]> rtime.felk.cvut.cz Git - can-eth-gw-linux.git/blob - net/ipv6/addrconf.c
Merge branch 'akpm' (Andrew's patch-bomb)
[can-eth-gw-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/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
88 #endif
89
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
99
100 #if ACONF_DEBUG >= 3
101 #define ADBG(x) printk x
102 #else
103 #define ADBG(x)
104 #endif
105
106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
107
108 static inline u32 cstamp_delta(unsigned long cstamp)
109 {
110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 }
112
113 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
114 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
115 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
116
117 #ifdef CONFIG_SYSCTL
118 static void addrconf_sysctl_register(struct inet6_dev *idev);
119 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
120 #else
121 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
122 {
123 }
124
125 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
126 {
127 }
128 #endif
129
130 #ifdef CONFIG_IPV6_PRIVACY
131 static void __ipv6_regen_rndid(struct inet6_dev *idev);
132 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
133 static void ipv6_regen_rndid(unsigned long data);
134 #endif
135
136 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
137 static int ipv6_count_addresses(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(unsigned long);
146
147 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
148 static DEFINE_SPINLOCK(addrconf_verify_lock);
149
150 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
151 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
152
153 static void addrconf_type_change(struct net_device *dev,
154                                  unsigned long event);
155 static int addrconf_ifdown(struct net_device *dev, int how);
156
157 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_timer(unsigned long data);
159 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_run(struct inet6_dev *idev);
161 static void addrconf_rs_timer(unsigned long data);
162 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164
165 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
166                                 struct prefix_info *pinfo);
167 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
168                                struct net_device *dev);
169
170 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
171
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
173         .forwarding             = 0,
174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
175         .mtu6                   = IPV6_MIN_MTU,
176         .accept_ra              = 1,
177         .accept_redirects       = 1,
178         .autoconf               = 1,
179         .force_mld_version      = 0,
180         .dad_transmits          = 1,
181         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
182         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
183         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
184 #ifdef CONFIG_IPV6_PRIVACY
185         .use_tempaddr           = 0,
186         .temp_valid_lft         = TEMP_VALID_LIFETIME,
187         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
188         .regen_max_retry        = REGEN_MAX_RETRY,
189         .max_desync_factor      = MAX_DESYNC_FACTOR,
190 #endif
191         .max_addresses          = IPV6_MAX_ADDRESSES,
192         .accept_ra_defrtr       = 1,
193         .accept_ra_pinfo        = 1,
194 #ifdef CONFIG_IPV6_ROUTER_PREF
195         .accept_ra_rtr_pref     = 1,
196         .rtr_probe_interval     = 60 * HZ,
197 #ifdef CONFIG_IPV6_ROUTE_INFO
198         .accept_ra_rt_info_max_plen = 0,
199 #endif
200 #endif
201         .proxy_ndp              = 0,
202         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
203         .disable_ipv6           = 0,
204         .accept_dad             = 1,
205 };
206
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
208         .forwarding             = 0,
209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
210         .mtu6                   = IPV6_MIN_MTU,
211         .accept_ra              = 1,
212         .accept_redirects       = 1,
213         .autoconf               = 1,
214         .dad_transmits          = 1,
215         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
216         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
217         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
218 #ifdef CONFIG_IPV6_PRIVACY
219         .use_tempaddr           = 0,
220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
222         .regen_max_retry        = REGEN_MAX_RETRY,
223         .max_desync_factor      = MAX_DESYNC_FACTOR,
224 #endif
225         .max_addresses          = IPV6_MAX_ADDRESSES,
226         .accept_ra_defrtr       = 1,
227         .accept_ra_pinfo        = 1,
228 #ifdef CONFIG_IPV6_ROUTER_PREF
229         .accept_ra_rtr_pref     = 1,
230         .rtr_probe_interval     = 60 * HZ,
231 #ifdef CONFIG_IPV6_ROUTE_INFO
232         .accept_ra_rt_info_max_plen = 0,
233 #endif
234 #endif
235         .proxy_ndp              = 0,
236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
237         .disable_ipv6           = 0,
238         .accept_dad             = 1,
239 };
240
241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
246
247 /* Check if a valid qdisc is available */
248 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
249 {
250         return !qdisc_tx_is_noop(dev);
251 }
252
253 /* Check if a route is valid prefix route */
254 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
255 {
256         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
257 }
258
259 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
260 {
261         if (del_timer(&ifp->timer))
262                 __in6_ifa_put(ifp);
263 }
264
265 enum addrconf_timer_t {
266         AC_NONE,
267         AC_DAD,
268         AC_RS,
269 };
270
271 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
272                                enum addrconf_timer_t what,
273                                unsigned long when)
274 {
275         if (!del_timer(&ifp->timer))
276                 in6_ifa_hold(ifp);
277
278         switch (what) {
279         case AC_DAD:
280                 ifp->timer.function = addrconf_dad_timer;
281                 break;
282         case AC_RS:
283                 ifp->timer.function = addrconf_rs_timer;
284                 break;
285         default:
286                 break;
287         }
288         ifp->timer.expires = jiffies + when;
289         add_timer(&ifp->timer);
290 }
291
292 static int snmp6_alloc_dev(struct inet6_dev *idev)
293 {
294         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
295                           sizeof(struct ipstats_mib),
296                           __alignof__(struct ipstats_mib)) < 0)
297                 goto err_ip;
298         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
299                                         GFP_KERNEL);
300         if (!idev->stats.icmpv6dev)
301                 goto err_icmp;
302         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
303                                            GFP_KERNEL);
304         if (!idev->stats.icmpv6msgdev)
305                 goto err_icmpmsg;
306
307         return 0;
308
309 err_icmpmsg:
310         kfree(idev->stats.icmpv6dev);
311 err_icmp:
312         snmp_mib_free((void __percpu **)idev->stats.ipv6);
313 err_ip:
314         return -ENOMEM;
315 }
316
317 static void snmp6_free_dev(struct inet6_dev *idev)
318 {
319         kfree(idev->stats.icmpv6msgdev);
320         kfree(idev->stats.icmpv6dev);
321         snmp_mib_free((void __percpu **)idev->stats.ipv6);
322 }
323
324 /* Nobody refers to this device, we may destroy it. */
325
326 void in6_dev_finish_destroy(struct inet6_dev *idev)
327 {
328         struct net_device *dev = idev->dev;
329
330         WARN_ON(!list_empty(&idev->addr_list));
331         WARN_ON(idev->mc_list != NULL);
332
333 #ifdef NET_REFCNT_DEBUG
334         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
335 #endif
336         dev_put(dev);
337         if (!idev->dead) {
338                 pr_warn("Freeing alive inet6 device %p\n", idev);
339                 return;
340         }
341         snmp6_free_dev(idev);
342         kfree_rcu(idev, rcu);
343 }
344 EXPORT_SYMBOL(in6_dev_finish_destroy);
345
346 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
347 {
348         struct inet6_dev *ndev;
349
350         ASSERT_RTNL();
351
352         if (dev->mtu < IPV6_MIN_MTU)
353                 return NULL;
354
355         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
356
357         if (ndev == NULL)
358                 return NULL;
359
360         rwlock_init(&ndev->lock);
361         ndev->dev = dev;
362         INIT_LIST_HEAD(&ndev->addr_list);
363
364         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
365         ndev->cnf.mtu6 = dev->mtu;
366         ndev->cnf.sysctl = NULL;
367         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
368         if (ndev->nd_parms == NULL) {
369                 kfree(ndev);
370                 return NULL;
371         }
372         if (ndev->cnf.forwarding)
373                 dev_disable_lro(dev);
374         /* We refer to the device */
375         dev_hold(dev);
376
377         if (snmp6_alloc_dev(ndev) < 0) {
378                 ADBG((KERN_WARNING
379                         "%s: cannot allocate memory for statistics; dev=%s.\n",
380                         __func__, dev->name));
381                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
382                 dev_put(dev);
383                 kfree(ndev);
384                 return NULL;
385         }
386
387         if (snmp6_register_dev(ndev) < 0) {
388                 ADBG((KERN_WARNING
389                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
390                         __func__, dev->name));
391                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
392                 ndev->dead = 1;
393                 in6_dev_finish_destroy(ndev);
394                 return NULL;
395         }
396
397         /* One reference from device.  We must do this before
398          * we invoke __ipv6_regen_rndid().
399          */
400         in6_dev_hold(ndev);
401
402         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
403                 ndev->cnf.accept_dad = -1;
404
405 #if IS_ENABLED(CONFIG_IPV6_SIT)
406         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
407                 pr_info("%s: Disabled Multicast RS\n", dev->name);
408                 ndev->cnf.rtr_solicits = 0;
409         }
410 #endif
411
412 #ifdef CONFIG_IPV6_PRIVACY
413         INIT_LIST_HEAD(&ndev->tempaddr_list);
414         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
415         if ((dev->flags&IFF_LOOPBACK) ||
416             dev->type == ARPHRD_TUNNEL ||
417             dev->type == ARPHRD_TUNNEL6 ||
418             dev->type == ARPHRD_SIT ||
419             dev->type == ARPHRD_NONE) {
420                 ndev->cnf.use_tempaddr = -1;
421         } else {
422                 in6_dev_hold(ndev);
423                 ipv6_regen_rndid((unsigned long) ndev);
424         }
425 #endif
426
427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
428                 ndev->if_flags |= IF_READY;
429
430         ipv6_mc_init_dev(ndev);
431         ndev->tstamp = jiffies;
432         addrconf_sysctl_register(ndev);
433         /* protected by rtnl_lock */
434         rcu_assign_pointer(dev->ip6_ptr, ndev);
435
436         /* Join all-node multicast group */
437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
438
439         /* Join all-router multicast group if forwarding is set */
440         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442
443         return ndev;
444 }
445
446 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
447 {
448         struct inet6_dev *idev;
449
450         ASSERT_RTNL();
451
452         idev = __in6_dev_get(dev);
453         if (!idev) {
454                 idev = ipv6_add_dev(dev);
455                 if (!idev)
456                         return NULL;
457         }
458
459         if (dev->flags&IFF_UP)
460                 ipv6_mc_up(idev);
461         return idev;
462 }
463
464 static int inet6_netconf_msgsize_devconf(int type)
465 {
466         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
467                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
468
469         /* type -1 is used for ALL */
470         if (type == -1 || type == NETCONFA_FORWARDING)
471                 size += nla_total_size(4);
472 #ifdef CONFIG_IPV6_MROUTE
473         if (type == -1 || type == NETCONFA_MC_FORWARDING)
474                 size += nla_total_size(4);
475 #endif
476
477         return size;
478 }
479
480 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
481                                       struct ipv6_devconf *devconf, u32 portid,
482                                       u32 seq, int event, unsigned int flags,
483                                       int type)
484 {
485         struct nlmsghdr  *nlh;
486         struct netconfmsg *ncm;
487
488         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
489                         flags);
490         if (nlh == NULL)
491                 return -EMSGSIZE;
492
493         ncm = nlmsg_data(nlh);
494         ncm->ncm_family = AF_INET6;
495
496         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
497                 goto nla_put_failure;
498
499         /* type -1 is used for ALL */
500         if ((type == -1 || type == NETCONFA_FORWARDING) &&
501             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
502                 goto nla_put_failure;
503 #ifdef CONFIG_IPV6_MROUTE
504         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
505             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
506                         devconf->mc_forwarding) < 0)
507                 goto nla_put_failure;
508 #endif
509         return nlmsg_end(skb, nlh);
510
511 nla_put_failure:
512         nlmsg_cancel(skb, nlh);
513         return -EMSGSIZE;
514 }
515
516 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
517                                   struct ipv6_devconf *devconf)
518 {
519         struct sk_buff *skb;
520         int err = -ENOBUFS;
521
522         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
523         if (skb == NULL)
524                 goto errout;
525
526         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
527                                          RTM_NEWNETCONF, 0, type);
528         if (err < 0) {
529                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
530                 WARN_ON(err == -EMSGSIZE);
531                 kfree_skb(skb);
532                 goto errout;
533         }
534         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
535         return;
536 errout:
537         if (err < 0)
538                 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
539 }
540
541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
542         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
543         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
544 };
545
546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
547                                      struct nlmsghdr *nlh,
548                                      void *arg)
549 {
550         struct net *net = sock_net(in_skb->sk);
551         struct nlattr *tb[NETCONFA_MAX+1];
552         struct netconfmsg *ncm;
553         struct sk_buff *skb;
554         struct ipv6_devconf *devconf;
555         struct inet6_dev *in6_dev;
556         struct net_device *dev;
557         int ifindex;
558         int err;
559
560         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
561                           devconf_ipv6_policy);
562         if (err < 0)
563                 goto errout;
564
565         err = EINVAL;
566         if (!tb[NETCONFA_IFINDEX])
567                 goto errout;
568
569         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
570         switch (ifindex) {
571         case NETCONFA_IFINDEX_ALL:
572                 devconf = net->ipv6.devconf_all;
573                 break;
574         case NETCONFA_IFINDEX_DEFAULT:
575                 devconf = net->ipv6.devconf_dflt;
576                 break;
577         default:
578                 dev = __dev_get_by_index(net, ifindex);
579                 if (dev == NULL)
580                         goto errout;
581                 in6_dev = __in6_dev_get(dev);
582                 if (in6_dev == NULL)
583                         goto errout;
584                 devconf = &in6_dev->cnf;
585                 break;
586         }
587
588         err = -ENOBUFS;
589         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
590         if (skb == NULL)
591                 goto errout;
592
593         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
594                                          NETLINK_CB(in_skb).portid,
595                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
596                                          -1);
597         if (err < 0) {
598                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
599                 WARN_ON(err == -EMSGSIZE);
600                 kfree_skb(skb);
601                 goto errout;
602         }
603         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
604 errout:
605         return err;
606 }
607
608 #ifdef CONFIG_SYSCTL
609 static void dev_forward_change(struct inet6_dev *idev)
610 {
611         struct net_device *dev;
612         struct inet6_ifaddr *ifa;
613
614         if (!idev)
615                 return;
616         dev = idev->dev;
617         if (idev->cnf.forwarding)
618                 dev_disable_lro(dev);
619         if (dev->flags & IFF_MULTICAST) {
620                 if (idev->cnf.forwarding)
621                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
622                 else
623                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
624         }
625
626         list_for_each_entry(ifa, &idev->addr_list, if_list) {
627                 if (ifa->flags&IFA_F_TENTATIVE)
628                         continue;
629                 if (idev->cnf.forwarding)
630                         addrconf_join_anycast(ifa);
631                 else
632                         addrconf_leave_anycast(ifa);
633         }
634         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
635                                      dev->ifindex, &idev->cnf);
636 }
637
638
639 static void addrconf_forward_change(struct net *net, __s32 newf)
640 {
641         struct net_device *dev;
642         struct inet6_dev *idev;
643
644         for_each_netdev(net, dev) {
645                 idev = __in6_dev_get(dev);
646                 if (idev) {
647                         int changed = (!idev->cnf.forwarding) ^ (!newf);
648                         idev->cnf.forwarding = newf;
649                         if (changed)
650                                 dev_forward_change(idev);
651                 }
652         }
653 }
654
655 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
656 {
657         struct net *net;
658         int old;
659
660         if (!rtnl_trylock())
661                 return restart_syscall();
662
663         net = (struct net *)table->extra2;
664         old = *p;
665         *p = newf;
666
667         if (p == &net->ipv6.devconf_dflt->forwarding) {
668                 if ((!newf) ^ (!old))
669                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
670                                                      NETCONFA_IFINDEX_DEFAULT,
671                                                      net->ipv6.devconf_dflt);
672                 rtnl_unlock();
673                 return 0;
674         }
675
676         if (p == &net->ipv6.devconf_all->forwarding) {
677                 net->ipv6.devconf_dflt->forwarding = newf;
678                 addrconf_forward_change(net, newf);
679                 if ((!newf) ^ (!old))
680                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
681                                                      NETCONFA_IFINDEX_ALL,
682                                                      net->ipv6.devconf_all);
683         } else if ((!newf) ^ (!old))
684                 dev_forward_change((struct inet6_dev *)table->extra1);
685         rtnl_unlock();
686
687         if (newf)
688                 rt6_purge_dflt_routers(net);
689         return 1;
690 }
691 #endif
692
693 /* Nobody refers to this ifaddr, destroy it */
694 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
695 {
696         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
697
698 #ifdef NET_REFCNT_DEBUG
699         pr_debug("%s\n", __func__);
700 #endif
701
702         in6_dev_put(ifp->idev);
703
704         if (del_timer(&ifp->timer))
705                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
706
707         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
708                 pr_warn("Freeing alive inet6 address %p\n", ifp);
709                 return;
710         }
711         ip6_rt_put(ifp->rt);
712
713         kfree_rcu(ifp, rcu);
714 }
715
716 static void
717 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
718 {
719         struct list_head *p;
720         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
721
722         /*
723          * Each device address list is sorted in order of scope -
724          * global before linklocal.
