ipv6: move DAD and addrconf_verify processing to workqueue
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob6c7fa0853fc74ef179b00de52d78aecee342e18b
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 void addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
113 #else
114 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
121 #endif
123 static void __ipv6_regen_rndid(struct inet6_dev *idev);
124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
125 static void ipv6_regen_rndid(unsigned long data);
127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
128 static int ipv6_count_addresses(struct inet6_dev *idev);
131 * Configured unicast address hash table
133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
134 static DEFINE_SPINLOCK(addrconf_hash_lock);
136 static void addrconf_verify(void);
137 static void addrconf_verify_rtnl(void);
138 static void addrconf_verify_work(struct work_struct *);
140 static struct workqueue_struct *addrconf_wq;
141 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
143 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
144 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_type_change(struct net_device *dev,
147 unsigned long event);
148 static int addrconf_ifdown(struct net_device *dev, int how);
150 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
151 int plen,
152 const struct net_device *dev,
153 u32 flags, u32 noflags);
155 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
156 static void addrconf_dad_work(struct work_struct *w);
157 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
158 static void addrconf_dad_run(struct inet6_dev *idev);
159 static void addrconf_rs_timer(unsigned long data);
160 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
161 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
163 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
164 struct prefix_info *pinfo);
165 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
166 struct net_device *dev);
168 static struct ipv6_devconf ipv6_devconf __read_mostly = {
169 .forwarding = 0,
170 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
171 .mtu6 = IPV6_MIN_MTU,
172 .accept_ra = 1,
173 .accept_redirects = 1,
174 .autoconf = 1,
175 .force_mld_version = 0,
176 .mldv1_unsolicited_report_interval = 10 * HZ,
177 .mldv2_unsolicited_report_interval = HZ,
178 .dad_transmits = 1,
179 .rtr_solicits = MAX_RTR_SOLICITATIONS,
180 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
181 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
182 .use_tempaddr = 0,
183 .temp_valid_lft = TEMP_VALID_LIFETIME,
184 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
185 .regen_max_retry = REGEN_MAX_RETRY,
186 .max_desync_factor = MAX_DESYNC_FACTOR,
187 .max_addresses = IPV6_MAX_ADDRESSES,
188 .accept_ra_defrtr = 1,
189 .accept_ra_pinfo = 1,
190 #ifdef CONFIG_IPV6_ROUTER_PREF
191 .accept_ra_rtr_pref = 1,
192 .rtr_probe_interval = 60 * HZ,
193 #ifdef CONFIG_IPV6_ROUTE_INFO
194 .accept_ra_rt_info_max_plen = 0,
195 #endif
196 #endif
197 .proxy_ndp = 0,
198 .accept_source_route = 0, /* we do not accept RH0 by default. */
199 .disable_ipv6 = 0,
200 .accept_dad = 1,
201 .suppress_frag_ndisc = 1,
204 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
205 .forwarding = 0,
206 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
207 .mtu6 = IPV6_MIN_MTU,
208 .accept_ra = 1,
209 .accept_redirects = 1,
210 .autoconf = 1,
211 .force_mld_version = 0,
212 .mldv1_unsolicited_report_interval = 10 * HZ,
213 .mldv2_unsolicited_report_interval = HZ,
214 .dad_transmits = 1,
215 .rtr_solicits = MAX_RTR_SOLICITATIONS,
216 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
217 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
218 .use_tempaddr = 0,
219 .temp_valid_lft = TEMP_VALID_LIFETIME,
220 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
221 .regen_max_retry = REGEN_MAX_RETRY,
222 .max_desync_factor = MAX_DESYNC_FACTOR,
223 .max_addresses = IPV6_MAX_ADDRESSES,
224 .accept_ra_defrtr = 1,
225 .accept_ra_pinfo = 1,
226 #ifdef CONFIG_IPV6_ROUTER_PREF
227 .accept_ra_rtr_pref = 1,
228 .rtr_probe_interval = 60 * HZ,
229 #ifdef CONFIG_IPV6_ROUTE_INFO
230 .accept_ra_rt_info_max_plen = 0,
231 #endif
232 #endif
233 .proxy_ndp = 0,
234 .accept_source_route = 0, /* we do not accept RH0 by default. */
235 .disable_ipv6 = 0,
236 .accept_dad = 1,
237 .suppress_frag_ndisc = 1,
240 /* Check if a valid qdisc is available */
241 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
243 return !qdisc_tx_is_noop(dev);
246 static void addrconf_del_rs_timer(struct inet6_dev *idev)
248 if (del_timer(&idev->rs_timer))
249 __in6_dev_put(idev);
252 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
254 if (cancel_delayed_work(&ifp->dad_work))
255 __in6_ifa_put(ifp);
258 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
259 unsigned long when)
261 if (!timer_pending(&idev->rs_timer))
262 in6_dev_hold(idev);
263 mod_timer(&idev->rs_timer, jiffies + when);
266 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
267 unsigned long delay)
269 if (!delayed_work_pending(&ifp->dad_work))
270 in6_ifa_hold(ifp);
271 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
274 static int snmp6_alloc_dev(struct inet6_dev *idev)
276 int i;
278 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
279 sizeof(struct ipstats_mib),
280 __alignof__(struct ipstats_mib)) < 0)
281 goto err_ip;
283 for_each_possible_cpu(i) {
284 struct ipstats_mib *addrconf_stats;
285 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
286 u64_stats_init(&addrconf_stats->syncp);
287 #if SNMP_ARRAY_SZ == 2
288 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
289 u64_stats_init(&addrconf_stats->syncp);
290 #endif
294 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
295 GFP_KERNEL);
296 if (!idev->stats.icmpv6dev)
297 goto err_icmp;
298 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
299 GFP_KERNEL);
300 if (!idev->stats.icmpv6msgdev)
301 goto err_icmpmsg;
303 return 0;
305 err_icmpmsg:
306 kfree(idev->stats.icmpv6dev);
307 err_icmp:
308 snmp_mib_free((void __percpu **)idev->stats.ipv6);
309 err_ip:
310 return -ENOMEM;
313 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
315 struct inet6_dev *ndev;
317 ASSERT_RTNL();
319 if (dev->mtu < IPV6_MIN_MTU)
320 return NULL;
322 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
324 if (ndev == NULL)
325 return NULL;
327 rwlock_init(&ndev->lock);
328 ndev->dev = dev;
329 INIT_LIST_HEAD(&ndev->addr_list);
330 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
331 (unsigned long)ndev);
332 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
333 ndev->cnf.mtu6 = dev->mtu;
334 ndev->cnf.sysctl = NULL;
335 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
336 if (ndev->nd_parms == NULL) {
337 kfree(ndev);
338 return NULL;
340 if (ndev->cnf.forwarding)
341 dev_disable_lro(dev);
342 /* We refer to the device */
343 dev_hold(dev);
345 if (snmp6_alloc_dev(ndev) < 0) {
346 ADBG(KERN_WARNING
347 "%s: cannot allocate memory for statistics; dev=%s.\n",
348 __func__, dev->name);
349 neigh_parms_release(&nd_tbl, ndev->nd_parms);
350 dev_put(dev);
351 kfree(ndev);
352 return NULL;
355 if (snmp6_register_dev(ndev) < 0) {
356 ADBG(KERN_WARNING
357 "%s: cannot create /proc/net/dev_snmp6/%s\n",
358 __func__, dev->name);
359 neigh_parms_release(&nd_tbl, ndev->nd_parms);
360 ndev->dead = 1;
361 in6_dev_finish_destroy(ndev);
362 return NULL;
365 /* One reference from device. We must do this before
366 * we invoke __ipv6_regen_rndid().
368 in6_dev_hold(ndev);
370 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
371 ndev->cnf.accept_dad = -1;
373 #if IS_ENABLED(CONFIG_IPV6_SIT)
374 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
375 pr_info("%s: Disabled Multicast RS\n", dev->name);
376 ndev->cnf.rtr_solicits = 0;
378 #endif
380 INIT_LIST_HEAD(&ndev->tempaddr_list);
381 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
382 if ((dev->flags&IFF_LOOPBACK) ||
383 dev->type == ARPHRD_TUNNEL ||
384 dev->type == ARPHRD_TUNNEL6 ||
385 dev->type == ARPHRD_SIT ||
386 dev->type == ARPHRD_NONE) {
387 ndev->cnf.use_tempaddr = -1;
388 } else {
389 in6_dev_hold(ndev);
390 ipv6_regen_rndid((unsigned long) ndev);
393 ndev->token = in6addr_any;
395 if (netif_running(dev) && addrconf_qdisc_ok(dev))
396 ndev->if_flags |= IF_READY;
398 ipv6_mc_init_dev(ndev);
399 ndev->tstamp = jiffies;
400 addrconf_sysctl_register(ndev);
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;
417 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
419 struct inet6_dev *idev;
421 ASSERT_RTNL();
423 idev = __in6_dev_get(dev);
424 if (!idev) {
425 idev = ipv6_add_dev(dev);
426 if (!idev)
427 return NULL;
430 if (dev->flags&IFF_UP)
431 ipv6_mc_up(idev);
432 return idev;
435 static int inet6_netconf_msgsize_devconf(int type)
437 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
438 + nla_total_size(4); /* NETCONFA_IFINDEX */
440 /* type -1 is used for ALL */
441 if (type == -1 || type == NETCONFA_FORWARDING)
442 size += nla_total_size(4);
443 #ifdef CONFIG_IPV6_MROUTE
444 if (type == -1 || type == NETCONFA_MC_FORWARDING)
445 size += nla_total_size(4);
446 #endif
447 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
448 size += nla_total_size(4);
450 return size;
453 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
454 struct ipv6_devconf *devconf, u32 portid,
455 u32 seq, int event, unsigned int flags,
456 int type)
458 struct nlmsghdr *nlh;
459 struct netconfmsg *ncm;
461 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
462 flags);
463 if (nlh == NULL)
464 return -EMSGSIZE;
466 ncm = nlmsg_data(nlh);
467 ncm->ncm_family = AF_INET6;
469 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
470 goto nla_put_failure;
472 /* type -1 is used for ALL */
473 if ((type == -1 || type == NETCONFA_FORWARDING) &&
474 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
475 goto nla_put_failure;
476 #ifdef CONFIG_IPV6_MROUTE
477 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
478 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
479 devconf->mc_forwarding) < 0)
480 goto nla_put_failure;
481 #endif
482 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
483 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
484 goto nla_put_failure;
486 return nlmsg_end(skb, nlh);
488 nla_put_failure:
489 nlmsg_cancel(skb, nlh);
490 return -EMSGSIZE;
493 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
494 struct ipv6_devconf *devconf)
496 struct sk_buff *skb;
497 int err = -ENOBUFS;
499 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
500 if (skb == NULL)
501 goto errout;
503 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
504 RTM_NEWNETCONF, 0, type);
505 if (err < 0) {
506 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
507 WARN_ON(err == -EMSGSIZE);
508 kfree_skb(skb);
509 goto errout;
511 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
512 return;
513 errout:
514 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
517 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
518 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
519 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
520 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
523 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
524 struct nlmsghdr *nlh)
526 struct net *net = sock_net(in_skb->sk);
527 struct nlattr *tb[NETCONFA_MAX+1];
528 struct netconfmsg *ncm;
529 struct sk_buff *skb;
530 struct ipv6_devconf *devconf;
531 struct inet6_dev *in6_dev;
532 struct net_device *dev;
533 int ifindex;
534 int err;
536 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
537 devconf_ipv6_policy);
538 if (err < 0)
539 goto errout;
541 err = EINVAL;
542 if (!tb[NETCONFA_IFINDEX])
543 goto errout;
545 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
546 switch (ifindex) {
547 case NETCONFA_IFINDEX_ALL:
548 devconf = net->ipv6.devconf_all;
549 break;
550 case NETCONFA_IFINDEX_DEFAULT:
551 devconf = net->ipv6.devconf_dflt;
552 break;
553 default:
554 dev = __dev_get_by_index(net, ifindex);
555 if (dev == NULL)
556 goto errout;
557 in6_dev = __in6_dev_get(dev);
558 if (in6_dev == NULL)
559 goto errout;
560 devconf = &in6_dev->cnf;
561 break;
564 err = -ENOBUFS;
565 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
566 if (skb == NULL)
567 goto errout;
569 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
570 NETLINK_CB(in_skb).portid,
571 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
572 -1);
573 if (err < 0) {
574 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
575 WARN_ON(err == -EMSGSIZE);
576 kfree_skb(skb);
577 goto errout;
579 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
580 errout:
581 return err;
584 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
585 struct netlink_callback *cb)
587 struct net *net = sock_net(skb->sk);
588 int h, s_h;
589 int idx, s_idx;
590 struct net_device *dev;
591 struct inet6_dev *idev;
592 struct hlist_head *head;
594 s_h = cb->args[0];
595 s_idx = idx = cb->args[1];
597 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
598 idx = 0;
599 head = &net->dev_index_head[h];
600 rcu_read_lock();
601 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
602 net->dev_base_seq;
603 hlist_for_each_entry_rcu(dev, head, index_hlist) {
604 if (idx < s_idx)
605 goto cont;
606 idev = __in6_dev_get(dev);
607 if (!idev)
608 goto cont;
610 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
611 &idev->cnf,
612 NETLINK_CB(cb->skb).portid,
613 cb->nlh->nlmsg_seq,
614 RTM_NEWNETCONF,
615 NLM_F_MULTI,
616 -1) <= 0) {
617 rcu_read_unlock();
618 goto done;
620 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
621 cont:
622 idx++;
624 rcu_read_unlock();
626 if (h == NETDEV_HASHENTRIES) {
627 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
628 net->ipv6.devconf_all,
629 NETLINK_CB(cb->skb).portid,
630 cb->nlh->nlmsg_seq,
631 RTM_NEWNETCONF, NLM_F_MULTI,
632 -1) <= 0)
633 goto done;
634 else
635 h++;
637 if (h == NETDEV_HASHENTRIES + 1) {
638 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
639 net->ipv6.devconf_dflt,
640 NETLINK_CB(cb->skb).portid,
641 cb->nlh->nlmsg_seq,
642 RTM_NEWNETCONF, NLM_F_MULTI,
643 -1) <= 0)
644 goto done;
645 else
646 h++;
648 done:
649 cb->args[0] = h;
650 cb->args[1] = idx;
652 return skb->len;
655 #ifdef CONFIG_SYSCTL
656 static void dev_forward_change(struct inet6_dev *idev)
658 struct net_device *dev;
659 struct inet6_ifaddr *ifa;
661 if (!idev)
662 return;
663 dev = idev->dev;
664 if (idev->cnf.forwarding)
665 dev_disable_lro(dev);
666 if (dev->flags & IFF_MULTICAST) {
667 if (idev->cnf.forwarding) {
668 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
669 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
670 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
671 } else {
672 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
673 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
674 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
678 list_for_each_entry(ifa, &idev->addr_list, if_list) {
679 if (ifa->flags&IFA_F_TENTATIVE)
680 continue;
681 if (idev->cnf.forwarding)
682 addrconf_join_anycast(ifa);
683 else
684 addrconf_leave_anycast(ifa);
686 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
687 dev->ifindex, &idev->cnf);
691 static void addrconf_forward_change(struct net *net, __s32 newf)
693 struct net_device *dev;
694 struct inet6_dev *idev;
696 for_each_netdev(net, dev) {
697 idev = __in6_dev_get(dev);
698 if (idev) {
699 int changed = (!idev->cnf.forwarding) ^ (!newf);
700 idev->cnf.forwarding = newf;
701 if (changed)
702 dev_forward_change(idev);
707 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
709 struct net *net;
710 int old;
712 if (!rtnl_trylock())
713 return restart_syscall();
715 net = (struct net *)table->extra2;
716 old = *p;
717 *p = newf;
719 if (p == &net->ipv6.devconf_dflt->forwarding) {
720 if ((!newf) ^ (!old))
721 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
722 NETCONFA_IFINDEX_DEFAULT,
723 net->ipv6.devconf_dflt);
724 rtnl_unlock();
725 return 0;
728 if (p == &net->ipv6.devconf_all->forwarding) {
729 net->ipv6.devconf_dflt->forwarding = newf;
730 addrconf_forward_change(net, newf);
731 if ((!newf) ^ (!old))
732 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
733 NETCONFA_IFINDEX_ALL,
734 net->ipv6.devconf_all);
735 } else if ((!newf) ^ (!old))
736 dev_forward_change((struct inet6_dev *)table->extra1);
737 rtnl_unlock();
739 if (newf)
740 rt6_purge_dflt_routers(net);
741 return 1;
743 #endif
745 /* Nobody refers to this ifaddr, destroy it */
746 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
748 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
750 #ifdef NET_REFCNT_DEBUG
751 pr_debug("%s\n", __func__);
752 #endif
754 in6_dev_put(ifp->idev);
756 if (cancel_delayed_work(&ifp->dad_work))
757 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
758 ifp);
760 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
761 pr_warn("Freeing alive inet6 address %p\n", ifp);
762 return;
764 ip6_rt_put(ifp->rt);
766 kfree_rcu(ifp, rcu);
769 static void
770 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
772 struct list_head *p;
773 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
776 * Each device address list is sorted in order of scope -
777 * global before linklocal.
