]> Pileus Git - ~andy/linux/blob - net/ipv6/addrconf.c
22ae75d5401794b0ddca846eccb863f957538a50
[~andy/linux] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
88 #endif
89
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
99
100 #if ACONF_DEBUG >= 3
101 #define ADBG(x) printk x
102 #else
103 #define ADBG(x)
104 #endif
105
106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
107
108 static inline u32 cstamp_delta(unsigned long cstamp)
109 {
110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 }
112
113 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
114 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
115 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
116
117 #ifdef CONFIG_SYSCTL
118 static void addrconf_sysctl_register(struct inet6_dev *idev);
119 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
120 #else
121 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
122 {
123 }
124
125 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
126 {
127 }
128 #endif
129
130 #ifdef CONFIG_IPV6_PRIVACY
131 static void __ipv6_regen_rndid(struct inet6_dev *idev);
132 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
133 static void ipv6_regen_rndid(unsigned long data);
134 #endif
135
136 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
137 static int ipv6_count_addresses(struct inet6_dev *idev);
138
139 /*
140  *      Configured unicast address hash table
141  */
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
144
145 static void addrconf_verify(unsigned long);
146
147 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
148 static DEFINE_SPINLOCK(addrconf_verify_lock);
149
150 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
151 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
152
153 static void addrconf_type_change(struct net_device *dev,
154                                  unsigned long event);
155 static int addrconf_ifdown(struct net_device *dev, int how);
156
157 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_timer(unsigned long data);
159 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_run(struct inet6_dev *idev);
161 static void addrconf_rs_timer(unsigned long data);
162 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164
165 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
166                                 struct prefix_info *pinfo);
167 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
168                                struct net_device *dev);
169
170 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
171
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
173         .forwarding             = 0,
174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
175         .mtu6                   = IPV6_MIN_MTU,
176         .accept_ra              = 1,
177         .accept_redirects       = 1,
178         .autoconf               = 1,
179         .force_mld_version      = 0,
180         .dad_transmits          = 1,
181         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
182         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
183         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
184 #ifdef CONFIG_IPV6_PRIVACY
185         .use_tempaddr           = 0,
186         .temp_valid_lft         = TEMP_VALID_LIFETIME,
187         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
188         .regen_max_retry        = REGEN_MAX_RETRY,
189         .max_desync_factor      = MAX_DESYNC_FACTOR,
190 #endif
191         .max_addresses          = IPV6_MAX_ADDRESSES,
192         .accept_ra_defrtr       = 1,
193         .accept_ra_pinfo        = 1,
194 #ifdef CONFIG_IPV6_ROUTER_PREF
195         .accept_ra_rtr_pref     = 1,
196         .rtr_probe_interval     = 60 * HZ,
197 #ifdef CONFIG_IPV6_ROUTE_INFO
198         .accept_ra_rt_info_max_plen = 0,
199 #endif
200 #endif
201         .proxy_ndp              = 0,
202         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
203         .disable_ipv6           = 0,
204         .accept_dad             = 1,
205 };
206
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
208         .forwarding             = 0,
209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
210         .mtu6                   = IPV6_MIN_MTU,
211         .accept_ra              = 1,
212         .accept_redirects       = 1,
213         .autoconf               = 1,
214         .dad_transmits          = 1,
215         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
216         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
217         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
218 #ifdef CONFIG_IPV6_PRIVACY
219         .use_tempaddr           = 0,
220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
222         .regen_max_retry        = REGEN_MAX_RETRY,
223         .max_desync_factor      = MAX_DESYNC_FACTOR,
224 #endif
225         .max_addresses          = IPV6_MAX_ADDRESSES,
226         .accept_ra_defrtr       = 1,
227         .accept_ra_pinfo        = 1,
228 #ifdef CONFIG_IPV6_ROUTER_PREF
229         .accept_ra_rtr_pref     = 1,
230         .rtr_probe_interval     = 60 * HZ,
231 #ifdef CONFIG_IPV6_ROUTE_INFO
232         .accept_ra_rt_info_max_plen = 0,
233 #endif
234 #endif
235         .proxy_ndp              = 0,
236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
237         .disable_ipv6           = 0,
238         .accept_dad             = 1,
239 };
240
241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
246
247 /* Check if a valid qdisc is available */
248 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
249 {
250         return !qdisc_tx_is_noop(dev);
251 }
252
253 /* Check if a route is valid prefix route */
254 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
255 {
256         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
257 }
258
259 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
260 {
261         if (del_timer(&ifp->timer))
262                 __in6_ifa_put(ifp);
263 }
264
265 enum addrconf_timer_t {
266         AC_NONE,
267         AC_DAD,
268         AC_RS,
269 };
270
271 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
272                                enum addrconf_timer_t what,
273                                unsigned long when)
274 {
275         if (!del_timer(&ifp->timer))
276                 in6_ifa_hold(ifp);
277
278         switch (what) {
279         case AC_DAD:
280                 ifp->timer.function = addrconf_dad_timer;
281                 break;
282         case AC_RS:
283                 ifp->timer.function = addrconf_rs_timer;
284                 break;
285         default:
286                 break;
287         }
288         ifp->timer.expires = jiffies + when;
289         add_timer(&ifp->timer);
290 }
291
292 static int snmp6_alloc_dev(struct inet6_dev *idev)
293 {
294         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
295                           sizeof(struct ipstats_mib),
296                           __alignof__(struct ipstats_mib)) < 0)
297                 goto err_ip;
298         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
299                                         GFP_KERNEL);
300         if (!idev->stats.icmpv6dev)
301                 goto err_icmp;
302         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
303                                            GFP_KERNEL);
304         if (!idev->stats.icmpv6msgdev)
305                 goto err_icmpmsg;
306
307         return 0;
308
309 err_icmpmsg:
310         kfree(idev->stats.icmpv6dev);
311 err_icmp:
312         snmp_mib_free((void __percpu **)idev->stats.ipv6);
313 err_ip:
314         return -ENOMEM;
315 }
316
317 static void snmp6_free_dev(struct inet6_dev *idev)
318 {
319         kfree(idev->stats.icmpv6msgdev);
320         kfree(idev->stats.icmpv6dev);
321         snmp_mib_free((void __percpu **)idev->stats.ipv6);
322 }
323
324 /* Nobody refers to this device, we may destroy it. */
325
326 void in6_dev_finish_destroy(struct inet6_dev *idev)
327 {
328         struct net_device *dev = idev->dev;
329
330         WARN_ON(!list_empty(&idev->addr_list));
331         WARN_ON(idev->mc_list != NULL);
332
333 #ifdef NET_REFCNT_DEBUG
334         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
335 #endif
336         dev_put(dev);
337         if (!idev->dead) {
338                 pr_warn("Freeing alive inet6 device %p\n", idev);
339                 return;
340         }
341         snmp6_free_dev(idev);
342         kfree_rcu(idev, rcu);
343 }
344 EXPORT_SYMBOL(in6_dev_finish_destroy);
345
346 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
347 {
348         struct inet6_dev *ndev;
349
350         ASSERT_RTNL();
351
352         if (dev->mtu < IPV6_MIN_MTU)
353                 return NULL;
354
355         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
356
357         if (ndev == NULL)
358                 return NULL;
359
360         rwlock_init(&ndev->lock);
361         ndev->dev = dev;
362         INIT_LIST_HEAD(&ndev->addr_list);
363
364         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
365         ndev->cnf.mtu6 = dev->mtu;
366         ndev->cnf.sysctl = NULL;
367         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
368         if (ndev->nd_parms == NULL) {
369                 kfree(ndev);
370                 return NULL;
371         }
372         if (ndev->cnf.forwarding)
373                 dev_disable_lro(dev);
374         /* We refer to the device */
375         dev_hold(dev);
376
377         if (snmp6_alloc_dev(ndev) < 0) {
378                 ADBG((KERN_WARNING
379                         "%s: cannot allocate memory for statistics; dev=%s.\n",
380                         __func__, dev->name));
381                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
382                 dev_put(dev);
383                 kfree(ndev);
384                 return NULL;
385         }
386
387         if (snmp6_register_dev(ndev) < 0) {
388                 ADBG((KERN_WARNING
389                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
390                         __func__, dev->name));
391                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
392                 ndev->dead = 1;
393                 in6_dev_finish_destroy(ndev);
394                 return NULL;
395         }
396
397         /* One reference from device.  We must do this before
398          * we invoke __ipv6_regen_rndid().
399          */
400         in6_dev_hold(ndev);
401
402         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
403                 ndev->cnf.accept_dad = -1;
404
405 #if IS_ENABLED(CONFIG_IPV6_SIT)
406         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
407                 pr_info("%s: Disabled Multicast RS\n", dev->name);
408                 ndev->cnf.rtr_solicits = 0;
409         }
410 #endif
411
412 #ifdef CONFIG_IPV6_PRIVACY
413         INIT_LIST_HEAD(&ndev->tempaddr_list);
414         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
415         if ((dev->flags&IFF_LOOPBACK) ||
416             dev->type == ARPHRD_TUNNEL ||
417             dev->type == ARPHRD_TUNNEL6 ||
418             dev->type == ARPHRD_SIT ||
419             dev->type == ARPHRD_NONE) {
420                 ndev->cnf.use_tempaddr = -1;
421         } else {
422                 in6_dev_hold(ndev);
423                 ipv6_regen_rndid((unsigned long) ndev);
424         }
425 #endif
426
427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
428                 ndev->if_flags |= IF_READY;
429
430         ipv6_mc_init_dev(ndev);
431         ndev->tstamp = jiffies;
432         addrconf_sysctl_register(ndev);
433         /* protected by rtnl_lock */
434         rcu_assign_pointer(dev->ip6_ptr, ndev);
435
436         /* Join all-node multicast group */
437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
438
439         /* Join all-router multicast group if forwarding is set */
440         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442
443         return ndev;
444 }
445
446 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
447 {
448         struct inet6_dev *idev;
449
450         ASSERT_RTNL();
451
452         idev = __in6_dev_get(dev);
453         if (!idev) {
454                 idev = ipv6_add_dev(dev);
455                 if (!idev)
456                         return NULL;
457         }
458
459         if (dev->flags&IFF_UP)
460                 ipv6_mc_up(idev);
461         return idev;
462 }
463
464 static int inet6_netconf_msgsize_devconf(int type)
465 {
466         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
467                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
468
469         /* type -1 is used for ALL */
470         if (type == -1 || type == NETCONFA_FORWARDING)
471                 size += nla_total_size(4);
472
473         return size;
474 }
475
476 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
477                                       struct ipv6_devconf *devconf, u32 portid,
478                                       u32 seq, int event, unsigned int flags,
479                                       int type)
480 {
481         struct nlmsghdr  *nlh;
482         struct netconfmsg *ncm;
483
484         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
485                         flags);
486         if (nlh == NULL)
487                 return -EMSGSIZE;
488
489         ncm = nlmsg_data(nlh);
490         ncm->ncm_family = AF_INET6;
491
492         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
493                 goto nla_put_failure;
494
495         /* type -1 is used for ALL */
496         if ((type == -1 || type == NETCONFA_FORWARDING) &&
497             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
498                 goto nla_put_failure;
499
500         return nlmsg_end(skb, nlh);
501
502 nla_put_failure:
503         nlmsg_cancel(skb, nlh);
504         return -EMSGSIZE;
505 }
506
507 static void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
508                                          struct ipv6_devconf *devconf)
509 {
510         struct sk_buff *skb;
511         int err = -ENOBUFS;
512
513         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
514         if (skb == NULL)
515                 goto errout;
516
517         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
518                                          RTM_NEWNETCONF, 0, type);
519         if (err < 0) {
520                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
521                 WARN_ON(err == -EMSGSIZE);
522                 kfree_skb(skb);
523                 goto errout;
524         }
525         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
526         return;
527 errout:
528         if (err < 0)
529                 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
530 }
531
532 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
533         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
534         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
535 };
536
537 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
538                                      struct nlmsghdr *nlh,
539                                      void *arg)
540 {
541         struct net *net = sock_net(in_skb->sk);
542         struct nlattr *tb[NETCONFA_MAX+1];
543         struct netconfmsg *ncm;
544         struct sk_buff *skb;
545         struct ipv6_devconf *devconf;
546         struct inet6_dev *in6_dev;
547         struct net_device *dev;
548         int ifindex;
549         int err;
550
551         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
552                           devconf_ipv6_policy);
553         if (err < 0)
554                 goto errout;
555
556         err = EINVAL;
557         if (!tb[NETCONFA_IFINDEX])
558                 goto errout;
559
560         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
561         switch (ifindex) {
562         case NETCONFA_IFINDEX_ALL:
563                 devconf = net->ipv6.devconf_all;
564                 break;
565         case NETCONFA_IFINDEX_DEFAULT:
566                 devconf = net->ipv6.devconf_dflt;
567                 break;
568         default:
569                 dev = __dev_get_by_index(net, ifindex);
570                 if (dev == NULL)
571                         goto errout;
572                 in6_dev = __in6_dev_get(dev);
573                 if (in6_dev == NULL)
574                         goto errout;
575                 devconf = &in6_dev->cnf;
576                 break;
577         }
578
579         err = -ENOBUFS;
580         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
581         if (skb == NULL)
582                 goto errout;
583
584         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
585                                          NETLINK_CB(in_skb).portid,
586                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
587                                          -1);
588         if (err < 0) {
589                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
590                 WARN_ON(err == -EMSGSIZE);
591                 kfree_skb(skb);
592                 goto errout;
593         }
594         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
595 errout:
596         return err;
597 }
598
599 #ifdef CONFIG_SYSCTL
600 static void dev_forward_change(struct inet6_dev *idev)
601 {
602         struct net_device *dev;
603         struct inet6_ifaddr *ifa;
604
605         if (!idev)
606                 return;
607         dev = idev->dev;
608         if (idev->cnf.forwarding)
609                 dev_disable_lro(dev);
610         if (dev->flags & IFF_MULTICAST) {
611                 if (idev->cnf.forwarding)
612                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
613                 else
614                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
615         }
616
617         list_for_each_entry(ifa, &idev->addr_list, if_list) {
618                 if (ifa->flags&IFA_F_TENTATIVE)
619                         continue;
620                 if (idev->cnf.forwarding)
621                         addrconf_join_anycast(ifa);
622                 else
623                         addrconf_leave_anycast(ifa);
624         }
625         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
626                                      dev->ifindex, &idev->cnf);
627 }
628
629
630 static void addrconf_forward_change(struct net *net, __s32 newf)
631 {
632         struct net_device *dev;
633         struct inet6_dev *idev;
634
635         for_each_netdev(net, dev) {
636                 idev = __in6_dev_get(dev);
637                 if (idev) {
638                         int changed = (!idev->cnf.forwarding) ^ (!newf);
639                         idev->cnf.forwarding = newf;
640                         if (changed)
641                                 dev_forward_change(idev);
642                 }
643         }
644 }
645
646 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
647 {
648         struct net *net;
649         int old;
650
651         if (!rtnl_trylock())
652                 return restart_syscall();
653
654         net = (struct net *)table->extra2;
655         old = *p;
656         *p = newf;
657
658         if (p == &net->ipv6.devconf_dflt->forwarding) {
659                 if ((!newf) ^ (!old))
660                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
661                                                      NETCONFA_IFINDEX_DEFAULT,
662                                                      net->ipv6.devconf_dflt);
663                 rtnl_unlock();
664                 return 0;
665         }
666
667         if (p == &net->ipv6.devconf_all->forwarding) {
668                 net->ipv6.devconf_dflt->forwarding = newf;
669                 addrconf_forward_change(net, newf);
670                 if ((!newf) ^ (!old))
671                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
672                                                      NETCONFA_IFINDEX_ALL,
673                                                      net->ipv6.devconf_all);
674         } else if ((!newf) ^ (!old))
675                 dev_forward_change((struct inet6_dev *)table->extra1);
676         rtnl_unlock();
677
678         if (newf)
679                 rt6_purge_dflt_routers(net);
680         return 1;
681 }
682 #endif
683
684 /* Nobody refers to this ifaddr, destroy it */
685 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
686 {
687         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
688
689 #ifdef NET_REFCNT_DEBUG
690         pr_debug("%s\n", __func__);
691 #endif
692
693         in6_dev_put(ifp->idev);
694
695         if (del_timer(&ifp->timer))
696                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
697
698         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
699                 pr_warn("Freeing alive inet6 address %p\n", ifp);
700                 return;
701         }
702         ip6_rt_put(ifp->rt);
703
704         kfree_rcu(ifp, rcu);
705 }
706
707 static void
708 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
709 {
710         struct list_head *p;
711         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
712
713         /*
714          * Each device address list is sorted in order of scope -
715          * global before linklocal.
