OSDN Git Service

ipv6: remove unnecessary check in addrconf_prefix_rcv_add_addr()
[uclinux-h8/linux.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/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
61 #ifdef CONFIG_SYSCTL
62 #include <linux/sysctl.h>
63 #endif
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
69
70 #include <net/net_namespace.h>
71 #include <net/sock.h>
72 #include <net/snmp.h>
73
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
76 #include <net/ipv6.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
81 #include <net/tcp.h>
82 #include <net/ip.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
92
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
96
97 #define INFINITY_LIFE_TIME      0xFFFFFFFF
98
99 #define IPV6_MAX_STRLEN \
100         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
101
102 static inline u32 cstamp_delta(unsigned long cstamp)
103 {
104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
105 }
106
107 static inline s32 rfc3315_s14_backoff_init(s32 irt)
108 {
109         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
110         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
111         do_div(tmp, 1000000);
112         return (s32)tmp;
113 }
114
115 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
116 {
117         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
118         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
119         do_div(tmp, 1000000);
120         if ((s32)tmp > mrt) {
121                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
122                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
123                 do_div(tmp, 1000000);
124         }
125         return (s32)tmp;
126 }
127
128 #ifdef CONFIG_SYSCTL
129 static int addrconf_sysctl_register(struct inet6_dev *idev);
130 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
131 #else
132 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
133 {
134         return 0;
135 }
136
137 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
138 {
139 }
140 #endif
141
142 static void ipv6_regen_rndid(struct inet6_dev *idev);
143 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
144
145 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
146 static int ipv6_count_addresses(const struct inet6_dev *idev);
147 static int ipv6_generate_stable_address(struct in6_addr *addr,
148                                         u8 dad_count,
149                                         const struct inet6_dev *idev);
150
151 #define IN6_ADDR_HSIZE_SHIFT    8
152 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
153 /*
154  *      Configured unicast address hash table
155  */
156 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
157 static DEFINE_SPINLOCK(addrconf_hash_lock);
158
159 static void addrconf_verify(void);
160 static void addrconf_verify_rtnl(void);
161 static void addrconf_verify_work(struct work_struct *);
162
163 static struct workqueue_struct *addrconf_wq;
164 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
165
166 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
167 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
168
169 static void addrconf_type_change(struct net_device *dev,
170                                  unsigned long event);
171 static int addrconf_ifdown(struct net_device *dev, int how);
172
173 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
174                                                   int plen,
175                                                   const struct net_device *dev,
176                                                   u32 flags, u32 noflags);
177
178 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
179 static void addrconf_dad_work(struct work_struct *w);
180 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
181                                    bool send_na);
182 static void addrconf_dad_run(struct inet6_dev *idev);
183 static void addrconf_rs_timer(struct timer_list *t);
184 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
185 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
186
187 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
188                                 struct prefix_info *pinfo);
189
190 static struct ipv6_devconf ipv6_devconf __read_mostly = {
191         .forwarding             = 0,
192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
193         .mtu6                   = IPV6_MIN_MTU,
194         .accept_ra              = 1,
195         .accept_redirects       = 1,
196         .autoconf               = 1,
197         .force_mld_version      = 0,
198         .mldv1_unsolicited_report_interval = 10 * HZ,
199         .mldv2_unsolicited_report_interval = HZ,
200         .dad_transmits          = 1,
201         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
202         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
203         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
205         .use_tempaddr           = 0,
206         .temp_valid_lft         = TEMP_VALID_LIFETIME,
207         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
208         .regen_max_retry        = REGEN_MAX_RETRY,
209         .max_desync_factor      = MAX_DESYNC_FACTOR,
210         .max_addresses          = IPV6_MAX_ADDRESSES,
211         .accept_ra_defrtr       = 1,
212         .accept_ra_from_local   = 0,
213         .accept_ra_min_hop_limit= 1,
214         .accept_ra_pinfo        = 1,
215 #ifdef CONFIG_IPV6_ROUTER_PREF
216         .accept_ra_rtr_pref     = 1,
217         .rtr_probe_interval     = 60 * HZ,
218 #ifdef CONFIG_IPV6_ROUTE_INFO
219         .accept_ra_rt_info_min_plen = 0,
220         .accept_ra_rt_info_max_plen = 0,
221 #endif
222 #endif
223         .proxy_ndp              = 0,
224         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
225         .disable_ipv6           = 0,
226         .accept_dad             = 0,
227         .suppress_frag_ndisc    = 1,
228         .accept_ra_mtu          = 1,
229         .stable_secret          = {
230                 .initialized = false,
231         },
232         .use_oif_addrs_only     = 0,
233         .ignore_routes_with_linkdown = 0,
234         .keep_addr_on_down      = 0,
235         .seg6_enabled           = 0,
236 #ifdef CONFIG_IPV6_SEG6_HMAC
237         .seg6_require_hmac      = 0,
238 #endif
239         .enhanced_dad           = 1,
240         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
241         .disable_policy         = 0,
242 };
243
244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
245         .forwarding             = 0,
246         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
247         .mtu6                   = IPV6_MIN_MTU,
248         .accept_ra              = 1,
249         .accept_redirects       = 1,
250         .autoconf               = 1,
251         .force_mld_version      = 0,
252         .mldv1_unsolicited_report_interval = 10 * HZ,
253         .mldv2_unsolicited_report_interval = HZ,
254         .dad_transmits          = 1,
255         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
256         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
257         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
258         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
259         .use_tempaddr           = 0,
260         .temp_valid_lft         = TEMP_VALID_LIFETIME,
261         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
262         .regen_max_retry        = REGEN_MAX_RETRY,
263         .max_desync_factor      = MAX_DESYNC_FACTOR,
264         .max_addresses          = IPV6_MAX_ADDRESSES,
265         .accept_ra_defrtr       = 1,
266         .accept_ra_from_local   = 0,
267         .accept_ra_min_hop_limit= 1,
268         .accept_ra_pinfo        = 1,
269 #ifdef CONFIG_IPV6_ROUTER_PREF
270         .accept_ra_rtr_pref     = 1,
271         .rtr_probe_interval     = 60 * HZ,
272 #ifdef CONFIG_IPV6_ROUTE_INFO
273         .accept_ra_rt_info_min_plen = 0,
274         .accept_ra_rt_info_max_plen = 0,
275 #endif
276 #endif
277         .proxy_ndp              = 0,
278         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
279         .disable_ipv6           = 0,
280         .accept_dad             = 1,
281         .suppress_frag_ndisc    = 1,
282         .accept_ra_mtu          = 1,
283         .stable_secret          = {
284                 .initialized = false,
285         },
286         .use_oif_addrs_only     = 0,
287         .ignore_routes_with_linkdown = 0,
288         .keep_addr_on_down      = 0,
289         .seg6_enabled           = 0,
290 #ifdef CONFIG_IPV6_SEG6_HMAC
291         .seg6_require_hmac      = 0,
292 #endif
293         .enhanced_dad           = 1,
294         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
295         .disable_policy         = 0,
296 };
297
298 /* Check if link is ready: is it up and is a valid qdisc available */
299 static inline bool addrconf_link_ready(const struct net_device *dev)
300 {
301         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
302 }
303
304 static void addrconf_del_rs_timer(struct inet6_dev *idev)
305 {
306         if (del_timer(&idev->rs_timer))
307                 __in6_dev_put(idev);
308 }
309
310 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
311 {
312         if (cancel_delayed_work(&ifp->dad_work))
313                 __in6_ifa_put(ifp);
314 }
315
316 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
317                                   unsigned long when)
318 {
319         if (!timer_pending(&idev->rs_timer))
320                 in6_dev_hold(idev);
321         mod_timer(&idev->rs_timer, jiffies + when);
322 }
323
324 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
325                                    unsigned long delay)
326 {
327         in6_ifa_hold(ifp);
328         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
329                 in6_ifa_put(ifp);
330 }
331
332 static int snmp6_alloc_dev(struct inet6_dev *idev)
333 {
334         int i;
335
336         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
337         if (!idev->stats.ipv6)
338                 goto err_ip;
339
340         for_each_possible_cpu(i) {
341                 struct ipstats_mib *addrconf_stats;
342                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
343                 u64_stats_init(&addrconf_stats->syncp);
344         }
345
346
347         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
348                                         GFP_KERNEL);
349         if (!idev->stats.icmpv6dev)
350                 goto err_icmp;
351         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
352                                            GFP_KERNEL);
353         if (!idev->stats.icmpv6msgdev)
354                 goto err_icmpmsg;
355
356         return 0;
357
358 err_icmpmsg:
359         kfree(idev->stats.icmpv6dev);
360 err_icmp:
361         free_percpu(idev->stats.ipv6);
362 err_ip:
363         return -ENOMEM;
364 }
365
366 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
367 {
368         struct inet6_dev *ndev;
369         int err = -ENOMEM;
370
371         ASSERT_RTNL();
372
373         if (dev->mtu < IPV6_MIN_MTU)
374                 return ERR_PTR(-EINVAL);
375
376         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
377         if (!ndev)
378                 return ERR_PTR(err);
379
380         rwlock_init(&ndev->lock);
381         ndev->dev = dev;
382         INIT_LIST_HEAD(&ndev->addr_list);
383         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
384         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
385
386         if (ndev->cnf.stable_secret.initialized)
387                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
388         else
389                 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
390
391         ndev->cnf.mtu6 = dev->mtu;
392         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
393         if (!ndev->nd_parms) {
394                 kfree(ndev);
395                 return ERR_PTR(err);
396         }
397         if (ndev->cnf.forwarding)
398                 dev_disable_lro(dev);
399         /* We refer to the device */
400         dev_hold(dev);
401
402         if (snmp6_alloc_dev(ndev) < 0) {
403                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
404                            __func__);
405                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
406                 dev_put(dev);
407                 kfree(ndev);
408                 return ERR_PTR(err);
409         }
410
411         if (snmp6_register_dev(ndev) < 0) {
412                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
413                            __func__, dev->name);
414                 goto err_release;
415         }
416
417         /* One reference from device. */
418         refcount_set(&ndev->refcnt, 1);
419
420         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
421                 ndev->cnf.accept_dad = -1;
422
423 #if IS_ENABLED(CONFIG_IPV6_SIT)
424         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
425                 pr_info("%s: Disabled Multicast RS\n", dev->name);
426                 ndev->cnf.rtr_solicits = 0;
427         }
428 #endif
429
430         INIT_LIST_HEAD(&ndev->tempaddr_list);
431         ndev->desync_factor = U32_MAX;
432         if ((dev->flags&IFF_LOOPBACK) ||
433             dev->type == ARPHRD_TUNNEL ||
434             dev->type == ARPHRD_TUNNEL6 ||
435             dev->type == ARPHRD_SIT ||
436             dev->type == ARPHRD_NONE) {
437                 ndev->cnf.use_tempaddr = -1;
438         } else
439                 ipv6_regen_rndid(ndev);
440
441         ndev->token = in6addr_any;
442
443         if (netif_running(dev) && addrconf_link_ready(dev))
444                 ndev->if_flags |= IF_READY;
445
446         ipv6_mc_init_dev(ndev);
447         ndev->tstamp = jiffies;
448         err = addrconf_sysctl_register(ndev);
449         if (err) {
450                 ipv6_mc_destroy_dev(ndev);
451                 snmp6_unregister_dev(ndev);
452                 goto err_release;
453         }
454         /* protected by rtnl_lock */
455         rcu_assign_pointer(dev->ip6_ptr, ndev);
456
457         /* Join interface-local all-node multicast group */
458         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
459
460         /* Join all-node multicast group */
461         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
462
463         /* Join all-router multicast group if forwarding is set */
464         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
465                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
466
467         return ndev;
468
469 err_release:
470         neigh_parms_release(&nd_tbl, ndev->nd_parms);
471         ndev->dead = 1;
472         in6_dev_finish_destroy(ndev);
473         return ERR_PTR(err);
474 }
475
476 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
477 {
478         struct inet6_dev *idev;
479
480         ASSERT_RTNL();
481
482         idev = __in6_dev_get(dev);
483         if (!idev) {
484                 idev = ipv6_add_dev(dev);
485                 if (IS_ERR(idev))
486                         return NULL;
487         }
488
489         if (dev->flags&IFF_UP)
490                 ipv6_mc_up(idev);
491         return idev;
492 }
493
494 static int inet6_netconf_msgsize_devconf(int type)
495 {
496         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
497                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
498         bool all = false;
499
500         if (type == NETCONFA_ALL)
501                 all = true;
502
503         if (all || type == NETCONFA_FORWARDING)
504                 size += nla_total_size(4);
505 #ifdef CONFIG_IPV6_MROUTE
506         if (all || type == NETCONFA_MC_FORWARDING)
507                 size += nla_total_size(4);
508 #endif
509         if (all || type == NETCONFA_PROXY_NEIGH)
510                 size += nla_total_size(4);
511
512         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
513                 size += nla_total_size(4);
514
515         return size;
516 }
517
518 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
519                                       struct ipv6_devconf *devconf, u32 portid,
520                                       u32 seq, int event, unsigned int flags,
521                                       int type)
522 {
523         struct nlmsghdr  *nlh;
524         struct netconfmsg *ncm;
525         bool all = false;
526
527         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
528                         flags);
529         if (!nlh)
530                 return -EMSGSIZE;
531
532         if (type == NETCONFA_ALL)
533                 all = true;
534
535         ncm = nlmsg_data(nlh);
536         ncm->ncm_family = AF_INET6;
537
538         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
539                 goto nla_put_failure;
540
541         if (!devconf)
542                 goto out;
543
544         if ((all || type == NETCONFA_FORWARDING) &&
545             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
546                 goto nla_put_failure;
547 #ifdef CONFIG_IPV6_MROUTE
548         if ((all || type == NETCONFA_MC_FORWARDING) &&
549             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
550                         devconf->mc_forwarding) < 0)
551                 goto nla_put_failure;
552 #endif
553         if ((all || type == NETCONFA_PROXY_NEIGH) &&
554             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
555                 goto nla_put_failure;
556
557         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
558             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
559                         devconf->ignore_routes_with_linkdown) < 0)
560                 goto nla_put_failure;
561
562 out:
563         nlmsg_end(skb, nlh);
564         return 0;
565
566 nla_put_failure:
567         nlmsg_cancel(skb, nlh);
568         return -EMSGSIZE;
569 }
570
571 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
572                                   int ifindex, struct ipv6_devconf *devconf)
573 {
574         struct sk_buff *skb;
575         int err = -ENOBUFS;
576
577         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
578         if (!skb)
579                 goto errout;
580
581         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
582                                          event, 0, type);
583         if (err < 0) {
584                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
585                 WARN_ON(err == -EMSGSIZE);
586                 kfree_skb(skb);
587                 goto errout;
588         }
589         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
590         return;
591 errout:
592         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
593 }
594
595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
596         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
597         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
598         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
599         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
600 };
601
602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
603                                      struct nlmsghdr *nlh,
604                                      struct netlink_ext_ack *extack)
605 {
606         struct net *net = sock_net(in_skb->sk);
607         struct nlattr *tb[NETCONFA_MAX+1];
608         struct inet6_dev *in6_dev = NULL;
609         struct net_device *dev = NULL;
610         struct netconfmsg *ncm;
611         struct sk_buff *skb;
612         struct ipv6_devconf *devconf;
613         int ifindex;
614         int err;
615
616         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
617                           devconf_ipv6_policy, extack);
618         if (err < 0)
619                 return err;
620
621         if (!tb[NETCONFA_IFINDEX])
622                 return -EINVAL;
623
624         err = -EINVAL;
625         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
626         switch (ifindex) {
627         case NETCONFA_IFINDEX_ALL:
628                 devconf = net->ipv6.devconf_all;
629                 break;
630         case NETCONFA_IFINDEX_DEFAULT:
631                 devconf = net->ipv6.devconf_dflt;
632                 break;
633         default:
634                 dev = dev_get_by_index(net, ifindex);
635                 if (!dev)
636                         return -EINVAL;
637                 in6_dev = in6_dev_get(dev);
638                 if (!in6_dev)
639                         goto errout;
640                 devconf = &in6_dev->cnf;
641                 break;
642         }
643
644         err = -ENOBUFS;
645         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
646         if (!skb)
647                 goto errout;
648
649         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
650                                          NETLINK_CB(in_skb).portid,
651                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
652                                          NETCONFA_ALL);
653         if (err < 0) {
654                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
655                 WARN_ON(err == -EMSGSIZE);
656                 kfree_skb(skb);
657                 goto errout;
658         }
659         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
660 errout:
661         if (in6_dev)
662                 in6_dev_put(in6_dev);
663         if (dev)
664                 dev_put(dev);
665         return err;
666 }
667
668 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
669                                       struct netlink_callback *cb)
670 {
671         struct net *net = sock_net(skb->sk);
672         int h, s_h;
673         int idx, s_idx;
674         struct net_device *dev;
675         struct inet6_dev *idev;
676         struct hlist_head *head;
677
678         s_h = cb->args[0];
679         s_idx = idx = cb->args[1];
680
681         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
682                 idx = 0;
683                 head = &net->dev_index_head[h];
684                 rcu_read_lock();
685                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
686                           net->dev_base_seq;
687                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
688                         if (idx < s_idx)
689                                 goto cont;
690                         idev = __in6_dev_get(dev);
691                         if (!idev)
692                                 goto cont;
693
694                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
695                                                        &idev->cnf,
696                                                        NETLINK_CB(cb->skb).portid,
697                                                        cb->nlh->nlmsg_seq,
698                                                        RTM_NEWNETCONF,
699                                                        NLM_F_MULTI,
700                                                        NETCONFA_ALL) < 0) {
701                                 rcu_read_unlock();
702                                 goto done;
703                         }
704                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
705 cont:
706                         idx++;
707                 }
708                 rcu_read_unlock();
709         }
710         if (h == NETDEV_HASHENTRIES) {
711                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
712                                                net->ipv6.devconf_all,
713                                                NETLINK_CB(cb->skb).portid,
714                                                cb->nlh->nlmsg_seq,
715                                                RTM_NEWNETCONF, NLM_F_MULTI,
716                                                NETCONFA_ALL) < 0)
717                         goto done;
718                 else
719                         h++;
720         }
721         if (h == NETDEV_HASHENTRIES + 1) {
722                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
723                                                net->ipv6.devconf_dflt,
724                                                NETLINK_CB(cb->skb).portid,
725                                                cb->nlh->nlmsg_seq,
726                                                RTM_NEWNETCONF, NLM_F_MULTI,
727                                                NETCONFA_ALL) < 0)
728                         goto done;
729                 else
730                         h++;
731         }
732 done:
733         cb->args[0] = h;
734         cb->args[1] = idx;
735
736         return skb->len;
737 }
738
739 #ifdef CONFIG_SYSCTL
740 static void dev_forward_change(struct inet6_dev *idev)
741 {
742         struct net_device *dev;
743         struct inet6_ifaddr *ifa;
744
745         if (!idev)
746                 return;
747         dev = idev->dev;
748         if (idev->cnf.forwarding)
749                 dev_disable_lro(dev);
750         if (dev->flags & IFF_MULTICAST) {
751                 if (idev->cnf.forwarding) {
752                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
753                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
754                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
755                 } else {
756                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
757                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
758                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
759                 }
760         }
761
762         list_for_each_entry(ifa, &idev->addr_list, if_list) {
763                 if (ifa->flags&IFA_F_TENTATIVE)
764                         continue;
765                 if (idev->cnf.forwarding)
766                         addrconf_join_anycast(ifa);
767                 else
768                         addrconf_leave_anycast(ifa);
769         }
770         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
771                                      NETCONFA_FORWARDING,
772                                      dev->ifindex, &idev->cnf);
773 }
774
775
776 static void addrconf_forward_change(struct net *net, __s32 newf)
777 {
778         struct net_device *dev;
779         struct inet6_dev *idev;
780
781         for_each_netdev(net, dev) {
782                 idev = __in6_dev_get(dev);
783                 if (idev) {
784                         int changed = (!idev->cnf.forwarding) ^ (!newf);
785                         idev->cnf.forwarding = newf;
786                         if (changed)
787                                 dev_forward_change(idev);
788                 }
789         }
790 }
791
792 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
793 {
794         struct net *net;
795         int old;
796
797         if (!rtnl_trylock())
798                 return restart_syscall();
799
800         net = (struct net *)table->extra2;
801         old = *p;
802         *p = newf;
803
804         if (p == &net->ipv6.devconf_dflt->forwarding) {
805                 if ((!newf) ^ (!old))
806                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
807                                                      NETCONFA_FORWARDING,
808                                                      NETCONFA_IFINDEX_DEFAULT,
809                                                      net->ipv6.devconf_dflt);
810                 rtnl_unlock();
811                 return 0;
812         }
813
814         if (p == &net->ipv6.devconf_all->forwarding) {
815                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
816
817                 net->ipv6.devconf_dflt->forwarding = newf;
818                 if ((!newf) ^ (!old_dflt))
819                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
820                                                      NETCONFA_FORWARDING,
821                                                      NETCONFA_IFINDEX_DEFAULT,
822                                                      net->ipv6.devconf_dflt);
823
824                 addrconf_forward_change(net, newf);
825                 if ((!newf) ^ (!old))
826                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
827                                                      NETCONFA_FORWARDING,
828                                                      NETCONFA_IFINDEX_ALL,
829                                                      net->ipv6.devconf_all);
830         } else if ((!newf) ^ (!old))
831                 dev_forward_change((struct inet6_dev *)table->extra1);
832         rtnl_unlock();
833
834         if (newf)
835                 rt6_purge_dflt_routers(net);
836         return 1;
837 }
838
839 static void addrconf_linkdown_change(struct net *net, __s32 newf)
840 {
841         struct net_device *dev;
842         struct inet6_dev *idev;
843
844         for_each_netdev(net, dev) {
845                 idev = __in6_dev_get(dev);
846                 if (idev) {
847                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
848
849                         idev->cnf.ignore_routes_with_linkdown = newf;
850                         if (changed)
851                                 inet6_netconf_notify_devconf(dev_net(dev),
852                                                              RTM_NEWNETCONF,
853                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
854                                                              dev->ifindex,
855                                                              &idev->cnf);
856                 }
857         }
858 }
859
860 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
861 {
862         struct net *net;
863         int old;
864
865         if (!rtnl_trylock())
866                 return restart_syscall();
867
868         net = (struct net *)table->extra2;
869         old = *p;
870         *p = newf;
871
872         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
873                 if ((!newf) ^ (!old))
874                         inet6_netconf_notify_devconf(net,
875                                                      RTM_NEWNETCONF,
876                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
877                                                      NETCONFA_IFINDEX_DEFAULT,
878                                                      net->ipv6.devconf_dflt);
879                 rtnl_unlock();
880                 return 0;
881         }
882
883         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
884                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
885                 addrconf_linkdown_change(net, newf);
886                 if ((!newf) ^ (!old))
887                         inet6_netconf_notify_devconf(net,
888                                                      RTM_NEWNETCONF,
889                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
890                                                      NETCONFA_IFINDEX_ALL,
891                                                      net->ipv6.devconf_all);
892         }
893         rtnl_unlock();
894
895         return 1;
896 }
897
898 #endif
899
900 /* Nobody refers to this ifaddr, destroy it */
901 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
902 {
903         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
904
905 #ifdef NET_REFCNT_DEBUG
906         pr_debug("%s\n", __func__);
907 #endif
908
909         in6_dev_put(ifp->idev);
910
911         if (cancel_delayed_work(&ifp->dad_work))
912                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
913                           ifp);
914
915         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
916                 pr_warn("Freeing alive inet6 address %p\n", ifp);
917                 return;
918         }
919         ip6_rt_put(ifp->rt);
920
921         kfree_rcu(ifp, rcu);
922 }
923
924 static void
925 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
926 {
927         struct list_head *p;
928         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
929
930         /*
931          * Each device address list is sorted in order of scope -
932          * global before linklocal.
