PM / OPP: Don't support OPP if it provides supported-hw but platform does not
[cascardo/linux.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
68
69 #include <net/net_namespace.h>
70 #include <net/sock.h>
71 #include <net/snmp.h>
72
73 #include <net/6lowpan.h>
74 #include <net/firewire.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/tcp.h>
81 #include <net/ip.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
91
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
95
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
98
99 #if ACONF_DEBUG >= 3
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101 #else
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #endif
104
105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
106
107 #define IPV6_MAX_STRLEN \
108         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
109
110 static inline u32 cstamp_delta(unsigned long cstamp)
111 {
112         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
113 }
114
115 #ifdef CONFIG_SYSCTL
116 static int addrconf_sysctl_register(struct inet6_dev *idev);
117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118 #else
119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
120 {
121         return 0;
122 }
123
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
125 {
126 }
127 #endif
128
129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131 static void ipv6_regen_rndid(unsigned long data);
132
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135 static int ipv6_generate_stable_address(struct in6_addr *addr,
136                                         u8 dad_count,
137                                         const struct inet6_dev *idev);
138
139 /*
140  *      Configured unicast address hash table
141  */
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
144
145 static void addrconf_verify(void);
146 static void addrconf_verify_rtnl(void);
147 static void addrconf_verify_work(struct work_struct *);
148
149 static struct workqueue_struct *addrconf_wq;
150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
151
152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
154
155 static void addrconf_type_change(struct net_device *dev,
156                                  unsigned long event);
157 static int addrconf_ifdown(struct net_device *dev, int how);
158
159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
160                                                   int plen,
161                                                   const struct net_device *dev,
162                                                   u32 flags, u32 noflags);
163
164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165 static void addrconf_dad_work(struct work_struct *w);
166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167 static void addrconf_dad_run(struct inet6_dev *idev);
168 static void addrconf_rs_timer(unsigned long data);
169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
171
172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173                                 struct prefix_info *pinfo);
174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175                                struct net_device *dev);
176
177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
178         .forwarding             = 0,
179         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
180         .mtu6                   = IPV6_MIN_MTU,
181         .accept_ra              = 1,
182         .accept_redirects       = 1,
183         .autoconf               = 1,
184         .force_mld_version      = 0,
185         .mldv1_unsolicited_report_interval = 10 * HZ,
186         .mldv2_unsolicited_report_interval = HZ,
187         .dad_transmits          = 1,
188         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
189         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
190         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
191         .use_tempaddr           = 0,
192         .temp_valid_lft         = TEMP_VALID_LIFETIME,
193         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
194         .regen_max_retry        = REGEN_MAX_RETRY,
195         .max_desync_factor      = MAX_DESYNC_FACTOR,
196         .max_addresses          = IPV6_MAX_ADDRESSES,
197         .accept_ra_defrtr       = 1,
198         .accept_ra_from_local   = 0,
199         .accept_ra_min_hop_limit= 1,
200         .accept_ra_pinfo        = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202         .accept_ra_rtr_pref     = 1,
203         .rtr_probe_interval     = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205         .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208         .proxy_ndp              = 0,
209         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
210         .disable_ipv6           = 0,
211         .accept_dad             = 1,
212         .suppress_frag_ndisc    = 1,
213         .accept_ra_mtu          = 1,
214         .stable_secret          = {
215                 .initialized = false,
216         },
217         .use_oif_addrs_only     = 0,
218         .ignore_routes_with_linkdown = 0,
219         .keep_addr_on_down      = 0,
220 };
221
222 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
223         .forwarding             = 0,
224         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
225         .mtu6                   = IPV6_MIN_MTU,
226         .accept_ra              = 1,
227         .accept_redirects       = 1,
228         .autoconf               = 1,
229         .force_mld_version      = 0,
230         .mldv1_unsolicited_report_interval = 10 * HZ,
231         .mldv2_unsolicited_report_interval = HZ,
232         .dad_transmits          = 1,
233         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
234         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
235         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
236         .use_tempaddr           = 0,
237         .temp_valid_lft         = TEMP_VALID_LIFETIME,
238         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
239         .regen_max_retry        = REGEN_MAX_RETRY,
240         .max_desync_factor      = MAX_DESYNC_FACTOR,
241         .max_addresses          = IPV6_MAX_ADDRESSES,
242         .accept_ra_defrtr       = 1,
243         .accept_ra_from_local   = 0,
244         .accept_ra_min_hop_limit= 1,
245         .accept_ra_pinfo        = 1,
246 #ifdef CONFIG_IPV6_ROUTER_PREF
247         .accept_ra_rtr_pref     = 1,
248         .rtr_probe_interval     = 60 * HZ,
249 #ifdef CONFIG_IPV6_ROUTE_INFO
250         .accept_ra_rt_info_max_plen = 0,
251 #endif
252 #endif
253         .proxy_ndp              = 0,
254         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
255         .disable_ipv6           = 0,
256         .accept_dad             = 1,
257         .suppress_frag_ndisc    = 1,
258         .accept_ra_mtu          = 1,
259         .stable_secret          = {
260                 .initialized = false,
261         },
262         .use_oif_addrs_only     = 0,
263         .ignore_routes_with_linkdown = 0,
264         .keep_addr_on_down      = 0,
265 };
266
267 /* Check if a valid qdisc is available */
268 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
269 {
270         return !qdisc_tx_is_noop(dev);
271 }
272
273 static void addrconf_del_rs_timer(struct inet6_dev *idev)
274 {
275         if (del_timer(&idev->rs_timer))
276                 __in6_dev_put(idev);
277 }
278
279 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
280 {
281         if (cancel_delayed_work(&ifp->dad_work))
282                 __in6_ifa_put(ifp);
283 }
284
285 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
286                                   unsigned long when)
287 {
288         if (!timer_pending(&idev->rs_timer))
289                 in6_dev_hold(idev);
290         mod_timer(&idev->rs_timer, jiffies + when);
291 }
292
293 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
294                                    unsigned long delay)
295 {
296         if (!delayed_work_pending(&ifp->dad_work))
297                 in6_ifa_hold(ifp);
298         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
299 }
300
301 static int snmp6_alloc_dev(struct inet6_dev *idev)
302 {
303         int i;
304
305         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
306         if (!idev->stats.ipv6)
307                 goto err_ip;
308
309         for_each_possible_cpu(i) {
310                 struct ipstats_mib *addrconf_stats;
311                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
312                 u64_stats_init(&addrconf_stats->syncp);
313         }
314
315
316         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
317                                         GFP_KERNEL);
318         if (!idev->stats.icmpv6dev)
319                 goto err_icmp;
320         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
321                                            GFP_KERNEL);
322         if (!idev->stats.icmpv6msgdev)
323                 goto err_icmpmsg;
324
325         return 0;
326
327 err_icmpmsg:
328         kfree(idev->stats.icmpv6dev);
329 err_icmp:
330         free_percpu(idev->stats.ipv6);
331 err_ip:
332         return -ENOMEM;
333 }
334
335 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
336 {
337         struct inet6_dev *ndev;
338         int err = -ENOMEM;
339
340         ASSERT_RTNL();
341
342         if (dev->mtu < IPV6_MIN_MTU)
343                 return ERR_PTR(-EINVAL);
344
345         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
346         if (!ndev)
347                 return ERR_PTR(err);
348
349         rwlock_init(&ndev->lock);
350         ndev->dev = dev;
351         INIT_LIST_HEAD(&ndev->addr_list);
352         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
353                     (unsigned long)ndev);
354         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
355
356         if (ndev->cnf.stable_secret.initialized)
357                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
358         else
359                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
360
361         ndev->cnf.mtu6 = dev->mtu;
362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
363         if (!ndev->nd_parms) {
364                 kfree(ndev);
365                 return ERR_PTR(err);
366         }
367         if (ndev->cnf.forwarding)
368                 dev_disable_lro(dev);
369         /* We refer to the device */
370         dev_hold(dev);
371
372         if (snmp6_alloc_dev(ndev) < 0) {
373                 ADBG(KERN_WARNING
374                         "%s: cannot allocate memory for statistics; dev=%s.\n",
375                         __func__, dev->name);
376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
377                 dev_put(dev);
378                 kfree(ndev);
379                 return ERR_PTR(err);
380         }
381
382         if (snmp6_register_dev(ndev) < 0) {
383                 ADBG(KERN_WARNING
384                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
385                         __func__, dev->name);
386                 goto err_release;
387         }
388
389         /* One reference from device.  We must do this before
390          * we invoke __ipv6_regen_rndid().
391          */
392         in6_dev_hold(ndev);
393
394         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
395                 ndev->cnf.accept_dad = -1;
396
397 #if IS_ENABLED(CONFIG_IPV6_SIT)
398         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
399                 pr_info("%s: Disabled Multicast RS\n", dev->name);
400                 ndev->cnf.rtr_solicits = 0;
401         }
402 #endif
403
404         INIT_LIST_HEAD(&ndev->tempaddr_list);
405         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
406         if ((dev->flags&IFF_LOOPBACK) ||
407             dev->type == ARPHRD_TUNNEL ||
408             dev->type == ARPHRD_TUNNEL6 ||
409             dev->type == ARPHRD_SIT ||
410             dev->type == ARPHRD_NONE) {
411                 ndev->cnf.use_tempaddr = -1;
412         } else {
413                 in6_dev_hold(ndev);
414                 ipv6_regen_rndid((unsigned long) ndev);
415         }
416
417         ndev->token = in6addr_any;
418
419         if (netif_running(dev) && addrconf_qdisc_ok(dev))
420                 ndev->if_flags |= IF_READY;
421
422         ipv6_mc_init_dev(ndev);
423         ndev->tstamp = jiffies;
424         err = addrconf_sysctl_register(ndev);
425         if (err) {
426                 ipv6_mc_destroy_dev(ndev);
427                 del_timer(&ndev->regen_timer);
428                 snmp6_unregister_dev(ndev);
429                 goto err_release;
430         }
431         /* protected by rtnl_lock */
432         rcu_assign_pointer(dev->ip6_ptr, ndev);
433
434         /* Join interface-local all-node multicast group */
435         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
436
437         /* Join all-node multicast group */
438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439
440         /* Join all-router multicast group if forwarding is set */
441         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
442                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
443
444         return ndev;
445
446 err_release:
447         neigh_parms_release(&nd_tbl, ndev->nd_parms);
448         ndev->dead = 1;
449         in6_dev_finish_destroy(ndev);
450         return ERR_PTR(err);
451 }
452
453 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
454 {
455         struct inet6_dev *idev;
456
457         ASSERT_RTNL();
458
459         idev = __in6_dev_get(dev);
460         if (!idev) {
461                 idev = ipv6_add_dev(dev);
462                 if (IS_ERR(idev))
463                         return NULL;
464         }
465
466         if (dev->flags&IFF_UP)
467                 ipv6_mc_up(idev);
468         return idev;
469 }
470
471 static int inet6_netconf_msgsize_devconf(int type)
472 {
473         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
474                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
475         bool all = false;
476
477         if (type == NETCONFA_ALL)
478                 all = true;
479
480         if (all || type == NETCONFA_FORWARDING)
481                 size += nla_total_size(4);
482 #ifdef CONFIG_IPV6_MROUTE
483         if (all || type == NETCONFA_MC_FORWARDING)
484                 size += nla_total_size(4);
485 #endif
486         if (all || type == NETCONFA_PROXY_NEIGH)
487                 size += nla_total_size(4);
488
489         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
490                 size += nla_total_size(4);
491
492         return size;
493 }
494
495 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
496                                       struct ipv6_devconf *devconf, u32 portid,
497                                       u32 seq, int event, unsigned int flags,
498                                       int type)
499 {
500         struct nlmsghdr  *nlh;
501         struct netconfmsg *ncm;
502         bool all = false;
503
504         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
505                         flags);
506         if (!nlh)
507                 return -EMSGSIZE;
508
509         if (type == NETCONFA_ALL)
510                 all = true;
511
512         ncm = nlmsg_data(nlh);
513         ncm->ncm_family = AF_INET6;
514
515         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
516                 goto nla_put_failure;
517
518         if ((all || type == NETCONFA_FORWARDING) &&
519             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
520                 goto nla_put_failure;
521 #ifdef CONFIG_IPV6_MROUTE
522         if ((all || type == NETCONFA_MC_FORWARDING) &&
523             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
524                         devconf->mc_forwarding) < 0)
525                 goto nla_put_failure;
526 #endif
527         if ((all || type == NETCONFA_PROXY_NEIGH) &&
528             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
529                 goto nla_put_failure;
530
531         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
532             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
533                         devconf->ignore_routes_with_linkdown) < 0)
534                 goto nla_put_failure;
535
536         nlmsg_end(skb, nlh);
537         return 0;
538
539 nla_put_failure:
540         nlmsg_cancel(skb, nlh);
541         return -EMSGSIZE;
542 }
543
544 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
545                                   struct ipv6_devconf *devconf)
546 {
547         struct sk_buff *skb;
548         int err = -ENOBUFS;
549
550         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
551         if (!skb)
552                 goto errout;
553
554         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
555                                          RTM_NEWNETCONF, 0, type);
556         if (err < 0) {
557                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
558                 WARN_ON(err == -EMSGSIZE);
559                 kfree_skb(skb);
560                 goto errout;
561         }
562         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
563         return;
564 errout:
565         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
566 }
567
568 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
569         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
570         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
571         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
572         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
573 };
574
575 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
576                                      struct nlmsghdr *nlh)
577 {
578         struct net *net = sock_net(in_skb->sk);
579         struct nlattr *tb[NETCONFA_MAX+1];
580         struct netconfmsg *ncm;
581         struct sk_buff *skb;
582         struct ipv6_devconf *devconf;
583         struct inet6_dev *in6_dev;
584         struct net_device *dev;
585         int ifindex;
586         int err;
587
588         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
589                           devconf_ipv6_policy);
590         if (err < 0)
591                 goto errout;
592
593         err = -EINVAL;
594         if (!tb[NETCONFA_IFINDEX])
595                 goto errout;
596
597         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
598         switch (ifindex) {
599         case NETCONFA_IFINDEX_ALL:
600                 devconf = net->ipv6.devconf_all;
601                 break;
602         case NETCONFA_IFINDEX_DEFAULT:
603                 devconf = net->ipv6.devconf_dflt;
604                 break;
605         default:
606                 dev = __dev_get_by_index(net, ifindex);
607                 if (!dev)
608                         goto errout;
609                 in6_dev = __in6_dev_get(dev);
610                 if (!in6_dev)
611                         goto errout;
612                 devconf = &in6_dev->cnf;
613                 break;
614         }
615
616         err = -ENOBUFS;
617         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
618         if (!skb)
619                 goto errout;
620
621         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
622                                          NETLINK_CB(in_skb).portid,
623                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
624                                          NETCONFA_ALL);
625         if (err < 0) {
626                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
627                 WARN_ON(err == -EMSGSIZE);
628                 kfree_skb(skb);
629                 goto errout;
630         }
631         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
632 errout:
633         return err;
634 }
635
636 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
637                                       struct netlink_callback *cb)
638 {
639         struct net *net = sock_net(skb->sk);
640         int h, s_h;
641         int idx, s_idx;
642         struct net_device *dev;
643         struct inet6_dev *idev;
644         struct hlist_head *head;
645
646         s_h = cb->args[0];
647         s_idx = idx = cb->args[1];
648
649         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
650                 idx = 0;
651                 head = &net->dev_index_head[h];
652                 rcu_read_lock();
653                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
654                           net->dev_base_seq;
655                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
656                         if (idx < s_idx)
657                                 goto cont;
658                         idev = __in6_dev_get(dev);
659                         if (!idev)
660                                 goto cont;
661
662                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
663                                                        &idev->cnf,
664                                                        NETLINK_CB(cb->skb).portid,
665                                                        cb->nlh->nlmsg_seq,
666                                                        RTM_NEWNETCONF,
667                                                        NLM_F_MULTI,
668                                                        NETCONFA_ALL) < 0) {
669                                 rcu_read_unlock();
670                                 goto done;
671                         }
672                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
673 cont:
674                         idx++;
675                 }
676                 rcu_read_unlock();
677         }
678         if (h == NETDEV_HASHENTRIES) {
679                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
680                                                net->ipv6.devconf_all,
681                                                NETLINK_CB(cb->skb).portid,
682                                                cb->nlh->nlmsg_seq,
683                                                RTM_NEWNETCONF, NLM_F_MULTI,
684                                                NETCONFA_ALL) < 0)
685                         goto done;
686                 else
687                         h++;
688         }
689         if (h == NETDEV_HASHENTRIES + 1) {
690                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
691                                                net->ipv6.devconf_dflt,
692                                                NETLINK_CB(cb->skb).portid,
693                                                cb->nlh->nlmsg_seq,
694                                                RTM_NEWNETCONF, NLM_F_MULTI,
695                                                NETCONFA_ALL) < 0)
696                         goto done;
697                 else
698                         h++;
699         }
700 done:
701         cb->args[0] = h;
702         cb->args[1] = idx;
703
704         return skb->len;
705 }
706
707 #ifdef CONFIG_SYSCTL
708 static void dev_forward_change(struct inet6_dev *idev)
709 {
710         struct net_device *dev;
711         struct inet6_ifaddr *ifa;
712
713         if (!idev)
714                 return;
715         dev = idev->dev;
716         if (idev->cnf.forwarding)
717                 dev_disable_lro(dev);
718         if (dev->flags & IFF_MULTICAST) {
719                 if (idev->cnf.forwarding) {
720                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
721                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
722                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
723                 } else {
724                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
725                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
726                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
727                 }
728         }
729
730         list_for_each_entry(ifa, &idev->addr_list, if_list) {
731                 if (ifa->flags&IFA_F_TENTATIVE)
732                         continue;
733                 if (idev->cnf.forwarding)
734                         addrconf_join_anycast(ifa);
735                 else
736                         addrconf_leave_anycast(ifa);
737         }
738         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
739                                      dev->ifindex, &idev->cnf);
740 }
741
742
743 static void addrconf_forward_change(struct net *net, __s32 newf)
744 {
745         struct net_device *dev;
746         struct inet6_dev *idev;
747
748         for_each_netdev(net, dev) {
749                 idev = __in6_dev_get(dev);
750                 if (idev) {
751                         int changed = (!idev->cnf.forwarding) ^ (!newf);
752                         idev->cnf.forwarding = newf;
753                         if (changed)
754                                 dev_forward_change(idev);
755                 }
756         }
757 }
758
759 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
760 {
761         struct net *net;
762         int old;
763
764         if (!rtnl_trylock())
765                 return restart_syscall();
766
767         net = (struct net *)table->extra2;
768         old = *p;
769         *p = newf;
770
771         if (p == &net->ipv6.devconf_dflt->forwarding) {
772                 if ((!newf) ^ (!old))
773                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
774                                                      NETCONFA_IFINDEX_DEFAULT,
775                                                      net->ipv6.devconf_dflt);
776                 rtnl_unlock();
777                 return 0;
778         }
779
780         if (p == &net->ipv6.devconf_all->forwarding) {
781                 net->ipv6.devconf_dflt->forwarding = newf;
782                 addrconf_forward_change(net, newf);
783                 if ((!newf) ^ (!old))
784                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
785                                                      NETCONFA_IFINDEX_ALL,
786                                                      net->ipv6.devconf_all);
787         } else if ((!newf) ^ (!old))
788                 dev_forward_change((struct inet6_dev *)table->extra1);
789         rtnl_unlock();
790
791         if (newf)
792                 rt6_purge_dflt_routers(net);
793         return 1;
794 }
795
796 static void addrconf_linkdown_change(struct net *net, __s32 newf)
797 {
798         struct net_device *dev;
799         struct inet6_dev *idev;
800
801         for_each_netdev(net, dev) {
802                 idev = __in6_dev_get(dev);
803                 if (idev) {
804                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
805
806                         idev->cnf.ignore_routes_with_linkdown = newf;
807                         if (changed)
808                                 inet6_netconf_notify_devconf(dev_net(dev),
809                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
810                                                              dev->ifindex,
811                                                              &idev->cnf);
812                 }
813         }
814 }
815
816 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
817 {
818         struct net *net;
819         int old;
820
821         if (!rtnl_trylock())
822                 return restart_syscall();
823
824         net = (struct net *)table->extra2;
825         old = *p;
826         *p = newf;
827
828         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
829                 if ((!newf) ^ (!old))
830                         inet6_netconf_notify_devconf(net,
831                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
832                                                      NETCONFA_IFINDEX_DEFAULT,
833                                                      net->ipv6.devconf_dflt);
834                 rtnl_unlock();
835                 return 0;
836         }
837
838         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
839                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
840                 addrconf_linkdown_change(net, newf);
841                 if ((!newf) ^ (!old))
842                         inet6_netconf_notify_devconf(net,
843                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
844                                                      NETCONFA_IFINDEX_ALL,
845                                                      net->ipv6.devconf_all);
846         }
847         rtnl_unlock();
848
849         return 1;
850 }
851
852 #endif
853
854 /* Nobody refers to this ifaddr, destroy it */
855 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
856 {
857         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
858
859 #ifdef NET_REFCNT_DEBUG
860         pr_debug("%s\n", __func__);
861 #endif
862
863         in6_dev_put(ifp->idev);
864
865         if (cancel_delayed_work(&ifp->dad_work))
866                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
867                           ifp);
868
869         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
870                 pr_warn("Freeing alive inet6 address %p\n", ifp);
871                 return;
872         }
873         ip6_rt_put(ifp->rt);
874
875         kfree_rcu(ifp, rcu);
876 }
877
878 static void
879 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
880 {
881         struct list_head *p;
882         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
883
884         /*
885          * Each device address list is sorted in order of scope -
886          * global before linklocal.