725          */
726         list_for_each(p, &idev->addr_list) {
727                 struct inet6_ifaddr *ifa
728                         = list_entry(p, struct inet6_ifaddr, if_list);
729                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
730                         break;
731         }
732
733         list_add_tail(&ifp->if_list, p);
734 }
735
736 static u32 inet6_addr_hash(const struct in6_addr *addr)
737 {
738         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
739 }
740
741 /* On success it returns ifp with increased reference count */
742
743 static struct inet6_ifaddr *
744 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
745               int scope, u32 flags)
746 {
747         struct inet6_ifaddr *ifa = NULL;
748         struct rt6_info *rt;
749         unsigned int hash;
750         int err = 0;
751         int addr_type = ipv6_addr_type(addr);
752
753         if (addr_type == IPV6_ADDR_ANY ||
754             addr_type & IPV6_ADDR_MULTICAST ||
755             (!(idev->dev->flags & IFF_LOOPBACK) &&
756              addr_type & IPV6_ADDR_LOOPBACK))
757                 return ERR_PTR(-EADDRNOTAVAIL);
758
759         rcu_read_lock_bh();
760         if (idev->dead) {
761                 err = -ENODEV;                  /*XXX*/
762                 goto out2;
763         }
764
765         if (idev->cnf.disable_ipv6) {
766                 err = -EACCES;
767                 goto out2;
768         }
769
770         spin_lock(&addrconf_hash_lock);
771
772         /* Ignore adding duplicate addresses on an interface */
773         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
774                 ADBG(("ipv6_add_addr: already assigned\n"));
775                 err = -EEXIST;
776                 goto out;
777         }
778
779         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
780
781         if (ifa == NULL) {
782                 ADBG(("ipv6_add_addr: malloc failed\n"));
783                 err = -ENOBUFS;
784                 goto out;
785         }
786
787         rt = addrconf_dst_alloc(idev, addr, false);
788         if (IS_ERR(rt)) {
789                 err = PTR_ERR(rt);
790                 goto out;
791         }
792
793         ifa->addr = *addr;
794
795         spin_lock_init(&ifa->lock);
796         spin_lock_init(&ifa->state_lock);
797         init_timer(&ifa->timer);
798         INIT_HLIST_NODE(&ifa->addr_lst);
799         ifa->timer.data = (unsigned long) ifa;
800         ifa->scope = scope;
801         ifa->prefix_len = pfxlen;
802         ifa->flags = flags | IFA_F_TENTATIVE;
803         ifa->cstamp = ifa->tstamp = jiffies;
804
805         ifa->rt = rt;
806
807         ifa->idev = idev;
808         in6_dev_hold(idev);
809         /* For caller */
810         in6_ifa_hold(ifa);
811
812         /* Add to big hash table */
813         hash = inet6_addr_hash(addr);
814
815         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
816         spin_unlock(&addrconf_hash_lock);
817
818         write_lock(&idev->lock);
819         /* Add to inet6_dev unicast addr list. */
820         ipv6_link_dev_addr(idev, ifa);
821
822 #ifdef CONFIG_IPV6_PRIVACY
823         if (ifa->flags&IFA_F_TEMPORARY) {
824                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
825                 in6_ifa_hold(ifa);
826         }
827 #endif
828
829         in6_ifa_hold(ifa);
830         write_unlock(&idev->lock);
831 out2:
832         rcu_read_unlock_bh();
833
834         if (likely(err == 0))
835                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
836         else {
837                 kfree(ifa);
838                 ifa = ERR_PTR(err);
839         }
840
841         return ifa;
842 out:
843         spin_unlock(&addrconf_hash_lock);
844         goto out2;
845 }
846
847 /* This function wants to get referenced ifp and releases it before return */
848
849 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
850 {
851         struct inet6_ifaddr *ifa, *ifn;
852         struct inet6_dev *idev = ifp->idev;
853         int state;
854         int deleted = 0, onlink = 0;
855         unsigned long expires = jiffies;
856
857         spin_lock_bh(&ifp->state_lock);
858         state = ifp->state;
859         ifp->state = INET6_IFADDR_STATE_DEAD;
860         spin_unlock_bh(&ifp->state_lock);
861
862         if (state == INET6_IFADDR_STATE_DEAD)
863                 goto out;
864
865         spin_lock_bh(&addrconf_hash_lock);
866         hlist_del_init_rcu(&ifp->addr_lst);
867         spin_unlock_bh(&addrconf_hash_lock);
868
869         write_lock_bh(&idev->lock);
870 #ifdef CONFIG_IPV6_PRIVACY
871         if (ifp->flags&IFA_F_TEMPORARY) {
872                 list_del(&ifp->tmp_list);
873                 if (ifp->ifpub) {
874                         in6_ifa_put(ifp->ifpub);
875                         ifp->ifpub = NULL;
876                 }
877                 __in6_ifa_put(ifp);
878         }
879 #endif
880
881         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
882                 if (ifa == ifp) {
883                         list_del_init(&ifp->if_list);
884                         __in6_ifa_put(ifp);
885
886                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
887                                 break;
888                         deleted = 1;
889                         continue;
890                 } else if (ifp->flags & IFA_F_PERMANENT) {
891                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
892                                               ifp->prefix_len)) {
893                                 if (ifa->flags & IFA_F_PERMANENT) {
894                                         onlink = 1;
895                                         if (deleted)
896                                                 break;
897                                 } else {
898                                         unsigned long lifetime;
899
900                                         if (!onlink)
901                                                 onlink = -1;
902
903                                         spin_lock(&ifa->lock);
904
905                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
906                                         /*
907                                          * Note: Because this address is
908                                          * not permanent, lifetime <
909                                          * LONG_MAX / HZ here.
910                                          */
911                                         if (time_before(expires,
912                                                         ifa->tstamp + lifetime * HZ))
913                                                 expires = ifa->tstamp + lifetime * HZ;
914                                         spin_unlock(&ifa->lock);
915                                 }
916                         }
917                 }
918         }
919         write_unlock_bh(&idev->lock);
920
921         addrconf_del_timer(ifp);
922
923         ipv6_ifa_notify(RTM_DELADDR, ifp);
924
925         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
926
927         /*
928          * Purge or update corresponding prefix
929          *
930          * 1) we don't purge prefix here if address was not permanent.
931          *    prefix is managed by its own lifetime.
932          * 2) if there're no addresses, delete prefix.
933          * 3) if there're still other permanent address(es),
934          *    corresponding prefix is still permanent.
935          * 4) otherwise, update prefix lifetime to the
936          *    longest valid lifetime among the corresponding
937          *    addresses on the device.
938          *    Note: subsequent RA will update lifetime.
939          *
940          * --yoshfuji
941          */
942         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
943                 struct in6_addr prefix;
944                 struct rt6_info *rt;
945                 struct net *net = dev_net(ifp->idev->dev);
946                 struct flowi6 fl6 = {};
947
948                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
949                 fl6.flowi6_oif = ifp->idev->dev->ifindex;
950                 fl6.daddr = prefix;
951                 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
952                                                          RT6_LOOKUP_F_IFACE);
953
954                 if (rt != net->ipv6.ip6_null_entry &&
955                     addrconf_is_prefix_route(rt)) {
956                         if (onlink == 0) {
957                                 ip6_del_rt(rt);
958                                 rt = NULL;
959                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
960                                 rt6_set_expires(rt, expires);
961                         }
962                 }
963                 ip6_rt_put(rt);
964         }
965
966         /* clean up prefsrc entries */
967         rt6_remove_prefsrc(ifp);
968 out:
969         in6_ifa_put(ifp);
970 }
971
972 #ifdef CONFIG_IPV6_PRIVACY
973 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
974 {
975         struct inet6_dev *idev = ifp->idev;
976         struct in6_addr addr, *tmpaddr;
977         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
978         unsigned long regen_advance;
979         int tmp_plen;
980         int ret = 0;
981         int max_addresses;
982         u32 addr_flags;
983         unsigned long now = jiffies;
984
985         write_lock(&idev->lock);
986         if (ift) {
987                 spin_lock_bh(&ift->lock);
988                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
989                 spin_unlock_bh(&ift->lock);
990                 tmpaddr = &addr;
991         } else {
992                 tmpaddr = NULL;
993         }
994 retry:
995         in6_dev_hold(idev);
996         if (idev->cnf.use_tempaddr <= 0) {
997                 write_unlock(&idev->lock);
998                 pr_info("%s: use_tempaddr is disabled\n", __func__);
999                 in6_dev_put(idev);
1000                 ret = -1;
1001                 goto out;
1002         }
1003         spin_lock_bh(&ifp->lock);
1004         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1005                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1006                 spin_unlock_bh(&ifp->lock);
1007                 write_unlock(&idev->lock);
1008                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1009                         __func__);
1010                 in6_dev_put(idev);
1011                 ret = -1;
1012                 goto out;
1013         }
1014         in6_ifa_hold(ifp);
1015         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1016         __ipv6_try_regen_rndid(idev, tmpaddr);
1017         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1018         age = (now - ifp->tstamp) / HZ;
1019         tmp_valid_lft = min_t(__u32,
1020                               ifp->valid_lft,
1021                               idev->cnf.temp_valid_lft + age);
1022         tmp_prefered_lft = min_t(__u32,
1023                                  ifp->prefered_lft,
1024                                  idev->cnf.temp_prefered_lft + age -
1025                                  idev->cnf.max_desync_factor);
1026         tmp_plen = ifp->prefix_len;
1027         max_addresses = idev->cnf.max_addresses;
1028         tmp_tstamp = ifp->tstamp;
1029         spin_unlock_bh(&ifp->lock);
1030
1031         regen_advance = idev->cnf.regen_max_retry *
1032                         idev->cnf.dad_transmits *
1033                         idev->nd_parms->retrans_time / HZ;
1034         write_unlock(&idev->lock);
1035
1036         /* A temporary address is created only if this calculated Preferred
1037          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1038          * an implementation must not create a temporary address with a zero
1039          * Preferred Lifetime.
1040          */
1041         if (tmp_prefered_lft <= regen_advance) {
1042                 in6_ifa_put(ifp);
1043                 in6_dev_put(idev);
1044                 ret = -1;
1045                 goto out;
1046         }
1047
1048         addr_flags = IFA_F_TEMPORARY;
1049         /* set in addrconf_prefix_rcv() */
1050         if (ifp->flags & IFA_F_OPTIMISTIC)
1051                 addr_flags |= IFA_F_OPTIMISTIC;
1052
1053         ift = !max_addresses ||
1054               ipv6_count_addresses(idev) < max_addresses ?
1055                 ipv6_add_addr(idev, &addr, tmp_plen,
1056                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1057                               addr_flags) : NULL;
1058         if (!ift || IS_ERR(ift)) {
1059                 in6_ifa_put(ifp);
1060                 in6_dev_put(idev);
1061                 pr_info("%s: retry temporary address regeneration\n", __func__);
1062                 tmpaddr = &addr;
1063                 write_lock(&idev->lock);
1064                 goto retry;
1065         }
1066
1067         spin_lock_bh(&ift->lock);
1068         ift->ifpub = ifp;
1069         ift->valid_lft = tmp_valid_lft;
1070         ift->prefered_lft = tmp_prefered_lft;
1071         ift->cstamp = now;
1072         ift->tstamp = tmp_tstamp;
1073         spin_unlock_bh(&ift->lock);
1074
1075         addrconf_dad_start(ift);
1076         in6_ifa_put(ift);
1077         in6_dev_put(idev);
1078 out:
1079         return ret;
1080 }
1081 #endif
1082
1083 /*
1084  *      Choose an appropriate source address (RFC3484)
1085  */
1086 enum {
1087         IPV6_SADDR_RULE_INIT = 0,
1088         IPV6_SADDR_RULE_LOCAL,
1089         IPV6_SADDR_RULE_SCOPE,
1090         IPV6_SADDR_RULE_PREFERRED,
1091 #ifdef CONFIG_IPV6_MIP6
1092         IPV6_SADDR_RULE_HOA,
1093 #endif
1094         IPV6_SADDR_RULE_OIF,
1095         IPV6_SADDR_RULE_LABEL,
1096 #ifdef CONFIG_IPV6_PRIVACY
1097         IPV6_SADDR_RULE_PRIVACY,
1098 #endif
1099         IPV6_SADDR_RULE_ORCHID,
1100         IPV6_SADDR_RULE_PREFIX,
1101         IPV6_SADDR_RULE_MAX
1102 };
1103
1104 struct ipv6_saddr_score {
1105         int                     rule;
1106         int                     addr_type;
1107         struct inet6_ifaddr     *ifa;
1108         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1109         int                     scopedist;
1110         int                     matchlen;
1111 };
1112
1113 struct ipv6_saddr_dst {
1114         const struct in6_addr *addr;
1115         int ifindex;
1116         int scope;
1117         int label;
1118         unsigned int prefs;
1119 };
1120
1121 static inline int ipv6_saddr_preferred(int type)
1122 {
1123         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1124                 return 1;
1125         return 0;
1126 }
1127
1128 static int ipv6_get_saddr_eval(struct net *net,
1129                                struct ipv6_saddr_score *score,
1130                                struct ipv6_saddr_dst *dst,
1131                                int i)
1132 {
1133         int ret;
1134
1135         if (i <= score->rule) {
1136                 switch (i) {
1137                 case IPV6_SADDR_RULE_SCOPE:
1138                         ret = score->scopedist;
1139                         break;
1140                 case IPV6_SADDR_RULE_PREFIX:
1141                         ret = score->matchlen;
1142                         break;
1143                 default:
1144                         ret = !!test_bit(i, score->scorebits);
1145                 }
1146                 goto out;
1147         }
1148
1149         switch (i) {
1150         case IPV6_SADDR_RULE_INIT:
1151                 /* Rule 0: remember if hiscore is not ready yet */
1152                 ret = !!score->ifa;
1153                 break;
1154         case IPV6_SADDR_RULE_LOCAL:
1155                 /* Rule 1: Prefer same address */
1156                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1157                 break;
1158         case IPV6_SADDR_RULE_SCOPE:
1159                 /* Rule 2: Prefer appropriate scope
1160                  *
1161                  *      ret
1162                  *       ^
1163                  *    -1 |  d 15
1164                  *    ---+--+-+---> scope
1165                  *       |
1166                  *       |             d is scope of the destination.
1167                  *  B-d  |  \
1168                  *       |   \      <- smaller scope is better if
1169                  *  B-15 |    \        if scope is enough for destinaion.
1170                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1171                  * d-C-1 | /
1172                  *       |/         <- greater is better
1173                  *   -C  /             if scope is not enough for destination.
1174                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1175                  *
1176                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1177                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1178                  * Assume B = 0 and we get C > 29.
1179                  */
1180                 ret = __ipv6_addr_src_scope(score->addr_type);
1181                 if (ret >= dst->scope)
1182                         ret = -ret;
1183                 else
1184                         ret -= 128;     /* 30 is enough */
1185                 score->scopedist = ret;
1186                 break;
1187         case IPV6_SADDR_RULE_PREFERRED:
1188                 /* Rule 3: Avoid deprecated and optimistic addresses */
1189                 ret = ipv6_saddr_preferred(score->addr_type) ||
1190                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1191                 break;
1192 #ifdef CONFIG_IPV6_MIP6
1193         case IPV6_SADDR_RULE_HOA:
1194             {
1195                 /* Rule 4: Prefer home address */
1196                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1197                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1198                 break;
1199             }
1200 #endif
1201         case IPV6_SADDR_RULE_OIF:
1202                 /* Rule 5: Prefer outgoing interface */
1203                 ret = (!dst->ifindex ||
1204                        dst->ifindex == score->ifa->idev->dev->ifindex);
1205                 break;
1206         case IPV6_SADDR_RULE_LABEL:
1207                 /* Rule 6: Prefer matching label */
1208                 ret = ipv6_addr_label(net,
1209                                       &score->ifa->addr, score->addr_type,
1210                                       score->ifa->idev->dev->ifindex) == dst->label;
1211                 break;
1212 #ifdef CONFIG_IPV6_PRIVACY
1213         case IPV6_SADDR_RULE_PRIVACY:
1214             {
1215                 /* Rule 7: Prefer public address
1216                  * Note: prefer temporary address if use_tempaddr >= 2
1217                  */
1218                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1219                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1220                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1221                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1222                 break;
1223             }
1224 #endif
1225         case IPV6_SADDR_RULE_ORCHID:
1226                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1227                  *          non-ORCHID vs non-ORCHID
1228                  */
1229                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1230                         ipv6_addr_orchid(dst->addr));
1231                 break;
1232         case IPV6_SADDR_RULE_PREFIX:
1233                 /* Rule 8: Use longest matching prefix */
1234                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1235                 if (ret > score->ifa->prefix_len)
1236                         ret = score->ifa->prefix_len;
1237                 score->matchlen = ret;
1238                 break;
1239         default:
1240                 ret = 0;
1241         }
1242
1243         if (ret)
1244                 __set_bit(i, score->scorebits);
1245         score->rule = i;
1246 out:
1247         return ret;
1248 }
1249
1250 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1251                        const struct in6_addr *daddr, unsigned int prefs,
1252                        struct in6_addr *saddr)
1253 {
1254         struct ipv6_saddr_score scores[2],
1255                                 *score = &scores[0], *hiscore = &scores[1];
1256         struct ipv6_saddr_dst dst;
1257         struct net_device *dev;
1258         int dst_type;
1259
1260         dst_type = __ipv6_addr_type(daddr);
1261         dst.addr = daddr;
1262         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1263         dst.scope = __ipv6_addr_src_scope(dst_type);
1264         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1265         dst.prefs = prefs;
1266
1267         hiscore->rule = -1;
1268         hiscore->ifa = NULL;
1269
1270         rcu_read_lock();
1271
1272         for_each_netdev_rcu(net, dev) {
1273                 struct inet6_dev *idev;
1274
1275                 /* Candidate Source Address (section 4)
1276                  *  - multicast and link-local destination address,
1277                  *    the set of candidate source address MUST only
1278                  *    include addresses assigned to interfaces
1279                  *    belonging to the same link as the outgoing
1280                  *    interface.
1281                  * (- For site-local destination addresses, the
1282                  *    set of candidate source addresses MUST only
1283                  *    include addresses assigned to interfaces
1284                  *    belonging to the same site as the outgoing
1285                  *    interface.)
1286                  */
1287                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1288                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1289                     dst.ifindex && dev->ifindex != dst.ifindex)
1290                         continue;
1291
1292                 idev = __in6_dev_get(dev);
1293                 if (!idev)
1294                         continue;
1295
1296                 read_lock_bh(&idev->lock);
1297                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1298                         int i;
1299
1300                         /*
1301                          * - Tentative Address (RFC2462 section 5.4)
1302                          *  - A tentative address is not considered
1303                          *    "assigned to an interface" in the traditional
1304                          *    sense, unless it is also flagged as optimistic.
1305                          * - Candidate Source Address (section 4)
1306                          *  - In any case, anycast addresses, multicast
1307                          *    addresses, and the unspecified address MUST
1308                          *    NOT be included in a candidate set.
