sit/gre6: don't try to add the same route two times
[linux-2.6.git] / net / ipv6 / addrconf.c
blobcdc14d0947f39c465234bdecf39aef34ce9898a9
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 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
275 sizeof(struct ipstats_mib),
276 __alignof__(struct ipstats_mib)) < 0)
277 goto err_ip;
278 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
279 GFP_KERNEL);
280 if (!idev->stats.icmpv6dev)
281 goto err_icmp;
282 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
283 GFP_KERNEL);
284 if (!idev->stats.icmpv6msgdev)
285 goto err_icmpmsg;
287 return 0;
289 err_icmpmsg:
290 kfree(idev->stats.icmpv6dev);
291 err_icmp:
292 snmp_mib_free((void __percpu **)idev->stats.ipv6);
293 err_ip:
294 return -ENOMEM;
297 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
299 struct inet6_dev *ndev;
301 ASSERT_RTNL();
303 if (dev->mtu < IPV6_MIN_MTU)
304 return NULL;
306 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
308 if (ndev == NULL)
309 return NULL;
311 rwlock_init(&ndev->lock);
312 ndev->dev = dev;
313 INIT_LIST_HEAD(&ndev->addr_list);
314 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
315 (unsigned long)ndev);
316 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
317 ndev->cnf.mtu6 = dev->mtu;
318 ndev->cnf.sysctl = NULL;
319 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
320 if (ndev->nd_parms == NULL) {
321 kfree(ndev);
322 return NULL;
324 if (ndev->cnf.forwarding)
325 dev_disable_lro(dev);
326 /* We refer to the device */
327 dev_hold(dev);
329 if (snmp6_alloc_dev(ndev) < 0) {
330 ADBG(KERN_WARNING
331 "%s: cannot allocate memory for statistics; dev=%s.\n",
332 __func__, dev->name);
333 neigh_parms_release(&nd_tbl, ndev->nd_parms);
334 dev_put(dev);
335 kfree(ndev);
336 return NULL;
339 if (snmp6_register_dev(ndev) < 0) {
340 ADBG(KERN_WARNING
341 "%s: cannot create /proc/net/dev_snmp6/%s\n",
342 __func__, dev->name);
343 neigh_parms_release(&nd_tbl, ndev->nd_parms);
344 ndev->dead = 1;
345 in6_dev_finish_destroy(ndev);
346 return NULL;
349 /* One reference from device. We must do this before
350 * we invoke __ipv6_regen_rndid().
352 in6_dev_hold(ndev);
354 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
355 ndev->cnf.accept_dad = -1;
357 #if IS_ENABLED(CONFIG_IPV6_SIT)
358 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
359 pr_info("%s: Disabled Multicast RS\n", dev->name);
360 ndev->cnf.rtr_solicits = 0;
362 #endif
364 INIT_LIST_HEAD(&ndev->tempaddr_list);
365 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
366 if ((dev->flags&IFF_LOOPBACK) ||
367 dev->type == ARPHRD_TUNNEL ||
368 dev->type == ARPHRD_TUNNEL6 ||
369 dev->type == ARPHRD_SIT ||
370 dev->type == ARPHRD_NONE) {
371 ndev->cnf.use_tempaddr = -1;
372 } else {
373 in6_dev_hold(ndev);
374 ipv6_regen_rndid((unsigned long) ndev);
377 ndev->token = in6addr_any;
379 if (netif_running(dev) && addrconf_qdisc_ok(dev))
380 ndev->if_flags |= IF_READY;
382 ipv6_mc_init_dev(ndev);
383 ndev->tstamp = jiffies;
384 addrconf_sysctl_register(ndev);
385 /* protected by rtnl_lock */
386 rcu_assign_pointer(dev->ip6_ptr, ndev);
388 /* Join interface-local all-node multicast group */
389 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
391 /* Join all-node multicast group */
392 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
394 /* Join all-router multicast group if forwarding is set */
395 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
396 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
398 return ndev;
401 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
403 struct inet6_dev *idev;
405 ASSERT_RTNL();
407 idev = __in6_dev_get(dev);
408 if (!idev) {
409 idev = ipv6_add_dev(dev);
410 if (!idev)
411 return NULL;
414 if (dev->flags&IFF_UP)
415 ipv6_mc_up(idev);
416 return idev;
419 static int inet6_netconf_msgsize_devconf(int type)
421 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
422 + nla_total_size(4); /* NETCONFA_IFINDEX */
424 /* type -1 is used for ALL */
425 if (type == -1 || type == NETCONFA_FORWARDING)
426 size += nla_total_size(4);
427 #ifdef CONFIG_IPV6_MROUTE
428 if (type == -1 || type == NETCONFA_MC_FORWARDING)
429 size += nla_total_size(4);
430 #endif
432 return size;
435 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
436 struct ipv6_devconf *devconf, u32 portid,
437 u32 seq, int event, unsigned int flags,
438 int type)
440 struct nlmsghdr *nlh;
441 struct netconfmsg *ncm;
443 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
444 flags);
445 if (nlh == NULL)
446 return -EMSGSIZE;
448 ncm = nlmsg_data(nlh);
449 ncm->ncm_family = AF_INET6;
451 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
452 goto nla_put_failure;
454 /* type -1 is used for ALL */
455 if ((type == -1 || type == NETCONFA_FORWARDING) &&
456 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
457 goto nla_put_failure;
458 #ifdef CONFIG_IPV6_MROUTE
459 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
460 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
461 devconf->mc_forwarding) < 0)
462 goto nla_put_failure;
463 #endif
464 return nlmsg_end(skb, nlh);
466 nla_put_failure:
467 nlmsg_cancel(skb, nlh);
468 return -EMSGSIZE;
471 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
472 struct ipv6_devconf *devconf)
474 struct sk_buff *skb;
475 int err = -ENOBUFS;
477 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
478 if (skb == NULL)
479 goto errout;
481 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
482 RTM_NEWNETCONF, 0, type);
483 if (err < 0) {
484 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
485 WARN_ON(err == -EMSGSIZE);
486 kfree_skb(skb);
487 goto errout;
489 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
490 return;
491 errout:
492 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
495 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
496 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
497 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
500 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
501 struct nlmsghdr *nlh)
503 struct net *net = sock_net(in_skb->sk);
504 struct nlattr *tb[NETCONFA_MAX+1];
505 struct netconfmsg *ncm;
506 struct sk_buff *skb;
507 struct ipv6_devconf *devconf;
508 struct inet6_dev *in6_dev;
509 struct net_device *dev;
510 int ifindex;
511 int err;
513 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
514 devconf_ipv6_policy);
515 if (err < 0)
516 goto errout;
518 err = EINVAL;
519 if (!tb[NETCONFA_IFINDEX])
520 goto errout;
522 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
523 switch (ifindex) {
524 case NETCONFA_IFINDEX_ALL:
525 devconf = net->ipv6.devconf_all;
526 break;
527 case NETCONFA_IFINDEX_DEFAULT:
528 devconf = net->ipv6.devconf_dflt;
529 break;
530 default:
531 dev = __dev_get_by_index(net, ifindex);
532 if (dev == NULL)
533 goto errout;
534 in6_dev = __in6_dev_get(dev);
535 if (in6_dev == NULL)
536 goto errout;
537 devconf = &in6_dev->cnf;
538 break;
541 err = -ENOBUFS;
542 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
543 if (skb == NULL)
544 goto errout;
546 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
547 NETLINK_CB(in_skb).portid,
548 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
549 -1);
550 if (err < 0) {
551 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
552 WARN_ON(err == -EMSGSIZE);
553 kfree_skb(skb);
554 goto errout;
556 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
557 errout:
558 return err;
561 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
562 struct netlink_callback *cb)
564 struct net *net = sock_net(skb->sk);
565 int h, s_h;
566 int idx, s_idx;
567 struct net_device *dev;
568 struct inet6_dev *idev;
569 struct hlist_head *head;
571 s_h = cb->args[0];
572 s_idx = idx = cb->args[1];
574 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
575 idx = 0;
576 head = &net->dev_index_head[h];
577 rcu_read_lock();
578 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
579 net->dev_base_seq;
580 hlist_for_each_entry_rcu(dev, head, index_hlist) {
581 if (idx < s_idx)
582 goto cont;
583 idev = __in6_dev_get(dev);
584 if (!idev)
585 goto cont;
587 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
588 &idev->cnf,
589 NETLINK_CB(cb->skb).portid,
590 cb->nlh->nlmsg_seq,
591 RTM_NEWNETCONF,
592 NLM_F_MULTI,
593 -1) <= 0) {
594 rcu_read_unlock();
595 goto done;
597 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
598 cont:
599 idx++;
601 rcu_read_unlock();
603 if (h == NETDEV_HASHENTRIES) {
604 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
605 net->ipv6.devconf_all,
606 NETLINK_CB(cb->skb).portid,
607 cb->nlh->nlmsg_seq,
608 RTM_NEWNETCONF, NLM_F_MULTI,
609 -1) <= 0)
610 goto done;
611 else
612 h++;
614 if (h == NETDEV_HASHENTRIES + 1) {
615 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
616 net->ipv6.devconf_dflt,
617 NETLINK_CB(cb->skb).portid,
618 cb->nlh->nlmsg_seq,
619 RTM_NEWNETCONF, NLM_F_MULTI,
620 -1) <= 0)
621 goto done;
622 else
623 h++;
625 done:
626 cb->args[0] = h;
627 cb->args[1] = idx;
629 return skb->len;
632 #ifdef CONFIG_SYSCTL
633 static void dev_forward_change(struct inet6_dev *idev)
635 struct net_device *dev;
636 struct inet6_ifaddr *ifa;
638 if (!idev)
639 return;
640 dev = idev->dev;
641 if (idev->cnf.forwarding)
642 dev_disable_lro(dev);
643 if (dev->flags & IFF_MULTICAST) {
644 if (idev->cnf.forwarding) {
645 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
646 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
647 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
648 } else {
649 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
650 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
651 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
655 list_for_each_entry(ifa, &idev->addr_list, if_list) {
656 if (ifa->flags&IFA_F_TENTATIVE)
657 continue;
658 if (idev->cnf.forwarding)
659 addrconf_join_anycast(ifa);
660 else
661 addrconf_leave_anycast(ifa);
663 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
664 dev->ifindex, &idev->cnf);
668 static void addrconf_forward_change(struct net *net, __s32 newf)
670 struct net_device *dev;
671 struct inet6_dev *idev;
673 for_each_netdev(net, dev) {
674 idev = __in6_dev_get(dev);
675 if (idev) {
676 int changed = (!idev->cnf.forwarding) ^ (!newf);
677 idev->cnf.forwarding = newf;
678 if (changed)
679 dev_forward_change(idev);
684 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
686 struct net *net;
687 int old;
689 if (!rtnl_trylock())
690 return restart_syscall();
692 net = (struct net *)table->extra2;
693 old = *p;
694 *p = newf;
696 if (p == &net->ipv6.devconf_dflt->forwarding) {
697 if ((!newf) ^ (!old))
698 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
699 NETCONFA_IFINDEX_DEFAULT,
700 net->ipv6.devconf_dflt);
701 rtnl_unlock();
702 return 0;
705 if (p == &net->ipv6.devconf_all->forwarding) {
706 net->ipv6.devconf_dflt->forwarding = newf;
707 addrconf_forward_change(net, newf);
708 if ((!newf) ^ (!old))
709 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
710 NETCONFA_IFINDEX_ALL,
711 net->ipv6.devconf_all);
712 } else if ((!newf) ^ (!old))
713 dev_forward_change((struct inet6_dev *)table->extra1);
714 rtnl_unlock();
716 if (newf)
717 rt6_purge_dflt_routers(net);
718 return 1;
720 #endif
722 /* Nobody refers to this ifaddr, destroy it */
723 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
725 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
727 #ifdef NET_REFCNT_DEBUG
728 pr_debug("%s\n", __func__);
729 #endif
731 in6_dev_put(ifp->idev);
733 if (del_timer(&ifp->dad_timer))
734 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
736 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
737 pr_warn("Freeing alive inet6 address %p\n", ifp);
738 return;
740 ip6_rt_put(ifp->rt);
742 kfree_rcu(ifp, rcu);
745 static void
746 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
748 struct list_head *p;
749 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
752 * Each device address list is sorted in order of scope -
753 * global before linklocal.
755 list_for_each(p, &idev->addr_list) {
756 struct inet6_ifaddr *ifa
757 = list_entry(p, struct inet6_ifaddr, if_list);
758 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
759 break;
762 list_add_tail(&ifp->if_list, p);
765 static u32 inet6_addr_hash(const struct in6_addr *addr)
767 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
770 /* On success it returns ifp with increased reference count */
772 static struct inet6_ifaddr *
773 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
774 const struct in6_addr *peer_addr, int pfxlen,
775 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
777 struct inet6_ifaddr *ifa = NULL;
778 struct rt6_info *rt;
779 unsigned int hash;
780 int err = 0;
781 int addr_type = ipv6_addr_type(addr);
783 if (addr_type == IPV6_ADDR_ANY ||
784 addr_type & IPV6_ADDR_MULTICAST ||
785 (!(idev->dev->flags & IFF_LOOPBACK) &&
786 addr_type & IPV6_ADDR_LOOPBACK))
787 return ERR_PTR(-EADDRNOTAVAIL);
789 rcu_read_lock_bh();
790 if (idev->dead) {
791 err = -ENODEV; /*XXX*/
792 goto out2;
795 if (idev->cnf.disable_ipv6) {
796 err = -EACCES;
797 goto out2;
800 spin_lock(&addrconf_hash_lock);
802 /* Ignore adding duplicate addresses on an interface */
803 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
804 ADBG("ipv6_add_addr: already assigned\n");
805 err = -EEXIST;
806 goto out;
809 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
811 if (ifa == NULL) {
812 ADBG("ipv6_add_addr: malloc failed\n");
813 err = -ENOBUFS;
814 goto out;
817 rt = addrconf_dst_alloc(idev, addr, false);
818 if (IS_ERR(rt)) {
819 err = PTR_ERR(rt);
820 goto out;
823 ifa->addr = *addr;
824 if (peer_addr)
825 ifa->peer_addr = *peer_addr;
827 spin_lock_init(&ifa->lock);
828 spin_lock_init(&ifa->state_lock);
829 setup_timer(&ifa->dad_timer, addrconf_dad_timer,
830 (unsigned long)ifa);
831 INIT_HLIST_NODE(&ifa->addr_lst);
832 ifa->scope = scope;
833 ifa->prefix_len = pfxlen;
834 ifa->flags = flags | IFA_F_TENTATIVE;
835 ifa->valid_lft = valid_lft;
836 ifa->prefered_lft = prefered_lft;
837 ifa->cstamp = ifa->tstamp = jiffies;
838 ifa->tokenized = false;
840 ifa->rt = rt;
842 ifa->idev = idev;
843 in6_dev_hold(idev);
844 /* For caller */
845 in6_ifa_hold(ifa);
847 /* Add to big hash table */
848 hash = inet6_addr_hash(addr);
850 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
851 spin_unlock(&addrconf_hash_lock);
853 write_lock(&idev->lock);
854 /* Add to inet6_dev unicast addr list. */
855 ipv6_link_dev_addr(idev, ifa);
857 if (ifa->flags&IFA_F_TEMPORARY) {
858 list_add(&ifa->tmp_list, &idev->tempaddr_list);
859 in6_ifa_hold(ifa);
862 in6_ifa_hold(ifa);
863 write_unlock(&idev->lock);
864 out2:
865 rcu_read_unlock_bh();
867 if (likely(err == 0))
868 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
869 else {
870 kfree(ifa);
871 ifa = ERR_PTR(err);
874 return ifa;
875 out:
876 spin_unlock(&addrconf_hash_lock);
877 goto out2;
880 /* This function wants to get referenced ifp and releases it before return */
882 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
884 struct inet6_ifaddr *ifa, *ifn;
885 struct inet6_dev *idev = ifp->idev;
886 int state;
887 int deleted = 0, onlink = 0;
888 unsigned long expires = jiffies;
890 spin_lock_bh(&ifp->state_lock);
891 state = ifp->state;
892 ifp->state = INET6_IFADDR_STATE_DEAD;
893 spin_unlock_bh(&ifp->state_lock);
895 if (state == INET6_IFADDR_STATE_DEAD)
896 goto out;
898 spin_lock_bh(&addrconf_hash_lock);
899 hlist_del_init_rcu(&ifp->addr_lst);
900 spin_unlock_bh(&addrconf_hash_lock);
902 write_lock_bh(&idev->lock);
904 if (ifp->flags&IFA_F_TEMPORARY) {
905 list_del(&ifp->tmp_list);
906 if (ifp->ifpub) {
907 in6_ifa_put(ifp->ifpub);
908 ifp->ifpub = NULL;
910 __in6_ifa_put(ifp);
913 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
914 if (ifa == ifp) {
915 list_del_init(&ifp->if_list);
916 __in6_ifa_put(ifp);
918 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
919 break;
920 deleted = 1;
921 continue;
922 } else if (ifp->flags & IFA_F_PERMANENT) {
923 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
924 ifp->prefix_len)) {
925 if (ifa->flags & IFA_F_PERMANENT) {
926 onlink = 1;
927 if (deleted)
928 break;
929 } else {
930 unsigned long lifetime;
932 if (!onlink)
933 onlink = -1;
935 spin_lock(&ifa->lock);
937 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
939 * Note: Because this address is
940 * not permanent, lifetime <
941 * LONG_MAX / HZ here.