779 list_for_each(p, &idev->addr_list) {
780 struct inet6_ifaddr *ifa
781 = list_entry(p, struct inet6_ifaddr, if_list);
782 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
783 break;
786 list_add_tail(&ifp->if_list, p);
789 static u32 inet6_addr_hash(const struct in6_addr *addr)
791 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
794 /* On success it returns ifp with increased reference count */
796 static struct inet6_ifaddr *
797 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
798 const struct in6_addr *peer_addr, int pfxlen,
799 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
801 struct inet6_ifaddr *ifa = NULL;
802 struct rt6_info *rt;
803 unsigned int hash;
804 int err = 0;
805 int addr_type = ipv6_addr_type(addr);
807 if (addr_type == IPV6_ADDR_ANY ||
808 addr_type & IPV6_ADDR_MULTICAST ||
809 (!(idev->dev->flags & IFF_LOOPBACK) &&
810 addr_type & IPV6_ADDR_LOOPBACK))
811 return ERR_PTR(-EADDRNOTAVAIL);
813 rcu_read_lock_bh();
814 if (idev->dead) {
815 err = -ENODEV; /*XXX*/
816 goto out2;
819 if (idev->cnf.disable_ipv6) {
820 err = -EACCES;
821 goto out2;
824 spin_lock(&addrconf_hash_lock);
826 /* Ignore adding duplicate addresses on an interface */
827 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
828 ADBG("ipv6_add_addr: already assigned\n");
829 err = -EEXIST;
830 goto out;
833 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
835 if (ifa == NULL) {
836 ADBG("ipv6_add_addr: malloc failed\n");
837 err = -ENOBUFS;
838 goto out;
841 rt = addrconf_dst_alloc(idev, addr, false);
842 if (IS_ERR(rt)) {
843 err = PTR_ERR(rt);
844 goto out;
847 neigh_parms_data_state_setall(idev->nd_parms);
849 ifa->addr = *addr;
850 if (peer_addr)
851 ifa->peer_addr = *peer_addr;
853 spin_lock_init(&ifa->lock);
854 spin_lock_init(&ifa->state_lock);
855 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
856 INIT_HLIST_NODE(&ifa->addr_lst);
857 ifa->scope = scope;
858 ifa->prefix_len = pfxlen;
859 ifa->flags = flags | IFA_F_TENTATIVE;
860 ifa->valid_lft = valid_lft;
861 ifa->prefered_lft = prefered_lft;
862 ifa->cstamp = ifa->tstamp = jiffies;
863 ifa->tokenized = false;
865 ifa->rt = rt;
867 ifa->idev = idev;
868 in6_dev_hold(idev);
869 /* For caller */
870 in6_ifa_hold(ifa);
872 /* Add to big hash table */
873 hash = inet6_addr_hash(addr);
875 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
876 spin_unlock(&addrconf_hash_lock);
878 write_lock(&idev->lock);
879 /* Add to inet6_dev unicast addr list. */
880 ipv6_link_dev_addr(idev, ifa);
882 if (ifa->flags&IFA_F_TEMPORARY) {
883 list_add(&ifa->tmp_list, &idev->tempaddr_list);
884 in6_ifa_hold(ifa);
887 in6_ifa_hold(ifa);
888 write_unlock(&idev->lock);
889 out2:
890 rcu_read_unlock_bh();
892 if (likely(err == 0))
893 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
894 else {
895 kfree(ifa);
896 ifa = ERR_PTR(err);
899 return ifa;
900 out:
901 spin_unlock(&addrconf_hash_lock);
902 goto out2;
905 enum cleanup_prefix_rt_t {
906 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
907 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
908 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
912 * Check, whether the prefix for ifp would still need a prefix route
913 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
914 * constants.
916 * 1) we don't purge prefix if address was not permanent.
917 * prefix is managed by its own lifetime.
918 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
919 * 3) if there are no addresses, delete prefix.
920 * 4) if there are still other permanent address(es),
921 * corresponding prefix is still permanent.
922 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
923 * don't purge the prefix, assume user space is managing it.
924 * 6) otherwise, update prefix lifetime to the
925 * longest valid lifetime among the corresponding
926 * addresses on the device.
927 * Note: subsequent RA will update lifetime.
929 static enum cleanup_prefix_rt_t
930 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
932 struct inet6_ifaddr *ifa;
933 struct inet6_dev *idev = ifp->idev;
934 unsigned long lifetime;
935 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
937 *expires = jiffies;
939 list_for_each_entry(ifa, &idev->addr_list, if_list) {
940 if (ifa == ifp)
941 continue;
942 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
943 ifp->prefix_len))
944 continue;
945 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
946 return CLEANUP_PREFIX_RT_NOP;
948 action = CLEANUP_PREFIX_RT_EXPIRE;
950 spin_lock(&ifa->lock);
952 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
954 * Note: Because this address is
955 * not permanent, lifetime <
956 * LONG_MAX / HZ here.
958 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
959 *expires = ifa->tstamp + lifetime * HZ;
960 spin_unlock(&ifa->lock);
963 return action;
966 static void
967 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
969 struct rt6_info *rt;
971 rt = addrconf_get_prefix_route(&ifp->addr,
972 ifp->prefix_len,
973 ifp->idev->dev,
974 0, RTF_GATEWAY | RTF_DEFAULT);
975 if (rt) {
976 if (del_rt)
977 ip6_del_rt(rt);
978 else {
979 if (!(rt->rt6i_flags & RTF_EXPIRES))
980 rt6_set_expires(rt, expires);
981 ip6_rt_put(rt);
987 /* This function wants to get referenced ifp and releases it before return */
989 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
991 int state;
992 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
993 unsigned long expires;
995 ASSERT_RTNL();
997 spin_lock_bh(&ifp->state_lock);
998 state = ifp->state;
999 ifp->state = INET6_IFADDR_STATE_DEAD;
1000 spin_unlock_bh(&ifp->state_lock);
1002 if (state == INET6_IFADDR_STATE_DEAD)
1003 goto out;
1005 spin_lock_bh(&addrconf_hash_lock);
1006 hlist_del_init_rcu(&ifp->addr_lst);
1007 spin_unlock_bh(&addrconf_hash_lock);
1009 write_lock_bh(&ifp->idev->lock);
1011 if (ifp->flags&IFA_F_TEMPORARY) {
1012 list_del(&ifp->tmp_list);
1013 if (ifp->ifpub) {
1014 in6_ifa_put(ifp->ifpub);
1015 ifp->ifpub = NULL;
1017 __in6_ifa_put(ifp);
1020 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1021 action = check_cleanup_prefix_route(ifp, &expires);
1023 list_del_init(&ifp->if_list);
1024 __in6_ifa_put(ifp);
1026 write_unlock_bh(&ifp->idev->lock);
1028 addrconf_del_dad_work(ifp);
1030 ipv6_ifa_notify(RTM_DELADDR, ifp);
1032 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1034 if (action != CLEANUP_PREFIX_RT_NOP) {
1035 cleanup_prefix_route(ifp, expires,
1036 action == CLEANUP_PREFIX_RT_DEL);
1039 /* clean up prefsrc entries */
1040 rt6_remove_prefsrc(ifp);
1041 out:
1042 in6_ifa_put(ifp);
1045 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1047 struct inet6_dev *idev = ifp->idev;
1048 struct in6_addr addr, *tmpaddr;
1049 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1050 unsigned long regen_advance;
1051 int tmp_plen;
1052 int ret = 0;
1053 u32 addr_flags;
1054 unsigned long now = jiffies;
1056 write_lock_bh(&idev->lock);
1057 if (ift) {
1058 spin_lock_bh(&ift->lock);
1059 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1060 spin_unlock_bh(&ift->lock);
1061 tmpaddr = &addr;
1062 } else {
1063 tmpaddr = NULL;
1065 retry:
1066 in6_dev_hold(idev);
1067 if (idev->cnf.use_tempaddr <= 0) {
1068 write_unlock_bh(&idev->lock);
1069 pr_info("%s: use_tempaddr is disabled\n", __func__);
1070 in6_dev_put(idev);
1071 ret = -1;
1072 goto out;
1074 spin_lock_bh(&ifp->lock);
1075 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1076 idev->cnf.use_tempaddr = -1; /*XXX*/
1077 spin_unlock_bh(&ifp->lock);
1078 write_unlock_bh(&idev->lock);
1079 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1080 __func__);
1081 in6_dev_put(idev);
1082 ret = -1;
1083 goto out;
1085 in6_ifa_hold(ifp);
1086 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1087 __ipv6_try_regen_rndid(idev, tmpaddr);
1088 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1089 age = (now - ifp->tstamp) / HZ;
1090 tmp_valid_lft = min_t(__u32,
1091 ifp->valid_lft,
1092 idev->cnf.temp_valid_lft + age);
1093 tmp_prefered_lft = min_t(__u32,
1094 ifp->prefered_lft,
1095 idev->cnf.temp_prefered_lft + age -
1096 idev->cnf.max_desync_factor);
1097 tmp_plen = ifp->prefix_len;
1098 tmp_tstamp = ifp->tstamp;
1099 spin_unlock_bh(&ifp->lock);
1101 regen_advance = idev->cnf.regen_max_retry *
1102 idev->cnf.dad_transmits *
1103 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1104 write_unlock_bh(&idev->lock);
1106 /* A temporary address is created only if this calculated Preferred
1107 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1108 * an implementation must not create a temporary address with a zero
1109 * Preferred Lifetime.
1110 * Use age calculation as in addrconf_verify to avoid unnecessary
1111 * temporary addresses being generated.
1113 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1114 if (tmp_prefered_lft <= regen_advance + age) {
1115 in6_ifa_put(ifp);
1116 in6_dev_put(idev);
1117 ret = -1;
1118 goto out;
1121 addr_flags = IFA_F_TEMPORARY;
1122 /* set in addrconf_prefix_rcv() */
1123 if (ifp->flags & IFA_F_OPTIMISTIC)
1124 addr_flags |= IFA_F_OPTIMISTIC;
1126 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1127 ipv6_addr_scope(&addr), addr_flags,
1128 tmp_valid_lft, tmp_prefered_lft);
1129 if (IS_ERR(ift)) {
1130 in6_ifa_put(ifp);
1131 in6_dev_put(idev);
1132 pr_info("%s: retry temporary address regeneration\n", __func__);
1133 tmpaddr = &addr;
1134 write_lock_bh(&idev->lock);
1135 goto retry;
1138 spin_lock_bh(&ift->lock);
1139 ift->ifpub = ifp;
1140 ift->cstamp = now;
1141 ift->tstamp = tmp_tstamp;
1142 spin_unlock_bh(&ift->lock);
1144 addrconf_dad_start(ift);
1145 in6_ifa_put(ift);
1146 in6_dev_put(idev);
1147 out:
1148 return ret;
1152 * Choose an appropriate source address (RFC3484)
1154 enum {
1155 IPV6_SADDR_RULE_INIT = 0,
1156 IPV6_SADDR_RULE_LOCAL,
1157 IPV6_SADDR_RULE_SCOPE,
1158 IPV6_SADDR_RULE_PREFERRED,
1159 #ifdef CONFIG_IPV6_MIP6
1160 IPV6_SADDR_RULE_HOA,
1161 #endif
1162 IPV6_SADDR_RULE_OIF,
1163 IPV6_SADDR_RULE_LABEL,
1164 IPV6_SADDR_RULE_PRIVACY,
1165 IPV6_SADDR_RULE_ORCHID,
1166 IPV6_SADDR_RULE_PREFIX,
1167 IPV6_SADDR_RULE_MAX
1170 struct ipv6_saddr_score {
1171 int rule;
1172 int addr_type;
1173 struct inet6_ifaddr *ifa;
1174 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1175 int scopedist;
1176 int matchlen;
1179 struct ipv6_saddr_dst {
1180 const struct in6_addr *addr;
1181 int ifindex;
1182 int scope;
1183 int label;
1184 unsigned int prefs;
1187 static inline int ipv6_saddr_preferred(int type)
1189 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1190 return 1;
1191 return 0;
1194 static int ipv6_get_saddr_eval(struct net *net,
1195 struct ipv6_saddr_score *score,
1196 struct ipv6_saddr_dst *dst,
1197 int i)
1199 int ret;
1201 if (i <= score->rule) {
1202 switch (i) {
1203 case IPV6_SADDR_RULE_SCOPE:
1204 ret = score->scopedist;
1205 break;
1206 case IPV6_SADDR_RULE_PREFIX:
1207 ret = score->matchlen;
1208 break;
1209 default:
1210 ret = !!test_bit(i, score->scorebits);
1212 goto out;
1215 switch (i) {
1216 case IPV6_SADDR_RULE_INIT:
1217 /* Rule 0: remember if hiscore is not ready yet */
1218 ret = !!score->ifa;
1219 break;
1220 case IPV6_SADDR_RULE_LOCAL:
1221 /* Rule 1: Prefer same address */
1222 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1223 break;
1224 case IPV6_SADDR_RULE_SCOPE:
1225 /* Rule 2: Prefer appropriate scope
1227 * ret
1229 * -1 | d 15
1230 * ---+--+-+---> scope
1232 * | d is scope of the destination.
1233 * B-d | \
1234 * | \ <- smaller scope is better if
1235 * B-15 | \ if scope is enough for destination.
1236 * | ret = B - scope (-1 <= scope >= d <= 15).
1237 * d-C-1 | /
1238 * |/ <- greater is better
1239 * -C / if scope is not enough for destination.
1240 * /| ret = scope - C (-1 <= d < scope <= 15).
1242 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1243 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1244 * Assume B = 0 and we get C > 29.
1246 ret = __ipv6_addr_src_scope(score->addr_type);
1247 if (ret >= dst->scope)
1248 ret = -ret;
1249 else
1250 ret -= 128; /* 30 is enough */
1251 score->scopedist = ret;
1252 break;
1253 case IPV6_SADDR_RULE_PREFERRED:
1254 /* Rule 3: Avoid deprecated and optimistic addresses */
1255 ret = ipv6_saddr_preferred(score->addr_type) ||
1256 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1257 break;
1258 #ifdef CONFIG_IPV6_MIP6
1259 case IPV6_SADDR_RULE_HOA:
1261 /* Rule 4: Prefer home address */
1262 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1263 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1264 break;
1266 #endif
1267 case IPV6_SADDR_RULE_OIF:
1268 /* Rule 5: Prefer outgoing interface */
1269 ret = (!dst->ifindex ||
1270 dst->ifindex == score->ifa->idev->dev->ifindex);
1271 break;
1272 case IPV6_SADDR_RULE_LABEL:
1273 /* Rule 6: Prefer matching label */
1274 ret = ipv6_addr_label(net,
1275 &score->ifa->addr, score->addr_type,
1276 score->ifa->idev->dev->ifindex) == dst->label;
1277 break;
1278 case IPV6_SADDR_RULE_PRIVACY:
1280 /* Rule 7: Prefer public address
1281 * Note: prefer temporary address if use_tempaddr >= 2
1283 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1284 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1285 score->ifa->idev->cnf.use_tempaddr >= 2;
1286 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1287 break;
1289 case IPV6_SADDR_RULE_ORCHID:
1290 /* Rule 8-: Prefer ORCHID vs ORCHID or
1291 * non-ORCHID vs non-ORCHID
1293 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1294 ipv6_addr_orchid(dst->addr));
1295 break;
1296 case IPV6_SADDR_RULE_PREFIX:
1297 /* Rule 8: Use longest matching prefix */
1298 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1299 if (ret > score->ifa->prefix_len)
1300 ret = score->ifa->prefix_len;
1301 score->matchlen = ret;
1302 break;
1303 default:
1304 ret = 0;
1307 if (ret)
1308 __set_bit(i, score->scorebits);
1309 score->rule = i;
1310 out:
1311 return ret;
1314 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1315 const struct in6_addr *daddr, unsigned int prefs,
1316 struct in6_addr *saddr)
1318 struct ipv6_saddr_score scores[2],
1319 *score = &scores[0], *hiscore = &scores[1];
1320 struct ipv6_saddr_dst dst;
1321 struct net_device *dev;
1322 int dst_type;
1324 dst_type = __ipv6_addr_type(daddr);
1325 dst.addr = daddr;
1326 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1327 dst.scope = __ipv6_addr_src_scope(dst_type);
1328 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1329 dst.prefs = prefs;
1331 hiscore->rule = -1;
1332 hiscore->ifa = NULL;
1334 rcu_read_lock();
1336 for_each_netdev_rcu(net, dev) {
1337 struct inet6_dev *idev;
1339 /* Candidate Source Address (section 4)
1340 * - multicast and link-local destination address,
1341 * the set of candidate source address MUST only
1342 * include addresses assigned to interfaces
1343 * belonging to the same link as the outgoing
1344 * interface.
1345 * (- For site-local destination addresses, the
1346 * set of candidate source addresses MUST only
1347 * include addresses assigned to interfaces
1348 * belonging to the same site as the outgoing
1349 * interface.)
1351 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1352 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1353 dst.ifindex && dev->ifindex != dst.ifindex)
1354 continue;
1356 idev = __in6_dev_get(dev);
1357 if (!idev)
1358 continue;
1360 read_lock_bh(&idev->lock);
1361 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1362 int i;
1365 * - Tentative Address (RFC2462 section 5.4)
1366 * - A tentative address is not considered
1367 * "assigned to an interface" in the traditional
1368 * sense, unless it is also flagged as optimistic.
1369 * - Candidate Source Address (section 4)
1370 * - In any case, anycast addresses, multicast
1371 * addresses, and the unspecified address MUST
1372 * NOT be included in a candidate set.
