]> Pileus Git - ~andy/linux/blob - net/ipv6/addrconf.c
4dc0d44a5d31cb54788b7d2df076617b46b4e2f5
[~andy/linux] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/ipv6.h>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
78 #include <net/tcp.h>
79 #include <net/ip.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
88 #endif
89
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
99
100 #if ACONF_DEBUG >= 3
101 #define ADBG(x) printk x
102 #else
103 #define ADBG(x)
104 #endif
105
106 #define INFINITY_LIFE_TIME      0xFFFFFFFF
107
108 static inline u32 cstamp_delta(unsigned long cstamp)
109 {
110         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 }
112
113 #ifdef CONFIG_SYSCTL
114 static void addrconf_sysctl_register(struct inet6_dev *idev);
115 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
116 #else
117 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
118 {
119 }
120
121 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
122 {
123 }
124 #endif
125
126 #ifdef CONFIG_IPV6_PRIVACY
127 static void __ipv6_regen_rndid(struct inet6_dev *idev);
128 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
129 static void ipv6_regen_rndid(unsigned long data);
130 #endif
131
132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
133 static int ipv6_count_addresses(struct inet6_dev *idev);
134
135 /*
136  *      Configured unicast address hash table
137  */
138 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
139 static DEFINE_SPINLOCK(addrconf_hash_lock);
140
141 static void addrconf_verify(unsigned long);
142
143 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
144 static DEFINE_SPINLOCK(addrconf_verify_lock);
145
146 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
148
149 static void addrconf_type_change(struct net_device *dev,
150                                  unsigned long event);
151 static int addrconf_ifdown(struct net_device *dev, int how);
152
153 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
154                                                   int plen,
155                                                   const struct net_device *dev,
156                                                   u32 flags, u32 noflags);
157
158 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_timer(unsigned long data);
160 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
161 static void addrconf_dad_run(struct inet6_dev *idev);
162 static void addrconf_rs_timer(unsigned long data);
163 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
165
166 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
167                                 struct prefix_info *pinfo);
168 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
169                                struct net_device *dev);
170
171 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
172
173 static struct ipv6_devconf ipv6_devconf __read_mostly = {
174         .forwarding             = 0,
175         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
176         .mtu6                   = IPV6_MIN_MTU,
177         .accept_ra              = 1,
178         .accept_redirects       = 1,
179         .autoconf               = 1,
180         .force_mld_version      = 0,
181         .dad_transmits          = 1,
182         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
183         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
184         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
185 #ifdef CONFIG_IPV6_PRIVACY
186         .use_tempaddr           = 0,
187         .temp_valid_lft         = TEMP_VALID_LIFETIME,
188         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
189         .regen_max_retry        = REGEN_MAX_RETRY,
190         .max_desync_factor      = MAX_DESYNC_FACTOR,
191 #endif
192         .max_addresses          = IPV6_MAX_ADDRESSES,
193         .accept_ra_defrtr       = 1,
194         .accept_ra_pinfo        = 1,
195 #ifdef CONFIG_IPV6_ROUTER_PREF
196         .accept_ra_rtr_pref     = 1,
197         .rtr_probe_interval     = 60 * HZ,
198 #ifdef CONFIG_IPV6_ROUTE_INFO
199         .accept_ra_rt_info_max_plen = 0,
200 #endif
201 #endif
202         .proxy_ndp              = 0,
203         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
204         .disable_ipv6           = 0,
205         .accept_dad             = 1,
206 };
207
208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
209         .forwarding             = 0,
210         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
211         .mtu6                   = IPV6_MIN_MTU,
212         .accept_ra              = 1,
213         .accept_redirects       = 1,
214         .autoconf               = 1,
215         .dad_transmits          = 1,
216         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
217         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
218         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
219 #ifdef CONFIG_IPV6_PRIVACY
220         .use_tempaddr           = 0,
221         .temp_valid_lft         = TEMP_VALID_LIFETIME,
222         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
223         .regen_max_retry        = REGEN_MAX_RETRY,
224         .max_desync_factor      = MAX_DESYNC_FACTOR,
225 #endif
226         .max_addresses          = IPV6_MAX_ADDRESSES,
227         .accept_ra_defrtr       = 1,
228         .accept_ra_pinfo        = 1,
229 #ifdef CONFIG_IPV6_ROUTER_PREF
230         .accept_ra_rtr_pref     = 1,
231         .rtr_probe_interval     = 60 * HZ,
232 #ifdef CONFIG_IPV6_ROUTE_INFO
233         .accept_ra_rt_info_max_plen = 0,
234 #endif
235 #endif
236         .proxy_ndp              = 0,
237         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
238         .disable_ipv6           = 0,
239         .accept_dad             = 1,
240 };
241
242 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
243 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
244 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
245 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
246 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
247 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
248 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
249 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
250
251 /* Check if a valid qdisc is available */
252 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
253 {
254         return !qdisc_tx_is_noop(dev);
255 }
256
257 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
258 {
259         if (del_timer(&ifp->timer))
260                 __in6_ifa_put(ifp);
261 }
262
263 enum addrconf_timer_t {
264         AC_NONE,
265         AC_DAD,
266         AC_RS,
267 };
268
269 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
270                                enum addrconf_timer_t what,
271                                unsigned long when)
272 {
273         if (!del_timer(&ifp->timer))
274                 in6_ifa_hold(ifp);
275
276         switch (what) {
277         case AC_DAD:
278                 ifp->timer.function = addrconf_dad_timer;
279                 break;
280         case AC_RS:
281                 ifp->timer.function = addrconf_rs_timer;
282                 break;
283         default:
284                 break;
285         }
286         ifp->timer.expires = jiffies + when;
287         add_timer(&ifp->timer);
288 }
289
290 static int snmp6_alloc_dev(struct inet6_dev *idev)
291 {
292         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
293                           sizeof(struct ipstats_mib),
294                           __alignof__(struct ipstats_mib)) < 0)
295                 goto err_ip;
296         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
297                                         GFP_KERNEL);
298         if (!idev->stats.icmpv6dev)
299                 goto err_icmp;
300         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
301                                            GFP_KERNEL);
302         if (!idev->stats.icmpv6msgdev)
303                 goto err_icmpmsg;
304
305         return 0;
306
307 err_icmpmsg:
308         kfree(idev->stats.icmpv6dev);
309 err_icmp:
310         snmp_mib_free((void __percpu **)idev->stats.ipv6);
311 err_ip:
312         return -ENOMEM;
313 }
314
315 static void snmp6_free_dev(struct inet6_dev *idev)
316 {
317         kfree(idev->stats.icmpv6msgdev);
318         kfree(idev->stats.icmpv6dev);
319         snmp_mib_free((void __percpu **)idev->stats.ipv6);
320 }
321
322 /* Nobody refers to this device, we may destroy it. */
323
324 void in6_dev_finish_destroy(struct inet6_dev *idev)
325 {
326         struct net_device *dev = idev->dev;
327
328         WARN_ON(!list_empty(&idev->addr_list));
329         WARN_ON(idev->mc_list != NULL);
330
331 #ifdef NET_REFCNT_DEBUG
332         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
333 #endif
334         dev_put(dev);
335         if (!idev->dead) {
336                 pr_warn("Freeing alive inet6 device %p\n", idev);
337                 return;
338         }
339         snmp6_free_dev(idev);
340         kfree_rcu(idev, rcu);
341 }
342 EXPORT_SYMBOL(in6_dev_finish_destroy);
343
344 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
345 {
346         struct inet6_dev *ndev;
347
348         ASSERT_RTNL();
349
350         if (dev->mtu < IPV6_MIN_MTU)
351                 return NULL;
352
353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
354
355         if (ndev == NULL)
356                 return NULL;
357
358         rwlock_init(&ndev->lock);
359         ndev->dev = dev;
360         INIT_LIST_HEAD(&ndev->addr_list);
361
362         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
363         ndev->cnf.mtu6 = dev->mtu;
364         ndev->cnf.sysctl = NULL;
365         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
366         if (ndev->nd_parms == NULL) {
367                 kfree(ndev);
368                 return NULL;
369         }
370         if (ndev->cnf.forwarding)
371                 dev_disable_lro(dev);
372         /* We refer to the device */
373         dev_hold(dev);
374
375         if (snmp6_alloc_dev(ndev) < 0) {
376                 ADBG((KERN_WARNING
377                         "%s: cannot allocate memory for statistics; dev=%s.\n",
378                         __func__, dev->name));
379                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
380                 dev_put(dev);
381                 kfree(ndev);
382                 return NULL;
383         }
384
385         if (snmp6_register_dev(ndev) < 0) {
386                 ADBG((KERN_WARNING
387                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
388                         __func__, dev->name));
389                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
390                 ndev->dead = 1;
391                 in6_dev_finish_destroy(ndev);
392                 return NULL;
393         }
394
395         /* One reference from device.  We must do this before
396          * we invoke __ipv6_regen_rndid().
397          */
398         in6_dev_hold(ndev);
399
400         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
401                 ndev->cnf.accept_dad = -1;
402
403 #if IS_ENABLED(CONFIG_IPV6_SIT)
404         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
405                 pr_info("%s: Disabled Multicast RS\n", dev->name);
406                 ndev->cnf.rtr_solicits = 0;
407         }
408 #endif
409
410 #ifdef CONFIG_IPV6_PRIVACY
411         INIT_LIST_HEAD(&ndev->tempaddr_list);
412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413         if ((dev->flags&IFF_LOOPBACK) ||
414             dev->type == ARPHRD_TUNNEL ||
415             dev->type == ARPHRD_TUNNEL6 ||
416             dev->type == ARPHRD_SIT ||
417             dev->type == ARPHRD_NONE) {
418                 ndev->cnf.use_tempaddr = -1;
419         } else {
420                 in6_dev_hold(ndev);
421                 ipv6_regen_rndid((unsigned long) ndev);
422         }
423 #endif
424
425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
426                 ndev->if_flags |= IF_READY;
427
428         ipv6_mc_init_dev(ndev);
429         ndev->tstamp = jiffies;
430         addrconf_sysctl_register(ndev);
431         /* protected by rtnl_lock */
432         rcu_assign_pointer(dev->ip6_ptr, ndev);
433
434         /* Join interface-local all-node multicast group */
435         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
436
437         /* Join all-node multicast group */
438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439
440         /* Join all-router multicast group if forwarding is set */
441         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
442                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
443
444         return ndev;
445 }
446
447 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
448 {
449         struct inet6_dev *idev;
450
451         ASSERT_RTNL();
452
453         idev = __in6_dev_get(dev);
454         if (!idev) {
455                 idev = ipv6_add_dev(dev);
456                 if (!idev)
457                         return NULL;
458         }
459
460         if (dev->flags&IFF_UP)
461                 ipv6_mc_up(idev);
462         return idev;
463 }
464
465 static int inet6_netconf_msgsize_devconf(int type)
466 {
467         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
468                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
469
470         /* type -1 is used for ALL */
471         if (type == -1 || type == NETCONFA_FORWARDING)
472                 size += nla_total_size(4);
473 #ifdef CONFIG_IPV6_MROUTE
474         if (type == -1 || type == NETCONFA_MC_FORWARDING)
475                 size += nla_total_size(4);
476 #endif
477
478         return size;
479 }
480
481 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
482                                       struct ipv6_devconf *devconf, u32 portid,
483                                       u32 seq, int event, unsigned int flags,
484                                       int type)
485 {
486         struct nlmsghdr  *nlh;
487         struct netconfmsg *ncm;
488
489         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
490                         flags);
491         if (nlh == NULL)
492                 return -EMSGSIZE;
493
494         ncm = nlmsg_data(nlh);
495         ncm->ncm_family = AF_INET6;
496
497         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
498                 goto nla_put_failure;
499
500         /* type -1 is used for ALL */
501         if ((type == -1 || type == NETCONFA_FORWARDING) &&
502             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
503                 goto nla_put_failure;
504 #ifdef CONFIG_IPV6_MROUTE
505         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
506             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
507                         devconf->mc_forwarding) < 0)
508                 goto nla_put_failure;
509 #endif
510         return nlmsg_end(skb, nlh);
511
512 nla_put_failure:
513         nlmsg_cancel(skb, nlh);
514         return -EMSGSIZE;
515 }
516
517 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
518                                   struct ipv6_devconf *devconf)
519 {
520         struct sk_buff *skb;
521         int err = -ENOBUFS;
522
523         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
524         if (skb == NULL)
525                 goto errout;
526
527         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
528                                          RTM_NEWNETCONF, 0, type);
529         if (err < 0) {
530                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
531                 WARN_ON(err == -EMSGSIZE);
532                 kfree_skb(skb);
533                 goto errout;
534         }
535         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
536         return;
537 errout:
538         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
539 }
540
541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
542         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
543         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
544 };
545
546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
547                                      struct nlmsghdr *nlh,
548                                      void *arg)
549 {
550         struct net *net = sock_net(in_skb->sk);
551         struct nlattr *tb[NETCONFA_MAX+1];
552         struct netconfmsg *ncm;
553         struct sk_buff *skb;
554         struct ipv6_devconf *devconf;
555         struct inet6_dev *in6_dev;
556         struct net_device *dev;
557         int ifindex;
558         int err;
559
560         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
561                           devconf_ipv6_policy);
562         if (err < 0)
563                 goto errout;
564
565         err = EINVAL;
566         if (!tb[NETCONFA_IFINDEX])
567                 goto errout;
568
569         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
570         switch (ifindex) {
571         case NETCONFA_IFINDEX_ALL:
572                 devconf = net->ipv6.devconf_all;
573                 break;
574         case NETCONFA_IFINDEX_DEFAULT:
575                 devconf = net->ipv6.devconf_dflt;
576                 break;
577         default:
578                 dev = __dev_get_by_index(net, ifindex);
579                 if (dev == NULL)
580                         goto errout;
581                 in6_dev = __in6_dev_get(dev);
582                 if (in6_dev == NULL)
583                         goto errout;
584                 devconf = &in6_dev->cnf;
585                 break;
586         }
587
588         err = -ENOBUFS;
589         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
590         if (skb == NULL)
591                 goto errout;
592
593         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
594                                          NETLINK_CB(in_skb).portid,
595                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
596                                          -1);
597         if (err < 0) {
598                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
599                 WARN_ON(err == -EMSGSIZE);
600                 kfree_skb(skb);
601                 goto errout;
602         }
603         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
604 errout:
605         return err;
606 }
607
608 #ifdef CONFIG_SYSCTL
609 static void dev_forward_change(struct inet6_dev *idev)
610 {
611         struct net_device *dev;
612         struct inet6_ifaddr *ifa;
613
614         if (!idev)
615                 return;
616         dev = idev->dev;
617         if (idev->cnf.forwarding)
618                 dev_disable_lro(dev);
619         if (dev->flags & IFF_MULTICAST) {
620                 if (idev->cnf.forwarding) {
621                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
622                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
623                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
624                 } else {
625                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
626                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
627                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
628                 }
629         }
630
631         list_for_each_entry(ifa, &idev->addr_list, if_list) {
632                 if (ifa->flags&IFA_F_TENTATIVE)
633                         continue;
634                 if (idev->cnf.forwarding)
635                         addrconf_join_anycast(ifa);
636                 else
637                         addrconf_leave_anycast(ifa);
638         }
639         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
640                                      dev->ifindex, &idev->cnf);
641 }
642
643
644 static void addrconf_forward_change(struct net *net, __s32 newf)
645 {
646         struct net_device *dev;
647         struct inet6_dev *idev;
648
649         for_each_netdev(net, dev) {
650                 idev = __in6_dev_get(dev);
651                 if (idev) {
652                         int changed = (!idev->cnf.forwarding) ^ (!newf);
653                         idev->cnf.forwarding = newf;
654                         if (changed)
655                                 dev_forward_change(idev);
656                 }
657         }
658 }
659
660 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
661 {
662         struct net *net;
663         int old;
664
665         if (!rtnl_trylock())
666                 return restart_syscall();
667
668         net = (struct net *)table->extra2;
669         old = *p;
670         *p = newf;
671
672         if (p == &net->ipv6.devconf_dflt->forwarding) {
673                 if ((!newf) ^ (!old))
674                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
675                                                      NETCONFA_IFINDEX_DEFAULT,
676                                                      net->ipv6.devconf_dflt);
677                 rtnl_unlock();
678                 return 0;
679         }
680
681         if (p == &net->ipv6.devconf_all->forwarding) {
682                 net->ipv6.devconf_dflt->forwarding = newf;
683                 addrconf_forward_change(net, newf);
684                 if ((!newf) ^ (!old))
685                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
686                                                      NETCONFA_IFINDEX_ALL,
687                                                      net->ipv6.devconf_all);
688         } else if ((!newf) ^ (!old))
689                 dev_forward_change((struct inet6_dev *)table->extra1);
690         rtnl_unlock();
691
692         if (newf)
693                 rt6_purge_dflt_routers(net);
694         return 1;
695 }
696 #endif
697
698 /* Nobody refers to this ifaddr, destroy it */
699 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
700 {
701         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
702
703 #ifdef NET_REFCNT_DEBUG
704         pr_debug("%s\n", __func__);
705 #endif
706
707         in6_dev_put(ifp->idev);
708
709         if (del_timer(&ifp->timer))
710                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
711
712         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
713                 pr_warn("Freeing alive inet6 address %p\n", ifp);
714                 return;
715         }
716         ip6_rt_put(ifp->rt);
717
718         kfree_rcu(ifp, rcu);
719 }
720
721 static void
722 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
723 {
724         struct list_head *p;
725         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
726
727         /*
728          * Each device address list is sorted in order of scope -
729          * global before linklocal.
