]> Git Repo - linux.git/blob - net/ipv6/addrconf.c
x86/config: Fix warning for 'make ARCH=x86_64 tinyconfig'
[linux.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             <[email protected]>
8  *      Alexey Kuznetsov        <[email protected]>
9  */
10
11 /*
12  *      Changes:
13  *
14  *      Janos Farkas                    :       delete timer on ifdown
15  *      <[email protected]>
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 = get_random_u32_inclusive(900000, 1100000) * (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 = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116         do_div(tmp, 1000000);
117         if ((s32)tmp > mrt) {
118                 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119                 tmp = get_random_u32_inclusive(900000, 1100000) * (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 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
152
153 static struct workqueue_struct *addrconf_wq;
154
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
157
158 static void addrconf_type_change(struct net_device *dev,
159                                  unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
161
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
163                                                   int plen,
164                                                   const struct net_device *dev,
165                                                   u32 flags, u32 noflags,
166                                                   bool no_gw);
167
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
171                                    bool send_na);
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
176
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178                                 struct prefix_info *pinfo);
179
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
181         .forwarding             = 0,
182         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
183         .mtu6                   = IPV6_MIN_MTU,
184         .accept_ra              = 1,
185         .accept_redirects       = 1,
186         .autoconf               = 1,
187         .force_mld_version      = 0,
188         .mldv1_unsolicited_report_interval = 10 * HZ,
189         .mldv2_unsolicited_report_interval = HZ,
190         .dad_transmits          = 1,
191         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
192         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
193         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
195         .use_tempaddr           = 0,
196         .temp_valid_lft         = TEMP_VALID_LIFETIME,
197         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
198         .regen_min_advance      = REGEN_MIN_ADVANCE,
199         .regen_max_retry        = REGEN_MAX_RETRY,
200         .max_desync_factor      = MAX_DESYNC_FACTOR,
201         .max_addresses          = IPV6_MAX_ADDRESSES,
202         .accept_ra_defrtr       = 1,
203         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
204         .accept_ra_from_local   = 0,
205         .accept_ra_min_hop_limit= 1,
206         .accept_ra_min_lft      = 0,
207         .accept_ra_pinfo        = 1,
208 #ifdef CONFIG_IPV6_ROUTER_PREF
209         .accept_ra_rtr_pref     = 1,
210         .rtr_probe_interval     = 60 * HZ,
211 #ifdef CONFIG_IPV6_ROUTE_INFO
212         .accept_ra_rt_info_min_plen = 0,
213         .accept_ra_rt_info_max_plen = 0,
214 #endif
215 #endif
216         .proxy_ndp              = 0,
217         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
218         .disable_ipv6           = 0,
219         .accept_dad             = 0,
220         .suppress_frag_ndisc    = 1,
221         .accept_ra_mtu          = 1,
222         .stable_secret          = {
223                 .initialized = false,
224         },
225         .use_oif_addrs_only     = 0,
226         .ignore_routes_with_linkdown = 0,
227         .keep_addr_on_down      = 0,
228         .seg6_enabled           = 0,
229 #ifdef CONFIG_IPV6_SEG6_HMAC
230         .seg6_require_hmac      = 0,
231 #endif
232         .enhanced_dad           = 1,
233         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
234         .disable_policy         = 0,
235         .rpl_seg_enabled        = 0,
236         .ioam6_enabled          = 0,
237         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
238         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
239         .ndisc_evict_nocarrier  = 1,
240         .ra_honor_pio_life      = 0,
241 };
242
243 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
244         .forwarding             = 0,
245         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
246         .mtu6                   = IPV6_MIN_MTU,
247         .accept_ra              = 1,
248         .accept_redirects       = 1,
249         .autoconf               = 1,
250         .force_mld_version      = 0,
251         .mldv1_unsolicited_report_interval = 10 * HZ,
252         .mldv2_unsolicited_report_interval = HZ,
253         .dad_transmits          = 1,
254         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
255         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
256         .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
257         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
258         .use_tempaddr           = 0,
259         .temp_valid_lft         = TEMP_VALID_LIFETIME,
260         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
261         .regen_min_advance      = REGEN_MIN_ADVANCE,
262         .regen_max_retry        = REGEN_MAX_RETRY,
263         .max_desync_factor      = MAX_DESYNC_FACTOR,
264         .max_addresses          = IPV6_MAX_ADDRESSES,
265         .accept_ra_defrtr       = 1,
266         .ra_defrtr_metric       = IP6_RT_PRIO_USER,
267         .accept_ra_from_local   = 0,
268         .accept_ra_min_hop_limit= 1,
269         .accept_ra_min_lft      = 0,
270         .accept_ra_pinfo        = 1,
271 #ifdef CONFIG_IPV6_ROUTER_PREF
272         .accept_ra_rtr_pref     = 1,
273         .rtr_probe_interval     = 60 * HZ,
274 #ifdef CONFIG_IPV6_ROUTE_INFO
275         .accept_ra_rt_info_min_plen = 0,
276         .accept_ra_rt_info_max_plen = 0,
277 #endif
278 #endif
279         .proxy_ndp              = 0,
280         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
281         .disable_ipv6           = 0,
282         .accept_dad             = 1,
283         .suppress_frag_ndisc    = 1,
284         .accept_ra_mtu          = 1,
285         .stable_secret          = {
286                 .initialized = false,
287         },
288         .use_oif_addrs_only     = 0,
289         .ignore_routes_with_linkdown = 0,
290         .keep_addr_on_down      = 0,
291         .seg6_enabled           = 0,
292 #ifdef CONFIG_IPV6_SEG6_HMAC
293         .seg6_require_hmac      = 0,
294 #endif
295         .enhanced_dad           = 1,
296         .addr_gen_mode          = IN6_ADDR_GEN_MODE_EUI64,
297         .disable_policy         = 0,
298         .rpl_seg_enabled        = 0,
299         .ioam6_enabled          = 0,
300         .ioam6_id               = IOAM6_DEFAULT_IF_ID,
301         .ioam6_id_wide          = IOAM6_DEFAULT_IF_ID_WIDE,
302         .ndisc_evict_nocarrier  = 1,
303         .ra_honor_pio_life      = 0,
304 };
305
306 /* Check if link is ready: is it up and is a valid qdisc available */
307 static inline bool addrconf_link_ready(const struct net_device *dev)
308 {
309         return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
310 }
311
312 static void addrconf_del_rs_timer(struct inet6_dev *idev)
313 {
314         if (del_timer(&idev->rs_timer))
315                 __in6_dev_put(idev);
316 }
317
318 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
319 {
320         if (cancel_delayed_work(&ifp->dad_work))
321                 __in6_ifa_put(ifp);
322 }
323
324 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
325                                   unsigned long when)
326 {
327         if (!mod_timer(&idev->rs_timer, jiffies + when))
328                 in6_dev_hold(idev);
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_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
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_ACCOUNT);
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 && dev != blackhole_netdev)
381                 return ERR_PTR(-EINVAL);
382
383         ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
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         netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
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                 netdev_put(dev, &ndev->dev_tracker);
413                 kfree(ndev);
414                 return ERR_PTR(err);
415         }
416
417         if (dev != blackhole_netdev) {
418                 if (snmp6_register_dev(ndev) < 0) {
419                         netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
420                                    __func__, dev->name);
421                         goto err_release;
422                 }
423         }
424         /* One reference from device. */
425         refcount_set(&ndev->refcnt, 1);
426
427         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
428                 ndev->cnf.accept_dad = -1;
429
430 #if IS_ENABLED(CONFIG_IPV6_SIT)
431         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
432                 pr_info("%s: Disabled Multicast RS\n", dev->name);
433                 ndev->cnf.rtr_solicits = 0;
434         }
435 #endif
436
437         INIT_LIST_HEAD(&ndev->tempaddr_list);
438         ndev->desync_factor = U32_MAX;
439         if ((dev->flags&IFF_LOOPBACK) ||
440             dev->type == ARPHRD_TUNNEL ||
441             dev->type == ARPHRD_TUNNEL6 ||
442             dev->type == ARPHRD_SIT ||
443             dev->type == ARPHRD_NONE) {
444                 ndev->cnf.use_tempaddr = -1;
445         }
446
447         ndev->token = in6addr_any;
448
449         if (netif_running(dev) && addrconf_link_ready(dev))
450                 ndev->if_flags |= IF_READY;
451
452         ipv6_mc_init_dev(ndev);
453         ndev->tstamp = jiffies;
454         if (dev != blackhole_netdev) {
455                 err = addrconf_sysctl_register(ndev);
456                 if (err) {
457                         ipv6_mc_destroy_dev(ndev);
458                         snmp6_unregister_dev(ndev);
459                         goto err_release;
460                 }
461         }
462         /* protected by rtnl_lock */
463         rcu_assign_pointer(dev->ip6_ptr, ndev);
464
465         if (dev != blackhole_netdev) {
466                 /* Join interface-local all-node multicast group */
467                 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
468
469                 /* Join all-node multicast group */
470                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
471
472                 /* Join all-router multicast group if forwarding is set */
473                 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
474                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
475         }
476         return ndev;
477
478 err_release:
479         neigh_parms_release(&nd_tbl, ndev->nd_parms);
480         ndev->dead = 1;
481         in6_dev_finish_destroy(ndev);
482         return ERR_PTR(err);
483 }
484
485 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
486 {
487         struct inet6_dev *idev;
488
489         ASSERT_RTNL();
490
491         idev = __in6_dev_get(dev);
492         if (!idev) {
493                 idev = ipv6_add_dev(dev);
494                 if (IS_ERR(idev))
495                         return idev;
496         }
497
498         if (dev->flags&IFF_UP)
499                 ipv6_mc_up(idev);
500         return idev;
501 }
502
503 static int inet6_netconf_msgsize_devconf(int type)
504 {
505         int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
506                     + nla_total_size(4);        /* NETCONFA_IFINDEX */
507         bool all = false;
508
509         if (type == NETCONFA_ALL)
510                 all = true;
511
512         if (all || type == NETCONFA_FORWARDING)
513                 size += nla_total_size(4);
514 #ifdef CONFIG_IPV6_MROUTE
515         if (all || type == NETCONFA_MC_FORWARDING)
516                 size += nla_total_size(4);
517 #endif
518         if (all || type == NETCONFA_PROXY_NEIGH)
519                 size += nla_total_size(4);
520
521         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
522                 size += nla_total_size(4);
523
524         return size;
525 }
526
527 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
528                                       struct ipv6_devconf *devconf, u32 portid,
529                                       u32 seq, int event, unsigned int flags,
530                                       int type)
531 {
532         struct nlmsghdr  *nlh;
533         struct netconfmsg *ncm;
534         bool all = false;
535
536         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
537                         flags);
538         if (!nlh)
539                 return -EMSGSIZE;
540
541         if (type == NETCONFA_ALL)
542                 all = true;
543
544         ncm = nlmsg_data(nlh);
545         ncm->ncm_family = AF_INET6;
546
547         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
548                 goto nla_put_failure;
549
550         if (!devconf)
551                 goto out;
552
553         if ((all || type == NETCONFA_FORWARDING) &&
554             nla_put_s32(skb, NETCONFA_FORWARDING,
555                         READ_ONCE(devconf->forwarding)) < 0)
556                 goto nla_put_failure;
557 #ifdef CONFIG_IPV6_MROUTE
558         if ((all || type == NETCONFA_MC_FORWARDING) &&
559             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
560                         atomic_read(&devconf->mc_forwarding)) < 0)
561                 goto nla_put_failure;
562 #endif
563         if ((all || type == NETCONFA_PROXY_NEIGH) &&
564             nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
565                         READ_ONCE(devconf->proxy_ndp)) < 0)
566                 goto nla_put_failure;
567
568         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
569             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
570                         READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
571                 goto nla_put_failure;
572
573 out:
574         nlmsg_end(skb, nlh);
575         return 0;
576
577 nla_put_failure:
578         nlmsg_cancel(skb, nlh);
579         return -EMSGSIZE;
580 }
581
582 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
583                                   int ifindex, struct ipv6_devconf *devconf)
584 {
585         struct sk_buff *skb;
586         int err = -ENOBUFS;
587
588         skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
589         if (!skb)
590                 goto errout;
591
592         err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
593                                          event, 0, type);
594         if (err < 0) {
595                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
596                 WARN_ON(err == -EMSGSIZE);
597                 kfree_skb(skb);
598                 goto errout;
599         }
600         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
601         return;
602 errout:
603         rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
604 }
605
606 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
607         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
608         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
609         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
610         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
611 };
612
613 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
614                                        const struct nlmsghdr *nlh,
615                                        struct nlattr **tb,
616                                        struct netlink_ext_ack *extack)
617 {
618         int i, err;
619
620         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
621                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
622                 return -EINVAL;
623         }
624
625         if (!netlink_strict_get_check(skb))
626                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
627                                               tb, NETCONFA_MAX,
628                                               devconf_ipv6_policy, extack);
629
630         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
631                                             tb, NETCONFA_MAX,
632                                             devconf_ipv6_policy, extack);
633         if (err)
634                 return err;
635
636         for (i = 0; i <= NETCONFA_MAX; i++) {
637                 if (!tb[i])
638                         continue;
639
640                 switch (i) {
641                 case NETCONFA_IFINDEX:
642                         break;
643                 default:
644                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
645                         return -EINVAL;
646                 }
647         }
648
649         return 0;
650 }
651
652 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
653                                      struct nlmsghdr *nlh,
654                                      struct netlink_ext_ack *extack)
655 {
656         struct net *net = sock_net(in_skb->sk);
657         struct nlattr *tb[NETCONFA_MAX+1];
658         struct inet6_dev *in6_dev = NULL;
659         struct net_device *dev = NULL;
660         struct sk_buff *skb;
661         struct ipv6_devconf *devconf;
662         int ifindex;
663         int err;
664
665         err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
666         if (err < 0)
667                 return err;
668
669         if (!tb[NETCONFA_IFINDEX])
670                 return -EINVAL;
671
672         err = -EINVAL;
673         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
674         switch (ifindex) {
675         case NETCONFA_IFINDEX_ALL:
676                 devconf = net->ipv6.devconf_all;
677                 break;
678         case NETCONFA_IFINDEX_DEFAULT:
679                 devconf = net->ipv6.devconf_dflt;
680                 break;
681         default:
682                 dev = dev_get_by_index(net, ifindex);
683                 if (!dev)
684                         return -EINVAL;
685                 in6_dev = in6_dev_get(dev);
686                 if (!in6_dev)
687                         goto errout;
688                 devconf = &in6_dev->cnf;
689                 break;
690         }
691
692         err = -ENOBUFS;
693         skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
694         if (!skb)
695                 goto errout;
696
697         err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
698                                          NETLINK_CB(in_skb).portid,
699                                          nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
700                                          NETCONFA_ALL);
701         if (err < 0) {
702                 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
703                 WARN_ON(err == -EMSGSIZE);
704                 kfree_skb(skb);
705                 goto errout;
706         }
707         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
708 errout:
709         if (in6_dev)
710                 in6_dev_put(in6_dev);
711         dev_put(dev);
712         return err;
713 }
714
715 /* Combine dev_addr_genid and dev_base_seq to detect changes.
716  */
717 static u32 inet6_base_seq(const struct net *net)
718 {
719         u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
720                   READ_ONCE(net->dev_base_seq);
721
722         /* Must not return 0 (see nl_dump_check_consistent()).
723          * Chose a value far away from 0.
724          */
725         if (!res)
726                 res = 0x80000000;
727         return res;
728 }
729
730 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
731                                       struct netlink_callback *cb)
732 {
733         const struct nlmsghdr *nlh = cb->nlh;
734         struct net *net = sock_net(skb->sk);
735         struct {
736                 unsigned long ifindex;
737                 unsigned int all_default;
738         } *ctx = (void *)cb->ctx;
739         struct net_device *dev;
740         struct inet6_dev *idev;
741         int err = 0;
742
743         if (cb->strict_check) {
744                 struct netlink_ext_ack *extack = cb->extack;
745                 struct netconfmsg *ncm;
746
747                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
748                         NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
749                         return -EINVAL;
750                 }
751
752                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
753                         NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
754                         return -EINVAL;
755                 }
756         }
757
758         rcu_read_lock();
759         for_each_netdev_dump(net, dev, ctx->ifindex) {
760                 idev = __in6_dev_get(dev);
761                 if (!idev)
762                         continue;
763                 err = inet6_netconf_fill_devconf(skb, dev->ifindex,
764                                                  &idev->cnf,
765                                                  NETLINK_CB(cb->skb).portid,
766                                                  nlh->nlmsg_seq,
767                                                  RTM_NEWNETCONF,
768                                                  NLM_F_MULTI,
769                                                  NETCONFA_ALL);
770                 if (err < 0)
771                         goto done;
772         }
773         if (ctx->all_default == 0) {
774                 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
775                                                  net->ipv6.devconf_all,
776                                                  NETLINK_CB(cb->skb).portid,
777                                                  nlh->nlmsg_seq,
778                                                  RTM_NEWNETCONF, NLM_F_MULTI,
779                                                  NETCONFA_ALL);
780                 if (err < 0)
781                         goto done;
782                 ctx->all_default++;
783         }
784         if (ctx->all_default == 1) {
785                 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
786                                                  net->ipv6.devconf_dflt,
787                                                  NETLINK_CB(cb->skb).portid,
788                                                  nlh->nlmsg_seq,
789                                                  RTM_NEWNETCONF, NLM_F_MULTI,
790                                                  NETCONFA_ALL);
791                 if (err < 0)
792                         goto done;
793                 ctx->all_default++;
794         }
795 done:
796         rcu_read_unlock();
797         return err;
798 }
799
800 #ifdef CONFIG_SYSCTL
801 static void dev_forward_change(struct inet6_dev *idev)
802 {
803         struct net_device *dev;
804         struct inet6_ifaddr *ifa;
805         LIST_HEAD(tmp_addr_list);
806
807         if (!idev)
808                 return;
809         dev = idev->dev;
810         if (idev->cnf.forwarding)
811                 dev_disable_lro(dev);
812         if (dev->flags & IFF_MULTICAST) {
813                 if (idev->cnf.forwarding) {
814                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
815                         ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
816                         ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
817                 } else {
818                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
819                         ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
820                         ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
821                 }
822         }
823
824         read_lock_bh(&idev->lock);
825         list_for_each_entry(ifa, &idev->addr_list, if_list) {
826                 if (ifa->flags&IFA_F_TENTATIVE)
827                         continue;
828                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
829         }
830         read_unlock_bh(&idev->lock);
831
832         while (!list_empty(&tmp_addr_list)) {
833                 ifa = list_first_entry(&tmp_addr_list,
834                                        struct inet6_ifaddr, if_list_aux);
835                 list_del(&ifa->if_list_aux);
836                 if (idev->cnf.forwarding)
837                         addrconf_join_anycast(ifa);
838                 else
839                         addrconf_leave_anycast(ifa);
840         }
841
842         inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
843                                      NETCONFA_FORWARDING,
844                                      dev->ifindex, &idev->cnf);
845 }
846
847
848 static void addrconf_forward_change(struct net *net, __s32 newf)
849 {
850         struct net_device *dev;
851         struct inet6_dev *idev;
852
853         for_each_netdev(net, dev) {
854                 idev = __in6_dev_get(dev);
855                 if (idev) {
856                         int changed = (!idev->cnf.forwarding) ^ (!newf);
857
858                         WRITE_ONCE(idev->cnf.forwarding, newf);
859                         if (changed)
860                                 dev_forward_change(idev);
861                 }
862         }
863 }
864
865 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
866 {
867         struct net *net;
868         int old;
869
870         if (!rtnl_trylock())
871                 return restart_syscall();
872
873         net = (struct net *)table->extra2;
874         old = *p;
875         WRITE_ONCE(*p, newf);
876
877         if (p == &net->ipv6.devconf_dflt->forwarding) {
878                 if ((!newf) ^ (!old))
879                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
880                                                      NETCONFA_FORWARDING,
881                                                      NETCONFA_IFINDEX_DEFAULT,
882                                                      net->ipv6.devconf_dflt);
883                 rtnl_unlock();
884                 return 0;
885         }
886
887         if (p == &net->ipv6.devconf_all->forwarding) {
888                 int old_dflt = net->ipv6.devconf_dflt->forwarding;
889
890                 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
891                 if ((!newf) ^ (!old_dflt))
892                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
893                                                      NETCONFA_FORWARDING,
894                                                      NETCONFA_IFINDEX_DEFAULT,
895                                                      net->ipv6.devconf_dflt);
896
897                 addrconf_forward_change(net, newf);
898                 if ((!newf) ^ (!old))
899                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
900                                                      NETCONFA_FORWARDING,
901                                                      NETCONFA_IFINDEX_ALL,
902                                                      net->ipv6.devconf_all);
903         } else if ((!newf) ^ (!old))
904                 dev_forward_change((struct inet6_dev *)table->extra1);
905         rtnl_unlock();
906
907         if (newf)
908                 rt6_purge_dflt_routers(net);
909         return 1;
910 }
911
912 static void addrconf_linkdown_change(struct net *net, __s32 newf)
913 {
914         struct net_device *dev;
915         struct inet6_dev *idev;
916
917         for_each_netdev(net, dev) {
918                 idev = __in6_dev_get(dev);
919                 if (idev) {
920                         int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
921
922                         WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
923                         if (changed)
924                                 inet6_netconf_notify_devconf(dev_net(dev),
925                                                              RTM_NEWNETCONF,
926                                                              NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
927                                                              dev->ifindex,
928                                                              &idev->cnf);
929                 }
930         }
931 }
932
933 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
934 {
935         struct net *net;
936         int old;
937
938         if (!rtnl_trylock())
939                 return restart_syscall();
940
941         net = (struct net *)table->extra2;
942         old = *p;
943         WRITE_ONCE(*p, newf);
944
945         if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
946                 if ((!newf) ^ (!old))
947                         inet6_netconf_notify_devconf(net,
948                                                      RTM_NEWNETCONF,
949                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
950                                                      NETCONFA_IFINDEX_DEFAULT,
951                                                      net->ipv6.devconf_dflt);
952                 rtnl_unlock();
953                 return 0;
954         }
955
956         if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
957                 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
958                 addrconf_linkdown_change(net, newf);
959                 if ((!newf) ^ (!old))
960                         inet6_netconf_notify_devconf(net,
961                                                      RTM_NEWNETCONF,
962                                                      NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
963                                                      NETCONFA_IFINDEX_ALL,
964                                                      net->ipv6.devconf_all);
965         }
966         rtnl_unlock();
967
968         return 1;
969 }
970
971 #endif
972
973 /* Nobody refers to this ifaddr, destroy it */
974 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
975 {
976         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
977
978 #ifdef NET_REFCNT_DEBUG
979         pr_debug("%s\n", __func__);
980 #endif
981
982         in6_dev_put(ifp->idev);
983
984         if (cancel_delayed_work(&ifp->dad_work))
985                 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
986                           ifp);
987
988         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
989                 pr_warn("Freeing alive inet6 address %p\n", ifp);
990                 return;
991         }
992
993         kfree_rcu(ifp, rcu);
994 }
995
996 static void
997 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
998 {
999         struct list_head *p;
1000         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1001
1002         /*
1003          * Each device address list is sorted in order of scope -
1004          * global before linklocal.
1005          */
1006         list_for_each(p, &idev->addr_list) {
1007                 struct inet6_ifaddr *ifa
1008                         = list_entry(p, struct inet6_ifaddr, if_list);
1009                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1010                         break;
1011         }
1012
1013         list_add_tail_rcu(&ifp->if_list, p);
1014 }
1015
1016 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1017 {
1018         u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1019
1020         return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1021 }
1022
1023 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1024                                struct net_device *dev, unsigned int hash)
1025 {
1026         struct inet6_ifaddr *ifp;
1027
1028         hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1029                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1030                         if (!dev || ifp->idev->dev == dev)
1031                                 return true;
1032                 }
1033         }
1034         return false;
1035 }
1036
1037 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1038 {
1039         struct net *net = dev_net(dev);
1040         unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1041         int err = 0;
1042
1043         spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1044
1045         /* Ignore adding duplicate addresses on an interface */
1046         if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1047                 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1048                 err = -EEXIST;
1049         } else {
1050                 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1051         }
1052
1053         spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1054
1055         return err;
1056 }
1057
1058 /* On success it returns ifp with increased reference count */
1059
1060 static struct inet6_ifaddr *
1061 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1062               bool can_block, struct netlink_ext_ack *extack)
1063 {
1064         gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1065         int addr_type = ipv6_addr_type(cfg->pfx);
1066         struct net *net = dev_net(idev->dev);
1067         struct inet6_ifaddr *ifa = NULL;
1068         struct fib6_info *f6i = NULL;
1069         int err = 0;
1070
1071         if (addr_type == IPV6_ADDR_ANY) {
1072                 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1073                 return ERR_PTR(-EADDRNOTAVAIL);
1074         } else if (addr_type & IPV6_ADDR_MULTICAST &&
1075                    !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1076                 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1077                 return ERR_PTR(-EADDRNOTAVAIL);
1078         } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1079                    !netif_is_l3_master(idev->dev) &&
1080                    addr_type & IPV6_ADDR_LOOPBACK) {
1081                 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1082                 return ERR_PTR(-EADDRNOTAVAIL);
1083         }
1084
1085         if (idev->dead) {
1086                 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1087                 err = -ENODEV;
1088                 goto out;
1089         }
1090
1091         if (idev->cnf.disable_ipv6) {
1092                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1093                 err = -EACCES;
1094                 goto out;
1095         }
1096
1097         /* validator notifier needs to be blocking;
1098          * do not call in atomic context
1099          */
1100         if (can_block) {
1101                 struct in6_validator_info i6vi = {
1102                         .i6vi_addr = *cfg->pfx,
1103                         .i6vi_dev = idev,
1104                         .extack = extack,
1105                 };
1106
1107                 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1108                 err = notifier_to_errno(err);
1109                 if (err < 0)
1110                         goto out;
1111         }
1112
1113         ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1114         if (!ifa) {
1115                 err = -ENOBUFS;
1116                 goto out;
1117         }
1118
1119         f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1120         if (IS_ERR(f6i)) {
1121                 err = PTR_ERR(f6i);
1122                 f6i = NULL;
1123                 goto out;
1124         }
1125
1126         neigh_parms_data_state_setall(idev->nd_parms);
1127
1128         ifa->addr = *cfg->pfx;
1129         if (cfg->peer_pfx)
1130                 ifa->peer_addr = *cfg->peer_pfx;
1131
1132         spin_lock_init(&ifa->lock);
1133         INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1134         INIT_HLIST_NODE(&ifa->addr_lst);
1135         ifa->scope = cfg->scope;
1136         ifa->prefix_len = cfg->plen;
1137         ifa->rt_priority = cfg->rt_priority;
1138         ifa->flags = cfg->ifa_flags;
1139         ifa->ifa_proto = cfg->ifa_proto;
1140         /* No need to add the TENTATIVE flag for addresses with NODAD */
1141         if (!(cfg->ifa_flags & IFA_F_NODAD))
1142                 ifa->flags |= IFA_F_TENTATIVE;
1143         ifa->valid_lft = cfg->valid_lft;
1144         ifa->prefered_lft = cfg->preferred_lft;
1145         ifa->cstamp = ifa->tstamp = jiffies;
1146         ifa->tokenized = false;
1147
1148         ifa->rt = f6i;
1149
1150         ifa->idev = idev;
1151         in6_dev_hold(idev);
1152
1153         /* For caller */
1154         refcount_set(&ifa->refcnt, 1);
1155
1156         rcu_read_lock();
1157
1158         err = ipv6_add_addr_hash(idev->dev, ifa);
1159         if (err < 0) {
1160                 rcu_read_unlock();
1161                 goto out;
1162         }
1163
1164         write_lock_bh(&idev->lock);
1165
1166         /* Add to inet6_dev unicast addr list. */
1167         ipv6_link_dev_addr(idev, ifa);
1168
1169         if (ifa->flags&IFA_F_TEMPORARY) {
1170                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1171                 in6_ifa_hold(ifa);
1172         }
1173
1174         in6_ifa_hold(ifa);
1175         write_unlock_bh(&idev->lock);
1176
1177         rcu_read_unlock();
1178
1179         inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1180 out:
1181         if (unlikely(err < 0)) {
1182                 fib6_info_release(f6i);
1183
1184                 if (ifa) {
1185                         if (ifa->idev)
1186                                 in6_dev_put(ifa->idev);
1187                         kfree(ifa);
1188                 }
1189                 ifa = ERR_PTR(err);
1190         }
1191
1192         return ifa;
1193 }
1194
1195 enum cleanup_prefix_rt_t {
1196         CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
1197         CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
1198         CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1199 };
1200
1201 /*
1202  * Check, whether the prefix for ifp would still need a prefix route
1203  * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1204  * constants.
