c63ccce6425fb221170b058e701c11b7ba9a497e
[muen/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 fib6_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
389         ndev->cnf.mtu6 = dev->mtu;
390         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
391         if (!ndev->nd_parms) {
392                 kfree(ndev);
393                 return ERR_PTR(err);
394         }
395         if (ndev->cnf.forwarding)
396                 dev_disable_lro(dev);
397         /* We refer to the device */
398         dev_hold(dev);
399
400         if (snmp6_alloc_dev(ndev) < 0) {
401                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
402                            __func__);
403                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
404                 dev_put(dev);
405                 kfree(ndev);
406                 return ERR_PTR(err);
407         }
408
409         if (snmp6_register_dev(ndev) < 0) {
410                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
411                            __func__, dev->name);
412                 goto err_release;
413         }
414
415         /* One reference from device. */
416         refcount_set(&ndev->refcnt, 1);
417
418         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
419                 ndev->cnf.accept_dad = -1;
420
421 #if IS_ENABLED(CONFIG_IPV6_SIT)
422         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
423                 pr_info("%s: Disabled Multicast RS\n", dev->name);
424                 ndev->cnf.rtr_solicits = 0;
425         }
426 #endif
427
428         INIT_LIST_HEAD(&ndev->tempaddr_list);
429         ndev->desync_factor = U32_MAX;
430         if ((dev->flags&IFF_LOOPBACK) ||
431             dev->type == ARPHRD_TUNNEL ||
432             dev->type == ARPHRD_TUNNEL6 ||
433             dev->type == ARPHRD_SIT ||
434             dev->type == ARPHRD_NONE) {
435                 ndev->cnf.use_tempaddr = -1;
436         } else
437                 ipv6_regen_rndid(ndev);
438
439         ndev->token = in6addr_any;
440
441         if (netif_running(dev) && addrconf_link_ready(dev))
442                 ndev->if_flags |= IF_READY;
443
444         ipv6_mc_init_dev(ndev);
445         ndev->tstamp = jiffies;
446         err = addrconf_sysctl_register(ndev);
447         if (err) {
448                 ipv6_mc_destroy_dev(ndev);
449                 snmp6_unregister_dev(ndev);
450                 goto err_release;
451         }
452         /* protected by rtnl_lock */
453         rcu_assign_pointer(dev->ip6_ptr, ndev);
454
455         /* Join interface-local all-node multicast group */
456         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
457
458         /* Join all-node multicast group */
459         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
460
461         /* Join all-router multicast group if forwarding is set */
462         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
463                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
464
465         return ndev;
466
467 err_release:
468         neigh_parms_release(&nd_tbl, ndev->nd_parms);
469         ndev->dead = 1;
470         in6_dev_finish_destroy(ndev);
471         return ERR_PTR(err);
472 }
473
474 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
475 {
476         struct inet6_dev *idev;
477
478         ASSERT_RTNL();
479
480         idev = __in6_dev_get(dev);
481         if (!idev) {
482                 idev = ipv6_add_dev(dev);
483                 if (IS_ERR(idev))
484                         return NULL;
485         }
486
487         if (dev->flags&IFF_UP)
488                 ipv6_mc_up(idev);
489         return idev;
490 }
491
492 static int inet6_netconf_msgsize_devconf(int type)
493 {
494         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
495                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
496         bool all = false;
497
498         if (type == NETCONFA_ALL)
499                 all = true;
500
501         if (all || type == NETCONFA_FORWARDING)
502                 size += nla_total_size(4);
503 #ifdef CONFIG_IPV6_MROUTE
504         if (all || type == NETCONFA_MC_FORWARDING)
505                 size += nla_total_size(4);
506 #endif
507         if (all || type == NETCONFA_PROXY_NEIGH)
508                 size += nla_total_size(4);
509
510         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
511                 size += nla_total_size(4);
512
513         return size;
514 }
515
516 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
517                                       struct ipv6_devconf *devconf, u32 portid,
518                                       u32 seq, int event, unsigned int flags,
519                                       int type)
520 {
521         struct nlmsghdr  *nlh;
522         struct netconfmsg *ncm;
523         bool all = false;
524
525         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
526                         flags);
527         if (!nlh)
528                 return -EMSGSIZE;
529
530         if (type == NETCONFA_ALL)
531                 all = true;
532
533         ncm = nlmsg_data(nlh);
534         ncm->ncm_family = AF_INET6;
535
536         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
537                 goto nla_put_failure;
538
539         if (!devconf)
540                 goto out;
541
542         if ((all || type == NETCONFA_FORWARDING) &&
543             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
544                 goto nla_put_failure;
545 #ifdef CONFIG_IPV6_MROUTE
546         if ((all || type == NETCONFA_MC_FORWARDING) &&
547             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
548                         devconf->mc_forwarding) < 0)
549                 goto nla_put_failure;
550 #endif
551         if ((all || type == NETCONFA_PROXY_NEIGH) &&
552             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
553                 goto nla_put_failure;
554
555         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
556             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
557                         devconf->ignore_routes_with_linkdown) < 0)
558                 goto nla_put_failure;
559
560 out:
561         nlmsg_end(skb, nlh);
562         return 0;
563
564 nla_put_failure:
565         nlmsg_cancel(skb, nlh);
566         return -EMSGSIZE;
567 }
568
569 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
570                                   int ifindex, struct ipv6_devconf *devconf)
571 {
572         struct sk_buff *skb;
573         int err = -ENOBUFS;
574
575         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
576         if (!skb)
577                 goto errout;
578
579         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
580                                          event, 0, type);
581         if (err < 0) {
582                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
583                 WARN_ON(err == -EMSGSIZE);
584                 kfree_skb(skb);
585                 goto errout;
586         }
587         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
588         return;
589 errout:
590         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
591 }
592
593 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
594         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
595         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
596         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
597         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
598 };
599
600 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
601                                      struct nlmsghdr *nlh,
602                                      struct netlink_ext_ack *extack)
603 {
604         struct net *net = sock_net(in_skb->sk);
605         struct nlattr *tb[NETCONFA_MAX+1];
606         struct inet6_dev *in6_dev = NULL;
607         struct net_device *dev = NULL;
608         struct netconfmsg *ncm;
609         struct sk_buff *skb;
610         struct ipv6_devconf *devconf;
611         int ifindex;
612         int err;
613
614         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
615                           devconf_ipv6_policy, extack);
616         if (err < 0)
617                 return err;
618
619         if (!tb[NETCONFA_IFINDEX])
620                 return -EINVAL;
621
622         err = -EINVAL;
623         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
624         switch (ifindex) {
625         case NETCONFA_IFINDEX_ALL:
626                 devconf = net->ipv6.devconf_all;
627                 break;
628         case NETCONFA_IFINDEX_DEFAULT:
629                 devconf = net->ipv6.devconf_dflt;
630                 break;
631         default:
632                 dev = dev_get_by_index(net, ifindex);
633                 if (!dev)
634                         return -EINVAL;
635                 in6_dev = in6_dev_get(dev);
636                 if (!in6_dev)
637                         goto errout;
638                 devconf = &in6_dev->cnf;
639                 break;
640         }
641
642         err = -ENOBUFS;
643         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
644         if (!skb)
645                 goto errout;
646
647         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
648                                          NETLINK_CB(in_skb).portid,
649                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
650                                          NETCONFA_ALL);
651         if (err < 0) {
652                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
653                 WARN_ON(err == -EMSGSIZE);
654                 kfree_skb(skb);
655                 goto errout;
656         }
657         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
658 errout:
659         if (in6_dev)
660                 in6_dev_put(in6_dev);
661         if (dev)
662                 dev_put(dev);
663         return err;
664 }
665
666 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
667                                       struct netlink_callback *cb)
668 {
669         struct net *net = sock_net(skb->sk);
670         int h, s_h;
671         int idx, s_idx;
672         struct net_device *dev;
673         struct inet6_dev *idev;
674         struct hlist_head *head;
675
676         s_h = cb->args[0];
677         s_idx = idx = cb->args[1];
678
679         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
680                 idx = 0;
681                 head = &net->dev_index_head[h];
682                 rcu_read_lock();
683                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
684                           net->dev_base_seq;
685                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
686                         if (idx < s_idx)
687                                 goto cont;
688                         idev = __in6_dev_get(dev);
689                         if (!idev)
690                                 goto cont;
691
692                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
693                                                        &idev->cnf,
694                                                        NETLINK_CB(cb->skb).portid,
695                                                        cb->nlh->nlmsg_seq,
696                                                        RTM_NEWNETCONF,
697                                                        NLM_F_MULTI,
698                                                        NETCONFA_ALL) < 0) {
699                                 rcu_read_unlock();
700                                 goto done;
701                         }
702                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
703 cont:
704                         idx++;
705                 }
706                 rcu_read_unlock();
707         }
708         if (h == NETDEV_HASHENTRIES) {
709                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
710                                                net->ipv6.devconf_all,
711                                                NETLINK_CB(cb->skb).portid,
712                                                cb->nlh->nlmsg_seq,
713                                                RTM_NEWNETCONF, NLM_F_MULTI,
714                                                NETCONFA_ALL) < 0)
715                         goto done;
716                 else
717                         h++;
718         }
719         if (h == NETDEV_HASHENTRIES + 1) {
720                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
721                                                net->ipv6.devconf_dflt,
722                                                NETLINK_CB(cb->skb).portid,
723                                                cb->nlh->nlmsg_seq,
724                                                RTM_NEWNETCONF, NLM_F_MULTI,
725                                                NETCONFA_ALL) < 0)
726                         goto done;
727                 else
728                         h++;
729         }
730 done:
731         cb->args[0] = h;
732         cb->args[1] = idx;
733
734         return skb->len;
735 }
736
737 #ifdef CONFIG_SYSCTL
738 static void dev_forward_change(struct inet6_dev *idev)
739 {
740         struct net_device *dev;
741         struct inet6_ifaddr *ifa;
742
743         if (!idev)
744                 return;
745         dev = idev->dev;
746         if (idev->cnf.forwarding)
747                 dev_disable_lro(dev);
748         if (dev->flags & IFF_MULTICAST) {
749                 if (idev->cnf.forwarding) {
750                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
751                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
752                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
753                 } else {
754                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
755                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
756                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
757                 }
758         }
759
760         list_for_each_entry(ifa, &idev->addr_list, if_list) {
761                 if (ifa->flags&IFA_F_TENTATIVE)
762                         continue;
763                 if (idev->cnf.forwarding)
764                         addrconf_join_anycast(ifa);
765                 else
766                         addrconf_leave_anycast(ifa);
767         }
768         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
769                                      NETCONFA_FORWARDING,
770                                      dev->ifindex, &idev->cnf);
771 }
772
773
774 static void addrconf_forward_change(struct net *net, __s32 newf)
775 {
776         struct net_device *dev;
777         struct inet6_dev *idev;
778
779         for_each_netdev(net, dev) {
780                 idev = __in6_dev_get(dev);
781                 if (idev) {
782                         int changed = (!idev->cnf.forwarding) ^ (!newf);
783                         idev->cnf.forwarding = newf;
784                         if (changed)
785                                 dev_forward_change(idev);
786                 }
787         }
788 }
789
790 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
791 {
792         struct net *net;
793         int old;
794
795         if (!rtnl_trylock())
796                 return restart_syscall();
797
798         net = (struct net *)table->extra2;
799         old = *p;
800         *p = newf;
801
802         if (p == &net->ipv6.devconf_dflt->forwarding) {
803                 if ((!newf) ^ (!old))
804                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
805                                                      NETCONFA_FORWARDING,
806                                                      NETCONFA_IFINDEX_DEFAULT,
807                                                      net->ipv6.devconf_dflt);
808                 rtnl_unlock();
809                 return 0;
810         }
811
812         if (p == &net->ipv6.devconf_all->forwarding) {
813                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
814
815                 net->ipv6.devconf_dflt->forwarding = newf;
816                 if ((!newf) ^ (!old_dflt))
817                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
818                                                      NETCONFA_FORWARDING,
819                                                      NETCONFA_IFINDEX_DEFAULT,
820                                                      net->ipv6.devconf_dflt);
821
822                 addrconf_forward_change(net, newf);
823                 if ((!newf) ^ (!old))
824                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
825                                                      NETCONFA_FORWARDING,
826                                                      NETCONFA_IFINDEX_ALL,
827                                                      net->ipv6.devconf_all);
828         } else if ((!newf) ^ (!old))
829                 dev_forward_change((struct inet6_dev *)table->extra1);
830         rtnl_unlock();
831
832         if (newf)
833                 rt6_purge_dflt_routers(net);
834         return 1;
835 }
836
837 static void addrconf_linkdown_change(struct net *net, __s32 newf)
838 {
839         struct net_device *dev;
840         struct inet6_dev *idev;
841
842         for_each_netdev(net, dev) {
843                 idev = __in6_dev_get(dev);
844                 if (idev) {
845                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
846
847                         idev->cnf.ignore_routes_with_linkdown = newf;
848                         if (changed)
849                                 inet6_netconf_notify_devconf(dev_net(dev),
850                                                              RTM_NEWNETCONF,
851                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
852                                                              dev->ifindex,
853                                                              &idev->cnf);
854                 }
855         }
856 }
857
858 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
859 {
860         struct net *net;
861         int old;
862
863         if (!rtnl_trylock())
864                 return restart_syscall();
865
866         net = (struct net *)table->extra2;
867         old = *p;
868         *p = newf;
869
870         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
871                 if ((!newf) ^ (!old))
872                         inet6_netconf_notify_devconf(net,
873                                                      RTM_NEWNETCONF,
874                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
875                                                      NETCONFA_IFINDEX_DEFAULT,
876                                                      net->ipv6.devconf_dflt);
877                 rtnl_unlock();
878                 return 0;
879         }
880
881         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
882                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
883                 addrconf_linkdown_change(net, newf);
884                 if ((!newf) ^ (!old))
885                         inet6_netconf_notify_devconf(net,
886                                                      RTM_NEWNETCONF,
887                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
888                                                      NETCONFA_IFINDEX_ALL,
889                                                      net->ipv6.devconf_all);
890         }
891         rtnl_unlock();
892
893         return 1;
894 }
895
896 #endif
897
898 /* Nobody refers to this ifaddr, destroy it */
899 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
900 {
901         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
902
903 #ifdef NET_REFCNT_DEBUG
904         pr_debug("%s\n", __func__);
905 #endif
906
907         in6_dev_put(ifp->idev);
908
909         if (cancel_delayed_work(&ifp->dad_work))
910                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
911                           ifp);
912
913         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
914                 pr_warn("Freeing alive inet6 address %p\n", ifp);
915                 return;
916         }
917
918         kfree_rcu(ifp, rcu);
919 }
920
921 static void
922 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
923 {
924         struct list_head *p;
925         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
926
927         /*
928          * Each device address list is sorted in order of scope -
929          * global before linklocal.
