Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blobf7c8bbeb27b704c0106f714d5a0677c27d3346e0
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
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.
16 * Changes:
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.
41 #define pr_fmt(fmt) "IPv6: " fmt
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>
68 #include <net/net_namespace.h>
69 #include <net/sock.h>
70 #include <net/snmp.h>
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
74 #include <net/ipv6.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
79 #include <net/tcp.h>
80 #include <net/ip.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
94 /* Set to 3 to get tracing... */
95 #define ACONF_DEBUG 2
97 #if ACONF_DEBUG >= 3
98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
99 #else
100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
101 #endif
103 #define INFINITY_LIFE_TIME 0xFFFFFFFF
105 static inline u32 cstamp_delta(unsigned long cstamp)
107 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
110 #ifdef CONFIG_SYSCTL
111 static int addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
113 #else
114 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
116 return 0;
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
122 #endif
124 static void __ipv6_regen_rndid(struct inet6_dev *idev);
125 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
126 static void ipv6_regen_rndid(unsigned long data);
128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
129 static int ipv6_count_addresses(struct inet6_dev *idev);
132 * Configured unicast address hash table
134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
135 static DEFINE_SPINLOCK(addrconf_hash_lock);
137 static void addrconf_verify(void);
138 static void addrconf_verify_rtnl(void);
139 static void addrconf_verify_work(struct work_struct *);
141 static struct workqueue_struct *addrconf_wq;
142 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_type_change(struct net_device *dev,
148 unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
151 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
152 int plen,
153 const struct net_device *dev,
154 u32 flags, u32 noflags);
156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
157 static void addrconf_dad_work(struct work_struct *w);
158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_run(struct inet6_dev *idev);
160 static void addrconf_rs_timer(unsigned long data);
161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
165 struct prefix_info *pinfo);
166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
167 struct net_device *dev);
169 static struct ipv6_devconf ipv6_devconf __read_mostly = {
170 .forwarding = 0,
171 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
172 .mtu6 = IPV6_MIN_MTU,
173 .accept_ra = 1,
174 .accept_redirects = 1,
175 .autoconf = 1,
176 .force_mld_version = 0,
177 .mldv1_unsolicited_report_interval = 10 * HZ,
178 .mldv2_unsolicited_report_interval = HZ,
179 .dad_transmits = 1,
180 .rtr_solicits = MAX_RTR_SOLICITATIONS,
181 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
182 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
183 .use_tempaddr = 0,
184 .temp_valid_lft = TEMP_VALID_LIFETIME,
185 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
186 .regen_max_retry = REGEN_MAX_RETRY,
187 .max_desync_factor = MAX_DESYNC_FACTOR,
188 .max_addresses = IPV6_MAX_ADDRESSES,
189 .accept_ra_defrtr = 1,
190 .accept_ra_from_local = 0,
191 .accept_ra_pinfo = 1,
192 #ifdef CONFIG_IPV6_ROUTER_PREF
193 .accept_ra_rtr_pref = 1,
194 .rtr_probe_interval = 60 * HZ,
195 #ifdef CONFIG_IPV6_ROUTE_INFO
196 .accept_ra_rt_info_max_plen = 0,
197 #endif
198 #endif
199 .proxy_ndp = 0,
200 .accept_source_route = 0, /* we do not accept RH0 by default. */
201 .disable_ipv6 = 0,
202 .accept_dad = 1,
203 .suppress_frag_ndisc = 1,
206 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
207 .forwarding = 0,
208 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
209 .mtu6 = IPV6_MIN_MTU,
210 .accept_ra = 1,
211 .accept_redirects = 1,
212 .autoconf = 1,
213 .force_mld_version = 0,
214 .mldv1_unsolicited_report_interval = 10 * HZ,
215 .mldv2_unsolicited_report_interval = HZ,
216 .dad_transmits = 1,
217 .rtr_solicits = MAX_RTR_SOLICITATIONS,
218 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
219 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
220 .use_tempaddr = 0,
221 .temp_valid_lft = TEMP_VALID_LIFETIME,
222 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
223 .regen_max_retry = REGEN_MAX_RETRY,
224 .max_desync_factor = MAX_DESYNC_FACTOR,
225 .max_addresses = IPV6_MAX_ADDRESSES,
226 .accept_ra_defrtr = 1,
227 .accept_ra_from_local = 0,
228 .accept_ra_pinfo = 1,
229 #ifdef CONFIG_IPV6_ROUTER_PREF
230 .accept_ra_rtr_pref = 1,
231 .rtr_probe_interval = 60 * HZ,
232 #ifdef CONFIG_IPV6_ROUTE_INFO
233 .accept_ra_rt_info_max_plen = 0,
234 #endif
235 #endif
236 .proxy_ndp = 0,
237 .accept_source_route = 0, /* we do not accept RH0 by default. */
238 .disable_ipv6 = 0,
239 .accept_dad = 1,
240 .suppress_frag_ndisc = 1,
243 /* Check if a valid qdisc is available */
244 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
246 return !qdisc_tx_is_noop(dev);
249 static void addrconf_del_rs_timer(struct inet6_dev *idev)
251 if (del_timer(&idev->rs_timer))
252 __in6_dev_put(idev);
255 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
257 if (cancel_delayed_work(&ifp->dad_work))
258 __in6_ifa_put(ifp);
261 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
262 unsigned long when)
264 if (!timer_pending(&idev->rs_timer))
265 in6_dev_hold(idev);
266 mod_timer(&idev->rs_timer, jiffies + when);
269 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
270 unsigned long delay)
272 if (!delayed_work_pending(&ifp->dad_work))
273 in6_ifa_hold(ifp);
274 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
277 static int snmp6_alloc_dev(struct inet6_dev *idev)
279 int i;
281 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
282 if (!idev->stats.ipv6)
283 goto err_ip;
285 for_each_possible_cpu(i) {
286 struct ipstats_mib *addrconf_stats;
287 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
288 u64_stats_init(&addrconf_stats->syncp);
292 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
293 GFP_KERNEL);
294 if (!idev->stats.icmpv6dev)
295 goto err_icmp;
296 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
297 GFP_KERNEL);
298 if (!idev->stats.icmpv6msgdev)
299 goto err_icmpmsg;
301 return 0;
303 err_icmpmsg:
304 kfree(idev->stats.icmpv6dev);
305 err_icmp:
306 free_percpu(idev->stats.ipv6);
307 err_ip:
308 return -ENOMEM;
311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
313 struct inet6_dev *ndev;
314 int err = -ENOMEM;
316 ASSERT_RTNL();
318 if (dev->mtu < IPV6_MIN_MTU)
319 return ERR_PTR(-EINVAL);
321 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
322 if (ndev == NULL)
323 return ERR_PTR(err);
325 rwlock_init(&ndev->lock);
326 ndev->dev = dev;
327 INIT_LIST_HEAD(&ndev->addr_list);
328 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
329 (unsigned long)ndev);
330 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
331 ndev->cnf.mtu6 = dev->mtu;
332 ndev->cnf.sysctl = NULL;
333 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
334 if (ndev->nd_parms == NULL) {
335 kfree(ndev);
336 return ERR_PTR(err);
338 if (ndev->cnf.forwarding)
339 dev_disable_lro(dev);
340 /* We refer to the device */
341 dev_hold(dev);
343 if (snmp6_alloc_dev(ndev) < 0) {
344 ADBG(KERN_WARNING
345 "%s: cannot allocate memory for statistics; dev=%s.\n",
346 __func__, dev->name);
347 neigh_parms_release(&nd_tbl, ndev->nd_parms);
348 dev_put(dev);
349 kfree(ndev);
350 return ERR_PTR(err);
353 if (snmp6_register_dev(ndev) < 0) {
354 ADBG(KERN_WARNING
355 "%s: cannot create /proc/net/dev_snmp6/%s\n",
356 __func__, dev->name);
357 goto err_release;
360 /* One reference from device. We must do this before
361 * we invoke __ipv6_regen_rndid().
363 in6_dev_hold(ndev);
365 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
366 ndev->cnf.accept_dad = -1;
368 #if IS_ENABLED(CONFIG_IPV6_SIT)
369 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
370 pr_info("%s: Disabled Multicast RS\n", dev->name);
371 ndev->cnf.rtr_solicits = 0;
373 #endif
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);
388 ndev->token = in6addr_any;
390 if (netif_running(dev) && addrconf_qdisc_ok(dev))
391 ndev->if_flags |= IF_READY;
393 ipv6_mc_init_dev(ndev);
394 ndev->tstamp = jiffies;
395 err = addrconf_sysctl_register(ndev);
396 if (err) {
397 ipv6_mc_destroy_dev(ndev);
398 del_timer(&ndev->regen_timer);
399 goto err_release;
401 /* protected by rtnl_lock */
402 rcu_assign_pointer(dev->ip6_ptr, ndev);
404 /* Join interface-local all-node multicast group */
405 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
407 /* Join all-node multicast group */
408 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
410 /* Join all-router multicast group if forwarding is set */
411 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
412 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
414 return ndev;
416 err_release:
417 neigh_parms_release(&nd_tbl, ndev->nd_parms);
418 ndev->dead = 1;
419 in6_dev_finish_destroy(ndev);
420 return ERR_PTR(err);
423 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
425 struct inet6_dev *idev;
427 ASSERT_RTNL();
429 idev = __in6_dev_get(dev);
430 if (!idev) {
431 idev = ipv6_add_dev(dev);
432 if (IS_ERR(idev))
433 return NULL;
436 if (dev->flags&IFF_UP)
437 ipv6_mc_up(idev);
438 return idev;
441 static int inet6_netconf_msgsize_devconf(int type)
443 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
444 + nla_total_size(4); /* NETCONFA_IFINDEX */
446 /* type -1 is used for ALL */
447 if (type == -1 || type == NETCONFA_FORWARDING)
448 size += nla_total_size(4);
449 #ifdef CONFIG_IPV6_MROUTE
450 if (type == -1 || type == NETCONFA_MC_FORWARDING)
451 size += nla_total_size(4);
452 #endif
453 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
454 size += nla_total_size(4);
456 return size;
459 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
460 struct ipv6_devconf *devconf, u32 portid,
461 u32 seq, int event, unsigned int flags,
462 int type)
464 struct nlmsghdr *nlh;
465 struct netconfmsg *ncm;
467 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
468 flags);
469 if (nlh == NULL)
470 return -EMSGSIZE;
472 ncm = nlmsg_data(nlh);
473 ncm->ncm_family = AF_INET6;
475 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
476 goto nla_put_failure;
478 /* type -1 is used for ALL */
479 if ((type == -1 || type == NETCONFA_FORWARDING) &&
480 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
481 goto nla_put_failure;
482 #ifdef CONFIG_IPV6_MROUTE
483 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
484 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
485 devconf->mc_forwarding) < 0)
486 goto nla_put_failure;
487 #endif
488 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
489 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
490 goto nla_put_failure;
492 return nlmsg_end(skb, nlh);
494 nla_put_failure:
495 nlmsg_cancel(skb, nlh);
496 return -EMSGSIZE;
499 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
500 struct ipv6_devconf *devconf)
502 struct sk_buff *skb;
503 int err = -ENOBUFS;
505 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
506 if (skb == NULL)
507 goto errout;
509 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
510 RTM_NEWNETCONF, 0, type);
511 if (err < 0) {
512 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
513 WARN_ON(err == -EMSGSIZE);
514 kfree_skb(skb);
515 goto errout;
517 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
518 return;
519 errout:
520 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
523 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
524 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
525 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
526 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
529 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
530 struct nlmsghdr *nlh)
532 struct net *net = sock_net(in_skb->sk);
533 struct nlattr *tb[NETCONFA_MAX+1];
534 struct netconfmsg *ncm;
535 struct sk_buff *skb;
536 struct ipv6_devconf *devconf;
537 struct inet6_dev *in6_dev;
538 struct net_device *dev;
539 int ifindex;
540 int err;
542 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
543 devconf_ipv6_policy);
544 if (err < 0)
545 goto errout;
547 err = EINVAL;
548 if (!tb[NETCONFA_IFINDEX])
549 goto errout;
551 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
552 switch (ifindex) {
553 case NETCONFA_IFINDEX_ALL:
554 devconf = net->ipv6.devconf_all;
555 break;
556 case NETCONFA_IFINDEX_DEFAULT:
557 devconf = net->ipv6.devconf_dflt;
558 break;
559 default:
560 dev = __dev_get_by_index(net, ifindex);
561 if (dev == NULL)
562 goto errout;
563 in6_dev = __in6_dev_get(dev);
564 if (in6_dev == NULL)
565 goto errout;
566 devconf = &in6_dev->cnf;
567 break;
570 err = -ENOBUFS;
571 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
572 if (skb == NULL)
573 goto errout;
575 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
576 NETLINK_CB(in_skb).portid,
577 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
578 -1);
579 if (err < 0) {
580 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
581 WARN_ON(err == -EMSGSIZE);
582 kfree_skb(skb);
583 goto errout;
585 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
586 errout:
587 return err;
590 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
591 struct netlink_callback *cb)
593 struct net *net = sock_net(skb->sk);
594 int h, s_h;
595 int idx, s_idx;
596 struct net_device *dev;
597 struct inet6_dev *idev;
598 struct hlist_head *head;
600 s_h = cb->args[0];
601 s_idx = idx = cb->args[1];
603 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
604 idx = 0;
605 head = &net->dev_index_head[h];
606 rcu_read_lock();
607 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
608 net->dev_base_seq;
609 hlist_for_each_entry_rcu(dev, head, index_hlist) {
610 if (idx < s_idx)
611 goto cont;
612 idev = __in6_dev_get(dev);
613 if (!idev)
614 goto cont;
616 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
617 &idev->cnf,
618 NETLINK_CB(cb->skb).portid,
619 cb->nlh->nlmsg_seq,
620 RTM_NEWNETCONF,
621 NLM_F_MULTI,
622 -1) <= 0) {
623 rcu_read_unlock();
624 goto done;
626 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
627 cont:
628 idx++;
630 rcu_read_unlock();
632 if (h == NETDEV_HASHENTRIES) {
633 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
634 net->ipv6.devconf_all,
635 NETLINK_CB(cb->skb).portid,
636 cb->nlh->nlmsg_seq,
637 RTM_NEWNETCONF, NLM_F_MULTI,
638 -1) <= 0)
639 goto done;
640 else
641 h++;
643 if (h == NETDEV_HASHENTRIES + 1) {
644 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
645 net->ipv6.devconf_dflt,
646 NETLINK_CB(cb->skb).portid,
647 cb->nlh->nlmsg_seq,
648 RTM_NEWNETCONF, NLM_F_MULTI,
649 -1) <= 0)
650 goto done;
651 else
652 h++;
654 done:
655 cb->args[0] = h;
656 cb->args[1] = idx;
658 return skb->len;
661 #ifdef CONFIG_SYSCTL
662 static void dev_forward_change(struct inet6_dev *idev)
664 struct net_device *dev;
665 struct inet6_ifaddr *ifa;
667 if (!idev)
668 return;
669 dev = idev->dev;
670 if (idev->cnf.forwarding)
671 dev_disable_lro(dev);
672 if (dev->flags & IFF_MULTICAST) {
673 if (idev->cnf.forwarding) {
674 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
675 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
676 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
677 } else {
678 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
679 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
680 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
684 list_for_each_entry(ifa, &idev->addr_list, if_list) {
685 if (ifa->flags&IFA_F_TENTATIVE)
686 continue;
687 if (idev->cnf.forwarding)
688 addrconf_join_anycast(ifa);
689 else
690 addrconf_leave_anycast(ifa);
692 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
693 dev->ifindex, &idev->cnf);
697 static void addrconf_forward_change(struct net *net, __s32 newf)
699 struct net_device *dev;
700 struct inet6_dev *idev;
702 for_each_netdev(net, dev) {
703 idev = __in6_dev_get(dev);
704 if (idev) {
705 int changed = (!idev->cnf.forwarding) ^ (!newf);
706 idev->cnf.forwarding = newf;
707 if (changed)
708 dev_forward_change(idev);
713 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
715 struct net *net;
716 int old;
718 if (!rtnl_trylock())
719 return restart_syscall();
721 net = (struct net *)table->extra2;
722 old = *p;
723 *p = newf;
725 if (p == &net->ipv6.devconf_dflt->forwarding) {
726 if ((!newf) ^ (!old))
727 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
728 NETCONFA_IFINDEX_DEFAULT,
729 net->ipv6.devconf_dflt);
730 rtnl_unlock();
731 return 0;
734 if (p == &net->ipv6.devconf_all->forwarding) {
735 net->ipv6.devconf_dflt->forwarding = newf;
736 addrconf_forward_change(net, newf);
737 if ((!newf) ^ (!old))
738 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
739 NETCONFA_IFINDEX_ALL,
740 net->ipv6.devconf_all);
741 } else if ((!newf) ^ (!old))
742 dev_forward_change((struct inet6_dev *)table->extra1);
743 rtnl_unlock();
745 if (newf)
746 rt6_purge_dflt_routers(net);
747 return 1;
749 #endif
751 /* Nobody refers to this ifaddr, destroy it */
752 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
754 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
756 #ifdef NET_REFCNT_DEBUG
757 pr_debug("%s\n", __func__);
758 #endif
760 in6_dev_put(ifp->idev);
762 if (cancel_delayed_work(&ifp->dad_work))
763 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
764 ifp);
766 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
767 pr_warn("Freeing alive inet6 address %p\n", ifp);
768 return;
770 ip6_rt_put(ifp->rt);
772 kfree_rcu(ifp, rcu);
775 static void
776 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
778 struct list_head *p;
779 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
782 * Each device address list is sorted in order of scope -
783 * global before linklocal.