716          */
717         list_for_each(p, &idev->addr_list) {
718                 struct inet6_ifaddr *ifa
719                         = list_entry(p, struct inet6_ifaddr, if_list);
720                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
721                         break;
722         }
723
724         list_add_tail(&ifp->if_list, p);
725 }
726
727 static u32 inet6_addr_hash(const struct in6_addr *addr)
728 {
729         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
730 }
731
732 /* On success it returns ifp with increased reference count */
733
734 static struct inet6_ifaddr *
735 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
736               int scope, u32 flags)
737 {
738         struct inet6_ifaddr *ifa = NULL;
739         struct rt6_info *rt;
740         unsigned int hash;
741         int err = 0;
742         int addr_type = ipv6_addr_type(addr);
743
744         if (addr_type == IPV6_ADDR_ANY ||
745             addr_type & IPV6_ADDR_MULTICAST ||
746             (!(idev->dev->flags & IFF_LOOPBACK) &&
747              addr_type & IPV6_ADDR_LOOPBACK))
748                 return ERR_PTR(-EADDRNOTAVAIL);
749
750         rcu_read_lock_bh();
751         if (idev->dead) {
752                 err = -ENODEV;                  /*XXX*/
753                 goto out2;
754         }
755
756         if (idev->cnf.disable_ipv6) {
757                 err = -EACCES;
758                 goto out2;
759         }
760
761         spin_lock(&addrconf_hash_lock);
762
763         /* Ignore adding duplicate addresses on an interface */
764         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
765                 ADBG(("ipv6_add_addr: already assigned\n"));
766                 err = -EEXIST;
767                 goto out;
768         }
769
770         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
771
772         if (ifa == NULL) {
773                 ADBG(("ipv6_add_addr: malloc failed\n"));
774                 err = -ENOBUFS;
775                 goto out;
776         }
777
778         rt = addrconf_dst_alloc(idev, addr, false);
779         if (IS_ERR(rt)) {
780                 err = PTR_ERR(rt);
781                 goto out;
782         }
783
784         ifa->addr = *addr;
785
786         spin_lock_init(&ifa->lock);
787         spin_lock_init(&ifa->state_lock);
788         init_timer(&ifa->timer);
789         INIT_HLIST_NODE(&ifa->addr_lst);
790         ifa->timer.data = (unsigned long) ifa;
791         ifa->scope = scope;
792         ifa->prefix_len = pfxlen;
793         ifa->flags = flags | IFA_F_TENTATIVE;
794         ifa->cstamp = ifa->tstamp = jiffies;
795
796         ifa->rt = rt;
797
798         ifa->idev = idev;
799         in6_dev_hold(idev);
800         /* For caller */
801         in6_ifa_hold(ifa);
802
803         /* Add to big hash table */
804         hash = inet6_addr_hash(addr);
805
806         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
807         spin_unlock(&addrconf_hash_lock);
808
809         write_lock(&idev->lock);
810         /* Add to inet6_dev unicast addr list. */
811         ipv6_link_dev_addr(idev, ifa);
812
813 #ifdef CONFIG_IPV6_PRIVACY
814         if (ifa->flags&IFA_F_TEMPORARY) {
815                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
816                 in6_ifa_hold(ifa);
817         }
818 #endif
819
820         in6_ifa_hold(ifa);
821         write_unlock(&idev->lock);
822 out2:
823         rcu_read_unlock_bh();
824
825         if (likely(err == 0))
826                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
827         else {
828                 kfree(ifa);
829                 ifa = ERR_PTR(err);
830         }
831
832         return ifa;
833 out:
834         spin_unlock(&addrconf_hash_lock);
835         goto out2;
836 }
837
838 /* This function wants to get referenced ifp and releases it before return */
839
840 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
841 {
842         struct inet6_ifaddr *ifa, *ifn;
843         struct inet6_dev *idev = ifp->idev;
844         int state;
845         int deleted = 0, onlink = 0;
846         unsigned long expires = jiffies;
847
848         spin_lock_bh(&ifp->state_lock);
849         state = ifp->state;
850         ifp->state = INET6_IFADDR_STATE_DEAD;
851         spin_unlock_bh(&ifp->state_lock);
852
853         if (state == INET6_IFADDR_STATE_DEAD)
854                 goto out;
855
856         spin_lock_bh(&addrconf_hash_lock);
857         hlist_del_init_rcu(&ifp->addr_lst);
858         spin_unlock_bh(&addrconf_hash_lock);
859
860         write_lock_bh(&idev->lock);
861 #ifdef CONFIG_IPV6_PRIVACY
862         if (ifp->flags&IFA_F_TEMPORARY) {
863                 list_del(&ifp->tmp_list);
864                 if (ifp->ifpub) {
865                         in6_ifa_put(ifp->ifpub);
866                         ifp->ifpub = NULL;
867                 }
868                 __in6_ifa_put(ifp);
869         }
870 #endif
871
872         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
873                 if (ifa == ifp) {
874                         list_del_init(&ifp->if_list);
875                         __in6_ifa_put(ifp);
876
877                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
878                                 break;
879                         deleted = 1;
880                         continue;
881                 } else if (ifp->flags & IFA_F_PERMANENT) {
882                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
883                                               ifp->prefix_len)) {
884                                 if (ifa->flags & IFA_F_PERMANENT) {
885                                         onlink = 1;
886                                         if (deleted)
887                                                 break;
888                                 } else {
889                                         unsigned long lifetime;
890
891                                         if (!onlink)
892                                                 onlink = -1;
893
894                                         spin_lock(&ifa->lock);
895
896                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
897                                         /*
898                                          * Note: Because this address is
899                                          * not permanent, lifetime <
900                                          * LONG_MAX / HZ here.
901                                          */
902                                         if (time_before(expires,
903                                                         ifa->tstamp + lifetime * HZ))
904                                                 expires = ifa->tstamp + lifetime * HZ;
905                                         spin_unlock(&ifa->lock);
906                                 }
907                         }
908                 }
909         }
910         write_unlock_bh(&idev->lock);
911
912         addrconf_del_timer(ifp);
913
914         ipv6_ifa_notify(RTM_DELADDR, ifp);
915
916         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
917
918         /*
919          * Purge or update corresponding prefix
920          *
921          * 1) we don't purge prefix here if address was not permanent.
922          *    prefix is managed by its own lifetime.
923          * 2) if there're no addresses, delete prefix.
924          * 3) if there're still other permanent address(es),
925          *    corresponding prefix is still permanent.
926          * 4) otherwise, update prefix lifetime to the
927          *    longest valid lifetime among the corresponding
928          *    addresses on the device.
929          *    Note: subsequent RA will update lifetime.
930          *
931          * --yoshfuji
932          */
933         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
934                 struct in6_addr prefix;
935                 struct rt6_info *rt;
936                 struct net *net = dev_net(ifp->idev->dev);
937                 struct flowi6 fl6 = {};
938
939                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
940                 fl6.flowi6_oif = ifp->idev->dev->ifindex;
941                 fl6.daddr = prefix;
942                 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
943                                                          RT6_LOOKUP_F_IFACE);
944
945                 if (rt != net->ipv6.ip6_null_entry &&
946                     addrconf_is_prefix_route(rt)) {
947                         if (onlink == 0) {
948                                 ip6_del_rt(rt);
949                                 rt = NULL;
950                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
951                                 rt6_set_expires(rt, expires);
952                         }
953                 }
954                 ip6_rt_put(rt);
955         }
956
957         /* clean up prefsrc entries */
958         rt6_remove_prefsrc(ifp);
959 out:
960         in6_ifa_put(ifp);
961 }
962
963 #ifdef CONFIG_IPV6_PRIVACY
964 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
965 {
966         struct inet6_dev *idev = ifp->idev;
967         struct in6_addr addr, *tmpaddr;
968         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
969         unsigned long regen_advance;
970         int tmp_plen;
971         int ret = 0;
972         int max_addresses;
973         u32 addr_flags;
974         unsigned long now = jiffies;
975
976         write_lock(&idev->lock);
977         if (ift) {
978                 spin_lock_bh(&ift->lock);
979                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
980                 spin_unlock_bh(&ift->lock);
981                 tmpaddr = &addr;
982         } else {
983                 tmpaddr = NULL;
984         }
985 retry:
986         in6_dev_hold(idev);
987         if (idev->cnf.use_tempaddr <= 0) {
988                 write_unlock(&idev->lock);
989                 pr_info("%s: use_tempaddr is disabled\n", __func__);
990                 in6_dev_put(idev);
991                 ret = -1;
992                 goto out;
993         }
994         spin_lock_bh(&ifp->lock);
995         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
996                 idev->cnf.use_tempaddr = -1;    /*XXX*/
997                 spin_unlock_bh(&ifp->lock);
998                 write_unlock(&idev->lock);
999                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1000                         __func__);
1001                 in6_dev_put(idev);
1002                 ret = -1;
1003                 goto out;
1004         }
1005         in6_ifa_hold(ifp);
1006         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1007         __ipv6_try_regen_rndid(idev, tmpaddr);
1008         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1009         age = (now - ifp->tstamp) / HZ;
1010         tmp_valid_lft = min_t(__u32,
1011                               ifp->valid_lft,
1012                               idev->cnf.temp_valid_lft + age);
1013         tmp_prefered_lft = min_t(__u32,
1014                                  ifp->prefered_lft,
1015                                  idev->cnf.temp_prefered_lft + age -
1016                                  idev->cnf.max_desync_factor);
1017         tmp_plen = ifp->prefix_len;
1018         max_addresses = idev->cnf.max_addresses;
1019         tmp_tstamp = ifp->tstamp;
1020         spin_unlock_bh(&ifp->lock);
1021
1022         regen_advance = idev->cnf.regen_max_retry *
1023                         idev->cnf.dad_transmits *
1024                         idev->nd_parms->retrans_time / HZ;
1025         write_unlock(&idev->lock);
1026
1027         /* A temporary address is created only if this calculated Preferred
1028          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1029          * an implementation must not create a temporary address with a zero
1030          * Preferred Lifetime.
1031          */
1032         if (tmp_prefered_lft <= regen_advance) {
1033                 in6_ifa_put(ifp);
1034                 in6_dev_put(idev);
1035                 ret = -1;
1036                 goto out;
1037         }
1038
1039         addr_flags = IFA_F_TEMPORARY;
1040         /* set in addrconf_prefix_rcv() */
1041         if (ifp->flags & IFA_F_OPTIMISTIC)
1042                 addr_flags |= IFA_F_OPTIMISTIC;
1043
1044         ift = !max_addresses ||
1045               ipv6_count_addresses(idev) < max_addresses ?
1046                 ipv6_add_addr(idev, &addr, tmp_plen,
1047                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1048                               addr_flags) : NULL;
1049         if (!ift || IS_ERR(ift)) {
1050                 in6_ifa_put(ifp);
1051                 in6_dev_put(idev);
1052                 pr_info("%s: retry temporary address regeneration\n", __func__);
1053                 tmpaddr = &addr;
1054                 write_lock(&idev->lock);
1055                 goto retry;
1056         }
1057
1058         spin_lock_bh(&ift->lock);
1059         ift->ifpub = ifp;
1060         ift->valid_lft = tmp_valid_lft;
1061         ift->prefered_lft = tmp_prefered_lft;
1062         ift->cstamp = now;
1063         ift->tstamp = tmp_tstamp;
1064         spin_unlock_bh(&ift->lock);
1065
1066         addrconf_dad_start(ift);
1067         in6_ifa_put(ift);
1068         in6_dev_put(idev);
1069 out:
1070         return ret;
1071 }
1072 #endif
1073
1074 /*
1075  *      Choose an appropriate source address (RFC3484)
1076  */
1077 enum {
1078         IPV6_SADDR_RULE_INIT = 0,
1079         IPV6_SADDR_RULE_LOCAL,
1080         IPV6_SADDR_RULE_SCOPE,
1081         IPV6_SADDR_RULE_PREFERRED,
1082 #ifdef CONFIG_IPV6_MIP6
1083         IPV6_SADDR_RULE_HOA,
1084 #endif
1085         IPV6_SADDR_RULE_OIF,
1086         IPV6_SADDR_RULE_LABEL,
1087 #ifdef CONFIG_IPV6_PRIVACY
1088         IPV6_SADDR_RULE_PRIVACY,
1089 #endif
1090         IPV6_SADDR_RULE_ORCHID,
1091         IPV6_SADDR_RULE_PREFIX,
1092         IPV6_SADDR_RULE_MAX
1093 };
1094
1095 struct ipv6_saddr_score {
1096         int                     rule;
1097         int                     addr_type;
1098         struct inet6_ifaddr     *ifa;
1099         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1100         int                     scopedist;
1101         int                     matchlen;
1102 };
1103
1104 struct ipv6_saddr_dst {
1105         const struct in6_addr *addr;
1106         int ifindex;
1107         int scope;
1108         int label;
1109         unsigned int prefs;
1110 };
1111
1112 static inline int ipv6_saddr_preferred(int type)
1113 {
1114         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1115                 return 1;
1116         return 0;
1117 }
1118
1119 static int ipv6_get_saddr_eval(struct net *net,
1120                                struct ipv6_saddr_score *score,
1121                                struct ipv6_saddr_dst *dst,
1122                                int i)
1123 {
1124         int ret;
1125
1126         if (i <= score->rule) {
1127                 switch (i) {
1128                 case IPV6_SADDR_RULE_SCOPE:
1129                         ret = score->scopedist;
1130                         break;
1131                 case IPV6_SADDR_RULE_PREFIX:
1132                         ret = score->matchlen;
1133                         break;
1134                 default:
1135                         ret = !!test_bit(i, score->scorebits);
1136                 }
1137                 goto out;
1138         }
1139
1140         switch (i) {
1141         case IPV6_SADDR_RULE_INIT:
1142                 /* Rule 0: remember if hiscore is not ready yet */
1143                 ret = !!score->ifa;
1144                 break;
1145         case IPV6_SADDR_RULE_LOCAL:
1146                 /* Rule 1: Prefer same address */
1147                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1148                 break;
1149         case IPV6_SADDR_RULE_SCOPE:
1150                 /* Rule 2: Prefer appropriate scope
1151                  *
1152                  *      ret
1153                  *       ^
1154                  *    -1 |  d 15
1155                  *    ---+--+-+---> scope
1156                  *       |
1157                  *       |             d is scope of the destination.
1158                  *  B-d  |  \
1159                  *       |   \      <- smaller scope is better if
1160                  *  B-15 |    \        if scope is enough for destinaion.
1161                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1162                  * d-C-1 | /
1163                  *       |/         <- greater is better
1164                  *   -C  /             if scope is not enough for destination.
1165                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1166                  *
1167                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1168                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1169                  * Assume B = 0 and we get C > 29.
1170                  */
1171                 ret = __ipv6_addr_src_scope(score->addr_type);
1172                 if (ret >= dst->scope)
1173                         ret = -ret;
1174                 else
1175                         ret -= 128;     /* 30 is enough */
1176                 score->scopedist = ret;
1177                 break;
1178         case IPV6_SADDR_RULE_PREFERRED:
1179                 /* Rule 3: Avoid deprecated and optimistic addresses */
1180                 ret = ipv6_saddr_preferred(score->addr_type) ||
1181                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1182                 break;
1183 #ifdef CONFIG_IPV6_MIP6
1184         case IPV6_SADDR_RULE_HOA:
1185             {
1186                 /* Rule 4: Prefer home address */
1187                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1188                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1189                 break;
1190             }
1191 #endif
1192         case IPV6_SADDR_RULE_OIF:
1193                 /* Rule 5: Prefer outgoing interface */
1194                 ret = (!dst->ifindex ||
1195                        dst->ifindex == score->ifa->idev->dev->ifindex);
1196                 break;
1197         case IPV6_SADDR_RULE_LABEL:
1198                 /* Rule 6: Prefer matching label */
1199                 ret = ipv6_addr_label(net,
1200                                       &score->ifa->addr, score->addr_type,
1201                                       score->ifa->idev->dev->ifindex) == dst->label;
1202                 break;
1203 #ifdef CONFIG_IPV6_PRIVACY
1204         case IPV6_SADDR_RULE_PRIVACY:
1205             {
1206                 /* Rule 7: Prefer public address
1207                  * Note: prefer temporary address if use_tempaddr >= 2
1208                  */
1209                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1210                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1211                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1212                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1213                 break;
1214             }
1215 #endif
1216         case IPV6_SADDR_RULE_ORCHID:
1217                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1218                  *          non-ORCHID vs non-ORCHID
1219                  */
1220                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1221                         ipv6_addr_orchid(dst->addr));
1222                 break;
1223         case IPV6_SADDR_RULE_PREFIX:
1224                 /* Rule 8: Use longest matching prefix */
1225                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1226                 if (ret > score->ifa->prefix_len)
1227                         ret = score->ifa->prefix_len;
1228                 score->matchlen = ret;
1229                 break;
1230         default:
1231                 ret = 0;
1232         }
1233
1234         if (ret)
1235                 __set_bit(i, score->scorebits);
1236         score->rule = i;
1237 out:
1238         return ret;
1239 }
1240
1241 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1242                        const struct in6_addr *daddr, unsigned int prefs,
1243                        struct in6_addr *saddr)
1244 {
1245         struct ipv6_saddr_score scores[2],
1246                                 *score = &scores[0], *hiscore = &scores[1];
1247         struct ipv6_saddr_dst dst;
1248         struct net_device *dev;
1249         int dst_type;
1250
1251         dst_type = __ipv6_addr_type(daddr);
1252         dst.addr = daddr;
1253         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1254         dst.scope = __ipv6_addr_src_scope(dst_type);
1255         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1256         dst.prefs = prefs;
1257
1258         hiscore->rule = -1;
1259         hiscore->ifa = NULL;
1260
1261         rcu_read_lock();
1262
1263         for_each_netdev_rcu(net, dev) {
1264                 struct inet6_dev *idev;
1265
1266                 /* Candidate Source Address (section 4)
1267                  *  - multicast and link-local destination address,
1268                  *    the set of candidate source address MUST only
1269                  *    include addresses assigned to interfaces
1270                  *    belonging to the same link as the outgoing
1271                  *    interface.
1272                  * (- For site-local destination addresses, the
1273                  *    set of candidate source addresses MUST only
1274                  *    include addresses assigned to interfaces
1275                  *    belonging to the same site as the outgoing
1276                  *    interface.)
1277                  */
1278                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1279                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1280                     dst.ifindex && dev->ifindex != dst.ifindex)
1281                         continue;
1282
1283                 idev = __in6_dev_get(dev);
1284                 if (!idev)
1285                         continue;
1286
1287                 read_lock_bh(&idev->lock);
1288                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1289                         int i;
1290
1291                         /*
1292                          * - Tentative Address (RFC2462 section 5.4)
1293                          *  - A tentative address is not considered
1294                          *    "assigned to an interface" in the traditional
1295                          *    sense, unless it is also flagged as optimistic.
1296                          * - Candidate Source Address (section 4)
1297                          *  - In any case, anycast addresses, multicast
1298                          *    addresses, and the unspecified address MUST
1299                          *    NOT be included in a candidate set.
