]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - net/ipv6/addrconf.c
ced58e1d91b6593a4e6df6259b6c8c77127a7c60
[karo-tx-linux.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
88 #endif
89
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
99
100 #if ACONF_DEBUG >= 3
101 #define ADBG(x) printk x
102 #else
103 #define ADBG(x)
104 #endif
105
106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
107
108 static inline u32 cstamp_delta(unsigned long cstamp)
109 {
110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 }
112
113 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
114 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
115 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
116
117 #ifdef CONFIG_SYSCTL
118 static void addrconf_sysctl_register(struct inet6_dev *idev);
119 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
120 #else
121 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
122 {
123 }
124
125 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
126 {
127 }
128 #endif
129
130 #ifdef CONFIG_IPV6_PRIVACY
131 static void __ipv6_regen_rndid(struct inet6_dev *idev);
132 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
133 static void ipv6_regen_rndid(unsigned long data);
134 #endif
135
136 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
137 static int ipv6_count_addresses(struct inet6_dev *idev);
138
139 /*
140  *      Configured unicast address hash table
141  */
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
144
145 static void addrconf_verify(unsigned long);
146
147 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
148 static DEFINE_SPINLOCK(addrconf_verify_lock);
149
150 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
151 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
152
153 static void addrconf_type_change(struct net_device *dev,
154                                  unsigned long event);
155 static int addrconf_ifdown(struct net_device *dev, int how);
156
157 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_timer(unsigned long data);
159 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_run(struct inet6_dev *idev);
161 static void addrconf_rs_timer(unsigned long data);
162 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164
165 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
166                                 struct prefix_info *pinfo);
167 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
168                                struct net_device *dev);
169
170 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
171
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
173         .forwarding             = 0,
174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
175         .mtu6                   = IPV6_MIN_MTU,
176         .accept_ra              = 1,
177         .accept_redirects       = 1,
178         .autoconf               = 1,
179         .force_mld_version      = 0,
180         .dad_transmits          = 1,
181         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
182         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
183         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
184 #ifdef CONFIG_IPV6_PRIVACY
185         .use_tempaddr           = 0,
186         .temp_valid_lft         = TEMP_VALID_LIFETIME,
187         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
188         .regen_max_retry        = REGEN_MAX_RETRY,
189         .max_desync_factor      = MAX_DESYNC_FACTOR,
190 #endif
191         .max_addresses          = IPV6_MAX_ADDRESSES,
192         .accept_ra_defrtr       = 1,
193         .accept_ra_pinfo        = 1,
194 #ifdef CONFIG_IPV6_ROUTER_PREF
195         .accept_ra_rtr_pref     = 1,
196         .rtr_probe_interval     = 60 * HZ,
197 #ifdef CONFIG_IPV6_ROUTE_INFO
198         .accept_ra_rt_info_max_plen = 0,
199 #endif
200 #endif
201         .proxy_ndp              = 0,
202         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
203         .disable_ipv6           = 0,
204         .accept_dad             = 1,
205 };
206
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
208         .forwarding             = 0,
209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
210         .mtu6                   = IPV6_MIN_MTU,
211         .accept_ra              = 1,
212         .accept_redirects       = 1,
213         .autoconf               = 1,
214         .dad_transmits          = 1,
215         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
216         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
217         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
218 #ifdef CONFIG_IPV6_PRIVACY
219         .use_tempaddr           = 0,
220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
222         .regen_max_retry        = REGEN_MAX_RETRY,
223         .max_desync_factor      = MAX_DESYNC_FACTOR,
224 #endif
225         .max_addresses          = IPV6_MAX_ADDRESSES,
226         .accept_ra_defrtr       = 1,
227         .accept_ra_pinfo        = 1,
228 #ifdef CONFIG_IPV6_ROUTER_PREF
229         .accept_ra_rtr_pref     = 1,
230         .rtr_probe_interval     = 60 * HZ,
231 #ifdef CONFIG_IPV6_ROUTE_INFO
232         .accept_ra_rt_info_max_plen = 0,
233 #endif
234 #endif
235         .proxy_ndp              = 0,
236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
237         .disable_ipv6           = 0,
238         .accept_dad             = 1,
239 };
240
241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
246
247 /* Check if a valid qdisc is available */
248 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
249 {
250         return !qdisc_tx_is_noop(dev);
251 }
252
253 /* Check if a route is valid prefix route */
254 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
255 {
256         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
257 }
258
259 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
260 {
261         if (del_timer(&ifp->timer))
262                 __in6_ifa_put(ifp);
263 }
264
265 enum addrconf_timer_t {
266         AC_NONE,
267         AC_DAD,
268         AC_RS,
269 };
270
271 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
272                                enum addrconf_timer_t what,
273                                unsigned long when)
274 {
275         if (!del_timer(&ifp->timer))
276                 in6_ifa_hold(ifp);
277
278         switch (what) {
279         case AC_DAD:
280                 ifp->timer.function = addrconf_dad_timer;
281                 break;
282         case AC_RS:
283                 ifp->timer.function = addrconf_rs_timer;
284                 break;
285         default:
286                 break;
287         }
288         ifp->timer.expires = jiffies + when;
289         add_timer(&ifp->timer);
290 }
291
292 static int snmp6_alloc_dev(struct inet6_dev *idev)
293 {
294         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
295                           sizeof(struct ipstats_mib),
296                           __alignof__(struct ipstats_mib)) < 0)
297                 goto err_ip;
298         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
299                                         GFP_KERNEL);
300         if (!idev->stats.icmpv6dev)
301                 goto err_icmp;
302         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
303                                            GFP_KERNEL);
304         if (!idev->stats.icmpv6msgdev)
305                 goto err_icmpmsg;
306
307         return 0;
308
309 err_icmpmsg:
310         kfree(idev->stats.icmpv6dev);
311 err_icmp:
312         snmp_mib_free((void __percpu **)idev->stats.ipv6);
313 err_ip:
314         return -ENOMEM;
315 }
316
317 static void snmp6_free_dev(struct inet6_dev *idev)
318 {
319         kfree(idev->stats.icmpv6msgdev);
320         kfree(idev->stats.icmpv6dev);
321         snmp_mib_free((void __percpu **)idev->stats.ipv6);
322 }
323
324 /* Nobody refers to this device, we may destroy it. */
325
326 void in6_dev_finish_destroy(struct inet6_dev *idev)
327 {
328         struct net_device *dev = idev->dev;
329
330         WARN_ON(!list_empty(&idev->addr_list));
331         WARN_ON(idev->mc_list != NULL);
332
333 #ifdef NET_REFCNT_DEBUG
334         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
335 #endif
336         dev_put(dev);
337         if (!idev->dead) {
338                 pr_warn("Freeing alive inet6 device %p\n", idev);
339                 return;
340         }
341         snmp6_free_dev(idev);
342         kfree_rcu(idev, rcu);
343 }
344 EXPORT_SYMBOL(in6_dev_finish_destroy);
345
346 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
347 {
348         struct inet6_dev *ndev;
349
350         ASSERT_RTNL();
351
352         if (dev->mtu < IPV6_MIN_MTU)
353                 return NULL;
354
355         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
356
357         if (ndev == NULL)
358                 return NULL;
359
360         rwlock_init(&ndev->lock);
361         ndev->dev = dev;
362         INIT_LIST_HEAD(&ndev->addr_list);
363
364         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
365         ndev->cnf.mtu6 = dev->mtu;
366         ndev->cnf.sysctl = NULL;
367         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
368         if (ndev->nd_parms == NULL) {
369                 kfree(ndev);
370                 return NULL;
371         }
372         if (ndev->cnf.forwarding)
373                 dev_disable_lro(dev);
374         /* We refer to the device */
375         dev_hold(dev);
376
377         if (snmp6_alloc_dev(ndev) < 0) {
378                 ADBG((KERN_WARNING
379                         "%s: cannot allocate memory for statistics; dev=%s.\n",
380                         __func__, dev->name));
381                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
382                 dev_put(dev);
383                 kfree(ndev);
384                 return NULL;
385         }
386
387         if (snmp6_register_dev(ndev) < 0) {
388                 ADBG((KERN_WARNING
389                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
390                         __func__, dev->name));
391                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
392                 ndev->dead = 1;
393                 in6_dev_finish_destroy(ndev);
394                 return NULL;
395         }
396
397         /* One reference from device.  We must do this before
398          * we invoke __ipv6_regen_rndid().
399          */
400         in6_dev_hold(ndev);
401
402         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
403                 ndev->cnf.accept_dad = -1;
404
405 #if IS_ENABLED(CONFIG_IPV6_SIT)
406         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
407                 pr_info("%s: Disabled Multicast RS\n", dev->name);
408                 ndev->cnf.rtr_solicits = 0;
409         }
410 #endif
411
412 #ifdef CONFIG_IPV6_PRIVACY
413         INIT_LIST_HEAD(&ndev->tempaddr_list);
414         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
415         if ((dev->flags&IFF_LOOPBACK) ||
416             dev->type == ARPHRD_TUNNEL ||
417             dev->type == ARPHRD_TUNNEL6 ||
418             dev->type == ARPHRD_SIT ||
419             dev->type == ARPHRD_NONE) {
420                 ndev->cnf.use_tempaddr = -1;
421         } else {
422                 in6_dev_hold(ndev);
423                 ipv6_regen_rndid((unsigned long) ndev);
424         }
425 #endif
426
427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
428                 ndev->if_flags |= IF_READY;
429
430         ipv6_mc_init_dev(ndev);
431         ndev->tstamp = jiffies;
432         addrconf_sysctl_register(ndev);
433         /* protected by rtnl_lock */
434         rcu_assign_pointer(dev->ip6_ptr, ndev);
435
436         /* Join all-node multicast group */
437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
438
439         /* Join all-router multicast group if forwarding is set */
440         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442
443         return ndev;
444 }
445
446 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
447 {
448         struct inet6_dev *idev;
449
450         ASSERT_RTNL();
451
452         idev = __in6_dev_get(dev);
453         if (!idev) {
454                 idev = ipv6_add_dev(dev);
455                 if (!idev)
456                         return NULL;
457         }
458
459         if (dev->flags&IFF_UP)
460                 ipv6_mc_up(idev);
461         return idev;
462 }
463
464 static int inet6_netconf_msgsize_devconf(int type)
465 {
466         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
467                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
468
469         /* type -1 is used for ALL */
470         if (type == -1 || type == NETCONFA_FORWARDING)
471                 size += nla_total_size(4);
472
473         return size;
474 }
475
476 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
477                                       struct ipv6_devconf *devconf, u32 portid,
478                                       u32 seq, int event, unsigned int flags,
479                                       int type)
480 {
481         struct nlmsghdr  *nlh;
482         struct netconfmsg *ncm;
483
484         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
485                         flags);
486         if (nlh == NULL)
487                 return -EMSGSIZE;
488
489         ncm = nlmsg_data(nlh);
490         ncm->ncm_family = AF_INET6;
491
492         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
493                 goto nla_put_failure;
494
495         /* type -1 is used for ALL */
496         if ((type == -1 || type == NETCONFA_FORWARDING) &&
497             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
498                 goto nla_put_failure;
499
500         return nlmsg_end(skb, nlh);
501
502 nla_put_failure:
503         nlmsg_cancel(skb, nlh);
504         return -EMSGSIZE;
505 }
506
507 static void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
508                                          struct ipv6_devconf *devconf)
509 {
510         struct sk_buff *skb;
511         int err = -ENOBUFS;
512
513         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
514         if (skb == NULL)
515                 goto errout;
516
517         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
518                                          RTM_NEWNETCONF, 0, type);
519         if (err < 0) {
520                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
521                 WARN_ON(err == -EMSGSIZE);
522                 kfree_skb(skb);
523                 goto errout;
524         }
525         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
526         return;
527 errout:
528         if (err < 0)
529                 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
530 }
531
532 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
533         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
534         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
535 };
536
537 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
538                                      struct nlmsghdr *nlh,
539                                      void *arg)
540 {
541         struct net *net = sock_net(in_skb->sk);
542         struct nlattr *tb[NETCONFA_MAX+1];
543         struct netconfmsg *ncm;
544         struct sk_buff *skb;
545         struct ipv6_devconf *devconf;
546         struct inet6_dev *in6_dev;
547         struct net_device *dev;
548         int ifindex;
549         int err;
550
551         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
552                           devconf_ipv6_policy);
553         if (err < 0)
554                 goto errout;
555
556         err = EINVAL;
557         if (!tb[NETCONFA_IFINDEX])
558                 goto errout;
559
560         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
561         switch (ifindex) {
562         case NETCONFA_IFINDEX_ALL:
563                 devconf = net->ipv6.devconf_all;
564                 break;
565         case NETCONFA_IFINDEX_DEFAULT:
566                 devconf = net->ipv6.devconf_dflt;
567                 break;
568         default:
569                 dev = __dev_get_by_index(net, ifindex);
570                 if (dev == NULL)
571                         goto errout;
572                 in6_dev = __in6_dev_get(dev);
573                 if (in6_dev == NULL)
574                         goto errout;
575                 devconf = &in6_dev->cnf;
576                 break;
577         }
578
579         err = -ENOBUFS;
580         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
581         if (skb == NULL)
582                 goto errout;
583
584         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
585                                          NETLINK_CB(in_skb).portid,
586                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
587                                          -1);
588         if (err < 0) {
589                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
590                 WARN_ON(err == -EMSGSIZE);
591                 kfree_skb(skb);
592                 goto errout;
593         }
594         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
595 errout:
596         return err;
597 }
598
599 #ifdef CONFIG_SYSCTL
600 static void dev_forward_change(struct inet6_dev *idev)
601 {
602         struct net_device *dev;
603         struct inet6_ifaddr *ifa;
604
605         if (!idev)
606                 return;
607         dev = idev->dev;
608         if (idev->cnf.forwarding)
609                 dev_disable_lro(dev);
610         if (dev->flags & IFF_MULTICAST) {
611                 if (idev->cnf.forwarding)
612                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
613                 else
614                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
615         }
616
617         list_for_each_entry(ifa, &idev->addr_list, if_list) {
618                 if (ifa->flags&IFA_F_TENTATIVE)
619                         continue;
620                 if (idev->cnf.forwarding)
621                         addrconf_join_anycast(ifa);
622                 else
623                         addrconf_leave_anycast(ifa);
624         }
625         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
626                                      dev->ifindex, &idev->cnf);
627 }
628
629
630 static void addrconf_forward_change(struct net *net, __s32 newf)
631 {
632         struct net_device *dev;
633         struct inet6_dev *idev;
634
635         for_each_netdev(net, dev) {
636                 idev = __in6_dev_get(dev);
637                 if (idev) {
638                         int changed = (!idev->cnf.forwarding) ^ (!newf);
639                         idev->cnf.forwarding = newf;
640                         if (changed)
641                                 dev_forward_change(idev);
642                 }
643         }
644 }
645
646 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
647 {
648         struct net *net;
649         int old;
650
651         if (!rtnl_trylock())
652                 return restart_syscall();
653
654         net = (struct net *)table->extra2;
655         old = *p;
656         *p = newf;
657
658         if (p == &net->ipv6.devconf_dflt->forwarding) {
659                 if ((!newf) ^ (!old))
660                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
661                                                      NETCONFA_IFINDEX_DEFAULT,
662                                                      net->ipv6.devconf_dflt);
663                 rtnl_unlock();
664                 return 0;
665         }
666
667         if (p == &net->ipv6.devconf_all->forwarding) {
668                 net->ipv6.devconf_dflt->forwarding = newf;
669                 addrconf_forward_change(net, newf);
670                 if ((!newf) ^ (!old))
671                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
672                                                      NETCONFA_IFINDEX_ALL,
673                                                      net->ipv6.devconf_all);
674         } else if ((!newf) ^ (!old))
675                 dev_forward_change((struct inet6_dev *)table->extra1);
676         rtnl_unlock();
677
678         if (newf)
679                 rt6_purge_dflt_routers(net);
680         return 1;
681 }
682 #endif
683
684 /* Nobody refers to this ifaddr, destroy it */
685 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
686 {
687         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
688
689 #ifdef NET_REFCNT_DEBUG
690         pr_debug("%s\n", __func__);
691 #endif
692
693         in6_dev_put(ifp->idev);
694
695         if (del_timer(&ifp->timer))
696                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
697
698         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
699                 pr_warn("Freeing alive inet6 address %p\n", ifp);
700                 return;
701         }
702         dst_release(&ifp->rt->dst);
703
704         kfree_rcu(ifp, rcu);
705 }
706
707 static void
708 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
709 {
710         struct list_head *p;
711         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
712
713         /*
714          * Each device address list is sorted in order of scope -
715          * global before linklocal.