1309                          */
1310                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1311                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1312                                 continue;
1313
1314                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1315
1316                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1317                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1318                                 LIMIT_NETDEBUG(KERN_DEBUG
1319                                                "ADDRCONF: unspecified / multicast address "
1320                                                "assigned as unicast address on %s",
1321                                                dev->name);
1322                                 continue;
1323                         }
1324
1325                         score->rule = -1;
1326                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1327
1328                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1329                                 int minihiscore, miniscore;
1330
1331                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1332                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1333
1334                                 if (minihiscore > miniscore) {
1335                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1336                                             score->scopedist > 0) {
1337                                                 /*
1338                                                  * special case:
1339                                                  * each remaining entry
1340                                                  * has too small (not enough)
1341                                                  * scope, because ifa entries
1342                                                  * are sorted by their scope
1343                                                  * values.
1344                                                  */
1345                                                 goto try_nextdev;
1346                                         }
1347                                         break;
1348                                 } else if (minihiscore < miniscore) {
1349                                         if (hiscore->ifa)
1350                                                 in6_ifa_put(hiscore->ifa);
1351
1352                                         in6_ifa_hold(score->ifa);
1353
1354                                         swap(hiscore, score);
1355
1356                                         /* restore our iterator */
1357                                         score->ifa = hiscore->ifa;
1358
1359                                         break;
1360                                 }
1361                         }
1362                 }
1363 try_nextdev:
1364                 read_unlock_bh(&idev->lock);
1365         }
1366         rcu_read_unlock();
1367
1368         if (!hiscore->ifa)
1369                 return -EADDRNOTAVAIL;
1370
1371         *saddr = hiscore->ifa->addr;
1372         in6_ifa_put(hiscore->ifa);
1373         return 0;
1374 }
1375 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1376
1377 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1378                     unsigned char banned_flags)
1379 {
1380         struct inet6_dev *idev;
1381         int err = -EADDRNOTAVAIL;
1382
1383         rcu_read_lock();
1384         idev = __in6_dev_get(dev);
1385         if (idev) {
1386                 struct inet6_ifaddr *ifp;
1387
1388                 read_lock_bh(&idev->lock);
1389                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1390                         if (ifp->scope == IFA_LINK &&
1391                             !(ifp->flags & banned_flags)) {
1392                                 *addr = ifp->addr;
1393                                 err = 0;
1394                                 break;
1395                         }
1396                 }
1397                 read_unlock_bh(&idev->lock);
1398         }
1399         rcu_read_unlock();
1400         return err;
1401 }
1402
1403 static int ipv6_count_addresses(struct inet6_dev *idev)
1404 {
1405         int cnt = 0;
1406         struct inet6_ifaddr *ifp;
1407
1408         read_lock_bh(&idev->lock);
1409         list_for_each_entry(ifp, &idev->addr_list, if_list)
1410                 cnt++;
1411         read_unlock_bh(&idev->lock);
1412         return cnt;
1413 }
1414
1415 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1416                   struct net_device *dev, int strict)
1417 {
1418         struct inet6_ifaddr *ifp;
1419         struct hlist_node *node;
1420         unsigned int hash = inet6_addr_hash(addr);
1421
1422         rcu_read_lock_bh();
1423         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1424                 if (!net_eq(dev_net(ifp->idev->dev), net))
1425                         continue;
1426                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1427                     !(ifp->flags&IFA_F_TENTATIVE) &&
1428                     (dev == NULL || ifp->idev->dev == dev ||
1429                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1430                         rcu_read_unlock_bh();
1431                         return 1;
1432                 }
1433         }
1434
1435         rcu_read_unlock_bh();
1436         return 0;
1437 }
1438 EXPORT_SYMBOL(ipv6_chk_addr);
1439
1440 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1441                                struct net_device *dev)
1442 {
1443         unsigned int hash = inet6_addr_hash(addr);
1444         struct inet6_ifaddr *ifp;
1445         struct hlist_node *node;
1446
1447         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1448                 if (!net_eq(dev_net(ifp->idev->dev), net))
1449                         continue;
1450                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1451                         if (dev == NULL || ifp->idev->dev == dev)
1452                                 return true;
1453                 }
1454         }
1455         return false;
1456 }
1457
1458 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1459 {
1460         struct inet6_dev *idev;
1461         struct inet6_ifaddr *ifa;
1462         int     onlink;
1463
1464         onlink = 0;
1465         rcu_read_lock();
1466         idev = __in6_dev_get(dev);
1467         if (idev) {
1468                 read_lock_bh(&idev->lock);
1469                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1470                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1471                                                    ifa->prefix_len);
1472                         if (onlink)
1473                                 break;
1474                 }
1475                 read_unlock_bh(&idev->lock);
1476         }
1477         rcu_read_unlock();
1478         return onlink;
1479 }
1480 EXPORT_SYMBOL(ipv6_chk_prefix);
1481
1482 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1483                                      struct net_device *dev, int strict)
1484 {
1485         struct inet6_ifaddr *ifp, *result = NULL;
1486         unsigned int hash = inet6_addr_hash(addr);
1487         struct hlist_node *node;
1488
1489         rcu_read_lock_bh();
1490         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1491                 if (!net_eq(dev_net(ifp->idev->dev), net))
1492                         continue;
1493                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1494                         if (dev == NULL || ifp->idev->dev == dev ||
1495                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1496                                 result = ifp;
1497                                 in6_ifa_hold(ifp);
1498                                 break;
1499                         }
1500                 }
1501         }
1502         rcu_read_unlock_bh();
1503
1504         return result;
1505 }
1506
1507 /* Gets referenced address, destroys ifaddr */
1508
1509 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1510 {
1511         if (ifp->flags&IFA_F_PERMANENT) {
1512                 spin_lock_bh(&ifp->lock);
1513                 addrconf_del_timer(ifp);
1514                 ifp->flags |= IFA_F_TENTATIVE;
1515                 if (dad_failed)
1516                         ifp->flags |= IFA_F_DADFAILED;
1517                 spin_unlock_bh(&ifp->lock);
1518                 if (dad_failed)
1519                         ipv6_ifa_notify(0, ifp);
1520                 in6_ifa_put(ifp);
1521 #ifdef CONFIG_IPV6_PRIVACY
1522         } else if (ifp->flags&IFA_F_TEMPORARY) {
1523                 struct inet6_ifaddr *ifpub;
1524                 spin_lock_bh(&ifp->lock);
1525                 ifpub = ifp->ifpub;
1526                 if (ifpub) {
1527                         in6_ifa_hold(ifpub);
1528                         spin_unlock_bh(&ifp->lock);
1529                         ipv6_create_tempaddr(ifpub, ifp);
1530                         in6_ifa_put(ifpub);
1531                 } else {
1532                         spin_unlock_bh(&ifp->lock);
1533                 }
1534                 ipv6_del_addr(ifp);
1535 #endif
1536         } else
1537                 ipv6_del_addr(ifp);
1538 }
1539
1540 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1541 {
1542         int err = -ENOENT;
1543
1544         spin_lock(&ifp->state_lock);
1545         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1546                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1547                 err = 0;
1548         }
1549         spin_unlock(&ifp->state_lock);
1550
1551         return err;
1552 }
1553
1554 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1555 {
1556         struct inet6_dev *idev = ifp->idev;
1557
1558         if (addrconf_dad_end(ifp)) {
1559                 in6_ifa_put(ifp);
1560                 return;
1561         }
1562
1563         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1564                              ifp->idev->dev->name, &ifp->addr);
1565
1566         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1567                 struct in6_addr addr;
1568
1569                 addr.s6_addr32[0] = htonl(0xfe800000);
1570                 addr.s6_addr32[1] = 0;
1571
1572                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1573                     ipv6_addr_equal(&ifp->addr, &addr)) {
1574                         /* DAD failed for link-local based on MAC address */
1575                         idev->cnf.disable_ipv6 = 1;
1576
1577                         pr_info("%s: IPv6 being disabled!\n",
1578                                 ifp->idev->dev->name);
1579                 }
1580         }
1581
1582         addrconf_dad_stop(ifp, 1);
1583 }
1584
1585 /* Join to solicited addr multicast group. */
1586
1587 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1588 {
1589         struct in6_addr maddr;
1590
1591         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1592                 return;
1593
1594         addrconf_addr_solict_mult(addr, &maddr);
1595         ipv6_dev_mc_inc(dev, &maddr);
1596 }
1597
1598 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1599 {
1600         struct in6_addr maddr;
1601
1602         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1603                 return;
1604
1605         addrconf_addr_solict_mult(addr, &maddr);
1606         __ipv6_dev_mc_dec(idev, &maddr);
1607 }
1608
1609 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1610 {
1611         struct in6_addr addr;
1612         if (ifp->prefix_len == 127) /* RFC 6164 */
1613                 return;
1614         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1615         if (ipv6_addr_any(&addr))
1616                 return;
1617         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1618 }
1619
1620 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1621 {
1622         struct in6_addr addr;
1623         if (ifp->prefix_len == 127) /* RFC 6164 */
1624                 return;
1625         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1626         if (ipv6_addr_any(&addr))
1627                 return;
1628         __ipv6_dev_ac_dec(ifp->idev, &addr);
1629 }
1630
1631 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1632 {
1633         if (dev->addr_len != ETH_ALEN)
1634                 return -1;
1635         memcpy(eui, dev->dev_addr, 3);
1636         memcpy(eui + 5, dev->dev_addr + 3, 3);
1637
1638         /*
1639          * The zSeries OSA network cards can be shared among various
1640          * OS instances, but the OSA cards have only one MAC address.
1641          * This leads to duplicate address conflicts in conjunction
1642          * with IPv6 if more than one instance uses the same card.
1643          *
1644          * The driver for these cards can deliver a unique 16-bit
1645          * identifier for each instance sharing the same card.  It is
1646          * placed instead of 0xFFFE in the interface identifier.  The
1647          * "u" bit of the interface identifier is not inverted in this
1648          * case.  Hence the resulting interface identifier has local
1649          * scope according to RFC2373.
1650          */
1651         if (dev->dev_id) {
1652                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1653                 eui[4] = dev->dev_id & 0xFF;
1654         } else {
1655                 eui[3] = 0xFF;
1656                 eui[4] = 0xFE;
1657                 eui[0] ^= 2;
1658         }
1659         return 0;
1660 }
1661
1662 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1663 {
1664         if (dev->addr_len != IEEE802154_ADDR_LEN)
1665                 return -1;
1666         memcpy(eui, dev->dev_addr, 8);
1667         return 0;
1668 }
1669
1670 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1671 {
1672         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1673         if (dev->addr_len != ARCNET_ALEN)
1674                 return -1;
1675         memset(eui, 0, 7);
1676         eui[7] = *(u8 *)dev->dev_addr;
1677         return 0;
1678 }
1679
1680 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1681 {
1682         if (dev->addr_len != INFINIBAND_ALEN)
1683                 return -1;
1684         memcpy(eui, dev->dev_addr + 12, 8);
1685         eui[0] |= 2;
1686         return 0;
1687 }
1688
1689 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1690 {
1691         if (addr == 0)
1692                 return -1;
1693         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1694                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1695                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1696                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1697                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1698                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1699         eui[1] = 0;
1700         eui[2] = 0x5E;
1701         eui[3] = 0xFE;
1702         memcpy(eui + 4, &addr, 4);
1703         return 0;
1704 }
1705
1706 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1707 {
1708         if (dev->priv_flags & IFF_ISATAP)
1709                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1710         return -1;
1711 }
1712
1713 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1714 {
1715         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1716 }
1717
1718 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1719 {
1720         switch (dev->type) {
1721         case ARPHRD_ETHER:
1722         case ARPHRD_FDDI:
1723                 return addrconf_ifid_eui48(eui, dev);
1724         case ARPHRD_ARCNET:
1725                 return addrconf_ifid_arcnet(eui, dev);
1726         case ARPHRD_INFINIBAND:
1727                 return addrconf_ifid_infiniband(eui, dev);
1728         case ARPHRD_SIT:
1729                 return addrconf_ifid_sit(eui, dev);
1730         case ARPHRD_IPGRE:
1731                 return addrconf_ifid_gre(eui, dev);
1732         case ARPHRD_IEEE802154:
1733                 return addrconf_ifid_eui64(eui, dev);
1734         }
1735         return -1;
1736 }
1737
1738 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1739 {
1740         int err = -1;
1741         struct inet6_ifaddr *ifp;
1742
1743         read_lock_bh(&idev->lock);
1744         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1745                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1746                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1747                         err = 0;
1748                         break;
1749                 }
1750         }
1751         read_unlock_bh(&idev->lock);
1752         return err;
1753 }
1754
1755 #ifdef CONFIG_IPV6_PRIVACY
1756 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1757 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1758 {
1759 regen:
1760         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1761         idev->rndid[0] &= ~0x02;
1762
1763         /*
1764          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1765          * check if generated address is not inappropriate
1766          *
1767          *  - Reserved subnet anycast (RFC 2526)
1768          *      11111101 11....11 1xxxxxxx
1769          *  - ISATAP (RFC4214) 6.1
1770          *      00-00-5E-FE-xx-xx-xx-xx
1771          *  - value 0
1772          *  - XXX: already assigned to an address on the device
1773          */
1774         if (idev->rndid[0] == 0xfd &&
1775             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1776             (idev->rndid[7]&0x80))
1777                 goto regen;
1778         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1779                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1780                         goto regen;
1781                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1782                         goto regen;
1783         }
1784 }
1785
1786 static void ipv6_regen_rndid(unsigned long data)
1787 {
1788         struct inet6_dev *idev = (struct inet6_dev *) data;
1789         unsigned long expires;
1790
1791         rcu_read_lock_bh();
1792         write_lock_bh(&idev->lock);
1793
1794         if (idev->dead)
1795                 goto out;
1796
1797         __ipv6_regen_rndid(idev);
1798
1799         expires = jiffies +
1800                 idev->cnf.temp_prefered_lft * HZ -
1801                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1802                 idev->cnf.max_desync_factor * HZ;
1803         if (time_before(expires, jiffies)) {
1804                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1805                         __func__, idev->dev->name);
1806                 goto out;
1807         }
1808
1809         if (!mod_timer(&idev->regen_timer, expires))
1810                 in6_dev_hold(idev);
1811
1812 out:
1813         write_unlock_bh(&idev->lock);
1814         rcu_read_unlock_bh();
1815         in6_dev_put(idev);
1816 }
1817
1818 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1819 {
1820         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1821                 __ipv6_regen_rndid(idev);
1822 }
1823 #endif
1824
1825 /*
1826  *      Add prefix route.
1827  */
1828
1829 static void
1830 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1831                       unsigned long expires, u32 flags)
1832 {
1833         struct fib6_config cfg = {
1834                 .fc_table = RT6_TABLE_PREFIX,
1835                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1836                 .fc_ifindex = dev->ifindex,
1837                 .fc_expires = expires,
1838                 .fc_dst_len = plen,
1839                 .fc_flags = RTF_UP | flags,
1840                 .fc_nlinfo.nl_net = dev_net(dev),
1841                 .fc_protocol = RTPROT_KERNEL,
1842         };
1843
1844         cfg.fc_dst = *pfx;
1845
1846         /* Prevent useless cloning on PtP SIT.
1847            This thing is done here expecting that the whole
1848            class of non-broadcast devices need not cloning.
1849          */
1850 #if IS_ENABLED(CONFIG_IPV6_SIT)
1851         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1852                 cfg.fc_flags |= RTF_NONEXTHOP;
1853 #endif
1854
1855         ip6_route_add(&cfg);
1856 }
1857
1858
1859 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1860                                                   int plen,
1861                                                   const struct net_device *dev,
1862                                                   u32 flags, u32 noflags)
1863 {
1864         struct fib6_node *fn;
1865         struct rt6_info *rt = NULL;
1866         struct fib6_table *table;
1867
1868         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1869         if (table == NULL)
1870                 return NULL;
1871
1872         read_lock_bh(&table->tb6_lock);
1873         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1874         if (!fn)
1875                 goto out;
1876         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1877                 if (rt->dst.dev->ifindex != dev->ifindex)
1878                         continue;
1879                 if ((rt->rt6i_flags & flags) != flags)
1880                         continue;
1881                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1882                         continue;
1883                 dst_hold(&rt->dst);
1884                 break;
1885         }
1886 out:
1887         read_unlock_bh(&table->tb6_lock);
1888         return rt;
1889 }
1890
1891
1892 /* Create "default" multicast route to the interface */
1893
1894 static void addrconf_add_mroute(struct net_device *dev)
1895 {
1896         struct fib6_config cfg = {
1897                 .fc_table = RT6_TABLE_LOCAL,
1898                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1899                 .fc_ifindex = dev->ifindex,
1900                 .fc_dst_len = 8,
1901                 .fc_flags = RTF_UP,
1902                 .fc_nlinfo.nl_net = dev_net(dev),
1903         };
1904
1905         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1906
1907         ip6_route_add(&cfg);
1908 }
1909
1910 #if IS_ENABLED(CONFIG_IPV6_SIT)
1911 static void sit_route_add(struct net_device *dev)
1912 {
1913         struct fib6_config cfg = {
1914                 .fc_table = RT6_TABLE_MAIN,
1915                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1916                 .fc_ifindex = dev->ifindex,
1917                 .fc_dst_len = 96,
1918                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1919                 .fc_nlinfo.nl_net = dev_net(dev),
1920         };
1921
1922         /* prefix length - 96 bits "::d.d.d.d" */
1923         ip6_route_add(&cfg);
1924 }
1925 #endif
1926
1927 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1928 {
1929         struct inet6_dev *idev;
1930
1931         ASSERT_RTNL();
1932
1933         idev = ipv6_find_idev(dev);
1934         if (!idev)
1935                 return ERR_PTR(-ENOBUFS);
1936
1937         if (idev->cnf.disable_ipv6)
1938                 return ERR_PTR(-EACCES);
1939
1940         /* Add default multicast route */
1941         if (!(dev->flags & IFF_LOOPBACK))
1942                 addrconf_add_mroute(dev);
1943
1944         return idev;
1945 }
1946
1947 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1948 {
1949         struct prefix_info *pinfo;
1950         __u32 valid_lft;
1951         __u32 prefered_lft;
1952         int addr_type;
1953         struct inet6_dev *in6_dev;
1954         struct net *net = dev_net(dev);
1955
1956         pinfo = (struct prefix_info *) opt;
1957
1958         if (len < sizeof(struct prefix_info)) {
1959                 ADBG(("addrconf: prefix option too short\n"));
1960                 return;
1961         }
1962
1963         /*
1964          *      Validation checks ([ADDRCONF], page 19)
1965          */
1966
1967         addr_type = ipv6_addr_type(&pinfo->prefix);
1968
1969         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1970                 return;
1971
1972         valid_lft = ntohl(pinfo->valid);
1973         prefered_lft = ntohl(pinfo->prefered);
1974
1975         if (prefered_lft > valid_lft) {
1976                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1977                 return;
1978         }
1979
1980         in6_dev = in6_dev_get(dev);
1981
1982         if (in6_dev == NULL) {
1983                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1984                                     dev->name);
1985                 return;
1986         }
1987
1988         /*
1989          *      Two things going on here:
1990          *      1) Add routes for on-link prefixes
1991          *      2) Configure prefixes with the auto flag set
1992          */
1993
1994         if (pinfo->onlink) {
1995                 struct rt6_info *rt;
1996                 unsigned long rt_expires;
1997
1998                 /* Avoid arithmetic overflow. Really, we could
1999                  * save rt_expires in seconds, likely valid_lft,
2000                  * but it would require division in fib gc, that it
2001                  * not good.