1300                          */
1301                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1302                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1303                                 continue;
1304
1305                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1306
1307                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1308                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1309                                 LIMIT_NETDEBUG(KERN_DEBUG
1310                                                "ADDRCONF: unspecified / multicast address "
1311                                                "assigned as unicast address on %s",
1312                                                dev->name);
1313                                 continue;
1314                         }
1315
1316                         score->rule = -1;
1317                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1318
1319                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1320                                 int minihiscore, miniscore;
1321
1322                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1323                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1324
1325                                 if (minihiscore > miniscore) {
1326                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1327                                             score->scopedist > 0) {
1328                                                 /*
1329                                                  * special case:
1330                                                  * each remaining entry
1331                                                  * has too small (not enough)
1332                                                  * scope, because ifa entries
1333                                                  * are sorted by their scope
1334                                                  * values.
1335                                                  */
1336                                                 goto try_nextdev;
1337                                         }
1338                                         break;
1339                                 } else if (minihiscore < miniscore) {
1340                                         if (hiscore->ifa)
1341                                                 in6_ifa_put(hiscore->ifa);
1342
1343                                         in6_ifa_hold(score->ifa);
1344
1345                                         swap(hiscore, score);
1346
1347                                         /* restore our iterator */
1348                                         score->ifa = hiscore->ifa;
1349
1350                                         break;
1351                                 }
1352                         }
1353                 }
1354 try_nextdev:
1355                 read_unlock_bh(&idev->lock);
1356         }
1357         rcu_read_unlock();
1358
1359         if (!hiscore->ifa)
1360                 return -EADDRNOTAVAIL;
1361
1362         *saddr = hiscore->ifa->addr;
1363         in6_ifa_put(hiscore->ifa);
1364         return 0;
1365 }
1366 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1367
1368 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1369                     unsigned char banned_flags)
1370 {
1371         struct inet6_dev *idev;
1372         int err = -EADDRNOTAVAIL;
1373
1374         rcu_read_lock();
1375         idev = __in6_dev_get(dev);
1376         if (idev) {
1377                 struct inet6_ifaddr *ifp;
1378
1379                 read_lock_bh(&idev->lock);
1380                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1381                         if (ifp->scope == IFA_LINK &&
1382                             !(ifp->flags & banned_flags)) {
1383                                 *addr = ifp->addr;
1384                                 err = 0;
1385                                 break;
1386                         }
1387                 }
1388                 read_unlock_bh(&idev->lock);
1389         }
1390         rcu_read_unlock();
1391         return err;
1392 }
1393
1394 static int ipv6_count_addresses(struct inet6_dev *idev)
1395 {
1396         int cnt = 0;
1397         struct inet6_ifaddr *ifp;
1398
1399         read_lock_bh(&idev->lock);
1400         list_for_each_entry(ifp, &idev->addr_list, if_list)
1401                 cnt++;
1402         read_unlock_bh(&idev->lock);
1403         return cnt;
1404 }
1405
1406 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1407                   struct net_device *dev, int strict)
1408 {
1409         struct inet6_ifaddr *ifp;
1410         struct hlist_node *node;
1411         unsigned int hash = inet6_addr_hash(addr);
1412
1413         rcu_read_lock_bh();
1414         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1415                 if (!net_eq(dev_net(ifp->idev->dev), net))
1416                         continue;
1417                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1418                     !(ifp->flags&IFA_F_TENTATIVE) &&
1419                     (dev == NULL || ifp->idev->dev == dev ||
1420                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1421                         rcu_read_unlock_bh();
1422                         return 1;
1423                 }
1424         }
1425
1426         rcu_read_unlock_bh();
1427         return 0;
1428 }
1429 EXPORT_SYMBOL(ipv6_chk_addr);
1430
1431 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1432                                struct net_device *dev)
1433 {
1434         unsigned int hash = inet6_addr_hash(addr);
1435         struct inet6_ifaddr *ifp;
1436         struct hlist_node *node;
1437
1438         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1439                 if (!net_eq(dev_net(ifp->idev->dev), net))
1440                         continue;
1441                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1442                         if (dev == NULL || ifp->idev->dev == dev)
1443                                 return true;
1444                 }
1445         }
1446         return false;
1447 }
1448
1449 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1450 {
1451         struct inet6_dev *idev;
1452         struct inet6_ifaddr *ifa;
1453         int     onlink;
1454
1455         onlink = 0;
1456         rcu_read_lock();
1457         idev = __in6_dev_get(dev);
1458         if (idev) {
1459                 read_lock_bh(&idev->lock);
1460                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1461                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1462                                                    ifa->prefix_len);
1463                         if (onlink)
1464                                 break;
1465                 }
1466                 read_unlock_bh(&idev->lock);
1467         }
1468         rcu_read_unlock();
1469         return onlink;
1470 }
1471 EXPORT_SYMBOL(ipv6_chk_prefix);
1472
1473 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1474                                      struct net_device *dev, int strict)
1475 {
1476         struct inet6_ifaddr *ifp, *result = NULL;
1477         unsigned int hash = inet6_addr_hash(addr);
1478         struct hlist_node *node;
1479
1480         rcu_read_lock_bh();
1481         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1482                 if (!net_eq(dev_net(ifp->idev->dev), net))
1483                         continue;
1484                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1485                         if (dev == NULL || ifp->idev->dev == dev ||
1486                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1487                                 result = ifp;
1488                                 in6_ifa_hold(ifp);
1489                                 break;
1490                         }
1491                 }
1492         }
1493         rcu_read_unlock_bh();
1494
1495         return result;
1496 }
1497
1498 /* Gets referenced address, destroys ifaddr */
1499
1500 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1501 {
1502         if (ifp->flags&IFA_F_PERMANENT) {
1503                 spin_lock_bh(&ifp->lock);
1504                 addrconf_del_timer(ifp);
1505                 ifp->flags |= IFA_F_TENTATIVE;
1506                 if (dad_failed)
1507                         ifp->flags |= IFA_F_DADFAILED;
1508                 spin_unlock_bh(&ifp->lock);
1509                 if (dad_failed)
1510                         ipv6_ifa_notify(0, ifp);
1511                 in6_ifa_put(ifp);
1512 #ifdef CONFIG_IPV6_PRIVACY
1513         } else if (ifp->flags&IFA_F_TEMPORARY) {
1514                 struct inet6_ifaddr *ifpub;
1515                 spin_lock_bh(&ifp->lock);
1516                 ifpub = ifp->ifpub;
1517                 if (ifpub) {
1518                         in6_ifa_hold(ifpub);
1519                         spin_unlock_bh(&ifp->lock);
1520                         ipv6_create_tempaddr(ifpub, ifp);
1521                         in6_ifa_put(ifpub);
1522                 } else {
1523                         spin_unlock_bh(&ifp->lock);
1524                 }
1525                 ipv6_del_addr(ifp);
1526 #endif
1527         } else
1528                 ipv6_del_addr(ifp);
1529 }
1530
1531 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1532 {
1533         int err = -ENOENT;
1534
1535         spin_lock(&ifp->state_lock);
1536         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1537                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1538                 err = 0;
1539         }
1540         spin_unlock(&ifp->state_lock);
1541
1542         return err;
1543 }
1544
1545 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1546 {
1547         struct inet6_dev *idev = ifp->idev;
1548
1549         if (addrconf_dad_end(ifp)) {
1550                 in6_ifa_put(ifp);
1551                 return;
1552         }
1553
1554         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1555                              ifp->idev->dev->name, &ifp->addr);
1556
1557         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1558                 struct in6_addr addr;
1559
1560                 addr.s6_addr32[0] = htonl(0xfe800000);
1561                 addr.s6_addr32[1] = 0;
1562
1563                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1564                     ipv6_addr_equal(&ifp->addr, &addr)) {
1565                         /* DAD failed for link-local based on MAC address */
1566                         idev->cnf.disable_ipv6 = 1;
1567
1568                         pr_info("%s: IPv6 being disabled!\n",
1569                                 ifp->idev->dev->name);
1570                 }
1571         }
1572
1573         addrconf_dad_stop(ifp, 1);
1574 }
1575
1576 /* Join to solicited addr multicast group. */
1577
1578 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1579 {
1580         struct in6_addr maddr;
1581
1582         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1583                 return;
1584
1585         addrconf_addr_solict_mult(addr, &maddr);
1586         ipv6_dev_mc_inc(dev, &maddr);
1587 }
1588
1589 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1590 {
1591         struct in6_addr maddr;
1592
1593         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1594                 return;
1595
1596         addrconf_addr_solict_mult(addr, &maddr);
1597         __ipv6_dev_mc_dec(idev, &maddr);
1598 }
1599
1600 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1601 {
1602         struct in6_addr addr;
1603         if (ifp->prefix_len == 127) /* RFC 6164 */
1604                 return;
1605         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1606         if (ipv6_addr_any(&addr))
1607                 return;
1608         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1609 }
1610
1611 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1612 {
1613         struct in6_addr addr;
1614         if (ifp->prefix_len == 127) /* RFC 6164 */
1615                 return;
1616         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1617         if (ipv6_addr_any(&addr))
1618                 return;
1619         __ipv6_dev_ac_dec(ifp->idev, &addr);
1620 }
1621
1622 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1623 {
1624         if (dev->addr_len != ETH_ALEN)
1625                 return -1;
1626         memcpy(eui, dev->dev_addr, 3);
1627         memcpy(eui + 5, dev->dev_addr + 3, 3);
1628
1629         /*
1630          * The zSeries OSA network cards can be shared among various
1631          * OS instances, but the OSA cards have only one MAC address.
1632          * This leads to duplicate address conflicts in conjunction
1633          * with IPv6 if more than one instance uses the same card.
1634          *
1635          * The driver for these cards can deliver a unique 16-bit
1636          * identifier for each instance sharing the same card.  It is
1637          * placed instead of 0xFFFE in the interface identifier.  The
1638          * "u" bit of the interface identifier is not inverted in this
1639          * case.  Hence the resulting interface identifier has local
1640          * scope according to RFC2373.
1641          */
1642         if (dev->dev_id) {
1643                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1644                 eui[4] = dev->dev_id & 0xFF;
1645         } else {
1646                 eui[3] = 0xFF;
1647                 eui[4] = 0xFE;
1648                 eui[0] ^= 2;
1649         }
1650         return 0;
1651 }
1652
1653 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1654 {
1655         if (dev->addr_len != IEEE802154_ADDR_LEN)
1656                 return -1;
1657         memcpy(eui, dev->dev_addr, 8);
1658         return 0;
1659 }
1660
1661 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1662 {
1663         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1664         if (dev->addr_len != ARCNET_ALEN)
1665                 return -1;
1666         memset(eui, 0, 7);
1667         eui[7] = *(u8 *)dev->dev_addr;
1668         return 0;
1669 }
1670
1671 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1672 {
1673         if (dev->addr_len != INFINIBAND_ALEN)
1674                 return -1;
1675         memcpy(eui, dev->dev_addr + 12, 8);
1676         eui[0] |= 2;
1677         return 0;
1678 }
1679
1680 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1681 {
1682         if (addr == 0)
1683                 return -1;
1684         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1685                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1686                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1687                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1688                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1689                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1690         eui[1] = 0;
1691         eui[2] = 0x5E;
1692         eui[3] = 0xFE;
1693         memcpy(eui + 4, &addr, 4);
1694         return 0;
1695 }
1696
1697 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1698 {
1699         if (dev->priv_flags & IFF_ISATAP)
1700                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1701         return -1;
1702 }
1703
1704 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1705 {
1706         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1707 }
1708
1709 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1710 {
1711         switch (dev->type) {
1712         case ARPHRD_ETHER:
1713         case ARPHRD_FDDI:
1714                 return addrconf_ifid_eui48(eui, dev);
1715         case ARPHRD_ARCNET:
1716                 return addrconf_ifid_arcnet(eui, dev);
1717         case ARPHRD_INFINIBAND:
1718                 return addrconf_ifid_infiniband(eui, dev);
1719         case ARPHRD_SIT:
1720                 return addrconf_ifid_sit(eui, dev);
1721         case ARPHRD_IPGRE:
1722                 return addrconf_ifid_gre(eui, dev);
1723         case ARPHRD_IEEE802154:
1724                 return addrconf_ifid_eui64(eui, dev);
1725         }
1726         return -1;
1727 }
1728
1729 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1730 {
1731         int err = -1;
1732         struct inet6_ifaddr *ifp;
1733
1734         read_lock_bh(&idev->lock);
1735         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1736                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1737                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1738                         err = 0;
1739                         break;
1740                 }
1741         }
1742         read_unlock_bh(&idev->lock);
1743         return err;
1744 }
1745
1746 #ifdef CONFIG_IPV6_PRIVACY
1747 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1748 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1749 {
1750 regen:
1751         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1752         idev->rndid[0] &= ~0x02;
1753
1754         /*
1755          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1756          * check if generated address is not inappropriate
1757          *
1758          *  - Reserved subnet anycast (RFC 2526)
1759          *      11111101 11....11 1xxxxxxx
1760          *  - ISATAP (RFC4214) 6.1
1761          *      00-00-5E-FE-xx-xx-xx-xx
1762          *  - value 0
1763          *  - XXX: already assigned to an address on the device
1764          */
1765         if (idev->rndid[0] == 0xfd &&
1766             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1767             (idev->rndid[7]&0x80))
1768                 goto regen;
1769         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1770                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1771                         goto regen;
1772                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1773                         goto regen;
1774         }
1775 }
1776
1777 static void ipv6_regen_rndid(unsigned long data)
1778 {
1779         struct inet6_dev *idev = (struct inet6_dev *) data;
1780         unsigned long expires;
1781
1782         rcu_read_lock_bh();
1783         write_lock_bh(&idev->lock);
1784
1785         if (idev->dead)
1786                 goto out;
1787
1788         __ipv6_regen_rndid(idev);
1789
1790         expires = jiffies +
1791                 idev->cnf.temp_prefered_lft * HZ -
1792                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1793                 idev->cnf.max_desync_factor * HZ;
1794         if (time_before(expires, jiffies)) {
1795                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1796                         __func__, idev->dev->name);
1797                 goto out;
1798         }
1799
1800         if (!mod_timer(&idev->regen_timer, expires))
1801                 in6_dev_hold(idev);
1802
1803 out:
1804         write_unlock_bh(&idev->lock);
1805         rcu_read_unlock_bh();
1806         in6_dev_put(idev);
1807 }
1808
1809 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1810 {
1811         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1812                 __ipv6_regen_rndid(idev);
1813 }
1814 #endif
1815
1816 /*
1817  *      Add prefix route.
1818  */
1819
1820 static void
1821 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1822                       unsigned long expires, u32 flags)
1823 {
1824         struct fib6_config cfg = {
1825                 .fc_table = RT6_TABLE_PREFIX,
1826                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1827                 .fc_ifindex = dev->ifindex,
1828                 .fc_expires = expires,
1829                 .fc_dst_len = plen,
1830                 .fc_flags = RTF_UP | flags,
1831                 .fc_nlinfo.nl_net = dev_net(dev),
1832                 .fc_protocol = RTPROT_KERNEL,
1833         };
1834
1835         cfg.fc_dst = *pfx;
1836
1837         /* Prevent useless cloning on PtP SIT.
1838            This thing is done here expecting that the whole
1839            class of non-broadcast devices need not cloning.
1840          */
1841 #if IS_ENABLED(CONFIG_IPV6_SIT)
1842         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1843                 cfg.fc_flags |= RTF_NONEXTHOP;
1844 #endif
1845
1846         ip6_route_add(&cfg);
1847 }
1848
1849
1850 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1851                                                   int plen,
1852                                                   const struct net_device *dev,
1853                                                   u32 flags, u32 noflags)
1854 {
1855         struct fib6_node *fn;
1856         struct rt6_info *rt = NULL;
1857         struct fib6_table *table;
1858
1859         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1860         if (table == NULL)
1861                 return NULL;
1862
1863         read_lock_bh(&table->tb6_lock);
1864         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1865         if (!fn)
1866                 goto out;
1867         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1868                 if (rt->dst.dev->ifindex != dev->ifindex)
1869                         continue;
1870                 if ((rt->rt6i_flags & flags) != flags)
1871                         continue;
1872                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1873                         continue;
1874                 dst_hold(&rt->dst);
1875                 break;
1876         }
1877 out:
1878         read_unlock_bh(&table->tb6_lock);
1879         return rt;
1880 }
1881
1882
1883 /* Create "default" multicast route to the interface */
1884
1885 static void addrconf_add_mroute(struct net_device *dev)
1886 {
1887         struct fib6_config cfg = {
1888                 .fc_table = RT6_TABLE_LOCAL,
1889                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1890                 .fc_ifindex = dev->ifindex,
1891                 .fc_dst_len = 8,
1892                 .fc_flags = RTF_UP,
1893                 .fc_nlinfo.nl_net = dev_net(dev),
1894         };
1895
1896         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1897
1898         ip6_route_add(&cfg);
1899 }
1900
1901 #if IS_ENABLED(CONFIG_IPV6_SIT)
1902 static void sit_route_add(struct net_device *dev)
1903 {
1904         struct fib6_config cfg = {
1905                 .fc_table = RT6_TABLE_MAIN,
1906                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1907                 .fc_ifindex = dev->ifindex,
1908                 .fc_dst_len = 96,
1909                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1910                 .fc_nlinfo.nl_net = dev_net(dev),
1911         };
1912
1913         /* prefix length - 96 bits "::d.d.d.d" */
1914         ip6_route_add(&cfg);
1915 }
1916 #endif
1917
1918 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1919 {
1920         struct inet6_dev *idev;
1921
1922         ASSERT_RTNL();
1923
1924         idev = ipv6_find_idev(dev);
1925         if (!idev)
1926                 return ERR_PTR(-ENOBUFS);
1927
1928         if (idev->cnf.disable_ipv6)
1929                 return ERR_PTR(-EACCES);
1930
1931         /* Add default multicast route */
1932         if (!(dev->flags & IFF_LOOPBACK))
1933                 addrconf_add_mroute(dev);
1934
1935         return idev;
1936 }
1937
1938 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1939 {
1940         struct prefix_info *pinfo;
1941         __u32 valid_lft;
1942         __u32 prefered_lft;
1943         int addr_type;
1944         struct inet6_dev *in6_dev;
1945         struct net *net = dev_net(dev);
1946
1947         pinfo = (struct prefix_info *) opt;
1948
1949         if (len < sizeof(struct prefix_info)) {
1950                 ADBG(("addrconf: prefix option too short\n"));
1951                 return;
1952         }
1953
1954         /*
1955          *      Validation checks ([ADDRCONF], page 19)
1956          */
1957
1958         addr_type = ipv6_addr_type(&pinfo->prefix);
1959
1960         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1961                 return;
1962
1963         valid_lft = ntohl(pinfo->valid);
1964         prefered_lft = ntohl(pinfo->prefered);
1965
1966         if (prefered_lft > valid_lft) {
1967                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1968                 return;
1969         }
1970
1971         in6_dev = in6_dev_get(dev);
1972
1973         if (in6_dev == NULL) {
1974                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1975                                     dev->name);
1976                 return;
1977         }
1978
1979         /*
1980          *      Two things going on here:
1981          *      1) Add routes for on-link prefixes
1982          *      2) Configure prefixes with the auto flag set
1983          */
1984
1985         if (pinfo->onlink) {
1986                 struct rt6_info *rt;
1987                 unsigned long rt_expires;
1988
1989                 /* Avoid arithmetic overflow. Really, we could
1990                  * save rt_expires in seconds, likely valid_lft,
1991                  * but it would require division in fib gc, that it
1992                  * not good.
