Merge branch 'for-linux-next' of git://people.freedesktop.org/~danvet/drm-intel into...
[pandora-kernel.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
88 #endif
89
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
99
100 #if ACONF_DEBUG >= 3
101 #define ADBG(x) printk x
102 #else
103 #define ADBG(x)
104 #endif
105
106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
107
108 static inline u32 cstamp_delta(unsigned long cstamp)
109 {
110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 }
112
113 #ifdef CONFIG_SYSCTL
114 static void addrconf_sysctl_register(struct inet6_dev *idev);
115 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
116 #else
117 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
118 {
119 }
120
121 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
122 {
123 }
124 #endif
125
126 #ifdef CONFIG_IPV6_PRIVACY
127 static void __ipv6_regen_rndid(struct inet6_dev *idev);
128 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
129 static void ipv6_regen_rndid(unsigned long data);
130 #endif
131
132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
133 static int ipv6_count_addresses(struct inet6_dev *idev);
134
135 /*
136  *      Configured unicast address hash table
137  */
138 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
139 static DEFINE_SPINLOCK(addrconf_hash_lock);
140
141 static void addrconf_verify(unsigned long);
142
143 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
144 static DEFINE_SPINLOCK(addrconf_verify_lock);
145
146 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
148
149 static void addrconf_type_change(struct net_device *dev,
150                                  unsigned long event);
151 static int addrconf_ifdown(struct net_device *dev, int how);
152
153 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
154                                                   int plen,
155                                                   const struct net_device *dev,
156                                                   u32 flags, u32 noflags);
157
158 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_timer(unsigned long data);
160 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
161 static void addrconf_dad_run(struct inet6_dev *idev);
162 static void addrconf_rs_timer(unsigned long data);
163 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
165
166 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
167                                 struct prefix_info *pinfo);
168 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
169                                struct net_device *dev);
170
171 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
172
173 static struct ipv6_devconf ipv6_devconf __read_mostly = {
174         .forwarding             = 0,
175         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
176         .mtu6                   = IPV6_MIN_MTU,
177         .accept_ra              = 1,
178         .accept_redirects       = 1,
179         .autoconf               = 1,
180         .force_mld_version      = 0,
181         .dad_transmits          = 1,
182         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
183         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
184         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
185 #ifdef CONFIG_IPV6_PRIVACY
186         .use_tempaddr           = 0,
187         .temp_valid_lft         = TEMP_VALID_LIFETIME,
188         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
189         .regen_max_retry        = REGEN_MAX_RETRY,
190         .max_desync_factor      = MAX_DESYNC_FACTOR,
191 #endif
192         .max_addresses          = IPV6_MAX_ADDRESSES,
193         .accept_ra_defrtr       = 1,
194         .accept_ra_pinfo        = 1,
195 #ifdef CONFIG_IPV6_ROUTER_PREF
196         .accept_ra_rtr_pref     = 1,
197         .rtr_probe_interval     = 60 * HZ,
198 #ifdef CONFIG_IPV6_ROUTE_INFO
199         .accept_ra_rt_info_max_plen = 0,
200 #endif
201 #endif
202         .proxy_ndp              = 0,
203         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
204         .disable_ipv6           = 0,
205         .accept_dad             = 1,
206 };
207
208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
209         .forwarding             = 0,
210         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
211         .mtu6                   = IPV6_MIN_MTU,
212         .accept_ra              = 1,
213         .accept_redirects       = 1,
214         .autoconf               = 1,
215         .dad_transmits          = 1,
216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
219 #ifdef CONFIG_IPV6_PRIVACY
220         .use_tempaddr           = 0,
221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
223         .regen_max_retry        = REGEN_MAX_RETRY,
224         .max_desync_factor      = MAX_DESYNC_FACTOR,
225 #endif
226         .max_addresses          = IPV6_MAX_ADDRESSES,
227         .accept_ra_defrtr       = 1,
228         .accept_ra_pinfo        = 1,
229 #ifdef CONFIG_IPV6_ROUTER_PREF
230         .accept_ra_rtr_pref     = 1,
231         .rtr_probe_interval     = 60 * HZ,
232 #ifdef CONFIG_IPV6_ROUTE_INFO
233         .accept_ra_rt_info_max_plen = 0,
234 #endif
235 #endif
236         .proxy_ndp              = 0,
237         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
238         .disable_ipv6           = 0,
239         .accept_dad             = 1,
240 };
241
242 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
243 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
244 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
245 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
246 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
247 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
248 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
249 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
250
251 /* Check if a valid qdisc is available */
252 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
253 {
254         return !qdisc_tx_is_noop(dev);
255 }
256
257 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
258 {
259         if (del_timer(&ifp->timer))
260                 __in6_ifa_put(ifp);
261 }
262
263 enum addrconf_timer_t {
264         AC_NONE,
265         AC_DAD,
266         AC_RS,
267 };
268
269 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
270                                enum addrconf_timer_t what,
271                                unsigned long when)
272 {
273         if (!del_timer(&ifp->timer))
274                 in6_ifa_hold(ifp);
275
276         switch (what) {
277         case AC_DAD:
278                 ifp->timer.function = addrconf_dad_timer;
279                 break;
280         case AC_RS:
281                 ifp->timer.function = addrconf_rs_timer;
282                 break;
283         default:
284                 break;
285         }
286         ifp->timer.expires = jiffies + when;
287         add_timer(&ifp->timer);
288 }
289
290 static int snmp6_alloc_dev(struct inet6_dev *idev)
291 {
292         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
293                           sizeof(struct ipstats_mib),
294                           __alignof__(struct ipstats_mib)) < 0)
295                 goto err_ip;
296         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
297                                         GFP_KERNEL);
298         if (!idev->stats.icmpv6dev)
299                 goto err_icmp;
300         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
301                                            GFP_KERNEL);
302         if (!idev->stats.icmpv6msgdev)
303                 goto err_icmpmsg;
304
305         return 0;
306
307 err_icmpmsg:
308         kfree(idev->stats.icmpv6dev);
309 err_icmp:
310         snmp_mib_free((void __percpu **)idev->stats.ipv6);
311 err_ip:
312         return -ENOMEM;
313 }
314
315 static void snmp6_free_dev(struct inet6_dev *idev)
316 {
317         kfree(idev->stats.icmpv6msgdev);
318         kfree(idev->stats.icmpv6dev);
319         snmp_mib_free((void __percpu **)idev->stats.ipv6);
320 }
321
322 /* Nobody refers to this device, we may destroy it. */
323
324 void in6_dev_finish_destroy(struct inet6_dev *idev)
325 {
326         struct net_device *dev = idev->dev;
327
328         WARN_ON(!list_empty(&idev->addr_list));
329         WARN_ON(idev->mc_list != NULL);
330
331 #ifdef NET_REFCNT_DEBUG
332         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
333 #endif
334         dev_put(dev);
335         if (!idev->dead) {
336                 pr_warn("Freeing alive inet6 device %p\n", idev);
337                 return;
338         }
339         snmp6_free_dev(idev);
340         kfree_rcu(idev, rcu);
341 }
342 EXPORT_SYMBOL(in6_dev_finish_destroy);
343
344 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
345 {
346         struct inet6_dev *ndev;
347
348         ASSERT_RTNL();
349
350         if (dev->mtu < IPV6_MIN_MTU)
351                 return NULL;
352
353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
354
355         if (ndev == NULL)
356                 return NULL;
357
358         rwlock_init(&ndev->lock);
359         ndev->dev = dev;
360         INIT_LIST_HEAD(&ndev->addr_list);
361
362         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
363         ndev->cnf.mtu6 = dev->mtu;
364         ndev->cnf.sysctl = NULL;
365         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
366         if (ndev->nd_parms == NULL) {
367                 kfree(ndev);
368                 return NULL;
369         }
370         if (ndev->cnf.forwarding)
371                 dev_disable_lro(dev);
372         /* We refer to the device */
373         dev_hold(dev);
374
375         if (snmp6_alloc_dev(ndev) < 0) {
376                 ADBG((KERN_WARNING
377                         "%s: cannot allocate memory for statistics; dev=%s.\n",
378                         __func__, dev->name));
379                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
380                 dev_put(dev);
381                 kfree(ndev);
382                 return NULL;
383         }
384
385         if (snmp6_register_dev(ndev) < 0) {
386                 ADBG((KERN_WARNING
387                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
388                         __func__, dev->name));
389                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
390                 ndev->dead = 1;
391                 in6_dev_finish_destroy(ndev);
392                 return NULL;
393         }
394
395         /* One reference from device.  We must do this before
396          * we invoke __ipv6_regen_rndid().
397          */
398         in6_dev_hold(ndev);
399
400         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
401                 ndev->cnf.accept_dad = -1;
402
403 #if IS_ENABLED(CONFIG_IPV6_SIT)
404         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
405                 pr_info("%s: Disabled Multicast RS\n", dev->name);
406                 ndev->cnf.rtr_solicits = 0;
407         }
408 #endif
409
410 #ifdef CONFIG_IPV6_PRIVACY
411         INIT_LIST_HEAD(&ndev->tempaddr_list);
412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413         if ((dev->flags&IFF_LOOPBACK) ||
414             dev->type == ARPHRD_TUNNEL ||
415             dev->type == ARPHRD_TUNNEL6 ||
416             dev->type == ARPHRD_SIT ||
417             dev->type == ARPHRD_NONE) {
418                 ndev->cnf.use_tempaddr = -1;
419         } else {
420                 in6_dev_hold(ndev);
421                 ipv6_regen_rndid((unsigned long) ndev);
422         }
423 #endif
424
425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
426                 ndev->if_flags |= IF_READY;
427
428         ipv6_mc_init_dev(ndev);
429         ndev->tstamp = jiffies;
430         addrconf_sysctl_register(ndev);
431         /* protected by rtnl_lock */
432         rcu_assign_pointer(dev->ip6_ptr, ndev);
433
434         /* Join interface-local all-node multicast group */
435         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
436
437         /* Join all-node multicast group */
438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439
440         /* Join all-router multicast group if forwarding is set */
441         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
442                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
443
444         return ndev;
445 }
446
447 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
448 {
449         struct inet6_dev *idev;
450
451         ASSERT_RTNL();
452
453         idev = __in6_dev_get(dev);
454         if (!idev) {
455                 idev = ipv6_add_dev(dev);
456                 if (!idev)
457                         return NULL;
458         }
459
460         if (dev->flags&IFF_UP)
461                 ipv6_mc_up(idev);
462         return idev;
463 }
464
465 static int inet6_netconf_msgsize_devconf(int type)
466 {
467         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
468                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
469
470         /* type -1 is used for ALL */
471         if (type == -1 || type == NETCONFA_FORWARDING)
472                 size += nla_total_size(4);
473 #ifdef CONFIG_IPV6_MROUTE
474         if (type == -1 || type == NETCONFA_MC_FORWARDING)
475                 size += nla_total_size(4);
476 #endif
477
478         return size;
479 }
480
481 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
482                                       struct ipv6_devconf *devconf, u32 portid,
483                                       u32 seq, int event, unsigned int flags,
484                                       int type)
485 {
486         struct nlmsghdr  *nlh;
487         struct netconfmsg *ncm;
488
489         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
490                         flags);
491         if (nlh == NULL)
492                 return -EMSGSIZE;
493
494         ncm = nlmsg_data(nlh);
495         ncm->ncm_family = AF_INET6;
496
497         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
498                 goto nla_put_failure;
499
500         /* type -1 is used for ALL */
501         if ((type == -1 || type == NETCONFA_FORWARDING) &&
502             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
503                 goto nla_put_failure;
504 #ifdef CONFIG_IPV6_MROUTE
505         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
506             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
507                         devconf->mc_forwarding) < 0)
508                 goto nla_put_failure;
509 #endif
510         return nlmsg_end(skb, nlh);
511
512 nla_put_failure:
513         nlmsg_cancel(skb, nlh);
514         return -EMSGSIZE;
515 }
516
517 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
518                                   struct ipv6_devconf *devconf)
519 {
520         struct sk_buff *skb;
521         int err = -ENOBUFS;
522
523         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
524         if (skb == NULL)
525                 goto errout;
526
527         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
528                                          RTM_NEWNETCONF, 0, type);
529         if (err < 0) {
530                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
531                 WARN_ON(err == -EMSGSIZE);
532                 kfree_skb(skb);
533                 goto errout;
534         }
535         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
536         return;
537 errout:
538         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
539 }
540
541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
542         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
543         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
544 };
545
546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
547                                      struct nlmsghdr *nlh,
548                                      void *arg)
549 {
550         struct net *net = sock_net(in_skb->sk);
551         struct nlattr *tb[NETCONFA_MAX+1];
552         struct netconfmsg *ncm;
553         struct sk_buff *skb;
554         struct ipv6_devconf *devconf;
555         struct inet6_dev *in6_dev;
556         struct net_device *dev;
557         int ifindex;
558         int err;
559
560         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
561                           devconf_ipv6_policy);
562         if (err < 0)
563                 goto errout;
564
565         err = EINVAL;
566         if (!tb[NETCONFA_IFINDEX])
567                 goto errout;
568
569         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
570         switch (ifindex) {
571         case NETCONFA_IFINDEX_ALL:
572                 devconf = net->ipv6.devconf_all;
573                 break;
574         case NETCONFA_IFINDEX_DEFAULT:
575                 devconf = net->ipv6.devconf_dflt;
576                 break;
577         default:
578                 dev = __dev_get_by_index(net, ifindex);
579                 if (dev == NULL)
580                         goto errout;
581                 in6_dev = __in6_dev_get(dev);
582                 if (in6_dev == NULL)
583                         goto errout;
584                 devconf = &in6_dev->cnf;
585                 break;
586         }
587
588         err = -ENOBUFS;
589         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
590         if (skb == NULL)
591                 goto errout;
592
593         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
594                                          NETLINK_CB(in_skb).portid,
595                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
596                                          -1);
597         if (err < 0) {
598                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
599                 WARN_ON(err == -EMSGSIZE);
600                 kfree_skb(skb);
601                 goto errout;
602         }
603         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
604 errout:
605         return err;
606 }
607
608 #ifdef CONFIG_SYSCTL
609 static void dev_forward_change(struct inet6_dev *idev)
610 {
611         struct net_device *dev;
612         struct inet6_ifaddr *ifa;
613
614         if (!idev)
615                 return;
616         dev = idev->dev;
617         if (idev->cnf.forwarding)
618                 dev_disable_lro(dev);
619         if (dev->flags & IFF_MULTICAST) {
620                 if (idev->cnf.forwarding) {
621                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
622                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
623                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
624                 } else {
625                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
626                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
627                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
628                 }
629         }
630
631         list_for_each_entry(ifa, &idev->addr_list, if_list) {
632                 if (ifa->flags&IFA_F_TENTATIVE)
633                         continue;
634                 if (idev->cnf.forwarding)
635                         addrconf_join_anycast(ifa);
636                 else
637                         addrconf_leave_anycast(ifa);
638         }
639         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
640                                      dev->ifindex, &idev->cnf);
641 }
642
643
644 static void addrconf_forward_change(struct net *net, __s32 newf)
645 {
646         struct net_device *dev;
647         struct inet6_dev *idev;
648
649         for_each_netdev(net, dev) {
650                 idev = __in6_dev_get(dev);
651                 if (idev) {
652                         int changed = (!idev->cnf.forwarding) ^ (!newf);
653                         idev->cnf.forwarding = newf;
654                         if (changed)
655                                 dev_forward_change(idev);
656                 }
657         }
658 }
659
660 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
661 {
662         struct net *net;
663         int old;
664
665         if (!rtnl_trylock())
666                 return restart_syscall();
667
668         net = (struct net *)table->extra2;
669         old = *p;
670         *p = newf;
671
672         if (p == &net->ipv6.devconf_dflt->forwarding) {
673                 if ((!newf) ^ (!old))
674                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
675                                                      NETCONFA_IFINDEX_DEFAULT,
676                                                      net->ipv6.devconf_dflt);
677                 rtnl_unlock();
678                 return 0;
679         }
680
681         if (p == &net->ipv6.devconf_all->forwarding) {
682                 net->ipv6.devconf_dflt->forwarding = newf;
683                 addrconf_forward_change(net, newf);
684                 if ((!newf) ^ (!old))
685                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
686                                                      NETCONFA_IFINDEX_ALL,
687                                                      net->ipv6.devconf_all);
688         } else if ((!newf) ^ (!old))
689                 dev_forward_change((struct inet6_dev *)table->extra1);
690         rtnl_unlock();
691
692         if (newf)
693                 rt6_purge_dflt_routers(net);
694         return 1;
695 }
696 #endif
697
698 /* Nobody refers to this ifaddr, destroy it */
699 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
700 {
701         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
702
703 #ifdef NET_REFCNT_DEBUG
704         pr_debug("%s\n", __func__);
705 #endif
706
707         in6_dev_put(ifp->idev);
708
709         if (del_timer(&ifp->timer))
710                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
711
712         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
713                 pr_warn("Freeing alive inet6 address %p\n", ifp);
714                 return;
715         }
716         ip6_rt_put(ifp->rt);
717
718         kfree_rcu(ifp, rcu);
719 }
720
721 static void
722 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
723 {
724         struct list_head *p;
725         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
726
727         /*
728          * Each device address list is sorted in order of scope -
729          * global before linklocal.
