hyperv: Add processing of MTU reduced by the host
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob251fcb48b216054298b93c558c56402d804841b2
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 if ((idev = __in6_dev_get(dev)) == NULL)
2547 return -ENXIO;
2549 read_lock_bh(&idev->lock);
2550 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2551 if (ifp->prefix_len == plen &&
2552 ipv6_addr_equal(pfx, &ifp->addr)) {
2553 in6_ifa_hold(ifp);
2554 read_unlock_bh(&idev->lock);
2556 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2557 (ifa_flags & IFA_F_MANAGETEMPADDR))
2558 manage_tempaddrs(idev, ifp, 0, 0, false,
2559 jiffies);
2560 ipv6_del_addr(ifp);
2561 addrconf_verify_rtnl();
2562 return 0;
2565 read_unlock_bh(&idev->lock);
2566 return -EADDRNOTAVAIL;
2570 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2572 struct in6_ifreq ireq;
2573 int err;
2575 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2576 return -EPERM;
2578 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2579 return -EFAULT;
2581 rtnl_lock();
2582 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2583 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2584 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2585 rtnl_unlock();
2586 return err;
2589 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2591 struct in6_ifreq ireq;
2592 int err;
2594 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2595 return -EPERM;
2597 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2598 return -EFAULT;
2600 rtnl_lock();
2601 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2602 ireq.ifr6_prefixlen);
2603 rtnl_unlock();
2604 return err;
2607 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2608 int plen, int scope)
2610 struct inet6_ifaddr *ifp;
2612 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2613 scope, IFA_F_PERMANENT,
2614 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2615 if (!IS_ERR(ifp)) {
2616 spin_lock_bh(&ifp->lock);
2617 ifp->flags &= ~IFA_F_TENTATIVE;
2618 spin_unlock_bh(&ifp->lock);
2619 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2620 in6_ifa_put(ifp);
2624 #if IS_ENABLED(CONFIG_IPV6_SIT)
2625 static void sit_add_v4_addrs(struct inet6_dev *idev)
2627 struct in6_addr addr;
2628 struct net_device *dev;
2629 struct net *net = dev_net(idev->dev);
2630 int scope, plen;
2631 u32 pflags = 0;
2633 ASSERT_RTNL();
2635 memset(&addr, 0, sizeof(struct in6_addr));
2636 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2638 if (idev->dev->flags&IFF_POINTOPOINT) {
2639 addr.s6_addr32[0] = htonl(0xfe800000);
2640 scope = IFA_LINK;
2641 plen = 64;
2642 } else {
2643 scope = IPV6_ADDR_COMPATv4;
2644 plen = 96;
2645 pflags |= RTF_NONEXTHOP;
2648 if (addr.s6_addr32[3]) {
2649 add_addr(idev, &addr, plen, scope);
2650 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2651 return;
2654 for_each_netdev(net, dev) {
2655 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2656 if (in_dev && (dev->flags & IFF_UP)) {
2657 struct in_ifaddr *ifa;
2659 int flag = scope;
2661 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2663 addr.s6_addr32[3] = ifa->ifa_local;
2665 if (ifa->ifa_scope == RT_SCOPE_LINK)
2666 continue;
2667 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2668 if (idev->dev->flags&IFF_POINTOPOINT)
2669 continue;
2670 flag |= IFA_HOST;
2673 add_addr(idev, &addr, plen, flag);
2674 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2675 pflags);
2680 #endif
2682 static void init_loopback(struct net_device *dev)
2684 struct inet6_dev *idev;
2685 struct net_device *sp_dev;
2686 struct inet6_ifaddr *sp_ifa;
2687 struct rt6_info *sp_rt;
2689 /* ::1 */
2691 ASSERT_RTNL();
2693 if ((idev = ipv6_find_idev(dev)) == NULL) {
2694 pr_debug("%s: add_dev failed\n", __func__);
2695 return;
2698 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2700 /* Add routes to other interface's IPv6 addresses */
2701 for_each_netdev(dev_net(dev), sp_dev) {
2702 if (!strcmp(sp_dev->name, dev->name))
2703 continue;
2705 idev = __in6_dev_get(sp_dev);
2706 if (!idev)
2707 continue;
2709 read_lock_bh(&idev->lock);
2710 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2712 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2713 continue;
2715 if (sp_ifa->rt) {
2716 /* This dst has been added to garbage list when
2717 * lo device down, release this obsolete dst and
2718 * reallocate a new router for ifa.
2720 if (sp_ifa->rt->dst.obsolete > 0) {
2721 ip6_rt_put(sp_ifa->rt);
2722 sp_ifa->rt = NULL;
2723 } else {
2724 continue;
2728 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2730 /* Failure cases are ignored */
2731 if (!IS_ERR(sp_rt)) {
2732 sp_ifa->rt = sp_rt;
2733 ip6_ins_rt(sp_rt);
2736 read_unlock_bh(&idev->lock);
2740 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2742 struct inet6_ifaddr *ifp;
2743 u32 addr_flags = IFA_F_PERMANENT;
2745 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2746 if (idev->cnf.optimistic_dad &&
2747 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2748 addr_flags |= IFA_F_OPTIMISTIC;
2749 #endif
2752 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2753 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2754 if (!IS_ERR(ifp)) {
2755 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2756 addrconf_dad_start(ifp);
2757 in6_ifa_put(ifp);
2761 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2763 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2764 struct in6_addr addr;
2766 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2767 /* addrconf_add_linklocal also adds a prefix_route and we
2768 * only need to care about prefix routes if ipv6_generate_eui64
2769 * couldn't generate one.
2771 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2772 addrconf_add_linklocal(idev, &addr);
2773 else if (prefix_route)
2774 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2778 static void addrconf_dev_config(struct net_device *dev)
2780 struct inet6_dev *idev;
2782 ASSERT_RTNL();
2784 if ((dev->type != ARPHRD_ETHER) &&
2785 (dev->type != ARPHRD_FDDI) &&
2786 (dev->type != ARPHRD_ARCNET) &&
2787 (dev->type != ARPHRD_INFINIBAND) &&
2788 (dev->type != ARPHRD_IEEE802154) &&
2789 (dev->type != ARPHRD_IEEE1394) &&
2790 (dev->type != ARPHRD_TUNNEL6) &&
2791 (dev->type != ARPHRD_6LOWPAN)) {
2792 /* Alas, we support only Ethernet autoconfiguration. */
2793 return;
2796 idev = addrconf_add_dev(dev);
2797 if (IS_ERR(idev))
2798 return;
2800 addrconf_addr_gen(idev, false);
2803 #if IS_ENABLED(CONFIG_IPV6_SIT)
2804 static void addrconf_sit_config(struct net_device *dev)
2806 struct inet6_dev *idev;
2808 ASSERT_RTNL();
2811 * Configure the tunnel with one of our IPv4
2812 * addresses... we should configure all of
2813 * our v4 addrs in the tunnel
2816 if ((idev = ipv6_find_idev(dev)) == NULL) {
2817 pr_debug("%s: add_dev failed\n", __func__);
2818 return;
2821 if (dev->priv_flags & IFF_ISATAP) {
2822 addrconf_addr_gen(idev, false);
2823 return;
2826 sit_add_v4_addrs(idev);
2828 if (dev->flags&IFF_POINTOPOINT)
2829 addrconf_add_mroute(dev);
2831 #endif
2833 #if IS_ENABLED(CONFIG_NET_IPGRE)
2834 static void addrconf_gre_config(struct net_device *dev)
2836 struct inet6_dev *idev;
2838 ASSERT_RTNL();
2840 if ((idev = ipv6_find_idev(dev)) == NULL) {
2841 pr_debug("%s: add_dev failed\n", __func__);
2842 return;
2845 addrconf_addr_gen(idev, true);
2847 #endif
2849 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2850 void *ptr)
2852 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2853 struct inet6_dev *idev = __in6_dev_get(dev);
2854 int run_pending = 0;
2855 int err;
2857 switch (event) {
2858 case NETDEV_REGISTER:
2859 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2860 idev = ipv6_add_dev(dev);
2861 if (IS_ERR(idev))
2862 return notifier_from_errno(PTR_ERR(idev));
2864 break;
2866 case NETDEV_UP:
2867 case NETDEV_CHANGE:
2868 if (dev->flags & IFF_SLAVE)
2869 break;
2871 if (idev && idev->cnf.disable_ipv6)
2872 break;
2874 if (event == NETDEV_UP) {
2875 if (!addrconf_qdisc_ok(dev)) {
2876 /* device is not ready yet. */
2877 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2878 dev->name);
2879 break;
2882 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2883 idev = ipv6_add_dev(dev);
2885 if (!IS_ERR_OR_NULL(idev)) {
2886 idev->if_flags |= IF_READY;
2887 run_pending = 1;
2889 } else {
2890 if (!addrconf_qdisc_ok(dev)) {
2891 /* device is still not ready. */
2892 break;
2895 if (idev) {
2896 if (idev->if_flags & IF_READY)
2897 /* device is already configured. */
2898 break;
2899 idev->if_flags |= IF_READY;
2902 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2903 dev->name);
2905 run_pending = 1;
2908 switch (dev->type) {
2909 #if IS_ENABLED(CONFIG_IPV6_SIT)
2910 case ARPHRD_SIT:
2911 addrconf_sit_config(dev);
2912 break;
2913 #endif
2914 #if IS_ENABLED(CONFIG_NET_IPGRE)
2915 case ARPHRD_IPGRE:
2916 addrconf_gre_config(dev);
2917 break;
2918 #endif
2919 case ARPHRD_LOOPBACK:
2920 init_loopback(dev);
2921 break;
2923 default:
2924 addrconf_dev_config(dev);
2925 break;
2928 if (!IS_ERR_OR_NULL(idev)) {
2929 if (run_pending)
2930 addrconf_dad_run(idev);
2933 * If the MTU changed during the interface down,
2934 * when the interface up, the changed MTU must be
2935 * reflected in the idev as well as routers.
