Merge branch 'misc.namei' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
[linux-2.6-microblaze.git] / net / ipv6 / addrconf.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *      IPv6 Address [auto]configuration
4  *      Linux INET6 implementation
5  *
6  *      Authors:
7  *      Pedro Roque             <roque@di.fc.ul.pt>
8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
9  */
10
11 /*
12  *      Changes:
13  *
14  *      Janos Farkas                    :       delete timer on ifdown
15  *      <chexum@bankinf.banki.hu>
16  *      Andi Kleen                      :       kill double kfree on module
17  *                                              unload.
18  *      Maciej W. Rozycki               :       FDDI support
19  *      sekiya@USAGI                    :       Don't send too many RS
20  *                                              packets.
21  *      yoshfuji@USAGI                  :       Fixed interval between DAD
22  *                                              packets.
23  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
24  *                                              address validation timer.
25  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
26  *                                              support.
27  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
28  *                                              address on a same interface.
29  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
30  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
31  *                                              seq_file.
32  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
33  *                                              selection; consider scope,
34  *                                              status etc.
35  */
36
37 #define pr_fmt(fmt) "IPv6: " fmt
38
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
65
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
69
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
88
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
92 #include <linux/ioam6.h>
93
94 #define INFINITY_LIFE_TIME      0xFFFFFFFF
95
96 #define IPV6_MAX_STRLEN \
97         sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98
99 static inline u32 cstamp_delta(unsigned long cstamp)
100 {
101         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
102 }
103
104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 {
106         /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107         u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
108         do_div(tmp, 1000000);
109         return (s32)tmp;
110 }
111
112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 {
114         /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115         u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
116         do_div(tmp, 1000000);
117         if ((s32)tmp > mrt) {
118                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119                 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
120                 do_div(tmp, 1000000);
121         }
122         return (s32)tmp;
123 }
124
125 #ifdef CONFIG_SYSCTL
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
128 #else
129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 {
131         return 0;
132 }
133
134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
135 {
136 }
137 #endif
138
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
140
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144                                         u8 dad_count,
145                                         const struct inet6_dev *idev);
146
147 #define IN6_ADDR_HSIZE_SHIFT    8
148 #define IN6_ADDR_HSIZE          (1 << IN6_ADDR_HSIZE_SHIFT)
149 /*
150  *      Configured unicast address hash table
151  */
152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
153 static DEFINE_SPINLOCK(addrconf_hash_lock);
154
155 static void addrconf_verify(void);
156 static void addrconf_verify_rtnl(void);
157 static void addrconf_verify_work(struct work_struct *);
158
159 static struct workqueue_struct *addrconf_wq;
160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
161
162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
164
165 static void addrconf_type_change(struct net_device *dev,
166                                  unsigned long event);
167 static int addrconf_ifdown(struct net_device *dev, bool unregister);
168
169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
170                                                   int plen,
171                                                   const struct net_device *dev,
172                                                   u32 flags, u32 noflags,
173                                                   bool no_gw);
174
175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
176 static void addrconf_dad_work(struct work_struct *w);
177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
178                                    bool send_na);
179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
180 static void addrconf_rs_timer(struct timer_list *t);
181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
183
184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
185                                 struct prefix_info *pinfo);
186
187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
188         .forwarding             = 0,
189         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
190         .mtu6                   = IPV6_MIN_MTU,
191         .accept_ra              = 1,
192         .accept_redirects       = 1,
193         .autoconf               = 1,
194         .force_mld_version      = 0,
195         .mldv1_unsolicited_report_interval = 10 * HZ,
196         .mldv2_unsolicited_report_interval = HZ,
197         .dad_transmits          = 1,
198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
200         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
202         .use_tempaddr           = 0,
203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
205         .regen_max_retry        = REGEN_MAX_RETRY,
206         .max_desync_factor      = MAX_DESYNC_FACTOR,
207         .max_addresses          = IPV6_MAX_ADDRESSES,
208         .accept_ra_defrtr       = 1,
209         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
210         .accept_ra_from_local   = 0,
211         .accept_ra_min_hop_limit= 1,
212         .accept_ra_pinfo        = 1,
213 #ifdef CONFIG_IPV6_ROUTER_PREF
214         .accept_ra_rtr_pref     = 1,
215         .rtr_probe_interval     = 60 * HZ,
216 #ifdef CONFIG_IPV6_ROUTE_INFO
217         .accept_ra_rt_info_min_plen = 0,
218         .accept_ra_rt_info_max_plen = 0,
219 #endif
220 #endif
221         .proxy_ndp              = 0,
222         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
223         .disable_ipv6           = 0,
224         .accept_dad             = 0,
225         .suppress_frag_ndisc    = 1,
226         .accept_ra_mtu          = 1,
227         .stable_secret          = {
228                 .initialized = false,
229         },
230         .use_oif_addrs_only     = 0,
231         .ignore_routes_with_linkdown = 0,
232         .keep_addr_on_down      = 0,
233         .seg6_enabled           = 0,
234 #ifdef CONFIG_IPV6_SEG6_HMAC
235         .seg6_require_hmac      = 0,
236 #endif
237         .enhanced_dad           = 1,
238         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
239         .disable_policy         = 0,
240         .rpl_seg_enabled        = 0,
241         .ioam6_enabled          = 0,
242         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
243         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
244 };
245
246 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
247         .forwarding             = 0,
248         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
249         .mtu6                   = IPV6_MIN_MTU,
250         .accept_ra              = 1,
251         .accept_redirects       = 1,
252         .autoconf               = 1,
253         .force_mld_version      = 0,
254         .mldv1_unsolicited_report_interval = 10 * HZ,
255         .mldv2_unsolicited_report_interval = HZ,
256         .dad_transmits          = 1,
257         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
258         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
259         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
260         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
261         .use_tempaddr           = 0,
262         .temp_valid_lft         = TEMP_VALID_LIFETIME,
263         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
264         .regen_max_retry        = REGEN_MAX_RETRY,
265         .max_desync_factor      = MAX_DESYNC_FACTOR,
266         .max_addresses          = IPV6_MAX_ADDRESSES,
267         .accept_ra_defrtr       = 1,
268         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
269         .accept_ra_from_local   = 0,
270         .accept_ra_min_hop_limit= 1,
271         .accept_ra_pinfo        = 1,
272 #ifdef CONFIG_IPV6_ROUTER_PREF
273         .accept_ra_rtr_pref     = 1,
274         .rtr_probe_interval     = 60 * HZ,
275 #ifdef CONFIG_IPV6_ROUTE_INFO
276         .accept_ra_rt_info_min_plen = 0,
277         .accept_ra_rt_info_max_plen = 0,
278 #endif
279 #endif
280         .proxy_ndp              = 0,
281         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
282         .disable_ipv6           = 0,
283         .accept_dad             = 1,
284         .suppress_frag_ndisc    = 1,
285         .accept_ra_mtu          = 1,
286         .stable_secret          = {
287                 .initialized = false,
288         },
289         .use_oif_addrs_only     = 0,
290         .ignore_routes_with_linkdown = 0,
291         .keep_addr_on_down      = 0,
292         .seg6_enabled           = 0,
293 #ifdef CONFIG_IPV6_SEG6_HMAC
294         .seg6_require_hmac      = 0,
295 #endif
296         .enhanced_dad           = 1,
297         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
298         .disable_policy         = 0,
299         .rpl_seg_enabled        = 0,
300         .ioam6_enabled          = 0,
301         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
302         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
303 };
304
305 /* Check if link is ready: is it up and is a valid qdisc available */
306 static inline bool addrconf_link_ready(const struct net_device *dev)
307 {
308         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
309 }
310
311 static void addrconf_del_rs_timer(struct inet6_dev *idev)
312 {
313         if (del_timer(&idev->rs_timer))
314                 __in6_dev_put(idev);
315 }
316
317 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
318 {
319         if (cancel_delayed_work(&ifp->dad_work))
320                 __in6_ifa_put(ifp);
321 }
322
323 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
324                                   unsigned long when)
325 {
326         if (!timer_pending(&idev->rs_timer))
327                 in6_dev_hold(idev);
328         mod_timer(&idev->rs_timer, jiffies + when);
329 }
330
331 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
332                                    unsigned long delay)
333 {
334         in6_ifa_hold(ifp);
335         if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
336                 in6_ifa_put(ifp);
337 }
338
339 static int snmp6_alloc_dev(struct inet6_dev *idev)
340 {
341         int i;
342
343         idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
344         if (!idev->stats.ipv6)
345                 goto err_ip;
346
347         for_each_possible_cpu(i) {
348                 struct ipstats_mib *addrconf_stats;
349                 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
350                 u64_stats_init(&addrconf_stats->syncp);
351         }
352
353
354         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
355                                         GFP_KERNEL);
356         if (!idev->stats.icmpv6dev)
357                 goto err_icmp;
358         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
359                                            GFP_KERNEL);
360         if (!idev->stats.icmpv6msgdev)
361                 goto err_icmpmsg;
362
363         return 0;
364
365 err_icmpmsg:
366         kfree(idev->stats.icmpv6dev);
367 err_icmp:
368         free_percpu(idev->stats.ipv6);
369 err_ip:
370         return -ENOMEM;
371 }
372
373 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
374 {
375         struct inet6_dev *ndev;
376         int err = -ENOMEM;
377
378         ASSERT_RTNL();
379
380         if (dev->mtu < IPV6_MIN_MTU)
381                 return ERR_PTR(-EINVAL);
382
383         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
384         if (!ndev)
385                 return ERR_PTR(err);
386
387         rwlock_init(&ndev->lock);
388         ndev->dev = dev;
389         INIT_LIST_HEAD(&ndev->addr_list);
390         timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
391         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
392
393         if (ndev->cnf.stable_secret.initialized)
394                 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
395
396         ndev->cnf.mtu6 = dev->mtu;
397         ndev->ra_mtu = 0;
398         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
399         if (!ndev->nd_parms) {
400                 kfree(ndev);
401                 return ERR_PTR(err);
402         }
403         if (ndev->cnf.forwarding)
404                 dev_disable_lro(dev);
405         /* We refer to the device */
406         dev_hold(dev);
407
408         if (snmp6_alloc_dev(ndev) < 0) {
409                 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
410                            __func__);
411                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
412                 dev_put(dev);
413                 kfree(ndev);
414                 return ERR_PTR(err);
415         }
416
417         if (snmp6_register_dev(ndev) < 0) {
418                 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
419                            __func__, dev->name);
420                 goto err_release;
421         }
422
423         /* One reference from device. */
424         refcount_set(&ndev->refcnt, 1);
425
426         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
427                 ndev->cnf.accept_dad = -1;
428
429 #if IS_ENABLED(CONFIG_IPV6_SIT)
430         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
431                 pr_info("%s: Disabled Multicast RS\n", dev->name);
432                 ndev->cnf.rtr_solicits = 0;
433         }
434 #endif
435
436         INIT_LIST_HEAD(&ndev->tempaddr_list);
437         ndev->desync_factor = U32_MAX;
438         if ((dev->flags&IFF_LOOPBACK) ||
439             dev->type == ARPHRD_TUNNEL ||
440             dev->type == ARPHRD_TUNNEL6 ||
441             dev->type == ARPHRD_SIT ||
442             dev->type == ARPHRD_NONE) {
443                 ndev->cnf.use_tempaddr = -1;
444         }
445
446         ndev->token = in6addr_any;
447
448         if (netif_running(dev) && addrconf_link_ready(dev))
449                 ndev->if_flags |= IF_READY;
450
451         ipv6_mc_init_dev(ndev);
452         ndev->tstamp = jiffies;
453         err = addrconf_sysctl_register(ndev);
454         if (err) {
455                 ipv6_mc_destroy_dev(ndev);
456                 snmp6_unregister_dev(ndev);
457                 goto err_release;
458         }
459         /* protected by rtnl_lock */
460         rcu_assign_pointer(dev->ip6_ptr, ndev);
461
462         /* Join interface-local all-node multicast group */
463         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
464
465         /* Join all-node multicast group */
466         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
467
468         /* Join all-router multicast group if forwarding is set */
469         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
470                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
471
472         return ndev;
473
474 err_release:
475         neigh_parms_release(&nd_tbl, ndev->nd_parms);
476         ndev->dead = 1;
477         in6_dev_finish_destroy(ndev);
478         return ERR_PTR(err);
479 }
480
481 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
482 {
483         struct inet6_dev *idev;
484
485         ASSERT_RTNL();
486
487         idev = __in6_dev_get(dev);
488         if (!idev) {
489                 idev = ipv6_add_dev(dev);
490                 if (IS_ERR(idev))
491                         return idev;
492         }
493
494         if (dev->flags&IFF_UP)
495                 ipv6_mc_up(idev);
496         return idev;
497 }
498
499 static int inet6_netconf_msgsize_devconf(int type)
500 {
501         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
502                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
503         bool all = false;
504
505         if (type == NETCONFA_ALL)
506                 all = true;
507
508         if (all || type == NETCONFA_FORWARDING)
509                 size += nla_total_size(4);
510 #ifdef CONFIG_IPV6_MROUTE
511         if (all || type == NETCONFA_MC_FORWARDING)
512                 size += nla_total_size(4);
513 #endif
514         if (all || type == NETCONFA_PROXY_NEIGH)
515                 size += nla_total_size(4);
516
517         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
518                 size += nla_total_size(4);
519
520         return size;
521 }
522
523 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
524                                       struct ipv6_devconf *devconf, u32 portid,
525                                       u32 seq, int event, unsigned int flags,
526                                       int type)
527 {
528         struct nlmsghdr  *nlh;
529         struct netconfmsg *ncm;
530         bool all = false;
531
532         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
533                         flags);
534         if (!nlh)
535                 return -EMSGSIZE;
536
537         if (type == NETCONFA_ALL)
538                 all = true;
539
540         ncm = nlmsg_data(nlh);
541         ncm->ncm_family = AF_INET6;
542
543         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
544                 goto nla_put_failure;
545
546         if (!devconf)
547                 goto out;
548
549         if ((all || type == NETCONFA_FORWARDING) &&
550             nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
551                 goto nla_put_failure;
552 #ifdef CONFIG_IPV6_MROUTE
553         if ((all || type == NETCONFA_MC_FORWARDING) &&
554             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
555                         devconf->mc_forwarding) < 0)
556                 goto nla_put_failure;
557 #endif
558         if ((all || type == NETCONFA_PROXY_NEIGH) &&
559             nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
560                 goto nla_put_failure;
561
562         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
563             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
564                         devconf->ignore_routes_with_linkdown) < 0)
565                 goto nla_put_failure;
566
567 out:
568         nlmsg_end(skb, nlh);
569         return 0;
570
571 nla_put_failure:
572         nlmsg_cancel(skb, nlh);
573         return -EMSGSIZE;
574 }
575
576 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
577                                   int ifindex, struct ipv6_devconf *devconf)
578 {
579         struct sk_buff *skb;
580         int err = -ENOBUFS;
581
582         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
583         if (!skb)
584                 goto errout;
585
586         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
587                                          event, 0, type);
588         if (err < 0) {
589                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
590                 WARN_ON(err == -EMSGSIZE);
591                 kfree_skb(skb);
592                 goto errout;
593         }
594         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
595         return;
596 errout:
597         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
598 }
599
600 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
601         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
602         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
603         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
604         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
605 };
606
607 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
608                                        const struct nlmsghdr *nlh,
609                                        struct nlattr **tb,
610                                        struct netlink_ext_ack *extack)
611 {
612         int i, err;
613
614         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
615                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
616                 return -EINVAL;
617         }
618
619         if (!netlink_strict_get_check(skb))
620                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
621                                               tb, NETCONFA_MAX,
622                                               devconf_ipv6_policy, extack);
623
624         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
625                                             tb, NETCONFA_MAX,
626                                             devconf_ipv6_policy, extack);
627         if (err)
628                 return err;
629
630         for (i = 0; i <= NETCONFA_MAX; i++) {
631                 if (!tb[i])
632                         continue;
633
634                 switch (i) {
635                 case NETCONFA_IFINDEX:
636                         break;
637                 default:
638                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
639                         return -EINVAL;
640                 }
641         }
642
643         return 0;
644 }
645
646 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
647                                      struct nlmsghdr *nlh,
648                                      struct netlink_ext_ack *extack)
649 {
650         struct net *net = sock_net(in_skb->sk);
651         struct nlattr *tb[NETCONFA_MAX+1];
652         struct inet6_dev *in6_dev = NULL;
653         struct net_device *dev = NULL;
654         struct sk_buff *skb;
655         struct ipv6_devconf *devconf;
656         int ifindex;
657         int err;
658
659         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
660         if (err < 0)
661                 return err;
662
663         if (!tb[NETCONFA_IFINDEX])
664                 return -EINVAL;
665
666         err = -EINVAL;
667         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
668         switch (ifindex) {
669         case NETCONFA_IFINDEX_ALL:
670                 devconf = net->ipv6.devconf_all;
671                 break;
672         case NETCONFA_IFINDEX_DEFAULT:
673                 devconf = net->ipv6.devconf_dflt;
674                 break;
675         default:
676                 dev = dev_get_by_index(net, ifindex);
677                 if (!dev)
678                         return -EINVAL;
679                 in6_dev = in6_dev_get(dev);
680                 if (!in6_dev)
681                         goto errout;
682                 devconf = &in6_dev->cnf;
683                 break;
684         }
685
686         err = -ENOBUFS;
687         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
688         if (!skb)
689                 goto errout;
690
691         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
692                                          NETLINK_CB(in_skb).portid,
693                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
694                                          NETCONFA_ALL);
695         if (err < 0) {
696                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
697                 WARN_ON(err == -EMSGSIZE);
698                 kfree_skb(skb);
699                 goto errout;
700         }
701         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
702 errout:
703         if (in6_dev)
704                 in6_dev_put(in6_dev);
705         dev_put(dev);
706         return err;
707 }
708
709 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
710                                       struct netlink_callback *cb)
711 {
712         const struct nlmsghdr *nlh = cb->nlh;
713         struct net *net = sock_net(skb->sk);
714         int h, s_h;
715         int idx, s_idx;
716         struct net_device *dev;
717         struct inet6_dev *idev;
718         struct hlist_head *head;
719
720         if (cb->strict_check) {
721                 struct netlink_ext_ack *extack = cb->extack;
722                 struct netconfmsg *ncm;
723
724                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
725                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
726                         return -EINVAL;
727                 }
728
729                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
730                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
731                         return -EINVAL;
732                 }
733         }
734
735         s_h = cb->args[0];
736         s_idx = idx = cb->args[1];
737
738         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
739                 idx = 0;
740                 head = &net->dev_index_head[h];
741                 rcu_read_lock();
742                 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
743                           net->dev_base_seq;
744                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
745                         if (idx < s_idx)
746                                 goto cont;
747                         idev = __in6_dev_get(dev);
748                         if (!idev)
749                                 goto cont;
750
751                         if (inet6_netconf_fill_devconf(skb, dev->ifindex,
752                                                        &idev->cnf,
753                                                        NETLINK_CB(cb->skb).portid,
754                                                        nlh->nlmsg_seq,
755                                                        RTM_NEWNETCONF,
756                                                        NLM_F_MULTI,
757                                                        NETCONFA_ALL) < 0) {
758                                 rcu_read_unlock();
759                                 goto done;
760                         }
761                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
762 cont:
763                         idx++;
764                 }
765                 rcu_read_unlock();
766         }
767         if (h == NETDEV_HASHENTRIES) {
768                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
769                                                net->ipv6.devconf_all,
770                                                NETLINK_CB(cb->skb).portid,
771                                                nlh->nlmsg_seq,
772                                                RTM_NEWNETCONF, NLM_F_MULTI,
773                                                NETCONFA_ALL) < 0)
774                         goto done;
775                 else
776                         h++;
777         }
778         if (h == NETDEV_HASHENTRIES + 1) {
779                 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
780                                                net->ipv6.devconf_dflt,
781                                                NETLINK_CB(cb->skb).portid,
782                                                nlh->nlmsg_seq,
783                                                RTM_NEWNETCONF, NLM_F_MULTI,
784                                                NETCONFA_ALL) < 0)
785                         goto done;
786                 else
787                         h++;
788         }
789 done:
790         cb->args[0] = h;
791         cb->args[1] = idx;
792
793         return skb->len;
794 }
795
796 #ifdef CONFIG_SYSCTL
797 static void dev_forward_change(struct inet6_dev *idev)
798 {
799         struct net_device *dev;
800         struct inet6_ifaddr *ifa;
801
802         if (!idev)
803                 return;
804         dev = idev->dev;
805         if (idev->cnf.forwarding)
806                 dev_disable_lro(dev);
807         if (dev->flags & IFF_MULTICAST) {
808                 if (idev->cnf.forwarding) {
809                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
810                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
811                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
812                 } else {
813                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
814                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
815                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
816                 }
817         }
818
819         list_for_each_entry(ifa, &idev->addr_list, if_list) {
820                 if (ifa->flags&IFA_F_TENTATIVE)
821                         continue;
822                 if (idev->cnf.forwarding)
823                         addrconf_join_anycast(ifa);
824                 else
825                         addrconf_leave_anycast(ifa);
826         }
827         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
828                                      NETCONFA_FORWARDING,
829                                      dev->ifindex, &idev->cnf);
830 }
831
832
833 static void addrconf_forward_change(struct net *net, __s32 newf)
834 {
835         struct net_device *dev;
836         struct inet6_dev *idev;
837
838         for_each_netdev(net, dev) {
839                 idev = __in6_dev_get(dev);
840                 if (idev) {
841                         int changed = (!idev->cnf.forwarding) ^ (!newf);
842                         idev->cnf.forwarding = newf;
843                         if (changed)
844                                 dev_forward_change(idev);
845                 }
846         }
847 }
848
849 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
850 {
851         struct net *net;
852         int old;
853
854         if (!rtnl_trylock())
855                 return restart_syscall();
856
857         net = (struct net *)table->extra2;
858         old = *p;
859         *p = newf;
860
861         if (p == &net->ipv6.devconf_dflt->forwarding) {
862                 if ((!newf) ^ (!old))
863                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
864                                                      NETCONFA_FORWARDING,
865                                                      NETCONFA_IFINDEX_DEFAULT,
866                                                      net->ipv6.devconf_dflt);
867                 rtnl_unlock();
868                 return 0;
869         }
870
871         if (p == &net->ipv6.devconf_all->forwarding) {
872                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
873
874                 net->ipv6.devconf_dflt->forwarding = newf;
875                 if ((!newf) ^ (!old_dflt))
876                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
877                                                      NETCONFA_FORWARDING,
878                                                      NETCONFA_IFINDEX_DEFAULT,
879                                                      net->ipv6.devconf_dflt);
880
881                 addrconf_forward_change(net, newf);
882                 if ((!newf) ^ (!old))
883                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
884                                                      NETCONFA_FORWARDING,
885                                                      NETCONFA_IFINDEX_ALL,
886                                                      net->ipv6.devconf_all);
887         } else if ((!newf) ^ (!old))
888                 dev_forward_change((struct inet6_dev *)table->extra1);
889         rtnl_unlock();
890
891         if (newf)
892                 rt6_purge_dflt_routers(net);
893         return 1;
894 }
895
896 static void addrconf_linkdown_change(struct net *net, __s32 newf)
897 {
898         struct net_device *dev;
899         struct inet6_dev *idev;
900
901         for_each_netdev(net, dev) {
902                 idev = __in6_dev_get(dev);
903                 if (idev) {
904                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
905
906                         idev->cnf.ignore_routes_with_linkdown = newf;
907                         if (changed)
908                                 inet6_netconf_notify_devconf(dev_net(dev),
909                                                              RTM_NEWNETCONF,
910                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
911                                                              dev->ifindex,
912                                                              &idev->cnf);
913                 }
914         }
915 }
916
917 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
918 {
919         struct net *net;
920         int old;
921
922         if (!rtnl_trylock())
923                 return restart_syscall();
924
925         net = (struct net *)table->extra2;
926         old = *p;
927         *p = newf;
928
929         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
930                 if ((!newf) ^ (!old))
931                         inet6_netconf_notify_devconf(net,
932                                                      RTM_NEWNETCONF,
933                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
934                                                      NETCONFA_IFINDEX_DEFAULT,
935                                                      net->ipv6.devconf_dflt);
936                 rtnl_unlock();
937                 return 0;
938         }
939
940         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
941                 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
942                 addrconf_linkdown_change(net, newf);
943                 if ((!newf) ^ (!old))
944                         inet6_netconf_notify_devconf(net,
945                                                      RTM_NEWNETCONF,
946                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
947                                                      NETCONFA_IFINDEX_ALL,
948                                                      net->ipv6.devconf_all);
949         }
950         rtnl_unlock();
951
952         return 1;
953 }
954
955 #endif
956
957 /* Nobody refers to this ifaddr, destroy it */
958 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
959 {
960         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
961
962 #ifdef NET_REFCNT_DEBUG
963         pr_debug("%s\n", __func__);
964 #endif
965
966         in6_dev_put(ifp->idev);
967
968         if (cancel_delayed_work(&ifp->dad_work))
969                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
970                           ifp);
971
972         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
973                 pr_warn("Freeing alive inet6 address %p\n", ifp);
974                 return;
975         }
976
977         kfree_rcu(ifp, rcu);
978 }
979
980 static void
981 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
982 {
983         struct list_head *p;
984         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
985
986         /*
987          * Each device address list is sorted in order of scope -
988          * global before linklocal.
