Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blobfdbfeca36d6344c22db31886c3661796eb9e8f86
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(unsigned long);
138 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
139 static DEFINE_SPINLOCK(addrconf_verify_lock);
141 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
142 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
144 static void addrconf_type_change(struct net_device *dev,
145 unsigned long event);
146 static int addrconf_ifdown(struct net_device *dev, int how);
148 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
149 int plen,
150 const struct net_device *dev,
151 u32 flags, u32 noflags);
153 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
154 static void addrconf_dad_timer(unsigned long data);
155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
156 static void addrconf_dad_run(struct inet6_dev *idev);
157 static void addrconf_rs_timer(unsigned long data);
158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
159 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
161 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
162 struct prefix_info *pinfo);
163 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
164 struct net_device *dev);
166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
167 .forwarding = 0,
168 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
169 .mtu6 = IPV6_MIN_MTU,
170 .accept_ra = 1,
171 .accept_redirects = 1,
172 .autoconf = 1,
173 .force_mld_version = 0,
174 .mldv1_unsolicited_report_interval = 10 * HZ,
175 .mldv2_unsolicited_report_interval = HZ,
176 .dad_transmits = 1,
177 .rtr_solicits = MAX_RTR_SOLICITATIONS,
178 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
179 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
180 .use_tempaddr = 0,
181 .temp_valid_lft = TEMP_VALID_LIFETIME,
182 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
183 .regen_max_retry = REGEN_MAX_RETRY,
184 .max_desync_factor = MAX_DESYNC_FACTOR,
185 .max_addresses = IPV6_MAX_ADDRESSES,
186 .accept_ra_defrtr = 1,
187 .accept_ra_pinfo = 1,
188 #ifdef CONFIG_IPV6_ROUTER_PREF
189 .accept_ra_rtr_pref = 1,
190 .rtr_probe_interval = 60 * HZ,
191 #ifdef CONFIG_IPV6_ROUTE_INFO
192 .accept_ra_rt_info_max_plen = 0,
193 #endif
194 #endif
195 .proxy_ndp = 0,
196 .accept_source_route = 0, /* we do not accept RH0 by default. */
197 .disable_ipv6 = 0,
198 .accept_dad = 1,
199 .suppress_frag_ndisc = 1,
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 .forwarding = 0,
204 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
205 .mtu6 = IPV6_MIN_MTU,
206 .accept_ra = 1,
207 .accept_redirects = 1,
208 .autoconf = 1,
209 .force_mld_version = 0,
210 .mldv1_unsolicited_report_interval = 10 * HZ,
211 .mldv2_unsolicited_report_interval = HZ,
212 .dad_transmits = 1,
213 .rtr_solicits = MAX_RTR_SOLICITATIONS,
214 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
215 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
216 .use_tempaddr = 0,
217 .temp_valid_lft = TEMP_VALID_LIFETIME,
218 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
219 .regen_max_retry = REGEN_MAX_RETRY,
220 .max_desync_factor = MAX_DESYNC_FACTOR,
221 .max_addresses = IPV6_MAX_ADDRESSES,
222 .accept_ra_defrtr = 1,
223 .accept_ra_pinfo = 1,
224 #ifdef CONFIG_IPV6_ROUTER_PREF
225 .accept_ra_rtr_pref = 1,
226 .rtr_probe_interval = 60 * HZ,
227 #ifdef CONFIG_IPV6_ROUTE_INFO
228 .accept_ra_rt_info_max_plen = 0,
229 #endif
230 #endif
231 .proxy_ndp = 0,
232 .accept_source_route = 0, /* we do not accept RH0 by default. */
233 .disable_ipv6 = 0,
234 .accept_dad = 1,
235 .suppress_frag_ndisc = 1,
238 /* Check if a valid qdisc is available */
239 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
241 return !qdisc_tx_is_noop(dev);
244 static void addrconf_del_rs_timer(struct inet6_dev *idev)
246 if (del_timer(&idev->rs_timer))
247 __in6_dev_put(idev);
250 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
252 if (del_timer(&ifp->dad_timer))
253 __in6_ifa_put(ifp);
256 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
257 unsigned long when)
259 if (!timer_pending(&idev->rs_timer))
260 in6_dev_hold(idev);
261 mod_timer(&idev->rs_timer, jiffies + when);
264 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
265 unsigned long when)
267 if (!timer_pending(&ifp->dad_timer))
268 in6_ifa_hold(ifp);
269 mod_timer(&ifp->dad_timer, jiffies + when);
272 static int snmp6_alloc_dev(struct inet6_dev *idev)
274 int i;
276 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
277 sizeof(struct ipstats_mib),
278 __alignof__(struct ipstats_mib)) < 0)
279 goto err_ip;
281 for_each_possible_cpu(i) {
282 struct ipstats_mib *addrconf_stats;
283 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i);
284 u64_stats_init(&addrconf_stats->syncp);
285 #if SNMP_ARRAY_SZ == 2
286 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i);
287 u64_stats_init(&addrconf_stats->syncp);
288 #endif
292 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
293 GFP_KERNEL);
294 if (!idev->stats.icmpv6dev)
295 goto err_icmp;
296 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
297 GFP_KERNEL);
298 if (!idev->stats.icmpv6msgdev)
299 goto err_icmpmsg;
301 return 0;
303 err_icmpmsg:
304 kfree(idev->stats.icmpv6dev);
305 err_icmp:
306 snmp_mib_free((void __percpu **)idev->stats.ipv6);
307 err_ip:
308 return -ENOMEM;
311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
313 struct inet6_dev *ndev;
315 ASSERT_RTNL();
317 if (dev->mtu < IPV6_MIN_MTU)
318 return NULL;
320 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
322 if (ndev == NULL)
323 return NULL;
325 rwlock_init(&ndev->lock);
326 ndev->dev = dev;
327 INIT_LIST_HEAD(&ndev->addr_list);
328 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
329 (unsigned long)ndev);
330 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
331 ndev->cnf.mtu6 = dev->mtu;
332 ndev->cnf.sysctl = NULL;
333 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
334 if (ndev->nd_parms == NULL) {
335 kfree(ndev);
336 return NULL;
338 if (ndev->cnf.forwarding)
339 dev_disable_lro(dev);
340 /* We refer to the device */
341 dev_hold(dev);
343 if (snmp6_alloc_dev(ndev) < 0) {
344 ADBG(KERN_WARNING
345 "%s: cannot allocate memory for statistics; dev=%s.\n",
346 __func__, dev->name);
347 neigh_parms_release(&nd_tbl, ndev->nd_parms);
348 dev_put(dev);
349 kfree(ndev);
350 return NULL;
353 if (snmp6_register_dev(ndev) < 0) {
354 ADBG(KERN_WARNING
355 "%s: cannot create /proc/net/dev_snmp6/%s\n",
356 __func__, dev->name);
357 neigh_parms_release(&nd_tbl, ndev->nd_parms);
358 ndev->dead = 1;
359 in6_dev_finish_destroy(ndev);
360 return NULL;
363 /* One reference from device. We must do this before
364 * we invoke __ipv6_regen_rndid().
366 in6_dev_hold(ndev);
368 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
369 ndev->cnf.accept_dad = -1;
371 #if IS_ENABLED(CONFIG_IPV6_SIT)
372 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
373 pr_info("%s: Disabled Multicast RS\n", dev->name);
374 ndev->cnf.rtr_solicits = 0;
376 #endif
378 INIT_LIST_HEAD(&ndev->tempaddr_list);
379 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
380 if ((dev->flags&IFF_LOOPBACK) ||
381 dev->type == ARPHRD_TUNNEL ||
382 dev->type == ARPHRD_TUNNEL6 ||
383 dev->type == ARPHRD_SIT ||
384 dev->type == ARPHRD_NONE) {
385 ndev->cnf.use_tempaddr = -1;
386 } else {
387 in6_dev_hold(ndev);
388 ipv6_regen_rndid((unsigned long) ndev);
391 ndev->token = in6addr_any;
393 if (netif_running(dev) && addrconf_qdisc_ok(dev))
394 ndev->if_flags |= IF_READY;
396 ipv6_mc_init_dev(ndev);
397 ndev->tstamp = jiffies;
398 addrconf_sysctl_register(ndev);
399 /* protected by rtnl_lock */
400 rcu_assign_pointer(dev->ip6_ptr, ndev);
402 /* Join interface-local all-node multicast group */
403 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
405 /* Join all-node multicast group */
406 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
408 /* Join all-router multicast group if forwarding is set */
409 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
410 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
412 return ndev;
415 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
417 struct inet6_dev *idev;
419 ASSERT_RTNL();
421 idev = __in6_dev_get(dev);
422 if (!idev) {
423 idev = ipv6_add_dev(dev);
424 if (!idev)
425 return NULL;
428 if (dev->flags&IFF_UP)
429 ipv6_mc_up(idev);
430 return idev;
433 static int inet6_netconf_msgsize_devconf(int type)
435 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
436 + nla_total_size(4); /* NETCONFA_IFINDEX */
438 /* type -1 is used for ALL */
439 if (type == -1 || type == NETCONFA_FORWARDING)
440 size += nla_total_size(4);
441 #ifdef CONFIG_IPV6_MROUTE
442 if (type == -1 || type == NETCONFA_MC_FORWARDING)
443 size += nla_total_size(4);
444 #endif
445 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
446 size += nla_total_size(4);
448 return size;
451 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
452 struct ipv6_devconf *devconf, u32 portid,
453 u32 seq, int event, unsigned int flags,
454 int type)
456 struct nlmsghdr *nlh;
457 struct netconfmsg *ncm;
459 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
460 flags);
461 if (nlh == NULL)
462 return -EMSGSIZE;
464 ncm = nlmsg_data(nlh);
465 ncm->ncm_family = AF_INET6;
467 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
468 goto nla_put_failure;
470 /* type -1 is used for ALL */
471 if ((type == -1 || type == NETCONFA_FORWARDING) &&
472 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
473 goto nla_put_failure;
474 #ifdef CONFIG_IPV6_MROUTE
475 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
476 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
477 devconf->mc_forwarding) < 0)
478 goto nla_put_failure;
479 #endif
480 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
481 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
482 goto nla_put_failure;
484 return nlmsg_end(skb, nlh);
486 nla_put_failure:
487 nlmsg_cancel(skb, nlh);
488 return -EMSGSIZE;
491 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
492 struct ipv6_devconf *devconf)
494 struct sk_buff *skb;
495 int err = -ENOBUFS;
497 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
498 if (skb == NULL)
499 goto errout;
501 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
502 RTM_NEWNETCONF, 0, type);
503 if (err < 0) {
504 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
505 WARN_ON(err == -EMSGSIZE);
506 kfree_skb(skb);
507 goto errout;
509 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
510 return;
511 errout:
512 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
515 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
516 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
517 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
518 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
521 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
522 struct nlmsghdr *nlh)
524 struct net *net = sock_net(in_skb->sk);
525 struct nlattr *tb[NETCONFA_MAX+1];
526 struct netconfmsg *ncm;
527 struct sk_buff *skb;
528 struct ipv6_devconf *devconf;
529 struct inet6_dev *in6_dev;
530 struct net_device *dev;
531 int ifindex;
532 int err;
534 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
535 devconf_ipv6_policy);
536 if (err < 0)
537 goto errout;
539 err = EINVAL;
540 if (!tb[NETCONFA_IFINDEX])
541 goto errout;
543 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
544 switch (ifindex) {
545 case NETCONFA_IFINDEX_ALL:
546 devconf = net->ipv6.devconf_all;
547 break;
548 case NETCONFA_IFINDEX_DEFAULT:
549 devconf = net->ipv6.devconf_dflt;
550 break;
551 default:
552 dev = __dev_get_by_index(net, ifindex);
553 if (dev == NULL)
554 goto errout;
555 in6_dev = __in6_dev_get(dev);
556 if (in6_dev == NULL)
557 goto errout;
558 devconf = &in6_dev->cnf;
559 break;
562 err = -ENOBUFS;
563 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
564 if (skb == NULL)
565 goto errout;
567 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
568 NETLINK_CB(in_skb).portid,
569 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
570 -1);
571 if (err < 0) {
572 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
573 WARN_ON(err == -EMSGSIZE);
574 kfree_skb(skb);
575 goto errout;
577 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
578 errout:
579 return err;
582 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
583 struct netlink_callback *cb)
585 struct net *net = sock_net(skb->sk);
586 int h, s_h;
587 int idx, s_idx;
588 struct net_device *dev;
589 struct inet6_dev *idev;
590 struct hlist_head *head;
592 s_h = cb->args[0];
593 s_idx = idx = cb->args[1];
595 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
596 idx = 0;
597 head = &net->dev_index_head[h];
598 rcu_read_lock();
599 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
600 net->dev_base_seq;
601 hlist_for_each_entry_rcu(dev, head, index_hlist) {
602 if (idx < s_idx)
603 goto cont;
604 idev = __in6_dev_get(dev);
605 if (!idev)
606 goto cont;
608 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
609 &idev->cnf,
610 NETLINK_CB(cb->skb).portid,
611 cb->nlh->nlmsg_seq,
612 RTM_NEWNETCONF,
613 NLM_F_MULTI,
614 -1) <= 0) {
615 rcu_read_unlock();
616 goto done;
618 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
619 cont:
620 idx++;
622 rcu_read_unlock();
624 if (h == NETDEV_HASHENTRIES) {
625 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
626 net->ipv6.devconf_all,
627 NETLINK_CB(cb->skb).portid,
628 cb->nlh->nlmsg_seq,
629 RTM_NEWNETCONF, NLM_F_MULTI,
630 -1) <= 0)
631 goto done;
632 else
633 h++;
635 if (h == NETDEV_HASHENTRIES + 1) {
636 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
637 net->ipv6.devconf_dflt,
638 NETLINK_CB(cb->skb).portid,
639 cb->nlh->nlmsg_seq,
640 RTM_NEWNETCONF, NLM_F_MULTI,
641 -1) <= 0)
642 goto done;
643 else
644 h++;
646 done:
647 cb->args[0] = h;
648 cb->args[1] = idx;
650 return skb->len;
653 #ifdef CONFIG_SYSCTL
654 static void dev_forward_change(struct inet6_dev *idev)
656 struct net_device *dev;
657 struct inet6_ifaddr *ifa;
659 if (!idev)
660 return;
661 dev = idev->dev;
662 if (idev->cnf.forwarding)
663 dev_disable_lro(dev);
664 if (dev->flags & IFF_MULTICAST) {
665 if (idev->cnf.forwarding) {
666 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
667 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
668 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
669 } else {
670 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
671 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
672 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
676 list_for_each_entry(ifa, &idev->addr_list, if_list) {
677 if (ifa->flags&IFA_F_TENTATIVE)
678 continue;
679 if (idev->cnf.forwarding)
680 addrconf_join_anycast(ifa);
681 else
682 addrconf_leave_anycast(ifa);
684 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
685 dev->ifindex, &idev->cnf);
689 static void addrconf_forward_change(struct net *net, __s32 newf)
691 struct net_device *dev;
692 struct inet6_dev *idev;
694 for_each_netdev(net, dev) {
695 idev = __in6_dev_get(dev);
696 if (idev) {
697 int changed = (!idev->cnf.forwarding) ^ (!newf);
698 idev->cnf.forwarding = newf;
699 if (changed)
700 dev_forward_change(idev);
705 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
707 struct net *net;
708 int old;
710 if (!rtnl_trylock())
711 return restart_syscall();
713 net = (struct net *)table->extra2;
714 old = *p;
715 *p = newf;
717 if (p == &net->ipv6.devconf_dflt->forwarding) {
718 if ((!newf) ^ (!old))
719 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
720 NETCONFA_IFINDEX_DEFAULT,
721 net->ipv6.devconf_dflt);
722 rtnl_unlock();
723 return 0;
726 if (p == &net->ipv6.devconf_all->forwarding) {
727 net->ipv6.devconf_dflt->forwarding = newf;
728 addrconf_forward_change(net, newf);
729 if ((!newf) ^ (!old))
730 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
731 NETCONFA_IFINDEX_ALL,
732 net->ipv6.devconf_all);
733 } else if ((!newf) ^ (!old))
734 dev_forward_change((struct inet6_dev *)table->extra1);
735 rtnl_unlock();
737 if (newf)
738 rt6_purge_dflt_routers(net);
739 return 1;
741 #endif
743 /* Nobody refers to this ifaddr, destroy it */
744 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
746 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
748 #ifdef NET_REFCNT_DEBUG
749 pr_debug("%s\n", __func__);
750 #endif
752 in6_dev_put(ifp->idev);
754 if (del_timer(&ifp->dad_timer))
755 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
757 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
758 pr_warn("Freeing alive inet6 address %p\n", ifp);
759 return;
761 ip6_rt_put(ifp->rt);
763 kfree_rcu(ifp, rcu);
766 static void
767 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
769 struct list_head *p;
770 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
773 * Each device address list is sorted in order of scope -
774 * global before linklocal.