730          */
731         list_for_each(p, &idev->addr_list) {
732                 struct inet6_ifaddr *ifa
733                         = list_entry(p, struct inet6_ifaddr, if_list);
734                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
735                         break;
736         }
737
738         list_add_tail(&ifp->if_list, p);
739 }
740
741 static u32 inet6_addr_hash(const struct in6_addr *addr)
742 {
743         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
744 }
745
746 /* On success it returns ifp with increased reference count */
747
748 static struct inet6_ifaddr *
749 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
750               int scope, u32 flags)
751 {
752         struct inet6_ifaddr *ifa = NULL;
753         struct rt6_info *rt;
754         unsigned int hash;
755         int err = 0;
756         int addr_type = ipv6_addr_type(addr);
757
758         if (addr_type == IPV6_ADDR_ANY ||
759             addr_type & IPV6_ADDR_MULTICAST ||
760             (!(idev->dev->flags & IFF_LOOPBACK) &&
761              addr_type & IPV6_ADDR_LOOPBACK))
762                 return ERR_PTR(-EADDRNOTAVAIL);
763
764         rcu_read_lock_bh();
765         if (idev->dead) {
766                 err = -ENODEV;                  /*XXX*/
767                 goto out2;
768         }
769
770         if (idev->cnf.disable_ipv6) {
771                 err = -EACCES;
772                 goto out2;
773         }
774
775         spin_lock(&addrconf_hash_lock);
776
777         /* Ignore adding duplicate addresses on an interface */
778         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
779                 ADBG(("ipv6_add_addr: already assigned\n"));
780                 err = -EEXIST;
781                 goto out;
782         }
783
784         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
785
786         if (ifa == NULL) {
787                 ADBG(("ipv6_add_addr: malloc failed\n"));
788                 err = -ENOBUFS;
789                 goto out;
790         }
791
792         rt = addrconf_dst_alloc(idev, addr, false);
793         if (IS_ERR(rt)) {
794                 err = PTR_ERR(rt);
795                 goto out;
796         }
797
798         ifa->addr = *addr;
799
800         spin_lock_init(&ifa->lock);
801         spin_lock_init(&ifa->state_lock);
802         init_timer(&ifa->timer);
803         INIT_HLIST_NODE(&ifa->addr_lst);
804         ifa->timer.data = (unsigned long) ifa;
805         ifa->scope = scope;
806         ifa->prefix_len = pfxlen;
807         ifa->flags = flags | IFA_F_TENTATIVE;
808         ifa->cstamp = ifa->tstamp = jiffies;
809
810         ifa->rt = rt;
811
812         ifa->idev = idev;
813         in6_dev_hold(idev);
814         /* For caller */
815         in6_ifa_hold(ifa);
816
817         /* Add to big hash table */
818         hash = inet6_addr_hash(addr);
819
820         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
821         spin_unlock(&addrconf_hash_lock);
822
823         write_lock(&idev->lock);
824         /* Add to inet6_dev unicast addr list. */
825         ipv6_link_dev_addr(idev, ifa);
826
827 #ifdef CONFIG_IPV6_PRIVACY
828         if (ifa->flags&IFA_F_TEMPORARY) {
829                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
830                 in6_ifa_hold(ifa);
831         }
832 #endif
833
834         in6_ifa_hold(ifa);
835         write_unlock(&idev->lock);
836 out2:
837         rcu_read_unlock_bh();
838
839         if (likely(err == 0))
840                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
841         else {
842                 kfree(ifa);
843                 ifa = ERR_PTR(err);
844         }
845
846         return ifa;
847 out:
848         spin_unlock(&addrconf_hash_lock);
849         goto out2;
850 }
851
852 /* This function wants to get referenced ifp and releases it before return */
853
854 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
855 {
856         struct inet6_ifaddr *ifa, *ifn;
857         struct inet6_dev *idev = ifp->idev;
858         int state;
859         int deleted = 0, onlink = 0;
860         unsigned long expires = jiffies;
861
862         spin_lock_bh(&ifp->state_lock);
863         state = ifp->state;
864         ifp->state = INET6_IFADDR_STATE_DEAD;
865         spin_unlock_bh(&ifp->state_lock);
866
867         if (state == INET6_IFADDR_STATE_DEAD)
868                 goto out;
869
870         spin_lock_bh(&addrconf_hash_lock);
871         hlist_del_init_rcu(&ifp->addr_lst);
872         spin_unlock_bh(&addrconf_hash_lock);
873
874         write_lock_bh(&idev->lock);
875 #ifdef CONFIG_IPV6_PRIVACY
876         if (ifp->flags&IFA_F_TEMPORARY) {
877                 list_del(&ifp->tmp_list);
878                 if (ifp->ifpub) {
879                         in6_ifa_put(ifp->ifpub);
880                         ifp->ifpub = NULL;
881                 }
882                 __in6_ifa_put(ifp);
883         }
884 #endif
885
886         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
887                 if (ifa == ifp) {
888                         list_del_init(&ifp->if_list);
889                         __in6_ifa_put(ifp);
890
891                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
892                                 break;
893                         deleted = 1;
894                         continue;
895                 } else if (ifp->flags & IFA_F_PERMANENT) {
896                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
897                                               ifp->prefix_len)) {
898                                 if (ifa->flags & IFA_F_PERMANENT) {
899                                         onlink = 1;
900                                         if (deleted)
901                                                 break;
902                                 } else {
903                                         unsigned long lifetime;
904
905                                         if (!onlink)
906                                                 onlink = -1;
907
908                                         spin_lock(&ifa->lock);
909
910                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
911                                         /*
912                                          * Note: Because this address is
913                                          * not permanent, lifetime <
914                                          * LONG_MAX / HZ here.
915                                          */
916                                         if (time_before(expires,
917                                                         ifa->tstamp + lifetime * HZ))
918                                                 expires = ifa->tstamp + lifetime * HZ;
919                                         spin_unlock(&ifa->lock);
920                                 }
921                         }
922                 }
923         }
924         write_unlock_bh(&idev->lock);
925
926         addrconf_del_timer(ifp);
927
928         ipv6_ifa_notify(RTM_DELADDR, ifp);
929
930         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
931
932         /*
933          * Purge or update corresponding prefix
934          *
935          * 1) we don't purge prefix here if address was not permanent.
936          *    prefix is managed by its own lifetime.
937          * 2) if there're no addresses, delete prefix.
938          * 3) if there're still other permanent address(es),
939          *    corresponding prefix is still permanent.
940          * 4) otherwise, update prefix lifetime to the
941          *    longest valid lifetime among the corresponding
942          *    addresses on the device.
943          *    Note: subsequent RA will update lifetime.
944          *
945          * --yoshfuji
946          */
947         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
948                 struct in6_addr prefix;
949                 struct rt6_info *rt;
950
951                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
952
953                 rt = addrconf_get_prefix_route(&prefix,
954                                                ifp->prefix_len,
955                                                ifp->idev->dev,
956                                                0, RTF_GATEWAY | RTF_DEFAULT);
957
958                 if (rt) {
959                         if (onlink == 0) {
960                                 ip6_del_rt(rt);
961                                 rt = NULL;
962                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
963                                 rt6_set_expires(rt, expires);
964                         }
965                 }
966                 ip6_rt_put(rt);
967         }
968
969         /* clean up prefsrc entries */
970         rt6_remove_prefsrc(ifp);
971 out:
972         in6_ifa_put(ifp);
973 }
974
975 #ifdef CONFIG_IPV6_PRIVACY
976 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
977 {
978         struct inet6_dev *idev = ifp->idev;
979         struct in6_addr addr, *tmpaddr;
980         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
981         unsigned long regen_advance;
982         int tmp_plen;
983         int ret = 0;
984         int max_addresses;
985         u32 addr_flags;
986         unsigned long now = jiffies;
987
988         write_lock(&idev->lock);
989         if (ift) {
990                 spin_lock_bh(&ift->lock);
991                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
992                 spin_unlock_bh(&ift->lock);
993                 tmpaddr = &addr;
994         } else {
995                 tmpaddr = NULL;
996         }
997 retry:
998         in6_dev_hold(idev);
999         if (idev->cnf.use_tempaddr <= 0) {
1000                 write_unlock(&idev->lock);
1001                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1002                 in6_dev_put(idev);
1003                 ret = -1;
1004                 goto out;
1005         }
1006         spin_lock_bh(&ifp->lock);
1007         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1008                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1009                 spin_unlock_bh(&ifp->lock);
1010                 write_unlock(&idev->lock);
1011                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1012                         __func__);
1013                 in6_dev_put(idev);
1014                 ret = -1;
1015                 goto out;
1016         }
1017         in6_ifa_hold(ifp);
1018         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1019         __ipv6_try_regen_rndid(idev, tmpaddr);
1020         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1021         age = (now - ifp->tstamp) / HZ;
1022         tmp_valid_lft = min_t(__u32,
1023                               ifp->valid_lft,
1024                               idev->cnf.temp_valid_lft + age);
1025         tmp_prefered_lft = min_t(__u32,
1026                                  ifp->prefered_lft,
1027                                  idev->cnf.temp_prefered_lft + age -
1028                                  idev->cnf.max_desync_factor);
1029         tmp_plen = ifp->prefix_len;
1030         max_addresses = idev->cnf.max_addresses;
1031         tmp_tstamp = ifp->tstamp;
1032         spin_unlock_bh(&ifp->lock);
1033
1034         regen_advance = idev->cnf.regen_max_retry *
1035                         idev->cnf.dad_transmits *
1036                         idev->nd_parms->retrans_time / HZ;
1037         write_unlock(&idev->lock);
1038
1039         /* A temporary address is created only if this calculated Preferred
1040          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1041          * an implementation must not create a temporary address with a zero
1042          * Preferred Lifetime.
1043          */
1044         if (tmp_prefered_lft <= regen_advance) {
1045                 in6_ifa_put(ifp);
1046                 in6_dev_put(idev);
1047                 ret = -1;
1048                 goto out;
1049         }
1050
1051         addr_flags = IFA_F_TEMPORARY;
1052         /* set in addrconf_prefix_rcv() */
1053         if (ifp->flags & IFA_F_OPTIMISTIC)
1054                 addr_flags |= IFA_F_OPTIMISTIC;
1055
1056         ift = !max_addresses ||
1057               ipv6_count_addresses(idev) < max_addresses ?
1058                 ipv6_add_addr(idev, &addr, tmp_plen,
1059                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1060                               addr_flags) : NULL;
1061         if (IS_ERR_OR_NULL(ift)) {
1062                 in6_ifa_put(ifp);
1063                 in6_dev_put(idev);
1064                 pr_info("%s: retry temporary address regeneration\n", __func__);
1065                 tmpaddr = &addr;
1066                 write_lock(&idev->lock);
1067                 goto retry;
1068         }
1069
1070         spin_lock_bh(&ift->lock);
1071         ift->ifpub = ifp;
1072         ift->valid_lft = tmp_valid_lft;
1073         ift->prefered_lft = tmp_prefered_lft;
1074         ift->cstamp = now;
1075         ift->tstamp = tmp_tstamp;
1076         spin_unlock_bh(&ift->lock);
1077
1078         addrconf_dad_start(ift);
1079         in6_ifa_put(ift);
1080         in6_dev_put(idev);
1081 out:
1082         return ret;
1083 }
1084 #endif
1085
1086 /*
1087  *      Choose an appropriate source address (RFC3484)
1088  */
1089 enum {
1090         IPV6_SADDR_RULE_INIT = 0,
1091         IPV6_SADDR_RULE_LOCAL,
1092         IPV6_SADDR_RULE_SCOPE,
1093         IPV6_SADDR_RULE_PREFERRED,
1094 #ifdef CONFIG_IPV6_MIP6
1095         IPV6_SADDR_RULE_HOA,
1096 #endif
1097         IPV6_SADDR_RULE_OIF,
1098         IPV6_SADDR_RULE_LABEL,
1099 #ifdef CONFIG_IPV6_PRIVACY
1100         IPV6_SADDR_RULE_PRIVACY,
1101 #endif
1102         IPV6_SADDR_RULE_ORCHID,
1103         IPV6_SADDR_RULE_PREFIX,
1104         IPV6_SADDR_RULE_MAX
1105 };
1106
1107 struct ipv6_saddr_score {
1108         int                     rule;
1109         int                     addr_type;
1110         struct inet6_ifaddr     *ifa;
1111         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1112         int                     scopedist;
1113         int                     matchlen;
1114 };
1115
1116 struct ipv6_saddr_dst {
1117         const struct in6_addr *addr;
1118         int ifindex;
1119         int scope;
1120         int label;
1121         unsigned int prefs;
1122 };
1123
1124 static inline int ipv6_saddr_preferred(int type)
1125 {
1126         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1127                 return 1;
1128         return 0;
1129 }
1130
1131 static int ipv6_get_saddr_eval(struct net *net,
1132                                struct ipv6_saddr_score *score,
1133                                struct ipv6_saddr_dst *dst,
1134                                int i)
1135 {
1136         int ret;
1137
1138         if (i <= score->rule) {
1139                 switch (i) {
1140                 case IPV6_SADDR_RULE_SCOPE:
1141                         ret = score->scopedist;
1142                         break;
1143                 case IPV6_SADDR_RULE_PREFIX:
1144                         ret = score->matchlen;
1145                         break;
1146                 default:
1147                         ret = !!test_bit(i, score->scorebits);
1148                 }
1149                 goto out;
1150         }
1151
1152         switch (i) {
1153         case IPV6_SADDR_RULE_INIT:
1154                 /* Rule 0: remember if hiscore is not ready yet */
1155                 ret = !!score->ifa;
1156                 break;
1157         case IPV6_SADDR_RULE_LOCAL:
1158                 /* Rule 1: Prefer same address */
1159                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1160                 break;
1161         case IPV6_SADDR_RULE_SCOPE:
1162                 /* Rule 2: Prefer appropriate scope
1163                  *
1164                  *      ret
1165                  *       ^
1166                  *    -1 |  d 15
1167                  *    ---+--+-+---> scope
1168                  *       |
1169                  *       |             d is scope of the destination.
1170                  *  B-d  |  \
1171                  *       |   \      <- smaller scope is better if
1172                  *  B-15 |    \        if scope is enough for destinaion.
1173                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1174                  * d-C-1 | /
1175                  *       |/         <- greater is better
1176                  *   -C  /             if scope is not enough for destination.
1177                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1178                  *
1179                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1180                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1181                  * Assume B = 0 and we get C > 29.
1182                  */
1183                 ret = __ipv6_addr_src_scope(score->addr_type);
1184                 if (ret >= dst->scope)
1185                         ret = -ret;
1186                 else
1187                         ret -= 128;     /* 30 is enough */
1188                 score->scopedist = ret;
1189                 break;
1190         case IPV6_SADDR_RULE_PREFERRED:
1191                 /* Rule 3: Avoid deprecated and optimistic addresses */
1192                 ret = ipv6_saddr_preferred(score->addr_type) ||
1193                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1194                 break;
1195 #ifdef CONFIG_IPV6_MIP6
1196         case IPV6_SADDR_RULE_HOA:
1197             {
1198                 /* Rule 4: Prefer home address */
1199                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1200                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1201                 break;
1202             }
1203 #endif
1204         case IPV6_SADDR_RULE_OIF:
1205                 /* Rule 5: Prefer outgoing interface */
1206                 ret = (!dst->ifindex ||
1207                        dst->ifindex == score->ifa->idev->dev->ifindex);
1208                 break;
1209         case IPV6_SADDR_RULE_LABEL:
1210                 /* Rule 6: Prefer matching label */
1211                 ret = ipv6_addr_label(net,
1212                                       &score->ifa->addr, score->addr_type,
1213                                       score->ifa->idev->dev->ifindex) == dst->label;
1214                 break;
1215 #ifdef CONFIG_IPV6_PRIVACY
1216         case IPV6_SADDR_RULE_PRIVACY:
1217             {
1218                 /* Rule 7: Prefer public address
1219                  * Note: prefer temporary address if use_tempaddr >= 2
1220                  */
1221                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1222                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1223                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1224                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1225                 break;
1226             }
1227 #endif
1228         case IPV6_SADDR_RULE_ORCHID:
1229                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1230                  *          non-ORCHID vs non-ORCHID
1231                  */
1232                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1233                         ipv6_addr_orchid(dst->addr));
1234                 break;
1235         case IPV6_SADDR_RULE_PREFIX:
1236                 /* Rule 8: Use longest matching prefix */
1237                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1238                 if (ret > score->ifa->prefix_len)
1239                         ret = score->ifa->prefix_len;
1240                 score->matchlen = ret;
1241                 break;
1242         default:
1243                 ret = 0;
1244         }
1245
1246         if (ret)
1247                 __set_bit(i, score->scorebits);
1248         score->rule = i;
1249 out:
1250         return ret;
1251 }
1252
1253 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1254                        const struct in6_addr *daddr, unsigned int prefs,
1255                        struct in6_addr *saddr)
1256 {
1257         struct ipv6_saddr_score scores[2],
1258                                 *score = &scores[0], *hiscore = &scores[1];
1259         struct ipv6_saddr_dst dst;
1260         struct net_device *dev;
1261         int dst_type;
1262
1263         dst_type = __ipv6_addr_type(daddr);
1264         dst.addr = daddr;
1265         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1266         dst.scope = __ipv6_addr_src_scope(dst_type);
1267         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1268         dst.prefs = prefs;
1269
1270         hiscore->rule = -1;
1271         hiscore->ifa = NULL;
1272
1273         rcu_read_lock();
1274
1275         for_each_netdev_rcu(net, dev) {
1276                 struct inet6_dev *idev;
1277
1278                 /* Candidate Source Address (section 4)
1279                  *  - multicast and link-local destination address,
1280                  *    the set of candidate source address MUST only
1281                  *    include addresses assigned to interfaces
1282                  *    belonging to the same link as the outgoing
1283                  *    interface.
1284                  * (- For site-local destination addresses, the
1285                  *    set of candidate source addresses MUST only
1286                  *    include addresses assigned to interfaces
1287                  *    belonging to the same site as the outgoing
1288                  *    interface.)
1289                  */
1290                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1291                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1292                     dst.ifindex && dev->ifindex != dst.ifindex)
1293                         continue;
1294
1295                 idev = __in6_dev_get(dev);
1296                 if (!idev)
1297                         continue;
1298
1299                 read_lock_bh(&idev->lock);
1300                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1301                         int i;
1302
1303                         /*
1304                          * - Tentative Address (RFC2462 section 5.4)
1305                          *  - A tentative address is not considered
1306                          *    "assigned to an interface" in the traditional
1307                          *    sense, unless it is also flagged as optimistic.
