]> rtime.felk.cvut.cz Git - linux-imx.git/blob - net/ipv6/addrconf.c
ipv6: addrconf.c: remove unnecessary "if"
[linux-imx.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
88 #endif
89
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
99
100 #if ACONF_DEBUG >= 3
101 #define ADBG(x) printk x
102 #else
103 #define ADBG(x)
104 #endif
105
106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
107
108 static inline u32 cstamp_delta(unsigned long cstamp)
109 {
110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 }
112
113 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
114 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
115 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
116
117 #ifdef CONFIG_SYSCTL
118 static void addrconf_sysctl_register(struct inet6_dev *idev);
119 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
120 #else
121 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
122 {
123 }
124
125 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
126 {
127 }
128 #endif
129
130 #ifdef CONFIG_IPV6_PRIVACY
131 static void __ipv6_regen_rndid(struct inet6_dev *idev);
132 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
133 static void ipv6_regen_rndid(unsigned long data);
134 #endif
135
136 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
137 static int ipv6_count_addresses(struct inet6_dev *idev);
138
139 /*
140  *      Configured unicast address hash table
141  */
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
144
145 static void addrconf_verify(unsigned long);
146
147 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
148 static DEFINE_SPINLOCK(addrconf_verify_lock);
149
150 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
151 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
152
153 static void addrconf_type_change(struct net_device *dev,
154                                  unsigned long event);
155 static int addrconf_ifdown(struct net_device *dev, int how);
156
157 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_timer(unsigned long data);
159 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_run(struct inet6_dev *idev);
161 static void addrconf_rs_timer(unsigned long data);
162 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164
165 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
166                                 struct prefix_info *pinfo);
167 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
168                                struct net_device *dev);
169
170 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
171
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
173         .forwarding             = 0,
174         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
175         .mtu6                   = IPV6_MIN_MTU,
176         .accept_ra              = 1,
177         .accept_redirects       = 1,
178         .autoconf               = 1,
179         .force_mld_version      = 0,
180         .dad_transmits          = 1,
181         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
182         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
183         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
184 #ifdef CONFIG_IPV6_PRIVACY
185         .use_tempaddr           = 0,
186         .temp_valid_lft         = TEMP_VALID_LIFETIME,
187         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
188         .regen_max_retry        = REGEN_MAX_RETRY,
189         .max_desync_factor      = MAX_DESYNC_FACTOR,
190 #endif
191         .max_addresses          = IPV6_MAX_ADDRESSES,
192         .accept_ra_defrtr       = 1,
193         .accept_ra_pinfo        = 1,
194 #ifdef CONFIG_IPV6_ROUTER_PREF
195         .accept_ra_rtr_pref     = 1,
196         .rtr_probe_interval     = 60 * HZ,
197 #ifdef CONFIG_IPV6_ROUTE_INFO
198         .accept_ra_rt_info_max_plen = 0,
199 #endif
200 #endif
201         .proxy_ndp              = 0,
202         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
203         .disable_ipv6           = 0,
204         .accept_dad             = 1,
205 };
206
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
208         .forwarding             = 0,
209         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
210         .mtu6                   = IPV6_MIN_MTU,
211         .accept_ra              = 1,
212         .accept_redirects       = 1,
213         .autoconf               = 1,
214         .dad_transmits          = 1,
215         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
216         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
217         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
218 #ifdef CONFIG_IPV6_PRIVACY
219         .use_tempaddr           = 0,
220         .temp_valid_lft         = TEMP_VALID_LIFETIME,
221         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
222         .regen_max_retry        = REGEN_MAX_RETRY,
223         .max_desync_factor      = MAX_DESYNC_FACTOR,
224 #endif
225         .max_addresses          = IPV6_MAX_ADDRESSES,
226         .accept_ra_defrtr       = 1,
227         .accept_ra_pinfo        = 1,
228 #ifdef CONFIG_IPV6_ROUTER_PREF
229         .accept_ra_rtr_pref     = 1,
230         .rtr_probe_interval     = 60 * HZ,
231 #ifdef CONFIG_IPV6_ROUTE_INFO
232         .accept_ra_rt_info_max_plen = 0,
233 #endif
234 #endif
235         .proxy_ndp              = 0,
236         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
237         .disable_ipv6           = 0,
238         .accept_dad             = 1,
239 };
240
241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
246
247 /* Check if a valid qdisc is available */
248 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
249 {
250         return !qdisc_tx_is_noop(dev);
251 }
252
253 /* Check if a route is valid prefix route */
254 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
255 {
256         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
257 }
258
259 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
260 {
261         if (del_timer(&ifp->timer))
262                 __in6_ifa_put(ifp);
263 }
264
265 enum addrconf_timer_t {
266         AC_NONE,
267         AC_DAD,
268         AC_RS,
269 };
270
271 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
272                                enum addrconf_timer_t what,
273                                unsigned long when)
274 {
275         if (!del_timer(&ifp->timer))
276                 in6_ifa_hold(ifp);
277
278         switch (what) {
279         case AC_DAD:
280                 ifp->timer.function = addrconf_dad_timer;
281                 break;
282         case AC_RS:
283                 ifp->timer.function = addrconf_rs_timer;
284                 break;
285         default:
286                 break;
287         }
288         ifp->timer.expires = jiffies + when;
289         add_timer(&ifp->timer);
290 }
291
292 static int snmp6_alloc_dev(struct inet6_dev *idev)
293 {
294         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
295                           sizeof(struct ipstats_mib),
296                           __alignof__(struct ipstats_mib)) < 0)
297                 goto err_ip;
298         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
299                                         GFP_KERNEL);
300         if (!idev->stats.icmpv6dev)
301                 goto err_icmp;
302         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
303                                            GFP_KERNEL);
304         if (!idev->stats.icmpv6msgdev)
305                 goto err_icmpmsg;
306
307         return 0;
308
309 err_icmpmsg:
310         kfree(idev->stats.icmpv6dev);
311 err_icmp:
312         snmp_mib_free((void __percpu **)idev->stats.ipv6);
313 err_ip:
314         return -ENOMEM;
315 }
316
317 static void snmp6_free_dev(struct inet6_dev *idev)
318 {
319         kfree(idev->stats.icmpv6msgdev);
320         kfree(idev->stats.icmpv6dev);
321         snmp_mib_free((void __percpu **)idev->stats.ipv6);
322 }
323
324 /* Nobody refers to this device, we may destroy it. */
325
326 void in6_dev_finish_destroy(struct inet6_dev *idev)
327 {
328         struct net_device *dev = idev->dev;
329
330         WARN_ON(!list_empty(&idev->addr_list));
331         WARN_ON(idev->mc_list != NULL);
332
333 #ifdef NET_REFCNT_DEBUG
334         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
335 #endif
336         dev_put(dev);
337         if (!idev->dead) {
338                 pr_warn("Freeing alive inet6 device %p\n", idev);
339                 return;
340         }
341         snmp6_free_dev(idev);
342         kfree_rcu(idev, rcu);
343 }
344 EXPORT_SYMBOL(in6_dev_finish_destroy);
345
346 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
347 {
348         struct inet6_dev *ndev;
349
350         ASSERT_RTNL();
351
352         if (dev->mtu < IPV6_MIN_MTU)
353                 return NULL;
354
355         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
356
357         if (ndev == NULL)
358                 return NULL;
359
360         rwlock_init(&ndev->lock);
361         ndev->dev = dev;
362         INIT_LIST_HEAD(&ndev->addr_list);
363
364         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
365         ndev->cnf.mtu6 = dev->mtu;
366         ndev->cnf.sysctl = NULL;
367         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
368         if (ndev->nd_parms == NULL) {
369                 kfree(ndev);
370                 return NULL;
371         }
372         if (ndev->cnf.forwarding)
373                 dev_disable_lro(dev);
374         /* We refer to the device */
375         dev_hold(dev);
376
377         if (snmp6_alloc_dev(ndev) < 0) {
378                 ADBG((KERN_WARNING
379                         "%s: cannot allocate memory for statistics; dev=%s.\n",
380                         __func__, dev->name));
381                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
382                 dev_put(dev);
383                 kfree(ndev);
384                 return NULL;
385         }
386
387         if (snmp6_register_dev(ndev) < 0) {
388                 ADBG((KERN_WARNING
389                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
390                         __func__, dev->name));
391                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
392                 ndev->dead = 1;
393                 in6_dev_finish_destroy(ndev);
394                 return NULL;
395         }
396
397         /* One reference from device.  We must do this before
398          * we invoke __ipv6_regen_rndid().
399          */
400         in6_dev_hold(ndev);
401
402         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
403                 ndev->cnf.accept_dad = -1;
404
405 #if IS_ENABLED(CONFIG_IPV6_SIT)
406         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
407                 pr_info("%s: Disabled Multicast RS\n", dev->name);
408                 ndev->cnf.rtr_solicits = 0;
409         }
410 #endif
411
412 #ifdef CONFIG_IPV6_PRIVACY
413         INIT_LIST_HEAD(&ndev->tempaddr_list);
414         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
415         if ((dev->flags&IFF_LOOPBACK) ||
416             dev->type == ARPHRD_TUNNEL ||
417             dev->type == ARPHRD_TUNNEL6 ||
418             dev->type == ARPHRD_SIT ||
419             dev->type == ARPHRD_NONE) {
420                 ndev->cnf.use_tempaddr = -1;
421         } else {
422                 in6_dev_hold(ndev);
423                 ipv6_regen_rndid((unsigned long) ndev);
424         }
425 #endif
426
427         if (netif_running(dev) && addrconf_qdisc_ok(dev))
428                 ndev->if_flags |= IF_READY;
429
430         ipv6_mc_init_dev(ndev);
431         ndev->tstamp = jiffies;
432         addrconf_sysctl_register(ndev);
433         /* protected by rtnl_lock */
434         rcu_assign_pointer(dev->ip6_ptr, ndev);
435
436         /* Join all-node multicast group */
437         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
438
439         /* Join all-router multicast group if forwarding is set */
440         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442
443         return ndev;
444 }
445
446 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
447 {
448         struct inet6_dev *idev;
449
450         ASSERT_RTNL();
451
452         idev = __in6_dev_get(dev);
453         if (!idev) {
454                 idev = ipv6_add_dev(dev);
455                 if (!idev)
456                         return NULL;
457         }
458
459         if (dev->flags&IFF_UP)
460                 ipv6_mc_up(idev);
461         return idev;
462 }
463
464 static int inet6_netconf_msgsize_devconf(int type)
465 {
466         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
467                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
468
469         /* type -1 is used for ALL */
470         if (type == -1 || type == NETCONFA_FORWARDING)
471                 size += nla_total_size(4);
472 #ifdef CONFIG_IPV6_MROUTE
473         if (type == -1 || type == NETCONFA_MC_FORWARDING)
474                 size += nla_total_size(4);
475 #endif
476
477         return size;
478 }
479
480 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
481                                       struct ipv6_devconf *devconf, u32 portid,
482                                       u32 seq, int event, unsigned int flags,
483                                       int type)
484 {
485         struct nlmsghdr  *nlh;
486         struct netconfmsg *ncm;
487
488         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
489                         flags);
490         if (nlh == NULL)
491                 return -EMSGSIZE;
492
493         ncm = nlmsg_data(nlh);
494         ncm->ncm_family = AF_INET6;
495
496         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
497                 goto nla_put_failure;
498
499         /* type -1 is used for ALL */
500         if ((type == -1 || type == NETCONFA_FORWARDING) &&
501             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
502                 goto nla_put_failure;
503 #ifdef CONFIG_IPV6_MROUTE
504         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
505             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
506                         devconf->mc_forwarding) < 0)
507                 goto nla_put_failure;
508 #endif
509         return nlmsg_end(skb, nlh);
510
511 nla_put_failure:
512         nlmsg_cancel(skb, nlh);
513         return -EMSGSIZE;
514 }
515
516 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
517                                   struct ipv6_devconf *devconf)
518 {
519         struct sk_buff *skb;
520         int err = -ENOBUFS;
521
522         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
523         if (skb == NULL)
524                 goto errout;
525
526         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
527                                          RTM_NEWNETCONF, 0, type);
528         if (err < 0) {
529                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
530                 WARN_ON(err == -EMSGSIZE);
531                 kfree_skb(skb);
532                 goto errout;
533         }
534         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
535         return;
536 errout:
537         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
538 }
539
540 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
541         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
542         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
543 };
544
545 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
546                                      struct nlmsghdr *nlh,
547                                      void *arg)
548 {
549         struct net *net = sock_net(in_skb->sk);
550         struct nlattr *tb[NETCONFA_MAX+1];
551         struct netconfmsg *ncm;
552         struct sk_buff *skb;
553         struct ipv6_devconf *devconf;
554         struct inet6_dev *in6_dev;
555         struct net_device *dev;
556         int ifindex;
557         int err;
558
559         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
560                           devconf_ipv6_policy);
561         if (err < 0)
562                 goto errout;
563
564         err = EINVAL;
565         if (!tb[NETCONFA_IFINDEX])
566                 goto errout;
567
568         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
569         switch (ifindex) {
570         case NETCONFA_IFINDEX_ALL:
571                 devconf = net->ipv6.devconf_all;
572                 break;
573         case NETCONFA_IFINDEX_DEFAULT:
574                 devconf = net->ipv6.devconf_dflt;
575                 break;
576         default:
577                 dev = __dev_get_by_index(net, ifindex);
578                 if (dev == NULL)
579                         goto errout;
580                 in6_dev = __in6_dev_get(dev);
581                 if (in6_dev == NULL)
582                         goto errout;
583                 devconf = &in6_dev->cnf;
584                 break;
585         }
586
587         err = -ENOBUFS;
588         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
589         if (skb == NULL)
590                 goto errout;
591
592         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
593                                          NETLINK_CB(in_skb).portid,
594                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
595                                          -1);
596         if (err < 0) {
597                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
598                 WARN_ON(err == -EMSGSIZE);
599                 kfree_skb(skb);
600                 goto errout;
601         }
602         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
603 errout:
604         return err;
605 }
606
607 #ifdef CONFIG_SYSCTL
608 static void dev_forward_change(struct inet6_dev *idev)
609 {
610         struct net_device *dev;
611         struct inet6_ifaddr *ifa;
612
613         if (!idev)
614                 return;
615         dev = idev->dev;
616         if (idev->cnf.forwarding)
617                 dev_disable_lro(dev);
618         if (dev->flags & IFF_MULTICAST) {
619                 if (idev->cnf.forwarding)
620                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
621                 else
622                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
623         }
624
625         list_for_each_entry(ifa, &idev->addr_list, if_list) {
626                 if (ifa->flags&IFA_F_TENTATIVE)
627                         continue;
628                 if (idev->cnf.forwarding)
629                         addrconf_join_anycast(ifa);
630                 else
631                         addrconf_leave_anycast(ifa);
632         }
633         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
634                                      dev->ifindex, &idev->cnf);
635 }
636
637
638 static void addrconf_forward_change(struct net *net, __s32 newf)
639 {
640         struct net_device *dev;
641         struct inet6_dev *idev;
642
643         for_each_netdev(net, dev) {
644                 idev = __in6_dev_get(dev);
645                 if (idev) {
646                         int changed = (!idev->cnf.forwarding) ^ (!newf);
647                         idev->cnf.forwarding = newf;
648                         if (changed)
649                                 dev_forward_change(idev);
650                 }
651         }
652 }
653
654 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
655 {
656         struct net *net;
657         int old;
658
659         if (!rtnl_trylock())
660                 return restart_syscall();
661
662         net = (struct net *)table->extra2;
663         old = *p;
664         *p = newf;
665
666         if (p == &net->ipv6.devconf_dflt->forwarding) {
667                 if ((!newf) ^ (!old))
668                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
669                                                      NETCONFA_IFINDEX_DEFAULT,
670                                                      net->ipv6.devconf_dflt);
671                 rtnl_unlock();
672                 return 0;
673         }
674
675         if (p == &net->ipv6.devconf_all->forwarding) {
676                 net->ipv6.devconf_dflt->forwarding = newf;
677                 addrconf_forward_change(net, newf);
678                 if ((!newf) ^ (!old))
679                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
680                                                      NETCONFA_IFINDEX_ALL,
681                                                      net->ipv6.devconf_all);
682         } else if ((!newf) ^ (!old))
683                 dev_forward_change((struct inet6_dev *)table->extra1);
684         rtnl_unlock();
685
686         if (newf)
687                 rt6_purge_dflt_routers(net);
688         return 1;
689 }
690 #endif
691
692 /* Nobody refers to this ifaddr, destroy it */
693 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
694 {
695         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
696
697 #ifdef NET_REFCNT_DEBUG
698         pr_debug("%s\n", __func__);
699 #endif
700
701         in6_dev_put(ifp->idev);
702
703         if (del_timer(&ifp->timer))
704                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
705
706         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
707                 pr_warn("Freeing alive inet6 address %p\n", ifp);
708                 return;
709         }
710         ip6_rt_put(ifp->rt);
711
712         kfree_rcu(ifp, rcu);
713 }
714
715 static void
716 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
717 {
718         struct list_head *p;
719         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
720
721         /*
722          * Each device address list is sorted in order of scope -
723          * global before linklocal.
