?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'devicetree-for-linus' of git://git.secretlab.ca/git/linux
[linux-2.6.git] / net / ipv6 / addrconf.c
blobcd3fb301da38a970cd48302386428a8923f21c91
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
87 #ifdef CONFIG_IPV6_PRIVACY
88 #include <linux/random.h>
89 #endif
90
91 #include <linux/uaccess.h>
92 #include <asm/unaligned.h>
93
94 #include <linux/proc_fs.h>
95 #include <linux/seq_file.h>
96 #include <linux/export.h>
97
98 /* Set to 3 to get tracing... */
99 #define ACONF_DEBUG 2
100
101 #if ACONF_DEBUG >= 3
102 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
103 #else
104 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
105 #endif
106
107 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108
109 static inline u32 cstamp_delta(unsigned long cstamp)
110 {
111 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
112 }
113
114 #ifdef CONFIG_SYSCTL
115 static void addrconf_sysctl_register(struct inet6_dev *idev);
116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
117 #else
118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
119 {
120 }
121
122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
123 {
124 }
125 #endif
126
127 #ifdef CONFIG_IPV6_PRIVACY
128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
130 static void ipv6_regen_rndid(unsigned long data);
131 #endif
132
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135
136 /*
137 * Configured unicast address hash table
138 */
139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
140 static DEFINE_SPINLOCK(addrconf_hash_lock);
141
142 static void addrconf_verify(unsigned long);
143
144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
145 static DEFINE_SPINLOCK(addrconf_verify_lock);
146
147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
149
150 static void addrconf_type_change(struct net_device *dev,
151 unsigned long event);
152 static int addrconf_ifdown(struct net_device *dev, int how);
153
154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
155 int plen,
156 const struct net_device *dev,
157 u32 flags, u32 noflags);
158
159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_timer(unsigned long data);
161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
162 static void addrconf_dad_run(struct inet6_dev *idev);
163 static void addrconf_rs_timer(unsigned long data);
164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
166
167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
168 struct prefix_info *pinfo);
169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
170 struct net_device *dev);
171
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
173 .forwarding = 0,
174 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
175 .mtu6 = IPV6_MIN_MTU,
176 .accept_ra = 1,
177 .accept_redirects = 1,
178 .autoconf = 1,
179 .force_mld_version = 0,
180 .mldv1_unsolicited_report_interval = 10 * HZ,
181 .mldv2_unsolicited_report_interval = HZ,
182 .dad_transmits = 1,
183 .rtr_solicits = MAX_RTR_SOLICITATIONS,
184 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
185 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
186 #ifdef CONFIG_IPV6_PRIVACY
187 .use_tempaddr = 0,
188 .temp_valid_lft = TEMP_VALID_LIFETIME,
189 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
190 .regen_max_retry = REGEN_MAX_RETRY,
191 .max_desync_factor = MAX_DESYNC_FACTOR,
192 #endif
193 .max_addresses = IPV6_MAX_ADDRESSES,
194 .accept_ra_defrtr = 1,
195 .accept_ra_pinfo = 1,
196 #ifdef CONFIG_IPV6_ROUTER_PREF
197 .accept_ra_rtr_pref = 1,
198 .rtr_probe_interval = 60 * HZ,
199 #ifdef CONFIG_IPV6_ROUTE_INFO
200 .accept_ra_rt_info_max_plen = 0,
201 #endif
202 #endif
203 .proxy_ndp = 0,
204 .accept_source_route = 0, /* we do not accept RH0 by default. */
205 .disable_ipv6 = 0,
206 .accept_dad = 1,
207 .suppress_frag_ndisc = 1,
208 };
209
210 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
211 .forwarding = 0,
212 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
213 .mtu6 = IPV6_MIN_MTU,
214 .accept_ra = 1,
215 .accept_redirects = 1,
216 .autoconf = 1,
217 .force_mld_version = 0,
218 .mldv1_unsolicited_report_interval = 10 * HZ,
219 .mldv2_unsolicited_report_interval = HZ,
220 .dad_transmits = 1,
221 .rtr_solicits = MAX_RTR_SOLICITATIONS,
222 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
223 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
224 #ifdef CONFIG_IPV6_PRIVACY
225 .use_tempaddr = 0,
226 .temp_valid_lft = TEMP_VALID_LIFETIME,
227 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
228 .regen_max_retry = REGEN_MAX_RETRY,
229 .max_desync_factor = MAX_DESYNC_FACTOR,
230 #endif
231 .max_addresses = IPV6_MAX_ADDRESSES,
232 .accept_ra_defrtr = 1,
233 .accept_ra_pinfo = 1,
234 #ifdef CONFIG_IPV6_ROUTER_PREF
235 .accept_ra_rtr_pref = 1,
236 .rtr_probe_interval = 60 * HZ,
237 #ifdef CONFIG_IPV6_ROUTE_INFO
238 .accept_ra_rt_info_max_plen = 0,
239 #endif
240 #endif
241 .proxy_ndp = 0,
242 .accept_source_route = 0, /* we do not accept RH0 by default. */
243 .disable_ipv6 = 0,
244 .accept_dad = 1,
245 .suppress_frag_ndisc = 1,
246 };
247
248 /* Check if a valid qdisc is available */
249 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
250 {
251 return !qdisc_tx_is_noop(dev);
252 }
253
254 static void addrconf_del_rs_timer(struct inet6_dev *idev)
255 {
256 if (del_timer(&idev->rs_timer))
257 __in6_dev_put(idev);
258 }
259
260 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
261 {
262 if (del_timer(&ifp->dad_timer))
263 __in6_ifa_put(ifp);
264 }
265
266 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
267 unsigned long when)
268 {
269 if (!timer_pending(&idev->rs_timer))
270 in6_dev_hold(idev);
271 mod_timer(&idev->rs_timer, jiffies + when);
272 }
273
274 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
275 unsigned long when)
276 {
277 if (!timer_pending(&ifp->dad_timer))
278 in6_ifa_hold(ifp);
279 mod_timer(&ifp->dad_timer, jiffies + when);
280 }
281
282 static int snmp6_alloc_dev(struct inet6_dev *idev)
283 {
284 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
285 sizeof(struct ipstats_mib),
286 __alignof__(struct ipstats_mib)) < 0)
287 goto err_ip;
288 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
289 GFP_KERNEL);
290 if (!idev->stats.icmpv6dev)
291 goto err_icmp;
292 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
293 GFP_KERNEL);
294 if (!idev->stats.icmpv6msgdev)
295 goto err_icmpmsg;
296
297 return 0;
298
299 err_icmpmsg:
300 kfree(idev->stats.icmpv6dev);
301 err_icmp:
302 snmp_mib_free((void __percpu **)idev->stats.ipv6);
303 err_ip:
304 return -ENOMEM;
305 }
306
307 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
308 {
309 struct inet6_dev *ndev;
310
311 ASSERT_RTNL();
312
313 if (dev->mtu < IPV6_MIN_MTU)
314 return NULL;
315
316 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
317
318 if (ndev == NULL)
319 return NULL;
320
321 rwlock_init(&ndev->lock);
322 ndev->dev = dev;
323 INIT_LIST_HEAD(&ndev->addr_list);
324 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
325 (unsigned long)ndev);
326 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
327 ndev->cnf.mtu6 = dev->mtu;
328 ndev->cnf.sysctl = NULL;
329 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
330 if (ndev->nd_parms == NULL) {
331 kfree(ndev);
332 return NULL;
333 }
334 if (ndev->cnf.forwarding)
335 dev_disable_lro(dev);
336 /* We refer to the device */
337 dev_hold(dev);
338
339 if (snmp6_alloc_dev(ndev) < 0) {
340 ADBG(KERN_WARNING
341 "%s: cannot allocate memory for statistics; dev=%s.\n",
342 __func__, dev->name);
343 neigh_parms_release(&nd_tbl, ndev->nd_parms);
344 dev_put(dev);
345 kfree(ndev);
346 return NULL;
347 }
348
349 if (snmp6_register_dev(ndev) < 0) {
350 ADBG(KERN_WARNING
351 "%s: cannot create /proc/net/dev_snmp6/%s\n",
352 __func__, dev->name);
353 neigh_parms_release(&nd_tbl, ndev->nd_parms);
354 ndev->dead = 1;
355 in6_dev_finish_destroy(ndev);
356 return NULL;
357 }
358
359 /* One reference from device. We must do this before
360 * we invoke __ipv6_regen_rndid().
361 */
362 in6_dev_hold(ndev);
363
364 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
365 ndev->cnf.accept_dad = -1;
366
367 #if IS_ENABLED(CONFIG_IPV6_SIT)
368 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
369 pr_info("%s: Disabled Multicast RS\n", dev->name);
370 ndev->cnf.rtr_solicits = 0;
371 }
372 #endif
373
374 #ifdef CONFIG_IPV6_PRIVACY
375 INIT_LIST_HEAD(&ndev->tempaddr_list);
376 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
377 if ((dev->flags&IFF_LOOPBACK) ||
378 dev->type == ARPHRD_TUNNEL ||
379 dev->type == ARPHRD_TUNNEL6 ||
380 dev->type == ARPHRD_SIT ||
381 dev->type == ARPHRD_NONE) {
382 ndev->cnf.use_tempaddr = -1;
383 } else {
384 in6_dev_hold(ndev);
385 ipv6_regen_rndid((unsigned long) ndev);
386 }
387 #endif
388 ndev->token = in6addr_any;
389
390 if (netif_running(dev) && addrconf_qdisc_ok(dev))
391 ndev->if_flags |= IF_READY;
392
393 ipv6_mc_init_dev(ndev);
394 ndev->tstamp = jiffies;
395 addrconf_sysctl_register(ndev);
396 /* protected by rtnl_lock */
397 rcu_assign_pointer(dev->ip6_ptr, ndev);
398
399 /* Join interface-local all-node multicast group */
400 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
401
402 /* Join all-node multicast group */
403 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
404
405 /* Join all-router multicast group if forwarding is set */
406 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
407 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
408
409 return ndev;
410 }
411
412 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
413 {
414 struct inet6_dev *idev;
415
416 ASSERT_RTNL();
417
418 idev = __in6_dev_get(dev);
419 if (!idev) {
420 idev = ipv6_add_dev(dev);
421 if (!idev)
422 return NULL;
423 }
424
425 if (dev->flags&IFF_UP)
426 ipv6_mc_up(idev);
427 return idev;
428 }
429
430 static int inet6_netconf_msgsize_devconf(int type)
431 {
432 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
433 + nla_total_size(4); /* NETCONFA_IFINDEX */
434
435 /* type -1 is used for ALL */
436 if (type == -1 || type == NETCONFA_FORWARDING)
437 size += nla_total_size(4);
438 #ifdef CONFIG_IPV6_MROUTE
439 if (type == -1 || type == NETCONFA_MC_FORWARDING)
440 size += nla_total_size(4);
441 #endif
442
443 return size;
444 }
445
446 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
447 struct ipv6_devconf *devconf, u32 portid,
448 u32 seq, int event, unsigned int flags,
449 int type)
450 {
451 struct nlmsghdr *nlh;
452 struct netconfmsg *ncm;
453
454 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
455 flags);
456 if (nlh == NULL)
457 return -EMSGSIZE;
458
459 ncm = nlmsg_data(nlh);
460 ncm->ncm_family = AF_INET6;
461
462 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
463 goto nla_put_failure;
464
465 /* type -1 is used for ALL */
466 if ((type == -1 || type == NETCONFA_FORWARDING) &&
467 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
468 goto nla_put_failure;
469 #ifdef CONFIG_IPV6_MROUTE
470 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
471 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
472 devconf->mc_forwarding) < 0)
473 goto nla_put_failure;
474 #endif
475 return nlmsg_end(skb, nlh);
476
477 nla_put_failure:
478 nlmsg_cancel(skb, nlh);
479 return -EMSGSIZE;
480 }
481
482 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
483 struct ipv6_devconf *devconf)
484 {
485 struct sk_buff *skb;
486 int err = -ENOBUFS;
487
488 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
489 if (skb == NULL)
490 goto errout;
491
492 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
493 RTM_NEWNETCONF, 0, type);
494 if (err < 0) {
495 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
496 WARN_ON(err == -EMSGSIZE);
497 kfree_skb(skb);
498 goto errout;
499 }
500 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
501 return;
502 errout:
503 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
504 }
505
506 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
507 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
508 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
509 };
510
511 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
512 struct nlmsghdr *nlh)
513 {
514 struct net *net = sock_net(in_skb->sk);
515 struct nlattr *tb[NETCONFA_MAX+1];
516 struct netconfmsg *ncm;
517 struct sk_buff *skb;
518 struct ipv6_devconf *devconf;
519 struct inet6_dev *in6_dev;
520 struct net_device *dev;
521 int ifindex;
522 int err;
523
524 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
525 devconf_ipv6_policy);
526 if (err < 0)
527 goto errout;
528
529 err = EINVAL;
530 if (!tb[NETCONFA_IFINDEX])
531 goto errout;
532
533 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
534 switch (ifindex) {
535 case NETCONFA_IFINDEX_ALL:
536 devconf = net->ipv6.devconf_all;
537 break;
538 case NETCONFA_IFINDEX_DEFAULT:
539 devconf = net->ipv6.devconf_dflt;
540 break;
541 default:
542 dev = __dev_get_by_index(net, ifindex);
543 if (dev == NULL)
544 goto errout;
545 in6_dev = __in6_dev_get(dev);
546 if (in6_dev == NULL)
547 goto errout;
548 devconf = &in6_dev->cnf;
549 break;
550 }
551
552 err = -ENOBUFS;
553 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
554 if (skb == NULL)
555 goto errout;
556
557 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
558 NETLINK_CB(in_skb).portid,
559 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
560 -1);
561 if (err < 0) {
562 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
563 WARN_ON(err == -EMSGSIZE);
564 kfree_skb(skb);
565 goto errout;
566 }
567 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
568 errout:
569 return err;
570 }
571
572 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
573 struct netlink_callback *cb)
574 {
575 struct net *net = sock_net(skb->sk);
576 int h, s_h;
577 int idx, s_idx;
578 struct net_device *dev;
579 struct inet6_dev *idev;
580 struct hlist_head *head;
581
582 s_h = cb->args[0];
583 s_idx = idx = cb->args[1];
584
585 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
586 idx = 0;
587 head = &net->dev_index_head[h];
588 rcu_read_lock();
589 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
590 net->dev_base_seq;
591 hlist_for_each_entry_rcu(dev, head, index_hlist) {
592 if (idx < s_idx)
593 goto cont;
594 idev = __in6_dev_get(dev);
595 if (!idev)
596 goto cont;
597
598 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
599 &idev->cnf,
600 NETLINK_CB(cb->skb).portid,
601 cb->nlh->nlmsg_seq,
602 RTM_NEWNETCONF,
603 NLM_F_MULTI,
604 -1) <= 0) {
605 rcu_read_unlock();
606 goto done;
607 }
608 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
609 cont:
610 idx++;
611 }
612 rcu_read_unlock();
613 }
614 if (h == NETDEV_HASHENTRIES) {
615 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
616 net->ipv6.devconf_all,
617 NETLINK_CB(cb->skb).portid,
618 cb->nlh->nlmsg_seq,
619 RTM_NEWNETCONF, NLM_F_MULTI,
620 -1) <= 0)
621 goto done;
622 else
623 h++;
624 }
625 if (h == NETDEV_HASHENTRIES + 1) {
626 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
627 net->ipv6.devconf_dflt,
628 NETLINK_CB(cb->skb).portid,
629 cb->nlh->nlmsg_seq,
630 RTM_NEWNETCONF, NLM_F_MULTI,
631 -1) <= 0)
632 goto done;
633 else
634 h++;
635 }
636 done:
637 cb->args[0] = h;
638 cb->args[1] = idx;
639
640 return skb->len;
641 }
642
643 #ifdef CONFIG_SYSCTL
644 static void dev_forward_change(struct inet6_dev *idev)
645 {
646 struct net_device *dev;
647 struct inet6_ifaddr *ifa;
648
649 if (!idev)
650 return;
651 dev = idev->dev;
652 if (idev->cnf.forwarding)
653 dev_disable_lro(dev);
654 if (dev->flags & IFF_MULTICAST) {
655 if (idev->cnf.forwarding) {
656 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
657 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
658 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
659 } else {
660 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
661 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
662 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
663 }
664 }
665
666 list_for_each_entry(ifa, &idev->addr_list, if_list) {
667 if (ifa->flags&IFA_F_TENTATIVE)
668 continue;
669 if (idev->cnf.forwarding)
670 addrconf_join_anycast(ifa);
671 else
672 addrconf_leave_anycast(ifa);
673 }
674 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
675 dev->ifindex, &idev->cnf);
676 }
677
678
679 static void addrconf_forward_change(struct net *net, __s32 newf)
680 {
681 struct net_device *dev;
682 struct inet6_dev *idev;
683
684 for_each_netdev(net, dev) {
685 idev = __in6_dev_get(dev);
686 if (idev) {
687 int changed = (!idev->cnf.forwarding) ^ (!newf);
688 idev->cnf.forwarding = newf;
689 if (changed)
690 dev_forward_change(idev);
691 }
692 }
693 }
694
695 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
696 {
697 struct net *net;
698 int old;
699
700 if (!rtnl_trylock())
701 return restart_syscall();
702
703 net = (struct net *)table->extra2;
704 old = *p;
705 *p = newf;
706
707 if (p == &net->ipv6.devconf_dflt->forwarding) {
708 if ((!newf) ^ (!old))
709 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
710 NETCONFA_IFINDEX_DEFAULT,
711 net->ipv6.devconf_dflt);
712 rtnl_unlock();
713 return 0;
714 }
715
716 if (p == &net->ipv6.devconf_all->forwarding) {
717 net->ipv6.devconf_dflt->forwarding = newf;
718 addrconf_forward_change(net, newf);
719 if ((!newf) ^ (!old))
720 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
721 NETCONFA_IFINDEX_ALL,
722 net->ipv6.devconf_all);
723 } else if ((!newf) ^ (!old))
724 dev_forward_change((struct inet6_dev *)table->extra1);
725 rtnl_unlock();
726
727 if (newf)
728 rt6_purge_dflt_routers(net);
729 return 1;
730 }
731 #endif
732
733 /* Nobody refers to this ifaddr, destroy it */
734 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
735 {
736 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
737
738 #ifdef NET_REFCNT_DEBUG
739 pr_debug("%s\n", __func__);
740 #endif
741
742 in6_dev_put(ifp->idev);
743
744 if (del_timer(&ifp->dad_timer))
745 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
746
747 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
748 pr_warn("Freeing alive inet6 address %p\n", ifp);
749 return;
750 }
751 ip6_rt_put(ifp->rt);
752
753 kfree_rcu(ifp, rcu);
754 }
755
756 static void
757 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
758 {
759 struct list_head *p;
760 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
761
762 /*
763 * Each device address list is sorted in order of scope -
764 * global before linklocal.
