644,
5587                         .proc_handler   = proc_dointvec,
5588                 },
5589                 {
5590                         .procname       = "accept_ra_defrtr",
5591                         .data           = &ipv6_devconf.accept_ra_defrtr,
5592                         .maxlen         = sizeof(int),
5593                         .mode           = 0644,
5594                         .proc_handler   = proc_dointvec,
5595                 },
5596                 {
5597                         .procname       = "accept_ra_min_hop_limit",
5598                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5599                         .maxlen         = sizeof(int),
5600                         .mode           = 0644,
5601                         .proc_handler   = proc_dointvec,
5602                 },
5603                 {
5604                         .procname       = "accept_ra_pinfo",
5605                         .data           = &ipv6_devconf.accept_ra_pinfo,
5606                         .maxlen         = sizeof(int),
5607                         .mode           = 0644,
5608                         .proc_handler   = proc_dointvec,
5609                 },
5610 #ifdef CONFIG_IPV6_ROUTER_PREF
5611                 {
5612                         .procname       = "accept_ra_rtr_pref",
5613                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5614                         .maxlen         = sizeof(int),
5615                         .mode           = 0644,
5616                         .proc_handler   = proc_dointvec,
5617                 },
5618                 {
5619                         .procname       = "router_probe_interval",
5620                         .data           = &ipv6_devconf.rtr_probe_interval,
5621                         .maxlen         = sizeof(int),
5622                         .mode           = 0644,
5623                         .proc_handler   = proc_dointvec_jiffies,
5624                 },
5625 #ifdef CONFIG_IPV6_ROUTE_INFO
5626                 {
5627                         .procname       = "accept_ra_rt_info_max_plen",
5628                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5629                         .maxlen         = sizeof(int),
5630                         .mode           = 0644,
5631                         .proc_handler   = proc_dointvec,
5632                 },
5633 #endif
5634 #endif
5635                 {
5636                         .procname       = "proxy_ndp",
5637                         .data           = &ipv6_devconf.proxy_ndp,
5638                         .maxlen         = sizeof(int),
5639                         .mode           = 0644,
5640                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5641                 },
5642                 {
5643                         .procname       = "accept_source_route",
5644                         .data           = &ipv6_devconf.accept_source_route,
5645                         .maxlen         = sizeof(int),
5646                         .mode           = 0644,
5647                         .proc_handler   = proc_dointvec,
5648                 },
5649 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5650                 {
5651                         .procname       = "optimistic_dad",
5652                         .data           = &ipv6_devconf.optimistic_dad,
5653                         .maxlen         = sizeof(int),
5654                         .mode           = 0644,
5655                         .proc_handler   = proc_dointvec,
5656
5657                 },
5658                 {
5659                         .procname       = "use_optimistic",
5660                         .data           = &ipv6_devconf.use_optimistic,
5661                         .maxlen         = sizeof(int),
5662                         .mode           = 0644,
5663                         .proc_handler   = proc_dointvec,
5664
5665                 },
5666 #endif
5667 #ifdef CONFIG_IPV6_MROUTE
5668                 {
5669                         .procname       = "mc_forwarding",
5670                         .data           = &ipv6_devconf.mc_forwarding,
5671                         .maxlen         = sizeof(int),
5672                         .mode           = 0444,
5673                         .proc_handler   = proc_dointvec,
5674                 },
5675 #endif
5676                 {
5677                         .procname       = "disable_ipv6",
5678                         .data           = &ipv6_devconf.disable_ipv6,
5679                         .maxlen         = sizeof(int),
5680                         .mode           = 0644,
5681                         .proc_handler   = addrconf_sysctl_disable,
5682                 },
5683                 {
5684                         .procname       = "accept_dad",
5685                         .data           = &ipv6_devconf.accept_dad,
5686                         .maxlen         = sizeof(int),
5687                         .mode           = 0644,
5688                         .proc_handler   = proc_dointvec,
5689                 },
5690                 {
5691                         .procname       = "force_tllao",
5692                         .data           = &ipv6_devconf.force_tllao,
5693                         .maxlen         = sizeof(int),
5694                         .mode           = 0644,
5695                         .proc_handler   = proc_dointvec
5696                 },
5697                 {
5698                         .procname       = "ndisc_notify",
5699                         .data           = &ipv6_devconf.ndisc_notify,
5700                         .maxlen         = sizeof(int),
5701                         .mode           = 0644,
5702                         .proc_handler   = proc_dointvec