1308                          * - Candidate Source Address (section 4)
1309                          *  - In any case, anycast addresses, multicast
1310                          *    addresses, and the unspecified address MUST
1311                          *    NOT be included in a candidate set.
1312                          */
1313                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1314                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1315                                 continue;
1316
1317                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1318
1319                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1320                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1321                                 LIMIT_NETDEBUG(KERN_DEBUG
1322                                                "ADDRCONF: unspecified / multicast address "
1323                                                "assigned as unicast address on %s",
1324                                                dev->name);
1325                                 continue;
1326                         }
1327
1328                         score->rule = -1;
1329                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1330
1331                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1332                                 int minihiscore, miniscore;
1333
1334                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1335                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1336
1337                                 if (minihiscore > miniscore) {
1338                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1339                                             score->scopedist > 0) {
1340                                                 /*
1341                                                  * special case:
1342                                                  * each remaining entry
1343                                                  * has too small (not enough)
1344                                                  * scope, because ifa entries
1345                                                  * are sorted by their scope
1346                                                  * values.
1347                                                  */
1348                                                 goto try_nextdev;
1349                                         }
1350                                         break;
1351                                 } else if (minihiscore < miniscore) {
1352                                         if (hiscore->ifa)
1353                                                 in6_ifa_put(hiscore->ifa);
1354
1355                                         in6_ifa_hold(score->ifa);
1356
1357                                         swap(hiscore, score);
1358
1359                                         /* restore our iterator */
1360                                         score->ifa = hiscore->ifa;
1361
1362                                         break;
1363                                 }
1364                         }
1365                 }
1366 try_nextdev:
1367                 read_unlock_bh(&idev->lock);
1368         }
1369         rcu_read_unlock();
1370
1371         if (!hiscore->ifa)
1372                 return -EADDRNOTAVAIL;
1373
1374         *saddr = hiscore->ifa->addr;
1375         in6_ifa_put(hiscore->ifa);
1376         return 0;
1377 }
1378 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1379
1380 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1381                     unsigned char banned_flags)
1382 {
1383         struct inet6_dev *idev;
1384         int err = -EADDRNOTAVAIL;
1385
1386         rcu_read_lock();
1387         idev = __in6_dev_get(dev);
1388         if (idev) {
1389                 struct inet6_ifaddr *ifp;
1390
1391                 read_lock_bh(&idev->lock);
1392                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1393                         if (ifp->scope == IFA_LINK &&
1394                             !(ifp->flags & banned_flags)) {
1395                                 *addr = ifp->addr;
1396                                 err = 0;
1397                                 break;
1398                         }
1399                 }
1400                 read_unlock_bh(&idev->lock);
1401         }
1402         rcu_read_unlock();
1403         return err;
1404 }
1405
1406 static int ipv6_count_addresses(struct inet6_dev *idev)
1407 {
1408         int cnt = 0;
1409         struct inet6_ifaddr *ifp;
1410
1411         read_lock_bh(&idev->lock);
1412         list_for_each_entry(ifp, &idev->addr_list, if_list)
1413                 cnt++;
1414         read_unlock_bh(&idev->lock);
1415         return cnt;
1416 }
1417
1418 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1419                   struct net_device *dev, int strict)
1420 {
1421         struct inet6_ifaddr *ifp;
1422         struct hlist_node *node;
1423         unsigned int hash = inet6_addr_hash(addr);
1424
1425         rcu_read_lock_bh();
1426         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1427                 if (!net_eq(dev_net(ifp->idev->dev), net))
1428                         continue;
1429                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1430                     !(ifp->flags&IFA_F_TENTATIVE) &&
1431                     (dev == NULL || ifp->idev->dev == dev ||
1432                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1433                         rcu_read_unlock_bh();
1434                         return 1;
1435                 }
1436         }
1437
1438         rcu_read_unlock_bh();
1439         return 0;
1440 }
1441 EXPORT_SYMBOL(ipv6_chk_addr);
1442
1443 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1444                                struct net_device *dev)
1445 {
1446         unsigned int hash = inet6_addr_hash(addr);
1447         struct inet6_ifaddr *ifp;
1448         struct hlist_node *node;
1449
1450         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1451                 if (!net_eq(dev_net(ifp->idev->dev), net))
1452                         continue;
1453                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1454                         if (dev == NULL || ifp->idev->dev == dev)
1455                                 return true;
1456                 }
1457         }
1458         return false;
1459 }
1460
1461 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1462 {
1463         struct inet6_dev *idev;
1464         struct inet6_ifaddr *ifa;
1465         int     onlink;
1466
1467         onlink = 0;
1468         rcu_read_lock();
1469         idev = __in6_dev_get(dev);
1470         if (idev) {
1471                 read_lock_bh(&idev->lock);
1472                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1473                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1474                                                    ifa->prefix_len);
1475                         if (onlink)
1476                                 break;
1477                 }
1478                 read_unlock_bh(&idev->lock);
1479         }
1480         rcu_read_unlock();
1481         return onlink;
1482 }
1483 EXPORT_SYMBOL(ipv6_chk_prefix);
1484
1485 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1486                                      struct net_device *dev, int strict)
1487 {
1488         struct inet6_ifaddr *ifp, *result = NULL;
1489         unsigned int hash = inet6_addr_hash(addr);
1490         struct hlist_node *node;
1491
1492         rcu_read_lock_bh();
1493         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1494                 if (!net_eq(dev_net(ifp->idev->dev), net))
1495                         continue;
1496                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1497                         if (dev == NULL || ifp->idev->dev == dev ||
1498                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1499                                 result = ifp;
1500                                 in6_ifa_hold(ifp);
1501                                 break;
1502                         }
1503                 }
1504         }
1505         rcu_read_unlock_bh();
1506
1507         return result;
1508 }
1509
1510 /* Gets referenced address, destroys ifaddr */
1511
1512 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1513 {
1514         if (ifp->flags&IFA_F_PERMANENT) {
1515                 spin_lock_bh(&ifp->lock);
1516                 addrconf_del_timer(ifp);
1517                 ifp->flags |= IFA_F_TENTATIVE;
1518                 if (dad_failed)
1519                         ifp->flags |= IFA_F_DADFAILED;
1520                 spin_unlock_bh(&ifp->lock);
1521                 if (dad_failed)
1522                         ipv6_ifa_notify(0, ifp);
1523                 in6_ifa_put(ifp);
1524 #ifdef CONFIG_IPV6_PRIVACY
1525         } else if (ifp->flags&IFA_F_TEMPORARY) {
1526                 struct inet6_ifaddr *ifpub;
1527                 spin_lock_bh(&ifp->lock);
1528                 ifpub = ifp->ifpub;
1529                 if (ifpub) {
1530                         in6_ifa_hold(ifpub);
1531                         spin_unlock_bh(&ifp->lock);
1532                         ipv6_create_tempaddr(ifpub, ifp);
1533                         in6_ifa_put(ifpub);
1534                 } else {
1535                         spin_unlock_bh(&ifp->lock);
1536                 }
1537                 ipv6_del_addr(ifp);
1538 #endif
1539         } else
1540                 ipv6_del_addr(ifp);
1541 }
1542
1543 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1544 {
1545         int err = -ENOENT;
1546
1547         spin_lock(&ifp->state_lock);
1548         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1549                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1550                 err = 0;
1551         }
1552         spin_unlock(&ifp->state_lock);
1553
1554         return err;
1555 }
1556
1557 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1558 {
1559         struct inet6_dev *idev = ifp->idev;
1560
1561         if (addrconf_dad_end(ifp)) {
1562                 in6_ifa_put(ifp);
1563                 return;
1564         }
1565
1566         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1567                              ifp->idev->dev->name, &ifp->addr);
1568
1569         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1570                 struct in6_addr addr;
1571
1572                 addr.s6_addr32[0] = htonl(0xfe800000);
1573                 addr.s6_addr32[1] = 0;
1574
1575                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1576                     ipv6_addr_equal(&ifp->addr, &addr)) {
1577                         /* DAD failed for link-local based on MAC address */
1578                         idev->cnf.disable_ipv6 = 1;
1579
1580                         pr_info("%s: IPv6 being disabled!\n",
1581                                 ifp->idev->dev->name);
1582                 }
1583         }
1584
1585         addrconf_dad_stop(ifp, 1);
1586 }
1587
1588 /* Join to solicited addr multicast group. */
1589
1590 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1591 {
1592         struct in6_addr maddr;
1593
1594         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1595                 return;
1596
1597         addrconf_addr_solict_mult(addr, &maddr);
1598         ipv6_dev_mc_inc(dev, &maddr);
1599 }
1600
1601 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1602 {
1603         struct in6_addr maddr;
1604
1605         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1606                 return;
1607
1608         addrconf_addr_solict_mult(addr, &maddr);
1609         __ipv6_dev_mc_dec(idev, &maddr);
1610 }
1611
1612 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1613 {
1614         struct in6_addr addr;
1615         if (ifp->prefix_len == 127) /* RFC 6164 */
1616                 return;
1617         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1618         if (ipv6_addr_any(&addr))
1619                 return;
1620         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1621 }
1622
1623 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1624 {
1625         struct in6_addr addr;
1626         if (ifp->prefix_len == 127) /* RFC 6164 */
1627                 return;
1628         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1629         if (ipv6_addr_any(&addr))
1630                 return;
1631         __ipv6_dev_ac_dec(ifp->idev, &addr);
1632 }
1633
1634 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1635 {
1636         if (dev->addr_len != ETH_ALEN)
1637                 return -1;
1638         memcpy(eui, dev->dev_addr, 3);
1639         memcpy(eui + 5, dev->dev_addr + 3, 3);
1640
1641         /*
1642          * The zSeries OSA network cards can be shared among various
1643          * OS instances, but the OSA cards have only one MAC address.
1644          * This leads to duplicate address conflicts in conjunction
1645          * with IPv6 if more than one instance uses the same card.
1646          *
1647          * The driver for these cards can deliver a unique 16-bit
1648          * identifier for each instance sharing the same card.  It is
1649          * placed instead of 0xFFFE in the interface identifier.  The
1650          * "u" bit of the interface identifier is not inverted in this
1651          * case.  Hence the resulting interface identifier has local
1652          * scope according to RFC2373.
1653          */
1654         if (dev->dev_id) {
1655                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1656                 eui[4] = dev->dev_id & 0xFF;
1657         } else {
1658                 eui[3] = 0xFF;
1659                 eui[4] = 0xFE;
1660                 eui[0] ^= 2;
1661         }
1662         return 0;
1663 }
1664
1665 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1666 {
1667         if (dev->addr_len != IEEE802154_ADDR_LEN)
1668                 return -1;
1669         memcpy(eui, dev->dev_addr, 8);
1670         eui[0] ^= 2;
1671         return 0;
1672 }
1673
1674 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1675 {
1676         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1677         if (dev->addr_len != ARCNET_ALEN)
1678                 return -1;
1679         memset(eui, 0, 7);
1680         eui[7] = *(u8 *)dev->dev_addr;
1681         return 0;
1682 }
1683
1684 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1685 {
1686         if (dev->addr_len != INFINIBAND_ALEN)
1687                 return -1;
1688         memcpy(eui, dev->dev_addr + 12, 8);
1689         eui[0] |= 2;
1690         return 0;
1691 }
1692
1693 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1694 {
1695         if (addr == 0)
1696                 return -1;
1697         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1698                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1699                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1700                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1701                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1702                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1703         eui[1] = 0;
1704         eui[2] = 0x5E;
1705         eui[3] = 0xFE;
1706         memcpy(eui + 4, &addr, 4);
1707         return 0;
1708 }
1709
1710 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1711 {
1712         if (dev->priv_flags & IFF_ISATAP)
1713                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1714         return -1;
1715 }
1716
1717 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1718 {
1719         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1720 }
1721
1722 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1723 {
1724         switch (dev->type) {
1725         case ARPHRD_ETHER:
1726         case ARPHRD_FDDI:
1727                 return addrconf_ifid_eui48(eui, dev);
1728         case ARPHRD_ARCNET:
1729                 return addrconf_ifid_arcnet(eui, dev);
1730         case ARPHRD_INFINIBAND:
1731                 return addrconf_ifid_infiniband(eui, dev);
1732         case ARPHRD_SIT:
1733                 return addrconf_ifid_sit(eui, dev);
1734         case ARPHRD_IPGRE:
1735                 return addrconf_ifid_gre(eui, dev);
1736         case ARPHRD_IEEE802154:
1737                 return addrconf_ifid_eui64(eui, dev);
1738         }
1739         return -1;
1740 }
1741
1742 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1743 {
1744         int err = -1;
1745         struct inet6_ifaddr *ifp;
1746
1747         read_lock_bh(&idev->lock);
1748         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1749                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1750                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1751                         err = 0;
1752                         break;
1753                 }
1754         }
1755         read_unlock_bh(&idev->lock);
1756         return err;
1757 }
1758
1759 #ifdef CONFIG_IPV6_PRIVACY
1760 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1761 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1762 {
1763 regen:
1764         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1765         idev->rndid[0] &= ~0x02;
1766
1767         /*
1768          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1769          * check if generated address is not inappropriate
1770          *
1771          *  - Reserved subnet anycast (RFC 2526)
1772          *      11111101 11....11 1xxxxxxx
1773          *  - ISATAP (RFC4214) 6.1
1774          *      00-00-5E-FE-xx-xx-xx-xx
1775          *  - value 0
1776          *  - XXX: already assigned to an address on the device
1777          */
1778         if (idev->rndid[0] == 0xfd &&
1779             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1780             (idev->rndid[7]&0x80))
1781                 goto regen;
1782         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1783                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1784                         goto regen;
1785                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1786                         goto regen;
1787         }
1788 }
1789
1790 static void ipv6_regen_rndid(unsigned long data)
1791 {
1792         struct inet6_dev *idev = (struct inet6_dev *) data;
1793         unsigned long expires;
1794
1795         rcu_read_lock_bh();
1796         write_lock_bh(&idev->lock);
1797
1798         if (idev->dead)
1799                 goto out;
1800
1801         __ipv6_regen_rndid(idev);
1802
1803         expires = jiffies +
1804                 idev->cnf.temp_prefered_lft * HZ -
1805                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1806                 idev->cnf.max_desync_factor * HZ;
1807         if (time_before(expires, jiffies)) {
1808                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1809                         __func__, idev->dev->name);
1810                 goto out;
1811         }
1812
1813         if (!mod_timer(&idev->regen_timer, expires))
1814                 in6_dev_hold(idev);
1815
1816 out:
1817         write_unlock_bh(&idev->lock);
1818         rcu_read_unlock_bh();
1819         in6_dev_put(idev);
1820 }
1821
1822 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1823 {
1824         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1825                 __ipv6_regen_rndid(idev);
1826 }
1827 #endif
1828
1829 /*
1830  *      Add prefix route.
1831  */
1832
1833 static void
1834 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1835                       unsigned long expires, u32 flags)
1836 {
1837         struct fib6_config cfg = {
1838                 .fc_table = RT6_TABLE_PREFIX,
1839                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1840                 .fc_ifindex = dev->ifindex,
1841                 .fc_expires = expires,
1842                 .fc_dst_len = plen,
1843                 .fc_flags = RTF_UP | flags,
1844                 .fc_nlinfo.nl_net = dev_net(dev),
1845                 .fc_protocol = RTPROT_KERNEL,
1846         };
1847
1848         cfg.fc_dst = *pfx;
1849
1850         /* Prevent useless cloning on PtP SIT.
1851            This thing is done here expecting that the whole
1852            class of non-broadcast devices need not cloning.
1853          */
1854 #if IS_ENABLED(CONFIG_IPV6_SIT)
1855         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1856                 cfg.fc_flags |= RTF_NONEXTHOP;
1857 #endif
1858
1859         ip6_route_add(&cfg);
1860 }
1861
1862
1863 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1864                                                   int plen,
1865                                                   const struct net_device *dev,
1866                                                   u32 flags, u32 noflags)
1867 {
1868         struct fib6_node *fn;
1869         struct rt6_info *rt = NULL;
1870         struct fib6_table *table;
1871
1872         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1873         if (table == NULL)
1874                 return NULL;
1875
1876         read_lock_bh(&table->tb6_lock);
1877         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1878         if (!fn)
1879                 goto out;
1880         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1881                 if (rt->dst.dev->ifindex != dev->ifindex)
1882                         continue;
1883                 if ((rt->rt6i_flags & flags) != flags)
1884                         continue;
1885                 if ((rt->rt6i_flags & noflags) != 0)
1886                         continue;
1887                 dst_hold(&rt->dst);
1888                 break;
1889         }
1890 out:
1891         read_unlock_bh(&table->tb6_lock);
1892         return rt;
1893 }
1894
1895
1896 /* Create "default" multicast route to the interface */
1897
1898 static void addrconf_add_mroute(struct net_device *dev)
1899 {
1900         struct fib6_config cfg = {
1901                 .fc_table = RT6_TABLE_LOCAL,
1902                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1903                 .fc_ifindex = dev->ifindex,
1904                 .fc_dst_len = 8,
1905                 .fc_flags = RTF_UP,
1906                 .fc_nlinfo.nl_net = dev_net(dev),
1907         };
1908
1909         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1910
1911         ip6_route_add(&cfg);
1912 }
1913
1914 #if IS_ENABLED(CONFIG_IPV6_SIT)
1915 static void sit_route_add(struct net_device *dev)
1916 {
1917         struct fib6_config cfg = {
1918                 .fc_table = RT6_TABLE_MAIN,
1919                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1920                 .fc_ifindex = dev->ifindex,
1921                 .fc_dst_len = 96,
1922                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1923                 .fc_nlinfo.nl_net = dev_net(dev),
1924         };
1925
1926         /* prefix length - 96 bits "::d.d.d.d" */
1927         ip6_route_add(&cfg);
1928 }
1929 #endif
1930
1931 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1932 {
1933         struct inet6_dev *idev;
1934
1935         ASSERT_RTNL();
1936
1937         idev = ipv6_find_idev(dev);
1938         if (!idev)
1939                 return ERR_PTR(-ENOBUFS);
1940
1941         if (idev->cnf.disable_ipv6)
1942                 return ERR_PTR(-EACCES);
1943
1944         /* Add default multicast route */
1945         if (!(dev->flags & IFF_LOOPBACK))
1946                 addrconf_add_mroute(dev);
1947
1948         return idev;
1949 }
1950
1951 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1952 {
1953         struct prefix_info *pinfo;
1954         __u32 valid_lft;
1955         __u32 prefered_lft;
1956         int addr_type;
1957         struct inet6_dev *in6_dev;
1958         struct net *net = dev_net(dev);
1959
1960         pinfo = (struct prefix_info *) opt;
1961
1962         if (len < sizeof(struct prefix_info)) {
1963                 ADBG(("addrconf: prefix option too short\n"));
1964                 return;
1965         }
1966
1967         /*
1968          *      Validation checks ([ADDRCONF], page 19)
1969          */
1970
1971         addr_type = ipv6_addr_type(&pinfo->prefix);
1972
1973         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1974                 return;
1975
1976         valid_lft = ntohl(pinfo->valid);
1977         prefered_lft = ntohl(pinfo->prefered);
1978
1979         if (prefered_lft > valid_lft) {
1980                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1981                 return;
1982         }
1983
1984         in6_dev = in6_dev_get(dev);
1985
1986         if (in6_dev == NULL) {
1987                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1988                                     dev->name);
1989                 return;
1990         }
1991
1992         /*
1993          *      Two things going on here:
1994          *      1) Add routes for on-link prefixes
1995          *      2) Configure prefixes with the auto flag set
1996          */
1997
1998         if (pinfo->onlink) {
1999                 struct rt6_info *rt;
2000                 unsigned long rt_expires;
2001
2002                 /* Avoid arithmetic overflow. Really, we could
2003                  * save rt_expires in seconds, likely valid_lft,
2004                  * but it would require division in fib gc, that it
2005                  * not good.