765 */
766 list_for_each(p, &idev->addr_list) {
767 struct inet6_ifaddr *ifa
768 = list_entry(p, struct inet6_ifaddr, if_list);
769 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
770 break;
771 }
772
773 list_add_tail(&ifp->if_list, p);
774 }
775
776 static u32 inet6_addr_hash(const struct in6_addr *addr)
777 {
778 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
779 }
780
781 /* On success it returns ifp with increased reference count */
782
783 static struct inet6_ifaddr *
784 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
785 const struct in6_addr *peer_addr, int pfxlen,
786 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
787 {
788 struct inet6_ifaddr *ifa = NULL;
789 struct rt6_info *rt;
790 unsigned int hash;
791 int err = 0;
792 int addr_type = ipv6_addr_type(addr);
793
794 if (addr_type == IPV6_ADDR_ANY ||
795 addr_type & IPV6_ADDR_MULTICAST ||
796 (!(idev->dev->flags & IFF_LOOPBACK) &&
797 addr_type & IPV6_ADDR_LOOPBACK))
798 return ERR_PTR(-EADDRNOTAVAIL);
799
800 rcu_read_lock_bh();
801 if (idev->dead) {
802 err = -ENODEV; /*XXX*/
803 goto out2;
804 }
805
806 if (idev->cnf.disable_ipv6) {
807 err = -EACCES;
808 goto out2;
809 }
810
811 spin_lock(&addrconf_hash_lock);
812
813 /* Ignore adding duplicate addresses on an interface */
814 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
815 ADBG("ipv6_add_addr: already assigned\n");
816 err = -EEXIST;
817 goto out;
818 }
819
820 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
821
822 if (ifa == NULL) {
823 ADBG("ipv6_add_addr: malloc failed\n");
824 err = -ENOBUFS;
825 goto out;
826 }
827
828 rt = addrconf_dst_alloc(idev, addr, false);
829 if (IS_ERR(rt)) {
830 err = PTR_ERR(rt);
831 goto out;
832 }
833
834 ifa->addr = *addr;
835 if (peer_addr)
836 ifa->peer_addr = *peer_addr;
837
838 spin_lock_init(&ifa->lock);
839 spin_lock_init(&ifa->state_lock);
840 setup_timer(&ifa->dad_timer, addrconf_dad_timer,
841 (unsigned long)ifa);
842 INIT_HLIST_NODE(&ifa->addr_lst);
843 ifa->scope = scope;
844 ifa->prefix_len = pfxlen;
845 ifa->flags = flags | IFA_F_TENTATIVE;
846 ifa->valid_lft = valid_lft;
847 ifa->prefered_lft = prefered_lft;
848 ifa->cstamp = ifa->tstamp = jiffies;
849 ifa->tokenized = false;
850
851 ifa->rt = rt;
852
853 ifa->idev = idev;
854 in6_dev_hold(idev);
855 /* For caller */
856 in6_ifa_hold(ifa);
857
858 /* Add to big hash table */
859 hash = inet6_addr_hash(addr);
860
861 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
862 spin_unlock(&addrconf_hash_lock);
863
864 write_lock(&idev->lock);
865 /* Add to inet6_dev unicast addr list. */
866 ipv6_link_dev_addr(idev, ifa);
867
868 #ifdef CONFIG_IPV6_PRIVACY
869 if (ifa->flags&IFA_F_TEMPORARY) {
870 list_add(&ifa->tmp_list, &idev->tempaddr_list);
871 in6_ifa_hold(ifa);
872 }
873 #endif
874
875 in6_ifa_hold(ifa);
876 write_unlock(&idev->lock);
877 out2:
878 rcu_read_unlock_bh();
879
880 if (likely(err == 0))
881 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
882 else {
883 kfree(ifa);
884 ifa = ERR_PTR(err);
885 }
886
887 return ifa;
888 out:
889 spin_unlock(&addrconf_hash_lock);
890 goto out2;
891 }
892
893 /* This function wants to get referenced ifp and releases it before return */
894
895 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
896 {
897 struct inet6_ifaddr *ifa, *ifn;
898 struct inet6_dev *idev = ifp->idev;
899 int state;
900 int deleted = 0, onlink = 0;
901 unsigned long expires = jiffies;
902
903 spin_lock_bh(&ifp->state_lock);
904 state = ifp->state;
905 ifp->state = INET6_IFADDR_STATE_DEAD;
906 spin_unlock_bh(&ifp->state_lock);
907
908 if (state == INET6_IFADDR_STATE_DEAD)
909 goto out;
910
911 spin_lock_bh(&addrconf_hash_lock);
912 hlist_del_init_rcu(&ifp->addr_lst);
913 spin_unlock_bh(&addrconf_hash_lock);
914
915 write_lock_bh(&idev->lock);
916 #ifdef CONFIG_IPV6_PRIVACY
917 if (ifp->flags&IFA_F_TEMPORARY) {
918 list_del(&ifp->tmp_list);
919 if (ifp->ifpub) {
920 in6_ifa_put(ifp->ifpub);
921 ifp->ifpub = NULL;
922 }
923 __in6_ifa_put(ifp);
924 }
925 #endif
926
927 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
928 if (ifa == ifp) {
929 list_del_init(&ifp->if_list);
930 __in6_ifa_put(ifp);
931
932 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
933 break;
934 deleted = 1;
935 continue;
936 } else if (ifp->flags & IFA_F_PERMANENT) {
937 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
938 ifp->prefix_len)) {
939 if (ifa->flags & IFA_F_PERMANENT) {
940 onlink = 1;
941 if (deleted)
942 break;
943 } else {
944 unsigned long lifetime;
945
946 if (!onlink)
947 onlink = -1;
948
949 spin_lock(&ifa->lock);
950
951 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
952 /*
953 * Note: Because this address is
954 * not permanent, lifetime <
955 * LONG_MAX / HZ here.
956 */
957 if (time_before(expires,
958 ifa->tstamp + lifetime * HZ))
959 expires = ifa->tstamp + lifetime * HZ;
960 spin_unlock(&ifa->lock);
961 }
962 }
963 }
964 }
965 write_unlock_bh(&idev->lock);
966
967 addrconf_del_dad_timer(ifp);
968
969 ipv6_ifa_notify(RTM_DELADDR, ifp);
970
971 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
972
973 /*
974 * Purge or update corresponding prefix
975 *
976 * 1) we don't purge prefix here if address was not permanent.
977 * prefix is managed by its own lifetime.
978 * 2) if there're no addresses, delete prefix.
979 * 3) if there're still other permanent address(es),
980 * corresponding prefix is still permanent.
981 * 4) otherwise, update prefix lifetime to the
982 * longest valid lifetime among the corresponding
983 * addresses on the device.
984 * Note: subsequent RA will update lifetime.
985 *
986 * --yoshfuji
987 */
988 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
989 struct in6_addr prefix;
990 struct rt6_info *rt;
991
992 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
993
994 rt = addrconf_get_prefix_route(&prefix,
995 ifp->prefix_len,
996 ifp->idev->dev,
997 0, RTF_GATEWAY | RTF_DEFAULT);
998
999 if (rt) {
1000 if (onlink == 0) {
1001 ip6_del_rt(rt);
1002 rt = NULL;
1003 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1004 rt6_set_expires(rt, expires);
1005 }
1006 }
1007 ip6_rt_put(rt);
1008 }
1009
1010 /* clean up prefsrc entries */
1011 rt6_remove_prefsrc(ifp);
1012 out:
1013 in6_ifa_put(ifp);
1014 }
1015
1016 #ifdef CONFIG_IPV6_PRIVACY
1017 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1018 {
1019 struct inet6_dev *idev = ifp->idev;
1020 struct in6_addr addr, *tmpaddr;
1021 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1022 unsigned long regen_advance;
1023 int tmp_plen;
1024 int ret = 0;
1025 u32 addr_flags;
1026 unsigned long now = jiffies;
1027
1028 write_lock(&idev->lock);
1029 if (ift) {
1030 spin_lock_bh(&ift->lock);
1031 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1032 spin_unlock_bh(&ift->lock);
1033 tmpaddr = &addr;
1034 } else {
1035 tmpaddr = NULL;
1036 }
1037 retry:
1038 in6_dev_hold(idev);
1039 if (idev->cnf.use_tempaddr <= 0) {
1040 write_unlock(&idev->lock);
1041 pr_info("%s: use_tempaddr is disabled\n", __func__);
1042 in6_dev_put(idev);
1043 ret = -1;
1044 goto out;
1045 }
1046 spin_lock_bh(&ifp->lock);
1047 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1048 idev->cnf.use_tempaddr = -1; /*XXX*/
1049 spin_unlock_bh(&ifp->lock);
1050 write_unlock(&idev->lock);
1051 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1052 __func__);
1053 in6_dev_put(idev);
1054 ret = -1;
1055 goto out;
1056 }
1057 in6_ifa_hold(ifp);
1058 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1059 __ipv6_try_regen_rndid(idev, tmpaddr);
1060 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1061 age = (now - ifp->tstamp) / HZ;
1062 tmp_valid_lft = min_t(__u32,
1063 ifp->valid_lft,
1064 idev->cnf.temp_valid_lft + age);
1065 tmp_prefered_lft = min_t(__u32,
1066 ifp->prefered_lft,
1067 idev->cnf.temp_prefered_lft + age -
1068 idev->cnf.max_desync_factor);
1069 tmp_plen = ifp->prefix_len;
1070 tmp_tstamp = ifp->tstamp;
1071 spin_unlock_bh(&ifp->lock);
1072
1073 regen_advance = idev->cnf.regen_max_retry *
1074 idev->cnf.dad_transmits *
1075 idev->nd_parms->retrans_time / HZ;
1076 write_unlock(&idev->lock);
1077
1078 /* A temporary address is created only if this calculated Preferred
1079 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1080 * an implementation must not create a temporary address with a zero
1081 * Preferred Lifetime.
1082 */
1083 if (tmp_prefered_lft <= regen_advance) {
1084 in6_ifa_put(ifp);
1085 in6_dev_put(idev);
1086 ret = -1;
1087 goto out;
1088 }
1089
1090 addr_flags = IFA_F_TEMPORARY;
1091 /* set in addrconf_prefix_rcv() */
1092 if (ifp->flags & IFA_F_OPTIMISTIC)
1093 addr_flags |= IFA_F_OPTIMISTIC;
1094
1095 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1096 ipv6_addr_scope(&addr), addr_flags,
1097 tmp_valid_lft, tmp_prefered_lft);
1098 if (IS_ERR(ift)) {
1099 in6_ifa_put(ifp);
1100 in6_dev_put(idev);
1101 pr_info("%s: retry temporary address regeneration\n", __func__);
1102 tmpaddr = &addr;
1103 write_lock(&idev->lock);
1104 goto retry;
1105 }
1106
1107 spin_lock_bh(&ift->lock);
1108 ift->ifpub = ifp;
1109 ift->cstamp = now;
1110 ift->tstamp = tmp_tstamp;
1111 spin_unlock_bh(&ift->lock);
1112
1113 addrconf_dad_start(ift);
1114 in6_ifa_put(ift);
1115 in6_dev_put(idev);
1116 out:
1117 return ret;
1118 }
1119 #endif
1120
1121 /*
1122 * Choose an appropriate source address (RFC3484)
1123 */
1124 enum {
1125 IPV6_SADDR_RULE_INIT = 0,
1126 IPV6_SADDR_RULE_LOCAL,
1127 IPV6_SADDR_RULE_SCOPE,
1128 IPV6_SADDR_RULE_PREFERRED,
1129 #ifdef CONFIG_IPV6_MIP6
1130 IPV6_SADDR_RULE_HOA,
1131 #endif
1132 IPV6_SADDR_RULE_OIF,
1133 IPV6_SADDR_RULE_LABEL,
1134 #ifdef CONFIG_IPV6_PRIVACY
1135 IPV6_SADDR_RULE_PRIVACY,
1136 #endif
1137 IPV6_SADDR_RULE_ORCHID,
1138 IPV6_SADDR_RULE_PREFIX,
1139 IPV6_SADDR_RULE_MAX
1140 };
1141
1142 struct ipv6_saddr_score {
1143 int rule;
1144 int addr_type;
1145 struct inet6_ifaddr *ifa;
1146 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1147 int scopedist;
1148 int matchlen;
1149 };
1150
1151 struct ipv6_saddr_dst {
1152 const struct in6_addr *addr;
1153 int ifindex;
1154 int scope;
1155 int label;
1156 unsigned int prefs;
1157 };
1158
1159 static inline int ipv6_saddr_preferred(int type)
1160 {
1161 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1162 return 1;
1163 return 0;
1164 }
1165
1166 static int ipv6_get_saddr_eval(struct net *net,
1167 struct ipv6_saddr_score *score,
1168 struct ipv6_saddr_dst *dst,
1169 int i)
1170 {
1171 int ret;
1172
1173 if (i <= score->rule) {
1174 switch (i) {
1175 case IPV6_SADDR_RULE_SCOPE:
1176 ret = score->scopedist;
1177 break;
1178 case IPV6_SADDR_RULE_PREFIX:
1179 ret = score->matchlen;
1180 break;
1181 default:
1182 ret = !!test_bit(i, score->scorebits);
1183 }
1184 goto out;
1185 }
1186
1187 switch (i) {
1188 case IPV6_SADDR_RULE_INIT:
1189 /* Rule 0: remember if hiscore is not ready yet */
1190 ret = !!score->ifa;
1191 break;
1192 case IPV6_SADDR_RULE_LOCAL:
1193 /* Rule 1: Prefer same address */
1194 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1195 break;
1196 case IPV6_SADDR_RULE_SCOPE:
1197 /* Rule 2: Prefer appropriate scope
1198 *
1199 * ret
1200 * ^
1201 * -1 | d 15
1202 * ---+--+-+---> scope
1203 * |
1204 * | d is scope of the destination.
1205 * B-d | \
1206 * | \ <- smaller scope is better if
1207 * B-15 | \ if scope is enough for destinaion.
1208 * | ret = B - scope (-1 <= scope >= d <= 15).
1209 * d-C-1 | /
1210 * |/ <- greater is better
1211 * -C / if scope is not enough for destination.
1212 * /| ret = scope - C (-1 <= d < scope <= 15).
1213 *
1214 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1215 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1216 * Assume B = 0 and we get C > 29.
1217 */
1218 ret = __ipv6_addr_src_scope(score->addr_type);
1219 if (ret >= dst->scope)
1220 ret = -ret;
1221 else
1222 ret -= 128; /* 30 is enough */
1223 score->scopedist = ret;
1224 break;
1225 case IPV6_SADDR_RULE_PREFERRED:
1226 /* Rule 3: Avoid deprecated and optimistic addresses */
1227 ret = ipv6_saddr_preferred(score->addr_type) ||
1228 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1229 break;
1230 #ifdef CONFIG_IPV6_MIP6
1231 case IPV6_SADDR_RULE_HOA:
1232 {
1233 /* Rule 4: Prefer home address */
1234 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1235 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1236 break;
1237 }
1238 #endif
1239 case IPV6_SADDR_RULE_OIF:
1240 /* Rule 5: Prefer outgoing interface */
1241 ret = (!dst->ifindex ||
1242 dst->ifindex == score->ifa->idev->dev->ifindex);
1243 break;
1244 case IPV6_SADDR_RULE_LABEL:
1245 /* Rule 6: Prefer matching label */
1246 ret = ipv6_addr_label(net,
1247 &score->ifa->addr, score->addr_type,
1248 score->ifa->idev->dev->ifindex) == dst->label;
1249 break;
1250 #ifdef CONFIG_IPV6_PRIVACY
1251 case IPV6_SADDR_RULE_PRIVACY:
1252 {
1253 /* Rule 7: Prefer public address
1254 * Note: prefer temporary address if use_tempaddr >= 2
1255 */
1256 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1257 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1258 score->ifa->idev->cnf.use_tempaddr >= 2;
1259 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1260 break;
1261 }
1262 #endif
1263 case IPV6_SADDR_RULE_ORCHID:
1264 /* Rule 8-: Prefer ORCHID vs ORCHID or
1265 * non-ORCHID vs non-ORCHID
1266 */
1267 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1268 ipv6_addr_orchid(dst->addr));
1269 break;
1270 case IPV6_SADDR_RULE_PREFIX:
1271 /* Rule 8: Use longest matching prefix */
1272 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1273 if (ret > score->ifa->prefix_len)
1274 ret = score->ifa->prefix_len;
1275 score->matchlen = ret;
1276 break;
1277 default:
1278 ret = 0;
1279 }
1280
1281 if (ret)
1282 __set_bit(i, score->scorebits);
1283 score->rule = i;
1284 out:
1285 return ret;
1286 }
1287
1288 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1289 const struct in6_addr *daddr, unsigned int prefs,
1290 struct in6_addr *saddr)
1291 {
1292 struct ipv6_saddr_score scores[2],
1293 *score = &scores[0], *hiscore = &scores[1];
1294 struct ipv6_saddr_dst dst;
1295 struct net_device *dev;
1296 int dst_type;
1297
1298 dst_type = __ipv6_addr_type(daddr);
1299 dst.addr = daddr;
1300 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1301 dst.scope = __ipv6_addr_src_scope(dst_type);
1302 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1303 dst.prefs = prefs;
1304
1305 hiscore->rule = -1;
1306 hiscore->ifa = NULL;
1307
1308 rcu_read_lock();
1309
1310 for_each_netdev_rcu(net, dev) {
1311 struct inet6_dev *idev;
1312
1313 /* Candidate Source Address (section 4)
1314 * - multicast and link-local destination address,
1315 * the set of candidate source address MUST only
1316 * include addresses assigned to interfaces
1317 * belonging to the same link as the outgoing
1318 * interface.
1319 * (- For site-local destination addresses, the
1320 * set of candidate source addresses MUST only
1321 * include addresses assigned to interfaces
1322 * belonging to the same site as the outgoing
1323 * interface.)
1324 */
1325 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1326 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1327 dst.ifindex && dev->ifindex != dst.ifindex)
1328 continue;
1329
1330 idev = __in6_dev_get(dev);
1331 if (!idev)
1332 continue;
1333
1334 read_lock_bh(&idev->lock);
1335 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1336 int i;
1337
1338 /*
1339 * - Tentative Address (RFC2462 section 5.4)
1340 * - A tentative address is not considered
1341 * "assigned to an interface" in the traditional
1342 * sense, unless it is also flagged as optimistic.