1374 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1375 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1376 continue;
1378 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1380 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1381 score->addr_type & IPV6_ADDR_MULTICAST)) {
1382 LIMIT_NETDEBUG(KERN_DEBUG
1383 "ADDRCONF: unspecified / multicast address "
1384 "assigned as unicast address on %s",
1385 dev->name);
1386 continue;
1389 score->rule = -1;
1390 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1392 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1393 int minihiscore, miniscore;
1395 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1396 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1398 if (minihiscore > miniscore) {
1399 if (i == IPV6_SADDR_RULE_SCOPE &&
1400 score->scopedist > 0) {
1402 * special case:
1403 * each remaining entry
1404 * has too small (not enough)
1405 * scope, because ifa entries
1406 * are sorted by their scope
1407 * values.
1409 goto try_nextdev;
1411 break;
1412 } else if (minihiscore < miniscore) {
1413 if (hiscore->ifa)
1414 in6_ifa_put(hiscore->ifa);
1416 in6_ifa_hold(score->ifa);
1418 swap(hiscore, score);
1420 /* restore our iterator */
1421 score->ifa = hiscore->ifa;
1423 break;
1427 try_nextdev:
1428 read_unlock_bh(&idev->lock);
1430 rcu_read_unlock();
1432 if (!hiscore->ifa)
1433 return -EADDRNOTAVAIL;
1435 *saddr = hiscore->ifa->addr;
1436 in6_ifa_put(hiscore->ifa);
1437 return 0;
1439 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1441 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1442 u32 banned_flags)
1444 struct inet6_ifaddr *ifp;
1445 int err = -EADDRNOTAVAIL;
1447 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1448 if (ifp->scope > IFA_LINK)
1449 break;
1450 if (ifp->scope == IFA_LINK &&
1451 !(ifp->flags & banned_flags)) {
1452 *addr = ifp->addr;
1453 err = 0;
1454 break;
1457 return err;
1460 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1461 u32 banned_flags)
1463 struct inet6_dev *idev;
1464 int err = -EADDRNOTAVAIL;
1466 rcu_read_lock();
1467 idev = __in6_dev_get(dev);
1468 if (idev) {
1469 read_lock_bh(&idev->lock);
1470 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1471 read_unlock_bh(&idev->lock);
1473 rcu_read_unlock();
1474 return err;
1477 static int ipv6_count_addresses(struct inet6_dev *idev)
1479 int cnt = 0;
1480 struct inet6_ifaddr *ifp;
1482 read_lock_bh(&idev->lock);
1483 list_for_each_entry(ifp, &idev->addr_list, if_list)
1484 cnt++;
1485 read_unlock_bh(&idev->lock);
1486 return cnt;
1489 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1490 const struct net_device *dev, int strict)
1492 struct inet6_ifaddr *ifp;
1493 unsigned int hash = inet6_addr_hash(addr);
1495 rcu_read_lock_bh();
1496 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1497 if (!net_eq(dev_net(ifp->idev->dev), net))
1498 continue;
1499 if (ipv6_addr_equal(&ifp->addr, addr) &&
1500 !(ifp->flags&IFA_F_TENTATIVE) &&
1501 (dev == NULL || ifp->idev->dev == dev ||
1502 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1503 rcu_read_unlock_bh();
1504 return 1;
1508 rcu_read_unlock_bh();
1509 return 0;
1511 EXPORT_SYMBOL(ipv6_chk_addr);
1513 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1514 struct net_device *dev)
1516 unsigned int hash = inet6_addr_hash(addr);
1517 struct inet6_ifaddr *ifp;
1519 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1520 if (!net_eq(dev_net(ifp->idev->dev), net))
1521 continue;
1522 if (ipv6_addr_equal(&ifp->addr, addr)) {
1523 if (dev == NULL || ifp->idev->dev == dev)
1524 return true;
1527 return false;
1530 /* Compares an address/prefix_len with addresses on device @dev.
1531 * If one is found it returns true.
1533 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1534 const unsigned int prefix_len, struct net_device *dev)
1536 struct inet6_dev *idev;
1537 struct inet6_ifaddr *ifa;
1538 bool ret = false;
1540 rcu_read_lock();
1541 idev = __in6_dev_get(dev);
1542 if (idev) {
1543 read_lock_bh(&idev->lock);
1544 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1545 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1546 if (ret)
1547 break;
1549 read_unlock_bh(&idev->lock);
1551 rcu_read_unlock();
1553 return ret;
1555 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1557 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1559 struct inet6_dev *idev;
1560 struct inet6_ifaddr *ifa;
1561 int onlink;
1563 onlink = 0;
1564 rcu_read_lock();
1565 idev = __in6_dev_get(dev);
1566 if (idev) {
1567 read_lock_bh(&idev->lock);
1568 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1569 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1570 ifa->prefix_len);
1571 if (onlink)
1572 break;
1574 read_unlock_bh(&idev->lock);
1576 rcu_read_unlock();
1577 return onlink;
1579 EXPORT_SYMBOL(ipv6_chk_prefix);
1581 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1582 struct net_device *dev, int strict)
1584 struct inet6_ifaddr *ifp, *result = NULL;
1585 unsigned int hash = inet6_addr_hash(addr);
1587 rcu_read_lock_bh();
1588 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1589 if (!net_eq(dev_net(ifp->idev->dev), net))
1590 continue;
1591 if (ipv6_addr_equal(&ifp->addr, addr)) {
1592 if (dev == NULL || ifp->idev->dev == dev ||
1593 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1594 result = ifp;
1595 in6_ifa_hold(ifp);
1596 break;
1600 rcu_read_unlock_bh();
1602 return result;
1605 /* Gets referenced address, destroys ifaddr */
1607 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1609 if (ifp->flags&IFA_F_PERMANENT) {
1610 spin_lock_bh(&ifp->lock);
1611 addrconf_del_dad_work(ifp);
1612 ifp->flags |= IFA_F_TENTATIVE;
1613 if (dad_failed)
1614 ifp->flags |= IFA_F_DADFAILED;
1615 spin_unlock_bh(&ifp->lock);
1616 if (dad_failed)
1617 ipv6_ifa_notify(0, ifp);
1618 in6_ifa_put(ifp);
1619 } else if (ifp->flags&IFA_F_TEMPORARY) {
1620 struct inet6_ifaddr *ifpub;
1621 spin_lock_bh(&ifp->lock);
1622 ifpub = ifp->ifpub;
1623 if (ifpub) {
1624 in6_ifa_hold(ifpub);
1625 spin_unlock_bh(&ifp->lock);
1626 ipv6_create_tempaddr(ifpub, ifp);
1627 in6_ifa_put(ifpub);
1628 } else {
1629 spin_unlock_bh(&ifp->lock);
1631 ipv6_del_addr(ifp);
1632 } else {
1633 ipv6_del_addr(ifp);
1637 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1639 int err = -ENOENT;
1641 spin_lock_bh(&ifp->state_lock);
1642 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1643 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1644 err = 0;
1646 spin_unlock_bh(&ifp->state_lock);
1648 return err;
1651 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1653 struct inet6_dev *idev = ifp->idev;
1655 if (addrconf_dad_end(ifp)) {
1656 in6_ifa_put(ifp);
1657 return;
1660 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1661 ifp->idev->dev->name, &ifp->addr);
1663 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1664 struct in6_addr addr;
1666 addr.s6_addr32[0] = htonl(0xfe800000);
1667 addr.s6_addr32[1] = 0;
1669 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1670 ipv6_addr_equal(&ifp->addr, &addr)) {
1671 /* DAD failed for link-local based on MAC address */
1672 idev->cnf.disable_ipv6 = 1;
1674 pr_info("%s: IPv6 being disabled!\n",
1675 ifp->idev->dev->name);
1679 spin_lock_bh(&ifp->state_lock);
1680 /* transition from _POSTDAD to _ERRDAD */
1681 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1682 spin_unlock_bh(&ifp->state_lock);
1684 addrconf_mod_dad_work(ifp, 0);
1687 /* Join to solicited addr multicast group. */
1689 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1691 struct in6_addr maddr;
1693 ASSERT_RTNL();
1695 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1696 return;
1698 addrconf_addr_solict_mult(addr, &maddr);
1699 ipv6_dev_mc_inc(dev, &maddr);
1702 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1704 struct in6_addr maddr;
1706 ASSERT_RTNL();
1708 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1709 return;
1711 addrconf_addr_solict_mult(addr, &maddr);
1712 __ipv6_dev_mc_dec(idev, &maddr);
1715 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1717 struct in6_addr addr;
1719 ASSERT_RTNL();
1721 if (ifp->prefix_len >= 127) /* RFC 6164 */
1722 return;
1723 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1724 if (ipv6_addr_any(&addr))
1725 return;
1726 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1729 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1731 struct in6_addr addr;
1733 ASSERT_RTNL();
1735 if (ifp->prefix_len >= 127) /* RFC 6164 */
1736 return;
1737 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1738 if (ipv6_addr_any(&addr))
1739 return;
1740 __ipv6_dev_ac_dec(ifp->idev, &addr);
1743 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1745 if (dev->addr_len != ETH_ALEN)
1746 return -1;
1747 memcpy(eui, dev->dev_addr, 3);
1748 memcpy(eui + 5, dev->dev_addr + 3, 3);
1751 * The zSeries OSA network cards can be shared among various
1752 * OS instances, but the OSA cards have only one MAC address.
1753 * This leads to duplicate address conflicts in conjunction
1754 * with IPv6 if more than one instance uses the same card.
1756 * The driver for these cards can deliver a unique 16-bit
1757 * identifier for each instance sharing the same card. It is
1758 * placed instead of 0xFFFE in the interface identifier. The
1759 * "u" bit of the interface identifier is not inverted in this
1760 * case. Hence the resulting interface identifier has local
1761 * scope according to RFC2373.
1763 if (dev->dev_id) {
1764 eui[3] = (dev->dev_id >> 8) & 0xFF;
1765 eui[4] = dev->dev_id & 0xFF;
1766 } else {
1767 eui[3] = 0xFF;
1768 eui[4] = 0xFE;
1769 eui[0] ^= 2;
1771 return 0;
1774 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1776 if (dev->addr_len != IEEE802154_ADDR_LEN)
1777 return -1;
1778 memcpy(eui, dev->dev_addr, 8);
1779 eui[0] ^= 2;
1780 return 0;
1783 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1785 union fwnet_hwaddr *ha;
1787 if (dev->addr_len != FWNET_ALEN)
1788 return -1;
1790 ha = (union fwnet_hwaddr *)dev->dev_addr;
1792 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1793 eui[0] ^= 2;
1794 return 0;
1797 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1799 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1800 if (dev->addr_len != ARCNET_ALEN)
1801 return -1;
1802 memset(eui, 0, 7);
1803 eui[7] = *(u8 *)dev->dev_addr;
1804 return 0;
1807 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1809 if (dev->addr_len != INFINIBAND_ALEN)
1810 return -1;
1811 memcpy(eui, dev->dev_addr + 12, 8);
1812 eui[0] |= 2;
1813 return 0;
1816 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1818 if (addr == 0)
1819 return -1;
1820 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1821 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1822 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1823 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1824 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1825 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1826 eui[1] = 0;
1827 eui[2] = 0x5E;
1828 eui[3] = 0xFE;
1829 memcpy(eui + 4, &addr, 4);
1830 return 0;
1833 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1835 if (dev->priv_flags & IFF_ISATAP)
1836 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1837 return -1;
1840 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1842 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1845 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1847 memcpy(eui, dev->perm_addr, 3);
1848 memcpy(eui + 5, dev->perm_addr + 3, 3);
1849 eui[3] = 0xFF;
1850 eui[4] = 0xFE;
1851 eui[0] ^= 2;
1852 return 0;
1855 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1857 switch (dev->type) {
1858 case ARPHRD_ETHER:
1859 case ARPHRD_FDDI:
1860 return addrconf_ifid_eui48(eui, dev);
1861 case ARPHRD_ARCNET:
1862 return addrconf_ifid_arcnet(eui, dev);
1863 case ARPHRD_INFINIBAND:
1864 return addrconf_ifid_infiniband(eui, dev);
1865 case ARPHRD_SIT:
1866 return addrconf_ifid_sit(eui, dev);
1867 case ARPHRD_IPGRE:
1868 return addrconf_ifid_gre(eui, dev);
1869 case ARPHRD_6LOWPAN:
1870 case ARPHRD_IEEE802154:
1871 return addrconf_ifid_eui64(eui, dev);
1872 case ARPHRD_IEEE1394:
1873 return addrconf_ifid_ieee1394(eui, dev);
1874 case ARPHRD_TUNNEL6:
1875 return addrconf_ifid_ip6tnl(eui, dev);
1877 return -1;
1880 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1882 int err = -1;
1883 struct inet6_ifaddr *ifp;
1885 read_lock_bh(&idev->lock);
1886 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1887 if (ifp->scope > IFA_LINK)
1888 break;
1889 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1890 memcpy(eui, ifp->addr.s6_addr+8, 8);
1891 err = 0;
1892 break;
1895 read_unlock_bh(&idev->lock);
1896 return err;
1899 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1900 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1902 regen:
1903 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1904 idev->rndid[0] &= ~0x02;
1907 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1908 * check if generated address is not inappropriate
1910 * - Reserved subnet anycast (RFC 2526)
1911 * 11111101 11....11 1xxxxxxx
1912 * - ISATAP (RFC4214) 6.1
1913 * 00-00-5E-FE-xx-xx-xx-xx
1914 * - value 0
1915 * - XXX: already assigned to an address on the device
1917 if (idev->rndid[0] == 0xfd &&
1918 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1919 (idev->rndid[7]&0x80))
1920 goto regen;
1921 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1922 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1923 goto regen;
1924 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1925 goto regen;
1929 static void ipv6_regen_rndid(unsigned long data)
1931 struct inet6_dev *idev = (struct inet6_dev *) data;
1932 unsigned long expires;
1934 rcu_read_lock_bh();
1935 write_lock_bh(&idev->lock);
1937 if (idev->dead)
1938 goto out;
1940 __ipv6_regen_rndid(idev);
1942 expires = jiffies +
1943 idev->cnf.temp_prefered_lft * HZ -
1944 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1945 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1946 idev->cnf.max_desync_factor * HZ;
1947 if (time_before(expires, jiffies)) {
1948 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1949 __func__, idev->dev->name);
1950 goto out;
1953 if (!mod_timer(&idev->regen_timer, expires))
1954 in6_dev_hold(idev);
1956 out:
1957 write_unlock_bh(&idev->lock);
1958 rcu_read_unlock_bh();
1959 in6_dev_put(idev);
1962 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1964 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1965 __ipv6_regen_rndid(idev);
1969 * Add prefix route.
1972 static void
1973 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1974 unsigned long expires, u32 flags)
1976 struct fib6_config cfg = {
1977 .fc_table = RT6_TABLE_PREFIX,
1978 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1979 .fc_ifindex = dev->ifindex,
1980 .fc_expires = expires,
1981 .fc_dst_len = plen,
1982 .fc_flags = RTF_UP | flags,
1983 .fc_nlinfo.nl_net = dev_net(dev),
1984 .fc_protocol = RTPROT_KERNEL,
1987 cfg.fc_dst = *pfx;
1989 /* Prevent useless cloning on PtP SIT.
1990 This thing is done here expecting that the whole
1991 class of non-broadcast devices need not cloning.
1993 #if IS_ENABLED(CONFIG_IPV6_SIT)
1994 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1995 cfg.fc_flags |= RTF_NONEXTHOP;
1996 #endif
1998 ip6_route_add(&cfg);
2002 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2003 int plen,
2004 const struct net_device *dev,
2005 u32 flags, u32 noflags)
2007 struct fib6_node *fn;
2008 struct rt6_info *rt = NULL;
2009 struct fib6_table *table;
2011 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2012 if (table == NULL)
2013 return NULL;
2015 read_lock_bh(&table->tb6_lock);
2016 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2017 if (!fn)
2018 goto out;
2019 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2020 if (rt->dst.dev->ifindex != dev->ifindex)
2021 continue;
2022 if ((rt->rt6i_flags & flags) != flags)
2023 continue;
2024 if ((rt->rt6i_flags & noflags) != 0)
2025 continue;
2026 dst_hold(&rt->dst);
2027 break;
2029 out:
2030 read_unlock_bh(&table->tb6_lock);
2031 return rt;
2035 /* Create "default" multicast route to the interface */
2037 static void addrconf_add_mroute(struct net_device *dev)
2039 struct fib6_config cfg = {
2040 .fc_table = RT6_TABLE_LOCAL,
2041 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2042 .fc_ifindex = dev->ifindex,
2043 .fc_dst_len = 8,
2044 .fc_flags = RTF_UP,
2045 .fc_nlinfo.nl_net = dev_net(dev),
2048 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2050 ip6_route_add(&cfg);
2053 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2055 struct inet6_dev *idev;
2057 ASSERT_RTNL();
2059 idev = ipv6_find_idev(dev);
2060 if (!idev)
2061 return ERR_PTR(-ENOBUFS);
2063 if (idev->cnf.disable_ipv6)
2064 return ERR_PTR(-EACCES);
2066 /* Add default multicast route */
2067 if (!(dev->flags & IFF_LOOPBACK))
2068 addrconf_add_mroute(dev);
2070 return idev;
2073 static void manage_tempaddrs(struct inet6_dev *idev,
2074 struct inet6_ifaddr *ifp,
2075 __u32 valid_lft, __u32 prefered_lft,
2076 bool create, unsigned long now)
2078 u32 flags;
2079 struct inet6_ifaddr *ift;
2081 read_lock_bh(&idev->lock);
2082 /* update all temporary addresses in the list */
2083 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2084 int age, max_valid, max_prefered;
2086 if (ifp != ift->ifpub)
2087 continue;
2089 /* RFC 4941 section 3.3:
2090 * If a received option will extend the lifetime of a public
2091 * address, the lifetimes of temporary addresses should
2092 * be extended, subject to the overall constraint that no
2093 * temporary addresses should ever remain "valid" or "preferred"
2094 * for a time longer than (TEMP_VALID_LIFETIME) or
2095 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2097 age = (now - ift->cstamp) / HZ;
2098 max_valid = idev->cnf.temp_valid_lft - age;
2099 if (max_valid < 0)
2100 max_valid = 0;
2102 max_prefered = idev->cnf.temp_prefered_lft -
2103 idev->cnf.max_desync_factor - age;
2104 if (max_prefered < 0)
2105 max_prefered = 0;
2107 if (valid_lft > max_valid)
2108 valid_lft = max_valid;
2110 if (prefered_lft > max_prefered)
2111 prefered_lft = max_prefered;
2113 spin_lock(&ift->lock);
2114 flags = ift->flags;
2115 ift->valid_lft = valid_lft;
2116 ift->prefered_lft = prefered_lft;
2117 ift->tstamp = now;
2118 if (prefered_lft > 0)
2119 ift->flags &= ~IFA_F_DEPRECATED;
2121 spin_unlock(&ift->lock);
2122 if (!(flags&IFA_F_TENTATIVE))
2123 ipv6_ifa_notify(0, ift);
2126 if ((create || list_empty(&idev->tempaddr_list)) &&
2127 idev->cnf.use_tempaddr > 0) {
2128 /* When a new public address is created as described
2129 * in [ADDRCONF], also create a new temporary address.