887          */
888         list_for_each(p, &idev->addr_list) {
889                 struct inet6_ifaddr *ifa
890                         = list_entry(p, struct inet6_ifaddr, if_list);
891                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
892                         break;
893         }
894
895         list_add_tail(&ifp->if_list, p);
896 }
897
898 static u32 inet6_addr_hash(const struct in6_addr *addr)
899 {
900         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
901 }
902
903 /* On success it returns ifp with increased reference count */
904
905 static struct inet6_ifaddr *
906 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
907               const struct in6_addr *peer_addr, int pfxlen,
908               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
909 {
910         struct inet6_ifaddr *ifa = NULL;
911         struct rt6_info *rt;
912         unsigned int hash;
913         int err = 0;
914         int addr_type = ipv6_addr_type(addr);
915
916         if (addr_type == IPV6_ADDR_ANY ||
917             addr_type & IPV6_ADDR_MULTICAST ||
918             (!(idev->dev->flags & IFF_LOOPBACK) &&
919              addr_type & IPV6_ADDR_LOOPBACK))
920                 return ERR_PTR(-EADDRNOTAVAIL);
921
922         rcu_read_lock_bh();
923         if (idev->dead) {
924                 err = -ENODEV;                  /*XXX*/
925                 goto out2;
926         }
927
928         if (idev->cnf.disable_ipv6) {
929                 err = -EACCES;
930                 goto out2;
931         }
932
933         spin_lock(&addrconf_hash_lock);
934
935         /* Ignore adding duplicate addresses on an interface */
936         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
937                 ADBG("ipv6_add_addr: already assigned\n");
938                 err = -EEXIST;
939                 goto out;
940         }
941
942         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
943
944         if (!ifa) {
945                 ADBG("ipv6_add_addr: malloc failed\n");
946                 err = -ENOBUFS;
947                 goto out;
948         }
949
950         rt = addrconf_dst_alloc(idev, addr, false);
951         if (IS_ERR(rt)) {
952                 err = PTR_ERR(rt);
953                 goto out;
954         }
955
956         neigh_parms_data_state_setall(idev->nd_parms);
957
958         ifa->addr = *addr;
959         if (peer_addr)
960                 ifa->peer_addr = *peer_addr;
961
962         spin_lock_init(&ifa->lock);
963         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
964         INIT_HLIST_NODE(&ifa->addr_lst);
965         ifa->scope = scope;
966         ifa->prefix_len = pfxlen;
967         ifa->flags = flags | IFA_F_TENTATIVE;
968         ifa->valid_lft = valid_lft;
969         ifa->prefered_lft = prefered_lft;
970         ifa->cstamp = ifa->tstamp = jiffies;
971         ifa->tokenized = false;
972
973         ifa->rt = rt;
974
975         ifa->idev = idev;
976         in6_dev_hold(idev);
977         /* For caller */
978         in6_ifa_hold(ifa);
979
980         /* Add to big hash table */
981         hash = inet6_addr_hash(addr);
982
983         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
984         spin_unlock(&addrconf_hash_lock);
985
986         write_lock(&idev->lock);
987         /* Add to inet6_dev unicast addr list. */
988         ipv6_link_dev_addr(idev, ifa);
989
990         if (ifa->flags&IFA_F_TEMPORARY) {
991                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
992                 in6_ifa_hold(ifa);
993         }
994
995         in6_ifa_hold(ifa);
996         write_unlock(&idev->lock);
997 out2:
998         rcu_read_unlock_bh();
999
1000         if (likely(err == 0))
1001                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1002         else {
1003                 kfree(ifa);
1004                 ifa = ERR_PTR(err);
1005         }
1006
1007         return ifa;
1008 out:
1009         spin_unlock(&addrconf_hash_lock);
1010         goto out2;
1011 }
1012
1013 enum cleanup_prefix_rt_t {
1014         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1015         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1016         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1017 };
1018
1019 /*
1020  * Check, whether the prefix for ifp would still need a prefix route
1021  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1022  * constants.
1023  *
1024  * 1) we don't purge prefix if address was not permanent.
1025  *    prefix is managed by its own lifetime.
1026  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1027  * 3) if there are no addresses, delete prefix.
1028  * 4) if there are still other permanent address(es),
1029  *    corresponding prefix is still permanent.
1030  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1031  *    don't purge the prefix, assume user space is managing it.
1032  * 6) otherwise, update prefix lifetime to the
1033  *    longest valid lifetime among the corresponding
1034  *    addresses on the device.
1035  *    Note: subsequent RA will update lifetime.
1036  **/
1037 static enum cleanup_prefix_rt_t
1038 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1039 {
1040         struct inet6_ifaddr *ifa;
1041         struct inet6_dev *idev = ifp->idev;
1042         unsigned long lifetime;
1043         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1044
1045         *expires = jiffies;
1046
1047         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1048                 if (ifa == ifp)
1049                         continue;
1050                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1051                                        ifp->prefix_len))
1052                         continue;
1053                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1054                         return CLEANUP_PREFIX_RT_NOP;
1055
1056                 action = CLEANUP_PREFIX_RT_EXPIRE;
1057
1058                 spin_lock(&ifa->lock);
1059
1060                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1061                 /*
1062                  * Note: Because this address is
1063                  * not permanent, lifetime <
1064                  * LONG_MAX / HZ here.
1065                  */
1066                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1067                         *expires = ifa->tstamp + lifetime * HZ;
1068                 spin_unlock(&ifa->lock);
1069         }
1070
1071         return action;
1072 }
1073
1074 static void
1075 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1076 {
1077         struct rt6_info *rt;
1078
1079         rt = addrconf_get_prefix_route(&ifp->addr,
1080                                        ifp->prefix_len,
1081                                        ifp->idev->dev,
1082                                        0, RTF_GATEWAY | RTF_DEFAULT);
1083         if (rt) {
1084                 if (del_rt)
1085                         ip6_del_rt(rt);
1086                 else {
1087                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1088                                 rt6_set_expires(rt, expires);
1089                         ip6_rt_put(rt);
1090                 }
1091         }
1092 }
1093
1094
1095 /* This function wants to get referenced ifp and releases it before return */
1096
1097 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1098 {
1099         int state;
1100         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1101         unsigned long expires;
1102
1103         ASSERT_RTNL();
1104
1105         spin_lock_bh(&ifp->lock);
1106         state = ifp->state;
1107         ifp->state = INET6_IFADDR_STATE_DEAD;
1108         spin_unlock_bh(&ifp->lock);
1109
1110         if (state == INET6_IFADDR_STATE_DEAD)
1111                 goto out;
1112
1113         spin_lock_bh(&addrconf_hash_lock);
1114         hlist_del_init_rcu(&ifp->addr_lst);
1115         spin_unlock_bh(&addrconf_hash_lock);
1116
1117         write_lock_bh(&ifp->idev->lock);
1118
1119         if (ifp->flags&IFA_F_TEMPORARY) {
1120                 list_del(&ifp->tmp_list);
1121                 if (ifp->ifpub) {
1122                         in6_ifa_put(ifp->ifpub);
1123                         ifp->ifpub = NULL;
1124                 }
1125                 __in6_ifa_put(ifp);
1126         }
1127
1128         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1129                 action = check_cleanup_prefix_route(ifp, &expires);
1130
1131         list_del_init(&ifp->if_list);
1132         __in6_ifa_put(ifp);
1133
1134         write_unlock_bh(&ifp->idev->lock);
1135
1136         addrconf_del_dad_work(ifp);
1137
1138         ipv6_ifa_notify(RTM_DELADDR, ifp);
1139
1140         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1141
1142         if (action != CLEANUP_PREFIX_RT_NOP) {
1143                 cleanup_prefix_route(ifp, expires,
1144                         action == CLEANUP_PREFIX_RT_DEL);
1145         }
1146
1147         /* clean up prefsrc entries */
1148         rt6_remove_prefsrc(ifp);
1149 out:
1150         in6_ifa_put(ifp);
1151 }
1152
1153 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1154 {
1155         struct inet6_dev *idev = ifp->idev;
1156         struct in6_addr addr, *tmpaddr;
1157         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1158         unsigned long regen_advance;
1159         int tmp_plen;
1160         int ret = 0;
1161         u32 addr_flags;
1162         unsigned long now = jiffies;
1163
1164         write_lock_bh(&idev->lock);
1165         if (ift) {
1166                 spin_lock_bh(&ift->lock);
1167                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1168                 spin_unlock_bh(&ift->lock);
1169                 tmpaddr = &addr;
1170         } else {
1171                 tmpaddr = NULL;
1172         }
1173 retry:
1174         in6_dev_hold(idev);
1175         if (idev->cnf.use_tempaddr <= 0) {
1176                 write_unlock_bh(&idev->lock);
1177                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1178                 in6_dev_put(idev);
1179                 ret = -1;
1180                 goto out;
1181         }
1182         spin_lock_bh(&ifp->lock);
1183         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1184                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1185                 spin_unlock_bh(&ifp->lock);
1186                 write_unlock_bh(&idev->lock);
1187                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1188                         __func__);
1189                 in6_dev_put(idev);
1190                 ret = -1;
1191                 goto out;
1192         }
1193         in6_ifa_hold(ifp);
1194         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1195         __ipv6_try_regen_rndid(idev, tmpaddr);
1196         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1197         age = (now - ifp->tstamp) / HZ;
1198         tmp_valid_lft = min_t(__u32,
1199                               ifp->valid_lft,
1200                               idev->cnf.temp_valid_lft + age);
1201         tmp_prefered_lft = min_t(__u32,
1202                                  ifp->prefered_lft,
1203                                  idev->cnf.temp_prefered_lft + age -
1204                                  idev->cnf.max_desync_factor);
1205         tmp_plen = ifp->prefix_len;
1206         tmp_tstamp = ifp->tstamp;
1207         spin_unlock_bh(&ifp->lock);
1208
1209         regen_advance = idev->cnf.regen_max_retry *
1210                         idev->cnf.dad_transmits *
1211                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1212         write_unlock_bh(&idev->lock);
1213
1214         /* A temporary address is created only if this calculated Preferred
1215          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1216          * an implementation must not create a temporary address with a zero
1217          * Preferred Lifetime.
1218          * Use age calculation as in addrconf_verify to avoid unnecessary
1219          * temporary addresses being generated.
1220          */
1221         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1222         if (tmp_prefered_lft <= regen_advance + age) {
1223                 in6_ifa_put(ifp);
1224                 in6_dev_put(idev);
1225                 ret = -1;
1226                 goto out;
1227         }
1228
1229         addr_flags = IFA_F_TEMPORARY;
1230         /* set in addrconf_prefix_rcv() */
1231         if (ifp->flags & IFA_F_OPTIMISTIC)
1232                 addr_flags |= IFA_F_OPTIMISTIC;
1233
1234         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1235                             ipv6_addr_scope(&addr), addr_flags,
1236                             tmp_valid_lft, tmp_prefered_lft);
1237         if (IS_ERR(ift)) {
1238                 in6_ifa_put(ifp);
1239                 in6_dev_put(idev);
1240                 pr_info("%s: retry temporary address regeneration\n", __func__);
1241                 tmpaddr = &addr;
1242                 write_lock_bh(&idev->lock);
1243                 goto retry;
1244         }
1245
1246         spin_lock_bh(&ift->lock);
1247         ift->ifpub = ifp;
1248         ift->cstamp = now;
1249         ift->tstamp = tmp_tstamp;
1250         spin_unlock_bh(&ift->lock);
1251
1252         addrconf_dad_start(ift);
1253         in6_ifa_put(ift);
1254         in6_dev_put(idev);
1255 out:
1256         return ret;
1257 }
1258
1259 /*
1260  *      Choose an appropriate source address (RFC3484)
1261  */
1262 enum {
1263         IPV6_SADDR_RULE_INIT = 0,
1264         IPV6_SADDR_RULE_LOCAL,
1265         IPV6_SADDR_RULE_SCOPE,
1266         IPV6_SADDR_RULE_PREFERRED,
1267 #ifdef CONFIG_IPV6_MIP6
1268         IPV6_SADDR_RULE_HOA,
1269 #endif
1270         IPV6_SADDR_RULE_OIF,
1271         IPV6_SADDR_RULE_LABEL,
1272         IPV6_SADDR_RULE_PRIVACY,
1273         IPV6_SADDR_RULE_ORCHID,
1274         IPV6_SADDR_RULE_PREFIX,
1275 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1276         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1277 #endif
1278         IPV6_SADDR_RULE_MAX
1279 };
1280
1281 struct ipv6_saddr_score {
1282         int                     rule;
1283         int                     addr_type;
1284         struct inet6_ifaddr     *ifa;
1285         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1286         int                     scopedist;
1287         int                     matchlen;
1288 };
1289
1290 struct ipv6_saddr_dst {
1291         const struct in6_addr *addr;
1292         int ifindex;
1293         int scope;
1294         int label;
1295         unsigned int prefs;
1296 };
1297
1298 static inline int ipv6_saddr_preferred(int type)
1299 {
1300         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1301                 return 1;
1302         return 0;
1303 }
1304
1305 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1306 {
1307 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1308         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1309 #else
1310         return false;
1311 #endif
1312 }
1313
1314 static int ipv6_get_saddr_eval(struct net *net,
1315                                struct ipv6_saddr_score *score,
1316                                struct ipv6_saddr_dst *dst,
1317                                int i)
1318 {
1319         int ret;
1320
1321         if (i <= score->rule) {
1322                 switch (i) {
1323                 case IPV6_SADDR_RULE_SCOPE:
1324                         ret = score->scopedist;
1325                         break;
1326                 case IPV6_SADDR_RULE_PREFIX:
1327                         ret = score->matchlen;
1328                         break;
1329                 default:
1330                         ret = !!test_bit(i, score->scorebits);
1331                 }
1332                 goto out;
1333         }
1334
1335         switch (i) {
1336         case IPV6_SADDR_RULE_INIT:
1337                 /* Rule 0: remember if hiscore is not ready yet */
1338                 ret = !!score->ifa;
1339                 break;
1340         case IPV6_SADDR_RULE_LOCAL:
1341                 /* Rule 1: Prefer same address */
1342                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1343                 break;
1344         case IPV6_SADDR_RULE_SCOPE:
1345                 /* Rule 2: Prefer appropriate scope
1346                  *
1347                  *      ret
1348                  *       ^
1349                  *    -1 |  d 15
1350                  *    ---+--+-+---> scope
1351                  *       |
1352                  *       |             d is scope of the destination.
1353                  *  B-d  |  \
1354                  *       |   \      <- smaller scope is better if
1355                  *  B-15 |    \        if scope is enough for destination.
1356                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1357                  * d-C-1 | /
1358                  *       |/         <- greater is better
1359                  *   -C  /             if scope is not enough for destination.
1360                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1361                  *
1362                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1363                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1364                  * Assume B = 0 and we get C > 29.
1365                  */
1366                 ret = __ipv6_addr_src_scope(score->addr_type);
1367                 if (ret >= dst->scope)
1368                         ret = -ret;
1369                 else
1370                         ret -= 128;     /* 30 is enough */
1371                 score->scopedist = ret;
1372                 break;
1373         case IPV6_SADDR_RULE_PREFERRED:
1374             {
1375                 /* Rule 3: Avoid deprecated and optimistic addresses */
1376                 u8 avoid = IFA_F_DEPRECATED;
1377
1378                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1379                         avoid |= IFA_F_OPTIMISTIC;
1380                 ret = ipv6_saddr_preferred(score->addr_type) ||
1381                       !(score->ifa->flags & avoid);
1382                 break;
1383             }
1384 #ifdef CONFIG_IPV6_MIP6
1385         case IPV6_SADDR_RULE_HOA:
1386             {
1387                 /* Rule 4: Prefer home address */
1388                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1389                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1390                 break;
1391             }
1392 #endif
1393         case IPV6_SADDR_RULE_OIF:
1394                 /* Rule 5: Prefer outgoing interface */
1395                 ret = (!dst->ifindex ||
1396                        dst->ifindex == score->ifa->idev->dev->ifindex);
1397                 break;
1398         case IPV6_SADDR_RULE_LABEL:
1399                 /* Rule 6: Prefer matching label */
1400                 ret = ipv6_addr_label(net,
1401                                       &score->ifa->addr, score->addr_type,
1402                                       score->ifa->idev->dev->ifindex) == dst->label;
1403                 break;
1404         case IPV6_SADDR_RULE_PRIVACY:
1405             {
1406                 /* Rule 7: Prefer public address
1407                  * Note: prefer temporary address if use_tempaddr >= 2
1408                  */
1409                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1410                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1411                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1412                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1413                 break;
1414             }
1415         case IPV6_SADDR_RULE_ORCHID:
1416                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1417                  *          non-ORCHID vs non-ORCHID
1418                  */
1419                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1420                         ipv6_addr_orchid(dst->addr));
1421                 break;
1422         case IPV6_SADDR_RULE_PREFIX:
1423                 /* Rule 8: Use longest matching prefix */
1424                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1425                 if (ret > score->ifa->prefix_len)
1426                         ret = score->ifa->prefix_len;
1427                 score->matchlen = ret;
1428                 break;
1429 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1430         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1431                 /* Optimistic addresses still have lower precedence than other
1432                  * preferred addresses.
1433                  */
1434                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1435                 break;
1436 #endif
1437         default:
1438                 ret = 0;
1439         }
1440
1441         if (ret)
1442                 __set_bit(i, score->scorebits);
1443         score->rule = i;
1444 out:
1445         return ret;
1446 }
1447
1448 static int __ipv6_dev_get_saddr(struct net *net,
1449                                 struct ipv6_saddr_dst *dst,
1450                                 struct inet6_dev *idev,
1451                                 struct ipv6_saddr_score *scores,
1452                                 int hiscore_idx)
1453 {
1454         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1455
1456         read_lock_bh(&idev->lock);
1457         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1458                 int i;
1459
1460                 /*
1461                  * - Tentative Address (RFC2462 section 5.4)
1462                  *  - A tentative address is not considered
1463                  *    "assigned to an interface" in the traditional
1464                  *    sense, unless it is also flagged as optimistic.
1465                  * - Candidate Source Address (section 4)
1466                  *  - In any case, anycast addresses, multicast
1467                  *    addresses, and the unspecified address MUST
1468                  *    NOT be included in a candidate set.
1469                  */
1470                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1471                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1472                         continue;
1473
1474                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1475
1476                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1477                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1478                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1479                                             idev->dev->name);
1480                         continue;
1481                 }
1482
1483                 score->rule = -1;
1484                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1485
1486                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1487                         int minihiscore, miniscore;
1488
1489                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1490                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1491
1492                         if (minihiscore > miniscore) {
1493                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1494                                     score->scopedist > 0) {
1495                                         /*
1496                                          * special case:
1497                                          * each remaining entry
1498                                          * has too small (not enough)
1499                                          * scope, because ifa entries
1500                                          * are sorted by their scope
1501                                          * values.
1502                                          */
1503                                         goto out;
1504                                 }
1505                                 break;
1506                         } else if (minihiscore < miniscore) {
1507                                 if (hiscore->ifa)
1508                                         in6_ifa_put(hiscore->ifa);
1509
1510                                 in6_ifa_hold(score->ifa);
1511
1512                                 swap(hiscore, score);
1513                                 hiscore_idx = 1 - hiscore_idx;
1514
1515                                 /* restore our iterator */
1516                                 score->ifa = hiscore->ifa;
1517
1518                                 break;
1519                         }
1520                 }
1521         }
1522 out:
1523         read_unlock_bh(&idev->lock);
1524         return hiscore_idx;
1525 }
1526
1527 static int ipv6_get_saddr_master(struct net *net,
1528                                  const struct net_device *dst_dev,
1529                                  const struct net_device *master,
1530                                  struct ipv6_saddr_dst *dst,
1531                                  struct ipv6_saddr_score *scores,
1532                                  int hiscore_idx)
1533 {
1534         struct inet6_dev *idev;
1535
1536         idev = __in6_dev_get(dst_dev);
1537         if (idev)
1538                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1539                                                    scores, hiscore_idx);
1540
1541         idev = __in6_dev_get(master);
1542         if (idev)
1543                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1544                                                    scores, hiscore_idx);
1545
1546         return hiscore_idx;
1547 }
1548
1549 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1550                        const struct in6_addr *daddr, unsigned int prefs,
1551                        struct in6_addr *saddr)
1552 {
1553         struct ipv6_saddr_score scores[2], *hiscore;
1554         struct ipv6_saddr_dst dst;
1555         struct inet6_dev *idev;
1556         struct net_device *dev;
1557         int dst_type;
1558         bool use_oif_addr = false;
1559         int hiscore_idx = 0;
1560
1561         dst_type = __ipv6_addr_type(daddr);
1562         dst.addr = daddr;
1563         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1564         dst.scope = __ipv6_addr_src_scope(dst_type);
1565         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1566         dst.prefs = prefs;
1567
1568         scores[hiscore_idx].rule = -1;
1569         scores[hiscore_idx].ifa = NULL;
1570
1571         rcu_read_lock();
1572
1573         /* Candidate Source Address (section 4)
1574          *  - multicast and link-local destination address,
1575          *    the set of candidate source address MUST only
1576          *    include addresses assigned to interfaces
1577          *    belonging to the same link as the outgoing
1578          *    interface.
1579          * (- For site-local destination addresses, the
1580          *    set of candidate source addresses MUST only
1581          *    include addresses assigned to interfaces
1582          *    belonging to the same site as the outgoing
1583          *    interface.)
1584          *  - "It is RECOMMENDED that the candidate source addresses
1585          *    be the set of unicast addresses assigned to the
1586          *    interface that will be used to send to the destination
1587          *    (the 'outgoing' interface)." (RFC 6724)
1588          */
1589         if (dst_dev) {
1590                 idev = __in6_dev_get(dst_dev);
1591                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1592                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1593                     (idev && idev->cnf.use_oif_addrs_only)) {
1594                         use_oif_addr = true;
1595                 }
1596         }
1597
1598         if (use_oif_addr) {
1599                 if (idev)
1600                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1601         } else {
1602                 const struct net_device *master;
1603                 int master_idx = 0;
1604
1605                 /* if dst_dev exists and is enslaved to an L3 device, then
1606                  * prefer addresses from dst_dev and then the master over
1607                  * any other enslaved devices in the L3 domain.
