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