2002                  */
2003                 if (HZ > USER_HZ)
2004                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2005                 else
2006                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2007
2008                 if (addrconf_finite_timeout(rt_expires))
2009                         rt_expires *= HZ;
2010
2011                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2012                                                pinfo->prefix_len,
2013                                                dev,
2014                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2015                                                RTF_GATEWAY | RTF_DEFAULT);
2016
2017                 if (rt) {
2018                         /* Autoconf prefix route */
2019                         if (valid_lft == 0) {
2020                                 ip6_del_rt(rt);
2021                                 rt = NULL;
2022                         } else if (addrconf_finite_timeout(rt_expires)) {
2023                                 /* not infinity */
2024                                 rt6_set_expires(rt, jiffies + rt_expires);
2025                         } else {
2026                                 rt6_clean_expires(rt);
2027                         }
2028                 } else if (valid_lft) {
2029                         clock_t expires = 0;
2030                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2031                         if (addrconf_finite_timeout(rt_expires)) {
2032                                 /* not infinity */
2033                                 flags |= RTF_EXPIRES;
2034                                 expires = jiffies_to_clock_t(rt_expires);
2035                         }
2036                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2037                                               dev, expires, flags);
2038                 }
2039                 ip6_rt_put(rt);
2040         }
2041
2042         /* Try to figure out our local address for this prefix */
2043
2044         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2045                 struct inet6_ifaddr *ifp;
2046                 struct in6_addr addr;
2047                 int create = 0, update_lft = 0;
2048
2049                 if (pinfo->prefix_len == 64) {
2050                         memcpy(&addr, &pinfo->prefix, 8);
2051                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2052                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2053                                 in6_dev_put(in6_dev);
2054                                 return;
2055                         }
2056                         goto ok;
2057                 }
2058                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2059                                     pinfo->prefix_len);
2060                 in6_dev_put(in6_dev);
2061                 return;
2062
2063 ok:
2064
2065                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2066
2067                 if (ifp == NULL && valid_lft) {
2068                         int max_addresses = in6_dev->cnf.max_addresses;
2069                         u32 addr_flags = 0;
2070
2071 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2072                         if (in6_dev->cnf.optimistic_dad &&
2073                             !net->ipv6.devconf_all->forwarding && sllao)
2074                                 addr_flags = IFA_F_OPTIMISTIC;
2075 #endif
2076
2077                         /* Do not allow to create too much of autoconfigured
2078                          * addresses; this would be too easy way to crash kernel.
2079                          */
2080                         if (!max_addresses ||
2081                             ipv6_count_addresses(in6_dev) < max_addresses)
2082                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2083                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2084                                                     addr_flags);
2085
2086                         if (!ifp || IS_ERR(ifp)) {
2087                                 in6_dev_put(in6_dev);
2088                                 return;
2089                         }
2090
2091                         update_lft = create = 1;
2092                         ifp->cstamp = jiffies;
2093                         addrconf_dad_start(ifp);
2094                 }
2095
2096                 if (ifp) {
2097                         int flags;
2098                         unsigned long now;
2099 #ifdef CONFIG_IPV6_PRIVACY
2100                         struct inet6_ifaddr *ift;
2101 #endif
2102                         u32 stored_lft;
2103
2104                         /* update lifetime (RFC2462 5.5.3 e) */
2105                         spin_lock(&ifp->lock);
2106                         now = jiffies;
2107                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2108                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2109                         else
2110                                 stored_lft = 0;
2111                         if (!update_lft && stored_lft) {
2112                                 if (valid_lft > MIN_VALID_LIFETIME ||
2113                                     valid_lft > stored_lft)
2114                                         update_lft = 1;
2115                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2116                                         /* valid_lft <= stored_lft is always true */
2117                                         /*
2118                                          * RFC 4862 Section 5.5.3e:
2119                                          * "Note that the preferred lifetime of
2120                                          *  the corresponding address is always
2121                                          *  reset to the Preferred Lifetime in
2122                                          *  the received Prefix Information
2123                                          *  option, regardless of whether the
2124                                          *  valid lifetime is also reset or
2125                                          *  ignored."
2126                                          *
2127                                          *  So if the preferred lifetime in
2128                                          *  this advertisement is different
2129                                          *  than what we have stored, but the
2130                                          *  valid lifetime is invalid, just
2131                                          *  reset prefered_lft.
2132                                          *
2133                                          *  We must set the valid lifetime
2134                                          *  to the stored lifetime since we'll
2135                                          *  be updating the timestamp below,
2136                                          *  else we'll set it back to the
2137                                          *  minimum.
2138                                          */
2139                                         if (prefered_lft != ifp->prefered_lft) {
2140                                                 valid_lft = stored_lft;
2141                                                 update_lft = 1;
2142                                         }
2143                                 } else {
2144                                         valid_lft = MIN_VALID_LIFETIME;
2145                                         if (valid_lft < prefered_lft)
2146                                                 prefered_lft = valid_lft;
2147                                         update_lft = 1;
2148                                 }
2149                         }
2150
2151                         if (update_lft) {
2152                                 ifp->valid_lft = valid_lft;
2153                                 ifp->prefered_lft = prefered_lft;
2154                                 ifp->tstamp = now;
2155                                 flags = ifp->flags;
2156                                 ifp->flags &= ~IFA_F_DEPRECATED;
2157                                 spin_unlock(&ifp->lock);
2158
2159                                 if (!(flags&IFA_F_TENTATIVE))
2160                                         ipv6_ifa_notify(0, ifp);
2161                         } else
2162                                 spin_unlock(&ifp->lock);
2163
2164 #ifdef CONFIG_IPV6_PRIVACY
2165                         read_lock_bh(&in6_dev->lock);
2166                         /* update all temporary addresses in the list */
2167                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2168                                             tmp_list) {
2169                                 int age, max_valid, max_prefered;
2170
2171                                 if (ifp != ift->ifpub)
2172                                         continue;
2173
2174                                 /*
2175                                  * RFC 4941 section 3.3:
2176                                  * If a received option will extend the lifetime
2177                                  * of a public address, the lifetimes of
2178                                  * temporary addresses should be extended,
2179                                  * subject to the overall constraint that no
2180                                  * temporary addresses should ever remain
2181                                  * "valid" or "preferred" for a time longer than
2182                                  * (TEMP_VALID_LIFETIME) or
2183                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2184                                  * respectively.
2185                                  */
2186                                 age = (now - ift->cstamp) / HZ;
2187                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2188                                 if (max_valid < 0)
2189                                         max_valid = 0;
2190
2191                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2192                                                in6_dev->cnf.max_desync_factor -
2193                                                age;
2194                                 if (max_prefered < 0)
2195                                         max_prefered = 0;
2196
2197                                 if (valid_lft > max_valid)
2198                                         valid_lft = max_valid;
2199
2200                                 if (prefered_lft > max_prefered)
2201                                         prefered_lft = max_prefered;
2202
2203                                 spin_lock(&ift->lock);
2204                                 flags = ift->flags;
2205                                 ift->valid_lft = valid_lft;
2206                                 ift->prefered_lft = prefered_lft;
2207                                 ift->tstamp = now;
2208                                 if (prefered_lft > 0)
2209                                         ift->flags &= ~IFA_F_DEPRECATED;
2210
2211                                 spin_unlock(&ift->lock);
2212                                 if (!(flags&IFA_F_TENTATIVE))
2213                                         ipv6_ifa_notify(0, ift);
2214                         }
2215
2216                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2217                                 /*
2218                                  * When a new public address is created as
2219                                  * described in [ADDRCONF], also create a new
2220                                  * temporary address. Also create a temporary
2221                                  * address if it's enabled but no temporary
2222                                  * address currently exists.
2223                                  */
2224                                 read_unlock_bh(&in6_dev->lock);
2225                                 ipv6_create_tempaddr(ifp, NULL);
2226                         } else {
2227                                 read_unlock_bh(&in6_dev->lock);
2228                         }
2229 #endif
2230                         in6_ifa_put(ifp);
2231                         addrconf_verify(0);
2232                 }
2233         }
2234         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2235         in6_dev_put(in6_dev);
2236 }
2237
2238 /*
2239  *      Set destination address.
2240  *      Special case for SIT interfaces where we create a new "virtual"
2241  *      device.
2242  */
2243 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2244 {
2245         struct in6_ifreq ireq;
2246         struct net_device *dev;
2247         int err = -EINVAL;
2248
2249         rtnl_lock();
2250
2251         err = -EFAULT;
2252         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2253                 goto err_exit;
2254
2255         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2256
2257         err = -ENODEV;
2258         if (dev == NULL)
2259                 goto err_exit;
2260
2261 #if IS_ENABLED(CONFIG_IPV6_SIT)
2262         if (dev->type == ARPHRD_SIT) {
2263                 const struct net_device_ops *ops = dev->netdev_ops;
2264                 struct ifreq ifr;
2265                 struct ip_tunnel_parm p;
2266
2267                 err = -EADDRNOTAVAIL;
2268                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2269                         goto err_exit;
2270
2271                 memset(&p, 0, sizeof(p));
2272                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2273                 p.iph.saddr = 0;
2274                 p.iph.version = 4;
2275                 p.iph.ihl = 5;
2276                 p.iph.protocol = IPPROTO_IPV6;
2277                 p.iph.ttl = 64;
2278                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2279
2280                 if (ops->ndo_do_ioctl) {
2281                         mm_segment_t oldfs = get_fs();
2282
2283                         set_fs(KERNEL_DS);
2284                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2285                         set_fs(oldfs);
2286                 } else
2287                         err = -EOPNOTSUPP;
2288
2289                 if (err == 0) {
2290                         err = -ENOBUFS;
2291                         dev = __dev_get_by_name(net, p.name);
2292                         if (!dev)
2293                                 goto err_exit;
2294                         err = dev_open(dev);
2295                 }
2296         }
2297 #endif
2298
2299 err_exit:
2300         rtnl_unlock();
2301         return err;
2302 }
2303
2304 /*
2305  *      Manual configuration of address on an interface
2306  */
2307 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2308                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2309                           __u32 valid_lft)
2310 {
2311         struct inet6_ifaddr *ifp;
2312         struct inet6_dev *idev;
2313         struct net_device *dev;
2314         int scope;
2315         u32 flags;
2316         clock_t expires;
2317         unsigned long timeout;
2318
2319         ASSERT_RTNL();
2320
2321         if (plen > 128)
2322                 return -EINVAL;
2323
2324         /* check the lifetime */
2325         if (!valid_lft || prefered_lft > valid_lft)
2326                 return -EINVAL;
2327
2328         dev = __dev_get_by_index(net, ifindex);
2329         if (!dev)
2330                 return -ENODEV;
2331
2332         idev = addrconf_add_dev(dev);
2333         if (IS_ERR(idev))
2334                 return PTR_ERR(idev);
2335
2336         scope = ipv6_addr_scope(pfx);
2337
2338         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2339         if (addrconf_finite_timeout(timeout)) {
2340                 expires = jiffies_to_clock_t(timeout * HZ);
2341                 valid_lft = timeout;
2342                 flags = RTF_EXPIRES;
2343         } else {
2344                 expires = 0;
2345                 flags = 0;
2346                 ifa_flags |= IFA_F_PERMANENT;
2347         }
2348
2349         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2350         if (addrconf_finite_timeout(timeout)) {
2351                 if (timeout == 0)
2352                         ifa_flags |= IFA_F_DEPRECATED;
2353                 prefered_lft = timeout;
2354         }
2355
2356         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2357
2358         if (!IS_ERR(ifp)) {
2359                 spin_lock_bh(&ifp->lock);
2360                 ifp->valid_lft = valid_lft;
2361                 ifp->prefered_lft = prefered_lft;
2362                 ifp->tstamp = jiffies;
2363                 spin_unlock_bh(&ifp->lock);
2364
2365                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2366                                       expires, flags);
2367                 /*
2368                  * Note that section 3.1 of RFC 4429 indicates
2369                  * that the Optimistic flag should not be set for
2370                  * manually configured addresses
2371                  */
2372                 addrconf_dad_start(ifp);
2373                 in6_ifa_put(ifp);
2374                 addrconf_verify(0);
2375                 return 0;
2376         }
2377
2378         return PTR_ERR(ifp);
2379 }
2380
2381 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2382                           unsigned int plen)
2383 {
2384         struct inet6_ifaddr *ifp;
2385         struct inet6_dev *idev;
2386         struct net_device *dev;
2387
2388         if (plen > 128)
2389                 return -EINVAL;
2390
2391         dev = __dev_get_by_index(net, ifindex);
2392         if (!dev)
2393                 return -ENODEV;
2394
2395         if ((idev = __in6_dev_get(dev)) == NULL)
2396                 return -ENXIO;
2397
2398         read_lock_bh(&idev->lock);
2399         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2400                 if (ifp->prefix_len == plen &&
2401                     ipv6_addr_equal(pfx, &ifp->addr)) {
2402                         in6_ifa_hold(ifp);
2403                         read_unlock_bh(&idev->lock);
2404
2405                         ipv6_del_addr(ifp);
2406
2407                         /* If the last address is deleted administratively,
2408                            disable IPv6 on this interface.