2130 * Also create a temporary address if it's enabled but
2131 * no temporary address currently exists.
2133 read_unlock_bh(&idev->lock);
2134 ipv6_create_tempaddr(ifp, NULL);
2135 } else {
2136 read_unlock_bh(&idev->lock);
2140 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2142 struct prefix_info *pinfo;
2143 __u32 valid_lft;
2144 __u32 prefered_lft;
2145 int addr_type;
2146 struct inet6_dev *in6_dev;
2147 struct net *net = dev_net(dev);
2149 pinfo = (struct prefix_info *) opt;
2151 if (len < sizeof(struct prefix_info)) {
2152 ADBG("addrconf: prefix option too short\n");
2153 return;
2157 * Validation checks ([ADDRCONF], page 19)
2160 addr_type = ipv6_addr_type(&pinfo->prefix);
2162 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2163 return;
2165 valid_lft = ntohl(pinfo->valid);
2166 prefered_lft = ntohl(pinfo->prefered);
2168 if (prefered_lft > valid_lft) {
2169 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2170 return;
2173 in6_dev = in6_dev_get(dev);
2175 if (in6_dev == NULL) {
2176 net_dbg_ratelimited("addrconf: device %s not configured\n",
2177 dev->name);
2178 return;
2182 * Two things going on here:
2183 * 1) Add routes for on-link prefixes
2184 * 2) Configure prefixes with the auto flag set
2187 if (pinfo->onlink) {
2188 struct rt6_info *rt;
2189 unsigned long rt_expires;
2191 /* Avoid arithmetic overflow. Really, we could
2192 * save rt_expires in seconds, likely valid_lft,
2193 * but it would require division in fib gc, that it
2194 * not good.
2196 if (HZ > USER_HZ)
2197 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2198 else
2199 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2201 if (addrconf_finite_timeout(rt_expires))
2202 rt_expires *= HZ;
2204 rt = addrconf_get_prefix_route(&pinfo->prefix,
2205 pinfo->prefix_len,
2206 dev,
2207 RTF_ADDRCONF | RTF_PREFIX_RT,
2208 RTF_GATEWAY | RTF_DEFAULT);
2210 if (rt) {
2211 /* Autoconf prefix route */
2212 if (valid_lft == 0) {
2213 ip6_del_rt(rt);
2214 rt = NULL;
2215 } else if (addrconf_finite_timeout(rt_expires)) {
2216 /* not infinity */
2217 rt6_set_expires(rt, jiffies + rt_expires);
2218 } else {
2219 rt6_clean_expires(rt);
2221 } else if (valid_lft) {
2222 clock_t expires = 0;
2223 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2224 if (addrconf_finite_timeout(rt_expires)) {
2225 /* not infinity */
2226 flags |= RTF_EXPIRES;
2227 expires = jiffies_to_clock_t(rt_expires);
2229 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2230 dev, expires, flags);
2232 ip6_rt_put(rt);
2235 /* Try to figure out our local address for this prefix */
2237 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2238 struct inet6_ifaddr *ifp;
2239 struct in6_addr addr;
2240 int create = 0, update_lft = 0;
2241 bool tokenized = false;
2243 if (pinfo->prefix_len == 64) {
2244 memcpy(&addr, &pinfo->prefix, 8);
2246 if (!ipv6_addr_any(&in6_dev->token)) {
2247 read_lock_bh(&in6_dev->lock);
2248 memcpy(addr.s6_addr + 8,
2249 in6_dev->token.s6_addr + 8, 8);
2250 read_unlock_bh(&in6_dev->lock);
2251 tokenized = true;
2252 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2253 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2254 in6_dev_put(in6_dev);
2255 return;
2257 goto ok;
2259 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2260 pinfo->prefix_len);
2261 in6_dev_put(in6_dev);
2262 return;
2266 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2268 if (ifp == NULL && valid_lft) {
2269 int max_addresses = in6_dev->cnf.max_addresses;
2270 u32 addr_flags = 0;
2272 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2273 if (in6_dev->cnf.optimistic_dad &&
2274 !net->ipv6.devconf_all->forwarding && sllao)
2275 addr_flags = IFA_F_OPTIMISTIC;
2276 #endif
2278 /* Do not allow to create too much of autoconfigured
2279 * addresses; this would be too easy way to crash kernel.
2281 if (!max_addresses ||
2282 ipv6_count_addresses(in6_dev) < max_addresses)
2283 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2284 pinfo->prefix_len,
2285 addr_type&IPV6_ADDR_SCOPE_MASK,
2286 addr_flags, valid_lft,
2287 prefered_lft);
2289 if (IS_ERR_OR_NULL(ifp)) {
2290 in6_dev_put(in6_dev);
2291 return;
2294 update_lft = 0;
2295 create = 1;
2296 spin_lock_bh(&ifp->lock);
2297 ifp->flags |= IFA_F_MANAGETEMPADDR;
2298 ifp->cstamp = jiffies;
2299 ifp->tokenized = tokenized;
2300 spin_unlock_bh(&ifp->lock);
2301 addrconf_dad_start(ifp);
2304 if (ifp) {
2305 u32 flags;
2306 unsigned long now;
2307 u32 stored_lft;
2309 /* update lifetime (RFC2462 5.5.3 e) */
2310 spin_lock(&ifp->lock);
2311 now = jiffies;
2312 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2313 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2314 else
2315 stored_lft = 0;
2316 if (!update_lft && !create && stored_lft) {
2317 const u32 minimum_lft = min(
2318 stored_lft, (u32)MIN_VALID_LIFETIME);
2319 valid_lft = max(valid_lft, minimum_lft);
2321 /* RFC4862 Section 5.5.3e:
2322 * "Note that the preferred lifetime of the
2323 * corresponding address is always reset to
2324 * the Preferred Lifetime in the received
2325 * Prefix Information option, regardless of
2326 * whether the valid lifetime is also reset or
2327 * ignored."
2329 * So we should always update prefered_lft here.
2331 update_lft = 1;
2334 if (update_lft) {
2335 ifp->valid_lft = valid_lft;
2336 ifp->prefered_lft = prefered_lft;
2337 ifp->tstamp = now;
2338 flags = ifp->flags;
2339 ifp->flags &= ~IFA_F_DEPRECATED;
2340 spin_unlock(&ifp->lock);
2342 if (!(flags&IFA_F_TENTATIVE))
2343 ipv6_ifa_notify(0, ifp);
2344 } else
2345 spin_unlock(&ifp->lock);
2347 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2348 create, now);
2350 in6_ifa_put(ifp);
2351 addrconf_verify();
2354 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2355 in6_dev_put(in6_dev);
2359 * Set destination address.
2360 * Special case for SIT interfaces where we create a new "virtual"
2361 * device.
2363 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2365 struct in6_ifreq ireq;
2366 struct net_device *dev;
2367 int err = -EINVAL;
2369 rtnl_lock();
2371 err = -EFAULT;
2372 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2373 goto err_exit;
2375 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2377 err = -ENODEV;
2378 if (dev == NULL)
2379 goto err_exit;
2381 #if IS_ENABLED(CONFIG_IPV6_SIT)
2382 if (dev->type == ARPHRD_SIT) {
2383 const struct net_device_ops *ops = dev->netdev_ops;
2384 struct ifreq ifr;
2385 struct ip_tunnel_parm p;
2387 err = -EADDRNOTAVAIL;
2388 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2389 goto err_exit;
2391 memset(&p, 0, sizeof(p));
2392 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2393 p.iph.saddr = 0;
2394 p.iph.version = 4;
2395 p.iph.ihl = 5;
2396 p.iph.protocol = IPPROTO_IPV6;
2397 p.iph.ttl = 64;
2398 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2400 if (ops->ndo_do_ioctl) {
2401 mm_segment_t oldfs = get_fs();
2403 set_fs(KERNEL_DS);
2404 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2405 set_fs(oldfs);
2406 } else
2407 err = -EOPNOTSUPP;
2409 if (err == 0) {
2410 err = -ENOBUFS;
2411 dev = __dev_get_by_name(net, p.name);
2412 if (!dev)
2413 goto err_exit;
2414 err = dev_open(dev);
2417 #endif
2419 err_exit:
2420 rtnl_unlock();
2421 return err;
2425 * Manual configuration of address on an interface
2427 static int inet6_addr_add(struct net *net, int ifindex,
2428 const struct in6_addr *pfx,
2429 const struct in6_addr *peer_pfx,
2430 unsigned int plen, __u32 ifa_flags,
2431 __u32 prefered_lft, __u32 valid_lft)
2433 struct inet6_ifaddr *ifp;
2434 struct inet6_dev *idev;
2435 struct net_device *dev;
2436 int scope;
2437 u32 flags;
2438 clock_t expires;
2439 unsigned long timeout;
2441 ASSERT_RTNL();
2443 if (plen > 128)
2444 return -EINVAL;
2446 /* check the lifetime */
2447 if (!valid_lft || prefered_lft > valid_lft)
2448 return -EINVAL;
2450 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2451 return -EINVAL;
2453 dev = __dev_get_by_index(net, ifindex);
2454 if (!dev)
2455 return -ENODEV;
2457 idev = addrconf_add_dev(dev);
2458 if (IS_ERR(idev))
2459 return PTR_ERR(idev);
2461 scope = ipv6_addr_scope(pfx);
2463 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2464 if (addrconf_finite_timeout(timeout)) {
2465 expires = jiffies_to_clock_t(timeout * HZ);
2466 valid_lft = timeout;
2467 flags = RTF_EXPIRES;
2468 } else {
2469 expires = 0;
2470 flags = 0;
2471 ifa_flags |= IFA_F_PERMANENT;
2474 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2475 if (addrconf_finite_timeout(timeout)) {
2476 if (timeout == 0)
2477 ifa_flags |= IFA_F_DEPRECATED;
2478 prefered_lft = timeout;
2481 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2482 valid_lft, prefered_lft);
2484 if (!IS_ERR(ifp)) {
2485 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2486 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2487 expires, flags);
2491 * Note that section 3.1 of RFC 4429 indicates
2492 * that the Optimistic flag should not be set for
2493 * manually configured addresses
2495 addrconf_dad_start(ifp);
2496 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2497 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2498 true, jiffies);
2499 in6_ifa_put(ifp);
2500 addrconf_verify_rtnl();
2501 return 0;
2504 return PTR_ERR(ifp);
2507 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2508 unsigned int plen)
2510 struct inet6_ifaddr *ifp;
2511 struct inet6_dev *idev;
2512 struct net_device *dev;
2514 if (plen > 128)
2515 return -EINVAL;
2517 dev = __dev_get_by_index(net, ifindex);
2518 if (!dev)
2519 return -ENODEV;
2521 if ((idev = __in6_dev_get(dev)) == NULL)
2522 return -ENXIO;
2524 read_lock_bh(&idev->lock);
2525 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2526 if (ifp->prefix_len == plen &&
2527 ipv6_addr_equal(pfx, &ifp->addr)) {
2528 in6_ifa_hold(ifp);
2529 read_unlock_bh(&idev->lock);
2531 ipv6_del_addr(ifp);
2532 return 0;
2535 read_unlock_bh(&idev->lock);
2536 return -EADDRNOTAVAIL;
2540 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2542 struct in6_ifreq ireq;
2543 int err;
2545 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2546 return -EPERM;
2548 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2549 return -EFAULT;
2551 rtnl_lock();
2552 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2553 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2554 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2555 rtnl_unlock();
2556 return err;
2559 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2561 struct in6_ifreq ireq;
2562 int err;
2564 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2565 return -EPERM;
2567 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2568 return -EFAULT;
2570 rtnl_lock();
2571 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2572 ireq.ifr6_prefixlen);
2573 rtnl_unlock();
2574 return err;
2577 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2578 int plen, int scope)
2580 struct inet6_ifaddr *ifp;
2582 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2583 scope, IFA_F_PERMANENT,
2584 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2585 if (!IS_ERR(ifp)) {
2586 spin_lock_bh(&ifp->lock);
2587 ifp->flags &= ~IFA_F_TENTATIVE;
2588 spin_unlock_bh(&ifp->lock);
2589 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2590 in6_ifa_put(ifp);
2594 #if IS_ENABLED(CONFIG_IPV6_SIT)
2595 static void sit_add_v4_addrs(struct inet6_dev *idev)
2597 struct in6_addr addr;
2598 struct net_device *dev;
2599 struct net *net = dev_net(idev->dev);
2600 int scope, plen;
2601 u32 pflags = 0;
2603 ASSERT_RTNL();
2605 memset(&addr, 0, sizeof(struct in6_addr));
2606 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2608 if (idev->dev->flags&IFF_POINTOPOINT) {
2609 addr.s6_addr32[0] = htonl(0xfe800000);
2610 scope = IFA_LINK;
2611 plen = 64;
2612 } else {
2613 scope = IPV6_ADDR_COMPATv4;
2614 plen = 96;
2615 pflags |= RTF_NONEXTHOP;
2618 if (addr.s6_addr32[3]) {
2619 add_addr(idev, &addr, plen, scope);
2620 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2621 return;
2624 for_each_netdev(net, dev) {
2625 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2626 if (in_dev && (dev->flags & IFF_UP)) {
2627 struct in_ifaddr *ifa;
2629 int flag = scope;
2631 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2633 addr.s6_addr32[3] = ifa->ifa_local;
2635 if (ifa->ifa_scope == RT_SCOPE_LINK)
2636 continue;
2637 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2638 if (idev->dev->flags&IFF_POINTOPOINT)
2639 continue;
2640 flag |= IFA_HOST;
2643 add_addr(idev, &addr, plen, flag);
2644 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2645 pflags);
2650 #endif
2652 static void init_loopback(struct net_device *dev)
2654 struct inet6_dev *idev;
2655 struct net_device *sp_dev;
2656 struct inet6_ifaddr *sp_ifa;
2657 struct rt6_info *sp_rt;
2659 /* ::1 */
2661 ASSERT_RTNL();
2663 if ((idev = ipv6_find_idev(dev)) == NULL) {
2664 pr_debug("%s: add_dev failed\n", __func__);
2665 return;
2668 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2670 /* Add routes to other interface's IPv6 addresses */
2671 for_each_netdev(dev_net(dev), sp_dev) {
2672 if (!strcmp(sp_dev->name, dev->name))
2673 continue;
2675 idev = __in6_dev_get(sp_dev);
2676 if (!idev)
2677 continue;
2679 read_lock_bh(&idev->lock);
2680 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2682 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2683 continue;
2685 if (sp_ifa->rt) {
2686 /* This dst has been added to garbage list when
2687 * lo device down, release this obsolete dst and
2688 * reallocate a new router for ifa.