1993                  */
1994                 if (HZ > USER_HZ)
1995                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1996                 else
1997                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1998
1999                 if (addrconf_finite_timeout(rt_expires))
2000                         rt_expires *= HZ;
2001
2002                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2003                                                pinfo->prefix_len,
2004                                                dev,
2005                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2006                                                RTF_GATEWAY | RTF_DEFAULT);
2007
2008                 if (rt) {
2009                         /* Autoconf prefix route */
2010                         if (valid_lft == 0) {
2011                                 ip6_del_rt(rt);
2012                                 rt = NULL;
2013                         } else if (addrconf_finite_timeout(rt_expires)) {
2014                                 /* not infinity */
2015                                 rt6_set_expires(rt, jiffies + rt_expires);
2016                         } else {
2017                                 rt6_clean_expires(rt);
2018                         }
2019                 } else if (valid_lft) {
2020                         clock_t expires = 0;
2021                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2022                         if (addrconf_finite_timeout(rt_expires)) {
2023                                 /* not infinity */
2024                                 flags |= RTF_EXPIRES;
2025                                 expires = jiffies_to_clock_t(rt_expires);
2026                         }
2027                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2028                                               dev, expires, flags);
2029                 }
2030                 ip6_rt_put(rt);
2031         }
2032
2033         /* Try to figure out our local address for this prefix */
2034
2035         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2036                 struct inet6_ifaddr *ifp;
2037                 struct in6_addr addr;
2038                 int create = 0, update_lft = 0;
2039
2040                 if (pinfo->prefix_len == 64) {
2041                         memcpy(&addr, &pinfo->prefix, 8);
2042                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2043                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2044                                 in6_dev_put(in6_dev);
2045                                 return;
2046                         }
2047                         goto ok;
2048                 }
2049                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2050                                     pinfo->prefix_len);
2051                 in6_dev_put(in6_dev);
2052                 return;
2053
2054 ok:
2055
2056                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2057
2058                 if (ifp == NULL && valid_lft) {
2059                         int max_addresses = in6_dev->cnf.max_addresses;
2060                         u32 addr_flags = 0;
2061
2062 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2063                         if (in6_dev->cnf.optimistic_dad &&
2064                             !net->ipv6.devconf_all->forwarding && sllao)
2065                                 addr_flags = IFA_F_OPTIMISTIC;
2066 #endif
2067
2068                         /* Do not allow to create too much of autoconfigured
2069                          * addresses; this would be too easy way to crash kernel.
2070                          */
2071                         if (!max_addresses ||
2072                             ipv6_count_addresses(in6_dev) < max_addresses)
2073                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2074                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2075                                                     addr_flags);
2076
2077                         if (!ifp || IS_ERR(ifp)) {
2078                                 in6_dev_put(in6_dev);
2079                                 return;
2080                         }
2081
2082                         update_lft = create = 1;
2083                         ifp->cstamp = jiffies;
2084                         addrconf_dad_start(ifp);
2085                 }
2086
2087                 if (ifp) {
2088                         int flags;
2089                         unsigned long now;
2090 #ifdef CONFIG_IPV6_PRIVACY
2091                         struct inet6_ifaddr *ift;
2092 #endif
2093                         u32 stored_lft;
2094
2095                         /* update lifetime (RFC2462 5.5.3 e) */
2096                         spin_lock(&ifp->lock);
2097                         now = jiffies;
2098                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2099                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2100                         else
2101                                 stored_lft = 0;
2102                         if (!update_lft && stored_lft) {
2103                                 if (valid_lft > MIN_VALID_LIFETIME ||
2104                                     valid_lft > stored_lft)
2105                                         update_lft = 1;
2106                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2107                                         /* valid_lft <= stored_lft is always true */
2108                                         /*
2109                                          * RFC 4862 Section 5.5.3e:
2110                                          * "Note that the preferred lifetime of
2111                                          *  the corresponding address is always
2112                                          *  reset to the Preferred Lifetime in
2113                                          *  the received Prefix Information
2114                                          *  option, regardless of whether the
2115                                          *  valid lifetime is also reset or
2116                                          *  ignored."
2117                                          *
2118                                          *  So if the preferred lifetime in
2119                                          *  this advertisement is different
2120                                          *  than what we have stored, but the
2121                                          *  valid lifetime is invalid, just
2122                                          *  reset prefered_lft.
2123                                          *
2124                                          *  We must set the valid lifetime
2125                                          *  to the stored lifetime since we'll
2126                                          *  be updating the timestamp below,
2127                                          *  else we'll set it back to the
2128                                          *  minimum.
2129                                          */
2130                                         if (prefered_lft != ifp->prefered_lft) {
2131                                                 valid_lft = stored_lft;
2132                                                 update_lft = 1;
2133                                         }
2134                                 } else {
2135                                         valid_lft = MIN_VALID_LIFETIME;
2136                                         if (valid_lft < prefered_lft)
2137                                                 prefered_lft = valid_lft;
2138                                         update_lft = 1;
2139                                 }
2140                         }
2141
2142                         if (update_lft) {
2143                                 ifp->valid_lft = valid_lft;
2144                                 ifp->prefered_lft = prefered_lft;
2145                                 ifp->tstamp = now;
2146                                 flags = ifp->flags;
2147                                 ifp->flags &= ~IFA_F_DEPRECATED;
2148                                 spin_unlock(&ifp->lock);
2149
2150                                 if (!(flags&IFA_F_TENTATIVE))
2151                                         ipv6_ifa_notify(0, ifp);
2152                         } else
2153                                 spin_unlock(&ifp->lock);
2154
2155 #ifdef CONFIG_IPV6_PRIVACY
2156                         read_lock_bh(&in6_dev->lock);
2157                         /* update all temporary addresses in the list */
2158                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2159                                             tmp_list) {
2160                                 int age, max_valid, max_prefered;
2161
2162                                 if (ifp != ift->ifpub)
2163                                         continue;
2164
2165                                 /*
2166                                  * RFC 4941 section 3.3:
2167                                  * If a received option will extend the lifetime
2168                                  * of a public address, the lifetimes of
2169                                  * temporary addresses should be extended,
2170                                  * subject to the overall constraint that no
2171                                  * temporary addresses should ever remain
2172                                  * "valid" or "preferred" for a time longer than
2173                                  * (TEMP_VALID_LIFETIME) or
2174                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2175                                  * respectively.
2176                                  */
2177                                 age = (now - ift->cstamp) / HZ;
2178                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2179                                 if (max_valid < 0)
2180                                         max_valid = 0;
2181
2182                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2183                                                in6_dev->cnf.max_desync_factor -
2184                                                age;
2185                                 if (max_prefered < 0)
2186                                         max_prefered = 0;
2187
2188                                 if (valid_lft > max_valid)
2189                                         valid_lft = max_valid;
2190
2191                                 if (prefered_lft > max_prefered)
2192                                         prefered_lft = max_prefered;
2193
2194                                 spin_lock(&ift->lock);
2195                                 flags = ift->flags;
2196                                 ift->valid_lft = valid_lft;
2197                                 ift->prefered_lft = prefered_lft;
2198                                 ift->tstamp = now;
2199                                 if (prefered_lft > 0)
2200                                         ift->flags &= ~IFA_F_DEPRECATED;
2201
2202                                 spin_unlock(&ift->lock);
2203                                 if (!(flags&IFA_F_TENTATIVE))
2204                                         ipv6_ifa_notify(0, ift);
2205                         }
2206
2207                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2208                                 /*
2209                                  * When a new public address is created as
2210                                  * described in [ADDRCONF], also create a new
2211                                  * temporary address. Also create a temporary
2212                                  * address if it's enabled but no temporary
2213                                  * address currently exists.
2214                                  */
2215                                 read_unlock_bh(&in6_dev->lock);
2216                                 ipv6_create_tempaddr(ifp, NULL);
2217                         } else {
2218                                 read_unlock_bh(&in6_dev->lock);
2219                         }
2220 #endif
2221                         in6_ifa_put(ifp);
2222                         addrconf_verify(0);
2223                 }
2224         }
2225         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2226         in6_dev_put(in6_dev);
2227 }
2228
2229 /*
2230  *      Set destination address.
2231  *      Special case for SIT interfaces where we create a new "virtual"
2232  *      device.
2233  */
2234 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2235 {
2236         struct in6_ifreq ireq;
2237         struct net_device *dev;
2238         int err = -EINVAL;
2239
2240         rtnl_lock();
2241
2242         err = -EFAULT;
2243         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2244                 goto err_exit;
2245
2246         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2247
2248         err = -ENODEV;
2249         if (dev == NULL)
2250                 goto err_exit;
2251
2252 #if IS_ENABLED(CONFIG_IPV6_SIT)
2253         if (dev->type == ARPHRD_SIT) {
2254                 const struct net_device_ops *ops = dev->netdev_ops;
2255                 struct ifreq ifr;
2256                 struct ip_tunnel_parm p;
2257
2258                 err = -EADDRNOTAVAIL;
2259                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2260                         goto err_exit;
2261
2262                 memset(&p, 0, sizeof(p));
2263                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2264                 p.iph.saddr = 0;
2265                 p.iph.version = 4;
2266                 p.iph.ihl = 5;
2267                 p.iph.protocol = IPPROTO_IPV6;
2268                 p.iph.ttl = 64;
2269                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2270
2271                 if (ops->ndo_do_ioctl) {
2272                         mm_segment_t oldfs = get_fs();
2273
2274                         set_fs(KERNEL_DS);
2275                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2276                         set_fs(oldfs);
2277                 } else
2278                         err = -EOPNOTSUPP;
2279
2280                 if (err == 0) {
2281                         err = -ENOBUFS;
2282                         dev = __dev_get_by_name(net, p.name);
2283                         if (!dev)
2284                                 goto err_exit;
2285                         err = dev_open(dev);
2286                 }
2287         }
2288 #endif
2289
2290 err_exit:
2291         rtnl_unlock();
2292         return err;
2293 }
2294
2295 /*
2296  *      Manual configuration of address on an interface
2297  */
2298 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2299                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2300                           __u32 valid_lft)
2301 {
2302         struct inet6_ifaddr *ifp;
2303         struct inet6_dev *idev;
2304         struct net_device *dev;
2305         int scope;
2306         u32 flags;
2307         clock_t expires;
2308         unsigned long timeout;
2309
2310         ASSERT_RTNL();
2311
2312         if (plen > 128)
2313                 return -EINVAL;
2314
2315         /* check the lifetime */
2316         if (!valid_lft || prefered_lft > valid_lft)
2317                 return -EINVAL;
2318
2319         dev = __dev_get_by_index(net, ifindex);
2320         if (!dev)
2321                 return -ENODEV;
2322
2323         idev = addrconf_add_dev(dev);
2324         if (IS_ERR(idev))
2325                 return PTR_ERR(idev);
2326
2327         scope = ipv6_addr_scope(pfx);
2328
2329         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2330         if (addrconf_finite_timeout(timeout)) {
2331                 expires = jiffies_to_clock_t(timeout * HZ);
2332                 valid_lft = timeout;
2333                 flags = RTF_EXPIRES;
2334         } else {
2335                 expires = 0;
2336                 flags = 0;
2337                 ifa_flags |= IFA_F_PERMANENT;
2338         }
2339
2340         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2341         if (addrconf_finite_timeout(timeout)) {
2342                 if (timeout == 0)
2343                         ifa_flags |= IFA_F_DEPRECATED;
2344                 prefered_lft = timeout;
2345         }
2346
2347         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2348
2349         if (!IS_ERR(ifp)) {
2350                 spin_lock_bh(&ifp->lock);
2351                 ifp->valid_lft = valid_lft;
2352                 ifp->prefered_lft = prefered_lft;
2353                 ifp->tstamp = jiffies;
2354                 spin_unlock_bh(&ifp->lock);
2355
2356                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2357                                       expires, flags);
2358                 /*
2359                  * Note that section 3.1 of RFC 4429 indicates
2360                  * that the Optimistic flag should not be set for
2361                  * manually configured addresses
2362                  */
2363                 addrconf_dad_start(ifp);
2364                 in6_ifa_put(ifp);
2365                 addrconf_verify(0);
2366                 return 0;
2367         }
2368
2369         return PTR_ERR(ifp);
2370 }
2371
2372 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2373                           unsigned int plen)
2374 {
2375         struct inet6_ifaddr *ifp;
2376         struct inet6_dev *idev;
2377         struct net_device *dev;
2378
2379         if (plen > 128)
2380                 return -EINVAL;
2381
2382         dev = __dev_get_by_index(net, ifindex);
2383         if (!dev)
2384                 return -ENODEV;
2385
2386         if ((idev = __in6_dev_get(dev)) == NULL)
2387                 return -ENXIO;
2388
2389         read_lock_bh(&idev->lock);
2390         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2391                 if (ifp->prefix_len == plen &&
2392                     ipv6_addr_equal(pfx, &ifp->addr)) {
2393                         in6_ifa_hold(ifp);
2394                         read_unlock_bh(&idev->lock);
2395
2396                         ipv6_del_addr(ifp);
2397
2398                         /* If the last address is deleted administratively,
2399                            disable IPv6 on this interface.