2006                  */
2007                 if (HZ > USER_HZ)
2008                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2009                 else
2010                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2011
2012                 if (addrconf_finite_timeout(rt_expires))
2013                         rt_expires *= HZ;
2014
2015                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2016                                                pinfo->prefix_len,
2017                                                dev,
2018                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2019                                                RTF_GATEWAY | RTF_DEFAULT);
2020
2021                 if (rt) {
2022                         /* Autoconf prefix route */
2023                         if (valid_lft == 0) {
2024                                 ip6_del_rt(rt);
2025                                 rt = NULL;
2026                         } else if (addrconf_finite_timeout(rt_expires)) {
2027                                 /* not infinity */
2028                                 rt6_set_expires(rt, jiffies + rt_expires);
2029                         } else {
2030                                 rt6_clean_expires(rt);
2031                         }
2032                 } else if (valid_lft) {
2033                         clock_t expires = 0;
2034                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2035                         if (addrconf_finite_timeout(rt_expires)) {
2036                                 /* not infinity */
2037                                 flags |= RTF_EXPIRES;
2038                                 expires = jiffies_to_clock_t(rt_expires);
2039                         }
2040                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2041                                               dev, expires, flags);
2042                 }
2043                 ip6_rt_put(rt);
2044         }
2045
2046         /* Try to figure out our local address for this prefix */
2047
2048         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2049                 struct inet6_ifaddr *ifp;
2050                 struct in6_addr addr;
2051                 int create = 0, update_lft = 0;
2052
2053                 if (pinfo->prefix_len == 64) {
2054                         memcpy(&addr, &pinfo->prefix, 8);
2055                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2056                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2057                                 in6_dev_put(in6_dev);
2058                                 return;
2059                         }
2060                         goto ok;
2061                 }
2062                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2063                                     pinfo->prefix_len);
2064                 in6_dev_put(in6_dev);
2065                 return;
2066
2067 ok:
2068
2069                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2070
2071                 if (ifp == NULL && valid_lft) {
2072                         int max_addresses = in6_dev->cnf.max_addresses;
2073                         u32 addr_flags = 0;
2074
2075 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2076                         if (in6_dev->cnf.optimistic_dad &&
2077                             !net->ipv6.devconf_all->forwarding && sllao)
2078                                 addr_flags = IFA_F_OPTIMISTIC;
2079 #endif
2080
2081                         /* Do not allow to create too much of autoconfigured
2082                          * addresses; this would be too easy way to crash kernel.
2083                          */
2084                         if (!max_addresses ||
2085                             ipv6_count_addresses(in6_dev) < max_addresses)
2086                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2087                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2088                                                     addr_flags);
2089
2090                         if (IS_ERR_OR_NULL(ifp)) {
2091                                 in6_dev_put(in6_dev);
2092                                 return;
2093                         }
2094
2095                         update_lft = create = 1;
2096                         ifp->cstamp = jiffies;
2097                         addrconf_dad_start(ifp);
2098                 }
2099
2100                 if (ifp) {
2101                         int flags;
2102                         unsigned long now;
2103 #ifdef CONFIG_IPV6_PRIVACY
2104                         struct inet6_ifaddr *ift;
2105 #endif
2106                         u32 stored_lft;
2107
2108                         /* update lifetime (RFC2462 5.5.3 e) */
2109                         spin_lock(&ifp->lock);
2110                         now = jiffies;
2111                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2112                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2113                         else
2114                                 stored_lft = 0;
2115                         if (!update_lft && stored_lft) {
2116                                 if (valid_lft > MIN_VALID_LIFETIME ||
2117                                     valid_lft > stored_lft)
2118                                         update_lft = 1;
2119                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2120                                         /* valid_lft <= stored_lft is always true */
2121                                         /*
2122                                          * RFC 4862 Section 5.5.3e:
2123                                          * "Note that the preferred lifetime of
2124                                          *  the corresponding address is always
2125                                          *  reset to the Preferred Lifetime in
2126                                          *  the received Prefix Information
2127                                          *  option, regardless of whether the
2128                                          *  valid lifetime is also reset or
2129                                          *  ignored."
2130                                          *
2131                                          *  So if the preferred lifetime in
2132                                          *  this advertisement is different
2133                                          *  than what we have stored, but the
2134                                          *  valid lifetime is invalid, just
2135                                          *  reset prefered_lft.
2136                                          *
2137                                          *  We must set the valid lifetime
2138                                          *  to the stored lifetime since we'll
2139                                          *  be updating the timestamp below,
2140                                          *  else we'll set it back to the
2141                                          *  minimum.
2142                                          */
2143                                         if (prefered_lft != ifp->prefered_lft) {
2144                                                 valid_lft = stored_lft;
2145                                                 update_lft = 1;
2146                                         }
2147                                 } else {
2148                                         valid_lft = MIN_VALID_LIFETIME;
2149                                         if (valid_lft < prefered_lft)
2150                                                 prefered_lft = valid_lft;
2151                                         update_lft = 1;
2152                                 }
2153                         }
2154
2155                         if (update_lft) {
2156                                 ifp->valid_lft = valid_lft;
2157                                 ifp->prefered_lft = prefered_lft;
2158                                 ifp->tstamp = now;
2159                                 flags = ifp->flags;
2160                                 ifp->flags &= ~IFA_F_DEPRECATED;
2161                                 spin_unlock(&ifp->lock);
2162
2163                                 if (!(flags&IFA_F_TENTATIVE))
2164                                         ipv6_ifa_notify(0, ifp);
2165                         } else
2166                                 spin_unlock(&ifp->lock);
2167
2168 #ifdef CONFIG_IPV6_PRIVACY
2169                         read_lock_bh(&in6_dev->lock);
2170                         /* update all temporary addresses in the list */
2171                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2172                                             tmp_list) {
2173                                 int age, max_valid, max_prefered;
2174
2175                                 if (ifp != ift->ifpub)
2176                                         continue;
2177
2178                                 /*
2179                                  * RFC 4941 section 3.3:
2180                                  * If a received option will extend the lifetime
2181                                  * of a public address, the lifetimes of
2182                                  * temporary addresses should be extended,
2183                                  * subject to the overall constraint that no
2184                                  * temporary addresses should ever remain
2185                                  * "valid" or "preferred" for a time longer than
2186                                  * (TEMP_VALID_LIFETIME) or
2187                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2188                                  * respectively.
2189                                  */
2190                                 age = (now - ift->cstamp) / HZ;
2191                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2192                                 if (max_valid < 0)
2193                                         max_valid = 0;
2194
2195                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2196                                                in6_dev->cnf.max_desync_factor -
2197                                                age;
2198                                 if (max_prefered < 0)
2199                                         max_prefered = 0;
2200
2201                                 if (valid_lft > max_valid)
2202                                         valid_lft = max_valid;
2203
2204                                 if (prefered_lft > max_prefered)
2205                                         prefered_lft = max_prefered;
2206
2207                                 spin_lock(&ift->lock);
2208                                 flags = ift->flags;
2209                                 ift->valid_lft = valid_lft;
2210                                 ift->prefered_lft = prefered_lft;
2211                                 ift->tstamp = now;
2212                                 if (prefered_lft > 0)
2213                                         ift->flags &= ~IFA_F_DEPRECATED;
2214
2215                                 spin_unlock(&ift->lock);
2216                                 if (!(flags&IFA_F_TENTATIVE))
2217                                         ipv6_ifa_notify(0, ift);
2218                         }
2219
2220                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2221                                 /*
2222                                  * When a new public address is created as
2223                                  * described in [ADDRCONF], also create a new
2224                                  * temporary address. Also create a temporary
2225                                  * address if it's enabled but no temporary
2226                                  * address currently exists.
2227                                  */
2228                                 read_unlock_bh(&in6_dev->lock);
2229                                 ipv6_create_tempaddr(ifp, NULL);
2230                         } else {
2231                                 read_unlock_bh(&in6_dev->lock);
2232                         }
2233 #endif
2234                         in6_ifa_put(ifp);
2235                         addrconf_verify(0);
2236                 }
2237         }
2238         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2239         in6_dev_put(in6_dev);
2240 }
2241
2242 /*
2243  *      Set destination address.
2244  *      Special case for SIT interfaces where we create a new "virtual"
2245  *      device.
2246  */
2247 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2248 {
2249         struct in6_ifreq ireq;
2250         struct net_device *dev;
2251         int err = -EINVAL;
2252
2253         rtnl_lock();
2254
2255         err = -EFAULT;
2256         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2257                 goto err_exit;
2258
2259         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2260
2261         err = -ENODEV;
2262         if (dev == NULL)
2263                 goto err_exit;
2264
2265 #if IS_ENABLED(CONFIG_IPV6_SIT)
2266         if (dev->type == ARPHRD_SIT) {
2267                 const struct net_device_ops *ops = dev->netdev_ops;
2268                 struct ifreq ifr;
2269                 struct ip_tunnel_parm p;
2270
2271                 err = -EADDRNOTAVAIL;
2272                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2273                         goto err_exit;
2274
2275                 memset(&p, 0, sizeof(p));
2276                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2277                 p.iph.saddr = 0;
2278                 p.iph.version = 4;
2279                 p.iph.ihl = 5;
2280                 p.iph.protocol = IPPROTO_IPV6;
2281                 p.iph.ttl = 64;
2282                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2283
2284                 if (ops->ndo_do_ioctl) {
2285                         mm_segment_t oldfs = get_fs();
2286
2287                         set_fs(KERNEL_DS);
2288                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2289                         set_fs(oldfs);
2290                 } else
2291                         err = -EOPNOTSUPP;
2292
2293                 if (err == 0) {
2294                         err = -ENOBUFS;
2295                         dev = __dev_get_by_name(net, p.name);
2296                         if (!dev)
2297                                 goto err_exit;
2298                         err = dev_open(dev);
2299                 }
2300         }
2301 #endif
2302
2303 err_exit:
2304         rtnl_unlock();
2305         return err;
2306 }
2307
2308 /*
2309  *      Manual configuration of address on an interface
2310  */
2311 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2312                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2313                           __u32 valid_lft)
2314 {
2315         struct inet6_ifaddr *ifp;
2316         struct inet6_dev *idev;
2317         struct net_device *dev;
2318         int scope;
2319         u32 flags;
2320         clock_t expires;
2321         unsigned long timeout;
2322
2323         ASSERT_RTNL();
2324
2325         if (plen > 128)
2326                 return -EINVAL;
2327
2328         /* check the lifetime */
2329         if (!valid_lft || prefered_lft > valid_lft)
2330                 return -EINVAL;
2331
2332         dev = __dev_get_by_index(net, ifindex);
2333         if (!dev)
2334                 return -ENODEV;
2335
2336         idev = addrconf_add_dev(dev);
2337         if (IS_ERR(idev))
2338                 return PTR_ERR(idev);
2339
2340         scope = ipv6_addr_scope(pfx);
2341
2342         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2343         if (addrconf_finite_timeout(timeout)) {
2344                 expires = jiffies_to_clock_t(timeout * HZ);
2345                 valid_lft = timeout;
2346                 flags = RTF_EXPIRES;
2347         } else {
2348                 expires = 0;
2349                 flags = 0;
2350                 ifa_flags |= IFA_F_PERMANENT;
2351         }
2352
2353         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2354         if (addrconf_finite_timeout(timeout)) {
2355                 if (timeout == 0)
2356                         ifa_flags |= IFA_F_DEPRECATED;
2357                 prefered_lft = timeout;
2358         }
2359
2360         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2361
2362         if (!IS_ERR(ifp)) {
2363                 spin_lock_bh(&ifp->lock);
2364                 ifp->valid_lft = valid_lft;
2365                 ifp->prefered_lft = prefered_lft;
2366                 ifp->tstamp = jiffies;
2367                 spin_unlock_bh(&ifp->lock);
2368
2369                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2370                                       expires, flags);
2371                 /*
2372                  * Note that section 3.1 of RFC 4429 indicates
2373                  * that the Optimistic flag should not be set for
2374                  * manually configured addresses
2375                  */
2376                 addrconf_dad_start(ifp);
2377                 in6_ifa_put(ifp);
2378                 addrconf_verify(0);
2379                 return 0;
2380         }
2381
2382         return PTR_ERR(ifp);
2383 }
2384
2385 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2386                           unsigned int plen)
2387 {
2388         struct inet6_ifaddr *ifp;
2389         struct inet6_dev *idev;
2390         struct net_device *dev;
2391
2392         if (plen > 128)
2393                 return -EINVAL;
2394
2395         dev = __dev_get_by_index(net, ifindex);
2396         if (!dev)
2397                 return -ENODEV;
2398
2399         if ((idev = __in6_dev_get(dev)) == NULL)
2400                 return -ENXIO;
2401
2402         read_lock_bh(&idev->lock);
2403         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2404                 if (ifp->prefix_len == plen &&
2405                     ipv6_addr_equal(pfx, &ifp->addr)) {
2406                         in6_ifa_hold(ifp);
2407                         read_unlock_bh(&idev->lock);
2408
2409                         ipv6_del_addr(ifp);
2410
2411                         /* If the last address is deleted administratively,
2412                            disable IPv6 on this interface.
