Documentation: bindings: add dt documentation for Rockchip PCIe PHY
[firefly-linux-kernel-4.4.55.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 <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
91
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
95
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
98
99 #if ACONF_DEBUG >= 3
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101 #else
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #endif
104
105 #define INFINITY_LIFE_TIME      0xFFFFFFFF
106
107 #define IPV6_MAX_STRLEN \
108         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
109
110 static inline u32 cstamp_delta(unsigned long cstamp)
111 {
112         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
113 }
114
115 #ifdef CONFIG_SYSCTL
116 static int addrconf_sysctl_register(struct inet6_dev *idev);
117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118 #else
119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
120 {
121         return 0;
122 }
123
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
125 {
126 }
127 #endif
128
129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131 static void ipv6_regen_rndid(unsigned long data);
132
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135 static int ipv6_generate_stable_address(struct in6_addr *addr,
136                                         u8 dad_count,
137                                         const struct inet6_dev *idev);
138
139 /*
140  *      Configured unicast address hash table
141  */
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
144
145 static void addrconf_verify(void);
146 static void addrconf_verify_rtnl(void);
147 static void addrconf_verify_work(struct work_struct *);
148
149 static struct workqueue_struct *addrconf_wq;
150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
151
152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
154
155 static void addrconf_type_change(struct net_device *dev,
156                                  unsigned long event);
157 static int addrconf_ifdown(struct net_device *dev, int how);
158
159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
160                                                   int plen,
161                                                   const struct net_device *dev,
162                                                   u32 flags, u32 noflags);
163
164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165 static void addrconf_dad_work(struct work_struct *w);
166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167 static void addrconf_dad_run(struct inet6_dev *idev);
168 static void addrconf_rs_timer(unsigned long data);
169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
171
172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173                                 struct prefix_info *pinfo);
174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175                                struct net_device *dev);
176
177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
178         .forwarding             = 0,
179         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
180         .mtu6                   = IPV6_MIN_MTU,
181         .accept_ra              = 1,
182         .accept_redirects       = 1,
183         .autoconf               = 1,
184         .force_mld_version      = 0,
185         .mldv1_unsolicited_report_interval = 10 * HZ,
186         .mldv2_unsolicited_report_interval = HZ,
187         .dad_transmits          = 1,
188         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
189         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
190         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
191         .use_tempaddr           = 0,
192         .temp_valid_lft         = TEMP_VALID_LIFETIME,
193         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
194         .regen_max_retry        = REGEN_MAX_RETRY,
195         .max_desync_factor      = MAX_DESYNC_FACTOR,
196         .max_addresses          = IPV6_MAX_ADDRESSES,
197         .accept_ra_defrtr       = 1,
198         .accept_ra_from_local   = 0,
199         .accept_ra_min_hop_limit= 1,
200         .accept_ra_pinfo        = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202         .accept_ra_rtr_pref     = 1,
203         .rtr_probe_interval     = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205         .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208         .accept_ra_rt_table     = 0,
209         .proxy_ndp              = 0,
210         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
211         .disable_ipv6           = 0,
212         .accept_dad             = 1,
213         .suppress_frag_ndisc    = 1,
214         .accept_ra_mtu          = 1,
215         .stable_secret          = {
216                 .initialized = false,
217         },
218         .use_oif_addrs_only     = 0,
219         .ignore_routes_with_linkdown = 0,
220 };
221
222 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
223         .forwarding             = 0,
224         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
225         .mtu6                   = IPV6_MIN_MTU,
226         .accept_ra              = 1,
227         .accept_redirects       = 1,
228         .autoconf               = 1,
229         .force_mld_version      = 0,
230         .mldv1_unsolicited_report_interval = 10 * HZ,
231         .mldv2_unsolicited_report_interval = HZ,
232         .dad_transmits          = 1,
233         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
234         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
235         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
236         .use_tempaddr           = 0,
237         .temp_valid_lft         = TEMP_VALID_LIFETIME,
238         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
239         .regen_max_retry        = REGEN_MAX_RETRY,
240         .max_desync_factor      = MAX_DESYNC_FACTOR,
241         .max_addresses          = IPV6_MAX_ADDRESSES,
242         .accept_ra_defrtr       = 1,
243         .accept_ra_from_local   = 0,
244         .accept_ra_min_hop_limit= 1,
245         .accept_ra_pinfo        = 1,
246 #ifdef CONFIG_IPV6_ROUTER_PREF
247         .accept_ra_rtr_pref     = 1,
248         .rtr_probe_interval     = 60 * HZ,
249 #ifdef CONFIG_IPV6_ROUTE_INFO
250         .accept_ra_rt_info_max_plen = 0,
251 #endif
252 #endif
253         .accept_ra_rt_table     = 0,
254         .proxy_ndp              = 0,
255         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
256         .disable_ipv6           = 0,
257         .accept_dad             = 1,
258         .suppress_frag_ndisc    = 1,
259         .accept_ra_mtu          = 1,
260         .stable_secret          = {
261                 .initialized = false,
262         },
263         .use_oif_addrs_only     = 0,
264         .ignore_routes_with_linkdown = 0,
265 };
266
267 /* Check if a valid qdisc is available */
268 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
269 {
270         return !qdisc_tx_is_noop(dev);
271 }
272
273 static void addrconf_del_rs_timer(struct inet6_dev *idev)
274 {
275         if (del_timer(&idev->rs_timer))
276                 __in6_dev_put(idev);
277 }
278
279 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
280 {
281         if (cancel_delayed_work(&ifp->dad_work))
282                 __in6_ifa_put(ifp);
283 }
284
285 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
286                                   unsigned long when)
287 {
288         if (!timer_pending(&idev->rs_timer))
289                 in6_dev_hold(idev);
290         mod_timer(&idev->rs_timer, jiffies + when);
291 }
292
293 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
294                                    unsigned long delay)
295 {
296         if (!delayed_work_pending(&ifp->dad_work))
297                 in6_ifa_hold(ifp);
298         mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
299 }
300
301 static int snmp6_alloc_dev(struct inet6_dev *idev)
302 {
303         int i;
304
305         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
306         if (!idev->stats.ipv6)
307                 goto err_ip;
308
309         for_each_possible_cpu(i) {
310                 struct ipstats_mib *addrconf_stats;
311                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
312                 u64_stats_init(&addrconf_stats->syncp);
313         }
314
315
316         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
317                                         GFP_KERNEL);
318         if (!idev->stats.icmpv6dev)
319                 goto err_icmp;
320         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
321                                            GFP_KERNEL);
322         if (!idev->stats.icmpv6msgdev)
323                 goto err_icmpmsg;
324
325         return 0;
326
327 err_icmpmsg:
328         kfree(idev->stats.icmpv6dev);
329 err_icmp:
330         free_percpu(idev->stats.ipv6);
331 err_ip:
332         return -ENOMEM;
333 }
334
335 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
336 {
337         struct inet6_dev *ndev;
338         int err = -ENOMEM;
339
340         ASSERT_RTNL();
341
342         if (dev->mtu < IPV6_MIN_MTU)
343                 return ERR_PTR(-EINVAL);
344
345         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
346         if (!ndev)
347                 return ERR_PTR(err);
348
349         rwlock_init(&ndev->lock);
350         ndev->dev = dev;
351         INIT_LIST_HEAD(&ndev->addr_list);
352         setup_timer(&ndev->rs_timer, addrconf_rs_timer,
353                     (unsigned long)ndev);
354         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
355
356         if (ndev->cnf.stable_secret.initialized)
357                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
358         else
359                 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
360
361         ndev->cnf.mtu6 = dev->mtu;
362         ndev->cnf.sysctl = NULL;
363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364         if (!ndev->nd_parms) {
365                 kfree(ndev);
366                 return ERR_PTR(err);
367         }
368         if (ndev->cnf.forwarding)
369                 dev_disable_lro(dev);
370         /* We refer to the device */
371         dev_hold(dev);
372
373         if (snmp6_alloc_dev(ndev) < 0) {
374                 ADBG(KERN_WARNING
375                         "%s: cannot allocate memory for statistics; dev=%s.\n",
376                         __func__, dev->name);
377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378                 dev_put(dev);
379                 kfree(ndev);
380                 return ERR_PTR(err);
381         }
382
383         if (snmp6_register_dev(ndev) < 0) {
384                 ADBG(KERN_WARNING
385                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
386                         __func__, dev->name);
387                 goto err_release;
388         }
389
390         /* One reference from device.  We must do this before
391          * we invoke __ipv6_regen_rndid().
392          */
393         in6_dev_hold(ndev);
394
395         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
396                 ndev->cnf.accept_dad = -1;
397
398 #if IS_ENABLED(CONFIG_IPV6_SIT)
399         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
400                 pr_info("%s: Disabled Multicast RS\n", dev->name);
401                 ndev->cnf.rtr_solicits = 0;
402         }
403 #endif
404
405         INIT_LIST_HEAD(&ndev->tempaddr_list);
406         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
407         if ((dev->flags&IFF_LOOPBACK) ||
408             dev->type == ARPHRD_TUNNEL ||
409             dev->type == ARPHRD_TUNNEL6 ||
410             dev->type == ARPHRD_SIT ||
411             dev->type == ARPHRD_NONE) {
412                 ndev->cnf.use_tempaddr = -1;
413         } else {
414                 in6_dev_hold(ndev);
415                 ipv6_regen_rndid((unsigned long) ndev);
416         }
417
418         ndev->token = in6addr_any;
419
420         if (netif_running(dev) && addrconf_qdisc_ok(dev))
421                 ndev->if_flags |= IF_READY;
422
423         ipv6_mc_init_dev(ndev);
424         ndev->tstamp = jiffies;
425         err = addrconf_sysctl_register(ndev);
426         if (err) {
427                 ipv6_mc_destroy_dev(ndev);
428                 del_timer(&ndev->regen_timer);
429                 snmp6_unregister_dev(ndev);
430                 goto err_release;
431         }
432         /* protected by rtnl_lock */
433         rcu_assign_pointer(dev->ip6_ptr, ndev);
434
435         /* Join interface-local all-node multicast group */
436         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
437
438         /* Join all-node multicast group */
439         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
440
441         /* Join all-router multicast group if forwarding is set */
442         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
443                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
444
445         return ndev;
446
447 err_release:
448         neigh_parms_release(&nd_tbl, ndev->nd_parms);
449         ndev->dead = 1;
450         in6_dev_finish_destroy(ndev);
451         return ERR_PTR(err);
452 }
453
454 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
455 {
456         struct inet6_dev *idev;
457
458         ASSERT_RTNL();
459
460         idev = __in6_dev_get(dev);
461         if (!idev) {
462                 idev = ipv6_add_dev(dev);
463                 if (IS_ERR(idev))
464                         return NULL;
465         }
466
467         if (dev->flags&IFF_UP)
468                 ipv6_mc_up(idev);
469         return idev;
470 }
471
472 static int inet6_netconf_msgsize_devconf(int type)
473 {
474         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
475                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
476
477         /* type -1 is used for ALL */
478         if (type == -1 || type == NETCONFA_FORWARDING)
479                 size += nla_total_size(4);
480 #ifdef CONFIG_IPV6_MROUTE
481         if (type == -1 || type == NETCONFA_MC_FORWARDING)
482                 size += nla_total_size(4);
483 #endif
484         if (type == -1 || type == NETCONFA_PROXY_NEIGH)
485                 size += nla_total_size(4);
486
487         if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
488                 size += nla_total_size(4);
489
490         return size;
491 }
492
493 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
494                                       struct ipv6_devconf *devconf, u32 portid,
495                                       u32 seq, int event, unsigned int flags,
496                                       int type)
497 {
498         struct nlmsghdr  *nlh;
499         struct netconfmsg *ncm;
500
501         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
502                         flags);
503         if (!nlh)
504                 return -EMSGSIZE;
505
506         ncm = nlmsg_data(nlh);
507         ncm->ncm_family = AF_INET6;
508
509         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
510                 goto nla_put_failure;
511
512         /* type -1 is used for ALL */
513         if ((type == -1 || type == NETCONFA_FORWARDING) &&
514             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
515                 goto nla_put_failure;
516 #ifdef CONFIG_IPV6_MROUTE
517         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
518             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
519                         devconf->mc_forwarding) < 0)
520                 goto nla_put_failure;
521 #endif
522         if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
523             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
524                 goto nla_put_failure;
525
526         if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
527             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
528                         devconf->ignore_routes_with_linkdown) < 0)
529                 goto nla_put_failure;
530
531         nlmsg_end(skb, nlh);
532         return 0;
533
534 nla_put_failure:
535         nlmsg_cancel(skb, nlh);
536         return -EMSGSIZE;
537 }
538
539 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
540                                   struct ipv6_devconf *devconf)
541 {
542         struct sk_buff *skb;
543         int err = -ENOBUFS;
544
545         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
546         if (!skb)
547                 goto errout;
548
549         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
550                                          RTM_NEWNETCONF, 0, type);
551         if (err < 0) {
552                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
553                 WARN_ON(err == -EMSGSIZE);
554                 kfree_skb(skb);
555                 goto errout;
556         }
557         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
558         return;
559 errout:
560         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
561 }
562
563 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
564         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
565         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
566         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
567         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
568 };
569
570 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
571                                      struct nlmsghdr *nlh)
572 {
573         struct net *net = sock_net(in_skb->sk);
574         struct nlattr *tb[NETCONFA_MAX+1];
575         struct netconfmsg *ncm;
576         struct sk_buff *skb;
577         struct ipv6_devconf *devconf;
578         struct inet6_dev *in6_dev;
579         struct net_device *dev;
580         int ifindex;
581         int err;
582
583         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
584                           devconf_ipv6_policy);
585         if (err < 0)
586                 goto errout;
587
588         err = -EINVAL;
589         if (!tb[NETCONFA_IFINDEX])
590                 goto errout;
591
592         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
593         switch (ifindex) {
594         case NETCONFA_IFINDEX_ALL:
595                 devconf = net->ipv6.devconf_all;
596                 break;
597         case NETCONFA_IFINDEX_DEFAULT:
598                 devconf = net->ipv6.devconf_dflt;
599                 break;
600         default:
601                 dev = __dev_get_by_index(net, ifindex);
602                 if (!dev)
603                         goto errout;
604                 in6_dev = __in6_dev_get(dev);
605                 if (!in6_dev)
606                         goto errout;
607                 devconf = &in6_dev->cnf;
608                 break;
609         }
610
611         err = -ENOBUFS;
612         skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
613         if (!skb)
614                 goto errout;
615
616         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
617                                          NETLINK_CB(in_skb).portid,
618                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
619                                          -1);
620         if (err < 0) {
621                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
622                 WARN_ON(err == -EMSGSIZE);
623                 kfree_skb(skb);
624                 goto errout;
625         }
626         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
627 errout:
628         return err;
629 }
630
631 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
632                                       struct netlink_callback *cb)
633 {
634         struct net *net = sock_net(skb->sk);
635         int h, s_h;
636         int idx, s_idx;
637         struct net_device *dev;
638         struct inet6_dev *idev;
639         struct hlist_head *head;
640
641         s_h = cb->args[0];
642         s_idx = idx = cb->args[1];
643
644         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
645                 idx = 0;
646                 head = &net->dev_index_head[h];
647                 rcu_read_lock();
648                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
649                           net->dev_base_seq;
650                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
651                         if (idx < s_idx)
652                                 goto cont;
653                         idev = __in6_dev_get(dev);
654                         if (!idev)
655                                 goto cont;
656
657                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
658                                                        &idev->cnf,
659                                                        NETLINK_CB(cb->skb).portid,
660                                                        cb->nlh->nlmsg_seq,
661                                                        RTM_NEWNETCONF,
662                                                        NLM_F_MULTI,
663                                                        -1) < 0) {
664                                 rcu_read_unlock();
665                                 goto done;
666                         }
667                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
668 cont:
669                         idx++;
670                 }
671                 rcu_read_unlock();
672         }
673         if (h == NETDEV_HASHENTRIES) {
674                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
675                                                net->ipv6.devconf_all,
676                                                NETLINK_CB(cb->skb).portid,
677                                                cb->nlh->nlmsg_seq,
678                                                RTM_NEWNETCONF, NLM_F_MULTI,
679                                                -1) < 0)
680                         goto done;
681                 else
682                         h++;
683         }
684         if (h == NETDEV_HASHENTRIES + 1) {
685                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
686                                                net->ipv6.devconf_dflt,
687                                                NETLINK_CB(cb->skb).portid,
688                                                cb->nlh->nlmsg_seq,
689                                                RTM_NEWNETCONF, NLM_F_MULTI,
690                                                -1) < 0)
691                         goto done;
692                 else
693                         h++;
694         }
695 done:
696         cb->args[0] = h;
697         cb->args[1] = idx;
698
699         return skb->len;
700 }
701
702 #ifdef CONFIG_SYSCTL
703 static void dev_forward_change(struct inet6_dev *idev)
704 {
705         struct net_device *dev;
706         struct inet6_ifaddr *ifa;
707
708         if (!idev)
709                 return;
710         dev = idev->dev;
711         if (idev->cnf.forwarding)
712                 dev_disable_lro(dev);
713         if (dev->flags & IFF_MULTICAST) {
714                 if (idev->cnf.forwarding) {
715                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
716                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
717                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
718                 } else {
719                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
720                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
721                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
722                 }
723         }
724
725         list_for_each_entry(ifa, &idev->addr_list, if_list) {
726                 if (ifa->flags&IFA_F_TENTATIVE)
727                         continue;
728                 if (idev->cnf.forwarding)
729                         addrconf_join_anycast(ifa);
730                 else
731                         addrconf_leave_anycast(ifa);
732         }
733         inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
734                                      dev->ifindex, &idev->cnf);
735 }
736
737
738 static void addrconf_forward_change(struct net *net, __s32 newf)
739 {
740         struct net_device *dev;
741         struct inet6_dev *idev;
742
743         for_each_netdev(net, dev) {
744                 idev = __in6_dev_get(dev);
745                 if (idev) {
746                         int changed = (!idev->cnf.forwarding) ^ (!newf);
747                         idev->cnf.forwarding = newf;
748                         if (changed)
749                                 dev_forward_change(idev);
750                 }
751         }
752 }
753
754 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
755 {
756         struct net *net;
757         int old;
758
759         if (!rtnl_trylock())
760                 return restart_syscall();
761
762         net = (struct net *)table->extra2;
763         old = *p;
764         *p = newf;
765
766         if (p == &net->ipv6.devconf_dflt->forwarding) {
767                 if ((!newf) ^ (!old))
768                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
769                                                      NETCONFA_IFINDEX_DEFAULT,
770                                                      net->ipv6.devconf_dflt);
771                 rtnl_unlock();
772                 return 0;
773         }
774
775         if (p == &net->ipv6.devconf_all->forwarding) {
776                 net->ipv6.devconf_dflt->forwarding = newf;
777                 addrconf_forward_change(net, newf);
778                 if ((!newf) ^ (!old))
779                         inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
780                                                      NETCONFA_IFINDEX_ALL,
781                                                      net->ipv6.devconf_all);
782         } else if ((!newf) ^ (!old))
783                 dev_forward_change((struct inet6_dev *)table->extra1);
784         rtnl_unlock();
785
786         if (newf)
787                 rt6_purge_dflt_routers(net);
788         return 1;
789 }
790
791 static void addrconf_linkdown_change(struct net *net, __s32 newf)
792 {
793         struct net_device *dev;
794         struct inet6_dev *idev;
795
796         for_each_netdev(net, dev) {
797                 idev = __in6_dev_get(dev);
798                 if (idev) {
799                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
800
801                         idev->cnf.ignore_routes_with_linkdown = newf;
802                         if (changed)
803                                 inet6_netconf_notify_devconf(dev_net(dev),
804                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
805                                                              dev->ifindex,
806                                                              &idev->cnf);
807                 }
808         }
809 }
810
811 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
812 {
813         struct net *net;
814         int old;
815
816         if (!rtnl_trylock())
817                 return restart_syscall();
818
819         net = (struct net *)table->extra2;
820         old = *p;
821         *p = newf;
822
823         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
824                 if ((!newf) ^ (!old))
825                         inet6_netconf_notify_devconf(net,
826                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
827                                                      NETCONFA_IFINDEX_DEFAULT,
828                                                      net->ipv6.devconf_dflt);
829                 rtnl_unlock();
830                 return 0;
831         }
832
833         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
834                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
835                 addrconf_linkdown_change(net, newf);
836                 if ((!newf) ^ (!old))
837                         inet6_netconf_notify_devconf(net,
838                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
839                                                      NETCONFA_IFINDEX_ALL,
840                                                      net->ipv6.devconf_all);
841         }
842         rtnl_unlock();
843
844         return 1;
845 }
846
847 #endif
848
849 /* Nobody refers to this ifaddr, destroy it */
850 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
851 {
852         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
853
854 #ifdef NET_REFCNT_DEBUG
855         pr_debug("%s\n", __func__);
856 #endif
857
858         in6_dev_put(ifp->idev);
859
860         if (cancel_delayed_work(&ifp->dad_work))
861                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
862                           ifp);
863
864         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
865                 pr_warn("Freeing alive inet6 address %p\n", ifp);
866                 return;
867         }
868         ip6_rt_put(ifp->rt);
869
870         kfree_rcu(ifp, rcu);
871 }
872
873 static void
874 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
875 {
876         struct list_head *p;
877         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
878
879         /*
880          * Each device address list is sorted in order of scope -
881          * global before linklocal.
