Merge tag 'sunxi-fixes-for-4.3' of https://git.kernel.org/pub/scm/linux/kernel/git...
[linux-drm-fsl-dcu.git] / net / ipv6 / addrconf.c
1 /*
2  *      IPv6 Address [auto]configuration
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
9  *      This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  */
14
15 /*
16  *      Changes:
17  *
18  *      Janos Farkas                    :       delete timer on ifdown
19  *      <chexum@bankinf.banki.hu>
20  *      Andi Kleen                      :       kill double kfree on module
21  *                                              unload.
22  *      Maciej W. Rozycki               :       FDDI support
23  *      sekiya@USAGI                    :       Don't send too many RS
24  *                                              packets.
25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
26  *                                              packets.
27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
28  *                                              address validation timer.
29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
30  *                                              support.
31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
32  *                                              address on a same interface.
33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
35  *                                              seq_file.
36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
37  *                                              selection; consider scope,
38  *                                              status etc.
39  */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
68
69 #include <net/net_namespace.h>
70 #include <net/sock.h>
71 #include <net/snmp.h>
72
73 #include <net/af_ieee802154.h>
74 #include <net/firewire.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/tcp.h>
81 #include <net/ip.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <linux/if_tunnel.h>
85 #include <linux/rtnetlink.h>
86 #include <linux/netconf.h>
87 #include <linux/random.h>
88 #include <linux/uaccess.h>
89 #include <asm/unaligned.h>
90
91 #include <linux/proc_fs.h>
92 #include <linux/seq_file.h>
93 #include <linux/export.h>
94
95 /* Set to 3 to get tracing... */
96 #define ACONF_DEBUG 2
97
98 #if ACONF_DEBUG >= 3
99 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
100 #else
101 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
102 #endif
103
104 #define INFINITY_LIFE_TIME      0xFFFFFFFF
105
106 #define IPV6_MAX_STRLEN \
107         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
108
109 static inline u32 cstamp_delta(unsigned long cstamp)
110 {
111         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
112 }
113
114 #ifdef CONFIG_SYSCTL
115 static int addrconf_sysctl_register(struct inet6_dev *idev);
116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
117 #else
118 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
119 {
120         return 0;
121 }
122
123 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
124 {
125 }
126 #endif
127
128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
130 static void ipv6_regen_rndid(unsigned long data);
131
132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
133 static int ipv6_count_addresses(struct inet6_dev *idev);
134 static int ipv6_generate_stable_address(struct in6_addr *addr,
135                                         u8 dad_count,
136                                         const struct inet6_dev *idev);
137
138 /*
139  *      Configured unicast address hash table
140  */
141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
142 static DEFINE_SPINLOCK(addrconf_hash_lock);
143
144 static void addrconf_verify(void);
145 static void addrconf_verify_rtnl(void);
146 static void addrconf_verify_work(struct work_struct *);
147
148 static struct workqueue_struct *addrconf_wq;
149 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
150
151 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
152 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
153
154 static void addrconf_type_change(struct net_device *dev,
155                                  unsigned long event);
156 static int addrconf_ifdown(struct net_device *dev, int how);
157
158 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
159                                                   int plen,
160                                                   const struct net_device *dev,
161                                                   u32 flags, u32 noflags);
162
163 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
164 static void addrconf_dad_work(struct work_struct *w);
165 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
166 static void addrconf_dad_run(struct inet6_dev *idev);
167 static void addrconf_rs_timer(unsigned long data);
168 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
169 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170
171 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
172                                 struct prefix_info *pinfo);
173 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
174                                struct net_device *dev);
175
176 static struct ipv6_devconf ipv6_devconf __read_mostly = {
177         .forwarding             = 0,
178         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
179         .mtu6                   = IPV6_MIN_MTU,
180         .accept_ra              = 1,
181         .accept_redirects       = 1,
182         .autoconf               = 1,
183         .force_mld_version      = 0,
184         .mldv1_unsolicited_report_interval = 10 * HZ,
185         .mldv2_unsolicited_report_interval = HZ,
186         .dad_transmits          = 1,
187         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
188         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
189         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
190         .use_tempaddr           = 0,
191         .temp_valid_lft         = TEMP_VALID_LIFETIME,
192         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
193         .regen_max_retry        = REGEN_MAX_RETRY,
194         .max_desync_factor      = MAX_DESYNC_FACTOR,
195         .max_addresses          = IPV6_MAX_ADDRESSES,
196         .accept_ra_defrtr       = 1,
197         .accept_ra_from_local   = 0,
198         .accept_ra_min_hop_limit= 1,
199         .accept_ra_pinfo        = 1,
200 #ifdef CONFIG_IPV6_ROUTER_PREF
201         .accept_ra_rtr_pref     = 1,
202         .rtr_probe_interval     = 60 * HZ,
203 #ifdef CONFIG_IPV6_ROUTE_INFO
204         .accept_ra_rt_info_max_plen = 0,
205 #endif
206 #endif
207         .proxy_ndp              = 0,
208         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
209         .disable_ipv6           = 0,
210         .accept_dad             = 1,
211         .suppress_frag_ndisc    = 1,
212         .accept_ra_mtu          = 1,
213         .stable_secret          = {
214                 .initialized = false,
215         },
216         .use_oif_addrs_only     = 0,
217         .ignore_routes_with_linkdown = 0,
218 };
219
220 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
221         .forwarding             = 0,
222         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
223         .mtu6                   = IPV6_MIN_MTU,
224         .accept_ra              = 1,
225         .accept_redirects       = 1,
226         .autoconf               = 1,
227         .force_mld_version      = 0,
228         .mldv1_unsolicited_report_interval = 10 * HZ,
229         .mldv2_unsolicited_report_interval = HZ,
230         .dad_transmits          = 1,
231         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
232         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
233         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
234         .use_tempaddr           = 0,
235         .temp_valid_lft         = TEMP_VALID_LIFETIME,
236         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
237         .regen_max_retry        = REGEN_MAX_RETRY,
238         .max_desync_factor      = MAX_DESYNC_FACTOR,
239         .max_addresses          = IPV6_MAX_ADDRESSES,
240         .accept_ra_defrtr       = 1,
241         .accept_ra_from_local   = 0,
242         .accept_ra_min_hop_limit= 1,
243         .accept_ra_pinfo        = 1,
244 #ifdef CONFIG_IPV6_ROUTER_PREF
245         .accept_ra_rtr_pref     = 1,
246         .rtr_probe_interval     = 60 * HZ,
247 #ifdef CONFIG_IPV6_ROUTE_INFO
248         .accept_ra_rt_info_max_plen = 0,
249 #endif
250 #endif
251         .proxy_ndp              = 0,
252         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
253         .disable_ipv6           = 0,
254         .accept_dad             = 1,
255         .suppress_frag_ndisc    = 1,
256         .accept_ra_mtu          = 1,
257         .stable_secret          = {
258                 .initialized = false,
259         },
260         .use_oif_addrs_only     = 0,
261         .ignore_routes_with_linkdown = 0,
262 };
263
264 /* Check if a valid qdisc is available */
265 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
266 {
267         return !qdisc_tx_is_noop(dev);
268 }
269
270 static void addrconf_del_rs_timer(struct inet6_dev *idev)
271 {
272         if (del_timer(&idev->rs_timer))
273                 __in6_dev_put(idev);
274 }
275
276 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
277 {
278         if (cancel_delayed_work(&ifp->dad_work))
279                 __in6_ifa_put(ifp);
280 }
281
282 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
283                                   unsigned long when)
284 {
285         if (!timer_pending(&idev->rs_timer))
286                 in6_dev_hold(idev);
287         mod_timer(&idev->rs_timer, jiffies + when);
288 }
289
290 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
291                                    unsigned long delay)
292 {
293         if (!delayed_work_pending(&ifp->dad_work))
294                 in6_ifa_hold(ifp);
295         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
296 }
297
298 static int snmp6_alloc_dev(struct inet6_dev *idev)
299 {
300         int i;
301
302         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
303         if (!idev->stats.ipv6)
304                 goto err_ip;
305
306         for_each_possible_cpu(i) {
307                 struct ipstats_mib *addrconf_stats;
308                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
309                 u64_stats_init(&addrconf_stats->syncp);
310         }
311
312
313         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
314                                         GFP_KERNEL);
315         if (!idev->stats.icmpv6dev)
316                 goto err_icmp;
317         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
318                                            GFP_KERNEL);
319         if (!idev->stats.icmpv6msgdev)
320                 goto err_icmpmsg;
321
322         return 0;
323
324 err_icmpmsg:
325         kfree(idev->stats.icmpv6dev);
326 err_icmp:
327         free_percpu(idev->stats.ipv6);
328 err_ip:
329         return -ENOMEM;
330 }
331
332 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
333 {
334         struct inet6_dev *ndev;
335         int err = -ENOMEM;
336
337         ASSERT_RTNL();
338
339         if (dev->mtu < IPV6_MIN_MTU)
340                 return ERR_PTR(-EINVAL);
341
342         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
343         if (!ndev)
344                 return ERR_PTR(err);
345
346         rwlock_init(&ndev->lock);
347         ndev->dev = dev;
348         INIT_LIST_HEAD(&ndev->addr_list);
349         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
350                     (unsigned long)ndev);
351         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
352         ndev->cnf.mtu6 = dev->mtu;
353         ndev->cnf.sysctl = NULL;
354         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
355         if (!ndev->nd_parms) {
356                 kfree(ndev);
357                 return ERR_PTR(err);
358         }
359         if (ndev->cnf.forwarding)
360                 dev_disable_lro(dev);
361         /* We refer to the device */
362         dev_hold(dev);
363
364         if (snmp6_alloc_dev(ndev) < 0) {
365                 ADBG(KERN_WARNING
366                         "%s: cannot allocate memory for statistics; dev=%s.\n",
367                         __func__, dev->name);
368                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
369                 dev_put(dev);
370                 kfree(ndev);
371                 return ERR_PTR(err);
372         }
373
374         if (snmp6_register_dev(ndev) < 0) {
375                 ADBG(KERN_WARNING
376                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
377                         __func__, dev->name);
378                 goto err_release;
379         }
380
381         /* One reference from device.  We must do this before
382          * we invoke __ipv6_regen_rndid().
383          */
384         in6_dev_hold(ndev);
385
386         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
387                 ndev->cnf.accept_dad = -1;
388
389 #if IS_ENABLED(CONFIG_IPV6_SIT)
390         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
391                 pr_info("%s: Disabled Multicast RS\n", dev->name);
392                 ndev->cnf.rtr_solicits = 0;
393         }
394 #endif
395
396         INIT_LIST_HEAD(&ndev->tempaddr_list);
397         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
398         if ((dev->flags&IFF_LOOPBACK) ||
399             dev->type == ARPHRD_TUNNEL ||
400             dev->type == ARPHRD_TUNNEL6 ||
401             dev->type == ARPHRD_SIT ||
402             dev->type == ARPHRD_NONE) {
403                 ndev->cnf.use_tempaddr = -1;
404         } else {
405                 in6_dev_hold(ndev);
406                 ipv6_regen_rndid((unsigned long) ndev);
407         }
408
409         ndev->token = in6addr_any;
410
411         if (netif_running(dev) && addrconf_qdisc_ok(dev))
412                 ndev->if_flags |= IF_READY;
413
414         ipv6_mc_init_dev(ndev);
415         ndev->tstamp = jiffies;
416         err = addrconf_sysctl_register(ndev);
417         if (err) {
418                 ipv6_mc_destroy_dev(ndev);
419                 del_timer(&ndev->regen_timer);
420                 goto err_release;
421         }
422         /* protected by rtnl_lock */
423         rcu_assign_pointer(dev->ip6_ptr, ndev);
424
425         /* Join interface-local all-node multicast group */
426         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
427
428         /* Join all-node multicast group */
429         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
430
431         /* Join all-router multicast group if forwarding is set */
432         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
433                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
434
435         return ndev;
436
437 err_release:
438         neigh_parms_release(&nd_tbl, ndev->nd_parms);
439         ndev->dead = 1;
440         in6_dev_finish_destroy(ndev);
441         return ERR_PTR(err);
442 }
443
444 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
445 {
446         struct inet6_dev *idev;
447
448         ASSERT_RTNL();
449
450         idev = __in6_dev_get(dev);
451         if (!idev) {
452                 idev = ipv6_add_dev(dev);
453                 if (IS_ERR(idev))
454                         return NULL;
455         }
456
457         if (dev->flags&IFF_UP)
458                 ipv6_mc_up(idev);
459         return idev;
460 }
461
462 static int inet6_netconf_msgsize_devconf(int type)
463 {
464         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
465                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
466
467         /* type -1 is used for ALL */
468         if (type == -1 || type == NETCONFA_FORWARDING)
469                 size += nla_total_size(4);
470 #ifdef CONFIG_IPV6_MROUTE
471         if (type == -1 || type == NETCONFA_MC_FORWARDING)
472                 size += nla_total_size(4);
473 #endif
474         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
475                 size += nla_total_size(4);
476
477         if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
478                 size += nla_total_size(4);
479
480         return size;
481 }
482
483 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
484                                       struct ipv6_devconf *devconf, u32 portid,
485                                       u32 seq, int event, unsigned int flags,
486                                       int type)
487 {
488         struct nlmsghdr  *nlh;
489         struct netconfmsg *ncm;
490
491         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
492                         flags);
493         if (!nlh)
494                 return -EMSGSIZE;
495
496         ncm = nlmsg_data(nlh);
497         ncm->ncm_family = AF_INET6;
498
499         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
500                 goto nla_put_failure;
501
502         /* type -1 is used for ALL */
503         if ((type == -1 || type == NETCONFA_FORWARDING) &&
504             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
505                 goto nla_put_failure;
506 #ifdef CONFIG_IPV6_MROUTE
507         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
508             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
509                         devconf->mc_forwarding) < 0)
510                 goto nla_put_failure;
511 #endif
512         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
513             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
514                 goto nla_put_failure;
515
516         if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
517             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
518                         devconf->ignore_routes_with_linkdown) < 0)
519                 goto nla_put_failure;
520
521         nlmsg_end(skb, nlh);
522         return 0;
523
524 nla_put_failure:
525         nlmsg_cancel(skb, nlh);
526         return -EMSGSIZE;
527 }
528
529 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
530                                   struct ipv6_devconf *devconf)
531 {
532         struct sk_buff *skb;
533         int err = -ENOBUFS;
534
535         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
536         if (!skb)
537                 goto errout;
538
539         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
540                                          RTM_NEWNETCONF, 0, type);
541         if (err < 0) {
542                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
543                 WARN_ON(err == -EMSGSIZE);
544                 kfree_skb(skb);
545                 goto errout;
546         }
547         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
548         return;
549 errout:
550         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
551 }
552
553 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
554         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
555         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
556         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
557         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
558 };
559
560 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
561                                      struct nlmsghdr *nlh)
562 {
563         struct net *net = sock_net(in_skb->sk);
564         struct nlattr *tb[NETCONFA_MAX+1];
565         struct netconfmsg *ncm;
566         struct sk_buff *skb;
567         struct ipv6_devconf *devconf;
568         struct inet6_dev *in6_dev;
569         struct net_device *dev;
570         int ifindex;
571         int err;
572
573         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
574                           devconf_ipv6_policy);
575         if (err < 0)
576                 goto errout;
577
578         err = EINVAL;
579         if (!tb[NETCONFA_IFINDEX])
580                 goto errout;
581
582         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
583         switch (ifindex) {
584         case NETCONFA_IFINDEX_ALL:
585                 devconf = net->ipv6.devconf_all;
586                 break;
587         case NETCONFA_IFINDEX_DEFAULT:
588                 devconf = net->ipv6.devconf_dflt;
589                 break;
590         default:
591                 dev = __dev_get_by_index(net, ifindex);
592                 if (!dev)
593                         goto errout;
594                 in6_dev = __in6_dev_get(dev);
595                 if (!in6_dev)
596                         goto errout;
597                 devconf = &in6_dev->cnf;
598                 break;
599         }
600
601         err = -ENOBUFS;
602         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
603         if (!skb)
604                 goto errout;
605
606         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
607                                          NETLINK_CB(in_skb).portid,
608                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
609                                          -1);
610         if (err < 0) {
611                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
612                 WARN_ON(err == -EMSGSIZE);
613                 kfree_skb(skb);
614                 goto errout;
615         }
616         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
617 errout:
618         return err;
619 }
620
621 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
622                                       struct netlink_callback *cb)
623 {
624         struct net *net = sock_net(skb->sk);
625         int h, s_h;
626         int idx, s_idx;
627         struct net_device *dev;
628         struct inet6_dev *idev;
629         struct hlist_head *head;
630
631         s_h = cb->args[0];
632         s_idx = idx = cb->args[1];
633
634         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
635                 idx = 0;
636                 head = &net->dev_index_head[h];
637                 rcu_read_lock();
638                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
639                           net->dev_base_seq;
640                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
641                         if (idx < s_idx)
642                                 goto cont;
643                         idev = __in6_dev_get(dev);
644                         if (!idev)
645                                 goto cont;
646
647                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
648                                                        &idev->cnf,
649                                                        NETLINK_CB(cb->skb).portid,
650                                                        cb->nlh->nlmsg_seq,
651                                                        RTM_NEWNETCONF,
652                                                        NLM_F_MULTI,
653                                                        -1) < 0) {
654                                 rcu_read_unlock();
655                                 goto done;
656                         }
657                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
658 cont:
659                         idx++;
660                 }
661                 rcu_read_unlock();
662         }
663         if (h == NETDEV_HASHENTRIES) {
664                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
665                                                net->ipv6.devconf_all,
666                                                NETLINK_CB(cb->skb).portid,
667                                                cb->nlh->nlmsg_seq,
668                                                RTM_NEWNETCONF, NLM_F_MULTI,
669                                                -1) < 0)
670                         goto done;
671                 else
672                         h++;
673         }
674         if (h == NETDEV_HASHENTRIES + 1) {
675                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
676                                                net->ipv6.devconf_dflt,
677                                                NETLINK_CB(cb->skb).portid,
678                                                cb->nlh->nlmsg_seq,
679                                                RTM_NEWNETCONF, NLM_F_MULTI,
680                                                -1) < 0)
681                         goto done;
682                 else
683                         h++;
684         }
685 done:
686         cb->args[0] = h;
687         cb->args[1] = idx;
688
689         return skb->len;
690 }
691
692 #ifdef CONFIG_SYSCTL
693 static void dev_forward_change(struct inet6_dev *idev)
694 {
695         struct net_device *dev;
696         struct inet6_ifaddr *ifa;
697
698         if (!idev)
699                 return;
700         dev = idev->dev;
701         if (idev->cnf.forwarding)
702                 dev_disable_lro(dev);
703         if (dev->flags & IFF_MULTICAST) {
704                 if (idev->cnf.forwarding) {
705                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
706                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
707                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
708                 } else {
709                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
710                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
711                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
712                 }
713         }
714
715         list_for_each_entry(ifa, &idev->addr_list, if_list) {
716                 if (ifa->flags&IFA_F_TENTATIVE)
717                         continue;
718                 if (idev->cnf.forwarding)
719                         addrconf_join_anycast(ifa);
720                 else
721                         addrconf_leave_anycast(ifa);
722         }
723         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
724                                      dev->ifindex, &idev->cnf);
725 }
726
727
728 static void addrconf_forward_change(struct net *net, __s32 newf)
729 {
730         struct net_device *dev;
731         struct inet6_dev *idev;
732
733         for_each_netdev(net, dev) {
734                 idev = __in6_dev_get(dev);
735                 if (idev) {
736                         int changed = (!idev->cnf.forwarding) ^ (!newf);
737                         idev->cnf.forwarding = newf;
738                         if (changed)
739                                 dev_forward_change(idev);
740                 }
741         }
742 }
743
744 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
745 {
746         struct net *net;
747         int old;
748
749         if (!rtnl_trylock())
750                 return restart_syscall();
751
752         net = (struct net *)table->extra2;
753         old = *p;
754         *p = newf;
755
756         if (p == &net->ipv6.devconf_dflt->forwarding) {
757                 if ((!newf) ^ (!old))
758                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
759                                                      NETCONFA_IFINDEX_DEFAULT,
760                                                      net->ipv6.devconf_dflt);
761                 rtnl_unlock();
762                 return 0;
763         }
764
765         if (p == &net->ipv6.devconf_all->forwarding) {
766                 net->ipv6.devconf_dflt->forwarding = newf;
767                 addrconf_forward_change(net, newf);
768                 if ((!newf) ^ (!old))
769                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
770                                                      NETCONFA_IFINDEX_ALL,
771                                                      net->ipv6.devconf_all);
772         } else if ((!newf) ^ (!old))
773                 dev_forward_change((struct inet6_dev *)table->extra1);
774         rtnl_unlock();
775
776         if (newf)
777                 rt6_purge_dflt_routers(net);
778         return 1;
779 }
780
781 static void addrconf_linkdown_change(struct net *net, __s32 newf)
782 {
783         struct net_device *dev;
784         struct inet6_dev *idev;
785
786         for_each_netdev(net, dev) {
787                 idev = __in6_dev_get(dev);
788                 if (idev) {
789                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
790
791                         idev->cnf.ignore_routes_with_linkdown = newf;
792                         if (changed)
793                                 inet6_netconf_notify_devconf(dev_net(dev),
794                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
795                                                              dev->ifindex,
796                                                              &idev->cnf);
797                 }
798         }
799 }
800
801 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
802 {
803         struct net *net;
804         int old;
805
806         if (!rtnl_trylock())
807                 return restart_syscall();
808
809         net = (struct net *)table->extra2;
810         old = *p;
811         *p = newf;
812
813         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
814                 if ((!newf) ^ (!old))
815                         inet6_netconf_notify_devconf(net,
816                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
817                                                      NETCONFA_IFINDEX_DEFAULT,
818                                                      net->ipv6.devconf_dflt);
819                 rtnl_unlock();
820                 return 0;
821         }
822
823         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
824                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
825                 addrconf_linkdown_change(net, newf);
826                 if ((!newf) ^ (!old))
827                         inet6_netconf_notify_devconf(net,
828                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
829                                                      NETCONFA_IFINDEX_ALL,
830                                                      net->ipv6.devconf_all);
831         }
832         rtnl_unlock();
833
834         return 1;
835 }
836
837 #endif
838
839 /* Nobody refers to this ifaddr, destroy it */
840 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
841 {
842         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
843
844 #ifdef NET_REFCNT_DEBUG
845         pr_debug("%s\n", __func__);
846 #endif
847
848         in6_dev_put(ifp->idev);
849
850         if (cancel_delayed_work(&ifp->dad_work))
851                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
852                           ifp);
853
854         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
855                 pr_warn("Freeing alive inet6 address %p\n", ifp);
856                 return;
857         }
858         ip6_rt_put(ifp->rt);
859
860         kfree_rcu(ifp, rcu);
861 }
862
863 static void
864 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
865 {
866         struct list_head *p;
867         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
868
869         /*
870          * Each device address list is sorted in order of scope -
871          * global before linklocal.