933          */
934         list_for_each(p, &idev->addr_list) {
935                 struct inet6_ifaddr *ifa
936                         = list_entry(p, struct inet6_ifaddr, if_list);
937                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
938                         break;
939         }
940
941         list_add_tail_rcu(&ifp->if_list, p);
942 }
943
944 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
945 {
946         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
947
948         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
949 }
950
951 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
952                                struct net_device *dev, unsigned int hash)
953 {
954         struct inet6_ifaddr *ifp;
955
956         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
957                 if (!net_eq(dev_net(ifp->idev->dev), net))
958                         continue;
959                 if (ipv6_addr_equal(&ifp->addr, addr)) {
960                         if (!dev || ifp->idev->dev == dev)
961                                 return true;
962                 }
963         }
964         return false;
965 }
966
967 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
968 {
969         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
970         int err = 0;
971
972         spin_lock(&addrconf_hash_lock);
973
974         /* Ignore adding duplicate addresses on an interface */
975         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
976                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
977                 err = -EEXIST;
978         } else {
979                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
980         }
981
982         spin_unlock(&addrconf_hash_lock);
983
984         return err;
985 }
986
987 /* On success it returns ifp with increased reference count */
988
989 static struct inet6_ifaddr *
990 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
991               const struct in6_addr *peer_addr, int pfxlen,
992               int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
993               bool can_block, struct netlink_ext_ack *extack)
994 {
995         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
996         struct net *net = dev_net(idev->dev);
997         struct inet6_ifaddr *ifa = NULL;
998         struct rt6_info *rt = NULL;
999         int err = 0;
1000         int addr_type = ipv6_addr_type(addr);
1001
1002         if (addr_type == IPV6_ADDR_ANY ||
1003             addr_type & IPV6_ADDR_MULTICAST ||
1004             (!(idev->dev->flags & IFF_LOOPBACK) &&
1005              addr_type & IPV6_ADDR_LOOPBACK))
1006                 return ERR_PTR(-EADDRNOTAVAIL);
1007
1008         if (idev->dead) {
1009                 err = -ENODEV;                  /*XXX*/
1010                 goto out;
1011         }
1012
1013         if (idev->cnf.disable_ipv6) {
1014                 err = -EACCES;
1015                 goto out;
1016         }
1017
1018         /* validator notifier needs to be blocking;
1019          * do not call in atomic context
1020          */
1021         if (can_block) {
1022                 struct in6_validator_info i6vi = {
1023                         .i6vi_addr = *addr,
1024                         .i6vi_dev = idev,
1025                         .extack = extack,
1026                 };
1027
1028                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1029                 err = notifier_to_errno(err);
1030                 if (err < 0)
1031                         goto out;
1032         }
1033
1034         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1035         if (!ifa) {
1036                 err = -ENOBUFS;
1037                 goto out;
1038         }
1039
1040         rt = addrconf_dst_alloc(idev, addr, false);
1041         if (IS_ERR(rt)) {
1042                 err = PTR_ERR(rt);
1043                 rt = NULL;
1044                 goto out;
1045         }
1046
1047         if (net->ipv6.devconf_all->disable_policy ||
1048             idev->cnf.disable_policy)
1049                 rt->dst.flags |= DST_NOPOLICY;
1050
1051         neigh_parms_data_state_setall(idev->nd_parms);
1052
1053         ifa->addr = *addr;
1054         if (peer_addr)
1055                 ifa->peer_addr = *peer_addr;
1056
1057         spin_lock_init(&ifa->lock);
1058         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1059         INIT_HLIST_NODE(&ifa->addr_lst);
1060         ifa->scope = scope;
1061         ifa->prefix_len = pfxlen;
1062         ifa->flags = flags;
1063         /* No need to add the TENTATIVE flag for addresses with NODAD */
1064         if (!(flags & IFA_F_NODAD))
1065                 ifa->flags |= IFA_F_TENTATIVE;
1066         ifa->valid_lft = valid_lft;
1067         ifa->prefered_lft = prefered_lft;
1068         ifa->cstamp = ifa->tstamp = jiffies;
1069         ifa->tokenized = false;
1070
1071         ifa->rt = rt;
1072
1073         ifa->idev = idev;
1074         in6_dev_hold(idev);
1075
1076         /* For caller */
1077         refcount_set(&ifa->refcnt, 1);
1078
1079         rcu_read_lock_bh();
1080
1081         err = ipv6_add_addr_hash(idev->dev, ifa);
1082         if (err < 0) {
1083                 rcu_read_unlock_bh();
1084                 goto out;
1085         }
1086
1087         write_lock(&idev->lock);
1088
1089         /* Add to inet6_dev unicast addr list. */
1090         ipv6_link_dev_addr(idev, ifa);
1091
1092         if (ifa->flags&IFA_F_TEMPORARY) {
1093                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1094                 in6_ifa_hold(ifa);
1095         }
1096
1097         in6_ifa_hold(ifa);
1098         write_unlock(&idev->lock);
1099
1100         rcu_read_unlock_bh();
1101
1102         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1103 out:
1104         if (unlikely(err < 0)) {
1105                 if (rt)
1106                         ip6_rt_put(rt);
1107                 if (ifa) {
1108                         if (ifa->idev)
1109                                 in6_dev_put(ifa->idev);
1110                         kfree(ifa);
1111                 }
1112                 ifa = ERR_PTR(err);
1113         }
1114
1115         return ifa;
1116 }
1117
1118 enum cleanup_prefix_rt_t {
1119         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1120         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1121         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1122 };
1123
1124 /*
1125  * Check, whether the prefix for ifp would still need a prefix route
1126  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1127  * constants.
1128  *
1129  * 1) we don't purge prefix if address was not permanent.
1130  *    prefix is managed by its own lifetime.
1131  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1132  * 3) if there are no addresses, delete prefix.
1133  * 4) if there are still other permanent address(es),
1134  *    corresponding prefix is still permanent.
1135  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1136  *    don't purge the prefix, assume user space is managing it.
1137  * 6) otherwise, update prefix lifetime to the
1138  *    longest valid lifetime among the corresponding
1139  *    addresses on the device.
1140  *    Note: subsequent RA will update lifetime.
1141  **/
1142 static enum cleanup_prefix_rt_t
1143 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1144 {
1145         struct inet6_ifaddr *ifa;
1146         struct inet6_dev *idev = ifp->idev;
1147         unsigned long lifetime;
1148         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1149
1150         *expires = jiffies;
1151
1152         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1153                 if (ifa == ifp)
1154                         continue;
1155                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1156                                        ifp->prefix_len))
1157                         continue;
1158                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1159                         return CLEANUP_PREFIX_RT_NOP;
1160
1161                 action = CLEANUP_PREFIX_RT_EXPIRE;
1162
1163                 spin_lock(&ifa->lock);
1164
1165                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1166                 /*
1167                  * Note: Because this address is
1168                  * not permanent, lifetime <
1169                  * LONG_MAX / HZ here.
1170                  */
1171                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1172                         *expires = ifa->tstamp + lifetime * HZ;
1173                 spin_unlock(&ifa->lock);
1174         }
1175
1176         return action;
1177 }
1178
1179 static void
1180 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1181 {
1182         struct rt6_info *rt;
1183
1184         rt = addrconf_get_prefix_route(&ifp->addr,
1185                                        ifp->prefix_len,
1186                                        ifp->idev->dev,
1187                                        0, RTF_GATEWAY | RTF_DEFAULT);
1188         if (rt) {
1189                 if (del_rt)
1190                         ip6_del_rt(rt);
1191                 else {
1192                         if (!(rt->rt6i_flags & RTF_EXPIRES))
1193                                 rt6_set_expires(rt, expires);
1194                         ip6_rt_put(rt);
1195                 }
1196         }
1197 }
1198
1199
1200 /* This function wants to get referenced ifp and releases it before return */
1201
1202 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1203 {
1204         int state;
1205         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1206         unsigned long expires;
1207
1208         ASSERT_RTNL();
1209
1210         spin_lock_bh(&ifp->lock);
1211         state = ifp->state;
1212         ifp->state = INET6_IFADDR_STATE_DEAD;
1213         spin_unlock_bh(&ifp->lock);
1214
1215         if (state == INET6_IFADDR_STATE_DEAD)
1216                 goto out;
1217
1218         spin_lock_bh(&addrconf_hash_lock);
1219         hlist_del_init_rcu(&ifp->addr_lst);
1220         spin_unlock_bh(&addrconf_hash_lock);
1221
1222         write_lock_bh(&ifp->idev->lock);
1223
1224         if (ifp->flags&IFA_F_TEMPORARY) {
1225                 list_del(&ifp->tmp_list);
1226                 if (ifp->ifpub) {
1227                         in6_ifa_put(ifp->ifpub);
1228                         ifp->ifpub = NULL;
1229                 }
1230                 __in6_ifa_put(ifp);
1231         }
1232
1233         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1234                 action = check_cleanup_prefix_route(ifp, &expires);
1235
1236         list_del_rcu(&ifp->if_list);
1237         __in6_ifa_put(ifp);
1238
1239         write_unlock_bh(&ifp->idev->lock);
1240
1241         addrconf_del_dad_work(ifp);
1242
1243         ipv6_ifa_notify(RTM_DELADDR, ifp);
1244
1245         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1246
1247         if (action != CLEANUP_PREFIX_RT_NOP) {
1248                 cleanup_prefix_route(ifp, expires,
1249                         action == CLEANUP_PREFIX_RT_DEL);
1250         }
1251
1252         /* clean up prefsrc entries */
1253         rt6_remove_prefsrc(ifp);
1254 out:
1255         in6_ifa_put(ifp);
1256 }
1257
1258 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1259                                 struct inet6_ifaddr *ift,
1260                                 bool block)
1261 {
1262         struct inet6_dev *idev = ifp->idev;
1263         struct in6_addr addr, *tmpaddr;
1264         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1265         unsigned long regen_advance;
1266         int tmp_plen;
1267         int ret = 0;
1268         u32 addr_flags;
1269         unsigned long now = jiffies;
1270         long max_desync_factor;
1271         s32 cnf_temp_preferred_lft;
1272
1273         write_lock_bh(&idev->lock);
1274         if (ift) {
1275                 spin_lock_bh(&ift->lock);
1276                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1277                 spin_unlock_bh(&ift->lock);
1278                 tmpaddr = &addr;
1279         } else {
1280                 tmpaddr = NULL;
1281         }
1282 retry:
1283         in6_dev_hold(idev);
1284         if (idev->cnf.use_tempaddr <= 0) {
1285                 write_unlock_bh(&idev->lock);
1286                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1287                 in6_dev_put(idev);
1288                 ret = -1;
1289                 goto out;
1290         }
1291         spin_lock_bh(&ifp->lock);
1292         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1293                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1294                 spin_unlock_bh(&ifp->lock);
1295                 write_unlock_bh(&idev->lock);
1296                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1297                         __func__);
1298                 in6_dev_put(idev);
1299                 ret = -1;
1300                 goto out;
1301         }
1302         in6_ifa_hold(ifp);
1303         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1304         ipv6_try_regen_rndid(idev, tmpaddr);
1305         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1306         age = (now - ifp->tstamp) / HZ;
1307
1308         regen_advance = idev->cnf.regen_max_retry *
1309                         idev->cnf.dad_transmits *
1310                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1311
1312         /* recalculate max_desync_factor each time and update
1313          * idev->desync_factor if it's larger
1314          */
1315         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1316         max_desync_factor = min_t(__u32,
1317                                   idev->cnf.max_desync_factor,
1318                                   cnf_temp_preferred_lft - regen_advance);
1319
1320         if (unlikely(idev->desync_factor > max_desync_factor)) {
1321                 if (max_desync_factor > 0) {
1322                         get_random_bytes(&idev->desync_factor,
1323                                          sizeof(idev->desync_factor));
1324                         idev->desync_factor %= max_desync_factor;
1325                 } else {
1326                         idev->desync_factor = 0;
1327                 }
1328         }
1329
1330         tmp_valid_lft = min_t(__u32,
1331                               ifp->valid_lft,
1332                               idev->cnf.temp_valid_lft + age);
1333         tmp_prefered_lft = cnf_temp_preferred_lft + age -
1334                             idev->desync_factor;
1335         tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1336         tmp_plen = ifp->prefix_len;
1337         tmp_tstamp = ifp->tstamp;
1338         spin_unlock_bh(&ifp->lock);
1339
1340         write_unlock_bh(&idev->lock);
1341
1342         /* A temporary address is created only if this calculated Preferred
1343          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1344          * an implementation must not create a temporary address with a zero
1345          * Preferred Lifetime.
1346          * Use age calculation as in addrconf_verify to avoid unnecessary
1347          * temporary addresses being generated.
1348          */
1349         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1350         if (tmp_prefered_lft <= regen_advance + age) {
1351                 in6_ifa_put(ifp);
1352                 in6_dev_put(idev);
1353                 ret = -1;
1354                 goto out;
1355         }
1356
1357         addr_flags = IFA_F_TEMPORARY;
1358         /* set in addrconf_prefix_rcv() */
1359         if (ifp->flags & IFA_F_OPTIMISTIC)
1360                 addr_flags |= IFA_F_OPTIMISTIC;
1361
1362         ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1363                             ipv6_addr_scope(&addr), addr_flags,
1364                             tmp_valid_lft, tmp_prefered_lft, block, NULL);
1365         if (IS_ERR(ift)) {
1366                 in6_ifa_put(ifp);
1367                 in6_dev_put(idev);
1368                 pr_info("%s: retry temporary address regeneration\n", __func__);
1369                 tmpaddr = &addr;
1370                 write_lock_bh(&idev->lock);
1371                 goto retry;
1372         }
1373
1374         spin_lock_bh(&ift->lock);
1375         ift->ifpub = ifp;
1376         ift->cstamp = now;
1377         ift->tstamp = tmp_tstamp;
1378         spin_unlock_bh(&ift->lock);
1379
1380         addrconf_dad_start(ift);
1381         in6_ifa_put(ift);
1382         in6_dev_put(idev);
1383 out:
1384         return ret;
1385 }
1386
1387 /*
1388  *      Choose an appropriate source address (RFC3484)
1389  */
1390 enum {
1391         IPV6_SADDR_RULE_INIT = 0,
1392         IPV6_SADDR_RULE_LOCAL,
1393         IPV6_SADDR_RULE_SCOPE,
1394         IPV6_SADDR_RULE_PREFERRED,
1395 #ifdef CONFIG_IPV6_MIP6
1396         IPV6_SADDR_RULE_HOA,
1397 #endif
1398         IPV6_SADDR_RULE_OIF,
1399         IPV6_SADDR_RULE_LABEL,
1400         IPV6_SADDR_RULE_PRIVACY,
1401         IPV6_SADDR_RULE_ORCHID,
1402         IPV6_SADDR_RULE_PREFIX,
1403 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1404         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1405 #endif
1406         IPV6_SADDR_RULE_MAX
1407 };
1408
1409 struct ipv6_saddr_score {
1410         int                     rule;
1411         int                     addr_type;
1412         struct inet6_ifaddr     *ifa;
1413         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1414         int                     scopedist;
1415         int                     matchlen;
1416 };
1417
1418 struct ipv6_saddr_dst {
1419         const struct in6_addr *addr;
1420         int ifindex;
1421         int scope;
1422         int label;
1423         unsigned int prefs;
1424 };
1425
1426 static inline int ipv6_saddr_preferred(int type)
1427 {
1428         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1429                 return 1;
1430         return 0;
1431 }
1432
1433 static bool ipv6_use_optimistic_addr(struct net *net,
1434                                      struct inet6_dev *idev)
1435 {
1436 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1437         if (!idev)
1438                 return false;
1439         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1440                 return false;
1441         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1442                 return false;
1443
1444         return true;
1445 #else
1446         return false;
1447 #endif
1448 }
1449
1450 static bool ipv6_allow_optimistic_dad(struct net *net,
1451                                       struct inet6_dev *idev)
1452 {
1453 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1454         if (!idev)
1455                 return false;
1456         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1457                 return false;
1458
1459         return true;
1460 #else
1461         return false;
1462 #endif
1463 }
1464
1465 static int ipv6_get_saddr_eval(struct net *net,
1466                                struct ipv6_saddr_score *score,
1467                                struct ipv6_saddr_dst *dst,
1468                                int i)
1469 {
1470         int ret;
1471
1472         if (i <= score->rule) {
1473                 switch (i) {
1474                 case IPV6_SADDR_RULE_SCOPE:
1475                         ret = score->scopedist;
1476                         break;
1477                 case IPV6_SADDR_RULE_PREFIX:
1478                         ret = score->matchlen;
1479                         break;
1480                 default:
1481                         ret = !!test_bit(i, score->scorebits);
1482                 }
1483                 goto out;
1484         }
1485
1486         switch (i) {
1487         case IPV6_SADDR_RULE_INIT:
1488                 /* Rule 0: remember if hiscore is not ready yet */
1489                 ret = !!score->ifa;
1490                 break;
1491         case IPV6_SADDR_RULE_LOCAL:
1492                 /* Rule 1: Prefer same address */
1493                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1494                 break;
1495         case IPV6_SADDR_RULE_SCOPE:
1496                 /* Rule 2: Prefer appropriate scope
1497                  *
1498                  *      ret
1499                  *       ^
1500                  *    -1 |  d 15
1501                  *    ---+--+-+---> scope
1502                  *       |
1503                  *       |             d is scope of the destination.
1504                  *  B-d  |  \
1505                  *       |   \      <- smaller scope is better if
1506                  *  B-15 |    \        if scope is enough for destination.
1507                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1508                  * d-C-1 | /
1509                  *       |/         <- greater is better
1510                  *   -C  /             if scope is not enough for destination.
1511                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1512                  *
1513                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1514                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1515                  * Assume B = 0 and we get C > 29.
1516                  */
1517                 ret = __ipv6_addr_src_scope(score->addr_type);
1518                 if (ret >= dst->scope)
1519                         ret = -ret;
1520                 else
1521                         ret -= 128;     /* 30 is enough */
1522                 score->scopedist = ret;
1523                 break;
1524         case IPV6_SADDR_RULE_PREFERRED:
1525             {
1526                 /* Rule 3: Avoid deprecated and optimistic addresses */
1527                 u8 avoid = IFA_F_DEPRECATED;
1528
1529                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1530                         avoid |= IFA_F_OPTIMISTIC;
1531                 ret = ipv6_saddr_preferred(score->addr_type) ||
1532                       !(score->ifa->flags & avoid);
1533                 break;
1534             }
1535 #ifdef CONFIG_IPV6_MIP6
1536         case IPV6_SADDR_RULE_HOA:
1537             {
1538                 /* Rule 4: Prefer home address */
1539                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1540                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1541                 break;
1542             }
1543 #endif
1544         case IPV6_SADDR_RULE_OIF:
1545                 /* Rule 5: Prefer outgoing interface */
1546                 ret = (!dst->ifindex ||
1547                        dst->ifindex == score->ifa->idev->dev->ifindex);
1548                 break;
1549         case IPV6_SADDR_RULE_LABEL:
1550                 /* Rule 6: Prefer matching label */
1551                 ret = ipv6_addr_label(net,
1552                                       &score->ifa->addr, score->addr_type,
1553                                       score->ifa->idev->dev->ifindex) == dst->label;
1554                 break;
1555         case IPV6_SADDR_RULE_PRIVACY:
1556             {
1557                 /* Rule 7: Prefer public address
1558                  * Note: prefer temporary address if use_tempaddr >= 2
1559                  */
1560                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1561                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1562                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1563                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1564                 break;
1565             }
1566         case IPV6_SADDR_RULE_ORCHID:
1567                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1568                  *          non-ORCHID vs non-ORCHID
1569                  */
1570                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1571                         ipv6_addr_orchid(dst->addr));
1572                 break;
1573         case IPV6_SADDR_RULE_PREFIX:
1574                 /* Rule 8: Use longest matching prefix */
1575                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1576                 if (ret > score->ifa->prefix_len)
1577                         ret = score->ifa->prefix_len;
1578                 score->matchlen = ret;
1579                 break;
1580 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1581         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1582                 /* Optimistic addresses still have lower precedence than other
1583                  * preferred addresses.
1584                  */
1585                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1586                 break;
1587 #endif
1588         default:
1589                 ret = 0;
1590         }
1591
1592         if (ret)
1593                 __set_bit(i, score->scorebits);
1594         score->rule = i;
1595 out:
1596         return ret;
1597 }
1598
1599 static int __ipv6_dev_get_saddr(struct net *net,
1600                                 struct ipv6_saddr_dst *dst,
1601                                 struct inet6_dev *idev,
1602                                 struct ipv6_saddr_score *scores,
1603                                 int hiscore_idx)
1604 {
1605         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1606
1607         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1608                 int i;
1609
1610                 /*
1611                  * - Tentative Address (RFC2462 section 5.4)
1612                  *  - A tentative address is not considered
1613                  *    "assigned to an interface" in the traditional
1614                  *    sense, unless it is also flagged as optimistic.
1615                  * - Candidate Source Address (section 4)
1616                  *  - In any case, anycast addresses, multicast
1617                  *    addresses, and the unspecified address MUST
1618                  *    NOT be included in a candidate set.
1619                  */
1620                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1621                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1622                         continue;
1623
1624                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1625
1626                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1627                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1628                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1629                                             idev->dev->name);
1630                         continue;
1631                 }
1632
1633                 score->rule = -1;
1634                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1635
1636                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1637                         int minihiscore, miniscore;
1638
1639                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1640                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1641
1642                         if (minihiscore > miniscore) {
1643                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1644                                     score->scopedist > 0) {
1645                                         /*
1646                                          * special case:
1647                                          * each remaining entry
1648                                          * has too small (not enough)
1649                                          * scope, because ifa entries
1650                                          * are sorted by their scope
1651                                          * values.
1652                                          */
1653                                         goto out;
1654                                 }
1655                                 break;
1656                         } else if (minihiscore < miniscore) {
1657                                 swap(hiscore, score);
1658                                 hiscore_idx = 1 - hiscore_idx;
1659
1660                                 /* restore our iterator */
1661                                 score->ifa = hiscore->ifa;
1662
1663                                 break;
1664                         }
1665                 }
1666         }
1667 out:
1668         return hiscore_idx;
1669 }
1670
1671 static int ipv6_get_saddr_master(struct net *net,
1672                                  const struct net_device *dst_dev,
1673                                  const struct net_device *master,
1674                                  struct ipv6_saddr_dst *dst,
1675                                  struct ipv6_saddr_score *scores,
1676                                  int hiscore_idx)
1677 {
1678         struct inet6_dev *idev;
1679
1680         idev = __in6_dev_get(dst_dev);
1681         if (idev)
1682                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1683                                                    scores, hiscore_idx);
1684
1685         idev = __in6_dev_get(master);
1686         if (idev)
1687                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1688                                                    scores, hiscore_idx);
1689
1690         return hiscore_idx;
1691 }
1692
1693 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1694                        const struct in6_addr *daddr, unsigned int prefs,
1695                        struct in6_addr *saddr)
1696 {
1697         struct ipv6_saddr_score scores[2], *hiscore;
1698         struct ipv6_saddr_dst dst;
1699         struct inet6_dev *idev;
1700         struct net_device *dev;
1701         int dst_type;
1702         bool use_oif_addr = false;
1703         int hiscore_idx = 0;
1704         int ret = 0;
1705
1706         dst_type = __ipv6_addr_type(daddr);
1707         dst.addr = daddr;
1708         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1709         dst.scope = __ipv6_addr_src_scope(dst_type);
1710         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1711         dst.prefs = prefs;
1712
1713         scores[hiscore_idx].rule = -1;
1714         scores[hiscore_idx].ifa = NULL;
1715
1716         rcu_read_lock();
1717
1718         /* Candidate Source Address (section 4)
1719          *  - multicast and link-local destination address,
1720          *    the set of candidate source address MUST only
1721          *    include addresses assigned to interfaces
1722          *    belonging to the same link as the outgoing
1723          *    interface.
1724          * (- For site-local destination addresses, the
1725          *    set of candidate source addresses MUST only
1726          *    include addresses assigned to interfaces
1727          *    belonging to the same site as the outgoing
1728          *    interface.)
1729          *  - "It is RECOMMENDED that the candidate source addresses
1730          *    be the set of unicast addresses assigned to the
1731          *    interface that will be used to send to the destination
1732          *    (the 'outgoing' interface)." (RFC 6724)
1733          */
1734         if (dst_dev) {
1735                 idev = __in6_dev_get(dst_dev);
1736                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1737                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1738                     (idev && idev->cnf.use_oif_addrs_only)) {
1739                         use_oif_addr = true;
1740                 }
1741         }
1742
1743         if (use_oif_addr) {
1744                 if (idev)
1745                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1746         } else {
1747                 const struct net_device *master;
1748                 int master_idx = 0;
1749
1750                 /* if dst_dev exists and is enslaved to an L3 device, then
1751                  * prefer addresses from dst_dev and then the master over
1752                  * any other enslaved devices in the L3 domain.