2413                          */
2414                         if (list_empty(&idev->addr_list))
2415                                 addrconf_ifdown(idev->dev, 1);
2416                         return 0;
2417                 }
2418         }
2419         read_unlock_bh(&idev->lock);
2420         return -EADDRNOTAVAIL;
2421 }
2422
2423
2424 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2425 {
2426         struct in6_ifreq ireq;
2427         int err;
2428
2429         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2430                 return -EPERM;
2431
2432         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2433                 return -EFAULT;
2434
2435         rtnl_lock();
2436         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2437                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2438                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2439         rtnl_unlock();
2440         return err;
2441 }
2442
2443 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2444 {
2445         struct in6_ifreq ireq;
2446         int err;
2447
2448         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2449                 return -EPERM;
2450
2451         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2452                 return -EFAULT;
2453
2454         rtnl_lock();
2455         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2456                              ireq.ifr6_prefixlen);
2457         rtnl_unlock();
2458         return err;
2459 }
2460
2461 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2462                      int plen, int scope)
2463 {
2464         struct inet6_ifaddr *ifp;
2465
2466         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2467         if (!IS_ERR(ifp)) {
2468                 spin_lock_bh(&ifp->lock);
2469                 ifp->flags &= ~IFA_F_TENTATIVE;
2470                 spin_unlock_bh(&ifp->lock);
2471                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2472                 in6_ifa_put(ifp);
2473         }
2474 }
2475
2476 #if IS_ENABLED(CONFIG_IPV6_SIT)
2477 static void sit_add_v4_addrs(struct inet6_dev *idev)
2478 {
2479         struct in6_addr addr;
2480         struct net_device *dev;
2481         struct net *net = dev_net(idev->dev);
2482         int scope;
2483
2484         ASSERT_RTNL();
2485
2486         memset(&addr, 0, sizeof(struct in6_addr));
2487         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2488
2489         if (idev->dev->flags&IFF_POINTOPOINT) {
2490                 addr.s6_addr32[0] = htonl(0xfe800000);
2491                 scope = IFA_LINK;
2492         } else {
2493                 scope = IPV6_ADDR_COMPATv4;
2494         }
2495
2496         if (addr.s6_addr32[3]) {
2497                 add_addr(idev, &addr, 128, scope);
2498                 return;
2499         }
2500
2501         for_each_netdev(net, dev) {
2502                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2503                 if (in_dev && (dev->flags & IFF_UP)) {
2504                         struct in_ifaddr *ifa;
2505
2506                         int flag = scope;
2507
2508                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2509                                 int plen;
2510
2511                                 addr.s6_addr32[3] = ifa->ifa_local;
2512
2513                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2514                                         continue;
2515                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2516                                         if (idev->dev->flags&IFF_POINTOPOINT)
2517                                                 continue;
2518                                         flag |= IFA_HOST;
2519                                 }
2520                                 if (idev->dev->flags&IFF_POINTOPOINT)
2521                                         plen = 64;
2522                                 else
2523                                         plen = 96;
2524
2525                                 add_addr(idev, &addr, plen, flag);
2526                         }
2527                 }
2528         }
2529 }
2530 #endif
2531
2532 static void init_loopback(struct net_device *dev)
2533 {
2534         struct inet6_dev  *idev;
2535
2536         /* ::1 */
2537
2538         ASSERT_RTNL();
2539
2540         if ((idev = ipv6_find_idev(dev)) == NULL) {
2541                 pr_debug("%s: add_dev failed\n", __func__);
2542                 return;
2543         }
2544
2545         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2546 }
2547
2548 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2549 {
2550         struct inet6_ifaddr *ifp;
2551         u32 addr_flags = IFA_F_PERMANENT;
2552
2553 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2554         if (idev->cnf.optimistic_dad &&
2555             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2556                 addr_flags |= IFA_F_OPTIMISTIC;
2557 #endif
2558
2559
2560         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2561         if (!IS_ERR(ifp)) {
2562                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2563                 addrconf_dad_start(ifp);
2564                 in6_ifa_put(ifp);
2565         }
2566 }
2567
2568 static void addrconf_dev_config(struct net_device *dev)
2569 {
2570         struct in6_addr addr;
2571         struct inet6_dev *idev;
2572
2573         ASSERT_RTNL();
2574
2575         if ((dev->type != ARPHRD_ETHER) &&
2576             (dev->type != ARPHRD_FDDI) &&
2577             (dev->type != ARPHRD_ARCNET) &&
2578             (dev->type != ARPHRD_INFINIBAND) &&
2579             (dev->type != ARPHRD_IEEE802154)) {
2580                 /* Alas, we support only Ethernet autoconfiguration. */
2581                 return;
2582         }
2583
2584         idev = addrconf_add_dev(dev);
2585         if (IS_ERR(idev))
2586                 return;
2587
2588         memset(&addr, 0, sizeof(struct in6_addr));
2589         addr.s6_addr32[0] = htonl(0xFE800000);
2590
2591         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2592                 addrconf_add_linklocal(idev, &addr);
2593 }
2594
2595 #if IS_ENABLED(CONFIG_IPV6_SIT)
2596 static void addrconf_sit_config(struct net_device *dev)
2597 {
2598         struct inet6_dev *idev;
2599
2600         ASSERT_RTNL();
2601
2602         /*
2603          * Configure the tunnel with one of our IPv4
2604          * addresses... we should configure all of
2605          * our v4 addrs in the tunnel
2606          */
2607
2608         if ((idev = ipv6_find_idev(dev)) == NULL) {
2609                 pr_debug("%s: add_dev failed\n", __func__);
2610                 return;
2611         }
2612
2613         if (dev->priv_flags & IFF_ISATAP) {
2614                 struct in6_addr addr;
2615
2616                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2617                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2618                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2619                         addrconf_add_linklocal(idev, &addr);
2620                 return;
2621         }
2622
2623         sit_add_v4_addrs(idev);
2624
2625         if (dev->flags&IFF_POINTOPOINT)
2626                 addrconf_add_mroute(dev);
2627         else
2628                 sit_route_add(dev);
2629 }
2630 #endif
2631
2632 #if IS_ENABLED(CONFIG_NET_IPGRE)
2633 static void addrconf_gre_config(struct net_device *dev)
2634 {
2635         struct inet6_dev *idev;
2636         struct in6_addr addr;
2637
2638         pr_info("%s(%s)\n", __func__, dev->name);
2639
2640         ASSERT_RTNL();
2641
2642         if ((idev = ipv6_find_idev(dev)) == NULL) {
2643                 pr_debug("%s: add_dev failed\n", __func__);
2644                 return;
2645         }
2646
2647         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2648         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2649
2650         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2651                 addrconf_add_linklocal(idev, &addr);
2652 }
2653 #endif
2654
2655 static inline int
2656 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2657 {
2658         struct in6_addr lladdr;
2659
2660         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2661                 addrconf_add_linklocal(idev, &lladdr);
2662                 return 0;
2663         }
2664         return -1;
2665 }
2666
2667 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2668 {
2669         struct net_device *link_dev;
2670         struct net *net = dev_net(idev->dev);
2671
2672         /* first try to inherit the link-local address from the link device */
2673         if (idev->dev->iflink &&
2674             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2675                 if (!ipv6_inherit_linklocal(idev, link_dev))
2676                         return;
2677         }
2678         /* then try to inherit it from any device */
2679         for_each_netdev(net, link_dev) {
2680                 if (!ipv6_inherit_linklocal(idev, link_dev))
2681                         return;
2682         }
2683         pr_debug("init ip6-ip6: add_linklocal failed\n");
2684 }
2685
2686 /*
2687  * Autoconfigure tunnel with a link-local address so routing protocols,
2688  * DHCPv6, MLD etc. can be run over the virtual link
2689  */
2690
2691 static void addrconf_ip6_tnl_config(struct net_device *dev)
2692 {
2693         struct inet6_dev *idev;
2694
2695         ASSERT_RTNL();
2696
2697         idev = addrconf_add_dev(dev);
2698         if (IS_ERR(idev)) {
2699                 pr_debug("init ip6-ip6: add_dev failed\n");
2700                 return;
2701         }
2702         ip6_tnl_add_linklocal(idev);
2703 }
2704
2705 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2706                            void *data)
2707 {
2708         struct net_device *dev = (struct net_device *) data;
2709         struct inet6_dev *idev = __in6_dev_get(dev);
2710         int run_pending = 0;
2711         int err;
2712
2713         switch (event) {
2714         case NETDEV_REGISTER:
2715                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2716                         idev = ipv6_add_dev(dev);
2717                         if (!idev)
2718                                 return notifier_from_errno(-ENOMEM);
2719                 }
2720                 break;
2721
2722         case NETDEV_UP:
2723         case NETDEV_CHANGE:
2724                 if (dev->flags & IFF_SLAVE)
2725                         break;
2726
2727                 if (event == NETDEV_UP) {
2728                         if (!addrconf_qdisc_ok(dev)) {
2729                                 /* device is not ready yet. */
2730                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2731                                         dev->name);
2732                                 break;
2733                         }
2734
2735                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2736                                 idev = ipv6_add_dev(dev);
2737
2738                         if (idev) {
2739                                 idev->if_flags |= IF_READY;
2740                                 run_pending = 1;
2741                         }
2742                 } else {
2743                         if (!addrconf_qdisc_ok(dev)) {
2744                                 /* device is still not ready. */
2745                                 break;
2746                         }
2747
2748                         if (idev) {
2749                                 if (idev->if_flags & IF_READY)
2750                                         /* device is already configured. */
2751                                         break;
2752                                 idev->if_flags |= IF_READY;
2753                         }
2754
2755                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2756                                 dev->name);
2757
2758                         run_pending = 1;
2759                 }
2760
2761                 switch (dev->type) {
2762 #if IS_ENABLED(CONFIG_IPV6_SIT)
2763                 case ARPHRD_SIT:
2764                         addrconf_sit_config(dev);
2765                         break;
2766 #endif
2767 #if IS_ENABLED(CONFIG_NET_IPGRE)
2768                 case ARPHRD_IPGRE:
2769                         addrconf_gre_config(dev);
2770                         break;
2771 #endif
2772                 case ARPHRD_TUNNEL6:
2773                         addrconf_ip6_tnl_config(dev);
2774                         break;
2775                 case ARPHRD_LOOPBACK:
2776                         init_loopback(dev);
2777                         break;
2778
2779                 default:
2780                         addrconf_dev_config(dev);
2781                         break;
2782                 }
2783
2784                 if (idev) {
2785                         if (run_pending)
2786                                 addrconf_dad_run(idev);
2787
2788                         /*
2789                          * If the MTU changed during the interface down,
2790                          * when the interface up, the changed MTU must be
2791                          * reflected in the idev as well as routers.
2792                          */
2793                         if (idev->cnf.mtu6 != dev->mtu &&
2794                             dev->mtu >= IPV6_MIN_MTU) {
2795                                 rt6_mtu_change(dev, dev->mtu);
2796                                 idev->cnf.mtu6 = dev->mtu;
2797                         }
2798                         idev->tstamp = jiffies;
2799                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2800
2801                         /*
2802                          * If the changed mtu during down is lower than
2803                          * IPV6_MIN_MTU stop IPv6 on this interface.
2804                          */
2805                         if (dev->mtu < IPV6_MIN_MTU)
2806                                 addrconf_ifdown(dev, 1);
2807                 }
2808                 break;
2809
2810         case NETDEV_CHANGEMTU:
2811                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2812                         rt6_mtu_change(dev, dev->mtu);
2813                         idev->cnf.mtu6 = dev->mtu;
2814                         break;
2815                 }
2816
2817                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2818                         idev = ipv6_add_dev(dev);
2819                         if (idev)
2820                                 break;
2821                 }
2822
2823                 /*
2824                  * MTU falled under IPV6_MIN_MTU.
2825                  * Stop IPv6 on this interface.
2826                  */
2827
2828         case NETDEV_DOWN:
2829         case NETDEV_UNREGISTER:
2830                 /*
2831                  *      Remove all addresses from this interface.
2832                  */
2833                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2834                 break;
2835
2836         case NETDEV_CHANGENAME:
2837                 if (idev) {
2838                         snmp6_unregister_dev(idev);
2839                         addrconf_sysctl_unregister(idev);
2840                         addrconf_sysctl_register(idev);
2841                         err = snmp6_register_dev(idev);
2842                         if (err)
2843                                 return notifier_from_errno(err);
2844                 }
2845                 break;
2846
2847         case NETDEV_PRE_TYPE_CHANGE:
2848         case NETDEV_POST_TYPE_CHANGE:
2849                 addrconf_type_change(dev, event);
2850                 break;
2851         }
2852
2853         return NOTIFY_OK;
2854 }
2855
2856 /*
2857  *      addrconf module should be notified of a device going up
2858  */
2859 static struct notifier_block ipv6_dev_notf = {
2860         .notifier_call = addrconf_notify,
2861 };
2862
2863 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2864 {
2865         struct inet6_dev *idev;
2866         ASSERT_RTNL();
2867
2868         idev = __in6_dev_get(dev);
2869
2870         if (event == NETDEV_POST_TYPE_CHANGE)
2871                 ipv6_mc_remap(idev);
2872         else if (event == NETDEV_PRE_TYPE_CHANGE)
2873                 ipv6_mc_unmap(idev);
2874 }
2875
2876 static int addrconf_ifdown(struct net_device *dev, int how)
2877 {
2878         struct net *net = dev_net(dev);
2879         struct inet6_dev *idev;
2880         struct inet6_ifaddr *ifa;
2881         int state, i;
2882
2883         ASSERT_RTNL();
2884
2885         rt6_ifdown(net, dev);
2886         neigh_ifdown(&nd_tbl, dev);
2887
2888         idev = __in6_dev_get(dev);
2889         if (idev == NULL)
2890                 return -ENODEV;
2891
2892         /*
2893          * Step 1: remove reference to ipv6 device from parent device.
2894          *         Do not dev_put!
2895          */
2896         if (how) {
2897                 idev->dead = 1;
2898
2899                 /* protected by rtnl_lock */
2900                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2901
2902                 /* Step 1.5: remove snmp6 entry */
2903                 snmp6_unregister_dev(idev);
2904
2905         }
2906
2907         /* Step 2: clear hash table */
2908         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2909                 struct hlist_head *h = &inet6_addr_lst[i];
2910                 struct hlist_node *n;
2911
2912                 spin_lock_bh(&addrconf_hash_lock);
2913         restart:
2914                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2915                         if (ifa->idev == idev) {
2916                                 hlist_del_init_rcu(&ifa->addr_lst);
2917                                 addrconf_del_timer(ifa);
2918                                 goto restart;
2919                         }
2920                 }
2921                 spin_unlock_bh(&addrconf_hash_lock);
2922         }
2923
2924         write_lock_bh(&idev->lock);
2925
2926         /* Step 2: clear flags for stateless addrconf */
2927         if (!how)
2928                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2929
2930 #ifdef CONFIG_IPV6_PRIVACY
2931         if (how && del_timer(&idev->regen_timer))
2932                 in6_dev_put(idev);
2933
2934         /* Step 3: clear tempaddr list */
2935         while (!list_empty(&idev->tempaddr_list)) {
2936                 ifa = list_first_entry(&idev->tempaddr_list,
2937                                        struct inet6_ifaddr, tmp_list);
2938                 list_del(&ifa->tmp_list);
2939                 write_unlock_bh(&idev->lock);
2940                 spin_lock_bh(&ifa->lock);
2941
2942                 if (ifa->ifpub) {
2943                         in6_ifa_put(ifa->ifpub);
2944                         ifa->ifpub = NULL;
2945                 }
2946                 spin_unlock_bh(&ifa->lock);
2947                 in6_ifa_put(ifa);
2948                 write_lock_bh(&idev->lock);
2949         }
2950 #endif
2951
2952         while (!list_empty(&idev->addr_list)) {
2953                 ifa = list_first_entry(&idev->addr_list,
2954                                        struct inet6_ifaddr, if_list);
2955                 addrconf_del_timer(ifa);
2956
2957                 list_del(&ifa->if_list);
2958
2959                 write_unlock_bh(&idev->lock);
2960
2961                 spin_lock_bh(&ifa->state_lock);
2962                 state = ifa->state;
2963                 ifa->state = INET6_IFADDR_STATE_DEAD;
2964                 spin_unlock_bh(&ifa->state_lock);
2965
2966                 if (state != INET6_IFADDR_STATE_DEAD) {
2967                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2968                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2969                 }
2970                 in6_ifa_put(ifa);
2971
2972                 write_lock_bh(&idev->lock);
2973         }
2974
2975         write_unlock_bh(&idev->lock);
2976
2977         /* Step 5: Discard multicast list */
2978         if (how)
2979                 ipv6_mc_destroy_dev(idev);
2980         else
2981                 ipv6_mc_down(idev);
2982
2983         idev->tstamp = jiffies;
2984
2985         /* Last: Shot the device (if unregistered) */
2986         if (how) {
2987                 addrconf_sysctl_unregister(idev);
2988                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2989                 neigh_ifdown(&nd_tbl, dev);
2990                 in6_dev_put(idev);
2991         }
2992         return 0;
2993 }
2994
2995 static void addrconf_rs_timer(unsigned long data)
2996 {
2997         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2998         struct inet6_dev *idev = ifp->idev;
2999
3000         read_lock(&idev->lock);
3001         if (idev->dead || !(idev->if_flags & IF_READY))
3002                 goto out;
3003
3004         if (!ipv6_accept_ra(idev))
3005                 goto out;
3006
3007         /* Announcement received after solicitation was sent */
3008         if (idev->if_flags & IF_RA_RCVD)
3009                 goto out;
3010
3011         spin_lock(&ifp->lock);
3012         if (ifp->probes++ < idev->cnf.rtr_solicits) {
3013                 /* The wait after the last probe can be shorter */
3014                 addrconf_mod_timer(ifp, AC_RS,
3015                                    (ifp->probes == idev->cnf.rtr_solicits) ?
3016                                    idev->cnf.rtr_solicit_delay :
3017                                    idev->cnf.rtr_solicit_interval);
3018                 spin_unlock(&ifp->lock);
3019
3020                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3021         } else {
3022                 spin_unlock(&ifp->lock);
3023                 /*
3024                  * Note: we do not support deprecated "all on-link"
3025                  * assumption any longer.
3026                  */
3027                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3028         }
3029
3030 out:
3031         read_unlock(&idev->lock);
3032         in6_ifa_put(ifp);
3033 }
3034
3035 /*
3036  *      Duplicate Address Detection
3037  */
3038 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3039 {
3040         unsigned long rand_num;
3041         struct inet6_dev *idev = ifp->idev;
3042
3043         if (ifp->flags & IFA_F_OPTIMISTIC)
3044                 rand_num = 0;
3045         else
3046                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3047
3048         ifp->probes = idev->cnf.dad_transmits;
3049         addrconf_mod_timer(ifp, AC_DAD, rand_num);
3050 }
3051
3052 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3053 {
3054         struct inet6_dev *idev = ifp->idev;
3055         struct net_device *dev = idev->dev;
3056
3057         addrconf_join_solict(dev, &ifp->addr);
3058
3059         net_srandom(ifp->addr.s6_addr32[3]);
3060
3061         read_lock_bh(&idev->lock);
3062         spin_lock(&ifp->lock);
3063         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3064                 goto out;
3065
3066         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3067             idev->cnf.accept_dad < 1 ||
3068             !(ifp->flags&IFA_F_TENTATIVE) ||
3069             ifp->flags & IFA_F_NODAD) {
3070                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3071                 spin_unlock(&ifp->lock);
3072                 read_unlock_bh(&idev->lock);
3073
3074                 addrconf_dad_completed(ifp);
3075                 return;
3076         }
3077
3078         if (!(idev->if_flags & IF_READY)) {
3079                 spin_unlock(&ifp->lock);
3080                 read_unlock_bh(&idev->lock);
3081                 /*
3082                  * If the device is not ready:
3083                  * - keep it tentative if it is a permanent address.
3084                  * - otherwise, kill it.
3085                  */
3086                 in6_ifa_hold(ifp);
3087                 addrconf_dad_stop(ifp, 0);
3088                 return;
3089         }
3090
3091         /*
3092          * Optimistic nodes can start receiving
3093          * Frames right away
3094          */
3095         if (ifp->flags & IFA_F_OPTIMISTIC)
3096                 ip6_ins_rt(ifp->rt);
3097
3098         addrconf_dad_kick(ifp);
3099 out:
3100         spin_unlock(&ifp->lock);
3101         read_unlock_bh(&idev->lock);
3102 }
3103
3104 static void addrconf_dad_timer(unsigned long data)
3105 {
3106         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3107         struct inet6_dev *idev = ifp->idev;
3108         struct in6_addr mcaddr;
3109
3110         if (!ifp->probes && addrconf_dad_end(ifp))
3111                 goto out;
3112
3113         read_lock(&idev->lock);
3114         if (idev->dead || !(idev->if_flags & IF_READY)) {
3115                 read_unlock(&idev->lock);
3116                 goto out;
3117         }
3118
3119         spin_lock(&ifp->lock);
3120         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3121                 spin_unlock(&ifp->lock);
3122                 read_unlock(&idev->lock);
3123                 goto out;
3124         }
3125
3126         if (ifp->probes == 0) {
3127                 /*
3128                  * DAD was successful
3129                  */
3130
3131                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3132                 spin_unlock(&ifp->lock);
3133                 read_unlock(&idev->lock);
3134
3135                 addrconf_dad_completed(ifp);
3136
3137                 goto out;
3138         }
3139
3140         ifp->probes--;
3141         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3142         spin_unlock(&ifp->lock);
3143         read_unlock(&idev->lock);
3144
3145         /* send a neighbour solicitation for our addr */
3146         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3147         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3148 out:
3149         in6_ifa_put(ifp);
3150 }
3151
3152 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3153 {
3154         struct net_device *dev = ifp->idev->dev;
3155
3156         /*
3157          *      Configure the address for reception. Now it is valid.
