Merge tag 'vfio-v3.20-rc1' of git://github.com/awilliam/linux-vfio
[deliverable/linux.git] / net / ipv6 / addrconf.c
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 */
14
15 /*
16 * Changes:
17 *
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
39 */
40
41 #define pr_fmt(fmt) "IPv6: " fmt
42
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
59 #ifdef CONFIG_SYSCTL
60 #include <linux/sysctl.h>
61 #endif
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
67
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
71
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
74 #include <net/ipv6.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
79 #include <net/tcp.h>
80 #include <net/ip.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
89
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
93
94 /* Set to 3 to get tracing... */
95 #define ACONF_DEBUG 2
96
97 #if ACONF_DEBUG >= 3
98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
99 #else
100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
101 #endif
102
103 #define INFINITY_LIFE_TIME 0xFFFFFFFF
104
105 static inline u32 cstamp_delta(unsigned long cstamp)
106 {
107 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
108 }
109
110 #ifdef CONFIG_SYSCTL
111 static int addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
113 #else
114 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
115 {
116 return 0;
117 }
118
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
120 {
121 }
122 #endif
123
124 static void __ipv6_regen_rndid(struct inet6_dev *idev);
125 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
126 static void ipv6_regen_rndid(unsigned long data);
127
128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
129 static int ipv6_count_addresses(struct inet6_dev *idev);
130
131 /*
132 * Configured unicast address hash table
133 */
134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
135 static DEFINE_SPINLOCK(addrconf_hash_lock);
136
137 static void addrconf_verify(void);
138 static void addrconf_verify_rtnl(void);
139 static void addrconf_verify_work(struct work_struct *);
140
141 static struct workqueue_struct *addrconf_wq;
142 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
143
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146
147 static void addrconf_type_change(struct net_device *dev,
148 unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
150
151 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
152 int plen,
153 const struct net_device *dev,
154 u32 flags, u32 noflags);
155
156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
157 static void addrconf_dad_work(struct work_struct *w);
158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_run(struct inet6_dev *idev);
160 static void addrconf_rs_timer(unsigned long data);
161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163
164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
165 struct prefix_info *pinfo);
166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
167 struct net_device *dev);
168
169 static struct ipv6_devconf ipv6_devconf __read_mostly = {
170 .forwarding = 0,
171 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
172 .mtu6 = IPV6_MIN_MTU,
173 .accept_ra = 1,
174 .accept_redirects = 1,
175 .autoconf = 1,
176 .force_mld_version = 0,
177 .mldv1_unsolicited_report_interval = 10 * HZ,
178 .mldv2_unsolicited_report_interval = HZ,
179 .dad_transmits = 1,
180 .rtr_solicits = MAX_RTR_SOLICITATIONS,
181 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
182 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
183 .use_tempaddr = 0,
184 .temp_valid_lft = TEMP_VALID_LIFETIME,
185 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
186 .regen_max_retry = REGEN_MAX_RETRY,
187 .max_desync_factor = MAX_DESYNC_FACTOR,
188 .max_addresses = IPV6_MAX_ADDRESSES,
189 .accept_ra_defrtr = 1,
190 .accept_ra_from_local = 0,
191 .accept_ra_pinfo = 1,
192 #ifdef CONFIG_IPV6_ROUTER_PREF
193 .accept_ra_rtr_pref = 1,
194 .rtr_probe_interval = 60 * HZ,
195 #ifdef CONFIG_IPV6_ROUTE_INFO
196 .accept_ra_rt_info_max_plen = 0,
197 #endif
198 #endif
199 .proxy_ndp = 0,
200 .accept_source_route = 0, /* we do not accept RH0 by default. */
201 .disable_ipv6 = 0,
202 .accept_dad = 1,
203 .suppress_frag_ndisc = 1,
204 .accept_ra_mtu = 1,
205 };
206
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
208 .forwarding = 0,
209 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
210 .mtu6 = IPV6_MIN_MTU,
211 .accept_ra = 1,
212 .accept_redirects = 1,
213 .autoconf = 1,
214 .force_mld_version = 0,
215 .mldv1_unsolicited_report_interval = 10 * HZ,
216 .mldv2_unsolicited_report_interval = HZ,
217 .dad_transmits = 1,
218 .rtr_solicits = MAX_RTR_SOLICITATIONS,
219 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
220 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
221 .use_tempaddr = 0,
222 .temp_valid_lft = TEMP_VALID_LIFETIME,
223 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
224 .regen_max_retry = REGEN_MAX_RETRY,
225 .max_desync_factor = MAX_DESYNC_FACTOR,
226 .max_addresses = IPV6_MAX_ADDRESSES,
227 .accept_ra_defrtr = 1,
228 .accept_ra_from_local = 0,
229 .accept_ra_pinfo = 1,
230 #ifdef CONFIG_IPV6_ROUTER_PREF
231 .accept_ra_rtr_pref = 1,
232 .rtr_probe_interval = 60 * HZ,
233 #ifdef CONFIG_IPV6_ROUTE_INFO
234 .accept_ra_rt_info_max_plen = 0,
235 #endif
236 #endif
237 .proxy_ndp = 0,
238 .accept_source_route = 0, /* we do not accept RH0 by default. */
239 .disable_ipv6 = 0,
240 .accept_dad = 1,
241 .suppress_frag_ndisc = 1,
242 .accept_ra_mtu = 1,
243 };
244
245 /* Check if a valid qdisc is available */
246 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
247 {
248 return !qdisc_tx_is_noop(dev);
249 }
250
251 static void addrconf_del_rs_timer(struct inet6_dev *idev)
252 {
253 if (del_timer(&idev->rs_timer))
254 __in6_dev_put(idev);
255 }
256
257 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
258 {
259 if (cancel_delayed_work(&ifp->dad_work))
260 __in6_ifa_put(ifp);
261 }
262
263 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
264 unsigned long when)
265 {
266 if (!timer_pending(&idev->rs_timer))
267 in6_dev_hold(idev);
268 mod_timer(&idev->rs_timer, jiffies + when);
269 }
270
271 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
272 unsigned long delay)
273 {
274 if (!delayed_work_pending(&ifp->dad_work))
275 in6_ifa_hold(ifp);
276 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
277 }
278
279 static int snmp6_alloc_dev(struct inet6_dev *idev)
280 {
281 int i;
282
283 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
284 if (!idev->stats.ipv6)
285 goto err_ip;
286
287 for_each_possible_cpu(i) {
288 struct ipstats_mib *addrconf_stats;
289 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
290 u64_stats_init(&addrconf_stats->syncp);
291 }
292
293
294 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
295 GFP_KERNEL);
296 if (!idev->stats.icmpv6dev)
297 goto err_icmp;
298 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
299 GFP_KERNEL);
300 if (!idev->stats.icmpv6msgdev)
301 goto err_icmpmsg;
302
303 return 0;
304
305 err_icmpmsg:
306 kfree(idev->stats.icmpv6dev);
307 err_icmp:
308 free_percpu(idev->stats.ipv6);
309 err_ip:
310 return -ENOMEM;
311 }
312
313 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
314 {
315 struct inet6_dev *ndev;
316 int err = -ENOMEM;
317
318 ASSERT_RTNL();
319
320 if (dev->mtu < IPV6_MIN_MTU)
321 return ERR_PTR(-EINVAL);
322
323 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
324 if (ndev == NULL)
325 return ERR_PTR(err);
326
327 rwlock_init(&ndev->lock);
328 ndev->dev = dev;
329 INIT_LIST_HEAD(&ndev->addr_list);
330 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
331 (unsigned long)ndev);
332 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
333 ndev->cnf.mtu6 = dev->mtu;
334 ndev->cnf.sysctl = NULL;
335 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
336 if (ndev->nd_parms == NULL) {
337 kfree(ndev);
338 return ERR_PTR(err);
339 }
340 if (ndev->cnf.forwarding)
341 dev_disable_lro(dev);
342 /* We refer to the device */
343 dev_hold(dev);
344
345 if (snmp6_alloc_dev(ndev) < 0) {
346 ADBG(KERN_WARNING
347 "%s: cannot allocate memory for statistics; dev=%s.\n",
348 __func__, dev->name);
349 neigh_parms_release(&nd_tbl, ndev->nd_parms);
350 dev_put(dev);
351 kfree(ndev);
352 return ERR_PTR(err);
353 }
354
355 if (snmp6_register_dev(ndev) < 0) {
356 ADBG(KERN_WARNING
357 "%s: cannot create /proc/net/dev_snmp6/%s\n",
358 __func__, dev->name);
359 goto err_release;
360 }
361
362 /* One reference from device. We must do this before
363 * we invoke __ipv6_regen_rndid().
364 */
365 in6_dev_hold(ndev);
366
367 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
368 ndev->cnf.accept_dad = -1;
369
370 #if IS_ENABLED(CONFIG_IPV6_SIT)
371 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
372 pr_info("%s: Disabled Multicast RS\n", dev->name);
373 ndev->cnf.rtr_solicits = 0;
374 }
375 #endif
376
377 INIT_LIST_HEAD(&ndev->tempaddr_list);
378 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
379 if ((dev->flags&IFF_LOOPBACK) ||
380 dev->type == ARPHRD_TUNNEL ||
381 dev->type == ARPHRD_TUNNEL6 ||
382 dev->type == ARPHRD_SIT ||
383 dev->type == ARPHRD_NONE) {
384 ndev->cnf.use_tempaddr = -1;
385 } else {
386 in6_dev_hold(ndev);
387 ipv6_regen_rndid((unsigned long) ndev);
388 }
389
390 ndev->token = in6addr_any;
391
392 if (netif_running(dev) && addrconf_qdisc_ok(dev))
393 ndev->if_flags |= IF_READY;
394
395 ipv6_mc_init_dev(ndev);
396 ndev->tstamp = jiffies;
397 err = addrconf_sysctl_register(ndev);
398 if (err) {
399 ipv6_mc_destroy_dev(ndev);
400 del_timer(&ndev->regen_timer);
401 goto err_release;
402 }
403 /* protected by rtnl_lock */
404 rcu_assign_pointer(dev->ip6_ptr, ndev);
405
406 /* Join interface-local all-node multicast group */
407 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
408
409 /* Join all-node multicast group */
410 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
411
412 /* Join all-router multicast group if forwarding is set */
413 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
414 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
415
416 return ndev;
417
418 err_release:
419 neigh_parms_release(&nd_tbl, ndev->nd_parms);
420 ndev->dead = 1;
421 in6_dev_finish_destroy(ndev);
422 return ERR_PTR(err);
423 }
424
425 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
426 {
427 struct inet6_dev *idev;
428
429 ASSERT_RTNL();
430
431 idev = __in6_dev_get(dev);
432 if (!idev) {
433 idev = ipv6_add_dev(dev);
434 if (IS_ERR(idev))
435 return NULL;
436 }
437
438 if (dev->flags&IFF_UP)
439 ipv6_mc_up(idev);
440 return idev;
441 }
442
443 static int inet6_netconf_msgsize_devconf(int type)
444 {
445 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
446 + nla_total_size(4); /* NETCONFA_IFINDEX */
447
448 /* type -1 is used for ALL */
449 if (type == -1 || type == NETCONFA_FORWARDING)
450 size += nla_total_size(4);
451 #ifdef CONFIG_IPV6_MROUTE
452 if (type == -1 || type == NETCONFA_MC_FORWARDING)
453 size += nla_total_size(4);
454 #endif
455 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
456 size += nla_total_size(4);
457
458 return size;
459 }
460
461 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
462 struct ipv6_devconf *devconf, u32 portid,
463 u32 seq, int event, unsigned int flags,
464 int type)
465 {
466 struct nlmsghdr *nlh;
467 struct netconfmsg *ncm;
468
469 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
470 flags);
471 if (nlh == NULL)
472 return -EMSGSIZE;
473
474 ncm = nlmsg_data(nlh);
475 ncm->ncm_family = AF_INET6;
476
477 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
478 goto nla_put_failure;
479
480 /* type -1 is used for ALL */
481 if ((type == -1 || type == NETCONFA_FORWARDING) &&
482 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
483 goto nla_put_failure;
484 #ifdef CONFIG_IPV6_MROUTE
485 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
486 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
487 devconf->mc_forwarding) < 0)
488 goto nla_put_failure;
489 #endif
490 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
491 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
492 goto nla_put_failure;
493
494 nlmsg_end(skb, nlh);
495 return 0;
496
497 nla_put_failure:
498 nlmsg_cancel(skb, nlh);
499 return -EMSGSIZE;
500 }
501
502 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
503 struct ipv6_devconf *devconf)
504 {
505 struct sk_buff *skb;
506 int err = -ENOBUFS;
507
508 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
509 if (skb == NULL)
510 goto errout;
511
512 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
513 RTM_NEWNETCONF, 0, type);
514 if (err < 0) {
515 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
516 WARN_ON(err == -EMSGSIZE);
517 kfree_skb(skb);
518 goto errout;
519 }
520 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
521 return;
522 errout:
523 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
524 }
525
526 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
527 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
528 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
529 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
530 };
531
532 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
533 struct nlmsghdr *nlh)
534 {
535 struct net *net = sock_net(in_skb->sk);
536 struct nlattr *tb[NETCONFA_MAX+1];
537 struct netconfmsg *ncm;
538 struct sk_buff *skb;
539 struct ipv6_devconf *devconf;
540 struct inet6_dev *in6_dev;
541 struct net_device *dev;
542 int ifindex;
543 int err;
544
545 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
546 devconf_ipv6_policy);
547 if (err < 0)
548 goto errout;
549
550 err = EINVAL;
551 if (!tb[NETCONFA_IFINDEX])
552 goto errout;
553
554 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
555 switch (ifindex) {
556 case NETCONFA_IFINDEX_ALL:
557 devconf = net->ipv6.devconf_all;
558 break;
559 case NETCONFA_IFINDEX_DEFAULT:
560 devconf = net->ipv6.devconf_dflt;
561 break;
562 default:
563 dev = __dev_get_by_index(net, ifindex);
564 if (dev == NULL)
565 goto errout;
566 in6_dev = __in6_dev_get(dev);
567 if (in6_dev == NULL)
568 goto errout;
569 devconf = &in6_dev->cnf;
570 break;
571 }
572
573 err = -ENOBUFS;
574 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
575 if (skb == NULL)
576 goto errout;
577
578 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
579 NETLINK_CB(in_skb).portid,
580 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
581 -1);
582 if (err < 0) {
583 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
584 WARN_ON(err == -EMSGSIZE);
585 kfree_skb(skb);
586 goto errout;
587 }
588 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
589 errout:
590 return err;
591 }
592
593 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
594 struct netlink_callback *cb)
595 {
596 struct net *net = sock_net(skb->sk);
597 int h, s_h;
598 int idx, s_idx;
599 struct net_device *dev;
600 struct inet6_dev *idev;
601 struct hlist_head *head;
602
603 s_h = cb->args[0];
604 s_idx = idx = cb->args[1];
605
606 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
607 idx = 0;
608 head = &net->dev_index_head[h];
609 rcu_read_lock();
610 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
611 net->dev_base_seq;
612 hlist_for_each_entry_rcu(dev, head, index_hlist) {
613 if (idx < s_idx)
614 goto cont;
615 idev = __in6_dev_get(dev);
616 if (!idev)
617 goto cont;
618
619 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
620 &idev->cnf,
621 NETLINK_CB(cb->skb).portid,
622 cb->nlh->nlmsg_seq,
623 RTM_NEWNETCONF,
624 NLM_F_MULTI,
625 -1) < 0) {
626 rcu_read_unlock();
627 goto done;
628 }
629 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
630 cont:
631 idx++;
632 }
633 rcu_read_unlock();
634 }
635 if (h == NETDEV_HASHENTRIES) {
636 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
637 net->ipv6.devconf_all,
638 NETLINK_CB(cb->skb).portid,
639 cb->nlh->nlmsg_seq,
640 RTM_NEWNETCONF, NLM_F_MULTI,
641 -1) < 0)
642 goto done;
643 else
644 h++;
645 }
646 if (h == NETDEV_HASHENTRIES + 1) {
647 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
648 net->ipv6.devconf_dflt,
649 NETLINK_CB(cb->skb).portid,
650 cb->nlh->nlmsg_seq,
651 RTM_NEWNETCONF, NLM_F_MULTI,
652 -1) < 0)
653 goto done;
654 else
655 h++;
656 }
657 done:
658 cb->args[0] = h;
659 cb->args[1] = idx;
660
661 return skb->len;
662 }
663
664 #ifdef CONFIG_SYSCTL
665 static void dev_forward_change(struct inet6_dev *idev)
666 {
667 struct net_device *dev;
668 struct inet6_ifaddr *ifa;
669
670 if (!idev)
671 return;
672 dev = idev->dev;
673 if (idev->cnf.forwarding)
674 dev_disable_lro(dev);
675 if (dev->flags & IFF_MULTICAST) {
676 if (idev->cnf.forwarding) {
677 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
678 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
679 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
680 } else {
681 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
682 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
683 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
684 }
685 }
686
687 list_for_each_entry(ifa, &idev->addr_list, if_list) {
688 if (ifa->flags&IFA_F_TENTATIVE)
689 continue;
690 if (idev->cnf.forwarding)
691 addrconf_join_anycast(ifa);
692 else
693 addrconf_leave_anycast(ifa);
694 }
695 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
696 dev->ifindex, &idev->cnf);
697 }
698
699
700 static void addrconf_forward_change(struct net *net, __s32 newf)
701 {
702 struct net_device *dev;
703 struct inet6_dev *idev;
704
705 for_each_netdev(net, dev) {
706 idev = __in6_dev_get(dev);
707 if (idev) {
708 int changed = (!idev->cnf.forwarding) ^ (!newf);
709 idev->cnf.forwarding = newf;
710 if (changed)
711 dev_forward_change(idev);
712 }
713 }
714 }
715
716 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
717 {
718 struct net *net;
719 int old;
720
721 if (!rtnl_trylock())
722 return restart_syscall();
723
724 net = (struct net *)table->extra2;
725 old = *p;
726 *p = newf;
727
728 if (p == &net->ipv6.devconf_dflt->forwarding) {
729 if ((!newf) ^ (!old))
730 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
731 NETCONFA_IFINDEX_DEFAULT,
732 net->ipv6.devconf_dflt);
733 rtnl_unlock();
734 return 0;
735 }
736
737 if (p == &net->ipv6.devconf_all->forwarding) {
738 net->ipv6.devconf_dflt->forwarding = newf;
739 addrconf_forward_change(net, newf);
740 if ((!newf) ^ (!old))
741 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
742 NETCONFA_IFINDEX_ALL,
743 net->ipv6.devconf_all);
744 } else if ((!newf) ^ (!old))
745 dev_forward_change((struct inet6_dev *)table->extra1);
746 rtnl_unlock();
747
748 if (newf)
749 rt6_purge_dflt_routers(net);
750 return 1;
751 }
752 #endif
753
754 /* Nobody refers to this ifaddr, destroy it */
755 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
756 {
757 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
758
759 #ifdef NET_REFCNT_DEBUG
760 pr_debug("%s\n", __func__);
761 #endif
762
763 in6_dev_put(ifp->idev);
764
765 if (cancel_delayed_work(&ifp->dad_work))
766 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
767 ifp);
768
769 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
770 pr_warn("Freeing alive inet6 address %p\n", ifp);
771 return;
772 }
773 ip6_rt_put(ifp->rt);
774
775 kfree_rcu(ifp, rcu);
776 }
777
778 static void
779 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
780 {
781 struct list_head *p;
782 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
783
784 /*
785 * Each device address list is sorted in order of scope -
786 * global before linklocal.