1753                  */
1754                 master = l3mdev_master_dev_rcu(dst_dev);
1755                 if (master) {
1756                         master_idx = master->ifindex;
1757
1758                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1759                                                             master, &dst,
1760                                                             scores, hiscore_idx);
1761
1762                         if (scores[hiscore_idx].ifa)
1763                                 goto out;
1764                 }
1765
1766                 for_each_netdev_rcu(net, dev) {
1767                         /* only consider addresses on devices in the
1768                          * same L3 domain
1769                          */
1770                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1771                                 continue;
1772                         idev = __in6_dev_get(dev);
1773                         if (!idev)
1774                                 continue;
1775                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1776                 }
1777         }
1778
1779 out:
1780         hiscore = &scores[hiscore_idx];
1781         if (!hiscore->ifa)
1782                 ret = -EADDRNOTAVAIL;
1783         else
1784                 *saddr = hiscore->ifa->addr;
1785
1786         rcu_read_unlock();
1787         return ret;
1788 }
1789 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1790
1791 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1792                       u32 banned_flags)
1793 {
1794         struct inet6_ifaddr *ifp;
1795         int err = -EADDRNOTAVAIL;
1796
1797         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1798                 if (ifp->scope > IFA_LINK)
1799                         break;
1800                 if (ifp->scope == IFA_LINK &&
1801                     !(ifp->flags & banned_flags)) {
1802                         *addr = ifp->addr;
1803                         err = 0;
1804                         break;
1805                 }
1806         }
1807         return err;
1808 }
1809
1810 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1811                     u32 banned_flags)
1812 {
1813         struct inet6_dev *idev;
1814         int err = -EADDRNOTAVAIL;
1815
1816         rcu_read_lock();
1817         idev = __in6_dev_get(dev);
1818         if (idev) {
1819                 read_lock_bh(&idev->lock);
1820                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1821                 read_unlock_bh(&idev->lock);
1822         }
1823         rcu_read_unlock();
1824         return err;
1825 }
1826
1827 static int ipv6_count_addresses(const struct inet6_dev *idev)
1828 {
1829         const struct inet6_ifaddr *ifp;
1830         int cnt = 0;
1831
1832         rcu_read_lock();
1833         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1834                 cnt++;
1835         rcu_read_unlock();
1836         return cnt;
1837 }
1838
1839 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1840                   const struct net_device *dev, int strict)
1841 {
1842         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1843                                        strict, IFA_F_TENTATIVE);
1844 }
1845 EXPORT_SYMBOL(ipv6_chk_addr);
1846
1847 /* device argument is used to find the L3 domain of interest. If
1848  * skip_dev_check is set, then the ifp device is not checked against
1849  * the passed in dev argument. So the 2 cases for addresses checks are:
1850  *   1. does the address exist in the L3 domain that dev is part of
1851  *      (skip_dev_check = true), or
1852  *
1853  *   2. does the address exist on the specific device
1854  *      (skip_dev_check = false)
1855  */
1856 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1857                             const struct net_device *dev, bool skip_dev_check,
1858                             int strict, u32 banned_flags)
1859 {
1860         unsigned int hash = inet6_addr_hash(net, addr);
1861         const struct net_device *l3mdev;
1862         struct inet6_ifaddr *ifp;
1863         u32 ifp_flags;
1864
1865         rcu_read_lock();
1866
1867         l3mdev = l3mdev_master_dev_rcu(dev);
1868         if (skip_dev_check)
1869                 dev = NULL;
1870
1871         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1872                 if (!net_eq(dev_net(ifp->idev->dev), net))
1873                         continue;
1874
1875                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1876                         continue;
1877
1878                 /* Decouple optimistic from tentative for evaluation here.
1879                  * Ban optimistic addresses explicitly, when required.
1880                  */
1881                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1882                             ? (ifp->flags&~IFA_F_TENTATIVE)
1883                             : ifp->flags;
1884                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1885                     !(ifp_flags&banned_flags) &&
1886                     (!dev || ifp->idev->dev == dev ||
1887                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1888                         rcu_read_unlock();
1889                         return 1;
1890                 }
1891         }
1892
1893         rcu_read_unlock();
1894         return 0;
1895 }
1896 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1897
1898
1899 /* Compares an address/prefix_len with addresses on device @dev.
1900  * If one is found it returns true.
1901  */
1902 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1903         const unsigned int prefix_len, struct net_device *dev)
1904 {
1905         const struct inet6_ifaddr *ifa;
1906         const struct inet6_dev *idev;
1907         bool ret = false;
1908
1909         rcu_read_lock();
1910         idev = __in6_dev_get(dev);
1911         if (idev) {
1912                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1913                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1914                         if (ret)
1915                                 break;
1916                 }
1917         }
1918         rcu_read_unlock();
1919
1920         return ret;
1921 }
1922 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1923
1924 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1925 {
1926         const struct inet6_ifaddr *ifa;
1927         const struct inet6_dev *idev;
1928         int     onlink;
1929
1930         onlink = 0;
1931         rcu_read_lock();
1932         idev = __in6_dev_get(dev);
1933         if (idev) {
1934                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1935                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1936                                                    ifa->prefix_len);
1937                         if (onlink)
1938                                 break;
1939                 }
1940         }
1941         rcu_read_unlock();
1942         return onlink;
1943 }
1944 EXPORT_SYMBOL(ipv6_chk_prefix);
1945
1946 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1947                                      struct net_device *dev, int strict)
1948 {
1949         unsigned int hash = inet6_addr_hash(net, addr);
1950         struct inet6_ifaddr *ifp, *result = NULL;
1951
1952         rcu_read_lock();
1953         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1954                 if (!net_eq(dev_net(ifp->idev->dev), net))
1955                         continue;
1956                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1957                         if (!dev || ifp->idev->dev == dev ||
1958                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1959                                 result = ifp;
1960                                 in6_ifa_hold(ifp);
1961                                 break;
1962                         }
1963                 }
1964         }
1965         rcu_read_unlock();
1966
1967         return result;
1968 }
1969
1970 /* Gets referenced address, destroys ifaddr */
1971
1972 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1973 {
1974         if (dad_failed)
1975                 ifp->flags |= IFA_F_DADFAILED;
1976
1977         if (ifp->flags&IFA_F_TEMPORARY) {
1978                 struct inet6_ifaddr *ifpub;
1979                 spin_lock_bh(&ifp->lock);
1980                 ifpub = ifp->ifpub;
1981                 if (ifpub) {
1982                         in6_ifa_hold(ifpub);
1983                         spin_unlock_bh(&ifp->lock);
1984                         ipv6_create_tempaddr(ifpub, ifp, true);
1985                         in6_ifa_put(ifpub);
1986                 } else {
1987                         spin_unlock_bh(&ifp->lock);
1988                 }
1989                 ipv6_del_addr(ifp);
1990         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1991                 spin_lock_bh(&ifp->lock);
1992                 addrconf_del_dad_work(ifp);
1993                 ifp->flags |= IFA_F_TENTATIVE;
1994                 if (dad_failed)
1995                         ifp->flags &= ~IFA_F_OPTIMISTIC;
1996                 spin_unlock_bh(&ifp->lock);
1997                 if (dad_failed)
1998                         ipv6_ifa_notify(0, ifp);
1999                 in6_ifa_put(ifp);
2000         } else {
2001                 ipv6_del_addr(ifp);
2002         }
2003 }
2004
2005 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2006 {
2007         int err = -ENOENT;
2008
2009         spin_lock_bh(&ifp->lock);
2010         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2011                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2012                 err = 0;
2013         }
2014         spin_unlock_bh(&ifp->lock);
2015
2016         return err;
2017 }
2018
2019 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2020 {
2021         struct inet6_dev *idev = ifp->idev;
2022         struct net *net = dev_net(ifp->idev->dev);
2023
2024         if (addrconf_dad_end(ifp)) {
2025                 in6_ifa_put(ifp);
2026                 return;
2027         }
2028
2029         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2030                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2031
2032         spin_lock_bh(&ifp->lock);
2033
2034         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2035                 int scope = ifp->scope;
2036                 u32 flags = ifp->flags;
2037                 struct in6_addr new_addr;
2038                 struct inet6_ifaddr *ifp2;
2039                 u32 valid_lft, preferred_lft;
2040                 int pfxlen = ifp->prefix_len;
2041                 int retries = ifp->stable_privacy_retry + 1;
2042
2043                 if (retries > net->ipv6.sysctl.idgen_retries) {
2044                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2045                                              ifp->idev->dev->name);
2046                         goto errdad;
2047                 }
2048
2049                 new_addr = ifp->addr;
2050                 if (ipv6_generate_stable_address(&new_addr, retries,
2051                                                  idev))
2052                         goto errdad;
2053
2054                 valid_lft = ifp->valid_lft;
2055                 preferred_lft = ifp->prefered_lft;
2056
2057                 spin_unlock_bh(&ifp->lock);
2058
2059                 if (idev->cnf.max_addresses &&
2060                     ipv6_count_addresses(idev) >=
2061                     idev->cnf.max_addresses)
2062                         goto lock_errdad;
2063
2064                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2065                                      ifp->idev->dev->name);
2066
2067                 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2068                                      scope, flags, valid_lft,
2069                                      preferred_lft, false, NULL);
2070                 if (IS_ERR(ifp2))
2071                         goto lock_errdad;
2072
2073                 spin_lock_bh(&ifp2->lock);
2074                 ifp2->stable_privacy_retry = retries;
2075                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2076                 spin_unlock_bh(&ifp2->lock);
2077
2078                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2079                 in6_ifa_put(ifp2);
2080 lock_errdad:
2081                 spin_lock_bh(&ifp->lock);
2082         }
2083
2084 errdad:
2085         /* transition from _POSTDAD to _ERRDAD */
2086         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2087         spin_unlock_bh(&ifp->lock);
2088
2089         addrconf_mod_dad_work(ifp, 0);
2090         in6_ifa_put(ifp);
2091 }
2092
2093 /* Join to solicited addr multicast group.
2094  * caller must hold RTNL */
2095 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2096 {
2097         struct in6_addr maddr;
2098
2099         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2100                 return;
2101
2102         addrconf_addr_solict_mult(addr, &maddr);
2103         ipv6_dev_mc_inc(dev, &maddr);
2104 }
2105
2106 /* caller must hold RTNL */
2107 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2108 {
2109         struct in6_addr maddr;
2110
2111         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2112                 return;
2113
2114         addrconf_addr_solict_mult(addr, &maddr);
2115         __ipv6_dev_mc_dec(idev, &maddr);
2116 }
2117
2118 /* caller must hold RTNL */
2119 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2120 {
2121         struct in6_addr addr;
2122
2123         if (ifp->prefix_len >= 127) /* RFC 6164 */
2124                 return;
2125         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2126         if (ipv6_addr_any(&addr))
2127                 return;
2128         __ipv6_dev_ac_inc(ifp->idev, &addr);
2129 }
2130
2131 /* caller must hold RTNL */
2132 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2133 {
2134         struct in6_addr addr;
2135
2136         if (ifp->prefix_len >= 127) /* RFC 6164 */
2137                 return;
2138         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2139         if (ipv6_addr_any(&addr))
2140                 return;
2141         __ipv6_dev_ac_dec(ifp->idev, &addr);
2142 }
2143
2144 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2145 {
2146         switch (dev->addr_len) {
2147         case ETH_ALEN:
2148                 memcpy(eui, dev->dev_addr, 3);
2149                 eui[3] = 0xFF;
2150                 eui[4] = 0xFE;
2151                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2152                 break;
2153         case EUI64_ADDR_LEN:
2154                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2155                 eui[0] ^= 2;
2156                 break;
2157         default:
2158                 return -1;
2159         }
2160
2161         return 0;
2162 }
2163
2164 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2165 {
2166         union fwnet_hwaddr *ha;
2167
2168         if (dev->addr_len != FWNET_ALEN)
2169                 return -1;
2170
2171         ha = (union fwnet_hwaddr *)dev->dev_addr;
2172
2173         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2174         eui[0] ^= 2;
2175         return 0;
2176 }
2177
2178 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2179 {
2180         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2181         if (dev->addr_len != ARCNET_ALEN)
2182                 return -1;
2183         memset(eui, 0, 7);
2184         eui[7] = *(u8 *)dev->dev_addr;
2185         return 0;
2186 }
2187
2188 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2189 {
2190         if (dev->addr_len != INFINIBAND_ALEN)
2191                 return -1;
2192         memcpy(eui, dev->dev_addr + 12, 8);
2193         eui[0] |= 2;
2194         return 0;
2195 }
2196
2197 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2198 {
2199         if (addr == 0)
2200                 return -1;
2201         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2202                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2203                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2204                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2205                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2206                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2207         eui[1] = 0;
2208         eui[2] = 0x5E;
2209         eui[3] = 0xFE;
2210         memcpy(eui + 4, &addr, 4);
2211         return 0;
2212 }
2213
2214 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2215 {
2216         if (dev->priv_flags & IFF_ISATAP)
2217                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2218         return -1;
2219 }
2220
2221 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2222 {
2223         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2224 }
2225
2226 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2227 {
2228         memcpy(eui, dev->perm_addr, 3);
2229         memcpy(eui + 5, dev->perm_addr + 3, 3);
2230         eui[3] = 0xFF;
2231         eui[4] = 0xFE;
2232         eui[0] ^= 2;
2233         return 0;
2234 }
2235
2236 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2237 {
2238         switch (dev->type) {
2239         case ARPHRD_ETHER:
2240         case ARPHRD_FDDI:
2241                 return addrconf_ifid_eui48(eui, dev);
2242         case ARPHRD_ARCNET:
2243                 return addrconf_ifid_arcnet(eui, dev);
2244         case ARPHRD_INFINIBAND:
2245                 return addrconf_ifid_infiniband(eui, dev);
2246         case ARPHRD_SIT:
2247                 return addrconf_ifid_sit(eui, dev);
2248         case ARPHRD_IPGRE:
2249         case ARPHRD_TUNNEL:
2250                 return addrconf_ifid_gre(eui, dev);
2251         case ARPHRD_6LOWPAN:
2252                 return addrconf_ifid_6lowpan(eui, dev);
2253         case ARPHRD_IEEE1394:
2254                 return addrconf_ifid_ieee1394(eui, dev);
2255         case ARPHRD_TUNNEL6:
2256         case ARPHRD_IP6GRE:
2257                 return addrconf_ifid_ip6tnl(eui, dev);
2258         }
2259         return -1;
2260 }
2261
2262 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2263 {
2264         int err = -1;
2265         struct inet6_ifaddr *ifp;
2266
2267         read_lock_bh(&idev->lock);
2268         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2269                 if (ifp->scope > IFA_LINK)
2270                         break;
2271                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2272                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2273                         err = 0;
2274                         break;
2275                 }
2276         }
2277         read_unlock_bh(&idev->lock);
2278         return err;
2279 }
2280
2281 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2282 static void ipv6_regen_rndid(struct inet6_dev *idev)
2283 {
2284 regen:
2285         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2286         idev->rndid[0] &= ~0x02;
2287
2288         /*
2289          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2290          * check if generated address is not inappropriate
2291          *
2292          *  - Reserved subnet anycast (RFC 2526)
2293          *      11111101 11....11 1xxxxxxx
2294          *  - ISATAP (RFC4214) 6.1
2295          *      00-00-5E-FE-xx-xx-xx-xx
2296          *  - value 0
2297          *  - XXX: already assigned to an address on the device
2298          */
2299         if (idev->rndid[0] == 0xfd &&
2300             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2301             (idev->rndid[7]&0x80))
2302                 goto regen;
2303         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2304                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2305                         goto regen;
2306                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2307                         goto regen;
2308         }
2309 }
2310
2311 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2312 {
2313         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2314                 ipv6_regen_rndid(idev);
2315 }
2316
2317 /*
2318  *      Add prefix route.
2319  */
2320
2321 static void
2322 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2323                       unsigned long expires, u32 flags)
2324 {
2325         struct fib6_config cfg = {
2326                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2327                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2328                 .fc_ifindex = dev->ifindex,
2329                 .fc_expires = expires,
2330                 .fc_dst_len = plen,
2331                 .fc_flags = RTF_UP | flags,
2332                 .fc_nlinfo.nl_net = dev_net(dev),
2333                 .fc_protocol = RTPROT_KERNEL,
2334         };
2335
2336         cfg.fc_dst = *pfx;
2337
2338         /* Prevent useless cloning on PtP SIT.
2339            This thing is done here expecting that the whole
2340            class of non-broadcast devices need not cloning.
2341          */
2342 #if IS_ENABLED(CONFIG_IPV6_SIT)
2343         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2344                 cfg.fc_flags |= RTF_NONEXTHOP;
2345 #endif
2346
2347         ip6_route_add(&cfg, NULL);
2348 }
2349
2350
2351 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2352                                                   int plen,
2353                                                   const struct net_device *dev,
2354                                                   u32 flags, u32 noflags)
2355 {
2356         struct fib6_node *fn;
2357         struct rt6_info *rt = NULL;
2358         struct fib6_table *table;
2359         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2360
2361         table = fib6_get_table(dev_net(dev), tb_id);
2362         if (!table)
2363                 return NULL;
2364
2365         rcu_read_lock();
2366         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2367         if (!fn)
2368                 goto out;
2369
2370         for_each_fib6_node_rt_rcu(fn) {
2371                 if (rt->dst.dev->ifindex != dev->ifindex)
2372                         continue;
2373                 if ((rt->rt6i_flags & flags) != flags)
2374                         continue;
2375                 if ((rt->rt6i_flags & noflags) != 0)
2376                         continue;
2377                 if (!dst_hold_safe(&rt->dst))
2378                         rt = NULL;
2379                 break;
2380         }
2381 out:
2382         rcu_read_unlock();
2383         return rt;
2384 }
2385
2386
2387 /* Create "default" multicast route to the interface */
2388
2389 static void addrconf_add_mroute(struct net_device *dev)
2390 {
2391         struct fib6_config cfg = {
2392                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2393                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2394                 .fc_ifindex = dev->ifindex,
2395                 .fc_dst_len = 8,
2396                 .fc_flags = RTF_UP,
2397                 .fc_nlinfo.nl_net = dev_net(dev),
2398         };
2399
2400         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2401
2402         ip6_route_add(&cfg, NULL);
2403 }
2404
2405 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2406 {
2407         struct inet6_dev *idev;
2408
2409         ASSERT_RTNL();
2410
2411         idev = ipv6_find_idev(dev);
2412         if (!idev)
2413                 return ERR_PTR(-ENOBUFS);
2414
2415         if (idev->cnf.disable_ipv6)
2416                 return ERR_PTR(-EACCES);
2417
2418         /* Add default multicast route */
2419         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2420                 addrconf_add_mroute(dev);
2421
2422         return idev;
2423 }
2424
2425 static void manage_tempaddrs(struct inet6_dev *idev,
2426                              struct inet6_ifaddr *ifp,
2427                              __u32 valid_lft, __u32 prefered_lft,
2428                              bool create, unsigned long now)
2429 {
2430         u32 flags;
2431         struct inet6_ifaddr *ift;
2432
2433         read_lock_bh(&idev->lock);
2434         /* update all temporary addresses in the list */
2435         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2436                 int age, max_valid, max_prefered;
2437
2438                 if (ifp != ift->ifpub)
2439                         continue;
2440
2441                 /* RFC 4941 section 3.3:
2442                  * If a received option will extend the lifetime of a public
2443                  * address, the lifetimes of temporary addresses should
2444                  * be extended, subject to the overall constraint that no
2445                  * temporary addresses should ever remain "valid" or "preferred"
2446                  * for a time longer than (TEMP_VALID_LIFETIME) or
2447                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2448                  */
2449                 age = (now - ift->cstamp) / HZ;
2450                 max_valid = idev->cnf.temp_valid_lft - age;
2451                 if (max_valid < 0)
2452                         max_valid = 0;
2453
2454                 max_prefered = idev->cnf.temp_prefered_lft -
2455                                idev->desync_factor - age;
2456                 if (max_prefered < 0)
2457                         max_prefered = 0;
2458
2459                 if (valid_lft > max_valid)
2460                         valid_lft = max_valid;
2461
2462                 if (prefered_lft > max_prefered)
2463                         prefered_lft = max_prefered;
2464
2465                 spin_lock(&ift->lock);
2466                 flags = ift->flags;
2467                 ift->valid_lft = valid_lft;
2468                 ift->prefered_lft = prefered_lft;
2469                 ift->tstamp = now;
2470                 if (prefered_lft > 0)
2471                         ift->flags &= ~IFA_F_DEPRECATED;
2472
2473                 spin_unlock(&ift->lock);
2474                 if (!(flags&IFA_F_TENTATIVE))
2475                         ipv6_ifa_notify(0, ift);
2476         }
2477
2478         if ((create || list_empty(&idev->tempaddr_list)) &&
2479             idev->cnf.use_tempaddr > 0) {
2480                 /* When a new public address is created as described
2481                  * in [ADDRCONF], also create a new temporary address.
2482                  * Also create a temporary address if it's enabled but
2483                  * no temporary address currently exists.
2484                  */
2485                 read_unlock_bh(&idev->lock);
2486                 ipv6_create_tempaddr(ifp, NULL, false);
2487         } else {
2488                 read_unlock_bh(&idev->lock);
2489         }
2490 }
2491
2492 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2493 {
2494         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2495                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2496 }
2497
2498 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2499                                  const struct prefix_info *pinfo,
2500                                  struct inet6_dev *in6_dev,
2501                                  const struct in6_addr *addr, int addr_type,
2502                                  u32 addr_flags, bool sllao, bool tokenized,
2503                                  __u32 valid_lft, u32 prefered_lft)
2504 {
2505         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2506         int create = 0, update_lft = 0;
2507
2508         if (!ifp && valid_lft) {
2509                 int max_addresses = in6_dev->cnf.max_addresses;
2510
2511 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2512                 if ((net->ipv6.devconf_all->optimistic_dad ||
2513                      in6_dev->cnf.optimistic_dad) &&
2514                     !net->ipv6.devconf_all->forwarding && sllao)
2515                         addr_flags |= IFA_F_OPTIMISTIC;
2516 #endif
2517
2518                 /* Do not allow to create too much of autoconfigured
2519                  * addresses; this would be too easy way to crash kernel.
2520                  */
2521                 if (!max_addresses ||
2522                     ipv6_count_addresses(in6_dev) < max_addresses)
2523                         ifp = ipv6_add_addr(in6_dev, addr, NULL,
2524                                             pinfo->prefix_len,
2525                                             addr_type&IPV6_ADDR_SCOPE_MASK,
2526                                             addr_flags, valid_lft,
2527                                             prefered_lft, false, NULL);
2528
2529                 if (IS_ERR_OR_NULL(ifp))
2530                         return -1;
2531
2532                 create = 1;
2533                 spin_lock_bh(&ifp->lock);
2534                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2535                 ifp->cstamp = jiffies;
2536                 ifp->tokenized = tokenized;
2537                 spin_unlock_bh(&ifp->lock);
2538                 addrconf_dad_start(ifp);
2539         }
2540
2541         if (ifp) {
2542                 u32 flags;
2543                 unsigned long now;
2544                 u32 stored_lft;
2545
2546                 /* update lifetime (RFC2462 5.5.3 e) */
2547                 spin_lock_bh(&ifp->lock);
2548                 now = jiffies;
2549                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2550                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2551                 else
2552                         stored_lft = 0;
2553                 if (!create && stored_lft) {
2554                         const u32 minimum_lft = min_t(u32,
2555                                 stored_lft, MIN_VALID_LIFETIME);
2556                         valid_lft = max(valid_lft, minimum_lft);
2557
2558                         /* RFC4862 Section 5.5.3e:
2559                          * "Note that the preferred lifetime of the
2560                          *  corresponding address is always reset to
2561                          *  the Preferred Lifetime in the received
2562                          *  Prefix Information option, regardless of
2563                          *  whether the valid lifetime is also reset or
2564                          *  ignored."
2565                          *
2566                          * So we should always update prefered_lft here.