3158          */
3159
3160         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3161
3162         /* If added prefix is link local and we are prepared to process
3163            router advertisements, start sending router solicitations.
3164          */
3165
3166         if (ipv6_accept_ra(ifp->idev) &&
3167             ifp->idev->cnf.rtr_solicits > 0 &&
3168             (dev->flags&IFF_LOOPBACK) == 0 &&
3169             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3170                 /*
3171                  *      If a host as already performed a random delay
3172                  *      [...] as part of DAD [...] there is no need
3173                  *      to delay again before sending the first RS
3174                  */
3175                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3176
3177                 spin_lock_bh(&ifp->lock);
3178                 ifp->probes = 1;
3179                 ifp->idev->if_flags |= IF_RS_SENT;
3180                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3181                 spin_unlock_bh(&ifp->lock);
3182         }
3183 }
3184
3185 static void addrconf_dad_run(struct inet6_dev *idev)
3186 {
3187         struct inet6_ifaddr *ifp;
3188
3189         read_lock_bh(&idev->lock);
3190         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3191                 spin_lock(&ifp->lock);
3192                 if (ifp->flags & IFA_F_TENTATIVE &&
3193                     ifp->state == INET6_IFADDR_STATE_DAD)
3194                         addrconf_dad_kick(ifp);
3195                 spin_unlock(&ifp->lock);
3196         }
3197         read_unlock_bh(&idev->lock);
3198 }
3199
3200 #ifdef CONFIG_PROC_FS
3201 struct if6_iter_state {
3202         struct seq_net_private p;
3203         int bucket;
3204         int offset;
3205 };
3206
3207 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3208 {
3209         struct inet6_ifaddr *ifa = NULL;
3210         struct if6_iter_state *state = seq->private;
3211         struct net *net = seq_file_net(seq);
3212         int p = 0;
3213
3214         /* initial bucket if pos is 0 */
3215         if (pos == 0) {
3216                 state->bucket = 0;
3217                 state->offset = 0;
3218         }
3219
3220         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3221                 struct hlist_node *n;
3222                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3223                                          addr_lst) {
3224                         if (!net_eq(dev_net(ifa->idev->dev), net))
3225                                 continue;
3226                         /* sync with offset */
3227                         if (p < state->offset) {
3228                                 p++;
3229                                 continue;
3230                         }
3231                         state->offset++;
3232                         return ifa;
3233                 }
3234
3235                 /* prepare for next bucket */
3236                 state->offset = 0;
3237                 p = 0;
3238         }
3239         return NULL;
3240 }
3241
3242 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3243                                          struct inet6_ifaddr *ifa)
3244 {
3245         struct if6_iter_state *state = seq->private;
3246         struct net *net = seq_file_net(seq);
3247         struct hlist_node *n = &ifa->addr_lst;
3248
3249         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3250                 if (!net_eq(dev_net(ifa->idev->dev), net))
3251                         continue;
3252                 state->offset++;
3253                 return ifa;
3254         }
3255
3256         while (++state->bucket < IN6_ADDR_HSIZE) {
3257                 state->offset = 0;
3258                 hlist_for_each_entry_rcu_bh(ifa, n,
3259                                      &inet6_addr_lst[state->bucket], addr_lst) {
3260                         if (!net_eq(dev_net(ifa->idev->dev), net))
3261                                 continue;
3262                         state->offset++;
3263                         return ifa;
3264                 }
3265         }
3266
3267         return NULL;
3268 }
3269
3270 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3271         __acquires(rcu_bh)
3272 {
3273         rcu_read_lock_bh();
3274         return if6_get_first(seq, *pos);
3275 }
3276
3277 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3278 {
3279         struct inet6_ifaddr *ifa;
3280
3281         ifa = if6_get_next(seq, v);
3282         ++*pos;
3283         return ifa;
3284 }
3285
3286 static void if6_seq_stop(struct seq_file *seq, void *v)
3287         __releases(rcu_bh)
3288 {
3289         rcu_read_unlock_bh();
3290 }
3291
3292 static int if6_seq_show(struct seq_file *seq, void *v)
3293 {
3294         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3295         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3296                    &ifp->addr,
3297                    ifp->idev->dev->ifindex,
3298                    ifp->prefix_len,
3299                    ifp->scope,
3300                    ifp->flags,
3301                    ifp->idev->dev->name);
3302         return 0;
3303 }
3304
3305 static const struct seq_operations if6_seq_ops = {
3306         .start  = if6_seq_start,
3307         .next   = if6_seq_next,
3308         .show   = if6_seq_show,
3309         .stop   = if6_seq_stop,
3310 };
3311
3312 static int if6_seq_open(struct inode *inode, struct file *file)
3313 {
3314         return seq_open_net(inode, file, &if6_seq_ops,
3315                             sizeof(struct if6_iter_state));
3316 }
3317
3318 static const struct file_operations if6_fops = {
3319         .owner          = THIS_MODULE,
3320         .open           = if6_seq_open,
3321         .read           = seq_read,
3322         .llseek         = seq_lseek,
3323         .release        = seq_release_net,
3324 };
3325
3326 static int __net_init if6_proc_net_init(struct net *net)
3327 {
3328         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3329                 return -ENOMEM;
3330         return 0;
3331 }
3332
3333 static void __net_exit if6_proc_net_exit(struct net *net)
3334 {
3335         remove_proc_entry("if_inet6", net->proc_net);
3336 }
3337
3338 static struct pernet_operations if6_proc_net_ops = {
3339        .init = if6_proc_net_init,
3340        .exit = if6_proc_net_exit,
3341 };
3342
3343 int __init if6_proc_init(void)
3344 {
3345         return register_pernet_subsys(&if6_proc_net_ops);
3346 }
3347
3348 void if6_proc_exit(void)
3349 {
3350         unregister_pernet_subsys(&if6_proc_net_ops);
3351 }
3352 #endif  /* CONFIG_PROC_FS */
3353
3354 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3355 /* Check if address is a home address configured on any interface. */
3356 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3357 {
3358         int ret = 0;
3359         struct inet6_ifaddr *ifp = NULL;
3360         struct hlist_node *n;
3361         unsigned int hash = inet6_addr_hash(addr);
3362
3363         rcu_read_lock_bh();
3364         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3365                 if (!net_eq(dev_net(ifp->idev->dev), net))
3366                         continue;
3367                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3368                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3369                         ret = 1;
3370                         break;
3371                 }
3372         }
3373         rcu_read_unlock_bh();
3374         return ret;
3375 }
3376 #endif
3377
3378 /*
3379  *      Periodic address status verification
3380  */
3381
3382 static void addrconf_verify(unsigned long foo)
3383 {
3384         unsigned long now, next, next_sec, next_sched;
3385         struct inet6_ifaddr *ifp;
3386         struct hlist_node *node;
3387         int i;
3388
3389         rcu_read_lock_bh();
3390         spin_lock(&addrconf_verify_lock);
3391         now = jiffies;
3392         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3393
3394         del_timer(&addr_chk_timer);
3395
3396         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3397 restart:
3398                 hlist_for_each_entry_rcu_bh(ifp, node,
3399                                          &inet6_addr_lst[i], addr_lst) {
3400                         unsigned long age;
3401
3402                         if (ifp->flags & IFA_F_PERMANENT)
3403                                 continue;
3404
3405                         spin_lock(&ifp->lock);
3406                         /* We try to batch several events at once. */
3407                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3408
3409                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3410                             age >= ifp->valid_lft) {
3411                                 spin_unlock(&ifp->lock);
3412                                 in6_ifa_hold(ifp);
3413                                 ipv6_del_addr(ifp);
3414                                 goto restart;
3415                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3416                                 spin_unlock(&ifp->lock);
3417                                 continue;
3418                         } else if (age >= ifp->prefered_lft) {
3419                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3420                                 int deprecate = 0;
3421
3422                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3423                                         deprecate = 1;
3424                                         ifp->flags |= IFA_F_DEPRECATED;
3425                                 }
3426
3427                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3428                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3429
3430                                 spin_unlock(&ifp->lock);
3431
3432                                 if (deprecate) {
3433                                         in6_ifa_hold(ifp);
3434
3435                                         ipv6_ifa_notify(0, ifp);
3436                                         in6_ifa_put(ifp);
3437                                         goto restart;
3438                                 }
3439 #ifdef CONFIG_IPV6_PRIVACY
3440                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3441                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3442                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3443                                         ifp->idev->cnf.dad_transmits *
3444                                         ifp->idev->nd_parms->retrans_time / HZ;
3445
3446                                 if (age >= ifp->prefered_lft - regen_advance) {
3447                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3448                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3449                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3450                                         if (!ifp->regen_count && ifpub) {
3451                                                 ifp->regen_count++;
3452                                                 in6_ifa_hold(ifp);
3453                                                 in6_ifa_hold(ifpub);
3454                                                 spin_unlock(&ifp->lock);
3455
3456                                                 spin_lock(&ifpub->lock);
3457                                                 ifpub->regen_count = 0;
3458                                                 spin_unlock(&ifpub->lock);
3459                                                 ipv6_create_tempaddr(ifpub, ifp);
3460                                                 in6_ifa_put(ifpub);
3461                                                 in6_ifa_put(ifp);
3462                                                 goto restart;
3463                                         }
3464                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3465                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3466                                 spin_unlock(&ifp->lock);
3467 #endif
3468                         } else {
3469                                 /* ifp->prefered_lft <= ifp->valid_lft */
3470                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3471                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3472                                 spin_unlock(&ifp->lock);
3473                         }
3474                 }
3475         }
3476
3477         next_sec = round_jiffies_up(next);
3478         next_sched = next;
3479
3480         /* If rounded timeout is accurate enough, accept it. */
3481         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3482                 next_sched = next_sec;
3483
3484         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3485         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3486                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3487
3488         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3489               now, next, next_sec, next_sched));
3490
3491         addr_chk_timer.expires = next_sched;
3492         add_timer(&addr_chk_timer);
3493         spin_unlock(&addrconf_verify_lock);
3494         rcu_read_unlock_bh();
3495 }
3496
3497 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3498 {
3499         struct in6_addr *pfx = NULL;
3500
3501         if (addr)
3502                 pfx = nla_data(addr);
3503
3504         if (local) {
3505                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3506                         pfx = NULL;
3507                 else
3508                         pfx = nla_data(local);
3509         }
3510
3511         return pfx;
3512 }
3513
3514 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3515         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3516         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3517         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3518 };
3519
3520 static int
3521 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3522 {
3523         struct net *net = sock_net(skb->sk);
3524         struct ifaddrmsg *ifm;
3525         struct nlattr *tb[IFA_MAX+1];
3526         struct in6_addr *pfx;
3527         int err;
3528
3529         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3530         if (err < 0)
3531                 return err;
3532
3533         ifm = nlmsg_data(nlh);
3534         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3535         if (pfx == NULL)
3536                 return -EINVAL;
3537
3538         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3539 }
3540
3541 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3542                              u32 prefered_lft, u32 valid_lft)
3543 {
3544         u32 flags;
3545         clock_t expires;
3546         unsigned long timeout;
3547
3548         if (!valid_lft || (prefered_lft > valid_lft))
3549                 return -EINVAL;
3550
3551         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3552         if (addrconf_finite_timeout(timeout)) {
3553                 expires = jiffies_to_clock_t(timeout * HZ);
3554                 valid_lft = timeout;
3555                 flags = RTF_EXPIRES;
3556         } else {
3557                 expires = 0;
3558                 flags = 0;
3559                 ifa_flags |= IFA_F_PERMANENT;
3560         }
3561
3562         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3563         if (addrconf_finite_timeout(timeout)) {
3564                 if (timeout == 0)
3565                         ifa_flags |= IFA_F_DEPRECATED;
3566                 prefered_lft = timeout;
3567         }
3568
3569         spin_lock_bh(&ifp->lock);
3570         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3571         ifp->tstamp = jiffies;
3572         ifp->valid_lft = valid_lft;
3573         ifp->prefered_lft = prefered_lft;
3574
3575         spin_unlock_bh(&ifp->lock);
3576         if (!(ifp->flags&IFA_F_TENTATIVE))
3577                 ipv6_ifa_notify(0, ifp);
3578
3579         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3580                               expires, flags);
3581         addrconf_verify(0);
3582
3583         return 0;
3584 }
3585
3586 static int
3587 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3588 {
3589         struct net *net = sock_net(skb->sk);
3590         struct ifaddrmsg *ifm;
3591         struct nlattr *tb[IFA_MAX+1];
3592         struct in6_addr *pfx;
3593         struct inet6_ifaddr *ifa;
3594         struct net_device *dev;
3595         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3596         u8 ifa_flags;
3597         int err;
3598
3599         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3600         if (err < 0)
3601                 return err;
3602
3603         ifm = nlmsg_data(nlh);
3604         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3605         if (pfx == NULL)
3606                 return -EINVAL;
3607
3608         if (tb[IFA_CACHEINFO]) {
3609                 struct ifa_cacheinfo *ci;
3610
3611                 ci = nla_data(tb[IFA_CACHEINFO]);
3612                 valid_lft = ci->ifa_valid;
3613                 preferred_lft = ci->ifa_prefered;
3614         } else {
3615                 preferred_lft = INFINITY_LIFE_TIME;
3616                 valid_lft = INFINITY_LIFE_TIME;
3617         }
3618
3619         dev =  __dev_get_by_index(net, ifm->ifa_index);
3620         if (dev == NULL)
3621                 return -ENODEV;
3622
3623         /* We ignore other flags so far. */
3624         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3625
3626         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3627         if (ifa == NULL) {
3628                 /*
3629                  * It would be best to check for !NLM_F_CREATE here but
3630                  * userspace alreay relies on not having to provide this.