930          */
931         list_for_each(p, &idev->addr_list) {
932                 struct inet6_ifaddr *ifa
933                         = list_entry(p, struct inet6_ifaddr, if_list);
934                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
935                         break;
936         }
937
938         list_add_tail_rcu(&ifp->if_list, p);
939 }
940
941 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
942 {
943         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
944
945         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
946 }
947
948 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
949                                struct net_device *dev, unsigned int hash)
950 {
951         struct inet6_ifaddr *ifp;
952
953         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
954                 if (!net_eq(dev_net(ifp->idev->dev), net))
955                         continue;
956                 if (ipv6_addr_equal(&ifp->addr, addr)) {
957                         if (!dev || ifp->idev->dev == dev)
958                                 return true;
959                 }
960         }
961         return false;
962 }
963
964 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
965 {
966         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
967         int err = 0;
968
969         spin_lock(&addrconf_hash_lock);
970
971         /* Ignore adding duplicate addresses on an interface */
972         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
973                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
974                 err = -EEXIST;
975         } else {
976                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
977         }
978
979         spin_unlock(&addrconf_hash_lock);
980
981         return err;
982 }
983
984 /* On success it returns ifp with increased reference count */
985
986 static struct inet6_ifaddr *
987 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
988               bool can_block, struct netlink_ext_ack *extack)
989 {
990         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
991         int addr_type = ipv6_addr_type(cfg->pfx);
992         struct net *net = dev_net(idev->dev);
993         struct inet6_ifaddr *ifa = NULL;
994         struct fib6_info *f6i = NULL;
995         int err = 0;
996
997         if (addr_type == IPV6_ADDR_ANY ||
998             addr_type & IPV6_ADDR_MULTICAST ||
999             (!(idev->dev->flags & IFF_LOOPBACK) &&
1000              addr_type & IPV6_ADDR_LOOPBACK))
1001                 return ERR_PTR(-EADDRNOTAVAIL);
1002
1003         if (idev->dead) {
1004                 err = -ENODEV;                  /*XXX*/
1005                 goto out;
1006         }
1007
1008         if (idev->cnf.disable_ipv6) {
1009                 err = -EACCES;
1010                 goto out;
1011         }
1012
1013         /* validator notifier needs to be blocking;
1014          * do not call in atomic context
1015          */
1016         if (can_block) {
1017                 struct in6_validator_info i6vi = {
1018                         .i6vi_addr = *cfg->pfx,
1019                         .i6vi_dev = idev,
1020                         .extack = extack,
1021                 };
1022
1023                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1024                 err = notifier_to_errno(err);
1025                 if (err < 0)
1026                         goto out;
1027         }
1028
1029         ifa = kzalloc(sizeof(*ifa), gfp_flags);
1030         if (!ifa) {
1031                 err = -ENOBUFS;
1032                 goto out;
1033         }
1034
1035         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1036         if (IS_ERR(f6i)) {
1037                 err = PTR_ERR(f6i);
1038                 f6i = NULL;
1039                 goto out;
1040         }
1041
1042         if (net->ipv6.devconf_all->disable_policy ||
1043             idev->cnf.disable_policy)
1044                 f6i->dst_nopolicy = true;
1045
1046         neigh_parms_data_state_setall(idev->nd_parms);
1047
1048         ifa->addr = *cfg->pfx;
1049         if (cfg->peer_pfx)
1050                 ifa->peer_addr = *cfg->peer_pfx;
1051
1052         spin_lock_init(&ifa->lock);
1053         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1054         INIT_HLIST_NODE(&ifa->addr_lst);
1055         ifa->scope = cfg->scope;
1056         ifa->prefix_len = cfg->plen;
1057         ifa->rt_priority = cfg->rt_priority;
1058         ifa->flags = cfg->ifa_flags;
1059         /* No need to add the TENTATIVE flag for addresses with NODAD */
1060         if (!(cfg->ifa_flags & IFA_F_NODAD))
1061                 ifa->flags |= IFA_F_TENTATIVE;
1062         ifa->valid_lft = cfg->valid_lft;
1063         ifa->prefered_lft = cfg->preferred_lft;
1064         ifa->cstamp = ifa->tstamp = jiffies;
1065         ifa->tokenized = false;
1066
1067         ifa->rt = f6i;
1068
1069         ifa->idev = idev;
1070         in6_dev_hold(idev);
1071
1072         /* For caller */
1073         refcount_set(&ifa->refcnt, 1);
1074
1075         rcu_read_lock_bh();
1076
1077         err = ipv6_add_addr_hash(idev->dev, ifa);
1078         if (err < 0) {
1079                 rcu_read_unlock_bh();
1080                 goto out;
1081         }
1082
1083         write_lock(&idev->lock);
1084
1085         /* Add to inet6_dev unicast addr list. */
1086         ipv6_link_dev_addr(idev, ifa);
1087
1088         if (ifa->flags&IFA_F_TEMPORARY) {
1089                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1090                 in6_ifa_hold(ifa);
1091         }
1092
1093         in6_ifa_hold(ifa);
1094         write_unlock(&idev->lock);
1095
1096         rcu_read_unlock_bh();
1097
1098         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1099 out:
1100         if (unlikely(err < 0)) {
1101                 fib6_info_release(f6i);
1102
1103                 if (ifa) {
1104                         if (ifa->idev)
1105                                 in6_dev_put(ifa->idev);
1106                         kfree(ifa);
1107                 }
1108                 ifa = ERR_PTR(err);
1109         }
1110
1111         return ifa;
1112 }
1113
1114 enum cleanup_prefix_rt_t {
1115         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1116         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1117         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1118 };
1119
1120 /*
1121  * Check, whether the prefix for ifp would still need a prefix route
1122  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1123  * constants.
1124  *
1125  * 1) we don't purge prefix if address was not permanent.
1126  *    prefix is managed by its own lifetime.
1127  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1128  * 3) if there are no addresses, delete prefix.
1129  * 4) if there are still other permanent address(es),
1130  *    corresponding prefix is still permanent.
1131  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1132  *    don't purge the prefix, assume user space is managing it.
1133  * 6) otherwise, update prefix lifetime to the
1134  *    longest valid lifetime among the corresponding
1135  *    addresses on the device.
1136  *    Note: subsequent RA will update lifetime.
1137  **/
1138 static enum cleanup_prefix_rt_t
1139 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1140 {
1141         struct inet6_ifaddr *ifa;
1142         struct inet6_dev *idev = ifp->idev;
1143         unsigned long lifetime;
1144         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1145
1146         *expires = jiffies;
1147
1148         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1149                 if (ifa == ifp)
1150                         continue;
1151                 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1152                                        ifp->prefix_len))
1153                         continue;
1154                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1155                         return CLEANUP_PREFIX_RT_NOP;
1156
1157                 action = CLEANUP_PREFIX_RT_EXPIRE;
1158
1159                 spin_lock(&ifa->lock);
1160
1161                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1162                 /*
1163                  * Note: Because this address is
1164                  * not permanent, lifetime <
1165                  * LONG_MAX / HZ here.
1166                  */
1167                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1168                         *expires = ifa->tstamp + lifetime * HZ;
1169                 spin_unlock(&ifa->lock);
1170         }
1171
1172         return action;
1173 }
1174
1175 static void
1176 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1177 {
1178         struct fib6_info *f6i;
1179
1180         f6i = addrconf_get_prefix_route(&ifp->addr,
1181                                        ifp->prefix_len,
1182                                        ifp->idev->dev,
1183                                        0, RTF_GATEWAY | RTF_DEFAULT);
1184         if (f6i) {
1185                 if (del_rt)
1186                         ip6_del_rt(dev_net(ifp->idev->dev), f6i);
1187                 else {
1188                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1189                                 fib6_set_expires(f6i, expires);
1190                         fib6_info_release(f6i);
1191                 }
1192         }
1193 }
1194
1195
1196 /* This function wants to get referenced ifp and releases it before return */
1197
1198 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1199 {
1200         int state;
1201         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1202         unsigned long expires;
1203
1204         ASSERT_RTNL();
1205
1206         spin_lock_bh(&ifp->lock);
1207         state = ifp->state;
1208         ifp->state = INET6_IFADDR_STATE_DEAD;
1209         spin_unlock_bh(&ifp->lock);
1210
1211         if (state == INET6_IFADDR_STATE_DEAD)
1212                 goto out;
1213
1214         spin_lock_bh(&addrconf_hash_lock);
1215         hlist_del_init_rcu(&ifp->addr_lst);
1216         spin_unlock_bh(&addrconf_hash_lock);
1217
1218         write_lock_bh(&ifp->idev->lock);
1219
1220         if (ifp->flags&IFA_F_TEMPORARY) {
1221                 list_del(&ifp->tmp_list);
1222                 if (ifp->ifpub) {
1223                         in6_ifa_put(ifp->ifpub);
1224                         ifp->ifpub = NULL;
1225                 }
1226                 __in6_ifa_put(ifp);
1227         }
1228
1229         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1230                 action = check_cleanup_prefix_route(ifp, &expires);
1231
1232         list_del_rcu(&ifp->if_list);
1233         __in6_ifa_put(ifp);
1234
1235         write_unlock_bh(&ifp->idev->lock);
1236
1237         addrconf_del_dad_work(ifp);
1238
1239         ipv6_ifa_notify(RTM_DELADDR, ifp);
1240
1241         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1242
1243         if (action != CLEANUP_PREFIX_RT_NOP) {
1244                 cleanup_prefix_route(ifp, expires,
1245                         action == CLEANUP_PREFIX_RT_DEL);
1246         }
1247
1248         /* clean up prefsrc entries */
1249         rt6_remove_prefsrc(ifp);
1250 out:
1251         in6_ifa_put(ifp);
1252 }
1253
1254 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1255                                 struct inet6_ifaddr *ift,
1256                                 bool block)
1257 {
1258         struct inet6_dev *idev = ifp->idev;
1259         struct in6_addr addr, *tmpaddr;
1260         unsigned long tmp_tstamp, age;
1261         unsigned long regen_advance;
1262         struct ifa6_config cfg;
1263         int ret = 0;
1264         unsigned long now = jiffies;
1265         long max_desync_factor;
1266         s32 cnf_temp_preferred_lft;
1267
1268         write_lock_bh(&idev->lock);
1269         if (ift) {
1270                 spin_lock_bh(&ift->lock);
1271                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1272                 spin_unlock_bh(&ift->lock);
1273                 tmpaddr = &addr;
1274         } else {
1275                 tmpaddr = NULL;
1276         }
1277 retry:
1278         in6_dev_hold(idev);
1279         if (idev->cnf.use_tempaddr <= 0) {
1280                 write_unlock_bh(&idev->lock);
1281                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1282                 in6_dev_put(idev);
1283                 ret = -1;
1284                 goto out;
1285         }
1286         spin_lock_bh(&ifp->lock);
1287         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1288                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1289                 spin_unlock_bh(&ifp->lock);
1290                 write_unlock_bh(&idev->lock);
1291                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1292                         __func__);
1293                 in6_dev_put(idev);
1294                 ret = -1;
1295                 goto out;
1296         }
1297         in6_ifa_hold(ifp);
1298         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1299         ipv6_try_regen_rndid(idev, tmpaddr);
1300         memcpy(&addr.s6_addr[8], idev->rndid, 8);
1301         age = (now - ifp->tstamp) / HZ;
1302
1303         regen_advance = idev->cnf.regen_max_retry *
1304                         idev->cnf.dad_transmits *
1305                         NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1306
1307         /* recalculate max_desync_factor each time and update
1308          * idev->desync_factor if it's larger
1309          */
1310         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1311         max_desync_factor = min_t(__u32,
1312                                   idev->cnf.max_desync_factor,
1313                                   cnf_temp_preferred_lft - regen_advance);
1314
1315         if (unlikely(idev->desync_factor > max_desync_factor)) {
1316                 if (max_desync_factor > 0) {
1317                         get_random_bytes(&idev->desync_factor,
1318                                          sizeof(idev->desync_factor));
1319                         idev->desync_factor %= max_desync_factor;
1320                 } else {
1321                         idev->desync_factor = 0;
1322                 }
1323         }
1324
1325         memset(&cfg, 0, sizeof(cfg));
1326         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1327                               idev->cnf.temp_valid_lft + age);
1328         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1329         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1330
1331         cfg.plen = ifp->prefix_len;
1332         tmp_tstamp = ifp->tstamp;
1333         spin_unlock_bh(&ifp->lock);
1334
1335         write_unlock_bh(&idev->lock);
1336
1337         /* A temporary address is created only if this calculated Preferred
1338          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1339          * an implementation must not create a temporary address with a zero
1340          * Preferred Lifetime.