1205  *
1206  * 1) we don't purge prefix if address was not permanent.
1207  *    prefix is managed by its own lifetime.
1208  * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1209  * 3) if there are no addresses, delete prefix.
1210  * 4) if there are still other permanent address(es),
1211  *    corresponding prefix is still permanent.
1212  * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1213  *    don't purge the prefix, assume user space is managing it.
1214  * 6) otherwise, update prefix lifetime to the
1215  *    longest valid lifetime among the corresponding
1216  *    addresses on the device.
1217  *    Note: subsequent RA will update lifetime.
1218  **/
1219 static enum cleanup_prefix_rt_t
1220 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1221 {
1222         struct inet6_ifaddr *ifa;
1223         struct inet6_dev *idev = ifp->idev;
1224         unsigned long lifetime;
1225         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1226
1227         *expires = jiffies;
1228
1229         list_for_each_entry(ifa, &idev->addr_list, if_list) {
1230                 if (ifa == ifp)
1231                         continue;
1232                 if (ifa->prefix_len != ifp->prefix_len ||
1233                     !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1234                                        ifp->prefix_len))
1235                         continue;
1236                 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1237                         return CLEANUP_PREFIX_RT_NOP;
1238
1239                 action = CLEANUP_PREFIX_RT_EXPIRE;
1240
1241                 spin_lock(&ifa->lock);
1242
1243                 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1244                 /*
1245                  * Note: Because this address is
1246                  * not permanent, lifetime <
1247                  * LONG_MAX / HZ here.
1248                  */
1249                 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1250                         *expires = ifa->tstamp + lifetime * HZ;
1251                 spin_unlock(&ifa->lock);
1252         }
1253
1254         return action;
1255 }
1256
1257 static void
1258 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1259                      bool del_rt, bool del_peer)
1260 {
1261         struct fib6_table *table;
1262         struct fib6_info *f6i;
1263
1264         f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1265                                         ifp->prefix_len,
1266                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
1267         if (f6i) {
1268                 if (del_rt)
1269                         ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1270                 else {
1271                         if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1272                                 table = f6i->fib6_table;
1273                                 spin_lock_bh(&table->tb6_lock);
1274
1275                                 fib6_set_expires(f6i, expires);
1276                                 fib6_add_gc_list(f6i);
1277
1278                                 spin_unlock_bh(&table->tb6_lock);
1279                         }
1280                         fib6_info_release(f6i);
1281                 }
1282         }
1283 }
1284
1285
1286 /* This function wants to get referenced ifp and releases it before return */
1287
1288 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1289 {
1290         enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1291         struct net *net = dev_net(ifp->idev->dev);
1292         unsigned long expires;
1293         int state;
1294
1295         ASSERT_RTNL();
1296
1297         spin_lock_bh(&ifp->lock);
1298         state = ifp->state;
1299         ifp->state = INET6_IFADDR_STATE_DEAD;
1300         spin_unlock_bh(&ifp->lock);
1301
1302         if (state == INET6_IFADDR_STATE_DEAD)
1303                 goto out;
1304
1305         spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1306         hlist_del_init_rcu(&ifp->addr_lst);
1307         spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1308
1309         write_lock_bh(&ifp->idev->lock);
1310
1311         if (ifp->flags&IFA_F_TEMPORARY) {
1312                 list_del(&ifp->tmp_list);
1313                 if (ifp->ifpub) {
1314                         in6_ifa_put(ifp->ifpub);
1315                         ifp->ifpub = NULL;
1316                 }
1317                 __in6_ifa_put(ifp);
1318         }
1319
1320         if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1321                 action = check_cleanup_prefix_route(ifp, &expires);
1322
1323         list_del_rcu(&ifp->if_list);
1324         __in6_ifa_put(ifp);
1325
1326         write_unlock_bh(&ifp->idev->lock);
1327
1328         addrconf_del_dad_work(ifp);
1329
1330         ipv6_ifa_notify(RTM_DELADDR, ifp);
1331
1332         inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1333
1334         if (action != CLEANUP_PREFIX_RT_NOP) {
1335                 cleanup_prefix_route(ifp, expires,
1336                         action == CLEANUP_PREFIX_RT_DEL, false);
1337         }
1338
1339         /* clean up prefsrc entries */
1340         rt6_remove_prefsrc(ifp);
1341 out:
1342         in6_ifa_put(ifp);
1343 }
1344
1345 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1346 {
1347         return READ_ONCE(idev->cnf.regen_min_advance) +
1348                 READ_ONCE(idev->cnf.regen_max_retry) *
1349                 READ_ONCE(idev->cnf.dad_transmits) *
1350                 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1351 }
1352
1353 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1354 {
1355         struct inet6_dev *idev = ifp->idev;
1356         unsigned long tmp_tstamp, age;
1357         unsigned long regen_advance;
1358         unsigned long now = jiffies;
1359         u32 if_public_preferred_lft;
1360         s32 cnf_temp_preferred_lft;
1361         struct inet6_ifaddr *ift;
1362         struct ifa6_config cfg;
1363         long max_desync_factor;
1364         struct in6_addr addr;
1365         int ret = 0;
1366
1367         write_lock_bh(&idev->lock);
1368
1369 retry:
1370         in6_dev_hold(idev);
1371         if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1372                 write_unlock_bh(&idev->lock);
1373                 pr_info("%s: use_tempaddr is disabled\n", __func__);
1374                 in6_dev_put(idev);
1375                 ret = -1;
1376                 goto out;
1377         }
1378         spin_lock_bh(&ifp->lock);
1379         if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1380                 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/
1381                 spin_unlock_bh(&ifp->lock);
1382                 write_unlock_bh(&idev->lock);
1383                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1384                         __func__);
1385                 in6_dev_put(idev);
1386                 ret = -1;
1387                 goto out;
1388         }
1389         in6_ifa_hold(ifp);
1390         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1391         ipv6_gen_rnd_iid(&addr);
1392
1393         age = (now - ifp->tstamp) / HZ;
1394
1395         regen_advance = ipv6_get_regen_advance(idev);
1396
1397         /* recalculate max_desync_factor each time and update
1398          * idev->desync_factor if it's larger
1399          */
1400         cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1401         max_desync_factor = min_t(long,
1402                                   READ_ONCE(idev->cnf.max_desync_factor),
1403                                   cnf_temp_preferred_lft - regen_advance);
1404
1405         if (unlikely(idev->desync_factor > max_desync_factor)) {
1406                 if (max_desync_factor > 0) {
1407                         get_random_bytes(&idev->desync_factor,
1408                                          sizeof(idev->desync_factor));
1409                         idev->desync_factor %= max_desync_factor;
1410                 } else {
1411                         idev->desync_factor = 0;
1412                 }
1413         }
1414
1415         if_public_preferred_lft = ifp->prefered_lft;
1416
1417         memset(&cfg, 0, sizeof(cfg));
1418         cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1419                               READ_ONCE(idev->cnf.temp_valid_lft) + age);
1420         cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1421         cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1422         cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1423
1424         cfg.plen = ifp->prefix_len;
1425         tmp_tstamp = ifp->tstamp;
1426         spin_unlock_bh(&ifp->lock);
1427
1428         write_unlock_bh(&idev->lock);
1429
1430         /* From RFC 4941:
1431          *
1432          *     A temporary address is created only if this calculated Preferred
1433          *     Lifetime is greater than REGEN_ADVANCE time units.  In
1434          *     particular, an implementation must not create a temporary address
1435          *     with a zero Preferred Lifetime.
1436          *
1437          *     ...
1438          *
1439          *     When creating a temporary address, the lifetime values MUST be
1440          *     derived from the corresponding prefix as follows:
1441          *
1442          *     ...
1443          *
1444          *     *  Its Preferred Lifetime is the lower of the Preferred Lifetime
1445          *        of the public address or TEMP_PREFERRED_LIFETIME -
1446          *        DESYNC_FACTOR.
1447          *
1448          * To comply with the RFC's requirements, clamp the preferred lifetime
1449          * to a minimum of regen_advance, unless that would exceed valid_lft or
1450          * ifp->prefered_lft.
1451          *
1452          * Use age calculation as in addrconf_verify to avoid unnecessary
1453          * temporary addresses being generated.
1454          */
1455         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1456         if (cfg.preferred_lft <= regen_advance + age) {
1457                 cfg.preferred_lft = regen_advance + age + 1;
1458                 if (cfg.preferred_lft > cfg.valid_lft ||
1459                     cfg.preferred_lft > if_public_preferred_lft) {
1460                         in6_ifa_put(ifp);
1461                         in6_dev_put(idev);
1462                         ret = -1;
1463                         goto out;
1464                 }
1465         }
1466
1467         cfg.ifa_flags = IFA_F_TEMPORARY;
1468         /* set in addrconf_prefix_rcv() */
1469         if (ifp->flags & IFA_F_OPTIMISTIC)
1470                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1471
1472         cfg.pfx = &addr;
1473         cfg.scope = ipv6_addr_scope(cfg.pfx);
1474
1475         ift = ipv6_add_addr(idev, &cfg, block, NULL);
1476         if (IS_ERR(ift)) {
1477                 in6_ifa_put(ifp);
1478                 in6_dev_put(idev);
1479                 pr_info("%s: retry temporary address regeneration\n", __func__);
1480                 write_lock_bh(&idev->lock);
1481                 goto retry;
1482         }
1483
1484         spin_lock_bh(&ift->lock);
1485         ift->ifpub = ifp;
1486         ift->cstamp = now;
1487         ift->tstamp = tmp_tstamp;
1488         spin_unlock_bh(&ift->lock);
1489
1490         addrconf_dad_start(ift);
1491         in6_ifa_put(ift);
1492         in6_dev_put(idev);
1493 out:
1494         return ret;
1495 }
1496
1497 /*
1498  *      Choose an appropriate source address (RFC3484)
1499  */
1500 enum {
1501         IPV6_SADDR_RULE_INIT = 0,
1502         IPV6_SADDR_RULE_LOCAL,
1503         IPV6_SADDR_RULE_SCOPE,
1504         IPV6_SADDR_RULE_PREFERRED,
1505 #ifdef CONFIG_IPV6_MIP6
1506         IPV6_SADDR_RULE_HOA,
1507 #endif
1508         IPV6_SADDR_RULE_OIF,
1509         IPV6_SADDR_RULE_LABEL,
1510         IPV6_SADDR_RULE_PRIVACY,
1511         IPV6_SADDR_RULE_ORCHID,
1512         IPV6_SADDR_RULE_PREFIX,
1513 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1514         IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1515 #endif
1516         IPV6_SADDR_RULE_MAX
1517 };
1518
1519 struct ipv6_saddr_score {
1520         int                     rule;
1521         int                     addr_type;
1522         struct inet6_ifaddr     *ifa;
1523         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1524         int                     scopedist;
1525         int                     matchlen;
1526 };
1527
1528 struct ipv6_saddr_dst {
1529         const struct in6_addr *addr;
1530         int ifindex;
1531         int scope;
1532         int label;
1533         unsigned int prefs;
1534 };
1535
1536 static inline int ipv6_saddr_preferred(int type)
1537 {
1538         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1539                 return 1;
1540         return 0;
1541 }
1542
1543 static bool ipv6_use_optimistic_addr(const struct net *net,
1544                                      const struct inet6_dev *idev)
1545 {
1546 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1547         if (!idev)
1548                 return false;
1549         if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1550             !READ_ONCE(idev->cnf.optimistic_dad))
1551                 return false;
1552         if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1553             !READ_ONCE(idev->cnf.use_optimistic))
1554                 return false;
1555
1556         return true;
1557 #else
1558         return false;
1559 #endif
1560 }
1561
1562 static bool ipv6_allow_optimistic_dad(const struct net *net,
1563                                       const struct inet6_dev *idev)
1564 {
1565 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1566         if (!idev)
1567                 return false;
1568         if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1569             !READ_ONCE(idev->cnf.optimistic_dad))
1570                 return false;
1571
1572         return true;
1573 #else
1574         return false;
1575 #endif
1576 }
1577
1578 static int ipv6_get_saddr_eval(struct net *net,
1579                                struct ipv6_saddr_score *score,
1580                                struct ipv6_saddr_dst *dst,
1581                                int i)
1582 {
1583         int ret;
1584
1585         if (i <= score->rule) {
1586                 switch (i) {
1587                 case IPV6_SADDR_RULE_SCOPE:
1588                         ret = score->scopedist;
1589                         break;
1590                 case IPV6_SADDR_RULE_PREFIX:
1591                         ret = score->matchlen;
1592                         break;
1593                 default:
1594                         ret = !!test_bit(i, score->scorebits);
1595                 }
1596                 goto out;
1597         }
1598
1599         switch (i) {
1600         case IPV6_SADDR_RULE_INIT:
1601                 /* Rule 0: remember if hiscore is not ready yet */
1602                 ret = !!score->ifa;
1603                 break;
1604         case IPV6_SADDR_RULE_LOCAL:
1605                 /* Rule 1: Prefer same address */
1606                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1607                 break;
1608         case IPV6_SADDR_RULE_SCOPE:
1609                 /* Rule 2: Prefer appropriate scope
1610                  *
1611                  *      ret
1612                  *       ^
1613                  *    -1 |  d 15
1614                  *    ---+--+-+---> scope
1615                  *       |
1616                  *       |             d is scope of the destination.
1617                  *  B-d  |  \
1618                  *       |   \      <- smaller scope is better if
1619                  *  B-15 |    \        if scope is enough for destination.
1620                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1621                  * d-C-1 | /
1622                  *       |/         <- greater is better
1623                  *   -C  /             if scope is not enough for destination.
1624                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1625                  *
1626                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1627                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1628                  * Assume B = 0 and we get C > 29.
1629                  */
1630                 ret = __ipv6_addr_src_scope(score->addr_type);
1631                 if (ret >= dst->scope)
1632                         ret = -ret;
1633                 else
1634                         ret -= 128;     /* 30 is enough */
1635                 score->scopedist = ret;
1636                 break;
1637         case IPV6_SADDR_RULE_PREFERRED:
1638             {
1639                 /* Rule 3: Avoid deprecated and optimistic addresses */
1640                 u8 avoid = IFA_F_DEPRECATED;
1641
1642                 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1643                         avoid |= IFA_F_OPTIMISTIC;
1644                 ret = ipv6_saddr_preferred(score->addr_type) ||
1645                       !(score->ifa->flags & avoid);
1646                 break;
1647             }
1648 #ifdef CONFIG_IPV6_MIP6
1649         case IPV6_SADDR_RULE_HOA:
1650             {
1651                 /* Rule 4: Prefer home address */
1652                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1653                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1654                 break;
1655             }
1656 #endif
1657         case IPV6_SADDR_RULE_OIF:
1658                 /* Rule 5: Prefer outgoing interface */
1659                 ret = (!dst->ifindex ||
1660                        dst->ifindex == score->ifa->idev->dev->ifindex);
1661                 break;
1662         case IPV6_SADDR_RULE_LABEL:
1663                 /* Rule 6: Prefer matching label */
1664                 ret = ipv6_addr_label(net,
1665                                       &score->ifa->addr, score->addr_type,
1666                                       score->ifa->idev->dev->ifindex) == dst->label;
1667                 break;
1668         case IPV6_SADDR_RULE_PRIVACY:
1669             {
1670                 /* Rule 7: Prefer public address
1671                  * Note: prefer temporary address if use_tempaddr >= 2
1672                  */
1673                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1674                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1675                                 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1676                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1677                 break;
1678             }
1679         case IPV6_SADDR_RULE_ORCHID:
1680                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1681                  *          non-ORCHID vs non-ORCHID
1682                  */
1683                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1684                         ipv6_addr_orchid(dst->addr));
1685                 break;
1686         case IPV6_SADDR_RULE_PREFIX:
1687                 /* Rule 8: Use longest matching prefix */
1688                 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1689                 if (ret > score->ifa->prefix_len)
1690                         ret = score->ifa->prefix_len;
1691                 score->matchlen = ret;
1692                 break;
1693 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1694         case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1695                 /* Optimistic addresses still have lower precedence than other
1696                  * preferred addresses.
1697                  */
1698                 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1699                 break;
1700 #endif
1701         default:
1702                 ret = 0;
1703         }
1704
1705         if (ret)
1706                 __set_bit(i, score->scorebits);
1707         score->rule = i;
1708 out:
1709         return ret;
1710 }
1711
1712 static int __ipv6_dev_get_saddr(struct net *net,
1713                                 struct ipv6_saddr_dst *dst,
1714                                 struct inet6_dev *idev,
1715                                 struct ipv6_saddr_score *scores,
1716                                 int hiscore_idx)
1717 {
1718         struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1719
1720         list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1721                 int i;
1722
1723                 /*
1724                  * - Tentative Address (RFC2462 section 5.4)
1725                  *  - A tentative address is not considered
1726                  *    "assigned to an interface" in the traditional
1727                  *    sense, unless it is also flagged as optimistic.
1728                  * - Candidate Source Address (section 4)
1729                  *  - In any case, anycast addresses, multicast
1730                  *    addresses, and the unspecified address MUST
1731                  *    NOT be included in a candidate set.
1732                  */
1733                 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1734                     (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1735                         continue;
1736
1737                 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1738
1739                 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1740                              score->addr_type & IPV6_ADDR_MULTICAST)) {
1741                         net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1742                                             idev->dev->name);
1743                         continue;
1744                 }
1745
1746                 score->rule = -1;
1747                 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1748
1749                 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1750                         int minihiscore, miniscore;
1751
1752                         minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1753                         miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1754
1755                         if (minihiscore > miniscore) {
1756                                 if (i == IPV6_SADDR_RULE_SCOPE &&
1757                                     score->scopedist > 0) {
1758                                         /*
1759                                          * special case:
1760                                          * each remaining entry
1761                                          * has too small (not enough)
1762                                          * scope, because ifa entries
1763                                          * are sorted by their scope
1764                                          * values.
1765                                          */
1766                                         goto out;
1767                                 }
1768                                 break;
1769                         } else if (minihiscore < miniscore) {
1770                                 swap(hiscore, score);
1771                                 hiscore_idx = 1 - hiscore_idx;
1772
1773                                 /* restore our iterator */
1774                                 score->ifa = hiscore->ifa;
1775
1776                                 break;
1777                         }
1778                 }
1779         }
1780 out:
1781         return hiscore_idx;
1782 }
1783
1784 static int ipv6_get_saddr_master(struct net *net,
1785                                  const struct net_device *dst_dev,
1786                                  const struct net_device *master,
1787                                  struct ipv6_saddr_dst *dst,
1788                                  struct ipv6_saddr_score *scores,
1789                                  int hiscore_idx)
1790 {
1791         struct inet6_dev *idev;
1792
1793         idev = __in6_dev_get(dst_dev);
1794         if (idev)
1795                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1796                                                    scores, hiscore_idx);
1797
1798         idev = __in6_dev_get(master);
1799         if (idev)
1800                 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1801                                                    scores, hiscore_idx);
1802
1803         return hiscore_idx;
1804 }
1805
1806 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1807                        const struct in6_addr *daddr, unsigned int prefs,
1808                        struct in6_addr *saddr)
1809 {
1810         struct ipv6_saddr_score scores[2], *hiscore;
1811         struct ipv6_saddr_dst dst;
1812         struct inet6_dev *idev;
1813         struct net_device *dev;
1814         int dst_type;
1815         bool use_oif_addr = false;
1816         int hiscore_idx = 0;
1817         int ret = 0;
1818
1819         dst_type = __ipv6_addr_type(daddr);
1820         dst.addr = daddr;
1821         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1822         dst.scope = __ipv6_addr_src_scope(dst_type);
1823         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1824         dst.prefs = prefs;
1825
1826         scores[hiscore_idx].rule = -1;
1827         scores[hiscore_idx].ifa = NULL;
1828
1829         rcu_read_lock();
1830
1831         /* Candidate Source Address (section 4)
1832          *  - multicast and link-local destination address,
1833          *    the set of candidate source address MUST only
1834          *    include addresses assigned to interfaces
1835          *    belonging to the same link as the outgoing
1836          *    interface.
1837          * (- For site-local destination addresses, the
1838          *    set of candidate source addresses MUST only
1839          *    include addresses assigned to interfaces
1840          *    belonging to the same site as the outgoing
1841          *    interface.)
1842          *  - "It is RECOMMENDED that the candidate source addresses
1843          *    be the set of unicast addresses assigned to the
1844          *    interface that will be used to send to the destination
1845          *    (the 'outgoing' interface)." (RFC 6724)
1846          */
1847         if (dst_dev) {
1848                 idev = __in6_dev_get(dst_dev);
1849                 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1850                     dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1851                     (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1852                         use_oif_addr = true;
1853                 }
1854         }
1855
1856         if (use_oif_addr) {
1857                 if (idev)
1858                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1859         } else {
1860                 const struct net_device *master;
1861                 int master_idx = 0;
1862
1863                 /* if dst_dev exists and is enslaved to an L3 device, then
1864                  * prefer addresses from dst_dev and then the master over
1865                  * any other enslaved devices in the L3 domain.
1866                  */
1867                 master = l3mdev_master_dev_rcu(dst_dev);
1868                 if (master) {
1869                         master_idx = master->ifindex;
1870
1871                         hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1872                                                             master, &dst,
1873                                                             scores, hiscore_idx);
1874
1875                         if (scores[hiscore_idx].ifa)
1876                                 goto out;
1877                 }
1878
1879                 for_each_netdev_rcu(net, dev) {
1880                         /* only consider addresses on devices in the
1881                          * same L3 domain
1882                          */
1883                         if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1884                                 continue;
1885                         idev = __in6_dev_get(dev);
1886                         if (!idev)
1887                                 continue;
1888                         hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1889                 }
1890         }
1891
1892 out:
1893         hiscore = &scores[hiscore_idx];
1894         if (!hiscore->ifa)
1895                 ret = -EADDRNOTAVAIL;
1896         else
1897                 *saddr = hiscore->ifa->addr;
1898
1899         rcu_read_unlock();
1900         return ret;
1901 }
1902 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1903
1904 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1905                               u32 banned_flags)
1906 {
1907         struct inet6_ifaddr *ifp;
1908         int err = -EADDRNOTAVAIL;
1909
1910         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1911                 if (ifp->scope > IFA_LINK)
1912                         break;
1913                 if (ifp->scope == IFA_LINK &&
1914                     !(ifp->flags & banned_flags)) {
1915                         *addr = ifp->addr;
1916                         err = 0;
1917                         break;
1918                 }
1919         }
1920         return err;
1921 }
1922
1923 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1924                     u32 banned_flags)
1925 {
1926         struct inet6_dev *idev;
1927         int err = -EADDRNOTAVAIL;
1928
1929         rcu_read_lock();
1930         idev = __in6_dev_get(dev);
1931         if (idev) {
1932                 read_lock_bh(&idev->lock);
1933                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1934                 read_unlock_bh(&idev->lock);
1935         }
1936         rcu_read_unlock();
1937         return err;
1938 }
1939
1940 static int ipv6_count_addresses(const struct inet6_dev *idev)
1941 {
1942         const struct inet6_ifaddr *ifp;
1943         int cnt = 0;
1944
1945         rcu_read_lock();
1946         list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1947                 cnt++;
1948         rcu_read_unlock();
1949         return cnt;
1950 }
1951
1952 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1953                   const struct net_device *dev, int strict)
1954 {
1955         return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1956                                        strict, IFA_F_TENTATIVE);
1957 }
1958 EXPORT_SYMBOL(ipv6_chk_addr);
1959
1960 /* device argument is used to find the L3 domain of interest. If
1961  * skip_dev_check is set, then the ifp device is not checked against
1962  * the passed in dev argument. So the 2 cases for addresses checks are:
1963  *   1. does the address exist in the L3 domain that dev is part of
1964  *      (skip_dev_check = true), or
1965  *
1966  *   2. does the address exist on the specific device
1967  *      (skip_dev_check = false)
1968  */
1969 static struct net_device *
1970 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1971                           const struct net_device *dev, bool skip_dev_check,
1972                           int strict, u32 banned_flags)
1973 {
1974         unsigned int hash = inet6_addr_hash(net, addr);
1975         struct net_device *l3mdev, *ndev;
1976         struct inet6_ifaddr *ifp;
1977         u32 ifp_flags;
1978
1979         rcu_read_lock();
1980
1981         l3mdev = l3mdev_master_dev_rcu(dev);
1982         if (skip_dev_check)
1983                 dev = NULL;
1984
1985         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1986                 ndev = ifp->idev->dev;
1987
1988                 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1989                         continue;
1990
1991                 /* Decouple optimistic from tentative for evaluation here.
1992                  * Ban optimistic addresses explicitly, when required.
1993                  */
1994                 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1995                             ? (ifp->flags&~IFA_F_TENTATIVE)
1996                             : ifp->flags;
1997                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1998                     !(ifp_flags&banned_flags) &&
1999                     (!dev || ndev == dev ||
2000                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2001                         rcu_read_unlock();
2002                         return ndev;
2003                 }
2004         }
2005
2006         rcu_read_unlock();
2007         return NULL;
2008 }
2009
2010 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2011                             const struct net_device *dev, bool skip_dev_check,
2012                             int strict, u32 banned_flags)
2013 {
2014         return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2015                                          strict, banned_flags) ? 1 : 0;
2016 }
2017 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2018
2019
2020 /* Compares an address/prefix_len with addresses on device @dev.
2021  * If one is found it returns true.
2022  */
2023 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2024         const unsigned int prefix_len, struct net_device *dev)
2025 {
2026         const struct inet6_ifaddr *ifa;
2027         const struct inet6_dev *idev;
2028         bool ret = false;
2029
2030         rcu_read_lock();
2031         idev = __in6_dev_get(dev);
2032         if (idev) {
2033                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2034                         ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2035                         if (ret)
2036                                 break;
2037                 }
2038         }
2039         rcu_read_unlock();
2040
2041         return ret;
2042 }
2043 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2044
2045 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2046 {
2047         const struct inet6_ifaddr *ifa;
2048         const struct inet6_dev *idev;
2049         int     onlink;
2050
2051         onlink = 0;
2052         rcu_read_lock();
2053         idev = __in6_dev_get(dev);
2054         if (idev) {
2055                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2056                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
2057                                                    ifa->prefix_len);
2058                         if (onlink)
2059                                 break;
2060                 }
2061         }
2062         rcu_read_unlock();
2063         return onlink;
2064 }
2065 EXPORT_SYMBOL(ipv6_chk_prefix);
2066
2067 /**
2068  * ipv6_dev_find - find the first device with a given source address.