882          */
883         list_for_each(p, &idev->addr_list) {
884                 struct inet6_ifaddr *ifa
885                         = list_entry(p, struct inet6_ifaddr, if_list);
886                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
887                         break;
888         }
889
890         list_add_tail(&ifp->if_list, p);
891 }
892
893 static u32 inet6_addr_hash(const struct in6_addr *addr)
894 {
895         return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
896 }
897
898 /* On success it returns ifp with increased reference count */
899
900 static struct inet6_ifaddr *
901 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
902               const struct in6_addr *peer_addr, int pfxlen,
903               int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
904 {
905         struct inet6_ifaddr *ifa = NULL;
906         struct rt6_info *rt;
907         unsigned int hash;
908         int err = 0;
909         int addr_type = ipv6_addr_type(addr);
910
911         if (addr_type == IPV6_ADDR_ANY ||
912             addr_type & IPV6_ADDR_MULTICAST ||
913             (!(idev->dev->flags & IFF_LOOPBACK) &&
914              addr_type & IPV6_ADDR_LOOPBACK))
915                 return ERR_PTR(-EADDRNOTAVAIL);
916
917         rcu_read_lock_bh();
918         if (idev->dead) {
919                 err = -ENODEV;                  /*XXX*/
920                 goto out2;
921         }
922
923         if (idev->cnf.disable_ipv6) {
924                 err = -EACCES;
925                 goto out2;
926         }
927
928         spin_lock(&addrconf_hash_lock);
929
930         /* Ignore adding duplicate addresses on an interface */
931         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
932                 ADBG("ipv6_add_addr: already assigned\n");
933                 err = -EEXIST;
934                 goto out;
935         }
936
937         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
938
939         if (!ifa) {
940                 ADBG("ipv6_add_addr: malloc failed\n");
941                 err = -ENOBUFS;
942                 goto out;
943         }
944
945         rt = addrconf_dst_alloc(idev, addr, false);
946         if (IS_ERR(rt)) {
947                 err = PTR_ERR(rt);
948                 goto out;
949         }
950
951         neigh_parms_data_state_setall(idev->nd_parms);
952
953         ifa->addr = *addr;
954         if (peer_addr)
955                 ifa->peer_addr = *peer_addr;
956
957         spin_lock_init(&ifa->lock);
958         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
959         INIT_HLIST_NODE(&ifa->addr_lst);
960         ifa->scope = scope;
961         ifa->prefix_len = pfxlen;
962         ifa->flags = flags | IFA_F_TENTATIVE;
963         ifa->valid_lft = valid_lft;
964         ifa->prefered_lft = prefered_lft;
965         ifa->cstamp = ifa->tstamp = jiffies;
966         ifa->tokenized = false;
967
968         ifa->rt = rt;
969
970         ifa->idev = idev;
971         in6_dev_hold(idev);
972         /* For caller */
973         in6_ifa_hold(ifa);
974
975         /* Add to big hash table */
976         hash = inet6_addr_hash(addr);
977
978         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
979         spin_unlock(&addrconf_hash_lock);
980
981         write_lock(&idev->lock);
982         /* Add to inet6_dev unicast addr list. */
983         ipv6_link_dev_addr(idev, ifa);
984
985         if (ifa->flags&IFA_F_TEMPORARY) {
986                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
987                 in6_ifa_hold(ifa);
988         }
989
990         in6_ifa_hold(ifa);
991         write_unlock(&idev->lock);
992 out2:
993         rcu_read_unlock_bh();
994
995         if (likely(err == 0))
996                 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
997         else {
998                 kfree(ifa);
999                 ifa = ERR_PTR(err);
1000         }
1001
1002         return ifa;
1003 out:
1004         spin_unlock(&addrconf_hash_lock);
1005         goto out2;
1006 }
1007
1008 enum cleanup_prefix_rt_t {
1009         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1010         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1011         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1012 };
1013
1014 /*
1015  * Check, whether the prefix for ifp would still need a prefix route
1016  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1017  * constants.
1018  *
1019  * 1) we don't purge prefix if address was not permanent.
1020  *    prefix is managed by its own lifetime.
1021  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1022  * 3) if there are no addresses, delete prefix.
1023  * 4) if there are still other permanent address(es),
1024  *    corresponding prefix is still permanent.
1025  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1026  *    don't purge the prefix, assume user space is managing it.
1027  * 6) otherwise, update prefix lifetime to the
1028  *    longest valid lifetime among the corresponding
1029  *    addresses on the device.
1030  *    Note: subsequent RA will update lifetime.
1031  **/
1032 static enum cleanup_prefix_rt_t
1033 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1034 {
1035         struct inet6_ifaddr *ifa;
1036         struct inet6_dev *idev = ifp->idev;
1037         unsigned long lifetime;
1038         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1039
1040         *expires = jiffies;
1041
1042         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1043                 if (ifa == ifp)
1044                         continue;
1045                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1046                                        ifp->prefix_len))
1047                         continue;
1048                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1049                         return CLEANUP_PREFIX_RT_NOP;
1050
1051                 action = CLEANUP_PREFIX_RT_EXPIRE;
1052
1053                 spin_lock(&ifa->lock);
1054
1055                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1056                 /*
1057                  * Note: Because this address is
1058                  * not permanent, lifetime <
1059                  * LONG_MAX / HZ here.
1060                  */
1061                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1062                         *expires = ifa->tstamp + lifetime * HZ;
1063                 spin_unlock(&ifa->lock);
1064         }
1065
1066         return action;
1067 }
1068
1069 static void
1070 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1071 {
1072         struct rt6_info *rt;
1073
1074         rt = addrconf_get_prefix_route(&ifp->addr,
1075                                        ifp->prefix_len,
1076                                        ifp->idev->dev,
1077                                        0, RTF_GATEWAY | RTF_DEFAULT);
1078         if (rt) {
1079                 if (del_rt)
1080                         ip6_del_rt(rt);
1081                 else {
1082                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1083                                 rt6_set_expires(rt, expires);
1084                         ip6_rt_put(rt);
1085                 }
1086         }
1087 }
1088
1089
1090 /* This function wants to get referenced ifp and releases it before return */
1091
1092 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1093 {
1094         int state;
1095         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1096         unsigned long expires;
1097
1098         ASSERT_RTNL();
1099
1100         spin_lock_bh(&ifp->lock);
1101         state = ifp->state;
1102         ifp->state = INET6_IFADDR_STATE_DEAD;
1103         spin_unlock_bh(&ifp->lock);
1104
1105         if (state == INET6_IFADDR_STATE_DEAD)
1106                 goto out;
1107
1108         spin_lock_bh(&addrconf_hash_lock);
1109         hlist_del_init_rcu(&ifp->addr_lst);
1110         spin_unlock_bh(&addrconf_hash_lock);
1111
1112         write_lock_bh(&ifp->idev->lock);
1113
1114         if (ifp->flags&IFA_F_TEMPORARY) {
1115                 list_del(&ifp->tmp_list);
1116                 if (ifp->ifpub) {
1117                         in6_ifa_put(ifp->ifpub);
1118                         ifp->ifpub = NULL;
1119                 }
1120                 __in6_ifa_put(ifp);
1121         }
1122
1123         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1124                 action = check_cleanup_prefix_route(ifp, &expires);
1125
1126         list_del_init(&ifp->if_list);
1127         __in6_ifa_put(ifp);
1128
1129         write_unlock_bh(&ifp->idev->lock);
1130
1131         addrconf_del_dad_work(ifp);
1132
1133         ipv6_ifa_notify(RTM_DELADDR, ifp);
1134
1135         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1136
1137         if (action != CLEANUP_PREFIX_RT_NOP) {
1138                 cleanup_prefix_route(ifp, expires,
1139                         action == CLEANUP_PREFIX_RT_DEL);
1140         }
1141
1142         /* clean up prefsrc entries */
1143         rt6_remove_prefsrc(ifp);
1144 out:
1145         in6_ifa_put(ifp);
1146 }
1147
1148 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1149 {
1150         struct inet6_dev *idev = ifp->idev;
1151         struct in6_addr addr, *tmpaddr;
1152         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1153         unsigned long regen_advance;
1154         int tmp_plen;
1155         int ret = 0;
1156         u32 addr_flags;
1157         unsigned long now = jiffies;
1158
1159         write_lock_bh(&idev->lock);
1160         if (ift) {
1161                 spin_lock_bh(&ift->lock);
1162                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1163                 spin_unlock_bh(&ift->lock);
1164                 tmpaddr = &addr;
1165         } else {
1166                 tmpaddr = NULL;
1167         }
1168 retry:
1169         in6_dev_hold(idev);
1170         if (idev->cnf.use_tempaddr <= 0) {
1171                 write_unlock_bh(&idev->lock);
1172                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1173                 in6_dev_put(idev);
1174                 ret = -1;
1175                 goto out;
1176         }
1177         spin_lock_bh(&ifp->lock);
1178         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1179                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1180                 spin_unlock_bh(&ifp->lock);
1181                 write_unlock_bh(&idev->lock);
1182                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1183                         __func__);
1184                 in6_dev_put(idev);
1185                 ret = -1;
1186                 goto out;
1187         }
1188         in6_ifa_hold(ifp);
1189         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1190         __ipv6_try_regen_rndid(idev, tmpaddr);
1191         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1192         age = (now - ifp->tstamp) / HZ;
1193         tmp_valid_lft = min_t(__u32,
1194                               ifp->valid_lft,
1195                               idev->cnf.temp_valid_lft + age);
1196         tmp_prefered_lft = min_t(__u32,
1197                                  ifp->prefered_lft,
1198                                  idev->cnf.temp_prefered_lft + age -
1199                                  idev->cnf.max_desync_factor);
1200         tmp_plen = ifp->prefix_len;
1201         tmp_tstamp = ifp->tstamp;
1202         spin_unlock_bh(&ifp->lock);
1203
1204         regen_advance = idev->cnf.regen_max_retry *
1205                         idev->cnf.dad_transmits *
1206                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1207         write_unlock_bh(&idev->lock);
1208
1209         /* A temporary address is created only if this calculated Preferred
1210          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1211          * an implementation must not create a temporary address with a zero
1212          * Preferred Lifetime.
1213          * Use age calculation as in addrconf_verify to avoid unnecessary
1214          * temporary addresses being generated.
1215          */
1216         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1217         if (tmp_prefered_lft <= regen_advance + age) {
1218                 in6_ifa_put(ifp);
1219                 in6_dev_put(idev);
1220                 ret = -1;
1221                 goto out;
1222         }
1223
1224         addr_flags = IFA_F_TEMPORARY;
1225         /* set in addrconf_prefix_rcv() */
1226         if (ifp->flags & IFA_F_OPTIMISTIC)
1227                 addr_flags |= IFA_F_OPTIMISTIC;
1228
1229         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1230                             ipv6_addr_scope(&addr), addr_flags,
1231                             tmp_valid_lft, tmp_prefered_lft);
1232         if (IS_ERR(ift)) {
1233                 in6_ifa_put(ifp);
1234                 in6_dev_put(idev);
1235                 pr_info("%s: retry temporary address regeneration\n", __func__);
1236                 tmpaddr = &addr;
1237                 write_lock_bh(&idev->lock);
1238                 goto retry;
1239         }
1240
1241         spin_lock_bh(&ift->lock);
1242         ift->ifpub = ifp;
1243         ift->cstamp = now;
1244         ift->tstamp = tmp_tstamp;
1245         spin_unlock_bh(&ift->lock);
1246
1247         addrconf_dad_start(ift);
1248         in6_ifa_put(ift);
1249         in6_dev_put(idev);
1250 out:
1251         return ret;
1252 }
1253
1254 /*
1255  *      Choose an appropriate source address (RFC3484)
1256  */
1257 enum {
1258         IPV6_SADDR_RULE_INIT = 0,
1259         IPV6_SADDR_RULE_LOCAL,
1260         IPV6_SADDR_RULE_SCOPE,
1261         IPV6_SADDR_RULE_PREFERRED,
1262 #ifdef CONFIG_IPV6_MIP6
1263         IPV6_SADDR_RULE_HOA,
1264 #endif
1265         IPV6_SADDR_RULE_OIF,
1266         IPV6_SADDR_RULE_LABEL,
1267         IPV6_SADDR_RULE_PRIVACY,
1268         IPV6_SADDR_RULE_ORCHID,
1269         IPV6_SADDR_RULE_PREFIX,
1270 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1271         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1272 #endif
1273         IPV6_SADDR_RULE_MAX
1274 };
1275
1276 struct ipv6_saddr_score {
1277         int                     rule;
1278         int                     addr_type;
1279         struct inet6_ifaddr     *ifa;
1280         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1281         int                     scopedist;
1282         int                     matchlen;
1283 };
1284
1285 struct ipv6_saddr_dst {
1286         const struct in6_addr *addr;
1287         int ifindex;
1288         int scope;
1289         int label;
1290         unsigned int prefs;
1291 };
1292
1293 static inline int ipv6_saddr_preferred(int type)
1294 {
1295         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1296                 return 1;
1297         return 0;
1298 }
1299
1300 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1301 {
1302 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1303         return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1304 #else
1305         return false;
1306 #endif
1307 }
1308
1309 static int ipv6_get_saddr_eval(struct net *net,
1310                                struct ipv6_saddr_score *score,
1311                                struct ipv6_saddr_dst *dst,
1312                                int i)
1313 {
1314         int ret;
1315
1316         if (i <= score->rule) {
1317                 switch (i) {
1318                 case IPV6_SADDR_RULE_SCOPE:
1319                         ret = score->scopedist;
1320                         break;
1321                 case IPV6_SADDR_RULE_PREFIX:
1322                         ret = score->matchlen;
1323                         break;
1324                 default:
1325                         ret = !!test_bit(i, score->scorebits);
1326                 }
1327                 goto out;
1328         }
1329
1330         switch (i) {
1331         case IPV6_SADDR_RULE_INIT:
1332                 /* Rule 0: remember if hiscore is not ready yet */
1333                 ret = !!score->ifa;
1334                 break;
1335         case IPV6_SADDR_RULE_LOCAL:
1336                 /* Rule 1: Prefer same address */
1337                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1338                 break;
1339         case IPV6_SADDR_RULE_SCOPE:
1340                 /* Rule 2: Prefer appropriate scope
1341                  *
1342                  *      ret
1343                  *       ^
1344                  *    -1 |  d 15
1345                  *    ---+--+-+---> scope
1346                  *       |
1347                  *       |             d is scope of the destination.
1348                  *  B-d  |  \
1349                  *       |   \      <- smaller scope is better if
1350                  *  B-15 |    \        if scope is enough for destination.
1351                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1352                  * d-C-1 | /
1353                  *       |/         <- greater is better
1354                  *   -C  /             if scope is not enough for destination.
1355                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1356                  *
1357                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1358                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1359                  * Assume B = 0 and we get C > 29.
1360                  */
1361                 ret = __ipv6_addr_src_scope(score->addr_type);
1362                 if (ret >= dst->scope)
1363                         ret = -ret;
1364                 else
1365                         ret -= 128;     /* 30 is enough */
1366                 score->scopedist = ret;
1367                 break;
1368         case IPV6_SADDR_RULE_PREFERRED:
1369             {
1370                 /* Rule 3: Avoid deprecated and optimistic addresses */
1371                 u8 avoid = IFA_F_DEPRECATED;
1372
1373                 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1374                         avoid |= IFA_F_OPTIMISTIC;
1375                 ret = ipv6_saddr_preferred(score->addr_type) ||
1376                       !(score->ifa->flags & avoid);
1377                 break;
1378             }
1379 #ifdef CONFIG_IPV6_MIP6
1380         case IPV6_SADDR_RULE_HOA:
1381             {
1382                 /* Rule 4: Prefer home address */
1383                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1384                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1385                 break;
1386             }
1387 #endif
1388         case IPV6_SADDR_RULE_OIF:
1389                 /* Rule 5: Prefer outgoing interface */
1390                 ret = (!dst->ifindex ||
1391                        dst->ifindex == score->ifa->idev->dev->ifindex);
1392                 break;
1393         case IPV6_SADDR_RULE_LABEL:
1394                 /* Rule 6: Prefer matching label */
1395                 ret = ipv6_addr_label(net,
1396                                       &score->ifa->addr, score->addr_type,
1397                                       score->ifa->idev->dev->ifindex) == dst->label;
1398                 break;
1399         case IPV6_SADDR_RULE_PRIVACY:
1400             {
1401                 /* Rule 7: Prefer public address
1402                  * Note: prefer temporary address if use_tempaddr >= 2
1403                  */
1404                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1405                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1406                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1407                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1408                 break;
1409             }
1410         case IPV6_SADDR_RULE_ORCHID:
1411                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1412                  *          non-ORCHID vs non-ORCHID
1413                  */
1414                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1415                         ipv6_addr_orchid(dst->addr));
1416                 break;
1417         case IPV6_SADDR_RULE_PREFIX:
1418                 /* Rule 8: Use longest matching prefix */
1419                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1420                 if (ret > score->ifa->prefix_len)
1421                         ret = score->ifa->prefix_len;
1422                 score->matchlen = ret;
1423                 break;
1424 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1425         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1426                 /* Optimistic addresses still have lower precedence than other
1427                  * preferred addresses.
1428                  */
1429                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1430                 break;
1431 #endif
1432         default:
1433                 ret = 0;
1434         }
1435
1436         if (ret)
1437                 __set_bit(i, score->scorebits);
1438         score->rule = i;
1439 out:
1440         return ret;
1441 }
1442
1443 static int __ipv6_dev_get_saddr(struct net *net,
1444                                 struct ipv6_saddr_dst *dst,
1445                                 struct inet6_dev *idev,
1446                                 struct ipv6_saddr_score *scores,
1447                                 int hiscore_idx)
1448 {
1449         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1450
1451         read_lock_bh(&idev->lock);
1452         list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1453                 int i;
1454
1455                 /*
1456                  * - Tentative Address (RFC2462 section 5.4)
1457                  *  - A tentative address is not considered
1458                  *    "assigned to an interface" in the traditional
1459                  *    sense, unless it is also flagged as optimistic.
1460                  * - Candidate Source Address (section 4)
1461                  *  - In any case, anycast addresses, multicast
1462                  *    addresses, and the unspecified address MUST
1463                  *    NOT be included in a candidate set.
1464                  */
1465                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1466                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1467                         continue;
1468
1469                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1470
1471                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1472                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1473                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1474                                             idev->dev->name);
1475                         continue;
1476                 }
1477
1478                 score->rule = -1;
1479                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1480
1481                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1482                         int minihiscore, miniscore;
1483
1484                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1485                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1486
1487                         if (minihiscore > miniscore) {
1488                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1489                                     score->scopedist > 0) {
1490                                         /*
1491                                          * special case:
1492                                          * each remaining entry
1493                                          * has too small (not enough)
1494                                          * scope, because ifa entries
1495                                          * are sorted by their scope
1496                                          * values.
1497                                          */
1498                                         goto out;
1499                                 }
1500                                 break;
1501                         } else if (minihiscore < miniscore) {
1502                                 if (hiscore->ifa)
1503                                         in6_ifa_put(hiscore->ifa);
1504
1505                                 in6_ifa_hold(score->ifa);
1506
1507                                 swap(hiscore, score);
1508                                 hiscore_idx = 1 - hiscore_idx;
1509
1510                                 /* restore our iterator */
1511                                 score->ifa = hiscore->ifa;
1512
1513                                 break;
1514                         }
1515                 }
1516         }
1517 out:
1518         read_unlock_bh(&idev->lock);
1519         return hiscore_idx;
1520 }
1521
1522 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1523                        const struct in6_addr *daddr, unsigned int prefs,
1524                        struct in6_addr *saddr)
1525 {
1526         struct ipv6_saddr_score scores[2], *hiscore;
1527         struct ipv6_saddr_dst dst;
1528         struct inet6_dev *idev;
1529         struct net_device *dev;
1530         int dst_type;
1531         bool use_oif_addr = false;
1532         int hiscore_idx = 0;
1533
1534         dst_type = __ipv6_addr_type(daddr);
1535         dst.addr = daddr;
1536         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1537         dst.scope = __ipv6_addr_src_scope(dst_type);
1538         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1539         dst.prefs = prefs;
1540
1541         scores[hiscore_idx].rule = -1;
1542         scores[hiscore_idx].ifa = NULL;
1543
1544         rcu_read_lock();
1545
1546         /* Candidate Source Address (section 4)
1547          *  - multicast and link-local destination address,
1548          *    the set of candidate source address MUST only
1549          *    include addresses assigned to interfaces
1550          *    belonging to the same link as the outgoing
1551          *    interface.
1552          * (- For site-local destination addresses, the
1553          *    set of candidate source addresses MUST only
1554          *    include addresses assigned to interfaces
1555          *    belonging to the same site as the outgoing
1556          *    interface.)