730          */
731         list_for_each(p, &idev->addr_list) {
732                 struct inet6_ifaddr *ifa
733                         = list_entry(p, struct inet6_ifaddr, if_list);
734                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
735                         break;
736         }
737
738         list_add_tail(&ifp->if_list, p);
739 }
740
741 static u32 inet6_addr_hash(const struct in6_addr *addr)
742 {
743         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
744 }
745
746 /* On success it returns ifp with increased reference count */
747
748 static struct inet6_ifaddr *
749 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
750               int scope, u32 flags)
751 {
752         struct inet6_ifaddr *ifa = NULL;
753         struct rt6_info *rt;
754         unsigned int hash;
755         int err = 0;
756         int addr_type = ipv6_addr_type(addr);
757
758         if (addr_type == IPV6_ADDR_ANY ||
759             addr_type & IPV6_ADDR_MULTICAST ||
760             (!(idev->dev->flags & IFF_LOOPBACK) &&
761              addr_type & IPV6_ADDR_LOOPBACK))
762                 return ERR_PTR(-EADDRNOTAVAIL);
763
764         rcu_read_lock_bh();
765         if (idev->dead) {
766                 err = -ENODEV;                  /*XXX*/
767                 goto out2;
768         }
769
770         if (idev->cnf.disable_ipv6) {
771                 err = -EACCES;
772                 goto out2;
773         }
774
775         spin_lock(&addrconf_hash_lock);
776
777         /* Ignore adding duplicate addresses on an interface */
778         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
779                 ADBG(("ipv6_add_addr: already assigned\n"));
780                 err = -EEXIST;
781                 goto out;
782         }
783
784         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
785
786         if (ifa == NULL) {
787                 ADBG(("ipv6_add_addr: malloc failed\n"));
788                 err = -ENOBUFS;
789                 goto out;
790         }
791
792         rt = addrconf_dst_alloc(idev, addr, false);
793         if (IS_ERR(rt)) {
794                 err = PTR_ERR(rt);
795                 goto out;
796         }
797
798         ifa->addr = *addr;
799
800         spin_lock_init(&ifa->lock);
801         spin_lock_init(&ifa->state_lock);
802         init_timer(&ifa->timer);
803         INIT_HLIST_NODE(&ifa->addr_lst);
804         ifa->timer.data = (unsigned long) ifa;
805         ifa->scope = scope;
806         ifa->prefix_len = pfxlen;
807         ifa->flags = flags | IFA_F_TENTATIVE;
808         ifa->cstamp = ifa->tstamp = jiffies;
809
810         ifa->rt = rt;
811
812         ifa->idev = idev;
813         in6_dev_hold(idev);
814         /* For caller */
815         in6_ifa_hold(ifa);
816
817         /* Add to big hash table */
818         hash = inet6_addr_hash(addr);
819
820         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
821         spin_unlock(&addrconf_hash_lock);
822
823         write_lock(&idev->lock);
824         /* Add to inet6_dev unicast addr list. */
825         ipv6_link_dev_addr(idev, ifa);
826
827 #ifdef CONFIG_IPV6_PRIVACY
828         if (ifa->flags&IFA_F_TEMPORARY) {
829                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
830                 in6_ifa_hold(ifa);
831         }
832 #endif
833
834         in6_ifa_hold(ifa);
835         write_unlock(&idev->lock);
836 out2:
837         rcu_read_unlock_bh();
838
839         if (likely(err == 0))
840                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
841         else {
842                 kfree(ifa);
843                 ifa = ERR_PTR(err);
844         }
845
846         return ifa;
847 out:
848         spin_unlock(&addrconf_hash_lock);
849         goto out2;
850 }
851
852 /* This function wants to get referenced ifp and releases it before return */
853
854 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
855 {
856         struct inet6_ifaddr *ifa, *ifn;
857         struct inet6_dev *idev = ifp->idev;
858         int state;
859         int deleted = 0, onlink = 0;
860         unsigned long expires = jiffies;
861
862         spin_lock_bh(&ifp->state_lock);
863         state = ifp->state;
864         ifp->state = INET6_IFADDR_STATE_DEAD;
865         spin_unlock_bh(&ifp->state_lock);
866
867         if (state == INET6_IFADDR_STATE_DEAD)
868                 goto out;
869
870         spin_lock_bh(&addrconf_hash_lock);
871         hlist_del_init_rcu(&ifp->addr_lst);
872         spin_unlock_bh(&addrconf_hash_lock);
873
874         write_lock_bh(&idev->lock);
875 #ifdef CONFIG_IPV6_PRIVACY
876         if (ifp->flags&IFA_F_TEMPORARY) {
877                 list_del(&ifp->tmp_list);
878                 if (ifp->ifpub) {
879                         in6_ifa_put(ifp->ifpub);
880                         ifp->ifpub = NULL;
881                 }
882                 __in6_ifa_put(ifp);
883         }
884 #endif
885
886         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
887                 if (ifa == ifp) {
888                         list_del_init(&ifp->if_list);
889                         __in6_ifa_put(ifp);
890
891                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
892                                 break;
893                         deleted = 1;
894                         continue;
895                 } else if (ifp->flags & IFA_F_PERMANENT) {
896                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
897                                               ifp->prefix_len)) {
898                                 if (ifa->flags & IFA_F_PERMANENT) {
899                                         onlink = 1;
900                                         if (deleted)
901                                                 break;
902                                 } else {
903                                         unsigned long lifetime;
904
905                                         if (!onlink)
906                                                 onlink = -1;
907
908                                         spin_lock(&ifa->lock);
909
910                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
911                                         /*
912                                          * Note: Because this address is
913                                          * not permanent, lifetime <
914                                          * LONG_MAX / HZ here.
915                                          */
916                                         if (time_before(expires,
917                                                         ifa->tstamp + lifetime * HZ))
918                                                 expires = ifa->tstamp + lifetime * HZ;
919                                         spin_unlock(&ifa->lock);
920                                 }
921                         }
922                 }
923         }
924         write_unlock_bh(&idev->lock);
925
926         addrconf_del_timer(ifp);
927
928         ipv6_ifa_notify(RTM_DELADDR, ifp);
929
930         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
931
932         /*
933          * Purge or update corresponding prefix
934          *
935          * 1) we don't purge prefix here if address was not permanent.
936          *    prefix is managed by its own lifetime.
937          * 2) if there're no addresses, delete prefix.
938          * 3) if there're still other permanent address(es),
939          *    corresponding prefix is still permanent.
940          * 4) otherwise, update prefix lifetime to the
941          *    longest valid lifetime among the corresponding
942          *    addresses on the device.
943          *    Note: subsequent RA will update lifetime.
944          *
945          * --yoshfuji
946          */
947         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
948                 struct in6_addr prefix;
949                 struct rt6_info *rt;
950
951                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
952
953                 rt = addrconf_get_prefix_route(&prefix,
954                                                ifp->prefix_len,
955                                                ifp->idev->dev,
956                                                0, RTF_GATEWAY | RTF_DEFAULT);
957
958                 if (rt) {
959                         if (onlink == 0) {
960                                 ip6_del_rt(rt);
961                                 rt = NULL;
962                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
963                                 rt6_set_expires(rt, expires);
964                         }
965                 }
966                 ip6_rt_put(rt);
967         }
968
969         /* clean up prefsrc entries */
970         rt6_remove_prefsrc(ifp);
971 out:
972         in6_ifa_put(ifp);
973 }
974
975 #ifdef CONFIG_IPV6_PRIVACY
976 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
977 {
978         struct inet6_dev *idev = ifp->idev;
979         struct in6_addr addr, *tmpaddr;
980         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
981         unsigned long regen_advance;
982         int tmp_plen;
983         int ret = 0;
984         int max_addresses;
985         u32 addr_flags;
986         unsigned long now = jiffies;
987
988         write_lock(&idev->lock);
989         if (ift) {
990                 spin_lock_bh(&ift->lock);
991                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
992                 spin_unlock_bh(&ift->lock);
993                 tmpaddr = &addr;
994         } else {
995                 tmpaddr = NULL;
996         }
997 retry:
998         in6_dev_hold(idev);
999         if (idev->cnf.use_tempaddr <= 0) {
1000                 write_unlock(&idev->lock);
1001                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1002                 in6_dev_put(idev);
1003                 ret = -1;
1004                 goto out;
1005         }
1006         spin_lock_bh(&ifp->lock);
1007         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1008                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1009                 spin_unlock_bh(&ifp->lock);
1010                 write_unlock(&idev->lock);
1011                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1012                         __func__);
1013                 in6_dev_put(idev);
1014                 ret = -1;
1015                 goto out;
1016         }
1017         in6_ifa_hold(ifp);
1018         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1019         __ipv6_try_regen_rndid(idev, tmpaddr);
1020         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1021         age = (now - ifp->tstamp) / HZ;
1022         tmp_valid_lft = min_t(__u32,
1023                               ifp->valid_lft,
1024                               idev->cnf.temp_valid_lft + age);
1025         tmp_prefered_lft = min_t(__u32,
1026                                  ifp->prefered_lft,
1027                                  idev->cnf.temp_prefered_lft + age -
1028                                  idev->cnf.max_desync_factor);
1029         tmp_plen = ifp->prefix_len;
1030         max_addresses = idev->cnf.max_addresses;
1031         tmp_tstamp = ifp->tstamp;
1032         spin_unlock_bh(&ifp->lock);
1033
1034         regen_advance = idev->cnf.regen_max_retry *
1035                         idev->cnf.dad_transmits *
1036                         idev->nd_parms->retrans_time / HZ;
1037         write_unlock(&idev->lock);
1038
1039         /* A temporary address is created only if this calculated Preferred
1040          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1041          * an implementation must not create a temporary address with a zero
1042          * Preferred Lifetime.
1043          */
1044         if (tmp_prefered_lft <= regen_advance) {
1045                 in6_ifa_put(ifp);
1046                 in6_dev_put(idev);
1047                 ret = -1;
1048                 goto out;
1049         }
1050
1051         addr_flags = IFA_F_TEMPORARY;
1052         /* set in addrconf_prefix_rcv() */
1053         if (ifp->flags & IFA_F_OPTIMISTIC)
1054                 addr_flags |= IFA_F_OPTIMISTIC;
1055
1056         ift = !max_addresses ||
1057               ipv6_count_addresses(idev) < max_addresses ?
1058                 ipv6_add_addr(idev, &addr, tmp_plen,
1059                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1060                               addr_flags) : NULL;
1061         if (IS_ERR_OR_NULL(ift)) {
1062                 in6_ifa_put(ifp);
1063                 in6_dev_put(idev);
1064                 pr_info("%s: retry temporary address regeneration\n", __func__);
1065                 tmpaddr = &addr;
1066                 write_lock(&idev->lock);
1067                 goto retry;
1068         }
1069
1070         spin_lock_bh(&ift->lock);
1071         ift->ifpub = ifp;
1072         ift->valid_lft = tmp_valid_lft;
1073         ift->prefered_lft = tmp_prefered_lft;
1074         ift->cstamp = now;
1075         ift->tstamp = tmp_tstamp;
1076         spin_unlock_bh(&ift->lock);
1077
1078         addrconf_dad_start(ift);
1079         in6_ifa_put(ift);
1080         in6_dev_put(idev);
1081 out:
1082         return ret;
1083 }
1084 #endif
1085
1086 /*
1087  *      Choose an appropriate source address (RFC3484)
1088  */
1089 enum {
1090         IPV6_SADDR_RULE_INIT = 0,
1091         IPV6_SADDR_RULE_LOCAL,
1092         IPV6_SADDR_RULE_SCOPE,
1093         IPV6_SADDR_RULE_PREFERRED,
1094 #ifdef CONFIG_IPV6_MIP6
1095         IPV6_SADDR_RULE_HOA,
1096 #endif
1097         IPV6_SADDR_RULE_OIF,
1098         IPV6_SADDR_RULE_LABEL,
1099 #ifdef CONFIG_IPV6_PRIVACY
1100         IPV6_SADDR_RULE_PRIVACY,
1101 #endif
1102         IPV6_SADDR_RULE_ORCHID,
1103         IPV6_SADDR_RULE_PREFIX,
1104         IPV6_SADDR_RULE_MAX
1105 };
1106
1107 struct ipv6_saddr_score {
1108         int                     rule;
1109         int                     addr_type;
1110         struct inet6_ifaddr     *ifa;
1111         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1112         int                     scopedist;
1113         int                     matchlen;
1114 };
1115
1116 struct ipv6_saddr_dst {
1117         const struct in6_addr *addr;
1118         int ifindex;
1119         int scope;
1120         int label;
1121         unsigned int prefs;
1122 };
1123
1124 static inline int ipv6_saddr_preferred(int type)
1125 {
1126         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1127                 return 1;
1128         return 0;
1129 }
1130
1131 static int ipv6_get_saddr_eval(struct net *net,
1132                                struct ipv6_saddr_score *score,
1133                                struct ipv6_saddr_dst *dst,
1134                                int i)
1135 {
1136         int ret;
1137
1138         if (i <= score->rule) {
1139                 switch (i) {
1140                 case IPV6_SADDR_RULE_SCOPE:
1141                         ret = score->scopedist;
1142                         break;
1143                 case IPV6_SADDR_RULE_PREFIX:
1144                         ret = score->matchlen;
1145                         break;
1146                 default:
1147                         ret = !!test_bit(i, score->scorebits);
1148                 }
1149                 goto out;
1150         }
1151
1152         switch (i) {
1153         case IPV6_SADDR_RULE_INIT:
1154                 /* Rule 0: remember if hiscore is not ready yet */
1155                 ret = !!score->ifa;
1156                 break;
1157         case IPV6_SADDR_RULE_LOCAL:
1158                 /* Rule 1: Prefer same address */
1159                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1160                 break;
1161         case IPV6_SADDR_RULE_SCOPE:
1162                 /* Rule 2: Prefer appropriate scope
1163                  *
1164                  *      ret
1165                  *       ^
1166                  *    -1 |  d 15
1167                  *    ---+--+-+---> scope
1168                  *       |
1169                  *       |             d is scope of the destination.
1170                  *  B-d  |  \
1171                  *       |   \      <- smaller scope is better if
1172                  *  B-15 |    \        if scope is enough for destinaion.
1173                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1174                  * d-C-1 | /
1175                  *       |/         <- greater is better
1176                  *   -C  /             if scope is not enough for destination.
1177                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1178                  *
1179                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1180                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1181                  * Assume B = 0 and we get C > 29.
1182                  */
1183                 ret = __ipv6_addr_src_scope(score->addr_type);
1184                 if (ret >= dst->scope)
1185                         ret = -ret;
1186                 else
1187                         ret -= 128;     /* 30 is enough */
1188                 score->scopedist = ret;
1189                 break;
1190         case IPV6_SADDR_RULE_PREFERRED:
1191                 /* Rule 3: Avoid deprecated and optimistic addresses */
1192                 ret = ipv6_saddr_preferred(score->addr_type) ||
1193                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1194                 break;
1195 #ifdef CONFIG_IPV6_MIP6
1196         case IPV6_SADDR_RULE_HOA:
1197             {
1198                 /* Rule 4: Prefer home address */
1199                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1200                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1201                 break;
1202             }
1203 #endif
1204         case IPV6_SADDR_RULE_OIF:
1205                 /* Rule 5: Prefer outgoing interface */
1206                 ret = (!dst->ifindex ||
1207                        dst->ifindex == score->ifa->idev->dev->ifindex);
1208                 break;
1209         case IPV6_SADDR_RULE_LABEL:
1210                 /* Rule 6: Prefer matching label */
1211                 ret = ipv6_addr_label(net,
1212                                       &score->ifa->addr, score->addr_type,
1213                                       score->ifa->idev->dev->ifindex) == dst->label;
1214                 break;
1215 #ifdef CONFIG_IPV6_PRIVACY
1216         case IPV6_SADDR_RULE_PRIVACY:
1217             {
1218                 /* Rule 7: Prefer public address
1219                  * Note: prefer temporary address if use_tempaddr >= 2
1220                  */
1221                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1222                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1223                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1224                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1225                 break;
1226             }
1227 #endif
1228         case IPV6_SADDR_RULE_ORCHID:
1229                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1230                  *          non-ORCHID vs non-ORCHID
1231                  */
1232                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1233                         ipv6_addr_orchid(dst->addr));
1234                 break;
1235         case IPV6_SADDR_RULE_PREFIX:
1236                 /* Rule 8: Use longest matching prefix */
1237                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1238                 if (ret > score->ifa->prefix_len)
1239                         ret = score->ifa->prefix_len;
1240                 score->matchlen = ret;
1241                 break;
1242         default:
1243                 ret = 0;
1244         }
1245
1246         if (ret)
1247                 __set_bit(i, score->scorebits);
1248         score->rule = i;
1249 out:
1250         return ret;
1251 }
1252
1253 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1254                        const struct in6_addr *daddr, unsigned int prefs,
1255                        struct in6_addr *saddr)
1256 {
1257         struct ipv6_saddr_score scores[2],
1258                                 *score = &scores[0], *hiscore = &scores[1];
1259         struct ipv6_saddr_dst dst;
1260         struct net_device *dev;
1261         int dst_type;
1262
1263         dst_type = __ipv6_addr_type(daddr);
1264         dst.addr = daddr;
1265         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1266         dst.scope = __ipv6_addr_src_scope(dst_type);
1267         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1268         dst.prefs = prefs;
1269
1270         hiscore->rule = -1;
1271         hiscore->ifa = NULL;
1272
1273         rcu_read_lock();
1274
1275         for_each_netdev_rcu(net, dev) {
1276                 struct inet6_dev *idev;
1277
1278                 /* Candidate Source Address (section 4)
1279                  *  - multicast and link-local destination address,
1280                  *    the set of candidate source address MUST only
1281                  *    include addresses assigned to interfaces
1282                  *    belonging to the same link as the outgoing
1283                  *    interface.
1284                  * (- For site-local destination addresses, the
1285                  *    set of candidate source addresses MUST only
1286                  *    include addresses assigned to interfaces
1287                  *    belonging to the same site as the outgoing
1288                  *    interface.)
1289                  */
1290                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1291                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1292                     dst.ifindex && dev->ifindex != dst.ifindex)
1293                         continue;
1294
1295                 idev = __in6_dev_get(dev);
1296                 if (!idev)
1297                         continue;
1298
1299                 read_lock_bh(&idev->lock);
1300                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1301                         int i;
1302
1303                         /*
1304                          * - Tentative Address (RFC2462 section 5.4)
1305                          *  - A tentative address is not considered
1306                          *    "assigned to an interface" in the traditional
1307                          *    sense, unless it is also flagged as optimistic.
1308                          * - Candidate Source Address (section 4)
1309                          *  - In any case, anycast addresses, multicast
1310                          *    addresses, and the unspecified address MUST
1311                          *    NOT be included in a candidate set.