2400                          */
2401                         if (list_empty(&idev->addr_list))
2402                                 addrconf_ifdown(idev->dev, 1);
2403                         return 0;
2404                 }
2405         }
2406         read_unlock_bh(&idev->lock);
2407         return -EADDRNOTAVAIL;
2408 }
2409
2410
2411 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2412 {
2413         struct in6_ifreq ireq;
2414         int err;
2415
2416         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2417                 return -EPERM;
2418
2419         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2420                 return -EFAULT;
2421
2422         rtnl_lock();
2423         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2424                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2425                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2426         rtnl_unlock();
2427         return err;
2428 }
2429
2430 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2431 {
2432         struct in6_ifreq ireq;
2433         int err;
2434
2435         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2436                 return -EPERM;
2437
2438         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2439                 return -EFAULT;
2440
2441         rtnl_lock();
2442         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2443                              ireq.ifr6_prefixlen);
2444         rtnl_unlock();
2445         return err;
2446 }
2447
2448 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2449                      int plen, int scope)
2450 {
2451         struct inet6_ifaddr *ifp;
2452
2453         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2454         if (!IS_ERR(ifp)) {
2455                 spin_lock_bh(&ifp->lock);
2456                 ifp->flags &= ~IFA_F_TENTATIVE;
2457                 spin_unlock_bh(&ifp->lock);
2458                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2459                 in6_ifa_put(ifp);
2460         }
2461 }
2462
2463 #if IS_ENABLED(CONFIG_IPV6_SIT)
2464 static void sit_add_v4_addrs(struct inet6_dev *idev)
2465 {
2466         struct in6_addr addr;
2467         struct net_device *dev;
2468         struct net *net = dev_net(idev->dev);
2469         int scope;
2470
2471         ASSERT_RTNL();
2472
2473         memset(&addr, 0, sizeof(struct in6_addr));
2474         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2475
2476         if (idev->dev->flags&IFF_POINTOPOINT) {
2477                 addr.s6_addr32[0] = htonl(0xfe800000);
2478                 scope = IFA_LINK;
2479         } else {
2480                 scope = IPV6_ADDR_COMPATv4;
2481         }
2482
2483         if (addr.s6_addr32[3]) {
2484                 add_addr(idev, &addr, 128, scope);
2485                 return;
2486         }
2487
2488         for_each_netdev(net, dev) {
2489                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2490                 if (in_dev && (dev->flags & IFF_UP)) {
2491                         struct in_ifaddr *ifa;
2492
2493                         int flag = scope;
2494
2495                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2496                                 int plen;
2497
2498                                 addr.s6_addr32[3] = ifa->ifa_local;
2499
2500                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2501                                         continue;
2502                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2503                                         if (idev->dev->flags&IFF_POINTOPOINT)
2504                                                 continue;
2505                                         flag |= IFA_HOST;
2506                                 }
2507                                 if (idev->dev->flags&IFF_POINTOPOINT)
2508                                         plen = 64;
2509                                 else
2510                                         plen = 96;
2511
2512                                 add_addr(idev, &addr, plen, flag);
2513                         }
2514                 }
2515         }
2516 }
2517 #endif
2518
2519 static void init_loopback(struct net_device *dev)
2520 {
2521         struct inet6_dev  *idev;
2522
2523         /* ::1 */
2524
2525         ASSERT_RTNL();
2526
2527         if ((idev = ipv6_find_idev(dev)) == NULL) {
2528                 pr_debug("%s: add_dev failed\n", __func__);
2529                 return;
2530         }
2531
2532         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2533 }
2534
2535 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2536 {
2537         struct inet6_ifaddr *ifp;
2538         u32 addr_flags = IFA_F_PERMANENT;
2539
2540 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2541         if (idev->cnf.optimistic_dad &&
2542             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2543                 addr_flags |= IFA_F_OPTIMISTIC;
2544 #endif
2545
2546
2547         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2548         if (!IS_ERR(ifp)) {
2549                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2550                 addrconf_dad_start(ifp);
2551                 in6_ifa_put(ifp);
2552         }
2553 }
2554
2555 static void addrconf_dev_config(struct net_device *dev)
2556 {
2557         struct in6_addr addr;
2558         struct inet6_dev *idev;
2559
2560         ASSERT_RTNL();
2561
2562         if ((dev->type != ARPHRD_ETHER) &&
2563             (dev->type != ARPHRD_FDDI) &&
2564             (dev->type != ARPHRD_ARCNET) &&
2565             (dev->type != ARPHRD_INFINIBAND) &&
2566             (dev->type != ARPHRD_IEEE802154)) {
2567                 /* Alas, we support only Ethernet autoconfiguration. */
2568                 return;
2569         }
2570
2571         idev = addrconf_add_dev(dev);
2572         if (IS_ERR(idev))
2573                 return;
2574
2575         memset(&addr, 0, sizeof(struct in6_addr));
2576         addr.s6_addr32[0] = htonl(0xFE800000);
2577
2578         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2579                 addrconf_add_linklocal(idev, &addr);
2580 }
2581
2582 #if IS_ENABLED(CONFIG_IPV6_SIT)
2583 static void addrconf_sit_config(struct net_device *dev)
2584 {
2585         struct inet6_dev *idev;
2586
2587         ASSERT_RTNL();
2588
2589         /*
2590          * Configure the tunnel with one of our IPv4
2591          * addresses... we should configure all of
2592          * our v4 addrs in the tunnel
2593          */
2594
2595         if ((idev = ipv6_find_idev(dev)) == NULL) {
2596                 pr_debug("%s: add_dev failed\n", __func__);
2597                 return;
2598         }
2599
2600         if (dev->priv_flags & IFF_ISATAP) {
2601                 struct in6_addr addr;
2602
2603                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2604                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2605                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2606                         addrconf_add_linklocal(idev, &addr);
2607                 return;
2608         }
2609
2610         sit_add_v4_addrs(idev);
2611
2612         if (dev->flags&IFF_POINTOPOINT)
2613                 addrconf_add_mroute(dev);
2614         else
2615                 sit_route_add(dev);
2616 }
2617 #endif
2618
2619 #if IS_ENABLED(CONFIG_NET_IPGRE)
2620 static void addrconf_gre_config(struct net_device *dev)
2621 {
2622         struct inet6_dev *idev;
2623         struct in6_addr addr;
2624
2625         pr_info("%s(%s)\n", __func__, dev->name);
2626
2627         ASSERT_RTNL();
2628
2629         if ((idev = ipv6_find_idev(dev)) == NULL) {
2630                 pr_debug("%s: add_dev failed\n", __func__);
2631                 return;
2632         }
2633
2634         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2635         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2636
2637         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2638                 addrconf_add_linklocal(idev, &addr);
2639 }
2640 #endif
2641
2642 static inline int
2643 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2644 {
2645         struct in6_addr lladdr;
2646
2647         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2648                 addrconf_add_linklocal(idev, &lladdr);
2649                 return 0;
2650         }
2651         return -1;
2652 }
2653
2654 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2655 {
2656         struct net_device *link_dev;
2657         struct net *net = dev_net(idev->dev);
2658
2659         /* first try to inherit the link-local address from the link device */
2660         if (idev->dev->iflink &&
2661             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2662                 if (!ipv6_inherit_linklocal(idev, link_dev))
2663                         return;
2664         }
2665         /* then try to inherit it from any device */
2666         for_each_netdev(net, link_dev) {
2667                 if (!ipv6_inherit_linklocal(idev, link_dev))
2668                         return;
2669         }
2670         pr_debug("init ip6-ip6: add_linklocal failed\n");
2671 }
2672
2673 /*
2674  * Autoconfigure tunnel with a link-local address so routing protocols,
2675  * DHCPv6, MLD etc. can be run over the virtual link
2676  */
2677
2678 static void addrconf_ip6_tnl_config(struct net_device *dev)
2679 {
2680         struct inet6_dev *idev;
2681
2682         ASSERT_RTNL();
2683
2684         idev = addrconf_add_dev(dev);
2685         if (IS_ERR(idev)) {
2686                 pr_debug("init ip6-ip6: add_dev failed\n");
2687                 return;
2688         }
2689         ip6_tnl_add_linklocal(idev);
2690 }
2691
2692 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2693                            void *data)
2694 {
2695         struct net_device *dev = (struct net_device *) data;
2696         struct inet6_dev *idev = __in6_dev_get(dev);
2697         int run_pending = 0;
2698         int err;
2699
2700         switch (event) {
2701         case NETDEV_REGISTER:
2702                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2703                         idev = ipv6_add_dev(dev);
2704                         if (!idev)
2705                                 return notifier_from_errno(-ENOMEM);
2706                 }
2707                 break;
2708
2709         case NETDEV_UP:
2710         case NETDEV_CHANGE:
2711                 if (dev->flags & IFF_SLAVE)
2712                         break;
2713
2714                 if (event == NETDEV_UP) {
2715                         if (!addrconf_qdisc_ok(dev)) {
2716                                 /* device is not ready yet. */
2717                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2718                                         dev->name);
2719                                 break;
2720                         }
2721
2722                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2723                                 idev = ipv6_add_dev(dev);
2724
2725                         if (idev) {
2726                                 idev->if_flags |= IF_READY;
2727                                 run_pending = 1;
2728                         }
2729                 } else {
2730                         if (!addrconf_qdisc_ok(dev)) {
2731                                 /* device is still not ready. */
2732                                 break;
2733                         }
2734
2735                         if (idev) {
2736                                 if (idev->if_flags & IF_READY)
2737                                         /* device is already configured. */
2738                                         break;
2739                                 idev->if_flags |= IF_READY;
2740                         }
2741
2742                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2743                                 dev->name);
2744
2745                         run_pending = 1;
2746                 }
2747
2748                 switch (dev->type) {
2749 #if IS_ENABLED(CONFIG_IPV6_SIT)
2750                 case ARPHRD_SIT:
2751                         addrconf_sit_config(dev);
2752                         break;
2753 #endif
2754 #if IS_ENABLED(CONFIG_NET_IPGRE)
2755                 case ARPHRD_IPGRE:
2756                         addrconf_gre_config(dev);
2757                         break;
2758 #endif
2759                 case ARPHRD_TUNNEL6:
2760                         addrconf_ip6_tnl_config(dev);
2761                         break;
2762                 case ARPHRD_LOOPBACK:
2763                         init_loopback(dev);
2764                         break;
2765
2766                 default:
2767                         addrconf_dev_config(dev);
2768                         break;
2769                 }
2770
2771                 if (idev) {
2772                         if (run_pending)
2773                                 addrconf_dad_run(idev);
2774
2775                         /*
2776                          * If the MTU changed during the interface down,
2777                          * when the interface up, the changed MTU must be
2778                          * reflected in the idev as well as routers.
2779                          */
2780                         if (idev->cnf.mtu6 != dev->mtu &&
2781                             dev->mtu >= IPV6_MIN_MTU) {
2782                                 rt6_mtu_change(dev, dev->mtu);
2783                                 idev->cnf.mtu6 = dev->mtu;
2784                         }
2785                         idev->tstamp = jiffies;
2786                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2787
2788                         /*
2789                          * If the changed mtu during down is lower than
2790                          * IPV6_MIN_MTU stop IPv6 on this interface.
2791                          */
2792                         if (dev->mtu < IPV6_MIN_MTU)
2793                                 addrconf_ifdown(dev, 1);
2794                 }
2795                 break;
2796
2797         case NETDEV_CHANGEMTU:
2798                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2799                         rt6_mtu_change(dev, dev->mtu);
2800                         idev->cnf.mtu6 = dev->mtu;
2801                         break;
2802                 }
2803
2804                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2805                         idev = ipv6_add_dev(dev);
2806                         if (idev)
2807                                 break;
2808                 }
2809
2810                 /*
2811                  * MTU falled under IPV6_MIN_MTU.
2812                  * Stop IPv6 on this interface.
2813                  */
2814
2815         case NETDEV_DOWN:
2816         case NETDEV_UNREGISTER:
2817                 /*
2818                  *      Remove all addresses from this interface.
2819                  */
2820                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2821                 break;
2822
2823         case NETDEV_CHANGENAME:
2824                 if (idev) {
2825                         snmp6_unregister_dev(idev);
2826                         addrconf_sysctl_unregister(idev);
2827                         addrconf_sysctl_register(idev);
2828                         err = snmp6_register_dev(idev);
2829                         if (err)
2830                                 return notifier_from_errno(err);
2831                 }
2832                 break;
2833
2834         case NETDEV_PRE_TYPE_CHANGE:
2835         case NETDEV_POST_TYPE_CHANGE:
2836                 addrconf_type_change(dev, event);
2837                 break;
2838         }
2839
2840         return NOTIFY_OK;
2841 }
2842
2843 /*
2844  *      addrconf module should be notified of a device going up
2845  */
2846 static struct notifier_block ipv6_dev_notf = {
2847         .notifier_call = addrconf_notify,
2848 };
2849
2850 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2851 {
2852         struct inet6_dev *idev;
2853         ASSERT_RTNL();
2854
2855         idev = __in6_dev_get(dev);
2856
2857         if (event == NETDEV_POST_TYPE_CHANGE)
2858                 ipv6_mc_remap(idev);
2859         else if (event == NETDEV_PRE_TYPE_CHANGE)
2860                 ipv6_mc_unmap(idev);
2861 }
2862
2863 static int addrconf_ifdown(struct net_device *dev, int how)
2864 {
2865         struct net *net = dev_net(dev);
2866         struct inet6_dev *idev;
2867         struct inet6_ifaddr *ifa;
2868         int state, i;
2869
2870         ASSERT_RTNL();
2871
2872         rt6_ifdown(net, dev);
2873         neigh_ifdown(&nd_tbl, dev);
2874
2875         idev = __in6_dev_get(dev);
2876         if (idev == NULL)
2877                 return -ENODEV;
2878
2879         /*
2880          * Step 1: remove reference to ipv6 device from parent device.
2881          *         Do not dev_put!
2882          */
2883         if (how) {
2884                 idev->dead = 1;
2885
2886                 /* protected by rtnl_lock */
2887                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2888
2889                 /* Step 1.5: remove snmp6 entry */
2890                 snmp6_unregister_dev(idev);
2891
2892         }
2893
2894         /* Step 2: clear hash table */
2895         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2896                 struct hlist_head *h = &inet6_addr_lst[i];
2897                 struct hlist_node *n;
2898
2899                 spin_lock_bh(&addrconf_hash_lock);
2900         restart:
2901                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2902                         if (ifa->idev == idev) {
2903                                 hlist_del_init_rcu(&ifa->addr_lst);
2904                                 addrconf_del_timer(ifa);
2905                                 goto restart;
2906                         }
2907                 }
2908                 spin_unlock_bh(&addrconf_hash_lock);
2909         }
2910
2911         write_lock_bh(&idev->lock);
2912
2913         /* Step 2: clear flags for stateless addrconf */
2914         if (!how)
2915                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2916
2917 #ifdef CONFIG_IPV6_PRIVACY
2918         if (how && del_timer(&idev->regen_timer))
2919                 in6_dev_put(idev);
2920
2921         /* Step 3: clear tempaddr list */
2922         while (!list_empty(&idev->tempaddr_list)) {
2923                 ifa = list_first_entry(&idev->tempaddr_list,
2924                                        struct inet6_ifaddr, tmp_list);
2925                 list_del(&ifa->tmp_list);
2926                 write_unlock_bh(&idev->lock);
2927                 spin_lock_bh(&ifa->lock);
2928
2929                 if (ifa->ifpub) {
2930                         in6_ifa_put(ifa->ifpub);
2931                         ifa->ifpub = NULL;
2932                 }
2933                 spin_unlock_bh(&ifa->lock);
2934                 in6_ifa_put(ifa);
2935                 write_lock_bh(&idev->lock);
2936         }
2937 #endif
2938
2939         while (!list_empty(&idev->addr_list)) {
2940                 ifa = list_first_entry(&idev->addr_list,
2941                                        struct inet6_ifaddr, if_list);
2942                 addrconf_del_timer(ifa);
2943
2944                 list_del(&ifa->if_list);
2945
2946                 write_unlock_bh(&idev->lock);
2947
2948                 spin_lock_bh(&ifa->state_lock);
2949                 state = ifa->state;
2950                 ifa->state = INET6_IFADDR_STATE_DEAD;
2951                 spin_unlock_bh(&ifa->state_lock);
2952
2953                 if (state != INET6_IFADDR_STATE_DEAD) {
2954                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2955                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2956                 }
2957                 in6_ifa_put(ifa);
2958
2959                 write_lock_bh(&idev->lock);
2960         }
2961
2962         write_unlock_bh(&idev->lock);
2963
2964         /* Step 5: Discard multicast list */
2965         if (how)
2966                 ipv6_mc_destroy_dev(idev);
2967         else
2968                 ipv6_mc_down(idev);
2969
2970         idev->tstamp = jiffies;
2971
2972         /* Last: Shot the device (if unregistered) */
2973         if (how) {
2974                 addrconf_sysctl_unregister(idev);
2975                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2976                 neigh_ifdown(&nd_tbl, dev);
2977                 in6_dev_put(idev);
2978         }
2979         return 0;
2980 }
2981
2982 static void addrconf_rs_timer(unsigned long data)
2983 {
2984         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2985         struct inet6_dev *idev = ifp->idev;
2986
2987         read_lock(&idev->lock);
2988         if (idev->dead || !(idev->if_flags & IF_READY))
2989                 goto out;
2990
2991         if (!ipv6_accept_ra(idev))
2992                 goto out;
2993
2994         /* Announcement received after solicitation was sent */
2995         if (idev->if_flags & IF_RA_RCVD)
2996                 goto out;
2997
2998         spin_lock(&ifp->lock);
2999         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3000                 /* The wait after the last probe can be shorter */
3001                 addrconf_mod_timer(ifp, AC_RS,
3002                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3003                                    idev->cnf.rtr_solicit_delay :
3004                                    idev->cnf.rtr_solicit_interval);
3005                 spin_unlock(&ifp->lock);
3006
3007                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3008         } else {
3009                 spin_unlock(&ifp->lock);
3010                 /*
3011                  * Note: we do not support deprecated "all on-link"
3012                  * assumption any longer.
3013                  */
3014                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3015         }
3016
3017 out:
3018         read_unlock(&idev->lock);
3019         in6_ifa_put(ifp);
3020 }
3021
3022 /*
3023  *      Duplicate Address Detection
3024  */
3025 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3026 {
3027         unsigned long rand_num;
3028         struct inet6_dev *idev = ifp->idev;
3029
3030         if (ifp->flags & IFA_F_OPTIMISTIC)
3031                 rand_num = 0;
3032         else
3033                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3034
3035         ifp->probes = idev->cnf.dad_transmits;
3036         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3037 }
3038
3039 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3040 {
3041         struct inet6_dev *idev = ifp->idev;
3042         struct net_device *dev = idev->dev;
3043
3044         addrconf_join_solict(dev, &ifp->addr);
3045
3046         net_srandom(ifp->addr.s6_addr32[3]);
3047
3048         read_lock_bh(&idev->lock);
3049         spin_lock(&ifp->lock);
3050         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3051                 goto out;
3052
3053         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3054             idev->cnf.accept_dad < 1 ||
3055             !(ifp->flags&IFA_F_TENTATIVE) ||
3056             ifp->flags & IFA_F_NODAD) {
3057                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3058                 spin_unlock(&ifp->lock);
3059                 read_unlock_bh(&idev->lock);
3060
3061                 addrconf_dad_completed(ifp);
3062                 return;
3063         }
3064
3065         if (!(idev->if_flags & IF_READY)) {
3066                 spin_unlock(&ifp->lock);
3067                 read_unlock_bh(&idev->lock);
3068                 /*
3069                  * If the device is not ready:
3070                  * - keep it tentative if it is a permanent address.
3071                  * - otherwise, kill it.
3072                  */
3073                 in6_ifa_hold(ifp);
3074                 addrconf_dad_stop(ifp, 0);
3075                 return;
3076         }
3077
3078         /*
3079          * Optimistic nodes can start receiving
3080          * Frames right away
3081          */
3082         if (ifp->flags & IFA_F_OPTIMISTIC)
3083                 ip6_ins_rt(ifp->rt);
3084
3085         addrconf_dad_kick(ifp);
3086 out:
3087         spin_unlock(&ifp->lock);
3088         read_unlock_bh(&idev->lock);
3089 }
3090
3091 static void addrconf_dad_timer(unsigned long data)
3092 {
3093         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3094         struct inet6_dev *idev = ifp->idev;
3095         struct in6_addr mcaddr;
3096
3097         if (!ifp->probes && addrconf_dad_end(ifp))
3098                 goto out;
3099
3100         read_lock(&idev->lock);
3101         if (idev->dead || !(idev->if_flags & IF_READY)) {
3102                 read_unlock(&idev->lock);
3103                 goto out;
3104         }
3105
3106         spin_lock(&ifp->lock);
3107         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3108                 spin_unlock(&ifp->lock);
3109                 read_unlock(&idev->lock);
3110                 goto out;
3111         }
3112
3113         if (ifp->probes == 0) {
3114                 /*
3115                  * DAD was successful
3116                  */
3117
3118                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3119                 spin_unlock(&ifp->lock);
3120                 read_unlock(&idev->lock);
3121
3122                 addrconf_dad_completed(ifp);
3123
3124                 goto out;
3125         }
3126
3127         ifp->probes--;
3128         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3129         spin_unlock(&ifp->lock);
3130         read_unlock(&idev->lock);
3131
3132         /* send a neighbour solicitation for our addr */
3133         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3134         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3135 out:
3136         in6_ifa_put(ifp);
3137 }
3138
3139 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3140 {
3141         struct net_device *dev = ifp->idev->dev;
3142
3143         /*
3144          *      Configure the address for reception. Now it is valid.