2409                          */
2410                         if (list_empty(&idev->addr_list))
2411                                 addrconf_ifdown(idev->dev, 1);
2412                         return 0;
2413                 }
2414         }
2415         read_unlock_bh(&idev->lock);
2416         return -EADDRNOTAVAIL;
2417 }
2418
2419
2420 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2421 {
2422         struct in6_ifreq ireq;
2423         int err;
2424
2425         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2426                 return -EPERM;
2427
2428         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2429                 return -EFAULT;
2430
2431         rtnl_lock();
2432         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2433                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2434                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2435         rtnl_unlock();
2436         return err;
2437 }
2438
2439 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2440 {
2441         struct in6_ifreq ireq;
2442         int err;
2443
2444         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2445                 return -EPERM;
2446
2447         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2448                 return -EFAULT;
2449
2450         rtnl_lock();
2451         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2452                              ireq.ifr6_prefixlen);
2453         rtnl_unlock();
2454         return err;
2455 }
2456
2457 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2458                      int plen, int scope)
2459 {
2460         struct inet6_ifaddr *ifp;
2461
2462         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2463         if (!IS_ERR(ifp)) {
2464                 spin_lock_bh(&ifp->lock);
2465                 ifp->flags &= ~IFA_F_TENTATIVE;
2466                 spin_unlock_bh(&ifp->lock);
2467                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2468                 in6_ifa_put(ifp);
2469         }
2470 }
2471
2472 #if IS_ENABLED(CONFIG_IPV6_SIT)
2473 static void sit_add_v4_addrs(struct inet6_dev *idev)
2474 {
2475         struct in6_addr addr;
2476         struct net_device *dev;
2477         struct net *net = dev_net(idev->dev);
2478         int scope;
2479
2480         ASSERT_RTNL();
2481
2482         memset(&addr, 0, sizeof(struct in6_addr));
2483         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2484
2485         if (idev->dev->flags&IFF_POINTOPOINT) {
2486                 addr.s6_addr32[0] = htonl(0xfe800000);
2487                 scope = IFA_LINK;
2488         } else {
2489                 scope = IPV6_ADDR_COMPATv4;
2490         }
2491
2492         if (addr.s6_addr32[3]) {
2493                 add_addr(idev, &addr, 128, scope);
2494                 return;
2495         }
2496
2497         for_each_netdev(net, dev) {
2498                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2499                 if (in_dev && (dev->flags & IFF_UP)) {
2500                         struct in_ifaddr *ifa;
2501
2502                         int flag = scope;
2503
2504                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2505                                 int plen;
2506
2507                                 addr.s6_addr32[3] = ifa->ifa_local;
2508
2509                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2510                                         continue;
2511                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2512                                         if (idev->dev->flags&IFF_POINTOPOINT)
2513                                                 continue;
2514                                         flag |= IFA_HOST;
2515                                 }
2516                                 if (idev->dev->flags&IFF_POINTOPOINT)
2517                                         plen = 64;
2518                                 else
2519                                         plen = 96;
2520
2521                                 add_addr(idev, &addr, plen, flag);
2522                         }
2523                 }
2524         }
2525 }
2526 #endif
2527
2528 static void init_loopback(struct net_device *dev)
2529 {
2530         struct inet6_dev  *idev;
2531
2532         /* ::1 */
2533
2534         ASSERT_RTNL();
2535
2536         if ((idev = ipv6_find_idev(dev)) == NULL) {
2537                 pr_debug("%s: add_dev failed\n", __func__);
2538                 return;
2539         }
2540
2541         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2542 }
2543
2544 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2545 {
2546         struct inet6_ifaddr *ifp;
2547         u32 addr_flags = IFA_F_PERMANENT;
2548
2549 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2550         if (idev->cnf.optimistic_dad &&
2551             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2552                 addr_flags |= IFA_F_OPTIMISTIC;
2553 #endif
2554
2555
2556         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2557         if (!IS_ERR(ifp)) {
2558                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2559                 addrconf_dad_start(ifp);
2560                 in6_ifa_put(ifp);
2561         }
2562 }
2563
2564 static void addrconf_dev_config(struct net_device *dev)
2565 {
2566         struct in6_addr addr;
2567         struct inet6_dev *idev;
2568
2569         ASSERT_RTNL();
2570
2571         if ((dev->type != ARPHRD_ETHER) &&
2572             (dev->type != ARPHRD_FDDI) &&
2573             (dev->type != ARPHRD_ARCNET) &&
2574             (dev->type != ARPHRD_INFINIBAND) &&
2575             (dev->type != ARPHRD_IEEE802154)) {
2576                 /* Alas, we support only Ethernet autoconfiguration. */
2577                 return;
2578         }
2579
2580         idev = addrconf_add_dev(dev);
2581         if (IS_ERR(idev))
2582                 return;
2583
2584         memset(&addr, 0, sizeof(struct in6_addr));
2585         addr.s6_addr32[0] = htonl(0xFE800000);
2586
2587         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2588                 addrconf_add_linklocal(idev, &addr);
2589 }
2590
2591 #if IS_ENABLED(CONFIG_IPV6_SIT)
2592 static void addrconf_sit_config(struct net_device *dev)
2593 {
2594         struct inet6_dev *idev;
2595
2596         ASSERT_RTNL();
2597
2598         /*
2599          * Configure the tunnel with one of our IPv4
2600          * addresses... we should configure all of
2601          * our v4 addrs in the tunnel
2602          */
2603
2604         if ((idev = ipv6_find_idev(dev)) == NULL) {
2605                 pr_debug("%s: add_dev failed\n", __func__);
2606                 return;
2607         }
2608
2609         if (dev->priv_flags & IFF_ISATAP) {
2610                 struct in6_addr addr;
2611
2612                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2613                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2614                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2615                         addrconf_add_linklocal(idev, &addr);
2616                 return;
2617         }
2618
2619         sit_add_v4_addrs(idev);
2620
2621         if (dev->flags&IFF_POINTOPOINT)
2622                 addrconf_add_mroute(dev);
2623         else
2624                 sit_route_add(dev);
2625 }
2626 #endif
2627
2628 #if IS_ENABLED(CONFIG_NET_IPGRE)
2629 static void addrconf_gre_config(struct net_device *dev)
2630 {
2631         struct inet6_dev *idev;
2632         struct in6_addr addr;
2633
2634         pr_info("%s(%s)\n", __func__, dev->name);
2635
2636         ASSERT_RTNL();
2637
2638         if ((idev = ipv6_find_idev(dev)) == NULL) {
2639                 pr_debug("%s: add_dev failed\n", __func__);
2640                 return;
2641         }
2642
2643         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2644         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2645
2646         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2647                 addrconf_add_linklocal(idev, &addr);
2648 }
2649 #endif
2650
2651 static inline int
2652 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2653 {
2654         struct in6_addr lladdr;
2655
2656         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2657                 addrconf_add_linklocal(idev, &lladdr);
2658                 return 0;
2659         }
2660         return -1;
2661 }
2662
2663 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2664 {
2665         struct net_device *link_dev;
2666         struct net *net = dev_net(idev->dev);
2667
2668         /* first try to inherit the link-local address from the link device */
2669         if (idev->dev->iflink &&
2670             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2671                 if (!ipv6_inherit_linklocal(idev, link_dev))
2672                         return;
2673         }
2674         /* then try to inherit it from any device */
2675         for_each_netdev(net, link_dev) {
2676                 if (!ipv6_inherit_linklocal(idev, link_dev))
2677                         return;
2678         }
2679         pr_debug("init ip6-ip6: add_linklocal failed\n");
2680 }
2681
2682 /*
2683  * Autoconfigure tunnel with a link-local address so routing protocols,
2684  * DHCPv6, MLD etc. can be run over the virtual link
2685  */
2686
2687 static void addrconf_ip6_tnl_config(struct net_device *dev)
2688 {
2689         struct inet6_dev *idev;
2690
2691         ASSERT_RTNL();
2692
2693         idev = addrconf_add_dev(dev);
2694         if (IS_ERR(idev)) {
2695                 pr_debug("init ip6-ip6: add_dev failed\n");
2696                 return;
2697         }
2698         ip6_tnl_add_linklocal(idev);
2699 }
2700
2701 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2702                            void *data)
2703 {
2704         struct net_device *dev = (struct net_device *) data;
2705         struct inet6_dev *idev = __in6_dev_get(dev);
2706         int run_pending = 0;
2707         int err;
2708
2709         switch (event) {
2710         case NETDEV_REGISTER:
2711                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2712                         idev = ipv6_add_dev(dev);
2713                         if (!idev)
2714                                 return notifier_from_errno(-ENOMEM);
2715                 }
2716                 break;
2717
2718         case NETDEV_UP:
2719         case NETDEV_CHANGE:
2720                 if (dev->flags & IFF_SLAVE)
2721                         break;
2722
2723                 if (event == NETDEV_UP) {
2724                         if (!addrconf_qdisc_ok(dev)) {
2725                                 /* device is not ready yet. */
2726                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2727                                         dev->name);
2728                                 break;
2729                         }
2730
2731                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2732                                 idev = ipv6_add_dev(dev);
2733
2734                         if (idev) {
2735                                 idev->if_flags |= IF_READY;
2736                                 run_pending = 1;
2737                         }
2738                 } else {
2739                         if (!addrconf_qdisc_ok(dev)) {
2740                                 /* device is still not ready. */
2741                                 break;
2742                         }
2743
2744                         if (idev) {
2745                                 if (idev->if_flags & IF_READY)
2746                                         /* device is already configured. */
2747                                         break;
2748                                 idev->if_flags |= IF_READY;
2749                         }
2750
2751                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2752                                 dev->name);
2753
2754                         run_pending = 1;
2755                 }
2756
2757                 switch (dev->type) {
2758 #if IS_ENABLED(CONFIG_IPV6_SIT)
2759                 case ARPHRD_SIT:
2760                         addrconf_sit_config(dev);
2761                         break;
2762 #endif
2763 #if IS_ENABLED(CONFIG_NET_IPGRE)
2764                 case ARPHRD_IPGRE:
2765                         addrconf_gre_config(dev);
2766                         break;
2767 #endif
2768                 case ARPHRD_TUNNEL6:
2769                         addrconf_ip6_tnl_config(dev);
2770                         break;
2771                 case ARPHRD_LOOPBACK:
2772                         init_loopback(dev);
2773                         break;
2774
2775                 default:
2776                         addrconf_dev_config(dev);
2777                         break;
2778                 }
2779
2780                 if (idev) {
2781                         if (run_pending)
2782                                 addrconf_dad_run(idev);
2783
2784                         /*
2785                          * If the MTU changed during the interface down,
2786                          * when the interface up, the changed MTU must be
2787                          * reflected in the idev as well as routers.
2788                          */
2789                         if (idev->cnf.mtu6 != dev->mtu &&
2790                             dev->mtu >= IPV6_MIN_MTU) {
2791                                 rt6_mtu_change(dev, dev->mtu);
2792                                 idev->cnf.mtu6 = dev->mtu;
2793                         }
2794                         idev->tstamp = jiffies;
2795                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2796
2797                         /*
2798                          * If the changed mtu during down is lower than
2799                          * IPV6_MIN_MTU stop IPv6 on this interface.
2800                          */
2801                         if (dev->mtu < IPV6_MIN_MTU)
2802                                 addrconf_ifdown(dev, 1);
2803                 }
2804                 break;
2805
2806         case NETDEV_CHANGEMTU:
2807                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2808                         rt6_mtu_change(dev, dev->mtu);
2809                         idev->cnf.mtu6 = dev->mtu;
2810                         break;
2811                 }
2812
2813                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2814                         idev = ipv6_add_dev(dev);
2815                         if (idev)
2816                                 break;
2817                 }
2818
2819                 /*
2820                  * MTU falled under IPV6_MIN_MTU.
2821                  * Stop IPv6 on this interface.
2822                  */
2823
2824         case NETDEV_DOWN:
2825         case NETDEV_UNREGISTER:
2826                 /*
2827                  *      Remove all addresses from this interface.
2828                  */
2829                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2830                 break;
2831
2832         case NETDEV_CHANGENAME:
2833                 if (idev) {
2834                         snmp6_unregister_dev(idev);
2835                         addrconf_sysctl_unregister(idev);
2836                         addrconf_sysctl_register(idev);
2837                         err = snmp6_register_dev(idev);
2838                         if (err)
2839                                 return notifier_from_errno(err);
2840                 }
2841                 break;
2842
2843         case NETDEV_PRE_TYPE_CHANGE:
2844         case NETDEV_POST_TYPE_CHANGE:
2845                 addrconf_type_change(dev, event);
2846                 break;
2847         }
2848
2849         return NOTIFY_OK;
2850 }
2851
2852 /*
2853  *      addrconf module should be notified of a device going up
2854  */
2855 static struct notifier_block ipv6_dev_notf = {
2856         .notifier_call = addrconf_notify,
2857 };
2858
2859 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2860 {
2861         struct inet6_dev *idev;
2862         ASSERT_RTNL();
2863
2864         idev = __in6_dev_get(dev);
2865
2866         if (event == NETDEV_POST_TYPE_CHANGE)
2867                 ipv6_mc_remap(idev);
2868         else if (event == NETDEV_PRE_TYPE_CHANGE)
2869                 ipv6_mc_unmap(idev);
2870 }
2871
2872 static int addrconf_ifdown(struct net_device *dev, int how)
2873 {
2874         struct net *net = dev_net(dev);
2875         struct inet6_dev *idev;
2876         struct inet6_ifaddr *ifa;
2877         int state, i;
2878
2879         ASSERT_RTNL();
2880
2881         rt6_ifdown(net, dev);
2882         neigh_ifdown(&nd_tbl, dev);
2883
2884         idev = __in6_dev_get(dev);
2885         if (idev == NULL)
2886                 return -ENODEV;
2887
2888         /*
2889          * Step 1: remove reference to ipv6 device from parent device.
2890          *         Do not dev_put!
2891          */
2892         if (how) {
2893                 idev->dead = 1;
2894
2895                 /* protected by rtnl_lock */
2896                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2897
2898                 /* Step 1.5: remove snmp6 entry */
2899                 snmp6_unregister_dev(idev);
2900
2901         }
2902
2903         /* Step 2: clear hash table */
2904         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2905                 struct hlist_head *h = &inet6_addr_lst[i];
2906                 struct hlist_node *n;
2907
2908                 spin_lock_bh(&addrconf_hash_lock);
2909         restart:
2910                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2911                         if (ifa->idev == idev) {
2912                                 hlist_del_init_rcu(&ifa->addr_lst);
2913                                 addrconf_del_timer(ifa);
2914                                 goto restart;
2915                         }
2916                 }
2917                 spin_unlock_bh(&addrconf_hash_lock);
2918         }
2919
2920         write_lock_bh(&idev->lock);
2921
2922         /* Step 2: clear flags for stateless addrconf */
2923         if (!how)
2924                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2925
2926 #ifdef CONFIG_IPV6_PRIVACY
2927         if (how && del_timer(&idev->regen_timer))
2928                 in6_dev_put(idev);
2929
2930         /* Step 3: clear tempaddr list */
2931         while (!list_empty(&idev->tempaddr_list)) {
2932                 ifa = list_first_entry(&idev->tempaddr_list,
2933                                        struct inet6_ifaddr, tmp_list);
2934                 list_del(&ifa->tmp_list);
2935                 write_unlock_bh(&idev->lock);
2936                 spin_lock_bh(&ifa->lock);
2937
2938                 if (ifa->ifpub) {
2939                         in6_ifa_put(ifa->ifpub);
2940                         ifa->ifpub = NULL;
2941                 }
2942                 spin_unlock_bh(&ifa->lock);
2943                 in6_ifa_put(ifa);
2944                 write_lock_bh(&idev->lock);
2945         }
2946 #endif
2947
2948         while (!list_empty(&idev->addr_list)) {
2949                 ifa = list_first_entry(&idev->addr_list,
2950                                        struct inet6_ifaddr, if_list);
2951                 addrconf_del_timer(ifa);
2952
2953                 list_del(&ifa->if_list);
2954
2955                 write_unlock_bh(&idev->lock);
2956
2957                 spin_lock_bh(&ifa->state_lock);
2958                 state = ifa->state;
2959                 ifa->state = INET6_IFADDR_STATE_DEAD;
2960                 spin_unlock_bh(&ifa->state_lock);
2961
2962                 if (state != INET6_IFADDR_STATE_DEAD) {
2963                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2964                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2965                 }
2966                 in6_ifa_put(ifa);
2967
2968                 write_lock_bh(&idev->lock);
2969         }
2970
2971         write_unlock_bh(&idev->lock);
2972
2973         /* Step 5: Discard multicast list */
2974         if (how)
2975                 ipv6_mc_destroy_dev(idev);
2976         else
2977                 ipv6_mc_down(idev);
2978
2979         idev->tstamp = jiffies;
2980
2981         /* Last: Shot the device (if unregistered) */
2982         if (how) {
2983                 addrconf_sysctl_unregister(idev);
2984                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2985                 neigh_ifdown(&nd_tbl, dev);
2986                 in6_dev_put(idev);
2987         }
2988         return 0;
2989 }
2990
2991 static void addrconf_rs_timer(unsigned long data)
2992 {
2993         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2994         struct inet6_dev *idev = ifp->idev;
2995
2996         read_lock(&idev->lock);
2997         if (idev->dead || !(idev->if_flags & IF_READY))
2998                 goto out;
2999
3000         if (!ipv6_accept_ra(idev))
3001                 goto out;
3002
3003         /* Announcement received after solicitation was sent */
3004         if (idev->if_flags & IF_RA_RCVD)
3005                 goto out;
3006
3007         spin_lock(&ifp->lock);
3008         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3009                 /* The wait after the last probe can be shorter */
3010                 addrconf_mod_timer(ifp, AC_RS,
3011                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3012                                    idev->cnf.rtr_solicit_delay :
3013                                    idev->cnf.rtr_solicit_interval);
3014                 spin_unlock(&ifp->lock);
3015
3016                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3017         } else {
3018                 spin_unlock(&ifp->lock);
3019                 /*
3020                  * Note: we do not support deprecated "all on-link"
3021                  * assumption any longer.
3022                  */
3023                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3024         }
3025
3026 out:
3027         read_unlock(&idev->lock);
3028         in6_ifa_put(ifp);
3029 }
3030
3031 /*
3032  *      Duplicate Address Detection
3033  */
3034 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3035 {
3036         unsigned long rand_num;
3037         struct inet6_dev *idev = ifp->idev;
3038
3039         if (ifp->flags & IFA_F_OPTIMISTIC)
3040                 rand_num = 0;
3041         else
3042                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3043
3044         ifp->probes = idev->cnf.dad_transmits;
3045         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3046 }
3047
3048 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3049 {
3050         struct inet6_dev *idev = ifp->idev;
3051         struct net_device *dev = idev->dev;
3052
3053         addrconf_join_solict(dev, &ifp->addr);
3054
3055         net_srandom(ifp->addr.s6_addr32[3]);
3056
3057         read_lock_bh(&idev->lock);
3058         spin_lock(&ifp->lock);
3059         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3060                 goto out;
3061
3062         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3063             idev->cnf.accept_dad < 1 ||
3064             !(ifp->flags&IFA_F_TENTATIVE) ||
3065             ifp->flags & IFA_F_NODAD) {
3066                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3067                 spin_unlock(&ifp->lock);
3068                 read_unlock_bh(&idev->lock);
3069
3070                 addrconf_dad_completed(ifp);
3071                 return;
3072         }
3073
3074         if (!(idev->if_flags & IF_READY)) {
3075                 spin_unlock(&ifp->lock);
3076                 read_unlock_bh(&idev->lock);
3077                 /*
3078                  * If the device is not ready:
3079                  * - keep it tentative if it is a permanent address.
3080                  * - otherwise, kill it.
3081                  */
3082                 in6_ifa_hold(ifp);
3083                 addrconf_dad_stop(ifp, 0);
3084                 return;
3085         }
3086
3087         /*
3088          * Optimistic nodes can start receiving
3089          * Frames right away
3090          */
3091         if (ifp->flags & IFA_F_OPTIMISTIC)
3092                 ip6_ins_rt(ifp->rt);
3093
3094         addrconf_dad_kick(ifp);
3095 out:
3096         spin_unlock(&ifp->lock);
3097         read_unlock_bh(&idev->lock);
3098 }
3099
3100 static void addrconf_dad_timer(unsigned long data)
3101 {
3102         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3103         struct inet6_dev *idev = ifp->idev;
3104         struct in6_addr mcaddr;
3105
3106         if (!ifp->probes && addrconf_dad_end(ifp))
3107                 goto out;
3108
3109         read_lock(&idev->lock);
3110         if (idev->dead || !(idev->if_flags & IF_READY)) {
3111                 read_unlock(&idev->lock);
3112                 goto out;
3113         }
3114
3115         spin_lock(&ifp->lock);
3116         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3117                 spin_unlock(&ifp->lock);
3118                 read_unlock(&idev->lock);
3119                 goto out;
3120         }
3121
3122         if (ifp->probes == 0) {
3123                 /*
3124                  * DAD was successful
3125                  */
3126
3127                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3128                 spin_unlock(&ifp->lock);
3129                 read_unlock(&idev->lock);
3130
3131                 addrconf_dad_completed(ifp);
3132
3133                 goto out;
3134         }
3135
3136         ifp->probes--;
3137         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3138         spin_unlock(&ifp->lock);
3139         read_unlock(&idev->lock);
3140
3141         /* send a neighbour solicitation for our addr */
3142         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3143         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3144 out:
3145         in6_ifa_put(ifp);
3146 }
3147
3148 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3149 {
3150         struct net_device *dev = ifp->idev->dev;
3151
3152         /*
3153          *      Configure the address for reception. Now it is valid.