787 */
788 list_for_each(p, &idev->addr_list) {
789 struct inet6_ifaddr *ifa
790 = list_entry(p, struct inet6_ifaddr, if_list);
791 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
792 break;
793 }
794
795 list_add_tail(&ifp->if_list, p);
796 }
797
798 static u32 inet6_addr_hash(const struct in6_addr *addr)
799 {
800 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
801 }
802
803 /* On success it returns ifp with increased reference count */
804
805 static struct inet6_ifaddr *
806 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
807 const struct in6_addr *peer_addr, int pfxlen,
808 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
809 {
810 struct inet6_ifaddr *ifa = NULL;
811 struct rt6_info *rt;
812 unsigned int hash;
813 int err = 0;
814 int addr_type = ipv6_addr_type(addr);
815
816 if (addr_type == IPV6_ADDR_ANY ||
817 addr_type & IPV6_ADDR_MULTICAST ||
818 (!(idev->dev->flags & IFF_LOOPBACK) &&
819 addr_type & IPV6_ADDR_LOOPBACK))
820 return ERR_PTR(-EADDRNOTAVAIL);
821
822 rcu_read_lock_bh();
823 if (idev->dead) {
824 err = -ENODEV; /*XXX*/
825 goto out2;
826 }
827
828 if (idev->cnf.disable_ipv6) {
829 err = -EACCES;
830 goto out2;
831 }
832
833 spin_lock(&addrconf_hash_lock);
834
835 /* Ignore adding duplicate addresses on an interface */
836 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
837 ADBG("ipv6_add_addr: already assigned\n");
838 err = -EEXIST;
839 goto out;
840 }
841
842 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
843
844 if (ifa == NULL) {
845 ADBG("ipv6_add_addr: malloc failed\n");
846 err = -ENOBUFS;
847 goto out;
848 }
849
850 rt = addrconf_dst_alloc(idev, addr, false);
851 if (IS_ERR(rt)) {
852 err = PTR_ERR(rt);
853 goto out;
854 }
855
856 neigh_parms_data_state_setall(idev->nd_parms);
857
858 ifa->addr = *addr;
859 if (peer_addr)
860 ifa->peer_addr = *peer_addr;
861
862 spin_lock_init(&ifa->lock);
863 spin_lock_init(&ifa->state_lock);
864 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
865 INIT_HLIST_NODE(&ifa->addr_lst);
866 ifa->scope = scope;
867 ifa->prefix_len = pfxlen;
868 ifa->flags = flags | IFA_F_TENTATIVE;
869 ifa->valid_lft = valid_lft;
870 ifa->prefered_lft = prefered_lft;
871 ifa->cstamp = ifa->tstamp = jiffies;
872 ifa->tokenized = false;
873
874 ifa->rt = rt;
875
876 ifa->idev = idev;
877 in6_dev_hold(idev);
878 /* For caller */
879 in6_ifa_hold(ifa);
880
881 /* Add to big hash table */
882 hash = inet6_addr_hash(addr);
883
884 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
885 spin_unlock(&addrconf_hash_lock);
886
887 write_lock(&idev->lock);
888 /* Add to inet6_dev unicast addr list. */
889 ipv6_link_dev_addr(idev, ifa);
890
891 if (ifa->flags&IFA_F_TEMPORARY) {
892 list_add(&ifa->tmp_list, &idev->tempaddr_list);
893 in6_ifa_hold(ifa);
894 }
895
896 in6_ifa_hold(ifa);
897 write_unlock(&idev->lock);
898 out2:
899 rcu_read_unlock_bh();
900
901 if (likely(err == 0))
902 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
903 else {
904 kfree(ifa);
905 ifa = ERR_PTR(err);
906 }
907
908 return ifa;
909 out:
910 spin_unlock(&addrconf_hash_lock);
911 goto out2;
912 }
913
914 enum cleanup_prefix_rt_t {
915 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
916 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
917 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
918 };
919
920 /*
921 * Check, whether the prefix for ifp would still need a prefix route
922 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
923 * constants.
924 *
925 * 1) we don't purge prefix if address was not permanent.
926 * prefix is managed by its own lifetime.
927 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
928 * 3) if there are no addresses, delete prefix.
929 * 4) if there are still other permanent address(es),
930 * corresponding prefix is still permanent.
931 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
932 * don't purge the prefix, assume user space is managing it.
933 * 6) otherwise, update prefix lifetime to the
934 * longest valid lifetime among the corresponding
935 * addresses on the device.
936 * Note: subsequent RA will update lifetime.
937 **/
938 static enum cleanup_prefix_rt_t
939 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
940 {
941 struct inet6_ifaddr *ifa;
942 struct inet6_dev *idev = ifp->idev;
943 unsigned long lifetime;
944 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
945
946 *expires = jiffies;
947
948 list_for_each_entry(ifa, &idev->addr_list, if_list) {
949 if (ifa == ifp)
950 continue;
951 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
952 ifp->prefix_len))
953 continue;
954 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
955 return CLEANUP_PREFIX_RT_NOP;
956
957 action = CLEANUP_PREFIX_RT_EXPIRE;
958
959 spin_lock(&ifa->lock);
960
961 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
962 /*
963 * Note: Because this address is
964 * not permanent, lifetime <
965 * LONG_MAX / HZ here.
966 */
967 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
968 *expires = ifa->tstamp + lifetime * HZ;
969 spin_unlock(&ifa->lock);
970 }
971
972 return action;
973 }
974
975 static void
976 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
977 {
978 struct rt6_info *rt;
979
980 rt = addrconf_get_prefix_route(&ifp->addr,
981 ifp->prefix_len,
982 ifp->idev->dev,
983 0, RTF_GATEWAY | RTF_DEFAULT);
984 if (rt) {
985 if (del_rt)
986 ip6_del_rt(rt);
987 else {
988 if (!(rt->rt6i_flags & RTF_EXPIRES))
989 rt6_set_expires(rt, expires);
990 ip6_rt_put(rt);
991 }
992 }
993 }
994
995
996 /* This function wants to get referenced ifp and releases it before return */
997
998 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
999 {
1000 int state;
1001 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1002 unsigned long expires;
1003
1004 ASSERT_RTNL();
1005
1006 spin_lock_bh(&ifp->state_lock);
1007 state = ifp->state;
1008 ifp->state = INET6_IFADDR_STATE_DEAD;
1009 spin_unlock_bh(&ifp->state_lock);
1010
1011 if (state == INET6_IFADDR_STATE_DEAD)
1012 goto out;
1013
1014 spin_lock_bh(&addrconf_hash_lock);
1015 hlist_del_init_rcu(&ifp->addr_lst);
1016 spin_unlock_bh(&addrconf_hash_lock);
1017
1018 write_lock_bh(&ifp->idev->lock);
1019
1020 if (ifp->flags&IFA_F_TEMPORARY) {
1021 list_del(&ifp->tmp_list);
1022 if (ifp->ifpub) {
1023 in6_ifa_put(ifp->ifpub);
1024 ifp->ifpub = NULL;
1025 }
1026 __in6_ifa_put(ifp);
1027 }
1028
1029 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1030 action = check_cleanup_prefix_route(ifp, &expires);
1031
1032 list_del_init(&ifp->if_list);
1033 __in6_ifa_put(ifp);
1034
1035 write_unlock_bh(&ifp->idev->lock);
1036
1037 addrconf_del_dad_work(ifp);
1038
1039 ipv6_ifa_notify(RTM_DELADDR, ifp);
1040
1041 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1042
1043 if (action != CLEANUP_PREFIX_RT_NOP) {
1044 cleanup_prefix_route(ifp, expires,
1045 action == CLEANUP_PREFIX_RT_DEL);
1046 }
1047
1048 /* clean up prefsrc entries */
1049 rt6_remove_prefsrc(ifp);
1050 out:
1051 in6_ifa_put(ifp);
1052 }
1053
1054 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1055 {
1056 struct inet6_dev *idev = ifp->idev;
1057 struct in6_addr addr, *tmpaddr;
1058 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1059 unsigned long regen_advance;
1060 int tmp_plen;
1061 int ret = 0;
1062 u32 addr_flags;
1063 unsigned long now = jiffies;
1064
1065 write_lock_bh(&idev->lock);
1066 if (ift) {
1067 spin_lock_bh(&ift->lock);
1068 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1069 spin_unlock_bh(&ift->lock);
1070 tmpaddr = &addr;
1071 } else {
1072 tmpaddr = NULL;
1073 }
1074 retry:
1075 in6_dev_hold(idev);
1076 if (idev->cnf.use_tempaddr <= 0) {
1077 write_unlock_bh(&idev->lock);
1078 pr_info("%s: use_tempaddr is disabled\n", __func__);
1079 in6_dev_put(idev);
1080 ret = -1;
1081 goto out;
1082 }
1083 spin_lock_bh(&ifp->lock);
1084 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1085 idev->cnf.use_tempaddr = -1; /*XXX*/
1086 spin_unlock_bh(&ifp->lock);
1087 write_unlock_bh(&idev->lock);
1088 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1089 __func__);
1090 in6_dev_put(idev);
1091 ret = -1;
1092 goto out;
1093 }
1094 in6_ifa_hold(ifp);
1095 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1096 __ipv6_try_regen_rndid(idev, tmpaddr);
1097 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1098 age = (now - ifp->tstamp) / HZ;
1099 tmp_valid_lft = min_t(__u32,
1100 ifp->valid_lft,
1101 idev->cnf.temp_valid_lft + age);
1102 tmp_prefered_lft = min_t(__u32,
1103 ifp->prefered_lft,
1104 idev->cnf.temp_prefered_lft + age -
1105 idev->cnf.max_desync_factor);
1106 tmp_plen = ifp->prefix_len;
1107 tmp_tstamp = ifp->tstamp;
1108 spin_unlock_bh(&ifp->lock);
1109
1110 regen_advance = idev->cnf.regen_max_retry *
1111 idev->cnf.dad_transmits *
1112 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1113 write_unlock_bh(&idev->lock);
1114
1115 /* A temporary address is created only if this calculated Preferred
1116 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1117 * an implementation must not create a temporary address with a zero
1118 * Preferred Lifetime.
1119 * Use age calculation as in addrconf_verify to avoid unnecessary
1120 * temporary addresses being generated.
1121 */
1122 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1123 if (tmp_prefered_lft <= regen_advance + age) {
1124 in6_ifa_put(ifp);
1125 in6_dev_put(idev);
1126 ret = -1;
1127 goto out;
1128 }
1129
1130 addr_flags = IFA_F_TEMPORARY;
1131 /* set in addrconf_prefix_rcv() */
1132 if (ifp->flags & IFA_F_OPTIMISTIC)
1133 addr_flags |= IFA_F_OPTIMISTIC;
1134
1135 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1136 ipv6_addr_scope(&addr), addr_flags,
1137 tmp_valid_lft, tmp_prefered_lft);
1138 if (IS_ERR(ift)) {
1139 in6_ifa_put(ifp);
1140 in6_dev_put(idev);
1141 pr_info("%s: retry temporary address regeneration\n", __func__);
1142 tmpaddr = &addr;
1143 write_lock_bh(&idev->lock);
1144 goto retry;
1145 }
1146
1147 spin_lock_bh(&ift->lock);
1148 ift->ifpub = ifp;
1149 ift->cstamp = now;
1150 ift->tstamp = tmp_tstamp;
1151 spin_unlock_bh(&ift->lock);
1152
1153 addrconf_dad_start(ift);
1154 in6_ifa_put(ift);
1155 in6_dev_put(idev);
1156 out:
1157 return ret;
1158 }
1159
1160 /*
1161 * Choose an appropriate source address (RFC3484)
1162 */
1163 enum {
1164 IPV6_SADDR_RULE_INIT = 0,
1165 IPV6_SADDR_RULE_LOCAL,
1166 IPV6_SADDR_RULE_SCOPE,
1167 IPV6_SADDR_RULE_PREFERRED,
1168 #ifdef CONFIG_IPV6_MIP6
1169 IPV6_SADDR_RULE_HOA,
1170 #endif
1171 IPV6_SADDR_RULE_OIF,
1172 IPV6_SADDR_RULE_LABEL,
1173 IPV6_SADDR_RULE_PRIVACY,
1174 IPV6_SADDR_RULE_ORCHID,
1175 IPV6_SADDR_RULE_PREFIX,
1176 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1177 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1178 #endif
1179 IPV6_SADDR_RULE_MAX
1180 };
1181
1182 struct ipv6_saddr_score {
1183 int rule;
1184 int addr_type;
1185 struct inet6_ifaddr *ifa;
1186 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1187 int scopedist;
1188 int matchlen;
1189 };
1190
1191 struct ipv6_saddr_dst {
1192 const struct in6_addr *addr;
1193 int ifindex;
1194 int scope;
1195 int label;
1196 unsigned int prefs;
1197 };
1198
1199 static inline int ipv6_saddr_preferred(int type)
1200 {
1201 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1202 return 1;
1203 return 0;
1204 }
1205
1206 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1207 {
1208 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1209 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1210 #else
1211 return false;
1212 #endif
1213 }
1214
1215 static int ipv6_get_saddr_eval(struct net *net,
1216 struct ipv6_saddr_score *score,
1217 struct ipv6_saddr_dst *dst,
1218 int i)
1219 {
1220 int ret;
1221
1222 if (i <= score->rule) {
1223 switch (i) {
1224 case IPV6_SADDR_RULE_SCOPE:
1225 ret = score->scopedist;
1226 break;
1227 case IPV6_SADDR_RULE_PREFIX:
1228 ret = score->matchlen;
1229 break;
1230 default:
1231 ret = !!test_bit(i, score->scorebits);
1232 }
1233 goto out;
1234 }
1235
1236 switch (i) {
1237 case IPV6_SADDR_RULE_INIT:
1238 /* Rule 0: remember if hiscore is not ready yet */
1239 ret = !!score->ifa;
1240 break;
1241 case IPV6_SADDR_RULE_LOCAL:
1242 /* Rule 1: Prefer same address */
1243 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1244 break;
1245 case IPV6_SADDR_RULE_SCOPE:
1246 /* Rule 2: Prefer appropriate scope
1247 *
1248 * ret
1249 * ^
1250 * -1 | d 15
1251 * ---+--+-+---> scope
1252 * |
1253 * | d is scope of the destination.
1254 * B-d | \
1255 * | \ <- smaller scope is better if
1256 * B-15 | \ if scope is enough for destination.
1257 * | ret = B - scope (-1 <= scope >= d <= 15).
1258 * d-C-1 | /
1259 * |/ <- greater is better
1260 * -C / if scope is not enough for destination.
1261 * /| ret = scope - C (-1 <= d < scope <= 15).
1262 *
1263 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1264 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1265 * Assume B = 0 and we get C > 29.
1266 */
1267 ret = __ipv6_addr_src_scope(score->addr_type);
1268 if (ret >= dst->scope)
1269 ret = -ret;
1270 else
1271 ret -= 128; /* 30 is enough */
1272 score->scopedist = ret;
1273 break;
1274 case IPV6_SADDR_RULE_PREFERRED:
1275 {
1276 /* Rule 3: Avoid deprecated and optimistic addresses */
1277 u8 avoid = IFA_F_DEPRECATED;
1278
1279 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1280 avoid |= IFA_F_OPTIMISTIC;
1281 ret = ipv6_saddr_preferred(score->addr_type) ||
1282 !(score->ifa->flags & avoid);
1283 break;
1284 }
1285 #ifdef CONFIG_IPV6_MIP6
1286 case IPV6_SADDR_RULE_HOA:
1287 {
1288 /* Rule 4: Prefer home address */
1289 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1290 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1291 break;
1292 }
1293 #endif
1294 case IPV6_SADDR_RULE_OIF:
1295 /* Rule 5: Prefer outgoing interface */
1296 ret = (!dst->ifindex ||
1297 dst->ifindex == score->ifa->idev->dev->ifindex);
1298 break;
1299 case IPV6_SADDR_RULE_LABEL:
1300 /* Rule 6: Prefer matching label */
1301 ret = ipv6_addr_label(net,
1302 &score->ifa->addr, score->addr_type,
1303 score->ifa->idev->dev->ifindex) == dst->label;
1304 break;
1305 case IPV6_SADDR_RULE_PRIVACY:
1306 {
1307 /* Rule 7: Prefer public address
1308 * Note: prefer temporary address if use_tempaddr >= 2
1309 */
1310 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1311 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1312 score->ifa->idev->cnf.use_tempaddr >= 2;
1313 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1314 break;
1315 }
1316 case IPV6_SADDR_RULE_ORCHID:
1317 /* Rule 8-: Prefer ORCHID vs ORCHID or
1318 * non-ORCHID vs non-ORCHID
1319 */
1320 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1321 ipv6_addr_orchid(dst->addr));
1322 break;
1323 case IPV6_SADDR_RULE_PREFIX:
1324 /* Rule 8: Use longest matching prefix */
1325 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1326 if (ret > score->ifa->prefix_len)
1327 ret = score->ifa->prefix_len;
1328 score->matchlen = ret;
1329 break;
1330 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1331 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1332 /* Optimistic addresses still have lower precedence than other
1333 * preferred addresses.
1334 */
1335 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1336 break;
1337 #endif
1338 default:
1339 ret = 0;
1340 }
1341
1342 if (ret)
1343 __set_bit(i, score->scorebits);
1344 score->rule = i;
1345 out:
1346 return ret;
1347 }
1348
1349 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1350 const struct in6_addr *daddr, unsigned int prefs,
1351 struct in6_addr *saddr)
1352 {
1353 struct ipv6_saddr_score scores[2],
1354 *score = &scores[0], *hiscore = &scores[1];
1355 struct ipv6_saddr_dst dst;
1356 struct net_device *dev;
1357 int dst_type;
1358
1359 dst_type = __ipv6_addr_type(daddr);
1360 dst.addr = daddr;
1361 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1362 dst.scope = __ipv6_addr_src_scope(dst_type);
1363 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1364 dst.prefs = prefs;
1365
1366 hiscore->rule = -1;
1367 hiscore->ifa = NULL;
1368
1369 rcu_read_lock();
1370
1371 for_each_netdev_rcu(net, dev) {
1372 struct inet6_dev *idev;
1373
1374 /* Candidate Source Address (section 4)
1375 * - multicast and link-local destination address,
1376 * the set of candidate source address MUST only
1377 * include addresses assigned to interfaces
1378 * belonging to the same link as the outgoing
1379 * interface.
1380 * (- For site-local destination addresses, the
1381 * set of candidate source addresses MUST only
1382 * include addresses assigned to interfaces
1383 * belonging to the same site as the outgoing
1384 * interface.)
1385 */
1386 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1387 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1388 dst.ifindex && dev->ifindex != dst.ifindex)
1389 continue;
1390
1391 idev = __in6_dev_get(dev);
1392 if (!idev)
1393 continue;
1394
1395 read_lock_bh(&idev->lock);
1396 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1397 int i;
1398
1399 /*
1400 * - Tentative Address (RFC2462 section 5.4)
1401 * - A tentative address is not considered
1402 * "assigned to an interface" in the traditional
1403 * sense, unless it is also flagged as optimistic.
1404 * - Candidate Source Address (section 4)
1405 * - In any case, anycast addresses, multicast
1406 * addresses, and the unspecified address MUST
1407 * NOT be included in a candidate set.
1408 */
1409 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1410 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1411 continue;
1412
1413 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1414
1415 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1416 score->addr_type & IPV6_ADDR_MULTICAST)) {
1417 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1418 dev->name);
1419 continue;
1420 }
1421
1422 score->rule = -1;
1423 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1424
1425 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1426 int minihiscore, miniscore;
1427
1428 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1429 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1430
1431 if (minihiscore > miniscore) {
1432 if (i == IPV6_SADDR_RULE_SCOPE &&
1433 score->scopedist > 0) {
1434 /*
1435 * special case:
1436 * each remaining entry
1437 * has too small (not enough)
1438 * scope, because ifa entries
1439 * are sorted by their scope
1440 * values.