724          */
725         list_for_each(p, &idev->addr_list) {
726                 struct inet6_ifaddr *ifa
727                         = list_entry(p, struct inet6_ifaddr, if_list);
728                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
729                         break;
730         }
731
732         list_add_tail(&ifp->if_list, p);
733 }
734
735 static u32 inet6_addr_hash(const struct in6_addr *addr)
736 {
737         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
738 }
739
740 /* On success it returns ifp with increased reference count */
741
742 static struct inet6_ifaddr *
743 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
744               int scope, u32 flags)
745 {
746         struct inet6_ifaddr *ifa = NULL;
747         struct rt6_info *rt;
748         unsigned int hash;
749         int err = 0;
750         int addr_type = ipv6_addr_type(addr);
751
752         if (addr_type == IPV6_ADDR_ANY ||
753             addr_type & IPV6_ADDR_MULTICAST ||
754             (!(idev->dev->flags & IFF_LOOPBACK) &&
755              addr_type & IPV6_ADDR_LOOPBACK))
756                 return ERR_PTR(-EADDRNOTAVAIL);
757
758         rcu_read_lock_bh();
759         if (idev->dead) {
760                 err = -ENODEV;                  /*XXX*/
761                 goto out2;
762         }
763
764         if (idev->cnf.disable_ipv6) {
765                 err = -EACCES;
766                 goto out2;
767         }
768
769         spin_lock(&addrconf_hash_lock);
770
771         /* Ignore adding duplicate addresses on an interface */
772         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
773                 ADBG(("ipv6_add_addr: already assigned\n"));
774                 err = -EEXIST;
775                 goto out;
776         }
777
778         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
779
780         if (ifa == NULL) {
781                 ADBG(("ipv6_add_addr: malloc failed\n"));
782                 err = -ENOBUFS;
783                 goto out;
784         }
785
786         rt = addrconf_dst_alloc(idev, addr, false);
787         if (IS_ERR(rt)) {
788                 err = PTR_ERR(rt);
789                 goto out;
790         }
791
792         ifa->addr = *addr;
793
794         spin_lock_init(&ifa->lock);
795         spin_lock_init(&ifa->state_lock);
796         init_timer(&ifa->timer);
797         INIT_HLIST_NODE(&ifa->addr_lst);
798         ifa->timer.data = (unsigned long) ifa;
799         ifa->scope = scope;
800         ifa->prefix_len = pfxlen;
801         ifa->flags = flags | IFA_F_TENTATIVE;
802         ifa->cstamp = ifa->tstamp = jiffies;
803
804         ifa->rt = rt;
805
806         ifa->idev = idev;
807         in6_dev_hold(idev);
808         /* For caller */
809         in6_ifa_hold(ifa);
810
811         /* Add to big hash table */
812         hash = inet6_addr_hash(addr);
813
814         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
815         spin_unlock(&addrconf_hash_lock);
816
817         write_lock(&idev->lock);
818         /* Add to inet6_dev unicast addr list. */
819         ipv6_link_dev_addr(idev, ifa);
820
821 #ifdef CONFIG_IPV6_PRIVACY
822         if (ifa->flags&IFA_F_TEMPORARY) {
823                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
824                 in6_ifa_hold(ifa);
825         }
826 #endif
827
828         in6_ifa_hold(ifa);
829         write_unlock(&idev->lock);
830 out2:
831         rcu_read_unlock_bh();
832
833         if (likely(err == 0))
834                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
835         else {
836                 kfree(ifa);
837                 ifa = ERR_PTR(err);
838         }
839
840         return ifa;
841 out:
842         spin_unlock(&addrconf_hash_lock);
843         goto out2;
844 }
845
846 /* This function wants to get referenced ifp and releases it before return */
847
848 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
849 {
850         struct inet6_ifaddr *ifa, *ifn;
851         struct inet6_dev *idev = ifp->idev;
852         int state;
853         int deleted = 0, onlink = 0;
854         unsigned long expires = jiffies;
855
856         spin_lock_bh(&ifp->state_lock);
857         state = ifp->state;
858         ifp->state = INET6_IFADDR_STATE_DEAD;
859         spin_unlock_bh(&ifp->state_lock);
860
861         if (state == INET6_IFADDR_STATE_DEAD)
862                 goto out;
863
864         spin_lock_bh(&addrconf_hash_lock);
865         hlist_del_init_rcu(&ifp->addr_lst);
866         spin_unlock_bh(&addrconf_hash_lock);
867
868         write_lock_bh(&idev->lock);
869 #ifdef CONFIG_IPV6_PRIVACY
870         if (ifp->flags&IFA_F_TEMPORARY) {
871                 list_del(&ifp->tmp_list);
872                 if (ifp->ifpub) {
873                         in6_ifa_put(ifp->ifpub);
874                         ifp->ifpub = NULL;
875                 }
876                 __in6_ifa_put(ifp);
877         }
878 #endif
879
880         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
881                 if (ifa == ifp) {
882                         list_del_init(&ifp->if_list);
883                         __in6_ifa_put(ifp);
884
885                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
886                                 break;
887                         deleted = 1;
888                         continue;
889                 } else if (ifp->flags & IFA_F_PERMANENT) {
890                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
891                                               ifp->prefix_len)) {
892                                 if (ifa->flags & IFA_F_PERMANENT) {
893                                         onlink = 1;
894                                         if (deleted)
895                                                 break;
896                                 } else {
897                                         unsigned long lifetime;
898
899                                         if (!onlink)
900                                                 onlink = -1;
901
902                                         spin_lock(&ifa->lock);
903
904                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
905                                         /*
906                                          * Note: Because this address is
907                                          * not permanent, lifetime <
908                                          * LONG_MAX / HZ here.
909                                          */
910                                         if (time_before(expires,
911                                                         ifa->tstamp + lifetime * HZ))
912                                                 expires = ifa->tstamp + lifetime * HZ;
913                                         spin_unlock(&ifa->lock);
914                                 }
915                         }
916                 }
917         }
918         write_unlock_bh(&idev->lock);
919
920         addrconf_del_timer(ifp);
921
922         ipv6_ifa_notify(RTM_DELADDR, ifp);
923
924         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
925
926         /*
927          * Purge or update corresponding prefix
928          *
929          * 1) we don't purge prefix here if address was not permanent.
930          *    prefix is managed by its own lifetime.
931          * 2) if there're no addresses, delete prefix.
932          * 3) if there're still other permanent address(es),
933          *    corresponding prefix is still permanent.
934          * 4) otherwise, update prefix lifetime to the
935          *    longest valid lifetime among the corresponding
936          *    addresses on the device.
937          *    Note: subsequent RA will update lifetime.
938          *
939          * --yoshfuji
940          */
941         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
942                 struct in6_addr prefix;
943                 struct rt6_info *rt;
944                 struct net *net = dev_net(ifp->idev->dev);
945                 struct flowi6 fl6 = {};
946
947                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
948                 fl6.flowi6_oif = ifp->idev->dev->ifindex;
949                 fl6.daddr = prefix;
950                 rt = (struct rt6_info *)ip6_route_lookup(net, &fl6,
951                                                          RT6_LOOKUP_F_IFACE);
952
953                 if (rt != net->ipv6.ip6_null_entry &&
954                     addrconf_is_prefix_route(rt)) {
955                         if (onlink == 0) {
956                                 ip6_del_rt(rt);
957                                 rt = NULL;
958                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
959                                 rt6_set_expires(rt, expires);
960                         }
961                 }
962                 ip6_rt_put(rt);
963         }
964
965         /* clean up prefsrc entries */
966         rt6_remove_prefsrc(ifp);
967 out:
968         in6_ifa_put(ifp);
969 }
970
971 #ifdef CONFIG_IPV6_PRIVACY
972 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
973 {
974         struct inet6_dev *idev = ifp->idev;
975         struct in6_addr addr, *tmpaddr;
976         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
977         unsigned long regen_advance;
978         int tmp_plen;
979         int ret = 0;
980         int max_addresses;
981         u32 addr_flags;
982         unsigned long now = jiffies;
983
984         write_lock(&idev->lock);
985         if (ift) {
986                 spin_lock_bh(&ift->lock);
987                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
988                 spin_unlock_bh(&ift->lock);
989                 tmpaddr = &addr;
990         } else {
991                 tmpaddr = NULL;
992         }
993 retry:
994         in6_dev_hold(idev);
995         if (idev->cnf.use_tempaddr <= 0) {
996                 write_unlock(&idev->lock);
997                 pr_info("%s: use_tempaddr is disabled\n", __func__);
998                 in6_dev_put(idev);
999                 ret = -1;
1000                 goto out;
1001         }
1002         spin_lock_bh(&ifp->lock);
1003         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1004                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1005                 spin_unlock_bh(&ifp->lock);
1006                 write_unlock(&idev->lock);
1007                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1008                         __func__);
1009                 in6_dev_put(idev);
1010                 ret = -1;
1011                 goto out;
1012         }
1013         in6_ifa_hold(ifp);
1014         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1015         __ipv6_try_regen_rndid(idev, tmpaddr);
1016         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1017         age = (now - ifp->tstamp) / HZ;
1018         tmp_valid_lft = min_t(__u32,
1019                               ifp->valid_lft,
1020                               idev->cnf.temp_valid_lft + age);
1021         tmp_prefered_lft = min_t(__u32,
1022                                  ifp->prefered_lft,
1023                                  idev->cnf.temp_prefered_lft + age -
1024                                  idev->cnf.max_desync_factor);
1025         tmp_plen = ifp->prefix_len;
1026         max_addresses = idev->cnf.max_addresses;
1027         tmp_tstamp = ifp->tstamp;
1028         spin_unlock_bh(&ifp->lock);
1029
1030         regen_advance = idev->cnf.regen_max_retry *
1031                         idev->cnf.dad_transmits *
1032                         idev->nd_parms->retrans_time / HZ;
1033         write_unlock(&idev->lock);
1034
1035         /* A temporary address is created only if this calculated Preferred
1036          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1037          * an implementation must not create a temporary address with a zero
1038          * Preferred Lifetime.
1039          */
1040         if (tmp_prefered_lft <= regen_advance) {
1041                 in6_ifa_put(ifp);
1042                 in6_dev_put(idev);
1043                 ret = -1;
1044                 goto out;
1045         }
1046
1047         addr_flags = IFA_F_TEMPORARY;
1048         /* set in addrconf_prefix_rcv() */
1049         if (ifp->flags & IFA_F_OPTIMISTIC)
1050                 addr_flags |= IFA_F_OPTIMISTIC;
1051
1052         ift = !max_addresses ||
1053               ipv6_count_addresses(idev) < max_addresses ?
1054                 ipv6_add_addr(idev, &addr, tmp_plen,
1055                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1056                               addr_flags) : NULL;
1057         if (!ift || IS_ERR(ift)) {
1058                 in6_ifa_put(ifp);
1059                 in6_dev_put(idev);
1060                 pr_info("%s: retry temporary address regeneration\n", __func__);
1061                 tmpaddr = &addr;
1062                 write_lock(&idev->lock);
1063                 goto retry;
1064         }
1065
1066         spin_lock_bh(&ift->lock);
1067         ift->ifpub = ifp;
1068         ift->valid_lft = tmp_valid_lft;
1069         ift->prefered_lft = tmp_prefered_lft;
1070         ift->cstamp = now;
1071         ift->tstamp = tmp_tstamp;
1072         spin_unlock_bh(&ift->lock);
1073
1074         addrconf_dad_start(ift);
1075         in6_ifa_put(ift);
1076         in6_dev_put(idev);
1077 out:
1078         return ret;
1079 }
1080 #endif
1081
1082 /*
1083  *      Choose an appropriate source address (RFC3484)
1084  */
1085 enum {
1086         IPV6_SADDR_RULE_INIT = 0,
1087         IPV6_SADDR_RULE_LOCAL,
1088         IPV6_SADDR_RULE_SCOPE,
1089         IPV6_SADDR_RULE_PREFERRED,
1090 #ifdef CONFIG_IPV6_MIP6
1091         IPV6_SADDR_RULE_HOA,
1092 #endif
1093         IPV6_SADDR_RULE_OIF,
1094         IPV6_SADDR_RULE_LABEL,
1095 #ifdef CONFIG_IPV6_PRIVACY
1096         IPV6_SADDR_RULE_PRIVACY,
1097 #endif
1098         IPV6_SADDR_RULE_ORCHID,
1099         IPV6_SADDR_RULE_PREFIX,
1100         IPV6_SADDR_RULE_MAX
1101 };
1102
1103 struct ipv6_saddr_score {
1104         int                     rule;
1105         int                     addr_type;
1106         struct inet6_ifaddr     *ifa;
1107         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1108         int                     scopedist;
1109         int                     matchlen;
1110 };
1111
1112 struct ipv6_saddr_dst {
1113         const struct in6_addr *addr;
1114         int ifindex;
1115         int scope;
1116         int label;
1117         unsigned int prefs;
1118 };
1119
1120 static inline int ipv6_saddr_preferred(int type)
1121 {
1122         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1123                 return 1;
1124         return 0;
1125 }
1126
1127 static int ipv6_get_saddr_eval(struct net *net,
1128                                struct ipv6_saddr_score *score,
1129                                struct ipv6_saddr_dst *dst,
1130                                int i)
1131 {
1132         int ret;
1133
1134         if (i <= score->rule) {
1135                 switch (i) {
1136                 case IPV6_SADDR_RULE_SCOPE:
1137                         ret = score->scopedist;
1138                         break;
1139                 case IPV6_SADDR_RULE_PREFIX:
1140                         ret = score->matchlen;
1141                         break;
1142                 default:
1143                         ret = !!test_bit(i, score->scorebits);
1144                 }
1145                 goto out;
1146         }
1147
1148         switch (i) {
1149         case IPV6_SADDR_RULE_INIT:
1150                 /* Rule 0: remember if hiscore is not ready yet */
1151                 ret = !!score->ifa;
1152                 break;
1153         case IPV6_SADDR_RULE_LOCAL:
1154                 /* Rule 1: Prefer same address */
1155                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1156                 break;
1157         case IPV6_SADDR_RULE_SCOPE:
1158                 /* Rule 2: Prefer appropriate scope
1159                  *
1160                  *      ret
1161                  *       ^
1162                  *    -1 |  d 15
1163                  *    ---+--+-+---> scope
1164                  *       |
1165                  *       |             d is scope of the destination.
1166                  *  B-d  |  \
1167                  *       |   \      <- smaller scope is better if
1168                  *  B-15 |    \        if scope is enough for destinaion.
1169                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1170                  * d-C-1 | /
1171                  *       |/         <- greater is better
1172                  *   -C  /             if scope is not enough for destination.
1173                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1174                  *
1175                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1176                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1177                  * Assume B = 0 and we get C > 29.
1178                  */
1179                 ret = __ipv6_addr_src_scope(score->addr_type);
1180                 if (ret >= dst->scope)
1181                         ret = -ret;
1182                 else
1183                         ret -= 128;     /* 30 is enough */
1184                 score->scopedist = ret;
1185                 break;
1186         case IPV6_SADDR_RULE_PREFERRED:
1187                 /* Rule 3: Avoid deprecated and optimistic addresses */
1188                 ret = ipv6_saddr_preferred(score->addr_type) ||
1189                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1190                 break;
1191 #ifdef CONFIG_IPV6_MIP6
1192         case IPV6_SADDR_RULE_HOA:
1193             {
1194                 /* Rule 4: Prefer home address */
1195                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1196                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1197                 break;
1198             }
1199 #endif
1200         case IPV6_SADDR_RULE_OIF:
1201                 /* Rule 5: Prefer outgoing interface */
1202                 ret = (!dst->ifindex ||
1203                        dst->ifindex == score->ifa->idev->dev->ifindex);
1204                 break;
1205         case IPV6_SADDR_RULE_LABEL:
1206                 /* Rule 6: Prefer matching label */
1207                 ret = ipv6_addr_label(net,
1208                                       &score->ifa->addr, score->addr_type,
1209                                       score->ifa->idev->dev->ifindex) == dst->label;
1210                 break;
1211 #ifdef CONFIG_IPV6_PRIVACY
1212         case IPV6_SADDR_RULE_PRIVACY:
1213             {
1214                 /* Rule 7: Prefer public address
1215                  * Note: prefer temporary address if use_tempaddr >= 2
1216                  */
1217                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1218                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1219                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1220                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1221                 break;
1222             }
1223 #endif
1224         case IPV6_SADDR_RULE_ORCHID:
1225                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1226                  *          non-ORCHID vs non-ORCHID
1227                  */
1228                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1229                         ipv6_addr_orchid(dst->addr));
1230                 break;
1231         case IPV6_SADDR_RULE_PREFIX:
1232                 /* Rule 8: Use longest matching prefix */
1233                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1234                 if (ret > score->ifa->prefix_len)
1235                         ret = score->ifa->prefix_len;
1236                 score->matchlen = ret;
1237                 break;
1238         default:
1239                 ret = 0;
1240         }
1241
1242         if (ret)
1243                 __set_bit(i, score->scorebits);
1244         score->rule = i;
1245 out:
1246         return ret;
1247 }
1248
1249 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1250                        const struct in6_addr *daddr, unsigned int prefs,
1251                        struct in6_addr *saddr)
1252 {
1253         struct ipv6_saddr_score scores[2],
1254                                 *score = &scores[0], *hiscore = &scores[1];
1255         struct ipv6_saddr_dst dst;
1256         struct net_device *dev;
1257         int dst_type;
1258
1259         dst_type = __ipv6_addr_type(daddr);
1260         dst.addr = daddr;
1261         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1262         dst.scope = __ipv6_addr_src_scope(dst_type);
1263         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1264         dst.prefs = prefs;
1265
1266         hiscore->rule = -1;
1267         hiscore->ifa = NULL;
1268
1269         rcu_read_lock();
1270
1271         for_each_netdev_rcu(net, dev) {
1272                 struct inet6_dev *idev;
1273
1274                 /* Candidate Source Address (section 4)
1275                  *  - multicast and link-local destination address,
1276                  *    the set of candidate source address MUST only
1277                  *    include addresses assigned to interfaces
1278                  *    belonging to the same link as the outgoing
1279                  *    interface.
1280                  * (- For site-local destination addresses, the
1281                  *    set of candidate source addresses MUST only
1282                  *    include addresses assigned to interfaces
1283                  *    belonging to the same site as the outgoing
1284                  *    interface.)
1285                  */
1286                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1287                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1288                     dst.ifindex && dev->ifindex != dst.ifindex)
1289                         continue;
1290
1291                 idev = __in6_dev_get(dev);
1292                 if (!idev)
1293                         continue;
1294
1295                 read_lock_bh(&idev->lock);
1296                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1297                         int i;
1298
1299                         /*
1300                          * - Tentative Address (RFC2462 section 5.4)
1301                          *  - A tentative address is not considered
1302                          *    "assigned to an interface" in the traditional
1303                          *    sense, unless it is also flagged as optimistic.
1304                          * - Candidate Source Address (section 4)
1305                          *  - In any case, anycast addresses, multicast
1306                          *    addresses, and the unspecified address MUST
1307                          *    NOT be included in a candidate set.