2690 if (sp_ifa->rt->dst.obsolete > 0) {
2691 ip6_rt_put(sp_ifa->rt);
2692 sp_ifa->rt = NULL;
2693 } else {
2694 continue;
2698 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2700 /* Failure cases are ignored */
2701 if (!IS_ERR(sp_rt)) {
2702 sp_ifa->rt = sp_rt;
2703 ip6_ins_rt(sp_rt);
2706 read_unlock_bh(&idev->lock);
2710 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2712 struct inet6_ifaddr *ifp;
2713 u32 addr_flags = IFA_F_PERMANENT;
2715 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2716 if (idev->cnf.optimistic_dad &&
2717 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2718 addr_flags |= IFA_F_OPTIMISTIC;
2719 #endif
2722 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2723 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2724 if (!IS_ERR(ifp)) {
2725 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2726 addrconf_dad_start(ifp);
2727 in6_ifa_put(ifp);
2731 static void addrconf_dev_config(struct net_device *dev)
2733 struct in6_addr addr;
2734 struct inet6_dev *idev;
2736 ASSERT_RTNL();
2738 if ((dev->type != ARPHRD_ETHER) &&
2739 (dev->type != ARPHRD_FDDI) &&
2740 (dev->type != ARPHRD_ARCNET) &&
2741 (dev->type != ARPHRD_INFINIBAND) &&
2742 (dev->type != ARPHRD_IEEE802154) &&
2743 (dev->type != ARPHRD_IEEE1394) &&
2744 (dev->type != ARPHRD_TUNNEL6) &&
2745 (dev->type != ARPHRD_6LOWPAN)) {
2746 /* Alas, we support only Ethernet autoconfiguration. */
2747 return;
2750 idev = addrconf_add_dev(dev);
2751 if (IS_ERR(idev))
2752 return;
2754 memset(&addr, 0, sizeof(struct in6_addr));
2755 addr.s6_addr32[0] = htonl(0xFE800000);
2757 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2758 addrconf_add_linklocal(idev, &addr);
2761 #if IS_ENABLED(CONFIG_IPV6_SIT)
2762 static void addrconf_sit_config(struct net_device *dev)
2764 struct inet6_dev *idev;
2766 ASSERT_RTNL();
2769 * Configure the tunnel with one of our IPv4
2770 * addresses... we should configure all of
2771 * our v4 addrs in the tunnel
2774 if ((idev = ipv6_find_idev(dev)) == NULL) {
2775 pr_debug("%s: add_dev failed\n", __func__);
2776 return;
2779 if (dev->priv_flags & IFF_ISATAP) {
2780 struct in6_addr addr;
2782 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2783 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2784 addrconf_add_linklocal(idev, &addr);
2785 return;
2788 sit_add_v4_addrs(idev);
2790 if (dev->flags&IFF_POINTOPOINT)
2791 addrconf_add_mroute(dev);
2793 #endif
2795 #if IS_ENABLED(CONFIG_NET_IPGRE)
2796 static void addrconf_gre_config(struct net_device *dev)
2798 struct inet6_dev *idev;
2799 struct in6_addr addr;
2801 ASSERT_RTNL();
2803 if ((idev = ipv6_find_idev(dev)) == NULL) {
2804 pr_debug("%s: add_dev failed\n", __func__);
2805 return;
2808 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2809 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2810 addrconf_add_linklocal(idev, &addr);
2811 else
2812 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2814 #endif
2816 static inline int
2817 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2819 struct in6_addr lladdr;
2821 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2822 addrconf_add_linklocal(idev, &lladdr);
2823 return 0;
2825 return -1;
2828 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2829 void *ptr)
2831 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2832 struct inet6_dev *idev = __in6_dev_get(dev);
2833 int run_pending = 0;
2834 int err;
2836 switch (event) {
2837 case NETDEV_REGISTER:
2838 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2839 idev = ipv6_add_dev(dev);
2840 if (!idev)
2841 return notifier_from_errno(-ENOMEM);
2843 break;
2845 case NETDEV_UP:
2846 case NETDEV_CHANGE:
2847 if (dev->flags & IFF_SLAVE)
2848 break;
2850 if (event == NETDEV_UP) {
2851 if (!addrconf_qdisc_ok(dev)) {
2852 /* device is not ready yet. */
2853 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2854 dev->name);
2855 break;
2858 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2859 idev = ipv6_add_dev(dev);
2861 if (idev) {
2862 idev->if_flags |= IF_READY;
2863 run_pending = 1;
2865 } else {
2866 if (!addrconf_qdisc_ok(dev)) {
2867 /* device is still not ready. */
2868 break;
2871 if (idev) {
2872 if (idev->if_flags & IF_READY)
2873 /* device is already configured. */
2874 break;
2875 idev->if_flags |= IF_READY;
2878 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2879 dev->name);
2881 run_pending = 1;
2884 switch (dev->type) {
2885 #if IS_ENABLED(CONFIG_IPV6_SIT)
2886 case ARPHRD_SIT:
2887 addrconf_sit_config(dev);
2888 break;
2889 #endif
2890 #if IS_ENABLED(CONFIG_NET_IPGRE)
2891 case ARPHRD_IPGRE:
2892 addrconf_gre_config(dev);
2893 break;
2894 #endif
2895 case ARPHRD_LOOPBACK:
2896 init_loopback(dev);
2897 break;
2899 default:
2900 addrconf_dev_config(dev);
2901 break;
2904 if (idev) {
2905 if (run_pending)
2906 addrconf_dad_run(idev);
2909 * If the MTU changed during the interface down,
2910 * when the interface up, the changed MTU must be
2911 * reflected in the idev as well as routers.
2913 if (idev->cnf.mtu6 != dev->mtu &&
2914 dev->mtu >= IPV6_MIN_MTU) {
2915 rt6_mtu_change(dev, dev->mtu);
2916 idev->cnf.mtu6 = dev->mtu;
2918 idev->tstamp = jiffies;
2919 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2922 * If the changed mtu during down is lower than
2923 * IPV6_MIN_MTU stop IPv6 on this interface.
2925 if (dev->mtu < IPV6_MIN_MTU)
2926 addrconf_ifdown(dev, 1);
2928 break;
2930 case NETDEV_CHANGEMTU:
2931 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2932 rt6_mtu_change(dev, dev->mtu);
2933 idev->cnf.mtu6 = dev->mtu;
2934 break;
2937 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2938 idev = ipv6_add_dev(dev);
2939 if (idev)
2940 break;
2944 * if MTU under IPV6_MIN_MTU.
2945 * Stop IPv6 on this interface.
2948 case NETDEV_DOWN:
2949 case NETDEV_UNREGISTER:
2951 * Remove all addresses from this interface.
2953 addrconf_ifdown(dev, event != NETDEV_DOWN);
2954 break;
2956 case NETDEV_CHANGENAME:
2957 if (idev) {
2958 snmp6_unregister_dev(idev);
2959 addrconf_sysctl_unregister(idev);
2960 addrconf_sysctl_register(idev);
2961 err = snmp6_register_dev(idev);
2962 if (err)
2963 return notifier_from_errno(err);
2965 break;
2967 case NETDEV_PRE_TYPE_CHANGE:
2968 case NETDEV_POST_TYPE_CHANGE:
2969 addrconf_type_change(dev, event);
2970 break;
2973 return NOTIFY_OK;
2977 * addrconf module should be notified of a device going up
2979 static struct notifier_block ipv6_dev_notf = {
2980 .notifier_call = addrconf_notify,
2983 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2985 struct inet6_dev *idev;
2986 ASSERT_RTNL();
2988 idev = __in6_dev_get(dev);
2990 if (event == NETDEV_POST_TYPE_CHANGE)
2991 ipv6_mc_remap(idev);
2992 else if (event == NETDEV_PRE_TYPE_CHANGE)
2993 ipv6_mc_unmap(idev);
2996 static int addrconf_ifdown(struct net_device *dev, int how)
2998 struct net *net = dev_net(dev);
2999 struct inet6_dev *idev;
3000 struct inet6_ifaddr *ifa;
3001 int state, i;
3003 ASSERT_RTNL();
3005 rt6_ifdown(net, dev);
3006 neigh_ifdown(&nd_tbl, dev);
3008 idev = __in6_dev_get(dev);
3009 if (idev == NULL)
3010 return -ENODEV;
3013 * Step 1: remove reference to ipv6 device from parent device.
3014 * Do not dev_put!
3016 if (how) {
3017 idev->dead = 1;
3019 /* protected by rtnl_lock */
3020 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3022 /* Step 1.5: remove snmp6 entry */
3023 snmp6_unregister_dev(idev);
3027 /* Step 2: clear hash table */
3028 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3029 struct hlist_head *h = &inet6_addr_lst[i];
3031 spin_lock_bh(&addrconf_hash_lock);
3032 restart:
3033 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3034 if (ifa->idev == idev) {
3035 hlist_del_init_rcu(&ifa->addr_lst);
3036 addrconf_del_dad_work(ifa);
3037 goto restart;
3040 spin_unlock_bh(&addrconf_hash_lock);
3043 write_lock_bh(&idev->lock);
3045 addrconf_del_rs_timer(idev);
3047 /* Step 2: clear flags for stateless addrconf */
3048 if (!how)
3049 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3051 if (how && del_timer(&idev->regen_timer))
3052 in6_dev_put(idev);
3054 /* Step 3: clear tempaddr list */
3055 while (!list_empty(&idev->tempaddr_list)) {
3056 ifa = list_first_entry(&idev->tempaddr_list,
3057 struct inet6_ifaddr, tmp_list);
3058 list_del(&ifa->tmp_list);
3059 write_unlock_bh(&idev->lock);
3060 spin_lock_bh(&ifa->lock);
3062 if (ifa->ifpub) {
3063 in6_ifa_put(ifa->ifpub);
3064 ifa->ifpub = NULL;
3066 spin_unlock_bh(&ifa->lock);
3067 in6_ifa_put(ifa);
3068 write_lock_bh(&idev->lock);
3071 while (!list_empty(&idev->addr_list)) {
3072 ifa = list_first_entry(&idev->addr_list,
3073 struct inet6_ifaddr, if_list);
3074 addrconf_del_dad_work(ifa);
3076 list_del(&ifa->if_list);
3078 write_unlock_bh(&idev->lock);
3080 spin_lock_bh(&ifa->state_lock);
3081 state = ifa->state;
3082 ifa->state = INET6_IFADDR_STATE_DEAD;
3083 spin_unlock_bh(&ifa->state_lock);
3085 if (state != INET6_IFADDR_STATE_DEAD) {
3086 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3087 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3089 in6_ifa_put(ifa);
3091 write_lock_bh(&idev->lock);
3094 write_unlock_bh(&idev->lock);
3096 /* Step 5: Discard multicast list */
3097 if (how)
3098 ipv6_mc_destroy_dev(idev);
3099 else
3100 ipv6_mc_down(idev);
3102 idev->tstamp = jiffies;
3104 /* Last: Shot the device (if unregistered) */
3105 if (how) {
3106 addrconf_sysctl_unregister(idev);
3107 neigh_parms_release(&nd_tbl, idev->nd_parms);
3108 neigh_ifdown(&nd_tbl, dev);
3109 in6_dev_put(idev);
3111 return 0;
3114 static void addrconf_rs_timer(unsigned long data)
3116 struct inet6_dev *idev = (struct inet6_dev *)data;
3117 struct net_device *dev = idev->dev;
3118 struct in6_addr lladdr;
3120 write_lock(&idev->lock);
3121 if (idev->dead || !(idev->if_flags & IF_READY))
3122 goto out;
3124 if (!ipv6_accept_ra(idev))
3125 goto out;
3127 /* Announcement received after solicitation was sent */
3128 if (idev->if_flags & IF_RA_RCVD)
3129 goto out;
3131 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3132 write_unlock(&idev->lock);
3133 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3134 ndisc_send_rs(dev, &lladdr,
3135 &in6addr_linklocal_allrouters);
3136 else
3137 goto put;
3139 write_lock(&idev->lock);
3140 /* The wait after the last probe can be shorter */
3141 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3142 idev->cnf.rtr_solicits) ?
3143 idev->cnf.rtr_solicit_delay :
3144 idev->cnf.rtr_solicit_interval);
3145 } else {
3147 * Note: we do not support deprecated "all on-link"
3148 * assumption any longer.
3150 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3153 out:
3154 write_unlock(&idev->lock);
3155 put:
3156 in6_dev_put(idev);
3160 * Duplicate Address Detection
3162 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3164 unsigned long rand_num;
3165 struct inet6_dev *idev = ifp->idev;
3167 if (ifp->flags & IFA_F_OPTIMISTIC)
3168 rand_num = 0;
3169 else
3170 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3172 ifp->dad_probes = idev->cnf.dad_transmits;
3173 addrconf_mod_dad_work(ifp, rand_num);
3176 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3178 struct inet6_dev *idev = ifp->idev;
3179 struct net_device *dev = idev->dev;
3181 addrconf_join_solict(dev, &ifp->addr);
3183 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3185 read_lock_bh(&idev->lock);
3186 spin_lock(&ifp->lock);
3187 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3188 goto out;
3190 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3191 idev->cnf.accept_dad < 1 ||
3192 !(ifp->flags&IFA_F_TENTATIVE) ||
3193 ifp->flags & IFA_F_NODAD) {
3194 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3195 spin_unlock(&ifp->lock);
3196 read_unlock_bh(&idev->lock);
3198 addrconf_dad_completed(ifp);
3199 return;
3202 if (!(idev->if_flags & IF_READY)) {
3203 spin_unlock(&ifp->lock);
3204 read_unlock_bh(&idev->lock);
3206 * If the device is not ready:
3207 * - keep it tentative if it is a permanent address.
3208 * - otherwise, kill it.
3210 in6_ifa_hold(ifp);
3211 addrconf_dad_stop(ifp, 0);
3212 return;
3216 * Optimistic nodes can start receiving
3217 * Frames right away
3219 if (ifp->flags & IFA_F_OPTIMISTIC)
3220 ip6_ins_rt(ifp->rt);
3222 addrconf_dad_kick(ifp);
3223 out:
3224 spin_unlock(&ifp->lock);
3225 read_unlock_bh(&idev->lock);
3228 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3230 bool begin_dad = false;
3232 spin_lock_bh(&ifp->state_lock);
3233 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3234 ifp->state = INET6_IFADDR_STATE_PREDAD;
3235 begin_dad = true;
3237 spin_unlock_bh(&ifp->state_lock);
3239 if (begin_dad)
3240 addrconf_mod_dad_work(ifp, 0);
3243 static void addrconf_dad_work(struct work_struct *w)
3245 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3246 struct inet6_ifaddr,
3247 dad_work);
3248 struct inet6_dev *idev = ifp->idev;
3249 struct in6_addr mcaddr;
3251 enum {
3252 DAD_PROCESS,
3253 DAD_BEGIN,
3254 DAD_ABORT,
3255 } action = DAD_PROCESS;
3257 rtnl_lock();
3259 spin_lock_bh(&ifp->state_lock);
3260 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3261 action = DAD_BEGIN;
3262 ifp->state = INET6_IFADDR_STATE_DAD;
3263 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3264 action = DAD_ABORT;
3265 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3267 spin_unlock_bh(&ifp->state_lock);
3269 if (action == DAD_BEGIN) {
3270 addrconf_dad_begin(ifp);
3271 goto out;
3272 } else if (action == DAD_ABORT) {
3273 addrconf_dad_stop(ifp, 1);
3274 goto out;
3277 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3278 goto out;
3280 write_lock_bh(&idev->lock);
3281 if (idev->dead || !(idev->if_flags & IF_READY)) {
3282 write_unlock_bh(&idev->lock);
3283 goto out;
3286 spin_lock(&ifp->lock);
3287 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3288 spin_unlock(&ifp->lock);
3289 write_unlock_bh(&idev->lock);
3290 goto out;
3293 if (ifp->dad_probes == 0) {
3295 * DAD was successful
3298 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3299 spin_unlock(&ifp->lock);
3300 write_unlock_bh(&idev->lock);
3302 addrconf_dad_completed(ifp);
3304 goto out;
3307 ifp->dad_probes--;
3308 addrconf_mod_dad_work(ifp,
3309 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3310 spin_unlock(&ifp->lock);
3311 write_unlock_bh(&idev->lock);
3313 /* send a neighbour solicitation for our addr */
3314 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3315 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3316 out:
3317 in6_ifa_put(ifp);
3318 rtnl_unlock();
3321 /* ifp->idev must be at least read locked */
3322 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3324 struct inet6_ifaddr *ifpiter;
3325 struct inet6_dev *idev = ifp->idev;
3327 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3328 if (ifpiter->scope > IFA_LINK)
3329 break;
3330 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3331 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3332 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3333 IFA_F_PERMANENT)
3334 return false;
3336 return true;
3339 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3341 struct net_device *dev = ifp->idev->dev;
3342 struct in6_addr lladdr;
3343 bool send_rs, send_mld;
3345 addrconf_del_dad_work(ifp);
3348 * Configure the address for reception. Now it is valid.
3351 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3353 /* If added prefix is link local and we are prepared to process
3354 router advertisements, start sending router solicitations.
3357 read_lock_bh(&ifp->idev->lock);
3358 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3359 send_rs = send_mld &&
3360 ipv6_accept_ra(ifp->idev) &&
3361 ifp->idev->cnf.rtr_solicits > 0 &&
3362 (dev->flags&IFF_LOOPBACK) == 0;
3363 read_unlock_bh(&ifp->idev->lock);
3365 /* While dad is in progress mld report's source address is in6_addrany.
3366 * Resend with proper ll now.