943 if (time_before(expires,
944 ifa->tstamp + lifetime * HZ))
945 expires = ifa->tstamp + lifetime * HZ;
946 spin_unlock(&ifa->lock);
951 write_unlock_bh(&idev->lock);
953 addrconf_del_dad_timer(ifp);
955 ipv6_ifa_notify(RTM_DELADDR, ifp);
957 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
960 * Purge or update corresponding prefix
962 * 1) we don't purge prefix here if address was not permanent.
963 * prefix is managed by its own lifetime.
964 * 2) if there're no addresses, delete prefix.
965 * 3) if there're still other permanent address(es),
966 * corresponding prefix is still permanent.
967 * 4) otherwise, update prefix lifetime to the
968 * longest valid lifetime among the corresponding
969 * addresses on the device.
970 * Note: subsequent RA will update lifetime.
972 * --yoshfuji
974 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
975 struct in6_addr prefix;
976 struct rt6_info *rt;
978 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
980 rt = addrconf_get_prefix_route(&prefix,
981 ifp->prefix_len,
982 ifp->idev->dev,
983 0, RTF_GATEWAY | RTF_DEFAULT);
985 if (rt) {
986 if (onlink == 0) {
987 ip6_del_rt(rt);
988 rt = NULL;
989 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
990 rt6_set_expires(rt, expires);
993 ip6_rt_put(rt);
996 /* clean up prefsrc entries */
997 rt6_remove_prefsrc(ifp);
998 out:
999 in6_ifa_put(ifp);
1002 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1004 struct inet6_dev *idev = ifp->idev;
1005 struct in6_addr addr, *tmpaddr;
1006 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1007 unsigned long regen_advance;
1008 int tmp_plen;
1009 int ret = 0;
1010 u32 addr_flags;
1011 unsigned long now = jiffies;
1013 write_lock(&idev->lock);
1014 if (ift) {
1015 spin_lock_bh(&ift->lock);
1016 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1017 spin_unlock_bh(&ift->lock);
1018 tmpaddr = &addr;
1019 } else {
1020 tmpaddr = NULL;
1022 retry:
1023 in6_dev_hold(idev);
1024 if (idev->cnf.use_tempaddr <= 0) {
1025 write_unlock(&idev->lock);
1026 pr_info("%s: use_tempaddr is disabled\n", __func__);
1027 in6_dev_put(idev);
1028 ret = -1;
1029 goto out;
1031 spin_lock_bh(&ifp->lock);
1032 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1033 idev->cnf.use_tempaddr = -1; /*XXX*/
1034 spin_unlock_bh(&ifp->lock);
1035 write_unlock(&idev->lock);
1036 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1037 __func__);
1038 in6_dev_put(idev);
1039 ret = -1;
1040 goto out;
1042 in6_ifa_hold(ifp);
1043 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1044 __ipv6_try_regen_rndid(idev, tmpaddr);
1045 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1046 age = (now - ifp->tstamp) / HZ;
1047 tmp_valid_lft = min_t(__u32,
1048 ifp->valid_lft,
1049 idev->cnf.temp_valid_lft + age);
1050 tmp_prefered_lft = min_t(__u32,
1051 ifp->prefered_lft,
1052 idev->cnf.temp_prefered_lft + age -
1053 idev->cnf.max_desync_factor);
1054 tmp_plen = ifp->prefix_len;
1055 tmp_tstamp = ifp->tstamp;
1056 spin_unlock_bh(&ifp->lock);
1058 regen_advance = idev->cnf.regen_max_retry *
1059 idev->cnf.dad_transmits *
1060 idev->nd_parms->retrans_time / HZ;
1061 write_unlock(&idev->lock);
1063 /* A temporary address is created only if this calculated Preferred
1064 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1065 * an implementation must not create a temporary address with a zero
1066 * Preferred Lifetime.
1068 if (tmp_prefered_lft <= regen_advance) {
1069 in6_ifa_put(ifp);
1070 in6_dev_put(idev);
1071 ret = -1;
1072 goto out;
1075 addr_flags = IFA_F_TEMPORARY;
1076 /* set in addrconf_prefix_rcv() */
1077 if (ifp->flags & IFA_F_OPTIMISTIC)
1078 addr_flags |= IFA_F_OPTIMISTIC;
1080 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1081 ipv6_addr_scope(&addr), addr_flags,
1082 tmp_valid_lft, tmp_prefered_lft);
1083 if (IS_ERR(ift)) {
1084 in6_ifa_put(ifp);
1085 in6_dev_put(idev);
1086 pr_info("%s: retry temporary address regeneration\n", __func__);
1087 tmpaddr = &addr;
1088 write_lock(&idev->lock);
1089 goto retry;
1092 spin_lock_bh(&ift->lock);
1093 ift->ifpub = ifp;
1094 ift->cstamp = now;
1095 ift->tstamp = tmp_tstamp;
1096 spin_unlock_bh(&ift->lock);
1098 addrconf_dad_start(ift);
1099 in6_ifa_put(ift);
1100 in6_dev_put(idev);
1101 out:
1102 return ret;
1106 * Choose an appropriate source address (RFC3484)
1108 enum {
1109 IPV6_SADDR_RULE_INIT = 0,
1110 IPV6_SADDR_RULE_LOCAL,
1111 IPV6_SADDR_RULE_SCOPE,
1112 IPV6_SADDR_RULE_PREFERRED,
1113 #ifdef CONFIG_IPV6_MIP6
1114 IPV6_SADDR_RULE_HOA,
1115 #endif
1116 IPV6_SADDR_RULE_OIF,
1117 IPV6_SADDR_RULE_LABEL,
1118 IPV6_SADDR_RULE_PRIVACY,
1119 IPV6_SADDR_RULE_ORCHID,
1120 IPV6_SADDR_RULE_PREFIX,
1121 IPV6_SADDR_RULE_MAX
1124 struct ipv6_saddr_score {
1125 int rule;
1126 int addr_type;
1127 struct inet6_ifaddr *ifa;
1128 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1129 int scopedist;
1130 int matchlen;
1133 struct ipv6_saddr_dst {
1134 const struct in6_addr *addr;
1135 int ifindex;
1136 int scope;
1137 int label;
1138 unsigned int prefs;
1141 static inline int ipv6_saddr_preferred(int type)
1143 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1144 return 1;
1145 return 0;
1148 static int ipv6_get_saddr_eval(struct net *net,
1149 struct ipv6_saddr_score *score,
1150 struct ipv6_saddr_dst *dst,
1151 int i)
1153 int ret;
1155 if (i <= score->rule) {
1156 switch (i) {
1157 case IPV6_SADDR_RULE_SCOPE:
1158 ret = score->scopedist;
1159 break;
1160 case IPV6_SADDR_RULE_PREFIX:
1161 ret = score->matchlen;
1162 break;
1163 default:
1164 ret = !!test_bit(i, score->scorebits);
1166 goto out;
1169 switch (i) {
1170 case IPV6_SADDR_RULE_INIT:
1171 /* Rule 0: remember if hiscore is not ready yet */
1172 ret = !!score->ifa;
1173 break;
1174 case IPV6_SADDR_RULE_LOCAL:
1175 /* Rule 1: Prefer same address */
1176 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1177 break;
1178 case IPV6_SADDR_RULE_SCOPE:
1179 /* Rule 2: Prefer appropriate scope
1181 * ret
1183 * -1 | d 15
1184 * ---+--+-+---> scope
1186 * | d is scope of the destination.
1187 * B-d | \
1188 * | \ <- smaller scope is better if
1189 * B-15 | \ if scope is enough for destinaion.
1190 * | ret = B - scope (-1 <= scope >= d <= 15).
1191 * d-C-1 | /
1192 * |/ <- greater is better
1193 * -C / if scope is not enough for destination.
1194 * /| ret = scope - C (-1 <= d < scope <= 15).
1196 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1197 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1198 * Assume B = 0 and we get C > 29.
1200 ret = __ipv6_addr_src_scope(score->addr_type);
1201 if (ret >= dst->scope)
1202 ret = -ret;
1203 else
1204 ret -= 128; /* 30 is enough */
1205 score->scopedist = ret;
1206 break;
1207 case IPV6_SADDR_RULE_PREFERRED:
1208 /* Rule 3: Avoid deprecated and optimistic addresses */
1209 ret = ipv6_saddr_preferred(score->addr_type) ||
1210 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1211 break;
1212 #ifdef CONFIG_IPV6_MIP6
1213 case IPV6_SADDR_RULE_HOA:
1215 /* Rule 4: Prefer home address */
1216 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1217 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1218 break;
1220 #endif
1221 case IPV6_SADDR_RULE_OIF:
1222 /* Rule 5: Prefer outgoing interface */
1223 ret = (!dst->ifindex ||
1224 dst->ifindex == score->ifa->idev->dev->ifindex);
1225 break;
1226 case IPV6_SADDR_RULE_LABEL:
1227 /* Rule 6: Prefer matching label */
1228 ret = ipv6_addr_label(net,
1229 &score->ifa->addr, score->addr_type,
1230 score->ifa->idev->dev->ifindex) == dst->label;
1231 break;
1232 case IPV6_SADDR_RULE_PRIVACY:
1234 /* Rule 7: Prefer public address
1235 * Note: prefer temporary address if use_tempaddr >= 2
1237 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1238 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1239 score->ifa->idev->cnf.use_tempaddr >= 2;
1240 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1241 break;
1243 case IPV6_SADDR_RULE_ORCHID:
1244 /* Rule 8-: Prefer ORCHID vs ORCHID or
1245 * non-ORCHID vs non-ORCHID
1247 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1248 ipv6_addr_orchid(dst->addr));
1249 break;
1250 case IPV6_SADDR_RULE_PREFIX:
1251 /* Rule 8: Use longest matching prefix */
1252 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1253 if (ret > score->ifa->prefix_len)
1254 ret = score->ifa->prefix_len;
1255 score->matchlen = ret;
1256 break;
1257 default:
1258 ret = 0;
1261 if (ret)
1262 __set_bit(i, score->scorebits);
1263 score->rule = i;
1264 out:
1265 return ret;
1268 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1269 const struct in6_addr *daddr, unsigned int prefs,
1270 struct in6_addr *saddr)
1272 struct ipv6_saddr_score scores[2],
1273 *score = &scores[0], *hiscore = &scores[1];
1274 struct ipv6_saddr_dst dst;
1275 struct net_device *dev;
1276 int dst_type;
1278 dst_type = __ipv6_addr_type(daddr);
1279 dst.addr = daddr;
1280 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1281 dst.scope = __ipv6_addr_src_scope(dst_type);
1282 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1283 dst.prefs = prefs;
1285 hiscore->rule = -1;
1286 hiscore->ifa = NULL;
1288 rcu_read_lock();
1290 for_each_netdev_rcu(net, dev) {
1291 struct inet6_dev *idev;
1293 /* Candidate Source Address (section 4)
1294 * - multicast and link-local destination address,
1295 * the set of candidate source address MUST only
1296 * include addresses assigned to interfaces
1297 * belonging to the same link as the outgoing
1298 * interface.
1299 * (- For site-local destination addresses, the
1300 * set of candidate source addresses MUST only
1301 * include addresses assigned to interfaces
1302 * belonging to the same site as the outgoing
1303 * interface.)
1305 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1306 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1307 dst.ifindex && dev->ifindex != dst.ifindex)
1308 continue;
1310 idev = __in6_dev_get(dev);
1311 if (!idev)
1312 continue;
1314 read_lock_bh(&idev->lock);
1315 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1316 int i;
1319 * - Tentative Address (RFC2462 section 5.4)
1320 * - A tentative address is not considered
1321 * "assigned to an interface" in the traditional
1322 * sense, unless it is also flagged as optimistic.
1323 * - Candidate Source Address (section 4)
1324 * - In any case, anycast addresses, multicast
1325 * addresses, and the unspecified address MUST
1326 * NOT be included in a candidate set.
1328 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1329 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1330 continue;
1332 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1334 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1335 score->addr_type & IPV6_ADDR_MULTICAST)) {
1336 LIMIT_NETDEBUG(KERN_DEBUG
1337 "ADDRCONF: unspecified / multicast address "
1338 "assigned as unicast address on %s",
1339 dev->name);
1340 continue;
1343 score->rule = -1;
1344 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1346 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1347 int minihiscore, miniscore;
1349 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1350 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1352 if (minihiscore > miniscore) {
1353 if (i == IPV6_SADDR_RULE_SCOPE &&
1354 score->scopedist > 0) {
1356 * special case:
1357 * each remaining entry
1358 * has too small (not enough)
1359 * scope, because ifa entries
1360 * are sorted by their scope
1361 * values.
1363 goto try_nextdev;
1365 break;
1366 } else if (minihiscore < miniscore) {
1367 if (hiscore->ifa)
1368 in6_ifa_put(hiscore->ifa);
1370 in6_ifa_hold(score->ifa);
1372 swap(hiscore, score);
1374 /* restore our iterator */
1375 score->ifa = hiscore->ifa;
1377 break;
1381 try_nextdev:
1382 read_unlock_bh(&idev->lock);
1384 rcu_read_unlock();
1386 if (!hiscore->ifa)
1387 return -EADDRNOTAVAIL;
1389 *saddr = hiscore->ifa->addr;
1390 in6_ifa_put(hiscore->ifa);
1391 return 0;
1393 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1395 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1396 unsigned char banned_flags)
1398 struct inet6_ifaddr *ifp;
1399 int err = -EADDRNOTAVAIL;
1401 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1402 if (ifp->scope == IFA_LINK &&
1403 !(ifp->flags & banned_flags)) {
1404 *addr = ifp->addr;
1405 err = 0;
1406 break;
1409 return err;
1412 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1413 unsigned char banned_flags)
1415 struct inet6_dev *idev;
1416 int err = -EADDRNOTAVAIL;
1418 rcu_read_lock();
1419 idev = __in6_dev_get(dev);
1420 if (idev) {
1421 read_lock_bh(&idev->lock);
1422 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1423 read_unlock_bh(&idev->lock);
1425 rcu_read_unlock();
1426 return err;
1429 static int ipv6_count_addresses(struct inet6_dev *idev)
1431 int cnt = 0;
1432 struct inet6_ifaddr *ifp;
1434 read_lock_bh(&idev->lock);
1435 list_for_each_entry(ifp, &idev->addr_list, if_list)
1436 cnt++;
1437 read_unlock_bh(&idev->lock);
1438 return cnt;
1441 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1442 const struct net_device *dev, int strict)
1444 struct inet6_ifaddr *ifp;
1445 unsigned int hash = inet6_addr_hash(addr);
1447 rcu_read_lock_bh();
1448 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1449 if (!net_eq(dev_net(ifp->idev->dev), net))
1450 continue;
1451 if (ipv6_addr_equal(&ifp->addr, addr) &&
1452 !(ifp->flags&IFA_F_TENTATIVE) &&
1453 (dev == NULL || ifp->idev->dev == dev ||
1454 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1455 rcu_read_unlock_bh();
1456 return 1;
1460 rcu_read_unlock_bh();
1461 return 0;
1463 EXPORT_SYMBOL(ipv6_chk_addr);
1465 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1466 struct net_device *dev)
1468 unsigned int hash = inet6_addr_hash(addr);
1469 struct inet6_ifaddr *ifp;
1471 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1472 if (!net_eq(dev_net(ifp->idev->dev), net))
1473 continue;
1474 if (ipv6_addr_equal(&ifp->addr, addr)) {
1475 if (dev == NULL || ifp->idev->dev == dev)
1476 return true;
1479 return false;
1482 /* Compares an address/prefix_len with addresses on device @dev.