1441 */
1442 goto try_nextdev;
1443 }
1444 break;
1445 } else if (minihiscore < miniscore) {
1446 if (hiscore->ifa)
1447 in6_ifa_put(hiscore->ifa);
1448
1449 in6_ifa_hold(score->ifa);
1450
1451 swap(hiscore, score);
1452
1453 /* restore our iterator */
1454 score->ifa = hiscore->ifa;
1455
1456 break;
1457 }
1458 }
1459 }
1460 try_nextdev:
1461 read_unlock_bh(&idev->lock);
1462 }
1463 rcu_read_unlock();
1464
1465 if (!hiscore->ifa)
1466 return -EADDRNOTAVAIL;
1467
1468 *saddr = hiscore->ifa->addr;
1469 in6_ifa_put(hiscore->ifa);
1470 return 0;
1471 }
1472 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1473
1474 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1475 u32 banned_flags)
1476 {
1477 struct inet6_ifaddr *ifp;
1478 int err = -EADDRNOTAVAIL;
1479
1480 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1481 if (ifp->scope > IFA_LINK)
1482 break;
1483 if (ifp->scope == IFA_LINK &&
1484 !(ifp->flags & banned_flags)) {
1485 *addr = ifp->addr;
1486 err = 0;
1487 break;
1488 }
1489 }
1490 return err;
1491 }
1492
1493 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1494 u32 banned_flags)
1495 {
1496 struct inet6_dev *idev;
1497 int err = -EADDRNOTAVAIL;
1498
1499 rcu_read_lock();
1500 idev = __in6_dev_get(dev);
1501 if (idev) {
1502 read_lock_bh(&idev->lock);
1503 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1504 read_unlock_bh(&idev->lock);
1505 }
1506 rcu_read_unlock();
1507 return err;
1508 }
1509
1510 static int ipv6_count_addresses(struct inet6_dev *idev)
1511 {
1512 int cnt = 0;
1513 struct inet6_ifaddr *ifp;
1514
1515 read_lock_bh(&idev->lock);
1516 list_for_each_entry(ifp, &idev->addr_list, if_list)
1517 cnt++;
1518 read_unlock_bh(&idev->lock);
1519 return cnt;
1520 }
1521
1522 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1523 const struct net_device *dev, int strict)
1524 {
1525 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1526 }
1527 EXPORT_SYMBOL(ipv6_chk_addr);
1528
1529 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1530 const struct net_device *dev, int strict,
1531 u32 banned_flags)
1532 {
1533 struct inet6_ifaddr *ifp;
1534 unsigned int hash = inet6_addr_hash(addr);
1535 u32 ifp_flags;
1536
1537 rcu_read_lock_bh();
1538 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1539 if (!net_eq(dev_net(ifp->idev->dev), net))
1540 continue;
1541 /* Decouple optimistic from tentative for evaluation here.
1542 * Ban optimistic addresses explicitly, when required.
1543 */
1544 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1545 ? (ifp->flags&~IFA_F_TENTATIVE)
1546 : ifp->flags;
1547 if (ipv6_addr_equal(&ifp->addr, addr) &&
1548 !(ifp_flags&banned_flags) &&
1549 (dev == NULL || ifp->idev->dev == dev ||
1550 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1551 rcu_read_unlock_bh();
1552 return 1;
1553 }
1554 }
1555
1556 rcu_read_unlock_bh();
1557 return 0;
1558 }
1559 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1560
1561 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1562 struct net_device *dev)
1563 {
1564 unsigned int hash = inet6_addr_hash(addr);
1565 struct inet6_ifaddr *ifp;
1566
1567 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1568 if (!net_eq(dev_net(ifp->idev->dev), net))
1569 continue;
1570 if (ipv6_addr_equal(&ifp->addr, addr)) {
1571 if (dev == NULL || ifp->idev->dev == dev)
1572 return true;
1573 }
1574 }
1575 return false;
1576 }
1577
1578 /* Compares an address/prefix_len with addresses on device @dev.
1579 * If one is found it returns true.
1580 */
1581 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1582 const unsigned int prefix_len, struct net_device *dev)
1583 {
1584 struct inet6_dev *idev;
1585 struct inet6_ifaddr *ifa;
1586 bool ret = false;
1587
1588 rcu_read_lock();
1589 idev = __in6_dev_get(dev);
1590 if (idev) {
1591 read_lock_bh(&idev->lock);
1592 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1593 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1594 if (ret)
1595 break;
1596 }
1597 read_unlock_bh(&idev->lock);
1598 }
1599 rcu_read_unlock();
1600
1601 return ret;
1602 }
1603 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1604
1605 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1606 {
1607 struct inet6_dev *idev;
1608 struct inet6_ifaddr *ifa;
1609 int onlink;
1610
1611 onlink = 0;
1612 rcu_read_lock();
1613 idev = __in6_dev_get(dev);
1614 if (idev) {
1615 read_lock_bh(&idev->lock);
1616 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1617 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1618 ifa->prefix_len);
1619 if (onlink)
1620 break;
1621 }
1622 read_unlock_bh(&idev->lock);
1623 }
1624 rcu_read_unlock();
1625 return onlink;
1626 }
1627 EXPORT_SYMBOL(ipv6_chk_prefix);
1628
1629 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1630 struct net_device *dev, int strict)
1631 {
1632 struct inet6_ifaddr *ifp, *result = NULL;
1633 unsigned int hash = inet6_addr_hash(addr);
1634
1635 rcu_read_lock_bh();
1636 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1637 if (!net_eq(dev_net(ifp->idev->dev), net))
1638 continue;
1639 if (ipv6_addr_equal(&ifp->addr, addr)) {
1640 if (dev == NULL || ifp->idev->dev == dev ||
1641 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1642 result = ifp;
1643 in6_ifa_hold(ifp);
1644 break;
1645 }
1646 }
1647 }
1648 rcu_read_unlock_bh();
1649
1650 return result;
1651 }
1652
1653 /* Gets referenced address, destroys ifaddr */
1654
1655 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1656 {
1657 if (ifp->flags&IFA_F_PERMANENT) {
1658 spin_lock_bh(&ifp->lock);
1659 addrconf_del_dad_work(ifp);
1660 ifp->flags |= IFA_F_TENTATIVE;
1661 if (dad_failed)
1662 ifp->flags |= IFA_F_DADFAILED;
1663 spin_unlock_bh(&ifp->lock);
1664 if (dad_failed)
1665 ipv6_ifa_notify(0, ifp);
1666 in6_ifa_put(ifp);
1667 } else if (ifp->flags&IFA_F_TEMPORARY) {
1668 struct inet6_ifaddr *ifpub;
1669 spin_lock_bh(&ifp->lock);
1670 ifpub = ifp->ifpub;
1671 if (ifpub) {
1672 in6_ifa_hold(ifpub);
1673 spin_unlock_bh(&ifp->lock);
1674 ipv6_create_tempaddr(ifpub, ifp);
1675 in6_ifa_put(ifpub);
1676 } else {
1677 spin_unlock_bh(&ifp->lock);
1678 }
1679 ipv6_del_addr(ifp);
1680 } else {
1681 ipv6_del_addr(ifp);
1682 }
1683 }
1684
1685 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1686 {
1687 int err = -ENOENT;
1688
1689 spin_lock_bh(&ifp->state_lock);
1690 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1691 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1692 err = 0;
1693 }
1694 spin_unlock_bh(&ifp->state_lock);
1695
1696 return err;
1697 }
1698
1699 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1700 {
1701 struct inet6_dev *idev = ifp->idev;
1702
1703 if (addrconf_dad_end(ifp)) {
1704 in6_ifa_put(ifp);
1705 return;
1706 }
1707
1708 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1709 ifp->idev->dev->name, &ifp->addr);
1710
1711 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1712 struct in6_addr addr;
1713
1714 addr.s6_addr32[0] = htonl(0xfe800000);
1715 addr.s6_addr32[1] = 0;
1716
1717 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1718 ipv6_addr_equal(&ifp->addr, &addr)) {
1719 /* DAD failed for link-local based on MAC address */
1720 idev->cnf.disable_ipv6 = 1;
1721
1722 pr_info("%s: IPv6 being disabled!\n",
1723 ifp->idev->dev->name);
1724 }
1725 }
1726
1727 spin_lock_bh(&ifp->state_lock);
1728 /* transition from _POSTDAD to _ERRDAD */
1729 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1730 spin_unlock_bh(&ifp->state_lock);
1731
1732 addrconf_mod_dad_work(ifp, 0);
1733 }
1734
1735 /* Join to solicited addr multicast group.
1736 * caller must hold RTNL */
1737 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1738 {
1739 struct in6_addr maddr;
1740
1741 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1742 return;
1743
1744 addrconf_addr_solict_mult(addr, &maddr);
1745 ipv6_dev_mc_inc(dev, &maddr);
1746 }
1747
1748 /* caller must hold RTNL */
1749 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1750 {
1751 struct in6_addr maddr;
1752
1753 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1754 return;
1755
1756 addrconf_addr_solict_mult(addr, &maddr);
1757 __ipv6_dev_mc_dec(idev, &maddr);
1758 }
1759
1760 /* caller must hold RTNL */
1761 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1762 {
1763 struct in6_addr addr;
1764
1765 if (ifp->prefix_len >= 127) /* RFC 6164 */
1766 return;
1767 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1768 if (ipv6_addr_any(&addr))
1769 return;
1770 __ipv6_dev_ac_inc(ifp->idev, &addr);
1771 }
1772
1773 /* caller must hold RTNL */
1774 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1775 {
1776 struct in6_addr addr;
1777
1778 if (ifp->prefix_len >= 127) /* RFC 6164 */
1779 return;
1780 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1781 if (ipv6_addr_any(&addr))
1782 return;
1783 __ipv6_dev_ac_dec(ifp->idev, &addr);
1784 }
1785
1786 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1787 {
1788 if (dev->addr_len != ETH_ALEN)
1789 return -1;
1790 memcpy(eui, dev->dev_addr, 3);
1791 memcpy(eui + 5, dev->dev_addr + 3, 3);
1792
1793 /*
1794 * The zSeries OSA network cards can be shared among various
1795 * OS instances, but the OSA cards have only one MAC address.
1796 * This leads to duplicate address conflicts in conjunction
1797 * with IPv6 if more than one instance uses the same card.
1798 *
1799 * The driver for these cards can deliver a unique 16-bit
1800 * identifier for each instance sharing the same card. It is
1801 * placed instead of 0xFFFE in the interface identifier. The
1802 * "u" bit of the interface identifier is not inverted in this
1803 * case. Hence the resulting interface identifier has local
1804 * scope according to RFC2373.
1805 */
1806 if (dev->dev_id) {
1807 eui[3] = (dev->dev_id >> 8) & 0xFF;
1808 eui[4] = dev->dev_id & 0xFF;
1809 } else {
1810 eui[3] = 0xFF;
1811 eui[4] = 0xFE;
1812 eui[0] ^= 2;
1813 }
1814 return 0;
1815 }
1816
1817 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1818 {
1819 if (dev->addr_len != IEEE802154_ADDR_LEN)
1820 return -1;
1821 memcpy(eui, dev->dev_addr, 8);
1822 eui[0] ^= 2;
1823 return 0;
1824 }
1825
1826 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1827 {
1828 union fwnet_hwaddr *ha;
1829
1830 if (dev->addr_len != FWNET_ALEN)
1831 return -1;
1832
1833 ha = (union fwnet_hwaddr *)dev->dev_addr;
1834
1835 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1836 eui[0] ^= 2;
1837 return 0;
1838 }
1839
1840 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1841 {
1842 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1843 if (dev->addr_len != ARCNET_ALEN)
1844 return -1;
1845 memset(eui, 0, 7);
1846 eui[7] = *(u8 *)dev->dev_addr;
1847 return 0;
1848 }
1849
1850 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1851 {
1852 if (dev->addr_len != INFINIBAND_ALEN)
1853 return -1;
1854 memcpy(eui, dev->dev_addr + 12, 8);
1855 eui[0] |= 2;
1856 return 0;
1857 }
1858
1859 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1860 {
1861 if (addr == 0)
1862 return -1;
1863 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1864 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1865 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1866 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1867 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1868 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1869 eui[1] = 0;
1870 eui[2] = 0x5E;
1871 eui[3] = 0xFE;
1872 memcpy(eui + 4, &addr, 4);
1873 return 0;
1874 }
1875
1876 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1877 {
1878 if (dev->priv_flags & IFF_ISATAP)
1879 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1880 return -1;
1881 }
1882
1883 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1884 {
1885 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1886 }
1887
1888 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1889 {
1890 memcpy(eui, dev->perm_addr, 3);
1891 memcpy(eui + 5, dev->perm_addr + 3, 3);
1892 eui[3] = 0xFF;
1893 eui[4] = 0xFE;
1894 eui[0] ^= 2;
1895 return 0;
1896 }
1897
1898 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1899 {
1900 switch (dev->type) {
1901 case ARPHRD_ETHER:
1902 case ARPHRD_FDDI:
1903 return addrconf_ifid_eui48(eui, dev);
1904 case ARPHRD_ARCNET:
1905 return addrconf_ifid_arcnet(eui, dev);
1906 case ARPHRD_INFINIBAND:
1907 return addrconf_ifid_infiniband(eui, dev);
1908 case ARPHRD_SIT:
1909 return addrconf_ifid_sit(eui, dev);
1910 case ARPHRD_IPGRE:
1911 return addrconf_ifid_gre(eui, dev);
1912 case ARPHRD_6LOWPAN:
1913 case ARPHRD_IEEE802154:
1914 return addrconf_ifid_eui64(eui, dev);
1915 case ARPHRD_IEEE1394:
1916 return addrconf_ifid_ieee1394(eui, dev);
1917 case ARPHRD_TUNNEL6:
1918 return addrconf_ifid_ip6tnl(eui, dev);
1919 }
1920 return -1;
1921 }
1922
1923 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1924 {
1925 int err = -1;
1926 struct inet6_ifaddr *ifp;
1927
1928 read_lock_bh(&idev->lock);
1929 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1930 if (ifp->scope > IFA_LINK)
1931 break;
1932 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1933 memcpy(eui, ifp->addr.s6_addr+8, 8);
1934 err = 0;
1935 break;
1936 }
1937 }
1938 read_unlock_bh(&idev->lock);
1939 return err;
1940 }
1941
1942 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1943 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1944 {
1945 regen:
1946 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1947 idev->rndid[0] &= ~0x02;
1948
1949 /*
1950 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1951 * check if generated address is not inappropriate
1952 *
1953 * - Reserved subnet anycast (RFC 2526)
1954 * 11111101 11....11 1xxxxxxx
1955 * - ISATAP (RFC4214) 6.1
1956 * 00-00-5E-FE-xx-xx-xx-xx
1957 * - value 0
1958 * - XXX: already assigned to an address on the device
1959 */
1960 if (idev->rndid[0] == 0xfd &&
1961 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1962 (idev->rndid[7]&0x80))
1963 goto regen;
1964 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1965 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1966 goto regen;
1967 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1968 goto regen;
1969 }
1970 }
1971
1972 static void ipv6_regen_rndid(unsigned long data)
1973 {
1974 struct inet6_dev *idev = (struct inet6_dev *) data;
1975 unsigned long expires;
1976
1977 rcu_read_lock_bh();
1978 write_lock_bh(&idev->lock);
1979
1980 if (idev->dead)
1981 goto out;
1982
1983 __ipv6_regen_rndid(idev);
1984
1985 expires = jiffies +
1986 idev->cnf.temp_prefered_lft * HZ -
1987 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1988 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1989 idev->cnf.max_desync_factor * HZ;
1990 if (time_before(expires, jiffies)) {
1991 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1992 __func__, idev->dev->name);
1993 goto out;
1994 }
1995
1996 if (!mod_timer(&idev->regen_timer, expires))
1997 in6_dev_hold(idev);
1998
1999 out:
2000 write_unlock_bh(&idev->lock);
2001 rcu_read_unlock_bh();
2002 in6_dev_put(idev);
2003 }
2004
2005 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2006 {
2007 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2008 __ipv6_regen_rndid(idev);
2009 }
2010
2011 /*
2012 * Add prefix route.
2013 */
2014
2015 static void
2016 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2017 unsigned long expires, u32 flags)
2018 {
2019 struct fib6_config cfg = {
2020 .fc_table = RT6_TABLE_PREFIX,
2021 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2022 .fc_ifindex = dev->ifindex,
2023 .fc_expires = expires,
2024 .fc_dst_len = plen,
2025 .fc_flags = RTF_UP | flags,
2026 .fc_nlinfo.nl_net = dev_net(dev),
2027 .fc_protocol = RTPROT_KERNEL,
2028 };
2029
2030 cfg.fc_dst = *pfx;
2031
2032 /* Prevent useless cloning on PtP SIT.
2033 This thing is done here expecting that the whole
2034 class of non-broadcast devices need not cloning.
2035 */
2036 #if IS_ENABLED(CONFIG_IPV6_SIT)
2037 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2038 cfg.fc_flags |= RTF_NONEXTHOP;
2039 #endif
2040
2041 ip6_route_add(&cfg);
2042 }
2043
2044
2045 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2046 int plen,
2047 const struct net_device *dev,
2048 u32 flags, u32 noflags)
2049 {
2050 struct fib6_node *fn;
2051 struct rt6_info *rt = NULL;
2052 struct fib6_table *table;
2053
2054 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2055 if (table == NULL)
2056 return NULL;
2057
2058 read_lock_bh(&table->tb6_lock);
2059 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2060 if (!fn)
2061 goto out;
2062 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2063 if (rt->dst.dev->ifindex != dev->ifindex)
2064 continue;
2065 if ((rt->rt6i_flags & flags) != flags)
2066 continue;
2067 if ((rt->rt6i_flags & noflags) != 0)
2068 continue;
2069 dst_hold(&rt->dst);
2070 break;
2071 }
2072 out:
2073 read_unlock_bh(&table->tb6_lock);
2074 return rt;
2075 }
2076
2077
2078 /* Create "default" multicast route to the interface */
2079
2080 static void addrconf_add_mroute(struct net_device *dev)
2081 {
2082 struct fib6_config cfg = {
2083 .fc_table = RT6_TABLE_LOCAL,
2084 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2085 .fc_ifindex = dev->ifindex,
2086 .fc_dst_len = 8,
2087 .fc_flags = RTF_UP,
2088 .fc_nlinfo.nl_net = dev_net(dev),
2089 };
2090
2091 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2092
2093 ip6_route_add(&cfg);
2094 }
2095
2096 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2097 {
2098 struct inet6_dev *idev;
2099
2100 ASSERT_RTNL();
2101
2102 idev = ipv6_find_idev(dev);
2103 if (!idev)
2104 return ERR_PTR(-ENOBUFS);
2105
2106 if (idev->cnf.disable_ipv6)
2107 return ERR_PTR(-EACCES);
2108
2109 /* Add default multicast route */
2110 if (!(dev->flags & IFF_LOOPBACK))
2111 addrconf_add_mroute(dev);
2112
2113 return idev;
2114 }
2115
2116 static void manage_tempaddrs(struct inet6_dev *idev,
2117 struct inet6_ifaddr *ifp,
2118 __u32 valid_lft, __u32 prefered_lft,
2119 bool create, unsigned long now)
2120 {
2121 u32 flags;
2122 struct inet6_ifaddr *ift;
2123
2124 read_lock_bh(&idev->lock);
2125 /* update all temporary addresses in the list */
2126 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2127 int age, max_valid, max_prefered;
2128
2129 if (ifp != ift->ifpub)
2130 continue;
2131
2132 /* RFC 4941 section 3.3:
2133 * If a received option will extend the lifetime of a public
2134 * address, the lifetimes of temporary addresses should
2135 * be extended, subject to the overall constraint that no
2136 * temporary addresses should ever remain "valid" or "preferred"
2137 * for a time longer than (TEMP_VALID_LIFETIME) or
2138 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2139 */
2140 age = (now - ift->cstamp) / HZ;
2141 max_valid = idev->cnf.temp_valid_lft - age;
2142 if (max_valid < 0)
2143 max_valid = 0;
2144
2145 max_prefered = idev->cnf.temp_prefered_lft -
2146 idev->cnf.max_desync_factor - age;
2147 if (max_prefered < 0)
2148 max_prefered = 0;
2149
2150 if (valid_lft > max_valid)
2151 valid_lft = max_valid;
2152
2153 if (prefered_lft > max_prefered)
2154 prefered_lft = max_prefered;
2155
2156 spin_lock(&ift->lock);
2157 flags = ift->flags;
2158 ift->valid_lft = valid_lft;
2159 ift->prefered_lft = prefered_lft;
2160 ift->tstamp = now;
2161 if (prefered_lft > 0)
2162 ift->flags &= ~IFA_F_DEPRECATED;
2163
2164 spin_unlock(&ift->lock);
2165 if (!(flags&IFA_F_TENTATIVE))
2166 ipv6_ifa_notify(0, ift);
2167 }
2168
2169 if ((create || list_empty(&idev->tempaddr_list)) &&
2170 idev->cnf.use_tempaddr > 0) {
2171 /* When a new public address is created as described
2172 * in [ADDRCONF], also create a new temporary address.