3368 if (send_mld)
3369 ipv6_mc_dad_complete(ifp->idev);
3371 if (send_rs) {
3373 * If a host as already performed a random delay
3374 * [...] as part of DAD [...] there is no need
3375 * to delay again before sending the first RS
3377 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3378 return;
3379 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3381 write_lock_bh(&ifp->idev->lock);
3382 spin_lock(&ifp->lock);
3383 ifp->idev->rs_probes = 1;
3384 ifp->idev->if_flags |= IF_RS_SENT;
3385 addrconf_mod_rs_timer(ifp->idev,
3386 ifp->idev->cnf.rtr_solicit_interval);
3387 spin_unlock(&ifp->lock);
3388 write_unlock_bh(&ifp->idev->lock);
3392 static void addrconf_dad_run(struct inet6_dev *idev)
3394 struct inet6_ifaddr *ifp;
3396 read_lock_bh(&idev->lock);
3397 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3398 spin_lock(&ifp->lock);
3399 if (ifp->flags & IFA_F_TENTATIVE &&
3400 ifp->state == INET6_IFADDR_STATE_DAD)
3401 addrconf_dad_kick(ifp);
3402 spin_unlock(&ifp->lock);
3404 read_unlock_bh(&idev->lock);
3407 #ifdef CONFIG_PROC_FS
3408 struct if6_iter_state {
3409 struct seq_net_private p;
3410 int bucket;
3411 int offset;
3414 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3416 struct inet6_ifaddr *ifa = NULL;
3417 struct if6_iter_state *state = seq->private;
3418 struct net *net = seq_file_net(seq);
3419 int p = 0;
3421 /* initial bucket if pos is 0 */
3422 if (pos == 0) {
3423 state->bucket = 0;
3424 state->offset = 0;
3427 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3428 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3429 addr_lst) {
3430 if (!net_eq(dev_net(ifa->idev->dev), net))
3431 continue;
3432 /* sync with offset */
3433 if (p < state->offset) {
3434 p++;
3435 continue;
3437 state->offset++;
3438 return ifa;
3441 /* prepare for next bucket */
3442 state->offset = 0;
3443 p = 0;
3445 return NULL;
3448 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3449 struct inet6_ifaddr *ifa)
3451 struct if6_iter_state *state = seq->private;
3452 struct net *net = seq_file_net(seq);
3454 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3455 if (!net_eq(dev_net(ifa->idev->dev), net))
3456 continue;
3457 state->offset++;
3458 return ifa;
3461 while (++state->bucket < IN6_ADDR_HSIZE) {
3462 state->offset = 0;
3463 hlist_for_each_entry_rcu_bh(ifa,
3464 &inet6_addr_lst[state->bucket], addr_lst) {
3465 if (!net_eq(dev_net(ifa->idev->dev), net))
3466 continue;
3467 state->offset++;
3468 return ifa;
3472 return NULL;
3475 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3476 __acquires(rcu_bh)
3478 rcu_read_lock_bh();
3479 return if6_get_first(seq, *pos);
3482 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3484 struct inet6_ifaddr *ifa;
3486 ifa = if6_get_next(seq, v);
3487 ++*pos;
3488 return ifa;
3491 static void if6_seq_stop(struct seq_file *seq, void *v)
3492 __releases(rcu_bh)
3494 rcu_read_unlock_bh();
3497 static int if6_seq_show(struct seq_file *seq, void *v)
3499 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3500 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3501 &ifp->addr,
3502 ifp->idev->dev->ifindex,
3503 ifp->prefix_len,
3504 ifp->scope,
3505 (u8) ifp->flags,
3506 ifp->idev->dev->name);
3507 return 0;
3510 static const struct seq_operations if6_seq_ops = {
3511 .start = if6_seq_start,
3512 .next = if6_seq_next,
3513 .show = if6_seq_show,
3514 .stop = if6_seq_stop,
3517 static int if6_seq_open(struct inode *inode, struct file *file)
3519 return seq_open_net(inode, file, &if6_seq_ops,
3520 sizeof(struct if6_iter_state));
3523 static const struct file_operations if6_fops = {
3524 .owner = THIS_MODULE,
3525 .open = if6_seq_open,
3526 .read = seq_read,
3527 .llseek = seq_lseek,
3528 .release = seq_release_net,
3531 static int __net_init if6_proc_net_init(struct net *net)
3533 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3534 return -ENOMEM;
3535 return 0;
3538 static void __net_exit if6_proc_net_exit(struct net *net)
3540 remove_proc_entry("if_inet6", net->proc_net);
3543 static struct pernet_operations if6_proc_net_ops = {
3544 .init = if6_proc_net_init,
3545 .exit = if6_proc_net_exit,
3548 int __init if6_proc_init(void)
3550 return register_pernet_subsys(&if6_proc_net_ops);
3553 void if6_proc_exit(void)
3555 unregister_pernet_subsys(&if6_proc_net_ops);
3557 #endif /* CONFIG_PROC_FS */
3559 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3560 /* Check if address is a home address configured on any interface. */
3561 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3563 int ret = 0;
3564 struct inet6_ifaddr *ifp = NULL;
3565 unsigned int hash = inet6_addr_hash(addr);
3567 rcu_read_lock_bh();
3568 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3569 if (!net_eq(dev_net(ifp->idev->dev), net))
3570 continue;
3571 if (ipv6_addr_equal(&ifp->addr, addr) &&
3572 (ifp->flags & IFA_F_HOMEADDRESS)) {
3573 ret = 1;
3574 break;
3577 rcu_read_unlock_bh();
3578 return ret;
3580 #endif
3583 * Periodic address status verification
3586 static void addrconf_verify_rtnl(void)
3588 unsigned long now, next, next_sec, next_sched;
3589 struct inet6_ifaddr *ifp;
3590 int i;
3592 ASSERT_RTNL();
3594 rcu_read_lock_bh();
3595 now = jiffies;
3596 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3598 cancel_delayed_work(&addr_chk_work);
3600 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3601 restart:
3602 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3603 unsigned long age;
3605 /* When setting preferred_lft to a value not zero or
3606 * infinity, while valid_lft is infinity
3607 * IFA_F_PERMANENT has a non-infinity life time.
3609 if ((ifp->flags & IFA_F_PERMANENT) &&
3610 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3611 continue;
3613 spin_lock(&ifp->lock);
3614 /* We try to batch several events at once. */
3615 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3617 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3618 age >= ifp->valid_lft) {
3619 spin_unlock(&ifp->lock);
3620 in6_ifa_hold(ifp);
3621 ipv6_del_addr(ifp);
3622 goto restart;
3623 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3624 spin_unlock(&ifp->lock);
3625 continue;
3626 } else if (age >= ifp->prefered_lft) {
3627 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3628 int deprecate = 0;
3630 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3631 deprecate = 1;
3632 ifp->flags |= IFA_F_DEPRECATED;
3635 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3636 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3637 next = ifp->tstamp + ifp->valid_lft * HZ;
3639 spin_unlock(&ifp->lock);
3641 if (deprecate) {
3642 in6_ifa_hold(ifp);
3644 ipv6_ifa_notify(0, ifp);
3645 in6_ifa_put(ifp);
3646 goto restart;
3648 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3649 !(ifp->flags&IFA_F_TENTATIVE)) {
3650 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3651 ifp->idev->cnf.dad_transmits *
3652 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3654 if (age >= ifp->prefered_lft - regen_advance) {
3655 struct inet6_ifaddr *ifpub = ifp->ifpub;
3656 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3657 next = ifp->tstamp + ifp->prefered_lft * HZ;
3658 if (!ifp->regen_count && ifpub) {
3659 ifp->regen_count++;
3660 in6_ifa_hold(ifp);
3661 in6_ifa_hold(ifpub);
3662 spin_unlock(&ifp->lock);
3664 spin_lock(&ifpub->lock);
3665 ifpub->regen_count = 0;
3666 spin_unlock(&ifpub->lock);
3667 ipv6_create_tempaddr(ifpub, ifp);
3668 in6_ifa_put(ifpub);
3669 in6_ifa_put(ifp);
3670 goto restart;
3672 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3673 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3674 spin_unlock(&ifp->lock);
3675 } else {
3676 /* ifp->prefered_lft <= ifp->valid_lft */
3677 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3678 next = ifp->tstamp + ifp->prefered_lft * HZ;
3679 spin_unlock(&ifp->lock);
3684 next_sec = round_jiffies_up(next);
3685 next_sched = next;
3687 /* If rounded timeout is accurate enough, accept it. */
3688 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3689 next_sched = next_sec;
3691 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3692 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3693 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3695 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3696 now, next, next_sec, next_sched);
3697 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3698 rcu_read_unlock_bh();
3701 static void addrconf_verify_work(struct work_struct *w)
3703 rtnl_lock();
3704 addrconf_verify_rtnl();
3705 rtnl_unlock();
3708 static void addrconf_verify(void)
3710 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3713 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3714 struct in6_addr **peer_pfx)
3716 struct in6_addr *pfx = NULL;
3718 *peer_pfx = NULL;
3720 if (addr)
3721 pfx = nla_data(addr);
3723 if (local) {
3724 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3725 *peer_pfx = pfx;
3726 pfx = nla_data(local);
3729 return pfx;
3732 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3733 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3734 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3735 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3736 [IFA_FLAGS] = { .len = sizeof(u32) },
3739 static int
3740 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3742 struct net *net = sock_net(skb->sk);
3743 struct ifaddrmsg *ifm;
3744 struct nlattr *tb[IFA_MAX+1];
3745 struct in6_addr *pfx, *peer_pfx;
3746 int err;
3748 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3749 if (err < 0)
3750 return err;
3752 ifm = nlmsg_data(nlh);
3753 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3754 if (pfx == NULL)
3755 return -EINVAL;
3757 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3760 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3761 u32 prefered_lft, u32 valid_lft)
3763 u32 flags;
3764 clock_t expires;
3765 unsigned long timeout;
3766 bool was_managetempaddr;
3767 bool had_prefixroute;
3769 ASSERT_RTNL();
3771 if (!valid_lft || (prefered_lft > valid_lft))
3772 return -EINVAL;
3774 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3775 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3776 return -EINVAL;
3778 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3779 if (addrconf_finite_timeout(timeout)) {
3780 expires = jiffies_to_clock_t(timeout * HZ);
3781 valid_lft = timeout;
3782 flags = RTF_EXPIRES;
3783 } else {
3784 expires = 0;
3785 flags = 0;
3786 ifa_flags |= IFA_F_PERMANENT;
3789 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3790 if (addrconf_finite_timeout(timeout)) {
3791 if (timeout == 0)
3792 ifa_flags |= IFA_F_DEPRECATED;
3793 prefered_lft = timeout;
3796 spin_lock_bh(&ifp->lock);
3797 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3798 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3799 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3800 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3801 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3802 IFA_F_NOPREFIXROUTE);
3803 ifp->flags |= ifa_flags;
3804 ifp->tstamp = jiffies;
3805 ifp->valid_lft = valid_lft;
3806 ifp->prefered_lft = prefered_lft;
3808 spin_unlock_bh(&ifp->lock);
3809 if (!(ifp->flags&IFA_F_TENTATIVE))
3810 ipv6_ifa_notify(0, ifp);
3812 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3813 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3814 expires, flags);
3815 } else if (had_prefixroute) {
3816 enum cleanup_prefix_rt_t action;
3817 unsigned long rt_expires;
3819 write_lock_bh(&ifp->idev->lock);
3820 action = check_cleanup_prefix_route(ifp, &rt_expires);
3821 write_unlock_bh(&ifp->idev->lock);
3823 if (action != CLEANUP_PREFIX_RT_NOP) {
3824 cleanup_prefix_route(ifp, rt_expires,
3825 action == CLEANUP_PREFIX_RT_DEL);
3829 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3830 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3831 valid_lft = prefered_lft = 0;
3832 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3833 !was_managetempaddr, jiffies);
3836 addrconf_verify_rtnl();
3838 return 0;
3841 static int
3842 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3844 struct net *net = sock_net(skb->sk);
3845 struct ifaddrmsg *ifm;
3846 struct nlattr *tb[IFA_MAX+1];
3847 struct in6_addr *pfx, *peer_pfx;
3848 struct inet6_ifaddr *ifa;
3849 struct net_device *dev;
3850 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3851 u32 ifa_flags;
3852 int err;
3854 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3855 if (err < 0)
3856 return err;
3858 ifm = nlmsg_data(nlh);
3859 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3860 if (pfx == NULL)
3861 return -EINVAL;
3863 if (tb[IFA_CACHEINFO]) {
3864 struct ifa_cacheinfo *ci;
3866 ci = nla_data(tb[IFA_CACHEINFO]);
3867 valid_lft = ci->ifa_valid;
3868 preferred_lft = ci->ifa_prefered;
3869 } else {
3870 preferred_lft = INFINITY_LIFE_TIME;
3871 valid_lft = INFINITY_LIFE_TIME;
3874 dev = __dev_get_by_index(net, ifm->ifa_index);
3875 if (dev == NULL)
3876 return -ENODEV;
3878 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3880 /* We ignore other flags so far. */
3881 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3882 IFA_F_NOPREFIXROUTE;
3884 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3885 if (ifa == NULL) {
3887 * It would be best to check for !NLM_F_CREATE here but
3888 * userspace already relies on not having to provide this.