1343 * - Candidate Source Address (section 4)
1344 * - In any case, anycast addresses, multicast
1345 * addresses, and the unspecified address MUST
1346 * NOT be included in a candidate set.
1347 */
1348 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1349 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1350 continue;
1351
1352 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1353
1354 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1355 score->addr_type & IPV6_ADDR_MULTICAST)) {
1356 LIMIT_NETDEBUG(KERN_DEBUG
1357 "ADDRCONF: unspecified / multicast address "
1358 "assigned as unicast address on %s",
1359 dev->name);
1360 continue;
1361 }
1362
1363 score->rule = -1;
1364 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1365
1366 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1367 int minihiscore, miniscore;
1368
1369 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1370 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1371
1372 if (minihiscore > miniscore) {
1373 if (i == IPV6_SADDR_RULE_SCOPE &&
1374 score->scopedist > 0) {
1375 /*
1376 * special case:
1377 * each remaining entry
1378 * has too small (not enough)
1379 * scope, because ifa entries
1380 * are sorted by their scope
1381 * values.
1382 */
1383 goto try_nextdev;
1384 }
1385 break;
1386 } else if (minihiscore < miniscore) {
1387 if (hiscore->ifa)
1388 in6_ifa_put(hiscore->ifa);
1389
1390 in6_ifa_hold(score->ifa);
1391
1392 swap(hiscore, score);
1393
1394 /* restore our iterator */
1395 score->ifa = hiscore->ifa;
1396
1397 break;
1398 }
1399 }
1400 }
1401 try_nextdev:
1402 read_unlock_bh(&idev->lock);
1403 }
1404 rcu_read_unlock();
1405
1406 if (!hiscore->ifa)
1407 return -EADDRNOTAVAIL;
1408
1409 *saddr = hiscore->ifa->addr;
1410 in6_ifa_put(hiscore->ifa);
1411 return 0;
1412 }
1413 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1414
1415 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1416 unsigned char banned_flags)
1417 {
1418 struct inet6_ifaddr *ifp;
1419 int err = -EADDRNOTAVAIL;
1420
1421 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1422 if (ifp->scope == IFA_LINK &&
1423 !(ifp->flags & banned_flags)) {
1424 *addr = ifp->addr;
1425 err = 0;
1426 break;
1427 }
1428 }
1429 return err;
1430 }
1431
1432 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1433 unsigned char banned_flags)
1434 {
1435 struct inet6_dev *idev;
1436 int err = -EADDRNOTAVAIL;
1437
1438 rcu_read_lock();
1439 idev = __in6_dev_get(dev);
1440 if (idev) {
1441 read_lock_bh(&idev->lock);
1442 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1443 read_unlock_bh(&idev->lock);
1444 }
1445 rcu_read_unlock();
1446 return err;
1447 }
1448
1449 static int ipv6_count_addresses(struct inet6_dev *idev)
1450 {
1451 int cnt = 0;
1452 struct inet6_ifaddr *ifp;
1453
1454 read_lock_bh(&idev->lock);
1455 list_for_each_entry(ifp, &idev->addr_list, if_list)
1456 cnt++;
1457 read_unlock_bh(&idev->lock);
1458 return cnt;
1459 }
1460
1461 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1462 const struct net_device *dev, int strict)
1463 {
1464 struct inet6_ifaddr *ifp;
1465 unsigned int hash = inet6_addr_hash(addr);
1466
1467 rcu_read_lock_bh();
1468 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1469 if (!net_eq(dev_net(ifp->idev->dev), net))
1470 continue;
1471 if (ipv6_addr_equal(&ifp->addr, addr) &&
1472 !(ifp->flags&IFA_F_TENTATIVE) &&
1473 (dev == NULL || ifp->idev->dev == dev ||
1474 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1475 rcu_read_unlock_bh();
1476 return 1;
1477 }
1478 }
1479
1480 rcu_read_unlock_bh();
1481 return 0;
1482 }
1483 EXPORT_SYMBOL(ipv6_chk_addr);
1484
1485 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1486 struct net_device *dev)
1487 {
1488 unsigned int hash = inet6_addr_hash(addr);
1489 struct inet6_ifaddr *ifp;
1490
1491 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1492 if (!net_eq(dev_net(ifp->idev->dev), net))
1493 continue;
1494 if (ipv6_addr_equal(&ifp->addr, addr)) {
1495 if (dev == NULL || ifp->idev->dev == dev)
1496 return true;
1497 }
1498 }
1499 return false;
1500 }
1501
1502 /* Compares an address/prefix_len with addresses on device @dev.
1503 * If one is found it returns true.
1504 */
1505 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1506 const unsigned int prefix_len, struct net_device *dev)
1507 {
1508 struct inet6_dev *idev;
1509 struct inet6_ifaddr *ifa;
1510 bool ret = false;
1511
1512 rcu_read_lock();
1513 idev = __in6_dev_get(dev);
1514 if (idev) {
1515 read_lock_bh(&idev->lock);
1516 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1517 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1518 if (ret)
1519 break;
1520 }
1521 read_unlock_bh(&idev->lock);
1522 }
1523 rcu_read_unlock();
1524
1525 return ret;
1526 }
1527 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1528
1529 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1530 {
1531 struct inet6_dev *idev;
1532 struct inet6_ifaddr *ifa;
1533 int onlink;
1534
1535 onlink = 0;
1536 rcu_read_lock();
1537 idev = __in6_dev_get(dev);
1538 if (idev) {
1539 read_lock_bh(&idev->lock);
1540 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1541 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1542 ifa->prefix_len);
1543 if (onlink)
1544 break;
1545 }
1546 read_unlock_bh(&idev->lock);
1547 }
1548 rcu_read_unlock();
1549 return onlink;
1550 }
1551 EXPORT_SYMBOL(ipv6_chk_prefix);
1552
1553 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1554 struct net_device *dev, int strict)
1555 {
1556 struct inet6_ifaddr *ifp, *result = NULL;
1557 unsigned int hash = inet6_addr_hash(addr);
1558
1559 rcu_read_lock_bh();
1560 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1561 if (!net_eq(dev_net(ifp->idev->dev), net))
1562 continue;
1563 if (ipv6_addr_equal(&ifp->addr, addr)) {
1564 if (dev == NULL || ifp->idev->dev == dev ||
1565 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1566 result = ifp;
1567 in6_ifa_hold(ifp);
1568 break;
1569 }
1570 }
1571 }
1572 rcu_read_unlock_bh();
1573
1574 return result;
1575 }
1576
1577 /* Gets referenced address, destroys ifaddr */
1578
1579 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1580 {
1581 if (ifp->flags&IFA_F_PERMANENT) {
1582 spin_lock_bh(&ifp->lock);
1583 addrconf_del_dad_timer(ifp);
1584 ifp->flags |= IFA_F_TENTATIVE;
1585 if (dad_failed)
1586 ifp->flags |= IFA_F_DADFAILED;
1587 spin_unlock_bh(&ifp->lock);
1588 if (dad_failed)
1589 ipv6_ifa_notify(0, ifp);
1590 in6_ifa_put(ifp);
1591 #ifdef CONFIG_IPV6_PRIVACY
1592 } else if (ifp->flags&IFA_F_TEMPORARY) {
1593 struct inet6_ifaddr *ifpub;
1594 spin_lock_bh(&ifp->lock);
1595 ifpub = ifp->ifpub;
1596 if (ifpub) {
1597 in6_ifa_hold(ifpub);
1598 spin_unlock_bh(&ifp->lock);
1599 ipv6_create_tempaddr(ifpub, ifp);
1600 in6_ifa_put(ifpub);
1601 } else {
1602 spin_unlock_bh(&ifp->lock);
1603 }
1604 ipv6_del_addr(ifp);
1605 #endif
1606 } else
1607 ipv6_del_addr(ifp);
1608 }
1609
1610 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1611 {
1612 int err = -ENOENT;
1613
1614 spin_lock(&ifp->state_lock);
1615 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1616 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1617 err = 0;
1618 }
1619 spin_unlock(&ifp->state_lock);
1620
1621 return err;
1622 }
1623
1624 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1625 {
1626 struct inet6_dev *idev = ifp->idev;
1627
1628 if (addrconf_dad_end(ifp)) {
1629 in6_ifa_put(ifp);
1630 return;
1631 }
1632
1633 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1634 ifp->idev->dev->name, &ifp->addr);
1635
1636 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1637 struct in6_addr addr;
1638
1639 addr.s6_addr32[0] = htonl(0xfe800000);
1640 addr.s6_addr32[1] = 0;
1641
1642 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1643 ipv6_addr_equal(&ifp->addr, &addr)) {
1644 /* DAD failed for link-local based on MAC address */
1645 idev->cnf.disable_ipv6 = 1;
1646
1647 pr_info("%s: IPv6 being disabled!\n",
1648 ifp->idev->dev->name);
1649 }
1650 }
1651
1652 addrconf_dad_stop(ifp, 1);
1653 }
1654
1655 /* Join to solicited addr multicast group. */
1656
1657 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1658 {
1659 struct in6_addr maddr;
1660
1661 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1662 return;
1663
1664 addrconf_addr_solict_mult(addr, &maddr);
1665 ipv6_dev_mc_inc(dev, &maddr);
1666 }
1667
1668 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1669 {
1670 struct in6_addr maddr;
1671
1672 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1673 return;
1674
1675 addrconf_addr_solict_mult(addr, &maddr);
1676 __ipv6_dev_mc_dec(idev, &maddr);
1677 }
1678
1679 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1680 {
1681 struct in6_addr addr;
1682 if (ifp->prefix_len == 127) /* RFC 6164 */
1683 return;
1684 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1685 if (ipv6_addr_any(&addr))
1686 return;
1687 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1688 }
1689
1690 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1691 {
1692 struct in6_addr addr;
1693 if (ifp->prefix_len == 127) /* RFC 6164 */
1694 return;
1695 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1696 if (ipv6_addr_any(&addr))
1697 return;
1698 __ipv6_dev_ac_dec(ifp->idev, &addr);
1699 }
1700
1701 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1702 {
1703 if (dev->addr_len != ETH_ALEN)
1704 return -1;
1705 memcpy(eui, dev->dev_addr, 3);
1706 memcpy(eui + 5, dev->dev_addr + 3, 3);
1707
1708 /*
1709 * The zSeries OSA network cards can be shared among various
1710 * OS instances, but the OSA cards have only one MAC address.
1711 * This leads to duplicate address conflicts in conjunction
1712 * with IPv6 if more than one instance uses the same card.
1713 *
1714 * The driver for these cards can deliver a unique 16-bit
1715 * identifier for each instance sharing the same card. It is
1716 * placed instead of 0xFFFE in the interface identifier. The
1717 * "u" bit of the interface identifier is not inverted in this
1718 * case. Hence the resulting interface identifier has local
1719 * scope according to RFC2373.
1720 */
1721 if (dev->dev_id) {
1722 eui[3] = (dev->dev_id >> 8) & 0xFF;
1723 eui[4] = dev->dev_id & 0xFF;
1724 } else {
1725 eui[3] = 0xFF;
1726 eui[4] = 0xFE;
1727 eui[0] ^= 2;
1728 }
1729 return 0;
1730 }
1731
1732 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1733 {
1734 if (dev->addr_len != IEEE802154_ADDR_LEN)
1735 return -1;
1736 memcpy(eui, dev->dev_addr, 8);
1737 eui[0] ^= 2;
1738 return 0;
1739 }
1740
1741 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1742 {
1743 union fwnet_hwaddr *ha;
1744
1745 if (dev->addr_len != FWNET_ALEN)
1746 return -1;
1747
1748 ha = (union fwnet_hwaddr *)dev->dev_addr;
1749
1750 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1751 eui[0] ^= 2;
1752 return 0;
1753 }
1754
1755 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1756 {
1757 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1758 if (dev->addr_len != ARCNET_ALEN)
1759 return -1;
1760 memset(eui, 0, 7);
1761 eui[7] = *(u8 *)dev->dev_addr;
1762 return 0;
1763 }
1764
1765 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1766 {
1767 if (dev->addr_len != INFINIBAND_ALEN)
1768 return -1;
1769 memcpy(eui, dev->dev_addr + 12, 8);
1770 eui[0] |= 2;
1771 return 0;
1772 }
1773
1774 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1775 {
1776 if (addr == 0)
1777 return -1;
1778 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1779 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1780 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1781 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1782 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1783 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1784 eui[1] = 0;
1785 eui[2] = 0x5E;
1786 eui[3] = 0xFE;
1787 memcpy(eui + 4, &addr, 4);
1788 return 0;
1789 }
1790
1791 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1792 {
1793 if (dev->priv_flags & IFF_ISATAP)
1794 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1795 return -1;
1796 }
1797
1798 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1799 {
1800 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1801 }
1802
1803 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1804 {
1805 memcpy(eui, dev->perm_addr, 3);
1806 memcpy(eui + 5, dev->perm_addr + 3, 3);
1807 eui[3] = 0xFF;
1808 eui[4] = 0xFE;
1809 eui[0] ^= 2;
1810 return 0;
1811 }
1812
1813 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1814 {
1815 switch (dev->type) {
1816 case ARPHRD_ETHER:
1817 case ARPHRD_FDDI:
1818 return addrconf_ifid_eui48(eui, dev);
1819 case ARPHRD_ARCNET:
1820 return addrconf_ifid_arcnet(eui, dev);
1821 case ARPHRD_INFINIBAND:
1822 return addrconf_ifid_infiniband(eui, dev);
1823 case ARPHRD_SIT:
1824 return addrconf_ifid_sit(eui, dev);
1825 case ARPHRD_IPGRE:
1826 return addrconf_ifid_gre(eui, dev);
1827 case ARPHRD_IEEE802154:
1828 return addrconf_ifid_eui64(eui, dev);
1829 case ARPHRD_IEEE1394:
1830 return addrconf_ifid_ieee1394(eui, dev);
1831 case ARPHRD_TUNNEL6:
1832 return addrconf_ifid_ip6tnl(eui, dev);
1833 }
1834 return -1;
1835 }
1836
1837 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1838 {
1839 int err = -1;
1840 struct inet6_ifaddr *ifp;
1841
1842 read_lock_bh(&idev->lock);
1843 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1844 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1845 memcpy(eui, ifp->addr.s6_addr+8, 8);
1846 err = 0;
1847 break;
1848 }
1849 }
1850 read_unlock_bh(&idev->lock);
1851 return err;
1852 }
1853
1854 #ifdef CONFIG_IPV6_PRIVACY
1855 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1856 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1857 {
1858 regen:
1859 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1860 idev->rndid[0] &= ~0x02;
1861
1862 /*
1863 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1864 * check if generated address is not inappropriate
1865 *
1866 * - Reserved subnet anycast (RFC 2526)
1867 * 11111101 11....11 1xxxxxxx
1868 * - ISATAP (RFC4214) 6.1
1869 * 00-00-5E-FE-xx-xx-xx-xx
1870 * - value 0
1871 * - XXX: already assigned to an address on the device
1872 */
1873 if (idev->rndid[0] == 0xfd &&
1874 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1875 (idev->rndid[7]&0x80))
1876 goto regen;
1877 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1878 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1879 goto regen;
1880 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1881 goto regen;
1882 }
1883 }
1884
1885 static void ipv6_regen_rndid(unsigned long data)
1886 {
1887 struct inet6_dev *idev = (struct inet6_dev *) data;
1888 unsigned long expires;
1889
1890 rcu_read_lock_bh();
1891 write_lock_bh(&idev->lock);
1892
1893 if (idev->dead)
1894 goto out;
1895
1896 __ipv6_regen_rndid(idev);
1897
1898 expires = jiffies +
1899 idev->cnf.temp_prefered_lft * HZ -
1900 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1901 idev->cnf.max_desync_factor * HZ;
1902 if (time_before(expires, jiffies)) {
1903 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1904 __func__, idev->dev->name);
1905 goto out;
1906 }
1907
1908 if (!mod_timer(&idev->regen_timer, expires))
1909 in6_dev_hold(idev);
1910
1911 out:
1912 write_unlock_bh(&idev->lock);
1913 rcu_read_unlock_bh();
1914 in6_dev_put(idev);
1915 }
1916
1917 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1918 {
1919 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1920 __ipv6_regen_rndid(idev);
1921 }
1922 #endif
1923
1924 /*
1925 * Add prefix route.
1926 */
1927
1928 static void
1929 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1930 unsigned long expires, u32 flags)
1931 {
1932 struct fib6_config cfg = {
1933 .fc_table = RT6_TABLE_PREFIX,
1934 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1935 .fc_ifindex = dev->ifindex,
1936 .fc_expires = expires,
1937 .fc_dst_len = plen,
1938 .fc_flags = RTF_UP | flags,
1939 .fc_nlinfo.nl_net = dev_net(dev),
1940 .fc_protocol = RTPROT_KERNEL,
1941 };
1942
1943 cfg.fc_dst = *pfx;
1944
1945 /* Prevent useless cloning on PtP SIT.
1946 This thing is done here expecting that the whole
1947 class of non-broadcast devices need not cloning.