1308                          */
1309                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1310                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1311                                 continue;
1312
1313                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1314
1315                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1316                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1317                                 LIMIT_NETDEBUG(KERN_DEBUG
1318                                                "ADDRCONF: unspecified / multicast address "
1319                                                "assigned as unicast address on %s",
1320                                                dev->name);
1321                                 continue;
1322                         }
1323
1324                         score->rule = -1;
1325                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1326
1327                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1328                                 int minihiscore, miniscore;
1329
1330                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1331                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1332
1333                                 if (minihiscore > miniscore) {
1334                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1335                                             score->scopedist > 0) {
1336                                                 /*
1337                                                  * special case:
1338                                                  * each remaining entry
1339                                                  * has too small (not enough)
1340                                                  * scope, because ifa entries
1341                                                  * are sorted by their scope
1342                                                  * values.
1343                                                  */
1344                                                 goto try_nextdev;
1345                                         }
1346                                         break;
1347                                 } else if (minihiscore < miniscore) {
1348                                         if (hiscore->ifa)
1349                                                 in6_ifa_put(hiscore->ifa);
1350
1351                                         in6_ifa_hold(score->ifa);
1352
1353                                         swap(hiscore, score);
1354
1355                                         /* restore our iterator */
1356                                         score->ifa = hiscore->ifa;
1357
1358                                         break;
1359                                 }
1360                         }
1361                 }
1362 try_nextdev:
1363                 read_unlock_bh(&idev->lock);
1364         }
1365         rcu_read_unlock();
1366
1367         if (!hiscore->ifa)
1368                 return -EADDRNOTAVAIL;
1369
1370         *saddr = hiscore->ifa->addr;
1371         in6_ifa_put(hiscore->ifa);
1372         return 0;
1373 }
1374 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1375
1376 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1377                     unsigned char banned_flags)
1378 {
1379         struct inet6_dev *idev;
1380         int err = -EADDRNOTAVAIL;
1381
1382         rcu_read_lock();
1383         idev = __in6_dev_get(dev);
1384         if (idev) {
1385                 struct inet6_ifaddr *ifp;
1386
1387                 read_lock_bh(&idev->lock);
1388                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1389                         if (ifp->scope == IFA_LINK &&
1390                             !(ifp->flags & banned_flags)) {
1391                                 *addr = ifp->addr;
1392                                 err = 0;
1393                                 break;
1394                         }
1395                 }
1396                 read_unlock_bh(&idev->lock);
1397         }
1398         rcu_read_unlock();
1399         return err;
1400 }
1401
1402 static int ipv6_count_addresses(struct inet6_dev *idev)
1403 {
1404         int cnt = 0;
1405         struct inet6_ifaddr *ifp;
1406
1407         read_lock_bh(&idev->lock);
1408         list_for_each_entry(ifp, &idev->addr_list, if_list)
1409                 cnt++;
1410         read_unlock_bh(&idev->lock);
1411         return cnt;
1412 }
1413
1414 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1415                   struct net_device *dev, int strict)
1416 {
1417         struct inet6_ifaddr *ifp;
1418         struct hlist_node *node;
1419         unsigned int hash = inet6_addr_hash(addr);
1420
1421         rcu_read_lock_bh();
1422         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1423                 if (!net_eq(dev_net(ifp->idev->dev), net))
1424                         continue;
1425                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1426                     !(ifp->flags&IFA_F_TENTATIVE) &&
1427                     (dev == NULL || ifp->idev->dev == dev ||
1428                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1429                         rcu_read_unlock_bh();
1430                         return 1;
1431                 }
1432         }
1433
1434         rcu_read_unlock_bh();
1435         return 0;
1436 }
1437 EXPORT_SYMBOL(ipv6_chk_addr);
1438
1439 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1440                                struct net_device *dev)
1441 {
1442         unsigned int hash = inet6_addr_hash(addr);
1443         struct inet6_ifaddr *ifp;
1444         struct hlist_node *node;
1445
1446         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1447                 if (!net_eq(dev_net(ifp->idev->dev), net))
1448                         continue;
1449                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1450                         if (dev == NULL || ifp->idev->dev == dev)
1451                                 return true;
1452                 }
1453         }
1454         return false;
1455 }
1456
1457 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1458 {
1459         struct inet6_dev *idev;
1460         struct inet6_ifaddr *ifa;
1461         int     onlink;
1462
1463         onlink = 0;
1464         rcu_read_lock();
1465         idev = __in6_dev_get(dev);
1466         if (idev) {
1467                 read_lock_bh(&idev->lock);
1468                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1469                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1470                                                    ifa->prefix_len);
1471                         if (onlink)
1472                                 break;
1473                 }
1474                 read_unlock_bh(&idev->lock);
1475         }
1476         rcu_read_unlock();
1477         return onlink;
1478 }
1479 EXPORT_SYMBOL(ipv6_chk_prefix);
1480
1481 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1482                                      struct net_device *dev, int strict)
1483 {
1484         struct inet6_ifaddr *ifp, *result = NULL;
1485         unsigned int hash = inet6_addr_hash(addr);
1486         struct hlist_node *node;
1487
1488         rcu_read_lock_bh();
1489         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1490                 if (!net_eq(dev_net(ifp->idev->dev), net))
1491                         continue;
1492                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1493                         if (dev == NULL || ifp->idev->dev == dev ||
1494                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1495                                 result = ifp;
1496                                 in6_ifa_hold(ifp);
1497                                 break;
1498                         }
1499                 }
1500         }
1501         rcu_read_unlock_bh();
1502
1503         return result;
1504 }
1505
1506 /* Gets referenced address, destroys ifaddr */
1507
1508 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1509 {
1510         if (ifp->flags&IFA_F_PERMANENT) {
1511                 spin_lock_bh(&ifp->lock);
1512                 addrconf_del_timer(ifp);
1513                 ifp->flags |= IFA_F_TENTATIVE;
1514                 if (dad_failed)
1515                         ifp->flags |= IFA_F_DADFAILED;
1516                 spin_unlock_bh(&ifp->lock);
1517                 if (dad_failed)
1518                         ipv6_ifa_notify(0, ifp);
1519                 in6_ifa_put(ifp);
1520 #ifdef CONFIG_IPV6_PRIVACY
1521         } else if (ifp->flags&IFA_F_TEMPORARY) {
1522                 struct inet6_ifaddr *ifpub;
1523                 spin_lock_bh(&ifp->lock);
1524                 ifpub = ifp->ifpub;
1525                 if (ifpub) {
1526                         in6_ifa_hold(ifpub);
1527                         spin_unlock_bh(&ifp->lock);
1528                         ipv6_create_tempaddr(ifpub, ifp);
1529                         in6_ifa_put(ifpub);
1530                 } else {
1531                         spin_unlock_bh(&ifp->lock);
1532                 }
1533                 ipv6_del_addr(ifp);
1534 #endif
1535         } else
1536                 ipv6_del_addr(ifp);
1537 }
1538
1539 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1540 {
1541         int err = -ENOENT;
1542
1543         spin_lock(&ifp->state_lock);
1544         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1545                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1546                 err = 0;
1547         }
1548         spin_unlock(&ifp->state_lock);
1549
1550         return err;
1551 }
1552
1553 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1554 {
1555         struct inet6_dev *idev = ifp->idev;
1556
1557         if (addrconf_dad_end(ifp)) {
1558                 in6_ifa_put(ifp);
1559                 return;
1560         }
1561
1562         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1563                              ifp->idev->dev->name, &ifp->addr);
1564
1565         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1566                 struct in6_addr addr;
1567
1568                 addr.s6_addr32[0] = htonl(0xfe800000);
1569                 addr.s6_addr32[1] = 0;
1570
1571                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1572                     ipv6_addr_equal(&ifp->addr, &addr)) {
1573                         /* DAD failed for link-local based on MAC address */
1574                         idev->cnf.disable_ipv6 = 1;
1575
1576                         pr_info("%s: IPv6 being disabled!\n",
1577                                 ifp->idev->dev->name);
1578                 }
1579         }
1580
1581         addrconf_dad_stop(ifp, 1);
1582 }
1583
1584 /* Join to solicited addr multicast group. */
1585
1586 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1587 {
1588         struct in6_addr maddr;
1589
1590         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1591                 return;
1592
1593         addrconf_addr_solict_mult(addr, &maddr);
1594         ipv6_dev_mc_inc(dev, &maddr);
1595 }
1596
1597 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1598 {
1599         struct in6_addr maddr;
1600
1601         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1602                 return;
1603
1604         addrconf_addr_solict_mult(addr, &maddr);
1605         __ipv6_dev_mc_dec(idev, &maddr);
1606 }
1607
1608 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1609 {
1610         struct in6_addr addr;
1611         if (ifp->prefix_len == 127) /* RFC 6164 */
1612                 return;
1613         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1614         if (ipv6_addr_any(&addr))
1615                 return;
1616         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1617 }
1618
1619 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1620 {
1621         struct in6_addr addr;
1622         if (ifp->prefix_len == 127) /* RFC 6164 */
1623                 return;
1624         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1625         if (ipv6_addr_any(&addr))
1626                 return;
1627         __ipv6_dev_ac_dec(ifp->idev, &addr);
1628 }
1629
1630 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1631 {
1632         if (dev->addr_len != ETH_ALEN)
1633                 return -1;
1634         memcpy(eui, dev->dev_addr, 3);
1635         memcpy(eui + 5, dev->dev_addr + 3, 3);
1636
1637         /*
1638          * The zSeries OSA network cards can be shared among various
1639          * OS instances, but the OSA cards have only one MAC address.
1640          * This leads to duplicate address conflicts in conjunction
1641          * with IPv6 if more than one instance uses the same card.
1642          *
1643          * The driver for these cards can deliver a unique 16-bit
1644          * identifier for each instance sharing the same card.  It is
1645          * placed instead of 0xFFFE in the interface identifier.  The
1646          * "u" bit of the interface identifier is not inverted in this
1647          * case.  Hence the resulting interface identifier has local
1648          * scope according to RFC2373.
1649          */
1650         if (dev->dev_id) {
1651                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1652                 eui[4] = dev->dev_id & 0xFF;
1653         } else {
1654                 eui[3] = 0xFF;
1655                 eui[4] = 0xFE;
1656                 eui[0] ^= 2;
1657         }
1658         return 0;
1659 }
1660
1661 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1662 {
1663         if (dev->addr_len != IEEE802154_ADDR_LEN)
1664                 return -1;
1665         memcpy(eui, dev->dev_addr, 8);
1666         return 0;
1667 }
1668
1669 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1670 {
1671         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1672         if (dev->addr_len != ARCNET_ALEN)
1673                 return -1;
1674         memset(eui, 0, 7);
1675         eui[7] = *(u8 *)dev->dev_addr;
1676         return 0;
1677 }
1678
1679 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1680 {
1681         if (dev->addr_len != INFINIBAND_ALEN)
1682                 return -1;
1683         memcpy(eui, dev->dev_addr + 12, 8);
1684         eui[0] |= 2;
1685         return 0;
1686 }
1687
1688 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1689 {
1690         if (addr == 0)
1691                 return -1;
1692         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1693                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1694                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1695                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1696                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1697                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1698         eui[1] = 0;
1699         eui[2] = 0x5E;
1700         eui[3] = 0xFE;
1701         memcpy(eui + 4, &addr, 4);
1702         return 0;
1703 }
1704
1705 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1706 {
1707         if (dev->priv_flags & IFF_ISATAP)
1708                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1709         return -1;
1710 }
1711
1712 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1713 {
1714         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1715 }
1716
1717 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1718 {
1719         switch (dev->type) {
1720         case ARPHRD_ETHER:
1721         case ARPHRD_FDDI:
1722                 return addrconf_ifid_eui48(eui, dev);
1723         case ARPHRD_ARCNET:
1724                 return addrconf_ifid_arcnet(eui, dev);
1725         case ARPHRD_INFINIBAND:
1726                 return addrconf_ifid_infiniband(eui, dev);
1727         case ARPHRD_SIT:
1728                 return addrconf_ifid_sit(eui, dev);
1729         case ARPHRD_IPGRE:
1730                 return addrconf_ifid_gre(eui, dev);
1731         case ARPHRD_IEEE802154:
1732                 return addrconf_ifid_eui64(eui, dev);
1733         }
1734         return -1;
1735 }
1736
1737 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1738 {
1739         int err = -1;
1740         struct inet6_ifaddr *ifp;
1741
1742         read_lock_bh(&idev->lock);
1743         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1744                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1745                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1746                         err = 0;
1747                         break;
1748                 }
1749         }
1750         read_unlock_bh(&idev->lock);
1751         return err;
1752 }
1753
1754 #ifdef CONFIG_IPV6_PRIVACY
1755 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1756 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1757 {
1758 regen:
1759         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1760         idev->rndid[0] &= ~0x02;
1761
1762         /*
1763          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1764          * check if generated address is not inappropriate
1765          *
1766          *  - Reserved subnet anycast (RFC 2526)
1767          *      11111101 11....11 1xxxxxxx
1768          *  - ISATAP (RFC4214) 6.1
1769          *      00-00-5E-FE-xx-xx-xx-xx
1770          *  - value 0
1771          *  - XXX: already assigned to an address on the device
1772          */
1773         if (idev->rndid[0] == 0xfd &&
1774             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1775             (idev->rndid[7]&0x80))
1776                 goto regen;
1777         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1778                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1779                         goto regen;
1780                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1781                         goto regen;
1782         }
1783 }
1784
1785 static void ipv6_regen_rndid(unsigned long data)
1786 {
1787         struct inet6_dev *idev = (struct inet6_dev *) data;
1788         unsigned long expires;
1789
1790         rcu_read_lock_bh();
1791         write_lock_bh(&idev->lock);
1792
1793         if (idev->dead)
1794                 goto out;
1795
1796         __ipv6_regen_rndid(idev);
1797
1798         expires = jiffies +
1799                 idev->cnf.temp_prefered_lft * HZ -
1800                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1801                 idev->cnf.max_desync_factor * HZ;
1802         if (time_before(expires, jiffies)) {
1803                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1804                         __func__, idev->dev->name);
1805                 goto out;
1806         }
1807
1808         if (!mod_timer(&idev->regen_timer, expires))
1809                 in6_dev_hold(idev);
1810
1811 out:
1812         write_unlock_bh(&idev->lock);
1813         rcu_read_unlock_bh();
1814         in6_dev_put(idev);
1815 }
1816
1817 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1818 {
1819         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1820                 __ipv6_regen_rndid(idev);
1821 }
1822 #endif
1823
1824 /*
1825  *      Add prefix route.
1826  */
1827
1828 static void
1829 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1830                       unsigned long expires, u32 flags)
1831 {
1832         struct fib6_config cfg = {
1833                 .fc_table = RT6_TABLE_PREFIX,
1834                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1835                 .fc_ifindex = dev->ifindex,
1836                 .fc_expires = expires,
1837                 .fc_dst_len = plen,
1838                 .fc_flags = RTF_UP | flags,
1839                 .fc_nlinfo.nl_net = dev_net(dev),
1840                 .fc_protocol = RTPROT_KERNEL,
1841         };
1842
1843         cfg.fc_dst = *pfx;
1844
1845         /* Prevent useless cloning on PtP SIT.
1846            This thing is done here expecting that the whole
1847            class of non-broadcast devices need not cloning.
1848          */
1849 #if IS_ENABLED(CONFIG_IPV6_SIT)
1850         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1851                 cfg.fc_flags |= RTF_NONEXTHOP;
1852 #endif
1853
1854         ip6_route_add(&cfg);
1855 }
1856
1857
1858 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1859                                                   int plen,
1860                                                   const struct net_device *dev,
1861                                                   u32 flags, u32 noflags)
1862 {
1863         struct fib6_node *fn;
1864         struct rt6_info *rt = NULL;
1865         struct fib6_table *table;
1866
1867         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1868         if (table == NULL)
1869                 return NULL;
1870
1871         read_lock_bh(&table->tb6_lock);
1872         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1873         if (!fn)
1874                 goto out;
1875         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1876                 if (rt->dst.dev->ifindex != dev->ifindex)
1877                         continue;
1878                 if ((rt->rt6i_flags & flags) != flags)
1879                         continue;
1880                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1881                         continue;
1882                 dst_hold(&rt->dst);
1883                 break;
1884         }
1885 out:
1886         read_unlock_bh(&table->tb6_lock);
1887         return rt;
1888 }
1889
1890
1891 /* Create "default" multicast route to the interface */
1892
1893 static void addrconf_add_mroute(struct net_device *dev)
1894 {
1895         struct fib6_config cfg = {
1896                 .fc_table = RT6_TABLE_LOCAL,
1897                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1898                 .fc_ifindex = dev->ifindex,
1899                 .fc_dst_len = 8,
1900                 .fc_flags = RTF_UP,
1901                 .fc_nlinfo.nl_net = dev_net(dev),
1902         };
1903
1904         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1905
1906         ip6_route_add(&cfg);
1907 }
1908
1909 #if IS_ENABLED(CONFIG_IPV6_SIT)
1910 static void sit_route_add(struct net_device *dev)
1911 {
1912         struct fib6_config cfg = {
1913                 .fc_table = RT6_TABLE_MAIN,
1914                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1915                 .fc_ifindex = dev->ifindex,
1916                 .fc_dst_len = 96,
1917                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1918                 .fc_nlinfo.nl_net = dev_net(dev),
1919         };
1920
1921         /* prefix length - 96 bits "::d.d.d.d" */
1922         ip6_route_add(&cfg);
1923 }
1924 #endif
1925
1926 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1927 {
1928         struct inet6_dev *idev;
1929
1930         ASSERT_RTNL();
1931
1932         idev = ipv6_find_idev(dev);
1933         if (!idev)
1934                 return ERR_PTR(-ENOBUFS);
1935
1936         if (idev->cnf.disable_ipv6)
1937                 return ERR_PTR(-EACCES);
1938
1939         /* Add default multicast route */
1940         if (!(dev->flags & IFF_LOOPBACK))
1941                 addrconf_add_mroute(dev);
1942
1943         return idev;
1944 }
1945
1946 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1947 {
1948         struct prefix_info *pinfo;
1949         __u32 valid_lft;
1950         __u32 prefered_lft;
1951         int addr_type;
1952         struct inet6_dev *in6_dev;
1953         struct net *net = dev_net(dev);
1954
1955         pinfo = (struct prefix_info *) opt;
1956
1957         if (len < sizeof(struct prefix_info)) {
1958                 ADBG(("addrconf: prefix option too short\n"));
1959                 return;
1960         }
1961
1962         /*
1963          *      Validation checks ([ADDRCONF], page 19)
1964          */
1965
1966         addr_type = ipv6_addr_type(&pinfo->prefix);
1967
1968         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1969                 return;
1970
1971         valid_lft = ntohl(pinfo->valid);
1972         prefered_lft = ntohl(pinfo->prefered);
1973
1974         if (prefered_lft > valid_lft) {
1975                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1976                 return;
1977         }
1978
1979         in6_dev = in6_dev_get(dev);
1980
1981         if (in6_dev == NULL) {
1982                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1983                                     dev->name);
1984                 return;
1985         }
1986
1987         /*
1988          *      Two things going on here:
1989          *      1) Add routes for on-link prefixes
1990          *      2) Configure prefixes with the auto flag set
1991          */
1992
1993         if (pinfo->onlink) {
1994                 struct rt6_info *rt;
1995                 unsigned long rt_expires;
1996
1997                 /* Avoid arithmetic overflow. Really, we could
1998                  * save rt_expires in seconds, likely valid_lft,
1999                  * but it would require division in fib gc, that it
2000                  * not good.