1341          * Use age calculation as in addrconf_verify to avoid unnecessary
1342          * temporary addresses being generated.
1343          */
1344         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1345         if (cfg.preferred_lft <= regen_advance + age) {
1346                 in6_ifa_put(ifp);
1347                 in6_dev_put(idev);
1348                 ret = -1;
1349                 goto out;
1350         }
1351
1352         cfg.ifa_flags = IFA_F_TEMPORARY;
1353         /* set in addrconf_prefix_rcv() */
1354         if (ifp->flags & IFA_F_OPTIMISTIC)
1355                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1356
1357         cfg.pfx = &addr;
1358         cfg.scope = ipv6_addr_scope(cfg.pfx);
1359
1360         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1361         if (IS_ERR(ift)) {
1362                 in6_ifa_put(ifp);
1363                 in6_dev_put(idev);
1364                 pr_info("%s: retry temporary address regeneration\n", __func__);
1365                 tmpaddr = &addr;
1366                 write_lock_bh(&idev->lock);
1367                 goto retry;
1368         }
1369
1370         spin_lock_bh(&ift->lock);
1371         ift->ifpub = ifp;
1372         ift->cstamp = now;
1373         ift->tstamp = tmp_tstamp;
1374         spin_unlock_bh(&ift->lock);
1375
1376         addrconf_dad_start(ift);
1377         in6_ifa_put(ift);
1378         in6_dev_put(idev);
1379 out:
1380         return ret;
1381 }
1382
1383 /*
1384  *      Choose an appropriate source address (RFC3484)
1385  */
1386 enum {
1387         IPV6_SADDR_RULE_INIT = 0,
1388         IPV6_SADDR_RULE_LOCAL,
1389         IPV6_SADDR_RULE_SCOPE,
1390         IPV6_SADDR_RULE_PREFERRED,
1391 #ifdef CONFIG_IPV6_MIP6
1392         IPV6_SADDR_RULE_HOA,
1393 #endif
1394         IPV6_SADDR_RULE_OIF,
1395         IPV6_SADDR_RULE_LABEL,
1396         IPV6_SADDR_RULE_PRIVACY,
1397         IPV6_SADDR_RULE_ORCHID,
1398         IPV6_SADDR_RULE_PREFIX,
1399 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1400         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1401 #endif
1402         IPV6_SADDR_RULE_MAX
1403 };
1404
1405 struct ipv6_saddr_score {
1406         int                     rule;
1407         int                     addr_type;
1408         struct inet6_ifaddr     *ifa;
1409         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1410         int                     scopedist;
1411         int                     matchlen;
1412 };
1413
1414 struct ipv6_saddr_dst {
1415         const struct in6_addr *addr;
1416         int ifindex;
1417         int scope;
1418         int label;
1419         unsigned int prefs;
1420 };
1421
1422 static inline int ipv6_saddr_preferred(int type)
1423 {
1424         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1425                 return 1;
1426         return 0;
1427 }
1428
1429 static bool ipv6_use_optimistic_addr(struct net *net,
1430                                      struct inet6_dev *idev)
1431 {
1432 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1433         if (!idev)
1434                 return false;
1435         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1436                 return false;
1437         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1438                 return false;
1439
1440         return true;
1441 #else
1442         return false;
1443 #endif
1444 }
1445
1446 static bool ipv6_allow_optimistic_dad(struct net *net,
1447                                       struct inet6_dev *idev)
1448 {
1449 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1450         if (!idev)
1451                 return false;
1452         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1453                 return false;
1454
1455         return true;
1456 #else
1457         return false;
1458 #endif
1459 }
1460
1461 static int ipv6_get_saddr_eval(struct net *net,
1462                                struct ipv6_saddr_score *score,
1463                                struct ipv6_saddr_dst *dst,
1464                                int i)
1465 {
1466         int ret;
1467
1468         if (i <= score->rule) {
1469                 switch (i) {
1470                 case IPV6_SADDR_RULE_SCOPE:
1471                         ret = score->scopedist;
1472                         break;
1473                 case IPV6_SADDR_RULE_PREFIX:
1474                         ret = score->matchlen;
1475                         break;
1476                 default:
1477                         ret = !!test_bit(i, score->scorebits);
1478                 }
1479                 goto out;
1480         }
1481
1482         switch (i) {
1483         case IPV6_SADDR_RULE_INIT:
1484                 /* Rule 0: remember if hiscore is not ready yet */
1485                 ret = !!score->ifa;
1486                 break;
1487         case IPV6_SADDR_RULE_LOCAL:
1488                 /* Rule 1: Prefer same address */
1489                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1490                 break;
1491         case IPV6_SADDR_RULE_SCOPE:
1492                 /* Rule 2: Prefer appropriate scope
1493                  *
1494                  *      ret
1495                  *       ^
1496                  *    -1 |  d 15
1497                  *    ---+--+-+---> scope
1498                  *       |
1499                  *       |             d is scope of the destination.
1500                  *  B-d  |  \
1501                  *       |   \      <- smaller scope is better if
1502                  *  B-15 |    \        if scope is enough for destination.
1503                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1504                  * d-C-1 | /
1505                  *       |/         <- greater is better
1506                  *   -C  /             if scope is not enough for destination.
1507                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1508                  *
1509                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1510                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1511                  * Assume B = 0 and we get C > 29.
1512                  */
1513                 ret = __ipv6_addr_src_scope(score->addr_type);
1514                 if (ret >= dst->scope)
1515                         ret = -ret;
1516                 else
1517                         ret -= 128;     /* 30 is enough */
1518                 score->scopedist = ret;
1519                 break;
1520         case IPV6_SADDR_RULE_PREFERRED:
1521             {
1522                 /* Rule 3: Avoid deprecated and optimistic addresses */
1523                 u8 avoid = IFA_F_DEPRECATED;
1524
1525                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1526                         avoid |= IFA_F_OPTIMISTIC;
1527                 ret = ipv6_saddr_preferred(score->addr_type) ||
1528                       !(score->ifa->flags & avoid);
1529                 break;
1530             }
1531 #ifdef CONFIG_IPV6_MIP6
1532         case IPV6_SADDR_RULE_HOA:
1533             {
1534                 /* Rule 4: Prefer home address */
1535                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1536                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1537                 break;
1538             }
1539 #endif
1540         case IPV6_SADDR_RULE_OIF:
1541                 /* Rule 5: Prefer outgoing interface */
1542                 ret = (!dst->ifindex ||
1543                        dst->ifindex == score->ifa->idev->dev->ifindex);
1544                 break;
1545         case IPV6_SADDR_RULE_LABEL:
1546                 /* Rule 6: Prefer matching label */
1547                 ret = ipv6_addr_label(net,
1548                                       &score->ifa->addr, score->addr_type,
1549                                       score->ifa->idev->dev->ifindex) == dst->label;
1550                 break;
1551         case IPV6_SADDR_RULE_PRIVACY:
1552             {
1553                 /* Rule 7: Prefer public address
1554                  * Note: prefer temporary address if use_tempaddr >= 2
1555                  */
1556                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1557                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1558                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1559                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1560                 break;
1561             }
1562         case IPV6_SADDR_RULE_ORCHID:
1563                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1564                  *          non-ORCHID vs non-ORCHID
1565                  */
1566                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1567                         ipv6_addr_orchid(dst->addr));
1568                 break;
1569         case IPV6_SADDR_RULE_PREFIX:
1570                 /* Rule 8: Use longest matching prefix */
1571                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1572                 if (ret > score->ifa->prefix_len)
1573                         ret = score->ifa->prefix_len;
1574                 score->matchlen = ret;
1575                 break;
1576 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1577         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1578                 /* Optimistic addresses still have lower precedence than other
1579                  * preferred addresses.
1580                  */
1581                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1582                 break;
1583 #endif
1584         default:
1585                 ret = 0;
1586         }
1587
1588         if (ret)
1589                 __set_bit(i, score->scorebits);
1590         score->rule = i;
1591 out:
1592         return ret;
1593 }
1594
1595 static int __ipv6_dev_get_saddr(struct net *net,
1596                                 struct ipv6_saddr_dst *dst,
1597                                 struct inet6_dev *idev,
1598                                 struct ipv6_saddr_score *scores,
1599                                 int hiscore_idx)
1600 {
1601         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1602
1603         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1604                 int i;
1605
1606                 /*
1607                  * - Tentative Address (RFC2462 section 5.4)
1608                  *  - A tentative address is not considered
1609                  *    "assigned to an interface" in the traditional
1610                  *    sense, unless it is also flagged as optimistic.
1611                  * - Candidate Source Address (section 4)
1612                  *  - In any case, anycast addresses, multicast
1613                  *    addresses, and the unspecified address MUST
1614                  *    NOT be included in a candidate set.
1615                  */
1616                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1617                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1618                         continue;
1619
1620                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1621
1622                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1623                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1624                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1625                                             idev->dev->name);
1626                         continue;
1627                 }
1628
1629                 score->rule = -1;
1630                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1631
1632                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1633                         int minihiscore, miniscore;
1634
1635                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1636                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1637
1638                         if (minihiscore > miniscore) {
1639                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1640                                     score->scopedist > 0) {
1641                                         /*
1642                                          * special case:
1643                                          * each remaining entry
1644                                          * has too small (not enough)
1645                                          * scope, because ifa entries
1646                                          * are sorted by their scope
1647                                          * values.
1648                                          */
1649                                         goto out;
1650                                 }
1651                                 break;
1652                         } else if (minihiscore < miniscore) {
1653                                 swap(hiscore, score);
1654                                 hiscore_idx = 1 - hiscore_idx;
1655
1656                                 /* restore our iterator */
1657                                 score->ifa = hiscore->ifa;
1658
1659                                 break;
1660                         }
1661                 }
1662         }
1663 out:
1664         return hiscore_idx;
1665 }
1666
1667 static int ipv6_get_saddr_master(struct net *net,
1668                                  const struct net_device *dst_dev,
1669                                  const struct net_device *master,
1670                                  struct ipv6_saddr_dst *dst,
1671                                  struct ipv6_saddr_score *scores,
1672                                  int hiscore_idx)
1673 {
1674         struct inet6_dev *idev;
1675
1676         idev = __in6_dev_get(dst_dev);
1677         if (idev)
1678                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1679                                                    scores, hiscore_idx);
1680
1681         idev = __in6_dev_get(master);
1682         if (idev)
1683                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1684                                                    scores, hiscore_idx);
1685
1686         return hiscore_idx;
1687 }
1688
1689 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1690                        const struct in6_addr *daddr, unsigned int prefs,
1691                        struct in6_addr *saddr)
1692 {
1693         struct ipv6_saddr_score scores[2], *hiscore;
1694         struct ipv6_saddr_dst dst;
1695         struct inet6_dev *idev;
1696         struct net_device *dev;
1697         int dst_type;
1698         bool use_oif_addr = false;
1699         int hiscore_idx = 0;
1700         int ret = 0;
1701
1702         dst_type = __ipv6_addr_type(daddr);
1703         dst.addr = daddr;
1704         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1705         dst.scope = __ipv6_addr_src_scope(dst_type);
1706         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1707         dst.prefs = prefs;
1708
1709         scores[hiscore_idx].rule = -1;
1710         scores[hiscore_idx].ifa = NULL;
1711
1712         rcu_read_lock();
1713
1714         /* Candidate Source Address (section 4)
1715          *  - multicast and link-local destination address,
1716          *    the set of candidate source address MUST only
1717          *    include addresses assigned to interfaces
1718          *    belonging to the same link as the outgoing
1719          *    interface.
1720          * (- For site-local destination addresses, the
1721          *    set of candidate source addresses MUST only
1722          *    include addresses assigned to interfaces
1723          *    belonging to the same site as the outgoing
1724          *    interface.)