1312                          */
1313                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1314                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1315                                 continue;
1316
1317                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1318
1319                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1320                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1321                                 LIMIT_NETDEBUG(KERN_DEBUG
1322                                                "ADDRCONF: unspecified / multicast address "
1323                                                "assigned as unicast address on %s",
1324                                                dev->name);
1325                                 continue;
1326                         }
1327
1328                         score->rule = -1;
1329                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1330
1331                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1332                                 int minihiscore, miniscore;
1333
1334                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1335                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1336
1337                                 if (minihiscore > miniscore) {
1338                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1339                                             score->scopedist > 0) {
1340                                                 /*
1341                                                  * special case:
1342                                                  * each remaining entry
1343                                                  * has too small (not enough)
1344                                                  * scope, because ifa entries
1345                                                  * are sorted by their scope
1346                                                  * values.
1347                                                  */
1348                                                 goto try_nextdev;
1349                                         }
1350                                         break;
1351                                 } else if (minihiscore < miniscore) {
1352                                         if (hiscore->ifa)
1353                                                 in6_ifa_put(hiscore->ifa);
1354
1355                                         in6_ifa_hold(score->ifa);
1356
1357                                         swap(hiscore, score);
1358
1359                                         /* restore our iterator */
1360                                         score->ifa = hiscore->ifa;
1361
1362                                         break;
1363                                 }
1364                         }
1365                 }
1366 try_nextdev:
1367                 read_unlock_bh(&idev->lock);
1368         }
1369         rcu_read_unlock();
1370
1371         if (!hiscore->ifa)
1372                 return -EADDRNOTAVAIL;
1373
1374         *saddr = hiscore->ifa->addr;
1375         in6_ifa_put(hiscore->ifa);
1376         return 0;
1377 }
1378 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1379
1380 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1381                     unsigned char banned_flags)
1382 {
1383         struct inet6_dev *idev;
1384         int err = -EADDRNOTAVAIL;
1385
1386         rcu_read_lock();
1387         idev = __in6_dev_get(dev);
1388         if (idev) {
1389                 struct inet6_ifaddr *ifp;
1390
1391                 read_lock_bh(&idev->lock);
1392                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1393                         if (ifp->scope == IFA_LINK &&
1394                             !(ifp->flags & banned_flags)) {
1395                                 *addr = ifp->addr;
1396                                 err = 0;
1397                                 break;
1398                         }
1399                 }
1400                 read_unlock_bh(&idev->lock);
1401         }
1402         rcu_read_unlock();
1403         return err;
1404 }
1405
1406 static int ipv6_count_addresses(struct inet6_dev *idev)
1407 {
1408         int cnt = 0;
1409         struct inet6_ifaddr *ifp;
1410
1411         read_lock_bh(&idev->lock);
1412         list_for_each_entry(ifp, &idev->addr_list, if_list)
1413                 cnt++;
1414         read_unlock_bh(&idev->lock);
1415         return cnt;
1416 }
1417
1418 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1419                   struct net_device *dev, int strict)
1420 {
1421         struct inet6_ifaddr *ifp;
1422         unsigned int hash = inet6_addr_hash(addr);
1423
1424         rcu_read_lock_bh();
1425         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1426                 if (!net_eq(dev_net(ifp->idev->dev), net))
1427                         continue;
1428                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1429                     !(ifp->flags&IFA_F_TENTATIVE) &&
1430                     (dev == NULL || ifp->idev->dev == dev ||
1431                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1432                         rcu_read_unlock_bh();
1433                         return 1;
1434                 }
1435         }
1436
1437         rcu_read_unlock_bh();
1438         return 0;
1439 }
1440 EXPORT_SYMBOL(ipv6_chk_addr);
1441
1442 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1443                                struct net_device *dev)
1444 {
1445         unsigned int hash = inet6_addr_hash(addr);
1446         struct inet6_ifaddr *ifp;
1447
1448         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1449                 if (!net_eq(dev_net(ifp->idev->dev), net))
1450                         continue;
1451                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1452                         if (dev == NULL || ifp->idev->dev == dev)
1453                                 return true;
1454                 }
1455         }
1456         return false;
1457 }
1458
1459 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1460 {
1461         struct inet6_dev *idev;
1462         struct inet6_ifaddr *ifa;
1463         int     onlink;
1464
1465         onlink = 0;
1466         rcu_read_lock();
1467         idev = __in6_dev_get(dev);
1468         if (idev) {
1469                 read_lock_bh(&idev->lock);
1470                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1471                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1472                                                    ifa->prefix_len);
1473                         if (onlink)
1474                                 break;
1475                 }
1476                 read_unlock_bh(&idev->lock);
1477         }
1478         rcu_read_unlock();
1479         return onlink;
1480 }
1481 EXPORT_SYMBOL(ipv6_chk_prefix);
1482
1483 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1484                                      struct net_device *dev, int strict)
1485 {
1486         struct inet6_ifaddr *ifp, *result = NULL;
1487         unsigned int hash = inet6_addr_hash(addr);
1488
1489         rcu_read_lock_bh();
1490         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1491                 if (!net_eq(dev_net(ifp->idev->dev), net))
1492                         continue;
1493                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1494                         if (dev == NULL || ifp->idev->dev == dev ||
1495                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1496                                 result = ifp;
1497                                 in6_ifa_hold(ifp);
1498                                 break;
1499                         }
1500                 }
1501         }
1502         rcu_read_unlock_bh();
1503
1504         return result;
1505 }
1506
1507 /* Gets referenced address, destroys ifaddr */
1508
1509 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1510 {
1511         if (ifp->flags&IFA_F_PERMANENT) {
1512                 spin_lock_bh(&ifp->lock);
1513                 addrconf_del_timer(ifp);
1514                 ifp->flags |= IFA_F_TENTATIVE;
1515                 if (dad_failed)
1516                         ifp->flags |= IFA_F_DADFAILED;
1517                 spin_unlock_bh(&ifp->lock);
1518                 if (dad_failed)
1519                         ipv6_ifa_notify(0, ifp);
1520                 in6_ifa_put(ifp);
1521 #ifdef CONFIG_IPV6_PRIVACY
1522         } else if (ifp->flags&IFA_F_TEMPORARY) {
1523                 struct inet6_ifaddr *ifpub;
1524                 spin_lock_bh(&ifp->lock);
1525                 ifpub = ifp->ifpub;
1526                 if (ifpub) {
1527                         in6_ifa_hold(ifpub);
1528                         spin_unlock_bh(&ifp->lock);
1529                         ipv6_create_tempaddr(ifpub, ifp);
1530                         in6_ifa_put(ifpub);
1531                 } else {
1532                         spin_unlock_bh(&ifp->lock);
1533                 }
1534                 ipv6_del_addr(ifp);
1535 #endif
1536         } else
1537                 ipv6_del_addr(ifp);
1538 }
1539
1540 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1541 {
1542         int err = -ENOENT;
1543
1544         spin_lock(&ifp->state_lock);
1545         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1546                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1547                 err = 0;
1548         }
1549         spin_unlock(&ifp->state_lock);
1550
1551         return err;
1552 }
1553
1554 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1555 {
1556         struct inet6_dev *idev = ifp->idev;
1557
1558         if (addrconf_dad_end(ifp)) {
1559                 in6_ifa_put(ifp);
1560                 return;
1561         }
1562
1563         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1564                              ifp->idev->dev->name, &ifp->addr);
1565
1566         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1567                 struct in6_addr addr;
1568
1569                 addr.s6_addr32[0] = htonl(0xfe800000);
1570                 addr.s6_addr32[1] = 0;
1571
1572                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1573                     ipv6_addr_equal(&ifp->addr, &addr)) {
1574                         /* DAD failed for link-local based on MAC address */
1575                         idev->cnf.disable_ipv6 = 1;
1576
1577                         pr_info("%s: IPv6 being disabled!\n",
1578                                 ifp->idev->dev->name);
1579                 }
1580         }
1581
1582         addrconf_dad_stop(ifp, 1);
1583 }
1584
1585 /* Join to solicited addr multicast group. */
1586
1587 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1588 {
1589         struct in6_addr maddr;
1590
1591         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1592                 return;
1593
1594         addrconf_addr_solict_mult(addr, &maddr);
1595         ipv6_dev_mc_inc(dev, &maddr);
1596 }
1597
1598 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1599 {
1600         struct in6_addr maddr;
1601
1602         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1603                 return;
1604
1605         addrconf_addr_solict_mult(addr, &maddr);
1606         __ipv6_dev_mc_dec(idev, &maddr);
1607 }
1608
1609 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1610 {
1611         struct in6_addr addr;
1612         if (ifp->prefix_len == 127) /* RFC 6164 */
1613                 return;
1614         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1615         if (ipv6_addr_any(&addr))
1616                 return;
1617         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1618 }
1619
1620 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1621 {
1622         struct in6_addr addr;
1623         if (ifp->prefix_len == 127) /* RFC 6164 */
1624                 return;
1625         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1626         if (ipv6_addr_any(&addr))
1627                 return;
1628         __ipv6_dev_ac_dec(ifp->idev, &addr);
1629 }
1630
1631 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1632 {
1633         if (dev->addr_len != ETH_ALEN)
1634                 return -1;
1635         memcpy(eui, dev->dev_addr, 3);
1636         memcpy(eui + 5, dev->dev_addr + 3, 3);
1637
1638         /*
1639          * The zSeries OSA network cards can be shared among various
1640          * OS instances, but the OSA cards have only one MAC address.
1641          * This leads to duplicate address conflicts in conjunction
1642          * with IPv6 if more than one instance uses the same card.
1643          *
1644          * The driver for these cards can deliver a unique 16-bit
1645          * identifier for each instance sharing the same card.  It is
1646          * placed instead of 0xFFFE in the interface identifier.  The
1647          * "u" bit of the interface identifier is not inverted in this
1648          * case.  Hence the resulting interface identifier has local
1649          * scope according to RFC2373.
1650          */
1651         if (dev->dev_id) {
1652                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1653                 eui[4] = dev->dev_id & 0xFF;
1654         } else {
1655                 eui[3] = 0xFF;
1656                 eui[4] = 0xFE;
1657                 eui[0] ^= 2;
1658         }
1659         return 0;
1660 }
1661
1662 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1663 {
1664         if (dev->addr_len != IEEE802154_ADDR_LEN)
1665                 return -1;
1666         memcpy(eui, dev->dev_addr, 8);
1667         eui[0] ^= 2;
1668         return 0;
1669 }
1670
1671 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1672 {
1673         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1674         if (dev->addr_len != ARCNET_ALEN)
1675                 return -1;
1676         memset(eui, 0, 7);
1677         eui[7] = *(u8 *)dev->dev_addr;
1678         return 0;
1679 }
1680
1681 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1682 {
1683         if (dev->addr_len != INFINIBAND_ALEN)
1684                 return -1;
1685         memcpy(eui, dev->dev_addr + 12, 8);
1686         eui[0] |= 2;
1687         return 0;
1688 }
1689
1690 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1691 {
1692         if (addr == 0)
1693                 return -1;
1694         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1695                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1696                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1697                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1698                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1699                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1700         eui[1] = 0;
1701         eui[2] = 0x5E;
1702         eui[3] = 0xFE;
1703         memcpy(eui + 4, &addr, 4);
1704         return 0;
1705 }
1706
1707 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1708 {
1709         if (dev->priv_flags & IFF_ISATAP)
1710                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1711         return -1;
1712 }
1713
1714 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1715 {
1716         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1717 }
1718
1719 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1720 {
1721         switch (dev->type) {
1722         case ARPHRD_ETHER:
1723         case ARPHRD_FDDI:
1724                 return addrconf_ifid_eui48(eui, dev);
1725         case ARPHRD_ARCNET:
1726                 return addrconf_ifid_arcnet(eui, dev);
1727         case ARPHRD_INFINIBAND:
1728                 return addrconf_ifid_infiniband(eui, dev);
1729         case ARPHRD_SIT:
1730                 return addrconf_ifid_sit(eui, dev);
1731         case ARPHRD_IPGRE:
1732                 return addrconf_ifid_gre(eui, dev);
1733         case ARPHRD_IEEE802154:
1734                 return addrconf_ifid_eui64(eui, dev);
1735         }
1736         return -1;
1737 }
1738
1739 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1740 {
1741         int err = -1;
1742         struct inet6_ifaddr *ifp;
1743
1744         read_lock_bh(&idev->lock);
1745         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1746                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1747                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1748                         err = 0;
1749                         break;
1750                 }
1751         }
1752         read_unlock_bh(&idev->lock);
1753         return err;
1754 }
1755
1756 #ifdef CONFIG_IPV6_PRIVACY
1757 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1758 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1759 {
1760 regen:
1761         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1762         idev->rndid[0] &= ~0x02;
1763
1764         /*
1765          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1766          * check if generated address is not inappropriate
1767          *
1768          *  - Reserved subnet anycast (RFC 2526)
1769          *      11111101 11....11 1xxxxxxx
1770          *  - ISATAP (RFC4214) 6.1
1771          *      00-00-5E-FE-xx-xx-xx-xx
1772          *  - value 0
1773          *  - XXX: already assigned to an address on the device
1774          */
1775         if (idev->rndid[0] == 0xfd &&
1776             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1777             (idev->rndid[7]&0x80))
1778                 goto regen;
1779         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1780                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1781                         goto regen;
1782                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1783                         goto regen;
1784         }
1785 }
1786
1787 static void ipv6_regen_rndid(unsigned long data)
1788 {
1789         struct inet6_dev *idev = (struct inet6_dev *) data;
1790         unsigned long expires;
1791
1792         rcu_read_lock_bh();
1793         write_lock_bh(&idev->lock);
1794
1795         if (idev->dead)
1796                 goto out;
1797
1798         __ipv6_regen_rndid(idev);
1799
1800         expires = jiffies +
1801                 idev->cnf.temp_prefered_lft * HZ -
1802                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1803                 idev->cnf.max_desync_factor * HZ;
1804         if (time_before(expires, jiffies)) {
1805                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1806                         __func__, idev->dev->name);
1807                 goto out;
1808         }
1809
1810         if (!mod_timer(&idev->regen_timer, expires))
1811                 in6_dev_hold(idev);
1812
1813 out:
1814         write_unlock_bh(&idev->lock);
1815         rcu_read_unlock_bh();
1816         in6_dev_put(idev);
1817 }
1818
1819 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1820 {
1821         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1822                 __ipv6_regen_rndid(idev);
1823 }
1824 #endif
1825
1826 /*
1827  *      Add prefix route.
1828  */
1829
1830 static void
1831 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1832                       unsigned long expires, u32 flags)
1833 {
1834         struct fib6_config cfg = {
1835                 .fc_table = RT6_TABLE_PREFIX,
1836                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1837                 .fc_ifindex = dev->ifindex,
1838                 .fc_expires = expires,
1839                 .fc_dst_len = plen,
1840                 .fc_flags = RTF_UP | flags,
1841                 .fc_nlinfo.nl_net = dev_net(dev),
1842                 .fc_protocol = RTPROT_KERNEL,
1843         };
1844
1845         cfg.fc_dst = *pfx;
1846
1847         /* Prevent useless cloning on PtP SIT.
1848            This thing is done here expecting that the whole
1849            class of non-broadcast devices need not cloning.
1850          */
1851 #if IS_ENABLED(CONFIG_IPV6_SIT)
1852         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1853                 cfg.fc_flags |= RTF_NONEXTHOP;
1854 #endif
1855
1856         ip6_route_add(&cfg);
1857 }
1858
1859
1860 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1861                                                   int plen,
1862                                                   const struct net_device *dev,
1863                                                   u32 flags, u32 noflags)
1864 {
1865         struct fib6_node *fn;
1866         struct rt6_info *rt = NULL;
1867         struct fib6_table *table;
1868
1869         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1870         if (table == NULL)
1871                 return NULL;
1872
1873         read_lock_bh(&table->tb6_lock);
1874         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1875         if (!fn)
1876                 goto out;
1877         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1878                 if (rt->dst.dev->ifindex != dev->ifindex)
1879                         continue;
1880                 if ((rt->rt6i_flags & flags) != flags)
1881                         continue;
1882                 if ((rt->rt6i_flags & noflags) != 0)
1883                         continue;
1884                 dst_hold(&rt->dst);
1885                 break;
1886         }
1887 out:
1888         read_unlock_bh(&table->tb6_lock);
1889         return rt;
1890 }
1891
1892
1893 /* Create "default" multicast route to the interface */
1894
1895 static void addrconf_add_mroute(struct net_device *dev)
1896 {
1897         struct fib6_config cfg = {
1898                 .fc_table = RT6_TABLE_LOCAL,
1899                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1900                 .fc_ifindex = dev->ifindex,
1901                 .fc_dst_len = 8,
1902                 .fc_flags = RTF_UP,
1903                 .fc_nlinfo.nl_net = dev_net(dev),
1904         };
1905
1906         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1907
1908         ip6_route_add(&cfg);
1909 }
1910
1911 #if IS_ENABLED(CONFIG_IPV6_SIT)
1912 static void sit_route_add(struct net_device *dev)
1913 {
1914         struct fib6_config cfg = {
1915                 .fc_table = RT6_TABLE_MAIN,
1916                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1917                 .fc_ifindex = dev->ifindex,
1918                 .fc_dst_len = 96,
1919                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1920                 .fc_nlinfo.nl_net = dev_net(dev),
1921         };
1922
1923         /* prefix length - 96 bits "::d.d.d.d" */
1924         ip6_route_add(&cfg);
1925 }
1926 #endif
1927
1928 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1929 {
1930         struct inet6_dev *idev;
1931
1932         ASSERT_RTNL();
1933
1934         idev = ipv6_find_idev(dev);
1935         if (!idev)
1936                 return ERR_PTR(-ENOBUFS);
1937
1938         if (idev->cnf.disable_ipv6)
1939                 return ERR_PTR(-EACCES);
1940
1941         /* Add default multicast route */
1942         if (!(dev->flags & IFF_LOOPBACK))
1943                 addrconf_add_mroute(dev);
1944
1945         return idev;
1946 }
1947
1948 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1949 {
1950         struct prefix_info *pinfo;
1951         __u32 valid_lft;
1952         __u32 prefered_lft;
1953         int addr_type;
1954         struct inet6_dev *in6_dev;
1955         struct net *net = dev_net(dev);
1956
1957         pinfo = (struct prefix_info *) opt;
1958
1959         if (len < sizeof(struct prefix_info)) {
1960                 ADBG(("addrconf: prefix option too short\n"));
1961                 return;
1962         }
1963
1964         /*
1965          *      Validation checks ([ADDRCONF], page 19)
1966          */
1967
1968         addr_type = ipv6_addr_type(&pinfo->prefix);
1969
1970         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1971                 return;
1972
1973         valid_lft = ntohl(pinfo->valid);
1974         prefered_lft = ntohl(pinfo->prefered);
1975
1976         if (prefered_lft > valid_lft) {
1977                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1978                 return;
1979         }
1980
1981         in6_dev = in6_dev_get(dev);
1982
1983         if (in6_dev == NULL) {
1984                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1985                                     dev->name);
1986                 return;
1987         }
1988
1989         /*
1990          *      Two things going on here:
1991          *      1) Add routes for on-link prefixes
1992          *      2) Configure prefixes with the auto flag set
1993          */
1994
1995         if (pinfo->onlink) {
1996                 struct rt6_info *rt;
1997                 unsigned long rt_expires;
1998
1999                 /* Avoid arithmetic overflow. Really, we could
2000                  * save rt_expires in seconds, likely valid_lft,
2001                  * but it would require division in fib gc, that it
2002                  * not good.