785 list_for_each(p, &idev->addr_list) {
786 struct inet6_ifaddr *ifa
787 = list_entry(p, struct inet6_ifaddr, if_list);
788 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
789 break;
792 list_add_tail(&ifp->if_list, p);
795 static u32 inet6_addr_hash(const struct in6_addr *addr)
797 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
800 /* On success it returns ifp with increased reference count */
802 static struct inet6_ifaddr *
803 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
804 const struct in6_addr *peer_addr, int pfxlen,
805 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
807 struct inet6_ifaddr *ifa = NULL;
808 struct rt6_info *rt;
809 unsigned int hash;
810 int err = 0;
811 int addr_type = ipv6_addr_type(addr);
813 if (addr_type == IPV6_ADDR_ANY ||
814 addr_type & IPV6_ADDR_MULTICAST ||
815 (!(idev->dev->flags & IFF_LOOPBACK) &&
816 addr_type & IPV6_ADDR_LOOPBACK))
817 return ERR_PTR(-EADDRNOTAVAIL);
819 rcu_read_lock_bh();
820 if (idev->dead) {
821 err = -ENODEV; /*XXX*/
822 goto out2;
825 if (idev->cnf.disable_ipv6) {
826 err = -EACCES;
827 goto out2;
830 spin_lock(&addrconf_hash_lock);
832 /* Ignore adding duplicate addresses on an interface */
833 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
834 ADBG("ipv6_add_addr: already assigned\n");
835 err = -EEXIST;
836 goto out;
839 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
841 if (ifa == NULL) {
842 ADBG("ipv6_add_addr: malloc failed\n");
843 err = -ENOBUFS;
844 goto out;
847 rt = addrconf_dst_alloc(idev, addr, false);
848 if (IS_ERR(rt)) {
849 err = PTR_ERR(rt);
850 goto out;
853 neigh_parms_data_state_setall(idev->nd_parms);
855 ifa->addr = *addr;
856 if (peer_addr)
857 ifa->peer_addr = *peer_addr;
859 spin_lock_init(&ifa->lock);
860 spin_lock_init(&ifa->state_lock);
861 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
862 INIT_HLIST_NODE(&ifa->addr_lst);
863 ifa->scope = scope;
864 ifa->prefix_len = pfxlen;
865 ifa->flags = flags | IFA_F_TENTATIVE;
866 ifa->valid_lft = valid_lft;
867 ifa->prefered_lft = prefered_lft;
868 ifa->cstamp = ifa->tstamp = jiffies;
869 ifa->tokenized = false;
871 ifa->rt = rt;
873 ifa->idev = idev;
874 in6_dev_hold(idev);
875 /* For caller */
876 in6_ifa_hold(ifa);
878 /* Add to big hash table */
879 hash = inet6_addr_hash(addr);
881 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
882 spin_unlock(&addrconf_hash_lock);
884 write_lock(&idev->lock);
885 /* Add to inet6_dev unicast addr list. */
886 ipv6_link_dev_addr(idev, ifa);
888 if (ifa->flags&IFA_F_TEMPORARY) {
889 list_add(&ifa->tmp_list, &idev->tempaddr_list);
890 in6_ifa_hold(ifa);
893 in6_ifa_hold(ifa);
894 write_unlock(&idev->lock);
895 out2:
896 rcu_read_unlock_bh();
898 if (likely(err == 0))
899 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
900 else {
901 kfree(ifa);
902 ifa = ERR_PTR(err);
905 return ifa;
906 out:
907 spin_unlock(&addrconf_hash_lock);
908 goto out2;
911 enum cleanup_prefix_rt_t {
912 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
913 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
914 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
918 * Check, whether the prefix for ifp would still need a prefix route
919 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
920 * constants.
922 * 1) we don't purge prefix if address was not permanent.
923 * prefix is managed by its own lifetime.
924 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
925 * 3) if there are no addresses, delete prefix.
926 * 4) if there are still other permanent address(es),
927 * corresponding prefix is still permanent.
928 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
929 * don't purge the prefix, assume user space is managing it.
930 * 6) otherwise, update prefix lifetime to the
931 * longest valid lifetime among the corresponding
932 * addresses on the device.
933 * Note: subsequent RA will update lifetime.
935 static enum cleanup_prefix_rt_t
936 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
938 struct inet6_ifaddr *ifa;
939 struct inet6_dev *idev = ifp->idev;
940 unsigned long lifetime;
941 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
943 *expires = jiffies;
945 list_for_each_entry(ifa, &idev->addr_list, if_list) {
946 if (ifa == ifp)
947 continue;
948 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
949 ifp->prefix_len))
950 continue;
951 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
952 return CLEANUP_PREFIX_RT_NOP;
954 action = CLEANUP_PREFIX_RT_EXPIRE;
956 spin_lock(&ifa->lock);
958 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
960 * Note: Because this address is
961 * not permanent, lifetime <
962 * LONG_MAX / HZ here.
964 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
965 *expires = ifa->tstamp + lifetime * HZ;
966 spin_unlock(&ifa->lock);
969 return action;
972 static void
973 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
975 struct rt6_info *rt;
977 rt = addrconf_get_prefix_route(&ifp->addr,
978 ifp->prefix_len,
979 ifp->idev->dev,
980 0, RTF_GATEWAY | RTF_DEFAULT);
981 if (rt) {
982 if (del_rt)
983 ip6_del_rt(rt);
984 else {
985 if (!(rt->rt6i_flags & RTF_EXPIRES))
986 rt6_set_expires(rt, expires);
987 ip6_rt_put(rt);
993 /* This function wants to get referenced ifp and releases it before return */
995 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
997 int state;
998 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
999 unsigned long expires;
1001 ASSERT_RTNL();
1003 spin_lock_bh(&ifp->state_lock);
1004 state = ifp->state;
1005 ifp->state = INET6_IFADDR_STATE_DEAD;
1006 spin_unlock_bh(&ifp->state_lock);
1008 if (state == INET6_IFADDR_STATE_DEAD)
1009 goto out;
1011 spin_lock_bh(&addrconf_hash_lock);
1012 hlist_del_init_rcu(&ifp->addr_lst);
1013 spin_unlock_bh(&addrconf_hash_lock);
1015 write_lock_bh(&ifp->idev->lock);
1017 if (ifp->flags&IFA_F_TEMPORARY) {
1018 list_del(&ifp->tmp_list);
1019 if (ifp->ifpub) {
1020 in6_ifa_put(ifp->ifpub);
1021 ifp->ifpub = NULL;
1023 __in6_ifa_put(ifp);
1026 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1027 action = check_cleanup_prefix_route(ifp, &expires);
1029 list_del_init(&ifp->if_list);
1030 __in6_ifa_put(ifp);
1032 write_unlock_bh(&ifp->idev->lock);
1034 addrconf_del_dad_work(ifp);
1036 ipv6_ifa_notify(RTM_DELADDR, ifp);
1038 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1040 if (action != CLEANUP_PREFIX_RT_NOP) {
1041 cleanup_prefix_route(ifp, expires,
1042 action == CLEANUP_PREFIX_RT_DEL);
1045 /* clean up prefsrc entries */
1046 rt6_remove_prefsrc(ifp);
1047 out:
1048 in6_ifa_put(ifp);
1051 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1053 struct inet6_dev *idev = ifp->idev;
1054 struct in6_addr addr, *tmpaddr;
1055 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1056 unsigned long regen_advance;
1057 int tmp_plen;
1058 int ret = 0;
1059 u32 addr_flags;
1060 unsigned long now = jiffies;
1062 write_lock_bh(&idev->lock);
1063 if (ift) {
1064 spin_lock_bh(&ift->lock);
1065 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1066 spin_unlock_bh(&ift->lock);
1067 tmpaddr = &addr;
1068 } else {
1069 tmpaddr = NULL;
1071 retry:
1072 in6_dev_hold(idev);
1073 if (idev->cnf.use_tempaddr <= 0) {
1074 write_unlock_bh(&idev->lock);
1075 pr_info("%s: use_tempaddr is disabled\n", __func__);
1076 in6_dev_put(idev);
1077 ret = -1;
1078 goto out;
1080 spin_lock_bh(&ifp->lock);
1081 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1082 idev->cnf.use_tempaddr = -1; /*XXX*/
1083 spin_unlock_bh(&ifp->lock);
1084 write_unlock_bh(&idev->lock);
1085 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1086 __func__);
1087 in6_dev_put(idev);
1088 ret = -1;
1089 goto out;
1091 in6_ifa_hold(ifp);
1092 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1093 __ipv6_try_regen_rndid(idev, tmpaddr);
1094 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1095 age = (now - ifp->tstamp) / HZ;
1096 tmp_valid_lft = min_t(__u32,
1097 ifp->valid_lft,
1098 idev->cnf.temp_valid_lft + age);
1099 tmp_prefered_lft = min_t(__u32,
1100 ifp->prefered_lft,
1101 idev->cnf.temp_prefered_lft + age -
1102 idev->cnf.max_desync_factor);
1103 tmp_plen = ifp->prefix_len;
1104 tmp_tstamp = ifp->tstamp;
1105 spin_unlock_bh(&ifp->lock);
1107 regen_advance = idev->cnf.regen_max_retry *
1108 idev->cnf.dad_transmits *
1109 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1110 write_unlock_bh(&idev->lock);
1112 /* A temporary address is created only if this calculated Preferred
1113 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1114 * an implementation must not create a temporary address with a zero
1115 * Preferred Lifetime.
1116 * Use age calculation as in addrconf_verify to avoid unnecessary
1117 * temporary addresses being generated.
1119 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1120 if (tmp_prefered_lft <= regen_advance + age) {
1121 in6_ifa_put(ifp);
1122 in6_dev_put(idev);
1123 ret = -1;
1124 goto out;
1127 addr_flags = IFA_F_TEMPORARY;
1128 /* set in addrconf_prefix_rcv() */
1129 if (ifp->flags & IFA_F_OPTIMISTIC)
1130 addr_flags |= IFA_F_OPTIMISTIC;
1132 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1133 ipv6_addr_scope(&addr), addr_flags,
1134 tmp_valid_lft, tmp_prefered_lft);
1135 if (IS_ERR(ift)) {
1136 in6_ifa_put(ifp);
1137 in6_dev_put(idev);
1138 pr_info("%s: retry temporary address regeneration\n", __func__);
1139 tmpaddr = &addr;
1140 write_lock_bh(&idev->lock);
1141 goto retry;
1144 spin_lock_bh(&ift->lock);
1145 ift->ifpub = ifp;
1146 ift->cstamp = now;
1147 ift->tstamp = tmp_tstamp;
1148 spin_unlock_bh(&ift->lock);
1150 addrconf_dad_start(ift);
1151 in6_ifa_put(ift);
1152 in6_dev_put(idev);
1153 out:
1154 return ret;
1158 * Choose an appropriate source address (RFC3484)
1160 enum {
1161 IPV6_SADDR_RULE_INIT = 0,
1162 IPV6_SADDR_RULE_LOCAL,
1163 IPV6_SADDR_RULE_SCOPE,
1164 IPV6_SADDR_RULE_PREFERRED,
1165 #ifdef CONFIG_IPV6_MIP6
1166 IPV6_SADDR_RULE_HOA,
1167 #endif
1168 IPV6_SADDR_RULE_OIF,
1169 IPV6_SADDR_RULE_LABEL,
1170 IPV6_SADDR_RULE_PRIVACY,
1171 IPV6_SADDR_RULE_ORCHID,
1172 IPV6_SADDR_RULE_PREFIX,
1173 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1174 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1175 #endif
1176 IPV6_SADDR_RULE_MAX
1179 struct ipv6_saddr_score {
1180 int rule;
1181 int addr_type;
1182 struct inet6_ifaddr *ifa;
1183 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1184 int scopedist;
1185 int matchlen;
1188 struct ipv6_saddr_dst {
1189 const struct in6_addr *addr;
1190 int ifindex;
1191 int scope;
1192 int label;
1193 unsigned int prefs;
1196 static inline int ipv6_saddr_preferred(int type)
1198 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1199 return 1;
1200 return 0;
1203 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1205 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1206 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1207 #else
1208 return false;
1209 #endif
1212 static int ipv6_get_saddr_eval(struct net *net,
1213 struct ipv6_saddr_score *score,
1214 struct ipv6_saddr_dst *dst,
1215 int i)
1217 int ret;
1219 if (i <= score->rule) {
1220 switch (i) {
1221 case IPV6_SADDR_RULE_SCOPE:
1222 ret = score->scopedist;
1223 break;
1224 case IPV6_SADDR_RULE_PREFIX:
1225 ret = score->matchlen;
1226 break;
1227 default:
1228 ret = !!test_bit(i, score->scorebits);
1230 goto out;
1233 switch (i) {
1234 case IPV6_SADDR_RULE_INIT:
1235 /* Rule 0: remember if hiscore is not ready yet */
1236 ret = !!score->ifa;
1237 break;
1238 case IPV6_SADDR_RULE_LOCAL:
1239 /* Rule 1: Prefer same address */
1240 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1241 break;
1242 case IPV6_SADDR_RULE_SCOPE:
1243 /* Rule 2: Prefer appropriate scope
1245 * ret
1247 * -1 | d 15
1248 * ---+--+-+---> scope
1250 * | d is scope of the destination.
1251 * B-d | \
1252 * | \ <- smaller scope is better if
1253 * B-15 | \ if scope is enough for destination.
1254 * | ret = B - scope (-1 <= scope >= d <= 15).
1255 * d-C-1 | /
1256 * |/ <- greater is better
1257 * -C / if scope is not enough for destination.
1258 * /| ret = scope - C (-1 <= d < scope <= 15).
1260 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1261 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1262 * Assume B = 0 and we get C > 29.
1264 ret = __ipv6_addr_src_scope(score->addr_type);
1265 if (ret >= dst->scope)
1266 ret = -ret;
1267 else
1268 ret -= 128; /* 30 is enough */
1269 score->scopedist = ret;
1270 break;
1271 case IPV6_SADDR_RULE_PREFERRED:
1273 /* Rule 3: Avoid deprecated and optimistic addresses */
1274 u8 avoid = IFA_F_DEPRECATED;
1276 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1277 avoid |= IFA_F_OPTIMISTIC;
1278 ret = ipv6_saddr_preferred(score->addr_type) ||
1279 !(score->ifa->flags & avoid);
1280 break;
1282 #ifdef CONFIG_IPV6_MIP6
1283 case IPV6_SADDR_RULE_HOA:
1285 /* Rule 4: Prefer home address */
1286 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1287 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1288 break;
1290 #endif
1291 case IPV6_SADDR_RULE_OIF:
1292 /* Rule 5: Prefer outgoing interface */
1293 ret = (!dst->ifindex ||
1294 dst->ifindex == score->ifa->idev->dev->ifindex);
1295 break;
1296 case IPV6_SADDR_RULE_LABEL:
1297 /* Rule 6: Prefer matching label */
1298 ret = ipv6_addr_label(net,
1299 &score->ifa->addr, score->addr_type,
1300 score->ifa->idev->dev->ifindex) == dst->label;
1301 break;
1302 case IPV6_SADDR_RULE_PRIVACY:
1304 /* Rule 7: Prefer public address
1305 * Note: prefer temporary address if use_tempaddr >= 2
1307 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1308 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1309 score->ifa->idev->cnf.use_tempaddr >= 2;
1310 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1311 break;
1313 case IPV6_SADDR_RULE_ORCHID:
1314 /* Rule 8-: Prefer ORCHID vs ORCHID or
1315 * non-ORCHID vs non-ORCHID
1317 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1318 ipv6_addr_orchid(dst->addr));
1319 break;
1320 case IPV6_SADDR_RULE_PREFIX:
1321 /* Rule 8: Use longest matching prefix */
1322 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1323 if (ret > score->ifa->prefix_len)
1324 ret = score->ifa->prefix_len;
1325 score->matchlen = ret;
1326 break;
1327 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1328 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1329 /* Optimistic addresses still have lower precedence than other
1330 * preferred addresses.
1332 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1333 break;
1334 #endif
1335 default:
1336 ret = 0;
1339 if (ret)
1340 __set_bit(i, score->scorebits);
1341 score->rule = i;
1342 out:
1343 return ret;
1346 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1347 const struct in6_addr *daddr, unsigned int prefs,
1348 struct in6_addr *saddr)
1350 struct ipv6_saddr_score scores[2],
1351 *score = &scores[0], *hiscore = &scores[1];
1352 struct ipv6_saddr_dst dst;
1353 struct net_device *dev;
1354 int dst_type;
1356 dst_type = __ipv6_addr_type(daddr);
1357 dst.addr = daddr;
1358 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1359 dst.scope = __ipv6_addr_src_scope(dst_type);
1360 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1361 dst.prefs = prefs;
1363 hiscore->rule = -1;
1364 hiscore->ifa = NULL;
1366 rcu_read_lock();
1368 for_each_netdev_rcu(net, dev) {
1369 struct inet6_dev *idev;
1371 /* Candidate Source Address (section 4)
1372 * - multicast and link-local destination address,
1373 * the set of candidate source address MUST only
1374 * include addresses assigned to interfaces
1375 * belonging to the same link as the outgoing
1376 * interface.
1377 * (- For site-local destination addresses, the
1378 * set of candidate source addresses MUST only
1379 * include addresses assigned to interfaces
1380 * belonging to the same site as the outgoing
1381 * interface.)
1383 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1384 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1385 dst.ifindex && dev->ifindex != dst.ifindex)
1386 continue;
1388 idev = __in6_dev_get(dev);
1389 if (!idev)
1390 continue;
1392 read_lock_bh(&idev->lock);
1393 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1394 int i;
1397 * - Tentative Address (RFC2462 section 5.4)
1398 * - A tentative address is not considered
1399 * "assigned to an interface" in the traditional
1400 * sense, unless it is also flagged as optimistic.