1948 */
1949 #if IS_ENABLED(CONFIG_IPV6_SIT)
1950 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1951 cfg.fc_flags |= RTF_NONEXTHOP;
1952 #endif
1953
1954 ip6_route_add(&cfg);
1955 }
1956
1957
1958 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1959 int plen,
1960 const struct net_device *dev,
1961 u32 flags, u32 noflags)
1962 {
1963 struct fib6_node *fn;
1964 struct rt6_info *rt = NULL;
1965 struct fib6_table *table;
1966
1967 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1968 if (table == NULL)
1969 return NULL;
1970
1971 read_lock_bh(&table->tb6_lock);
1972 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1973 if (!fn)
1974 goto out;
1975 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1976 if (rt->dst.dev->ifindex != dev->ifindex)
1977 continue;
1978 if ((rt->rt6i_flags & flags) != flags)
1979 continue;
1980 if ((rt->rt6i_flags & noflags) != 0)
1981 continue;
1982 dst_hold(&rt->dst);
1983 break;
1984 }
1985 out:
1986 read_unlock_bh(&table->tb6_lock);
1987 return rt;
1988 }
1989
1990
1991 /* Create "default" multicast route to the interface */
1992
1993 static void addrconf_add_mroute(struct net_device *dev)
1994 {
1995 struct fib6_config cfg = {
1996 .fc_table = RT6_TABLE_LOCAL,
1997 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1998 .fc_ifindex = dev->ifindex,
1999 .fc_dst_len = 8,
2000 .fc_flags = RTF_UP,
2001 .fc_nlinfo.nl_net = dev_net(dev),
2002 };
2003
2004 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2005
2006 ip6_route_add(&cfg);
2007 }
2008
2009 #if IS_ENABLED(CONFIG_IPV6_SIT)
2010 static void sit_route_add(struct net_device *dev)
2011 {
2012 struct fib6_config cfg = {
2013 .fc_table = RT6_TABLE_MAIN,
2014 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2015 .fc_ifindex = dev->ifindex,
2016 .fc_dst_len = 96,
2017 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2018 .fc_nlinfo.nl_net = dev_net(dev),
2019 };
2020
2021 /* prefix length - 96 bits "::d.d.d.d" */
2022 ip6_route_add(&cfg);
2023 }
2024 #endif
2025
2026 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2027 {
2028 struct inet6_dev *idev;
2029
2030 ASSERT_RTNL();
2031
2032 idev = ipv6_find_idev(dev);
2033 if (!idev)
2034 return ERR_PTR(-ENOBUFS);
2035
2036 if (idev->cnf.disable_ipv6)
2037 return ERR_PTR(-EACCES);
2038
2039 /* Add default multicast route */
2040 if (!(dev->flags & IFF_LOOPBACK))
2041 addrconf_add_mroute(dev);
2042
2043 return idev;
2044 }
2045
2046 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2047 {
2048 struct prefix_info *pinfo;
2049 __u32 valid_lft;
2050 __u32 prefered_lft;
2051 int addr_type;
2052 struct inet6_dev *in6_dev;
2053 struct net *net = dev_net(dev);
2054
2055 pinfo = (struct prefix_info *) opt;
2056
2057 if (len < sizeof(struct prefix_info)) {
2058 ADBG("addrconf: prefix option too short\n");
2059 return;
2060 }
2061
2062 /*
2063 * Validation checks ([ADDRCONF], page 19)
2064 */
2065
2066 addr_type = ipv6_addr_type(&pinfo->prefix);
2067
2068 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2069 return;
2070
2071 valid_lft = ntohl(pinfo->valid);
2072 prefered_lft = ntohl(pinfo->prefered);
2073
2074 if (prefered_lft > valid_lft) {
2075 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2076 return;
2077 }
2078
2079 in6_dev = in6_dev_get(dev);
2080
2081 if (in6_dev == NULL) {
2082 net_dbg_ratelimited("addrconf: device %s not configured\n",
2083 dev->name);
2084 return;
2085 }
2086
2087 /*
2088 * Two things going on here:
2089 * 1) Add routes for on-link prefixes
2090 * 2) Configure prefixes with the auto flag set
2091 */
2092
2093 if (pinfo->onlink) {
2094 struct rt6_info *rt;
2095 unsigned long rt_expires;
2096
2097 /* Avoid arithmetic overflow. Really, we could
2098 * save rt_expires in seconds, likely valid_lft,
2099 * but it would require division in fib gc, that it
2100 * not good.
2101 */
2102 if (HZ > USER_HZ)
2103 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2104 else
2105 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2106
2107 if (addrconf_finite_timeout(rt_expires))
2108 rt_expires *= HZ;
2109
2110 rt = addrconf_get_prefix_route(&pinfo->prefix,
2111 pinfo->prefix_len,
2112 dev,
2113 RTF_ADDRCONF | RTF_PREFIX_RT,
2114 RTF_GATEWAY | RTF_DEFAULT);
2115
2116 if (rt) {
2117 /* Autoconf prefix route */
2118 if (valid_lft == 0) {
2119 ip6_del_rt(rt);
2120 rt = NULL;
2121 } else if (addrconf_finite_timeout(rt_expires)) {
2122 /* not infinity */
2123 rt6_set_expires(rt, jiffies + rt_expires);
2124 } else {
2125 rt6_clean_expires(rt);
2126 }
2127 } else if (valid_lft) {
2128 clock_t expires = 0;
2129 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2130 if (addrconf_finite_timeout(rt_expires)) {
2131 /* not infinity */
2132 flags |= RTF_EXPIRES;
2133 expires = jiffies_to_clock_t(rt_expires);
2134 }
2135 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2136 dev, expires, flags);
2137 }
2138 ip6_rt_put(rt);
2139 }
2140
2141 /* Try to figure out our local address for this prefix */
2142
2143 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2144 struct inet6_ifaddr *ifp;
2145 struct in6_addr addr;
2146 int create = 0, update_lft = 0;
2147 bool tokenized = false;
2148
2149 if (pinfo->prefix_len == 64) {
2150 memcpy(&addr, &pinfo->prefix, 8);
2151
2152 if (!ipv6_addr_any(&in6_dev->token)) {
2153 read_lock_bh(&in6_dev->lock);
2154 memcpy(addr.s6_addr + 8,
2155 in6_dev->token.s6_addr + 8, 8);
2156 read_unlock_bh(&in6_dev->lock);
2157 tokenized = true;
2158 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2159 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2160 in6_dev_put(in6_dev);
2161 return;
2162 }
2163 goto ok;
2164 }
2165 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2166 pinfo->prefix_len);
2167 in6_dev_put(in6_dev);
2168 return;
2169
2170 ok:
2171
2172 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2173
2174 if (ifp == NULL && valid_lft) {
2175 int max_addresses = in6_dev->cnf.max_addresses;
2176 u32 addr_flags = 0;
2177
2178 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2179 if (in6_dev->cnf.optimistic_dad &&
2180 !net->ipv6.devconf_all->forwarding && sllao)
2181 addr_flags = IFA_F_OPTIMISTIC;
2182 #endif
2183
2184 /* Do not allow to create too much of autoconfigured
2185 * addresses; this would be too easy way to crash kernel.
2186 */
2187 if (!max_addresses ||
2188 ipv6_count_addresses(in6_dev) < max_addresses)
2189 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2190 pinfo->prefix_len,
2191 addr_type&IPV6_ADDR_SCOPE_MASK,
2192 addr_flags, valid_lft,
2193 prefered_lft);
2194
2195 if (IS_ERR_OR_NULL(ifp)) {
2196 in6_dev_put(in6_dev);
2197 return;
2198 }
2199
2200 update_lft = 0;
2201 create = 1;
2202 ifp->cstamp = jiffies;
2203 ifp->tokenized = tokenized;
2204 addrconf_dad_start(ifp);
2205 }
2206
2207 if (ifp) {
2208 int flags;
2209 unsigned long now;
2210 #ifdef CONFIG_IPV6_PRIVACY
2211 struct inet6_ifaddr *ift;
2212 #endif
2213 u32 stored_lft;
2214
2215 /* update lifetime (RFC2462 5.5.3 e) */
2216 spin_lock(&ifp->lock);
2217 now = jiffies;
2218 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2219 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2220 else
2221 stored_lft = 0;
2222 if (!update_lft && !create && stored_lft) {
2223 const u32 minimum_lft = min(
2224 stored_lft, (u32)MIN_VALID_LIFETIME);
2225 valid_lft = max(valid_lft, minimum_lft);
2226
2227 /* RFC4862 Section 5.5.3e:
2228 * "Note that the preferred lifetime of the
2229 * corresponding address is always reset to
2230 * the Preferred Lifetime in the received
2231 * Prefix Information option, regardless of
2232 * whether the valid lifetime is also reset or
2233 * ignored."
2234 *
2235 * So we should always update prefered_lft here.
2236 */
2237 update_lft = 1;
2238 }
2239
2240 if (update_lft) {
2241 ifp->valid_lft = valid_lft;
2242 ifp->prefered_lft = prefered_lft;
2243 ifp->tstamp = now;
2244 flags = ifp->flags;
2245 ifp->flags &= ~IFA_F_DEPRECATED;
2246 spin_unlock(&ifp->lock);
2247
2248 if (!(flags&IFA_F_TENTATIVE))
2249 ipv6_ifa_notify(0, ifp);
2250 } else
2251 spin_unlock(&ifp->lock);
2252
2253 #ifdef CONFIG_IPV6_PRIVACY
2254 read_lock_bh(&in6_dev->lock);
2255 /* update all temporary addresses in the list */
2256 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2257 tmp_list) {
2258 int age, max_valid, max_prefered;
2259
2260 if (ifp != ift->ifpub)
2261 continue;
2262
2263 /*
2264 * RFC 4941 section 3.3:
2265 * If a received option will extend the lifetime
2266 * of a public address, the lifetimes of
2267 * temporary addresses should be extended,
2268 * subject to the overall constraint that no
2269 * temporary addresses should ever remain
2270 * "valid" or "preferred" for a time longer than
2271 * (TEMP_VALID_LIFETIME) or
2272 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2273 * respectively.
2274 */
2275 age = (now - ift->cstamp) / HZ;
2276 max_valid = in6_dev->cnf.temp_valid_lft - age;
2277 if (max_valid < 0)
2278 max_valid = 0;
2279
2280 max_prefered = in6_dev->cnf.temp_prefered_lft -
2281 in6_dev->cnf.max_desync_factor -
2282 age;
2283 if (max_prefered < 0)
2284 max_prefered = 0;
2285
2286 if (valid_lft > max_valid)
2287 valid_lft = max_valid;
2288
2289 if (prefered_lft > max_prefered)
2290 prefered_lft = max_prefered;
2291
2292 spin_lock(&ift->lock);
2293 flags = ift->flags;
2294 ift->valid_lft = valid_lft;
2295 ift->prefered_lft = prefered_lft;
2296 ift->tstamp = now;
2297 if (prefered_lft > 0)
2298 ift->flags &= ~IFA_F_DEPRECATED;
2299
2300 spin_unlock(&ift->lock);
2301 if (!(flags&IFA_F_TENTATIVE))
2302 ipv6_ifa_notify(0, ift);
2303 }
2304
2305 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2306 /*
2307 * When a new public address is created as
2308 * described in [ADDRCONF], also create a new
2309 * temporary address. Also create a temporary
2310 * address if it's enabled but no temporary
2311 * address currently exists.
2312 */
2313 read_unlock_bh(&in6_dev->lock);
2314 ipv6_create_tempaddr(ifp, NULL);
2315 } else {
2316 read_unlock_bh(&in6_dev->lock);
2317 }
2318 #endif
2319 in6_ifa_put(ifp);
2320 addrconf_verify(0);
2321 }
2322 }
2323 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2324 in6_dev_put(in6_dev);
2325 }
2326
2327 /*
2328 * Set destination address.
2329 * Special case for SIT interfaces where we create a new "virtual"
2330 * device.
2331 */
2332 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2333 {
2334 struct in6_ifreq ireq;
2335 struct net_device *dev;
2336 int err = -EINVAL;
2337
2338 rtnl_lock();
2339
2340 err = -EFAULT;
2341 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2342 goto err_exit;
2343
2344 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2345
2346 err = -ENODEV;
2347 if (dev == NULL)
2348 goto err_exit;
2349
2350 #if IS_ENABLED(CONFIG_IPV6_SIT)
2351 if (dev->type == ARPHRD_SIT) {
2352 const struct net_device_ops *ops = dev->netdev_ops;
2353 struct ifreq ifr;
2354 struct ip_tunnel_parm p;
2355
2356 err = -EADDRNOTAVAIL;
2357 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2358 goto err_exit;
2359
2360 memset(&p, 0, sizeof(p));
2361 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2362 p.iph.saddr = 0;
2363 p.iph.version = 4;
2364 p.iph.ihl = 5;
2365 p.iph.protocol = IPPROTO_IPV6;
2366 p.iph.ttl = 64;
2367 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2368
2369 if (ops->ndo_do_ioctl) {
2370 mm_segment_t oldfs = get_fs();
2371
2372 set_fs(KERNEL_DS);
2373 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2374 set_fs(oldfs);
2375 } else
2376 err = -EOPNOTSUPP;
2377
2378 if (err == 0) {
2379 err = -ENOBUFS;
2380 dev = __dev_get_by_name(net, p.name);
2381 if (!dev)
2382 goto err_exit;
2383 err = dev_open(dev);
2384 }
2385 }
2386 #endif
2387
2388 err_exit:
2389 rtnl_unlock();
2390 return err;
2391 }
2392
2393 /*
2394 * Manual configuration of address on an interface
2395 */
2396 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2397 const struct in6_addr *peer_pfx,
2398 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2399 __u32 valid_lft)
2400 {
2401 struct inet6_ifaddr *ifp;
2402 struct inet6_dev *idev;
2403 struct net_device *dev;
2404 int scope;
2405 u32 flags;
2406 clock_t expires;
2407 unsigned long timeout;
2408
2409 ASSERT_RTNL();
2410
2411 if (plen > 128)
2412 return -EINVAL;
2413
2414 /* check the lifetime */
2415 if (!valid_lft || prefered_lft > valid_lft)
2416 return -EINVAL;
2417
2418 dev = __dev_get_by_index(net, ifindex);
2419 if (!dev)
2420 return -ENODEV;
2421
2422 idev = addrconf_add_dev(dev);
2423 if (IS_ERR(idev))
2424 return PTR_ERR(idev);
2425
2426 scope = ipv6_addr_scope(pfx);
2427
2428 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2429 if (addrconf_finite_timeout(timeout)) {
2430 expires = jiffies_to_clock_t(timeout * HZ);
2431 valid_lft = timeout;
2432 flags = RTF_EXPIRES;
2433 } else {
2434 expires = 0;
2435 flags = 0;
2436 ifa_flags |= IFA_F_PERMANENT;
2437 }
2438
2439 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2440 if (addrconf_finite_timeout(timeout)) {
2441 if (timeout == 0)
2442 ifa_flags |= IFA_F_DEPRECATED;
2443 prefered_lft = timeout;
2444 }
2445
2446 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2447 valid_lft, prefered_lft);
2448
2449 if (!IS_ERR(ifp)) {
2450 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2451 expires, flags);
2452 /*
2453 * Note that section 3.1 of RFC 4429 indicates
2454 * that the Optimistic flag should not be set for
2455 * manually configured addresses
2456 */
2457 addrconf_dad_start(ifp);
2458 in6_ifa_put(ifp);
2459 addrconf_verify(0);
2460 return 0;
2461 }
2462
2463 return PTR_ERR(ifp);
2464 }
2465
2466 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2467 unsigned int plen)
2468 {
2469 struct inet6_ifaddr *ifp;
2470 struct inet6_dev *idev;
2471 struct net_device *dev;
2472
2473 if (plen > 128)
2474 return -EINVAL;
2475
2476 dev = __dev_get_by_index(net, ifindex);
2477 if (!dev)
2478 return -ENODEV;
2479
2480 if ((idev = __in6_dev_get(dev)) == NULL)
2481 return -ENXIO;
2482
2483 read_lock_bh(&idev->lock);
2484 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2485 if (ifp->prefix_len == plen &&
2486 ipv6_addr_equal(pfx, &ifp->addr)) {
2487 in6_ifa_hold(ifp);
2488 read_unlock_bh(&idev->lock);
2489
2490 ipv6_del_addr(ifp);
2491 return 0;
2492 }
2493 }
2494 read_unlock_bh(&idev->lock);
2495 return -EADDRNOTAVAIL;
2496 }
2497
2498
2499 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2500 {
2501 struct in6_ifreq ireq;
2502 int err;
2503
2504 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2505 return -EPERM;
2506
2507 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2508 return -EFAULT;
2509
2510 rtnl_lock();
2511 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2512 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2513 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2514 rtnl_unlock();
2515 return err;
2516 }
2517
2518 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2519 {
2520 struct in6_ifreq ireq;
2521 int err;
2522
2523 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2524 return -EPERM;
2525
2526 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2527 return -EFAULT;
2528
2529 rtnl_lock();
2530 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2531 ireq.ifr6_prefixlen);
2532 rtnl_unlock();
2533 return err;
2534 }
2535
2536 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2537 int plen, int scope)
2538 {
2539 struct inet6_ifaddr *ifp;
2540
2541 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2542 scope, IFA_F_PERMANENT, 0, 0);
2543 if (!IS_ERR(ifp)) {
2544 spin_lock_bh(&ifp->lock);
2545 ifp->flags &= ~IFA_F_TENTATIVE;
2546 spin_unlock_bh(&ifp->lock);
2547 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2548 in6_ifa_put(ifp);
2549 }
2550 }
2551
2552 #if IS_ENABLED(CONFIG_IPV6_SIT)
2553 static void sit_add_v4_addrs(struct inet6_dev *idev)
2554 {
2555 struct in6_addr addr;
2556 struct net_device *dev;
2557 struct net *net = dev_net(idev->dev);
2558 int scope;
2559
2560 ASSERT_RTNL();
2561
2562 memset(&addr, 0, sizeof(struct in6_addr));
2563 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2564
2565 if (idev->dev->flags&IFF_POINTOPOINT) {
2566 addr.s6_addr32[0] = htonl(0xfe800000);
2567 scope = IFA_LINK;
2568 } else {
2569 scope = IPV6_ADDR_COMPATv4;
2570 }
2571
2572 if (addr.s6_addr32[3]) {
2573 add_addr(idev, &addr, 128, scope);
2574 return;
2575 }
2576
2577 for_each_netdev(net, dev) {
2578 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2579 if (in_dev && (dev->flags & IFF_UP)) {
2580 struct in_ifaddr *ifa;
2581
2582 int flag = scope;
2583
2584 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2585 int plen;
2586
2587 addr.s6_addr32[3] = ifa->ifa_local;
2588
2589 if (ifa->ifa_scope == RT_SCOPE_LINK)
2590 continue;