989          */
990         list_for_each(p, &idev->addr_list) {
991                 struct inet6_ifaddr *ifa
992                         = list_entry(p, struct inet6_ifaddr, if_list);
993                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
994                         break;
995         }
996
997         list_add_tail_rcu(&ifp->if_list, p);
998 }
999
1000 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1001 {
1002         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1003
1004         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1005 }
1006
1007 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1008                                struct net_device *dev, unsigned int hash)
1009 {
1010         struct inet6_ifaddr *ifp;
1011
1012         hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1013                 if (!net_eq(dev_net(ifp->idev->dev), net))
1014                         continue;
1015                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1016                         if (!dev || ifp->idev->dev == dev)
1017                                 return true;
1018                 }
1019         }
1020         return false;
1021 }
1022
1023 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1024 {
1025         unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1026         int err = 0;
1027
1028         spin_lock(&addrconf_hash_lock);
1029
1030         /* Ignore adding duplicate addresses on an interface */
1031         if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1032                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1033                 err = -EEXIST;
1034         } else {
1035                 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1036         }
1037
1038         spin_unlock(&addrconf_hash_lock);
1039
1040         return err;
1041 }
1042
1043 /* On success it returns ifp with increased reference count */
1044
1045 static struct inet6_ifaddr *
1046 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1047               bool can_block, struct netlink_ext_ack *extack)
1048 {
1049         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1050         int addr_type = ipv6_addr_type(cfg->pfx);
1051         struct net *net = dev_net(idev->dev);
1052         struct inet6_ifaddr *ifa = NULL;
1053         struct fib6_info *f6i = NULL;
1054         int err = 0;
1055
1056         if (addr_type == IPV6_ADDR_ANY ||
1057             (addr_type & IPV6_ADDR_MULTICAST &&
1058              !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1059             (!(idev->dev->flags & IFF_LOOPBACK) &&
1060              !netif_is_l3_master(idev->dev) &&
1061              addr_type & IPV6_ADDR_LOOPBACK))
1062                 return ERR_PTR(-EADDRNOTAVAIL);
1063
1064         if (idev->dead) {
1065                 err = -ENODEV;                  /*XXX*/
1066                 goto out;
1067         }
1068
1069         if (idev->cnf.disable_ipv6) {
1070                 err = -EACCES;
1071                 goto out;
1072         }
1073
1074         /* validator notifier needs to be blocking;
1075          * do not call in atomic context
1076          */
1077         if (can_block) {
1078                 struct in6_validator_info i6vi = {
1079                         .i6vi_addr = *cfg->pfx,
1080                         .i6vi_dev = idev,
1081                         .extack = extack,
1082                 };
1083
1084                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1085                 err = notifier_to_errno(err);
1086                 if (err < 0)
1087                         goto out;
1088         }
1089
1090         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1091         if (!ifa) {
1092                 err = -ENOBUFS;
1093                 goto out;
1094         }
1095
1096         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1097         if (IS_ERR(f6i)) {
1098                 err = PTR_ERR(f6i);
1099                 f6i = NULL;
1100                 goto out;
1101         }
1102
1103         if (net->ipv6.devconf_all->disable_policy ||
1104             idev->cnf.disable_policy)
1105                 f6i->dst_nopolicy = true;
1106
1107         neigh_parms_data_state_setall(idev->nd_parms);
1108
1109         ifa->addr = *cfg->pfx;
1110         if (cfg->peer_pfx)
1111                 ifa->peer_addr = *cfg->peer_pfx;
1112
1113         spin_lock_init(&ifa->lock);
1114         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1115         INIT_HLIST_NODE(&ifa->addr_lst);
1116         ifa->scope = cfg->scope;
1117         ifa->prefix_len = cfg->plen;
1118         ifa->rt_priority = cfg->rt_priority;
1119         ifa->flags = cfg->ifa_flags;
1120         /* No need to add the TENTATIVE flag for addresses with NODAD */
1121         if (!(cfg->ifa_flags & IFA_F_NODAD))
1122                 ifa->flags |= IFA_F_TENTATIVE;
1123         ifa->valid_lft = cfg->valid_lft;
1124         ifa->prefered_lft = cfg->preferred_lft;
1125         ifa->cstamp = ifa->tstamp = jiffies;
1126         ifa->tokenized = false;
1127
1128         ifa->rt = f6i;
1129
1130         ifa->idev = idev;
1131         in6_dev_hold(idev);
1132
1133         /* For caller */
1134         refcount_set(&ifa->refcnt, 1);
1135
1136         rcu_read_lock_bh();
1137
1138         err = ipv6_add_addr_hash(idev->dev, ifa);
1139         if (err < 0) {
1140                 rcu_read_unlock_bh();
1141                 goto out;
1142         }
1143
1144         write_lock(&idev->lock);
1145
1146         /* Add to inet6_dev unicast addr list. */
1147         ipv6_link_dev_addr(idev, ifa);
1148
1149         if (ifa->flags&IFA_F_TEMPORARY) {
1150                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1151                 in6_ifa_hold(ifa);
1152         }
1153
1154         in6_ifa_hold(ifa);
1155         write_unlock(&idev->lock);
1156
1157         rcu_read_unlock_bh();
1158
1159         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1160 out:
1161         if (unlikely(err < 0)) {
1162                 fib6_info_release(f6i);
1163
1164                 if (ifa) {
1165                         if (ifa->idev)
1166                                 in6_dev_put(ifa->idev);
1167                         kfree(ifa);
1168                 }
1169                 ifa = ERR_PTR(err);
1170         }
1171
1172         return ifa;
1173 }
1174
1175 enum cleanup_prefix_rt_t {
1176         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1177         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1178         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1179 };
1180
1181 /*
1182  * Check, whether the prefix for ifp would still need a prefix route
1183  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1184  * constants.
1185  *
1186  * 1) we don't purge prefix if address was not permanent.
1187  *    prefix is managed by its own lifetime.
1188  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1189  * 3) if there are no addresses, delete prefix.
1190  * 4) if there are still other permanent address(es),
1191  *    corresponding prefix is still permanent.
1192  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1193  *    don't purge the prefix, assume user space is managing it.
1194  * 6) otherwise, update prefix lifetime to the
1195  *    longest valid lifetime among the corresponding
1196  *    addresses on the device.
1197  *    Note: subsequent RA will update lifetime.
1198  **/
1199 static enum cleanup_prefix_rt_t
1200 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1201 {
1202         struct inet6_ifaddr *ifa;
1203         struct inet6_dev *idev = ifp->idev;
1204         unsigned long lifetime;
1205         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1206
1207         *expires = jiffies;
1208
1209         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1210                 if (ifa == ifp)
1211                         continue;
1212                 if (ifa->prefix_len != ifp->prefix_len ||
1213                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1214                                        ifp->prefix_len))
1215                         continue;
1216                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1217                         return CLEANUP_PREFIX_RT_NOP;
1218
1219                 action = CLEANUP_PREFIX_RT_EXPIRE;
1220
1221                 spin_lock(&ifa->lock);
1222
1223                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1224                 /*
1225                  * Note: Because this address is
1226                  * not permanent, lifetime <
1227                  * LONG_MAX / HZ here.
1228                  */
1229                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1230                         *expires = ifa->tstamp + lifetime * HZ;
1231                 spin_unlock(&ifa->lock);
1232         }
1233
1234         return action;
1235 }
1236
1237 static void
1238 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1239                      bool del_rt, bool del_peer)
1240 {
1241         struct fib6_info *f6i;
1242
1243         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1244                                         ifp->prefix_len,
1245                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1246         if (f6i) {
1247                 if (del_rt)
1248                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1249                 else {
1250                         if (!(f6i->fib6_flags & RTF_EXPIRES))
1251                                 fib6_set_expires(f6i, expires);
1252                         fib6_info_release(f6i);
1253                 }
1254         }
1255 }
1256
1257
1258 /* This function wants to get referenced ifp and releases it before return */
1259
1260 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1261 {
1262         int state;
1263         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1264         unsigned long expires;
1265
1266         ASSERT_RTNL();
1267
1268         spin_lock_bh(&ifp->lock);
1269         state = ifp->state;
1270         ifp->state = INET6_IFADDR_STATE_DEAD;
1271         spin_unlock_bh(&ifp->lock);
1272
1273         if (state == INET6_IFADDR_STATE_DEAD)
1274                 goto out;
1275
1276         spin_lock_bh(&addrconf_hash_lock);
1277         hlist_del_init_rcu(&ifp->addr_lst);
1278         spin_unlock_bh(&addrconf_hash_lock);
1279
1280         write_lock_bh(&ifp->idev->lock);
1281
1282         if (ifp->flags&IFA_F_TEMPORARY) {
1283                 list_del(&ifp->tmp_list);
1284                 if (ifp->ifpub) {
1285                         in6_ifa_put(ifp->ifpub);
1286                         ifp->ifpub = NULL;
1287                 }
1288                 __in6_ifa_put(ifp);
1289         }
1290
1291         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1292                 action = check_cleanup_prefix_route(ifp, &expires);
1293
1294         list_del_rcu(&ifp->if_list);
1295         __in6_ifa_put(ifp);
1296
1297         write_unlock_bh(&ifp->idev->lock);
1298
1299         addrconf_del_dad_work(ifp);
1300
1301         ipv6_ifa_notify(RTM_DELADDR, ifp);
1302
1303         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1304
1305         if (action != CLEANUP_PREFIX_RT_NOP) {
1306                 cleanup_prefix_route(ifp, expires,
1307                         action == CLEANUP_PREFIX_RT_DEL, false);
1308         }
1309
1310         /* clean up prefsrc entries */
1311         rt6_remove_prefsrc(ifp);
1312 out:
1313         in6_ifa_put(ifp);
1314 }
1315
1316 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1317 {
1318         struct inet6_dev *idev = ifp->idev;
1319         unsigned long tmp_tstamp, age;
1320         unsigned long regen_advance;
1321         unsigned long now = jiffies;
1322         s32 cnf_temp_preferred_lft;
1323         struct inet6_ifaddr *ift;
1324         struct ifa6_config cfg;
1325         long max_desync_factor;
1326         struct in6_addr addr;
1327         int ret = 0;
1328
1329         write_lock_bh(&idev->lock);
1330
1331 retry:
1332         in6_dev_hold(idev);
1333         if (idev->cnf.use_tempaddr <= 0) {
1334                 write_unlock_bh(&idev->lock);
1335                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1336                 in6_dev_put(idev);
1337                 ret = -1;
1338                 goto out;
1339         }
1340         spin_lock_bh(&ifp->lock);
1341         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1342                 idev->cnf.use_tempaddr = -1;    /*XXX*/
1343                 spin_unlock_bh(&ifp->lock);
1344                 write_unlock_bh(&idev->lock);
1345                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1346                         __func__);
1347                 in6_dev_put(idev);
1348                 ret = -1;
1349                 goto out;
1350         }
1351         in6_ifa_hold(ifp);
1352         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1353         ipv6_gen_rnd_iid(&addr);
1354
1355         age = (now - ifp->tstamp) / HZ;
1356
1357         regen_advance = idev->cnf.regen_max_retry *
1358                         idev->cnf.dad_transmits *
1359                         max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1360
1361         /* recalculate max_desync_factor each time and update
1362          * idev->desync_factor if it's larger
1363          */
1364         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1365         max_desync_factor = min_t(__u32,
1366                                   idev->cnf.max_desync_factor,
1367                                   cnf_temp_preferred_lft - regen_advance);
1368
1369         if (unlikely(idev->desync_factor > max_desync_factor)) {
1370                 if (max_desync_factor > 0) {
1371                         get_random_bytes(&idev->desync_factor,
1372                                          sizeof(idev->desync_factor));
1373                         idev->desync_factor %= max_desync_factor;
1374                 } else {
1375                         idev->desync_factor = 0;
1376                 }
1377         }
1378
1379         memset(&cfg, 0, sizeof(cfg));
1380         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1381                               idev->cnf.temp_valid_lft + age);
1382         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1383         cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1384
1385         cfg.plen = ifp->prefix_len;
1386         tmp_tstamp = ifp->tstamp;
1387         spin_unlock_bh(&ifp->lock);
1388
1389         write_unlock_bh(&idev->lock);
1390
1391         /* A temporary address is created only if this calculated Preferred
1392          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
1393          * an implementation must not create a temporary address with a zero
1394          * Preferred Lifetime.
1395          * Use age calculation as in addrconf_verify to avoid unnecessary
1396          * temporary addresses being generated.
1397          */
1398         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1399         if (cfg.preferred_lft <= regen_advance + age) {
1400                 in6_ifa_put(ifp);
1401                 in6_dev_put(idev);
1402                 ret = -1;
1403                 goto out;
1404         }
1405
1406         cfg.ifa_flags = IFA_F_TEMPORARY;
1407         /* set in addrconf_prefix_rcv() */
1408         if (ifp->flags & IFA_F_OPTIMISTIC)
1409                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1410
1411         cfg.pfx = &addr;
1412         cfg.scope = ipv6_addr_scope(cfg.pfx);
1413
1414         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1415         if (IS_ERR(ift)) {
1416                 in6_ifa_put(ifp);
1417                 in6_dev_put(idev);
1418                 pr_info("%s: retry temporary address regeneration\n", __func__);
1419                 write_lock_bh(&idev->lock);
1420                 goto retry;
1421         }
1422
1423         spin_lock_bh(&ift->lock);
1424         ift->ifpub = ifp;
1425         ift->cstamp = now;
1426         ift->tstamp = tmp_tstamp;
1427         spin_unlock_bh(&ift->lock);
1428
1429         addrconf_dad_start(ift);
1430         in6_ifa_put(ift);
1431         in6_dev_put(idev);
1432 out:
1433         return ret;
1434 }
1435
1436 /*
1437  *      Choose an appropriate source address (RFC3484)
1438  */
1439 enum {
1440         IPV6_SADDR_RULE_INIT = 0,
1441         IPV6_SADDR_RULE_LOCAL,
1442         IPV6_SADDR_RULE_SCOPE,
1443         IPV6_SADDR_RULE_PREFERRED,
1444 #ifdef CONFIG_IPV6_MIP6
1445         IPV6_SADDR_RULE_HOA,
1446 #endif
1447         IPV6_SADDR_RULE_OIF,
1448         IPV6_SADDR_RULE_LABEL,
1449         IPV6_SADDR_RULE_PRIVACY,
1450         IPV6_SADDR_RULE_ORCHID,
1451         IPV6_SADDR_RULE_PREFIX,
1452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1453         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1454 #endif
1455         IPV6_SADDR_RULE_MAX
1456 };
1457
1458 struct ipv6_saddr_score {
1459         int                     rule;
1460         int                     addr_type;
1461         struct inet6_ifaddr     *ifa;
1462         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1463         int                     scopedist;
1464         int                     matchlen;
1465 };
1466
1467 struct ipv6_saddr_dst {
1468         const struct in6_addr *addr;
1469         int ifindex;
1470         int scope;
1471         int label;
1472         unsigned int prefs;
1473 };
1474
1475 static inline int ipv6_saddr_preferred(int type)
1476 {
1477         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1478                 return 1;
1479         return 0;
1480 }
1481
1482 static bool ipv6_use_optimistic_addr(struct net *net,
1483                                      struct inet6_dev *idev)
1484 {
1485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1486         if (!idev)
1487                 return false;
1488         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1489                 return false;
1490         if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1491                 return false;
1492
1493         return true;
1494 #else
1495         return false;
1496 #endif
1497 }
1498
1499 static bool ipv6_allow_optimistic_dad(struct net *net,
1500                                       struct inet6_dev *idev)
1501 {
1502 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1503         if (!idev)
1504                 return false;
1505         if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1506                 return false;
1507
1508         return true;
1509 #else
1510         return false;
1511 #endif
1512 }
1513
1514 static int ipv6_get_saddr_eval(struct net *net,
1515                                struct ipv6_saddr_score *score,
1516                                struct ipv6_saddr_dst *dst,
1517                                int i)
1518 {
1519         int ret;
1520
1521         if (i <= score->rule) {
1522                 switch (i) {
1523                 case IPV6_SADDR_RULE_SCOPE:
1524                         ret = score->scopedist;
1525                         break;
1526                 case IPV6_SADDR_RULE_PREFIX:
1527                         ret = score->matchlen;
1528                         break;
1529                 default:
1530                         ret = !!test_bit(i, score->scorebits);
1531                 }
1532                 goto out;
1533         }
1534
1535         switch (i) {
1536         case IPV6_SADDR_RULE_INIT:
1537                 /* Rule 0: remember if hiscore is not ready yet */
1538                 ret = !!score->ifa;
1539                 break;
1540         case IPV6_SADDR_RULE_LOCAL:
1541                 /* Rule 1: Prefer same address */
1542                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1543                 break;
1544         case IPV6_SADDR_RULE_SCOPE:
1545                 /* Rule 2: Prefer appropriate scope
1546                  *
1547                  *      ret
1548                  *       ^
1549                  *    -1 |  d 15
1550                  *    ---+--+-+---> scope
1551                  *       |
1552                  *       |             d is scope of the destination.
1553                  *  B-d  |  \
1554                  *       |   \      <- smaller scope is better if
1555                  *  B-15 |    \        if scope is enough for destination.
1556                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1557                  * d-C-1 | /
1558                  *       |/         <- greater is better
1559                  *   -C  /             if scope is not enough for destination.
1560                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1561                  *
1562                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1563                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1564                  * Assume B = 0 and we get C > 29.
1565                  */
1566                 ret = __ipv6_addr_src_scope(score->addr_type);
1567                 if (ret >= dst->scope)
1568                         ret = -ret;
1569                 else
1570                         ret -= 128;     /* 30 is enough */
1571                 score->scopedist = ret;
1572                 break;
1573         case IPV6_SADDR_RULE_PREFERRED:
1574             {
1575                 /* Rule 3: Avoid deprecated and optimistic addresses */
1576                 u8 avoid = IFA_F_DEPRECATED;
1577
1578                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1579                         avoid |= IFA_F_OPTIMISTIC;
1580                 ret = ipv6_saddr_preferred(score->addr_type) ||
1581                       !(score->ifa->flags & avoid);
1582                 break;
1583             }
1584 #ifdef CONFIG_IPV6_MIP6
1585         case IPV6_SADDR_RULE_HOA:
1586             {
1587                 /* Rule 4: Prefer home address */
1588                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1589                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1590                 break;
1591             }
1592 #endif
1593         case IPV6_SADDR_RULE_OIF:
1594                 /* Rule 5: Prefer outgoing interface */
1595                 ret = (!dst->ifindex ||
1596                        dst->ifindex == score->ifa->idev->dev->ifindex);
1597                 break;
1598         case IPV6_SADDR_RULE_LABEL:
1599                 /* Rule 6: Prefer matching label */
1600                 ret = ipv6_addr_label(net,
1601                                       &score->ifa->addr, score->addr_type,
1602                                       score->ifa->idev->dev->ifindex) == dst->label;
1603                 break;
1604         case IPV6_SADDR_RULE_PRIVACY:
1605             {
1606                 /* Rule 7: Prefer public address
1607                  * Note: prefer temporary address if use_tempaddr >= 2
1608                  */
1609                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1610                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1611                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1612                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1613                 break;
1614             }
1615         case IPV6_SADDR_RULE_ORCHID:
1616                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1617                  *          non-ORCHID vs non-ORCHID
1618                  */
1619                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1620                         ipv6_addr_orchid(dst->addr));
1621                 break;
1622         case IPV6_SADDR_RULE_PREFIX:
1623                 /* Rule 8: Use longest matching prefix */
1624                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1625                 if (ret > score->ifa->prefix_len)
1626                         ret = score->ifa->prefix_len;
1627                 score->matchlen = ret;
1628                 break;
1629 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1630         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1631                 /* Optimistic addresses still have lower precedence than other
1632                  * preferred addresses.
1633                  */
1634                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1635                 break;
1636 #endif
1637         default:
1638                 ret = 0;
1639         }
1640
1641         if (ret)
1642                 __set_bit(i, score->scorebits);
1643         score->rule = i;
1644 out:
1645         return ret;
1646 }
1647
1648 static int __ipv6_dev_get_saddr(struct net *net,
1649                                 struct ipv6_saddr_dst *dst,
1650                                 struct inet6_dev *idev,
1651                                 struct ipv6_saddr_score *scores,
1652                                 int hiscore_idx)
1653 {
1654         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1655
1656         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1657                 int i;
1658
1659                 /*
1660                  * - Tentative Address (RFC2462 section 5.4)
1661                  *  - A tentative address is not considered
1662                  *    "assigned to an interface" in the traditional
1663                  *    sense, unless it is also flagged as optimistic.
1664                  * - Candidate Source Address (section 4)
1665                  *  - In any case, anycast addresses, multicast
1666                  *    addresses, and the unspecified address MUST
1667                  *    NOT be included in a candidate set.
1668                  */
1669                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1670                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1671                         continue;
1672
1673                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1674
1675                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1676                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1677                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1678                                             idev->dev->name);
1679                         continue;
1680                 }
1681
1682                 score->rule = -1;
1683                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1684
1685                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1686                         int minihiscore, miniscore;
1687
1688                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1689                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1690
1691                         if (minihiscore > miniscore) {
1692                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1693                                     score->scopedist > 0) {
1694                                         /*
1695                                          * special case:
1696                                          * each remaining entry
1697                                          * has too small (not enough)
1698                                          * scope, because ifa entries
1699                                          * are sorted by their scope
1700                                          * values.
1701                                          */
1702                                         goto out;
1703                                 }
1704                                 break;
1705                         } else if (minihiscore < miniscore) {
1706                                 swap(hiscore, score);
1707                                 hiscore_idx = 1 - hiscore_idx;
1708
1709                                 /* restore our iterator */
1710                                 score->ifa = hiscore->ifa;
1711
1712                                 break;
1713                         }
1714                 }
1715         }
1716 out:
1717         return hiscore_idx;
1718 }
1719
1720 static int ipv6_get_saddr_master(struct net *net,
1721                                  const struct net_device *dst_dev,
1722                                  const struct net_device *master,
1723                                  struct ipv6_saddr_dst *dst,
1724                                  struct ipv6_saddr_score *scores,
1725                                  int hiscore_idx)
1726 {
1727         struct inet6_dev *idev;
1728
1729         idev = __in6_dev_get(dst_dev);
1730         if (idev)
1731                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1732                                                    scores, hiscore_idx);
1733
1734         idev = __in6_dev_get(master);
1735         if (idev)
1736                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1737                                                    scores, hiscore_idx);
1738
1739         return hiscore_idx;
1740 }
1741
1742 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1743                        const struct in6_addr *daddr, unsigned int prefs,
1744                        struct in6_addr *saddr)
1745 {
1746         struct ipv6_saddr_score scores[2], *hiscore;
1747         struct ipv6_saddr_dst dst;
1748         struct inet6_dev *idev;
1749         struct net_device *dev;
1750         int dst_type;
1751         bool use_oif_addr = false;
1752         int hiscore_idx = 0;
1753         int ret = 0;
1754
1755         dst_type = __ipv6_addr_type(daddr);
1756         dst.addr = daddr;
1757         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1758         dst.scope = __ipv6_addr_src_scope(dst_type);
1759         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1760         dst.prefs = prefs;
1761
1762         scores[hiscore_idx].rule = -1;
1763         scores[hiscore_idx].ifa = NULL;
1764
1765         rcu_read_lock();
1766
1767         /* Candidate Source Address (section 4)
1768          *  - multicast and link-local destination address,
1769          *    the set of candidate source address MUST only
1770          *    include addresses assigned to interfaces
1771          *    belonging to the same link as the outgoing
1772          *    interface.
1773          * (- For site-local destination addresses, the
1774          *    set of candidate source addresses MUST only
1775          *    include addresses assigned to interfaces
1776          *    belonging to the same site as the outgoing
1777          *    interface.)
1778          *  - "It is RECOMMENDED that the candidate source addresses
1779          *    be the set of unicast addresses assigned to the
1780          *    interface that will be used to send to the destination
1781          *    (the 'outgoing' interface)." (RFC 6724)
1782          */
1783         if (dst_dev) {
1784                 idev = __in6_dev_get(dst_dev);
1785                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1786                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1787                     (idev && idev->cnf.use_oif_addrs_only)) {
1788                         use_oif_addr = true;
1789                 }
1790         }
1791
1792         if (use_oif_addr) {
1793                 if (idev)
1794                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1795         } else {
1796                 const struct net_device *master;
1797                 int master_idx = 0;
1798
1799                 /* if dst_dev exists and is enslaved to an L3 device, then
1800                  * prefer addresses from dst_dev and then the master over
1801                  * any other enslaved devices in the L3 domain.
1802                  */
1803                 master = l3mdev_master_dev_rcu(dst_dev);
1804                 if (master) {
1805                         master_idx = master->ifindex;
1806
1807                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1808                                                             master, &dst,
1809                                                             scores, hiscore_idx);
1810
1811                         if (scores[hiscore_idx].ifa)
1812                                 goto out;
1813                 }
1814
1815                 for_each_netdev_rcu(net, dev) {
1816                         /* only consider addresses on devices in the
1817                          * same L3 domain
1818                          */
1819                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1820                                 continue;
1821                         idev = __in6_dev_get(dev);
1822                         if (!idev)
1823                                 continue;
1824                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1825                 }
1826         }
1827
1828 out:
1829         hiscore = &scores[hiscore_idx];
1830         if (!hiscore->ifa)
1831                 ret = -EADDRNOTAVAIL;
1832         else
1833                 *saddr = hiscore->ifa->addr;
1834
1835         rcu_read_unlock();
1836         return ret;
1837 }
1838 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1839
1840 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1841                       u32 banned_flags)
1842 {
1843         struct inet6_ifaddr *ifp;
1844         int err = -EADDRNOTAVAIL;
1845
1846         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1847                 if (ifp->scope > IFA_LINK)
1848                         break;
1849                 if (ifp->scope == IFA_LINK &&
1850                     !(ifp->flags & banned_flags)) {
1851                         *addr = ifp->addr;
1852                         err = 0;
1853                         break;
1854                 }
1855         }
1856         return err;
1857 }
1858
1859 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1860                     u32 banned_flags)
1861 {
1862         struct inet6_dev *idev;
1863         int err = -EADDRNOTAVAIL;
1864
1865         rcu_read_lock();
1866         idev = __in6_dev_get(dev);
1867         if (idev) {
1868                 read_lock_bh(&idev->lock);
1869                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1870                 read_unlock_bh(&idev->lock);
1871         }
1872         rcu_read_unlock();
1873         return err;
1874 }
1875
1876 static int ipv6_count_addresses(const struct inet6_dev *idev)
1877 {
1878         const struct inet6_ifaddr *ifp;
1879         int cnt = 0;
1880
1881         rcu_read_lock();
1882         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1883                 cnt++;
1884         rcu_read_unlock();
1885         return cnt;
1886 }
1887
1888 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1889                   const struct net_device *dev, int strict)
1890 {
1891         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1892                                        strict, IFA_F_TENTATIVE);
1893 }
1894 EXPORT_SYMBOL(ipv6_chk_addr);
1895
1896 /* device argument is used to find the L3 domain of interest. If
1897  * skip_dev_check is set, then the ifp device is not checked against
1898  * the passed in dev argument. So the 2 cases for addresses checks are:
1899  *   1. does the address exist in the L3 domain that dev is part of
1900  *      (skip_dev_check = true), or
1901  *
1902  *   2. does the address exist on the specific device
1903  *      (skip_dev_check = false)
1904  */
1905 static struct net_device *
1906 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1907                           const struct net_device *dev, bool skip_dev_check,
1908                           int strict, u32 banned_flags)
1909 {
1910         unsigned int hash = inet6_addr_hash(net, addr);
1911         struct net_device *l3mdev, *ndev;
1912         struct inet6_ifaddr *ifp;
1913         u32 ifp_flags;
1914
1915         rcu_read_lock();
1916
1917         l3mdev = l3mdev_master_dev_rcu(dev);
1918         if (skip_dev_check)
1919                 dev = NULL;
1920
1921         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1922                 ndev = ifp->idev->dev;
1923                 if (!net_eq(dev_net(ndev), net))
1924                         continue;
1925
1926                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1927                         continue;
1928
1929                 /* Decouple optimistic from tentative for evaluation here.