1483 * If one is found it returns true.
1485 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1486 const unsigned int prefix_len, struct net_device *dev)
1488 struct inet6_dev *idev;
1489 struct inet6_ifaddr *ifa;
1490 bool ret = false;
1492 rcu_read_lock();
1493 idev = __in6_dev_get(dev);
1494 if (idev) {
1495 read_lock_bh(&idev->lock);
1496 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1497 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1498 if (ret)
1499 break;
1501 read_unlock_bh(&idev->lock);
1503 rcu_read_unlock();
1505 return ret;
1507 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1509 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1511 struct inet6_dev *idev;
1512 struct inet6_ifaddr *ifa;
1513 int onlink;
1515 onlink = 0;
1516 rcu_read_lock();
1517 idev = __in6_dev_get(dev);
1518 if (idev) {
1519 read_lock_bh(&idev->lock);
1520 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1521 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1522 ifa->prefix_len);
1523 if (onlink)
1524 break;
1526 read_unlock_bh(&idev->lock);
1528 rcu_read_unlock();
1529 return onlink;
1531 EXPORT_SYMBOL(ipv6_chk_prefix);
1533 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1534 struct net_device *dev, int strict)
1536 struct inet6_ifaddr *ifp, *result = NULL;
1537 unsigned int hash = inet6_addr_hash(addr);
1539 rcu_read_lock_bh();
1540 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1541 if (!net_eq(dev_net(ifp->idev->dev), net))
1542 continue;
1543 if (ipv6_addr_equal(&ifp->addr, addr)) {
1544 if (dev == NULL || ifp->idev->dev == dev ||
1545 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1546 result = ifp;
1547 in6_ifa_hold(ifp);
1548 break;
1552 rcu_read_unlock_bh();
1554 return result;
1557 /* Gets referenced address, destroys ifaddr */
1559 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1561 if (ifp->flags&IFA_F_PERMANENT) {
1562 spin_lock_bh(&ifp->lock);
1563 addrconf_del_dad_timer(ifp);
1564 ifp->flags |= IFA_F_TENTATIVE;
1565 if (dad_failed)
1566 ifp->flags |= IFA_F_DADFAILED;
1567 spin_unlock_bh(&ifp->lock);
1568 if (dad_failed)
1569 ipv6_ifa_notify(0, ifp);
1570 in6_ifa_put(ifp);
1571 } else if (ifp->flags&IFA_F_TEMPORARY) {
1572 struct inet6_ifaddr *ifpub;
1573 spin_lock_bh(&ifp->lock);
1574 ifpub = ifp->ifpub;
1575 if (ifpub) {
1576 in6_ifa_hold(ifpub);
1577 spin_unlock_bh(&ifp->lock);
1578 ipv6_create_tempaddr(ifpub, ifp);
1579 in6_ifa_put(ifpub);
1580 } else {
1581 spin_unlock_bh(&ifp->lock);
1583 ipv6_del_addr(ifp);
1584 } else
1585 ipv6_del_addr(ifp);
1588 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1590 int err = -ENOENT;
1592 spin_lock(&ifp->state_lock);
1593 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1594 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1595 err = 0;
1597 spin_unlock(&ifp->state_lock);
1599 return err;
1602 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1604 struct inet6_dev *idev = ifp->idev;
1606 if (addrconf_dad_end(ifp)) {
1607 in6_ifa_put(ifp);
1608 return;
1611 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1612 ifp->idev->dev->name, &ifp->addr);
1614 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1615 struct in6_addr addr;
1617 addr.s6_addr32[0] = htonl(0xfe800000);
1618 addr.s6_addr32[1] = 0;
1620 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1621 ipv6_addr_equal(&ifp->addr, &addr)) {
1622 /* DAD failed for link-local based on MAC address */
1623 idev->cnf.disable_ipv6 = 1;
1625 pr_info("%s: IPv6 being disabled!\n",
1626 ifp->idev->dev->name);
1630 addrconf_dad_stop(ifp, 1);
1633 /* Join to solicited addr multicast group. */
1635 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1637 struct in6_addr maddr;
1639 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1640 return;
1642 addrconf_addr_solict_mult(addr, &maddr);
1643 ipv6_dev_mc_inc(dev, &maddr);
1646 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1648 struct in6_addr maddr;
1650 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1651 return;
1653 addrconf_addr_solict_mult(addr, &maddr);
1654 __ipv6_dev_mc_dec(idev, &maddr);
1657 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1659 struct in6_addr addr;
1660 if (ifp->prefix_len == 127) /* RFC 6164 */
1661 return;
1662 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1663 if (ipv6_addr_any(&addr))
1664 return;
1665 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1668 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1670 struct in6_addr addr;
1671 if (ifp->prefix_len == 127) /* RFC 6164 */
1672 return;
1673 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1674 if (ipv6_addr_any(&addr))
1675 return;
1676 __ipv6_dev_ac_dec(ifp->idev, &addr);
1679 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1681 if (dev->addr_len != ETH_ALEN)
1682 return -1;
1683 memcpy(eui, dev->dev_addr, 3);
1684 memcpy(eui + 5, dev->dev_addr + 3, 3);
1687 * The zSeries OSA network cards can be shared among various
1688 * OS instances, but the OSA cards have only one MAC address.
1689 * This leads to duplicate address conflicts in conjunction
1690 * with IPv6 if more than one instance uses the same card.
1692 * The driver for these cards can deliver a unique 16-bit
1693 * identifier for each instance sharing the same card. It is
1694 * placed instead of 0xFFFE in the interface identifier. The
1695 * "u" bit of the interface identifier is not inverted in this
1696 * case. Hence the resulting interface identifier has local
1697 * scope according to RFC2373.
1699 if (dev->dev_id) {
1700 eui[3] = (dev->dev_id >> 8) & 0xFF;
1701 eui[4] = dev->dev_id & 0xFF;
1702 } else {
1703 eui[3] = 0xFF;
1704 eui[4] = 0xFE;
1705 eui[0] ^= 2;
1707 return 0;
1710 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1712 if (dev->addr_len != IEEE802154_ADDR_LEN)
1713 return -1;
1714 memcpy(eui, dev->dev_addr, 8);
1715 eui[0] ^= 2;
1716 return 0;
1719 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1721 union fwnet_hwaddr *ha;
1723 if (dev->addr_len != FWNET_ALEN)
1724 return -1;
1726 ha = (union fwnet_hwaddr *)dev->dev_addr;
1728 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1729 eui[0] ^= 2;
1730 return 0;
1733 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1735 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1736 if (dev->addr_len != ARCNET_ALEN)
1737 return -1;
1738 memset(eui, 0, 7);
1739 eui[7] = *(u8 *)dev->dev_addr;
1740 return 0;
1743 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1745 if (dev->addr_len != INFINIBAND_ALEN)
1746 return -1;
1747 memcpy(eui, dev->dev_addr + 12, 8);
1748 eui[0] |= 2;
1749 return 0;
1752 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1754 if (addr == 0)
1755 return -1;
1756 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1757 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1758 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1759 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1760 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1761 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1762 eui[1] = 0;
1763 eui[2] = 0x5E;
1764 eui[3] = 0xFE;
1765 memcpy(eui + 4, &addr, 4);
1766 return 0;
1769 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1771 if (dev->priv_flags & IFF_ISATAP)
1772 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1773 return -1;
1776 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1778 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1781 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1783 memcpy(eui, dev->perm_addr, 3);
1784 memcpy(eui + 5, dev->perm_addr + 3, 3);
1785 eui[3] = 0xFF;
1786 eui[4] = 0xFE;
1787 eui[0] ^= 2;
1788 return 0;
1791 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1793 switch (dev->type) {
1794 case ARPHRD_ETHER:
1795 case ARPHRD_FDDI:
1796 return addrconf_ifid_eui48(eui, dev);
1797 case ARPHRD_ARCNET:
1798 return addrconf_ifid_arcnet(eui, dev);
1799 case ARPHRD_INFINIBAND:
1800 return addrconf_ifid_infiniband(eui, dev);
1801 case ARPHRD_SIT:
1802 return addrconf_ifid_sit(eui, dev);
1803 case ARPHRD_IPGRE:
1804 return addrconf_ifid_gre(eui, dev);
1805 case ARPHRD_IEEE802154:
1806 return addrconf_ifid_eui64(eui, dev);
1807 case ARPHRD_IEEE1394:
1808 return addrconf_ifid_ieee1394(eui, dev);
1809 case ARPHRD_TUNNEL6:
1810 return addrconf_ifid_ip6tnl(eui, dev);
1812 return -1;
1815 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1817 int err = -1;
1818 struct inet6_ifaddr *ifp;
1820 read_lock_bh(&idev->lock);
1821 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1822 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1823 memcpy(eui, ifp->addr.s6_addr+8, 8);
1824 err = 0;
1825 break;
1828 read_unlock_bh(&idev->lock);
1829 return err;
1832 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1833 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1835 regen:
1836 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1837 idev->rndid[0] &= ~0x02;
1840 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1841 * check if generated address is not inappropriate
1843 * - Reserved subnet anycast (RFC 2526)
1844 * 11111101 11....11 1xxxxxxx
1845 * - ISATAP (RFC4214) 6.1
1846 * 00-00-5E-FE-xx-xx-xx-xx
1847 * - value 0
1848 * - XXX: already assigned to an address on the device
1850 if (idev->rndid[0] == 0xfd &&
1851 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1852 (idev->rndid[7]&0x80))
1853 goto regen;
1854 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1855 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1856 goto regen;
1857 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1858 goto regen;
1862 static void ipv6_regen_rndid(unsigned long data)
1864 struct inet6_dev *idev = (struct inet6_dev *) data;
1865 unsigned long expires;
1867 rcu_read_lock_bh();
1868 write_lock_bh(&idev->lock);
1870 if (idev->dead)
1871 goto out;
1873 __ipv6_regen_rndid(idev);
1875 expires = jiffies +
1876 idev->cnf.temp_prefered_lft * HZ -
1877 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1878 idev->cnf.max_desync_factor * HZ;
1879 if (time_before(expires, jiffies)) {
1880 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1881 __func__, idev->dev->name);
1882 goto out;
1885 if (!mod_timer(&idev->regen_timer, expires))
1886 in6_dev_hold(idev);
1888 out:
1889 write_unlock_bh(&idev->lock);
1890 rcu_read_unlock_bh();
1891 in6_dev_put(idev);
1894 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1896 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1897 __ipv6_regen_rndid(idev);
1901 * Add prefix route.
1904 static void
1905 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1906 unsigned long expires, u32 flags)
1908 struct fib6_config cfg = {
1909 .fc_table = RT6_TABLE_PREFIX,
1910 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1911 .fc_ifindex = dev->ifindex,
1912 .fc_expires = expires,
1913 .fc_dst_len = plen,
1914 .fc_flags = RTF_UP | flags,
1915 .fc_nlinfo.nl_net = dev_net(dev),
1916 .fc_protocol = RTPROT_KERNEL,
1919 cfg.fc_dst = *pfx;
1921 /* Prevent useless cloning on PtP SIT.
1922 This thing is done here expecting that the whole
1923 class of non-broadcast devices need not cloning.
1925 #if IS_ENABLED(CONFIG_IPV6_SIT)
1926 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1927 cfg.fc_flags |= RTF_NONEXTHOP;
1928 #endif
1930 ip6_route_add(&cfg);
1934 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1935 int plen,
1936 const struct net_device *dev,
1937 u32 flags, u32 noflags)
1939 struct fib6_node *fn;
1940 struct rt6_info *rt = NULL;
1941 struct fib6_table *table;
1943 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1944 if (table == NULL)
1945 return NULL;
1947 read_lock_bh(&table->tb6_lock);
1948 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1949 if (!fn)
1950 goto out;
1951 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1952 if (rt->dst.dev->ifindex != dev->ifindex)
1953 continue;
1954 if ((rt->rt6i_flags & flags) != flags)
1955 continue;
1956 if ((rt->rt6i_flags & noflags) != 0)
1957 continue;
1958 dst_hold(&rt->dst);
1959 break;
1961 out:
1962 read_unlock_bh(&table->tb6_lock);
1963 return rt;
1967 /* Create "default" multicast route to the interface */
1969 static void addrconf_add_mroute(struct net_device *dev)
1971 struct fib6_config cfg = {
1972 .fc_table = RT6_TABLE_LOCAL,
1973 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1974 .fc_ifindex = dev->ifindex,
1975 .fc_dst_len = 8,
1976 .fc_flags = RTF_UP,
1977 .fc_nlinfo.nl_net = dev_net(dev),
1980 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1982 ip6_route_add(&cfg);
1985 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1987 struct inet6_dev *idev;
1989 ASSERT_RTNL();
1991 idev = ipv6_find_idev(dev);
1992 if (!idev)
1993 return ERR_PTR(-ENOBUFS);
1995 if (idev->cnf.disable_ipv6)
1996 return ERR_PTR(-EACCES);
1998 /* Add default multicast route */
1999 if (!(dev->flags & IFF_LOOPBACK))
2000 addrconf_add_mroute(dev);
2002 return idev;
2005 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2007 struct prefix_info *pinfo;
2008 __u32 valid_lft;
2009 __u32 prefered_lft;
2010 int addr_type;
2011 struct inet6_dev *in6_dev;
2012 struct net *net = dev_net(dev);
2014 pinfo = (struct prefix_info *) opt;
2016 if (len < sizeof(struct prefix_info)) {
2017 ADBG("addrconf: prefix option too short\n");
2018 return;
2022 * Validation checks ([ADDRCONF], page 19)
2025 addr_type = ipv6_addr_type(&pinfo->prefix);
2027 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2028 return;
2030 valid_lft = ntohl(pinfo->valid);
2031 prefered_lft = ntohl(pinfo->prefered);
2033 if (prefered_lft > valid_lft) {
2034 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2035 return;
2038 in6_dev = in6_dev_get(dev);
2040 if (in6_dev == NULL) {
2041 net_dbg_ratelimited("addrconf: device %s not configured\n",
2042 dev->name);
2043 return;
2047 * Two things going on here:
2048 * 1) Add routes for on-link prefixes
2049 * 2) Configure prefixes with the auto flag set
2052 if (pinfo->onlink) {
2053 struct rt6_info *rt;
2054 unsigned long rt_expires;
2056 /* Avoid arithmetic overflow. Really, we could
2057 * save rt_expires in seconds, likely valid_lft,
2058 * but it would require division in fib gc, that it
2059 * not good.