2591 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2592 if (idev->dev->flags&IFF_POINTOPOINT)
2593 continue;
2594 flag |= IFA_HOST;
2595 }
2596 if (idev->dev->flags&IFF_POINTOPOINT)
2597 plen = 64;
2598 else
2599 plen = 96;
2600
2601 add_addr(idev, &addr, plen, flag);
2602 }
2603 }
2604 }
2605 }
2606 #endif
2607
2608 static void init_loopback(struct net_device *dev)
2609 {
2610 struct inet6_dev *idev;
2611 struct net_device *sp_dev;
2612 struct inet6_ifaddr *sp_ifa;
2613 struct rt6_info *sp_rt;
2614
2615 /* ::1 */
2616
2617 ASSERT_RTNL();
2618
2619 if ((idev = ipv6_find_idev(dev)) == NULL) {
2620 pr_debug("%s: add_dev failed\n", __func__);
2621 return;
2622 }
2623
2624 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2625
2626 /* Add routes to other interface's IPv6 addresses */
2627 for_each_netdev(dev_net(dev), sp_dev) {
2628 if (!strcmp(sp_dev->name, dev->name))
2629 continue;
2630
2631 idev = __in6_dev_get(sp_dev);
2632 if (!idev)
2633 continue;
2634
2635 read_lock_bh(&idev->lock);
2636 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2637
2638 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2639 continue;
2640
2641 if (sp_ifa->rt)
2642 continue;
2643
2644 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2645
2646 /* Failure cases are ignored */
2647 if (!IS_ERR(sp_rt)) {
2648 sp_ifa->rt = sp_rt;
2649 ip6_ins_rt(sp_rt);
2650 }
2651 }
2652 read_unlock_bh(&idev->lock);
2653 }
2654 }
2655
2656 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2657 {
2658 struct inet6_ifaddr *ifp;
2659 u32 addr_flags = IFA_F_PERMANENT;
2660
2661 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2662 if (idev->cnf.optimistic_dad &&
2663 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2664 addr_flags |= IFA_F_OPTIMISTIC;
2665 #endif
2666
2667
2668 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2669 if (!IS_ERR(ifp)) {
2670 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2671 addrconf_dad_start(ifp);
2672 in6_ifa_put(ifp);
2673 }
2674 }
2675
2676 static void addrconf_dev_config(struct net_device *dev)
2677 {
2678 struct in6_addr addr;
2679 struct inet6_dev *idev;
2680
2681 ASSERT_RTNL();
2682
2683 if ((dev->type != ARPHRD_ETHER) &&
2684 (dev->type != ARPHRD_FDDI) &&
2685 (dev->type != ARPHRD_ARCNET) &&
2686 (dev->type != ARPHRD_INFINIBAND) &&
2687 (dev->type != ARPHRD_IEEE802154) &&
2688 (dev->type != ARPHRD_IEEE1394) &&
2689 (dev->type != ARPHRD_TUNNEL6)) {
2690 /* Alas, we support only Ethernet autoconfiguration. */
2691 return;
2692 }
2693
2694 idev = addrconf_add_dev(dev);
2695 if (IS_ERR(idev))
2696 return;
2697
2698 memset(&addr, 0, sizeof(struct in6_addr));
2699 addr.s6_addr32[0] = htonl(0xFE800000);
2700
2701 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2702 addrconf_add_linklocal(idev, &addr);
2703 }
2704
2705 #if IS_ENABLED(CONFIG_IPV6_SIT)
2706 static void addrconf_sit_config(struct net_device *dev)
2707 {
2708 struct inet6_dev *idev;
2709
2710 ASSERT_RTNL();
2711
2712 /*
2713 * Configure the tunnel with one of our IPv4
2714 * addresses... we should configure all of
2715 * our v4 addrs in the tunnel
2716 */
2717
2718 if ((idev = ipv6_find_idev(dev)) == NULL) {
2719 pr_debug("%s: add_dev failed\n", __func__);
2720 return;
2721 }
2722
2723 if (dev->priv_flags & IFF_ISATAP) {
2724 struct in6_addr addr;
2725
2726 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2727 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2728 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2729 addrconf_add_linklocal(idev, &addr);
2730 return;
2731 }
2732
2733 sit_add_v4_addrs(idev);
2734
2735 if (dev->flags&IFF_POINTOPOINT)
2736 addrconf_add_mroute(dev);
2737 else
2738 sit_route_add(dev);
2739 }
2740 #endif
2741
2742 #if IS_ENABLED(CONFIG_NET_IPGRE)
2743 static void addrconf_gre_config(struct net_device *dev)
2744 {
2745 struct inet6_dev *idev;
2746 struct in6_addr addr;
2747
2748 ASSERT_RTNL();
2749
2750 if ((idev = ipv6_find_idev(dev)) == NULL) {
2751 pr_debug("%s: add_dev failed\n", __func__);
2752 return;
2753 }
2754
2755 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2756 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2757
2758 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2759 addrconf_add_linklocal(idev, &addr);
2760 }
2761 #endif
2762
2763 static inline int
2764 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2765 {
2766 struct in6_addr lladdr;
2767
2768 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2769 addrconf_add_linklocal(idev, &lladdr);
2770 return 0;
2771 }
2772 return -1;
2773 }
2774
2775 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2776 void *ptr)
2777 {
2778 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2779 struct inet6_dev *idev = __in6_dev_get(dev);
2780 int run_pending = 0;
2781 int err;
2782
2783 switch (event) {
2784 case NETDEV_REGISTER:
2785 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2786 idev = ipv6_add_dev(dev);
2787 if (!idev)
2788 return notifier_from_errno(-ENOMEM);
2789 }
2790 break;
2791
2792 case NETDEV_UP:
2793 case NETDEV_CHANGE:
2794 if (dev->flags & IFF_SLAVE)
2795 break;
2796
2797 if (event == NETDEV_UP) {
2798 if (!addrconf_qdisc_ok(dev)) {
2799 /* device is not ready yet. */
2800 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2801 dev->name);
2802 break;
2803 }
2804
2805 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2806 idev = ipv6_add_dev(dev);
2807
2808 if (idev) {
2809 idev->if_flags |= IF_READY;
2810 run_pending = 1;
2811 }
2812 } else {
2813 if (!addrconf_qdisc_ok(dev)) {
2814 /* device is still not ready. */
2815 break;
2816 }
2817
2818 if (idev) {
2819 if (idev->if_flags & IF_READY)
2820 /* device is already configured. */
2821 break;
2822 idev->if_flags |= IF_READY;
2823 }
2824
2825 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2826 dev->name);
2827
2828 run_pending = 1;
2829 }
2830
2831 switch (dev->type) {
2832 #if IS_ENABLED(CONFIG_IPV6_SIT)
2833 case ARPHRD_SIT:
2834 addrconf_sit_config(dev);
2835 break;
2836 #endif
2837 #if IS_ENABLED(CONFIG_NET_IPGRE)
2838 case ARPHRD_IPGRE:
2839 addrconf_gre_config(dev);
2840 break;
2841 #endif
2842 case ARPHRD_LOOPBACK:
2843 init_loopback(dev);
2844 break;
2845
2846 default:
2847 addrconf_dev_config(dev);
2848 break;
2849 }
2850
2851 if (idev) {
2852 if (run_pending)
2853 addrconf_dad_run(idev);
2854
2855 /*
2856 * If the MTU changed during the interface down,
2857 * when the interface up, the changed MTU must be
2858 * reflected in the idev as well as routers.
2859 */
2860 if (idev->cnf.mtu6 != dev->mtu &&
2861 dev->mtu >= IPV6_MIN_MTU) {
2862 rt6_mtu_change(dev, dev->mtu);
2863 idev->cnf.mtu6 = dev->mtu;
2864 }
2865 idev->tstamp = jiffies;
2866 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2867
2868 /*
2869 * If the changed mtu during down is lower than
2870 * IPV6_MIN_MTU stop IPv6 on this interface.
2871 */
2872 if (dev->mtu < IPV6_MIN_MTU)
2873 addrconf_ifdown(dev, 1);
2874 }
2875 break;
2876
2877 case NETDEV_CHANGEMTU:
2878 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2879 rt6_mtu_change(dev, dev->mtu);
2880 idev->cnf.mtu6 = dev->mtu;
2881 break;
2882 }
2883
2884 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2885 idev = ipv6_add_dev(dev);
2886 if (idev)
2887 break;
2888 }
2889
2890 /*
2891 * MTU falled under IPV6_MIN_MTU.
2892 * Stop IPv6 on this interface.
2893 */
2894
2895 case NETDEV_DOWN:
2896 case NETDEV_UNREGISTER:
2897 /*
2898 * Remove all addresses from this interface.
2899 */
2900 addrconf_ifdown(dev, event != NETDEV_DOWN);
2901 break;
2902
2903 case NETDEV_CHANGENAME:
2904 if (idev) {
2905 snmp6_unregister_dev(idev);
2906 addrconf_sysctl_unregister(idev);
2907 addrconf_sysctl_register(idev);
2908 err = snmp6_register_dev(idev);
2909 if (err)
2910 return notifier_from_errno(err);
2911 }
2912 break;
2913
2914 case NETDEV_PRE_TYPE_CHANGE:
2915 case NETDEV_POST_TYPE_CHANGE:
2916 addrconf_type_change(dev, event);
2917 break;
2918 }
2919
2920 return NOTIFY_OK;
2921 }
2922
2923 /*
2924 * addrconf module should be notified of a device going up
2925 */
2926 static struct notifier_block ipv6_dev_notf = {
2927 .notifier_call = addrconf_notify,
2928 };
2929
2930 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2931 {
2932 struct inet6_dev *idev;
2933 ASSERT_RTNL();
2934
2935 idev = __in6_dev_get(dev);
2936
2937 if (event == NETDEV_POST_TYPE_CHANGE)
2938 ipv6_mc_remap(idev);
2939 else if (event == NETDEV_PRE_TYPE_CHANGE)
2940 ipv6_mc_unmap(idev);
2941 }
2942
2943 static int addrconf_ifdown(struct net_device *dev, int how)
2944 {
2945 struct net *net = dev_net(dev);
2946 struct inet6_dev *idev;
2947 struct inet6_ifaddr *ifa;
2948 int state, i;
2949
2950 ASSERT_RTNL();
2951
2952 rt6_ifdown(net, dev);
2953 neigh_ifdown(&nd_tbl, dev);
2954
2955 idev = __in6_dev_get(dev);
2956 if (idev == NULL)
2957 return -ENODEV;
2958
2959 /*
2960 * Step 1: remove reference to ipv6 device from parent device.
2961 * Do not dev_put!
2962 */
2963 if (how) {
2964 idev->dead = 1;
2965
2966 /* protected by rtnl_lock */
2967 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2968
2969 /* Step 1.5: remove snmp6 entry */
2970 snmp6_unregister_dev(idev);
2971
2972 }
2973
2974 /* Step 2: clear hash table */
2975 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2976 struct hlist_head *h = &inet6_addr_lst[i];
2977
2978 spin_lock_bh(&addrconf_hash_lock);
2979 restart:
2980 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2981 if (ifa->idev == idev) {
2982 hlist_del_init_rcu(&ifa->addr_lst);
2983 addrconf_del_dad_timer(ifa);
2984 goto restart;
2985 }
2986 }
2987 spin_unlock_bh(&addrconf_hash_lock);
2988 }
2989
2990 write_lock_bh(&idev->lock);
2991
2992 addrconf_del_rs_timer(idev);
2993
2994 /* Step 2: clear flags for stateless addrconf */
2995 if (!how)
2996 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2997
2998 #ifdef CONFIG_IPV6_PRIVACY
2999 if (how && del_timer(&idev->regen_timer))
3000 in6_dev_put(idev);
3001
3002 /* Step 3: clear tempaddr list */
3003 while (!list_empty(&idev->tempaddr_list)) {
3004 ifa = list_first_entry(&idev->tempaddr_list,
3005 struct inet6_ifaddr, tmp_list);
3006 list_del(&ifa->tmp_list);
3007 write_unlock_bh(&idev->lock);
3008 spin_lock_bh(&ifa->lock);
3009
3010 if (ifa->ifpub) {
3011 in6_ifa_put(ifa->ifpub);
3012 ifa->ifpub = NULL;
3013 }
3014 spin_unlock_bh(&ifa->lock);
3015 in6_ifa_put(ifa);
3016 write_lock_bh(&idev->lock);
3017 }
3018 #endif
3019
3020 while (!list_empty(&idev->addr_list)) {
3021 ifa = list_first_entry(&idev->addr_list,
3022 struct inet6_ifaddr, if_list);
3023 addrconf_del_dad_timer(ifa);
3024
3025 list_del(&ifa->if_list);
3026
3027 write_unlock_bh(&idev->lock);
3028
3029 spin_lock_bh(&ifa->state_lock);
3030 state = ifa->state;
3031 ifa->state = INET6_IFADDR_STATE_DEAD;
3032 spin_unlock_bh(&ifa->state_lock);
3033
3034 if (state != INET6_IFADDR_STATE_DEAD) {
3035 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3036 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3037 }
3038 in6_ifa_put(ifa);
3039
3040 write_lock_bh(&idev->lock);
3041 }
3042
3043 write_unlock_bh(&idev->lock);
3044
3045 /* Step 5: Discard multicast list */
3046 if (how)
3047 ipv6_mc_destroy_dev(idev);
3048 else
3049 ipv6_mc_down(idev);
3050
3051 idev->tstamp = jiffies;
3052
3053 /* Last: Shot the device (if unregistered) */
3054 if (how) {
3055 addrconf_sysctl_unregister(idev);
3056 neigh_parms_release(&nd_tbl, idev->nd_parms);
3057 neigh_ifdown(&nd_tbl, dev);
3058 in6_dev_put(idev);
3059 }
3060 return 0;
3061 }
3062
3063 static void addrconf_rs_timer(unsigned long data)
3064 {
3065 struct inet6_dev *idev = (struct inet6_dev *)data;
3066 struct net_device *dev = idev->dev;
3067 struct in6_addr lladdr;
3068
3069 write_lock(&idev->lock);
3070 if (idev->dead || !(idev->if_flags & IF_READY))
3071 goto out;
3072
3073 if (!ipv6_accept_ra(idev))
3074 goto out;
3075
3076 /* Announcement received after solicitation was sent */
3077 if (idev->if_flags & IF_RA_RCVD)
3078 goto out;
3079
3080 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3081 write_unlock(&idev->lock);
3082 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3083 ndisc_send_rs(dev, &lladdr,
3084 &in6addr_linklocal_allrouters);
3085 else
3086 goto put;
3087
3088 write_lock(&idev->lock);
3089 /* The wait after the last probe can be shorter */
3090 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3091 idev->cnf.rtr_solicits) ?
3092 idev->cnf.rtr_solicit_delay :
3093 idev->cnf.rtr_solicit_interval);
3094 } else {
3095 /*
3096 * Note: we do not support deprecated "all on-link"
3097 * assumption any longer.
3098 */
3099 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3100 }
3101
3102 out:
3103 write_unlock(&idev->lock);
3104 put:
3105 in6_dev_put(idev);
3106 }
3107
3108 /*
3109 * Duplicate Address Detection
3110 */
3111 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3112 {
3113 unsigned long rand_num;
3114 struct inet6_dev *idev = ifp->idev;
3115
3116 if (ifp->flags & IFA_F_OPTIMISTIC)
3117 rand_num = 0;
3118 else
3119 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3120
3121 ifp->dad_probes = idev->cnf.dad_transmits;
3122 addrconf_mod_dad_timer(ifp, rand_num);
3123 }
3124
3125 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3126 {
3127 struct inet6_dev *idev = ifp->idev;
3128 struct net_device *dev = idev->dev;
3129
3130 addrconf_join_solict(dev, &ifp->addr);
3131
3132 net_srandom(ifp->addr.s6_addr32[3]);
3133
3134 read_lock_bh(&idev->lock);
3135 spin_lock(&ifp->lock);
3136 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3137 goto out;
3138
3139 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3140 idev->cnf.accept_dad < 1 ||
3141 !(ifp->flags&IFA_F_TENTATIVE) ||
3142 ifp->flags & IFA_F_NODAD) {
3143 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3144 spin_unlock(&ifp->lock);
3145 read_unlock_bh(&idev->lock);
3146
3147 addrconf_dad_completed(ifp);
3148 return;
3149 }
3150
3151 if (!(idev->if_flags & IF_READY)) {
3152 spin_unlock(&ifp->lock);
3153 read_unlock_bh(&idev->lock);
3154 /*
3155 * If the device is not ready:
3156 * - keep it tentative if it is a permanent address.
3157 * - otherwise, kill it.
3158 */
3159 in6_ifa_hold(ifp);
3160 addrconf_dad_stop(ifp, 0);
3161 return;
3162 }
3163
3164 /*
3165 * Optimistic nodes can start receiving
3166 * Frames right away
3167 */
3168 if (ifp->flags & IFA_F_OPTIMISTIC)
3169 ip6_ins_rt(ifp->rt);
3170
3171 addrconf_dad_kick(ifp);
3172 out:
3173 spin_unlock(&ifp->lock);
3174 read_unlock_bh(&idev->lock);
3175 }
3176
3177 static void addrconf_dad_timer(unsigned long data)
3178 {
3179 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3180 struct inet6_dev *idev = ifp->idev;
3181 struct in6_addr mcaddr;
3182
3183 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3184 goto out;
3185
3186 write_lock(&idev->lock);
3187 if (idev->dead || !(idev->if_flags & IF_READY)) {
3188 write_unlock(&idev->lock);
3189 goto out;
3190 }
3191
3192 spin_lock(&ifp->lock);
3193 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3194 spin_unlock(&ifp->lock);
3195 write_unlock(&idev->lock);
3196 goto out;
3197 }
3198
3199 if (ifp->dad_probes == 0) {
3200 /*
3201 * DAD was successful
3202 */
3203
3204 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3205 spin_unlock(&ifp->lock);
3206 write_unlock(&idev->lock);
3207
3208 addrconf_dad_completed(ifp);
3209
3210 goto out;
3211 }
3212
3213 ifp->dad_probes--;
3214 addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3215 spin_unlock(&ifp->lock);
3216 write_unlock(&idev->lock);
3217
3218 /* send a neighbour solicitation for our addr */
3219 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3220 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3221 out:
3222 in6_ifa_put(ifp);
3223 }
3224
3225 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3226 {
3227 struct net_device *dev = ifp->idev->dev;
3228 struct in6_addr lladdr;
3229 bool send_rs, send_mld;
3230
3231 addrconf_del_dad_timer(ifp);
3232
3233 /*
3234 * Configure the address for reception. Now it is valid.
3235 */
3236
3237 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3238
3239 /* If added prefix is link local and we are prepared to process
3240 router advertisements, start sending router solicitations.
3241 */
3242
3243 read_lock_bh(&ifp->idev->lock);
3244 spin_lock(&ifp->lock);
3245 send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3246 ifp->idev->valid_ll_addr_cnt == 1;
3247 send_rs = send_mld &&
3248 ipv6_accept_ra(ifp->idev) &&
3249 ifp->idev->cnf.rtr_solicits > 0 &&
3250 (dev->flags&IFF_LOOPBACK) == 0;
3251 spin_unlock(&ifp->lock);
3252 read_unlock_bh(&ifp->idev->lock);
3253
3254 /* While dad is in progress mld report's source address is in6_addrany.
3255 * Resend with proper ll now.