3145          */
3146
3147         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3148
3149         /* If added prefix is link local and we are prepared to process
3150            router advertisements, start sending router solicitations.
3151          */
3152
3153         if (ipv6_accept_ra(ifp->idev) &&
3154             ifp->idev->cnf.rtr_solicits > 0 &&
3155             (dev->flags&IFF_LOOPBACK) == 0 &&
3156             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3157                 /*
3158                  *      If a host as already performed a random delay
3159                  *      [...] as part of DAD [...] there is no need
3160                  *      to delay again before sending the first RS
3161                  */
3162                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3163
3164                 spin_lock_bh(&ifp->lock);
3165                 ifp->probes = 1;
3166                 ifp->idev->if_flags |= IF_RS_SENT;
3167                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3168                 spin_unlock_bh(&ifp->lock);
3169         }
3170 }
3171
3172 static void addrconf_dad_run(struct inet6_dev *idev)
3173 {
3174         struct inet6_ifaddr *ifp;
3175
3176         read_lock_bh(&idev->lock);
3177         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3178                 spin_lock(&ifp->lock);
3179                 if (ifp->flags & IFA_F_TENTATIVE &&
3180                     ifp->state == INET6_IFADDR_STATE_DAD)
3181                         addrconf_dad_kick(ifp);
3182                 spin_unlock(&ifp->lock);
3183         }
3184         read_unlock_bh(&idev->lock);
3185 }
3186
3187 #ifdef CONFIG_PROC_FS
3188 struct if6_iter_state {
3189         struct seq_net_private p;
3190         int bucket;
3191         int offset;
3192 };
3193
3194 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3195 {
3196         struct inet6_ifaddr *ifa = NULL;
3197         struct if6_iter_state *state = seq->private;
3198         struct net *net = seq_file_net(seq);
3199         int p = 0;
3200
3201         /* initial bucket if pos is 0 */
3202         if (pos == 0) {
3203                 state->bucket = 0;
3204                 state->offset = 0;
3205         }
3206
3207         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3208                 struct hlist_node *n;
3209                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3210                                          addr_lst) {
3211                         if (!net_eq(dev_net(ifa->idev->dev), net))
3212                                 continue;
3213                         /* sync with offset */
3214                         if (p < state->offset) {
3215                                 p++;
3216                                 continue;
3217                         }
3218                         state->offset++;
3219                         return ifa;
3220                 }
3221
3222                 /* prepare for next bucket */
3223                 state->offset = 0;
3224                 p = 0;
3225         }
3226         return NULL;
3227 }
3228
3229 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3230                                          struct inet6_ifaddr *ifa)
3231 {
3232         struct if6_iter_state *state = seq->private;
3233         struct net *net = seq_file_net(seq);
3234         struct hlist_node *n = &ifa->addr_lst;
3235
3236         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3237                 if (!net_eq(dev_net(ifa->idev->dev), net))
3238                         continue;
3239                 state->offset++;
3240                 return ifa;
3241         }
3242
3243         while (++state->bucket < IN6_ADDR_HSIZE) {
3244                 state->offset = 0;
3245                 hlist_for_each_entry_rcu_bh(ifa, n,
3246                                      &inet6_addr_lst[state->bucket], addr_lst) {
3247                         if (!net_eq(dev_net(ifa->idev->dev), net))
3248                                 continue;
3249                         state->offset++;
3250                         return ifa;
3251                 }
3252         }
3253
3254         return NULL;
3255 }
3256
3257 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3258         __acquires(rcu_bh)
3259 {
3260         rcu_read_lock_bh();
3261         return if6_get_first(seq, *pos);
3262 }
3263
3264 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3265 {
3266         struct inet6_ifaddr *ifa;
3267
3268         ifa = if6_get_next(seq, v);
3269         ++*pos;
3270         return ifa;
3271 }
3272
3273 static void if6_seq_stop(struct seq_file *seq, void *v)
3274         __releases(rcu_bh)
3275 {
3276         rcu_read_unlock_bh();
3277 }
3278
3279 static int if6_seq_show(struct seq_file *seq, void *v)
3280 {
3281         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3282         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3283                    &ifp->addr,
3284                    ifp->idev->dev->ifindex,
3285                    ifp->prefix_len,
3286                    ifp->scope,
3287                    ifp->flags,
3288                    ifp->idev->dev->name);
3289         return 0;
3290 }
3291
3292 static const struct seq_operations if6_seq_ops = {
3293         .start  = if6_seq_start,
3294         .next   = if6_seq_next,
3295         .show   = if6_seq_show,
3296         .stop   = if6_seq_stop,
3297 };
3298
3299 static int if6_seq_open(struct inode *inode, struct file *file)
3300 {
3301         return seq_open_net(inode, file, &if6_seq_ops,
3302                             sizeof(struct if6_iter_state));
3303 }
3304
3305 static const struct file_operations if6_fops = {
3306         .owner          = THIS_MODULE,
3307         .open           = if6_seq_open,
3308         .read           = seq_read,
3309         .llseek         = seq_lseek,
3310         .release        = seq_release_net,
3311 };
3312
3313 static int __net_init if6_proc_net_init(struct net *net)
3314 {
3315         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3316                 return -ENOMEM;
3317         return 0;
3318 }
3319
3320 static void __net_exit if6_proc_net_exit(struct net *net)
3321 {
3322        proc_net_remove(net, "if_inet6");
3323 }
3324
3325 static struct pernet_operations if6_proc_net_ops = {
3326        .init = if6_proc_net_init,
3327        .exit = if6_proc_net_exit,
3328 };
3329
3330 int __init if6_proc_init(void)
3331 {
3332         return register_pernet_subsys(&if6_proc_net_ops);
3333 }
3334
3335 void if6_proc_exit(void)
3336 {
3337         unregister_pernet_subsys(&if6_proc_net_ops);
3338 }
3339 #endif  /* CONFIG_PROC_FS */
3340
3341 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3342 /* Check if address is a home address configured on any interface. */
3343 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3344 {
3345         int ret = 0;
3346         struct inet6_ifaddr *ifp = NULL;
3347         struct hlist_node *n;
3348         unsigned int hash = inet6_addr_hash(addr);
3349
3350         rcu_read_lock_bh();
3351         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3352                 if (!net_eq(dev_net(ifp->idev->dev), net))
3353                         continue;
3354                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3355                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3356                         ret = 1;
3357                         break;
3358                 }
3359         }
3360         rcu_read_unlock_bh();
3361         return ret;
3362 }
3363 #endif
3364
3365 /*
3366  *      Periodic address status verification
3367  */
3368
3369 static void addrconf_verify(unsigned long foo)
3370 {
3371         unsigned long now, next, next_sec, next_sched;
3372         struct inet6_ifaddr *ifp;
3373         struct hlist_node *node;
3374         int i;
3375
3376         rcu_read_lock_bh();
3377         spin_lock(&addrconf_verify_lock);
3378         now = jiffies;
3379         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3380
3381         del_timer(&addr_chk_timer);
3382
3383         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3384 restart:
3385                 hlist_for_each_entry_rcu_bh(ifp, node,
3386                                          &inet6_addr_lst[i], addr_lst) {
3387                         unsigned long age;
3388
3389                         if (ifp->flags & IFA_F_PERMANENT)
3390                                 continue;
3391
3392                         spin_lock(&ifp->lock);
3393                         /* We try to batch several events at once. */
3394                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3395
3396                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3397                             age >= ifp->valid_lft) {
3398                                 spin_unlock(&ifp->lock);
3399                                 in6_ifa_hold(ifp);
3400                                 ipv6_del_addr(ifp);
3401                                 goto restart;
3402                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3403                                 spin_unlock(&ifp->lock);
3404                                 continue;
3405                         } else if (age >= ifp->prefered_lft) {
3406                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3407                                 int deprecate = 0;
3408
3409                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3410                                         deprecate = 1;
3411                                         ifp->flags |= IFA_F_DEPRECATED;
3412                                 }
3413
3414                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3415                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3416
3417                                 spin_unlock(&ifp->lock);
3418
3419                                 if (deprecate) {
3420                                         in6_ifa_hold(ifp);
3421
3422                                         ipv6_ifa_notify(0, ifp);
3423                                         in6_ifa_put(ifp);
3424                                         goto restart;
3425                                 }
3426 #ifdef CONFIG_IPV6_PRIVACY
3427                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3428                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3429                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3430                                         ifp->idev->cnf.dad_transmits *
3431                                         ifp->idev->nd_parms->retrans_time / HZ;
3432
3433                                 if (age >= ifp->prefered_lft - regen_advance) {
3434                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3435                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3436                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3437                                         if (!ifp->regen_count && ifpub) {
3438                                                 ifp->regen_count++;
3439                                                 in6_ifa_hold(ifp);
3440                                                 in6_ifa_hold(ifpub);
3441                                                 spin_unlock(&ifp->lock);
3442
3443                                                 spin_lock(&ifpub->lock);
3444                                                 ifpub->regen_count = 0;
3445                                                 spin_unlock(&ifpub->lock);
3446                                                 ipv6_create_tempaddr(ifpub, ifp);
3447                                                 in6_ifa_put(ifpub);
3448                                                 in6_ifa_put(ifp);
3449                                                 goto restart;
3450                                         }
3451                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3452                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3453                                 spin_unlock(&ifp->lock);
3454 #endif
3455                         } else {
3456                                 /* ifp->prefered_lft <= ifp->valid_lft */
3457                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3458                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3459                                 spin_unlock(&ifp->lock);
3460                         }
3461                 }
3462         }
3463
3464         next_sec = round_jiffies_up(next);
3465         next_sched = next;
3466
3467         /* If rounded timeout is accurate enough, accept it. */
3468         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3469                 next_sched = next_sec;
3470
3471         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3472         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3473                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3474
3475         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3476               now, next, next_sec, next_sched));
3477
3478         addr_chk_timer.expires = next_sched;
3479         add_timer(&addr_chk_timer);
3480         spin_unlock(&addrconf_verify_lock);
3481         rcu_read_unlock_bh();
3482 }
3483
3484 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3485 {
3486         struct in6_addr *pfx = NULL;
3487
3488         if (addr)
3489                 pfx = nla_data(addr);
3490
3491         if (local) {
3492                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3493                         pfx = NULL;
3494                 else
3495                         pfx = nla_data(local);
3496         }
3497
3498         return pfx;
3499 }
3500
3501 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3502         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3503         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3504         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3505 };
3506
3507 static int
3508 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3509 {
3510         struct net *net = sock_net(skb->sk);
3511         struct ifaddrmsg *ifm;
3512         struct nlattr *tb[IFA_MAX+1];
3513         struct in6_addr *pfx;
3514         int err;
3515
3516         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3517         if (err < 0)
3518                 return err;
3519
3520         ifm = nlmsg_data(nlh);
3521         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3522         if (pfx == NULL)
3523                 return -EINVAL;
3524
3525         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3526 }
3527
3528 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3529                              u32 prefered_lft, u32 valid_lft)
3530 {
3531         u32 flags;
3532         clock_t expires;
3533         unsigned long timeout;
3534
3535         if (!valid_lft || (prefered_lft > valid_lft))
3536                 return -EINVAL;
3537
3538         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3539         if (addrconf_finite_timeout(timeout)) {
3540                 expires = jiffies_to_clock_t(timeout * HZ);
3541                 valid_lft = timeout;
3542                 flags = RTF_EXPIRES;
3543         } else {
3544                 expires = 0;
3545                 flags = 0;
3546                 ifa_flags |= IFA_F_PERMANENT;
3547         }
3548
3549         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3550         if (addrconf_finite_timeout(timeout)) {
3551                 if (timeout == 0)
3552                         ifa_flags |= IFA_F_DEPRECATED;
3553                 prefered_lft = timeout;
3554         }
3555
3556         spin_lock_bh(&ifp->lock);
3557         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3558         ifp->tstamp = jiffies;
3559         ifp->valid_lft = valid_lft;
3560         ifp->prefered_lft = prefered_lft;
3561
3562         spin_unlock_bh(&ifp->lock);
3563         if (!(ifp->flags&IFA_F_TENTATIVE))
3564                 ipv6_ifa_notify(0, ifp);
3565
3566         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3567                               expires, flags);
3568         addrconf_verify(0);
3569
3570         return 0;
3571 }
3572
3573 static int
3574 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3575 {
3576         struct net *net = sock_net(skb->sk);
3577         struct ifaddrmsg *ifm;
3578         struct nlattr *tb[IFA_MAX+1];
3579         struct in6_addr *pfx;
3580         struct inet6_ifaddr *ifa;
3581         struct net_device *dev;
3582         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3583         u8 ifa_flags;
3584         int err;
3585
3586         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3587         if (err < 0)
3588                 return err;
3589
3590         ifm = nlmsg_data(nlh);
3591         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3592         if (pfx == NULL)
3593                 return -EINVAL;
3594
3595         if (tb[IFA_CACHEINFO]) {
3596                 struct ifa_cacheinfo *ci;
3597
3598                 ci = nla_data(tb[IFA_CACHEINFO]);
3599                 valid_lft = ci->ifa_valid;
3600                 preferred_lft = ci->ifa_prefered;
3601         } else {
3602                 preferred_lft = INFINITY_LIFE_TIME;
3603                 valid_lft = INFINITY_LIFE_TIME;
3604         }
3605
3606         dev =  __dev_get_by_index(net, ifm->ifa_index);
3607         if (dev == NULL)
3608                 return -ENODEV;
3609
3610         /* We ignore other flags so far. */
3611         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3612
3613         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3614         if (ifa == NULL) {
3615                 /*
3616                  * It would be best to check for !NLM_F_CREATE here but
3617                  * userspace alreay relies on not having to provide this.