2069  * @net: the net namespace
2070  * @addr: the source address
2071  * @dev: used to find the L3 domain of interest
2072  *
2073  * The caller should be protected by RCU, or RTNL.
2074  */
2075 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2076                                  struct net_device *dev)
2077 {
2078         return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2079                                          IFA_F_TENTATIVE);
2080 }
2081 EXPORT_SYMBOL(ipv6_dev_find);
2082
2083 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2084                                      struct net_device *dev, int strict)
2085 {
2086         unsigned int hash = inet6_addr_hash(net, addr);
2087         struct inet6_ifaddr *ifp, *result = NULL;
2088
2089         rcu_read_lock();
2090         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2091                 if (ipv6_addr_equal(&ifp->addr, addr)) {
2092                         if (!dev || ifp->idev->dev == dev ||
2093                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2094                                 result = ifp;
2095                                 in6_ifa_hold(ifp);
2096                                 break;
2097                         }
2098                 }
2099         }
2100         rcu_read_unlock();
2101
2102         return result;
2103 }
2104
2105 /* Gets referenced address, destroys ifaddr */
2106
2107 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2108 {
2109         if (dad_failed)
2110                 ifp->flags |= IFA_F_DADFAILED;
2111
2112         if (ifp->flags&IFA_F_TEMPORARY) {
2113                 struct inet6_ifaddr *ifpub;
2114                 spin_lock_bh(&ifp->lock);
2115                 ifpub = ifp->ifpub;
2116                 if (ifpub) {
2117                         in6_ifa_hold(ifpub);
2118                         spin_unlock_bh(&ifp->lock);
2119                         ipv6_create_tempaddr(ifpub, true);
2120                         in6_ifa_put(ifpub);
2121                 } else {
2122                         spin_unlock_bh(&ifp->lock);
2123                 }
2124                 ipv6_del_addr(ifp);
2125         } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2126                 spin_lock_bh(&ifp->lock);
2127                 addrconf_del_dad_work(ifp);
2128                 ifp->flags |= IFA_F_TENTATIVE;
2129                 if (dad_failed)
2130                         ifp->flags &= ~IFA_F_OPTIMISTIC;
2131                 spin_unlock_bh(&ifp->lock);
2132                 if (dad_failed)
2133                         ipv6_ifa_notify(0, ifp);
2134                 in6_ifa_put(ifp);
2135         } else {
2136                 ipv6_del_addr(ifp);
2137         }
2138 }
2139
2140 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2141 {
2142         int err = -ENOENT;
2143
2144         spin_lock_bh(&ifp->lock);
2145         if (ifp->state == INET6_IFADDR_STATE_DAD) {
2146                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2147                 err = 0;
2148         }
2149         spin_unlock_bh(&ifp->lock);
2150
2151         return err;
2152 }
2153
2154 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2155 {
2156         struct inet6_dev *idev = ifp->idev;
2157         struct net *net = dev_net(idev->dev);
2158         int max_addresses;
2159
2160         if (addrconf_dad_end(ifp)) {
2161                 in6_ifa_put(ifp);
2162                 return;
2163         }
2164
2165         net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2166                              ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2167
2168         spin_lock_bh(&ifp->lock);
2169
2170         if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2171                 struct in6_addr new_addr;
2172                 struct inet6_ifaddr *ifp2;
2173                 int retries = ifp->stable_privacy_retry + 1;
2174                 struct ifa6_config cfg = {
2175                         .pfx = &new_addr,
2176                         .plen = ifp->prefix_len,
2177                         .ifa_flags = ifp->flags,
2178                         .valid_lft = ifp->valid_lft,
2179                         .preferred_lft = ifp->prefered_lft,
2180                         .scope = ifp->scope,
2181                 };
2182
2183                 if (retries > net->ipv6.sysctl.idgen_retries) {
2184                         net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2185                                              ifp->idev->dev->name);
2186                         goto errdad;
2187                 }
2188
2189                 new_addr = ifp->addr;
2190                 if (ipv6_generate_stable_address(&new_addr, retries,
2191                                                  idev))
2192                         goto errdad;
2193
2194                 spin_unlock_bh(&ifp->lock);
2195
2196                 max_addresses = READ_ONCE(idev->cnf.max_addresses);
2197                 if (max_addresses &&
2198                     ipv6_count_addresses(idev) >= max_addresses)
2199                         goto lock_errdad;
2200
2201                 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2202                                      ifp->idev->dev->name);
2203
2204                 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2205                 if (IS_ERR(ifp2))
2206                         goto lock_errdad;
2207
2208                 spin_lock_bh(&ifp2->lock);
2209                 ifp2->stable_privacy_retry = retries;
2210                 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2211                 spin_unlock_bh(&ifp2->lock);
2212
2213                 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2214                 in6_ifa_put(ifp2);
2215 lock_errdad:
2216                 spin_lock_bh(&ifp->lock);
2217         }
2218
2219 errdad:
2220         /* transition from _POSTDAD to _ERRDAD */
2221         ifp->state = INET6_IFADDR_STATE_ERRDAD;
2222         spin_unlock_bh(&ifp->lock);
2223
2224         addrconf_mod_dad_work(ifp, 0);
2225         in6_ifa_put(ifp);
2226 }
2227
2228 /* Join to solicited addr multicast group.
2229  * caller must hold RTNL */
2230 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2231 {
2232         struct in6_addr maddr;
2233
2234         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2235                 return;
2236
2237         addrconf_addr_solict_mult(addr, &maddr);
2238         ipv6_dev_mc_inc(dev, &maddr);
2239 }
2240
2241 /* caller must hold RTNL */
2242 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2243 {
2244         struct in6_addr maddr;
2245
2246         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2247                 return;
2248
2249         addrconf_addr_solict_mult(addr, &maddr);
2250         __ipv6_dev_mc_dec(idev, &maddr);
2251 }
2252
2253 /* caller must hold RTNL */
2254 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2255 {
2256         struct in6_addr addr;
2257
2258         if (ifp->prefix_len >= 127) /* RFC 6164 */
2259                 return;
2260         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2261         if (ipv6_addr_any(&addr))
2262                 return;
2263         __ipv6_dev_ac_inc(ifp->idev, &addr);
2264 }
2265
2266 /* caller must hold RTNL */
2267 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2268 {
2269         struct in6_addr addr;
2270
2271         if (ifp->prefix_len >= 127) /* RFC 6164 */
2272                 return;
2273         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2274         if (ipv6_addr_any(&addr))
2275                 return;
2276         __ipv6_dev_ac_dec(ifp->idev, &addr);
2277 }
2278
2279 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2280 {
2281         switch (dev->addr_len) {
2282         case ETH_ALEN:
2283                 memcpy(eui, dev->dev_addr, 3);
2284                 eui[3] = 0xFF;
2285                 eui[4] = 0xFE;
2286                 memcpy(eui + 5, dev->dev_addr + 3, 3);
2287                 break;
2288         case EUI64_ADDR_LEN:
2289                 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2290                 eui[0] ^= 2;
2291                 break;
2292         default:
2293                 return -1;
2294         }
2295
2296         return 0;
2297 }
2298
2299 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2300 {
2301         const union fwnet_hwaddr *ha;
2302
2303         if (dev->addr_len != FWNET_ALEN)
2304                 return -1;
2305
2306         ha = (const union fwnet_hwaddr *)dev->dev_addr;
2307
2308         memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2309         eui[0] ^= 2;
2310         return 0;
2311 }
2312
2313 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2314 {
2315         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2316         if (dev->addr_len != ARCNET_ALEN)
2317                 return -1;
2318         memset(eui, 0, 7);
2319         eui[7] = *(u8 *)dev->dev_addr;
2320         return 0;
2321 }
2322
2323 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2324 {
2325         if (dev->addr_len != INFINIBAND_ALEN)
2326                 return -1;
2327         memcpy(eui, dev->dev_addr + 12, 8);
2328         eui[0] |= 2;
2329         return 0;
2330 }
2331
2332 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2333 {
2334         if (addr == 0)
2335                 return -1;
2336         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2337                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2338                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2339                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2340                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2341                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2342         eui[1] = 0;
2343         eui[2] = 0x5E;
2344         eui[3] = 0xFE;
2345         memcpy(eui + 4, &addr, 4);
2346         return 0;
2347 }
2348
2349 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2350 {
2351         if (dev->priv_flags & IFF_ISATAP)
2352                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2353         return -1;
2354 }
2355
2356 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2357 {
2358         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2359 }
2360
2361 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2362 {
2363         memcpy(eui, dev->perm_addr, 3);
2364         memcpy(eui + 5, dev->perm_addr + 3, 3);
2365         eui[3] = 0xFF;
2366         eui[4] = 0xFE;
2367         eui[0] ^= 2;
2368         return 0;
2369 }
2370
2371 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2372 {
2373         switch (dev->type) {
2374         case ARPHRD_ETHER:
2375         case ARPHRD_FDDI:
2376                 return addrconf_ifid_eui48(eui, dev);
2377         case ARPHRD_ARCNET:
2378                 return addrconf_ifid_arcnet(eui, dev);
2379         case ARPHRD_INFINIBAND:
2380                 return addrconf_ifid_infiniband(eui, dev);
2381         case ARPHRD_SIT:
2382                 return addrconf_ifid_sit(eui, dev);
2383         case ARPHRD_IPGRE:
2384         case ARPHRD_TUNNEL:
2385                 return addrconf_ifid_gre(eui, dev);
2386         case ARPHRD_6LOWPAN:
2387                 return addrconf_ifid_6lowpan(eui, dev);
2388         case ARPHRD_IEEE1394:
2389                 return addrconf_ifid_ieee1394(eui, dev);
2390         case ARPHRD_TUNNEL6:
2391         case ARPHRD_IP6GRE:
2392         case ARPHRD_RAWIP:
2393                 return addrconf_ifid_ip6tnl(eui, dev);
2394         }
2395         return -1;
2396 }
2397
2398 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2399 {
2400         int err = -1;
2401         struct inet6_ifaddr *ifp;
2402
2403         read_lock_bh(&idev->lock);
2404         list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2405                 if (ifp->scope > IFA_LINK)
2406                         break;
2407                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2408                         memcpy(eui, ifp->addr.s6_addr+8, 8);
2409                         err = 0;
2410                         break;
2411                 }
2412         }
2413         read_unlock_bh(&idev->lock);
2414         return err;
2415 }
2416
2417 /* Generation of a randomized Interface Identifier
2418  * draft-ietf-6man-rfc4941bis, Section 3.3.1
2419  */
2420
2421 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2422 {
2423 regen:
2424         get_random_bytes(&addr->s6_addr[8], 8);
2425
2426         /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2427          * check if generated address is not inappropriate:
2428          *
2429          * - Reserved IPv6 Interface Identifiers
2430          * - XXX: already assigned to an address on the device
2431          */
2432
2433         /* Subnet-router anycast: 0000:0000:0000:0000 */
2434         if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2435                 goto regen;
2436
2437         /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2438          * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
2439          * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2440          */
2441         if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2442             (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2443                 goto regen;
2444
2445         /* Reserved subnet anycast addresses */
2446         if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2447             ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2448                 goto regen;
2449 }
2450
2451 /*
2452  *      Add prefix route.
2453  */
2454
2455 static void
2456 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2457                       struct net_device *dev, unsigned long expires,
2458                       u32 flags, gfp_t gfp_flags)
2459 {
2460         struct fib6_config cfg = {
2461                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2462                 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2463                 .fc_ifindex = dev->ifindex,
2464                 .fc_expires = expires,
2465                 .fc_dst_len = plen,
2466                 .fc_flags = RTF_UP | flags,
2467                 .fc_nlinfo.nl_net = dev_net(dev),
2468                 .fc_protocol = RTPROT_KERNEL,
2469                 .fc_type = RTN_UNICAST,
2470         };
2471
2472         cfg.fc_dst = *pfx;
2473
2474         /* Prevent useless cloning on PtP SIT.
2475            This thing is done here expecting that the whole
2476            class of non-broadcast devices need not cloning.
2477          */
2478 #if IS_ENABLED(CONFIG_IPV6_SIT)
2479         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2480                 cfg.fc_flags |= RTF_NONEXTHOP;
2481 #endif
2482
2483         ip6_route_add(&cfg, gfp_flags, NULL);
2484 }
2485
2486
2487 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2488                                                   int plen,
2489                                                   const struct net_device *dev,
2490                                                   u32 flags, u32 noflags,
2491                                                   bool no_gw)
2492 {
2493         struct fib6_node *fn;
2494         struct fib6_info *rt = NULL;
2495         struct fib6_table *table;
2496         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2497
2498         table = fib6_get_table(dev_net(dev), tb_id);
2499         if (!table)
2500                 return NULL;
2501
2502         rcu_read_lock();
2503         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2504         if (!fn)
2505                 goto out;
2506
2507         for_each_fib6_node_rt_rcu(fn) {
2508                 /* prefix routes only use builtin fib6_nh */
2509                 if (rt->nh)
2510                         continue;
2511
2512                 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2513                         continue;
2514                 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2515                         continue;
2516                 if ((rt->fib6_flags & flags) != flags)
2517                         continue;
2518                 if ((rt->fib6_flags & noflags) != 0)
2519                         continue;
2520                 if (!fib6_info_hold_safe(rt))
2521                         continue;
2522                 break;
2523         }
2524 out:
2525         rcu_read_unlock();
2526         return rt;
2527 }
2528
2529
2530 /* Create "default" multicast route to the interface */
2531
2532 static void addrconf_add_mroute(struct net_device *dev)
2533 {
2534         struct fib6_config cfg = {
2535                 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2536                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2537                 .fc_ifindex = dev->ifindex,
2538                 .fc_dst_len = 8,
2539                 .fc_flags = RTF_UP,
2540                 .fc_type = RTN_MULTICAST,
2541                 .fc_nlinfo.nl_net = dev_net(dev),
2542                 .fc_protocol = RTPROT_KERNEL,
2543         };
2544
2545         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2546
2547         ip6_route_add(&cfg, GFP_KERNEL, NULL);
2548 }
2549
2550 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2551 {
2552         struct inet6_dev *idev;
2553
2554         ASSERT_RTNL();
2555
2556         idev = ipv6_find_idev(dev);
2557         if (IS_ERR(idev))
2558                 return idev;
2559
2560         if (idev->cnf.disable_ipv6)
2561                 return ERR_PTR(-EACCES);
2562
2563         /* Add default multicast route */
2564         if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2565                 addrconf_add_mroute(dev);
2566
2567         return idev;
2568 }
2569
2570 static void manage_tempaddrs(struct inet6_dev *idev,
2571                              struct inet6_ifaddr *ifp,
2572                              __u32 valid_lft, __u32 prefered_lft,
2573                              bool create, unsigned long now)
2574 {
2575         u32 flags;
2576         struct inet6_ifaddr *ift;
2577
2578         read_lock_bh(&idev->lock);
2579         /* update all temporary addresses in the list */
2580         list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2581                 int age, max_valid, max_prefered;
2582
2583                 if (ifp != ift->ifpub)
2584                         continue;
2585
2586                 /* RFC 4941 section 3.3:
2587                  * If a received option will extend the lifetime of a public
2588                  * address, the lifetimes of temporary addresses should
2589                  * be extended, subject to the overall constraint that no
2590                  * temporary addresses should ever remain "valid" or "preferred"
2591                  * for a time longer than (TEMP_VALID_LIFETIME) or
2592                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2593                  */
2594                 age = (now - ift->cstamp) / HZ;
2595                 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2596                 if (max_valid < 0)
2597                         max_valid = 0;
2598
2599                 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2600                                idev->desync_factor - age;
2601                 if (max_prefered < 0)
2602                         max_prefered = 0;
2603
2604                 if (valid_lft > max_valid)
2605                         valid_lft = max_valid;
2606
2607                 if (prefered_lft > max_prefered)
2608                         prefered_lft = max_prefered;
2609
2610                 spin_lock(&ift->lock);
2611                 flags = ift->flags;
2612                 ift->valid_lft = valid_lft;
2613                 ift->prefered_lft = prefered_lft;
2614                 ift->tstamp = now;
2615                 if (prefered_lft > 0)
2616                         ift->flags &= ~IFA_F_DEPRECATED;
2617
2618                 spin_unlock(&ift->lock);
2619                 if (!(flags&IFA_F_TENTATIVE))
2620                         ipv6_ifa_notify(0, ift);
2621         }
2622
2623         /* Also create a temporary address if it's enabled but no temporary
2624          * address currently exists.
2625          * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2626          * as part of cleanup (ie. deleting the mngtmpaddr).
2627          * We don't want that to result in creating a new temporary ip address.
2628          */
2629         if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2630                 create = true;
2631
2632         if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2633                 /* When a new public address is created as described
2634                  * in [ADDRCONF], also create a new temporary address.
2635                  */
2636                 read_unlock_bh(&idev->lock);
2637                 ipv6_create_tempaddr(ifp, false);
2638         } else {
2639                 read_unlock_bh(&idev->lock);
2640         }
2641 }
2642
2643 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2644 {
2645         return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2646                idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2647 }
2648
2649 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2650                                  const struct prefix_info *pinfo,
2651                                  struct inet6_dev *in6_dev,
2652                                  const struct in6_addr *addr, int addr_type,
2653                                  u32 addr_flags, bool sllao, bool tokenized,
2654                                  __u32 valid_lft, u32 prefered_lft)
2655 {
2656         struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2657         int create = 0, update_lft = 0;
2658
2659         if (!ifp && valid_lft) {
2660                 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2661                 struct ifa6_config cfg = {
2662                         .pfx = addr,
2663                         .plen = pinfo->prefix_len,
2664                         .ifa_flags = addr_flags,
2665                         .valid_lft = valid_lft,
2666                         .preferred_lft = prefered_lft,
2667                         .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2668                         .ifa_proto = IFAPROT_KERNEL_RA
2669                 };
2670
2671 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2672                 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2673                      READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2674                     !net->ipv6.devconf_all->forwarding && sllao)
2675                         cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2676 #endif
2677
2678                 /* Do not allow to create too much of autoconfigured
2679                  * addresses; this would be too easy way to crash kernel.
2680                  */
2681                 if (!max_addresses ||
2682                     ipv6_count_addresses(in6_dev) < max_addresses)
2683                         ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2684
2685                 if (IS_ERR_OR_NULL(ifp))
2686                         return -1;
2687
2688                 create = 1;
2689                 spin_lock_bh(&ifp->lock);
2690                 ifp->flags |= IFA_F_MANAGETEMPADDR;
2691                 ifp->cstamp = jiffies;
2692                 ifp->tokenized = tokenized;
2693                 spin_unlock_bh(&ifp->lock);
2694                 addrconf_dad_start(ifp);
2695         }
2696
2697         if (ifp) {
2698                 u32 flags;
2699                 unsigned long now;
2700                 u32 stored_lft;
2701
2702                 /* update lifetime (RFC2462 5.5.3 e) */
2703                 spin_lock_bh(&ifp->lock);
2704                 now = jiffies;
2705                 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2706                         stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2707                 else
2708                         stored_lft = 0;
2709
2710                 /* RFC4862 Section 5.5.3e:
2711                  * "Note that the preferred lifetime of the
2712                  *  corresponding address is always reset to
2713                  *  the Preferred Lifetime in the received
2714                  *  Prefix Information option, regardless of
2715                  *  whether the valid lifetime is also reset or
2716                  *  ignored."
2717                  *
2718                  * So we should always update prefered_lft here.
2719                  */
2720                 update_lft = !create && stored_lft;
2721
2722                 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2723                         const u32 minimum_lft = min_t(u32,
2724                                 stored_lft, MIN_VALID_LIFETIME);
2725                         valid_lft = max(valid_lft, minimum_lft);
2726                 }
2727
2728                 if (update_lft) {
2729                         ifp->valid_lft = valid_lft;
2730                         ifp->prefered_lft = prefered_lft;
2731                         WRITE_ONCE(ifp->tstamp, now);
2732                         flags = ifp->flags;
2733                         ifp->flags &= ~IFA_F_DEPRECATED;
2734                         spin_unlock_bh(&ifp->lock);
2735
2736                         if (!(flags&IFA_F_TENTATIVE))
2737                                 ipv6_ifa_notify(0, ifp);
2738                 } else
2739                         spin_unlock_bh(&ifp->lock);
2740
2741                 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2742                                  create, now);
2743
2744                 in6_ifa_put(ifp);
2745                 addrconf_verify(net);
2746         }
2747
2748         return 0;
2749 }
2750 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2751
2752 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2753 {
2754         struct prefix_info *pinfo;
2755         struct fib6_table *table;
2756         __u32 valid_lft;
2757         __u32 prefered_lft;
2758         int addr_type, err;
2759         u32 addr_flags = 0;
2760         struct inet6_dev *in6_dev;
2761         struct net *net = dev_net(dev);
2762
2763         pinfo = (struct prefix_info *) opt;
2764
2765         if (len < sizeof(struct prefix_info)) {
2766                 netdev_dbg(dev, "addrconf: prefix option too short\n");
2767                 return;
2768         }
2769
2770         /*
2771          *      Validation checks ([ADDRCONF], page 19)
2772          */
2773
2774         addr_type = ipv6_addr_type(&pinfo->prefix);
2775
2776         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2777                 return;
2778
2779         valid_lft = ntohl(pinfo->valid);
2780         prefered_lft = ntohl(pinfo->prefered);
2781
2782         if (prefered_lft > valid_lft) {
2783                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2784                 return;
2785         }
2786
2787         in6_dev = in6_dev_get(dev);
2788
2789         if (!in6_dev) {
2790                 net_dbg_ratelimited("addrconf: device %s not configured\n",
2791                                     dev->name);
2792                 return;
2793         }
2794
2795         if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2796                 goto put;
2797
2798         /*
2799          *      Two things going on here:
2800          *      1) Add routes for on-link prefixes
2801          *      2) Configure prefixes with the auto flag set
2802          */
2803
2804         if (pinfo->onlink) {
2805                 struct fib6_info *rt;
2806                 unsigned long rt_expires;
2807
2808                 /* Avoid arithmetic overflow. Really, we could
2809                  * save rt_expires in seconds, likely valid_lft,
2810                  * but it would require division in fib gc, that it
2811                  * not good.
2812                  */
2813                 if (HZ > USER_HZ)
2814                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2815                 else
2816                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2817
2818                 if (addrconf_finite_timeout(rt_expires))
2819                         rt_expires *= HZ;
2820
2821                 rt = addrconf_get_prefix_route(&pinfo->prefix,
2822                                                pinfo->prefix_len,
2823                                                dev,
2824                                                RTF_ADDRCONF | RTF_PREFIX_RT,
2825                                                RTF_DEFAULT, true);
2826
2827                 if (rt) {
2828                         /* Autoconf prefix route */
2829                         if (valid_lft == 0) {
2830                                 ip6_del_rt(net, rt, false);
2831                                 rt = NULL;
2832                         } else {
2833                                 table = rt->fib6_table;
2834                                 spin_lock_bh(&table->tb6_lock);
2835
2836                                 if (addrconf_finite_timeout(rt_expires)) {
2837                                         /* not infinity */
2838                                         fib6_set_expires(rt, jiffies + rt_expires);
2839                                         fib6_add_gc_list(rt);
2840                                 } else {
2841                                         fib6_clean_expires(rt);
2842                                         fib6_remove_gc_list(rt);
2843                                 }
2844
2845                                 spin_unlock_bh(&table->tb6_lock);
2846                         }
2847                 } else if (valid_lft) {
2848                         clock_t expires = 0;
2849                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2850                         if (addrconf_finite_timeout(rt_expires)) {
2851                                 /* not infinity */
2852                                 flags |= RTF_EXPIRES;
2853                                 expires = jiffies_to_clock_t(rt_expires);
2854                         }
2855                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2856                                               0, dev, expires, flags,
2857                                               GFP_ATOMIC);
2858                 }
2859                 fib6_info_release(rt);
2860         }
2861
2862         /* Try to figure out our local address for this prefix */
2863
2864         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2865                 struct in6_addr addr;
2866                 bool tokenized = false, dev_addr_generated = false;
2867
2868                 if (pinfo->prefix_len == 64) {
2869                         memcpy(&addr, &pinfo->prefix, 8);
2870
2871                         if (!ipv6_addr_any(&in6_dev->token)) {
2872                                 read_lock_bh(&in6_dev->lock);
2873                                 memcpy(addr.s6_addr + 8,
2874                                        in6_dev->token.s6_addr + 8, 8);
2875                                 read_unlock_bh(&in6_dev->lock);
2876                                 tokenized = true;
2877                         } else if (is_addr_mode_generate_stable(in6_dev) &&
2878                                    !ipv6_generate_stable_address(&addr, 0,
2879                                                                  in6_dev)) {
2880                                 addr_flags |= IFA_F_STABLE_PRIVACY;
2881                                 goto ok;
2882                         } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2883                                    ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2884                                 goto put;
2885                         } else {
2886                                 dev_addr_generated = true;
2887                         }
2888                         goto ok;
2889                 }
2890                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2891                                     pinfo->prefix_len);
2892                 goto put;
2893
2894 ok:
2895                 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2896                                                    &addr, addr_type,
2897                                                    addr_flags, sllao,
2898                                                    tokenized, valid_lft,
2899                                                    prefered_lft);
2900                 if (err)
2901                         goto put;
2902
2903                 /* Ignore error case here because previous prefix add addr was
2904                  * successful which will be notified.
2905                  */
2906                 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2907                                               addr_type, addr_flags, sllao,
2908                                               tokenized, valid_lft,
2909                                               prefered_lft,
2910                                               dev_addr_generated);
2911         }
2912         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2913 put:
2914         in6_dev_put(in6_dev);
2915 }
2916
2917 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2918                 struct in6_ifreq *ireq)
2919 {
2920         struct ip_tunnel_parm p = { };
2921         int err;
2922
2923         if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2924                 return -EADDRNOTAVAIL;
2925
2926         p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2927         p.iph.version = 4;
2928         p.iph.ihl = 5;
2929         p.iph.protocol = IPPROTO_IPV6;
2930         p.iph.ttl = 64;
2931
2932         if (!dev->netdev_ops->ndo_tunnel_ctl)
2933                 return -EOPNOTSUPP;
2934         err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2935         if (err)
2936                 return err;
2937
2938         dev = __dev_get_by_name(net, p.name);
2939         if (!dev)
2940                 return -ENOBUFS;
2941         return dev_open(dev, NULL);
2942 }
2943
2944 /*
2945  *      Set destination address.
2946  *      Special case for SIT interfaces where we create a new "virtual"
2947  *      device.