2937 if (idev->cnf.mtu6 != dev->mtu &&
2938 dev->mtu >= IPV6_MIN_MTU) {
2939 rt6_mtu_change(dev, dev->mtu);
2940 idev->cnf.mtu6 = dev->mtu;
2942 idev->tstamp = jiffies;
2943 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2946 * If the changed mtu during down is lower than
2947 * IPV6_MIN_MTU stop IPv6 on this interface.
2949 if (dev->mtu < IPV6_MIN_MTU)
2950 addrconf_ifdown(dev, 1);
2952 break;
2954 case NETDEV_CHANGEMTU:
2955 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2956 rt6_mtu_change(dev, dev->mtu);
2957 idev->cnf.mtu6 = dev->mtu;
2958 break;
2961 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2962 idev = ipv6_add_dev(dev);
2963 if (!IS_ERR(idev))
2964 break;
2968 * if MTU under IPV6_MIN_MTU.
2969 * Stop IPv6 on this interface.
2972 case NETDEV_DOWN:
2973 case NETDEV_UNREGISTER:
2975 * Remove all addresses from this interface.
2977 addrconf_ifdown(dev, event != NETDEV_DOWN);
2978 break;
2980 case NETDEV_CHANGENAME:
2981 if (idev) {
2982 snmp6_unregister_dev(idev);
2983 addrconf_sysctl_unregister(idev);
2984 err = addrconf_sysctl_register(idev);
2985 if (err)
2986 return notifier_from_errno(err);
2987 err = snmp6_register_dev(idev);
2988 if (err) {
2989 addrconf_sysctl_unregister(idev);
2990 return notifier_from_errno(err);
2993 break;
2995 case NETDEV_PRE_TYPE_CHANGE:
2996 case NETDEV_POST_TYPE_CHANGE:
2997 addrconf_type_change(dev, event);
2998 break;
3001 return NOTIFY_OK;
3005 * addrconf module should be notified of a device going up
3007 static struct notifier_block ipv6_dev_notf = {
3008 .notifier_call = addrconf_notify,
3011 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3013 struct inet6_dev *idev;
3014 ASSERT_RTNL();
3016 idev = __in6_dev_get(dev);
3018 if (event == NETDEV_POST_TYPE_CHANGE)
3019 ipv6_mc_remap(idev);
3020 else if (event == NETDEV_PRE_TYPE_CHANGE)
3021 ipv6_mc_unmap(idev);
3024 static int addrconf_ifdown(struct net_device *dev, int how)
3026 struct net *net = dev_net(dev);
3027 struct inet6_dev *idev;
3028 struct inet6_ifaddr *ifa;
3029 int state, i;
3031 ASSERT_RTNL();
3033 rt6_ifdown(net, dev);
3034 neigh_ifdown(&nd_tbl, dev);
3036 idev = __in6_dev_get(dev);
3037 if (idev == NULL)
3038 return -ENODEV;
3041 * Step 1: remove reference to ipv6 device from parent device.
3042 * Do not dev_put!
3044 if (how) {
3045 idev->dead = 1;
3047 /* protected by rtnl_lock */
3048 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3050 /* Step 1.5: remove snmp6 entry */
3051 snmp6_unregister_dev(idev);
3055 /* Step 2: clear hash table */
3056 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3057 struct hlist_head *h = &inet6_addr_lst[i];
3059 spin_lock_bh(&addrconf_hash_lock);
3060 restart:
3061 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3062 if (ifa->idev == idev) {
3063 hlist_del_init_rcu(&ifa->addr_lst);
3064 addrconf_del_dad_work(ifa);
3065 goto restart;
3068 spin_unlock_bh(&addrconf_hash_lock);
3071 write_lock_bh(&idev->lock);
3073 addrconf_del_rs_timer(idev);
3075 /* Step 2: clear flags for stateless addrconf */
3076 if (!how)
3077 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3079 if (how && del_timer(&idev->regen_timer))
3080 in6_dev_put(idev);
3082 /* Step 3: clear tempaddr list */
3083 while (!list_empty(&idev->tempaddr_list)) {
3084 ifa = list_first_entry(&idev->tempaddr_list,
3085 struct inet6_ifaddr, tmp_list);
3086 list_del(&ifa->tmp_list);
3087 write_unlock_bh(&idev->lock);
3088 spin_lock_bh(&ifa->lock);
3090 if (ifa->ifpub) {
3091 in6_ifa_put(ifa->ifpub);
3092 ifa->ifpub = NULL;
3094 spin_unlock_bh(&ifa->lock);
3095 in6_ifa_put(ifa);
3096 write_lock_bh(&idev->lock);
3099 while (!list_empty(&idev->addr_list)) {
3100 ifa = list_first_entry(&idev->addr_list,
3101 struct inet6_ifaddr, if_list);
3102 addrconf_del_dad_work(ifa);
3104 list_del(&ifa->if_list);
3106 write_unlock_bh(&idev->lock);
3108 spin_lock_bh(&ifa->state_lock);
3109 state = ifa->state;
3110 ifa->state = INET6_IFADDR_STATE_DEAD;
3111 spin_unlock_bh(&ifa->state_lock);
3113 if (state != INET6_IFADDR_STATE_DEAD) {
3114 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3115 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3117 in6_ifa_put(ifa);
3119 write_lock_bh(&idev->lock);
3122 write_unlock_bh(&idev->lock);
3124 /* Step 5: Discard anycast and multicast list */
3125 if (how) {
3126 ipv6_ac_destroy_dev(idev);
3127 ipv6_mc_destroy_dev(idev);
3128 } else {
3129 ipv6_mc_down(idev);
3132 idev->tstamp = jiffies;
3134 /* Last: Shot the device (if unregistered) */
3135 if (how) {
3136 addrconf_sysctl_unregister(idev);
3137 neigh_parms_release(&nd_tbl, idev->nd_parms);
3138 neigh_ifdown(&nd_tbl, dev);
3139 in6_dev_put(idev);
3141 return 0;
3144 static void addrconf_rs_timer(unsigned long data)
3146 struct inet6_dev *idev = (struct inet6_dev *)data;
3147 struct net_device *dev = idev->dev;
3148 struct in6_addr lladdr;
3150 write_lock(&idev->lock);
3151 if (idev->dead || !(idev->if_flags & IF_READY))
3152 goto out;
3154 if (!ipv6_accept_ra(idev))
3155 goto out;
3157 /* Announcement received after solicitation was sent */
3158 if (idev->if_flags & IF_RA_RCVD)
3159 goto out;
3161 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3162 write_unlock(&idev->lock);
3163 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3164 ndisc_send_rs(dev, &lladdr,
3165 &in6addr_linklocal_allrouters);
3166 else
3167 goto put;
3169 write_lock(&idev->lock);
3170 /* The wait after the last probe can be shorter */
3171 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3172 idev->cnf.rtr_solicits) ?
3173 idev->cnf.rtr_solicit_delay :
3174 idev->cnf.rtr_solicit_interval);
3175 } else {
3177 * Note: we do not support deprecated "all on-link"
3178 * assumption any longer.
3180 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3183 out:
3184 write_unlock(&idev->lock);
3185 put:
3186 in6_dev_put(idev);
3190 * Duplicate Address Detection
3192 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3194 unsigned long rand_num;
3195 struct inet6_dev *idev = ifp->idev;
3197 if (ifp->flags & IFA_F_OPTIMISTIC)
3198 rand_num = 0;
3199 else
3200 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3202 ifp->dad_probes = idev->cnf.dad_transmits;
3203 addrconf_mod_dad_work(ifp, rand_num);
3206 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3208 struct inet6_dev *idev = ifp->idev;
3209 struct net_device *dev = idev->dev;
3211 addrconf_join_solict(dev, &ifp->addr);
3213 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3215 read_lock_bh(&idev->lock);
3216 spin_lock(&ifp->lock);
3217 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3218 goto out;
3220 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3221 idev->cnf.accept_dad < 1 ||
3222 !(ifp->flags&IFA_F_TENTATIVE) ||
3223 ifp->flags & IFA_F_NODAD) {
3224 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3225 spin_unlock(&ifp->lock);
3226 read_unlock_bh(&idev->lock);
3228 addrconf_dad_completed(ifp);
3229 return;
3232 if (!(idev->if_flags & IF_READY)) {
3233 spin_unlock(&ifp->lock);
3234 read_unlock_bh(&idev->lock);
3236 * If the device is not ready:
3237 * - keep it tentative if it is a permanent address.
3238 * - otherwise, kill it.