1930                  * Ban optimistic addresses explicitly, when required.
1931                  */
1932                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1933                             ? (ifp->flags&~IFA_F_TENTATIVE)
1934                             : ifp->flags;
1935                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1936                     !(ifp_flags&banned_flags) &&
1937                     (!dev || ndev == dev ||
1938                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1939                         rcu_read_unlock();
1940                         return ndev;
1941                 }
1942         }
1943
1944         rcu_read_unlock();
1945         return NULL;
1946 }
1947
1948 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1949                             const struct net_device *dev, bool skip_dev_check,
1950                             int strict, u32 banned_flags)
1951 {
1952         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
1953                                          strict, banned_flags) ? 1 : 0;
1954 }
1955 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1956
1957
1958 /* Compares an address/prefix_len with addresses on device @dev.
1959  * If one is found it returns true.
1960  */
1961 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1962         const unsigned int prefix_len, struct net_device *dev)
1963 {
1964         const struct inet6_ifaddr *ifa;
1965         const struct inet6_dev *idev;
1966         bool ret = false;
1967
1968         rcu_read_lock();
1969         idev = __in6_dev_get(dev);
1970         if (idev) {
1971                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1972                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1973                         if (ret)
1974                                 break;
1975                 }
1976         }
1977         rcu_read_unlock();
1978
1979         return ret;
1980 }
1981 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1982
1983 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1984 {
1985         const struct inet6_ifaddr *ifa;
1986         const struct inet6_dev *idev;
1987         int     onlink;
1988
1989         onlink = 0;
1990         rcu_read_lock();
1991         idev = __in6_dev_get(dev);
1992         if (idev) {
1993                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1994                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1995                                                    ifa->prefix_len);
1996                         if (onlink)
1997                                 break;
1998                 }
1999         }
2000         rcu_read_unlock();
2001         return onlink;
2002 }
2003 EXPORT_SYMBOL(ipv6_chk_prefix);
2004
2005 /**
2006  * ipv6_dev_find - find the first device with a given source address.
2007  * @net: the net namespace
2008  * @addr: the source address
2009  * @dev: used to find the L3 domain of interest
2010  *
2011  * The caller should be protected by RCU, or RTNL.
2012  */
2013 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2014                                  struct net_device *dev)
2015 {
2016         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2017                                          IFA_F_TENTATIVE);
2018 }
2019 EXPORT_SYMBOL(ipv6_dev_find);
2020
2021 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2022                                      struct net_device *dev, int strict)
2023 {
2024         unsigned int hash = inet6_addr_hash(net, addr);
2025         struct inet6_ifaddr *ifp, *result = NULL;
2026
2027         rcu_read_lock();
2028         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2029                 if (!net_eq(dev_net(ifp->idev->dev), net))
2030                         continue;
2031                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2032                         if (!dev || ifp->idev->dev == dev ||
2033                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2034                                 result = ifp;
2035                                 in6_ifa_hold(ifp);
2036                                 break;
2037                         }
2038                 }
2039         }
2040         rcu_read_unlock();
2041
2042         return result;
2043 }
2044
2045 /* Gets referenced address, destroys ifaddr */
2046
2047 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2048 {
2049         if (dad_failed)
2050                 ifp->flags |= IFA_F_DADFAILED;
2051
2052         if (ifp->flags&IFA_F_TEMPORARY) {
2053                 struct inet6_ifaddr *ifpub;
2054                 spin_lock_bh(&ifp->lock);
2055                 ifpub = ifp->ifpub;
2056                 if (ifpub) {
2057                         in6_ifa_hold(ifpub);
2058                         spin_unlock_bh(&ifp->lock);
2059                         ipv6_create_tempaddr(ifpub, true);
2060                         in6_ifa_put(ifpub);
2061                 } else {
2062                         spin_unlock_bh(&ifp->lock);
2063                 }
2064                 ipv6_del_addr(ifp);
2065         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2066                 spin_lock_bh(&ifp->lock);
2067                 addrconf_del_dad_work(ifp);
2068                 ifp->flags |= IFA_F_TENTATIVE;
2069                 if (dad_failed)
2070                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2071                 spin_unlock_bh(&ifp->lock);
2072                 if (dad_failed)
2073                         ipv6_ifa_notify(0, ifp);
2074                 in6_ifa_put(ifp);
2075         } else {
2076                 ipv6_del_addr(ifp);
2077         }
2078 }
2079
2080 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2081 {
2082         int err = -ENOENT;
2083
2084         spin_lock_bh(&ifp->lock);
2085         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2086                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2087                 err = 0;
2088         }
2089         spin_unlock_bh(&ifp->lock);
2090
2091         return err;
2092 }
2093
2094 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2095 {
2096         struct inet6_dev *idev = ifp->idev;
2097         struct net *net = dev_net(ifp->idev->dev);
2098
2099         if (addrconf_dad_end(ifp)) {
2100                 in6_ifa_put(ifp);
2101                 return;
2102         }
2103
2104         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2105                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2106
2107         spin_lock_bh(&ifp->lock);
2108
2109         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2110                 struct in6_addr new_addr;
2111                 struct inet6_ifaddr *ifp2;
2112                 int retries = ifp->stable_privacy_retry + 1;
2113                 struct ifa6_config cfg = {
2114                         .pfx = &new_addr,
2115                         .plen = ifp->prefix_len,
2116                         .ifa_flags = ifp->flags,
2117                         .valid_lft = ifp->valid_lft,
2118                         .preferred_lft = ifp->prefered_lft,
2119                         .scope = ifp->scope,
2120                 };
2121
2122                 if (retries > net->ipv6.sysctl.idgen_retries) {
2123                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2124                                              ifp->idev->dev->name);
2125                         goto errdad;
2126                 }
2127
2128                 new_addr = ifp->addr;
2129                 if (ipv6_generate_stable_address(&new_addr, retries,
2130                                                  idev))
2131                         goto errdad;
2132
2133                 spin_unlock_bh(&ifp->lock);
2134
2135                 if (idev->cnf.max_addresses &&
2136                     ipv6_count_addresses(idev) >=
2137                     idev->cnf.max_addresses)
2138                         goto lock_errdad;
2139
2140                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2141                                      ifp->idev->dev->name);
2142
2143                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2144                 if (IS_ERR(ifp2))
2145                         goto lock_errdad;
2146
2147                 spin_lock_bh(&ifp2->lock);
2148                 ifp2->stable_privacy_retry = retries;
2149                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2150                 spin_unlock_bh(&ifp2->lock);
2151
2152                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2153                 in6_ifa_put(ifp2);
2154 lock_errdad:
2155                 spin_lock_bh(&ifp->lock);
2156         }
2157
2158 errdad:
2159         /* transition from _POSTDAD to _ERRDAD */
2160         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2161         spin_unlock_bh(&ifp->lock);
2162
2163         addrconf_mod_dad_work(ifp, 0);
2164         in6_ifa_put(ifp);
2165 }
2166
2167 /* Join to solicited addr multicast group.
2168  * caller must hold RTNL */
2169 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2170 {
2171         struct in6_addr maddr;
2172
2173         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2174                 return;
2175
2176         addrconf_addr_solict_mult(addr, &maddr);
2177         ipv6_dev_mc_inc(dev, &maddr);
2178 }
2179
2180 /* caller must hold RTNL */
2181 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2182 {
2183         struct in6_addr maddr;
2184
2185         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2186                 return;
2187
2188         addrconf_addr_solict_mult(addr, &maddr);
2189         __ipv6_dev_mc_dec(idev, &maddr);
2190 }
2191
2192 /* caller must hold RTNL */
2193 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2194 {
2195         struct in6_addr addr;
2196
2197         if (ifp->prefix_len >= 127) /* RFC 6164 */
2198                 return;
2199         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2200         if (ipv6_addr_any(&addr))
2201                 return;
2202         __ipv6_dev_ac_inc(ifp->idev, &addr);
2203 }
2204
2205 /* caller must hold RTNL */
2206 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2207 {
2208         struct in6_addr addr;
2209
2210         if (ifp->prefix_len >= 127) /* RFC 6164 */
2211                 return;
2212         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2213         if (ipv6_addr_any(&addr))
2214                 return;
2215         __ipv6_dev_ac_dec(ifp->idev, &addr);
2216 }
2217
2218 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2219 {
2220         switch (dev->addr_len) {
2221         case ETH_ALEN:
2222                 memcpy(eui, dev->dev_addr, 3);
2223                 eui[3] = 0xFF;
2224                 eui[4] = 0xFE;
2225                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2226                 break;
2227         case EUI64_ADDR_LEN:
2228                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2229                 eui[0] ^= 2;
2230                 break;
2231         default:
2232                 return -1;
2233         }
2234
2235         return 0;
2236 }
2237
2238 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2239 {
2240         union fwnet_hwaddr *ha;
2241
2242         if (dev->addr_len != FWNET_ALEN)
2243                 return -1;
2244
2245         ha = (union fwnet_hwaddr *)dev->dev_addr;
2246
2247         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2248         eui[0] ^= 2;
2249         return 0;
2250 }
2251
2252 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2253 {
2254         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2255         if (dev->addr_len != ARCNET_ALEN)
2256                 return -1;
2257         memset(eui, 0, 7);
2258         eui[7] = *(u8 *)dev->dev_addr;
2259         return 0;
2260 }
2261
2262 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2263 {
2264         if (dev->addr_len != INFINIBAND_ALEN)
2265                 return -1;
2266         memcpy(eui, dev->dev_addr + 12, 8);
2267         eui[0] |= 2;
2268         return 0;
2269 }
2270
2271 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2272 {
2273         if (addr == 0)
2274                 return -1;
2275         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2276                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2277                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2278                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2279                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2280                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2281         eui[1] = 0;
2282         eui[2] = 0x5E;
2283         eui[3] = 0xFE;
2284         memcpy(eui + 4, &addr, 4);
2285         return 0;
2286 }
2287
2288 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2289 {
2290         if (dev->priv_flags & IFF_ISATAP)
2291                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2292         return -1;
2293 }
2294
2295 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2296 {
2297         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2298 }
2299
2300 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2301 {
2302         memcpy(eui, dev->perm_addr, 3);
2303         memcpy(eui + 5, dev->perm_addr + 3, 3);
2304         eui[3] = 0xFF;
2305         eui[4] = 0xFE;
2306         eui[0] ^= 2;
2307         return 0;
2308 }
2309
2310 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2311 {
2312         switch (dev->type) {
2313         case ARPHRD_ETHER:
2314         case ARPHRD_FDDI:
2315                 return addrconf_ifid_eui48(eui, dev);
2316         case ARPHRD_ARCNET:
2317                 return addrconf_ifid_arcnet(eui, dev);
2318         case ARPHRD_INFINIBAND:
2319                 return addrconf_ifid_infiniband(eui, dev);
2320         case ARPHRD_SIT:
2321                 return addrconf_ifid_sit(eui, dev);
2322         case ARPHRD_IPGRE:
2323         case ARPHRD_TUNNEL:
2324                 return addrconf_ifid_gre(eui, dev);
2325         case ARPHRD_6LOWPAN:
2326                 return addrconf_ifid_6lowpan(eui, dev);
2327         case ARPHRD_IEEE1394:
2328                 return addrconf_ifid_ieee1394(eui, dev);
2329         case ARPHRD_TUNNEL6:
2330         case ARPHRD_IP6GRE:
2331         case ARPHRD_RAWIP:
2332                 return addrconf_ifid_ip6tnl(eui, dev);
2333         }
2334         return -1;
2335 }
2336
2337 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2338 {
2339         int err = -1;
2340         struct inet6_ifaddr *ifp;
2341
2342         read_lock_bh(&idev->lock);
2343         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2344                 if (ifp->scope > IFA_LINK)
2345                         break;
2346                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2347                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2348                         err = 0;
2349                         break;
2350                 }
2351         }
2352         read_unlock_bh(&idev->lock);
2353         return err;
2354 }
2355
2356 /* Generation of a randomized Interface Identifier
2357  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2358  */
2359
2360 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2361 {
2362 regen:
2363         get_random_bytes(&addr->s6_addr[8], 8);
2364
2365         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2366          * check if generated address is not inappropriate:
2367          *
2368          * - Reserved IPv6 Interface Identifiers
2369          * - XXX: already assigned to an address on the device
2370          */
2371
2372         /* Subnet-router anycast: 0000:0000:0000:0000 */
2373         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2374                 goto regen;
2375
2376         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2377          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2378          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2379          */
2380         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2381             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2382                 goto regen;
2383
2384         /* Reserved subnet anycast addresses */
2385         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2386             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2387                 goto regen;
2388 }
2389
2390 /*
2391  *      Add prefix route.
2392  */
2393
2394 static void
2395 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2396                       struct net_device *dev, unsigned long expires,
2397                       u32 flags, gfp_t gfp_flags)
2398 {
2399         struct fib6_config cfg = {
2400                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2401                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2402                 .fc_ifindex = dev->ifindex,
2403                 .fc_expires = expires,
2404                 .fc_dst_len = plen,
2405                 .fc_flags = RTF_UP | flags,
2406                 .fc_nlinfo.nl_net = dev_net(dev),
2407                 .fc_protocol = RTPROT_KERNEL,
2408                 .fc_type = RTN_UNICAST,
2409         };
2410
2411         cfg.fc_dst = *pfx;
2412
2413         /* Prevent useless cloning on PtP SIT.
2414            This thing is done here expecting that the whole
2415            class of non-broadcast devices need not cloning.
2416          */
2417 #if IS_ENABLED(CONFIG_IPV6_SIT)
2418         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2419                 cfg.fc_flags |= RTF_NONEXTHOP;
2420 #endif
2421
2422         ip6_route_add(&cfg, gfp_flags, NULL);
2423 }
2424
2425
2426 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2427                                                   int plen,
2428                                                   const struct net_device *dev,
2429                                                   u32 flags, u32 noflags,
2430                                                   bool no_gw)
2431 {
2432         struct fib6_node *fn;
2433         struct fib6_info *rt = NULL;
2434         struct fib6_table *table;
2435         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2436
2437         table = fib6_get_table(dev_net(dev), tb_id);
2438         if (!table)
2439                 return NULL;
2440
2441         rcu_read_lock();
2442         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2443         if (!fn)
2444                 goto out;
2445
2446         for_each_fib6_node_rt_rcu(fn) {
2447                 /* prefix routes only use builtin fib6_nh */
2448                 if (rt->nh)
2449                         continue;
2450
2451                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2452                         continue;
2453                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2454                         continue;
2455                 if ((rt->fib6_flags & flags) != flags)
2456                         continue;
2457                 if ((rt->fib6_flags & noflags) != 0)
2458                         continue;
2459                 if (!fib6_info_hold_safe(rt))
2460                         continue;
2461                 break;
2462         }
2463 out:
2464         rcu_read_unlock();
2465         return rt;
2466 }
2467
2468
2469 /* Create "default" multicast route to the interface */
2470
2471 static void addrconf_add_mroute(struct net_device *dev)
2472 {
2473         struct fib6_config cfg = {
2474                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2475                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2476                 .fc_ifindex = dev->ifindex,
2477                 .fc_dst_len = 8,
2478                 .fc_flags = RTF_UP,
2479                 .fc_type = RTN_MULTICAST,
2480                 .fc_nlinfo.nl_net = dev_net(dev),
2481                 .fc_protocol = RTPROT_KERNEL,
2482         };
2483
2484         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2485
2486         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2487 }
2488
2489 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2490 {
2491         struct inet6_dev *idev;
2492
2493         ASSERT_RTNL();
2494
2495         idev = ipv6_find_idev(dev);
2496         if (IS_ERR(idev))
2497                 return idev;
2498
2499         if (idev->cnf.disable_ipv6)
2500                 return ERR_PTR(-EACCES);
2501
2502         /* Add default multicast route */
2503         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2504                 addrconf_add_mroute(dev);
2505
2506         return idev;
2507 }
2508
2509 static void manage_tempaddrs(struct inet6_dev *idev,
2510                              struct inet6_ifaddr *ifp,
2511                              __u32 valid_lft, __u32 prefered_lft,
2512                              bool create, unsigned long now)
2513 {
2514         u32 flags;
2515         struct inet6_ifaddr *ift;
2516
2517         read_lock_bh(&idev->lock);
2518         /* update all temporary addresses in the list */
2519         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2520                 int age, max_valid, max_prefered;
2521
2522                 if (ifp != ift->ifpub)
2523                         continue;
2524
2525                 /* RFC 4941 section 3.3:
2526                  * If a received option will extend the lifetime of a public
2527                  * address, the lifetimes of temporary addresses should
2528                  * be extended, subject to the overall constraint that no
2529                  * temporary addresses should ever remain "valid" or "preferred"
2530                  * for a time longer than (TEMP_VALID_LIFETIME) or
2531                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2532                  */
2533                 age = (now - ift->cstamp) / HZ;
2534                 max_valid = idev->cnf.temp_valid_lft - age;
2535                 if (max_valid < 0)
2536                         max_valid = 0;
2537
2538                 max_prefered = idev->cnf.temp_prefered_lft -
2539                                idev->desync_factor - age;
2540                 if (max_prefered < 0)
2541                         max_prefered = 0;
2542
2543                 if (valid_lft > max_valid)
2544                         valid_lft = max_valid;
2545
2546                 if (prefered_lft > max_prefered)
2547                         prefered_lft = max_prefered;
2548
2549                 spin_lock(&ift->lock);
2550                 flags = ift->flags;
2551                 ift->valid_lft = valid_lft;
2552                 ift->prefered_lft = prefered_lft;
2553                 ift->tstamp = now;
2554                 if (prefered_lft > 0)
2555                         ift->flags &= ~IFA_F_DEPRECATED;
2556
2557                 spin_unlock(&ift->lock);
2558                 if (!(flags&IFA_F_TENTATIVE))
2559                         ipv6_ifa_notify(0, ift);
2560         }
2561
2562         if ((create || list_empty(&idev->tempaddr_list)) &&
2563             idev->cnf.use_tempaddr > 0) {
2564                 /* When a new public address is created as described
2565                  * in [ADDRCONF], also create a new temporary address.
2566                  * Also create a temporary address if it's enabled but
2567                  * no temporary address currently exists.
2568                  */
2569                 read_unlock_bh(&idev->lock);
2570                 ipv6_create_tempaddr(ifp, false);
2571         } else {
2572                 read_unlock_bh(&idev->lock);
2573         }
2574 }
2575
2576 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2577 {
2578         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2579                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2580 }
2581
2582 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2583                                  const struct prefix_info *pinfo,
2584                                  struct inet6_dev *in6_dev,
2585                                  const struct in6_addr *addr, int addr_type,
2586                                  u32 addr_flags, bool sllao, bool tokenized,
2587                                  __u32 valid_lft, u32 prefered_lft)
2588 {
2589         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2590         int create = 0;
2591
2592         if (!ifp && valid_lft) {
2593                 int max_addresses = in6_dev->cnf.max_addresses;
2594                 struct ifa6_config cfg = {
2595                         .pfx = addr,
2596                         .plen = pinfo->prefix_len,
2597                         .ifa_flags = addr_flags,
2598                         .valid_lft = valid_lft,
2599                         .preferred_lft = prefered_lft,
2600                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2601                 };
2602
2603 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2604                 if ((net->ipv6.devconf_all->optimistic_dad ||
2605                      in6_dev->cnf.optimistic_dad) &&
2606                     !net->ipv6.devconf_all->forwarding && sllao)
2607                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2608 #endif
2609
2610                 /* Do not allow to create too much of autoconfigured
2611                  * addresses; this would be too easy way to crash kernel.
2612                  */
2613                 if (!max_addresses ||
2614                     ipv6_count_addresses(in6_dev) < max_addresses)
2615                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2616
2617                 if (IS_ERR_OR_NULL(ifp))
2618                         return -1;
2619
2620                 create = 1;
2621                 spin_lock_bh(&ifp->lock);
2622                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2623                 ifp->cstamp = jiffies;
2624                 ifp->tokenized = tokenized;
2625                 spin_unlock_bh(&ifp->lock);
2626                 addrconf_dad_start(ifp);
2627         }
2628
2629         if (ifp) {
2630                 u32 flags;
2631                 unsigned long now;
2632                 u32 stored_lft;
2633
2634                 /* Update lifetime (RFC4862 5.5.3 e)
2635                  * We deviate from RFC4862 by honoring all Valid Lifetimes to
2636                  * improve the reaction of SLAAC to renumbering events
2637                  * (draft-gont-6man-slaac-renum-06, Section 4.2)
2638                  */
2639                 spin_lock_bh(&ifp->lock);
2640                 now = jiffies;
2641                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2642                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2643                 else
2644                         stored_lft = 0;
2645
2646                 if (!create && stored_lft) {
2647                         ifp->valid_lft = valid_lft;
2648                         ifp->prefered_lft = prefered_lft;
2649                         ifp->tstamp = now;
2650                         flags = ifp->flags;
2651                         ifp->flags &= ~IFA_F_DEPRECATED;
2652                         spin_unlock_bh(&ifp->lock);
2653
2654                         if (!(flags&IFA_F_TENTATIVE))
2655                                 ipv6_ifa_notify(0, ifp);
2656                 } else
2657                         spin_unlock_bh(&ifp->lock);
2658
2659                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2660                                  create, now);
2661
2662                 in6_ifa_put(ifp);
2663                 addrconf_verify();
2664         }
2665
2666         return 0;
2667 }
2668 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2669
2670 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2671 {
2672         struct prefix_info *pinfo;
2673         __u32 valid_lft;
2674         __u32 prefered_lft;
2675         int addr_type, err;
2676         u32 addr_flags = 0;
2677         struct inet6_dev *in6_dev;
2678         struct net *net = dev_net(dev);
2679
2680         pinfo = (struct prefix_info *) opt;
2681
2682         if (len < sizeof(struct prefix_info)) {
2683                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2684                 return;
2685         }
2686
2687         /*
2688          *      Validation checks ([ADDRCONF], page 19)
2689          */
2690
2691         addr_type = ipv6_addr_type(&pinfo->prefix);
2692
2693         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2694                 return;
2695
2696         valid_lft = ntohl(pinfo->valid);
2697         prefered_lft = ntohl(pinfo->prefered);
2698
2699         if (prefered_lft > valid_lft) {
2700                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2701                 return;
2702         }
2703
2704         in6_dev = in6_dev_get(dev);
2705
2706         if (!in6_dev) {
2707                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2708                                     dev->name);
2709                 return;
2710         }
2711
2712         /*
2713          *      Two things going on here:
2714          *      1) Add routes for on-link prefixes
2715          *      2) Configure prefixes with the auto flag set
2716          */
2717
2718         if (pinfo->onlink) {
2719                 struct fib6_info *rt;
2720                 unsigned long rt_expires;
2721
2722                 /* Avoid arithmetic overflow. Really, we could
2723                  * save rt_expires in seconds, likely valid_lft,
2724                  * but it would require division in fib gc, that it
2725                  * not good.
2726                  */
2727                 if (HZ > USER_HZ)
2728                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2729                 else
2730                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2731
2732                 if (addrconf_finite_timeout(rt_expires))
2733                         rt_expires *= HZ;
2734
2735                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2736                                                pinfo->prefix_len,
2737                                                dev,
2738                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2739                                                RTF_DEFAULT, true);
2740
2741                 if (rt) {
2742                         /* Autoconf prefix route */
2743                         if (valid_lft == 0) {
2744                                 ip6_del_rt(net, rt, false);
2745                                 rt = NULL;
2746                         } else if (addrconf_finite_timeout(rt_expires)) {
2747                                 /* not infinity */
2748                                 fib6_set_expires(rt, jiffies + rt_expires);
2749                         } else {
2750                                 fib6_clean_expires(rt);
2751                         }
2752                 } else if (valid_lft) {
2753                         clock_t expires = 0;
2754                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2755                         if (addrconf_finite_timeout(rt_expires)) {
2756                                 /* not infinity */
2757                                 flags |= RTF_EXPIRES;
2758                                 expires = jiffies_to_clock_t(rt_expires);
2759                         }
2760                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2761                                               0, dev, expires, flags,
2762                                               GFP_ATOMIC);
2763                 }
2764                 fib6_info_release(rt);
2765         }
2766
2767         /* Try to figure out our local address for this prefix */
2768
2769         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2770                 struct in6_addr addr;
2771                 bool tokenized = false, dev_addr_generated = false;
2772
2773                 if (pinfo->prefix_len == 64) {
2774                         memcpy(&addr, &pinfo->prefix, 8);
2775
2776                         if (!ipv6_addr_any(&in6_dev->token)) {
2777                                 read_lock_bh(&in6_dev->lock);
2778                                 memcpy(addr.s6_addr + 8,
2779                                        in6_dev->token.s6_addr + 8, 8);
2780                                 read_unlock_bh(&in6_dev->lock);
2781                                 tokenized = true;
2782                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2783                                    !ipv6_generate_stable_address(&addr, 0,
2784                                                                  in6_dev)) {
2785                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2786                                 goto ok;
2787                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2788                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2789                                 goto put;
2790                         } else {
2791                                 dev_addr_generated = true;
2792                         }
2793                         goto ok;
2794                 }
2795                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2796                                     pinfo->prefix_len);
2797                 goto put;
2798
2799 ok:
2800                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2801                                                    &addr, addr_type,
2802                                                    addr_flags, sllao,
2803                                                    tokenized, valid_lft,
2804                                                    prefered_lft);
2805                 if (err)
2806                         goto put;
2807
2808                 /* Ignore error case here because previous prefix add addr was
2809                  * successful which will be notified.