3890 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3891 ifm->ifa_prefixlen, ifa_flags,
3892 preferred_lft, valid_lft);
3895 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3896 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3897 err = -EEXIST;
3898 else
3899 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3901 in6_ifa_put(ifa);
3903 return err;
3906 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3907 u8 scope, int ifindex)
3909 struct ifaddrmsg *ifm;
3911 ifm = nlmsg_data(nlh);
3912 ifm->ifa_family = AF_INET6;
3913 ifm->ifa_prefixlen = prefixlen;
3914 ifm->ifa_flags = flags;
3915 ifm->ifa_scope = scope;
3916 ifm->ifa_index = ifindex;
3919 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3920 unsigned long tstamp, u32 preferred, u32 valid)
3922 struct ifa_cacheinfo ci;
3924 ci.cstamp = cstamp_delta(cstamp);
3925 ci.tstamp = cstamp_delta(tstamp);
3926 ci.ifa_prefered = preferred;
3927 ci.ifa_valid = valid;
3929 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3932 static inline int rt_scope(int ifa_scope)
3934 if (ifa_scope & IFA_HOST)
3935 return RT_SCOPE_HOST;
3936 else if (ifa_scope & IFA_LINK)
3937 return RT_SCOPE_LINK;
3938 else if (ifa_scope & IFA_SITE)
3939 return RT_SCOPE_SITE;
3940 else
3941 return RT_SCOPE_UNIVERSE;
3944 static inline int inet6_ifaddr_msgsize(void)
3946 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3947 + nla_total_size(16) /* IFA_LOCAL */
3948 + nla_total_size(16) /* IFA_ADDRESS */
3949 + nla_total_size(sizeof(struct ifa_cacheinfo))
3950 + nla_total_size(4) /* IFA_FLAGS */;
3953 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3954 u32 portid, u32 seq, int event, unsigned int flags)
3956 struct nlmsghdr *nlh;
3957 u32 preferred, valid;
3959 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3960 if (nlh == NULL)
3961 return -EMSGSIZE;
3963 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3964 ifa->idev->dev->ifindex);
3966 if (!((ifa->flags&IFA_F_PERMANENT) &&
3967 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3968 preferred = ifa->prefered_lft;
3969 valid = ifa->valid_lft;
3970 if (preferred != INFINITY_LIFE_TIME) {
3971 long tval = (jiffies - ifa->tstamp)/HZ;
3972 if (preferred > tval)
3973 preferred -= tval;
3974 else
3975 preferred = 0;
3976 if (valid != INFINITY_LIFE_TIME) {
3977 if (valid > tval)
3978 valid -= tval;
3979 else
3980 valid = 0;
3983 } else {
3984 preferred = INFINITY_LIFE_TIME;
3985 valid = INFINITY_LIFE_TIME;
3988 if (!ipv6_addr_any(&ifa->peer_addr)) {
3989 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3990 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3991 goto error;
3992 } else
3993 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3994 goto error;
3996 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3997 goto error;
3999 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4000 goto error;
4002 return nlmsg_end(skb, nlh);
4004 error:
4005 nlmsg_cancel(skb, nlh);
4006 return -EMSGSIZE;
4009 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4010 u32 portid, u32 seq, int event, u16 flags)
4012 struct nlmsghdr *nlh;
4013 u8 scope = RT_SCOPE_UNIVERSE;
4014 int ifindex = ifmca->idev->dev->ifindex;
4016 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4017 scope = RT_SCOPE_SITE;
4019 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4020 if (nlh == NULL)
4021 return -EMSGSIZE;
4023 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4024 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4025 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4026 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4027 nlmsg_cancel(skb, nlh);
4028 return -EMSGSIZE;
4031 return nlmsg_end(skb, nlh);
4034 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4035 u32 portid, u32 seq, int event, unsigned int flags)
4037 struct nlmsghdr *nlh;
4038 u8 scope = RT_SCOPE_UNIVERSE;
4039 int ifindex = ifaca->aca_idev->dev->ifindex;
4041 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4042 scope = RT_SCOPE_SITE;
4044 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4045 if (nlh == NULL)
4046 return -EMSGSIZE;
4048 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4049 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4050 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4051 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4052 nlmsg_cancel(skb, nlh);
4053 return -EMSGSIZE;
4056 return nlmsg_end(skb, nlh);
4059 enum addr_type_t {
4060 UNICAST_ADDR,
4061 MULTICAST_ADDR,
4062 ANYCAST_ADDR,
4065 /* called with rcu_read_lock() */
4066 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4067 struct netlink_callback *cb, enum addr_type_t type,
4068 int s_ip_idx, int *p_ip_idx)
4070 struct ifmcaddr6 *ifmca;
4071 struct ifacaddr6 *ifaca;
4072 int err = 1;
4073 int ip_idx = *p_ip_idx;
4075 read_lock_bh(&idev->lock);
4076 switch (type) {
4077 case UNICAST_ADDR: {
4078 struct inet6_ifaddr *ifa;
4080 /* unicast address incl. temp addr */
4081 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4082 if (++ip_idx < s_ip_idx)
4083 continue;
4084 err = inet6_fill_ifaddr(skb, ifa,
4085 NETLINK_CB(cb->skb).portid,
4086 cb->nlh->nlmsg_seq,
4087 RTM_NEWADDR,
4088 NLM_F_MULTI);
4089 if (err <= 0)
4090 break;
4091 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4093 break;
4095 case MULTICAST_ADDR:
4096 /* multicast address */
4097 for (ifmca = idev->mc_list; ifmca;
4098 ifmca = ifmca->next, ip_idx++) {
4099 if (ip_idx < s_ip_idx)
4100 continue;
4101 err = inet6_fill_ifmcaddr(skb, ifmca,
4102 NETLINK_CB(cb->skb).portid,
4103 cb->nlh->nlmsg_seq,
4104 RTM_GETMULTICAST,
4105 NLM_F_MULTI);
4106 if (err <= 0)
4107 break;
4109 break;
4110 case ANYCAST_ADDR:
4111 /* anycast address */
4112 for (ifaca = idev->ac_list; ifaca;
4113 ifaca = ifaca->aca_next, ip_idx++) {
4114 if (ip_idx < s_ip_idx)
4115 continue;
4116 err = inet6_fill_ifacaddr(skb, ifaca,
4117 NETLINK_CB(cb->skb).portid,
4118 cb->nlh->nlmsg_seq,
4119 RTM_GETANYCAST,
4120 NLM_F_MULTI);
4121 if (err <= 0)
4122 break;
4124 break;
4125 default:
4126 break;
4128 read_unlock_bh(&idev->lock);
4129 *p_ip_idx = ip_idx;
4130 return err;
4133 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4134 enum addr_type_t type)
4136 struct net *net = sock_net(skb->sk);
4137 int h, s_h;
4138 int idx, ip_idx;
4139 int s_idx, s_ip_idx;
4140 struct net_device *dev;
4141 struct inet6_dev *idev;
4142 struct hlist_head *head;
4144 s_h = cb->args[0];
4145 s_idx = idx = cb->args[1];
4146 s_ip_idx = ip_idx = cb->args[2];
4148 rcu_read_lock();
4149 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4150 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4151 idx = 0;
4152 head = &net->dev_index_head[h];
4153 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4154 if (idx < s_idx)
4155 goto cont;
4156 if (h > s_h || idx > s_idx)
4157 s_ip_idx = 0;
4158 ip_idx = 0;
4159 idev = __in6_dev_get(dev);
4160 if (!idev)
4161 goto cont;
4163 if (in6_dump_addrs(idev, skb, cb, type,
4164 s_ip_idx, &ip_idx) <= 0)
4165 goto done;
4166 cont:
4167 idx++;
4170 done:
4171 rcu_read_unlock();
4172 cb->args[0] = h;
4173 cb->args[1] = idx;
4174 cb->args[2] = ip_idx;
4176 return skb->len;
4179 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4181 enum addr_type_t type = UNICAST_ADDR;
4183 return inet6_dump_addr(skb, cb, type);
4186 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4188 enum addr_type_t type = MULTICAST_ADDR;
4190 return inet6_dump_addr(skb, cb, type);
4194 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4196 enum addr_type_t type = ANYCAST_ADDR;
4198 return inet6_dump_addr(skb, cb, type);
4201 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4203 struct net *net = sock_net(in_skb->sk);
4204 struct ifaddrmsg *ifm;
4205 struct nlattr *tb[IFA_MAX+1];
4206 struct in6_addr *addr = NULL, *peer;
4207 struct net_device *dev = NULL;
4208 struct inet6_ifaddr *ifa;
4209 struct sk_buff *skb;
4210 int err;
4212 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4213 if (err < 0)
4214 goto errout;
4216 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4217 if (addr == NULL) {
4218 err = -EINVAL;
4219 goto errout;
4222 ifm = nlmsg_data(nlh);
4223 if (ifm->ifa_index)
4224 dev = __dev_get_by_index(net, ifm->ifa_index);
4226 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4227 if (!ifa) {
4228 err = -EADDRNOTAVAIL;
4229 goto errout;
4232 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4233 if (!skb) {
4234 err = -ENOBUFS;
4235 goto errout_ifa;
4238 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4239 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4240 if (err < 0) {
4241 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4242 WARN_ON(err == -EMSGSIZE);
4243 kfree_skb(skb);
4244 goto errout_ifa;
4246 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4247 errout_ifa:
4248 in6_ifa_put(ifa);
4249 errout:
4250 return err;
4253 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4255 struct sk_buff *skb;
4256 struct net *net = dev_net(ifa->idev->dev);
4257 int err = -ENOBUFS;
4259 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4260 if (skb == NULL)
4261 goto errout;
4263 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4264 if (err < 0) {
4265 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4266 WARN_ON(err == -EMSGSIZE);
4267 kfree_skb(skb);
4268 goto errout;
4270 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4271 return;
4272 errout:
4273 if (err < 0)
4274 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4277 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4278 __s32 *array, int bytes)
4280 BUG_ON(bytes < (DEVCONF_MAX * 4));
4282 memset(array, 0, bytes);
4283 array[DEVCONF_FORWARDING] = cnf->forwarding;
4284 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4285 array[DEVCONF_MTU6] = cnf->mtu6;
4286 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4287 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4288 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4289 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4290 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4291 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4292 jiffies_to_msecs(cnf->rtr_solicit_interval);
4293 array[DEVCONF_RTR_SOLICIT_DELAY] =
4294 jiffies_to_msecs(cnf->rtr_solicit_delay);
4295 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4296 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4297 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4298 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4299 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4300 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4301 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4302 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4303 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4304 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4305 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4306 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4307 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4308 #ifdef CONFIG_IPV6_ROUTER_PREF
4309 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4310 array[DEVCONF_RTR_PROBE_INTERVAL] =
4311 jiffies_to_msecs(cnf->rtr_probe_interval);
4312 #ifdef CONFIG_IPV6_ROUTE_INFO
4313 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4314 #endif
4315 #endif
4316 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4317 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4318 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4319 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4320 #endif
4321 #ifdef CONFIG_IPV6_MROUTE
4322 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4323 #endif
4324 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4325 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4326 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4327 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4328 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4331 static inline size_t inet6_ifla6_size(void)
4333 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4334 + nla_total_size(sizeof(struct ifla_cacheinfo))
4335 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4336 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4337 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4338 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4341 static inline size_t inet6_if_nlmsg_size(void)
4343 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4344 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4345 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4346 + nla_total_size(4) /* IFLA_MTU */
4347 + nla_total_size(4) /* IFLA_LINK */
4348 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4351 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4352 int items, int bytes)
4354 int i;
4355 int pad = bytes - sizeof(u64) * items;
4356 BUG_ON(pad < 0);
4358 /* Use put_unaligned() because stats may not be aligned for u64. */
4359 put_unaligned(items, &stats[0]);
4360 for (i = 1; i < items; i++)
4361 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4363 memset(&stats[items], 0, pad);
4366 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4367 int items, int bytes, size_t syncpoff)
4369 int i;
4370 int pad = bytes - sizeof(u64) * items;
4371 BUG_ON(pad < 0);
4373 /* Use put_unaligned() because stats may not be aligned for u64. */
4374 put_unaligned(items, &stats[0]);
4375 for (i = 1; i < items; i++)
4376 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4378 memset(&stats[items], 0, pad);
4381 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4382 int bytes)
4384 switch (attrtype) {
4385 case IFLA_INET6_STATS:
4386 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4387 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4388 break;
4389 case IFLA_INET6_ICMP6STATS:
4390 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4391 break;
4395 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4397 struct nlattr *nla;
4398 struct ifla_cacheinfo ci;
4400 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4401 goto nla_put_failure;
4402 ci.max_reasm_len = IPV6_MAXPLEN;
4403 ci.tstamp = cstamp_delta(idev->tstamp);
4404 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4405 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4406 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4407 goto nla_put_failure;
4408 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4409 if (nla == NULL)
4410 goto nla_put_failure;
4411 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4413 /* XXX - MC not implemented */
4415 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4416 if (nla == NULL)
4417 goto nla_put_failure;
4418 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4420 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4421 if (nla == NULL)
4422 goto nla_put_failure;
4423 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4425 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4426 if (nla == NULL)
4427 goto nla_put_failure;
4428 read_lock_bh(&idev->lock);
4429 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4430 read_unlock_bh(&idev->lock);
4432 return 0;
4434 nla_put_failure:
4435 return -EMSGSIZE;
4438 static size_t inet6_get_link_af_size(const struct net_device *dev)
4440 if (!__in6_dev_get(dev))
4441 return 0;
4443 return inet6_ifla6_size();
4446 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4448 struct inet6_dev *idev = __in6_dev_get(dev);
4450 if (!idev)
4451 return -ENODATA;
4453 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4454 return -EMSGSIZE;
4456 return 0;
4459 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4461 struct inet6_ifaddr *ifp;
4462 struct net_device *dev = idev->dev;
4463 bool update_rs = false;
4464 struct in6_addr ll_addr;
4466 ASSERT_RTNL();
4468 if (token == NULL)
4469 return -EINVAL;
4470 if (ipv6_addr_any(token))
4471 return -EINVAL;
4472 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4473 return -EINVAL;
4474 if (!ipv6_accept_ra(idev))
4475 return -EINVAL;
4476 if (idev->cnf.rtr_solicits <= 0)
4477 return -EINVAL;
4479 write_lock_bh(&idev->lock);
4481 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4482 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4484 write_unlock_bh(&idev->lock);
4486 if (!idev->dead && (idev->if_flags & IF_READY) &&
4487 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4488 IFA_F_OPTIMISTIC)) {
4490 /* If we're not ready, then normal ifup will take care
4491 * of this. Otherwise, we need to request our rs here.
4493 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4494 update_rs = true;
4497 write_lock_bh(&idev->lock);
4499 if (update_rs) {
4500 idev->if_flags |= IF_RS_SENT;
4501 idev->rs_probes = 1;
4502 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4505 /* Well, that's kinda nasty ... */
4506 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4507 spin_lock(&ifp->lock);
4508 if (ifp->tokenized) {
4509 ifp->valid_lft = 0;
4510 ifp->prefered_lft = 0;
4512 spin_unlock(&ifp->lock);
4515 write_unlock_bh(&idev->lock);
4516 addrconf_verify_rtnl();
4517 return 0;
4520 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4522 int err = -EINVAL;
4523 struct inet6_dev *idev = __in6_dev_get(dev);
4524 struct nlattr *tb[IFLA_INET6_MAX + 1];
4526 if (!idev)
4527 return -EAFNOSUPPORT;
4529 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4530 BUG();
4532 if (tb[IFLA_INET6_TOKEN])
4533 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4535 return err;
4538 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4539 u32 portid, u32 seq, int event, unsigned int flags)
4541 struct net_device *dev = idev->dev;
4542 struct ifinfomsg *hdr;
4543 struct nlmsghdr *nlh;
4544 void *protoinfo;
4546 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4547 if (nlh == NULL)
4548 return -EMSGSIZE;
4550 hdr = nlmsg_data(nlh);
4551 hdr->ifi_family = AF_INET6;
4552 hdr->__ifi_pad = 0;
4553 hdr->ifi_type = dev->type;
4554 hdr->ifi_index = dev->ifindex;
4555 hdr->ifi_flags = dev_get_flags(dev);
4556 hdr->ifi_change = 0;
4558 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4559 (dev->addr_len &&
4560 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4561 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4562 (dev->ifindex != dev->iflink &&
4563 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4564 goto nla_put_failure;
4565 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4566 if (protoinfo == NULL)
4567 goto nla_put_failure;
4569 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4570 goto nla_put_failure;
4572 nla_nest_end(skb, protoinfo);
4573 return nlmsg_end(skb, nlh);
4575 nla_put_failure:
4576 nlmsg_cancel(skb, nlh);
4577 return -EMSGSIZE;
4580 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4582 struct net *net = sock_net(skb->sk);
4583 int h, s_h;
4584 int idx = 0, s_idx;
4585 struct net_device *dev;
4586 struct inet6_dev *idev;
4587 struct hlist_head *head;
4589 s_h = cb->args[0];
4590 s_idx = cb->args[1];
4592 rcu_read_lock();
4593 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4594 idx = 0;
4595 head = &net->dev_index_head[h];
4596 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4597 if (idx < s_idx)
4598 goto cont;
4599 idev = __in6_dev_get(dev);
4600 if (!idev)
4601 goto cont;
4602 if (inet6_fill_ifinfo(skb, idev,
4603 NETLINK_CB(cb->skb).portid,
4604 cb->nlh->nlmsg_seq,
4605 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4606 goto out;
4607 cont:
4608 idx++;
4611 out:
4612 rcu_read_unlock();
4613 cb->args[1] = idx;
4614 cb->args[0] = h;
4616 return skb->len;
4619 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4621 struct sk_buff *skb;
4622 struct net *net = dev_net(idev->dev);
4623 int err = -ENOBUFS;
4625 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4626 if (skb == NULL)
4627 goto errout;
4629 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4630 if (err < 0) {
4631 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4632 WARN_ON(err == -EMSGSIZE);
4633 kfree_skb(skb);
4634 goto errout;
4636 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4637 return;
4638 errout:
4639 if (err < 0)
4640 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4643 static inline size_t inet6_prefix_nlmsg_size(void)
4645 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4646 + nla_total_size(sizeof(struct in6_addr))
4647 + nla_total_size(sizeof(struct prefix_cacheinfo));
4650 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4651 struct prefix_info *pinfo, u32 portid, u32 seq,
4652 int event, unsigned int flags)
4654 struct prefixmsg *pmsg;
4655 struct nlmsghdr *nlh;
4656 struct prefix_cacheinfo ci;
4658 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4659 if (nlh == NULL)
4660 return -EMSGSIZE;
4662 pmsg = nlmsg_data(nlh);
4663 pmsg->prefix_family = AF_INET6;
4664 pmsg->prefix_pad1 = 0;
4665 pmsg->prefix_pad2 = 0;
4666 pmsg->prefix_ifindex = idev->dev->ifindex;
4667 pmsg->prefix_len = pinfo->prefix_len;
4668 pmsg->prefix_type = pinfo->type;
4669 pmsg->prefix_pad3 = 0;
4670 pmsg->prefix_flags = 0;
4671 if (pinfo->onlink)
4672 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4673 if (pinfo->autoconf)
4674 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4676 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4677 goto nla_put_failure;
4678 ci.preferred_time = ntohl(pinfo->prefered);
4679 ci.valid_time = ntohl(pinfo->valid);
4680 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4681 goto nla_put_failure;
4682 return nlmsg_end(skb, nlh);
4684 nla_put_failure:
4685 nlmsg_cancel(skb, nlh);
4686 return -EMSGSIZE;
4689 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4690 struct prefix_info *pinfo)
4692 struct sk_buff *skb;
4693 struct net *net = dev_net(idev->dev);
4694 int err = -ENOBUFS;
4696 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4697 if (skb == NULL)
4698 goto errout;
4700 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4701 if (err < 0) {
4702 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4703 WARN_ON(err == -EMSGSIZE);
4704 kfree_skb(skb);
4705 goto errout;
4707 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4708 return;
4709 errout:
4710 if (err < 0)
4711 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4714 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4716 struct net *net = dev_net(ifp->idev->dev);
4718 if (event)
4719 ASSERT_RTNL();
4721 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4723 switch (event) {
4724 case RTM_NEWADDR:
4726 * If the address was optimistic
4727 * we inserted the route at the start of
4728 * our DAD process, so we don't need
4729 * to do it again
4731 if (!(ifp->rt->rt6i_node))
4732 ip6_ins_rt(ifp->rt);
4733 if (ifp->idev->cnf.forwarding)
4734 addrconf_join_anycast(ifp);
4735 if (!ipv6_addr_any(&ifp->peer_addr))
4736 addrconf_prefix_route(&ifp->peer_addr, 128,
4737 ifp->idev->dev, 0, 0);
4738 break;
4739 case RTM_DELADDR:
4740 if (ifp->idev->cnf.forwarding)
4741 addrconf_leave_anycast(ifp);
4742 addrconf_leave_solict(ifp->idev, &ifp->addr);
4743 if (!ipv6_addr_any(&ifp->peer_addr)) {
4744 struct rt6_info *rt;
4745 struct net_device *dev = ifp->idev->dev;
4747 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4748 dev->ifindex, 1);
4749 if (rt) {
4750 dst_hold(&rt->dst);
4751 if (ip6_del_rt(rt))
4752 dst_free(&rt->dst);
4755 dst_hold(&ifp->rt->dst);
4757 if (ip6_del_rt(ifp->rt))
4758 dst_free(&ifp->rt->dst);
4759 break;
4761 atomic_inc(&net->ipv6.dev_addr_genid);
4762 rt_genid_bump_ipv6(net);
4765 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4767 rcu_read_lock_bh();
4768 if (likely(ifp->idev->dead == 0))
4769 __ipv6_ifa_notify(event, ifp);
4770 rcu_read_unlock_bh();
4773 #ifdef CONFIG_SYSCTL
4775 static
4776 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4777 void __user *buffer, size_t *lenp, loff_t *ppos)
4779 int *valp = ctl->data;
4780 int val = *valp;
4781 loff_t pos = *ppos;
4782 struct ctl_table lctl;
4783 int ret;
4786 * ctl->data points to idev->cnf.forwarding, we should
4787 * not modify it until we get the rtnl lock.