3240 in6_ifa_hold(ifp);
3241 addrconf_dad_stop(ifp, 0);
3242 return;
3246 * Optimistic nodes can start receiving
3247 * Frames right away
3249 if (ifp->flags & IFA_F_OPTIMISTIC) {
3250 ip6_ins_rt(ifp->rt);
3251 if (ipv6_use_optimistic_addr(idev)) {
3252 /* Because optimistic nodes can use this address,
3253 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3255 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3259 addrconf_dad_kick(ifp);
3260 out:
3261 spin_unlock(&ifp->lock);
3262 read_unlock_bh(&idev->lock);
3265 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3267 bool begin_dad = false;
3269 spin_lock_bh(&ifp->state_lock);
3270 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3271 ifp->state = INET6_IFADDR_STATE_PREDAD;
3272 begin_dad = true;
3274 spin_unlock_bh(&ifp->state_lock);
3276 if (begin_dad)
3277 addrconf_mod_dad_work(ifp, 0);
3280 static void addrconf_dad_work(struct work_struct *w)
3282 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3283 struct inet6_ifaddr,
3284 dad_work);
3285 struct inet6_dev *idev = ifp->idev;
3286 struct in6_addr mcaddr;
3288 enum {
3289 DAD_PROCESS,
3290 DAD_BEGIN,
3291 DAD_ABORT,
3292 } action = DAD_PROCESS;
3294 rtnl_lock();
3296 spin_lock_bh(&ifp->state_lock);
3297 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3298 action = DAD_BEGIN;
3299 ifp->state = INET6_IFADDR_STATE_DAD;
3300 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3301 action = DAD_ABORT;
3302 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3304 spin_unlock_bh(&ifp->state_lock);
3306 if (action == DAD_BEGIN) {
3307 addrconf_dad_begin(ifp);
3308 goto out;
3309 } else if (action == DAD_ABORT) {
3310 addrconf_dad_stop(ifp, 1);
3311 goto out;
3314 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3315 goto out;
3317 write_lock_bh(&idev->lock);
3318 if (idev->dead || !(idev->if_flags & IF_READY)) {
3319 write_unlock_bh(&idev->lock);
3320 goto out;
3323 spin_lock(&ifp->lock);
3324 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3325 spin_unlock(&ifp->lock);
3326 write_unlock_bh(&idev->lock);
3327 goto out;
3330 if (ifp->dad_probes == 0) {
3332 * DAD was successful
3335 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3336 spin_unlock(&ifp->lock);
3337 write_unlock_bh(&idev->lock);
3339 addrconf_dad_completed(ifp);
3341 goto out;
3344 ifp->dad_probes--;
3345 addrconf_mod_dad_work(ifp,
3346 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3347 spin_unlock(&ifp->lock);
3348 write_unlock_bh(&idev->lock);
3350 /* send a neighbour solicitation for our addr */
3351 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3352 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3353 out:
3354 in6_ifa_put(ifp);
3355 rtnl_unlock();
3358 /* ifp->idev must be at least read locked */
3359 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3361 struct inet6_ifaddr *ifpiter;
3362 struct inet6_dev *idev = ifp->idev;
3364 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3365 if (ifpiter->scope > IFA_LINK)
3366 break;
3367 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3368 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3369 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3370 IFA_F_PERMANENT)
3371 return false;
3373 return true;
3376 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3378 struct net_device *dev = ifp->idev->dev;
3379 struct in6_addr lladdr;
3380 bool send_rs, send_mld;
3382 addrconf_del_dad_work(ifp);
3385 * Configure the address for reception. Now it is valid.
3388 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3390 /* If added prefix is link local and we are prepared to process
3391 router advertisements, start sending router solicitations.
3394 read_lock_bh(&ifp->idev->lock);
3395 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3396 send_rs = send_mld &&
3397 ipv6_accept_ra(ifp->idev) &&
3398 ifp->idev->cnf.rtr_solicits > 0 &&
3399 (dev->flags&IFF_LOOPBACK) == 0;
3400 read_unlock_bh(&ifp->idev->lock);
3402 /* While dad is in progress mld report's source address is in6_addrany.
3403 * Resend with proper ll now.
3405 if (send_mld)
3406 ipv6_mc_dad_complete(ifp->idev);
3408 if (send_rs) {
3410 * If a host as already performed a random delay
3411 * [...] as part of DAD [...] there is no need
3412 * to delay again before sending the first RS
3414 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3415 return;
3416 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3418 write_lock_bh(&ifp->idev->lock);
3419 spin_lock(&ifp->lock);
3420 ifp->idev->rs_probes = 1;
3421 ifp->idev->if_flags |= IF_RS_SENT;
3422 addrconf_mod_rs_timer(ifp->idev,
3423 ifp->idev->cnf.rtr_solicit_interval);
3424 spin_unlock(&ifp->lock);
3425 write_unlock_bh(&ifp->idev->lock);
3429 static void addrconf_dad_run(struct inet6_dev *idev)
3431 struct inet6_ifaddr *ifp;
3433 read_lock_bh(&idev->lock);
3434 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3435 spin_lock(&ifp->lock);
3436 if (ifp->flags & IFA_F_TENTATIVE &&
3437 ifp->state == INET6_IFADDR_STATE_DAD)
3438 addrconf_dad_kick(ifp);
3439 spin_unlock(&ifp->lock);
3441 read_unlock_bh(&idev->lock);
3444 #ifdef CONFIG_PROC_FS
3445 struct if6_iter_state {
3446 struct seq_net_private p;
3447 int bucket;
3448 int offset;
3451 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3453 struct inet6_ifaddr *ifa = NULL;
3454 struct if6_iter_state *state = seq->private;
3455 struct net *net = seq_file_net(seq);
3456 int p = 0;
3458 /* initial bucket if pos is 0 */
3459 if (pos == 0) {
3460 state->bucket = 0;
3461 state->offset = 0;
3464 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3465 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3466 addr_lst) {
3467 if (!net_eq(dev_net(ifa->idev->dev), net))
3468 continue;
3469 /* sync with offset */
3470 if (p < state->offset) {
3471 p++;
3472 continue;
3474 state->offset++;
3475 return ifa;
3478 /* prepare for next bucket */
3479 state->offset = 0;
3480 p = 0;
3482 return NULL;
3485 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3486 struct inet6_ifaddr *ifa)
3488 struct if6_iter_state *state = seq->private;
3489 struct net *net = seq_file_net(seq);
3491 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3492 if (!net_eq(dev_net(ifa->idev->dev), net))
3493 continue;
3494 state->offset++;
3495 return ifa;
3498 while (++state->bucket < IN6_ADDR_HSIZE) {
3499 state->offset = 0;
3500 hlist_for_each_entry_rcu_bh(ifa,
3501 &inet6_addr_lst[state->bucket], addr_lst) {
3502 if (!net_eq(dev_net(ifa->idev->dev), net))
3503 continue;
3504 state->offset++;
3505 return ifa;
3509 return NULL;
3512 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3513 __acquires(rcu_bh)
3515 rcu_read_lock_bh();
3516 return if6_get_first(seq, *pos);
3519 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3521 struct inet6_ifaddr *ifa;
3523 ifa = if6_get_next(seq, v);
3524 ++*pos;
3525 return ifa;
3528 static void if6_seq_stop(struct seq_file *seq, void *v)
3529 __releases(rcu_bh)
3531 rcu_read_unlock_bh();
3534 static int if6_seq_show(struct seq_file *seq, void *v)
3536 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3537 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3538 &ifp->addr,
3539 ifp->idev->dev->ifindex,
3540 ifp->prefix_len,
3541 ifp->scope,
3542 (u8) ifp->flags,
3543 ifp->idev->dev->name);
3544 return 0;
3547 static const struct seq_operations if6_seq_ops = {
3548 .start = if6_seq_start,
3549 .next = if6_seq_next,
3550 .show = if6_seq_show,
3551 .stop = if6_seq_stop,
3554 static int if6_seq_open(struct inode *inode, struct file *file)
3556 return seq_open_net(inode, file, &if6_seq_ops,
3557 sizeof(struct if6_iter_state));
3560 static const struct file_operations if6_fops = {
3561 .owner = THIS_MODULE,
3562 .open = if6_seq_open,
3563 .read = seq_read,
3564 .llseek = seq_lseek,
3565 .release = seq_release_net,
3568 static int __net_init if6_proc_net_init(struct net *net)
3570 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3571 return -ENOMEM;
3572 return 0;
3575 static void __net_exit if6_proc_net_exit(struct net *net)
3577 remove_proc_entry("if_inet6", net->proc_net);
3580 static struct pernet_operations if6_proc_net_ops = {
3581 .init = if6_proc_net_init,
3582 .exit = if6_proc_net_exit,
3585 int __init if6_proc_init(void)
3587 return register_pernet_subsys(&if6_proc_net_ops);
3590 void if6_proc_exit(void)
3592 unregister_pernet_subsys(&if6_proc_net_ops);
3594 #endif /* CONFIG_PROC_FS */
3596 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3597 /* Check if address is a home address configured on any interface. */
3598 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3600 int ret = 0;
3601 struct inet6_ifaddr *ifp = NULL;
3602 unsigned int hash = inet6_addr_hash(addr);
3604 rcu_read_lock_bh();
3605 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3606 if (!net_eq(dev_net(ifp->idev->dev), net))
3607 continue;
3608 if (ipv6_addr_equal(&ifp->addr, addr) &&
3609 (ifp->flags & IFA_F_HOMEADDRESS)) {
3610 ret = 1;
3611 break;
3614 rcu_read_unlock_bh();
3615 return ret;
3617 #endif
3620 * Periodic address status verification
3623 static void addrconf_verify_rtnl(void)
3625 unsigned long now, next, next_sec, next_sched;
3626 struct inet6_ifaddr *ifp;
3627 int i;
3629 ASSERT_RTNL();
3631 rcu_read_lock_bh();
3632 now = jiffies;
3633 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3635 cancel_delayed_work(&addr_chk_work);
3637 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3638 restart:
3639 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3640 unsigned long age;
3642 /* When setting preferred_lft to a value not zero or
3643 * infinity, while valid_lft is infinity
3644 * IFA_F_PERMANENT has a non-infinity life time.