872          */
873         list_for_each(p, &idev->addr_list) {
874                 struct inet6_ifaddr *ifa
875                         = list_entry(p, struct inet6_ifaddr, if_list);
876                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
877                         break;
878         }
879
880         list_add_tail(&ifp->if_list, p);
881 }
882
883 static u32 inet6_addr_hash(const struct in6_addr *addr)
884 {
885         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
886 }
887
888 /* On success it returns ifp with increased reference count */
889
890 static struct inet6_ifaddr *
891 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
892               const struct in6_addr *peer_addr, int pfxlen,
893               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
894 {
895         struct inet6_ifaddr *ifa = NULL;
896         struct rt6_info *rt;
897         unsigned int hash;
898         int err = 0;
899         int addr_type = ipv6_addr_type(addr);
900
901         if (addr_type == IPV6_ADDR_ANY ||
902             addr_type & IPV6_ADDR_MULTICAST ||
903             (!(idev->dev->flags & IFF_LOOPBACK) &&
904              addr_type & IPV6_ADDR_LOOPBACK))
905                 return ERR_PTR(-EADDRNOTAVAIL);
906
907         rcu_read_lock_bh();
908         if (idev->dead) {
909                 err = -ENODEV;                  /*XXX*/
910                 goto out2;
911         }
912
913         if (idev->cnf.disable_ipv6) {
914                 err = -EACCES;
915                 goto out2;
916         }
917
918         spin_lock(&addrconf_hash_lock);
919
920         /* Ignore adding duplicate addresses on an interface */
921         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
922                 ADBG("ipv6_add_addr: already assigned\n");
923                 err = -EEXIST;
924                 goto out;
925         }
926
927         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
928
929         if (!ifa) {
930                 ADBG("ipv6_add_addr: malloc failed\n");
931                 err = -ENOBUFS;
932                 goto out;
933         }
934
935         rt = addrconf_dst_alloc(idev, addr, false);
936         if (IS_ERR(rt)) {
937                 err = PTR_ERR(rt);
938                 goto out;
939         }
940
941         neigh_parms_data_state_setall(idev->nd_parms);
942
943         ifa->addr = *addr;
944         if (peer_addr)
945                 ifa->peer_addr = *peer_addr;
946
947         spin_lock_init(&ifa->lock);
948         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
949         INIT_HLIST_NODE(&ifa->addr_lst);
950         ifa->scope = scope;
951         ifa->prefix_len = pfxlen;
952         ifa->flags = flags | IFA_F_TENTATIVE;
953         ifa->valid_lft = valid_lft;
954         ifa->prefered_lft = prefered_lft;
955         ifa->cstamp = ifa->tstamp = jiffies;
956         ifa->tokenized = false;
957
958         ifa->rt = rt;
959
960         ifa->idev = idev;
961         in6_dev_hold(idev);
962         /* For caller */
963         in6_ifa_hold(ifa);
964
965         /* Add to big hash table */
966         hash = inet6_addr_hash(addr);
967
968         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
969         spin_unlock(&addrconf_hash_lock);
970
971         write_lock(&idev->lock);
972         /* Add to inet6_dev unicast addr list. */
973         ipv6_link_dev_addr(idev, ifa);
974
975         if (ifa->flags&IFA_F_TEMPORARY) {
976                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
977                 in6_ifa_hold(ifa);
978         }
979
980         in6_ifa_hold(ifa);
981         write_unlock(&idev->lock);
982 out2:
983         rcu_read_unlock_bh();
984
985         if (likely(err == 0))
986                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
987         else {
988                 kfree(ifa);
989                 ifa = ERR_PTR(err);
990         }
991
992         return ifa;
993 out:
994         spin_unlock(&addrconf_hash_lock);
995         goto out2;
996 }
997
998 enum cleanup_prefix_rt_t {
999         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1000         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1001         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1002 };
1003
1004 /*
1005  * Check, whether the prefix for ifp would still need a prefix route
1006  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1007  * constants.
1008  *
1009  * 1) we don't purge prefix if address was not permanent.
1010  *    prefix is managed by its own lifetime.
1011  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1012  * 3) if there are no addresses, delete prefix.
1013  * 4) if there are still other permanent address(es),
1014  *    corresponding prefix is still permanent.
1015  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1016  *    don't purge the prefix, assume user space is managing it.
1017  * 6) otherwise, update prefix lifetime to the
1018  *    longest valid lifetime among the corresponding
1019  *    addresses on the device.
1020  *    Note: subsequent RA will update lifetime.
1021  **/
1022 static enum cleanup_prefix_rt_t
1023 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1024 {
1025         struct inet6_ifaddr *ifa;
1026         struct inet6_dev *idev = ifp->idev;
1027         unsigned long lifetime;
1028         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1029
1030         *expires = jiffies;
1031
1032         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1033                 if (ifa == ifp)
1034                         continue;
1035                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1036                                        ifp->prefix_len))
1037                         continue;
1038                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1039                         return CLEANUP_PREFIX_RT_NOP;
1040
1041                 action = CLEANUP_PREFIX_RT_EXPIRE;
1042
1043                 spin_lock(&ifa->lock);
1044
1045                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1046                 /*
1047                  * Note: Because this address is
1048                  * not permanent, lifetime <
1049                  * LONG_MAX / HZ here.
1050                  */
1051                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1052                         *expires = ifa->tstamp + lifetime * HZ;
1053                 spin_unlock(&ifa->lock);
1054         }
1055
1056         return action;
1057 }
1058
1059 static void
1060 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1061 {
1062         struct rt6_info *rt;
1063
1064         rt = addrconf_get_prefix_route(&ifp->addr,
1065                                        ifp->prefix_len,
1066                                        ifp->idev->dev,
1067                                        0, RTF_GATEWAY | RTF_DEFAULT);
1068         if (rt) {
1069                 if (del_rt)
1070                         ip6_del_rt(rt);
1071                 else {
1072                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1073                                 rt6_set_expires(rt, expires);
1074                         ip6_rt_put(rt);
1075                 }
1076         }
1077 }
1078
1079
1080 /* This function wants to get referenced ifp and releases it before return */
1081
1082 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1083 {
1084         int state;
1085         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1086         unsigned long expires;
1087
1088         ASSERT_RTNL();
1089
1090         spin_lock_bh(&ifp->lock);
1091         state = ifp->state;
1092         ifp->state = INET6_IFADDR_STATE_DEAD;
1093         spin_unlock_bh(&ifp->lock);
1094
1095         if (state == INET6_IFADDR_STATE_DEAD)
1096                 goto out;
1097
1098         spin_lock_bh(&addrconf_hash_lock);
1099         hlist_del_init_rcu(&ifp->addr_lst);
1100         spin_unlock_bh(&addrconf_hash_lock);
1101
1102         write_lock_bh(&ifp->idev->lock);
1103
1104         if (ifp->flags&IFA_F_TEMPORARY) {
1105                 list_del(&ifp->tmp_list);
1106                 if (ifp->ifpub) {
1107                         in6_ifa_put(ifp->ifpub);
1108                         ifp->ifpub = NULL;
1109                 }
1110                 __in6_ifa_put(ifp);
1111         }
1112
1113         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1114                 action = check_cleanup_prefix_route(ifp, &expires);
1115
1116         list_del_init(&ifp->if_list);
1117         __in6_ifa_put(ifp);
1118
1119         write_unlock_bh(&ifp->idev->lock);
1120
1121         addrconf_del_dad_work(ifp);
1122
1123         ipv6_ifa_notify(RTM_DELADDR, ifp);
1124
1125         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1126
1127         if (action != CLEANUP_PREFIX_RT_NOP) {
1128                 cleanup_prefix_route(ifp, expires,
1129                         action == CLEANUP_PREFIX_RT_DEL);
1130         }
1131
1132         /* clean up prefsrc entries */
1133         rt6_remove_prefsrc(ifp);
1134 out:
1135         in6_ifa_put(ifp);
1136 }
1137
1138 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1139 {
1140         struct inet6_dev *idev = ifp->idev;
1141         struct in6_addr addr, *tmpaddr;
1142         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1143         unsigned long regen_advance;
1144         int tmp_plen;
1145         int ret = 0;
1146         u32 addr_flags;
1147         unsigned long now = jiffies;
1148
1149         write_lock_bh(&idev->lock);
1150         if (ift) {
1151                 spin_lock_bh(&ift->lock);
1152                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1153                 spin_unlock_bh(&ift->lock);
1154                 tmpaddr = &addr;
1155         } else {
1156                 tmpaddr = NULL;
1157         }
1158 retry:
1159         in6_dev_hold(idev);
1160         if (idev->cnf.use_tempaddr <= 0) {
1161                 write_unlock_bh(&idev->lock);
1162                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1163                 in6_dev_put(idev);
1164                 ret = -1;
1165                 goto out;
1166         }
1167         spin_lock_bh(&ifp->lock);
1168         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1169                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1170                 spin_unlock_bh(&ifp->lock);
1171                 write_unlock_bh(&idev->lock);
1172                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1173                         __func__);
1174                 in6_dev_put(idev);
1175                 ret = -1;
1176                 goto out;
1177         }
1178         in6_ifa_hold(ifp);
1179         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1180         __ipv6_try_regen_rndid(idev, tmpaddr);
1181         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1182         age = (now - ifp->tstamp) / HZ;
1183         tmp_valid_lft = min_t(__u32,
1184                               ifp->valid_lft,
1185                               idev->cnf.temp_valid_lft + age);
1186         tmp_prefered_lft = min_t(__u32,
1187                                  ifp->prefered_lft,
1188                                  idev->cnf.temp_prefered_lft + age -
1189                                  idev->cnf.max_desync_factor);
1190         tmp_plen = ifp->prefix_len;
1191         tmp_tstamp = ifp->tstamp;
1192         spin_unlock_bh(&ifp->lock);
1193
1194         regen_advance = idev->cnf.regen_max_retry *
1195                         idev->cnf.dad_transmits *
1196                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1197         write_unlock_bh(&idev->lock);
1198
1199         /* A temporary address is created only if this calculated Preferred
1200          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1201          * an implementation must not create a temporary address with a zero
1202          * Preferred Lifetime.
1203          * Use age calculation as in addrconf_verify to avoid unnecessary
1204          * temporary addresses being generated.
1205          */
1206         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1207         if (tmp_prefered_lft <= regen_advance + age) {
1208                 in6_ifa_put(ifp);
1209                 in6_dev_put(idev);
1210                 ret = -1;
1211                 goto out;
1212         }
1213
1214         addr_flags = IFA_F_TEMPORARY;
1215         /* set in addrconf_prefix_rcv() */
1216         if (ifp->flags & IFA_F_OPTIMISTIC)
1217                 addr_flags |= IFA_F_OPTIMISTIC;
1218
1219         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1220                             ipv6_addr_scope(&addr), addr_flags,
1221                             tmp_valid_lft, tmp_prefered_lft);
1222         if (IS_ERR(ift)) {
1223                 in6_ifa_put(ifp);
1224                 in6_dev_put(idev);
1225                 pr_info("%s: retry temporary address regeneration\n", __func__);
1226                 tmpaddr = &addr;
1227                 write_lock_bh(&idev->lock);
1228                 goto retry;
1229         }
1230
1231         spin_lock_bh(&ift->lock);
1232         ift->ifpub = ifp;
1233         ift->cstamp = now;
1234         ift->tstamp = tmp_tstamp;
1235         spin_unlock_bh(&ift->lock);
1236
1237         addrconf_dad_start(ift);
1238         in6_ifa_put(ift);
1239         in6_dev_put(idev);
1240 out:
1241         return ret;
1242 }
1243
1244 /*
1245  *      Choose an appropriate source address (RFC3484)
1246  */
1247 enum {
1248         IPV6_SADDR_RULE_INIT = 0,
1249         IPV6_SADDR_RULE_LOCAL,
1250         IPV6_SADDR_RULE_SCOPE,
1251         IPV6_SADDR_RULE_PREFERRED,
1252 #ifdef CONFIG_IPV6_MIP6
1253         IPV6_SADDR_RULE_HOA,
1254 #endif
1255         IPV6_SADDR_RULE_OIF,
1256         IPV6_SADDR_RULE_LABEL,
1257         IPV6_SADDR_RULE_PRIVACY,
1258         IPV6_SADDR_RULE_ORCHID,
1259         IPV6_SADDR_RULE_PREFIX,
1260 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1261         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1262 #endif
1263         IPV6_SADDR_RULE_MAX
1264 };
1265
1266 struct ipv6_saddr_score {
1267         int                     rule;
1268         int                     addr_type;
1269         struct inet6_ifaddr     *ifa;
1270         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1271         int                     scopedist;
1272         int                     matchlen;
1273 };
1274
1275 struct ipv6_saddr_dst {
1276         const struct in6_addr *addr;
1277         int ifindex;
1278         int scope;
1279         int label;
1280         unsigned int prefs;
1281 };
1282
1283 static inline int ipv6_saddr_preferred(int type)
1284 {
1285         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1286                 return 1;
1287         return 0;
1288 }
1289
1290 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1291 {
1292 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1293         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1294 #else
1295         return false;
1296 #endif
1297 }
1298
1299 static int ipv6_get_saddr_eval(struct net *net,
1300                                struct ipv6_saddr_score *score,
1301                                struct ipv6_saddr_dst *dst,
1302                                int i)
1303 {
1304         int ret;
1305
1306         if (i <= score->rule) {
1307                 switch (i) {
1308                 case IPV6_SADDR_RULE_SCOPE:
1309                         ret = score->scopedist;
1310                         break;
1311                 case IPV6_SADDR_RULE_PREFIX:
1312                         ret = score->matchlen;
1313                         break;
1314                 default:
1315                         ret = !!test_bit(i, score->scorebits);
1316                 }
1317                 goto out;
1318         }
1319
1320         switch (i) {
1321         case IPV6_SADDR_RULE_INIT:
1322                 /* Rule 0: remember if hiscore is not ready yet */
1323                 ret = !!score->ifa;
1324                 break;
1325         case IPV6_SADDR_RULE_LOCAL:
1326                 /* Rule 1: Prefer same address */
1327                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1328                 break;
1329         case IPV6_SADDR_RULE_SCOPE:
1330                 /* Rule 2: Prefer appropriate scope
1331                  *
1332                  *      ret
1333                  *       ^
1334                  *    -1 |  d 15
1335                  *    ---+--+-+---> scope
1336                  *       |
1337                  *       |             d is scope of the destination.
1338                  *  B-d  |  \
1339                  *       |   \      <- smaller scope is better if
1340                  *  B-15 |    \        if scope is enough for destination.
1341                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1342                  * d-C-1 | /
1343                  *       |/         <- greater is better
1344                  *   -C  /             if scope is not enough for destination.
1345                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1346                  *
1347                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1348                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1349                  * Assume B = 0 and we get C > 29.
1350                  */
1351                 ret = __ipv6_addr_src_scope(score->addr_type);
1352                 if (ret >= dst->scope)
1353                         ret = -ret;
1354                 else
1355                         ret -= 128;     /* 30 is enough */
1356                 score->scopedist = ret;
1357                 break;
1358         case IPV6_SADDR_RULE_PREFERRED:
1359             {
1360                 /* Rule 3: Avoid deprecated and optimistic addresses */
1361                 u8 avoid = IFA_F_DEPRECATED;
1362
1363                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1364                         avoid |= IFA_F_OPTIMISTIC;
1365                 ret = ipv6_saddr_preferred(score->addr_type) ||
1366                       !(score->ifa->flags & avoid);
1367                 break;
1368             }
1369 #ifdef CONFIG_IPV6_MIP6
1370         case IPV6_SADDR_RULE_HOA:
1371             {
1372                 /* Rule 4: Prefer home address */
1373                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1374                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1375                 break;
1376             }
1377 #endif
1378         case IPV6_SADDR_RULE_OIF:
1379                 /* Rule 5: Prefer outgoing interface */
1380                 ret = (!dst->ifindex ||
1381                        dst->ifindex == score->ifa->idev->dev->ifindex);
1382                 break;
1383         case IPV6_SADDR_RULE_LABEL:
1384                 /* Rule 6: Prefer matching label */
1385                 ret = ipv6_addr_label(net,
1386                                       &score->ifa->addr, score->addr_type,
1387                                       score->ifa->idev->dev->ifindex) == dst->label;
1388                 break;
1389         case IPV6_SADDR_RULE_PRIVACY:
1390             {
1391                 /* Rule 7: Prefer public address
1392                  * Note: prefer temporary address if use_tempaddr >= 2
1393                  */
1394                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1395                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1396                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1397                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1398                 break;
1399             }
1400         case IPV6_SADDR_RULE_ORCHID:
1401                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1402                  *          non-ORCHID vs non-ORCHID
1403                  */
1404                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1405                         ipv6_addr_orchid(dst->addr));
1406                 break;
1407         case IPV6_SADDR_RULE_PREFIX:
1408                 /* Rule 8: Use longest matching prefix */
1409                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1410                 if (ret > score->ifa->prefix_len)
1411                         ret = score->ifa->prefix_len;
1412                 score->matchlen = ret;
1413                 break;
1414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1415         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1416                 /* Optimistic addresses still have lower precedence than other
1417                  * preferred addresses.
1418                  */
1419                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1420                 break;
1421 #endif
1422         default:
1423                 ret = 0;
1424         }
1425
1426         if (ret)
1427                 __set_bit(i, score->scorebits);
1428         score->rule = i;
1429 out:
1430         return ret;
1431 }
1432
1433 static int __ipv6_dev_get_saddr(struct net *net,
1434                                 struct ipv6_saddr_dst *dst,
1435                                 struct inet6_dev *idev,
1436                                 struct ipv6_saddr_score *scores,
1437                                 int hiscore_idx)
1438 {
1439         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1440
1441         read_lock_bh(&idev->lock);
1442         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1443                 int i;
1444
1445                 /*
1446                  * - Tentative Address (RFC2462 section 5.4)
1447                  *  - A tentative address is not considered
1448                  *    "assigned to an interface" in the traditional
1449                  *    sense, unless it is also flagged as optimistic.
1450                  * - Candidate Source Address (section 4)
1451                  *  - In any case, anycast addresses, multicast
1452                  *    addresses, and the unspecified address MUST
1453                  *    NOT be included in a candidate set.
1454                  */
1455                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1456                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1457                         continue;
1458
1459                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1460
1461                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1462                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1463                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1464                                             idev->dev->name);
1465                         continue;
1466                 }
1467
1468                 score->rule = -1;
1469                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1470
1471                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1472                         int minihiscore, miniscore;
1473
1474                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1475                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1476
1477                         if (minihiscore > miniscore) {
1478                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1479                                     score->scopedist > 0) {
1480                                         /*
1481                                          * special case:
1482                                          * each remaining entry
1483                                          * has too small (not enough)
1484                                          * scope, because ifa entries
1485                                          * are sorted by their scope
1486                                          * values.
1487                                          */
1488                                         goto out;
1489                                 }
1490                                 break;
1491                         } else if (minihiscore < miniscore) {
1492                                 if (hiscore->ifa)
1493                                         in6_ifa_put(hiscore->ifa);
1494
1495                                 in6_ifa_hold(score->ifa);
1496
1497                                 swap(hiscore, score);
1498                                 hiscore_idx = 1 - hiscore_idx;
1499
1500                                 /* restore our iterator */
1501                                 score->ifa = hiscore->ifa;
1502
1503                                 break;
1504                         }
1505                 }
1506         }
1507 out:
1508         read_unlock_bh(&idev->lock);
1509         return hiscore_idx;
1510 }
1511
1512 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1513                        const struct in6_addr *daddr, unsigned int prefs,
1514                        struct in6_addr *saddr)
1515 {
1516         struct ipv6_saddr_score scores[2], *hiscore;
1517         struct ipv6_saddr_dst dst;
1518         struct inet6_dev *idev;
1519         struct net_device *dev;
1520         int dst_type;
1521         bool use_oif_addr = false;
1522         int hiscore_idx = 0;
1523
1524         dst_type = __ipv6_addr_type(daddr);
1525         dst.addr = daddr;
1526         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1527         dst.scope = __ipv6_addr_src_scope(dst_type);
1528         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1529         dst.prefs = prefs;
1530
1531         scores[hiscore_idx].rule = -1;
1532         scores[hiscore_idx].ifa = NULL;
1533
1534         rcu_read_lock();
1535
1536         /* Candidate Source Address (section 4)
1537          *  - multicast and link-local destination address,
1538          *    the set of candidate source address MUST only
1539          *    include addresses assigned to interfaces
1540          *    belonging to the same link as the outgoing
1541          *    interface.