2810                  */
2811                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2812                                               addr_type, addr_flags, sllao,
2813                                               tokenized, valid_lft,
2814                                               prefered_lft,
2815                                               dev_addr_generated);
2816         }
2817         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2818 put:
2819         in6_dev_put(in6_dev);
2820 }
2821
2822 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2823                 struct in6_ifreq *ireq)
2824 {
2825         struct ip_tunnel_parm p = { };
2826         int err;
2827
2828         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2829                 return -EADDRNOTAVAIL;
2830
2831         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2832         p.iph.version = 4;
2833         p.iph.ihl = 5;
2834         p.iph.protocol = IPPROTO_IPV6;
2835         p.iph.ttl = 64;
2836
2837         if (!dev->netdev_ops->ndo_tunnel_ctl)
2838                 return -EOPNOTSUPP;
2839         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2840         if (err)
2841                 return err;
2842
2843         dev = __dev_get_by_name(net, p.name);
2844         if (!dev)
2845                 return -ENOBUFS;
2846         return dev_open(dev, NULL);
2847 }
2848
2849 /*
2850  *      Set destination address.
2851  *      Special case for SIT interfaces where we create a new "virtual"
2852  *      device.
2853  */
2854 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2855 {
2856         struct net_device *dev;
2857         struct in6_ifreq ireq;
2858         int err = -ENODEV;
2859
2860         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2861                 return -ENODEV;
2862         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2863                 return -EFAULT;
2864
2865         rtnl_lock();
2866         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2867         if (dev && dev->type == ARPHRD_SIT)
2868                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2869         rtnl_unlock();
2870         return err;
2871 }
2872
2873 static int ipv6_mc_config(struct sock *sk, bool join,
2874                           const struct in6_addr *addr, int ifindex)
2875 {
2876         int ret;
2877
2878         ASSERT_RTNL();
2879
2880         lock_sock(sk);
2881         if (join)
2882                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2883         else
2884                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2885         release_sock(sk);
2886
2887         return ret;
2888 }
2889
2890 /*
2891  *      Manual configuration of address on an interface
2892  */
2893 static int inet6_addr_add(struct net *net, int ifindex,
2894                           struct ifa6_config *cfg,
2895                           struct netlink_ext_ack *extack)
2896 {
2897         struct inet6_ifaddr *ifp;
2898         struct inet6_dev *idev;
2899         struct net_device *dev;
2900         unsigned long timeout;
2901         clock_t expires;
2902         u32 flags;
2903
2904         ASSERT_RTNL();
2905
2906         if (cfg->plen > 128)
2907                 return -EINVAL;
2908
2909         /* check the lifetime */
2910         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2911                 return -EINVAL;
2912
2913         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2914                 return -EINVAL;
2915
2916         dev = __dev_get_by_index(net, ifindex);
2917         if (!dev)
2918                 return -ENODEV;
2919
2920         idev = addrconf_add_dev(dev);
2921         if (IS_ERR(idev))
2922                 return PTR_ERR(idev);
2923
2924         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2925                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2926                                          true, cfg->pfx, ifindex);
2927
2928                 if (ret < 0)
2929                         return ret;
2930         }
2931
2932         cfg->scope = ipv6_addr_scope(cfg->pfx);
2933
2934         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2935         if (addrconf_finite_timeout(timeout)) {
2936                 expires = jiffies_to_clock_t(timeout * HZ);
2937                 cfg->valid_lft = timeout;
2938                 flags = RTF_EXPIRES;
2939         } else {
2940                 expires = 0;
2941                 flags = 0;
2942                 cfg->ifa_flags |= IFA_F_PERMANENT;
2943         }
2944
2945         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2946         if (addrconf_finite_timeout(timeout)) {
2947                 if (timeout == 0)
2948                         cfg->ifa_flags |= IFA_F_DEPRECATED;
2949                 cfg->preferred_lft = timeout;
2950         }
2951
2952         ifp = ipv6_add_addr(idev, cfg, true, extack);
2953         if (!IS_ERR(ifp)) {
2954                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2955                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2956                                               ifp->rt_priority, dev, expires,
2957                                               flags, GFP_KERNEL);
2958                 }
2959
2960                 /* Send a netlink notification if DAD is enabled and
2961                  * optimistic flag is not set
2962                  */
2963                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2964                         ipv6_ifa_notify(0, ifp);
2965                 /*
2966                  * Note that section 3.1 of RFC 4429 indicates
2967                  * that the Optimistic flag should not be set for
2968                  * manually configured addresses
2969                  */
2970                 addrconf_dad_start(ifp);
2971                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2972                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
2973                                          cfg->preferred_lft, true, jiffies);
2974                 in6_ifa_put(ifp);
2975                 addrconf_verify_rtnl();
2976                 return 0;
2977         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2978                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2979                                cfg->pfx, ifindex);
2980         }
2981
2982         return PTR_ERR(ifp);
2983 }
2984
2985 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2986                           const struct in6_addr *pfx, unsigned int plen)
2987 {
2988         struct inet6_ifaddr *ifp;
2989         struct inet6_dev *idev;
2990         struct net_device *dev;
2991
2992         if (plen > 128)
2993                 return -EINVAL;
2994
2995         dev = __dev_get_by_index(net, ifindex);
2996         if (!dev)
2997                 return -ENODEV;
2998
2999         idev = __in6_dev_get(dev);
3000         if (!idev)
3001                 return -ENXIO;
3002
3003         read_lock_bh(&idev->lock);
3004         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3005                 if (ifp->prefix_len == plen &&
3006                     ipv6_addr_equal(pfx, &ifp->addr)) {
3007                         in6_ifa_hold(ifp);
3008                         read_unlock_bh(&idev->lock);
3009
3010                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3011                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3012                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3013                                                  jiffies);
3014                         ipv6_del_addr(ifp);
3015                         addrconf_verify_rtnl();
3016                         if (ipv6_addr_is_multicast(pfx)) {
3017                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3018                                                false, pfx, dev->ifindex);
3019                         }
3020                         return 0;
3021                 }
3022         }
3023         read_unlock_bh(&idev->lock);
3024         return -EADDRNOTAVAIL;
3025 }
3026
3027
3028 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3029 {
3030         struct ifa6_config cfg = {
3031                 .ifa_flags = IFA_F_PERMANENT,
3032                 .preferred_lft = INFINITY_LIFE_TIME,
3033                 .valid_lft = INFINITY_LIFE_TIME,
3034         };
3035         struct in6_ifreq ireq;
3036         int err;
3037
3038         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3039                 return -EPERM;
3040
3041         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3042                 return -EFAULT;
3043
3044         cfg.pfx = &ireq.ifr6_addr;
3045         cfg.plen = ireq.ifr6_prefixlen;
3046
3047         rtnl_lock();
3048         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3049         rtnl_unlock();
3050         return err;
3051 }
3052
3053 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3054 {
3055         struct in6_ifreq ireq;
3056         int err;
3057
3058         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3059                 return -EPERM;
3060
3061         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3062                 return -EFAULT;
3063
3064         rtnl_lock();
3065         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3066                              ireq.ifr6_prefixlen);
3067         rtnl_unlock();
3068         return err;
3069 }
3070
3071 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3072                      int plen, int scope)
3073 {
3074         struct inet6_ifaddr *ifp;
3075         struct ifa6_config cfg = {
3076                 .pfx = addr,
3077                 .plen = plen,
3078                 .ifa_flags = IFA_F_PERMANENT,
3079                 .valid_lft = INFINITY_LIFE_TIME,
3080                 .preferred_lft = INFINITY_LIFE_TIME,
3081                 .scope = scope
3082         };
3083
3084         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3085         if (!IS_ERR(ifp)) {
3086                 spin_lock_bh(&ifp->lock);
3087                 ifp->flags &= ~IFA_F_TENTATIVE;
3088                 spin_unlock_bh(&ifp->lock);
3089                 rt_genid_bump_ipv6(dev_net(idev->dev));
3090                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3091                 in6_ifa_put(ifp);
3092         }
3093 }
3094
3095 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3096 static void add_v4_addrs(struct inet6_dev *idev)
3097 {
3098         struct in6_addr addr;
3099         struct net_device *dev;
3100         struct net *net = dev_net(idev->dev);
3101         int scope, plen, offset = 0;
3102         u32 pflags = 0;
3103
3104         ASSERT_RTNL();
3105
3106         memset(&addr, 0, sizeof(struct in6_addr));
3107         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3108         if (idev->dev->addr_len == sizeof(struct in6_addr))
3109                 offset = sizeof(struct in6_addr) - 4;
3110         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3111
3112         if (idev->dev->flags&IFF_POINTOPOINT) {
3113                 addr.s6_addr32[0] = htonl(0xfe800000);
3114                 scope = IFA_LINK;
3115                 plen = 64;
3116         } else {
3117                 scope = IPV6_ADDR_COMPATv4;
3118                 plen = 96;
3119                 pflags |= RTF_NONEXTHOP;
3120         }
3121
3122         if (addr.s6_addr32[3]) {
3123                 add_addr(idev, &addr, plen, scope);
3124                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3125                                       GFP_KERNEL);
3126                 return;
3127         }
3128
3129         for_each_netdev(net, dev) {
3130                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3131                 if (in_dev && (dev->flags & IFF_UP)) {
3132                         struct in_ifaddr *ifa;
3133                         int flag = scope;
3134
3135                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3136                                 addr.s6_addr32[3] = ifa->ifa_local;
3137
3138                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3139                                         continue;
3140                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3141                                         if (idev->dev->flags&IFF_POINTOPOINT)
3142                                                 continue;
3143                                         flag |= IFA_HOST;
3144                                 }
3145
3146                                 add_addr(idev, &addr, plen, flag);
3147                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3148                                                       0, pflags, GFP_KERNEL);
3149                         }
3150                 }
3151         }
3152 }
3153 #endif
3154
3155 static void init_loopback(struct net_device *dev)
3156 {
3157         struct inet6_dev  *idev;
3158
3159         /* ::1 */
3160
3161         ASSERT_RTNL();
3162
3163         idev = ipv6_find_idev(dev);
3164         if (IS_ERR(idev)) {
3165                 pr_debug("%s: add_dev failed\n", __func__);
3166                 return;
3167         }
3168
3169         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3170 }
3171
3172 void addrconf_add_linklocal(struct inet6_dev *idev,
3173                             const struct in6_addr *addr, u32 flags)
3174 {
3175         struct ifa6_config cfg = {
3176                 .pfx = addr,
3177                 .plen = 64,
3178                 .ifa_flags = flags | IFA_F_PERMANENT,
3179                 .valid_lft = INFINITY_LIFE_TIME,
3180                 .preferred_lft = INFINITY_LIFE_TIME,
3181                 .scope = IFA_LINK
3182         };
3183         struct inet6_ifaddr *ifp;
3184
3185 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3186         if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3187              idev->cnf.optimistic_dad) &&
3188             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3189                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3190 #endif
3191
3192         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3193         if (!IS_ERR(ifp)) {
3194                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3195                                       0, 0, GFP_ATOMIC);
3196                 addrconf_dad_start(ifp);
3197                 in6_ifa_put(ifp);
3198         }
3199 }
3200 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3201
3202 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3203 {
3204         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3205                 return true;
3206
3207         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3208             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3209                 return true;
3210
3211         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3212             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3213                 return true;
3214
3215         return false;
3216 }
3217
3218 static int ipv6_generate_stable_address(struct in6_addr *address,
3219                                         u8 dad_count,
3220                                         const struct inet6_dev *idev)
3221 {
3222         static DEFINE_SPINLOCK(lock);
3223         static __u32 digest[SHA1_DIGEST_WORDS];
3224         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3225
3226         static union {
3227                 char __data[SHA1_BLOCK_SIZE];
3228                 struct {
3229                         struct in6_addr secret;
3230                         __be32 prefix[2];
3231                         unsigned char hwaddr[MAX_ADDR_LEN];
3232                         u8 dad_count;
3233                 } __packed;
3234         } data;
3235
3236         struct in6_addr secret;
3237         struct in6_addr temp;
3238         struct net *net = dev_net(idev->dev);
3239
3240         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3241
3242         if (idev->cnf.stable_secret.initialized)
3243                 secret = idev->cnf.stable_secret.secret;
3244         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3245                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3246         else
3247                 return -1;
3248
3249 retry:
3250         spin_lock_bh(&lock);
3251
3252         sha1_init(digest);
3253         memset(&data, 0, sizeof(data));
3254         memset(workspace, 0, sizeof(workspace));
3255         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3256         data.prefix[0] = address->s6_addr32[0];
3257         data.prefix[1] = address->s6_addr32[1];
3258         data.secret = secret;
3259         data.dad_count = dad_count;
3260
3261         sha1_transform(digest, data.__data, workspace);
3262
3263         temp = *address;
3264         temp.s6_addr32[2] = (__force __be32)digest[0];
3265         temp.s6_addr32[3] = (__force __be32)digest[1];
3266
3267         spin_unlock_bh(&lock);
3268
3269         if (ipv6_reserved_interfaceid(temp)) {
3270                 dad_count++;
3271                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3272                         return -1;
3273                 goto retry;
3274         }
3275
3276         *address = temp;
3277         return 0;
3278 }
3279
3280 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3281 {
3282         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3283
3284         if (s->initialized)
3285                 return;
3286         s = &idev->cnf.stable_secret;
3287         get_random_bytes(&s->secret, sizeof(s->secret));
3288         s->initialized = true;
3289 }
3290
3291 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3292 {
3293         struct in6_addr addr;
3294
3295         /* no link local addresses on L3 master devices */
3296         if (netif_is_l3_master(idev->dev))
3297                 return;
3298
3299         /* no link local addresses on devices flagged as slaves */
3300         if (idev->dev->flags & IFF_SLAVE)
3301                 return;
3302
3303         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3304
3305         switch (idev->cnf.addr_gen_mode) {
3306         case IN6_ADDR_GEN_MODE_RANDOM:
3307                 ipv6_gen_mode_random_init(idev);
3308                 fallthrough;
3309         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3310                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3311                         addrconf_add_linklocal(idev, &addr,
3312                                                IFA_F_STABLE_PRIVACY);
3313                 else if (prefix_route)
3314                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3315                                               0, 0, GFP_KERNEL);
3316                 break;
3317         case IN6_ADDR_GEN_MODE_EUI64:
3318                 /* addrconf_add_linklocal also adds a prefix_route and we
3319                  * only need to care about prefix routes if ipv6_generate_eui64
3320                  * couldn't generate one.
3321                  */
3322                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3323                         addrconf_add_linklocal(idev, &addr, 0);
3324                 else if (prefix_route)
3325                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3326                                               0, 0, GFP_KERNEL);
3327                 break;
3328         case IN6_ADDR_GEN_MODE_NONE:
3329         default:
3330                 /* will not add any link local address */
3331                 break;
3332         }
3333 }
3334
3335 static void addrconf_dev_config(struct net_device *dev)
3336 {
3337         struct inet6_dev *idev;
3338
3339         ASSERT_RTNL();
3340
3341         if ((dev->type != ARPHRD_ETHER) &&
3342             (dev->type != ARPHRD_FDDI) &&
3343             (dev->type != ARPHRD_ARCNET) &&
3344             (dev->type != ARPHRD_INFINIBAND) &&
3345             (dev->type != ARPHRD_IEEE1394) &&
3346             (dev->type != ARPHRD_TUNNEL6) &&
3347             (dev->type != ARPHRD_6LOWPAN) &&
3348             (dev->type != ARPHRD_TUNNEL) &&
3349             (dev->type != ARPHRD_NONE) &&
3350             (dev->type != ARPHRD_RAWIP)) {
3351                 /* Alas, we support only Ethernet autoconfiguration. */
3352                 idev = __in6_dev_get(dev);
3353                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3354                     dev->flags & IFF_MULTICAST)
3355                         ipv6_mc_up(idev);
3356                 return;
3357         }
3358
3359         idev = addrconf_add_dev(dev);
3360         if (IS_ERR(idev))
3361                 return;
3362
3363         /* this device type has no EUI support */
3364         if (dev->type == ARPHRD_NONE &&
3365             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3366                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3367
3368         addrconf_addr_gen(idev, false);
3369 }
3370
3371 #if IS_ENABLED(CONFIG_IPV6_SIT)
3372 static void addrconf_sit_config(struct net_device *dev)
3373 {
3374         struct inet6_dev *idev;
3375
3376         ASSERT_RTNL();
3377
3378         /*
3379          * Configure the tunnel with one of our IPv4
3380          * addresses... we should configure all of
3381          * our v4 addrs in the tunnel
3382          */
3383
3384         idev = ipv6_find_idev(dev);
3385         if (IS_ERR(idev)) {
3386                 pr_debug("%s: add_dev failed\n", __func__);
3387                 return;
3388         }
3389
3390         if (dev->priv_flags & IFF_ISATAP) {
3391                 addrconf_addr_gen(idev, false);
3392                 return;
3393         }
3394
3395         add_v4_addrs(idev);
3396
3397         if (dev->flags&IFF_POINTOPOINT)
3398                 addrconf_add_mroute(dev);
3399 }
3400 #endif
3401
3402 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3403 static void addrconf_gre_config(struct net_device *dev)
3404 {
3405         struct inet6_dev *idev;
3406
3407         ASSERT_RTNL();
3408
3409         idev = ipv6_find_idev(dev);
3410         if (IS_ERR(idev)) {
3411                 pr_debug("%s: add_dev failed\n", __func__);
3412                 return;
3413         }
3414
3415         if (dev->type == ARPHRD_ETHER) {
3416                 addrconf_addr_gen(idev, true);
3417                 return;
3418         }
3419
3420         add_v4_addrs(idev);
3421
3422         if (dev->flags & IFF_POINTOPOINT)
3423                 addrconf_add_mroute(dev);
3424 }
3425 #endif
3426
3427 static int fixup_permanent_addr(struct net *net,
3428                                 struct inet6_dev *idev,
3429                                 struct inet6_ifaddr *ifp)
3430 {
3431         /* !fib6_node means the host route was removed from the
3432          * FIB, for example, if 'lo' device is taken down. In that
3433          * case regenerate the host route.
3434          */
3435         if (!ifp->rt || !ifp->rt->fib6_node) {
3436                 struct fib6_info *f6i, *prev;
3437
3438                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3439                                          GFP_ATOMIC);
3440                 if (IS_ERR(f6i))
3441                         return PTR_ERR(f6i);
3442
3443                 /* ifp->rt can be accessed outside of rtnl */
3444                 spin_lock(&ifp->lock);
3445                 prev = ifp->rt;
3446                 ifp->rt = f6i;
3447                 spin_unlock(&ifp->lock);
3448
3449                 fib6_info_release(prev);
3450         }
3451
3452         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3453                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3454                                       ifp->rt_priority, idev->dev, 0, 0,
3455                                       GFP_ATOMIC);
3456         }
3457
3458         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3459                 addrconf_dad_start(ifp);
3460
3461         return 0;
3462 }
3463
3464 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3465 {
3466         struct inet6_ifaddr *ifp, *tmp;
3467         struct inet6_dev *idev;
3468
3469         idev = __in6_dev_get(dev);
3470         if (!idev)
3471                 return;
3472
3473         write_lock_bh(&idev->lock);
3474
3475         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3476                 if ((ifp->flags & IFA_F_PERMANENT) &&
3477                     fixup_permanent_addr(net, idev, ifp) < 0) {
3478                         write_unlock_bh(&idev->lock);
3479                         in6_ifa_hold(ifp);
3480                         ipv6_del_addr(ifp);
3481                         write_lock_bh(&idev->lock);
3482
3483                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3484                                              idev->dev->name, &ifp->addr);
3485                 }
3486         }
3487
3488         write_unlock_bh(&idev->lock);
3489 }
3490
3491 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3492                            void *ptr)
3493 {
3494         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3495         struct netdev_notifier_change_info *change_info;
3496         struct netdev_notifier_changeupper_info *info;
3497         struct inet6_dev *idev = __in6_dev_get(dev);
3498         struct net *net = dev_net(dev);
3499         int run_pending = 0;
3500         int err;
3501
3502         switch (event) {
3503         case NETDEV_REGISTER:
3504                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3505                         idev = ipv6_add_dev(dev);
3506                         if (IS_ERR(idev))
3507                                 return notifier_from_errno(PTR_ERR(idev));
3508                 }
3509                 break;
3510
3511         case NETDEV_CHANGEMTU:
3512                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3513                 if (dev->mtu < IPV6_MIN_MTU) {
3514                         addrconf_ifdown(dev, dev != net->loopback_dev);
3515                         break;
3516                 }
3517
3518                 if (idev) {
3519                         rt6_mtu_change(dev, dev->mtu);
3520                         idev->cnf.mtu6 = dev->mtu;
3521                         break;
3522                 }
3523
3524                 /* allocate new idev */
3525                 idev = ipv6_add_dev(dev);
3526                 if (IS_ERR(idev))
3527                         break;
3528
3529                 /* device is still not ready */
3530                 if (!(idev->if_flags & IF_READY))
3531                         break;
3532
3533                 run_pending = 1;
3534                 fallthrough;
3535         case NETDEV_UP:
3536         case NETDEV_CHANGE:
3537                 if (dev->flags & IFF_SLAVE)
3538                         break;
3539
3540                 if (idev && idev->cnf.disable_ipv6)
3541                         break;
3542
3543                 if (event == NETDEV_UP) {
3544                         /* restore routes for permanent addresses */
3545                         addrconf_permanent_addr(net, dev);
3546
3547                         if (!addrconf_link_ready(dev)) {
3548                                 /* device is not ready yet. */
3549                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3550                                          dev->name);
3551                                 break;
3552                         }
3553
3554                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3555                                 idev = ipv6_add_dev(dev);
3556
3557                         if (!IS_ERR_OR_NULL(idev)) {
3558                                 idev->if_flags |= IF_READY;
3559                                 run_pending = 1;
3560                         }
3561                 } else if (event == NETDEV_CHANGE) {
3562                         if (!addrconf_link_ready(dev)) {
3563                                 /* device is still not ready. */
3564                                 rt6_sync_down_dev(dev, event);
3565                                 break;
3566                         }
3567
3568                         if (!IS_ERR_OR_NULL(idev)) {
3569                                 if (idev->if_flags & IF_READY) {
3570                                         /* device is already configured -
3571                                          * but resend MLD reports, we might
3572                                          * have roamed and need to update
3573                                          * multicast snooping switches
3574                                          */
3575                                         ipv6_mc_up(idev);
3576                                         change_info = ptr;
3577                                         if (change_info->flags_changed & IFF_NOARP)
3578                                                 addrconf_dad_run(idev, true);
3579                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3580                                         break;
3581                                 }
3582                                 idev->if_flags |= IF_READY;
3583                         }
3584
3585                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3586                                 dev->name);
3587
3588                         run_pending = 1;
3589                 }
3590
3591                 switch (dev->type) {
3592 #if IS_ENABLED(CONFIG_IPV6_SIT)
3593                 case ARPHRD_SIT:
3594                         addrconf_sit_config(dev);
3595                         break;
3596 #endif
3597 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3598                 case ARPHRD_IP6GRE:
3599                 case ARPHRD_IPGRE:
3600                         addrconf_gre_config(dev);
3601                         break;
3602 #endif
3603                 case ARPHRD_LOOPBACK:
3604                         init_loopback(dev);
3605                         break;
3606
3607                 default:
3608                         addrconf_dev_config(dev);
3609                         break;
3610                 }
3611
3612                 if (!IS_ERR_OR_NULL(idev)) {
3613                         if (run_pending)
3614                                 addrconf_dad_run(idev, false);
3615
3616                         /* Device has an address by now */
3617                         rt6_sync_up(dev, RTNH_F_DEAD);
3618
3619                         /*
3620                          * If the MTU changed during the interface down,
3621                          * when the interface up, the changed MTU must be
3622                          * reflected in the idev as well as routers.
3623                          */
3624                         if (idev->cnf.mtu6 != dev->mtu &&
3625                             dev->mtu >= IPV6_MIN_MTU) {
3626                                 rt6_mtu_change(dev, dev->mtu);
3627                                 idev->cnf.mtu6 = dev->mtu;
3628                         }
3629                         idev->tstamp = jiffies;
3630                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3631
3632                         /*
3633                          * If the changed mtu during down is lower than
3634                          * IPV6_MIN_MTU stop IPv6 on this interface.
3635                          */
3636                         if (dev->mtu < IPV6_MIN_MTU)
3637                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3638                 }
3639                 break;
3640
3641         case NETDEV_DOWN:
3642         case NETDEV_UNREGISTER:
3643                 /*
3644                  *      Remove all addresses from this interface.
3645                  */
3646                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3647                 break;
3648
3649         case NETDEV_CHANGENAME:
3650                 if (idev) {
3651                         snmp6_unregister_dev(idev);
3652                         addrconf_sysctl_unregister(idev);
3653                         err = addrconf_sysctl_register(idev);
3654                         if (err)
3655                                 return notifier_from_errno(err);
3656                         err = snmp6_register_dev(idev);
3657                         if (err) {
3658                                 addrconf_sysctl_unregister(idev);
3659                                 return notifier_from_errno(err);
3660                         }
3661                 }
3662                 break;
3663
3664         case NETDEV_PRE_TYPE_CHANGE:
3665         case NETDEV_POST_TYPE_CHANGE:
3666                 if (idev)
3667                         addrconf_type_change(dev, event);
3668                 break;
3669
3670         case NETDEV_CHANGEUPPER:
3671                 info = ptr;
3672
3673                 /* flush all routes if dev is linked to or unlinked from
3674                  * an L3 master device (e.g., VRF)
3675                  */
3676                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3677                         addrconf_ifdown(dev, false);
3678         }
3679
3680         return NOTIFY_OK;
3681 }
3682
3683 /*
3684  *      addrconf module should be notified of a device going up
3685  */
3686 static struct notifier_block ipv6_dev_notf = {
3687         .notifier_call = addrconf_notify,
3688         .priority = ADDRCONF_NOTIFY_PRIORITY,
3689 };
3690
3691 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3692 {
3693         struct inet6_dev *idev;
3694         ASSERT_RTNL();
3695
3696         idev = __in6_dev_get(dev);
3697
3698         if (event == NETDEV_POST_TYPE_CHANGE)
3699                 ipv6_mc_remap(idev);
3700         else if (event == NETDEV_PRE_TYPE_CHANGE)
3701                 ipv6_mc_unmap(idev);
3702 }
3703
3704 static bool addr_is_local(const struct in6_addr *addr)
3705 {
3706         return ipv6_addr_type(addr) &
3707                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3708 }
3709
3710 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3711 {
3712         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3713         struct net *net = dev_net(dev);
3714         struct inet6_dev *idev;
3715         struct inet6_ifaddr *ifa, *tmp;
3716         bool keep_addr = false;
3717         int state, i;
3718
3719         ASSERT_RTNL();
3720
3721         rt6_disable_ip(dev, event);
3722
3723         idev = __in6_dev_get(dev);
3724         if (!idev)
3725                 return -ENODEV;
3726
3727         /*
3728          * Step 1: remove reference to ipv6 device from parent device.