716          */
717         list_for_each(p, &idev->addr_list) {
718                 struct inet6_ifaddr *ifa
719                         = list_entry(p, struct inet6_ifaddr, if_list);
720                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
721                         break;
722         }
723
724         list_add_tail(&ifp->if_list, p);
725 }
726
727 static u32 inet6_addr_hash(const struct in6_addr *addr)
728 {
729         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
730 }
731
732 /* On success it returns ifp with increased reference count */
733
734 static struct inet6_ifaddr *
735 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
736               int scope, u32 flags)
737 {
738         struct inet6_ifaddr *ifa = NULL;
739         struct rt6_info *rt;
740         unsigned int hash;
741         int err = 0;
742         int addr_type = ipv6_addr_type(addr);
743
744         if (addr_type == IPV6_ADDR_ANY ||
745             addr_type & IPV6_ADDR_MULTICAST ||
746             (!(idev->dev->flags & IFF_LOOPBACK) &&
747              addr_type & IPV6_ADDR_LOOPBACK))
748                 return ERR_PTR(-EADDRNOTAVAIL);
749
750         rcu_read_lock_bh();
751         if (idev->dead) {
752                 err = -ENODEV;                  /*XXX*/
753                 goto out2;
754         }
755
756         if (idev->cnf.disable_ipv6) {
757                 err = -EACCES;
758                 goto out2;
759         }
760
761         spin_lock(&addrconf_hash_lock);
762
763         /* Ignore adding duplicate addresses on an interface */
764         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
765                 ADBG(("ipv6_add_addr: already assigned\n"));
766                 err = -EEXIST;
767                 goto out;
768         }
769
770         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
771
772         if (ifa == NULL) {
773                 ADBG(("ipv6_add_addr: malloc failed\n"));
774                 err = -ENOBUFS;
775                 goto out;
776         }
777
778         rt = addrconf_dst_alloc(idev, addr, false);
779         if (IS_ERR(rt)) {
780                 err = PTR_ERR(rt);
781                 goto out;
782         }
783
784         ifa->addr = *addr;
785
786         spin_lock_init(&ifa->lock);
787         spin_lock_init(&ifa->state_lock);
788         init_timer(&ifa->timer);
789         INIT_HLIST_NODE(&ifa->addr_lst);
790         ifa->timer.data = (unsigned long) ifa;
791         ifa->scope = scope;
792         ifa->prefix_len = pfxlen;
793         ifa->flags = flags | IFA_F_TENTATIVE;
794         ifa->cstamp = ifa->tstamp = jiffies;
795
796         ifa->rt = rt;
797
798         ifa->idev = idev;
799         in6_dev_hold(idev);
800         /* For caller */
801         in6_ifa_hold(ifa);
802
803         /* Add to big hash table */
804         hash = inet6_addr_hash(addr);
805
806         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
807         spin_unlock(&addrconf_hash_lock);
808
809         write_lock(&idev->lock);
810         /* Add to inet6_dev unicast addr list. */
811         ipv6_link_dev_addr(idev, ifa);
812
813 #ifdef CONFIG_IPV6_PRIVACY
814         if (ifa->flags&IFA_F_TEMPORARY) {
815                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
816                 in6_ifa_hold(ifa);
817         }
818 #endif
819
820         in6_ifa_hold(ifa);
821         write_unlock(&idev->lock);
822 out2:
823         rcu_read_unlock_bh();
824
825         if (likely(err == 0))
826                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
827         else {
828                 kfree(ifa);
829                 ifa = ERR_PTR(err);
830         }
831
832         return ifa;
833 out:
834         spin_unlock(&addrconf_hash_lock);
835         goto out2;
836 }
837
838 /* This function wants to get referenced ifp and releases it before return */
839
840 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
841 {
842         struct inet6_ifaddr *ifa, *ifn;
843         struct inet6_dev *idev = ifp->idev;
844         int state;
845         int deleted = 0, onlink = 0;
846         unsigned long expires = jiffies;
847
848         spin_lock_bh(&ifp->state_lock);
849         state = ifp->state;
850         ifp->state = INET6_IFADDR_STATE_DEAD;
851         spin_unlock_bh(&ifp->state_lock);
852
853         if (state == INET6_IFADDR_STATE_DEAD)
854                 goto out;
855
856         spin_lock_bh(&addrconf_hash_lock);
857         hlist_del_init_rcu(&ifp->addr_lst);
858         spin_unlock_bh(&addrconf_hash_lock);
859
860         write_lock_bh(&idev->lock);
861 #ifdef CONFIG_IPV6_PRIVACY
862         if (ifp->flags&IFA_F_TEMPORARY) {
863                 list_del(&ifp->tmp_list);
864                 if (ifp->ifpub) {
865                         in6_ifa_put(ifp->ifpub);
866                         ifp->ifpub = NULL;
867                 }
868                 __in6_ifa_put(ifp);
869         }
870 #endif
871
872         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
873                 if (ifa == ifp) {
874                         list_del_init(&ifp->if_list);
875                         __in6_ifa_put(ifp);
876
877                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
878                                 break;
879                         deleted = 1;
880                         continue;
881                 } else if (ifp->flags & IFA_F_PERMANENT) {
882                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
883                                               ifp->prefix_len)) {
884                                 if (ifa->flags & IFA_F_PERMANENT) {
885                                         onlink = 1;
886                                         if (deleted)
887                                                 break;
888                                 } else {
889                                         unsigned long lifetime;
890
891                                         if (!onlink)
892                                                 onlink = -1;
893
894                                         spin_lock(&ifa->lock);
895
896                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
897                                         /*
898                                          * Note: Because this address is
899                                          * not permanent, lifetime <
900                                          * LONG_MAX / HZ here.
901                                          */
902                                         if (time_before(expires,
903                                                         ifa->tstamp + lifetime * HZ))
904                                                 expires = ifa->tstamp + lifetime * HZ;
905                                         spin_unlock(&ifa->lock);
906                                 }
907                         }
908                 }
909         }
910         write_unlock_bh(&idev->lock);
911
912         addrconf_del_timer(ifp);
913
914         ipv6_ifa_notify(RTM_DELADDR, ifp);
915
916         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
917
918         /*
919          * Purge or update corresponding prefix
920          *
921          * 1) we don't purge prefix here if address was not permanent.
922          *    prefix is managed by its own lifetime.
923          * 2) if there're no addresses, delete prefix.
924          * 3) if there're still other permanent address(es),
925          *    corresponding prefix is still permanent.
926          * 4) otherwise, update prefix lifetime to the
927          *    longest valid lifetime among the corresponding
928          *    addresses on the device.
929          *    Note: subsequent RA will update lifetime.
930          *
931          * --yoshfuji
932          */
933         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
934                 struct in6_addr prefix;
935                 struct rt6_info *rt;
936                 struct net *net = dev_net(ifp->idev->dev);
937                 struct flowi6 fl6 = {};
938
939                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
940                 fl6.flowi6_oif = ifp->idev->dev->ifindex;
941                 fl6.daddr = prefix;
942                 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
943                                                          RT6_LOOKUP_F_IFACE);
944
945                 if (rt != net->ipv6.ip6_null_entry &&
946                     addrconf_is_prefix_route(rt)) {
947                         if (onlink == 0) {
948                                 ip6_del_rt(rt);
949                                 rt = NULL;
950                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
951                                 rt6_set_expires(rt, expires);
952                         }
953                 }
954                 dst_release(&rt->dst);
955         }
956
957         /* clean up prefsrc entries */
958         rt6_remove_prefsrc(ifp);
959 out:
960         in6_ifa_put(ifp);
961 }
962
963 #ifdef CONFIG_IPV6_PRIVACY
964 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
965 {
966         struct inet6_dev *idev = ifp->idev;
967         struct in6_addr addr, *tmpaddr;
968         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
969         unsigned long regen_advance;
970         int tmp_plen;
971         int ret = 0;
972         int max_addresses;
973         u32 addr_flags;
974         unsigned long now = jiffies;
975
976         write_lock(&idev->lock);
977         if (ift) {
978                 spin_lock_bh(&ift->lock);
979                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
980                 spin_unlock_bh(&ift->lock);
981                 tmpaddr = &addr;
982         } else {
983                 tmpaddr = NULL;
984         }
985 retry:
986         in6_dev_hold(idev);
987         if (idev->cnf.use_tempaddr <= 0) {
988                 write_unlock(&idev->lock);
989                 pr_info("%s: use_tempaddr is disabled\n", __func__);
990                 in6_dev_put(idev);
991                 ret = -1;
992                 goto out;
993         }
994         spin_lock_bh(&ifp->lock);
995         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
996                 idev->cnf.use_tempaddr = -1;    /*XXX*/
997                 spin_unlock_bh(&ifp->lock);
998                 write_unlock(&idev->lock);
999                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1000                         __func__);
1001                 in6_dev_put(idev);
1002                 ret = -1;
1003                 goto out;
1004         }
1005         in6_ifa_hold(ifp);
1006         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1007         __ipv6_try_regen_rndid(idev, tmpaddr);
1008         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1009         age = (now - ifp->tstamp) / HZ;
1010         tmp_valid_lft = min_t(__u32,
1011                               ifp->valid_lft,
1012                               idev->cnf.temp_valid_lft + age);
1013         tmp_prefered_lft = min_t(__u32,
1014                                  ifp->prefered_lft,
1015                                  idev->cnf.temp_prefered_lft + age -
1016                                  idev->cnf.max_desync_factor);
1017         tmp_plen = ifp->prefix_len;
1018         max_addresses = idev->cnf.max_addresses;
1019         tmp_tstamp = ifp->tstamp;
1020         spin_unlock_bh(&ifp->lock);
1021
1022         regen_advance = idev->cnf.regen_max_retry *
1023                         idev->cnf.dad_transmits *
1024                         idev->nd_parms->retrans_time / HZ;
1025         write_unlock(&idev->lock);
1026
1027         /* A temporary address is created only if this calculated Preferred
1028          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1029          * an implementation must not create a temporary address with a zero
1030          * Preferred Lifetime.
1031          */
1032         if (tmp_prefered_lft <= regen_advance) {
1033                 in6_ifa_put(ifp);
1034                 in6_dev_put(idev);
1035                 ret = -1;
1036                 goto out;
1037         }
1038
1039         addr_flags = IFA_F_TEMPORARY;
1040         /* set in addrconf_prefix_rcv() */
1041         if (ifp->flags & IFA_F_OPTIMISTIC)
1042                 addr_flags |= IFA_F_OPTIMISTIC;
1043
1044         ift = !max_addresses ||
1045               ipv6_count_addresses(idev) < max_addresses ?
1046                 ipv6_add_addr(idev, &addr, tmp_plen,
1047                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1048                               addr_flags) : NULL;
1049         if (!ift || IS_ERR(ift)) {
1050                 in6_ifa_put(ifp);
1051                 in6_dev_put(idev);
1052                 pr_info("%s: retry temporary address regeneration\n", __func__);
1053                 tmpaddr = &addr;
1054                 write_lock(&idev->lock);
1055                 goto retry;
1056         }
1057
1058         spin_lock_bh(&ift->lock);
1059         ift->ifpub = ifp;
1060         ift->valid_lft = tmp_valid_lft;
1061         ift->prefered_lft = tmp_prefered_lft;
1062         ift->cstamp = now;
1063         ift->tstamp = tmp_tstamp;
1064         spin_unlock_bh(&ift->lock);
1065
1066         addrconf_dad_start(ift);
1067         in6_ifa_put(ift);
1068         in6_dev_put(idev);
1069 out:
1070         return ret;
1071 }
1072 #endif
1073
1074 /*
1075  *      Choose an appropriate source address (RFC3484)
1076  */
1077 enum {
1078         IPV6_SADDR_RULE_INIT = 0,
1079         IPV6_SADDR_RULE_LOCAL,
1080         IPV6_SADDR_RULE_SCOPE,
1081         IPV6_SADDR_RULE_PREFERRED,
1082 #ifdef CONFIG_IPV6_MIP6
1083         IPV6_SADDR_RULE_HOA,
1084 #endif
1085         IPV6_SADDR_RULE_OIF,
1086         IPV6_SADDR_RULE_LABEL,
1087 #ifdef CONFIG_IPV6_PRIVACY
1088         IPV6_SADDR_RULE_PRIVACY,
1089 #endif
1090         IPV6_SADDR_RULE_ORCHID,
1091         IPV6_SADDR_RULE_PREFIX,
1092         IPV6_SADDR_RULE_MAX
1093 };
1094
1095 struct ipv6_saddr_score {
1096         int                     rule;
1097         int                     addr_type;
1098         struct inet6_ifaddr     *ifa;
1099         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1100         int                     scopedist;
1101         int                     matchlen;
1102 };
1103
1104 struct ipv6_saddr_dst {
1105         const struct in6_addr *addr;
1106         int ifindex;
1107         int scope;
1108         int label;
1109         unsigned int prefs;
1110 };
1111
1112 static inline int ipv6_saddr_preferred(int type)
1113 {
1114         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1115                 return 1;
1116         return 0;
1117 }
1118
1119 static int ipv6_get_saddr_eval(struct net *net,
1120                                struct ipv6_saddr_score *score,
1121                                struct ipv6_saddr_dst *dst,
1122                                int i)
1123 {
1124         int ret;
1125
1126         if (i <= score->rule) {
1127                 switch (i) {
1128                 case IPV6_SADDR_RULE_SCOPE:
1129                         ret = score->scopedist;
1130                         break;
1131                 case IPV6_SADDR_RULE_PREFIX:
1132                         ret = score->matchlen;
1133                         break;
1134                 default:
1135                         ret = !!test_bit(i, score->scorebits);
1136                 }
1137                 goto out;
1138         }
1139
1140         switch (i) {
1141         case IPV6_SADDR_RULE_INIT:
1142                 /* Rule 0: remember if hiscore is not ready yet */
1143                 ret = !!score->ifa;
1144                 break;
1145         case IPV6_SADDR_RULE_LOCAL:
1146                 /* Rule 1: Prefer same address */
1147                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1148                 break;
1149         case IPV6_SADDR_RULE_SCOPE:
1150                 /* Rule 2: Prefer appropriate scope
1151                  *
1152                  *      ret
1153                  *       ^
1154                  *    -1 |  d 15
1155                  *    ---+--+-+---> scope
1156                  *       |
1157                  *       |             d is scope of the destination.
1158                  *  B-d  |  \
1159                  *       |   \      <- smaller scope is better if
1160                  *  B-15 |    \        if scope is enough for destinaion.
1161                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1162                  * d-C-1 | /
1163                  *       |/         <- greater is better
1164                  *   -C  /             if scope is not enough for destination.
1165                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1166                  *
1167                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1168                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1169                  * Assume B = 0 and we get C > 29.
1170                  */
1171                 ret = __ipv6_addr_src_scope(score->addr_type);
1172                 if (ret >= dst->scope)
1173                         ret = -ret;
1174                 else
1175                         ret -= 128;     /* 30 is enough */
1176                 score->scopedist = ret;
1177                 break;
1178         case IPV6_SADDR_RULE_PREFERRED:
1179                 /* Rule 3: Avoid deprecated and optimistic addresses */
1180                 ret = ipv6_saddr_preferred(score->addr_type) ||
1181                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1182                 break;
1183 #ifdef CONFIG_IPV6_MIP6
1184         case IPV6_SADDR_RULE_HOA:
1185             {
1186                 /* Rule 4: Prefer home address */
1187                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1188                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1189                 break;
1190             }
1191 #endif
1192         case IPV6_SADDR_RULE_OIF:
1193                 /* Rule 5: Prefer outgoing interface */
1194                 ret = (!dst->ifindex ||
1195                        dst->ifindex == score->ifa->idev->dev->ifindex);
1196                 break;
1197         case IPV6_SADDR_RULE_LABEL:
1198                 /* Rule 6: Prefer matching label */
1199                 ret = ipv6_addr_label(net,
1200                                       &score->ifa->addr, score->addr_type,
1201                                       score->ifa->idev->dev->ifindex) == dst->label;
1202                 break;
1203 #ifdef CONFIG_IPV6_PRIVACY
1204         case IPV6_SADDR_RULE_PRIVACY:
1205             {
1206                 /* Rule 7: Prefer public address
1207                  * Note: prefer temporary address if use_tempaddr >= 2
1208                  */
1209                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1210                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1211                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1212                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1213                 break;
1214             }
1215 #endif
1216         case IPV6_SADDR_RULE_ORCHID:
1217                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1218                  *          non-ORCHID vs non-ORCHID
1219                  */
1220                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1221                         ipv6_addr_orchid(dst->addr));
1222                 break;
1223         case IPV6_SADDR_RULE_PREFIX:
1224                 /* Rule 8: Use longest matching prefix */
1225                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1226                 if (ret > score->ifa->prefix_len)
1227                         ret = score->ifa->prefix_len;
1228                 score->matchlen = ret;
1229                 break;
1230         default:
1231                 ret = 0;
1232         }
1233
1234         if (ret)
1235                 __set_bit(i, score->scorebits);
1236         score->rule = i;
1237 out:
1238         return ret;
1239 }
1240
1241 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1242                        const struct in6_addr *daddr, unsigned int prefs,
1243                        struct in6_addr *saddr)
1244 {
1245         struct ipv6_saddr_score scores[2],
1246                                 *score = &scores[0], *hiscore = &scores[1];
1247         struct ipv6_saddr_dst dst;
1248         struct net_device *dev;
1249         int dst_type;
1250
1251         dst_type = __ipv6_addr_type(daddr);
1252         dst.addr = daddr;
1253         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1254         dst.scope = __ipv6_addr_src_scope(dst_type);
1255         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1256         dst.prefs = prefs;
1257
1258         hiscore->rule = -1;
1259         hiscore->ifa = NULL;
1260
1261         rcu_read_lock();
1262
1263         for_each_netdev_rcu(net, dev) {
1264                 struct inet6_dev *idev;
1265
1266                 /* Candidate Source Address (section 4)
1267                  *  - multicast and link-local destination address,
1268                  *    the set of candidate source address MUST only
1269                  *    include addresses assigned to interfaces
1270                  *    belonging to the same link as the outgoing
1271                  *    interface.
1272                  * (- For site-local destination addresses, the
1273                  *    set of candidate source addresses MUST only
1274                  *    include addresses assigned to interfaces
1275                  *    belonging to the same site as the outgoing
1276                  *    interface.)
1277                  */
1278                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1279                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1280                     dst.ifindex && dev->ifindex != dst.ifindex)
1281                         continue;
1282
1283                 idev = __in6_dev_get(dev);
1284                 if (!idev)
1285                         continue;
1286
1287                 read_lock_bh(&idev->lock);
1288                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1289                         int i;
1290
1291                         /*
1292                          * - Tentative Address (RFC2462 section 5.4)
1293                          *  - A tentative address is not considered
1294                          *    "assigned to an interface" in the traditional
1295                          *    sense, unless it is also flagged as optimistic.
1296                          * - Candidate Source Address (section 4)
1297                          *  - In any case, anycast addresses, multicast
1298                          *    addresses, and the unspecified address MUST
1299                          *    NOT be included in a candidate set.