2061 if (HZ > USER_HZ)
2062 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2063 else
2064 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2066 if (addrconf_finite_timeout(rt_expires))
2067 rt_expires *= HZ;
2069 rt = addrconf_get_prefix_route(&pinfo->prefix,
2070 pinfo->prefix_len,
2071 dev,
2072 RTF_ADDRCONF | RTF_PREFIX_RT,
2073 RTF_GATEWAY | RTF_DEFAULT);
2075 if (rt) {
2076 /* Autoconf prefix route */
2077 if (valid_lft == 0) {
2078 ip6_del_rt(rt);
2079 rt = NULL;
2080 } else if (addrconf_finite_timeout(rt_expires)) {
2081 /* not infinity */
2082 rt6_set_expires(rt, jiffies + rt_expires);
2083 } else {
2084 rt6_clean_expires(rt);
2086 } else if (valid_lft) {
2087 clock_t expires = 0;
2088 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2089 if (addrconf_finite_timeout(rt_expires)) {
2090 /* not infinity */
2091 flags |= RTF_EXPIRES;
2092 expires = jiffies_to_clock_t(rt_expires);
2094 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2095 dev, expires, flags);
2097 ip6_rt_put(rt);
2100 /* Try to figure out our local address for this prefix */
2102 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2103 struct inet6_ifaddr *ifp;
2104 struct in6_addr addr;
2105 int create = 0, update_lft = 0;
2106 bool tokenized = false;
2108 if (pinfo->prefix_len == 64) {
2109 memcpy(&addr, &pinfo->prefix, 8);
2111 if (!ipv6_addr_any(&in6_dev->token)) {
2112 read_lock_bh(&in6_dev->lock);
2113 memcpy(addr.s6_addr + 8,
2114 in6_dev->token.s6_addr + 8, 8);
2115 read_unlock_bh(&in6_dev->lock);
2116 tokenized = true;
2117 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2118 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2119 in6_dev_put(in6_dev);
2120 return;
2122 goto ok;
2124 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2125 pinfo->prefix_len);
2126 in6_dev_put(in6_dev);
2127 return;
2131 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2133 if (ifp == NULL && valid_lft) {
2134 int max_addresses = in6_dev->cnf.max_addresses;
2135 u32 addr_flags = 0;
2137 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2138 if (in6_dev->cnf.optimistic_dad &&
2139 !net->ipv6.devconf_all->forwarding && sllao)
2140 addr_flags = IFA_F_OPTIMISTIC;
2141 #endif
2143 /* Do not allow to create too much of autoconfigured
2144 * addresses; this would be too easy way to crash kernel.
2146 if (!max_addresses ||
2147 ipv6_count_addresses(in6_dev) < max_addresses)
2148 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2149 pinfo->prefix_len,
2150 addr_type&IPV6_ADDR_SCOPE_MASK,
2151 addr_flags, valid_lft,
2152 prefered_lft);
2154 if (IS_ERR_OR_NULL(ifp)) {
2155 in6_dev_put(in6_dev);
2156 return;
2159 update_lft = 0;
2160 create = 1;
2161 ifp->cstamp = jiffies;
2162 ifp->tokenized = tokenized;
2163 addrconf_dad_start(ifp);
2166 if (ifp) {
2167 int flags;
2168 unsigned long now;
2169 struct inet6_ifaddr *ift;
2170 u32 stored_lft;
2172 /* update lifetime (RFC2462 5.5.3 e) */
2173 spin_lock(&ifp->lock);
2174 now = jiffies;
2175 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2176 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2177 else
2178 stored_lft = 0;
2179 if (!update_lft && !create && stored_lft) {
2180 const u32 minimum_lft = min(
2181 stored_lft, (u32)MIN_VALID_LIFETIME);
2182 valid_lft = max(valid_lft, minimum_lft);
2184 /* RFC4862 Section 5.5.3e:
2185 * "Note that the preferred lifetime of the
2186 * corresponding address is always reset to
2187 * the Preferred Lifetime in the received
2188 * Prefix Information option, regardless of
2189 * whether the valid lifetime is also reset or
2190 * ignored."
2192 * So we should always update prefered_lft here.
2194 update_lft = 1;
2197 if (update_lft) {
2198 ifp->valid_lft = valid_lft;
2199 ifp->prefered_lft = prefered_lft;
2200 ifp->tstamp = now;
2201 flags = ifp->flags;
2202 ifp->flags &= ~IFA_F_DEPRECATED;
2203 spin_unlock(&ifp->lock);
2205 if (!(flags&IFA_F_TENTATIVE))
2206 ipv6_ifa_notify(0, ifp);
2207 } else
2208 spin_unlock(&ifp->lock);
2210 read_lock_bh(&in6_dev->lock);
2211 /* update all temporary addresses in the list */
2212 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2213 tmp_list) {
2214 int age, max_valid, max_prefered;
2216 if (ifp != ift->ifpub)
2217 continue;
2220 * RFC 4941 section 3.3:
2221 * If a received option will extend the lifetime
2222 * of a public address, the lifetimes of
2223 * temporary addresses should be extended,
2224 * subject to the overall constraint that no
2225 * temporary addresses should ever remain
2226 * "valid" or "preferred" for a time longer than
2227 * (TEMP_VALID_LIFETIME) or
2228 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2229 * respectively.
2231 age = (now - ift->cstamp) / HZ;
2232 max_valid = in6_dev->cnf.temp_valid_lft - age;
2233 if (max_valid < 0)
2234 max_valid = 0;
2236 max_prefered = in6_dev->cnf.temp_prefered_lft -
2237 in6_dev->cnf.max_desync_factor -
2238 age;
2239 if (max_prefered < 0)
2240 max_prefered = 0;
2242 if (valid_lft > max_valid)
2243 valid_lft = max_valid;
2245 if (prefered_lft > max_prefered)
2246 prefered_lft = max_prefered;
2248 spin_lock(&ift->lock);
2249 flags = ift->flags;
2250 ift->valid_lft = valid_lft;
2251 ift->prefered_lft = prefered_lft;
2252 ift->tstamp = now;
2253 if (prefered_lft > 0)
2254 ift->flags &= ~IFA_F_DEPRECATED;
2256 spin_unlock(&ift->lock);
2257 if (!(flags&IFA_F_TENTATIVE))
2258 ipv6_ifa_notify(0, ift);
2261 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2263 * When a new public address is created as
2264 * described in [ADDRCONF], also create a new
2265 * temporary address. Also create a temporary
2266 * address if it's enabled but no temporary
2267 * address currently exists.
2269 read_unlock_bh(&in6_dev->lock);
2270 ipv6_create_tempaddr(ifp, NULL);
2271 } else {
2272 read_unlock_bh(&in6_dev->lock);
2275 in6_ifa_put(ifp);
2276 addrconf_verify(0);
2279 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2280 in6_dev_put(in6_dev);
2284 * Set destination address.
2285 * Special case for SIT interfaces where we create a new "virtual"
2286 * device.
2288 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2290 struct in6_ifreq ireq;
2291 struct net_device *dev;
2292 int err = -EINVAL;
2294 rtnl_lock();
2296 err = -EFAULT;
2297 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2298 goto err_exit;
2300 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2302 err = -ENODEV;
2303 if (dev == NULL)
2304 goto err_exit;
2306 #if IS_ENABLED(CONFIG_IPV6_SIT)
2307 if (dev->type == ARPHRD_SIT) {
2308 const struct net_device_ops *ops = dev->netdev_ops;
2309 struct ifreq ifr;
2310 struct ip_tunnel_parm p;
2312 err = -EADDRNOTAVAIL;
2313 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2314 goto err_exit;
2316 memset(&p, 0, sizeof(p));
2317 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2318 p.iph.saddr = 0;
2319 p.iph.version = 4;
2320 p.iph.ihl = 5;
2321 p.iph.protocol = IPPROTO_IPV6;
2322 p.iph.ttl = 64;
2323 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2325 if (ops->ndo_do_ioctl) {
2326 mm_segment_t oldfs = get_fs();
2328 set_fs(KERNEL_DS);
2329 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2330 set_fs(oldfs);
2331 } else
2332 err = -EOPNOTSUPP;
2334 if (err == 0) {
2335 err = -ENOBUFS;
2336 dev = __dev_get_by_name(net, p.name);
2337 if (!dev)
2338 goto err_exit;
2339 err = dev_open(dev);
2342 #endif
2344 err_exit:
2345 rtnl_unlock();
2346 return err;
2350 * Manual configuration of address on an interface
2352 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2353 const struct in6_addr *peer_pfx,
2354 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2355 __u32 valid_lft)
2357 struct inet6_ifaddr *ifp;
2358 struct inet6_dev *idev;
2359 struct net_device *dev;
2360 int scope;
2361 u32 flags;
2362 clock_t expires;
2363 unsigned long timeout;
2365 ASSERT_RTNL();
2367 if (plen > 128)
2368 return -EINVAL;
2370 /* check the lifetime */
2371 if (!valid_lft || prefered_lft > valid_lft)
2372 return -EINVAL;
2374 dev = __dev_get_by_index(net, ifindex);
2375 if (!dev)
2376 return -ENODEV;
2378 idev = addrconf_add_dev(dev);
2379 if (IS_ERR(idev))
2380 return PTR_ERR(idev);
2382 scope = ipv6_addr_scope(pfx);
2384 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2385 if (addrconf_finite_timeout(timeout)) {
2386 expires = jiffies_to_clock_t(timeout * HZ);
2387 valid_lft = timeout;
2388 flags = RTF_EXPIRES;
2389 } else {
2390 expires = 0;
2391 flags = 0;
2392 ifa_flags |= IFA_F_PERMANENT;
2395 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2396 if (addrconf_finite_timeout(timeout)) {
2397 if (timeout == 0)
2398 ifa_flags |= IFA_F_DEPRECATED;
2399 prefered_lft = timeout;
2402 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2403 valid_lft, prefered_lft);
2405 if (!IS_ERR(ifp)) {
2406 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2407 expires, flags);
2409 * Note that section 3.1 of RFC 4429 indicates
2410 * that the Optimistic flag should not be set for
2411 * manually configured addresses
2413 addrconf_dad_start(ifp);
2414 in6_ifa_put(ifp);
2415 addrconf_verify(0);
2416 return 0;
2419 return PTR_ERR(ifp);
2422 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2423 unsigned int plen)
2425 struct inet6_ifaddr *ifp;
2426 struct inet6_dev *idev;
2427 struct net_device *dev;
2429 if (plen > 128)
2430 return -EINVAL;
2432 dev = __dev_get_by_index(net, ifindex);
2433 if (!dev)
2434 return -ENODEV;
2436 if ((idev = __in6_dev_get(dev)) == NULL)
2437 return -ENXIO;
2439 read_lock_bh(&idev->lock);
2440 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2441 if (ifp->prefix_len == plen &&
2442 ipv6_addr_equal(pfx, &ifp->addr)) {
2443 in6_ifa_hold(ifp);
2444 read_unlock_bh(&idev->lock);
2446 ipv6_del_addr(ifp);
2447 return 0;
2450 read_unlock_bh(&idev->lock);
2451 return -EADDRNOTAVAIL;
2455 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2457 struct in6_ifreq ireq;
2458 int err;
2460 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2461 return -EPERM;
2463 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2464 return -EFAULT;
2466 rtnl_lock();
2467 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2468 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2469 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2470 rtnl_unlock();
2471 return err;
2474 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2476 struct in6_ifreq ireq;
2477 int err;
2479 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2480 return -EPERM;
2482 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2483 return -EFAULT;
2485 rtnl_lock();
2486 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2487 ireq.ifr6_prefixlen);
2488 rtnl_unlock();
2489 return err;
2492 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2493 int plen, int scope)
2495 struct inet6_ifaddr *ifp;
2497 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2498 scope, IFA_F_PERMANENT, 0, 0);
2499 if (!IS_ERR(ifp)) {
2500 spin_lock_bh(&ifp->lock);
2501 ifp->flags &= ~IFA_F_TENTATIVE;
2502 spin_unlock_bh(&ifp->lock);
2503 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2504 in6_ifa_put(ifp);
2508 #if IS_ENABLED(CONFIG_IPV6_SIT)
2509 static void sit_add_v4_addrs(struct inet6_dev *idev)
2511 struct in6_addr addr;
2512 struct net_device *dev;
2513 struct net *net = dev_net(idev->dev);
2514 int scope, plen;
2515 u32 pflags = 0;
2517 ASSERT_RTNL();
2519 memset(&addr, 0, sizeof(struct in6_addr));
2520 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2522 if (idev->dev->flags&IFF_POINTOPOINT) {
2523 addr.s6_addr32[0] = htonl(0xfe800000);
2524 scope = IFA_LINK;
2525 plen = 64;
2526 } else {
2527 scope = IPV6_ADDR_COMPATv4;
2528 plen = 96;
2529 pflags |= RTF_NONEXTHOP;
2532 if (addr.s6_addr32[3]) {
2533 add_addr(idev, &addr, plen, scope);
2534 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2535 return;
2538 for_each_netdev(net, dev) {
2539 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2540 if (in_dev && (dev->flags & IFF_UP)) {
2541 struct in_ifaddr *ifa;
2543 int flag = scope;
2545 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2547 addr.s6_addr32[3] = ifa->ifa_local;
2549 if (ifa->ifa_scope == RT_SCOPE_LINK)
2550 continue;
2551 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2552 if (idev->dev->flags&IFF_POINTOPOINT)
2553 continue;
2554 flag |= IFA_HOST;
2557 add_addr(idev, &addr, plen, flag);
2558 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2559 pflags);
2564 #endif
2566 static void init_loopback(struct net_device *dev)
2568 struct inet6_dev *idev;
2569 struct net_device *sp_dev;
2570 struct inet6_ifaddr *sp_ifa;
2571 struct rt6_info *sp_rt;
2573 /* ::1 */
2575 ASSERT_RTNL();
2577 if ((idev = ipv6_find_idev(dev)) == NULL) {
2578 pr_debug("%s: add_dev failed\n", __func__);
2579 return;
2582 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2584 /* Add routes to other interface's IPv6 addresses */
2585 for_each_netdev(dev_net(dev), sp_dev) {
2586 if (!strcmp(sp_dev->name, dev->name))
2587 continue;
2589 idev = __in6_dev_get(sp_dev);
2590 if (!idev)
2591 continue;
2593 read_lock_bh(&idev->lock);
2594 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2596 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2597 continue;
2599 if (sp_ifa->rt)
2600 continue;
2602 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2604 /* Failure cases are ignored */
2605 if (!IS_ERR(sp_rt)) {
2606 sp_ifa->rt = sp_rt;
2607 ip6_ins_rt(sp_rt);
2610 read_unlock_bh(&idev->lock);
2614 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2616 struct inet6_ifaddr *ifp;
2617 u32 addr_flags = IFA_F_PERMANENT;
2619 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2620 if (idev->cnf.optimistic_dad &&
2621 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2622 addr_flags |= IFA_F_OPTIMISTIC;
2623 #endif
2626 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2627 if (!IS_ERR(ifp)) {
2628 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2629 addrconf_dad_start(ifp);
2630 in6_ifa_put(ifp);
2634 static void addrconf_dev_config(struct net_device *dev)
2636 struct in6_addr addr;
2637 struct inet6_dev *idev;
2639 ASSERT_RTNL();
2641 if ((dev->type != ARPHRD_ETHER) &&
2642 (dev->type != ARPHRD_FDDI) &&
2643 (dev->type != ARPHRD_ARCNET) &&
2644 (dev->type != ARPHRD_INFINIBAND) &&
2645 (dev->type != ARPHRD_IEEE802154) &&
2646 (dev->type != ARPHRD_IEEE1394) &&
2647 (dev->type != ARPHRD_TUNNEL6)) {
2648 /* Alas, we support only Ethernet autoconfiguration. */
2649 return;
2652 idev = addrconf_add_dev(dev);
2653 if (IS_ERR(idev))
2654 return;
2656 memset(&addr, 0, sizeof(struct in6_addr));
2657 addr.s6_addr32[0] = htonl(0xFE800000);
2659 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2660 addrconf_add_linklocal(idev, &addr);
2663 #if IS_ENABLED(CONFIG_IPV6_SIT)
2664 static void addrconf_sit_config(struct net_device *dev)
2666 struct inet6_dev *idev;
2668 ASSERT_RTNL();
2671 * Configure the tunnel with one of our IPv4
2672 * addresses... we should configure all of
2673 * our v4 addrs in the tunnel
2676 if ((idev = ipv6_find_idev(dev)) == NULL) {
2677 pr_debug("%s: add_dev failed\n", __func__);
2678 return;
2681 if (dev->priv_flags & IFF_ISATAP) {
2682 struct in6_addr addr;
2684 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2685 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2686 addrconf_add_linklocal(idev, &addr);
2687 return;
2690 sit_add_v4_addrs(idev);
2692 if (dev->flags&IFF_POINTOPOINT)
2693 addrconf_add_mroute(dev);
2695 #endif
2697 #if IS_ENABLED(CONFIG_NET_IPGRE)
2698 static void addrconf_gre_config(struct net_device *dev)
2700 struct inet6_dev *idev;
2701 struct in6_addr addr;
2703 ASSERT_RTNL();
2705 if ((idev = ipv6_find_idev(dev)) == NULL) {
2706 pr_debug("%s: add_dev failed\n", __func__);
2707 return;
2710 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2711 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2712 addrconf_add_linklocal(idev, &addr);
2714 #endif
2716 static inline int
2717 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2719 struct in6_addr lladdr;
2721 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2722 addrconf_add_linklocal(idev, &lladdr);
2723 return 0;
2725 return -1;
2728 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2729 void *ptr)
2731 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2732 struct inet6_dev *idev = __in6_dev_get(dev);
2733 int run_pending = 0;
2734 int err;
2736 switch (event) {
2737 case NETDEV_REGISTER:
2738 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2739 idev = ipv6_add_dev(dev);
2740 if (!idev)
2741 return notifier_from_errno(-ENOMEM);
2743 break;
2745 case NETDEV_UP:
2746 case NETDEV_CHANGE:
2747 if (dev->flags & IFF_SLAVE)
2748 break;
2750 if (event == NETDEV_UP) {
2751 if (!addrconf_qdisc_ok(dev)) {
2752 /* device is not ready yet. */
2753 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2754 dev->name);
2755 break;
2758 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2759 idev = ipv6_add_dev(dev);
2761 if (idev) {
2762 idev->if_flags |= IF_READY;
2763 run_pending = 1;
2765 } else {
2766 if (!addrconf_qdisc_ok(dev)) {
2767 /* device is still not ready. */
2768 break;
2771 if (idev) {
2772 if (idev->if_flags & IF_READY)
2773 /* device is already configured. */
2774 break;
2775 idev->if_flags |= IF_READY;
2778 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2779 dev->name);
2781 run_pending = 1;
2784 switch (dev->type) {
2785 #if IS_ENABLED(CONFIG_IPV6_SIT)
2786 case ARPHRD_SIT:
2787 addrconf_sit_config(dev);
2788 break;
2789 #endif
2790 #if IS_ENABLED(CONFIG_NET_IPGRE)
2791 case ARPHRD_IPGRE:
2792 addrconf_gre_config(dev);
2793 break;
2794 #endif
2795 case ARPHRD_LOOPBACK:
2796 init_loopback(dev);
2797 break;
2799 default:
2800 addrconf_dev_config(dev);
2801 break;
2804 if (idev) {
2805 if (run_pending)
2806 addrconf_dad_run(idev);
2809 * If the MTU changed during the interface down,
2810 * when the interface up, the changed MTU must be
2811 * reflected in the idev as well as routers.