1557          *  - "It is RECOMMENDED that the candidate source addresses
1558          *    be the set of unicast addresses assigned to the
1559          *    interface that will be used to send to the destination
1560          *    (the 'outgoing' interface)." (RFC 6724)
1561          */
1562         if (dst_dev) {
1563                 idev = __in6_dev_get(dst_dev);
1564                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1565                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1566                     (idev && idev->cnf.use_oif_addrs_only)) {
1567                         use_oif_addr = true;
1568                 }
1569         }
1570
1571         if (use_oif_addr) {
1572                 if (idev)
1573                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1574         } else {
1575                 for_each_netdev_rcu(net, dev) {
1576                         idev = __in6_dev_get(dev);
1577                         if (!idev)
1578                                 continue;
1579                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1580                 }
1581         }
1582         rcu_read_unlock();
1583
1584         hiscore = &scores[hiscore_idx];
1585         if (!hiscore->ifa)
1586                 return -EADDRNOTAVAIL;
1587
1588         *saddr = hiscore->ifa->addr;
1589         in6_ifa_put(hiscore->ifa);
1590         return 0;
1591 }
1592 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1593
1594 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1595                       u32 banned_flags)
1596 {
1597         struct inet6_ifaddr *ifp;
1598         int err = -EADDRNOTAVAIL;
1599
1600         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1601                 if (ifp->scope > IFA_LINK)
1602                         break;
1603                 if (ifp->scope == IFA_LINK &&
1604                     !(ifp->flags & banned_flags)) {
1605                         *addr = ifp->addr;
1606                         err = 0;
1607                         break;
1608                 }
1609         }
1610         return err;
1611 }
1612
1613 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1614                     u32 banned_flags)
1615 {
1616         struct inet6_dev *idev;
1617         int err = -EADDRNOTAVAIL;
1618
1619         rcu_read_lock();
1620         idev = __in6_dev_get(dev);
1621         if (idev) {
1622                 read_lock_bh(&idev->lock);
1623                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1624                 read_unlock_bh(&idev->lock);
1625         }
1626         rcu_read_unlock();
1627         return err;
1628 }
1629
1630 static int ipv6_count_addresses(struct inet6_dev *idev)
1631 {
1632         int cnt = 0;
1633         struct inet6_ifaddr *ifp;
1634
1635         read_lock_bh(&idev->lock);
1636         list_for_each_entry(ifp, &idev->addr_list, if_list)
1637                 cnt++;
1638         read_unlock_bh(&idev->lock);
1639         return cnt;
1640 }
1641
1642 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1643                   const struct net_device *dev, int strict)
1644 {
1645         return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1646 }
1647 EXPORT_SYMBOL(ipv6_chk_addr);
1648
1649 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1650                             const struct net_device *dev, int strict,
1651                             u32 banned_flags)
1652 {
1653         struct inet6_ifaddr *ifp;
1654         unsigned int hash = inet6_addr_hash(addr);
1655         u32 ifp_flags;
1656
1657         rcu_read_lock_bh();
1658         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1659                 if (!net_eq(dev_net(ifp->idev->dev), net))
1660                         continue;
1661                 /* Decouple optimistic from tentative for evaluation here.
1662                  * Ban optimistic addresses explicitly, when required.
1663                  */
1664                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1665                             ? (ifp->flags&~IFA_F_TENTATIVE)
1666                             : ifp->flags;
1667                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1668                     !(ifp_flags&banned_flags) &&
1669                     (!dev || ifp->idev->dev == dev ||
1670                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1671                         rcu_read_unlock_bh();
1672                         return 1;
1673                 }
1674         }
1675
1676         rcu_read_unlock_bh();
1677         return 0;
1678 }
1679 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1680
1681 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1682                                struct net_device *dev)
1683 {
1684         unsigned int hash = inet6_addr_hash(addr);
1685         struct inet6_ifaddr *ifp;
1686
1687         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1688                 if (!net_eq(dev_net(ifp->idev->dev), net))
1689                         continue;
1690                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1691                         if (!dev || ifp->idev->dev == dev)
1692                                 return true;
1693                 }
1694         }
1695         return false;
1696 }
1697
1698 /* Compares an address/prefix_len with addresses on device @dev.
1699  * If one is found it returns true.
1700  */
1701 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1702         const unsigned int prefix_len, struct net_device *dev)
1703 {
1704         struct inet6_dev *idev;
1705         struct inet6_ifaddr *ifa;
1706         bool ret = false;
1707
1708         rcu_read_lock();
1709         idev = __in6_dev_get(dev);
1710         if (idev) {
1711                 read_lock_bh(&idev->lock);
1712                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1713                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1714                         if (ret)
1715                                 break;
1716                 }
1717                 read_unlock_bh(&idev->lock);
1718         }
1719         rcu_read_unlock();
1720
1721         return ret;
1722 }
1723 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1724
1725 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1726 {
1727         struct inet6_dev *idev;
1728         struct inet6_ifaddr *ifa;
1729         int     onlink;
1730
1731         onlink = 0;
1732         rcu_read_lock();
1733         idev = __in6_dev_get(dev);
1734         if (idev) {
1735                 read_lock_bh(&idev->lock);
1736                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1737                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1738                                                    ifa->prefix_len);
1739                         if (onlink)
1740                                 break;
1741                 }
1742                 read_unlock_bh(&idev->lock);
1743         }
1744         rcu_read_unlock();
1745         return onlink;
1746 }
1747 EXPORT_SYMBOL(ipv6_chk_prefix);
1748
1749 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1750                                      struct net_device *dev, int strict)
1751 {
1752         struct inet6_ifaddr *ifp, *result = NULL;
1753         unsigned int hash = inet6_addr_hash(addr);
1754
1755         rcu_read_lock_bh();
1756         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1757                 if (!net_eq(dev_net(ifp->idev->dev), net))
1758                         continue;
1759                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1760                         if (!dev || ifp->idev->dev == dev ||
1761                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1762                                 result = ifp;
1763                                 in6_ifa_hold(ifp);
1764                                 break;
1765                         }
1766                 }
1767         }
1768         rcu_read_unlock_bh();
1769
1770         return result;
1771 }
1772
1773 /* Gets referenced address, destroys ifaddr */
1774
1775 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1776 {
1777         if (ifp->flags&IFA_F_PERMANENT) {
1778                 spin_lock_bh(&ifp->lock);
1779                 addrconf_del_dad_work(ifp);
1780                 ifp->flags |= IFA_F_TENTATIVE;
1781                 if (dad_failed)
1782                         ifp->flags |= IFA_F_DADFAILED;
1783                 spin_unlock_bh(&ifp->lock);
1784                 if (dad_failed)
1785                         ipv6_ifa_notify(0, ifp);
1786                 in6_ifa_put(ifp);
1787         } else if (ifp->flags&IFA_F_TEMPORARY) {
1788                 struct inet6_ifaddr *ifpub;
1789                 spin_lock_bh(&ifp->lock);
1790                 ifpub = ifp->ifpub;
1791                 if (ifpub) {
1792                         in6_ifa_hold(ifpub);
1793                         spin_unlock_bh(&ifp->lock);
1794                         ipv6_create_tempaddr(ifpub, ifp);
1795                         in6_ifa_put(ifpub);
1796                 } else {
1797                         spin_unlock_bh(&ifp->lock);
1798                 }
1799                 ipv6_del_addr(ifp);
1800         } else {
1801                 ipv6_del_addr(ifp);
1802         }
1803 }
1804
1805 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1806 {
1807         int err = -ENOENT;
1808
1809         spin_lock_bh(&ifp->lock);
1810         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1811                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1812                 err = 0;
1813         }
1814         spin_unlock_bh(&ifp->lock);
1815
1816         return err;
1817 }
1818
1819 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1820 {
1821         struct in6_addr addr;
1822         struct inet6_dev *idev = ifp->idev;
1823         struct net *net = dev_net(ifp->idev->dev);
1824
1825         if (addrconf_dad_end(ifp)) {
1826                 in6_ifa_put(ifp);
1827                 return;
1828         }
1829
1830         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1831                              ifp->idev->dev->name, &ifp->addr);
1832
1833         spin_lock_bh(&ifp->lock);
1834
1835         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1836                 int scope = ifp->scope;
1837                 u32 flags = ifp->flags;
1838                 struct in6_addr new_addr;
1839                 struct inet6_ifaddr *ifp2;
1840                 u32 valid_lft, preferred_lft;
1841                 int pfxlen = ifp->prefix_len;
1842                 int retries = ifp->stable_privacy_retry + 1;
1843
1844                 if (retries > net->ipv6.sysctl.idgen_retries) {
1845                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1846                                              ifp->idev->dev->name);
1847                         goto errdad;
1848                 }
1849
1850                 new_addr = ifp->addr;
1851                 if (ipv6_generate_stable_address(&new_addr, retries,
1852                                                  idev))
1853                         goto errdad;
1854
1855                 valid_lft = ifp->valid_lft;
1856                 preferred_lft = ifp->prefered_lft;
1857
1858                 spin_unlock_bh(&ifp->lock);
1859
1860                 if (idev->cnf.max_addresses &&
1861                     ipv6_count_addresses(idev) >=
1862                     idev->cnf.max_addresses)
1863                         goto lock_errdad;
1864
1865                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1866                                      ifp->idev->dev->name);
1867
1868                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1869                                      scope, flags, valid_lft,
1870                                      preferred_lft);
1871                 if (IS_ERR(ifp2))
1872                         goto lock_errdad;
1873
1874                 spin_lock_bh(&ifp2->lock);
1875                 ifp2->stable_privacy_retry = retries;
1876                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1877                 spin_unlock_bh(&ifp2->lock);
1878
1879                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1880                 in6_ifa_put(ifp2);
1881 lock_errdad:
1882                 spin_lock_bh(&ifp->lock);
1883         } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1884                 addr.s6_addr32[0] = htonl(0xfe800000);
1885                 addr.s6_addr32[1] = 0;
1886
1887                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1888                     ipv6_addr_equal(&ifp->addr, &addr)) {
1889                         /* DAD failed for link-local based on MAC address */
1890                         idev->cnf.disable_ipv6 = 1;
1891
1892                         pr_info("%s: IPv6 being disabled!\n",
1893                                 ifp->idev->dev->name);
1894                 }
1895         }
1896
1897 errdad:
1898         /* transition from _POSTDAD to _ERRDAD */
1899         ifp->state = INET6_IFADDR_STATE_ERRDAD;
1900         spin_unlock_bh(&ifp->lock);
1901
1902         addrconf_mod_dad_work(ifp, 0);
1903 }
1904
1905 /* Join to solicited addr multicast group.
1906  * caller must hold RTNL */
1907 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1908 {
1909         struct in6_addr maddr;
1910
1911         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1912                 return;
1913
1914         addrconf_addr_solict_mult(addr, &maddr);
1915         ipv6_dev_mc_inc(dev, &maddr);
1916 }
1917
1918 /* caller must hold RTNL */
1919 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1920 {
1921         struct in6_addr maddr;
1922
1923         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1924                 return;
1925
1926         addrconf_addr_solict_mult(addr, &maddr);
1927         __ipv6_dev_mc_dec(idev, &maddr);
1928 }
1929
1930 /* caller must hold RTNL */
1931 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1932 {
1933         struct in6_addr addr;
1934
1935         if (ifp->prefix_len >= 127) /* RFC 6164 */
1936                 return;
1937         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1938         if (ipv6_addr_any(&addr))
1939                 return;
1940         __ipv6_dev_ac_inc(ifp->idev, &addr);
1941 }
1942
1943 /* caller must hold RTNL */
1944 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1945 {
1946         struct in6_addr addr;
1947
1948         if (ifp->prefix_len >= 127) /* RFC 6164 */
1949                 return;
1950         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1951         if (ipv6_addr_any(&addr))
1952                 return;
1953         __ipv6_dev_ac_dec(ifp->idev, &addr);
1954 }
1955
1956 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1957 {
1958         if (dev->addr_len != IEEE802154_ADDR_LEN)
1959                 return -1;
1960         memcpy(eui, dev->dev_addr, 8);
1961         eui[0] ^= 2;
1962         return 0;
1963 }
1964
1965 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1966 {
1967         union fwnet_hwaddr *ha;
1968
1969         if (dev->addr_len != FWNET_ALEN)
1970                 return -1;
1971
1972         ha = (union fwnet_hwaddr *)dev->dev_addr;
1973
1974         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1975         eui[0] ^= 2;
1976         return 0;
1977 }
1978
1979 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1980 {
1981         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1982         if (dev->addr_len != ARCNET_ALEN)
1983                 return -1;
1984         memset(eui, 0, 7);
1985         eui[7] = *(u8 *)dev->dev_addr;
1986         return 0;
1987 }
1988
1989 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1990 {
1991         if (dev->addr_len != INFINIBAND_ALEN)
1992                 return -1;
1993         memcpy(eui, dev->dev_addr + 12, 8);
1994         eui[0] |= 2;
1995         return 0;
1996 }
1997
1998 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1999 {
2000         if (addr == 0)
2001                 return -1;
2002         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2003                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2004                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2005                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2006                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2007                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2008         eui[1] = 0;
2009         eui[2] = 0x5E;
2010         eui[3] = 0xFE;
2011         memcpy(eui + 4, &addr, 4);
2012         return 0;
2013 }
2014
2015 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2016 {
2017         if (dev->priv_flags & IFF_ISATAP)
2018                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2019         return -1;
2020 }
2021
2022 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2023 {
2024         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2025 }
2026
2027 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2028 {
2029         memcpy(eui, dev->perm_addr, 3);
2030         memcpy(eui + 5, dev->perm_addr + 3, 3);
2031         eui[3] = 0xFF;
2032         eui[4] = 0xFE;
2033         eui[0] ^= 2;
2034         return 0;
2035 }
2036
2037 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2038 {
2039         switch (dev->type) {
2040         case ARPHRD_ETHER:
2041         case ARPHRD_FDDI:
2042                 return addrconf_ifid_eui48(eui, dev);
2043         case ARPHRD_ARCNET:
2044                 return addrconf_ifid_arcnet(eui, dev);
2045         case ARPHRD_INFINIBAND:
2046                 return addrconf_ifid_infiniband(eui, dev);
2047         case ARPHRD_SIT:
2048                 return addrconf_ifid_sit(eui, dev);
2049         case ARPHRD_IPGRE:
2050                 return addrconf_ifid_gre(eui, dev);
2051         case ARPHRD_6LOWPAN:
2052         case ARPHRD_IEEE802154:
2053                 return addrconf_ifid_eui64(eui, dev);
2054         case ARPHRD_IEEE1394:
2055                 return addrconf_ifid_ieee1394(eui, dev);
2056         case ARPHRD_TUNNEL6:
2057                 return addrconf_ifid_ip6tnl(eui, dev);
2058         }
2059         return -1;
2060 }
2061
2062 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2063 {
2064         int err = -1;
2065         struct inet6_ifaddr *ifp;
2066
2067         read_lock_bh(&idev->lock);
2068         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2069                 if (ifp->scope > IFA_LINK)
2070                         break;
2071                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2072                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2073                         err = 0;
2074                         break;
2075                 }
2076         }
2077         read_unlock_bh(&idev->lock);
2078         return err;
2079 }
2080
2081 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2082 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2083 {
2084 regen:
2085         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2086         idev->rndid[0] &= ~0x02;
2087
2088         /*
2089          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2090          * check if generated address is not inappropriate
2091          *
2092          *  - Reserved subnet anycast (RFC 2526)
2093          *      11111101 11....11 1xxxxxxx
2094          *  - ISATAP (RFC4214) 6.1
2095          *      00-00-5E-FE-xx-xx-xx-xx
2096          *  - value 0
2097          *  - XXX: already assigned to an address on the device
2098          */
2099         if (idev->rndid[0] == 0xfd &&
2100             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2101             (idev->rndid[7]&0x80))
2102                 goto regen;
2103         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2104                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2105                         goto regen;
2106                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2107                         goto regen;
2108         }
2109 }
2110
2111 static void ipv6_regen_rndid(unsigned long data)
2112 {
2113         struct inet6_dev *idev = (struct inet6_dev *) data;
2114         unsigned long expires;
2115
2116         rcu_read_lock_bh();
2117         write_lock_bh(&idev->lock);
2118
2119         if (idev->dead)
2120                 goto out;
2121
2122         __ipv6_regen_rndid(idev);
2123
2124         expires = jiffies +
2125                 idev->cnf.temp_prefered_lft * HZ -
2126                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2127                 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2128                 idev->cnf.max_desync_factor * HZ;
2129         if (time_before(expires, jiffies)) {
2130                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2131                         __func__, idev->dev->name);
2132                 goto out;
2133         }
2134
2135         if (!mod_timer(&idev->regen_timer, expires))
2136                 in6_dev_hold(idev);
2137
2138 out:
2139         write_unlock_bh(&idev->lock);
2140         rcu_read_unlock_bh();
2141         in6_dev_put(idev);
2142 }
2143
2144 static void  __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2145 {
2146         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2147                 __ipv6_regen_rndid(idev);
2148 }
2149
2150 u32 addrconf_rt_table(const struct net_device *dev, u32 default_table) {
2151         /* Determines into what table to put autoconf PIO/RIO/default routes
2152          * learned on this device.
2153          *
2154          * - If 0, use the same table for every device. This puts routes into
2155          *   one of RT_TABLE_{PREFIX,INFO,DFLT} depending on the type of route
2156          *   (but note that these three are currently all equal to
2157          *   RT6_TABLE_MAIN).
2158          * - If > 0, use the specified table.
2159          * - If < 0, put routes into table dev->ifindex + (-rt_table).
2160          */
2161         struct inet6_dev *idev = in6_dev_get(dev);
2162         u32 table;
2163         int sysctl = idev->cnf.accept_ra_rt_table;
2164         if (sysctl == 0) {
2165                 table = default_table;
2166         } else if (sysctl > 0) {
2167                 table = (u32) sysctl;
2168         } else {
2169                 table = (unsigned) dev->ifindex + (-sysctl);
2170         }
2171         in6_dev_put(idev);
2172         return table;
2173 }
2174
2175 /*
2176  *      Add prefix route.
2177  */
2178
2179 static void
2180 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2181                       unsigned long expires, u32 flags)
2182 {
2183         struct fib6_config cfg = {
2184                 .fc_table = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_PREFIX),
2185                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2186                 .fc_ifindex = dev->ifindex,
2187                 .fc_expires = expires,
2188                 .fc_dst_len = plen,
2189                 .fc_flags = RTF_UP | flags,
2190                 .fc_nlinfo.nl_net = dev_net(dev),
2191                 .fc_protocol = RTPROT_KERNEL,
2192         };
2193
2194         cfg.fc_dst = *pfx;
2195
2196         /* Prevent useless cloning on PtP SIT.
2197            This thing is done here expecting that the whole
2198            class of non-broadcast devices need not cloning.
2199          */
2200 #if IS_ENABLED(CONFIG_IPV6_SIT)
2201         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2202                 cfg.fc_flags |= RTF_NONEXTHOP;
2203 #endif
2204
2205         ip6_route_add(&cfg);
2206 }
2207
2208
2209 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2210                                                   int plen,
2211                                                   const struct net_device *dev,
2212                                                   u32 flags, u32 noflags)
2213 {
2214         struct fib6_node *fn;
2215         struct rt6_info *rt = NULL;
2216         struct fib6_table *table;
2217         u32 tb_id = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_PREFIX);
2218
2219         table = fib6_get_table(dev_net(dev), tb_id);
2220         if (!table)
2221                 return NULL;
2222
2223         read_lock_bh(&table->tb6_lock);
2224         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2225         if (!fn)
2226                 goto out;
2227
2228         noflags |= RTF_CACHE;
2229         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2230                 if (rt->dst.dev->ifindex != dev->ifindex)
2231                         continue;
2232                 if ((rt->rt6i_flags & flags) != flags)
2233                         continue;
2234                 if ((rt->rt6i_flags & noflags) != 0)
2235                         continue;
2236                 dst_hold(&rt->dst);
2237                 break;
2238         }
2239 out:
2240         read_unlock_bh(&table->tb6_lock);
2241         return rt;
2242 }
2243
2244
2245 /* Create "default" multicast route to the interface */
2246
2247 static void addrconf_add_mroute(struct net_device *dev)
2248 {
2249         struct fib6_config cfg = {
2250                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2251                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2252                 .fc_ifindex = dev->ifindex,
2253                 .fc_dst_len = 8,
2254                 .fc_flags = RTF_UP,
2255                 .fc_nlinfo.nl_net = dev_net(dev),
2256         };
2257
2258         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2259
2260         ip6_route_add(&cfg);
2261 }
2262
2263 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2264 {
2265         struct inet6_dev *idev;
2266
2267         ASSERT_RTNL();
2268
2269         idev = ipv6_find_idev(dev);
2270         if (!idev)
2271                 return ERR_PTR(-ENOBUFS);
2272
2273         if (idev->cnf.disable_ipv6)
2274                 return ERR_PTR(-EACCES);
2275
2276         /* Add default multicast route */
2277         if (!(dev->flags & IFF_LOOPBACK))
2278                 addrconf_add_mroute(dev);
2279
2280         return idev;
2281 }
2282
2283 static void manage_tempaddrs(struct inet6_dev *idev,
2284                              struct inet6_ifaddr *ifp,
2285                              __u32 valid_lft, __u32 prefered_lft,
2286                              bool create, unsigned long now)
2287 {
2288         u32 flags;
2289         struct inet6_ifaddr *ift;
2290
2291         read_lock_bh(&idev->lock);
2292         /* update all temporary addresses in the list */
2293         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2294                 int age, max_valid, max_prefered;
2295
2296                 if (ifp != ift->ifpub)
2297                         continue;
2298
2299                 /* RFC 4941 section 3.3:
2300                  * If a received option will extend the lifetime of a public
2301                  * address, the lifetimes of temporary addresses should
2302                  * be extended, subject to the overall constraint that no
2303                  * temporary addresses should ever remain "valid" or "preferred"
2304                  * for a time longer than (TEMP_VALID_LIFETIME) or
2305                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2306                  */
2307                 age = (now - ift->cstamp) / HZ;
2308                 max_valid = idev->cnf.temp_valid_lft - age;
2309                 if (max_valid < 0)
2310                         max_valid = 0;
2311
2312                 max_prefered = idev->cnf.temp_prefered_lft -
2313                                idev->cnf.max_desync_factor - age;
2314                 if (max_prefered < 0)
2315                         max_prefered = 0;
2316
2317                 if (valid_lft > max_valid)
2318                         valid_lft = max_valid;
2319
2320                 if (prefered_lft > max_prefered)
2321                         prefered_lft = max_prefered;
2322
2323                 spin_lock(&ift->lock);
2324                 flags = ift->flags;
2325                 ift->valid_lft = valid_lft;
2326                 ift->prefered_lft = prefered_lft;
2327                 ift->tstamp = now;
2328                 if (prefered_lft > 0)
2329                         ift->flags &= ~IFA_F_DEPRECATED;
2330
2331                 spin_unlock(&ift->lock);
2332                 if (!(flags&IFA_F_TENTATIVE))
2333                         ipv6_ifa_notify(0, ift);
2334         }
2335
2336         if ((create || list_empty(&idev->tempaddr_list)) &&
2337             idev->cnf.use_tempaddr > 0) {
2338                 /* When a new public address is created as described
2339                  * in [ADDRCONF], also create a new temporary address.