2567                          */
2568                         update_lft = 1;
2569                 }
2570
2571                 if (update_lft) {
2572                         ifp->valid_lft = valid_lft;
2573                         ifp->prefered_lft = prefered_lft;
2574                         ifp->tstamp = now;
2575                         flags = ifp->flags;
2576                         ifp->flags &= ~IFA_F_DEPRECATED;
2577                         spin_unlock_bh(&ifp->lock);
2578
2579                         if (!(flags&IFA_F_TENTATIVE))
2580                                 ipv6_ifa_notify(0, ifp);
2581                 } else
2582                         spin_unlock_bh(&ifp->lock);
2583
2584                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2585                                  create, now);
2586
2587                 in6_ifa_put(ifp);
2588                 addrconf_verify();
2589         }
2590
2591         return 0;
2592 }
2593 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2594
2595 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2596 {
2597         struct prefix_info *pinfo;
2598         __u32 valid_lft;
2599         __u32 prefered_lft;
2600         int addr_type, err;
2601         u32 addr_flags = 0;
2602         struct inet6_dev *in6_dev;
2603         struct net *net = dev_net(dev);
2604
2605         pinfo = (struct prefix_info *) opt;
2606
2607         if (len < sizeof(struct prefix_info)) {
2608                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2609                 return;
2610         }
2611
2612         /*
2613          *      Validation checks ([ADDRCONF], page 19)
2614          */
2615
2616         addr_type = ipv6_addr_type(&pinfo->prefix);
2617
2618         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2619                 return;
2620
2621         valid_lft = ntohl(pinfo->valid);
2622         prefered_lft = ntohl(pinfo->prefered);
2623
2624         if (prefered_lft > valid_lft) {
2625                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2626                 return;
2627         }
2628
2629         in6_dev = in6_dev_get(dev);
2630
2631         if (!in6_dev) {
2632                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2633                                     dev->name);
2634                 return;
2635         }
2636
2637         /*
2638          *      Two things going on here:
2639          *      1) Add routes for on-link prefixes
2640          *      2) Configure prefixes with the auto flag set
2641          */
2642
2643         if (pinfo->onlink) {
2644                 struct rt6_info *rt;
2645                 unsigned long rt_expires;
2646
2647                 /* Avoid arithmetic overflow. Really, we could
2648                  * save rt_expires in seconds, likely valid_lft,
2649                  * but it would require division in fib gc, that it
2650                  * not good.
2651                  */
2652                 if (HZ > USER_HZ)
2653                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2654                 else
2655                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2656
2657                 if (addrconf_finite_timeout(rt_expires))
2658                         rt_expires *= HZ;
2659
2660                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2661                                                pinfo->prefix_len,
2662                                                dev,
2663                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2664                                                RTF_GATEWAY | RTF_DEFAULT);
2665
2666                 if (rt) {
2667                         /* Autoconf prefix route */
2668                         if (valid_lft == 0) {
2669                                 ip6_del_rt(rt);
2670                                 rt = NULL;
2671                         } else if (addrconf_finite_timeout(rt_expires)) {
2672                                 /* not infinity */
2673                                 rt6_set_expires(rt, jiffies + rt_expires);
2674                         } else {
2675                                 rt6_clean_expires(rt);
2676                         }
2677                 } else if (valid_lft) {
2678                         clock_t expires = 0;
2679                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2680                         if (addrconf_finite_timeout(rt_expires)) {
2681                                 /* not infinity */
2682                                 flags |= RTF_EXPIRES;
2683                                 expires = jiffies_to_clock_t(rt_expires);
2684                         }
2685                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2686                                               dev, expires, flags);
2687                 }
2688                 ip6_rt_put(rt);
2689         }
2690
2691         /* Try to figure out our local address for this prefix */
2692
2693         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2694                 struct in6_addr addr;
2695                 bool tokenized = false, dev_addr_generated = false;
2696
2697                 if (pinfo->prefix_len == 64) {
2698                         memcpy(&addr, &pinfo->prefix, 8);
2699
2700                         if (!ipv6_addr_any(&in6_dev->token)) {
2701                                 read_lock_bh(&in6_dev->lock);
2702                                 memcpy(addr.s6_addr + 8,
2703                                        in6_dev->token.s6_addr + 8, 8);
2704                                 read_unlock_bh(&in6_dev->lock);
2705                                 tokenized = true;
2706                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2707                                    !ipv6_generate_stable_address(&addr, 0,
2708                                                                  in6_dev)) {
2709                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2710                                 goto ok;
2711                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2712                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2713                                 goto put;
2714                         } else {
2715                                 dev_addr_generated = true;
2716                         }
2717                         goto ok;
2718                 }
2719                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2720                                     pinfo->prefix_len);
2721                 goto put;
2722
2723 ok:
2724                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2725                                                    &addr, addr_type,
2726                                                    addr_flags, sllao,
2727                                                    tokenized, valid_lft,
2728                                                    prefered_lft);
2729                 if (err)
2730                         goto put;
2731
2732                 /* Ignore error case here because previous prefix add addr was
2733                  * successful which will be notified.
2734                  */
2735                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2736                                               addr_type, addr_flags, sllao,
2737                                               tokenized, valid_lft,
2738                                               prefered_lft,
2739                                               dev_addr_generated);
2740         }
2741         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2742 put:
2743         in6_dev_put(in6_dev);
2744 }
2745
2746 /*
2747  *      Set destination address.
2748  *      Special case for SIT interfaces where we create a new "virtual"
2749  *      device.
2750  */
2751 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2752 {
2753         struct in6_ifreq ireq;
2754         struct net_device *dev;
2755         int err = -EINVAL;
2756
2757         rtnl_lock();
2758
2759         err = -EFAULT;
2760         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2761                 goto err_exit;
2762
2763         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2764
2765         err = -ENODEV;
2766         if (!dev)
2767                 goto err_exit;
2768
2769 #if IS_ENABLED(CONFIG_IPV6_SIT)
2770         if (dev->type == ARPHRD_SIT) {
2771                 const struct net_device_ops *ops = dev->netdev_ops;
2772                 struct ifreq ifr;
2773                 struct ip_tunnel_parm p;
2774
2775                 err = -EADDRNOTAVAIL;
2776                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2777                         goto err_exit;
2778
2779                 memset(&p, 0, sizeof(p));
2780                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2781                 p.iph.saddr = 0;
2782                 p.iph.version = 4;
2783                 p.iph.ihl = 5;
2784                 p.iph.protocol = IPPROTO_IPV6;
2785                 p.iph.ttl = 64;
2786                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2787
2788                 if (ops->ndo_do_ioctl) {
2789                         mm_segment_t oldfs = get_fs();
2790
2791                         set_fs(KERNEL_DS);
2792                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2793                         set_fs(oldfs);
2794                 } else
2795                         err = -EOPNOTSUPP;
2796
2797                 if (err == 0) {
2798                         err = -ENOBUFS;
2799                         dev = __dev_get_by_name(net, p.name);
2800                         if (!dev)
2801                                 goto err_exit;
2802                         err = dev_open(dev);
2803                 }
2804         }
2805 #endif
2806
2807 err_exit:
2808         rtnl_unlock();
2809         return err;
2810 }
2811
2812 static int ipv6_mc_config(struct sock *sk, bool join,
2813                           const struct in6_addr *addr, int ifindex)
2814 {
2815         int ret;
2816
2817         ASSERT_RTNL();
2818
2819         lock_sock(sk);
2820         if (join)
2821                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2822         else
2823                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2824         release_sock(sk);
2825
2826         return ret;
2827 }
2828
2829 /*
2830  *      Manual configuration of address on an interface
2831  */
2832 static int inet6_addr_add(struct net *net, int ifindex,
2833                           const struct in6_addr *pfx,
2834                           const struct in6_addr *peer_pfx,
2835                           unsigned int plen, __u32 ifa_flags,
2836                           __u32 prefered_lft, __u32 valid_lft,
2837                           struct netlink_ext_ack *extack)
2838 {
2839         struct inet6_ifaddr *ifp;
2840         struct inet6_dev *idev;
2841         struct net_device *dev;
2842         unsigned long timeout;
2843         clock_t expires;
2844         int scope;
2845         u32 flags;
2846
2847         ASSERT_RTNL();
2848
2849         if (plen > 128)
2850                 return -EINVAL;
2851
2852         /* check the lifetime */
2853         if (!valid_lft || prefered_lft > valid_lft)
2854                 return -EINVAL;
2855
2856         if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2857                 return -EINVAL;
2858
2859         dev = __dev_get_by_index(net, ifindex);
2860         if (!dev)
2861                 return -ENODEV;
2862
2863         idev = addrconf_add_dev(dev);
2864         if (IS_ERR(idev))
2865                 return PTR_ERR(idev);
2866
2867         if (ifa_flags & IFA_F_MCAUTOJOIN) {
2868                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2869                                          true, pfx, ifindex);
2870
2871                 if (ret < 0)
2872                         return ret;
2873         }
2874
2875         scope = ipv6_addr_scope(pfx);
2876
2877         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2878         if (addrconf_finite_timeout(timeout)) {
2879                 expires = jiffies_to_clock_t(timeout * HZ);
2880                 valid_lft = timeout;
2881                 flags = RTF_EXPIRES;
2882         } else {
2883                 expires = 0;
2884                 flags = 0;
2885                 ifa_flags |= IFA_F_PERMANENT;
2886         }
2887
2888         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2889         if (addrconf_finite_timeout(timeout)) {
2890                 if (timeout == 0)
2891                         ifa_flags |= IFA_F_DEPRECATED;
2892                 prefered_lft = timeout;
2893         }
2894
2895         ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2896                             valid_lft, prefered_lft, true, extack);
2897
2898         if (!IS_ERR(ifp)) {
2899                 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2900                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2901                                               expires, flags);
2902                 }
2903
2904                 /*
2905                  * Note that section 3.1 of RFC 4429 indicates
2906                  * that the Optimistic flag should not be set for
2907                  * manually configured addresses
2908                  */
2909                 addrconf_dad_start(ifp);
2910                 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2911                         manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2912                                          true, jiffies);
2913                 in6_ifa_put(ifp);
2914                 addrconf_verify_rtnl();
2915                 return 0;
2916         } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2917                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2918                                false, pfx, ifindex);
2919         }
2920
2921         return PTR_ERR(ifp);
2922 }
2923
2924 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2925                           const struct in6_addr *pfx, unsigned int plen)
2926 {
2927         struct inet6_ifaddr *ifp;
2928         struct inet6_dev *idev;
2929         struct net_device *dev;
2930
2931         if (plen > 128)
2932                 return -EINVAL;
2933
2934         dev = __dev_get_by_index(net, ifindex);
2935         if (!dev)
2936                 return -ENODEV;
2937
2938         idev = __in6_dev_get(dev);
2939         if (!idev)
2940                 return -ENXIO;
2941
2942         read_lock_bh(&idev->lock);
2943         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2944                 if (ifp->prefix_len == plen &&
2945                     ipv6_addr_equal(pfx, &ifp->addr)) {
2946                         in6_ifa_hold(ifp);
2947                         read_unlock_bh(&idev->lock);
2948
2949                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2950                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2951                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2952                                                  jiffies);
2953                         ipv6_del_addr(ifp);
2954                         addrconf_verify_rtnl();
2955                         if (ipv6_addr_is_multicast(pfx)) {
2956                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2957                                                false, pfx, dev->ifindex);
2958                         }
2959                         return 0;
2960                 }
2961         }
2962         read_unlock_bh(&idev->lock);
2963         return -EADDRNOTAVAIL;
2964 }
2965
2966
2967 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2968 {
2969         struct in6_ifreq ireq;
2970         int err;
2971
2972         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2973                 return -EPERM;
2974
2975         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2976                 return -EFAULT;
2977
2978         rtnl_lock();
2979         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2980                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2981                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2982         rtnl_unlock();
2983         return err;
2984 }
2985
2986 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2987 {
2988         struct in6_ifreq ireq;
2989         int err;
2990
2991         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2992                 return -EPERM;
2993
2994         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2995                 return -EFAULT;
2996
2997         rtnl_lock();
2998         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2999                              ireq.ifr6_prefixlen);
3000         rtnl_unlock();
3001         return err;
3002 }
3003
3004 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3005                      int plen, int scope)
3006 {
3007         struct inet6_ifaddr *ifp;
3008
3009         ifp = ipv6_add_addr(idev, addr, NULL, plen,
3010                             scope, IFA_F_PERMANENT,
3011                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
3012                             true, NULL);
3013         if (!IS_ERR(ifp)) {
3014                 spin_lock_bh(&ifp->lock);
3015                 ifp->flags &= ~IFA_F_TENTATIVE;
3016                 spin_unlock_bh(&ifp->lock);
3017                 rt_genid_bump_ipv6(dev_net(idev->dev));
3018                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3019                 in6_ifa_put(ifp);
3020         }
3021 }
3022
3023 #if IS_ENABLED(CONFIG_IPV6_SIT)
3024 static void sit_add_v4_addrs(struct inet6_dev *idev)
3025 {
3026         struct in6_addr addr;
3027         struct net_device *dev;
3028         struct net *net = dev_net(idev->dev);
3029         int scope, plen;
3030         u32 pflags = 0;
3031
3032         ASSERT_RTNL();
3033
3034         memset(&addr, 0, sizeof(struct in6_addr));
3035         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3036
3037         if (idev->dev->flags&IFF_POINTOPOINT) {
3038                 addr.s6_addr32[0] = htonl(0xfe800000);
3039                 scope = IFA_LINK;
3040                 plen = 64;
3041         } else {
3042                 scope = IPV6_ADDR_COMPATv4;
3043                 plen = 96;
3044                 pflags |= RTF_NONEXTHOP;
3045         }
3046
3047         if (addr.s6_addr32[3]) {
3048                 add_addr(idev, &addr, plen, scope);
3049                 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3050                 return;
3051         }
3052
3053         for_each_netdev(net, dev) {
3054                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3055                 if (in_dev && (dev->flags & IFF_UP)) {
3056                         struct in_ifaddr *ifa;
3057
3058                         int flag = scope;
3059
3060                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3061
3062                                 addr.s6_addr32[3] = ifa->ifa_local;
3063
3064                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3065                                         continue;
3066                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3067                                         if (idev->dev->flags&IFF_POINTOPOINT)
3068                                                 continue;
3069                                         flag |= IFA_HOST;
3070                                 }
3071
3072                                 add_addr(idev, &addr, plen, flag);
3073                                 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3074                                                       pflags);
3075                         }
3076                 }
3077         }
3078 }
3079 #endif
3080
3081 static void init_loopback(struct net_device *dev)
3082 {
3083         struct inet6_dev  *idev;
3084
3085         /* ::1 */
3086
3087         ASSERT_RTNL();
3088
3089         idev = ipv6_find_idev(dev);
3090         if (!idev) {
3091                 pr_debug("%s: add_dev failed\n", __func__);
3092                 return;
3093         }
3094
3095         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3096 }
3097
3098 void addrconf_add_linklocal(struct inet6_dev *idev,
3099                             const struct in6_addr *addr, u32 flags)
3100 {
3101         struct inet6_ifaddr *ifp;
3102         u32 addr_flags = flags | IFA_F_PERMANENT;
3103
3104 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3105         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3106              idev->cnf.optimistic_dad) &&
3107             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3108                 addr_flags |= IFA_F_OPTIMISTIC;
3109 #endif
3110
3111         ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3112                             INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3113         if (!IS_ERR(ifp)) {
3114                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3115                 addrconf_dad_start(ifp);
3116                 in6_ifa_put(ifp);
3117         }
3118 }
3119 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3120
3121 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3122 {
3123         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3124                 return true;
3125
3126         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3127             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3128                 return true;
3129
3130         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3131             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3132                 return true;
3133
3134         return false;
3135 }
3136
3137 static int ipv6_generate_stable_address(struct in6_addr *address,
3138                                         u8 dad_count,
3139                                         const struct inet6_dev *idev)
3140 {
3141         static DEFINE_SPINLOCK(lock);
3142         static __u32 digest[SHA_DIGEST_WORDS];
3143         static __u32 workspace[SHA_WORKSPACE_WORDS];
3144
3145         static union {
3146                 char __data[SHA_MESSAGE_BYTES];
3147                 struct {
3148                         struct in6_addr secret;
3149                         __be32 prefix[2];
3150                         unsigned char hwaddr[MAX_ADDR_LEN];
3151                         u8 dad_count;
3152                 } __packed;
3153         } data;
3154
3155         struct in6_addr secret;
3156         struct in6_addr temp;
3157         struct net *net = dev_net(idev->dev);
3158
3159         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3160
3161         if (idev->cnf.stable_secret.initialized)
3162                 secret = idev->cnf.stable_secret.secret;
3163         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3164                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3165         else
3166                 return -1;
3167
3168 retry:
3169         spin_lock_bh(&lock);
3170
3171         sha_init(digest);
3172         memset(&data, 0, sizeof(data));
3173         memset(workspace, 0, sizeof(workspace));
3174         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3175         data.prefix[0] = address->s6_addr32[0];
3176         data.prefix[1] = address->s6_addr32[1];
3177         data.secret = secret;
3178         data.dad_count = dad_count;
3179
3180         sha_transform(digest, data.__data, workspace);
3181
3182         temp = *address;
3183         temp.s6_addr32[2] = (__force __be32)digest[0];
3184         temp.s6_addr32[3] = (__force __be32)digest[1];
3185
3186         spin_unlock_bh(&lock);
3187
3188         if (ipv6_reserved_interfaceid(temp)) {
3189                 dad_count++;
3190                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3191                         return -1;
3192                 goto retry;
3193         }
3194
3195         *address = temp;
3196         return 0;
3197 }
3198
3199 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3200 {
3201         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3202
3203         if (s->initialized)
3204                 return;
3205         s = &idev->cnf.stable_secret;
3206         get_random_bytes(&s->secret, sizeof(s->secret));
3207         s->initialized = true;
3208 }
3209
3210 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3211 {
3212         struct in6_addr addr;
3213
3214         /* no link local addresses on L3 master devices */
3215         if (netif_is_l3_master(idev->dev))
3216                 return;
3217
3218         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3219
3220         switch (idev->cnf.addr_gen_mode) {
3221         case IN6_ADDR_GEN_MODE_RANDOM:
3222                 ipv6_gen_mode_random_init(idev);
3223                 /* fallthrough */
3224         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3225                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3226                         addrconf_add_linklocal(idev, &addr,
3227                                                IFA_F_STABLE_PRIVACY);
3228                 else if (prefix_route)
3229                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3230                 break;
3231         case IN6_ADDR_GEN_MODE_EUI64:
3232                 /* addrconf_add_linklocal also adds a prefix_route and we
3233                  * only need to care about prefix routes if ipv6_generate_eui64
3234                  * couldn't generate one.
3235                  */
3236                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3237                         addrconf_add_linklocal(idev, &addr, 0);
3238                 else if (prefix_route)
3239                         addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3240                 break;
3241         case IN6_ADDR_GEN_MODE_NONE:
3242         default:
3243                 /* will not add any link local address */
3244                 break;
3245         }
3246 }
3247
3248 static void addrconf_dev_config(struct net_device *dev)
3249 {
3250         struct inet6_dev *idev;
3251
3252         ASSERT_RTNL();
3253
3254         if ((dev->type != ARPHRD_ETHER) &&
3255             (dev->type != ARPHRD_FDDI) &&
3256             (dev->type != ARPHRD_ARCNET) &&
3257             (dev->type != ARPHRD_INFINIBAND) &&
3258             (dev->type != ARPHRD_IEEE1394) &&
3259             (dev->type != ARPHRD_TUNNEL6) &&
3260             (dev->type != ARPHRD_6LOWPAN) &&
3261             (dev->type != ARPHRD_IP6GRE) &&
3262             (dev->type != ARPHRD_IPGRE) &&
3263             (dev->type != ARPHRD_TUNNEL) &&
3264             (dev->type != ARPHRD_NONE)) {
3265                 /* Alas, we support only Ethernet autoconfiguration. */
3266                 return;
3267         }
3268
3269         idev = addrconf_add_dev(dev);
3270         if (IS_ERR(idev))
3271                 return;
3272
3273         /* this device type has no EUI support */
3274         if (dev->type == ARPHRD_NONE &&
3275             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3276                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3277
3278         addrconf_addr_gen(idev, false);
3279 }
3280
3281 #if IS_ENABLED(CONFIG_IPV6_SIT)
3282 static void addrconf_sit_config(struct net_device *dev)
3283 {
3284         struct inet6_dev *idev;
3285
3286         ASSERT_RTNL();
3287
3288         /*
3289          * Configure the tunnel with one of our IPv4
3290          * addresses... we should configure all of
3291          * our v4 addrs in the tunnel
3292          */
3293
3294         idev = ipv6_find_idev(dev);
3295         if (!idev) {
3296                 pr_debug("%s: add_dev failed\n", __func__);
3297                 return;
3298         }
3299
3300         if (dev->priv_flags & IFF_ISATAP) {
3301                 addrconf_addr_gen(idev, false);
3302                 return;
3303         }
3304
3305         sit_add_v4_addrs(idev);
3306
3307         if (dev->flags&IFF_POINTOPOINT)
3308                 addrconf_add_mroute(dev);
3309 }
3310 #endif
3311
3312 #if IS_ENABLED(CONFIG_NET_IPGRE)
3313 static void addrconf_gre_config(struct net_device *dev)
3314 {
3315         struct inet6_dev *idev;
3316
3317         ASSERT_RTNL();
3318
3319         idev = ipv6_find_idev(dev);
3320         if (!idev) {
3321                 pr_debug("%s: add_dev failed\n", __func__);
3322                 return;
3323         }
3324
3325         addrconf_addr_gen(idev, true);
3326         if (dev->flags & IFF_POINTOPOINT)
3327                 addrconf_add_mroute(dev);
3328 }
3329 #endif
3330
3331 static int fixup_permanent_addr(struct inet6_dev *idev,
3332                                 struct inet6_ifaddr *ifp)
3333 {
3334         /* !rt6i_node means the host route was removed from the
3335          * FIB, for example, if 'lo' device is taken down. In that
3336          * case regenerate the host route.
3337          */
3338         if (!ifp->rt || !ifp->rt->rt6i_node) {
3339                 struct rt6_info *rt, *prev;
3340
3341                 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3342                 if (IS_ERR(rt))
3343                         return PTR_ERR(rt);
3344
3345                 /* ifp->rt can be accessed outside of rtnl */
3346                 spin_lock(&ifp->lock);
3347                 prev = ifp->rt;
3348                 ifp->rt = rt;
3349                 spin_unlock(&ifp->lock);
3350
3351                 ip6_rt_put(prev);
3352         }
3353
3354         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3355                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3356                                       idev->dev, 0, 0);
3357         }
3358
3359         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3360                 addrconf_dad_start(ifp);
3361
3362         return 0;
3363 }
3364
3365 static void addrconf_permanent_addr(struct net_device *dev)
3366 {
3367         struct inet6_ifaddr *ifp, *tmp;
3368         struct inet6_dev *idev;
3369
3370         idev = __in6_dev_get(dev);
3371         if (!idev)
3372                 return;
3373
3374         write_lock_bh(&idev->lock);
3375
3376         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3377                 if ((ifp->flags & IFA_F_PERMANENT) &&
3378                     fixup_permanent_addr(idev, ifp) < 0) {
3379                         write_unlock_bh(&idev->lock);
3380                         in6_ifa_hold(ifp);
3381                         ipv6_del_addr(ifp);
3382                         write_lock_bh(&idev->lock);
3383
3384                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3385                                              idev->dev->name, &ifp->addr);
3386                 }
3387         }
3388
3389         write_unlock_bh(&idev->lock);
3390 }
3391
3392 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3393                            void *ptr)
3394 {
3395         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3396         struct netdev_notifier_changeupper_info *info;
3397         struct inet6_dev *idev = __in6_dev_get(dev);
3398         struct net *net = dev_net(dev);
3399         int run_pending = 0;
3400         int err;
3401
3402         switch (event) {
3403         case NETDEV_REGISTER:
3404                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3405                         idev = ipv6_add_dev(dev);
3406                         if (IS_ERR(idev))
3407                                 return notifier_from_errno(PTR_ERR(idev));
3408                 }
3409                 break;
3410
3411         case NETDEV_CHANGEMTU:
3412                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3413                 if (dev->mtu < IPV6_MIN_MTU) {
3414                         addrconf_ifdown(dev, dev != net->loopback_dev);
3415                         break;
3416                 }
3417
3418                 if (idev) {
3419                         rt6_mtu_change(dev, dev->mtu);
3420                         idev->cnf.mtu6 = dev->mtu;
3421                         break;
3422                 }
3423
3424                 /* allocate new idev */
3425                 idev = ipv6_add_dev(dev);
3426                 if (IS_ERR(idev))
3427                         break;
3428
3429                 /* device is still not ready */
3430                 if (!(idev->if_flags & IF_READY))
3431                         break;
3432
3433                 run_pending = 1;
3434
3435                 /* fall through */
3436
3437         case NETDEV_UP:
3438         case NETDEV_CHANGE:
3439                 if (dev->flags & IFF_SLAVE)
3440                         break;
3441
3442                 if (idev && idev->cnf.disable_ipv6)
3443                         break;
3444
3445                 if (event == NETDEV_UP) {
3446                         /* restore routes for permanent addresses */
3447                         addrconf_permanent_addr(dev);
3448
3449                         if (!addrconf_link_ready(dev)) {
3450                                 /* device is not ready yet. */
3451                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3452                                         dev->name);
3453                                 break;
3454                         }
3455
3456                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3457                                 idev = ipv6_add_dev(dev);
3458
3459                         if (!IS_ERR_OR_NULL(idev)) {
3460                                 idev->if_flags |= IF_READY;
3461                                 run_pending = 1;
3462                         }
3463                 } else if (event == NETDEV_CHANGE) {
3464                         if (!addrconf_link_ready(dev)) {
3465                                 /* device is still not ready. */
3466                                 rt6_sync_down_dev(dev, event);
3467                                 break;
3468                         }
3469
3470                         if (idev) {
3471                                 if (idev->if_flags & IF_READY) {
3472                                         /* device is already configured -
3473                                          * but resend MLD reports, we might
3474                                          * have roamed and need to update
3475                                          * multicast snooping switches
3476                                          */
3477                                         ipv6_mc_up(idev);
3478                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3479                                         break;
3480                                 }
3481                                 idev->if_flags |= IF_READY;
3482                         }
3483
3484                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3485                                 dev->name);
3486
3487                         run_pending = 1;
3488                 }
3489
3490                 switch (dev->type) {
3491 #if IS_ENABLED(CONFIG_IPV6_SIT)
3492                 case ARPHRD_SIT:
3493                         addrconf_sit_config(dev);
3494                         break;
3495 #endif
3496 #if IS_ENABLED(CONFIG_NET_IPGRE)
3497                 case ARPHRD_IPGRE:
3498                         addrconf_gre_config(dev);
3499                         break;
3500 #endif
3501                 case ARPHRD_LOOPBACK:
3502                         init_loopback(dev);
3503                         break;
3504
3505                 default:
3506                         addrconf_dev_config(dev);
3507                         break;
3508                 }
3509
3510                 if (!IS_ERR_OR_NULL(idev)) {
3511                         if (run_pending)
3512                                 addrconf_dad_run(idev);
3513
3514                         /* Device has an address by now */
3515                         rt6_sync_up(dev, RTNH_F_DEAD);
3516
3517                         /*
3518                          * If the MTU changed during the interface down,
3519                          * when the interface up, the changed MTU must be
3520                          * reflected in the idev as well as routers.