2173 * Also create a temporary address if it's enabled but
2174 * no temporary address currently exists.
2175 */
2176 read_unlock_bh(&idev->lock);
2177 ipv6_create_tempaddr(ifp, NULL);
2178 } else {
2179 read_unlock_bh(&idev->lock);
2180 }
2181 }
2182
2183 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2184 {
2185 struct prefix_info *pinfo;
2186 __u32 valid_lft;
2187 __u32 prefered_lft;
2188 int addr_type;
2189 struct inet6_dev *in6_dev;
2190 struct net *net = dev_net(dev);
2191
2192 pinfo = (struct prefix_info *) opt;
2193
2194 if (len < sizeof(struct prefix_info)) {
2195 ADBG("addrconf: prefix option too short\n");
2196 return;
2197 }
2198
2199 /*
2200 * Validation checks ([ADDRCONF], page 19)
2201 */
2202
2203 addr_type = ipv6_addr_type(&pinfo->prefix);
2204
2205 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2206 return;
2207
2208 valid_lft = ntohl(pinfo->valid);
2209 prefered_lft = ntohl(pinfo->prefered);
2210
2211 if (prefered_lft > valid_lft) {
2212 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2213 return;
2214 }
2215
2216 in6_dev = in6_dev_get(dev);
2217
2218 if (in6_dev == NULL) {
2219 net_dbg_ratelimited("addrconf: device %s not configured\n",
2220 dev->name);
2221 return;
2222 }
2223
2224 /*
2225 * Two things going on here:
2226 * 1) Add routes for on-link prefixes
2227 * 2) Configure prefixes with the auto flag set
2228 */
2229
2230 if (pinfo->onlink) {
2231 struct rt6_info *rt;
2232 unsigned long rt_expires;
2233
2234 /* Avoid arithmetic overflow. Really, we could
2235 * save rt_expires in seconds, likely valid_lft,
2236 * but it would require division in fib gc, that it
2237 * not good.
2238 */
2239 if (HZ > USER_HZ)
2240 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2241 else
2242 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2243
2244 if (addrconf_finite_timeout(rt_expires))
2245 rt_expires *= HZ;
2246
2247 rt = addrconf_get_prefix_route(&pinfo->prefix,
2248 pinfo->prefix_len,
2249 dev,
2250 RTF_ADDRCONF | RTF_PREFIX_RT,
2251 RTF_GATEWAY | RTF_DEFAULT);
2252
2253 if (rt) {
2254 /* Autoconf prefix route */
2255 if (valid_lft == 0) {
2256 ip6_del_rt(rt);
2257 rt = NULL;
2258 } else if (addrconf_finite_timeout(rt_expires)) {
2259 /* not infinity */
2260 rt6_set_expires(rt, jiffies + rt_expires);
2261 } else {
2262 rt6_clean_expires(rt);
2263 }
2264 } else if (valid_lft) {
2265 clock_t expires = 0;
2266 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2267 if (addrconf_finite_timeout(rt_expires)) {
2268 /* not infinity */
2269 flags |= RTF_EXPIRES;
2270 expires = jiffies_to_clock_t(rt_expires);
2271 }
2272 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2273 dev, expires, flags);
2274 }
2275 ip6_rt_put(rt);
2276 }
2277
2278 /* Try to figure out our local address for this prefix */
2279
2280 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2281 struct inet6_ifaddr *ifp;
2282 struct in6_addr addr;
2283 int create = 0, update_lft = 0;
2284 bool tokenized = false;
2285
2286 if (pinfo->prefix_len == 64) {
2287 memcpy(&addr, &pinfo->prefix, 8);
2288
2289 if (!ipv6_addr_any(&in6_dev->token)) {
2290 read_lock_bh(&in6_dev->lock);
2291 memcpy(addr.s6_addr + 8,
2292 in6_dev->token.s6_addr + 8, 8);
2293 read_unlock_bh(&in6_dev->lock);
2294 tokenized = true;
2295 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2296 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2297 in6_dev_put(in6_dev);
2298 return;
2299 }
2300 goto ok;
2301 }
2302 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2303 pinfo->prefix_len);
2304 in6_dev_put(in6_dev);
2305 return;
2306
2307 ok:
2308
2309 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2310
2311 if (ifp == NULL && valid_lft) {
2312 int max_addresses = in6_dev->cnf.max_addresses;
2313 u32 addr_flags = 0;
2314
2315 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2316 if (in6_dev->cnf.optimistic_dad &&
2317 !net->ipv6.devconf_all->forwarding && sllao)
2318 addr_flags = IFA_F_OPTIMISTIC;
2319 #endif
2320
2321 /* Do not allow to create too much of autoconfigured
2322 * addresses; this would be too easy way to crash kernel.
2323 */
2324 if (!max_addresses ||
2325 ipv6_count_addresses(in6_dev) < max_addresses)
2326 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2327 pinfo->prefix_len,
2328 addr_type&IPV6_ADDR_SCOPE_MASK,
2329 addr_flags, valid_lft,
2330 prefered_lft);
2331
2332 if (IS_ERR_OR_NULL(ifp)) {
2333 in6_dev_put(in6_dev);
2334 return;
2335 }
2336
2337 update_lft = 0;
2338 create = 1;
2339 spin_lock_bh(&ifp->lock);
2340 ifp->flags |= IFA_F_MANAGETEMPADDR;
2341 ifp->cstamp = jiffies;
2342 ifp->tokenized = tokenized;
2343 spin_unlock_bh(&ifp->lock);
2344 addrconf_dad_start(ifp);
2345 }
2346
2347 if (ifp) {
2348 u32 flags;
2349 unsigned long now;
2350 u32 stored_lft;
2351
2352 /* update lifetime (RFC2462 5.5.3 e) */
2353 spin_lock(&ifp->lock);
2354 now = jiffies;
2355 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2356 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2357 else
2358 stored_lft = 0;
2359 if (!update_lft && !create && stored_lft) {
2360 const u32 minimum_lft = min_t(u32,
2361 stored_lft, MIN_VALID_LIFETIME);
2362 valid_lft = max(valid_lft, minimum_lft);
2363
2364 /* RFC4862 Section 5.5.3e:
2365 * "Note that the preferred lifetime of the
2366 * corresponding address is always reset to
2367 * the Preferred Lifetime in the received
2368 * Prefix Information option, regardless of
2369 * whether the valid lifetime is also reset or
2370 * ignored."
2371 *
2372 * So we should always update prefered_lft here.
2373 */
2374 update_lft = 1;
2375 }
2376
2377 if (update_lft) {
2378 ifp->valid_lft = valid_lft;
2379 ifp->prefered_lft = prefered_lft;
2380 ifp->tstamp = now;
2381 flags = ifp->flags;
2382 ifp->flags &= ~IFA_F_DEPRECATED;
2383 spin_unlock(&ifp->lock);
2384
2385 if (!(flags&IFA_F_TENTATIVE))
2386 ipv6_ifa_notify(0, ifp);
2387 } else
2388 spin_unlock(&ifp->lock);
2389
2390 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2391 create, now);
2392
2393 in6_ifa_put(ifp);
2394 addrconf_verify();
2395 }
2396 }
2397 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2398 in6_dev_put(in6_dev);
2399 }
2400
2401 /*
2402 * Set destination address.
2403 * Special case for SIT interfaces where we create a new "virtual"
2404 * device.
2405 */
2406 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2407 {
2408 struct in6_ifreq ireq;
2409 struct net_device *dev;
2410 int err = -EINVAL;
2411
2412 rtnl_lock();
2413
2414 err = -EFAULT;
2415 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2416 goto err_exit;
2417
2418 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2419
2420 err = -ENODEV;
2421 if (dev == NULL)
2422 goto err_exit;
2423
2424 #if IS_ENABLED(CONFIG_IPV6_SIT)
2425 if (dev->type == ARPHRD_SIT) {
2426 const struct net_device_ops *ops = dev->netdev_ops;
2427 struct ifreq ifr;
2428 struct ip_tunnel_parm p;
2429
2430 err = -EADDRNOTAVAIL;
2431 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2432 goto err_exit;
2433
2434 memset(&p, 0, sizeof(p));
2435 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2436 p.iph.saddr = 0;
2437 p.iph.version = 4;
2438 p.iph.ihl = 5;
2439 p.iph.protocol = IPPROTO_IPV6;
2440 p.iph.ttl = 64;
2441 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2442
2443 if (ops->ndo_do_ioctl) {
2444 mm_segment_t oldfs = get_fs();
2445
2446 set_fs(KERNEL_DS);
2447 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2448 set_fs(oldfs);
2449 } else
2450 err = -EOPNOTSUPP;
2451
2452 if (err == 0) {
2453 err = -ENOBUFS;
2454 dev = __dev_get_by_name(net, p.name);
2455 if (!dev)
2456 goto err_exit;
2457 err = dev_open(dev);
2458 }
2459 }
2460 #endif
2461
2462 err_exit:
2463 rtnl_unlock();
2464 return err;
2465 }
2466
2467 /*
2468 * Manual configuration of address on an interface
2469 */
2470 static int inet6_addr_add(struct net *net, int ifindex,
2471 const struct in6_addr *pfx,
2472 const struct in6_addr *peer_pfx,
2473 unsigned int plen, __u32 ifa_flags,
2474 __u32 prefered_lft, __u32 valid_lft)
2475 {
2476 struct inet6_ifaddr *ifp;
2477 struct inet6_dev *idev;
2478 struct net_device *dev;
2479 int scope;
2480 u32 flags;
2481 clock_t expires;
2482 unsigned long timeout;
2483
2484 ASSERT_RTNL();
2485
2486 if (plen > 128)
2487 return -EINVAL;
2488
2489 /* check the lifetime */
2490 if (!valid_lft || prefered_lft > valid_lft)
2491 return -EINVAL;
2492
2493 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2494 return -EINVAL;
2495
2496 dev = __dev_get_by_index(net, ifindex);
2497 if (!dev)
2498 return -ENODEV;
2499
2500 idev = addrconf_add_dev(dev);
2501 if (IS_ERR(idev))
2502 return PTR_ERR(idev);
2503
2504 scope = ipv6_addr_scope(pfx);
2505
2506 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2507 if (addrconf_finite_timeout(timeout)) {
2508 expires = jiffies_to_clock_t(timeout * HZ);
2509 valid_lft = timeout;
2510 flags = RTF_EXPIRES;
2511 } else {
2512 expires = 0;
2513 flags = 0;
2514 ifa_flags |= IFA_F_PERMANENT;
2515 }
2516
2517 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2518 if (addrconf_finite_timeout(timeout)) {
2519 if (timeout == 0)
2520 ifa_flags |= IFA_F_DEPRECATED;
2521 prefered_lft = timeout;
2522 }
2523
2524 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2525 valid_lft, prefered_lft);
2526
2527 if (!IS_ERR(ifp)) {
2528 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2529 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2530 expires, flags);
2531 }
2532
2533 /*
2534 * Note that section 3.1 of RFC 4429 indicates
2535 * that the Optimistic flag should not be set for
2536 * manually configured addresses
2537 */
2538 addrconf_dad_start(ifp);
2539 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2540 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2541 true, jiffies);
2542 in6_ifa_put(ifp);
2543 addrconf_verify_rtnl();
2544 return 0;
2545 }
2546
2547 return PTR_ERR(ifp);
2548 }
2549
2550 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2551 const struct in6_addr *pfx, unsigned int plen)
2552 {
2553 struct inet6_ifaddr *ifp;
2554 struct inet6_dev *idev;
2555 struct net_device *dev;
2556
2557 if (plen > 128)
2558 return -EINVAL;
2559
2560 dev = __dev_get_by_index(net, ifindex);
2561 if (!dev)
2562 return -ENODEV;
2563
2564 idev = __in6_dev_get(dev);
2565 if (idev == NULL)
2566 return -ENXIO;
2567
2568 read_lock_bh(&idev->lock);
2569 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2570 if (ifp->prefix_len == plen &&
2571 ipv6_addr_equal(pfx, &ifp->addr)) {
2572 in6_ifa_hold(ifp);
2573 read_unlock_bh(&idev->lock);
2574
2575 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2576 (ifa_flags & IFA_F_MANAGETEMPADDR))
2577 manage_tempaddrs(idev, ifp, 0, 0, false,
2578 jiffies);
2579 ipv6_del_addr(ifp);
2580 addrconf_verify_rtnl();
2581 return 0;
2582 }
2583 }
2584 read_unlock_bh(&idev->lock);
2585 return -EADDRNOTAVAIL;
2586 }
2587
2588
2589 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2590 {
2591 struct in6_ifreq ireq;
2592 int err;
2593
2594 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2595 return -EPERM;
2596
2597 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2598 return -EFAULT;
2599
2600 rtnl_lock();
2601 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2602 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2603 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2604 rtnl_unlock();
2605 return err;
2606 }
2607
2608 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2609 {
2610 struct in6_ifreq ireq;
2611 int err;
2612
2613 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2614 return -EPERM;
2615
2616 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2617 return -EFAULT;
2618
2619 rtnl_lock();
2620 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2621 ireq.ifr6_prefixlen);
2622 rtnl_unlock();
2623 return err;
2624 }
2625
2626 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2627 int plen, int scope)
2628 {
2629 struct inet6_ifaddr *ifp;
2630
2631 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2632 scope, IFA_F_PERMANENT,
2633 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2634 if (!IS_ERR(ifp)) {
2635 spin_lock_bh(&ifp->lock);
2636 ifp->flags &= ~IFA_F_TENTATIVE;
2637 spin_unlock_bh(&ifp->lock);
2638 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2639 in6_ifa_put(ifp);
2640 }
2641 }
2642
2643 #if IS_ENABLED(CONFIG_IPV6_SIT)
2644 static void sit_add_v4_addrs(struct inet6_dev *idev)
2645 {
2646 struct in6_addr addr;
2647 struct net_device *dev;
2648 struct net *net = dev_net(idev->dev);
2649 int scope, plen;
2650 u32 pflags = 0;
2651
2652 ASSERT_RTNL();
2653
2654 memset(&addr, 0, sizeof(struct in6_addr));
2655 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2656
2657 if (idev->dev->flags&IFF_POINTOPOINT) {
2658 addr.s6_addr32[0] = htonl(0xfe800000);
2659 scope = IFA_LINK;
2660 plen = 64;
2661 } else {
2662 scope = IPV6_ADDR_COMPATv4;
2663 plen = 96;
2664 pflags |= RTF_NONEXTHOP;
2665 }
2666
2667 if (addr.s6_addr32[3]) {
2668 add_addr(idev, &addr, plen, scope);
2669 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2670 return;
2671 }
2672
2673 for_each_netdev(net, dev) {
2674 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2675 if (in_dev && (dev->flags & IFF_UP)) {
2676 struct in_ifaddr *ifa;
2677
2678 int flag = scope;
2679
2680 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2681
2682 addr.s6_addr32[3] = ifa->ifa_local;
2683
2684 if (ifa->ifa_scope == RT_SCOPE_LINK)
2685 continue;
2686 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2687 if (idev->dev->flags&IFF_POINTOPOINT)
2688 continue;
2689 flag |= IFA_HOST;
2690 }
2691
2692 add_addr(idev, &addr, plen, flag);
2693 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2694 pflags);
2695 }
2696 }
2697 }
2698 }
2699 #endif
2700
2701 static void init_loopback(struct net_device *dev)
2702 {
2703 struct inet6_dev *idev;
2704 struct net_device *sp_dev;
2705 struct inet6_ifaddr *sp_ifa;
2706 struct rt6_info *sp_rt;
2707
2708 /* ::1 */
2709
2710 ASSERT_RTNL();
2711
2712 idev = ipv6_find_idev(dev);
2713 if (idev == NULL) {
2714 pr_debug("%s: add_dev failed\n", __func__);
2715 return;
2716 }
2717
2718 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2719
2720 /* Add routes to other interface's IPv6 addresses */
2721 for_each_netdev(dev_net(dev), sp_dev) {
2722 if (!strcmp(sp_dev->name, dev->name))
2723 continue;
2724
2725 idev = __in6_dev_get(sp_dev);
2726 if (!idev)
2727 continue;
2728
2729 read_lock_bh(&idev->lock);
2730 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2731
2732 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2733 continue;
2734
2735 if (sp_ifa->rt) {
2736 /* This dst has been added to garbage list when
2737 * lo device down, release this obsolete dst and
2738 * reallocate a new router for ifa.
2739 */
2740 if (sp_ifa->rt->dst.obsolete > 0) {
2741 ip6_rt_put(sp_ifa->rt);
2742 sp_ifa->rt = NULL;
2743 } else {
2744 continue;
2745 }
2746 }
2747
2748 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2749
2750 /* Failure cases are ignored */
2751 if (!IS_ERR(sp_rt)) {
2752 sp_ifa->rt = sp_rt;
2753 ip6_ins_rt(sp_rt);
2754 }
2755 }
2756 read_unlock_bh(&idev->lock);
2757 }
2758 }
2759
2760 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2761 {
2762 struct inet6_ifaddr *ifp;
2763 u32 addr_flags = IFA_F_PERMANENT;
2764
2765 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2766 if (idev->cnf.optimistic_dad &&
2767 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2768 addr_flags |= IFA_F_OPTIMISTIC;
2769 #endif
2770
2771
2772 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2773 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2774 if (!IS_ERR(ifp)) {
2775 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2776 addrconf_dad_start(ifp);
2777 in6_ifa_put(ifp);
2778 }
2779 }
2780
2781 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2782 {
2783 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2784 struct in6_addr addr;
2785
2786 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2787 /* addrconf_add_linklocal also adds a prefix_route and we
2788 * only need to care about prefix routes if ipv6_generate_eui64
2789 * couldn't generate one.