1608                  */
1609                 master = l3mdev_master_dev_rcu(dst_dev);
1610                 if (master) {
1611                         master_idx = master->ifindex;
1612
1613                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1614                                                             master, &dst,
1615                                                             scores, hiscore_idx);
1616
1617                         if (scores[hiscore_idx].ifa)
1618                                 goto out;
1619                 }
1620
1621                 for_each_netdev_rcu(net, dev) {
1622                         /* only consider addresses on devices in the
1623                          * same L3 domain
1624                          */
1625                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1626                                 continue;
1627                         idev = __in6_dev_get(dev);
1628                         if (!idev)
1629                                 continue;
1630                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1631                 }
1632         }
1633
1634 out:
1635         rcu_read_unlock();
1636
1637         hiscore = &scores[hiscore_idx];
1638         if (!hiscore->ifa)
1639                 return -EADDRNOTAVAIL;
1640
1641         *saddr = hiscore->ifa->addr;
1642         in6_ifa_put(hiscore->ifa);
1643         return 0;
1644 }
1645 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1646
1647 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1648                       u32 banned_flags)
1649 {
1650         struct inet6_ifaddr *ifp;
1651         int err = -EADDRNOTAVAIL;
1652
1653         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1654                 if (ifp->scope > IFA_LINK)
1655                         break;
1656                 if (ifp->scope == IFA_LINK &&
1657                     !(ifp->flags & banned_flags)) {
1658                         *addr = ifp->addr;
1659                         err = 0;
1660                         break;
1661                 }
1662         }
1663         return err;
1664 }
1665
1666 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1667                     u32 banned_flags)
1668 {
1669         struct inet6_dev *idev;
1670         int err = -EADDRNOTAVAIL;
1671
1672         rcu_read_lock();
1673         idev = __in6_dev_get(dev);
1674         if (idev) {
1675                 read_lock_bh(&idev->lock);
1676                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1677                 read_unlock_bh(&idev->lock);
1678         }
1679         rcu_read_unlock();
1680         return err;
1681 }
1682
1683 static int ipv6_count_addresses(struct inet6_dev *idev)
1684 {
1685         int cnt = 0;
1686         struct inet6_ifaddr *ifp;
1687
1688         read_lock_bh(&idev->lock);
1689         list_for_each_entry(ifp, &idev->addr_list, if_list)
1690                 cnt++;
1691         read_unlock_bh(&idev->lock);
1692         return cnt;
1693 }
1694
1695 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1696                   const struct net_device *dev, int strict)
1697 {
1698         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1699 }
1700 EXPORT_SYMBOL(ipv6_chk_addr);
1701
1702 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1703                             const struct net_device *dev, int strict,
1704                             u32 banned_flags)
1705 {
1706         struct inet6_ifaddr *ifp;
1707         unsigned int hash = inet6_addr_hash(addr);
1708         u32 ifp_flags;
1709
1710         rcu_read_lock_bh();
1711         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1712                 if (!net_eq(dev_net(ifp->idev->dev), net))
1713                         continue;
1714                 /* Decouple optimistic from tentative for evaluation here.
1715                  * Ban optimistic addresses explicitly, when required.
1716                  */
1717                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1718                             ? (ifp->flags&~IFA_F_TENTATIVE)
1719                             : ifp->flags;
1720                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1721                     !(ifp_flags&banned_flags) &&
1722                     (!dev || ifp->idev->dev == dev ||
1723                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1724                         rcu_read_unlock_bh();
1725                         return 1;
1726                 }
1727         }
1728
1729         rcu_read_unlock_bh();
1730         return 0;
1731 }
1732 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1733
1734 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1735                                struct net_device *dev)
1736 {
1737         unsigned int hash = inet6_addr_hash(addr);
1738         struct inet6_ifaddr *ifp;
1739
1740         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1741                 if (!net_eq(dev_net(ifp->idev->dev), net))
1742                         continue;
1743                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1744                         if (!dev || ifp->idev->dev == dev)
1745                                 return true;
1746                 }
1747         }
1748         return false;
1749 }
1750
1751 /* Compares an address/prefix_len with addresses on device @dev.
1752  * If one is found it returns true.
1753  */
1754 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1755         const unsigned int prefix_len, struct net_device *dev)
1756 {
1757         struct inet6_dev *idev;
1758         struct inet6_ifaddr *ifa;
1759         bool ret = false;
1760
1761         rcu_read_lock();
1762         idev = __in6_dev_get(dev);
1763         if (idev) {
1764                 read_lock_bh(&idev->lock);
1765                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1766                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1767                         if (ret)
1768                                 break;
1769                 }
1770                 read_unlock_bh(&idev->lock);
1771         }
1772         rcu_read_unlock();
1773
1774         return ret;
1775 }
1776 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1777
1778 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1779 {
1780         struct inet6_dev *idev;
1781         struct inet6_ifaddr *ifa;
1782         int     onlink;
1783
1784         onlink = 0;
1785         rcu_read_lock();
1786         idev = __in6_dev_get(dev);
1787         if (idev) {
1788                 read_lock_bh(&idev->lock);
1789                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1790                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1791                                                    ifa->prefix_len);
1792                         if (onlink)
1793                                 break;
1794                 }
1795                 read_unlock_bh(&idev->lock);
1796         }
1797         rcu_read_unlock();
1798         return onlink;
1799 }
1800 EXPORT_SYMBOL(ipv6_chk_prefix);
1801
1802 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1803                                      struct net_device *dev, int strict)
1804 {
1805         struct inet6_ifaddr *ifp, *result = NULL;
1806         unsigned int hash = inet6_addr_hash(addr);
1807
1808         rcu_read_lock_bh();
1809         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1810                 if (!net_eq(dev_net(ifp->idev->dev), net))
1811                         continue;
1812                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1813                         if (!dev || ifp->idev->dev == dev ||
1814                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1815                                 result = ifp;
1816                                 in6_ifa_hold(ifp);
1817                                 break;
1818                         }
1819                 }
1820         }
1821         rcu_read_unlock_bh();
1822
1823         return result;
1824 }
1825
1826 /* Gets referenced address, destroys ifaddr */
1827
1828 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1829 {
1830         if (dad_failed)
1831                 ifp->flags |= IFA_F_DADFAILED;
1832
1833         if (ifp->flags&IFA_F_PERMANENT) {
1834                 spin_lock_bh(&ifp->lock);
1835                 addrconf_del_dad_work(ifp);
1836                 ifp->flags |= IFA_F_TENTATIVE;
1837                 spin_unlock_bh(&ifp->lock);
1838                 if (dad_failed)
1839                         ipv6_ifa_notify(0, ifp);
1840                 in6_ifa_put(ifp);
1841         } else if (ifp->flags&IFA_F_TEMPORARY) {
1842                 struct inet6_ifaddr *ifpub;
1843                 spin_lock_bh(&ifp->lock);
1844                 ifpub = ifp->ifpub;
1845                 if (ifpub) {
1846                         in6_ifa_hold(ifpub);
1847                         spin_unlock_bh(&ifp->lock);
1848                         ipv6_create_tempaddr(ifpub, ifp);
1849                         in6_ifa_put(ifpub);
1850                 } else {
1851                         spin_unlock_bh(&ifp->lock);
1852                 }
1853                 ipv6_del_addr(ifp);
1854         } else {
1855                 ipv6_del_addr(ifp);
1856         }
1857 }
1858
1859 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1860 {
1861         int err = -ENOENT;
1862
1863         spin_lock_bh(&ifp->lock);
1864         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1865                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1866                 err = 0;
1867         }
1868         spin_unlock_bh(&ifp->lock);
1869
1870         return err;
1871 }
1872
1873 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1874 {
1875         struct inet6_dev *idev = ifp->idev;
1876         struct net *net = dev_net(ifp->idev->dev);
1877
1878         if (addrconf_dad_end(ifp)) {
1879                 in6_ifa_put(ifp);
1880                 return;
1881         }
1882
1883         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1884                              ifp->idev->dev->name, &ifp->addr);
1885
1886         spin_lock_bh(&ifp->lock);
1887
1888         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1889                 int scope = ifp->scope;
1890                 u32 flags = ifp->flags;
1891                 struct in6_addr new_addr;
1892                 struct inet6_ifaddr *ifp2;
1893                 u32 valid_lft, preferred_lft;
1894                 int pfxlen = ifp->prefix_len;
1895                 int retries = ifp->stable_privacy_retry + 1;
1896
1897                 if (retries > net->ipv6.sysctl.idgen_retries) {
1898                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1899                                              ifp->idev->dev->name);
1900                         goto errdad;
1901                 }
1902
1903                 new_addr = ifp->addr;
1904                 if (ipv6_generate_stable_address(&new_addr, retries,
1905                                                  idev))
1906                         goto errdad;
1907
1908                 valid_lft = ifp->valid_lft;
1909                 preferred_lft = ifp->prefered_lft;
1910
1911                 spin_unlock_bh(&ifp->lock);
1912
1913                 if (idev->cnf.max_addresses &&
1914                     ipv6_count_addresses(idev) >=
1915                     idev->cnf.max_addresses)
1916                         goto lock_errdad;
1917
1918                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1919                                      ifp->idev->dev->name);
1920
1921                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1922                                      scope, flags, valid_lft,
1923                                      preferred_lft);
1924                 if (IS_ERR(ifp2))
1925                         goto lock_errdad;
1926
1927                 spin_lock_bh(&ifp2->lock);
1928                 ifp2->stable_privacy_retry = retries;
1929                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1930                 spin_unlock_bh(&ifp2->lock);
1931
1932                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1933                 in6_ifa_put(ifp2);
1934 lock_errdad:
1935                 spin_lock_bh(&ifp->lock);
1936         }
1937
1938 errdad:
1939         /* transition from _POSTDAD to _ERRDAD */
1940         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1941         spin_unlock_bh(&ifp->lock);
1942
1943         addrconf_mod_dad_work(ifp, 0);
1944 }
1945
1946 /* Join to solicited addr multicast group.
1947  * caller must hold RTNL */
1948 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1949 {
1950         struct in6_addr maddr;
1951
1952         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1953                 return;
1954
1955         addrconf_addr_solict_mult(addr, &maddr);
1956         ipv6_dev_mc_inc(dev, &maddr);
1957 }
1958
1959 /* caller must hold RTNL */
1960 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1961 {
1962         struct in6_addr maddr;
1963
1964         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1965                 return;
1966
1967         addrconf_addr_solict_mult(addr, &maddr);
1968         __ipv6_dev_mc_dec(idev, &maddr);
1969 }
1970
1971 /* caller must hold RTNL */
1972 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1973 {
1974         struct in6_addr addr;
1975
1976         if (ifp->prefix_len >= 127) /* RFC 6164 */
1977                 return;
1978         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1979         if (ipv6_addr_any(&addr))
1980                 return;
1981         __ipv6_dev_ac_inc(ifp->idev, &addr);
1982 }
1983
1984 /* caller must hold RTNL */
1985 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1986 {
1987         struct in6_addr addr;
1988
1989         if (ifp->prefix_len >= 127) /* RFC 6164 */
1990                 return;
1991         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1992         if (ipv6_addr_any(&addr))
1993                 return;
1994         __ipv6_dev_ac_dec(ifp->idev, &addr);
1995 }
1996
1997 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1998 {
1999         if (dev->addr_len != EUI64_ADDR_LEN)
2000                 return -1;
2001         memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2002         eui[0] ^= 2;
2003         return 0;
2004 }
2005
2006 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2007 {
2008         union fwnet_hwaddr *ha;
2009
2010         if (dev->addr_len != FWNET_ALEN)
2011                 return -1;
2012
2013         ha = (union fwnet_hwaddr *)dev->dev_addr;
2014
2015         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2016         eui[0] ^= 2;
2017         return 0;
2018 }
2019
2020 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2021 {
2022         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2023         if (dev->addr_len != ARCNET_ALEN)
2024                 return -1;
2025         memset(eui, 0, 7);
2026         eui[7] = *(u8 *)dev->dev_addr;
2027         return 0;
2028 }
2029
2030 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2031 {
2032         if (dev->addr_len != INFINIBAND_ALEN)
2033                 return -1;
2034         memcpy(eui, dev->dev_addr + 12, 8);
2035         eui[0] |= 2;
2036         return 0;
2037 }
2038
2039 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2040 {
2041         if (addr == 0)
2042                 return -1;
2043         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2044                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2045                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2046                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2047                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2048                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2049         eui[1] = 0;
2050         eui[2] = 0x5E;
2051         eui[3] = 0xFE;
2052         memcpy(eui + 4, &addr, 4);
2053         return 0;
2054 }
2055
2056 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2057 {
2058         if (dev->priv_flags & IFF_ISATAP)
2059                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2060         return -1;
2061 }
2062
2063 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2064 {
2065         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2066 }
2067
2068 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2069 {
2070         memcpy(eui, dev->perm_addr, 3);
2071         memcpy(eui + 5, dev->perm_addr + 3, 3);
2072         eui[3] = 0xFF;
2073         eui[4] = 0xFE;
2074         eui[0] ^= 2;
2075         return 0;
2076 }
2077
2078 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2079 {
2080         switch (dev->type) {
2081         case ARPHRD_ETHER:
2082         case ARPHRD_FDDI:
2083                 return addrconf_ifid_eui48(eui, dev);
2084         case ARPHRD_ARCNET:
2085                 return addrconf_ifid_arcnet(eui, dev);
2086         case ARPHRD_INFINIBAND:
2087                 return addrconf_ifid_infiniband(eui, dev);
2088         case ARPHRD_SIT:
2089                 return addrconf_ifid_sit(eui, dev);
2090         case ARPHRD_IPGRE:
2091                 return addrconf_ifid_gre(eui, dev);
2092         case ARPHRD_6LOWPAN:
2093                 return addrconf_ifid_eui64(eui, dev);
2094         case ARPHRD_IEEE1394:
2095                 return addrconf_ifid_ieee1394(eui, dev);
2096         case ARPHRD_TUNNEL6:
2097                 return addrconf_ifid_ip6tnl(eui, dev);
2098         }
2099         return -1;
2100 }
2101
2102 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2103 {
2104         int err = -1;
2105         struct inet6_ifaddr *ifp;
2106
2107         read_lock_bh(&idev->lock);
2108         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2109                 if (ifp->scope > IFA_LINK)
2110                         break;
2111                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2112                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2113                         err = 0;
2114                         break;
2115                 }
2116         }
2117         read_unlock_bh(&idev->lock);
2118         return err;
2119 }
2120
2121 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2122 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2123 {
2124 regen:
2125         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2126         idev->rndid[0] &= ~0x02;
2127
2128         /*
2129          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2130          * check if generated address is not inappropriate
2131          *
2132          *  - Reserved subnet anycast (RFC 2526)
2133          *      11111101 11....11 1xxxxxxx
2134          *  - ISATAP (RFC4214) 6.1
2135          *      00-00-5E-FE-xx-xx-xx-xx
2136          *  - value 0
2137          *  - XXX: already assigned to an address on the device
2138          */
2139         if (idev->rndid[0] == 0xfd &&
2140             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2141             (idev->rndid[7]&0x80))
2142                 goto regen;
2143         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2144                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2145                         goto regen;
2146                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2147                         goto regen;
2148         }
2149 }
2150
2151 static void ipv6_regen_rndid(unsigned long data)
2152 {
2153         struct inet6_dev *idev = (struct inet6_dev *) data;
2154         unsigned long expires;
2155
2156         rcu_read_lock_bh();
2157         write_lock_bh(&idev->lock);
2158
2159         if (idev->dead)
2160                 goto out;
2161
2162         __ipv6_regen_rndid(idev);
2163
2164         expires = jiffies +
2165                 idev->cnf.temp_prefered_lft * HZ -
2166                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2167                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2168                 idev->cnf.max_desync_factor * HZ;
2169         if (time_before(expires, jiffies)) {
2170                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2171                         __func__, idev->dev->name);
2172                 goto out;
2173         }
2174
2175         if (!mod_timer(&idev->regen_timer, expires))
2176                 in6_dev_hold(idev);
2177
2178 out:
2179         write_unlock_bh(&idev->lock);
2180         rcu_read_unlock_bh();
2181         in6_dev_put(idev);
2182 }
2183
2184 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2185 {
2186         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2187                 __ipv6_regen_rndid(idev);
2188 }
2189
2190 /*
2191  *      Add prefix route.
2192  */
2193
2194 static void
2195 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2196                       unsigned long expires, u32 flags)
2197 {
2198         struct fib6_config cfg = {
2199                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2200                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2201                 .fc_ifindex = dev->ifindex,
2202                 .fc_expires = expires,
2203                 .fc_dst_len = plen,
2204                 .fc_flags = RTF_UP | flags,
2205                 .fc_nlinfo.nl_net = dev_net(dev),
2206                 .fc_protocol = RTPROT_KERNEL,
2207         };
2208
2209         cfg.fc_dst = *pfx;
2210
2211         /* Prevent useless cloning on PtP SIT.
2212            This thing is done here expecting that the whole
2213            class of non-broadcast devices need not cloning.
2214          */
2215 #if IS_ENABLED(CONFIG_IPV6_SIT)
2216         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2217                 cfg.fc_flags |= RTF_NONEXTHOP;
2218 #endif
2219
2220         ip6_route_add(&cfg);
2221 }
2222
2223
2224 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2225                                                   int plen,
2226                                                   const struct net_device *dev,
2227                                                   u32 flags, u32 noflags)
2228 {
2229         struct fib6_node *fn;
2230         struct rt6_info *rt = NULL;
2231         struct fib6_table *table;
2232         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2233
2234         table = fib6_get_table(dev_net(dev), tb_id);
2235         if (!table)
2236                 return NULL;
2237
2238         read_lock_bh(&table->tb6_lock);
2239         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2240         if (!fn)
2241                 goto out;
2242
2243         noflags |= RTF_CACHE;
2244         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2245                 if (rt->dst.dev->ifindex != dev->ifindex)
2246                         continue;
2247                 if ((rt->rt6i_flags & flags) != flags)
2248                         continue;
2249                 if ((rt->rt6i_flags & noflags) != 0)
2250                         continue;
2251                 dst_hold(&rt->dst);
2252                 break;
2253         }
2254 out:
2255         read_unlock_bh(&table->tb6_lock);
2256         return rt;
2257 }
2258
2259
2260 /* Create "default" multicast route to the interface */
2261
2262 static void addrconf_add_mroute(struct net_device *dev)
2263 {
2264         struct fib6_config cfg = {
2265                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2266                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2267                 .fc_ifindex = dev->ifindex,
2268                 .fc_dst_len = 8,
2269                 .fc_flags = RTF_UP,
2270                 .fc_nlinfo.nl_net = dev_net(dev),
2271         };
2272
2273         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2274
2275         ip6_route_add(&cfg);
2276 }
2277
2278 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2279 {
2280         struct inet6_dev *idev;
2281
2282         ASSERT_RTNL();
2283
2284         idev = ipv6_find_idev(dev);
2285         if (!idev)
2286                 return ERR_PTR(-ENOBUFS);
2287
2288         if (idev->cnf.disable_ipv6)
2289                 return ERR_PTR(-EACCES);
2290
2291         /* Add default multicast route */
2292         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2293                 addrconf_add_mroute(dev);
2294
2295         return idev;
2296 }
2297
2298 static void manage_tempaddrs(struct inet6_dev *idev,
2299                              struct inet6_ifaddr *ifp,
2300                              __u32 valid_lft, __u32 prefered_lft,
2301                              bool create, unsigned long now)
2302 {
2303         u32 flags;
2304         struct inet6_ifaddr *ift;
2305
2306         read_lock_bh(&idev->lock);
2307         /* update all temporary addresses in the list */
2308         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2309                 int age, max_valid, max_prefered;
2310
2311                 if (ifp != ift->ifpub)
2312                         continue;
2313
2314                 /* RFC 4941 section 3.3:
2315                  * If a received option will extend the lifetime of a public
2316                  * address, the lifetimes of temporary addresses should
2317                  * be extended, subject to the overall constraint that no
2318                  * temporary addresses should ever remain "valid" or "preferred"
2319                  * for a time longer than (TEMP_VALID_LIFETIME) or
2320                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2321                  */
2322                 age = (now - ift->cstamp) / HZ;
2323                 max_valid = idev->cnf.temp_valid_lft - age;
2324                 if (max_valid < 0)
2325                         max_valid = 0;
2326
2327                 max_prefered = idev->cnf.temp_prefered_lft -
2328                                idev->cnf.max_desync_factor - age;
2329                 if (max_prefered < 0)
2330                         max_prefered = 0;
2331
2332                 if (valid_lft > max_valid)
2333                         valid_lft = max_valid;
2334
2335                 if (prefered_lft > max_prefered)
2336                         prefered_lft = max_prefered;
2337
2338                 spin_lock(&ift->lock);
2339                 flags = ift->flags;
2340                 ift->valid_lft = valid_lft;
2341                 ift->prefered_lft = prefered_lft;
2342                 ift->tstamp = now;
2343                 if (prefered_lft > 0)
2344                         ift->flags &= ~IFA_F_DEPRECATED;
2345
2346                 spin_unlock(&ift->lock);
2347                 if (!(flags&IFA_F_TENTATIVE))
2348                         ipv6_ifa_notify(0, ift);
2349         }
2350
2351         if ((create || list_empty(&idev->tempaddr_list)) &&
2352             idev->cnf.use_tempaddr > 0) {
2353                 /* When a new public address is created as described
2354                  * in [ADDRCONF], also create a new temporary address.
2355                  * Also create a temporary address if it's enabled but
2356                  * no temporary address currently exists.
2357                  */
2358                 read_unlock_bh(&idev->lock);
2359                 ipv6_create_tempaddr(ifp, NULL);
2360         } else {
2361                 read_unlock_bh(&idev->lock);
2362         }
2363 }
2364
2365 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2366 {
2367         return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2368                idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2369 }
2370
2371 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2372                                  const struct prefix_info *pinfo,
2373                                  struct inet6_dev *in6_dev,
2374                                  const struct in6_addr *addr, int addr_type,
2375                                  u32 addr_flags, bool sllao, bool tokenized,
2376                                  __u32 valid_lft, u32 prefered_lft)
2377 {
2378         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2379         int create = 0, update_lft = 0;
2380
2381         if (!ifp && valid_lft) {
2382                 int max_addresses = in6_dev->cnf.max_addresses;
2383
2384 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2385                 if (in6_dev->cnf.optimistic_dad &&
2386                     !net->ipv6.devconf_all->forwarding && sllao)
2387                         addr_flags |= IFA_F_OPTIMISTIC;
2388 #endif
2389
2390                 /* Do not allow to create too much of autoconfigured
2391                  * addresses; this would be too easy way to crash kernel.
2392                  */
2393                 if (!max_addresses ||
2394                     ipv6_count_addresses(in6_dev) < max_addresses)
2395                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2396                                             pinfo->prefix_len,
2397                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2398                                             addr_flags, valid_lft,
2399                                             prefered_lft);
2400
2401                 if (IS_ERR_OR_NULL(ifp))
2402                         return -1;
2403
2404                 update_lft = 0;
2405                 create = 1;
2406                 spin_lock_bh(&ifp->lock);
2407                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2408                 ifp->cstamp = jiffies;
2409                 ifp->tokenized = tokenized;
2410                 spin_unlock_bh(&ifp->lock);
2411                 addrconf_dad_start(ifp);
2412         }
2413
2414         if (ifp) {
2415                 u32 flags;
2416                 unsigned long now;
2417                 u32 stored_lft;
2418
2419                 /* update lifetime (RFC2462 5.5.3 e) */
2420                 spin_lock_bh(&ifp->lock);
2421                 now = jiffies;
2422                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2423                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2424                 else
2425                         stored_lft = 0;
2426                 if (!update_lft && !create && stored_lft) {
2427                         const u32 minimum_lft = min_t(u32,
2428                                 stored_lft, MIN_VALID_LIFETIME);
2429                         valid_lft = max(valid_lft, minimum_lft);
2430
2431                         /* RFC4862 Section 5.5.3e:
2432                          * "Note that the preferred lifetime of the
2433                          *  corresponding address is always reset to
2434                          *  the Preferred Lifetime in the received
2435                          *  Prefix Information option, regardless of
2436                          *  whether the valid lifetime is also reset or
2437                          *  ignored."