2003                  */
2004                 if (HZ > USER_HZ)
2005                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2006                 else
2007                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2008
2009                 if (addrconf_finite_timeout(rt_expires))
2010                         rt_expires *= HZ;
2011
2012                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2013                                                pinfo->prefix_len,
2014                                                dev,
2015                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2016                                                RTF_GATEWAY | RTF_DEFAULT);
2017
2018                 if (rt) {
2019                         /* Autoconf prefix route */
2020                         if (valid_lft == 0) {
2021                                 ip6_del_rt(rt);
2022                                 rt = NULL;
2023                         } else if (addrconf_finite_timeout(rt_expires)) {
2024                                 /* not infinity */
2025                                 rt6_set_expires(rt, jiffies + rt_expires);
2026                         } else {
2027                                 rt6_clean_expires(rt);
2028                         }
2029                 } else if (valid_lft) {
2030                         clock_t expires = 0;
2031                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2032                         if (addrconf_finite_timeout(rt_expires)) {
2033                                 /* not infinity */
2034                                 flags |= RTF_EXPIRES;
2035                                 expires = jiffies_to_clock_t(rt_expires);
2036                         }
2037                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2038                                               dev, expires, flags);
2039                 }
2040                 ip6_rt_put(rt);
2041         }
2042
2043         /* Try to figure out our local address for this prefix */
2044
2045         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2046                 struct inet6_ifaddr *ifp;
2047                 struct in6_addr addr;
2048                 int create = 0, update_lft = 0;
2049
2050                 if (pinfo->prefix_len == 64) {
2051                         memcpy(&addr, &pinfo->prefix, 8);
2052                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2053                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2054                                 in6_dev_put(in6_dev);
2055                                 return;
2056                         }
2057                         goto ok;
2058                 }
2059                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2060                                     pinfo->prefix_len);
2061                 in6_dev_put(in6_dev);
2062                 return;
2063
2064 ok:
2065
2066                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2067
2068                 if (ifp == NULL && valid_lft) {
2069                         int max_addresses = in6_dev->cnf.max_addresses;
2070                         u32 addr_flags = 0;
2071
2072 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2073                         if (in6_dev->cnf.optimistic_dad &&
2074                             !net->ipv6.devconf_all->forwarding && sllao)
2075                                 addr_flags = IFA_F_OPTIMISTIC;
2076 #endif
2077
2078                         /* Do not allow to create too much of autoconfigured
2079                          * addresses; this would be too easy way to crash kernel.
2080                          */
2081                         if (!max_addresses ||
2082                             ipv6_count_addresses(in6_dev) < max_addresses)
2083                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2084                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2085                                                     addr_flags);
2086
2087                         if (IS_ERR_OR_NULL(ifp)) {
2088                                 in6_dev_put(in6_dev);
2089                                 return;
2090                         }
2091
2092                         update_lft = create = 1;
2093                         ifp->cstamp = jiffies;
2094                         addrconf_dad_start(ifp);
2095                 }
2096
2097                 if (ifp) {
2098                         int flags;
2099                         unsigned long now;
2100 #ifdef CONFIG_IPV6_PRIVACY
2101                         struct inet6_ifaddr *ift;
2102 #endif
2103                         u32 stored_lft;
2104
2105                         /* update lifetime (RFC2462 5.5.3 e) */
2106                         spin_lock(&ifp->lock);
2107                         now = jiffies;
2108                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2109                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2110                         else
2111                                 stored_lft = 0;
2112                         if (!update_lft && stored_lft) {
2113                                 if (valid_lft > MIN_VALID_LIFETIME ||
2114                                     valid_lft > stored_lft)
2115                                         update_lft = 1;
2116                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2117                                         /* valid_lft <= stored_lft is always true */
2118                                         /*
2119                                          * RFC 4862 Section 5.5.3e:
2120                                          * "Note that the preferred lifetime of
2121                                          *  the corresponding address is always
2122                                          *  reset to the Preferred Lifetime in
2123                                          *  the received Prefix Information
2124                                          *  option, regardless of whether the
2125                                          *  valid lifetime is also reset or
2126                                          *  ignored."
2127                                          *
2128                                          *  So if the preferred lifetime in
2129                                          *  this advertisement is different
2130                                          *  than what we have stored, but the
2131                                          *  valid lifetime is invalid, just
2132                                          *  reset prefered_lft.
2133                                          *
2134                                          *  We must set the valid lifetime
2135                                          *  to the stored lifetime since we'll
2136                                          *  be updating the timestamp below,
2137                                          *  else we'll set it back to the
2138                                          *  minimum.
2139                                          */
2140                                         if (prefered_lft != ifp->prefered_lft) {
2141                                                 valid_lft = stored_lft;
2142                                                 update_lft = 1;
2143                                         }
2144                                 } else {
2145                                         valid_lft = MIN_VALID_LIFETIME;
2146                                         if (valid_lft < prefered_lft)
2147                                                 prefered_lft = valid_lft;
2148                                         update_lft = 1;
2149                                 }
2150                         }
2151
2152                         if (update_lft) {
2153                                 ifp->valid_lft = valid_lft;
2154                                 ifp->prefered_lft = prefered_lft;
2155                                 ifp->tstamp = now;
2156                                 flags = ifp->flags;
2157                                 ifp->flags &= ~IFA_F_DEPRECATED;
2158                                 spin_unlock(&ifp->lock);
2159
2160                                 if (!(flags&IFA_F_TENTATIVE))
2161                                         ipv6_ifa_notify(0, ifp);
2162                         } else
2163                                 spin_unlock(&ifp->lock);
2164
2165 #ifdef CONFIG_IPV6_PRIVACY
2166                         read_lock_bh(&in6_dev->lock);
2167                         /* update all temporary addresses in the list */
2168                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2169                                             tmp_list) {
2170                                 int age, max_valid, max_prefered;
2171
2172                                 if (ifp != ift->ifpub)
2173                                         continue;
2174
2175                                 /*
2176                                  * RFC 4941 section 3.3:
2177                                  * If a received option will extend the lifetime
2178                                  * of a public address, the lifetimes of
2179                                  * temporary addresses should be extended,
2180                                  * subject to the overall constraint that no
2181                                  * temporary addresses should ever remain
2182                                  * "valid" or "preferred" for a time longer than
2183                                  * (TEMP_VALID_LIFETIME) or
2184                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2185                                  * respectively.
2186                                  */
2187                                 age = (now - ift->cstamp) / HZ;
2188                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2189                                 if (max_valid < 0)
2190                                         max_valid = 0;
2191
2192                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2193                                                in6_dev->cnf.max_desync_factor -
2194                                                age;
2195                                 if (max_prefered < 0)
2196                                         max_prefered = 0;
2197
2198                                 if (valid_lft > max_valid)
2199                                         valid_lft = max_valid;
2200
2201                                 if (prefered_lft > max_prefered)
2202                                         prefered_lft = max_prefered;
2203
2204                                 spin_lock(&ift->lock);
2205                                 flags = ift->flags;
2206                                 ift->valid_lft = valid_lft;
2207                                 ift->prefered_lft = prefered_lft;
2208                                 ift->tstamp = now;
2209                                 if (prefered_lft > 0)
2210                                         ift->flags &= ~IFA_F_DEPRECATED;
2211
2212                                 spin_unlock(&ift->lock);
2213                                 if (!(flags&IFA_F_TENTATIVE))
2214                                         ipv6_ifa_notify(0, ift);
2215                         }
2216
2217                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2218                                 /*
2219                                  * When a new public address is created as
2220                                  * described in [ADDRCONF], also create a new
2221                                  * temporary address. Also create a temporary
2222                                  * address if it's enabled but no temporary
2223                                  * address currently exists.
2224                                  */
2225                                 read_unlock_bh(&in6_dev->lock);
2226                                 ipv6_create_tempaddr(ifp, NULL);
2227                         } else {
2228                                 read_unlock_bh(&in6_dev->lock);
2229                         }
2230 #endif
2231                         in6_ifa_put(ifp);
2232                         addrconf_verify(0);
2233                 }
2234         }
2235         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2236         in6_dev_put(in6_dev);
2237 }
2238
2239 /*
2240  *      Set destination address.
2241  *      Special case for SIT interfaces where we create a new "virtual"
2242  *      device.
2243  */
2244 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2245 {
2246         struct in6_ifreq ireq;
2247         struct net_device *dev;
2248         int err = -EINVAL;
2249
2250         rtnl_lock();
2251
2252         err = -EFAULT;
2253         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2254                 goto err_exit;
2255
2256         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2257
2258         err = -ENODEV;
2259         if (dev == NULL)
2260                 goto err_exit;
2261
2262 #if IS_ENABLED(CONFIG_IPV6_SIT)
2263         if (dev->type == ARPHRD_SIT) {
2264                 const struct net_device_ops *ops = dev->netdev_ops;
2265                 struct ifreq ifr;
2266                 struct ip_tunnel_parm p;
2267
2268                 err = -EADDRNOTAVAIL;
2269                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2270                         goto err_exit;
2271
2272                 memset(&p, 0, sizeof(p));
2273                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2274                 p.iph.saddr = 0;
2275                 p.iph.version = 4;
2276                 p.iph.ihl = 5;
2277                 p.iph.protocol = IPPROTO_IPV6;
2278                 p.iph.ttl = 64;
2279                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2280
2281                 if (ops->ndo_do_ioctl) {
2282                         mm_segment_t oldfs = get_fs();
2283
2284                         set_fs(KERNEL_DS);
2285                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2286                         set_fs(oldfs);
2287                 } else
2288                         err = -EOPNOTSUPP;
2289
2290                 if (err == 0) {
2291                         err = -ENOBUFS;
2292                         dev = __dev_get_by_name(net, p.name);
2293                         if (!dev)
2294                                 goto err_exit;
2295                         err = dev_open(dev);
2296                 }
2297         }
2298 #endif
2299
2300 err_exit:
2301         rtnl_unlock();
2302         return err;
2303 }
2304
2305 /*
2306  *      Manual configuration of address on an interface
2307  */
2308 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2309                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2310                           __u32 valid_lft)
2311 {
2312         struct inet6_ifaddr *ifp;
2313         struct inet6_dev *idev;
2314         struct net_device *dev;
2315         int scope;
2316         u32 flags;
2317         clock_t expires;
2318         unsigned long timeout;
2319
2320         ASSERT_RTNL();
2321
2322         if (plen > 128)
2323                 return -EINVAL;
2324
2325         /* check the lifetime */
2326         if (!valid_lft || prefered_lft > valid_lft)
2327                 return -EINVAL;
2328
2329         dev = __dev_get_by_index(net, ifindex);
2330         if (!dev)
2331                 return -ENODEV;
2332
2333         idev = addrconf_add_dev(dev);
2334         if (IS_ERR(idev))
2335                 return PTR_ERR(idev);
2336
2337         scope = ipv6_addr_scope(pfx);
2338
2339         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2340         if (addrconf_finite_timeout(timeout)) {
2341                 expires = jiffies_to_clock_t(timeout * HZ);
2342                 valid_lft = timeout;
2343                 flags = RTF_EXPIRES;
2344         } else {
2345                 expires = 0;
2346                 flags = 0;
2347                 ifa_flags |= IFA_F_PERMANENT;
2348         }
2349
2350         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2351         if (addrconf_finite_timeout(timeout)) {
2352                 if (timeout == 0)
2353                         ifa_flags |= IFA_F_DEPRECATED;
2354                 prefered_lft = timeout;
2355         }
2356
2357         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2358
2359         if (!IS_ERR(ifp)) {
2360                 spin_lock_bh(&ifp->lock);
2361                 ifp->valid_lft = valid_lft;
2362                 ifp->prefered_lft = prefered_lft;
2363                 ifp->tstamp = jiffies;
2364                 spin_unlock_bh(&ifp->lock);
2365
2366                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2367                                       expires, flags);
2368                 /*
2369                  * Note that section 3.1 of RFC 4429 indicates
2370                  * that the Optimistic flag should not be set for
2371                  * manually configured addresses
2372                  */
2373                 addrconf_dad_start(ifp);
2374                 in6_ifa_put(ifp);
2375                 addrconf_verify(0);
2376                 return 0;
2377         }
2378
2379         return PTR_ERR(ifp);
2380 }
2381
2382 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2383                           unsigned int plen)
2384 {
2385         struct inet6_ifaddr *ifp;
2386         struct inet6_dev *idev;
2387         struct net_device *dev;
2388
2389         if (plen > 128)
2390                 return -EINVAL;
2391
2392         dev = __dev_get_by_index(net, ifindex);
2393         if (!dev)
2394                 return -ENODEV;
2395
2396         if ((idev = __in6_dev_get(dev)) == NULL)
2397                 return -ENXIO;
2398
2399         read_lock_bh(&idev->lock);
2400         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2401                 if (ifp->prefix_len == plen &&
2402                     ipv6_addr_equal(pfx, &ifp->addr)) {
2403                         in6_ifa_hold(ifp);
2404                         read_unlock_bh(&idev->lock);
2405
2406                         ipv6_del_addr(ifp);
2407
2408                         /* If the last address is deleted administratively,
2409                            disable IPv6 on this interface.