3256 */
3257 if (send_mld)
3258 ipv6_mc_dad_complete(ifp->idev);
3259
3260 if (send_rs) {
3261 /*
3262 * If a host as already performed a random delay
3263 * [...] as part of DAD [...] there is no need
3264 * to delay again before sending the first RS
3265 */
3266 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3267 return;
3268 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3269
3270 write_lock_bh(&ifp->idev->lock);
3271 spin_lock(&ifp->lock);
3272 ifp->idev->rs_probes = 1;
3273 ifp->idev->if_flags |= IF_RS_SENT;
3274 addrconf_mod_rs_timer(ifp->idev,
3275 ifp->idev->cnf.rtr_solicit_interval);
3276 spin_unlock(&ifp->lock);
3277 write_unlock_bh(&ifp->idev->lock);
3278 }
3279 }
3280
3281 static void addrconf_dad_run(struct inet6_dev *idev)
3282 {
3283 struct inet6_ifaddr *ifp;
3284
3285 read_lock_bh(&idev->lock);
3286 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3287 spin_lock(&ifp->lock);
3288 if (ifp->flags & IFA_F_TENTATIVE &&
3289 ifp->state == INET6_IFADDR_STATE_DAD)
3290 addrconf_dad_kick(ifp);
3291 spin_unlock(&ifp->lock);
3292 }
3293 read_unlock_bh(&idev->lock);
3294 }
3295
3296 #ifdef CONFIG_PROC_FS
3297 struct if6_iter_state {
3298 struct seq_net_private p;
3299 int bucket;
3300 int offset;
3301 };
3302
3303 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3304 {
3305 struct inet6_ifaddr *ifa = NULL;
3306 struct if6_iter_state *state = seq->private;
3307 struct net *net = seq_file_net(seq);
3308 int p = 0;
3309
3310 /* initial bucket if pos is 0 */
3311 if (pos == 0) {
3312 state->bucket = 0;
3313 state->offset = 0;
3314 }
3315
3316 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3317 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3318 addr_lst) {
3319 if (!net_eq(dev_net(ifa->idev->dev), net))
3320 continue;
3321 /* sync with offset */
3322 if (p < state->offset) {
3323 p++;
3324 continue;
3325 }
3326 state->offset++;
3327 return ifa;
3328 }
3329
3330 /* prepare for next bucket */
3331 state->offset = 0;
3332 p = 0;
3333 }
3334 return NULL;
3335 }
3336
3337 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3338 struct inet6_ifaddr *ifa)
3339 {
3340 struct if6_iter_state *state = seq->private;
3341 struct net *net = seq_file_net(seq);
3342
3343 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3344 if (!net_eq(dev_net(ifa->idev->dev), net))
3345 continue;
3346 state->offset++;
3347 return ifa;
3348 }
3349
3350 while (++state->bucket < IN6_ADDR_HSIZE) {
3351 state->offset = 0;
3352 hlist_for_each_entry_rcu_bh(ifa,
3353 &inet6_addr_lst[state->bucket], addr_lst) {
3354 if (!net_eq(dev_net(ifa->idev->dev), net))
3355 continue;
3356 state->offset++;
3357 return ifa;
3358 }
3359 }
3360
3361 return NULL;
3362 }
3363
3364 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3365 __acquires(rcu_bh)
3366 {
3367 rcu_read_lock_bh();
3368 return if6_get_first(seq, *pos);
3369 }
3370
3371 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3372 {
3373 struct inet6_ifaddr *ifa;
3374
3375 ifa = if6_get_next(seq, v);
3376 ++*pos;
3377 return ifa;
3378 }
3379
3380 static void if6_seq_stop(struct seq_file *seq, void *v)
3381 __releases(rcu_bh)
3382 {
3383 rcu_read_unlock_bh();
3384 }
3385
3386 static int if6_seq_show(struct seq_file *seq, void *v)
3387 {
3388 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3389 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3390 &ifp->addr,
3391 ifp->idev->dev->ifindex,
3392 ifp->prefix_len,
3393 ifp->scope,
3394 ifp->flags,
3395 ifp->idev->dev->name);
3396 return 0;
3397 }
3398
3399 static const struct seq_operations if6_seq_ops = {
3400 .start = if6_seq_start,
3401 .next = if6_seq_next,
3402 .show = if6_seq_show,
3403 .stop = if6_seq_stop,
3404 };
3405
3406 static int if6_seq_open(struct inode *inode, struct file *file)
3407 {
3408 return seq_open_net(inode, file, &if6_seq_ops,
3409 sizeof(struct if6_iter_state));
3410 }
3411
3412 static const struct file_operations if6_fops = {
3413 .owner = THIS_MODULE,
3414 .open = if6_seq_open,
3415 .read = seq_read,
3416 .llseek = seq_lseek,
3417 .release = seq_release_net,
3418 };
3419
3420 static int __net_init if6_proc_net_init(struct net *net)
3421 {
3422 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3423 return -ENOMEM;
3424 return 0;
3425 }
3426
3427 static void __net_exit if6_proc_net_exit(struct net *net)
3428 {
3429 remove_proc_entry("if_inet6", net->proc_net);
3430 }
3431
3432 static struct pernet_operations if6_proc_net_ops = {
3433 .init = if6_proc_net_init,
3434 .exit = if6_proc_net_exit,
3435 };
3436
3437 int __init if6_proc_init(void)
3438 {
3439 return register_pernet_subsys(&if6_proc_net_ops);
3440 }
3441
3442 void if6_proc_exit(void)
3443 {
3444 unregister_pernet_subsys(&if6_proc_net_ops);
3445 }
3446 #endif /* CONFIG_PROC_FS */
3447
3448 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3449 /* Check if address is a home address configured on any interface. */
3450 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3451 {
3452 int ret = 0;
3453 struct inet6_ifaddr *ifp = NULL;
3454 unsigned int hash = inet6_addr_hash(addr);
3455
3456 rcu_read_lock_bh();
3457 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3458 if (!net_eq(dev_net(ifp->idev->dev), net))
3459 continue;
3460 if (ipv6_addr_equal(&ifp->addr, addr) &&
3461 (ifp->flags & IFA_F_HOMEADDRESS)) {
3462 ret = 1;
3463 break;
3464 }
3465 }
3466 rcu_read_unlock_bh();
3467 return ret;
3468 }
3469 #endif
3470
3471 /*
3472 * Periodic address status verification
3473 */
3474
3475 static void addrconf_verify(unsigned long foo)
3476 {
3477 unsigned long now, next, next_sec, next_sched;
3478 struct inet6_ifaddr *ifp;
3479 int i;
3480
3481 rcu_read_lock_bh();
3482 spin_lock(&addrconf_verify_lock);
3483 now = jiffies;
3484 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3485
3486 del_timer(&addr_chk_timer);
3487
3488 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3489 restart:
3490 hlist_for_each_entry_rcu_bh(ifp,
3491 &inet6_addr_lst[i], addr_lst) {
3492 unsigned long age;
3493
3494 if (ifp->flags & IFA_F_PERMANENT)
3495 continue;
3496
3497 spin_lock(&ifp->lock);
3498 /* We try to batch several events at once. */
3499 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3500
3501 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3502 age >= ifp->valid_lft) {
3503 spin_unlock(&ifp->lock);
3504 in6_ifa_hold(ifp);
3505 ipv6_del_addr(ifp);
3506 goto restart;
3507 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3508 spin_unlock(&ifp->lock);
3509 continue;
3510 } else if (age >= ifp->prefered_lft) {
3511 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3512 int deprecate = 0;
3513
3514 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3515 deprecate = 1;
3516 ifp->flags |= IFA_F_DEPRECATED;
3517 }
3518
3519 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3520 next = ifp->tstamp + ifp->valid_lft * HZ;
3521
3522 spin_unlock(&ifp->lock);
3523
3524 if (deprecate) {
3525 in6_ifa_hold(ifp);
3526
3527 ipv6_ifa_notify(0, ifp);
3528 in6_ifa_put(ifp);
3529 goto restart;
3530 }
3531 #ifdef CONFIG_IPV6_PRIVACY
3532 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3533 !(ifp->flags&IFA_F_TENTATIVE)) {
3534 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3535 ifp->idev->cnf.dad_transmits *
3536 ifp->idev->nd_parms->retrans_time / HZ;
3537
3538 if (age >= ifp->prefered_lft - regen_advance) {
3539 struct inet6_ifaddr *ifpub = ifp->ifpub;
3540 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3541 next = ifp->tstamp + ifp->prefered_lft * HZ;
3542 if (!ifp->regen_count && ifpub) {
3543 ifp->regen_count++;
3544 in6_ifa_hold(ifp);
3545 in6_ifa_hold(ifpub);
3546 spin_unlock(&ifp->lock);
3547
3548 spin_lock(&ifpub->lock);
3549 ifpub->regen_count = 0;
3550 spin_unlock(&ifpub->lock);
3551 ipv6_create_tempaddr(ifpub, ifp);
3552 in6_ifa_put(ifpub);
3553 in6_ifa_put(ifp);
3554 goto restart;
3555 }
3556 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3557 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3558 spin_unlock(&ifp->lock);
3559 #endif
3560 } else {
3561 /* ifp->prefered_lft <= ifp->valid_lft */
3562 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3563 next = ifp->tstamp + ifp->prefered_lft * HZ;
3564 spin_unlock(&ifp->lock);
3565 }
3566 }
3567 }
3568
3569 next_sec = round_jiffies_up(next);
3570 next_sched = next;
3571
3572 /* If rounded timeout is accurate enough, accept it. */
3573 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3574 next_sched = next_sec;
3575
3576 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3577 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3578 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3579
3580 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3581 now, next, next_sec, next_sched);
3582
3583 addr_chk_timer.expires = next_sched;
3584 add_timer(&addr_chk_timer);
3585 spin_unlock(&addrconf_verify_lock);
3586 rcu_read_unlock_bh();
3587 }
3588
3589 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3590 struct in6_addr **peer_pfx)
3591 {
3592 struct in6_addr *pfx = NULL;
3593
3594 *peer_pfx = NULL;
3595
3596 if (addr)
3597 pfx = nla_data(addr);
3598
3599 if (local) {
3600 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3601 *peer_pfx = pfx;
3602 pfx = nla_data(local);
3603 }
3604
3605 return pfx;
3606 }
3607
3608 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3609 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3610 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3611 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3612 };
3613
3614 static int
3615 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3616 {
3617 struct net *net = sock_net(skb->sk);
3618 struct ifaddrmsg *ifm;
3619 struct nlattr *tb[IFA_MAX+1];
3620 struct in6_addr *pfx, *peer_pfx;
3621 int err;
3622
3623 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3624 if (err < 0)
3625 return err;
3626
3627 ifm = nlmsg_data(nlh);
3628 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3629 if (pfx == NULL)
3630 return -EINVAL;
3631
3632 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3633 }
3634
3635 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3636 u32 prefered_lft, u32 valid_lft)
3637 {
3638 u32 flags;
3639 clock_t expires;
3640 unsigned long timeout;
3641
3642 if (!valid_lft || (prefered_lft > valid_lft))
3643 return -EINVAL;
3644
3645 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3646 if (addrconf_finite_timeout(timeout)) {
3647 expires = jiffies_to_clock_t(timeout * HZ);
3648 valid_lft = timeout;
3649 flags = RTF_EXPIRES;
3650 } else {
3651 expires = 0;
3652 flags = 0;
3653 ifa_flags |= IFA_F_PERMANENT;
3654 }
3655
3656 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3657 if (addrconf_finite_timeout(timeout)) {
3658 if (timeout == 0)
3659 ifa_flags |= IFA_F_DEPRECATED;
3660 prefered_lft = timeout;
3661 }
3662
3663 spin_lock_bh(&ifp->lock);
3664 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3665 ifp->tstamp = jiffies;
3666 ifp->valid_lft = valid_lft;
3667 ifp->prefered_lft = prefered_lft;
3668
3669 spin_unlock_bh(&ifp->lock);
3670 if (!(ifp->flags&IFA_F_TENTATIVE))
3671 ipv6_ifa_notify(0, ifp);
3672
3673 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3674 expires, flags);
3675 addrconf_verify(0);
3676
3677 return 0;
3678 }
3679
3680 static int
3681 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3682 {
3683 struct net *net = sock_net(skb->sk);
3684 struct ifaddrmsg *ifm;
3685 struct nlattr *tb[IFA_MAX+1];
3686 struct in6_addr *pfx, *peer_pfx;
3687 struct inet6_ifaddr *ifa;
3688 struct net_device *dev;
3689 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3690 u8 ifa_flags;
3691 int err;
3692
3693 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3694 if (err < 0)
3695 return err;
3696
3697 ifm = nlmsg_data(nlh);
3698 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3699 if (pfx == NULL)
3700 return -EINVAL;
3701
3702 if (tb[IFA_CACHEINFO]) {
3703 struct ifa_cacheinfo *ci;
3704
3705 ci = nla_data(tb[IFA_CACHEINFO]);
3706 valid_lft = ci->ifa_valid;
3707 preferred_lft = ci->ifa_prefered;
3708 } else {
3709 preferred_lft = INFINITY_LIFE_TIME;
3710 valid_lft = INFINITY_LIFE_TIME;
3711 }
3712
3713 dev = __dev_get_by_index(net, ifm->ifa_index);
3714 if (dev == NULL)
3715 return -ENODEV;
3716
3717 /* We ignore other flags so far. */
3718 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3719
3720 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3721 if (ifa == NULL) {
3722 /*
3723 * It would be best to check for !NLM_F_CREATE here but
3724 * userspace alreay relies on not having to provide this.
3725 */
3726 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3727 ifm->ifa_prefixlen, ifa_flags,
3728 preferred_lft, valid_lft);
3729 }
3730
3731 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3732 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3733 err = -EEXIST;
3734 else
3735 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3736
3737 in6_ifa_put(ifa);
3738
3739 return err;
3740 }
3741
3742 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3743 u8 scope, int ifindex)
3744 {
3745 struct ifaddrmsg *ifm;
3746
3747 ifm = nlmsg_data(nlh);
3748 ifm->ifa_family = AF_INET6;
3749 ifm->ifa_prefixlen = prefixlen;
3750 ifm->ifa_flags = flags;
3751 ifm->ifa_scope = scope;
3752 ifm->ifa_index = ifindex;
3753 }
3754
3755 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3756 unsigned long tstamp, u32 preferred, u32 valid)
3757 {
3758 struct ifa_cacheinfo ci;
3759
3760 ci.cstamp = cstamp_delta(cstamp);
3761 ci.tstamp = cstamp_delta(tstamp);
3762 ci.ifa_prefered = preferred;
3763 ci.ifa_valid = valid;
3764
3765 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3766 }
3767
3768 static inline int rt_scope(int ifa_scope)
3769 {
3770 if (ifa_scope & IFA_HOST)
3771 return RT_SCOPE_HOST;
3772 else if (ifa_scope & IFA_LINK)
3773 return RT_SCOPE_LINK;
3774 else if (ifa_scope & IFA_SITE)
3775 return RT_SCOPE_SITE;
3776 else
3777 return RT_SCOPE_UNIVERSE;
3778 }
3779
3780 static inline int inet6_ifaddr_msgsize(void)
3781 {
3782 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3783 + nla_total_size(16) /* IFA_LOCAL */
3784 + nla_total_size(16) /* IFA_ADDRESS */
3785 + nla_total_size(sizeof(struct ifa_cacheinfo));
3786 }
3787
3788 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3789 u32 portid, u32 seq, int event, unsigned int flags)
3790 {
3791 struct nlmsghdr *nlh;
3792 u32 preferred, valid;
3793
3794 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3795 if (nlh == NULL)
3796 return -EMSGSIZE;
3797
3798 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3799 ifa->idev->dev->ifindex);
3800
3801 if (!(ifa->flags&IFA_F_PERMANENT)) {
3802 preferred = ifa->prefered_lft;
3803 valid = ifa->valid_lft;
3804 if (preferred != INFINITY_LIFE_TIME) {
3805 long tval = (jiffies - ifa->tstamp)/HZ;
3806 if (preferred > tval)
3807 preferred -= tval;
3808 else
3809 preferred = 0;
3810 if (valid != INFINITY_LIFE_TIME) {
3811 if (valid > tval)
3812 valid -= tval;
3813 else
3814 valid = 0;
3815 }
3816 }
3817 } else {
3818 preferred = INFINITY_LIFE_TIME;
3819 valid = INFINITY_LIFE_TIME;
3820 }
3821
3822 if (!ipv6_addr_any(&ifa->peer_addr)) {
3823 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3824 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3825 goto error;
3826 } else
3827 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3828 goto error;
3829
3830 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3831 goto error;
3832
3833 return nlmsg_end(skb, nlh);
3834
3835 error:
3836 nlmsg_cancel(skb, nlh);
3837 return -EMSGSIZE;
3838 }
3839
3840 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3841 u32 portid, u32 seq, int event, u16 flags)
3842 {
3843 struct nlmsghdr *nlh;
3844 u8 scope = RT_SCOPE_UNIVERSE;
3845 int ifindex = ifmca->idev->dev->ifindex;
3846
3847 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3848 scope = RT_SCOPE_SITE;
3849
3850 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3851 if (nlh == NULL)
3852 return -EMSGSIZE;
3853
3854 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3855 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3856 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3857 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3858 nlmsg_cancel(skb, nlh);
3859 return -EMSGSIZE;
3860 }
3861
3862 return nlmsg_end(skb, nlh);
3863 }
3864
3865 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3866 u32 portid, u32 seq, int event, unsigned int flags)
3867 {
3868 struct nlmsghdr *nlh;
3869 u8 scope = RT_SCOPE_UNIVERSE;
3870 int ifindex = ifaca->aca_idev->dev->ifindex;
3871
3872 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3873 scope = RT_SCOPE_SITE;
3874
3875 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3876 if (nlh == NULL)
3877 return -EMSGSIZE;
3878
3879 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3880 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3881 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3882 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3883 nlmsg_cancel(skb, nlh);
3884 return -EMSGSIZE;
3885 }
3886
3887 return nlmsg_end(skb, nlh);
3888 }
3889
3890 enum addr_type_t {
3891 UNICAST_ADDR,
3892 MULTICAST_ADDR,
3893 ANYCAST_ADDR,
3894 };
3895
3896 /* called with rcu_read_lock() */
3897 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3898 struct netlink_callback *cb, enum addr_type_t type,
3899 int s_ip_idx, int *p_ip_idx)
3900 {
3901 struct ifmcaddr6 *ifmca;
3902 struct ifacaddr6 *ifaca;
3903 int err = 1;
3904 int ip_idx = *p_ip_idx;
3905
3906 read_lock_bh(&idev->lock);
3907 switch (type) {
3908 case UNICAST_ADDR: {
3909 struct inet6_ifaddr *ifa;
3910
3911 /* unicast address incl. temp addr */
3912 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3913 if (++ip_idx < s_ip_idx)
3914 continue;
3915 err = inet6_fill_ifaddr(skb, ifa,
3916 NETLINK_CB(cb->skb).portid,
3917 cb->nlh->nlmsg_seq,
3918 RTM_NEWADDR,
3919 NLM_F_MULTI);
3920 if (err <= 0)
3921 break;
3922 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3923 }
3924 break;
3925 }
3926 case MULTICAST_ADDR:
3927 /* multicast address */
3928 for (ifmca = idev->mc_list; ifmca;
3929 ifmca = ifmca->next, ip_idx++) {
3930 if (ip_idx < s_ip_idx)
3931 continue;
3932 err = inet6_fill_ifmcaddr(skb, ifmca,
3933 NETLINK_CB(cb->skb).portid,
3934 cb->nlh->nlmsg_seq,
3935 RTM_GETMULTICAST,
3936 NLM_F_MULTI);
3937 if (err <= 0)
3938 break;
3939 }
3940 break;
3941 case ANYCAST_ADDR:
3942 /* anycast address */
3943 for (ifaca = idev->ac_list; ifaca;
3944 ifaca = ifaca->aca_next, ip_idx++) {
3945 if (ip_idx < s_ip_idx)
3946 continue;
3947 err = inet6_fill_ifacaddr(skb, ifaca,
3948 NETLINK_CB(cb->skb).portid,
3949 cb->nlh->nlmsg_seq,
3950 RTM_GETANYCAST,
3951 NLM_F_MULTI);
3952 if (err <= 0)
3953 break;
3954 }
3955 break;
3956 default:
3957 break;
3958 }
3959 read_unlock_bh(&idev->lock);
3960 *p_ip_idx = ip_idx;
3961 return err;
3962 }
3963
3964 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3965 enum addr_type_t type)
3966 {
3967 struct net *net = sock_net(skb->sk);
3968 int h, s_h;
3969 int idx, ip_idx;
3970 int s_idx, s_ip_idx;
3971 struct net_device *dev;
3972 struct inet6_dev *idev;
3973 struct hlist_head *head;
3974
3975 s_h = cb->args[0];