3618                  */
3619                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3620                                       ifm->ifa_prefixlen, ifa_flags,
3621                                       preferred_lft, valid_lft);
3622         }
3623
3624         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3625             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3626                 err = -EEXIST;
3627         else
3628                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3629
3630         in6_ifa_put(ifa);
3631
3632         return err;
3633 }
3634
3635 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3636                           u8 scope, int ifindex)
3637 {
3638         struct ifaddrmsg *ifm;
3639
3640         ifm = nlmsg_data(nlh);
3641         ifm->ifa_family = AF_INET6;
3642         ifm->ifa_prefixlen = prefixlen;
3643         ifm->ifa_flags = flags;
3644         ifm->ifa_scope = scope;
3645         ifm->ifa_index = ifindex;
3646 }
3647
3648 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3649                          unsigned long tstamp, u32 preferred, u32 valid)
3650 {
3651         struct ifa_cacheinfo ci;
3652
3653         ci.cstamp = cstamp_delta(cstamp);
3654         ci.tstamp = cstamp_delta(tstamp);
3655         ci.ifa_prefered = preferred;
3656         ci.ifa_valid = valid;
3657
3658         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3659 }
3660
3661 static inline int rt_scope(int ifa_scope)
3662 {
3663         if (ifa_scope & IFA_HOST)
3664                 return RT_SCOPE_HOST;
3665         else if (ifa_scope & IFA_LINK)
3666                 return RT_SCOPE_LINK;
3667         else if (ifa_scope & IFA_SITE)
3668                 return RT_SCOPE_SITE;
3669         else
3670                 return RT_SCOPE_UNIVERSE;
3671 }
3672
3673 static inline int inet6_ifaddr_msgsize(void)
3674 {
3675         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3676                + nla_total_size(16) /* IFA_ADDRESS */
3677                + nla_total_size(sizeof(struct ifa_cacheinfo));
3678 }
3679
3680 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3681                              u32 portid, u32 seq, int event, unsigned int flags)
3682 {
3683         struct nlmsghdr  *nlh;
3684         u32 preferred, valid;
3685
3686         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3687         if (nlh == NULL)
3688                 return -EMSGSIZE;
3689
3690         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3691                       ifa->idev->dev->ifindex);
3692
3693         if (!(ifa->flags&IFA_F_PERMANENT)) {
3694                 preferred = ifa->prefered_lft;
3695                 valid = ifa->valid_lft;
3696                 if (preferred != INFINITY_LIFE_TIME) {
3697                         long tval = (jiffies - ifa->tstamp)/HZ;
3698                         if (preferred > tval)
3699                                 preferred -= tval;
3700                         else
3701                                 preferred = 0;
3702                         if (valid != INFINITY_LIFE_TIME) {
3703                                 if (valid > tval)
3704                                         valid -= tval;
3705                                 else
3706                                         valid = 0;
3707                         }
3708                 }
3709         } else {
3710                 preferred = INFINITY_LIFE_TIME;
3711                 valid = INFINITY_LIFE_TIME;
3712         }
3713
3714         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3715             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3716                 nlmsg_cancel(skb, nlh);
3717                 return -EMSGSIZE;
3718         }
3719
3720         return nlmsg_end(skb, nlh);
3721 }
3722
3723 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3724                                 u32 portid, u32 seq, int event, u16 flags)
3725 {
3726         struct nlmsghdr  *nlh;
3727         u8 scope = RT_SCOPE_UNIVERSE;
3728         int ifindex = ifmca->idev->dev->ifindex;
3729
3730         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3731                 scope = RT_SCOPE_SITE;
3732
3733         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3734         if (nlh == NULL)
3735                 return -EMSGSIZE;
3736
3737         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3738         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3739             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3740                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3741                 nlmsg_cancel(skb, nlh);
3742                 return -EMSGSIZE;
3743         }
3744
3745         return nlmsg_end(skb, nlh);
3746 }
3747
3748 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3749                                 u32 portid, u32 seq, int event, unsigned int flags)
3750 {
3751         struct nlmsghdr  *nlh;
3752         u8 scope = RT_SCOPE_UNIVERSE;
3753         int ifindex = ifaca->aca_idev->dev->ifindex;
3754
3755         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3756                 scope = RT_SCOPE_SITE;
3757
3758         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3759         if (nlh == NULL)
3760                 return -EMSGSIZE;
3761
3762         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3763         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3764             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3765                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3766                 nlmsg_cancel(skb, nlh);
3767                 return -EMSGSIZE;
3768         }
3769
3770         return nlmsg_end(skb, nlh);
3771 }
3772
3773 enum addr_type_t {
3774         UNICAST_ADDR,
3775         MULTICAST_ADDR,
3776         ANYCAST_ADDR,
3777 };
3778
3779 /* called with rcu_read_lock() */
3780 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3781                           struct netlink_callback *cb, enum addr_type_t type,
3782                           int s_ip_idx, int *p_ip_idx)
3783 {
3784         struct ifmcaddr6 *ifmca;
3785         struct ifacaddr6 *ifaca;
3786         int err = 1;
3787         int ip_idx = *p_ip_idx;
3788
3789         read_lock_bh(&idev->lock);
3790         switch (type) {
3791         case UNICAST_ADDR: {
3792                 struct inet6_ifaddr *ifa;
3793
3794                 /* unicast address incl. temp addr */
3795                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3796                         if (++ip_idx < s_ip_idx)
3797                                 continue;
3798                         err = inet6_fill_ifaddr(skb, ifa,
3799                                                 NETLINK_CB(cb->skb).portid,
3800                                                 cb->nlh->nlmsg_seq,
3801                                                 RTM_NEWADDR,
3802                                                 NLM_F_MULTI);
3803                         if (err <= 0)
3804                                 break;
3805                 }
3806                 break;
3807         }
3808         case MULTICAST_ADDR:
3809                 /* multicast address */
3810                 for (ifmca = idev->mc_list; ifmca;
3811                      ifmca = ifmca->next, ip_idx++) {
3812                         if (ip_idx < s_ip_idx)
3813                                 continue;
3814                         err = inet6_fill_ifmcaddr(skb, ifmca,
3815                                                   NETLINK_CB(cb->skb).portid,
3816                                                   cb->nlh->nlmsg_seq,
3817                                                   RTM_GETMULTICAST,
3818                                                   NLM_F_MULTI);
3819                         if (err <= 0)
3820                                 break;
3821                 }
3822                 break;
3823         case ANYCAST_ADDR:
3824                 /* anycast address */
3825                 for (ifaca = idev->ac_list; ifaca;
3826                      ifaca = ifaca->aca_next, ip_idx++) {
3827                         if (ip_idx < s_ip_idx)
3828                                 continue;
3829                         err = inet6_fill_ifacaddr(skb, ifaca,
3830                                                   NETLINK_CB(cb->skb).portid,
3831                                                   cb->nlh->nlmsg_seq,
3832                                                   RTM_GETANYCAST,
3833                                                   NLM_F_MULTI);
3834                         if (err <= 0)
3835                                 break;
3836                 }
3837                 break;
3838         default:
3839                 break;
3840         }
3841         read_unlock_bh(&idev->lock);
3842         *p_ip_idx = ip_idx;
3843         return err;
3844 }
3845
3846 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3847                            enum addr_type_t type)
3848 {
3849         struct net *net = sock_net(skb->sk);
3850         int h, s_h;
3851         int idx, ip_idx;
3852         int s_idx, s_ip_idx;
3853         struct net_device *dev;
3854         struct inet6_dev *idev;
3855         struct hlist_head *head;
3856         struct hlist_node *node;
3857
3858         s_h = cb->args[0];
3859         s_idx = idx = cb->args[1];
3860         s_ip_idx = ip_idx = cb->args[2];
3861
3862         rcu_read_lock();
3863         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3864                 idx = 0;
3865                 head = &net->dev_index_head[h];
3866                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3867                         if (idx < s_idx)
3868                                 goto cont;
3869                         if (h > s_h || idx > s_idx)
3870                                 s_ip_idx = 0;
3871                         ip_idx = 0;
3872                         idev = __in6_dev_get(dev);
3873                         if (!idev)
3874                                 goto cont;
3875
3876                         if (in6_dump_addrs(idev, skb, cb, type,
3877                                            s_ip_idx, &ip_idx) <= 0)
3878                                 goto done;
3879 cont:
3880                         idx++;
3881                 }
3882         }
3883 done:
3884         rcu_read_unlock();
3885         cb->args[0] = h;
3886         cb->args[1] = idx;
3887         cb->args[2] = ip_idx;
3888
3889         return skb->len;
3890 }
3891
3892 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3893 {
3894         enum addr_type_t type = UNICAST_ADDR;
3895
3896         return inet6_dump_addr(skb, cb, type);
3897 }
3898
3899 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3900 {
3901         enum addr_type_t type = MULTICAST_ADDR;
3902
3903         return inet6_dump_addr(skb, cb, type);
3904 }
3905
3906
3907 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3908 {
3909         enum addr_type_t type = ANYCAST_ADDR;
3910
3911         return inet6_dump_addr(skb, cb, type);
3912 }
3913
3914 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3915                              void *arg)
3916 {
3917         struct net *net = sock_net(in_skb->sk);
3918         struct ifaddrmsg *ifm;
3919         struct nlattr *tb[IFA_MAX+1];
3920         struct in6_addr *addr = NULL;
3921         struct net_device *dev = NULL;
3922         struct inet6_ifaddr *ifa;
3923         struct sk_buff *skb;
3924         int err;
3925
3926         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3927         if (err < 0)
3928                 goto errout;
3929
3930         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3931         if (addr == NULL) {
3932                 err = -EINVAL;
3933                 goto errout;
3934         }
3935
3936         ifm = nlmsg_data(nlh);
3937         if (ifm->ifa_index)
3938                 dev = __dev_get_by_index(net, ifm->ifa_index);
3939
3940         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3941         if (!ifa) {
3942                 err = -EADDRNOTAVAIL;
3943                 goto errout;
3944         }
3945
3946         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3947         if (!skb) {
3948                 err = -ENOBUFS;
3949                 goto errout_ifa;
3950         }
3951
3952         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3953                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3954         if (err < 0) {
3955                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3956                 WARN_ON(err == -EMSGSIZE);
3957                 kfree_skb(skb);
3958                 goto errout_ifa;
3959         }
3960         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3961 errout_ifa:
3962         in6_ifa_put(ifa);
3963 errout:
3964         return err;
3965 }
3966
3967 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3968 {
3969         struct sk_buff *skb;
3970         struct net *net = dev_net(ifa->idev->dev);
3971         int err = -ENOBUFS;
3972
3973         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3974         if (skb == NULL)
3975                 goto errout;
3976
3977         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3978         if (err < 0) {
3979                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3980                 WARN_ON(err == -EMSGSIZE);
3981                 kfree_skb(skb);
3982                 goto errout;
3983         }
3984         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3985         return;
3986 errout:
3987         if (err < 0)
3988                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3989 }
3990
3991 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3992                                 __s32 *array, int bytes)
3993 {
3994         BUG_ON(bytes < (DEVCONF_MAX * 4));
3995
3996         memset(array, 0, bytes);
3997         array[DEVCONF_FORWARDING] = cnf->forwarding;
3998         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3999         array[DEVCONF_MTU6] = cnf->mtu6;
4000         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4001         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4002         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4003         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4004         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4005         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4006                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4007         array[DEVCONF_RTR_SOLICIT_DELAY] =
4008                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4009         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4010 #ifdef CONFIG_IPV6_PRIVACY
4011         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4012         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4013         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4014         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4015         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4016 #endif
4017         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4018         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4019         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4020 #ifdef CONFIG_IPV6_ROUTER_PREF
4021         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4022         array[DEVCONF_RTR_PROBE_INTERVAL] =
4023                 jiffies_to_msecs(cnf->rtr_probe_interval);
4024 #ifdef CONFIG_IPV6_ROUTE_INFO
4025         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4026 #endif
4027 #endif
4028         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4029         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4030 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4031         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4032 #endif
4033 #ifdef CONFIG_IPV6_MROUTE
4034         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4035 #endif
4036         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4037         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4038         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4039         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4040 }
4041
4042 static inline size_t inet6_ifla6_size(void)
4043 {
4044         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4045              + nla_total_size(sizeof(struct ifla_cacheinfo))
4046              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4047              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4048              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4049 }
4050
4051 static inline size_t inet6_if_nlmsg_size(void)
4052 {
4053         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4054                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4055                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4056                + nla_total_size(4) /* IFLA_MTU */
4057                + nla_total_size(4) /* IFLA_LINK */
4058                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4059 }
4060
4061 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4062                                       int items, int bytes)
4063 {
4064         int i;
4065         int pad = bytes - sizeof(u64) * items;
4066         BUG_ON(pad < 0);
4067
4068         /* Use put_unaligned() because stats may not be aligned for u64. */
4069         put_unaligned(items, &stats[0]);
4070         for (i = 1; i < items; i++)
4071                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4072
4073         memset(&stats[items], 0, pad);
4074 }
4075
4076 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4077                                       int items, int bytes, size_t syncpoff)
4078 {
4079         int i;
4080         int pad = bytes - sizeof(u64) * items;
4081         BUG_ON(pad < 0);
4082
4083         /* Use put_unaligned() because stats may not be aligned for u64. */
4084         put_unaligned(items, &stats[0]);
4085         for (i = 1; i < items; i++)
4086                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4087
4088         memset(&stats[items], 0, pad);
4089 }
4090
4091 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4092                              int bytes)
4093 {
4094         switch (attrtype) {
4095         case IFLA_INET6_STATS:
4096                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4097                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4098                 break;
4099         case IFLA_INET6_ICMP6STATS:
4100                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4101                 break;
4102         }
4103 }
4104
4105 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4106 {
4107         struct nlattr *nla;
4108         struct ifla_cacheinfo ci;
4109
4110         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4111                 goto nla_put_failure;
4112         ci.max_reasm_len = IPV6_MAXPLEN;
4113         ci.tstamp = cstamp_delta(idev->tstamp);
4114         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4115         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4116         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4117                 goto nla_put_failure;
4118         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4119         if (nla == NULL)
4120                 goto nla_put_failure;
4121         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4122
4123         /* XXX - MC not implemented */
4124
4125         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4126         if (nla == NULL)
4127                 goto nla_put_failure;
4128         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4129
4130         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4131         if (nla == NULL)
4132                 goto nla_put_failure;
4133         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4134
4135         return 0;
4136
4137 nla_put_failure:
4138         return -EMSGSIZE;
4139 }
4140
4141 static size_t inet6_get_link_af_size(const struct net_device *dev)
4142 {
4143         if (!__in6_dev_get(dev))
4144                 return 0;
4145
4146         return inet6_ifla6_size();
4147 }
4148
4149 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4150 {
4151         struct inet6_dev *idev = __in6_dev_get(dev);
4152
4153         if (!idev)
4154                 return -ENODATA;
4155
4156         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4157                 return -EMSGSIZE;
4158
4159         return 0;
4160 }
4161
4162 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4163                              u32 portid, u32 seq, int event, unsigned int flags)
4164 {
4165         struct net_device *dev = idev->dev;
4166         struct ifinfomsg *hdr;
4167         struct nlmsghdr *nlh;
4168         void *protoinfo;
4169
4170         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4171         if (nlh == NULL)
4172                 return -EMSGSIZE;
4173
4174         hdr = nlmsg_data(nlh);
4175         hdr->ifi_family = AF_INET6;
4176         hdr->__ifi_pad = 0;
4177         hdr->ifi_type = dev->type;
4178         hdr->ifi_index = dev->ifindex;
4179         hdr->ifi_flags = dev_get_flags(dev);
4180         hdr->ifi_change = 0;
4181
4182         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4183             (dev->addr_len &&
4184              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4185             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4186             (dev->ifindex != dev->iflink &&
4187              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4188                 goto nla_put_failure;
4189         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4190         if (protoinfo == NULL)
4191                 goto nla_put_failure;
4192
4193         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4194                 goto nla_put_failure;
4195
4196         nla_nest_end(skb, protoinfo);
4197         return nlmsg_end(skb, nlh);
4198
4199 nla_put_failure:
4200         nlmsg_cancel(skb, nlh);
4201         return -EMSGSIZE;
4202 }
4203
4204 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4205 {
4206         struct net *net = sock_net(skb->sk);
4207         int h, s_h;
4208         int idx = 0, s_idx;
4209         struct net_device *dev;
4210         struct inet6_dev *idev;
4211         struct hlist_head *head;
4212         struct hlist_node *node;
4213
4214         s_h = cb->args[0];
4215         s_idx = cb->args[1];
4216
4217         rcu_read_lock();
4218         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4219                 idx = 0;
4220                 head = &net->dev_index_head[h];
4221                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4222                         if (idx < s_idx)
4223                                 goto cont;
4224                         idev = __in6_dev_get(dev);
4225                         if (!idev)
4226                                 goto cont;
4227                         if (inet6_fill_ifinfo(skb, idev,
4228                                               NETLINK_CB(cb->skb).portid,
4229                                               cb->nlh->nlmsg_seq,
4230                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4231                                 goto out;
4232 cont:
4233                         idx++;
4234                 }
4235         }
4236 out:
4237         rcu_read_unlock();
4238         cb->args[1] = idx;
4239         cb->args[0] = h;
4240
4241         return skb->len;
4242 }
4243
4244 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4245 {
4246         struct sk_buff *skb;
4247         struct net *net = dev_net(idev->dev);
4248         int err = -ENOBUFS;
4249
4250         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4251         if (skb == NULL)
4252                 goto errout;
4253
4254         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4255         if (err < 0) {
4256                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4257                 WARN_ON(err == -EMSGSIZE);
4258                 kfree_skb(skb);
4259                 goto errout;
4260         }
4261         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4262         return;
4263 errout:
4264         if (err < 0)
4265                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4266 }
4267
4268 static inline size_t inet6_prefix_nlmsg_size(void)
4269 {
4270         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4271                + nla_total_size(sizeof(struct in6_addr))
4272                + nla_total_size(sizeof(struct prefix_cacheinfo));
4273 }
4274
4275 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4276                              struct prefix_info *pinfo, u32 portid, u32 seq,
4277                              int event, unsigned int flags)
4278 {
4279         struct prefixmsg *pmsg;
4280         struct nlmsghdr *nlh;
4281         struct prefix_cacheinfo ci;
4282
4283         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4284         if (nlh == NULL)
4285                 return -EMSGSIZE;
4286
4287         pmsg = nlmsg_data(nlh);
4288         pmsg->prefix_family = AF_INET6;
4289         pmsg->prefix_pad1 = 0;
4290         pmsg->prefix_pad2 = 0;
4291         pmsg->prefix_ifindex = idev->dev->ifindex;
4292         pmsg->prefix_len = pinfo->prefix_len;
4293         pmsg->prefix_type = pinfo->type;
4294         pmsg->prefix_pad3 = 0;
4295         pmsg->prefix_flags = 0;
4296         if (pinfo->onlink)
4297                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4298         if (pinfo->autoconf)
4299                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4300
4301         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4302                 goto nla_put_failure;
4303         ci.preferred_time = ntohl(pinfo->prefered);
4304         ci.valid_time = ntohl(pinfo->valid);
4305         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4306                 goto nla_put_failure;
4307         return nlmsg_end(skb, nlh);
4308
4309 nla_put_failure:
4310         nlmsg_cancel(skb, nlh);
4311         return -EMSGSIZE;
4312 }
4313
4314 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4315                          struct prefix_info *pinfo)
4316 {
4317         struct sk_buff *skb;
4318         struct net *net = dev_net(idev->dev);
4319         int err = -ENOBUFS;
4320
4321         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4322         if (skb == NULL)
4323                 goto errout;
4324
4325         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4326         if (err < 0) {
4327                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4328                 WARN_ON(err == -EMSGSIZE);
4329                 kfree_skb(skb);
4330                 goto errout;
4331         }
4332         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4333         return;
4334 errout:
4335         if (err < 0)
4336                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4337 }
4338
4339 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4340 {
4341         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4342
4343         switch (event) {
4344         case RTM_NEWADDR:
4345                 /*
4346                  * If the address was optimistic
4347                  * we inserted the route at the start of
4348                  * our DAD process, so we don't need
4349                  * to do it again
4350                  */
4351                 if (!(ifp->rt->rt6i_node))
4352                         ip6_ins_rt(ifp->rt);
4353                 if (ifp->idev->cnf.forwarding)
4354                         addrconf_join_anycast(ifp);
4355                 break;
4356         case RTM_DELADDR:
4357                 if (ifp->idev->cnf.forwarding)
4358                         addrconf_leave_anycast(ifp);
4359                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4360                 dst_hold(&ifp->rt->dst);
4361
4362                 if (ip6_del_rt(ifp->rt))
4363                         dst_free(&ifp->rt->dst);
4364                 break;
4365         }
4366 }
4367
4368 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4369 {
4370         rcu_read_lock_bh();
4371         if (likely(ifp->idev->dead == 0))
4372                 __ipv6_ifa_notify(event, ifp);
4373         rcu_read_unlock_bh();
4374 }
4375
4376 #ifdef CONFIG_SYSCTL
4377
4378 static
4379 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4380                            void __user *buffer, size_t *lenp, loff_t *ppos)
4381 {
4382         int *valp = ctl->data;
4383         int val = *valp;
4384         loff_t pos = *ppos;
4385         ctl_table lctl;
4386         int ret;
4387
4388         /*
4389          * ctl->data points to idev->cnf.forwarding, we should
4390          * not modify it until we get the rtnl lock.