1401 * - Candidate Source Address (section 4)
1402 * - In any case, anycast addresses, multicast
1403 * addresses, and the unspecified address MUST
1404 * NOT be included in a candidate set.
1406 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1407 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1408 continue;
1410 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1412 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1413 score->addr_type & IPV6_ADDR_MULTICAST)) {
1414 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1415 dev->name);
1416 continue;
1419 score->rule = -1;
1420 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1422 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1423 int minihiscore, miniscore;
1425 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1426 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1428 if (minihiscore > miniscore) {
1429 if (i == IPV6_SADDR_RULE_SCOPE &&
1430 score->scopedist > 0) {
1432 * special case:
1433 * each remaining entry
1434 * has too small (not enough)
1435 * scope, because ifa entries
1436 * are sorted by their scope
1437 * values.
1439 goto try_nextdev;
1441 break;
1442 } else if (minihiscore < miniscore) {
1443 if (hiscore->ifa)
1444 in6_ifa_put(hiscore->ifa);
1446 in6_ifa_hold(score->ifa);
1448 swap(hiscore, score);
1450 /* restore our iterator */
1451 score->ifa = hiscore->ifa;
1453 break;
1457 try_nextdev:
1458 read_unlock_bh(&idev->lock);
1460 rcu_read_unlock();
1462 if (!hiscore->ifa)
1463 return -EADDRNOTAVAIL;
1465 *saddr = hiscore->ifa->addr;
1466 in6_ifa_put(hiscore->ifa);
1467 return 0;
1469 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1471 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1472 u32 banned_flags)
1474 struct inet6_ifaddr *ifp;
1475 int err = -EADDRNOTAVAIL;
1477 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1478 if (ifp->scope > IFA_LINK)
1479 break;
1480 if (ifp->scope == IFA_LINK &&
1481 !(ifp->flags & banned_flags)) {
1482 *addr = ifp->addr;
1483 err = 0;
1484 break;
1487 return err;
1490 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1491 u32 banned_flags)
1493 struct inet6_dev *idev;
1494 int err = -EADDRNOTAVAIL;
1496 rcu_read_lock();
1497 idev = __in6_dev_get(dev);
1498 if (idev) {
1499 read_lock_bh(&idev->lock);
1500 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1501 read_unlock_bh(&idev->lock);
1503 rcu_read_unlock();
1504 return err;
1507 static int ipv6_count_addresses(struct inet6_dev *idev)
1509 int cnt = 0;
1510 struct inet6_ifaddr *ifp;
1512 read_lock_bh(&idev->lock);
1513 list_for_each_entry(ifp, &idev->addr_list, if_list)
1514 cnt++;
1515 read_unlock_bh(&idev->lock);
1516 return cnt;
1519 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1520 const struct net_device *dev, int strict)
1522 struct inet6_ifaddr *ifp;
1523 unsigned int hash = inet6_addr_hash(addr);
1525 rcu_read_lock_bh();
1526 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1527 if (!net_eq(dev_net(ifp->idev->dev), net))
1528 continue;
1529 if (ipv6_addr_equal(&ifp->addr, addr) &&
1530 !(ifp->flags&IFA_F_TENTATIVE) &&
1531 (dev == NULL || ifp->idev->dev == dev ||
1532 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1533 rcu_read_unlock_bh();
1534 return 1;
1538 rcu_read_unlock_bh();
1539 return 0;
1541 EXPORT_SYMBOL(ipv6_chk_addr);
1543 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1544 struct net_device *dev)
1546 unsigned int hash = inet6_addr_hash(addr);
1547 struct inet6_ifaddr *ifp;
1549 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1550 if (!net_eq(dev_net(ifp->idev->dev), net))
1551 continue;
1552 if (ipv6_addr_equal(&ifp->addr, addr)) {
1553 if (dev == NULL || ifp->idev->dev == dev)
1554 return true;
1557 return false;
1560 /* Compares an address/prefix_len with addresses on device @dev.
1561 * If one is found it returns true.
1563 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1564 const unsigned int prefix_len, struct net_device *dev)
1566 struct inet6_dev *idev;
1567 struct inet6_ifaddr *ifa;
1568 bool ret = false;
1570 rcu_read_lock();
1571 idev = __in6_dev_get(dev);
1572 if (idev) {
1573 read_lock_bh(&idev->lock);
1574 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1575 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1576 if (ret)
1577 break;
1579 read_unlock_bh(&idev->lock);
1581 rcu_read_unlock();
1583 return ret;
1585 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1587 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1589 struct inet6_dev *idev;
1590 struct inet6_ifaddr *ifa;
1591 int onlink;
1593 onlink = 0;
1594 rcu_read_lock();
1595 idev = __in6_dev_get(dev);
1596 if (idev) {
1597 read_lock_bh(&idev->lock);
1598 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1599 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1600 ifa->prefix_len);
1601 if (onlink)
1602 break;
1604 read_unlock_bh(&idev->lock);
1606 rcu_read_unlock();
1607 return onlink;
1609 EXPORT_SYMBOL(ipv6_chk_prefix);
1611 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1612 struct net_device *dev, int strict)
1614 struct inet6_ifaddr *ifp, *result = NULL;
1615 unsigned int hash = inet6_addr_hash(addr);
1617 rcu_read_lock_bh();
1618 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1619 if (!net_eq(dev_net(ifp->idev->dev), net))
1620 continue;
1621 if (ipv6_addr_equal(&ifp->addr, addr)) {
1622 if (dev == NULL || ifp->idev->dev == dev ||
1623 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1624 result = ifp;
1625 in6_ifa_hold(ifp);
1626 break;
1630 rcu_read_unlock_bh();
1632 return result;
1635 /* Gets referenced address, destroys ifaddr */
1637 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1639 if (ifp->flags&IFA_F_PERMANENT) {
1640 spin_lock_bh(&ifp->lock);
1641 addrconf_del_dad_work(ifp);
1642 ifp->flags |= IFA_F_TENTATIVE;
1643 if (dad_failed)
1644 ifp->flags |= IFA_F_DADFAILED;
1645 spin_unlock_bh(&ifp->lock);
1646 if (dad_failed)
1647 ipv6_ifa_notify(0, ifp);
1648 in6_ifa_put(ifp);
1649 } else if (ifp->flags&IFA_F_TEMPORARY) {
1650 struct inet6_ifaddr *ifpub;
1651 spin_lock_bh(&ifp->lock);
1652 ifpub = ifp->ifpub;
1653 if (ifpub) {
1654 in6_ifa_hold(ifpub);
1655 spin_unlock_bh(&ifp->lock);
1656 ipv6_create_tempaddr(ifpub, ifp);
1657 in6_ifa_put(ifpub);
1658 } else {
1659 spin_unlock_bh(&ifp->lock);
1661 ipv6_del_addr(ifp);
1662 } else {
1663 ipv6_del_addr(ifp);
1667 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1669 int err = -ENOENT;
1671 spin_lock_bh(&ifp->state_lock);
1672 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1673 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1674 err = 0;
1676 spin_unlock_bh(&ifp->state_lock);
1678 return err;
1681 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1683 struct inet6_dev *idev = ifp->idev;
1685 if (addrconf_dad_end(ifp)) {
1686 in6_ifa_put(ifp);
1687 return;
1690 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1691 ifp->idev->dev->name, &ifp->addr);
1693 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1694 struct in6_addr addr;
1696 addr.s6_addr32[0] = htonl(0xfe800000);
1697 addr.s6_addr32[1] = 0;
1699 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1700 ipv6_addr_equal(&ifp->addr, &addr)) {
1701 /* DAD failed for link-local based on MAC address */
1702 idev->cnf.disable_ipv6 = 1;
1704 pr_info("%s: IPv6 being disabled!\n",
1705 ifp->idev->dev->name);
1709 spin_lock_bh(&ifp->state_lock);
1710 /* transition from _POSTDAD to _ERRDAD */
1711 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1712 spin_unlock_bh(&ifp->state_lock);
1714 addrconf_mod_dad_work(ifp, 0);
1717 /* Join to solicited addr multicast group.
1718 * caller must hold RTNL */
1719 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1721 struct in6_addr maddr;
1723 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1724 return;
1726 addrconf_addr_solict_mult(addr, &maddr);
1727 ipv6_dev_mc_inc(dev, &maddr);
1730 /* caller must hold RTNL */
1731 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1733 struct in6_addr maddr;
1735 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1736 return;
1738 addrconf_addr_solict_mult(addr, &maddr);
1739 __ipv6_dev_mc_dec(idev, &maddr);
1742 /* caller must hold RTNL */
1743 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1745 struct in6_addr addr;
1747 if (ifp->prefix_len >= 127) /* RFC 6164 */
1748 return;
1749 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1750 if (ipv6_addr_any(&addr))
1751 return;
1752 __ipv6_dev_ac_inc(ifp->idev, &addr);
1755 /* caller must hold RTNL */
1756 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1758 struct in6_addr addr;
1760 if (ifp->prefix_len >= 127) /* RFC 6164 */
1761 return;
1762 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1763 if (ipv6_addr_any(&addr))
1764 return;
1765 __ipv6_dev_ac_dec(ifp->idev, &addr);
1768 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1770 if (dev->addr_len != ETH_ALEN)
1771 return -1;
1772 memcpy(eui, dev->dev_addr, 3);
1773 memcpy(eui + 5, dev->dev_addr + 3, 3);
1776 * The zSeries OSA network cards can be shared among various
1777 * OS instances, but the OSA cards have only one MAC address.
1778 * This leads to duplicate address conflicts in conjunction
1779 * with IPv6 if more than one instance uses the same card.
1781 * The driver for these cards can deliver a unique 16-bit
1782 * identifier for each instance sharing the same card. It is
1783 * placed instead of 0xFFFE in the interface identifier. The
1784 * "u" bit of the interface identifier is not inverted in this
1785 * case. Hence the resulting interface identifier has local
1786 * scope according to RFC2373.
1788 if (dev->dev_id) {
1789 eui[3] = (dev->dev_id >> 8) & 0xFF;
1790 eui[4] = dev->dev_id & 0xFF;
1791 } else {
1792 eui[3] = 0xFF;
1793 eui[4] = 0xFE;
1794 eui[0] ^= 2;
1796 return 0;
1799 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1801 if (dev->addr_len != IEEE802154_ADDR_LEN)
1802 return -1;
1803 memcpy(eui, dev->dev_addr, 8);
1804 eui[0] ^= 2;
1805 return 0;
1808 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1810 union fwnet_hwaddr *ha;
1812 if (dev->addr_len != FWNET_ALEN)
1813 return -1;
1815 ha = (union fwnet_hwaddr *)dev->dev_addr;
1817 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1818 eui[0] ^= 2;
1819 return 0;
1822 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1824 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1825 if (dev->addr_len != ARCNET_ALEN)
1826 return -1;
1827 memset(eui, 0, 7);
1828 eui[7] = *(u8 *)dev->dev_addr;
1829 return 0;
1832 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1834 if (dev->addr_len != INFINIBAND_ALEN)
1835 return -1;
1836 memcpy(eui, dev->dev_addr + 12, 8);
1837 eui[0] |= 2;
1838 return 0;
1841 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1843 if (addr == 0)
1844 return -1;
1845 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1846 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1847 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1848 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1849 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1850 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1851 eui[1] = 0;
1852 eui[2] = 0x5E;
1853 eui[3] = 0xFE;
1854 memcpy(eui + 4, &addr, 4);
1855 return 0;
1858 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1860 if (dev->priv_flags & IFF_ISATAP)
1861 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1862 return -1;
1865 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1867 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1870 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1872 memcpy(eui, dev->perm_addr, 3);
1873 memcpy(eui + 5, dev->perm_addr + 3, 3);
1874 eui[3] = 0xFF;
1875 eui[4] = 0xFE;
1876 eui[0] ^= 2;
1877 return 0;
1880 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1882 switch (dev->type) {
1883 case ARPHRD_ETHER:
1884 case ARPHRD_FDDI:
1885 return addrconf_ifid_eui48(eui, dev);
1886 case ARPHRD_ARCNET:
1887 return addrconf_ifid_arcnet(eui, dev);
1888 case ARPHRD_INFINIBAND:
1889 return addrconf_ifid_infiniband(eui, dev);
1890 case ARPHRD_SIT:
1891 return addrconf_ifid_sit(eui, dev);
1892 case ARPHRD_IPGRE:
1893 return addrconf_ifid_gre(eui, dev);
1894 case ARPHRD_6LOWPAN:
1895 case ARPHRD_IEEE802154:
1896 return addrconf_ifid_eui64(eui, dev);
1897 case ARPHRD_IEEE1394:
1898 return addrconf_ifid_ieee1394(eui, dev);
1899 case ARPHRD_TUNNEL6:
1900 return addrconf_ifid_ip6tnl(eui, dev);
1902 return -1;
1905 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1907 int err = -1;
1908 struct inet6_ifaddr *ifp;
1910 read_lock_bh(&idev->lock);
1911 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1912 if (ifp->scope > IFA_LINK)
1913 break;
1914 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1915 memcpy(eui, ifp->addr.s6_addr+8, 8);
1916 err = 0;
1917 break;
1920 read_unlock_bh(&idev->lock);
1921 return err;
1924 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1925 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1927 regen:
1928 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1929 idev->rndid[0] &= ~0x02;
1932 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1933 * check if generated address is not inappropriate
1935 * - Reserved subnet anycast (RFC 2526)
1936 * 11111101 11....11 1xxxxxxx
1937 * - ISATAP (RFC4214) 6.1
1938 * 00-00-5E-FE-xx-xx-xx-xx
1939 * - value 0
1940 * - XXX: already assigned to an address on the device
1942 if (idev->rndid[0] == 0xfd &&
1943 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1944 (idev->rndid[7]&0x80))
1945 goto regen;
1946 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1947 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1948 goto regen;
1949 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1950 goto regen;
1954 static void ipv6_regen_rndid(unsigned long data)
1956 struct inet6_dev *idev = (struct inet6_dev *) data;
1957 unsigned long expires;
1959 rcu_read_lock_bh();
1960 write_lock_bh(&idev->lock);
1962 if (idev->dead)
1963 goto out;
1965 __ipv6_regen_rndid(idev);
1967 expires = jiffies +
1968 idev->cnf.temp_prefered_lft * HZ -
1969 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1970 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1971 idev->cnf.max_desync_factor * HZ;
1972 if (time_before(expires, jiffies)) {
1973 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1974 __func__, idev->dev->name);
1975 goto out;
1978 if (!mod_timer(&idev->regen_timer, expires))
1979 in6_dev_hold(idev);
1981 out:
1982 write_unlock_bh(&idev->lock);
1983 rcu_read_unlock_bh();
1984 in6_dev_put(idev);
1987 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1989 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1990 __ipv6_regen_rndid(idev);
1994 * Add prefix route.
1997 static void
1998 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1999 unsigned long expires, u32 flags)
2001 struct fib6_config cfg = {
2002 .fc_table = RT6_TABLE_PREFIX,
2003 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2004 .fc_ifindex = dev->ifindex,
2005 .fc_expires = expires,
2006 .fc_dst_len = plen,
2007 .fc_flags = RTF_UP | flags,
2008 .fc_nlinfo.nl_net = dev_net(dev),
2009 .fc_protocol = RTPROT_KERNEL,
2012 cfg.fc_dst = *pfx;
2014 /* Prevent useless cloning on PtP SIT.
2015 This thing is done here expecting that the whole
2016 class of non-broadcast devices need not cloning.
2018 #if IS_ENABLED(CONFIG_IPV6_SIT)
2019 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2020 cfg.fc_flags |= RTF_NONEXTHOP;
2021 #endif
2023 ip6_route_add(&cfg);
2027 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2028 int plen,
2029 const struct net_device *dev,
2030 u32 flags, u32 noflags)
2032 struct fib6_node *fn;
2033 struct rt6_info *rt = NULL;
2034 struct fib6_table *table;
2036 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2037 if (table == NULL)
2038 return NULL;
2040 read_lock_bh(&table->tb6_lock);
2041 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2042 if (!fn)
2043 goto out;
2044 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2045 if (rt->dst.dev->ifindex != dev->ifindex)
2046 continue;
2047 if ((rt->rt6i_flags & flags) != flags)
2048 continue;
2049 if ((rt->rt6i_flags & noflags) != 0)
2050 continue;
2051 dst_hold(&rt->dst);
2052 break;
2054 out:
2055 read_unlock_bh(&table->tb6_lock);
2056 return rt;
2060 /* Create "default" multicast route to the interface */
2062 static void addrconf_add_mroute(struct net_device *dev)
2064 struct fib6_config cfg = {
2065 .fc_table = RT6_TABLE_LOCAL,
2066 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2067 .fc_ifindex = dev->ifindex,
2068 .fc_dst_len = 8,
2069 .fc_flags = RTF_UP,
2070 .fc_nlinfo.nl_net = dev_net(dev),
2073 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2075 ip6_route_add(&cfg);
2078 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2080 struct inet6_dev *idev;
2082 ASSERT_RTNL();
2084 idev = ipv6_find_idev(dev);
2085 if (!idev)
2086 return ERR_PTR(-ENOBUFS);
2088 if (idev->cnf.disable_ipv6)
2089 return ERR_PTR(-EACCES);
2091 /* Add default multicast route */
2092 if (!(dev->flags & IFF_LOOPBACK))
2093 addrconf_add_mroute(dev);
2095 return idev;
2098 static void manage_tempaddrs(struct inet6_dev *idev,
2099 struct inet6_ifaddr *ifp,
2100 __u32 valid_lft, __u32 prefered_lft,
2101 bool create, unsigned long now)
2103 u32 flags;
2104 struct inet6_ifaddr *ift;
2106 read_lock_bh(&idev->lock);
2107 /* update all temporary addresses in the list */
2108 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2109 int age, max_valid, max_prefered;
2111 if (ifp != ift->ifpub)
2112 continue;
2114 /* RFC 4941 section 3.3:
2115 * If a received option will extend the lifetime of a public
2116 * address, the lifetimes of temporary addresses should
2117 * be extended, subject to the overall constraint that no
2118 * temporary addresses should ever remain "valid" or "preferred"
2119 * for a time longer than (TEMP_VALID_LIFETIME) or
2120 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2122 age = (now - ift->cstamp) / HZ;
2123 max_valid = idev->cnf.temp_valid_lft - age;
2124 if (max_valid < 0)
2125 max_valid = 0;
2127 max_prefered = idev->cnf.temp_prefered_lft -
2128 idev->cnf.max_desync_factor - age;
2129 if (max_prefered < 0)
2130 max_prefered = 0;
2132 if (valid_lft > max_valid)
2133 valid_lft = max_valid;
2135 if (prefered_lft > max_prefered)
2136 prefered_lft = max_prefered;
2138 spin_lock(&ift->lock);
2139 flags = ift->flags;
2140 ift->valid_lft = valid_lft;
2141 ift->prefered_lft = prefered_lft;
2142 ift->tstamp = now;
2143 if (prefered_lft > 0)
2144 ift->flags &= ~IFA_F_DEPRECATED;
2146 spin_unlock(&ift->lock);
2147 if (!(flags&IFA_F_TENTATIVE))
2148 ipv6_ifa_notify(0, ift);
2151 if ((create || list_empty(&idev->tempaddr_list)) &&
2152 idev->cnf.use_tempaddr > 0) {
2153 /* When a new public address is created as described
2154 * in [ADDRCONF], also create a new temporary address.