3646 if ((ifp->flags & IFA_F_PERMANENT) &&
3647 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3648 continue;
3650 spin_lock(&ifp->lock);
3651 /* We try to batch several events at once. */
3652 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3654 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3655 age >= ifp->valid_lft) {
3656 spin_unlock(&ifp->lock);
3657 in6_ifa_hold(ifp);
3658 ipv6_del_addr(ifp);
3659 goto restart;
3660 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3661 spin_unlock(&ifp->lock);
3662 continue;
3663 } else if (age >= ifp->prefered_lft) {
3664 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3665 int deprecate = 0;
3667 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3668 deprecate = 1;
3669 ifp->flags |= IFA_F_DEPRECATED;
3672 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3673 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3674 next = ifp->tstamp + ifp->valid_lft * HZ;
3676 spin_unlock(&ifp->lock);
3678 if (deprecate) {
3679 in6_ifa_hold(ifp);
3681 ipv6_ifa_notify(0, ifp);
3682 in6_ifa_put(ifp);
3683 goto restart;
3685 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3686 !(ifp->flags&IFA_F_TENTATIVE)) {
3687 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3688 ifp->idev->cnf.dad_transmits *
3689 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3691 if (age >= ifp->prefered_lft - regen_advance) {
3692 struct inet6_ifaddr *ifpub = ifp->ifpub;
3693 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3694 next = ifp->tstamp + ifp->prefered_lft * HZ;
3695 if (!ifp->regen_count && ifpub) {
3696 ifp->regen_count++;
3697 in6_ifa_hold(ifp);
3698 in6_ifa_hold(ifpub);
3699 spin_unlock(&ifp->lock);
3701 spin_lock(&ifpub->lock);
3702 ifpub->regen_count = 0;
3703 spin_unlock(&ifpub->lock);
3704 ipv6_create_tempaddr(ifpub, ifp);
3705 in6_ifa_put(ifpub);
3706 in6_ifa_put(ifp);
3707 goto restart;
3709 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3710 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3711 spin_unlock(&ifp->lock);
3712 } else {
3713 /* ifp->prefered_lft <= ifp->valid_lft */
3714 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3715 next = ifp->tstamp + ifp->prefered_lft * HZ;
3716 spin_unlock(&ifp->lock);
3721 next_sec = round_jiffies_up(next);
3722 next_sched = next;
3724 /* If rounded timeout is accurate enough, accept it. */
3725 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3726 next_sched = next_sec;
3728 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3729 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3730 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3732 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3733 now, next, next_sec, next_sched);
3734 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3735 rcu_read_unlock_bh();
3738 static void addrconf_verify_work(struct work_struct *w)
3740 rtnl_lock();
3741 addrconf_verify_rtnl();
3742 rtnl_unlock();
3745 static void addrconf_verify(void)
3747 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3750 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3751 struct in6_addr **peer_pfx)
3753 struct in6_addr *pfx = NULL;
3755 *peer_pfx = NULL;
3757 if (addr)
3758 pfx = nla_data(addr);
3760 if (local) {
3761 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3762 *peer_pfx = pfx;
3763 pfx = nla_data(local);
3766 return pfx;
3769 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3770 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3771 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3772 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3773 [IFA_FLAGS] = { .len = sizeof(u32) },
3776 static int
3777 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3779 struct net *net = sock_net(skb->sk);
3780 struct ifaddrmsg *ifm;
3781 struct nlattr *tb[IFA_MAX+1];
3782 struct in6_addr *pfx, *peer_pfx;
3783 u32 ifa_flags;
3784 int err;
3786 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3787 if (err < 0)
3788 return err;
3790 ifm = nlmsg_data(nlh);
3791 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3792 if (pfx == NULL)
3793 return -EINVAL;
3795 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3797 /* We ignore other flags so far. */
3798 ifa_flags &= IFA_F_MANAGETEMPADDR;
3800 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3801 ifm->ifa_prefixlen);
3804 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3805 u32 prefered_lft, u32 valid_lft)
3807 u32 flags;
3808 clock_t expires;
3809 unsigned long timeout;
3810 bool was_managetempaddr;
3811 bool had_prefixroute;
3813 ASSERT_RTNL();
3815 if (!valid_lft || (prefered_lft > valid_lft))
3816 return -EINVAL;
3818 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3819 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3820 return -EINVAL;
3822 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3823 if (addrconf_finite_timeout(timeout)) {
3824 expires = jiffies_to_clock_t(timeout * HZ);
3825 valid_lft = timeout;
3826 flags = RTF_EXPIRES;
3827 } else {
3828 expires = 0;
3829 flags = 0;
3830 ifa_flags |= IFA_F_PERMANENT;
3833 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3834 if (addrconf_finite_timeout(timeout)) {
3835 if (timeout == 0)
3836 ifa_flags |= IFA_F_DEPRECATED;
3837 prefered_lft = timeout;
3840 spin_lock_bh(&ifp->lock);
3841 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3842 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3843 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3844 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3845 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3846 IFA_F_NOPREFIXROUTE);
3847 ifp->flags |= ifa_flags;
3848 ifp->tstamp = jiffies;
3849 ifp->valid_lft = valid_lft;
3850 ifp->prefered_lft = prefered_lft;
3852 spin_unlock_bh(&ifp->lock);
3853 if (!(ifp->flags&IFA_F_TENTATIVE))
3854 ipv6_ifa_notify(0, ifp);
3856 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3857 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3858 expires, flags);
3859 } else if (had_prefixroute) {
3860 enum cleanup_prefix_rt_t action;
3861 unsigned long rt_expires;
3863 write_lock_bh(&ifp->idev->lock);
3864 action = check_cleanup_prefix_route(ifp, &rt_expires);
3865 write_unlock_bh(&ifp->idev->lock);
3867 if (action != CLEANUP_PREFIX_RT_NOP) {
3868 cleanup_prefix_route(ifp, rt_expires,
3869 action == CLEANUP_PREFIX_RT_DEL);
3873 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3874 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3875 valid_lft = prefered_lft = 0;
3876 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3877 !was_managetempaddr, jiffies);
3880 addrconf_verify_rtnl();
3882 return 0;
3885 static int
3886 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3888 struct net *net = sock_net(skb->sk);
3889 struct ifaddrmsg *ifm;
3890 struct nlattr *tb[IFA_MAX+1];
3891 struct in6_addr *pfx, *peer_pfx;
3892 struct inet6_ifaddr *ifa;
3893 struct net_device *dev;
3894 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3895 u32 ifa_flags;
3896 int err;
3898 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3899 if (err < 0)
3900 return err;
3902 ifm = nlmsg_data(nlh);
3903 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3904 if (pfx == NULL)
3905 return -EINVAL;
3907 if (tb[IFA_CACHEINFO]) {
3908 struct ifa_cacheinfo *ci;
3910 ci = nla_data(tb[IFA_CACHEINFO]);
3911 valid_lft = ci->ifa_valid;
3912 preferred_lft = ci->ifa_prefered;
3913 } else {
3914 preferred_lft = INFINITY_LIFE_TIME;
3915 valid_lft = INFINITY_LIFE_TIME;
3918 dev = __dev_get_by_index(net, ifm->ifa_index);
3919 if (dev == NULL)
3920 return -ENODEV;
3922 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3924 /* We ignore other flags so far. */
3925 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3926 IFA_F_NOPREFIXROUTE;
3928 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3929 if (ifa == NULL) {
3931 * It would be best to check for !NLM_F_CREATE here but
3932 * userspace already relies on not having to provide this.