3154          */
3155
3156         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3157
3158         /* If added prefix is link local and we are prepared to process
3159            router advertisements, start sending router solicitations.
3160          */
3161
3162         if (ipv6_accept_ra(ifp->idev) &&
3163             ifp->idev->cnf.rtr_solicits > 0 &&
3164             (dev->flags&IFF_LOOPBACK) == 0 &&
3165             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3166                 /*
3167                  *      If a host as already performed a random delay
3168                  *      [...] as part of DAD [...] there is no need
3169                  *      to delay again before sending the first RS
3170                  */
3171                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3172
3173                 spin_lock_bh(&ifp->lock);
3174                 ifp->probes = 1;
3175                 ifp->idev->if_flags |= IF_RS_SENT;
3176                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3177                 spin_unlock_bh(&ifp->lock);
3178         }
3179 }
3180
3181 static void addrconf_dad_run(struct inet6_dev *idev)
3182 {
3183         struct inet6_ifaddr *ifp;
3184
3185         read_lock_bh(&idev->lock);
3186         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3187                 spin_lock(&ifp->lock);
3188                 if (ifp->flags & IFA_F_TENTATIVE &&
3189                     ifp->state == INET6_IFADDR_STATE_DAD)
3190                         addrconf_dad_kick(ifp);
3191                 spin_unlock(&ifp->lock);
3192         }
3193         read_unlock_bh(&idev->lock);
3194 }
3195
3196 #ifdef CONFIG_PROC_FS
3197 struct if6_iter_state {
3198         struct seq_net_private p;
3199         int bucket;
3200         int offset;
3201 };
3202
3203 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3204 {
3205         struct inet6_ifaddr *ifa = NULL;
3206         struct if6_iter_state *state = seq->private;
3207         struct net *net = seq_file_net(seq);
3208         int p = 0;
3209
3210         /* initial bucket if pos is 0 */
3211         if (pos == 0) {
3212                 state->bucket = 0;
3213                 state->offset = 0;
3214         }
3215
3216         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3217                 struct hlist_node *n;
3218                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3219                                          addr_lst) {
3220                         if (!net_eq(dev_net(ifa->idev->dev), net))
3221                                 continue;
3222                         /* sync with offset */
3223                         if (p < state->offset) {
3224                                 p++;
3225                                 continue;
3226                         }
3227                         state->offset++;
3228                         return ifa;
3229                 }
3230
3231                 /* prepare for next bucket */
3232                 state->offset = 0;
3233                 p = 0;
3234         }
3235         return NULL;
3236 }
3237
3238 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3239                                          struct inet6_ifaddr *ifa)
3240 {
3241         struct if6_iter_state *state = seq->private;
3242         struct net *net = seq_file_net(seq);
3243         struct hlist_node *n = &ifa->addr_lst;
3244
3245         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3246                 if (!net_eq(dev_net(ifa->idev->dev), net))
3247                         continue;
3248                 state->offset++;
3249                 return ifa;
3250         }
3251
3252         while (++state->bucket < IN6_ADDR_HSIZE) {
3253                 state->offset = 0;
3254                 hlist_for_each_entry_rcu_bh(ifa, n,
3255                                      &inet6_addr_lst[state->bucket], addr_lst) {
3256                         if (!net_eq(dev_net(ifa->idev->dev), net))
3257                                 continue;
3258                         state->offset++;
3259                         return ifa;
3260                 }
3261         }
3262
3263         return NULL;
3264 }
3265
3266 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3267         __acquires(rcu_bh)
3268 {
3269         rcu_read_lock_bh();
3270         return if6_get_first(seq, *pos);
3271 }
3272
3273 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3274 {
3275         struct inet6_ifaddr *ifa;
3276
3277         ifa = if6_get_next(seq, v);
3278         ++*pos;
3279         return ifa;
3280 }
3281
3282 static void if6_seq_stop(struct seq_file *seq, void *v)
3283         __releases(rcu_bh)
3284 {
3285         rcu_read_unlock_bh();
3286 }
3287
3288 static int if6_seq_show(struct seq_file *seq, void *v)
3289 {
3290         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3291         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3292                    &ifp->addr,
3293                    ifp->idev->dev->ifindex,
3294                    ifp->prefix_len,
3295                    ifp->scope,
3296                    ifp->flags,
3297                    ifp->idev->dev->name);
3298         return 0;
3299 }
3300
3301 static const struct seq_operations if6_seq_ops = {
3302         .start  = if6_seq_start,
3303         .next   = if6_seq_next,
3304         .show   = if6_seq_show,
3305         .stop   = if6_seq_stop,
3306 };
3307
3308 static int if6_seq_open(struct inode *inode, struct file *file)
3309 {
3310         return seq_open_net(inode, file, &if6_seq_ops,
3311                             sizeof(struct if6_iter_state));
3312 }
3313
3314 static const struct file_operations if6_fops = {
3315         .owner          = THIS_MODULE,
3316         .open           = if6_seq_open,
3317         .read           = seq_read,
3318         .llseek         = seq_lseek,
3319         .release        = seq_release_net,
3320 };
3321
3322 static int __net_init if6_proc_net_init(struct net *net)
3323 {
3324         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3325                 return -ENOMEM;
3326         return 0;
3327 }
3328
3329 static void __net_exit if6_proc_net_exit(struct net *net)
3330 {
3331        proc_net_remove(net, "if_inet6");
3332 }
3333
3334 static struct pernet_operations if6_proc_net_ops = {
3335        .init = if6_proc_net_init,
3336        .exit = if6_proc_net_exit,
3337 };
3338
3339 int __init if6_proc_init(void)
3340 {
3341         return register_pernet_subsys(&if6_proc_net_ops);
3342 }
3343
3344 void if6_proc_exit(void)
3345 {
3346         unregister_pernet_subsys(&if6_proc_net_ops);
3347 }
3348 #endif  /* CONFIG_PROC_FS */
3349
3350 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3351 /* Check if address is a home address configured on any interface. */
3352 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3353 {
3354         int ret = 0;
3355         struct inet6_ifaddr *ifp = NULL;
3356         struct hlist_node *n;
3357         unsigned int hash = inet6_addr_hash(addr);
3358
3359         rcu_read_lock_bh();
3360         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3361                 if (!net_eq(dev_net(ifp->idev->dev), net))
3362                         continue;
3363                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3364                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3365                         ret = 1;
3366                         break;
3367                 }
3368         }
3369         rcu_read_unlock_bh();
3370         return ret;
3371 }
3372 #endif
3373
3374 /*
3375  *      Periodic address status verification
3376  */
3377
3378 static void addrconf_verify(unsigned long foo)
3379 {
3380         unsigned long now, next, next_sec, next_sched;
3381         struct inet6_ifaddr *ifp;
3382         struct hlist_node *node;
3383         int i;
3384
3385         rcu_read_lock_bh();
3386         spin_lock(&addrconf_verify_lock);
3387         now = jiffies;
3388         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3389
3390         del_timer(&addr_chk_timer);
3391
3392         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3393 restart:
3394                 hlist_for_each_entry_rcu_bh(ifp, node,
3395                                          &inet6_addr_lst[i], addr_lst) {
3396                         unsigned long age;
3397
3398                         if (ifp->flags & IFA_F_PERMANENT)
3399                                 continue;
3400
3401                         spin_lock(&ifp->lock);
3402                         /* We try to batch several events at once. */
3403                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3404
3405                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3406                             age >= ifp->valid_lft) {
3407                                 spin_unlock(&ifp->lock);
3408                                 in6_ifa_hold(ifp);
3409                                 ipv6_del_addr(ifp);
3410                                 goto restart;
3411                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3412                                 spin_unlock(&ifp->lock);
3413                                 continue;
3414                         } else if (age >= ifp->prefered_lft) {
3415                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3416                                 int deprecate = 0;
3417
3418                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3419                                         deprecate = 1;
3420                                         ifp->flags |= IFA_F_DEPRECATED;
3421                                 }
3422
3423                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3424                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3425
3426                                 spin_unlock(&ifp->lock);
3427
3428                                 if (deprecate) {
3429                                         in6_ifa_hold(ifp);
3430
3431                                         ipv6_ifa_notify(0, ifp);
3432                                         in6_ifa_put(ifp);
3433                                         goto restart;
3434                                 }
3435 #ifdef CONFIG_IPV6_PRIVACY
3436                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3437                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3438                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3439                                         ifp->idev->cnf.dad_transmits *
3440                                         ifp->idev->nd_parms->retrans_time / HZ;
3441
3442                                 if (age >= ifp->prefered_lft - regen_advance) {
3443                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3444                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3445                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3446                                         if (!ifp->regen_count && ifpub) {
3447                                                 ifp->regen_count++;
3448                                                 in6_ifa_hold(ifp);
3449                                                 in6_ifa_hold(ifpub);
3450                                                 spin_unlock(&ifp->lock);
3451
3452                                                 spin_lock(&ifpub->lock);
3453                                                 ifpub->regen_count = 0;
3454                                                 spin_unlock(&ifpub->lock);
3455                                                 ipv6_create_tempaddr(ifpub, ifp);
3456                                                 in6_ifa_put(ifpub);
3457                                                 in6_ifa_put(ifp);
3458                                                 goto restart;
3459                                         }
3460                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3461                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3462                                 spin_unlock(&ifp->lock);
3463 #endif
3464                         } else {
3465                                 /* ifp->prefered_lft <= ifp->valid_lft */
3466                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3467                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3468                                 spin_unlock(&ifp->lock);
3469                         }
3470                 }
3471         }
3472
3473         next_sec = round_jiffies_up(next);
3474         next_sched = next;
3475
3476         /* If rounded timeout is accurate enough, accept it. */
3477         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3478                 next_sched = next_sec;
3479
3480         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3481         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3482                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3483
3484         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3485               now, next, next_sec, next_sched));
3486
3487         addr_chk_timer.expires = next_sched;
3488         add_timer(&addr_chk_timer);
3489         spin_unlock(&addrconf_verify_lock);
3490         rcu_read_unlock_bh();
3491 }
3492
3493 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3494 {
3495         struct in6_addr *pfx = NULL;
3496
3497         if (addr)
3498                 pfx = nla_data(addr);
3499
3500         if (local) {
3501                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3502                         pfx = NULL;
3503                 else
3504                         pfx = nla_data(local);
3505         }
3506
3507         return pfx;
3508 }
3509
3510 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3511         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3512         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3513         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3514 };
3515
3516 static int
3517 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3518 {
3519         struct net *net = sock_net(skb->sk);
3520         struct ifaddrmsg *ifm;
3521         struct nlattr *tb[IFA_MAX+1];
3522         struct in6_addr *pfx;
3523         int err;
3524
3525         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3526         if (err < 0)
3527                 return err;
3528
3529         ifm = nlmsg_data(nlh);
3530         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3531         if (pfx == NULL)
3532                 return -EINVAL;
3533
3534         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3535 }
3536
3537 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3538                              u32 prefered_lft, u32 valid_lft)
3539 {
3540         u32 flags;
3541         clock_t expires;
3542         unsigned long timeout;
3543
3544         if (!valid_lft || (prefered_lft > valid_lft))
3545                 return -EINVAL;
3546
3547         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3548         if (addrconf_finite_timeout(timeout)) {
3549                 expires = jiffies_to_clock_t(timeout * HZ);
3550                 valid_lft = timeout;
3551                 flags = RTF_EXPIRES;
3552         } else {
3553                 expires = 0;
3554                 flags = 0;
3555                 ifa_flags |= IFA_F_PERMANENT;
3556         }
3557
3558         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3559         if (addrconf_finite_timeout(timeout)) {
3560                 if (timeout == 0)
3561                         ifa_flags |= IFA_F_DEPRECATED;
3562                 prefered_lft = timeout;
3563         }
3564
3565         spin_lock_bh(&ifp->lock);
3566         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3567         ifp->tstamp = jiffies;
3568         ifp->valid_lft = valid_lft;
3569         ifp->prefered_lft = prefered_lft;
3570
3571         spin_unlock_bh(&ifp->lock);
3572         if (!(ifp->flags&IFA_F_TENTATIVE))
3573                 ipv6_ifa_notify(0, ifp);
3574
3575         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3576                               expires, flags);
3577         addrconf_verify(0);
3578
3579         return 0;
3580 }
3581
3582 static int
3583 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3584 {
3585         struct net *net = sock_net(skb->sk);
3586         struct ifaddrmsg *ifm;
3587         struct nlattr *tb[IFA_MAX+1];
3588         struct in6_addr *pfx;
3589         struct inet6_ifaddr *ifa;
3590         struct net_device *dev;
3591         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3592         u8 ifa_flags;
3593         int err;
3594
3595         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3596         if (err < 0)
3597                 return err;
3598
3599         ifm = nlmsg_data(nlh);
3600         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3601         if (pfx == NULL)
3602                 return -EINVAL;
3603
3604         if (tb[IFA_CACHEINFO]) {
3605                 struct ifa_cacheinfo *ci;
3606
3607                 ci = nla_data(tb[IFA_CACHEINFO]);
3608                 valid_lft = ci->ifa_valid;
3609                 preferred_lft = ci->ifa_prefered;
3610         } else {
3611                 preferred_lft = INFINITY_LIFE_TIME;
3612                 valid_lft = INFINITY_LIFE_TIME;
3613         }
3614
3615         dev =  __dev_get_by_index(net, ifm->ifa_index);
3616         if (dev == NULL)
3617                 return -ENODEV;
3618
3619         /* We ignore other flags so far. */
3620         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3621
3622         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3623         if (ifa == NULL) {
3624                 /*
3625                  * It would be best to check for !NLM_F_CREATE here but
3626                  * userspace alreay relies on not having to provide this.