1542          * (- For site-local destination addresses, the
1543          *    set of candidate source addresses MUST only
1544          *    include addresses assigned to interfaces
1545          *    belonging to the same site as the outgoing
1546          *    interface.)
1547          *  - "It is RECOMMENDED that the candidate source addresses
1548          *    be the set of unicast addresses assigned to the
1549          *    interface that will be used to send to the destination
1550          *    (the 'outgoing' interface)." (RFC 6724)
1551          */
1552         if (dst_dev) {
1553                 idev = __in6_dev_get(dst_dev);
1554                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1555                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1556                     (idev && idev->cnf.use_oif_addrs_only)) {
1557                         use_oif_addr = true;
1558                 }
1559         }
1560
1561         if (use_oif_addr) {
1562                 if (idev)
1563                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1564         } else {
1565                 for_each_netdev_rcu(net, dev) {
1566                         idev = __in6_dev_get(dev);
1567                         if (!idev)
1568                                 continue;
1569                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1570                 }
1571         }
1572         rcu_read_unlock();
1573
1574         hiscore = &scores[hiscore_idx];
1575         if (!hiscore->ifa)
1576                 return -EADDRNOTAVAIL;
1577
1578         *saddr = hiscore->ifa->addr;
1579         in6_ifa_put(hiscore->ifa);
1580         return 0;
1581 }
1582 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1583
1584 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1585                       u32 banned_flags)
1586 {
1587         struct inet6_ifaddr *ifp;
1588         int err = -EADDRNOTAVAIL;
1589
1590         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1591                 if (ifp->scope > IFA_LINK)
1592                         break;
1593                 if (ifp->scope == IFA_LINK &&
1594                     !(ifp->flags & banned_flags)) {
1595                         *addr = ifp->addr;
1596                         err = 0;
1597                         break;
1598                 }
1599         }
1600         return err;
1601 }
1602
1603 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1604                     u32 banned_flags)
1605 {
1606         struct inet6_dev *idev;
1607         int err = -EADDRNOTAVAIL;
1608
1609         rcu_read_lock();
1610         idev = __in6_dev_get(dev);
1611         if (idev) {
1612                 read_lock_bh(&idev->lock);
1613                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1614                 read_unlock_bh(&idev->lock);
1615         }
1616         rcu_read_unlock();
1617         return err;
1618 }
1619
1620 static int ipv6_count_addresses(struct inet6_dev *idev)
1621 {
1622         int cnt = 0;
1623         struct inet6_ifaddr *ifp;
1624
1625         read_lock_bh(&idev->lock);
1626         list_for_each_entry(ifp, &idev->addr_list, if_list)
1627                 cnt++;
1628         read_unlock_bh(&idev->lock);
1629         return cnt;
1630 }
1631
1632 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1633                   const struct net_device *dev, int strict)
1634 {
1635         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1636 }
1637 EXPORT_SYMBOL(ipv6_chk_addr);
1638
1639 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1640                             const struct net_device *dev, int strict,
1641                             u32 banned_flags)
1642 {
1643         struct inet6_ifaddr *ifp;
1644         unsigned int hash = inet6_addr_hash(addr);
1645         u32 ifp_flags;
1646
1647         rcu_read_lock_bh();
1648         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1649                 if (!net_eq(dev_net(ifp->idev->dev), net))
1650                         continue;
1651                 /* Decouple optimistic from tentative for evaluation here.
1652                  * Ban optimistic addresses explicitly, when required.
1653                  */
1654                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1655                             ? (ifp->flags&~IFA_F_TENTATIVE)
1656                             : ifp->flags;
1657                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1658                     !(ifp_flags&banned_flags) &&
1659                     (!dev || ifp->idev->dev == dev ||
1660                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1661                         rcu_read_unlock_bh();
1662                         return 1;
1663                 }
1664         }
1665
1666         rcu_read_unlock_bh();
1667         return 0;
1668 }
1669 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1670
1671 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1672                                struct net_device *dev)
1673 {
1674         unsigned int hash = inet6_addr_hash(addr);
1675         struct inet6_ifaddr *ifp;
1676
1677         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1678                 if (!net_eq(dev_net(ifp->idev->dev), net))
1679                         continue;
1680                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1681                         if (!dev || ifp->idev->dev == dev)
1682                                 return true;
1683                 }
1684         }
1685         return false;
1686 }
1687
1688 /* Compares an address/prefix_len with addresses on device @dev.
1689  * If one is found it returns true.
1690  */
1691 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1692         const unsigned int prefix_len, struct net_device *dev)
1693 {
1694         struct inet6_dev *idev;
1695         struct inet6_ifaddr *ifa;
1696         bool ret = false;
1697
1698         rcu_read_lock();
1699         idev = __in6_dev_get(dev);
1700         if (idev) {
1701                 read_lock_bh(&idev->lock);
1702                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1703                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1704                         if (ret)
1705                                 break;
1706                 }
1707                 read_unlock_bh(&idev->lock);
1708         }
1709         rcu_read_unlock();
1710
1711         return ret;
1712 }
1713 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1714
1715 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1716 {
1717         struct inet6_dev *idev;
1718         struct inet6_ifaddr *ifa;
1719         int     onlink;
1720
1721         onlink = 0;
1722         rcu_read_lock();
1723         idev = __in6_dev_get(dev);
1724         if (idev) {
1725                 read_lock_bh(&idev->lock);
1726                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1727                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1728                                                    ifa->prefix_len);
1729                         if (onlink)
1730                                 break;
1731                 }
1732                 read_unlock_bh(&idev->lock);
1733         }
1734         rcu_read_unlock();
1735         return onlink;
1736 }
1737 EXPORT_SYMBOL(ipv6_chk_prefix);
1738
1739 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1740                                      struct net_device *dev, int strict)
1741 {
1742         struct inet6_ifaddr *ifp, *result = NULL;
1743         unsigned int hash = inet6_addr_hash(addr);
1744
1745         rcu_read_lock_bh();
1746         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1747                 if (!net_eq(dev_net(ifp->idev->dev), net))
1748                         continue;
1749                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1750                         if (!dev || ifp->idev->dev == dev ||
1751                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1752                                 result = ifp;
1753                                 in6_ifa_hold(ifp);
1754                                 break;
1755                         }
1756                 }
1757         }
1758         rcu_read_unlock_bh();
1759
1760         return result;
1761 }
1762
1763 /* Gets referenced address, destroys ifaddr */
1764
1765 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1766 {
1767         if (ifp->flags&IFA_F_PERMANENT) {
1768                 spin_lock_bh(&ifp->lock);
1769                 addrconf_del_dad_work(ifp);
1770                 ifp->flags |= IFA_F_TENTATIVE;
1771                 if (dad_failed)
1772                         ifp->flags |= IFA_F_DADFAILED;
1773                 spin_unlock_bh(&ifp->lock);
1774                 if (dad_failed)
1775                         ipv6_ifa_notify(0, ifp);
1776                 in6_ifa_put(ifp);
1777         } else if (ifp->flags&IFA_F_TEMPORARY) {
1778                 struct inet6_ifaddr *ifpub;
1779                 spin_lock_bh(&ifp->lock);
1780                 ifpub = ifp->ifpub;
1781                 if (ifpub) {
1782                         in6_ifa_hold(ifpub);
1783                         spin_unlock_bh(&ifp->lock);
1784                         ipv6_create_tempaddr(ifpub, ifp);
1785                         in6_ifa_put(ifpub);
1786                 } else {
1787                         spin_unlock_bh(&ifp->lock);
1788                 }
1789                 ipv6_del_addr(ifp);
1790         } else {
1791                 ipv6_del_addr(ifp);
1792         }
1793 }
1794
1795 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1796 {
1797         int err = -ENOENT;
1798
1799         spin_lock_bh(&ifp->lock);
1800         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1801                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1802                 err = 0;
1803         }
1804         spin_unlock_bh(&ifp->lock);
1805
1806         return err;
1807 }
1808
1809 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1810 {
1811         struct in6_addr addr;
1812         struct inet6_dev *idev = ifp->idev;
1813         struct net *net = dev_net(ifp->idev->dev);
1814
1815         if (addrconf_dad_end(ifp)) {
1816                 in6_ifa_put(ifp);
1817                 return;
1818         }
1819
1820         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1821                              ifp->idev->dev->name, &ifp->addr);
1822
1823         spin_lock_bh(&ifp->lock);
1824
1825         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1826                 int scope = ifp->scope;
1827                 u32 flags = ifp->flags;
1828                 struct in6_addr new_addr;
1829                 struct inet6_ifaddr *ifp2;
1830                 u32 valid_lft, preferred_lft;
1831                 int pfxlen = ifp->prefix_len;
1832                 int retries = ifp->stable_privacy_retry + 1;
1833
1834                 if (retries > net->ipv6.sysctl.idgen_retries) {
1835                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1836                                              ifp->idev->dev->name);
1837                         goto errdad;
1838                 }
1839
1840                 new_addr = ifp->addr;
1841                 if (ipv6_generate_stable_address(&new_addr, retries,
1842                                                  idev))
1843                         goto errdad;
1844
1845                 valid_lft = ifp->valid_lft;
1846                 preferred_lft = ifp->prefered_lft;
1847
1848                 spin_unlock_bh(&ifp->lock);
1849
1850                 if (idev->cnf.max_addresses &&
1851                     ipv6_count_addresses(idev) >=
1852                     idev->cnf.max_addresses)
1853                         goto lock_errdad;
1854
1855                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1856                                      ifp->idev->dev->name);
1857
1858                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1859                                      scope, flags, valid_lft,
1860                                      preferred_lft);
1861                 if (IS_ERR(ifp2))
1862                         goto lock_errdad;
1863
1864                 spin_lock_bh(&ifp2->lock);
1865                 ifp2->stable_privacy_retry = retries;
1866                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1867                 spin_unlock_bh(&ifp2->lock);
1868
1869                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1870                 in6_ifa_put(ifp2);
1871 lock_errdad:
1872                 spin_lock_bh(&ifp->lock);
1873         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1874                 addr.s6_addr32[0] = htonl(0xfe800000);
1875                 addr.s6_addr32[1] = 0;
1876
1877                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1878                     ipv6_addr_equal(&ifp->addr, &addr)) {
1879                         /* DAD failed for link-local based on MAC address */
1880                         idev->cnf.disable_ipv6 = 1;
1881
1882                         pr_info("%s: IPv6 being disabled!\n",
1883                                 ifp->idev->dev->name);
1884                 }
1885         }
1886
1887 errdad:
1888         /* transition from _POSTDAD to _ERRDAD */
1889         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1890         spin_unlock_bh(&ifp->lock);
1891
1892         addrconf_mod_dad_work(ifp, 0);
1893 }
1894
1895 /* Join to solicited addr multicast group.
1896  * caller must hold RTNL */
1897 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1898 {
1899         struct in6_addr maddr;
1900
1901         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1902                 return;
1903
1904         addrconf_addr_solict_mult(addr, &maddr);
1905         ipv6_dev_mc_inc(dev, &maddr);
1906 }
1907
1908 /* caller must hold RTNL */
1909 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1910 {
1911         struct in6_addr maddr;
1912
1913         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1914                 return;
1915
1916         addrconf_addr_solict_mult(addr, &maddr);
1917         __ipv6_dev_mc_dec(idev, &maddr);
1918 }
1919
1920 /* caller must hold RTNL */
1921 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1922 {
1923         struct in6_addr addr;
1924
1925         if (ifp->prefix_len >= 127) /* RFC 6164 */
1926                 return;
1927         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1928         if (ipv6_addr_any(&addr))
1929                 return;
1930         __ipv6_dev_ac_inc(ifp->idev, &addr);
1931 }
1932
1933 /* caller must hold RTNL */
1934 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1935 {
1936         struct in6_addr addr;
1937
1938         if (ifp->prefix_len >= 127) /* RFC 6164 */
1939                 return;
1940         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1941         if (ipv6_addr_any(&addr))
1942                 return;
1943         __ipv6_dev_ac_dec(ifp->idev, &addr);
1944 }
1945
1946 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1947 {
1948         if (dev->addr_len != IEEE802154_ADDR_LEN)
1949                 return -1;
1950         memcpy(eui, dev->dev_addr, 8);
1951         eui[0] ^= 2;
1952         return 0;
1953 }
1954
1955 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1956 {
1957         union fwnet_hwaddr *ha;
1958
1959         if (dev->addr_len != FWNET_ALEN)
1960                 return -1;
1961
1962         ha = (union fwnet_hwaddr *)dev->dev_addr;
1963
1964         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1965         eui[0] ^= 2;
1966         return 0;
1967 }
1968
1969 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1970 {
1971         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1972         if (dev->addr_len != ARCNET_ALEN)
1973                 return -1;
1974         memset(eui, 0, 7);
1975         eui[7] = *(u8 *)dev->dev_addr;
1976         return 0;
1977 }
1978
1979 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1980 {
1981         if (dev->addr_len != INFINIBAND_ALEN)
1982                 return -1;
1983         memcpy(eui, dev->dev_addr + 12, 8);
1984         eui[0] |= 2;
1985         return 0;
1986 }
1987
1988 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1989 {
1990         if (addr == 0)
1991                 return -1;
1992         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1993                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1994                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1995                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1996                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1997                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1998         eui[1] = 0;
1999         eui[2] = 0x5E;
2000         eui[3] = 0xFE;
2001         memcpy(eui + 4, &addr, 4);
2002         return 0;
2003 }
2004
2005 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2006 {
2007         if (dev->priv_flags & IFF_ISATAP)
2008                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2009         return -1;
2010 }
2011
2012 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2013 {
2014         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2015 }
2016
2017 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2018 {
2019         memcpy(eui, dev->perm_addr, 3);
2020         memcpy(eui + 5, dev->perm_addr + 3, 3);
2021         eui[3] = 0xFF;
2022         eui[4] = 0xFE;
2023         eui[0] ^= 2;
2024         return 0;
2025 }
2026
2027 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2028 {
2029         switch (dev->type) {
2030         case ARPHRD_ETHER:
2031         case ARPHRD_FDDI:
2032                 return addrconf_ifid_eui48(eui, dev);
2033         case ARPHRD_ARCNET:
2034                 return addrconf_ifid_arcnet(eui, dev);
2035         case ARPHRD_INFINIBAND:
2036                 return addrconf_ifid_infiniband(eui, dev);
2037         case ARPHRD_SIT:
2038                 return addrconf_ifid_sit(eui, dev);
2039         case ARPHRD_IPGRE:
2040                 return addrconf_ifid_gre(eui, dev);
2041         case ARPHRD_6LOWPAN:
2042         case ARPHRD_IEEE802154:
2043                 return addrconf_ifid_eui64(eui, dev);
2044         case ARPHRD_IEEE1394:
2045                 return addrconf_ifid_ieee1394(eui, dev);
2046         case ARPHRD_TUNNEL6:
2047                 return addrconf_ifid_ip6tnl(eui, dev);
2048         }
2049         return -1;
2050 }
2051
2052 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2053 {
2054         int err = -1;
2055         struct inet6_ifaddr *ifp;
2056
2057         read_lock_bh(&idev->lock);
2058         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2059                 if (ifp->scope > IFA_LINK)
2060                         break;
2061                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2062                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2063                         err = 0;
2064                         break;
2065                 }
2066         }
2067         read_unlock_bh(&idev->lock);
2068         return err;
2069 }
2070
2071 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2072 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2073 {
2074 regen:
2075         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2076         idev->rndid[0] &= ~0x02;
2077
2078         /*
2079          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2080          * check if generated address is not inappropriate
2081          *
2082          *  - Reserved subnet anycast (RFC 2526)
2083          *      11111101 11....11 1xxxxxxx
2084          *  - ISATAP (RFC4214) 6.1
2085          *      00-00-5E-FE-xx-xx-xx-xx
2086          *  - value 0
2087          *  - XXX: already assigned to an address on the device
2088          */
2089         if (idev->rndid[0] == 0xfd &&
2090             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2091             (idev->rndid[7]&0x80))
2092                 goto regen;
2093         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2094                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2095                         goto regen;
2096                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2097                         goto regen;
2098         }
2099 }
2100
2101 static void ipv6_regen_rndid(unsigned long data)
2102 {
2103         struct inet6_dev *idev = (struct inet6_dev *) data;
2104         unsigned long expires;
2105
2106         rcu_read_lock_bh();
2107         write_lock_bh(&idev->lock);
2108
2109         if (idev->dead)
2110                 goto out;
2111
2112         __ipv6_regen_rndid(idev);
2113
2114         expires = jiffies +
2115                 idev->cnf.temp_prefered_lft * HZ -
2116                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2117                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2118                 idev->cnf.max_desync_factor * HZ;
2119         if (time_before(expires, jiffies)) {
2120                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2121                         __func__, idev->dev->name);
2122                 goto out;
2123         }
2124
2125         if (!mod_timer(&idev->regen_timer, expires))
2126                 in6_dev_hold(idev);
2127
2128 out:
2129         write_unlock_bh(&idev->lock);
2130         rcu_read_unlock_bh();
2131         in6_dev_put(idev);
2132 }
2133
2134 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2135 {
2136         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2137                 __ipv6_regen_rndid(idev);
2138 }
2139
2140 /*
2141  *      Add prefix route.
2142  */
2143
2144 static void
2145 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2146                       unsigned long expires, u32 flags)
2147 {
2148         struct fib6_config cfg = {
2149                 .fc_table = RT6_TABLE_PREFIX,
2150                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2151                 .fc_ifindex = dev->ifindex,
2152                 .fc_expires = expires,
2153                 .fc_dst_len = plen,
2154                 .fc_flags = RTF_UP | flags,
2155                 .fc_nlinfo.nl_net = dev_net(dev),
2156                 .fc_protocol = RTPROT_KERNEL,
2157         };
2158
2159         cfg.fc_dst = *pfx;
2160
2161         /* Prevent useless cloning on PtP SIT.
2162            This thing is done here expecting that the whole
2163            class of non-broadcast devices need not cloning.
2164          */
2165 #if IS_ENABLED(CONFIG_IPV6_SIT)
2166         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2167                 cfg.fc_flags |= RTF_NONEXTHOP;
2168 #endif
2169
2170         ip6_route_add(&cfg);
2171 }
2172
2173
2174 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2175                                                   int plen,
2176                                                   const struct net_device *dev,
2177                                                   u32 flags, u32 noflags)
2178 {
2179         struct fib6_node *fn;
2180         struct rt6_info *rt = NULL;
2181         struct fib6_table *table;
2182
2183         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2184         if (!table)
2185                 return NULL;
2186
2187         read_lock_bh(&table->tb6_lock);
2188         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2189         if (!fn)
2190                 goto out;
2191
2192         noflags |= RTF_CACHE;
2193         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2194                 if (rt->dst.dev->ifindex != dev->ifindex)
2195                         continue;
2196                 if ((rt->rt6i_flags & flags) != flags)
2197                         continue;
2198                 if ((rt->rt6i_flags & noflags) != 0)
2199                         continue;
2200                 dst_hold(&rt->dst);
2201                 break;
2202         }
2203 out:
2204         read_unlock_bh(&table->tb6_lock);
2205         return rt;
2206 }
2207
2208
2209 /* Create "default" multicast route to the interface */
2210
2211 static void addrconf_add_mroute(struct net_device *dev)
2212 {
2213         struct fib6_config cfg = {
2214                 .fc_table = RT6_TABLE_LOCAL,
2215                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2216                 .fc_ifindex = dev->ifindex,
2217                 .fc_dst_len = 8,
2218                 .fc_flags = RTF_UP,
2219                 .fc_nlinfo.nl_net = dev_net(dev),
2220         };
2221
2222         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2223
2224         ip6_route_add(&cfg);
2225 }
2226
2227 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2228 {
2229         struct inet6_dev *idev;
2230
2231         ASSERT_RTNL();
2232
2233         idev = ipv6_find_idev(dev);
2234         if (!idev)
2235                 return ERR_PTR(-ENOBUFS);
2236
2237         if (idev->cnf.disable_ipv6)
2238                 return ERR_PTR(-EACCES);
2239
2240         /* Add default multicast route */
2241         if (!(dev->flags & IFF_LOOPBACK))
2242                 addrconf_add_mroute(dev);
2243
2244         return idev;
2245 }
2246
2247 static void manage_tempaddrs(struct inet6_dev *idev,
2248                              struct inet6_ifaddr *ifp,
2249                              __u32 valid_lft, __u32 prefered_lft,
2250                              bool create, unsigned long now)
2251 {
2252         u32 flags;
2253         struct inet6_ifaddr *ift;
2254
2255         read_lock_bh(&idev->lock);
2256         /* update all temporary addresses in the list */
2257         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2258                 int age, max_valid, max_prefered;
2259
2260                 if (ifp != ift->ifpub)
2261                         continue;
2262
2263                 /* RFC 4941 section 3.3:
2264                  * If a received option will extend the lifetime of a public
2265                  * address, the lifetimes of temporary addresses should
2266                  * be extended, subject to the overall constraint that no
2267                  * temporary addresses should ever remain "valid" or "preferred"
2268                  * for a time longer than (TEMP_VALID_LIFETIME) or
2269                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2270                  */
2271                 age = (now - ift->cstamp) / HZ;
2272                 max_valid = idev->cnf.temp_valid_lft - age;
2273                 if (max_valid < 0)
2274                         max_valid = 0;
2275
2276                 max_prefered = idev->cnf.temp_prefered_lft -
2277                                idev->cnf.max_desync_factor - age;
2278                 if (max_prefered < 0)
2279                         max_prefered = 0;
2280
2281                 if (valid_lft > max_valid)
2282                         valid_lft = max_valid;
2283
2284                 if (prefered_lft > max_prefered)
2285                         prefered_lft = max_prefered;
2286
2287                 spin_lock(&ift->lock);
2288                 flags = ift->flags;
2289                 ift->valid_lft = valid_lft;
2290                 ift->prefered_lft = prefered_lft;
2291                 ift->tstamp = now;
2292                 if (prefered_lft > 0)
2293                         ift->flags &= ~IFA_F_DEPRECATED;
2294
2295                 spin_unlock(&ift->lock);
2296                 if (!(flags&IFA_F_TENTATIVE))
2297                         ipv6_ifa_notify(0, ift);
2298         }
2299
2300         if ((create || list_empty(&idev->tempaddr_list)) &&
2301             idev->cnf.use_tempaddr > 0) {
2302                 /* When a new public address is created as described
2303                  * in [ADDRCONF], also create a new temporary address.