776 list_for_each(p, &idev->addr_list) {
777 struct inet6_ifaddr *ifa
778 = list_entry(p, struct inet6_ifaddr, if_list);
779 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
780 break;
783 list_add_tail(&ifp->if_list, p);
786 static u32 inet6_addr_hash(const struct in6_addr *addr)
788 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
791 /* On success it returns ifp with increased reference count */
793 static struct inet6_ifaddr *
794 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
795 const struct in6_addr *peer_addr, int pfxlen,
796 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
798 struct inet6_ifaddr *ifa = NULL;
799 struct rt6_info *rt;
800 unsigned int hash;
801 int err = 0;
802 int addr_type = ipv6_addr_type(addr);
804 if (addr_type == IPV6_ADDR_ANY ||
805 addr_type & IPV6_ADDR_MULTICAST ||
806 (!(idev->dev->flags & IFF_LOOPBACK) &&
807 addr_type & IPV6_ADDR_LOOPBACK))
808 return ERR_PTR(-EADDRNOTAVAIL);
810 rcu_read_lock_bh();
811 if (idev->dead) {
812 err = -ENODEV; /*XXX*/
813 goto out2;
816 if (idev->cnf.disable_ipv6) {
817 err = -EACCES;
818 goto out2;
821 spin_lock(&addrconf_hash_lock);
823 /* Ignore adding duplicate addresses on an interface */
824 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
825 ADBG("ipv6_add_addr: already assigned\n");
826 err = -EEXIST;
827 goto out;
830 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
832 if (ifa == NULL) {
833 ADBG("ipv6_add_addr: malloc failed\n");
834 err = -ENOBUFS;
835 goto out;
838 rt = addrconf_dst_alloc(idev, addr, false);
839 if (IS_ERR(rt)) {
840 err = PTR_ERR(rt);
841 goto out;
844 neigh_parms_data_state_setall(idev->nd_parms);
846 ifa->addr = *addr;
847 if (peer_addr)
848 ifa->peer_addr = *peer_addr;
850 spin_lock_init(&ifa->lock);
851 spin_lock_init(&ifa->state_lock);
852 setup_timer(&ifa->dad_timer, addrconf_dad_timer,
853 (unsigned long)ifa);
854 INIT_HLIST_NODE(&ifa->addr_lst);
855 ifa->scope = scope;
856 ifa->prefix_len = pfxlen;
857 ifa->flags = flags | IFA_F_TENTATIVE;
858 ifa->valid_lft = valid_lft;
859 ifa->prefered_lft = prefered_lft;
860 ifa->cstamp = ifa->tstamp = jiffies;
861 ifa->tokenized = false;
863 ifa->rt = rt;
865 ifa->idev = idev;
866 in6_dev_hold(idev);
867 /* For caller */
868 in6_ifa_hold(ifa);
870 /* Add to big hash table */
871 hash = inet6_addr_hash(addr);
873 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
874 spin_unlock(&addrconf_hash_lock);
876 write_lock(&idev->lock);
877 /* Add to inet6_dev unicast addr list. */
878 ipv6_link_dev_addr(idev, ifa);
880 if (ifa->flags&IFA_F_TEMPORARY) {
881 list_add(&ifa->tmp_list, &idev->tempaddr_list);
882 in6_ifa_hold(ifa);
885 in6_ifa_hold(ifa);
886 write_unlock(&idev->lock);
887 out2:
888 rcu_read_unlock_bh();
890 if (likely(err == 0))
891 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
892 else {
893 kfree(ifa);
894 ifa = ERR_PTR(err);
897 return ifa;
898 out:
899 spin_unlock(&addrconf_hash_lock);
900 goto out2;
903 enum cleanup_prefix_rt_t {
904 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
905 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
906 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
910 * Check, whether the prefix for ifp would still need a prefix route
911 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
912 * constants.
914 * 1) we don't purge prefix if address was not permanent.
915 * prefix is managed by its own lifetime.
916 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
917 * 3) if there are no addresses, delete prefix.
918 * 4) if there are still other permanent address(es),
919 * corresponding prefix is still permanent.
920 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
921 * don't purge the prefix, assume user space is managing it.
922 * 6) otherwise, update prefix lifetime to the
923 * longest valid lifetime among the corresponding
924 * addresses on the device.
925 * Note: subsequent RA will update lifetime.
927 static enum cleanup_prefix_rt_t
928 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
930 struct inet6_ifaddr *ifa;
931 struct inet6_dev *idev = ifp->idev;
932 unsigned long lifetime;
933 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
935 *expires = jiffies;
937 list_for_each_entry(ifa, &idev->addr_list, if_list) {
938 if (ifa == ifp)
939 continue;
940 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
941 ifp->prefix_len))
942 continue;
943 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
944 return CLEANUP_PREFIX_RT_NOP;
946 action = CLEANUP_PREFIX_RT_EXPIRE;
948 spin_lock(&ifa->lock);
950 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
952 * Note: Because this address is
953 * not permanent, lifetime <
954 * LONG_MAX / HZ here.
956 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
957 *expires = ifa->tstamp + lifetime * HZ;
958 spin_unlock(&ifa->lock);
961 return action;
964 static void
965 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
967 struct rt6_info *rt;
969 rt = addrconf_get_prefix_route(&ifp->addr,
970 ifp->prefix_len,
971 ifp->idev->dev,
972 0, RTF_GATEWAY | RTF_DEFAULT);
973 if (rt) {
974 if (del_rt)
975 ip6_del_rt(rt);
976 else {
977 if (!(rt->rt6i_flags & RTF_EXPIRES))
978 rt6_set_expires(rt, expires);
979 ip6_rt_put(rt);
985 /* This function wants to get referenced ifp and releases it before return */
987 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
989 int state;
990 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
991 unsigned long expires;
993 spin_lock_bh(&ifp->state_lock);
994 state = ifp->state;
995 ifp->state = INET6_IFADDR_STATE_DEAD;
996 spin_unlock_bh(&ifp->state_lock);
998 if (state == INET6_IFADDR_STATE_DEAD)
999 goto out;
1001 spin_lock_bh(&addrconf_hash_lock);
1002 hlist_del_init_rcu(&ifp->addr_lst);
1003 spin_unlock_bh(&addrconf_hash_lock);
1005 write_lock_bh(&ifp->idev->lock);
1007 if (ifp->flags&IFA_F_TEMPORARY) {
1008 list_del(&ifp->tmp_list);
1009 if (ifp->ifpub) {
1010 in6_ifa_put(ifp->ifpub);
1011 ifp->ifpub = NULL;
1013 __in6_ifa_put(ifp);
1016 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1017 action = check_cleanup_prefix_route(ifp, &expires);
1019 list_del_init(&ifp->if_list);
1020 __in6_ifa_put(ifp);
1022 write_unlock_bh(&ifp->idev->lock);
1024 addrconf_del_dad_timer(ifp);
1026 ipv6_ifa_notify(RTM_DELADDR, ifp);
1028 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1030 if (action != CLEANUP_PREFIX_RT_NOP) {
1031 cleanup_prefix_route(ifp, expires,
1032 action == CLEANUP_PREFIX_RT_DEL);
1035 /* clean up prefsrc entries */
1036 rt6_remove_prefsrc(ifp);
1037 out:
1038 in6_ifa_put(ifp);
1041 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1043 struct inet6_dev *idev = ifp->idev;
1044 struct in6_addr addr, *tmpaddr;
1045 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1046 unsigned long regen_advance;
1047 int tmp_plen;
1048 int ret = 0;
1049 u32 addr_flags;
1050 unsigned long now = jiffies;
1052 write_lock_bh(&idev->lock);
1053 if (ift) {
1054 spin_lock_bh(&ift->lock);
1055 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1056 spin_unlock_bh(&ift->lock);
1057 tmpaddr = &addr;
1058 } else {
1059 tmpaddr = NULL;
1061 retry:
1062 in6_dev_hold(idev);
1063 if (idev->cnf.use_tempaddr <= 0) {
1064 write_unlock_bh(&idev->lock);
1065 pr_info("%s: use_tempaddr is disabled\n", __func__);
1066 in6_dev_put(idev);
1067 ret = -1;
1068 goto out;
1070 spin_lock_bh(&ifp->lock);
1071 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1072 idev->cnf.use_tempaddr = -1; /*XXX*/
1073 spin_unlock_bh(&ifp->lock);
1074 write_unlock_bh(&idev->lock);
1075 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1076 __func__);
1077 in6_dev_put(idev);
1078 ret = -1;
1079 goto out;
1081 in6_ifa_hold(ifp);
1082 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1083 __ipv6_try_regen_rndid(idev, tmpaddr);
1084 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1085 age = (now - ifp->tstamp) / HZ;
1086 tmp_valid_lft = min_t(__u32,
1087 ifp->valid_lft,
1088 idev->cnf.temp_valid_lft + age);
1089 tmp_prefered_lft = min_t(__u32,
1090 ifp->prefered_lft,
1091 idev->cnf.temp_prefered_lft + age -
1092 idev->cnf.max_desync_factor);
1093 tmp_plen = ifp->prefix_len;
1094 tmp_tstamp = ifp->tstamp;
1095 spin_unlock_bh(&ifp->lock);
1097 regen_advance = idev->cnf.regen_max_retry *
1098 idev->cnf.dad_transmits *
1099 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1100 write_unlock_bh(&idev->lock);
1102 /* A temporary address is created only if this calculated Preferred
1103 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1104 * an implementation must not create a temporary address with a zero
1105 * Preferred Lifetime.
1107 if (tmp_prefered_lft <= regen_advance) {
1108 in6_ifa_put(ifp);
1109 in6_dev_put(idev);
1110 ret = -1;
1111 goto out;
1114 addr_flags = IFA_F_TEMPORARY;
1115 /* set in addrconf_prefix_rcv() */
1116 if (ifp->flags & IFA_F_OPTIMISTIC)
1117 addr_flags |= IFA_F_OPTIMISTIC;
1119 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1120 ipv6_addr_scope(&addr), addr_flags,
1121 tmp_valid_lft, tmp_prefered_lft);
1122 if (IS_ERR(ift)) {
1123 in6_ifa_put(ifp);
1124 in6_dev_put(idev);
1125 pr_info("%s: retry temporary address regeneration\n", __func__);
1126 tmpaddr = &addr;
1127 write_lock_bh(&idev->lock);
1128 goto retry;
1131 spin_lock_bh(&ift->lock);
1132 ift->ifpub = ifp;
1133 ift->cstamp = now;
1134 ift->tstamp = tmp_tstamp;
1135 spin_unlock_bh(&ift->lock);
1137 addrconf_dad_start(ift);
1138 in6_ifa_put(ift);
1139 in6_dev_put(idev);
1140 out:
1141 return ret;
1145 * Choose an appropriate source address (RFC3484)
1147 enum {
1148 IPV6_SADDR_RULE_INIT = 0,
1149 IPV6_SADDR_RULE_LOCAL,
1150 IPV6_SADDR_RULE_SCOPE,
1151 IPV6_SADDR_RULE_PREFERRED,
1152 #ifdef CONFIG_IPV6_MIP6
1153 IPV6_SADDR_RULE_HOA,
1154 #endif
1155 IPV6_SADDR_RULE_OIF,
1156 IPV6_SADDR_RULE_LABEL,
1157 IPV6_SADDR_RULE_PRIVACY,
1158 IPV6_SADDR_RULE_ORCHID,
1159 IPV6_SADDR_RULE_PREFIX,
1160 IPV6_SADDR_RULE_MAX
1163 struct ipv6_saddr_score {
1164 int rule;
1165 int addr_type;
1166 struct inet6_ifaddr *ifa;
1167 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1168 int scopedist;
1169 int matchlen;
1172 struct ipv6_saddr_dst {
1173 const struct in6_addr *addr;
1174 int ifindex;
1175 int scope;
1176 int label;
1177 unsigned int prefs;
1180 static inline int ipv6_saddr_preferred(int type)
1182 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1183 return 1;
1184 return 0;
1187 static int ipv6_get_saddr_eval(struct net *net,
1188 struct ipv6_saddr_score *score,
1189 struct ipv6_saddr_dst *dst,
1190 int i)
1192 int ret;
1194 if (i <= score->rule) {
1195 switch (i) {
1196 case IPV6_SADDR_RULE_SCOPE:
1197 ret = score->scopedist;
1198 break;
1199 case IPV6_SADDR_RULE_PREFIX:
1200 ret = score->matchlen;
1201 break;
1202 default:
1203 ret = !!test_bit(i, score->scorebits);
1205 goto out;
1208 switch (i) {
1209 case IPV6_SADDR_RULE_INIT:
1210 /* Rule 0: remember if hiscore is not ready yet */
1211 ret = !!score->ifa;
1212 break;
1213 case IPV6_SADDR_RULE_LOCAL:
1214 /* Rule 1: Prefer same address */
1215 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1216 break;
1217 case IPV6_SADDR_RULE_SCOPE:
1218 /* Rule 2: Prefer appropriate scope
1220 * ret
1222 * -1 | d 15
1223 * ---+--+-+---> scope
1225 * | d is scope of the destination.
1226 * B-d | \
1227 * | \ <- smaller scope is better if
1228 * B-15 | \ if scope is enough for destination.
1229 * | ret = B - scope (-1 <= scope >= d <= 15).
1230 * d-C-1 | /
1231 * |/ <- greater is better
1232 * -C / if scope is not enough for destination.
1233 * /| ret = scope - C (-1 <= d < scope <= 15).
1235 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1236 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1237 * Assume B = 0 and we get C > 29.
1239 ret = __ipv6_addr_src_scope(score->addr_type);
1240 if (ret >= dst->scope)
1241 ret = -ret;
1242 else
1243 ret -= 128; /* 30 is enough */
1244 score->scopedist = ret;
1245 break;
1246 case IPV6_SADDR_RULE_PREFERRED:
1247 /* Rule 3: Avoid deprecated and optimistic addresses */
1248 ret = ipv6_saddr_preferred(score->addr_type) ||
1249 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1250 break;
1251 #ifdef CONFIG_IPV6_MIP6
1252 case IPV6_SADDR_RULE_HOA:
1254 /* Rule 4: Prefer home address */
1255 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1256 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1257 break;
1259 #endif
1260 case IPV6_SADDR_RULE_OIF:
1261 /* Rule 5: Prefer outgoing interface */
1262 ret = (!dst->ifindex ||
1263 dst->ifindex == score->ifa->idev->dev->ifindex);
1264 break;
1265 case IPV6_SADDR_RULE_LABEL:
1266 /* Rule 6: Prefer matching label */
1267 ret = ipv6_addr_label(net,
1268 &score->ifa->addr, score->addr_type,
1269 score->ifa->idev->dev->ifindex) == dst->label;
1270 break;
1271 case IPV6_SADDR_RULE_PRIVACY:
1273 /* Rule 7: Prefer public address
1274 * Note: prefer temporary address if use_tempaddr >= 2
1276 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1277 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1278 score->ifa->idev->cnf.use_tempaddr >= 2;
1279 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1280 break;
1282 case IPV6_SADDR_RULE_ORCHID:
1283 /* Rule 8-: Prefer ORCHID vs ORCHID or
1284 * non-ORCHID vs non-ORCHID
1286 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1287 ipv6_addr_orchid(dst->addr));
1288 break;
1289 case IPV6_SADDR_RULE_PREFIX:
1290 /* Rule 8: Use longest matching prefix */
1291 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1292 if (ret > score->ifa->prefix_len)
1293 ret = score->ifa->prefix_len;
1294 score->matchlen = ret;
1295 break;
1296 default:
1297 ret = 0;
1300 if (ret)
1301 __set_bit(i, score->scorebits);
1302 score->rule = i;
1303 out:
1304 return ret;
1307 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1308 const struct in6_addr *daddr, unsigned int prefs,
1309 struct in6_addr *saddr)
1311 struct ipv6_saddr_score scores[2],
1312 *score = &scores[0], *hiscore = &scores[1];
1313 struct ipv6_saddr_dst dst;
1314 struct net_device *dev;
1315 int dst_type;
1317 dst_type = __ipv6_addr_type(daddr);
1318 dst.addr = daddr;
1319 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1320 dst.scope = __ipv6_addr_src_scope(dst_type);
1321 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1322 dst.prefs = prefs;
1324 hiscore->rule = -1;
1325 hiscore->ifa = NULL;
1327 rcu_read_lock();
1329 for_each_netdev_rcu(net, dev) {
1330 struct inet6_dev *idev;
1332 /* Candidate Source Address (section 4)
1333 * - multicast and link-local destination address,
1334 * the set of candidate source address MUST only
1335 * include addresses assigned to interfaces
1336 * belonging to the same link as the outgoing
1337 * interface.
1338 * (- For site-local destination addresses, the
1339 * set of candidate source addresses MUST only
1340 * include addresses assigned to interfaces
1341 * belonging to the same site as the outgoing
1342 * interface.)
1344 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1345 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1346 dst.ifindex && dev->ifindex != dst.ifindex)
1347 continue;
1349 idev = __in6_dev_get(dev);
1350 if (!idev)
1351 continue;
1353 read_lock_bh(&idev->lock);
1354 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1355 int i;
1358 * - Tentative Address (RFC2462 section 5.4)
1359 * - A tentative address is not considered
1360 * "assigned to an interface" in the traditional
1361 * sense, unless it is also flagged as optimistic.