2410                          */
2411                         if (list_empty(&idev->addr_list))
2412                                 addrconf_ifdown(idev->dev, 1);
2413                         return 0;
2414                 }
2415         }
2416         read_unlock_bh(&idev->lock);
2417         return -EADDRNOTAVAIL;
2418 }
2419
2420
2421 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2422 {
2423         struct in6_ifreq ireq;
2424         int err;
2425
2426         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2427                 return -EPERM;
2428
2429         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2430                 return -EFAULT;
2431
2432         rtnl_lock();
2433         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2434                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2435                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2436         rtnl_unlock();
2437         return err;
2438 }
2439
2440 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2441 {
2442         struct in6_ifreq ireq;
2443         int err;
2444
2445         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2446                 return -EPERM;
2447
2448         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2449                 return -EFAULT;
2450
2451         rtnl_lock();
2452         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2453                              ireq.ifr6_prefixlen);
2454         rtnl_unlock();
2455         return err;
2456 }
2457
2458 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2459                      int plen, int scope)
2460 {
2461         struct inet6_ifaddr *ifp;
2462
2463         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2464         if (!IS_ERR(ifp)) {
2465                 spin_lock_bh(&ifp->lock);
2466                 ifp->flags &= ~IFA_F_TENTATIVE;
2467                 spin_unlock_bh(&ifp->lock);
2468                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2469                 in6_ifa_put(ifp);
2470         }
2471 }
2472
2473 #if IS_ENABLED(CONFIG_IPV6_SIT)
2474 static void sit_add_v4_addrs(struct inet6_dev *idev)
2475 {
2476         struct in6_addr addr;
2477         struct net_device *dev;
2478         struct net *net = dev_net(idev->dev);
2479         int scope;
2480
2481         ASSERT_RTNL();
2482
2483         memset(&addr, 0, sizeof(struct in6_addr));
2484         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2485
2486         if (idev->dev->flags&IFF_POINTOPOINT) {
2487                 addr.s6_addr32[0] = htonl(0xfe800000);
2488                 scope = IFA_LINK;
2489         } else {
2490                 scope = IPV6_ADDR_COMPATv4;
2491         }
2492
2493         if (addr.s6_addr32[3]) {
2494                 add_addr(idev, &addr, 128, scope);
2495                 return;
2496         }
2497
2498         for_each_netdev(net, dev) {
2499                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2500                 if (in_dev && (dev->flags & IFF_UP)) {
2501                         struct in_ifaddr *ifa;
2502
2503                         int flag = scope;
2504
2505                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2506                                 int plen;
2507
2508                                 addr.s6_addr32[3] = ifa->ifa_local;
2509
2510                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2511                                         continue;
2512                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2513                                         if (idev->dev->flags&IFF_POINTOPOINT)
2514                                                 continue;
2515                                         flag |= IFA_HOST;
2516                                 }
2517                                 if (idev->dev->flags&IFF_POINTOPOINT)
2518                                         plen = 64;
2519                                 else
2520                                         plen = 96;
2521
2522                                 add_addr(idev, &addr, plen, flag);
2523                         }
2524                 }
2525         }
2526 }
2527 #endif
2528
2529 static void init_loopback(struct net_device *dev)
2530 {
2531         struct inet6_dev  *idev;
2532         struct net_device *sp_dev;
2533         struct inet6_ifaddr *sp_ifa;
2534         struct rt6_info *sp_rt;
2535
2536         /* ::1 */
2537
2538         ASSERT_RTNL();
2539
2540         if ((idev = ipv6_find_idev(dev)) == NULL) {
2541                 pr_debug("%s: add_dev failed\n", __func__);
2542                 return;
2543         }
2544
2545         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2546
2547         /* Add routes to other interface's IPv6 addresses */
2548         for_each_netdev(dev_net(dev), sp_dev) {
2549                 if (!strcmp(sp_dev->name, dev->name))
2550                         continue;
2551
2552                 idev = __in6_dev_get(sp_dev);
2553                 if (!idev)
2554                         continue;
2555
2556                 read_lock_bh(&idev->lock);
2557                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2558
2559                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2560                                 continue;
2561
2562                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2563
2564                         /* Failure cases are ignored */
2565                         if (!IS_ERR(sp_rt))
2566                                 ip6_ins_rt(sp_rt);
2567                 }
2568                 read_unlock_bh(&idev->lock);
2569         }
2570 }
2571
2572 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2573 {
2574         struct inet6_ifaddr *ifp;
2575         u32 addr_flags = IFA_F_PERMANENT;
2576
2577 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2578         if (idev->cnf.optimistic_dad &&
2579             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2580                 addr_flags |= IFA_F_OPTIMISTIC;
2581 #endif
2582
2583
2584         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2585         if (!IS_ERR(ifp)) {
2586                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2587                 addrconf_dad_start(ifp);
2588                 in6_ifa_put(ifp);
2589         }
2590 }
2591
2592 static void addrconf_dev_config(struct net_device *dev)
2593 {
2594         struct in6_addr addr;
2595         struct inet6_dev *idev;
2596
2597         ASSERT_RTNL();
2598
2599         if ((dev->type != ARPHRD_ETHER) &&
2600             (dev->type != ARPHRD_FDDI) &&
2601             (dev->type != ARPHRD_ARCNET) &&
2602             (dev->type != ARPHRD_INFINIBAND) &&
2603             (dev->type != ARPHRD_IEEE802154)) {
2604                 /* Alas, we support only Ethernet autoconfiguration. */
2605                 return;
2606         }
2607
2608         idev = addrconf_add_dev(dev);
2609         if (IS_ERR(idev))
2610                 return;
2611
2612         memset(&addr, 0, sizeof(struct in6_addr));
2613         addr.s6_addr32[0] = htonl(0xFE800000);
2614
2615         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2616                 addrconf_add_linklocal(idev, &addr);
2617 }
2618
2619 #if IS_ENABLED(CONFIG_IPV6_SIT)
2620 static void addrconf_sit_config(struct net_device *dev)
2621 {
2622         struct inet6_dev *idev;
2623
2624         ASSERT_RTNL();
2625
2626         /*
2627          * Configure the tunnel with one of our IPv4
2628          * addresses... we should configure all of
2629          * our v4 addrs in the tunnel
2630          */
2631
2632         if ((idev = ipv6_find_idev(dev)) == NULL) {
2633                 pr_debug("%s: add_dev failed\n", __func__);
2634                 return;
2635         }
2636
2637         if (dev->priv_flags & IFF_ISATAP) {
2638                 struct in6_addr addr;
2639
2640                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2641                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2642                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2643                         addrconf_add_linklocal(idev, &addr);
2644                 return;
2645         }
2646
2647         sit_add_v4_addrs(idev);
2648
2649         if (dev->flags&IFF_POINTOPOINT)
2650                 addrconf_add_mroute(dev);
2651         else
2652                 sit_route_add(dev);
2653 }
2654 #endif
2655
2656 #if IS_ENABLED(CONFIG_NET_IPGRE)
2657 static void addrconf_gre_config(struct net_device *dev)
2658 {
2659         struct inet6_dev *idev;
2660         struct in6_addr addr;
2661
2662         pr_info("%s(%s)\n", __func__, dev->name);
2663
2664         ASSERT_RTNL();
2665
2666         if ((idev = ipv6_find_idev(dev)) == NULL) {
2667                 pr_debug("%s: add_dev failed\n", __func__);
2668                 return;
2669         }
2670
2671         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2672         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2673
2674         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2675                 addrconf_add_linklocal(idev, &addr);
2676 }
2677 #endif
2678
2679 static inline int
2680 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2681 {
2682         struct in6_addr lladdr;
2683
2684         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2685                 addrconf_add_linklocal(idev, &lladdr);
2686                 return 0;
2687         }
2688         return -1;
2689 }
2690
2691 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2692 {
2693         struct net_device *link_dev;
2694         struct net *net = dev_net(idev->dev);
2695
2696         /* first try to inherit the link-local address from the link device */
2697         if (idev->dev->iflink &&
2698             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2699                 if (!ipv6_inherit_linklocal(idev, link_dev))
2700                         return;
2701         }
2702         /* then try to inherit it from any device */
2703         for_each_netdev(net, link_dev) {
2704                 if (!ipv6_inherit_linklocal(idev, link_dev))
2705                         return;
2706         }
2707         pr_debug("init ip6-ip6: add_linklocal failed\n");
2708 }
2709
2710 /*
2711  * Autoconfigure tunnel with a link-local address so routing protocols,
2712  * DHCPv6, MLD etc. can be run over the virtual link
2713  */
2714
2715 static void addrconf_ip6_tnl_config(struct net_device *dev)
2716 {
2717         struct inet6_dev *idev;
2718
2719         ASSERT_RTNL();
2720
2721         idev = addrconf_add_dev(dev);
2722         if (IS_ERR(idev)) {
2723                 pr_debug("init ip6-ip6: add_dev failed\n");
2724                 return;
2725         }
2726         ip6_tnl_add_linklocal(idev);
2727 }
2728
2729 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2730                            void *data)
2731 {
2732         struct net_device *dev = (struct net_device *) data;
2733         struct inet6_dev *idev = __in6_dev_get(dev);
2734         int run_pending = 0;
2735         int err;
2736
2737         switch (event) {
2738         case NETDEV_REGISTER:
2739                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2740                         idev = ipv6_add_dev(dev);
2741                         if (!idev)
2742                                 return notifier_from_errno(-ENOMEM);
2743                 }
2744                 break;
2745
2746         case NETDEV_UP:
2747         case NETDEV_CHANGE:
2748                 if (dev->flags & IFF_SLAVE)
2749                         break;
2750
2751                 if (event == NETDEV_UP) {
2752                         if (!addrconf_qdisc_ok(dev)) {
2753                                 /* device is not ready yet. */
2754                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2755                                         dev->name);
2756                                 break;
2757                         }
2758
2759                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2760                                 idev = ipv6_add_dev(dev);
2761
2762                         if (idev) {
2763                                 idev->if_flags |= IF_READY;
2764                                 run_pending = 1;
2765                         }
2766                 } else {
2767                         if (!addrconf_qdisc_ok(dev)) {
2768                                 /* device is still not ready. */
2769                                 break;
2770                         }
2771
2772                         if (idev) {
2773                                 if (idev->if_flags & IF_READY)
2774                                         /* device is already configured. */
2775                                         break;
2776                                 idev->if_flags |= IF_READY;
2777                         }
2778
2779                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2780                                 dev->name);
2781
2782                         run_pending = 1;
2783                 }
2784
2785                 switch (dev->type) {
2786 #if IS_ENABLED(CONFIG_IPV6_SIT)
2787                 case ARPHRD_SIT:
2788                         addrconf_sit_config(dev);
2789                         break;
2790 #endif
2791 #if IS_ENABLED(CONFIG_NET_IPGRE)
2792                 case ARPHRD_IPGRE:
2793                         addrconf_gre_config(dev);
2794                         break;
2795 #endif
2796                 case ARPHRD_TUNNEL6:
2797                         addrconf_ip6_tnl_config(dev);
2798                         break;
2799                 case ARPHRD_LOOPBACK:
2800                         init_loopback(dev);
2801                         break;
2802
2803                 default:
2804                         addrconf_dev_config(dev);
2805                         break;
2806                 }
2807
2808                 if (idev) {
2809                         if (run_pending)
2810                                 addrconf_dad_run(idev);
2811
2812                         /*
2813                          * If the MTU changed during the interface down,
2814                          * when the interface up, the changed MTU must be
2815                          * reflected in the idev as well as routers.
2816                          */
2817                         if (idev->cnf.mtu6 != dev->mtu &&
2818                             dev->mtu >= IPV6_MIN_MTU) {
2819                                 rt6_mtu_change(dev, dev->mtu);
2820                                 idev->cnf.mtu6 = dev->mtu;
2821                         }
2822                         idev->tstamp = jiffies;
2823                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2824
2825                         /*
2826                          * If the changed mtu during down is lower than
2827                          * IPV6_MIN_MTU stop IPv6 on this interface.
2828                          */
2829                         if (dev->mtu < IPV6_MIN_MTU)
2830                                 addrconf_ifdown(dev, 1);
2831                 }
2832                 break;
2833
2834         case NETDEV_CHANGEMTU:
2835                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2836                         rt6_mtu_change(dev, dev->mtu);
2837                         idev->cnf.mtu6 = dev->mtu;
2838                         break;
2839                 }
2840
2841                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2842                         idev = ipv6_add_dev(dev);
2843                         if (idev)
2844                                 break;
2845                 }
2846
2847                 /*
2848                  * MTU falled under IPV6_MIN_MTU.
2849                  * Stop IPv6 on this interface.
2850                  */
2851
2852         case NETDEV_DOWN:
2853         case NETDEV_UNREGISTER:
2854                 /*
2855                  *      Remove all addresses from this interface.
2856                  */
2857                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2858                 break;
2859
2860         case NETDEV_CHANGENAME:
2861                 if (idev) {
2862                         snmp6_unregister_dev(idev);
2863                         addrconf_sysctl_unregister(idev);
2864                         addrconf_sysctl_register(idev);
2865                         err = snmp6_register_dev(idev);
2866                         if (err)
2867                                 return notifier_from_errno(err);
2868                 }
2869                 break;
2870
2871         case NETDEV_PRE_TYPE_CHANGE:
2872         case NETDEV_POST_TYPE_CHANGE:
2873                 addrconf_type_change(dev, event);
2874                 break;
2875         }
2876
2877         return NOTIFY_OK;
2878 }
2879
2880 /*
2881  *      addrconf module should be notified of a device going up
2882  */
2883 static struct notifier_block ipv6_dev_notf = {
2884         .notifier_call = addrconf_notify,
2885 };
2886
2887 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2888 {
2889         struct inet6_dev *idev;
2890         ASSERT_RTNL();
2891
2892         idev = __in6_dev_get(dev);
2893
2894         if (event == NETDEV_POST_TYPE_CHANGE)
2895                 ipv6_mc_remap(idev);
2896         else if (event == NETDEV_PRE_TYPE_CHANGE)
2897                 ipv6_mc_unmap(idev);
2898 }
2899
2900 static int addrconf_ifdown(struct net_device *dev, int how)
2901 {
2902         struct net *net = dev_net(dev);
2903         struct inet6_dev *idev;
2904         struct inet6_ifaddr *ifa;
2905         int state, i;
2906
2907         ASSERT_RTNL();
2908
2909         rt6_ifdown(net, dev);
2910         neigh_ifdown(&nd_tbl, dev);
2911
2912         idev = __in6_dev_get(dev);
2913         if (idev == NULL)
2914                 return -ENODEV;
2915
2916         /*
2917          * Step 1: remove reference to ipv6 device from parent device.
2918          *         Do not dev_put!
2919          */
2920         if (how) {
2921                 idev->dead = 1;
2922
2923                 /* protected by rtnl_lock */
2924                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2925
2926                 /* Step 1.5: remove snmp6 entry */
2927                 snmp6_unregister_dev(idev);
2928
2929         }
2930
2931         /* Step 2: clear hash table */
2932         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2933                 struct hlist_head *h = &inet6_addr_lst[i];
2934
2935                 spin_lock_bh(&addrconf_hash_lock);
2936         restart:
2937                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2938                         if (ifa->idev == idev) {
2939                                 hlist_del_init_rcu(&ifa->addr_lst);
2940                                 addrconf_del_timer(ifa);
2941                                 goto restart;
2942                         }
2943                 }
2944                 spin_unlock_bh(&addrconf_hash_lock);
2945         }
2946
2947         write_lock_bh(&idev->lock);
2948
2949         /* Step 2: clear flags for stateless addrconf */
2950         if (!how)
2951                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2952
2953 #ifdef CONFIG_IPV6_PRIVACY
2954         if (how && del_timer(&idev->regen_timer))
2955                 in6_dev_put(idev);
2956
2957         /* Step 3: clear tempaddr list */
2958         while (!list_empty(&idev->tempaddr_list)) {
2959                 ifa = list_first_entry(&idev->tempaddr_list,
2960                                        struct inet6_ifaddr, tmp_list);
2961                 list_del(&ifa->tmp_list);
2962                 write_unlock_bh(&idev->lock);
2963                 spin_lock_bh(&ifa->lock);
2964
2965                 if (ifa->ifpub) {
2966                         in6_ifa_put(ifa->ifpub);
2967                         ifa->ifpub = NULL;
2968                 }
2969                 spin_unlock_bh(&ifa->lock);
2970                 in6_ifa_put(ifa);
2971                 write_lock_bh(&idev->lock);
2972         }
2973 #endif
2974
2975         while (!list_empty(&idev->addr_list)) {
2976                 ifa = list_first_entry(&idev->addr_list,
2977                                        struct inet6_ifaddr, if_list);
2978                 addrconf_del_timer(ifa);
2979
2980                 list_del(&ifa->if_list);
2981
2982                 write_unlock_bh(&idev->lock);
2983
2984                 spin_lock_bh(&ifa->state_lock);
2985                 state = ifa->state;
2986                 ifa->state = INET6_IFADDR_STATE_DEAD;
2987                 spin_unlock_bh(&ifa->state_lock);
2988
2989                 if (state != INET6_IFADDR_STATE_DEAD) {
2990                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2991                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2992                 }
2993                 in6_ifa_put(ifa);
2994
2995                 write_lock_bh(&idev->lock);
2996         }
2997
2998         write_unlock_bh(&idev->lock);
2999
3000         /* Step 5: Discard multicast list */
3001         if (how)
3002                 ipv6_mc_destroy_dev(idev);
3003         else
3004                 ipv6_mc_down(idev);
3005
3006         idev->tstamp = jiffies;
3007
3008         /* Last: Shot the device (if unregistered) */
3009         if (how) {
3010                 addrconf_sysctl_unregister(idev);
3011                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3012                 neigh_ifdown(&nd_tbl, dev);
3013                 in6_dev_put(idev);
3014         }
3015         return 0;
3016 }
3017
3018 static void addrconf_rs_timer(unsigned long data)
3019 {
3020         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3021         struct inet6_dev *idev = ifp->idev;
3022
3023         read_lock(&idev->lock);
3024         if (idev->dead || !(idev->if_flags & IF_READY))
3025                 goto out;
3026
3027         if (!ipv6_accept_ra(idev))
3028                 goto out;
3029
3030         /* Announcement received after solicitation was sent */
3031         if (idev->if_flags & IF_RA_RCVD)
3032                 goto out;
3033
3034         spin_lock(&ifp->lock);
3035         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3036                 /* The wait after the last probe can be shorter */
3037                 addrconf_mod_timer(ifp, AC_RS,
3038                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3039                                    idev->cnf.rtr_solicit_delay :
3040                                    idev->cnf.rtr_solicit_interval);
3041                 spin_unlock(&ifp->lock);
3042
3043                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3044         } else {
3045                 spin_unlock(&ifp->lock);
3046                 /*
3047                  * Note: we do not support deprecated "all on-link"
3048                  * assumption any longer.
3049                  */
3050                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3051         }
3052
3053 out:
3054         read_unlock(&idev->lock);
3055         in6_ifa_put(ifp);
3056 }
3057
3058 /*
3059  *      Duplicate Address Detection
3060  */
3061 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3062 {
3063         unsigned long rand_num;
3064         struct inet6_dev *idev = ifp->idev;
3065
3066         if (ifp->flags & IFA_F_OPTIMISTIC)
3067                 rand_num = 0;
3068         else
3069                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3070
3071         ifp->probes = idev->cnf.dad_transmits;
3072         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3073 }
3074
3075 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3076 {
3077         struct inet6_dev *idev = ifp->idev;
3078         struct net_device *dev = idev->dev;
3079
3080         addrconf_join_solict(dev, &ifp->addr);
3081
3082         net_srandom(ifp->addr.s6_addr32[3]);
3083
3084         read_lock_bh(&idev->lock);
3085         spin_lock(&ifp->lock);
3086         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3087                 goto out;
3088
3089         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3090             idev->cnf.accept_dad < 1 ||
3091             !(ifp->flags&IFA_F_TENTATIVE) ||
3092             ifp->flags & IFA_F_NODAD) {
3093                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3094                 spin_unlock(&ifp->lock);
3095                 read_unlock_bh(&idev->lock);
3096
3097                 addrconf_dad_completed(ifp);
3098                 return;
3099         }
3100
3101         if (!(idev->if_flags & IF_READY)) {
3102                 spin_unlock(&ifp->lock);
3103                 read_unlock_bh(&idev->lock);
3104                 /*
3105                  * If the device is not ready:
3106                  * - keep it tentative if it is a permanent address.
3107                  * - otherwise, kill it.
3108                  */
3109                 in6_ifa_hold(ifp);
3110                 addrconf_dad_stop(ifp, 0);
3111                 return;
3112         }
3113
3114         /*
3115          * Optimistic nodes can start receiving
3116          * Frames right away
3117          */
3118         if (ifp->flags & IFA_F_OPTIMISTIC)
3119                 ip6_ins_rt(ifp->rt);
3120
3121         addrconf_dad_kick(ifp);
3122 out:
3123         spin_unlock(&ifp->lock);
3124         read_unlock_bh(&idev->lock);
3125 }
3126
3127 static void addrconf_dad_timer(unsigned long data)
3128 {
3129         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3130         struct inet6_dev *idev = ifp->idev;
3131         struct in6_addr mcaddr;
3132
3133         if (!ifp->probes && addrconf_dad_end(ifp))
3134                 goto out;
3135
3136         read_lock(&idev->lock);
3137         if (idev->dead || !(idev->if_flags & IF_READY)) {
3138                 read_unlock(&idev->lock);
3139                 goto out;
3140         }
3141
3142         spin_lock(&ifp->lock);
3143         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3144                 spin_unlock(&ifp->lock);
3145                 read_unlock(&idev->lock);
3146                 goto out;
3147         }
3148
3149         if (ifp->probes == 0) {
3150                 /*
3151                  * DAD was successful
3152                  */
3153
3154                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3155                 spin_unlock(&ifp->lock);
3156                 read_unlock(&idev->lock);
3157
3158                 addrconf_dad_completed(ifp);
3159
3160                 goto out;
3161         }
3162
3163         ifp->probes--;
3164         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3165         spin_unlock(&ifp->lock);
3166         read_unlock(&idev->lock);
3167
3168         /* send a neighbour solicitation for our addr */
3169         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3170         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3171 out:
3172         in6_ifa_put(ifp);
3173 }
3174
3175 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3176 {
3177         struct net_device *dev = ifp->idev->dev;
3178
3179         /*
3180          *      Configure the address for reception. Now it is valid.