3521                          */
3522                         if (idev->cnf.mtu6 != dev->mtu &&
3523                             dev->mtu >= IPV6_MIN_MTU) {
3524                                 rt6_mtu_change(dev, dev->mtu);
3525                                 idev->cnf.mtu6 = dev->mtu;
3526                         }
3527                         idev->tstamp = jiffies;
3528                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3529
3530                         /*
3531                          * If the changed mtu during down is lower than
3532                          * IPV6_MIN_MTU stop IPv6 on this interface.
3533                          */
3534                         if (dev->mtu < IPV6_MIN_MTU)
3535                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3536                 }
3537                 break;
3538
3539         case NETDEV_DOWN:
3540         case NETDEV_UNREGISTER:
3541                 /*
3542                  *      Remove all addresses from this interface.
3543                  */
3544                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3545                 break;
3546
3547         case NETDEV_CHANGENAME:
3548                 if (idev) {
3549                         snmp6_unregister_dev(idev);
3550                         addrconf_sysctl_unregister(idev);
3551                         err = addrconf_sysctl_register(idev);
3552                         if (err)
3553                                 return notifier_from_errno(err);
3554                         err = snmp6_register_dev(idev);
3555                         if (err) {
3556                                 addrconf_sysctl_unregister(idev);
3557                                 return notifier_from_errno(err);
3558                         }
3559                 }
3560                 break;
3561
3562         case NETDEV_PRE_TYPE_CHANGE:
3563         case NETDEV_POST_TYPE_CHANGE:
3564                 if (idev)
3565                         addrconf_type_change(dev, event);
3566                 break;
3567
3568         case NETDEV_CHANGEUPPER:
3569                 info = ptr;
3570
3571                 /* flush all routes if dev is linked to or unlinked from
3572                  * an L3 master device (e.g., VRF)
3573                  */
3574                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3575                         addrconf_ifdown(dev, 0);
3576         }
3577
3578         return NOTIFY_OK;
3579 }
3580
3581 /*
3582  *      addrconf module should be notified of a device going up
3583  */
3584 static struct notifier_block ipv6_dev_notf = {
3585         .notifier_call = addrconf_notify,
3586         .priority = ADDRCONF_NOTIFY_PRIORITY,
3587 };
3588
3589 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3590 {
3591         struct inet6_dev *idev;
3592         ASSERT_RTNL();
3593
3594         idev = __in6_dev_get(dev);
3595
3596         if (event == NETDEV_POST_TYPE_CHANGE)
3597                 ipv6_mc_remap(idev);
3598         else if (event == NETDEV_PRE_TYPE_CHANGE)
3599                 ipv6_mc_unmap(idev);
3600 }
3601
3602 static bool addr_is_local(const struct in6_addr *addr)
3603 {
3604         return ipv6_addr_type(addr) &
3605                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3606 }
3607
3608 static int addrconf_ifdown(struct net_device *dev, int how)
3609 {
3610         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3611         struct net *net = dev_net(dev);
3612         struct inet6_dev *idev;
3613         struct inet6_ifaddr *ifa, *tmp;
3614         int _keep_addr;
3615         bool keep_addr;
3616         int state, i;
3617
3618         ASSERT_RTNL();
3619
3620         rt6_disable_ip(dev, event);
3621
3622         idev = __in6_dev_get(dev);
3623         if (!idev)
3624                 return -ENODEV;
3625
3626         /*
3627          * Step 1: remove reference to ipv6 device from parent device.
3628          *         Do not dev_put!
3629          */
3630         if (how) {
3631                 idev->dead = 1;
3632
3633                 /* protected by rtnl_lock */
3634                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3635
3636                 /* Step 1.5: remove snmp6 entry */
3637                 snmp6_unregister_dev(idev);
3638
3639         }
3640
3641         /* aggregate the system setting and interface setting */
3642         _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3643         if (!_keep_addr)
3644                 _keep_addr = idev->cnf.keep_addr_on_down;
3645
3646         /* combine the user config with event to determine if permanent
3647          * addresses are to be removed from address hash table
3648          */
3649         keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3650
3651         /* Step 2: clear hash table */
3652         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3653                 struct hlist_head *h = &inet6_addr_lst[i];
3654
3655                 spin_lock_bh(&addrconf_hash_lock);
3656 restart:
3657                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3658                         if (ifa->idev == idev) {
3659                                 addrconf_del_dad_work(ifa);
3660                                 /* combined flag + permanent flag decide if
3661                                  * address is retained on a down event
3662                                  */
3663                                 if (!keep_addr ||
3664                                     !(ifa->flags & IFA_F_PERMANENT) ||
3665                                     addr_is_local(&ifa->addr)) {
3666                                         hlist_del_init_rcu(&ifa->addr_lst);
3667                                         goto restart;
3668                                 }
3669                         }
3670                 }
3671                 spin_unlock_bh(&addrconf_hash_lock);
3672         }
3673
3674         write_lock_bh(&idev->lock);
3675
3676         addrconf_del_rs_timer(idev);
3677
3678         /* Step 2: clear flags for stateless addrconf */
3679         if (!how)
3680                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3681
3682         /* Step 3: clear tempaddr list */
3683         while (!list_empty(&idev->tempaddr_list)) {
3684                 ifa = list_first_entry(&idev->tempaddr_list,
3685                                        struct inet6_ifaddr, tmp_list);
3686                 list_del(&ifa->tmp_list);
3687                 write_unlock_bh(&idev->lock);
3688                 spin_lock_bh(&ifa->lock);
3689
3690                 if (ifa->ifpub) {
3691                         in6_ifa_put(ifa->ifpub);
3692                         ifa->ifpub = NULL;
3693                 }
3694                 spin_unlock_bh(&ifa->lock);
3695                 in6_ifa_put(ifa);
3696                 write_lock_bh(&idev->lock);
3697         }
3698
3699         /* re-combine the user config with event to determine if permanent
3700          * addresses are to be removed from the interface list
3701          */
3702         keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3703
3704         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3705                 struct rt6_info *rt = NULL;
3706                 bool keep;
3707
3708                 addrconf_del_dad_work(ifa);
3709
3710                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3711                         !addr_is_local(&ifa->addr);
3712
3713                 write_unlock_bh(&idev->lock);
3714                 spin_lock_bh(&ifa->lock);
3715
3716                 if (keep) {
3717                         /* set state to skip the notifier below */
3718                         state = INET6_IFADDR_STATE_DEAD;
3719                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3720                         if (!(ifa->flags & IFA_F_NODAD))
3721                                 ifa->flags |= IFA_F_TENTATIVE;
3722
3723                         rt = ifa->rt;
3724                         ifa->rt = NULL;
3725                 } else {
3726                         state = ifa->state;
3727                         ifa->state = INET6_IFADDR_STATE_DEAD;
3728                 }
3729
3730                 spin_unlock_bh(&ifa->lock);
3731
3732                 if (rt)
3733                         ip6_del_rt(rt);
3734
3735                 if (state != INET6_IFADDR_STATE_DEAD) {
3736                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3737                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3738                 } else {
3739                         if (idev->cnf.forwarding)
3740                                 addrconf_leave_anycast(ifa);
3741                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3742                 }
3743
3744                 write_lock_bh(&idev->lock);
3745                 if (!keep) {
3746                         list_del_rcu(&ifa->if_list);
3747                         in6_ifa_put(ifa);
3748                 }
3749         }
3750
3751         write_unlock_bh(&idev->lock);
3752
3753         /* Step 5: Discard anycast and multicast list */
3754         if (how) {
3755                 ipv6_ac_destroy_dev(idev);
3756                 ipv6_mc_destroy_dev(idev);
3757         } else {
3758                 ipv6_mc_down(idev);
3759         }
3760
3761         idev->tstamp = jiffies;
3762
3763         /* Last: Shot the device (if unregistered) */
3764         if (how) {
3765                 addrconf_sysctl_unregister(idev);
3766                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3767                 neigh_ifdown(&nd_tbl, dev);
3768                 in6_dev_put(idev);
3769         }
3770         return 0;
3771 }
3772
3773 static void addrconf_rs_timer(struct timer_list *t)
3774 {
3775         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3776         struct net_device *dev = idev->dev;
3777         struct in6_addr lladdr;
3778
3779         write_lock(&idev->lock);
3780         if (idev->dead || !(idev->if_flags & IF_READY))
3781                 goto out;
3782
3783         if (!ipv6_accept_ra(idev))
3784                 goto out;
3785
3786         /* Announcement received after solicitation was sent */
3787         if (idev->if_flags & IF_RA_RCVD)
3788                 goto out;
3789
3790         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3791                 write_unlock(&idev->lock);
3792                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3793                         ndisc_send_rs(dev, &lladdr,
3794                                       &in6addr_linklocal_allrouters);
3795                 else
3796                         goto put;
3797
3798                 write_lock(&idev->lock);
3799                 idev->rs_interval = rfc3315_s14_backoff_update(
3800                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3801                 /* The wait after the last probe can be shorter */
3802                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3803                                              idev->cnf.rtr_solicits) ?
3804                                       idev->cnf.rtr_solicit_delay :
3805                                       idev->rs_interval);
3806         } else {
3807                 /*
3808                  * Note: we do not support deprecated "all on-link"
3809                  * assumption any longer.
3810                  */
3811                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3812         }
3813
3814 out:
3815         write_unlock(&idev->lock);
3816 put:
3817         in6_dev_put(idev);
3818 }
3819
3820 /*
3821  *      Duplicate Address Detection
3822  */
3823 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3824 {
3825         unsigned long rand_num;
3826         struct inet6_dev *idev = ifp->idev;
3827         u64 nonce;
3828
3829         if (ifp->flags & IFA_F_OPTIMISTIC)
3830                 rand_num = 0;
3831         else
3832                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3833
3834         nonce = 0;
3835         if (idev->cnf.enhanced_dad ||
3836             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3837                 do
3838                         get_random_bytes(&nonce, 6);
3839                 while (nonce == 0);
3840         }
3841         ifp->dad_nonce = nonce;
3842         ifp->dad_probes = idev->cnf.dad_transmits;
3843         addrconf_mod_dad_work(ifp, rand_num);
3844 }
3845
3846 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3847 {
3848         struct inet6_dev *idev = ifp->idev;
3849         struct net_device *dev = idev->dev;
3850         bool bump_id, notify = false;
3851
3852         addrconf_join_solict(dev, &ifp->addr);
3853
3854         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3855
3856         read_lock_bh(&idev->lock);
3857         spin_lock(&ifp->lock);
3858         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3859                 goto out;
3860
3861         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3862             (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3863              idev->cnf.accept_dad < 1) ||
3864             !(ifp->flags&IFA_F_TENTATIVE) ||
3865             ifp->flags & IFA_F_NODAD) {
3866                 bool send_na = false;
3867
3868                 if (ifp->flags & IFA_F_TENTATIVE &&
3869                     !(ifp->flags & IFA_F_OPTIMISTIC))
3870                         send_na = true;
3871                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3872                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3873                 spin_unlock(&ifp->lock);
3874                 read_unlock_bh(&idev->lock);
3875
3876                 addrconf_dad_completed(ifp, bump_id, send_na);
3877                 return;
3878         }
3879
3880         if (!(idev->if_flags & IF_READY)) {
3881                 spin_unlock(&ifp->lock);
3882                 read_unlock_bh(&idev->lock);
3883                 /*
3884                  * If the device is not ready:
3885                  * - keep it tentative if it is a permanent address.
3886                  * - otherwise, kill it.
3887                  */
3888                 in6_ifa_hold(ifp);
3889                 addrconf_dad_stop(ifp, 0);
3890                 return;
3891         }
3892
3893         /*
3894          * Optimistic nodes can start receiving
3895          * Frames right away
3896          */
3897         if (ifp->flags & IFA_F_OPTIMISTIC) {
3898                 ip6_ins_rt(ifp->rt);
3899                 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3900                         /* Because optimistic nodes can use this address,
3901                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
3902                          */
3903                         notify = true;
3904                 }
3905         }
3906
3907         addrconf_dad_kick(ifp);
3908 out:
3909         spin_unlock(&ifp->lock);
3910         read_unlock_bh(&idev->lock);
3911         if (notify)
3912                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3913 }
3914
3915 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3916 {
3917         bool begin_dad = false;
3918
3919         spin_lock_bh(&ifp->lock);
3920         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3921                 ifp->state = INET6_IFADDR_STATE_PREDAD;
3922                 begin_dad = true;
3923         }
3924         spin_unlock_bh(&ifp->lock);
3925
3926         if (begin_dad)
3927                 addrconf_mod_dad_work(ifp, 0);
3928 }
3929
3930 static void addrconf_dad_work(struct work_struct *w)
3931 {
3932         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3933                                                 struct inet6_ifaddr,
3934                                                 dad_work);
3935         struct inet6_dev *idev = ifp->idev;
3936         bool bump_id, disable_ipv6 = false;
3937         struct in6_addr mcaddr;
3938
3939         enum {
3940                 DAD_PROCESS,
3941                 DAD_BEGIN,
3942                 DAD_ABORT,
3943         } action = DAD_PROCESS;
3944
3945         rtnl_lock();
3946
3947         spin_lock_bh(&ifp->lock);
3948         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3949                 action = DAD_BEGIN;
3950                 ifp->state = INET6_IFADDR_STATE_DAD;
3951         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3952                 action = DAD_ABORT;
3953                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3954
3955                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3956                      idev->cnf.accept_dad > 1) &&
3957                     !idev->cnf.disable_ipv6 &&
3958                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3959                         struct in6_addr addr;
3960
3961                         addr.s6_addr32[0] = htonl(0xfe800000);
3962                         addr.s6_addr32[1] = 0;
3963
3964                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3965                             ipv6_addr_equal(&ifp->addr, &addr)) {
3966                                 /* DAD failed for link-local based on MAC */
3967                                 idev->cnf.disable_ipv6 = 1;
3968
3969                                 pr_info("%s: IPv6 being disabled!\n",
3970                                         ifp->idev->dev->name);
3971                                 disable_ipv6 = true;
3972                         }
3973                 }
3974         }
3975         spin_unlock_bh(&ifp->lock);
3976
3977         if (action == DAD_BEGIN) {
3978                 addrconf_dad_begin(ifp);
3979                 goto out;
3980         } else if (action == DAD_ABORT) {
3981                 in6_ifa_hold(ifp);
3982                 addrconf_dad_stop(ifp, 1);
3983                 if (disable_ipv6)
3984                         addrconf_ifdown(idev->dev, 0);
3985                 goto out;
3986         }
3987
3988         if (!ifp->dad_probes && addrconf_dad_end(ifp))
3989                 goto out;
3990
3991         write_lock_bh(&idev->lock);
3992         if (idev->dead || !(idev->if_flags & IF_READY)) {
3993                 write_unlock_bh(&idev->lock);
3994                 goto out;
3995         }
3996
3997         spin_lock(&ifp->lock);
3998         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3999                 spin_unlock(&ifp->lock);
4000                 write_unlock_bh(&idev->lock);
4001                 goto out;
4002         }
4003
4004         if (ifp->dad_probes == 0) {
4005                 bool send_na = false;
4006
4007                 /*
4008                  * DAD was successful
4009                  */
4010
4011                 if (ifp->flags & IFA_F_TENTATIVE &&
4012                     !(ifp->flags & IFA_F_OPTIMISTIC))
4013                         send_na = true;
4014                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4015                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4016                 spin_unlock(&ifp->lock);
4017                 write_unlock_bh(&idev->lock);
4018
4019                 addrconf_dad_completed(ifp, bump_id, send_na);
4020
4021                 goto out;
4022         }
4023
4024         ifp->dad_probes--;
4025         addrconf_mod_dad_work(ifp,
4026                               NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4027         spin_unlock(&ifp->lock);
4028         write_unlock_bh(&idev->lock);
4029
4030         /* send a neighbour solicitation for our addr */
4031         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4032         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4033                       ifp->dad_nonce);
4034 out:
4035         in6_ifa_put(ifp);
4036         rtnl_unlock();
4037 }
4038
4039 /* ifp->idev must be at least read locked */
4040 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4041 {
4042         struct inet6_ifaddr *ifpiter;
4043         struct inet6_dev *idev = ifp->idev;
4044
4045         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4046                 if (ifpiter->scope > IFA_LINK)
4047                         break;
4048                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4049                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4050                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4051                     IFA_F_PERMANENT)
4052                         return false;
4053         }
4054         return true;
4055 }
4056
4057 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4058                                    bool send_na)
4059 {
4060         struct net_device *dev = ifp->idev->dev;
4061         struct in6_addr lladdr;
4062         bool send_rs, send_mld;
4063
4064         addrconf_del_dad_work(ifp);
4065
4066         /*
4067          *      Configure the address for reception. Now it is valid.
4068          */
4069
4070         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4071
4072         /* If added prefix is link local and we are prepared to process
4073            router advertisements, start sending router solicitations.
4074          */
4075
4076         read_lock_bh(&ifp->idev->lock);
4077         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4078         send_rs = send_mld &&
4079                   ipv6_accept_ra(ifp->idev) &&
4080                   ifp->idev->cnf.rtr_solicits != 0 &&
4081                   (dev->flags&IFF_LOOPBACK) == 0;
4082         read_unlock_bh(&ifp->idev->lock);
4083
4084         /* While dad is in progress mld report's source address is in6_addrany.
4085          * Resend with proper ll now.
4086          */
4087         if (send_mld)
4088                 ipv6_mc_dad_complete(ifp->idev);
4089
4090         /* send unsolicited NA if enabled */
4091         if (send_na &&
4092             (ifp->idev->cnf.ndisc_notify ||
4093              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4094                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4095                               /*router=*/ !!ifp->idev->cnf.forwarding,
4096                               /*solicited=*/ false, /*override=*/ true,
4097                               /*inc_opt=*/ true);
4098         }
4099
4100         if (send_rs) {
4101                 /*
4102                  *      If a host as already performed a random delay
4103                  *      [...] as part of DAD [...] there is no need
4104                  *      to delay again before sending the first RS
4105                  */
4106                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4107                         return;
4108                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4109
4110                 write_lock_bh(&ifp->idev->lock);
4111                 spin_lock(&ifp->lock);
4112                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4113                         ifp->idev->cnf.rtr_solicit_interval);
4114                 ifp->idev->rs_probes = 1;
4115                 ifp->idev->if_flags |= IF_RS_SENT;
4116                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4117                 spin_unlock(&ifp->lock);
4118                 write_unlock_bh(&ifp->idev->lock);
4119         }
4120
4121         if (bump_id)
4122                 rt_genid_bump_ipv6(dev_net(dev));
4123
4124         /* Make sure that a new temporary address will be created
4125          * before this temporary address becomes deprecated.
4126          */
4127         if (ifp->flags & IFA_F_TEMPORARY)
4128                 addrconf_verify_rtnl();
4129 }
4130
4131 static void addrconf_dad_run(struct inet6_dev *idev)
4132 {
4133         struct inet6_ifaddr *ifp;
4134
4135         read_lock_bh(&idev->lock);
4136         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4137                 spin_lock(&ifp->lock);
4138                 if (ifp->flags & IFA_F_TENTATIVE &&
4139                     ifp->state == INET6_IFADDR_STATE_DAD)
4140                         addrconf_dad_kick(ifp);
4141                 spin_unlock(&ifp->lock);
4142         }
4143         read_unlock_bh(&idev->lock);
4144 }
4145
4146 #ifdef CONFIG_PROC_FS
4147 struct if6_iter_state {
4148         struct seq_net_private p;
4149         int bucket;
4150         int offset;
4151 };
4152
4153 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4154 {
4155         struct if6_iter_state *state = seq->private;
4156         struct net *net = seq_file_net(seq);
4157         struct inet6_ifaddr *ifa = NULL;
4158         int p = 0;
4159
4160         /* initial bucket if pos is 0 */
4161         if (pos == 0) {
4162                 state->bucket = 0;
4163                 state->offset = 0;
4164         }
4165
4166         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4167                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4168                                          addr_lst) {
4169                         if (!net_eq(dev_net(ifa->idev->dev), net))
4170                                 continue;
4171                         /* sync with offset */
4172                         if (p < state->offset) {
4173                                 p++;
4174                                 continue;
4175                         }
4176                         state->offset++;
4177                         return ifa;
4178                 }
4179
4180                 /* prepare for next bucket */
4181                 state->offset = 0;
4182                 p = 0;
4183         }
4184         return NULL;
4185 }
4186
4187 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4188                                          struct inet6_ifaddr *ifa)
4189 {
4190         struct if6_iter_state *state = seq->private;
4191         struct net *net = seq_file_net(seq);
4192
4193         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4194                 if (!net_eq(dev_net(ifa->idev->dev), net))
4195                         continue;
4196                 state->offset++;
4197                 return ifa;
4198         }
4199
4200         while (++state->bucket < IN6_ADDR_HSIZE) {
4201                 state->offset = 0;
4202                 hlist_for_each_entry_rcu(ifa,
4203                                      &inet6_addr_lst[state->bucket], addr_lst) {
4204                         if (!net_eq(dev_net(ifa->idev->dev), net))
4205                                 continue;
4206                         state->offset++;
4207                         return ifa;
4208                 }
4209         }
4210
4211         return NULL;
4212 }
4213
4214 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4215         __acquires(rcu)
4216 {
4217         rcu_read_lock();
4218         return if6_get_first(seq, *pos);
4219 }
4220
4221 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4222 {
4223         struct inet6_ifaddr *ifa;
4224
4225         ifa = if6_get_next(seq, v);
4226         ++*pos;
4227         return ifa;
4228 }
4229
4230 static void if6_seq_stop(struct seq_file *seq, void *v)
4231         __releases(rcu)
4232 {
4233         rcu_read_unlock();
4234 }
4235
4236 static int if6_seq_show(struct seq_file *seq, void *v)
4237 {
4238         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4239         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4240                    &ifp->addr,
4241                    ifp->idev->dev->ifindex,
4242                    ifp->prefix_len,
4243                    ifp->scope,
4244                    (u8) ifp->flags,
4245                    ifp->idev->dev->name);
4246         return 0;
4247 }
4248
4249 static const struct seq_operations if6_seq_ops = {
4250         .start  = if6_seq_start,
4251         .next   = if6_seq_next,
4252         .show   = if6_seq_show,
4253         .stop   = if6_seq_stop,
4254 };
4255
4256 static int if6_seq_open(struct inode *inode, struct file *file)
4257 {
4258         return seq_open_net(inode, file, &if6_seq_ops,
4259                             sizeof(struct if6_iter_state));
4260 }
4261
4262 static const struct file_operations if6_fops = {
4263         .open           = if6_seq_open,
4264         .read           = seq_read,
4265         .llseek         = seq_lseek,
4266         .release        = seq_release_net,
4267 };
4268
4269 static int __net_init if6_proc_net_init(struct net *net)
4270 {
4271         if (!proc_create("if_inet6", 0444, net->proc_net, &if6_fops))
4272                 return -ENOMEM;
4273         return 0;
4274 }
4275
4276 static void __net_exit if6_proc_net_exit(struct net *net)
4277 {
4278         remove_proc_entry("if_inet6", net->proc_net);
4279 }
4280
4281 static struct pernet_operations if6_proc_net_ops = {
4282         .init = if6_proc_net_init,
4283         .exit = if6_proc_net_exit,
4284 };
4285
4286 int __init if6_proc_init(void)
4287 {
4288         return register_pernet_subsys(&if6_proc_net_ops);
4289 }
4290
4291 void if6_proc_exit(void)
4292 {
4293         unregister_pernet_subsys(&if6_proc_net_ops);
4294 }
4295 #endif  /* CONFIG_PROC_FS */
4296
4297 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4298 /* Check if address is a home address configured on any interface. */
4299 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4300 {
4301         unsigned int hash = inet6_addr_hash(net, addr);
4302         struct inet6_ifaddr *ifp = NULL;
4303         int ret = 0;
4304
4305         rcu_read_lock();
4306         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4307                 if (!net_eq(dev_net(ifp->idev->dev), net))
4308                         continue;
4309                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4310                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4311                         ret = 1;
4312                         break;
4313                 }
4314         }
4315         rcu_read_unlock();
4316         return ret;
4317 }
4318 #endif
4319
4320 /*
4321  *      Periodic address status verification
4322  */
4323
4324 static void addrconf_verify_rtnl(void)
4325 {
4326         unsigned long now, next, next_sec, next_sched;
4327         struct inet6_ifaddr *ifp;
4328         int i;
4329
4330         ASSERT_RTNL();
4331
4332         rcu_read_lock_bh();
4333         now = jiffies;
4334         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4335
4336         cancel_delayed_work(&addr_chk_work);
4337
4338         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4339 restart:
4340                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4341                         unsigned long age;
4342
4343                         /* When setting preferred_lft to a value not zero or
4344                          * infinity, while valid_lft is infinity
4345                          * IFA_F_PERMANENT has a non-infinity life time.