2155 * Also create a temporary address if it's enabled but
2156 * no temporary address currently exists.
2158 read_unlock_bh(&idev->lock);
2159 ipv6_create_tempaddr(ifp, NULL);
2160 } else {
2161 read_unlock_bh(&idev->lock);
2165 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2167 struct prefix_info *pinfo;
2168 __u32 valid_lft;
2169 __u32 prefered_lft;
2170 int addr_type;
2171 struct inet6_dev *in6_dev;
2172 struct net *net = dev_net(dev);
2174 pinfo = (struct prefix_info *) opt;
2176 if (len < sizeof(struct prefix_info)) {
2177 ADBG("addrconf: prefix option too short\n");
2178 return;
2182 * Validation checks ([ADDRCONF], page 19)
2185 addr_type = ipv6_addr_type(&pinfo->prefix);
2187 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2188 return;
2190 valid_lft = ntohl(pinfo->valid);
2191 prefered_lft = ntohl(pinfo->prefered);
2193 if (prefered_lft > valid_lft) {
2194 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2195 return;
2198 in6_dev = in6_dev_get(dev);
2200 if (in6_dev == NULL) {
2201 net_dbg_ratelimited("addrconf: device %s not configured\n",
2202 dev->name);
2203 return;
2207 * Two things going on here:
2208 * 1) Add routes for on-link prefixes
2209 * 2) Configure prefixes with the auto flag set
2212 if (pinfo->onlink) {
2213 struct rt6_info *rt;
2214 unsigned long rt_expires;
2216 /* Avoid arithmetic overflow. Really, we could
2217 * save rt_expires in seconds, likely valid_lft,
2218 * but it would require division in fib gc, that it
2219 * not good.
2221 if (HZ > USER_HZ)
2222 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2223 else
2224 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2226 if (addrconf_finite_timeout(rt_expires))
2227 rt_expires *= HZ;
2229 rt = addrconf_get_prefix_route(&pinfo->prefix,
2230 pinfo->prefix_len,
2231 dev,
2232 RTF_ADDRCONF | RTF_PREFIX_RT,
2233 RTF_GATEWAY | RTF_DEFAULT);
2235 if (rt) {
2236 /* Autoconf prefix route */
2237 if (valid_lft == 0) {
2238 ip6_del_rt(rt);
2239 rt = NULL;
2240 } else if (addrconf_finite_timeout(rt_expires)) {
2241 /* not infinity */
2242 rt6_set_expires(rt, jiffies + rt_expires);
2243 } else {
2244 rt6_clean_expires(rt);
2246 } else if (valid_lft) {
2247 clock_t expires = 0;
2248 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2249 if (addrconf_finite_timeout(rt_expires)) {
2250 /* not infinity */
2251 flags |= RTF_EXPIRES;
2252 expires = jiffies_to_clock_t(rt_expires);
2254 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2255 dev, expires, flags);
2257 ip6_rt_put(rt);
2260 /* Try to figure out our local address for this prefix */
2262 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2263 struct inet6_ifaddr *ifp;
2264 struct in6_addr addr;
2265 int create = 0, update_lft = 0;
2266 bool tokenized = false;
2268 if (pinfo->prefix_len == 64) {
2269 memcpy(&addr, &pinfo->prefix, 8);
2271 if (!ipv6_addr_any(&in6_dev->token)) {
2272 read_lock_bh(&in6_dev->lock);
2273 memcpy(addr.s6_addr + 8,
2274 in6_dev->token.s6_addr + 8, 8);
2275 read_unlock_bh(&in6_dev->lock);
2276 tokenized = true;
2277 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2278 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2279 in6_dev_put(in6_dev);
2280 return;
2282 goto ok;
2284 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2285 pinfo->prefix_len);
2286 in6_dev_put(in6_dev);
2287 return;
2291 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2293 if (ifp == NULL && valid_lft) {
2294 int max_addresses = in6_dev->cnf.max_addresses;
2295 u32 addr_flags = 0;
2297 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2298 if (in6_dev->cnf.optimistic_dad &&
2299 !net->ipv6.devconf_all->forwarding && sllao)
2300 addr_flags = IFA_F_OPTIMISTIC;
2301 #endif
2303 /* Do not allow to create too much of autoconfigured
2304 * addresses; this would be too easy way to crash kernel.
2306 if (!max_addresses ||
2307 ipv6_count_addresses(in6_dev) < max_addresses)
2308 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2309 pinfo->prefix_len,
2310 addr_type&IPV6_ADDR_SCOPE_MASK,
2311 addr_flags, valid_lft,
2312 prefered_lft);
2314 if (IS_ERR_OR_NULL(ifp)) {
2315 in6_dev_put(in6_dev);
2316 return;
2319 update_lft = 0;
2320 create = 1;
2321 spin_lock_bh(&ifp->lock);
2322 ifp->flags |= IFA_F_MANAGETEMPADDR;
2323 ifp->cstamp = jiffies;
2324 ifp->tokenized = tokenized;
2325 spin_unlock_bh(&ifp->lock);
2326 addrconf_dad_start(ifp);
2329 if (ifp) {
2330 u32 flags;
2331 unsigned long now;
2332 u32 stored_lft;
2334 /* update lifetime (RFC2462 5.5.3 e) */
2335 spin_lock(&ifp->lock);
2336 now = jiffies;
2337 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2338 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2339 else
2340 stored_lft = 0;
2341 if (!update_lft && !create && stored_lft) {
2342 const u32 minimum_lft = min_t(u32,
2343 stored_lft, MIN_VALID_LIFETIME);
2344 valid_lft = max(valid_lft, minimum_lft);
2346 /* RFC4862 Section 5.5.3e:
2347 * "Note that the preferred lifetime of the
2348 * corresponding address is always reset to
2349 * the Preferred Lifetime in the received
2350 * Prefix Information option, regardless of
2351 * whether the valid lifetime is also reset or
2352 * ignored."
2354 * So we should always update prefered_lft here.
2356 update_lft = 1;
2359 if (update_lft) {
2360 ifp->valid_lft = valid_lft;
2361 ifp->prefered_lft = prefered_lft;
2362 ifp->tstamp = now;
2363 flags = ifp->flags;
2364 ifp->flags &= ~IFA_F_DEPRECATED;
2365 spin_unlock(&ifp->lock);
2367 if (!(flags&IFA_F_TENTATIVE))
2368 ipv6_ifa_notify(0, ifp);
2369 } else
2370 spin_unlock(&ifp->lock);
2372 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2373 create, now);
2375 in6_ifa_put(ifp);
2376 addrconf_verify();
2379 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2380 in6_dev_put(in6_dev);
2384 * Set destination address.
2385 * Special case for SIT interfaces where we create a new "virtual"
2386 * device.
2388 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2390 struct in6_ifreq ireq;
2391 struct net_device *dev;
2392 int err = -EINVAL;
2394 rtnl_lock();
2396 err = -EFAULT;
2397 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2398 goto err_exit;
2400 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2402 err = -ENODEV;
2403 if (dev == NULL)
2404 goto err_exit;
2406 #if IS_ENABLED(CONFIG_IPV6_SIT)
2407 if (dev->type == ARPHRD_SIT) {
2408 const struct net_device_ops *ops = dev->netdev_ops;
2409 struct ifreq ifr;
2410 struct ip_tunnel_parm p;
2412 err = -EADDRNOTAVAIL;
2413 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2414 goto err_exit;
2416 memset(&p, 0, sizeof(p));
2417 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2418 p.iph.saddr = 0;
2419 p.iph.version = 4;
2420 p.iph.ihl = 5;
2421 p.iph.protocol = IPPROTO_IPV6;
2422 p.iph.ttl = 64;
2423 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2425 if (ops->ndo_do_ioctl) {
2426 mm_segment_t oldfs = get_fs();
2428 set_fs(KERNEL_DS);
2429 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2430 set_fs(oldfs);
2431 } else
2432 err = -EOPNOTSUPP;
2434 if (err == 0) {
2435 err = -ENOBUFS;
2436 dev = __dev_get_by_name(net, p.name);
2437 if (!dev)
2438 goto err_exit;
2439 err = dev_open(dev);
2442 #endif
2444 err_exit:
2445 rtnl_unlock();
2446 return err;
2450 * Manual configuration of address on an interface
2452 static int inet6_addr_add(struct net *net, int ifindex,
2453 const struct in6_addr *pfx,
2454 const struct in6_addr *peer_pfx,
2455 unsigned int plen, __u32 ifa_flags,
2456 __u32 prefered_lft, __u32 valid_lft)
2458 struct inet6_ifaddr *ifp;
2459 struct inet6_dev *idev;
2460 struct net_device *dev;
2461 int scope;
2462 u32 flags;
2463 clock_t expires;
2464 unsigned long timeout;
2466 ASSERT_RTNL();
2468 if (plen > 128)
2469 return -EINVAL;
2471 /* check the lifetime */
2472 if (!valid_lft || prefered_lft > valid_lft)
2473 return -EINVAL;
2475 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2476 return -EINVAL;
2478 dev = __dev_get_by_index(net, ifindex);
2479 if (!dev)
2480 return -ENODEV;
2482 idev = addrconf_add_dev(dev);
2483 if (IS_ERR(idev))
2484 return PTR_ERR(idev);
2486 scope = ipv6_addr_scope(pfx);
2488 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2489 if (addrconf_finite_timeout(timeout)) {
2490 expires = jiffies_to_clock_t(timeout * HZ);
2491 valid_lft = timeout;
2492 flags = RTF_EXPIRES;
2493 } else {
2494 expires = 0;
2495 flags = 0;
2496 ifa_flags |= IFA_F_PERMANENT;
2499 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2500 if (addrconf_finite_timeout(timeout)) {
2501 if (timeout == 0)
2502 ifa_flags |= IFA_F_DEPRECATED;
2503 prefered_lft = timeout;
2506 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2507 valid_lft, prefered_lft);
2509 if (!IS_ERR(ifp)) {
2510 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2511 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2512 expires, flags);
2516 * Note that section 3.1 of RFC 4429 indicates
2517 * that the Optimistic flag should not be set for
2518 * manually configured addresses
2520 addrconf_dad_start(ifp);
2521 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2522 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2523 true, jiffies);
2524 in6_ifa_put(ifp);
2525 addrconf_verify_rtnl();
2526 return 0;
2529 return PTR_ERR(ifp);
2532 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2533 const struct in6_addr *pfx, unsigned int plen)
2535 struct inet6_ifaddr *ifp;
2536 struct inet6_dev *idev;
2537 struct net_device *dev;
2539 if (plen > 128)
2540 return -EINVAL;
2542 dev = __dev_get_by_index(net, ifindex);
2543 if (!dev)
2544 return -ENODEV;
2546 idev = __in6_dev_get(dev);
2547 if (idev == NULL)
2548 return -ENXIO;
2550 read_lock_bh(&idev->lock);
2551 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2552 if (ifp->prefix_len == plen &&
2553 ipv6_addr_equal(pfx, &ifp->addr)) {
2554 in6_ifa_hold(ifp);
2555 read_unlock_bh(&idev->lock);
2557 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2558 (ifa_flags & IFA_F_MANAGETEMPADDR))
2559 manage_tempaddrs(idev, ifp, 0, 0, false,
2560 jiffies);
2561 ipv6_del_addr(ifp);
2562 addrconf_verify_rtnl();
2563 return 0;
2566 read_unlock_bh(&idev->lock);
2567 return -EADDRNOTAVAIL;
2571 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2573 struct in6_ifreq ireq;
2574 int err;
2576 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2577 return -EPERM;
2579 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2580 return -EFAULT;
2582 rtnl_lock();
2583 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2584 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2585 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2586 rtnl_unlock();
2587 return err;
2590 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2592 struct in6_ifreq ireq;
2593 int err;
2595 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2596 return -EPERM;
2598 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2599 return -EFAULT;
2601 rtnl_lock();
2602 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2603 ireq.ifr6_prefixlen);
2604 rtnl_unlock();
2605 return err;
2608 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2609 int plen, int scope)
2611 struct inet6_ifaddr *ifp;
2613 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2614 scope, IFA_F_PERMANENT,
2615 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2616 if (!IS_ERR(ifp)) {
2617 spin_lock_bh(&ifp->lock);
2618 ifp->flags &= ~IFA_F_TENTATIVE;
2619 spin_unlock_bh(&ifp->lock);
2620 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2621 in6_ifa_put(ifp);
2625 #if IS_ENABLED(CONFIG_IPV6_SIT)
2626 static void sit_add_v4_addrs(struct inet6_dev *idev)
2628 struct in6_addr addr;
2629 struct net_device *dev;
2630 struct net *net = dev_net(idev->dev);
2631 int scope, plen;
2632 u32 pflags = 0;
2634 ASSERT_RTNL();
2636 memset(&addr, 0, sizeof(struct in6_addr));
2637 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2639 if (idev->dev->flags&IFF_POINTOPOINT) {
2640 addr.s6_addr32[0] = htonl(0xfe800000);
2641 scope = IFA_LINK;
2642 plen = 64;
2643 } else {
2644 scope = IPV6_ADDR_COMPATv4;
2645 plen = 96;
2646 pflags |= RTF_NONEXTHOP;
2649 if (addr.s6_addr32[3]) {
2650 add_addr(idev, &addr, plen, scope);
2651 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2652 return;
2655 for_each_netdev(net, dev) {
2656 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2657 if (in_dev && (dev->flags & IFF_UP)) {
2658 struct in_ifaddr *ifa;
2660 int flag = scope;
2662 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2664 addr.s6_addr32[3] = ifa->ifa_local;
2666 if (ifa->ifa_scope == RT_SCOPE_LINK)
2667 continue;
2668 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2669 if (idev->dev->flags&IFF_POINTOPOINT)
2670 continue;
2671 flag |= IFA_HOST;
2674 add_addr(idev, &addr, plen, flag);
2675 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2676 pflags);
2681 #endif
2683 static void init_loopback(struct net_device *dev)
2685 struct inet6_dev *idev;
2686 struct net_device *sp_dev;
2687 struct inet6_ifaddr *sp_ifa;
2688 struct rt6_info *sp_rt;
2690 /* ::1 */
2692 ASSERT_RTNL();
2694 idev = ipv6_find_idev(dev);
2695 if (idev == NULL) {
2696 pr_debug("%s: add_dev failed\n", __func__);
2697 return;
2700 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2702 /* Add routes to other interface's IPv6 addresses */
2703 for_each_netdev(dev_net(dev), sp_dev) {
2704 if (!strcmp(sp_dev->name, dev->name))
2705 continue;
2707 idev = __in6_dev_get(sp_dev);
2708 if (!idev)
2709 continue;
2711 read_lock_bh(&idev->lock);
2712 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2714 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2715 continue;
2717 if (sp_ifa->rt) {
2718 /* This dst has been added to garbage list when
2719 * lo device down, release this obsolete dst and
2720 * reallocate a new router for ifa.
2722 if (sp_ifa->rt->dst.obsolete > 0) {
2723 ip6_rt_put(sp_ifa->rt);
2724 sp_ifa->rt = NULL;
2725 } else {
2726 continue;
2730 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2732 /* Failure cases are ignored */
2733 if (!IS_ERR(sp_rt)) {
2734 sp_ifa->rt = sp_rt;
2735 ip6_ins_rt(sp_rt);
2738 read_unlock_bh(&idev->lock);
2742 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2744 struct inet6_ifaddr *ifp;
2745 u32 addr_flags = IFA_F_PERMANENT;
2747 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2748 if (idev->cnf.optimistic_dad &&
2749 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2750 addr_flags |= IFA_F_OPTIMISTIC;
2751 #endif
2754 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2755 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2756 if (!IS_ERR(ifp)) {
2757 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2758 addrconf_dad_start(ifp);
2759 in6_ifa_put(ifp);
2763 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2765 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2766 struct in6_addr addr;
2768 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2769 /* addrconf_add_linklocal also adds a prefix_route and we
2770 * only need to care about prefix routes if ipv6_generate_eui64
2771 * couldn't generate one.