3627                  */
3628                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3629                                       ifm->ifa_prefixlen, ifa_flags,
3630                                       preferred_lft, valid_lft);
3631         }
3632
3633         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3634             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3635                 err = -EEXIST;
3636         else
3637                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3638
3639         in6_ifa_put(ifa);
3640
3641         return err;
3642 }
3643
3644 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3645                           u8 scope, int ifindex)
3646 {
3647         struct ifaddrmsg *ifm;
3648
3649         ifm = nlmsg_data(nlh);
3650         ifm->ifa_family = AF_INET6;
3651         ifm->ifa_prefixlen = prefixlen;
3652         ifm->ifa_flags = flags;
3653         ifm->ifa_scope = scope;
3654         ifm->ifa_index = ifindex;
3655 }
3656
3657 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3658                          unsigned long tstamp, u32 preferred, u32 valid)
3659 {
3660         struct ifa_cacheinfo ci;
3661
3662         ci.cstamp = cstamp_delta(cstamp);
3663         ci.tstamp = cstamp_delta(tstamp);
3664         ci.ifa_prefered = preferred;
3665         ci.ifa_valid = valid;
3666
3667         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3668 }
3669
3670 static inline int rt_scope(int ifa_scope)
3671 {
3672         if (ifa_scope & IFA_HOST)
3673                 return RT_SCOPE_HOST;
3674         else if (ifa_scope & IFA_LINK)
3675                 return RT_SCOPE_LINK;
3676         else if (ifa_scope & IFA_SITE)
3677                 return RT_SCOPE_SITE;
3678         else
3679                 return RT_SCOPE_UNIVERSE;
3680 }
3681
3682 static inline int inet6_ifaddr_msgsize(void)
3683 {
3684         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3685                + nla_total_size(16) /* IFA_ADDRESS */
3686                + nla_total_size(sizeof(struct ifa_cacheinfo));
3687 }
3688
3689 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3690                              u32 portid, u32 seq, int event, unsigned int flags)
3691 {
3692         struct nlmsghdr  *nlh;
3693         u32 preferred, valid;
3694
3695         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3696         if (nlh == NULL)
3697                 return -EMSGSIZE;
3698
3699         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3700                       ifa->idev->dev->ifindex);
3701
3702         if (!(ifa->flags&IFA_F_PERMANENT)) {
3703                 preferred = ifa->prefered_lft;
3704                 valid = ifa->valid_lft;
3705                 if (preferred != INFINITY_LIFE_TIME) {
3706                         long tval = (jiffies - ifa->tstamp)/HZ;
3707                         if (preferred > tval)
3708                                 preferred -= tval;
3709                         else
3710                                 preferred = 0;
3711                         if (valid != INFINITY_LIFE_TIME) {
3712                                 if (valid > tval)
3713                                         valid -= tval;
3714                                 else
3715                                         valid = 0;
3716                         }
3717                 }
3718         } else {
3719                 preferred = INFINITY_LIFE_TIME;
3720                 valid = INFINITY_LIFE_TIME;
3721         }
3722
3723         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3724             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3725                 nlmsg_cancel(skb, nlh);
3726                 return -EMSGSIZE;
3727         }
3728
3729         return nlmsg_end(skb, nlh);
3730 }
3731
3732 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3733                                 u32 portid, u32 seq, int event, u16 flags)
3734 {
3735         struct nlmsghdr  *nlh;
3736         u8 scope = RT_SCOPE_UNIVERSE;
3737         int ifindex = ifmca->idev->dev->ifindex;
3738
3739         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3740                 scope = RT_SCOPE_SITE;
3741
3742         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3743         if (nlh == NULL)
3744                 return -EMSGSIZE;
3745
3746         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3747         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3748             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3749                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3750                 nlmsg_cancel(skb, nlh);
3751                 return -EMSGSIZE;
3752         }
3753
3754         return nlmsg_end(skb, nlh);
3755 }
3756
3757 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3758                                 u32 portid, u32 seq, int event, unsigned int flags)
3759 {
3760         struct nlmsghdr  *nlh;
3761         u8 scope = RT_SCOPE_UNIVERSE;
3762         int ifindex = ifaca->aca_idev->dev->ifindex;
3763
3764         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3765                 scope = RT_SCOPE_SITE;
3766
3767         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3768         if (nlh == NULL)
3769                 return -EMSGSIZE;
3770
3771         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3772         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3773             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3774                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3775                 nlmsg_cancel(skb, nlh);
3776                 return -EMSGSIZE;
3777         }
3778
3779         return nlmsg_end(skb, nlh);
3780 }
3781
3782 enum addr_type_t {
3783         UNICAST_ADDR,
3784         MULTICAST_ADDR,
3785         ANYCAST_ADDR,
3786 };
3787
3788 /* called with rcu_read_lock() */
3789 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3790                           struct netlink_callback *cb, enum addr_type_t type,
3791                           int s_ip_idx, int *p_ip_idx)
3792 {
3793         struct ifmcaddr6 *ifmca;
3794         struct ifacaddr6 *ifaca;
3795         int err = 1;
3796         int ip_idx = *p_ip_idx;
3797
3798         read_lock_bh(&idev->lock);
3799         switch (type) {
3800         case UNICAST_ADDR: {
3801                 struct inet6_ifaddr *ifa;
3802
3803                 /* unicast address incl. temp addr */
3804                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3805                         if (++ip_idx < s_ip_idx)
3806                                 continue;
3807                         err = inet6_fill_ifaddr(skb, ifa,
3808                                                 NETLINK_CB(cb->skb).portid,
3809                                                 cb->nlh->nlmsg_seq,
3810                                                 RTM_NEWADDR,
3811                                                 NLM_F_MULTI);
3812                         if (err <= 0)
3813                                 break;
3814                 }
3815                 break;
3816         }
3817         case MULTICAST_ADDR:
3818                 /* multicast address */
3819                 for (ifmca = idev->mc_list; ifmca;
3820                      ifmca = ifmca->next, ip_idx++) {
3821                         if (ip_idx < s_ip_idx)
3822                                 continue;
3823                         err = inet6_fill_ifmcaddr(skb, ifmca,
3824                                                   NETLINK_CB(cb->skb).portid,
3825                                                   cb->nlh->nlmsg_seq,
3826                                                   RTM_GETMULTICAST,
3827                                                   NLM_F_MULTI);
3828                         if (err <= 0)
3829                                 break;
3830                 }
3831                 break;
3832         case ANYCAST_ADDR:
3833                 /* anycast address */
3834                 for (ifaca = idev->ac_list; ifaca;
3835                      ifaca = ifaca->aca_next, ip_idx++) {
3836                         if (ip_idx < s_ip_idx)
3837                                 continue;
3838                         err = inet6_fill_ifacaddr(skb, ifaca,
3839                                                   NETLINK_CB(cb->skb).portid,
3840                                                   cb->nlh->nlmsg_seq,
3841                                                   RTM_GETANYCAST,
3842                                                   NLM_F_MULTI);
3843                         if (err <= 0)
3844                                 break;
3845                 }
3846                 break;
3847         default:
3848                 break;
3849         }
3850         read_unlock_bh(&idev->lock);
3851         *p_ip_idx = ip_idx;
3852         return err;
3853 }
3854
3855 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3856                            enum addr_type_t type)
3857 {
3858         struct net *net = sock_net(skb->sk);
3859         int h, s_h;
3860         int idx, ip_idx;
3861         int s_idx, s_ip_idx;
3862         struct net_device *dev;
3863         struct inet6_dev *idev;
3864         struct hlist_head *head;
3865         struct hlist_node *node;
3866
3867         s_h = cb->args[0];
3868         s_idx = idx = cb->args[1];
3869         s_ip_idx = ip_idx = cb->args[2];
3870
3871         rcu_read_lock();
3872         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3873                 idx = 0;
3874                 head = &net->dev_index_head[h];
3875                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3876                         if (idx < s_idx)
3877                                 goto cont;
3878                         if (h > s_h || idx > s_idx)
3879                                 s_ip_idx = 0;
3880                         ip_idx = 0;
3881                         idev = __in6_dev_get(dev);
3882                         if (!idev)
3883                                 goto cont;
3884
3885                         if (in6_dump_addrs(idev, skb, cb, type,
3886                                            s_ip_idx, &ip_idx) <= 0)
3887                                 goto done;
3888 cont:
3889                         idx++;
3890                 }
3891         }
3892 done:
3893         rcu_read_unlock();
3894         cb->args[0] = h;
3895         cb->args[1] = idx;
3896         cb->args[2] = ip_idx;
3897
3898         return skb->len;
3899 }
3900
3901 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3902 {
3903         enum addr_type_t type = UNICAST_ADDR;
3904
3905         return inet6_dump_addr(skb, cb, type);
3906 }
3907
3908 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3909 {
3910         enum addr_type_t type = MULTICAST_ADDR;
3911
3912         return inet6_dump_addr(skb, cb, type);
3913 }
3914
3915
3916 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3917 {
3918         enum addr_type_t type = ANYCAST_ADDR;
3919
3920         return inet6_dump_addr(skb, cb, type);
3921 }
3922
3923 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3924                              void *arg)
3925 {
3926         struct net *net = sock_net(in_skb->sk);
3927         struct ifaddrmsg *ifm;
3928         struct nlattr *tb[IFA_MAX+1];
3929         struct in6_addr *addr = NULL;
3930         struct net_device *dev = NULL;
3931         struct inet6_ifaddr *ifa;
3932         struct sk_buff *skb;
3933         int err;
3934
3935         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3936         if (err < 0)
3937                 goto errout;
3938
3939         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3940         if (addr == NULL) {
3941                 err = -EINVAL;
3942                 goto errout;
3943         }
3944
3945         ifm = nlmsg_data(nlh);
3946         if (ifm->ifa_index)
3947                 dev = __dev_get_by_index(net, ifm->ifa_index);
3948
3949         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3950         if (!ifa) {
3951                 err = -EADDRNOTAVAIL;
3952                 goto errout;
3953         }
3954
3955         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3956         if (!skb) {
3957                 err = -ENOBUFS;
3958                 goto errout_ifa;
3959         }
3960
3961         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3962                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3963         if (err < 0) {
3964                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3965                 WARN_ON(err == -EMSGSIZE);
3966                 kfree_skb(skb);
3967                 goto errout_ifa;
3968         }
3969         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3970 errout_ifa:
3971         in6_ifa_put(ifa);
3972 errout:
3973         return err;
3974 }
3975
3976 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3977 {
3978         struct sk_buff *skb;
3979         struct net *net = dev_net(ifa->idev->dev);
3980         int err = -ENOBUFS;
3981
3982         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3983         if (skb == NULL)
3984                 goto errout;
3985
3986         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3987         if (err < 0) {
3988                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3989                 WARN_ON(err == -EMSGSIZE);
3990                 kfree_skb(skb);
3991                 goto errout;
3992         }
3993         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3994         return;
3995 errout:
3996         if (err < 0)
3997                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3998 }
3999
4000 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4001                                 __s32 *array, int bytes)
4002 {
4003         BUG_ON(bytes < (DEVCONF_MAX * 4));
4004
4005         memset(array, 0, bytes);
4006         array[DEVCONF_FORWARDING] = cnf->forwarding;
4007         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4008         array[DEVCONF_MTU6] = cnf->mtu6;
4009         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4010         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4011         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4012         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4013         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4014         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4015                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4016         array[DEVCONF_RTR_SOLICIT_DELAY] =
4017                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4018         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4019 #ifdef CONFIG_IPV6_PRIVACY
4020         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4021         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4022         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4023         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4024         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4025 #endif
4026         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4027         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4028         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4029 #ifdef CONFIG_IPV6_ROUTER_PREF
4030         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4031         array[DEVCONF_RTR_PROBE_INTERVAL] =
4032                 jiffies_to_msecs(cnf->rtr_probe_interval);
4033 #ifdef CONFIG_IPV6_ROUTE_INFO
4034         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4035 #endif
4036 #endif
4037         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4038         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4039 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4040         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4041 #endif
4042 #ifdef CONFIG_IPV6_MROUTE
4043         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4044 #endif
4045         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4046         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4047         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4048         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4049 }
4050
4051 static inline size_t inet6_ifla6_size(void)
4052 {
4053         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4054              + nla_total_size(sizeof(struct ifla_cacheinfo))
4055              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4056              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4057              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4058 }
4059
4060 static inline size_t inet6_if_nlmsg_size(void)
4061 {
4062         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4063                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4064                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4065                + nla_total_size(4) /* IFLA_MTU */
4066                + nla_total_size(4) /* IFLA_LINK */
4067                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4068 }
4069
4070 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4071                                       int items, int bytes)
4072 {
4073         int i;
4074         int pad = bytes - sizeof(u64) * items;
4075         BUG_ON(pad < 0);
4076
4077         /* Use put_unaligned() because stats may not be aligned for u64. */
4078         put_unaligned(items, &stats[0]);
4079         for (i = 1; i < items; i++)
4080                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4081
4082         memset(&stats[items], 0, pad);
4083 }
4084
4085 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4086                                       int items, int bytes, size_t syncpoff)
4087 {
4088         int i;
4089         int pad = bytes - sizeof(u64) * items;
4090         BUG_ON(pad < 0);
4091
4092         /* Use put_unaligned() because stats may not be aligned for u64. */
4093         put_unaligned(items, &stats[0]);
4094         for (i = 1; i < items; i++)
4095                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4096
4097         memset(&stats[items], 0, pad);
4098 }
4099
4100 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4101                              int bytes)
4102 {
4103         switch (attrtype) {
4104         case IFLA_INET6_STATS:
4105                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4106                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4107                 break;
4108         case IFLA_INET6_ICMP6STATS:
4109                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4110                 break;
4111         }
4112 }
4113
4114 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4115 {
4116         struct nlattr *nla;
4117         struct ifla_cacheinfo ci;
4118
4119         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4120                 goto nla_put_failure;
4121         ci.max_reasm_len = IPV6_MAXPLEN;
4122         ci.tstamp = cstamp_delta(idev->tstamp);
4123         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4124         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4125         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4126                 goto nla_put_failure;
4127         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4128         if (nla == NULL)
4129                 goto nla_put_failure;
4130         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4131
4132         /* XXX - MC not implemented */
4133
4134         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4135         if (nla == NULL)
4136                 goto nla_put_failure;
4137         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4138
4139         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4140         if (nla == NULL)
4141                 goto nla_put_failure;
4142         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4143
4144         return 0;
4145
4146 nla_put_failure:
4147         return -EMSGSIZE;
4148 }
4149
4150 static size_t inet6_get_link_af_size(const struct net_device *dev)
4151 {
4152         if (!__in6_dev_get(dev))
4153                 return 0;
4154
4155         return inet6_ifla6_size();
4156 }
4157
4158 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4159 {
4160         struct inet6_dev *idev = __in6_dev_get(dev);
4161
4162         if (!idev)
4163                 return -ENODATA;
4164
4165         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4166                 return -EMSGSIZE;
4167
4168         return 0;
4169 }
4170
4171 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4172                              u32 portid, u32 seq, int event, unsigned int flags)
4173 {
4174         struct net_device *dev = idev->dev;
4175         struct ifinfomsg *hdr;
4176         struct nlmsghdr *nlh;
4177         void *protoinfo;
4178
4179         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4180         if (nlh == NULL)
4181                 return -EMSGSIZE;
4182
4183         hdr = nlmsg_data(nlh);
4184         hdr->ifi_family = AF_INET6;
4185         hdr->__ifi_pad = 0;
4186         hdr->ifi_type = dev->type;
4187         hdr->ifi_index = dev->ifindex;
4188         hdr->ifi_flags = dev_get_flags(dev);
4189         hdr->ifi_change = 0;
4190
4191         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4192             (dev->addr_len &&
4193              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4194             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4195             (dev->ifindex != dev->iflink &&
4196              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4197                 goto nla_put_failure;
4198         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4199         if (protoinfo == NULL)
4200                 goto nla_put_failure;
4201
4202         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4203                 goto nla_put_failure;
4204
4205         nla_nest_end(skb, protoinfo);
4206         return nlmsg_end(skb, nlh);
4207
4208 nla_put_failure:
4209         nlmsg_cancel(skb, nlh);
4210         return -EMSGSIZE;
4211 }
4212
4213 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4214 {
4215         struct net *net = sock_net(skb->sk);
4216         int h, s_h;
4217         int idx = 0, s_idx;
4218         struct net_device *dev;
4219         struct inet6_dev *idev;
4220         struct hlist_head *head;
4221         struct hlist_node *node;
4222
4223         s_h = cb->args[0];
4224         s_idx = cb->args[1];
4225
4226         rcu_read_lock();
4227         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4228                 idx = 0;
4229                 head = &net->dev_index_head[h];
4230                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4231                         if (idx < s_idx)
4232                                 goto cont;
4233                         idev = __in6_dev_get(dev);
4234                         if (!idev)
4235                                 goto cont;
4236                         if (inet6_fill_ifinfo(skb, idev,
4237                                               NETLINK_CB(cb->skb).portid,
4238                                               cb->nlh->nlmsg_seq,
4239                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4240                                 goto out;
4241 cont:
4242                         idx++;
4243                 }
4244         }
4245 out:
4246         rcu_read_unlock();
4247         cb->args[1] = idx;
4248         cb->args[0] = h;
4249
4250         return skb->len;
4251 }
4252
4253 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4254 {
4255         struct sk_buff *skb;
4256         struct net *net = dev_net(idev->dev);
4257         int err = -ENOBUFS;
4258
4259         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4260         if (skb == NULL)
4261                 goto errout;
4262
4263         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4264         if (err < 0) {
4265                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4266                 WARN_ON(err == -EMSGSIZE);
4267                 kfree_skb(skb);
4268                 goto errout;
4269         }
4270         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4271         return;
4272 errout:
4273         if (err < 0)
4274                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4275 }
4276
4277 static inline size_t inet6_prefix_nlmsg_size(void)
4278 {
4279         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4280                + nla_total_size(sizeof(struct in6_addr))
4281                + nla_total_size(sizeof(struct prefix_cacheinfo));
4282 }
4283
4284 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4285                              struct prefix_info *pinfo, u32 portid, u32 seq,
4286                              int event, unsigned int flags)
4287 {
4288         struct prefixmsg *pmsg;
4289         struct nlmsghdr *nlh;
4290         struct prefix_cacheinfo ci;
4291
4292         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4293         if (nlh == NULL)
4294                 return -EMSGSIZE;
4295
4296         pmsg = nlmsg_data(nlh);
4297         pmsg->prefix_family = AF_INET6;
4298         pmsg->prefix_pad1 = 0;
4299         pmsg->prefix_pad2 = 0;
4300         pmsg->prefix_ifindex = idev->dev->ifindex;
4301         pmsg->prefix_len = pinfo->prefix_len;
4302         pmsg->prefix_type = pinfo->type;
4303         pmsg->prefix_pad3 = 0;
4304         pmsg->prefix_flags = 0;
4305         if (pinfo->onlink)
4306                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4307         if (pinfo->autoconf)
4308                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4309
4310         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4311                 goto nla_put_failure;
4312         ci.preferred_time = ntohl(pinfo->prefered);
4313         ci.valid_time = ntohl(pinfo->valid);
4314         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4315                 goto nla_put_failure;
4316         return nlmsg_end(skb, nlh);
4317
4318 nla_put_failure:
4319         nlmsg_cancel(skb, nlh);
4320         return -EMSGSIZE;
4321 }
4322
4323 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4324                          struct prefix_info *pinfo)
4325 {
4326         struct sk_buff *skb;
4327         struct net *net = dev_net(idev->dev);
4328         int err = -ENOBUFS;
4329
4330         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4331         if (skb == NULL)
4332                 goto errout;
4333
4334         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4335         if (err < 0) {
4336                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4337                 WARN_ON(err == -EMSGSIZE);
4338                 kfree_skb(skb);
4339                 goto errout;
4340         }
4341         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4342         return;
4343 errout:
4344         if (err < 0)
4345                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4346 }
4347
4348 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4349 {
4350         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4351
4352         switch (event) {
4353         case RTM_NEWADDR:
4354                 /*
4355                  * If the address was optimistic
4356                  * we inserted the route at the start of
4357                  * our DAD process, so we don't need
4358                  * to do it again
4359                  */
4360                 if (!(ifp->rt->rt6i_node))
4361                         ip6_ins_rt(ifp->rt);
4362                 if (ifp->idev->cnf.forwarding)
4363                         addrconf_join_anycast(ifp);
4364                 break;
4365         case RTM_DELADDR:
4366                 if (ifp->idev->cnf.forwarding)
4367                         addrconf_leave_anycast(ifp);
4368                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4369                 dst_hold(&ifp->rt->dst);
4370
4371                 if (ip6_del_rt(ifp->rt))
4372                         dst_free(&ifp->rt->dst);
4373                 break;
4374         }
4375 }
4376
4377 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4378 {
4379         rcu_read_lock_bh();
4380         if (likely(ifp->idev->dead == 0))
4381                 __ipv6_ifa_notify(event, ifp);
4382         rcu_read_unlock_bh();
4383 }
4384
4385 #ifdef CONFIG_SYSCTL
4386
4387 static
4388 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4389                            void __user *buffer, size_t *lenp, loff_t *ppos)
4390 {
4391         int *valp = ctl->data;
4392         int val = *valp;
4393         loff_t pos = *ppos;
4394         ctl_table lctl;
4395         int ret;
4396
4397         /*
4398          * ctl->data points to idev->cnf.forwarding, we should
4399          * not modify it until we get the rtnl lock.