3976 s_idx = idx = cb->args[1];
3977 s_ip_idx = ip_idx = cb->args[2];
3978
3979 rcu_read_lock();
3980 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3981 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3982 idx = 0;
3983 head = &net->dev_index_head[h];
3984 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3985 if (idx < s_idx)
3986 goto cont;
3987 if (h > s_h || idx > s_idx)
3988 s_ip_idx = 0;
3989 ip_idx = 0;
3990 idev = __in6_dev_get(dev);
3991 if (!idev)
3992 goto cont;
3993
3994 if (in6_dump_addrs(idev, skb, cb, type,
3995 s_ip_idx, &ip_idx) <= 0)
3996 goto done;
3997 cont:
3998 idx++;
3999 }
4000 }
4001 done:
4002 rcu_read_unlock();
4003 cb->args[0] = h;
4004 cb->args[1] = idx;
4005 cb->args[2] = ip_idx;
4006
4007 return skb->len;
4008 }
4009
4010 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4011 {
4012 enum addr_type_t type = UNICAST_ADDR;
4013
4014 return inet6_dump_addr(skb, cb, type);
4015 }
4016
4017 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4018 {
4019 enum addr_type_t type = MULTICAST_ADDR;
4020
4021 return inet6_dump_addr(skb, cb, type);
4022 }
4023
4024
4025 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4026 {
4027 enum addr_type_t type = ANYCAST_ADDR;
4028
4029 return inet6_dump_addr(skb, cb, type);
4030 }
4031
4032 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4033 {
4034 struct net *net = sock_net(in_skb->sk);
4035 struct ifaddrmsg *ifm;
4036 struct nlattr *tb[IFA_MAX+1];
4037 struct in6_addr *addr = NULL, *peer;
4038 struct net_device *dev = NULL;
4039 struct inet6_ifaddr *ifa;
4040 struct sk_buff *skb;
4041 int err;
4042
4043 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4044 if (err < 0)
4045 goto errout;
4046
4047 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4048 if (addr == NULL) {
4049 err = -EINVAL;
4050 goto errout;
4051 }
4052
4053 ifm = nlmsg_data(nlh);
4054 if (ifm->ifa_index)
4055 dev = __dev_get_by_index(net, ifm->ifa_index);
4056
4057 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4058 if (!ifa) {
4059 err = -EADDRNOTAVAIL;
4060 goto errout;
4061 }
4062
4063 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4064 if (!skb) {
4065 err = -ENOBUFS;
4066 goto errout_ifa;
4067 }
4068
4069 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4070 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4071 if (err < 0) {
4072 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4073 WARN_ON(err == -EMSGSIZE);
4074 kfree_skb(skb);
4075 goto errout_ifa;
4076 }
4077 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4078 errout_ifa:
4079 in6_ifa_put(ifa);
4080 errout:
4081 return err;
4082 }
4083
4084 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4085 {
4086 struct sk_buff *skb;
4087 struct net *net = dev_net(ifa->idev->dev);
4088 int err = -ENOBUFS;
4089
4090 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4091 if (skb == NULL)
4092 goto errout;
4093
4094 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4095 if (err < 0) {
4096 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4097 WARN_ON(err == -EMSGSIZE);
4098 kfree_skb(skb);
4099 goto errout;
4100 }
4101 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4102 return;
4103 errout:
4104 if (err < 0)
4105 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4106 }
4107
4108 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4109 __s32 *array, int bytes)
4110 {
4111 BUG_ON(bytes < (DEVCONF_MAX * 4));
4112
4113 memset(array, 0, bytes);
4114 array[DEVCONF_FORWARDING] = cnf->forwarding;
4115 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4116 array[DEVCONF_MTU6] = cnf->mtu6;
4117 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4118 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4119 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4120 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4121 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4122 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4123 jiffies_to_msecs(cnf->rtr_solicit_interval);
4124 array[DEVCONF_RTR_SOLICIT_DELAY] =
4125 jiffies_to_msecs(cnf->rtr_solicit_delay);
4126 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4127 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4128 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4129 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4130 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4131 #ifdef CONFIG_IPV6_PRIVACY
4132 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4133 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4134 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4135 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4136 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4137 #endif
4138 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4139 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4140 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4141 #ifdef CONFIG_IPV6_ROUTER_PREF
4142 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4143 array[DEVCONF_RTR_PROBE_INTERVAL] =
4144 jiffies_to_msecs(cnf->rtr_probe_interval);
4145 #ifdef CONFIG_IPV6_ROUTE_INFO
4146 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4147 #endif
4148 #endif
4149 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4150 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4151 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4152 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4153 #endif
4154 #ifdef CONFIG_IPV6_MROUTE
4155 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4156 #endif
4157 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4158 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4159 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4160 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4161 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4162 }
4163
4164 static inline size_t inet6_ifla6_size(void)
4165 {
4166 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4167 + nla_total_size(sizeof(struct ifla_cacheinfo))
4168 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4169 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4170 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4171 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4172 }
4173
4174 static inline size_t inet6_if_nlmsg_size(void)
4175 {
4176 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4177 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4178 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4179 + nla_total_size(4) /* IFLA_MTU */
4180 + nla_total_size(4) /* IFLA_LINK */
4181 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4182 }
4183
4184 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4185 int items, int bytes)
4186 {
4187 int i;
4188 int pad = bytes - sizeof(u64) * items;
4189 BUG_ON(pad < 0);
4190
4191 /* Use put_unaligned() because stats may not be aligned for u64. */
4192 put_unaligned(items, &stats[0]);
4193 for (i = 1; i < items; i++)
4194 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4195
4196 memset(&stats[items], 0, pad);
4197 }
4198
4199 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4200 int items, int bytes, size_t syncpoff)
4201 {
4202 int i;
4203 int pad = bytes - sizeof(u64) * items;
4204 BUG_ON(pad < 0);
4205
4206 /* Use put_unaligned() because stats may not be aligned for u64. */
4207 put_unaligned(items, &stats[0]);
4208 for (i = 1; i < items; i++)
4209 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4210
4211 memset(&stats[items], 0, pad);
4212 }
4213
4214 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4215 int bytes)
4216 {
4217 switch (attrtype) {
4218 case IFLA_INET6_STATS:
4219 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4220 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4221 break;
4222 case IFLA_INET6_ICMP6STATS:
4223 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4224 break;
4225 }
4226 }
4227
4228 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4229 {
4230 struct nlattr *nla;
4231 struct ifla_cacheinfo ci;
4232
4233 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4234 goto nla_put_failure;
4235 ci.max_reasm_len = IPV6_MAXPLEN;
4236 ci.tstamp = cstamp_delta(idev->tstamp);
4237 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4238 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4239 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4240 goto nla_put_failure;
4241 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4242 if (nla == NULL)
4243 goto nla_put_failure;
4244 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4245
4246 /* XXX - MC not implemented */
4247
4248 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4249 if (nla == NULL)
4250 goto nla_put_failure;
4251 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4252
4253 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4254 if (nla == NULL)
4255 goto nla_put_failure;
4256 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4257
4258 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4259 if (nla == NULL)
4260 goto nla_put_failure;
4261 read_lock_bh(&idev->lock);
4262 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4263 read_unlock_bh(&idev->lock);
4264
4265 return 0;
4266
4267 nla_put_failure:
4268 return -EMSGSIZE;
4269 }
4270
4271 static size_t inet6_get_link_af_size(const struct net_device *dev)
4272 {
4273 if (!__in6_dev_get(dev))
4274 return 0;
4275
4276 return inet6_ifla6_size();
4277 }
4278
4279 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4280 {
4281 struct inet6_dev *idev = __in6_dev_get(dev);
4282
4283 if (!idev)
4284 return -ENODATA;
4285
4286 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4287 return -EMSGSIZE;
4288
4289 return 0;
4290 }
4291
4292 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4293 {
4294 struct inet6_ifaddr *ifp;
4295 struct net_device *dev = idev->dev;
4296 bool update_rs = false;
4297 struct in6_addr ll_addr;
4298
4299 if (token == NULL)
4300 return -EINVAL;
4301 if (ipv6_addr_any(token))
4302 return -EINVAL;
4303 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4304 return -EINVAL;
4305 if (!ipv6_accept_ra(idev))
4306 return -EINVAL;
4307 if (idev->cnf.rtr_solicits <= 0)
4308 return -EINVAL;
4309
4310 write_lock_bh(&idev->lock);
4311
4312 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4313 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4314
4315 write_unlock_bh(&idev->lock);
4316
4317 if (!idev->dead && (idev->if_flags & IF_READY) &&
4318 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4319 IFA_F_OPTIMISTIC)) {
4320
4321 /* If we're not ready, then normal ifup will take care
4322 * of this. Otherwise, we need to request our rs here.
4323 */
4324 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4325 update_rs = true;
4326 }
4327
4328 write_lock_bh(&idev->lock);
4329
4330 if (update_rs) {
4331 idev->if_flags |= IF_RS_SENT;
4332 idev->rs_probes = 1;
4333 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4334 }
4335
4336 /* Well, that's kinda nasty ... */
4337 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4338 spin_lock(&ifp->lock);
4339 if (ifp->tokenized) {
4340 ifp->valid_lft = 0;
4341 ifp->prefered_lft = 0;
4342 }
4343 spin_unlock(&ifp->lock);
4344 }
4345
4346 write_unlock_bh(&idev->lock);
4347 addrconf_verify(0);
4348 return 0;
4349 }
4350
4351 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4352 {
4353 int err = -EINVAL;
4354 struct inet6_dev *idev = __in6_dev_get(dev);
4355 struct nlattr *tb[IFLA_INET6_MAX + 1];
4356
4357 if (!idev)
4358 return -EAFNOSUPPORT;
4359
4360 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4361 BUG();
4362
4363 if (tb[IFLA_INET6_TOKEN])
4364 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4365
4366 return err;
4367 }
4368
4369 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4370 u32 portid, u32 seq, int event, unsigned int flags)
4371 {
4372 struct net_device *dev = idev->dev;
4373 struct ifinfomsg *hdr;
4374 struct nlmsghdr *nlh;
4375 void *protoinfo;
4376
4377 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4378 if (nlh == NULL)
4379 return -EMSGSIZE;
4380
4381 hdr = nlmsg_data(nlh);
4382 hdr->ifi_family = AF_INET6;
4383 hdr->__ifi_pad = 0;
4384 hdr->ifi_type = dev->type;
4385 hdr->ifi_index = dev->ifindex;
4386 hdr->ifi_flags = dev_get_flags(dev);
4387 hdr->ifi_change = 0;
4388
4389 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4390 (dev->addr_len &&
4391 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4392 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4393 (dev->ifindex != dev->iflink &&
4394 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4395 goto nla_put_failure;
4396 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4397 if (protoinfo == NULL)
4398 goto nla_put_failure;
4399
4400 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4401 goto nla_put_failure;
4402
4403 nla_nest_end(skb, protoinfo);
4404 return nlmsg_end(skb, nlh);
4405
4406 nla_put_failure:
4407 nlmsg_cancel(skb, nlh);
4408 return -EMSGSIZE;
4409 }
4410
4411 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4412 {
4413 struct net *net = sock_net(skb->sk);
4414 int h, s_h;
4415 int idx = 0, s_idx;
4416 struct net_device *dev;
4417 struct inet6_dev *idev;
4418 struct hlist_head *head;
4419
4420 s_h = cb->args[0];
4421 s_idx = cb->args[1];
4422
4423 rcu_read_lock();
4424 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4425 idx = 0;
4426 head = &net->dev_index_head[h];
4427 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4428 if (idx < s_idx)
4429 goto cont;
4430 idev = __in6_dev_get(dev);
4431 if (!idev)
4432 goto cont;
4433 if (inet6_fill_ifinfo(skb, idev,
4434 NETLINK_CB(cb->skb).portid,
4435 cb->nlh->nlmsg_seq,
4436 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4437 goto out;
4438 cont:
4439 idx++;
4440 }
4441 }
4442 out:
4443 rcu_read_unlock();
4444 cb->args[1] = idx;
4445 cb->args[0] = h;
4446
4447 return skb->len;
4448 }
4449
4450 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4451 {
4452 struct sk_buff *skb;
4453 struct net *net = dev_net(idev->dev);
4454 int err = -ENOBUFS;
4455
4456 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4457 if (skb == NULL)
4458 goto errout;
4459
4460 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4461 if (err < 0) {
4462 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4463 WARN_ON(err == -EMSGSIZE);
4464 kfree_skb(skb);
4465 goto errout;
4466 }
4467 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4468 return;
4469 errout:
4470 if (err < 0)
4471 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4472 }
4473
4474 static inline size_t inet6_prefix_nlmsg_size(void)
4475 {
4476 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4477 + nla_total_size(sizeof(struct in6_addr))
4478 + nla_total_size(sizeof(struct prefix_cacheinfo));
4479 }
4480
4481 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4482 struct prefix_info *pinfo, u32 portid, u32 seq,
4483 int event, unsigned int flags)
4484 {
4485 struct prefixmsg *pmsg;
4486 struct nlmsghdr *nlh;
4487 struct prefix_cacheinfo ci;
4488
4489 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4490 if (nlh == NULL)
4491 return -EMSGSIZE;
4492
4493 pmsg = nlmsg_data(nlh);
4494 pmsg->prefix_family = AF_INET6;
4495 pmsg->prefix_pad1 = 0;
4496 pmsg->prefix_pad2 = 0;
4497 pmsg->prefix_ifindex = idev->dev->ifindex;
4498 pmsg->prefix_len = pinfo->prefix_len;
4499 pmsg->prefix_type = pinfo->type;
4500 pmsg->prefix_pad3 = 0;
4501 pmsg->prefix_flags = 0;
4502 if (pinfo->onlink)
4503 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4504 if (pinfo->autoconf)
4505 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4506
4507 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4508 goto nla_put_failure;
4509 ci.preferred_time = ntohl(pinfo->prefered);
4510 ci.valid_time = ntohl(pinfo->valid);
4511 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4512 goto nla_put_failure;
4513 return nlmsg_end(skb, nlh);
4514
4515 nla_put_failure:
4516 nlmsg_cancel(skb, nlh);
4517 return -EMSGSIZE;
4518 }
4519
4520 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4521 struct prefix_info *pinfo)
4522 {
4523 struct sk_buff *skb;
4524 struct net *net = dev_net(idev->dev);
4525 int err = -ENOBUFS;
4526
4527 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4528 if (skb == NULL)
4529 goto errout;
4530
4531 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4532 if (err < 0) {
4533 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4534 WARN_ON(err == -EMSGSIZE);
4535 kfree_skb(skb);
4536 goto errout;
4537 }
4538 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4539 return;
4540 errout:
4541 if (err < 0)
4542 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4543 }
4544
4545 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4546 {
4547 write_lock_bh(&ifp->idev->lock);
4548 spin_lock(&ifp->lock);
4549 if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4550 IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4551 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4552 ifp->idev->valid_ll_addr_cnt += count;
4553 WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4554 spin_unlock(&ifp->lock);
4555 write_unlock_bh(&ifp->idev->lock);
4556 }
4557
4558 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4559 {
4560 struct net *net = dev_net(ifp->idev->dev);
4561
4562 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4563
4564 switch (event) {
4565 case RTM_NEWADDR:
4566 update_valid_ll_addr_cnt(ifp, 1);
4567
4568 /*
4569 * If the address was optimistic
4570 * we inserted the route at the start of
4571 * our DAD process, so we don't need
4572 * to do it again
4573 */
4574 if (!(ifp->rt->rt6i_node))
4575 ip6_ins_rt(ifp->rt);
4576 if (ifp->idev->cnf.forwarding)
4577 addrconf_join_anycast(ifp);
4578 if (!ipv6_addr_any(&ifp->peer_addr))
4579 addrconf_prefix_route(&ifp->peer_addr, 128,
4580 ifp->idev->dev, 0, 0);
4581 break;
4582 case RTM_DELADDR:
4583 update_valid_ll_addr_cnt(ifp, -1);
4584
4585 if (ifp->idev->cnf.forwarding)
4586 addrconf_leave_anycast(ifp);
4587 addrconf_leave_solict(ifp->idev, &ifp->addr);
4588 if (!ipv6_addr_any(&ifp->peer_addr)) {
4589 struct rt6_info *rt;
4590 struct net_device *dev = ifp->idev->dev;
4591
4592 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4593 dev->ifindex, 1);
4594 if (rt) {
4595 dst_hold(&rt->dst);
4596 if (ip6_del_rt(rt))
4597 dst_free(&rt->dst);
4598 }
4599 }
4600 dst_hold(&ifp->rt->dst);
4601
4602 if (ip6_del_rt(ifp->rt))
4603 dst_free(&ifp->rt->dst);
4604 break;
4605 }
4606 atomic_inc(&net->ipv6.dev_addr_genid);
4607 rt_genid_bump_ipv6(net);
4608 }
4609
4610 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4611 {
4612 rcu_read_lock_bh();
4613 if (likely(ifp->idev->dead == 0))
4614 __ipv6_ifa_notify(event, ifp);
4615 rcu_read_unlock_bh();
4616 }
4617
4618 #ifdef CONFIG_SYSCTL
4619
4620 static
4621 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4622 void __user *buffer, size_t *lenp, loff_t *ppos)
4623 {
4624 int *valp = ctl->data;
4625 int val = *valp;
4626 loff_t pos = *ppos;
4627 struct ctl_table lctl;
4628 int ret;
4629
4630 /*
4631 * ctl->data points to idev->cnf.forwarding, we should
4632 * not modify it until we get the rtnl lock.