2773 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2774 addrconf_add_linklocal(idev, &addr);
2775 else if (prefix_route)
2776 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2780 static void addrconf_dev_config(struct net_device *dev)
2782 struct inet6_dev *idev;
2784 ASSERT_RTNL();
2786 if ((dev->type != ARPHRD_ETHER) &&
2787 (dev->type != ARPHRD_FDDI) &&
2788 (dev->type != ARPHRD_ARCNET) &&
2789 (dev->type != ARPHRD_INFINIBAND) &&
2790 (dev->type != ARPHRD_IEEE802154) &&
2791 (dev->type != ARPHRD_IEEE1394) &&
2792 (dev->type != ARPHRD_TUNNEL6) &&
2793 (dev->type != ARPHRD_6LOWPAN)) {
2794 /* Alas, we support only Ethernet autoconfiguration. */
2795 return;
2798 idev = addrconf_add_dev(dev);
2799 if (IS_ERR(idev))
2800 return;
2802 addrconf_addr_gen(idev, false);
2805 #if IS_ENABLED(CONFIG_IPV6_SIT)
2806 static void addrconf_sit_config(struct net_device *dev)
2808 struct inet6_dev *idev;
2810 ASSERT_RTNL();
2813 * Configure the tunnel with one of our IPv4
2814 * addresses... we should configure all of
2815 * our v4 addrs in the tunnel
2818 idev = ipv6_find_idev(dev);
2819 if (idev == NULL) {
2820 pr_debug("%s: add_dev failed\n", __func__);
2821 return;
2824 if (dev->priv_flags & IFF_ISATAP) {
2825 addrconf_addr_gen(idev, false);
2826 return;
2829 sit_add_v4_addrs(idev);
2831 if (dev->flags&IFF_POINTOPOINT)
2832 addrconf_add_mroute(dev);
2834 #endif
2836 #if IS_ENABLED(CONFIG_NET_IPGRE)
2837 static void addrconf_gre_config(struct net_device *dev)
2839 struct inet6_dev *idev;
2841 ASSERT_RTNL();
2843 idev = ipv6_find_idev(dev);
2844 if (idev == NULL) {
2845 pr_debug("%s: add_dev failed\n", __func__);
2846 return;
2849 addrconf_addr_gen(idev, true);
2851 #endif
2853 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2854 void *ptr)
2856 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2857 struct inet6_dev *idev = __in6_dev_get(dev);
2858 int run_pending = 0;
2859 int err;
2861 switch (event) {
2862 case NETDEV_REGISTER:
2863 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2864 idev = ipv6_add_dev(dev);
2865 if (IS_ERR(idev))
2866 return notifier_from_errno(PTR_ERR(idev));
2868 break;
2870 case NETDEV_UP:
2871 case NETDEV_CHANGE:
2872 if (dev->flags & IFF_SLAVE)
2873 break;
2875 if (idev && idev->cnf.disable_ipv6)
2876 break;
2878 if (event == NETDEV_UP) {
2879 if (!addrconf_qdisc_ok(dev)) {
2880 /* device is not ready yet. */
2881 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2882 dev->name);
2883 break;
2886 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2887 idev = ipv6_add_dev(dev);
2889 if (!IS_ERR_OR_NULL(idev)) {
2890 idev->if_flags |= IF_READY;
2891 run_pending = 1;
2893 } else {
2894 if (!addrconf_qdisc_ok(dev)) {
2895 /* device is still not ready. */
2896 break;
2899 if (idev) {
2900 if (idev->if_flags & IF_READY)
2901 /* device is already configured. */
2902 break;
2903 idev->if_flags |= IF_READY;
2906 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2907 dev->name);
2909 run_pending = 1;
2912 switch (dev->type) {
2913 #if IS_ENABLED(CONFIG_IPV6_SIT)
2914 case ARPHRD_SIT:
2915 addrconf_sit_config(dev);
2916 break;
2917 #endif
2918 #if IS_ENABLED(CONFIG_NET_IPGRE)
2919 case ARPHRD_IPGRE:
2920 addrconf_gre_config(dev);
2921 break;
2922 #endif
2923 case ARPHRD_LOOPBACK:
2924 init_loopback(dev);
2925 break;
2927 default:
2928 addrconf_dev_config(dev);
2929 break;
2932 if (!IS_ERR_OR_NULL(idev)) {
2933 if (run_pending)
2934 addrconf_dad_run(idev);
2937 * If the MTU changed during the interface down,
2938 * when the interface up, the changed MTU must be
2939 * reflected in the idev as well as routers.
2941 if (idev->cnf.mtu6 != dev->mtu &&
2942 dev->mtu >= IPV6_MIN_MTU) {
2943 rt6_mtu_change(dev, dev->mtu);
2944 idev->cnf.mtu6 = dev->mtu;
2946 idev->tstamp = jiffies;
2947 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2950 * If the changed mtu during down is lower than
2951 * IPV6_MIN_MTU stop IPv6 on this interface.
2953 if (dev->mtu < IPV6_MIN_MTU)
2954 addrconf_ifdown(dev, 1);
2956 break;
2958 case NETDEV_CHANGEMTU:
2959 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2960 rt6_mtu_change(dev, dev->mtu);
2961 idev->cnf.mtu6 = dev->mtu;
2962 break;
2965 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2966 idev = ipv6_add_dev(dev);
2967 if (!IS_ERR(idev))
2968 break;
2972 * if MTU under IPV6_MIN_MTU.
2973 * Stop IPv6 on this interface.
2976 case NETDEV_DOWN:
2977 case NETDEV_UNREGISTER:
2979 * Remove all addresses from this interface.
2981 addrconf_ifdown(dev, event != NETDEV_DOWN);
2982 break;
2984 case NETDEV_CHANGENAME:
2985 if (idev) {
2986 snmp6_unregister_dev(idev);
2987 addrconf_sysctl_unregister(idev);
2988 err = addrconf_sysctl_register(idev);
2989 if (err)
2990 return notifier_from_errno(err);
2991 err = snmp6_register_dev(idev);
2992 if (err) {
2993 addrconf_sysctl_unregister(idev);
2994 return notifier_from_errno(err);
2997 break;
2999 case NETDEV_PRE_TYPE_CHANGE:
3000 case NETDEV_POST_TYPE_CHANGE:
3001 addrconf_type_change(dev, event);
3002 break;
3005 return NOTIFY_OK;
3009 * addrconf module should be notified of a device going up
3011 static struct notifier_block ipv6_dev_notf = {
3012 .notifier_call = addrconf_notify,
3015 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3017 struct inet6_dev *idev;
3018 ASSERT_RTNL();
3020 idev = __in6_dev_get(dev);
3022 if (event == NETDEV_POST_TYPE_CHANGE)
3023 ipv6_mc_remap(idev);
3024 else if (event == NETDEV_PRE_TYPE_CHANGE)
3025 ipv6_mc_unmap(idev);
3028 static int addrconf_ifdown(struct net_device *dev, int how)
3030 struct net *net = dev_net(dev);
3031 struct inet6_dev *idev;
3032 struct inet6_ifaddr *ifa;
3033 int state, i;
3035 ASSERT_RTNL();
3037 rt6_ifdown(net, dev);
3038 neigh_ifdown(&nd_tbl, dev);
3040 idev = __in6_dev_get(dev);
3041 if (idev == NULL)
3042 return -ENODEV;
3045 * Step 1: remove reference to ipv6 device from parent device.
3046 * Do not dev_put!
3048 if (how) {
3049 idev->dead = 1;
3051 /* protected by rtnl_lock */
3052 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3054 /* Step 1.5: remove snmp6 entry */
3055 snmp6_unregister_dev(idev);
3059 /* Step 2: clear hash table */
3060 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3061 struct hlist_head *h = &inet6_addr_lst[i];
3063 spin_lock_bh(&addrconf_hash_lock);
3064 restart:
3065 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3066 if (ifa->idev == idev) {
3067 hlist_del_init_rcu(&ifa->addr_lst);
3068 addrconf_del_dad_work(ifa);
3069 goto restart;
3072 spin_unlock_bh(&addrconf_hash_lock);
3075 write_lock_bh(&idev->lock);
3077 addrconf_del_rs_timer(idev);
3079 /* Step 2: clear flags for stateless addrconf */
3080 if (!how)
3081 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3083 if (how && del_timer(&idev->regen_timer))
3084 in6_dev_put(idev);
3086 /* Step 3: clear tempaddr list */
3087 while (!list_empty(&idev->tempaddr_list)) {
3088 ifa = list_first_entry(&idev->tempaddr_list,
3089 struct inet6_ifaddr, tmp_list);
3090 list_del(&ifa->tmp_list);
3091 write_unlock_bh(&idev->lock);
3092 spin_lock_bh(&ifa->lock);
3094 if (ifa->ifpub) {
3095 in6_ifa_put(ifa->ifpub);
3096 ifa->ifpub = NULL;
3098 spin_unlock_bh(&ifa->lock);
3099 in6_ifa_put(ifa);
3100 write_lock_bh(&idev->lock);
3103 while (!list_empty(&idev->addr_list)) {
3104 ifa = list_first_entry(&idev->addr_list,
3105 struct inet6_ifaddr, if_list);
3106 addrconf_del_dad_work(ifa);
3108 list_del(&ifa->if_list);
3110 write_unlock_bh(&idev->lock);
3112 spin_lock_bh(&ifa->state_lock);
3113 state = ifa->state;
3114 ifa->state = INET6_IFADDR_STATE_DEAD;
3115 spin_unlock_bh(&ifa->state_lock);
3117 if (state != INET6_IFADDR_STATE_DEAD) {
3118 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3119 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3121 in6_ifa_put(ifa);
3123 write_lock_bh(&idev->lock);
3126 write_unlock_bh(&idev->lock);
3128 /* Step 5: Discard anycast and multicast list */
3129 if (how) {
3130 ipv6_ac_destroy_dev(idev);
3131 ipv6_mc_destroy_dev(idev);
3132 } else {
3133 ipv6_mc_down(idev);
3136 idev->tstamp = jiffies;
3138 /* Last: Shot the device (if unregistered) */
3139 if (how) {
3140 addrconf_sysctl_unregister(idev);
3141 neigh_parms_release(&nd_tbl, idev->nd_parms);
3142 neigh_ifdown(&nd_tbl, dev);
3143 in6_dev_put(idev);
3145 return 0;
3148 static void addrconf_rs_timer(unsigned long data)
3150 struct inet6_dev *idev = (struct inet6_dev *)data;
3151 struct net_device *dev = idev->dev;
3152 struct in6_addr lladdr;
3154 write_lock(&idev->lock);
3155 if (idev->dead || !(idev->if_flags & IF_READY))
3156 goto out;
3158 if (!ipv6_accept_ra(idev))
3159 goto out;
3161 /* Announcement received after solicitation was sent */
3162 if (idev->if_flags & IF_RA_RCVD)
3163 goto out;
3165 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3166 write_unlock(&idev->lock);
3167 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3168 ndisc_send_rs(dev, &lladdr,
3169 &in6addr_linklocal_allrouters);
3170 else
3171 goto put;
3173 write_lock(&idev->lock);
3174 /* The wait after the last probe can be shorter */
3175 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3176 idev->cnf.rtr_solicits) ?
3177 idev->cnf.rtr_solicit_delay :
3178 idev->cnf.rtr_solicit_interval);
3179 } else {
3181 * Note: we do not support deprecated "all on-link"
3182 * assumption any longer.
3184 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3187 out:
3188 write_unlock(&idev->lock);
3189 put:
3190 in6_dev_put(idev);
3194 * Duplicate Address Detection
3196 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3198 unsigned long rand_num;
3199 struct inet6_dev *idev = ifp->idev;
3201 if (ifp->flags & IFA_F_OPTIMISTIC)
3202 rand_num = 0;
3203 else
3204 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3206 ifp->dad_probes = idev->cnf.dad_transmits;
3207 addrconf_mod_dad_work(ifp, rand_num);
3210 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3212 struct inet6_dev *idev = ifp->idev;
3213 struct net_device *dev = idev->dev;
3215 addrconf_join_solict(dev, &ifp->addr);
3217 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3219 read_lock_bh(&idev->lock);
3220 spin_lock(&ifp->lock);
3221 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3222 goto out;
3224 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3225 idev->cnf.accept_dad < 1 ||
3226 !(ifp->flags&IFA_F_TENTATIVE) ||
3227 ifp->flags & IFA_F_NODAD) {
3228 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3229 spin_unlock(&ifp->lock);
3230 read_unlock_bh(&idev->lock);
3232 addrconf_dad_completed(ifp);
3233 return;
3236 if (!(idev->if_flags & IF_READY)) {
3237 spin_unlock(&ifp->lock);
3238 read_unlock_bh(&idev->lock);
3240 * If the device is not ready:
3241 * - keep it tentative if it is a permanent address.
3242 * - otherwise, kill it.
3244 in6_ifa_hold(ifp);
3245 addrconf_dad_stop(ifp, 0);
3246 return;
3250 * Optimistic nodes can start receiving
3251 * Frames right away
3253 if (ifp->flags & IFA_F_OPTIMISTIC) {
3254 ip6_ins_rt(ifp->rt);
3255 if (ipv6_use_optimistic_addr(idev)) {
3256 /* Because optimistic nodes can use this address,
3257 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3259 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3263 addrconf_dad_kick(ifp);
3264 out:
3265 spin_unlock(&ifp->lock);
3266 read_unlock_bh(&idev->lock);
3269 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3271 bool begin_dad = false;
3273 spin_lock_bh(&ifp->state_lock);
3274 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3275 ifp->state = INET6_IFADDR_STATE_PREDAD;
3276 begin_dad = true;
3278 spin_unlock_bh(&ifp->state_lock);
3280 if (begin_dad)
3281 addrconf_mod_dad_work(ifp, 0);
3284 static void addrconf_dad_work(struct work_struct *w)
3286 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3287 struct inet6_ifaddr,
3288 dad_work);
3289 struct inet6_dev *idev = ifp->idev;
3290 struct in6_addr mcaddr;
3292 enum {
3293 DAD_PROCESS,
3294 DAD_BEGIN,
3295 DAD_ABORT,
3296 } action = DAD_PROCESS;
3298 rtnl_lock();
3300 spin_lock_bh(&ifp->state_lock);
3301 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3302 action = DAD_BEGIN;
3303 ifp->state = INET6_IFADDR_STATE_DAD;
3304 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3305 action = DAD_ABORT;
3306 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3308 spin_unlock_bh(&ifp->state_lock);
3310 if (action == DAD_BEGIN) {
3311 addrconf_dad_begin(ifp);
3312 goto out;
3313 } else if (action == DAD_ABORT) {
3314 addrconf_dad_stop(ifp, 1);
3315 goto out;
3318 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3319 goto out;
3321 write_lock_bh(&idev->lock);
3322 if (idev->dead || !(idev->if_flags & IF_READY)) {
3323 write_unlock_bh(&idev->lock);
3324 goto out;
3327 spin_lock(&ifp->lock);
3328 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3329 spin_unlock(&ifp->lock);
3330 write_unlock_bh(&idev->lock);
3331 goto out;
3334 if (ifp->dad_probes == 0) {
3336 * DAD was successful
3339 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3340 spin_unlock(&ifp->lock);
3341 write_unlock_bh(&idev->lock);
3343 addrconf_dad_completed(ifp);
3345 goto out;
3348 ifp->dad_probes--;
3349 addrconf_mod_dad_work(ifp,
3350 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3351 spin_unlock(&ifp->lock);
3352 write_unlock_bh(&idev->lock);
3354 /* send a neighbour solicitation for our addr */
3355 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3356 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3357 out:
3358 in6_ifa_put(ifp);
3359 rtnl_unlock();
3362 /* ifp->idev must be at least read locked */
3363 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3365 struct inet6_ifaddr *ifpiter;
3366 struct inet6_dev *idev = ifp->idev;
3368 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3369 if (ifpiter->scope > IFA_LINK)
3370 break;
3371 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3372 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3373 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3374 IFA_F_PERMANENT)
3375 return false;
3377 return true;
3380 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3382 struct net_device *dev = ifp->idev->dev;
3383 struct in6_addr lladdr;
3384 bool send_rs, send_mld;
3386 addrconf_del_dad_work(ifp);
3389 * Configure the address for reception. Now it is valid.
3392 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3394 /* If added prefix is link local and we are prepared to process
3395 router advertisements, start sending router solicitations.
3398 read_lock_bh(&ifp->idev->lock);
3399 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3400 send_rs = send_mld &&
3401 ipv6_accept_ra(ifp->idev) &&
3402 ifp->idev->cnf.rtr_solicits > 0 &&
3403 (dev->flags&IFF_LOOPBACK) == 0;
3404 read_unlock_bh(&ifp->idev->lock);
3406 /* While dad is in progress mld report's source address is in6_addrany.
3407 * Resend with proper ll now.