1362 * - Candidate Source Address (section 4)
1363 * - In any case, anycast addresses, multicast
1364 * addresses, and the unspecified address MUST
1365 * NOT be included in a candidate set.
1367 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1368 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1369 continue;
1371 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1373 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1374 score->addr_type & IPV6_ADDR_MULTICAST)) {
1375 LIMIT_NETDEBUG(KERN_DEBUG
1376 "ADDRCONF: unspecified / multicast address "
1377 "assigned as unicast address on %s",
1378 dev->name);
1379 continue;
1382 score->rule = -1;
1383 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1385 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1386 int minihiscore, miniscore;
1388 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1389 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1391 if (minihiscore > miniscore) {
1392 if (i == IPV6_SADDR_RULE_SCOPE &&
1393 score->scopedist > 0) {
1395 * special case:
1396 * each remaining entry
1397 * has too small (not enough)
1398 * scope, because ifa entries
1399 * are sorted by their scope
1400 * values.
1402 goto try_nextdev;
1404 break;
1405 } else if (minihiscore < miniscore) {
1406 if (hiscore->ifa)
1407 in6_ifa_put(hiscore->ifa);
1409 in6_ifa_hold(score->ifa);
1411 swap(hiscore, score);
1413 /* restore our iterator */
1414 score->ifa = hiscore->ifa;
1416 break;
1420 try_nextdev:
1421 read_unlock_bh(&idev->lock);
1423 rcu_read_unlock();
1425 if (!hiscore->ifa)
1426 return -EADDRNOTAVAIL;
1428 *saddr = hiscore->ifa->addr;
1429 in6_ifa_put(hiscore->ifa);
1430 return 0;
1432 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1434 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1435 u32 banned_flags)
1437 struct inet6_ifaddr *ifp;
1438 int err = -EADDRNOTAVAIL;
1440 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1441 if (ifp->scope > IFA_LINK)
1442 break;
1443 if (ifp->scope == IFA_LINK &&
1444 !(ifp->flags & banned_flags)) {
1445 *addr = ifp->addr;
1446 err = 0;
1447 break;
1450 return err;
1453 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1454 u32 banned_flags)
1456 struct inet6_dev *idev;
1457 int err = -EADDRNOTAVAIL;
1459 rcu_read_lock();
1460 idev = __in6_dev_get(dev);
1461 if (idev) {
1462 read_lock_bh(&idev->lock);
1463 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1464 read_unlock_bh(&idev->lock);
1466 rcu_read_unlock();
1467 return err;
1470 static int ipv6_count_addresses(struct inet6_dev *idev)
1472 int cnt = 0;
1473 struct inet6_ifaddr *ifp;
1475 read_lock_bh(&idev->lock);
1476 list_for_each_entry(ifp, &idev->addr_list, if_list)
1477 cnt++;
1478 read_unlock_bh(&idev->lock);
1479 return cnt;
1482 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1483 const struct net_device *dev, int strict)
1485 struct inet6_ifaddr *ifp;
1486 unsigned int hash = inet6_addr_hash(addr);
1488 rcu_read_lock_bh();
1489 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1490 if (!net_eq(dev_net(ifp->idev->dev), net))
1491 continue;
1492 if (ipv6_addr_equal(&ifp->addr, addr) &&
1493 !(ifp->flags&IFA_F_TENTATIVE) &&
1494 (dev == NULL || ifp->idev->dev == dev ||
1495 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1496 rcu_read_unlock_bh();
1497 return 1;
1501 rcu_read_unlock_bh();
1502 return 0;
1504 EXPORT_SYMBOL(ipv6_chk_addr);
1506 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1507 struct net_device *dev)
1509 unsigned int hash = inet6_addr_hash(addr);
1510 struct inet6_ifaddr *ifp;
1512 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1513 if (!net_eq(dev_net(ifp->idev->dev), net))
1514 continue;
1515 if (ipv6_addr_equal(&ifp->addr, addr)) {
1516 if (dev == NULL || ifp->idev->dev == dev)
1517 return true;
1520 return false;
1523 /* Compares an address/prefix_len with addresses on device @dev.
1524 * If one is found it returns true.
1526 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1527 const unsigned int prefix_len, struct net_device *dev)
1529 struct inet6_dev *idev;
1530 struct inet6_ifaddr *ifa;
1531 bool ret = false;
1533 rcu_read_lock();
1534 idev = __in6_dev_get(dev);
1535 if (idev) {
1536 read_lock_bh(&idev->lock);
1537 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1538 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1539 if (ret)
1540 break;
1542 read_unlock_bh(&idev->lock);
1544 rcu_read_unlock();
1546 return ret;
1548 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1550 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1552 struct inet6_dev *idev;
1553 struct inet6_ifaddr *ifa;
1554 int onlink;
1556 onlink = 0;
1557 rcu_read_lock();
1558 idev = __in6_dev_get(dev);
1559 if (idev) {
1560 read_lock_bh(&idev->lock);
1561 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1562 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1563 ifa->prefix_len);
1564 if (onlink)
1565 break;
1567 read_unlock_bh(&idev->lock);
1569 rcu_read_unlock();
1570 return onlink;
1572 EXPORT_SYMBOL(ipv6_chk_prefix);
1574 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1575 struct net_device *dev, int strict)
1577 struct inet6_ifaddr *ifp, *result = NULL;
1578 unsigned int hash = inet6_addr_hash(addr);
1580 rcu_read_lock_bh();
1581 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1582 if (!net_eq(dev_net(ifp->idev->dev), net))
1583 continue;
1584 if (ipv6_addr_equal(&ifp->addr, addr)) {
1585 if (dev == NULL || ifp->idev->dev == dev ||
1586 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1587 result = ifp;
1588 in6_ifa_hold(ifp);
1589 break;
1593 rcu_read_unlock_bh();
1595 return result;
1598 /* Gets referenced address, destroys ifaddr */
1600 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1602 if (ifp->flags&IFA_F_PERMANENT) {
1603 spin_lock_bh(&ifp->lock);
1604 addrconf_del_dad_timer(ifp);
1605 ifp->flags |= IFA_F_TENTATIVE;
1606 if (dad_failed)
1607 ifp->flags |= IFA_F_DADFAILED;
1608 spin_unlock_bh(&ifp->lock);
1609 if (dad_failed)
1610 ipv6_ifa_notify(0, ifp);
1611 in6_ifa_put(ifp);
1612 } else if (ifp->flags&IFA_F_TEMPORARY) {
1613 struct inet6_ifaddr *ifpub;
1614 spin_lock_bh(&ifp->lock);
1615 ifpub = ifp->ifpub;
1616 if (ifpub) {
1617 in6_ifa_hold(ifpub);
1618 spin_unlock_bh(&ifp->lock);
1619 ipv6_create_tempaddr(ifpub, ifp);
1620 in6_ifa_put(ifpub);
1621 } else {
1622 spin_unlock_bh(&ifp->lock);
1624 ipv6_del_addr(ifp);
1625 } else
1626 ipv6_del_addr(ifp);
1629 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1631 int err = -ENOENT;
1633 spin_lock(&ifp->state_lock);
1634 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1635 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1636 err = 0;
1638 spin_unlock(&ifp->state_lock);
1640 return err;
1643 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1645 struct inet6_dev *idev = ifp->idev;
1647 if (addrconf_dad_end(ifp)) {
1648 in6_ifa_put(ifp);
1649 return;
1652 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1653 ifp->idev->dev->name, &ifp->addr);
1655 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1656 struct in6_addr addr;
1658 addr.s6_addr32[0] = htonl(0xfe800000);
1659 addr.s6_addr32[1] = 0;
1661 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1662 ipv6_addr_equal(&ifp->addr, &addr)) {
1663 /* DAD failed for link-local based on MAC address */
1664 idev->cnf.disable_ipv6 = 1;
1666 pr_info("%s: IPv6 being disabled!\n",
1667 ifp->idev->dev->name);
1671 addrconf_dad_stop(ifp, 1);
1674 /* Join to solicited addr multicast group. */
1676 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1678 struct in6_addr maddr;
1680 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1681 return;
1683 addrconf_addr_solict_mult(addr, &maddr);
1684 ipv6_dev_mc_inc(dev, &maddr);
1687 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1689 struct in6_addr maddr;
1691 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1692 return;
1694 addrconf_addr_solict_mult(addr, &maddr);
1695 __ipv6_dev_mc_dec(idev, &maddr);
1698 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1700 struct in6_addr addr;
1701 if (ifp->prefix_len >= 127) /* RFC 6164 */
1702 return;
1703 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1704 if (ipv6_addr_any(&addr))
1705 return;
1706 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1709 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1711 struct in6_addr addr;
1712 if (ifp->prefix_len >= 127) /* RFC 6164 */
1713 return;
1714 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1715 if (ipv6_addr_any(&addr))
1716 return;
1717 __ipv6_dev_ac_dec(ifp->idev, &addr);
1720 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1722 if (dev->addr_len != ETH_ALEN)
1723 return -1;
1724 memcpy(eui, dev->dev_addr, 3);
1725 memcpy(eui + 5, dev->dev_addr + 3, 3);
1728 * The zSeries OSA network cards can be shared among various
1729 * OS instances, but the OSA cards have only one MAC address.
1730 * This leads to duplicate address conflicts in conjunction
1731 * with IPv6 if more than one instance uses the same card.
1733 * The driver for these cards can deliver a unique 16-bit
1734 * identifier for each instance sharing the same card. It is
1735 * placed instead of 0xFFFE in the interface identifier. The
1736 * "u" bit of the interface identifier is not inverted in this
1737 * case. Hence the resulting interface identifier has local
1738 * scope according to RFC2373.
1740 if (dev->dev_id) {
1741 eui[3] = (dev->dev_id >> 8) & 0xFF;
1742 eui[4] = dev->dev_id & 0xFF;
1743 } else {
1744 eui[3] = 0xFF;
1745 eui[4] = 0xFE;
1746 eui[0] ^= 2;
1748 return 0;
1751 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1753 if (dev->addr_len != IEEE802154_ADDR_LEN)
1754 return -1;
1755 memcpy(eui, dev->dev_addr, 8);
1756 eui[0] ^= 2;
1757 return 0;
1760 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1762 union fwnet_hwaddr *ha;
1764 if (dev->addr_len != FWNET_ALEN)
1765 return -1;
1767 ha = (union fwnet_hwaddr *)dev->dev_addr;
1769 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1770 eui[0] ^= 2;
1771 return 0;
1774 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1776 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1777 if (dev->addr_len != ARCNET_ALEN)
1778 return -1;
1779 memset(eui, 0, 7);
1780 eui[7] = *(u8 *)dev->dev_addr;
1781 return 0;
1784 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1786 if (dev->addr_len != INFINIBAND_ALEN)
1787 return -1;
1788 memcpy(eui, dev->dev_addr + 12, 8);
1789 eui[0] |= 2;
1790 return 0;
1793 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1795 if (addr == 0)
1796 return -1;
1797 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1798 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1799 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1800 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1801 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1802 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1803 eui[1] = 0;
1804 eui[2] = 0x5E;
1805 eui[3] = 0xFE;
1806 memcpy(eui + 4, &addr, 4);
1807 return 0;
1810 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1812 if (dev->priv_flags & IFF_ISATAP)
1813 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1814 return -1;
1817 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1819 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1822 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1824 memcpy(eui, dev->perm_addr, 3);
1825 memcpy(eui + 5, dev->perm_addr + 3, 3);
1826 eui[3] = 0xFF;
1827 eui[4] = 0xFE;
1828 eui[0] ^= 2;
1829 return 0;
1832 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1834 switch (dev->type) {
1835 case ARPHRD_ETHER:
1836 case ARPHRD_FDDI:
1837 return addrconf_ifid_eui48(eui, dev);
1838 case ARPHRD_ARCNET:
1839 return addrconf_ifid_arcnet(eui, dev);
1840 case ARPHRD_INFINIBAND:
1841 return addrconf_ifid_infiniband(eui, dev);
1842 case ARPHRD_SIT:
1843 return addrconf_ifid_sit(eui, dev);
1844 case ARPHRD_IPGRE:
1845 return addrconf_ifid_gre(eui, dev);
1846 case ARPHRD_6LOWPAN:
1847 case ARPHRD_IEEE802154:
1848 return addrconf_ifid_eui64(eui, dev);
1849 case ARPHRD_IEEE1394:
1850 return addrconf_ifid_ieee1394(eui, dev);
1851 case ARPHRD_TUNNEL6:
1852 return addrconf_ifid_ip6tnl(eui, dev);
1854 return -1;
1857 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1859 int err = -1;
1860 struct inet6_ifaddr *ifp;
1862 read_lock_bh(&idev->lock);
1863 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1864 if (ifp->scope > IFA_LINK)
1865 break;
1866 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1867 memcpy(eui, ifp->addr.s6_addr+8, 8);
1868 err = 0;
1869 break;
1872 read_unlock_bh(&idev->lock);
1873 return err;
1876 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1877 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1879 regen:
1880 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1881 idev->rndid[0] &= ~0x02;
1884 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1885 * check if generated address is not inappropriate
1887 * - Reserved subnet anycast (RFC 2526)
1888 * 11111101 11....11 1xxxxxxx
1889 * - ISATAP (RFC4214) 6.1
1890 * 00-00-5E-FE-xx-xx-xx-xx
1891 * - value 0
1892 * - XXX: already assigned to an address on the device
1894 if (idev->rndid[0] == 0xfd &&
1895 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1896 (idev->rndid[7]&0x80))
1897 goto regen;
1898 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1899 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1900 goto regen;
1901 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1902 goto regen;
1906 static void ipv6_regen_rndid(unsigned long data)
1908 struct inet6_dev *idev = (struct inet6_dev *) data;
1909 unsigned long expires;
1911 rcu_read_lock_bh();
1912 write_lock_bh(&idev->lock);
1914 if (idev->dead)
1915 goto out;
1917 __ipv6_regen_rndid(idev);
1919 expires = jiffies +
1920 idev->cnf.temp_prefered_lft * HZ -
1921 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1922 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1923 idev->cnf.max_desync_factor * HZ;
1924 if (time_before(expires, jiffies)) {
1925 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1926 __func__, idev->dev->name);
1927 goto out;
1930 if (!mod_timer(&idev->regen_timer, expires))
1931 in6_dev_hold(idev);
1933 out:
1934 write_unlock_bh(&idev->lock);
1935 rcu_read_unlock_bh();
1936 in6_dev_put(idev);
1939 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1941 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1942 __ipv6_regen_rndid(idev);
1946 * Add prefix route.
1949 static void
1950 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1951 unsigned long expires, u32 flags)
1953 struct fib6_config cfg = {
1954 .fc_table = RT6_TABLE_PREFIX,
1955 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1956 .fc_ifindex = dev->ifindex,
1957 .fc_expires = expires,
1958 .fc_dst_len = plen,
1959 .fc_flags = RTF_UP | flags,
1960 .fc_nlinfo.nl_net = dev_net(dev),
1961 .fc_protocol = RTPROT_KERNEL,
1964 cfg.fc_dst = *pfx;
1966 /* Prevent useless cloning on PtP SIT.
1967 This thing is done here expecting that the whole
1968 class of non-broadcast devices need not cloning.