1725          *  - "It is RECOMMENDED that the candidate source addresses
1726          *    be the set of unicast addresses assigned to the
1727          *    interface that will be used to send to the destination
1728          *    (the 'outgoing' interface)." (RFC 6724)
1729          */
1730         if (dst_dev) {
1731                 idev = __in6_dev_get(dst_dev);
1732                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1733                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1734                     (idev && idev->cnf.use_oif_addrs_only)) {
1735                         use_oif_addr = true;
1736                 }
1737         }
1738
1739         if (use_oif_addr) {
1740                 if (idev)
1741                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1742         } else {
1743                 const struct net_device *master;
1744                 int master_idx = 0;
1745
1746                 /* if dst_dev exists and is enslaved to an L3 device, then
1747                  * prefer addresses from dst_dev and then the master over
1748                  * any other enslaved devices in the L3 domain.
1749                  */
1750                 master = l3mdev_master_dev_rcu(dst_dev);
1751                 if (master) {
1752                         master_idx = master->ifindex;
1753
1754                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1755                                                             master, &dst,
1756                                                             scores, hiscore_idx);
1757
1758                         if (scores[hiscore_idx].ifa)
1759                                 goto out;
1760                 }
1761
1762                 for_each_netdev_rcu(net, dev) {
1763                         /* only consider addresses on devices in the
1764                          * same L3 domain
1765                          */
1766                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1767                                 continue;
1768                         idev = __in6_dev_get(dev);
1769                         if (!idev)
1770                                 continue;
1771                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1772                 }
1773         }
1774
1775 out:
1776         hiscore = &scores[hiscore_idx];
1777         if (!hiscore->ifa)
1778                 ret = -EADDRNOTAVAIL;
1779         else
1780                 *saddr = hiscore->ifa->addr;
1781
1782         rcu_read_unlock();
1783         return ret;
1784 }
1785 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1786
1787 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1788                       u32 banned_flags)
1789 {
1790         struct inet6_ifaddr *ifp;
1791         int err = -EADDRNOTAVAIL;
1792
1793         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1794                 if (ifp->scope > IFA_LINK)
1795                         break;
1796                 if (ifp->scope == IFA_LINK &&
1797                     !(ifp->flags & banned_flags)) {
1798                         *addr = ifp->addr;
1799                         err = 0;
1800                         break;
1801                 }
1802         }
1803         return err;
1804 }
1805
1806 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1807                     u32 banned_flags)
1808 {
1809         struct inet6_dev *idev;
1810         int err = -EADDRNOTAVAIL;
1811
1812         rcu_read_lock();
1813         idev = __in6_dev_get(dev);
1814         if (idev) {
1815                 read_lock_bh(&idev->lock);
1816                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1817                 read_unlock_bh(&idev->lock);
1818         }
1819         rcu_read_unlock();
1820         return err;
1821 }
1822
1823 static int ipv6_count_addresses(const struct inet6_dev *idev)
1824 {
1825         const struct inet6_ifaddr *ifp;
1826         int cnt = 0;
1827
1828         rcu_read_lock();
1829         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1830                 cnt++;
1831         rcu_read_unlock();
1832         return cnt;
1833 }
1834
1835 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1836                   const struct net_device *dev, int strict)
1837 {
1838         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1839                                        strict, IFA_F_TENTATIVE);
1840 }
1841 EXPORT_SYMBOL(ipv6_chk_addr);
1842
1843 /* device argument is used to find the L3 domain of interest. If
1844  * skip_dev_check is set, then the ifp device is not checked against
1845  * the passed in dev argument. So the 2 cases for addresses checks are:
1846  *   1. does the address exist in the L3 domain that dev is part of
1847  *      (skip_dev_check = true), or
1848  *
1849  *   2. does the address exist on the specific device
1850  *      (skip_dev_check = false)
1851  */
1852 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1853                             const struct net_device *dev, bool skip_dev_check,
1854                             int strict, u32 banned_flags)
1855 {
1856         unsigned int hash = inet6_addr_hash(net, addr);
1857         const struct net_device *l3mdev;
1858         struct inet6_ifaddr *ifp;
1859         u32 ifp_flags;
1860
1861         rcu_read_lock();
1862
1863         l3mdev = l3mdev_master_dev_rcu(dev);
1864         if (skip_dev_check)
1865                 dev = NULL;
1866
1867         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1868                 if (!net_eq(dev_net(ifp->idev->dev), net))
1869                         continue;
1870
1871                 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1872                         continue;
1873
1874                 /* Decouple optimistic from tentative for evaluation here.
1875                  * Ban optimistic addresses explicitly, when required.
1876                  */
1877                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1878                             ? (ifp->flags&~IFA_F_TENTATIVE)
1879                             : ifp->flags;
1880                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1881                     !(ifp_flags&banned_flags) &&
1882                     (!dev || ifp->idev->dev == dev ||
1883                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1884                         rcu_read_unlock();
1885                         return 1;
1886                 }
1887         }
1888
1889         rcu_read_unlock();
1890         return 0;
1891 }
1892 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1893
1894
1895 /* Compares an address/prefix_len with addresses on device @dev.
1896  * If one is found it returns true.
1897  */
1898 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1899         const unsigned int prefix_len, struct net_device *dev)
1900 {
1901         const struct inet6_ifaddr *ifa;
1902         const struct inet6_dev *idev;
1903         bool ret = false;
1904
1905         rcu_read_lock();
1906         idev = __in6_dev_get(dev);
1907         if (idev) {
1908                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1909                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1910                         if (ret)
1911                                 break;
1912                 }
1913         }
1914         rcu_read_unlock();
1915
1916         return ret;
1917 }
1918 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1919
1920 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1921 {
1922         const struct inet6_ifaddr *ifa;
1923         const struct inet6_dev *idev;
1924         int     onlink;
1925
1926         onlink = 0;
1927         rcu_read_lock();
1928         idev = __in6_dev_get(dev);
1929         if (idev) {
1930                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1931                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1932                                                    ifa->prefix_len);
1933                         if (onlink)
1934                                 break;
1935                 }
1936         }
1937         rcu_read_unlock();
1938         return onlink;
1939 }
1940 EXPORT_SYMBOL(ipv6_chk_prefix);
1941
1942 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1943                                      struct net_device *dev, int strict)
1944 {
1945         unsigned int hash = inet6_addr_hash(net, addr);
1946         struct inet6_ifaddr *ifp, *result = NULL;
1947
1948         rcu_read_lock();
1949         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1950                 if (!net_eq(dev_net(ifp->idev->dev), net))
1951                         continue;
1952                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1953                         if (!dev || ifp->idev->dev == dev ||
1954                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1955                                 result = ifp;
1956                                 in6_ifa_hold(ifp);
1957                                 break;
1958                         }
1959                 }
1960         }
1961         rcu_read_unlock();
1962
1963         return result;
1964 }
1965
1966 /* Gets referenced address, destroys ifaddr */
1967
1968 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1969 {
1970         if (dad_failed)
1971                 ifp->flags |= IFA_F_DADFAILED;
1972
1973         if (ifp->flags&IFA_F_TEMPORARY) {
1974                 struct inet6_ifaddr *ifpub;
1975                 spin_lock_bh(&ifp->lock);
1976                 ifpub = ifp->ifpub;
1977                 if (ifpub) {
1978                         in6_ifa_hold(ifpub);
1979                         spin_unlock_bh(&ifp->lock);
1980                         ipv6_create_tempaddr(ifpub, ifp, true);
1981                         in6_ifa_put(ifpub);
1982                 } else {
1983                         spin_unlock_bh(&ifp->lock);
1984                 }
1985                 ipv6_del_addr(ifp);
1986         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1987                 spin_lock_bh(&ifp->lock);
1988                 addrconf_del_dad_work(ifp);
1989                 ifp->flags |= IFA_F_TENTATIVE;
1990                 if (dad_failed)
1991                         ifp->flags &= ~IFA_F_OPTIMISTIC;
1992                 spin_unlock_bh(&ifp->lock);
1993                 if (dad_failed)
1994                         ipv6_ifa_notify(0, ifp);
1995                 in6_ifa_put(ifp);
1996         } else {
1997                 ipv6_del_addr(ifp);
1998         }
1999 }
2000
2001 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2002 {
2003         int err = -ENOENT;
2004
2005         spin_lock_bh(&ifp->lock);
2006         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2007                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2008                 err = 0;
2009         }
2010         spin_unlock_bh(&ifp->lock);
2011
2012         return err;
2013 }
2014
2015 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2016 {
2017         struct inet6_dev *idev = ifp->idev;
2018         struct net *net = dev_net(ifp->idev->dev);
2019
2020         if (addrconf_dad_end(ifp)) {
2021                 in6_ifa_put(ifp);
2022                 return;
2023         }
2024
2025         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2026                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2027
2028         spin_lock_bh(&ifp->lock);
2029
2030         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2031                 struct in6_addr new_addr;
2032                 struct inet6_ifaddr *ifp2;
2033                 int retries = ifp->stable_privacy_retry + 1;
2034                 struct ifa6_config cfg = {
2035                         .pfx = &new_addr,
2036                         .plen = ifp->prefix_len,
2037                         .ifa_flags = ifp->flags,
2038                         .valid_lft = ifp->valid_lft,
2039                         .preferred_lft = ifp->prefered_lft,
2040                         .scope = ifp->scope,
2041                 };
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                 spin_unlock_bh(&ifp->lock);
2055
2056                 if (idev->cnf.max_addresses &&
2057                     ipv6_count_addresses(idev) >=
2058                     idev->cnf.max_addresses)
2059                         goto lock_errdad;
2060
2061                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2062                                      ifp->idev->dev->name);
2063
2064                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2065                 if (IS_ERR(ifp2))
2066                         goto lock_errdad;
2067
2068                 spin_lock_bh(&ifp2->lock);
2069                 ifp2->stable_privacy_retry = retries;
2070                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2071                 spin_unlock_bh(&ifp2->lock);
2072
2073                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2074                 in6_ifa_put(ifp2);
2075 lock_errdad:
2076                 spin_lock_bh(&ifp->lock);
2077         }
2078
2079 errdad:
2080         /* transition from _POSTDAD to _ERRDAD */
2081         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2082         spin_unlock_bh(&ifp->lock);
2083
2084         addrconf_mod_dad_work(ifp, 0);
2085         in6_ifa_put(ifp);
2086 }
2087
2088 /* Join to solicited addr multicast group.