3409 if (send_mld)
3410 ipv6_mc_dad_complete(ifp->idev);
3412 if (send_rs) {
3414 * If a host as already performed a random delay
3415 * [...] as part of DAD [...] there is no need
3416 * to delay again before sending the first RS
3418 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3419 return;
3420 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3422 write_lock_bh(&ifp->idev->lock);
3423 spin_lock(&ifp->lock);
3424 ifp->idev->rs_probes = 1;
3425 ifp->idev->if_flags |= IF_RS_SENT;
3426 addrconf_mod_rs_timer(ifp->idev,
3427 ifp->idev->cnf.rtr_solicit_interval);
3428 spin_unlock(&ifp->lock);
3429 write_unlock_bh(&ifp->idev->lock);
3433 static void addrconf_dad_run(struct inet6_dev *idev)
3435 struct inet6_ifaddr *ifp;
3437 read_lock_bh(&idev->lock);
3438 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3439 spin_lock(&ifp->lock);
3440 if (ifp->flags & IFA_F_TENTATIVE &&
3441 ifp->state == INET6_IFADDR_STATE_DAD)
3442 addrconf_dad_kick(ifp);
3443 spin_unlock(&ifp->lock);
3445 read_unlock_bh(&idev->lock);
3448 #ifdef CONFIG_PROC_FS
3449 struct if6_iter_state {
3450 struct seq_net_private p;
3451 int bucket;
3452 int offset;
3455 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3457 struct inet6_ifaddr *ifa = NULL;
3458 struct if6_iter_state *state = seq->private;
3459 struct net *net = seq_file_net(seq);
3460 int p = 0;
3462 /* initial bucket if pos is 0 */
3463 if (pos == 0) {
3464 state->bucket = 0;
3465 state->offset = 0;
3468 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3469 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3470 addr_lst) {
3471 if (!net_eq(dev_net(ifa->idev->dev), net))
3472 continue;
3473 /* sync with offset */
3474 if (p < state->offset) {
3475 p++;
3476 continue;
3478 state->offset++;
3479 return ifa;
3482 /* prepare for next bucket */
3483 state->offset = 0;
3484 p = 0;
3486 return NULL;
3489 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3490 struct inet6_ifaddr *ifa)
3492 struct if6_iter_state *state = seq->private;
3493 struct net *net = seq_file_net(seq);
3495 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3496 if (!net_eq(dev_net(ifa->idev->dev), net))
3497 continue;
3498 state->offset++;
3499 return ifa;
3502 while (++state->bucket < IN6_ADDR_HSIZE) {
3503 state->offset = 0;
3504 hlist_for_each_entry_rcu_bh(ifa,
3505 &inet6_addr_lst[state->bucket], addr_lst) {
3506 if (!net_eq(dev_net(ifa->idev->dev), net))
3507 continue;
3508 state->offset++;
3509 return ifa;
3513 return NULL;
3516 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3517 __acquires(rcu_bh)
3519 rcu_read_lock_bh();
3520 return if6_get_first(seq, *pos);
3523 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3525 struct inet6_ifaddr *ifa;
3527 ifa = if6_get_next(seq, v);
3528 ++*pos;
3529 return ifa;
3532 static void if6_seq_stop(struct seq_file *seq, void *v)
3533 __releases(rcu_bh)
3535 rcu_read_unlock_bh();
3538 static int if6_seq_show(struct seq_file *seq, void *v)
3540 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3541 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3542 &ifp->addr,
3543 ifp->idev->dev->ifindex,
3544 ifp->prefix_len,
3545 ifp->scope,
3546 (u8) ifp->flags,
3547 ifp->idev->dev->name);
3548 return 0;
3551 static const struct seq_operations if6_seq_ops = {
3552 .start = if6_seq_start,
3553 .next = if6_seq_next,
3554 .show = if6_seq_show,
3555 .stop = if6_seq_stop,
3558 static int if6_seq_open(struct inode *inode, struct file *file)
3560 return seq_open_net(inode, file, &if6_seq_ops,
3561 sizeof(struct if6_iter_state));
3564 static const struct file_operations if6_fops = {
3565 .owner = THIS_MODULE,
3566 .open = if6_seq_open,
3567 .read = seq_read,
3568 .llseek = seq_lseek,
3569 .release = seq_release_net,
3572 static int __net_init if6_proc_net_init(struct net *net)
3574 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3575 return -ENOMEM;
3576 return 0;
3579 static void __net_exit if6_proc_net_exit(struct net *net)
3581 remove_proc_entry("if_inet6", net->proc_net);
3584 static struct pernet_operations if6_proc_net_ops = {
3585 .init = if6_proc_net_init,
3586 .exit = if6_proc_net_exit,
3589 int __init if6_proc_init(void)
3591 return register_pernet_subsys(&if6_proc_net_ops);
3594 void if6_proc_exit(void)
3596 unregister_pernet_subsys(&if6_proc_net_ops);
3598 #endif /* CONFIG_PROC_FS */
3600 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3601 /* Check if address is a home address configured on any interface. */
3602 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3604 int ret = 0;
3605 struct inet6_ifaddr *ifp = NULL;
3606 unsigned int hash = inet6_addr_hash(addr);
3608 rcu_read_lock_bh();
3609 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3610 if (!net_eq(dev_net(ifp->idev->dev), net))
3611 continue;
3612 if (ipv6_addr_equal(&ifp->addr, addr) &&
3613 (ifp->flags & IFA_F_HOMEADDRESS)) {
3614 ret = 1;
3615 break;
3618 rcu_read_unlock_bh();
3619 return ret;
3621 #endif
3624 * Periodic address status verification
3627 static void addrconf_verify_rtnl(void)
3629 unsigned long now, next, next_sec, next_sched;
3630 struct inet6_ifaddr *ifp;
3631 int i;
3633 ASSERT_RTNL();
3635 rcu_read_lock_bh();
3636 now = jiffies;
3637 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3639 cancel_delayed_work(&addr_chk_work);
3641 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3642 restart:
3643 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3644 unsigned long age;
3646 /* When setting preferred_lft to a value not zero or
3647 * infinity, while valid_lft is infinity
3648 * IFA_F_PERMANENT has a non-infinity life time.
3650 if ((ifp->flags & IFA_F_PERMANENT) &&
3651 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3652 continue;
3654 spin_lock(&ifp->lock);
3655 /* We try to batch several events at once. */
3656 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3658 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3659 age >= ifp->valid_lft) {
3660 spin_unlock(&ifp->lock);
3661 in6_ifa_hold(ifp);
3662 ipv6_del_addr(ifp);
3663 goto restart;
3664 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3665 spin_unlock(&ifp->lock);
3666 continue;
3667 } else if (age >= ifp->prefered_lft) {
3668 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3669 int deprecate = 0;
3671 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3672 deprecate = 1;
3673 ifp->flags |= IFA_F_DEPRECATED;
3676 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3677 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3678 next = ifp->tstamp + ifp->valid_lft * HZ;
3680 spin_unlock(&ifp->lock);
3682 if (deprecate) {
3683 in6_ifa_hold(ifp);
3685 ipv6_ifa_notify(0, ifp);
3686 in6_ifa_put(ifp);
3687 goto restart;
3689 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3690 !(ifp->flags&IFA_F_TENTATIVE)) {
3691 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3692 ifp->idev->cnf.dad_transmits *
3693 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3695 if (age >= ifp->prefered_lft - regen_advance) {
3696 struct inet6_ifaddr *ifpub = ifp->ifpub;
3697 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3698 next = ifp->tstamp + ifp->prefered_lft * HZ;
3699 if (!ifp->regen_count && ifpub) {
3700 ifp->regen_count++;
3701 in6_ifa_hold(ifp);
3702 in6_ifa_hold(ifpub);
3703 spin_unlock(&ifp->lock);
3705 spin_lock(&ifpub->lock);
3706 ifpub->regen_count = 0;
3707 spin_unlock(&ifpub->lock);
3708 ipv6_create_tempaddr(ifpub, ifp);
3709 in6_ifa_put(ifpub);
3710 in6_ifa_put(ifp);
3711 goto restart;
3713 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3714 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3715 spin_unlock(&ifp->lock);
3716 } else {
3717 /* ifp->prefered_lft <= ifp->valid_lft */
3718 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3719 next = ifp->tstamp + ifp->prefered_lft * HZ;
3720 spin_unlock(&ifp->lock);
3725 next_sec = round_jiffies_up(next);
3726 next_sched = next;
3728 /* If rounded timeout is accurate enough, accept it. */
3729 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3730 next_sched = next_sec;
3732 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3733 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3734 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3736 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3737 now, next, next_sec, next_sched);
3738 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3739 rcu_read_unlock_bh();
3742 static void addrconf_verify_work(struct work_struct *w)
3744 rtnl_lock();
3745 addrconf_verify_rtnl();
3746 rtnl_unlock();
3749 static void addrconf_verify(void)
3751 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3754 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3755 struct in6_addr **peer_pfx)
3757 struct in6_addr *pfx = NULL;
3759 *peer_pfx = NULL;
3761 if (addr)
3762 pfx = nla_data(addr);
3764 if (local) {
3765 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3766 *peer_pfx = pfx;
3767 pfx = nla_data(local);
3770 return pfx;
3773 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3774 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3775 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3776 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3777 [IFA_FLAGS] = { .len = sizeof(u32) },
3780 static int
3781 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3783 struct net *net = sock_net(skb->sk);
3784 struct ifaddrmsg *ifm;
3785 struct nlattr *tb[IFA_MAX+1];
3786 struct in6_addr *pfx, *peer_pfx;
3787 u32 ifa_flags;
3788 int err;
3790 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3791 if (err < 0)
3792 return err;
3794 ifm = nlmsg_data(nlh);
3795 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3796 if (pfx == NULL)
3797 return -EINVAL;
3799 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3801 /* We ignore other flags so far. */
3802 ifa_flags &= IFA_F_MANAGETEMPADDR;
3804 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3805 ifm->ifa_prefixlen);
3808 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3809 u32 prefered_lft, u32 valid_lft)
3811 u32 flags;
3812 clock_t expires;
3813 unsigned long timeout;
3814 bool was_managetempaddr;
3815 bool had_prefixroute;
3817 ASSERT_RTNL();
3819 if (!valid_lft || (prefered_lft > valid_lft))
3820 return -EINVAL;
3822 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3823 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3824 return -EINVAL;
3826 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3827 if (addrconf_finite_timeout(timeout)) {
3828 expires = jiffies_to_clock_t(timeout * HZ);
3829 valid_lft = timeout;
3830 flags = RTF_EXPIRES;
3831 } else {
3832 expires = 0;
3833 flags = 0;
3834 ifa_flags |= IFA_F_PERMANENT;
3837 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3838 if (addrconf_finite_timeout(timeout)) {
3839 if (timeout == 0)
3840 ifa_flags |= IFA_F_DEPRECATED;
3841 prefered_lft = timeout;
3844 spin_lock_bh(&ifp->lock);
3845 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3846 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3847 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3848 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3849 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3850 IFA_F_NOPREFIXROUTE);
3851 ifp->flags |= ifa_flags;
3852 ifp->tstamp = jiffies;
3853 ifp->valid_lft = valid_lft;
3854 ifp->prefered_lft = prefered_lft;
3856 spin_unlock_bh(&ifp->lock);
3857 if (!(ifp->flags&IFA_F_TENTATIVE))
3858 ipv6_ifa_notify(0, ifp);
3860 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3861 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3862 expires, flags);
3863 } else if (had_prefixroute) {
3864 enum cleanup_prefix_rt_t action;
3865 unsigned long rt_expires;
3867 write_lock_bh(&ifp->idev->lock);
3868 action = check_cleanup_prefix_route(ifp, &rt_expires);
3869 write_unlock_bh(&ifp->idev->lock);
3871 if (action != CLEANUP_PREFIX_RT_NOP) {
3872 cleanup_prefix_route(ifp, rt_expires,
3873 action == CLEANUP_PREFIX_RT_DEL);
3877 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3878 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3879 valid_lft = prefered_lft = 0;
3880 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3881 !was_managetempaddr, jiffies);
3884 addrconf_verify_rtnl();
3886 return 0;
3889 static int
3890 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3892 struct net *net = sock_net(skb->sk);
3893 struct ifaddrmsg *ifm;
3894 struct nlattr *tb[IFA_MAX+1];
3895 struct in6_addr *pfx, *peer_pfx;
3896 struct inet6_ifaddr *ifa;
3897 struct net_device *dev;
3898 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3899 u32 ifa_flags;
3900 int err;
3902 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3903 if (err < 0)
3904 return err;
3906 ifm = nlmsg_data(nlh);
3907 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3908 if (pfx == NULL)
3909 return -EINVAL;
3911 if (tb[IFA_CACHEINFO]) {
3912 struct ifa_cacheinfo *ci;
3914 ci = nla_data(tb[IFA_CACHEINFO]);
3915 valid_lft = ci->ifa_valid;
3916 preferred_lft = ci->ifa_prefered;
3917 } else {
3918 preferred_lft = INFINITY_LIFE_TIME;
3919 valid_lft = INFINITY_LIFE_TIME;
3922 dev = __dev_get_by_index(net, ifm->ifa_index);
3923 if (dev == NULL)
3924 return -ENODEV;
3926 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3928 /* We ignore other flags so far. */
3929 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3930 IFA_F_NOPREFIXROUTE;
3932 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3933 if (ifa == NULL) {
3935 * It would be best to check for !NLM_F_CREATE here but
3936 * userspace already relies on not having to provide this.
3938 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3939 ifm->ifa_prefixlen, ifa_flags,
3940 preferred_lft, valid_lft);
3943 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3944 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3945 err = -EEXIST;
3946 else
3947 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3949 in6_ifa_put(ifa);
3951 return err;
3954 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3955 u8 scope, int ifindex)
3957 struct ifaddrmsg *ifm;
3959 ifm = nlmsg_data(nlh);
3960 ifm->ifa_family = AF_INET6;
3961 ifm->ifa_prefixlen = prefixlen;
3962 ifm->ifa_flags = flags;
3963 ifm->ifa_scope = scope;
3964 ifm->ifa_index = ifindex;
3967 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3968 unsigned long tstamp, u32 preferred, u32 valid)
3970 struct ifa_cacheinfo ci;
3972 ci.cstamp = cstamp_delta(cstamp);
3973 ci.tstamp = cstamp_delta(tstamp);
3974 ci.ifa_prefered = preferred;
3975 ci.ifa_valid = valid;
3977 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3980 static inline int rt_scope(int ifa_scope)
3982 if (ifa_scope & IFA_HOST)
3983 return RT_SCOPE_HOST;
3984 else if (ifa_scope & IFA_LINK)
3985 return RT_SCOPE_LINK;
3986 else if (ifa_scope & IFA_SITE)
3987 return RT_SCOPE_SITE;
3988 else
3989 return RT_SCOPE_UNIVERSE;
3992 static inline int inet6_ifaddr_msgsize(void)
3994 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3995 + nla_total_size(16) /* IFA_LOCAL */
3996 + nla_total_size(16) /* IFA_ADDRESS */
3997 + nla_total_size(sizeof(struct ifa_cacheinfo))
3998 + nla_total_size(4) /* IFA_FLAGS */;
4001 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4002 u32 portid, u32 seq, int event, unsigned int flags)
4004 struct nlmsghdr *nlh;
4005 u32 preferred, valid;
4007 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4008 if (nlh == NULL)
4009 return -EMSGSIZE;
4011 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4012 ifa->idev->dev->ifindex);
4014 if (!((ifa->flags&IFA_F_PERMANENT) &&
4015 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4016 preferred = ifa->prefered_lft;
4017 valid = ifa->valid_lft;
4018 if (preferred != INFINITY_LIFE_TIME) {
4019 long tval = (jiffies - ifa->tstamp)/HZ;
4020 if (preferred > tval)
4021 preferred -= tval;
4022 else
4023 preferred = 0;
4024 if (valid != INFINITY_LIFE_TIME) {
4025 if (valid > tval)
4026 valid -= tval;
4027 else
4028 valid = 0;
4031 } else {
4032 preferred = INFINITY_LIFE_TIME;
4033 valid = INFINITY_LIFE_TIME;
4036 if (!ipv6_addr_any(&ifa->peer_addr)) {
4037 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
4038 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4039 goto error;
4040 } else
4041 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4042 goto error;
4044 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4045 goto error;
4047 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4048 goto error;
4050 return nlmsg_end(skb, nlh);
4052 error:
4053 nlmsg_cancel(skb, nlh);
4054 return -EMSGSIZE;
4057 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4058 u32 portid, u32 seq, int event, u16 flags)
4060 struct nlmsghdr *nlh;
4061 u8 scope = RT_SCOPE_UNIVERSE;
4062 int ifindex = ifmca->idev->dev->ifindex;
4064 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4065 scope = RT_SCOPE_SITE;
4067 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4068 if (nlh == NULL)
4069 return -EMSGSIZE;
4071 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4072 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4073 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4074 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4075 nlmsg_cancel(skb, nlh);
4076 return -EMSGSIZE;
4079 return nlmsg_end(skb, nlh);
4082 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4083 u32 portid, u32 seq, int event, unsigned int flags)
4085 struct nlmsghdr *nlh;
4086 u8 scope = RT_SCOPE_UNIVERSE;
4087 int ifindex = ifaca->aca_idev->dev->ifindex;
4089 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4090 scope = RT_SCOPE_SITE;
4092 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4093 if (nlh == NULL)
4094 return -EMSGSIZE;
4096 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4097 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4098 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4099 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4100 nlmsg_cancel(skb, nlh);
4101 return -EMSGSIZE;
4104 return nlmsg_end(skb, nlh);
4107 enum addr_type_t {
4108 UNICAST_ADDR,
4109 MULTICAST_ADDR,
4110 ANYCAST_ADDR,
4113 /* called with rcu_read_lock() */
4114 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4115 struct netlink_callback *cb, enum addr_type_t type,
4116 int s_ip_idx, int *p_ip_idx)
4118 struct ifmcaddr6 *ifmca;
4119 struct ifacaddr6 *ifaca;
4120 int err = 1;
4121 int ip_idx = *p_ip_idx;
4123 read_lock_bh(&idev->lock);
4124 switch (type) {
4125 case UNICAST_ADDR: {
4126 struct inet6_ifaddr *ifa;
4128 /* unicast address incl. temp addr */
4129 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4130 if (++ip_idx < s_ip_idx)
4131 continue;
4132 err = inet6_fill_ifaddr(skb, ifa,
4133 NETLINK_CB(cb->skb).portid,
4134 cb->nlh->nlmsg_seq,
4135 RTM_NEWADDR,
4136 NLM_F_MULTI);
4137 if (err <= 0)
4138 break;
4139 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4141 break;
4143 case MULTICAST_ADDR:
4144 /* multicast address */
4145 for (ifmca = idev->mc_list; ifmca;
4146 ifmca = ifmca->next, ip_idx++) {
4147 if (ip_idx < s_ip_idx)
4148 continue;