2438                          *
2439                          * So we should always update prefered_lft here.
2440                          */
2441                         update_lft = 1;
2442                 }
2443
2444                 if (update_lft) {
2445                         ifp->valid_lft = valid_lft;
2446                         ifp->prefered_lft = prefered_lft;
2447                         ifp->tstamp = now;
2448                         flags = ifp->flags;
2449                         ifp->flags &= ~IFA_F_DEPRECATED;
2450                         spin_unlock_bh(&ifp->lock);
2451
2452                         if (!(flags&IFA_F_TENTATIVE))
2453                                 ipv6_ifa_notify(0, ifp);
2454                 } else
2455                         spin_unlock_bh(&ifp->lock);
2456
2457                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2458                                  create, now);
2459
2460                 in6_ifa_put(ifp);
2461                 addrconf_verify();
2462         }
2463
2464         return 0;
2465 }
2466 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2467
2468 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2469 {
2470         struct prefix_info *pinfo;
2471         __u32 valid_lft;
2472         __u32 prefered_lft;
2473         int addr_type, err;
2474         u32 addr_flags = 0;
2475         struct inet6_dev *in6_dev;
2476         struct net *net = dev_net(dev);
2477
2478         pinfo = (struct prefix_info *) opt;
2479
2480         if (len < sizeof(struct prefix_info)) {
2481                 ADBG("addrconf: prefix option too short\n");
2482                 return;
2483         }
2484
2485         /*
2486          *      Validation checks ([ADDRCONF], page 19)
2487          */
2488
2489         addr_type = ipv6_addr_type(&pinfo->prefix);
2490
2491         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2492                 return;
2493
2494         valid_lft = ntohl(pinfo->valid);
2495         prefered_lft = ntohl(pinfo->prefered);
2496
2497         if (prefered_lft > valid_lft) {
2498                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2499                 return;
2500         }
2501
2502         in6_dev = in6_dev_get(dev);
2503
2504         if (!in6_dev) {
2505                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2506                                     dev->name);
2507                 return;
2508         }
2509
2510         /*
2511          *      Two things going on here:
2512          *      1) Add routes for on-link prefixes
2513          *      2) Configure prefixes with the auto flag set
2514          */
2515
2516         if (pinfo->onlink) {
2517                 struct rt6_info *rt;
2518                 unsigned long rt_expires;
2519
2520                 /* Avoid arithmetic overflow. Really, we could
2521                  * save rt_expires in seconds, likely valid_lft,
2522                  * but it would require division in fib gc, that it
2523                  * not good.
2524                  */
2525                 if (HZ > USER_HZ)
2526                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2527                 else
2528                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2529
2530                 if (addrconf_finite_timeout(rt_expires))
2531                         rt_expires *= HZ;
2532
2533                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2534                                                pinfo->prefix_len,
2535                                                dev,
2536                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2537                                                RTF_GATEWAY | RTF_DEFAULT);
2538
2539                 if (rt) {
2540                         /* Autoconf prefix route */
2541                         if (valid_lft == 0) {
2542                                 ip6_del_rt(rt);
2543                                 rt = NULL;
2544                         } else if (addrconf_finite_timeout(rt_expires)) {
2545                                 /* not infinity */
2546                                 rt6_set_expires(rt, jiffies + rt_expires);
2547                         } else {
2548                                 rt6_clean_expires(rt);
2549                         }
2550                 } else if (valid_lft) {
2551                         clock_t expires = 0;
2552                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2553                         if (addrconf_finite_timeout(rt_expires)) {
2554                                 /* not infinity */
2555                                 flags |= RTF_EXPIRES;
2556                                 expires = jiffies_to_clock_t(rt_expires);
2557                         }
2558                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2559                                               dev, expires, flags);
2560                 }
2561                 ip6_rt_put(rt);
2562         }
2563
2564         /* Try to figure out our local address for this prefix */
2565
2566         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2567                 struct in6_addr addr;
2568                 bool tokenized = false, dev_addr_generated = false;
2569
2570                 if (pinfo->prefix_len == 64) {
2571                         memcpy(&addr, &pinfo->prefix, 8);
2572
2573                         if (!ipv6_addr_any(&in6_dev->token)) {
2574                                 read_lock_bh(&in6_dev->lock);
2575                                 memcpy(addr.s6_addr + 8,
2576                                        in6_dev->token.s6_addr + 8, 8);
2577                                 read_unlock_bh(&in6_dev->lock);
2578                                 tokenized = true;
2579                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2580                                    !ipv6_generate_stable_address(&addr, 0,
2581                                                                  in6_dev)) {
2582                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2583                                 goto ok;
2584                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2585                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2586                                 goto put;
2587                         } else {
2588                                 dev_addr_generated = true;
2589                         }
2590                         goto ok;
2591                 }
2592                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2593                                     pinfo->prefix_len);
2594                 goto put;
2595
2596 ok:
2597                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2598                                                    &addr, addr_type,
2599                                                    addr_flags, sllao,
2600                                                    tokenized, valid_lft,
2601                                                    prefered_lft);
2602                 if (err)
2603                         goto put;
2604
2605                 /* Ignore error case here because previous prefix add addr was
2606                  * successful which will be notified.
2607                  */
2608                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2609                                               addr_type, addr_flags, sllao,
2610                                               tokenized, valid_lft,
2611                                               prefered_lft,
2612                                               dev_addr_generated);
2613         }
2614         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2615 put:
2616         in6_dev_put(in6_dev);
2617 }
2618
2619 /*
2620  *      Set destination address.
2621  *      Special case for SIT interfaces where we create a new "virtual"
2622  *      device.
2623  */
2624 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2625 {
2626         struct in6_ifreq ireq;
2627         struct net_device *dev;
2628         int err = -EINVAL;
2629
2630         rtnl_lock();
2631
2632         err = -EFAULT;
2633         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2634                 goto err_exit;
2635
2636         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2637
2638         err = -ENODEV;
2639         if (!dev)
2640                 goto err_exit;
2641
2642 #if IS_ENABLED(CONFIG_IPV6_SIT)
2643         if (dev->type == ARPHRD_SIT) {
2644                 const struct net_device_ops *ops = dev->netdev_ops;
2645                 struct ifreq ifr;
2646                 struct ip_tunnel_parm p;
2647
2648                 err = -EADDRNOTAVAIL;
2649                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2650                         goto err_exit;
2651
2652                 memset(&p, 0, sizeof(p));
2653                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2654                 p.iph.saddr = 0;
2655                 p.iph.version = 4;
2656                 p.iph.ihl = 5;
2657                 p.iph.protocol = IPPROTO_IPV6;
2658                 p.iph.ttl = 64;
2659                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2660
2661                 if (ops->ndo_do_ioctl) {
2662                         mm_segment_t oldfs = get_fs();
2663
2664                         set_fs(KERNEL_DS);
2665                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2666                         set_fs(oldfs);
2667                 } else
2668                         err = -EOPNOTSUPP;
2669
2670                 if (err == 0) {
2671                         err = -ENOBUFS;
2672                         dev = __dev_get_by_name(net, p.name);
2673                         if (!dev)
2674                                 goto err_exit;
2675                         err = dev_open(dev);
2676                 }
2677         }
2678 #endif
2679
2680 err_exit:
2681         rtnl_unlock();
2682         return err;
2683 }
2684
2685 static int ipv6_mc_config(struct sock *sk, bool join,
2686                           const struct in6_addr *addr, int ifindex)
2687 {
2688         int ret;
2689
2690         ASSERT_RTNL();
2691
2692         lock_sock(sk);
2693         if (join)
2694                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2695         else
2696                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2697         release_sock(sk);
2698
2699         return ret;
2700 }
2701
2702 /*
2703  *      Manual configuration of address on an interface
2704  */
2705 static int inet6_addr_add(struct net *net, int ifindex,
2706                           const struct in6_addr *pfx,
2707                           const struct in6_addr *peer_pfx,
2708                           unsigned int plen, __u32 ifa_flags,
2709                           __u32 prefered_lft, __u32 valid_lft)
2710 {
2711         struct inet6_ifaddr *ifp;
2712         struct inet6_dev *idev;
2713         struct net_device *dev;
2714         unsigned long timeout;
2715         clock_t expires;
2716         int scope;
2717         u32 flags;
2718
2719         ASSERT_RTNL();
2720
2721         if (plen > 128)
2722                 return -EINVAL;
2723
2724         /* check the lifetime */
2725         if (!valid_lft || prefered_lft > valid_lft)
2726                 return -EINVAL;
2727
2728         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2729                 return -EINVAL;
2730
2731         dev = __dev_get_by_index(net, ifindex);
2732         if (!dev)
2733                 return -ENODEV;
2734
2735         idev = addrconf_add_dev(dev);
2736         if (IS_ERR(idev))
2737                 return PTR_ERR(idev);
2738
2739         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2740                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2741                                          true, pfx, ifindex);
2742
2743                 if (ret < 0)
2744                         return ret;
2745         }
2746
2747         scope = ipv6_addr_scope(pfx);
2748
2749         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2750         if (addrconf_finite_timeout(timeout)) {
2751                 expires = jiffies_to_clock_t(timeout * HZ);
2752                 valid_lft = timeout;
2753                 flags = RTF_EXPIRES;
2754         } else {
2755                 expires = 0;
2756                 flags = 0;
2757                 ifa_flags |= IFA_F_PERMANENT;
2758         }
2759
2760         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2761         if (addrconf_finite_timeout(timeout)) {
2762                 if (timeout == 0)
2763                         ifa_flags |= IFA_F_DEPRECATED;
2764                 prefered_lft = timeout;
2765         }
2766
2767         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2768                             valid_lft, prefered_lft);
2769
2770         if (!IS_ERR(ifp)) {
2771                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2772                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2773                                               expires, flags);
2774                 }
2775
2776                 /*
2777                  * Note that section 3.1 of RFC 4429 indicates
2778                  * that the Optimistic flag should not be set for
2779                  * manually configured addresses
2780                  */
2781                 addrconf_dad_start(ifp);
2782                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2783                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2784                                          true, jiffies);
2785                 in6_ifa_put(ifp);
2786                 addrconf_verify_rtnl();
2787                 return 0;
2788         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2789                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2790                                false, pfx, ifindex);
2791         }
2792
2793         return PTR_ERR(ifp);
2794 }
2795
2796 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2797                           const struct in6_addr *pfx, unsigned int plen)
2798 {
2799         struct inet6_ifaddr *ifp;
2800         struct inet6_dev *idev;
2801         struct net_device *dev;
2802
2803         if (plen > 128)
2804                 return -EINVAL;
2805
2806         dev = __dev_get_by_index(net, ifindex);
2807         if (!dev)
2808                 return -ENODEV;
2809
2810         idev = __in6_dev_get(dev);
2811         if (!idev)
2812                 return -ENXIO;
2813
2814         read_lock_bh(&idev->lock);
2815         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2816                 if (ifp->prefix_len == plen &&
2817                     ipv6_addr_equal(pfx, &ifp->addr)) {
2818                         in6_ifa_hold(ifp);
2819                         read_unlock_bh(&idev->lock);
2820
2821                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2822                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2823                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2824                                                  jiffies);
2825                         ipv6_del_addr(ifp);
2826                         addrconf_verify_rtnl();
2827                         if (ipv6_addr_is_multicast(pfx)) {
2828                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2829                                                false, pfx, dev->ifindex);
2830                         }
2831                         return 0;
2832                 }
2833         }
2834         read_unlock_bh(&idev->lock);
2835         return -EADDRNOTAVAIL;
2836 }
2837
2838
2839 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2840 {
2841         struct in6_ifreq ireq;
2842         int err;
2843
2844         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2845                 return -EPERM;
2846
2847         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2848                 return -EFAULT;
2849
2850         rtnl_lock();
2851         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2852                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2853                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2854         rtnl_unlock();
2855         return err;
2856 }
2857
2858 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2859 {
2860         struct in6_ifreq ireq;
2861         int err;
2862
2863         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2864                 return -EPERM;
2865
2866         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2867                 return -EFAULT;
2868
2869         rtnl_lock();
2870         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2871                              ireq.ifr6_prefixlen);
2872         rtnl_unlock();
2873         return err;
2874 }
2875
2876 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2877                      int plen, int scope)
2878 {
2879         struct inet6_ifaddr *ifp;
2880
2881         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2882                             scope, IFA_F_PERMANENT,
2883                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2884         if (!IS_ERR(ifp)) {
2885                 spin_lock_bh(&ifp->lock);
2886                 ifp->flags &= ~IFA_F_TENTATIVE;
2887                 spin_unlock_bh(&ifp->lock);
2888                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2889                 in6_ifa_put(ifp);
2890         }
2891 }
2892
2893 #if IS_ENABLED(CONFIG_IPV6_SIT)
2894 static void sit_add_v4_addrs(struct inet6_dev *idev)
2895 {
2896         struct in6_addr addr;
2897         struct net_device *dev;
2898         struct net *net = dev_net(idev->dev);
2899         int scope, plen;
2900         u32 pflags = 0;
2901
2902         ASSERT_RTNL();
2903
2904         memset(&addr, 0, sizeof(struct in6_addr));
2905         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2906
2907         if (idev->dev->flags&IFF_POINTOPOINT) {
2908                 addr.s6_addr32[0] = htonl(0xfe800000);
2909                 scope = IFA_LINK;
2910                 plen = 64;
2911         } else {
2912                 scope = IPV6_ADDR_COMPATv4;
2913                 plen = 96;
2914                 pflags |= RTF_NONEXTHOP;
2915         }
2916
2917         if (addr.s6_addr32[3]) {
2918                 add_addr(idev, &addr, plen, scope);
2919                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2920                 return;
2921         }
2922
2923         for_each_netdev(net, dev) {
2924                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2925                 if (in_dev && (dev->flags & IFF_UP)) {
2926                         struct in_ifaddr *ifa;
2927
2928                         int flag = scope;
2929
2930                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2931
2932                                 addr.s6_addr32[3] = ifa->ifa_local;
2933
2934                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2935                                         continue;
2936                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2937                                         if (idev->dev->flags&IFF_POINTOPOINT)
2938                                                 continue;
2939                                         flag |= IFA_HOST;
2940                                 }
2941
2942                                 add_addr(idev, &addr, plen, flag);
2943                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2944                                                       pflags);
2945                         }
2946                 }
2947         }
2948 }
2949 #endif
2950
2951 static void init_loopback(struct net_device *dev)
2952 {
2953         struct inet6_dev  *idev;
2954         struct net_device *sp_dev;
2955         struct inet6_ifaddr *sp_ifa;
2956         struct rt6_info *sp_rt;
2957
2958         /* ::1 */
2959
2960         ASSERT_RTNL();
2961
2962         idev = ipv6_find_idev(dev);
2963         if (!idev) {
2964                 pr_debug("%s: add_dev failed\n", __func__);
2965                 return;
2966         }
2967
2968         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2969
2970         /* Add routes to other interface's IPv6 addresses */
2971         for_each_netdev(dev_net(dev), sp_dev) {
2972                 if (!strcmp(sp_dev->name, dev->name))
2973                         continue;
2974
2975                 idev = __in6_dev_get(sp_dev);
2976                 if (!idev)
2977                         continue;
2978
2979                 read_lock_bh(&idev->lock);
2980                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2981
2982                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2983                                 continue;
2984
2985                         if (sp_ifa->rt) {
2986                                 /* This dst has been added to garbage list when
2987                                  * lo device down, release this obsolete dst and
2988                                  * reallocate a new router for ifa.
2989                                  */
2990                                 if (sp_ifa->rt->dst.obsolete > 0) {
2991                                         ip6_rt_put(sp_ifa->rt);
2992                                         sp_ifa->rt = NULL;
2993                                 } else {
2994                                         continue;
2995                                 }
2996                         }
2997
2998                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2999
3000                         /* Failure cases are ignored */
3001                         if (!IS_ERR(sp_rt)) {
3002                                 sp_ifa->rt = sp_rt;
3003                                 ip6_ins_rt(sp_rt);
3004                         }
3005                 }
3006                 read_unlock_bh(&idev->lock);
3007         }
3008 }
3009
3010 void addrconf_add_linklocal(struct inet6_dev *idev,
3011                             const struct in6_addr *addr, u32 flags)
3012 {
3013         struct inet6_ifaddr *ifp;
3014         u32 addr_flags = flags | IFA_F_PERMANENT;
3015
3016 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3017         if (idev->cnf.optimistic_dad &&
3018             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3019                 addr_flags |= IFA_F_OPTIMISTIC;
3020 #endif
3021
3022         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3023                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3024         if (!IS_ERR(ifp)) {
3025                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3026                 addrconf_dad_start(ifp);
3027                 in6_ifa_put(ifp);
3028         }
3029 }
3030 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3031
3032 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3033 {
3034         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3035                 return true;
3036
3037         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3038             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3039                 return true;
3040
3041         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3042             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3043                 return true;
3044
3045         return false;
3046 }
3047
3048 static int ipv6_generate_stable_address(struct in6_addr *address,
3049                                         u8 dad_count,
3050                                         const struct inet6_dev *idev)
3051 {
3052         static DEFINE_SPINLOCK(lock);
3053         static __u32 digest[SHA_DIGEST_WORDS];
3054         static __u32 workspace[SHA_WORKSPACE_WORDS];
3055
3056         static union {
3057                 char __data[SHA_MESSAGE_BYTES];
3058                 struct {
3059                         struct in6_addr secret;
3060                         __be32 prefix[2];
3061                         unsigned char hwaddr[MAX_ADDR_LEN];
3062                         u8 dad_count;
3063                 } __packed;
3064         } data;
3065
3066         struct in6_addr secret;
3067         struct in6_addr temp;
3068         struct net *net = dev_net(idev->dev);
3069
3070         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3071
3072         if (idev->cnf.stable_secret.initialized)
3073                 secret = idev->cnf.stable_secret.secret;
3074         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3075                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3076         else
3077                 return -1;
3078
3079 retry:
3080         spin_lock_bh(&lock);
3081
3082         sha_init(digest);
3083         memset(&data, 0, sizeof(data));
3084         memset(workspace, 0, sizeof(workspace));
3085         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3086         data.prefix[0] = address->s6_addr32[0];
3087         data.prefix[1] = address->s6_addr32[1];
3088         data.secret = secret;
3089         data.dad_count = dad_count;
3090
3091         sha_transform(digest, data.__data, workspace);
3092
3093         temp = *address;
3094         temp.s6_addr32[2] = (__force __be32)digest[0];
3095         temp.s6_addr32[3] = (__force __be32)digest[1];
3096
3097         spin_unlock_bh(&lock);
3098
3099         if (ipv6_reserved_interfaceid(temp)) {
3100                 dad_count++;
3101                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3102                         return -1;
3103                 goto retry;
3104         }
3105
3106         *address = temp;
3107         return 0;
3108 }
3109
3110 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3111 {
3112         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3113
3114         if (s->initialized)
3115                 return;
3116         s = &idev->cnf.stable_secret;
3117         get_random_bytes(&s->secret, sizeof(s->secret));
3118         s->initialized = true;
3119 }
3120
3121 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3122 {
3123         struct in6_addr addr;
3124
3125         /* no link local addresses on L3 master devices */
3126         if (netif_is_l3_master(idev->dev))
3127                 return;
3128
3129         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3130
3131         switch (idev->addr_gen_mode) {
3132         case IN6_ADDR_GEN_MODE_RANDOM:
3133                 ipv6_gen_mode_random_init(idev);
3134                 /* fallthrough */
3135         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3136                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3137                         addrconf_add_linklocal(idev, &addr,
3138                                                IFA_F_STABLE_PRIVACY);
3139                 else if (prefix_route)
3140                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3141                 break;
3142         case IN6_ADDR_GEN_MODE_EUI64:
3143                 /* addrconf_add_linklocal also adds a prefix_route and we
3144                  * only need to care about prefix routes if ipv6_generate_eui64
3145                  * couldn't generate one.