1300                          */
1301                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1302                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1303                                 continue;
1304
1305                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1306
1307                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1308                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1309                                 LIMIT_NETDEBUG(KERN_DEBUG
1310                                                "ADDRCONF: unspecified / multicast address "
1311                                                "assigned as unicast address on %s",
1312                                                dev->name);
1313                                 continue;
1314                         }
1315
1316                         score->rule = -1;
1317                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1318
1319                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1320                                 int minihiscore, miniscore;
1321
1322                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1323                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1324
1325                                 if (minihiscore > miniscore) {
1326                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1327                                             score->scopedist > 0) {
1328                                                 /*
1329                                                  * special case:
1330                                                  * each remaining entry
1331                                                  * has too small (not enough)
1332                                                  * scope, because ifa entries
1333                                                  * are sorted by their scope
1334                                                  * values.
1335                                                  */
1336                                                 goto try_nextdev;
1337                                         }
1338                                         break;
1339                                 } else if (minihiscore < miniscore) {
1340                                         if (hiscore->ifa)
1341                                                 in6_ifa_put(hiscore->ifa);
1342
1343                                         in6_ifa_hold(score->ifa);
1344
1345                                         swap(hiscore, score);
1346
1347                                         /* restore our iterator */
1348                                         score->ifa = hiscore->ifa;
1349
1350                                         break;
1351                                 }
1352                         }
1353                 }
1354 try_nextdev:
1355                 read_unlock_bh(&idev->lock);
1356         }
1357         rcu_read_unlock();
1358
1359         if (!hiscore->ifa)
1360                 return -EADDRNOTAVAIL;
1361
1362         *saddr = hiscore->ifa->addr;
1363         in6_ifa_put(hiscore->ifa);
1364         return 0;
1365 }
1366 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1367
1368 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1369                     unsigned char banned_flags)
1370 {
1371         struct inet6_dev *idev;
1372         int err = -EADDRNOTAVAIL;
1373
1374         rcu_read_lock();
1375         idev = __in6_dev_get(dev);
1376         if (idev) {
1377                 struct inet6_ifaddr *ifp;
1378
1379                 read_lock_bh(&idev->lock);
1380                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1381                         if (ifp->scope == IFA_LINK &&
1382                             !(ifp->flags & banned_flags)) {
1383                                 *addr = ifp->addr;
1384                                 err = 0;
1385                                 break;
1386                         }
1387                 }
1388                 read_unlock_bh(&idev->lock);
1389         }
1390         rcu_read_unlock();
1391         return err;
1392 }
1393
1394 static int ipv6_count_addresses(struct inet6_dev *idev)
1395 {
1396         int cnt = 0;
1397         struct inet6_ifaddr *ifp;
1398
1399         read_lock_bh(&idev->lock);
1400         list_for_each_entry(ifp, &idev->addr_list, if_list)
1401                 cnt++;
1402         read_unlock_bh(&idev->lock);
1403         return cnt;
1404 }
1405
1406 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1407                   struct net_device *dev, int strict)
1408 {
1409         struct inet6_ifaddr *ifp;
1410         struct hlist_node *node;
1411         unsigned int hash = inet6_addr_hash(addr);
1412
1413         rcu_read_lock_bh();
1414         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1415                 if (!net_eq(dev_net(ifp->idev->dev), net))
1416                         continue;
1417                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1418                     !(ifp->flags&IFA_F_TENTATIVE) &&
1419                     (dev == NULL || ifp->idev->dev == dev ||
1420                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1421                         rcu_read_unlock_bh();
1422                         return 1;
1423                 }
1424         }
1425
1426         rcu_read_unlock_bh();
1427         return 0;
1428 }
1429 EXPORT_SYMBOL(ipv6_chk_addr);
1430
1431 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1432                                struct net_device *dev)
1433 {
1434         unsigned int hash = inet6_addr_hash(addr);
1435         struct inet6_ifaddr *ifp;
1436         struct hlist_node *node;
1437
1438         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1439                 if (!net_eq(dev_net(ifp->idev->dev), net))
1440                         continue;
1441                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1442                         if (dev == NULL || ifp->idev->dev == dev)
1443                                 return true;
1444                 }
1445         }
1446         return false;
1447 }
1448
1449 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1450 {
1451         struct inet6_dev *idev;
1452         struct inet6_ifaddr *ifa;
1453         int     onlink;
1454
1455         onlink = 0;
1456         rcu_read_lock();
1457         idev = __in6_dev_get(dev);
1458         if (idev) {
1459                 read_lock_bh(&idev->lock);
1460                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1461                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1462                                                    ifa->prefix_len);
1463                         if (onlink)
1464                                 break;
1465                 }
1466                 read_unlock_bh(&idev->lock);
1467         }
1468         rcu_read_unlock();
1469         return onlink;
1470 }
1471 EXPORT_SYMBOL(ipv6_chk_prefix);
1472
1473 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1474                                      struct net_device *dev, int strict)
1475 {
1476         struct inet6_ifaddr *ifp, *result = NULL;
1477         unsigned int hash = inet6_addr_hash(addr);
1478         struct hlist_node *node;
1479
1480         rcu_read_lock_bh();
1481         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1482                 if (!net_eq(dev_net(ifp->idev->dev), net))
1483                         continue;
1484                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1485                         if (dev == NULL || ifp->idev->dev == dev ||
1486                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1487                                 result = ifp;
1488                                 in6_ifa_hold(ifp);
1489                                 break;
1490                         }
1491                 }
1492         }
1493         rcu_read_unlock_bh();
1494
1495         return result;
1496 }
1497
1498 /* Gets referenced address, destroys ifaddr */
1499
1500 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1501 {
1502         if (ifp->flags&IFA_F_PERMANENT) {
1503                 spin_lock_bh(&ifp->lock);
1504                 addrconf_del_timer(ifp);
1505                 ifp->flags |= IFA_F_TENTATIVE;
1506                 if (dad_failed)
1507                         ifp->flags |= IFA_F_DADFAILED;
1508                 spin_unlock_bh(&ifp->lock);
1509                 if (dad_failed)
1510                         ipv6_ifa_notify(0, ifp);
1511                 in6_ifa_put(ifp);
1512 #ifdef CONFIG_IPV6_PRIVACY
1513         } else if (ifp->flags&IFA_F_TEMPORARY) {
1514                 struct inet6_ifaddr *ifpub;
1515                 spin_lock_bh(&ifp->lock);
1516                 ifpub = ifp->ifpub;
1517                 if (ifpub) {
1518                         in6_ifa_hold(ifpub);
1519                         spin_unlock_bh(&ifp->lock);
1520                         ipv6_create_tempaddr(ifpub, ifp);
1521                         in6_ifa_put(ifpub);
1522                 } else {
1523                         spin_unlock_bh(&ifp->lock);
1524                 }
1525                 ipv6_del_addr(ifp);
1526 #endif
1527         } else
1528                 ipv6_del_addr(ifp);
1529 }
1530
1531 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1532 {
1533         int err = -ENOENT;
1534
1535         spin_lock(&ifp->state_lock);
1536         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1537                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1538                 err = 0;
1539         }
1540         spin_unlock(&ifp->state_lock);
1541
1542         return err;
1543 }
1544
1545 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1546 {
1547         struct inet6_dev *idev = ifp->idev;
1548
1549         if (addrconf_dad_end(ifp)) {
1550                 in6_ifa_put(ifp);
1551                 return;
1552         }
1553
1554         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1555                              ifp->idev->dev->name, &ifp->addr);
1556
1557         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1558                 struct in6_addr addr;
1559
1560                 addr.s6_addr32[0] = htonl(0xfe800000);
1561                 addr.s6_addr32[1] = 0;
1562
1563                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1564                     ipv6_addr_equal(&ifp->addr, &addr)) {
1565                         /* DAD failed for link-local based on MAC address */
1566                         idev->cnf.disable_ipv6 = 1;
1567
1568                         pr_info("%s: IPv6 being disabled!\n",
1569                                 ifp->idev->dev->name);
1570                 }
1571         }
1572
1573         addrconf_dad_stop(ifp, 1);
1574 }
1575
1576 /* Join to solicited addr multicast group. */
1577
1578 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1579 {
1580         struct in6_addr maddr;
1581
1582         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1583                 return;
1584
1585         addrconf_addr_solict_mult(addr, &maddr);
1586         ipv6_dev_mc_inc(dev, &maddr);
1587 }
1588
1589 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1590 {
1591         struct in6_addr maddr;
1592
1593         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1594                 return;
1595
1596         addrconf_addr_solict_mult(addr, &maddr);
1597         __ipv6_dev_mc_dec(idev, &maddr);
1598 }
1599
1600 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1601 {
1602         struct in6_addr addr;
1603         if (ifp->prefix_len == 127) /* RFC 6164 */
1604                 return;
1605         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1606         if (ipv6_addr_any(&addr))
1607                 return;
1608         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1609 }
1610
1611 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1612 {
1613         struct in6_addr addr;
1614         if (ifp->prefix_len == 127) /* RFC 6164 */
1615                 return;
1616         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1617         if (ipv6_addr_any(&addr))
1618                 return;
1619         __ipv6_dev_ac_dec(ifp->idev, &addr);
1620 }
1621
1622 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1623 {
1624         if (dev->addr_len != ETH_ALEN)
1625                 return -1;
1626         memcpy(eui, dev->dev_addr, 3);
1627         memcpy(eui + 5, dev->dev_addr + 3, 3);
1628
1629         /*
1630          * The zSeries OSA network cards can be shared among various
1631          * OS instances, but the OSA cards have only one MAC address.
1632          * This leads to duplicate address conflicts in conjunction
1633          * with IPv6 if more than one instance uses the same card.
1634          *
1635          * The driver for these cards can deliver a unique 16-bit
1636          * identifier for each instance sharing the same card.  It is
1637          * placed instead of 0xFFFE in the interface identifier.  The
1638          * "u" bit of the interface identifier is not inverted in this
1639          * case.  Hence the resulting interface identifier has local
1640          * scope according to RFC2373.
1641          */
1642         if (dev->dev_id) {
1643                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1644                 eui[4] = dev->dev_id & 0xFF;
1645         } else {
1646                 eui[3] = 0xFF;
1647                 eui[4] = 0xFE;
1648                 eui[0] ^= 2;
1649         }
1650         return 0;
1651 }
1652
1653 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1654 {
1655         if (dev->addr_len != IEEE802154_ADDR_LEN)
1656                 return -1;
1657         memcpy(eui, dev->dev_addr, 8);
1658         return 0;
1659 }
1660
1661 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1662 {
1663         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1664         if (dev->addr_len != ARCNET_ALEN)
1665                 return -1;
1666         memset(eui, 0, 7);
1667         eui[7] = *(u8 *)dev->dev_addr;
1668         return 0;
1669 }
1670
1671 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1672 {
1673         if (dev->addr_len != INFINIBAND_ALEN)
1674                 return -1;
1675         memcpy(eui, dev->dev_addr + 12, 8);
1676         eui[0] |= 2;
1677         return 0;
1678 }
1679
1680 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1681 {
1682         if (addr == 0)
1683                 return -1;
1684         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1685                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1686                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1687                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1688                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1689                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1690         eui[1] = 0;
1691         eui[2] = 0x5E;
1692         eui[3] = 0xFE;
1693         memcpy(eui + 4, &addr, 4);
1694         return 0;
1695 }
1696
1697 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1698 {
1699         if (dev->priv_flags & IFF_ISATAP)
1700                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1701         return -1;
1702 }
1703
1704 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1705 {
1706         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1707 }
1708
1709 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1710 {
1711         switch (dev->type) {
1712         case ARPHRD_ETHER:
1713         case ARPHRD_FDDI:
1714                 return addrconf_ifid_eui48(eui, dev);
1715         case ARPHRD_ARCNET:
1716                 return addrconf_ifid_arcnet(eui, dev);
1717         case ARPHRD_INFINIBAND:
1718                 return addrconf_ifid_infiniband(eui, dev);
1719         case ARPHRD_SIT:
1720                 return addrconf_ifid_sit(eui, dev);
1721         case ARPHRD_IPGRE:
1722                 return addrconf_ifid_gre(eui, dev);
1723         case ARPHRD_IEEE802154:
1724                 return addrconf_ifid_eui64(eui, dev);
1725         }
1726         return -1;
1727 }
1728
1729 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1730 {
1731         int err = -1;
1732         struct inet6_ifaddr *ifp;
1733
1734         read_lock_bh(&idev->lock);
1735         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1736                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1737                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1738                         err = 0;
1739                         break;
1740                 }
1741         }
1742         read_unlock_bh(&idev->lock);
1743         return err;
1744 }
1745
1746 #ifdef CONFIG_IPV6_PRIVACY
1747 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1748 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1749 {
1750 regen:
1751         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1752         idev->rndid[0] &= ~0x02;
1753
1754         /*
1755          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1756          * check if generated address is not inappropriate
1757          *
1758          *  - Reserved subnet anycast (RFC 2526)
1759          *      11111101 11....11 1xxxxxxx
1760          *  - ISATAP (RFC4214) 6.1
1761          *      00-00-5E-FE-xx-xx-xx-xx
1762          *  - value 0
1763          *  - XXX: already assigned to an address on the device
1764          */
1765         if (idev->rndid[0] == 0xfd &&
1766             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1767             (idev->rndid[7]&0x80))
1768                 goto regen;
1769         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1770                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1771                         goto regen;
1772                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1773                         goto regen;
1774         }
1775 }
1776
1777 static void ipv6_regen_rndid(unsigned long data)
1778 {
1779         struct inet6_dev *idev = (struct inet6_dev *) data;
1780         unsigned long expires;
1781
1782         rcu_read_lock_bh();
1783         write_lock_bh(&idev->lock);
1784
1785         if (idev->dead)
1786                 goto out;
1787
1788         __ipv6_regen_rndid(idev);
1789
1790         expires = jiffies +
1791                 idev->cnf.temp_prefered_lft * HZ -
1792                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1793                 idev->cnf.max_desync_factor * HZ;
1794         if (time_before(expires, jiffies)) {
1795                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1796                         __func__, idev->dev->name);
1797                 goto out;
1798         }
1799
1800         if (!mod_timer(&idev->regen_timer, expires))
1801                 in6_dev_hold(idev);
1802
1803 out:
1804         write_unlock_bh(&idev->lock);
1805         rcu_read_unlock_bh();
1806         in6_dev_put(idev);
1807 }
1808
1809 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1810 {
1811         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1812                 __ipv6_regen_rndid(idev);
1813 }
1814 #endif
1815
1816 /*
1817  *      Add prefix route.
1818  */
1819
1820 static void
1821 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1822                       unsigned long expires, u32 flags)
1823 {
1824         struct fib6_config cfg = {
1825                 .fc_table = RT6_TABLE_PREFIX,
1826                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1827                 .fc_ifindex = dev->ifindex,
1828                 .fc_expires = expires,
1829                 .fc_dst_len = plen,
1830                 .fc_flags = RTF_UP | flags,
1831                 .fc_nlinfo.nl_net = dev_net(dev),
1832                 .fc_protocol = RTPROT_KERNEL,
1833         };
1834
1835         cfg.fc_dst = *pfx;
1836
1837         /* Prevent useless cloning on PtP SIT.
1838            This thing is done here expecting that the whole
1839            class of non-broadcast devices need not cloning.
1840          */
1841 #if IS_ENABLED(CONFIG_IPV6_SIT)
1842         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1843                 cfg.fc_flags |= RTF_NONEXTHOP;
1844 #endif
1845
1846         ip6_route_add(&cfg);
1847 }
1848
1849
1850 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1851                                                   int plen,
1852                                                   const struct net_device *dev,
1853                                                   u32 flags, u32 noflags)
1854 {
1855         struct fib6_node *fn;
1856         struct rt6_info *rt = NULL;
1857         struct fib6_table *table;
1858
1859         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1860         if (table == NULL)
1861                 return NULL;
1862
1863         read_lock_bh(&table->tb6_lock);
1864         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1865         if (!fn)
1866                 goto out;
1867         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1868                 if (rt->dst.dev->ifindex != dev->ifindex)
1869                         continue;
1870                 if ((rt->rt6i_flags & flags) != flags)
1871                         continue;
1872                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1873                         continue;
1874                 dst_hold(&rt->dst);
1875                 break;
1876         }
1877 out:
1878         read_unlock_bh(&table->tb6_lock);
1879         return rt;
1880 }
1881
1882
1883 /* Create "default" multicast route to the interface */
1884
1885 static void addrconf_add_mroute(struct net_device *dev)
1886 {
1887         struct fib6_config cfg = {
1888                 .fc_table = RT6_TABLE_LOCAL,
1889                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1890                 .fc_ifindex = dev->ifindex,
1891                 .fc_dst_len = 8,
1892                 .fc_flags = RTF_UP,
1893                 .fc_nlinfo.nl_net = dev_net(dev),
1894         };
1895
1896         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1897
1898         ip6_route_add(&cfg);
1899 }
1900
1901 #if IS_ENABLED(CONFIG_IPV6_SIT)
1902 static void sit_route_add(struct net_device *dev)
1903 {
1904         struct fib6_config cfg = {
1905                 .fc_table = RT6_TABLE_MAIN,
1906                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1907                 .fc_ifindex = dev->ifindex,
1908                 .fc_dst_len = 96,
1909                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1910                 .fc_nlinfo.nl_net = dev_net(dev),
1911         };
1912
1913         /* prefix length - 96 bits "::d.d.d.d" */
1914         ip6_route_add(&cfg);
1915 }
1916 #endif
1917
1918 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1919 {
1920         struct inet6_dev *idev;
1921
1922         ASSERT_RTNL();
1923
1924         idev = ipv6_find_idev(dev);
1925         if (!idev)
1926                 return ERR_PTR(-ENOBUFS);
1927
1928         if (idev->cnf.disable_ipv6)
1929                 return ERR_PTR(-EACCES);
1930
1931         /* Add default multicast route */
1932         if (!(dev->flags & IFF_LOOPBACK))
1933                 addrconf_add_mroute(dev);
1934
1935         return idev;
1936 }
1937
1938 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1939 {
1940         struct prefix_info *pinfo;
1941         __u32 valid_lft;
1942         __u32 prefered_lft;
1943         int addr_type;
1944         struct inet6_dev *in6_dev;
1945         struct net *net = dev_net(dev);
1946
1947         pinfo = (struct prefix_info *) opt;
1948
1949         if (len < sizeof(struct prefix_info)) {
1950                 ADBG(("addrconf: prefix option too short\n"));
1951                 return;
1952         }
1953
1954         /*
1955          *      Validation checks ([ADDRCONF], page 19)
1956          */
1957
1958         addr_type = ipv6_addr_type(&pinfo->prefix);
1959
1960         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1961                 return;
1962
1963         valid_lft = ntohl(pinfo->valid);
1964         prefered_lft = ntohl(pinfo->prefered);
1965
1966         if (prefered_lft > valid_lft) {
1967                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1968                 return;
1969         }
1970
1971         in6_dev = in6_dev_get(dev);
1972
1973         if (in6_dev == NULL) {
1974                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1975                                     dev->name);
1976                 return;
1977         }
1978
1979         /*
1980          *      Two things going on here:
1981          *      1) Add routes for on-link prefixes
1982          *      2) Configure prefixes with the auto flag set
1983          */
1984
1985         if (pinfo->onlink) {
1986                 struct rt6_info *rt;
1987                 unsigned long rt_expires;
1988
1989                 /* Avoid arithmetic overflow. Really, we could
1990                  * save rt_expires in seconds, likely valid_lft,
1991                  * but it would require division in fib gc, that it
1992                  * not good.