2089  * caller must hold RTNL */
2090 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2091 {
2092         struct in6_addr maddr;
2093
2094         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2095                 return;
2096
2097         addrconf_addr_solict_mult(addr, &maddr);
2098         ipv6_dev_mc_inc(dev, &maddr);
2099 }
2100
2101 /* caller must hold RTNL */
2102 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2103 {
2104         struct in6_addr maddr;
2105
2106         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2107                 return;
2108
2109         addrconf_addr_solict_mult(addr, &maddr);
2110         __ipv6_dev_mc_dec(idev, &maddr);
2111 }
2112
2113 /* caller must hold RTNL */
2114 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2115 {
2116         struct in6_addr addr;
2117
2118         if (ifp->prefix_len >= 127) /* RFC 6164 */
2119                 return;
2120         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2121         if (ipv6_addr_any(&addr))
2122                 return;
2123         __ipv6_dev_ac_inc(ifp->idev, &addr);
2124 }
2125
2126 /* caller must hold RTNL */
2127 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2128 {
2129         struct in6_addr addr;
2130
2131         if (ifp->prefix_len >= 127) /* RFC 6164 */
2132                 return;
2133         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2134         if (ipv6_addr_any(&addr))
2135                 return;
2136         __ipv6_dev_ac_dec(ifp->idev, &addr);
2137 }
2138
2139 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2140 {
2141         switch (dev->addr_len) {
2142         case ETH_ALEN:
2143                 memcpy(eui, dev->dev_addr, 3);
2144                 eui[3] = 0xFF;
2145                 eui[4] = 0xFE;
2146                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2147                 break;
2148         case EUI64_ADDR_LEN:
2149                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2150                 eui[0] ^= 2;
2151                 break;
2152         default:
2153                 return -1;
2154         }
2155
2156         return 0;
2157 }
2158
2159 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2160 {
2161         union fwnet_hwaddr *ha;
2162
2163         if (dev->addr_len != FWNET_ALEN)
2164                 return -1;
2165
2166         ha = (union fwnet_hwaddr *)dev->dev_addr;
2167
2168         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2169         eui[0] ^= 2;
2170         return 0;
2171 }
2172
2173 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2174 {
2175         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2176         if (dev->addr_len != ARCNET_ALEN)
2177                 return -1;
2178         memset(eui, 0, 7);
2179         eui[7] = *(u8 *)dev->dev_addr;
2180         return 0;
2181 }
2182
2183 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2184 {
2185         if (dev->addr_len != INFINIBAND_ALEN)
2186                 return -1;
2187         memcpy(eui, dev->dev_addr + 12, 8);
2188         eui[0] |= 2;
2189         return 0;
2190 }
2191
2192 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2193 {
2194         if (addr == 0)
2195                 return -1;
2196         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2197                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2198                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2199                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2200                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2201                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2202         eui[1] = 0;
2203         eui[2] = 0x5E;
2204         eui[3] = 0xFE;
2205         memcpy(eui + 4, &addr, 4);
2206         return 0;
2207 }
2208
2209 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2210 {
2211         if (dev->priv_flags & IFF_ISATAP)
2212                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2213         return -1;
2214 }
2215
2216 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2217 {
2218         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2219 }
2220
2221 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2222 {
2223         memcpy(eui, dev->perm_addr, 3);
2224         memcpy(eui + 5, dev->perm_addr + 3, 3);
2225         eui[3] = 0xFF;
2226         eui[4] = 0xFE;
2227         eui[0] ^= 2;
2228         return 0;
2229 }
2230
2231 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2232 {
2233         switch (dev->type) {
2234         case ARPHRD_ETHER:
2235         case ARPHRD_FDDI:
2236                 return addrconf_ifid_eui48(eui, dev);
2237         case ARPHRD_ARCNET:
2238                 return addrconf_ifid_arcnet(eui, dev);
2239         case ARPHRD_INFINIBAND:
2240                 return addrconf_ifid_infiniband(eui, dev);
2241         case ARPHRD_SIT:
2242                 return addrconf_ifid_sit(eui, dev);
2243         case ARPHRD_IPGRE:
2244         case ARPHRD_TUNNEL:
2245                 return addrconf_ifid_gre(eui, dev);
2246         case ARPHRD_6LOWPAN:
2247                 return addrconf_ifid_6lowpan(eui, dev);
2248         case ARPHRD_IEEE1394:
2249                 return addrconf_ifid_ieee1394(eui, dev);
2250         case ARPHRD_TUNNEL6:
2251         case ARPHRD_IP6GRE:
2252         case ARPHRD_RAWIP:
2253                 return addrconf_ifid_ip6tnl(eui, dev);
2254         }
2255         return -1;
2256 }
2257
2258 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2259 {
2260         int err = -1;
2261         struct inet6_ifaddr *ifp;
2262
2263         read_lock_bh(&idev->lock);
2264         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2265                 if (ifp->scope > IFA_LINK)
2266                         break;
2267                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2268                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2269                         err = 0;
2270                         break;
2271                 }
2272         }
2273         read_unlock_bh(&idev->lock);
2274         return err;
2275 }
2276
2277 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2278 static void ipv6_regen_rndid(struct inet6_dev *idev)
2279 {
2280 regen:
2281         get_random_bytes(idev->rndid, sizeof(idev->rndid));
2282         idev->rndid[0] &= ~0x02;
2283
2284         /*
2285          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2286          * check if generated address is not inappropriate
2287          *
2288          *  - Reserved subnet anycast (RFC 2526)
2289          *      11111101 11....11 1xxxxxxx
2290          *  - ISATAP (RFC4214) 6.1
2291          *      00-00-5E-FE-xx-xx-xx-xx
2292          *  - value 0
2293          *  - XXX: already assigned to an address on the device
2294          */
2295         if (idev->rndid[0] == 0xfd &&
2296             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2297             (idev->rndid[7]&0x80))
2298                 goto regen;
2299         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2300                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2301                         goto regen;
2302                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2303                         goto regen;
2304         }
2305 }
2306
2307 static void  ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2308 {
2309         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2310                 ipv6_regen_rndid(idev);
2311 }
2312
2313 /*
2314  *      Add prefix route.
2315  */
2316
2317 static void
2318 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2319                       struct net_device *dev, unsigned long expires,
2320                       u32 flags, gfp_t gfp_flags)
2321 {
2322         struct fib6_config cfg = {
2323                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2324                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2325                 .fc_ifindex = dev->ifindex,
2326                 .fc_expires = expires,
2327                 .fc_dst_len = plen,
2328                 .fc_flags = RTF_UP | flags,
2329                 .fc_nlinfo.nl_net = dev_net(dev),
2330                 .fc_protocol = RTPROT_KERNEL,
2331                 .fc_type = RTN_UNICAST,
2332         };
2333
2334         cfg.fc_dst = *pfx;
2335
2336         /* Prevent useless cloning on PtP SIT.
2337            This thing is done here expecting that the whole
2338            class of non-broadcast devices need not cloning.
2339          */
2340 #if IS_ENABLED(CONFIG_IPV6_SIT)
2341         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2342                 cfg.fc_flags |= RTF_NONEXTHOP;
2343 #endif
2344
2345         ip6_route_add(&cfg, gfp_flags, NULL);
2346 }
2347
2348
2349 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2350                                                   int plen,
2351                                                   const struct net_device *dev,
2352                                                   u32 flags, u32 noflags)
2353 {
2354         struct fib6_node *fn;
2355         struct fib6_info *rt = NULL;
2356         struct fib6_table *table;
2357         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2358
2359         table = fib6_get_table(dev_net(dev), tb_id);
2360         if (!table)
2361                 return NULL;
2362
2363         rcu_read_lock();
2364         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2365         if (!fn)
2366                 goto out;
2367
2368         for_each_fib6_node_rt_rcu(fn) {
2369                 if (rt->fib6_nh.nh_dev->ifindex != dev->ifindex)
2370                         continue;
2371                 if ((rt->fib6_flags & flags) != flags)
2372                         continue;
2373                 if ((rt->fib6_flags & noflags) != 0)
2374                         continue;
2375                 if (!fib6_info_hold_safe(rt))
2376                         continue;
2377                 break;
2378         }
2379 out:
2380         rcu_read_unlock();
2381         return rt;
2382 }
2383
2384
2385 /* Create "default" multicast route to the interface */
2386
2387 static void addrconf_add_mroute(struct net_device *dev)
2388 {
2389         struct fib6_config cfg = {
2390                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2391                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2392                 .fc_ifindex = dev->ifindex,
2393                 .fc_dst_len = 8,
2394                 .fc_flags = RTF_UP,
2395                 .fc_type = RTN_UNICAST,
2396                 .fc_nlinfo.nl_net = dev_net(dev),
2397         };
2398
2399         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2400
2401         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2402 }
2403
2404 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2405 {
2406         struct inet6_dev *idev;
2407
2408         ASSERT_RTNL();
2409
2410         idev = ipv6_find_idev(dev);
2411         if (!idev)
2412                 return ERR_PTR(-ENOBUFS);
2413
2414         if (idev->cnf.disable_ipv6)
2415                 return ERR_PTR(-EACCES);
2416
2417         /* Add default multicast route */
2418         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2419                 addrconf_add_mroute(dev);
2420
2421         return idev;
2422 }
2423
2424 static void manage_tempaddrs(struct inet6_dev *idev,
2425                              struct inet6_ifaddr *ifp,
2426                              __u32 valid_lft, __u32 prefered_lft,
2427                              bool create, unsigned long now)
2428 {
2429         u32 flags;
2430         struct inet6_ifaddr *ift;
2431
2432         read_lock_bh(&idev->lock);
2433         /* update all temporary addresses in the list */
2434         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2435                 int age, max_valid, max_prefered;
2436
2437                 if (ifp != ift->ifpub)
2438                         continue;
2439
2440                 /* RFC 4941 section 3.3:
2441                  * If a received option will extend the lifetime of a public
2442                  * address, the lifetimes of temporary addresses should
2443                  * be extended, subject to the overall constraint that no
2444                  * temporary addresses should ever remain "valid" or "preferred"
2445                  * for a time longer than (TEMP_VALID_LIFETIME) or
2446                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2447                  */
2448                 age = (now - ift->cstamp) / HZ;
2449                 max_valid = idev->cnf.temp_valid_lft - age;
2450                 if (max_valid < 0)
2451                         max_valid = 0;
2452
2453                 max_prefered = idev->cnf.temp_prefered_lft -
2454                                idev->desync_factor - age;
2455                 if (max_prefered < 0)
2456                         max_prefered = 0;
2457
2458                 if (valid_lft > max_valid)
2459                         valid_lft = max_valid;
2460
2461                 if (prefered_lft > max_prefered)
2462                         prefered_lft = max_prefered;
2463
2464                 spin_lock(&ift->lock);
2465                 flags = ift->flags;
2466                 ift->valid_lft = valid_lft;
2467                 ift->prefered_lft = prefered_lft;
2468                 ift->tstamp = now;
2469                 if (prefered_lft > 0)
2470                         ift->flags &= ~IFA_F_DEPRECATED;
2471
2472                 spin_unlock(&ift->lock);
2473                 if (!(flags&IFA_F_TENTATIVE))
2474                         ipv6_ifa_notify(0, ift);
2475         }
2476
2477         if ((create || list_empty(&idev->tempaddr_list)) &&
2478             idev->cnf.use_tempaddr > 0) {
2479                 /* When a new public address is created as described
2480                  * in [ADDRCONF], also create a new temporary address.
2481                  * Also create a temporary address if it's enabled but
2482                  * no temporary address currently exists.
2483                  */
2484                 read_unlock_bh(&idev->lock);
2485                 ipv6_create_tempaddr(ifp, NULL, false);
2486         } else {
2487                 read_unlock_bh(&idev->lock);
2488         }
2489 }
2490
2491 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2492 {
2493         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2494                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2495 }
2496
2497 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2498                                  const struct prefix_info *pinfo,
2499                                  struct inet6_dev *in6_dev,
2500                                  const struct in6_addr *addr, int addr_type,
2501                                  u32 addr_flags, bool sllao, bool tokenized,
2502                                  __u32 valid_lft, u32 prefered_lft)
2503 {
2504         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2505         int create = 0, update_lft = 0;
2506
2507         if (!ifp && valid_lft) {
2508                 int max_addresses = in6_dev->cnf.max_addresses;
2509                 struct ifa6_config cfg = {
2510                         .pfx = addr,
2511                         .plen = pinfo->prefix_len,
2512                         .ifa_flags = addr_flags,
2513                         .valid_lft = valid_lft,
2514                         .preferred_lft = prefered_lft,
2515                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2516                 };
2517
2518 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2519                 if ((net->ipv6.devconf_all->optimistic_dad ||
2520                      in6_dev->cnf.optimistic_dad) &&
2521                     !net->ipv6.devconf_all->forwarding && sllao)
2522                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2523 #endif
2524
2525                 /* Do not allow to create too much of autoconfigured
2526                  * addresses; this would be too easy way to crash kernel.
2527                  */
2528                 if (!max_addresses ||
2529                     ipv6_count_addresses(in6_dev) < max_addresses)
2530                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2531
2532                 if (IS_ERR_OR_NULL(ifp))
2533                         return -1;
2534
2535                 create = 1;
2536                 spin_lock_bh(&ifp->lock);
2537                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2538                 ifp->cstamp = jiffies;
2539                 ifp->tokenized = tokenized;
2540                 spin_unlock_bh(&ifp->lock);
2541                 addrconf_dad_start(ifp);
2542         }
2543
2544         if (ifp) {
2545                 u32 flags;
2546                 unsigned long now;
2547                 u32 stored_lft;
2548
2549                 /* update lifetime (RFC2462 5.5.3 e) */
2550                 spin_lock_bh(&ifp->lock);
2551                 now = jiffies;
2552                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2553                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2554                 else
2555                         stored_lft = 0;
2556                 if (!create && stored_lft) {
2557                         const u32 minimum_lft = min_t(u32,
2558                                 stored_lft, MIN_VALID_LIFETIME);
2559                         valid_lft = max(valid_lft, minimum_lft);
2560
2561                         /* RFC4862 Section 5.5.3e:
2562                          * "Note that the preferred lifetime of the
2563                          *  corresponding address is always reset to
2564                          *  the Preferred Lifetime in the received
2565                          *  Prefix Information option, regardless of
2566                          *  whether the valid lifetime is also reset or
2567                          *  ignored."
2568                          *
2569                          * So we should always update prefered_lft here.
2570                          */
2571                         update_lft = 1;
2572                 }
2573
2574                 if (update_lft) {
2575                         ifp->valid_lft = valid_lft;
2576                         ifp->prefered_lft = prefered_lft;
2577                         ifp->tstamp = now;
2578                         flags = ifp->flags;
2579                         ifp->flags &= ~IFA_F_DEPRECATED;
2580                         spin_unlock_bh(&ifp->lock);
2581
2582                         if (!(flags&IFA_F_TENTATIVE))
2583                                 ipv6_ifa_notify(0, ifp);
2584                 } else
2585                         spin_unlock_bh(&ifp->lock);
2586
2587                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2588                                  create, now);
2589
2590                 in6_ifa_put(ifp);
2591                 addrconf_verify();
2592         }
2593
2594         return 0;
2595 }
2596 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2597
2598 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2599 {
2600         struct prefix_info *pinfo;
2601         __u32 valid_lft;
2602         __u32 prefered_lft;
2603         int addr_type, err;
2604         u32 addr_flags = 0;
2605         struct inet6_dev *in6_dev;
2606         struct net *net = dev_net(dev);
2607
2608         pinfo = (struct prefix_info *) opt;
2609
2610         if (len < sizeof(struct prefix_info)) {
2611                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2612                 return;
2613         }
2614
2615         /*
2616          *      Validation checks ([ADDRCONF], page 19)
2617          */
2618
2619         addr_type = ipv6_addr_type(&pinfo->prefix);
2620
2621         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2622                 return;
2623
2624         valid_lft = ntohl(pinfo->valid);
2625         prefered_lft = ntohl(pinfo->prefered);
2626
2627         if (prefered_lft > valid_lft) {
2628                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2629                 return;
2630         }
2631
2632         in6_dev = in6_dev_get(dev);
2633
2634         if (!in6_dev) {
2635                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2636                                     dev->name);
2637                 return;
2638         }
2639
2640         /*
2641          *      Two things going on here:
2642          *      1) Add routes for on-link prefixes
2643          *      2) Configure prefixes with the auto flag set
2644          */
2645
2646         if (pinfo->onlink) {
2647                 struct fib6_info *rt;
2648                 unsigned long rt_expires;
2649
2650                 /* Avoid arithmetic overflow. Really, we could
2651                  * save rt_expires in seconds, likely valid_lft,
2652                  * but it would require division in fib gc, that it
2653                  * not good.