2813 if (idev->cnf.mtu6 != dev->mtu &&
2814 dev->mtu >= IPV6_MIN_MTU) {
2815 rt6_mtu_change(dev, dev->mtu);
2816 idev->cnf.mtu6 = dev->mtu;
2818 idev->tstamp = jiffies;
2819 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2822 * If the changed mtu during down is lower than
2823 * IPV6_MIN_MTU stop IPv6 on this interface.
2825 if (dev->mtu < IPV6_MIN_MTU)
2826 addrconf_ifdown(dev, 1);
2828 break;
2830 case NETDEV_CHANGEMTU:
2831 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2832 rt6_mtu_change(dev, dev->mtu);
2833 idev->cnf.mtu6 = dev->mtu;
2834 break;
2837 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2838 idev = ipv6_add_dev(dev);
2839 if (idev)
2840 break;
2844 * MTU falled under IPV6_MIN_MTU.
2845 * Stop IPv6 on this interface.
2848 case NETDEV_DOWN:
2849 case NETDEV_UNREGISTER:
2851 * Remove all addresses from this interface.
2853 addrconf_ifdown(dev, event != NETDEV_DOWN);
2854 break;
2856 case NETDEV_CHANGENAME:
2857 if (idev) {
2858 snmp6_unregister_dev(idev);
2859 addrconf_sysctl_unregister(idev);
2860 addrconf_sysctl_register(idev);
2861 err = snmp6_register_dev(idev);
2862 if (err)
2863 return notifier_from_errno(err);
2865 break;
2867 case NETDEV_PRE_TYPE_CHANGE:
2868 case NETDEV_POST_TYPE_CHANGE:
2869 addrconf_type_change(dev, event);
2870 break;
2873 return NOTIFY_OK;
2877 * addrconf module should be notified of a device going up
2879 static struct notifier_block ipv6_dev_notf = {
2880 .notifier_call = addrconf_notify,
2883 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2885 struct inet6_dev *idev;
2886 ASSERT_RTNL();
2888 idev = __in6_dev_get(dev);
2890 if (event == NETDEV_POST_TYPE_CHANGE)
2891 ipv6_mc_remap(idev);
2892 else if (event == NETDEV_PRE_TYPE_CHANGE)
2893 ipv6_mc_unmap(idev);
2896 static int addrconf_ifdown(struct net_device *dev, int how)
2898 struct net *net = dev_net(dev);
2899 struct inet6_dev *idev;
2900 struct inet6_ifaddr *ifa;
2901 int state, i;
2903 ASSERT_RTNL();
2905 rt6_ifdown(net, dev);
2906 neigh_ifdown(&nd_tbl, dev);
2908 idev = __in6_dev_get(dev);
2909 if (idev == NULL)
2910 return -ENODEV;
2913 * Step 1: remove reference to ipv6 device from parent device.
2914 * Do not dev_put!
2916 if (how) {
2917 idev->dead = 1;
2919 /* protected by rtnl_lock */
2920 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2922 /* Step 1.5: remove snmp6 entry */
2923 snmp6_unregister_dev(idev);
2927 /* Step 2: clear hash table */
2928 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2929 struct hlist_head *h = &inet6_addr_lst[i];
2931 spin_lock_bh(&addrconf_hash_lock);
2932 restart:
2933 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
2934 if (ifa->idev == idev) {
2935 hlist_del_init_rcu(&ifa->addr_lst);
2936 addrconf_del_dad_timer(ifa);
2937 goto restart;
2940 spin_unlock_bh(&addrconf_hash_lock);
2943 write_lock_bh(&idev->lock);
2945 addrconf_del_rs_timer(idev);
2947 /* Step 2: clear flags for stateless addrconf */
2948 if (!how)
2949 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2951 if (how && del_timer(&idev->regen_timer))
2952 in6_dev_put(idev);
2954 /* Step 3: clear tempaddr list */
2955 while (!list_empty(&idev->tempaddr_list)) {
2956 ifa = list_first_entry(&idev->tempaddr_list,
2957 struct inet6_ifaddr, tmp_list);
2958 list_del(&ifa->tmp_list);
2959 write_unlock_bh(&idev->lock);
2960 spin_lock_bh(&ifa->lock);
2962 if (ifa->ifpub) {
2963 in6_ifa_put(ifa->ifpub);
2964 ifa->ifpub = NULL;
2966 spin_unlock_bh(&ifa->lock);
2967 in6_ifa_put(ifa);
2968 write_lock_bh(&idev->lock);
2971 while (!list_empty(&idev->addr_list)) {
2972 ifa = list_first_entry(&idev->addr_list,
2973 struct inet6_ifaddr, if_list);
2974 addrconf_del_dad_timer(ifa);
2976 list_del(&ifa->if_list);
2978 write_unlock_bh(&idev->lock);
2980 spin_lock_bh(&ifa->state_lock);
2981 state = ifa->state;
2982 ifa->state = INET6_IFADDR_STATE_DEAD;
2983 spin_unlock_bh(&ifa->state_lock);
2985 if (state != INET6_IFADDR_STATE_DEAD) {
2986 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2987 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
2989 in6_ifa_put(ifa);
2991 write_lock_bh(&idev->lock);
2994 write_unlock_bh(&idev->lock);
2996 /* Step 5: Discard multicast list */
2997 if (how)
2998 ipv6_mc_destroy_dev(idev);
2999 else
3000 ipv6_mc_down(idev);
3002 idev->tstamp = jiffies;
3004 /* Last: Shot the device (if unregistered) */
3005 if (how) {
3006 addrconf_sysctl_unregister(idev);
3007 neigh_parms_release(&nd_tbl, idev->nd_parms);
3008 neigh_ifdown(&nd_tbl, dev);
3009 in6_dev_put(idev);
3011 return 0;
3014 static void addrconf_rs_timer(unsigned long data)
3016 struct inet6_dev *idev = (struct inet6_dev *)data;
3017 struct net_device *dev = idev->dev;
3018 struct in6_addr lladdr;
3020 write_lock(&idev->lock);
3021 if (idev->dead || !(idev->if_flags & IF_READY))
3022 goto out;
3024 if (!ipv6_accept_ra(idev))
3025 goto out;
3027 /* Announcement received after solicitation was sent */
3028 if (idev->if_flags & IF_RA_RCVD)
3029 goto out;
3031 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3032 write_unlock(&idev->lock);
3033 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3034 ndisc_send_rs(dev, &lladdr,
3035 &in6addr_linklocal_allrouters);
3036 else
3037 goto put;
3039 write_lock(&idev->lock);
3040 /* The wait after the last probe can be shorter */
3041 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3042 idev->cnf.rtr_solicits) ?
3043 idev->cnf.rtr_solicit_delay :
3044 idev->cnf.rtr_solicit_interval);
3045 } else {
3047 * Note: we do not support deprecated "all on-link"
3048 * assumption any longer.
3050 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3053 out:
3054 write_unlock(&idev->lock);
3055 put:
3056 in6_dev_put(idev);
3060 * Duplicate Address Detection
3062 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3064 unsigned long rand_num;
3065 struct inet6_dev *idev = ifp->idev;
3067 if (ifp->flags & IFA_F_OPTIMISTIC)
3068 rand_num = 0;
3069 else
3070 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3072 ifp->dad_probes = idev->cnf.dad_transmits;
3073 addrconf_mod_dad_timer(ifp, rand_num);
3076 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3078 struct inet6_dev *idev = ifp->idev;
3079 struct net_device *dev = idev->dev;
3081 addrconf_join_solict(dev, &ifp->addr);
3083 net_srandom(ifp->addr.s6_addr32[3]);
3085 read_lock_bh(&idev->lock);
3086 spin_lock(&ifp->lock);
3087 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3088 goto out;
3090 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3091 idev->cnf.accept_dad < 1 ||
3092 !(ifp->flags&IFA_F_TENTATIVE) ||
3093 ifp->flags & IFA_F_NODAD) {
3094 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3095 spin_unlock(&ifp->lock);
3096 read_unlock_bh(&idev->lock);
3098 addrconf_dad_completed(ifp);
3099 return;
3102 if (!(idev->if_flags & IF_READY)) {
3103 spin_unlock(&ifp->lock);
3104 read_unlock_bh(&idev->lock);
3106 * If the device is not ready:
3107 * - keep it tentative if it is a permanent address.
3108 * - otherwise, kill it.
3110 in6_ifa_hold(ifp);
3111 addrconf_dad_stop(ifp, 0);
3112 return;
3116 * Optimistic nodes can start receiving
3117 * Frames right away
3119 if (ifp->flags & IFA_F_OPTIMISTIC)
3120 ip6_ins_rt(ifp->rt);
3122 addrconf_dad_kick(ifp);
3123 out:
3124 spin_unlock(&ifp->lock);
3125 read_unlock_bh(&idev->lock);
3128 static void addrconf_dad_timer(unsigned long data)
3130 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3131 struct inet6_dev *idev = ifp->idev;
3132 struct in6_addr mcaddr;
3134 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3135 goto out;
3137 write_lock(&idev->lock);
3138 if (idev->dead || !(idev->if_flags & IF_READY)) {
3139 write_unlock(&idev->lock);
3140 goto out;
3143 spin_lock(&ifp->lock);
3144 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3145 spin_unlock(&ifp->lock);
3146 write_unlock(&idev->lock);
3147 goto out;
3150 if (ifp->dad_probes == 0) {
3152 * DAD was successful
3155 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3156 spin_unlock(&ifp->lock);
3157 write_unlock(&idev->lock);
3159 addrconf_dad_completed(ifp);
3161 goto out;
3164 ifp->dad_probes--;
3165 addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3166 spin_unlock(&ifp->lock);
3167 write_unlock(&idev->lock);
3169 /* send a neighbour solicitation for our addr */
3170 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3171 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3172 out:
3173 in6_ifa_put(ifp);
3176 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3178 struct net_device *dev = ifp->idev->dev;
3179 struct in6_addr lladdr;
3180 bool send_rs, send_mld;
3182 addrconf_del_dad_timer(ifp);
3185 * Configure the address for reception. Now it is valid.
3188 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3190 /* If added prefix is link local and we are prepared to process
3191 router advertisements, start sending router solicitations.
3194 read_lock_bh(&ifp->idev->lock);
3195 spin_lock(&ifp->lock);
3196 send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3197 ifp->idev->valid_ll_addr_cnt == 1;
3198 send_rs = send_mld &&
3199 ipv6_accept_ra(ifp->idev) &&
3200 ifp->idev->cnf.rtr_solicits > 0 &&
3201 (dev->flags&IFF_LOOPBACK) == 0;
3202 spin_unlock(&ifp->lock);
3203 read_unlock_bh(&ifp->idev->lock);
3205 /* While dad is in progress mld report's source address is in6_addrany.
3206 * Resend with proper ll now.