4391          */
4392         lctl = *ctl;
4393         lctl.data = &val;
4394
4395         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4396
4397         if (write)
4398                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4399         if (ret)
4400                 *ppos = pos;
4401         return ret;
4402 }
4403
4404 static void dev_disable_change(struct inet6_dev *idev)
4405 {
4406         if (!idev || !idev->dev)
4407                 return;
4408
4409         if (idev->cnf.disable_ipv6)
4410                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4411         else
4412                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4413 }
4414
4415 static void addrconf_disable_change(struct net *net, __s32 newf)
4416 {
4417         struct net_device *dev;
4418         struct inet6_dev *idev;
4419
4420         rcu_read_lock();
4421         for_each_netdev_rcu(net, dev) {
4422                 idev = __in6_dev_get(dev);
4423                 if (idev) {
4424                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4425                         idev->cnf.disable_ipv6 = newf;
4426                         if (changed)
4427                                 dev_disable_change(idev);
4428                 }
4429         }
4430         rcu_read_unlock();
4431 }
4432
4433 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4434 {
4435         struct net *net;
4436         int old;
4437
4438         if (!rtnl_trylock())
4439                 return restart_syscall();
4440
4441         net = (struct net *)table->extra2;
4442         old = *p;
4443         *p = newf;
4444
4445         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4446                 rtnl_unlock();
4447                 return 0;
4448         }
4449
4450         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4451                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4452                 addrconf_disable_change(net, newf);
4453         } else if ((!newf) ^ (!old))
4454                 dev_disable_change((struct inet6_dev *)table->extra1);
4455
4456         rtnl_unlock();
4457         return 0;
4458 }
4459
4460 static
4461 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4462                             void __user *buffer, size_t *lenp, loff_t *ppos)
4463 {
4464         int *valp = ctl->data;
4465         int val = *valp;
4466         loff_t pos = *ppos;
4467         ctl_table lctl;
4468         int ret;
4469
4470         /*
4471          * ctl->data points to idev->cnf.disable_ipv6, we should
4472          * not modify it until we get the rtnl lock.
4473          */
4474         lctl = *ctl;
4475         lctl.data = &val;
4476
4477         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4478
4479         if (write)
4480                 ret = addrconf_disable_ipv6(ctl, valp, val);
4481         if (ret)
4482                 *ppos = pos;
4483         return ret;
4484 }
4485
4486 static struct addrconf_sysctl_table
4487 {
4488         struct ctl_table_header *sysctl_header;
4489         ctl_table addrconf_vars[DEVCONF_MAX+1];
4490 } addrconf_sysctl __read_mostly = {
4491         .sysctl_header = NULL,
4492         .addrconf_vars = {
4493                 {
4494                         .procname       = "forwarding",
4495                         .data           = &ipv6_devconf.forwarding,
4496                         .maxlen         = sizeof(int),
4497                         .mode           = 0644,
4498                         .proc_handler   = addrconf_sysctl_forward,
4499                 },
4500                 {
4501                         .procname       = "hop_limit",
4502                         .data           = &ipv6_devconf.hop_limit,
4503                         .maxlen         = sizeof(int),
4504                         .mode           = 0644,
4505                         .proc_handler   = proc_dointvec,
4506                 },
4507                 {
4508                         .procname       = "mtu",
4509                         .data           = &ipv6_devconf.mtu6,
4510                         .maxlen         = sizeof(int),
4511                         .mode           = 0644,
4512                         .proc_handler   = proc_dointvec,
4513                 },
4514                 {
4515                         .procname       = "accept_ra",
4516                         .data           = &ipv6_devconf.accept_ra,
4517                         .maxlen         = sizeof(int),
4518                         .mode           = 0644,
4519                         .proc_handler   = proc_dointvec,
4520                 },
4521                 {
4522                         .procname       = "accept_redirects",
4523                         .data           = &ipv6_devconf.accept_redirects,
4524                         .maxlen         = sizeof(int),
4525                         .mode           = 0644,
4526                         .proc_handler   = proc_dointvec,
4527                 },
4528                 {
4529                         .procname       = "autoconf",
4530                         .data           = &ipv6_devconf.autoconf,
4531                         .maxlen         = sizeof(int),
4532                         .mode           = 0644,
4533                         .proc_handler   = proc_dointvec,
4534                 },
4535                 {
4536                         .procname       = "dad_transmits",
4537                         .data           = &ipv6_devconf.dad_transmits,
4538                         .maxlen         = sizeof(int),
4539                         .mode           = 0644,
4540                         .proc_handler   = proc_dointvec,
4541                 },
4542                 {
4543                         .procname       = "router_solicitations",
4544                         .data           = &ipv6_devconf.rtr_solicits,
4545                         .maxlen         = sizeof(int),
4546                         .mode           = 0644,
4547                         .proc_handler   = proc_dointvec,
4548                 },
4549                 {
4550                         .procname       = "router_solicitation_interval",
4551                         .data           = &ipv6_devconf.rtr_solicit_interval,
4552                         .maxlen         = sizeof(int),
4553                         .mode           = 0644,
4554                         .proc_handler   = proc_dointvec_jiffies,
4555                 },
4556                 {
4557                         .procname       = "router_solicitation_delay",
4558                         .data           = &ipv6_devconf.rtr_solicit_delay,
4559                         .maxlen         = sizeof(int),
4560                         .mode           = 0644,
4561                         .proc_handler   = proc_dointvec_jiffies,
4562                 },
4563                 {
4564                         .procname       = "force_mld_version",
4565                         .data           = &ipv6_devconf.force_mld_version,
4566                         .maxlen         = sizeof(int),
4567                         .mode           = 0644,
4568                         .proc_handler   = proc_dointvec,
4569                 },
4570 #ifdef CONFIG_IPV6_PRIVACY
4571                 {
4572                         .procname       = "use_tempaddr",
4573                         .data           = &ipv6_devconf.use_tempaddr,
4574                         .maxlen         = sizeof(int),
4575                         .mode           = 0644,
4576                         .proc_handler   = proc_dointvec,
4577                 },
4578                 {
4579                         .procname       = "temp_valid_lft",
4580                         .data           = &ipv6_devconf.temp_valid_lft,
4581                         .maxlen         = sizeof(int),
4582                         .mode           = 0644,
4583                         .proc_handler   = proc_dointvec,
4584                 },
4585                 {
4586                         .procname       = "temp_prefered_lft",
4587                         .data           = &ipv6_devconf.temp_prefered_lft,
4588                         .maxlen         = sizeof(int),
4589                         .mode           = 0644,
4590                         .proc_handler   = proc_dointvec,
4591                 },
4592                 {
4593                         .procname       = "regen_max_retry",
4594                         .data           = &ipv6_devconf.regen_max_retry,
4595                         .maxlen         = sizeof(int),
4596                         .mode           = 0644,
4597                         .proc_handler   = proc_dointvec,
4598                 },
4599                 {
4600                         .procname       = "max_desync_factor",
4601                         .data           = &ipv6_devconf.max_desync_factor,
4602                         .maxlen         = sizeof(int),
4603                         .mode           = 0644,
4604                         .proc_handler   = proc_dointvec,
4605                 },
4606 #endif
4607                 {
4608                         .procname       = "max_addresses",
4609                         .data           = &ipv6_devconf.max_addresses,
4610                         .maxlen         = sizeof(int),
4611                         .mode           = 0644,
4612                         .proc_handler   = proc_dointvec,
4613                 },
4614                 {
4615                         .procname       = "accept_ra_defrtr",
4616                         .data           = &ipv6_devconf.accept_ra_defrtr,
4617                         .maxlen         = sizeof(int),
4618                         .mode           = 0644,
4619                         .proc_handler   = proc_dointvec,
4620                 },
4621                 {
4622                         .procname       = "accept_ra_pinfo",
4623                         .data           = &ipv6_devconf.accept_ra_pinfo,
4624                         .maxlen         = sizeof(int),
4625                         .mode           = 0644,
4626                         .proc_handler   = proc_dointvec,
4627                 },
4628 #ifdef CONFIG_IPV6_ROUTER_PREF
4629                 {
4630                         .procname       = "accept_ra_rtr_pref",
4631                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4632                         .maxlen         = sizeof(int),
4633                         .mode           = 0644,
4634                         .proc_handler   = proc_dointvec,
4635                 },
4636                 {
4637                         .procname       = "router_probe_interval",
4638                         .data           = &ipv6_devconf.rtr_probe_interval,
4639                         .maxlen         = sizeof(int),
4640                         .mode           = 0644,
4641                         .proc_handler   = proc_dointvec_jiffies,
4642                 },
4643 #ifdef CONFIG_IPV6_ROUTE_INFO
4644                 {
4645                         .procname       = "accept_ra_rt_info_max_plen",
4646                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4647                         .maxlen         = sizeof(int),
4648                         .mode           = 0644,
4649                         .proc_handler   = proc_dointvec,
4650                 },
4651 #endif
4652 #endif
4653                 {
4654                         .procname       = "proxy_ndp",
4655                         .data           = &ipv6_devconf.proxy_ndp,
4656                         .maxlen         = sizeof(int),
4657                         .mode           = 0644,
4658                         .proc_handler   = proc_dointvec,
4659                 },
4660                 {
4661                         .procname       = "accept_source_route",
4662                         .data           = &ipv6_devconf.accept_source_route,
4663                         .maxlen         = sizeof(int),
4664                         .mode           = 0644,
4665                         .proc_handler   = proc_dointvec,
4666                 },
4667 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4668                 {
4669                         .procname       = "optimistic_dad",
4670                         .data           = &ipv6_devconf.optimistic_dad,
4671                         .maxlen         = sizeof(int),
4672                         .mode           = 0644,
4673                         .proc_handler   = proc_dointvec,
4674
4675                 },
4676 #endif
4677 #ifdef CONFIG_IPV6_MROUTE
4678                 {
4679                         .procname       = "mc_forwarding",
4680                         .data           = &ipv6_devconf.mc_forwarding,
4681                         .maxlen         = sizeof(int),
4682                         .mode           = 0444,
4683                         .proc_handler   = proc_dointvec,
4684                 },
4685 #endif
4686                 {
4687                         .procname       = "disable_ipv6",
4688                         .data           = &ipv6_devconf.disable_ipv6,
4689                         .maxlen         = sizeof(int),
4690                         .mode           = 0644,
4691                         .proc_handler   = addrconf_sysctl_disable,
4692                 },
4693                 {
4694                         .procname       = "accept_dad",
4695                         .data           = &ipv6_devconf.accept_dad,
4696                         .maxlen         = sizeof(int),
4697                         .mode           = 0644,
4698                         .proc_handler   = proc_dointvec,
4699                 },
4700                 {
4701                         .procname       = "force_tllao",
4702                         .data           = &ipv6_devconf.force_tllao,
4703                         .maxlen         = sizeof(int),
4704                         .mode           = 0644,
4705                         .proc_handler   = proc_dointvec
4706                 },
4707                 {
4708                         .procname       = "ndisc_notify",
4709                         .data           = &ipv6_devconf.ndisc_notify,
4710                         .maxlen         = sizeof(int),
4711                         .mode           = 0644,
4712                         .proc_handler   = proc_dointvec
4713                 },
4714                 {
4715                         /* sentinel */
4716                 }
4717         },
4718 };
4719
4720 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4721                 struct inet6_dev *idev, struct ipv6_devconf *p)
4722 {
4723         int i;
4724         struct addrconf_sysctl_table *t;
4725         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4726
4727         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4728         if (t == NULL)
4729                 goto out;
4730
4731         for (i = 0; t->addrconf_vars[i].data; i++) {
4732                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4733                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4734                 t->addrconf_vars[i].extra2 = net;
4735         }
4736
4737         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4738
4739         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4740         if (t->sysctl_header == NULL)
4741                 goto free;
4742
4743         p->sysctl = t;
4744         return 0;
4745
4746 free:
4747         kfree(t);
4748 out:
4749         return -ENOBUFS;
4750 }
4751
4752 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4753 {
4754         struct addrconf_sysctl_table *t;
4755
4756         if (p->sysctl == NULL)
4757                 return;
4758
4759         t = p->sysctl;
4760         p->sysctl = NULL;
4761         unregister_net_sysctl_table(t->sysctl_header);
4762         kfree(t);
4763 }
4764
4765 static void addrconf_sysctl_register(struct inet6_dev *idev)
4766 {
4767         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4768                               &ndisc_ifinfo_sysctl_change);
4769         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4770                                         idev, &idev->cnf);
4771 }
4772
4773 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4774 {
4775         __addrconf_sysctl_unregister(&idev->cnf);
4776         neigh_sysctl_unregister(idev->nd_parms);
4777 }
4778
4779
4780 #endif
4781
4782 static int __net_init addrconf_init_net(struct net *net)
4783 {
4784         int err;
4785         struct ipv6_devconf *all, *dflt;
4786
4787         err = -ENOMEM;
4788         all = &ipv6_devconf;
4789         dflt = &ipv6_devconf_dflt;
4790
4791         if (!net_eq(net, &init_net)) {
4792                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4793                 if (all == NULL)
4794                         goto err_alloc_all;
4795
4796                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4797                 if (dflt == NULL)
4798                         goto err_alloc_dflt;
4799         } else {
4800                 /* these will be inherited by all namespaces */
4801                 dflt->autoconf = ipv6_defaults.autoconf;
4802                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4803         }
4804
4805         net->ipv6.devconf_all = all;
4806         net->ipv6.devconf_dflt = dflt;
4807
4808 #ifdef CONFIG_SYSCTL
4809         err = __addrconf_sysctl_register(net, "all", NULL, all);
4810         if (err < 0)
4811                 goto err_reg_all;
4812
4813         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4814         if (err < 0)
4815                 goto err_reg_dflt;
4816 #endif
4817         return 0;
4818
4819 #ifdef CONFIG_SYSCTL
4820 err_reg_dflt:
4821         __addrconf_sysctl_unregister(all);
4822 err_reg_all:
4823         kfree(dflt);
4824 #endif
4825 err_alloc_dflt:
4826         kfree(all);
4827 err_alloc_all:
4828         return err;
4829 }
4830
4831 static void __net_exit addrconf_exit_net(struct net *net)
4832 {
4833 #ifdef CONFIG_SYSCTL
4834         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4835         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4836 #endif
4837         if (!net_eq(net, &init_net)) {
4838                 kfree(net->ipv6.devconf_dflt);
4839                 kfree(net->ipv6.devconf_all);
4840         }
4841 }
4842
4843 static struct pernet_operations addrconf_ops = {
4844         .init = addrconf_init_net,
4845         .exit = addrconf_exit_net,
4846 };
4847
4848 /*
4849  *      Device notifier
4850  */
4851
4852 int register_inet6addr_notifier(struct notifier_block *nb)
4853 {
4854         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4855 }
4856 EXPORT_SYMBOL(register_inet6addr_notifier);
4857
4858 int unregister_inet6addr_notifier(struct notifier_block *nb)
4859 {
4860         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4861 }
4862 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4863
4864 static struct rtnl_af_ops inet6_ops = {
4865         .family           = AF_INET6,
4866         .fill_link_af     = inet6_fill_link_af,
4867         .get_link_af_size = inet6_get_link_af_size,
4868 };
4869
4870 /*
4871  *      Init / cleanup code
4872  */
4873
4874 int __init addrconf_init(void)
4875 {
4876         int i, err;
4877
4878         err = ipv6_addr_label_init();
4879         if (err < 0) {
4880                 pr_crit("%s: cannot initialize default policy table: %d\n",
4881                         __func__, err);
4882                 goto out;
4883         }
4884
4885         err = register_pernet_subsys(&addrconf_ops);
4886         if (err < 0)
4887                 goto out_addrlabel;
4888
4889         /* The addrconf netdev notifier requires that loopback_dev
4890          * has it's ipv6 private information allocated and setup
4891          * before it can bring up and give link-local addresses
4892          * to other devices which are up.
4893          *
4894          * Unfortunately, loopback_dev is not necessarily the first
4895          * entry in the global dev_base list of net devices.  In fact,
4896          * it is likely to be the very last entry on that list.
4897          * So this causes the notifier registry below to try and
4898          * give link-local addresses to all devices besides loopback_dev
4899          * first, then loopback_dev, which cases all the non-loopback_dev
4900          * devices to fail to get a link-local address.
4901          *
4902          * So, as a temporary fix, allocate the ipv6 structure for
4903          * loopback_dev first by hand.
4904          * Longer term, all of the dependencies ipv6 has upon the loopback
4905          * device and it being up should be removed.
4906          */
4907         rtnl_lock();
4908         if (!ipv6_add_dev(init_net.loopback_dev))
4909                 err = -ENOMEM;
4910         rtnl_unlock();
4911         if (err)
4912                 goto errlo;
4913
4914         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4915                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4916
4917         register_netdevice_notifier(&ipv6_dev_notf);
4918
4919         addrconf_verify(0);
4920
4921         err = rtnl_af_register(&inet6_ops);
4922         if (err < 0)
4923                 goto errout_af;
4924
4925         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4926                               NULL);
4927         if (err < 0)
4928                 goto errout;
4929
4930         /* Only the first call to __rtnl_register can fail */
4931         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4932         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4933         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4934                         inet6_dump_ifaddr, NULL);
4935         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4936                         inet6_dump_ifmcaddr, NULL);
4937         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4938                         inet6_dump_ifacaddr, NULL);
4939         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4940                         NULL, NULL);
4941
4942         ipv6_addr_label_rtnl_register();
4943
4944         return 0;
4945 errout:
4946         rtnl_af_unregister(&inet6_ops);
4947 errout_af:
4948         unregister_netdevice_notifier(&ipv6_dev_notf);
4949 errlo:
4950         unregister_pernet_subsys(&addrconf_ops);
4951 out_addrlabel:
4952         ipv6_addr_label_cleanup();
4953 out:
4954         return err;
4955 }
4956
4957 void addrconf_cleanup(void)
4958 {
4959         struct net_device *dev;
4960         int i;
4961
4962         unregister_netdevice_notifier(&ipv6_dev_notf);
4963         unregister_pernet_subsys(&addrconf_ops);
4964         ipv6_addr_label_cleanup();
4965
4966         rtnl_lock();
4967
4968         __rtnl_af_unregister(&inet6_ops);
4969
4970         /* clean dev list */
4971         for_each_netdev(&init_net, dev) {
4972                 if (__in6_dev_get(dev) == NULL)
4973                         continue;
4974                 addrconf_ifdown(dev, 1);
4975         }
4976         addrconf_ifdown(init_net.loopback_dev, 2);
4977
4978         /*
4979          *      Check hash table.
4980          */
4981         spin_lock_bh(&addrconf_hash_lock);
4982         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4983                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4984         spin_unlock_bh(&addrconf_hash_lock);
4985
4986         del_timer(&addr_chk_timer);
4987         rtnl_unlock();
4988 }