2304                  * Also create a temporary address if it's enabled but
2305                  * no temporary address currently exists.
2306                  */
2307                 read_unlock_bh(&idev->lock);
2308                 ipv6_create_tempaddr(ifp, NULL);
2309         } else {
2310                 read_unlock_bh(&idev->lock);
2311         }
2312 }
2313
2314 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2315 {
2316         struct prefix_info *pinfo;
2317         __u32 valid_lft;
2318         __u32 prefered_lft;
2319         int addr_type;
2320         u32 addr_flags = 0;
2321         struct inet6_dev *in6_dev;
2322         struct net *net = dev_net(dev);
2323
2324         pinfo = (struct prefix_info *) opt;
2325
2326         if (len < sizeof(struct prefix_info)) {
2327                 ADBG("addrconf: prefix option too short\n");
2328                 return;
2329         }
2330
2331         /*
2332          *      Validation checks ([ADDRCONF], page 19)
2333          */
2334
2335         addr_type = ipv6_addr_type(&pinfo->prefix);
2336
2337         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2338                 return;
2339
2340         valid_lft = ntohl(pinfo->valid);
2341         prefered_lft = ntohl(pinfo->prefered);
2342
2343         if (prefered_lft > valid_lft) {
2344                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2345                 return;
2346         }
2347
2348         in6_dev = in6_dev_get(dev);
2349
2350         if (!in6_dev) {
2351                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2352                                     dev->name);
2353                 return;
2354         }
2355
2356         /*
2357          *      Two things going on here:
2358          *      1) Add routes for on-link prefixes
2359          *      2) Configure prefixes with the auto flag set
2360          */
2361
2362         if (pinfo->onlink) {
2363                 struct rt6_info *rt;
2364                 unsigned long rt_expires;
2365
2366                 /* Avoid arithmetic overflow. Really, we could
2367                  * save rt_expires in seconds, likely valid_lft,
2368                  * but it would require division in fib gc, that it
2369                  * not good.
2370                  */
2371                 if (HZ > USER_HZ)
2372                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2373                 else
2374                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2375
2376                 if (addrconf_finite_timeout(rt_expires))
2377                         rt_expires *= HZ;
2378
2379                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2380                                                pinfo->prefix_len,
2381                                                dev,
2382                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2383                                                RTF_GATEWAY | RTF_DEFAULT);
2384
2385                 if (rt) {
2386                         /* Autoconf prefix route */
2387                         if (valid_lft == 0) {
2388                                 ip6_del_rt(rt);
2389                                 rt = NULL;
2390                         } else if (addrconf_finite_timeout(rt_expires)) {
2391                                 /* not infinity */
2392                                 rt6_set_expires(rt, jiffies + rt_expires);
2393                         } else {
2394                                 rt6_clean_expires(rt);
2395                         }
2396                 } else if (valid_lft) {
2397                         clock_t expires = 0;
2398                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2399                         if (addrconf_finite_timeout(rt_expires)) {
2400                                 /* not infinity */
2401                                 flags |= RTF_EXPIRES;
2402                                 expires = jiffies_to_clock_t(rt_expires);
2403                         }
2404                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2405                                               dev, expires, flags);
2406                 }
2407                 ip6_rt_put(rt);
2408         }
2409
2410         /* Try to figure out our local address for this prefix */
2411
2412         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2413                 struct inet6_ifaddr *ifp;
2414                 struct in6_addr addr;
2415                 int create = 0, update_lft = 0;
2416                 bool tokenized = false;
2417
2418                 if (pinfo->prefix_len == 64) {
2419                         memcpy(&addr, &pinfo->prefix, 8);
2420
2421                         if (!ipv6_addr_any(&in6_dev->token)) {
2422                                 read_lock_bh(&in6_dev->lock);
2423                                 memcpy(addr.s6_addr + 8,
2424                                        in6_dev->token.s6_addr + 8, 8);
2425                                 read_unlock_bh(&in6_dev->lock);
2426                                 tokenized = true;
2427                         } else if (in6_dev->addr_gen_mode ==
2428                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2429                                    !ipv6_generate_stable_address(&addr, 0,
2430                                                                  in6_dev)) {
2431                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2432                                 goto ok;
2433                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2434                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2435                                 in6_dev_put(in6_dev);
2436                                 return;
2437                         }
2438                         goto ok;
2439                 }
2440                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2441                                     pinfo->prefix_len);
2442                 in6_dev_put(in6_dev);
2443                 return;
2444
2445 ok:
2446
2447                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2448
2449                 if (!ifp && valid_lft) {
2450                         int max_addresses = in6_dev->cnf.max_addresses;
2451
2452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2453                         if (in6_dev->cnf.optimistic_dad &&
2454                             !net->ipv6.devconf_all->forwarding && sllao)
2455                                 addr_flags = IFA_F_OPTIMISTIC;
2456 #endif
2457
2458                         /* Do not allow to create too much of autoconfigured
2459                          * addresses; this would be too easy way to crash kernel.
2460                          */
2461                         if (!max_addresses ||
2462                             ipv6_count_addresses(in6_dev) < max_addresses)
2463                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2464                                                     pinfo->prefix_len,
2465                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2466                                                     addr_flags, valid_lft,
2467                                                     prefered_lft);
2468
2469                         if (IS_ERR_OR_NULL(ifp)) {
2470                                 in6_dev_put(in6_dev);
2471                                 return;
2472                         }
2473
2474                         update_lft = 0;
2475                         create = 1;
2476                         spin_lock_bh(&ifp->lock);
2477                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2478                         ifp->cstamp = jiffies;
2479                         ifp->tokenized = tokenized;
2480                         spin_unlock_bh(&ifp->lock);
2481                         addrconf_dad_start(ifp);
2482                 }
2483
2484                 if (ifp) {
2485                         u32 flags;
2486                         unsigned long now;
2487                         u32 stored_lft;
2488
2489                         /* update lifetime (RFC2462 5.5.3 e) */
2490                         spin_lock_bh(&ifp->lock);
2491                         now = jiffies;
2492                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2493                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2494                         else
2495                                 stored_lft = 0;
2496                         if (!update_lft && !create && stored_lft) {
2497                                 const u32 minimum_lft = min_t(u32,
2498                                         stored_lft, MIN_VALID_LIFETIME);
2499                                 valid_lft = max(valid_lft, minimum_lft);
2500
2501                                 /* RFC4862 Section 5.5.3e:
2502                                  * "Note that the preferred lifetime of the
2503                                  *  corresponding address is always reset to
2504                                  *  the Preferred Lifetime in the received
2505                                  *  Prefix Information option, regardless of
2506                                  *  whether the valid lifetime is also reset or
2507                                  *  ignored."
2508                                  *
2509                                  * So we should always update prefered_lft here.
2510                                  */
2511                                 update_lft = 1;
2512                         }
2513
2514                         if (update_lft) {
2515                                 ifp->valid_lft = valid_lft;
2516                                 ifp->prefered_lft = prefered_lft;
2517                                 ifp->tstamp = now;
2518                                 flags = ifp->flags;
2519                                 ifp->flags &= ~IFA_F_DEPRECATED;
2520                                 spin_unlock_bh(&ifp->lock);
2521
2522                                 if (!(flags&IFA_F_TENTATIVE))
2523                                         ipv6_ifa_notify(0, ifp);
2524                         } else
2525                                 spin_unlock_bh(&ifp->lock);
2526
2527                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2528                                          create, now);
2529
2530                         in6_ifa_put(ifp);
2531                         addrconf_verify();
2532                 }
2533         }
2534         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2535         in6_dev_put(in6_dev);
2536 }
2537
2538 /*
2539  *      Set destination address.
2540  *      Special case for SIT interfaces where we create a new "virtual"
2541  *      device.
2542  */
2543 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2544 {
2545         struct in6_ifreq ireq;
2546         struct net_device *dev;
2547         int err = -EINVAL;
2548
2549         rtnl_lock();
2550
2551         err = -EFAULT;
2552         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2553                 goto err_exit;
2554
2555         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2556
2557         err = -ENODEV;
2558         if (!dev)
2559                 goto err_exit;
2560
2561 #if IS_ENABLED(CONFIG_IPV6_SIT)
2562         if (dev->type == ARPHRD_SIT) {
2563                 const struct net_device_ops *ops = dev->netdev_ops;
2564                 struct ifreq ifr;
2565                 struct ip_tunnel_parm p;
2566
2567                 err = -EADDRNOTAVAIL;
2568                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2569                         goto err_exit;
2570
2571                 memset(&p, 0, sizeof(p));
2572                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2573                 p.iph.saddr = 0;
2574                 p.iph.version = 4;
2575                 p.iph.ihl = 5;
2576                 p.iph.protocol = IPPROTO_IPV6;
2577                 p.iph.ttl = 64;
2578                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2579
2580                 if (ops->ndo_do_ioctl) {
2581                         mm_segment_t oldfs = get_fs();
2582
2583                         set_fs(KERNEL_DS);
2584                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2585                         set_fs(oldfs);
2586                 } else
2587                         err = -EOPNOTSUPP;
2588
2589                 if (err == 0) {
2590                         err = -ENOBUFS;
2591                         dev = __dev_get_by_name(net, p.name);
2592                         if (!dev)
2593                                 goto err_exit;
2594                         err = dev_open(dev);
2595                 }
2596         }
2597 #endif
2598
2599 err_exit:
2600         rtnl_unlock();
2601         return err;
2602 }
2603
2604 static int ipv6_mc_config(struct sock *sk, bool join,
2605                           const struct in6_addr *addr, int ifindex)
2606 {
2607         int ret;
2608
2609         ASSERT_RTNL();
2610
2611         lock_sock(sk);
2612         if (join)
2613                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2614         else
2615                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2616         release_sock(sk);
2617
2618         return ret;
2619 }
2620
2621 /*
2622  *      Manual configuration of address on an interface
2623  */
2624 static int inet6_addr_add(struct net *net, int ifindex,
2625                           const struct in6_addr *pfx,
2626                           const struct in6_addr *peer_pfx,
2627                           unsigned int plen, __u32 ifa_flags,
2628                           __u32 prefered_lft, __u32 valid_lft)
2629 {
2630         struct inet6_ifaddr *ifp;
2631         struct inet6_dev *idev;
2632         struct net_device *dev;
2633         unsigned long timeout;
2634         clock_t expires;
2635         int scope;
2636         u32 flags;
2637
2638         ASSERT_RTNL();
2639
2640         if (plen > 128)
2641                 return -EINVAL;
2642
2643         /* check the lifetime */
2644         if (!valid_lft || prefered_lft > valid_lft)
2645                 return -EINVAL;
2646
2647         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2648                 return -EINVAL;
2649
2650         dev = __dev_get_by_index(net, ifindex);
2651         if (!dev)
2652                 return -ENODEV;
2653
2654         idev = addrconf_add_dev(dev);
2655         if (IS_ERR(idev))
2656                 return PTR_ERR(idev);
2657
2658         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2659                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2660                                          true, pfx, ifindex);
2661
2662                 if (ret < 0)
2663                         return ret;
2664         }
2665
2666         scope = ipv6_addr_scope(pfx);
2667
2668         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2669         if (addrconf_finite_timeout(timeout)) {
2670                 expires = jiffies_to_clock_t(timeout * HZ);
2671                 valid_lft = timeout;
2672                 flags = RTF_EXPIRES;
2673         } else {
2674                 expires = 0;
2675                 flags = 0;
2676                 ifa_flags |= IFA_F_PERMANENT;
2677         }
2678
2679         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2680         if (addrconf_finite_timeout(timeout)) {
2681                 if (timeout == 0)
2682                         ifa_flags |= IFA_F_DEPRECATED;
2683                 prefered_lft = timeout;
2684         }
2685
2686         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2687                             valid_lft, prefered_lft);
2688
2689         if (!IS_ERR(ifp)) {
2690                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2691                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2692                                               expires, flags);
2693                 }
2694
2695                 /*
2696                  * Note that section 3.1 of RFC 4429 indicates
2697                  * that the Optimistic flag should not be set for
2698                  * manually configured addresses
2699                  */
2700                 addrconf_dad_start(ifp);
2701                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2702                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2703                                          true, jiffies);
2704                 in6_ifa_put(ifp);
2705                 addrconf_verify_rtnl();
2706                 return 0;
2707         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2708                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2709                                false, pfx, ifindex);
2710         }
2711
2712         return PTR_ERR(ifp);
2713 }
2714
2715 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2716                           const struct in6_addr *pfx, unsigned int plen)
2717 {
2718         struct inet6_ifaddr *ifp;
2719         struct inet6_dev *idev;
2720         struct net_device *dev;
2721
2722         if (plen > 128)
2723                 return -EINVAL;
2724
2725         dev = __dev_get_by_index(net, ifindex);
2726         if (!dev)
2727                 return -ENODEV;
2728
2729         idev = __in6_dev_get(dev);
2730         if (!idev)
2731                 return -ENXIO;
2732
2733         read_lock_bh(&idev->lock);
2734         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2735                 if (ifp->prefix_len == plen &&
2736                     ipv6_addr_equal(pfx, &ifp->addr)) {
2737                         in6_ifa_hold(ifp);
2738                         read_unlock_bh(&idev->lock);
2739
2740                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2741                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2742                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2743                                                  jiffies);
2744                         ipv6_del_addr(ifp);
2745                         addrconf_verify_rtnl();
2746                         if (ipv6_addr_is_multicast(pfx)) {
2747                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2748                                                false, pfx, dev->ifindex);
2749                         }
2750                         return 0;
2751                 }
2752         }
2753         read_unlock_bh(&idev->lock);
2754         return -EADDRNOTAVAIL;
2755 }
2756
2757
2758 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2759 {
2760         struct in6_ifreq ireq;
2761         int err;
2762
2763         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2764                 return -EPERM;
2765
2766         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2767                 return -EFAULT;
2768
2769         rtnl_lock();
2770         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2771                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2772                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2773         rtnl_unlock();
2774         return err;
2775 }
2776
2777 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2778 {
2779         struct in6_ifreq ireq;
2780         int err;
2781
2782         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2783                 return -EPERM;
2784
2785         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2786                 return -EFAULT;
2787
2788         rtnl_lock();
2789         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2790                              ireq.ifr6_prefixlen);
2791         rtnl_unlock();
2792         return err;
2793 }
2794
2795 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2796                      int plen, int scope)
2797 {
2798         struct inet6_ifaddr *ifp;
2799
2800         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2801                             scope, IFA_F_PERMANENT,
2802                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2803         if (!IS_ERR(ifp)) {
2804                 spin_lock_bh(&ifp->lock);
2805                 ifp->flags &= ~IFA_F_TENTATIVE;
2806                 spin_unlock_bh(&ifp->lock);
2807                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2808                 in6_ifa_put(ifp);
2809         }
2810 }
2811
2812 #if IS_ENABLED(CONFIG_IPV6_SIT)
2813 static void sit_add_v4_addrs(struct inet6_dev *idev)
2814 {
2815         struct in6_addr addr;
2816         struct net_device *dev;
2817         struct net *net = dev_net(idev->dev);
2818         int scope, plen;
2819         u32 pflags = 0;
2820
2821         ASSERT_RTNL();
2822
2823         memset(&addr, 0, sizeof(struct in6_addr));
2824         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2825
2826         if (idev->dev->flags&IFF_POINTOPOINT) {
2827                 addr.s6_addr32[0] = htonl(0xfe800000);
2828                 scope = IFA_LINK;
2829                 plen = 64;
2830         } else {
2831                 scope = IPV6_ADDR_COMPATv4;
2832                 plen = 96;
2833                 pflags |= RTF_NONEXTHOP;
2834         }
2835
2836         if (addr.s6_addr32[3]) {
2837                 add_addr(idev, &addr, plen, scope);
2838                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2839                 return;
2840         }
2841
2842         for_each_netdev(net, dev) {
2843                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2844                 if (in_dev && (dev->flags & IFF_UP)) {
2845                         struct in_ifaddr *ifa;
2846
2847                         int flag = scope;
2848
2849                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2850
2851                                 addr.s6_addr32[3] = ifa->ifa_local;
2852
2853                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2854                                         continue;
2855                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2856                                         if (idev->dev->flags&IFF_POINTOPOINT)
2857                                                 continue;
2858                                         flag |= IFA_HOST;
2859                                 }
2860
2861                                 add_addr(idev, &addr, plen, flag);
2862                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2863                                                       pflags);
2864                         }
2865                 }
2866         }
2867 }
2868 #endif
2869
2870 static void init_loopback(struct net_device *dev)
2871 {
2872         struct inet6_dev  *idev;
2873         struct net_device *sp_dev;
2874         struct inet6_ifaddr *sp_ifa;
2875         struct rt6_info *sp_rt;
2876
2877         /* ::1 */
2878
2879         ASSERT_RTNL();
2880
2881         idev = ipv6_find_idev(dev);
2882         if (!idev) {
2883                 pr_debug("%s: add_dev failed\n", __func__);
2884                 return;
2885         }
2886
2887         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2888
2889         /* Add routes to other interface's IPv6 addresses */
2890         for_each_netdev(dev_net(dev), sp_dev) {
2891                 if (!strcmp(sp_dev->name, dev->name))
2892                         continue;
2893
2894                 idev = __in6_dev_get(sp_dev);
2895                 if (!idev)
2896                         continue;
2897
2898                 read_lock_bh(&idev->lock);
2899                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2900
2901                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2902                                 continue;
2903
2904                         if (sp_ifa->rt) {
2905                                 /* This dst has been added to garbage list when
2906                                  * lo device down, release this obsolete dst and
2907                                  * reallocate a new router for ifa.
2908                                  */
2909                                 if (sp_ifa->rt->dst.obsolete > 0) {
2910                                         ip6_rt_put(sp_ifa->rt);
2911                                         sp_ifa->rt = NULL;
2912                                 } else {
2913                                         continue;
2914                                 }
2915                         }
2916
2917                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2918
2919                         /* Failure cases are ignored */
2920                         if (!IS_ERR(sp_rt)) {
2921                                 sp_ifa->rt = sp_rt;
2922                                 ip6_ins_rt(sp_rt);
2923                         }
2924                 }
2925                 read_unlock_bh(&idev->lock);
2926         }
2927 }
2928
2929 static void addrconf_add_linklocal(struct inet6_dev *idev,
2930                                    const struct in6_addr *addr, u32 flags)
2931 {
2932         struct inet6_ifaddr *ifp;
2933         u32 addr_flags = flags | IFA_F_PERMANENT;
2934
2935 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2936         if (idev->cnf.optimistic_dad &&
2937             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2938                 addr_flags |= IFA_F_OPTIMISTIC;
2939 #endif
2940
2941         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2942                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2943         if (!IS_ERR(ifp)) {
2944                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2945                 addrconf_dad_start(ifp);
2946                 in6_ifa_put(ifp);
2947         }
2948 }
2949
2950 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2951 {
2952         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2953                 return true;
2954
2955         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2956             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2957                 return true;
2958
2959         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2960             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2961                 return true;
2962
2963         return false;
2964 }
2965
2966 static int ipv6_generate_stable_address(struct in6_addr *address,
2967                                         u8 dad_count,
2968                                         const struct inet6_dev *idev)
2969 {
2970         static DEFINE_SPINLOCK(lock);
2971         static __u32 digest[SHA_DIGEST_WORDS];
2972         static __u32 workspace[SHA_WORKSPACE_WORDS];
2973
2974         static union {
2975                 char __data[SHA_MESSAGE_BYTES];
2976                 struct {
2977                         struct in6_addr secret;
2978                         __be32 prefix[2];
2979                         unsigned char hwaddr[MAX_ADDR_LEN];
2980                         u8 dad_count;
2981                 } __packed;
2982         } data;
2983
2984         struct in6_addr secret;
2985         struct in6_addr temp;
2986         struct net *net = dev_net(idev->dev);
2987
2988         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2989
2990         if (idev->cnf.stable_secret.initialized)
2991                 secret = idev->cnf.stable_secret.secret;
2992         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
2993                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
2994         else
2995                 return -1;
2996
2997 retry:
2998         spin_lock_bh(&lock);
2999
3000         sha_init(digest);
3001         memset(&data, 0, sizeof(data));
3002         memset(workspace, 0, sizeof(workspace));
3003         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3004         data.prefix[0] = address->s6_addr32[0];
3005         data.prefix[1] = address->s6_addr32[1];
3006         data.secret = secret;
3007         data.dad_count = dad_count;
3008
3009         sha_transform(digest, data.__data, workspace);
3010
3011         temp = *address;
3012         temp.s6_addr32[2] = (__force __be32)digest[0];
3013         temp.s6_addr32[3] = (__force __be32)digest[1];
3014
3015         spin_unlock_bh(&lock);
3016
3017         if (ipv6_reserved_interfaceid(temp)) {
3018                 dad_count++;
3019                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3020                         return -1;
3021                 goto retry;
3022         }
3023
3024         *address = temp;
3025         return 0;
3026 }
3027
3028 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3029 {
3030         struct in6_addr addr;
3031
3032         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3033
3034         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3035                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3036                         addrconf_add_linklocal(idev, &addr,
3037                                                IFA_F_STABLE_PRIVACY);
3038                 else if (prefix_route)
3039                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3040         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3041                 /* addrconf_add_linklocal also adds a prefix_route and we
3042                  * only need to care about prefix routes if ipv6_generate_eui64
3043                  * couldn't generate one.