2948  */
2949 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2950 {
2951         struct net_device *dev;
2952         struct in6_ifreq ireq;
2953         int err = -ENODEV;
2954
2955         if (!IS_ENABLED(CONFIG_IPV6_SIT))
2956                 return -ENODEV;
2957         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2958                 return -EFAULT;
2959
2960         rtnl_lock();
2961         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2962         if (dev && dev->type == ARPHRD_SIT)
2963                 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2964         rtnl_unlock();
2965         return err;
2966 }
2967
2968 static int ipv6_mc_config(struct sock *sk, bool join,
2969                           const struct in6_addr *addr, int ifindex)
2970 {
2971         int ret;
2972
2973         ASSERT_RTNL();
2974
2975         lock_sock(sk);
2976         if (join)
2977                 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2978         else
2979                 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2980         release_sock(sk);
2981
2982         return ret;
2983 }
2984
2985 /*
2986  *      Manual configuration of address on an interface
2987  */
2988 static int inet6_addr_add(struct net *net, int ifindex,
2989                           struct ifa6_config *cfg,
2990                           struct netlink_ext_ack *extack)
2991 {
2992         struct inet6_ifaddr *ifp;
2993         struct inet6_dev *idev;
2994         struct net_device *dev;
2995         unsigned long timeout;
2996         clock_t expires;
2997         u32 flags;
2998
2999         ASSERT_RTNL();
3000
3001         if (cfg->plen > 128) {
3002                 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3003                 return -EINVAL;
3004         }
3005
3006         /* check the lifetime */
3007         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
3008                 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
3009                 return -EINVAL;
3010         }
3011
3012         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3013                 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3014                 return -EINVAL;
3015         }
3016
3017         dev = __dev_get_by_index(net, ifindex);
3018         if (!dev)
3019                 return -ENODEV;
3020
3021         idev = addrconf_add_dev(dev);
3022         if (IS_ERR(idev)) {
3023                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3024                 return PTR_ERR(idev);
3025         }
3026
3027         if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3028                 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3029                                          true, cfg->pfx, ifindex);
3030
3031                 if (ret < 0) {
3032                         NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3033                         return ret;
3034                 }
3035         }
3036
3037         cfg->scope = ipv6_addr_scope(cfg->pfx);
3038
3039         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
3040         if (addrconf_finite_timeout(timeout)) {
3041                 expires = jiffies_to_clock_t(timeout * HZ);
3042                 cfg->valid_lft = timeout;
3043                 flags = RTF_EXPIRES;
3044         } else {
3045                 expires = 0;
3046                 flags = 0;
3047                 cfg->ifa_flags |= IFA_F_PERMANENT;
3048         }
3049
3050         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3051         if (addrconf_finite_timeout(timeout)) {
3052                 if (timeout == 0)
3053                         cfg->ifa_flags |= IFA_F_DEPRECATED;
3054                 cfg->preferred_lft = timeout;
3055         }
3056
3057         ifp = ipv6_add_addr(idev, cfg, true, extack);
3058         if (!IS_ERR(ifp)) {
3059                 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3060                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3061                                               ifp->rt_priority, dev, expires,
3062                                               flags, GFP_KERNEL);
3063                 }
3064
3065                 /* Send a netlink notification if DAD is enabled and
3066                  * optimistic flag is not set
3067                  */
3068                 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3069                         ipv6_ifa_notify(0, ifp);
3070                 /*
3071                  * Note that section 3.1 of RFC 4429 indicates
3072                  * that the Optimistic flag should not be set for
3073                  * manually configured addresses
3074                  */
3075                 addrconf_dad_start(ifp);
3076                 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3077                         manage_tempaddrs(idev, ifp, cfg->valid_lft,
3078                                          cfg->preferred_lft, true, jiffies);
3079                 in6_ifa_put(ifp);
3080                 addrconf_verify_rtnl(net);
3081                 return 0;
3082         } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3083                 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3084                                cfg->pfx, ifindex);
3085         }
3086
3087         return PTR_ERR(ifp);
3088 }
3089
3090 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3091                           const struct in6_addr *pfx, unsigned int plen,
3092                           struct netlink_ext_ack *extack)
3093 {
3094         struct inet6_ifaddr *ifp;
3095         struct inet6_dev *idev;
3096         struct net_device *dev;
3097
3098         if (plen > 128) {
3099                 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3100                 return -EINVAL;
3101         }
3102
3103         dev = __dev_get_by_index(net, ifindex);
3104         if (!dev) {
3105                 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3106                 return -ENODEV;
3107         }
3108
3109         idev = __in6_dev_get(dev);
3110         if (!idev) {
3111                 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3112                 return -ENXIO;
3113         }
3114
3115         read_lock_bh(&idev->lock);
3116         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3117                 if (ifp->prefix_len == plen &&
3118                     ipv6_addr_equal(pfx, &ifp->addr)) {
3119                         in6_ifa_hold(ifp);
3120                         read_unlock_bh(&idev->lock);
3121
3122                         if (!(ifp->flags & IFA_F_TEMPORARY) &&
3123                             (ifa_flags & IFA_F_MANAGETEMPADDR))
3124                                 manage_tempaddrs(idev, ifp, 0, 0, false,
3125                                                  jiffies);
3126                         ipv6_del_addr(ifp);
3127                         addrconf_verify_rtnl(net);
3128                         if (ipv6_addr_is_multicast(pfx)) {
3129                                 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3130                                                false, pfx, dev->ifindex);
3131                         }
3132                         return 0;
3133                 }
3134         }
3135         read_unlock_bh(&idev->lock);
3136
3137         NL_SET_ERR_MSG_MOD(extack, "address not found");
3138         return -EADDRNOTAVAIL;
3139 }
3140
3141
3142 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3143 {
3144         struct ifa6_config cfg = {
3145                 .ifa_flags = IFA_F_PERMANENT,
3146                 .preferred_lft = INFINITY_LIFE_TIME,
3147                 .valid_lft = INFINITY_LIFE_TIME,
3148         };
3149         struct in6_ifreq ireq;
3150         int err;
3151
3152         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3153                 return -EPERM;
3154
3155         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3156                 return -EFAULT;
3157
3158         cfg.pfx = &ireq.ifr6_addr;
3159         cfg.plen = ireq.ifr6_prefixlen;
3160
3161         rtnl_lock();
3162         err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3163         rtnl_unlock();
3164         return err;
3165 }
3166
3167 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3168 {
3169         struct in6_ifreq ireq;
3170         int err;
3171
3172         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3173                 return -EPERM;
3174
3175         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3176                 return -EFAULT;
3177
3178         rtnl_lock();
3179         err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3180                              ireq.ifr6_prefixlen, NULL);
3181         rtnl_unlock();
3182         return err;
3183 }
3184
3185 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3186                      int plen, int scope, u8 proto)
3187 {
3188         struct inet6_ifaddr *ifp;
3189         struct ifa6_config cfg = {
3190                 .pfx = addr,
3191                 .plen = plen,
3192                 .ifa_flags = IFA_F_PERMANENT,
3193                 .valid_lft = INFINITY_LIFE_TIME,
3194                 .preferred_lft = INFINITY_LIFE_TIME,
3195                 .scope = scope,
3196                 .ifa_proto = proto
3197         };
3198
3199         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3200         if (!IS_ERR(ifp)) {
3201                 spin_lock_bh(&ifp->lock);
3202                 ifp->flags &= ~IFA_F_TENTATIVE;
3203                 spin_unlock_bh(&ifp->lock);
3204                 rt_genid_bump_ipv6(dev_net(idev->dev));
3205                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3206                 in6_ifa_put(ifp);
3207         }
3208 }
3209
3210 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3211 static void add_v4_addrs(struct inet6_dev *idev)
3212 {
3213         struct in6_addr addr;
3214         struct net_device *dev;
3215         struct net *net = dev_net(idev->dev);
3216         int scope, plen, offset = 0;
3217         u32 pflags = 0;
3218
3219         ASSERT_RTNL();
3220
3221         memset(&addr, 0, sizeof(struct in6_addr));
3222         /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3223         if (idev->dev->addr_len == sizeof(struct in6_addr))
3224                 offset = sizeof(struct in6_addr) - 4;
3225         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3226
3227         if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3228                 scope = IPV6_ADDR_COMPATv4;
3229                 plen = 96;
3230                 pflags |= RTF_NONEXTHOP;
3231         } else {
3232                 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3233                         return;
3234
3235                 addr.s6_addr32[0] = htonl(0xfe800000);
3236                 scope = IFA_LINK;
3237                 plen = 64;
3238         }
3239
3240         if (addr.s6_addr32[3]) {
3241                 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3242                 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3243                                       GFP_KERNEL);
3244                 return;
3245         }
3246
3247         for_each_netdev(net, dev) {
3248                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3249                 if (in_dev && (dev->flags & IFF_UP)) {
3250                         struct in_ifaddr *ifa;
3251                         int flag = scope;
3252
3253                         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3254                                 addr.s6_addr32[3] = ifa->ifa_local;
3255
3256                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
3257                                         continue;
3258                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3259                                         if (idev->dev->flags&IFF_POINTOPOINT)
3260                                                 continue;
3261                                         flag |= IFA_HOST;
3262                                 }
3263
3264                                 add_addr(idev, &addr, plen, flag,
3265                                          IFAPROT_UNSPEC);
3266                                 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3267                                                       0, pflags, GFP_KERNEL);
3268                         }
3269                 }
3270         }
3271 }
3272 #endif
3273
3274 static void init_loopback(struct net_device *dev)
3275 {
3276         struct inet6_dev  *idev;
3277
3278         /* ::1 */
3279
3280         ASSERT_RTNL();
3281
3282         idev = ipv6_find_idev(dev);
3283         if (IS_ERR(idev)) {
3284                 pr_debug("%s: add_dev failed\n", __func__);
3285                 return;
3286         }
3287
3288         add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3289 }
3290
3291 void addrconf_add_linklocal(struct inet6_dev *idev,
3292                             const struct in6_addr *addr, u32 flags)
3293 {
3294         struct ifa6_config cfg = {
3295                 .pfx = addr,
3296                 .plen = 64,
3297                 .ifa_flags = flags | IFA_F_PERMANENT,
3298                 .valid_lft = INFINITY_LIFE_TIME,
3299                 .preferred_lft = INFINITY_LIFE_TIME,
3300                 .scope = IFA_LINK,
3301                 .ifa_proto = IFAPROT_KERNEL_LL
3302         };
3303         struct inet6_ifaddr *ifp;
3304
3305 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3306         if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3307              READ_ONCE(idev->cnf.optimistic_dad)) &&
3308             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3309                 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3310 #endif
3311
3312         ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3313         if (!IS_ERR(ifp)) {
3314                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3315                                       0, 0, GFP_ATOMIC);
3316                 addrconf_dad_start(ifp);
3317                 in6_ifa_put(ifp);
3318         }
3319 }
3320 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3321
3322 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3323 {
3324         if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3325                 return true;
3326
3327         if (address.s6_addr32[2] == htonl(0x02005eff) &&
3328             ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3329                 return true;
3330
3331         if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3332             ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3333                 return true;
3334
3335         return false;
3336 }
3337
3338 static int ipv6_generate_stable_address(struct in6_addr *address,
3339                                         u8 dad_count,
3340                                         const struct inet6_dev *idev)
3341 {
3342         static DEFINE_SPINLOCK(lock);
3343         static __u32 digest[SHA1_DIGEST_WORDS];
3344         static __u32 workspace[SHA1_WORKSPACE_WORDS];
3345
3346         static union {
3347                 char __data[SHA1_BLOCK_SIZE];
3348                 struct {
3349                         struct in6_addr secret;
3350                         __be32 prefix[2];
3351                         unsigned char hwaddr[MAX_ADDR_LEN];
3352                         u8 dad_count;
3353                 } __packed;
3354         } data;
3355
3356         struct in6_addr secret;
3357         struct in6_addr temp;
3358         struct net *net = dev_net(idev->dev);
3359
3360         BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3361
3362         if (idev->cnf.stable_secret.initialized)
3363                 secret = idev->cnf.stable_secret.secret;
3364         else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3365                 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3366         else
3367                 return -1;
3368
3369 retry:
3370         spin_lock_bh(&lock);
3371
3372         sha1_init(digest);
3373         memset(&data, 0, sizeof(data));
3374         memset(workspace, 0, sizeof(workspace));
3375         memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3376         data.prefix[0] = address->s6_addr32[0];
3377         data.prefix[1] = address->s6_addr32[1];
3378         data.secret = secret;
3379         data.dad_count = dad_count;
3380
3381         sha1_transform(digest, data.__data, workspace);
3382
3383         temp = *address;
3384         temp.s6_addr32[2] = (__force __be32)digest[0];
3385         temp.s6_addr32[3] = (__force __be32)digest[1];
3386
3387         spin_unlock_bh(&lock);
3388
3389         if (ipv6_reserved_interfaceid(temp)) {
3390                 dad_count++;
3391                 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3392                         return -1;
3393                 goto retry;
3394         }
3395
3396         *address = temp;
3397         return 0;
3398 }
3399
3400 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3401 {
3402         struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3403
3404         if (s->initialized)
3405                 return;
3406         s = &idev->cnf.stable_secret;
3407         get_random_bytes(&s->secret, sizeof(s->secret));
3408         s->initialized = true;
3409 }
3410
3411 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3412 {
3413         struct in6_addr addr;
3414
3415         /* no link local addresses on L3 master devices */
3416         if (netif_is_l3_master(idev->dev))
3417                 return;
3418
3419         /* no link local addresses on devices flagged as slaves */
3420         if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3421                 return;
3422
3423         ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3424
3425         switch (idev->cnf.addr_gen_mode) {
3426         case IN6_ADDR_GEN_MODE_RANDOM:
3427                 ipv6_gen_mode_random_init(idev);
3428                 fallthrough;
3429         case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3430                 if (!ipv6_generate_stable_address(&addr, 0, idev))
3431                         addrconf_add_linklocal(idev, &addr,
3432                                                IFA_F_STABLE_PRIVACY);
3433                 else if (prefix_route)
3434                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3435                                               0, 0, GFP_KERNEL);
3436                 break;
3437         case IN6_ADDR_GEN_MODE_EUI64:
3438                 /* addrconf_add_linklocal also adds a prefix_route and we
3439                  * only need to care about prefix routes if ipv6_generate_eui64
3440                  * couldn't generate one.
3441                  */
3442                 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3443                         addrconf_add_linklocal(idev, &addr, 0);
3444                 else if (prefix_route)
3445                         addrconf_prefix_route(&addr, 64, 0, idev->dev,
3446                                               0, 0, GFP_KERNEL);
3447                 break;
3448         case IN6_ADDR_GEN_MODE_NONE:
3449         default:
3450                 /* will not add any link local address */
3451                 break;
3452         }
3453 }
3454
3455 static void addrconf_dev_config(struct net_device *dev)
3456 {
3457         struct inet6_dev *idev;
3458
3459         ASSERT_RTNL();
3460
3461         if ((dev->type != ARPHRD_ETHER) &&
3462             (dev->type != ARPHRD_FDDI) &&
3463             (dev->type != ARPHRD_ARCNET) &&
3464             (dev->type != ARPHRD_INFINIBAND) &&
3465             (dev->type != ARPHRD_IEEE1394) &&
3466             (dev->type != ARPHRD_TUNNEL6) &&
3467             (dev->type != ARPHRD_6LOWPAN) &&
3468             (dev->type != ARPHRD_TUNNEL) &&
3469             (dev->type != ARPHRD_NONE) &&
3470             (dev->type != ARPHRD_RAWIP)) {
3471                 /* Alas, we support only Ethernet autoconfiguration. */
3472                 idev = __in6_dev_get(dev);
3473                 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3474                     dev->flags & IFF_MULTICAST)
3475                         ipv6_mc_up(idev);
3476                 return;
3477         }
3478
3479         idev = addrconf_add_dev(dev);
3480         if (IS_ERR(idev))
3481                 return;
3482
3483         /* this device type has no EUI support */
3484         if (dev->type == ARPHRD_NONE &&
3485             idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3486                 WRITE_ONCE(idev->cnf.addr_gen_mode,
3487                            IN6_ADDR_GEN_MODE_RANDOM);
3488
3489         addrconf_addr_gen(idev, false);
3490 }
3491
3492 #if IS_ENABLED(CONFIG_IPV6_SIT)
3493 static void addrconf_sit_config(struct net_device *dev)
3494 {
3495         struct inet6_dev *idev;
3496
3497         ASSERT_RTNL();
3498
3499         /*
3500          * Configure the tunnel with one of our IPv4
3501          * addresses... we should configure all of
3502          * our v4 addrs in the tunnel
3503          */
3504
3505         idev = ipv6_find_idev(dev);
3506         if (IS_ERR(idev)) {
3507                 pr_debug("%s: add_dev failed\n", __func__);
3508                 return;
3509         }
3510
3511         if (dev->priv_flags & IFF_ISATAP) {
3512                 addrconf_addr_gen(idev, false);
3513                 return;
3514         }
3515
3516         add_v4_addrs(idev);
3517
3518         if (dev->flags&IFF_POINTOPOINT)
3519                 addrconf_add_mroute(dev);
3520 }
3521 #endif
3522
3523 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3524 static void addrconf_gre_config(struct net_device *dev)
3525 {
3526         struct inet6_dev *idev;
3527
3528         ASSERT_RTNL();
3529
3530         idev = ipv6_find_idev(dev);
3531         if (IS_ERR(idev)) {
3532                 pr_debug("%s: add_dev failed\n", __func__);
3533                 return;
3534         }
3535
3536         if (dev->type == ARPHRD_ETHER) {
3537                 addrconf_addr_gen(idev, true);
3538                 return;
3539         }
3540
3541         add_v4_addrs(idev);
3542
3543         if (dev->flags & IFF_POINTOPOINT)
3544                 addrconf_add_mroute(dev);
3545 }
3546 #endif
3547
3548 static void addrconf_init_auto_addrs(struct net_device *dev)
3549 {
3550         switch (dev->type) {
3551 #if IS_ENABLED(CONFIG_IPV6_SIT)
3552         case ARPHRD_SIT:
3553                 addrconf_sit_config(dev);
3554                 break;
3555 #endif
3556 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3557         case ARPHRD_IP6GRE:
3558         case ARPHRD_IPGRE:
3559                 addrconf_gre_config(dev);
3560                 break;
3561 #endif
3562         case ARPHRD_LOOPBACK:
3563                 init_loopback(dev);
3564                 break;
3565
3566         default:
3567                 addrconf_dev_config(dev);
3568                 break;
3569         }
3570 }
3571
3572 static int fixup_permanent_addr(struct net *net,
3573                                 struct inet6_dev *idev,
3574                                 struct inet6_ifaddr *ifp)
3575 {
3576         /* !fib6_node means the host route was removed from the
3577          * FIB, for example, if 'lo' device is taken down. In that
3578          * case regenerate the host route.
3579          */
3580         if (!ifp->rt || !ifp->rt->fib6_node) {
3581                 struct fib6_info *f6i, *prev;
3582
3583                 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3584                                          GFP_ATOMIC, NULL);
3585                 if (IS_ERR(f6i))
3586                         return PTR_ERR(f6i);
3587
3588                 /* ifp->rt can be accessed outside of rtnl */
3589                 spin_lock(&ifp->lock);
3590                 prev = ifp->rt;
3591                 ifp->rt = f6i;
3592                 spin_unlock(&ifp->lock);
3593
3594                 fib6_info_release(prev);
3595         }
3596
3597         if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3598                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3599                                       ifp->rt_priority, idev->dev, 0, 0,
3600                                       GFP_ATOMIC);
3601         }
3602
3603         if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3604                 addrconf_dad_start(ifp);
3605
3606         return 0;
3607 }
3608
3609 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3610 {
3611         struct inet6_ifaddr *ifp, *tmp;
3612         struct inet6_dev *idev;
3613
3614         idev = __in6_dev_get(dev);
3615         if (!idev)
3616                 return;
3617
3618         write_lock_bh(&idev->lock);
3619
3620         list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3621                 if ((ifp->flags & IFA_F_PERMANENT) &&
3622                     fixup_permanent_addr(net, idev, ifp) < 0) {
3623                         write_unlock_bh(&idev->lock);
3624                         in6_ifa_hold(ifp);
3625                         ipv6_del_addr(ifp);
3626                         write_lock_bh(&idev->lock);
3627
3628                         net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3629                                              idev->dev->name, &ifp->addr);
3630                 }
3631         }
3632
3633         write_unlock_bh(&idev->lock);
3634 }
3635
3636 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3637                            void *ptr)
3638 {
3639         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3640         struct netdev_notifier_change_info *change_info;
3641         struct netdev_notifier_changeupper_info *info;
3642         struct inet6_dev *idev = __in6_dev_get(dev);
3643         struct net *net = dev_net(dev);
3644         int run_pending = 0;
3645         int err;
3646
3647         switch (event) {
3648         case NETDEV_REGISTER:
3649                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3650                         idev = ipv6_add_dev(dev);
3651                         if (IS_ERR(idev))
3652                                 return notifier_from_errno(PTR_ERR(idev));
3653                 }
3654                 break;
3655
3656         case NETDEV_CHANGEMTU:
3657                 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3658                 if (dev->mtu < IPV6_MIN_MTU) {
3659                         addrconf_ifdown(dev, dev != net->loopback_dev);
3660                         break;
3661                 }
3662
3663                 if (idev) {
3664                         rt6_mtu_change(dev, dev->mtu);
3665                         WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3666                         break;
3667                 }
3668
3669                 /* allocate new idev */
3670                 idev = ipv6_add_dev(dev);
3671                 if (IS_ERR(idev))
3672                         break;
3673
3674                 /* device is still not ready */
3675                 if (!(idev->if_flags & IF_READY))
3676                         break;
3677
3678                 run_pending = 1;
3679                 fallthrough;
3680         case NETDEV_UP:
3681         case NETDEV_CHANGE:
3682                 if (idev && idev->cnf.disable_ipv6)
3683                         break;
3684
3685                 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3686                         if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3687                             dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3688                                 ipv6_mc_up(idev);
3689                         break;
3690                 }
3691
3692                 if (event == NETDEV_UP) {
3693                         /* restore routes for permanent addresses */
3694                         addrconf_permanent_addr(net, dev);
3695
3696                         if (!addrconf_link_ready(dev)) {
3697                                 /* device is not ready yet. */
3698                                 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3699                                          dev->name);
3700                                 break;
3701                         }
3702
3703                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
3704                                 idev = ipv6_add_dev(dev);
3705
3706                         if (!IS_ERR_OR_NULL(idev)) {
3707                                 idev->if_flags |= IF_READY;
3708                                 run_pending = 1;
3709                         }
3710                 } else if (event == NETDEV_CHANGE) {
3711                         if (!addrconf_link_ready(dev)) {
3712                                 /* device is still not ready. */
3713                                 rt6_sync_down_dev(dev, event);
3714                                 break;
3715                         }
3716
3717                         if (!IS_ERR_OR_NULL(idev)) {
3718                                 if (idev->if_flags & IF_READY) {
3719                                         /* device is already configured -
3720                                          * but resend MLD reports, we might
3721                                          * have roamed and need to update
3722                                          * multicast snooping switches
3723                                          */
3724                                         ipv6_mc_up(idev);
3725                                         change_info = ptr;
3726                                         if (change_info->flags_changed & IFF_NOARP)
3727                                                 addrconf_dad_run(idev, true);
3728                                         rt6_sync_up(dev, RTNH_F_LINKDOWN);
3729                                         break;
3730                                 }
3731                                 idev->if_flags |= IF_READY;
3732                         }
3733
3734                         pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3735                                  dev->name);
3736
3737                         run_pending = 1;
3738                 }
3739
3740                 addrconf_init_auto_addrs(dev);
3741
3742                 if (!IS_ERR_OR_NULL(idev)) {
3743                         if (run_pending)
3744                                 addrconf_dad_run(idev, false);
3745
3746                         /* Device has an address by now */
3747                         rt6_sync_up(dev, RTNH_F_DEAD);
3748
3749                         /*
3750                          * If the MTU changed during the interface down,
3751                          * when the interface up, the changed MTU must be
3752                          * reflected in the idev as well as routers.
3753                          */
3754                         if (idev->cnf.mtu6 != dev->mtu &&
3755                             dev->mtu >= IPV6_MIN_MTU) {
3756                                 rt6_mtu_change(dev, dev->mtu);
3757                                 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3758                         }
3759                         WRITE_ONCE(idev->tstamp, jiffies);
3760                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
3761
3762                         /*
3763                          * If the changed mtu during down is lower than
3764                          * IPV6_MIN_MTU stop IPv6 on this interface.
3765                          */
3766                         if (dev->mtu < IPV6_MIN_MTU)
3767                                 addrconf_ifdown(dev, dev != net->loopback_dev);
3768                 }
3769                 break;
3770
3771         case NETDEV_DOWN:
3772         case NETDEV_UNREGISTER:
3773                 /*
3774                  *      Remove all addresses from this interface.
3775                  */
3776                 addrconf_ifdown(dev, event != NETDEV_DOWN);
3777                 break;
3778
3779         case NETDEV_CHANGENAME:
3780                 if (idev) {
3781                         snmp6_unregister_dev(idev);
3782                         addrconf_sysctl_unregister(idev);
3783                         err = addrconf_sysctl_register(idev);
3784                         if (err)
3785                                 return notifier_from_errno(err);
3786                         err = snmp6_register_dev(idev);
3787                         if (err) {
3788                                 addrconf_sysctl_unregister(idev);
3789                                 return notifier_from_errno(err);
3790                         }
3791                 }
3792                 break;
3793
3794         case NETDEV_PRE_TYPE_CHANGE:
3795         case NETDEV_POST_TYPE_CHANGE:
3796                 if (idev)
3797                         addrconf_type_change(dev, event);
3798                 break;
3799
3800         case NETDEV_CHANGEUPPER:
3801                 info = ptr;
3802
3803                 /* flush all routes if dev is linked to or unlinked from
3804                  * an L3 master device (e.g., VRF)
3805                  */
3806                 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3807                         addrconf_ifdown(dev, false);
3808         }
3809
3810         return NOTIFY_OK;
3811 }
3812
3813 /*
3814  *      addrconf module should be notified of a device going up
3815  */
3816 static struct notifier_block ipv6_dev_notf = {
3817         .notifier_call = addrconf_notify,
3818         .priority = ADDRCONF_NOTIFY_PRIORITY,
3819 };
3820
3821 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3822 {
3823         struct inet6_dev *idev;
3824         ASSERT_RTNL();
3825
3826         idev = __in6_dev_get(dev);
3827
3828         if (event == NETDEV_POST_TYPE_CHANGE)
3829                 ipv6_mc_remap(idev);
3830         else if (event == NETDEV_PRE_TYPE_CHANGE)
3831                 ipv6_mc_unmap(idev);
3832 }
3833
3834 static bool addr_is_local(const struct in6_addr *addr)
3835 {
3836         return ipv6_addr_type(addr) &
3837                 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3838 }
3839
3840 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3841 {
3842         unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3843         struct net *net = dev_net(dev);
3844         struct inet6_dev *idev;
3845         struct inet6_ifaddr *ifa;
3846         LIST_HEAD(tmp_addr_list);
3847         bool keep_addr = false;
3848         bool was_ready;
3849         int state, i;
3850
3851         ASSERT_RTNL();
3852
3853         rt6_disable_ip(dev, event);
3854
3855         idev = __in6_dev_get(dev);
3856         if (!idev)
3857                 return -ENODEV;
3858
3859         /*
3860          * Step 1: remove reference to ipv6 device from parent device.
3861          *         Do not dev_put!