3181          */
3182
3183         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3184
3185         /* If added prefix is link local and we are prepared to process
3186            router advertisements, start sending router solicitations.
3187          */
3188
3189         if (ipv6_accept_ra(ifp->idev) &&
3190             ifp->idev->cnf.rtr_solicits > 0 &&
3191             (dev->flags&IFF_LOOPBACK) == 0 &&
3192             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3193                 /*
3194                  *      If a host as already performed a random delay
3195                  *      [...] as part of DAD [...] there is no need
3196                  *      to delay again before sending the first RS
3197                  */
3198                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3199
3200                 spin_lock_bh(&ifp->lock);
3201                 ifp->probes = 1;
3202                 ifp->idev->if_flags |= IF_RS_SENT;
3203                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3204                 spin_unlock_bh(&ifp->lock);
3205         }
3206 }
3207
3208 static void addrconf_dad_run(struct inet6_dev *idev)
3209 {
3210         struct inet6_ifaddr *ifp;
3211
3212         read_lock_bh(&idev->lock);
3213         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3214                 spin_lock(&ifp->lock);
3215                 if (ifp->flags & IFA_F_TENTATIVE &&
3216                     ifp->state == INET6_IFADDR_STATE_DAD)
3217                         addrconf_dad_kick(ifp);
3218                 spin_unlock(&ifp->lock);
3219         }
3220         read_unlock_bh(&idev->lock);
3221 }
3222
3223 #ifdef CONFIG_PROC_FS
3224 struct if6_iter_state {
3225         struct seq_net_private p;
3226         int bucket;
3227         int offset;
3228 };
3229
3230 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3231 {
3232         struct inet6_ifaddr *ifa = NULL;
3233         struct if6_iter_state *state = seq->private;
3234         struct net *net = seq_file_net(seq);
3235         int p = 0;
3236
3237         /* initial bucket if pos is 0 */
3238         if (pos == 0) {
3239                 state->bucket = 0;
3240                 state->offset = 0;
3241         }
3242
3243         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3244                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3245                                          addr_lst) {
3246                         if (!net_eq(dev_net(ifa->idev->dev), net))
3247                                 continue;
3248                         /* sync with offset */
3249                         if (p < state->offset) {
3250                                 p++;
3251                                 continue;
3252                         }
3253                         state->offset++;
3254                         return ifa;
3255                 }
3256
3257                 /* prepare for next bucket */
3258                 state->offset = 0;
3259                 p = 0;
3260         }
3261         return NULL;
3262 }
3263
3264 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3265                                          struct inet6_ifaddr *ifa)
3266 {
3267         struct if6_iter_state *state = seq->private;
3268         struct net *net = seq_file_net(seq);
3269
3270         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3271                 if (!net_eq(dev_net(ifa->idev->dev), net))
3272                         continue;
3273                 state->offset++;
3274                 return ifa;
3275         }
3276
3277         while (++state->bucket < IN6_ADDR_HSIZE) {
3278                 state->offset = 0;
3279                 hlist_for_each_entry_rcu_bh(ifa,
3280                                      &inet6_addr_lst[state->bucket], addr_lst) {
3281                         if (!net_eq(dev_net(ifa->idev->dev), net))
3282                                 continue;
3283                         state->offset++;
3284                         return ifa;
3285                 }
3286         }
3287
3288         return NULL;
3289 }
3290
3291 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3292         __acquires(rcu_bh)
3293 {
3294         rcu_read_lock_bh();
3295         return if6_get_first(seq, *pos);
3296 }
3297
3298 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3299 {
3300         struct inet6_ifaddr *ifa;
3301
3302         ifa = if6_get_next(seq, v);
3303         ++*pos;
3304         return ifa;
3305 }
3306
3307 static void if6_seq_stop(struct seq_file *seq, void *v)
3308         __releases(rcu_bh)
3309 {
3310         rcu_read_unlock_bh();
3311 }
3312
3313 static int if6_seq_show(struct seq_file *seq, void *v)
3314 {
3315         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3316         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3317                    &ifp->addr,
3318                    ifp->idev->dev->ifindex,
3319                    ifp->prefix_len,
3320                    ifp->scope,
3321                    ifp->flags,
3322                    ifp->idev->dev->name);
3323         return 0;
3324 }
3325
3326 static const struct seq_operations if6_seq_ops = {
3327         .start  = if6_seq_start,
3328         .next   = if6_seq_next,
3329         .show   = if6_seq_show,
3330         .stop   = if6_seq_stop,
3331 };
3332
3333 static int if6_seq_open(struct inode *inode, struct file *file)
3334 {
3335         return seq_open_net(inode, file, &if6_seq_ops,
3336                             sizeof(struct if6_iter_state));
3337 }
3338
3339 static const struct file_operations if6_fops = {
3340         .owner          = THIS_MODULE,
3341         .open           = if6_seq_open,
3342         .read           = seq_read,
3343         .llseek         = seq_lseek,
3344         .release        = seq_release_net,
3345 };
3346
3347 static int __net_init if6_proc_net_init(struct net *net)
3348 {
3349         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3350                 return -ENOMEM;
3351         return 0;
3352 }
3353
3354 static void __net_exit if6_proc_net_exit(struct net *net)
3355 {
3356         remove_proc_entry("if_inet6", net->proc_net);
3357 }
3358
3359 static struct pernet_operations if6_proc_net_ops = {
3360        .init = if6_proc_net_init,
3361        .exit = if6_proc_net_exit,
3362 };
3363
3364 int __init if6_proc_init(void)
3365 {
3366         return register_pernet_subsys(&if6_proc_net_ops);
3367 }
3368
3369 void if6_proc_exit(void)
3370 {
3371         unregister_pernet_subsys(&if6_proc_net_ops);
3372 }
3373 #endif  /* CONFIG_PROC_FS */
3374
3375 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3376 /* Check if address is a home address configured on any interface. */
3377 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3378 {
3379         int ret = 0;
3380         struct inet6_ifaddr *ifp = NULL;
3381         unsigned int hash = inet6_addr_hash(addr);
3382
3383         rcu_read_lock_bh();
3384         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3385                 if (!net_eq(dev_net(ifp->idev->dev), net))
3386                         continue;
3387                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3388                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3389                         ret = 1;
3390                         break;
3391                 }
3392         }
3393         rcu_read_unlock_bh();
3394         return ret;
3395 }
3396 #endif
3397
3398 /*
3399  *      Periodic address status verification
3400  */
3401
3402 static void addrconf_verify(unsigned long foo)
3403 {
3404         unsigned long now, next, next_sec, next_sched;
3405         struct inet6_ifaddr *ifp;
3406         int i;
3407
3408         rcu_read_lock_bh();
3409         spin_lock(&addrconf_verify_lock);
3410         now = jiffies;
3411         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3412
3413         del_timer(&addr_chk_timer);
3414
3415         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3416 restart:
3417                 hlist_for_each_entry_rcu_bh(ifp,
3418                                          &inet6_addr_lst[i], addr_lst) {
3419                         unsigned long age;
3420
3421                         if (ifp->flags & IFA_F_PERMANENT)
3422                                 continue;
3423
3424                         spin_lock(&ifp->lock);
3425                         /* We try to batch several events at once. */
3426                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3427
3428                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3429                             age >= ifp->valid_lft) {
3430                                 spin_unlock(&ifp->lock);
3431                                 in6_ifa_hold(ifp);
3432                                 ipv6_del_addr(ifp);
3433                                 goto restart;
3434                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3435                                 spin_unlock(&ifp->lock);
3436                                 continue;
3437                         } else if (age >= ifp->prefered_lft) {
3438                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3439                                 int deprecate = 0;
3440
3441                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3442                                         deprecate = 1;
3443                                         ifp->flags |= IFA_F_DEPRECATED;
3444                                 }
3445
3446                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3447                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3448
3449                                 spin_unlock(&ifp->lock);
3450
3451                                 if (deprecate) {
3452                                         in6_ifa_hold(ifp);
3453
3454                                         ipv6_ifa_notify(0, ifp);
3455                                         in6_ifa_put(ifp);
3456                                         goto restart;
3457                                 }
3458 #ifdef CONFIG_IPV6_PRIVACY
3459                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3460                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3461                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3462                                         ifp->idev->cnf.dad_transmits *
3463                                         ifp->idev->nd_parms->retrans_time / HZ;
3464
3465                                 if (age >= ifp->prefered_lft - regen_advance) {
3466                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3467                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3468                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3469                                         if (!ifp->regen_count && ifpub) {
3470                                                 ifp->regen_count++;
3471                                                 in6_ifa_hold(ifp);
3472                                                 in6_ifa_hold(ifpub);
3473                                                 spin_unlock(&ifp->lock);
3474
3475                                                 spin_lock(&ifpub->lock);
3476                                                 ifpub->regen_count = 0;
3477                                                 spin_unlock(&ifpub->lock);
3478                                                 ipv6_create_tempaddr(ifpub, ifp);
3479                                                 in6_ifa_put(ifpub);
3480                                                 in6_ifa_put(ifp);
3481                                                 goto restart;
3482                                         }
3483                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3484                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3485                                 spin_unlock(&ifp->lock);
3486 #endif
3487                         } else {
3488                                 /* ifp->prefered_lft <= ifp->valid_lft */
3489                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3490                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3491                                 spin_unlock(&ifp->lock);
3492                         }
3493                 }
3494         }
3495
3496         next_sec = round_jiffies_up(next);
3497         next_sched = next;
3498
3499         /* If rounded timeout is accurate enough, accept it. */
3500         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3501                 next_sched = next_sec;
3502
3503         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3504         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3505                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3506
3507         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3508               now, next, next_sec, next_sched));
3509
3510         addr_chk_timer.expires = next_sched;
3511         add_timer(&addr_chk_timer);
3512         spin_unlock(&addrconf_verify_lock);
3513         rcu_read_unlock_bh();
3514 }
3515
3516 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3517 {
3518         struct in6_addr *pfx = NULL;
3519
3520         if (addr)
3521                 pfx = nla_data(addr);
3522
3523         if (local) {
3524                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3525                         pfx = NULL;
3526                 else
3527                         pfx = nla_data(local);
3528         }
3529
3530         return pfx;
3531 }
3532
3533 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3534         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3535         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3536         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3537 };
3538
3539 static int
3540 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3541 {
3542         struct net *net = sock_net(skb->sk);
3543         struct ifaddrmsg *ifm;
3544         struct nlattr *tb[IFA_MAX+1];
3545         struct in6_addr *pfx;
3546         int err;
3547
3548         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3549         if (err < 0)
3550                 return err;
3551
3552         ifm = nlmsg_data(nlh);
3553         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3554         if (pfx == NULL)
3555                 return -EINVAL;
3556
3557         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3558 }
3559
3560 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3561                              u32 prefered_lft, u32 valid_lft)
3562 {
3563         u32 flags;
3564         clock_t expires;
3565         unsigned long timeout;
3566
3567         if (!valid_lft || (prefered_lft > valid_lft))
3568                 return -EINVAL;
3569
3570         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3571         if (addrconf_finite_timeout(timeout)) {
3572                 expires = jiffies_to_clock_t(timeout * HZ);
3573                 valid_lft = timeout;
3574                 flags = RTF_EXPIRES;
3575         } else {
3576                 expires = 0;
3577                 flags = 0;
3578                 ifa_flags |= IFA_F_PERMANENT;
3579         }
3580
3581         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3582         if (addrconf_finite_timeout(timeout)) {
3583                 if (timeout == 0)
3584                         ifa_flags |= IFA_F_DEPRECATED;
3585                 prefered_lft = timeout;
3586         }
3587
3588         spin_lock_bh(&ifp->lock);
3589         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3590         ifp->tstamp = jiffies;
3591         ifp->valid_lft = valid_lft;
3592         ifp->prefered_lft = prefered_lft;
3593
3594         spin_unlock_bh(&ifp->lock);
3595         if (!(ifp->flags&IFA_F_TENTATIVE))
3596                 ipv6_ifa_notify(0, ifp);
3597
3598         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3599                               expires, flags);
3600         addrconf_verify(0);
3601
3602         return 0;
3603 }
3604
3605 static int
3606 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3607 {
3608         struct net *net = sock_net(skb->sk);
3609         struct ifaddrmsg *ifm;
3610         struct nlattr *tb[IFA_MAX+1];
3611         struct in6_addr *pfx;
3612         struct inet6_ifaddr *ifa;
3613         struct net_device *dev;
3614         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3615         u8 ifa_flags;
3616         int err;
3617
3618         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3619         if (err < 0)
3620                 return err;
3621
3622         ifm = nlmsg_data(nlh);
3623         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3624         if (pfx == NULL)
3625                 return -EINVAL;
3626
3627         if (tb[IFA_CACHEINFO]) {
3628                 struct ifa_cacheinfo *ci;
3629
3630                 ci = nla_data(tb[IFA_CACHEINFO]);
3631                 valid_lft = ci->ifa_valid;
3632                 preferred_lft = ci->ifa_prefered;
3633         } else {
3634                 preferred_lft = INFINITY_LIFE_TIME;
3635                 valid_lft = INFINITY_LIFE_TIME;
3636         }
3637
3638         dev =  __dev_get_by_index(net, ifm->ifa_index);
3639         if (dev == NULL)
3640                 return -ENODEV;
3641
3642         /* We ignore other flags so far. */
3643         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3644
3645         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3646         if (ifa == NULL) {
3647                 /*
3648                  * It would be best to check for !NLM_F_CREATE here but
3649                  * userspace alreay relies on not having to provide this.