4346                          */
4347                         if ((ifp->flags & IFA_F_PERMANENT) &&
4348                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4349                                 continue;
4350
4351                         spin_lock(&ifp->lock);
4352                         /* We try to batch several events at once. */
4353                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4354
4355                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4356                             age >= ifp->valid_lft) {
4357                                 spin_unlock(&ifp->lock);
4358                                 in6_ifa_hold(ifp);
4359                                 ipv6_del_addr(ifp);
4360                                 goto restart;
4361                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4362                                 spin_unlock(&ifp->lock);
4363                                 continue;
4364                         } else if (age >= ifp->prefered_lft) {
4365                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4366                                 int deprecate = 0;
4367
4368                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4369                                         deprecate = 1;
4370                                         ifp->flags |= IFA_F_DEPRECATED;
4371                                 }
4372
4373                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4374                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4375                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4376
4377                                 spin_unlock(&ifp->lock);
4378
4379                                 if (deprecate) {
4380                                         in6_ifa_hold(ifp);
4381
4382                                         ipv6_ifa_notify(0, ifp);
4383                                         in6_ifa_put(ifp);
4384                                         goto restart;
4385                                 }
4386                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4387                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4388                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4389                                         ifp->idev->cnf.dad_transmits *
4390                                         NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4391
4392                                 if (age >= ifp->prefered_lft - regen_advance) {
4393                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4394                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4395                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4396                                         if (!ifp->regen_count && ifpub) {
4397                                                 ifp->regen_count++;
4398                                                 in6_ifa_hold(ifp);
4399                                                 in6_ifa_hold(ifpub);
4400                                                 spin_unlock(&ifp->lock);
4401
4402                                                 spin_lock(&ifpub->lock);
4403                                                 ifpub->regen_count = 0;
4404                                                 spin_unlock(&ifpub->lock);
4405                                                 rcu_read_unlock_bh();
4406                                                 ipv6_create_tempaddr(ifpub, ifp, true);
4407                                                 in6_ifa_put(ifpub);
4408                                                 in6_ifa_put(ifp);
4409                                                 rcu_read_lock_bh();
4410                                                 goto restart;
4411                                         }
4412                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4413                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4414                                 spin_unlock(&ifp->lock);
4415                         } else {
4416                                 /* ifp->prefered_lft <= ifp->valid_lft */
4417                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4418                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4419                                 spin_unlock(&ifp->lock);
4420                         }
4421                 }
4422         }
4423
4424         next_sec = round_jiffies_up(next);
4425         next_sched = next;
4426
4427         /* If rounded timeout is accurate enough, accept it. */
4428         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4429                 next_sched = next_sec;
4430
4431         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4432         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4433                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4434
4435         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4436                  now, next, next_sec, next_sched);
4437         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4438         rcu_read_unlock_bh();
4439 }
4440
4441 static void addrconf_verify_work(struct work_struct *w)
4442 {
4443         rtnl_lock();
4444         addrconf_verify_rtnl();
4445         rtnl_unlock();
4446 }
4447
4448 static void addrconf_verify(void)
4449 {
4450         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4451 }
4452
4453 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4454                                      struct in6_addr **peer_pfx)
4455 {
4456         struct in6_addr *pfx = NULL;
4457
4458         *peer_pfx = NULL;
4459
4460         if (addr)
4461                 pfx = nla_data(addr);
4462
4463         if (local) {
4464                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4465                         *peer_pfx = pfx;
4466                 pfx = nla_data(local);
4467         }
4468
4469         return pfx;
4470 }
4471
4472 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4473         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4474         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4475         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4476         [IFA_FLAGS]             = { .len = sizeof(u32) },
4477 };
4478
4479 static int
4480 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4481                   struct netlink_ext_ack *extack)
4482 {
4483         struct net *net = sock_net(skb->sk);
4484         struct ifaddrmsg *ifm;
4485         struct nlattr *tb[IFA_MAX+1];
4486         struct in6_addr *pfx, *peer_pfx;
4487         u32 ifa_flags;
4488         int err;
4489
4490         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4491                           extack);
4492         if (err < 0)
4493                 return err;
4494
4495         ifm = nlmsg_data(nlh);
4496         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4497         if (!pfx)
4498                 return -EINVAL;
4499
4500         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4501
4502         /* We ignore other flags so far. */
4503         ifa_flags &= IFA_F_MANAGETEMPADDR;
4504
4505         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4506                               ifm->ifa_prefixlen);
4507 }
4508
4509 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4510                              u32 prefered_lft, u32 valid_lft)
4511 {
4512         u32 flags;
4513         clock_t expires;
4514         unsigned long timeout;
4515         bool was_managetempaddr;
4516         bool had_prefixroute;
4517
4518         ASSERT_RTNL();
4519
4520         if (!valid_lft || (prefered_lft > valid_lft))
4521                 return -EINVAL;
4522
4523         if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4524             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4525                 return -EINVAL;
4526
4527         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4528                 ifa_flags &= ~IFA_F_OPTIMISTIC;
4529
4530         timeout = addrconf_timeout_fixup(valid_lft, HZ);
4531         if (addrconf_finite_timeout(timeout)) {
4532                 expires = jiffies_to_clock_t(timeout * HZ);
4533                 valid_lft = timeout;
4534                 flags = RTF_EXPIRES;
4535         } else {
4536                 expires = 0;
4537                 flags = 0;
4538                 ifa_flags |= IFA_F_PERMANENT;
4539         }
4540
4541         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4542         if (addrconf_finite_timeout(timeout)) {
4543                 if (timeout == 0)
4544                         ifa_flags |= IFA_F_DEPRECATED;
4545                 prefered_lft = timeout;
4546         }
4547
4548         spin_lock_bh(&ifp->lock);
4549         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4550         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4551                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4552         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4553                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4554                         IFA_F_NOPREFIXROUTE);
4555         ifp->flags |= ifa_flags;
4556         ifp->tstamp = jiffies;
4557         ifp->valid_lft = valid_lft;
4558         ifp->prefered_lft = prefered_lft;
4559
4560         spin_unlock_bh(&ifp->lock);
4561         if (!(ifp->flags&IFA_F_TENTATIVE))
4562                 ipv6_ifa_notify(0, ifp);
4563
4564         if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4565                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4566                                       expires, flags);
4567         } else if (had_prefixroute) {
4568                 enum cleanup_prefix_rt_t action;
4569                 unsigned long rt_expires;
4570
4571                 write_lock_bh(&ifp->idev->lock);
4572                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4573                 write_unlock_bh(&ifp->idev->lock);
4574
4575                 if (action != CLEANUP_PREFIX_RT_NOP) {
4576                         cleanup_prefix_route(ifp, rt_expires,
4577                                 action == CLEANUP_PREFIX_RT_DEL);
4578                 }
4579         }
4580
4581         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4582                 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4583                         valid_lft = prefered_lft = 0;
4584                 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4585                                  !was_managetempaddr, jiffies);
4586         }
4587
4588         addrconf_verify_rtnl();
4589
4590         return 0;
4591 }
4592
4593 static int
4594 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4595                   struct netlink_ext_ack *extack)
4596 {
4597         struct net *net = sock_net(skb->sk);
4598         struct ifaddrmsg *ifm;
4599         struct nlattr *tb[IFA_MAX+1];
4600         struct in6_addr *pfx, *peer_pfx;
4601         struct inet6_ifaddr *ifa;
4602         struct net_device *dev;
4603         struct inet6_dev *idev;
4604         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4605         u32 ifa_flags;
4606         int err;
4607
4608         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4609                           extack);
4610         if (err < 0)
4611                 return err;
4612
4613         ifm = nlmsg_data(nlh);
4614         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4615         if (!pfx)
4616                 return -EINVAL;
4617
4618         if (tb[IFA_CACHEINFO]) {
4619                 struct ifa_cacheinfo *ci;
4620
4621                 ci = nla_data(tb[IFA_CACHEINFO]);
4622                 valid_lft = ci->ifa_valid;
4623                 preferred_lft = ci->ifa_prefered;
4624         } else {
4625                 preferred_lft = INFINITY_LIFE_TIME;
4626                 valid_lft = INFINITY_LIFE_TIME;
4627         }
4628
4629         dev =  __dev_get_by_index(net, ifm->ifa_index);
4630         if (!dev)
4631                 return -ENODEV;
4632
4633         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4634
4635         /* We ignore other flags so far. */
4636         ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4637                      IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4638
4639         idev = ipv6_find_idev(dev);
4640         if (IS_ERR(idev))
4641                 return PTR_ERR(idev);
4642
4643         if (!ipv6_allow_optimistic_dad(net, idev))
4644                 ifa_flags &= ~IFA_F_OPTIMISTIC;
4645
4646         if (ifa_flags & IFA_F_NODAD && ifa_flags & IFA_F_OPTIMISTIC) {
4647                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4648                 return -EINVAL;
4649         }
4650
4651         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4652         if (!ifa) {
4653                 /*
4654                  * It would be best to check for !NLM_F_CREATE here but
4655                  * userspace already relies on not having to provide this.
4656                  */
4657                 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4658                                       ifm->ifa_prefixlen, ifa_flags,
4659                                       preferred_lft, valid_lft, extack);
4660         }
4661
4662         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4663             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4664                 err = -EEXIST;
4665         else
4666                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4667
4668         in6_ifa_put(ifa);
4669
4670         return err;
4671 }
4672
4673 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4674                           u8 scope, int ifindex)
4675 {
4676         struct ifaddrmsg *ifm;
4677
4678         ifm = nlmsg_data(nlh);
4679         ifm->ifa_family = AF_INET6;
4680         ifm->ifa_prefixlen = prefixlen;
4681         ifm->ifa_flags = flags;
4682         ifm->ifa_scope = scope;
4683         ifm->ifa_index = ifindex;
4684 }
4685
4686 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4687                          unsigned long tstamp, u32 preferred, u32 valid)
4688 {
4689         struct ifa_cacheinfo ci;
4690
4691         ci.cstamp = cstamp_delta(cstamp);
4692         ci.tstamp = cstamp_delta(tstamp);
4693         ci.ifa_prefered = preferred;
4694         ci.ifa_valid = valid;
4695
4696         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4697 }
4698
4699 static inline int rt_scope(int ifa_scope)
4700 {
4701         if (ifa_scope & IFA_HOST)
4702                 return RT_SCOPE_HOST;
4703         else if (ifa_scope & IFA_LINK)
4704                 return RT_SCOPE_LINK;
4705         else if (ifa_scope & IFA_SITE)
4706                 return RT_SCOPE_SITE;
4707         else
4708                 return RT_SCOPE_UNIVERSE;
4709 }
4710
4711 static inline int inet6_ifaddr_msgsize(void)
4712 {
4713         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4714                + nla_total_size(16) /* IFA_LOCAL */
4715                + nla_total_size(16) /* IFA_ADDRESS */
4716                + nla_total_size(sizeof(struct ifa_cacheinfo))
4717                + nla_total_size(4)  /* IFA_FLAGS */;
4718 }
4719
4720 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4721                              u32 portid, u32 seq, int event, unsigned int flags)
4722 {
4723         struct nlmsghdr  *nlh;
4724         u32 preferred, valid;
4725
4726         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4727         if (!nlh)
4728                 return -EMSGSIZE;
4729
4730         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4731                       ifa->idev->dev->ifindex);
4732
4733         if (!((ifa->flags&IFA_F_PERMANENT) &&
4734               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4735                 preferred = ifa->prefered_lft;
4736                 valid = ifa->valid_lft;
4737                 if (preferred != INFINITY_LIFE_TIME) {
4738                         long tval = (jiffies - ifa->tstamp)/HZ;
4739                         if (preferred > tval)
4740                                 preferred -= tval;
4741                         else
4742                                 preferred = 0;
4743                         if (valid != INFINITY_LIFE_TIME) {
4744                                 if (valid > tval)
4745                                         valid -= tval;
4746                                 else
4747                                         valid = 0;
4748                         }
4749                 }
4750         } else {
4751                 preferred = INFINITY_LIFE_TIME;
4752                 valid = INFINITY_LIFE_TIME;
4753         }
4754
4755         if (!ipv6_addr_any(&ifa->peer_addr)) {
4756                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4757                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4758                         goto error;
4759         } else
4760                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4761                         goto error;
4762
4763         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4764                 goto error;
4765
4766         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4767                 goto error;
4768
4769         nlmsg_end(skb, nlh);
4770         return 0;
4771
4772 error:
4773         nlmsg_cancel(skb, nlh);
4774         return -EMSGSIZE;
4775 }
4776
4777 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4778                                 u32 portid, u32 seq, int event, u16 flags)
4779 {
4780         struct nlmsghdr  *nlh;
4781         u8 scope = RT_SCOPE_UNIVERSE;
4782         int ifindex = ifmca->idev->dev->ifindex;
4783
4784         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4785                 scope = RT_SCOPE_SITE;
4786
4787         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4788         if (!nlh)
4789                 return -EMSGSIZE;
4790
4791         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4792         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4793             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4794                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4795                 nlmsg_cancel(skb, nlh);
4796                 return -EMSGSIZE;
4797         }
4798
4799         nlmsg_end(skb, nlh);
4800         return 0;
4801 }
4802
4803 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4804                                 u32 portid, u32 seq, int event, unsigned int flags)
4805 {
4806         struct nlmsghdr  *nlh;
4807         u8 scope = RT_SCOPE_UNIVERSE;
4808         int ifindex = ifaca->aca_idev->dev->ifindex;
4809
4810         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4811                 scope = RT_SCOPE_SITE;
4812
4813         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4814         if (!nlh)
4815                 return -EMSGSIZE;
4816
4817         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4818         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4819             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4820                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4821                 nlmsg_cancel(skb, nlh);
4822                 return -EMSGSIZE;
4823         }
4824
4825         nlmsg_end(skb, nlh);
4826         return 0;
4827 }
4828
4829 enum addr_type_t {
4830         UNICAST_ADDR,
4831         MULTICAST_ADDR,
4832         ANYCAST_ADDR,
4833 };
4834
4835 /* called with rcu_read_lock() */
4836 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4837                           struct netlink_callback *cb, enum addr_type_t type,
4838                           int s_ip_idx, int *p_ip_idx)
4839 {
4840         struct ifmcaddr6 *ifmca;
4841         struct ifacaddr6 *ifaca;
4842         int err = 1;
4843         int ip_idx = *p_ip_idx;
4844
4845         read_lock_bh(&idev->lock);
4846         switch (type) {
4847         case UNICAST_ADDR: {
4848                 struct inet6_ifaddr *ifa;
4849
4850                 /* unicast address incl. temp addr */
4851                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4852                         if (++ip_idx < s_ip_idx)
4853                                 continue;
4854                         err = inet6_fill_ifaddr(skb, ifa,
4855                                                 NETLINK_CB(cb->skb).portid,
4856                                                 cb->nlh->nlmsg_seq,
4857                                                 RTM_NEWADDR,
4858                                                 NLM_F_MULTI);
4859                         if (err < 0)
4860                                 break;
4861                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4862                 }
4863                 break;
4864         }
4865         case MULTICAST_ADDR:
4866                 /* multicast address */
4867                 for (ifmca = idev->mc_list; ifmca;
4868                      ifmca = ifmca->next, ip_idx++) {
4869                         if (ip_idx < s_ip_idx)
4870                                 continue;
4871                         err = inet6_fill_ifmcaddr(skb, ifmca,
4872                                                   NETLINK_CB(cb->skb).portid,
4873                                                   cb->nlh->nlmsg_seq,
4874                                                   RTM_GETMULTICAST,
4875                                                   NLM_F_MULTI);
4876                         if (err < 0)
4877                                 break;
4878                 }
4879                 break;
4880         case ANYCAST_ADDR:
4881                 /* anycast address */
4882                 for (ifaca = idev->ac_list; ifaca;
4883                      ifaca = ifaca->aca_next, ip_idx++) {
4884                         if (ip_idx < s_ip_idx)
4885                                 continue;
4886                         err = inet6_fill_ifacaddr(skb, ifaca,
4887                                                   NETLINK_CB(cb->skb).portid,
4888                                                   cb->nlh->nlmsg_seq,
4889                                                   RTM_GETANYCAST,
4890                                                   NLM_F_MULTI);
4891                         if (err < 0)
4892                                 break;
4893                 }
4894                 break;
4895         default:
4896                 break;
4897         }
4898         read_unlock_bh(&idev->lock);
4899         *p_ip_idx = ip_idx;
4900         return err;
4901 }
4902
4903 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4904                            enum addr_type_t type)
4905 {
4906         struct net *net = sock_net(skb->sk);
4907         int h, s_h;
4908         int idx, ip_idx;
4909         int s_idx, s_ip_idx;
4910         struct net_device *dev;
4911         struct inet6_dev *idev;
4912         struct hlist_head *head;
4913
4914         s_h = cb->args[0];
4915         s_idx = idx = cb->args[1];
4916         s_ip_idx = ip_idx = cb->args[2];
4917
4918         rcu_read_lock();
4919         cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4920         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4921                 idx = 0;
4922                 head = &net->dev_index_head[h];
4923                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4924                         if (idx < s_idx)
4925                                 goto cont;
4926                         if (h > s_h || idx > s_idx)
4927                                 s_ip_idx = 0;
4928                         ip_idx = 0;
4929                         idev = __in6_dev_get(dev);
4930                         if (!idev)
4931                                 goto cont;
4932
4933                         if (in6_dump_addrs(idev, skb, cb, type,
4934                                            s_ip_idx, &ip_idx) < 0)
4935                                 goto done;
4936 cont:
4937                         idx++;
4938                 }
4939         }
4940 done:
4941         rcu_read_unlock();
4942         cb->args[0] = h;
4943         cb->args[1] = idx;
4944         cb->args[2] = ip_idx;
4945
4946         return skb->len;
4947 }
4948
4949 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4950 {
4951         enum addr_type_t type = UNICAST_ADDR;
4952
4953         return inet6_dump_addr(skb, cb, type);
4954 }
4955
4956 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4957 {
4958         enum addr_type_t type = MULTICAST_ADDR;
4959
4960         return inet6_dump_addr(skb, cb, type);
4961 }
4962
4963
4964 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4965 {
4966         enum addr_type_t type = ANYCAST_ADDR;
4967
4968         return inet6_dump_addr(skb, cb, type);
4969 }
4970
4971 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4972                              struct netlink_ext_ack *extack)
4973 {
4974         struct net *net = sock_net(in_skb->sk);
4975         struct ifaddrmsg *ifm;
4976         struct nlattr *tb[IFA_MAX+1];
4977         struct in6_addr *addr = NULL, *peer;
4978         struct net_device *dev = NULL;
4979         struct inet6_ifaddr *ifa;
4980         struct sk_buff *skb;
4981         int err;
4982
4983         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4984                           extack);
4985         if (err < 0)
4986                 return err;
4987
4988         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4989         if (!addr)
4990                 return -EINVAL;
4991
4992         ifm = nlmsg_data(nlh);
4993         if (ifm->ifa_index)
4994                 dev = dev_get_by_index(net, ifm->ifa_index);
4995
4996         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4997         if (!ifa) {
4998                 err = -EADDRNOTAVAIL;
4999                 goto errout;
5000         }
5001
5002         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5003         if (!skb) {
5004                 err = -ENOBUFS;
5005                 goto errout_ifa;
5006         }
5007
5008         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
5009                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
5010         if (err < 0) {
5011                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5012                 WARN_ON(err == -EMSGSIZE);
5013                 kfree_skb(skb);
5014                 goto errout_ifa;
5015         }
5016         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5017 errout_ifa:
5018         in6_ifa_put(ifa);
5019 errout:
5020         if (dev)
5021                 dev_put(dev);
5022         return err;
5023 }
5024
5025 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5026 {
5027         struct sk_buff *skb;
5028         struct net *net = dev_net(ifa->idev->dev);
5029         int err = -ENOBUFS;
5030
5031         /* Don't send DELADDR notification for TENTATIVE address,
5032          * since NEWADDR notification is sent only after removing
5033          * TENTATIVE flag, if DAD has not failed.