3729          *         Do not dev_put!
3730          */
3731         if (unregister) {
3732                 idev->dead = 1;
3733
3734                 /* protected by rtnl_lock */
3735                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3736
3737                 /* Step 1.5: remove snmp6 entry */
3738                 snmp6_unregister_dev(idev);
3739
3740         }
3741
3742         /* combine the user config with event to determine if permanent
3743          * addresses are to be removed from address hash table
3744          */
3745         if (!unregister && !idev->cnf.disable_ipv6) {
3746                 /* aggregate the system setting and interface setting */
3747                 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3748
3749                 if (!_keep_addr)
3750                         _keep_addr = idev->cnf.keep_addr_on_down;
3751
3752                 keep_addr = (_keep_addr > 0);
3753         }
3754
3755         /* Step 2: clear hash table */
3756         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3757                 struct hlist_head *h = &inet6_addr_lst[i];
3758
3759                 spin_lock_bh(&addrconf_hash_lock);
3760 restart:
3761                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3762                         if (ifa->idev == idev) {
3763                                 addrconf_del_dad_work(ifa);
3764                                 /* combined flag + permanent flag decide if
3765                                  * address is retained on a down event
3766                                  */
3767                                 if (!keep_addr ||
3768                                     !(ifa->flags & IFA_F_PERMANENT) ||
3769                                     addr_is_local(&ifa->addr)) {
3770                                         hlist_del_init_rcu(&ifa->addr_lst);
3771                                         goto restart;
3772                                 }
3773                         }
3774                 }
3775                 spin_unlock_bh(&addrconf_hash_lock);
3776         }
3777
3778         write_lock_bh(&idev->lock);
3779
3780         addrconf_del_rs_timer(idev);
3781
3782         /* Step 2: clear flags for stateless addrconf */
3783         if (!unregister)
3784                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3785
3786         /* Step 3: clear tempaddr list */
3787         while (!list_empty(&idev->tempaddr_list)) {
3788                 ifa = list_first_entry(&idev->tempaddr_list,
3789                                        struct inet6_ifaddr, tmp_list);
3790                 list_del(&ifa->tmp_list);
3791                 write_unlock_bh(&idev->lock);
3792                 spin_lock_bh(&ifa->lock);
3793
3794                 if (ifa->ifpub) {
3795                         in6_ifa_put(ifa->ifpub);
3796                         ifa->ifpub = NULL;
3797                 }
3798                 spin_unlock_bh(&ifa->lock);
3799                 in6_ifa_put(ifa);
3800                 write_lock_bh(&idev->lock);
3801         }
3802
3803         list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3804                 struct fib6_info *rt = NULL;
3805                 bool keep;
3806
3807                 addrconf_del_dad_work(ifa);
3808
3809                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3810                         !addr_is_local(&ifa->addr);
3811
3812                 write_unlock_bh(&idev->lock);
3813                 spin_lock_bh(&ifa->lock);
3814
3815                 if (keep) {
3816                         /* set state to skip the notifier below */
3817                         state = INET6_IFADDR_STATE_DEAD;
3818                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3819                         if (!(ifa->flags & IFA_F_NODAD))
3820                                 ifa->flags |= IFA_F_TENTATIVE;
3821
3822                         rt = ifa->rt;
3823                         ifa->rt = NULL;
3824                 } else {
3825                         state = ifa->state;
3826                         ifa->state = INET6_IFADDR_STATE_DEAD;
3827                 }
3828
3829                 spin_unlock_bh(&ifa->lock);
3830
3831                 if (rt)
3832                         ip6_del_rt(net, rt, false);
3833
3834                 if (state != INET6_IFADDR_STATE_DEAD) {
3835                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3836                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3837                 } else {
3838                         if (idev->cnf.forwarding)
3839                                 addrconf_leave_anycast(ifa);
3840                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3841                 }
3842
3843                 write_lock_bh(&idev->lock);
3844                 if (!keep) {
3845                         list_del_rcu(&ifa->if_list);
3846                         in6_ifa_put(ifa);
3847                 }
3848         }
3849
3850         write_unlock_bh(&idev->lock);
3851
3852         /* Step 5: Discard anycast and multicast list */
3853         if (unregister) {
3854                 ipv6_ac_destroy_dev(idev);
3855                 ipv6_mc_destroy_dev(idev);
3856         } else {
3857                 ipv6_mc_down(idev);
3858         }
3859
3860         idev->tstamp = jiffies;
3861         idev->ra_mtu = 0;
3862
3863         /* Last: Shot the device (if unregistered) */
3864         if (unregister) {
3865                 addrconf_sysctl_unregister(idev);
3866                 neigh_parms_release(&nd_tbl, idev->nd_parms);
3867                 neigh_ifdown(&nd_tbl, dev);
3868                 in6_dev_put(idev);
3869         }
3870         return 0;
3871 }
3872
3873 static void addrconf_rs_timer(struct timer_list *t)
3874 {
3875         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3876         struct net_device *dev = idev->dev;
3877         struct in6_addr lladdr;
3878
3879         write_lock(&idev->lock);
3880         if (idev->dead || !(idev->if_flags & IF_READY))
3881                 goto out;
3882
3883         if (!ipv6_accept_ra(idev))
3884                 goto out;
3885
3886         /* Announcement received after solicitation was sent */
3887         if (idev->if_flags & IF_RA_RCVD)
3888                 goto out;
3889
3890         if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3891                 write_unlock(&idev->lock);
3892                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3893                         ndisc_send_rs(dev, &lladdr,
3894                                       &in6addr_linklocal_allrouters);
3895                 else
3896                         goto put;
3897
3898                 write_lock(&idev->lock);
3899                 idev->rs_interval = rfc3315_s14_backoff_update(
3900                         idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3901                 /* The wait after the last probe can be shorter */
3902                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3903                                              idev->cnf.rtr_solicits) ?
3904                                       idev->cnf.rtr_solicit_delay :
3905                                       idev->rs_interval);
3906         } else {
3907                 /*
3908                  * Note: we do not support deprecated "all on-link"
3909                  * assumption any longer.
3910                  */
3911                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3912         }
3913
3914 out:
3915         write_unlock(&idev->lock);
3916 put:
3917         in6_dev_put(idev);
3918 }
3919
3920 /*
3921  *      Duplicate Address Detection
3922  */
3923 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3924 {
3925         unsigned long rand_num;
3926         struct inet6_dev *idev = ifp->idev;
3927         u64 nonce;
3928
3929         if (ifp->flags & IFA_F_OPTIMISTIC)
3930                 rand_num = 0;
3931         else
3932                 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3933
3934         nonce = 0;
3935         if (idev->cnf.enhanced_dad ||
3936             dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3937                 do
3938                         get_random_bytes(&nonce, 6);
3939                 while (nonce == 0);
3940         }
3941         ifp->dad_nonce = nonce;
3942         ifp->dad_probes = idev->cnf.dad_transmits;
3943         addrconf_mod_dad_work(ifp, rand_num);
3944 }
3945
3946 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3947 {
3948         struct inet6_dev *idev = ifp->idev;
3949         struct net_device *dev = idev->dev;
3950         bool bump_id, notify = false;
3951         struct net *net;
3952
3953         addrconf_join_solict(dev, &ifp->addr);
3954
3955         prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3956
3957         read_lock_bh(&idev->lock);
3958         spin_lock(&ifp->lock);
3959         if (ifp->state == INET6_IFADDR_STATE_DEAD)
3960                 goto out;
3961
3962         net = dev_net(dev);
3963         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3964             (net->ipv6.devconf_all->accept_dad < 1 &&
3965              idev->cnf.accept_dad < 1) ||
3966             !(ifp->flags&IFA_F_TENTATIVE) ||
3967             ifp->flags & IFA_F_NODAD) {
3968                 bool send_na = false;
3969
3970                 if (ifp->flags & IFA_F_TENTATIVE &&
3971                     !(ifp->flags & IFA_F_OPTIMISTIC))
3972                         send_na = true;
3973                 bump_id = ifp->flags & IFA_F_TENTATIVE;
3974                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3975                 spin_unlock(&ifp->lock);
3976                 read_unlock_bh(&idev->lock);
3977
3978                 addrconf_dad_completed(ifp, bump_id, send_na);
3979                 return;
3980         }
3981
3982         if (!(idev->if_flags & IF_READY)) {
3983                 spin_unlock(&ifp->lock);
3984                 read_unlock_bh(&idev->lock);
3985                 /*
3986                  * If the device is not ready:
3987                  * - keep it tentative if it is a permanent address.
3988                  * - otherwise, kill it.
3989                  */
3990                 in6_ifa_hold(ifp);
3991                 addrconf_dad_stop(ifp, 0);
3992                 return;
3993         }
3994
3995         /*
3996          * Optimistic nodes can start receiving
3997          * Frames right away
3998          */
3999         if (ifp->flags & IFA_F_OPTIMISTIC) {
4000                 ip6_ins_rt(net, ifp->rt);
4001                 if (ipv6_use_optimistic_addr(net, idev)) {
4002                         /* Because optimistic nodes can use this address,
4003                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4004                          */
4005                         notify = true;
4006                 }
4007         }
4008
4009         addrconf_dad_kick(ifp);
4010 out:
4011         spin_unlock(&ifp->lock);
4012         read_unlock_bh(&idev->lock);
4013         if (notify)
4014                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4015 }
4016
4017 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4018 {
4019         bool begin_dad = false;
4020
4021         spin_lock_bh(&ifp->lock);
4022         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4023                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4024                 begin_dad = true;
4025         }
4026         spin_unlock_bh(&ifp->lock);
4027
4028         if (begin_dad)
4029                 addrconf_mod_dad_work(ifp, 0);
4030 }
4031
4032 static void addrconf_dad_work(struct work_struct *w)
4033 {
4034         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4035                                                 struct inet6_ifaddr,
4036                                                 dad_work);
4037         struct inet6_dev *idev = ifp->idev;
4038         bool bump_id, disable_ipv6 = false;
4039         struct in6_addr mcaddr;
4040
4041         enum {
4042                 DAD_PROCESS,
4043                 DAD_BEGIN,
4044                 DAD_ABORT,
4045         } action = DAD_PROCESS;
4046
4047         rtnl_lock();
4048
4049         spin_lock_bh(&ifp->lock);
4050         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4051                 action = DAD_BEGIN;
4052                 ifp->state = INET6_IFADDR_STATE_DAD;
4053         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4054                 action = DAD_ABORT;
4055                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4056
4057                 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4058                      idev->cnf.accept_dad > 1) &&
4059                     !idev->cnf.disable_ipv6 &&
4060                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4061                         struct in6_addr addr;
4062
4063                         addr.s6_addr32[0] = htonl(0xfe800000);
4064                         addr.s6_addr32[1] = 0;
4065
4066                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4067                             ipv6_addr_equal(&ifp->addr, &addr)) {
4068                                 /* DAD failed for link-local based on MAC */
4069                                 idev->cnf.disable_ipv6 = 1;
4070
4071                                 pr_info("%s: IPv6 being disabled!\n",
4072                                         ifp->idev->dev->name);
4073                                 disable_ipv6 = true;
4074                         }
4075                 }
4076         }
4077         spin_unlock_bh(&ifp->lock);
4078
4079         if (action == DAD_BEGIN) {
4080                 addrconf_dad_begin(ifp);
4081                 goto out;
4082         } else if (action == DAD_ABORT) {
4083                 in6_ifa_hold(ifp);
4084                 addrconf_dad_stop(ifp, 1);
4085                 if (disable_ipv6)
4086                         addrconf_ifdown(idev->dev, false);
4087                 goto out;
4088         }
4089
4090         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4091                 goto out;
4092
4093         write_lock_bh(&idev->lock);
4094         if (idev->dead || !(idev->if_flags & IF_READY)) {
4095                 write_unlock_bh(&idev->lock);
4096                 goto out;
4097         }
4098
4099         spin_lock(&ifp->lock);
4100         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4101                 spin_unlock(&ifp->lock);
4102                 write_unlock_bh(&idev->lock);
4103                 goto out;
4104         }
4105
4106         if (ifp->dad_probes == 0) {
4107                 bool send_na = false;
4108
4109                 /*
4110                  * DAD was successful
4111                  */
4112
4113                 if (ifp->flags & IFA_F_TENTATIVE &&
4114                     !(ifp->flags & IFA_F_OPTIMISTIC))
4115                         send_na = true;
4116                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4117                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4118                 spin_unlock(&ifp->lock);
4119                 write_unlock_bh(&idev->lock);
4120
4121                 addrconf_dad_completed(ifp, bump_id, send_na);
4122
4123                 goto out;
4124         }
4125
4126         ifp->dad_probes--;
4127         addrconf_mod_dad_work(ifp,
4128                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4129                                   HZ/100));
4130         spin_unlock(&ifp->lock);
4131         write_unlock_bh(&idev->lock);
4132
4133         /* send a neighbour solicitation for our addr */
4134         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4135         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4136                       ifp->dad_nonce);
4137 out:
4138         in6_ifa_put(ifp);
4139         rtnl_unlock();
4140 }
4141
4142 /* ifp->idev must be at least read locked */
4143 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4144 {
4145         struct inet6_ifaddr *ifpiter;
4146         struct inet6_dev *idev = ifp->idev;
4147
4148         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4149                 if (ifpiter->scope > IFA_LINK)
4150                         break;
4151                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4152                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4153                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4154                     IFA_F_PERMANENT)
4155                         return false;
4156         }
4157         return true;
4158 }
4159
4160 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4161                                    bool send_na)
4162 {
4163         struct net_device *dev = ifp->idev->dev;
4164         struct in6_addr lladdr;
4165         bool send_rs, send_mld;
4166
4167         addrconf_del_dad_work(ifp);
4168
4169         /*
4170          *      Configure the address for reception. Now it is valid.
4171          */
4172
4173         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4174
4175         /* If added prefix is link local and we are prepared to process
4176            router advertisements, start sending router solicitations.
4177          */
4178
4179         read_lock_bh(&ifp->idev->lock);
4180         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4181         send_rs = send_mld &&
4182                   ipv6_accept_ra(ifp->idev) &&
4183                   ifp->idev->cnf.rtr_solicits != 0 &&
4184                   (dev->flags&IFF_LOOPBACK) == 0;
4185         read_unlock_bh(&ifp->idev->lock);
4186
4187         /* While dad is in progress mld report's source address is in6_addrany.
4188          * Resend with proper ll now.
4189          */
4190         if (send_mld)
4191                 ipv6_mc_dad_complete(ifp->idev);
4192
4193         /* send unsolicited NA if enabled */
4194         if (send_na &&
4195             (ifp->idev->cnf.ndisc_notify ||
4196              dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4197                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4198                               /*router=*/ !!ifp->idev->cnf.forwarding,
4199                               /*solicited=*/ false, /*override=*/ true,
4200                               /*inc_opt=*/ true);
4201         }
4202
4203         if (send_rs) {
4204                 /*
4205                  *      If a host as already performed a random delay
4206                  *      [...] as part of DAD [...] there is no need
4207                  *      to delay again before sending the first RS
4208                  */
4209                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4210                         return;
4211                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4212
4213                 write_lock_bh(&ifp->idev->lock);
4214                 spin_lock(&ifp->lock);
4215                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4216                         ifp->idev->cnf.rtr_solicit_interval);
4217                 ifp->idev->rs_probes = 1;
4218                 ifp->idev->if_flags |= IF_RS_SENT;
4219                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4220                 spin_unlock(&ifp->lock);
4221                 write_unlock_bh(&ifp->idev->lock);
4222         }
4223
4224         if (bump_id)
4225                 rt_genid_bump_ipv6(dev_net(dev));
4226
4227         /* Make sure that a new temporary address will be created
4228          * before this temporary address becomes deprecated.
4229          */
4230         if (ifp->flags & IFA_F_TEMPORARY)
4231                 addrconf_verify_rtnl();
4232 }
4233
4234 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4235 {
4236         struct inet6_ifaddr *ifp;
4237
4238         read_lock_bh(&idev->lock);
4239         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4240                 spin_lock(&ifp->lock);
4241                 if ((ifp->flags & IFA_F_TENTATIVE &&
4242                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4243                         if (restart)
4244                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4245                         addrconf_dad_kick(ifp);
4246                 }
4247                 spin_unlock(&ifp->lock);
4248         }
4249         read_unlock_bh(&idev->lock);
4250 }
4251
4252 #ifdef CONFIG_PROC_FS
4253 struct if6_iter_state {
4254         struct seq_net_private p;
4255         int bucket;
4256         int offset;
4257 };
4258
4259 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4260 {
4261         struct if6_iter_state *state = seq->private;
4262         struct net *net = seq_file_net(seq);
4263         struct inet6_ifaddr *ifa = NULL;
4264         int p = 0;
4265
4266         /* initial bucket if pos is 0 */
4267         if (pos == 0) {
4268                 state->bucket = 0;
4269                 state->offset = 0;
4270         }
4271
4272         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4273                 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4274                                          addr_lst) {
4275                         if (!net_eq(dev_net(ifa->idev->dev), net))
4276                                 continue;
4277                         /* sync with offset */
4278                         if (p < state->offset) {
4279                                 p++;
4280                                 continue;
4281                         }
4282                         return ifa;
4283                 }
4284
4285                 /* prepare for next bucket */
4286                 state->offset = 0;
4287                 p = 0;
4288         }
4289         return NULL;
4290 }
4291
4292 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4293                                          struct inet6_ifaddr *ifa)
4294 {
4295         struct if6_iter_state *state = seq->private;
4296         struct net *net = seq_file_net(seq);
4297
4298         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4299                 if (!net_eq(dev_net(ifa->idev->dev), net))
4300                         continue;
4301                 state->offset++;
4302                 return ifa;
4303         }
4304
4305         state->offset = 0;
4306         while (++state->bucket < IN6_ADDR_HSIZE) {
4307                 hlist_for_each_entry_rcu(ifa,
4308                                      &inet6_addr_lst[state->bucket], addr_lst) {
4309                         if (!net_eq(dev_net(ifa->idev->dev), net))
4310                                 continue;
4311                         return ifa;
4312                 }
4313         }
4314
4315         return NULL;
4316 }
4317
4318 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4319         __acquires(rcu)
4320 {
4321         rcu_read_lock();
4322         return if6_get_first(seq, *pos);
4323 }
4324
4325 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4326 {
4327         struct inet6_ifaddr *ifa;
4328
4329         ifa = if6_get_next(seq, v);
4330         ++*pos;
4331         return ifa;
4332 }
4333
4334 static void if6_seq_stop(struct seq_file *seq, void *v)
4335         __releases(rcu)
4336 {
4337         rcu_read_unlock();
4338 }
4339
4340 static int if6_seq_show(struct seq_file *seq, void *v)
4341 {
4342         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4343         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4344                    &ifp->addr,
4345                    ifp->idev->dev->ifindex,
4346                    ifp->prefix_len,
4347                    ifp->scope,
4348                    (u8) ifp->flags,
4349                    ifp->idev->dev->name);
4350         return 0;
4351 }
4352
4353 static const struct seq_operations if6_seq_ops = {
4354         .start  = if6_seq_start,
4355         .next   = if6_seq_next,
4356         .show   = if6_seq_show,
4357         .stop   = if6_seq_stop,
4358 };
4359
4360 static int __net_init if6_proc_net_init(struct net *net)
4361 {
4362         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4363                         sizeof(struct if6_iter_state)))
4364                 return -ENOMEM;
4365         return 0;
4366 }
4367
4368 static void __net_exit if6_proc_net_exit(struct net *net)
4369 {
4370         remove_proc_entry("if_inet6", net->proc_net);
4371 }
4372
4373 static struct pernet_operations if6_proc_net_ops = {
4374         .init = if6_proc_net_init,
4375         .exit = if6_proc_net_exit,
4376 };
4377
4378 int __init if6_proc_init(void)
4379 {
4380         return register_pernet_subsys(&if6_proc_net_ops);
4381 }
4382
4383 void if6_proc_exit(void)
4384 {
4385         unregister_pernet_subsys(&if6_proc_net_ops);
4386 }
4387 #endif  /* CONFIG_PROC_FS */
4388
4389 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4390 /* Check if address is a home address configured on any interface. */
4391 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4392 {
4393         unsigned int hash = inet6_addr_hash(net, addr);
4394         struct inet6_ifaddr *ifp = NULL;
4395         int ret = 0;
4396
4397         rcu_read_lock();
4398         hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4399                 if (!net_eq(dev_net(ifp->idev->dev), net))
4400                         continue;
4401                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4402                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4403                         ret = 1;
4404                         break;
4405                 }
4406         }
4407         rcu_read_unlock();
4408         return ret;
4409 }
4410 #endif
4411
4412 /* RFC6554 has some algorithm to avoid loops in segment routing by
4413  * checking if the segments contains any of a local interface address.
4414  *
4415  * Quote:
4416  *
4417  * To detect loops in the SRH, a router MUST determine if the SRH
4418  * includes multiple addresses assigned to any interface on that router.
4419  * If such addresses appear more than once and are separated by at least
4420  * one address not assigned to that router.
4421  */
4422 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4423                           unsigned char nsegs)
4424 {
4425         const struct in6_addr *addr;
4426         int i, ret = 0, found = 0;
4427         struct inet6_ifaddr *ifp;
4428         bool separated = false;
4429         unsigned int hash;
4430         bool hash_found;
4431
4432         rcu_read_lock();
4433         for (i = 0; i < nsegs; i++) {
4434                 addr = &segs[i];
4435                 hash = inet6_addr_hash(net, addr);
4436
4437                 hash_found = false;
4438                 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4439                         if (!net_eq(dev_net(ifp->idev->dev), net))
4440                                 continue;
4441
4442                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4443                                 hash_found = true;
4444                                 break;
4445                         }
4446                 }
4447
4448                 if (hash_found) {
4449                         if (found > 1 && separated) {
4450                                 ret = 1;
4451                                 break;
4452                         }
4453
4454                         separated = false;
4455                         found++;
4456                 } else {
4457                         separated = true;
4458                 }
4459         }
4460         rcu_read_unlock();
4461
4462         return ret;
4463 }
4464
4465 /*
4466  *      Periodic address status verification
4467  */
4468
4469 static void addrconf_verify_rtnl(void)
4470 {
4471         unsigned long now, next, next_sec, next_sched;
4472         struct inet6_ifaddr *ifp;
4473         int i;
4474
4475         ASSERT_RTNL();
4476
4477         rcu_read_lock_bh();
4478         now = jiffies;
4479         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4480
4481         cancel_delayed_work(&addr_chk_work);
4482
4483         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4484 restart:
4485                 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4486                         unsigned long age;
4487
4488                         /* When setting preferred_lft to a value not zero or
4489                          * infinity, while valid_lft is infinity
4490                          * IFA_F_PERMANENT has a non-infinity life time.