3934 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3935 ifm->ifa_prefixlen, ifa_flags,
3936 preferred_lft, valid_lft);
3939 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3940 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3941 err = -EEXIST;
3942 else
3943 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3945 in6_ifa_put(ifa);
3947 return err;
3950 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3951 u8 scope, int ifindex)
3953 struct ifaddrmsg *ifm;
3955 ifm = nlmsg_data(nlh);
3956 ifm->ifa_family = AF_INET6;
3957 ifm->ifa_prefixlen = prefixlen;
3958 ifm->ifa_flags = flags;
3959 ifm->ifa_scope = scope;
3960 ifm->ifa_index = ifindex;
3963 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3964 unsigned long tstamp, u32 preferred, u32 valid)
3966 struct ifa_cacheinfo ci;
3968 ci.cstamp = cstamp_delta(cstamp);
3969 ci.tstamp = cstamp_delta(tstamp);
3970 ci.ifa_prefered = preferred;
3971 ci.ifa_valid = valid;
3973 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3976 static inline int rt_scope(int ifa_scope)
3978 if (ifa_scope & IFA_HOST)
3979 return RT_SCOPE_HOST;
3980 else if (ifa_scope & IFA_LINK)
3981 return RT_SCOPE_LINK;
3982 else if (ifa_scope & IFA_SITE)
3983 return RT_SCOPE_SITE;
3984 else
3985 return RT_SCOPE_UNIVERSE;
3988 static inline int inet6_ifaddr_msgsize(void)
3990 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3991 + nla_total_size(16) /* IFA_LOCAL */
3992 + nla_total_size(16) /* IFA_ADDRESS */
3993 + nla_total_size(sizeof(struct ifa_cacheinfo))
3994 + nla_total_size(4) /* IFA_FLAGS */;
3997 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3998 u32 portid, u32 seq, int event, unsigned int flags)
4000 struct nlmsghdr *nlh;
4001 u32 preferred, valid;
4003 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4004 if (nlh == NULL)
4005 return -EMSGSIZE;
4007 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4008 ifa->idev->dev->ifindex);
4010 if (!((ifa->flags&IFA_F_PERMANENT) &&
4011 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4012 preferred = ifa->prefered_lft;
4013 valid = ifa->valid_lft;
4014 if (preferred != INFINITY_LIFE_TIME) {
4015 long tval = (jiffies - ifa->tstamp)/HZ;
4016 if (preferred > tval)
4017 preferred -= tval;
4018 else
4019 preferred = 0;
4020 if (valid != INFINITY_LIFE_TIME) {
4021 if (valid > tval)
4022 valid -= tval;
4023 else
4024 valid = 0;
4027 } else {
4028 preferred = INFINITY_LIFE_TIME;
4029 valid = INFINITY_LIFE_TIME;
4032 if (!ipv6_addr_any(&ifa->peer_addr)) {
4033 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
4034 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4035 goto error;
4036 } else
4037 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4038 goto error;
4040 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4041 goto error;
4043 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4044 goto error;
4046 return nlmsg_end(skb, nlh);
4048 error:
4049 nlmsg_cancel(skb, nlh);
4050 return -EMSGSIZE;
4053 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4054 u32 portid, u32 seq, int event, u16 flags)
4056 struct nlmsghdr *nlh;
4057 u8 scope = RT_SCOPE_UNIVERSE;
4058 int ifindex = ifmca->idev->dev->ifindex;
4060 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4061 scope = RT_SCOPE_SITE;
4063 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4064 if (nlh == NULL)
4065 return -EMSGSIZE;
4067 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4068 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4069 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4070 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4071 nlmsg_cancel(skb, nlh);
4072 return -EMSGSIZE;
4075 return nlmsg_end(skb, nlh);
4078 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4079 u32 portid, u32 seq, int event, unsigned int flags)
4081 struct nlmsghdr *nlh;
4082 u8 scope = RT_SCOPE_UNIVERSE;
4083 int ifindex = ifaca->aca_idev->dev->ifindex;
4085 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4086 scope = RT_SCOPE_SITE;
4088 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4089 if (nlh == NULL)
4090 return -EMSGSIZE;
4092 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4093 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4094 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4095 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4096 nlmsg_cancel(skb, nlh);
4097 return -EMSGSIZE;
4100 return nlmsg_end(skb, nlh);
4103 enum addr_type_t {
4104 UNICAST_ADDR,
4105 MULTICAST_ADDR,
4106 ANYCAST_ADDR,
4109 /* called with rcu_read_lock() */
4110 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4111 struct netlink_callback *cb, enum addr_type_t type,
4112 int s_ip_idx, int *p_ip_idx)
4114 struct ifmcaddr6 *ifmca;
4115 struct ifacaddr6 *ifaca;
4116 int err = 1;
4117 int ip_idx = *p_ip_idx;
4119 read_lock_bh(&idev->lock);
4120 switch (type) {
4121 case UNICAST_ADDR: {
4122 struct inet6_ifaddr *ifa;
4124 /* unicast address incl. temp addr */
4125 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4126 if (++ip_idx < s_ip_idx)
4127 continue;
4128 err = inet6_fill_ifaddr(skb, ifa,
4129 NETLINK_CB(cb->skb).portid,
4130 cb->nlh->nlmsg_seq,
4131 RTM_NEWADDR,
4132 NLM_F_MULTI);
4133 if (err <= 0)
4134 break;
4135 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4137 break;
4139 case MULTICAST_ADDR:
4140 /* multicast address */
4141 for (ifmca = idev->mc_list; ifmca;
4142 ifmca = ifmca->next, ip_idx++) {
4143 if (ip_idx < s_ip_idx)
4144 continue;
4145 err = inet6_fill_ifmcaddr(skb, ifmca,
4146 NETLINK_CB(cb->skb).portid,
4147 cb->nlh->nlmsg_seq,
4148 RTM_GETMULTICAST,
4149 NLM_F_MULTI);
4150 if (err <= 0)
4151 break;
4153 break;
4154 case ANYCAST_ADDR:
4155 /* anycast address */
4156 for (ifaca = idev->ac_list; ifaca;
4157 ifaca = ifaca->aca_next, ip_idx++) {
4158 if (ip_idx < s_ip_idx)
4159 continue;
4160 err = inet6_fill_ifacaddr(skb, ifaca,
4161 NETLINK_CB(cb->skb).portid,
4162 cb->nlh->nlmsg_seq,
4163 RTM_GETANYCAST,
4164 NLM_F_MULTI);
4165 if (err <= 0)
4166 break;
4168 break;
4169 default:
4170 break;
4172 read_unlock_bh(&idev->lock);
4173 *p_ip_idx = ip_idx;
4174 return err;
4177 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4178 enum addr_type_t type)
4180 struct net *net = sock_net(skb->sk);
4181 int h, s_h;
4182 int idx, ip_idx;
4183 int s_idx, s_ip_idx;
4184 struct net_device *dev;
4185 struct inet6_dev *idev;
4186 struct hlist_head *head;
4188 s_h = cb->args[0];
4189 s_idx = idx = cb->args[1];
4190 s_ip_idx = ip_idx = cb->args[2];
4192 rcu_read_lock();
4193 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4194 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4195 idx = 0;
4196 head = &net->dev_index_head[h];
4197 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4198 if (idx < s_idx)
4199 goto cont;
4200 if (h > s_h || idx > s_idx)
4201 s_ip_idx = 0;
4202 ip_idx = 0;
4203 idev = __in6_dev_get(dev);
4204 if (!idev)
4205 goto cont;
4207 if (in6_dump_addrs(idev, skb, cb, type,
4208 s_ip_idx, &ip_idx) <= 0)
4209 goto done;
4210 cont:
4211 idx++;
4214 done:
4215 rcu_read_unlock();
4216 cb->args[0] = h;
4217 cb->args[1] = idx;
4218 cb->args[2] = ip_idx;
4220 return skb->len;
4223 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4225 enum addr_type_t type = UNICAST_ADDR;
4227 return inet6_dump_addr(skb, cb, type);
4230 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4232 enum addr_type_t type = MULTICAST_ADDR;
4234 return inet6_dump_addr(skb, cb, type);
4238 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4240 enum addr_type_t type = ANYCAST_ADDR;
4242 return inet6_dump_addr(skb, cb, type);
4245 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4247 struct net *net = sock_net(in_skb->sk);
4248 struct ifaddrmsg *ifm;
4249 struct nlattr *tb[IFA_MAX+1];
4250 struct in6_addr *addr = NULL, *peer;
4251 struct net_device *dev = NULL;
4252 struct inet6_ifaddr *ifa;
4253 struct sk_buff *skb;
4254 int err;
4256 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4257 if (err < 0)
4258 goto errout;
4260 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4261 if (addr == NULL) {
4262 err = -EINVAL;
4263 goto errout;
4266 ifm = nlmsg_data(nlh);
4267 if (ifm->ifa_index)
4268 dev = __dev_get_by_index(net, ifm->ifa_index);
4270 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4271 if (!ifa) {
4272 err = -EADDRNOTAVAIL;
4273 goto errout;
4276 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4277 if (!skb) {
4278 err = -ENOBUFS;
4279 goto errout_ifa;
4282 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4283 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4284 if (err < 0) {
4285 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4286 WARN_ON(err == -EMSGSIZE);
4287 kfree_skb(skb);
4288 goto errout_ifa;
4290 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4291 errout_ifa:
4292 in6_ifa_put(ifa);
4293 errout:
4294 return err;
4297 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4299 struct sk_buff *skb;
4300 struct net *net = dev_net(ifa->idev->dev);
4301 int err = -ENOBUFS;
4303 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4304 if (skb == NULL)
4305 goto errout;
4307 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4308 if (err < 0) {
4309 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4310 WARN_ON(err == -EMSGSIZE);
4311 kfree_skb(skb);
4312 goto errout;
4314 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4315 return;
4316 errout:
4317 if (err < 0)
4318 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4321 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4322 __s32 *array, int bytes)
4324 BUG_ON(bytes < (DEVCONF_MAX * 4));
4326 memset(array, 0, bytes);
4327 array[DEVCONF_FORWARDING] = cnf->forwarding;
4328 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4329 array[DEVCONF_MTU6] = cnf->mtu6;
4330 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4331 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4332 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4333 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4334 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4335 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4336 jiffies_to_msecs(cnf->rtr_solicit_interval);
4337 array[DEVCONF_RTR_SOLICIT_DELAY] =
4338 jiffies_to_msecs(cnf->rtr_solicit_delay);
4339 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4340 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4341 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4342 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4343 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4344 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4345 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4346 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4347 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4348 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4349 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4350 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4351 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4352 #ifdef CONFIG_IPV6_ROUTER_PREF
4353 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4354 array[DEVCONF_RTR_PROBE_INTERVAL] =
4355 jiffies_to_msecs(cnf->rtr_probe_interval);
4356 #ifdef CONFIG_IPV6_ROUTE_INFO
4357 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4358 #endif
4359 #endif
4360 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4361 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4362 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4363 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4364 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4365 #endif
4366 #ifdef CONFIG_IPV6_MROUTE
4367 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4368 #endif
4369 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4370 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4371 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4372 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4373 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4374 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4377 static inline size_t inet6_ifla6_size(void)
4379 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4380 + nla_total_size(sizeof(struct ifla_cacheinfo))
4381 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4382 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4383 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4384 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4387 static inline size_t inet6_if_nlmsg_size(void)
4389 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4390 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4391 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4392 + nla_total_size(4) /* IFLA_MTU */
4393 + nla_total_size(4) /* IFLA_LINK */
4394 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4397 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4398 int items, int bytes)
4400 int i;
4401 int pad = bytes - sizeof(u64) * items;
4402 BUG_ON(pad < 0);
4404 /* Use put_unaligned() because stats may not be aligned for u64. */
4405 put_unaligned(items, &stats[0]);
4406 for (i = 1; i < items; i++)
4407 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4409 memset(&stats[items], 0, pad);
4412 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4413 int items, int bytes, size_t syncpoff)
4415 int i;
4416 int pad = bytes - sizeof(u64) * items;
4417 BUG_ON(pad < 0);
4419 /* Use put_unaligned() because stats may not be aligned for u64. */
4420 put_unaligned(items, &stats[0]);
4421 for (i = 1; i < items; i++)
4422 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4424 memset(&stats[items], 0, pad);
4427 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4428 int bytes)
4430 switch (attrtype) {
4431 case IFLA_INET6_STATS:
4432 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4433 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4434 break;
4435 case IFLA_INET6_ICMP6STATS:
4436 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4437 break;
4441 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4443 struct nlattr *nla;
4444 struct ifla_cacheinfo ci;
4446 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4447 goto nla_put_failure;
4448 ci.max_reasm_len = IPV6_MAXPLEN;
4449 ci.tstamp = cstamp_delta(idev->tstamp);
4450 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4451 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4452 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4453 goto nla_put_failure;
4454 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4455 if (nla == NULL)
4456 goto nla_put_failure;
4457 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4459 /* XXX - MC not implemented */
4461 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4462 if (nla == NULL)
4463 goto nla_put_failure;
4464 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4466 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4467 if (nla == NULL)
4468 goto nla_put_failure;
4469 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4471 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4472 if (nla == NULL)
4473 goto nla_put_failure;
4475 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4476 goto nla_put_failure;
4478 read_lock_bh(&idev->lock);
4479 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4480 read_unlock_bh(&idev->lock);
4482 return 0;
4484 nla_put_failure:
4485 return -EMSGSIZE;
4488 static size_t inet6_get_link_af_size(const struct net_device *dev)
4490 if (!__in6_dev_get(dev))
4491 return 0;
4493 return inet6_ifla6_size();
4496 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4498 struct inet6_dev *idev = __in6_dev_get(dev);
4500 if (!idev)
4501 return -ENODATA;
4503 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4504 return -EMSGSIZE;
4506 return 0;
4509 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4511 struct inet6_ifaddr *ifp;
4512 struct net_device *dev = idev->dev;
4513 bool update_rs = false;
4514 struct in6_addr ll_addr;
4516 ASSERT_RTNL();
4518 if (token == NULL)
4519 return -EINVAL;
4520 if (ipv6_addr_any(token))
4521 return -EINVAL;
4522 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4523 return -EINVAL;
4524 if (!ipv6_accept_ra(idev))
4525 return -EINVAL;
4526 if (idev->cnf.rtr_solicits <= 0)
4527 return -EINVAL;
4529 write_lock_bh(&idev->lock);
4531 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4532 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4534 write_unlock_bh(&idev->lock);
4536 if (!idev->dead && (idev->if_flags & IF_READY) &&
4537 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4538 IFA_F_OPTIMISTIC)) {
4540 /* If we're not ready, then normal ifup will take care
4541 * of this. Otherwise, we need to request our rs here.