3862          */
3863         if (unregister) {
3864                 idev->dead = 1;
3865
3866                 /* protected by rtnl_lock */
3867                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3868
3869                 /* Step 1.5: remove snmp6 entry */
3870                 snmp6_unregister_dev(idev);
3871
3872         }
3873
3874         /* combine the user config with event to determine if permanent
3875          * addresses are to be removed from address hash table
3876          */
3877         if (!unregister && !idev->cnf.disable_ipv6) {
3878                 /* aggregate the system setting and interface setting */
3879                 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3880
3881                 if (!_keep_addr)
3882                         _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3883
3884                 keep_addr = (_keep_addr > 0);
3885         }
3886
3887         /* Step 2: clear hash table */
3888         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3889                 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3890
3891                 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3892 restart:
3893                 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3894                         if (ifa->idev == idev) {
3895                                 addrconf_del_dad_work(ifa);
3896                                 /* combined flag + permanent flag decide if
3897                                  * address is retained on a down event
3898                                  */
3899                                 if (!keep_addr ||
3900                                     !(ifa->flags & IFA_F_PERMANENT) ||
3901                                     addr_is_local(&ifa->addr)) {
3902                                         hlist_del_init_rcu(&ifa->addr_lst);
3903                                         goto restart;
3904                                 }
3905                         }
3906                 }
3907                 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3908         }
3909
3910         write_lock_bh(&idev->lock);
3911
3912         addrconf_del_rs_timer(idev);
3913
3914         /* Step 2: clear flags for stateless addrconf, repeated down
3915          *         detection
3916          */
3917         was_ready = idev->if_flags & IF_READY;
3918         if (!unregister)
3919                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3920
3921         /* Step 3: clear tempaddr list */
3922         while (!list_empty(&idev->tempaddr_list)) {
3923                 ifa = list_first_entry(&idev->tempaddr_list,
3924                                        struct inet6_ifaddr, tmp_list);
3925                 list_del(&ifa->tmp_list);
3926                 write_unlock_bh(&idev->lock);
3927                 spin_lock_bh(&ifa->lock);
3928
3929                 if (ifa->ifpub) {
3930                         in6_ifa_put(ifa->ifpub);
3931                         ifa->ifpub = NULL;
3932                 }
3933                 spin_unlock_bh(&ifa->lock);
3934                 in6_ifa_put(ifa);
3935                 write_lock_bh(&idev->lock);
3936         }
3937
3938         list_for_each_entry(ifa, &idev->addr_list, if_list)
3939                 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3940         write_unlock_bh(&idev->lock);
3941
3942         while (!list_empty(&tmp_addr_list)) {
3943                 struct fib6_info *rt = NULL;
3944                 bool keep;
3945
3946                 ifa = list_first_entry(&tmp_addr_list,
3947                                        struct inet6_ifaddr, if_list_aux);
3948                 list_del(&ifa->if_list_aux);
3949
3950                 addrconf_del_dad_work(ifa);
3951
3952                 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3953                         !addr_is_local(&ifa->addr);
3954
3955                 spin_lock_bh(&ifa->lock);
3956
3957                 if (keep) {
3958                         /* set state to skip the notifier below */
3959                         state = INET6_IFADDR_STATE_DEAD;
3960                         ifa->state = INET6_IFADDR_STATE_PREDAD;
3961                         if (!(ifa->flags & IFA_F_NODAD))
3962                                 ifa->flags |= IFA_F_TENTATIVE;
3963
3964                         rt = ifa->rt;
3965                         ifa->rt = NULL;
3966                 } else {
3967                         state = ifa->state;
3968                         ifa->state = INET6_IFADDR_STATE_DEAD;
3969                 }
3970
3971                 spin_unlock_bh(&ifa->lock);
3972
3973                 if (rt)
3974                         ip6_del_rt(net, rt, false);
3975
3976                 if (state != INET6_IFADDR_STATE_DEAD) {
3977                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
3978                         inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3979                 } else {
3980                         if (idev->cnf.forwarding)
3981                                 addrconf_leave_anycast(ifa);
3982                         addrconf_leave_solict(ifa->idev, &ifa->addr);
3983                 }
3984
3985                 if (!keep) {
3986                         write_lock_bh(&idev->lock);
3987                         list_del_rcu(&ifa->if_list);
3988                         write_unlock_bh(&idev->lock);
3989                         in6_ifa_put(ifa);
3990                 }
3991         }
3992
3993         /* Step 5: Discard anycast and multicast list */
3994         if (unregister) {
3995                 ipv6_ac_destroy_dev(idev);
3996                 ipv6_mc_destroy_dev(idev);
3997         } else if (was_ready) {
3998                 ipv6_mc_down(idev);
3999         }
4000
4001         WRITE_ONCE(idev->tstamp, jiffies);
4002         idev->ra_mtu = 0;
4003
4004         /* Last: Shot the device (if unregistered) */
4005         if (unregister) {
4006                 addrconf_sysctl_unregister(idev);
4007                 neigh_parms_release(&nd_tbl, idev->nd_parms);
4008                 neigh_ifdown(&nd_tbl, dev);
4009                 in6_dev_put(idev);
4010         }
4011         return 0;
4012 }
4013
4014 static void addrconf_rs_timer(struct timer_list *t)
4015 {
4016         struct inet6_dev *idev = from_timer(idev, t, rs_timer);
4017         struct net_device *dev = idev->dev;
4018         struct in6_addr lladdr;
4019         int rtr_solicits;
4020
4021         write_lock(&idev->lock);
4022         if (idev->dead || !(idev->if_flags & IF_READY))
4023                 goto out;
4024
4025         if (!ipv6_accept_ra(idev))
4026                 goto out;
4027
4028         /* Announcement received after solicitation was sent */
4029         if (idev->if_flags & IF_RA_RCVD)
4030                 goto out;
4031
4032         rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4033
4034         if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4035                 write_unlock(&idev->lock);
4036                 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4037                         ndisc_send_rs(dev, &lladdr,
4038                                       &in6addr_linklocal_allrouters);
4039                 else
4040                         goto put;
4041
4042                 write_lock(&idev->lock);
4043                 idev->rs_interval = rfc3315_s14_backoff_update(
4044                                 idev->rs_interval,
4045                                 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4046                 /* The wait after the last probe can be shorter */
4047                 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4048                                              READ_ONCE(idev->cnf.rtr_solicits)) ?
4049                                       READ_ONCE(idev->cnf.rtr_solicit_delay) :
4050                                       idev->rs_interval);
4051         } else {
4052                 /*
4053                  * Note: we do not support deprecated "all on-link"
4054                  * assumption any longer.
4055                  */
4056                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4057         }
4058
4059 out:
4060         write_unlock(&idev->lock);
4061 put:
4062         in6_dev_put(idev);
4063 }
4064
4065 /*
4066  *      Duplicate Address Detection
4067  */
4068 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4069 {
4070         struct inet6_dev *idev = ifp->idev;
4071         unsigned long rand_num;
4072         u64 nonce;
4073
4074         if (ifp->flags & IFA_F_OPTIMISTIC)
4075                 rand_num = 0;
4076         else
4077                 rand_num = get_random_u32_below(
4078                                 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4079
4080         nonce = 0;
4081         if (READ_ONCE(idev->cnf.enhanced_dad) ||
4082             READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4083                 do
4084                         get_random_bytes(&nonce, 6);
4085                 while (nonce == 0);
4086         }
4087         ifp->dad_nonce = nonce;
4088         ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4089         addrconf_mod_dad_work(ifp, rand_num);
4090 }
4091
4092 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4093 {
4094         struct inet6_dev *idev = ifp->idev;
4095         struct net_device *dev = idev->dev;
4096         bool bump_id, notify = false;
4097         struct net *net;
4098
4099         addrconf_join_solict(dev, &ifp->addr);
4100
4101         read_lock_bh(&idev->lock);
4102         spin_lock(&ifp->lock);
4103         if (ifp->state == INET6_IFADDR_STATE_DEAD)
4104                 goto out;
4105
4106         net = dev_net(dev);
4107         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4108             (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4109              READ_ONCE(idev->cnf.accept_dad) < 1) ||
4110             !(ifp->flags&IFA_F_TENTATIVE) ||
4111             ifp->flags & IFA_F_NODAD) {
4112                 bool send_na = false;
4113
4114                 if (ifp->flags & IFA_F_TENTATIVE &&
4115                     !(ifp->flags & IFA_F_OPTIMISTIC))
4116                         send_na = true;
4117                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4118                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4119                 spin_unlock(&ifp->lock);
4120                 read_unlock_bh(&idev->lock);
4121
4122                 addrconf_dad_completed(ifp, bump_id, send_na);
4123                 return;
4124         }
4125
4126         if (!(idev->if_flags & IF_READY)) {
4127                 spin_unlock(&ifp->lock);
4128                 read_unlock_bh(&idev->lock);
4129                 /*
4130                  * If the device is not ready:
4131                  * - keep it tentative if it is a permanent address.
4132                  * - otherwise, kill it.
4133                  */
4134                 in6_ifa_hold(ifp);
4135                 addrconf_dad_stop(ifp, 0);
4136                 return;
4137         }
4138
4139         /*
4140          * Optimistic nodes can start receiving
4141          * Frames right away
4142          */
4143         if (ifp->flags & IFA_F_OPTIMISTIC) {
4144                 ip6_ins_rt(net, ifp->rt);
4145                 if (ipv6_use_optimistic_addr(net, idev)) {
4146                         /* Because optimistic nodes can use this address,
4147                          * notify listeners. If DAD fails, RTM_DELADDR is sent.
4148                          */
4149                         notify = true;
4150                 }
4151         }
4152
4153         addrconf_dad_kick(ifp);
4154 out:
4155         spin_unlock(&ifp->lock);
4156         read_unlock_bh(&idev->lock);
4157         if (notify)
4158                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4159 }
4160
4161 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4162 {
4163         bool begin_dad = false;
4164
4165         spin_lock_bh(&ifp->lock);
4166         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4167                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4168                 begin_dad = true;
4169         }
4170         spin_unlock_bh(&ifp->lock);
4171
4172         if (begin_dad)
4173                 addrconf_mod_dad_work(ifp, 0);
4174 }
4175
4176 static void addrconf_dad_work(struct work_struct *w)
4177 {
4178         struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4179                                                 struct inet6_ifaddr,
4180                                                 dad_work);
4181         struct inet6_dev *idev = ifp->idev;
4182         bool bump_id, disable_ipv6 = false;
4183         struct in6_addr mcaddr;
4184
4185         enum {
4186                 DAD_PROCESS,
4187                 DAD_BEGIN,
4188                 DAD_ABORT,
4189         } action = DAD_PROCESS;
4190
4191         rtnl_lock();
4192
4193         spin_lock_bh(&ifp->lock);
4194         if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4195                 action = DAD_BEGIN;
4196                 ifp->state = INET6_IFADDR_STATE_DAD;
4197         } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4198                 action = DAD_ABORT;
4199                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4200
4201                 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->accept_dad) > 1 ||
4202                      READ_ONCE(idev->cnf.accept_dad) > 1) &&
4203                     !idev->cnf.disable_ipv6 &&
4204                     !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4205                         struct in6_addr addr;
4206
4207                         addr.s6_addr32[0] = htonl(0xfe800000);
4208                         addr.s6_addr32[1] = 0;
4209
4210                         if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4211                             ipv6_addr_equal(&ifp->addr, &addr)) {
4212                                 /* DAD failed for link-local based on MAC */
4213                                 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4214
4215                                 pr_info("%s: IPv6 being disabled!\n",
4216                                         ifp->idev->dev->name);
4217                                 disable_ipv6 = true;
4218                         }
4219                 }
4220         }
4221         spin_unlock_bh(&ifp->lock);
4222
4223         if (action == DAD_BEGIN) {
4224                 addrconf_dad_begin(ifp);
4225                 goto out;
4226         } else if (action == DAD_ABORT) {
4227                 in6_ifa_hold(ifp);
4228                 addrconf_dad_stop(ifp, 1);
4229                 if (disable_ipv6)
4230                         addrconf_ifdown(idev->dev, false);
4231                 goto out;
4232         }
4233
4234         if (!ifp->dad_probes && addrconf_dad_end(ifp))
4235                 goto out;
4236
4237         write_lock_bh(&idev->lock);
4238         if (idev->dead || !(idev->if_flags & IF_READY)) {
4239                 write_unlock_bh(&idev->lock);
4240                 goto out;
4241         }
4242
4243         spin_lock(&ifp->lock);
4244         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4245                 spin_unlock(&ifp->lock);
4246                 write_unlock_bh(&idev->lock);
4247                 goto out;
4248         }
4249
4250         if (ifp->dad_probes == 0) {
4251                 bool send_na = false;
4252
4253                 /*
4254                  * DAD was successful
4255                  */
4256
4257                 if (ifp->flags & IFA_F_TENTATIVE &&
4258                     !(ifp->flags & IFA_F_OPTIMISTIC))
4259                         send_na = true;
4260                 bump_id = ifp->flags & IFA_F_TENTATIVE;
4261                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4262                 spin_unlock(&ifp->lock);
4263                 write_unlock_bh(&idev->lock);
4264
4265                 addrconf_dad_completed(ifp, bump_id, send_na);
4266
4267                 goto out;
4268         }
4269
4270         ifp->dad_probes--;
4271         addrconf_mod_dad_work(ifp,
4272                               max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4273                                   HZ/100));
4274         spin_unlock(&ifp->lock);
4275         write_unlock_bh(&idev->lock);
4276
4277         /* send a neighbour solicitation for our addr */
4278         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4279         ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4280                       ifp->dad_nonce);
4281 out:
4282         in6_ifa_put(ifp);
4283         rtnl_unlock();
4284 }
4285
4286 /* ifp->idev must be at least read locked */
4287 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4288 {
4289         struct inet6_ifaddr *ifpiter;
4290         struct inet6_dev *idev = ifp->idev;
4291
4292         list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4293                 if (ifpiter->scope > IFA_LINK)
4294                         break;
4295                 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4296                     (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4297                                        IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4298                     IFA_F_PERMANENT)
4299                         return false;
4300         }
4301         return true;
4302 }
4303
4304 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4305                                    bool send_na)
4306 {
4307         struct net_device *dev = ifp->idev->dev;
4308         struct in6_addr lladdr;
4309         bool send_rs, send_mld;
4310
4311         addrconf_del_dad_work(ifp);
4312
4313         /*
4314          *      Configure the address for reception. Now it is valid.
4315          */
4316
4317         ipv6_ifa_notify(RTM_NEWADDR, ifp);
4318
4319         /* If added prefix is link local and we are prepared to process
4320            router advertisements, start sending router solicitations.
4321          */
4322
4323         read_lock_bh(&ifp->idev->lock);
4324         send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4325         send_rs = send_mld &&
4326                   ipv6_accept_ra(ifp->idev) &&
4327                   READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4328                   (dev->flags & IFF_LOOPBACK) == 0 &&
4329                   (dev->type != ARPHRD_TUNNEL) &&
4330                   !netif_is_team_port(dev);
4331         read_unlock_bh(&ifp->idev->lock);
4332
4333         /* While dad is in progress mld report's source address is in6_addrany.
4334          * Resend with proper ll now.
4335          */
4336         if (send_mld)
4337                 ipv6_mc_dad_complete(ifp->idev);
4338
4339         /* send unsolicited NA if enabled */
4340         if (send_na &&
4341             (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4342              READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4343                 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4344                               /*router=*/ !!ifp->idev->cnf.forwarding,
4345                               /*solicited=*/ false, /*override=*/ true,
4346                               /*inc_opt=*/ true);
4347         }
4348
4349         if (send_rs) {
4350                 /*
4351                  *      If a host as already performed a random delay
4352                  *      [...] as part of DAD [...] there is no need
4353                  *      to delay again before sending the first RS
4354                  */
4355                 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4356                         return;
4357                 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4358
4359                 write_lock_bh(&ifp->idev->lock);
4360                 spin_lock(&ifp->lock);
4361                 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4362                         READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4363                 ifp->idev->rs_probes = 1;
4364                 ifp->idev->if_flags |= IF_RS_SENT;
4365                 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4366                 spin_unlock(&ifp->lock);
4367                 write_unlock_bh(&ifp->idev->lock);
4368         }
4369
4370         if (bump_id)
4371                 rt_genid_bump_ipv6(dev_net(dev));
4372
4373         /* Make sure that a new temporary address will be created
4374          * before this temporary address becomes deprecated.
4375          */
4376         if (ifp->flags & IFA_F_TEMPORARY)
4377                 addrconf_verify_rtnl(dev_net(dev));
4378 }
4379
4380 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4381 {
4382         struct inet6_ifaddr *ifp;
4383
4384         read_lock_bh(&idev->lock);
4385         list_for_each_entry(ifp, &idev->addr_list, if_list) {
4386                 spin_lock(&ifp->lock);
4387                 if ((ifp->flags & IFA_F_TENTATIVE &&
4388                      ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4389                         if (restart)
4390                                 ifp->state = INET6_IFADDR_STATE_PREDAD;
4391                         addrconf_dad_kick(ifp);
4392                 }
4393                 spin_unlock(&ifp->lock);
4394         }
4395         read_unlock_bh(&idev->lock);
4396 }
4397
4398 #ifdef CONFIG_PROC_FS
4399 struct if6_iter_state {
4400         struct seq_net_private p;
4401         int bucket;
4402         int offset;
4403 };
4404
4405 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4406 {
4407         struct if6_iter_state *state = seq->private;
4408         struct net *net = seq_file_net(seq);
4409         struct inet6_ifaddr *ifa = NULL;
4410         int p = 0;
4411
4412         /* initial bucket if pos is 0 */
4413         if (pos == 0) {
4414                 state->bucket = 0;
4415                 state->offset = 0;
4416         }
4417
4418         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4419                 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4420                                          addr_lst) {
4421                         /* sync with offset */
4422                         if (p < state->offset) {
4423                                 p++;
4424                                 continue;
4425                         }
4426                         return ifa;
4427                 }
4428
4429                 /* prepare for next bucket */
4430                 state->offset = 0;
4431                 p = 0;
4432         }
4433         return NULL;
4434 }
4435
4436 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4437                                          struct inet6_ifaddr *ifa)
4438 {
4439         struct if6_iter_state *state = seq->private;
4440         struct net *net = seq_file_net(seq);
4441
4442         hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4443                 state->offset++;
4444                 return ifa;
4445         }
4446
4447         state->offset = 0;
4448         while (++state->bucket < IN6_ADDR_HSIZE) {
4449                 hlist_for_each_entry_rcu(ifa,
4450                                      &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4451                         return ifa;
4452                 }
4453         }
4454
4455         return NULL;
4456 }
4457
4458 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4459         __acquires(rcu)
4460 {
4461         rcu_read_lock();
4462         return if6_get_first(seq, *pos);
4463 }
4464
4465 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4466 {
4467         struct inet6_ifaddr *ifa;
4468
4469         ifa = if6_get_next(seq, v);
4470         ++*pos;
4471         return ifa;
4472 }
4473
4474 static void if6_seq_stop(struct seq_file *seq, void *v)
4475         __releases(rcu)
4476 {
4477         rcu_read_unlock();
4478 }
4479
4480 static int if6_seq_show(struct seq_file *seq, void *v)
4481 {
4482         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4483         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4484                    &ifp->addr,
4485                    ifp->idev->dev->ifindex,
4486                    ifp->prefix_len,
4487                    ifp->scope,
4488                    (u8) ifp->flags,
4489                    ifp->idev->dev->name);
4490         return 0;
4491 }
4492
4493 static const struct seq_operations if6_seq_ops = {
4494         .start  = if6_seq_start,
4495         .next   = if6_seq_next,
4496         .show   = if6_seq_show,
4497         .stop   = if6_seq_stop,
4498 };
4499
4500 static int __net_init if6_proc_net_init(struct net *net)
4501 {
4502         if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4503                         sizeof(struct if6_iter_state)))
4504                 return -ENOMEM;
4505         return 0;
4506 }
4507
4508 static void __net_exit if6_proc_net_exit(struct net *net)
4509 {
4510         remove_proc_entry("if_inet6", net->proc_net);
4511 }
4512
4513 static struct pernet_operations if6_proc_net_ops = {
4514         .init = if6_proc_net_init,
4515         .exit = if6_proc_net_exit,
4516 };
4517
4518 int __init if6_proc_init(void)
4519 {
4520         return register_pernet_subsys(&if6_proc_net_ops);
4521 }
4522
4523 void if6_proc_exit(void)
4524 {
4525         unregister_pernet_subsys(&if6_proc_net_ops);
4526 }
4527 #endif  /* CONFIG_PROC_FS */
4528
4529 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4530 /* Check if address is a home address configured on any interface. */
4531 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4532 {
4533         unsigned int hash = inet6_addr_hash(net, addr);
4534         struct inet6_ifaddr *ifp = NULL;
4535         int ret = 0;
4536
4537         rcu_read_lock();
4538         hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4539                 if (ipv6_addr_equal(&ifp->addr, addr) &&
4540                     (ifp->flags & IFA_F_HOMEADDRESS)) {
4541                         ret = 1;
4542                         break;
4543                 }
4544         }
4545         rcu_read_unlock();
4546         return ret;
4547 }
4548 #endif
4549
4550 /* RFC6554 has some algorithm to avoid loops in segment routing by
4551  * checking if the segments contains any of a local interface address.
4552  *
4553  * Quote:
4554  *
4555  * To detect loops in the SRH, a router MUST determine if the SRH
4556  * includes multiple addresses assigned to any interface on that router.
4557  * If such addresses appear more than once and are separated by at least
4558  * one address not assigned to that router.
4559  */
4560 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4561                           unsigned char nsegs)
4562 {
4563         const struct in6_addr *addr;
4564         int i, ret = 0, found = 0;
4565         struct inet6_ifaddr *ifp;
4566         bool separated = false;
4567         unsigned int hash;
4568         bool hash_found;
4569
4570         rcu_read_lock();
4571         for (i = 0; i < nsegs; i++) {
4572                 addr = &segs[i];
4573                 hash = inet6_addr_hash(net, addr);
4574
4575                 hash_found = false;
4576                 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4577
4578                         if (ipv6_addr_equal(&ifp->addr, addr)) {
4579                                 hash_found = true;
4580                                 break;
4581                         }
4582                 }
4583
4584                 if (hash_found) {
4585                         if (found > 1 && separated) {
4586                                 ret = 1;
4587                                 break;
4588                         }
4589
4590                         separated = false;
4591                         found++;
4592                 } else {
4593                         separated = true;
4594                 }
4595         }
4596         rcu_read_unlock();
4597
4598         return ret;
4599 }
4600
4601 /*
4602  *      Periodic address status verification
4603  */
4604
4605 static void addrconf_verify_rtnl(struct net *net)
4606 {
4607         unsigned long now, next, next_sec, next_sched;
4608         struct inet6_ifaddr *ifp;
4609         int i;
4610
4611         ASSERT_RTNL();
4612
4613         rcu_read_lock_bh();
4614         now = jiffies;
4615         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4616
4617         cancel_delayed_work(&net->ipv6.addr_chk_work);
4618
4619         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4620 restart:
4621                 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4622                         unsigned long age;
4623
4624                         /* When setting preferred_lft to a value not zero or
4625                          * infinity, while valid_lft is infinity
4626                          * IFA_F_PERMANENT has a non-infinity life time.