3044                  */
3045                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3046                         addrconf_add_linklocal(idev, &addr, 0);
3047                 else if (prefix_route)
3048                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3049         }
3050 }
3051
3052 static void addrconf_dev_config(struct net_device *dev)
3053 {
3054         struct inet6_dev *idev;
3055
3056         ASSERT_RTNL();
3057
3058         if ((dev->type != ARPHRD_ETHER) &&
3059             (dev->type != ARPHRD_FDDI) &&
3060             (dev->type != ARPHRD_ARCNET) &&
3061             (dev->type != ARPHRD_INFINIBAND) &&
3062             (dev->type != ARPHRD_IEEE802154) &&
3063             (dev->type != ARPHRD_IEEE1394) &&
3064             (dev->type != ARPHRD_TUNNEL6) &&
3065             (dev->type != ARPHRD_6LOWPAN)) {
3066                 /* Alas, we support only Ethernet autoconfiguration. */
3067                 return;
3068         }
3069
3070         idev = addrconf_add_dev(dev);
3071         if (IS_ERR(idev))
3072                 return;
3073
3074         addrconf_addr_gen(idev, false);
3075 }
3076
3077 #if IS_ENABLED(CONFIG_IPV6_SIT)
3078 static void addrconf_sit_config(struct net_device *dev)
3079 {
3080         struct inet6_dev *idev;
3081
3082         ASSERT_RTNL();
3083
3084         /*
3085          * Configure the tunnel with one of our IPv4
3086          * addresses... we should configure all of
3087          * our v4 addrs in the tunnel
3088          */
3089
3090         idev = ipv6_find_idev(dev);
3091         if (!idev) {
3092                 pr_debug("%s: add_dev failed\n", __func__);
3093                 return;
3094         }
3095
3096         if (dev->priv_flags & IFF_ISATAP) {
3097                 addrconf_addr_gen(idev, false);
3098                 return;
3099         }
3100
3101         sit_add_v4_addrs(idev);
3102
3103         if (dev->flags&IFF_POINTOPOINT)
3104                 addrconf_add_mroute(dev);
3105 }
3106 #endif
3107
3108 #if IS_ENABLED(CONFIG_NET_IPGRE)
3109 static void addrconf_gre_config(struct net_device *dev)
3110 {
3111         struct inet6_dev *idev;
3112
3113         ASSERT_RTNL();
3114
3115         idev = ipv6_find_idev(dev);
3116         if (!idev) {
3117                 pr_debug("%s: add_dev failed\n", __func__);
3118                 return;
3119         }
3120
3121         addrconf_addr_gen(idev, true);
3122 }
3123 #endif
3124
3125 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3126                            void *ptr)
3127 {
3128         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3129         struct inet6_dev *idev = __in6_dev_get(dev);
3130         int run_pending = 0;
3131         int err;
3132
3133         switch (event) {
3134         case NETDEV_REGISTER:
3135                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3136                         idev = ipv6_add_dev(dev);
3137                         if (IS_ERR(idev))
3138                                 return notifier_from_errno(PTR_ERR(idev));
3139                 }
3140                 break;
3141
3142         case NETDEV_UP:
3143         case NETDEV_CHANGE:
3144                 if (dev->flags & IFF_SLAVE)
3145                         break;
3146
3147                 if (idev && idev->cnf.disable_ipv6)
3148                         break;
3149
3150                 if (event == NETDEV_UP) {
3151                         if (!addrconf_qdisc_ok(dev)) {
3152                                 /* device is not ready yet. */
3153                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3154                                         dev->name);
3155                                 break;
3156                         }
3157
3158                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3159                                 idev = ipv6_add_dev(dev);
3160
3161                         if (!IS_ERR_OR_NULL(idev)) {
3162                                 idev->if_flags |= IF_READY;
3163                                 run_pending = 1;
3164                         }
3165                 } else {
3166                         if (!addrconf_qdisc_ok(dev)) {
3167                                 /* device is still not ready. */
3168                                 break;
3169                         }
3170
3171                         if (idev) {
3172                                 if (idev->if_flags & IF_READY)
3173                                         /* device is already configured. */
3174                                         break;
3175                                 idev->if_flags |= IF_READY;
3176                         }
3177
3178                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3179                                 dev->name);
3180
3181                         run_pending = 1;
3182                 }
3183
3184                 switch (dev->type) {
3185 #if IS_ENABLED(CONFIG_IPV6_SIT)
3186                 case ARPHRD_SIT:
3187                         addrconf_sit_config(dev);
3188                         break;
3189 #endif
3190 #if IS_ENABLED(CONFIG_NET_IPGRE)
3191                 case ARPHRD_IPGRE:
3192                         addrconf_gre_config(dev);
3193                         break;
3194 #endif
3195                 case ARPHRD_LOOPBACK:
3196                         init_loopback(dev);
3197                         break;
3198
3199                 default:
3200                         addrconf_dev_config(dev);
3201                         break;
3202                 }
3203
3204                 if (!IS_ERR_OR_NULL(idev)) {
3205                         if (run_pending)
3206                                 addrconf_dad_run(idev);
3207
3208                         /*
3209                          * If the MTU changed during the interface down,
3210                          * when the interface up, the changed MTU must be
3211                          * reflected in the idev as well as routers.
3212                          */
3213                         if (idev->cnf.mtu6 != dev->mtu &&
3214                             dev->mtu >= IPV6_MIN_MTU) {
3215                                 rt6_mtu_change(dev, dev->mtu);
3216                                 idev->cnf.mtu6 = dev->mtu;
3217                         }
3218                         idev->tstamp = jiffies;
3219                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3220
3221                         /*
3222                          * If the changed mtu during down is lower than
3223                          * IPV6_MIN_MTU stop IPv6 on this interface.
3224                          */
3225                         if (dev->mtu < IPV6_MIN_MTU)
3226                                 addrconf_ifdown(dev, 1);
3227                 }
3228                 break;
3229
3230         case NETDEV_CHANGEMTU:
3231                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3232                         rt6_mtu_change(dev, dev->mtu);
3233                         idev->cnf.mtu6 = dev->mtu;
3234                         break;
3235                 }
3236
3237                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3238                         idev = ipv6_add_dev(dev);
3239                         if (!IS_ERR(idev))
3240                                 break;
3241                 }
3242
3243                 /*
3244                  * if MTU under IPV6_MIN_MTU.
3245                  * Stop IPv6 on this interface.
3246                  */
3247
3248         case NETDEV_DOWN:
3249         case NETDEV_UNREGISTER:
3250                 /*
3251                  *      Remove all addresses from this interface.
3252                  */
3253                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3254                 break;
3255
3256         case NETDEV_CHANGENAME:
3257                 if (idev) {
3258                         snmp6_unregister_dev(idev);
3259                         addrconf_sysctl_unregister(idev);
3260                         err = addrconf_sysctl_register(idev);
3261                         if (err)
3262                                 return notifier_from_errno(err);
3263                         err = snmp6_register_dev(idev);
3264                         if (err) {
3265                                 addrconf_sysctl_unregister(idev);
3266                                 return notifier_from_errno(err);
3267                         }
3268                 }
3269                 break;
3270
3271         case NETDEV_PRE_TYPE_CHANGE:
3272         case NETDEV_POST_TYPE_CHANGE:
3273                 addrconf_type_change(dev, event);
3274                 break;
3275         }
3276
3277         return NOTIFY_OK;
3278 }
3279
3280 /*
3281  *      addrconf module should be notified of a device going up
3282  */
3283 static struct notifier_block ipv6_dev_notf = {
3284         .notifier_call = addrconf_notify,
3285 };
3286
3287 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3288 {
3289         struct inet6_dev *idev;
3290         ASSERT_RTNL();
3291
3292         idev = __in6_dev_get(dev);
3293
3294         if (event == NETDEV_POST_TYPE_CHANGE)
3295                 ipv6_mc_remap(idev);
3296         else if (event == NETDEV_PRE_TYPE_CHANGE)
3297                 ipv6_mc_unmap(idev);
3298 }
3299
3300 static int addrconf_ifdown(struct net_device *dev, int how)
3301 {
3302         struct net *net = dev_net(dev);
3303         struct inet6_dev *idev;
3304         struct inet6_ifaddr *ifa;
3305         int state, i;
3306
3307         ASSERT_RTNL();
3308
3309         rt6_ifdown(net, dev);
3310         neigh_ifdown(&nd_tbl, dev);
3311
3312         idev = __in6_dev_get(dev);
3313         if (!idev)
3314                 return -ENODEV;
3315
3316         /*
3317          * Step 1: remove reference to ipv6 device from parent device.
3318          *         Do not dev_put!
3319          */
3320         if (how) {
3321                 idev->dead = 1;
3322
3323                 /* protected by rtnl_lock */
3324                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3325
3326                 /* Step 1.5: remove snmp6 entry */
3327                 snmp6_unregister_dev(idev);
3328
3329         }
3330
3331         /* Step 2: clear hash table */
3332         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3333                 struct hlist_head *h = &inet6_addr_lst[i];
3334
3335                 spin_lock_bh(&addrconf_hash_lock);
3336 restart:
3337                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3338                         if (ifa->idev == idev) {
3339                                 hlist_del_init_rcu(&ifa->addr_lst);
3340                                 addrconf_del_dad_work(ifa);
3341                                 goto restart;
3342                         }
3343                 }
3344                 spin_unlock_bh(&addrconf_hash_lock);
3345         }
3346
3347         write_lock_bh(&idev->lock);
3348
3349         addrconf_del_rs_timer(idev);
3350
3351         /* Step 2: clear flags for stateless addrconf */
3352         if (!how)
3353                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3354
3355         if (how && del_timer(&idev->regen_timer))
3356                 in6_dev_put(idev);
3357
3358         /* Step 3: clear tempaddr list */
3359         while (!list_empty(&idev->tempaddr_list)) {
3360                 ifa = list_first_entry(&idev->tempaddr_list,
3361                                        struct inet6_ifaddr, tmp_list);
3362                 list_del(&ifa->tmp_list);
3363                 write_unlock_bh(&idev->lock);
3364                 spin_lock_bh(&ifa->lock);
3365
3366                 if (ifa->ifpub) {
3367                         in6_ifa_put(ifa->ifpub);
3368                         ifa->ifpub = NULL;
3369                 }
3370                 spin_unlock_bh(&ifa->lock);
3371                 in6_ifa_put(ifa);
3372                 write_lock_bh(&idev->lock);
3373         }
3374
3375         while (!list_empty(&idev->addr_list)) {
3376                 ifa = list_first_entry(&idev->addr_list,
3377                                        struct inet6_ifaddr, if_list);
3378                 addrconf_del_dad_work(ifa);
3379
3380                 list_del(&ifa->if_list);
3381
3382                 write_unlock_bh(&idev->lock);
3383
3384                 spin_lock_bh(&ifa->lock);
3385                 state = ifa->state;
3386                 ifa->state = INET6_IFADDR_STATE_DEAD;
3387                 spin_unlock_bh(&ifa->lock);
3388
3389                 if (state != INET6_IFADDR_STATE_DEAD) {
3390                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3391                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3392                 }
3393                 in6_ifa_put(ifa);
3394
3395                 write_lock_bh(&idev->lock);
3396         }
3397
3398         write_unlock_bh(&idev->lock);
3399
3400         /* Step 5: Discard anycast and multicast list */
3401         if (how) {
3402                 ipv6_ac_destroy_dev(idev);
3403                 ipv6_mc_destroy_dev(idev);
3404         } else {
3405                 ipv6_mc_down(idev);
3406         }
3407
3408         idev->tstamp = jiffies;
3409
3410         /* Last: Shot the device (if unregistered) */
3411         if (how) {
3412                 addrconf_sysctl_unregister(idev);
3413                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3414                 neigh_ifdown(&nd_tbl, dev);
3415                 in6_dev_put(idev);
3416         }
3417         return 0;
3418 }
3419
3420 static void addrconf_rs_timer(unsigned long data)
3421 {
3422         struct inet6_dev *idev = (struct inet6_dev *)data;
3423         struct net_device *dev = idev->dev;
3424         struct in6_addr lladdr;
3425
3426         write_lock(&idev->lock);
3427         if (idev->dead || !(idev->if_flags & IF_READY))
3428                 goto out;
3429
3430         if (!ipv6_accept_ra(idev))
3431                 goto out;
3432
3433         /* Announcement received after solicitation was sent */
3434         if (idev->if_flags & IF_RA_RCVD)
3435                 goto out;
3436
3437         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3438                 write_unlock(&idev->lock);
3439                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3440                         ndisc_send_rs(dev, &lladdr,
3441                                       &in6addr_linklocal_allrouters);
3442                 else
3443                         goto put;
3444
3445                 write_lock(&idev->lock);
3446                 /* The wait after the last probe can be shorter */
3447                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3448                                              idev->cnf.rtr_solicits) ?
3449                                       idev->cnf.rtr_solicit_delay :
3450                                       idev->cnf.rtr_solicit_interval);
3451         } else {
3452                 /*
3453                  * Note: we do not support deprecated "all on-link"
3454                  * assumption any longer.
3455                  */
3456                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3457         }
3458
3459 out:
3460         write_unlock(&idev->lock);
3461 put:
3462         in6_dev_put(idev);
3463 }
3464
3465 /*
3466  *      Duplicate Address Detection
3467  */
3468 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3469 {
3470         unsigned long rand_num;
3471         struct inet6_dev *idev = ifp->idev;
3472
3473         if (ifp->flags & IFA_F_OPTIMISTIC)
3474                 rand_num = 0;
3475         else
3476                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3477
3478         ifp->dad_probes = idev->cnf.dad_transmits;
3479         addrconf_mod_dad_work(ifp, rand_num);
3480 }
3481
3482 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3483 {
3484         struct inet6_dev *idev = ifp->idev;
3485         struct net_device *dev = idev->dev;
3486
3487         addrconf_join_solict(dev, &ifp->addr);
3488
3489         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3490
3491         read_lock_bh(&idev->lock);
3492         spin_lock(&ifp->lock);
3493         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3494                 goto out;
3495
3496         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3497             idev->cnf.accept_dad < 1 ||
3498             !(ifp->flags&IFA_F_TENTATIVE) ||
3499             ifp->flags & IFA_F_NODAD) {
3500                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3501                 spin_unlock(&ifp->lock);
3502                 read_unlock_bh(&idev->lock);
3503
3504                 addrconf_dad_completed(ifp);
3505                 return;
3506         }
3507
3508         if (!(idev->if_flags & IF_READY)) {
3509                 spin_unlock(&ifp->lock);
3510                 read_unlock_bh(&idev->lock);
3511                 /*
3512                  * If the device is not ready:
3513                  * - keep it tentative if it is a permanent address.
3514                  * - otherwise, kill it.
3515                  */
3516                 in6_ifa_hold(ifp);
3517                 addrconf_dad_stop(ifp, 0);
3518                 return;
3519         }
3520
3521         /*
3522          * Optimistic nodes can start receiving
3523          * Frames right away
3524          */
3525         if (ifp->flags & IFA_F_OPTIMISTIC) {
3526                 ip6_ins_rt(ifp->rt);
3527                 if (ipv6_use_optimistic_addr(idev)) {
3528                         /* Because optimistic nodes can use this address,
3529                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3530                          */
3531                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3532                 }
3533         }
3534
3535         addrconf_dad_kick(ifp);
3536 out:
3537         spin_unlock(&ifp->lock);
3538         read_unlock_bh(&idev->lock);
3539 }
3540
3541 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3542 {
3543         bool begin_dad = false;
3544
3545         spin_lock_bh(&ifp->lock);
3546         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3547                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3548                 begin_dad = true;
3549         }
3550         spin_unlock_bh(&ifp->lock);
3551
3552         if (begin_dad)
3553                 addrconf_mod_dad_work(ifp, 0);
3554 }
3555
3556 static void addrconf_dad_work(struct work_struct *w)
3557 {
3558         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3559                                                 struct inet6_ifaddr,
3560                                                 dad_work);
3561         struct inet6_dev *idev = ifp->idev;
3562         struct in6_addr mcaddr;
3563
3564         enum {
3565                 DAD_PROCESS,
3566                 DAD_BEGIN,
3567                 DAD_ABORT,
3568         } action = DAD_PROCESS;
3569
3570         rtnl_lock();
3571
3572         spin_lock_bh(&ifp->lock);
3573         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3574                 action = DAD_BEGIN;
3575                 ifp->state = INET6_IFADDR_STATE_DAD;
3576         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3577                 action = DAD_ABORT;
3578                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3579         }
3580         spin_unlock_bh(&ifp->lock);
3581
3582         if (action == DAD_BEGIN) {
3583                 addrconf_dad_begin(ifp);
3584                 goto out;
3585         } else if (action == DAD_ABORT) {
3586                 addrconf_dad_stop(ifp, 1);
3587                 goto out;
3588         }
3589
3590         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3591                 goto out;
3592
3593         write_lock_bh(&idev->lock);
3594         if (idev->dead || !(idev->if_flags & IF_READY)) {
3595                 write_unlock_bh(&idev->lock);
3596                 goto out;
3597         }
3598
3599         spin_lock(&ifp->lock);
3600         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3601                 spin_unlock(&ifp->lock);
3602                 write_unlock_bh(&idev->lock);
3603                 goto out;
3604         }
3605
3606         if (ifp->dad_probes == 0) {
3607                 /*
3608                  * DAD was successful
3609                  */
3610
3611                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3612                 spin_unlock(&ifp->lock);
3613                 write_unlock_bh(&idev->lock);
3614
3615                 addrconf_dad_completed(ifp);
3616
3617                 goto out;
3618         }
3619
3620         ifp->dad_probes--;
3621         addrconf_mod_dad_work(ifp,
3622                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3623         spin_unlock(&ifp->lock);
3624         write_unlock_bh(&idev->lock);
3625
3626         /* send a neighbour solicitation for our addr */
3627         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3628         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any, NULL);
3629 out:
3630         in6_ifa_put(ifp);
3631         rtnl_unlock();
3632 }
3633
3634 /* ifp->idev must be at least read locked */
3635 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3636 {
3637         struct inet6_ifaddr *ifpiter;
3638         struct inet6_dev *idev = ifp->idev;
3639
3640         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3641                 if (ifpiter->scope > IFA_LINK)
3642                         break;
3643                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3644                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3645                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3646                     IFA_F_PERMANENT)
3647                         return false;
3648         }
3649         return true;
3650 }
3651
3652 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3653 {
3654         struct net_device *dev = ifp->idev->dev;
3655         struct in6_addr lladdr;
3656         bool send_rs, send_mld;
3657
3658         addrconf_del_dad_work(ifp);
3659
3660         /*
3661          *      Configure the address for reception. Now it is valid.
3662          */
3663
3664         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3665
3666         /* If added prefix is link local and we are prepared to process
3667            router advertisements, start sending router solicitations.
3668          */
3669
3670         read_lock_bh(&ifp->idev->lock);
3671         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3672         send_rs = send_mld &&
3673                   ipv6_accept_ra(ifp->idev) &&
3674                   ifp->idev->cnf.rtr_solicits > 0 &&
3675                   (dev->flags&IFF_LOOPBACK) == 0;
3676         read_unlock_bh(&ifp->idev->lock);
3677
3678         /* While dad is in progress mld report's source address is in6_addrany.
3679          * Resend with proper ll now.