5703                 },
5704                 {
5705                         .procname       = "suppress_frag_ndisc",
5706                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5707                         .maxlen         = sizeof(int),
5708                         .mode           = 0644,
5709                         .proc_handler   = proc_dointvec
5710                 },
5711                 {
5712                         .procname       = "accept_ra_from_local",
5713                         .data           = &ipv6_devconf.accept_ra_from_local,
5714                         .maxlen         = sizeof(int),
5715                         .mode           = 0644,
5716                         .proc_handler   = proc_dointvec,
5717                 },
5718                 {
5719                         .procname       = "accept_ra_mtu",
5720                         .data           = &ipv6_devconf.accept_ra_mtu,
5721                         .maxlen         = sizeof(int),
5722                         .mode           = 0644,
5723                         .proc_handler   = proc_dointvec,
5724                 },
5725                 {
5726                         .procname       = "stable_secret",
5727                         .data           = &ipv6_devconf.stable_secret,
5728                         .maxlen         = IPV6_MAX_STRLEN,
5729                         .mode           = 0600,
5730                         .proc_handler   = addrconf_sysctl_stable_secret,
5731                 },
5732                 {
5733                         .procname       = "use_oif_addrs_only",
5734                         .data           = &ipv6_devconf.use_oif_addrs_only,
5735                         .maxlen         = sizeof(int),
5736                         .mode           = 0644,
5737                         .proc_handler   = proc_dointvec,
5738                 },
5739                 {
5740                         .procname       = "ignore_routes_with_linkdown",
5741                         .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5742                         .maxlen         = sizeof(int),
5743                         .mode           = 0644,
5744                         .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5745                 },
5746                 {
5747                         /* sentinel */
5748                 }
5749         },
5750 };
5751
5752 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5753                 struct inet6_dev *idev, struct ipv6_devconf *p)
5754 {
5755         int i;
5756         struct addrconf_sysctl_table *t;
5757         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5758
5759         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5760         if (!t)
5761                 goto out;
5762
5763         for (i = 0; t->addrconf_vars[i].data; i++) {
5764                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5765                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5766                 t->addrconf_vars[i].extra2 = net;
5767         }
5768
5769         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5770
5771         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5772         if (!t->sysctl_header)
5773                 goto free;
5774
5775         p->sysctl = t;
5776         return 0;
5777
5778 free:
5779         kfree(t);
5780 out:
5781         return -ENOBUFS;
5782 }
5783
5784 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5785 {
5786         struct addrconf_sysctl_table *t;
5787
5788         if (!p->sysctl)
5789                 return;
5790
5791         t = p->sysctl;
5792         p->sysctl = NULL;
5793         unregister_net_sysctl_table(t->sysctl_header);
5794         kfree(t);
5795 }
5796
5797 static int addrconf_sysctl_register(struct inet6_dev *idev)
5798 {
5799         int err;
5800
5801         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5802                 return -EINVAL;
5803
5804         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5805                                     &ndisc_ifinfo_sysctl_change);
5806         if (err)
5807                 return err;
5808         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5809                                          idev, &idev->cnf);
5810         if (err)
5811                 neigh_sysctl_unregister(idev->nd_parms);
5812
5813         return err;
5814 }
5815
5816 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5817 {
5818         __addrconf_sysctl_unregister(&idev->cnf);
5819         neigh_sysctl_unregister(idev->nd_parms);
5820 }
5821
5822
5823 #endif
5824
5825 static int __net_init addrconf_init_net(struct net *net)
5826 {
5827         int err = -ENOMEM;
5828         struct ipv6_devconf *all, *dflt;
5829
5830         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5831         if (!all)
5832                 goto err_alloc_all;
5833
5834         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5835         if (!dflt)
5836                 goto err_alloc_dflt;
5837
5838         /* these will be inherited by all namespaces */
5839         dflt->autoconf = ipv6_defaults.autoconf;
5840         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5841
5842         dflt->stable_secret.initialized = false;
5843         all->stable_secret.initialized = false;
5844
5845         net->ipv6.devconf_all = all;
5846         net->ipv6.devconf_dflt = dflt;
5847
5848 #ifdef CONFIG_SYSCTL
5849         err = __addrconf_sysctl_register(net, "all", NULL, all);
5850         if (err < 0)
5851                 goto err_reg_all;
5852
5853         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5854         if (err < 0)
5855                 goto err_reg_dflt;
5856 #endif
5857         return 0;
5858
