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