3146                  */
3147                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3148                         addrconf_add_linklocal(idev, &addr, 0);
3149                 else if (prefix_route)
3150                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3151                 break;
3152         case IN6_ADDR_GEN_MODE_NONE:
3153         default:
3154                 /* will not add any link local address */
3155                 break;
3156         }
3157 }
3158
3159 static void addrconf_dev_config(struct net_device *dev)
3160 {
3161         struct inet6_dev *idev;
3162
3163         ASSERT_RTNL();
3164
3165         if ((dev->type != ARPHRD_ETHER) &&
3166             (dev->type != ARPHRD_FDDI) &&
3167             (dev->type != ARPHRD_ARCNET) &&
3168             (dev->type != ARPHRD_INFINIBAND) &&
3169             (dev->type != ARPHRD_IEEE1394) &&
3170             (dev->type != ARPHRD_TUNNEL6) &&
3171             (dev->type != ARPHRD_6LOWPAN) &&
3172             (dev->type != ARPHRD_NONE)) {
3173                 /* Alas, we support only Ethernet autoconfiguration. */
3174                 return;
3175         }
3176
3177         idev = addrconf_add_dev(dev);
3178         if (IS_ERR(idev))
3179                 return;
3180
3181         /* this device type has no EUI support */
3182         if (dev->type == ARPHRD_NONE &&
3183             idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3184                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3185
3186         addrconf_addr_gen(idev, false);
3187 }
3188
3189 #if IS_ENABLED(CONFIG_IPV6_SIT)
3190 static void addrconf_sit_config(struct net_device *dev)
3191 {
3192         struct inet6_dev *idev;
3193
3194         ASSERT_RTNL();
3195
3196         /*
3197          * Configure the tunnel with one of our IPv4
3198          * addresses... we should configure all of
3199          * our v4 addrs in the tunnel
3200          */
3201
3202         idev = ipv6_find_idev(dev);
3203         if (!idev) {
3204                 pr_debug("%s: add_dev failed\n", __func__);
3205                 return;
3206         }
3207
3208         if (dev->priv_flags & IFF_ISATAP) {
3209                 addrconf_addr_gen(idev, false);
3210                 return;
3211         }
3212
3213         sit_add_v4_addrs(idev);
3214
3215         if (dev->flags&IFF_POINTOPOINT)
3216                 addrconf_add_mroute(dev);
3217 }
3218 #endif
3219
3220 #if IS_ENABLED(CONFIG_NET_IPGRE)
3221 static void addrconf_gre_config(struct net_device *dev)
3222 {
3223         struct inet6_dev *idev;
3224
3225         ASSERT_RTNL();
3226
3227         idev = ipv6_find_idev(dev);
3228         if (!idev) {
3229                 pr_debug("%s: add_dev failed\n", __func__);
3230                 return;
3231         }
3232
3233         addrconf_addr_gen(idev, true);
3234         if (dev->flags & IFF_POINTOPOINT)
3235                 addrconf_add_mroute(dev);
3236 }
3237 #endif
3238
3239 static int fixup_permanent_addr(struct inet6_dev *idev,
3240                                 struct inet6_ifaddr *ifp)
3241 {
3242         if (!ifp->rt) {
3243                 struct rt6_info *rt;
3244
3245                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3246                 if (unlikely(IS_ERR(rt)))
3247                         return PTR_ERR(rt);
3248
3249                 ifp->rt = rt;
3250         }
3251
3252         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3253                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3254                                       idev->dev, 0, 0);
3255         }
3256
3257         addrconf_dad_start(ifp);
3258
3259         return 0;
3260 }
3261
3262 static void addrconf_permanent_addr(struct net_device *dev)
3263 {
3264         struct inet6_ifaddr *ifp, *tmp;
3265         struct inet6_dev *idev;
3266
3267         idev = __in6_dev_get(dev);
3268         if (!idev)
3269                 return;
3270
3271         write_lock_bh(&idev->lock);
3272
3273         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3274                 if ((ifp->flags & IFA_F_PERMANENT) &&
3275                     fixup_permanent_addr(idev, ifp) < 0) {
3276                         write_unlock_bh(&idev->lock);
3277                         ipv6_del_addr(ifp);
3278                         write_lock_bh(&idev->lock);
3279
3280                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3281                                              idev->dev->name, &ifp->addr);
3282                 }
3283         }
3284
3285         write_unlock_bh(&idev->lock);
3286 }
3287
3288 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3289                            void *ptr)
3290 {
3291         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3292         struct netdev_notifier_changeupper_info *info;
3293         struct inet6_dev *idev = __in6_dev_get(dev);
3294         int run_pending = 0;
3295         int err;
3296
3297         switch (event) {
3298         case NETDEV_REGISTER:
3299                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3300                         idev = ipv6_add_dev(dev);
3301                         if (IS_ERR(idev))
3302                                 return notifier_from_errno(PTR_ERR(idev));
3303                 }
3304                 break;
3305
3306         case NETDEV_CHANGEMTU:
3307                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3308                 if (dev->mtu < IPV6_MIN_MTU) {
3309                         addrconf_ifdown(dev, 1);
3310                         break;
3311                 }
3312
3313                 if (idev) {
3314                         rt6_mtu_change(dev, dev->mtu);
3315                         idev->cnf.mtu6 = dev->mtu;
3316                         break;
3317                 }
3318
3319                 /* allocate new idev */
3320                 idev = ipv6_add_dev(dev);
3321                 if (IS_ERR(idev))
3322                         break;
3323
3324                 /* device is still not ready */
3325                 if (!(idev->if_flags & IF_READY))
3326                         break;
3327
3328                 run_pending = 1;
3329
3330                 /* fall through */
3331
3332         case NETDEV_UP:
3333         case NETDEV_CHANGE:
3334                 if (dev->flags & IFF_SLAVE)
3335                         break;
3336
3337                 if (idev && idev->cnf.disable_ipv6)
3338                         break;
3339
3340                 if (event == NETDEV_UP) {
3341                         /* restore routes for permanent addresses */
3342                         addrconf_permanent_addr(dev);
3343
3344                         if (!addrconf_qdisc_ok(dev)) {
3345                                 /* device is not ready yet. */
3346                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3347                                         dev->name);
3348                                 break;
3349                         }
3350
3351                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3352                                 idev = ipv6_add_dev(dev);
3353
3354                         if (!IS_ERR_OR_NULL(idev)) {
3355                                 idev->if_flags |= IF_READY;
3356                                 run_pending = 1;
3357                         }
3358                 } else if (event == NETDEV_CHANGE) {
3359                         if (!addrconf_qdisc_ok(dev)) {
3360                                 /* device is still not ready. */
3361                                 break;
3362                         }
3363
3364                         if (idev) {
3365                                 if (idev->if_flags & IF_READY)
3366                                         /* device is already configured. */
3367                                         break;
3368                                 idev->if_flags |= IF_READY;
3369                         }
3370
3371                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3372                                 dev->name);
3373
3374                         run_pending = 1;
3375                 }
3376
3377                 switch (dev->type) {
3378 #if IS_ENABLED(CONFIG_IPV6_SIT)
3379                 case ARPHRD_SIT:
3380                         addrconf_sit_config(dev);
3381                         break;
3382 #endif
3383 #if IS_ENABLED(CONFIG_NET_IPGRE)
3384                 case ARPHRD_IPGRE:
3385                         addrconf_gre_config(dev);
3386                         break;
3387 #endif
3388                 case ARPHRD_LOOPBACK:
3389                         init_loopback(dev);
3390                         break;
3391
3392                 default:
3393                         addrconf_dev_config(dev);
3394                         break;
3395                 }
3396
3397                 if (!IS_ERR_OR_NULL(idev)) {
3398                         if (run_pending)
3399                                 addrconf_dad_run(idev);
3400
3401                         /*
3402                          * If the MTU changed during the interface down,
3403                          * when the interface up, the changed MTU must be
3404                          * reflected in the idev as well as routers.
3405                          */
3406                         if (idev->cnf.mtu6 != dev->mtu &&
3407                             dev->mtu >= IPV6_MIN_MTU) {
3408                                 rt6_mtu_change(dev, dev->mtu);
3409                                 idev->cnf.mtu6 = dev->mtu;
3410                         }
3411                         idev->tstamp = jiffies;
3412                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3413
3414                         /*
3415                          * If the changed mtu during down is lower than
3416                          * IPV6_MIN_MTU stop IPv6 on this interface.
3417                          */
3418                         if (dev->mtu < IPV6_MIN_MTU)
3419                                 addrconf_ifdown(dev, 1);
3420                 }
3421                 break;
3422
3423         case NETDEV_DOWN:
3424         case NETDEV_UNREGISTER:
3425                 /*
3426                  *      Remove all addresses from this interface.
3427                  */
3428                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3429                 break;
3430
3431         case NETDEV_CHANGENAME:
3432                 if (idev) {
3433                         snmp6_unregister_dev(idev);
3434                         addrconf_sysctl_unregister(idev);
3435                         err = addrconf_sysctl_register(idev);
3436                         if (err)
3437                                 return notifier_from_errno(err);
3438                         err = snmp6_register_dev(idev);
3439                         if (err) {
3440                                 addrconf_sysctl_unregister(idev);
3441                                 return notifier_from_errno(err);
3442                         }
3443                 }
3444                 break;
3445
3446         case NETDEV_PRE_TYPE_CHANGE:
3447         case NETDEV_POST_TYPE_CHANGE:
3448                 if (idev)
3449                         addrconf_type_change(dev, event);
3450                 break;
3451
3452         case NETDEV_CHANGEUPPER:
3453                 info = ptr;
3454
3455                 /* flush all routes if dev is linked to or unlinked from
3456                  * an L3 master device (e.g., VRF)
3457                  */
3458                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3459                         addrconf_ifdown(dev, 0);
3460         }
3461
3462         return NOTIFY_OK;
3463 }
3464
3465 /*
3466  *      addrconf module should be notified of a device going up
3467  */
3468 static struct notifier_block ipv6_dev_notf = {
3469         .notifier_call = addrconf_notify,
3470 };
3471
3472 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3473 {
3474         struct inet6_dev *idev;
3475         ASSERT_RTNL();
3476
3477         idev = __in6_dev_get(dev);
3478
3479         if (event == NETDEV_POST_TYPE_CHANGE)
3480                 ipv6_mc_remap(idev);
3481         else if (event == NETDEV_PRE_TYPE_CHANGE)
3482                 ipv6_mc_unmap(idev);
3483 }
3484
3485 static bool addr_is_local(const struct in6_addr *addr)
3486 {
3487         return ipv6_addr_type(addr) &
3488                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3489 }
3490
3491 static int addrconf_ifdown(struct net_device *dev, int how)
3492 {
3493         struct net *net = dev_net(dev);
3494         struct inet6_dev *idev;
3495         struct inet6_ifaddr *ifa, *tmp;
3496         struct list_head del_list;
3497         int _keep_addr;
3498         bool keep_addr;
3499         int state, i;
3500
3501         ASSERT_RTNL();
3502
3503         rt6_ifdown(net, dev);
3504         neigh_ifdown(&nd_tbl, dev);
3505
3506         idev = __in6_dev_get(dev);
3507         if (!idev)
3508                 return -ENODEV;
3509
3510         /*
3511          * Step 1: remove reference to ipv6 device from parent device.
3512          *         Do not dev_put!
3513          */
3514         if (how) {
3515                 idev->dead = 1;
3516
3517                 /* protected by rtnl_lock */
3518                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3519
3520                 /* Step 1.5: remove snmp6 entry */
3521                 snmp6_unregister_dev(idev);
3522
3523         }
3524
3525         /* aggregate the system setting and interface setting */
3526         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3527         if (!_keep_addr)
3528                 _keep_addr = idev->cnf.keep_addr_on_down;
3529
3530         /* combine the user config with event to determine if permanent
3531          * addresses are to be removed from address hash table
3532          */
3533         keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3534
3535         /* Step 2: clear hash table */
3536         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3537                 struct hlist_head *h = &inet6_addr_lst[i];
3538
3539                 spin_lock_bh(&addrconf_hash_lock);
3540 restart:
3541                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3542                         if (ifa->idev == idev) {
3543                                 addrconf_del_dad_work(ifa);
3544                                 /* combined flag + permanent flag decide if
3545                                  * address is retained on a down event
3546                                  */
3547                                 if (!keep_addr ||
3548                                     !(ifa->flags & IFA_F_PERMANENT) ||
3549                                     addr_is_local(&ifa->addr)) {
3550                                         hlist_del_init_rcu(&ifa->addr_lst);
3551                                         goto restart;
3552                                 }
3553                         }
3554                 }
3555                 spin_unlock_bh(&addrconf_hash_lock);
3556         }
3557
3558         write_lock_bh(&idev->lock);
3559
3560         addrconf_del_rs_timer(idev);
3561
3562         /* Step 2: clear flags for stateless addrconf */
3563         if (!how)
3564                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3565
3566         if (how && del_timer(&idev->regen_timer))
3567                 in6_dev_put(idev);
3568
3569         /* Step 3: clear tempaddr list */
3570         while (!list_empty(&idev->tempaddr_list)) {
3571                 ifa = list_first_entry(&idev->tempaddr_list,
3572                                        struct inet6_ifaddr, tmp_list);
3573                 list_del(&ifa->tmp_list);
3574                 write_unlock_bh(&idev->lock);
3575                 spin_lock_bh(&ifa->lock);
3576
3577                 if (ifa->ifpub) {
3578                         in6_ifa_put(ifa->ifpub);
3579                         ifa->ifpub = NULL;
3580                 }
3581                 spin_unlock_bh(&ifa->lock);
3582                 in6_ifa_put(ifa);
3583                 write_lock_bh(&idev->lock);
3584         }
3585
3586         /* re-combine the user config with event to determine if permanent
3587          * addresses are to be removed from the interface list
3588          */
3589         keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3590
3591         INIT_LIST_HEAD(&del_list);
3592         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3593                 struct rt6_info *rt = NULL;
3594
3595                 addrconf_del_dad_work(ifa);
3596
3597                 write_unlock_bh(&idev->lock);
3598                 spin_lock_bh(&ifa->lock);
3599
3600                 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3601                     !addr_is_local(&ifa->addr)) {
3602                         /* set state to skip the notifier below */
3603                         state = INET6_IFADDR_STATE_DEAD;
3604                         ifa->state = 0;
3605                         if (!(ifa->flags & IFA_F_NODAD))
3606                                 ifa->flags |= IFA_F_TENTATIVE;
3607
3608                         rt = ifa->rt;
3609                         ifa->rt = NULL;
3610                 } else {
3611                         state = ifa->state;
3612                         ifa->state = INET6_IFADDR_STATE_DEAD;
3613
3614                         list_move(&ifa->if_list, &del_list);
3615                 }
3616
3617                 spin_unlock_bh(&ifa->lock);
3618
3619                 if (rt)
3620                         ip6_del_rt(rt);
3621
3622                 if (state != INET6_IFADDR_STATE_DEAD) {
3623                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3624                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3625                 } else {
3626                         if (idev->cnf.forwarding)
3627                                 addrconf_leave_anycast(ifa);
3628                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3629                 }
3630
3631                 write_lock_bh(&idev->lock);
3632         }
3633
3634         write_unlock_bh(&idev->lock);
3635
3636         /* now clean up addresses to be removed */
3637         while (!list_empty(&del_list)) {
3638                 ifa = list_first_entry(&del_list,
3639                                        struct inet6_ifaddr, if_list);
3640                 list_del(&ifa->if_list);
3641
3642                 in6_ifa_put(ifa);
3643         }
3644
3645         /* Step 5: Discard anycast and multicast list */
3646         if (how) {
3647                 ipv6_ac_destroy_dev(idev);
3648                 ipv6_mc_destroy_dev(idev);
3649         } else {
3650                 ipv6_mc_down(idev);
3651         }
3652
3653         idev->tstamp = jiffies;
3654
3655         /* Last: Shot the device (if unregistered) */
3656         if (how) {
3657                 addrconf_sysctl_unregister(idev);
3658                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3659                 neigh_ifdown(&nd_tbl, dev);
3660                 in6_dev_put(idev);
3661         }
3662         return 0;
3663 }
3664
3665 static void addrconf_rs_timer(unsigned long data)
3666 {
3667         struct inet6_dev *idev = (struct inet6_dev *)data;
3668         struct net_device *dev = idev->dev;
3669         struct in6_addr lladdr;
3670
3671         write_lock(&idev->lock);
3672         if (idev->dead || !(idev->if_flags & IF_READY))
3673                 goto out;
3674
3675         if (!ipv6_accept_ra(idev))
3676                 goto out;
3677
3678         /* Announcement received after solicitation was sent */
3679         if (idev->if_flags & IF_RA_RCVD)
3680                 goto out;
3681
3682         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3683                 write_unlock(&idev->lock);
3684                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3685                         ndisc_send_rs(dev, &lladdr,
3686                                       &in6addr_linklocal_allrouters);
3687                 else
3688                         goto put;
3689
3690                 write_lock(&idev->lock);
3691                 /* The wait after the last probe can be shorter */
3692                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3693                                              idev->cnf.rtr_solicits) ?
3694                                       idev->cnf.rtr_solicit_delay :
3695                                       idev->cnf.rtr_solicit_interval);
3696         } else {
3697                 /*
3698                  * Note: we do not support deprecated "all on-link"
3699                  * assumption any longer.
3700                  */
3701                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3702         }
3703
3704 out:
3705         write_unlock(&idev->lock);
3706 put:
3707         in6_dev_put(idev);
3708 }
3709
3710 /*
3711  *      Duplicate Address Detection
3712  */
3713 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3714 {
3715         unsigned long rand_num;
3716         struct inet6_dev *idev = ifp->idev;
3717
3718         if (ifp->flags & IFA_F_OPTIMISTIC)
3719                 rand_num = 0;
3720         else
3721                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3722
3723         ifp->dad_probes = idev->cnf.dad_transmits;
3724         addrconf_mod_dad_work(ifp, rand_num);
3725 }
3726
3727 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3728 {
3729         struct inet6_dev *idev = ifp->idev;
3730         struct net_device *dev = idev->dev;
3731         bool notify = false;
3732
3733         addrconf_join_solict(dev, &ifp->addr);
3734
3735         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3736
3737         read_lock_bh(&idev->lock);
3738         spin_lock(&ifp->lock);
3739         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3740                 goto out;
3741
3742         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3743             idev->cnf.accept_dad < 1 ||
3744             !(ifp->flags&IFA_F_TENTATIVE) ||
3745             ifp->flags & IFA_F_NODAD) {
3746                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3747                 spin_unlock(&ifp->lock);
3748                 read_unlock_bh(&idev->lock);
3749
3750                 addrconf_dad_completed(ifp);
3751                 return;
3752         }
3753
3754         if (!(idev->if_flags & IF_READY)) {
3755                 spin_unlock(&ifp->lock);
3756                 read_unlock_bh(&idev->lock);
3757                 /*
3758                  * If the device is not ready:
3759                  * - keep it tentative if it is a permanent address.
3760                  * - otherwise, kill it.
3761                  */
3762                 in6_ifa_hold(ifp);
3763                 addrconf_dad_stop(ifp, 0);
3764                 return;
3765         }
3766
3767         /*
3768          * Optimistic nodes can start receiving
3769          * Frames right away
3770          */
3771         if (ifp->flags & IFA_F_OPTIMISTIC) {
3772                 ip6_ins_rt(ifp->rt);
3773                 if (ipv6_use_optimistic_addr(idev)) {
3774                         /* Because optimistic nodes can use this address,
3775                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3776                          */
3777                         notify = true;
3778                 }
3779         }
3780
3781         addrconf_dad_kick(ifp);
3782 out:
3783         spin_unlock(&ifp->lock);
3784         read_unlock_bh(&idev->lock);
3785         if (notify)
3786                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3787 }
3788
3789 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3790 {
3791         bool begin_dad = false;
3792
3793         spin_lock_bh(&ifp->lock);
3794         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3795                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3796                 begin_dad = true;
3797         }
3798         spin_unlock_bh(&ifp->lock);
3799
3800         if (begin_dad)
3801                 addrconf_mod_dad_work(ifp, 0);
3802 }
3803
3804 static void addrconf_dad_work(struct work_struct *w)
3805 {
3806         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3807                                                 struct inet6_ifaddr,
3808                                                 dad_work);
3809         struct inet6_dev *idev = ifp->idev;
3810         struct in6_addr mcaddr;
3811         bool disable_ipv6 = false;
3812
3813         enum {
3814                 DAD_PROCESS,
3815                 DAD_BEGIN,
3816                 DAD_ABORT,
3817         } action = DAD_PROCESS;
3818
3819         rtnl_lock();
3820
3821         spin_lock_bh(&ifp->lock);
3822         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3823                 action = DAD_BEGIN;
3824                 ifp->state = INET6_IFADDR_STATE_DAD;
3825         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3826                 action = DAD_ABORT;
3827                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3828
3829                 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3830                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3831                         struct in6_addr addr;
3832
3833                         addr.s6_addr32[0] = htonl(0xfe800000);
3834                         addr.s6_addr32[1] = 0;
3835
3836                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3837                             ipv6_addr_equal(&ifp->addr, &addr)) {
3838                                 /* DAD failed for link-local based on MAC */
3839                                 idev->cnf.disable_ipv6 = 1;
3840
3841                                 pr_info("%s: IPv6 being disabled!\n",
3842                                         ifp->idev->dev->name);
3843                                 disable_ipv6 = true;
3844                         }
3845                 }
3846         }
3847         spin_unlock_bh(&ifp->lock);
3848
3849         if (action == DAD_BEGIN) {
3850                 addrconf_dad_begin(ifp);
3851                 goto out;
3852         } else if (action == DAD_ABORT) {
3853                 addrconf_dad_stop(ifp, 1);
3854                 if (disable_ipv6)
3855                         addrconf_ifdown(idev->dev, 0);
3856                 goto out;
3857         }
3858
3859         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3860                 goto out;
3861
3862         write_lock_bh(&idev->lock);
3863         if (idev->dead || !(idev->if_flags & IF_READY)) {
3864                 write_unlock_bh(&idev->lock);
3865                 goto out;
3866         }
3867
3868         spin_lock(&ifp->lock);
3869         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3870                 spin_unlock(&ifp->lock);
3871                 write_unlock_bh(&idev->lock);
3872                 goto out;
3873         }
3874
3875         if (ifp->dad_probes == 0) {
3876                 /*
3877                  * DAD was successful
3878                  */
3879
3880                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3881                 spin_unlock(&ifp->lock);
3882                 write_unlock_bh(&idev->lock);
3883
3884                 addrconf_dad_completed(ifp);
3885
3886                 goto out;
3887         }
3888
3889         ifp->dad_probes--;
3890         addrconf_mod_dad_work(ifp,
3891                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3892         spin_unlock(&ifp->lock);
3893         write_unlock_bh(&idev->lock);
3894
3895         /* send a neighbour solicitation for our addr */
3896         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3897         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3898 out:
3899         in6_ifa_put(ifp);
3900         rtnl_unlock();
3901 }
3902
3903 /* ifp->idev must be at least read locked */
3904 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3905 {
3906         struct inet6_ifaddr *ifpiter;
3907         struct inet6_dev *idev = ifp->idev;
3908
3909         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3910                 if (ifpiter->scope > IFA_LINK)
3911                         break;
3912                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3913                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3914                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3915                     IFA_F_PERMANENT)
3916                         return false;
3917         }
3918         return true;
3919 }
3920
3921 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3922 {
3923         struct net_device *dev = ifp->idev->dev;
3924         struct in6_addr lladdr;
3925         bool send_rs, send_mld;
3926
3927         addrconf_del_dad_work(ifp);
3928
3929         /*
3930          *      Configure the address for reception. Now it is valid.
3931          */
3932
3933         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3934
3935         /* If added prefix is link local and we are prepared to process
3936            router advertisements, start sending router solicitations.
3937          */
3938
3939         read_lock_bh(&ifp->idev->lock);
3940         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3941         send_rs = send_mld &&
3942                   ipv6_accept_ra(ifp->idev) &&
3943                   ifp->idev->cnf.rtr_solicits > 0 &&
3944                   (dev->flags&IFF_LOOPBACK) == 0;
3945         read_unlock_bh(&ifp->idev->lock);
3946
3947         /* While dad is in progress mld report's source address is in6_addrany.