2654                  */
2655                 if (HZ > USER_HZ)
2656                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2657                 else
2658                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2659
2660                 if (addrconf_finite_timeout(rt_expires))
2661                         rt_expires *= HZ;
2662
2663                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2664                                                pinfo->prefix_len,
2665                                                dev,
2666                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2667                                                RTF_GATEWAY | RTF_DEFAULT);
2668
2669                 if (rt) {
2670                         /* Autoconf prefix route */
2671                         if (valid_lft == 0) {
2672                                 ip6_del_rt(net, rt);
2673                                 rt = NULL;
2674                         } else if (addrconf_finite_timeout(rt_expires)) {
2675                                 /* not infinity */
2676                                 fib6_set_expires(rt, jiffies + rt_expires);
2677                         } else {
2678                                 fib6_clean_expires(rt);
2679                         }
2680                 } else if (valid_lft) {
2681                         clock_t expires = 0;
2682                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2683                         if (addrconf_finite_timeout(rt_expires)) {
2684                                 /* not infinity */
2685                                 flags |= RTF_EXPIRES;
2686                                 expires = jiffies_to_clock_t(rt_expires);
2687                         }
2688                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2689                                               0, dev, expires, flags,
2690                                               GFP_ATOMIC);
2691                 }
2692                 fib6_info_release(rt);
2693         }
2694
2695         /* Try to figure out our local address for this prefix */
2696
2697         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2698                 struct in6_addr addr;
2699                 bool tokenized = false, dev_addr_generated = false;
2700
2701                 if (pinfo->prefix_len == 64) {
2702                         memcpy(&addr, &pinfo->prefix, 8);
2703
2704                         if (!ipv6_addr_any(&in6_dev->token)) {
2705                                 read_lock_bh(&in6_dev->lock);
2706                                 memcpy(addr.s6_addr + 8,
2707                                        in6_dev->token.s6_addr + 8, 8);
2708                                 read_unlock_bh(&in6_dev->lock);
2709                                 tokenized = true;
2710                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2711                                    !ipv6_generate_stable_address(&addr, 0,
2712                                                                  in6_dev)) {
2713                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2714                                 goto ok;
2715                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2716                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2717                                 goto put;
2718                         } else {
2719                                 dev_addr_generated = true;
2720                         }
2721                         goto ok;
2722                 }
2723                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2724                                     pinfo->prefix_len);
2725                 goto put;
2726
2727 ok:
2728                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2729                                                    &addr, addr_type,
2730                                                    addr_flags, sllao,
2731                                                    tokenized, valid_lft,
2732                                                    prefered_lft);
2733                 if (err)
2734                         goto put;
2735
2736                 /* Ignore error case here because previous prefix add addr was
2737                  * successful which will be notified.
2738                  */
2739                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2740                                               addr_type, addr_flags, sllao,
2741                                               tokenized, valid_lft,
2742                                               prefered_lft,
2743                                               dev_addr_generated);
2744         }
2745         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2746 put:
2747         in6_dev_put(in6_dev);
2748 }
2749
2750 /*
2751  *      Set destination address.
2752  *      Special case for SIT interfaces where we create a new "virtual"
2753  *      device.
2754  */
2755 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2756 {
2757         struct in6_ifreq ireq;
2758         struct net_device *dev;
2759         int err = -EINVAL;
2760
2761         rtnl_lock();
2762
2763         err = -EFAULT;
2764         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2765                 goto err_exit;
2766
2767         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2768
2769         err = -ENODEV;
2770         if (!dev)
2771                 goto err_exit;
2772
2773 #if IS_ENABLED(CONFIG_IPV6_SIT)
2774         if (dev->type == ARPHRD_SIT) {
2775                 const struct net_device_ops *ops = dev->netdev_ops;
2776                 struct ifreq ifr;
2777                 struct ip_tunnel_parm p;
2778
2779                 err = -EADDRNOTAVAIL;
2780                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2781                         goto err_exit;
2782
2783                 memset(&p, 0, sizeof(p));
2784                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2785                 p.iph.saddr = 0;
2786                 p.iph.version = 4;
2787                 p.iph.ihl = 5;
2788                 p.iph.protocol = IPPROTO_IPV6;
2789                 p.iph.ttl = 64;
2790                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2791
2792                 if (ops->ndo_do_ioctl) {
2793                         mm_segment_t oldfs = get_fs();
2794
2795                         set_fs(KERNEL_DS);
2796                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2797                         set_fs(oldfs);
2798                 } else
2799                         err = -EOPNOTSUPP;
2800
2801                 if (err == 0) {
2802                         err = -ENOBUFS;
2803                         dev = __dev_get_by_name(net, p.name);
2804                         if (!dev)
2805                                 goto err_exit;
2806                         err = dev_open(dev);
2807                 }
2808         }
2809 #endif
2810
2811 err_exit:
2812         rtnl_unlock();
2813         return err;
2814 }
2815
2816 static int ipv6_mc_config(struct sock *sk, bool join,
2817                           const struct in6_addr *addr, int ifindex)
2818 {
2819         int ret;
2820
2821         ASSERT_RTNL();
2822
2823         lock_sock(sk);
2824         if (join)
2825                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2826         else
2827                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2828         release_sock(sk);
2829
2830         return ret;
2831 }
2832
2833 /*
2834  *      Manual configuration of address on an interface
2835  */
2836 static int inet6_addr_add(struct net *net, int ifindex,
2837                           struct ifa6_config *cfg,
2838                           struct netlink_ext_ack *extack)
2839 {
2840         struct inet6_ifaddr *ifp;
2841         struct inet6_dev *idev;
2842         struct net_device *dev;
2843         unsigned long timeout;
2844         clock_t expires;
2845         u32 flags;
2846
2847         ASSERT_RTNL();
2848
2849         if (cfg->plen > 128)
2850                 return -EINVAL;
2851
2852         /* check the lifetime */
2853         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2854                 return -EINVAL;
2855
2856         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->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 (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2868                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2869                                          true, cfg->pfx, ifindex);
2870
2871                 if (ret < 0)
2872                         return ret;
2873         }
2874
2875         cfg->scope = ipv6_addr_scope(cfg->pfx);
2876
2877         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2878         if (addrconf_finite_timeout(timeout)) {
2879                 expires = jiffies_to_clock_t(timeout * HZ);
2880                 cfg->valid_lft = timeout;
2881                 flags = RTF_EXPIRES;
2882         } else {
2883                 expires = 0;
2884                 flags = 0;
2885                 cfg->ifa_flags |= IFA_F_PERMANENT;
2886         }
2887
2888         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2889         if (addrconf_finite_timeout(timeout)) {
2890                 if (timeout == 0)
2891                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2892                 cfg->preferred_lft = timeout;
2893         }
2894
2895         ifp = ipv6_add_addr(idev, cfg, true, extack);
2896         if (!IS_ERR(ifp)) {
2897                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2898                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2899                                               ifp->rt_priority, dev, expires,
2900                                               flags, GFP_KERNEL);
2901                 }
2902
2903                 /* Send a netlink notification if DAD is enabled and
2904                  * optimistic flag is not set
2905                  */
2906                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2907                         ipv6_ifa_notify(0, ifp);
2908                 /*
2909                  * Note that section 3.1 of RFC 4429 indicates
2910                  * that the Optimistic flag should not be set for
2911                  * manually configured addresses
2912                  */
2913                 addrconf_dad_start(ifp);
2914                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2915                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2916                                          cfg->preferred_lft, true, jiffies);
2917                 in6_ifa_put(ifp);
2918                 addrconf_verify_rtnl();
2919                 return 0;
2920         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2921                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2922                                cfg->pfx, ifindex);
2923         }
2924
2925         return PTR_ERR(ifp);
2926 }
2927
2928 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2929                           const struct in6_addr *pfx, unsigned int plen)
2930 {
2931         struct inet6_ifaddr *ifp;
2932         struct inet6_dev *idev;
2933         struct net_device *dev;
2934
2935         if (plen > 128)
2936                 return -EINVAL;
2937
2938         dev = __dev_get_by_index(net, ifindex);
2939         if (!dev)
2940                 return -ENODEV;
2941
2942         idev = __in6_dev_get(dev);
2943         if (!idev)
2944                 return -ENXIO;
2945
2946         read_lock_bh(&idev->lock);
2947         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2948                 if (ifp->prefix_len == plen &&
2949                     ipv6_addr_equal(pfx, &ifp->addr)) {
2950                         in6_ifa_hold(ifp);
2951                         read_unlock_bh(&idev->lock);
2952
2953                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
2954                             (ifa_flags & IFA_F_MANAGETEMPADDR))
2955                                 manage_tempaddrs(idev, ifp, 0, 0, false,
2956                                                  jiffies);
2957                         ipv6_del_addr(ifp);
2958                         addrconf_verify_rtnl();
2959                         if (ipv6_addr_is_multicast(pfx)) {
2960                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2961                                                false, pfx, dev->ifindex);
2962                         }
2963                         return 0;
2964                 }
2965         }
2966         read_unlock_bh(&idev->lock);
2967         return -EADDRNOTAVAIL;
2968 }
2969
2970
2971 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2972 {
2973         struct ifa6_config cfg = {
2974                 .ifa_flags = IFA_F_PERMANENT,
2975                 .preferred_lft = INFINITY_LIFE_TIME,
2976                 .valid_lft = INFINITY_LIFE_TIME,
2977         };
2978         struct in6_ifreq ireq;
2979         int err;
2980
2981         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2982                 return -EPERM;
2983
2984         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2985                 return -EFAULT;
2986
2987         cfg.pfx = &ireq.ifr6_addr;
2988         cfg.plen = ireq.ifr6_prefixlen;
2989
2990         rtnl_lock();
2991         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
2992         rtnl_unlock();
2993         return err;
2994 }
2995
2996 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2997 {
2998         struct in6_ifreq ireq;
2999         int err;
3000
3001         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3002                 return -EPERM;
3003
3004         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3005                 return -EFAULT;
3006
3007         rtnl_lock();
3008         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3009                              ireq.ifr6_prefixlen);
3010         rtnl_unlock();
3011         return err;
3012 }
3013
3014 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3015                      int plen, int scope)
3016 {
3017         struct inet6_ifaddr *ifp;
3018         struct ifa6_config cfg = {
3019                 .pfx = addr,
3020                 .plen = plen,
3021                 .ifa_flags = IFA_F_PERMANENT,
3022                 .valid_lft = INFINITY_LIFE_TIME,
3023                 .preferred_lft = INFINITY_LIFE_TIME,
3024                 .scope = scope
3025         };
3026
3027         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3028         if (!IS_ERR(ifp)) {
3029                 spin_lock_bh(&ifp->lock);
3030                 ifp->flags &= ~IFA_F_TENTATIVE;
3031                 spin_unlock_bh(&ifp->lock);
3032                 rt_genid_bump_ipv6(dev_net(idev->dev));
3033                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3034                 in6_ifa_put(ifp);
3035         }
3036 }
3037
3038 #if IS_ENABLED(CONFIG_IPV6_SIT)
3039 static void sit_add_v4_addrs(struct inet6_dev *idev)
3040 {
3041         struct in6_addr addr;
3042         struct net_device *dev;
3043         struct net *net = dev_net(idev->dev);
3044         int scope, plen;
3045         u32 pflags = 0;
3046
3047         ASSERT_RTNL();
3048
3049         memset(&addr, 0, sizeof(struct in6_addr));
3050         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3051
3052         if (idev->dev->flags&IFF_POINTOPOINT) {
3053                 addr.s6_addr32[0] = htonl(0xfe800000);
3054                 scope = IFA_LINK;
3055                 plen = 64;
3056         } else {
3057                 scope = IPV6_ADDR_COMPATv4;
3058                 plen = 96;
3059                 pflags |= RTF_NONEXTHOP;
3060         }
3061
3062         if (addr.s6_addr32[3]) {
3063                 add_addr(idev, &addr, plen, scope);
3064                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3065                                       GFP_KERNEL);
3066                 return;
3067         }
3068
3069         for_each_netdev(net, dev) {
3070                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3071                 if (in_dev && (dev->flags & IFF_UP)) {
3072                         struct in_ifaddr *ifa;
3073
3074                         int flag = scope;
3075
3076                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3077
3078                                 addr.s6_addr32[3] = ifa->ifa_local;
3079
3080                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3081                                         continue;
3082                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3083                                         if (idev->dev->flags&IFF_POINTOPOINT)
3084                                                 continue;
3085                                         flag |= IFA_HOST;
3086                                 }
3087
3088                                 add_addr(idev, &addr, plen, flag);
3089                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3090                                                       0, pflags, GFP_KERNEL);