3680          */
3681         if (send_mld)
3682                 ipv6_mc_dad_complete(ifp->idev);
3683
3684         if (send_rs) {
3685                 /*
3686                  *      If a host as already performed a random delay
3687                  *      [...] as part of DAD [...] there is no need
3688                  *      to delay again before sending the first RS
3689                  */
3690                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3691                         return;
3692                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3693
3694                 write_lock_bh(&ifp->idev->lock);
3695                 spin_lock(&ifp->lock);
3696                 ifp->idev->rs_probes = 1;
3697                 ifp->idev->if_flags |= IF_RS_SENT;
3698                 addrconf_mod_rs_timer(ifp->idev,
3699                                       ifp->idev->cnf.rtr_solicit_interval);
3700                 spin_unlock(&ifp->lock);
3701                 write_unlock_bh(&ifp->idev->lock);
3702         }
3703 }
3704
3705 static void addrconf_dad_run(struct inet6_dev *idev)
3706 {
3707         struct inet6_ifaddr *ifp;
3708
3709         read_lock_bh(&idev->lock);
3710         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3711                 spin_lock(&ifp->lock);
3712                 if (ifp->flags & IFA_F_TENTATIVE &&
3713                     ifp->state == INET6_IFADDR_STATE_DAD)
3714                         addrconf_dad_kick(ifp);
3715                 spin_unlock(&ifp->lock);
3716         }
3717         read_unlock_bh(&idev->lock);
3718 }
3719
3720 #ifdef CONFIG_PROC_FS
3721 struct if6_iter_state {
3722         struct seq_net_private p;
3723         int bucket;
3724         int offset;
3725 };
3726
3727 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3728 {
3729         struct inet6_ifaddr *ifa = NULL;
3730         struct if6_iter_state *state = seq->private;
3731         struct net *net = seq_file_net(seq);
3732         int p = 0;
3733
3734         /* initial bucket if pos is 0 */
3735         if (pos == 0) {
3736                 state->bucket = 0;
3737                 state->offset = 0;
3738         }
3739
3740         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3741                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3742                                          addr_lst) {
3743                         if (!net_eq(dev_net(ifa->idev->dev), net))
3744                                 continue;
3745                         /* sync with offset */
3746                         if (p < state->offset) {
3747                                 p++;
3748                                 continue;
3749                         }
3750                         state->offset++;
3751                         return ifa;
3752                 }
3753
3754                 /* prepare for next bucket */
3755                 state->offset = 0;
3756                 p = 0;
3757         }
3758         return NULL;
3759 }
3760
3761 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3762                                          struct inet6_ifaddr *ifa)
3763 {
3764         struct if6_iter_state *state = seq->private;
3765         struct net *net = seq_file_net(seq);
3766
3767         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3768                 if (!net_eq(dev_net(ifa->idev->dev), net))
3769                         continue;
3770                 state->offset++;
3771                 return ifa;
3772         }
3773
3774         while (++state->bucket < IN6_ADDR_HSIZE) {
3775                 state->offset = 0;
3776                 hlist_for_each_entry_rcu_bh(ifa,
3777                                      &inet6_addr_lst[state->bucket], addr_lst) {
3778                         if (!net_eq(dev_net(ifa->idev->dev), net))
3779                                 continue;
3780                         state->offset++;
3781                         return ifa;
3782                 }
3783         }
3784
3785         return NULL;
3786 }
3787
3788 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3789         __acquires(rcu_bh)
3790 {
3791         rcu_read_lock_bh();
3792         return if6_get_first(seq, *pos);
3793 }
3794
3795 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3796 {
3797         struct inet6_ifaddr *ifa;
3798
3799         ifa = if6_get_next(seq, v);
3800         ++*pos;
3801         return ifa;
3802 }
3803
3804 static void if6_seq_stop(struct seq_file *seq, void *v)
3805         __releases(rcu_bh)
3806 {
3807         rcu_read_unlock_bh();
3808 }
3809
3810 static int if6_seq_show(struct seq_file *seq, void *v)
3811 {
3812         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3813         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3814                    &ifp->addr,
3815                    ifp->idev->dev->ifindex,
3816                    ifp->prefix_len,
3817                    ifp->scope,
3818                    (u8) ifp->flags,
3819                    ifp->idev->dev->name);
3820         return 0;
3821 }
3822
3823 static const struct seq_operations if6_seq_ops = {
3824         .start  = if6_seq_start,
3825         .next   = if6_seq_next,
3826         .show   = if6_seq_show,
3827         .stop   = if6_seq_stop,
3828 };
3829
3830 static int if6_seq_open(struct inode *inode, struct file *file)
3831 {
3832         return seq_open_net(inode, file, &if6_seq_ops,
3833                             sizeof(struct if6_iter_state));
3834 }
3835
3836 static const struct file_operations if6_fops = {
3837         .owner          = THIS_MODULE,
3838         .open           = if6_seq_open,
3839         .read           = seq_read,
3840         .llseek         = seq_lseek,
3841         .release        = seq_release_net,
3842 };
3843
3844 static int __net_init if6_proc_net_init(struct net *net)
3845 {
3846         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3847                 return -ENOMEM;
3848         return 0;
3849 }
3850
3851 static void __net_exit if6_proc_net_exit(struct net *net)
3852 {
3853         remove_proc_entry("if_inet6", net->proc_net);
3854 }
3855
3856 static struct pernet_operations if6_proc_net_ops = {
3857         .init = if6_proc_net_init,
3858         .exit = if6_proc_net_exit,
3859 };
3860
3861 int __init if6_proc_init(void)
3862 {
3863         return register_pernet_subsys(&if6_proc_net_ops);
3864 }
3865
3866 void if6_proc_exit(void)
3867 {
3868         unregister_pernet_subsys(&if6_proc_net_ops);
3869 }
3870 #endif  /* CONFIG_PROC_FS */
3871
3872 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3873 /* Check if address is a home address configured on any interface. */
3874 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3875 {
3876         int ret = 0;
3877         struct inet6_ifaddr *ifp = NULL;
3878         unsigned int hash = inet6_addr_hash(addr);
3879
3880         rcu_read_lock_bh();
3881         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3882                 if (!net_eq(dev_net(ifp->idev->dev), net))
3883                         continue;
3884                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3885                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3886                         ret = 1;
3887                         break;
3888                 }
3889         }
3890         rcu_read_unlock_bh();
3891         return ret;
3892 }
3893 #endif
3894
3895 /*
3896  *      Periodic address status verification
3897  */
3898
3899 static void addrconf_verify_rtnl(void)
3900 {
3901         unsigned long now, next, next_sec, next_sched;
3902         struct inet6_ifaddr *ifp;
3903         int i;
3904
3905         ASSERT_RTNL();
3906
3907         rcu_read_lock_bh();
3908         now = jiffies;
3909         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3910
3911         cancel_delayed_work(&addr_chk_work);
3912
3913         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3914 restart:
3915                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3916                         unsigned long age;
3917
3918                         /* When setting preferred_lft to a value not zero or
3919                          * infinity, while valid_lft is infinity
3920                          * IFA_F_PERMANENT has a non-infinity life time.
3921                          */
3922                         if ((ifp->flags & IFA_F_PERMANENT) &&
3923                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3924                                 continue;
3925
3926                         spin_lock(&ifp->lock);
3927                         /* We try to batch several events at once. */
3928                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3929
3930                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3931                             age >= ifp->valid_lft) {
3932                                 spin_unlock(&ifp->lock);
3933                                 in6_ifa_hold(ifp);
3934                                 ipv6_del_addr(ifp);
3935                                 goto restart;
3936                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3937                                 spin_unlock(&ifp->lock);
3938                                 continue;
3939                         } else if (age >= ifp->prefered_lft) {
3940                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3941                                 int deprecate = 0;
3942
3943                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3944                                         deprecate = 1;
3945                                         ifp->flags |= IFA_F_DEPRECATED;
3946                                 }
3947
3948                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3949                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3950                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3951
3952                                 spin_unlock(&ifp->lock);
3953
3954                                 if (deprecate) {
3955                                         in6_ifa_hold(ifp);
3956
3957                                         ipv6_ifa_notify(0, ifp);
3958                                         in6_ifa_put(ifp);
3959                                         goto restart;
3960                                 }
3961                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3962                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3963                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3964                                         ifp->idev->cnf.dad_transmits *
3965                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3966
3967                                 if (age >= ifp->prefered_lft - regen_advance) {
3968                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3969                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3970                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3971                                         if (!ifp->regen_count && ifpub) {
3972                                                 ifp->regen_count++;
3973                                                 in6_ifa_hold(ifp);
3974                                                 in6_ifa_hold(ifpub);
3975                                                 spin_unlock(&ifp->lock);
3976
3977                                                 spin_lock(&ifpub->lock);
3978                                                 ifpub->regen_count = 0;
3979                                                 spin_unlock(&ifpub->lock);
3980                                                 ipv6_create_tempaddr(ifpub, ifp);
3981                                                 in6_ifa_put(ifpub);
3982                                                 in6_ifa_put(ifp);
3983                                                 goto restart;
3984                                         }
3985                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3986                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3987                                 spin_unlock(&ifp->lock);
3988                         } else {
3989                                 /* ifp->prefered_lft <= ifp->valid_lft */
3990                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3991                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3992                                 spin_unlock(&ifp->lock);
3993                         }
3994                 }
3995         }
3996
3997         next_sec = round_jiffies_up(next);
3998         next_sched = next;
3999
4000         /* If rounded timeout is accurate enough, accept it. */
4001         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4002                 next_sched = next_sec;
4003
4004         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4005         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4006                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4007
4008         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4009               now, next, next_sec, next_sched);
4010         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4011         rcu_read_unlock_bh();
4012 }
4013
4014 static void addrconf_verify_work(struct work_struct *w)
4015 {
4016         rtnl_lock();
4017         addrconf_verify_rtnl();
4018         rtnl_unlock();
4019 }
4020
4021 static void addrconf_verify(void)
4022 {
4023         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4024 }
4025
4026 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4027                                      struct in6_addr **peer_pfx)
4028 {
4029         struct in6_addr *pfx = NULL;
4030
4031         *peer_pfx = NULL;
4032
4033         if (addr)
4034                 pfx = nla_data(addr);
4035
4036         if (local) {
4037                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4038                         *peer_pfx = pfx;
4039                 pfx = nla_data(local);
4040         }
4041
4042         return pfx;
4043 }
4044
4045 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4046         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4047         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4048         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4049         [IFA_FLAGS]             = { .len = sizeof(u32) },
4050 };
4051
4052 static int
4053 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4054 {
4055         struct net *net = sock_net(skb->sk);
4056         struct ifaddrmsg *ifm;
4057         struct nlattr *tb[IFA_MAX+1];
4058         struct in6_addr *pfx, *peer_pfx;
4059         u32 ifa_flags;
4060         int err;
4061
4062         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4063         if (err < 0)
4064                 return err;
4065
4066         ifm = nlmsg_data(nlh);
4067         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4068         if (!pfx)
4069                 return -EINVAL;
4070
4071         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4072
4073         /* We ignore other flags so far. */
4074         ifa_flags &= IFA_F_MANAGETEMPADDR;
4075
4076         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4077                               ifm->ifa_prefixlen);
4078 }
4079
4080 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4081                              u32 prefered_lft, u32 valid_lft)
4082 {
4083         u32 flags;
4084         clock_t expires;
4085         unsigned long timeout;
4086         bool was_managetempaddr;
4087         bool had_prefixroute;
4088
4089         ASSERT_RTNL();
4090
4091         if (!valid_lft || (prefered_lft > valid_lft))
4092                 return -EINVAL;
4093
4094         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4095             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4096                 return -EINVAL;
4097
4098         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4099         if (addrconf_finite_timeout(timeout)) {
4100                 expires = jiffies_to_clock_t(timeout * HZ);
4101                 valid_lft = timeout;
4102                 flags = RTF_EXPIRES;
4103         } else {
4104                 expires = 0;
4105                 flags = 0;
4106                 ifa_flags |= IFA_F_PERMANENT;
4107         }
4108
4109         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4110         if (addrconf_finite_timeout(timeout)) {
4111                 if (timeout == 0)
4112                         ifa_flags |= IFA_F_DEPRECATED;
4113                 prefered_lft = timeout;
4114         }
4115
4116         spin_lock_bh(&ifp->lock);
4117         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4118         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4119                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4120         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4121                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4122                         IFA_F_NOPREFIXROUTE);
4123         ifp->flags |= ifa_flags;
4124         ifp->tstamp = jiffies;
4125         ifp->valid_lft = valid_lft;
4126         ifp->prefered_lft = prefered_lft;
4127
4128         spin_unlock_bh(&ifp->lock);
4129         if (!(ifp->flags&IFA_F_TENTATIVE))
4130                 ipv6_ifa_notify(0, ifp);
4131
4132         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4133                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4134                                       expires, flags);
4135         } else if (had_prefixroute) {
4136                 enum cleanup_prefix_rt_t action;
4137                 unsigned long rt_expires;
4138
4139                 write_lock_bh(&ifp->idev->lock);
4140                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4141                 write_unlock_bh(&ifp->idev->lock);
4142
4143                 if (action != CLEANUP_PREFIX_RT_NOP) {
4144                         cleanup_prefix_route(ifp, rt_expires,
4145                                 action == CLEANUP_PREFIX_RT_DEL);
4146                 }
4147         }
4148
4149         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4150                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4151                         valid_lft = prefered_lft = 0;
4152                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4153                                  !was_managetempaddr, jiffies);
4154         }
4155
4156         addrconf_verify_rtnl();
4157
4158         return 0;
4159 }
4160
4161 static int
4162 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4163 {
4164         struct net *net = sock_net(skb->sk);
4165         struct ifaddrmsg *ifm;
4166         struct nlattr *tb[IFA_MAX+1];
4167         struct in6_addr *pfx, *peer_pfx;
4168         struct inet6_ifaddr *ifa;
4169         struct net_device *dev;
4170         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4171         u32 ifa_flags;
4172         int err;
4173
4174         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4175         if (err < 0)
4176                 return err;
4177
4178         ifm = nlmsg_data(nlh);
4179         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4180         if (!pfx)
4181                 return -EINVAL;
4182
4183         if (tb[IFA_CACHEINFO]) {
4184                 struct ifa_cacheinfo *ci;
4185
4186                 ci = nla_data(tb[IFA_CACHEINFO]);
4187                 valid_lft = ci->ifa_valid;
4188                 preferred_lft = ci->ifa_prefered;
4189         } else {
4190                 preferred_lft = INFINITY_LIFE_TIME;
4191                 valid_lft = INFINITY_LIFE_TIME;
4192         }
4193
4194         dev =  __dev_get_by_index(net, ifm->ifa_index);
4195         if (!dev)
4196                 return -ENODEV;
4197
4198         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4199
4200         /* We ignore other flags so far. */
4201         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4202                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4203
4204         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4205         if (!ifa) {
4206                 /*
4207                  * It would be best to check for !NLM_F_CREATE here but
4208                  * userspace already relies on not having to provide this.
4209                  */
4210                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4211                                       ifm->ifa_prefixlen, ifa_flags,
4212                                       preferred_lft, valid_lft);
4213         }
4214
4215         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4216             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4217                 err = -EEXIST;
4218         else
4219                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4220
4221         in6_ifa_put(ifa);
4222
4223         return err;
4224 }
4225
4226 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4227                           u8 scope, int ifindex)
4228 {
4229         struct ifaddrmsg *ifm;
4230
4231         ifm = nlmsg_data(nlh);
4232         ifm->ifa_family = AF_INET6;
4233         ifm->ifa_prefixlen = prefixlen;
4234         ifm->ifa_flags = flags;
4235         ifm->ifa_scope = scope;
4236         ifm->ifa_index = ifindex;
4237 }
4238
4239 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4240                          unsigned long tstamp, u32 preferred, u32 valid)
4241 {
4242         struct ifa_cacheinfo ci;
4243
4244         ci.cstamp = cstamp_delta(cstamp);
4245         ci.tstamp = cstamp_delta(tstamp);
4246         ci.ifa_prefered = preferred;
4247         ci.ifa_valid = valid;
4248
4249         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4250 }
4251
4252 static inline int rt_scope(int ifa_scope)
4253 {
4254         if (ifa_scope & IFA_HOST)
4255                 return RT_SCOPE_HOST;
4256         else if (ifa_scope & IFA_LINK)
4257                 return RT_SCOPE_LINK;
4258         else if (ifa_scope & IFA_SITE)
4259                 return RT_SCOPE_SITE;
4260         else
4261                 return RT_SCOPE_UNIVERSE;
4262 }
4263
4264 static inline int inet6_ifaddr_msgsize(void)
4265 {
4266         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4267                + nla_total_size(16) /* IFA_LOCAL */
4268                + nla_total_size(16) /* IFA_ADDRESS */
4269                + nla_total_size(sizeof(struct ifa_cacheinfo))
4270                + nla_total_size(4)  /* IFA_FLAGS */;
4271 }
4272
4273 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4274                              u32 portid, u32 seq, int event, unsigned int flags)
4275 {
4276         struct nlmsghdr  *nlh;
4277         u32 preferred, valid;
4278
4279         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4280         if (!nlh)
4281                 return -EMSGSIZE;
4282
4283         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4284                       ifa->idev->dev->ifindex);
4285
4286         if (!((ifa->flags&IFA_F_PERMANENT) &&
4287               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4288                 preferred = ifa->prefered_lft;
4289                 valid = ifa->valid_lft;
4290                 if (preferred != INFINITY_LIFE_TIME) {
4291                         long tval = (jiffies - ifa->tstamp)/HZ;
4292                         if (preferred > tval)
4293                                 preferred -= tval;
4294                         else
4295                                 preferred = 0;
4296                         if (valid != INFINITY_LIFE_TIME) {
4297                                 if (valid > tval)
4298                                         valid -= tval;
4299                                 else
4300                                         valid = 0;
4301                         }
4302                 }
4303         } else {
4304                 preferred = INFINITY_LIFE_TIME;
4305                 valid = INFINITY_LIFE_TIME;
4306         }
4307
4308         if (!ipv6_addr_any(&ifa->peer_addr)) {
4309                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4310                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4311                         goto error;
4312         } else
4313                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4314                         goto error;
4315
4316         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4317                 goto error;
4318
4319         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4320                 goto error;
4321
4322         nlmsg_end(skb, nlh);
4323         return 0;
4324
4325 error:
4326         nlmsg_cancel(skb, nlh);
4327         return -EMSGSIZE;
4328 }
4329
4330 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4331                                 u32 portid, u32 seq, int event, u16 flags)
4332 {
4333         struct nlmsghdr  *nlh;
4334         u8 scope = RT_SCOPE_UNIVERSE;
4335         int ifindex = ifmca->idev->dev->ifindex;
4336
4337         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4338                 scope = RT_SCOPE_SITE;
4339
4340         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4341         if (!nlh)
4342                 return -EMSGSIZE;
4343
4344         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4345         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4346             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4347                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4348                 nlmsg_cancel(skb, nlh);
4349                 return -EMSGSIZE;
4350         }
4351
4352         nlmsg_end(skb, nlh);
4353         return 0;
4354 }
4355
4356 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4357                                 u32 portid, u32 seq, int event, unsigned int flags)
4358 {
4359         struct nlmsghdr  *nlh;
4360         u8 scope = RT_SCOPE_UNIVERSE;
4361         int ifindex = ifaca->aca_idev->dev->ifindex;
4362
4363         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4364                 scope = RT_SCOPE_SITE;
4365
4366         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4367         if (!nlh)
4368                 return -EMSGSIZE;
4369
4370         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4371         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4372             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4373                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4374                 nlmsg_cancel(skb, nlh);
4375                 return -EMSGSIZE;
4376         }
4377
4378         nlmsg_end(skb, nlh);
4379         return 0;
4380 }
4381
4382 enum addr_type_t {
4383         UNICAST_ADDR,
4384         MULTICAST_ADDR,
4385         ANYCAST_ADDR,
4386 };
4387
4388 /* called with rcu_read_lock() */
4389 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4390                           struct netlink_callback *cb, enum addr_type_t type,
4391                           int s_ip_idx, int *p_ip_idx)
4392 {
4393         struct ifmcaddr6 *ifmca;
4394         struct ifacaddr6 *ifaca;
4395         int err = 1;
4396         int ip_idx = *p_ip_idx;
4397
4398         read_lock_bh(&idev->lock);
4399         switch (type) {
4400         case UNICAST_ADDR: {
4401                 struct inet6_ifaddr *ifa;
4402
4403                 /* unicast address incl. temp addr */
4404                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4405                         if (++ip_idx < s_ip_idx)
4406                                 continue;
4407                         err = inet6_fill_ifaddr(skb, ifa,
4408                                                 NETLINK_CB(cb->skb).portid,
4409                                                 cb->nlh->nlmsg_seq,
4410                                                 RTM_NEWADDR,
4411                                                 NLM_F_MULTI);
4412                         if (err < 0)
4413                                 break;
4414                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4415                 }
4416                 break;
4417         }
4418         case MULTICAST_ADDR:
4419                 /* multicast address */
4420                 for (ifmca = idev->mc_list; ifmca;
4421                      ifmca = ifmca->next, ip_idx++) {
4422                         if (ip_idx < s_ip_idx)
4423                                 continue;
4424                         err = inet6_fill_ifmcaddr(skb, ifmca,
4425                                                   NETLINK_CB(cb->skb).portid,
4426                                                   cb->nlh->nlmsg_seq,
4427                                                   RTM_GETMULTICAST,
4428                                                   NLM_F_MULTI);
4429                         if (err < 0)
4430                                 break;
4431                 }
4432                 break;
4433         case ANYCAST_ADDR:
4434                 /* anycast address */
4435                 for (ifaca = idev->ac_list; ifaca;
4436                      ifaca = ifaca->aca_next, ip_idx++) {
4437                         if (ip_idx < s_ip_idx)
4438                                 continue;
4439                         err = inet6_fill_ifacaddr(skb, ifaca,
4440                                                   NETLINK_CB(cb->skb).portid,
4441                                                   cb->nlh->nlmsg_seq,
4442                                                   RTM_GETANYCAST,
4443                                                   NLM_F_MULTI);
4444                         if (err < 0)
4445                                 break;
4446                 }
4447                 break;
4448         default:
4449                 break;
4450         }
4451         read_unlock_bh(&idev->lock);