4633 */
4634 lctl = *ctl;
4635 lctl.data = &val;
4636
4637 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4638
4639 if (write)
4640 ret = addrconf_fixup_forwarding(ctl, valp, val);
4641 if (ret)
4642 *ppos = pos;
4643 return ret;
4644 }
4645
4646 static void dev_disable_change(struct inet6_dev *idev)
4647 {
4648 struct netdev_notifier_info info;
4649
4650 if (!idev || !idev->dev)
4651 return;
4652
4653 netdev_notifier_info_init(&info, idev->dev);
4654 if (idev->cnf.disable_ipv6)
4655 addrconf_notify(NULL, NETDEV_DOWN, &info);
4656 else
4657 addrconf_notify(NULL, NETDEV_UP, &info);
4658 }
4659
4660 static void addrconf_disable_change(struct net *net, __s32 newf)
4661 {
4662 struct net_device *dev;
4663 struct inet6_dev *idev;
4664
4665 rcu_read_lock();
4666 for_each_netdev_rcu(net, dev) {
4667 idev = __in6_dev_get(dev);
4668 if (idev) {
4669 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4670 idev->cnf.disable_ipv6 = newf;
4671 if (changed)
4672 dev_disable_change(idev);
4673 }
4674 }
4675 rcu_read_unlock();
4676 }
4677
4678 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4679 {
4680 struct net *net;
4681 int old;
4682
4683 if (!rtnl_trylock())
4684 return restart_syscall();
4685
4686 net = (struct net *)table->extra2;
4687 old = *p;
4688 *p = newf;
4689
4690 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4691 rtnl_unlock();
4692 return 0;
4693 }
4694
4695 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4696 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4697 addrconf_disable_change(net, newf);
4698 } else if ((!newf) ^ (!old))
4699 dev_disable_change((struct inet6_dev *)table->extra1);
4700
4701 rtnl_unlock();
4702 return 0;
4703 }
4704
4705 static
4706 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4707 void __user *buffer, size_t *lenp, loff_t *ppos)
4708 {
4709 int *valp = ctl->data;
4710 int val = *valp;
4711 loff_t pos = *ppos;
4712 struct ctl_table lctl;
4713 int ret;
4714
4715 /*
4716 * ctl->data points to idev->cnf.disable_ipv6, we should
4717 * not modify it until we get the rtnl lock.
4718 */
4719 lctl = *ctl;
4720 lctl.data = &val;
4721
4722 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4723
4724 if (write)
4725 ret = addrconf_disable_ipv6(ctl, valp, val);
4726 if (ret)
4727 *ppos = pos;
4728 return ret;
4729 }
4730
4731 static struct addrconf_sysctl_table
4732 {
4733 struct ctl_table_header *sysctl_header;
4734 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4735 } addrconf_sysctl __read_mostly = {
4736 .sysctl_header = NULL,
4737 .addrconf_vars = {
4738 {
4739 .procname = "forwarding",
4740 .data = &ipv6_devconf.forwarding,
4741 .maxlen = sizeof(int),
4742 .mode = 0644,
4743 .proc_handler = addrconf_sysctl_forward,
4744 },
4745 {
4746 .procname = "hop_limit",
4747 .data = &ipv6_devconf.hop_limit,
4748 .maxlen = sizeof(int),
4749 .mode = 0644,
4750 .proc_handler = proc_dointvec,
4751 },
4752 {
4753 .procname = "mtu",
4754 .data = &ipv6_devconf.mtu6,
4755 .maxlen = sizeof(int),
4756 .mode = 0644,
4757 .proc_handler = proc_dointvec,
4758 },
4759 {
4760 .procname = "accept_ra",
4761 .data = &ipv6_devconf.accept_ra,
4762 .maxlen = sizeof(int),
4763 .mode = 0644,
4764 .proc_handler = proc_dointvec,
4765 },
4766 {
4767 .procname = "accept_redirects",
4768 .data = &ipv6_devconf.accept_redirects,
4769 .maxlen = sizeof(int),
4770 .mode = 0644,
4771 .proc_handler = proc_dointvec,
4772 },
4773 {
4774 .procname = "autoconf",
4775 .data = &ipv6_devconf.autoconf,
4776 .maxlen = sizeof(int),
4777 .mode = 0644,
4778 .proc_handler = proc_dointvec,
4779 },
4780 {
4781 .procname = "dad_transmits",
4782 .data = &ipv6_devconf.dad_transmits,
4783 .maxlen = sizeof(int),
4784 .mode = 0644,
4785 .proc_handler = proc_dointvec,
4786 },
4787 {
4788 .procname = "router_solicitations",
4789 .data = &ipv6_devconf.rtr_solicits,
4790 .maxlen = sizeof(int),
4791 .mode = 0644,
4792 .proc_handler = proc_dointvec,
4793 },
4794 {
4795 .procname = "router_solicitation_interval",
4796 .data = &ipv6_devconf.rtr_solicit_interval,
4797 .maxlen = sizeof(int),
4798 .mode = 0644,
4799 .proc_handler = proc_dointvec_jiffies,
4800 },
4801 {
4802 .procname = "router_solicitation_delay",
4803 .data = &ipv6_devconf.rtr_solicit_delay,
4804 .maxlen = sizeof(int),
4805 .mode = 0644,
4806 .proc_handler = proc_dointvec_jiffies,
4807 },
4808 {
4809 .procname = "force_mld_version",
4810 .data = &ipv6_devconf.force_mld_version,
4811 .maxlen = sizeof(int),
4812 .mode = 0644,
4813 .proc_handler = proc_dointvec,
4814 },
4815 {
4816 .procname = "mldv1_unsolicited_report_interval",
4817 .data =
4818 &ipv6_devconf.mldv1_unsolicited_report_interval,
4819 .maxlen = sizeof(int),
4820 .mode = 0644,
4821 .proc_handler = proc_dointvec_ms_jiffies,
4822 },
4823 {
4824 .procname = "mldv2_unsolicited_report_interval",
4825 .data =
4826 &ipv6_devconf.mldv2_unsolicited_report_interval,
4827 .maxlen = sizeof(int),
4828 .mode = 0644,
4829 .proc_handler = proc_dointvec_ms_jiffies,
4830 },
4831 #ifdef CONFIG_IPV6_PRIVACY
4832 {
4833 .procname = "use_tempaddr",
4834 .data = &ipv6_devconf.use_tempaddr,
4835 .maxlen = sizeof(int),
4836 .mode = 0644,
4837 .proc_handler = proc_dointvec,
4838 },
4839 {
4840 .procname = "temp_valid_lft",
4841 .data = &ipv6_devconf.temp_valid_lft,
4842 .maxlen = sizeof(int),
4843 .mode = 0644,
4844 .proc_handler = proc_dointvec,
4845 },
4846 {
4847 .procname = "temp_prefered_lft",
4848 .data = &ipv6_devconf.temp_prefered_lft,
4849 .maxlen = sizeof(int),
4850 .mode = 0644,
4851 .proc_handler = proc_dointvec,
4852 },
4853 {
4854 .procname = "regen_max_retry",
4855 .data = &ipv6_devconf.regen_max_retry,
4856 .maxlen = sizeof(int),
4857 .mode = 0644,
4858 .proc_handler = proc_dointvec,
4859 },
4860 {
4861 .procname = "max_desync_factor",
4862 .data = &ipv6_devconf.max_desync_factor,
4863 .maxlen = sizeof(int),
4864 .mode = 0644,
4865 .proc_handler = proc_dointvec,
4866 },
4867 #endif
4868 {
4869 .procname = "max_addresses",
4870 .data = &ipv6_devconf.max_addresses,
4871 .maxlen = sizeof(int),
4872 .mode = 0644,
4873 .proc_handler = proc_dointvec,
4874 },
4875 {
4876 .procname = "accept_ra_defrtr",
4877 .data = &ipv6_devconf.accept_ra_defrtr,
4878 .maxlen = sizeof(int),
4879 .mode = 0644,
4880 .proc_handler = proc_dointvec,
4881 },
4882 {
4883 .procname = "accept_ra_pinfo",
4884 .data = &ipv6_devconf.accept_ra_pinfo,
4885 .maxlen = sizeof(int),
4886 .mode = 0644,
4887 .proc_handler = proc_dointvec,
4888 },
4889 #ifdef CONFIG_IPV6_ROUTER_PREF
4890 {
4891 .procname = "accept_ra_rtr_pref",
4892 .data = &ipv6_devconf.accept_ra_rtr_pref,
4893 .maxlen = sizeof(int),
4894 .mode = 0644,
4895 .proc_handler = proc_dointvec,
4896 },
4897 {
4898 .procname = "router_probe_interval",
4899 .data = &ipv6_devconf.rtr_probe_interval,
4900 .maxlen = sizeof(int),
4901 .mode = 0644,
4902 .proc_handler = proc_dointvec_jiffies,
4903 },
4904 #ifdef CONFIG_IPV6_ROUTE_INFO
4905 {
4906 .procname = "accept_ra_rt_info_max_plen",
4907 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4908 .maxlen = sizeof(int),
4909 .mode = 0644,
4910 .proc_handler = proc_dointvec,
4911 },
4912 #endif
4913 #endif
4914 {
4915 .procname = "proxy_ndp",
4916 .data = &ipv6_devconf.proxy_ndp,
4917 .maxlen = sizeof(int),
4918 .mode = 0644,
4919 .proc_handler = proc_dointvec,
4920 },
4921 {
4922 .procname = "accept_source_route",
4923 .data = &ipv6_devconf.accept_source_route,
4924 .maxlen = sizeof(int),
4925 .mode = 0644,
4926 .proc_handler = proc_dointvec,
4927 },
4928 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4929 {
4930 .procname = "optimistic_dad",
4931 .data = &ipv6_devconf.optimistic_dad,
4932 .maxlen = sizeof(int),
4933 .mode = 0644,
4934 .proc_handler = proc_dointvec,
4935
4936 },
4937 #endif
4938 #ifdef CONFIG_IPV6_MROUTE
4939 {
4940 .procname = "mc_forwarding",
4941 .data = &ipv6_devconf.mc_forwarding,
4942 .maxlen = sizeof(int),
4943 .mode = 0444,
4944 .proc_handler = proc_dointvec,
4945 },
4946 #endif
4947 {
4948 .procname = "disable_ipv6",
4949 .data = &ipv6_devconf.disable_ipv6,
4950 .maxlen = sizeof(int),
4951 .mode = 0644,
4952 .proc_handler = addrconf_sysctl_disable,
4953 },
4954 {
4955 .procname = "accept_dad",
4956 .data = &ipv6_devconf.accept_dad,
4957 .maxlen = sizeof(int),
4958 .mode = 0644,
4959 .proc_handler = proc_dointvec,
4960 },
4961 {
4962 .procname = "force_tllao",
4963 .data = &ipv6_devconf.force_tllao,
4964 .maxlen = sizeof(int),
4965 .mode = 0644,
4966 .proc_handler = proc_dointvec
4967 },
4968 {
4969 .procname = "ndisc_notify",
4970 .data = &ipv6_devconf.ndisc_notify,
4971 .maxlen = sizeof(int),
4972 .mode = 0644,
4973 .proc_handler = proc_dointvec
4974 },
4975 {
4976 .procname = "suppress_frag_ndisc",
4977 .data = &ipv6_devconf.suppress_frag_ndisc,
4978 .maxlen = sizeof(int),
4979 .mode = 0644,
4980 .proc_handler = proc_dointvec
4981 },
4982 {
4983 /* sentinel */
4984 }
4985 },
4986 };
4987
4988 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4989 struct inet6_dev *idev, struct ipv6_devconf *p)
4990 {
4991 int i;
4992 struct addrconf_sysctl_table *t;
4993 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4994
4995 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4996 if (t == NULL)
4997 goto out;
4998
4999 for (i = 0; t->addrconf_vars[i].data; i++) {
5000 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5001 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5002 t->addrconf_vars[i].extra2 = net;
5003 }
5004
5005 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5006
5007 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5008 if (t->sysctl_header == NULL)
5009 goto free;
5010
5011 p->sysctl = t;
5012 return 0;
5013
5014 free:
5015 kfree(t);
5016 out:
5017 return -ENOBUFS;
5018 }
5019
5020 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5021 {
5022 struct addrconf_sysctl_table *t;
5023
5024 if (p->sysctl == NULL)
5025 return;
5026
5027 t = p->sysctl;
5028 p->sysctl = NULL;
5029 unregister_net_sysctl_table(t->sysctl_header);
5030 kfree(t);
5031 }
5032
5033 static void addrconf_sysctl_register(struct inet6_dev *idev)
5034 {
5035 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5036 &ndisc_ifinfo_sysctl_change);
5037 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5038 idev, &idev->cnf);
5039 }
5040
5041 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5042 {
5043 __addrconf_sysctl_unregister(&idev->cnf);
5044 neigh_sysctl_unregister(idev->nd_parms);
5045 }
5046
5047
5048 #endif
5049
5050 static int __net_init addrconf_init_net(struct net *net)
5051 {
5052 int err = -ENOMEM;
5053 struct ipv6_devconf *all, *dflt;
5054
5055 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5056 if (all == NULL)
5057 goto err_alloc_all;
5058
5059 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5060 if (dflt == NULL)
5061 goto err_alloc_dflt;
5062
5063 /* these will be inherited by all namespaces */
5064 dflt->autoconf = ipv6_defaults.autoconf;
5065 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5066
5067 net->ipv6.devconf_all = all;
5068 net->ipv6.devconf_dflt = dflt;
5069
5070 #ifdef CONFIG_SYSCTL
5071 err = __addrconf_sysctl_register(net, "all", NULL, all);
5072 if (err < 0)
5073 goto err_reg_all;
5074
5075 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5076 if (err < 0)
5077 goto err_reg_dflt;
5078 #endif
5079 return 0;
5080
5081 #ifdef CONFIG_SYSCTL
5082 err_reg_dflt:
5083 __addrconf_sysctl_unregister(all);
5084 err_reg_all:
5085 kfree(dflt);
5086 #endif
5087 err_alloc_dflt:
5088 kfree(all);
5089 err_alloc_all:
5090 return err;
5091 }
5092
5093 static void __net_exit addrconf_exit_net(struct net *net)
5094 {
5095 #ifdef CONFIG_SYSCTL
5096 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5097 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5098 #endif
5099 if (!net_eq(net, &init_net)) {
5100 kfree(net->ipv6.devconf_dflt);
5101 kfree(net->ipv6.devconf_all);
5102 }
5103 }
5104
5105 static struct pernet_operations addrconf_ops = {
5106 .init = addrconf_init_net,
5107 .exit = addrconf_exit_net,
5108 };
5109
5110 static struct rtnl_af_ops inet6_ops = {
5111 .family = AF_INET6,
5112 .fill_link_af = inet6_fill_link_af,
5113 .get_link_af_size = inet6_get_link_af_size,
5114 .set_link_af = inet6_set_link_af,
5115 };
5116
5117 /*
5118 * Init / cleanup code
5119 */
5120
5121 int __init addrconf_init(void)
5122 {
5123 int i, err;
5124
5125 err = ipv6_addr_label_init();
5126 if (err < 0) {
5127 pr_crit("%s: cannot initialize default policy table: %d\n",
5128 __func__, err);
5129 goto out;
5130 }
5131
5132 err = register_pernet_subsys(&addrconf_ops);
5133 if (err < 0)
5134 goto out_addrlabel;
5135
5136 /* The addrconf netdev notifier requires that loopback_dev
5137 * has it's ipv6 private information allocated and setup
5138 * before it can bring up and give link-local addresses
5139 * to other devices which are up.
5140 *
5141 * Unfortunately, loopback_dev is not necessarily the first
5142 * entry in the global dev_base list of net devices. In fact,
5143 * it is likely to be the very last entry on that list.
5144 * So this causes the notifier registry below to try and
5145 * give link-local addresses to all devices besides loopback_dev
5146 * first, then loopback_dev, which cases all the non-loopback_dev
5147 * devices to fail to get a link-local address.
5148 *
5149 * So, as a temporary fix, allocate the ipv6 structure for
5150 * loopback_dev first by hand.
5151 * Longer term, all of the dependencies ipv6 has upon the loopback
5152 * device and it being up should be removed.
5153 */
5154 rtnl_lock();
5155 if (!ipv6_add_dev(init_net.loopback_dev))
5156 err = -ENOMEM;
5157 rtnl_unlock();
5158 if (err)
5159 goto errlo;
5160
5161 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5162 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5163
5164 register_netdevice_notifier(&ipv6_dev_notf);
5165
5166 addrconf_verify(0);
5167
5168 err = rtnl_af_register(&inet6_ops);
5169 if (err < 0)
5170 goto errout_af;
5171
5172 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5173 NULL);
5174 if (err < 0)
5175 goto errout;
5176
5177 /* Only the first call to __rtnl_register can fail */
5178 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5179 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5180 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5181 inet6_dump_ifaddr, NULL);
5182 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5183 inet6_dump_ifmcaddr, NULL);
5184 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5185 inet6_dump_ifacaddr, NULL);
5186 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5187 inet6_netconf_dump_devconf, NULL);
5188
5189 ipv6_addr_label_rtnl_register();
5190
5191 return 0;
5192 errout:
5193 rtnl_af_unregister(&inet6_ops);
5194 errout_af:
5195 unregister_netdevice_notifier(&ipv6_dev_notf);
5196 errlo:
5197 unregister_pernet_subsys(&addrconf_ops);
5198 out_addrlabel:
5199 ipv6_addr_label_cleanup();
5200 out:
5201 return err;
5202 }
5203
5204 void addrconf_cleanup(void)
5205 {
5206 struct net_device *dev;
5207 int i;
5208
5209 unregister_netdevice_notifier(&ipv6_dev_notf);
5210 unregister_pernet_subsys(&addrconf_ops);
5211 ipv6_addr_label_cleanup();
5212
5213 rtnl_lock();
5214
5215 __rtnl_af_unregister(&inet6_ops);
5216
5217 /* clean dev list */
5218 for_each_netdev(&init_net, dev) {
5219 if (__in6_dev_get(dev) == NULL)
5220 continue;
5221 addrconf_ifdown(dev, 1);
5222 }
5223 addrconf_ifdown(init_net.loopback_dev, 2);
5224
5225 /*
5226 * Check hash table.
5227 */
5228 spin_lock_bh(&addrconf_hash_lock);
5229 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5230 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5231 spin_unlock_bh(&addrconf_hash_lock);
5232
5233 del_timer(&addr_chk_timer);
5234 rtnl_unlock();
5235 }
Linus' kernel tree