3631                  */
3632                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3633                                       ifm->ifa_prefixlen, ifa_flags,
3634                                       preferred_lft, valid_lft);
3635         }
3636
3637         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3638             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3639                 err = -EEXIST;
3640         else
3641                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3642
3643         in6_ifa_put(ifa);
3644
3645         return err;
3646 }
3647
3648 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3649                           u8 scope, int ifindex)
3650 {
3651         struct ifaddrmsg *ifm;
3652
3653         ifm = nlmsg_data(nlh);
3654         ifm->ifa_family = AF_INET6;
3655         ifm->ifa_prefixlen = prefixlen;
3656         ifm->ifa_flags = flags;
3657         ifm->ifa_scope = scope;
3658         ifm->ifa_index = ifindex;
3659 }
3660
3661 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3662                          unsigned long tstamp, u32 preferred, u32 valid)
3663 {
3664         struct ifa_cacheinfo ci;
3665
3666         ci.cstamp = cstamp_delta(cstamp);
3667         ci.tstamp = cstamp_delta(tstamp);
3668         ci.ifa_prefered = preferred;
3669         ci.ifa_valid = valid;
3670
3671         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3672 }
3673
3674 static inline int rt_scope(int ifa_scope)
3675 {
3676         if (ifa_scope & IFA_HOST)
3677                 return RT_SCOPE_HOST;
3678         else if (ifa_scope & IFA_LINK)
3679                 return RT_SCOPE_LINK;
3680         else if (ifa_scope & IFA_SITE)
3681                 return RT_SCOPE_SITE;
3682         else
3683                 return RT_SCOPE_UNIVERSE;
3684 }
3685
3686 static inline int inet6_ifaddr_msgsize(void)
3687 {
3688         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3689                + nla_total_size(16) /* IFA_ADDRESS */
3690                + nla_total_size(sizeof(struct ifa_cacheinfo));
3691 }
3692
3693 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3694                              u32 portid, u32 seq, int event, unsigned int flags)
3695 {
3696         struct nlmsghdr  *nlh;
3697         u32 preferred, valid;
3698
3699         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3700         if (nlh == NULL)
3701                 return -EMSGSIZE;
3702
3703         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3704                       ifa->idev->dev->ifindex);
3705
3706         if (!(ifa->flags&IFA_F_PERMANENT)) {
3707                 preferred = ifa->prefered_lft;
3708                 valid = ifa->valid_lft;
3709                 if (preferred != INFINITY_LIFE_TIME) {
3710                         long tval = (jiffies - ifa->tstamp)/HZ;
3711                         if (preferred > tval)
3712                                 preferred -= tval;
3713                         else
3714                                 preferred = 0;
3715                         if (valid != INFINITY_LIFE_TIME) {
3716                                 if (valid > tval)
3717                                         valid -= tval;
3718                                 else
3719                                         valid = 0;
3720                         }
3721                 }
3722         } else {
3723                 preferred = INFINITY_LIFE_TIME;
3724                 valid = INFINITY_LIFE_TIME;
3725         }
3726
3727         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3728             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3729                 nlmsg_cancel(skb, nlh);
3730                 return -EMSGSIZE;
3731         }
3732
3733         return nlmsg_end(skb, nlh);
3734 }
3735
3736 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3737                                 u32 portid, u32 seq, int event, u16 flags)
3738 {
3739         struct nlmsghdr  *nlh;
3740         u8 scope = RT_SCOPE_UNIVERSE;
3741         int ifindex = ifmca->idev->dev->ifindex;
3742
3743         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3744                 scope = RT_SCOPE_SITE;
3745
3746         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3747         if (nlh == NULL)
3748                 return -EMSGSIZE;
3749
3750         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3751         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3752             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3753                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3754                 nlmsg_cancel(skb, nlh);
3755                 return -EMSGSIZE;
3756         }
3757
3758         return nlmsg_end(skb, nlh);
3759 }
3760
3761 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3762                                 u32 portid, u32 seq, int event, unsigned int flags)
3763 {
3764         struct nlmsghdr  *nlh;
3765         u8 scope = RT_SCOPE_UNIVERSE;
3766         int ifindex = ifaca->aca_idev->dev->ifindex;
3767
3768         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3769                 scope = RT_SCOPE_SITE;
3770
3771         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3772         if (nlh == NULL)
3773                 return -EMSGSIZE;
3774
3775         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3776         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3777             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3778                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3779                 nlmsg_cancel(skb, nlh);
3780                 return -EMSGSIZE;
3781         }
3782
3783         return nlmsg_end(skb, nlh);
3784 }
3785
3786 enum addr_type_t {
3787         UNICAST_ADDR,
3788         MULTICAST_ADDR,
3789         ANYCAST_ADDR,
3790 };
3791
3792 /* called with rcu_read_lock() */
3793 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3794                           struct netlink_callback *cb, enum addr_type_t type,
3795                           int s_ip_idx, int *p_ip_idx)
3796 {
3797         struct ifmcaddr6 *ifmca;
3798         struct ifacaddr6 *ifaca;
3799         int err = 1;
3800         int ip_idx = *p_ip_idx;
3801
3802         read_lock_bh(&idev->lock);
3803         switch (type) {
3804         case UNICAST_ADDR: {
3805                 struct inet6_ifaddr *ifa;
3806
3807                 /* unicast address incl. temp addr */
3808                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3809                         if (++ip_idx < s_ip_idx)
3810                                 continue;
3811                         err = inet6_fill_ifaddr(skb, ifa,
3812                                                 NETLINK_CB(cb->skb).portid,
3813                                                 cb->nlh->nlmsg_seq,
3814                                                 RTM_NEWADDR,
3815                                                 NLM_F_MULTI);
3816                         if (err <= 0)
3817                                 break;
3818                 }
3819                 break;
3820         }
3821         case MULTICAST_ADDR:
3822                 /* multicast address */
3823                 for (ifmca = idev->mc_list; ifmca;
3824                      ifmca = ifmca->next, ip_idx++) {
3825                         if (ip_idx < s_ip_idx)
3826                                 continue;
3827                         err = inet6_fill_ifmcaddr(skb, ifmca,
3828                                                   NETLINK_CB(cb->skb).portid,
3829                                                   cb->nlh->nlmsg_seq,
3830                                                   RTM_GETMULTICAST,
3831                                                   NLM_F_MULTI);
3832                         if (err <= 0)
3833                                 break;
3834                 }
3835                 break;
3836         case ANYCAST_ADDR:
3837                 /* anycast address */
3838                 for (ifaca = idev->ac_list; ifaca;
3839                      ifaca = ifaca->aca_next, ip_idx++) {
3840                         if (ip_idx < s_ip_idx)
3841                                 continue;
3842                         err = inet6_fill_ifacaddr(skb, ifaca,
3843                                                   NETLINK_CB(cb->skb).portid,
3844                                                   cb->nlh->nlmsg_seq,
3845                                                   RTM_GETANYCAST,
3846                                                   NLM_F_MULTI);
3847                         if (err <= 0)
3848                                 break;
3849                 }
3850                 break;
3851         default:
3852                 break;
3853         }
3854         read_unlock_bh(&idev->lock);
3855         *p_ip_idx = ip_idx;
3856         return err;
3857 }
3858
3859 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3860                            enum addr_type_t type)
3861 {
3862         struct net *net = sock_net(skb->sk);
3863         int h, s_h;
3864         int idx, ip_idx;
3865         int s_idx, s_ip_idx;
3866         struct net_device *dev;
3867         struct inet6_dev *idev;
3868         struct hlist_head *head;
3869         struct hlist_node *node;
3870
3871         s_h = cb->args[0];
3872         s_idx = idx = cb->args[1];
3873         s_ip_idx = ip_idx = cb->args[2];
3874
3875         rcu_read_lock();
3876         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3877                 idx = 0;
3878                 head = &net->dev_index_head[h];
3879                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3880                         if (idx < s_idx)
3881                                 goto cont;
3882                         if (h > s_h || idx > s_idx)
3883                                 s_ip_idx = 0;
3884                         ip_idx = 0;
3885                         idev = __in6_dev_get(dev);
3886                         if (!idev)
3887                                 goto cont;
3888
3889                         if (in6_dump_addrs(idev, skb, cb, type,
3890                                            s_ip_idx, &ip_idx) <= 0)
3891                                 goto done;
3892 cont:
3893                         idx++;
3894                 }
3895         }
3896 done:
3897         rcu_read_unlock();
3898         cb->args[0] = h;
3899         cb->args[1] = idx;
3900         cb->args[2] = ip_idx;
3901
3902         return skb->len;
3903 }
3904
3905 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3906 {
3907         enum addr_type_t type = UNICAST_ADDR;
3908
3909         return inet6_dump_addr(skb, cb, type);
3910 }
3911
3912 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3913 {
3914         enum addr_type_t type = MULTICAST_ADDR;
3915
3916         return inet6_dump_addr(skb, cb, type);
3917 }
3918
3919
3920 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3921 {
3922         enum addr_type_t type = ANYCAST_ADDR;
3923
3924         return inet6_dump_addr(skb, cb, type);
3925 }
3926
3927 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3928                              void *arg)
3929 {
3930         struct net *net = sock_net(in_skb->sk);
3931         struct ifaddrmsg *ifm;
3932         struct nlattr *tb[IFA_MAX+1];
3933         struct in6_addr *addr = NULL;
3934         struct net_device *dev = NULL;
3935         struct inet6_ifaddr *ifa;
3936         struct sk_buff *skb;
3937         int err;
3938
3939         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3940         if (err < 0)
3941                 goto errout;
3942
3943         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3944         if (addr == NULL) {
3945                 err = -EINVAL;
3946                 goto errout;
3947         }
3948
3949         ifm = nlmsg_data(nlh);
3950         if (ifm->ifa_index)
3951                 dev = __dev_get_by_index(net, ifm->ifa_index);
3952
3953         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3954         if (!ifa) {
3955                 err = -EADDRNOTAVAIL;
3956                 goto errout;
3957         }
3958
3959         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3960         if (!skb) {
3961                 err = -ENOBUFS;
3962                 goto errout_ifa;
3963         }
3964
3965         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3966                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3967         if (err < 0) {
3968                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3969                 WARN_ON(err == -EMSGSIZE);
3970                 kfree_skb(skb);
3971                 goto errout_ifa;
3972         }
3973         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3974 errout_ifa:
3975         in6_ifa_put(ifa);
3976 errout:
3977         return err;
3978 }
3979
3980 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3981 {
3982         struct sk_buff *skb;
3983         struct net *net = dev_net(ifa->idev->dev);
3984         int err = -ENOBUFS;
3985
3986         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3987         if (skb == NULL)
3988                 goto errout;
3989
3990         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3991         if (err < 0) {
3992                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3993                 WARN_ON(err == -EMSGSIZE);
3994                 kfree_skb(skb);
3995                 goto errout;
3996         }
3997         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3998         return;
3999 errout:
4000         if (err < 0)
4001                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4002 }
4003
4004 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4005                                 __s32 *array, int bytes)
4006 {
4007         BUG_ON(bytes < (DEVCONF_MAX * 4));
4008
4009         memset(array, 0, bytes);
4010         array[DEVCONF_FORWARDING] = cnf->forwarding;
4011         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4012         array[DEVCONF_MTU6] = cnf->mtu6;
4013         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4014         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4015         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4016         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4017         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4018         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4019                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4020         array[DEVCONF_RTR_SOLICIT_DELAY] =
4021                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4022         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4023 #ifdef CONFIG_IPV6_PRIVACY
4024         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4025         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4026         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4027         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4028         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4029 #endif
4030         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4031         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4032         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4033 #ifdef CONFIG_IPV6_ROUTER_PREF
4034         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4035         array[DEVCONF_RTR_PROBE_INTERVAL] =
4036                 jiffies_to_msecs(cnf->rtr_probe_interval);
4037 #ifdef CONFIG_IPV6_ROUTE_INFO
4038         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4039 #endif
4040 #endif
4041         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4042         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4043 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4044         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4045 #endif
4046 #ifdef CONFIG_IPV6_MROUTE
4047         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4048 #endif
4049         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4050         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4051         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4052         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4053 }
4054
4055 static inline size_t inet6_ifla6_size(void)
4056 {
4057         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4058              + nla_total_size(sizeof(struct ifla_cacheinfo))
4059              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4060              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4061              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4062 }
4063
4064 static inline size_t inet6_if_nlmsg_size(void)
4065 {
4066         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4067                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4068                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4069                + nla_total_size(4) /* IFLA_MTU */
4070                + nla_total_size(4) /* IFLA_LINK */
4071                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4072 }
4073
4074 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4075                                       int items, int bytes)
4076 {
4077         int i;
4078         int pad = bytes - sizeof(u64) * items;
4079         BUG_ON(pad < 0);
4080
4081         /* Use put_unaligned() because stats may not be aligned for u64. */
4082         put_unaligned(items, &stats[0]);
4083         for (i = 1; i < items; i++)
4084                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4085
4086         memset(&stats[items], 0, pad);
4087 }
4088
4089 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4090                                       int items, int bytes, size_t syncpoff)
4091 {
4092         int i;
4093         int pad = bytes - sizeof(u64) * items;
4094         BUG_ON(pad < 0);
4095
4096         /* Use put_unaligned() because stats may not be aligned for u64. */
4097         put_unaligned(items, &stats[0]);
4098         for (i = 1; i < items; i++)
4099                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4100
4101         memset(&stats[items], 0, pad);
4102 }
4103
4104 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4105                              int bytes)
4106 {
4107         switch (attrtype) {
4108         case IFLA_INET6_STATS:
4109                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4110                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4111                 break;
4112         case IFLA_INET6_ICMP6STATS:
4113                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4114                 break;
4115         }
4116 }
4117
4118 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4119 {
4120         struct nlattr *nla;
4121         struct ifla_cacheinfo ci;
4122
4123         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4124                 goto nla_put_failure;
4125         ci.max_reasm_len = IPV6_MAXPLEN;
4126         ci.tstamp = cstamp_delta(idev->tstamp);
4127         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4128         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4129         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4130                 goto nla_put_failure;
4131         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4132         if (nla == NULL)
4133                 goto nla_put_failure;
4134         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4135
4136         /* XXX - MC not implemented */
4137
4138         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4139         if (nla == NULL)
4140                 goto nla_put_failure;
4141         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4142
4143         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4144         if (nla == NULL)
4145                 goto nla_put_failure;
4146         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4147
4148         return 0;
4149
4150 nla_put_failure:
4151         return -EMSGSIZE;
4152 }
4153
4154 static size_t inet6_get_link_af_size(const struct net_device *dev)
4155 {
4156         if (!__in6_dev_get(dev))
4157                 return 0;
4158
4159         return inet6_ifla6_size();
4160 }
4161
4162 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4163 {
4164         struct inet6_dev *idev = __in6_dev_get(dev);
4165
4166         if (!idev)
4167                 return -ENODATA;
4168
4169         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4170                 return -EMSGSIZE;
4171
4172         return 0;
4173 }
4174
4175 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4176                              u32 portid, u32 seq, int event, unsigned int flags)
4177 {
4178         struct net_device *dev = idev->dev;
4179         struct ifinfomsg *hdr;
4180         struct nlmsghdr *nlh;
4181         void *protoinfo;
4182
4183         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4184         if (nlh == NULL)
4185                 return -EMSGSIZE;
4186
4187         hdr = nlmsg_data(nlh);
4188         hdr->ifi_family = AF_INET6;
4189         hdr->__ifi_pad = 0;
4190         hdr->ifi_type = dev->type;
4191         hdr->ifi_index = dev->ifindex;
4192         hdr->ifi_flags = dev_get_flags(dev);
4193         hdr->ifi_change = 0;
4194
4195         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4196             (dev->addr_len &&
4197              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4198             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4199             (dev->ifindex != dev->iflink &&
4200              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4201                 goto nla_put_failure;
4202         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4203         if (protoinfo == NULL)
4204                 goto nla_put_failure;
4205
4206         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4207                 goto nla_put_failure;
4208
4209         nla_nest_end(skb, protoinfo);
4210         return nlmsg_end(skb, nlh);
4211
4212 nla_put_failure:
4213         nlmsg_cancel(skb, nlh);
4214         return -EMSGSIZE;
4215 }
4216
4217 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4218 {
4219         struct net *net = sock_net(skb->sk);
4220         int h, s_h;
4221         int idx = 0, s_idx;
4222         struct net_device *dev;
4223         struct inet6_dev *idev;
4224         struct hlist_head *head;
4225         struct hlist_node *node;
4226
4227         s_h = cb->args[0];
4228         s_idx = cb->args[1];
4229
4230         rcu_read_lock();
4231         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4232                 idx = 0;
4233                 head = &net->dev_index_head[h];
4234                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4235                         if (idx < s_idx)
4236                                 goto cont;
4237                         idev = __in6_dev_get(dev);
4238                         if (!idev)
4239                                 goto cont;
4240                         if (inet6_fill_ifinfo(skb, idev,
4241                                               NETLINK_CB(cb->skb).portid,
4242                                               cb->nlh->nlmsg_seq,
4243                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4244                                 goto out;
4245 cont:
4246                         idx++;
4247                 }
4248         }
4249 out:
4250         rcu_read_unlock();
4251         cb->args[1] = idx;
4252         cb->args[0] = h;
4253
4254         return skb->len;
4255 }
4256
4257 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4258 {
4259         struct sk_buff *skb;
4260         struct net *net = dev_net(idev->dev);
4261         int err = -ENOBUFS;
4262
4263         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4264         if (skb == NULL)
4265                 goto errout;
4266
4267         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4268         if (err < 0) {
4269                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4270                 WARN_ON(err == -EMSGSIZE);
4271                 kfree_skb(skb);
4272                 goto errout;
4273         }
4274         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4275         return;
4276 errout:
4277         if (err < 0)
4278                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4279 }
4280
4281 static inline size_t inet6_prefix_nlmsg_size(void)
4282 {
4283         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4284                + nla_total_size(sizeof(struct in6_addr))
4285                + nla_total_size(sizeof(struct prefix_cacheinfo));
4286 }
4287
4288 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4289                              struct prefix_info *pinfo, u32 portid, u32 seq,
4290                              int event, unsigned int flags)
4291 {
4292         struct prefixmsg *pmsg;
4293         struct nlmsghdr *nlh;
4294         struct prefix_cacheinfo ci;
4295
4296         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4297         if (nlh == NULL)
4298                 return -EMSGSIZE;
4299
4300         pmsg = nlmsg_data(nlh);
4301         pmsg->prefix_family = AF_INET6;
4302         pmsg->prefix_pad1 = 0;
4303         pmsg->prefix_pad2 = 0;
4304         pmsg->prefix_ifindex = idev->dev->ifindex;
4305         pmsg->prefix_len = pinfo->prefix_len;
4306         pmsg->prefix_type = pinfo->type;
4307         pmsg->prefix_pad3 = 0;
4308         pmsg->prefix_flags = 0;
4309         if (pinfo->onlink)
4310                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4311         if (pinfo->autoconf)
4312                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4313
4314         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4315                 goto nla_put_failure;
4316         ci.preferred_time = ntohl(pinfo->prefered);
4317         ci.valid_time = ntohl(pinfo->valid);
4318         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4319                 goto nla_put_failure;
4320         return nlmsg_end(skb, nlh);
4321
4322 nla_put_failure:
4323         nlmsg_cancel(skb, nlh);
4324         return -EMSGSIZE;
4325 }
4326
4327 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4328                          struct prefix_info *pinfo)
4329 {
4330         struct sk_buff *skb;
4331         struct net *net = dev_net(idev->dev);
4332         int err = -ENOBUFS;
4333
4334         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4335         if (skb == NULL)
4336                 goto errout;
4337
4338         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4339         if (err < 0) {
4340                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4341                 WARN_ON(err == -EMSGSIZE);
4342                 kfree_skb(skb);
4343                 goto errout;
4344         }
4345         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4346         return;
4347 errout:
4348         if (err < 0)
4349                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4350 }
4351
4352 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4353 {
4354         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4355
4356         switch (event) {
4357         case RTM_NEWADDR:
4358                 /*
4359                  * If the address was optimistic
4360                  * we inserted the route at the start of
4361                  * our DAD process, so we don't need
4362                  * to do it again
4363                  */
4364                 if (!(ifp->rt->rt6i_node))
4365                         ip6_ins_rt(ifp->rt);
4366                 if (ifp->idev->cnf.forwarding)
4367                         addrconf_join_anycast(ifp);
4368                 break;
4369         case RTM_DELADDR:
4370                 if (ifp->idev->cnf.forwarding)
4371                         addrconf_leave_anycast(ifp);
4372                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4373                 dst_hold(&ifp->rt->dst);
4374
4375                 if (ip6_del_rt(ifp->rt))
4376                         dst_free(&ifp->rt->dst);
4377                 break;
4378         }
4379 }
4380
4381 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4382 {
4383         rcu_read_lock_bh();
4384         if (likely(ifp->idev->dead == 0))
4385                 __ipv6_ifa_notify(event, ifp);
4386         rcu_read_unlock_bh();
4387 }
4388
4389 #ifdef CONFIG_SYSCTL
4390
4391 static
4392 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4393                            void __user *buffer, size_t *lenp, loff_t *ppos)
4394 {
4395         int *valp = ctl->data;
4396         int val = *valp;
4397         loff_t pos = *ppos;
4398         ctl_table lctl;
4399         int ret;
4400
4401         /*
4402          * ctl->data points to idev->cnf.forwarding, we should
4403          * not modify it until we get the rtnl lock.