4543 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4544 update_rs = true;
4547 write_lock_bh(&idev->lock);
4549 if (update_rs) {
4550 idev->if_flags |= IF_RS_SENT;
4551 idev->rs_probes = 1;
4552 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4555 /* Well, that's kinda nasty ... */
4556 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4557 spin_lock(&ifp->lock);
4558 if (ifp->tokenized) {
4559 ifp->valid_lft = 0;
4560 ifp->prefered_lft = 0;
4562 spin_unlock(&ifp->lock);
4565 write_unlock_bh(&idev->lock);
4566 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4567 addrconf_verify_rtnl();
4568 return 0;
4571 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4573 int err = -EINVAL;
4574 struct inet6_dev *idev = __in6_dev_get(dev);
4575 struct nlattr *tb[IFLA_INET6_MAX + 1];
4577 if (!idev)
4578 return -EAFNOSUPPORT;
4580 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4581 BUG();
4583 if (tb[IFLA_INET6_TOKEN]) {
4584 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4585 if (err)
4586 return err;
4589 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4590 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4592 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4593 mode != IN6_ADDR_GEN_MODE_NONE)
4594 return -EINVAL;
4595 idev->addr_gen_mode = mode;
4596 err = 0;
4599 return err;
4602 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4603 u32 portid, u32 seq, int event, unsigned int flags)
4605 struct net_device *dev = idev->dev;
4606 struct ifinfomsg *hdr;
4607 struct nlmsghdr *nlh;
4608 void *protoinfo;
4610 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4611 if (nlh == NULL)
4612 return -EMSGSIZE;
4614 hdr = nlmsg_data(nlh);
4615 hdr->ifi_family = AF_INET6;
4616 hdr->__ifi_pad = 0;
4617 hdr->ifi_type = dev->type;
4618 hdr->ifi_index = dev->ifindex;
4619 hdr->ifi_flags = dev_get_flags(dev);
4620 hdr->ifi_change = 0;
4622 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4623 (dev->addr_len &&
4624 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4625 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4626 (dev->ifindex != dev->iflink &&
4627 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4628 goto nla_put_failure;
4629 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4630 if (protoinfo == NULL)
4631 goto nla_put_failure;
4633 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4634 goto nla_put_failure;
4636 nla_nest_end(skb, protoinfo);
4637 return nlmsg_end(skb, nlh);
4639 nla_put_failure:
4640 nlmsg_cancel(skb, nlh);
4641 return -EMSGSIZE;
4644 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4646 struct net *net = sock_net(skb->sk);
4647 int h, s_h;
4648 int idx = 0, s_idx;
4649 struct net_device *dev;
4650 struct inet6_dev *idev;
4651 struct hlist_head *head;
4653 s_h = cb->args[0];
4654 s_idx = cb->args[1];
4656 rcu_read_lock();
4657 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4658 idx = 0;
4659 head = &net->dev_index_head[h];
4660 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4661 if (idx < s_idx)
4662 goto cont;
4663 idev = __in6_dev_get(dev);
4664 if (!idev)
4665 goto cont;
4666 if (inet6_fill_ifinfo(skb, idev,
4667 NETLINK_CB(cb->skb).portid,
4668 cb->nlh->nlmsg_seq,
4669 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4670 goto out;
4671 cont:
4672 idx++;
4675 out:
4676 rcu_read_unlock();
4677 cb->args[1] = idx;
4678 cb->args[0] = h;
4680 return skb->len;
4683 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4685 struct sk_buff *skb;
4686 struct net *net = dev_net(idev->dev);
4687 int err = -ENOBUFS;
4689 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4690 if (skb == NULL)
4691 goto errout;
4693 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4694 if (err < 0) {
4695 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4696 WARN_ON(err == -EMSGSIZE);
4697 kfree_skb(skb);
4698 goto errout;
4700 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4701 return;
4702 errout:
4703 if (err < 0)
4704 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4707 static inline size_t inet6_prefix_nlmsg_size(void)
4709 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4710 + nla_total_size(sizeof(struct in6_addr))
4711 + nla_total_size(sizeof(struct prefix_cacheinfo));
4714 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4715 struct prefix_info *pinfo, u32 portid, u32 seq,
4716 int event, unsigned int flags)
4718 struct prefixmsg *pmsg;
4719 struct nlmsghdr *nlh;
4720 struct prefix_cacheinfo ci;
4722 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4723 if (nlh == NULL)
4724 return -EMSGSIZE;
4726 pmsg = nlmsg_data(nlh);
4727 pmsg->prefix_family = AF_INET6;
4728 pmsg->prefix_pad1 = 0;
4729 pmsg->prefix_pad2 = 0;
4730 pmsg->prefix_ifindex = idev->dev->ifindex;
4731 pmsg->prefix_len = pinfo->prefix_len;
4732 pmsg->prefix_type = pinfo->type;
4733 pmsg->prefix_pad3 = 0;
4734 pmsg->prefix_flags = 0;
4735 if (pinfo->onlink)
4736 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4737 if (pinfo->autoconf)
4738 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4740 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4741 goto nla_put_failure;
4742 ci.preferred_time = ntohl(pinfo->prefered);
4743 ci.valid_time = ntohl(pinfo->valid);
4744 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4745 goto nla_put_failure;
4746 return nlmsg_end(skb, nlh);
4748 nla_put_failure:
4749 nlmsg_cancel(skb, nlh);
4750 return -EMSGSIZE;
4753 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4754 struct prefix_info *pinfo)
4756 struct sk_buff *skb;
4757 struct net *net = dev_net(idev->dev);
4758 int err = -ENOBUFS;
4760 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4761 if (skb == NULL)
4762 goto errout;
4764 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4765 if (err < 0) {
4766 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4767 WARN_ON(err == -EMSGSIZE);
4768 kfree_skb(skb);
4769 goto errout;
4771 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4772 return;
4773 errout:
4774 if (err < 0)
4775 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4778 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4780 struct net *net = dev_net(ifp->idev->dev);
4782 if (event)
4783 ASSERT_RTNL();
4785 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4787 switch (event) {
4788 case RTM_NEWADDR:
4790 * If the address was optimistic
4791 * we inserted the route at the start of
4792 * our DAD process, so we don't need
4793 * to do it again
4795 if (!(ifp->rt->rt6i_node))
4796 ip6_ins_rt(ifp->rt);
4797 if (ifp->idev->cnf.forwarding)
4798 addrconf_join_anycast(ifp);
4799 if (!ipv6_addr_any(&ifp->peer_addr))
4800 addrconf_prefix_route(&ifp->peer_addr, 128,
4801 ifp->idev->dev, 0, 0);
4802 break;
4803 case RTM_DELADDR:
4804 if (ifp->idev->cnf.forwarding)
4805 addrconf_leave_anycast(ifp);
4806 addrconf_leave_solict(ifp->idev, &ifp->addr);
4807 if (!ipv6_addr_any(&ifp->peer_addr)) {
4808 struct rt6_info *rt;
4810 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
4811 ifp->idev->dev, 0, 0);
4812 if (rt && ip6_del_rt(rt))
4813 dst_free(&rt->dst);
4815 dst_hold(&ifp->rt->dst);
4817 if (ip6_del_rt(ifp->rt))
4818 dst_free(&ifp->rt->dst);
4820 rt_genid_bump_ipv6(net);
4821 break;
4823 atomic_inc(&net->ipv6.dev_addr_genid);
4826 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4828 rcu_read_lock_bh();
4829 if (likely(ifp->idev->dead == 0))
4830 __ipv6_ifa_notify(event, ifp);
4831 rcu_read_unlock_bh();
4834 #ifdef CONFIG_SYSCTL
4836 static
4837 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4838 void __user *buffer, size_t *lenp, loff_t *ppos)
4840 int *valp = ctl->data;
4841 int val = *valp;
4842 loff_t pos = *ppos;
4843 struct ctl_table lctl;
4844 int ret;
4847 * ctl->data points to idev->cnf.forwarding, we should
4848 * not modify it until we get the rtnl lock.