2001                  */
2002                 if (HZ > USER_HZ)
2003                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2004                 else
2005                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2006
2007                 if (addrconf_finite_timeout(rt_expires))
2008                         rt_expires *= HZ;
2009
2010                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2011                                                pinfo->prefix_len,
2012                                                dev,
2013                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2014                                                RTF_GATEWAY | RTF_DEFAULT);
2015
2016                 if (rt) {
2017                         /* Autoconf prefix route */
2018                         if (valid_lft == 0) {
2019                                 ip6_del_rt(rt);
2020                                 rt = NULL;
2021                         } else if (addrconf_finite_timeout(rt_expires)) {
2022                                 /* not infinity */
2023                                 rt6_set_expires(rt, jiffies + rt_expires);
2024                         } else {
2025                                 rt6_clean_expires(rt);
2026                         }
2027                 } else if (valid_lft) {
2028                         clock_t expires = 0;
2029                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2030                         if (addrconf_finite_timeout(rt_expires)) {
2031                                 /* not infinity */
2032                                 flags |= RTF_EXPIRES;
2033                                 expires = jiffies_to_clock_t(rt_expires);
2034                         }
2035                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2036                                               dev, expires, flags);
2037                 }
2038                 ip6_rt_put(rt);
2039         }
2040
2041         /* Try to figure out our local address for this prefix */
2042
2043         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2044                 struct inet6_ifaddr *ifp;
2045                 struct in6_addr addr;
2046                 int create = 0, update_lft = 0;
2047
2048                 if (pinfo->prefix_len == 64) {
2049                         memcpy(&addr, &pinfo->prefix, 8);
2050                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2051                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2052                                 in6_dev_put(in6_dev);
2053                                 return;
2054                         }
2055                         goto ok;
2056                 }
2057                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2058                                     pinfo->prefix_len);
2059                 in6_dev_put(in6_dev);
2060                 return;
2061
2062 ok:
2063
2064                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2065
2066                 if (ifp == NULL && valid_lft) {
2067                         int max_addresses = in6_dev->cnf.max_addresses;
2068                         u32 addr_flags = 0;
2069
2070 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2071                         if (in6_dev->cnf.optimistic_dad &&
2072                             !net->ipv6.devconf_all->forwarding && sllao)
2073                                 addr_flags = IFA_F_OPTIMISTIC;
2074 #endif
2075
2076                         /* Do not allow to create too much of autoconfigured
2077                          * addresses; this would be too easy way to crash kernel.
2078                          */
2079                         if (!max_addresses ||
2080                             ipv6_count_addresses(in6_dev) < max_addresses)
2081                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2082                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2083                                                     addr_flags);
2084
2085                         if (!ifp || IS_ERR(ifp)) {
2086                                 in6_dev_put(in6_dev);
2087                                 return;
2088                         }
2089
2090                         update_lft = create = 1;
2091                         ifp->cstamp = jiffies;
2092                         addrconf_dad_start(ifp);
2093                 }
2094
2095                 if (ifp) {
2096                         int flags;
2097                         unsigned long now;
2098 #ifdef CONFIG_IPV6_PRIVACY
2099                         struct inet6_ifaddr *ift;
2100 #endif
2101                         u32 stored_lft;
2102
2103                         /* update lifetime (RFC2462 5.5.3 e) */
2104                         spin_lock(&ifp->lock);
2105                         now = jiffies;
2106                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2107                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2108                         else
2109                                 stored_lft = 0;
2110                         if (!update_lft && stored_lft) {
2111                                 if (valid_lft > MIN_VALID_LIFETIME ||
2112                                     valid_lft > stored_lft)
2113                                         update_lft = 1;
2114                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2115                                         /* valid_lft <= stored_lft is always true */
2116                                         /*
2117                                          * RFC 4862 Section 5.5.3e:
2118                                          * "Note that the preferred lifetime of
2119                                          *  the corresponding address is always
2120                                          *  reset to the Preferred Lifetime in
2121                                          *  the received Prefix Information
2122                                          *  option, regardless of whether the
2123                                          *  valid lifetime is also reset or
2124                                          *  ignored."
2125                                          *
2126                                          *  So if the preferred lifetime in
2127                                          *  this advertisement is different
2128                                          *  than what we have stored, but the
2129                                          *  valid lifetime is invalid, just
2130                                          *  reset prefered_lft.
2131                                          *
2132                                          *  We must set the valid lifetime
2133                                          *  to the stored lifetime since we'll
2134                                          *  be updating the timestamp below,
2135                                          *  else we'll set it back to the
2136                                          *  minimum.
2137                                          */
2138                                         if (prefered_lft != ifp->prefered_lft) {
2139                                                 valid_lft = stored_lft;
2140                                                 update_lft = 1;
2141                                         }
2142                                 } else {
2143                                         valid_lft = MIN_VALID_LIFETIME;
2144                                         if (valid_lft < prefered_lft)
2145                                                 prefered_lft = valid_lft;
2146                                         update_lft = 1;
2147                                 }
2148                         }
2149
2150                         if (update_lft) {
2151                                 ifp->valid_lft = valid_lft;
2152                                 ifp->prefered_lft = prefered_lft;
2153                                 ifp->tstamp = now;
2154                                 flags = ifp->flags;
2155                                 ifp->flags &= ~IFA_F_DEPRECATED;
2156                                 spin_unlock(&ifp->lock);
2157
2158                                 if (!(flags&IFA_F_TENTATIVE))
2159                                         ipv6_ifa_notify(0, ifp);
2160                         } else
2161                                 spin_unlock(&ifp->lock);
2162
2163 #ifdef CONFIG_IPV6_PRIVACY
2164                         read_lock_bh(&in6_dev->lock);
2165                         /* update all temporary addresses in the list */
2166                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2167                                             tmp_list) {
2168                                 int age, max_valid, max_prefered;
2169
2170                                 if (ifp != ift->ifpub)
2171                                         continue;
2172
2173                                 /*
2174                                  * RFC 4941 section 3.3:
2175                                  * If a received option will extend the lifetime
2176                                  * of a public address, the lifetimes of
2177                                  * temporary addresses should be extended,
2178                                  * subject to the overall constraint that no
2179                                  * temporary addresses should ever remain
2180                                  * "valid" or "preferred" for a time longer than
2181                                  * (TEMP_VALID_LIFETIME) or
2182                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2183                                  * respectively.
2184                                  */
2185                                 age = (now - ift->cstamp) / HZ;
2186                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2187                                 if (max_valid < 0)
2188                                         max_valid = 0;
2189
2190                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2191                                                in6_dev->cnf.max_desync_factor -
2192                                                age;
2193                                 if (max_prefered < 0)
2194                                         max_prefered = 0;
2195
2196                                 if (valid_lft > max_valid)
2197                                         valid_lft = max_valid;
2198
2199                                 if (prefered_lft > max_prefered)
2200                                         prefered_lft = max_prefered;
2201
2202                                 spin_lock(&ift->lock);
2203                                 flags = ift->flags;
2204                                 ift->valid_lft = valid_lft;
2205                                 ift->prefered_lft = prefered_lft;
2206                                 ift->tstamp = now;
2207                                 if (prefered_lft > 0)
2208                                         ift->flags &= ~IFA_F_DEPRECATED;
2209
2210                                 spin_unlock(&ift->lock);
2211                                 if (!(flags&IFA_F_TENTATIVE))
2212                                         ipv6_ifa_notify(0, ift);
2213                         }
2214
2215                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2216                                 /*
2217                                  * When a new public address is created as
2218                                  * described in [ADDRCONF], also create a new
2219                                  * temporary address. Also create a temporary
2220                                  * address if it's enabled but no temporary
2221                                  * address currently exists.
2222                                  */
2223                                 read_unlock_bh(&in6_dev->lock);
2224                                 ipv6_create_tempaddr(ifp, NULL);
2225                         } else {
2226                                 read_unlock_bh(&in6_dev->lock);
2227                         }
2228 #endif
2229                         in6_ifa_put(ifp);
2230                         addrconf_verify(0);
2231                 }
2232         }
2233         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2234         in6_dev_put(in6_dev);
2235 }
2236
2237 /*
2238  *      Set destination address.
2239  *      Special case for SIT interfaces where we create a new "virtual"
2240  *      device.
2241  */
2242 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2243 {
2244         struct in6_ifreq ireq;
2245         struct net_device *dev;
2246         int err = -EINVAL;
2247
2248         rtnl_lock();
2249
2250         err = -EFAULT;
2251         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2252                 goto err_exit;
2253
2254         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2255
2256         err = -ENODEV;
2257         if (dev == NULL)
2258                 goto err_exit;
2259
2260 #if IS_ENABLED(CONFIG_IPV6_SIT)
2261         if (dev->type == ARPHRD_SIT) {
2262                 const struct net_device_ops *ops = dev->netdev_ops;
2263                 struct ifreq ifr;
2264                 struct ip_tunnel_parm p;
2265
2266                 err = -EADDRNOTAVAIL;
2267                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2268                         goto err_exit;
2269
2270                 memset(&p, 0, sizeof(p));
2271                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2272                 p.iph.saddr = 0;
2273                 p.iph.version = 4;
2274                 p.iph.ihl = 5;
2275                 p.iph.protocol = IPPROTO_IPV6;
2276                 p.iph.ttl = 64;
2277                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2278
2279                 if (ops->ndo_do_ioctl) {
2280                         mm_segment_t oldfs = get_fs();
2281
2282                         set_fs(KERNEL_DS);
2283                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2284                         set_fs(oldfs);
2285                 } else
2286                         err = -EOPNOTSUPP;
2287
2288                 if (err == 0) {
2289                         err = -ENOBUFS;
2290                         dev = __dev_get_by_name(net, p.name);
2291                         if (!dev)
2292                                 goto err_exit;
2293                         err = dev_open(dev);
2294                 }
2295         }
2296 #endif
2297
2298 err_exit:
2299         rtnl_unlock();
2300         return err;
2301 }
2302
2303 /*
2304  *      Manual configuration of address on an interface
2305  */
2306 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2307                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2308                           __u32 valid_lft)
2309 {
2310         struct inet6_ifaddr *ifp;
2311         struct inet6_dev *idev;
2312         struct net_device *dev;
2313         int scope;
2314         u32 flags;
2315         clock_t expires;
2316         unsigned long timeout;
2317
2318         ASSERT_RTNL();
2319
2320         if (plen > 128)
2321                 return -EINVAL;
2322
2323         /* check the lifetime */
2324         if (!valid_lft || prefered_lft > valid_lft)
2325                 return -EINVAL;
2326
2327         dev = __dev_get_by_index(net, ifindex);
2328         if (!dev)
2329                 return -ENODEV;
2330
2331         idev = addrconf_add_dev(dev);
2332         if (IS_ERR(idev))
2333                 return PTR_ERR(idev);
2334
2335         scope = ipv6_addr_scope(pfx);
2336
2337         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2338         if (addrconf_finite_timeout(timeout)) {
2339                 expires = jiffies_to_clock_t(timeout * HZ);
2340                 valid_lft = timeout;
2341                 flags = RTF_EXPIRES;
2342         } else {
2343                 expires = 0;
2344                 flags = 0;
2345                 ifa_flags |= IFA_F_PERMANENT;
2346         }
2347
2348         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2349         if (addrconf_finite_timeout(timeout)) {
2350                 if (timeout == 0)
2351                         ifa_flags |= IFA_F_DEPRECATED;
2352                 prefered_lft = timeout;
2353         }
2354
2355         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2356
2357         if (!IS_ERR(ifp)) {
2358                 spin_lock_bh(&ifp->lock);
2359                 ifp->valid_lft = valid_lft;
2360                 ifp->prefered_lft = prefered_lft;
2361                 ifp->tstamp = jiffies;
2362                 spin_unlock_bh(&ifp->lock);
2363
2364                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2365                                       expires, flags);
2366                 /*
2367                  * Note that section 3.1 of RFC 4429 indicates
2368                  * that the Optimistic flag should not be set for
2369                  * manually configured addresses
2370                  */
2371                 addrconf_dad_start(ifp);
2372                 in6_ifa_put(ifp);
2373                 addrconf_verify(0);
2374                 return 0;
2375         }
2376
2377         return PTR_ERR(ifp);
2378 }
2379
2380 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2381                           unsigned int plen)
2382 {
2383         struct inet6_ifaddr *ifp;
2384         struct inet6_dev *idev;
2385         struct net_device *dev;
2386
2387         if (plen > 128)
2388                 return -EINVAL;
2389
2390         dev = __dev_get_by_index(net, ifindex);
2391         if (!dev)
2392                 return -ENODEV;
2393
2394         if ((idev = __in6_dev_get(dev)) == NULL)
2395                 return -ENXIO;
2396
2397         read_lock_bh(&idev->lock);
2398         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2399                 if (ifp->prefix_len == plen &&
2400                     ipv6_addr_equal(pfx, &ifp->addr)) {
2401                         in6_ifa_hold(ifp);
2402                         read_unlock_bh(&idev->lock);
2403
2404                         ipv6_del_addr(ifp);
2405
2406                         /* If the last address is deleted administratively,
2407                            disable IPv6 on this interface.