5034          */
5035         if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
5036             event == RTM_DELADDR)
5037                 return;
5038
5039         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5040         if (!skb)
5041                 goto errout;
5042
5043         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5044         if (err < 0) {
5045                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5046                 WARN_ON(err == -EMSGSIZE);
5047                 kfree_skb(skb);
5048                 goto errout;
5049         }
5050         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5051         return;
5052 errout:
5053         if (err < 0)
5054                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5055 }
5056
5057 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5058                                 __s32 *array, int bytes)
5059 {
5060         BUG_ON(bytes < (DEVCONF_MAX * 4));
5061
5062         memset(array, 0, bytes);
5063         array[DEVCONF_FORWARDING] = cnf->forwarding;
5064         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5065         array[DEVCONF_MTU6] = cnf->mtu6;
5066         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5067         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5068         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5069         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5070         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5071         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5072                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5073         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5074                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5075         array[DEVCONF_RTR_SOLICIT_DELAY] =
5076                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5077         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5078         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5079                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5080         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5081                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5082         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5083         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5084         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5085         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5086         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5087         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5088         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5089         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5090         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5091 #ifdef CONFIG_IPV6_ROUTER_PREF
5092         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5093         array[DEVCONF_RTR_PROBE_INTERVAL] =
5094                 jiffies_to_msecs(cnf->rtr_probe_interval);
5095 #ifdef CONFIG_IPV6_ROUTE_INFO
5096         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5097         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5098 #endif
5099 #endif
5100         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5101         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5102 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5103         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5104         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5105 #endif
5106 #ifdef CONFIG_IPV6_MROUTE
5107         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5108 #endif
5109         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5110         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5111         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5112         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5113         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5114         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5115         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5116         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5117         /* we omit DEVCONF_STABLE_SECRET for now */
5118         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5119         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5120         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5121         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5122         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5123 #ifdef CONFIG_IPV6_SEG6_HMAC
5124         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5125 #endif
5126         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5127         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5128         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5129         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5130 }
5131
5132 static inline size_t inet6_ifla6_size(void)
5133 {
5134         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5135              + nla_total_size(sizeof(struct ifla_cacheinfo))
5136              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5137              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5138              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5139              + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5140 }
5141
5142 static inline size_t inet6_if_nlmsg_size(void)
5143 {
5144         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5145                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5146                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5147                + nla_total_size(4) /* IFLA_MTU */
5148                + nla_total_size(4) /* IFLA_LINK */
5149                + nla_total_size(1) /* IFLA_OPERSTATE */
5150                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5151 }
5152
5153 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5154                                         int bytes)
5155 {
5156         int i;
5157         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5158         BUG_ON(pad < 0);
5159
5160         /* Use put_unaligned() because stats may not be aligned for u64. */
5161         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5162         for (i = 1; i < ICMP6_MIB_MAX; i++)
5163                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5164
5165         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5166 }
5167
5168 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5169                                         int bytes, size_t syncpoff)
5170 {
5171         int i, c;
5172         u64 buff[IPSTATS_MIB_MAX];
5173         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5174
5175         BUG_ON(pad < 0);
5176
5177         memset(buff, 0, sizeof(buff));
5178         buff[0] = IPSTATS_MIB_MAX;
5179
5180         for_each_possible_cpu(c) {
5181                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5182                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5183         }
5184
5185         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5186         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5187 }
5188
5189 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5190                              int bytes)
5191 {
5192         switch (attrtype) {
5193         case IFLA_INET6_STATS:
5194                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5195                                      offsetof(struct ipstats_mib, syncp));
5196                 break;
5197         case IFLA_INET6_ICMP6STATS:
5198                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5199                 break;
5200         }
5201 }
5202
5203 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5204                                   u32 ext_filter_mask)
5205 {
5206         struct nlattr *nla;
5207         struct ifla_cacheinfo ci;
5208
5209         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5210                 goto nla_put_failure;
5211         ci.max_reasm_len = IPV6_MAXPLEN;
5212         ci.tstamp = cstamp_delta(idev->tstamp);
5213         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5214         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5215         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5216                 goto nla_put_failure;
5217         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5218         if (!nla)
5219                 goto nla_put_failure;
5220         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5221
5222         /* XXX - MC not implemented */
5223
5224         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5225                 return 0;
5226
5227         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5228         if (!nla)
5229                 goto nla_put_failure;
5230         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5231
5232         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5233         if (!nla)
5234                 goto nla_put_failure;
5235         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5236
5237         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5238         if (!nla)
5239                 goto nla_put_failure;
5240
5241         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5242                 goto nla_put_failure;
5243
5244         read_lock_bh(&idev->lock);
5245         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5246         read_unlock_bh(&idev->lock);
5247
5248         return 0;
5249
5250 nla_put_failure:
5251         return -EMSGSIZE;
5252 }
5253
5254 static size_t inet6_get_link_af_size(const struct net_device *dev,
5255                                      u32 ext_filter_mask)
5256 {
5257         if (!__in6_dev_get(dev))
5258                 return 0;
5259
5260         return inet6_ifla6_size();
5261 }
5262
5263 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5264                               u32 ext_filter_mask)
5265 {
5266         struct inet6_dev *idev = __in6_dev_get(dev);
5267
5268         if (!idev)
5269                 return -ENODATA;
5270
5271         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5272                 return -EMSGSIZE;
5273
5274         return 0;
5275 }
5276
5277 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5278 {
5279         struct inet6_ifaddr *ifp;
5280         struct net_device *dev = idev->dev;
5281         bool clear_token, update_rs = false;
5282         struct in6_addr ll_addr;
5283
5284         ASSERT_RTNL();
5285
5286         if (!token)
5287                 return -EINVAL;
5288         if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5289                 return -EINVAL;
5290         if (!ipv6_accept_ra(idev))
5291                 return -EINVAL;
5292         if (idev->cnf.rtr_solicits == 0)
5293                 return -EINVAL;
5294
5295         write_lock_bh(&idev->lock);
5296
5297         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5298         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5299
5300         write_unlock_bh(&idev->lock);
5301
5302         clear_token = ipv6_addr_any(token);
5303         if (clear_token)
5304                 goto update_lft;
5305
5306         if (!idev->dead && (idev->if_flags & IF_READY) &&
5307             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5308                              IFA_F_OPTIMISTIC)) {
5309                 /* If we're not ready, then normal ifup will take care
5310                  * of this. Otherwise, we need to request our rs here.
5311                  */
5312                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5313                 update_rs = true;
5314         }
5315
5316 update_lft:
5317         write_lock_bh(&idev->lock);
5318
5319         if (update_rs) {
5320                 idev->if_flags |= IF_RS_SENT;
5321                 idev->rs_interval = rfc3315_s14_backoff_init(
5322                         idev->cnf.rtr_solicit_interval);
5323                 idev->rs_probes = 1;
5324                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5325         }
5326
5327         /* Well, that's kinda nasty ... */
5328         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5329                 spin_lock(&ifp->lock);
5330                 if (ifp->tokenized) {
5331                         ifp->valid_lft = 0;
5332                         ifp->prefered_lft = 0;
5333                 }
5334                 spin_unlock(&ifp->lock);
5335         }
5336
5337         write_unlock_bh(&idev->lock);
5338         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5339         addrconf_verify_rtnl();
5340         return 0;
5341 }
5342
5343 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5344         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5345         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5346 };
5347
5348 static int inet6_validate_link_af(const struct net_device *dev,
5349                                   const struct nlattr *nla)
5350 {
5351         struct nlattr *tb[IFLA_INET6_MAX + 1];
5352
5353         if (dev && !__in6_dev_get(dev))
5354                 return -EAFNOSUPPORT;
5355
5356         return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5357                                 NULL);
5358 }
5359
5360 static int check_addr_gen_mode(int mode)
5361 {
5362         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5363             mode != IN6_ADDR_GEN_MODE_NONE &&
5364             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5365             mode != IN6_ADDR_GEN_MODE_RANDOM)
5366                 return -EINVAL;
5367         return 1;
5368 }
5369
5370 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5371                                 int mode)
5372 {
5373         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5374             !idev->cnf.stable_secret.initialized &&
5375             !net->ipv6.devconf_dflt->stable_secret.initialized)
5376                 return -EINVAL;
5377         return 1;
5378 }
5379
5380 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5381 {
5382         int err = -EINVAL;
5383         struct inet6_dev *idev = __in6_dev_get(dev);
5384         struct nlattr *tb[IFLA_INET6_MAX + 1];
5385
5386         if (!idev)
5387                 return -EAFNOSUPPORT;
5388
5389         if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5390                 BUG();
5391
5392         if (tb[IFLA_INET6_TOKEN]) {
5393                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5394                 if (err)
5395                         return err;
5396         }
5397
5398         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5399                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5400
5401                 if (check_addr_gen_mode(mode) < 0 ||
5402                     check_stable_privacy(idev, dev_net(dev), mode) < 0)
5403                         return -EINVAL;
5404
5405                 idev->cnf.addr_gen_mode = mode;
5406                 err = 0;
5407         }
5408
5409         return err;
5410 }
5411
5412 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5413                              u32 portid, u32 seq, int event, unsigned int flags)
5414 {
5415         struct net_device *dev = idev->dev;
5416         struct ifinfomsg *hdr;
5417         struct nlmsghdr *nlh;
5418         void *protoinfo;
5419
5420         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5421         if (!nlh)
5422                 return -EMSGSIZE;
5423
5424         hdr = nlmsg_data(nlh);
5425         hdr->ifi_family = AF_INET6;
5426         hdr->__ifi_pad = 0;
5427         hdr->ifi_type = dev->type;
5428         hdr->ifi_index = dev->ifindex;
5429         hdr->ifi_flags = dev_get_flags(dev);
5430         hdr->ifi_change = 0;
5431
5432         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5433             (dev->addr_len &&
5434              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5435             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5436             (dev->ifindex != dev_get_iflink(dev) &&
5437              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5438             nla_put_u8(skb, IFLA_OPERSTATE,
5439                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5440                 goto nla_put_failure;
5441         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5442         if (!protoinfo)
5443                 goto nla_put_failure;
5444
5445         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5446                 goto nla_put_failure;
5447
5448         nla_nest_end(skb, protoinfo);
5449         nlmsg_end(skb, nlh);
5450         return 0;
5451
5452 nla_put_failure:
5453         nlmsg_cancel(skb, nlh);
5454         return -EMSGSIZE;
5455 }
5456
5457 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5458 {
5459         struct net *net = sock_net(skb->sk);
5460         int h, s_h;
5461         int idx = 0, s_idx;
5462         struct net_device *dev;
5463         struct inet6_dev *idev;
5464         struct hlist_head *head;
5465
5466         s_h = cb->args[0];
5467         s_idx = cb->args[1];
5468
5469         rcu_read_lock();
5470         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5471                 idx = 0;
5472                 head = &net->dev_index_head[h];
5473                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5474                         if (idx < s_idx)
5475                                 goto cont;
5476                         idev = __in6_dev_get(dev);
5477                         if (!idev)
5478                                 goto cont;
5479                         if (inet6_fill_ifinfo(skb, idev,
5480                                               NETLINK_CB(cb->skb).portid,
5481                                               cb->nlh->nlmsg_seq,
5482                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5483                                 goto out;
5484 cont:
5485                         idx++;
5486                 }
5487         }
5488 out:
5489         rcu_read_unlock();
5490         cb->args[1] = idx;
5491         cb->args[0] = h;
5492
5493         return skb->len;
5494 }
5495
5496 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5497 {
5498         struct sk_buff *skb;
5499         struct net *net = dev_net(idev->dev);
5500         int err = -ENOBUFS;
5501
5502         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5503         if (!skb)
5504                 goto errout;
5505
5506         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5507         if (err < 0) {
5508                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5509                 WARN_ON(err == -EMSGSIZE);
5510                 kfree_skb(skb);
5511                 goto errout;
5512         }
5513         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5514         return;
5515 errout:
5516         if (err < 0)
5517                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5518 }
5519
5520 static inline size_t inet6_prefix_nlmsg_size(void)
5521 {
5522         return NLMSG_ALIGN(sizeof(struct prefixmsg))
5523                + nla_total_size(sizeof(struct in6_addr))
5524                + nla_total_size(sizeof(struct prefix_cacheinfo));
5525 }
5526
5527 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5528                              struct prefix_info *pinfo, u32 portid, u32 seq,
5529                              int event, unsigned int flags)
5530 {
5531         struct prefixmsg *pmsg;
5532         struct nlmsghdr *nlh;
5533         struct prefix_cacheinfo ci;
5534
5535         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5536         if (!nlh)
5537                 return -EMSGSIZE;
5538
5539         pmsg = nlmsg_data(nlh);
5540         pmsg->prefix_family = AF_INET6;
5541         pmsg->prefix_pad1 = 0;
5542         pmsg->prefix_pad2 = 0;
5543         pmsg->prefix_ifindex = idev->dev->ifindex;
5544         pmsg->prefix_len = pinfo->prefix_len;
5545         pmsg->prefix_type = pinfo->type;
5546         pmsg->prefix_pad3 = 0;
5547         pmsg->prefix_flags = 0;
5548         if (pinfo->onlink)
5549                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5550         if (pinfo->autoconf)
5551                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5552
5553         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5554                 goto nla_put_failure;
5555         ci.preferred_time = ntohl(pinfo->prefered);
5556         ci.valid_time = ntohl(pinfo->valid);
5557         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5558                 goto nla_put_failure;
5559         nlmsg_end(skb, nlh);
5560         return 0;
5561
5562 nla_put_failure:
5563         nlmsg_cancel(skb, nlh);
5564         return -EMSGSIZE;
5565 }
5566
5567 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5568                          struct prefix_info *pinfo)
5569 {
5570         struct sk_buff *skb;
5571         struct net *net = dev_net(idev->dev);
5572         int err = -ENOBUFS;
5573
5574         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5575         if (!skb)
5576                 goto errout;
5577
5578         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5579         if (err < 0) {
5580                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5581                 WARN_ON(err == -EMSGSIZE);
5582                 kfree_skb(skb);
5583                 goto errout;
5584         }
5585         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5586         return;
5587 errout:
5588         if (err < 0)
5589                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5590 }
5591
5592 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5593 {
5594         struct net *net = dev_net(ifp->idev->dev);
5595
5596         if (event)
5597                 ASSERT_RTNL();
5598
5599         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5600
5601         switch (event) {
5602         case RTM_NEWADDR:
5603                 /*
5604                  * If the address was optimistic
5605                  * we inserted the route at the start of
5606                  * our DAD process, so we don't need
5607                  * to do it again
5608                  */
5609                 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5610                         ip6_ins_rt(ifp->rt);
5611                 if (ifp->idev->cnf.forwarding)
5612                         addrconf_join_anycast(ifp);
5613                 if (!ipv6_addr_any(&ifp->peer_addr))
5614                         addrconf_prefix_route(&ifp->peer_addr, 128,
5615                                               ifp->idev->dev, 0, 0);
5616                 break;
5617         case RTM_DELADDR:
5618                 if (ifp->idev->cnf.forwarding)
5619                         addrconf_leave_anycast(ifp);
5620                 addrconf_leave_solict(ifp->idev, &ifp->addr);
5621                 if (!ipv6_addr_any(&ifp->peer_addr)) {
5622                         struct rt6_info *rt;
5623
5624                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5625                                                        ifp->idev->dev, 0, 0);
5626                         if (rt)
5627                                 ip6_del_rt(rt);
5628                 }
5629                 if (ifp->rt) {
5630                         if (dst_hold_safe(&ifp->rt->dst))
5631                                 ip6_del_rt(ifp->rt);
5632                 }
5633                 rt_genid_bump_ipv6(net);
5634                 break;
5635         }
5636         atomic_inc(&net->ipv6.dev_addr_genid);
5637 }
5638
5639 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5640 {
5641         rcu_read_lock_bh();
5642         if (likely(ifp->idev->dead == 0))
5643                 __ipv6_ifa_notify(event, ifp);
5644         rcu_read_unlock_bh();
5645 }
5646
5647 #ifdef CONFIG_SYSCTL
5648
5649 static
5650 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5651                            void __user *buffer, size_t *lenp, loff_t *ppos)
5652 {
5653         int *valp = ctl->data;
5654         int val = *valp;
5655         loff_t pos = *ppos;
5656         struct ctl_table lctl;
5657         int ret;
5658
5659         /*
5660          * ctl->data points to idev->cnf.forwarding, we should
5661          * not modify it until we get the rtnl lock.
5662          */
5663         lctl = *ctl;
5664         lctl.data = &val;
5665
5666         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5667
5668         if (write)
5669                 ret = addrconf_fixup_forwarding(ctl, valp, val);
5670         if (ret)
5671                 *ppos = pos;
5672         return ret;
5673 }
5674
5675 static
5676 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5677                         void __user *buffer, size_t *lenp, loff_t *ppos)
5678 {
5679         struct inet6_dev *idev = ctl->extra1;
5680         int min_mtu = IPV6_MIN_MTU;
5681         struct ctl_table lctl;
5682
5683         lctl = *ctl;
5684         lctl.extra1 = &min_mtu;
5685         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5686
5687         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5688 }
5689
5690 static void dev_disable_change(struct inet6_dev *idev)
5691 {
5692         struct netdev_notifier_info info;
5693
5694         if (!idev || !idev->dev)
5695                 return;
5696
5697         netdev_notifier_info_init(&info, idev->dev);
5698         if (idev->cnf.disable_ipv6)
5699                 addrconf_notify(NULL, NETDEV_DOWN, &info);
5700         else
5701                 addrconf_notify(NULL, NETDEV_UP, &info);
5702 }
5703
5704 static void addrconf_disable_change(struct net *net, __s32 newf)
5705 {
5706         struct net_device *dev;
5707         struct inet6_dev *idev;
5708
5709         for_each_netdev(net, dev) {
5710                 idev = __in6_dev_get(dev);
5711                 if (idev) {
5712                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5713                         idev->cnf.disable_ipv6 = newf;
5714                         if (changed)
5715                                 dev_disable_change(idev);
5716                 }
5717         }
5718 }
5719
5720 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5721 {
5722         struct net *net;
5723         int old;
5724
5725         if (!rtnl_trylock())
5726                 return restart_syscall();
5727
5728         net = (struct net *)table->extra2;
5729         old = *p;
5730         *p = newf;
5731
5732         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5733                 rtnl_unlock();
5734                 return 0;
5735         }
5736
5737         if (p == &net->ipv6.devconf_all->disable_ipv6) {
5738                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5739                 addrconf_disable_change(net, newf);
5740         } else if ((!newf) ^ (!old))
5741                 dev_disable_change((struct inet6_dev *)table->extra1);
5742
5743         rtnl_unlock();
5744         return 0;
5745 }
5746
5747 static
5748 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5749                             void __user *buffer, size_t *lenp, loff_t *ppos)
5750 {
5751         int *valp = ctl->data;
5752         int val = *valp;
5753         loff_t pos = *ppos;
5754         struct ctl_table lctl;
5755         int ret;
5756
5757         /*
5758          * ctl->data points to idev->cnf.disable_ipv6, we should
5759          * not modify it until we get the rtnl lock.
5760          */
5761         lctl = *ctl;
5762         lctl.data = &val;
5763
5764         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5765
5766         if (write)
5767                 ret = addrconf_disable_ipv6(ctl, valp, val);
5768         if (ret)
5769                 *ppos = pos;
5770         return ret;
5771 }
5772
5773 static
5774 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5775                               void __user *buffer, size_t *lenp, loff_t *ppos)
5776 {
5777         int *valp = ctl->data;
5778         int ret;
5779         int old, new;
5780
5781         old = *valp;
5782         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5783         new = *valp;
5784
5785         if (write && old != new) {
5786                 struct net *net = ctl->extra2;
5787
5788                 if (!rtnl_trylock())
5789                         return restart_syscall();
5790
5791                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5792                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5793                                                      NETCONFA_PROXY_NEIGH,
5794                                                      NETCONFA_IFINDEX_DEFAULT,
5795                                                      net->ipv6.devconf_dflt);
5796                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5797                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5798                                                      NETCONFA_PROXY_NEIGH,
5799                                                      NETCONFA_IFINDEX_ALL,
5800                                                      net->ipv6.devconf_all);
5801                 else {
5802                         struct inet6_dev *idev = ctl->extra1;
5803
5804                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5805                                                      NETCONFA_PROXY_NEIGH,
5806                                                      idev->dev->ifindex,
5807                                                      &idev->cnf);
5808                 }
5809                 rtnl_unlock();
5810         }
5811
5812         return ret;
5813 }
5814
5815 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5816                                          void __user *buffer, size_t *lenp,
5817                                          loff_t *ppos)
5818 {
5819         int ret = 0;
5820         int new_val;
5821         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5822         struct net *net = (struct net *)ctl->extra2;
5823
5824         if (!rtnl_trylock())
5825                 return restart_syscall();
5826
5827         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5828
5829         if (write) {
5830                 new_val = *((int *)ctl->data);
5831
5832                 if (check_addr_gen_mode(new_val) < 0) {
5833                         ret = -EINVAL;
5834                         goto out;
5835                 }
5836
5837                 /* request for default */
5838                 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5839                         ipv6_devconf_dflt.addr_gen_mode = new_val;
5840
5841                 /* request for individual net device */
5842                 } else {
5843                         if (!idev)
5844                                 goto out;
5845
5846                         if (check_stable_privacy(idev, net, new_val) < 0) {
5847                                 ret = -EINVAL;
5848                                 goto out;
5849                         }
5850
5851                         if (idev->cnf.addr_gen_mode != new_val) {
5852                                 idev->cnf.addr_gen_mode = new_val;
5853                                 addrconf_dev_config(idev->dev);
5854                         }
5855                 }
5856         }
5857
5858 out:
5859         rtnl_unlock();
5860
5861         return ret;
5862 }
5863
5864 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5865                                          void __user *buffer, size_t *lenp,
5866                                          loff_t *ppos)
5867 {
5868         int err;
5869         struct in6_addr addr;
5870         char str[IPV6_MAX_STRLEN];
5871         struct ctl_table lctl = *ctl;
5872         struct net *net = ctl->extra2;
5873         struct ipv6_stable_secret *secret = ctl->data;
5874
5875         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5876                 return -EIO;
5877
5878         lctl.maxlen = IPV6_MAX_STRLEN;
5879         lctl.data = str;
5880
5881         if (!rtnl_trylock())
5882                 return restart_syscall();
5883
5884         if (!write && !secret->initialized) {
5885                 err = -EIO;
5886                 goto out;
5887         }
5888
5889         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5890         if (err >= sizeof(str)) {
5891                 err = -EIO;
5892                 goto out;
5893         }
5894
5895         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5896         if (err || !write)
5897                 goto out;
5898
5899         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5900                 err = -EIO;
5901                 goto out;
5902         }
5903
5904         secret->initialized = true;
5905         secret->secret = addr;
5906
5907         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5908                 struct net_device *dev;
5909
5910                 for_each_netdev(net, dev) {
5911                         struct inet6_dev *idev = __in6_dev_get(dev);
5912
5913                         if (idev) {
5914                                 idev->cnf.addr_gen_mode =
5915                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5916                         }
5917                 }
5918         } else {
5919                 struct inet6_dev *idev = ctl->extra1;
5920
5921                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5922         }
5923
5924 out:
5925         rtnl_unlock();
5926
5927         return err;
5928 }
5929
5930 static
5931 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5932                                                 int write,
5933                                                 void __user *buffer,
5934                                                 size_t *lenp,
5935                                                 loff_t *ppos)
5936 {
5937         int *valp = ctl->data;
5938         int val = *valp;
5939         loff_t pos = *ppos;
5940         struct ctl_table lctl;
5941         int ret;
5942
5943         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5944          * we should not modify it until we get the rtnl lock.