4850 lctl = *ctl;
4851 lctl.data = &val;
4853 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4855 if (write)
4856 ret = addrconf_fixup_forwarding(ctl, valp, val);
4857 if (ret)
4858 *ppos = pos;
4859 return ret;
4862 static void dev_disable_change(struct inet6_dev *idev)
4864 struct netdev_notifier_info info;
4866 if (!idev || !idev->dev)
4867 return;
4869 netdev_notifier_info_init(&info, idev->dev);
4870 if (idev->cnf.disable_ipv6)
4871 addrconf_notify(NULL, NETDEV_DOWN, &info);
4872 else
4873 addrconf_notify(NULL, NETDEV_UP, &info);
4876 static void addrconf_disable_change(struct net *net, __s32 newf)
4878 struct net_device *dev;
4879 struct inet6_dev *idev;
4881 rcu_read_lock();
4882 for_each_netdev_rcu(net, dev) {
4883 idev = __in6_dev_get(dev);
4884 if (idev) {
4885 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4886 idev->cnf.disable_ipv6 = newf;
4887 if (changed)
4888 dev_disable_change(idev);
4891 rcu_read_unlock();
4894 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4896 struct net *net;
4897 int old;
4899 if (!rtnl_trylock())
4900 return restart_syscall();
4902 net = (struct net *)table->extra2;
4903 old = *p;
4904 *p = newf;
4906 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4907 rtnl_unlock();
4908 return 0;
4911 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4912 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4913 addrconf_disable_change(net, newf);
4914 } else if ((!newf) ^ (!old))
4915 dev_disable_change((struct inet6_dev *)table->extra1);
4917 rtnl_unlock();
4918 return 0;
4921 static
4922 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4923 void __user *buffer, size_t *lenp, loff_t *ppos)
4925 int *valp = ctl->data;
4926 int val = *valp;
4927 loff_t pos = *ppos;
4928 struct ctl_table lctl;
4929 int ret;
4932 * ctl->data points to idev->cnf.disable_ipv6, we should
4933 * not modify it until we get the rtnl lock.
4935 lctl = *ctl;
4936 lctl.data = &val;
4938 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4940 if (write)
4941 ret = addrconf_disable_ipv6(ctl, valp, val);
4942 if (ret)
4943 *ppos = pos;
4944 return ret;
4947 static
4948 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4949 void __user *buffer, size_t *lenp, loff_t *ppos)
4951 int *valp = ctl->data;
4952 int ret;
4953 int old, new;
4955 old = *valp;
4956 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4957 new = *valp;
4959 if (write && old != new) {
4960 struct net *net = ctl->extra2;
4962 if (!rtnl_trylock())
4963 return restart_syscall();
4965 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4966 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4967 NETCONFA_IFINDEX_DEFAULT,
4968 net->ipv6.devconf_dflt);
4969 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4970 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4971 NETCONFA_IFINDEX_ALL,
4972 net->ipv6.devconf_all);
4973 else {
4974 struct inet6_dev *idev = ctl->extra1;
4976 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4977 idev->dev->ifindex,
4978 &idev->cnf);
4980 rtnl_unlock();
4983 return ret;
4987 static struct addrconf_sysctl_table
4989 struct ctl_table_header *sysctl_header;
4990 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4991 } addrconf_sysctl __read_mostly = {
4992 .sysctl_header = NULL,
4993 .addrconf_vars = {
4995 .procname = "forwarding",
4996 .data = &ipv6_devconf.forwarding,
4997 .maxlen = sizeof(int),
4998 .mode = 0644,
4999 .proc_handler = addrconf_sysctl_forward,
5002 .procname = "hop_limit",
5003 .data = &ipv6_devconf.hop_limit,
5004 .maxlen = sizeof(int),
5005 .mode = 0644,
5006 .proc_handler = proc_dointvec,
5009 .procname = "mtu",
5010 .data = &ipv6_devconf.mtu6,
5011 .maxlen = sizeof(int),
5012 .mode = 0644,
5013 .proc_handler = proc_dointvec,
5016 .procname = "accept_ra",
5017 .data = &ipv6_devconf.accept_ra,
5018 .maxlen = sizeof(int),
5019 .mode = 0644,
5020 .proc_handler = proc_dointvec,
5023 .procname = "accept_redirects",
5024 .data = &ipv6_devconf.accept_redirects,
5025 .maxlen = sizeof(int),
5026 .mode = 0644,
5027 .proc_handler = proc_dointvec,
5030 .procname = "autoconf",
5031 .data = &ipv6_devconf.autoconf,
5032 .maxlen = sizeof(int),
5033 .mode = 0644,
5034 .proc_handler = proc_dointvec,
5037 .procname = "dad_transmits",
5038 .data = &ipv6_devconf.dad_transmits,
5039 .maxlen = sizeof(int),
5040 .mode = 0644,
5041 .proc_handler = proc_dointvec,
5044 .procname = "router_solicitations",
5045 .data = &ipv6_devconf.rtr_solicits,
5046 .maxlen = sizeof(int),
5047 .mode = 0644,
5048 .proc_handler = proc_dointvec,
5051 .procname = "router_solicitation_interval",
5052 .data = &ipv6_devconf.rtr_solicit_interval,
5053 .maxlen = sizeof(int),
5054 .mode = 0644,
5055 .proc_handler = proc_dointvec_jiffies,
5058 .procname = "router_solicitation_delay",
5059 .data = &ipv6_devconf.rtr_solicit_delay,
5060 .maxlen = sizeof(int),
5061 .mode = 0644,
5062 .proc_handler = proc_dointvec_jiffies,
5065 .procname = "force_mld_version",
5066 .data = &ipv6_devconf.force_mld_version,
5067 .maxlen = sizeof(int),
5068 .mode = 0644,
5069 .proc_handler = proc_dointvec,
5072 .procname = "mldv1_unsolicited_report_interval",
5073 .data =
5074 &ipv6_devconf.mldv1_unsolicited_report_interval,
5075 .maxlen = sizeof(int),
5076 .mode = 0644,
5077 .proc_handler = proc_dointvec_ms_jiffies,
5080 .procname = "mldv2_unsolicited_report_interval",
5081 .data =
5082 &ipv6_devconf.mldv2_unsolicited_report_interval,
5083 .maxlen = sizeof(int),
5084 .mode = 0644,
5085 .proc_handler = proc_dointvec_ms_jiffies,
5088 .procname = "use_tempaddr",
5089 .data = &ipv6_devconf.use_tempaddr,
5090 .maxlen = sizeof(int),
5091 .mode = 0644,
5092 .proc_handler = proc_dointvec,
5095 .procname = "temp_valid_lft",
5096 .data = &ipv6_devconf.temp_valid_lft,
5097 .maxlen = sizeof(int),
5098 .mode = 0644,
5099 .proc_handler = proc_dointvec,
5102 .procname = "temp_prefered_lft",
5103 .data = &ipv6_devconf.temp_prefered_lft,
5104 .maxlen = sizeof(int),
5105 .mode = 0644,
5106 .proc_handler = proc_dointvec,
5109 .procname = "regen_max_retry",
5110 .data = &ipv6_devconf.regen_max_retry,
5111 .maxlen = sizeof(int),
5112 .mode = 0644,
5113 .proc_handler = proc_dointvec,
5116 .procname = "max_desync_factor",
5117 .data = &ipv6_devconf.max_desync_factor,
5118 .maxlen = sizeof(int),
5119 .mode = 0644,
5120 .proc_handler = proc_dointvec,
5123 .procname = "max_addresses",
5124 .data = &ipv6_devconf.max_addresses,
5125 .maxlen = sizeof(int),
5126 .mode = 0644,
5127 .proc_handler = proc_dointvec,
5130 .procname = "accept_ra_defrtr",
5131 .data = &ipv6_devconf.accept_ra_defrtr,
5132 .maxlen = sizeof(int),
5133 .mode = 0644,
5134 .proc_handler = proc_dointvec,
5137 .procname = "accept_ra_pinfo",
5138 .data = &ipv6_devconf.accept_ra_pinfo,
5139 .maxlen = sizeof(int),
5140 .mode = 0644,
5141 .proc_handler = proc_dointvec,
5143 #ifdef CONFIG_IPV6_ROUTER_PREF
5145 .procname = "accept_ra_rtr_pref",
5146 .data = &ipv6_devconf.accept_ra_rtr_pref,
5147 .maxlen = sizeof(int),
5148 .mode = 0644,
5149 .proc_handler = proc_dointvec,
5152 .procname = "router_probe_interval",
5153 .data = &ipv6_devconf.rtr_probe_interval,
5154 .maxlen = sizeof(int),
5155 .mode = 0644,
5156 .proc_handler = proc_dointvec_jiffies,
5158 #ifdef CONFIG_IPV6_ROUTE_INFO
5160 .procname = "accept_ra_rt_info_max_plen",
5161 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5162 .maxlen = sizeof(int),
5163 .mode = 0644,
5164 .proc_handler = proc_dointvec,
5166 #endif
5167 #endif
5169 .procname = "proxy_ndp",
5170 .data = &ipv6_devconf.proxy_ndp,
5171 .maxlen = sizeof(int),
5172 .mode = 0644,
5173 .proc_handler = addrconf_sysctl_proxy_ndp,
5176 .procname = "accept_source_route",
5177 .data = &ipv6_devconf.accept_source_route,
5178 .maxlen = sizeof(int),
5179 .mode = 0644,
5180 .proc_handler = proc_dointvec,
5182 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5184 .procname = "optimistic_dad",
5185 .data = &ipv6_devconf.optimistic_dad,
5186 .maxlen = sizeof(int),
5187 .mode = 0644,
5188 .proc_handler = proc_dointvec,
5192 .procname = "use_optimistic",
5193 .data = &ipv6_devconf.use_optimistic,