2340                  * Also create a temporary address if it's enabled but
2341                  * no temporary address currently exists.
2342                  */
2343                 read_unlock_bh(&idev->lock);
2344                 ipv6_create_tempaddr(ifp, NULL);
2345         } else {
2346                 read_unlock_bh(&idev->lock);
2347         }
2348 }
2349
2350 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2351 {
2352         struct prefix_info *pinfo;
2353         __u32 valid_lft;
2354         __u32 prefered_lft;
2355         int addr_type;
2356         u32 addr_flags = 0;
2357         struct inet6_dev *in6_dev;
2358         struct net *net = dev_net(dev);
2359
2360         pinfo = (struct prefix_info *) opt;
2361
2362         if (len < sizeof(struct prefix_info)) {
2363                 ADBG("addrconf: prefix option too short\n");
2364                 return;
2365         }
2366
2367         /*
2368          *      Validation checks ([ADDRCONF], page 19)
2369          */
2370
2371         addr_type = ipv6_addr_type(&pinfo->prefix);
2372
2373         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2374                 return;
2375
2376         valid_lft = ntohl(pinfo->valid);
2377         prefered_lft = ntohl(pinfo->prefered);
2378
2379         if (prefered_lft > valid_lft) {
2380                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2381                 return;
2382         }
2383
2384         in6_dev = in6_dev_get(dev);
2385
2386         if (!in6_dev) {
2387                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2388                                     dev->name);
2389                 return;
2390         }
2391
2392         /*
2393          *      Two things going on here:
2394          *      1) Add routes for on-link prefixes
2395          *      2) Configure prefixes with the auto flag set
2396          */
2397
2398         if (pinfo->onlink) {
2399                 struct rt6_info *rt;
2400                 unsigned long rt_expires;
2401
2402                 /* Avoid arithmetic overflow. Really, we could
2403                  * save rt_expires in seconds, likely valid_lft,
2404                  * but it would require division in fib gc, that it
2405                  * not good.
2406                  */
2407                 if (HZ > USER_HZ)
2408                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2409                 else
2410                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2411
2412                 if (addrconf_finite_timeout(rt_expires))
2413                         rt_expires *= HZ;
2414
2415                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2416                                                pinfo->prefix_len,
2417                                                dev,
2418                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2419                                                RTF_GATEWAY | RTF_DEFAULT);
2420
2421                 if (rt) {
2422                         /* Autoconf prefix route */
2423                         if (valid_lft == 0) {
2424                                 ip6_del_rt(rt);
2425                                 rt = NULL;
2426                         } else if (addrconf_finite_timeout(rt_expires)) {
2427                                 /* not infinity */
2428                                 rt6_set_expires(rt, jiffies + rt_expires);
2429                         } else {
2430                                 rt6_clean_expires(rt);
2431                         }
2432                 } else if (valid_lft) {
2433                         clock_t expires = 0;
2434                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2435                         if (addrconf_finite_timeout(rt_expires)) {
2436                                 /* not infinity */
2437                                 flags |= RTF_EXPIRES;
2438                                 expires = jiffies_to_clock_t(rt_expires);
2439                         }
2440                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2441                                               dev, expires, flags);
2442                 }
2443                 ip6_rt_put(rt);
2444         }
2445
2446         /* Try to figure out our local address for this prefix */
2447
2448         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2449                 struct inet6_ifaddr *ifp;
2450                 struct in6_addr addr;
2451                 int create = 0, update_lft = 0;
2452                 bool tokenized = false;
2453
2454                 if (pinfo->prefix_len == 64) {
2455                         memcpy(&addr, &pinfo->prefix, 8);
2456
2457                         if (!ipv6_addr_any(&in6_dev->token)) {
2458                                 read_lock_bh(&in6_dev->lock);
2459                                 memcpy(addr.s6_addr + 8,
2460                                        in6_dev->token.s6_addr + 8, 8);
2461                                 read_unlock_bh(&in6_dev->lock);
2462                                 tokenized = true;
2463                         } else if (in6_dev->addr_gen_mode ==
2464                                    IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2465                                    !ipv6_generate_stable_address(&addr, 0,
2466                                                                  in6_dev)) {
2467                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2468                                 goto ok;
2469                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2470                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2471                                 in6_dev_put(in6_dev);
2472                                 return;
2473                         }
2474                         goto ok;
2475                 }
2476                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2477                                     pinfo->prefix_len);
2478                 in6_dev_put(in6_dev);
2479                 return;
2480
2481 ok:
2482
2483                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2484
2485                 if (!ifp && valid_lft) {
2486                         int max_addresses = in6_dev->cnf.max_addresses;
2487
2488 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2489                         if (in6_dev->cnf.optimistic_dad &&
2490                             !net->ipv6.devconf_all->forwarding && sllao)
2491                                 addr_flags |= IFA_F_OPTIMISTIC;
2492 #endif
2493
2494                         /* Do not allow to create too much of autoconfigured
2495                          * addresses; this would be too easy way to crash kernel.
2496                          */
2497                         if (!max_addresses ||
2498                             ipv6_count_addresses(in6_dev) < max_addresses)
2499                                 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2500                                                     pinfo->prefix_len,
2501                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
2502                                                     addr_flags, valid_lft,
2503                                                     prefered_lft);
2504
2505                         if (IS_ERR_OR_NULL(ifp)) {
2506                                 in6_dev_put(in6_dev);
2507                                 return;
2508                         }
2509
2510                         update_lft = 0;
2511                         create = 1;
2512                         spin_lock_bh(&ifp->lock);
2513                         ifp->flags |= IFA_F_MANAGETEMPADDR;
2514                         ifp->cstamp = jiffies;
2515                         ifp->tokenized = tokenized;
2516                         spin_unlock_bh(&ifp->lock);
2517                         addrconf_dad_start(ifp);
2518                 }
2519
2520                 if (ifp) {
2521                         u32 flags;
2522                         unsigned long now;
2523                         u32 stored_lft;
2524
2525                         /* update lifetime (RFC2462 5.5.3 e) */
2526                         spin_lock_bh(&ifp->lock);
2527                         now = jiffies;
2528                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2529                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2530                         else
2531                                 stored_lft = 0;
2532                         if (!update_lft && !create && stored_lft) {
2533                                 const u32 minimum_lft = min_t(u32,
2534                                         stored_lft, MIN_VALID_LIFETIME);
2535                                 valid_lft = max(valid_lft, minimum_lft);
2536
2537                                 /* RFC4862 Section 5.5.3e:
2538                                  * "Note that the preferred lifetime of the
2539                                  *  corresponding address is always reset to
2540                                  *  the Preferred Lifetime in the received
2541                                  *  Prefix Information option, regardless of
2542                                  *  whether the valid lifetime is also reset or
2543                                  *  ignored."
2544                                  *
2545                                  * So we should always update prefered_lft here.
2546                                  */
2547                                 update_lft = 1;
2548                         }
2549
2550                         if (update_lft) {
2551                                 ifp->valid_lft = valid_lft;
2552                                 ifp->prefered_lft = prefered_lft;
2553                                 ifp->tstamp = now;
2554                                 flags = ifp->flags;
2555                                 ifp->flags &= ~IFA_F_DEPRECATED;
2556                                 spin_unlock_bh(&ifp->lock);
2557
2558                                 if (!(flags&IFA_F_TENTATIVE))
2559                                         ipv6_ifa_notify(0, ifp);
2560                         } else
2561                                 spin_unlock_bh(&ifp->lock);
2562
2563                         manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2564                                          create, now);
2565
2566                         in6_ifa_put(ifp);
2567                         addrconf_verify();
2568                 }
2569         }
2570         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2571         in6_dev_put(in6_dev);
2572 }
2573
2574 /*
2575  *      Set destination address.
2576  *      Special case for SIT interfaces where we create a new "virtual"
2577  *      device.
2578  */
2579 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2580 {
2581         struct in6_ifreq ireq;
2582         struct net_device *dev;
2583         int err = -EINVAL;
2584
2585         rtnl_lock();
2586
2587         err = -EFAULT;
2588         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2589                 goto err_exit;
2590
2591         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2592
2593         err = -ENODEV;
2594         if (!dev)
2595                 goto err_exit;
2596
2597 #if IS_ENABLED(CONFIG_IPV6_SIT)
2598         if (dev->type == ARPHRD_SIT) {
2599                 const struct net_device_ops *ops = dev->netdev_ops;
2600                 struct ifreq ifr;
2601                 struct ip_tunnel_parm p;
2602
2603                 err = -EADDRNOTAVAIL;
2604                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2605                         goto err_exit;
2606
2607                 memset(&p, 0, sizeof(p));
2608                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2609                 p.iph.saddr = 0;
2610                 p.iph.version = 4;
2611                 p.iph.ihl = 5;
2612                 p.iph.protocol = IPPROTO_IPV6;
2613                 p.iph.ttl = 64;
2614                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2615
2616                 if (ops->ndo_do_ioctl) {
2617                         mm_segment_t oldfs = get_fs();
2618
2619                         set_fs(KERNEL_DS);
2620                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2621                         set_fs(oldfs);
2622                 } else
2623                         err = -EOPNOTSUPP;
2624
2625                 if (err == 0) {
2626                         err = -ENOBUFS;
2627                         dev = __dev_get_by_name(net, p.name);
2628                         if (!dev)
2629                                 goto err_exit;
2630                         err = dev_open(dev);
2631                 }
2632         }
2633 #endif
2634
2635 err_exit:
2636         rtnl_unlock();
2637         return err;
2638 }
2639
2640 static int ipv6_mc_config(struct sock *sk, bool join,
2641                           const struct in6_addr *addr, int ifindex)
2642 {
2643         int ret;
2644
2645         ASSERT_RTNL();
2646
2647         lock_sock(sk);
2648         if (join)
2649                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2650         else
2651                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2652         release_sock(sk);
2653
2654         return ret;
2655 }
2656
2657 /*
2658  *      Manual configuration of address on an interface
2659  */
2660 static int inet6_addr_add(struct net *net, int ifindex,
2661                           const struct in6_addr *pfx,
2662                           const struct in6_addr *peer_pfx,
2663                           unsigned int plen, __u32 ifa_flags,
2664                           __u32 prefered_lft, __u32 valid_lft)
2665 {
2666         struct inet6_ifaddr *ifp;
2667         struct inet6_dev *idev;
2668         struct net_device *dev;
2669         unsigned long timeout;
2670         clock_t expires;
2671         int scope;
2672         u32 flags;
2673
2674         ASSERT_RTNL();
2675
2676         if (plen > 128)
2677                 return -EINVAL;
2678
2679         /* check the lifetime */
2680         if (!valid_lft || prefered_lft > valid_lft)
2681                 return -EINVAL;
2682
2683         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2684                 return -EINVAL;
2685
2686         dev = __dev_get_by_index(net, ifindex);
2687         if (!dev)
2688                 return -ENODEV;
2689
2690         idev = addrconf_add_dev(dev);
2691         if (IS_ERR(idev))
2692                 return PTR_ERR(idev);
2693
2694         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2695                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2696                                          true, pfx, ifindex);
2697
2698                 if (ret < 0)
2699                         return ret;
2700         }
2701
2702         scope = ipv6_addr_scope(pfx);
2703
2704         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2705         if (addrconf_finite_timeout(timeout)) {
2706                 expires = jiffies_to_clock_t(timeout * HZ);
2707                 valid_lft = timeout;
2708                 flags = RTF_EXPIRES;
2709         } else {
2710                 expires = 0;
2711                 flags = 0;
2712                 ifa_flags |= IFA_F_PERMANENT;
2713         }
2714
2715         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2716         if (addrconf_finite_timeout(timeout)) {
2717                 if (timeout == 0)
2718                         ifa_flags |= IFA_F_DEPRECATED;
2719                 prefered_lft = timeout;
2720         }
2721
2722         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2723                             valid_lft, prefered_lft);
2724
2725         if (!IS_ERR(ifp)) {
2726                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2727                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2728                                               expires, flags);
2729                 }
2730
2731                 /*
2732                  * Note that section 3.1 of RFC 4429 indicates
2733                  * that the Optimistic flag should not be set for
2734                  * manually configured addresses
2735                  */
2736                 addrconf_dad_start(ifp);
2737                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2738                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2739                                          true, jiffies);
2740                 in6_ifa_put(ifp);
2741                 addrconf_verify_rtnl();
2742                 return 0;
2743         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2744                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2745                                false, pfx, ifindex);
2746         }
2747
2748         return PTR_ERR(ifp);
2749 }
2750
2751 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2752                           const struct in6_addr *pfx, unsigned int plen)
2753 {
2754         struct inet6_ifaddr *ifp;
2755         struct inet6_dev *idev;
2756         struct net_device *dev;
2757
2758         if (plen > 128)
2759                 return -EINVAL;
2760
2761         dev = __dev_get_by_index(net, ifindex);
2762         if (!dev)
2763                 return -ENODEV;
2764
2765         idev = __in6_dev_get(dev);
2766         if (!idev)
2767                 return -ENXIO;
2768
2769         read_lock_bh(&idev->lock);
2770         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2771                 if (ifp->prefix_len == plen &&
2772                     ipv6_addr_equal(pfx, &ifp->addr)) {
2773                         in6_ifa_hold(ifp);
2774                         read_unlock_bh(&idev->lock);
2775
2776                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2777                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2778                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2779                                                  jiffies);
2780                         ipv6_del_addr(ifp);
2781                         addrconf_verify_rtnl();
2782                         if (ipv6_addr_is_multicast(pfx)) {
2783                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2784                                                false, pfx, dev->ifindex);
2785                         }
2786                         return 0;
2787                 }
2788         }
2789         read_unlock_bh(&idev->lock);
2790         return -EADDRNOTAVAIL;
2791 }
2792
2793
2794 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2795 {
2796         struct in6_ifreq ireq;
2797         int err;
2798
2799         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2800                 return -EPERM;
2801
2802         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2803                 return -EFAULT;
2804
2805         rtnl_lock();
2806         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2807                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2808                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2809         rtnl_unlock();
2810         return err;
2811 }
2812
2813 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2814 {
2815         struct in6_ifreq ireq;
2816         int err;
2817
2818         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2819                 return -EPERM;
2820
2821         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2822                 return -EFAULT;
2823
2824         rtnl_lock();
2825         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2826                              ireq.ifr6_prefixlen);
2827         rtnl_unlock();
2828         return err;
2829 }
2830
2831 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2832                      int plen, int scope)
2833 {
2834         struct inet6_ifaddr *ifp;
2835
2836         ifp = ipv6_add_addr(idev, addr, NULL, plen,
2837                             scope, IFA_F_PERMANENT,
2838                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2839         if (!IS_ERR(ifp)) {
2840                 spin_lock_bh(&ifp->lock);
2841                 ifp->flags &= ~IFA_F_TENTATIVE;
2842                 spin_unlock_bh(&ifp->lock);
2843                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2844                 in6_ifa_put(ifp);
2845         }
2846 }
2847
2848 #if IS_ENABLED(CONFIG_IPV6_SIT)
2849 static void sit_add_v4_addrs(struct inet6_dev *idev)
2850 {
2851         struct in6_addr addr;
2852         struct net_device *dev;
2853         struct net *net = dev_net(idev->dev);
2854         int scope, plen;
2855         u32 pflags = 0;
2856
2857         ASSERT_RTNL();
2858
2859         memset(&addr, 0, sizeof(struct in6_addr));
2860         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2861
2862         if (idev->dev->flags&IFF_POINTOPOINT) {
2863                 addr.s6_addr32[0] = htonl(0xfe800000);
2864                 scope = IFA_LINK;
2865                 plen = 64;
2866         } else {
2867                 scope = IPV6_ADDR_COMPATv4;
2868                 plen = 96;
2869                 pflags |= RTF_NONEXTHOP;
2870         }
2871
2872         if (addr.s6_addr32[3]) {
2873                 add_addr(idev, &addr, plen, scope);
2874                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2875                 return;
2876         }
2877
2878         for_each_netdev(net, dev) {
2879                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2880                 if (in_dev && (dev->flags & IFF_UP)) {
2881                         struct in_ifaddr *ifa;
2882
2883                         int flag = scope;
2884
2885                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2886
2887                                 addr.s6_addr32[3] = ifa->ifa_local;
2888
2889                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2890                                         continue;
2891                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2892                                         if (idev->dev->flags&IFF_POINTOPOINT)
2893                                                 continue;
2894                                         flag |= IFA_HOST;
2895                                 }
2896
2897                                 add_addr(idev, &addr, plen, flag);
2898                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2899                                                       pflags);
2900                         }
2901                 }
2902         }
2903 }
2904 #endif
2905
2906 static void init_loopback(struct net_device *dev)
2907 {
2908         struct inet6_dev  *idev;
2909         struct net_device *sp_dev;
2910         struct inet6_ifaddr *sp_ifa;
2911         struct rt6_info *sp_rt;
2912
2913         /* ::1 */
2914
2915         ASSERT_RTNL();
2916
2917         idev = ipv6_find_idev(dev);
2918         if (!idev) {
2919                 pr_debug("%s: add_dev failed\n", __func__);
2920                 return;
2921         }
2922
2923         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2924
2925         /* Add routes to other interface's IPv6 addresses */
2926         for_each_netdev(dev_net(dev), sp_dev) {
2927                 if (!strcmp(sp_dev->name, dev->name))
2928                         continue;
2929
2930                 idev = __in6_dev_get(sp_dev);
2931                 if (!idev)
2932                         continue;
2933
2934                 read_lock_bh(&idev->lock);
2935                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2936
2937                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2938                                 continue;
2939
2940                         if (sp_ifa->rt) {
2941                                 /* This dst has been added to garbage list when
2942                                  * lo device down, release this obsolete dst and
2943                                  * reallocate a new router for ifa.
2944                                  */
2945                                 if (sp_ifa->rt->dst.obsolete > 0) {
2946                                         ip6_rt_put(sp_ifa->rt);
2947                                         sp_ifa->rt = NULL;
2948                                 } else {
2949                                         continue;
2950                                 }
2951                         }
2952
2953                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2954
2955                         /* Failure cases are ignored */
2956                         if (!IS_ERR(sp_rt)) {
2957                                 sp_ifa->rt = sp_rt;
2958                                 ip6_ins_rt(sp_rt);
2959                         }
2960                 }
2961                 read_unlock_bh(&idev->lock);
2962         }
2963 }
2964
2965 static void addrconf_add_linklocal(struct inet6_dev *idev,
2966                                    const struct in6_addr *addr, u32 flags)
2967 {
2968         struct inet6_ifaddr *ifp;
2969         u32 addr_flags = flags | IFA_F_PERMANENT;
2970
2971 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2972         if (idev->cnf.optimistic_dad &&
2973             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2974                 addr_flags |= IFA_F_OPTIMISTIC;
2975 #endif
2976
2977         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2978                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2979         if (!IS_ERR(ifp)) {
2980                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2981                 addrconf_dad_start(ifp);
2982                 in6_ifa_put(ifp);
2983         }
2984 }
2985
2986 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2987 {
2988         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2989                 return true;
2990
2991         if (address.s6_addr32[2] == htonl(0x02005eff) &&
2992             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2993                 return true;
2994
2995         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2996             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2997                 return true;
2998
2999         return false;
3000 }
3001
3002 static int ipv6_generate_stable_address(struct in6_addr *address,
3003                                         u8 dad_count,
3004                                         const struct inet6_dev *idev)
3005 {
3006         static DEFINE_SPINLOCK(lock);
3007         static __u32 digest[SHA_DIGEST_WORDS];
3008         static __u32 workspace[SHA_WORKSPACE_WORDS];
3009
3010         static union {
3011                 char __data[SHA_MESSAGE_BYTES];
3012                 struct {
3013                         struct in6_addr secret;
3014                         __be32 prefix[2];
3015                         unsigned char hwaddr[MAX_ADDR_LEN];
3016                         u8 dad_count;
3017                 } __packed;
3018         } data;
3019
3020         struct in6_addr secret;
3021         struct in6_addr temp;
3022         struct net *net = dev_net(idev->dev);
3023
3024         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3025
3026         if (idev->cnf.stable_secret.initialized)
3027                 secret = idev->cnf.stable_secret.secret;
3028         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3029                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3030         else
3031                 return -1;
3032
3033 retry:
3034         spin_lock_bh(&lock);
3035
3036         sha_init(digest);
3037         memset(&data, 0, sizeof(data));
3038         memset(workspace, 0, sizeof(workspace));
3039         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3040         data.prefix[0] = address->s6_addr32[0];
3041         data.prefix[1] = address->s6_addr32[1];
3042         data.secret = secret;
3043         data.dad_count = dad_count;
3044
3045         sha_transform(digest, data.__data, workspace);
3046
3047         temp = *address;
3048         temp.s6_addr32[2] = (__force __be32)digest[0];
3049         temp.s6_addr32[3] = (__force __be32)digest[1];
3050
3051         spin_unlock_bh(&lock);
3052
3053         if (ipv6_reserved_interfaceid(temp)) {
3054                 dad_count++;
3055                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3056                         return -1;
3057                 goto retry;
3058         }
3059
3060         *address = temp;
3061         return 0;
3062 }
3063
3064 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3065 {
3066         struct in6_addr addr;
3067
3068         /* no link local addresses on L3 master devices */
3069         if (netif_is_l3_master(idev->dev))
3070                 return;
3071
3072         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3073
3074         if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3075                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3076                         addrconf_add_linklocal(idev, &addr,
3077                                                IFA_F_STABLE_PRIVACY);
3078                 else if (prefix_route)
3079                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3080         } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3081                 /* addrconf_add_linklocal also adds a prefix_route and we
3082                  * only need to care about prefix routes if ipv6_generate_eui64
3083                  * couldn't generate one.