1993                  */
1994                 if (HZ > USER_HZ)
1995                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1996                 else
1997                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1998
1999                 if (addrconf_finite_timeout(rt_expires))
2000                         rt_expires *= HZ;
2001
2002                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2003                                                pinfo->prefix_len,
2004                                                dev,
2005                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2006                                                RTF_GATEWAY | RTF_DEFAULT);
2007
2008                 if (rt) {
2009                         /* Autoconf prefix route */
2010                         if (valid_lft == 0) {
2011                                 ip6_del_rt(rt);
2012                                 rt = NULL;
2013                         } else if (addrconf_finite_timeout(rt_expires)) {
2014                                 /* not infinity */
2015                                 rt6_set_expires(rt, jiffies + rt_expires);
2016                         } else {
2017                                 rt6_clean_expires(rt);
2018                         }
2019                 } else if (valid_lft) {
2020                         clock_t expires = 0;
2021                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2022                         if (addrconf_finite_timeout(rt_expires)) {
2023                                 /* not infinity */
2024                                 flags |= RTF_EXPIRES;
2025                                 expires = jiffies_to_clock_t(rt_expires);
2026                         }
2027                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2028                                               dev, expires, flags);
2029                 }
2030                 if (rt)
2031                         dst_release(&rt->dst);
2032         }
2033
2034         /* Try to figure out our local address for this prefix */
2035
2036         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2037                 struct inet6_ifaddr *ifp;
2038                 struct in6_addr addr;
2039                 int create = 0, update_lft = 0;
2040
2041                 if (pinfo->prefix_len == 64) {
2042                         memcpy(&addr, &pinfo->prefix, 8);
2043                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2044                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2045                                 in6_dev_put(in6_dev);
2046                                 return;
2047                         }
2048                         goto ok;
2049                 }
2050                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2051                                     pinfo->prefix_len);
2052                 in6_dev_put(in6_dev);
2053                 return;
2054
2055 ok:
2056
2057                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2058
2059                 if (ifp == NULL && valid_lft) {
2060                         int max_addresses = in6_dev->cnf.max_addresses;
2061                         u32 addr_flags = 0;
2062
2063 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2064                         if (in6_dev->cnf.optimistic_dad &&
2065                             !net->ipv6.devconf_all->forwarding && sllao)
2066                                 addr_flags = IFA_F_OPTIMISTIC;
2067 #endif
2068
2069                         /* Do not allow to create too much of autoconfigured
2070                          * addresses; this would be too easy way to crash kernel.
2071                          */
2072                         if (!max_addresses ||
2073                             ipv6_count_addresses(in6_dev) < max_addresses)
2074                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2075                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2076                                                     addr_flags);
2077
2078                         if (!ifp || IS_ERR(ifp)) {
2079                                 in6_dev_put(in6_dev);
2080                                 return;
2081                         }
2082
2083                         update_lft = create = 1;
2084                         ifp->cstamp = jiffies;
2085                         addrconf_dad_start(ifp);
2086                 }
2087
2088                 if (ifp) {
2089                         int flags;
2090                         unsigned long now;
2091 #ifdef CONFIG_IPV6_PRIVACY
2092                         struct inet6_ifaddr *ift;
2093 #endif
2094                         u32 stored_lft;
2095
2096                         /* update lifetime (RFC2462 5.5.3 e) */
2097                         spin_lock(&ifp->lock);
2098                         now = jiffies;
2099                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2100                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2101                         else
2102                                 stored_lft = 0;
2103                         if (!update_lft && stored_lft) {
2104                                 if (valid_lft > MIN_VALID_LIFETIME ||
2105                                     valid_lft > stored_lft)
2106                                         update_lft = 1;
2107                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2108                                         /* valid_lft <= stored_lft is always true */
2109                                         /*
2110                                          * RFC 4862 Section 5.5.3e:
2111                                          * "Note that the preferred lifetime of
2112                                          *  the corresponding address is always
2113                                          *  reset to the Preferred Lifetime in
2114                                          *  the received Prefix Information
2115                                          *  option, regardless of whether the
2116                                          *  valid lifetime is also reset or
2117                                          *  ignored."
2118                                          *
2119                                          *  So if the preferred lifetime in
2120                                          *  this advertisement is different
2121                                          *  than what we have stored, but the
2122                                          *  valid lifetime is invalid, just
2123                                          *  reset prefered_lft.
2124                                          *
2125                                          *  We must set the valid lifetime
2126                                          *  to the stored lifetime since we'll
2127                                          *  be updating the timestamp below,
2128                                          *  else we'll set it back to the
2129                                          *  minimum.
2130                                          */
2131                                         if (prefered_lft != ifp->prefered_lft) {
2132                                                 valid_lft = stored_lft;
2133                                                 update_lft = 1;
2134                                         }
2135                                 } else {
2136                                         valid_lft = MIN_VALID_LIFETIME;
2137                                         if (valid_lft < prefered_lft)
2138                                                 prefered_lft = valid_lft;
2139                                         update_lft = 1;
2140                                 }
2141                         }
2142
2143                         if (update_lft) {
2144                                 ifp->valid_lft = valid_lft;
2145                                 ifp->prefered_lft = prefered_lft;
2146                                 ifp->tstamp = now;
2147                                 flags = ifp->flags;
2148                                 ifp->flags &= ~IFA_F_DEPRECATED;
2149                                 spin_unlock(&ifp->lock);
2150
2151                                 if (!(flags&IFA_F_TENTATIVE))
2152                                         ipv6_ifa_notify(0, ifp);
2153                         } else
2154                                 spin_unlock(&ifp->lock);
2155
2156 #ifdef CONFIG_IPV6_PRIVACY
2157                         read_lock_bh(&in6_dev->lock);
2158                         /* update all temporary addresses in the list */
2159                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2160                                             tmp_list) {
2161                                 int age, max_valid, max_prefered;
2162
2163                                 if (ifp != ift->ifpub)
2164                                         continue;
2165
2166                                 /*
2167                                  * RFC 4941 section 3.3:
2168                                  * If a received option will extend the lifetime
2169                                  * of a public address, the lifetimes of
2170                                  * temporary addresses should be extended,
2171                                  * subject to the overall constraint that no
2172                                  * temporary addresses should ever remain
2173                                  * "valid" or "preferred" for a time longer than
2174                                  * (TEMP_VALID_LIFETIME) or
2175                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2176                                  * respectively.
2177                                  */
2178                                 age = (now - ift->cstamp) / HZ;
2179                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2180                                 if (max_valid < 0)
2181                                         max_valid = 0;
2182
2183                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2184                                                in6_dev->cnf.max_desync_factor -
2185                                                age;
2186                                 if (max_prefered < 0)
2187                                         max_prefered = 0;
2188
2189                                 if (valid_lft > max_valid)
2190                                         valid_lft = max_valid;
2191
2192                                 if (prefered_lft > max_prefered)
2193                                         prefered_lft = max_prefered;
2194
2195                                 spin_lock(&ift->lock);
2196                                 flags = ift->flags;
2197                                 ift->valid_lft = valid_lft;
2198                                 ift->prefered_lft = prefered_lft;
2199                                 ift->tstamp = now;
2200                                 if (prefered_lft > 0)
2201                                         ift->flags &= ~IFA_F_DEPRECATED;
2202
2203                                 spin_unlock(&ift->lock);
2204                                 if (!(flags&IFA_F_TENTATIVE))
2205                                         ipv6_ifa_notify(0, ift);
2206                         }
2207
2208                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2209                                 /*
2210                                  * When a new public address is created as
2211                                  * described in [ADDRCONF], also create a new
2212                                  * temporary address. Also create a temporary
2213                                  * address if it's enabled but no temporary
2214                                  * address currently exists.
2215                                  */
2216                                 read_unlock_bh(&in6_dev->lock);
2217                                 ipv6_create_tempaddr(ifp, NULL);
2218                         } else {
2219                                 read_unlock_bh(&in6_dev->lock);
2220                         }
2221 #endif
2222                         in6_ifa_put(ifp);
2223                         addrconf_verify(0);
2224                 }
2225         }
2226         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2227         in6_dev_put(in6_dev);
2228 }
2229
2230 /*
2231  *      Set destination address.
2232  *      Special case for SIT interfaces where we create a new "virtual"
2233  *      device.
2234  */
2235 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2236 {
2237         struct in6_ifreq ireq;
2238         struct net_device *dev;
2239         int err = -EINVAL;
2240
2241         rtnl_lock();
2242
2243         err = -EFAULT;
2244         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2245                 goto err_exit;
2246
2247         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2248
2249         err = -ENODEV;
2250         if (dev == NULL)
2251                 goto err_exit;
2252
2253 #if IS_ENABLED(CONFIG_IPV6_SIT)
2254         if (dev->type == ARPHRD_SIT) {
2255                 const struct net_device_ops *ops = dev->netdev_ops;
2256                 struct ifreq ifr;
2257                 struct ip_tunnel_parm p;
2258
2259                 err = -EADDRNOTAVAIL;
2260                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2261                         goto err_exit;
2262
2263                 memset(&p, 0, sizeof(p));
2264                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2265                 p.iph.saddr = 0;
2266                 p.iph.version = 4;
2267                 p.iph.ihl = 5;
2268                 p.iph.protocol = IPPROTO_IPV6;
2269                 p.iph.ttl = 64;
2270                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2271
2272                 if (ops->ndo_do_ioctl) {
2273                         mm_segment_t oldfs = get_fs();
2274
2275                         set_fs(KERNEL_DS);
2276                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2277                         set_fs(oldfs);
2278                 } else
2279                         err = -EOPNOTSUPP;
2280
2281                 if (err == 0) {
2282                         err = -ENOBUFS;
2283                         dev = __dev_get_by_name(net, p.name);
2284                         if (!dev)
2285                                 goto err_exit;
2286                         err = dev_open(dev);
2287                 }
2288         }
2289 #endif
2290
2291 err_exit:
2292         rtnl_unlock();
2293         return err;
2294 }
2295
2296 /*
2297  *      Manual configuration of address on an interface
2298  */
2299 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2300                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2301                           __u32 valid_lft)
2302 {
2303         struct inet6_ifaddr *ifp;
2304         struct inet6_dev *idev;
2305         struct net_device *dev;
2306         int scope;
2307         u32 flags;
2308         clock_t expires;
2309         unsigned long timeout;
2310
2311         ASSERT_RTNL();
2312
2313         if (plen > 128)
2314                 return -EINVAL;
2315
2316         /* check the lifetime */
2317         if (!valid_lft || prefered_lft > valid_lft)
2318                 return -EINVAL;
2319
2320         dev = __dev_get_by_index(net, ifindex);
2321         if (!dev)
2322                 return -ENODEV;
2323
2324         idev = addrconf_add_dev(dev);
2325         if (IS_ERR(idev))
2326                 return PTR_ERR(idev);
2327
2328         scope = ipv6_addr_scope(pfx);
2329
2330         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2331         if (addrconf_finite_timeout(timeout)) {
2332                 expires = jiffies_to_clock_t(timeout * HZ);
2333                 valid_lft = timeout;
2334                 flags = RTF_EXPIRES;
2335         } else {
2336                 expires = 0;
2337                 flags = 0;
2338                 ifa_flags |= IFA_F_PERMANENT;
2339         }
2340
2341         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2342         if (addrconf_finite_timeout(timeout)) {
2343                 if (timeout == 0)
2344                         ifa_flags |= IFA_F_DEPRECATED;
2345                 prefered_lft = timeout;
2346         }
2347
2348         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2349
2350         if (!IS_ERR(ifp)) {
2351                 spin_lock_bh(&ifp->lock);
2352                 ifp->valid_lft = valid_lft;
2353                 ifp->prefered_lft = prefered_lft;
2354                 ifp->tstamp = jiffies;
2355                 spin_unlock_bh(&ifp->lock);
2356
2357                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2358                                       expires, flags);
2359                 /*
2360                  * Note that section 3.1 of RFC 4429 indicates
2361                  * that the Optimistic flag should not be set for
2362                  * manually configured addresses
2363                  */
2364                 addrconf_dad_start(ifp);
2365                 in6_ifa_put(ifp);
2366                 addrconf_verify(0);
2367                 return 0;
2368         }
2369
2370         return PTR_ERR(ifp);
2371 }
2372
2373 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2374                           unsigned int plen)
2375 {
2376         struct inet6_ifaddr *ifp;
2377         struct inet6_dev *idev;
2378         struct net_device *dev;
2379
2380         if (plen > 128)
2381                 return -EINVAL;
2382
2383         dev = __dev_get_by_index(net, ifindex);
2384         if (!dev)
2385                 return -ENODEV;
2386
2387         if ((idev = __in6_dev_get(dev)) == NULL)
2388                 return -ENXIO;
2389
2390         read_lock_bh(&idev->lock);
2391         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2392                 if (ifp->prefix_len == plen &&
2393                     ipv6_addr_equal(pfx, &ifp->addr)) {
2394                         in6_ifa_hold(ifp);
2395                         read_unlock_bh(&idev->lock);
2396
2397                         ipv6_del_addr(ifp);
2398
2399                         /* If the last address is deleted administratively,
2400                            disable IPv6 on this interface.