5945          */
5946         lctl = *ctl;
5947         lctl.data = &val;
5948
5949         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5950
5951         if (write)
5952                 ret = addrconf_fixup_linkdown(ctl, valp, val);
5953         if (ret)
5954                 *ppos = pos;
5955         return ret;
5956 }
5957
5958 static
5959 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5960 {
5961         if (rt) {
5962                 if (action)
5963                         rt->dst.flags |= DST_NOPOLICY;
5964                 else
5965                         rt->dst.flags &= ~DST_NOPOLICY;
5966         }
5967 }
5968
5969 static
5970 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5971 {
5972         struct inet6_ifaddr *ifa;
5973
5974         read_lock_bh(&idev->lock);
5975         list_for_each_entry(ifa, &idev->addr_list, if_list) {
5976                 spin_lock(&ifa->lock);
5977                 if (ifa->rt) {
5978                         struct rt6_info *rt = ifa->rt;
5979                         int cpu;
5980
5981                         rcu_read_lock();
5982                         addrconf_set_nopolicy(ifa->rt, val);
5983                         if (rt->rt6i_pcpu) {
5984                                 for_each_possible_cpu(cpu) {
5985                                         struct rt6_info **rtp;
5986
5987                                         rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5988                                         addrconf_set_nopolicy(*rtp, val);
5989                                 }
5990                         }
5991                         rcu_read_unlock();
5992                 }
5993                 spin_unlock(&ifa->lock);
5994         }
5995         read_unlock_bh(&idev->lock);
5996 }
5997
5998 static
5999 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6000 {
6001         struct inet6_dev *idev;
6002         struct net *net;
6003
6004         if (!rtnl_trylock())
6005                 return restart_syscall();
6006
6007         *valp = val;
6008
6009         net = (struct net *)ctl->extra2;
6010         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6011                 rtnl_unlock();
6012                 return 0;
6013         }
6014
6015         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6016                 struct net_device *dev;
6017
6018                 for_each_netdev(net, dev) {
6019                         idev = __in6_dev_get(dev);
6020                         if (idev)
6021                                 addrconf_disable_policy_idev(idev, val);
6022                 }
6023         } else {
6024                 idev = (struct inet6_dev *)ctl->extra1;
6025                 addrconf_disable_policy_idev(idev, val);
6026         }
6027
6028         rtnl_unlock();
6029         return 0;
6030 }
6031
6032 static
6033 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6034                                    void __user *buffer, size_t *lenp,
6035                                    loff_t *ppos)
6036 {
6037         int *valp = ctl->data;
6038         int val = *valp;
6039         loff_t pos = *ppos;
6040         struct ctl_table lctl;
6041         int ret;
6042
6043         lctl = *ctl;
6044         lctl.data = &val;
6045         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6046
6047         if (write && (*valp != val))
6048                 ret = addrconf_disable_policy(ctl, valp, val);
6049
6050         if (ret)
6051                 *ppos = pos;
6052
6053         return ret;
6054 }
6055
6056 static int minus_one = -1;
6057 static const int zero = 0;
6058 static const int one = 1;
6059 static const int two_five_five = 255;
6060
6061 static const struct ctl_table addrconf_sysctl[] = {
6062         {
6063                 .procname       = "forwarding",
6064                 .data           = &ipv6_devconf.forwarding,
6065                 .maxlen         = sizeof(int),
6066                 .mode           = 0644,
6067                 .proc_handler   = addrconf_sysctl_forward,
6068         },
6069         {
6070                 .procname       = "hop_limit",
6071                 .data           = &ipv6_devconf.hop_limit,
6072                 .maxlen         = sizeof(int),
6073                 .mode           = 0644,
6074                 .proc_handler   = proc_dointvec_minmax,
6075                 .extra1         = (void *)&one,
6076                 .extra2         = (void *)&two_five_five,
6077         },
6078         {
6079                 .procname       = "mtu",
6080                 .data           = &ipv6_devconf.mtu6,
6081                 .maxlen         = sizeof(int),
6082                 .mode           = 0644,
6083                 .proc_handler   = addrconf_sysctl_mtu,
6084         },
6085         {
6086                 .procname       = "accept_ra",
6087                 .data           = &ipv6_devconf.accept_ra,
6088                 .maxlen         = sizeof(int),
6089                 .mode           = 0644,
6090                 .proc_handler   = proc_dointvec,
6091         },
6092         {
6093                 .procname       = "accept_redirects",
6094                 .data           = &ipv6_devconf.accept_redirects,
6095                 .maxlen         = sizeof(int),
6096                 .mode           = 0644,
6097                 .proc_handler   = proc_dointvec,
6098         },
6099         {
6100                 .procname       = "autoconf",
6101                 .data           = &ipv6_devconf.autoconf,
6102                 .maxlen         = sizeof(int),
6103                 .mode           = 0644,
6104                 .proc_handler   = proc_dointvec,
6105         },
6106         {
6107                 .procname       = "dad_transmits",
6108                 .data           = &ipv6_devconf.dad_transmits,
6109                 .maxlen         = sizeof(int),
6110                 .mode           = 0644,
6111                 .proc_handler   = proc_dointvec,
6112         },
6113         {
6114                 .procname       = "router_solicitations",
6115                 .data           = &ipv6_devconf.rtr_solicits,
6116                 .maxlen         = sizeof(int),
6117                 .mode           = 0644,
6118                 .proc_handler   = proc_dointvec_minmax,
6119                 .extra1         = &minus_one,
6120         },
6121         {
6122                 .procname       = "router_solicitation_interval",
6123                 .data           = &ipv6_devconf.rtr_solicit_interval,
6124                 .maxlen         = sizeof(int),
6125                 .mode           = 0644,
6126                 .proc_handler   = proc_dointvec_jiffies,
6127         },
6128         {
6129                 .procname       = "router_solicitation_max_interval",
6130                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6131                 .maxlen         = sizeof(int),
6132                 .mode           = 0644,
6133                 .proc_handler   = proc_dointvec_jiffies,
6134         },
6135         {
6136                 .procname       = "router_solicitation_delay",
6137                 .data           = &ipv6_devconf.rtr_solicit_delay,
6138                 .maxlen         = sizeof(int),
6139                 .mode           = 0644,
6140                 .proc_handler   = proc_dointvec_jiffies,
6141         },
6142         {
6143                 .procname       = "force_mld_version",
6144                 .data           = &ipv6_devconf.force_mld_version,
6145                 .maxlen         = sizeof(int),
6146                 .mode           = 0644,
6147                 .proc_handler   = proc_dointvec,
6148         },
6149         {
6150                 .procname       = "mldv1_unsolicited_report_interval",
6151                 .data           =
6152                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6153                 .maxlen         = sizeof(int),
6154                 .mode           = 0644,
6155                 .proc_handler   = proc_dointvec_ms_jiffies,
6156         },
6157         {
6158                 .procname       = "mldv2_unsolicited_report_interval",
6159                 .data           =
6160                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6161                 .maxlen         = sizeof(int),
6162                 .mode           = 0644,
6163                 .proc_handler   = proc_dointvec_ms_jiffies,
6164         },
6165         {
6166                 .procname       = "use_tempaddr",
6167                 .data           = &ipv6_devconf.use_tempaddr,
6168                 .maxlen         = sizeof(int),
6169                 .mode           = 0644,
6170                 .proc_handler   = proc_dointvec,
6171         },
6172         {
6173                 .procname       = "temp_valid_lft",
6174                 .data           = &ipv6_devconf.temp_valid_lft,
6175                 .maxlen         = sizeof(int),
6176                 .mode           = 0644,
6177                 .proc_handler   = proc_dointvec,
6178         },
6179         {
6180                 .procname       = "temp_prefered_lft",
6181                 .data           = &ipv6_devconf.temp_prefered_lft,
6182                 .maxlen         = sizeof(int),
6183                 .mode           = 0644,
6184                 .proc_handler   = proc_dointvec,
6185         },
6186         {
6187                 .procname       = "regen_max_retry",
6188                 .data           = &ipv6_devconf.regen_max_retry,
6189                 .maxlen         = sizeof(int),
6190                 .mode           = 0644,
6191                 .proc_handler   = proc_dointvec,
6192         },
6193         {
6194                 .procname       = "max_desync_factor",
6195                 .data           = &ipv6_devconf.max_desync_factor,
6196                 .maxlen         = sizeof(int),
6197                 .mode           = 0644,
6198                 .proc_handler   = proc_dointvec,
6199         },
6200         {
6201                 .procname       = "max_addresses",
6202                 .data           = &ipv6_devconf.max_addresses,
6203                 .maxlen         = sizeof(int),
6204                 .mode           = 0644,
6205                 .proc_handler   = proc_dointvec,
6206         },
6207         {
6208                 .procname       = "accept_ra_defrtr",
6209                 .data           = &ipv6_devconf.accept_ra_defrtr,
6210                 .maxlen         = sizeof(int),
6211                 .mode           = 0644,
6212                 .proc_handler   = proc_dointvec,
6213         },
6214         {
6215                 .procname       = "accept_ra_min_hop_limit",
6216                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6217                 .maxlen         = sizeof(int),
6218                 .mode           = 0644,
6219                 .proc_handler   = proc_dointvec,
6220         },
6221         {
6222                 .procname       = "accept_ra_pinfo",
6223                 .data           = &ipv6_devconf.accept_ra_pinfo,
6224                 .maxlen         = sizeof(int),
6225                 .mode           = 0644,
6226                 .proc_handler   = proc_dointvec,
6227         },
6228 #ifdef CONFIG_IPV6_ROUTER_PREF
6229         {
6230                 .procname       = "accept_ra_rtr_pref",
6231                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6232                 .maxlen         = sizeof(int),
6233                 .mode           = 0644,
6234                 .proc_handler   = proc_dointvec,
6235         },
6236         {
6237                 .procname       = "router_probe_interval",
6238                 .data           = &ipv6_devconf.rtr_probe_interval,
6239                 .maxlen         = sizeof(int),
6240                 .mode           = 0644,
6241                 .proc_handler   = proc_dointvec_jiffies,
6242         },
6243 #ifdef CONFIG_IPV6_ROUTE_INFO
6244         {
6245                 .procname       = "accept_ra_rt_info_min_plen",
6246                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6247                 .maxlen         = sizeof(int),
6248                 .mode           = 0644,
6249                 .proc_handler   = proc_dointvec,
6250         },
6251         {
6252                 .procname       = "accept_ra_rt_info_max_plen",
6253                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6254                 .maxlen         = sizeof(int),
6255                 .mode           = 0644,
6256                 .proc_handler   = proc_dointvec,
6257         },
6258 #endif
6259 #endif
6260         {
6261                 .procname       = "proxy_ndp",
6262                 .data           = &ipv6_devconf.proxy_ndp,
6263                 .maxlen         = sizeof(int),
6264                 .mode           = 0644,
6265                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6266         },
6267         {
6268                 .procname       = "accept_source_route",
6269                 .data           = &ipv6_devconf.accept_source_route,
6270                 .maxlen         = sizeof(int),
6271                 .mode           = 0644,
6272                 .proc_handler   = proc_dointvec,
6273         },
6274 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6275         {
6276                 .procname       = "optimistic_dad",
6277                 .data           = &ipv6_devconf.optimistic_dad,
6278                 .maxlen         = sizeof(int),
6279                 .mode           = 0644,
6280                 .proc_handler   = proc_dointvec,
6281         },
6282         {
6283                 .procname       = "use_optimistic",
6284                 .data           = &ipv6_devconf.use_optimistic,
6285                 .maxlen         = sizeof(int),
6286                 .mode           = 0644,
6287                 .proc_handler   = proc_dointvec,
6288         },
6289 #endif
6290 #ifdef CONFIG_IPV6_MROUTE
6291         {
6292                 .procname       = "mc_forwarding",
6293                 .data           = &ipv6_devconf.mc_forwarding,
6294                 .maxlen         = sizeof(int),
6295                 .mode           = 0444,
6296                 .proc_handler   = proc_dointvec,
6297         },
6298 #endif
6299         {
6300                 .procname       = "disable_ipv6",
6301                 .data           = &ipv6_devconf.disable_ipv6,
6302                 .maxlen         = sizeof(int),
6303                 .mode           = 0644,
6304                 .proc_handler   = addrconf_sysctl_disable,
6305         },
6306         {
6307                 .procname       = "accept_dad",
6308                 .data           = &ipv6_devconf.accept_dad,
6309                 .maxlen         = sizeof(int),
6310                 .mode           = 0644,
6311                 .proc_handler   = proc_dointvec,
6312         },
6313         {
6314                 .procname       = "force_tllao",
6315                 .data           = &ipv6_devconf.force_tllao,
6316                 .maxlen         = sizeof(int),
6317                 .mode           = 0644,
6318                 .proc_handler   = proc_dointvec
6319         },
6320         {
6321                 .procname       = "ndisc_notify",
6322                 .data           = &ipv6_devconf.ndisc_notify,
6323                 .maxlen         = sizeof(int),
6324                 .mode           = 0644,
6325                 .proc_handler   = proc_dointvec
6326         },
6327         {
6328                 .procname       = "suppress_frag_ndisc",
6329                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6330                 .maxlen         = sizeof(int),
6331                 .mode           = 0644,
6332                 .proc_handler   = proc_dointvec
6333         },
6334         {
6335                 .procname       = "accept_ra_from_local",
6336                 .data           = &ipv6_devconf.accept_ra_from_local,
6337                 .maxlen         = sizeof(int),
6338                 .mode           = 0644,
6339                 .proc_handler   = proc_dointvec,
6340         },
6341         {
6342                 .procname       = "accept_ra_mtu",
6343                 .data           = &ipv6_devconf.accept_ra_mtu,
6344                 .maxlen         = sizeof(int),
6345                 .mode           = 0644,
6346                 .proc_handler   = proc_dointvec,
6347         },
6348         {
6349                 .procname       = "stable_secret",
6350                 .data           = &ipv6_devconf.stable_secret,
6351                 .maxlen         = IPV6_MAX_STRLEN,
6352                 .mode           = 0600,
6353                 .proc_handler   = addrconf_sysctl_stable_secret,
6354         },
6355         {
6356                 .procname       = "use_oif_addrs_only",
6357                 .data           = &ipv6_devconf.use_oif_addrs_only,
6358                 .maxlen         = sizeof(int),
6359                 .mode           = 0644,
6360                 .proc_handler   = proc_dointvec,
6361         },
6362         {
6363                 .procname       = "ignore_routes_with_linkdown",
6364                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6365                 .maxlen         = sizeof(int),
6366                 .mode           = 0644,
6367                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6368         },
6369         {
6370                 .procname       = "drop_unicast_in_l2_multicast",
6371                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6372                 .maxlen         = sizeof(int),
6373                 .mode           = 0644,
6374                 .proc_handler   = proc_dointvec,
6375         },
6376         {
6377                 .procname       = "drop_unsolicited_na",
6378                 .data           = &ipv6_devconf.drop_unsolicited_na,
6379                 .maxlen         = sizeof(int),
6380                 .mode           = 0644,
6381                 .proc_handler   = proc_dointvec,
6382         },
6383         {
6384                 .procname       = "keep_addr_on_down",
6385                 .data           = &ipv6_devconf.keep_addr_on_down,
6386                 .maxlen         = sizeof(int),
6387                 .mode           = 0644,
6388                 .proc_handler   = proc_dointvec,
6389
6390         },
6391         {
6392                 .procname       = "seg6_enabled",
6393                 .data           = &ipv6_devconf.seg6_enabled,
6394                 .maxlen         = sizeof(int),
6395                 .mode           = 0644,
6396                 .proc_handler   = proc_dointvec,
6397         },
6398 #ifdef CONFIG_IPV6_SEG6_HMAC
6399         {
6400                 .procname       = "seg6_require_hmac",
6401                 .data           = &ipv6_devconf.seg6_require_hmac,
6402                 .maxlen         = sizeof(int),
6403                 .mode           = 0644,
6404                 .proc_handler   = proc_dointvec,
6405         },
6406 #endif
6407         {
6408                 .procname       = "enhanced_dad",
6409                 .data           = &ipv6_devconf.enhanced_dad,
6410                 .maxlen         = sizeof(int),
6411                 .mode           = 0644,
6412                 .proc_handler   = proc_dointvec,
6413         },
6414         {
6415                 .procname               = "addr_gen_mode",
6416                 .data                   = &ipv6_devconf.addr_gen_mode,
6417                 .maxlen                 = sizeof(int),
6418                 .mode                   = 0644,
6419                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6420         },
6421         {
6422                 .procname       = "disable_policy",
6423                 .data           = &ipv6_devconf.disable_policy,
6424                 .maxlen         = sizeof(int),
6425                 .mode           = 0644,
6426                 .proc_handler   = addrconf_sysctl_disable_policy,
6427         },
6428         {
6429                 .procname       = "ndisc_tclass",
6430                 .data           = &ipv6_devconf.ndisc_tclass,
6431                 .maxlen         = sizeof(int),
6432                 .mode           = 0644,
6433                 .proc_handler   = proc_dointvec_minmax,
6434                 .extra1         = (void *)&zero,
6435                 .extra2         = (void *)&two_five_five,
6436         },
6437         {
6438                 /* sentinel */
6439         }
6440 };
6441
6442 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6443                 struct inet6_dev *idev, struct ipv6_devconf *p)
6444 {
6445         int i, ifindex;
6446         struct ctl_table *table;
6447         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6448
6449         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6450         if (!table)
6451                 goto out;
6452
6453         for (i = 0; table[i].data; i++) {
6454                 table[i].data += (char *)p - (char *)&ipv6_devconf;
6455                 /* If one of these is already set, then it is not safe to
6456                  * overwrite either of them: this makes proc_dointvec_minmax
6457                  * usable.
6458                  */
6459                 if (!table[i].extra1 && !table[i].extra2) {
6460                         table[i].extra1 = idev; /* embedded; no ref */
6461                         table[i].extra2 = net;
6462                 }
6463         }
6464
6465         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6466
6467         p->sysctl_header = register_net_sysctl(net, path, table);
6468         if (!p->sysctl_header)
6469                 goto free;
6470
6471         if (!strcmp(dev_name, "all"))
6472                 ifindex = NETCONFA_IFINDEX_ALL;
6473         else if (!strcmp(dev_name, "default"))
6474                 ifindex = NETCONFA_IFINDEX_DEFAULT;
6475         else
6476                 ifindex = idev->dev->ifindex;
6477         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6478                                      ifindex, p);
6479         return 0;
6480
6481 free:
6482         kfree(table);
6483 out:
6484         return -ENOBUFS;
6485 }
6486
6487 static void __addrconf_sysctl_unregister(struct net *net,
6488                                          struct ipv6_devconf *p, int ifindex)
6489 {
6490         struct ctl_table *table;
6491
6492         if (!p->sysctl_header)
6493                 return;
6494
6495         table = p->sysctl_header->ctl_table_arg;
6496         unregister_net_sysctl_table(p->sysctl_header);
6497         p->sysctl_header = NULL;
6498         kfree(table);
6499
6500         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6501 }
6502
6503 static int addrconf_sysctl_register(struct inet6_dev *idev)
6504 {
6505         int err;
6506
6507         if (!sysctl_dev_name_is_allowed(idev->dev->name))
6508                 return -EINVAL;
6509
6510         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6511                                     &ndisc_ifinfo_sysctl_change);
6512         if (err)
6513                 return err;
6514         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6515                                          idev, &idev->cnf);
6516         if (err)
6517                 neigh_sysctl_unregister(idev->nd_parms);
6518
6519         return err;
6520 }
6521
6522 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6523 {
6524         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6525                                      idev->dev->ifindex);
6526         neigh_sysctl_unregister(idev->nd_parms);
6527 }
6528
6529
6530 #endif
6531
6532 static int __net_init addrconf_init_net(struct net *net)
6533 {
6534         int err = -ENOMEM;
6535         struct ipv6_devconf *all, *dflt;
6536
6537         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6538         if (!all)
6539                 goto err_alloc_all;
6540
6541         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6542         if (!dflt)
6543                 goto err_alloc_dflt;
6544
6545         /* these will be inherited by all namespaces */
6546         dflt->autoconf = ipv6_defaults.autoconf;
6547         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6548
6549         dflt->stable_secret.initialized = false;
6550         all->stable_secret.initialized = false;
6551
6552         net->ipv6.devconf_all = all;
6553         net->ipv6.devconf_dflt = dflt;
6554
6555 #ifdef CONFIG_SYSCTL
6556         err = __addrconf_sysctl_register(net, "all", NULL, all);
6557         if (err < 0)
6558                 goto err_reg_all;
6559
6560         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6561         if (err < 0)
6562                 goto err_reg_dflt;
6563 #endif
6564         return 0;
6565
6566 #ifdef CONFIG_SYSCTL
6567 err_reg_dflt:
6568         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6569 err_reg_all:
6570         kfree(dflt);
6571 #endif
6572 err_alloc_dflt:
6573         kfree(all);
6574 err_alloc_all:
6575         return err;
6576 }
6577
6578 static void __net_exit addrconf_exit_net(struct net *net)
6579 {
6580 #ifdef CONFIG_SYSCTL
6581         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6582                                      NETCONFA_IFINDEX_DEFAULT);
6583         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6584                                      NETCONFA_IFINDEX_ALL);
6585 #endif
6586         kfree(net->ipv6.devconf_dflt);
6587         kfree(net->ipv6.devconf_all);
6588 }
6589
6590 static struct pernet_operations addrconf_ops = {
6591         .init = addrconf_init_net,
6592         .exit = addrconf_exit_net,
6593 };
6594
6595 static struct rtnl_af_ops inet6_ops __read_mostly = {
6596         .family           = AF_INET6,
6597         .fill_link_af     = inet6_fill_link_af,
6598         .get_link_af_size = inet6_get_link_af_size,
6599         .validate_link_af = inet6_validate_link_af,
6600         .set_link_af      = inet6_set_link_af,
6601 };
6602
6603 /*
6604  *      Init / cleanup code
6605  */
6606
6607 int __init addrconf_init(void)
6608 {
6609         struct inet6_dev *idev;
6610         int i, err;
6611
6612         err = ipv6_addr_label_init();
6613         if (err < 0) {
6614                 pr_crit("%s: cannot initialize default policy table: %d\n",
6615                         __func__, err);
6616                 goto out;
6617         }
6618
6619         err = register_pernet_subsys(&addrconf_ops);
6620         if (err < 0)
6621                 goto out_addrlabel;
6622
6623         addrconf_wq = create_workqueue("ipv6_addrconf");
6624         if (!addrconf_wq) {
6625                 err = -ENOMEM;
6626                 goto out_nowq;
6627         }
6628
6629         /* The addrconf netdev notifier requires that loopback_dev
6630          * has it's ipv6 private information allocated and setup
6631          * before it can bring up and give link-local addresses
6632          * to other devices which are up.
6633          *
6634          * Unfortunately, loopback_dev is not necessarily the first
6635          * entry in the global dev_base list of net devices.  In fact,
6636          * it is likely to be the very last entry on that list.
6637          * So this causes the notifier registry below to try and
6638          * give link-local addresses to all devices besides loopback_dev
6639          * first, then loopback_dev, which cases all the non-loopback_dev
6640          * devices to fail to get a link-local address.
6641          *
6642          * So, as a temporary fix, allocate the ipv6 structure for
6643          * loopback_dev first by hand.
6644          * Longer term, all of the dependencies ipv6 has upon the loopback
6645          * device and it being up should be removed.
6646          */
6647         rtnl_lock();
6648         idev = ipv6_add_dev(init_net.loopback_dev);
6649         rtnl_unlock();
6650         if (IS_ERR(idev)) {
6651                 err = PTR_ERR(idev);
6652                 goto errlo;
6653         }
6654
6655         ip6_route_init_special_entries();
6656
6657         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6658                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6659
6660         register_netdevice_notifier(&ipv6_dev_notf);
6661
6662         addrconf_verify();
6663
6664         rtnl_af_register(&inet6_ops);
6665
6666         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6667                                    NULL, inet6_dump_ifinfo, 0);
6668         if (err < 0)
6669                 goto errout;
6670
6671         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6672                                    inet6_rtm_newaddr, NULL, 0);
6673         if (err < 0)
6674                 goto errout;
6675         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6676                                    inet6_rtm_deladdr, NULL, 0);
6677         if (err < 0)
6678                 goto errout;
6679         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6680                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
6681                                    RTNL_FLAG_DOIT_UNLOCKED);
6682         if (err < 0)
6683                 goto errout;
6684         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6685                                    NULL, inet6_dump_ifmcaddr, 0);
6686         if (err < 0)
6687                 goto errout;
6688         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6689                                    NULL, inet6_dump_ifacaddr, 0);
6690         if (err < 0)
6691                 goto errout;
6692         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6693                                    inet6_netconf_get_devconf,
6694                                    inet6_netconf_dump_devconf,
6695                                    RTNL_FLAG_DOIT_UNLOCKED);
6696         if (err < 0)
6697                 goto errout;
6698         err = ipv6_addr_label_rtnl_register();
6699         if (err < 0)
6700                 goto errout;
6701
6702         return 0;
6703 errout:
6704         rtnl_unregister_all(PF_INET6);
6705         rtnl_af_unregister(&inet6_ops);
6706         unregister_netdevice_notifier(&ipv6_dev_notf);
6707 errlo:
6708         destroy_workqueue(addrconf_wq);
6709 out_nowq:
6710         unregister_pernet_subsys(&addrconf_ops);
6711 out_addrlabel:
6712         ipv6_addr_label_cleanup();
6713 out:
6714         return err;
6715 }
6716
6717 void addrconf_cleanup(void)
6718 {
6719         struct net_device *dev;
6720         int i;
6721
6722         unregister_netdevice_notifier(&ipv6_dev_notf);
6723         unregister_pernet_subsys(&addrconf_ops);
6724         ipv6_addr_label_cleanup();
6725
6726         rtnl_af_unregister(&inet6_ops);
6727
6728         rtnl_lock();
6729
6730         /* clean dev list */
6731         for_each_netdev(&init_net, dev) {
6732                 if (__in6_dev_get(dev) == NULL)
6733                         continue;
6734                 addrconf_ifdown(dev, 1);
6735         }
6736         addrconf_ifdown(init_net.loopback_dev, 2);
6737
6738         /*
6739          *      Check hash table.
6740          */
6741         spin_lock_bh(&addrconf_hash_lock);
6742         for (i = 0; i < IN6_ADDR_HSIZE; i++)
6743                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6744         spin_unlock_bh(&addrconf_hash_lock);
6745         cancel_delayed_work(&addr_chk_work);
6746         rtnl_unlock();
6747
6748         destroy_workqueue(addrconf_wq);
6749 }