2408                          */
2409                         if (list_empty(&idev->addr_list))
2410                                 addrconf_ifdown(idev->dev, 1);
2411                         return 0;
2412                 }
2413         }
2414         read_unlock_bh(&idev->lock);
2415         return -EADDRNOTAVAIL;
2416 }
2417
2418
2419 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2420 {
2421         struct in6_ifreq ireq;
2422         int err;
2423
2424         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2425                 return -EPERM;
2426
2427         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2428                 return -EFAULT;
2429
2430         rtnl_lock();
2431         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2432                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2433                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2434         rtnl_unlock();
2435         return err;
2436 }
2437
2438 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2439 {
2440         struct in6_ifreq ireq;
2441         int err;
2442
2443         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2444                 return -EPERM;
2445
2446         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2447                 return -EFAULT;
2448
2449         rtnl_lock();
2450         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2451                              ireq.ifr6_prefixlen);
2452         rtnl_unlock();
2453         return err;
2454 }
2455
2456 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2457                      int plen, int scope)
2458 {
2459         struct inet6_ifaddr *ifp;
2460
2461         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2462         if (!IS_ERR(ifp)) {
2463                 spin_lock_bh(&ifp->lock);
2464                 ifp->flags &= ~IFA_F_TENTATIVE;
2465                 spin_unlock_bh(&ifp->lock);
2466                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2467                 in6_ifa_put(ifp);
2468         }
2469 }
2470
2471 #if IS_ENABLED(CONFIG_IPV6_SIT)
2472 static void sit_add_v4_addrs(struct inet6_dev *idev)
2473 {
2474         struct in6_addr addr;
2475         struct net_device *dev;
2476         struct net *net = dev_net(idev->dev);
2477         int scope;
2478
2479         ASSERT_RTNL();
2480
2481         memset(&addr, 0, sizeof(struct in6_addr));
2482         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2483
2484         if (idev->dev->flags&IFF_POINTOPOINT) {
2485                 addr.s6_addr32[0] = htonl(0xfe800000);
2486                 scope = IFA_LINK;
2487         } else {
2488                 scope = IPV6_ADDR_COMPATv4;
2489         }
2490
2491         if (addr.s6_addr32[3]) {
2492                 add_addr(idev, &addr, 128, scope);
2493                 return;
2494         }
2495
2496         for_each_netdev(net, dev) {
2497                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2498                 if (in_dev && (dev->flags & IFF_UP)) {
2499                         struct in_ifaddr *ifa;
2500
2501                         int flag = scope;
2502
2503                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2504                                 int plen;
2505
2506                                 addr.s6_addr32[3] = ifa->ifa_local;
2507
2508                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2509                                         continue;
2510                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2511                                         if (idev->dev->flags&IFF_POINTOPOINT)
2512                                                 continue;
2513                                         flag |= IFA_HOST;
2514                                 }
2515                                 if (idev->dev->flags&IFF_POINTOPOINT)
2516                                         plen = 64;
2517                                 else
2518                                         plen = 96;
2519
2520                                 add_addr(idev, &addr, plen, flag);
2521                         }
2522                 }
2523         }
2524 }
2525 #endif
2526
2527 static void init_loopback(struct net_device *dev)
2528 {
2529         struct inet6_dev  *idev;
2530
2531         /* ::1 */
2532
2533         ASSERT_RTNL();
2534
2535         if ((idev = ipv6_find_idev(dev)) == NULL) {
2536                 pr_debug("%s: add_dev failed\n", __func__);
2537                 return;
2538         }
2539
2540         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2541 }
2542
2543 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2544 {
2545         struct inet6_ifaddr *ifp;
2546         u32 addr_flags = IFA_F_PERMANENT;
2547
2548 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2549         if (idev->cnf.optimistic_dad &&
2550             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2551                 addr_flags |= IFA_F_OPTIMISTIC;
2552 #endif
2553
2554
2555         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2556         if (!IS_ERR(ifp)) {
2557                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2558                 addrconf_dad_start(ifp);
2559                 in6_ifa_put(ifp);
2560         }
2561 }
2562
2563 static void addrconf_dev_config(struct net_device *dev)
2564 {
2565         struct in6_addr addr;
2566         struct inet6_dev *idev;
2567
2568         ASSERT_RTNL();
2569
2570         if ((dev->type != ARPHRD_ETHER) &&
2571             (dev->type != ARPHRD_FDDI) &&
2572             (dev->type != ARPHRD_ARCNET) &&
2573             (dev->type != ARPHRD_INFINIBAND) &&
2574             (dev->type != ARPHRD_IEEE802154)) {
2575                 /* Alas, we support only Ethernet autoconfiguration. */
2576                 return;
2577         }
2578
2579         idev = addrconf_add_dev(dev);
2580         if (IS_ERR(idev))
2581                 return;
2582
2583         memset(&addr, 0, sizeof(struct in6_addr));
2584         addr.s6_addr32[0] = htonl(0xFE800000);
2585
2586         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2587                 addrconf_add_linklocal(idev, &addr);
2588 }
2589
2590 #if IS_ENABLED(CONFIG_IPV6_SIT)
2591 static void addrconf_sit_config(struct net_device *dev)
2592 {
2593         struct inet6_dev *idev;
2594
2595         ASSERT_RTNL();
2596
2597         /*
2598          * Configure the tunnel with one of our IPv4
2599          * addresses... we should configure all of
2600          * our v4 addrs in the tunnel
2601          */
2602
2603         if ((idev = ipv6_find_idev(dev)) == NULL) {
2604                 pr_debug("%s: add_dev failed\n", __func__);
2605                 return;
2606         }
2607
2608         if (dev->priv_flags & IFF_ISATAP) {
2609                 struct in6_addr addr;
2610
2611                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2612                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2613                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2614                         addrconf_add_linklocal(idev, &addr);
2615                 return;
2616         }
2617
2618         sit_add_v4_addrs(idev);
2619
2620         if (dev->flags&IFF_POINTOPOINT)
2621                 addrconf_add_mroute(dev);
2622         else
2623                 sit_route_add(dev);
2624 }
2625 #endif
2626
2627 #if IS_ENABLED(CONFIG_NET_IPGRE)
2628 static void addrconf_gre_config(struct net_device *dev)
2629 {
2630         struct inet6_dev *idev;
2631         struct in6_addr addr;
2632
2633         pr_info("%s(%s)\n", __func__, dev->name);
2634
2635         ASSERT_RTNL();
2636
2637         if ((idev = ipv6_find_idev(dev)) == NULL) {
2638                 pr_debug("%s: add_dev failed\n", __func__);
2639                 return;
2640         }
2641
2642         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2643         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2644
2645         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2646                 addrconf_add_linklocal(idev, &addr);
2647 }
2648 #endif
2649
2650 static inline int
2651 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2652 {
2653         struct in6_addr lladdr;
2654
2655         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2656                 addrconf_add_linklocal(idev, &lladdr);
2657                 return 0;
2658         }
2659         return -1;
2660 }
2661
2662 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2663 {
2664         struct net_device *link_dev;
2665         struct net *net = dev_net(idev->dev);
2666
2667         /* first try to inherit the link-local address from the link device */
2668         if (idev->dev->iflink &&
2669             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2670                 if (!ipv6_inherit_linklocal(idev, link_dev))
2671                         return;
2672         }
2673         /* then try to inherit it from any device */
2674         for_each_netdev(net, link_dev) {
2675                 if (!ipv6_inherit_linklocal(idev, link_dev))
2676                         return;
2677         }
2678         pr_debug("init ip6-ip6: add_linklocal failed\n");
2679 }
2680
2681 /*
2682  * Autoconfigure tunnel with a link-local address so routing protocols,
2683  * DHCPv6, MLD etc. can be run over the virtual link
2684  */
2685
2686 static void addrconf_ip6_tnl_config(struct net_device *dev)
2687 {
2688         struct inet6_dev *idev;
2689
2690         ASSERT_RTNL();
2691
2692         idev = addrconf_add_dev(dev);
2693         if (IS_ERR(idev)) {
2694                 pr_debug("init ip6-ip6: add_dev failed\n");
2695                 return;
2696         }
2697         ip6_tnl_add_linklocal(idev);
2698 }
2699
2700 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2701                            void *data)
2702 {
2703         struct net_device *dev = (struct net_device *) data;
2704         struct inet6_dev *idev = __in6_dev_get(dev);
2705         int run_pending = 0;
2706         int err;
2707
2708         switch (event) {
2709         case NETDEV_REGISTER:
2710                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2711                         idev = ipv6_add_dev(dev);
2712                         if (!idev)
2713                                 return notifier_from_errno(-ENOMEM);
2714                 }
2715                 break;
2716
2717         case NETDEV_UP:
2718         case NETDEV_CHANGE:
2719                 if (dev->flags & IFF_SLAVE)
2720                         break;
2721
2722                 if (event == NETDEV_UP) {
2723                         if (!addrconf_qdisc_ok(dev)) {
2724                                 /* device is not ready yet. */
2725                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2726                                         dev->name);
2727                                 break;
2728                         }
2729
2730                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2731                                 idev = ipv6_add_dev(dev);
2732
2733                         if (idev) {
2734                                 idev->if_flags |= IF_READY;
2735                                 run_pending = 1;
2736                         }
2737                 } else {
2738                         if (!addrconf_qdisc_ok(dev)) {
2739                                 /* device is still not ready. */
2740                                 break;
2741                         }
2742
2743                         if (idev) {
2744                                 if (idev->if_flags & IF_READY)
2745                                         /* device is already configured. */
2746                                         break;
2747                                 idev->if_flags |= IF_READY;
2748                         }
2749
2750                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2751                                 dev->name);
2752
2753                         run_pending = 1;
2754                 }
2755
2756                 switch (dev->type) {
2757 #if IS_ENABLED(CONFIG_IPV6_SIT)
2758                 case ARPHRD_SIT:
2759                         addrconf_sit_config(dev);
2760                         break;
2761 #endif
2762 #if IS_ENABLED(CONFIG_NET_IPGRE)
2763                 case ARPHRD_IPGRE:
2764                         addrconf_gre_config(dev);
2765                         break;
2766 #endif
2767                 case ARPHRD_TUNNEL6:
2768                         addrconf_ip6_tnl_config(dev);
2769                         break;
2770                 case ARPHRD_LOOPBACK:
2771                         init_loopback(dev);
2772                         break;
2773
2774                 default:
2775                         addrconf_dev_config(dev);
2776                         break;
2777                 }
2778
2779                 if (idev) {
2780                         if (run_pending)
2781                                 addrconf_dad_run(idev);
2782
2783                         /*
2784                          * If the MTU changed during the interface down,
2785                          * when the interface up, the changed MTU must be
2786                          * reflected in the idev as well as routers.
2787                          */
2788                         if (idev->cnf.mtu6 != dev->mtu &&
2789                             dev->mtu >= IPV6_MIN_MTU) {
2790                                 rt6_mtu_change(dev, dev->mtu);
2791                                 idev->cnf.mtu6 = dev->mtu;
2792                         }
2793                         idev->tstamp = jiffies;
2794                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2795
2796                         /*
2797                          * If the changed mtu during down is lower than
2798                          * IPV6_MIN_MTU stop IPv6 on this interface.
2799                          */
2800                         if (dev->mtu < IPV6_MIN_MTU)
2801                                 addrconf_ifdown(dev, 1);
2802                 }
2803                 break;
2804
2805         case NETDEV_CHANGEMTU:
2806                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2807                         rt6_mtu_change(dev, dev->mtu);
2808                         idev->cnf.mtu6 = dev->mtu;
2809                         break;
2810                 }
2811
2812                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2813                         idev = ipv6_add_dev(dev);
2814                         if (idev)
2815                                 break;
2816                 }
2817
2818                 /*
2819                  * MTU falled under IPV6_MIN_MTU.
2820                  * Stop IPv6 on this interface.
2821                  */
2822
2823         case NETDEV_DOWN:
2824         case NETDEV_UNREGISTER:
2825                 /*
2826                  *      Remove all addresses from this interface.
2827                  */
2828                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2829                 break;
2830
2831         case NETDEV_CHANGENAME:
2832                 if (idev) {
2833                         snmp6_unregister_dev(idev);
2834                         addrconf_sysctl_unregister(idev);
2835                         addrconf_sysctl_register(idev);
2836                         err = snmp6_register_dev(idev);
2837                         if (err)
2838                                 return notifier_from_errno(err);
2839                 }
2840                 break;
2841
2842         case NETDEV_PRE_TYPE_CHANGE:
2843         case NETDEV_POST_TYPE_CHANGE:
2844                 addrconf_type_change(dev, event);
2845                 break;
2846         }
2847
2848         return NOTIFY_OK;
2849 }
2850
2851 /*
2852  *      addrconf module should be notified of a device going up
2853  */
2854 static struct notifier_block ipv6_dev_notf = {
2855         .notifier_call = addrconf_notify,
2856 };
2857
2858 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2859 {
2860         struct inet6_dev *idev;
2861         ASSERT_RTNL();
2862
2863         idev = __in6_dev_get(dev);
2864
2865         if (event == NETDEV_POST_TYPE_CHANGE)
2866                 ipv6_mc_remap(idev);
2867         else if (event == NETDEV_PRE_TYPE_CHANGE)
2868                 ipv6_mc_unmap(idev);
2869 }
2870
2871 static int addrconf_ifdown(struct net_device *dev, int how)
2872 {
2873         struct net *net = dev_net(dev);
2874         struct inet6_dev *idev;
2875         struct inet6_ifaddr *ifa;
2876         int state, i;
2877
2878         ASSERT_RTNL();
2879
2880         rt6_ifdown(net, dev);
2881         neigh_ifdown(&nd_tbl, dev);
2882
2883         idev = __in6_dev_get(dev);
2884         if (idev == NULL)
2885                 return -ENODEV;
2886
2887         /*
2888          * Step 1: remove reference to ipv6 device from parent device.
2889          *         Do not dev_put!
2890          */
2891         if (how) {
2892                 idev->dead = 1;
2893
2894                 /* protected by rtnl_lock */
2895                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2896
2897                 /* Step 1.5: remove snmp6 entry */
2898                 snmp6_unregister_dev(idev);
2899
2900         }
2901
2902         /* Step 2: clear hash table */
2903         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2904                 struct hlist_head *h = &inet6_addr_lst[i];
2905                 struct hlist_node *n;
2906
2907                 spin_lock_bh(&addrconf_hash_lock);
2908         restart:
2909                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2910                         if (ifa->idev == idev) {
2911                                 hlist_del_init_rcu(&ifa->addr_lst);
2912                                 addrconf_del_timer(ifa);
2913                                 goto restart;
2914                         }
2915                 }
2916                 spin_unlock_bh(&addrconf_hash_lock);
2917         }
2918
2919         write_lock_bh(&idev->lock);
2920
2921         /* Step 2: clear flags for stateless addrconf */
2922         if (!how)
2923                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2924
2925 #ifdef CONFIG_IPV6_PRIVACY
2926         if (how && del_timer(&idev->regen_timer))
2927                 in6_dev_put(idev);
2928
2929         /* Step 3: clear tempaddr list */
2930         while (!list_empty(&idev->tempaddr_list)) {
2931                 ifa = list_first_entry(&idev->tempaddr_list,
2932                                        struct inet6_ifaddr, tmp_list);
2933                 list_del(&ifa->tmp_list);
2934                 write_unlock_bh(&idev->lock);
2935                 spin_lock_bh(&ifa->lock);
2936
2937                 if (ifa->ifpub) {
2938                         in6_ifa_put(ifa->ifpub);
2939                         ifa->ifpub = NULL;
2940                 }
2941                 spin_unlock_bh(&ifa->lock);
2942                 in6_ifa_put(ifa);
2943                 write_lock_bh(&idev->lock);
2944         }
2945 #endif
2946
2947         while (!list_empty(&idev->addr_list)) {
2948                 ifa = list_first_entry(&idev->addr_list,
2949                                        struct inet6_ifaddr, if_list);
2950                 addrconf_del_timer(ifa);
2951
2952                 list_del(&ifa->if_list);
2953
2954                 write_unlock_bh(&idev->lock);
2955
2956                 spin_lock_bh(&ifa->state_lock);
2957                 state = ifa->state;
2958                 ifa->state = INET6_IFADDR_STATE_DEAD;
2959                 spin_unlock_bh(&ifa->state_lock);
2960
2961                 if (state != INET6_IFADDR_STATE_DEAD) {
2962                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2963                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2964                 }
2965                 in6_ifa_put(ifa);
2966
2967                 write_lock_bh(&idev->lock);
2968         }
2969
2970         write_unlock_bh(&idev->lock);
2971
2972         /* Step 5: Discard multicast list */
2973         if (how)
2974                 ipv6_mc_destroy_dev(idev);
2975         else
2976                 ipv6_mc_down(idev);
2977
2978         idev->tstamp = jiffies;
2979
2980         /* Last: Shot the device (if unregistered) */
2981         if (how) {
2982                 addrconf_sysctl_unregister(idev);
2983                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2984                 neigh_ifdown(&nd_tbl, dev);
2985                 in6_dev_put(idev);
2986         }
2987         return 0;
2988 }
2989
2990 static void addrconf_rs_timer(unsigned long data)
2991 {
2992         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2993         struct inet6_dev *idev = ifp->idev;
2994
2995         read_lock(&idev->lock);
2996         if (idev->dead || !(idev->if_flags & IF_READY))
2997                 goto out;
2998
2999         if (!ipv6_accept_ra(idev))
3000                 goto out;
3001
3002         /* Announcement received after solicitation was sent */
3003         if (idev->if_flags & IF_RA_RCVD)
3004                 goto out;
3005
3006         spin_lock(&ifp->lock);
3007         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3008                 /* The wait after the last probe can be shorter */
3009                 addrconf_mod_timer(ifp, AC_RS,
3010                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3011                                    idev->cnf.rtr_solicit_delay :
3012                                    idev->cnf.rtr_solicit_interval);
3013                 spin_unlock(&ifp->lock);
3014
3015                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3016         } else {
3017                 spin_unlock(&ifp->lock);
3018                 /*
3019                  * Note: we do not support deprecated "all on-link"
3020                  * assumption any longer.
3021                  */
3022                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3023         }
3024
3025 out:
3026         read_unlock(&idev->lock);
3027         in6_ifa_put(ifp);
3028 }
3029
3030 /*
3031  *      Duplicate Address Detection
3032  */
3033 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3034 {
3035         unsigned long rand_num;
3036         struct inet6_dev *idev = ifp->idev;
3037
3038         if (ifp->flags & IFA_F_OPTIMISTIC)
3039                 rand_num = 0;
3040         else
3041                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3042
3043         ifp->probes = idev->cnf.dad_transmits;
3044         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3045 }
3046
3047 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3048 {
3049         struct inet6_dev *idev = ifp->idev;
3050         struct net_device *dev = idev->dev;
3051
3052         addrconf_join_solict(dev, &ifp->addr);
3053
3054         net_srandom(ifp->addr.s6_addr32[3]);
3055
3056         read_lock_bh(&idev->lock);
3057         spin_lock(&ifp->lock);
3058         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3059                 goto out;
3060
3061         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3062             idev->cnf.accept_dad < 1 ||
3063             !(ifp->flags&IFA_F_TENTATIVE) ||
3064             ifp->flags & IFA_F_NODAD) {
3065                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3066                 spin_unlock(&ifp->lock);
3067                 read_unlock_bh(&idev->lock);
3068
3069                 addrconf_dad_completed(ifp);
3070                 return;
3071         }
3072
3073         if (!(idev->if_flags & IF_READY)) {
3074                 spin_unlock(&ifp->lock);
3075                 read_unlock_bh(&idev->lock);
3076                 /*
3077                  * If the device is not ready:
3078                  * - keep it tentative if it is a permanent address.
3079                  * - otherwise, kill it.
3080                  */
3081                 in6_ifa_hold(ifp);
3082                 addrconf_dad_stop(ifp, 0);
3083                 return;
3084         }
3085
3086         /*
3087          * Optimistic nodes can start receiving
3088          * Frames right away
3089          */
3090         if (ifp->flags & IFA_F_OPTIMISTIC)
3091                 ip6_ins_rt(ifp->rt);
3092
3093         addrconf_dad_kick(ifp);
3094 out:
3095         spin_unlock(&ifp->lock);
3096         read_unlock_bh(&idev->lock);
3097 }
3098
3099 static void addrconf_dad_timer(unsigned long data)
3100 {
3101         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3102         struct inet6_dev *idev = ifp->idev;
3103         struct in6_addr mcaddr;
3104
3105         if (!ifp->probes && addrconf_dad_end(ifp))
3106                 goto out;
3107
3108         read_lock(&idev->lock);
3109         if (idev->dead || !(idev->if_flags & IF_READY)) {
3110                 read_unlock(&idev->lock);
3111                 goto out;
3112         }
3113
3114         spin_lock(&ifp->lock);
3115         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3116                 spin_unlock(&ifp->lock);
3117                 read_unlock(&idev->lock);
3118                 goto out;
3119         }
3120
3121         if (ifp->probes == 0) {
3122                 /*
3123                  * DAD was successful
3124                  */
3125
3126                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3127                 spin_unlock(&ifp->lock);
3128                 read_unlock(&idev->lock);
3129
3130                 addrconf_dad_completed(ifp);
3131
3132                 goto out;
3133         }
3134
3135         ifp->probes--;
3136         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3137         spin_unlock(&ifp->lock);
3138         read_unlock(&idev->lock);
3139
3140         /* send a neighbour solicitation for our addr */
3141         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3142         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3143 out:
3144         in6_ifa_put(ifp);
3145 }
3146
3147 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3148 {
3149         struct net_device *dev = ifp->idev->dev;
3150
3151         /*
3152          *      Configure the address for reception. Now it is valid.