4149 err = inet6_fill_ifmcaddr(skb, ifmca,
4150 NETLINK_CB(cb->skb).portid,
4151 cb->nlh->nlmsg_seq,
4152 RTM_GETMULTICAST,
4153 NLM_F_MULTI);
4154 if (err <= 0)
4155 break;
4157 break;
4158 case ANYCAST_ADDR:
4159 /* anycast address */
4160 for (ifaca = idev->ac_list; ifaca;
4161 ifaca = ifaca->aca_next, ip_idx++) {
4162 if (ip_idx < s_ip_idx)
4163 continue;
4164 err = inet6_fill_ifacaddr(skb, ifaca,
4165 NETLINK_CB(cb->skb).portid,
4166 cb->nlh->nlmsg_seq,
4167 RTM_GETANYCAST,
4168 NLM_F_MULTI);
4169 if (err <= 0)
4170 break;
4172 break;
4173 default:
4174 break;
4176 read_unlock_bh(&idev->lock);
4177 *p_ip_idx = ip_idx;
4178 return err;
4181 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4182 enum addr_type_t type)
4184 struct net *net = sock_net(skb->sk);
4185 int h, s_h;
4186 int idx, ip_idx;
4187 int s_idx, s_ip_idx;
4188 struct net_device *dev;
4189 struct inet6_dev *idev;
4190 struct hlist_head *head;
4192 s_h = cb->args[0];
4193 s_idx = idx = cb->args[1];
4194 s_ip_idx = ip_idx = cb->args[2];
4196 rcu_read_lock();
4197 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4198 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4199 idx = 0;
4200 head = &net->dev_index_head[h];
4201 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4202 if (idx < s_idx)
4203 goto cont;
4204 if (h > s_h || idx > s_idx)
4205 s_ip_idx = 0;
4206 ip_idx = 0;
4207 idev = __in6_dev_get(dev);
4208 if (!idev)
4209 goto cont;
4211 if (in6_dump_addrs(idev, skb, cb, type,
4212 s_ip_idx, &ip_idx) <= 0)
4213 goto done;
4214 cont:
4215 idx++;
4218 done:
4219 rcu_read_unlock();
4220 cb->args[0] = h;
4221 cb->args[1] = idx;
4222 cb->args[2] = ip_idx;
4224 return skb->len;
4227 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4229 enum addr_type_t type = UNICAST_ADDR;
4231 return inet6_dump_addr(skb, cb, type);
4234 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4236 enum addr_type_t type = MULTICAST_ADDR;
4238 return inet6_dump_addr(skb, cb, type);
4242 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4244 enum addr_type_t type = ANYCAST_ADDR;
4246 return inet6_dump_addr(skb, cb, type);
4249 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4251 struct net *net = sock_net(in_skb->sk);
4252 struct ifaddrmsg *ifm;
4253 struct nlattr *tb[IFA_MAX+1];
4254 struct in6_addr *addr = NULL, *peer;
4255 struct net_device *dev = NULL;
4256 struct inet6_ifaddr *ifa;
4257 struct sk_buff *skb;
4258 int err;
4260 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4261 if (err < 0)
4262 goto errout;
4264 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4265 if (addr == NULL) {
4266 err = -EINVAL;
4267 goto errout;
4270 ifm = nlmsg_data(nlh);
4271 if (ifm->ifa_index)
4272 dev = __dev_get_by_index(net, ifm->ifa_index);
4274 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4275 if (!ifa) {
4276 err = -EADDRNOTAVAIL;
4277 goto errout;
4280 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4281 if (!skb) {
4282 err = -ENOBUFS;
4283 goto errout_ifa;
4286 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4287 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4288 if (err < 0) {
4289 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4290 WARN_ON(err == -EMSGSIZE);
4291 kfree_skb(skb);
4292 goto errout_ifa;
4294 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4295 errout_ifa:
4296 in6_ifa_put(ifa);
4297 errout:
4298 return err;
4301 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4303 struct sk_buff *skb;
4304 struct net *net = dev_net(ifa->idev->dev);
4305 int err = -ENOBUFS;
4307 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4308 if (skb == NULL)
4309 goto errout;
4311 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4312 if (err < 0) {
4313 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4314 WARN_ON(err == -EMSGSIZE);
4315 kfree_skb(skb);
4316 goto errout;
4318 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4319 return;
4320 errout:
4321 if (err < 0)
4322 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4325 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4326 __s32 *array, int bytes)
4328 BUG_ON(bytes < (DEVCONF_MAX * 4));
4330 memset(array, 0, bytes);
4331 array[DEVCONF_FORWARDING] = cnf->forwarding;
4332 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4333 array[DEVCONF_MTU6] = cnf->mtu6;
4334 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4335 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4336 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4337 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4338 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4339 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4340 jiffies_to_msecs(cnf->rtr_solicit_interval);
4341 array[DEVCONF_RTR_SOLICIT_DELAY] =
4342 jiffies_to_msecs(cnf->rtr_solicit_delay);
4343 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4344 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4345 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4346 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4347 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4348 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4349 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4350 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4351 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4352 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4353 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4354 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4355 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4356 #ifdef CONFIG_IPV6_ROUTER_PREF
4357 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4358 array[DEVCONF_RTR_PROBE_INTERVAL] =
4359 jiffies_to_msecs(cnf->rtr_probe_interval);
4360 #ifdef CONFIG_IPV6_ROUTE_INFO
4361 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4362 #endif
4363 #endif
4364 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4365 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4366 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4367 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4368 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4369 #endif
4370 #ifdef CONFIG_IPV6_MROUTE
4371 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4372 #endif
4373 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4374 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4375 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4376 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4377 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4378 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4381 static inline size_t inet6_ifla6_size(void)
4383 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4384 + nla_total_size(sizeof(struct ifla_cacheinfo))
4385 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4386 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4387 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4388 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4391 static inline size_t inet6_if_nlmsg_size(void)
4393 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4394 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4395 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4396 + nla_total_size(4) /* IFLA_MTU */
4397 + nla_total_size(4) /* IFLA_LINK */
4398 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4401 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4402 int items, int bytes)
4404 int i;
4405 int pad = bytes - sizeof(u64) * items;
4406 BUG_ON(pad < 0);
4408 /* Use put_unaligned() because stats may not be aligned for u64. */
4409 put_unaligned(items, &stats[0]);
4410 for (i = 1; i < items; i++)
4411 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4413 memset(&stats[items], 0, pad);
4416 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4417 int items, int bytes, size_t syncpoff)
4419 int i;
4420 int pad = bytes - sizeof(u64) * items;
4421 BUG_ON(pad < 0);
4423 /* Use put_unaligned() because stats may not be aligned for u64. */
4424 put_unaligned(items, &stats[0]);
4425 for (i = 1; i < items; i++)
4426 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4428 memset(&stats[items], 0, pad);
4431 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4432 int bytes)
4434 switch (attrtype) {
4435 case IFLA_INET6_STATS:
4436 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4437 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4438 break;
4439 case IFLA_INET6_ICMP6STATS:
4440 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4441 break;
4445 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4447 struct nlattr *nla;
4448 struct ifla_cacheinfo ci;
4450 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4451 goto nla_put_failure;
4452 ci.max_reasm_len = IPV6_MAXPLEN;
4453 ci.tstamp = cstamp_delta(idev->tstamp);
4454 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4455 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4456 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4457 goto nla_put_failure;
4458 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4459 if (nla == NULL)
4460 goto nla_put_failure;
4461 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4463 /* XXX - MC not implemented */
4465 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4466 if (nla == NULL)
4467 goto nla_put_failure;
4468 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4470 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4471 if (nla == NULL)
4472 goto nla_put_failure;
4473 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4475 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4476 if (nla == NULL)
4477 goto nla_put_failure;
4479 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4480 goto nla_put_failure;
4482 read_lock_bh(&idev->lock);
4483 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4484 read_unlock_bh(&idev->lock);
4486 return 0;
4488 nla_put_failure:
4489 return -EMSGSIZE;
4492 static size_t inet6_get_link_af_size(const struct net_device *dev)
4494 if (!__in6_dev_get(dev))
4495 return 0;
4497 return inet6_ifla6_size();
4500 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4502 struct inet6_dev *idev = __in6_dev_get(dev);
4504 if (!idev)
4505 return -ENODATA;
4507 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4508 return -EMSGSIZE;
4510 return 0;
4513 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4515 struct inet6_ifaddr *ifp;
4516 struct net_device *dev = idev->dev;
4517 bool update_rs = false;
4518 struct in6_addr ll_addr;
4520 ASSERT_RTNL();
4522 if (token == NULL)
4523 return -EINVAL;
4524 if (ipv6_addr_any(token))
4525 return -EINVAL;
4526 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4527 return -EINVAL;
4528 if (!ipv6_accept_ra(idev))
4529 return -EINVAL;
4530 if (idev->cnf.rtr_solicits <= 0)
4531 return -EINVAL;
4533 write_lock_bh(&idev->lock);
4535 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4536 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4538 write_unlock_bh(&idev->lock);
4540 if (!idev->dead && (idev->if_flags & IF_READY) &&
4541 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4542 IFA_F_OPTIMISTIC)) {
4544 /* If we're not ready, then normal ifup will take care
4545 * of this. Otherwise, we need to request our rs here.
4547 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4548 update_rs = true;
4551 write_lock_bh(&idev->lock);
4553 if (update_rs) {
4554 idev->if_flags |= IF_RS_SENT;
4555 idev->rs_probes = 1;
4556 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4559 /* Well, that's kinda nasty ... */
4560 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4561 spin_lock(&ifp->lock);
4562 if (ifp->tokenized) {
4563 ifp->valid_lft = 0;
4564 ifp->prefered_lft = 0;
4566 spin_unlock(&ifp->lock);
4569 write_unlock_bh(&idev->lock);
4570 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4571 addrconf_verify_rtnl();
4572 return 0;
4575 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4577 int err = -EINVAL;
4578 struct inet6_dev *idev = __in6_dev_get(dev);
4579 struct nlattr *tb[IFLA_INET6_MAX + 1];
4581 if (!idev)
4582 return -EAFNOSUPPORT;
4584 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4585 BUG();
4587 if (tb[IFLA_INET6_TOKEN]) {
4588 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4589 if (err)
4590 return err;
4593 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4594 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4596 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4597 mode != IN6_ADDR_GEN_MODE_NONE)
4598 return -EINVAL;
4599 idev->addr_gen_mode = mode;
4600 err = 0;
4603 return err;
4606 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4607 u32 portid, u32 seq, int event, unsigned int flags)
4609 struct net_device *dev = idev->dev;
4610 struct ifinfomsg *hdr;
4611 struct nlmsghdr *nlh;
4612 void *protoinfo;
4614 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4615 if (nlh == NULL)
4616 return -EMSGSIZE;
4618 hdr = nlmsg_data(nlh);
4619 hdr->ifi_family = AF_INET6;
4620 hdr->__ifi_pad = 0;
4621 hdr->ifi_type = dev->type;
4622 hdr->ifi_index = dev->ifindex;
4623 hdr->ifi_flags = dev_get_flags(dev);
4624 hdr->ifi_change = 0;
4626 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4627 (dev->addr_len &&
4628 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4629 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4630 (dev->ifindex != dev->iflink &&
4631 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4632 goto nla_put_failure;
4633 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4634 if (protoinfo == NULL)
4635 goto nla_put_failure;
4637 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4638 goto nla_put_failure;
4640 nla_nest_end(skb, protoinfo);
4641 return nlmsg_end(skb, nlh);
4643 nla_put_failure:
4644 nlmsg_cancel(skb, nlh);
4645 return -EMSGSIZE;
4648 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4650 struct net *net = sock_net(skb->sk);
4651 int h, s_h;
4652 int idx = 0, s_idx;
4653 struct net_device *dev;
4654 struct inet6_dev *idev;
4655 struct hlist_head *head;
4657 s_h = cb->args[0];
4658 s_idx = cb->args[1];
4660 rcu_read_lock();
4661 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4662 idx = 0;
4663 head = &net->dev_index_head[h];
4664 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4665 if (idx < s_idx)
4666 goto cont;
4667 idev = __in6_dev_get(dev);
4668 if (!idev)
4669 goto cont;
4670 if (inet6_fill_ifinfo(skb, idev,
4671 NETLINK_CB(cb->skb).portid,
4672 cb->nlh->nlmsg_seq,
4673 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4674 goto out;
4675 cont:
4676 idx++;
4679 out:
4680 rcu_read_unlock();
4681 cb->args[1] = idx;
4682 cb->args[0] = h;
4684 return skb->len;
4687 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4689 struct sk_buff *skb;
4690 struct net *net = dev_net(idev->dev);
4691 int err = -ENOBUFS;
4693 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4694 if (skb == NULL)
4695 goto errout;
4697 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4698 if (err < 0) {
4699 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4700 WARN_ON(err == -EMSGSIZE);
4701 kfree_skb(skb);
4702 goto errout;
4704 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4705 return;
4706 errout:
4707 if (err < 0)
4708 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4711 static inline size_t inet6_prefix_nlmsg_size(void)
4713 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4714 + nla_total_size(sizeof(struct in6_addr))
4715 + nla_total_size(sizeof(struct prefix_cacheinfo));
4718 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4719 struct prefix_info *pinfo, u32 portid, u32 seq,
4720 int event, unsigned int flags)
4722 struct prefixmsg *pmsg;
4723 struct nlmsghdr *nlh;
4724 struct prefix_cacheinfo ci;
4726 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4727 if (nlh == NULL)
4728 return -EMSGSIZE;
4730 pmsg = nlmsg_data(nlh);
4731 pmsg->prefix_family = AF_INET6;
4732 pmsg->prefix_pad1 = 0;
4733 pmsg->prefix_pad2 = 0;
4734 pmsg->prefix_ifindex = idev->dev->ifindex;
4735 pmsg->prefix_len = pinfo->prefix_len;
4736 pmsg->prefix_type = pinfo->type;
4737 pmsg->prefix_pad3 = 0;
4738 pmsg->prefix_flags = 0;
4739 if (pinfo->onlink)
4740 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4741 if (pinfo->autoconf)
4742 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4744 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4745 goto nla_put_failure;
4746 ci.preferred_time = ntohl(pinfo->prefered);
4747 ci.valid_time = ntohl(pinfo->valid);
4748 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4749 goto nla_put_failure;
4750 return nlmsg_end(skb, nlh);
4752 nla_put_failure:
4753 nlmsg_cancel(skb, nlh);
4754 return -EMSGSIZE;
4757 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4758 struct prefix_info *pinfo)
4760 struct sk_buff *skb;
4761 struct net *net = dev_net(idev->dev);
4762 int err = -ENOBUFS;
4764 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4765 if (skb == NULL)
4766 goto errout;
4768 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4769 if (err < 0) {
4770 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4771 WARN_ON(err == -EMSGSIZE);
4772 kfree_skb(skb);
4773 goto errout;
4775 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4776 return;
4777 errout:
4778 if (err < 0)
4779 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4782 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4784 struct net *net = dev_net(ifp->idev->dev);
4786 if (event)
4787 ASSERT_RTNL();
4789 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4791 switch (event) {
4792 case RTM_NEWADDR:
4794 * If the address was optimistic
4795 * we inserted the route at the start of
4796 * our DAD process, so we don't need
4797 * to do it again
4799 if (!(ifp->rt->rt6i_node))
4800 ip6_ins_rt(ifp->rt);
4801 if (ifp->idev->cnf.forwarding)
4802 addrconf_join_anycast(ifp);
4803 if (!ipv6_addr_any(&ifp->peer_addr))
4804 addrconf_prefix_route(&ifp->peer_addr, 128,
4805 ifp->idev->dev, 0, 0);
4806 break;
4807 case RTM_DELADDR:
4808 if (ifp->idev->cnf.forwarding)
4809 addrconf_leave_anycast(ifp);
4810 addrconf_leave_solict(ifp->idev, &ifp->addr);
4811 if (!ipv6_addr_any(&ifp->peer_addr)) {
4812 struct rt6_info *rt;
4814 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
4815 ifp->idev->dev, 0, 0);
4816 if (rt && ip6_del_rt(rt))
4817 dst_free(&rt->dst);
4819 dst_hold(&ifp->rt->dst);
4821 if (ip6_del_rt(ifp->rt))
4822 dst_free(&ifp->rt->dst);
4824 rt_genid_bump_ipv6(net);
4825 break;
4827 atomic_inc(&net->ipv6.dev_addr_genid);
4830 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4832 rcu_read_lock_bh();
4833 if (likely(ifp->idev->dead == 0))
4834 __ipv6_ifa_notify(event, ifp);
4835 rcu_read_unlock_bh();
4838 #ifdef CONFIG_SYSCTL
4840 static
4841 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4842 void __user *buffer, size_t *lenp, loff_t *ppos)
4844 int *valp = ctl->data;
4845 int val = *valp;
4846 loff_t pos = *ppos;
4847 struct ctl_table lctl;
4848 int ret;
4851 * ctl->data points to idev->cnf.forwarding, we should
4852 * not modify it until we get the rtnl lock.