2790 */
2791 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2792 addrconf_add_linklocal(idev, &addr);
2793 else if (prefix_route)
2794 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2795 }
2796 }
2797
2798 static void addrconf_dev_config(struct net_device *dev)
2799 {
2800 struct inet6_dev *idev;
2801
2802 ASSERT_RTNL();
2803
2804 if ((dev->type != ARPHRD_ETHER) &&
2805 (dev->type != ARPHRD_FDDI) &&
2806 (dev->type != ARPHRD_ARCNET) &&
2807 (dev->type != ARPHRD_INFINIBAND) &&
2808 (dev->type != ARPHRD_IEEE802154) &&
2809 (dev->type != ARPHRD_IEEE1394) &&
2810 (dev->type != ARPHRD_TUNNEL6) &&
2811 (dev->type != ARPHRD_6LOWPAN)) {
2812 /* Alas, we support only Ethernet autoconfiguration. */
2813 return;
2814 }
2815
2816 idev = addrconf_add_dev(dev);
2817 if (IS_ERR(idev))
2818 return;
2819
2820 addrconf_addr_gen(idev, false);
2821 }
2822
2823 #if IS_ENABLED(CONFIG_IPV6_SIT)
2824 static void addrconf_sit_config(struct net_device *dev)
2825 {
2826 struct inet6_dev *idev;
2827
2828 ASSERT_RTNL();
2829
2830 /*
2831 * Configure the tunnel with one of our IPv4
2832 * addresses... we should configure all of
2833 * our v4 addrs in the tunnel
2834 */
2835
2836 idev = ipv6_find_idev(dev);
2837 if (idev == NULL) {
2838 pr_debug("%s: add_dev failed\n", __func__);
2839 return;
2840 }
2841
2842 if (dev->priv_flags & IFF_ISATAP) {
2843 addrconf_addr_gen(idev, false);
2844 return;
2845 }
2846
2847 sit_add_v4_addrs(idev);
2848
2849 if (dev->flags&IFF_POINTOPOINT)
2850 addrconf_add_mroute(dev);
2851 }
2852 #endif
2853
2854 #if IS_ENABLED(CONFIG_NET_IPGRE)
2855 static void addrconf_gre_config(struct net_device *dev)
2856 {
2857 struct inet6_dev *idev;
2858
2859 ASSERT_RTNL();
2860
2861 idev = ipv6_find_idev(dev);
2862 if (idev == NULL) {
2863 pr_debug("%s: add_dev failed\n", __func__);
2864 return;
2865 }
2866
2867 addrconf_addr_gen(idev, true);
2868 }
2869 #endif
2870
2871 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2872 void *ptr)
2873 {
2874 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2875 struct inet6_dev *idev = __in6_dev_get(dev);
2876 int run_pending = 0;
2877 int err;
2878
2879 switch (event) {
2880 case NETDEV_REGISTER:
2881 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2882 idev = ipv6_add_dev(dev);
2883 if (IS_ERR(idev))
2884 return notifier_from_errno(PTR_ERR(idev));
2885 }
2886 break;
2887
2888 case NETDEV_UP:
2889 case NETDEV_CHANGE:
2890 if (dev->flags & IFF_SLAVE)
2891 break;
2892
2893 if (idev && idev->cnf.disable_ipv6)
2894 break;
2895
2896 if (event == NETDEV_UP) {
2897 if (!addrconf_qdisc_ok(dev)) {
2898 /* device is not ready yet. */
2899 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2900 dev->name);
2901 break;
2902 }
2903
2904 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2905 idev = ipv6_add_dev(dev);
2906
2907 if (!IS_ERR_OR_NULL(idev)) {
2908 idev->if_flags |= IF_READY;
2909 run_pending = 1;
2910 }
2911 } else {
2912 if (!addrconf_qdisc_ok(dev)) {
2913 /* device is still not ready. */
2914 break;
2915 }
2916
2917 if (idev) {
2918 if (idev->if_flags & IF_READY)
2919 /* device is already configured. */
2920 break;
2921 idev->if_flags |= IF_READY;
2922 }
2923
2924 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2925 dev->name);
2926
2927 run_pending = 1;
2928 }
2929
2930 switch (dev->type) {
2931 #if IS_ENABLED(CONFIG_IPV6_SIT)
2932 case ARPHRD_SIT:
2933 addrconf_sit_config(dev);
2934 break;
2935 #endif
2936 #if IS_ENABLED(CONFIG_NET_IPGRE)
2937 case ARPHRD_IPGRE:
2938 addrconf_gre_config(dev);
2939 break;
2940 #endif
2941 case ARPHRD_LOOPBACK:
2942 init_loopback(dev);
2943 break;
2944
2945 default:
2946 addrconf_dev_config(dev);
2947 break;
2948 }
2949
2950 if (!IS_ERR_OR_NULL(idev)) {
2951 if (run_pending)
2952 addrconf_dad_run(idev);
2953
2954 /*
2955 * If the MTU changed during the interface down,
2956 * when the interface up, the changed MTU must be
2957 * reflected in the idev as well as routers.
2958 */
2959 if (idev->cnf.mtu6 != dev->mtu &&
2960 dev->mtu >= IPV6_MIN_MTU) {
2961 rt6_mtu_change(dev, dev->mtu);
2962 idev->cnf.mtu6 = dev->mtu;
2963 }
2964 idev->tstamp = jiffies;
2965 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2966
2967 /*
2968 * If the changed mtu during down is lower than
2969 * IPV6_MIN_MTU stop IPv6 on this interface.
2970 */
2971 if (dev->mtu < IPV6_MIN_MTU)
2972 addrconf_ifdown(dev, 1);
2973 }
2974 break;
2975
2976 case NETDEV_CHANGEMTU:
2977 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2978 rt6_mtu_change(dev, dev->mtu);
2979 idev->cnf.mtu6 = dev->mtu;
2980 break;
2981 }
2982
2983 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2984 idev = ipv6_add_dev(dev);
2985 if (!IS_ERR(idev))
2986 break;
2987 }
2988
2989 /*
2990 * if MTU under IPV6_MIN_MTU.
2991 * Stop IPv6 on this interface.
2992 */
2993
2994 case NETDEV_DOWN:
2995 case NETDEV_UNREGISTER:
2996 /*
2997 * Remove all addresses from this interface.
2998 */
2999 addrconf_ifdown(dev, event != NETDEV_DOWN);
3000 break;
3001
3002 case NETDEV_CHANGENAME:
3003 if (idev) {
3004 snmp6_unregister_dev(idev);
3005 addrconf_sysctl_unregister(idev);
3006 err = addrconf_sysctl_register(idev);
3007 if (err)
3008 return notifier_from_errno(err);
3009 err = snmp6_register_dev(idev);
3010 if (err) {
3011 addrconf_sysctl_unregister(idev);
3012 return notifier_from_errno(err);
3013 }
3014 }
3015 break;
3016
3017 case NETDEV_PRE_TYPE_CHANGE:
3018 case NETDEV_POST_TYPE_CHANGE:
3019 addrconf_type_change(dev, event);
3020 break;
3021 }
3022
3023 return NOTIFY_OK;
3024 }
3025
3026 /*
3027 * addrconf module should be notified of a device going up
3028 */
3029 static struct notifier_block ipv6_dev_notf = {
3030 .notifier_call = addrconf_notify,
3031 };
3032
3033 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3034 {
3035 struct inet6_dev *idev;
3036 ASSERT_RTNL();
3037
3038 idev = __in6_dev_get(dev);
3039
3040 if (event == NETDEV_POST_TYPE_CHANGE)
3041 ipv6_mc_remap(idev);
3042 else if (event == NETDEV_PRE_TYPE_CHANGE)
3043 ipv6_mc_unmap(idev);
3044 }
3045
3046 static int addrconf_ifdown(struct net_device *dev, int how)
3047 {
3048 struct net *net = dev_net(dev);
3049 struct inet6_dev *idev;
3050 struct inet6_ifaddr *ifa;
3051 int state, i;
3052
3053 ASSERT_RTNL();
3054
3055 rt6_ifdown(net, dev);
3056 neigh_ifdown(&nd_tbl, dev);
3057
3058 idev = __in6_dev_get(dev);
3059 if (idev == NULL)
3060 return -ENODEV;
3061
3062 /*
3063 * Step 1: remove reference to ipv6 device from parent device.
3064 * Do not dev_put!
3065 */
3066 if (how) {
3067 idev->dead = 1;
3068
3069 /* protected by rtnl_lock */
3070 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3071
3072 /* Step 1.5: remove snmp6 entry */
3073 snmp6_unregister_dev(idev);
3074
3075 }
3076
3077 /* Step 2: clear hash table */
3078 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3079 struct hlist_head *h = &inet6_addr_lst[i];
3080
3081 spin_lock_bh(&addrconf_hash_lock);
3082 restart:
3083 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3084 if (ifa->idev == idev) {
3085 hlist_del_init_rcu(&ifa->addr_lst);
3086 addrconf_del_dad_work(ifa);
3087 goto restart;
3088 }
3089 }
3090 spin_unlock_bh(&addrconf_hash_lock);
3091 }
3092
3093 write_lock_bh(&idev->lock);
3094
3095 addrconf_del_rs_timer(idev);
3096
3097 /* Step 2: clear flags for stateless addrconf */
3098 if (!how)
3099 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3100
3101 if (how && del_timer(&idev->regen_timer))
3102 in6_dev_put(idev);
3103
3104 /* Step 3: clear tempaddr list */
3105 while (!list_empty(&idev->tempaddr_list)) {
3106 ifa = list_first_entry(&idev->tempaddr_list,
3107 struct inet6_ifaddr, tmp_list);
3108 list_del(&ifa->tmp_list);
3109 write_unlock_bh(&idev->lock);
3110 spin_lock_bh(&ifa->lock);
3111
3112 if (ifa->ifpub) {
3113 in6_ifa_put(ifa->ifpub);
3114 ifa->ifpub = NULL;
3115 }
3116 spin_unlock_bh(&ifa->lock);
3117 in6_ifa_put(ifa);
3118 write_lock_bh(&idev->lock);
3119 }
3120
3121 while (!list_empty(&idev->addr_list)) {
3122 ifa = list_first_entry(&idev->addr_list,
3123 struct inet6_ifaddr, if_list);
3124 addrconf_del_dad_work(ifa);
3125
3126 list_del(&ifa->if_list);
3127
3128 write_unlock_bh(&idev->lock);
3129
3130 spin_lock_bh(&ifa->state_lock);
3131 state = ifa->state;
3132 ifa->state = INET6_IFADDR_STATE_DEAD;
3133 spin_unlock_bh(&ifa->state_lock);
3134
3135 if (state != INET6_IFADDR_STATE_DEAD) {
3136 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3137 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3138 }
3139 in6_ifa_put(ifa);
3140
3141 write_lock_bh(&idev->lock);
3142 }
3143
3144 write_unlock_bh(&idev->lock);
3145
3146 /* Step 5: Discard anycast and multicast list */
3147 if (how) {
3148 ipv6_ac_destroy_dev(idev);
3149 ipv6_mc_destroy_dev(idev);
3150 } else {
3151 ipv6_mc_down(idev);
3152 }
3153
3154 idev->tstamp = jiffies;
3155
3156 /* Last: Shot the device (if unregistered) */
3157 if (how) {
3158 addrconf_sysctl_unregister(idev);
3159 neigh_parms_release(&nd_tbl, idev->nd_parms);
3160 neigh_ifdown(&nd_tbl, dev);
3161 in6_dev_put(idev);
3162 }
3163 return 0;
3164 }
3165
3166 static void addrconf_rs_timer(unsigned long data)
3167 {
3168 struct inet6_dev *idev = (struct inet6_dev *)data;
3169 struct net_device *dev = idev->dev;
3170 struct in6_addr lladdr;
3171
3172 write_lock(&idev->lock);
3173 if (idev->dead || !(idev->if_flags & IF_READY))
3174 goto out;
3175
3176 if (!ipv6_accept_ra(idev))
3177 goto out;
3178
3179 /* Announcement received after solicitation was sent */
3180 if (idev->if_flags & IF_RA_RCVD)
3181 goto out;
3182
3183 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3184 write_unlock(&idev->lock);
3185 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3186 ndisc_send_rs(dev, &lladdr,
3187 &in6addr_linklocal_allrouters);
3188 else
3189 goto put;
3190
3191 write_lock(&idev->lock);
3192 /* The wait after the last probe can be shorter */
3193 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3194 idev->cnf.rtr_solicits) ?
3195 idev->cnf.rtr_solicit_delay :
3196 idev->cnf.rtr_solicit_interval);
3197 } else {
3198 /*
3199 * Note: we do not support deprecated "all on-link"
3200 * assumption any longer.
3201 */
3202 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3203 }
3204
3205 out:
3206 write_unlock(&idev->lock);
3207 put:
3208 in6_dev_put(idev);
3209 }
3210
3211 /*
3212 * Duplicate Address Detection
3213 */
3214 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3215 {
3216 unsigned long rand_num;
3217 struct inet6_dev *idev = ifp->idev;
3218
3219 if (ifp->flags & IFA_F_OPTIMISTIC)
3220 rand_num = 0;
3221 else
3222 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3223
3224 ifp->dad_probes = idev->cnf.dad_transmits;
3225 addrconf_mod_dad_work(ifp, rand_num);
3226 }
3227
3228 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3229 {
3230 struct inet6_dev *idev = ifp->idev;
3231 struct net_device *dev = idev->dev;
3232
3233 addrconf_join_solict(dev, &ifp->addr);
3234
3235 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3236
3237 read_lock_bh(&idev->lock);
3238 spin_lock(&ifp->lock);
3239 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3240 goto out;
3241
3242 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3243 idev->cnf.accept_dad < 1 ||
3244 !(ifp->flags&IFA_F_TENTATIVE) ||
3245 ifp->flags & IFA_F_NODAD) {
3246 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3247 spin_unlock(&ifp->lock);
3248 read_unlock_bh(&idev->lock);
3249
3250 addrconf_dad_completed(ifp);
3251 return;
3252 }
3253
3254 if (!(idev->if_flags & IF_READY)) {
3255 spin_unlock(&ifp->lock);
3256 read_unlock_bh(&idev->lock);
3257 /*
3258 * If the device is not ready:
3259 * - keep it tentative if it is a permanent address.
3260 * - otherwise, kill it.
3261 */
3262 in6_ifa_hold(ifp);
3263 addrconf_dad_stop(ifp, 0);
3264 return;
3265 }
3266
3267 /*
3268 * Optimistic nodes can start receiving
3269 * Frames right away
3270 */
3271 if (ifp->flags & IFA_F_OPTIMISTIC) {
3272 ip6_ins_rt(ifp->rt);
3273 if (ipv6_use_optimistic_addr(idev)) {
3274 /* Because optimistic nodes can use this address,
3275 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3276 */
3277 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3278 }
3279 }
3280
3281 addrconf_dad_kick(ifp);
3282 out:
3283 spin_unlock(&ifp->lock);
3284 read_unlock_bh(&idev->lock);
3285 }
3286
3287 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3288 {
3289 bool begin_dad = false;
3290
3291 spin_lock_bh(&ifp->state_lock);
3292 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3293 ifp->state = INET6_IFADDR_STATE_PREDAD;
3294 begin_dad = true;
3295 }
3296 spin_unlock_bh(&ifp->state_lock);
3297
3298 if (begin_dad)
3299 addrconf_mod_dad_work(ifp, 0);
3300 }
3301
3302 static void addrconf_dad_work(struct work_struct *w)
3303 {
3304 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3305 struct inet6_ifaddr,
3306 dad_work);
3307 struct inet6_dev *idev = ifp->idev;
3308 struct in6_addr mcaddr;
3309
3310 enum {
3311 DAD_PROCESS,
3312 DAD_BEGIN,
3313 DAD_ABORT,
3314 } action = DAD_PROCESS;
3315
3316 rtnl_lock();
3317
3318 spin_lock_bh(&ifp->state_lock);
3319 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3320 action = DAD_BEGIN;
3321 ifp->state = INET6_IFADDR_STATE_DAD;
3322 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3323 action = DAD_ABORT;
3324 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3325 }
3326 spin_unlock_bh(&ifp->state_lock);
3327
3328 if (action == DAD_BEGIN) {
3329 addrconf_dad_begin(ifp);
3330 goto out;
3331 } else if (action == DAD_ABORT) {
3332 addrconf_dad_stop(ifp, 1);
3333 goto out;
3334 }
3335
3336 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3337 goto out;
3338
3339 write_lock_bh(&idev->lock);
3340 if (idev->dead || !(idev->if_flags & IF_READY)) {
3341 write_unlock_bh(&idev->lock);
3342 goto out;
3343 }
3344
3345 spin_lock(&ifp->lock);
3346 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3347 spin_unlock(&ifp->lock);
3348 write_unlock_bh(&idev->lock);
3349 goto out;
3350 }
3351
3352 if (ifp->dad_probes == 0) {
3353 /*
3354 * DAD was successful
3355 */
3356
3357 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3358 spin_unlock(&ifp->lock);
3359 write_unlock_bh(&idev->lock);
3360
3361 addrconf_dad_completed(ifp);
3362
3363 goto out;
3364 }
3365
3366 ifp->dad_probes--;
3367 addrconf_mod_dad_work(ifp,
3368 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3369 spin_unlock(&ifp->lock);
3370 write_unlock_bh(&idev->lock);
3371
3372 /* send a neighbour solicitation for our addr */
3373 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3374 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3375 out:
3376 in6_ifa_put(ifp);
3377 rtnl_unlock();
3378 }
3379
3380 /* ifp->idev must be at least read locked */
3381 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3382 {
3383 struct inet6_ifaddr *ifpiter;
3384 struct inet6_dev *idev = ifp->idev;
3385
3386 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3387 if (ifpiter->scope > IFA_LINK)
3388 break;
3389 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3390 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3391 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3392 IFA_F_PERMANENT)
3393 return false;
3394 }
3395 return true;
3396 }
3397
3398 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3399 {
3400 struct net_device *dev = ifp->idev->dev;
3401 struct in6_addr lladdr;
3402 bool send_rs, send_mld;
3403
3404 addrconf_del_dad_work(ifp);
3405
3406 /*
3407 * Configure the address for reception. Now it is valid.
3408 */
3409
3410 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3411
3412 /* If added prefix is link local and we are prepared to process
3413 router advertisements, start sending router solicitations.
3414 */
3415
3416 read_lock_bh(&ifp->idev->lock);
3417 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3418 send_rs = send_mld &&
3419 ipv6_accept_ra(ifp->idev) &&
3420 ifp->idev->cnf.rtr_solicits > 0 &&
3421 (dev->flags&IFF_LOOPBACK) == 0;
3422 read_unlock_bh(&ifp->idev->lock);
3423
3424 /* While dad is in progress mld report's source address is in6_addrany.
3425 * Resend with proper ll now.