3091                         }
3092                 }
3093         }
3094 }
3095 #endif
3096
3097 static void init_loopback(struct net_device *dev)
3098 {
3099         struct inet6_dev  *idev;
3100
3101         /* ::1 */
3102
3103         ASSERT_RTNL();
3104
3105         idev = ipv6_find_idev(dev);
3106         if (!idev) {
3107                 pr_debug("%s: add_dev failed\n", __func__);
3108                 return;
3109         }
3110
3111         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3112 }
3113
3114 void addrconf_add_linklocal(struct inet6_dev *idev,
3115                             const struct in6_addr *addr, u32 flags)
3116 {
3117         struct ifa6_config cfg = {
3118                 .pfx = addr,
3119                 .plen = 64,
3120                 .ifa_flags = flags | IFA_F_PERMANENT,
3121                 .valid_lft = INFINITY_LIFE_TIME,
3122                 .preferred_lft = INFINITY_LIFE_TIME,
3123                 .scope = IFA_LINK
3124         };
3125         struct inet6_ifaddr *ifp;
3126
3127 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3128         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3129              idev->cnf.optimistic_dad) &&
3130             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3131                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3132 #endif
3133
3134         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3135         if (!IS_ERR(ifp)) {
3136                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3137                                       0, 0, GFP_ATOMIC);
3138                 addrconf_dad_start(ifp);
3139                 in6_ifa_put(ifp);
3140         }
3141 }
3142 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3143
3144 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3145 {
3146         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3147                 return true;
3148
3149         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3150             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3151                 return true;
3152
3153         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3154             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3155                 return true;
3156
3157         return false;
3158 }
3159
3160 static int ipv6_generate_stable_address(struct in6_addr *address,
3161                                         u8 dad_count,
3162                                         const struct inet6_dev *idev)
3163 {
3164         static DEFINE_SPINLOCK(lock);
3165         static __u32 digest[SHA_DIGEST_WORDS];
3166         static __u32 workspace[SHA_WORKSPACE_WORDS];
3167
3168         static union {
3169                 char __data[SHA_MESSAGE_BYTES];
3170                 struct {
3171                         struct in6_addr secret;
3172                         __be32 prefix[2];
3173                         unsigned char hwaddr[MAX_ADDR_LEN];
3174                         u8 dad_count;
3175                 } __packed;
3176         } data;
3177
3178         struct in6_addr secret;
3179         struct in6_addr temp;
3180         struct net *net = dev_net(idev->dev);
3181
3182         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3183
3184         if (idev->cnf.stable_secret.initialized)
3185                 secret = idev->cnf.stable_secret.secret;
3186         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3187                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3188         else
3189                 return -1;
3190
3191 retry:
3192         spin_lock_bh(&lock);
3193
3194         sha_init(digest);
3195         memset(&data, 0, sizeof(data));
3196         memset(workspace, 0, sizeof(workspace));
3197         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3198         data.prefix[0] = address->s6_addr32[0];
3199         data.prefix[1] = address->s6_addr32[1];
3200         data.secret = secret;
3201         data.dad_count = dad_count;
3202
3203         sha_transform(digest, data.__data, workspace);
3204
3205         temp = *address;
3206         temp.s6_addr32[2] = (__force __be32)digest[0];
3207         temp.s6_addr32[3] = (__force __be32)digest[1];
3208
3209         spin_unlock_bh(&lock);
3210
3211         if (ipv6_reserved_interfaceid(temp)) {
3212                 dad_count++;
3213                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3214                         return -1;
3215                 goto retry;
3216         }
3217
3218         *address = temp;
3219         return 0;
3220 }
3221
3222 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3223 {
3224         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3225
3226         if (s->initialized)
3227                 return;
3228         s = &idev->cnf.stable_secret;
3229         get_random_bytes(&s->secret, sizeof(s->secret));
3230         s->initialized = true;
3231 }
3232
3233 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3234 {
3235         struct in6_addr addr;
3236
3237         /* no link local addresses on L3 master devices */
3238         if (netif_is_l3_master(idev->dev))
3239                 return;
3240
3241         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3242
3243         switch (idev->cnf.addr_gen_mode) {
3244         case IN6_ADDR_GEN_MODE_RANDOM:
3245                 ipv6_gen_mode_random_init(idev);
3246                 /* fallthrough */
3247         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3248                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3249                         addrconf_add_linklocal(idev, &addr,
3250                                                IFA_F_STABLE_PRIVACY);
3251                 else if (prefix_route)
3252                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3253                                               0, 0, GFP_KERNEL);
3254                 break;
3255         case IN6_ADDR_GEN_MODE_EUI64:
3256                 /* addrconf_add_linklocal also adds a prefix_route and we
3257                  * only need to care about prefix routes if ipv6_generate_eui64
3258                  * couldn't generate one.
3259                  */
3260                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3261                         addrconf_add_linklocal(idev, &addr, 0);
3262                 else if (prefix_route)
3263                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3264                                               0, 0, GFP_KERNEL);
3265                 break;
3266         case IN6_ADDR_GEN_MODE_NONE:
3267         default:
3268                 /* will not add any link local address */
3269                 break;
3270         }
3271 }
3272
3273 static void addrconf_dev_config(struct net_device *dev)
3274 {
3275         struct inet6_dev *idev;
3276
3277         ASSERT_RTNL();
3278
3279         if ((dev->type != ARPHRD_ETHER) &&
3280             (dev->type != ARPHRD_FDDI) &&
3281             (dev->type != ARPHRD_ARCNET) &&
3282             (dev->type != ARPHRD_INFINIBAND) &&
3283             (dev->type != ARPHRD_IEEE1394) &&
3284             (dev->type != ARPHRD_TUNNEL6) &&
3285             (dev->type != ARPHRD_6LOWPAN) &&
3286             (dev->type != ARPHRD_IP6GRE) &&
3287             (dev->type != ARPHRD_IPGRE) &&
3288             (dev->type != ARPHRD_TUNNEL) &&
3289             (dev->type != ARPHRD_NONE) &&
3290             (dev->type != ARPHRD_RAWIP)) {
3291                 /* Alas, we support only Ethernet autoconfiguration. */
3292                 return;
3293         }
3294
3295         idev = addrconf_add_dev(dev);
3296         if (IS_ERR(idev))
3297                 return;
3298
3299         /* this device type has no EUI support */
3300         if (dev->type == ARPHRD_NONE &&
3301             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3302                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3303
3304         addrconf_addr_gen(idev, false);
3305 }
3306
3307 #if IS_ENABLED(CONFIG_IPV6_SIT)
3308 static void addrconf_sit_config(struct net_device *dev)
3309 {
3310         struct inet6_dev *idev;
3311
3312         ASSERT_RTNL();
3313
3314         /*
3315          * Configure the tunnel with one of our IPv4
3316          * addresses... we should configure all of
3317          * our v4 addrs in the tunnel
3318          */
3319
3320         idev = ipv6_find_idev(dev);
3321         if (!idev) {
3322                 pr_debug("%s: add_dev failed\n", __func__);
3323                 return;
3324         }
3325
3326         if (dev->priv_flags & IFF_ISATAP) {
3327                 addrconf_addr_gen(idev, false);
3328                 return;
3329         }
3330
3331         sit_add_v4_addrs(idev);
3332
3333         if (dev->flags&IFF_POINTOPOINT)
3334                 addrconf_add_mroute(dev);
3335 }
3336 #endif
3337
3338 #if IS_ENABLED(CONFIG_NET_IPGRE)
3339 static void addrconf_gre_config(struct net_device *dev)
3340 {
3341         struct inet6_dev *idev;
3342
3343         ASSERT_RTNL();
3344
3345         idev = ipv6_find_idev(dev);
3346         if (!idev) {
3347                 pr_debug("%s: add_dev failed\n", __func__);
3348                 return;
3349         }
3350
3351         addrconf_addr_gen(idev, true);
3352         if (dev->flags & IFF_POINTOPOINT)
3353                 addrconf_add_mroute(dev);
3354 }
3355 #endif
3356
3357 static int fixup_permanent_addr(struct net *net,
3358                                 struct inet6_dev *idev,
3359                                 struct inet6_ifaddr *ifp)
3360 {
3361         /* !fib6_node means the host route was removed from the
3362          * FIB, for example, if 'lo' device is taken down. In that
3363          * case regenerate the host route.
3364          */
3365         if (!ifp->rt || !ifp->rt->fib6_node) {
3366                 struct fib6_info *f6i, *prev;
3367
3368                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3369                                          GFP_ATOMIC);
3370                 if (IS_ERR(f6i))
3371                         return PTR_ERR(f6i);
3372
3373                 /* ifp->rt can be accessed outside of rtnl */
3374                 spin_lock(&ifp->lock);
3375                 prev = ifp->rt;
3376                 ifp->rt = f6i;
3377                 spin_unlock(&ifp->lock);
3378
3379                 fib6_info_release(prev);
3380         }
3381
3382         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3383                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3384                                       ifp->rt_priority, idev->dev, 0, 0,
3385                                       GFP_ATOMIC);
3386         }
3387
3388         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3389                 addrconf_dad_start(ifp);
3390
3391         return 0;
3392 }
3393
3394 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3395 {
3396         struct inet6_ifaddr *ifp, *tmp;
3397         struct inet6_dev *idev;
3398
3399         idev = __in6_dev_get(dev);
3400         if (!idev)
3401                 return;
3402
3403         write_lock_bh(&idev->lock);
3404
3405         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3406                 if ((ifp->flags & IFA_F_PERMANENT) &&
3407                     fixup_permanent_addr(net, idev, ifp) < 0) {
3408                         write_unlock_bh(&idev->lock);
3409                         in6_ifa_hold(ifp);
3410                         ipv6_del_addr(ifp);
3411                         write_lock_bh(&idev->lock);
3412
3413                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3414                                              idev->dev->name, &ifp->addr);
3415                 }
3416         }
3417
3418         write_unlock_bh(&idev->lock);
3419 }
3420
3421 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3422                            void *ptr)
3423 {
3424         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3425         struct netdev_notifier_changeupper_info *info;
3426         struct inet6_dev *idev = __in6_dev_get(dev);
3427         struct net *net = dev_net(dev);
3428         int run_pending = 0;
3429         int err;
3430
3431         switch (event) {
3432         case NETDEV_REGISTER:
3433                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3434                         idev = ipv6_add_dev(dev);
3435                         if (IS_ERR(idev))
3436                                 return notifier_from_errno(PTR_ERR(idev));
3437                 }
3438                 break;
3439
3440         case NETDEV_CHANGEMTU:
3441                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3442                 if (dev->mtu < IPV6_MIN_MTU) {
3443                         addrconf_ifdown(dev, dev != net->loopback_dev);
3444                         break;
3445                 }
3446
3447                 if (idev) {
3448                         rt6_mtu_change(dev, dev->mtu);
3449                         idev->cnf.mtu6 = dev->mtu;
3450                         break;
3451                 }
3452
3453                 /* allocate new idev */
3454                 idev = ipv6_add_dev(dev);
3455                 if (IS_ERR(idev))
3456                         break;
3457
3458                 /* device is still not ready */
3459                 if (!(idev->if_flags & IF_READY))
3460                         break;
3461
3462                 run_pending = 1;
3463
3464                 /* fall through */
3465
3466         case NETDEV_UP:
3467         case NETDEV_CHANGE:
3468                 if (dev->flags & IFF_SLAVE)
3469                         break;
3470
3471                 if (idev && idev->cnf.disable_ipv6)
3472                         break;
3473
3474                 if (event == NETDEV_UP) {
3475                         /* restore routes for permanent addresses */
3476                         addrconf_permanent_addr(net, dev);
3477
3478                         if (!addrconf_link_ready(dev)) {
3479                                 /* device is not ready yet. */
3480                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3481                                         dev->name);
3482                                 break;
3483                         }
3484
3485                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3486                                 idev = ipv6_add_dev(dev);
3487
3488                         if (!IS_ERR_OR_NULL(idev)) {
3489                                 idev->if_flags |= IF_READY;
3490                                 run_pending = 1;
3491                         }
3492                 } else if (event == NETDEV_CHANGE) {
3493                         if (!addrconf_link_ready(dev)) {
3494                                 /* device is still not ready. */
3495                                 rt6_sync_down_dev(dev, event);
3496                                 break;
3497                         }
3498
3499                         if (idev) {
3500                                 if (idev->if_flags & IF_READY) {
3501                                         /* device is already configured -
3502                                          * but resend MLD reports, we might
3503                                          * have roamed and need to update
3504                                          * multicast snooping switches
3505                                          */
3506                                         ipv6_mc_up(idev);
3507                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3508                                         break;
3509                                 }
3510                                 idev->if_flags |= IF_READY;
3511                         }
3512
3513                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3514                                 dev->name);
3515
3516                         run_pending = 1;
3517                 }
3518
3519                 switch (dev->type) {
3520 #if IS_ENABLED(CONFIG_IPV6_SIT)
3521                 case ARPHRD_SIT:
3522                         addrconf_sit_config(dev);
3523                         break;
3524 #endif
3525 #if IS_ENABLED(CONFIG_NET_IPGRE)
3526                 case ARPHRD_IPGRE:
3527                         addrconf_gre_config(dev);
3528                         break;
3529 #endif
3530                 case ARPHRD_LOOPBACK:
3531                         init_loopback(dev);
3532                         break;
3533
3534                 default:
3535                         addrconf_dev_config(dev);
3536                         break;
3537                 }
3538
3539                 if (!IS_ERR_OR_NULL(idev)) {
3540                         if (run_pending)
3541                                 addrconf_dad_run(idev);
3542
3543                         /* Device has an address by now */
3544                         rt6_sync_up(dev, RTNH_F_DEAD);
3545
3546                         /*
3547                          * If the MTU changed during the interface down,
3548                          * when the interface up, the changed MTU must be
3549                          * reflected in the idev as well as routers.