1970 #if IS_ENABLED(CONFIG_IPV6_SIT)
1971 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1972 cfg.fc_flags |= RTF_NONEXTHOP;
1973 #endif
1975 ip6_route_add(&cfg);
1979 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1980 int plen,
1981 const struct net_device *dev,
1982 u32 flags, u32 noflags)
1984 struct fib6_node *fn;
1985 struct rt6_info *rt = NULL;
1986 struct fib6_table *table;
1988 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1989 if (table == NULL)
1990 return NULL;
1992 read_lock_bh(&table->tb6_lock);
1993 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1994 if (!fn)
1995 goto out;
1996 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1997 if (rt->dst.dev->ifindex != dev->ifindex)
1998 continue;
1999 if ((rt->rt6i_flags & flags) != flags)
2000 continue;
2001 if ((rt->rt6i_flags & noflags) != 0)
2002 continue;
2003 dst_hold(&rt->dst);
2004 break;
2006 out:
2007 read_unlock_bh(&table->tb6_lock);
2008 return rt;
2012 /* Create "default" multicast route to the interface */
2014 static void addrconf_add_mroute(struct net_device *dev)
2016 struct fib6_config cfg = {
2017 .fc_table = RT6_TABLE_LOCAL,
2018 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2019 .fc_ifindex = dev->ifindex,
2020 .fc_dst_len = 8,
2021 .fc_flags = RTF_UP,
2022 .fc_nlinfo.nl_net = dev_net(dev),
2025 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2027 ip6_route_add(&cfg);
2030 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2032 struct inet6_dev *idev;
2034 ASSERT_RTNL();
2036 idev = ipv6_find_idev(dev);
2037 if (!idev)
2038 return ERR_PTR(-ENOBUFS);
2040 if (idev->cnf.disable_ipv6)
2041 return ERR_PTR(-EACCES);
2043 /* Add default multicast route */
2044 if (!(dev->flags & IFF_LOOPBACK))
2045 addrconf_add_mroute(dev);
2047 return idev;
2050 static void manage_tempaddrs(struct inet6_dev *idev,
2051 struct inet6_ifaddr *ifp,
2052 __u32 valid_lft, __u32 prefered_lft,
2053 bool create, unsigned long now)
2055 u32 flags;
2056 struct inet6_ifaddr *ift;
2058 read_lock_bh(&idev->lock);
2059 /* update all temporary addresses in the list */
2060 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2061 int age, max_valid, max_prefered;
2063 if (ifp != ift->ifpub)
2064 continue;
2066 /* RFC 4941 section 3.3:
2067 * If a received option will extend the lifetime of a public
2068 * address, the lifetimes of temporary addresses should
2069 * be extended, subject to the overall constraint that no
2070 * temporary addresses should ever remain "valid" or "preferred"
2071 * for a time longer than (TEMP_VALID_LIFETIME) or
2072 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2074 age = (now - ift->cstamp) / HZ;
2075 max_valid = idev->cnf.temp_valid_lft - age;
2076 if (max_valid < 0)
2077 max_valid = 0;
2079 max_prefered = idev->cnf.temp_prefered_lft -
2080 idev->cnf.max_desync_factor - age;
2081 if (max_prefered < 0)
2082 max_prefered = 0;
2084 if (valid_lft > max_valid)
2085 valid_lft = max_valid;
2087 if (prefered_lft > max_prefered)
2088 prefered_lft = max_prefered;
2090 spin_lock(&ift->lock);
2091 flags = ift->flags;
2092 ift->valid_lft = valid_lft;
2093 ift->prefered_lft = prefered_lft;
2094 ift->tstamp = now;
2095 if (prefered_lft > 0)
2096 ift->flags &= ~IFA_F_DEPRECATED;
2098 spin_unlock(&ift->lock);
2099 if (!(flags&IFA_F_TENTATIVE))
2100 ipv6_ifa_notify(0, ift);
2103 if ((create || list_empty(&idev->tempaddr_list)) &&
2104 idev->cnf.use_tempaddr > 0) {
2105 /* When a new public address is created as described
2106 * in [ADDRCONF], also create a new temporary address.
2107 * Also create a temporary address if it's enabled but
2108 * no temporary address currently exists.
2110 read_unlock_bh(&idev->lock);
2111 ipv6_create_tempaddr(ifp, NULL);
2112 } else {
2113 read_unlock_bh(&idev->lock);
2117 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2119 struct prefix_info *pinfo;
2120 __u32 valid_lft;
2121 __u32 prefered_lft;
2122 int addr_type;
2123 struct inet6_dev *in6_dev;
2124 struct net *net = dev_net(dev);
2126 pinfo = (struct prefix_info *) opt;
2128 if (len < sizeof(struct prefix_info)) {
2129 ADBG("addrconf: prefix option too short\n");
2130 return;
2134 * Validation checks ([ADDRCONF], page 19)
2137 addr_type = ipv6_addr_type(&pinfo->prefix);
2139 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2140 return;
2142 valid_lft = ntohl(pinfo->valid);
2143 prefered_lft = ntohl(pinfo->prefered);
2145 if (prefered_lft > valid_lft) {
2146 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2147 return;
2150 in6_dev = in6_dev_get(dev);
2152 if (in6_dev == NULL) {
2153 net_dbg_ratelimited("addrconf: device %s not configured\n",
2154 dev->name);
2155 return;
2159 * Two things going on here:
2160 * 1) Add routes for on-link prefixes
2161 * 2) Configure prefixes with the auto flag set
2164 if (pinfo->onlink) {
2165 struct rt6_info *rt;
2166 unsigned long rt_expires;
2168 /* Avoid arithmetic overflow. Really, we could
2169 * save rt_expires in seconds, likely valid_lft,
2170 * but it would require division in fib gc, that it
2171 * not good.
2173 if (HZ > USER_HZ)
2174 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2175 else
2176 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2178 if (addrconf_finite_timeout(rt_expires))
2179 rt_expires *= HZ;
2181 rt = addrconf_get_prefix_route(&pinfo->prefix,
2182 pinfo->prefix_len,
2183 dev,
2184 RTF_ADDRCONF | RTF_PREFIX_RT,
2185 RTF_GATEWAY | RTF_DEFAULT);
2187 if (rt) {
2188 /* Autoconf prefix route */
2189 if (valid_lft == 0) {
2190 ip6_del_rt(rt);
2191 rt = NULL;
2192 } else if (addrconf_finite_timeout(rt_expires)) {
2193 /* not infinity */
2194 rt6_set_expires(rt, jiffies + rt_expires);
2195 } else {
2196 rt6_clean_expires(rt);
2198 } else if (valid_lft) {
2199 clock_t expires = 0;
2200 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2201 if (addrconf_finite_timeout(rt_expires)) {
2202 /* not infinity */
2203 flags |= RTF_EXPIRES;
2204 expires = jiffies_to_clock_t(rt_expires);
2206 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2207 dev, expires, flags);
2209 ip6_rt_put(rt);
2212 /* Try to figure out our local address for this prefix */
2214 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2215 struct inet6_ifaddr *ifp;
2216 struct in6_addr addr;
2217 int create = 0, update_lft = 0;
2218 bool tokenized = false;
2220 if (pinfo->prefix_len == 64) {
2221 memcpy(&addr, &pinfo->prefix, 8);
2223 if (!ipv6_addr_any(&in6_dev->token)) {
2224 read_lock_bh(&in6_dev->lock);
2225 memcpy(addr.s6_addr + 8,
2226 in6_dev->token.s6_addr + 8, 8);
2227 read_unlock_bh(&in6_dev->lock);
2228 tokenized = true;
2229 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2230 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2231 in6_dev_put(in6_dev);
2232 return;
2234 goto ok;
2236 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2237 pinfo->prefix_len);
2238 in6_dev_put(in6_dev);
2239 return;
2243 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2245 if (ifp == NULL && valid_lft) {
2246 int max_addresses = in6_dev->cnf.max_addresses;
2247 u32 addr_flags = 0;
2249 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2250 if (in6_dev->cnf.optimistic_dad &&
2251 !net->ipv6.devconf_all->forwarding && sllao)
2252 addr_flags = IFA_F_OPTIMISTIC;
2253 #endif
2255 /* Do not allow to create too much of autoconfigured
2256 * addresses; this would be too easy way to crash kernel.
2258 if (!max_addresses ||
2259 ipv6_count_addresses(in6_dev) < max_addresses)
2260 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2261 pinfo->prefix_len,
2262 addr_type&IPV6_ADDR_SCOPE_MASK,
2263 addr_flags, valid_lft,
2264 prefered_lft);
2266 if (IS_ERR_OR_NULL(ifp)) {
2267 in6_dev_put(in6_dev);
2268 return;
2271 ifp->flags |= IFA_F_MANAGETEMPADDR;
2272 update_lft = 0;
2273 create = 1;
2274 ifp->cstamp = jiffies;
2275 ifp->tokenized = tokenized;
2276 addrconf_dad_start(ifp);
2279 if (ifp) {
2280 u32 flags;
2281 unsigned long now;
2282 u32 stored_lft;
2284 /* update lifetime (RFC2462 5.5.3 e) */
2285 spin_lock(&ifp->lock);
2286 now = jiffies;
2287 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2288 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2289 else
2290 stored_lft = 0;
2291 if (!update_lft && !create && stored_lft) {
2292 const u32 minimum_lft = min(
2293 stored_lft, (u32)MIN_VALID_LIFETIME);
2294 valid_lft = max(valid_lft, minimum_lft);
2296 /* RFC4862 Section 5.5.3e:
2297 * "Note that the preferred lifetime of the
2298 * corresponding address is always reset to
2299 * the Preferred Lifetime in the received
2300 * Prefix Information option, regardless of
2301 * whether the valid lifetime is also reset or
2302 * ignored."
2304 * So we should always update prefered_lft here.
2306 update_lft = 1;
2309 if (update_lft) {
2310 ifp->valid_lft = valid_lft;
2311 ifp->prefered_lft = prefered_lft;
2312 ifp->tstamp = now;
2313 flags = ifp->flags;
2314 ifp->flags &= ~IFA_F_DEPRECATED;
2315 spin_unlock(&ifp->lock);
2317 if (!(flags&IFA_F_TENTATIVE))
2318 ipv6_ifa_notify(0, ifp);
2319 } else
2320 spin_unlock(&ifp->lock);
2322 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2323 create, now);
2325 in6_ifa_put(ifp);
2326 addrconf_verify(0);
2329 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2330 in6_dev_put(in6_dev);
2334 * Set destination address.
2335 * Special case for SIT interfaces where we create a new "virtual"
2336 * device.
2338 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2340 struct in6_ifreq ireq;
2341 struct net_device *dev;
2342 int err = -EINVAL;
2344 rtnl_lock();
2346 err = -EFAULT;
2347 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2348 goto err_exit;
2350 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2352 err = -ENODEV;
2353 if (dev == NULL)
2354 goto err_exit;
2356 #if IS_ENABLED(CONFIG_IPV6_SIT)
2357 if (dev->type == ARPHRD_SIT) {
2358 const struct net_device_ops *ops = dev->netdev_ops;
2359 struct ifreq ifr;
2360 struct ip_tunnel_parm p;
2362 err = -EADDRNOTAVAIL;
2363 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2364 goto err_exit;
2366 memset(&p, 0, sizeof(p));
2367 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2368 p.iph.saddr = 0;
2369 p.iph.version = 4;
2370 p.iph.ihl = 5;
2371 p.iph.protocol = IPPROTO_IPV6;
2372 p.iph.ttl = 64;
2373 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2375 if (ops->ndo_do_ioctl) {
2376 mm_segment_t oldfs = get_fs();
2378 set_fs(KERNEL_DS);
2379 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2380 set_fs(oldfs);
2381 } else
2382 err = -EOPNOTSUPP;
2384 if (err == 0) {
2385 err = -ENOBUFS;
2386 dev = __dev_get_by_name(net, p.name);
2387 if (!dev)
2388 goto err_exit;
2389 err = dev_open(dev);
2392 #endif
2394 err_exit:
2395 rtnl_unlock();
2396 return err;
2400 * Manual configuration of address on an interface
2402 static int inet6_addr_add(struct net *net, int ifindex,
2403 const struct in6_addr *pfx,
2404 const struct in6_addr *peer_pfx,
2405 unsigned int plen, __u32 ifa_flags,
2406 __u32 prefered_lft, __u32 valid_lft)
2408 struct inet6_ifaddr *ifp;
2409 struct inet6_dev *idev;
2410 struct net_device *dev;
2411 int scope;
2412 u32 flags;
2413 clock_t expires;
2414 unsigned long timeout;
2416 ASSERT_RTNL();
2418 if (plen > 128)
2419 return -EINVAL;
2421 /* check the lifetime */
2422 if (!valid_lft || prefered_lft > valid_lft)
2423 return -EINVAL;
2425 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2426 return -EINVAL;
2428 dev = __dev_get_by_index(net, ifindex);
2429 if (!dev)
2430 return -ENODEV;
2432 idev = addrconf_add_dev(dev);
2433 if (IS_ERR(idev))
2434 return PTR_ERR(idev);
2436 scope = ipv6_addr_scope(pfx);
2438 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2439 if (addrconf_finite_timeout(timeout)) {
2440 expires = jiffies_to_clock_t(timeout * HZ);
2441 valid_lft = timeout;
2442 flags = RTF_EXPIRES;
2443 } else {
2444 expires = 0;
2445 flags = 0;
2446 ifa_flags |= IFA_F_PERMANENT;
2449 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2450 if (addrconf_finite_timeout(timeout)) {
2451 if (timeout == 0)
2452 ifa_flags |= IFA_F_DEPRECATED;
2453 prefered_lft = timeout;
2456 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2457 valid_lft, prefered_lft);
2459 if (!IS_ERR(ifp)) {
2460 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2461 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2462 expires, flags);
2466 * Note that section 3.1 of RFC 4429 indicates
2467 * that the Optimistic flag should not be set for
2468 * manually configured addresses
2470 addrconf_dad_start(ifp);
2471 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2472 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2473 true, jiffies);
2474 in6_ifa_put(ifp);
2475 addrconf_verify(0);
2476 return 0;
2479 return PTR_ERR(ifp);
2482 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2483 unsigned int plen)
2485 struct inet6_ifaddr *ifp;
2486 struct inet6_dev *idev;
2487 struct net_device *dev;
2489 if (plen > 128)
2490 return -EINVAL;
2492 dev = __dev_get_by_index(net, ifindex);
2493 if (!dev)
2494 return -ENODEV;
2496 if ((idev = __in6_dev_get(dev)) == NULL)
2497 return -ENXIO;
2499 read_lock_bh(&idev->lock);
2500 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2501 if (ifp->prefix_len == plen &&
2502 ipv6_addr_equal(pfx, &ifp->addr)) {
2503 in6_ifa_hold(ifp);
2504 read_unlock_bh(&idev->lock);
2506 ipv6_del_addr(ifp);
2507 return 0;
2510 read_unlock_bh(&idev->lock);
2511 return -EADDRNOTAVAIL;
2515 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2517 struct in6_ifreq ireq;
2518 int err;
2520 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2521 return -EPERM;
2523 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2524 return -EFAULT;
2526 rtnl_lock();
2527 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2528 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2529 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2530 rtnl_unlock();
2531 return err;
2534 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2536 struct in6_ifreq ireq;
2537 int err;
2539 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2540 return -EPERM;
2542 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2543 return -EFAULT;
2545 rtnl_lock();
2546 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2547 ireq.ifr6_prefixlen);
2548 rtnl_unlock();
2549 return err;
2552 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2553 int plen, int scope)
2555 struct inet6_ifaddr *ifp;
2557 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2558 scope, IFA_F_PERMANENT,
2559 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2560 if (!IS_ERR(ifp)) {
2561 spin_lock_bh(&ifp->lock);
2562 ifp->flags &= ~IFA_F_TENTATIVE;
2563 spin_unlock_bh(&ifp->lock);
2564 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2565 in6_ifa_put(ifp);
2569 #if IS_ENABLED(CONFIG_IPV6_SIT)
2570 static void sit_add_v4_addrs(struct inet6_dev *idev)
2572 struct in6_addr addr;
2573 struct net_device *dev;
2574 struct net *net = dev_net(idev->dev);
2575 int scope, plen;
2576 u32 pflags = 0;
2578 ASSERT_RTNL();
2580 memset(&addr, 0, sizeof(struct in6_addr));
2581 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2583 if (idev->dev->flags&IFF_POINTOPOINT) {
2584 addr.s6_addr32[0] = htonl(0xfe800000);
2585 scope = IFA_LINK;
2586 plen = 64;
2587 } else {
2588 scope = IPV6_ADDR_COMPATv4;
2589 plen = 96;
2590 pflags |= RTF_NONEXTHOP;
2593 if (addr.s6_addr32[3]) {
2594 add_addr(idev, &addr, plen, scope);
2595 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2596 return;
2599 for_each_netdev(net, dev) {
2600 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2601 if (in_dev && (dev->flags & IFF_UP)) {
2602 struct in_ifaddr *ifa;
2604 int flag = scope;
2606 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2608 addr.s6_addr32[3] = ifa->ifa_local;
2610 if (ifa->ifa_scope == RT_SCOPE_LINK)
2611 continue;
2612 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2613 if (idev->dev->flags&IFF_POINTOPOINT)
2614 continue;
2615 flag |= IFA_HOST;
2618 add_addr(idev, &addr, plen, flag);
2619 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2620 pflags);
2625 #endif
2627 static void init_loopback(struct net_device *dev)
2629 struct inet6_dev *idev;
2630 struct net_device *sp_dev;
2631 struct inet6_ifaddr *sp_ifa;
2632 struct rt6_info *sp_rt;
2634 /* ::1 */
2636 ASSERT_RTNL();
2638 if ((idev = ipv6_find_idev(dev)) == NULL) {
2639 pr_debug("%s: add_dev failed\n", __func__);
2640 return;
2643 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2645 /* Add routes to other interface's IPv6 addresses */
2646 for_each_netdev(dev_net(dev), sp_dev) {
2647 if (!strcmp(sp_dev->name, dev->name))
2648 continue;
2650 idev = __in6_dev_get(sp_dev);
2651 if (!idev)
2652 continue;
2654 read_lock_bh(&idev->lock);
2655 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2657 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2658 continue;
2660 if (sp_ifa->rt) {
2661 /* This dst has been added to garbage list when
2662 * lo device down, release this obsolete dst and
2663 * reallocate a new router for ifa.