3426 */
3427 if (send_mld)
3428 ipv6_mc_dad_complete(ifp->idev);
3429
3430 if (send_rs) {
3431 /*
3432 * If a host as already performed a random delay
3433 * [...] as part of DAD [...] there is no need
3434 * to delay again before sending the first RS
3435 */
3436 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3437 return;
3438 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3439
3440 write_lock_bh(&ifp->idev->lock);
3441 spin_lock(&ifp->lock);
3442 ifp->idev->rs_probes = 1;
3443 ifp->idev->if_flags |= IF_RS_SENT;
3444 addrconf_mod_rs_timer(ifp->idev,
3445 ifp->idev->cnf.rtr_solicit_interval);
3446 spin_unlock(&ifp->lock);
3447 write_unlock_bh(&ifp->idev->lock);
3448 }
3449 }
3450
3451 static void addrconf_dad_run(struct inet6_dev *idev)
3452 {
3453 struct inet6_ifaddr *ifp;
3454
3455 read_lock_bh(&idev->lock);
3456 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3457 spin_lock(&ifp->lock);
3458 if (ifp->flags & IFA_F_TENTATIVE &&
3459 ifp->state == INET6_IFADDR_STATE_DAD)
3460 addrconf_dad_kick(ifp);
3461 spin_unlock(&ifp->lock);
3462 }
3463 read_unlock_bh(&idev->lock);
3464 }
3465
3466 #ifdef CONFIG_PROC_FS
3467 struct if6_iter_state {
3468 struct seq_net_private p;
3469 int bucket;
3470 int offset;
3471 };
3472
3473 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3474 {
3475 struct inet6_ifaddr *ifa = NULL;
3476 struct if6_iter_state *state = seq->private;
3477 struct net *net = seq_file_net(seq);
3478 int p = 0;
3479
3480 /* initial bucket if pos is 0 */
3481 if (pos == 0) {
3482 state->bucket = 0;
3483 state->offset = 0;
3484 }
3485
3486 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3487 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3488 addr_lst) {
3489 if (!net_eq(dev_net(ifa->idev->dev), net))
3490 continue;
3491 /* sync with offset */
3492 if (p < state->offset) {
3493 p++;
3494 continue;
3495 }
3496 state->offset++;
3497 return ifa;
3498 }
3499
3500 /* prepare for next bucket */
3501 state->offset = 0;
3502 p = 0;
3503 }
3504 return NULL;
3505 }
3506
3507 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3508 struct inet6_ifaddr *ifa)
3509 {
3510 struct if6_iter_state *state = seq->private;
3511 struct net *net = seq_file_net(seq);
3512
3513 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3514 if (!net_eq(dev_net(ifa->idev->dev), net))
3515 continue;
3516 state->offset++;
3517 return ifa;
3518 }
3519
3520 while (++state->bucket < IN6_ADDR_HSIZE) {
3521 state->offset = 0;
3522 hlist_for_each_entry_rcu_bh(ifa,
3523 &inet6_addr_lst[state->bucket], addr_lst) {
3524 if (!net_eq(dev_net(ifa->idev->dev), net))
3525 continue;
3526 state->offset++;
3527 return ifa;
3528 }
3529 }
3530
3531 return NULL;
3532 }
3533
3534 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3535 __acquires(rcu_bh)
3536 {
3537 rcu_read_lock_bh();
3538 return if6_get_first(seq, *pos);
3539 }
3540
3541 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3542 {
3543 struct inet6_ifaddr *ifa;
3544
3545 ifa = if6_get_next(seq, v);
3546 ++*pos;
3547 return ifa;
3548 }
3549
3550 static void if6_seq_stop(struct seq_file *seq, void *v)
3551 __releases(rcu_bh)
3552 {
3553 rcu_read_unlock_bh();
3554 }
3555
3556 static int if6_seq_show(struct seq_file *seq, void *v)
3557 {
3558 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3559 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3560 &ifp->addr,
3561 ifp->idev->dev->ifindex,
3562 ifp->prefix_len,
3563 ifp->scope,
3564 (u8) ifp->flags,
3565 ifp->idev->dev->name);
3566 return 0;
3567 }
3568
3569 static const struct seq_operations if6_seq_ops = {
3570 .start = if6_seq_start,
3571 .next = if6_seq_next,
3572 .show = if6_seq_show,
3573 .stop = if6_seq_stop,
3574 };
3575
3576 static int if6_seq_open(struct inode *inode, struct file *file)
3577 {
3578 return seq_open_net(inode, file, &if6_seq_ops,
3579 sizeof(struct if6_iter_state));
3580 }
3581
3582 static const struct file_operations if6_fops = {
3583 .owner = THIS_MODULE,
3584 .open = if6_seq_open,
3585 .read = seq_read,
3586 .llseek = seq_lseek,
3587 .release = seq_release_net,
3588 };
3589
3590 static int __net_init if6_proc_net_init(struct net *net)
3591 {
3592 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3593 return -ENOMEM;
3594 return 0;
3595 }
3596
3597 static void __net_exit if6_proc_net_exit(struct net *net)
3598 {
3599 remove_proc_entry("if_inet6", net->proc_net);
3600 }
3601
3602 static struct pernet_operations if6_proc_net_ops = {
3603 .init = if6_proc_net_init,
3604 .exit = if6_proc_net_exit,
3605 };
3606
3607 int __init if6_proc_init(void)
3608 {
3609 return register_pernet_subsys(&if6_proc_net_ops);
3610 }
3611
3612 void if6_proc_exit(void)
3613 {
3614 unregister_pernet_subsys(&if6_proc_net_ops);
3615 }
3616 #endif /* CONFIG_PROC_FS */
3617
3618 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3619 /* Check if address is a home address configured on any interface. */
3620 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3621 {
3622 int ret = 0;
3623 struct inet6_ifaddr *ifp = NULL;
3624 unsigned int hash = inet6_addr_hash(addr);
3625
3626 rcu_read_lock_bh();
3627 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3628 if (!net_eq(dev_net(ifp->idev->dev), net))
3629 continue;
3630 if (ipv6_addr_equal(&ifp->addr, addr) &&
3631 (ifp->flags & IFA_F_HOMEADDRESS)) {
3632 ret = 1;
3633 break;
3634 }
3635 }
3636 rcu_read_unlock_bh();
3637 return ret;
3638 }
3639 #endif
3640
3641 /*
3642 * Periodic address status verification
3643 */
3644
3645 static void addrconf_verify_rtnl(void)
3646 {
3647 unsigned long now, next, next_sec, next_sched;
3648 struct inet6_ifaddr *ifp;
3649 int i;
3650
3651 ASSERT_RTNL();
3652
3653 rcu_read_lock_bh();
3654 now = jiffies;
3655 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3656
3657 cancel_delayed_work(&addr_chk_work);
3658
3659 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3660 restart:
3661 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3662 unsigned long age;
3663
3664 /* When setting preferred_lft to a value not zero or
3665 * infinity, while valid_lft is infinity
3666 * IFA_F_PERMANENT has a non-infinity life time.
3667 */
3668 if ((ifp->flags & IFA_F_PERMANENT) &&
3669 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3670 continue;
3671
3672 spin_lock(&ifp->lock);
3673 /* We try to batch several events at once. */
3674 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3675
3676 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3677 age >= ifp->valid_lft) {
3678 spin_unlock(&ifp->lock);
3679 in6_ifa_hold(ifp);
3680 ipv6_del_addr(ifp);
3681 goto restart;
3682 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3683 spin_unlock(&ifp->lock);
3684 continue;
3685 } else if (age >= ifp->prefered_lft) {
3686 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3687 int deprecate = 0;
3688
3689 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3690 deprecate = 1;
3691 ifp->flags |= IFA_F_DEPRECATED;
3692 }
3693
3694 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3695 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3696 next = ifp->tstamp + ifp->valid_lft * HZ;
3697
3698 spin_unlock(&ifp->lock);
3699
3700 if (deprecate) {
3701 in6_ifa_hold(ifp);
3702
3703 ipv6_ifa_notify(0, ifp);
3704 in6_ifa_put(ifp);
3705 goto restart;
3706 }
3707 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3708 !(ifp->flags&IFA_F_TENTATIVE)) {
3709 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3710 ifp->idev->cnf.dad_transmits *
3711 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3712
3713 if (age >= ifp->prefered_lft - regen_advance) {
3714 struct inet6_ifaddr *ifpub = ifp->ifpub;
3715 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3716 next = ifp->tstamp + ifp->prefered_lft * HZ;
3717 if (!ifp->regen_count && ifpub) {
3718 ifp->regen_count++;
3719 in6_ifa_hold(ifp);
3720 in6_ifa_hold(ifpub);
3721 spin_unlock(&ifp->lock);
3722
3723 spin_lock(&ifpub->lock);
3724 ifpub->regen_count = 0;
3725 spin_unlock(&ifpub->lock);
3726 ipv6_create_tempaddr(ifpub, ifp);
3727 in6_ifa_put(ifpub);
3728 in6_ifa_put(ifp);
3729 goto restart;
3730 }
3731 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3732 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3733 spin_unlock(&ifp->lock);
3734 } else {
3735 /* ifp->prefered_lft <= ifp->valid_lft */
3736 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3737 next = ifp->tstamp + ifp->prefered_lft * HZ;
3738 spin_unlock(&ifp->lock);
3739 }
3740 }
3741 }
3742
3743 next_sec = round_jiffies_up(next);
3744 next_sched = next;
3745
3746 /* If rounded timeout is accurate enough, accept it. */
3747 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3748 next_sched = next_sec;
3749
3750 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3751 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3752 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3753
3754 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3755 now, next, next_sec, next_sched);
3756 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3757 rcu_read_unlock_bh();
3758 }
3759
3760 static void addrconf_verify_work(struct work_struct *w)
3761 {
3762 rtnl_lock();
3763 addrconf_verify_rtnl();
3764 rtnl_unlock();
3765 }
3766
3767 static void addrconf_verify(void)
3768 {
3769 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3770 }
3771
3772 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3773 struct in6_addr **peer_pfx)
3774 {
3775 struct in6_addr *pfx = NULL;
3776
3777 *peer_pfx = NULL;
3778
3779 if (addr)
3780 pfx = nla_data(addr);
3781
3782 if (local) {
3783 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3784 *peer_pfx = pfx;
3785 pfx = nla_data(local);
3786 }
3787
3788 return pfx;
3789 }
3790
3791 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3792 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3793 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3794 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3795 [IFA_FLAGS] = { .len = sizeof(u32) },
3796 };
3797
3798 static int
3799 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3800 {
3801 struct net *net = sock_net(skb->sk);
3802 struct ifaddrmsg *ifm;
3803 struct nlattr *tb[IFA_MAX+1];
3804 struct in6_addr *pfx, *peer_pfx;
3805 u32 ifa_flags;
3806 int err;
3807
3808 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3809 if (err < 0)
3810 return err;
3811
3812 ifm = nlmsg_data(nlh);
3813 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3814 if (pfx == NULL)
3815 return -EINVAL;
3816
3817 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3818
3819 /* We ignore other flags so far. */
3820 ifa_flags &= IFA_F_MANAGETEMPADDR;
3821
3822 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3823 ifm->ifa_prefixlen);
3824 }
3825
3826 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3827 u32 prefered_lft, u32 valid_lft)
3828 {
3829 u32 flags;
3830 clock_t expires;
3831 unsigned long timeout;
3832 bool was_managetempaddr;
3833 bool had_prefixroute;
3834
3835 ASSERT_RTNL();
3836
3837 if (!valid_lft || (prefered_lft > valid_lft))
3838 return -EINVAL;
3839
3840 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3841 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3842 return -EINVAL;
3843
3844 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3845 if (addrconf_finite_timeout(timeout)) {
3846 expires = jiffies_to_clock_t(timeout * HZ);
3847 valid_lft = timeout;
3848 flags = RTF_EXPIRES;
3849 } else {
3850 expires = 0;
3851 flags = 0;
3852 ifa_flags |= IFA_F_PERMANENT;
3853 }
3854
3855 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3856 if (addrconf_finite_timeout(timeout)) {
3857 if (timeout == 0)
3858 ifa_flags |= IFA_F_DEPRECATED;
3859 prefered_lft = timeout;
3860 }
3861
3862 spin_lock_bh(&ifp->lock);
3863 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3864 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3865 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3866 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3867 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3868 IFA_F_NOPREFIXROUTE);
3869 ifp->flags |= ifa_flags;
3870 ifp->tstamp = jiffies;
3871 ifp->valid_lft = valid_lft;
3872 ifp->prefered_lft = prefered_lft;
3873
3874 spin_unlock_bh(&ifp->lock);
3875 if (!(ifp->flags&IFA_F_TENTATIVE))
3876 ipv6_ifa_notify(0, ifp);
3877
3878 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3879 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3880 expires, flags);
3881 } else if (had_prefixroute) {
3882 enum cleanup_prefix_rt_t action;
3883 unsigned long rt_expires;
3884
3885 write_lock_bh(&ifp->idev->lock);
3886 action = check_cleanup_prefix_route(ifp, &rt_expires);
3887 write_unlock_bh(&ifp->idev->lock);
3888
3889 if (action != CLEANUP_PREFIX_RT_NOP) {
3890 cleanup_prefix_route(ifp, rt_expires,
3891 action == CLEANUP_PREFIX_RT_DEL);
3892 }
3893 }
3894
3895 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3896 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3897 valid_lft = prefered_lft = 0;
3898 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3899 !was_managetempaddr, jiffies);
3900 }
3901
3902 addrconf_verify_rtnl();
3903
3904 return 0;
3905 }
3906
3907 static int
3908 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3909 {
3910 struct net *net = sock_net(skb->sk);
3911 struct ifaddrmsg *ifm;
3912 struct nlattr *tb[IFA_MAX+1];
3913 struct in6_addr *pfx, *peer_pfx;
3914 struct inet6_ifaddr *ifa;
3915 struct net_device *dev;
3916 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3917 u32 ifa_flags;
3918 int err;
3919
3920 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3921 if (err < 0)
3922 return err;
3923
3924 ifm = nlmsg_data(nlh);
3925 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3926 if (pfx == NULL)
3927 return -EINVAL;
3928
3929 if (tb[IFA_CACHEINFO]) {
3930 struct ifa_cacheinfo *ci;
3931
3932 ci = nla_data(tb[IFA_CACHEINFO]);
3933 valid_lft = ci->ifa_valid;
3934 preferred_lft = ci->ifa_prefered;
3935 } else {
3936 preferred_lft = INFINITY_LIFE_TIME;
3937 valid_lft = INFINITY_LIFE_TIME;
3938 }
3939
3940 dev = __dev_get_by_index(net, ifm->ifa_index);
3941 if (dev == NULL)
3942 return -ENODEV;
3943
3944 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3945
3946 /* We ignore other flags so far. */
3947 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3948 IFA_F_NOPREFIXROUTE;
3949
3950 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3951 if (ifa == NULL) {
3952 /*
3953 * It would be best to check for !NLM_F_CREATE here but
3954 * userspace already relies on not having to provide this.
3955 */
3956 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3957 ifm->ifa_prefixlen, ifa_flags,
3958 preferred_lft, valid_lft);
3959 }
3960
3961 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3962 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3963 err = -EEXIST;
3964 else
3965 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3966
3967 in6_ifa_put(ifa);
3968
3969 return err;
3970 }
3971
3972 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3973 u8 scope, int ifindex)
3974 {
3975 struct ifaddrmsg *ifm;
3976
3977 ifm = nlmsg_data(nlh);
3978 ifm->ifa_family = AF_INET6;
3979 ifm->ifa_prefixlen = prefixlen;
3980 ifm->ifa_flags = flags;
3981 ifm->ifa_scope = scope;
3982 ifm->ifa_index = ifindex;
3983 }
3984
3985 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3986 unsigned long tstamp, u32 preferred, u32 valid)
3987 {
3988 struct ifa_cacheinfo ci;
3989
3990 ci.cstamp = cstamp_delta(cstamp);
3991 ci.tstamp = cstamp_delta(tstamp);
3992 ci.ifa_prefered = preferred;
3993 ci.ifa_valid = valid;
3994
3995 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3996 }
3997
3998 static inline int rt_scope(int ifa_scope)
3999 {
4000 if (ifa_scope & IFA_HOST)
4001 return RT_SCOPE_HOST;
4002 else if (ifa_scope & IFA_LINK)
4003 return RT_SCOPE_LINK;
4004 else if (ifa_scope & IFA_SITE)
4005 return RT_SCOPE_SITE;
4006 else
4007 return RT_SCOPE_UNIVERSE;
4008 }
4009
4010 static inline int inet6_ifaddr_msgsize(void)
4011 {
4012 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4013 + nla_total_size(16) /* IFA_LOCAL */
4014 + nla_total_size(16) /* IFA_ADDRESS */
4015 + nla_total_size(sizeof(struct ifa_cacheinfo))
4016 + nla_total_size(4) /* IFA_FLAGS */;
4017 }
4018
4019 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4020 u32 portid, u32 seq, int event, unsigned int flags)
4021 {
4022 struct nlmsghdr *nlh;
4023 u32 preferred, valid;
4024
4025 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4026 if (nlh == NULL)
4027 return -EMSGSIZE;
4028
4029 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4030 ifa->idev->dev->ifindex);
4031
4032 if (!((ifa->flags&IFA_F_PERMANENT) &&
4033 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4034 preferred = ifa->prefered_lft;
4035 valid = ifa->valid_lft;
4036 if (preferred != INFINITY_LIFE_TIME) {
4037 long tval = (jiffies - ifa->tstamp)/HZ;
4038 if (preferred > tval)
4039 preferred -= tval;
4040 else
4041 preferred = 0;
4042 if (valid != INFINITY_LIFE_TIME) {
4043 if (valid > tval)
4044 valid -= tval;
4045 else
4046 valid = 0;
4047 }
4048 }
4049 } else {
4050 preferred = INFINITY_LIFE_TIME;
4051 valid = INFINITY_LIFE_TIME;
4052 }
4053
4054 if (!ipv6_addr_any(&ifa->peer_addr)) {
4055 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
4056 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4057 goto error;
4058 } else
4059 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4060 goto error;
4061
4062 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4063 goto error;
4064
4065 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4066 goto error;
4067
4068 nlmsg_end(skb, nlh);
4069 return 0;
4070
4071 error:
4072 nlmsg_cancel(skb, nlh);
4073 return -EMSGSIZE;
4074 }
4075
4076 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4077 u32 portid, u32 seq, int event, u16 flags)
4078 {
4079 struct nlmsghdr *nlh;
4080 u8 scope = RT_SCOPE_UNIVERSE;
4081 int ifindex = ifmca->idev->dev->ifindex;
4082
4083 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4084 scope = RT_SCOPE_SITE;
4085
4086 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4087 if (nlh == NULL)
4088 return -EMSGSIZE;
4089
4090 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4091 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4092 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4093 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4094 nlmsg_cancel(skb, nlh);
4095 return -EMSGSIZE;
4096 }
4097
4098 nlmsg_end(skb, nlh);
4099 return 0;
4100 }
4101
4102 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4103 u32 portid, u32 seq, int event, unsigned int flags)
4104 {
4105 struct nlmsghdr *nlh;
4106 u8 scope = RT_SCOPE_UNIVERSE;
4107 int ifindex = ifaca->aca_idev->dev->ifindex;
4108
4109 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4110 scope = RT_SCOPE_SITE;
4111
4112 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4113 if (nlh == NULL)
4114 return -EMSGSIZE;
4115
4116 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4117 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4118 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4119 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4120 nlmsg_cancel(skb, nlh);
4121 return -EMSGSIZE;
4122 }
4123
4124 nlmsg_end(skb, nlh);
4125 return 0;
4126 }
4127
4128 enum addr_type_t {
4129 UNICAST_ADDR,
4130 MULTICAST_ADDR,
4131 ANYCAST_ADDR,
4132 };
4133
4134 /* called with rcu_read_lock() */
4135 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4136 struct netlink_callback *cb, enum addr_type_t type,
4137 int s_ip_idx, int *p_ip_idx)
4138 {
4139 struct ifmcaddr6 *ifmca;
4140 struct ifacaddr6 *ifaca;
4141 int err = 1;
4142 int ip_idx = *p_ip_idx;
4143
4144 read_lock_bh(&idev->lock);
4145 switch (type) {
4146 case UNICAST_ADDR: {
4147 struct inet6_ifaddr *ifa;
4148
4149 /* unicast address incl. temp addr */
4150 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4151 if (++ip_idx < s_ip_idx)
4152 continue;
4153 err = inet6_fill_ifaddr(skb, ifa,
4154 NETLINK_CB(cb->skb).portid,
4155 cb->nlh->nlmsg_seq,
4156 RTM_NEWADDR,
4157 NLM_F_MULTI);
4158 if (err < 0)
4159 break;
4160 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4161 }
4162 break;
4163 }
4164 case MULTICAST_ADDR:
4165 /* multicast address */
4166 for (ifmca = idev->mc_list; ifmca;
4167 ifmca = ifmca->next, ip_idx++) {
4168 if (ip_idx < s_ip_idx)
4169 continue;
4170 err = inet6_fill_ifmcaddr(skb, ifmca,
4171 NETLINK_CB(cb->skb).portid,
4172 cb->nlh->nlmsg_seq,
4173 RTM_GETMULTICAST,
4174 NLM_F_MULTI);