3084                  */
3085                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3086                         addrconf_add_linklocal(idev, &addr, 0);
3087                 else if (prefix_route)
3088                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3089         }
3090 }
3091
3092 static void addrconf_dev_config(struct net_device *dev)
3093 {
3094         struct inet6_dev *idev;
3095
3096         ASSERT_RTNL();
3097
3098         if ((dev->type != ARPHRD_ETHER) &&
3099             (dev->type != ARPHRD_FDDI) &&
3100             (dev->type != ARPHRD_ARCNET) &&
3101             (dev->type != ARPHRD_INFINIBAND) &&
3102             (dev->type != ARPHRD_IEEE802154) &&
3103             (dev->type != ARPHRD_IEEE1394) &&
3104             (dev->type != ARPHRD_TUNNEL6) &&
3105             (dev->type != ARPHRD_6LOWPAN)) {
3106                 /* Alas, we support only Ethernet autoconfiguration. */
3107                 return;
3108         }
3109
3110         idev = addrconf_add_dev(dev);
3111         if (IS_ERR(idev))
3112                 return;
3113
3114         addrconf_addr_gen(idev, false);
3115 }
3116
3117 #if IS_ENABLED(CONFIG_IPV6_SIT)
3118 static void addrconf_sit_config(struct net_device *dev)
3119 {
3120         struct inet6_dev *idev;
3121
3122         ASSERT_RTNL();
3123
3124         /*
3125          * Configure the tunnel with one of our IPv4
3126          * addresses... we should configure all of
3127          * our v4 addrs in the tunnel
3128          */
3129
3130         idev = ipv6_find_idev(dev);
3131         if (!idev) {
3132                 pr_debug("%s: add_dev failed\n", __func__);
3133                 return;
3134         }
3135
3136         if (dev->priv_flags & IFF_ISATAP) {
3137                 addrconf_addr_gen(idev, false);
3138                 return;
3139         }
3140
3141         sit_add_v4_addrs(idev);
3142
3143         if (dev->flags&IFF_POINTOPOINT)
3144                 addrconf_add_mroute(dev);
3145 }
3146 #endif
3147
3148 #if IS_ENABLED(CONFIG_NET_IPGRE)
3149 static void addrconf_gre_config(struct net_device *dev)
3150 {
3151         struct inet6_dev *idev;
3152
3153         ASSERT_RTNL();
3154
3155         idev = ipv6_find_idev(dev);
3156         if (!idev) {
3157                 pr_debug("%s: add_dev failed\n", __func__);
3158                 return;
3159         }
3160
3161         addrconf_addr_gen(idev, true);
3162         if (dev->flags & IFF_POINTOPOINT)
3163                 addrconf_add_mroute(dev);
3164 }
3165 #endif
3166
3167 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3168                            void *ptr)
3169 {
3170         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3171         struct inet6_dev *idev = __in6_dev_get(dev);
3172         int run_pending = 0;
3173         int err;
3174
3175         switch (event) {
3176         case NETDEV_REGISTER:
3177                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3178                         idev = ipv6_add_dev(dev);
3179                         if (IS_ERR(idev))
3180                                 return notifier_from_errno(PTR_ERR(idev));
3181                 }
3182                 break;
3183
3184         case NETDEV_CHANGEMTU:
3185                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3186                 if (dev->mtu < IPV6_MIN_MTU) {
3187                         addrconf_ifdown(dev, 1);
3188                         break;
3189                 }
3190
3191                 if (idev) {
3192                         rt6_mtu_change(dev, dev->mtu);
3193                         idev->cnf.mtu6 = dev->mtu;
3194                         break;
3195                 }
3196
3197                 /* allocate new idev */
3198                 idev = ipv6_add_dev(dev);
3199                 if (IS_ERR(idev))
3200                         break;
3201
3202                 /* device is still not ready */
3203                 if (!(idev->if_flags & IF_READY))
3204                         break;
3205
3206                 run_pending = 1;
3207
3208                 /* fall through */
3209
3210         case NETDEV_UP:
3211         case NETDEV_CHANGE:
3212                 if (dev->flags & IFF_SLAVE)
3213                         break;
3214
3215                 if (idev && idev->cnf.disable_ipv6)
3216                         break;
3217
3218                 if (event == NETDEV_UP) {
3219                         if (!addrconf_qdisc_ok(dev)) {
3220                                 /* device is not ready yet. */
3221                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3222                                         dev->name);
3223                                 break;
3224                         }
3225
3226                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3227                                 idev = ipv6_add_dev(dev);
3228
3229                         if (!IS_ERR_OR_NULL(idev)) {
3230                                 idev->if_flags |= IF_READY;
3231                                 run_pending = 1;
3232                         }
3233                 } else if (event == NETDEV_CHANGE) {
3234                         if (!addrconf_qdisc_ok(dev)) {
3235                                 /* device is still not ready. */
3236                                 break;
3237                         }
3238
3239                         if (idev) {
3240                                 if (idev->if_flags & IF_READY)
3241                                         /* device is already configured. */
3242                                         break;
3243                                 idev->if_flags |= IF_READY;
3244                         }
3245
3246                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3247                                 dev->name);
3248
3249                         run_pending = 1;
3250                 }
3251
3252                 switch (dev->type) {
3253 #if IS_ENABLED(CONFIG_IPV6_SIT)
3254                 case ARPHRD_SIT:
3255                         addrconf_sit_config(dev);
3256                         break;
3257 #endif
3258 #if IS_ENABLED(CONFIG_NET_IPGRE)
3259                 case ARPHRD_IPGRE:
3260                         addrconf_gre_config(dev);
3261                         break;
3262 #endif
3263                 case ARPHRD_LOOPBACK:
3264                         init_loopback(dev);
3265                         break;
3266
3267                 default:
3268                         addrconf_dev_config(dev);
3269                         break;
3270                 }
3271
3272                 if (!IS_ERR_OR_NULL(idev)) {
3273                         if (run_pending)
3274                                 addrconf_dad_run(idev);
3275
3276                         /*
3277                          * If the MTU changed during the interface down,
3278                          * when the interface up, the changed MTU must be
3279                          * reflected in the idev as well as routers.
3280                          */
3281                         if (idev->cnf.mtu6 != dev->mtu &&
3282                             dev->mtu >= IPV6_MIN_MTU) {
3283                                 rt6_mtu_change(dev, dev->mtu);
3284                                 idev->cnf.mtu6 = dev->mtu;
3285                         }
3286                         idev->tstamp = jiffies;
3287                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3288
3289                         /*
3290                          * If the changed mtu during down is lower than
3291                          * IPV6_MIN_MTU stop IPv6 on this interface.
3292                          */
3293                         if (dev->mtu < IPV6_MIN_MTU)
3294                                 addrconf_ifdown(dev, 1);
3295                 }
3296                 break;
3297
3298         case NETDEV_DOWN:
3299         case NETDEV_UNREGISTER:
3300                 /*
3301                  *      Remove all addresses from this interface.
3302                  */
3303                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3304                 break;
3305
3306         case NETDEV_CHANGENAME:
3307                 if (idev) {
3308                         snmp6_unregister_dev(idev);
3309                         addrconf_sysctl_unregister(idev);
3310                         err = addrconf_sysctl_register(idev);
3311                         if (err)
3312                                 return notifier_from_errno(err);
3313                         err = snmp6_register_dev(idev);
3314                         if (err) {
3315                                 addrconf_sysctl_unregister(idev);
3316                                 return notifier_from_errno(err);
3317                         }
3318                 }
3319                 break;
3320
3321         case NETDEV_PRE_TYPE_CHANGE:
3322         case NETDEV_POST_TYPE_CHANGE:
3323                 addrconf_type_change(dev, event);
3324                 break;
3325         }
3326
3327         return NOTIFY_OK;
3328 }
3329
3330 /*
3331  *      addrconf module should be notified of a device going up
3332  */
3333 static struct notifier_block ipv6_dev_notf = {
3334         .notifier_call = addrconf_notify,
3335 };
3336
3337 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3338 {
3339         struct inet6_dev *idev;
3340         ASSERT_RTNL();
3341
3342         idev = __in6_dev_get(dev);
3343
3344         if (event == NETDEV_POST_TYPE_CHANGE)
3345                 ipv6_mc_remap(idev);
3346         else if (event == NETDEV_PRE_TYPE_CHANGE)
3347                 ipv6_mc_unmap(idev);
3348 }
3349
3350 static int addrconf_ifdown(struct net_device *dev, int how)
3351 {
3352         struct net *net = dev_net(dev);
3353         struct inet6_dev *idev;
3354         struct inet6_ifaddr *ifa;
3355         int state, i;
3356
3357         ASSERT_RTNL();
3358
3359         rt6_ifdown(net, dev);
3360         neigh_ifdown(&nd_tbl, dev);
3361
3362         idev = __in6_dev_get(dev);
3363         if (!idev)
3364                 return -ENODEV;
3365
3366         /*
3367          * Step 1: remove reference to ipv6 device from parent device.
3368          *         Do not dev_put!
3369          */
3370         if (how) {
3371                 idev->dead = 1;
3372
3373                 /* protected by rtnl_lock */
3374                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3375
3376                 /* Step 1.5: remove snmp6 entry */
3377                 snmp6_unregister_dev(idev);
3378
3379         }
3380
3381         /* Step 2: clear hash table */
3382         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3383                 struct hlist_head *h = &inet6_addr_lst[i];
3384
3385                 spin_lock_bh(&addrconf_hash_lock);
3386 restart:
3387                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3388                         if (ifa->idev == idev) {
3389                                 hlist_del_init_rcu(&ifa->addr_lst);
3390                                 addrconf_del_dad_work(ifa);
3391                                 goto restart;
3392                         }
3393                 }
3394                 spin_unlock_bh(&addrconf_hash_lock);
3395         }
3396
3397         write_lock_bh(&idev->lock);
3398
3399         addrconf_del_rs_timer(idev);
3400
3401         /* Step 2: clear flags for stateless addrconf */
3402         if (!how)
3403                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3404
3405         if (how && del_timer(&idev->regen_timer))
3406                 in6_dev_put(idev);
3407
3408         /* Step 3: clear tempaddr list */
3409         while (!list_empty(&idev->tempaddr_list)) {
3410                 ifa = list_first_entry(&idev->tempaddr_list,
3411                                        struct inet6_ifaddr, tmp_list);
3412                 list_del(&ifa->tmp_list);
3413                 write_unlock_bh(&idev->lock);
3414                 spin_lock_bh(&ifa->lock);
3415
3416                 if (ifa->ifpub) {
3417                         in6_ifa_put(ifa->ifpub);
3418                         ifa->ifpub = NULL;
3419                 }
3420                 spin_unlock_bh(&ifa->lock);
3421                 in6_ifa_put(ifa);
3422                 write_lock_bh(&idev->lock);
3423         }
3424
3425         while (!list_empty(&idev->addr_list)) {
3426                 ifa = list_first_entry(&idev->addr_list,
3427                                        struct inet6_ifaddr, if_list);
3428                 addrconf_del_dad_work(ifa);
3429
3430                 list_del(&ifa->if_list);
3431
3432                 write_unlock_bh(&idev->lock);
3433
3434                 spin_lock_bh(&ifa->lock);
3435                 state = ifa->state;
3436                 ifa->state = INET6_IFADDR_STATE_DEAD;
3437                 spin_unlock_bh(&ifa->lock);
3438
3439                 if (state != INET6_IFADDR_STATE_DEAD) {
3440                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3441                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3442                 }
3443                 in6_ifa_put(ifa);
3444
3445                 write_lock_bh(&idev->lock);
3446         }
3447
3448         write_unlock_bh(&idev->lock);
3449
3450         /* Step 5: Discard anycast and multicast list */
3451         if (how) {
3452                 ipv6_ac_destroy_dev(idev);
3453                 ipv6_mc_destroy_dev(idev);
3454         } else {
3455                 ipv6_mc_down(idev);
3456         }
3457
3458         idev->tstamp = jiffies;
3459
3460         /* Last: Shot the device (if unregistered) */
3461         if (how) {
3462                 addrconf_sysctl_unregister(idev);
3463                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3464                 neigh_ifdown(&nd_tbl, dev);
3465                 in6_dev_put(idev);
3466         }
3467         return 0;
3468 }
3469
3470 static void addrconf_rs_timer(unsigned long data)
3471 {
3472         struct inet6_dev *idev = (struct inet6_dev *)data;
3473         struct net_device *dev = idev->dev;
3474         struct in6_addr lladdr;
3475
3476         write_lock(&idev->lock);
3477         if (idev->dead || !(idev->if_flags & IF_READY))
3478                 goto out;
3479
3480         if (!ipv6_accept_ra(idev))
3481                 goto out;
3482
3483         /* Announcement received after solicitation was sent */
3484         if (idev->if_flags & IF_RA_RCVD)
3485                 goto out;
3486
3487         if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3488                 write_unlock(&idev->lock);
3489                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3490                         ndisc_send_rs(dev, &lladdr,
3491                                       &in6addr_linklocal_allrouters);
3492                 else
3493                         goto put;
3494
3495                 write_lock(&idev->lock);
3496                 /* The wait after the last probe can be shorter */
3497                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3498                                              idev->cnf.rtr_solicits) ?
3499                                       idev->cnf.rtr_solicit_delay :
3500                                       idev->cnf.rtr_solicit_interval);
3501         } else {
3502                 /*
3503                  * Note: we do not support deprecated "all on-link"
3504                  * assumption any longer.
3505                  */
3506                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3507         }
3508
3509 out:
3510         write_unlock(&idev->lock);
3511 put:
3512         in6_dev_put(idev);
3513 }
3514
3515 /*
3516  *      Duplicate Address Detection
3517  */
3518 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3519 {
3520         unsigned long rand_num;
3521         struct inet6_dev *idev = ifp->idev;
3522
3523         if (ifp->flags & IFA_F_OPTIMISTIC)
3524                 rand_num = 0;
3525         else
3526                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3527
3528         ifp->dad_probes = idev->cnf.dad_transmits;
3529         addrconf_mod_dad_work(ifp, rand_num);
3530 }
3531
3532 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3533 {
3534         struct inet6_dev *idev = ifp->idev;
3535         struct net_device *dev = idev->dev;
3536         bool notify = false;
3537
3538         addrconf_join_solict(dev, &ifp->addr);
3539
3540         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3541
3542         read_lock_bh(&idev->lock);
3543         spin_lock(&ifp->lock);
3544         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3545                 goto out;
3546
3547         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3548             idev->cnf.accept_dad < 1 ||
3549             !(ifp->flags&IFA_F_TENTATIVE) ||
3550             ifp->flags & IFA_F_NODAD) {
3551                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3552                 spin_unlock(&ifp->lock);
3553                 read_unlock_bh(&idev->lock);
3554
3555                 addrconf_dad_completed(ifp);
3556                 return;
3557         }
3558
3559         if (!(idev->if_flags & IF_READY)) {
3560                 spin_unlock(&ifp->lock);
3561                 read_unlock_bh(&idev->lock);
3562                 /*
3563                  * If the device is not ready:
3564                  * - keep it tentative if it is a permanent address.
3565                  * - otherwise, kill it.
3566                  */
3567                 in6_ifa_hold(ifp);
3568                 addrconf_dad_stop(ifp, 0);
3569                 return;
3570         }
3571
3572         /*
3573          * Optimistic nodes can start receiving
3574          * Frames right away
3575          */
3576         if (ifp->flags & IFA_F_OPTIMISTIC) {
3577                 ip6_ins_rt(ifp->rt);
3578                 if (ipv6_use_optimistic_addr(idev)) {
3579                         /* Because optimistic nodes can use this address,
3580                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3581                          */
3582                         notify = true;
3583                 }
3584         }
3585
3586         addrconf_dad_kick(ifp);
3587 out:
3588         spin_unlock(&ifp->lock);
3589         read_unlock_bh(&idev->lock);
3590         if (notify)
3591                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3592 }
3593
3594 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3595 {
3596         bool begin_dad = false;
3597
3598         spin_lock_bh(&ifp->lock);
3599         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3600                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3601                 begin_dad = true;
3602         }
3603         spin_unlock_bh(&ifp->lock);
3604
3605         if (begin_dad)
3606                 addrconf_mod_dad_work(ifp, 0);
3607 }
3608
3609 static void addrconf_dad_work(struct work_struct *w)
3610 {
3611         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3612                                                 struct inet6_ifaddr,
3613                                                 dad_work);
3614         struct inet6_dev *idev = ifp->idev;
3615         struct in6_addr mcaddr;
3616
3617         enum {
3618                 DAD_PROCESS,
3619                 DAD_BEGIN,
3620                 DAD_ABORT,
3621         } action = DAD_PROCESS;
3622
3623         rtnl_lock();
3624
3625         spin_lock_bh(&ifp->lock);
3626         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3627                 action = DAD_BEGIN;
3628                 ifp->state = INET6_IFADDR_STATE_DAD;
3629         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3630                 action = DAD_ABORT;
3631                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3632         }
3633         spin_unlock_bh(&ifp->lock);
3634
3635         if (action == DAD_BEGIN) {
3636                 addrconf_dad_begin(ifp);
3637                 goto out;
3638         } else if (action == DAD_ABORT) {
3639                 addrconf_dad_stop(ifp, 1);
3640                 goto out;
3641         }
3642
3643         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3644                 goto out;
3645
3646         write_lock_bh(&idev->lock);
3647         if (idev->dead || !(idev->if_flags & IF_READY)) {
3648                 write_unlock_bh(&idev->lock);
3649                 goto out;
3650         }
3651
3652         spin_lock(&ifp->lock);
3653         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3654                 spin_unlock(&ifp->lock);
3655                 write_unlock_bh(&idev->lock);
3656                 goto out;
3657         }
3658
3659         if (ifp->dad_probes == 0) {
3660                 /*
3661                  * DAD was successful
3662                  */
3663
3664                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3665                 spin_unlock(&ifp->lock);
3666                 write_unlock_bh(&idev->lock);
3667
3668                 addrconf_dad_completed(ifp);
3669
3670                 goto out;
3671         }
3672
3673         ifp->dad_probes--;
3674         addrconf_mod_dad_work(ifp,
3675                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3676         spin_unlock(&ifp->lock);
3677         write_unlock_bh(&idev->lock);
3678
3679         /* send a neighbour solicitation for our addr */
3680         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3681         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3682 out:
3683         in6_ifa_put(ifp);
3684         rtnl_unlock();
3685 }
3686
3687 /* ifp->idev must be at least read locked */
3688 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3689 {
3690         struct inet6_ifaddr *ifpiter;
3691         struct inet6_dev *idev = ifp->idev;
3692
3693         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3694                 if (ifpiter->scope > IFA_LINK)
3695                         break;
3696                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3697                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3698                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3699                     IFA_F_PERMANENT)
3700                         return false;
3701         }
3702         return true;
3703 }
3704
3705 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3706 {
3707         struct net_device *dev = ifp->idev->dev;
3708         struct in6_addr lladdr;
3709         bool send_rs, send_mld;
3710
3711         addrconf_del_dad_work(ifp);
3712
3713         /*
3714          *      Configure the address for reception. Now it is valid.
3715          */
3716
3717         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3718
3719         /* If added prefix is link local and we are prepared to process
3720            router advertisements, start sending router solicitations.
3721          */
3722
3723         read_lock_bh(&ifp->idev->lock);
3724         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3725         send_rs = send_mld &&
3726                   ipv6_accept_ra(ifp->idev) &&
3727                   ifp->idev->cnf.rtr_solicits > 0 &&
3728                   (dev->flags&IFF_LOOPBACK) == 0;
3729         read_unlock_bh(&ifp->idev->lock);
3730
3731         /* While dad is in progress mld report's source address is in6_addrany.
3732          * Resend with proper ll now.