3650                  */
3651                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3652                                       ifm->ifa_prefixlen, ifa_flags,
3653                                       preferred_lft, valid_lft);
3654         }
3655
3656         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3657             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3658                 err = -EEXIST;
3659         else
3660                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3661
3662         in6_ifa_put(ifa);
3663
3664         return err;
3665 }
3666
3667 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3668                           u8 scope, int ifindex)
3669 {
3670         struct ifaddrmsg *ifm;
3671
3672         ifm = nlmsg_data(nlh);
3673         ifm->ifa_family = AF_INET6;
3674         ifm->ifa_prefixlen = prefixlen;
3675         ifm->ifa_flags = flags;
3676         ifm->ifa_scope = scope;
3677         ifm->ifa_index = ifindex;
3678 }
3679
3680 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3681                          unsigned long tstamp, u32 preferred, u32 valid)
3682 {
3683         struct ifa_cacheinfo ci;
3684
3685         ci.cstamp = cstamp_delta(cstamp);
3686         ci.tstamp = cstamp_delta(tstamp);
3687         ci.ifa_prefered = preferred;
3688         ci.ifa_valid = valid;
3689
3690         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3691 }
3692
3693 static inline int rt_scope(int ifa_scope)
3694 {
3695         if (ifa_scope & IFA_HOST)
3696                 return RT_SCOPE_HOST;
3697         else if (ifa_scope & IFA_LINK)
3698                 return RT_SCOPE_LINK;
3699         else if (ifa_scope & IFA_SITE)
3700                 return RT_SCOPE_SITE;
3701         else
3702                 return RT_SCOPE_UNIVERSE;
3703 }
3704
3705 static inline int inet6_ifaddr_msgsize(void)
3706 {
3707         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3708                + nla_total_size(16) /* IFA_ADDRESS */
3709                + nla_total_size(sizeof(struct ifa_cacheinfo));
3710 }
3711
3712 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3713                              u32 portid, u32 seq, int event, unsigned int flags)
3714 {
3715         struct nlmsghdr  *nlh;
3716         u32 preferred, valid;
3717
3718         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3719         if (nlh == NULL)
3720                 return -EMSGSIZE;
3721
3722         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3723                       ifa->idev->dev->ifindex);
3724
3725         if (!(ifa->flags&IFA_F_PERMANENT)) {
3726                 preferred = ifa->prefered_lft;
3727                 valid = ifa->valid_lft;
3728                 if (preferred != INFINITY_LIFE_TIME) {
3729                         long tval = (jiffies - ifa->tstamp)/HZ;
3730                         if (preferred > tval)
3731                                 preferred -= tval;
3732                         else
3733                                 preferred = 0;
3734                         if (valid != INFINITY_LIFE_TIME) {
3735                                 if (valid > tval)
3736                                         valid -= tval;
3737                                 else
3738                                         valid = 0;
3739                         }
3740                 }
3741         } else {
3742                 preferred = INFINITY_LIFE_TIME;
3743                 valid = INFINITY_LIFE_TIME;
3744         }
3745
3746         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3747             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3748                 nlmsg_cancel(skb, nlh);
3749                 return -EMSGSIZE;
3750         }
3751
3752         return nlmsg_end(skb, nlh);
3753 }
3754
3755 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3756                                 u32 portid, u32 seq, int event, u16 flags)
3757 {
3758         struct nlmsghdr  *nlh;
3759         u8 scope = RT_SCOPE_UNIVERSE;
3760         int ifindex = ifmca->idev->dev->ifindex;
3761
3762         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3763                 scope = RT_SCOPE_SITE;
3764
3765         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3766         if (nlh == NULL)
3767                 return -EMSGSIZE;
3768
3769         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3770         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3771             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3772                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3773                 nlmsg_cancel(skb, nlh);
3774                 return -EMSGSIZE;
3775         }
3776
3777         return nlmsg_end(skb, nlh);
3778 }
3779
3780 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3781                                 u32 portid, u32 seq, int event, unsigned int flags)
3782 {
3783         struct nlmsghdr  *nlh;
3784         u8 scope = RT_SCOPE_UNIVERSE;
3785         int ifindex = ifaca->aca_idev->dev->ifindex;
3786
3787         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3788                 scope = RT_SCOPE_SITE;
3789
3790         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3791         if (nlh == NULL)
3792                 return -EMSGSIZE;
3793
3794         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3795         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3796             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3797                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3798                 nlmsg_cancel(skb, nlh);
3799                 return -EMSGSIZE;
3800         }
3801
3802         return nlmsg_end(skb, nlh);
3803 }
3804
3805 enum addr_type_t {
3806         UNICAST_ADDR,
3807         MULTICAST_ADDR,
3808         ANYCAST_ADDR,
3809 };
3810
3811 /* called with rcu_read_lock() */
3812 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3813                           struct netlink_callback *cb, enum addr_type_t type,
3814                           int s_ip_idx, int *p_ip_idx)
3815 {
3816         struct ifmcaddr6 *ifmca;
3817         struct ifacaddr6 *ifaca;
3818         int err = 1;
3819         int ip_idx = *p_ip_idx;
3820
3821         read_lock_bh(&idev->lock);
3822         switch (type) {
3823         case UNICAST_ADDR: {
3824                 struct inet6_ifaddr *ifa;
3825
3826                 /* unicast address incl. temp addr */
3827                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3828                         if (++ip_idx < s_ip_idx)
3829                                 continue;
3830                         err = inet6_fill_ifaddr(skb, ifa,
3831                                                 NETLINK_CB(cb->skb).portid,
3832                                                 cb->nlh->nlmsg_seq,
3833                                                 RTM_NEWADDR,
3834                                                 NLM_F_MULTI);
3835                         if (err <= 0)
3836                                 break;
3837                 }
3838                 break;
3839         }
3840         case MULTICAST_ADDR:
3841                 /* multicast address */
3842                 for (ifmca = idev->mc_list; ifmca;
3843                      ifmca = ifmca->next, ip_idx++) {
3844                         if (ip_idx < s_ip_idx)
3845                                 continue;
3846                         err = inet6_fill_ifmcaddr(skb, ifmca,
3847                                                   NETLINK_CB(cb->skb).portid,
3848                                                   cb->nlh->nlmsg_seq,
3849                                                   RTM_GETMULTICAST,
3850                                                   NLM_F_MULTI);
3851                         if (err <= 0)
3852                                 break;
3853                 }
3854                 break;
3855         case ANYCAST_ADDR:
3856                 /* anycast address */
3857                 for (ifaca = idev->ac_list; ifaca;
3858                      ifaca = ifaca->aca_next, ip_idx++) {
3859                         if (ip_idx < s_ip_idx)
3860                                 continue;
3861                         err = inet6_fill_ifacaddr(skb, ifaca,
3862                                                   NETLINK_CB(cb->skb).portid,
3863                                                   cb->nlh->nlmsg_seq,
3864                                                   RTM_GETANYCAST,
3865                                                   NLM_F_MULTI);
3866                         if (err <= 0)
3867                                 break;
3868                 }
3869                 break;
3870         default:
3871                 break;
3872         }
3873         read_unlock_bh(&idev->lock);
3874         *p_ip_idx = ip_idx;
3875         return err;
3876 }
3877
3878 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3879                            enum addr_type_t type)
3880 {
3881         struct net *net = sock_net(skb->sk);
3882         int h, s_h;
3883         int idx, ip_idx;
3884         int s_idx, s_ip_idx;
3885         struct net_device *dev;
3886         struct inet6_dev *idev;
3887         struct hlist_head *head;
3888
3889         s_h = cb->args[0];
3890         s_idx = idx = cb->args[1];
3891         s_ip_idx = ip_idx = cb->args[2];
3892
3893         rcu_read_lock();
3894         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3895                 idx = 0;
3896                 head = &net->dev_index_head[h];
3897                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3898                         if (idx < s_idx)
3899                                 goto cont;
3900                         if (h > s_h || idx > s_idx)
3901                                 s_ip_idx = 0;
3902                         ip_idx = 0;
3903                         idev = __in6_dev_get(dev);
3904                         if (!idev)
3905                                 goto cont;
3906
3907                         if (in6_dump_addrs(idev, skb, cb, type,
3908                                            s_ip_idx, &ip_idx) <= 0)
3909                                 goto done;
3910 cont:
3911                         idx++;
3912                 }
3913         }
3914 done:
3915         rcu_read_unlock();
3916         cb->args[0] = h;
3917         cb->args[1] = idx;
3918         cb->args[2] = ip_idx;
3919
3920         return skb->len;
3921 }
3922
3923 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3924 {
3925         enum addr_type_t type = UNICAST_ADDR;
3926
3927         return inet6_dump_addr(skb, cb, type);
3928 }
3929
3930 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3931 {
3932         enum addr_type_t type = MULTICAST_ADDR;
3933
3934         return inet6_dump_addr(skb, cb, type);
3935 }
3936
3937
3938 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3939 {
3940         enum addr_type_t type = ANYCAST_ADDR;
3941
3942         return inet6_dump_addr(skb, cb, type);
3943 }
3944
3945 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3946                              void *arg)
3947 {
3948         struct net *net = sock_net(in_skb->sk);
3949         struct ifaddrmsg *ifm;
3950         struct nlattr *tb[IFA_MAX+1];
3951         struct in6_addr *addr = NULL;
3952         struct net_device *dev = NULL;
3953         struct inet6_ifaddr *ifa;
3954         struct sk_buff *skb;
3955         int err;
3956
3957         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3958         if (err < 0)
3959                 goto errout;
3960
3961         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3962         if (addr == NULL) {
3963                 err = -EINVAL;
3964                 goto errout;
3965         }
3966
3967         ifm = nlmsg_data(nlh);
3968         if (ifm->ifa_index)
3969                 dev = __dev_get_by_index(net, ifm->ifa_index);
3970
3971         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3972         if (!ifa) {
3973                 err = -EADDRNOTAVAIL;
3974                 goto errout;
3975         }
3976
3977         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3978         if (!skb) {
3979                 err = -ENOBUFS;
3980                 goto errout_ifa;
3981         }
3982
3983         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3984                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3985         if (err < 0) {
3986                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3987                 WARN_ON(err == -EMSGSIZE);
3988                 kfree_skb(skb);
3989                 goto errout_ifa;
3990         }
3991         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3992 errout_ifa:
3993         in6_ifa_put(ifa);
3994 errout:
3995         return err;
3996 }
3997
3998 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3999 {
4000         struct sk_buff *skb;
4001         struct net *net = dev_net(ifa->idev->dev);
4002         int err = -ENOBUFS;
4003
4004         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4005         if (skb == NULL)
4006                 goto errout;
4007
4008         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4009         if (err < 0) {
4010                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4011                 WARN_ON(err == -EMSGSIZE);
4012                 kfree_skb(skb);
4013                 goto errout;
4014         }
4015         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4016         return;
4017 errout:
4018         if (err < 0)
4019                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4020 }
4021
4022 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4023                                 __s32 *array, int bytes)
4024 {
4025         BUG_ON(bytes < (DEVCONF_MAX * 4));
4026
4027         memset(array, 0, bytes);
4028         array[DEVCONF_FORWARDING] = cnf->forwarding;
4029         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4030         array[DEVCONF_MTU6] = cnf->mtu6;
4031         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4032         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4033         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4034         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4035         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4036         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4037                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4038         array[DEVCONF_RTR_SOLICIT_DELAY] =
4039                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4040         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4041 #ifdef CONFIG_IPV6_PRIVACY
4042         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4043         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4044         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4045         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4046         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4047 #endif
4048         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4049         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4050         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4051 #ifdef CONFIG_IPV6_ROUTER_PREF
4052         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4053         array[DEVCONF_RTR_PROBE_INTERVAL] =
4054                 jiffies_to_msecs(cnf->rtr_probe_interval);
4055 #ifdef CONFIG_IPV6_ROUTE_INFO
4056         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4057 #endif
4058 #endif
4059         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4060         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4061 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4062         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4063 #endif
4064 #ifdef CONFIG_IPV6_MROUTE
4065         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4066 #endif
4067         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4068         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4069         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4070         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4071 }
4072
4073 static inline size_t inet6_ifla6_size(void)
4074 {
4075         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4076              + nla_total_size(sizeof(struct ifla_cacheinfo))
4077              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4078              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4079              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4080 }
4081
4082 static inline size_t inet6_if_nlmsg_size(void)
4083 {
4084         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4085                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4086                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4087                + nla_total_size(4) /* IFLA_MTU */
4088                + nla_total_size(4) /* IFLA_LINK */
4089                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4090 }
4091
4092 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4093                                       int items, int bytes)
4094 {
4095         int i;
4096         int pad = bytes - sizeof(u64) * items;
4097         BUG_ON(pad < 0);
4098
4099         /* Use put_unaligned() because stats may not be aligned for u64. */
4100         put_unaligned(items, &stats[0]);
4101         for (i = 1; i < items; i++)
4102                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4103
4104         memset(&stats[items], 0, pad);
4105 }
4106
4107 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4108                                       int items, int bytes, size_t syncpoff)
4109 {
4110         int i;
4111         int pad = bytes - sizeof(u64) * items;
4112         BUG_ON(pad < 0);
4113
4114         /* Use put_unaligned() because stats may not be aligned for u64. */
4115         put_unaligned(items, &stats[0]);
4116         for (i = 1; i < items; i++)
4117                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4118
4119         memset(&stats[items], 0, pad);
4120 }
4121
4122 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4123                              int bytes)
4124 {
4125         switch (attrtype) {
4126         case IFLA_INET6_STATS:
4127                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4128                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4129                 break;
4130         case IFLA_INET6_ICMP6STATS:
4131                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4132                 break;
4133         }
4134 }
4135
4136 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4137 {
4138         struct nlattr *nla;
4139         struct ifla_cacheinfo ci;
4140
4141         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4142                 goto nla_put_failure;
4143         ci.max_reasm_len = IPV6_MAXPLEN;
4144         ci.tstamp = cstamp_delta(idev->tstamp);
4145         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4146         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4147         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4148                 goto nla_put_failure;
4149         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4150         if (nla == NULL)
4151                 goto nla_put_failure;
4152         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4153
4154         /* XXX - MC not implemented */
4155
4156         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4157         if (nla == NULL)
4158                 goto nla_put_failure;
4159         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4160
4161         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4162         if (nla == NULL)
4163                 goto nla_put_failure;
4164         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4165
4166         return 0;
4167
4168 nla_put_failure:
4169         return -EMSGSIZE;
4170 }
4171
4172 static size_t inet6_get_link_af_size(const struct net_device *dev)
4173 {
4174         if (!__in6_dev_get(dev))
4175                 return 0;
4176
4177         return inet6_ifla6_size();
4178 }
4179
4180 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4181 {
4182         struct inet6_dev *idev = __in6_dev_get(dev);
4183
4184         if (!idev)
4185                 return -ENODATA;
4186
4187         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4188                 return -EMSGSIZE;
4189
4190         return 0;
4191 }
4192
4193 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4194                              u32 portid, u32 seq, int event, unsigned int flags)
4195 {
4196         struct net_device *dev = idev->dev;
4197         struct ifinfomsg *hdr;
4198         struct nlmsghdr *nlh;
4199         void *protoinfo;
4200
4201         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4202         if (nlh == NULL)
4203                 return -EMSGSIZE;
4204
4205         hdr = nlmsg_data(nlh);
4206         hdr->ifi_family = AF_INET6;
4207         hdr->__ifi_pad = 0;
4208         hdr->ifi_type = dev->type;
4209         hdr->ifi_index = dev->ifindex;
4210         hdr->ifi_flags = dev_get_flags(dev);
4211         hdr->ifi_change = 0;
4212
4213         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4214             (dev->addr_len &&
4215              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4216             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4217             (dev->ifindex != dev->iflink &&
4218              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4219                 goto nla_put_failure;
4220         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4221         if (protoinfo == NULL)
4222                 goto nla_put_failure;
4223
4224         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4225                 goto nla_put_failure;
4226
4227         nla_nest_end(skb, protoinfo);
4228         return nlmsg_end(skb, nlh);
4229
4230 nla_put_failure:
4231         nlmsg_cancel(skb, nlh);
4232         return -EMSGSIZE;
4233 }
4234
4235 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4236 {
4237         struct net *net = sock_net(skb->sk);
4238         int h, s_h;
4239         int idx = 0, s_idx;
4240         struct net_device *dev;
4241         struct inet6_dev *idev;
4242         struct hlist_head *head;
4243
4244         s_h = cb->args[0];
4245         s_idx = cb->args[1];
4246
4247         rcu_read_lock();
4248         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4249                 idx = 0;
4250                 head = &net->dev_index_head[h];
4251                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4252                         if (idx < s_idx)
4253                                 goto cont;
4254                         idev = __in6_dev_get(dev);
4255                         if (!idev)
4256                                 goto cont;
4257                         if (inet6_fill_ifinfo(skb, idev,
4258                                               NETLINK_CB(cb->skb).portid,
4259                                               cb->nlh->nlmsg_seq,
4260                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4261                                 goto out;
4262 cont:
4263                         idx++;
4264                 }
4265         }
4266 out:
4267         rcu_read_unlock();
4268         cb->args[1] = idx;
4269         cb->args[0] = h;
4270
4271         return skb->len;
4272 }
4273
4274 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4275 {
4276         struct sk_buff *skb;
4277         struct net *net = dev_net(idev->dev);
4278         int err = -ENOBUFS;
4279
4280         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4281         if (skb == NULL)
4282                 goto errout;
4283
4284         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4285         if (err < 0) {
4286                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4287                 WARN_ON(err == -EMSGSIZE);
4288                 kfree_skb(skb);
4289                 goto errout;
4290         }
4291         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4292         return;
4293 errout:
4294         if (err < 0)
4295                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4296 }
4297
4298 static inline size_t inet6_prefix_nlmsg_size(void)
4299 {
4300         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4301                + nla_total_size(sizeof(struct in6_addr))
4302                + nla_total_size(sizeof(struct prefix_cacheinfo));
4303 }
4304
4305 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4306                              struct prefix_info *pinfo, u32 portid, u32 seq,
4307                              int event, unsigned int flags)
4308 {
4309         struct prefixmsg *pmsg;
4310         struct nlmsghdr *nlh;
4311         struct prefix_cacheinfo ci;
4312
4313         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4314         if (nlh == NULL)
4315                 return -EMSGSIZE;
4316
4317         pmsg = nlmsg_data(nlh);
4318         pmsg->prefix_family = AF_INET6;
4319         pmsg->prefix_pad1 = 0;
4320         pmsg->prefix_pad2 = 0;
4321         pmsg->prefix_ifindex = idev->dev->ifindex;
4322         pmsg->prefix_len = pinfo->prefix_len;
4323         pmsg->prefix_type = pinfo->type;
4324         pmsg->prefix_pad3 = 0;
4325         pmsg->prefix_flags = 0;
4326         if (pinfo->onlink)
4327                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4328         if (pinfo->autoconf)
4329                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4330
4331         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4332                 goto nla_put_failure;
4333         ci.preferred_time = ntohl(pinfo->prefered);
4334         ci.valid_time = ntohl(pinfo->valid);
4335         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4336                 goto nla_put_failure;
4337         return nlmsg_end(skb, nlh);
4338
4339 nla_put_failure:
4340         nlmsg_cancel(skb, nlh);
4341         return -EMSGSIZE;
4342 }
4343
4344 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4345                          struct prefix_info *pinfo)
4346 {
4347         struct sk_buff *skb;
4348         struct net *net = dev_net(idev->dev);
4349         int err = -ENOBUFS;
4350
4351         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4352         if (skb == NULL)
4353                 goto errout;
4354
4355         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4356         if (err < 0) {
4357                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4358                 WARN_ON(err == -EMSGSIZE);
4359                 kfree_skb(skb);
4360                 goto errout;
4361         }
4362         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4363         return;
4364 errout:
4365         if (err < 0)
4366                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4367 }
4368
4369 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4370 {
4371         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4372
4373         switch (event) {
4374         case RTM_NEWADDR:
4375                 /*
4376                  * If the address was optimistic
4377                  * we inserted the route at the start of
4378                  * our DAD process, so we don't need
4379                  * to do it again
4380                  */
4381                 if (!(ifp->rt->rt6i_node))
4382                         ip6_ins_rt(ifp->rt);
4383                 if (ifp->idev->cnf.forwarding)
4384                         addrconf_join_anycast(ifp);
4385                 break;
4386         case RTM_DELADDR:
4387                 if (ifp->idev->cnf.forwarding)
4388                         addrconf_leave_anycast(ifp);
4389                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4390                 dst_hold(&ifp->rt->dst);
4391
4392                 if (ip6_del_rt(ifp->rt))
4393                         dst_free(&ifp->rt->dst);
4394                 break;
4395         }
4396 }
4397
4398 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4399 {
4400         rcu_read_lock_bh();
4401         if (likely(ifp->idev->dead == 0))
4402                 __ipv6_ifa_notify(event, ifp);
4403         rcu_read_unlock_bh();
4404 }
4405
4406 #ifdef CONFIG_SYSCTL
4407
4408 static
4409 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4410                            void __user *buffer, size_t *lenp, loff_t *ppos)
4411 {
4412         int *valp = ctl->data;
4413         int val = *valp;
4414         loff_t pos = *ppos;
4415         ctl_table lctl;
4416         int ret;
4417
4418         /*
4419          * ctl->data points to idev->cnf.forwarding, we should
4420          * not modify it until we get the rtnl lock.