4175 if (err < 0)
4176 break;
4177 }
4178 break;
4179 case ANYCAST_ADDR:
4180 /* anycast address */
4181 for (ifaca = idev->ac_list; ifaca;
4182 ifaca = ifaca->aca_next, ip_idx++) {
4183 if (ip_idx < s_ip_idx)
4184 continue;
4185 err = inet6_fill_ifacaddr(skb, ifaca,
4186 NETLINK_CB(cb->skb).portid,
4187 cb->nlh->nlmsg_seq,
4188 RTM_GETANYCAST,
4189 NLM_F_MULTI);
4190 if (err < 0)
4191 break;
4192 }
4193 break;
4194 default:
4195 break;
4196 }
4197 read_unlock_bh(&idev->lock);
4198 *p_ip_idx = ip_idx;
4199 return err;
4200 }
4201
4202 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4203 enum addr_type_t type)
4204 {
4205 struct net *net = sock_net(skb->sk);
4206 int h, s_h;
4207 int idx, ip_idx;
4208 int s_idx, s_ip_idx;
4209 struct net_device *dev;
4210 struct inet6_dev *idev;
4211 struct hlist_head *head;
4212
4213 s_h = cb->args[0];
4214 s_idx = idx = cb->args[1];
4215 s_ip_idx = ip_idx = cb->args[2];
4216
4217 rcu_read_lock();
4218 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4219 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4220 idx = 0;
4221 head = &net->dev_index_head[h];
4222 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4223 if (idx < s_idx)
4224 goto cont;
4225 if (h > s_h || idx > s_idx)
4226 s_ip_idx = 0;
4227 ip_idx = 0;
4228 idev = __in6_dev_get(dev);
4229 if (!idev)
4230 goto cont;
4231
4232 if (in6_dump_addrs(idev, skb, cb, type,
4233 s_ip_idx, &ip_idx) < 0)
4234 goto done;
4235 cont:
4236 idx++;
4237 }
4238 }
4239 done:
4240 rcu_read_unlock();
4241 cb->args[0] = h;
4242 cb->args[1] = idx;
4243 cb->args[2] = ip_idx;
4244
4245 return skb->len;
4246 }
4247
4248 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4249 {
4250 enum addr_type_t type = UNICAST_ADDR;
4251
4252 return inet6_dump_addr(skb, cb, type);
4253 }
4254
4255 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4256 {
4257 enum addr_type_t type = MULTICAST_ADDR;
4258
4259 return inet6_dump_addr(skb, cb, type);
4260 }
4261
4262
4263 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4264 {
4265 enum addr_type_t type = ANYCAST_ADDR;
4266
4267 return inet6_dump_addr(skb, cb, type);
4268 }
4269
4270 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4271 {
4272 struct net *net = sock_net(in_skb->sk);
4273 struct ifaddrmsg *ifm;
4274 struct nlattr *tb[IFA_MAX+1];
4275 struct in6_addr *addr = NULL, *peer;
4276 struct net_device *dev = NULL;
4277 struct inet6_ifaddr *ifa;
4278 struct sk_buff *skb;
4279 int err;
4280
4281 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4282 if (err < 0)
4283 goto errout;
4284
4285 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4286 if (addr == NULL) {
4287 err = -EINVAL;
4288 goto errout;
4289 }
4290
4291 ifm = nlmsg_data(nlh);
4292 if (ifm->ifa_index)
4293 dev = __dev_get_by_index(net, ifm->ifa_index);
4294
4295 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4296 if (!ifa) {
4297 err = -EADDRNOTAVAIL;
4298 goto errout;
4299 }
4300
4301 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4302 if (!skb) {
4303 err = -ENOBUFS;
4304 goto errout_ifa;
4305 }
4306
4307 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4308 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4309 if (err < 0) {
4310 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4311 WARN_ON(err == -EMSGSIZE);
4312 kfree_skb(skb);
4313 goto errout_ifa;
4314 }
4315 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4316 errout_ifa:
4317 in6_ifa_put(ifa);
4318 errout:
4319 return err;
4320 }
4321
4322 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4323 {
4324 struct sk_buff *skb;
4325 struct net *net = dev_net(ifa->idev->dev);
4326 int err = -ENOBUFS;
4327
4328 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4329 if (skb == NULL)
4330 goto errout;
4331
4332 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4333 if (err < 0) {
4334 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4335 WARN_ON(err == -EMSGSIZE);
4336 kfree_skb(skb);
4337 goto errout;
4338 }
4339 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4340 return;
4341 errout:
4342 if (err < 0)
4343 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4344 }
4345
4346 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4347 __s32 *array, int bytes)
4348 {
4349 BUG_ON(bytes < (DEVCONF_MAX * 4));
4350
4351 memset(array, 0, bytes);
4352 array[DEVCONF_FORWARDING] = cnf->forwarding;
4353 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4354 array[DEVCONF_MTU6] = cnf->mtu6;
4355 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4356 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4357 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4358 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4359 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4360 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4361 jiffies_to_msecs(cnf->rtr_solicit_interval);
4362 array[DEVCONF_RTR_SOLICIT_DELAY] =
4363 jiffies_to_msecs(cnf->rtr_solicit_delay);
4364 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4365 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4366 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4367 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4368 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4369 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4370 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4371 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4372 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4373 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4374 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4375 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4376 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4377 #ifdef CONFIG_IPV6_ROUTER_PREF
4378 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4379 array[DEVCONF_RTR_PROBE_INTERVAL] =
4380 jiffies_to_msecs(cnf->rtr_probe_interval);
4381 #ifdef CONFIG_IPV6_ROUTE_INFO
4382 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4383 #endif
4384 #endif
4385 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4386 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4387 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4388 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4389 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4390 #endif
4391 #ifdef CONFIG_IPV6_MROUTE
4392 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4393 #endif
4394 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4395 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4396 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4397 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4398 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4399 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4400 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4401 }
4402
4403 static inline size_t inet6_ifla6_size(void)
4404 {
4405 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4406 + nla_total_size(sizeof(struct ifla_cacheinfo))
4407 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4408 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4409 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4410 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4411 }
4412
4413 static inline size_t inet6_if_nlmsg_size(void)
4414 {
4415 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4416 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4417 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4418 + nla_total_size(4) /* IFLA_MTU */
4419 + nla_total_size(4) /* IFLA_LINK */
4420 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4421 }
4422
4423 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4424 int items, int bytes)
4425 {
4426 int i;
4427 int pad = bytes - sizeof(u64) * items;
4428 BUG_ON(pad < 0);
4429
4430 /* Use put_unaligned() because stats may not be aligned for u64. */
4431 put_unaligned(items, &stats[0]);
4432 for (i = 1; i < items; i++)
4433 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4434
4435 memset(&stats[items], 0, pad);
4436 }
4437
4438 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4439 int items, int bytes, size_t syncpoff)
4440 {
4441 int i;
4442 int pad = bytes - sizeof(u64) * items;
4443 BUG_ON(pad < 0);
4444
4445 /* Use put_unaligned() because stats may not be aligned for u64. */
4446 put_unaligned(items, &stats[0]);
4447 for (i = 1; i < items; i++)
4448 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4449
4450 memset(&stats[items], 0, pad);
4451 }
4452
4453 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4454 int bytes)
4455 {
4456 switch (attrtype) {
4457 case IFLA_INET6_STATS:
4458 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4459 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4460 break;
4461 case IFLA_INET6_ICMP6STATS:
4462 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4463 break;
4464 }
4465 }
4466
4467 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4468 {
4469 struct nlattr *nla;
4470 struct ifla_cacheinfo ci;
4471
4472 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4473 goto nla_put_failure;
4474 ci.max_reasm_len = IPV6_MAXPLEN;
4475 ci.tstamp = cstamp_delta(idev->tstamp);
4476 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4477 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4478 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4479 goto nla_put_failure;
4480 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4481 if (nla == NULL)
4482 goto nla_put_failure;
4483 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4484
4485 /* XXX - MC not implemented */
4486
4487 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4488 if (nla == NULL)
4489 goto nla_put_failure;
4490 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4491
4492 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4493 if (nla == NULL)
4494 goto nla_put_failure;
4495 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4496
4497 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4498 if (nla == NULL)
4499 goto nla_put_failure;
4500
4501 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4502 goto nla_put_failure;
4503
4504 read_lock_bh(&idev->lock);
4505 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4506 read_unlock_bh(&idev->lock);
4507
4508 return 0;
4509
4510 nla_put_failure:
4511 return -EMSGSIZE;
4512 }
4513
4514 static size_t inet6_get_link_af_size(const struct net_device *dev)
4515 {
4516 if (!__in6_dev_get(dev))
4517 return 0;
4518
4519 return inet6_ifla6_size();
4520 }
4521
4522 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4523 {
4524 struct inet6_dev *idev = __in6_dev_get(dev);
4525
4526 if (!idev)
4527 return -ENODATA;
4528
4529 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4530 return -EMSGSIZE;
4531
4532 return 0;
4533 }
4534
4535 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4536 {
4537 struct inet6_ifaddr *ifp;
4538 struct net_device *dev = idev->dev;
4539 bool update_rs = false;
4540 struct in6_addr ll_addr;
4541
4542 ASSERT_RTNL();
4543
4544 if (token == NULL)
4545 return -EINVAL;
4546 if (ipv6_addr_any(token))
4547 return -EINVAL;
4548 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4549 return -EINVAL;
4550 if (!ipv6_accept_ra(idev))
4551 return -EINVAL;
4552 if (idev->cnf.rtr_solicits <= 0)
4553 return -EINVAL;
4554
4555 write_lock_bh(&idev->lock);
4556
4557 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4558 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4559
4560 write_unlock_bh(&idev->lock);
4561
4562 if (!idev->dead && (idev->if_flags & IF_READY) &&
4563 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4564 IFA_F_OPTIMISTIC)) {
4565
4566 /* If we're not ready, then normal ifup will take care
4567 * of this. Otherwise, we need to request our rs here.
4568 */
4569 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4570 update_rs = true;
4571 }
4572
4573 write_lock_bh(&idev->lock);
4574
4575 if (update_rs) {
4576 idev->if_flags |= IF_RS_SENT;
4577 idev->rs_probes = 1;
4578 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4579 }
4580
4581 /* Well, that's kinda nasty ... */
4582 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4583 spin_lock(&ifp->lock);
4584 if (ifp->tokenized) {
4585 ifp->valid_lft = 0;
4586 ifp->prefered_lft = 0;
4587 }
4588 spin_unlock(&ifp->lock);
4589 }
4590
4591 write_unlock_bh(&idev->lock);
4592 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4593 addrconf_verify_rtnl();
4594 return 0;
4595 }
4596
4597 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4598 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
4599 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
4600 };
4601
4602 static int inet6_validate_link_af(const struct net_device *dev,
4603 const struct nlattr *nla)
4604 {
4605 struct nlattr *tb[IFLA_INET6_MAX + 1];
4606
4607 if (dev && !__in6_dev_get(dev))
4608 return -EAFNOSUPPORT;
4609
4610 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4611 }
4612
4613 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4614 {
4615 int err = -EINVAL;
4616 struct inet6_dev *idev = __in6_dev_get(dev);
4617 struct nlattr *tb[IFLA_INET6_MAX + 1];
4618
4619 if (!idev)
4620 return -EAFNOSUPPORT;
4621
4622 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4623 BUG();
4624
4625 if (tb[IFLA_INET6_TOKEN]) {
4626 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4627 if (err)
4628 return err;
4629 }
4630
4631 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4632 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4633
4634 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4635 mode != IN6_ADDR_GEN_MODE_NONE)
4636 return -EINVAL;
4637 idev->addr_gen_mode = mode;
4638 err = 0;
4639 }
4640
4641 return err;
4642 }
4643
4644 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4645 u32 portid, u32 seq, int event, unsigned int flags)
4646 {
4647 struct net_device *dev = idev->dev;
4648 struct ifinfomsg *hdr;
4649 struct nlmsghdr *nlh;
4650 void *protoinfo;
4651
4652 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4653 if (nlh == NULL)
4654 return -EMSGSIZE;
4655
4656 hdr = nlmsg_data(nlh);
4657 hdr->ifi_family = AF_INET6;
4658 hdr->__ifi_pad = 0;
4659 hdr->ifi_type = dev->type;
4660 hdr->ifi_index = dev->ifindex;
4661 hdr->ifi_flags = dev_get_flags(dev);
4662 hdr->ifi_change = 0;
4663
4664 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4665 (dev->addr_len &&
4666 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4667 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4668 (dev->ifindex != dev->iflink &&
4669 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4670 goto nla_put_failure;
4671 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4672 if (protoinfo == NULL)
4673 goto nla_put_failure;
4674
4675 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4676 goto nla_put_failure;
4677
4678 nla_nest_end(skb, protoinfo);
4679 nlmsg_end(skb, nlh);
4680 return 0;
4681
4682 nla_put_failure:
4683 nlmsg_cancel(skb, nlh);
4684 return -EMSGSIZE;
4685 }
4686
4687 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4688 {
4689 struct net *net = sock_net(skb->sk);
4690 int h, s_h;
4691 int idx = 0, s_idx;
4692 struct net_device *dev;
4693 struct inet6_dev *idev;
4694 struct hlist_head *head;
4695
4696 s_h = cb->args[0];
4697 s_idx = cb->args[1];
4698
4699 rcu_read_lock();
4700 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4701 idx = 0;
4702 head = &net->dev_index_head[h];
4703 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4704 if (idx < s_idx)
4705 goto cont;
4706 idev = __in6_dev_get(dev);
4707 if (!idev)
4708 goto cont;
4709 if (inet6_fill_ifinfo(skb, idev,
4710 NETLINK_CB(cb->skb).portid,
4711 cb->nlh->nlmsg_seq,
4712 RTM_NEWLINK, NLM_F_MULTI) < 0)
4713 goto out;
4714 cont:
4715 idx++;
4716 }
4717 }
4718 out:
4719 rcu_read_unlock();
4720 cb->args[1] = idx;
4721 cb->args[0] = h;
4722
4723 return skb->len;
4724 }
4725
4726 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4727 {
4728 struct sk_buff *skb;
4729 struct net *net = dev_net(idev->dev);
4730 int err = -ENOBUFS;
4731
4732 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4733 if (skb == NULL)
4734 goto errout;
4735
4736 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4737 if (err < 0) {
4738 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4739 WARN_ON(err == -EMSGSIZE);
4740 kfree_skb(skb);
4741 goto errout;
4742 }
4743 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4744 return;
4745 errout:
4746 if (err < 0)
4747 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4748 }
4749
4750 static inline size_t inet6_prefix_nlmsg_size(void)
4751 {
4752 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4753 + nla_total_size(sizeof(struct in6_addr))
4754 + nla_total_size(sizeof(struct prefix_cacheinfo));
4755 }
4756
4757 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4758 struct prefix_info *pinfo, u32 portid, u32 seq,
4759 int event, unsigned int flags)
4760 {
4761 struct prefixmsg *pmsg;
4762 struct nlmsghdr *nlh;
4763 struct prefix_cacheinfo ci;
4764
4765 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4766 if (nlh == NULL)
4767 return -EMSGSIZE;
4768
4769 pmsg = nlmsg_data(nlh);
4770 pmsg->prefix_family = AF_INET6;
4771 pmsg->prefix_pad1 = 0;
4772 pmsg->prefix_pad2 = 0;
4773 pmsg->prefix_ifindex = idev->dev->ifindex;
4774 pmsg->prefix_len = pinfo->prefix_len;
4775 pmsg->prefix_type = pinfo->type;
4776 pmsg->prefix_pad3 = 0;
4777 pmsg->prefix_flags = 0;
4778 if (pinfo->onlink)
4779 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4780 if (pinfo->autoconf)
4781 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4782
4783 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4784 goto nla_put_failure;
4785 ci.preferred_time = ntohl(pinfo->prefered);
4786 ci.valid_time = ntohl(pinfo->valid);
4787 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4788 goto nla_put_failure;
4789 nlmsg_end(skb, nlh);
4790 return 0;
4791
4792 nla_put_failure:
4793 nlmsg_cancel(skb, nlh);
4794 return -EMSGSIZE;
4795 }
4796
4797 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4798 struct prefix_info *pinfo)
4799 {
4800 struct sk_buff *skb;
4801 struct net *net = dev_net(idev->dev);
4802 int err = -ENOBUFS;
4803
4804 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4805 if (skb == NULL)
4806 goto errout;
4807
4808 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4809 if (err < 0) {
4810 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4811 WARN_ON(err == -EMSGSIZE);
4812 kfree_skb(skb);
4813 goto errout;
4814 }
4815 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4816 return;
4817 errout:
4818 if (err < 0)
4819 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4820 }
4821
4822 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4823 {
4824 struct net *net = dev_net(ifp->idev->dev);
4825
4826 if (event)
4827 ASSERT_RTNL();
4828
4829 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4830
4831 switch (event) {
4832 case RTM_NEWADDR:
4833 /*
4834 * If the address was optimistic
4835 * we inserted the route at the start of
4836 * our DAD process, so we don't need
4837 * to do it again
4838 */
4839 if (!(ifp->rt->rt6i_node))
4840 ip6_ins_rt(ifp->rt);
4841 if (ifp->idev->cnf.forwarding)
4842 addrconf_join_anycast(ifp);
4843 if (!ipv6_addr_any(&ifp->peer_addr))
4844 addrconf_prefix_route(&ifp->peer_addr, 128,
4845 ifp->idev->dev, 0, 0);
4846 break;
4847 case RTM_DELADDR:
4848 if (ifp->idev->cnf.forwarding)
4849 addrconf_leave_anycast(ifp);
4850 addrconf_leave_solict(ifp->idev, &ifp->addr);
4851 if (!ipv6_addr_any(&ifp->peer_addr)) {
4852 struct rt6_info *rt;
4853
4854 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
4855 ifp->idev->dev, 0, 0);
4856 if (rt && ip6_del_rt(rt))
4857 dst_free(&rt->dst);
4858 }
4859 dst_hold(&ifp->rt->dst);
4860
4861 if (ip6_del_rt(ifp->rt))
4862 dst_free(&ifp->rt->dst);
4863
4864 rt_genid_bump_ipv6(net);
4865 break;
4866 }
4867 atomic_inc(&net->ipv6.dev_addr_genid);
4868 }
4869
4870 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4871 {
4872 rcu_read_lock_bh();
4873 if (likely(ifp->idev->dead == 0))
4874 __ipv6_ifa_notify(event, ifp);
4875 rcu_read_unlock_bh();
4876 }
4877
4878 #ifdef CONFIG_SYSCTL
4879
4880 static
4881 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4882 void __user *buffer, size_t *lenp, loff_t *ppos)
4883 {
4884 int *valp = ctl->data;
4885 int val = *valp;
4886 loff_t pos = *ppos;
4887 struct ctl_table lctl;
4888 int ret;
4889
4890 /*
4891 * ctl->data points to idev->cnf.forwarding, we should
4892 * not modify it until we get the rtnl lock.