2665 if (sp_ifa->rt->dst.obsolete > 0) {
2666 ip6_rt_put(sp_ifa->rt);
2667 sp_ifa->rt = NULL;
2668 } else {
2669 continue;
2673 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2675 /* Failure cases are ignored */
2676 if (!IS_ERR(sp_rt)) {
2677 sp_ifa->rt = sp_rt;
2678 ip6_ins_rt(sp_rt);
2681 read_unlock_bh(&idev->lock);
2685 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2687 struct inet6_ifaddr *ifp;
2688 u32 addr_flags = IFA_F_PERMANENT;
2690 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2691 if (idev->cnf.optimistic_dad &&
2692 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2693 addr_flags |= IFA_F_OPTIMISTIC;
2694 #endif
2697 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2698 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2699 if (!IS_ERR(ifp)) {
2700 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2701 addrconf_dad_start(ifp);
2702 in6_ifa_put(ifp);
2706 static void addrconf_dev_config(struct net_device *dev)
2708 struct in6_addr addr;
2709 struct inet6_dev *idev;
2711 ASSERT_RTNL();
2713 if ((dev->type != ARPHRD_ETHER) &&
2714 (dev->type != ARPHRD_FDDI) &&
2715 (dev->type != ARPHRD_ARCNET) &&
2716 (dev->type != ARPHRD_INFINIBAND) &&
2717 (dev->type != ARPHRD_IEEE802154) &&
2718 (dev->type != ARPHRD_IEEE1394) &&
2719 (dev->type != ARPHRD_TUNNEL6) &&
2720 (dev->type != ARPHRD_6LOWPAN)) {
2721 /* Alas, we support only Ethernet autoconfiguration. */
2722 return;
2725 idev = addrconf_add_dev(dev);
2726 if (IS_ERR(idev))
2727 return;
2729 memset(&addr, 0, sizeof(struct in6_addr));
2730 addr.s6_addr32[0] = htonl(0xFE800000);
2732 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2733 addrconf_add_linklocal(idev, &addr);
2736 #if IS_ENABLED(CONFIG_IPV6_SIT)
2737 static void addrconf_sit_config(struct net_device *dev)
2739 struct inet6_dev *idev;
2741 ASSERT_RTNL();
2744 * Configure the tunnel with one of our IPv4
2745 * addresses... we should configure all of
2746 * our v4 addrs in the tunnel
2749 if ((idev = ipv6_find_idev(dev)) == NULL) {
2750 pr_debug("%s: add_dev failed\n", __func__);
2751 return;
2754 if (dev->priv_flags & IFF_ISATAP) {
2755 struct in6_addr addr;
2757 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2758 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2759 addrconf_add_linklocal(idev, &addr);
2760 return;
2763 sit_add_v4_addrs(idev);
2765 if (dev->flags&IFF_POINTOPOINT)
2766 addrconf_add_mroute(dev);
2768 #endif
2770 #if IS_ENABLED(CONFIG_NET_IPGRE)
2771 static void addrconf_gre_config(struct net_device *dev)
2773 struct inet6_dev *idev;
2774 struct in6_addr addr;
2776 ASSERT_RTNL();
2778 if ((idev = ipv6_find_idev(dev)) == NULL) {
2779 pr_debug("%s: add_dev failed\n", __func__);
2780 return;
2783 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2784 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2785 addrconf_add_linklocal(idev, &addr);
2786 else
2787 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2789 #endif
2791 static inline int
2792 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2794 struct in6_addr lladdr;
2796 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2797 addrconf_add_linklocal(idev, &lladdr);
2798 return 0;
2800 return -1;
2803 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2804 void *ptr)
2806 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2807 struct inet6_dev *idev = __in6_dev_get(dev);
2808 int run_pending = 0;
2809 int err;
2811 switch (event) {
2812 case NETDEV_REGISTER:
2813 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2814 idev = ipv6_add_dev(dev);
2815 if (!idev)
2816 return notifier_from_errno(-ENOMEM);
2818 break;
2820 case NETDEV_UP:
2821 case NETDEV_CHANGE:
2822 if (dev->flags & IFF_SLAVE)
2823 break;
2825 if (event == NETDEV_UP) {
2826 if (!addrconf_qdisc_ok(dev)) {
2827 /* device is not ready yet. */
2828 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2829 dev->name);
2830 break;
2833 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2834 idev = ipv6_add_dev(dev);
2836 if (idev) {
2837 idev->if_flags |= IF_READY;
2838 run_pending = 1;
2840 } else {
2841 if (!addrconf_qdisc_ok(dev)) {
2842 /* device is still not ready. */
2843 break;
2846 if (idev) {
2847 if (idev->if_flags & IF_READY)
2848 /* device is already configured. */
2849 break;
2850 idev->if_flags |= IF_READY;
2853 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2854 dev->name);
2856 run_pending = 1;
2859 switch (dev->type) {
2860 #if IS_ENABLED(CONFIG_IPV6_SIT)
2861 case ARPHRD_SIT:
2862 addrconf_sit_config(dev);
2863 break;
2864 #endif
2865 #if IS_ENABLED(CONFIG_NET_IPGRE)
2866 case ARPHRD_IPGRE:
2867 addrconf_gre_config(dev);
2868 break;
2869 #endif
2870 case ARPHRD_LOOPBACK:
2871 init_loopback(dev);
2872 break;
2874 default:
2875 addrconf_dev_config(dev);
2876 break;
2879 if (idev) {
2880 if (run_pending)
2881 addrconf_dad_run(idev);
2884 * If the MTU changed during the interface down,
2885 * when the interface up, the changed MTU must be
2886 * reflected in the idev as well as routers.
2888 if (idev->cnf.mtu6 != dev->mtu &&
2889 dev->mtu >= IPV6_MIN_MTU) {
2890 rt6_mtu_change(dev, dev->mtu);
2891 idev->cnf.mtu6 = dev->mtu;
2893 idev->tstamp = jiffies;
2894 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2897 * If the changed mtu during down is lower than
2898 * IPV6_MIN_MTU stop IPv6 on this interface.
2900 if (dev->mtu < IPV6_MIN_MTU)
2901 addrconf_ifdown(dev, 1);
2903 break;
2905 case NETDEV_CHANGEMTU:
2906 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2907 rt6_mtu_change(dev, dev->mtu);
2908 idev->cnf.mtu6 = dev->mtu;
2909 break;
2912 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2913 idev = ipv6_add_dev(dev);
2914 if (idev)
2915 break;
2919 * if MTU under IPV6_MIN_MTU.
2920 * Stop IPv6 on this interface.
2923 case NETDEV_DOWN:
2924 case NETDEV_UNREGISTER:
2926 * Remove all addresses from this interface.
2928 addrconf_ifdown(dev, event != NETDEV_DOWN);
2929 break;
2931 case NETDEV_CHANGENAME:
2932 if (idev) {
2933 snmp6_unregister_dev(idev);
2934 addrconf_sysctl_unregister(idev);
2935 addrconf_sysctl_register(idev);
2936 err = snmp6_register_dev(idev);
2937 if (err)
2938 return notifier_from_errno(err);
2940 break;
2942 case NETDEV_PRE_TYPE_CHANGE:
2943 case NETDEV_POST_TYPE_CHANGE:
2944 addrconf_type_change(dev, event);
2945 break;
2948 return NOTIFY_OK;
2952 * addrconf module should be notified of a device going up
2954 static struct notifier_block ipv6_dev_notf = {
2955 .notifier_call = addrconf_notify,
2958 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2960 struct inet6_dev *idev;
2961 ASSERT_RTNL();
2963 idev = __in6_dev_get(dev);
2965 if (event == NETDEV_POST_TYPE_CHANGE)
2966 ipv6_mc_remap(idev);
2967 else if (event == NETDEV_PRE_TYPE_CHANGE)
2968 ipv6_mc_unmap(idev);
2971 static int addrconf_ifdown(struct net_device *dev, int how)
2973 struct net *net = dev_net(dev);
2974 struct inet6_dev *idev;
2975 struct inet6_ifaddr *ifa;
2976 int state, i;
2978 ASSERT_RTNL();
2980 rt6_ifdown(net, dev);
2981 neigh_ifdown(&nd_tbl, dev);
2983 idev = __in6_dev_get(dev);
2984 if (idev == NULL)
2985 return -ENODEV;
2988 * Step 1: remove reference to ipv6 device from parent device.
2989 * Do not dev_put!
2991 if (how) {
2992 idev->dead = 1;
2994 /* protected by rtnl_lock */
2995 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2997 /* Step 1.5: remove snmp6 entry */
2998 snmp6_unregister_dev(idev);
3002 /* Step 2: clear hash table */
3003 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3004 struct hlist_head *h = &inet6_addr_lst[i];
3006 spin_lock_bh(&addrconf_hash_lock);
3007 restart:
3008 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3009 if (ifa->idev == idev) {
3010 hlist_del_init_rcu(&ifa->addr_lst);
3011 addrconf_del_dad_timer(ifa);
3012 goto restart;
3015 spin_unlock_bh(&addrconf_hash_lock);
3018 write_lock_bh(&idev->lock);
3020 addrconf_del_rs_timer(idev);
3022 /* Step 2: clear flags for stateless addrconf */
3023 if (!how)
3024 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3026 if (how && del_timer(&idev->regen_timer))
3027 in6_dev_put(idev);
3029 /* Step 3: clear tempaddr list */
3030 while (!list_empty(&idev->tempaddr_list)) {
3031 ifa = list_first_entry(&idev->tempaddr_list,
3032 struct inet6_ifaddr, tmp_list);
3033 list_del(&ifa->tmp_list);
3034 write_unlock_bh(&idev->lock);
3035 spin_lock_bh(&ifa->lock);
3037 if (ifa->ifpub) {
3038 in6_ifa_put(ifa->ifpub);
3039 ifa->ifpub = NULL;
3041 spin_unlock_bh(&ifa->lock);
3042 in6_ifa_put(ifa);
3043 write_lock_bh(&idev->lock);
3046 while (!list_empty(&idev->addr_list)) {
3047 ifa = list_first_entry(&idev->addr_list,
3048 struct inet6_ifaddr, if_list);
3049 addrconf_del_dad_timer(ifa);
3051 list_del(&ifa->if_list);
3053 write_unlock_bh(&idev->lock);
3055 spin_lock_bh(&ifa->state_lock);
3056 state = ifa->state;
3057 ifa->state = INET6_IFADDR_STATE_DEAD;
3058 spin_unlock_bh(&ifa->state_lock);
3060 if (state != INET6_IFADDR_STATE_DEAD) {
3061 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3062 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3064 in6_ifa_put(ifa);
3066 write_lock_bh(&idev->lock);
3069 write_unlock_bh(&idev->lock);
3071 /* Step 5: Discard multicast list */
3072 if (how)
3073 ipv6_mc_destroy_dev(idev);
3074 else
3075 ipv6_mc_down(idev);
3077 idev->tstamp = jiffies;
3079 /* Last: Shot the device (if unregistered) */
3080 if (how) {
3081 addrconf_sysctl_unregister(idev);
3082 neigh_parms_release(&nd_tbl, idev->nd_parms);
3083 neigh_ifdown(&nd_tbl, dev);
3084 in6_dev_put(idev);
3086 return 0;
3089 static void addrconf_rs_timer(unsigned long data)
3091 struct inet6_dev *idev = (struct inet6_dev *)data;
3092 struct net_device *dev = idev->dev;
3093 struct in6_addr lladdr;
3095 write_lock(&idev->lock);
3096 if (idev->dead || !(idev->if_flags & IF_READY))
3097 goto out;
3099 if (!ipv6_accept_ra(idev))
3100 goto out;
3102 /* Announcement received after solicitation was sent */
3103 if (idev->if_flags & IF_RA_RCVD)
3104 goto out;
3106 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3107 write_unlock(&idev->lock);
3108 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3109 ndisc_send_rs(dev, &lladdr,
3110 &in6addr_linklocal_allrouters);
3111 else
3112 goto put;
3114 write_lock(&idev->lock);
3115 /* The wait after the last probe can be shorter */
3116 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3117 idev->cnf.rtr_solicits) ?
3118 idev->cnf.rtr_solicit_delay :
3119 idev->cnf.rtr_solicit_interval);
3120 } else {
3122 * Note: we do not support deprecated "all on-link"
3123 * assumption any longer.
3125 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3128 out:
3129 write_unlock(&idev->lock);
3130 put:
3131 in6_dev_put(idev);
3135 * Duplicate Address Detection
3137 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3139 unsigned long rand_num;
3140 struct inet6_dev *idev = ifp->idev;
3142 if (ifp->flags & IFA_F_OPTIMISTIC)
3143 rand_num = 0;
3144 else
3145 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3147 ifp->dad_probes = idev->cnf.dad_transmits;
3148 addrconf_mod_dad_timer(ifp, rand_num);
3151 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3153 struct inet6_dev *idev = ifp->idev;
3154 struct net_device *dev = idev->dev;
3156 addrconf_join_solict(dev, &ifp->addr);
3158 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3160 read_lock_bh(&idev->lock);
3161 spin_lock(&ifp->lock);
3162 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3163 goto out;
3165 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3166 idev->cnf.accept_dad < 1 ||
3167 !(ifp->flags&IFA_F_TENTATIVE) ||
3168 ifp->flags & IFA_F_NODAD) {
3169 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3170 spin_unlock(&ifp->lock);
3171 read_unlock_bh(&idev->lock);
3173 addrconf_dad_completed(ifp);
3174 return;
3177 if (!(idev->if_flags & IF_READY)) {
3178 spin_unlock(&ifp->lock);
3179 read_unlock_bh(&idev->lock);
3181 * If the device is not ready:
3182 * - keep it tentative if it is a permanent address.
3183 * - otherwise, kill it.
3185 in6_ifa_hold(ifp);
3186 addrconf_dad_stop(ifp, 0);
3187 return;
3191 * Optimistic nodes can start receiving
3192 * Frames right away
3194 if (ifp->flags & IFA_F_OPTIMISTIC)
3195 ip6_ins_rt(ifp->rt);
3197 addrconf_dad_kick(ifp);
3198 out:
3199 spin_unlock(&ifp->lock);
3200 read_unlock_bh(&idev->lock);
3203 static void addrconf_dad_timer(unsigned long data)
3205 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3206 struct inet6_dev *idev = ifp->idev;
3207 struct in6_addr mcaddr;
3209 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3210 goto out;
3212 write_lock(&idev->lock);
3213 if (idev->dead || !(idev->if_flags & IF_READY)) {
3214 write_unlock(&idev->lock);
3215 goto out;
3218 spin_lock(&ifp->lock);
3219 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3220 spin_unlock(&ifp->lock);
3221 write_unlock(&idev->lock);
3222 goto out;
3225 if (ifp->dad_probes == 0) {
3227 * DAD was successful
3230 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3231 spin_unlock(&ifp->lock);
3232 write_unlock(&idev->lock);
3234 addrconf_dad_completed(ifp);
3236 goto out;
3239 ifp->dad_probes--;
3240 addrconf_mod_dad_timer(ifp,
3241 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3242 spin_unlock(&ifp->lock);
3243 write_unlock(&idev->lock);
3245 /* send a neighbour solicitation for our addr */
3246 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3247 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3248 out:
3249 in6_ifa_put(ifp);
3252 /* ifp->idev must be at least read locked */
3253 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3255 struct inet6_ifaddr *ifpiter;
3256 struct inet6_dev *idev = ifp->idev;
3258 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3259 if (ifpiter->scope > IFA_LINK)
3260 break;
3261 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3262 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3263 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3264 IFA_F_PERMANENT)
3265 return false;
3267 return true;
3270 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3272 struct net_device *dev = ifp->idev->dev;
3273 struct in6_addr lladdr;
3274 bool send_rs, send_mld;
3276 addrconf_del_dad_timer(ifp);
3279 * Configure the address for reception. Now it is valid.
3282 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3284 /* If added prefix is link local and we are prepared to process
3285 router advertisements, start sending router solicitations.
3288 read_lock_bh(&ifp->idev->lock);
3289 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3290 send_rs = send_mld &&
3291 ipv6_accept_ra(ifp->idev) &&
3292 ifp->idev->cnf.rtr_solicits > 0 &&
3293 (dev->flags&IFF_LOOPBACK) == 0;
3294 read_unlock_bh(&ifp->idev->lock);
3296 /* While dad is in progress mld report's source address is in6_addrany.
3297 * Resend with proper ll now.