4452         *p_ip_idx = ip_idx;
4453         return err;
4454 }
4455
4456 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4457                            enum addr_type_t type)
4458 {
4459         struct net *net = sock_net(skb->sk);
4460         int h, s_h;
4461         int idx, ip_idx;
4462         int s_idx, s_ip_idx;
4463         struct net_device *dev;
4464         struct inet6_dev *idev;
4465         struct hlist_head *head;
4466
4467         s_h = cb->args[0];
4468         s_idx = idx = cb->args[1];
4469         s_ip_idx = ip_idx = cb->args[2];
4470
4471         rcu_read_lock();
4472         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4473         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4474                 idx = 0;
4475                 head = &net->dev_index_head[h];
4476                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4477                         if (idx < s_idx)
4478                                 goto cont;
4479                         if (h > s_h || idx > s_idx)
4480                                 s_ip_idx = 0;
4481                         ip_idx = 0;
4482                         idev = __in6_dev_get(dev);
4483                         if (!idev)
4484                                 goto cont;
4485
4486                         if (in6_dump_addrs(idev, skb, cb, type,
4487                                            s_ip_idx, &ip_idx) < 0)
4488                                 goto done;
4489 cont:
4490                         idx++;
4491                 }
4492         }
4493 done:
4494         rcu_read_unlock();
4495         cb->args[0] = h;
4496         cb->args[1] = idx;
4497         cb->args[2] = ip_idx;
4498
4499         return skb->len;
4500 }
4501
4502 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4503 {
4504         enum addr_type_t type = UNICAST_ADDR;
4505
4506         return inet6_dump_addr(skb, cb, type);
4507 }
4508
4509 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4510 {
4511         enum addr_type_t type = MULTICAST_ADDR;
4512
4513         return inet6_dump_addr(skb, cb, type);
4514 }
4515
4516
4517 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4518 {
4519         enum addr_type_t type = ANYCAST_ADDR;
4520
4521         return inet6_dump_addr(skb, cb, type);
4522 }
4523
4524 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4525 {
4526         struct net *net = sock_net(in_skb->sk);
4527         struct ifaddrmsg *ifm;
4528         struct nlattr *tb[IFA_MAX+1];
4529         struct in6_addr *addr = NULL, *peer;
4530         struct net_device *dev = NULL;
4531         struct inet6_ifaddr *ifa;
4532         struct sk_buff *skb;
4533         int err;
4534
4535         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4536         if (err < 0)
4537                 goto errout;
4538
4539         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4540         if (!addr) {
4541                 err = -EINVAL;
4542                 goto errout;
4543         }
4544
4545         ifm = nlmsg_data(nlh);
4546         if (ifm->ifa_index)
4547                 dev = __dev_get_by_index(net, ifm->ifa_index);
4548
4549         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4550         if (!ifa) {
4551                 err = -EADDRNOTAVAIL;
4552                 goto errout;
4553         }
4554
4555         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4556         if (!skb) {
4557                 err = -ENOBUFS;
4558                 goto errout_ifa;
4559         }
4560
4561         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4562                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4563         if (err < 0) {
4564                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4565                 WARN_ON(err == -EMSGSIZE);
4566                 kfree_skb(skb);
4567                 goto errout_ifa;
4568         }
4569         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4570 errout_ifa:
4571         in6_ifa_put(ifa);
4572 errout:
4573         return err;
4574 }
4575
4576 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4577 {
4578         struct sk_buff *skb;
4579         struct net *net = dev_net(ifa->idev->dev);
4580         int err = -ENOBUFS;
4581
4582         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4583         if (!skb)
4584                 goto errout;
4585
4586         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4587         if (err < 0) {
4588                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4589                 WARN_ON(err == -EMSGSIZE);
4590                 kfree_skb(skb);
4591                 goto errout;
4592         }
4593         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4594         return;
4595 errout:
4596         if (err < 0)
4597                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4598 }
4599
4600 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4601                                 __s32 *array, int bytes)
4602 {
4603         BUG_ON(bytes < (DEVCONF_MAX * 4));
4604
4605         memset(array, 0, bytes);
4606         array[DEVCONF_FORWARDING] = cnf->forwarding;
4607         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4608         array[DEVCONF_MTU6] = cnf->mtu6;
4609         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4610         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4611         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4612         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4613         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4614         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4615                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4616         array[DEVCONF_RTR_SOLICIT_DELAY] =
4617                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4618         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4619         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4620                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4621         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4622                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4623         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4624         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4625         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4626         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4627         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4628         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4629         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4630         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4631         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4632 #ifdef CONFIG_IPV6_ROUTER_PREF
4633         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4634         array[DEVCONF_RTR_PROBE_INTERVAL] =
4635                 jiffies_to_msecs(cnf->rtr_probe_interval);
4636 #ifdef CONFIG_IPV6_ROUTE_INFO
4637         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4638 #endif
4639 #endif
4640         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4641         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4642 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4643         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4644         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4645 #endif
4646 #ifdef CONFIG_IPV6_MROUTE
4647         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4648 #endif
4649         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4650         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4651         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4652         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4653         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4654         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4655         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4656         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4657         /* we omit DEVCONF_STABLE_SECRET for now */
4658         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4659 }
4660
4661 static inline size_t inet6_ifla6_size(void)
4662 {
4663         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4664              + nla_total_size(sizeof(struct ifla_cacheinfo))
4665              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4666              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4667              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4668              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4669 }
4670
4671 static inline size_t inet6_if_nlmsg_size(void)
4672 {
4673         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4674                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4675                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4676                + nla_total_size(4) /* IFLA_MTU */
4677                + nla_total_size(4) /* IFLA_LINK */
4678                + nla_total_size(1) /* IFLA_OPERSTATE */
4679                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4680 }
4681
4682 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4683                                       int items, int bytes)
4684 {
4685         int i;
4686         int pad = bytes - sizeof(u64) * items;
4687         BUG_ON(pad < 0);
4688
4689         /* Use put_unaligned() because stats may not be aligned for u64. */
4690         put_unaligned(items, &stats[0]);
4691         for (i = 1; i < items; i++)
4692                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4693
4694         memset(&stats[items], 0, pad);
4695 }
4696
4697 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4698                                         int bytes, size_t syncpoff)
4699 {
4700         int i, c;
4701         u64 buff[IPSTATS_MIB_MAX];
4702         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4703
4704         BUG_ON(pad < 0);
4705
4706         memset(buff, 0, sizeof(buff));
4707         buff[0] = IPSTATS_MIB_MAX;
4708
4709         for_each_possible_cpu(c) {
4710                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4711                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4712         }
4713
4714         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4715         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4716 }
4717
4718 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4719                              int bytes)
4720 {
4721         switch (attrtype) {
4722         case IFLA_INET6_STATS:
4723                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4724                                      offsetof(struct ipstats_mib, syncp));
4725                 break;
4726         case IFLA_INET6_ICMP6STATS:
4727                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4728                 break;
4729         }
4730 }
4731
4732 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4733 {
4734         struct nlattr *nla;
4735         struct ifla_cacheinfo ci;
4736
4737         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4738                 goto nla_put_failure;
4739         ci.max_reasm_len = IPV6_MAXPLEN;
4740         ci.tstamp = cstamp_delta(idev->tstamp);
4741         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4742         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4743         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4744                 goto nla_put_failure;
4745         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4746         if (!nla)
4747                 goto nla_put_failure;
4748         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4749
4750         /* XXX - MC not implemented */
4751
4752         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4753         if (!nla)
4754                 goto nla_put_failure;
4755         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4756
4757         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4758         if (!nla)
4759                 goto nla_put_failure;
4760         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4761
4762         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4763         if (!nla)
4764                 goto nla_put_failure;
4765
4766         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4767                 goto nla_put_failure;
4768
4769         read_lock_bh(&idev->lock);
4770         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4771         read_unlock_bh(&idev->lock);
4772
4773         return 0;
4774
4775 nla_put_failure:
4776         return -EMSGSIZE;
4777 }
4778
4779 static size_t inet6_get_link_af_size(const struct net_device *dev)
4780 {
4781         if (!__in6_dev_get(dev))
4782                 return 0;
4783
4784         return inet6_ifla6_size();
4785 }
4786
4787 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4788 {
4789         struct inet6_dev *idev = __in6_dev_get(dev);
4790
4791         if (!idev)
4792                 return -ENODATA;
4793
4794         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4795                 return -EMSGSIZE;
4796
4797         return 0;
4798 }
4799
4800 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4801 {
4802         struct inet6_ifaddr *ifp;
4803         struct net_device *dev = idev->dev;
4804         bool update_rs = false;
4805         struct in6_addr ll_addr;
4806
4807         ASSERT_RTNL();
4808
4809         if (!token)
4810                 return -EINVAL;
4811         if (ipv6_addr_any(token))
4812                 return -EINVAL;
4813         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4814                 return -EINVAL;
4815         if (!ipv6_accept_ra(idev))
4816                 return -EINVAL;
4817         if (idev->cnf.rtr_solicits <= 0)
4818                 return -EINVAL;
4819
4820         write_lock_bh(&idev->lock);
4821
4822         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4823         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4824
4825         write_unlock_bh(&idev->lock);
4826
4827         if (!idev->dead && (idev->if_flags & IF_READY) &&
4828             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4829                              IFA_F_OPTIMISTIC)) {
4830
4831                 /* If we're not ready, then normal ifup will take care
4832                  * of this. Otherwise, we need to request our rs here.
4833                  */
4834                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4835                 update_rs = true;
4836         }
4837
4838         write_lock_bh(&idev->lock);
4839
4840         if (update_rs) {
4841                 idev->if_flags |= IF_RS_SENT;
4842                 idev->rs_probes = 1;
4843                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4844         }
4845
4846         /* Well, that's kinda nasty ... */
4847         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4848                 spin_lock(&ifp->lock);
4849                 if (ifp->tokenized) {
4850                         ifp->valid_lft = 0;
4851                         ifp->prefered_lft = 0;
4852                 }
4853                 spin_unlock(&ifp->lock);
4854         }
4855
4856         write_unlock_bh(&idev->lock);
4857         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4858         addrconf_verify_rtnl();
4859         return 0;
4860 }
4861
4862 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4863         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4864         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4865 };
4866
4867 static int inet6_validate_link_af(const struct net_device *dev,
4868                                   const struct nlattr *nla)
4869 {
4870         struct nlattr *tb[IFLA_INET6_MAX + 1];
4871
4872         if (dev && !__in6_dev_get(dev))
4873                 return -EAFNOSUPPORT;
4874
4875         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4876 }
4877
4878 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4879 {
4880         int err = -EINVAL;
4881         struct inet6_dev *idev = __in6_dev_get(dev);
4882         struct nlattr *tb[IFLA_INET6_MAX + 1];
4883
4884         if (!idev)
4885                 return -EAFNOSUPPORT;
4886
4887         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4888                 BUG();
4889
4890         if (tb[IFLA_INET6_TOKEN]) {
4891                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4892                 if (err)
4893                         return err;
4894         }
4895
4896         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4897                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4898
4899                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4900                     mode != IN6_ADDR_GEN_MODE_NONE &&
4901                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4902                         return -EINVAL;
4903
4904                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4905                     !idev->cnf.stable_secret.initialized &&
4906                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4907                         return -EINVAL;
4908
4909                 idev->addr_gen_mode = mode;
4910                 err = 0;
4911         }
4912
4913         return err;
4914 }
4915
4916 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4917                              u32 portid, u32 seq, int event, unsigned int flags)
4918 {
4919         struct net_device *dev = idev->dev;
4920         struct ifinfomsg *hdr;
4921         struct nlmsghdr *nlh;
4922         void *protoinfo;
4923
4924         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4925         if (!nlh)
4926                 return -EMSGSIZE;
4927
4928         hdr = nlmsg_data(nlh);
4929         hdr->ifi_family = AF_INET6;
4930         hdr->__ifi_pad = 0;
4931         hdr->ifi_type = dev->type;
4932         hdr->ifi_index = dev->ifindex;
4933         hdr->ifi_flags = dev_get_flags(dev);
4934         hdr->ifi_change = 0;
4935
4936         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4937             (dev->addr_len &&
4938              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4939             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4940             (dev->ifindex != dev_get_iflink(dev) &&
4941              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
4942             nla_put_u8(skb, IFLA_OPERSTATE,
4943                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
4944                 goto nla_put_failure;
4945         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4946         if (!protoinfo)
4947                 goto nla_put_failure;
4948
4949         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4950                 goto nla_put_failure;
4951
4952         nla_nest_end(skb, protoinfo);
4953         nlmsg_end(skb, nlh);
4954         return 0;
4955
4956 nla_put_failure:
4957         nlmsg_cancel(skb, nlh);
4958         return -EMSGSIZE;
4959 }
4960
4961 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4962 {
4963         struct net *net = sock_net(skb->sk);
4964         int h, s_h;
4965         int idx = 0, s_idx;
4966         struct net_device *dev;
4967         struct inet6_dev *idev;
4968         struct hlist_head *head;
4969
4970         s_h = cb->args[0];
4971         s_idx = cb->args[1];
4972
4973         rcu_read_lock();
4974         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4975                 idx = 0;
4976                 head = &net->dev_index_head[h];
4977                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4978                         if (idx < s_idx)
4979                                 goto cont;
4980                         idev = __in6_dev_get(dev);
4981                         if (!idev)
4982                                 goto cont;
4983                         if (inet6_fill_ifinfo(skb, idev,
4984                                               NETLINK_CB(cb->skb).portid,
4985                                               cb->nlh->nlmsg_seq,
4986                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
4987                                 goto out;
4988 cont:
4989                         idx++;
4990                 }
4991         }
4992 out:
4993         rcu_read_unlock();
4994         cb->args[1] = idx;
4995         cb->args[0] = h;
4996
4997         return skb->len;
4998 }
4999
5000 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5001 {
5002         struct sk_buff *skb;
5003         struct net *net = dev_net(idev->dev);
5004         int err = -ENOBUFS;
5005
5006         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5007         if (!skb)
5008                 goto errout;
5009
5010         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5011         if (err < 0) {
5012                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5013                 WARN_ON(err == -EMSGSIZE);
5014                 kfree_skb(skb);
5015                 goto errout;
5016         }
5017         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5018         return;
5019 errout:
5020         if (err < 0)
5021                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5022 }
5023
5024 static inline size_t inet6_prefix_nlmsg_size(void)
5025 {
5026         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5027                + nla_total_size(sizeof(struct in6_addr))
5028                + nla_total_size(sizeof(struct prefix_cacheinfo));
5029 }
5030
5031 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5032                              struct prefix_info *pinfo, u32 portid, u32 seq,
5033                              int event, unsigned int flags)
5034 {
5035         struct prefixmsg *pmsg;
5036         struct nlmsghdr *nlh;
5037         struct prefix_cacheinfo ci;
5038
5039         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5040         if (!nlh)
5041                 return -EMSGSIZE;
5042
5043         pmsg = nlmsg_data(nlh);
5044         pmsg->prefix_family = AF_INET6;
5045         pmsg->prefix_pad1 = 0;
5046         pmsg->prefix_pad2 = 0;
5047         pmsg->prefix_ifindex = idev->dev->ifindex;
5048         pmsg->prefix_len = pinfo->prefix_len;
5049         pmsg->prefix_type = pinfo->type;
5050         pmsg->prefix_pad3 = 0;
5051         pmsg->prefix_flags = 0;
5052         if (pinfo->onlink)
5053                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5054         if (pinfo->autoconf)
5055                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5056
5057         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5058                 goto nla_put_failure;
5059         ci.preferred_time = ntohl(pinfo->prefered);
5060         ci.valid_time = ntohl(pinfo->valid);
5061         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5062                 goto nla_put_failure;
5063         nlmsg_end(skb, nlh);
5064         return 0;
5065
5066 nla_put_failure:
5067         nlmsg_cancel(skb, nlh);
5068         return -EMSGSIZE;
5069 }
5070
5071 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5072                          struct prefix_info *pinfo)
5073 {
5074         struct sk_buff *skb;
5075         struct net *net = dev_net(idev->dev);
5076         int err = -ENOBUFS;
5077
5078         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5079         if (!skb)
5080                 goto errout;
5081
5082         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5083         if (err < 0) {
5084                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5085                 WARN_ON(err == -EMSGSIZE);
5086                 kfree_skb(skb);
5087                 goto errout;
5088         }
5089         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5090         return;
5091 errout:
5092         if (err < 0)
5093                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5094 }
5095
5096 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5097 {
5098         struct net *net = dev_net(ifp->idev->dev);
5099
5100         if (event)
5101                 ASSERT_RTNL();
5102
5103         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5104
5105         switch (event) {
5106         case RTM_NEWADDR:
5107                 /*
5108                  * If the address was optimistic
5109                  * we inserted the route at the start of
5110                  * our DAD process, so we don't need
5111                  * to do it again
5112                  */
5113                 if (!(ifp->rt->rt6i_node))
5114                         ip6_ins_rt(ifp->rt);
5115                 if (ifp->idev->cnf.forwarding)
5116                         addrconf_join_anycast(ifp);
5117                 if (!ipv6_addr_any(&ifp->peer_addr))
5118                         addrconf_prefix_route(&ifp->peer_addr, 128,
5119                                               ifp->idev->dev, 0, 0);
5120                 break;
5121         case RTM_DELADDR:
5122                 if (ifp->idev->cnf.forwarding)
5123                         addrconf_leave_anycast(ifp);
5124                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5125                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5126                         struct rt6_info *rt;
5127
5128                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5129                                                        ifp->idev->dev, 0, 0);
5130                         if (rt)
5131                                 ip6_del_rt(rt);
5132                 }
5133                 dst_hold(&ifp->rt->dst);
5134
5135                 ip6_del_rt(ifp->rt);
5136
5137                 rt_genid_bump_ipv6(net);
5138                 break;
5139         }
5140         atomic_inc(&net->ipv6.dev_addr_genid);
5141 }
5142
5143 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5144 {
5145         rcu_read_lock_bh();
5146         if (likely(ifp->idev->dead == 0))
5147                 __ipv6_ifa_notify(event, ifp);
5148         rcu_read_unlock_bh();
5149 }
5150
5151 #ifdef CONFIG_SYSCTL
5152
5153 static
5154 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5155                            void __user *buffer, size_t *lenp, loff_t *ppos)
5156 {
5157         int *valp = ctl->data;
5158         int val = *valp;
5159         loff_t pos = *ppos;
5160         struct ctl_table lctl;
5161         int ret;
5162
5163         /*
5164          * ctl->data points to idev->cnf.forwarding, we should
5165          * not modify it until we get the rtnl lock.
5166          */
5167         lctl = *ctl;
5168         lctl.data = &val;
5169
5170         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5171
5172         if (write)
5173                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5174         if (ret)
5175                 *ppos = pos;
5176         return ret;
5177 }
5178
5179 static
5180 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5181                         void __user *buffer, size_t *lenp, loff_t *ppos)
5182 {
5183         struct inet6_dev *idev = ctl->extra1;
5184         int min_mtu = IPV6_MIN_MTU;
5185         struct ctl_table lctl;
5186
5187         lctl = *ctl;
5188         lctl.extra1 = &min_mtu;
5189         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5190
5191         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5192 }
5193
5194 static void dev_disable_change(struct inet6_dev *idev)
5195 {
5196         struct netdev_notifier_info info;
5197
5198         if (!idev || !idev->dev)
5199                 return;
5200
5201         netdev_notifier_info_init(&info, idev->dev);
5202         if (idev->cnf.disable_ipv6)
5203                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5204         else
5205                 addrconf_notify(NULL, NETDEV_UP, &info);
5206 }
5207
5208 static void addrconf_disable_change(struct net *net, __s32 newf)
5209 {
5210         struct net_device *dev;
5211         struct inet6_dev *idev;
5212
5213         rcu_read_lock();
5214         for_each_netdev_rcu(net, dev) {
5215                 idev = __in6_dev_get(dev);
5216                 if (idev) {
5217                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5218                         idev->cnf.disable_ipv6 = newf;
5219                         if (changed)
5220                                 dev_disable_change(idev);
5221                 }
5222         }
5223         rcu_read_unlock();
5224 }
5225
5226 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5227 {
5228         struct net *net;
5229         int old;
5230
5231         if (!rtnl_trylock())
5232                 return restart_syscall();
5233
5234         net = (struct net *)table->extra2;
5235         old = *p;
5236         *p = newf;
5237
5238         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5239                 rtnl_unlock();
5240                 return 0;
5241         }
5242
5243         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5244                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5245                 addrconf_disable_change(net, newf);
5246         } else if ((!newf) ^ (!old))
5247                 dev_disable_change((struct inet6_dev *)table->extra1);
5248
5249         rtnl_unlock();
5250         return 0;
5251 }
5252
5253 static
5254 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5255                             void __user *buffer, size_t *lenp, loff_t *ppos)
5256 {
5257         int *valp = ctl->data;
5258         int val = *valp;
5259         loff_t pos = *ppos;
5260         struct ctl_table lctl;
5261         int ret;
5262
5263         /*
5264          * ctl->data points to idev->cnf.disable_ipv6, we should
5265          * not modify it until we get the rtnl lock.