4854 lctl = *ctl;
4855 lctl.data = &val;
4857 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4859 if (write)
4860 ret = addrconf_fixup_forwarding(ctl, valp, val);
4861 if (ret)
4862 *ppos = pos;
4863 return ret;
4866 static void dev_disable_change(struct inet6_dev *idev)
4868 struct netdev_notifier_info info;
4870 if (!idev || !idev->dev)
4871 return;
4873 netdev_notifier_info_init(&info, idev->dev);
4874 if (idev->cnf.disable_ipv6)
4875 addrconf_notify(NULL, NETDEV_DOWN, &info);
4876 else
4877 addrconf_notify(NULL, NETDEV_UP, &info);
4880 static void addrconf_disable_change(struct net *net, __s32 newf)
4882 struct net_device *dev;
4883 struct inet6_dev *idev;
4885 rcu_read_lock();
4886 for_each_netdev_rcu(net, dev) {
4887 idev = __in6_dev_get(dev);
4888 if (idev) {
4889 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4890 idev->cnf.disable_ipv6 = newf;
4891 if (changed)
4892 dev_disable_change(idev);
4895 rcu_read_unlock();
4898 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4900 struct net *net;
4901 int old;
4903 if (!rtnl_trylock())
4904 return restart_syscall();
4906 net = (struct net *)table->extra2;
4907 old = *p;
4908 *p = newf;
4910 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4911 rtnl_unlock();
4912 return 0;
4915 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4916 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4917 addrconf_disable_change(net, newf);
4918 } else if ((!newf) ^ (!old))
4919 dev_disable_change((struct inet6_dev *)table->extra1);
4921 rtnl_unlock();
4922 return 0;
4925 static
4926 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4927 void __user *buffer, size_t *lenp, loff_t *ppos)
4929 int *valp = ctl->data;
4930 int val = *valp;
4931 loff_t pos = *ppos;
4932 struct ctl_table lctl;
4933 int ret;
4936 * ctl->data points to idev->cnf.disable_ipv6, we should
4937 * not modify it until we get the rtnl lock.
4939 lctl = *ctl;
4940 lctl.data = &val;
4942 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4944 if (write)
4945 ret = addrconf_disable_ipv6(ctl, valp, val);
4946 if (ret)
4947 *ppos = pos;
4948 return ret;
4951 static
4952 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4953 void __user *buffer, size_t *lenp, loff_t *ppos)
4955 int *valp = ctl->data;
4956 int ret;
4957 int old, new;
4959 old = *valp;
4960 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4961 new = *valp;
4963 if (write && old != new) {
4964 struct net *net = ctl->extra2;
4966 if (!rtnl_trylock())
4967 return restart_syscall();
4969 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4970 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4971 NETCONFA_IFINDEX_DEFAULT,
4972 net->ipv6.devconf_dflt);
4973 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4974 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4975 NETCONFA_IFINDEX_ALL,
4976 net->ipv6.devconf_all);
4977 else {
4978 struct inet6_dev *idev = ctl->extra1;
4980 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4981 idev->dev->ifindex,
4982 &idev->cnf);
4984 rtnl_unlock();
4987 return ret;
4991 static struct addrconf_sysctl_table
4993 struct ctl_table_header *sysctl_header;
4994 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4995 } addrconf_sysctl __read_mostly = {
4996 .sysctl_header = NULL,
4997 .addrconf_vars = {
4999 .procname = "forwarding",
5000 .data = &ipv6_devconf.forwarding,
5001 .maxlen = sizeof(int),
5002 .mode = 0644,
5003 .proc_handler = addrconf_sysctl_forward,
5006 .procname = "hop_limit",
5007 .data = &ipv6_devconf.hop_limit,
5008 .maxlen = sizeof(int),
5009 .mode = 0644,
5010 .proc_handler = proc_dointvec,
5013 .procname = "mtu",
5014 .data = &ipv6_devconf.mtu6,
5015 .maxlen = sizeof(int),
5016 .mode = 0644,
5017 .proc_handler = proc_dointvec,
5020 .procname = "accept_ra",
5021 .data = &ipv6_devconf.accept_ra,
5022 .maxlen = sizeof(int),
5023 .mode = 0644,
5024 .proc_handler = proc_dointvec,
5027 .procname = "accept_redirects",
5028 .data = &ipv6_devconf.accept_redirects,
5029 .maxlen = sizeof(int),
5030 .mode = 0644,
5031 .proc_handler = proc_dointvec,
5034 .procname = "autoconf",
5035 .data = &ipv6_devconf.autoconf,
5036 .maxlen = sizeof(int),
5037 .mode = 0644,
5038 .proc_handler = proc_dointvec,
5041 .procname = "dad_transmits",
5042 .data = &ipv6_devconf.dad_transmits,
5043 .maxlen = sizeof(int),
5044 .mode = 0644,
5045 .proc_handler = proc_dointvec,
5048 .procname = "router_solicitations",
5049 .data = &ipv6_devconf.rtr_solicits,
5050 .maxlen = sizeof(int),
5051 .mode = 0644,
5052 .proc_handler = proc_dointvec,
5055 .procname = "router_solicitation_interval",
5056 .data = &ipv6_devconf.rtr_solicit_interval,
5057 .maxlen = sizeof(int),
5058 .mode = 0644,
5059 .proc_handler = proc_dointvec_jiffies,
5062 .procname = "router_solicitation_delay",
5063 .data = &ipv6_devconf.rtr_solicit_delay,
5064 .maxlen = sizeof(int),
5065 .mode = 0644,
5066 .proc_handler = proc_dointvec_jiffies,
5069 .procname = "force_mld_version",
5070 .data = &ipv6_devconf.force_mld_version,
5071 .maxlen = sizeof(int),
5072 .mode = 0644,
5073 .proc_handler = proc_dointvec,
5076 .procname = "mldv1_unsolicited_report_interval",
5077 .data =
5078 &ipv6_devconf.mldv1_unsolicited_report_interval,
5079 .maxlen = sizeof(int),
5080 .mode = 0644,
5081 .proc_handler = proc_dointvec_ms_jiffies,
5084 .procname = "mldv2_unsolicited_report_interval",
5085 .data =
5086 &ipv6_devconf.mldv2_unsolicited_report_interval,
5087 .maxlen = sizeof(int),
5088 .mode = 0644,
5089 .proc_handler = proc_dointvec_ms_jiffies,
5092 .procname = "use_tempaddr",
5093 .data = &ipv6_devconf.use_tempaddr,
5094 .maxlen = sizeof(int),
5095 .mode = 0644,
5096 .proc_handler = proc_dointvec,
5099 .procname = "temp_valid_lft",
5100 .data = &ipv6_devconf.temp_valid_lft,
5101 .maxlen = sizeof(int),
5102 .mode = 0644,
5103 .proc_handler = proc_dointvec,
5106 .procname = "temp_prefered_lft",
5107 .data = &ipv6_devconf.temp_prefered_lft,
5108 .maxlen = sizeof(int),
5109 .mode = 0644,
5110 .proc_handler = proc_dointvec,
5113 .procname = "regen_max_retry",
5114 .data = &ipv6_devconf.regen_max_retry,
5115 .maxlen = sizeof(int),
5116 .mode = 0644,
5117 .proc_handler = proc_dointvec,
5120 .procname = "max_desync_factor",
5121 .data = &ipv6_devconf.max_desync_factor,
5122 .maxlen = sizeof(int),
5123 .mode = 0644,
5124 .proc_handler = proc_dointvec,
5127 .procname = "max_addresses",
5128 .data = &ipv6_devconf.max_addresses,
5129 .maxlen = sizeof(int),
5130 .mode = 0644,
5131 .proc_handler = proc_dointvec,
5134 .procname = "accept_ra_defrtr",
5135 .data = &ipv6_devconf.accept_ra_defrtr,
5136 .maxlen = sizeof(int),
5137 .mode = 0644,
5138 .proc_handler = proc_dointvec,
5141 .procname = "accept_ra_pinfo",
5142 .data = &ipv6_devconf.accept_ra_pinfo,
5143 .maxlen = sizeof(int),
5144 .mode = 0644,
5145 .proc_handler = proc_dointvec,
5147 #ifdef CONFIG_IPV6_ROUTER_PREF
5149 .procname = "accept_ra_rtr_pref",
5150 .data = &ipv6_devconf.accept_ra_rtr_pref,
5151 .maxlen = sizeof(int),
5152 .mode = 0644,
5153 .proc_handler = proc_dointvec,
5156 .procname = "router_probe_interval",
5157 .data = &ipv6_devconf.rtr_probe_interval,
5158 .maxlen = sizeof(int),
5159 .mode = 0644,
5160 .proc_handler = proc_dointvec_jiffies,
5162 #ifdef CONFIG_IPV6_ROUTE_INFO
5164 .procname = "accept_ra_rt_info_max_plen",
5165 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5166 .maxlen = sizeof(int),
5167 .mode = 0644,
5168 .proc_handler = proc_dointvec,
5170 #endif
5171 #endif
5173 .procname = "proxy_ndp",
5174 .data = &ipv6_devconf.proxy_ndp,
5175 .maxlen = sizeof(int),
5176 .mode = 0644,
5177 .proc_handler = addrconf_sysctl_proxy_ndp,
5180 .procname = "accept_source_route",
5181 .data = &ipv6_devconf.accept_source_route,
5182 .maxlen = sizeof(int),
5183 .mode = 0644,
5184 .proc_handler = proc_dointvec,
5186 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5188 .procname = "optimistic_dad",
5189 .data = &ipv6_devconf.optimistic_dad,
5190 .maxlen = sizeof(int),
5191 .mode = 0644,
5192 .proc_handler = proc_dointvec,
5196 .procname = "use_optimistic",
5197 .data = &ipv6_devconf.use_optimistic,
5198 .maxlen = sizeof(int),
5199 .mode = 0644,
5200 .proc_handler = proc_dointvec,
5203 #endif
5204 #ifdef CONFIG_IPV6_MROUTE
5206 .procname = "mc_forwarding",
5207 .data = &ipv6_devconf.mc_forwarding,
5208 .maxlen = sizeof(int),
5209 .mode = 0444,
5210 .proc_handler = proc_dointvec,
5212 #endif
5214 .procname = "disable_ipv6",
5215 .data = &ipv6_devconf.disable_ipv6,
5216 .maxlen = sizeof(int),
5217 .mode = 0644,
5218 .proc_handler = addrconf_sysctl_disable,
5221 .procname = "accept_dad",
5222 .data = &ipv6_devconf.accept_dad,
5223 .maxlen = sizeof(int),
5224 .mode = 0644,
5225 .proc_handler = proc_dointvec,
5228 .procname = "force_tllao",
5229 .data = &ipv6_devconf.force_tllao,
5230 .maxlen = sizeof(int),
5231 .mode = 0644,
5232 .proc_handler = proc_dointvec
5235 .procname = "ndisc_notify",
5236 .data = &ipv6_devconf.ndisc_notify,
5237 .maxlen = sizeof(int),
5238 .mode = 0644,
5239 .proc_handler = proc_dointvec
5242 .procname = "suppress_frag_ndisc",
5243 .data = &ipv6_devconf.suppress_frag_ndisc,
5244 .maxlen = sizeof(int),
5245 .mode = 0644,
5246 .proc_handler = proc_dointvec
5249 .procname = "accept_ra_from_local",
5250 .data = &ipv6_devconf.accept_ra_from_local,
5251 .maxlen = sizeof(int),
5252 .mode = 0644,
5253 .proc_handler = proc_dointvec,
5256 /* sentinel */
5261 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5262 struct inet6_dev *idev, struct ipv6_devconf *p)
5264 int i;
5265 struct addrconf_sysctl_table *t;
5266 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5268 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5269 if (t == NULL)
5270 goto out;
5272 for (i = 0; t->addrconf_vars[i].data; i++) {
5273 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5274 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5275 t->addrconf_vars[i].extra2 = net;
5278 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5280 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5281 if (t->sysctl_header == NULL)
5282 goto free;
5284 p->sysctl = t;
5285 return 0;
5287 free:
5288 kfree(t);
5289 out:
5290 return -ENOBUFS;
5293 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5295 struct addrconf_sysctl_table *t;
5297 if (p->sysctl == NULL)
5298 return;
5300 t = p->sysctl;
5301 p->sysctl = NULL;
5302 unregister_net_sysctl_table(t->sysctl_header);
5303 kfree(t);
5306 static int addrconf_sysctl_register(struct inet6_dev *idev)
5308 int err;
5310 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5311 return -EINVAL;
5313 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5314 &ndisc_ifinfo_sysctl_change);
5315 if (err)
5316 return err;
5317 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5318 idev, &idev->cnf);
5319 if (err)
5320 neigh_sysctl_unregister(idev->nd_parms);
5322 return err;
5325 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5327 __addrconf_sysctl_unregister(&idev->cnf);
5328 neigh_sysctl_unregister(idev->nd_parms);
5332 #endif
5334 static int __net_init addrconf_init_net(struct net *net)
5336 int err = -ENOMEM;
5337 struct ipv6_devconf *all, *dflt;
5339 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5340 if (all == NULL)
5341 goto err_alloc_all;
5343 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5344 if (dflt == NULL)
5345 goto err_alloc_dflt;
5347 /* these will be inherited by all namespaces */
5348 dflt->autoconf = ipv6_defaults.autoconf;
5349 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5351 net->ipv6.devconf_all = all;
5352 net->ipv6.devconf_dflt = dflt;
5354 #ifdef CONFIG_SYSCTL
5355 err = __addrconf_sysctl_register(net, "all", NULL, all);
5356 if (err < 0)
5357 goto err_reg_all;
5359 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5360 if (err < 0)
5361 goto err_reg_dflt;
5362 #endif
5363 return 0;
5365 #ifdef CONFIG_SYSCTL
5366 err_reg_dflt:
5367 __addrconf_sysctl_unregister(all);
5368 err_reg_all:
5369 kfree(dflt);
5370 #endif
5371 err_alloc_dflt:
5372 kfree(all);
5373 err_alloc_all:
5374 return err;
5377 static void __net_exit addrconf_exit_net(struct net *net)
5379 #ifdef CONFIG_SYSCTL
5380 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5381 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5382 #endif
5383 kfree(net->ipv6.devconf_dflt);
5384 kfree(net->ipv6.devconf_all);
5387 static struct pernet_operations addrconf_ops = {
5388 .init = addrconf_init_net,
5389 .exit = addrconf_exit_net,
5392 static struct rtnl_af_ops inet6_ops = {
5393 .family = AF_INET6,
5394 .fill_link_af = inet6_fill_link_af,
5395 .get_link_af_size = inet6_get_link_af_size,
5396 .set_link_af = inet6_set_link_af,
5400 * Init / cleanup code
5403 int __init addrconf_init(void)
5405 struct inet6_dev *idev;
5406 int i, err;
5408 err = ipv6_addr_label_init();
5409 if (err < 0) {
5410 pr_crit("%s: cannot initialize default policy table: %d\n",
5411 __func__, err);
5412 goto out;
5415 err = register_pernet_subsys(&addrconf_ops);
5416 if (err < 0)
5417 goto out_addrlabel;
5419 addrconf_wq = create_workqueue("ipv6_addrconf");
5420 if (!addrconf_wq) {
5421 err = -ENOMEM;
5422 goto out_nowq;
5425 /* The addrconf netdev notifier requires that loopback_dev
5426 * has it's ipv6 private information allocated and setup
5427 * before it can bring up and give link-local addresses
5428 * to other devices which are up.
5430 * Unfortunately, loopback_dev is not necessarily the first
5431 * entry in the global dev_base list of net devices. In fact,
5432 * it is likely to be the very last entry on that list.
5433 * So this causes the notifier registry below to try and
5434 * give link-local addresses to all devices besides loopback_dev
5435 * first, then loopback_dev, which cases all the non-loopback_dev
5436 * devices to fail to get a link-local address.
5438 * So, as a temporary fix, allocate the ipv6 structure for
5439 * loopback_dev first by hand.
5440 * Longer term, all of the dependencies ipv6 has upon the loopback
5441 * device and it being up should be removed.
5443 rtnl_lock();
5444 idev = ipv6_add_dev(init_net.loopback_dev);
5445 rtnl_unlock();
5446 if (IS_ERR(idev)) {
5447 err = PTR_ERR(idev);
5448 goto errlo;
5451 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5452 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5454 register_netdevice_notifier(&ipv6_dev_notf);
5456 addrconf_verify();
5458 rtnl_af_register(&inet6_ops);
5460 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5461 NULL);
5462 if (err < 0)
5463 goto errout;
5465 /* Only the first call to __rtnl_register can fail */
5466 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5467 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5468 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5469 inet6_dump_ifaddr, NULL);
5470 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5471 inet6_dump_ifmcaddr, NULL);
5472 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5473 inet6_dump_ifacaddr, NULL);
5474 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5475 inet6_netconf_dump_devconf, NULL);
5477 ipv6_addr_label_rtnl_register();
5479 return 0;
5480 errout:
5481 rtnl_af_unregister(&inet6_ops);
5482 unregister_netdevice_notifier(&ipv6_dev_notf);
5483 errlo:
5484 destroy_workqueue(addrconf_wq);
5485 out_nowq:
5486 unregister_pernet_subsys(&addrconf_ops);
5487 out_addrlabel:
5488 ipv6_addr_label_cleanup();
5489 out:
5490 return err;
5493 void addrconf_cleanup(void)
5495 struct net_device *dev;
5496 int i;
5498 unregister_netdevice_notifier(&ipv6_dev_notf);
5499 unregister_pernet_subsys(&addrconf_ops);
5500 ipv6_addr_label_cleanup();
5502 rtnl_lock();
5504 __rtnl_af_unregister(&inet6_ops);
5506 /* clean dev list */
5507 for_each_netdev(&init_net, dev) {
5508 if (__in6_dev_get(dev) == NULL)
5509 continue;
5510 addrconf_ifdown(dev, 1);
5512 addrconf_ifdown(init_net.loopback_dev, 2);
5515 * Check hash table.
5517 spin_lock_bh(&addrconf_hash_lock);
5518 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5519 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5520 spin_unlock_bh(&addrconf_hash_lock);
5521 cancel_delayed_work(&addr_chk_work);
5522 rtnl_unlock();
5524 destroy_workqueue(addrconf_wq);