3299 if (send_mld)
3300 ipv6_mc_dad_complete(ifp->idev);
3302 if (send_rs) {
3304 * If a host as already performed a random delay
3305 * [...] as part of DAD [...] there is no need
3306 * to delay again before sending the first RS
3308 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3309 return;
3310 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3312 write_lock_bh(&ifp->idev->lock);
3313 spin_lock(&ifp->lock);
3314 ifp->idev->rs_probes = 1;
3315 ifp->idev->if_flags |= IF_RS_SENT;
3316 addrconf_mod_rs_timer(ifp->idev,
3317 ifp->idev->cnf.rtr_solicit_interval);
3318 spin_unlock(&ifp->lock);
3319 write_unlock_bh(&ifp->idev->lock);
3323 static void addrconf_dad_run(struct inet6_dev *idev)
3325 struct inet6_ifaddr *ifp;
3327 read_lock_bh(&idev->lock);
3328 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3329 spin_lock(&ifp->lock);
3330 if (ifp->flags & IFA_F_TENTATIVE &&
3331 ifp->state == INET6_IFADDR_STATE_DAD)
3332 addrconf_dad_kick(ifp);
3333 spin_unlock(&ifp->lock);
3335 read_unlock_bh(&idev->lock);
3338 #ifdef CONFIG_PROC_FS
3339 struct if6_iter_state {
3340 struct seq_net_private p;
3341 int bucket;
3342 int offset;
3345 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3347 struct inet6_ifaddr *ifa = NULL;
3348 struct if6_iter_state *state = seq->private;
3349 struct net *net = seq_file_net(seq);
3350 int p = 0;
3352 /* initial bucket if pos is 0 */
3353 if (pos == 0) {
3354 state->bucket = 0;
3355 state->offset = 0;
3358 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3359 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3360 addr_lst) {
3361 if (!net_eq(dev_net(ifa->idev->dev), net))
3362 continue;
3363 /* sync with offset */
3364 if (p < state->offset) {
3365 p++;
3366 continue;
3368 state->offset++;
3369 return ifa;
3372 /* prepare for next bucket */
3373 state->offset = 0;
3374 p = 0;
3376 return NULL;
3379 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3380 struct inet6_ifaddr *ifa)
3382 struct if6_iter_state *state = seq->private;
3383 struct net *net = seq_file_net(seq);
3385 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3386 if (!net_eq(dev_net(ifa->idev->dev), net))
3387 continue;
3388 state->offset++;
3389 return ifa;
3392 while (++state->bucket < IN6_ADDR_HSIZE) {
3393 state->offset = 0;
3394 hlist_for_each_entry_rcu_bh(ifa,
3395 &inet6_addr_lst[state->bucket], addr_lst) {
3396 if (!net_eq(dev_net(ifa->idev->dev), net))
3397 continue;
3398 state->offset++;
3399 return ifa;
3403 return NULL;
3406 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3407 __acquires(rcu_bh)
3409 rcu_read_lock_bh();
3410 return if6_get_first(seq, *pos);
3413 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3415 struct inet6_ifaddr *ifa;
3417 ifa = if6_get_next(seq, v);
3418 ++*pos;
3419 return ifa;
3422 static void if6_seq_stop(struct seq_file *seq, void *v)
3423 __releases(rcu_bh)
3425 rcu_read_unlock_bh();
3428 static int if6_seq_show(struct seq_file *seq, void *v)
3430 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3431 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3432 &ifp->addr,
3433 ifp->idev->dev->ifindex,
3434 ifp->prefix_len,
3435 ifp->scope,
3436 (u8) ifp->flags,
3437 ifp->idev->dev->name);
3438 return 0;
3441 static const struct seq_operations if6_seq_ops = {
3442 .start = if6_seq_start,
3443 .next = if6_seq_next,
3444 .show = if6_seq_show,
3445 .stop = if6_seq_stop,
3448 static int if6_seq_open(struct inode *inode, struct file *file)
3450 return seq_open_net(inode, file, &if6_seq_ops,
3451 sizeof(struct if6_iter_state));
3454 static const struct file_operations if6_fops = {
3455 .owner = THIS_MODULE,
3456 .open = if6_seq_open,
3457 .read = seq_read,
3458 .llseek = seq_lseek,
3459 .release = seq_release_net,
3462 static int __net_init if6_proc_net_init(struct net *net)
3464 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3465 return -ENOMEM;
3466 return 0;
3469 static void __net_exit if6_proc_net_exit(struct net *net)
3471 remove_proc_entry("if_inet6", net->proc_net);
3474 static struct pernet_operations if6_proc_net_ops = {
3475 .init = if6_proc_net_init,
3476 .exit = if6_proc_net_exit,
3479 int __init if6_proc_init(void)
3481 return register_pernet_subsys(&if6_proc_net_ops);
3484 void if6_proc_exit(void)
3486 unregister_pernet_subsys(&if6_proc_net_ops);
3488 #endif /* CONFIG_PROC_FS */
3490 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3491 /* Check if address is a home address configured on any interface. */
3492 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3494 int ret = 0;
3495 struct inet6_ifaddr *ifp = NULL;
3496 unsigned int hash = inet6_addr_hash(addr);
3498 rcu_read_lock_bh();
3499 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3500 if (!net_eq(dev_net(ifp->idev->dev), net))
3501 continue;
3502 if (ipv6_addr_equal(&ifp->addr, addr) &&
3503 (ifp->flags & IFA_F_HOMEADDRESS)) {
3504 ret = 1;
3505 break;
3508 rcu_read_unlock_bh();
3509 return ret;
3511 #endif
3514 * Periodic address status verification
3517 static void addrconf_verify(unsigned long foo)
3519 unsigned long now, next, next_sec, next_sched;
3520 struct inet6_ifaddr *ifp;
3521 int i;
3523 rcu_read_lock_bh();
3524 spin_lock(&addrconf_verify_lock);
3525 now = jiffies;
3526 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3528 del_timer(&addr_chk_timer);
3530 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3531 restart:
3532 hlist_for_each_entry_rcu_bh(ifp,
3533 &inet6_addr_lst[i], addr_lst) {
3534 unsigned long age;
3536 /* When setting preferred_lft to a value not zero or
3537 * infinity, while valid_lft is infinity
3538 * IFA_F_PERMANENT has a non-infinity life time.
3540 if ((ifp->flags & IFA_F_PERMANENT) &&
3541 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3542 continue;
3544 spin_lock(&ifp->lock);
3545 /* We try to batch several events at once. */
3546 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3548 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3549 age >= ifp->valid_lft) {
3550 spin_unlock(&ifp->lock);
3551 in6_ifa_hold(ifp);
3552 ipv6_del_addr(ifp);
3553 goto restart;
3554 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3555 spin_unlock(&ifp->lock);
3556 continue;
3557 } else if (age >= ifp->prefered_lft) {
3558 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3559 int deprecate = 0;
3561 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3562 deprecate = 1;
3563 ifp->flags |= IFA_F_DEPRECATED;
3566 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3567 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3568 next = ifp->tstamp + ifp->valid_lft * HZ;
3570 spin_unlock(&ifp->lock);
3572 if (deprecate) {
3573 in6_ifa_hold(ifp);
3575 ipv6_ifa_notify(0, ifp);
3576 in6_ifa_put(ifp);
3577 goto restart;
3579 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3580 !(ifp->flags&IFA_F_TENTATIVE)) {
3581 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3582 ifp->idev->cnf.dad_transmits *
3583 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3585 if (age >= ifp->prefered_lft - regen_advance) {
3586 struct inet6_ifaddr *ifpub = ifp->ifpub;
3587 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3588 next = ifp->tstamp + ifp->prefered_lft * HZ;
3589 if (!ifp->regen_count && ifpub) {
3590 ifp->regen_count++;
3591 in6_ifa_hold(ifp);
3592 in6_ifa_hold(ifpub);
3593 spin_unlock(&ifp->lock);
3595 spin_lock(&ifpub->lock);
3596 ifpub->regen_count = 0;
3597 spin_unlock(&ifpub->lock);
3598 ipv6_create_tempaddr(ifpub, ifp);
3599 in6_ifa_put(ifpub);
3600 in6_ifa_put(ifp);
3601 goto restart;
3603 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3604 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3605 spin_unlock(&ifp->lock);
3606 } else {
3607 /* ifp->prefered_lft <= ifp->valid_lft */
3608 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3609 next = ifp->tstamp + ifp->prefered_lft * HZ;
3610 spin_unlock(&ifp->lock);
3615 next_sec = round_jiffies_up(next);
3616 next_sched = next;
3618 /* If rounded timeout is accurate enough, accept it. */
3619 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3620 next_sched = next_sec;
3622 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3623 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3624 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3626 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3627 now, next, next_sec, next_sched);
3629 addr_chk_timer.expires = next_sched;
3630 add_timer(&addr_chk_timer);
3631 spin_unlock(&addrconf_verify_lock);
3632 rcu_read_unlock_bh();
3635 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3636 struct in6_addr **peer_pfx)
3638 struct in6_addr *pfx = NULL;
3640 *peer_pfx = NULL;
3642 if (addr)
3643 pfx = nla_data(addr);
3645 if (local) {
3646 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3647 *peer_pfx = pfx;
3648 pfx = nla_data(local);
3651 return pfx;
3654 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3655 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3656 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3657 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3658 [IFA_FLAGS] = { .len = sizeof(u32) },
3661 static int
3662 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3664 struct net *net = sock_net(skb->sk);
3665 struct ifaddrmsg *ifm;
3666 struct nlattr *tb[IFA_MAX+1];
3667 struct in6_addr *pfx, *peer_pfx;
3668 int err;
3670 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3671 if (err < 0)
3672 return err;
3674 ifm = nlmsg_data(nlh);
3675 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3676 if (pfx == NULL)
3677 return -EINVAL;
3679 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3682 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3683 u32 prefered_lft, u32 valid_lft)
3685 u32 flags;
3686 clock_t expires;
3687 unsigned long timeout;
3688 bool was_managetempaddr;
3689 bool had_prefixroute;
3691 if (!valid_lft || (prefered_lft > valid_lft))
3692 return -EINVAL;
3694 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3695 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3696 return -EINVAL;
3698 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3699 if (addrconf_finite_timeout(timeout)) {
3700 expires = jiffies_to_clock_t(timeout * HZ);
3701 valid_lft = timeout;
3702 flags = RTF_EXPIRES;
3703 } else {
3704 expires = 0;
3705 flags = 0;
3706 ifa_flags |= IFA_F_PERMANENT;
3709 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3710 if (addrconf_finite_timeout(timeout)) {
3711 if (timeout == 0)
3712 ifa_flags |= IFA_F_DEPRECATED;
3713 prefered_lft = timeout;
3716 spin_lock_bh(&ifp->lock);
3717 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3718 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3719 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3720 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3721 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3722 IFA_F_NOPREFIXROUTE);
3723 ifp->flags |= ifa_flags;
3724 ifp->tstamp = jiffies;
3725 ifp->valid_lft = valid_lft;
3726 ifp->prefered_lft = prefered_lft;
3728 spin_unlock_bh(&ifp->lock);
3729 if (!(ifp->flags&IFA_F_TENTATIVE))
3730 ipv6_ifa_notify(0, ifp);
3732 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3733 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3734 expires, flags);
3735 } else if (had_prefixroute) {
3736 enum cleanup_prefix_rt_t action;
3737 unsigned long rt_expires;
3739 write_lock_bh(&ifp->idev->lock);
3740 action = check_cleanup_prefix_route(ifp, &rt_expires);
3741 write_unlock_bh(&ifp->idev->lock);
3743 if (action != CLEANUP_PREFIX_RT_NOP) {
3744 cleanup_prefix_route(ifp, rt_expires,
3745 action == CLEANUP_PREFIX_RT_DEL);
3749 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3750 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3751 valid_lft = prefered_lft = 0;
3752 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3753 !was_managetempaddr, jiffies);
3756 addrconf_verify(0);
3758 return 0;
3761 static int
3762 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3764 struct net *net = sock_net(skb->sk);
3765 struct ifaddrmsg *ifm;
3766 struct nlattr *tb[IFA_MAX+1];
3767 struct in6_addr *pfx, *peer_pfx;
3768 struct inet6_ifaddr *ifa;
3769 struct net_device *dev;
3770 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3771 u32 ifa_flags;
3772 int err;
3774 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3775 if (err < 0)
3776 return err;
3778 ifm = nlmsg_data(nlh);
3779 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3780 if (pfx == NULL)
3781 return -EINVAL;
3783 if (tb[IFA_CACHEINFO]) {
3784 struct ifa_cacheinfo *ci;
3786 ci = nla_data(tb[IFA_CACHEINFO]);
3787 valid_lft = ci->ifa_valid;
3788 preferred_lft = ci->ifa_prefered;
3789 } else {
3790 preferred_lft = INFINITY_LIFE_TIME;
3791 valid_lft = INFINITY_LIFE_TIME;
3794 dev = __dev_get_by_index(net, ifm->ifa_index);
3795 if (dev == NULL)
3796 return -ENODEV;
3798 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3800 /* We ignore other flags so far. */
3801 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3802 IFA_F_NOPREFIXROUTE;
3804 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3805 if (ifa == NULL) {
3807 * It would be best to check for !NLM_F_CREATE here but
3808 * userspace already relies on not having to provide this.