4627                          */
4628                         if ((ifp->flags & IFA_F_PERMANENT) &&
4629                             (ifp->prefered_lft == INFINITY_LIFE_TIME))
4630                                 continue;
4631
4632                         spin_lock(&ifp->lock);
4633                         /* We try to batch several events at once. */
4634                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4635
4636                         if ((ifp->flags&IFA_F_TEMPORARY) &&
4637                             !(ifp->flags&IFA_F_TENTATIVE) &&
4638                             ifp->prefered_lft != INFINITY_LIFE_TIME &&
4639                             !ifp->regen_count && ifp->ifpub) {
4640                                 /* This is a non-regenerated temporary addr. */
4641
4642                                 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4643
4644                                 if (age + regen_advance >= ifp->prefered_lft) {
4645                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
4646                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4647                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
4648
4649                                         ifp->regen_count++;
4650                                         in6_ifa_hold(ifp);
4651                                         in6_ifa_hold(ifpub);
4652                                         spin_unlock(&ifp->lock);
4653
4654                                         spin_lock(&ifpub->lock);
4655                                         ifpub->regen_count = 0;
4656                                         spin_unlock(&ifpub->lock);
4657                                         rcu_read_unlock_bh();
4658                                         ipv6_create_tempaddr(ifpub, true);
4659                                         in6_ifa_put(ifpub);
4660                                         in6_ifa_put(ifp);
4661                                         rcu_read_lock_bh();
4662                                         goto restart;
4663                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4664                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4665                         }
4666
4667                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4668                             age >= ifp->valid_lft) {
4669                                 spin_unlock(&ifp->lock);
4670                                 in6_ifa_hold(ifp);
4671                                 rcu_read_unlock_bh();
4672                                 ipv6_del_addr(ifp);
4673                                 rcu_read_lock_bh();
4674                                 goto restart;
4675                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4676                                 spin_unlock(&ifp->lock);
4677                                 continue;
4678                         } else if (age >= ifp->prefered_lft) {
4679                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4680                                 int deprecate = 0;
4681
4682                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4683                                         deprecate = 1;
4684                                         ifp->flags |= IFA_F_DEPRECATED;
4685                                 }
4686
4687                                 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4688                                     (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4689                                         next = ifp->tstamp + ifp->valid_lft * HZ;
4690
4691                                 spin_unlock(&ifp->lock);
4692
4693                                 if (deprecate) {
4694                                         in6_ifa_hold(ifp);
4695
4696                                         ipv6_ifa_notify(0, ifp);
4697                                         in6_ifa_put(ifp);
4698                                         goto restart;
4699                                 }
4700                         } else {
4701                                 /* ifp->prefered_lft <= ifp->valid_lft */
4702                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4703                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
4704                                 spin_unlock(&ifp->lock);
4705                         }
4706                 }
4707         }
4708
4709         next_sec = round_jiffies_up(next);
4710         next_sched = next;
4711
4712         /* If rounded timeout is accurate enough, accept it. */
4713         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4714                 next_sched = next_sec;
4715
4716         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4717         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4718                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4719
4720         pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4721                  now, next, next_sec, next_sched);
4722         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4723         rcu_read_unlock_bh();
4724 }
4725
4726 static void addrconf_verify_work(struct work_struct *w)
4727 {
4728         struct net *net = container_of(to_delayed_work(w), struct net,
4729                                        ipv6.addr_chk_work);
4730
4731         rtnl_lock();
4732         addrconf_verify_rtnl(net);
4733         rtnl_unlock();
4734 }
4735
4736 static void addrconf_verify(struct net *net)
4737 {
4738         mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4739 }
4740
4741 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4742                                      struct in6_addr **peer_pfx)
4743 {
4744         struct in6_addr *pfx = NULL;
4745
4746         *peer_pfx = NULL;
4747
4748         if (addr)
4749                 pfx = nla_data(addr);
4750
4751         if (local) {
4752                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4753                         *peer_pfx = pfx;
4754                 pfx = nla_data(local);
4755         }
4756
4757         return pfx;
4758 }
4759
4760 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4761         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
4762         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
4763         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
4764         [IFA_FLAGS]             = { .len = sizeof(u32) },
4765         [IFA_RT_PRIORITY]       = { .len = sizeof(u32) },
4766         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
4767         [IFA_PROTO]             = { .type = NLA_U8 },
4768 };
4769
4770 static int
4771 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4772                   struct netlink_ext_ack *extack)
4773 {
4774         struct net *net = sock_net(skb->sk);
4775         struct ifaddrmsg *ifm;
4776         struct nlattr *tb[IFA_MAX+1];
4777         struct in6_addr *pfx, *peer_pfx;
4778         u32 ifa_flags;
4779         int err;
4780
4781         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4782                                      ifa_ipv6_policy, extack);
4783         if (err < 0)
4784                 return err;
4785
4786         ifm = nlmsg_data(nlh);
4787         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4788         if (!pfx)
4789                 return -EINVAL;
4790
4791         ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4792
4793         /* We ignore other flags so far. */
4794         ifa_flags &= IFA_F_MANAGETEMPADDR;
4795
4796         return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4797                               ifm->ifa_prefixlen, extack);
4798 }
4799
4800 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4801                                unsigned long expires, u32 flags,
4802                                bool modify_peer)
4803 {
4804         struct fib6_table *table;
4805         struct fib6_info *f6i;
4806         u32 prio;
4807
4808         f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4809                                         ifp->prefix_len,
4810                                         ifp->idev->dev, 0, RTF_DEFAULT, true);
4811         if (!f6i)
4812                 return -ENOENT;
4813
4814         prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4815         if (f6i->fib6_metric != prio) {
4816                 /* delete old one */
4817                 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4818
4819                 /* add new one */
4820                 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4821                                       ifp->prefix_len,
4822                                       ifp->rt_priority, ifp->idev->dev,
4823                                       expires, flags, GFP_KERNEL);
4824         } else {
4825                 table = f6i->fib6_table;
4826                 spin_lock_bh(&table->tb6_lock);
4827
4828                 if (!(flags & RTF_EXPIRES)) {
4829                         fib6_clean_expires(f6i);
4830                         fib6_remove_gc_list(f6i);
4831                 } else {
4832                         fib6_set_expires(f6i, expires);
4833                         fib6_add_gc_list(f6i);
4834                 }
4835
4836                 spin_unlock_bh(&table->tb6_lock);
4837
4838                 fib6_info_release(f6i);
4839         }
4840
4841         return 0;
4842 }
4843
4844 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4845                              struct ifa6_config *cfg)
4846 {
4847         u32 flags;
4848         clock_t expires;
4849         unsigned long timeout;
4850         bool was_managetempaddr;
4851         bool had_prefixroute;
4852         bool new_peer = false;
4853
4854         ASSERT_RTNL();
4855
4856         if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4857                 return -EINVAL;
4858
4859         if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4860             (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4861                 return -EINVAL;
4862
4863         if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4864                 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4865
4866         timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4867         if (addrconf_finite_timeout(timeout)) {
4868                 expires = jiffies_to_clock_t(timeout * HZ);
4869                 cfg->valid_lft = timeout;
4870                 flags = RTF_EXPIRES;
4871         } else {
4872                 expires = 0;
4873                 flags = 0;
4874                 cfg->ifa_flags |= IFA_F_PERMANENT;
4875         }
4876
4877         timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4878         if (addrconf_finite_timeout(timeout)) {
4879                 if (timeout == 0)
4880                         cfg->ifa_flags |= IFA_F_DEPRECATED;
4881                 cfg->preferred_lft = timeout;
4882         }
4883
4884         if (cfg->peer_pfx &&
4885             memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4886                 if (!ipv6_addr_any(&ifp->peer_addr))
4887                         cleanup_prefix_route(ifp, expires, true, true);
4888                 new_peer = true;
4889         }
4890
4891         spin_lock_bh(&ifp->lock);
4892         was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4893         had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4894                           !(ifp->flags & IFA_F_NOPREFIXROUTE);
4895         ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4896                         IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4897                         IFA_F_NOPREFIXROUTE);
4898         ifp->flags |= cfg->ifa_flags;
4899         WRITE_ONCE(ifp->tstamp, jiffies);
4900         WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4901         WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4902         WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4903
4904         if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4905                 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4906
4907         if (new_peer)
4908                 ifp->peer_addr = *cfg->peer_pfx;
4909
4910         spin_unlock_bh(&ifp->lock);
4911         if (!(ifp->flags&IFA_F_TENTATIVE))
4912                 ipv6_ifa_notify(0, ifp);
4913
4914         if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4915                 int rc = -ENOENT;
4916
4917                 if (had_prefixroute)
4918                         rc = modify_prefix_route(ifp, expires, flags, false);
4919
4920                 /* prefix route could have been deleted; if so restore it */
4921                 if (rc == -ENOENT) {
4922                         addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4923                                               ifp->rt_priority, ifp->idev->dev,
4924                                               expires, flags, GFP_KERNEL);
4925                 }
4926
4927                 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4928                         rc = modify_prefix_route(ifp, expires, flags, true);
4929
4930                 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4931                         addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4932                                               ifp->rt_priority, ifp->idev->dev,
4933                                               expires, flags, GFP_KERNEL);
4934                 }
4935         } else if (had_prefixroute) {
4936                 enum cleanup_prefix_rt_t action;
4937                 unsigned long rt_expires;
4938
4939                 write_lock_bh(&ifp->idev->lock);
4940                 action = check_cleanup_prefix_route(ifp, &rt_expires);
4941                 write_unlock_bh(&ifp->idev->lock);
4942
4943                 if (action != CLEANUP_PREFIX_RT_NOP) {
4944                         cleanup_prefix_route(ifp, rt_expires,
4945                                 action == CLEANUP_PREFIX_RT_DEL, false);
4946                 }
4947         }
4948
4949         if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4950                 if (was_managetempaddr &&
4951                     !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4952                         cfg->valid_lft = 0;
4953                         cfg->preferred_lft = 0;
4954                 }
4955                 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4956                                  cfg->preferred_lft, !was_managetempaddr,
4957                                  jiffies);
4958         }
4959
4960         addrconf_verify_rtnl(net);
4961
4962         return 0;
4963 }
4964
4965 static int
4966 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4967                   struct netlink_ext_ack *extack)
4968 {
4969         struct net *net = sock_net(skb->sk);
4970         struct ifaddrmsg *ifm;
4971         struct nlattr *tb[IFA_MAX+1];
4972         struct in6_addr *peer_pfx;
4973         struct inet6_ifaddr *ifa;
4974         struct net_device *dev;
4975         struct inet6_dev *idev;
4976         struct ifa6_config cfg;
4977         int err;
4978
4979         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4980                                      ifa_ipv6_policy, extack);
4981         if (err < 0)
4982                 return err;
4983
4984         memset(&cfg, 0, sizeof(cfg));
4985
4986         ifm = nlmsg_data(nlh);
4987         cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4988         if (!cfg.pfx)
4989                 return -EINVAL;
4990
4991         cfg.peer_pfx = peer_pfx;
4992         cfg.plen = ifm->ifa_prefixlen;
4993         if (tb[IFA_RT_PRIORITY])
4994                 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4995
4996         if (tb[IFA_PROTO])
4997                 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4998
4999         cfg.valid_lft = INFINITY_LIFE_TIME;
5000         cfg.preferred_lft = INFINITY_LIFE_TIME;
5001
5002         if (tb[IFA_CACHEINFO]) {
5003                 struct ifa_cacheinfo *ci;
5004
5005                 ci = nla_data(tb[IFA_CACHEINFO]);
5006                 cfg.valid_lft = ci->ifa_valid;
5007                 cfg.preferred_lft = ci->ifa_prefered;
5008         }
5009
5010         dev =  __dev_get_by_index(net, ifm->ifa_index);
5011         if (!dev) {
5012                 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5013                 return -ENODEV;
5014         }
5015
5016         if (tb[IFA_FLAGS])
5017                 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
5018         else
5019                 cfg.ifa_flags = ifm->ifa_flags;
5020
5021         /* We ignore other flags so far. */
5022         cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
5023                          IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
5024                          IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
5025
5026         idev = ipv6_find_idev(dev);
5027         if (IS_ERR(idev))
5028                 return PTR_ERR(idev);
5029
5030         if (!ipv6_allow_optimistic_dad(net, idev))
5031                 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5032
5033         if (cfg.ifa_flags & IFA_F_NODAD &&
5034             cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5035                 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5036                 return -EINVAL;
5037         }
5038
5039         ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5040         if (!ifa) {
5041                 /*
5042                  * It would be best to check for !NLM_F_CREATE here but
5043                  * userspace already relies on not having to provide this.
5044                  */
5045                 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
5046         }
5047
5048         if (nlh->nlmsg_flags & NLM_F_EXCL ||
5049             !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5050                 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5051                 err = -EEXIST;
5052         } else {
5053                 err = inet6_addr_modify(net, ifa, &cfg);
5054         }
5055
5056         in6_ifa_put(ifa);
5057
5058         return err;
5059 }
5060
5061 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5062                           u8 scope, int ifindex)
5063 {
5064         struct ifaddrmsg *ifm;
5065
5066         ifm = nlmsg_data(nlh);
5067         ifm->ifa_family = AF_INET6;
5068         ifm->ifa_prefixlen = prefixlen;
5069         ifm->ifa_flags = flags;
5070         ifm->ifa_scope = scope;
5071         ifm->ifa_index = ifindex;
5072 }
5073
5074 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5075                          unsigned long tstamp, u32 preferred, u32 valid)
5076 {
5077         struct ifa_cacheinfo ci;
5078
5079         ci.cstamp = cstamp_delta(cstamp);
5080         ci.tstamp = cstamp_delta(tstamp);
5081         ci.ifa_prefered = preferred;
5082         ci.ifa_valid = valid;
5083
5084         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5085 }
5086
5087 static inline int rt_scope(int ifa_scope)
5088 {
5089         if (ifa_scope & IFA_HOST)
5090                 return RT_SCOPE_HOST;
5091         else if (ifa_scope & IFA_LINK)
5092                 return RT_SCOPE_LINK;
5093         else if (ifa_scope & IFA_SITE)
5094                 return RT_SCOPE_SITE;
5095         else
5096                 return RT_SCOPE_UNIVERSE;
5097 }
5098
5099 static inline int inet6_ifaddr_msgsize(void)
5100 {
5101         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5102                + nla_total_size(16) /* IFA_LOCAL */
5103                + nla_total_size(16) /* IFA_ADDRESS */
5104                + nla_total_size(sizeof(struct ifa_cacheinfo))
5105                + nla_total_size(4)  /* IFA_FLAGS */
5106                + nla_total_size(1)  /* IFA_PROTO */
5107                + nla_total_size(4)  /* IFA_RT_PRIORITY */;
5108 }
5109
5110 enum addr_type_t {
5111         UNICAST_ADDR,
5112         MULTICAST_ADDR,
5113         ANYCAST_ADDR,
5114 };
5115
5116 struct inet6_fill_args {
5117         u32 portid;
5118         u32 seq;
5119         int event;
5120         unsigned int flags;
5121         int netnsid;
5122         int ifindex;
5123         enum addr_type_t type;
5124 };
5125
5126 static int inet6_fill_ifaddr(struct sk_buff *skb,
5127                              const struct inet6_ifaddr *ifa,
5128                              struct inet6_fill_args *args)
5129 {
5130         struct nlmsghdr *nlh;
5131         u32 preferred, valid;
5132         u32 flags, priority;
5133         u8 proto;
5134
5135         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5136                         sizeof(struct ifaddrmsg), args->flags);
5137         if (!nlh)
5138                 return -EMSGSIZE;
5139
5140         flags = READ_ONCE(ifa->flags);
5141         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5142                       ifa->idev->dev->ifindex);
5143
5144         if (args->netnsid >= 0 &&
5145             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5146                 goto error;
5147
5148         preferred = READ_ONCE(ifa->prefered_lft);
5149         valid = READ_ONCE(ifa->valid_lft);
5150
5151         if (!((flags & IFA_F_PERMANENT) &&
5152               (preferred == INFINITY_LIFE_TIME))) {
5153                 if (preferred != INFINITY_LIFE_TIME) {
5154                         long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5155
5156                         if (preferred > tval)
5157                                 preferred -= tval;
5158                         else
5159                                 preferred = 0;
5160                         if (valid != INFINITY_LIFE_TIME) {
5161                                 if (valid > tval)
5162                                         valid -= tval;
5163                                 else
5164                                         valid = 0;
5165                         }
5166                 }
5167         } else {
5168                 preferred = INFINITY_LIFE_TIME;
5169                 valid = INFINITY_LIFE_TIME;
5170         }
5171
5172         if (!ipv6_addr_any(&ifa->peer_addr)) {
5173                 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5174                     nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5175                         goto error;
5176         } else {
5177                 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5178                         goto error;
5179         }
5180
5181         priority = READ_ONCE(ifa->rt_priority);
5182         if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5183                 goto error;
5184
5185         if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5186                           preferred, valid) < 0)
5187                 goto error;
5188
5189         if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5190                 goto error;
5191
5192         proto = READ_ONCE(ifa->ifa_proto);
5193         if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5194                 goto error;
5195
5196         nlmsg_end(skb, nlh);
5197         return 0;
5198
5199 error:
5200         nlmsg_cancel(skb, nlh);
5201         return -EMSGSIZE;
5202 }
5203
5204 static int inet6_fill_ifmcaddr(struct sk_buff *skb,
5205                                const struct ifmcaddr6 *ifmca,
5206                                struct inet6_fill_args *args)
5207 {
5208         int ifindex = ifmca->idev->dev->ifindex;
5209         u8 scope = RT_SCOPE_UNIVERSE;
5210         struct nlmsghdr *nlh;
5211
5212         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5213                 scope = RT_SCOPE_SITE;
5214
5215         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5216                         sizeof(struct ifaddrmsg), args->flags);
5217         if (!nlh)
5218                 return -EMSGSIZE;
5219
5220         if (args->netnsid >= 0 &&
5221             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5222                 nlmsg_cancel(skb, nlh);
5223                 return -EMSGSIZE;
5224         }
5225
5226         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5227         if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5228             put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5229                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5230                 nlmsg_cancel(skb, nlh);
5231                 return -EMSGSIZE;
5232         }
5233
5234         nlmsg_end(skb, nlh);
5235         return 0;
5236 }
5237
5238 static int inet6_fill_ifacaddr(struct sk_buff *skb,
5239                                const struct ifacaddr6 *ifaca,
5240                                struct inet6_fill_args *args)
5241 {
5242         struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5243         int ifindex = dev ? dev->ifindex : 1;
5244         u8 scope = RT_SCOPE_UNIVERSE;
5245         struct nlmsghdr *nlh;
5246
5247         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5248                 scope = RT_SCOPE_SITE;
5249
5250         nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5251                         sizeof(struct ifaddrmsg), args->flags);
5252         if (!nlh)
5253                 return -EMSGSIZE;
5254
5255         if (args->netnsid >= 0 &&
5256             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5257                 nlmsg_cancel(skb, nlh);
5258                 return -EMSGSIZE;
5259         }
5260
5261         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5262         if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5263             put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5264                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5265                 nlmsg_cancel(skb, nlh);
5266                 return -EMSGSIZE;
5267         }
5268
5269         nlmsg_end(skb, nlh);
5270         return 0;
5271 }
5272
5273 /* called with rcu_read_lock() */
5274 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5275                           struct netlink_callback *cb, int *s_ip_idx,
5276                           struct inet6_fill_args *fillargs)
5277 {
5278         const struct ifmcaddr6 *ifmca;
5279         const struct ifacaddr6 *ifaca;
5280         int ip_idx = 0;
5281         int err = 0;
5282
5283         switch (fillargs->type) {
5284         case UNICAST_ADDR: {
5285                 const struct inet6_ifaddr *ifa;
5286                 fillargs->event = RTM_NEWADDR;
5287
5288                 /* unicast address incl. temp addr */
5289                 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5290                         if (ip_idx < *s_ip_idx)
5291                                 goto next;
5292                         err = inet6_fill_ifaddr(skb, ifa, fillargs);
5293                         if (err < 0)
5294                                 break;
5295                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5296 next:
5297                         ip_idx++;
5298                 }
5299                 break;
5300         }
5301         case MULTICAST_ADDR:
5302                 fillargs->event = RTM_GETMULTICAST;
5303
5304                 /* multicast address */
5305                 for (ifmca = rcu_dereference(idev->mc_list);
5306                      ifmca;
5307                      ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5308                         if (ip_idx < *s_ip_idx)
5309                                 continue;
5310                         err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5311                         if (err < 0)
5312                                 break;
5313                 }
5314                 break;
5315         case ANYCAST_ADDR:
5316                 fillargs->event = RTM_GETANYCAST;
5317                 /* anycast address */
5318                 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5319                      ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5320                         if (ip_idx < *s_ip_idx)
5321                                 continue;
5322                         err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5323                         if (err < 0)
5324                                 break;
5325                 }
5326                 break;
5327         default:
5328                 break;
5329         }
5330         *s_ip_idx = err ? ip_idx : 0;
5331         return err;
5332 }
5333
5334 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5335                                        struct inet6_fill_args *fillargs,
5336                                        struct net **tgt_net, struct sock *sk,
5337                                        struct netlink_callback *cb)
5338 {
5339         struct netlink_ext_ack *extack = cb->extack;
5340         struct nlattr *tb[IFA_MAX+1];
5341         struct ifaddrmsg *ifm;
5342         int err, i;
5343
5344         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5345                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5346                 return -EINVAL;
5347         }
5348
5349         ifm = nlmsg_data(nlh);
5350         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5351                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5352                 return -EINVAL;
5353         }
5354
5355         fillargs->ifindex = ifm->ifa_index;
5356         if (fillargs->ifindex) {
5357                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5358                 fillargs->flags |= NLM_F_DUMP_FILTERED;
5359         }
5360
5361         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5362                                             ifa_ipv6_policy, extack);
5363         if (err < 0)
5364                 return err;
5365
5366         for (i = 0; i <= IFA_MAX; ++i) {
5367                 if (!tb[i])
5368                         continue;
5369
5370                 if (i == IFA_TARGET_NETNSID) {
5371                         struct net *net;
5372
5373                         fillargs->netnsid = nla_get_s32(tb[i]);
5374                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5375                         if (IS_ERR(net)) {
5376                                 fillargs->netnsid = -1;
5377                                 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5378                                 return PTR_ERR(net);
5379                         }
5380                         *tgt_net = net;
5381                 } else {
5382                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5383                         return -EINVAL;
5384                 }
5385         }
5386
5387         return 0;
5388 }
5389
5390 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5391                            enum addr_type_t type)
5392 {
5393         struct net *tgt_net = sock_net(skb->sk);
5394         const struct nlmsghdr *nlh = cb->nlh;
5395         struct inet6_fill_args fillargs = {
5396                 .portid = NETLINK_CB(cb->skb).portid,
5397                 .seq = cb->nlh->nlmsg_seq,
5398                 .flags = NLM_F_MULTI,
5399                 .netnsid = -1,
5400                 .type = type,
5401         };
5402         struct {
5403                 unsigned long ifindex;
5404                 int ip_idx;
5405         } *ctx = (void *)cb->ctx;
5406         struct net_device *dev;
5407         struct inet6_dev *idev;
5408         int err = 0;
5409
5410         rcu_read_lock();
5411         if (cb->strict_check) {
5412                 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5413                                                   skb->sk, cb);
5414                 if (err < 0)
5415                         goto done;
5416
5417                 err = 0;
5418                 if (fillargs.ifindex) {
5419                         err = -ENODEV;
5420                         dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5421                         if (!dev)
5422                                 goto done;
5423                         idev = __in6_dev_get(dev);
5424                         if (idev)
5425                                 err = in6_dump_addrs(idev, skb, cb,
5426                                                      &ctx->ip_idx,
5427                                                      &fillargs);
5428                         goto done;
5429                 }
5430         }
5431
5432         cb->seq = inet6_base_seq(tgt_net);
5433         for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5434                 idev = __in6_dev_get(dev);
5435                 if (!idev)
5436                         continue;
5437                 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5438                                      &fillargs);
5439                 if (err < 0)
5440                         goto done;
5441         }
5442 done:
5443         rcu_read_unlock();
5444         if (fillargs.netnsid >= 0)
5445                 put_net(tgt_net);
5446
5447         return err;
5448 }
5449
5450 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5451 {
5452         enum addr_type_t type = UNICAST_ADDR;
5453
5454         return inet6_dump_addr(skb, cb, type);
5455 }
5456
5457 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5458 {
5459         enum addr_type_t type = MULTICAST_ADDR;
5460
5461         return inet6_dump_addr(skb, cb, type);
5462 }
5463
5464
5465 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5466 {
5467         enum addr_type_t type = ANYCAST_ADDR;
5468
5469         return inet6_dump_addr(skb, cb, type);
5470 }
5471
5472 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5473                                        const struct nlmsghdr *nlh,
5474                                        struct nlattr **tb,
5475                                        struct netlink_ext_ack *extack)
5476 {
5477         struct ifaddrmsg *ifm;
5478         int i, err;
5479
5480         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5481                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5482                 return -EINVAL;
5483         }
5484
5485         if (!netlink_strict_get_check(skb))
5486                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5487                                               ifa_ipv6_policy, extack);
5488
5489         ifm = nlmsg_data(nlh);
5490         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5491                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5492                 return -EINVAL;
5493         }
5494
5495         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5496                                             ifa_ipv6_policy, extack);
5497         if (err)
5498                 return err;
5499
5500         for (i = 0; i <= IFA_MAX; i++) {
5501                 if (!tb[i])
5502                         continue;
5503
5504                 switch (i) {
5505                 case IFA_TARGET_NETNSID:
5506                 case IFA_ADDRESS:
5507                 case IFA_LOCAL:
5508                         break;
5509                 default:
5510                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5511                         return -EINVAL;
5512                 }
5513         }
5514
5515         return 0;
5516 }
5517
5518 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5519                              struct netlink_ext_ack *extack)
5520 {
5521         struct net *tgt_net = sock_net(in_skb->sk);
5522         struct inet6_fill_args fillargs = {
5523                 .portid = NETLINK_CB(in_skb).portid,
5524                 .seq = nlh->nlmsg_seq,
5525                 .event = RTM_NEWADDR,
5526                 .flags = 0,
5527                 .netnsid = -1,
5528         };
5529         struct ifaddrmsg *ifm;
5530         struct nlattr *tb[IFA_MAX+1];
5531         struct in6_addr *addr = NULL, *peer;
5532         struct net_device *dev = NULL;
5533         struct inet6_ifaddr *ifa;
5534         struct sk_buff *skb;
5535         int err;
5536
5537         err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5538         if (err < 0)
5539                 return err;
5540
5541         if (tb[IFA_TARGET_NETNSID]) {
5542                 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5543
5544                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5545                                                   fillargs.netnsid);
5546                 if (IS_ERR(tgt_net))
5547                         return PTR_ERR(tgt_net);
5548         }
5549
5550         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5551         if (!addr) {
5552                 err = -EINVAL;
5553                 goto errout;
5554         }
5555         ifm = nlmsg_data(nlh);
5556         if (ifm->ifa_index)
5557                 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5558
5559         ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5560         if (!ifa) {
5561                 err = -EADDRNOTAVAIL;
5562                 goto errout;
5563         }
5564
5565         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5566         if (!skb) {
5567                 err = -ENOBUFS;
5568                 goto errout_ifa;
5569         }
5570
5571         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5572         if (err < 0) {
5573                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5574                 WARN_ON(err == -EMSGSIZE);
5575                 kfree_skb(skb);
5576                 goto errout_ifa;
5577         }
5578         err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5579 errout_ifa:
5580         in6_ifa_put(ifa);
5581 errout:
5582         dev_put(dev);
5583         if (fillargs.netnsid >= 0)
5584                 put_net(tgt_net);
5585
5586         return err;
5587 }
5588
5589 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5590 {
5591         struct sk_buff *skb;
5592         struct net *net = dev_net(ifa->idev->dev);
5593         struct inet6_fill_args fillargs = {
5594                 .portid = 0,
5595                 .seq = 0,
5596                 .event = event,
5597                 .flags = 0,
5598                 .netnsid = -1,
5599         };
5600         int err = -ENOBUFS;
5601
5602         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5603         if (!skb)
5604                 goto errout;
5605
5606         err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5607         if (err < 0) {
5608                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5609                 WARN_ON(err == -EMSGSIZE);
5610                 kfree_skb(skb);
5611                 goto errout;
5612         }
5613         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5614         return;
5615 errout:
5616         if (err < 0)
5617                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5618 }
5619
5620 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5621                                __s32 *array, int bytes)
5622 {
5623         BUG_ON(bytes < (DEVCONF_MAX * 4));
5624
5625         memset(array, 0, bytes);
5626         array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5627         array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5628         array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5629         array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5630         array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5631         array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5632         array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5633         array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5634         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5635                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5636         array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5637                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5638         array[DEVCONF_RTR_SOLICIT_DELAY] =
5639                 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5640         array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5641         array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5642                 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5643         array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5644                 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5645         array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5646         array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5647         array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5648         array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5649         array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5650         array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5651         array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5652         array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5653         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5654                 READ_ONCE(cnf->accept_ra_min_hop_limit);
5655         array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5656 #ifdef CONFIG_IPV6_ROUTER_PREF
5657         array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5658         array[DEVCONF_RTR_PROBE_INTERVAL] =
5659                 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5660 #ifdef CONFIG_IPV6_ROUTE_INFO
5661         array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5662                 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5663         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5664                 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5665 #endif
5666 #endif
5667         array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5668         array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5669                 READ_ONCE(cnf->accept_source_route);
5670 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5671         array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5672         array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5673 #endif
5674 #ifdef CONFIG_IPV6_MROUTE
5675         array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5676 #endif
5677         array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5678         array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5679         array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5680         array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5681         array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5682                 READ_ONCE(cnf->suppress_frag_ndisc);
5683         array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5684                 READ_ONCE(cnf->accept_ra_from_local);
5685         array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5686         array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5687                 READ_ONCE(cnf->ignore_routes_with_linkdown);
5688         /* we omit DEVCONF_STABLE_SECRET for now */
5689         array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5690         array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5691                 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5692         array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5693         array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5694         array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5695 #ifdef CONFIG_IPV6_SEG6_HMAC
5696         array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5697 #endif
5698         array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5699         array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5700         array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5701         array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5702         array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5703         array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5704         array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5705         array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5706         array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5707                 READ_ONCE(cnf->ndisc_evict_nocarrier);
5708         array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5709                 READ_ONCE(cnf->accept_untracked_na);
5710         array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5711 }
5712
5713 static inline size_t inet6_ifla6_size(void)
5714 {
5715         return nla_total_size(4) /* IFLA_INET6_FLAGS */
5716              + nla_total_size(sizeof(struct ifla_cacheinfo))
5717              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5718              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5719              + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5720              + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5721              + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5722              + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5723              + 0;
5724 }
5725
5726 static inline size_t inet6_if_nlmsg_size(void)
5727 {
5728         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5729                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5730                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5731                + nla_total_size(4) /* IFLA_MTU */
5732                + nla_total_size(4) /* IFLA_LINK */
5733                + nla_total_size(1) /* IFLA_OPERSTATE */
5734                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5735 }
5736
5737 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5738                                         int bytes)
5739 {
5740         int i;
5741         int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5742         BUG_ON(pad < 0);
5743
5744         /* Use put_unaligned() because stats may not be aligned for u64. */
5745         put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5746         for (i = 1; i < ICMP6_MIB_MAX; i++)
5747                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5748
5749         memset(&stats[ICMP6_MIB_MAX], 0, pad);
5750 }
5751
5752 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5753                                         int bytes, size_t syncpoff)
5754 {
5755         int i, c;
5756         u64 buff[IPSTATS_MIB_MAX];
5757         int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5758
5759         BUG_ON(pad < 0);
5760
5761         memset(buff, 0, sizeof(buff));
5762         buff[0] = IPSTATS_MIB_MAX;
5763
5764         for_each_possible_cpu(c) {
5765                 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5766                         buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5767         }
5768
5769         memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5770         memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5771 }
5772
5773 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5774                              int bytes)
5775 {
5776         switch (attrtype) {
5777         case IFLA_INET6_STATS:
5778                 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5779                                      offsetof(struct ipstats_mib, syncp));
5780                 break;
5781         case IFLA_INET6_ICMP6STATS:
5782                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5783                 break;
5784         }
5785 }
5786
5787 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5788                                   u32 ext_filter_mask)
5789 {
5790         struct ifla_cacheinfo ci;
5791         struct nlattr *nla;
5792         u32 ra_mtu;
5793
5794         if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5795                 goto nla_put_failure;
5796         ci.max_reasm_len = IPV6_MAXPLEN;
5797         ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5798         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5799         ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5800         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5801                 goto nla_put_failure;
5802         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5803         if (!nla)
5804                 goto nla_put_failure;
5805         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5806
5807         /* XXX - MC not implemented */
5808
5809         if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5810                 return 0;
5811
5812         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5813         if (!nla)
5814                 goto nla_put_failure;
5815         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5816
5817         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5818         if (!nla)
5819                 goto nla_put_failure;
5820         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5821
5822         nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5823         if (!nla)
5824                 goto nla_put_failure;
5825         read_lock_bh(&idev->lock);
5826         memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5827         read_unlock_bh(&idev->lock);
5828
5829         if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5830                        READ_ONCE(idev->cnf.addr_gen_mode)))
5831                 goto nla_put_failure;
5832
5833         ra_mtu = READ_ONCE(idev->ra_mtu);
5834         if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5835                 goto nla_put_failure;
5836
5837         return 0;
5838
5839 nla_put_failure:
5840         return -EMSGSIZE;
5841 }
5842
5843 static size_t inet6_get_link_af_size(const struct net_device *dev,
5844                                      u32 ext_filter_mask)
5845 {
5846         if (!__in6_dev_get(dev))
5847                 return 0;
5848
5849         return inet6_ifla6_size();
5850 }
5851
5852 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5853                               u32 ext_filter_mask)
5854 {
5855         struct inet6_dev *idev = __in6_dev_get(dev);
5856
5857         if (!idev)
5858                 return -ENODATA;
5859
5860         if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5861                 return -EMSGSIZE;
5862
5863         return 0;
5864 }
5865
5866 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5867                              struct netlink_ext_ack *extack)
5868 {
5869         struct inet6_ifaddr *ifp;
5870         struct net_device *dev = idev->dev;
5871         bool clear_token, update_rs = false;
5872         struct in6_addr ll_addr;
5873
5874         ASSERT_RTNL();
5875
5876         if (!token)
5877                 return -EINVAL;
5878
5879         if (dev->flags & IFF_LOOPBACK) {
5880                 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5881                 return -EINVAL;
5882         }
5883
5884         if (dev->flags & IFF_NOARP) {
5885                 NL_SET_ERR_MSG_MOD(extack,
5886                                    "Device does not do neighbour discovery");
5887                 return -EINVAL;
5888         }
5889
5890         if (!ipv6_accept_ra(idev)) {
5891                 NL_SET_ERR_MSG_MOD(extack,
5892                                    "Router advertisement is disabled on device");
5893                 return -EINVAL;
5894         }
5895
5896         if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5897                 NL_SET_ERR_MSG(extack,
5898                                "Router solicitation is disabled on device");
5899                 return -EINVAL;
5900         }
5901
5902         write_lock_bh(&idev->lock);
5903
5904         BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5905         memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5906
5907         write_unlock_bh(&idev->lock);
5908
5909         clear_token = ipv6_addr_any(token);
5910         if (clear_token)
5911                 goto update_lft;
5912
5913         if (!idev->dead && (idev->if_flags & IF_READY) &&
5914             !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5915                              IFA_F_OPTIMISTIC)) {
5916                 /* If we're not ready, then normal ifup will take care
5917                  * of this. Otherwise, we need to request our rs here.