3948          * Resend with proper ll now.
3949          */
3950         if (send_mld)
3951                 ipv6_mc_dad_complete(ifp->idev);
3952
3953         if (send_rs) {
3954                 /*
3955                  *      If a host as already performed a random delay
3956                  *      [...] as part of DAD [...] there is no need
3957                  *      to delay again before sending the first RS
3958                  */
3959                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3960                         return;
3961                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3962
3963                 write_lock_bh(&ifp->idev->lock);
3964                 spin_lock(&ifp->lock);
3965                 ifp->idev->rs_probes = 1;
3966                 ifp->idev->if_flags |= IF_RS_SENT;
3967                 addrconf_mod_rs_timer(ifp->idev,
3968                                       ifp->idev->cnf.rtr_solicit_interval);
3969                 spin_unlock(&ifp->lock);
3970                 write_unlock_bh(&ifp->idev->lock);
3971         }
3972 }
3973
3974 static void addrconf_dad_run(struct inet6_dev *idev)
3975 {
3976         struct inet6_ifaddr *ifp;
3977
3978         read_lock_bh(&idev->lock);
3979         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3980                 spin_lock(&ifp->lock);
3981                 if (ifp->flags & IFA_F_TENTATIVE &&
3982                     ifp->state == INET6_IFADDR_STATE_DAD)
3983                         addrconf_dad_kick(ifp);
3984                 spin_unlock(&ifp->lock);
3985         }
3986         read_unlock_bh(&idev->lock);
3987 }
3988
3989 #ifdef CONFIG_PROC_FS
3990 struct if6_iter_state {
3991         struct seq_net_private p;
3992         int bucket;
3993         int offset;
3994 };
3995
3996 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3997 {
3998         struct inet6_ifaddr *ifa = NULL;
3999         struct if6_iter_state *state = seq->private;
4000         struct net *net = seq_file_net(seq);
4001         int p = 0;
4002
4003         /* initial bucket if pos is 0 */
4004         if (pos == 0) {
4005                 state->bucket = 0;
4006                 state->offset = 0;
4007         }
4008
4009         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4010                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4011                                          addr_lst) {
4012                         if (!net_eq(dev_net(ifa->idev->dev), net))
4013                                 continue;
4014                         /* sync with offset */
4015                         if (p < state->offset) {
4016                                 p++;
4017                                 continue;
4018                         }
4019                         state->offset++;
4020                         return ifa;
4021                 }
4022
4023                 /* prepare for next bucket */
4024                 state->offset = 0;
4025                 p = 0;
4026         }
4027         return NULL;
4028 }
4029
4030 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4031                                          struct inet6_ifaddr *ifa)
4032 {
4033         struct if6_iter_state *state = seq->private;
4034         struct net *net = seq_file_net(seq);
4035
4036         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4037                 if (!net_eq(dev_net(ifa->idev->dev), net))
4038                         continue;
4039                 state->offset++;
4040                 return ifa;
4041         }
4042
4043         while (++state->bucket < IN6_ADDR_HSIZE) {
4044                 state->offset = 0;
4045                 hlist_for_each_entry_rcu_bh(ifa,
4046                                      &inet6_addr_lst[state->bucket], addr_lst) {
4047                         if (!net_eq(dev_net(ifa->idev->dev), net))
4048                                 continue;
4049                         state->offset++;
4050                         return ifa;
4051                 }
4052         }
4053
4054         return NULL;
4055 }
4056
4057 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4058         __acquires(rcu_bh)
4059 {
4060         rcu_read_lock_bh();
4061         return if6_get_first(seq, *pos);
4062 }
4063
4064 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4065 {
4066         struct inet6_ifaddr *ifa;
4067
4068         ifa = if6_get_next(seq, v);
4069         ++*pos;
4070         return ifa;
4071 }
4072
4073 static void if6_seq_stop(struct seq_file *seq, void *v)
4074         __releases(rcu_bh)
4075 {
4076         rcu_read_unlock_bh();
4077 }
4078
4079 static int if6_seq_show(struct seq_file *seq, void *v)
4080 {
4081         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4082         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4083                    &ifp->addr,
4084                    ifp->idev->dev->ifindex,
4085                    ifp->prefix_len,
4086                    ifp->scope,
4087                    (u8) ifp->flags,
4088                    ifp->idev->dev->name);
4089         return 0;
4090 }
4091
4092 static const struct seq_operations if6_seq_ops = {
4093         .start  = if6_seq_start,
4094         .next   = if6_seq_next,
4095         .show   = if6_seq_show,
4096         .stop   = if6_seq_stop,
4097 };
4098
4099 static int if6_seq_open(struct inode *inode, struct file *file)
4100 {
4101         return seq_open_net(inode, file, &if6_seq_ops,
4102                             sizeof(struct if6_iter_state));
4103 }
4104
4105 static const struct file_operations if6_fops = {
4106         .owner          = THIS_MODULE,
4107         .open           = if6_seq_open,
4108         .read           = seq_read,
4109         .llseek         = seq_lseek,
4110         .release        = seq_release_net,
4111 };
4112
4113 static int __net_init if6_proc_net_init(struct net *net)
4114 {
4115         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4116                 return -ENOMEM;
4117         return 0;
4118 }
4119
4120 static void __net_exit if6_proc_net_exit(struct net *net)
4121 {
4122         remove_proc_entry("if_inet6", net->proc_net);
4123 }
4124
4125 static struct pernet_operations if6_proc_net_ops = {
4126         .init = if6_proc_net_init,
4127         .exit = if6_proc_net_exit,
4128 };
4129
4130 int __init if6_proc_init(void)
4131 {
4132         return register_pernet_subsys(&if6_proc_net_ops);
4133 }
4134
4135 void if6_proc_exit(void)
4136 {
4137         unregister_pernet_subsys(&if6_proc_net_ops);
4138 }
4139 #endif  /* CONFIG_PROC_FS */
4140
4141 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4142 /* Check if address is a home address configured on any interface. */
4143 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4144 {
4145         int ret = 0;
4146         struct inet6_ifaddr *ifp = NULL;
4147         unsigned int hash = inet6_addr_hash(addr);
4148
4149         rcu_read_lock_bh();
4150         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4151                 if (!net_eq(dev_net(ifp->idev->dev), net))
4152                         continue;
4153                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4154                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4155                         ret = 1;
4156                         break;
4157                 }
4158         }
4159         rcu_read_unlock_bh();
4160         return ret;
4161 }
4162 #endif
4163
4164 /*
4165  *      Periodic address status verification
4166  */
4167
4168 static void addrconf_verify_rtnl(void)
4169 {
4170         unsigned long now, next, next_sec, next_sched;
4171         struct inet6_ifaddr *ifp;
4172         int i;
4173
4174         ASSERT_RTNL();
4175
4176         rcu_read_lock_bh();
4177         now = jiffies;
4178         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4179
4180         cancel_delayed_work(&addr_chk_work);
4181
4182         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4183 restart:
4184                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4185                         unsigned long age;
4186
4187                         /* When setting preferred_lft to a value not zero or
4188                          * infinity, while valid_lft is infinity
4189                          * IFA_F_PERMANENT has a non-infinity life time.
4190                          */
4191                         if ((ifp->flags & IFA_F_PERMANENT) &&
4192                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4193                                 continue;
4194
4195                         spin_lock(&ifp->lock);
4196                         /* We try to batch several events at once. */
4197                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4198
4199                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4200                             age >= ifp->valid_lft) {
4201                                 spin_unlock(&ifp->lock);
4202                                 in6_ifa_hold(ifp);
4203                                 ipv6_del_addr(ifp);
4204                                 goto restart;
4205                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4206                                 spin_unlock(&ifp->lock);
4207                                 continue;
4208                         } else if (age >= ifp->prefered_lft) {
4209                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4210                                 int deprecate = 0;
4211
4212                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4213                                         deprecate = 1;
4214                                         ifp->flags |= IFA_F_DEPRECATED;
4215                                 }
4216
4217                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4218                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4219                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4220
4221                                 spin_unlock(&ifp->lock);
4222
4223                                 if (deprecate) {
4224                                         in6_ifa_hold(ifp);
4225
4226                                         ipv6_ifa_notify(0, ifp);
4227                                         in6_ifa_put(ifp);
4228                                         goto restart;
4229                                 }
4230                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4231                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4232                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4233                                         ifp->idev->cnf.dad_transmits *
4234                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4235
4236                                 if (age >= ifp->prefered_lft - regen_advance) {
4237                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4238                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4239                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4240                                         if (!ifp->regen_count && ifpub) {
4241                                                 ifp->regen_count++;
4242                                                 in6_ifa_hold(ifp);
4243                                                 in6_ifa_hold(ifpub);
4244                                                 spin_unlock(&ifp->lock);
4245
4246                                                 spin_lock(&ifpub->lock);
4247                                                 ifpub->regen_count = 0;
4248                                                 spin_unlock(&ifpub->lock);
4249                                                 ipv6_create_tempaddr(ifpub, ifp);
4250                                                 in6_ifa_put(ifpub);
4251                                                 in6_ifa_put(ifp);
4252                                                 goto restart;
4253                                         }
4254                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4255                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4256                                 spin_unlock(&ifp->lock);
4257                         } else {
4258                                 /* ifp->prefered_lft <= ifp->valid_lft */
4259                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4260                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4261                                 spin_unlock(&ifp->lock);
4262                         }
4263                 }
4264         }
4265
4266         next_sec = round_jiffies_up(next);
4267         next_sched = next;
4268
4269         /* If rounded timeout is accurate enough, accept it. */
4270         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4271                 next_sched = next_sec;
4272
4273         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4274         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4275                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4276
4277         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4278               now, next, next_sec, next_sched);
4279         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4280         rcu_read_unlock_bh();
4281 }
4282
4283 static void addrconf_verify_work(struct work_struct *w)
4284 {
4285         rtnl_lock();
4286         addrconf_verify_rtnl();
4287         rtnl_unlock();
4288 }
4289
4290 static void addrconf_verify(void)
4291 {
4292         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4293 }
4294
4295 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4296                                      struct in6_addr **peer_pfx)
4297 {
4298         struct in6_addr *pfx = NULL;
4299
4300         *peer_pfx = NULL;
4301
4302         if (addr)
4303                 pfx = nla_data(addr);
4304
4305         if (local) {
4306                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4307                         *peer_pfx = pfx;
4308                 pfx = nla_data(local);
4309         }
4310
4311         return pfx;
4312 }
4313
4314 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4315         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4316         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4317         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4318         [IFA_FLAGS]             = { .len = sizeof(u32) },
4319 };
4320
4321 static int
4322 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4323 {
4324         struct net *net = sock_net(skb->sk);
4325         struct ifaddrmsg *ifm;
4326         struct nlattr *tb[IFA_MAX+1];
4327         struct in6_addr *pfx, *peer_pfx;
4328         u32 ifa_flags;
4329         int err;
4330
4331         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4332         if (err < 0)
4333                 return err;
4334
4335         ifm = nlmsg_data(nlh);
4336         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4337         if (!pfx)
4338                 return -EINVAL;
4339
4340         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4341
4342         /* We ignore other flags so far. */
4343         ifa_flags &= IFA_F_MANAGETEMPADDR;
4344
4345         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4346                               ifm->ifa_prefixlen);
4347 }
4348
4349 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4350                              u32 prefered_lft, u32 valid_lft)
4351 {
4352         u32 flags;
4353         clock_t expires;
4354         unsigned long timeout;
4355         bool was_managetempaddr;
4356         bool had_prefixroute;
4357
4358         ASSERT_RTNL();
4359
4360         if (!valid_lft || (prefered_lft > valid_lft))
4361                 return -EINVAL;
4362
4363         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4364             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4365                 return -EINVAL;
4366
4367         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4368         if (addrconf_finite_timeout(timeout)) {
4369                 expires = jiffies_to_clock_t(timeout * HZ);
4370                 valid_lft = timeout;
4371                 flags = RTF_EXPIRES;
4372         } else {
4373                 expires = 0;
4374                 flags = 0;
4375                 ifa_flags |= IFA_F_PERMANENT;
4376         }
4377
4378         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4379         if (addrconf_finite_timeout(timeout)) {
4380                 if (timeout == 0)
4381                         ifa_flags |= IFA_F_DEPRECATED;
4382                 prefered_lft = timeout;
4383         }
4384
4385         spin_lock_bh(&ifp->lock);
4386         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4387         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4388                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4389         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4390                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4391                         IFA_F_NOPREFIXROUTE);
4392         ifp->flags |= ifa_flags;
4393         ifp->tstamp = jiffies;
4394         ifp->valid_lft = valid_lft;
4395         ifp->prefered_lft = prefered_lft;
4396
4397         spin_unlock_bh(&ifp->lock);
4398         if (!(ifp->flags&IFA_F_TENTATIVE))
4399                 ipv6_ifa_notify(0, ifp);
4400
4401         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4402                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4403                                       expires, flags);
4404         } else if (had_prefixroute) {
4405                 enum cleanup_prefix_rt_t action;
4406                 unsigned long rt_expires;
4407
4408                 write_lock_bh(&ifp->idev->lock);
4409                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4410                 write_unlock_bh(&ifp->idev->lock);
4411
4412                 if (action != CLEANUP_PREFIX_RT_NOP) {
4413                         cleanup_prefix_route(ifp, rt_expires,
4414                                 action == CLEANUP_PREFIX_RT_DEL);
4415                 }
4416         }
4417
4418         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4419                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4420                         valid_lft = prefered_lft = 0;
4421                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4422                                  !was_managetempaddr, jiffies);
4423         }
4424
4425         addrconf_verify_rtnl();
4426
4427         return 0;
4428 }
4429
4430 static int
4431 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4432 {
4433         struct net *net = sock_net(skb->sk);
4434         struct ifaddrmsg *ifm;
4435         struct nlattr *tb[IFA_MAX+1];
4436         struct in6_addr *pfx, *peer_pfx;
4437         struct inet6_ifaddr *ifa;
4438         struct net_device *dev;
4439         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4440         u32 ifa_flags;
4441         int err;
4442
4443         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4444         if (err < 0)
4445                 return err;
4446
4447         ifm = nlmsg_data(nlh);
4448         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4449         if (!pfx)
4450                 return -EINVAL;
4451
4452         if (tb[IFA_CACHEINFO]) {
4453                 struct ifa_cacheinfo *ci;
4454
4455                 ci = nla_data(tb[IFA_CACHEINFO]);
4456                 valid_lft = ci->ifa_valid;
4457                 preferred_lft = ci->ifa_prefered;
4458         } else {
4459                 preferred_lft = INFINITY_LIFE_TIME;
4460                 valid_lft = INFINITY_LIFE_TIME;
4461         }
4462
4463         dev =  __dev_get_by_index(net, ifm->ifa_index);
4464         if (!dev)
4465                 return -ENODEV;
4466
4467         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4468
4469         /* We ignore other flags so far. */
4470         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4471                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4472
4473         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4474         if (!ifa) {
4475                 /*
4476                  * It would be best to check for !NLM_F_CREATE here but
4477                  * userspace already relies on not having to provide this.
4478                  */
4479                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4480                                       ifm->ifa_prefixlen, ifa_flags,
4481                                       preferred_lft, valid_lft);
4482         }
4483
4484         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4485             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4486                 err = -EEXIST;
4487         else
4488                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4489
4490         in6_ifa_put(ifa);
4491
4492         return err;
4493 }
4494
4495 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4496                           u8 scope, int ifindex)
4497 {
4498         struct ifaddrmsg *ifm;
4499
4500         ifm = nlmsg_data(nlh);
4501         ifm->ifa_family = AF_INET6;
4502         ifm->ifa_prefixlen = prefixlen;
4503         ifm->ifa_flags = flags;
4504         ifm->ifa_scope = scope;
4505         ifm->ifa_index = ifindex;
4506 }
4507
4508 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4509                          unsigned long tstamp, u32 preferred, u32 valid)
4510 {
4511         struct ifa_cacheinfo ci;
4512
4513         ci.cstamp = cstamp_delta(cstamp);
4514         ci.tstamp = cstamp_delta(tstamp);
4515         ci.ifa_prefered = preferred;
4516         ci.ifa_valid = valid;
4517
4518         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4519 }
4520
4521 static inline int rt_scope(int ifa_scope)
4522 {
4523         if (ifa_scope & IFA_HOST)
4524                 return RT_SCOPE_HOST;
4525         else if (ifa_scope & IFA_LINK)
4526                 return RT_SCOPE_LINK;
4527         else if (ifa_scope & IFA_SITE)
4528                 return RT_SCOPE_SITE;
4529         else
4530                 return RT_SCOPE_UNIVERSE;
4531 }
4532
4533 static inline int inet6_ifaddr_msgsize(void)
4534 {
4535         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4536                + nla_total_size(16) /* IFA_LOCAL */
4537                + nla_total_size(16) /* IFA_ADDRESS */
4538                + nla_total_size(sizeof(struct ifa_cacheinfo))
4539                + nla_total_size(4)  /* IFA_FLAGS */;
4540 }
4541
4542 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4543                              u32 portid, u32 seq, int event, unsigned int flags)
4544 {
4545         struct nlmsghdr  *nlh;
4546         u32 preferred, valid;
4547
4548         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4549         if (!nlh)
4550                 return -EMSGSIZE;
4551
4552         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4553                       ifa->idev->dev->ifindex);
4554
4555         if (!((ifa->flags&IFA_F_PERMANENT) &&
4556               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4557                 preferred = ifa->prefered_lft;
4558                 valid = ifa->valid_lft;
4559                 if (preferred != INFINITY_LIFE_TIME) {
4560                         long tval = (jiffies - ifa->tstamp)/HZ;
4561                         if (preferred > tval)
4562                                 preferred -= tval;
4563                         else
4564                                 preferred = 0;
4565                         if (valid != INFINITY_LIFE_TIME) {
4566                                 if (valid > tval)
4567                                         valid -= tval;
4568                                 else
4569                                         valid = 0;
4570                         }
4571                 }
4572         } else {
4573                 preferred = INFINITY_LIFE_TIME;
4574                 valid = INFINITY_LIFE_TIME;
4575         }
4576
4577         if (!ipv6_addr_any(&ifa->peer_addr)) {
4578                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4579                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4580                         goto error;
4581         } else
4582                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4583                         goto error;
4584
4585         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4586                 goto error;
4587
4588         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4589                 goto error;
4590
4591         nlmsg_end(skb, nlh);
4592         return 0;
4593
4594 error:
4595         nlmsg_cancel(skb, nlh);
4596         return -EMSGSIZE;
4597 }
4598
4599 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4600                                 u32 portid, u32 seq, int event, u16 flags)
4601 {
4602         struct nlmsghdr  *nlh;
4603         u8 scope = RT_SCOPE_UNIVERSE;
4604         int ifindex = ifmca->idev->dev->ifindex;
4605
4606         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4607                 scope = RT_SCOPE_SITE;
4608
4609         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4610         if (!nlh)
4611                 return -EMSGSIZE;
4612
4613         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4614         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4615             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4616                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4617                 nlmsg_cancel(skb, nlh);
4618                 return -EMSGSIZE;
4619         }
4620
4621         nlmsg_end(skb, nlh);
4622         return 0;
4623 }
4624
4625 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4626                                 u32 portid, u32 seq, int event, unsigned int flags)
4627 {
4628         struct nlmsghdr  *nlh;
4629         u8 scope = RT_SCOPE_UNIVERSE;
4630         int ifindex = ifaca->aca_idev->dev->ifindex;
4631
4632         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4633                 scope = RT_SCOPE_SITE;
4634
4635         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4636         if (!nlh)
4637                 return -EMSGSIZE;
4638
4639         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4640         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4641             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4642                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4643                 nlmsg_cancel(skb, nlh);
4644                 return -EMSGSIZE;
4645         }
4646
4647         nlmsg_end(skb, nlh);
4648         return 0;
4649 }
4650
4651 enum addr_type_t {
4652         UNICAST_ADDR,
4653         MULTICAST_ADDR,
4654         ANYCAST_ADDR,
4655 };
4656
4657 /* called with rcu_read_lock() */
4658 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4659                           struct netlink_callback *cb, enum addr_type_t type,
4660                           int s_ip_idx, int *p_ip_idx)
4661 {
4662         struct ifmcaddr6 *ifmca;
4663         struct ifacaddr6 *ifaca;
4664         int err = 1;
4665         int ip_idx = *p_ip_idx;
4666
4667         read_lock_bh(&idev->lock);
4668         switch (type) {
4669         case UNICAST_ADDR: {
4670                 struct inet6_ifaddr *ifa;
4671
4672                 /* unicast address incl. temp addr */
4673                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4674                         if (++ip_idx < s_ip_idx)
4675                                 continue;
4676                         err = inet6_fill_ifaddr(skb, ifa,
4677                                                 NETLINK_CB(cb->skb).portid,
4678                                                 cb->nlh->nlmsg_seq,
4679                                                 RTM_NEWADDR,
4680                                                 NLM_F_MULTI);
4681                         if (err < 0)
4682                                 break;
4683                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4684                 }
4685                 break;
4686         }
4687         case MULTICAST_ADDR:
4688                 /* multicast address */
4689                 for (ifmca = idev->mc_list; ifmca;
4690                      ifmca = ifmca->next, ip_idx++) {
4691                         if (ip_idx < s_ip_idx)
4692                                 continue;
4693                         err = inet6_fill_ifmcaddr(skb, ifmca,
4694                                                   NETLINK_CB(cb->skb).portid,
4695                                                   cb->nlh->nlmsg_seq,
4696                                                   RTM_GETMULTICAST,
4697                                                   NLM_F_MULTI);
4698                         if (err < 0)
4699                                 break;
4700                 }
4701                 break;
4702         case ANYCAST_ADDR:
4703                 /* anycast address */
4704                 for (ifaca = idev->ac_list; ifaca;
4705                      ifaca = ifaca->aca_next, ip_idx++) {
4706                         if (ip_idx < s_ip_idx)
4707                                 continue;
4708                         err = inet6_fill_ifacaddr(skb, ifaca,
4709                                                   NETLINK_CB(cb->skb).portid,
4710                                                   cb->nlh->nlmsg_seq,
4711                                                   RTM_GETANYCAST,
4712                                                   NLM_F_MULTI);
4713                         if (err < 0)
4714                                 break;
4715                 }
4716                 break;
4717         default:
4718                 break;
4719         }
4720         read_unlock_bh(&idev->lock);
4721         *p_ip_idx = ip_idx;
4722         return err;
4723 }
4724
4725 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4726                            enum addr_type_t type)
4727 {
4728         struct net *net = sock_net(skb->sk);
4729         int h, s_h;
4730         int idx, ip_idx;
4731         int s_idx, s_ip_idx;
4732         struct net_device *dev;
4733         struct inet6_dev *idev;
4734         struct hlist_head *head;
4735
4736         s_h = cb->args[0];
4737         s_idx = idx = cb->args[1];