3208 if (send_mld)
3209 ipv6_mc_dad_complete(ifp->idev);
3211 if (send_rs) {
3213 * If a host as already performed a random delay
3214 * [...] as part of DAD [...] there is no need
3215 * to delay again before sending the first RS
3217 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3218 return;
3219 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3221 write_lock_bh(&ifp->idev->lock);
3222 spin_lock(&ifp->lock);
3223 ifp->idev->rs_probes = 1;
3224 ifp->idev->if_flags |= IF_RS_SENT;
3225 addrconf_mod_rs_timer(ifp->idev,
3226 ifp->idev->cnf.rtr_solicit_interval);
3227 spin_unlock(&ifp->lock);
3228 write_unlock_bh(&ifp->idev->lock);
3232 static void addrconf_dad_run(struct inet6_dev *idev)
3234 struct inet6_ifaddr *ifp;
3236 read_lock_bh(&idev->lock);
3237 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3238 spin_lock(&ifp->lock);
3239 if (ifp->flags & IFA_F_TENTATIVE &&
3240 ifp->state == INET6_IFADDR_STATE_DAD)
3241 addrconf_dad_kick(ifp);
3242 spin_unlock(&ifp->lock);
3244 read_unlock_bh(&idev->lock);
3247 #ifdef CONFIG_PROC_FS
3248 struct if6_iter_state {
3249 struct seq_net_private p;
3250 int bucket;
3251 int offset;
3254 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3256 struct inet6_ifaddr *ifa = NULL;
3257 struct if6_iter_state *state = seq->private;
3258 struct net *net = seq_file_net(seq);
3259 int p = 0;
3261 /* initial bucket if pos is 0 */
3262 if (pos == 0) {
3263 state->bucket = 0;
3264 state->offset = 0;
3267 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3268 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3269 addr_lst) {
3270 if (!net_eq(dev_net(ifa->idev->dev), net))
3271 continue;
3272 /* sync with offset */
3273 if (p < state->offset) {
3274 p++;
3275 continue;
3277 state->offset++;
3278 return ifa;
3281 /* prepare for next bucket */
3282 state->offset = 0;
3283 p = 0;
3285 return NULL;
3288 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3289 struct inet6_ifaddr *ifa)
3291 struct if6_iter_state *state = seq->private;
3292 struct net *net = seq_file_net(seq);
3294 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3295 if (!net_eq(dev_net(ifa->idev->dev), net))
3296 continue;
3297 state->offset++;
3298 return ifa;
3301 while (++state->bucket < IN6_ADDR_HSIZE) {
3302 state->offset = 0;
3303 hlist_for_each_entry_rcu_bh(ifa,
3304 &inet6_addr_lst[state->bucket], addr_lst) {
3305 if (!net_eq(dev_net(ifa->idev->dev), net))
3306 continue;
3307 state->offset++;
3308 return ifa;
3312 return NULL;
3315 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3316 __acquires(rcu_bh)
3318 rcu_read_lock_bh();
3319 return if6_get_first(seq, *pos);
3322 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3324 struct inet6_ifaddr *ifa;
3326 ifa = if6_get_next(seq, v);
3327 ++*pos;
3328 return ifa;
3331 static void if6_seq_stop(struct seq_file *seq, void *v)
3332 __releases(rcu_bh)
3334 rcu_read_unlock_bh();
3337 static int if6_seq_show(struct seq_file *seq, void *v)
3339 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3340 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3341 &ifp->addr,
3342 ifp->idev->dev->ifindex,
3343 ifp->prefix_len,
3344 ifp->scope,
3345 ifp->flags,
3346 ifp->idev->dev->name);
3347 return 0;
3350 static const struct seq_operations if6_seq_ops = {
3351 .start = if6_seq_start,
3352 .next = if6_seq_next,
3353 .show = if6_seq_show,
3354 .stop = if6_seq_stop,
3357 static int if6_seq_open(struct inode *inode, struct file *file)
3359 return seq_open_net(inode, file, &if6_seq_ops,
3360 sizeof(struct if6_iter_state));
3363 static const struct file_operations if6_fops = {
3364 .owner = THIS_MODULE,
3365 .open = if6_seq_open,
3366 .read = seq_read,
3367 .llseek = seq_lseek,
3368 .release = seq_release_net,
3371 static int __net_init if6_proc_net_init(struct net *net)
3373 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3374 return -ENOMEM;
3375 return 0;
3378 static void __net_exit if6_proc_net_exit(struct net *net)
3380 remove_proc_entry("if_inet6", net->proc_net);
3383 static struct pernet_operations if6_proc_net_ops = {
3384 .init = if6_proc_net_init,
3385 .exit = if6_proc_net_exit,
3388 int __init if6_proc_init(void)
3390 return register_pernet_subsys(&if6_proc_net_ops);
3393 void if6_proc_exit(void)
3395 unregister_pernet_subsys(&if6_proc_net_ops);
3397 #endif /* CONFIG_PROC_FS */
3399 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3400 /* Check if address is a home address configured on any interface. */
3401 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3403 int ret = 0;
3404 struct inet6_ifaddr *ifp = NULL;
3405 unsigned int hash = inet6_addr_hash(addr);
3407 rcu_read_lock_bh();
3408 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3409 if (!net_eq(dev_net(ifp->idev->dev), net))
3410 continue;
3411 if (ipv6_addr_equal(&ifp->addr, addr) &&
3412 (ifp->flags & IFA_F_HOMEADDRESS)) {
3413 ret = 1;
3414 break;
3417 rcu_read_unlock_bh();
3418 return ret;
3420 #endif
3423 * Periodic address status verification
3426 static void addrconf_verify(unsigned long foo)
3428 unsigned long now, next, next_sec, next_sched;
3429 struct inet6_ifaddr *ifp;
3430 int i;
3432 rcu_read_lock_bh();
3433 spin_lock(&addrconf_verify_lock);
3434 now = jiffies;
3435 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3437 del_timer(&addr_chk_timer);
3439 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3440 restart:
3441 hlist_for_each_entry_rcu_bh(ifp,
3442 &inet6_addr_lst[i], addr_lst) {
3443 unsigned long age;
3445 if (ifp->flags & IFA_F_PERMANENT)
3446 continue;
3448 spin_lock(&ifp->lock);
3449 /* We try to batch several events at once. */
3450 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3452 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3453 age >= ifp->valid_lft) {
3454 spin_unlock(&ifp->lock);
3455 in6_ifa_hold(ifp);
3456 ipv6_del_addr(ifp);
3457 goto restart;
3458 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3459 spin_unlock(&ifp->lock);
3460 continue;
3461 } else if (age >= ifp->prefered_lft) {
3462 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3463 int deprecate = 0;
3465 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3466 deprecate = 1;
3467 ifp->flags |= IFA_F_DEPRECATED;
3470 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3471 next = ifp->tstamp + ifp->valid_lft * HZ;
3473 spin_unlock(&ifp->lock);
3475 if (deprecate) {
3476 in6_ifa_hold(ifp);
3478 ipv6_ifa_notify(0, ifp);
3479 in6_ifa_put(ifp);
3480 goto restart;
3482 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3483 !(ifp->flags&IFA_F_TENTATIVE)) {
3484 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3485 ifp->idev->cnf.dad_transmits *
3486 ifp->idev->nd_parms->retrans_time / HZ;
3488 if (age >= ifp->prefered_lft - regen_advance) {
3489 struct inet6_ifaddr *ifpub = ifp->ifpub;
3490 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3491 next = ifp->tstamp + ifp->prefered_lft * HZ;
3492 if (!ifp->regen_count && ifpub) {
3493 ifp->regen_count++;
3494 in6_ifa_hold(ifp);
3495 in6_ifa_hold(ifpub);
3496 spin_unlock(&ifp->lock);
3498 spin_lock(&ifpub->lock);
3499 ifpub->regen_count = 0;
3500 spin_unlock(&ifpub->lock);
3501 ipv6_create_tempaddr(ifpub, ifp);
3502 in6_ifa_put(ifpub);
3503 in6_ifa_put(ifp);
3504 goto restart;
3506 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3507 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3508 spin_unlock(&ifp->lock);
3509 } else {
3510 /* ifp->prefered_lft <= ifp->valid_lft */
3511 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3512 next = ifp->tstamp + ifp->prefered_lft * HZ;
3513 spin_unlock(&ifp->lock);
3518 next_sec = round_jiffies_up(next);
3519 next_sched = next;
3521 /* If rounded timeout is accurate enough, accept it. */
3522 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3523 next_sched = next_sec;
3525 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3526 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3527 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3529 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3530 now, next, next_sec, next_sched);
3532 addr_chk_timer.expires = next_sched;
3533 add_timer(&addr_chk_timer);
3534 spin_unlock(&addrconf_verify_lock);
3535 rcu_read_unlock_bh();
3538 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3539 struct in6_addr **peer_pfx)
3541 struct in6_addr *pfx = NULL;
3543 *peer_pfx = NULL;
3545 if (addr)
3546 pfx = nla_data(addr);
3548 if (local) {
3549 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3550 *peer_pfx = pfx;
3551 pfx = nla_data(local);
3554 return pfx;
3557 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3558 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3559 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3560 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3563 static int
3564 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3566 struct net *net = sock_net(skb->sk);
3567 struct ifaddrmsg *ifm;
3568 struct nlattr *tb[IFA_MAX+1];
3569 struct in6_addr *pfx, *peer_pfx;
3570 int err;
3572 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3573 if (err < 0)
3574 return err;
3576 ifm = nlmsg_data(nlh);
3577 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3578 if (pfx == NULL)
3579 return -EINVAL;
3581 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3584 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3585 u32 prefered_lft, u32 valid_lft)
3587 u32 flags;
3588 clock_t expires;
3589 unsigned long timeout;
3591 if (!valid_lft || (prefered_lft > valid_lft))
3592 return -EINVAL;
3594 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3595 if (addrconf_finite_timeout(timeout)) {
3596 expires = jiffies_to_clock_t(timeout * HZ);
3597 valid_lft = timeout;
3598 flags = RTF_EXPIRES;
3599 } else {
3600 expires = 0;
3601 flags = 0;
3602 ifa_flags |= IFA_F_PERMANENT;
3605 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3606 if (addrconf_finite_timeout(timeout)) {
3607 if (timeout == 0)
3608 ifa_flags |= IFA_F_DEPRECATED;
3609 prefered_lft = timeout;
3612 spin_lock_bh(&ifp->lock);
3613 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3614 ifp->tstamp = jiffies;
3615 ifp->valid_lft = valid_lft;
3616 ifp->prefered_lft = prefered_lft;
3618 spin_unlock_bh(&ifp->lock);
3619 if (!(ifp->flags&IFA_F_TENTATIVE))
3620 ipv6_ifa_notify(0, ifp);
3622 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3623 expires, flags);
3624 addrconf_verify(0);
3626 return 0;
3629 static int
3630 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3632 struct net *net = sock_net(skb->sk);
3633 struct ifaddrmsg *ifm;
3634 struct nlattr *tb[IFA_MAX+1];
3635 struct in6_addr *pfx, *peer_pfx;
3636 struct inet6_ifaddr *ifa;
3637 struct net_device *dev;
3638 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3639 u8 ifa_flags;
3640 int err;
3642 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3643 if (err < 0)
3644 return err;
3646 ifm = nlmsg_data(nlh);
3647 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3648 if (pfx == NULL)
3649 return -EINVAL;
3651 if (tb[IFA_CACHEINFO]) {
3652 struct ifa_cacheinfo *ci;
3654 ci = nla_data(tb[IFA_CACHEINFO]);
3655 valid_lft = ci->ifa_valid;
3656 preferred_lft = ci->ifa_prefered;
3657 } else {
3658 preferred_lft = INFINITY_LIFE_TIME;
3659 valid_lft = INFINITY_LIFE_TIME;
3662 dev = __dev_get_by_index(net, ifm->ifa_index);
3663 if (dev == NULL)
3664 return -ENODEV;
3666 /* We ignore other flags so far. */
3667 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3669 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3670 if (ifa == NULL) {
3672 * It would be best to check for !NLM_F_CREATE here but
3673 * userspace alreay relies on not having to provide this.
3675 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3676 ifm->ifa_prefixlen, ifa_flags,
3677 preferred_lft, valid_lft);
3680 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3681 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3682 err = -EEXIST;
3683 else
3684 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3686 in6_ifa_put(ifa);
3688 return err;
3691 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3692 u8 scope, int ifindex)
3694 struct ifaddrmsg *ifm;
3696 ifm = nlmsg_data(nlh);
3697 ifm->ifa_family = AF_INET6;
3698 ifm->ifa_prefixlen = prefixlen;
3699 ifm->ifa_flags = flags;
3700 ifm->ifa_scope = scope;
3701 ifm->ifa_index = ifindex;
3704 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3705 unsigned long tstamp, u32 preferred, u32 valid)
3707 struct ifa_cacheinfo ci;
3709 ci.cstamp = cstamp_delta(cstamp);
3710 ci.tstamp = cstamp_delta(tstamp);
3711 ci.ifa_prefered = preferred;
3712 ci.ifa_valid = valid;
3714 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3717 static inline int rt_scope(int ifa_scope)
3719 if (ifa_scope & IFA_HOST)
3720 return RT_SCOPE_HOST;
3721 else if (ifa_scope & IFA_LINK)
3722 return RT_SCOPE_LINK;
3723 else if (ifa_scope & IFA_SITE)
3724 return RT_SCOPE_SITE;
3725 else
3726 return RT_SCOPE_UNIVERSE;
3729 static inline int inet6_ifaddr_msgsize(void)
3731 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3732 + nla_total_size(16) /* IFA_LOCAL */
3733 + nla_total_size(16) /* IFA_ADDRESS */
3734 + nla_total_size(sizeof(struct ifa_cacheinfo));
3737 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3738 u32 portid, u32 seq, int event, unsigned int flags)
3740 struct nlmsghdr *nlh;
3741 u32 preferred, valid;
3743 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3744 if (nlh == NULL)
3745 return -EMSGSIZE;
3747 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3748 ifa->idev->dev->ifindex);
3750 if (!(ifa->flags&IFA_F_PERMANENT)) {
3751 preferred = ifa->prefered_lft;
3752 valid = ifa->valid_lft;
3753 if (preferred != INFINITY_LIFE_TIME) {
3754 long tval = (jiffies - ifa->tstamp)/HZ;
3755 if (preferred > tval)
3756 preferred -= tval;
3757 else
3758 preferred = 0;
3759 if (valid != INFINITY_LIFE_TIME) {
3760 if (valid > tval)
3761 valid -= tval;
3762 else
3763 valid = 0;
3766 } else {
3767 preferred = INFINITY_LIFE_TIME;
3768 valid = INFINITY_LIFE_TIME;
3771 if (!ipv6_addr_any(&ifa->peer_addr)) {
3772 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3773 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3774 goto error;
3775 } else
3776 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3777 goto error;
3779 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3780 goto error;
3782 return nlmsg_end(skb, nlh);
3784 error:
3785 nlmsg_cancel(skb, nlh);
3786 return -EMSGSIZE;
3789 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3790 u32 portid, u32 seq, int event, u16 flags)
3792 struct nlmsghdr *nlh;
3793 u8 scope = RT_SCOPE_UNIVERSE;
3794 int ifindex = ifmca->idev->dev->ifindex;
3796 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3797 scope = RT_SCOPE_SITE;
3799 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3800 if (nlh == NULL)
3801 return -EMSGSIZE;
3803 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3804 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3805 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3806 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3807 nlmsg_cancel(skb, nlh);
3808 return -EMSGSIZE;
3811 return nlmsg_end(skb, nlh);
3814 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3815 u32 portid, u32 seq, int event, unsigned int flags)
3817 struct nlmsghdr *nlh;
3818 u8 scope = RT_SCOPE_UNIVERSE;
3819 int ifindex = ifaca->aca_idev->dev->ifindex;
3821 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3822 scope = RT_SCOPE_SITE;
3824 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3825 if (nlh == NULL)
3826 return -EMSGSIZE;
3828 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3829 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3830 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3831 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3832 nlmsg_cancel(skb, nlh);
3833 return -EMSGSIZE;
3836 return nlmsg_end(skb, nlh);
3839 enum addr_type_t {
3840 UNICAST_ADDR,
3841 MULTICAST_ADDR,
3842 ANYCAST_ADDR,
3845 /* called with rcu_read_lock() */
3846 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3847 struct netlink_callback *cb, enum addr_type_t type,
3848 int s_ip_idx, int *p_ip_idx)
3850 struct ifmcaddr6 *ifmca;
3851 struct ifacaddr6 *ifaca;
3852 int err = 1;
3853 int ip_idx = *p_ip_idx;
3855 read_lock_bh(&idev->lock);
3856 switch (type) {
3857 case UNICAST_ADDR: {
3858 struct inet6_ifaddr *ifa;
3860 /* unicast address incl. temp addr */
3861 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3862 if (++ip_idx < s_ip_idx)
3863 continue;
3864 err = inet6_fill_ifaddr(skb, ifa,
3865 NETLINK_CB(cb->skb).portid,
3866 cb->nlh->nlmsg_seq,
3867 RTM_NEWADDR,
3868 NLM_F_MULTI);
3869 if (err <= 0)
3870 break;
3871 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3873 break;
3875 case MULTICAST_ADDR:
3876 /* multicast address */
3877 for (ifmca = idev->mc_list; ifmca;
3878 ifmca = ifmca->next, ip_idx++) {
3879 if (ip_idx < s_ip_idx)
3880 continue;
3881 err = inet6_fill_ifmcaddr(skb, ifmca,
3882 NETLINK_CB(cb->skb).portid,
3883 cb->nlh->nlmsg_seq,
3884 RTM_GETMULTICAST,
3885 NLM_F_MULTI);
3886 if (err <= 0)
3887 break;
3889 break;
3890 case ANYCAST_ADDR:
3891 /* anycast address */
3892 for (ifaca = idev->ac_list; ifaca;
3893 ifaca = ifaca->aca_next, ip_idx++) {
3894 if (ip_idx < s_ip_idx)
3895 continue;
3896 err = inet6_fill_ifacaddr(skb, ifaca,
3897 NETLINK_CB(cb->skb).portid,
3898 cb->nlh->nlmsg_seq,
3899 RTM_GETANYCAST,
3900 NLM_F_MULTI);
3901 if (err <= 0)
3902 break;
3904 break;
3905 default:
3906 break;
3908 read_unlock_bh(&idev->lock);
3909 *p_ip_idx = ip_idx;
3910 return err;
3913 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3914 enum addr_type_t type)
3916 struct net *net = sock_net(skb->sk);
3917 int h, s_h;
3918 int idx, ip_idx;
3919 int s_idx, s_ip_idx;
3920 struct net_device *dev;
3921 struct inet6_dev *idev;
3922 struct hlist_head *head;
3924 s_h = cb->args[0];
3925 s_idx = idx = cb->args[1];
3926 s_ip_idx = ip_idx = cb->args[2];
3928 rcu_read_lock();