3153          */
3154
3155         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3156
3157         /* If added prefix is link local and we are prepared to process
3158            router advertisements, start sending router solicitations.
3159          */
3160
3161         if (ipv6_accept_ra(ifp->idev) &&
3162             ifp->idev->cnf.rtr_solicits > 0 &&
3163             (dev->flags&IFF_LOOPBACK) == 0 &&
3164             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3165                 /*
3166                  *      If a host as already performed a random delay
3167                  *      [...] as part of DAD [...] there is no need
3168                  *      to delay again before sending the first RS
3169                  */
3170                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3171
3172                 spin_lock_bh(&ifp->lock);
3173                 ifp->probes = 1;
3174                 ifp->idev->if_flags |= IF_RS_SENT;
3175                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3176                 spin_unlock_bh(&ifp->lock);
3177         }
3178 }
3179
3180 static void addrconf_dad_run(struct inet6_dev *idev)
3181 {
3182         struct inet6_ifaddr *ifp;
3183
3184         read_lock_bh(&idev->lock);
3185         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3186                 spin_lock(&ifp->lock);
3187                 if (ifp->flags & IFA_F_TENTATIVE &&
3188                     ifp->state == INET6_IFADDR_STATE_DAD)
3189                         addrconf_dad_kick(ifp);
3190                 spin_unlock(&ifp->lock);
3191         }
3192         read_unlock_bh(&idev->lock);
3193 }
3194
3195 #ifdef CONFIG_PROC_FS
3196 struct if6_iter_state {
3197         struct seq_net_private p;
3198         int bucket;
3199         int offset;
3200 };
3201
3202 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3203 {
3204         struct inet6_ifaddr *ifa = NULL;
3205         struct if6_iter_state *state = seq->private;
3206         struct net *net = seq_file_net(seq);
3207         int p = 0;
3208
3209         /* initial bucket if pos is 0 */
3210         if (pos == 0) {
3211                 state->bucket = 0;
3212                 state->offset = 0;
3213         }
3214
3215         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3216                 struct hlist_node *n;
3217                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3218                                          addr_lst) {
3219                         if (!net_eq(dev_net(ifa->idev->dev), net))
3220                                 continue;
3221                         /* sync with offset */
3222                         if (p < state->offset) {
3223                                 p++;
3224                                 continue;
3225                         }
3226                         state->offset++;
3227                         return ifa;
3228                 }
3229
3230                 /* prepare for next bucket */
3231                 state->offset = 0;
3232                 p = 0;
3233         }
3234         return NULL;
3235 }
3236
3237 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3238                                          struct inet6_ifaddr *ifa)
3239 {
3240         struct if6_iter_state *state = seq->private;
3241         struct net *net = seq_file_net(seq);
3242         struct hlist_node *n = &ifa->addr_lst;
3243
3244         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3245                 if (!net_eq(dev_net(ifa->idev->dev), net))
3246                         continue;
3247                 state->offset++;
3248                 return ifa;
3249         }
3250
3251         while (++state->bucket < IN6_ADDR_HSIZE) {
3252                 state->offset = 0;
3253                 hlist_for_each_entry_rcu_bh(ifa, n,
3254                                      &inet6_addr_lst[state->bucket], addr_lst) {
3255                         if (!net_eq(dev_net(ifa->idev->dev), net))
3256                                 continue;
3257                         state->offset++;
3258                         return ifa;
3259                 }
3260         }
3261
3262         return NULL;
3263 }
3264
3265 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3266         __acquires(rcu_bh)
3267 {
3268         rcu_read_lock_bh();
3269         return if6_get_first(seq, *pos);
3270 }
3271
3272 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3273 {
3274         struct inet6_ifaddr *ifa;
3275
3276         ifa = if6_get_next(seq, v);
3277         ++*pos;
3278         return ifa;
3279 }
3280
3281 static void if6_seq_stop(struct seq_file *seq, void *v)
3282         __releases(rcu_bh)
3283 {
3284         rcu_read_unlock_bh();
3285 }
3286
3287 static int if6_seq_show(struct seq_file *seq, void *v)
3288 {
3289         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3290         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3291                    &ifp->addr,
3292                    ifp->idev->dev->ifindex,
3293                    ifp->prefix_len,
3294                    ifp->scope,
3295                    ifp->flags,
3296                    ifp->idev->dev->name);
3297         return 0;
3298 }
3299
3300 static const struct seq_operations if6_seq_ops = {
3301         .start  = if6_seq_start,
3302         .next   = if6_seq_next,
3303         .show   = if6_seq_show,
3304         .stop   = if6_seq_stop,
3305 };
3306
3307 static int if6_seq_open(struct inode *inode, struct file *file)
3308 {
3309         return seq_open_net(inode, file, &if6_seq_ops,
3310                             sizeof(struct if6_iter_state));
3311 }
3312
3313 static const struct file_operations if6_fops = {
3314         .owner          = THIS_MODULE,
3315         .open           = if6_seq_open,
3316         .read           = seq_read,
3317         .llseek         = seq_lseek,
3318         .release        = seq_release_net,
3319 };
3320
3321 static int __net_init if6_proc_net_init(struct net *net)
3322 {
3323         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3324                 return -ENOMEM;
3325         return 0;
3326 }
3327
3328 static void __net_exit if6_proc_net_exit(struct net *net)
3329 {
3330        proc_net_remove(net, "if_inet6");
3331 }
3332
3333 static struct pernet_operations if6_proc_net_ops = {
3334        .init = if6_proc_net_init,
3335        .exit = if6_proc_net_exit,
3336 };
3337
3338 int __init if6_proc_init(void)
3339 {
3340         return register_pernet_subsys(&if6_proc_net_ops);
3341 }
3342
3343 void if6_proc_exit(void)
3344 {
3345         unregister_pernet_subsys(&if6_proc_net_ops);
3346 }
3347 #endif  /* CONFIG_PROC_FS */
3348
3349 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3350 /* Check if address is a home address configured on any interface. */
3351 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3352 {
3353         int ret = 0;
3354         struct inet6_ifaddr *ifp = NULL;
3355         struct hlist_node *n;
3356         unsigned int hash = inet6_addr_hash(addr);
3357
3358         rcu_read_lock_bh();
3359         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3360                 if (!net_eq(dev_net(ifp->idev->dev), net))
3361                         continue;
3362                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3363                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3364                         ret = 1;
3365                         break;
3366                 }
3367         }
3368         rcu_read_unlock_bh();
3369         return ret;
3370 }
3371 #endif
3372
3373 /*
3374  *      Periodic address status verification
3375  */
3376
3377 static void addrconf_verify(unsigned long foo)
3378 {
3379         unsigned long now, next, next_sec, next_sched;
3380         struct inet6_ifaddr *ifp;
3381         struct hlist_node *node;
3382         int i;
3383
3384         rcu_read_lock_bh();
3385         spin_lock(&addrconf_verify_lock);
3386         now = jiffies;
3387         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3388
3389         del_timer(&addr_chk_timer);
3390
3391         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3392 restart:
3393                 hlist_for_each_entry_rcu_bh(ifp, node,
3394                                          &inet6_addr_lst[i], addr_lst) {
3395                         unsigned long age;
3396
3397                         if (ifp->flags & IFA_F_PERMANENT)
3398                                 continue;
3399
3400                         spin_lock(&ifp->lock);
3401                         /* We try to batch several events at once. */
3402                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3403
3404                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3405                             age >= ifp->valid_lft) {
3406                                 spin_unlock(&ifp->lock);
3407                                 in6_ifa_hold(ifp);
3408                                 ipv6_del_addr(ifp);
3409                                 goto restart;
3410                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3411                                 spin_unlock(&ifp->lock);
3412                                 continue;
3413                         } else if (age >= ifp->prefered_lft) {
3414                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3415                                 int deprecate = 0;
3416
3417                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3418                                         deprecate = 1;
3419                                         ifp->flags |= IFA_F_DEPRECATED;
3420                                 }
3421
3422                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3423                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3424
3425                                 spin_unlock(&ifp->lock);
3426
3427                                 if (deprecate) {
3428                                         in6_ifa_hold(ifp);
3429
3430                                         ipv6_ifa_notify(0, ifp);
3431                                         in6_ifa_put(ifp);
3432                                         goto restart;
3433                                 }
3434 #ifdef CONFIG_IPV6_PRIVACY
3435                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3436                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3437                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3438                                         ifp->idev->cnf.dad_transmits *
3439                                         ifp->idev->nd_parms->retrans_time / HZ;
3440
3441                                 if (age >= ifp->prefered_lft - regen_advance) {
3442                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3443                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3444                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3445                                         if (!ifp->regen_count && ifpub) {
3446                                                 ifp->regen_count++;
3447                                                 in6_ifa_hold(ifp);
3448                                                 in6_ifa_hold(ifpub);
3449                                                 spin_unlock(&ifp->lock);
3450
3451                                                 spin_lock(&ifpub->lock);
3452                                                 ifpub->regen_count = 0;
3453                                                 spin_unlock(&ifpub->lock);
3454                                                 ipv6_create_tempaddr(ifpub, ifp);
3455                                                 in6_ifa_put(ifpub);
3456                                                 in6_ifa_put(ifp);
3457                                                 goto restart;
3458                                         }
3459                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3460                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3461                                 spin_unlock(&ifp->lock);
3462 #endif
3463                         } else {
3464                                 /* ifp->prefered_lft <= ifp->valid_lft */
3465                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3466                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3467                                 spin_unlock(&ifp->lock);
3468                         }
3469                 }
3470         }
3471
3472         next_sec = round_jiffies_up(next);
3473         next_sched = next;
3474
3475         /* If rounded timeout is accurate enough, accept it. */
3476         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3477                 next_sched = next_sec;
3478
3479         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3480         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3481                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3482
3483         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3484               now, next, next_sec, next_sched));
3485
3486         addr_chk_timer.expires = next_sched;
3487         add_timer(&addr_chk_timer);
3488         spin_unlock(&addrconf_verify_lock);
3489         rcu_read_unlock_bh();
3490 }
3491
3492 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3493 {
3494         struct in6_addr *pfx = NULL;
3495
3496         if (addr)
3497                 pfx = nla_data(addr);
3498
3499         if (local) {
3500                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3501                         pfx = NULL;
3502                 else
3503                         pfx = nla_data(local);
3504         }
3505
3506         return pfx;
3507 }
3508
3509 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3510         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3511         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3512         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3513 };
3514
3515 static int
3516 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3517 {
3518         struct net *net = sock_net(skb->sk);
3519         struct ifaddrmsg *ifm;
3520         struct nlattr *tb[IFA_MAX+1];
3521         struct in6_addr *pfx;
3522         int err;
3523
3524         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3525         if (err < 0)
3526                 return err;
3527
3528         ifm = nlmsg_data(nlh);
3529         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3530         if (pfx == NULL)
3531                 return -EINVAL;
3532
3533         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3534 }
3535
3536 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3537                              u32 prefered_lft, u32 valid_lft)
3538 {
3539         u32 flags;
3540         clock_t expires;
3541         unsigned long timeout;
3542
3543         if (!valid_lft || (prefered_lft > valid_lft))
3544                 return -EINVAL;
3545
3546         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3547         if (addrconf_finite_timeout(timeout)) {
3548                 expires = jiffies_to_clock_t(timeout * HZ);
3549                 valid_lft = timeout;
3550                 flags = RTF_EXPIRES;
3551         } else {
3552                 expires = 0;
3553                 flags = 0;
3554                 ifa_flags |= IFA_F_PERMANENT;
3555         }
3556
3557         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3558         if (addrconf_finite_timeout(timeout)) {
3559                 if (timeout == 0)
3560                         ifa_flags |= IFA_F_DEPRECATED;
3561                 prefered_lft = timeout;
3562         }
3563
3564         spin_lock_bh(&ifp->lock);
3565         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3566         ifp->tstamp = jiffies;
3567         ifp->valid_lft = valid_lft;
3568         ifp->prefered_lft = prefered_lft;
3569
3570         spin_unlock_bh(&ifp->lock);
3571         if (!(ifp->flags&IFA_F_TENTATIVE))
3572                 ipv6_ifa_notify(0, ifp);
3573
3574         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3575                               expires, flags);
3576         addrconf_verify(0);
3577
3578         return 0;
3579 }
3580
3581 static int
3582 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3583 {
3584         struct net *net = sock_net(skb->sk);
3585         struct ifaddrmsg *ifm;
3586         struct nlattr *tb[IFA_MAX+1];
3587         struct in6_addr *pfx;
3588         struct inet6_ifaddr *ifa;
3589         struct net_device *dev;
3590         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3591         u8 ifa_flags;
3592         int err;
3593
3594         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3595         if (err < 0)
3596                 return err;
3597
3598         ifm = nlmsg_data(nlh);
3599         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3600         if (pfx == NULL)
3601                 return -EINVAL;
3602
3603         if (tb[IFA_CACHEINFO]) {
3604                 struct ifa_cacheinfo *ci;
3605
3606                 ci = nla_data(tb[IFA_CACHEINFO]);
3607                 valid_lft = ci->ifa_valid;
3608                 preferred_lft = ci->ifa_prefered;
3609         } else {
3610                 preferred_lft = INFINITY_LIFE_TIME;
3611                 valid_lft = INFINITY_LIFE_TIME;
3612         }
3613
3614         dev =  __dev_get_by_index(net, ifm->ifa_index);
3615         if (dev == NULL)
3616                 return -ENODEV;
3617
3618         /* We ignore other flags so far. */
3619         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3620
3621         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3622         if (ifa == NULL) {
3623                 /*
3624                  * It would be best to check for !NLM_F_CREATE here but
3625                  * userspace alreay relies on not having to provide this.