4400          */
4401         lctl = *ctl;
4402         lctl.data = &val;
4403
4404         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4405
4406         if (write)
4407                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4408         if (ret)
4409                 *ppos = pos;
4410         return ret;
4411 }
4412
4413 static void dev_disable_change(struct inet6_dev *idev)
4414 {
4415         if (!idev || !idev->dev)
4416                 return;
4417
4418         if (idev->cnf.disable_ipv6)
4419                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4420         else
4421                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4422 }
4423
4424 static void addrconf_disable_change(struct net *net, __s32 newf)
4425 {
4426         struct net_device *dev;
4427         struct inet6_dev *idev;
4428
4429         rcu_read_lock();
4430         for_each_netdev_rcu(net, dev) {
4431                 idev = __in6_dev_get(dev);
4432                 if (idev) {
4433                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4434                         idev->cnf.disable_ipv6 = newf;
4435                         if (changed)
4436                                 dev_disable_change(idev);
4437                 }
4438         }
4439         rcu_read_unlock();
4440 }
4441
4442 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4443 {
4444         struct net *net;
4445         int old;
4446
4447         if (!rtnl_trylock())
4448                 return restart_syscall();
4449
4450         net = (struct net *)table->extra2;
4451         old = *p;
4452         *p = newf;
4453
4454         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4455                 rtnl_unlock();
4456                 return 0;
4457         }
4458
4459         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4460                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4461                 addrconf_disable_change(net, newf);
4462         } else if ((!newf) ^ (!old))
4463                 dev_disable_change((struct inet6_dev *)table->extra1);
4464
4465         rtnl_unlock();
4466         return 0;
4467 }
4468
4469 static
4470 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4471                             void __user *buffer, size_t *lenp, loff_t *ppos)
4472 {
4473         int *valp = ctl->data;
4474         int val = *valp;
4475         loff_t pos = *ppos;
4476         ctl_table lctl;
4477         int ret;
4478
4479         /*
4480          * ctl->data points to idev->cnf.disable_ipv6, we should
4481          * not modify it until we get the rtnl lock.
4482          */
4483         lctl = *ctl;
4484         lctl.data = &val;
4485
4486         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4487
4488         if (write)
4489                 ret = addrconf_disable_ipv6(ctl, valp, val);
4490         if (ret)
4491                 *ppos = pos;
4492         return ret;
4493 }
4494
4495 static struct addrconf_sysctl_table
4496 {
4497         struct ctl_table_header *sysctl_header;
4498         ctl_table addrconf_vars[DEVCONF_MAX+1];
4499 } addrconf_sysctl __read_mostly = {
4500         .sysctl_header = NULL,
4501         .addrconf_vars = {
4502                 {
4503                         .procname       = "forwarding",
4504                         .data           = &ipv6_devconf.forwarding,
4505                         .maxlen         = sizeof(int),
4506                         .mode           = 0644,
4507                         .proc_handler   = addrconf_sysctl_forward,
4508                 },
4509                 {
4510                         .procname       = "hop_limit",
4511                         .data           = &ipv6_devconf.hop_limit,
4512                         .maxlen         = sizeof(int),
4513                         .mode           = 0644,
4514                         .proc_handler   = proc_dointvec,
4515                 },
4516                 {
4517                         .procname       = "mtu",
4518                         .data           = &ipv6_devconf.mtu6,
4519                         .maxlen         = sizeof(int),
4520                         .mode           = 0644,
4521                         .proc_handler   = proc_dointvec,
4522                 },
4523                 {
4524                         .procname       = "accept_ra",
4525                         .data           = &ipv6_devconf.accept_ra,
4526                         .maxlen         = sizeof(int),
4527                         .mode           = 0644,
4528                         .proc_handler   = proc_dointvec,
4529                 },
4530                 {
4531                         .procname       = "accept_redirects",
4532                         .data           = &ipv6_devconf.accept_redirects,
4533                         .maxlen         = sizeof(int),
4534                         .mode           = 0644,
4535                         .proc_handler   = proc_dointvec,
4536                 },
4537                 {
4538                         .procname       = "autoconf",
4539                         .data           = &ipv6_devconf.autoconf,
4540                         .maxlen         = sizeof(int),
4541                         .mode           = 0644,
4542                         .proc_handler   = proc_dointvec,
4543                 },
4544                 {
4545                         .procname       = "dad_transmits",
4546                         .data           = &ipv6_devconf.dad_transmits,
4547                         .maxlen         = sizeof(int),
4548                         .mode           = 0644,
4549                         .proc_handler   = proc_dointvec,
4550                 },
4551                 {
4552                         .procname       = "router_solicitations",
4553                         .data           = &ipv6_devconf.rtr_solicits,
4554                         .maxlen         = sizeof(int),
4555                         .mode           = 0644,
4556                         .proc_handler   = proc_dointvec,
4557                 },
4558                 {
4559                         .procname       = "router_solicitation_interval",
4560                         .data           = &ipv6_devconf.rtr_solicit_interval,
4561                         .maxlen         = sizeof(int),
4562                         .mode           = 0644,
4563                         .proc_handler   = proc_dointvec_jiffies,
4564                 },
4565                 {
4566                         .procname       = "router_solicitation_delay",
4567                         .data           = &ipv6_devconf.rtr_solicit_delay,
4568                         .maxlen         = sizeof(int),
4569                         .mode           = 0644,
4570                         .proc_handler   = proc_dointvec_jiffies,
4571                 },
4572                 {
4573                         .procname       = "force_mld_version",
4574                         .data           = &ipv6_devconf.force_mld_version,
4575                         .maxlen         = sizeof(int),
4576                         .mode           = 0644,
4577                         .proc_handler   = proc_dointvec,
4578                 },
4579 #ifdef CONFIG_IPV6_PRIVACY
4580                 {
4581                         .procname       = "use_tempaddr",
4582                         .data           = &ipv6_devconf.use_tempaddr,
4583                         .maxlen         = sizeof(int),
4584                         .mode           = 0644,
4585                         .proc_handler   = proc_dointvec,
4586                 },
4587                 {
4588                         .procname       = "temp_valid_lft",
4589                         .data           = &ipv6_devconf.temp_valid_lft,
4590                         .maxlen         = sizeof(int),
4591                         .mode           = 0644,
4592                         .proc_handler   = proc_dointvec,
4593                 },
4594                 {
4595                         .procname       = "temp_prefered_lft",
4596                         .data           = &ipv6_devconf.temp_prefered_lft,
4597                         .maxlen         = sizeof(int),
4598                         .mode           = 0644,
4599                         .proc_handler   = proc_dointvec,
4600                 },
4601                 {
4602                         .procname       = "regen_max_retry",
4603                         .data           = &ipv6_devconf.regen_max_retry,
4604                         .maxlen         = sizeof(int),
4605                         .mode           = 0644,
4606                         .proc_handler   = proc_dointvec,
4607                 },
4608                 {
4609                         .procname       = "max_desync_factor",
4610                         .data           = &ipv6_devconf.max_desync_factor,
4611                         .maxlen         = sizeof(int),
4612                         .mode           = 0644,
4613                         .proc_handler   = proc_dointvec,
4614                 },
4615 #endif
4616                 {
4617                         .procname       = "max_addresses",
4618                         .data           = &ipv6_devconf.max_addresses,
4619                         .maxlen         = sizeof(int),
4620                         .mode           = 0644,
4621                         .proc_handler   = proc_dointvec,
4622                 },
4623                 {
4624                         .procname       = "accept_ra_defrtr",
4625                         .data           = &ipv6_devconf.accept_ra_defrtr,
4626                         .maxlen         = sizeof(int),
4627                         .mode           = 0644,
4628                         .proc_handler   = proc_dointvec,
4629                 },
4630                 {
4631                         .procname       = "accept_ra_pinfo",
4632                         .data           = &ipv6_devconf.accept_ra_pinfo,
4633                         .maxlen         = sizeof(int),
4634                         .mode           = 0644,
4635                         .proc_handler   = proc_dointvec,
4636                 },
4637 #ifdef CONFIG_IPV6_ROUTER_PREF
4638                 {
4639                         .procname       = "accept_ra_rtr_pref",
4640                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4641                         .maxlen         = sizeof(int),
4642                         .mode           = 0644,
4643                         .proc_handler   = proc_dointvec,
4644                 },
4645                 {
4646                         .procname       = "router_probe_interval",
4647                         .data           = &ipv6_devconf.rtr_probe_interval,
4648                         .maxlen         = sizeof(int),
4649                         .mode           = 0644,
4650                         .proc_handler   = proc_dointvec_jiffies,
4651                 },
4652 #ifdef CONFIG_IPV6_ROUTE_INFO
4653                 {
4654                         .procname       = "accept_ra_rt_info_max_plen",
4655                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4656                         .maxlen         = sizeof(int),
4657                         .mode           = 0644,
4658                         .proc_handler   = proc_dointvec,
4659                 },
4660 #endif
4661 #endif
4662                 {
4663                         .procname       = "proxy_ndp",
4664                         .data           = &ipv6_devconf.proxy_ndp,
4665                         .maxlen         = sizeof(int),
4666                         .mode           = 0644,
4667                         .proc_handler   = proc_dointvec,
4668                 },
4669                 {
4670                         .procname       = "accept_source_route",
4671                         .data           = &ipv6_devconf.accept_source_route,
4672                         .maxlen         = sizeof(int),
4673                         .mode           = 0644,
4674                         .proc_handler   = proc_dointvec,
4675                 },
4676 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4677                 {
4678                         .procname       = "optimistic_dad",
4679                         .data           = &ipv6_devconf.optimistic_dad,
4680                         .maxlen         = sizeof(int),
4681                         .mode           = 0644,
4682                         .proc_handler   = proc_dointvec,
4683
4684                 },
4685 #endif
4686 #ifdef CONFIG_IPV6_MROUTE
4687                 {
4688                         .procname       = "mc_forwarding",
4689                         .data           = &ipv6_devconf.mc_forwarding,
4690                         .maxlen         = sizeof(int),
4691                         .mode           = 0444,
4692                         .proc_handler   = proc_dointvec,
4693                 },
4694 #endif
4695                 {
4696                         .procname       = "disable_ipv6",
4697                         .data           = &ipv6_devconf.disable_ipv6,
4698                         .maxlen         = sizeof(int),
4699                         .mode           = 0644,
4700                         .proc_handler   = addrconf_sysctl_disable,
4701                 },
4702                 {
4703                         .procname       = "accept_dad",
4704                         .data           = &ipv6_devconf.accept_dad,
4705                         .maxlen         = sizeof(int),
4706                         .mode           = 0644,
4707                         .proc_handler   = proc_dointvec,
4708                 },
4709                 {
4710                         .procname       = "force_tllao",
4711                         .data           = &ipv6_devconf.force_tllao,
4712                         .maxlen         = sizeof(int),
4713                         .mode           = 0644,
4714                         .proc_handler   = proc_dointvec
4715                 },
4716                 {
4717                         .procname       = "ndisc_notify",
4718                         .data           = &ipv6_devconf.ndisc_notify,
4719                         .maxlen         = sizeof(int),
4720                         .mode           = 0644,
4721                         .proc_handler   = proc_dointvec
4722                 },
4723                 {
4724                         /* sentinel */
4725                 }
4726         },
4727 };
4728
4729 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4730                 struct inet6_dev *idev, struct ipv6_devconf *p)
4731 {
4732         int i;
4733         struct addrconf_sysctl_table *t;
4734         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4735
4736         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4737         if (t == NULL)
4738                 goto out;
4739
4740         for (i = 0; t->addrconf_vars[i].data; i++) {
4741                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4742                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4743                 t->addrconf_vars[i].extra2 = net;
4744         }
4745
4746         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4747
4748         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4749         if (t->sysctl_header == NULL)
4750                 goto free;
4751
4752         p->sysctl = t;
4753         return 0;
4754
4755 free:
4756         kfree(t);
4757 out:
4758         return -ENOBUFS;
4759 }
4760
4761 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4762 {
4763         struct addrconf_sysctl_table *t;
4764
4765         if (p->sysctl == NULL)
4766                 return;
4767
4768         t = p->sysctl;
4769         p->sysctl = NULL;
4770         unregister_net_sysctl_table(t->sysctl_header);
4771         kfree(t);
4772 }
4773
4774 static void addrconf_sysctl_register(struct inet6_dev *idev)
4775 {
4776         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4777                               &ndisc_ifinfo_sysctl_change);
4778         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4779                                         idev, &idev->cnf);
4780 }
4781
4782 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4783 {
4784         __addrconf_sysctl_unregister(&idev->cnf);
4785         neigh_sysctl_unregister(idev->nd_parms);
4786 }
4787
4788
4789 #endif
4790
4791 static int __net_init addrconf_init_net(struct net *net)
4792 {
4793         int err;
4794         struct ipv6_devconf *all, *dflt;
4795
4796         err = -ENOMEM;
4797         all = &ipv6_devconf;
4798         dflt = &ipv6_devconf_dflt;
4799
4800         if (!net_eq(net, &init_net)) {
4801                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4802                 if (all == NULL)
4803                         goto err_alloc_all;
4804
4805                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4806                 if (dflt == NULL)
4807                         goto err_alloc_dflt;
4808         } else {
4809                 /* these will be inherited by all namespaces */
4810                 dflt->autoconf = ipv6_defaults.autoconf;
4811                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4812         }
4813
4814         net->ipv6.devconf_all = all;
4815         net->ipv6.devconf_dflt = dflt;
4816
4817 #ifdef CONFIG_SYSCTL
4818         err = __addrconf_sysctl_register(net, "all", NULL, all);
4819         if (err < 0)
4820                 goto err_reg_all;
4821
4822         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4823         if (err < 0)
4824                 goto err_reg_dflt;
4825 #endif
4826         return 0;
4827
4828 #ifdef CONFIG_SYSCTL
4829 err_reg_dflt:
4830         __addrconf_sysctl_unregister(all);
4831 err_reg_all:
4832         kfree(dflt);
4833 #endif
4834 err_alloc_dflt:
4835         kfree(all);
4836 err_alloc_all:
4837         return err;
4838 }
4839
4840 static void __net_exit addrconf_exit_net(struct net *net)
4841 {
4842 #ifdef CONFIG_SYSCTL
4843         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4844         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4845 #endif
4846         if (!net_eq(net, &init_net)) {
4847                 kfree(net->ipv6.devconf_dflt);
4848                 kfree(net->ipv6.devconf_all);
4849         }
4850 }
4851
4852 static struct pernet_operations addrconf_ops = {
4853         .init = addrconf_init_net,
4854         .exit = addrconf_exit_net,
4855 };
4856
4857 /*
4858  *      Device notifier
4859  */
4860
4861 int register_inet6addr_notifier(struct notifier_block *nb)
4862 {
4863         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4864 }
4865 EXPORT_SYMBOL(register_inet6addr_notifier);
4866
4867 int unregister_inet6addr_notifier(struct notifier_block *nb)
4868 {
4869         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4870 }
4871 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4872
4873 static struct rtnl_af_ops inet6_ops = {
4874         .family           = AF_INET6,
4875         .fill_link_af     = inet6_fill_link_af,
4876         .get_link_af_size = inet6_get_link_af_size,
4877 };
4878
4879 /*
4880  *      Init / cleanup code
4881  */
4882
4883 int __init addrconf_init(void)
4884 {
4885         int i, err;
4886
4887         err = ipv6_addr_label_init();
4888         if (err < 0) {
4889                 pr_crit("%s: cannot initialize default policy table: %d\n",
4890                         __func__, err);
4891                 goto out;
4892         }
4893
4894         err = register_pernet_subsys(&addrconf_ops);
4895         if (err < 0)
4896                 goto out_addrlabel;
4897
4898         /* The addrconf netdev notifier requires that loopback_dev
4899          * has it's ipv6 private information allocated and setup
4900          * before it can bring up and give link-local addresses
4901          * to other devices which are up.
4902          *
4903          * Unfortunately, loopback_dev is not necessarily the first
4904          * entry in the global dev_base list of net devices.  In fact,
4905          * it is likely to be the very last entry on that list.
4906          * So this causes the notifier registry below to try and
4907          * give link-local addresses to all devices besides loopback_dev
4908          * first, then loopback_dev, which cases all the non-loopback_dev
4909          * devices to fail to get a link-local address.
4910          *
4911          * So, as a temporary fix, allocate the ipv6 structure for
4912          * loopback_dev first by hand.
4913          * Longer term, all of the dependencies ipv6 has upon the loopback
4914          * device and it being up should be removed.
4915          */
4916         rtnl_lock();
4917         if (!ipv6_add_dev(init_net.loopback_dev))
4918                 err = -ENOMEM;
4919         rtnl_unlock();
4920         if (err)
4921                 goto errlo;
4922
4923         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4924                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4925
4926         register_netdevice_notifier(&ipv6_dev_notf);
4927
4928         addrconf_verify(0);
4929
4930         err = rtnl_af_register(&inet6_ops);
4931         if (err < 0)
4932                 goto errout_af;
4933
4934         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4935                               NULL);
4936         if (err < 0)
4937                 goto errout;
4938
4939         /* Only the first call to __rtnl_register can fail */
4940         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4941         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4942         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4943                         inet6_dump_ifaddr, NULL);
4944         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4945                         inet6_dump_ifmcaddr, NULL);
4946         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4947                         inet6_dump_ifacaddr, NULL);
4948         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4949                         NULL, NULL);
4950
4951         ipv6_addr_label_rtnl_register();
4952
4953         return 0;
4954 errout:
4955         rtnl_af_unregister(&inet6_ops);
4956 errout_af:
4957         unregister_netdevice_notifier(&ipv6_dev_notf);
4958 errlo:
4959         unregister_pernet_subsys(&addrconf_ops);
4960 out_addrlabel:
4961         ipv6_addr_label_cleanup();
4962 out:
4963         return err;
4964 }
4965
4966 void addrconf_cleanup(void)
4967 {
4968         struct net_device *dev;
4969         int i;
4970
4971         unregister_netdevice_notifier(&ipv6_dev_notf);
4972         unregister_pernet_subsys(&addrconf_ops);
4973         ipv6_addr_label_cleanup();
4974
4975         rtnl_lock();
4976
4977         __rtnl_af_unregister(&inet6_ops);
4978
4979         /* clean dev list */
4980         for_each_netdev(&init_net, dev) {
4981                 if (__in6_dev_get(dev) == NULL)
4982                         continue;
4983                 addrconf_ifdown(dev, 1);
4984         }
4985         addrconf_ifdown(init_net.loopback_dev, 2);
4986
4987         /*
4988          *      Check hash table.
4989          */
4990         spin_lock_bh(&addrconf_hash_lock);
4991         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4992                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4993         spin_unlock_bh(&addrconf_hash_lock);
4994
4995         del_timer(&addr_chk_timer);
4996         rtnl_unlock();
4997 }