3733          */
3734         if (send_mld)
3735                 ipv6_mc_dad_complete(ifp->idev);
3736
3737         if (send_rs) {
3738                 /*
3739                  *      If a host as already performed a random delay
3740                  *      [...] as part of DAD [...] there is no need
3741                  *      to delay again before sending the first RS
3742                  */
3743                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3744                         return;
3745                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3746
3747                 write_lock_bh(&ifp->idev->lock);
3748                 spin_lock(&ifp->lock);
3749                 ifp->idev->rs_probes = 1;
3750                 ifp->idev->if_flags |= IF_RS_SENT;
3751                 addrconf_mod_rs_timer(ifp->idev,
3752                                       ifp->idev->cnf.rtr_solicit_interval);
3753                 spin_unlock(&ifp->lock);
3754                 write_unlock_bh(&ifp->idev->lock);
3755         }
3756 }
3757
3758 static void addrconf_dad_run(struct inet6_dev *idev)
3759 {
3760         struct inet6_ifaddr *ifp;
3761
3762         read_lock_bh(&idev->lock);
3763         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3764                 spin_lock(&ifp->lock);
3765                 if (ifp->flags & IFA_F_TENTATIVE &&
3766                     ifp->state == INET6_IFADDR_STATE_DAD)
3767                         addrconf_dad_kick(ifp);
3768                 spin_unlock(&ifp->lock);
3769         }
3770         read_unlock_bh(&idev->lock);
3771 }
3772
3773 #ifdef CONFIG_PROC_FS
3774 struct if6_iter_state {
3775         struct seq_net_private p;
3776         int bucket;
3777         int offset;
3778 };
3779
3780 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3781 {
3782         struct inet6_ifaddr *ifa = NULL;
3783         struct if6_iter_state *state = seq->private;
3784         struct net *net = seq_file_net(seq);
3785         int p = 0;
3786
3787         /* initial bucket if pos is 0 */
3788         if (pos == 0) {
3789                 state->bucket = 0;
3790                 state->offset = 0;
3791         }
3792
3793         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3794                 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3795                                          addr_lst) {
3796                         if (!net_eq(dev_net(ifa->idev->dev), net))
3797                                 continue;
3798                         /* sync with offset */
3799                         if (p < state->offset) {
3800                                 p++;
3801                                 continue;
3802                         }
3803                         state->offset++;
3804                         return ifa;
3805                 }
3806
3807                 /* prepare for next bucket */
3808                 state->offset = 0;
3809                 p = 0;
3810         }
3811         return NULL;
3812 }
3813
3814 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3815                                          struct inet6_ifaddr *ifa)
3816 {
3817         struct if6_iter_state *state = seq->private;
3818         struct net *net = seq_file_net(seq);
3819
3820         hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3821                 if (!net_eq(dev_net(ifa->idev->dev), net))
3822                         continue;
3823                 state->offset++;
3824                 return ifa;
3825         }
3826
3827         while (++state->bucket < IN6_ADDR_HSIZE) {
3828                 state->offset = 0;
3829                 hlist_for_each_entry_rcu_bh(ifa,
3830                                      &inet6_addr_lst[state->bucket], addr_lst) {
3831                         if (!net_eq(dev_net(ifa->idev->dev), net))
3832                                 continue;
3833                         state->offset++;
3834                         return ifa;
3835                 }
3836         }
3837
3838         return NULL;
3839 }
3840
3841 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3842         __acquires(rcu_bh)
3843 {
3844         rcu_read_lock_bh();
3845         return if6_get_first(seq, *pos);
3846 }
3847
3848 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3849 {
3850         struct inet6_ifaddr *ifa;
3851
3852         ifa = if6_get_next(seq, v);
3853         ++*pos;
3854         return ifa;
3855 }
3856
3857 static void if6_seq_stop(struct seq_file *seq, void *v)
3858         __releases(rcu_bh)
3859 {
3860         rcu_read_unlock_bh();
3861 }
3862
3863 static int if6_seq_show(struct seq_file *seq, void *v)
3864 {
3865         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3866         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3867                    &ifp->addr,
3868                    ifp->idev->dev->ifindex,
3869                    ifp->prefix_len,
3870                    ifp->scope,
3871                    (u8) ifp->flags,
3872                    ifp->idev->dev->name);
3873         return 0;
3874 }
3875
3876 static const struct seq_operations if6_seq_ops = {
3877         .start  = if6_seq_start,
3878         .next   = if6_seq_next,
3879         .show   = if6_seq_show,
3880         .stop   = if6_seq_stop,
3881 };
3882
3883 static int if6_seq_open(struct inode *inode, struct file *file)
3884 {
3885         return seq_open_net(inode, file, &if6_seq_ops,
3886                             sizeof(struct if6_iter_state));
3887 }
3888
3889 static const struct file_operations if6_fops = {
3890         .owner          = THIS_MODULE,
3891         .open           = if6_seq_open,
3892         .read           = seq_read,
3893         .llseek         = seq_lseek,
3894         .release        = seq_release_net,
3895 };
3896
3897 static int __net_init if6_proc_net_init(struct net *net)
3898 {
3899         if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3900                 return -ENOMEM;
3901         return 0;
3902 }
3903
3904 static void __net_exit if6_proc_net_exit(struct net *net)
3905 {
3906         remove_proc_entry("if_inet6", net->proc_net);
3907 }
3908
3909 static struct pernet_operations if6_proc_net_ops = {
3910         .init = if6_proc_net_init,
3911         .exit = if6_proc_net_exit,
3912 };
3913
3914 int __init if6_proc_init(void)
3915 {
3916         return register_pernet_subsys(&if6_proc_net_ops);
3917 }
3918
3919 void if6_proc_exit(void)
3920 {
3921         unregister_pernet_subsys(&if6_proc_net_ops);
3922 }
3923 #endif  /* CONFIG_PROC_FS */
3924
3925 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3926 /* Check if address is a home address configured on any interface. */
3927 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3928 {
3929         int ret = 0;
3930         struct inet6_ifaddr *ifp = NULL;
3931         unsigned int hash = inet6_addr_hash(addr);
3932
3933         rcu_read_lock_bh();
3934         hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3935                 if (!net_eq(dev_net(ifp->idev->dev), net))
3936                         continue;
3937                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3938                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3939                         ret = 1;
3940                         break;
3941                 }
3942         }
3943         rcu_read_unlock_bh();
3944         return ret;
3945 }
3946 #endif
3947
3948 /*
3949  *      Periodic address status verification
3950  */
3951
3952 static void addrconf_verify_rtnl(void)
3953 {
3954         unsigned long now, next, next_sec, next_sched;
3955         struct inet6_ifaddr *ifp;
3956         int i;
3957
3958         ASSERT_RTNL();
3959
3960         rcu_read_lock_bh();
3961         now = jiffies;
3962         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3963
3964         cancel_delayed_work(&addr_chk_work);
3965
3966         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3967 restart:
3968                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3969                         unsigned long age;
3970
3971                         /* When setting preferred_lft to a value not zero or
3972                          * infinity, while valid_lft is infinity
3973                          * IFA_F_PERMANENT has a non-infinity life time.
3974                          */
3975                         if ((ifp->flags & IFA_F_PERMANENT) &&
3976                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
3977                                 continue;
3978
3979                         spin_lock(&ifp->lock);
3980                         /* We try to batch several events at once. */
3981                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3982
3983                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3984                             age >= ifp->valid_lft) {
3985                                 spin_unlock(&ifp->lock);
3986                                 in6_ifa_hold(ifp);
3987                                 ipv6_del_addr(ifp);
3988                                 goto restart;
3989                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3990                                 spin_unlock(&ifp->lock);
3991                                 continue;
3992                         } else if (age >= ifp->prefered_lft) {
3993                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3994                                 int deprecate = 0;
3995
3996                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3997                                         deprecate = 1;
3998                                         ifp->flags |= IFA_F_DEPRECATED;
3999                                 }
4000
4001                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4002                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4003                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4004
4005                                 spin_unlock(&ifp->lock);
4006
4007                                 if (deprecate) {
4008                                         in6_ifa_hold(ifp);
4009
4010                                         ipv6_ifa_notify(0, ifp);
4011                                         in6_ifa_put(ifp);
4012                                         goto restart;
4013                                 }
4014                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4015                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4016                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4017                                         ifp->idev->cnf.dad_transmits *
4018                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4019
4020                                 if (age >= ifp->prefered_lft - regen_advance) {
4021                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4022                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4023                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4024                                         if (!ifp->regen_count && ifpub) {
4025                                                 ifp->regen_count++;
4026                                                 in6_ifa_hold(ifp);
4027                                                 in6_ifa_hold(ifpub);
4028                                                 spin_unlock(&ifp->lock);
4029
4030                                                 spin_lock(&ifpub->lock);
4031                                                 ifpub->regen_count = 0;
4032                                                 spin_unlock(&ifpub->lock);
4033                                                 ipv6_create_tempaddr(ifpub, ifp);
4034                                                 in6_ifa_put(ifpub);
4035                                                 in6_ifa_put(ifp);
4036                                                 goto restart;
4037                                         }
4038                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4039                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4040                                 spin_unlock(&ifp->lock);
4041                         } else {
4042                                 /* ifp->prefered_lft <= ifp->valid_lft */
4043                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4044                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4045                                 spin_unlock(&ifp->lock);
4046                         }
4047                 }
4048         }
4049
4050         next_sec = round_jiffies_up(next);
4051         next_sched = next;
4052
4053         /* If rounded timeout is accurate enough, accept it. */
4054         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4055                 next_sched = next_sec;
4056
4057         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4058         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4059                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4060
4061         ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4062               now, next, next_sec, next_sched);
4063         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4064         rcu_read_unlock_bh();
4065 }
4066
4067 static void addrconf_verify_work(struct work_struct *w)
4068 {
4069         rtnl_lock();
4070         addrconf_verify_rtnl();
4071         rtnl_unlock();
4072 }
4073
4074 static void addrconf_verify(void)
4075 {
4076         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4077 }
4078
4079 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4080                                      struct in6_addr **peer_pfx)
4081 {
4082         struct in6_addr *pfx = NULL;
4083
4084         *peer_pfx = NULL;
4085
4086         if (addr)
4087                 pfx = nla_data(addr);
4088
4089         if (local) {
4090                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4091                         *peer_pfx = pfx;
4092                 pfx = nla_data(local);
4093         }
4094
4095         return pfx;
4096 }
4097
4098 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4099         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4100         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4101         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4102         [IFA_FLAGS]             = { .len = sizeof(u32) },
4103 };
4104
4105 static int
4106 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4107 {
4108         struct net *net = sock_net(skb->sk);
4109         struct ifaddrmsg *ifm;
4110         struct nlattr *tb[IFA_MAX+1];
4111         struct in6_addr *pfx, *peer_pfx;
4112         u32 ifa_flags;
4113         int err;
4114
4115         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4116         if (err < 0)
4117                 return err;
4118
4119         ifm = nlmsg_data(nlh);
4120         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4121         if (!pfx)
4122                 return -EINVAL;
4123
4124         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4125
4126         /* We ignore other flags so far. */
4127         ifa_flags &= IFA_F_MANAGETEMPADDR;
4128
4129         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4130                               ifm->ifa_prefixlen);
4131 }
4132
4133 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4134                              u32 prefered_lft, u32 valid_lft)
4135 {
4136         u32 flags;
4137         clock_t expires;
4138         unsigned long timeout;
4139         bool was_managetempaddr;
4140         bool had_prefixroute;
4141
4142         ASSERT_RTNL();
4143
4144         if (!valid_lft || (prefered_lft > valid_lft))
4145                 return -EINVAL;
4146
4147         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4148             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4149                 return -EINVAL;
4150
4151         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4152         if (addrconf_finite_timeout(timeout)) {
4153                 expires = jiffies_to_clock_t(timeout * HZ);
4154                 valid_lft = timeout;
4155                 flags = RTF_EXPIRES;
4156         } else {
4157                 expires = 0;
4158                 flags = 0;
4159                 ifa_flags |= IFA_F_PERMANENT;
4160         }
4161
4162         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4163         if (addrconf_finite_timeout(timeout)) {
4164                 if (timeout == 0)
4165                         ifa_flags |= IFA_F_DEPRECATED;
4166                 prefered_lft = timeout;
4167         }
4168
4169         spin_lock_bh(&ifp->lock);
4170         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4171         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4172                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4173         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4174                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4175                         IFA_F_NOPREFIXROUTE);
4176         ifp->flags |= ifa_flags;
4177         ifp->tstamp = jiffies;
4178         ifp->valid_lft = valid_lft;
4179         ifp->prefered_lft = prefered_lft;
4180
4181         spin_unlock_bh(&ifp->lock);
4182         if (!(ifp->flags&IFA_F_TENTATIVE))
4183                 ipv6_ifa_notify(0, ifp);
4184
4185         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4186                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4187                                       expires, flags);
4188         } else if (had_prefixroute) {
4189                 enum cleanup_prefix_rt_t action;
4190                 unsigned long rt_expires;
4191
4192                 write_lock_bh(&ifp->idev->lock);
4193                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4194                 write_unlock_bh(&ifp->idev->lock);
4195
4196                 if (action != CLEANUP_PREFIX_RT_NOP) {
4197                         cleanup_prefix_route(ifp, rt_expires,
4198                                 action == CLEANUP_PREFIX_RT_DEL);
4199                 }
4200         }
4201
4202         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4203                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4204                         valid_lft = prefered_lft = 0;
4205                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4206                                  !was_managetempaddr, jiffies);
4207         }
4208
4209         addrconf_verify_rtnl();
4210
4211         return 0;
4212 }
4213
4214 static int
4215 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4216 {
4217         struct net *net = sock_net(skb->sk);
4218         struct ifaddrmsg *ifm;
4219         struct nlattr *tb[IFA_MAX+1];
4220         struct in6_addr *pfx, *peer_pfx;
4221         struct inet6_ifaddr *ifa;
4222         struct net_device *dev;
4223         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4224         u32 ifa_flags;
4225         int err;
4226
4227         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4228         if (err < 0)
4229                 return err;
4230
4231         ifm = nlmsg_data(nlh);
4232         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4233         if (!pfx)
4234                 return -EINVAL;
4235
4236         if (tb[IFA_CACHEINFO]) {
4237                 struct ifa_cacheinfo *ci;
4238
4239                 ci = nla_data(tb[IFA_CACHEINFO]);
4240                 valid_lft = ci->ifa_valid;
4241                 preferred_lft = ci->ifa_prefered;
4242         } else {
4243                 preferred_lft = INFINITY_LIFE_TIME;
4244                 valid_lft = INFINITY_LIFE_TIME;
4245         }
4246
4247         dev =  __dev_get_by_index(net, ifm->ifa_index);
4248         if (!dev)
4249                 return -ENODEV;
4250
4251         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4252
4253         /* We ignore other flags so far. */
4254         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4255                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4256
4257         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4258         if (!ifa) {
4259                 /*
4260                  * It would be best to check for !NLM_F_CREATE here but
4261                  * userspace already relies on not having to provide this.
4262                  */
4263                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4264                                       ifm->ifa_prefixlen, ifa_flags,
4265                                       preferred_lft, valid_lft);
4266         }
4267
4268         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4269             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4270                 err = -EEXIST;
4271         else
4272                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4273
4274         in6_ifa_put(ifa);
4275
4276         return err;
4277 }
4278
4279 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4280                           u8 scope, int ifindex)
4281 {
4282         struct ifaddrmsg *ifm;
4283
4284         ifm = nlmsg_data(nlh);
4285         ifm->ifa_family = AF_INET6;
4286         ifm->ifa_prefixlen = prefixlen;
4287         ifm->ifa_flags = flags;
4288         ifm->ifa_scope = scope;
4289         ifm->ifa_index = ifindex;
4290 }
4291
4292 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4293                          unsigned long tstamp, u32 preferred, u32 valid)
4294 {
4295         struct ifa_cacheinfo ci;
4296
4297         ci.cstamp = cstamp_delta(cstamp);
4298         ci.tstamp = cstamp_delta(tstamp);
4299         ci.ifa_prefered = preferred;
4300         ci.ifa_valid = valid;
4301
4302         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4303 }
4304
4305 static inline int rt_scope(int ifa_scope)
4306 {
4307         if (ifa_scope & IFA_HOST)
4308                 return RT_SCOPE_HOST;
4309         else if (ifa_scope & IFA_LINK)
4310                 return RT_SCOPE_LINK;
4311         else if (ifa_scope & IFA_SITE)
4312                 return RT_SCOPE_SITE;
4313         else
4314                 return RT_SCOPE_UNIVERSE;
4315 }
4316
4317 static inline int inet6_ifaddr_msgsize(void)
4318 {
4319         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4320                + nla_total_size(16) /* IFA_LOCAL */
4321                + nla_total_size(16) /* IFA_ADDRESS */
4322                + nla_total_size(sizeof(struct ifa_cacheinfo))
4323                + nla_total_size(4)  /* IFA_FLAGS */;
4324 }
4325
4326 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4327                              u32 portid, u32 seq, int event, unsigned int flags)
4328 {
4329         struct nlmsghdr  *nlh;
4330         u32 preferred, valid;
4331
4332         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4333         if (!nlh)
4334                 return -EMSGSIZE;
4335
4336         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4337                       ifa->idev->dev->ifindex);
4338
4339         if (!((ifa->flags&IFA_F_PERMANENT) &&
4340               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4341                 preferred = ifa->prefered_lft;
4342                 valid = ifa->valid_lft;
4343                 if (preferred != INFINITY_LIFE_TIME) {
4344                         long tval = (jiffies - ifa->tstamp)/HZ;
4345                         if (preferred > tval)
4346                                 preferred -= tval;
4347                         else
4348                                 preferred = 0;
4349                         if (valid != INFINITY_LIFE_TIME) {
4350                                 if (valid > tval)
4351                                         valid -= tval;
4352                                 else
4353                                         valid = 0;
4354                         }
4355                 }
4356         } else {
4357                 preferred = INFINITY_LIFE_TIME;
4358                 valid = INFINITY_LIFE_TIME;
4359         }
4360
4361         if (!ipv6_addr_any(&ifa->peer_addr)) {
4362                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4363                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4364                         goto error;
4365         } else
4366                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4367                         goto error;
4368
4369         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4370                 goto error;
4371
4372         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4373                 goto error;
4374
4375         nlmsg_end(skb, nlh);
4376         return 0;
4377
4378 error:
4379         nlmsg_cancel(skb, nlh);
4380         return -EMSGSIZE;
4381 }
4382
4383 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4384                                 u32 portid, u32 seq, int event, u16 flags)
4385 {
4386         struct nlmsghdr  *nlh;
4387         u8 scope = RT_SCOPE_UNIVERSE;
4388         int ifindex = ifmca->idev->dev->ifindex;
4389
4390         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4391                 scope = RT_SCOPE_SITE;
4392
4393         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4394         if (!nlh)
4395                 return -EMSGSIZE;
4396
4397         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4398         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4399             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4400                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4401                 nlmsg_cancel(skb, nlh);
4402                 return -EMSGSIZE;
4403         }
4404
4405         nlmsg_end(skb, nlh);
4406         return 0;
4407 }
4408
4409 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4410                                 u32 portid, u32 seq, int event, unsigned int flags)
4411 {
4412         struct nlmsghdr  *nlh;
4413         u8 scope = RT_SCOPE_UNIVERSE;
4414         int ifindex = ifaca->aca_idev->dev->ifindex;
4415
4416         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4417                 scope = RT_SCOPE_SITE;
4418
4419         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4420         if (!nlh)
4421                 return -EMSGSIZE;
4422
4423         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4424         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4425             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4426                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4427                 nlmsg_cancel(skb, nlh);
4428                 return -EMSGSIZE;
4429         }
4430
4431         nlmsg_end(skb, nlh);
4432         return 0;
4433 }
4434
4435 enum addr_type_t {
4436         UNICAST_ADDR,
4437         MULTICAST_ADDR,
4438         ANYCAST_ADDR,
4439 };
4440
4441 /* called with rcu_read_lock() */
4442 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4443                           struct netlink_callback *cb, enum addr_type_t type,
4444                           int s_ip_idx, int *p_ip_idx)
4445 {
4446         struct ifmcaddr6 *ifmca;
4447         struct ifacaddr6 *ifaca;
4448         int err = 1;
4449         int ip_idx = *p_ip_idx;
4450
4451         read_lock_bh(&idev->lock);
4452         switch (type) {
4453         case UNICAST_ADDR: {
4454                 struct inet6_ifaddr *ifa;
4455
4456                 /* unicast address incl. temp addr */
4457                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4458                         if (++ip_idx < s_ip_idx)
4459                                 continue;
4460                         err = inet6_fill_ifaddr(skb, ifa,
4461                                                 NETLINK_CB(cb->skb).portid,
4462                                                 cb->nlh->nlmsg_seq,
4463                                                 RTM_NEWADDR,
4464                                                 NLM_F_MULTI);
4465                         if (err < 0)
4466                                 break;
4467                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4468                 }
4469                 break;
4470         }
4471         case MULTICAST_ADDR:
4472                 /* multicast address */
4473                 for (ifmca = idev->mc_list; ifmca;
4474                      ifmca = ifmca->next, ip_idx++) {
4475                         if (ip_idx < s_ip_idx)
4476                                 continue;
4477                         err = inet6_fill_ifmcaddr(skb, ifmca,
4478                                                   NETLINK_CB(cb->skb).portid,
4479                                                   cb->nlh->nlmsg_seq,
4480                                                   RTM_GETMULTICAST,
4481                                                   NLM_F_MULTI);
4482                         if (err < 0)
4483                                 break;
4484                 }
4485                 break;
4486         case ANYCAST_ADDR:
4487                 /* anycast address */
4488                 for (ifaca = idev->ac_list; ifaca;
4489                      ifaca = ifaca->aca_next, ip_idx++) {
4490                         if (ip_idx < s_ip_idx)
4491                                 continue;
4492                         err = inet6_fill_ifacaddr(skb, ifaca,
4493                                                   NETLINK_CB(cb->skb).portid,
4494                                                   cb->nlh->nlmsg_seq,
4495                                                   RTM_GETANYCAST,
4496                                                   NLM_F_MULTI);
4497                         if (err < 0)
4498                                 break;
4499                 }
4500                 break;
4501         default:
4502                 break;
4503         }
4504         read_unlock_bh(&idev->lock);
4505         *p_ip_idx = ip_idx;
4506         return err;
4507 }
4508
4509 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4510                            enum addr_type_t type)