5918                  */
5919                 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5920                 update_rs = true;
5921         }
5922
5923 update_lft:
5924         write_lock_bh(&idev->lock);
5925
5926         if (update_rs) {
5927                 idev->if_flags |= IF_RS_SENT;
5928                 idev->rs_interval = rfc3315_s14_backoff_init(
5929                         READ_ONCE(idev->cnf.rtr_solicit_interval));
5930                 idev->rs_probes = 1;
5931                 addrconf_mod_rs_timer(idev, idev->rs_interval);
5932         }
5933
5934         /* Well, that's kinda nasty ... */
5935         list_for_each_entry(ifp, &idev->addr_list, if_list) {
5936                 spin_lock(&ifp->lock);
5937                 if (ifp->tokenized) {
5938                         ifp->valid_lft = 0;
5939                         ifp->prefered_lft = 0;
5940                 }
5941                 spin_unlock(&ifp->lock);
5942         }
5943
5944         write_unlock_bh(&idev->lock);
5945         inet6_ifinfo_notify(RTM_NEWLINK, idev);
5946         addrconf_verify_rtnl(dev_net(dev));
5947         return 0;
5948 }
5949
5950 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5951         [IFLA_INET6_ADDR_GEN_MODE]      = { .type = NLA_U8 },
5952         [IFLA_INET6_TOKEN]              = { .len = sizeof(struct in6_addr) },
5953         [IFLA_INET6_RA_MTU]             = { .type = NLA_REJECT,
5954                                             .reject_message =
5955                                                 "IFLA_INET6_RA_MTU can not be set" },
5956 };
5957
5958 static int check_addr_gen_mode(int mode)
5959 {
5960         if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5961             mode != IN6_ADDR_GEN_MODE_NONE &&
5962             mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5963             mode != IN6_ADDR_GEN_MODE_RANDOM)
5964                 return -EINVAL;
5965         return 1;
5966 }
5967
5968 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5969                                 int mode)
5970 {
5971         if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5972             !idev->cnf.stable_secret.initialized &&
5973             !net->ipv6.devconf_dflt->stable_secret.initialized)
5974                 return -EINVAL;
5975         return 1;
5976 }
5977
5978 static int inet6_validate_link_af(const struct net_device *dev,
5979                                   const struct nlattr *nla,
5980                                   struct netlink_ext_ack *extack)
5981 {
5982         struct nlattr *tb[IFLA_INET6_MAX + 1];
5983         struct inet6_dev *idev = NULL;
5984         int err;
5985
5986         if (dev) {
5987                 idev = __in6_dev_get(dev);
5988                 if (!idev)
5989                         return -EAFNOSUPPORT;
5990         }
5991
5992         err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5993                                           inet6_af_policy, extack);
5994         if (err)
5995                 return err;
5996
5997         if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5998                 return -EINVAL;
5999
6000         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6001                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6002
6003                 if (check_addr_gen_mode(mode) < 0)
6004                         return -EINVAL;
6005                 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6006                         return -EINVAL;
6007         }
6008
6009         return 0;
6010 }
6011
6012 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6013                              struct netlink_ext_ack *extack)
6014 {
6015         struct inet6_dev *idev = __in6_dev_get(dev);
6016         struct nlattr *tb[IFLA_INET6_MAX + 1];
6017         int err;
6018
6019         if (!idev)
6020                 return -EAFNOSUPPORT;
6021
6022         if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6023                 return -EINVAL;
6024
6025         if (tb[IFLA_INET6_TOKEN]) {
6026                 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6027                                         extack);
6028                 if (err)
6029                         return err;
6030         }
6031
6032         if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6033                 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6034
6035                 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6036         }
6037
6038         return 0;
6039 }
6040
6041 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6042                              u32 portid, u32 seq, int event, unsigned int flags)
6043 {
6044         struct net_device *dev = idev->dev;
6045         struct ifinfomsg *hdr;
6046         struct nlmsghdr *nlh;
6047         int ifindex, iflink;
6048         void *protoinfo;
6049
6050         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6051         if (!nlh)
6052                 return -EMSGSIZE;
6053
6054         hdr = nlmsg_data(nlh);
6055         hdr->ifi_family = AF_INET6;
6056         hdr->__ifi_pad = 0;
6057         hdr->ifi_type = dev->type;
6058         ifindex = READ_ONCE(dev->ifindex);
6059         hdr->ifi_index = ifindex;
6060         hdr->ifi_flags = dev_get_flags(dev);
6061         hdr->ifi_change = 0;
6062
6063         iflink = dev_get_iflink(dev);
6064         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6065             (dev->addr_len &&
6066              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6067             nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6068             (ifindex != iflink &&
6069              nla_put_u32(skb, IFLA_LINK, iflink)) ||
6070             nla_put_u8(skb, IFLA_OPERSTATE,
6071                        netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6072                 goto nla_put_failure;
6073         protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6074         if (!protoinfo)
6075                 goto nla_put_failure;
6076
6077         if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6078                 goto nla_put_failure;
6079
6080         nla_nest_end(skb, protoinfo);
6081         nlmsg_end(skb, nlh);
6082         return 0;
6083
6084 nla_put_failure:
6085         nlmsg_cancel(skb, nlh);
6086         return -EMSGSIZE;
6087 }
6088
6089 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6090                                    struct netlink_ext_ack *extack)
6091 {
6092         struct ifinfomsg *ifm;
6093
6094         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6095                 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6096                 return -EINVAL;
6097         }
6098
6099         if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6100                 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6101                 return -EINVAL;
6102         }
6103
6104         ifm = nlmsg_data(nlh);
6105         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6106             ifm->ifi_change || ifm->ifi_index) {
6107                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6108                 return -EINVAL;
6109         }
6110
6111         return 0;
6112 }
6113
6114 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6115 {
6116         struct net *net = sock_net(skb->sk);
6117         struct {
6118                 unsigned long ifindex;
6119         } *ctx = (void *)cb->ctx;
6120         struct net_device *dev;
6121         struct inet6_dev *idev;
6122         int err;
6123
6124         /* only requests using strict checking can pass data to
6125          * influence the dump
6126          */
6127         if (cb->strict_check) {
6128                 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6129
6130                 if (err < 0)
6131                         return err;
6132         }
6133
6134         err = 0;
6135         rcu_read_lock();
6136         for_each_netdev_dump(net, dev, ctx->ifindex) {
6137                 idev = __in6_dev_get(dev);
6138                 if (!idev)
6139                         continue;
6140                 err = inet6_fill_ifinfo(skb, idev,
6141                                         NETLINK_CB(cb->skb).portid,
6142                                         cb->nlh->nlmsg_seq,
6143                                         RTM_NEWLINK, NLM_F_MULTI);
6144                 if (err < 0)
6145                         break;
6146         }
6147         rcu_read_unlock();
6148
6149         return err;
6150 }
6151
6152 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6153 {
6154         struct sk_buff *skb;
6155         struct net *net = dev_net(idev->dev);
6156         int err = -ENOBUFS;
6157
6158         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6159         if (!skb)
6160                 goto errout;
6161
6162         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6163         if (err < 0) {
6164                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6165                 WARN_ON(err == -EMSGSIZE);
6166                 kfree_skb(skb);
6167                 goto errout;
6168         }
6169         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6170         return;
6171 errout:
6172         if (err < 0)
6173                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6174 }
6175
6176 static inline size_t inet6_prefix_nlmsg_size(void)
6177 {
6178         return NLMSG_ALIGN(sizeof(struct prefixmsg))
6179                + nla_total_size(sizeof(struct in6_addr))
6180                + nla_total_size(sizeof(struct prefix_cacheinfo));
6181 }
6182
6183 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6184                              struct prefix_info *pinfo, u32 portid, u32 seq,
6185                              int event, unsigned int flags)
6186 {
6187         struct prefixmsg *pmsg;
6188         struct nlmsghdr *nlh;
6189         struct prefix_cacheinfo ci;
6190
6191         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6192         if (!nlh)
6193                 return -EMSGSIZE;
6194
6195         pmsg = nlmsg_data(nlh);
6196         pmsg->prefix_family = AF_INET6;
6197         pmsg->prefix_pad1 = 0;
6198         pmsg->prefix_pad2 = 0;
6199         pmsg->prefix_ifindex = idev->dev->ifindex;
6200         pmsg->prefix_len = pinfo->prefix_len;
6201         pmsg->prefix_type = pinfo->type;
6202         pmsg->prefix_pad3 = 0;
6203         pmsg->prefix_flags = pinfo->flags;
6204
6205         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6206                 goto nla_put_failure;
6207         ci.preferred_time = ntohl(pinfo->prefered);
6208         ci.valid_time = ntohl(pinfo->valid);
6209         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6210                 goto nla_put_failure;
6211         nlmsg_end(skb, nlh);
6212         return 0;
6213
6214 nla_put_failure:
6215         nlmsg_cancel(skb, nlh);
6216         return -EMSGSIZE;
6217 }
6218
6219 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6220                          struct prefix_info *pinfo)
6221 {
6222         struct sk_buff *skb;
6223         struct net *net = dev_net(idev->dev);
6224         int err = -ENOBUFS;
6225
6226         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6227         if (!skb)
6228                 goto errout;
6229
6230         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6231         if (err < 0) {
6232                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6233                 WARN_ON(err == -EMSGSIZE);
6234                 kfree_skb(skb);
6235                 goto errout;
6236         }
6237         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6238         return;
6239 errout:
6240         if (err < 0)
6241                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6242 }
6243
6244 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6245 {
6246         struct net *net = dev_net(ifp->idev->dev);
6247
6248         if (event)
6249                 ASSERT_RTNL();
6250
6251         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6252
6253         switch (event) {
6254         case RTM_NEWADDR:
6255                 /*
6256                  * If the address was optimistic we inserted the route at the
6257                  * start of our DAD process, so we don't need to do it again.
6258                  * If the device was taken down in the middle of the DAD
6259                  * cycle there is a race where we could get here without a
6260                  * host route, so nothing to insert. That will be fixed when
6261                  * the device is brought up.
6262                  */
6263                 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6264                         ip6_ins_rt(net, ifp->rt);
6265                 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6266                         pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6267                                 &ifp->addr, ifp->idev->dev->name);
6268                 }
6269
6270                 if (ifp->idev->cnf.forwarding)
6271                         addrconf_join_anycast(ifp);
6272                 if (!ipv6_addr_any(&ifp->peer_addr))
6273                         addrconf_prefix_route(&ifp->peer_addr, 128,
6274                                               ifp->rt_priority, ifp->idev->dev,
6275                                               0, 0, GFP_ATOMIC);
6276                 break;
6277         case RTM_DELADDR:
6278                 if (ifp->idev->cnf.forwarding)
6279                         addrconf_leave_anycast(ifp);
6280                 addrconf_leave_solict(ifp->idev, &ifp->addr);
6281                 if (!ipv6_addr_any(&ifp->peer_addr)) {
6282                         struct fib6_info *rt;
6283
6284                         rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6285                                                        ifp->idev->dev, 0, 0,
6286                                                        false);
6287                         if (rt)
6288                                 ip6_del_rt(net, rt, false);
6289                 }
6290                 if (ifp->rt) {
6291                         ip6_del_rt(net, ifp->rt, false);
6292                         ifp->rt = NULL;
6293                 }
6294                 rt_genid_bump_ipv6(net);
6295                 break;
6296         }
6297         atomic_inc(&net->ipv6.dev_addr_genid);
6298 }
6299
6300 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6301 {
6302         if (likely(ifp->idev->dead == 0))
6303                 __ipv6_ifa_notify(event, ifp);
6304 }
6305
6306 #ifdef CONFIG_SYSCTL
6307
6308 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6309                 void *buffer, size_t *lenp, loff_t *ppos)
6310 {
6311         int *valp = ctl->data;
6312         int val = *valp;
6313         loff_t pos = *ppos;
6314         struct ctl_table lctl;
6315         int ret;
6316
6317         /*
6318          * ctl->data points to idev->cnf.forwarding, we should
6319          * not modify it until we get the rtnl lock.
6320          */
6321         lctl = *ctl;
6322         lctl.data = &val;
6323
6324         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6325
6326         if (write)
6327                 ret = addrconf_fixup_forwarding(ctl, valp, val);
6328         if (ret)
6329                 *ppos = pos;
6330         return ret;
6331 }
6332
6333 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6334                 void *buffer, size_t *lenp, loff_t *ppos)
6335 {
6336         struct inet6_dev *idev = ctl->extra1;
6337         int min_mtu = IPV6_MIN_MTU;
6338         struct ctl_table lctl;
6339
6340         lctl = *ctl;
6341         lctl.extra1 = &min_mtu;
6342         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6343
6344         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6345 }
6346
6347 static void dev_disable_change(struct inet6_dev *idev)
6348 {
6349         struct netdev_notifier_info info;
6350
6351         if (!idev || !idev->dev)
6352                 return;
6353
6354         netdev_notifier_info_init(&info, idev->dev);
6355         if (idev->cnf.disable_ipv6)
6356                 addrconf_notify(NULL, NETDEV_DOWN, &info);
6357         else
6358                 addrconf_notify(NULL, NETDEV_UP, &info);
6359 }
6360
6361 static void addrconf_disable_change(struct net *net, __s32 newf)
6362 {
6363         struct net_device *dev;
6364         struct inet6_dev *idev;
6365
6366         for_each_netdev(net, dev) {
6367                 idev = __in6_dev_get(dev);
6368                 if (idev) {
6369                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6370
6371                         WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6372                         if (changed)
6373                                 dev_disable_change(idev);
6374                 }
6375         }
6376 }
6377
6378 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6379 {
6380         struct net *net = (struct net *)table->extra2;
6381         int old;
6382
6383         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6384                 WRITE_ONCE(*p, newf);
6385                 return 0;
6386         }
6387
6388         if (!rtnl_trylock())
6389                 return restart_syscall();
6390
6391         old = *p;
6392         WRITE_ONCE(*p, newf);
6393
6394         if (p == &net->ipv6.devconf_all->disable_ipv6) {
6395                 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6396                 addrconf_disable_change(net, newf);
6397         } else if ((!newf) ^ (!old))
6398                 dev_disable_change((struct inet6_dev *)table->extra1);
6399
6400         rtnl_unlock();
6401         return 0;
6402 }
6403
6404 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6405                 void *buffer, size_t *lenp, loff_t *ppos)
6406 {
6407         int *valp = ctl->data;
6408         int val = *valp;
6409         loff_t pos = *ppos;
6410         struct ctl_table lctl;
6411         int ret;
6412
6413         /*
6414          * ctl->data points to idev->cnf.disable_ipv6, we should
6415          * not modify it until we get the rtnl lock.
6416          */
6417         lctl = *ctl;
6418         lctl.data = &val;
6419
6420         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6421
6422         if (write)
6423                 ret = addrconf_disable_ipv6(ctl, valp, val);
6424         if (ret)
6425                 *ppos = pos;
6426         return ret;
6427 }
6428
6429 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6430                 void *buffer, size_t *lenp, loff_t *ppos)
6431 {
6432         int *valp = ctl->data;
6433         int ret;
6434         int old, new;
6435
6436         old = *valp;
6437         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6438         new = *valp;
6439
6440         if (write && old != new) {
6441                 struct net *net = ctl->extra2;
6442
6443                 if (!rtnl_trylock())
6444                         return restart_syscall();
6445
6446                 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6447                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6448                                                      NETCONFA_PROXY_NEIGH,
6449                                                      NETCONFA_IFINDEX_DEFAULT,
6450                                                      net->ipv6.devconf_dflt);
6451                 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6452                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6453                                                      NETCONFA_PROXY_NEIGH,
6454                                                      NETCONFA_IFINDEX_ALL,
6455                                                      net->ipv6.devconf_all);
6456                 else {
6457                         struct inet6_dev *idev = ctl->extra1;
6458
6459                         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6460                                                      NETCONFA_PROXY_NEIGH,
6461                                                      idev->dev->ifindex,
6462                                                      &idev->cnf);
6463                 }
6464                 rtnl_unlock();
6465         }
6466
6467         return ret;
6468 }
6469
6470 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6471                                          void *buffer, size_t *lenp,
6472                                          loff_t *ppos)
6473 {
6474         int ret = 0;
6475         u32 new_val;
6476         struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6477         struct net *net = (struct net *)ctl->extra2;
6478         struct ctl_table tmp = {
6479                 .data = &new_val,
6480                 .maxlen = sizeof(new_val),
6481                 .mode = ctl->mode,
6482         };
6483
6484         if (!rtnl_trylock())
6485                 return restart_syscall();
6486
6487         new_val = *((u32 *)ctl->data);
6488
6489         ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6490         if (ret != 0)
6491                 goto out;
6492
6493         if (write) {
6494                 if (check_addr_gen_mode(new_val) < 0) {
6495                         ret = -EINVAL;
6496                         goto out;
6497                 }
6498
6499                 if (idev) {
6500                         if (check_stable_privacy(idev, net, new_val) < 0) {
6501                                 ret = -EINVAL;
6502                                 goto out;
6503                         }
6504
6505                         if (idev->cnf.addr_gen_mode != new_val) {
6506                                 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6507                                 addrconf_init_auto_addrs(idev->dev);
6508                         }
6509                 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6510                         struct net_device *dev;
6511
6512                         WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6513                         for_each_netdev(net, dev) {
6514                                 idev = __in6_dev_get(dev);
6515                                 if (idev &&
6516                                     idev->cnf.addr_gen_mode != new_val) {
6517                                         WRITE_ONCE(idev->cnf.addr_gen_mode,
6518                                                   new_val);
6519                                         addrconf_init_auto_addrs(idev->dev);
6520                                 }
6521                         }
6522                 }
6523
6524                 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6525         }
6526
6527 out:
6528         rtnl_unlock();
6529
6530         return ret;
6531 }
6532
6533 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6534                                          void *buffer, size_t *lenp,
6535                                          loff_t *ppos)
6536 {
6537         int err;
6538         struct in6_addr addr;
6539         char str[IPV6_MAX_STRLEN];
6540         struct ctl_table lctl = *ctl;
6541         struct net *net = ctl->extra2;
6542         struct ipv6_stable_secret *secret = ctl->data;
6543
6544         if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6545                 return -EIO;
6546
6547         lctl.maxlen = IPV6_MAX_STRLEN;
6548         lctl.data = str;
6549
6550         if (!rtnl_trylock())
6551                 return restart_syscall();
6552
6553         if (!write && !secret->initialized) {
6554                 err = -EIO;
6555                 goto out;
6556         }
6557
6558         err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6559         if (err >= sizeof(str)) {
6560                 err = -EIO;
6561                 goto out;
6562         }
6563
6564         err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6565         if (err || !write)
6566                 goto out;
6567
6568         if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6569                 err = -EIO;
6570                 goto out;
6571         }
6572
6573         secret->initialized = true;
6574         secret->secret = addr;
6575
6576         if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6577                 struct net_device *dev;
6578
6579                 for_each_netdev(net, dev) {
6580                         struct inet6_dev *idev = __in6_dev_get(dev);
6581
6582                         if (idev) {
6583                                 WRITE_ONCE(idev->cnf.addr_gen_mode,
6584                                            IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6585                         }
6586                 }
6587         } else {
6588                 struct inet6_dev *idev = ctl->extra1;
6589
6590                 WRITE_ONCE(idev->cnf.addr_gen_mode,
6591                            IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6592         }
6593
6594 out:
6595         rtnl_unlock();
6596
6597         return err;
6598 }
6599
6600 static
6601 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6602                                                 int write, void *buffer,
6603                                                 size_t *lenp,
6604                                                 loff_t *ppos)
6605 {
6606         int *valp = ctl->data;
6607         int val = *valp;
6608         loff_t pos = *ppos;
6609         struct ctl_table lctl;
6610         int ret;
6611
6612         /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6613          * we should not modify it until we get the rtnl lock.