3810 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3811 ifm->ifa_prefixlen, ifa_flags,
3812 preferred_lft, valid_lft);
3815 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3816 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3817 err = -EEXIST;
3818 else
3819 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3821 in6_ifa_put(ifa);
3823 return err;
3826 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3827 u8 scope, int ifindex)
3829 struct ifaddrmsg *ifm;
3831 ifm = nlmsg_data(nlh);
3832 ifm->ifa_family = AF_INET6;
3833 ifm->ifa_prefixlen = prefixlen;
3834 ifm->ifa_flags = flags;
3835 ifm->ifa_scope = scope;
3836 ifm->ifa_index = ifindex;
3839 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3840 unsigned long tstamp, u32 preferred, u32 valid)
3842 struct ifa_cacheinfo ci;
3844 ci.cstamp = cstamp_delta(cstamp);
3845 ci.tstamp = cstamp_delta(tstamp);
3846 ci.ifa_prefered = preferred;
3847 ci.ifa_valid = valid;
3849 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3852 static inline int rt_scope(int ifa_scope)
3854 if (ifa_scope & IFA_HOST)
3855 return RT_SCOPE_HOST;
3856 else if (ifa_scope & IFA_LINK)
3857 return RT_SCOPE_LINK;
3858 else if (ifa_scope & IFA_SITE)
3859 return RT_SCOPE_SITE;
3860 else
3861 return RT_SCOPE_UNIVERSE;
3864 static inline int inet6_ifaddr_msgsize(void)
3866 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3867 + nla_total_size(16) /* IFA_LOCAL */
3868 + nla_total_size(16) /* IFA_ADDRESS */
3869 + nla_total_size(sizeof(struct ifa_cacheinfo))
3870 + nla_total_size(4) /* IFA_FLAGS */;
3873 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3874 u32 portid, u32 seq, int event, unsigned int flags)
3876 struct nlmsghdr *nlh;
3877 u32 preferred, valid;
3879 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3880 if (nlh == NULL)
3881 return -EMSGSIZE;
3883 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3884 ifa->idev->dev->ifindex);
3886 if (!((ifa->flags&IFA_F_PERMANENT) &&
3887 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3888 preferred = ifa->prefered_lft;
3889 valid = ifa->valid_lft;
3890 if (preferred != INFINITY_LIFE_TIME) {
3891 long tval = (jiffies - ifa->tstamp)/HZ;
3892 if (preferred > tval)
3893 preferred -= tval;
3894 else
3895 preferred = 0;
3896 if (valid != INFINITY_LIFE_TIME) {
3897 if (valid > tval)
3898 valid -= tval;
3899 else
3900 valid = 0;
3903 } else {
3904 preferred = INFINITY_LIFE_TIME;
3905 valid = INFINITY_LIFE_TIME;
3908 if (!ipv6_addr_any(&ifa->peer_addr)) {
3909 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3910 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3911 goto error;
3912 } else
3913 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3914 goto error;
3916 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3917 goto error;
3919 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
3920 goto error;
3922 return nlmsg_end(skb, nlh);
3924 error:
3925 nlmsg_cancel(skb, nlh);
3926 return -EMSGSIZE;
3929 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3930 u32 portid, u32 seq, int event, u16 flags)
3932 struct nlmsghdr *nlh;
3933 u8 scope = RT_SCOPE_UNIVERSE;
3934 int ifindex = ifmca->idev->dev->ifindex;
3936 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3937 scope = RT_SCOPE_SITE;
3939 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3940 if (nlh == NULL)
3941 return -EMSGSIZE;
3943 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3944 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3945 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3946 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3947 nlmsg_cancel(skb, nlh);
3948 return -EMSGSIZE;
3951 return nlmsg_end(skb, nlh);
3954 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3955 u32 portid, u32 seq, int event, unsigned int flags)
3957 struct nlmsghdr *nlh;
3958 u8 scope = RT_SCOPE_UNIVERSE;
3959 int ifindex = ifaca->aca_idev->dev->ifindex;
3961 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3962 scope = RT_SCOPE_SITE;
3964 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3965 if (nlh == NULL)
3966 return -EMSGSIZE;
3968 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3969 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3970 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3971 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3972 nlmsg_cancel(skb, nlh);
3973 return -EMSGSIZE;
3976 return nlmsg_end(skb, nlh);
3979 enum addr_type_t {
3980 UNICAST_ADDR,
3981 MULTICAST_ADDR,
3982 ANYCAST_ADDR,
3985 /* called with rcu_read_lock() */
3986 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3987 struct netlink_callback *cb, enum addr_type_t type,
3988 int s_ip_idx, int *p_ip_idx)
3990 struct ifmcaddr6 *ifmca;
3991 struct ifacaddr6 *ifaca;
3992 int err = 1;
3993 int ip_idx = *p_ip_idx;
3995 read_lock_bh(&idev->lock);
3996 switch (type) {
3997 case UNICAST_ADDR: {
3998 struct inet6_ifaddr *ifa;
4000 /* unicast address incl. temp addr */
4001 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4002 if (++ip_idx < s_ip_idx)
4003 continue;
4004 err = inet6_fill_ifaddr(skb, ifa,
4005 NETLINK_CB(cb->skb).portid,
4006 cb->nlh->nlmsg_seq,
4007 RTM_NEWADDR,
4008 NLM_F_MULTI);
4009 if (err <= 0)
4010 break;
4011 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4013 break;
4015 case MULTICAST_ADDR:
4016 /* multicast address */
4017 for (ifmca = idev->mc_list; ifmca;
4018 ifmca = ifmca->next, ip_idx++) {
4019 if (ip_idx < s_ip_idx)
4020 continue;
4021 err = inet6_fill_ifmcaddr(skb, ifmca,
4022 NETLINK_CB(cb->skb).portid,
4023 cb->nlh->nlmsg_seq,
4024 RTM_GETMULTICAST,
4025 NLM_F_MULTI);
4026 if (err <= 0)
4027 break;
4029 break;
4030 case ANYCAST_ADDR:
4031 /* anycast address */
4032 for (ifaca = idev->ac_list; ifaca;
4033 ifaca = ifaca->aca_next, ip_idx++) {
4034 if (ip_idx < s_ip_idx)
4035 continue;
4036 err = inet6_fill_ifacaddr(skb, ifaca,
4037 NETLINK_CB(cb->skb).portid,
4038 cb->nlh->nlmsg_seq,
4039 RTM_GETANYCAST,
4040 NLM_F_MULTI);
4041 if (err <= 0)
4042 break;
4044 break;
4045 default:
4046 break;
4048 read_unlock_bh(&idev->lock);
4049 *p_ip_idx = ip_idx;
4050 return err;
4053 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4054 enum addr_type_t type)
4056 struct net *net = sock_net(skb->sk);
4057 int h, s_h;
4058 int idx, ip_idx;
4059 int s_idx, s_ip_idx;
4060 struct net_device *dev;
4061 struct inet6_dev *idev;
4062 struct hlist_head *head;
4064 s_h = cb->args[0];
4065 s_idx = idx = cb->args[1];
4066 s_ip_idx = ip_idx = cb->args[2];
4068 rcu_read_lock();
4069 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4070 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4071 idx = 0;
4072 head = &net->dev_index_head[h];
4073 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4074 if (idx < s_idx)
4075 goto cont;
4076 if (h > s_h || idx > s_idx)
4077 s_ip_idx = 0;
4078 ip_idx = 0;
4079 idev = __in6_dev_get(dev);
4080 if (!idev)
4081 goto cont;
4083 if (in6_dump_addrs(idev, skb, cb, type,
4084 s_ip_idx, &ip_idx) <= 0)
4085 goto done;
4086 cont:
4087 idx++;
4090 done:
4091 rcu_read_unlock();
4092 cb->args[0] = h;
4093 cb->args[1] = idx;
4094 cb->args[2] = ip_idx;
4096 return skb->len;
4099 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4101 enum addr_type_t type = UNICAST_ADDR;
4103 return inet6_dump_addr(skb, cb, type);
4106 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4108 enum addr_type_t type = MULTICAST_ADDR;
4110 return inet6_dump_addr(skb, cb, type);
4114 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4116 enum addr_type_t type = ANYCAST_ADDR;
4118 return inet6_dump_addr(skb, cb, type);
4121 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4123 struct net *net = sock_net(in_skb->sk);
4124 struct ifaddrmsg *ifm;
4125 struct nlattr *tb[IFA_MAX+1];
4126 struct in6_addr *addr = NULL, *peer;
4127 struct net_device *dev = NULL;
4128 struct inet6_ifaddr *ifa;
4129 struct sk_buff *skb;
4130 int err;
4132 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4133 if (err < 0)
4134 goto errout;
4136 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4137 if (addr == NULL) {
4138 err = -EINVAL;
4139 goto errout;
4142 ifm = nlmsg_data(nlh);
4143 if (ifm->ifa_index)
4144 dev = __dev_get_by_index(net, ifm->ifa_index);
4146 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4147 if (!ifa) {
4148 err = -EADDRNOTAVAIL;
4149 goto errout;
4152 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4153 if (!skb) {
4154 err = -ENOBUFS;
4155 goto errout_ifa;
4158 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4159 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4160 if (err < 0) {
4161 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4162 WARN_ON(err == -EMSGSIZE);
4163 kfree_skb(skb);
4164 goto errout_ifa;
4166 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4167 errout_ifa:
4168 in6_ifa_put(ifa);
4169 errout:
4170 return err;
4173 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4175 struct sk_buff *skb;
4176 struct net *net = dev_net(ifa->idev->dev);
4177 int err = -ENOBUFS;
4179 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4180 if (skb == NULL)
4181 goto errout;
4183 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4184 if (err < 0) {
4185 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4186 WARN_ON(err == -EMSGSIZE);
4187 kfree_skb(skb);
4188 goto errout;
4190 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4191 return;
4192 errout:
4193 if (err < 0)
4194 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4197 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4198 __s32 *array, int bytes)
4200 BUG_ON(bytes < (DEVCONF_MAX * 4));
4202 memset(array, 0, bytes);
4203 array[DEVCONF_FORWARDING] = cnf->forwarding;
4204 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4205 array[DEVCONF_MTU6] = cnf->mtu6;
4206 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4207 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4208 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4209 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4210 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4211 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4212 jiffies_to_msecs(cnf->rtr_solicit_interval);
4213 array[DEVCONF_RTR_SOLICIT_DELAY] =
4214 jiffies_to_msecs(cnf->rtr_solicit_delay);
4215 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4216 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4217 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4218 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4219 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4220 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4221 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4222 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4223 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4224 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4225 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4226 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4227 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4228 #ifdef CONFIG_IPV6_ROUTER_PREF
4229 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4230 array[DEVCONF_RTR_PROBE_INTERVAL] =
4231 jiffies_to_msecs(cnf->rtr_probe_interval);
4232 #ifdef CONFIG_IPV6_ROUTE_INFO
4233 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4234 #endif
4235 #endif
4236 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4237 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4238 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4239 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4240 #endif
4241 #ifdef CONFIG_IPV6_MROUTE
4242 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4243 #endif
4244 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4245 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4246 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4247 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4248 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4251 static inline size_t inet6_ifla6_size(void)
4253 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4254 + nla_total_size(sizeof(struct ifla_cacheinfo))
4255 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4256 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4257 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4258 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4261 static inline size_t inet6_if_nlmsg_size(void)
4263 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4264 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4265 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4266 + nla_total_size(4) /* IFLA_MTU */
4267 + nla_total_size(4) /* IFLA_LINK */
4268 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4271 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4272 int items, int bytes)
4274 int i;
4275 int pad = bytes - sizeof(u64) * items;
4276 BUG_ON(pad < 0);
4278 /* Use put_unaligned() because stats may not be aligned for u64. */
4279 put_unaligned(items, &stats[0]);
4280 for (i = 1; i < items; i++)
4281 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4283 memset(&stats[items], 0, pad);
4286 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4287 int items, int bytes, size_t syncpoff)
4289 int i;
4290 int pad = bytes - sizeof(u64) * items;
4291 BUG_ON(pad < 0);
4293 /* Use put_unaligned() because stats may not be aligned for u64. */
4294 put_unaligned(items, &stats[0]);
4295 for (i = 1; i < items; i++)
4296 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4298 memset(&stats[items], 0, pad);
4301 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4302 int bytes)
4304 switch (attrtype) {
4305 case IFLA_INET6_STATS:
4306 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4307 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4308 break;
4309 case IFLA_INET6_ICMP6STATS:
4310 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4311 break;
4315 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4317 struct nlattr *nla;
4318 struct ifla_cacheinfo ci;
4320 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4321 goto nla_put_failure;
4322 ci.max_reasm_len = IPV6_MAXPLEN;
4323 ci.tstamp = cstamp_delta(idev->tstamp);
4324 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4325 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4326 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4327 goto nla_put_failure;
4328 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4329 if (nla == NULL)
4330 goto nla_put_failure;
4331 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4333 /* XXX - MC not implemented */
4335 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4336 if (nla == NULL)
4337 goto nla_put_failure;
4338 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4340 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4341 if (nla == NULL)
4342 goto nla_put_failure;
4343 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4345 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4346 if (nla == NULL)
4347 goto nla_put_failure;
4348 read_lock_bh(&idev->lock);
4349 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4350 read_unlock_bh(&idev->lock);
4352 return 0;
4354 nla_put_failure:
4355 return -EMSGSIZE;
4358 static size_t inet6_get_link_af_size(const struct net_device *dev)
4360 if (!__in6_dev_get(dev))
4361 return 0;
4363 return inet6_ifla6_size();
4366 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4368 struct inet6_dev *idev = __in6_dev_get(dev);
4370 if (!idev)
4371 return -ENODATA;
4373 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4374 return -EMSGSIZE;
4376 return 0;
4379 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4381 struct inet6_ifaddr *ifp;
4382 struct net_device *dev = idev->dev;
4383 bool update_rs = false;
4384 struct in6_addr ll_addr;
4386 if (token == NULL)
4387 return -EINVAL;
4388 if (ipv6_addr_any(token))
4389 return -EINVAL;
4390 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4391 return -EINVAL;
4392 if (!ipv6_accept_ra(idev))
4393 return -EINVAL;
4394 if (idev->cnf.rtr_solicits <= 0)
4395 return -EINVAL;
4397 write_lock_bh(&idev->lock);
4399 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4400 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4402 write_unlock_bh(&idev->lock);
4404 if (!idev->dead && (idev->if_flags & IF_READY) &&
4405 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4406 IFA_F_OPTIMISTIC)) {
4408 /* If we're not ready, then normal ifup will take care
4409 * of this. Otherwise, we need to request our rs here.
4411 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4412 update_rs = true;
4415 write_lock_bh(&idev->lock);
4417 if (update_rs) {
4418 idev->if_flags |= IF_RS_SENT;
4419 idev->rs_probes = 1;
4420 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4423 /* Well, that's kinda nasty ... */
4424 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4425 spin_lock(&ifp->lock);
4426 if (ifp->tokenized) {
4427 ifp->valid_lft = 0;
4428 ifp->prefered_lft = 0;
4430 spin_unlock(&ifp->lock);
4433 write_unlock_bh(&idev->lock);
4434 addrconf_verify(0);
4435 return 0;
4438 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4440 int err = -EINVAL;
4441 struct inet6_dev *idev = __in6_dev_get(dev);
4442 struct nlattr *tb[IFLA_INET6_MAX + 1];
4444 if (!idev)
4445 return -EAFNOSUPPORT;
4447 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4448 BUG();
4450 if (tb[IFLA_INET6_TOKEN])
4451 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4453 return err;
4456 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4457 u32 portid, u32 seq, int event, unsigned int flags)
4459 struct net_device *dev = idev->dev;
4460 struct ifinfomsg *hdr;
4461 struct nlmsghdr *nlh;
4462 void *protoinfo;
4464 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4465 if (nlh == NULL)
4466 return -EMSGSIZE;
4468 hdr = nlmsg_data(nlh);
4469 hdr->ifi_family = AF_INET6;
4470 hdr->__ifi_pad = 0;
4471 hdr->ifi_type = dev->type;
4472 hdr->ifi_index = dev->ifindex;
4473 hdr->ifi_flags = dev_get_flags(dev);
4474 hdr->ifi_change = 0;
4476 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4477 (dev->addr_len &&
4478 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4479 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4480 (dev->ifindex != dev->iflink &&
4481 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4482 goto nla_put_failure;
4483 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4484 if (protoinfo == NULL)
4485 goto nla_put_failure;
4487 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4488 goto nla_put_failure;
4490 nla_nest_end(skb, protoinfo);
4491 return nlmsg_end(skb, nlh);
4493 nla_put_failure:
4494 nlmsg_cancel(skb, nlh);
4495 return -EMSGSIZE;
4498 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4500 struct net *net = sock_net(skb->sk);
4501 int h, s_h;
4502 int idx = 0, s_idx;
4503 struct net_device *dev;
4504 struct inet6_dev *idev;
4505 struct hlist_head *head;
4507 s_h = cb->args[0];
4508 s_idx = cb->args[1];
4510 rcu_read_lock();
4511 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4512 idx = 0;
4513 head = &net->dev_index_head[h];
4514 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4515 if (idx < s_idx)
4516 goto cont;
4517 idev = __in6_dev_get(dev);
4518 if (!idev)
4519 goto cont;
4520 if (inet6_fill_ifinfo(skb, idev,
4521 NETLINK_CB(cb->skb).portid,
4522 cb->nlh->nlmsg_seq,
4523 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4524 goto out;
4525 cont:
4526 idx++;
4529 out:
4530 rcu_read_unlock();
4531 cb->args[1] = idx;
4532 cb->args[0] = h;
4534 return skb->len;
4537 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4539 struct sk_buff *skb;
4540 struct net *net = dev_net(idev->dev);
4541 int err = -ENOBUFS;
4543 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4544 if (skb == NULL)
4545 goto errout;
4547 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4548 if (err < 0) {
4549 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4550 WARN_ON(err == -EMSGSIZE);
4551 kfree_skb(skb);
4552 goto errout;
4554 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4555 return;
4556 errout:
4557 if (err < 0)
4558 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4561 static inline size_t inet6_prefix_nlmsg_size(void)
4563 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4564 + nla_total_size(sizeof(struct in6_addr))
4565 + nla_total_size(sizeof(struct prefix_cacheinfo));
4568 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4569 struct prefix_info *pinfo, u32 portid, u32 seq,
4570 int event, unsigned int flags)
4572 struct prefixmsg *pmsg;
4573 struct nlmsghdr *nlh;
4574 struct prefix_cacheinfo ci;
4576 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4577 if (nlh == NULL)
4578 return -EMSGSIZE;
4580 pmsg = nlmsg_data(nlh);
4581 pmsg->prefix_family = AF_INET6;
4582 pmsg->prefix_pad1 = 0;
4583 pmsg->prefix_pad2 = 0;
4584 pmsg->prefix_ifindex = idev->dev->ifindex;
4585 pmsg->prefix_len = pinfo->prefix_len;
4586 pmsg->prefix_type = pinfo->type;
4587 pmsg->prefix_pad3 = 0;
4588 pmsg->prefix_flags = 0;
4589 if (pinfo->onlink)
4590 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4591 if (pinfo->autoconf)
4592 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4594 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4595 goto nla_put_failure;
4596 ci.preferred_time = ntohl(pinfo->prefered);
4597 ci.valid_time = ntohl(pinfo->valid);
4598 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4599 goto nla_put_failure;
4600 return nlmsg_end(skb, nlh);
4602 nla_put_failure:
4603 nlmsg_cancel(skb, nlh);
4604 return -EMSGSIZE;
4607 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4608 struct prefix_info *pinfo)
4610 struct sk_buff *skb;
4611 struct net *net = dev_net(idev->dev);
4612 int err = -ENOBUFS;
4614 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4615 if (skb == NULL)
4616 goto errout;
4618 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4619 if (err < 0) {
4620 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4621 WARN_ON(err == -EMSGSIZE);
4622 kfree_skb(skb);
4623 goto errout;
4625 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4626 return;
4627 errout:
4628 if (err < 0)
4629 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4632 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4634 struct net *net = dev_net(ifp->idev->dev);
4636 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4638 switch (event) {
4639 case RTM_NEWADDR:
4641 * If the address was optimistic
4642 * we inserted the route at the start of
4643 * our DAD process, so we don't need
4644 * to do it again
4646 if (!(ifp->rt->rt6i_node))
4647 ip6_ins_rt(ifp->rt);
4648 if (ifp->idev->cnf.forwarding)
4649 addrconf_join_anycast(ifp);
4650 if (!ipv6_addr_any(&ifp->peer_addr))
4651 addrconf_prefix_route(&ifp->peer_addr, 128,
4652 ifp->idev->dev, 0, 0);
4653 break;
4654 case RTM_DELADDR:
4655 if (ifp->idev->cnf.forwarding)
4656 addrconf_leave_anycast(ifp);
4657 addrconf_leave_solict(ifp->idev, &ifp->addr);
4658 if (!ipv6_addr_any(&ifp->peer_addr)) {
4659 struct rt6_info *rt;
4660 struct net_device *dev = ifp->idev->dev;
4662 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4663 dev->ifindex, 1);
4664 if (rt) {
4665 dst_hold(&rt->dst);
4666 if (ip6_del_rt(rt))
4667 dst_free(&rt->dst);
4670 dst_hold(&ifp->rt->dst);
4672 if (ip6_del_rt(ifp->rt))
4673 dst_free(&ifp->rt->dst);
4674 break;
4676 atomic_inc(&net->ipv6.dev_addr_genid);
4677 rt_genid_bump_ipv6(net);
4680 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4682 rcu_read_lock_bh();
4683 if (likely(ifp->idev->dead == 0))
4684 __ipv6_ifa_notify(event, ifp);
4685 rcu_read_unlock_bh();
4688 #ifdef CONFIG_SYSCTL
4690 static
4691 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4692 void __user *buffer, size_t *lenp, loff_t *ppos)
4694 int *valp = ctl->data;
4695 int val = *valp;
4696 loff_t pos = *ppos;
4697 struct ctl_table lctl;
4698 int ret;
4701 * ctl->data points to idev->cnf.forwarding, we should
4702 * not modify it until we get the rtnl lock.