3626                  */
3627                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3628                                       ifm->ifa_prefixlen, ifa_flags,
3629                                       preferred_lft, valid_lft);
3630         }
3631
3632         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3633             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3634                 err = -EEXIST;
3635         else
3636                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3637
3638         in6_ifa_put(ifa);
3639
3640         return err;
3641 }
3642
3643 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3644                           u8 scope, int ifindex)
3645 {
3646         struct ifaddrmsg *ifm;
3647
3648         ifm = nlmsg_data(nlh);
3649         ifm->ifa_family = AF_INET6;
3650         ifm->ifa_prefixlen = prefixlen;
3651         ifm->ifa_flags = flags;
3652         ifm->ifa_scope = scope;
3653         ifm->ifa_index = ifindex;
3654 }
3655
3656 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3657                          unsigned long tstamp, u32 preferred, u32 valid)
3658 {
3659         struct ifa_cacheinfo ci;
3660
3661         ci.cstamp = cstamp_delta(cstamp);
3662         ci.tstamp = cstamp_delta(tstamp);
3663         ci.ifa_prefered = preferred;
3664         ci.ifa_valid = valid;
3665
3666         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3667 }
3668
3669 static inline int rt_scope(int ifa_scope)
3670 {
3671         if (ifa_scope & IFA_HOST)
3672                 return RT_SCOPE_HOST;
3673         else if (ifa_scope & IFA_LINK)
3674                 return RT_SCOPE_LINK;
3675         else if (ifa_scope & IFA_SITE)
3676                 return RT_SCOPE_SITE;
3677         else
3678                 return RT_SCOPE_UNIVERSE;
3679 }
3680
3681 static inline int inet6_ifaddr_msgsize(void)
3682 {
3683         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3684                + nla_total_size(16) /* IFA_ADDRESS */
3685                + nla_total_size(sizeof(struct ifa_cacheinfo));
3686 }
3687
3688 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3689                              u32 portid, u32 seq, int event, unsigned int flags)
3690 {
3691         struct nlmsghdr  *nlh;
3692         u32 preferred, valid;
3693
3694         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3695         if (nlh == NULL)
3696                 return -EMSGSIZE;
3697
3698         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3699                       ifa->idev->dev->ifindex);
3700
3701         if (!(ifa->flags&IFA_F_PERMANENT)) {
3702                 preferred = ifa->prefered_lft;
3703                 valid = ifa->valid_lft;
3704                 if (preferred != INFINITY_LIFE_TIME) {
3705                         long tval = (jiffies - ifa->tstamp)/HZ;
3706                         if (preferred > tval)
3707                                 preferred -= tval;
3708                         else
3709                                 preferred = 0;
3710                         if (valid != INFINITY_LIFE_TIME) {
3711                                 if (valid > tval)
3712                                         valid -= tval;
3713                                 else
3714                                         valid = 0;
3715                         }
3716                 }
3717         } else {
3718                 preferred = INFINITY_LIFE_TIME;
3719                 valid = INFINITY_LIFE_TIME;
3720         }
3721
3722         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3723             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3724                 nlmsg_cancel(skb, nlh);
3725                 return -EMSGSIZE;
3726         }
3727
3728         return nlmsg_end(skb, nlh);
3729 }
3730
3731 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3732                                 u32 portid, u32 seq, int event, u16 flags)
3733 {
3734         struct nlmsghdr  *nlh;
3735         u8 scope = RT_SCOPE_UNIVERSE;
3736         int ifindex = ifmca->idev->dev->ifindex;
3737
3738         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3739                 scope = RT_SCOPE_SITE;
3740
3741         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3742         if (nlh == NULL)
3743                 return -EMSGSIZE;
3744
3745         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3746         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3747             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3748                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3749                 nlmsg_cancel(skb, nlh);
3750                 return -EMSGSIZE;
3751         }
3752
3753         return nlmsg_end(skb, nlh);
3754 }
3755
3756 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3757                                 u32 portid, u32 seq, int event, unsigned int flags)
3758 {
3759         struct nlmsghdr  *nlh;
3760         u8 scope = RT_SCOPE_UNIVERSE;
3761         int ifindex = ifaca->aca_idev->dev->ifindex;
3762
3763         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3764                 scope = RT_SCOPE_SITE;
3765
3766         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3767         if (nlh == NULL)
3768                 return -EMSGSIZE;
3769
3770         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3771         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3772             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3773                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3774                 nlmsg_cancel(skb, nlh);
3775                 return -EMSGSIZE;
3776         }
3777
3778         return nlmsg_end(skb, nlh);
3779 }
3780
3781 enum addr_type_t {
3782         UNICAST_ADDR,
3783         MULTICAST_ADDR,
3784         ANYCAST_ADDR,
3785 };
3786
3787 /* called with rcu_read_lock() */
3788 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3789                           struct netlink_callback *cb, enum addr_type_t type,
3790                           int s_ip_idx, int *p_ip_idx)
3791 {
3792         struct ifmcaddr6 *ifmca;
3793         struct ifacaddr6 *ifaca;
3794         int err = 1;
3795         int ip_idx = *p_ip_idx;
3796
3797         read_lock_bh(&idev->lock);
3798         switch (type) {
3799         case UNICAST_ADDR: {
3800                 struct inet6_ifaddr *ifa;
3801
3802                 /* unicast address incl. temp addr */
3803                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3804                         if (++ip_idx < s_ip_idx)
3805                                 continue;
3806                         err = inet6_fill_ifaddr(skb, ifa,
3807                                                 NETLINK_CB(cb->skb).portid,
3808                                                 cb->nlh->nlmsg_seq,
3809                                                 RTM_NEWADDR,
3810                                                 NLM_F_MULTI);
3811                         if (err <= 0)
3812                                 break;
3813                 }
3814                 break;
3815         }
3816         case MULTICAST_ADDR:
3817                 /* multicast address */
3818                 for (ifmca = idev->mc_list; ifmca;
3819                      ifmca = ifmca->next, ip_idx++) {
3820                         if (ip_idx < s_ip_idx)
3821                                 continue;
3822                         err = inet6_fill_ifmcaddr(skb, ifmca,
3823                                                   NETLINK_CB(cb->skb).portid,
3824                                                   cb->nlh->nlmsg_seq,
3825                                                   RTM_GETMULTICAST,
3826                                                   NLM_F_MULTI);
3827                         if (err <= 0)
3828                                 break;
3829                 }
3830                 break;
3831         case ANYCAST_ADDR:
3832                 /* anycast address */
3833                 for (ifaca = idev->ac_list; ifaca;
3834                      ifaca = ifaca->aca_next, ip_idx++) {
3835                         if (ip_idx < s_ip_idx)
3836                                 continue;
3837                         err = inet6_fill_ifacaddr(skb, ifaca,
3838                                                   NETLINK_CB(cb->skb).portid,
3839                                                   cb->nlh->nlmsg_seq,
3840                                                   RTM_GETANYCAST,
3841                                                   NLM_F_MULTI);
3842                         if (err <= 0)
3843                                 break;
3844                 }
3845                 break;
3846         default:
3847                 break;
3848         }
3849         read_unlock_bh(&idev->lock);
3850         *p_ip_idx = ip_idx;
3851         return err;
3852 }
3853
3854 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3855                            enum addr_type_t type)
3856 {
3857         struct net *net = sock_net(skb->sk);
3858         int h, s_h;
3859         int idx, ip_idx;
3860         int s_idx, s_ip_idx;
3861         struct net_device *dev;
3862         struct inet6_dev *idev;
3863         struct hlist_head *head;
3864         struct hlist_node *node;
3865
3866         s_h = cb->args[0];
3867         s_idx = idx = cb->args[1];
3868         s_ip_idx = ip_idx = cb->args[2];
3869
3870         rcu_read_lock();
3871         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3872                 idx = 0;
3873                 head = &net->dev_index_head[h];
3874                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3875                         if (idx < s_idx)
3876                                 goto cont;
3877                         if (h > s_h || idx > s_idx)
3878                                 s_ip_idx = 0;
3879                         ip_idx = 0;
3880                         idev = __in6_dev_get(dev);
3881                         if (!idev)
3882                                 goto cont;
3883
3884                         if (in6_dump_addrs(idev, skb, cb, type,
3885                                            s_ip_idx, &ip_idx) <= 0)
3886                                 goto done;
3887 cont:
3888                         idx++;
3889                 }
3890         }
3891 done:
3892         rcu_read_unlock();
3893         cb->args[0] = h;
3894         cb->args[1] = idx;
3895         cb->args[2] = ip_idx;
3896
3897         return skb->len;
3898 }
3899
3900 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3901 {
3902         enum addr_type_t type = UNICAST_ADDR;
3903
3904         return inet6_dump_addr(skb, cb, type);
3905 }
3906
3907 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3908 {
3909         enum addr_type_t type = MULTICAST_ADDR;
3910
3911         return inet6_dump_addr(skb, cb, type);
3912 }
3913
3914
3915 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3916 {
3917         enum addr_type_t type = ANYCAST_ADDR;
3918
3919         return inet6_dump_addr(skb, cb, type);
3920 }
3921
3922 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3923                              void *arg)
3924 {
3925         struct net *net = sock_net(in_skb->sk);
3926         struct ifaddrmsg *ifm;
3927         struct nlattr *tb[IFA_MAX+1];
3928         struct in6_addr *addr = NULL;
3929         struct net_device *dev = NULL;
3930         struct inet6_ifaddr *ifa;
3931         struct sk_buff *skb;
3932         int err;
3933
3934         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3935         if (err < 0)
3936                 goto errout;
3937
3938         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3939         if (addr == NULL) {
3940                 err = -EINVAL;
3941                 goto errout;
3942         }
3943
3944         ifm = nlmsg_data(nlh);
3945         if (ifm->ifa_index)
3946                 dev = __dev_get_by_index(net, ifm->ifa_index);
3947
3948         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3949         if (!ifa) {
3950                 err = -EADDRNOTAVAIL;
3951                 goto errout;
3952         }
3953
3954         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3955         if (!skb) {
3956                 err = -ENOBUFS;
3957                 goto errout_ifa;
3958         }
3959
3960         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3961                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3962         if (err < 0) {
3963                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3964                 WARN_ON(err == -EMSGSIZE);
3965                 kfree_skb(skb);
3966                 goto errout_ifa;
3967         }
3968         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3969 errout_ifa:
3970         in6_ifa_put(ifa);
3971 errout:
3972         return err;
3973 }
3974
3975 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3976 {
3977         struct sk_buff *skb;
3978         struct net *net = dev_net(ifa->idev->dev);
3979         int err = -ENOBUFS;
3980
3981         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3982         if (skb == NULL)
3983                 goto errout;
3984
3985         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3986         if (err < 0) {
3987                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3988                 WARN_ON(err == -EMSGSIZE);
3989                 kfree_skb(skb);
3990                 goto errout;
3991         }
3992         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3993         return;
3994 errout:
3995         if (err < 0)
3996                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3997 }
3998
3999 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4000                                 __s32 *array, int bytes)
4001 {
4002         BUG_ON(bytes < (DEVCONF_MAX * 4));
4003
4004         memset(array, 0, bytes);
4005         array[DEVCONF_FORWARDING] = cnf->forwarding;
4006         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4007         array[DEVCONF_MTU6] = cnf->mtu6;
4008         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4009         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4010         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4011         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4012         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4013         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4014                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4015         array[DEVCONF_RTR_SOLICIT_DELAY] =
4016                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4017         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4018 #ifdef CONFIG_IPV6_PRIVACY
4019         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4020         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4021         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4022         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4023         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4024 #endif
4025         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4026         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4027         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4028 #ifdef CONFIG_IPV6_ROUTER_PREF
4029         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4030         array[DEVCONF_RTR_PROBE_INTERVAL] =
4031                 jiffies_to_msecs(cnf->rtr_probe_interval);
4032 #ifdef CONFIG_IPV6_ROUTE_INFO
4033         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4034 #endif
4035 #endif
4036         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4037         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4038 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4039         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4040 #endif
4041 #ifdef CONFIG_IPV6_MROUTE
4042         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4043 #endif
4044         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4045         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4046         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4047         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4048 }
4049
4050 static inline size_t inet6_ifla6_size(void)
4051 {
4052         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4053              + nla_total_size(sizeof(struct ifla_cacheinfo))
4054              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4055              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4056              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4057 }
4058
4059 static inline size_t inet6_if_nlmsg_size(void)
4060 {
4061         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4062                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4063                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4064                + nla_total_size(4) /* IFLA_MTU */
4065                + nla_total_size(4) /* IFLA_LINK */
4066                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4067 }
4068
4069 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4070                                       int items, int bytes)
4071 {
4072         int i;
4073         int pad = bytes - sizeof(u64) * items;
4074         BUG_ON(pad < 0);
4075
4076         /* Use put_unaligned() because stats may not be aligned for u64. */
4077         put_unaligned(items, &stats[0]);
4078         for (i = 1; i < items; i++)
4079                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4080
4081         memset(&stats[items], 0, pad);
4082 }
4083
4084 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4085                                       int items, int bytes, size_t syncpoff)
4086 {
4087         int i;
4088         int pad = bytes - sizeof(u64) * items;
4089         BUG_ON(pad < 0);
4090
4091         /* Use put_unaligned() because stats may not be aligned for u64. */
4092         put_unaligned(items, &stats[0]);
4093         for (i = 1; i < items; i++)
4094                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4095
4096         memset(&stats[items], 0, pad);
4097 }
4098
4099 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4100                              int bytes)
4101 {
4102         switch (attrtype) {
4103         case IFLA_INET6_STATS:
4104                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4105                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4106                 break;
4107         case IFLA_INET6_ICMP6STATS:
4108                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4109                 break;
4110         }
4111 }
4112
4113 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4114 {
4115         struct nlattr *nla;
4116         struct ifla_cacheinfo ci;
4117
4118         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4119                 goto nla_put_failure;
4120         ci.max_reasm_len = IPV6_MAXPLEN;
4121         ci.tstamp = cstamp_delta(idev->tstamp);
4122         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4123         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4124         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4125                 goto nla_put_failure;
4126         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4127         if (nla == NULL)
4128                 goto nla_put_failure;
4129         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4130
4131         /* XXX - MC not implemented */
4132
4133         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4134         if (nla == NULL)
4135                 goto nla_put_failure;
4136         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4137
4138         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4139         if (nla == NULL)
4140                 goto nla_put_failure;
4141         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4142
4143         return 0;
4144
4145 nla_put_failure:
4146         return -EMSGSIZE;
4147 }
4148
4149 static size_t inet6_get_link_af_size(const struct net_device *dev)
4150 {
4151         if (!__in6_dev_get(dev))
4152                 return 0;
4153
4154         return inet6_ifla6_size();
4155 }
4156
4157 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4158 {
4159         struct inet6_dev *idev = __in6_dev_get(dev);
4160
4161         if (!idev)
4162                 return -ENODATA;
4163
4164         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4165                 return -EMSGSIZE;
4166
4167         return 0;
4168 }
4169
4170 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4171                              u32 portid, u32 seq, int event, unsigned int flags)
4172 {
4173         struct net_device *dev = idev->dev;
4174         struct ifinfomsg *hdr;
4175         struct nlmsghdr *nlh;
4176         void *protoinfo;
4177
4178         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4179         if (nlh == NULL)
4180                 return -EMSGSIZE;
4181
4182         hdr = nlmsg_data(nlh);
4183         hdr->ifi_family = AF_INET6;
4184         hdr->__ifi_pad = 0;
4185         hdr->ifi_type = dev->type;
4186         hdr->ifi_index = dev->ifindex;
4187         hdr->ifi_flags = dev_get_flags(dev);
4188         hdr->ifi_change = 0;
4189
4190         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4191             (dev->addr_len &&
4192              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4193             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4194             (dev->ifindex != dev->iflink &&
4195              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4196                 goto nla_put_failure;
4197         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4198         if (protoinfo == NULL)
4199                 goto nla_put_failure;
4200
4201         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4202                 goto nla_put_failure;
4203
4204         nla_nest_end(skb, protoinfo);
4205         return nlmsg_end(skb, nlh);
4206
4207 nla_put_failure:
4208         nlmsg_cancel(skb, nlh);
4209         return -EMSGSIZE;
4210 }
4211
4212 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4213 {
4214         struct net *net = sock_net(skb->sk);
4215         int h, s_h;
4216         int idx = 0, s_idx;
4217         struct net_device *dev;
4218         struct inet6_dev *idev;
4219         struct hlist_head *head;
4220         struct hlist_node *node;
4221
4222         s_h = cb->args[0];
4223         s_idx = cb->args[1];
4224
4225         rcu_read_lock();
4226         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4227                 idx = 0;
4228                 head = &net->dev_index_head[h];
4229                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4230                         if (idx < s_idx)
4231                                 goto cont;
4232                         idev = __in6_dev_get(dev);
4233                         if (!idev)
4234                                 goto cont;
4235                         if (inet6_fill_ifinfo(skb, idev,
4236                                               NETLINK_CB(cb->skb).portid,
4237                                               cb->nlh->nlmsg_seq,
4238                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4239                                 goto out;
4240 cont:
4241                         idx++;
4242                 }
4243         }
4244 out:
4245         rcu_read_unlock();
4246         cb->args[1] = idx;
4247         cb->args[0] = h;
4248
4249         return skb->len;
4250 }
4251
4252 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4253 {
4254         struct sk_buff *skb;
4255         struct net *net = dev_net(idev->dev);
4256         int err = -ENOBUFS;
4257
4258         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4259         if (skb == NULL)
4260                 goto errout;
4261
4262         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4263         if (err < 0) {
4264                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4265                 WARN_ON(err == -EMSGSIZE);
4266                 kfree_skb(skb);
4267                 goto errout;
4268         }
4269         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4270         return;
4271 errout:
4272         if (err < 0)
4273                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4274 }
4275
4276 static inline size_t inet6_prefix_nlmsg_size(void)
4277 {
4278         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4279                + nla_total_size(sizeof(struct in6_addr))
4280                + nla_total_size(sizeof(struct prefix_cacheinfo));
4281 }
4282
4283 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4284                              struct prefix_info *pinfo, u32 portid, u32 seq,
4285                              int event, unsigned int flags)
4286 {
4287         struct prefixmsg *pmsg;
4288         struct nlmsghdr *nlh;
4289         struct prefix_cacheinfo ci;
4290
4291         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4292         if (nlh == NULL)
4293                 return -EMSGSIZE;
4294
4295         pmsg = nlmsg_data(nlh);
4296         pmsg->prefix_family = AF_INET6;
4297         pmsg->prefix_pad1 = 0;
4298         pmsg->prefix_pad2 = 0;
4299         pmsg->prefix_ifindex = idev->dev->ifindex;
4300         pmsg->prefix_len = pinfo->prefix_len;
4301         pmsg->prefix_type = pinfo->type;
4302         pmsg->prefix_pad3 = 0;
4303         pmsg->prefix_flags = 0;
4304         if (pinfo->onlink)
4305                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4306         if (pinfo->autoconf)
4307                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4308
4309         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4310                 goto nla_put_failure;
4311         ci.preferred_time = ntohl(pinfo->prefered);
4312         ci.valid_time = ntohl(pinfo->valid);
4313         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4314                 goto nla_put_failure;
4315         return nlmsg_end(skb, nlh);
4316
4317 nla_put_failure:
4318         nlmsg_cancel(skb, nlh);
4319         return -EMSGSIZE;
4320 }
4321
4322 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4323                          struct prefix_info *pinfo)
4324 {
4325         struct sk_buff *skb;
4326         struct net *net = dev_net(idev->dev);
4327         int err = -ENOBUFS;
4328
4329         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4330         if (skb == NULL)
4331                 goto errout;
4332
4333         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4334         if (err < 0) {
4335                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4336                 WARN_ON(err == -EMSGSIZE);
4337                 kfree_skb(skb);
4338                 goto errout;
4339         }
4340         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4341         return;
4342 errout:
4343         if (err < 0)
4344                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4345 }
4346
4347 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4348 {
4349         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4350
4351         switch (event) {
4352         case RTM_NEWADDR:
4353                 /*
4354                  * If the address was optimistic
4355                  * we inserted the route at the start of
4356                  * our DAD process, so we don't need
4357                  * to do it again
4358                  */
4359                 if (!(ifp->rt->rt6i_node))
4360                         ip6_ins_rt(ifp->rt);
4361                 if (ifp->idev->cnf.forwarding)
4362                         addrconf_join_anycast(ifp);
4363                 break;
4364         case RTM_DELADDR:
4365                 if (ifp->idev->cnf.forwarding)
4366                         addrconf_leave_anycast(ifp);
4367                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4368                 dst_hold(&ifp->rt->dst);
4369
4370                 if (ip6_del_rt(ifp->rt))
4371                         dst_free(&ifp->rt->dst);
4372                 break;
4373         }
4374 }
4375
4376 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4377 {
4378         rcu_read_lock_bh();
4379         if (likely(ifp->idev->dead == 0))
4380                 __ipv6_ifa_notify(event, ifp);
4381         rcu_read_unlock_bh();
4382 }
4383
4384 #ifdef CONFIG_SYSCTL
4385
4386 static
4387 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4388                            void __user *buffer, size_t *lenp, loff_t *ppos)
4389 {
4390         int *valp = ctl->data;
4391         int val = *valp;
4392         loff_t pos = *ppos;
4393         ctl_table lctl;
4394         int ret;
4395
4396         /*
4397          * ctl->data points to idev->cnf.forwarding, we should
4398          * not modify it until we get the rtnl lock.