4789 lctl = *ctl;
4790 lctl.data = &val;
4792 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4794 if (write)
4795 ret = addrconf_fixup_forwarding(ctl, valp, val);
4796 if (ret)
4797 *ppos = pos;
4798 return ret;
4801 static void dev_disable_change(struct inet6_dev *idev)
4803 struct netdev_notifier_info info;
4805 if (!idev || !idev->dev)
4806 return;
4808 netdev_notifier_info_init(&info, idev->dev);
4809 if (idev->cnf.disable_ipv6)
4810 addrconf_notify(NULL, NETDEV_DOWN, &info);
4811 else
4812 addrconf_notify(NULL, NETDEV_UP, &info);
4815 static void addrconf_disable_change(struct net *net, __s32 newf)
4817 struct net_device *dev;
4818 struct inet6_dev *idev;
4820 rcu_read_lock();
4821 for_each_netdev_rcu(net, dev) {
4822 idev = __in6_dev_get(dev);
4823 if (idev) {
4824 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4825 idev->cnf.disable_ipv6 = newf;
4826 if (changed)
4827 dev_disable_change(idev);
4830 rcu_read_unlock();
4833 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4835 struct net *net;
4836 int old;
4838 if (!rtnl_trylock())
4839 return restart_syscall();
4841 net = (struct net *)table->extra2;
4842 old = *p;
4843 *p = newf;
4845 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4846 rtnl_unlock();
4847 return 0;
4850 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4851 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4852 addrconf_disable_change(net, newf);
4853 } else if ((!newf) ^ (!old))
4854 dev_disable_change((struct inet6_dev *)table->extra1);
4856 rtnl_unlock();
4857 return 0;
4860 static
4861 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4862 void __user *buffer, size_t *lenp, loff_t *ppos)
4864 int *valp = ctl->data;
4865 int val = *valp;
4866 loff_t pos = *ppos;
4867 struct ctl_table lctl;
4868 int ret;
4871 * ctl->data points to idev->cnf.disable_ipv6, we should
4872 * not modify it until we get the rtnl lock.
4874 lctl = *ctl;
4875 lctl.data = &val;
4877 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4879 if (write)
4880 ret = addrconf_disable_ipv6(ctl, valp, val);
4881 if (ret)
4882 *ppos = pos;
4883 return ret;
4886 static
4887 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4888 void __user *buffer, size_t *lenp, loff_t *ppos)
4890 int *valp = ctl->data;
4891 int ret;
4892 int old, new;
4894 old = *valp;
4895 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4896 new = *valp;
4898 if (write && old != new) {
4899 struct net *net = ctl->extra2;
4901 if (!rtnl_trylock())
4902 return restart_syscall();
4904 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4905 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4906 NETCONFA_IFINDEX_DEFAULT,
4907 net->ipv6.devconf_dflt);
4908 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4909 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4910 NETCONFA_IFINDEX_ALL,
4911 net->ipv6.devconf_all);
4912 else {
4913 struct inet6_dev *idev = ctl->extra1;
4915 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4916 idev->dev->ifindex,
4917 &idev->cnf);
4919 rtnl_unlock();
4922 return ret;
4926 static struct addrconf_sysctl_table
4928 struct ctl_table_header *sysctl_header;
4929 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4930 } addrconf_sysctl __read_mostly = {
4931 .sysctl_header = NULL,
4932 .addrconf_vars = {
4934 .procname = "forwarding",
4935 .data = &ipv6_devconf.forwarding,
4936 .maxlen = sizeof(int),
4937 .mode = 0644,
4938 .proc_handler = addrconf_sysctl_forward,
4941 .procname = "hop_limit",
4942 .data = &ipv6_devconf.hop_limit,
4943 .maxlen = sizeof(int),
4944 .mode = 0644,
4945 .proc_handler = proc_dointvec,
4948 .procname = "mtu",
4949 .data = &ipv6_devconf.mtu6,
4950 .maxlen = sizeof(int),
4951 .mode = 0644,
4952 .proc_handler = proc_dointvec,
4955 .procname = "accept_ra",
4956 .data = &ipv6_devconf.accept_ra,
4957 .maxlen = sizeof(int),
4958 .mode = 0644,
4959 .proc_handler = proc_dointvec,
4962 .procname = "accept_redirects",
4963 .data = &ipv6_devconf.accept_redirects,
4964 .maxlen = sizeof(int),
4965 .mode = 0644,
4966 .proc_handler = proc_dointvec,
4969 .procname = "autoconf",
4970 .data = &ipv6_devconf.autoconf,
4971 .maxlen = sizeof(int),
4972 .mode = 0644,
4973 .proc_handler = proc_dointvec,
4976 .procname = "dad_transmits",
4977 .data = &ipv6_devconf.dad_transmits,
4978 .maxlen = sizeof(int),
4979 .mode = 0644,
4980 .proc_handler = proc_dointvec,
4983 .procname = "router_solicitations",
4984 .data = &ipv6_devconf.rtr_solicits,
4985 .maxlen = sizeof(int),
4986 .mode = 0644,
4987 .proc_handler = proc_dointvec,
4990 .procname = "router_solicitation_interval",
4991 .data = &ipv6_devconf.rtr_solicit_interval,
4992 .maxlen = sizeof(int),
4993 .mode = 0644,
4994 .proc_handler = proc_dointvec_jiffies,
4997 .procname = "router_solicitation_delay",
4998 .data = &ipv6_devconf.rtr_solicit_delay,
4999 .maxlen = sizeof(int),
5000 .mode = 0644,
5001 .proc_handler = proc_dointvec_jiffies,
5004 .procname = "force_mld_version",
5005 .data = &ipv6_devconf.force_mld_version,
5006 .maxlen = sizeof(int),
5007 .mode = 0644,
5008 .proc_handler = proc_dointvec,
5011 .procname = "mldv1_unsolicited_report_interval",
5012 .data =
5013 &ipv6_devconf.mldv1_unsolicited_report_interval,
5014 .maxlen = sizeof(int),
5015 .mode = 0644,
5016 .proc_handler = proc_dointvec_ms_jiffies,
5019 .procname = "mldv2_unsolicited_report_interval",
5020 .data =
5021 &ipv6_devconf.mldv2_unsolicited_report_interval,
5022 .maxlen = sizeof(int),
5023 .mode = 0644,
5024 .proc_handler = proc_dointvec_ms_jiffies,
5027 .procname = "use_tempaddr",
5028 .data = &ipv6_devconf.use_tempaddr,
5029 .maxlen = sizeof(int),
5030 .mode = 0644,
5031 .proc_handler = proc_dointvec,
5034 .procname = "temp_valid_lft",
5035 .data = &ipv6_devconf.temp_valid_lft,
5036 .maxlen = sizeof(int),
5037 .mode = 0644,
5038 .proc_handler = proc_dointvec,
5041 .procname = "temp_prefered_lft",
5042 .data = &ipv6_devconf.temp_prefered_lft,
5043 .maxlen = sizeof(int),
5044 .mode = 0644,
5045 .proc_handler = proc_dointvec,
5048 .procname = "regen_max_retry",
5049 .data = &ipv6_devconf.regen_max_retry,
5050 .maxlen = sizeof(int),
5051 .mode = 0644,
5052 .proc_handler = proc_dointvec,
5055 .procname = "max_desync_factor",
5056 .data = &ipv6_devconf.max_desync_factor,
5057 .maxlen = sizeof(int),
5058 .mode = 0644,
5059 .proc_handler = proc_dointvec,
5062 .procname = "max_addresses",
5063 .data = &ipv6_devconf.max_addresses,
5064 .maxlen = sizeof(int),
5065 .mode = 0644,
5066 .proc_handler = proc_dointvec,
5069 .procname = "accept_ra_defrtr",
5070 .data = &ipv6_devconf.accept_ra_defrtr,
5071 .maxlen = sizeof(int),
5072 .mode = 0644,
5073 .proc_handler = proc_dointvec,
5076 .procname = "accept_ra_pinfo",
5077 .data = &ipv6_devconf.accept_ra_pinfo,
5078 .maxlen = sizeof(int),
5079 .mode = 0644,
5080 .proc_handler = proc_dointvec,
5082 #ifdef CONFIG_IPV6_ROUTER_PREF
5084 .procname = "accept_ra_rtr_pref",
5085 .data = &ipv6_devconf.accept_ra_rtr_pref,
5086 .maxlen = sizeof(int),
5087 .mode = 0644,
5088 .proc_handler = proc_dointvec,
5091 .procname = "router_probe_interval",
5092 .data = &ipv6_devconf.rtr_probe_interval,
5093 .maxlen = sizeof(int),
5094 .mode = 0644,
5095 .proc_handler = proc_dointvec_jiffies,
5097 #ifdef CONFIG_IPV6_ROUTE_INFO
5099 .procname = "accept_ra_rt_info_max_plen",
5100 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5101 .maxlen = sizeof(int),
5102 .mode = 0644,
5103 .proc_handler = proc_dointvec,
5105 #endif
5106 #endif
5108 .procname = "proxy_ndp",
5109 .data = &ipv6_devconf.proxy_ndp,
5110 .maxlen = sizeof(int),
5111 .mode = 0644,
5112 .proc_handler = addrconf_sysctl_proxy_ndp,
5115 .procname = "accept_source_route",
5116 .data = &ipv6_devconf.accept_source_route,
5117 .maxlen = sizeof(int),
5118 .mode = 0644,
5119 .proc_handler = proc_dointvec,
5121 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5123 .procname = "optimistic_dad",
5124 .data = &ipv6_devconf.optimistic_dad,
5125 .maxlen = sizeof(int),
5126 .mode = 0644,
5127 .proc_handler = proc_dointvec,
5130 #endif
5131 #ifdef CONFIG_IPV6_MROUTE
5133 .procname = "mc_forwarding",
5134 .data = &ipv6_devconf.mc_forwarding,
5135 .maxlen = sizeof(int),
5136 .mode = 0444,
5137 .proc_handler = proc_dointvec,
5139 #endif
5141 .procname = "disable_ipv6",
5142 .data = &ipv6_devconf.disable_ipv6,
5143 .maxlen = sizeof(int),
5144 .mode = 0644,
5145 .proc_handler = addrconf_sysctl_disable,
5148 .procname = "accept_dad",
5149 .data = &ipv6_devconf.accept_dad,
5150 .maxlen = sizeof(int),
5151 .mode = 0644,
5152 .proc_handler = proc_dointvec,
5155 .procname = "force_tllao",
5156 .data = &ipv6_devconf.force_tllao,
5157 .maxlen = sizeof(int),
5158 .mode = 0644,
5159 .proc_handler = proc_dointvec
5162 .procname = "ndisc_notify",
5163 .data = &ipv6_devconf.ndisc_notify,
5164 .maxlen = sizeof(int),
5165 .mode = 0644,
5166 .proc_handler = proc_dointvec
5169 .procname = "suppress_frag_ndisc",
5170 .data = &ipv6_devconf.suppress_frag_ndisc,
5171 .maxlen = sizeof(int),
5172 .mode = 0644,
5173 .proc_handler = proc_dointvec
5176 /* sentinel */
5181 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5182 struct inet6_dev *idev, struct ipv6_devconf *p)
5184 int i;
5185 struct addrconf_sysctl_table *t;
5186 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5188 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5189 if (t == NULL)
5190 goto out;
5192 for (i = 0; t->addrconf_vars[i].data; i++) {
5193 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5194 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5195 t->addrconf_vars[i].extra2 = net;
5198 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5200 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5201 if (t->sysctl_header == NULL)
5202 goto free;
5204 p->sysctl = t;
5205 return 0;
5207 free:
5208 kfree(t);
5209 out:
5210 return -ENOBUFS;
5213 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5215 struct addrconf_sysctl_table *t;
5217 if (p->sysctl == NULL)
5218 return;
5220 t = p->sysctl;
5221 p->sysctl = NULL;
5222 unregister_net_sysctl_table(t->sysctl_header);
5223 kfree(t);
5226 static void addrconf_sysctl_register(struct inet6_dev *idev)
5228 neigh_sysctl_register(idev->dev, idev->nd_parms,
5229 &ndisc_ifinfo_sysctl_change);
5230 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5231 idev, &idev->cnf);
5234 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5236 __addrconf_sysctl_unregister(&idev->cnf);
5237 neigh_sysctl_unregister(idev->nd_parms);
5241 #endif
5243 static int __net_init addrconf_init_net(struct net *net)
5245 int err = -ENOMEM;
5246 struct ipv6_devconf *all, *dflt;
5248 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5249 if (all == NULL)
5250 goto err_alloc_all;
5252 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5253 if (dflt == NULL)
5254 goto err_alloc_dflt;
5256 /* these will be inherited by all namespaces */
5257 dflt->autoconf = ipv6_defaults.autoconf;
5258 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5260 net->ipv6.devconf_all = all;
5261 net->ipv6.devconf_dflt = dflt;
5263 #ifdef CONFIG_SYSCTL
5264 err = __addrconf_sysctl_register(net, "all", NULL, all);
5265 if (err < 0)
5266 goto err_reg_all;
5268 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5269 if (err < 0)
5270 goto err_reg_dflt;
5271 #endif
5272 return 0;
5274 #ifdef CONFIG_SYSCTL
5275 err_reg_dflt:
5276 __addrconf_sysctl_unregister(all);
5277 err_reg_all:
5278 kfree(dflt);
5279 #endif
5280 err_alloc_dflt:
5281 kfree(all);
5282 err_alloc_all:
5283 return err;
5286 static void __net_exit addrconf_exit_net(struct net *net)
5288 #ifdef CONFIG_SYSCTL
5289 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5290 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5291 #endif
5292 if (!net_eq(net, &init_net)) {
5293 kfree(net->ipv6.devconf_dflt);
5294 kfree(net->ipv6.devconf_all);
5298 static struct pernet_operations addrconf_ops = {
5299 .init = addrconf_init_net,
5300 .exit = addrconf_exit_net,
5303 static struct rtnl_af_ops inet6_ops = {
5304 .family = AF_INET6,
5305 .fill_link_af = inet6_fill_link_af,
5306 .get_link_af_size = inet6_get_link_af_size,
5307 .set_link_af = inet6_set_link_af,
5311 * Init / cleanup code
5314 int __init addrconf_init(void)
5316 int i, err;
5318 err = ipv6_addr_label_init();
5319 if (err < 0) {
5320 pr_crit("%s: cannot initialize default policy table: %d\n",
5321 __func__, err);
5322 goto out;
5325 err = register_pernet_subsys(&addrconf_ops);
5326 if (err < 0)
5327 goto out_addrlabel;
5329 addrconf_wq = create_workqueue("ipv6_addrconf");
5330 if (!addrconf_wq) {
5331 err = -ENOMEM;
5332 goto out_nowq;
5335 /* The addrconf netdev notifier requires that loopback_dev
5336 * has it's ipv6 private information allocated and setup
5337 * before it can bring up and give link-local addresses
5338 * to other devices which are up.
5340 * Unfortunately, loopback_dev is not necessarily the first
5341 * entry in the global dev_base list of net devices. In fact,
5342 * it is likely to be the very last entry on that list.
5343 * So this causes the notifier registry below to try and
5344 * give link-local addresses to all devices besides loopback_dev
5345 * first, then loopback_dev, which cases all the non-loopback_dev
5346 * devices to fail to get a link-local address.
5348 * So, as a temporary fix, allocate the ipv6 structure for
5349 * loopback_dev first by hand.
5350 * Longer term, all of the dependencies ipv6 has upon the loopback
5351 * device and it being up should be removed.
5353 rtnl_lock();
5354 if (!ipv6_add_dev(init_net.loopback_dev))
5355 err = -ENOMEM;
5356 rtnl_unlock();
5357 if (err)
5358 goto errlo;
5360 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5361 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5363 register_netdevice_notifier(&ipv6_dev_notf);
5365 addrconf_verify();
5367 rtnl_af_register(&inet6_ops);
5369 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5370 NULL);
5371 if (err < 0)
5372 goto errout;
5374 /* Only the first call to __rtnl_register can fail */
5375 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5376 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5377 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5378 inet6_dump_ifaddr, NULL);
5379 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5380 inet6_dump_ifmcaddr, NULL);
5381 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5382 inet6_dump_ifacaddr, NULL);
5383 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5384 inet6_netconf_dump_devconf, NULL);
5386 ipv6_addr_label_rtnl_register();
5388 return 0;
5389 errout:
5390 rtnl_af_unregister(&inet6_ops);
5391 unregister_netdevice_notifier(&ipv6_dev_notf);
5392 errlo:
5393 destroy_workqueue(addrconf_wq);
5394 out_nowq:
5395 unregister_pernet_subsys(&addrconf_ops);
5396 out_addrlabel:
5397 ipv6_addr_label_cleanup();
5398 out:
5399 return err;
5402 void addrconf_cleanup(void)
5404 struct net_device *dev;
5405 int i;
5407 unregister_netdevice_notifier(&ipv6_dev_notf);
5408 unregister_pernet_subsys(&addrconf_ops);
5409 ipv6_addr_label_cleanup();
5411 rtnl_lock();
5413 __rtnl_af_unregister(&inet6_ops);
5415 /* clean dev list */
5416 for_each_netdev(&init_net, dev) {
5417 if (__in6_dev_get(dev) == NULL)
5418 continue;
5419 addrconf_ifdown(dev, 1);
5421 addrconf_ifdown(init_net.loopback_dev, 2);
5424 * Check hash table.
5426 spin_lock_bh(&addrconf_hash_lock);
5427 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5428 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5429 spin_unlock_bh(&addrconf_hash_lock);
5430 cancel_delayed_work(&addr_chk_work);
5431 rtnl_unlock();
5433 destroy_workqueue(addrconf_wq);