4704 lctl = *ctl;
4705 lctl.data = &val;
4707 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4709 if (write)
4710 ret = addrconf_fixup_forwarding(ctl, valp, val);
4711 if (ret)
4712 *ppos = pos;
4713 return ret;
4716 static void dev_disable_change(struct inet6_dev *idev)
4718 struct netdev_notifier_info info;
4720 if (!idev || !idev->dev)
4721 return;
4723 netdev_notifier_info_init(&info, idev->dev);
4724 if (idev->cnf.disable_ipv6)
4725 addrconf_notify(NULL, NETDEV_DOWN, &info);
4726 else
4727 addrconf_notify(NULL, NETDEV_UP, &info);
4730 static void addrconf_disable_change(struct net *net, __s32 newf)
4732 struct net_device *dev;
4733 struct inet6_dev *idev;
4735 rcu_read_lock();
4736 for_each_netdev_rcu(net, dev) {
4737 idev = __in6_dev_get(dev);
4738 if (idev) {
4739 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4740 idev->cnf.disable_ipv6 = newf;
4741 if (changed)
4742 dev_disable_change(idev);
4745 rcu_read_unlock();
4748 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4750 struct net *net;
4751 int old;
4753 if (!rtnl_trylock())
4754 return restart_syscall();
4756 net = (struct net *)table->extra2;
4757 old = *p;
4758 *p = newf;
4760 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4761 rtnl_unlock();
4762 return 0;
4765 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4766 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4767 addrconf_disable_change(net, newf);
4768 } else if ((!newf) ^ (!old))
4769 dev_disable_change((struct inet6_dev *)table->extra1);
4771 rtnl_unlock();
4772 return 0;
4775 static
4776 int addrconf_sysctl_disable(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.disable_ipv6, 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_disable_ipv6(ctl, valp, val);
4796 if (ret)
4797 *ppos = pos;
4798 return ret;
4801 static
4802 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4803 void __user *buffer, size_t *lenp, loff_t *ppos)
4805 int *valp = ctl->data;
4806 int ret;
4807 int old, new;
4809 old = *valp;
4810 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4811 new = *valp;
4813 if (write && old != new) {
4814 struct net *net = ctl->extra2;
4816 if (!rtnl_trylock())
4817 return restart_syscall();
4819 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4820 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4821 NETCONFA_IFINDEX_DEFAULT,
4822 net->ipv6.devconf_dflt);
4823 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4824 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4825 NETCONFA_IFINDEX_ALL,
4826 net->ipv6.devconf_all);
4827 else {
4828 struct inet6_dev *idev = ctl->extra1;
4830 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4831 idev->dev->ifindex,
4832 &idev->cnf);
4834 rtnl_unlock();
4837 return ret;
4841 static struct addrconf_sysctl_table
4843 struct ctl_table_header *sysctl_header;
4844 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4845 } addrconf_sysctl __read_mostly = {
4846 .sysctl_header = NULL,
4847 .addrconf_vars = {
4849 .procname = "forwarding",
4850 .data = &ipv6_devconf.forwarding,
4851 .maxlen = sizeof(int),
4852 .mode = 0644,
4853 .proc_handler = addrconf_sysctl_forward,
4856 .procname = "hop_limit",
4857 .data = &ipv6_devconf.hop_limit,
4858 .maxlen = sizeof(int),
4859 .mode = 0644,
4860 .proc_handler = proc_dointvec,
4863 .procname = "mtu",
4864 .data = &ipv6_devconf.mtu6,
4865 .maxlen = sizeof(int),
4866 .mode = 0644,
4867 .proc_handler = proc_dointvec,
4870 .procname = "accept_ra",
4871 .data = &ipv6_devconf.accept_ra,
4872 .maxlen = sizeof(int),
4873 .mode = 0644,
4874 .proc_handler = proc_dointvec,
4877 .procname = "accept_redirects",
4878 .data = &ipv6_devconf.accept_redirects,
4879 .maxlen = sizeof(int),
4880 .mode = 0644,
4881 .proc_handler = proc_dointvec,
4884 .procname = "autoconf",
4885 .data = &ipv6_devconf.autoconf,
4886 .maxlen = sizeof(int),
4887 .mode = 0644,
4888 .proc_handler = proc_dointvec,
4891 .procname = "dad_transmits",
4892 .data = &ipv6_devconf.dad_transmits,
4893 .maxlen = sizeof(int),
4894 .mode = 0644,
4895 .proc_handler = proc_dointvec,
4898 .procname = "router_solicitations",
4899 .data = &ipv6_devconf.rtr_solicits,
4900 .maxlen = sizeof(int),
4901 .mode = 0644,
4902 .proc_handler = proc_dointvec,
4905 .procname = "router_solicitation_interval",
4906 .data = &ipv6_devconf.rtr_solicit_interval,
4907 .maxlen = sizeof(int),
4908 .mode = 0644,
4909 .proc_handler = proc_dointvec_jiffies,
4912 .procname = "router_solicitation_delay",
4913 .data = &ipv6_devconf.rtr_solicit_delay,
4914 .maxlen = sizeof(int),
4915 .mode = 0644,
4916 .proc_handler = proc_dointvec_jiffies,
4919 .procname = "force_mld_version",
4920 .data = &ipv6_devconf.force_mld_version,
4921 .maxlen = sizeof(int),
4922 .mode = 0644,
4923 .proc_handler = proc_dointvec,
4926 .procname = "mldv1_unsolicited_report_interval",
4927 .data =
4928 &ipv6_devconf.mldv1_unsolicited_report_interval,
4929 .maxlen = sizeof(int),
4930 .mode = 0644,
4931 .proc_handler = proc_dointvec_ms_jiffies,
4934 .procname = "mldv2_unsolicited_report_interval",
4935 .data =
4936 &ipv6_devconf.mldv2_unsolicited_report_interval,
4937 .maxlen = sizeof(int),
4938 .mode = 0644,
4939 .proc_handler = proc_dointvec_ms_jiffies,
4942 .procname = "use_tempaddr",
4943 .data = &ipv6_devconf.use_tempaddr,
4944 .maxlen = sizeof(int),
4945 .mode = 0644,
4946 .proc_handler = proc_dointvec,
4949 .procname = "temp_valid_lft",
4950 .data = &ipv6_devconf.temp_valid_lft,
4951 .maxlen = sizeof(int),
4952 .mode = 0644,
4953 .proc_handler = proc_dointvec,
4956 .procname = "temp_prefered_lft",
4957 .data = &ipv6_devconf.temp_prefered_lft,
4958 .maxlen = sizeof(int),
4959 .mode = 0644,
4960 .proc_handler = proc_dointvec,
4963 .procname = "regen_max_retry",
4964 .data = &ipv6_devconf.regen_max_retry,
4965 .maxlen = sizeof(int),
4966 .mode = 0644,
4967 .proc_handler = proc_dointvec,
4970 .procname = "max_desync_factor",
4971 .data = &ipv6_devconf.max_desync_factor,
4972 .maxlen = sizeof(int),
4973 .mode = 0644,
4974 .proc_handler = proc_dointvec,
4977 .procname = "max_addresses",
4978 .data = &ipv6_devconf.max_addresses,
4979 .maxlen = sizeof(int),
4980 .mode = 0644,
4981 .proc_handler = proc_dointvec,
4984 .procname = "accept_ra_defrtr",
4985 .data = &ipv6_devconf.accept_ra_defrtr,
4986 .maxlen = sizeof(int),
4987 .mode = 0644,
4988 .proc_handler = proc_dointvec,
4991 .procname = "accept_ra_pinfo",
4992 .data = &ipv6_devconf.accept_ra_pinfo,
4993 .maxlen = sizeof(int),
4994 .mode = 0644,
4995 .proc_handler = proc_dointvec,
4997 #ifdef CONFIG_IPV6_ROUTER_PREF
4999 .procname = "accept_ra_rtr_pref",
5000 .data = &ipv6_devconf.accept_ra_rtr_pref,
5001 .maxlen = sizeof(int),
5002 .mode = 0644,
5003 .proc_handler = proc_dointvec,
5006 .procname = "router_probe_interval",
5007 .data = &ipv6_devconf.rtr_probe_interval,
5008 .maxlen = sizeof(int),
5009 .mode = 0644,
5010 .proc_handler = proc_dointvec_jiffies,
5012 #ifdef CONFIG_IPV6_ROUTE_INFO
5014 .procname = "accept_ra_rt_info_max_plen",
5015 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5016 .maxlen = sizeof(int),
5017 .mode = 0644,
5018 .proc_handler = proc_dointvec,
5020 #endif
5021 #endif
5023 .procname = "proxy_ndp",
5024 .data = &ipv6_devconf.proxy_ndp,
5025 .maxlen = sizeof(int),
5026 .mode = 0644,
5027 .proc_handler = addrconf_sysctl_proxy_ndp,
5030 .procname = "accept_source_route",
5031 .data = &ipv6_devconf.accept_source_route,
5032 .maxlen = sizeof(int),
5033 .mode = 0644,
5034 .proc_handler = proc_dointvec,
5036 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5038 .procname = "optimistic_dad",
5039 .data = &ipv6_devconf.optimistic_dad,
5040 .maxlen = sizeof(int),
5041 .mode = 0644,
5042 .proc_handler = proc_dointvec,
5045 #endif
5046 #ifdef CONFIG_IPV6_MROUTE
5048 .procname = "mc_forwarding",
5049 .data = &ipv6_devconf.mc_forwarding,
5050 .maxlen = sizeof(int),
5051 .mode = 0444,
5052 .proc_handler = proc_dointvec,
5054 #endif
5056 .procname = "disable_ipv6",
5057 .data = &ipv6_devconf.disable_ipv6,
5058 .maxlen = sizeof(int),
5059 .mode = 0644,
5060 .proc_handler = addrconf_sysctl_disable,
5063 .procname = "accept_dad",
5064 .data = &ipv6_devconf.accept_dad,
5065 .maxlen = sizeof(int),
5066 .mode = 0644,
5067 .proc_handler = proc_dointvec,
5070 .procname = "force_tllao",
5071 .data = &ipv6_devconf.force_tllao,
5072 .maxlen = sizeof(int),
5073 .mode = 0644,
5074 .proc_handler = proc_dointvec
5077 .procname = "ndisc_notify",
5078 .data = &ipv6_devconf.ndisc_notify,
5079 .maxlen = sizeof(int),
5080 .mode = 0644,
5081 .proc_handler = proc_dointvec
5084 .procname = "suppress_frag_ndisc",
5085 .data = &ipv6_devconf.suppress_frag_ndisc,
5086 .maxlen = sizeof(int),
5087 .mode = 0644,
5088 .proc_handler = proc_dointvec
5091 /* sentinel */
5096 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5097 struct inet6_dev *idev, struct ipv6_devconf *p)
5099 int i;
5100 struct addrconf_sysctl_table *t;
5101 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5103 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5104 if (t == NULL)
5105 goto out;
5107 for (i = 0; t->addrconf_vars[i].data; i++) {
5108 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5109 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5110 t->addrconf_vars[i].extra2 = net;
5113 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5115 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5116 if (t->sysctl_header == NULL)
5117 goto free;
5119 p->sysctl = t;
5120 return 0;
5122 free:
5123 kfree(t);
5124 out:
5125 return -ENOBUFS;
5128 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5130 struct addrconf_sysctl_table *t;
5132 if (p->sysctl == NULL)
5133 return;
5135 t = p->sysctl;
5136 p->sysctl = NULL;
5137 unregister_net_sysctl_table(t->sysctl_header);
5138 kfree(t);
5141 static void addrconf_sysctl_register(struct inet6_dev *idev)
5143 neigh_sysctl_register(idev->dev, idev->nd_parms,
5144 &ndisc_ifinfo_sysctl_change);
5145 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5146 idev, &idev->cnf);
5149 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5151 __addrconf_sysctl_unregister(&idev->cnf);
5152 neigh_sysctl_unregister(idev->nd_parms);
5156 #endif
5158 static int __net_init addrconf_init_net(struct net *net)
5160 int err = -ENOMEM;
5161 struct ipv6_devconf *all, *dflt;
5163 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5164 if (all == NULL)
5165 goto err_alloc_all;
5167 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5168 if (dflt == NULL)
5169 goto err_alloc_dflt;
5171 /* these will be inherited by all namespaces */
5172 dflt->autoconf = ipv6_defaults.autoconf;
5173 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5175 net->ipv6.devconf_all = all;
5176 net->ipv6.devconf_dflt = dflt;
5178 #ifdef CONFIG_SYSCTL
5179 err = __addrconf_sysctl_register(net, "all", NULL, all);
5180 if (err < 0)
5181 goto err_reg_all;
5183 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5184 if (err < 0)
5185 goto err_reg_dflt;
5186 #endif
5187 return 0;
5189 #ifdef CONFIG_SYSCTL
5190 err_reg_dflt:
5191 __addrconf_sysctl_unregister(all);
5192 err_reg_all:
5193 kfree(dflt);
5194 #endif
5195 err_alloc_dflt:
5196 kfree(all);
5197 err_alloc_all:
5198 return err;
5201 static void __net_exit addrconf_exit_net(struct net *net)
5203 #ifdef CONFIG_SYSCTL
5204 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5205 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5206 #endif
5207 if (!net_eq(net, &init_net)) {
5208 kfree(net->ipv6.devconf_dflt);
5209 kfree(net->ipv6.devconf_all);
5213 static struct pernet_operations addrconf_ops = {
5214 .init = addrconf_init_net,
5215 .exit = addrconf_exit_net,
5218 static struct rtnl_af_ops inet6_ops = {
5219 .family = AF_INET6,
5220 .fill_link_af = inet6_fill_link_af,
5221 .get_link_af_size = inet6_get_link_af_size,
5222 .set_link_af = inet6_set_link_af,
5226 * Init / cleanup code
5229 int __init addrconf_init(void)
5231 int i, err;
5233 err = ipv6_addr_label_init();
5234 if (err < 0) {
5235 pr_crit("%s: cannot initialize default policy table: %d\n",
5236 __func__, err);
5237 goto out;
5240 err = register_pernet_subsys(&addrconf_ops);
5241 if (err < 0)
5242 goto out_addrlabel;
5244 /* The addrconf netdev notifier requires that loopback_dev
5245 * has it's ipv6 private information allocated and setup
5246 * before it can bring up and give link-local addresses
5247 * to other devices which are up.
5249 * Unfortunately, loopback_dev is not necessarily the first
5250 * entry in the global dev_base list of net devices. In fact,
5251 * it is likely to be the very last entry on that list.
5252 * So this causes the notifier registry below to try and
5253 * give link-local addresses to all devices besides loopback_dev
5254 * first, then loopback_dev, which cases all the non-loopback_dev
5255 * devices to fail to get a link-local address.
5257 * So, as a temporary fix, allocate the ipv6 structure for
5258 * loopback_dev first by hand.
5259 * Longer term, all of the dependencies ipv6 has upon the loopback
5260 * device and it being up should be removed.
5262 rtnl_lock();
5263 if (!ipv6_add_dev(init_net.loopback_dev))
5264 err = -ENOMEM;
5265 rtnl_unlock();
5266 if (err)
5267 goto errlo;
5269 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5270 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5272 register_netdevice_notifier(&ipv6_dev_notf);
5274 addrconf_verify(0);
5276 rtnl_af_register(&inet6_ops);
5278 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5279 NULL);
5280 if (err < 0)
5281 goto errout;
5283 /* Only the first call to __rtnl_register can fail */
5284 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5285 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5286 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5287 inet6_dump_ifaddr, NULL);
5288 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5289 inet6_dump_ifmcaddr, NULL);
5290 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5291 inet6_dump_ifacaddr, NULL);
5292 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5293 inet6_netconf_dump_devconf, NULL);
5295 ipv6_addr_label_rtnl_register();
5297 return 0;
5298 errout:
5299 rtnl_af_unregister(&inet6_ops);
5300 unregister_netdevice_notifier(&ipv6_dev_notf);
5301 errlo:
5302 unregister_pernet_subsys(&addrconf_ops);
5303 out_addrlabel:
5304 ipv6_addr_label_cleanup();
5305 out:
5306 return err;
5309 void addrconf_cleanup(void)
5311 struct net_device *dev;
5312 int i;
5314 unregister_netdevice_notifier(&ipv6_dev_notf);
5315 unregister_pernet_subsys(&addrconf_ops);
5316 ipv6_addr_label_cleanup();
5318 rtnl_lock();
5320 __rtnl_af_unregister(&inet6_ops);
5322 /* clean dev list */
5323 for_each_netdev(&init_net, dev) {
5324 if (__in6_dev_get(dev) == NULL)
5325 continue;
5326 addrconf_ifdown(dev, 1);
5328 addrconf_ifdown(init_net.loopback_dev, 2);
5331 * Check hash table.
5333 spin_lock_bh(&addrconf_hash_lock);
5334 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5335 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5336 spin_unlock_bh(&addrconf_hash_lock);
5338 del_timer(&addr_chk_timer);
5339 rtnl_unlock();