6614          */
6615         lctl = *ctl;
6616         lctl.data = &val;
6617
6618         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6619
6620         if (write)
6621                 ret = addrconf_fixup_linkdown(ctl, valp, val);
6622         if (ret)
6623                 *ppos = pos;
6624         return ret;
6625 }
6626
6627 static
6628 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6629 {
6630         if (rt) {
6631                 if (action)
6632                         rt->dst.flags |= DST_NOPOLICY;
6633                 else
6634                         rt->dst.flags &= ~DST_NOPOLICY;
6635         }
6636 }
6637
6638 static
6639 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6640 {
6641         struct inet6_ifaddr *ifa;
6642
6643         read_lock_bh(&idev->lock);
6644         list_for_each_entry(ifa, &idev->addr_list, if_list) {
6645                 spin_lock(&ifa->lock);
6646                 if (ifa->rt) {
6647                         /* host routes only use builtin fib6_nh */
6648                         struct fib6_nh *nh = ifa->rt->fib6_nh;
6649                         int cpu;
6650
6651                         rcu_read_lock();
6652                         ifa->rt->dst_nopolicy = val ? true : false;
6653                         if (nh->rt6i_pcpu) {
6654                                 for_each_possible_cpu(cpu) {
6655                                         struct rt6_info **rtp;
6656
6657                                         rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6658                                         addrconf_set_nopolicy(*rtp, val);
6659                                 }
6660                         }
6661                         rcu_read_unlock();
6662                 }
6663                 spin_unlock(&ifa->lock);
6664         }
6665         read_unlock_bh(&idev->lock);
6666 }
6667
6668 static
6669 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6670 {
6671         struct net *net = (struct net *)ctl->extra2;
6672         struct inet6_dev *idev;
6673
6674         if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6675                 WRITE_ONCE(*valp, val);
6676                 return 0;
6677         }
6678
6679         if (!rtnl_trylock())
6680                 return restart_syscall();
6681
6682         WRITE_ONCE(*valp, val);
6683
6684         if (valp == &net->ipv6.devconf_all->disable_policy)  {
6685                 struct net_device *dev;
6686
6687                 for_each_netdev(net, dev) {
6688                         idev = __in6_dev_get(dev);
6689                         if (idev)
6690                                 addrconf_disable_policy_idev(idev, val);
6691                 }
6692         } else {
6693                 idev = (struct inet6_dev *)ctl->extra1;
6694                 addrconf_disable_policy_idev(idev, val);
6695         }
6696
6697         rtnl_unlock();
6698         return 0;
6699 }
6700
6701 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6702                                    void *buffer, size_t *lenp, loff_t *ppos)
6703 {
6704         int *valp = ctl->data;
6705         int val = *valp;
6706         loff_t pos = *ppos;
6707         struct ctl_table lctl;
6708         int ret;
6709
6710         lctl = *ctl;
6711         lctl.data = &val;
6712         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6713
6714         if (write && (*valp != val))
6715                 ret = addrconf_disable_policy(ctl, valp, val);
6716
6717         if (ret)
6718                 *ppos = pos;
6719
6720         return ret;
6721 }
6722
6723 static int minus_one = -1;
6724 static const int two_five_five = 255;
6725 static u32 ioam6_if_id_max = U16_MAX;
6726
6727 static const struct ctl_table addrconf_sysctl[] = {
6728         {
6729                 .procname       = "forwarding",
6730                 .data           = &ipv6_devconf.forwarding,
6731                 .maxlen         = sizeof(int),
6732                 .mode           = 0644,
6733                 .proc_handler   = addrconf_sysctl_forward,
6734         },
6735         {
6736                 .procname       = "hop_limit",
6737                 .data           = &ipv6_devconf.hop_limit,
6738                 .maxlen         = sizeof(int),
6739                 .mode           = 0644,
6740                 .proc_handler   = proc_dointvec_minmax,
6741                 .extra1         = (void *)SYSCTL_ONE,
6742                 .extra2         = (void *)&two_five_five,
6743         },
6744         {
6745                 .procname       = "mtu",
6746                 .data           = &ipv6_devconf.mtu6,
6747                 .maxlen         = sizeof(int),
6748                 .mode           = 0644,
6749                 .proc_handler   = addrconf_sysctl_mtu,
6750         },
6751         {
6752                 .procname       = "accept_ra",
6753                 .data           = &ipv6_devconf.accept_ra,
6754                 .maxlen         = sizeof(int),
6755                 .mode           = 0644,
6756                 .proc_handler   = proc_dointvec,
6757         },
6758         {
6759                 .procname       = "accept_redirects",
6760                 .data           = &ipv6_devconf.accept_redirects,
6761                 .maxlen         = sizeof(int),
6762                 .mode           = 0644,
6763                 .proc_handler   = proc_dointvec,
6764         },
6765         {
6766                 .procname       = "autoconf",
6767                 .data           = &ipv6_devconf.autoconf,
6768                 .maxlen         = sizeof(int),
6769                 .mode           = 0644,
6770                 .proc_handler   = proc_dointvec,
6771         },
6772         {
6773                 .procname       = "dad_transmits",
6774                 .data           = &ipv6_devconf.dad_transmits,
6775                 .maxlen         = sizeof(int),
6776                 .mode           = 0644,
6777                 .proc_handler   = proc_dointvec,
6778         },
6779         {
6780                 .procname       = "router_solicitations",
6781                 .data           = &ipv6_devconf.rtr_solicits,
6782                 .maxlen         = sizeof(int),
6783                 .mode           = 0644,
6784                 .proc_handler   = proc_dointvec_minmax,
6785                 .extra1         = &minus_one,
6786         },
6787         {
6788                 .procname       = "router_solicitation_interval",
6789                 .data           = &ipv6_devconf.rtr_solicit_interval,
6790                 .maxlen         = sizeof(int),
6791                 .mode           = 0644,
6792                 .proc_handler   = proc_dointvec_jiffies,
6793         },
6794         {
6795                 .procname       = "router_solicitation_max_interval",
6796                 .data           = &ipv6_devconf.rtr_solicit_max_interval,
6797                 .maxlen         = sizeof(int),
6798                 .mode           = 0644,
6799                 .proc_handler   = proc_dointvec_jiffies,
6800         },
6801         {
6802                 .procname       = "router_solicitation_delay",
6803                 .data           = &ipv6_devconf.rtr_solicit_delay,
6804                 .maxlen         = sizeof(int),
6805                 .mode           = 0644,
6806                 .proc_handler   = proc_dointvec_jiffies,
6807         },
6808         {
6809                 .procname       = "force_mld_version",
6810                 .data           = &ipv6_devconf.force_mld_version,
6811                 .maxlen         = sizeof(int),
6812                 .mode           = 0644,
6813                 .proc_handler   = proc_dointvec,
6814         },
6815         {
6816                 .procname       = "mldv1_unsolicited_report_interval",
6817                 .data           =
6818                         &ipv6_devconf.mldv1_unsolicited_report_interval,
6819                 .maxlen         = sizeof(int),
6820                 .mode           = 0644,
6821                 .proc_handler   = proc_dointvec_ms_jiffies,
6822         },
6823         {
6824                 .procname       = "mldv2_unsolicited_report_interval",
6825                 .data           =
6826                         &ipv6_devconf.mldv2_unsolicited_report_interval,
6827                 .maxlen         = sizeof(int),
6828                 .mode           = 0644,
6829                 .proc_handler   = proc_dointvec_ms_jiffies,
6830         },
6831         {
6832                 .procname       = "use_tempaddr",
6833                 .data           = &ipv6_devconf.use_tempaddr,
6834                 .maxlen         = sizeof(int),
6835                 .mode           = 0644,
6836                 .proc_handler   = proc_dointvec,
6837         },
6838         {
6839                 .procname       = "temp_valid_lft",
6840                 .data           = &ipv6_devconf.temp_valid_lft,
6841                 .maxlen         = sizeof(int),
6842                 .mode           = 0644,
6843                 .proc_handler   = proc_dointvec,
6844         },
6845         {
6846                 .procname       = "temp_prefered_lft",
6847                 .data           = &ipv6_devconf.temp_prefered_lft,
6848                 .maxlen         = sizeof(int),
6849                 .mode           = 0644,
6850                 .proc_handler   = proc_dointvec,
6851         },
6852         {
6853                 .procname       = "regen_min_advance",
6854                 .data           = &ipv6_devconf.regen_min_advance,
6855                 .maxlen         = sizeof(int),
6856                 .mode           = 0644,
6857                 .proc_handler   = proc_dointvec,
6858         },
6859         {
6860                 .procname       = "regen_max_retry",
6861                 .data           = &ipv6_devconf.regen_max_retry,
6862                 .maxlen         = sizeof(int),
6863                 .mode           = 0644,
6864                 .proc_handler   = proc_dointvec,
6865         },
6866         {
6867                 .procname       = "max_desync_factor",
6868                 .data           = &ipv6_devconf.max_desync_factor,
6869                 .maxlen         = sizeof(int),
6870                 .mode           = 0644,
6871                 .proc_handler   = proc_dointvec,
6872         },
6873         {
6874                 .procname       = "max_addresses",
6875                 .data           = &ipv6_devconf.max_addresses,
6876                 .maxlen         = sizeof(int),
6877                 .mode           = 0644,
6878                 .proc_handler   = proc_dointvec,
6879         },
6880         {
6881                 .procname       = "accept_ra_defrtr",
6882                 .data           = &ipv6_devconf.accept_ra_defrtr,
6883                 .maxlen         = sizeof(int),
6884                 .mode           = 0644,
6885                 .proc_handler   = proc_dointvec,
6886         },
6887         {
6888                 .procname       = "ra_defrtr_metric",
6889                 .data           = &ipv6_devconf.ra_defrtr_metric,
6890                 .maxlen         = sizeof(u32),
6891                 .mode           = 0644,
6892                 .proc_handler   = proc_douintvec_minmax,
6893                 .extra1         = (void *)SYSCTL_ONE,
6894         },
6895         {
6896                 .procname       = "accept_ra_min_hop_limit",
6897                 .data           = &ipv6_devconf.accept_ra_min_hop_limit,
6898                 .maxlen         = sizeof(int),
6899                 .mode           = 0644,
6900                 .proc_handler   = proc_dointvec,
6901         },
6902         {
6903                 .procname       = "accept_ra_min_lft",
6904                 .data           = &ipv6_devconf.accept_ra_min_lft,
6905                 .maxlen         = sizeof(int),
6906                 .mode           = 0644,
6907                 .proc_handler   = proc_dointvec,
6908         },
6909         {
6910                 .procname       = "accept_ra_pinfo",
6911                 .data           = &ipv6_devconf.accept_ra_pinfo,
6912                 .maxlen         = sizeof(int),
6913                 .mode           = 0644,
6914                 .proc_handler   = proc_dointvec,
6915         },
6916         {
6917                 .procname       = "ra_honor_pio_life",
6918                 .data           = &ipv6_devconf.ra_honor_pio_life,
6919                 .maxlen         = sizeof(u8),
6920                 .mode           = 0644,
6921                 .proc_handler   = proc_dou8vec_minmax,
6922                 .extra1         = SYSCTL_ZERO,
6923                 .extra2         = SYSCTL_ONE,
6924         },
6925 #ifdef CONFIG_IPV6_ROUTER_PREF
6926         {
6927                 .procname       = "accept_ra_rtr_pref",
6928                 .data           = &ipv6_devconf.accept_ra_rtr_pref,
6929                 .maxlen         = sizeof(int),
6930                 .mode           = 0644,
6931                 .proc_handler   = proc_dointvec,
6932         },
6933         {
6934                 .procname       = "router_probe_interval",
6935                 .data           = &ipv6_devconf.rtr_probe_interval,
6936                 .maxlen         = sizeof(int),
6937                 .mode           = 0644,
6938                 .proc_handler   = proc_dointvec_jiffies,
6939         },
6940 #ifdef CONFIG_IPV6_ROUTE_INFO
6941         {
6942                 .procname       = "accept_ra_rt_info_min_plen",
6943                 .data           = &ipv6_devconf.accept_ra_rt_info_min_plen,
6944                 .maxlen         = sizeof(int),
6945                 .mode           = 0644,
6946                 .proc_handler   = proc_dointvec,
6947         },
6948         {
6949                 .procname       = "accept_ra_rt_info_max_plen",
6950                 .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
6951                 .maxlen         = sizeof(int),
6952                 .mode           = 0644,
6953                 .proc_handler   = proc_dointvec,
6954         },
6955 #endif
6956 #endif
6957         {
6958                 .procname       = "proxy_ndp",
6959                 .data           = &ipv6_devconf.proxy_ndp,
6960                 .maxlen         = sizeof(int),
6961                 .mode           = 0644,
6962                 .proc_handler   = addrconf_sysctl_proxy_ndp,
6963         },
6964         {
6965                 .procname       = "accept_source_route",
6966                 .data           = &ipv6_devconf.accept_source_route,
6967                 .maxlen         = sizeof(int),
6968                 .mode           = 0644,
6969                 .proc_handler   = proc_dointvec,
6970         },
6971 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6972         {
6973                 .procname       = "optimistic_dad",
6974                 .data           = &ipv6_devconf.optimistic_dad,
6975                 .maxlen         = sizeof(int),
6976                 .mode           = 0644,
6977                 .proc_handler   = proc_dointvec,
6978         },
6979         {
6980                 .procname       = "use_optimistic",
6981                 .data           = &ipv6_devconf.use_optimistic,
6982                 .maxlen         = sizeof(int),
6983                 .mode           = 0644,
6984                 .proc_handler   = proc_dointvec,
6985         },
6986 #endif
6987 #ifdef CONFIG_IPV6_MROUTE
6988         {
6989                 .procname       = "mc_forwarding",
6990                 .data           = &ipv6_devconf.mc_forwarding,
6991                 .maxlen         = sizeof(int),
6992                 .mode           = 0444,
6993                 .proc_handler   = proc_dointvec,
6994         },
6995 #endif
6996         {
6997                 .procname       = "disable_ipv6",
6998                 .data           = &ipv6_devconf.disable_ipv6,
6999                 .maxlen         = sizeof(int),
7000                 .mode           = 0644,
7001                 .proc_handler   = addrconf_sysctl_disable,
7002         },
7003         {
7004                 .procname       = "accept_dad",
7005                 .data           = &ipv6_devconf.accept_dad,
7006                 .maxlen         = sizeof(int),
7007                 .mode           = 0644,
7008                 .proc_handler   = proc_dointvec,
7009         },
7010         {
7011                 .procname       = "force_tllao",
7012                 .data           = &ipv6_devconf.force_tllao,
7013                 .maxlen         = sizeof(int),
7014                 .mode           = 0644,
7015                 .proc_handler   = proc_dointvec
7016         },
7017         {
7018                 .procname       = "ndisc_notify",
7019                 .data           = &ipv6_devconf.ndisc_notify,
7020                 .maxlen         = sizeof(int),
7021                 .mode           = 0644,
7022                 .proc_handler   = proc_dointvec
7023         },
7024         {
7025                 .procname       = "suppress_frag_ndisc",
7026                 .data           = &ipv6_devconf.suppress_frag_ndisc,
7027                 .maxlen         = sizeof(int),
7028                 .mode           = 0644,
7029                 .proc_handler   = proc_dointvec
7030         },
7031         {
7032                 .procname       = "accept_ra_from_local",
7033                 .data           = &ipv6_devconf.accept_ra_from_local,
7034                 .maxlen         = sizeof(int),
7035                 .mode           = 0644,
7036                 .proc_handler   = proc_dointvec,
7037         },
7038         {
7039                 .procname       = "accept_ra_mtu",
7040                 .data           = &ipv6_devconf.accept_ra_mtu,
7041                 .maxlen         = sizeof(int),
7042                 .mode           = 0644,
7043                 .proc_handler   = proc_dointvec,
7044         },
7045         {
7046                 .procname       = "stable_secret",
7047                 .data           = &ipv6_devconf.stable_secret,
7048                 .maxlen         = IPV6_MAX_STRLEN,
7049                 .mode           = 0600,
7050                 .proc_handler   = addrconf_sysctl_stable_secret,
7051         },
7052         {
7053                 .procname       = "use_oif_addrs_only",
7054                 .data           = &ipv6_devconf.use_oif_addrs_only,
7055                 .maxlen         = sizeof(int),
7056                 .mode           = 0644,
7057                 .proc_handler   = proc_dointvec,
7058         },
7059         {
7060                 .procname       = "ignore_routes_with_linkdown",
7061                 .data           = &ipv6_devconf.ignore_routes_with_linkdown,
7062                 .maxlen         = sizeof(int),
7063                 .mode           = 0644,
7064                 .proc_handler   = addrconf_sysctl_ignore_routes_with_linkdown,
7065         },
7066         {
7067                 .procname       = "drop_unicast_in_l2_multicast",
7068                 .data           = &ipv6_devconf.drop_unicast_in_l2_multicast,
7069                 .maxlen         = sizeof(int),
7070                 .mode           = 0644,
7071                 .proc_handler   = proc_dointvec,
7072         },
7073         {
7074                 .procname       = "drop_unsolicited_na",
7075                 .data           = &ipv6_devconf.drop_unsolicited_na,
7076                 .maxlen         = sizeof(int),
7077                 .mode           = 0644,
7078                 .proc_handler   = proc_dointvec,
7079         },
7080         {
7081                 .procname       = "keep_addr_on_down",
7082                 .data           = &ipv6_devconf.keep_addr_on_down,
7083                 .maxlen         = sizeof(int),
7084                 .mode           = 0644,
7085                 .proc_handler   = proc_dointvec,
7086
7087         },
7088         {
7089                 .procname       = "seg6_enabled",
7090                 .data           = &ipv6_devconf.seg6_enabled,
7091                 .maxlen         = sizeof(int),
7092                 .mode           = 0644,
7093                 .proc_handler   = proc_dointvec,
7094         },
7095 #ifdef CONFIG_IPV6_SEG6_HMAC
7096         {
7097                 .procname       = "seg6_require_hmac",
7098                 .data           = &ipv6_devconf.seg6_require_hmac,
7099                 .maxlen         = sizeof(int),
7100                 .mode           = 0644,
7101                 .proc_handler   = proc_dointvec,
7102         },
7103 #endif
7104         {
7105                 .procname       = "enhanced_dad",
7106                 .data           = &ipv6_devconf.enhanced_dad,
7107                 .maxlen         = sizeof(int),
7108                 .mode           = 0644,
7109                 .proc_handler   = proc_dointvec,
7110         },
7111         {
7112                 .procname       = "addr_gen_mode",
7113                 .data           = &ipv6_devconf.addr_gen_mode,
7114                 .maxlen         = sizeof(int),
7115                 .mode           = 0644,
7116                 .proc_handler   = addrconf_sysctl_addr_gen_mode,
7117         },
7118         {
7119                 .procname       = "disable_policy",
7120                 .data           = &ipv6_devconf.disable_policy,
7121                 .maxlen         = sizeof(int),
7122                 .mode           = 0644,
7123                 .proc_handler   = addrconf_sysctl_disable_policy,
7124         },
7125         {
7126                 .procname       = "ndisc_tclass",
7127                 .data           = &ipv6_devconf.ndisc_tclass,
7128                 .maxlen         = sizeof(int),
7129                 .mode           = 0644,
7130                 .proc_handler   = proc_dointvec_minmax,
7131                 .extra1         = (void *)SYSCTL_ZERO,
7132                 .extra2         = (void *)&two_five_five,
7133         },
7134         {
7135                 .procname       = "rpl_seg_enabled",
7136                 .data           = &ipv6_devconf.rpl_seg_enabled,
7137                 .maxlen         = sizeof(int),
7138                 .mode           = 0644,
7139                 .proc_handler   = proc_dointvec,
7140         },
7141         {
7142                 .procname       = "ioam6_enabled",
7143                 .data           = &ipv6_devconf.ioam6_enabled,
7144                 .maxlen         = sizeof(u8),
7145                 .mode           = 0644,
7146                 .proc_handler   = proc_dou8vec_minmax,
7147                 .extra1         = (void *)SYSCTL_ZERO,
7148                 .extra2         = (void *)SYSCTL_ONE,
7149         },
7150         {
7151                 .procname       = "ioam6_id",
7152                 .data           = &ipv6_devconf.ioam6_id,
7153                 .maxlen         = sizeof(u32),
7154                 .mode           = 0644,
7155                 .proc_handler   = proc_douintvec_minmax,
7156                 .extra1         = (void *)SYSCTL_ZERO,
7157                 .extra2         = (void *)&ioam6_if_id_max,
7158         },
7159         {
7160                 .procname       = "ioam6_id_wide",
7161                 .data           = &ipv6_devconf.ioam6_id_wide,
7162                 .maxlen         = sizeof(u32),
7163                 .mode           = 0644,
7164                 .proc_handler   = proc_douintvec,
7165         },
7166         {
7167                 .procname       = "ndisc_evict_nocarrier",
7168                 .data           = &ipv6_devconf.ndisc_evict_nocarrier,
7169                 .maxlen         = sizeof(u8),
7170                 .mode           = 0644,
7171                 .proc_handler   = proc_dou8vec_minmax,
7172                 .extra1         = (void *)SYSCTL_ZERO,
7173                 .extra2         = (void *)SYSCTL_ONE,
7174         },
7175         {
7176                 .procname       = "accept_untracked_na",
7177                 .data           = &ipv6_devconf.accept_untracked_na,
7178                 .maxlen         = sizeof(int),
7179                 .mode           = 0644,
7180                 .proc_handler   = proc_dointvec_minmax,
7181                 .extra1         = SYSCTL_ZERO,
7182                 .extra2         = SYSCTL_TWO,
7183         },
7184         {
7185                 /* sentinel */
7186         }
7187 };
7188
7189 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7190                 struct inet6_dev *idev, struct ipv6_devconf *p)
7191 {
7192         int i, ifindex;
7193         struct ctl_table *table;
7194         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7195
7196         table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7197         if (!table)
7198                 goto out;
7199
7200         for (i = 0; table[i].data; i++) {
7201                 table[i].data += (char *)p - (char *)&ipv6_devconf;
7202                 /* If one of these is already set, then it is not safe to
7203                  * overwrite either of them: this makes proc_dointvec_minmax
7204                  * usable.
7205                  */
7206                 if (!table[i].extra1 && !table[i].extra2) {
7207                         table[i].extra1 = idev; /* embedded; no ref */
7208                         table[i].extra2 = net;
7209                 }
7210         }
7211
7212         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7213
7214         p->sysctl_header = register_net_sysctl_sz(net, path, table,
7215                                                   ARRAY_SIZE(addrconf_sysctl));
7216         if (!p->sysctl_header)
7217                 goto free;
7218
7219         if (!strcmp(dev_name, "all"))
7220                 ifindex = NETCONFA_IFINDEX_ALL;
7221         else if (!strcmp(dev_name, "default"))
7222                 ifindex = NETCONFA_IFINDEX_DEFAULT;
7223         else
7224                 ifindex = idev->dev->ifindex;
7225         inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7226                                      ifindex, p);
7227         return 0;
7228
7229 free:
7230         kfree(table);
7231 out:
7232         return -ENOBUFS;
7233 }
7234
7235 static void __addrconf_sysctl_unregister(struct net *net,
7236                                          struct ipv6_devconf *p, int ifindex)
7237 {
7238         struct ctl_table *table;
7239
7240         if (!p->sysctl_header)
7241                 return;
7242
7243         table = p->sysctl_header->ctl_table_arg;
7244         unregister_net_sysctl_table(p->sysctl_header);
7245         p->sysctl_header = NULL;
7246         kfree(table);
7247
7248         inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7249 }
7250
7251 static int addrconf_sysctl_register(struct inet6_dev *idev)
7252 {
7253         int err;
7254
7255         if (!sysctl_dev_name_is_allowed(idev->dev->name))
7256                 return -EINVAL;
7257
7258         err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7259                                     &ndisc_ifinfo_sysctl_change);
7260         if (err)
7261                 return err;
7262         err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7263                                          idev, &idev->cnf);
7264         if (err)
7265                 neigh_sysctl_unregister(idev->nd_parms);
7266
7267         return err;
7268 }
7269
7270 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7271 {
7272         __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7273                                      idev->dev->ifindex);
7274         neigh_sysctl_unregister(idev->nd_parms);
7275 }
7276
7277
7278 #endif
7279
7280 static int __net_init addrconf_init_net(struct net *net)
7281 {
7282         int err = -ENOMEM;
7283         struct ipv6_devconf *all, *dflt;
7284
7285         spin_lock_init(&net->ipv6.addrconf_hash_lock);
7286         INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7287         net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7288                                            sizeof(struct hlist_head),
7289                                            GFP_KERNEL);
7290         if (!net->ipv6.inet6_addr_lst)
7291                 goto err_alloc_addr;
7292
7293         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7294         if (!all)
7295                 goto err_alloc_all;
7296
7297         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7298         if (!dflt)
7299                 goto err_alloc_dflt;
7300
7301         if (!net_eq(net, &init_net)) {
7302                 switch (net_inherit_devconf()) {
7303                 case 1:  /* copy from init_net */
7304                         memcpy(all, init_net.ipv6.devconf_all,
7305                                sizeof(ipv6_devconf));
7306                         memcpy(dflt, init_net.ipv6.devconf_dflt,
7307                                sizeof(ipv6_devconf_dflt));
7308                         break;
7309                 case 3: /* copy from the current netns */
7310                         memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7311                                sizeof(ipv6_devconf));
7312                         memcpy(dflt,
7313                                current->nsproxy->net_ns->ipv6.devconf_dflt,
7314                                sizeof(ipv6_devconf_dflt));
7315                         break;
7316                 case 0:
7317                 case 2:
7318                         /* use compiled values */
7319                         break;
7320                 }
7321         }
7322
7323         /* these will be inherited by all namespaces */
7324         dflt->autoconf = ipv6_defaults.autoconf;
7325         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7326
7327         dflt->stable_secret.initialized = false;
7328         all->stable_secret.initialized = false;
7329
7330         net->ipv6.devconf_all = all;
7331         net->ipv6.devconf_dflt = dflt;
7332
7333 #ifdef CONFIG_SYSCTL
7334         err = __addrconf_sysctl_register(net, "all", NULL, all);
7335         if (err < 0)
7336                 goto err_reg_all;
7337
7338         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7339         if (err < 0)
7340                 goto err_reg_dflt;
7341 #endif
7342         return 0;
7343
7344 #ifdef CONFIG_SYSCTL
7345 err_reg_dflt:
7346         __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7347 err_reg_all:
7348         kfree(dflt);
7349         net->ipv6.devconf_dflt = NULL;
7350 #endif
7351 err_alloc_dflt:
7352         kfree(all);
7353         net->ipv6.devconf_all = NULL;
7354 err_alloc_all:
7355         kfree(net->ipv6.inet6_addr_lst);
7356 err_alloc_addr:
7357         return err;
7358 }
7359
7360 static void __net_exit addrconf_exit_net(struct net *net)
7361 {
7362         int i;
7363
7364 #ifdef CONFIG_SYSCTL
7365         __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7366                                      NETCONFA_IFINDEX_DEFAULT);
7367         __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7368                                      NETCONFA_IFINDEX_ALL);
7369 #endif
7370         kfree(net->ipv6.devconf_dflt);
7371         net->ipv6.devconf_dflt = NULL;
7372         kfree(net->ipv6.devconf_all);
7373         net->ipv6.devconf_all = NULL;
7374
7375         cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7376         /*
7377          *      Check hash table, then free it.
7378          */
7379         for (i = 0; i < IN6_ADDR_HSIZE; i++)
7380                 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7381
7382         kfree(net->ipv6.inet6_addr_lst);
7383         net->ipv6.inet6_addr_lst = NULL;
7384 }
7385
7386 static struct pernet_operations addrconf_ops = {
7387         .init = addrconf_init_net,
7388         .exit = addrconf_exit_net,
7389 };
7390
7391 static struct rtnl_af_ops inet6_ops __read_mostly = {
7392         .family           = AF_INET6,
7393         .fill_link_af     = inet6_fill_link_af,
7394         .get_link_af_size = inet6_get_link_af_size,
7395         .validate_link_af = inet6_validate_link_af,
7396         .set_link_af      = inet6_set_link_af,
7397 };
7398
7399 /*
7400  *      Init / cleanup code
7401  */
7402
7403 int __init addrconf_init(void)
7404 {
7405         struct inet6_dev *idev;
7406         int err;
7407
7408         err = ipv6_addr_label_init();
7409         if (err < 0) {
7410                 pr_crit("%s: cannot initialize default policy table: %d\n",
7411                         __func__, err);
7412                 goto out;
7413         }
7414
7415         err = register_pernet_subsys(&addrconf_ops);
7416         if (err < 0)
7417                 goto out_addrlabel;
7418
7419         /* All works using addrconf_wq need to lock rtnl. */
7420         addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7421         if (!addrconf_wq) {
7422                 err = -ENOMEM;
7423                 goto out_nowq;
7424         }
7425
7426         rtnl_lock();
7427         idev = ipv6_add_dev(blackhole_netdev);
7428         rtnl_unlock();
7429         if (IS_ERR(idev)) {
7430                 err = PTR_ERR(idev);
7431                 goto errlo;
7432         }
7433
7434         ip6_route_init_special_entries();
7435
7436         register_netdevice_notifier(&ipv6_dev_notf);
7437
7438         addrconf_verify(&init_net);
7439
7440         rtnl_af_register(&inet6_ops);
7441
7442         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7443                                    NULL, inet6_dump_ifinfo, RTNL_FLAG_DUMP_UNLOCKED);
7444         if (err < 0)
7445                 goto errout;
7446
7447         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7448                                    inet6_rtm_newaddr, NULL, 0);
7449         if (err < 0)
7450                 goto errout;
7451         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7452                                    inet6_rtm_deladdr, NULL, 0);
7453         if (err < 0)
7454                 goto errout;
7455         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7456                                    inet6_rtm_getaddr, inet6_dump_ifaddr,
7457                                    RTNL_FLAG_DOIT_UNLOCKED |
7458                                    RTNL_FLAG_DUMP_UNLOCKED);
7459         if (err < 0)
7460                 goto errout;
7461         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7462                                    NULL, inet6_dump_ifmcaddr,
7463                                    RTNL_FLAG_DUMP_UNLOCKED);
7464         if (err < 0)
7465                 goto errout;
7466         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7467                                    NULL, inet6_dump_ifacaddr,
7468                                    RTNL_FLAG_DUMP_UNLOCKED);
7469         if (err < 0)
7470                 goto errout;
7471         err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7472                                    inet6_netconf_get_devconf,
7473                                    inet6_netconf_dump_devconf,
7474                                    RTNL_FLAG_DOIT_UNLOCKED |
7475                                    RTNL_FLAG_DUMP_UNLOCKED);
7476         if (err < 0)
7477                 goto errout;
7478         err = ipv6_addr_label_rtnl_register();
7479         if (err < 0)
7480                 goto errout;
7481
7482         return 0;
7483 errout:
7484         rtnl_unregister_all(PF_INET6);
7485         rtnl_af_unregister(&inet6_ops);
7486         unregister_netdevice_notifier(&ipv6_dev_notf);
7487 errlo:
7488         destroy_workqueue(addrconf_wq);
7489 out_nowq:
7490         unregister_pernet_subsys(&addrconf_ops);
7491 out_addrlabel:
7492         ipv6_addr_label_cleanup();
7493 out:
7494         return err;
7495 }
7496
7497 void addrconf_cleanup(void)
7498 {
7499         struct net_device *dev;
7500
7501         unregister_netdevice_notifier(&ipv6_dev_notf);
7502         unregister_pernet_subsys(&addrconf_ops);
7503         ipv6_addr_label_cleanup();
7504
7505         rtnl_af_unregister(&inet6_ops);
7506
7507         rtnl_lock();
7508
7509         /* clean dev list */
7510         for_each_netdev(&init_net, dev) {
7511                 if (__in6_dev_get(dev) == NULL)
7512                         continue;
7513                 addrconf_ifdown(dev, true);
7514         }
7515         addrconf_ifdown(init_net.loopback_dev, true);
7516
7517         rtnl_unlock();
7518
7519         destroy_workqueue(addrconf_wq);
7520 }
This page took 0.458945 seconds and 4 git commands to generate.