4511 {
4512         struct net *net = sock_net(skb->sk);
4513         int h, s_h;
4514         int idx, ip_idx;
4515         int s_idx, s_ip_idx;
4516         struct net_device *dev;
4517         struct inet6_dev *idev;
4518         struct hlist_head *head;
4519
4520         s_h = cb->args[0];
4521         s_idx = idx = cb->args[1];
4522         s_ip_idx = ip_idx = cb->args[2];
4523
4524         rcu_read_lock();
4525         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4526         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4527                 idx = 0;
4528                 head = &net->dev_index_head[h];
4529                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4530                         if (idx < s_idx)
4531                                 goto cont;
4532                         if (h > s_h || idx > s_idx)
4533                                 s_ip_idx = 0;
4534                         ip_idx = 0;
4535                         idev = __in6_dev_get(dev);
4536                         if (!idev)
4537                                 goto cont;
4538
4539                         if (in6_dump_addrs(idev, skb, cb, type,
4540                                            s_ip_idx, &ip_idx) < 0)
4541                                 goto done;
4542 cont:
4543                         idx++;
4544                 }
4545         }
4546 done:
4547         rcu_read_unlock();
4548         cb->args[0] = h;
4549         cb->args[1] = idx;
4550         cb->args[2] = ip_idx;
4551
4552         return skb->len;
4553 }
4554
4555 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4556 {
4557         enum addr_type_t type = UNICAST_ADDR;
4558
4559         return inet6_dump_addr(skb, cb, type);
4560 }
4561
4562 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4563 {
4564         enum addr_type_t type = MULTICAST_ADDR;
4565
4566         return inet6_dump_addr(skb, cb, type);
4567 }
4568
4569
4570 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4571 {
4572         enum addr_type_t type = ANYCAST_ADDR;
4573
4574         return inet6_dump_addr(skb, cb, type);
4575 }
4576
4577 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4578 {
4579         struct net *net = sock_net(in_skb->sk);
4580         struct ifaddrmsg *ifm;
4581         struct nlattr *tb[IFA_MAX+1];
4582         struct in6_addr *addr = NULL, *peer;
4583         struct net_device *dev = NULL;
4584         struct inet6_ifaddr *ifa;
4585         struct sk_buff *skb;
4586         int err;
4587
4588         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4589         if (err < 0)
4590                 goto errout;
4591
4592         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4593         if (!addr) {
4594                 err = -EINVAL;
4595                 goto errout;
4596         }
4597
4598         ifm = nlmsg_data(nlh);
4599         if (ifm->ifa_index)
4600                 dev = __dev_get_by_index(net, ifm->ifa_index);
4601
4602         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4603         if (!ifa) {
4604                 err = -EADDRNOTAVAIL;
4605                 goto errout;
4606         }
4607
4608         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4609         if (!skb) {
4610                 err = -ENOBUFS;
4611                 goto errout_ifa;
4612         }
4613
4614         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4615                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4616         if (err < 0) {
4617                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4618                 WARN_ON(err == -EMSGSIZE);
4619                 kfree_skb(skb);
4620                 goto errout_ifa;
4621         }
4622         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4623 errout_ifa:
4624         in6_ifa_put(ifa);
4625 errout:
4626         return err;
4627 }
4628
4629 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4630 {
4631         struct sk_buff *skb;
4632         struct net *net = dev_net(ifa->idev->dev);
4633         int err = -ENOBUFS;
4634
4635         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4636         if (!skb)
4637                 goto errout;
4638
4639         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4640         if (err < 0) {
4641                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4642                 WARN_ON(err == -EMSGSIZE);
4643                 kfree_skb(skb);
4644                 goto errout;
4645         }
4646         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4647         return;
4648 errout:
4649         if (err < 0)
4650                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4651 }
4652
4653 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4654                                 __s32 *array, int bytes)
4655 {
4656         BUG_ON(bytes < (DEVCONF_MAX * 4));
4657
4658         memset(array, 0, bytes);
4659         array[DEVCONF_FORWARDING] = cnf->forwarding;
4660         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4661         array[DEVCONF_MTU6] = cnf->mtu6;
4662         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4663         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4664         array[DEVCONF_AUTOCONF] = cnf->autoconf;
4665         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4666         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4667         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4668                 jiffies_to_msecs(cnf->rtr_solicit_interval);
4669         array[DEVCONF_RTR_SOLICIT_DELAY] =
4670                 jiffies_to_msecs(cnf->rtr_solicit_delay);
4671         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4672         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4673                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4674         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4675                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4676         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4677         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4678         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4679         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4680         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4681         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4682         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4683         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4684         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4685 #ifdef CONFIG_IPV6_ROUTER_PREF
4686         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4687         array[DEVCONF_RTR_PROBE_INTERVAL] =
4688                 jiffies_to_msecs(cnf->rtr_probe_interval);
4689 #ifdef CONFIG_IPV6_ROUTE_INFO
4690         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4691 #endif
4692 #endif
4693         array[DEVCONF_ACCEPT_RA_RT_TABLE] = cnf->accept_ra_rt_table;
4694         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4695         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4696 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4697         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4698         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4699 #endif
4700 #ifdef CONFIG_IPV6_MROUTE
4701         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4702 #endif
4703         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4704         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4705         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4706         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4707         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4708         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4709         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4710         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4711         /* we omit DEVCONF_STABLE_SECRET for now */
4712         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4713 }
4714
4715 static inline size_t inet6_ifla6_size(void)
4716 {
4717         return nla_total_size(4) /* IFLA_INET6_FLAGS */
4718              + nla_total_size(sizeof(struct ifla_cacheinfo))
4719              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4720              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4721              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4722              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4723 }
4724
4725 static inline size_t inet6_if_nlmsg_size(void)
4726 {
4727         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4728                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4729                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4730                + nla_total_size(4) /* IFLA_MTU */
4731                + nla_total_size(4) /* IFLA_LINK */
4732                + nla_total_size(1) /* IFLA_OPERSTATE */
4733                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4734 }
4735
4736 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4737                                       int items, int bytes)
4738 {
4739         int i;
4740         int pad = bytes - sizeof(u64) * items;
4741         BUG_ON(pad < 0);
4742
4743         /* Use put_unaligned() because stats may not be aligned for u64. */
4744         put_unaligned(items, &stats[0]);
4745         for (i = 1; i < items; i++)
4746                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4747
4748         memset(&stats[items], 0, pad);
4749 }
4750
4751 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4752                                         int bytes, size_t syncpoff)
4753 {
4754         int i, c;
4755         u64 buff[IPSTATS_MIB_MAX];
4756         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4757
4758         BUG_ON(pad < 0);
4759
4760         memset(buff, 0, sizeof(buff));
4761         buff[0] = IPSTATS_MIB_MAX;
4762
4763         for_each_possible_cpu(c) {
4764                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4765                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4766         }
4767
4768         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4769         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4770 }
4771
4772 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4773                              int bytes)
4774 {
4775         switch (attrtype) {
4776         case IFLA_INET6_STATS:
4777                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4778                                      offsetof(struct ipstats_mib, syncp));
4779                 break;
4780         case IFLA_INET6_ICMP6STATS:
4781                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4782                 break;
4783         }
4784 }
4785
4786 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4787                                   u32 ext_filter_mask)
4788 {
4789         struct nlattr *nla;
4790         struct ifla_cacheinfo ci;
4791
4792         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4793                 goto nla_put_failure;
4794         ci.max_reasm_len = IPV6_MAXPLEN;
4795         ci.tstamp = cstamp_delta(idev->tstamp);
4796         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4797         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4798         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4799                 goto nla_put_failure;
4800         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4801         if (!nla)
4802                 goto nla_put_failure;
4803         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4804
4805         /* XXX - MC not implemented */
4806
4807         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4808                 return 0;
4809
4810         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4811         if (!nla)
4812                 goto nla_put_failure;
4813         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4814
4815         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4816         if (!nla)
4817                 goto nla_put_failure;
4818         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4819
4820         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4821         if (!nla)
4822                 goto nla_put_failure;
4823
4824         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4825                 goto nla_put_failure;
4826
4827         read_lock_bh(&idev->lock);
4828         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4829         read_unlock_bh(&idev->lock);
4830
4831         return 0;
4832
4833 nla_put_failure:
4834         return -EMSGSIZE;
4835 }
4836
4837 static size_t inet6_get_link_af_size(const struct net_device *dev,
4838                                      u32 ext_filter_mask)
4839 {
4840         if (!__in6_dev_get(dev))
4841                 return 0;
4842
4843         return inet6_ifla6_size();
4844 }
4845
4846 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4847                               u32 ext_filter_mask)
4848 {
4849         struct inet6_dev *idev = __in6_dev_get(dev);
4850
4851         if (!idev)
4852                 return -ENODATA;
4853
4854         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4855                 return -EMSGSIZE;
4856
4857         return 0;
4858 }
4859
4860 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4861 {
4862         struct inet6_ifaddr *ifp;
4863         struct net_device *dev = idev->dev;
4864         bool update_rs = false;
4865         struct in6_addr ll_addr;
4866
4867         ASSERT_RTNL();
4868
4869         if (!token)
4870                 return -EINVAL;
4871         if (ipv6_addr_any(token))
4872                 return -EINVAL;
4873         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4874                 return -EINVAL;
4875         if (!ipv6_accept_ra(idev))
4876                 return -EINVAL;
4877         if (idev->cnf.rtr_solicits <= 0)
4878                 return -EINVAL;
4879
4880         write_lock_bh(&idev->lock);
4881
4882         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4883         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4884
4885         write_unlock_bh(&idev->lock);
4886
4887         if (!idev->dead && (idev->if_flags & IF_READY) &&
4888             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4889                              IFA_F_OPTIMISTIC)) {
4890
4891                 /* If we're not ready, then normal ifup will take care
4892                  * of this. Otherwise, we need to request our rs here.
4893                  */
4894                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4895                 update_rs = true;
4896         }
4897
4898         write_lock_bh(&idev->lock);
4899
4900         if (update_rs) {
4901                 idev->if_flags |= IF_RS_SENT;
4902                 idev->rs_probes = 1;
4903                 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4904         }
4905
4906         /* Well, that's kinda nasty ... */
4907         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4908                 spin_lock(&ifp->lock);
4909                 if (ifp->tokenized) {
4910                         ifp->valid_lft = 0;
4911                         ifp->prefered_lft = 0;
4912                 }
4913                 spin_unlock(&ifp->lock);
4914         }
4915
4916         write_unlock_bh(&idev->lock);
4917         inet6_ifinfo_notify(RTM_NEWLINK, idev);
4918         addrconf_verify_rtnl();
4919         return 0;
4920 }
4921
4922 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4923         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
4924         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
4925 };
4926
4927 static int inet6_validate_link_af(const struct net_device *dev,
4928                                   const struct nlattr *nla)
4929 {
4930         struct nlattr *tb[IFLA_INET6_MAX + 1];
4931
4932         if (dev && !__in6_dev_get(dev))
4933                 return -EAFNOSUPPORT;
4934
4935         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4936 }
4937
4938 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4939 {
4940         int err = -EINVAL;
4941         struct inet6_dev *idev = __in6_dev_get(dev);
4942         struct nlattr *tb[IFLA_INET6_MAX + 1];
4943
4944         if (!idev)
4945                 return -EAFNOSUPPORT;
4946
4947         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4948                 BUG();
4949
4950         if (tb[IFLA_INET6_TOKEN]) {
4951                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4952                 if (err)
4953                         return err;
4954         }
4955
4956         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4957                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4958
4959                 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4960                     mode != IN6_ADDR_GEN_MODE_NONE &&
4961                     mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4962                         return -EINVAL;
4963
4964                 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4965                     !idev->cnf.stable_secret.initialized &&
4966                     !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4967                         return -EINVAL;
4968
4969                 idev->addr_gen_mode = mode;
4970                 err = 0;
4971         }
4972
4973         return err;
4974 }
4975
4976 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4977                              u32 portid, u32 seq, int event, unsigned int flags)
4978 {
4979         struct net_device *dev = idev->dev;
4980         struct ifinfomsg *hdr;
4981         struct nlmsghdr *nlh;
4982         void *protoinfo;
4983
4984         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4985         if (!nlh)
4986                 return -EMSGSIZE;
4987
4988         hdr = nlmsg_data(nlh);
4989         hdr->ifi_family = AF_INET6;
4990         hdr->__ifi_pad = 0;
4991         hdr->ifi_type = dev->type;
4992         hdr->ifi_index = dev->ifindex;
4993         hdr->ifi_flags = dev_get_flags(dev);
4994         hdr->ifi_change = 0;
4995
4996         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4997             (dev->addr_len &&
4998              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4999             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5000             (dev->ifindex != dev_get_iflink(dev) &&
5001              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5002             nla_put_u8(skb, IFLA_OPERSTATE,
5003                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5004                 goto nla_put_failure;
5005         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5006         if (!protoinfo)
5007                 goto nla_put_failure;
5008
5009         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5010                 goto nla_put_failure;
5011
5012         nla_nest_end(skb, protoinfo);
5013         nlmsg_end(skb, nlh);
5014         return 0;
5015
5016 nla_put_failure:
5017         nlmsg_cancel(skb, nlh);
5018         return -EMSGSIZE;
5019 }
5020
5021 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5022 {
5023         struct net *net = sock_net(skb->sk);
5024         int h, s_h;
5025         int idx = 0, s_idx;
5026         struct net_device *dev;
5027         struct inet6_dev *idev;
5028         struct hlist_head *head;
5029
5030         s_h = cb->args[0];
5031         s_idx = cb->args[1];
5032
5033         rcu_read_lock();
5034         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5035                 idx = 0;
5036                 head = &net->dev_index_head[h];
5037                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5038                         if (idx < s_idx)
5039                                 goto cont;
5040                         idev = __in6_dev_get(dev);
5041                         if (!idev)
5042                                 goto cont;
5043                         if (inet6_fill_ifinfo(skb, idev,
5044                                               NETLINK_CB(cb->skb).portid,
5045                                               cb->nlh->nlmsg_seq,
5046                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5047                                 goto out;
5048 cont:
5049                         idx++;
5050                 }
5051         }
5052 out:
5053         rcu_read_unlock();
5054         cb->args[1] = idx;
5055         cb->args[0] = h;
5056
5057         return skb->len;
5058 }
5059
5060 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5061 {
5062         struct sk_buff *skb;
5063         struct net *net = dev_net(idev->dev);
5064         int err = -ENOBUFS;
5065
5066         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5067         if (!skb)
5068                 goto errout;
5069
5070         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5071         if (err < 0) {
5072                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5073                 WARN_ON(err == -EMSGSIZE);
5074                 kfree_skb(skb);
5075                 goto errout;
5076         }
5077         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5078         return;
5079 errout:
5080         if (err < 0)
5081                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5082 }
5083
5084 static inline size_t inet6_prefix_nlmsg_size(void)
5085 {
5086         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5087                + nla_total_size(sizeof(struct in6_addr))
5088                + nla_total_size(sizeof(struct prefix_cacheinfo));
5089 }
5090
5091 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5092                              struct prefix_info *pinfo, u32 portid, u32 seq,
5093                              int event, unsigned int flags)
5094 {
5095         struct prefixmsg *pmsg;
5096         struct nlmsghdr *nlh;
5097         struct prefix_cacheinfo ci;
5098
5099         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5100         if (!nlh)
5101                 return -EMSGSIZE;
5102
5103         pmsg = nlmsg_data(nlh);
5104         pmsg->prefix_family = AF_INET6;
5105         pmsg->prefix_pad1 = 0;
5106         pmsg->prefix_pad2 = 0;
5107         pmsg->prefix_ifindex = idev->dev->ifindex;
5108         pmsg->prefix_len = pinfo->prefix_len;
5109         pmsg->prefix_type = pinfo->type;
5110         pmsg->prefix_pad3 = 0;
5111         pmsg->prefix_flags = 0;
5112         if (pinfo->onlink)
5113                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5114         if (pinfo->autoconf)
5115                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5116
5117         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5118                 goto nla_put_failure;
5119         ci.preferred_time = ntohl(pinfo->prefered);
5120         ci.valid_time = ntohl(pinfo->valid);
5121         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5122                 goto nla_put_failure;
5123         nlmsg_end(skb, nlh);
5124         return 0;
5125
5126 nla_put_failure:
5127         nlmsg_cancel(skb, nlh);
5128         return -EMSGSIZE;
5129 }
5130
5131 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5132                          struct prefix_info *pinfo)
5133 {
5134         struct sk_buff *skb;
5135         struct net *net = dev_net(idev->dev);
5136         int err = -ENOBUFS;
5137
5138         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5139         if (!skb)
5140                 goto errout;
5141
5142         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5143         if (err < 0) {
5144                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5145                 WARN_ON(err == -EMSGSIZE);
5146                 kfree_skb(skb);
5147                 goto errout;
5148         }
5149         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5150         return;
5151 errout:
5152         if (err < 0)
5153                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5154 }
5155
5156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5157 {
5158         struct net *net = dev_net(ifp->idev->dev);
5159
5160         if (event)
5161                 ASSERT_RTNL();
5162
5163         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5164
5165         switch (event) {
5166         case RTM_NEWADDR:
5167                 /*
5168                  * If the address was optimistic
5169                  * we inserted the route at the start of
5170                  * our DAD process, so we don't need
5171                  * to do it again
5172                  */
5173                 if (!(ifp->rt->rt6i_node))
5174                         ip6_ins_rt(ifp->rt);
5175                 if (ifp->idev->cnf.forwarding)
5176                         addrconf_join_anycast(ifp);
5177                 if (!ipv6_addr_any(&ifp->peer_addr))
5178                         addrconf_prefix_route(&ifp->peer_addr, 128,
5179                                               ifp->idev->dev, 0, 0);
5180                 break;
5181         case RTM_DELADDR:
5182                 if (ifp->idev->cnf.forwarding)
5183                         addrconf_leave_anycast(ifp);
5184                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5185                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5186                         struct rt6_info *rt;
5187
5188                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5189                                                        ifp->idev->dev, 0, 0);
5190                         if (rt)
5191                                 ip6_del_rt(rt);
5192                 }
5193                 dst_hold(&ifp->rt->dst);
5194
5195                 ip6_del_rt(ifp->rt);
5196
5197                 rt_genid_bump_ipv6(net);
5198                 break;
5199         }
5200         atomic_inc(&net->ipv6.dev_addr_genid);
5201 }
5202
5203 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5204 {
5205         rcu_read_lock_bh();
5206         if (likely(ifp->idev->dead == 0))
5207                 __ipv6_ifa_notify(event, ifp);
5208         rcu_read_unlock_bh();
5209 }
5210
5211 #ifdef CONFIG_SYSCTL
5212
5213 static
5214 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5215                            void __user *buffer, size_t *lenp, loff_t *ppos)
5216 {
5217         int *valp = ctl->data;
5218         int val = *valp;
5219         loff_t pos = *ppos;
5220         struct ctl_table lctl;
5221         int ret;
5222
5223         /*
5224          * ctl->data points to idev->cnf.forwarding, we should
5225          * not modify it until we get the rtnl lock.
5226          */
5227         lctl = *ctl;
5228         lctl.data = &val;
5229
5230         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5231
5232         if (write)
5233                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5234         if (ret)
5235                 *ppos = pos;
5236         return ret;
5237 }
5238
5239 static
5240 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5241                         void __user *buffer, size_t *lenp, loff_t *ppos)
5242 {
5243         struct inet6_dev *idev = ctl->extra1;
5244         int min_mtu = IPV6_MIN_MTU;
5245         struct ctl_table lctl;
5246
5247         lctl = *ctl;
5248         lctl.extra1 = &min_mtu;
5249         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5250
5251         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5252 }
5253
5254 static void dev_disable_change(struct inet6_dev *idev)
5255 {
5256         struct netdev_notifier_info info;
5257
5258         if (!idev || !idev->dev)
5259                 return;
5260
5261         netdev_notifier_info_init(&info, idev->dev);
5262         if (idev->cnf.disable_ipv6)
5263                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5264         else
5265                 addrconf_notify(NULL, NETDEV_UP, &info);
5266 }
5267
5268 static void addrconf_disable_change(struct net *net, __s32 newf)
5269 {
5270         struct net_device *dev;
5271         struct inet6_dev *idev;
5272
5273         rcu_read_lock();
5274         for_each_netdev_rcu(net, dev) {
5275                 idev = __in6_dev_get(dev);
5276                 if (idev) {
5277                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5278                         idev->cnf.disable_ipv6 = newf;
5279                         if (changed)
5280                                 dev_disable_change(idev);
5281                 }
5282         }
5283         rcu_read_unlock();
5284 }
5285
5286 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5287 {
5288         struct net *net;
5289         int old;
5290
5291         if (!rtnl_trylock())
5292                 return restart_syscall();
5293
5294         net = (struct net *)table->extra2;
5295         old = *p;
5296         *p = newf;
5297
5298         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5299                 rtnl_unlock();
5300                 return 0;
5301         }
5302
5303         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5304                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5305                 addrconf_disable_change(net, newf);
5306         } else if ((!newf) ^ (!old))
5307                 dev_disable_change((struct inet6_dev *)table->extra1);
5308
5309         rtnl_unlock();
5310         return 0;
5311 }
5312
5313 static
5314 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5315                             void __user *buffer, size_t *lenp, loff_t *ppos)
5316 {
5317         int *valp = ctl->data;
5318         int val = *valp;
5319         loff_t pos = *ppos;
5320         struct ctl_table lctl;
5321         int ret;
5322
5323         /*
5324          * ctl->data points to idev->cnf.disable_ipv6, we should
5325          * not modify it until we get the rtnl lock.