4399          */
4400         lctl = *ctl;
4401         lctl.data = &val;
4402
4403         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4404
4405         if (write)
4406                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4407         if (ret)
4408                 *ppos = pos;
4409         return ret;
4410 }
4411
4412 static void dev_disable_change(struct inet6_dev *idev)
4413 {
4414         if (!idev || !idev->dev)
4415                 return;
4416
4417         if (idev->cnf.disable_ipv6)
4418                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4419         else
4420                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4421 }
4422
4423 static void addrconf_disable_change(struct net *net, __s32 newf)
4424 {
4425         struct net_device *dev;
4426         struct inet6_dev *idev;
4427
4428         rcu_read_lock();
4429         for_each_netdev_rcu(net, dev) {
4430                 idev = __in6_dev_get(dev);
4431                 if (idev) {
4432                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4433                         idev->cnf.disable_ipv6 = newf;
4434                         if (changed)
4435                                 dev_disable_change(idev);
4436                 }
4437         }
4438         rcu_read_unlock();
4439 }
4440
4441 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4442 {
4443         struct net *net;
4444         int old;
4445
4446         if (!rtnl_trylock())
4447                 return restart_syscall();
4448
4449         net = (struct net *)table->extra2;
4450         old = *p;
4451         *p = newf;
4452
4453         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4454                 rtnl_unlock();
4455                 return 0;
4456         }
4457
4458         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4459                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4460                 addrconf_disable_change(net, newf);
4461         } else if ((!newf) ^ (!old))
4462                 dev_disable_change((struct inet6_dev *)table->extra1);
4463
4464         rtnl_unlock();
4465         return 0;
4466 }
4467
4468 static
4469 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4470                             void __user *buffer, size_t *lenp, loff_t *ppos)
4471 {
4472         int *valp = ctl->data;
4473         int val = *valp;
4474         loff_t pos = *ppos;
4475         ctl_table lctl;
4476         int ret;
4477
4478         /*
4479          * ctl->data points to idev->cnf.disable_ipv6, we should
4480          * not modify it until we get the rtnl lock.
4481          */
4482         lctl = *ctl;
4483         lctl.data = &val;
4484
4485         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4486
4487         if (write)
4488                 ret = addrconf_disable_ipv6(ctl, valp, val);
4489         if (ret)
4490                 *ppos = pos;
4491         return ret;
4492 }
4493
4494 static struct addrconf_sysctl_table
4495 {
4496         struct ctl_table_header *sysctl_header;
4497         ctl_table addrconf_vars[DEVCONF_MAX+1];
4498 } addrconf_sysctl __read_mostly = {
4499         .sysctl_header = NULL,
4500         .addrconf_vars = {
4501                 {
4502                         .procname       = "forwarding",
4503                         .data           = &ipv6_devconf.forwarding,
4504                         .maxlen         = sizeof(int),
4505                         .mode           = 0644,
4506                         .proc_handler   = addrconf_sysctl_forward,
4507                 },
4508                 {
4509                         .procname       = "hop_limit",
4510                         .data           = &ipv6_devconf.hop_limit,
4511                         .maxlen         = sizeof(int),
4512                         .mode           = 0644,
4513                         .proc_handler   = proc_dointvec,
4514                 },
4515                 {
4516                         .procname       = "mtu",
4517                         .data           = &ipv6_devconf.mtu6,
4518                         .maxlen         = sizeof(int),
4519                         .mode           = 0644,
4520                         .proc_handler   = proc_dointvec,
4521                 },
4522                 {
4523                         .procname       = "accept_ra",
4524                         .data           = &ipv6_devconf.accept_ra,
4525                         .maxlen         = sizeof(int),
4526                         .mode           = 0644,
4527                         .proc_handler   = proc_dointvec,
4528                 },
4529                 {
4530                         .procname       = "accept_redirects",
4531                         .data           = &ipv6_devconf.accept_redirects,
4532                         .maxlen         = sizeof(int),
4533                         .mode           = 0644,
4534                         .proc_handler   = proc_dointvec,
4535                 },
4536                 {
4537                         .procname       = "autoconf",
4538                         .data           = &ipv6_devconf.autoconf,
4539                         .maxlen         = sizeof(int),
4540                         .mode           = 0644,
4541                         .proc_handler   = proc_dointvec,
4542                 },
4543                 {
4544                         .procname       = "dad_transmits",
4545                         .data           = &ipv6_devconf.dad_transmits,
4546                         .maxlen         = sizeof(int),
4547                         .mode           = 0644,
4548                         .proc_handler   = proc_dointvec,
4549                 },
4550                 {
4551                         .procname       = "router_solicitations",
4552                         .data           = &ipv6_devconf.rtr_solicits,
4553                         .maxlen         = sizeof(int),
4554                         .mode           = 0644,
4555                         .proc_handler   = proc_dointvec,
4556                 },
4557                 {
4558                         .procname       = "router_solicitation_interval",
4559                         .data           = &ipv6_devconf.rtr_solicit_interval,
4560                         .maxlen         = sizeof(int),
4561                         .mode           = 0644,
4562                         .proc_handler   = proc_dointvec_jiffies,
4563                 },
4564                 {
4565                         .procname       = "router_solicitation_delay",
4566                         .data           = &ipv6_devconf.rtr_solicit_delay,
4567                         .maxlen         = sizeof(int),
4568                         .mode           = 0644,
4569                         .proc_handler   = proc_dointvec_jiffies,
4570                 },
4571                 {
4572                         .procname       = "force_mld_version",
4573                         .data           = &ipv6_devconf.force_mld_version,
4574                         .maxlen         = sizeof(int),
4575                         .mode           = 0644,
4576                         .proc_handler   = proc_dointvec,
4577                 },
4578 #ifdef CONFIG_IPV6_PRIVACY
4579                 {
4580                         .procname       = "use_tempaddr",
4581                         .data           = &ipv6_devconf.use_tempaddr,
4582                         .maxlen         = sizeof(int),
4583                         .mode           = 0644,
4584                         .proc_handler   = proc_dointvec,
4585                 },
4586                 {
4587                         .procname       = "temp_valid_lft",
4588                         .data           = &ipv6_devconf.temp_valid_lft,
4589                         .maxlen         = sizeof(int),
4590                         .mode           = 0644,
4591                         .proc_handler   = proc_dointvec,
4592                 },
4593                 {
4594                         .procname       = "temp_prefered_lft",
4595                         .data           = &ipv6_devconf.temp_prefered_lft,
4596                         .maxlen         = sizeof(int),
4597                         .mode           = 0644,
4598                         .proc_handler   = proc_dointvec,
4599                 },
4600                 {
4601                         .procname       = "regen_max_retry",
4602                         .data           = &ipv6_devconf.regen_max_retry,
4603                         .maxlen         = sizeof(int),
4604                         .mode           = 0644,
4605                         .proc_handler   = proc_dointvec,
4606                 },
4607                 {
4608                         .procname       = "max_desync_factor",
4609                         .data           = &ipv6_devconf.max_desync_factor,
4610                         .maxlen         = sizeof(int),
4611                         .mode           = 0644,
4612                         .proc_handler   = proc_dointvec,
4613                 },
4614 #endif
4615                 {
4616                         .procname       = "max_addresses",
4617                         .data           = &ipv6_devconf.max_addresses,
4618                         .maxlen         = sizeof(int),
4619                         .mode           = 0644,
4620                         .proc_handler   = proc_dointvec,
4621                 },
4622                 {
4623                         .procname       = "accept_ra_defrtr",
4624                         .data           = &ipv6_devconf.accept_ra_defrtr,
4625                         .maxlen         = sizeof(int),
4626                         .mode           = 0644,
4627                         .proc_handler   = proc_dointvec,
4628                 },
4629                 {
4630                         .procname       = "accept_ra_pinfo",
4631                         .data           = &ipv6_devconf.accept_ra_pinfo,
4632                         .maxlen         = sizeof(int),
4633                         .mode           = 0644,
4634                         .proc_handler   = proc_dointvec,
4635                 },
4636 #ifdef CONFIG_IPV6_ROUTER_PREF
4637                 {
4638                         .procname       = "accept_ra_rtr_pref",
4639                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4640                         .maxlen         = sizeof(int),
4641                         .mode           = 0644,
4642                         .proc_handler   = proc_dointvec,
4643                 },
4644                 {
4645                         .procname       = "router_probe_interval",
4646                         .data           = &ipv6_devconf.rtr_probe_interval,
4647                         .maxlen         = sizeof(int),
4648                         .mode           = 0644,
4649                         .proc_handler   = proc_dointvec_jiffies,
4650                 },
4651 #ifdef CONFIG_IPV6_ROUTE_INFO
4652                 {
4653                         .procname       = "accept_ra_rt_info_max_plen",
4654                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4655                         .maxlen         = sizeof(int),
4656                         .mode           = 0644,
4657                         .proc_handler   = proc_dointvec,
4658                 },
4659 #endif
4660 #endif
4661                 {
4662                         .procname       = "proxy_ndp",
4663                         .data           = &ipv6_devconf.proxy_ndp,
4664                         .maxlen         = sizeof(int),
4665                         .mode           = 0644,
4666                         .proc_handler   = proc_dointvec,
4667                 },
4668                 {
4669                         .procname       = "accept_source_route",
4670                         .data           = &ipv6_devconf.accept_source_route,
4671                         .maxlen         = sizeof(int),
4672                         .mode           = 0644,
4673                         .proc_handler   = proc_dointvec,
4674                 },
4675 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4676                 {
4677                         .procname       = "optimistic_dad",
4678                         .data           = &ipv6_devconf.optimistic_dad,
4679                         .maxlen         = sizeof(int),
4680                         .mode           = 0644,
4681                         .proc_handler   = proc_dointvec,
4682
4683                 },
4684 #endif
4685 #ifdef CONFIG_IPV6_MROUTE
4686                 {
4687                         .procname       = "mc_forwarding",
4688                         .data           = &ipv6_devconf.mc_forwarding,
4689                         .maxlen         = sizeof(int),
4690                         .mode           = 0444,
4691                         .proc_handler   = proc_dointvec,
4692                 },
4693 #endif
4694                 {
4695                         .procname       = "disable_ipv6",
4696                         .data           = &ipv6_devconf.disable_ipv6,
4697                         .maxlen         = sizeof(int),
4698                         .mode           = 0644,
4699                         .proc_handler   = addrconf_sysctl_disable,
4700                 },
4701                 {
4702                         .procname       = "accept_dad",
4703                         .data           = &ipv6_devconf.accept_dad,
4704                         .maxlen         = sizeof(int),
4705                         .mode           = 0644,
4706                         .proc_handler   = proc_dointvec,
4707                 },
4708                 {
4709                         .procname       = "force_tllao",
4710                         .data           = &ipv6_devconf.force_tllao,
4711                         .maxlen         = sizeof(int),
4712                         .mode           = 0644,
4713                         .proc_handler   = proc_dointvec
4714                 },
4715                 {
4716                         .procname       = "ndisc_notify",
4717                         .data           = &ipv6_devconf.ndisc_notify,
4718                         .maxlen         = sizeof(int),
4719                         .mode           = 0644,
4720                         .proc_handler   = proc_dointvec
4721                 },
4722                 {
4723                         /* sentinel */
4724                 }
4725         },
4726 };
4727
4728 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4729                 struct inet6_dev *idev, struct ipv6_devconf *p)
4730 {
4731         int i;
4732         struct addrconf_sysctl_table *t;
4733         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4734
4735         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4736         if (t == NULL)
4737                 goto out;
4738
4739         for (i = 0; t->addrconf_vars[i].data; i++) {
4740                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4741                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4742                 t->addrconf_vars[i].extra2 = net;
4743         }
4744
4745         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4746
4747         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4748         if (t->sysctl_header == NULL)
4749                 goto free;
4750
4751         p->sysctl = t;
4752         return 0;
4753
4754 free:
4755         kfree(t);
4756 out:
4757         return -ENOBUFS;
4758 }
4759
4760 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4761 {
4762         struct addrconf_sysctl_table *t;
4763
4764         if (p->sysctl == NULL)
4765                 return;
4766
4767         t = p->sysctl;
4768         p->sysctl = NULL;
4769         unregister_net_sysctl_table(t->sysctl_header);
4770         kfree(t);
4771 }
4772
4773 static void addrconf_sysctl_register(struct inet6_dev *idev)
4774 {
4775         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4776                               &ndisc_ifinfo_sysctl_change);
4777         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4778                                         idev, &idev->cnf);
4779 }
4780
4781 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4782 {
4783         __addrconf_sysctl_unregister(&idev->cnf);
4784         neigh_sysctl_unregister(idev->nd_parms);
4785 }
4786
4787
4788 #endif
4789
4790 static int __net_init addrconf_init_net(struct net *net)
4791 {
4792         int err;
4793         struct ipv6_devconf *all, *dflt;
4794
4795         err = -ENOMEM;
4796         all = &ipv6_devconf;
4797         dflt = &ipv6_devconf_dflt;
4798
4799         if (!net_eq(net, &init_net)) {
4800                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4801                 if (all == NULL)
4802                         goto err_alloc_all;
4803
4804                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4805                 if (dflt == NULL)
4806                         goto err_alloc_dflt;
4807         } else {
4808                 /* these will be inherited by all namespaces */
4809                 dflt->autoconf = ipv6_defaults.autoconf;
4810                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4811         }
4812
4813         net->ipv6.devconf_all = all;
4814         net->ipv6.devconf_dflt = dflt;
4815
4816 #ifdef CONFIG_SYSCTL
4817         err = __addrconf_sysctl_register(net, "all", NULL, all);
4818         if (err < 0)
4819                 goto err_reg_all;
4820
4821         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4822         if (err < 0)
4823                 goto err_reg_dflt;
4824 #endif
4825         return 0;
4826
4827 #ifdef CONFIG_SYSCTL
4828 err_reg_dflt:
4829         __addrconf_sysctl_unregister(all);
4830 err_reg_all:
4831         kfree(dflt);
4832 #endif
4833 err_alloc_dflt:
4834         kfree(all);
4835 err_alloc_all:
4836         return err;
4837 }
4838
4839 static void __net_exit addrconf_exit_net(struct net *net)
4840 {
4841 #ifdef CONFIG_SYSCTL
4842         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4843         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4844 #endif
4845         if (!net_eq(net, &init_net)) {
4846                 kfree(net->ipv6.devconf_dflt);
4847                 kfree(net->ipv6.devconf_all);
4848         }
4849 }
4850
4851 static struct pernet_operations addrconf_ops = {
4852         .init = addrconf_init_net,
4853         .exit = addrconf_exit_net,
4854 };
4855
4856 /*
4857  *      Device notifier
4858  */
4859
4860 int register_inet6addr_notifier(struct notifier_block *nb)
4861 {
4862         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4863 }
4864 EXPORT_SYMBOL(register_inet6addr_notifier);
4865
4866 int unregister_inet6addr_notifier(struct notifier_block *nb)
4867 {
4868         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4869 }
4870 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4871
4872 static struct rtnl_af_ops inet6_ops = {
4873         .family           = AF_INET6,
4874         .fill_link_af     = inet6_fill_link_af,
4875         .get_link_af_size = inet6_get_link_af_size,
4876 };
4877
4878 /*
4879  *      Init / cleanup code
4880  */
4881
4882 int __init addrconf_init(void)
4883 {
4884         int i, err;
4885
4886         err = ipv6_addr_label_init();
4887         if (err < 0) {
4888                 pr_crit("%s: cannot initialize default policy table: %d\n",
4889                         __func__, err);
4890                 goto out;
4891         }
4892
4893         err = register_pernet_subsys(&addrconf_ops);
4894         if (err < 0)
4895                 goto out_addrlabel;
4896
4897         /* The addrconf netdev notifier requires that loopback_dev
4898          * has it's ipv6 private information allocated and setup
4899          * before it can bring up and give link-local addresses
4900          * to other devices which are up.
4901          *
4902          * Unfortunately, loopback_dev is not necessarily the first
4903          * entry in the global dev_base list of net devices.  In fact,
4904          * it is likely to be the very last entry on that list.
4905          * So this causes the notifier registry below to try and
4906          * give link-local addresses to all devices besides loopback_dev
4907          * first, then loopback_dev, which cases all the non-loopback_dev
4908          * devices to fail to get a link-local address.
4909          *
4910          * So, as a temporary fix, allocate the ipv6 structure for
4911          * loopback_dev first by hand.
4912          * Longer term, all of the dependencies ipv6 has upon the loopback
4913          * device and it being up should be removed.
4914          */
4915         rtnl_lock();
4916         if (!ipv6_add_dev(init_net.loopback_dev))
4917                 err = -ENOMEM;
4918         rtnl_unlock();
4919         if (err)
4920                 goto errlo;
4921
4922         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4923                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4924
4925         register_netdevice_notifier(&ipv6_dev_notf);
4926
4927         addrconf_verify(0);
4928
4929         err = rtnl_af_register(&inet6_ops);
4930         if (err < 0)
4931                 goto errout_af;
4932
4933         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4934                               NULL);
4935         if (err < 0)
4936                 goto errout;
4937
4938         /* Only the first call to __rtnl_register can fail */
4939         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4940         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4941         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4942                         inet6_dump_ifaddr, NULL);
4943         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4944                         inet6_dump_ifmcaddr, NULL);
4945         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4946                         inet6_dump_ifacaddr, NULL);
4947         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4948                         NULL, NULL);
4949
4950         ipv6_addr_label_rtnl_register();
4951
4952         return 0;
4953 errout:
4954         rtnl_af_unregister(&inet6_ops);
4955 errout_af:
4956         unregister_netdevice_notifier(&ipv6_dev_notf);
4957 errlo:
4958         unregister_pernet_subsys(&addrconf_ops);
4959 out_addrlabel:
4960         ipv6_addr_label_cleanup();
4961 out:
4962         return err;
4963 }
4964
4965 void addrconf_cleanup(void)
4966 {
4967         struct net_device *dev;
4968         int i;
4969
4970         unregister_netdevice_notifier(&ipv6_dev_notf);
4971         unregister_pernet_subsys(&addrconf_ops);
4972         ipv6_addr_label_cleanup();
4973
4974         rtnl_lock();
4975
4976         __rtnl_af_unregister(&inet6_ops);
4977
4978         /* clean dev list */
4979         for_each_netdev(&init_net, dev) {
4980                 if (__in6_dev_get(dev) == NULL)
4981                         continue;
4982                 addrconf_ifdown(dev, 1);
4983         }
4984         addrconf_ifdown(init_net.loopback_dev, 2);
4985
4986         /*
4987          *      Check hash table.
4988          */
4989         spin_lock_bh(&addrconf_hash_lock);
4990         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4991                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4992         spin_unlock_bh(&addrconf_hash_lock);
4993
4994         del_timer(&addr_chk_timer);
4995         rtnl_unlock();
4996 }