3929 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3930 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3931 idx = 0;
3932 head = &net->dev_index_head[h];
3933 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3934 if (idx < s_idx)
3935 goto cont;
3936 if (h > s_h || idx > s_idx)
3937 s_ip_idx = 0;
3938 ip_idx = 0;
3939 idev = __in6_dev_get(dev);
3940 if (!idev)
3941 goto cont;
3943 if (in6_dump_addrs(idev, skb, cb, type,
3944 s_ip_idx, &ip_idx) <= 0)
3945 goto done;
3946 cont:
3947 idx++;
3950 done:
3951 rcu_read_unlock();
3952 cb->args[0] = h;
3953 cb->args[1] = idx;
3954 cb->args[2] = ip_idx;
3956 return skb->len;
3959 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3961 enum addr_type_t type = UNICAST_ADDR;
3963 return inet6_dump_addr(skb, cb, type);
3966 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3968 enum addr_type_t type = MULTICAST_ADDR;
3970 return inet6_dump_addr(skb, cb, type);
3974 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3976 enum addr_type_t type = ANYCAST_ADDR;
3978 return inet6_dump_addr(skb, cb, type);
3981 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3983 struct net *net = sock_net(in_skb->sk);
3984 struct ifaddrmsg *ifm;
3985 struct nlattr *tb[IFA_MAX+1];
3986 struct in6_addr *addr = NULL, *peer;
3987 struct net_device *dev = NULL;
3988 struct inet6_ifaddr *ifa;
3989 struct sk_buff *skb;
3990 int err;
3992 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3993 if (err < 0)
3994 goto errout;
3996 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
3997 if (addr == NULL) {
3998 err = -EINVAL;
3999 goto errout;
4002 ifm = nlmsg_data(nlh);
4003 if (ifm->ifa_index)
4004 dev = __dev_get_by_index(net, ifm->ifa_index);
4006 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4007 if (!ifa) {
4008 err = -EADDRNOTAVAIL;
4009 goto errout;
4012 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4013 if (!skb) {
4014 err = -ENOBUFS;
4015 goto errout_ifa;
4018 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4019 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4020 if (err < 0) {
4021 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4022 WARN_ON(err == -EMSGSIZE);
4023 kfree_skb(skb);
4024 goto errout_ifa;
4026 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4027 errout_ifa:
4028 in6_ifa_put(ifa);
4029 errout:
4030 return err;
4033 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4035 struct sk_buff *skb;
4036 struct net *net = dev_net(ifa->idev->dev);
4037 int err = -ENOBUFS;
4039 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4040 if (skb == NULL)
4041 goto errout;
4043 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4044 if (err < 0) {
4045 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4046 WARN_ON(err == -EMSGSIZE);
4047 kfree_skb(skb);
4048 goto errout;
4050 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4051 return;
4052 errout:
4053 if (err < 0)
4054 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4057 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4058 __s32 *array, int bytes)
4060 BUG_ON(bytes < (DEVCONF_MAX * 4));
4062 memset(array, 0, bytes);
4063 array[DEVCONF_FORWARDING] = cnf->forwarding;
4064 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4065 array[DEVCONF_MTU6] = cnf->mtu6;
4066 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4067 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4068 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4069 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4070 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4071 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4072 jiffies_to_msecs(cnf->rtr_solicit_interval);
4073 array[DEVCONF_RTR_SOLICIT_DELAY] =
4074 jiffies_to_msecs(cnf->rtr_solicit_delay);
4075 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4076 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4077 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4078 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4079 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4080 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4081 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4082 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4083 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4084 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4085 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4086 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4087 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4088 #ifdef CONFIG_IPV6_ROUTER_PREF
4089 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4090 array[DEVCONF_RTR_PROBE_INTERVAL] =
4091 jiffies_to_msecs(cnf->rtr_probe_interval);
4092 #ifdef CONFIG_IPV6_ROUTE_INFO
4093 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4094 #endif
4095 #endif
4096 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4097 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4098 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4099 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4100 #endif
4101 #ifdef CONFIG_IPV6_MROUTE
4102 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4103 #endif
4104 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4105 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4106 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4107 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4108 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4111 static inline size_t inet6_ifla6_size(void)
4113 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4114 + nla_total_size(sizeof(struct ifla_cacheinfo))
4115 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4116 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4117 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4118 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4121 static inline size_t inet6_if_nlmsg_size(void)
4123 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4124 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4125 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4126 + nla_total_size(4) /* IFLA_MTU */
4127 + nla_total_size(4) /* IFLA_LINK */
4128 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4131 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4132 int items, int bytes)
4134 int i;
4135 int pad = bytes - sizeof(u64) * items;
4136 BUG_ON(pad < 0);
4138 /* Use put_unaligned() because stats may not be aligned for u64. */
4139 put_unaligned(items, &stats[0]);
4140 for (i = 1; i < items; i++)
4141 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4143 memset(&stats[items], 0, pad);
4146 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4147 int items, int bytes, size_t syncpoff)
4149 int i;
4150 int pad = bytes - sizeof(u64) * items;
4151 BUG_ON(pad < 0);
4153 /* Use put_unaligned() because stats may not be aligned for u64. */
4154 put_unaligned(items, &stats[0]);
4155 for (i = 1; i < items; i++)
4156 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4158 memset(&stats[items], 0, pad);
4161 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4162 int bytes)
4164 switch (attrtype) {
4165 case IFLA_INET6_STATS:
4166 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4167 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4168 break;
4169 case IFLA_INET6_ICMP6STATS:
4170 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4171 break;
4175 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4177 struct nlattr *nla;
4178 struct ifla_cacheinfo ci;
4180 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4181 goto nla_put_failure;
4182 ci.max_reasm_len = IPV6_MAXPLEN;
4183 ci.tstamp = cstamp_delta(idev->tstamp);
4184 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4185 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4186 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4187 goto nla_put_failure;
4188 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4189 if (nla == NULL)
4190 goto nla_put_failure;
4191 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4193 /* XXX - MC not implemented */
4195 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4196 if (nla == NULL)
4197 goto nla_put_failure;
4198 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4200 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4201 if (nla == NULL)
4202 goto nla_put_failure;
4203 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4205 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4206 if (nla == NULL)
4207 goto nla_put_failure;
4208 read_lock_bh(&idev->lock);
4209 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4210 read_unlock_bh(&idev->lock);
4212 return 0;
4214 nla_put_failure:
4215 return -EMSGSIZE;
4218 static size_t inet6_get_link_af_size(const struct net_device *dev)
4220 if (!__in6_dev_get(dev))
4221 return 0;
4223 return inet6_ifla6_size();
4226 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4228 struct inet6_dev *idev = __in6_dev_get(dev);
4230 if (!idev)
4231 return -ENODATA;
4233 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4234 return -EMSGSIZE;
4236 return 0;
4239 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4241 struct inet6_ifaddr *ifp;
4242 struct net_device *dev = idev->dev;
4243 bool update_rs = false;
4244 struct in6_addr ll_addr;
4246 if (token == NULL)
4247 return -EINVAL;
4248 if (ipv6_addr_any(token))
4249 return -EINVAL;
4250 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4251 return -EINVAL;
4252 if (!ipv6_accept_ra(idev))
4253 return -EINVAL;
4254 if (idev->cnf.rtr_solicits <= 0)
4255 return -EINVAL;
4257 write_lock_bh(&idev->lock);
4259 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4260 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4262 write_unlock_bh(&idev->lock);
4264 if (!idev->dead && (idev->if_flags & IF_READY) &&
4265 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4266 IFA_F_OPTIMISTIC)) {
4268 /* If we're not ready, then normal ifup will take care
4269 * of this. Otherwise, we need to request our rs here.
4271 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4272 update_rs = true;
4275 write_lock_bh(&idev->lock);
4277 if (update_rs) {
4278 idev->if_flags |= IF_RS_SENT;
4279 idev->rs_probes = 1;
4280 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4283 /* Well, that's kinda nasty ... */
4284 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4285 spin_lock(&ifp->lock);
4286 if (ifp->tokenized) {
4287 ifp->valid_lft = 0;
4288 ifp->prefered_lft = 0;
4290 spin_unlock(&ifp->lock);
4293 write_unlock_bh(&idev->lock);
4294 addrconf_verify(0);
4295 return 0;
4298 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4300 int err = -EINVAL;
4301 struct inet6_dev *idev = __in6_dev_get(dev);
4302 struct nlattr *tb[IFLA_INET6_MAX + 1];
4304 if (!idev)
4305 return -EAFNOSUPPORT;
4307 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4308 BUG();
4310 if (tb[IFLA_INET6_TOKEN])
4311 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4313 return err;
4316 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4317 u32 portid, u32 seq, int event, unsigned int flags)
4319 struct net_device *dev = idev->dev;
4320 struct ifinfomsg *hdr;
4321 struct nlmsghdr *nlh;
4322 void *protoinfo;
4324 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4325 if (nlh == NULL)
4326 return -EMSGSIZE;
4328 hdr = nlmsg_data(nlh);
4329 hdr->ifi_family = AF_INET6;
4330 hdr->__ifi_pad = 0;
4331 hdr->ifi_type = dev->type;
4332 hdr->ifi_index = dev->ifindex;
4333 hdr->ifi_flags = dev_get_flags(dev);
4334 hdr->ifi_change = 0;
4336 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4337 (dev->addr_len &&
4338 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4339 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4340 (dev->ifindex != dev->iflink &&
4341 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4342 goto nla_put_failure;
4343 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4344 if (protoinfo == NULL)
4345 goto nla_put_failure;
4347 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4348 goto nla_put_failure;
4350 nla_nest_end(skb, protoinfo);
4351 return nlmsg_end(skb, nlh);
4353 nla_put_failure:
4354 nlmsg_cancel(skb, nlh);
4355 return -EMSGSIZE;
4358 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4360 struct net *net = sock_net(skb->sk);
4361 int h, s_h;
4362 int idx = 0, s_idx;
4363 struct net_device *dev;
4364 struct inet6_dev *idev;
4365 struct hlist_head *head;
4367 s_h = cb->args[0];
4368 s_idx = cb->args[1];
4370 rcu_read_lock();
4371 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4372 idx = 0;
4373 head = &net->dev_index_head[h];
4374 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4375 if (idx < s_idx)
4376 goto cont;
4377 idev = __in6_dev_get(dev);
4378 if (!idev)
4379 goto cont;
4380 if (inet6_fill_ifinfo(skb, idev,
4381 NETLINK_CB(cb->skb).portid,
4382 cb->nlh->nlmsg_seq,
4383 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4384 goto out;
4385 cont:
4386 idx++;
4389 out:
4390 rcu_read_unlock();
4391 cb->args[1] = idx;
4392 cb->args[0] = h;
4394 return skb->len;
4397 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4399 struct sk_buff *skb;
4400 struct net *net = dev_net(idev->dev);
4401 int err = -ENOBUFS;
4403 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4404 if (skb == NULL)
4405 goto errout;
4407 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4408 if (err < 0) {
4409 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4410 WARN_ON(err == -EMSGSIZE);
4411 kfree_skb(skb);
4412 goto errout;
4414 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4415 return;
4416 errout:
4417 if (err < 0)
4418 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4421 static inline size_t inet6_prefix_nlmsg_size(void)
4423 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4424 + nla_total_size(sizeof(struct in6_addr))
4425 + nla_total_size(sizeof(struct prefix_cacheinfo));
4428 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4429 struct prefix_info *pinfo, u32 portid, u32 seq,
4430 int event, unsigned int flags)
4432 struct prefixmsg *pmsg;
4433 struct nlmsghdr *nlh;
4434 struct prefix_cacheinfo ci;
4436 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4437 if (nlh == NULL)
4438 return -EMSGSIZE;
4440 pmsg = nlmsg_data(nlh);
4441 pmsg->prefix_family = AF_INET6;
4442 pmsg->prefix_pad1 = 0;
4443 pmsg->prefix_pad2 = 0;
4444 pmsg->prefix_ifindex = idev->dev->ifindex;
4445 pmsg->prefix_len = pinfo->prefix_len;
4446 pmsg->prefix_type = pinfo->type;
4447 pmsg->prefix_pad3 = 0;
4448 pmsg->prefix_flags = 0;
4449 if (pinfo->onlink)
4450 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4451 if (pinfo->autoconf)
4452 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4454 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4455 goto nla_put_failure;
4456 ci.preferred_time = ntohl(pinfo->prefered);
4457 ci.valid_time = ntohl(pinfo->valid);
4458 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4459 goto nla_put_failure;
4460 return nlmsg_end(skb, nlh);
4462 nla_put_failure:
4463 nlmsg_cancel(skb, nlh);
4464 return -EMSGSIZE;
4467 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4468 struct prefix_info *pinfo)
4470 struct sk_buff *skb;
4471 struct net *net = dev_net(idev->dev);
4472 int err = -ENOBUFS;
4474 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4475 if (skb == NULL)
4476 goto errout;
4478 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4479 if (err < 0) {
4480 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4481 WARN_ON(err == -EMSGSIZE);
4482 kfree_skb(skb);
4483 goto errout;
4485 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4486 return;
4487 errout:
4488 if (err < 0)
4489 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4492 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4494 write_lock_bh(&ifp->idev->lock);
4495 spin_lock(&ifp->lock);
4496 if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4497 IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4498 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4499 ifp->idev->valid_ll_addr_cnt += count;
4500 WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4501 spin_unlock(&ifp->lock);
4502 write_unlock_bh(&ifp->idev->lock);
4505 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4507 struct net *net = dev_net(ifp->idev->dev);
4509 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4511 switch (event) {
4512 case RTM_NEWADDR:
4513 update_valid_ll_addr_cnt(ifp, 1);
4516 * If the address was optimistic
4517 * we inserted the route at the start of
4518 * our DAD process, so we don't need
4519 * to do it again
4521 if (!(ifp->rt->rt6i_node))
4522 ip6_ins_rt(ifp->rt);
4523 if (ifp->idev->cnf.forwarding)
4524 addrconf_join_anycast(ifp);
4525 if (!ipv6_addr_any(&ifp->peer_addr))
4526 addrconf_prefix_route(&ifp->peer_addr, 128,
4527 ifp->idev->dev, 0, 0);
4528 break;
4529 case RTM_DELADDR:
4530 update_valid_ll_addr_cnt(ifp, -1);
4532 if (ifp->idev->cnf.forwarding)
4533 addrconf_leave_anycast(ifp);
4534 addrconf_leave_solict(ifp->idev, &ifp->addr);
4535 if (!ipv6_addr_any(&ifp->peer_addr)) {
4536 struct rt6_info *rt;
4537 struct net_device *dev = ifp->idev->dev;
4539 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4540 dev->ifindex, 1);
4541 if (rt) {
4542 dst_hold(&rt->dst);
4543 if (ip6_del_rt(rt))
4544 dst_free(&rt->dst);
4547 dst_hold(&ifp->rt->dst);
4549 if (ip6_del_rt(ifp->rt))
4550 dst_free(&ifp->rt->dst);
4551 break;
4553 atomic_inc(&net->ipv6.dev_addr_genid);
4554 rt_genid_bump_ipv6(net);
4557 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4559 rcu_read_lock_bh();
4560 if (likely(ifp->idev->dead == 0))
4561 __ipv6_ifa_notify(event, ifp);
4562 rcu_read_unlock_bh();
4565 #ifdef CONFIG_SYSCTL
4567 static
4568 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4569 void __user *buffer, size_t *lenp, loff_t *ppos)
4571 int *valp = ctl->data;
4572 int val = *valp;
4573 loff_t pos = *ppos;
4574 struct ctl_table lctl;
4575 int ret;
4578 * ctl->data points to idev->cnf.forwarding, we should
4579 * not modify it until we get the rtnl lock.