2401                          */
2402                         if (list_empty(&idev->addr_list))
2403                                 addrconf_ifdown(idev->dev, 1);
2404                         return 0;
2405                 }
2406         }
2407         read_unlock_bh(&idev->lock);
2408         return -EADDRNOTAVAIL;
2409 }
2410
2411
2412 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2413 {
2414         struct in6_ifreq ireq;
2415         int err;
2416
2417         if (!capable(CAP_NET_ADMIN))
2418                 return -EPERM;
2419
2420         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2421                 return -EFAULT;
2422
2423         rtnl_lock();
2424         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2425                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2426                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2427         rtnl_unlock();
2428         return err;
2429 }
2430
2431 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2432 {
2433         struct in6_ifreq ireq;
2434         int err;
2435
2436         if (!capable(CAP_NET_ADMIN))
2437                 return -EPERM;
2438
2439         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2440                 return -EFAULT;
2441
2442         rtnl_lock();
2443         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2444                              ireq.ifr6_prefixlen);
2445         rtnl_unlock();
2446         return err;
2447 }
2448
2449 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2450                      int plen, int scope)
2451 {
2452         struct inet6_ifaddr *ifp;
2453
2454         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2455         if (!IS_ERR(ifp)) {
2456                 spin_lock_bh(&ifp->lock);
2457                 ifp->flags &= ~IFA_F_TENTATIVE;
2458                 spin_unlock_bh(&ifp->lock);
2459                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2460                 in6_ifa_put(ifp);
2461         }
2462 }
2463
2464 #if IS_ENABLED(CONFIG_IPV6_SIT)
2465 static void sit_add_v4_addrs(struct inet6_dev *idev)
2466 {
2467         struct in6_addr addr;
2468         struct net_device *dev;
2469         struct net *net = dev_net(idev->dev);
2470         int scope;
2471
2472         ASSERT_RTNL();
2473
2474         memset(&addr, 0, sizeof(struct in6_addr));
2475         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2476
2477         if (idev->dev->flags&IFF_POINTOPOINT) {
2478                 addr.s6_addr32[0] = htonl(0xfe800000);
2479                 scope = IFA_LINK;
2480         } else {
2481                 scope = IPV6_ADDR_COMPATv4;
2482         }
2483
2484         if (addr.s6_addr32[3]) {
2485                 add_addr(idev, &addr, 128, scope);
2486                 return;
2487         }
2488
2489         for_each_netdev(net, dev) {
2490                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2491                 if (in_dev && (dev->flags & IFF_UP)) {
2492                         struct in_ifaddr *ifa;
2493
2494                         int flag = scope;
2495
2496                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2497                                 int plen;
2498
2499                                 addr.s6_addr32[3] = ifa->ifa_local;
2500
2501                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2502                                         continue;
2503                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2504                                         if (idev->dev->flags&IFF_POINTOPOINT)
2505                                                 continue;
2506                                         flag |= IFA_HOST;
2507                                 }
2508                                 if (idev->dev->flags&IFF_POINTOPOINT)
2509                                         plen = 64;
2510                                 else
2511                                         plen = 96;
2512
2513                                 add_addr(idev, &addr, plen, flag);
2514                         }
2515                 }
2516         }
2517 }
2518 #endif
2519
2520 static void init_loopback(struct net_device *dev)
2521 {
2522         struct inet6_dev  *idev;
2523
2524         /* ::1 */
2525
2526         ASSERT_RTNL();
2527
2528         if ((idev = ipv6_find_idev(dev)) == NULL) {
2529                 pr_debug("%s: add_dev failed\n", __func__);
2530                 return;
2531         }
2532
2533         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2534 }
2535
2536 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2537 {
2538         struct inet6_ifaddr *ifp;
2539         u32 addr_flags = IFA_F_PERMANENT;
2540
2541 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2542         if (idev->cnf.optimistic_dad &&
2543             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2544                 addr_flags |= IFA_F_OPTIMISTIC;
2545 #endif
2546
2547
2548         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2549         if (!IS_ERR(ifp)) {
2550                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2551                 addrconf_dad_start(ifp);
2552                 in6_ifa_put(ifp);
2553         }
2554 }
2555
2556 static void addrconf_dev_config(struct net_device *dev)
2557 {
2558         struct in6_addr addr;
2559         struct inet6_dev *idev;
2560
2561         ASSERT_RTNL();
2562
2563         if ((dev->type != ARPHRD_ETHER) &&
2564             (dev->type != ARPHRD_FDDI) &&
2565             (dev->type != ARPHRD_ARCNET) &&
2566             (dev->type != ARPHRD_INFINIBAND) &&
2567             (dev->type != ARPHRD_IEEE802154)) {
2568                 /* Alas, we support only Ethernet autoconfiguration. */
2569                 return;
2570         }
2571
2572         idev = addrconf_add_dev(dev);
2573         if (IS_ERR(idev))
2574                 return;
2575
2576         memset(&addr, 0, sizeof(struct in6_addr));
2577         addr.s6_addr32[0] = htonl(0xFE800000);
2578
2579         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2580                 addrconf_add_linklocal(idev, &addr);
2581 }
2582
2583 #if IS_ENABLED(CONFIG_IPV6_SIT)
2584 static void addrconf_sit_config(struct net_device *dev)
2585 {
2586         struct inet6_dev *idev;
2587
2588         ASSERT_RTNL();
2589
2590         /*
2591          * Configure the tunnel with one of our IPv4
2592          * addresses... we should configure all of
2593          * our v4 addrs in the tunnel
2594          */
2595
2596         if ((idev = ipv6_find_idev(dev)) == NULL) {
2597                 pr_debug("%s: add_dev failed\n", __func__);
2598                 return;
2599         }
2600
2601         if (dev->priv_flags & IFF_ISATAP) {
2602                 struct in6_addr addr;
2603
2604                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2605                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2606                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2607                         addrconf_add_linklocal(idev, &addr);
2608                 return;
2609         }
2610
2611         sit_add_v4_addrs(idev);
2612
2613         if (dev->flags&IFF_POINTOPOINT)
2614                 addrconf_add_mroute(dev);
2615         else
2616                 sit_route_add(dev);
2617 }
2618 #endif
2619
2620 #if IS_ENABLED(CONFIG_NET_IPGRE)
2621 static void addrconf_gre_config(struct net_device *dev)
2622 {
2623         struct inet6_dev *idev;
2624         struct in6_addr addr;
2625
2626         pr_info("%s(%s)\n", __func__, dev->name);
2627
2628         ASSERT_RTNL();
2629
2630         if ((idev = ipv6_find_idev(dev)) == NULL) {
2631                 pr_debug("%s: add_dev failed\n", __func__);
2632                 return;
2633         }
2634
2635         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2636         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2637
2638         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2639                 addrconf_add_linklocal(idev, &addr);
2640 }
2641 #endif
2642
2643 static inline int
2644 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2645 {
2646         struct in6_addr lladdr;
2647
2648         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2649                 addrconf_add_linklocal(idev, &lladdr);
2650                 return 0;
2651         }
2652         return -1;
2653 }
2654
2655 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2656 {
2657         struct net_device *link_dev;
2658         struct net *net = dev_net(idev->dev);
2659
2660         /* first try to inherit the link-local address from the link device */
2661         if (idev->dev->iflink &&
2662             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2663                 if (!ipv6_inherit_linklocal(idev, link_dev))
2664                         return;
2665         }
2666         /* then try to inherit it from any device */
2667         for_each_netdev(net, link_dev) {
2668                 if (!ipv6_inherit_linklocal(idev, link_dev))
2669                         return;
2670         }
2671         pr_debug("init ip6-ip6: add_linklocal failed\n");
2672 }
2673
2674 /*
2675  * Autoconfigure tunnel with a link-local address so routing protocols,
2676  * DHCPv6, MLD etc. can be run over the virtual link
2677  */
2678
2679 static void addrconf_ip6_tnl_config(struct net_device *dev)
2680 {
2681         struct inet6_dev *idev;
2682
2683         ASSERT_RTNL();
2684
2685         idev = addrconf_add_dev(dev);
2686         if (IS_ERR(idev)) {
2687                 pr_debug("init ip6-ip6: add_dev failed\n");
2688                 return;
2689         }
2690         ip6_tnl_add_linklocal(idev);
2691 }
2692
2693 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2694                            void *data)
2695 {
2696         struct net_device *dev = (struct net_device *) data;
2697         struct inet6_dev *idev = __in6_dev_get(dev);
2698         int run_pending = 0;
2699         int err;
2700
2701         switch (event) {
2702         case NETDEV_REGISTER:
2703                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2704                         idev = ipv6_add_dev(dev);
2705                         if (!idev)
2706                                 return notifier_from_errno(-ENOMEM);
2707                 }
2708                 break;
2709
2710         case NETDEV_UP:
2711         case NETDEV_CHANGE:
2712                 if (dev->flags & IFF_SLAVE)
2713                         break;
2714
2715                 if (event == NETDEV_UP) {
2716                         if (!addrconf_qdisc_ok(dev)) {
2717                                 /* device is not ready yet. */
2718                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2719                                         dev->name);
2720                                 break;
2721                         }
2722
2723                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2724                                 idev = ipv6_add_dev(dev);
2725
2726                         if (idev) {
2727                                 idev->if_flags |= IF_READY;
2728                                 run_pending = 1;
2729                         }
2730                 } else {
2731                         if (!addrconf_qdisc_ok(dev)) {
2732                                 /* device is still not ready. */
2733                                 break;
2734                         }
2735
2736                         if (idev) {
2737                                 if (idev->if_flags & IF_READY)
2738                                         /* device is already configured. */
2739                                         break;
2740                                 idev->if_flags |= IF_READY;
2741                         }
2742
2743                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2744                                 dev->name);
2745
2746                         run_pending = 1;
2747                 }
2748
2749                 switch (dev->type) {
2750 #if IS_ENABLED(CONFIG_IPV6_SIT)
2751                 case ARPHRD_SIT:
2752                         addrconf_sit_config(dev);
2753                         break;
2754 #endif
2755 #if IS_ENABLED(CONFIG_NET_IPGRE)
2756                 case ARPHRD_IPGRE:
2757                         addrconf_gre_config(dev);
2758                         break;
2759 #endif
2760                 case ARPHRD_TUNNEL6:
2761                         addrconf_ip6_tnl_config(dev);
2762                         break;
2763                 case ARPHRD_LOOPBACK:
2764                         init_loopback(dev);
2765                         break;
2766
2767                 default:
2768                         addrconf_dev_config(dev);
2769                         break;
2770                 }
2771
2772                 if (idev) {
2773                         if (run_pending)
2774                                 addrconf_dad_run(idev);
2775
2776                         /*
2777                          * If the MTU changed during the interface down,
2778                          * when the interface up, the changed MTU must be
2779                          * reflected in the idev as well as routers.
2780                          */
2781                         if (idev->cnf.mtu6 != dev->mtu &&
2782                             dev->mtu >= IPV6_MIN_MTU) {
2783                                 rt6_mtu_change(dev, dev->mtu);
2784                                 idev->cnf.mtu6 = dev->mtu;
2785                         }
2786                         idev->tstamp = jiffies;
2787                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2788
2789                         /*
2790                          * If the changed mtu during down is lower than
2791                          * IPV6_MIN_MTU stop IPv6 on this interface.
2792                          */
2793                         if (dev->mtu < IPV6_MIN_MTU)
2794                                 addrconf_ifdown(dev, 1);
2795                 }
2796                 break;
2797
2798         case NETDEV_CHANGEMTU:
2799                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2800                         rt6_mtu_change(dev, dev->mtu);
2801                         idev->cnf.mtu6 = dev->mtu;
2802                         break;
2803                 }
2804
2805                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2806                         idev = ipv6_add_dev(dev);
2807                         if (idev)
2808                                 break;
2809                 }
2810
2811                 /*
2812                  * MTU falled under IPV6_MIN_MTU.
2813                  * Stop IPv6 on this interface.
2814                  */
2815
2816         case NETDEV_DOWN:
2817         case NETDEV_UNREGISTER:
2818                 /*
2819                  *      Remove all addresses from this interface.
2820                  */
2821                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2822                 break;
2823
2824         case NETDEV_CHANGENAME:
2825                 if (idev) {
2826                         snmp6_unregister_dev(idev);
2827                         addrconf_sysctl_unregister(idev);
2828                         addrconf_sysctl_register(idev);
2829                         err = snmp6_register_dev(idev);
2830                         if (err)
2831                                 return notifier_from_errno(err);
2832                 }
2833                 break;
2834
2835         case NETDEV_PRE_TYPE_CHANGE:
2836         case NETDEV_POST_TYPE_CHANGE:
2837                 addrconf_type_change(dev, event);
2838                 break;
2839         }
2840
2841         return NOTIFY_OK;
2842 }
2843
2844 /*
2845  *      addrconf module should be notified of a device going up
2846  */
2847 static struct notifier_block ipv6_dev_notf = {
2848         .notifier_call = addrconf_notify,
2849 };
2850
2851 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2852 {
2853         struct inet6_dev *idev;
2854         ASSERT_RTNL();
2855
2856         idev = __in6_dev_get(dev);
2857
2858         if (event == NETDEV_POST_TYPE_CHANGE)
2859                 ipv6_mc_remap(idev);
2860         else if (event == NETDEV_PRE_TYPE_CHANGE)
2861                 ipv6_mc_unmap(idev);
2862 }
2863
2864 static int addrconf_ifdown(struct net_device *dev, int how)
2865 {
2866         struct net *net = dev_net(dev);
2867         struct inet6_dev *idev;
2868         struct inet6_ifaddr *ifa;
2869         int state, i;
2870
2871         ASSERT_RTNL();
2872
2873         rt6_ifdown(net, dev);
2874         neigh_ifdown(&nd_tbl, dev);
2875
2876         idev = __in6_dev_get(dev);
2877         if (idev == NULL)
2878                 return -ENODEV;
2879
2880         /*
2881          * Step 1: remove reference to ipv6 device from parent device.
2882          *         Do not dev_put!
2883          */
2884         if (how) {
2885                 idev->dead = 1;
2886
2887                 /* protected by rtnl_lock */
2888                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2889
2890                 /* Step 1.5: remove snmp6 entry */
2891                 snmp6_unregister_dev(idev);
2892
2893         }
2894
2895         /* Step 2: clear hash table */
2896         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2897                 struct hlist_head *h = &inet6_addr_lst[i];
2898                 struct hlist_node *n;
2899
2900                 spin_lock_bh(&addrconf_hash_lock);
2901         restart:
2902                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2903                         if (ifa->idev == idev) {
2904                                 hlist_del_init_rcu(&ifa->addr_lst);
2905                                 addrconf_del_timer(ifa);
2906                                 goto restart;
2907                         }
2908                 }
2909                 spin_unlock_bh(&addrconf_hash_lock);
2910         }
2911
2912         write_lock_bh(&idev->lock);
2913
2914         /* Step 2: clear flags for stateless addrconf */
2915         if (!how)
2916                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2917
2918 #ifdef CONFIG_IPV6_PRIVACY
2919         if (how && del_timer(&idev->regen_timer))
2920                 in6_dev_put(idev);
2921
2922         /* Step 3: clear tempaddr list */
2923         while (!list_empty(&idev->tempaddr_list)) {
2924                 ifa = list_first_entry(&idev->tempaddr_list,
2925                                        struct inet6_ifaddr, tmp_list);
2926                 list_del(&ifa->tmp_list);
2927                 write_unlock_bh(&idev->lock);
2928                 spin_lock_bh(&ifa->lock);
2929
2930                 if (ifa->ifpub) {
2931                         in6_ifa_put(ifa->ifpub);
2932                         ifa->ifpub = NULL;
2933                 }
2934                 spin_unlock_bh(&ifa->lock);
2935                 in6_ifa_put(ifa);
2936                 write_lock_bh(&idev->lock);
2937         }
2938 #endif
2939
2940         while (!list_empty(&idev->addr_list)) {
2941                 ifa = list_first_entry(&idev->addr_list,
2942                                        struct inet6_ifaddr, if_list);
2943                 addrconf_del_timer(ifa);
2944
2945                 list_del(&ifa->if_list);
2946
2947                 write_unlock_bh(&idev->lock);
2948
2949                 spin_lock_bh(&ifa->state_lock);
2950                 state = ifa->state;
2951                 ifa->state = INET6_IFADDR_STATE_DEAD;
2952                 spin_unlock_bh(&ifa->state_lock);
2953
2954                 if (state != INET6_IFADDR_STATE_DEAD) {
2955                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2956                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2957                 }
2958                 in6_ifa_put(ifa);
2959
2960                 write_lock_bh(&idev->lock);
2961         }
2962
2963         write_unlock_bh(&idev->lock);
2964
2965         /* Step 5: Discard multicast list */
2966         if (how)
2967                 ipv6_mc_destroy_dev(idev);
2968         else
2969                 ipv6_mc_down(idev);
2970
2971         idev->tstamp = jiffies;
2972
2973         /* Last: Shot the device (if unregistered) */
2974         if (how) {
2975                 addrconf_sysctl_unregister(idev);
2976                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2977                 neigh_ifdown(&nd_tbl, dev);
2978                 in6_dev_put(idev);
2979         }
2980         return 0;
2981 }
2982
2983 static void addrconf_rs_timer(unsigned long data)
2984 {
2985         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2986         struct inet6_dev *idev = ifp->idev;
2987
2988         read_lock(&idev->lock);
2989         if (idev->dead || !(idev->if_flags & IF_READY))
2990                 goto out;
2991
2992         if (idev->cnf.forwarding)
2993                 goto out;
2994
2995         /* Announcement received after solicitation was sent */
2996         if (idev->if_flags & IF_RA_RCVD)
2997                 goto out;
2998
2999         spin_lock(&ifp->lock);
3000         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3001                 /* The wait after the last probe can be shorter */
3002                 addrconf_mod_timer(ifp, AC_RS,
3003                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3004                                    idev->cnf.rtr_solicit_delay :
3005                                    idev->cnf.rtr_solicit_interval);
3006                 spin_unlock(&ifp->lock);
3007
3008                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3009         } else {
3010                 spin_unlock(&ifp->lock);
3011                 /*
3012                  * Note: we do not support deprecated "all on-link"
3013                  * assumption any longer.
3014                  */
3015                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3016         }
3017
3018 out:
3019         read_unlock(&idev->lock);
3020         in6_ifa_put(ifp);
3021 }
3022
3023 /*
3024  *      Duplicate Address Detection
3025  */
3026 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3027 {
3028         unsigned long rand_num;
3029         struct inet6_dev *idev = ifp->idev;
3030
3031         if (ifp->flags & IFA_F_OPTIMISTIC)
3032                 rand_num = 0;
3033         else
3034                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3035
3036         ifp->probes = idev->cnf.dad_transmits;
3037         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3038 }
3039
3040 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3041 {
3042         struct inet6_dev *idev = ifp->idev;
3043         struct net_device *dev = idev->dev;
3044
3045         addrconf_join_solict(dev, &ifp->addr);
3046
3047         net_srandom(ifp->addr.s6_addr32[3]);
3048
3049         read_lock_bh(&idev->lock);
3050         spin_lock(&ifp->lock);
3051         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3052                 goto out;
3053
3054         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3055             idev->cnf.accept_dad < 1 ||
3056             !(ifp->flags&IFA_F_TENTATIVE) ||
3057             ifp->flags & IFA_F_NODAD) {
3058                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3059                 spin_unlock(&ifp->lock);
3060                 read_unlock_bh(&idev->lock);
3061
3062                 addrconf_dad_completed(ifp);
3063                 return;
3064         }
3065
3066         if (!(idev->if_flags & IF_READY)) {
3067                 spin_unlock(&ifp->lock);
3068                 read_unlock_bh(&idev->lock);
3069                 /*
3070                  * If the device is not ready:
3071                  * - keep it tentative if it is a permanent address.
3072                  * - otherwise, kill it.
3073                  */
3074                 in6_ifa_hold(ifp);
3075                 addrconf_dad_stop(ifp, 0);
3076                 return;
3077         }
3078
3079         /*
3080          * Optimistic nodes can start receiving
3081          * Frames right away
3082          */
3083         if (ifp->flags & IFA_F_OPTIMISTIC)
3084                 ip6_ins_rt(ifp->rt);
3085
3086         addrconf_dad_kick(ifp);
3087 out:
3088         spin_unlock(&ifp->lock);
3089         read_unlock_bh(&idev->lock);
3090 }
3091
3092 static void addrconf_dad_timer(unsigned long data)
3093 {
3094         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3095         struct inet6_dev *idev = ifp->idev;
3096         struct in6_addr mcaddr;
3097
3098         if (!ifp->probes && addrconf_dad_end(ifp))
3099                 goto out;
3100
3101         read_lock(&idev->lock);
3102         if (idev->dead || !(idev->if_flags & IF_READY)) {
3103                 read_unlock(&idev->lock);
3104                 goto out;
3105         }
3106
3107         spin_lock(&ifp->lock);
3108         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3109                 spin_unlock(&ifp->lock);
3110                 read_unlock(&idev->lock);
3111                 goto out;
3112         }
3113
3114         if (ifp->probes == 0) {
3115                 /*
3116                  * DAD was successful
3117                  */
3118
3119                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3120                 spin_unlock(&ifp->lock);
3121                 read_unlock(&idev->lock);
3122
3123                 addrconf_dad_completed(ifp);
3124
3125                 goto out;
3126         }
3127
3128         ifp->probes--;
3129         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3130         spin_unlock(&ifp->lock);
3131         read_unlock(&idev->lock);
3132
3133         /* send a neighbour solicitation for our addr */
3134         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3135         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3136 out:
3137         in6_ifa_put(ifp);
3138 }
3139
3140 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3141 {
3142         struct net_device *dev = ifp->idev->dev;
3143
3144         /*
3145          *      Configure the address for reception. Now it is valid.