5859 #ifdef CONFIG_SYSCTL
5860 err_reg_dflt:
5861         __addrconf_sysctl_unregister(all);
5862 err_reg_all:
5863         kfree(dflt);
5864 #endif
5865 err_alloc_dflt:
5866         kfree(all);
5867 err_alloc_all:
5868         return err;
5869 }
5870
5871 static void __net_exit addrconf_exit_net(struct net *net)
5872 {
5873 #ifdef CONFIG_SYSCTL
5874         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5875         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5876 #endif
5877         kfree(net->ipv6.devconf_dflt);
5878         kfree(net->ipv6.devconf_all);
5879 }
5880
5881 static struct pernet_operations addrconf_ops = {
5882         .init = addrconf_init_net,
5883         .exit = addrconf_exit_net,
5884 };
5885
5886 static struct rtnl_af_ops inet6_ops __read_mostly = {
5887         .family           = AF_INET6,
5888         .fill_link_af     = inet6_fill_link_af,
5889         .get_link_af_size = inet6_get_link_af_size,
5890         .validate_link_af = inet6_validate_link_af,
5891         .set_link_af      = inet6_set_link_af,
5892 };
5893
5894 /*
5895  *      Init / cleanup code
5896  */
5897
5898 int __init addrconf_init(void)
5899 {
5900         struct inet6_dev *idev;
5901         int i, err;
5902
5903         err = ipv6_addr_label_init();
5904         if (err < 0) {
5905                 pr_crit("%s: cannot initialize default policy table: %d\n",
5906                         __func__, err);
5907                 goto out;
5908         }
5909
5910         err = register_pernet_subsys(&addrconf_ops);
5911         if (err < 0)
5912                 goto out_addrlabel;
5913
5914         addrconf_wq = create_workqueue("ipv6_addrconf");
5915         if (!addrconf_wq) {
5916                 err = -ENOMEM;
5917                 goto out_nowq;
5918         }
5919
5920         /* The addrconf netdev notifier requires that loopback_dev
5921          * has it's ipv6 private information allocated and setup
5922          * before it can bring up and give link-local addresses
5923          * to other devices which are up.
5924          *
5925          * Unfortunately, loopback_dev is not necessarily the first
5926          * entry in the global dev_base list of net devices.  In fact,
5927          * it is likely to be the very last entry on that list.
5928          * So this causes the notifier registry below to try and
5929          * give link-local addresses to all devices besides loopback_dev
5930          * first, then loopback_dev, which cases all the non-loopback_dev
5931          * devices to fail to get a link-local address.
5932          *
5933          * So, as a temporary fix, allocate the ipv6 structure for
5934          * loopback_dev first by hand.
5935          * Longer term, all of the dependencies ipv6 has upon the loopback
5936          * device and it being up should be removed.
5937          */
5938         rtnl_lock();
5939         idev = ipv6_add_dev(init_net.loopback_dev);
5940         rtnl_unlock();
5941         if (IS_ERR(idev)) {
5942                 err = PTR_ERR(idev);
5943                 goto errlo;
5944         }
5945
5946         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5947                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5948
5949         register_netdevice_notifier(&ipv6_dev_notf);
5950
5951         addrconf_verify();
5952
5953         rtnl_af_register(&inet6_ops);
5954
5955         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5956                               NULL);
5957         if (err < 0)
5958                 goto errout;
5959
5960         /* Only the first call to __rtnl_register can fail */
5961         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5962         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5963         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5964                         inet6_dump_ifaddr, NULL);
5965         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5966                         inet6_dump_ifmcaddr, NULL);
5967         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5968                         inet6_dump_ifacaddr, NULL);
5969         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5970                         inet6_netconf_dump_devconf, NULL);
5971
5972         ipv6_addr_label_rtnl_register();
5973
5974         return 0;
5975 errout:
5976         rtnl_af_unregister(&inet6_ops);
5977         unregister_netdevice_notifier(&ipv6_dev_notf);
5978 errlo:
5979         destroy_workqueue(addrconf_wq);
5980 out_nowq:
5981         unregister_pernet_subsys(&addrconf_ops);
5982 out_addrlabel:
5983         ipv6_addr_label_cleanup();
5984 out:
5985         return err;
5986 }
5987
5988 void addrconf_cleanup(void)
5989 {
5990         struct net_device *dev;
5991         int i;
5992
5993         unregister_netdevice_notifier(&ipv6_dev_notf);
5994         unregister_pernet_subsys(&addrconf_ops);
5995         ipv6_addr_label_cleanup();
5996
5997         rtnl_lock();
5998
5999         __rtnl_af_unregister(&inet6_ops);
6000
6001         /* clean dev list */
6002         for_each_netdev(&init_net, dev) {
6003                 if (__in6_dev_get(dev) == NULL)
6004                         continue;
6005                 addrconf_ifdown(dev, 1);
6006         }
6007         addrconf_ifdown(init_net.loopback_dev, 2);
6008
6009         /*
6010          *      Check hash table.
6011          */
6012         spin_lock_bh(&addrconf_hash_lock);
6013         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6014                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6015         spin_unlock_bh(&addrconf_hash_lock);
6016         cancel_delayed_work(&addr_chk_work);
6017         rtnl_unlock();
6018
6019         destroy_workqueue(addrconf_wq);
6020 }