4421          */
4422         lctl = *ctl;
4423         lctl.data = &val;
4424
4425         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4426
4427         if (write)
4428                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4429         if (ret)
4430                 *ppos = pos;
4431         return ret;
4432 }
4433
4434 static void dev_disable_change(struct inet6_dev *idev)
4435 {
4436         if (!idev || !idev->dev)
4437                 return;
4438
4439         if (idev->cnf.disable_ipv6)
4440                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4441         else
4442                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4443 }
4444
4445 static void addrconf_disable_change(struct net *net, __s32 newf)
4446 {
4447         struct net_device *dev;
4448         struct inet6_dev *idev;
4449
4450         rcu_read_lock();
4451         for_each_netdev_rcu(net, dev) {
4452                 idev = __in6_dev_get(dev);
4453                 if (idev) {
4454                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4455                         idev->cnf.disable_ipv6 = newf;
4456                         if (changed)
4457                                 dev_disable_change(idev);
4458                 }
4459         }
4460         rcu_read_unlock();
4461 }
4462
4463 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4464 {
4465         struct net *net;
4466         int old;
4467
4468         if (!rtnl_trylock())
4469                 return restart_syscall();
4470
4471         net = (struct net *)table->extra2;
4472         old = *p;
4473         *p = newf;
4474
4475         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4476                 rtnl_unlock();
4477                 return 0;
4478         }
4479
4480         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4481                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4482                 addrconf_disable_change(net, newf);
4483         } else if ((!newf) ^ (!old))
4484                 dev_disable_change((struct inet6_dev *)table->extra1);
4485
4486         rtnl_unlock();
4487         return 0;
4488 }
4489
4490 static
4491 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4492                             void __user *buffer, size_t *lenp, loff_t *ppos)
4493 {
4494         int *valp = ctl->data;
4495         int val = *valp;
4496         loff_t pos = *ppos;
4497         ctl_table lctl;
4498         int ret;
4499
4500         /*
4501          * ctl->data points to idev->cnf.disable_ipv6, we should
4502          * not modify it until we get the rtnl lock.
4503          */
4504         lctl = *ctl;
4505         lctl.data = &val;
4506
4507         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4508
4509         if (write)
4510                 ret = addrconf_disable_ipv6(ctl, valp, val);
4511         if (ret)
4512                 *ppos = pos;
4513         return ret;
4514 }
4515
4516 static struct addrconf_sysctl_table
4517 {
4518         struct ctl_table_header *sysctl_header;
4519         ctl_table addrconf_vars[DEVCONF_MAX+1];
4520 } addrconf_sysctl __read_mostly = {
4521         .sysctl_header = NULL,
4522         .addrconf_vars = {
4523                 {
4524                         .procname       = "forwarding",
4525                         .data           = &ipv6_devconf.forwarding,
4526                         .maxlen         = sizeof(int),
4527                         .mode           = 0644,
4528                         .proc_handler   = addrconf_sysctl_forward,
4529                 },
4530                 {
4531                         .procname       = "hop_limit",
4532                         .data           = &ipv6_devconf.hop_limit,
4533                         .maxlen         = sizeof(int),
4534                         .mode           = 0644,
4535                         .proc_handler   = proc_dointvec,
4536                 },
4537                 {
4538                         .procname       = "mtu",
4539                         .data           = &ipv6_devconf.mtu6,
4540                         .maxlen         = sizeof(int),
4541                         .mode           = 0644,
4542                         .proc_handler   = proc_dointvec,
4543                 },
4544                 {
4545                         .procname       = "accept_ra",
4546                         .data           = &ipv6_devconf.accept_ra,
4547                         .maxlen         = sizeof(int),
4548                         .mode           = 0644,
4549                         .proc_handler   = proc_dointvec,
4550                 },
4551                 {
4552                         .procname       = "accept_redirects",
4553                         .data           = &ipv6_devconf.accept_redirects,
4554                         .maxlen         = sizeof(int),
4555                         .mode           = 0644,
4556                         .proc_handler   = proc_dointvec,
4557                 },
4558                 {
4559                         .procname       = "autoconf",
4560                         .data           = &ipv6_devconf.autoconf,
4561                         .maxlen         = sizeof(int),
4562                         .mode           = 0644,
4563                         .proc_handler   = proc_dointvec,
4564                 },
4565                 {
4566                         .procname       = "dad_transmits",
4567                         .data           = &ipv6_devconf.dad_transmits,
4568                         .maxlen         = sizeof(int),
4569                         .mode           = 0644,
4570                         .proc_handler   = proc_dointvec,
4571                 },
4572                 {
4573                         .procname       = "router_solicitations",
4574                         .data           = &ipv6_devconf.rtr_solicits,
4575                         .maxlen         = sizeof(int),
4576                         .mode           = 0644,
4577                         .proc_handler   = proc_dointvec,
4578                 },
4579                 {
4580                         .procname       = "router_solicitation_interval",
4581                         .data           = &ipv6_devconf.rtr_solicit_interval,
4582                         .maxlen         = sizeof(int),
4583                         .mode           = 0644,
4584                         .proc_handler   = proc_dointvec_jiffies,
4585                 },
4586                 {
4587                         .procname       = "router_solicitation_delay",
4588                         .data           = &ipv6_devconf.rtr_solicit_delay,
4589                         .maxlen         = sizeof(int),
4590                         .mode           = 0644,
4591                         .proc_handler   = proc_dointvec_jiffies,
4592                 },
4593                 {
4594                         .procname       = "force_mld_version",
4595                         .data           = &ipv6_devconf.force_mld_version,
4596                         .maxlen         = sizeof(int),
4597                         .mode           = 0644,
4598                         .proc_handler   = proc_dointvec,
4599                 },
4600 #ifdef CONFIG_IPV6_PRIVACY
4601                 {
4602                         .procname       = "use_tempaddr",
4603                         .data           = &ipv6_devconf.use_tempaddr,
4604                         .maxlen         = sizeof(int),
4605                         .mode           = 0644,
4606                         .proc_handler   = proc_dointvec,
4607                 },
4608                 {
4609                         .procname       = "temp_valid_lft",
4610                         .data           = &ipv6_devconf.temp_valid_lft,
4611                         .maxlen         = sizeof(int),
4612                         .mode           = 0644,
4613                         .proc_handler   = proc_dointvec,
4614                 },
4615                 {
4616                         .procname       = "temp_prefered_lft",
4617                         .data           = &ipv6_devconf.temp_prefered_lft,
4618                         .maxlen         = sizeof(int),
4619                         .mode           = 0644,
4620                         .proc_handler   = proc_dointvec,
4621                 },
4622                 {
4623                         .procname       = "regen_max_retry",
4624                         .data           = &ipv6_devconf.regen_max_retry,
4625                         .maxlen         = sizeof(int),
4626                         .mode           = 0644,
4627                         .proc_handler   = proc_dointvec,
4628                 },
4629                 {
4630                         .procname       = "max_desync_factor",
4631                         .data           = &ipv6_devconf.max_desync_factor,
4632                         .maxlen         = sizeof(int),
4633                         .mode           = 0644,
4634                         .proc_handler   = proc_dointvec,
4635                 },
4636 #endif
4637                 {
4638                         .procname       = "max_addresses",
4639                         .data           = &ipv6_devconf.max_addresses,
4640                         .maxlen         = sizeof(int),
4641                         .mode           = 0644,
4642                         .proc_handler   = proc_dointvec,
4643                 },
4644                 {
4645                         .procname       = "accept_ra_defrtr",
4646                         .data           = &ipv6_devconf.accept_ra_defrtr,
4647                         .maxlen         = sizeof(int),
4648                         .mode           = 0644,
4649                         .proc_handler   = proc_dointvec,
4650                 },
4651                 {
4652                         .procname       = "accept_ra_pinfo",
4653                         .data           = &ipv6_devconf.accept_ra_pinfo,
4654                         .maxlen         = sizeof(int),
4655                         .mode           = 0644,
4656                         .proc_handler   = proc_dointvec,
4657                 },
4658 #ifdef CONFIG_IPV6_ROUTER_PREF
4659                 {
4660                         .procname       = "accept_ra_rtr_pref",
4661                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4662                         .maxlen         = sizeof(int),
4663                         .mode           = 0644,
4664                         .proc_handler   = proc_dointvec,
4665                 },
4666                 {
4667                         .procname       = "router_probe_interval",
4668                         .data           = &ipv6_devconf.rtr_probe_interval,
4669                         .maxlen         = sizeof(int),
4670                         .mode           = 0644,
4671                         .proc_handler   = proc_dointvec_jiffies,
4672                 },
4673 #ifdef CONFIG_IPV6_ROUTE_INFO
4674                 {
4675                         .procname       = "accept_ra_rt_info_max_plen",
4676                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4677                         .maxlen         = sizeof(int),
4678                         .mode           = 0644,
4679                         .proc_handler   = proc_dointvec,
4680                 },
4681 #endif
4682 #endif
4683                 {
4684                         .procname       = "proxy_ndp",
4685                         .data           = &ipv6_devconf.proxy_ndp,
4686                         .maxlen         = sizeof(int),
4687                         .mode           = 0644,
4688                         .proc_handler   = proc_dointvec,
4689                 },
4690                 {
4691                         .procname       = "accept_source_route",
4692                         .data           = &ipv6_devconf.accept_source_route,
4693                         .maxlen         = sizeof(int),
4694                         .mode           = 0644,
4695                         .proc_handler   = proc_dointvec,
4696                 },
4697 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4698                 {
4699                         .procname       = "optimistic_dad",
4700                         .data           = &ipv6_devconf.optimistic_dad,
4701                         .maxlen         = sizeof(int),
4702                         .mode           = 0644,
4703                         .proc_handler   = proc_dointvec,
4704
4705                 },
4706 #endif
4707 #ifdef CONFIG_IPV6_MROUTE
4708                 {
4709                         .procname       = "mc_forwarding",
4710                         .data           = &ipv6_devconf.mc_forwarding,
4711                         .maxlen         = sizeof(int),
4712                         .mode           = 0444,
4713                         .proc_handler   = proc_dointvec,
4714                 },
4715 #endif
4716                 {
4717                         .procname       = "disable_ipv6",
4718                         .data           = &ipv6_devconf.disable_ipv6,
4719                         .maxlen         = sizeof(int),
4720                         .mode           = 0644,
4721                         .proc_handler   = addrconf_sysctl_disable,
4722                 },
4723                 {
4724                         .procname       = "accept_dad",
4725                         .data           = &ipv6_devconf.accept_dad,
4726                         .maxlen         = sizeof(int),
4727                         .mode           = 0644,
4728                         .proc_handler   = proc_dointvec,
4729                 },
4730                 {
4731                         .procname       = "force_tllao",
4732                         .data           = &ipv6_devconf.force_tllao,
4733                         .maxlen         = sizeof(int),
4734                         .mode           = 0644,
4735                         .proc_handler   = proc_dointvec
4736                 },
4737                 {
4738                         .procname       = "ndisc_notify",
4739                         .data           = &ipv6_devconf.ndisc_notify,
4740                         .maxlen         = sizeof(int),
4741                         .mode           = 0644,
4742                         .proc_handler   = proc_dointvec
4743                 },
4744                 {
4745                         /* sentinel */
4746                 }
4747         },
4748 };
4749
4750 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4751                 struct inet6_dev *idev, struct ipv6_devconf *p)
4752 {
4753         int i;
4754         struct addrconf_sysctl_table *t;
4755         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4756
4757         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4758         if (t == NULL)
4759                 goto out;
4760
4761         for (i = 0; t->addrconf_vars[i].data; i++) {
4762                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4763                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4764                 t->addrconf_vars[i].extra2 = net;
4765         }
4766
4767         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4768
4769         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4770         if (t->sysctl_header == NULL)
4771                 goto free;
4772
4773         p->sysctl = t;
4774         return 0;
4775
4776 free:
4777         kfree(t);
4778 out:
4779         return -ENOBUFS;
4780 }
4781
4782 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4783 {
4784         struct addrconf_sysctl_table *t;
4785
4786         if (p->sysctl == NULL)
4787                 return;
4788
4789         t = p->sysctl;
4790         p->sysctl = NULL;
4791         unregister_net_sysctl_table(t->sysctl_header);
4792         kfree(t);
4793 }
4794
4795 static void addrconf_sysctl_register(struct inet6_dev *idev)
4796 {
4797         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4798                               &ndisc_ifinfo_sysctl_change);
4799         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4800                                         idev, &idev->cnf);
4801 }
4802
4803 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4804 {
4805         __addrconf_sysctl_unregister(&idev->cnf);
4806         neigh_sysctl_unregister(idev->nd_parms);
4807 }
4808
4809
4810 #endif
4811
4812 static int __net_init addrconf_init_net(struct net *net)
4813 {
4814         int err = -ENOMEM;
4815         struct ipv6_devconf *all, *dflt;
4816
4817         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
4818         if (all == NULL)
4819                 goto err_alloc_all;
4820
4821         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4822         if (dflt == NULL)
4823                 goto err_alloc_dflt;
4824
4825         /* these will be inherited by all namespaces */
4826         dflt->autoconf = ipv6_defaults.autoconf;
4827         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4828
4829         net->ipv6.devconf_all = all;
4830         net->ipv6.devconf_dflt = dflt;
4831
4832 #ifdef CONFIG_SYSCTL
4833         err = __addrconf_sysctl_register(net, "all", NULL, all);
4834         if (err < 0)
4835                 goto err_reg_all;
4836
4837         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4838         if (err < 0)
4839                 goto err_reg_dflt;
4840 #endif
4841         return 0;
4842
4843 #ifdef CONFIG_SYSCTL
4844 err_reg_dflt:
4845         __addrconf_sysctl_unregister(all);
4846 err_reg_all:
4847         kfree(dflt);
4848 #endif
4849 err_alloc_dflt:
4850         kfree(all);
4851 err_alloc_all:
4852         return err;
4853 }
4854
4855 static void __net_exit addrconf_exit_net(struct net *net)
4856 {
4857 #ifdef CONFIG_SYSCTL
4858         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4859         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4860 #endif
4861         if (!net_eq(net, &init_net)) {
4862                 kfree(net->ipv6.devconf_dflt);
4863                 kfree(net->ipv6.devconf_all);
4864         }
4865 }
4866
4867 static struct pernet_operations addrconf_ops = {
4868         .init = addrconf_init_net,
4869         .exit = addrconf_exit_net,
4870 };
4871
4872 /*
4873  *      Device notifier
4874  */
4875
4876 int register_inet6addr_notifier(struct notifier_block *nb)
4877 {
4878         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4879 }
4880 EXPORT_SYMBOL(register_inet6addr_notifier);
4881
4882 int unregister_inet6addr_notifier(struct notifier_block *nb)
4883 {
4884         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4885 }
4886 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4887
4888 static struct rtnl_af_ops inet6_ops = {
4889         .family           = AF_INET6,
4890         .fill_link_af     = inet6_fill_link_af,
4891         .get_link_af_size = inet6_get_link_af_size,
4892 };
4893
4894 /*
4895  *      Init / cleanup code
4896  */
4897
4898 int __init addrconf_init(void)
4899 {
4900         int i, err;
4901
4902         err = ipv6_addr_label_init();
4903         if (err < 0) {
4904                 pr_crit("%s: cannot initialize default policy table: %d\n",
4905                         __func__, err);
4906                 goto out;
4907         }
4908
4909         err = register_pernet_subsys(&addrconf_ops);
4910         if (err < 0)
4911                 goto out_addrlabel;
4912
4913         /* The addrconf netdev notifier requires that loopback_dev
4914          * has it's ipv6 private information allocated and setup
4915          * before it can bring up and give link-local addresses
4916          * to other devices which are up.
4917          *
4918          * Unfortunately, loopback_dev is not necessarily the first
4919          * entry in the global dev_base list of net devices.  In fact,
4920          * it is likely to be the very last entry on that list.
4921          * So this causes the notifier registry below to try and
4922          * give link-local addresses to all devices besides loopback_dev
4923          * first, then loopback_dev, which cases all the non-loopback_dev
4924          * devices to fail to get a link-local address.
4925          *
4926          * So, as a temporary fix, allocate the ipv6 structure for
4927          * loopback_dev first by hand.
4928          * Longer term, all of the dependencies ipv6 has upon the loopback
4929          * device and it being up should be removed.
4930          */
4931         rtnl_lock();
4932         if (!ipv6_add_dev(init_net.loopback_dev))
4933                 err = -ENOMEM;
4934         rtnl_unlock();
4935         if (err)
4936                 goto errlo;
4937
4938         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4939                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4940
4941         register_netdevice_notifier(&ipv6_dev_notf);
4942
4943         addrconf_verify(0);
4944
4945         err = rtnl_af_register(&inet6_ops);
4946         if (err < 0)
4947                 goto errout_af;
4948
4949         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4950                               NULL);
4951         if (err < 0)
4952                 goto errout;
4953
4954         /* Only the first call to __rtnl_register can fail */
4955         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4956         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4957         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4958                         inet6_dump_ifaddr, NULL);
4959         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4960                         inet6_dump_ifmcaddr, NULL);
4961         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4962                         inet6_dump_ifacaddr, NULL);
4963         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4964                         NULL, NULL);
4965
4966         ipv6_addr_label_rtnl_register();
4967
4968         return 0;
4969 errout:
4970         rtnl_af_unregister(&inet6_ops);
4971 errout_af:
4972         unregister_netdevice_notifier(&ipv6_dev_notf);
4973 errlo:
4974         unregister_pernet_subsys(&addrconf_ops);
4975 out_addrlabel:
4976         ipv6_addr_label_cleanup();
4977 out:
4978         return err;
4979 }
4980
4981 void addrconf_cleanup(void)
4982 {
4983         struct net_device *dev;
4984         int i;
4985
4986         unregister_netdevice_notifier(&ipv6_dev_notf);
4987         unregister_pernet_subsys(&addrconf_ops);
4988         ipv6_addr_label_cleanup();
4989
4990         rtnl_lock();
4991
4992         __rtnl_af_unregister(&inet6_ops);
4993
4994         /* clean dev list */
4995         for_each_netdev(&init_net, dev) {
4996                 if (__in6_dev_get(dev) == NULL)
4997                         continue;
4998                 addrconf_ifdown(dev, 1);
4999         }
5000         addrconf_ifdown(init_net.loopback_dev, 2);
5001
5002         /*
5003          *      Check hash table.
5004          */
5005         spin_lock_bh(&addrconf_hash_lock);
5006         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5007                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5008         spin_unlock_bh(&addrconf_hash_lock);
5009
5010         del_timer(&addr_chk_timer);
5011         rtnl_unlock();
5012 }