3146          */
3147
3148         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3149
3150         /* If added prefix is link local and we are prepared to process
3151            router advertisements, start sending router solicitations.
3152          */
3153
3154         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3155              ifp->idev->cnf.accept_ra == 2) &&
3156             ifp->idev->cnf.rtr_solicits > 0 &&
3157             (dev->flags&IFF_LOOPBACK) == 0 &&
3158             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3159                 /*
3160                  *      If a host as already performed a random delay
3161                  *      [...] as part of DAD [...] there is no need
3162                  *      to delay again before sending the first RS
3163                  */
3164                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3165
3166                 spin_lock_bh(&ifp->lock);
3167                 ifp->probes = 1;
3168                 ifp->idev->if_flags |= IF_RS_SENT;
3169                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3170                 spin_unlock_bh(&ifp->lock);
3171         }
3172 }
3173
3174 static void addrconf_dad_run(struct inet6_dev *idev)
3175 {
3176         struct inet6_ifaddr *ifp;
3177
3178         read_lock_bh(&idev->lock);
3179         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3180                 spin_lock(&ifp->lock);
3181                 if (ifp->flags & IFA_F_TENTATIVE &&
3182                     ifp->state == INET6_IFADDR_STATE_DAD)
3183                         addrconf_dad_kick(ifp);
3184                 spin_unlock(&ifp->lock);
3185         }
3186         read_unlock_bh(&idev->lock);
3187 }
3188
3189 #ifdef CONFIG_PROC_FS
3190 struct if6_iter_state {
3191         struct seq_net_private p;
3192         int bucket;
3193         int offset;
3194 };
3195
3196 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3197 {
3198         struct inet6_ifaddr *ifa = NULL;
3199         struct if6_iter_state *state = seq->private;
3200         struct net *net = seq_file_net(seq);
3201         int p = 0;
3202
3203         /* initial bucket if pos is 0 */
3204         if (pos == 0) {
3205                 state->bucket = 0;
3206                 state->offset = 0;
3207         }
3208
3209         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3210                 struct hlist_node *n;
3211                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3212                                          addr_lst) {
3213                         if (!net_eq(dev_net(ifa->idev->dev), net))
3214                                 continue;
3215                         /* sync with offset */
3216                         if (p < state->offset) {
3217                                 p++;
3218                                 continue;
3219                         }
3220                         state->offset++;
3221                         return ifa;
3222                 }
3223
3224                 /* prepare for next bucket */
3225                 state->offset = 0;
3226                 p = 0;
3227         }
3228         return NULL;
3229 }
3230
3231 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3232                                          struct inet6_ifaddr *ifa)
3233 {
3234         struct if6_iter_state *state = seq->private;
3235         struct net *net = seq_file_net(seq);
3236         struct hlist_node *n = &ifa->addr_lst;
3237
3238         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3239                 if (!net_eq(dev_net(ifa->idev->dev), net))
3240                         continue;
3241                 state->offset++;
3242                 return ifa;
3243         }
3244
3245         while (++state->bucket < IN6_ADDR_HSIZE) {
3246                 state->offset = 0;
3247                 hlist_for_each_entry_rcu_bh(ifa, n,
3248                                      &inet6_addr_lst[state->bucket], addr_lst) {
3249                         if (!net_eq(dev_net(ifa->idev->dev), net))
3250                                 continue;
3251                         state->offset++;
3252                         return ifa;
3253                 }
3254         }
3255
3256         return NULL;
3257 }
3258
3259 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3260         __acquires(rcu_bh)
3261 {
3262         rcu_read_lock_bh();
3263         return if6_get_first(seq, *pos);
3264 }
3265
3266 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3267 {
3268         struct inet6_ifaddr *ifa;
3269
3270         ifa = if6_get_next(seq, v);
3271         ++*pos;
3272         return ifa;
3273 }
3274
3275 static void if6_seq_stop(struct seq_file *seq, void *v)
3276         __releases(rcu_bh)
3277 {
3278         rcu_read_unlock_bh();
3279 }
3280
3281 static int if6_seq_show(struct seq_file *seq, void *v)
3282 {
3283         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3284         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3285                    &ifp->addr,
3286                    ifp->idev->dev->ifindex,
3287                    ifp->prefix_len,
3288                    ifp->scope,
3289                    ifp->flags,
3290                    ifp->idev->dev->name);
3291         return 0;
3292 }
3293
3294 static const struct seq_operations if6_seq_ops = {
3295         .start  = if6_seq_start,
3296         .next   = if6_seq_next,
3297         .show   = if6_seq_show,
3298         .stop   = if6_seq_stop,
3299 };
3300
3301 static int if6_seq_open(struct inode *inode, struct file *file)
3302 {
3303         return seq_open_net(inode, file, &if6_seq_ops,
3304                             sizeof(struct if6_iter_state));
3305 }
3306
3307 static const struct file_operations if6_fops = {
3308         .owner          = THIS_MODULE,
3309         .open           = if6_seq_open,
3310         .read           = seq_read,
3311         .llseek         = seq_lseek,
3312         .release        = seq_release_net,
3313 };
3314
3315 static int __net_init if6_proc_net_init(struct net *net)
3316 {
3317         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3318                 return -ENOMEM;
3319         return 0;
3320 }
3321
3322 static void __net_exit if6_proc_net_exit(struct net *net)
3323 {
3324        proc_net_remove(net, "if_inet6");
3325 }
3326
3327 static struct pernet_operations if6_proc_net_ops = {
3328        .init = if6_proc_net_init,
3329        .exit = if6_proc_net_exit,
3330 };
3331
3332 int __init if6_proc_init(void)
3333 {
3334         return register_pernet_subsys(&if6_proc_net_ops);
3335 }
3336
3337 void if6_proc_exit(void)
3338 {
3339         unregister_pernet_subsys(&if6_proc_net_ops);
3340 }
3341 #endif  /* CONFIG_PROC_FS */
3342
3343 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3344 /* Check if address is a home address configured on any interface. */
3345 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3346 {
3347         int ret = 0;
3348         struct inet6_ifaddr *ifp = NULL;
3349         struct hlist_node *n;
3350         unsigned int hash = inet6_addr_hash(addr);
3351
3352         rcu_read_lock_bh();
3353         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3354                 if (!net_eq(dev_net(ifp->idev->dev), net))
3355                         continue;
3356                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3357                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3358                         ret = 1;
3359                         break;
3360                 }
3361         }
3362         rcu_read_unlock_bh();
3363         return ret;
3364 }
3365 #endif
3366
3367 /*
3368  *      Periodic address status verification
3369  */
3370
3371 static void addrconf_verify(unsigned long foo)
3372 {
3373         unsigned long now, next, next_sec, next_sched;
3374         struct inet6_ifaddr *ifp;
3375         struct hlist_node *node;
3376         int i;
3377
3378         rcu_read_lock_bh();
3379         spin_lock(&addrconf_verify_lock);
3380         now = jiffies;
3381         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3382
3383         del_timer(&addr_chk_timer);
3384
3385         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3386 restart:
3387                 hlist_for_each_entry_rcu_bh(ifp, node,
3388                                          &inet6_addr_lst[i], addr_lst) {
3389                         unsigned long age;
3390
3391                         if (ifp->flags & IFA_F_PERMANENT)
3392                                 continue;
3393
3394                         spin_lock(&ifp->lock);
3395                         /* We try to batch several events at once. */
3396                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3397
3398                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3399                             age >= ifp->valid_lft) {
3400                                 spin_unlock(&ifp->lock);
3401                                 in6_ifa_hold(ifp);
3402                                 ipv6_del_addr(ifp);
3403                                 goto restart;
3404                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3405                                 spin_unlock(&ifp->lock);
3406                                 continue;
3407                         } else if (age >= ifp->prefered_lft) {
3408                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3409                                 int deprecate = 0;
3410
3411                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3412                                         deprecate = 1;
3413                                         ifp->flags |= IFA_F_DEPRECATED;
3414                                 }
3415
3416                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3417                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3418
3419                                 spin_unlock(&ifp->lock);
3420
3421                                 if (deprecate) {
3422                                         in6_ifa_hold(ifp);
3423
3424                                         ipv6_ifa_notify(0, ifp);
3425                                         in6_ifa_put(ifp);
3426                                         goto restart;
3427                                 }
3428 #ifdef CONFIG_IPV6_PRIVACY
3429                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3430                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3431                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3432                                         ifp->idev->cnf.dad_transmits *
3433                                         ifp->idev->nd_parms->retrans_time / HZ;
3434
3435                                 if (age >= ifp->prefered_lft - regen_advance) {
3436                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3437                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3438                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3439                                         if (!ifp->regen_count && ifpub) {
3440                                                 ifp->regen_count++;
3441                                                 in6_ifa_hold(ifp);
3442                                                 in6_ifa_hold(ifpub);
3443                                                 spin_unlock(&ifp->lock);
3444
3445                                                 spin_lock(&ifpub->lock);
3446                                                 ifpub->regen_count = 0;
3447                                                 spin_unlock(&ifpub->lock);
3448                                                 ipv6_create_tempaddr(ifpub, ifp);
3449                                                 in6_ifa_put(ifpub);
3450                                                 in6_ifa_put(ifp);
3451                                                 goto restart;
3452                                         }
3453                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3454                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3455                                 spin_unlock(&ifp->lock);
3456 #endif
3457                         } else {
3458                                 /* ifp->prefered_lft <= ifp->valid_lft */
3459                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3460                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3461                                 spin_unlock(&ifp->lock);
3462                         }
3463                 }
3464         }
3465
3466         next_sec = round_jiffies_up(next);
3467         next_sched = next;
3468
3469         /* If rounded timeout is accurate enough, accept it. */
3470         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3471                 next_sched = next_sec;
3472
3473         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3474         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3475                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3476
3477         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3478               now, next, next_sec, next_sched));
3479
3480         addr_chk_timer.expires = next_sched;
3481         add_timer(&addr_chk_timer);
3482         spin_unlock(&addrconf_verify_lock);
3483         rcu_read_unlock_bh();
3484 }
3485
3486 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3487 {
3488         struct in6_addr *pfx = NULL;
3489
3490         if (addr)
3491                 pfx = nla_data(addr);
3492
3493         if (local) {
3494                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3495                         pfx = NULL;
3496                 else
3497                         pfx = nla_data(local);
3498         }
3499
3500         return pfx;
3501 }
3502
3503 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3504         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3505         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3506         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3507 };
3508
3509 static int
3510 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3511 {
3512         struct net *net = sock_net(skb->sk);
3513         struct ifaddrmsg *ifm;
3514         struct nlattr *tb[IFA_MAX+1];
3515         struct in6_addr *pfx;
3516         int err;
3517
3518         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3519         if (err < 0)
3520                 return err;
3521
3522         ifm = nlmsg_data(nlh);
3523         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3524         if (pfx == NULL)
3525                 return -EINVAL;
3526
3527         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3528 }
3529
3530 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3531                              u32 prefered_lft, u32 valid_lft)
3532 {
3533         u32 flags;
3534         clock_t expires;
3535         unsigned long timeout;
3536
3537         if (!valid_lft || (prefered_lft > valid_lft))
3538                 return -EINVAL;
3539
3540         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3541         if (addrconf_finite_timeout(timeout)) {
3542                 expires = jiffies_to_clock_t(timeout * HZ);
3543                 valid_lft = timeout;
3544                 flags = RTF_EXPIRES;
3545         } else {
3546                 expires = 0;
3547                 flags = 0;
3548                 ifa_flags |= IFA_F_PERMANENT;
3549         }
3550
3551         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3552         if (addrconf_finite_timeout(timeout)) {
3553                 if (timeout == 0)
3554                         ifa_flags |= IFA_F_DEPRECATED;
3555                 prefered_lft = timeout;
3556         }
3557
3558         spin_lock_bh(&ifp->lock);
3559         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3560         ifp->tstamp = jiffies;
3561         ifp->valid_lft = valid_lft;
3562         ifp->prefered_lft = prefered_lft;
3563
3564         spin_unlock_bh(&ifp->lock);
3565         if (!(ifp->flags&IFA_F_TENTATIVE))
3566                 ipv6_ifa_notify(0, ifp);
3567
3568         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3569                               expires, flags);
3570         addrconf_verify(0);
3571
3572         return 0;
3573 }
3574
3575 static int
3576 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3577 {
3578         struct net *net = sock_net(skb->sk);
3579         struct ifaddrmsg *ifm;
3580         struct nlattr *tb[IFA_MAX+1];
3581         struct in6_addr *pfx;
3582         struct inet6_ifaddr *ifa;
3583         struct net_device *dev;
3584         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3585         u8 ifa_flags;
3586         int err;
3587
3588         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3589         if (err < 0)
3590                 return err;
3591
3592         ifm = nlmsg_data(nlh);
3593         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3594         if (pfx == NULL)
3595                 return -EINVAL;
3596
3597         if (tb[IFA_CACHEINFO]) {
3598                 struct ifa_cacheinfo *ci;
3599
3600                 ci = nla_data(tb[IFA_CACHEINFO]);
3601                 valid_lft = ci->ifa_valid;
3602                 preferred_lft = ci->ifa_prefered;
3603         } else {
3604                 preferred_lft = INFINITY_LIFE_TIME;
3605                 valid_lft = INFINITY_LIFE_TIME;
3606         }
3607
3608         dev =  __dev_get_by_index(net, ifm->ifa_index);
3609         if (dev == NULL)
3610                 return -ENODEV;
3611
3612         /* We ignore other flags so far. */
3613         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3614
3615         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3616         if (ifa == NULL) {
3617                 /*
3618                  * It would be best to check for !NLM_F_CREATE here but
3619                  * userspace alreay relies on not having to provide this.