4491                          */
4492                         if ((ifp->flags & IFA_F_PERMANENT) &&
4493                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4494                                 continue;
4495
4496                         spin_lock(&ifp->lock);
4497                         /* We try to batch several events at once. */
4498                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4499
4500                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4501                             age >= ifp->valid_lft) {
4502                                 spin_unlock(&ifp->lock);
4503                                 in6_ifa_hold(ifp);
4504                                 rcu_read_unlock_bh();
4505                                 ipv6_del_addr(ifp);
4506                                 rcu_read_lock_bh();
4507                                 goto restart;
4508                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4509                                 spin_unlock(&ifp->lock);
4510                                 continue;
4511                         } else if (age >= ifp->prefered_lft) {
4512                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4513                                 int deprecate = 0;
4514
4515                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4516                                         deprecate = 1;
4517                                         ifp->flags |= IFA_F_DEPRECATED;
4518                                 }
4519
4520                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4521                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4522                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4523
4524                                 spin_unlock(&ifp->lock);
4525
4526                                 if (deprecate) {
4527                                         in6_ifa_hold(ifp);
4528
4529                                         ipv6_ifa_notify(0, ifp);
4530                                         in6_ifa_put(ifp);
4531                                         goto restart;
4532                                 }
4533                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4534                                    !(ifp->flags&IFA_F_TENTATIVE)) {
4535                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4536                                         ifp->idev->cnf.dad_transmits *
4537                                         max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4538
4539                                 if (age >= ifp->prefered_lft - regen_advance) {
4540                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4541                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4542                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4543                                         if (!ifp->regen_count && ifpub) {
4544                                                 ifp->regen_count++;
4545                                                 in6_ifa_hold(ifp);
4546                                                 in6_ifa_hold(ifpub);
4547                                                 spin_unlock(&ifp->lock);
4548
4549                                                 spin_lock(&ifpub->lock);
4550                                                 ifpub->regen_count = 0;
4551                                                 spin_unlock(&ifpub->lock);
4552                                                 rcu_read_unlock_bh();
4553                                                 ipv6_create_tempaddr(ifpub, true);
4554                                                 in6_ifa_put(ifpub);
4555                                                 in6_ifa_put(ifp);
4556                                                 rcu_read_lock_bh();
4557                                                 goto restart;
4558                                         }
4559                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4560                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4561                                 spin_unlock(&ifp->lock);
4562                         } else {
4563                                 /* ifp->prefered_lft <= ifp->valid_lft */
4564                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4565                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4566                                 spin_unlock(&ifp->lock);
4567                         }
4568                 }
4569         }
4570
4571         next_sec = round_jiffies_up(next);
4572         next_sched = next;
4573
4574         /* If rounded timeout is accurate enough, accept it. */
4575         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4576                 next_sched = next_sec;
4577
4578         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4579         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4580                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4581
4582         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4583                  now, next, next_sec, next_sched);
4584         mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4585         rcu_read_unlock_bh();
4586 }
4587
4588 static void addrconf_verify_work(struct work_struct *w)
4589 {
4590         rtnl_lock();
4591         addrconf_verify_rtnl();
4592         rtnl_unlock();
4593 }
4594
4595 static void addrconf_verify(void)
4596 {
4597         mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4598 }
4599
4600 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4601                                      struct in6_addr **peer_pfx)
4602 {
4603         struct in6_addr *pfx = NULL;
4604
4605         *peer_pfx = NULL;
4606
4607         if (addr)
4608                 pfx = nla_data(addr);
4609
4610         if (local) {
4611                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4612                         *peer_pfx = pfx;
4613                 pfx = nla_data(local);
4614         }
4615
4616         return pfx;
4617 }
4618
4619 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4620         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4621         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4622         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4623         [IFA_FLAGS]             = { .len = sizeof(u32) },
4624         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4625         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4626 };
4627
4628 static int
4629 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4630                   struct netlink_ext_ack *extack)
4631 {
4632         struct net *net = sock_net(skb->sk);
4633         struct ifaddrmsg *ifm;
4634         struct nlattr *tb[IFA_MAX+1];
4635         struct in6_addr *pfx, *peer_pfx;
4636         u32 ifa_flags;
4637         int err;
4638
4639         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4640                                      ifa_ipv6_policy, extack);
4641         if (err < 0)
4642                 return err;
4643
4644         ifm = nlmsg_data(nlh);
4645         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4646         if (!pfx)
4647                 return -EINVAL;
4648
4649         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4650
4651         /* We ignore other flags so far. */
4652         ifa_flags &= IFA_F_MANAGETEMPADDR;
4653
4654         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4655                               ifm->ifa_prefixlen);
4656 }
4657
4658 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4659                                unsigned long expires, u32 flags,
4660                                bool modify_peer)
4661 {
4662         struct fib6_info *f6i;
4663         u32 prio;
4664
4665         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4666                                         ifp->prefix_len,
4667                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4668         if (!f6i)
4669                 return -ENOENT;
4670
4671         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4672         if (f6i->fib6_metric != prio) {
4673                 /* delete old one */
4674                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4675
4676                 /* add new one */
4677                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4678                                       ifp->prefix_len,
4679                                       ifp->rt_priority, ifp->idev->dev,
4680                                       expires, flags, GFP_KERNEL);
4681         } else {
4682                 if (!expires)
4683                         fib6_clean_expires(f6i);
4684                 else
4685                         fib6_set_expires(f6i, expires);
4686
4687                 fib6_info_release(f6i);
4688         }
4689
4690         return 0;
4691 }
4692
4693 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4694 {
4695         u32 flags;
4696         clock_t expires;
4697         unsigned long timeout;
4698         bool was_managetempaddr;
4699         bool had_prefixroute;
4700         bool new_peer = false;
4701
4702         ASSERT_RTNL();
4703
4704         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4705                 return -EINVAL;
4706
4707         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4708             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4709                 return -EINVAL;
4710
4711         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4712                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4713
4714         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4715         if (addrconf_finite_timeout(timeout)) {
4716                 expires = jiffies_to_clock_t(timeout * HZ);
4717                 cfg->valid_lft = timeout;
4718                 flags = RTF_EXPIRES;
4719         } else {
4720                 expires = 0;
4721                 flags = 0;
4722                 cfg->ifa_flags |= IFA_F_PERMANENT;
4723         }
4724
4725         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4726         if (addrconf_finite_timeout(timeout)) {
4727                 if (timeout == 0)
4728                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4729                 cfg->preferred_lft = timeout;
4730         }
4731
4732         if (cfg->peer_pfx &&
4733             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4734                 if (!ipv6_addr_any(&ifp->peer_addr))
4735                         cleanup_prefix_route(ifp, expires, true, true);
4736                 new_peer = true;
4737         }
4738
4739         spin_lock_bh(&ifp->lock);
4740         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4741         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4742                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4743         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4744                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4745                         IFA_F_NOPREFIXROUTE);
4746         ifp->flags |= cfg->ifa_flags;
4747         ifp->tstamp = jiffies;
4748         ifp->valid_lft = cfg->valid_lft;
4749         ifp->prefered_lft = cfg->preferred_lft;
4750
4751         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4752                 ifp->rt_priority = cfg->rt_priority;
4753
4754         if (new_peer)
4755                 ifp->peer_addr = *cfg->peer_pfx;
4756
4757         spin_unlock_bh(&ifp->lock);
4758         if (!(ifp->flags&IFA_F_TENTATIVE))
4759                 ipv6_ifa_notify(0, ifp);
4760
4761         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4762                 int rc = -ENOENT;
4763
4764                 if (had_prefixroute)
4765                         rc = modify_prefix_route(ifp, expires, flags, false);
4766
4767                 /* prefix route could have been deleted; if so restore it */
4768                 if (rc == -ENOENT) {
4769                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4770                                               ifp->rt_priority, ifp->idev->dev,
4771                                               expires, flags, GFP_KERNEL);
4772                 }
4773
4774                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4775                         rc = modify_prefix_route(ifp, expires, flags, true);
4776
4777                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4778                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4779                                               ifp->rt_priority, ifp->idev->dev,
4780                                               expires, flags, GFP_KERNEL);
4781                 }
4782         } else if (had_prefixroute) {
4783                 enum cleanup_prefix_rt_t action;
4784                 unsigned long rt_expires;
4785
4786                 write_lock_bh(&ifp->idev->lock);
4787                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4788                 write_unlock_bh(&ifp->idev->lock);
4789
4790                 if (action != CLEANUP_PREFIX_RT_NOP) {
4791                         cleanup_prefix_route(ifp, rt_expires,
4792                                 action == CLEANUP_PREFIX_RT_DEL, false);
4793                 }
4794         }
4795
4796         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4797                 if (was_managetempaddr &&
4798                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4799                         cfg->valid_lft = 0;
4800                         cfg->preferred_lft = 0;
4801                 }
4802                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4803                                  cfg->preferred_lft, !was_managetempaddr,
4804                                  jiffies);
4805         }
4806
4807         addrconf_verify_rtnl();
4808
4809         return 0;
4810 }
4811
4812 static int
4813 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4814                   struct netlink_ext_ack *extack)
4815 {
4816         struct net *net = sock_net(skb->sk);
4817         struct ifaddrmsg *ifm;
4818         struct nlattr *tb[IFA_MAX+1];
4819         struct in6_addr *peer_pfx;
4820         struct inet6_ifaddr *ifa;
4821         struct net_device *dev;
4822         struct inet6_dev *idev;
4823         struct ifa6_config cfg;
4824         int err;
4825
4826         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4827                                      ifa_ipv6_policy, extack);
4828         if (err < 0)
4829                 return err;
4830
4831         memset(&cfg, 0, sizeof(cfg));
4832
4833         ifm = nlmsg_data(nlh);
4834         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4835         if (!cfg.pfx)
4836                 return -EINVAL;
4837
4838         cfg.peer_pfx = peer_pfx;
4839         cfg.plen = ifm->ifa_prefixlen;
4840         if (tb[IFA_RT_PRIORITY])
4841                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4842
4843         cfg.valid_lft = INFINITY_LIFE_TIME;
4844         cfg.preferred_lft = INFINITY_LIFE_TIME;
4845
4846         if (tb[IFA_CACHEINFO]) {
4847                 struct ifa_cacheinfo *ci;
4848
4849                 ci = nla_data(tb[IFA_CACHEINFO]);
4850                 cfg.valid_lft = ci->ifa_valid;
4851                 cfg.preferred_lft = ci->ifa_prefered;
4852         }
4853
4854         dev =  __dev_get_by_index(net, ifm->ifa_index);
4855         if (!dev)
4856                 return -ENODEV;
4857
4858         if (tb[IFA_FLAGS])
4859                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4860         else
4861                 cfg.ifa_flags = ifm->ifa_flags;
4862
4863         /* We ignore other flags so far. */
4864         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4865                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4866                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4867
4868         idev = ipv6_find_idev(dev);
4869         if (IS_ERR(idev))
4870                 return PTR_ERR(idev);
4871
4872         if (!ipv6_allow_optimistic_dad(net, idev))
4873                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4874
4875         if (cfg.ifa_flags & IFA_F_NODAD &&
4876             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4877                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4878                 return -EINVAL;
4879         }
4880
4881         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4882         if (!ifa) {
4883                 /*
4884                  * It would be best to check for !NLM_F_CREATE here but
4885                  * userspace already relies on not having to provide this.
4886                  */
4887                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4888         }
4889
4890         if (nlh->nlmsg_flags & NLM_F_EXCL ||
4891             !(nlh->nlmsg_flags & NLM_F_REPLACE))
4892                 err = -EEXIST;
4893         else
4894                 err = inet6_addr_modify(ifa, &cfg);
4895
4896         in6_ifa_put(ifa);
4897
4898         return err;
4899 }
4900
4901 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4902                           u8 scope, int ifindex)
4903 {
4904         struct ifaddrmsg *ifm;
4905
4906         ifm = nlmsg_data(nlh);
4907         ifm->ifa_family = AF_INET6;
4908         ifm->ifa_prefixlen = prefixlen;
4909         ifm->ifa_flags = flags;
4910         ifm->ifa_scope = scope;
4911         ifm->ifa_index = ifindex;
4912 }
4913
4914 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4915                          unsigned long tstamp, u32 preferred, u32 valid)
4916 {
4917         struct ifa_cacheinfo ci;
4918
4919         ci.cstamp = cstamp_delta(cstamp);
4920         ci.tstamp = cstamp_delta(tstamp);
4921         ci.ifa_prefered = preferred;
4922         ci.ifa_valid = valid;
4923
4924         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4925 }
4926
4927 static inline int rt_scope(int ifa_scope)
4928 {
4929         if (ifa_scope & IFA_HOST)
4930                 return RT_SCOPE_HOST;
4931         else if (ifa_scope & IFA_LINK)
4932                 return RT_SCOPE_LINK;
4933         else if (ifa_scope & IFA_SITE)
4934                 return RT_SCOPE_SITE;
4935         else
4936                 return RT_SCOPE_UNIVERSE;
4937 }
4938
4939 static inline int inet6_ifaddr_msgsize(void)
4940 {
4941         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4942                + nla_total_size(16) /* IFA_LOCAL */
4943                + nla_total_size(16) /* IFA_ADDRESS */
4944                + nla_total_size(sizeof(struct ifa_cacheinfo))
4945                + nla_total_size(4)  /* IFA_FLAGS */
4946                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
4947 }
4948
4949 enum addr_type_t {
4950         UNICAST_ADDR,
4951         MULTICAST_ADDR,
4952         ANYCAST_ADDR,
4953 };
4954
4955 struct inet6_fill_args {
4956         u32 portid;
4957         u32 seq;
4958         int event;
4959         unsigned int flags;
4960         int netnsid;
4961         int ifindex;
4962         enum addr_type_t type;
4963 };
4964
4965 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4966                              struct inet6_fill_args *args)
4967 {
4968         struct nlmsghdr  *nlh;
4969         u32 preferred, valid;
4970
4971         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4972                         sizeof(struct ifaddrmsg), args->flags);
4973         if (!nlh)
4974                 return -EMSGSIZE;
4975
4976         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4977                       ifa->idev->dev->ifindex);
4978
4979         if (args->netnsid >= 0 &&
4980             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4981                 goto error;
4982
4983         if (!((ifa->flags&IFA_F_PERMANENT) &&
4984               (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4985                 preferred = ifa->prefered_lft;
4986                 valid = ifa->valid_lft;
4987                 if (preferred != INFINITY_LIFE_TIME) {
4988                         long tval = (jiffies - ifa->tstamp)/HZ;
4989                         if (preferred > tval)
4990                                 preferred -= tval;
4991                         else
4992                                 preferred = 0;
4993                         if (valid != INFINITY_LIFE_TIME) {
4994                                 if (valid > tval)
4995                                         valid -= tval;
4996                                 else
4997                                         valid = 0;
4998                         }
4999                 }
5000         } else {
5001                 preferred = INFINITY_LIFE_TIME;
5002                 valid = INFINITY_LIFE_TIME;
5003         }
5004
5005         if (!ipv6_addr_any(&ifa->peer_addr)) {
5006                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5007                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5008                         goto error;
5009         } else
5010                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5011                         goto error;
5012
5013         if (ifa->rt_priority &&
5014             nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
5015                 goto error;
5016
5017         if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
5018                 goto error;
5019
5020         if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5021                 goto error;
5022
5023         nlmsg_end(skb, nlh);
5024         return 0;
5025
5026 error:
5027         nlmsg_cancel(skb, nlh);
5028         return -EMSGSIZE;
5029 }
5030
5031 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5032                                struct inet6_fill_args *args)
5033 {
5034         struct nlmsghdr  *nlh;
5035         u8 scope = RT_SCOPE_UNIVERSE;
5036         int ifindex = ifmca->idev->dev->ifindex;
5037
5038         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5039                 scope = RT_SCOPE_SITE;
5040
5041         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5042                         sizeof(struct ifaddrmsg), args->flags);
5043         if (!nlh)
5044                 return -EMSGSIZE;
5045
5046         if (args->netnsid >= 0 &&
5047             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5048                 nlmsg_cancel(skb, nlh);
5049                 return -EMSGSIZE;
5050         }
5051
5052         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5053         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5054             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5055                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5056                 nlmsg_cancel(skb, nlh);
5057                 return -EMSGSIZE;
5058         }
5059
5060         nlmsg_end(skb, nlh);
5061         return 0;
5062 }
5063
5064 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5065                                struct inet6_fill_args *args)
5066 {
5067         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5068         int ifindex = dev ? dev->ifindex : 1;
5069         struct nlmsghdr  *nlh;
5070         u8 scope = RT_SCOPE_UNIVERSE;
5071
5072         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5073                 scope = RT_SCOPE_SITE;
5074
5075         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5076                         sizeof(struct ifaddrmsg), args->flags);
5077         if (!nlh)
5078                 return -EMSGSIZE;
5079
5080         if (args->netnsid >= 0 &&
5081             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5082                 nlmsg_cancel(skb, nlh);
5083                 return -EMSGSIZE;
5084         }
5085
5086         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5087         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5088             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5089                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5090                 nlmsg_cancel(skb, nlh);
5091                 return -EMSGSIZE;
5092         }
5093
5094         nlmsg_end(skb, nlh);
5095         return 0;
5096 }
5097
5098 /* called with rcu_read_lock() */
5099 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5100                           struct netlink_callback *cb, int s_ip_idx,
5101                           struct inet6_fill_args *fillargs)
5102 {
5103         struct ifmcaddr6 *ifmca;
5104         struct ifacaddr6 *ifaca;
5105         int ip_idx = 0;
5106         int err = 1;
5107
5108         read_lock_bh(&idev->lock);
5109         switch (fillargs->type) {
5110         case UNICAST_ADDR: {
5111                 struct inet6_ifaddr *ifa;
5112                 fillargs->event = RTM_NEWADDR;
5113
5114                 /* unicast address incl. temp addr */
5115                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5116                         if (ip_idx < s_ip_idx)
5117                                 goto next;
5118                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5119                         if (err < 0)
5120                                 break;
5121                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5122 next:
5123                         ip_idx++;
5124                 }
5125                 break;
5126         }
5127         case MULTICAST_ADDR:
5128                 read_unlock_bh(&idev->lock);
5129                 fillargs->event = RTM_GETMULTICAST;
5130
5131                 /* multicast address */
5132                 for (ifmca = rcu_dereference(idev->mc_list);
5133                      ifmca;
5134                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5135                         if (ip_idx < s_ip_idx)
5136                                 continue;
5137                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5138                         if (err < 0)
5139                                 break;
5140                 }
5141                 read_lock_bh(&idev->lock);
5142                 break;
5143         case ANYCAST_ADDR:
5144                 fillargs->event = RTM_GETANYCAST;
5145                 /* anycast address */
5146                 for (ifaca = idev->ac_list; ifaca;
5147                      ifaca = ifaca->aca_next, ip_idx++) {
5148                         if (ip_idx < s_ip_idx)
5149                                 continue;
5150                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5151                         if (err < 0)
5152                                 break;
5153                 }
5154                 break;
5155         default:
5156                 break;
5157         }
5158         read_unlock_bh(&idev->lock);
5159         cb->args[2] = ip_idx;
5160         return err;
5161 }
5162
5163 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5164                                        struct inet6_fill_args *fillargs,
5165                                        struct net **tgt_net, struct sock *sk,
5166                                        struct netlink_callback *cb)
5167 {
5168         struct netlink_ext_ack *extack = cb->extack;
5169         struct nlattr *tb[IFA_MAX+1];
5170         struct ifaddrmsg *ifm;
5171         int err, i;
5172
5173         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5174                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5175                 return -EINVAL;
5176         }
5177
5178         ifm = nlmsg_data(nlh);
5179         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5180                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5181                 return -EINVAL;
5182         }
5183
5184         fillargs->ifindex = ifm->ifa_index;
5185         if (fillargs->ifindex) {
5186                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5187                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5188         }
5189
5190         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5191                                             ifa_ipv6_policy, extack);
5192         if (err < 0)
5193                 return err;
5194
5195         for (i = 0; i <= IFA_MAX; ++i) {
5196                 if (!tb[i])
5197                         continue;
5198
5199                 if (i == IFA_TARGET_NETNSID) {
5200                         struct net *net;
5201
5202                         fillargs->netnsid = nla_get_s32(tb[i]);
5203                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5204                         if (IS_ERR(net)) {
5205                                 fillargs->netnsid = -1;
5206                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5207                                 return PTR_ERR(net);
5208                         }
5209                         *tgt_net = net;
5210                 } else {
5211                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5212                         return -EINVAL;
5213                 }
5214         }
5215
5216         return 0;
5217 }
5218
5219 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5220                            enum addr_type_t type)
5221 {
5222         const struct nlmsghdr *nlh = cb->nlh;
5223         struct inet6_fill_args fillargs = {
5224                 .portid = NETLINK_CB(cb->skb).portid,
5225                 .seq = cb->nlh->nlmsg_seq,
5226                 .flags = NLM_F_MULTI,
5227                 .netnsid = -1,
5228                 .type = type,
5229         };
5230         struct net *tgt_net = sock_net(skb->sk);
5231         int idx, s_idx, s_ip_idx;
5232         int h, s_h;
5233         struct net_device *dev;
5234         struct inet6_dev *idev;
5235         struct hlist_head *head;
5236         int err = 0;
5237
5238         s_h = cb->args[0];
5239         s_idx = idx = cb->args[1];
5240         s_ip_idx = cb->args[2];
5241
5242         if (cb->strict_check) {
5243                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5244                                                   skb->sk, cb);
5245                 if (err < 0)
5246                         goto put_tgt_net;
5247
5248                 err = 0;
5249                 if (fillargs.ifindex) {
5250                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5251                         if (!dev) {
5252                                 err = -ENODEV;
5253                                 goto put_tgt_net;
5254                         }
5255                         idev = __in6_dev_get(dev);
5256                         if (idev) {
5257                                 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5258                                                      &fillargs);
5259                                 if (err > 0)
5260                                         err = 0;
5261                         }
5262                         goto put_tgt_net;
5263                 }
5264         }
5265
5266         rcu_read_lock();
5267         cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5268         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5269                 idx = 0;
5270                 head = &tgt_net->dev_index_head[h];
5271                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5272                         if (idx < s_idx)
5273                                 goto cont;
5274                         if (h > s_h || idx > s_idx)
5275                                 s_ip_idx = 0;
5276                         idev = __in6_dev_get(dev);
5277                         if (!idev)
5278                                 goto cont;
5279
5280                         if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5281                                            &fillargs) < 0)
5282                                 goto done;
5283 cont:
5284                         idx++;
5285                 }
5286         }
5287 done:
5288         rcu_read_unlock();
5289         cb->args[0] = h;
5290         cb->args[1] = idx;
5291 put_tgt_net:
5292         if (fillargs.netnsid >= 0)
5293                 put_net(tgt_net);
5294
5295         return skb->len ? : err;
5296 }
5297
5298 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5299 {
5300         enum addr_type_t type = UNICAST_ADDR;
5301
5302         return inet6_dump_addr(skb, cb, type);
5303 }
5304
5305 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5306 {
5307         enum addr_type_t type = MULTICAST_ADDR;
5308
5309         return inet6_dump_addr(skb, cb, type);
5310 }
5311
5312
5313 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5314 {
5315         enum addr_type_t type = ANYCAST_ADDR;
5316
5317         return inet6_dump_addr(skb, cb, type);
5318 }
5319
5320 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5321                                        const struct nlmsghdr *nlh,
5322                                        struct nlattr **tb,
5323                                        struct netlink_ext_ack *extack)
5324 {
5325         struct ifaddrmsg *ifm;
5326         int i, err;
5327
5328         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5329                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5330                 return -EINVAL;
5331         }
5332
5333         if (!netlink_strict_get_check(skb))
5334                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5335                                               ifa_ipv6_policy, extack);
5336
5337         ifm = nlmsg_data(nlh);
5338         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5339                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5340                 return -EINVAL;
5341         }
5342
5343         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5344                                             ifa_ipv6_policy, extack);
5345         if (err)
5346                 return err;
5347
5348         for (i = 0; i <= IFA_MAX; i++) {
5349                 if (!tb[i])
5350                         continue;
5351
5352                 switch (i) {
5353                 case IFA_TARGET_NETNSID:
5354                 case IFA_ADDRESS:
5355                 case IFA_LOCAL:
5356                         break;
5357                 default:
5358                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5359                         return -EINVAL;
5360                 }
5361         }
5362
5363         return 0;
5364 }
5365
5366 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5367                              struct netlink_ext_ack *extack)
5368 {
5369         struct net *tgt_net = sock_net(in_skb->sk);
5370         struct inet6_fill_args fillargs = {
5371                 .portid = NETLINK_CB(in_skb).portid,
5372                 .seq = nlh->nlmsg_seq,
5373                 .event = RTM_NEWADDR,
5374                 .flags = 0,
5375                 .netnsid = -1,
5376         };
5377         struct ifaddrmsg *ifm;
5378         struct nlattr *tb[IFA_MAX+1];
5379         struct in6_addr *addr = NULL, *peer;
5380         struct net_device *dev = NULL;
5381         struct inet6_ifaddr *ifa;
5382         struct sk_buff *skb;
5383         int err;
5384
5385         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5386         if (err < 0)
5387                 return err;
5388
5389         if (tb[IFA_TARGET_NETNSID]) {
5390                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5391
5392                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5393                                                   fillargs.netnsid);
5394                 if (IS_ERR(tgt_net))
5395                         return PTR_ERR(tgt_net);
5396         }
5397
5398         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5399         if (!addr)
5400                 return -EINVAL;
5401
5402         ifm = nlmsg_data(nlh);
5403         if (ifm->ifa_index)
5404                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5405
5406         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5407         if (!ifa) {
5408                 err = -EADDRNOTAVAIL;
5409                 goto errout;
5410         }
5411
5412         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5413         if (!skb) {
5414                 err = -ENOBUFS;
5415                 goto errout_ifa;
5416         }
5417
5418         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5419         if (err < 0) {
5420                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5421                 WARN_ON(err == -EMSGSIZE);
5422                 kfree_skb(skb);
5423                 goto errout_ifa;
5424         }
5425         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5426 errout_ifa:
5427         in6_ifa_put(ifa);
5428 errout:
5429         dev_put(dev);
5430         if (fillargs.netnsid >= 0)
5431                 put_net(tgt_net);
5432
5433         return err;
5434 }
5435
5436 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5437 {
5438         struct sk_buff *skb;
5439         struct net *net = dev_net(ifa->idev->dev);
5440         struct inet6_fill_args fillargs = {
5441                 .portid = 0,
5442                 .seq = 0,
5443                 .event = event,
5444                 .flags = 0,
5445                 .netnsid = -1,
5446         };
5447         int err = -ENOBUFS;
5448
5449         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5450         if (!skb)
5451                 goto errout;
5452
5453         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5454         if (err < 0) {
5455                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5456                 WARN_ON(err == -EMSGSIZE);
5457                 kfree_skb(skb);
5458                 goto errout;
5459         }
5460         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5461         return;
5462 errout:
5463         if (err < 0)
5464                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5465 }
5466
5467 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5468                                 __s32 *array, int bytes)
5469 {
5470         BUG_ON(bytes < (DEVCONF_MAX * 4));
5471
5472         memset(array, 0, bytes);
5473         array[DEVCONF_FORWARDING] = cnf->forwarding;
5474         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5475         array[DEVCONF_MTU6] = cnf->mtu6;
5476         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5477         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5478         array[DEVCONF_AUTOCONF] = cnf->autoconf;
5479         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5480         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5481         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5482                 jiffies_to_msecs(cnf->rtr_solicit_interval);
5483         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5484                 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5485         array[DEVCONF_RTR_SOLICIT_DELAY] =
5486                 jiffies_to_msecs(cnf->rtr_solicit_delay);
5487         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5488         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5489                 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5490         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5491                 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5492         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5493         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5494         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5495         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5496         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5497         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5498         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5499         array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
5500         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5501         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5502 #ifdef CONFIG_IPV6_ROUTER_PREF
5503         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5504         array[DEVCONF_RTR_PROBE_INTERVAL] =
5505                 jiffies_to_msecs(cnf->rtr_probe_interval);
5506 #ifdef CONFIG_IPV6_ROUTE_INFO
5507         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5508         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5509 #endif
5510 #endif
5511         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5512         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5513 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5514         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5515         array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5516 #endif
5517 #ifdef CONFIG_IPV6_MROUTE
5518         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5519 #endif
5520         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5521         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5522         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5523         array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5524         array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5525         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5526         array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5527         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5528         /* we omit DEVCONF_STABLE_SECRET for now */
5529         array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5530         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5531         array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5532         array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5533         array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5534 #ifdef CONFIG_IPV6_SEG6_HMAC
5535         array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5536 #endif
5537         array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5538         array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5539         array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5540         array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5541         array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5542         array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
5543         array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
5544         array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
5545 }
5546
5547 static inline size_t inet6_ifla6_size(void)
5548 {
5549         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5550              + nla_total_size(sizeof(struct ifla_cacheinfo))
5551              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5552              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5553              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5554              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5555              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5556              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5557              + 0;
5558 }
5559
5560 static inline size_t inet6_if_nlmsg_size(void)
5561 {
5562         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5563                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5564                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5565                + nla_total_size(4) /* IFLA_MTU */
5566                + nla_total_size(4) /* IFLA_LINK */
5567                + nla_total_size(1) /* IFLA_OPERSTATE */
5568                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5569 }
5570
5571 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5572                                         int bytes)
5573 {
5574         int i;
5575         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5576         BUG_ON(pad < 0);
5577
5578         /* Use put_unaligned() because stats may not be aligned for u64. */
5579         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5580         for (i = 1; i < ICMP6_MIB_MAX; i++)
5581                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5582
5583         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5584 }
5585
5586 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5587                                         int bytes, size_t syncpoff)
5588 {
5589         int i, c;
5590         u64 buff[IPSTATS_MIB_MAX];
5591         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5592
5593         BUG_ON(pad < 0);
5594
5595         memset(buff, 0, sizeof(buff));
5596         buff[0] = IPSTATS_MIB_MAX;
5597
5598         for_each_possible_cpu(c) {
5599                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5600                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5601         }
5602
5603         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5604         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5605 }
5606
5607 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5608                              int bytes)
5609 {
5610         switch (attrtype) {
5611         case IFLA_INET6_STATS:
5612                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5613                                      offsetof(struct ipstats_mib, syncp));
5614                 break;
5615         case IFLA_INET6_ICMP6STATS:
5616                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5617                 break;
5618         }
5619 }
5620
5621 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5622                                   u32 ext_filter_mask)
5623 {
5624         struct nlattr *nla;
5625         struct ifla_cacheinfo ci;
5626
5627         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5628                 goto nla_put_failure;
5629         ci.max_reasm_len = IPV6_MAXPLEN;
5630         ci.tstamp = cstamp_delta(idev->tstamp);
5631         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5632         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5633         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5634                 goto nla_put_failure;
5635         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5636         if (!nla)
5637                 goto nla_put_failure;
5638         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5639
5640         /* XXX - MC not implemented */
5641
5642         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5643                 return 0;
5644
5645         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5646         if (!nla)
5647                 goto nla_put_failure;
5648         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5649
5650         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5651         if (!nla)
5652                 goto nla_put_failure;
5653         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5654
5655         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5656         if (!nla)
5657                 goto nla_put_failure;
5658         read_lock_bh(&idev->lock);
5659         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5660         read_unlock_bh(&idev->lock);
5661
5662         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5663                 goto nla_put_failure;
5664
5665         if (idev->ra_mtu &&
5666             nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
5667                 goto nla_put_failure;
5668
5669         return 0;
5670
5671 nla_put_failure:
5672         return -EMSGSIZE;
5673 }
5674
5675 static size_t inet6_get_link_af_size(const struct net_device *dev,
5676                                      u32 ext_filter_mask)
5677 {
5678         if (!__in6_dev_get(dev))
5679                 return 0;
5680
5681         return inet6_ifla6_size();
5682 }
5683
5684 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5685                               u32 ext_filter_mask)
5686 {
5687         struct inet6_dev *idev = __in6_dev_get(dev);
5688
5689         if (!idev)
5690                 return -ENODATA;
5691
5692         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5693                 return -EMSGSIZE;
5694
5695         return 0;
5696 }
5697
5698 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5699                              struct netlink_ext_ack *extack)
5700 {
5701         struct inet6_ifaddr *ifp;
5702         struct net_device *dev = idev->dev;
5703         bool clear_token, update_rs = false;
5704         struct in6_addr ll_addr;
5705
5706         ASSERT_RTNL();
5707
5708         if (!token)
5709                 return -EINVAL;
5710
5711         if (dev->flags & IFF_LOOPBACK) {
5712                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5713                 return -EINVAL;
5714         }
5715
5716         if (dev->flags & IFF_NOARP) {
5717                 NL_SET_ERR_MSG_MOD(extack,
5718                                    "Device does not do neighbour discovery");
5719                 return -EINVAL;
5720         }
5721
5722         if (!ipv6_accept_ra(idev)) {
5723                 NL_SET_ERR_MSG_MOD(extack,
5724                                    "Router advertisement is disabled on device");
5725                 return -EINVAL;
5726         }
5727
5728         if (idev->cnf.rtr_solicits == 0) {
5729                 NL_SET_ERR_MSG(extack,
5730                                "Router solicitation is disabled on device");
5731                 return -EINVAL;
5732         }
5733
5734         write_lock_bh(&idev->lock);
5735
5736         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5737         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5738
5739         write_unlock_bh(&idev->lock);
5740
5741         clear_token = ipv6_addr_any(token);
5742         if (clear_token)
5743                 goto update_lft;
5744
5745         if (!idev->dead && (idev->if_flags & IF_READY) &&
5746             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5747                              IFA_F_OPTIMISTIC)) {
5748                 /* If we're not ready, then normal ifup will take care
5749                  * of this. Otherwise, we need to request our rs here.