3550                          */
3551                         if (idev->cnf.mtu6 != dev->mtu &&
3552                             dev->mtu >= IPV6_MIN_MTU) {
3553                                 rt6_mtu_change(dev, dev->mtu);
3554                                 idev->cnf.mtu6 = dev->mtu;
3555                         }
3556                         idev->tstamp = jiffies;
3557                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3558
3559                         /*
3560                          * If the changed mtu during down is lower than
3561                          * IPV6_MIN_MTU stop IPv6 on this interface.
3562                          */
3563                         if (dev->mtu < IPV6_MIN_MTU)
3564                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3565                 }
3566                 break;
3567
3568         case NETDEV_DOWN:
3569         case NETDEV_UNREGISTER:
3570                 /*
3571                  *      Remove all addresses from this interface.
3572                  */
3573                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3574                 break;
3575
3576         case NETDEV_CHANGENAME:
3577                 if (idev) {
3578                         snmp6_unregister_dev(idev);
3579                         addrconf_sysctl_unregister(idev);
3580                         err = addrconf_sysctl_register(idev);
3581                         if (err)
3582                                 return notifier_from_errno(err);
3583                         err = snmp6_register_dev(idev);
3584                         if (err) {
3585                                 addrconf_sysctl_unregister(idev);
3586                                 return notifier_from_errno(err);
3587                         }
3588                 }
3589                 break;
3590
3591         case NETDEV_PRE_TYPE_CHANGE:
3592         case NETDEV_POST_TYPE_CHANGE:
3593                 if (idev)
3594                         addrconf_type_change(dev, event);
3595                 break;
3596
3597         case NETDEV_CHANGEUPPER:
3598                 info = ptr;
3599
3600                 /* flush all routes if dev is linked to or unlinked from
3601                  * an L3 master device (e.g., VRF)
3602                  */
3603                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3604                         addrconf_ifdown(dev, 0);
3605         }
3606
3607         return NOTIFY_OK;
3608 }
3609
3610 /*
3611  *      addrconf module should be notified of a device going up
3612  */
3613 static struct notifier_block ipv6_dev_notf = {
3614         .notifier_call = addrconf_notify,
3615         .priority = ADDRCONF_NOTIFY_PRIORITY,
3616 };
3617
3618 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3619 {
3620         struct inet6_dev *idev;
3621         ASSERT_RTNL();
3622
3623         idev = __in6_dev_get(dev);
3624
3625         if (event == NETDEV_POST_TYPE_CHANGE)
3626                 ipv6_mc_remap(idev);
3627         else if (event == NETDEV_PRE_TYPE_CHANGE)
3628                 ipv6_mc_unmap(idev);
3629 }
3630
3631 static bool addr_is_local(const struct in6_addr *addr)
3632 {
3633         return ipv6_addr_type(addr) &
3634                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3635 }
3636
3637 static int addrconf_ifdown(struct net_device *dev, int how)
3638 {
3639         unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3640         struct net *net = dev_net(dev);
3641         struct inet6_dev *idev;
3642         struct inet6_ifaddr *ifa, *tmp;
3643         bool keep_addr = false;
3644         int state, i;
3645
3646         ASSERT_RTNL();
3647
3648         rt6_disable_ip(dev, event);
3649
3650         idev = __in6_dev_get(dev);
3651         if (!idev)
3652                 return -ENODEV;
3653
3654         /*
3655          * Step 1: remove reference to ipv6 device from parent device.
3656          *         Do not dev_put!
3657          */
3658         if (how) {
3659                 idev->dead = 1;
3660
3661                 /* protected by rtnl_lock */
3662                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3663
3664                 /* Step 1.5: remove snmp6 entry */
3665                 snmp6_unregister_dev(idev);
3666
3667         }
3668
3669         /* combine the user config with event to determine if permanent
3670          * addresses are to be removed from address hash table
3671          */
3672         if (!how && !idev->cnf.disable_ipv6) {
3673                 /* aggregate the system setting and interface setting */
3674                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3675
3676                 if (!_keep_addr)
3677                         _keep_addr = idev->cnf.keep_addr_on_down;
3678
3679                 keep_addr = (_keep_addr > 0);
3680         }
3681
3682         /* Step 2: clear hash table */
3683         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3684                 struct hlist_head *h = &inet6_addr_lst[i];
3685
3686                 spin_lock_bh(&addrconf_hash_lock);
3687 restart:
3688                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3689                         if (ifa->idev == idev) {
3690                                 addrconf_del_dad_work(ifa);
3691                                 /* combined flag + permanent flag decide if
3692                                  * address is retained on a down event
3693                                  */
3694                                 if (!keep_addr ||
3695                                     !(ifa->flags & IFA_F_PERMANENT) ||
3696                                     addr_is_local(&ifa->addr)) {
3697                                         hlist_del_init_rcu(&ifa->addr_lst);
3698                                         goto restart;
3699                                 }
3700                         }
3701                 }
3702                 spin_unlock_bh(&addrconf_hash_lock);
3703         }
3704
3705         write_lock_bh(&idev->lock);
3706
3707         addrconf_del_rs_timer(idev);
3708
3709         /* Step 2: clear flags for stateless addrconf */
3710         if (!how)
3711                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3712
3713         /* Step 3: clear tempaddr list */
3714         while (!list_empty(&idev->tempaddr_list)) {
3715                 ifa = list_first_entry(&idev->tempaddr_list,
3716                                        struct inet6_ifaddr, tmp_list);
3717                 list_del(&ifa->tmp_list);
3718                 write_unlock_bh(&idev->lock);
3719                 spin_lock_bh(&ifa->lock);
3720
3721                 if (ifa->ifpub) {
3722                         in6_ifa_put(ifa->ifpub);
3723                         ifa->ifpub = NULL;
3724                 }
3725                 spin_unlock_bh(&ifa->lock);
3726                 in6_ifa_put(ifa);
3727                 write_lock_bh(&idev->lock);
3728         }
3729
3730         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3731                 struct fib6_info *rt = NULL;
3732                 bool keep;
3733
3734                 addrconf_del_dad_work(ifa);
3735
3736                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3737                         !addr_is_local(&ifa->addr);
3738
3739                 write_unlock_bh(&idev->lock);
3740                 spin_lock_bh(&ifa->lock);
3741
3742                 if (keep) {
3743                         /* set state to skip the notifier below */
3744                         state = INET6_IFADDR_STATE_DEAD;
3745                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3746                         if (!(ifa->flags & IFA_F_NODAD))
3747                                 ifa->flags |= IFA_F_TENTATIVE;
3748
3749                         rt = ifa->rt;
3750                         ifa->rt = NULL;
3751                 } else {
3752                         state = ifa->state;
3753                         ifa->state = INET6_IFADDR_STATE_DEAD;
3754                 }
3755
3756                 spin_unlock_bh(&ifa->lock);
3757
3758                 if (rt)
3759                         ip6_del_rt(net, rt);
3760
3761                 if (state != INET6_IFADDR_STATE_DEAD) {
3762                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3763                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3764                 } else {
3765                         if (idev->cnf.forwarding)
3766                                 addrconf_leave_anycast(ifa);
3767                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3768                 }
3769
3770                 write_lock_bh(&idev->lock);
3771                 if (!keep) {
3772                         list_del_rcu(&ifa->if_list);
3773                         in6_ifa_put(ifa);
3774                 }
3775         }
3776
3777         write_unlock_bh(&idev->lock);
3778
3779         /* Step 5: Discard anycast and multicast list */
3780         if (how) {
3781                 ipv6_ac_destroy_dev(idev);
3782                 ipv6_mc_destroy_dev(idev);
3783         } else {
3784                 ipv6_mc_down(idev);
3785         }
3786
3787         idev->tstamp = jiffies;
3788
3789         /* Last: Shot the device (if unregistered) */
3790         if (how) {
3791                 addrconf_sysctl_unregister(idev);
3792                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3793                 neigh_ifdown(&nd_tbl, dev);
3794                 in6_dev_put(idev);
3795         }
3796         return 0;
3797 }
3798
3799 static void addrconf_rs_timer(struct timer_list *t)
3800 {
3801         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3802         struct net_device *dev = idev->dev;
3803         struct in6_addr lladdr;
3804
3805         write_lock(&idev->lock);
3806         if (idev->dead || !(idev->if_flags & IF_READY))
3807                 goto out;
3808
3809         if (!ipv6_accept_ra(idev))
3810                 goto out;
3811
3812         /* Announcement received after solicitation was sent */
3813         if (idev->if_flags & IF_RA_RCVD)
3814                 goto out;
3815
3816         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3817                 write_unlock(&idev->lock);
3818                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3819                         ndisc_send_rs(dev, &lladdr,
3820                                       &in6addr_linklocal_allrouters);
3821                 else
3822                         goto put;
3823
3824                 write_lock(&idev->lock);
3825                 idev->rs_interval = rfc3315_s14_backoff_update(
3826                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3827                 /* The wait after the last probe can be shorter */
3828                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3829                                              idev->cnf.rtr_solicits) ?
3830                                       idev->cnf.rtr_solicit_delay :
3831                                       idev->rs_interval);
3832         } else {
3833                 /*
3834                  * Note: we do not support deprecated "all on-link"
3835                  * assumption any longer.
3836                  */
3837                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3838         }
3839
3840 out:
3841         write_unlock(&idev->lock);
3842 put:
3843         in6_dev_put(idev);
3844 }
3845
3846 /*
3847  *      Duplicate Address Detection
3848  */
3849 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3850 {
3851         unsigned long rand_num;
3852         struct inet6_dev *idev = ifp->idev;
3853         u64 nonce;
3854
3855         if (ifp->flags & IFA_F_OPTIMISTIC)
3856                 rand_num = 0;
3857         else
3858                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3859
3860         nonce = 0;
3861         if (idev->cnf.enhanced_dad ||
3862             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3863                 do
3864                         get_random_bytes(&nonce, 6);
3865                 while (nonce == 0);
3866         }
3867         ifp->dad_nonce = nonce;
3868         ifp->dad_probes = idev->cnf.dad_transmits;
3869         addrconf_mod_dad_work(ifp, rand_num);
3870 }
3871
3872 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3873 {
3874         struct inet6_dev *idev = ifp->idev;
3875         struct net_device *dev = idev->dev;
3876         bool bump_id, notify = false;
3877         struct net *net;
3878
3879         addrconf_join_solict(dev, &ifp->addr);
3880
3881         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3882
3883         read_lock_bh(&idev->lock);
3884         spin_lock(&ifp->lock);
3885         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3886                 goto out;
3887
3888         net = dev_net(dev);
3889         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3890             (net->ipv6.devconf_all->accept_dad < 1 &&
3891              idev->cnf.accept_dad < 1) ||
3892             !(ifp->flags&IFA_F_TENTATIVE) ||
3893             ifp->flags & IFA_F_NODAD) {
3894                 bool send_na = false;
3895
3896                 if (ifp->flags & IFA_F_TENTATIVE &&
3897                     !(ifp->flags & IFA_F_OPTIMISTIC))
3898                         send_na = true;
3899                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3900                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3901                 spin_unlock(&ifp->lock);
3902                 read_unlock_bh(&idev->lock);
3903
3904                 addrconf_dad_completed(ifp, bump_id, send_na);
3905                 return;
3906         }
3907
3908         if (!(idev->if_flags & IF_READY)) {
3909                 spin_unlock(&ifp->lock);
3910                 read_unlock_bh(&idev->lock);
3911                 /*
3912                  * If the device is not ready:
3913                  * - keep it tentative if it is a permanent address.
3914                  * - otherwise, kill it.
3915                  */
3916                 in6_ifa_hold(ifp);
3917                 a