3620                  */
3621                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3622                                       ifm->ifa_prefixlen, ifa_flags,
3623                                       preferred_lft, valid_lft);
3624         }
3625
3626         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3627             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3628                 err = -EEXIST;
3629         else
3630                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3631
3632         in6_ifa_put(ifa);
3633
3634         return err;
3635 }
3636
3637 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3638                           u8 scope, int ifindex)
3639 {
3640         struct ifaddrmsg *ifm;
3641
3642         ifm = nlmsg_data(nlh);
3643         ifm->ifa_family = AF_INET6;
3644         ifm->ifa_prefixlen = prefixlen;
3645         ifm->ifa_flags = flags;
3646         ifm->ifa_scope = scope;
3647         ifm->ifa_index = ifindex;
3648 }
3649
3650 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3651                          unsigned long tstamp, u32 preferred, u32 valid)
3652 {
3653         struct ifa_cacheinfo ci;
3654
3655         ci.cstamp = cstamp_delta(cstamp);
3656         ci.tstamp = cstamp_delta(tstamp);
3657         ci.ifa_prefered = preferred;
3658         ci.ifa_valid = valid;
3659
3660         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3661 }
3662
3663 static inline int rt_scope(int ifa_scope)
3664 {
3665         if (ifa_scope & IFA_HOST)
3666                 return RT_SCOPE_HOST;
3667         else if (ifa_scope & IFA_LINK)
3668                 return RT_SCOPE_LINK;
3669         else if (ifa_scope & IFA_SITE)
3670                 return RT_SCOPE_SITE;
3671         else
3672                 return RT_SCOPE_UNIVERSE;
3673 }
3674
3675 static inline int inet6_ifaddr_msgsize(void)
3676 {
3677         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3678                + nla_total_size(16) /* IFA_ADDRESS */
3679                + nla_total_size(sizeof(struct ifa_cacheinfo));
3680 }
3681
3682 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3683                              u32 portid, u32 seq, int event, unsigned int flags)
3684 {
3685         struct nlmsghdr  *nlh;
3686         u32 preferred, valid;
3687
3688         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3689         if (nlh == NULL)
3690                 return -EMSGSIZE;
3691
3692         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3693                       ifa->idev->dev->ifindex);
3694
3695         if (!(ifa->flags&IFA_F_PERMANENT)) {
3696                 preferred = ifa->prefered_lft;
3697                 valid = ifa->valid_lft;
3698                 if (preferred != INFINITY_LIFE_TIME) {
3699                         long tval = (jiffies - ifa->tstamp)/HZ;
3700                         if (preferred > tval)
3701                                 preferred -= tval;
3702                         else
3703                                 preferred = 0;
3704                         if (valid != INFINITY_LIFE_TIME) {
3705                                 if (valid > tval)
3706                                         valid -= tval;
3707                                 else
3708                                         valid = 0;
3709                         }
3710                 }
3711         } else {
3712                 preferred = INFINITY_LIFE_TIME;
3713                 valid = INFINITY_LIFE_TIME;
3714         }
3715
3716         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3717             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3718                 nlmsg_cancel(skb, nlh);
3719                 return -EMSGSIZE;
3720         }
3721
3722         return nlmsg_end(skb, nlh);
3723 }
3724
3725 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3726                                 u32 portid, u32 seq, int event, u16 flags)
3727 {
3728         struct nlmsghdr  *nlh;
3729         u8 scope = RT_SCOPE_UNIVERSE;
3730         int ifindex = ifmca->idev->dev->ifindex;
3731
3732         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3733                 scope = RT_SCOPE_SITE;
3734
3735         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3736         if (nlh == NULL)
3737                 return -EMSGSIZE;
3738
3739         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3740         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3741             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3742                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3743                 nlmsg_cancel(skb, nlh);
3744                 return -EMSGSIZE;
3745         }
3746
3747         return nlmsg_end(skb, nlh);
3748 }
3749
3750 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3751                                 u32 portid, u32 seq, int event, unsigned int flags)
3752 {
3753         struct nlmsghdr  *nlh;
3754         u8 scope = RT_SCOPE_UNIVERSE;
3755         int ifindex = ifaca->aca_idev->dev->ifindex;
3756
3757         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3758                 scope = RT_SCOPE_SITE;
3759
3760         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3761         if (nlh == NULL)
3762                 return -EMSGSIZE;
3763
3764         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3765         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3766             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3767                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3768                 nlmsg_cancel(skb, nlh);
3769                 return -EMSGSIZE;
3770         }
3771
3772         return nlmsg_end(skb, nlh);
3773 }
3774
3775 enum addr_type_t {
3776         UNICAST_ADDR,
3777         MULTICAST_ADDR,
3778         ANYCAST_ADDR,
3779 };
3780
3781 /* called with rcu_read_lock() */
3782 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3783                           struct netlink_callback *cb, enum addr_type_t type,
3784                           int s_ip_idx, int *p_ip_idx)
3785 {
3786         struct ifmcaddr6 *ifmca;
3787         struct ifacaddr6 *ifaca;
3788         int err = 1;
3789         int ip_idx = *p_ip_idx;
3790
3791         read_lock_bh(&idev->lock);
3792         switch (type) {
3793         case UNICAST_ADDR: {
3794                 struct inet6_ifaddr *ifa;
3795
3796                 /* unicast address incl. temp addr */
3797                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3798                         if (++ip_idx < s_ip_idx)
3799                                 continue;
3800                         err = inet6_fill_ifaddr(skb, ifa,
3801                                                 NETLINK_CB(cb->skb).portid,
3802                                                 cb->nlh->nlmsg_seq,
3803                                                 RTM_NEWADDR,
3804                                                 NLM_F_MULTI);
3805                         if (err <= 0)
3806                                 break;
3807                 }
3808                 break;
3809         }
3810         case MULTICAST_ADDR:
3811                 /* multicast address */
3812                 for (ifmca = idev->mc_list; ifmca;
3813                      ifmca = ifmca->next, ip_idx++) {
3814                         if (ip_idx < s_ip_idx)
3815                                 continue;
3816                         err = inet6_fill_ifmcaddr(skb, ifmca,
3817                                                   NETLINK_CB(cb->skb).portid,
3818                                                   cb->nlh->nlmsg_seq,
3819                                                   RTM_GETMULTICAST,
3820                                                   NLM_F_MULTI);
3821                         if (err <= 0)
3822                                 break;
3823                 }
3824                 break;
3825         case ANYCAST_ADDR:
3826                 /* anycast address */
3827                 for (ifaca = idev->ac_list; ifaca;
3828                      ifaca = ifaca->aca_next, ip_idx++) {
3829                         if (ip_idx < s_ip_idx)
3830                                 continue;
3831                         err = inet6_fill_ifacaddr(skb, ifaca,
3832                                                   NETLINK_CB(cb->skb).portid,
3833                                                   cb->nlh->nlmsg_seq,
3834                                                   RTM_GETANYCAST,
3835                                                   NLM_F_MULTI);
3836                         if (err <= 0)
3837                                 break;
3838                 }
3839                 break;
3840         default:
3841                 break;
3842         }
3843         read_unlock_bh(&idev->lock);
3844         *p_ip_idx = ip_idx;
3845         return err;
3846 }
3847
3848 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3849                            enum addr_type_t type)
3850 {
3851         struct net *net = sock_net(skb->sk);
3852         int h, s_h;
3853         int idx, ip_idx;
3854         int s_idx, s_ip_idx;
3855         struct net_device *dev;
3856         struct inet6_dev *idev;
3857         struct hlist_head *head;
3858         struct hlist_node *node;
3859
3860         s_h = cb->args[0];
3861         s_idx = idx = cb->args[1];
3862         s_ip_idx = ip_idx = cb->args[2];
3863
3864         rcu_read_lock();
3865         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3866                 idx = 0;
3867                 head = &net->dev_index_head[h];
3868                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3869                         if (idx < s_idx)
3870                                 goto cont;
3871                         if (h > s_h || idx > s_idx)
3872                                 s_ip_idx = 0;
3873                         ip_idx = 0;
3874                         idev = __in6_dev_get(dev);
3875                         if (!idev)
3876                                 goto cont;
3877
3878                         if (in6_dump_addrs(idev, skb, cb, type,
3879                                            s_ip_idx, &ip_idx) <= 0)
3880                                 goto done;
3881 cont:
3882                         idx++;
3883                 }
3884         }
3885 done:
3886         rcu_read_unlock();
3887         cb->args[0] = h;
3888         cb->args[1] = idx;
3889         cb->args[2] = ip_idx;
3890
3891         return skb->len;
3892 }
3893
3894 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3895 {
3896         enum addr_type_t type = UNICAST_ADDR;
3897
3898         return inet6_dump_addr(skb, cb, type);
3899 }
3900
3901 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3902 {
3903         enum addr_type_t type = MULTICAST_ADDR;
3904
3905         return inet6_dump_addr(skb, cb, type);
3906 }
3907
3908
3909 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3910 {
3911         enum addr_type_t type = ANYCAST_ADDR;
3912
3913         return inet6_dump_addr(skb, cb, type);
3914 }
3915
3916 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3917                              void *arg)
3918 {
3919         struct net *net = sock_net(in_skb->sk);
3920         struct ifaddrmsg *ifm;
3921         struct nlattr *tb[IFA_MAX+1];
3922         struct in6_addr *addr = NULL;
3923         struct net_device *dev = NULL;
3924         struct inet6_ifaddr *ifa;
3925         struct sk_buff *skb;
3926         int err;
3927
3928         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3929         if (err < 0)
3930                 goto errout;
3931
3932         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3933         if (addr == NULL) {
3934                 err = -EINVAL;
3935                 goto errout;
3936         }
3937
3938         ifm = nlmsg_data(nlh);
3939         if (ifm->ifa_index)
3940                 dev = __dev_get_by_index(net, ifm->ifa_index);
3941
3942         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3943         if (!ifa) {
3944                 err = -EADDRNOTAVAIL;
3945                 goto errout;
3946         }
3947
3948         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3949         if (!skb) {
3950                 err = -ENOBUFS;
3951                 goto errout_ifa;
3952         }
3953
3954         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3955                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3956         if (err < 0) {
3957                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3958                 WARN_ON(err == -EMSGSIZE);
3959                 kfree_skb(skb);
3960                 goto errout_ifa;
3961         }
3962         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3963 errout_ifa:
3964         in6_ifa_put(ifa);
3965 errout:
3966         return err;
3967 }
3968
3969 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3970 {
3971         struct sk_buff *skb;
3972         struct net *net = dev_net(ifa->idev->dev);
3973         int err = -ENOBUFS;
3974
3975         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3976         if (skb == NULL)
3977                 goto errout;
3978
3979         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3980         if (err < 0) {
3981                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3982                 WARN_ON(err == -EMSGSIZE);
3983                 kfree_skb(skb);
3984                 goto errout;
3985         }
3986         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3987         return;
3988 errout:
3989         if (err < 0)
3990                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3991 }
3992
3993 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3994                                 __s32 *array, int bytes)
3995 {
3996         BUG_ON(bytes < (DEVCONF_MAX * 4));
3997
3998         memset(array, 0, bytes);
3999         array[DEVCONF_FORWARDING] = cnf->forwarding;
4000         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4001         array[DEVCONF_MTU6] = cnf->mtu6;
4002         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4003         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4004         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4005         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4006         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4007         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4008                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4009         array[DEVCONF_RTR_SOLICIT_DELAY] =
4010                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4011         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4012 #ifdef CONFIG_IPV6_PRIVACY
4013         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4014         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4015         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4016         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4017         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4018 #endif
4019         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4020         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4021         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4022 #ifdef CONFIG_IPV6_ROUTER_PREF
4023         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4024         array[DEVCONF_RTR_PROBE_INTERVAL] =
4025                 jiffies_to_msecs(cnf->rtr_probe_interval);
4026 #ifdef CONFIG_IPV6_ROUTE_INFO
4027         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4028 #endif
4029 #endif
4030         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4031         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4032 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4033         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4034 #endif
4035 #ifdef CONFIG_IPV6_MROUTE
4036         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4037 #endif
4038         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4039         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4040         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4041 }
4042
4043 static inline size_t inet6_ifla6_size(void)
4044 {
4045         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4046              + nla_total_size(sizeof(struct ifla_cacheinfo))
4047              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4048              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4049              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4050 }
4051
4052 static inline size_t inet6_if_nlmsg_size(void)
4053 {
4054         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4055                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4056                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4057                + nla_total_size(4) /* IFLA_MTU */
4058                + nla_total_size(4) /* IFLA_LINK */
4059                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4060 }
4061
4062 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4063                                       int items, int bytes)
4064 {
4065         int i;
4066         int pad = bytes - sizeof(u64) * items;
4067         BUG_ON(pad < 0);
4068
4069         /* Use put_unaligned() because stats may not be aligned for u64. */
4070         put_unaligned(items, &stats[0]);
4071         for (i = 1; i < items; i++)
4072                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4073
4074         memset(&stats[items], 0, pad);
4075 }
4076
4077 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4078                                       int items, int bytes, size_t syncpoff)
4079 {
4080         int i;
4081         int pad = bytes - sizeof(u64) * items;
4082         BUG_ON(pad < 0);
4083
4084         /* Use put_unaligned() because stats may not be aligned for u64. */
4085         put_unaligned(items, &stats[0]);
4086         for (i = 1; i < items; i++)
4087                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4088
4089         memset(&stats[items], 0, pad);
4090 }
4091
4092 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4093                              int bytes)
4094 {
4095         switch (attrtype) {
4096         case IFLA_INET6_STATS:
4097                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4098                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4099                 break;
4100         case IFLA_INET6_ICMP6STATS:
4101                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4102                 break;
4103         }
4104 }
4105
4106 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4107 {
4108         struct nlattr *nla;
4109         struct ifla_cacheinfo ci;
4110
4111         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4112                 goto nla_put_failure;
4113         ci.max_reasm_len = IPV6_MAXPLEN;
4114         ci.tstamp = cstamp_delta(idev->tstamp);
4115         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4116         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4117         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4118                 goto nla_put_failure;
4119         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4120         if (nla == NULL)
4121                 goto nla_put_failure;
4122         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4123
4124         /* XXX - MC not implemented */
4125
4126         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4127         if (nla == NULL)
4128                 goto nla_put_failure;
4129         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4130
4131         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4132         if (nla == NULL)
4133                 goto nla_put_failure;
4134         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4135
4136         return 0;
4137
4138 nla_put_failure:
4139         return -EMSGSIZE;
4140 }
4141
4142 static size_t inet6_get_link_af_size(const struct net_device *dev)
4143 {
4144         if (!__in6_dev_get(dev))
4145                 return 0;
4146
4147         return inet6_ifla6_size();
4148 }
4149
4150 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4151 {
4152         struct inet6_dev *idev = __in6_dev_get(dev);
4153
4154         if (!idev)
4155                 return -ENODATA;
4156
4157         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4158                 return -EMSGSIZE;
4159
4160         return 0;
4161 }
4162
4163 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4164                              u32 portid, u32 seq, int event, unsigned int flags)
4165 {
4166         struct net_device *dev = idev->dev;
4167         struct ifinfomsg *hdr;
4168         struct nlmsghdr *nlh;
4169         void *protoinfo;
4170
4171         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4172         if (nlh == NULL)
4173                 return -EMSGSIZE;
4174
4175         hdr = nlmsg_data(nlh);
4176         hdr->ifi_family = AF_INET6;
4177         hdr->__ifi_pad = 0;
4178         hdr->ifi_type = dev->type;
4179         hdr->ifi_index = dev->ifindex;
4180         hdr->ifi_flags = dev_get_flags(dev);
4181         hdr->ifi_change = 0;
4182
4183         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4184             (dev->addr_len &&
4185              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4186             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4187             (dev->ifindex != dev->iflink &&
4188              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4189                 goto nla_put_failure;
4190         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4191         if (protoinfo == NULL)
4192                 goto nla_put_failure;
4193
4194         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4195                 goto nla_put_failure;
4196
4197         nla_nest_end(skb, protoinfo);
4198         return nlmsg_end(skb, nlh);
4199
4200 nla_put_failure:
4201         nlmsg_cancel(skb, nlh);
4202         return -EMSGSIZE;
4203 }
4204
4205 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4206 {
4207         struct net *net = sock_net(skb->sk);
4208         int h, s_h;
4209         int idx = 0, s_idx;
4210         struct net_device *dev;
4211         struct inet6_dev *idev;
4212         struct hlist_head *head;
4213         struct hlist_node *node;
4214
4215         s_h = cb->args[0];
4216         s_idx = cb->args[1];
4217
4218         rcu_read_lock();
4219         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4220                 idx = 0;
4221                 head = &net->dev_index_head[h];
4222                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4223                         if (idx < s_idx)
4224                                 goto cont;
4225                         idev = __in6_dev_get(dev);
4226                         if (!idev)
4227                                 goto cont;
4228                         if (inet6_fill_ifinfo(skb, idev,
4229                                               NETLINK_CB(cb->skb).portid,
4230                                               cb->nlh->nlmsg_seq,
4231                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4232                                 goto out;
4233 cont:
4234                         idx++;
4235                 }
4236         }
4237 out:
4238         rcu_read_unlock();
4239         cb->args[1] = idx;
4240         cb->args[0] = h;
4241
4242         return skb->len;
4243 }
4244
4245 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4246 {
4247         struct sk_buff *skb;
4248         struct net *net = dev_net(idev->dev);
4249         int err = -ENOBUFS;
4250
4251         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4252         if (skb == NULL)
4253                 goto errout;
4254
4255         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4256         if (err < 0) {
4257                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4258                 WARN_ON(err == -EMSGSIZE);
4259                 kfree_skb(skb);
4260                 goto errout;
4261         }
4262         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4263         return;
4264 errout:
4265         if (err < 0)
4266                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4267 }
4268
4269 static inline size_t inet6_prefix_nlmsg_size(void)
4270 {
4271         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4272                + nla_total_size(sizeof(struct in6_addr))
4273                + nla_total_size(sizeof(struct prefix_cacheinfo));
4274 }
4275
4276 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4277                              struct prefix_info *pinfo, u32 portid, u32 seq,
4278                              int event, unsigned int flags)
4279 {
4280         struct prefixmsg *pmsg;
4281         struct nlmsghdr *nlh;
4282         struct prefix_cacheinfo ci;
4283
4284         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4285         if (nlh == NULL)
4286                 return -EMSGSIZE;
4287
4288         pmsg = nlmsg_data(nlh);
4289         pmsg->prefix_family = AF_INET6;
4290         pmsg->prefix_pad1 = 0;
4291         pmsg->prefix_pad2 = 0;
4292         pmsg->prefix_ifindex = idev->dev->ifindex;
4293         pmsg->prefix_len = pinfo->prefix_len;
4294         pmsg->prefix_type = pinfo->type;
4295         pmsg->prefix_pad3 = 0;
4296         pmsg->prefix_flags = 0;
4297         if (pinfo->onlink)
4298                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4299         if (pinfo->autoconf)
4300                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4301
4302         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4303                 goto nla_put_failure;
4304         ci.preferred_time = ntohl(pinfo->prefered);
4305         ci.valid_time = ntohl(pinfo->valid);
4306         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4307                 goto nla_put_failure;
4308         return nlmsg_end(skb, nlh);
4309
4310 nla_put_failure:
4311         nlmsg_cancel(skb, nlh);
4312         return -EMSGSIZE;
4313 }
4314
4315 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4316                          struct prefix_info *pinfo)
4317 {
4318         struct sk_buff *skb;
4319         struct net *net = dev_net(idev->dev);
4320         int err = -ENOBUFS;
4321
4322         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4323         if (skb == NULL)
4324                 goto errout;
4325
4326         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4327         if (err < 0) {
4328                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4329                 WARN_ON(err == -EMSGSIZE);
4330                 kfree_skb(skb);
4331                 goto errout;
4332         }
4333         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4334         return;
4335 errout:
4336         if (err < 0)
4337                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4338 }
4339
4340 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4341 {
4342         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4343
4344         switch (event) {
4345         case RTM_NEWADDR:
4346                 /*
4347                  * If the address was optimistic
4348                  * we inserted the route at the start of
4349                  * our DAD process, so we don't need
4350                  * to do it again
4351                  */
4352                 if (!(ifp->rt->rt6i_node))
4353                         ip6_ins_rt(ifp->rt);
4354                 if (ifp->idev->cnf.forwarding)
4355                         addrconf_join_anycast(ifp);
4356                 break;
4357         case RTM_DELADDR:
4358                 if (ifp->idev->cnf.forwarding)
4359                         addrconf_leave_anycast(ifp);
4360                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4361                 dst_hold(&ifp->rt->dst);
4362
4363                 if (ip6_del_rt(ifp->rt))
4364                         dst_free(&ifp->rt->dst);
4365                 break;
4366         }
4367 }
4368
4369 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4370 {
4371         rcu_read_lock_bh();
4372         if (likely(ifp->idev->dead == 0))
4373                 __ipv6_ifa_notify(event, ifp);
4374         rcu_read_unlock_bh();
4375 }
4376
4377 #ifdef CONFIG_SYSCTL
4378
4379 static
4380 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4381                            void __user *buffer, size_t *lenp, loff_t *ppos)
4382 {
4383         int *valp = ctl->data;
4384         int val = *valp;
4385         loff_t pos = *ppos;
4386         ctl_table lctl;
4387         int ret;
4388
4389         /*
4390          * ctl->data points to idev->cnf.forwarding, we should
4391          * not modify it until we get the rtnl lock.