4738         s_ip_idx = ip_idx = cb->args[2];
4739
4740         rcu_read_lock();
4741         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4742         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4743                 idx = 0;
4744                 head = &net->dev_index_head[h];
4745                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4746                         if (idx < s_idx)
4747                                 goto cont;
4748                         if (h > s_h || idx > s_idx)
4749                                 s_ip_idx = 0;
4750                         ip_idx = 0;
4751                         idev = __in6_dev_get(dev);
4752                         if (!idev)
4753                                 goto cont;
4754
4755                         if (in6_dump_addrs(idev, skb, cb, type,
4756                                            s_ip_idx, &ip_idx) < 0)
4757                                 goto done;
4758 cont:
4759                         idx++;
4760                 }
4761         }
4762 done:
4763         rcu_read_unlock();
4764         cb->args[0] = h;
4765         cb->args[1] = idx;
4766         cb->args[2] = ip_idx;
4767
4768         return skb->len;
4769 }
4770
4771 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4772 {
4773         enum addr_type_t type = UNICAST_ADDR;
4774
4775         return inet6_dump_addr(skb, cb, type);
4776 }
4777
4778 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4779 {
4780         enum addr_type_t type = MULTICAST_ADDR;
4781
4782         return inet6_dump_addr(skb, cb, type);
4783 }
4784
4785
4786 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4787 {
4788         enum addr_type_t type = ANYCAST_ADDR;
4789
4790         return inet6_dump_addr(skb, cb, type);
4791 }
4792
4793 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4794 {
4795         struct net *net = sock_net(in_skb->sk);
4796         struct ifaddrmsg *ifm;
4797         struct nlattr *tb[IFA_MAX+1];
4798         struct in6_addr *addr = NULL, *peer;
4799         struct net_device *dev = NULL;
4800         struct inet6_ifaddr *ifa;
4801         struct sk_buff *skb;
4802         int err;
4803
4804         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4805         if (err < 0)
4806                 goto errout;
4807
4808         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4809         if (!addr) {
4810                 err = -EINVAL;
4811                 goto errout;
4812         }
4813
4814         ifm = nlmsg_data(nlh);
4815         if (ifm->ifa_index)
4816                 dev = __dev_get_by_index(net, ifm->ifa_index);
4817
4818         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4819         if (!ifa) {
4820                 err = -EADDRNOTAVAIL;
4821                 goto errout;
4822         }
4823
4824         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4825         if (!skb) {
4826                 err = -ENOBUFS;
4827                 goto errout_ifa;
4828         }
4829
4830         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4831                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4832         if (err < 0) {
4833                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4834                 WARN_ON(err == -EMSGSIZE);
4835                 kfree_skb(skb);
4836                 goto errout_ifa;
4837         }
4838         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4839 errout_ifa:
4840         in6_ifa_put(ifa);
4841 errout:
4842         return err;
4843 }
4844
4845 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4846 {
4847         struct sk_buff *skb;
4848         struct net *net = dev_net(ifa->idev->dev);
4849         int err = -ENOBUFS;
4850
4851         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4852         if (!skb)
4853                 goto errout;
4854
4855         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4856         if (err < 0) {
4857                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4858                 WARN_ON(err == -EMSGSIZE);
4859                 kfree_skb(skb);
4860                 goto errout;
4861         }
4862         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4863         return;
4864 errout:
4865         if (err < 0)
4866                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4867 }
4868
4869 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4870                                 __s32 *array, int bytes)
4871 {
4872         BUG_ON(bytes < (DEVCONF_MAX * 4));
4873
4874         memset(array, 0, bytes);
4875         array[DEVCONF_FORWARDING] = cnf->forwarding;
4876         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4877         array[DEVCONF_MTU6] = cnf->mtu6;
4878         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4879         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4880         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4881         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4882         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4883         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4884                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4885         array[DEVCONF_RTR_SOLICIT_DELAY] =
4886                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4887         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4888         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4889                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4890         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4891                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4892         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4893         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4894         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4895         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4896         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4897         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4898         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4899         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4900         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4901 #ifdef CONFIG_IPV6_ROUTER_PREF
4902         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4903         array[DEVCONF_RTR_PROBE_INTERVAL] =
4904                 jiffies_to_msecs(cnf->rtr_probe_interval);
4905 #ifdef CONFIG_IPV6_ROUTE_INFO
4906         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4907 #endif
4908 #endif
4909         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4910         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4911 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4912         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4913         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4914 #endif
4915 #ifdef CONFIG_IPV6_MROUTE
4916         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4917 #endif
4918         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4919         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4920         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4921         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4922         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4923         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4924         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4925         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4926         /* we omit DEVCONF_STABLE_SECRET for now */
4927         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4928         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4929         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4930         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4931 }
4932
4933 static inline size_t inet6_ifla6_size(void)
4934 {
4935         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4936              + nla_total_size(sizeof(struct ifla_cacheinfo))
4937              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4938              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4939              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4940              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4941 }
4942
4943 static inline size_t inet6_if_nlmsg_size(void)
4944 {
4945         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4946                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4947                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4948                + nla_total_size(4) /* IFLA_MTU */
4949                + nla_total_size(4) /* IFLA_LINK */
4950                + nla_total_size(1) /* IFLA_OPERSTATE */
4951                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4952 }
4953
4954 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4955                                       int items, int bytes)
4956 {
4957         int i;
4958         int pad = bytes - sizeof(u64) * items;
4959         BUG_ON(pad < 0);
4960
4961         /* Use put_unaligned() because stats may not be aligned for u64. */
4962         put_unaligned(items, &stats[0]);
4963         for (i = 1; i < items; i++)
4964                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4965
4966         memset(&stats[items], 0, pad);
4967 }
4968
4969 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4970                                         int bytes, size_t syncpoff)
4971 {
4972         int i, c;
4973         u64 buff[IPSTATS_MIB_MAX];
4974         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4975
4976         BUG_ON(pad < 0);
4977
4978         memset(buff, 0, sizeof(buff));
4979         buff[0] = IPSTATS_MIB_MAX;
4980
4981         for_each_possible_cpu(c) {
4982                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4983                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4984         }
4985
4986         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4987         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4988 }
4989
4990 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4991                              int bytes)
4992 {
4993         switch (attrtype) {
4994         case IFLA_INET6_STATS:
4995                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4996                                      offsetof(struct ipstats_mib, syncp));
4997                 break;
4998         case IFLA_INET6_ICMP6STATS:
4999                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
5000                 break;
5001         }
5002 }
5003
5004 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5005                                   u32 ext_filter_mask)
5006 {
5007         struct nlattr *nla;
5008         struct ifla_cacheinfo ci;
5009
5010         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5011                 goto nla_put_failure;
5012         ci.max_reasm_len = IPV6_MAXPLEN;
5013         ci.tstamp = cstamp_delta(idev->tstamp);
5014         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5015         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5016         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5017                 goto nla_put_failure;
5018         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5019         if (!nla)
5020                 goto nla_put_failure;
5021         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5022
5023         /* XXX - MC not implemented */
5024
5025         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5026                 return 0;
5027
5028         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5029         if (!nla)
5030                 goto nla_put_failure;
5031         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5032
5033         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5034         if (!nla)
5035                 goto nla_put_failure;
5036         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5037
5038         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5039         if (!nla)
5040                 goto nla_put_failure;
5041
5042         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5043                 goto nla_put_failure;
5044
5045         read_lock_bh(&idev->lock);
5046         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5047         read_unlock_bh(&idev->lock);
5048
5049         return 0;
5050
5051 nla_put_failure:
5052         return -EMSGSIZE;
5053 }
5054
5055 static size_t inet6_get_link_af_size(const struct net_device *dev,
5056                                      u32 ext_filter_mask)
5057 {
5058         if (!__in6_dev_get(dev))
5059                 return 0;
5060
5061         return inet6_ifla6_size();
5062 }
5063
5064 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5065                               u32 ext_filter_mask)
5066 {
5067         struct inet6_dev *idev = __in6_dev_get(dev);
5068
5069         if (!idev)
5070                 return -ENODATA;
5071
5072         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5073                 return -EMSGSIZE;
5074
5075         return 0;
5076 }
5077
5078 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5079 {
5080         struct inet6_ifaddr *ifp;
5081         struct net_device *dev = idev->dev;
5082         bool clear_token, update_rs = false;
5083         struct in6_addr ll_addr;
5084
5085         ASSERT_RTNL();
5086
5087         if (!token)
5088                 return -EINVAL;
5089         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5090                 return -EINVAL;
5091         if (!ipv6_accept_ra(idev))
5092                 return -EINVAL;
5093         if (idev->cnf.rtr_solicits <= 0)
5094                 return -EINVAL;
5095
5096         write_lock_bh(&idev->lock);
5097
5098         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5099         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5100
5101         write_unlock_bh(&idev->lock);
5102
5103         clear_token = ipv6_addr_any(token);
5104         if (clear_token)
5105                 goto update_lft;
5106
5107         if (!idev->dead && (idev->if_flags & IF_READY) &&
5108             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5109                              IFA_F_OPTIMISTIC)) {
5110                 /* If we're not ready, then normal ifup will take care
5111                  * of this. Otherwise, we need to request our rs here.
5112                  */
5113                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5114                 update_rs = true;
5115         }
5116
5117 update_lft:
5118         write_lock_bh(&idev->lock);
5119
5120         if (update_rs) {
5121                 idev->if_flags |= IF_RS_SENT;
5122                 idev->rs_probes = 1;
5123                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
5124         }
5125
5126         /* Well, that's kinda nasty ... */
5127         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5128                 spin_lock(&ifp->lock);
5129                 if (ifp->tokenized) {
5130                         ifp->valid_lft = 0;
5131                         ifp->prefered_lft = 0;
5132                 }
5133                 spin_unlock(&ifp->lock);
5134         }
5135
5136         write_unlock_bh(&idev->lock);
5137         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5138         addrconf_verify_rtnl();
5139         return 0;
5140 }
5141
5142 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5143         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5144         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5145 };
5146
5147 static int inet6_validate_link_af(const struct net_device *dev,
5148                                   const struct nlattr *nla)
5149 {
5150         struct nlattr *tb[IFLA_INET6_MAX + 1];
5151
5152         if (dev && !__in6_dev_get(dev))
5153                 return -EAFNOSUPPORT;
5154
5155         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5156 }
5157
5158 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5159 {
5160         int err = -EINVAL;
5161         struct inet6_dev *idev = __in6_dev_get(dev);
5162         struct nlattr *tb[IFLA_INET6_MAX + 1];
5163
5164         if (!idev)
5165                 return -EAFNOSUPPORT;
5166
5167         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5168                 BUG();
5169
5170         if (tb[IFLA_INET6_TOKEN]) {
5171                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5172                 if (err)
5173                         return err;
5174         }
5175
5176         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5177                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5178
5179                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5180                     mode != IN6_ADDR_GEN_MODE_NONE &&
5181                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5182                     mode != IN6_ADDR_GEN_MODE_RANDOM)
5183                         return -EINVAL;
5184
5185                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5186                     !idev->cnf.stable_secret.initialized &&
5187                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5188                         return -EINVAL;
5189
5190                 idev->addr_gen_mode = mode;
5191                 err = 0;
5192         }
5193
5194         return err;
5195 }
5196
5197 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5198                              u32 portid, u32 seq, int event, unsigned int flags)
5199 {
5200         struct net_device *dev = idev->dev;
5201         struct ifinfomsg *hdr;
5202         struct nlmsghdr *nlh;
5203         void *protoinfo;
5204
5205         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5206         if (!nlh)
5207                 return -EMSGSIZE;
5208
5209         hdr = nlmsg_data(nlh);
5210         hdr->ifi_family = AF_INET6;
5211         hdr->__ifi_pad = 0;
5212         hdr->ifi_type = dev->type;
5213         hdr->ifi_index = dev->ifindex;
5214         hdr->ifi_flags = dev_get_flags(dev);
5215         hdr->ifi_change = 0;
5216
5217         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5218             (dev->addr_len &&
5219              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5220             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5221             (dev->ifindex != dev_get_iflink(dev) &&
5222              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5223             nla_put_u8(skb, IFLA_OPERSTATE,
5224                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5225                 goto nla_put_failure;
5226         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5227         if (!protoinfo)
5228                 goto nla_put_failure;
5229
5230         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5231                 goto nla_put_failure;
5232
5233         nla_nest_end(skb, protoinfo);
5234         nlmsg_end(skb, nlh);
5235         return 0;
5236
5237 nla_put_failure:
5238         nlmsg_cancel(skb, nlh);
5239         return -EMSGSIZE;
5240 }
5241
5242 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5243 {
5244         struct net *net = sock_net(skb->sk);
5245         int h, s_h;
5246         int idx = 0, s_idx;
5247         struct net_device *dev;
5248         struct inet6_dev *idev;
5249         struct hlist_head *head;
5250
5251         s_h = cb->args[0];
5252         s_idx = cb->args[1];
5253
5254         rcu_read_lock();
5255         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5256                 idx = 0;
5257                 head = &net->dev_index_head[h];
5258                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5259                         if (idx < s_idx)
5260                                 goto cont;
5261                         idev = __in6_dev_get(dev);
5262                         if (!idev)
5263                                 goto cont;
5264                         if (inet6_fill_ifinfo(skb, idev,
5265                                               NETLINK_CB(cb->skb).portid,
5266                                               cb->nlh->nlmsg_seq,
5267                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5268                                 goto out;
5269 cont:
5270                         idx++;
5271                 }
5272         }
5273 out:
5274         rcu_read_unlock();
5275         cb->args[1] = idx;
5276         cb->args[0] = h;
5277
5278         return skb->len;
5279 }
5280
5281 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5282 {
5283         struct sk_buff *skb;
5284         struct net *net = dev_net(idev->dev);
5285         int err = -ENOBUFS;
5286
5287         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5288         if (!skb)
5289                 goto errout;
5290
5291         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5292         if (err < 0) {
5293                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5294                 WARN_ON(err == -EMSGSIZE);
5295                 kfree_skb(skb);
5296                 goto errout;
5297         }
5298         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5299         return;
5300 errout:
5301         if (err < 0)
5302                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5303 }
5304
5305 static inline size_t inet6_prefix_nlmsg_size(void)
5306 {
5307         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5308                + nla_total_size(sizeof(struct in6_addr))
5309                + nla_total_size(sizeof(struct prefix_cacheinfo));
5310 }
5311
5312 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5313                              struct prefix_info *pinfo, u32 portid, u32 seq,
5314                              int event, unsigned int flags)
5315 {
5316         struct prefixmsg *pmsg;
5317         struct nlmsghdr *nlh;
5318         struct prefix_cacheinfo ci;
5319
5320         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5321         if (!nlh)
5322                 return -EMSGSIZE;
5323
5324         pmsg = nlmsg_data(nlh);
5325         pmsg->prefix_family = AF_INET6;
5326         pmsg->prefix_pad1 = 0;
5327         pmsg->prefix_pad2 = 0;
5328         pmsg->prefix_ifindex = idev->dev->ifindex;
5329         pmsg->prefix_len = pinfo->prefix_len;
5330         pmsg->prefix_type = pinfo->type;
5331         pmsg->prefix_pad3 = 0;
5332         pmsg->prefix_flags = 0;
5333         if (pinfo->onlink)
5334                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5335         if (pinfo->autoconf)
5336                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5337
5338         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5339                 goto nla_put_failure;
5340         ci.preferred_time = ntohl(pinfo->prefered);
5341         ci.valid_time = ntohl(pinfo->valid);
5342         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5343                 goto nla_put_failure;
5344         nlmsg_end(skb, nlh);
5345         return 0;
5346
5347 nla_put_failure:
5348         nlmsg_cancel(skb, nlh);
5349         return -EMSGSIZE;
5350 }
5351
5352 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5353                          struct prefix_info *pinfo)
5354 {
5355         struct sk_buff *skb;
5356         struct net *net = dev_net(idev->dev);
5357         int err = -ENOBUFS;
5358
5359         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5360         if (!skb)
5361                 goto errout;
5362
5363         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5364         if (err < 0) {
5365                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5366                 WARN_ON(err == -EMSGSIZE);
5367                 kfree_skb(skb);
5368                 goto errout;
5369         }
5370         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5371         return;
5372 errout:
5373         if (err < 0)
5374                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5375 }
5376
5377 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5378 {
5379         struct net *net = dev_net(ifp->idev->dev);
5380
5381         if (event)
5382                 ASSERT_RTNL();
5383
5384         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5385
5386         switch (event) {
5387         case RTM_NEWADDR:
5388                 /*
5389                  * If the address was optimistic
5390                  * we inserted the route at the start of
5391                  * our DAD process, so we don't need
5392                  * to do it again
5393                  */
5394                 if (!(ifp->rt->rt6i_node))
5395                         ip6_ins_rt(ifp->rt);
5396                 if (ifp->idev->cnf.forwarding)
5397                         addrconf_join_anycast(ifp);
5398                 if (!ipv6_addr_any(&ifp->peer_addr))
5399                         addrconf_prefix_route(&ifp->peer_addr, 128,
5400                                               ifp->idev->dev, 0, 0);
5401                 break;
5402         case RTM_DELADDR:
5403                 if (ifp->idev->cnf.forwarding)
5404                         addrconf_leave_anycast(ifp);
5405                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5406                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5407                         struct rt6_info *rt;
5408
5409                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5410                                                        ifp->idev->dev, 0, 0);
5411                         if (rt)
5412                                 ip6_del_rt(rt);
5413                 }
5414                 if (ifp->rt) {
5415                         dst_hold(&ifp->rt->dst);
5416                         ip6_del_rt(ifp->rt);
5417                 }
5418                 rt_genid_bump_ipv6(net);
5419                 break;
5420         }
5421         atomic_inc(&net->ipv6.dev_addr_genid);
5422 }
5423
5424 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5425 {
5426         rcu_read_lock_bh();
5427         if (likely(ifp->idev->dead == 0))
5428                 __ipv6_ifa_notify(event, ifp);
5429         rcu_read_unlock_bh();
5430 }
5431
5432 #ifdef CONFIG_SYSCTL
5433
5434 static
5435 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5436                            void __user *buffer, size_t *lenp, loff_t *ppos)
5437 {
5438         int *valp = ctl->data;
5439         int val = *valp;
5440         loff_t pos = *ppos;
5441         struct ctl_table lctl;
5442         int ret;
5443
5444         /*
5445          * ctl->data points to idev->cnf.forwarding, we should
5446          * not modify it until we get the rtnl lock.
5447          */
5448         lctl = *ctl;
5449         lctl.data = &val;
5450
5451         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5452
5453         if (write)
5454                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5455         if (ret)
5456                 *ppos = pos;
5457         return ret;
5458 }
5459
5460 static
5461 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5462                               void __user *buffer, size_t *lenp, loff_t *ppos)
5463 {
5464         struct ctl_table lctl;
5465         int min_hl = 1, max_hl = 255;
5466
5467         lctl = *ctl;
5468         lctl.extra1 = &min_hl;
5469         lctl.extra2 = &max_hl;
5470
5471         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5472 }
5473
5474 static
5475 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5476                         void __user *buffer, size_t *lenp, loff_t *ppos)
5477 {
5478         struct inet6_dev *idev = ctl->extra1;
5479         int min_mtu = IPV6_MIN_MTU;
5480         struct ctl_table lctl;
5481
5482         lctl = *ctl;
5483         lctl.extra1 = &min_mtu;
5484         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5485
5486         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5487 }
5488
5489 static void dev_disable_change(struct inet6_dev *idev)
5490 {
5491         struct netdev_notifier_info info;
5492
5493         if (!idev || !idev->dev)
5494                 return;
5495
5496         netdev_notifier_info_init(&info, idev->dev);
5497         if (idev->cnf.disable_ipv6)
5498                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5499         else
5500                 addrconf_notify(NULL, NETDEV_UP, &info);
5501 }
5502
5503 static void addrconf_disable_change(struct net *net, __s32 newf)
5504 {
5505         struct net_device *dev;
5506         struct inet6_dev *idev;
5507
5508         rcu_read_lock();
5509         for_each_netdev_rcu(net, dev) {
5510                 idev = __in6_dev_get(dev);
5511                 if (idev) {
5512                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5513                         idev->cnf.disable_ipv6 = newf;
5514                         if (changed)
5515                                 dev_disable_change(idev);
5516                 }
5517         }
5518         rcu_read_unlock();
5519 }
5520
5521 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5522 {
5523         struct net *net;
5524         int old;
5525
5526         if (!rtnl_trylock())
5527                 return restart_syscall();
5528
5529         net = (struct net *)table->extra2;
5530         old = *p;
5531         *p = newf;
5532
5533         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5534                 rtnl_unlock();
5535                 return 0;
5536         }
5537
5538         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5539                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5540                 addrconf_disable_change(net, newf);
5541         } else if ((!newf) ^ (!old))
5542                 dev_disable_change((struct inet6_dev *)table->extra1);
5543
5544         rtnl_unlock();
5545         return 0;
5546 }
5547
5548 static
5549 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5550                             void __user *buffer, size_t *lenp, loff_t *ppos)
5551 {
5552         int *valp = ctl->data;
5553         int val = *valp;
5554         loff_t pos = *ppos;
5555         struct ctl_table lctl;
5556         int ret;
5557
5558         /*
5559          * ctl->data points to idev->cnf.disable_ipv6, we should
5560          * not modify it until we get the rtnl lock.