5326          */
5327         lctl = *ctl;
5328         lctl.data = &val;
5329
5330         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5331
5332         if (write)
5333                 ret = addrconf_disable_ipv6(ctl, valp, val);
5334         if (ret)
5335                 *ppos = pos;
5336         return ret;
5337 }
5338
5339 static
5340 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5341                               void __user *buffer, size_t *lenp, loff_t *ppos)
5342 {
5343         int *valp = ctl->data;
5344         int ret;
5345         int old, new;
5346
5347         old = *valp;
5348         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5349         new = *valp;
5350
5351         if (write && old != new) {
5352                 struct net *net = ctl->extra2;
5353
5354                 if (!rtnl_trylock())
5355                         return restart_syscall();
5356
5357                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5358                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5359                                                      NETCONFA_IFINDEX_DEFAULT,
5360                                                      net->ipv6.devconf_dflt);
5361                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5362                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5363                                                      NETCONFA_IFINDEX_ALL,
5364                                                      net->ipv6.devconf_all);
5365                 else {
5366                         struct inet6_dev *idev = ctl->extra1;
5367
5368                         inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5369                                                      idev->dev->ifindex,
5370                                                      &idev->cnf);
5371                 }
5372                 rtnl_unlock();
5373         }
5374
5375         return ret;
5376 }
5377
5378 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5379                                          void __user *buffer, size_t *lenp,
5380                                          loff_t *ppos)
5381 {
5382         int err;
5383         struct in6_addr addr;
5384         char str[IPV6_MAX_STRLEN];
5385         struct ctl_table lctl = *ctl;
5386         struct net *net = ctl->extra2;
5387         struct ipv6_stable_secret *secret = ctl->data;
5388
5389         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5390                 return -EIO;
5391
5392         lctl.maxlen = IPV6_MAX_STRLEN;
5393         lctl.data = str;
5394
5395         if (!rtnl_trylock())
5396                 return restart_syscall();
5397
5398         if (!write && !secret->initialized) {
5399                 err = -EIO;
5400                 goto out;
5401         }
5402
5403         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5404         if (err >= sizeof(str)) {
5405                 err = -EIO;
5406                 goto out;
5407         }
5408
5409         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5410         if (err || !write)
5411                 goto out;
5412
5413         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5414                 err = -EIO;
5415                 goto out;
5416         }
5417
5418         secret->initialized = true;
5419         secret->secret = addr;
5420
5421         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5422                 struct net_device *dev;
5423
5424                 for_each_netdev(net, dev) {
5425                         struct inet6_dev *idev = __in6_dev_get(dev);
5426
5427                         if (idev) {
5428                                 idev->addr_gen_mode =
5429                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5430                         }
5431                 }
5432         } else {
5433                 struct inet6_dev *idev = ctl->extra1;
5434
5435                 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5436         }
5437
5438 out:
5439         rtnl_unlock();
5440
5441         return err;
5442 }
5443
5444 static
5445 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5446                                                 int write,
5447                                                 void __user *buffer,
5448                                                 size_t *lenp,
5449                                                 loff_t *ppos)
5450 {
5451         int *valp = ctl->data;
5452         int val = *valp;
5453         loff_t pos = *ppos;
5454         struct ctl_table lctl;
5455         int ret;
5456
5457         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5458          * we should not modify it until we get the rtnl lock.
5459          */
5460         lctl = *ctl;
5461         lctl.data = &val;
5462
5463         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5464
5465         if (write)
5466                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5467         if (ret)
5468                 *ppos = pos;
5469         return ret;
5470 }
5471
5472 static struct addrconf_sysctl_table
5473 {
5474         struct ctl_table_header *sysctl_header;
5475         struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5476 } addrconf_sysctl __read_mostly = {
5477         .sysctl_header = NULL,
5478         .addrconf_vars = {
5479                 {
5480                         .procname       = "forwarding",
5481                         .data           = &ipv6_devconf.forwarding,
5482                         .maxlen         = sizeof(int),
5483                         .mode           = 0644,
5484                         .proc_handler   = addrconf_sysctl_forward,
5485                 },
5486                 {
5487                         .procname       = "hop_limit",
5488                         .data           = &ipv6_devconf.hop_limit,
5489                         .maxlen         = sizeof(int),
5490                         .mode           = 0644,
5491                         .proc_handler   = proc_dointvec,
5492                 },
5493                 {
5494                         .procname       = "mtu",
5495                         .data           = &ipv6_devconf.mtu6,
5496                         .maxlen         = sizeof(int),
5497                         .mode           = 0644,
5498                         .proc_handler   = addrconf_sysctl_mtu,
5499                 },
5500                 {
5501                         .procname       = "accept_ra",
5502                         .data           = &ipv6_devconf.accept_ra,
5503                         .maxlen         = sizeof(int),
5504                         .mode           = 0644,
5505                         .proc_handler   = proc_dointvec,
5506                 },
5507                 {
5508                         .procname       = "accept_redirects",
5509                         .data           = &ipv6_devconf.accept_redirects,
5510                         .maxlen         = sizeof(int),
5511                         .mode           = 0644,
5512                         .proc_handler   = proc_dointvec,
5513                 },
5514                 {
5515                         .procname       = "autoconf",
5516                         .data           = &ipv6_devconf.autoconf,
5517                         .maxlen         = sizeof(int),
5518                         .mode           = 0644,
5519                         .proc_handler   = proc_dointvec,
5520                 },
5521                 {
5522                         .procname       = "dad_transmits",
5523                         .data           = &ipv6_devconf.dad_transmits,
5524                         .maxlen         = sizeof(int),
5525                         .mode           = 0644,
5526                         .proc_handler   = proc_dointvec,
5527                 },
5528                 {
5529                         .procname       = "router_solicitations",
5530                         .data           = &ipv6_devconf.rtr_solicits,
5531                         .maxlen         = sizeof(int),
5532                         .mode           = 0644,
5533                         .proc_handler   = proc_dointvec,
5534                 },
5535                 {
5536                         .procname       = "router_solicitation_interval",
5537                         .data           = &ipv6_devconf.rtr_solicit_interval,
5538                         .maxlen         = sizeof(int),
5539                         .mode           = 0644,
5540                         .proc_handler   = proc_dointvec_jiffies,
5541                 },
5542                 {
5543                         .procname       = "router_solicitation_delay",
5544                         .data           = &ipv6_devconf.rtr_solicit_delay,
5545                         .maxlen         = sizeof(int),
5546                         .mode           = 0644,
5547                         .proc_handler   = proc_dointvec_jiffies,
5548                 },
5549                 {
5550                         .procname       = "force_mld_version",
5551                         .data           = &ipv6_devconf.force_mld_version,
5552                         .maxlen         = sizeof(int),
5553                         .mode           = 0644,
5554                         .proc_handler   = proc_dointvec,
5555                 },
5556                 {
5557                         .procname       = "mldv1_unsolicited_report_interval",
5558                         .data           =
5559                                 &ipv6_devconf.mldv1_unsolicited_report_interval,
5560                         .maxlen         = sizeof(int),
5561                         .mode           = 0644,
5562                         .proc_handler   = proc_dointvec_ms_jiffies,
5563                 },
5564                 {
5565                         .procname       = "mldv2_unsolicited_report_interval",
5566                         .data           =
5567                                 &ipv6_devconf.mldv2_unsolicited_report_interval,
5568                         .maxlen         = sizeof(int),
5569                         .mode           = 0644,
5570                         .proc_handler   = proc_dointvec_ms_jiffies,
5571                 },
5572                 {
5573                         .procname       = "use_tempaddr",
5574                         .data           = &ipv6_devconf.use_tempaddr,
5575                         .maxlen         = sizeof(int),
5576                         .mode           = 0644,
5577                         .proc_handler   = proc_dointvec,
5578                 },
5579                 {
5580                         .procname       = "temp_valid_lft",
5581                         .data           = &ipv6_devconf.temp_valid_lft,
5582                         .maxlen         = sizeof(int),
5583                         .mode           = 0644,
5584                         .proc_handler   = proc_dointvec,
5585                 },
5586                 {
5587                         .procname       = "temp_prefered_lft",
5588                         .data           = &ipv6_devconf.temp_prefered_lft,
5589                         .maxlen         = sizeof(int),
5590                         .mode           = 0644,
5591                         .proc_handler   = proc_dointvec,
5592                 },
5593                 {
5594                         .procname       = "regen_max_retry",
5595                         .data           = &ipv6_devconf.regen_max_retry,
5596                         .maxlen         = sizeof(int),
5597                         .mode           = 0644,
5598                         .proc_handler   = proc_dointvec,
5599                 },
5600                 {
5601                         .procname       = "max_desync_factor",
5602                         .data           = &ipv6_devconf.max_desync_factor,
5603                         .maxlen         = sizeof(int),
5604                         .mode           = 0644,
5605                         .proc_handler   = proc_dointvec,
5606                 },
5607                 {
5608                         .procname       = "max_addresses",
5609                         .data           = &ipv6_devconf.max_addresses,
5610                         .maxlen         = sizeof(int),
5611                         .mode           = 0644,
5612                         .proc_handler   = proc_dointvec,
5613                 },
5614                 {
5615                         .procname       = "accept_ra_defrtr",
5616                         .data           = &ipv6_devconf.accept_ra_defrtr,
5617                         .maxlen         = sizeof(int),
5618                         .mode           = 0644,
5619                         .proc_handler   = proc_dointvec,
5620                 },
5621                 {
5622                         .procname       = "accept_ra_min_hop_limit",
5623                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
5624                         .maxlen         = sizeof(int),
5625                         .mode           = 0644,
5626                         .proc_handler   = proc_dointvec,
5627                 },
5628                 {
5629                         .procname       = "accept_ra_pinfo",
5630                         .data           = &ipv6_devconf.accept_ra_pinfo,
5631                         .maxlen         = sizeof(int),
5632                         .mode           = 0644,
5633                         .proc_handler   = proc_dointvec,
5634                 },
5635 #ifdef CONFIG_IPV6_ROUTER_PREF
5636                 {
5637                         .procname       = "accept_ra_rtr_pref",
5638                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
5639                         .maxlen         = sizeof(int),
5640                         .mode           = 0644,
5641                         .proc_handler   = proc_dointvec,
5642                 },
5643                 {
5644                         .procname       = "router_probe_interval",
5645                         .data           = &ipv6_devconf.rtr_probe_interval,
5646                         .maxlen         = sizeof(int),
5647                         .mode           = 0644,
5648                         .proc_handler   = proc_dointvec_jiffies,
5649                 },
5650 #ifdef CONFIG_IPV6_ROUTE_INFO
5651                 {
5652                         .procname       = "accept_ra_rt_info_max_plen",
5653                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
5654                         .maxlen         = sizeof(int),
5655                         .mode           = 0644,
5656                         .proc_handler   = proc_dointvec,
5657                 },
5658 #endif
5659 #endif
5660                 {
5661                         .procname       = "accept_ra_rt_table",
5662                         .data           = &ipv6_devconf.accept_ra_rt_table,
5663                         .maxlen         = sizeof(int),
5664                         .mode           = 0644,
5665                         .proc_handler   = proc_dointvec,
5666                 },
5667                 {
5668                         .procname       = "proxy_ndp",
5669                         .data           = &ipv6_devconf.proxy_ndp,
5670                         .maxlen         = sizeof(int),
5671                         .mode           = 0644,
5672                         .proc_handler   = addrconf_sysctl_proxy_ndp,
5673                 },
5674                 {
5675                         .procname       = "accept_source_route",
5676                         .data           = &ipv6_devconf.accept_source_route,
5677                         .maxlen         = sizeof(int),
5678                         .mode           = 0644,
5679                         .proc_handler   = proc_dointvec,
5680                 },
5681 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5682                 {
5683                         .procname       = "optimistic_dad",
5684                         .data           = &ipv6_devconf.optimistic_dad,
5685                         .maxlen         = sizeof(int),
5686                         .mode           = 0644,
5687                         .proc_handler   = proc_dointvec,
5688
5689                 },
5690                 {
5691                         .procname       = "use_optimistic",
5692                         .data           = &ipv6_devconf.use_optimistic,
5693                         .maxlen         = sizeof(int),
5694                         .mode           = 0644,
5695                         .proc_handler   = proc_dointvec,
5696
5697                 },
5698 #endif
5699 #ifdef CONFIG_IPV6_MROUTE
5700                 {
5701                         .procname       = "mc_forwarding",
5702                         .data           = &ipv6_devconf.mc_forwarding,
5703                         .maxlen         = sizeof(int),
5704                         .mode           = 0444,
5705                         .proc_handler   = proc_dointvec,
5706                 },
5707 #endif
5708                 {
5709                         .procname       = "disable_ipv6",
5710                         .data           = &ipv6_devconf.disable_ipv6,
5711                         .maxlen         = sizeof(int),
5712                         .mode           = 0644,
5713                         .proc_handler   = addrconf_sysctl_disable,
5714                 },
5715                 {
5716                         .procname       = "accept_dad",
5717                         .data           = &ipv6_devconf.accept_dad,
5718                         .maxlen         = sizeof(int),
5719                         .mode           = 0644,
5720                         .proc_handler   = proc_dointvec,
5721                 },
5722                 {
5723                         .procname       = "force_tllao",
5724                         .data           = &ipv6_devconf.force_tllao,
5725                         .maxlen         = sizeof(int),
5726                         .mode           = 0644,
5727                         .proc_handler   = proc_dointvec
5728                 },
5729                 {
5730                         .procname       = "ndisc_notify",
5731                         .data           = &ipv6_devconf.ndisc_notify,
5732                         .maxlen         = sizeof(int),
5733                         .mode           = 0644,
5734                         .proc_handler   = proc_dointvec
5735                 },
5736                 {
5737                         .procname       = "suppress_frag_ndisc",
5738                         .data           = &ipv6_devconf.suppress_frag_ndisc,
5739                         .maxlen         = sizeof(int),
5740                         .mode           = 0644,
5741                         .proc_handler   = proc_dointvec
5742                 },
5743                 {
5744                         .procname       = "accept_ra_from_local",
5745                         .data           = &ipv6_devconf.accept_ra_from_local,
5746                         .maxlen         = sizeof(int),
5747                         .mode           = 0644,
5748                         .proc_handler   = proc_dointvec,
5749                 },
5750                 {
5751                         .procname       = "accept_ra_mtu",
5752                         .data           = &ipv6_devconf.accept_ra_mtu,
5753                         .maxlen         = sizeof(int),
5754                         .mode           = 0644,
5755                         .proc_handler   = proc_dointvec,
5756                 },
5757                 {
5758                         .procname       = "stable_secret",
5759                         .data           = &ipv6_devconf.stable_secret,
5760                         .maxlen         = IPV6_MAX_STRLEN,
5761                         .mode           = 0600,
5762                         .proc_handler   = addrconf_sysctl_stable_secret,
5763                 },
5764                 {
5765                         .procname       = "use_oif_addrs_only",
5766                         .data           = &ipv6_devconf.use_oif_addrs_only,
5767                         .maxlen         = sizeof(int),
5768                         .mode           = 0644,
5769                         .proc_handler   = proc_dointvec,
5770                 },
5771                 {
5772                         .procname       = "ignore_routes_with_linkdown",
5773                         .data           = &ipv6_devconf.ignore_routes_with_linkdown,
5774                         .maxlen         = sizeof(int),
5775                         .mode           = 0644,
5776                         .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
5777                 },
5778                 {
5779                         /* sentinel */
5780                 }
5781         },
5782 };
5783
5784 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5785                 struct inet6_dev *idev, struct ipv6_devconf *p)
5786 {
5787         int i;
5788         struct addrconf_sysctl_table *t;
5789         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5790
5791         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5792         if (!t)
5793                 goto out;
5794
5795         for (i = 0; t->addrconf_vars[i].data; i++) {
5796                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5797                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5798                 t->addrconf_vars[i].extra2 = net;
5799         }
5800
5801         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5802
5803         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5804         if (!t->sysctl_header)
5805                 goto free;
5806
5807         p->sysctl = t;
5808         return 0;
5809
5810 free:
5811         kfree(t);
5812 out:
5813         return -ENOBUFS;
5814 }
5815
5816 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5817 {
5818         struct addrconf_sysctl_table *t;
5819
5820         if (!p->sysctl)
5821                 return;
5822
5823         t = p->sysctl;
5824         p->sysctl = NULL;
5825         unregister_net_sysctl_table(t->sysctl_header);
5826         kfree(t);
5827 }
5828
5829 static int addrconf_sysctl_register(struct inet6_dev *idev)
5830 {
5831         int err;
5832
5833         if (!sysctl_dev_name_is_allowed(idev->dev->name))
5834                 return -EINVAL;
5835
5836         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5837                                     &ndisc_ifinfo_sysctl_change);
5838         if (err)
5839                 return err;
5840         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5841                                          idev, &idev->cnf);
5842         if (err)
5843                 neigh_sysctl_unregister(idev->nd_parms);
5844
5845         return err;
5846 }
5847
5848 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5849 {
5850         __addrconf_sysctl_unregister(&idev->cnf);
5851         neigh_sysctl_unregister(idev->nd_parms);
5852 }
5853
5854
5855 #endif
5856
5857 static int __net_init addrconf_init_net(struct net *net)
5858 {
5859         int err = -ENOMEM;
5860         struct ipv6_devconf *all, *dflt;
5861
5862         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5863         if (!all)
5864                 goto err_alloc_all;
5865
5866         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5867         if (!dflt)
5868                 goto err_alloc_dflt;
5869
5870         /* these will be inherited by all namespaces */
5871         dflt->autoconf = ipv6_defaults.autoconf;
5872         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5873
5874         dflt->stable_secret.initialized = false;
5875         all->stable_secret.initialized = false;
5876
5877         net->ipv6.devconf_all = all;
5878         net->ipv6.devconf_dflt = dflt;
5879
5880 #ifdef CONFIG_SYSCTL
5881         err = __addrconf_sysctl_register(net, "all", NULL, all);
5882         if (err < 0)
5883                 goto err_reg_all;
5884
5885         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5886         if (err < 0)
5887                 goto err_reg_dflt;
5888 #endif
5889         return 0;
5890
5891 #ifdef CONFIG_SYSCTL
5892 err_reg_dflt:
5893         __addrconf_sysctl_unregister(all);
5894 err_reg_all:
5895         kfree(dflt);
5896 #endif
5897 err_alloc_dflt:
5898         kfree(all);
5899 err_alloc_all:
5900         return err;
5901 }
5902
5903 static void __net_exit addrconf_exit_net(struct net *net)
5904 {
5905 #ifdef CONFIG_SYSCTL
5906         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5907         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5908 #endif
5909         kfree(net->ipv6.devconf_dflt);
5910         kfree(net->ipv6.devconf_all);
5911 }
5912
5913 static struct pernet_operations addrconf_ops = {
5914         .init = addrconf_init_net,
5915         .exit = addrconf_exit_net,
5916 };
5917
5918 static struct rtnl_af_ops inet6_ops __read_mostly = {
5919         .family           = AF_INET6,
5920         .fill_link_af     = inet6_fill_link_af,
5921         .get_link_af_size = inet6_get_link_af_size,
5922         .validate_link_af = inet6_validate_link_af,
5923         .set_link_af      = inet6_set_link_af,
5924 };
5925
5926 /*
5927  *      Init / cleanup code
5928  */
5929
5930 int __init addrconf_init(void)
5931 {
5932         struct inet6_dev *idev;
5933         int i, err;
5934
5935         err = ipv6_addr_label_init();
5936         if (err < 0) {
5937                 pr_crit("%s: cannot initialize default policy table: %d\n",
5938                         __func__, err);
5939                 goto out;
5940         }
5941
5942         err = register_pernet_subsys(&addrconf_ops);
5943         if (err < 0)
5944                 goto out_addrlabel;
5945
5946         addrconf_wq = create_workqueue("ipv6_addrconf");
5947         if (!addrconf_wq) {
5948                 err = -ENOMEM;
5949                 goto out_nowq;
5950         }
5951
5952         /* The addrconf netdev notifier requires that loopback_dev
5953          * has it's ipv6 private information allocated and setup
5954          * before it can bring up and give link-local addresses
5955          * to other devices which are up.
5956          *
5957          * Unfortunately, loopback_dev is not necessarily the first
5958          * entry in the global dev_base list of net devices.  In fact,
5959          * it is likely to be the very last entry on that list.
5960          * So this causes the notifier registry below to try and
5961          * give link-local addresses to all devices besides loopback_dev
5962          * first, then loopback_dev, which cases all the non-loopback_dev
5963          * devices to fail to get a link-local address.
5964          *
5965          * So, as a temporary fix, allocate the ipv6 structure for
5966          * loopback_dev first by hand.
5967          * Longer term, all of the dependencies ipv6 has upon the loopback
5968          * device and it being up should be removed.
5969          */
5970         rtnl_lock();
5971         idev = ipv6_add_dev(init_net.loopback_dev);
5972         rtnl_unlock();
5973         if (IS_ERR(idev)) {
5974                 err = PTR_ERR(idev);
5975                 goto errlo;
5976         }
5977
5978         for (i = 0; i < IN6_ADDR_HSIZE; i++)
5979                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5980
5981         register_netdevice_notifier(&ipv6_dev_notf);
5982
5983         addrconf_verify();
5984
5985         rtnl_af_register(&inet6_ops);
5986
5987         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5988                               NULL);
5989         if (err < 0)
5990                 goto errout;
5991
5992         /* Only the first call to __rtnl_register can fail */
5993         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5994         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5995         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5996                         inet6_dump_ifaddr, NULL);
5997         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5998                         inet6_dump_ifmcaddr, NULL);
5999         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6000                         inet6_dump_ifacaddr, NULL);
6001         __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6002                         inet6_netconf_dump_devconf, NULL);
6003
6004         ipv6_addr_label_rtnl_register();
6005
6006         return 0;
6007 errout:
6008         rtnl_af_unregister(&inet6_ops);
6009         unregister_netdevice_notifier(&ipv6_dev_notf);
6010 errlo:
6011         destroy_workqueue(addrconf_wq);
6012 out_nowq:
6013         unregister_pernet_subsys(&addrconf_ops);
6014 out_addrlabel:
6015         ipv6_addr_label_cleanup();
6016 out:
6017         return err;
6018 }
6019
6020 void addrconf_cleanup(void)
6021 {
6022         struct net_device *dev;
6023         int i;
6024
6025         unregister_netdevice_notifier(&ipv6_dev_notf);
6026         unregister_pernet_subsys(&addrconf_ops);
6027         ipv6_addr_label_cleanup();
6028
6029         rtnl_lock();
6030
6031         __rtnl_af_unregister(&inet6_ops);
6032
6033         /* clean dev list */
6034         for_each_netdev(&init_net, dev) {
6035                 if (__in6_dev_get(dev) == NULL)
6036                         continue;
6037                 addrconf_ifdown(dev, 1);
6038         }
6039         addrconf_ifdown(init_net.loopback_dev, 2);
6040
6041         /*
6042          *      Check hash table.
6043          */
6044         spin_lock_bh(&addrconf_hash_lock);
6045         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6046                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6047         spin_unlock_bh(&addrconf_hash_lock);
6048         cancel_delayed_work(&addr_chk_work);
6049         rtnl_unlock();
6050
6051         destroy_workqueue(addrconf_wq);
6052 }