4392          */
4393         lctl = *ctl;
4394         lctl.data = &val;
4395
4396         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4397
4398         if (write)
4399                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4400         if (ret)
4401                 *ppos = pos;
4402         return ret;
4403 }
4404
4405 static void dev_disable_change(struct inet6_dev *idev)
4406 {
4407         if (!idev || !idev->dev)
4408                 return;
4409
4410         if (idev->cnf.disable_ipv6)
4411                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4412         else
4413                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4414 }
4415
4416 static void addrconf_disable_change(struct net *net, __s32 newf)
4417 {
4418         struct net_device *dev;
4419         struct inet6_dev *idev;
4420
4421         rcu_read_lock();
4422         for_each_netdev_rcu(net, dev) {
4423                 idev = __in6_dev_get(dev);
4424                 if (idev) {
4425                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4426                         idev->cnf.disable_ipv6 = newf;
4427                         if (changed)
4428                                 dev_disable_change(idev);
4429                 }
4430         }
4431         rcu_read_unlock();
4432 }
4433
4434 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4435 {
4436         struct net *net;
4437         int old;
4438
4439         if (!rtnl_trylock())
4440                 return restart_syscall();
4441
4442         net = (struct net *)table->extra2;
4443         old = *p;
4444         *p = newf;
4445
4446         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4447                 rtnl_unlock();
4448                 return 0;
4449         }
4450
4451         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4452                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4453                 addrconf_disable_change(net, newf);
4454         } else if ((!newf) ^ (!old))
4455                 dev_disable_change((struct inet6_dev *)table->extra1);
4456
4457         rtnl_unlock();
4458         return 0;
4459 }
4460
4461 static
4462 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4463                             void __user *buffer, size_t *lenp, loff_t *ppos)
4464 {
4465         int *valp = ctl->data;
4466         int val = *valp;
4467         loff_t pos = *ppos;
4468         ctl_table lctl;
4469         int ret;
4470
4471         /*
4472          * ctl->data points to idev->cnf.disable_ipv6, we should
4473          * not modify it until we get the rtnl lock.
4474          */
4475         lctl = *ctl;
4476         lctl.data = &val;
4477
4478         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4479
4480         if (write)
4481                 ret = addrconf_disable_ipv6(ctl, valp, val);
4482         if (ret)
4483                 *ppos = pos;
4484         return ret;
4485 }
4486
4487 static struct addrconf_sysctl_table
4488 {
4489         struct ctl_table_header *sysctl_header;
4490         ctl_table addrconf_vars[DEVCONF_MAX+1];
4491 } addrconf_sysctl __read_mostly = {
4492         .sysctl_header = NULL,
4493         .addrconf_vars = {
4494                 {
4495                         .procname       = "forwarding",
4496                         .data           = &ipv6_devconf.forwarding,
4497                         .maxlen         = sizeof(int),
4498                         .mode           = 0644,
4499                         .proc_handler   = addrconf_sysctl_forward,
4500                 },
4501                 {
4502                         .procname       = "hop_limit",
4503                         .data           = &ipv6_devconf.hop_limit,
4504                         .maxlen         = sizeof(int),
4505                         .mode           = 0644,
4506                         .proc_handler   = proc_dointvec,
4507                 },
4508                 {
4509                         .procname       = "mtu",
4510                         .data           = &ipv6_devconf.mtu6,
4511                         .maxlen         = sizeof(int),
4512                         .mode           = 0644,
4513                         .proc_handler   = proc_dointvec,
4514                 },
4515                 {
4516                         .procname       = "accept_ra",
4517                         .data           = &ipv6_devconf.accept_ra,
4518                         .maxlen         = sizeof(int),
4519                         .mode           = 0644,
4520                         .proc_handler   = proc_dointvec,
4521                 },
4522                 {
4523                         .procname       = "accept_redirects",
4524                         .data           = &ipv6_devconf.accept_redirects,
4525                         .maxlen         = sizeof(int),
4526                         .mode           = 0644,
4527                         .proc_handler   = proc_dointvec,
4528                 },
4529                 {
4530                         .procname       = "autoconf",
4531                         .data           = &ipv6_devconf.autoconf,
4532                         .maxlen         = sizeof(int),
4533                         .mode           = 0644,
4534                         .proc_handler   = proc_dointvec,
4535                 },
4536                 {
4537                         .procname       = "dad_transmits",
4538                         .data           = &ipv6_devconf.dad_transmits,
4539                         .maxlen         = sizeof(int),
4540                         .mode           = 0644,
4541                         .proc_handler   = proc_dointvec,
4542                 },
4543                 {
4544                         .procname       = "router_solicitations",
4545                         .data           = &ipv6_devconf.rtr_solicits,
4546                         .maxlen         = sizeof(int),
4547                         .mode           = 0644,
4548                         .proc_handler   = proc_dointvec,
4549                 },
4550                 {
4551                         .procname       = "router_solicitation_interval",
4552                         .data           = &ipv6_devconf.rtr_solicit_interval,
4553                         .maxlen         = sizeof(int),
4554                         .mode           = 0644,
4555                         .proc_handler   = proc_dointvec_jiffies,
4556                 },
4557                 {
4558                         .procname       = "router_solicitation_delay",
4559                         .data           = &ipv6_devconf.rtr_solicit_delay,
4560                         .maxlen         = sizeof(int),
4561                         .mode           = 0644,
4562                         .proc_handler   = proc_dointvec_jiffies,
4563                 },
4564                 {
4565                         .procname       = "force_mld_version",
4566                         .data           = &ipv6_devconf.force_mld_version,
4567                         .maxlen         = sizeof(int),
4568                         .mode           = 0644,
4569                         .proc_handler   = proc_dointvec,
4570                 },
4571 #ifdef CONFIG_IPV6_PRIVACY
4572                 {
4573                         .procname       = "use_tempaddr",
4574                         .data           = &ipv6_devconf.use_tempaddr,
4575                         .maxlen         = sizeof(int),
4576                         .mode           = 0644,
4577                         .proc_handler   = proc_dointvec,
4578                 },
4579                 {
4580                         .procname       = "temp_valid_lft",
4581                         .data           = &ipv6_devconf.temp_valid_lft,
4582                         .maxlen         = sizeof(int),
4583                         .mode           = 0644,
4584                         .proc_handler   = proc_dointvec,
4585                 },
4586                 {
4587                         .procname       = "temp_prefered_lft",
4588                         .data           = &ipv6_devconf.temp_prefered_lft,
4589                         .maxlen         = sizeof(int),
4590                         .mode           = 0644,
4591                         .proc_handler   = proc_dointvec,
4592                 },
4593                 {
4594                         .procname       = "regen_max_retry",
4595                         .data           = &ipv6_devconf.regen_max_retry,
4596                         .maxlen         = sizeof(int),
4597                         .mode           = 0644,
4598                         .proc_handler   = proc_dointvec,
4599                 },
4600                 {
4601                         .procname       = "max_desync_factor",
4602                         .data           = &ipv6_devconf.max_desync_factor,
4603                         .maxlen         = sizeof(int),
4604                         .mode           = 0644,
4605                         .proc_handler   = proc_dointvec,
4606                 },
4607 #endif
4608                 {
4609                         .procname       = "max_addresses",
4610                         .data           = &ipv6_devconf.max_addresses,
4611                         .maxlen         = sizeof(int),
4612                         .mode           = 0644,
4613                         .proc_handler   = proc_dointvec,
4614                 },
4615                 {
4616                         .procname       = "accept_ra_defrtr",
4617                         .data           = &ipv6_devconf.accept_ra_defrtr,
4618                         .maxlen         = sizeof(int),
4619                         .mode           = 0644,
4620                         .proc_handler   = proc_dointvec,
4621                 },
4622                 {
4623                         .procname       = "accept_ra_pinfo",
4624                         .data           = &ipv6_devconf.accept_ra_pinfo,
4625                         .maxlen         = sizeof(int),
4626                         .mode           = 0644,
4627                         .proc_handler   = proc_dointvec,
4628                 },
4629 #ifdef CONFIG_IPV6_ROUTER_PREF
4630                 {
4631                         .procname       = "accept_ra_rtr_pref",
4632                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4633                         .maxlen         = sizeof(int),
4634                         .mode           = 0644,
4635                         .proc_handler   = proc_dointvec,
4636                 },
4637                 {
4638                         .procname       = "router_probe_interval",
4639                         .data           = &ipv6_devconf.rtr_probe_interval,
4640                         .maxlen         = sizeof(int),
4641                         .mode           = 0644,
4642                         .proc_handler   = proc_dointvec_jiffies,
4643                 },
4644 #ifdef CONFIG_IPV6_ROUTE_INFO
4645                 {
4646                         .procname       = "accept_ra_rt_info_max_plen",
4647                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4648                         .maxlen         = sizeof(int),
4649                         .mode           = 0644,
4650                         .proc_handler   = proc_dointvec,
4651                 },
4652 #endif
4653 #endif
4654                 {
4655                         .procname       = "proxy_ndp",
4656                         .data           = &ipv6_devconf.proxy_ndp,
4657                         .maxlen         = sizeof(int),
4658                         .mode           = 0644,
4659                         .proc_handler   = proc_dointvec,
4660                 },
4661                 {
4662                         .procname       = "accept_source_route",
4663                         .data           = &ipv6_devconf.accept_source_route,
4664                         .maxlen         = sizeof(int),
4665                         .mode           = 0644,
4666                         .proc_handler   = proc_dointvec,
4667                 },
4668 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4669                 {
4670                         .procname       = "optimistic_dad",
4671                         .data           = &ipv6_devconf.optimistic_dad,
4672                         .maxlen         = sizeof(int),
4673                         .mode           = 0644,
4674                         .proc_handler   = proc_dointvec,
4675
4676                 },
4677 #endif
4678 #ifdef CONFIG_IPV6_MROUTE
4679                 {
4680                         .procname       = "mc_forwarding",
4681                         .data           = &ipv6_devconf.mc_forwarding,
4682                         .maxlen         = sizeof(int),
4683                         .mode           = 0444,
4684                         .proc_handler   = proc_dointvec,
4685                 },
4686 #endif
4687                 {
4688                         .procname       = "disable_ipv6",
4689                         .data           = &ipv6_devconf.disable_ipv6,
4690                         .maxlen         = sizeof(int),
4691                         .mode           = 0644,
4692                         .proc_handler   = addrconf_sysctl_disable,
4693                 },
4694                 {
4695                         .procname       = "accept_dad",
4696                         .data           = &ipv6_devconf.accept_dad,
4697                         .maxlen         = sizeof(int),
4698                         .mode           = 0644,
4699                         .proc_handler   = proc_dointvec,
4700                 },
4701                 {
4702                         .procname       = "force_tllao",
4703                         .data           = &ipv6_devconf.force_tllao,
4704                         .maxlen         = sizeof(int),
4705                         .mode           = 0644,
4706                         .proc_handler   = proc_dointvec
4707                 },
4708                 {
4709                         /* sentinel */
4710                 }
4711         },
4712 };
4713
4714 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4715                 struct inet6_dev *idev, struct ipv6_devconf *p)
4716 {
4717         int i;
4718         struct addrconf_sysctl_table *t;
4719         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4720
4721         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4722         if (t == NULL)
4723                 goto out;
4724
4725         for (i = 0; t->addrconf_vars[i].data; i++) {
4726                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4727                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4728                 t->addrconf_vars[i].extra2 = net;
4729         }
4730
4731         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4732
4733         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4734         if (t->sysctl_header == NULL)
4735                 goto free;
4736
4737         p->sysctl = t;
4738         return 0;
4739
4740 free:
4741         kfree(t);
4742 out:
4743         return -ENOBUFS;
4744 }
4745
4746 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4747 {
4748         struct addrconf_sysctl_table *t;
4749
4750         if (p->sysctl == NULL)
4751                 return;
4752
4753         t = p->sysctl;
4754         p->sysctl = NULL;
4755         unregister_net_sysctl_table(t->sysctl_header);
4756         kfree(t);
4757 }
4758
4759 static void addrconf_sysctl_register(struct inet6_dev *idev)
4760 {
4761         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4762                               &ndisc_ifinfo_sysctl_change);
4763         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4764                                         idev, &idev->cnf);
4765 }
4766
4767 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4768 {
4769         __addrconf_sysctl_unregister(&idev->cnf);
4770         neigh_sysctl_unregister(idev->nd_parms);
4771 }
4772
4773
4774 #endif
4775
4776 static int __net_init addrconf_init_net(struct net *net)
4777 {
4778         int err;
4779         struct ipv6_devconf *all, *dflt;
4780
4781         err = -ENOMEM;
4782         all = &ipv6_devconf;
4783         dflt = &ipv6_devconf_dflt;
4784
4785         if (!net_eq(net, &init_net)) {
4786                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4787                 if (all == NULL)
4788                         goto err_alloc_all;
4789
4790                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4791                 if (dflt == NULL)
4792                         goto err_alloc_dflt;
4793         } else {
4794                 /* these will be inherited by all namespaces */
4795                 dflt->autoconf = ipv6_defaults.autoconf;
4796                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4797         }
4798
4799         net->ipv6.devconf_all = all;
4800         net->ipv6.devconf_dflt = dflt;
4801
4802 #ifdef CONFIG_SYSCTL
4803         err = __addrconf_sysctl_register(net, "all", NULL, all);
4804         if (err < 0)
4805                 goto err_reg_all;
4806
4807         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4808         if (err < 0)
4809                 goto err_reg_dflt;
4810 #endif
4811         return 0;
4812
4813 #ifdef CONFIG_SYSCTL
4814 err_reg_dflt:
4815         __addrconf_sysctl_unregister(all);
4816 err_reg_all:
4817         kfree(dflt);
4818 #endif
4819 err_alloc_dflt:
4820         kfree(all);
4821 err_alloc_all:
4822         return err;
4823 }
4824
4825 static void __net_exit addrconf_exit_net(struct net *net)
4826 {
4827 #ifdef CONFIG_SYSCTL
4828         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4829         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4830 #endif
4831         if (!net_eq(net, &init_net)) {
4832                 kfree(net->ipv6.devconf_dflt);
4833                 kfree(net->ipv6.devconf_all);
4834         }
4835 }
4836
4837 static struct pernet_operations addrconf_ops = {
4838         .init = addrconf_init_net,
4839         .exit = addrconf_exit_net,
4840 };
4841
4842 /*
4843  *      Device notifier
4844  */
4845
4846 int register_inet6addr_notifier(struct notifier_block *nb)
4847 {
4848         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4849 }
4850 EXPORT_SYMBOL(register_inet6addr_notifier);
4851
4852 int unregister_inet6addr_notifier(struct notifier_block *nb)
4853 {
4854         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4855 }
4856 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4857
4858 static struct rtnl_af_ops inet6_ops = {
4859         .family           = AF_INET6,
4860         .fill_link_af     = inet6_fill_link_af,
4861         .get_link_af_size = inet6_get_link_af_size,
4862 };
4863
4864 /*
4865  *      Init / cleanup code
4866  */
4867
4868 int __init addrconf_init(void)
4869 {
4870         int i, err;
4871
4872         err = ipv6_addr_label_init();
4873         if (err < 0) {
4874                 pr_crit("%s: cannot initialize default policy table: %d\n",
4875                         __func__, err);
4876                 goto out;
4877         }
4878
4879         err = register_pernet_subsys(&addrconf_ops);
4880         if (err < 0)
4881                 goto out_addrlabel;
4882
4883         /* The addrconf netdev notifier requires that loopback_dev
4884          * has it's ipv6 private information allocated and setup
4885          * before it can bring up and give link-local addresses
4886          * to other devices which are up.
4887          *
4888          * Unfortunately, loopback_dev is not necessarily the first
4889          * entry in the global dev_base list of net devices.  In fact,
4890          * it is likely to be the very last entry on that list.
4891          * So this causes the notifier registry below to try and
4892          * give link-local addresses to all devices besides loopback_dev
4893          * first, then loopback_dev, which cases all the non-loopback_dev
4894          * devices to fail to get a link-local address.
4895          *
4896          * So, as a temporary fix, allocate the ipv6 structure for
4897          * loopback_dev first by hand.
4898          * Longer term, all of the dependencies ipv6 has upon the loopback
4899          * device and it being up should be removed.
4900          */
4901         rtnl_lock();
4902         if (!ipv6_add_dev(init_net.loopback_dev))
4903                 err = -ENOMEM;
4904         rtnl_unlock();
4905         if (err)
4906                 goto errlo;
4907
4908         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4909                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4910
4911         register_netdevice_notifier(&ipv6_dev_notf);
4912
4913         addrconf_verify(0);
4914
4915         err = rtnl_af_register(&inet6_ops);
4916         if (err < 0)
4917                 goto errout_af;
4918
4919         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4920                               NULL);
4921         if (err < 0)
4922                 goto errout;
4923
4924         /* Only the first call to __rtnl_register can fail */
4925         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4926         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4927         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4928                         inet6_dump_ifaddr, NULL);
4929         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4930                         inet6_dump_ifmcaddr, NULL);
4931         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4932                         inet6_dump_ifacaddr, NULL);
4933         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4934                         NULL, NULL);
4935
4936         ipv6_addr_label_rtnl_register();
4937
4938         return 0;
4939 errout:
4940         rtnl_af_unregister(&inet6_ops);
4941 errout_af:
4942         unregister_netdevice_notifier(&ipv6_dev_notf);
4943 errlo:
4944         unregister_pernet_subsys(&addrconf_ops);
4945 out_addrlabel:
4946         ipv6_addr_label_cleanup();
4947 out:
4948         return err;
4949 }
4950
4951 void addrconf_cleanup(void)
4952 {
4953         struct net_device *dev;
4954         int i;
4955
4956         unregister_netdevice_notifier(&ipv6_dev_notf);
4957         unregister_pernet_subsys(&addrconf_ops);
4958         ipv6_addr_label_cleanup();
4959
4960         rtnl_lock();
4961
4962         __rtnl_af_unregister(&inet6_ops);
4963
4964         /* clean dev list */
4965         for_each_netdev(&init_net, dev) {
4966                 if (__in6_dev_get(dev) == NULL)
4967                         continue;
4968                 addrconf_ifdown(dev, 1);
4969         }
4970         addrconf_ifdown(init_net.loopback_dev, 2);
4971
4972         /*
4973          *      Check hash table.
4974          */
4975         spin_lock_bh(&addrconf_hash_lock);
4976         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4977                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4978         spin_unlock_bh(&addrconf_hash_lock);
4979
4980         del_timer(&addr_chk_timer);
4981         rtnl_unlock();
4982 }