5266          */
5267         lctl = *ctl;
5268         lctl.data = &val;
5269
5270         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5271
5272         if (write)
5273                 ret = addrconf_disable_ipv6(ctl, valp, val);
5274         if (ret)
5275                 *ppos = pos;
5276         return ret;
5277 }
5278
5279 static
5280 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5281                               void __user *buffer, size_t *lenp, loff_t *ppos)
5282 {
5283         int *valp = ctl->data;
5284         int ret;
5285         int old, new;
5286
5287         old = *valp;
5288         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5289         new = *valp;
5290
5291         if (write && old != new) {
5292                 struct net *net = ctl->extra2;
5293
5294                 if (!rtnl_trylock())
5295                         return restart_syscall();
5296
5297                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5298                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5299                                                      NETCONFA_IFINDEX_DEFAULT,
5300                                                      net->ipv6.devconf_dflt);
5301                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5302                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5303                                                      NETCONFA_IFINDEX_ALL,
5304                                                      net->ipv6.devconf_all);
5305                 else {
5306                         struct inet6_dev *idev = ctl->extra1;
5307
5308                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5309                                                      idev->dev->ifindex,
5310                                                      &idev->cnf);
5311                 }
5312                 rtnl_unlock();
5313         }
5314
5315         return ret;
5316 }
5317
5318 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5319                                          void __user *buffer, size_t *lenp,
5320                                          loff_t *ppos)
5321 {
5322         int err;
5323         struct in6_addr addr;
5324         char str[IPV6_MAX_STRLEN];
5325         struct ctl_table lctl = *ctl;
5326         struct net *net = ctl->extra2;
5327         struct ipv6_stable_secret *secret = ctl->data;
5328
5329         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5330                 return -EIO;
5331
5332         lctl.maxlen = IPV6_MAX_STRLEN;
5333         lctl.data = str;
5334
5335         if (!rtnl_trylock())
5336                 return restart_syscall();
5337
5338         if (!write && !secret->initialized) {
5339                 err = -EIO;
5340                 goto out;
5341         }
5342
5343         if (!write) {
5344                 err = snprintf(str, sizeof(str), "%pI6",
5345                                &secret->secret);
5346                 if (err >= sizeof(str)) {
5347                         err = -EIO;
5348                         goto out;
5349                 }
5350         }
5351
5352         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5353         if (err || !write)
5354                 goto out;
5355
5356         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5357                 err = -EIO;
5358                 goto out;
5359         }
5360
5361         secret->initialized = true;
5362         secret->secret = addr;
5363
5364         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5365                 struct net_device *dev;
5366
5367                 for_each_netdev(net, dev) {
5368                         struct inet6_dev *idev = __in6_dev_get(dev);
5369
5370                         if (idev) {
5371                                 idev->addr_gen_mode =
5372                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5373                         }
5374                 }
5375         } else {
5376                 struct inet6_dev *idev = ctl->extra1;
5377
5378                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5379         }
5380
5381 out:
5382         rtnl_unlock();
5383
5384         return err;
5385 }
5386
5387 static
5388 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5389                                                 int write,
5390                                                 void __user *buffer,
5391                                                 size_t *lenp,
5392                                                 loff_t *ppos)
5393 {
5394         int *valp = ctl->data;
5395         int val = *valp;
5396         loff_t pos = *ppos;
5397         struct ctl_table lctl;
5398         int ret;
5399
5400         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5401          * we should not modify it until we get the rtnl lock.
5402          */
5403         lctl = *ctl;
5404         lctl.data = &val;
5405
5406         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5407
5408         if (write)
5409                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5410         if (ret)
5411                 *ppos = pos;
5412         return ret;
5413 }
5414
5415 static struct addrconf_sysctl_table
5416 {
5417         struct ctl_table_header *sysctl_header;
5418         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5419 } addrconf_sysctl __read_mostly = {
5420         .sysctl_header = NULL,
5421         .addrconf_vars = {
5422                 {
5423                         .procname       = "forwarding",
5424                         .data           = &ipv6_devconf.forwarding,
5425                         .maxlen         = sizeof(int),
5426                         .mode           = 0644,
5427                         .proc_handler   = addrconf_sysctl_forward,
5428                 },
5429                 {
5430                         .procname       = "hop_limit",
5431                         .data           = &ipv6_devconf.hop_limit,
5432                         .maxlen         = sizeof(int),
5433                         .mode           = 0644,
5434                         .proc_handler   = proc_dointvec,
5435                 },
5436                 {
5437                         .procname       = "mtu",
5438                         .data           = &ipv6_devconf.mtu6,
5439                         .maxlen         = sizeof(int),
5440                         .mode           = 0644,
5441                         .proc_handler   = addrconf_sysctl_mtu,
5442                 },
5443                 {
5444                         .procname       = "accept_ra",
5445                         .data           = &ipv6_devconf.accept_ra,
5446                         .maxlen         = sizeof(int),
5447                         .mode           = 0644,
5448                         .proc_handler   = proc_dointvec,
5449                 },
5450                 {
5451                         .procname       = "accept_redirects",
5452                         .data           = &ipv6_devconf.accept_redirects,
5453                         .maxlen         = sizeof(int),
5454                         .mode           = 0644,
5455                         .proc_handler   = proc_dointvec,
5456                 },
5457                 {
5458                         .procname       = "autoconf",
5459                         .data           = &ipv6_devconf.autoconf,
5460                         .maxlen         = sizeof(int),
5461                         .mode           = 0644,
5462                         .proc_handler   = proc_dointvec,
5463                 },
5464                 {
5465                         .procname       = "dad_transmits",
5466                         .data           = &ipv6_devconf.dad_transmits,
5467                         .maxlen         = sizeof(int),
5468                         .mode           = 0644,
5469                         .proc_handler   = proc_dointvec,
5470                 },
5471                 {
5472                         .procname       = "router_solicitations",
5473                         .data           = &ipv6_devconf.rtr_solicits,
5474                         .maxlen         = sizeof(int),
5475                         .mode           = 0644,
5476                         .proc_handler   = proc_dointvec,
5477                 },
5478                 {
5479                         .procname       = "router_solicitation_interval",
5480                         .data           = &ipv6_devconf.rtr_solicit_interval,
5481                         .maxlen         = sizeof(int),
5482                         .mode           = 0644,
5483                         .proc_handler   = proc_dointvec_jiffies,
5484                 },
5485                 {
5486                         .procname       = "router_solicitation_delay",
5487                         .data           = &ipv6_devconf.rtr_solicit_delay,
5488                         .maxlen         = sizeof(int),
5489                         .mode           = 0644,
5490                         .proc_handler   = proc_dointvec_jiffies,
5491                 },
5492                 {
5493                         .procname       = "force_mld_version",
5494                         .data           = &ipv6_devconf.force_mld_version,
5495                         .maxlen         = sizeof(int),
5496                         .mode           = 0644,
5497                         .proc_handler   = proc_dointvec,
5498                 },
5499                 {
5500                         .procname       = "mldv1_unsolicited_report_interval",
5501                         .data           =
5502                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5503                         .maxlen         = sizeof(int),
5504                         .mode           = 0644,
5505                         .proc_handler   = proc_dointvec_ms_jiffies,
5506                 },
5507                 {
5508                         .procname       = "mldv2_unsolicited_report_interval",
5509                         .data           =
5510                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5511                         .maxlen         = sizeof(int),
5512                         .mode           = 0644,
5513                         .proc_handler   = proc_dointvec_ms_jiffies,
5514                 },
5515                 {
5516                         .procname       = "use_tempaddr",
5517                         .data           = &ipv6_devconf.use_tempaddr,
5518                         .maxlen         = sizeof(int),
5519                         .mode           = 0644,
5520                         .proc_handler   = proc_dointvec,
5521                 },
5522                 {
5523                         .procname       = "temp_valid_lft",
5524                         .data           = &ipv6_devconf.temp_valid_lft,
5525                         .maxlen         = sizeof(int),
5526                         .mode           = 0644,
5527                         .proc_handler   = proc_dointvec,
5528                 },
5529                 {
5530                         .procname       = "temp_prefered_lft",
5531                         .data           = &ipv6_devconf.temp_prefered_lft,
5532                         .maxlen         = sizeof(int),
5533                         .mode           = 0644,
5534                         .proc_handler   = proc_dointvec,
5535                 },
5536                 {
5537                         .procname       = "regen_max_retry",
5538                         .data           = &ipv6_devconf.regen_max_retry,
5539                         .maxlen         = sizeof(int),
5540                         .mode           = 0644,
5541                         .proc_handler   = proc_dointvec,
5542                 },
5543                 {
5544                         .procname       = "max_desync_factor",
5545                         .data           = &ipv6_devconf.max_desync_factor,
5546                         .maxlen         = sizeof(int),
5547                         .mode           = 0644,
5548                         .proc_handler   = proc_dointvec,
5549                 },
5550                 {
5551                         .procname       = "max_addresses",
5552                         .data           = &ipv6_devconf.max_addresses,
5553                         .maxlen         = sizeof(int),
5554                         .mode           = 0644,
5555                         .proc_handler   = proc_dointvec,
5556                 },
5557                 {
5558                         .procname       = "accept_ra_defrtr",
5559                         .data           = &ipv6_devconf.accept_ra_defrtr,
5560                         .maxlen         = sizeof(int),
5561                         .mode           = 0644,
5562                         .proc_handler   = proc_dointvec,
5563                 },
5564                 {
5565                         .procname       = "accept_ra_min_hop_limit",
5566                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5567                         .maxlen         = sizeof(int),
5568                         .mode           = 0644,
5569                         .proc_handler   = proc_dointvec,
5570                 },
5571                 {
5572                         .procname       = "accept_ra_pinfo",
5573                         .data           = &ipv6_devconf.accept_ra_pinfo,
5574                         .maxlen         = sizeof(int),
5575                         .mode           = 0644,
5576                         .proc_handler   = proc_dointvec,
5577                 },
5578 #ifdef CONFIG_IPV6_ROUTER_PREF
5579                 {
5580                         .procname       = "accept_ra_rtr_pref",
5581                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5582                         .maxlen         = sizeof(int),
5583                         .mode           = 0644,
5584                         .proc_handler   = proc_dointvec,
5585                 },
5586                 {
5587                         .procname       = "router_probe_interval",
5588                         .data           = &ipv6_devconf.rtr_probe_interval,
5589                         .maxlen         = sizeof(int),
5590                         .mode           = 0644,
5591                         .proc_handler   = proc_dointvec_jiffies,
5592                 },
5593 #ifdef CONFIG_IPV6_ROUTE_INFO
5594                 {
5595                         .procname       = "accept_ra_rt_info_max_plen",
5596                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5597                         .maxlen         = sizeof(int),
5598                         .mode           = 0644,
5599                         .proc_handler   = proc_dointvec,
5600                 },
5601 #endif
5602 #endif
5603                 {
5604                         .procname       = "proxy_ndp",
5605                         .data           = &ipv6_devconf.proxy_ndp,
5606                         .maxlen         = sizeof(int),
5607                         .mode           = 0644,
5608                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5609                 },
5610                 {
5611                         .procname       = "accept_source_route",
5612                         .data           = &ipv6_devconf.accept_source_route,
5613                         .maxlen         = sizeof(int),
5614                         .mode           = 0644,
5615                         .proc_handler   = proc_dointvec,
5616                 },
5617 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5618                 {
5619                         .procname       = "optimistic_dad",
5620                         .data           = &ipv6_devconf.optimistic_dad,
5621                         .maxlen         = sizeof(int),
5622                         .mode           = 0644,
5623                         .proc_handler   = proc_dointvec,
5624
5625                 },
5626                 {
5627                         .procname       = "use_optimistic",
5628                         .data           = &ipv6_devconf.use_optimistic,
5629                         .maxlen         = sizeof(int),
5630                         .mode           = 0644,
5631                         .proc_handler   = proc_dointvec,
5632
5633                 },
5634 #endif
5635 #ifdef CONFIG_IPV6_MROUTE
5636                 {
5637                         .procname       = "mc_forwarding",
5638                         .data           = &ipv6_devconf.mc_forwarding,
5639                         .maxlen         = sizeof(int),
5640                         .mode           = 0444,
5641                         .proc_handler   = proc_dointvec,
5642                 },
5643 #endif
5644                 {
5645                         .procname       = "disable_ipv6",
5646                         .data           = &ipv6_devconf.disable_ipv6,
5647                         .maxlen         = sizeof(int),
5648                         .mode           = 0644,
5649                         .proc_handler   = addrconf_sysctl_disable,
5650                 },
5651                 {
5652                         .procname       = "accept_dad",
5653                         .data           = &ipv6_devconf.accept_dad,
5654                         .maxlen         = sizeof(int),
5655                         .mode           = 0644,
5656                         .proc_handler   = proc_dointvec,
5657                 },
5658                 {
5659                         .procname       = "force_tllao",
5660                         .data           = &ipv6_devconf.force_tllao,
5661                         .maxlen         = sizeof(int),
5662                         .mode           = 0644,
5663                         .proc_handler   = proc_dointvec
5664                 },
5665                 {
5666                         .procname       = "ndisc_notify",
5667                         .data           = &ipv6_devconf.ndisc_notify,
5668                         .maxlen         = sizeof(int),
5669                         .mode           = 0644,
5670                         .proc_handler   = proc_dointvec
5671                 },
5672                 {
5673                         .procname       = "suppress_frag_ndisc",
5674                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5675                         .maxlen         = sizeof(int),
5676                         .mode           = 0644,
5677                         .proc_handler   = proc_dointvec
5678                 },
5679                 {
5680                         .procname       = "accept_ra_from_local",
5681                         .data           = &ipv6_devconf.accept_ra_from_local,
5682                         .maxlen         = sizeof(int),
5683                         .mode           = 0644,
5684                         .proc_handler   = proc_dointvec,
5685                 },
5686                 {
5687                         .procname       = "accept_ra_mtu",
5688                         .data           = &ipv6_devconf.accept_ra_mtu,
5689                         .maxlen         = sizeof(int),
5690                         .mode           = 0644,
5691                         .proc_handler   = proc_dointvec,
5692                 },
5693                 {
5694                         .procname       = "stable_secret",
5695                         .data           = &ipv6_devconf.stable_secret,
5696                         .maxlen         = IPV6_MAX_STRLEN,
5697                         .mode           = 0600,
5698                         .proc_handler   = addrconf_sysctl_stable_secret,
5699                 },
5700                 {
5701                         .procname       = "use_oif_addrs_only",
5702                         .data           = &ipv6_devconf.use_oif_addrs_only,
5703                         .maxlen         = sizeof(int),
5704                         .mode           = 0644,
5705                         .proc_handler   = proc_dointvec,
5706                 },
5707                 {
5708                         .procname       = "ignore_routes_with_linkdown",
5709                         .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5710                         .maxlen         = sizeof(int),
5711                         .mode           = 0644,
5712                         .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5713                 },
5714                 {
5715                         /* sentinel */
5716                 }
5717         },
5718 };
5719
5720 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5721                 struct inet6_dev *idev, struct ipv6_devconf *p)
5722 {
5723         int i;
5724         struct addrconf_sysctl_table *t;
5725         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5726
5727         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5728         if (!t)
5729                 goto out;
5730
5731         for (i = 0; t->addrconf_vars[i].data; i++) {
5732                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5733                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5734                 t->addrconf_vars[i].extra2 = net;
5735         }
5736
5737         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5738
5739         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5740         if (!t->sysctl_header)
5741                 goto free;
5742
5743         p->sysctl = t;
5744         return 0;
5745
5746 free:
5747         kfree(t);
5748 out:
5749         return -ENOBUFS;
5750 }
5751
5752 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5753 {
5754         struct addrconf_sysctl_table *t;
5755
5756         if (!p->sysctl)
5757                 return;
5758
5759         t = p->sysctl;
5760         p->sysctl = NULL;
5761         unregister_net_sysctl_table(t->sysctl_header);
5762         kfree(t);
5763 }
5764
5765 static int addrconf_sysctl_register(struct inet6_dev *idev)
5766 {
5767         int err;
5768
5769         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5770                 return -EINVAL;
5771
5772         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5773                                     &ndisc_ifinfo_sysctl_change);
5774         if (err)
5775                 return err;
5776         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5777                                          idev, &idev->cnf);
5778         if (err)
5779                 neigh_sysctl_unregister(idev->nd_parms);
5780
5781         return err;
5782 }
5783
5784 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5785 {
5786         __addrconf_sysctl_unregister(&idev->cnf);
5787         neigh_sysctl_unregister(idev->nd_parms);
5788 }
5789
5790
5791 #endif
5792
5793 static int __net_init addrconf_init_net(struct net *net)
5794 {
5795         int err = -ENOMEM;
5796         struct ipv6_devconf *all, *dflt;
5797
5798         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5799         if (!all)
5800                 goto err_alloc_all;
5801
5802         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5803         if (!dflt)
5804                 goto err_alloc_dflt;
5805
5806         /* these will be inherited by all namespaces */
5807         dflt->autoconf = ipv6_defaults.autoconf;
5808         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5809
5810         dflt->stable_secret.initialized = false;
5811         all->stable_secret.initialized = false;
5812
5813         net->ipv6.devconf_all = all;
5814         net->ipv6.devconf_dflt = dflt;
5815
5816 #ifdef CONFIG_SYSCTL
5817         err = __addrconf_sysctl_register(net, "all", NULL, all);
5818         if (err < 0)
5819                 goto err_reg_all;
5820
5821         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5822         if (err < 0)
5823                 goto err_reg_dflt;
5824 #endif
5825         return 0;
5826
5827 #ifdef CONFIG_SYSCTL
5828 err_reg_dflt:
5829         __addrconf_sysctl_unregister(all);
5830 err_reg_all:
5831         kfree(dflt);
5832 #endif
5833 err_alloc_dflt:
5834         kfree(all);
5835 err_alloc_all:
5836         return err;
5837 }
5838
5839 static void __net_exit addrconf_exit_net(struct net *net)
5840 {
5841 #ifdef CONFIG_SYSCTL
5842         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5843         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5844 #endif
5845         kfree(net->ipv6.devconf_dflt);
5846         kfree(net->ipv6.devconf_all);
5847 }
5848
5849 static struct pernet_operations addrconf_ops = {
5850         .init = addrconf_init_net,
5851         .exit = addrconf_exit_net,
5852 };
5853
5854 static struct rtnl_af_ops inet6_ops __read_mostly = {
5855         .family           = AF_INET6,
5856         .fill_link_af     = inet6_fill_link_af,
5857         .get_link_af_size = inet6_get_link_af_size,
5858         .validate_link_af = inet6_validate_link_af,
5859         .set_link_af      = inet6_set_link_af,
5860 };
5861
5862 /*
5863  *      Init / cleanup code
5864  */
5865
5866 int __init addrconf_init(void)
5867 {
5868         struct inet6_dev *idev;
5869         int i, err;
5870
5871         err = ipv6_addr_label_init();
5872         if (err < 0) {
5873                 pr_crit("%s: cannot initialize default policy table: %d\n",
5874                         __func__, err);
5875                 goto out;
5876         }
5877
5878         err = register_pernet_subsys(&addrconf_ops);
5879         if (err < 0)
5880                 goto out_addrlabel;
5881
5882         addrconf_wq = create_workqueue("ipv6_addrconf");
5883         if (!addrconf_wq) {
5884                 err = -ENOMEM;
5885                 goto out_nowq;
5886         }
5887
5888         /* The addrconf netdev notifier requires that loopback_dev
5889          * has it's ipv6 private information allocated and setup
5890          * before it can bring up and give link-local addresses
5891          * to other devices which are up.
5892          *
5893          * Unfortunately, loopback_dev is not necessarily the first
5894          * entry in the global dev_base list of net devices.  In fact,
5895          * it is likely to be the very last entry on that list.
5896          * So this causes the notifier registry below to try and
5897          * give link-local addresses to all devices besides loopback_dev
5898          * first, then loopback_dev, which cases all the non-loopback_dev
5899          * devices to fail to get a link-local address.
5900          *
5901          * So, as a temporary fix, allocate the ipv6 structure for
5902          * loopback_dev first by hand.
5903          * Longer term, all of the dependencies ipv6 has upon the loopback
5904          * device and it being up should be removed.
5905          */
5906         rtnl_lock();
5907         idev = ipv6_add_dev(init_net.loopback_dev);
5908         rtnl_unlock();
5909         if (IS_ERR(idev)) {
5910                 err = PTR_ERR(idev);
5911                 goto errlo;
5912         }
5913
5914         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5915                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5916
5917         register_netdevice_notifier(&ipv6_dev_notf);
5918
5919         addrconf_verify();
5920
5921         rtnl_af_register(&inet6_ops);
5922
5923         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5924                               NULL);
5925         if (err < 0)
5926                 goto errout;
5927
5928         /* Only the first call to __rtnl_register can fail */
5929         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5930         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5931         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5932                         inet6_dump_ifaddr, NULL);
5933         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5934                         inet6_dump_ifmcaddr, NULL);
5935         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5936                         inet6_dump_ifacaddr, NULL);
5937         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5938                         inet6_netconf_dump_devconf, NULL);
5939
5940         ipv6_addr_label_rtnl_register();
5941
5942         return 0;
5943 errout:
5944         rtnl_af_unregister(&inet6_ops);
5945         unregister_netdevice_notifier(&ipv6_dev_notf);
5946 errlo:
5947         destroy_workqueue(addrconf_wq);
5948 out_nowq:
5949         unregister_pernet_subsys(&addrconf_ops);
5950 out_addrlabel:
5951         ipv6_addr_label_cleanup();
5952 out:
5953         return err;
5954 }
5955
5956 void addrconf_cleanup(void)
5957 {
5958         struct net_device *dev;
5959         int i;
5960
5961         unregister_netdevice_notifier(&ipv6_dev_notf);
5962         unregister_pernet_subsys(&addrconf_ops);
5963         ipv6_addr_label_cleanup();
5964
5965         rtnl_lock();
5966
5967         __rtnl_af_unregister(&inet6_ops);
5968
5969         /* clean dev list */
5970         for_each_netdev(&init_net, dev) {
5971                 if (__in6_dev_get(dev) == NULL)
5972                         continue;
5973                 addrconf_ifdown(dev, 1);
5974         }
5975         addrconf_ifdown(init_net.loopback_dev, 2);
5976
5977         /*
5978          *      Check hash table.
5979          */
5980         spin_lock_bh(&addrconf_hash_lock);
5981         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5982                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5983         spin_unlock_bh(&addrconf_hash_lock);
5984         cancel_delayed_work(&addr_chk_work);
5985         rtnl_unlock();
5986
5987         destroy_workqueue(addrconf_wq);
5988 }