5194 .maxlen = sizeof(int),
5195 .mode = 0644,
5196 .proc_handler = proc_dointvec,
5199 #endif
5200 #ifdef CONFIG_IPV6_MROUTE
5202 .procname = "mc_forwarding",
5203 .data = &ipv6_devconf.mc_forwarding,
5204 .maxlen = sizeof(int),
5205 .mode = 0444,
5206 .proc_handler = proc_dointvec,
5208 #endif
5210 .procname = "disable_ipv6",
5211 .data = &ipv6_devconf.disable_ipv6,
5212 .maxlen = sizeof(int),
5213 .mode = 0644,
5214 .proc_handler = addrconf_sysctl_disable,
5217 .procname = "accept_dad",
5218 .data = &ipv6_devconf.accept_dad,
5219 .maxlen = sizeof(int),
5220 .mode = 0644,
5221 .proc_handler = proc_dointvec,
5224 .procname = "force_tllao",
5225 .data = &ipv6_devconf.force_tllao,
5226 .maxlen = sizeof(int),
5227 .mode = 0644,
5228 .proc_handler = proc_dointvec
5231 .procname = "ndisc_notify",
5232 .data = &ipv6_devconf.ndisc_notify,
5233 .maxlen = sizeof(int),
5234 .mode = 0644,
5235 .proc_handler = proc_dointvec
5238 .procname = "suppress_frag_ndisc",
5239 .data = &ipv6_devconf.suppress_frag_ndisc,
5240 .maxlen = sizeof(int),
5241 .mode = 0644,
5242 .proc_handler = proc_dointvec
5245 .procname = "accept_ra_from_local",
5246 .data = &ipv6_devconf.accept_ra_from_local,
5247 .maxlen = sizeof(int),
5248 .mode = 0644,
5249 .proc_handler = proc_dointvec,
5252 /* sentinel */
5257 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5258 struct inet6_dev *idev, struct ipv6_devconf *p)
5260 int i;
5261 struct addrconf_sysctl_table *t;
5262 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5264 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5265 if (t == NULL)
5266 goto out;
5268 for (i = 0; t->addrconf_vars[i].data; i++) {
5269 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5270 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5271 t->addrconf_vars[i].extra2 = net;
5274 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5276 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5277 if (t->sysctl_header == NULL)
5278 goto free;
5280 p->sysctl = t;
5281 return 0;
5283 free:
5284 kfree(t);
5285 out:
5286 return -ENOBUFS;
5289 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5291 struct addrconf_sysctl_table *t;
5293 if (p->sysctl == NULL)
5294 return;
5296 t = p->sysctl;
5297 p->sysctl = NULL;
5298 unregister_net_sysctl_table(t->sysctl_header);
5299 kfree(t);
5302 static int addrconf_sysctl_register(struct inet6_dev *idev)
5304 int err;
5306 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5307 return -EINVAL;
5309 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5310 &ndisc_ifinfo_sysctl_change);
5311 if (err)
5312 return err;
5313 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5314 idev, &idev->cnf);
5315 if (err)
5316 neigh_sysctl_unregister(idev->nd_parms);
5318 return err;
5321 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5323 __addrconf_sysctl_unregister(&idev->cnf);
5324 neigh_sysctl_unregister(idev->nd_parms);
5328 #endif
5330 static int __net_init addrconf_init_net(struct net *net)
5332 int err = -ENOMEM;
5333 struct ipv6_devconf *all, *dflt;
5335 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5336 if (all == NULL)
5337 goto err_alloc_all;
5339 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5340 if (dflt == NULL)
5341 goto err_alloc_dflt;
5343 /* these will be inherited by all namespaces */
5344 dflt->autoconf = ipv6_defaults.autoconf;
5345 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5347 net->ipv6.devconf_all = all;
5348 net->ipv6.devconf_dflt = dflt;
5350 #ifdef CONFIG_SYSCTL
5351 err = __addrconf_sysctl_register(net, "all", NULL, all);
5352 if (err < 0)
5353 goto err_reg_all;
5355 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5356 if (err < 0)
5357 goto err_reg_dflt;
5358 #endif
5359 return 0;
5361 #ifdef CONFIG_SYSCTL
5362 err_reg_dflt:
5363 __addrconf_sysctl_unregister(all);
5364 err_reg_all:
5365 kfree(dflt);
5366 #endif
5367 err_alloc_dflt:
5368 kfree(all);
5369 err_alloc_all:
5370 return err;
5373 static void __net_exit addrconf_exit_net(struct net *net)
5375 #ifdef CONFIG_SYSCTL
5376 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5377 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5378 #endif
5379 if (!net_eq(net, &init_net)) {
5380 kfree(net->ipv6.devconf_dflt);
5381 kfree(net->ipv6.devconf_all);
5385 static struct pernet_operations addrconf_ops = {
5386 .init = addrconf_init_net,
5387 .exit = addrconf_exit_net,
5390 static struct rtnl_af_ops inet6_ops = {
5391 .family = AF_INET6,
5392 .fill_link_af = inet6_fill_link_af,
5393 .get_link_af_size = inet6_get_link_af_size,
5394 .set_link_af = inet6_set_link_af,
5398 * Init / cleanup code
5401 int __init addrconf_init(void)
5403 struct inet6_dev *idev;
5404 int i, err;
5406 err = ipv6_addr_label_init();
5407 if (err < 0) {
5408 pr_crit("%s: cannot initialize default policy table: %d\n",
5409 __func__, err);
5410 goto out;
5413 err = register_pernet_subsys(&addrconf_ops);
5414 if (err < 0)
5415 goto out_addrlabel;
5417 addrconf_wq = create_workqueue("ipv6_addrconf");
5418 if (!addrconf_wq) {
5419 err = -ENOMEM;
5420 goto out_nowq;
5423 /* The addrconf netdev notifier requires that loopback_dev
5424 * has it's ipv6 private information allocated and setup
5425 * before it can bring up and give link-local addresses
5426 * to other devices which are up.
5428 * Unfortunately, loopback_dev is not necessarily the first
5429 * entry in the global dev_base list of net devices. In fact,
5430 * it is likely to be the very last entry on that list.
5431 * So this causes the notifier registry below to try and
5432 * give link-local addresses to all devices besides loopback_dev
5433 * first, then loopback_dev, which cases all the non-loopback_dev
5434 * devices to fail to get a link-local address.
5436 * So, as a temporary fix, allocate the ipv6 structure for
5437 * loopback_dev first by hand.
5438 * Longer term, all of the dependencies ipv6 has upon the loopback
5439 * device and it being up should be removed.
5441 rtnl_lock();
5442 idev = ipv6_add_dev(init_net.loopback_dev);
5443 rtnl_unlock();
5444 if (IS_ERR(idev)) {
5445 err = PTR_ERR(idev);
5446 goto errlo;
5449 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5450 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5452 register_netdevice_notifier(&ipv6_dev_notf);
5454 addrconf_verify();
5456 rtnl_af_register(&inet6_ops);
5458 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5459 NULL);
5460 if (err < 0)
5461 goto errout;
5463 /* Only the first call to __rtnl_register can fail */
5464 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5465 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5466 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5467 inet6_dump_ifaddr, NULL);
5468 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5469 inet6_dump_ifmcaddr, NULL);
5470 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5471 inet6_dump_ifacaddr, NULL);
5472 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5473 inet6_netconf_dump_devconf, NULL);
5475 ipv6_addr_label_rtnl_register();
5477 return 0;
5478 errout:
5479 rtnl_af_unregister(&inet6_ops);
5480 unregister_netdevice_notifier(&ipv6_dev_notf);
5481 errlo:
5482 destroy_workqueue(addrconf_wq);
5483 out_nowq:
5484 unregister_pernet_subsys(&addrconf_ops);
5485 out_addrlabel:
5486 ipv6_addr_label_cleanup();
5487 out:
5488 return err;
5491 void addrconf_cleanup(void)
5493 struct net_device *dev;
5494 int i;
5496 unregister_netdevice_notifier(&ipv6_dev_notf);
5497 unregister_pernet_subsys(&addrconf_ops);
5498 ipv6_addr_label_cleanup();
5500 rtnl_lock();
5502 __rtnl_af_unregister(&inet6_ops);
5504 /* clean dev list */
5505 for_each_netdev(&init_net, dev) {
5506 if (__in6_dev_get(dev) == NULL)
5507 continue;
5508 addrconf_ifdown(dev, 1);
5510 addrconf_ifdown(init_net.loopback_dev, 2);
5513 * Check hash table.
5515 spin_lock_bh(&addrconf_hash_lock);
5516 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5517 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5518 spin_unlock_bh(&addrconf_hash_lock);
5519 cancel_delayed_work(&addr_chk_work);
5520 rtnl_unlock();
5522 destroy_workqueue(addrconf_wq);