5750                  */
5751                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5752                 update_rs = true;
5753         }
5754
5755 update_lft:
5756         write_lock_bh(&idev->lock);
5757
5758         if (update_rs) {
5759                 idev->if_flags |= IF_RS_SENT;
5760                 idev->rs_interval = rfc3315_s14_backoff_init(
5761                         idev->cnf.rtr_solicit_interval);
5762                 idev->rs_probes = 1;
5763                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5764         }
5765
5766         /* Well, that's kinda nasty ... */
5767         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5768                 spin_lock(&ifp->lock);
5769                 if (ifp->tokenized) {
5770                         ifp->valid_lft = 0;
5771                         ifp->prefered_lft = 0;
5772                 }
5773                 spin_unlock(&ifp->lock);
5774         }
5775
5776         write_unlock_bh(&idev->lock);
5777         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5778         addrconf_verify_rtnl();
5779         return 0;
5780 }
5781
5782 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5783         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5784         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5785         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5786                                             .reject_message =
5787                                                 "IFLA_INET6_RA_MTU can not be set" },
5788 };
5789
5790 static int check_addr_gen_mode(int mode)
5791 {
5792         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5793             mode != IN6_ADDR_GEN_MODE_NONE &&
5794             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5795             mode != IN6_ADDR_GEN_MODE_RANDOM)
5796                 return -EINVAL;
5797         return 1;
5798 }
5799
5800 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5801                                 int mode)
5802 {
5803         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5804             !idev->cnf.stable_secret.initialized &&
5805             !net->ipv6.devconf_dflt->stable_secret.initialized)
5806                 return -EINVAL;
5807         return 1;
5808 }
5809
5810 static int inet6_validate_link_af(const struct net_device *dev,
5811                                   const struct nlattr *nla,
5812                                   struct netlink_ext_ack *extack)
5813 {
5814         struct nlattr *tb[IFLA_INET6_MAX + 1];
5815         struct inet6_dev *idev = NULL;
5816         int err;
5817
5818         if (dev) {
5819                 idev = __in6_dev_get(dev);
5820                 if (!idev)
5821                         return -EAFNOSUPPORT;
5822         }
5823
5824         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5825                                           inet6_af_policy, extack);
5826         if (err)
5827                 return err;
5828
5829         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5830                 return -EINVAL;
5831
5832         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5833                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5834
5835                 if (check_addr_gen_mode(mode) < 0)
5836                         return -EINVAL;
5837                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5838                         return -EINVAL;
5839         }
5840
5841         return 0;
5842 }
5843
5844 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
5845                              struct netlink_ext_ack *extack)
5846 {
5847         struct inet6_dev *idev = __in6_dev_get(dev);
5848         struct nlattr *tb[IFLA_INET6_MAX + 1];
5849         int err;
5850
5851         if (!idev)
5852                 return -EAFNOSUPPORT;
5853
5854         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5855                 return -EINVAL;
5856
5857         if (tb[IFLA_INET6_TOKEN]) {
5858                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
5859                                         extack);
5860                 if (err)
5861                         return err;
5862         }
5863
5864         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5865                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5866
5867                 idev->cnf.addr_gen_mode = mode;
5868         }
5869
5870         return 0;
5871 }
5872
5873 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5874                              u32 portid, u32 seq, int event, unsigned int flags)
5875 {
5876         struct net_device *dev = idev->dev;
5877         struct ifinfomsg *hdr;
5878         struct nlmsghdr *nlh;
5879         void *protoinfo;
5880
5881         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5882         if (!nlh)
5883                 return -EMSGSIZE;
5884
5885         hdr = nlmsg_data(nlh);
5886         hdr->ifi_family = AF_INET6;
5887         hdr->__ifi_pad = 0;
5888         hdr->ifi_type = dev->type;
5889         hdr->ifi_index = dev->ifindex;
5890         hdr->ifi_flags = dev_get_flags(dev);
5891         hdr->ifi_change = 0;
5892
5893         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5894             (dev->addr_len &&
5895              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5896             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5897             (dev->ifindex != dev_get_iflink(dev) &&
5898              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5899             nla_put_u8(skb, IFLA_OPERSTATE,
5900                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5901                 goto nla_put_failure;
5902         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5903         if (!protoinfo)
5904                 goto nla_put_failure;
5905
5906         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5907                 goto nla_put_failure;
5908
5909         nla_nest_end(skb, protoinfo);
5910         nlmsg_end(skb, nlh);
5911         return 0;
5912
5913 nla_put_failure:
5914         nlmsg_cancel(skb, nlh);
5915         return -EMSGSIZE;
5916 }
5917
5918 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5919                                    struct netlink_ext_ack *extack)
5920 {
5921         struct ifinfomsg *ifm;
5922
5923         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5924                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5925                 return -EINVAL;
5926         }
5927
5928         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5929                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5930                 return -EINVAL;
5931         }
5932
5933         ifm = nlmsg_data(nlh);
5934         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5935             ifm->ifi_change || ifm->ifi_index) {
5936                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5937                 return -EINVAL;
5938         }
5939
5940         return 0;
5941 }
5942
5943 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5944 {
5945         struct net *net = sock_net(skb->sk);
5946         int h, s_h;
5947         int idx = 0, s_idx;
5948         struct net_device *dev;
5949         struct inet6_dev *idev;
5950         struct hlist_head *head;
5951
5952         /* only requests using strict checking can pass data to
5953          * influence the dump
5954          */
5955         if (cb->strict_check) {
5956                 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5957
5958                 if (err < 0)
5959                         return err;
5960         }
5961
5962         s_h = cb->args[0];
5963         s_idx = cb->args[1];
5964
5965         rcu_read_lock();
5966         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5967                 idx = 0;
5968                 head = &net->dev_index_head[h];
5969                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5970                         if (idx < s_idx)
5971                                 goto cont;
5972                         idev = __in6_dev_get(dev);
5973                         if (!idev)
5974                                 goto cont;
5975                         if (inet6_fill_ifinfo(skb, idev,
5976                                               NETLINK_CB(cb->skb).portid,
5977                                               cb->nlh->nlmsg_seq,
5978                                               RTM_NEWLINK, NLM_F_MULTI) < 0)
5979                                 goto out;
5980 cont:
5981                         idx++;
5982                 }
5983         }
5984 out:
5985         rcu_read_unlock();
5986         cb->args[1] = idx;
5987         cb->args[0] = h;
5988
5989         return skb->len;
5990 }
5991
5992 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5993 {
5994         struct sk_buff *skb;
5995         struct net *net = dev_net(idev->dev);
5996         int err = -ENOBUFS;
5997
5998         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5999         if (!skb)
6000                 goto errout;
6001
6002         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6003         if (err < 0) {
6004                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6005                 WARN_ON(err == -EMSGSIZE);
6006                 kfree_skb(skb);
6007                 goto errout;
6008         }
6009         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6010         return;
6011 errout:
6012         if (err < 0)
6013                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6014 }
6015
6016 static inline size_t inet6_prefix_nlmsg_size(void)
6017 {
6018         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6019                + nla_total_size(sizeof(struct in6_addr))
6020                + nla_total_size(sizeof(struct prefix_cacheinfo));
6021 }
6022
6023 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6024                              struct prefix_info *pinfo, u32 portid, u32 seq,
6025                              int event, unsigned int flags)
6026 {
6027         struct prefixmsg *pmsg;
6028         struct nlmsghdr *nlh;
6029         struct prefix_cacheinfo ci;
6030
6031         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6032         if (!nlh)
6033                 return -EMSGSIZE;
6034
6035         pmsg = nlmsg_data(nlh);
6036         pmsg->prefix_family = AF_INET6;
6037         pmsg->prefix_pad1 = 0;
6038         pmsg->prefix_pad2 = 0;
6039         pmsg->prefix_ifindex = idev->dev->ifindex;
6040         pmsg->prefix_len = pinfo->prefix_len;
6041         pmsg->prefix_type = pinfo->type;
6042         pmsg->prefix_pad3 = 0;
6043         pmsg->prefix_flags = 0;
6044         if (pinfo->onlink)
6045                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
6046         if (pinfo->autoconf)
6047                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
6048
6049         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6050                 goto nla_put_failure;
6051         ci.preferred_time = ntohl(pinfo->prefered);
6052         ci.valid_time = ntohl(pinfo->valid);
6053         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6054                 goto nla_put_failure;
6055         nlmsg_end(skb, nlh);
6056         return 0;
6057
6058 nla_put_failure:
6059         nlmsg_cancel(skb, nlh);
6060         return -EMSGSIZE;
6061 }
6062
6063 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6064                          struct prefix_info *pinfo)
6065 {
6066         struct sk_buff *skb;
6067         struct net *net = dev_net(idev->dev);
6068         int err = -ENOBUFS;
6069
6070         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6071         if (!skb)
6072                 goto errout;
6073
6074         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6075         if (err < 0) {
6076                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6077                 WARN_ON(err == -EMSGSIZE);
6078                 kfree_skb(skb);
6079                 goto errout;
6080         }
6081         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6082         return;
6083 errout:
6084         if (err < 0)
6085                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6086 }
6087
6088 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6089 {
6090         struct net *net = dev_net(ifp->idev->dev);
6091
6092         if (event)
6093                 ASSERT_RTNL();
6094
6095         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6096
6097         switch (event) {
6098         case RTM_NEWADDR:
6099                 /*
6100                  * If the address was optimistic we inserted the route at the
6101                  * start of our DAD process, so we don't need to do it again.
6102                  * If the device was taken down in the middle of the DAD
6103                  * cycle there is a race where we could get here without a
6104                  * host route, so nothing to insert. That will be fixed when
6105                  * the device is brought up.
6106                  */
6107                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6108                         ip6_ins_rt(net, ifp->rt);
6109                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6110                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6111                                 &ifp->addr, ifp->idev->dev->name);
6112                 }
6113
6114                 if (ifp->idev->cnf.forwarding)
6115                         addrconf_join_anycast(ifp);
6116                 if (!ipv6_addr_any(&ifp->peer_addr))
6117                         addrconf_prefix_route(&ifp->peer_addr, 128,
6118                                               ifp->rt_priority, ifp->idev->dev,
6119                                               0, 0, GFP_ATOMIC);
6120                 break;
6121         case RTM_DELADDR:
6122                 if (ifp->idev->cnf.forwarding)
6123                         addrconf_leave_anycast(ifp);
6124                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6125                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6126                         struct fib6_info *rt;
6127
6128                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6129                                                        ifp->idev->dev, 0, 0,
6130                                                        false);
6131                         if (rt)
6132                                 ip6_del_rt(net, rt, false);
6133                 }
6134                 if (ifp->rt) {
6135                         ip6_del_rt(net, ifp->rt, false);
6136                         ifp->rt = NULL;
6137                 }
6138                 rt_genid_bump_ipv6(net);
6139                 break;
6140         }
6141         atomic_inc(&net->ipv6.dev_addr_genid);
6142 }
6143
6144 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6145 {
6146         if (likely(ifp->idev->dead == 0))
6147                 __ipv6_ifa_notify(event, ifp);
6148 }
6149
6150 #ifdef CONFIG_SYSCTL
6151
6152 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6153                 void *buffer, size_t *lenp, loff_t *ppos)
6154 {
6155         int *valp = ctl->data;
6156         int val = *valp;
6157         loff_t pos = *ppos;
6158         struct ctl_table lctl;
6159         int ret;
6160
6161         /*
6162          * ctl->data points to idev->cnf.forwarding, we should
6163          * not modify it until we get the rtnl lock.
6164          */
6165         lctl = *ctl;
6166         lctl.data = &val;
6167
6168         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6169
6170         if (write)
6171                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6172         if (ret)
6173                 *ppos = pos;
6174         return ret;
6175 }
6176
6177 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6178                 void *buffer, size_t *lenp, loff_t *ppos)
6179 {
6180         struct inet6_dev *idev = ctl->extra1;
6181         int min_mtu = IPV6_MIN_MTU;
6182         struct ctl_table lctl;
6183
6184         lctl = *ctl;
6185         lctl.extra1 = &min_mtu;
6186         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6187
6188         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6189 }
6190
6191 static void dev_disable_change(struct inet6_dev *idev)
6192 {
6193         struct netdev_notifier_info info;
6194
6195         if (!idev || !idev->dev)
6196                 return;
6197
6198         netdev_notifier_info_init(&info, idev->dev);
6199         if (idev->cnf.disable_ipv6)
6200                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6201         else
6202                 addrconf_notify(NULL, NETDEV_UP, &info);
6203 }
6204
6205 static void addrconf_disable_change(struct net *net, __s32 newf)
6206 {
6207         struct net_device *dev;
6208         struct inet6_dev *idev;
6209
6210         for_each_netdev(net, dev) {
6211                 idev = __in6_dev_get(dev);
6212                 if (idev) {
6213                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6214                         idev->cnf.disable_ipv6 = newf;
6215                         if (changed)
6216                                 dev_disable_change(idev);
6217                 }
6218         }
6219 }
6220
6221 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6222 {
6223         struct net *net;
6224         int old;
6225
6226         if (!rtnl_trylock())
6227                 return restart_syscall();
6228
6229         net = (struct net *)table->extra2;
6230         old = *p;
6231         *p = newf;
6232
6233         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6234                 rtnl_unlock();
6235                 return 0;
6236         }
6237
6238         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6239                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6240                 addrconf_disable_change(net, newf);
6241         } else if ((!newf) ^ (!old))
6242                 dev_disable_change((struct inet6_dev *)table->extra1);
6243
6244         rtnl_unlock();
6245         return 0;
6246 }
6247
6248 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6249                 void *buffer, size_t *lenp, loff_t *ppos)
6250 {
6251         int *valp = ctl->data;
6252         int val = *valp;
6253         loff_t pos = *ppos;
6254         struct ctl_table lctl;
6255         int ret;
6256
6257         /*
6258          * ctl->data points to idev->cnf.disable_ipv6, we should
6259          * not modify it until we get the rtnl lock.
6260          */
6261         lctl = *ctl;
6262         lctl.data = &val;
6263
6264         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6265
6266         if (write)
6267                 ret = addrconf_disable_ipv6(ctl, valp, val);
6268         if (ret)
6269                 *ppos = pos;
6270         return ret;
6271 }
6272
6273 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6274                 void *buffer, size_t *lenp, loff_t *ppos)
6275 {
6276         int *valp = ctl->data;
6277         int ret;
6278         int old, new;
6279
6280         old = *valp;
6281         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6282         new = *valp;
6283
6284         if (write && old != new) {
6285                 struct net *net = ctl->extra2;
6286
6287                 if (!rtnl_trylock())
6288                         return restart_syscall();
6289
6290                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6291                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6292                                                      NETCONFA_PROXY_NEIGH,
6293                                                      NETCONFA_IFINDEX_DEFAULT,
6294                                                      net->ipv6.devconf_dflt);
6295                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6296                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6297                                                      NETCONFA_PROXY_NEIGH,
6298                                                      NETCONFA_IFINDEX_ALL,
6299                                                      net->ipv6.devconf_all);
6300                 else {
6301                         struct inet6_dev *idev = ctl->extra1;
6302
6303                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6304                                                      NETCONFA_PROXY_NEIGH,
6305                                                      idev->dev->ifindex,
6306                                                      &idev->cnf);
6307                 }
6308                 rtnl_unlock();
6309         }
6310
6311         return ret;
6312 }
6313
6314 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6315                                          void *buffer, size_t *lenp,
6316                                          loff_t *ppos)
6317 {
6318         int ret = 0;
6319         u32 new_val;
6320         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6321         struct net *net = (struct net *)ctl->extra2;
6322         struct ctl_table tmp = {
6323                 .data = &new_val,
6324                 .maxlen = sizeof(new_val),
6325                 .mode = ctl->mode,
6326         };
6327
6328         if (!rtnl_trylock())
6329                 return restart_syscall();
6330
6331         new_val = *((u32 *)ctl->data);
6332
6333         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6334         if (ret != 0)
6335                 goto out;
6336
6337         if (write) {
6338                 if (check_addr_gen_mode(new_val) < 0) {
6339                         ret = -EINVAL;
6340                         goto out;
6341                 }
6342
6343                 if (idev) {
6344                         if (check_stable_privacy(idev, net, new_val) < 0) {
6345                                 ret = -EINVAL;
6346                                 goto out;
6347                         }
6348
6349                         if (idev->cnf.addr_gen_mode != new_val) {
6350                                 idev->cnf.addr_gen_mode = new_val;
6351                                 addrconf_dev_config(idev->dev);
6352                         }
6353                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6354                         struct net_device *dev;
6355
6356                         net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6357                         for_each_netdev(net, dev) {
6358                                 idev = __in6_dev_get(dev);
6359                                 if (idev &&
6360                                     idev->cnf.addr_gen_mode != new_val) {
6361                                         idev->cnf.addr_gen_mode = new_val;
6362                                         addrconf_dev_config(idev->dev);
6363                                 }
6364                         }
6365                 }
6366
6367                 *((u32 *)ctl->data) = new_val;
6368         }
6369
6370 out:
6371         rtnl_unlock();
6372
6373         return ret;
6374 }
6375
6376 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6377                                          void *buffer, size_t *lenp,
6378                                          loff_t *ppos)
6379 {
6380         int err;
6381         struct in6_addr addr;
6382         char str[IPV6_MAX_STRLEN];
6383         struct ctl_table lctl = *ctl;
6384         struct net *net = ctl->extra2;
6385         struct ipv6_stable_secret *secret = ctl->data;
6386
6387         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6388                 return -EIO;
6389
6390         lctl.maxlen = IPV6_MAX_STRLEN;
6391         lctl.data = str;
6392
6393         if (!rtnl_trylock())
6394                 return restart_syscall();
6395
6396         if (!write && !secret->initialized) {
6397                 err = -EIO;
6398                 goto out;
6399         }
6400
6401         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6402         if (err >= sizeof(str)) {
6403                 err = -EIO;
6404                 goto out;
6405         }
6406
6407         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6408         if (err || !write)
6409                 goto out;
6410
6411         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6412                 err = -EIO;
6413                 goto out;
6414         }
6415
6416         secret->initialized = true;
6417         secret->secret = addr;
6418
6419         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6420                 struct net_device *dev;
6421
6422                 for_each_netdev(net, dev) {
6423                         struct inet6_dev *idev = __in6_dev_get(dev);
6424
6425                         if (idev) {
6426                                 idev->cnf.addr_gen_mode =
6427                                         IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6428                         }
6429                 }
6430         } else {
6431                 struct inet6_dev *idev = ctl->extra1;
6432
6433                 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6434         }
6435
6436 out:
6437         rtnl_unlock();
6438
6439         return err;
6440 }
6441
6442 static
6443 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6444                                                 int write, void *buffer,
6445                                                 size_t *lenp,
6446                                                 loff_t *ppos)
6447 {
6448         int *valp = ctl->data;
6449         int val = *valp;
6450         loff_t pos = *ppos;
6451         struct ctl_table lctl;
6452         int ret;
6453
6454         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6455          * we should not modify it until we get the rtnl lock.