4893 */
4894 lctl = *ctl;
4895 lctl.data = &val;
4896
4897 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4898
4899 if (write)
4900 ret = addrconf_fixup_forwarding(ctl, valp, val);
4901 if (ret)
4902 *ppos = pos;
4903 return ret;
4904 }
4905
4906 static void dev_disable_change(struct inet6_dev *idev)
4907 {
4908 struct netdev_notifier_info info;
4909
4910 if (!idev || !idev->dev)
4911 return;
4912
4913 netdev_notifier_info_init(&info, idev->dev);
4914 if (idev->cnf.disable_ipv6)
4915 addrconf_notify(NULL, NETDEV_DOWN, &info);
4916 else
4917 addrconf_notify(NULL, NETDEV_UP, &info);
4918 }
4919
4920 static void addrconf_disable_change(struct net *net, __s32 newf)
4921 {
4922 struct net_device *dev;
4923 struct inet6_dev *idev;
4924
4925 rcu_read_lock();
4926 for_each_netdev_rcu(net, dev) {
4927 idev = __in6_dev_get(dev);
4928 if (idev) {
4929 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4930 idev->cnf.disable_ipv6 = newf;
4931 if (changed)
4932 dev_disable_change(idev);
4933 }
4934 }
4935 rcu_read_unlock();
4936 }
4937
4938 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4939 {
4940 struct net *net;
4941 int old;
4942
4943 if (!rtnl_trylock())
4944 return restart_syscall();
4945
4946 net = (struct net *)table->extra2;
4947 old = *p;
4948 *p = newf;
4949
4950 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4951 rtnl_unlock();
4952 return 0;
4953 }
4954
4955 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4956 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4957 addrconf_disable_change(net, newf);
4958 } else if ((!newf) ^ (!old))
4959 dev_disable_change((struct inet6_dev *)table->extra1);
4960
4961 rtnl_unlock();
4962 return 0;
4963 }
4964
4965 static
4966 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4967 void __user *buffer, size_t *lenp, loff_t *ppos)
4968 {
4969 int *valp = ctl->data;
4970 int val = *valp;
4971 loff_t pos = *ppos;
4972 struct ctl_table lctl;
4973 int ret;
4974
4975 /*
4976 * ctl->data points to idev->cnf.disable_ipv6, we should
4977 * not modify it until we get the rtnl lock.
4978 */
4979 lctl = *ctl;
4980 lctl.data = &val;
4981
4982 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4983
4984 if (write)
4985 ret = addrconf_disable_ipv6(ctl, valp, val);
4986 if (ret)
4987 *ppos = pos;
4988 return ret;
4989 }
4990
4991 static
4992 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4993 void __user *buffer, size_t *lenp, loff_t *ppos)
4994 {
4995 int *valp = ctl->data;
4996 int ret;
4997 int old, new;
4998
4999 old = *valp;
5000 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5001 new = *valp;
5002
5003 if (write && old != new) {
5004 struct net *net = ctl->extra2;
5005
5006 if (!rtnl_trylock())
5007 return restart_syscall();
5008
5009 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5010 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5011 NETCONFA_IFINDEX_DEFAULT,
5012 net->ipv6.devconf_dflt);
5013 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5014 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5015 NETCONFA_IFINDEX_ALL,
5016 net->ipv6.devconf_all);
5017 else {
5018 struct inet6_dev *idev = ctl->extra1;
5019
5020 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5021 idev->dev->ifindex,
5022 &idev->cnf);
5023 }
5024 rtnl_unlock();
5025 }
5026
5027 return ret;
5028 }
5029
5030
5031 static struct addrconf_sysctl_table
5032 {
5033 struct ctl_table_header *sysctl_header;
5034 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5035 } addrconf_sysctl __read_mostly = {
5036 .sysctl_header = NULL,
5037 .addrconf_vars = {
5038 {
5039 .procname = "forwarding",
5040 .data = &ipv6_devconf.forwarding,
5041 .maxlen = sizeof(int),
5042 .mode = 0644,
5043 .proc_handler = addrconf_sysctl_forward,
5044 },
5045 {
5046 .procname = "hop_limit",
5047 .data = &ipv6_devconf.hop_limit,
5048 .maxlen = sizeof(int),
5049 .mode = 0644,
5050 .proc_handler = proc_dointvec,
5051 },
5052 {
5053 .procname = "mtu",
5054 .data = &ipv6_devconf.mtu6,
5055 .maxlen = sizeof(int),
5056 .mode = 0644,
5057 .proc_handler = proc_dointvec,
5058 },
5059 {
5060 .procname = "accept_ra",
5061 .data = &ipv6_devconf.accept_ra,
5062 .maxlen = sizeof(int),
5063 .mode = 0644,
5064 .proc_handler = proc_dointvec,
5065 },
5066 {
5067 .procname = "accept_redirects",
5068 .data = &ipv6_devconf.accept_redirects,
5069 .maxlen = sizeof(int),
5070 .mode = 0644,
5071 .proc_handler = proc_dointvec,
5072 },
5073 {
5074 .procname = "autoconf",
5075 .data = &ipv6_devconf.autoconf,
5076 .maxlen = sizeof(int),
5077 .mode = 0644,
5078 .proc_handler = proc_dointvec,
5079 },
5080 {
5081 .procname = "dad_transmits",
5082 .data = &ipv6_devconf.dad_transmits,
5083 .maxlen = sizeof(int),
5084 .mode = 0644,
5085 .proc_handler = proc_dointvec,
5086 },
5087 {
5088 .procname = "router_solicitations",
5089 .data = &ipv6_devconf.rtr_solicits,
5090 .maxlen = sizeof(int),
5091 .mode = 0644,
5092 .proc_handler = proc_dointvec,
5093 },
5094 {
5095 .procname = "router_solicitation_interval",
5096 .data = &ipv6_devconf.rtr_solicit_interval,
5097 .maxlen = sizeof(int),
5098 .mode = 0644,
5099 .proc_handler = proc_dointvec_jiffies,
5100 },
5101 {
5102 .procname = "router_solicitation_delay",
5103 .data = &ipv6_devconf.rtr_solicit_delay,
5104 .maxlen = sizeof(int),
5105 .mode = 0644,
5106 .proc_handler = proc_dointvec_jiffies,
5107 },
5108 {
5109 .procname = "force_mld_version",
5110 .data = &ipv6_devconf.force_mld_version,
5111 .maxlen = sizeof(int),
5112 .mode = 0644,
5113 .proc_handler = proc_dointvec,
5114 },
5115 {
5116 .procname = "mldv1_unsolicited_report_interval",
5117 .data =
5118 &ipv6_devconf.mldv1_unsolicited_report_interval,
5119 .maxlen = sizeof(int),
5120 .mode = 0644,
5121 .proc_handler = proc_dointvec_ms_jiffies,
5122 },
5123 {
5124 .procname = "mldv2_unsolicited_report_interval",
5125 .data =
5126 &ipv6_devconf.mldv2_unsolicited_report_interval,
5127 .maxlen = sizeof(int),
5128 .mode = 0644,
5129 .proc_handler = proc_dointvec_ms_jiffies,
5130 },
5131 {
5132 .procname = "use_tempaddr",
5133 .data = &ipv6_devconf.use_tempaddr,
5134 .maxlen = sizeof(int),
5135 .mode = 0644,
5136 .proc_handler = proc_dointvec,
5137 },
5138 {
5139 .procname = "temp_valid_lft",
5140 .data = &ipv6_devconf.temp_valid_lft,
5141 .maxlen = sizeof(int),
5142 .mode = 0644,
5143 .proc_handler = proc_dointvec,
5144 },
5145 {
5146 .procname = "temp_prefered_lft",
5147 .data = &ipv6_devconf.temp_prefered_lft,
5148 .maxlen = sizeof(int),
5149 .mode = 0644,
5150 .proc_handler = proc_dointvec,
5151 },
5152 {
5153 .procname = "regen_max_retry",
5154 .data = &ipv6_devconf.regen_max_retry,
5155 .maxlen = sizeof(int),
5156 .mode = 0644,
5157 .proc_handler = proc_dointvec,
5158 },
5159 {
5160 .procname = "max_desync_factor",
5161 .data = &ipv6_devconf.max_desync_factor,
5162 .maxlen = sizeof(int),
5163 .mode = 0644,
5164 .proc_handler = proc_dointvec,
5165 },
5166 {
5167 .procname = "max_addresses",
5168 .data = &ipv6_devconf.max_addresses,
5169 .maxlen = sizeof(int),
5170 .mode = 0644,
5171 .proc_handler = proc_dointvec,
5172 },
5173 {
5174 .procname = "accept_ra_defrtr",
5175 .data = &ipv6_devconf.accept_ra_defrtr,
5176 .maxlen = sizeof(int),
5177 .mode = 0644,
5178 .proc_handler = proc_dointvec,
5179 },
5180 {
5181 .procname = "accept_ra_pinfo",
5182 .data = &ipv6_devconf.accept_ra_pinfo,
5183 .maxlen = sizeof(int),
5184 .mode = 0644,
5185 .proc_handler = proc_dointvec,
5186 },
5187 #ifdef CONFIG_IPV6_ROUTER_PREF
5188 {
5189 .procname = "accept_ra_rtr_pref",
5190 .data = &ipv6_devconf.accept_ra_rtr_pref,
5191 .maxlen = sizeof(int),
5192 .mode = 0644,
5193 .proc_handler = proc_dointvec,
5194 },
5195 {
5196 .procname = "router_probe_interval",
5197 .data = &ipv6_devconf.rtr_probe_interval,
5198 .maxlen = sizeof(int),
5199 .mode = 0644,
5200 .proc_handler = proc_dointvec_jiffies,
5201 },
5202 #ifdef CONFIG_IPV6_ROUTE_INFO
5203 {
5204 .procname = "accept_ra_rt_info_max_plen",
5205 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5206 .maxlen = sizeof(int),
5207 .mode = 0644,
5208 .proc_handler = proc_dointvec,
5209 },
5210 #endif
5211 #endif
5212 {
5213 .procname = "proxy_ndp",
5214 .data = &ipv6_devconf.proxy_ndp,
5215 .maxlen = sizeof(int),
5216 .mode = 0644,
5217 .proc_handler = addrconf_sysctl_proxy_ndp,
5218 },
5219 {
5220 .procname = "accept_source_route",
5221 .data = &ipv6_devconf.accept_source_route,
5222 .maxlen = sizeof(int),
5223 .mode = 0644,
5224 .proc_handler = proc_dointvec,
5225 },
5226 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5227 {
5228 .procname = "optimistic_dad",
5229 .data = &ipv6_devconf.optimistic_dad,
5230 .maxlen = sizeof(int),
5231 .mode = 0644,
5232 .proc_handler = proc_dointvec,
5233
5234 },
5235 {
5236 .procname = "use_optimistic",
5237 .data = &ipv6_devconf.use_optimistic,
5238 .maxlen = sizeof(int),
5239 .mode = 0644,
5240 .proc_handler = proc_dointvec,
5241
5242 },
5243 #endif
5244 #ifdef CONFIG_IPV6_MROUTE
5245 {
5246 .procname = "mc_forwarding",
5247 .data = &ipv6_devconf.mc_forwarding,
5248 .maxlen = sizeof(int),
5249 .mode = 0444,
5250 .proc_handler = proc_dointvec,
5251 },
5252 #endif
5253 {
5254 .procname = "disable_ipv6",
5255 .data = &ipv6_devconf.disable_ipv6,
5256 .maxlen = sizeof(int),
5257 .mode = 0644,
5258 .proc_handler = addrconf_sysctl_disable,
5259 },
5260 {
5261 .procname = "accept_dad",
5262 .data = &ipv6_devconf.accept_dad,
5263 .maxlen = sizeof(int),
5264 .mode = 0644,
5265 .proc_handler = proc_dointvec,
5266 },
5267 {
5268 .procname = "force_tllao",
5269 .data = &ipv6_devconf.force_tllao,
5270 .maxlen = sizeof(int),
5271 .mode = 0644,
5272 .proc_handler = proc_dointvec
5273 },
5274 {
5275 .procname = "ndisc_notify",
5276 .data = &ipv6_devconf.ndisc_notify,
5277 .maxlen = sizeof(int),
5278 .mode = 0644,
5279 .proc_handler = proc_dointvec
5280 },
5281 {
5282 .procname = "suppress_frag_ndisc",
5283 .data = &ipv6_devconf.suppress_frag_ndisc,
5284 .maxlen = sizeof(int),
5285 .mode = 0644,
5286 .proc_handler = proc_dointvec
5287 },
5288 {
5289 .procname = "accept_ra_from_local",
5290 .data = &ipv6_devconf.accept_ra_from_local,
5291 .maxlen = sizeof(int),
5292 .mode = 0644,
5293 .proc_handler = proc_dointvec,
5294 },
5295 {
5296 .procname = "accept_ra_mtu",
5297 .data = &ipv6_devconf.accept_ra_mtu,
5298 .maxlen = sizeof(int),
5299 .mode = 0644,
5300 .proc_handler = proc_dointvec,
5301 },
5302 {
5303 /* sentinel */
5304 }
5305 },
5306 };
5307
5308 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5309 struct inet6_dev *idev, struct ipv6_devconf *p)
5310 {
5311 int i;
5312 struct addrconf_sysctl_table *t;
5313 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5314
5315 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5316 if (t == NULL)
5317 goto out;
5318
5319 for (i = 0; t->addrconf_vars[i].data; i++) {
5320 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5321 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5322 t->addrconf_vars[i].extra2 = net;
5323 }
5324
5325 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5326
5327 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5328 if (t->sysctl_header == NULL)
5329 goto free;
5330
5331 p->sysctl = t;
5332 return 0;
5333
5334 free:
5335 kfree(t);
5336 out:
5337 return -ENOBUFS;
5338 }
5339
5340 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5341 {
5342 struct addrconf_sysctl_table *t;
5343
5344 if (p->sysctl == NULL)
5345 return;
5346
5347 t = p->sysctl;
5348 p->sysctl = NULL;
5349 unregister_net_sysctl_table(t->sysctl_header);
5350 kfree(t);
5351 }
5352
5353 static int addrconf_sysctl_register(struct inet6_dev *idev)
5354 {
5355 int err;
5356
5357 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5358 return -EINVAL;
5359
5360 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5361 &ndisc_ifinfo_sysctl_change);
5362 if (err)
5363 return err;
5364 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5365 idev, &idev->cnf);
5366 if (err)
5367 neigh_sysctl_unregister(idev->nd_parms);
5368
5369 return err;
5370 }
5371
5372 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5373 {
5374 __addrconf_sysctl_unregister(&idev->cnf);
5375 neigh_sysctl_unregister(idev->nd_parms);
5376 }
5377
5378
5379 #endif
5380
5381 static int __net_init addrconf_init_net(struct net *net)
5382 {
5383 int err = -ENOMEM;
5384 struct ipv6_devconf *all, *dflt;
5385
5386 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5387 if (all == NULL)
5388 goto err_alloc_all;
5389
5390 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5391 if (dflt == NULL)
5392 goto err_alloc_dflt;
5393
5394 /* these will be inherited by all namespaces */
5395 dflt->autoconf = ipv6_defaults.autoconf;
5396 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5397
5398 net->ipv6.devconf_all = all;
5399 net->ipv6.devconf_dflt = dflt;
5400
5401 #ifdef CONFIG_SYSCTL
5402 err = __addrconf_sysctl_register(net, "all", NULL, all);
5403 if (err < 0)
5404 goto err_reg_all;
5405
5406 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5407 if (err < 0)
5408 goto err_reg_dflt;
5409 #endif
5410 return 0;
5411
5412 #ifdef CONFIG_SYSCTL
5413 err_reg_dflt:
5414 __addrconf_sysctl_unregister(all);
5415 err_reg_all:
5416 kfree(dflt);
5417 #endif
5418 err_alloc_dflt:
5419 kfree(all);
5420 err_alloc_all:
5421 return err;
5422 }
5423
5424 static void __net_exit addrconf_exit_net(struct net *net)
5425 {
5426 #ifdef CONFIG_SYSCTL
5427 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5428 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5429 #endif
5430 kfree(net->ipv6.devconf_dflt);
5431 kfree(net->ipv6.devconf_all);
5432 }
5433
5434 static struct pernet_operations addrconf_ops = {
5435 .init = addrconf_init_net,
5436 .exit = addrconf_exit_net,
5437 };
5438
5439 static struct rtnl_af_ops inet6_ops __read_mostly = {
5440 .family = AF_INET6,
5441 .fill_link_af = inet6_fill_link_af,
5442 .get_link_af_size = inet6_get_link_af_size,
5443 .validate_link_af = inet6_validate_link_af,
5444 .set_link_af = inet6_set_link_af,
5445 };
5446
5447 /*
5448 * Init / cleanup code
5449 */
5450
5451 int __init addrconf_init(void)
5452 {
5453 struct inet6_dev *idev;
5454 int i, err;
5455
5456 err = ipv6_addr_label_init();
5457 if (err < 0) {
5458 pr_crit("%s: cannot initialize default policy table: %d\n",
5459 __func__, err);
5460 goto out;
5461 }
5462
5463 err = register_pernet_subsys(&addrconf_ops);
5464 if (err < 0)
5465 goto out_addrlabel;
5466
5467 addrconf_wq = create_workqueue("ipv6_addrconf");
5468 if (!addrconf_wq) {
5469 err = -ENOMEM;
5470 goto out_nowq;
5471 }
5472
5473 /* The addrconf netdev notifier requires that loopback_dev
5474 * has it's ipv6 private information allocated and setup
5475 * before it can bring up and give link-local addresses
5476 * to other devices which are up.
5477 *
5478 * Unfortunately, loopback_dev is not necessarily the first
5479 * entry in the global dev_base list of net devices. In fact,
5480 * it is likely to be the very last entry on that list.
5481 * So this causes the notifier registry below to try and
5482 * give link-local addresses to all devices besides loopback_dev
5483 * first, then loopback_dev, which cases all the non-loopback_dev
5484 * devices to fail to get a link-local address.
5485 *
5486 * So, as a temporary fix, allocate the ipv6 structure for
5487 * loopback_dev first by hand.
5488 * Longer term, all of the dependencies ipv6 has upon the loopback
5489 * device and it being up should be removed.
5490 */
5491 rtnl_lock();
5492 idev = ipv6_add_dev(init_net.loopback_dev);
5493 rtnl_unlock();
5494 if (IS_ERR(idev)) {
5495 err = PTR_ERR(idev);
5496 goto errlo;
5497 }
5498
5499 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5500 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5501
5502 register_netdevice_notifier(&ipv6_dev_notf);
5503
5504 addrconf_verify();
5505
5506 rtnl_af_register(&inet6_ops);
5507
5508 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5509 NULL);
5510 if (err < 0)
5511 goto errout;
5512
5513 /* Only the first call to __rtnl_register can fail */
5514 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5515 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5516 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5517 inet6_dump_ifaddr, NULL);
5518 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5519 inet6_dump_ifmcaddr, NULL);
5520 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5521 inet6_dump_ifacaddr, NULL);
5522 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5523 inet6_netconf_dump_devconf, NULL);
5524
5525 ipv6_addr_label_rtnl_register();
5526
5527 return 0;
5528 errout:
5529 rtnl_af_unregister(&inet6_ops);
5530 unregister_netdevice_notifier(&ipv6_dev_notf);
5531 errlo:
5532 destroy_workqueue(addrconf_wq);
5533 out_nowq:
5534 unregister_pernet_subsys(&addrconf_ops);
5535 out_addrlabel:
5536 ipv6_addr_label_cleanup();
5537 out:
5538 return err;
5539 }
5540
5541 void addrconf_cleanup(void)
5542 {
5543 struct net_device *dev;
5544 int i;
5545
5546 unregister_netdevice_notifier(&ipv6_dev_notf);
5547 unregister_pernet_subsys(&addrconf_ops);
5548 ipv6_addr_label_cleanup();
5549
5550 rtnl_lock();
5551
5552 __rtnl_af_unregister(&inet6_ops);
5553
5554 /* clean dev list */
5555 for_each_netdev(&init_net, dev) {
5556 if (__in6_dev_get(dev) == NULL)
5557 continue;
5558 addrconf_ifdown(dev, 1);
5559 }
5560 addrconf_ifdown(init_net.loopback_dev, 2);
5561
5562 /*
5563 * Check hash table.
5564 */
5565 spin_lock_bh(&addrconf_hash_lock);
5566 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5567 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5568 spin_unlock_bh(&addrconf_hash_lock);
5569 cancel_delayed_work(&addr_chk_work);
5570 rtnl_unlock();
5571
5572 destroy_workqueue(addrconf_wq);
5573 }
This page took 0.155009 seconds and 6 git commands to generate.