5561          */
5562         lctl = *ctl;
5563         lctl.data = &val;
5564
5565         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5566
5567         if (write)
5568                 ret = addrconf_disable_ipv6(ctl, valp, val);
5569         if (ret)
5570                 *ppos = pos;
5571         return ret;
5572 }
5573
5574 static
5575 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5576                               void __user *buffer, size_t *lenp, loff_t *ppos)
5577 {
5578         int *valp = ctl->data;
5579         int ret;
5580         int old, new;
5581
5582         old = *valp;
5583         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5584         new = *valp;
5585
5586         if (write && old != new) {
5587                 struct net *net = ctl->extra2;
5588
5589                 if (!rtnl_trylock())
5590                         return restart_syscall();
5591
5592                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5593                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5594                                                      NETCONFA_IFINDEX_DEFAULT,
5595                                                      net->ipv6.devconf_dflt);
5596                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5597                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5598                                                      NETCONFA_IFINDEX_ALL,
5599                                                      net->ipv6.devconf_all);
5600                 else {
5601                         struct inet6_dev *idev = ctl->extra1;
5602
5603                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5604                                                      idev->dev->ifindex,
5605                                                      &idev->cnf);
5606                 }
5607                 rtnl_unlock();
5608         }
5609
5610         return ret;
5611 }
5612
5613 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5614                                          void __user *buffer, size_t *lenp,
5615                                          loff_t *ppos)
5616 {
5617         int err;
5618         struct in6_addr addr;
5619         char str[IPV6_MAX_STRLEN];
5620         struct ctl_table lctl = *ctl;
5621         struct net *net = ctl->extra2;
5622         struct ipv6_stable_secret *secret = ctl->data;
5623
5624         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5625                 return -EIO;
5626
5627         lctl.maxlen = IPV6_MAX_STRLEN;
5628         lctl.data = str;
5629
5630         if (!rtnl_trylock())
5631                 return restart_syscall();
5632
5633         if (!write && !secret->initialized) {
5634                 err = -EIO;
5635                 goto out;
5636         }
5637
5638         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5639         if (err >= sizeof(str)) {
5640                 err = -EIO;
5641                 goto out;
5642         }
5643
5644         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5645         if (err || !write)
5646                 goto out;
5647
5648         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5649                 err = -EIO;
5650                 goto out;
5651         }
5652
5653         secret->initialized = true;
5654         secret->secret = addr;
5655
5656         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5657                 struct net_device *dev;
5658
5659                 for_each_netdev(net, dev) {
5660                         struct inet6_dev *idev = __in6_dev_get(dev);
5661
5662                         if (idev) {
5663                                 idev->addr_gen_mode =
5664                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5665                         }
5666                 }
5667         } else {
5668                 struct inet6_dev *idev = ctl->extra1;
5669
5670                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5671         }
5672
5673 out:
5674         rtnl_unlock();
5675
5676         return err;
5677 }
5678
5679 static
5680 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5681                                                 int write,
5682                                                 void __user *buffer,
5683                                                 size_t *lenp,
5684                                                 loff_t *ppos)
5685 {
5686         int *valp = ctl->data;
5687         int val = *valp;
5688         loff_t pos = *ppos;
5689         struct ctl_table lctl;
5690         int ret;
5691
5692         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5693          * we should not modify it until we get the rtnl lock.
5694          */
5695         lctl = *ctl;
5696         lctl.data = &val;
5697
5698         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5699
5700         if (write)
5701                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5702         if (ret)
5703                 *ppos = pos;
5704         return ret;
5705 }
5706
5707 static const struct ctl_table addrconf_sysctl[] = {
5708         {
5709                 .procname       = "forwarding",
5710                 .data           = &ipv6_devconf.forwarding,
5711                 .maxlen         = sizeof(int),
5712                 .mode           = 0644,
5713                 .proc_handler   = addrconf_sysctl_forward,
5714         },
5715         {
5716                 .procname       = "hop_limit",
5717                 .data           = &ipv6_devconf.hop_limit,
5718                 .maxlen         = sizeof(int),
5719                 .mode           = 0644,
5720                 .proc_handler   = addrconf_sysctl_hop_limit,
5721         },
5722         {
5723                 .procname       = "mtu",
5724                 .data           = &ipv6_devconf.mtu6,
5725                 .maxlen         = sizeof(int),
5726                 .mode           = 0644,
5727                 .proc_handler   = addrconf_sysctl_mtu,
5728         },
5729         {
5730                 .procname       = "accept_ra",
5731                 .data           = &ipv6_devconf.accept_ra,
5732                 .maxlen         = sizeof(int),
5733                 .mode           = 0644,
5734                 .proc_handler   = proc_dointvec,
5735         },
5736         {
5737                 .procname       = "accept_redirects",
5738                 .data           = &ipv6_devconf.accept_redirects,
5739                 .maxlen         = sizeof(int),
5740                 .mode           = 0644,
5741                 .proc_handler   = proc_dointvec,
5742         },
5743         {
5744                 .procname       = "autoconf",
5745                 .data           = &ipv6_devconf.autoconf,
5746                 .maxlen         = sizeof(int),
5747                 .mode           = 0644,
5748                 .proc_handler   = proc_dointvec,
5749         },
5750         {
5751                 .procname       = "dad_transmits",
5752                 .data           = &ipv6_devconf.dad_transmits,
5753                 .maxlen         = sizeof(int),
5754                 .mode           = 0644,
5755                 .proc_handler   = proc_dointvec,
5756         },
5757         {
5758                 .procname       = "router_solicitations",
5759                 .data           = &ipv6_devconf.rtr_solicits,
5760                 .maxlen         = sizeof(int),
5761                 .mode           = 0644,
5762                 .proc_handler   = proc_dointvec,
5763         },
5764         {
5765                 .procname       = "router_solicitation_interval",
5766                 .data           = &ipv6_devconf.rtr_solicit_interval,
5767                 .maxlen         = sizeof(int),
5768                 .mode           = 0644,
5769                 .proc_handler   = proc_dointvec_jiffies,
5770         },
5771         {
5772                 .procname       = "router_solicitation_delay",
5773                 .data           = &ipv6_devconf.rtr_solicit_delay,
5774                 .maxlen         = sizeof(int),
5775                 .mode           = 0644,
5776                 .proc_handler   = proc_dointvec_jiffies,
5777         },
5778         {
5779                 .procname       = "force_mld_version",
5780                 .data           = &ipv6_devconf.force_mld_version,
5781                 .maxlen         = sizeof(int),
5782                 .mode           = 0644,
5783                 .proc_handler   = proc_dointvec,
5784         },
5785         {
5786                 .procname       = "mldv1_unsolicited_report_interval",
5787                 .data           =
5788                         &ipv6_devconf.mldv1_unsolicited_report_interval,
5789                 .maxlen         = sizeof(int),
5790                 .mode           = 0644,
5791                 .proc_handler   = proc_dointvec_ms_jiffies,
5792         },
5793         {
5794                 .procname       = "mldv2_unsolicited_report_interval",
5795                 .data           =
5796                         &ipv6_devconf.mldv2_unsolicited_report_interval,
5797                 .maxlen         = sizeof(int),
5798                 .mode           = 0644,
5799                 .proc_handler   = proc_dointvec_ms_jiffies,
5800         },
5801         {
5802                 .procname       = "use_tempaddr",
5803                 .data           = &ipv6_devconf.use_tempaddr,
5804                 .maxlen         = sizeof(int),
5805                 .mode           = 0644,
5806                 .proc_handler   = proc_dointvec,
5807         },
5808         {
5809                 .procname       = "temp_valid_lft",
5810                 .data           = &ipv6_devconf.temp_valid_lft,
5811                 .maxlen         = sizeof(int),
5812                 .mode           = 0644,
5813                 .proc_handler   = proc_dointvec,
5814         },
5815         {
5816                 .procname       = "temp_prefered_lft",
5817                 .data           = &ipv6_devconf.temp_prefered_lft,
5818                 .maxlen         = sizeof(int),
5819                 .mode           = 0644,
5820                 .proc_handler   = proc_dointvec,
5821         },
5822         {
5823                 .procname       = "regen_max_retry",
5824                 .data           = &ipv6_devconf.regen_max_retry,
5825                 .maxlen         = sizeof(int),
5826                 .mode           = 0644,
5827                 .proc_handler   = proc_dointvec,
5828         },
5829         {
5830                 .procname       = "max_desync_factor",
5831                 .data           = &ipv6_devconf.max_desync_factor,
5832                 .maxlen         = sizeof(int),
5833                 .mode           = 0644,
5834                 .proc_handler   = proc_dointvec,
5835         },
5836         {
5837                 .procname       = "max_addresses",
5838                 .data           = &ipv6_devconf.max_addresses,
5839                 .maxlen         = sizeof(int),
5840                 .mode           = 0644,
5841                 .proc_handler   = proc_dointvec,
5842         },
5843         {
5844                 .procname       = "accept_ra_defrtr",
5845                 .data           = &ipv6_devconf.accept_ra_defrtr,
5846                 .maxlen         = sizeof(int),
5847                 .mode           = 0644,
5848                 .proc_handler   = proc_dointvec,
5849         },
5850         {
5851                 .procname       = "accept_ra_min_hop_limit",
5852                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5853                 .maxlen         = sizeof(int),
5854                 .mode           = 0644,
5855                 .proc_handler   = proc_dointvec,
5856         },
5857         {
5858                 .procname       = "accept_ra_pinfo",
5859                 .data           = &ipv6_devconf.accept_ra_pinfo,
5860                 .maxlen         = sizeof(int),
5861                 .mode           = 0644,
5862                 .proc_handler   = proc_dointvec,
5863         },
5864 #ifdef CONFIG_IPV6_ROUTER_PREF
5865         {
5866                 .procname       = "accept_ra_rtr_pref",
5867                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
5868                 .maxlen         = sizeof(int),
5869                 .mode           = 0644,
5870                 .proc_handler   = proc_dointvec,
5871         },
5872         {
5873                 .procname       = "router_probe_interval",
5874                 .data           = &ipv6_devconf.rtr_probe_interval,
5875                 .maxlen         = sizeof(int),
5876                 .mode           = 0644,
5877                 .proc_handler   = proc_dointvec_jiffies,
5878         },
5879 #ifdef CONFIG_IPV6_ROUTE_INFO
5880         {
5881                 .procname       = "accept_ra_rt_info_max_plen",
5882                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5883                 .maxlen         = sizeof(int),
5884                 .mode           = 0644,
5885                 .proc_handler   = proc_dointvec,
5886         },
5887 #endif
5888 #endif
5889         {
5890                 .procname       = "proxy_ndp",
5891                 .data           = &ipv6_devconf.proxy_ndp,
5892                 .maxlen         = sizeof(int),
5893                 .mode           = 0644,
5894                 .proc_handler   = addrconf_sysctl_proxy_ndp,
5895         },
5896         {
5897                 .procname       = "accept_source_route",
5898                 .data           = &ipv6_devconf.accept_source_route,
5899                 .maxlen         = sizeof(int),
5900                 .mode           = 0644,
5901                 .proc_handler   = proc_dointvec,
5902         },
5903 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5904         {
5905                 .procname       = "optimistic_dad",
5906                 .data           = &ipv6_devconf.optimistic_dad,
5907                 .maxlen         = sizeof(int),
5908                 .mode           = 0644,
5909                 .proc_handler   = proc_dointvec,
5910         },
5911         {
5912                 .procname       = "use_optimistic",
5913                 .data           = &ipv6_devconf.use_optimistic,
5914                 .maxlen         = sizeof(int),
5915                 .mode           = 0644,
5916                 .proc_handler   = proc_dointvec,
5917         },
5918 #endif
5919 #ifdef CONFIG_IPV6_MROUTE
5920         {
5921                 .procname       = "mc_forwarding",
5922                 .data           = &ipv6_devconf.mc_forwarding,
5923                 .maxlen         = sizeof(int),
5924                 .mode           = 0444,
5925                 .proc_handler   = proc_dointvec,
5926         },
5927 #endif
5928         {
5929                 .procname       = "disable_ipv6",
5930                 .data           = &ipv6_devconf.disable_ipv6,
5931                 .maxlen         = sizeof(int),
5932                 .mode           = 0644,
5933                 .proc_handler   = addrconf_sysctl_disable,
5934         },
5935         {
5936                 .procname       = "accept_dad",
5937                 .data           = &ipv6_devconf.accept_dad,
5938                 .maxlen         = sizeof(int),
5939                 .mode           = 0644,
5940                 .proc_handler   = proc_dointvec,
5941         },
5942         {
5943                 .procname       = "force_tllao",
5944                 .data           = &ipv6_devconf.force_tllao,
5945                 .maxlen         = sizeof(int),
5946                 .mode           = 0644,
5947                 .proc_handler   = proc_dointvec
5948         },
5949         {
5950                 .procname       = "ndisc_notify",
5951                 .data           = &ipv6_devconf.ndisc_notify,
5952                 .maxlen         = sizeof(int),
5953                 .mode           = 0644,
5954                 .proc_handler   = proc_dointvec
5955         },
5956         {
5957                 .procname       = "suppress_frag_ndisc",
5958                 .data           = &ipv6_devconf.suppress_frag_ndisc,
5959                 .maxlen         = sizeof(int),
5960                 .mode           = 0644,
5961                 .proc_handler   = proc_dointvec
5962         },
5963         {
5964                 .procname       = "accept_ra_from_local",
5965                 .data           = &ipv6_devconf.accept_ra_from_local,
5966                 .maxlen         = sizeof(int),
5967                 .mode           = 0644,
5968                 .proc_handler   = proc_dointvec,
5969         },
5970         {
5971                 .procname       = "accept_ra_mtu",
5972                 .data           = &ipv6_devconf.accept_ra_mtu,
5973                 .maxlen         = sizeof(int),
5974                 .mode           = 0644,
5975                 .proc_handler   = proc_dointvec,
5976         },
5977         {
5978                 .procname       = "stable_secret",
5979                 .data           = &ipv6_devconf.stable_secret,
5980                 .maxlen         = IPV6_MAX_STRLEN,
5981                 .mode           = 0600,
5982                 .proc_handler   = addrconf_sysctl_stable_secret,
5983         },
5984         {
5985                 .procname       = "use_oif_addrs_only",
5986                 .data           = &ipv6_devconf.use_oif_addrs_only,
5987                 .maxlen         = sizeof(int),
5988                 .mode           = 0644,
5989                 .proc_handler   = proc_dointvec,
5990         },
5991         {
5992                 .procname       = "ignore_routes_with_linkdown",
5993                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5994                 .maxlen         = sizeof(int),
5995                 .mode           = 0644,
5996                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5997         },
5998         {
5999                 .procname       = "drop_unicast_in_l2_multicast",
6000                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6001                 .maxlen         = sizeof(int),
6002                 .mode           = 0644,
6003                 .proc_handler   = proc_dointvec,
6004         },
6005         {
6006                 .procname       = "drop_unsolicited_na",
6007                 .data           = &ipv6_devconf.drop_unsolicited_na,
6008                 .maxlen         = sizeof(int),
6009                 .mode           = 0644,
6010                 .proc_handler   = proc_dointvec,
6011         },
6012         {
6013                 .procname       = "keep_addr_on_down",
6014                 .data           = &ipv6_devconf.keep_addr_on_down,
6015                 .maxlen         = sizeof(int),
6016                 .mode           = 0644,
6017                 .proc_handler   = proc_dointvec,
6018
6019         },
6020         {
6021                 /* sentinel */
6022         }
6023 };
6024
6025 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6026                 struct inet6_dev *idev, struct ipv6_devconf *p)
6027 {
6028         int i;
6029         struct ctl_table *table;
6030         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6031
6032         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6033         if (!table)
6034                 goto out;
6035
6036         for (i = 0; table[i].data; i++) {
6037                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6038                 table[i].extra1 = idev; /* embedded; no ref */
6039                 table[i].extra2 = net;
6040         }
6041
6042         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6043
6044         p->sysctl_header = register_net_sysctl(net, path, table);
6045         if (!p->sysctl_header)
6046                 goto free;
6047
6048         return 0;
6049
6050 free:
6051         kfree(table);
6052 out:
6053         return -ENOBUFS;
6054 }
6055
6056 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6057 {
6058         struct ctl_table *table;
6059
6060         if (!p->sysctl_header)
6061                 return;
6062
6063         table = p->sysctl_header->ctl_table_arg;
6064         unregister_net_sysctl_table(p->sysctl_header);
6065         p->sysctl_header = NULL;
6066         kfree(table);
6067 }
6068
6069 static int addrconf_sysctl_register(struct inet6_dev *idev)
6070 {
6071         int err;
6072
6073         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6074                 return -EINVAL;
6075
6076         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6077                                     &ndisc_ifinfo_sysctl_change);
6078         if (err)
6079                 return err;
6080         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6081                                          idev, &idev->cnf);
6082         if (err)
6083                 neigh_sysctl_unregister(idev->nd_parms);
6084
6085         return err;
6086 }
6087
6088 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6089 {
6090         __addrconf_sysctl_unregister(&idev->cnf);
6091         neigh_sysctl_unregister(idev->nd_parms);
6092 }
6093
6094
6095 #endif
6096
6097 static int __net_init addrconf_init_net(struct net *net)
6098 {
6099         int err = -ENOMEM;
6100         struct ipv6_devconf *all, *dflt;
6101
6102         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6103         if (!all)
6104                 goto err_alloc_all;
6105
6106         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6107         if (!dflt)
6108                 goto err_alloc_dflt;
6109
6110         /* these will be inherited by all namespaces */
6111         dflt->autoconf = ipv6_defaults.autoconf;
6112         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6113
6114         dflt->stable_secret.initialized = false;
6115         all->stable_secret.initialized = false;
6116
6117         net->ipv6.devconf_all = all;
6118         net->ipv6.devconf_dflt = dflt;
6119
6120 #ifdef CONFIG_SYSCTL
6121         err = __addrconf_sysctl_register(net, "all", NULL, all);
6122         if (err < 0)
6123                 goto err_reg_all;
6124
6125         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6126         if (err < 0)
6127                 goto err_reg_dflt;
6128 #endif
6129         return 0;
6130
6131 #ifdef CONFIG_SYSCTL
6132 err_reg_dflt:
6133         __addrconf_sysctl_unregister(all);
6134 err_reg_all:
6135         kfree(dflt);
6136 #endif
6137 err_alloc_dflt:
6138         kfree(all);
6139 err_alloc_all:
6140         return err;
6141 }
6142
6143 static void __net_exit addrconf_exit_net(struct net *net)
6144 {
6145 #ifdef CONFIG_SYSCTL
6146         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6147         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6148 #endif
6149         kfree(net->ipv6.devconf_dflt);
6150         kfree(net->ipv6.devconf_all);
6151 }
6152
6153 static struct pernet_operations addrconf_ops = {
6154         .init = addrconf_init_net,
6155         .exit = addrconf_exit_net,
6156 };
6157
6158 static struct rtnl_af_ops inet6_ops __read_mostly = {
6159         .family           = AF_INET6,
6160         .fill_link_af     = inet6_fill_link_af,
6161         .get_link_af_size = inet6_get_link_af_size,
6162         .validate_link_af = inet6_validate_link_af,
6163         .set_link_af      = inet6_set_link_af,
6164 };
6165
6166 /*
6167  *      Init / cleanup code
6168  */
6169
6170 int __init addrconf_init(void)
6171 {
6172         struct inet6_dev *idev;
6173         int i, err;
6174
6175         err = ipv6_addr_label_init();
6176         if (err < 0) {
6177                 pr_crit("%s: cannot initialize default policy table: %d\n",
6178                         __func__, err);
6179                 goto out;
6180         }
6181
6182         err = register_pernet_subsys(&addrconf_ops);
6183         if (err < 0)
6184                 goto out_addrlabel;
6185
6186         addrconf_wq = create_workqueue("ipv6_addrconf");
6187         if (!addrconf_wq) {
6188                 err = -ENOMEM;
6189                 goto out_nowq;
6190         }
6191
6192         /* The addrconf netdev notifier requires that loopback_dev
6193          * has it's ipv6 private information allocated and setup
6194          * before it can bring up and give link-local addresses
6195          * to other devices which are up.
6196          *
6197          * Unfortunately, loopback_dev is not necessarily the first
6198          * entry in the global dev_base list of net devices.  In fact,
6199          * it is likely to be the very last entry on that list.
6200          * So this causes the notifier registry below to try and
6201          * give link-local addresses to all devices besides loopback_dev
6202          * first, then loopback_dev, which cases all the non-loopback_dev
6203          * devices to fail to get a link-local address.
6204          *
6205          * So, as a temporary fix, allocate the ipv6 structure for
6206          * loopback_dev first by hand.
6207          * Longer term, all of the dependencies ipv6 has upon the loopback
6208          * device and it being up should be removed.
6209          */
6210         rtnl_lock();
6211         idev = ipv6_add_dev(init_net.loopback_dev);
6212         rtnl_unlock();
6213         if (IS_ERR(idev)) {
6214                 err = PTR_ERR(idev);
6215                 goto errlo;
6216         }
6217
6218         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6219                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6220
6221         register_netdevice_notifier(&ipv6_dev_notf);
6222
6223         addrconf_verify();
6224
6225         rtnl_af_register(&inet6_ops);
6226
6227         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6228                               NULL);
6229         if (err < 0)
6230                 goto errout;
6231
6232         /* Only the first call to __rtnl_register can fail */
6233         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6234         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6235         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6236                         inet6_dump_ifaddr, NULL);
6237         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6238                         inet6_dump_ifmcaddr, NULL);
6239         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6240                         inet6_dump_ifacaddr, NULL);
6241         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6242                         inet6_netconf_dump_devconf, NULL);
6243
6244         ipv6_addr_label_rtnl_register();
6245
6246         return 0;
6247 errout:
6248         rtnl_af_unregister(&inet6_ops);
6249         unregister_netdevice_notifier(&ipv6_dev_notf);
6250 errlo:
6251         destroy_workqueue(addrconf_wq);
6252 out_nowq:
6253         unregister_pernet_subsys(&addrconf_ops);
6254 out_addrlabel:
6255         ipv6_addr_label_cleanup();
6256 out:
6257         return err;
6258 }
6259
6260 void addrconf_cleanup(void)
6261 {
6262         struct net_device *dev;
6263         int i;
6264
6265         unregister_netdevice_notifier(&ipv6_dev_notf);
6266         unregister_pernet_subsys(&addrconf_ops);
6267         ipv6_addr_label_cleanup();
6268
6269         rtnl_lock();
6270
6271         __rtnl_af_unregister(&inet6_ops);
6272
6273         /* clean dev list */
6274         for_each_netdev(&init_net, dev) {
6275                 if (__in6_dev_get(dev) == NULL)
6276                         continue;
6277                 addrconf_ifdown(dev, 1);
6278         }
6279         addrconf_ifdown(init_net.loopback_dev, 2);
6280
6281         /*
6282          *      Check hash table.
6283          */
6284         spin_lock_bh(&addrconf_hash_lock);
6285         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6286                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6287         spin_unlock_bh(&addrconf_hash_lock);
6288         cancel_delayed_work(&addr_chk_work);
6289         rtnl_unlock();
6290
6291         destroy_workqueue(addrconf_wq);
6292 }