6456          */
6457         lctl = *ctl;
6458         lctl.data = &val;
6459
6460         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6461
6462         if (write)
6463                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6464         if (ret)
6465                 *ppos = pos;
6466         return ret;
6467 }
6468
6469 static
6470 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6471 {
6472         if (rt) {
6473                 if (action)
6474                         rt->dst.flags |= DST_NOPOLICY;
6475                 else
6476                         rt->dst.flags &= ~DST_NOPOLICY;
6477         }
6478 }
6479
6480 static
6481 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6482 {
6483         struct inet6_ifaddr *ifa;
6484
6485         read_lock_bh(&idev->lock);
6486         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6487                 spin_lock(&ifa->lock);
6488                 if (ifa->rt) {
6489                         /* host routes only use builtin fib6_nh */
6490                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6491                         int cpu;
6492
6493                         rcu_read_lock();
6494                         ifa->rt->dst_nopolicy = val ? true : false;
6495                         if (nh->rt6i_pcpu) {
6496                                 for_each_possible_cpu(cpu) {
6497                                         struct rt6_info **rtp;
6498
6499                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6500                                         addrconf_set_nopolicy(*rtp, val);
6501                                 }
6502                         }
6503                         rcu_read_unlock();
6504                 }
6505                 spin_unlock(&ifa->lock);
6506         }
6507         read_unlock_bh(&idev->lock);
6508 }
6509
6510 static
6511 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6512 {
6513         struct inet6_dev *idev;
6514         struct net *net;
6515
6516         if (!rtnl_trylock())
6517                 return restart_syscall();
6518
6519         *valp = val;
6520
6521         net = (struct net *)ctl->extra2;
6522         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6523                 rtnl_unlock();
6524                 return 0;
6525         }
6526
6527         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6528                 struct net_device *dev;
6529
6530                 for_each_netdev(net, dev) {
6531                         idev = __in6_dev_get(dev);
6532                         if (idev)
6533                                 addrconf_disable_policy_idev(idev, val);
6534                 }
6535         } else {
6536                 idev = (struct inet6_dev *)ctl->extra1;
6537                 addrconf_disable_policy_idev(idev, val);
6538         }
6539
6540         rtnl_unlock();
6541         return 0;
6542 }
6543
6544 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6545                                    void *buffer, size_t *lenp, loff_t *ppos)
6546 {
6547         int *valp = ctl->data;
6548         int val = *valp;
6549         loff_t pos = *ppos;
6550         struct ctl_table lctl;
6551         int ret;
6552
6553         lctl = *ctl;
6554         lctl.data = &val;
6555         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6556
6557         if (write && (*valp != val))
6558                 ret = addrconf_disable_policy(ctl, valp, val);
6559
6560         if (ret)
6561                 *ppos = pos;
6562
6563         return ret;
6564 }
6565
6566 static int minus_one = -1;
6567 static const int two_five_five = 255;
6568 static u32 ioam6_if_id_max = U16_MAX;
6569
6570 static const struct ctl_table addrconf_sysctl[] = {
6571         {
6572                 .procname       = "forwarding",
6573                 .data           = &ipv6_devconf.forwarding,
6574                 .maxlen         = sizeof(int),
6575                 .mode           = 0644,
6576                 .proc_handler   = addrconf_sysctl_forward,
6577         },
6578         {
6579                 .procname       = "hop_limit",
6580                 .data           = &ipv6_devconf.hop_limit,
6581                 .maxlen         = sizeof(int),
6582                 .mode           = 0644,
6583                 .proc_handler   = proc_dointvec_minmax,
6584                 .extra1         = (void *)SYSCTL_ONE,
6585                 .extra2         = (void *)&two_five_five,
6586         },
6587         {
6588                 .procname       = "mtu",
6589                 .data           = &ipv6_devconf.mtu6,
6590                 .maxlen         = sizeof(int),
6591                 .mode           = 0644,
6592                 .proc_handler   = addrconf_sysctl_mtu,
6593         },
6594         {
6595                 .procname       = "accept_ra",
6596                 .data           = &ipv6_devconf.accept_ra,
6597                 .maxlen         = sizeof(int),
6598                 .mode           = 0644,
6599                 .proc_handler   = proc_dointvec,
6600         },
6601         {
6602                 .procname       = "accept_redirects",
6603                 .data           = &ipv6_devconf.accept_redirects,
6604                 .maxlen         = sizeof(int),
6605                 .mode           = 0644,
6606                 .proc_handler   = proc_dointvec,
6607         },
6608         {
6609                 .procname       = "autoconf",
6610                 .data           = &ipv6_devconf.autoconf,
6611                 .maxlen         = sizeof(int),
6612                 .mode           = 0644,
6613                 .proc_handler   = proc_dointvec,
6614         },
6615         {
6616                 .procname       = "dad_transmits",
6617                 .data           = &ipv6_devconf.dad_transmits,
6618                 .maxlen         = sizeof(int),
6619                 .mode           = 0644,
6620                 .proc_handler   = proc_dointvec,
6621         },
6622         {
6623                 .procname       = "router_solicitations",
6624                 .data           = &ipv6_devconf.rtr_solicits,
6625                 .maxlen         = sizeof(int),
6626                 .mode           = 0644,
6627                 .proc_handler   = proc_dointvec_minmax,
6628                 .extra1         = &minus_one,
6629         },
6630         {
6631                 .procname       = "router_solicitation_interval",
6632                 .data           = &ipv6_devconf.rtr_solicit_interval,
6633                 .maxlen         = sizeof(int),
6634                 .mode           = 0644,
6635                 .proc_handler   = proc_dointvec_jiffies,
6636         },
6637         {
6638                 .procname       = "router_solicitation_max_interval",
6639                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6640                 .maxlen         = sizeof(int),
6641                 .mode           = 0644,
6642                 .proc_handler   = proc_dointvec_jiffies,
6643         },
6644         {
6645                 .procname       = "router_solicitation_delay",
6646                 .data           = &ipv6_devconf.rtr_solicit_delay,
6647                 .maxlen         = sizeof(int),
6648                 .mode           = 0644,
6649                 .proc_handler   = proc_dointvec_jiffies,
6650         },
6651         {
6652                 .procname       = "force_mld_version",
6653                 .data           = &ipv6_devconf.force_mld_version,
6654                 .maxlen         = sizeof(int),
6655                 .mode           = 0644,
6656                 .proc_handler   = proc_dointvec,
6657         },
6658         {
6659                 .procname       = "mldv1_unsolicited_report_interval",
6660                 .data           =
6661                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6662                 .maxlen         = sizeof(int),
6663                 .mode           = 0644,
6664                 .proc_handler   = proc_dointvec_ms_jiffies,
6665         },
6666         {
6667                 .procname       = "mldv2_unsolicited_report_interval",
6668                 .data           =
6669                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6670                 .maxlen         = sizeof(int),
6671                 .mode           = 0644,
6672                 .proc_handler   = proc_dointvec_ms_jiffies,
6673         },
6674         {
6675                 .procname       = "use_tempaddr",
6676                 .data           = &ipv6_devconf.use_tempaddr,
6677                 .maxlen         = sizeof(int),
6678                 .mode           = 0644,
6679                 .proc_handler   = proc_dointvec,
6680         },
6681         {
6682                 .procname       = "temp_valid_lft",
6683                 .data           = &ipv6_devconf.temp_valid_lft,
6684                 .maxlen         = sizeof(int),
6685                 .mode           = 0644,
6686                 .proc_handler   = proc_dointvec,
6687         },
6688         {
6689                 .procname       = "temp_prefered_lft",
6690                 .data           = &ipv6_devconf.temp_prefered_lft,
6691                 .maxlen         = sizeof(int),
6692                 .mode           = 0644,
6693                 .proc_handler   = proc_dointvec,
6694         },
6695         {
6696                 .procname       = "regen_max_retry",
6697                 .data           = &ipv6_devconf.regen_max_retry,
6698                 .maxlen         = sizeof(int),
6699                 .mode           = 0644,
6700                 .proc_handler   = proc_dointvec,
6701         },
6702         {
6703                 .procname       = "max_desync_factor",
6704                 .data           = &ipv6_devconf.max_desync_factor,
6705                 .maxlen         = sizeof(int),
6706                 .mode           = 0644,
6707                 .proc_handler   = proc_dointvec,
6708         },
6709         {
6710                 .procname       = "max_addresses",
6711                 .data           = &ipv6_devconf.max_addresses,
6712                 .maxlen         = sizeof(int),
6713                 .mode           = 0644,
6714                 .proc_handler   = proc_dointvec,
6715         },
6716         {
6717                 .procname       = "accept_ra_defrtr",
6718                 .data           = &ipv6_devconf.accept_ra_defrtr,
6719                 .maxlen         = sizeof(int),
6720                 .mode           = 0644,
6721                 .proc_handler   = proc_dointvec,
6722         },
6723         {
6724                 .procname       = "ra_defrtr_metric",
6725                 .data           = &ipv6_devconf.ra_defrtr_metric,
6726                 .maxlen         = sizeof(u32),
6727                 .mode           = 0644,
6728                 .proc_handler   = proc_douintvec_minmax,
6729                 .extra1         = (void *)SYSCTL_ONE,
6730         },
6731         {
6732                 .procname       = "accept_ra_min_hop_limit",
6733                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6734                 .maxlen         = sizeof(int),
6735                 .mode           = 0644,
6736                 .proc_handler   = proc_dointvec,
6737         },
6738         {
6739                 .procname       = "accept_ra_pinfo",
6740                 .data           = &ipv6_devconf.accept_ra_pinfo,
6741                 .maxlen         = sizeof(int),
6742                 .mode           = 0644,
6743                 .proc_handler   = proc_dointvec,
6744         },
6745 #ifdef CONFIG_IPV6_ROUTER_PREF
6746         {
6747                 .procname       = "accept_ra_rtr_pref",
6748                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6749                 .maxlen         = sizeof(int),
6750                 .mode           = 0644,
6751                 .proc_handler   = proc_dointvec,
6752         },
6753         {
6754                 .procname       = "router_probe_interval",
6755                 .data           = &ipv6_devconf.rtr_probe_interval,
6756                 .maxlen         = sizeof(int),
6757                 .mode           = 0644,
6758                 .proc_handler   = proc_dointvec_jiffies,
6759         },
6760 #ifdef CONFIG_IPV6_ROUTE_INFO
6761         {
6762                 .procname       = "accept_ra_rt_info_min_plen",
6763                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6764                 .maxlen         = sizeof(int),
6765                 .mode           = 0644,
6766                 .proc_handler   = proc_dointvec,
6767         },
6768         {
6769                 .procname       = "accept_ra_rt_info_max_plen",
6770                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6771                 .maxlen         = sizeof(int),
6772                 .mode           = 0644,
6773                 .proc_handler   = proc_dointvec,
6774         },
6775 #endif
6776 #endif
6777         {
6778                 .procname       = "proxy_ndp",
6779                 .data           = &ipv6_devconf.proxy_ndp,
6780                 .maxlen         = sizeof(int),
6781                 .mode           = 0644,
6782                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6783         },
6784         {
6785                 .procname       = "accept_source_route",
6786                 .data           = &ipv6_devconf.accept_source_route,
6787                 .maxlen         = sizeof(int),
6788                 .mode           = 0644,
6789                 .proc_handler   = proc_dointvec,
6790         },
6791 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6792         {
6793                 .procname       = "optimistic_dad",
6794                 .data           = &ipv6_devconf.optimistic_dad,
6795                 .maxlen         = sizeof(int),
6796                 .mode           = 0644,
6797                 .proc_handler   = proc_dointvec,
6798         },
6799         {
6800                 .procname       = "use_optimistic",
6801                 .data           = &ipv6_devconf.use_optimistic,
6802                 .maxlen         = sizeof(int),
6803                 .mode           = 0644,
6804                 .proc_handler   = proc_dointvec,
6805         },
6806 #endif
6807 #ifdef CONFIG_IPV6_MROUTE
6808         {
6809                 .procname       = "mc_forwarding",
6810                 .data           = &ipv6_devconf.mc_forwarding,
6811                 .maxlen         = sizeof(int),
6812                 .mode           = 0444,
6813                 .proc_handler   = proc_dointvec,
6814         },
6815 #endif
6816         {
6817                 .procname       = "disable_ipv6",
6818                 .data           = &ipv6_devconf.disable_ipv6,
6819                 .maxlen         = sizeof(int),
6820                 .mode           = 0644,
6821                 .proc_handler   = addrconf_sysctl_disable,
6822         },
6823         {
6824                 .procname       = "accept_dad",
6825                 .data           = &ipv6_devconf.accept_dad,
6826                 .maxlen         = sizeof(int),
6827                 .mode           = 0644,
6828                 .proc_handler   = proc_dointvec,
6829         },
6830         {
6831                 .procname       = "force_tllao",
6832                 .data           = &ipv6_devconf.force_tllao,
6833                 .maxlen         = sizeof(int),
6834                 .mode           = 0644,
6835                 .proc_handler   = proc_dointvec
6836         },
6837         {
6838                 .procname       = "ndisc_notify",
6839                 .data           = &ipv6_devconf.ndisc_notify,
6840                 .maxlen         = sizeof(int),
6841                 .mode           = 0644,
6842                 .proc_handler   = proc_dointvec
6843         },
6844         {
6845                 .procname       = "suppress_frag_ndisc",
6846                 .data           = &ipv6_devconf.suppress_frag_ndisc,
6847                 .maxlen         = sizeof(int),
6848                 .mode           = 0644,
6849                 .proc_handler   = proc_dointvec
6850         },
6851         {
6852                 .procname       = "accept_ra_from_local",
6853                 .data           = &ipv6_devconf.accept_ra_from_local,
6854                 .maxlen         = sizeof(int),
6855                 .mode           = 0644,
6856                 .proc_handler   = proc_dointvec,
6857         },
6858         {
6859                 .procname       = "accept_ra_mtu",
6860                 .data           = &ipv6_devconf.accept_ra_mtu,
6861                 .maxlen         = sizeof(int),
6862                 .mode           = 0644,
6863                 .proc_handler   = proc_dointvec,
6864         },
6865         {
6866                 .procname       = "stable_secret",
6867                 .data           = &ipv6_devconf.stable_secret,
6868                 .maxlen         = IPV6_MAX_STRLEN,
6869                 .mode           = 0600,
6870                 .proc_handler   = addrconf_sysctl_stable_secret,
6871         },
6872         {
6873                 .procname       = "use_oif_addrs_only",
6874                 .data           = &ipv6_devconf.use_oif_addrs_only,
6875                 .maxlen         = sizeof(int),
6876                 .mode           = 0644,
6877                 .proc_handler   = proc_dointvec,
6878         },
6879         {
6880                 .procname       = "ignore_routes_with_linkdown",
6881                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
6882                 .maxlen         = sizeof(int),
6883                 .mode           = 0644,
6884                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
6885         },
6886         {
6887                 .procname       = "drop_unicast_in_l2_multicast",
6888                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
6889                 .maxlen         = sizeof(int),
6890                 .mode           = 0644,
6891                 .proc_handler   = proc_dointvec,
6892         },
6893         {
6894                 .procname       = "drop_unsolicited_na",
6895                 .data           = &ipv6_devconf.drop_unsolicited_na,
6896                 .maxlen         = sizeof(int),
6897                 .mode           = 0644,
6898                 .proc_handler   = proc_dointvec,
6899         },
6900         {
6901                 .procname       = "keep_addr_on_down",
6902                 .data           = &ipv6_devconf.keep_addr_on_down,
6903                 .maxlen         = sizeof(int),
6904                 .mode           = 0644,
6905                 .proc_handler   = proc_dointvec,
6906
6907         },
6908         {
6909                 .procname       = "seg6_enabled",
6910                 .data           = &ipv6_devconf.seg6_enabled,
6911                 .maxlen         = sizeof(int),
6912                 .mode           = 0644,
6913                 .proc_handler   = proc_dointvec,
6914         },
6915 #ifdef CONFIG_IPV6_SEG6_HMAC
6916         {
6917                 .procname       = "seg6_require_hmac",
6918                 .data           = &ipv6_devconf.seg6_require_hmac,
6919                 .maxlen         = sizeof(int),
6920                 .mode           = 0644,
6921                 .proc_handler   = proc_dointvec,
6922         },
6923 #endif
6924         {
6925                 .procname       = "enhanced_dad",
6926                 .data           = &ipv6_devconf.enhanced_dad,
6927                 .maxlen         = sizeof(int),
6928                 .mode           = 0644,
6929                 .proc_handler   = proc_dointvec,
6930         },
6931         {
6932                 .procname       = "addr_gen_mode",
6933                 .data           = &ipv6_devconf.addr_gen_mode,
6934                 .maxlen         = sizeof(int),
6935                 .mode           = 0644,
6936                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
6937         },
6938         {
6939                 .procname       = "disable_policy",
6940                 .data           = &ipv6_devconf.disable_policy,
6941                 .maxlen         = sizeof(int),
6942                 .mode           = 0644,
6943                 .proc_handler   = addrconf_sysctl_disable_policy,
6944         },
6945         {
6946                 .procname       = "ndisc_tclass",
6947                 .data           = &ipv6_devconf.ndisc_tclass,
6948                 .maxlen         = sizeof(int),
6949                 .mode           = 0644,
6950                 .proc_handler   = proc_dointvec_minmax,
6951                 .extra1         = (void *)SYSCTL_ZERO,
6952                 .extra2         = (void *)&two_five_five,
6953         },
6954         {
6955                 .procname       = "rpl_seg_enabled",
6956                 .data           = &ipv6_devconf.rpl_seg_enabled,
6957                 .maxlen         = sizeof(int),
6958                 .mode           = 0644,
6959                 .proc_handler   = proc_dointvec,
6960         },
6961         {
6962                 .procname       = "ioam6_enabled",
6963                 .data           = &ipv6_devconf.ioam6_enabled,
6964                 .maxlen         = sizeof(u8),
6965                 .mode           = 0644,
6966                 .proc_handler   = proc_dou8vec_minmax,
6967                 .extra1         = (void *)SYSCTL_ZERO,
6968                 .extra2         = (void *)SYSCTL_ONE,
6969         },
6970         {
6971                 .procname       = "ioam6_id",
6972                 .data           = &ipv6_devconf.ioam6_id,
6973                 .maxlen         = sizeof(u32),
6974                 .mode           = 0644,
6975                 .proc_handler   = proc_douintvec_minmax,
6976                 .extra1         = (void *)SYSCTL_ZERO,
6977                 .extra2         = (void *)&ioam6_if_id_max,
6978         },
6979         {
6980                 .procname       = "ioam6_id_wide",
6981                 .data           = &ipv6_devconf.ioam6_id_wide,
6982                 .maxlen         = sizeof(u32),
6983                 .mode           = 0644,
6984                 .proc_handler   = proc_douintvec,
6985         },
6986         {
6987                 /* sentinel */
6988         }
6989 };
6990
6991 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6992                 struct inet6_dev *idev, struct ipv6_devconf *p)
6993 {
6994         int i, ifindex;
6995         struct ctl_table *table;
6996         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6997
6998         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6999         if (!table)
7000                 goto out;
7001
7002         for (i = 0; table[i].data; i++) {
7003                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7004                 /* If one of these is already set, then it is not safe to
7005                  * overwrite either of them: this makes proc_dointvec_minmax
7006                  * usable.
7007                  */
7008                 if (!table[i].extra1 && !table[i].extra2) {
7009                         table[i].extra1 = idev; /* embedded; no ref */
7010                         table[i].extra2 = net;
7011                 }
7012         }
7013
7014         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7015
7016         p->sysctl_header = register_net_sysctl(net, path, table);
7017         if (!p->sysctl_header)
7018                 goto free;
7019
7020         if (!strcmp(dev_name, "all"))
7021                 ifindex = NETCONFA_IFINDEX_ALL;
7022         else if (!strcmp(dev_name, "default"))
7023                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7024         else
7025                 ifindex = idev->dev->ifindex;
7026         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7027                                      ifindex, p);
7028         return 0;
7029
7030 free:
7031         kfree(table);
7032 out:
7033         return -ENOBUFS;
7034 }
7035
7036 static void __addrconf_sysctl_unregister(struct net *net,
7037                                          struct ipv6_devconf *p, int ifindex)
7038 {
7039         struct ctl_table *table;
7040
7041         if (!p->sysctl_header)
7042                 return;
7043
7044         table = p->sysctl_header->ctl_table_arg;
7045         unregister_net_sysctl_table(p->sysctl_header);
7046         p->sysctl_header = NULL;
7047         kfree(table);
7048
7049         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7050 }
7051
7052 static int addrconf_sysctl_register(struct inet6_dev *idev)
7053 {
7054         int err;
7055
7056         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7057                 return -EINVAL;
7058
7059         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7060                                     &ndisc_ifinfo_sysctl_change);
7061         if (err)
7062                 return err;
7063         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7064                                          idev, &idev->cnf);
7065         if (err)
7066                 neigh_sysctl_unregister(idev->nd_parms);
7067
7068         return err;
7069 }
7070
7071 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7072 {
7073         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7074                                      idev->dev->ifindex);
7075         neigh_sysctl_unregister(idev->nd_parms);
7076 }
7077
7078
7079 #endif
7080
7081 static int __net_init addrconf_init_net(struct net *net)
7082 {
7083         int err = -ENOMEM;
7084         struct ipv6_devconf *all, *dflt;
7085
7086         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7087         if (!all)
7088                 goto err_alloc_all;
7089
7090         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7091         if (!dflt)
7092                 goto err_alloc_dflt;
7093
7094         if (IS_ENABLED(CONFIG_SYSCTL) &&
7095             !net_eq(net, &init_net)) {
7096                 switch (sysctl_devconf_inherit_init_net) {
7097                 case 1:  /* copy from init_net */
7098                         memcpy(all, init_net.ipv6.devconf_all,
7099                                sizeof(ipv6_devconf));
7100                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7101                                sizeof(ipv6_devconf_dflt));
7102                         break;
7103                 case 3: /* copy from the current netns */
7104                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7105                                sizeof(ipv6_devconf));
7106                         memcpy(dflt,
7107                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7108                                sizeof(ipv6_devconf_dflt));
7109                         break;
7110                 case 0:
7111                 case 2:
7112                         /* use compiled values */
7113                         break;
7114                 }
7115         }
7116
7117         /* these will be inherited by all namespaces */
7118         dflt->autoconf = ipv6_defaults.autoconf;
7119         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7120
7121         dflt->stable_secret.initialized = false;
7122         all->stable_secret.initialized = false;
7123
7124         net->ipv6.devconf_all = all;
7125         net->ipv6.devconf_dflt = dflt;
7126
7127 #ifdef CONFIG_SYSCTL
7128         err = __addrconf_sysctl_register(net, "all", NULL, all);
7129         if (err < 0)
7130                 goto err_reg_all;
7131
7132         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7133         if (err < 0)
7134                 goto err_reg_dflt;
7135 #endif
7136         return 0;
7137
7138 #ifdef CONFIG_SYSCTL
7139 err_reg_dflt:
7140         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7141 err_reg_all:
7142         kfree(dflt);
7143 #endif
7144 err_alloc_dflt:
7145         kfree(all);
7146 err_alloc_all:
7147         return err;
7148 }
7149
7150 static void __net_exit addrconf_exit_net(struct net *net)
7151 {
7152 #ifdef CONFIG_SYSCTL
7153         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7154                                      NETCONFA_IFINDEX_DEFAULT);
7155         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7156                                      NETCONFA_IFINDEX_ALL);
7157 #endif
7158         kfree(net->ipv6.devconf_dflt);
7159         kfree(net->ipv6.devconf_all);
7160 }
7161
7162 static struct pernet_operations addrconf_ops = {
7163         .init = addrconf_init_net,
7164         .exit = addrconf_exit_net,
7165 };
7166
7167 static struct rtnl_af_ops inet6_ops __read_mostly = {
7168         .family           = AF_INET6,
7169         .fill_link_af     = inet6_fill_link_af,
7170         .get_link_af_size = inet6_get_link_af_size,
7171         .validate_link_af = inet6_validate_link_af,
7172         .set_link_af      = inet6_set_link_af,
7173 };
7174
7175 /*
7176  *      Init / cleanup code
7177  */
7178
7179 int __init addrconf_init(void)
7180 {
7181         struct inet6_dev *idev;
7182         int i, err;
7183
7184         err = ipv6_addr_label_init();
7185         if (err < 0) {
7186                 pr_crit("%s: cannot initialize default policy table: %d\n",
7187                         __func__, err);
7188                 goto out;
7189         }
7190
7191         err = register_pernet_subsys(&addrconf_ops);
7192         if (err < 0)
7193                 goto out_addrlabel;
7194
7195         addrconf_wq = create_workqueue("ipv6_addrconf");
7196         if (!addrconf_wq) {
7197                 err = -ENOMEM;
7198                 goto out_nowq;
7199         }
7200
7201         /* The addrconf netdev notifier requires that loopback_dev
7202          * has it's ipv6 private information allocated and setup
7203          * before it can bring up and give link-local addresses
7204          * to other devices which are up.
7205          *
7206          * Unfortunately, loopback_dev is not necessarily the first
7207          * entry in the global dev_base list of net devices.  In fact,
7208          * it is likely to be the very last entry on that list.
7209          * So this causes the notifier registry below to try and
7210          * give link-local addresses to all devices besides loopback_dev
7211          * first, then loopback_dev, which cases all the non-loopback_dev
7212          * devices to fail to get a link-local address.
7213          *
7214          * So, as a temporary fix, allocate the ipv6 structure for
7215          * loopback_dev first by hand.
7216          * Longer term, all of the dependencies ipv6 has upon the loopback
7217          * device and it being up should be removed.
7218          */
7219         rtnl_lock();
7220         idev = ipv6_add_dev(init_net.loopback_dev);
7221         rtnl_unlock();
7222         if (IS_ERR(idev)) {
7223                 err = PTR_ERR(idev);
7224                 goto errlo;
7225         }
7226
7227         ip6_route_init_special_entries();
7228
7229         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7230                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7231
7232         register_netdevice_notifier(&ipv6_dev_notf);
7233
7234         addrconf_verify();
7235
7236         rtnl_af_register(&inet6_ops);
7237
7238         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7239                                    NULL, inet6_dump_ifinfo, 0);
7240         if (err < 0)
7241                 goto errout;
7242
7243         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7244                                    inet6_rtm_newaddr, NULL, 0);
7245         if (err < 0)
7246                 goto errout;
7247         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7248                                    inet6_rtm_deladdr, NULL, 0);
7249         if (err < 0)
7250                 goto errout;
7251         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7252                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7253                                    RTNL_FLAG_DOIT_UNLOCKED);
7254         if (err < 0)
7255                 goto errout;
7256         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7257                                    NULL, inet6_dump_ifmcaddr, 0);
7258         if (err < 0)
7259                 goto errout;
7260         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7261                                    NULL, inet6_dump_ifacaddr, 0);
7262         if (err < 0)
7263                 goto errout;
7264         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7265                                    inet6_netconf_get_devconf,
7266                                    inet6_netconf_dump_devconf,
7267                                    RTNL_FLAG_DOIT_UNLOCKED);
7268         if (err < 0)
7269                 goto errout;
7270         err = ipv6_addr_label_rtnl_register();
7271         if (err < 0)
7272                 goto errout;
7273
7274         return 0;
7275 errout:
7276         rtnl_unregister_all(PF_INET6);
7277         rtnl_af_unregister(&inet6_ops);
7278         unregister_netdevice_notifier(&ipv6_dev_notf);
7279 errlo:
7280         destroy_workqueue(addrconf_wq);
7281 out_nowq:
7282         unregister_pernet_subsys(&addrconf_ops);
7283 out_addrlabel:
7284         ipv6_addr_label_cleanup();
7285 out:
7286         return err;
7287 }
7288
7289 void addrconf_cleanup(void)
7290 {
7291         struct net_device *dev;
7292         int i;
7293
7294         unregister_netdevice_notifier(&ipv6_dev_notf);
7295         unregister_pernet_subsys(&addrconf_ops);
7296         ipv6_addr_label_cleanup();
7297
7298         rtnl_af_unregister(&inet6_ops);
7299
7300         rtnl_lock();
7301
7302         /* clean dev list */
7303         for_each_netdev(&init_net, dev) {
7304                 if (__in6_dev_get(dev) == NULL)
7305                         continue;
7306                 addrconf_ifdown(dev, true);
7307         }
7308         addrconf_ifdown(init_net.loopback_dev, true);
7309
7310         /*
7311          *      Check hash table.
7312          */
7313         spin_lock_bh(&addrconf_hash_lock);
7314         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7315                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7316         spin_unlock_bh(&addrconf_hash_lock);
7317         cancel_delayed_work(&addr_chk_work);
7318         rtnl_unlock();
7319
7320         destroy_workqueue(addrconf_wq);
7321 }