4404          */
4405         lctl = *ctl;
4406         lctl.data = &val;
4407
4408         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4409
4410         if (write)
4411                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4412         if (ret)
4413                 *ppos = pos;
4414         return ret;
4415 }
4416
4417 static void dev_disable_change(struct inet6_dev *idev)
4418 {
4419         if (!idev || !idev->dev)
4420                 return;
4421
4422         if (idev->cnf.disable_ipv6)
4423                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4424         else
4425                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4426 }
4427
4428 static void addrconf_disable_change(struct net *net, __s32 newf)
4429 {
4430         struct net_device *dev;
4431         struct inet6_dev *idev;
4432
4433         rcu_read_lock();
4434         for_each_netdev_rcu(net, dev) {
4435                 idev = __in6_dev_get(dev);
4436                 if (idev) {
4437                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4438                         idev->cnf.disable_ipv6 = newf;
4439                         if (changed)
4440                                 dev_disable_change(idev);
4441                 }
4442         }
4443         rcu_read_unlock();
4444 }
4445
4446 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4447 {
4448         struct net *net;
4449         int old;
4450
4451         if (!rtnl_trylock())
4452                 return restart_syscall();
4453
4454         net = (struct net *)table->extra2;
4455         old = *p;
4456         *p = newf;
4457
4458         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4459                 rtnl_unlock();
4460                 return 0;
4461         }
4462
4463         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4464                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4465                 addrconf_disable_change(net, newf);
4466         } else if ((!newf) ^ (!old))
4467                 dev_disable_change((struct inet6_dev *)table->extra1);
4468
4469         rtnl_unlock();
4470         return 0;
4471 }
4472
4473 static
4474 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4475                             void __user *buffer, size_t *lenp, loff_t *ppos)
4476 {
4477         int *valp = ctl->data;
4478         int val = *valp;
4479         loff_t pos = *ppos;
4480         ctl_table lctl;
4481         int ret;
4482
4483         /*
4484          * ctl->data points to idev->cnf.disable_ipv6, we should
4485          * not modify it until we get the rtnl lock.
4486          */
4487         lctl = *ctl;
4488         lctl.data = &val;
4489
4490         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4491
4492         if (write)
4493                 ret = addrconf_disable_ipv6(ctl, valp, val);
4494         if (ret)
4495                 *ppos = pos;
4496         return ret;
4497 }
4498
4499 static struct addrconf_sysctl_table
4500 {
4501         struct ctl_table_header *sysctl_header;
4502         ctl_table addrconf_vars[DEVCONF_MAX+1];
4503 } addrconf_sysctl __read_mostly = {
4504         .sysctl_header = NULL,
4505         .addrconf_vars = {
4506                 {
4507                         .procname       = "forwarding",
4508                         .data           = &ipv6_devconf.forwarding,
4509                         .maxlen         = sizeof(int),
4510                         .mode           = 0644,
4511                         .proc_handler   = addrconf_sysctl_forward,
4512                 },
4513                 {
4514                         .procname       = "hop_limit",
4515                         .data           = &ipv6_devconf.hop_limit,
4516                         .maxlen         = sizeof(int),
4517                         .mode           = 0644,
4518                         .proc_handler   = proc_dointvec,
4519                 },
4520                 {
4521                         .procname       = "mtu",
4522                         .data           = &ipv6_devconf.mtu6,
4523                         .maxlen         = sizeof(int),
4524                         .mode           = 0644,
4525                         .proc_handler   = proc_dointvec,
4526                 },
4527                 {
4528                         .procname       = "accept_ra",
4529                         .data           = &ipv6_devconf.accept_ra,
4530                         .maxlen         = sizeof(int),
4531                         .mode           = 0644,
4532                         .proc_handler   = proc_dointvec,
4533                 },
4534                 {
4535                         .procname       = "accept_redirects",
4536                         .data           = &ipv6_devconf.accept_redirects,
4537                         .maxlen         = sizeof(int),
4538                         .mode           = 0644,
4539                         .proc_handler   = proc_dointvec,
4540                 },
4541                 {
4542                         .procname       = "autoconf",
4543                         .data           = &ipv6_devconf.autoconf,
4544                         .maxlen         = sizeof(int),
4545                         .mode           = 0644,
4546                         .proc_handler   = proc_dointvec,
4547                 },
4548                 {
4549                         .procname       = "dad_transmits",
4550                         .data           = &ipv6_devconf.dad_transmits,
4551                         .maxlen         = sizeof(int),
4552                         .mode           = 0644,
4553                         .proc_handler   = proc_dointvec,
4554                 },
4555                 {
4556                         .procname       = "router_solicitations",
4557                         .data           = &ipv6_devconf.rtr_solicits,
4558                         .maxlen         = sizeof(int),
4559                         .mode           = 0644,
4560                         .proc_handler   = proc_dointvec,
4561                 },
4562                 {
4563                         .procname       = "router_solicitation_interval",
4564                         .data           = &ipv6_devconf.rtr_solicit_interval,
4565                         .maxlen         = sizeof(int),
4566                         .mode           = 0644,
4567                         .proc_handler   = proc_dointvec_jiffies,
4568                 },
4569                 {
4570                         .procname       = "router_solicitation_delay",
4571                         .data           = &ipv6_devconf.rtr_solicit_delay,
4572                         .maxlen         = sizeof(int),
4573                         .mode           = 0644,
4574                         .proc_handler   = proc_dointvec_jiffies,
4575                 },
4576                 {
4577                         .procname       = "force_mld_version",
4578                         .data           = &ipv6_devconf.force_mld_version,
4579                         .maxlen         = sizeof(int),
4580                         .mode           = 0644,
4581                         .proc_handler   = proc_dointvec,
4582                 },
4583 #ifdef CONFIG_IPV6_PRIVACY
4584                 {
4585                         .procname       = "use_tempaddr",
4586                         .data           = &ipv6_devconf.use_tempaddr,
4587                         .maxlen         = sizeof(int),
4588                         .mode           = 0644,
4589                         .proc_handler   = proc_dointvec,
4590                 },
4591                 {
4592                         .procname       = "temp_valid_lft",
4593                         .data           = &ipv6_devconf.temp_valid_lft,
4594                         .maxlen         = sizeof(int),
4595                         .mode           = 0644,
4596                         .proc_handler   = proc_dointvec,
4597                 },
4598                 {
4599                         .procname       = "temp_prefered_lft",
4600                         .data           = &ipv6_devconf.temp_prefered_lft,
4601                         .maxlen         = sizeof(int),
4602                         .mode           = 0644,
4603                         .proc_handler   = proc_dointvec,
4604                 },
4605                 {
4606                         .procname       = "regen_max_retry",
4607                         .data           = &ipv6_devconf.regen_max_retry,
4608                         .maxlen         = sizeof(int),
4609                         .mode           = 0644,
4610                         .proc_handler   = proc_dointvec,
4611                 },
4612                 {
4613                         .procname       = "max_desync_factor",
4614                         .data           = &ipv6_devconf.max_desync_factor,
4615                         .maxlen         = sizeof(int),
4616                         .mode           = 0644,
4617                         .proc_handler   = proc_dointvec,
4618                 },
4619 #endif
4620                 {
4621                         .procname       = "max_addresses",
4622                         .data           = &ipv6_devconf.max_addresses,
4623                         .maxlen         = sizeof(int),
4624                         .mode           = 0644,
4625                         .proc_handler   = proc_dointvec,
4626                 },
4627                 {
4628                         .procname       = "accept_ra_defrtr",
4629                         .data           = &ipv6_devconf.accept_ra_defrtr,
4630                         .maxlen         = sizeof(int),
4631                         .mode           = 0644,
4632                         .proc_handler   = proc_dointvec,
4633                 },
4634                 {
4635                         .procname       = "accept_ra_pinfo",
4636                         .data           = &ipv6_devconf.accept_ra_pinfo,
4637                         .maxlen         = sizeof(int),
4638                         .mode           = 0644,
4639                         .proc_handler   = proc_dointvec,
4640                 },
4641 #ifdef CONFIG_IPV6_ROUTER_PREF
4642                 {
4643                         .procname       = "accept_ra_rtr_pref",
4644                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4645                         .maxlen         = sizeof(int),
4646                         .mode           = 0644,
4647                         .proc_handler   = proc_dointvec,
4648                 },
4649                 {
4650                         .procname       = "router_probe_interval",
4651                         .data           = &ipv6_devconf.rtr_probe_interval,
4652                         .maxlen         = sizeof(int),
4653                         .mode           = 0644,
4654                         .proc_handler   = proc_dointvec_jiffies,
4655                 },
4656 #ifdef CONFIG_IPV6_ROUTE_INFO
4657                 {
4658                         .procname       = "accept_ra_rt_info_max_plen",
4659                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4660                         .maxlen         = sizeof(int),
4661                         .mode           = 0644,
4662                         .proc_handler   = proc_dointvec,
4663                 },
4664 #endif
4665 #endif
4666                 {
4667                         .procname       = "proxy_ndp",
4668                         .data           = &ipv6_devconf.proxy_ndp,
4669                         .maxlen         = sizeof(int),
4670                         .mode           = 0644,
4671                         .proc_handler   = proc_dointvec,
4672                 },
4673                 {
4674                         .procname       = "accept_source_route",
4675                         .data           = &ipv6_devconf.accept_source_route,
4676                         .maxlen         = sizeof(int),
4677                         .mode           = 0644,
4678                         .proc_handler   = proc_dointvec,
4679                 },
4680 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4681                 {
4682                         .procname       = "optimistic_dad",
4683                         .data           = &ipv6_devconf.optimistic_dad,
4684                         .maxlen         = sizeof(int),
4685                         .mode           = 0644,
4686                         .proc_handler   = proc_dointvec,
4687
4688                 },
4689 #endif
4690 #ifdef CONFIG_IPV6_MROUTE
4691                 {
4692                         .procname       = "mc_forwarding",
4693                         .data           = &ipv6_devconf.mc_forwarding,
4694                         .maxlen         = sizeof(int),
4695                         .mode           = 0444,
4696                         .proc_handler   = proc_dointvec,
4697                 },
4698 #endif
4699                 {
4700                         .procname       = "disable_ipv6",
4701                         .data           = &ipv6_devconf.disable_ipv6,
4702                         .maxlen         = sizeof(int),
4703                         .mode           = 0644,
4704                         .proc_handler   = addrconf_sysctl_disable,
4705                 },
4706                 {
4707                         .procname       = "accept_dad",
4708                         .data           = &ipv6_devconf.accept_dad,
4709                         .maxlen         = sizeof(int),
4710                         .mode           = 0644,
4711                         .proc_handler   = proc_dointvec,
4712                 },
4713                 {
4714                         .procname       = "force_tllao",
4715                         .data           = &ipv6_devconf.force_tllao,
4716                         .maxlen         = sizeof(int),
4717                         .mode           = 0644,
4718                         .proc_handler   = proc_dointvec
4719                 },
4720                 {
4721                         .procname       = "ndisc_notify",
4722                         .data           = &ipv6_devconf.ndisc_notify,
4723                         .maxlen         = sizeof(int),
4724                         .mode           = 0644,
4725                         .proc_handler   = proc_dointvec
4726                 },
4727                 {
4728                         /* sentinel */
4729                 }
4730         },
4731 };
4732
4733 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4734                 struct inet6_dev *idev, struct ipv6_devconf *p)
4735 {
4736         int i;
4737         struct addrconf_sysctl_table *t;
4738         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4739
4740         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4741         if (t == NULL)
4742                 goto out;
4743
4744         for (i = 0; t->addrconf_vars[i].data; i++) {
4745                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4746                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4747                 t->addrconf_vars[i].extra2 = net;
4748         }
4749
4750         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4751
4752         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4753         if (t->sysctl_header == NULL)
4754                 goto free;
4755
4756         p->sysctl = t;
4757         return 0;
4758
4759 free:
4760         kfree(t);
4761 out:
4762         return -ENOBUFS;
4763 }
4764
4765 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4766 {
4767         struct addrconf_sysctl_table *t;
4768
4769         if (p->sysctl == NULL)
4770                 return;
4771
4772         t = p->sysctl;
4773         p->sysctl = NULL;
4774         unregister_net_sysctl_table(t->sysctl_header);
4775         kfree(t);
4776 }
4777
4778 static void addrconf_sysctl_register(struct inet6_dev *idev)
4779 {
4780         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4781                               &ndisc_ifinfo_sysctl_change);
4782         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4783                                         idev, &idev->cnf);
4784 }
4785
4786 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4787 {
4788         __addrconf_sysctl_unregister(&idev->cnf);
4789         neigh_sysctl_unregister(idev->nd_parms);
4790 }
4791
4792
4793 #endif
4794
4795 static int __net_init addrconf_init_net(struct net *net)
4796 {
4797         int err;
4798         struct ipv6_devconf *all, *dflt;
4799
4800         err = -ENOMEM;
4801         all = &ipv6_devconf;
4802         dflt = &ipv6_devconf_dflt;
4803
4804         if (!net_eq(net, &init_net)) {
4805                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4806                 if (all == NULL)
4807                         goto err_alloc_all;
4808
4809                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4810                 if (dflt == NULL)
4811                         goto err_alloc_dflt;
4812         } else {
4813                 /* these will be inherited by all namespaces */
4814                 dflt->autoconf = ipv6_defaults.autoconf;
4815                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4816         }
4817
4818         net->ipv6.devconf_all = all;
4819         net->ipv6.devconf_dflt = dflt;
4820
4821 #ifdef CONFIG_SYSCTL
4822         err = __addrconf_sysctl_register(net, "all", NULL, all);
4823         if (err < 0)
4824                 goto err_reg_all;
4825
4826         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4827         if (err < 0)
4828                 goto err_reg_dflt;
4829 #endif
4830         return 0;
4831
4832 #ifdef CONFIG_SYSCTL
4833 err_reg_dflt:
4834         __addrconf_sysctl_unregister(all);
4835 err_reg_all:
4836         kfree(dflt);
4837 #endif
4838 err_alloc_dflt:
4839         kfree(all);
4840 err_alloc_all:
4841         return err;
4842 }
4843
4844 static void __net_exit addrconf_exit_net(struct net *net)
4845 {
4846 #ifdef CONFIG_SYSCTL
4847         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4848         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4849 #endif
4850         if (!net_eq(net, &init_net)) {
4851                 kfree(net->ipv6.devconf_dflt);
4852                 kfree(net->ipv6.devconf_all);
4853         }
4854 }
4855
4856 static struct pernet_operations addrconf_ops = {
4857         .init = addrconf_init_net,
4858         .exit = addrconf_exit_net,
4859 };
4860
4861 /*
4862  *      Device notifier
4863  */
4864
4865 int register_inet6addr_notifier(struct notifier_block *nb)
4866 {
4867         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4868 }
4869 EXPORT_SYMBOL(register_inet6addr_notifier);
4870
4871 int unregister_inet6addr_notifier(struct notifier_block *nb)
4872 {
4873         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4874 }
4875 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4876
4877 static struct rtnl_af_ops inet6_ops = {
4878         .family           = AF_INET6,
4879         .fill_link_af     = inet6_fill_link_af,
4880         .get_link_af_size = inet6_get_link_af_size,
4881 };
4882
4883 /*
4884  *      Init / cleanup code
4885  */
4886
4887 int __init addrconf_init(void)
4888 {
4889         int i, err;
4890
4891         err = ipv6_addr_label_init();
4892         if (err < 0) {
4893                 pr_crit("%s: cannot initialize default policy table: %d\n",
4894                         __func__, err);
4895                 goto out;
4896         }
4897
4898         err = register_pernet_subsys(&addrconf_ops);
4899         if (err < 0)
4900                 goto out_addrlabel;
4901
4902         /* The addrconf netdev notifier requires that loopback_dev
4903          * has it's ipv6 private information allocated and setup
4904          * before it can bring up and give link-local addresses
4905          * to other devices which are up.
4906          *
4907          * Unfortunately, loopback_dev is not necessarily the first
4908          * entry in the global dev_base list of net devices.  In fact,
4909          * it is likely to be the very last entry on that list.
4910          * So this causes the notifier registry below to try and
4911          * give link-local addresses to all devices besides loopback_dev
4912          * first, then loopback_dev, which cases all the non-loopback_dev
4913          * devices to fail to get a link-local address.
4914          *
4915          * So, as a temporary fix, allocate the ipv6 structure for
4916          * loopback_dev first by hand.
4917          * Longer term, all of the dependencies ipv6 has upon the loopback
4918          * device and it being up should be removed.
4919          */
4920         rtnl_lock();
4921         if (!ipv6_add_dev(init_net.loopback_dev))
4922                 err = -ENOMEM;
4923         rtnl_unlock();
4924         if (err)
4925                 goto errlo;
4926
4927         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4928                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4929
4930         register_netdevice_notifier(&ipv6_dev_notf);
4931
4932         addrconf_verify(0);
4933
4934         err = rtnl_af_register(&inet6_ops);
4935         if (err < 0)
4936                 goto errout_af;
4937
4938         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4939                               NULL);
4940         if (err < 0)
4941                 goto errout;
4942
4943         /* Only the first call to __rtnl_register can fail */
4944         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4945         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4946         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4947                         inet6_dump_ifaddr, NULL);
4948         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4949                         inet6_dump_ifmcaddr, NULL);
4950         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4951                         inet6_dump_ifacaddr, NULL);
4952         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4953                         NULL, NULL);
4954
4955         ipv6_addr_label_rtnl_register();
4956
4957         return 0;
4958 errout:
4959         rtnl_af_unregister(&inet6_ops);
4960 errout_af:
4961         unregister_netdevice_notifier(&ipv6_dev_notf);
4962 errlo:
4963         unregister_pernet_subsys(&addrconf_ops);
4964 out_addrlabel:
4965         ipv6_addr_label_cleanup();
4966 out:
4967         return err;
4968 }
4969
4970 void addrconf_cleanup(void)
4971 {
4972         struct net_device *dev;
4973         int i;
4974
4975         unregister_netdevice_notifier(&ipv6_dev_notf);
4976         unregister_pernet_subsys(&addrconf_ops);
4977         ipv6_addr_label_cleanup();
4978
4979         rtnl_lock();
4980
4981         __rtnl_af_unregister(&inet6_ops);
4982
4983         /* clean dev list */
4984         for_each_netdev(&init_net, dev) {
4985                 if (__in6_dev_get(dev) == NULL)
4986                         continue;
4987                 addrconf_ifdown(dev, 1);
4988         }
4989         addrconf_ifdown(init_net.loopback_dev, 2);
4990
4991         /*
4992          *      Check hash table.
4993          */
4994         spin_lock_bh(&addrconf_hash_lock);
4995         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4996                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4997         spin_unlock_bh(&addrconf_hash_lock);
4998
4999         del_timer(&addr_chk_timer);
5000         rtnl_unlock();
5001 }