4581 lctl = *ctl;
4582 lctl.data = &val;
4584 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4586 if (write)
4587 ret = addrconf_fixup_forwarding(ctl, valp, val);
4588 if (ret)
4589 *ppos = pos;
4590 return ret;
4593 static void dev_disable_change(struct inet6_dev *idev)
4595 struct netdev_notifier_info info;
4597 if (!idev || !idev->dev)
4598 return;
4600 netdev_notifier_info_init(&info, idev->dev);
4601 if (idev->cnf.disable_ipv6)
4602 addrconf_notify(NULL, NETDEV_DOWN, &info);
4603 else
4604 addrconf_notify(NULL, NETDEV_UP, &info);
4607 static void addrconf_disable_change(struct net *net, __s32 newf)
4609 struct net_device *dev;
4610 struct inet6_dev *idev;
4612 rcu_read_lock();
4613 for_each_netdev_rcu(net, dev) {
4614 idev = __in6_dev_get(dev);
4615 if (idev) {
4616 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4617 idev->cnf.disable_ipv6 = newf;
4618 if (changed)
4619 dev_disable_change(idev);
4622 rcu_read_unlock();
4625 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4627 struct net *net;
4628 int old;
4630 if (!rtnl_trylock())
4631 return restart_syscall();
4633 net = (struct net *)table->extra2;
4634 old = *p;
4635 *p = newf;
4637 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4638 rtnl_unlock();
4639 return 0;
4642 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4643 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4644 addrconf_disable_change(net, newf);
4645 } else if ((!newf) ^ (!old))
4646 dev_disable_change((struct inet6_dev *)table->extra1);
4648 rtnl_unlock();
4649 return 0;
4652 static
4653 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4654 void __user *buffer, size_t *lenp, loff_t *ppos)
4656 int *valp = ctl->data;
4657 int val = *valp;
4658 loff_t pos = *ppos;
4659 struct ctl_table lctl;
4660 int ret;
4663 * ctl->data points to idev->cnf.disable_ipv6, we should
4664 * not modify it until we get the rtnl lock.
4666 lctl = *ctl;
4667 lctl.data = &val;
4669 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4671 if (write)
4672 ret = addrconf_disable_ipv6(ctl, valp, val);
4673 if (ret)
4674 *ppos = pos;
4675 return ret;
4678 static struct addrconf_sysctl_table
4680 struct ctl_table_header *sysctl_header;
4681 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4682 } addrconf_sysctl __read_mostly = {
4683 .sysctl_header = NULL,
4684 .addrconf_vars = {
4686 .procname = "forwarding",
4687 .data = &ipv6_devconf.forwarding,
4688 .maxlen = sizeof(int),
4689 .mode = 0644,
4690 .proc_handler = addrconf_sysctl_forward,
4693 .procname = "hop_limit",
4694 .data = &ipv6_devconf.hop_limit,
4695 .maxlen = sizeof(int),
4696 .mode = 0644,
4697 .proc_handler = proc_dointvec,
4700 .procname = "mtu",
4701 .data = &ipv6_devconf.mtu6,
4702 .maxlen = sizeof(int),
4703 .mode = 0644,
4704 .proc_handler = proc_dointvec,
4707 .procname = "accept_ra",
4708 .data = &ipv6_devconf.accept_ra,
4709 .maxlen = sizeof(int),
4710 .mode = 0644,
4711 .proc_handler = proc_dointvec,
4714 .procname = "accept_redirects",
4715 .data = &ipv6_devconf.accept_redirects,
4716 .maxlen = sizeof(int),
4717 .mode = 0644,
4718 .proc_handler = proc_dointvec,
4721 .procname = "autoconf",
4722 .data = &ipv6_devconf.autoconf,
4723 .maxlen = sizeof(int),
4724 .mode = 0644,
4725 .proc_handler = proc_dointvec,
4728 .procname = "dad_transmits",
4729 .data = &ipv6_devconf.dad_transmits,
4730 .maxlen = sizeof(int),
4731 .mode = 0644,
4732 .proc_handler = proc_dointvec,
4735 .procname = "router_solicitations",
4736 .data = &ipv6_devconf.rtr_solicits,
4737 .maxlen = sizeof(int),
4738 .mode = 0644,
4739 .proc_handler = proc_dointvec,
4742 .procname = "router_solicitation_interval",
4743 .data = &ipv6_devconf.rtr_solicit_interval,
4744 .maxlen = sizeof(int),
4745 .mode = 0644,
4746 .proc_handler = proc_dointvec_jiffies,
4749 .procname = "router_solicitation_delay",
4750 .data = &ipv6_devconf.rtr_solicit_delay,
4751 .maxlen = sizeof(int),
4752 .mode = 0644,
4753 .proc_handler = proc_dointvec_jiffies,
4756 .procname = "force_mld_version",
4757 .data = &ipv6_devconf.force_mld_version,
4758 .maxlen = sizeof(int),
4759 .mode = 0644,
4760 .proc_handler = proc_dointvec,
4763 .procname = "mldv1_unsolicited_report_interval",
4764 .data =
4765 &ipv6_devconf.mldv1_unsolicited_report_interval,
4766 .maxlen = sizeof(int),
4767 .mode = 0644,
4768 .proc_handler = proc_dointvec_ms_jiffies,
4771 .procname = "mldv2_unsolicited_report_interval",
4772 .data =
4773 &ipv6_devconf.mldv2_unsolicited_report_interval,
4774 .maxlen = sizeof(int),
4775 .mode = 0644,
4776 .proc_handler = proc_dointvec_ms_jiffies,
4779 .procname = "use_tempaddr",
4780 .data = &ipv6_devconf.use_tempaddr,
4781 .maxlen = sizeof(int),
4782 .mode = 0644,
4783 .proc_handler = proc_dointvec,
4786 .procname = "temp_valid_lft",
4787 .data = &ipv6_devconf.temp_valid_lft,
4788 .maxlen = sizeof(int),
4789 .mode = 0644,
4790 .proc_handler = proc_dointvec,
4793 .procname = "temp_prefered_lft",
4794 .data = &ipv6_devconf.temp_prefered_lft,
4795 .maxlen = sizeof(int),
4796 .mode = 0644,
4797 .proc_handler = proc_dointvec,
4800 .procname = "regen_max_retry",
4801 .data = &ipv6_devconf.regen_max_retry,
4802 .maxlen = sizeof(int),
4803 .mode = 0644,
4804 .proc_handler = proc_dointvec,
4807 .procname = "max_desync_factor",
4808 .data = &ipv6_devconf.max_desync_factor,
4809 .maxlen = sizeof(int),
4810 .mode = 0644,
4811 .proc_handler = proc_dointvec,
4814 .procname = "max_addresses",
4815 .data = &ipv6_devconf.max_addresses,
4816 .maxlen = sizeof(int),
4817 .mode = 0644,
4818 .proc_handler = proc_dointvec,
4821 .procname = "accept_ra_defrtr",
4822 .data = &ipv6_devconf.accept_ra_defrtr,
4823 .maxlen = sizeof(int),
4824 .mode = 0644,
4825 .proc_handler = proc_dointvec,
4828 .procname = "accept_ra_pinfo",
4829 .data = &ipv6_devconf.accept_ra_pinfo,
4830 .maxlen = sizeof(int),
4831 .mode = 0644,
4832 .proc_handler = proc_dointvec,
4834 #ifdef CONFIG_IPV6_ROUTER_PREF
4836 .procname = "accept_ra_rtr_pref",
4837 .data = &ipv6_devconf.accept_ra_rtr_pref,
4838 .maxlen = sizeof(int),
4839 .mode = 0644,
4840 .proc_handler = proc_dointvec,
4843 .procname = "router_probe_interval",
4844 .data = &ipv6_devconf.rtr_probe_interval,
4845 .maxlen = sizeof(int),
4846 .mode = 0644,
4847 .proc_handler = proc_dointvec_jiffies,
4849 #ifdef CONFIG_IPV6_ROUTE_INFO
4851 .procname = "accept_ra_rt_info_max_plen",
4852 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4853 .maxlen = sizeof(int),
4854 .mode = 0644,
4855 .proc_handler = proc_dointvec,
4857 #endif
4858 #endif
4860 .procname = "proxy_ndp",
4861 .data = &ipv6_devconf.proxy_ndp,
4862 .maxlen = sizeof(int),
4863 .mode = 0644,
4864 .proc_handler = proc_dointvec,
4867 .procname = "accept_source_route",
4868 .data = &ipv6_devconf.accept_source_route,
4869 .maxlen = sizeof(int),
4870 .mode = 0644,
4871 .proc_handler = proc_dointvec,
4873 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4875 .procname = "optimistic_dad",
4876 .data = &ipv6_devconf.optimistic_dad,
4877 .maxlen = sizeof(int),
4878 .mode = 0644,
4879 .proc_handler = proc_dointvec,
4882 #endif
4883 #ifdef CONFIG_IPV6_MROUTE
4885 .procname = "mc_forwarding",
4886 .data = &ipv6_devconf.mc_forwarding,
4887 .maxlen = sizeof(int),
4888 .mode = 0444,
4889 .proc_handler = proc_dointvec,
4891 #endif
4893 .procname = "disable_ipv6",
4894 .data = &ipv6_devconf.disable_ipv6,
4895 .maxlen = sizeof(int),
4896 .mode = 0644,
4897 .proc_handler = addrconf_sysctl_disable,
4900 .procname = "accept_dad",
4901 .data = &ipv6_devconf.accept_dad,
4902 .maxlen = sizeof(int),
4903 .mode = 0644,
4904 .proc_handler = proc_dointvec,
4907 .procname = "force_tllao",
4908 .data = &ipv6_devconf.force_tllao,
4909 .maxlen = sizeof(int),
4910 .mode = 0644,
4911 .proc_handler = proc_dointvec
4914 .procname = "ndisc_notify",
4915 .data = &ipv6_devconf.ndisc_notify,
4916 .maxlen = sizeof(int),
4917 .mode = 0644,
4918 .proc_handler = proc_dointvec
4921 .procname = "suppress_frag_ndisc",
4922 .data = &ipv6_devconf.suppress_frag_ndisc,
4923 .maxlen = sizeof(int),
4924 .mode = 0644,
4925 .proc_handler = proc_dointvec
4928 /* sentinel */
4933 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4934 struct inet6_dev *idev, struct ipv6_devconf *p)
4936 int i;
4937 struct addrconf_sysctl_table *t;
4938 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4940 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4941 if (t == NULL)
4942 goto out;
4944 for (i = 0; t->addrconf_vars[i].data; i++) {
4945 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4946 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4947 t->addrconf_vars[i].extra2 = net;
4950 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4952 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4953 if (t->sysctl_header == NULL)
4954 goto free;
4956 p->sysctl = t;
4957 return 0;
4959 free:
4960 kfree(t);
4961 out:
4962 return -ENOBUFS;
4965 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4967 struct addrconf_sysctl_table *t;
4969 if (p->sysctl == NULL)
4970 return;
4972 t = p->sysctl;
4973 p->sysctl = NULL;
4974 unregister_net_sysctl_table(t->sysctl_header);
4975 kfree(t);
4978 static void addrconf_sysctl_register(struct inet6_dev *idev)
4980 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4981 &ndisc_ifinfo_sysctl_change);
4982 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4983 idev, &idev->cnf);
4986 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4988 __addrconf_sysctl_unregister(&idev->cnf);
4989 neigh_sysctl_unregister(idev->nd_parms);
4993 #endif
4995 static int __net_init addrconf_init_net(struct net *net)
4997 int err = -ENOMEM;
4998 struct ipv6_devconf *all, *dflt;
5000 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5001 if (all == NULL)
5002 goto err_alloc_all;
5004 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5005 if (dflt == NULL)
5006 goto err_alloc_dflt;
5008 /* these will be inherited by all namespaces */
5009 dflt->autoconf = ipv6_defaults.autoconf;
5010 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5012 net->ipv6.devconf_all = all;
5013 net->ipv6.devconf_dflt = dflt;
5015 #ifdef CONFIG_SYSCTL
5016 err = __addrconf_sysctl_register(net, "all", NULL, all);
5017 if (err < 0)
5018 goto err_reg_all;
5020 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5021 if (err < 0)
5022 goto err_reg_dflt;
5023 #endif
5024 return 0;
5026 #ifdef CONFIG_SYSCTL
5027 err_reg_dflt:
5028 __addrconf_sysctl_unregister(all);
5029 err_reg_all:
5030 kfree(dflt);
5031 #endif
5032 err_alloc_dflt:
5033 kfree(all);
5034 err_alloc_all:
5035 return err;
5038 static void __net_exit addrconf_exit_net(struct net *net)
5040 #ifdef CONFIG_SYSCTL
5041 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5042 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5043 #endif
5044 if (!net_eq(net, &init_net)) {
5045 kfree(net->ipv6.devconf_dflt);
5046 kfree(net->ipv6.devconf_all);
5050 static struct pernet_operations addrconf_ops = {
5051 .init = addrconf_init_net,
5052 .exit = addrconf_exit_net,
5055 static struct rtnl_af_ops inet6_ops = {
5056 .family = AF_INET6,
5057 .fill_link_af = inet6_fill_link_af,
5058 .get_link_af_size = inet6_get_link_af_size,
5059 .set_link_af = inet6_set_link_af,
5063 * Init / cleanup code
5066 int __init addrconf_init(void)
5068 int i, err;
5070 err = ipv6_addr_label_init();
5071 if (err < 0) {
5072 pr_crit("%s: cannot initialize default policy table: %d\n",
5073 __func__, err);
5074 goto out;
5077 err = register_pernet_subsys(&addrconf_ops);
5078 if (err < 0)
5079 goto out_addrlabel;
5081 /* The addrconf netdev notifier requires that loopback_dev
5082 * has it's ipv6 private information allocated and setup
5083 * before it can bring up and give link-local addresses
5084 * to other devices which are up.
5086 * Unfortunately, loopback_dev is not necessarily the first
5087 * entry in the global dev_base list of net devices. In fact,
5088 * it is likely to be the very last entry on that list.
5089 * So this causes the notifier registry below to try and
5090 * give link-local addresses to all devices besides loopback_dev
5091 * first, then loopback_dev, which cases all the non-loopback_dev
5092 * devices to fail to get a link-local address.
5094 * So, as a temporary fix, allocate the ipv6 structure for
5095 * loopback_dev first by hand.
5096 * Longer term, all of the dependencies ipv6 has upon the loopback
5097 * device and it being up should be removed.
5099 rtnl_lock();
5100 if (!ipv6_add_dev(init_net.loopback_dev))
5101 err = -ENOMEM;
5102 rtnl_unlock();
5103 if (err)
5104 goto errlo;
5106 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5107 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5109 register_netdevice_notifier(&ipv6_dev_notf);
5111 addrconf_verify(0);
5113 err = rtnl_af_register(&inet6_ops);
5114 if (err < 0)
5115 goto errout_af;
5117 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5118 NULL);
5119 if (err < 0)
5120 goto errout;
5122 /* Only the first call to __rtnl_register can fail */
5123 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5124 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5125 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5126 inet6_dump_ifaddr, NULL);
5127 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5128 inet6_dump_ifmcaddr, NULL);
5129 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5130 inet6_dump_ifacaddr, NULL);
5131 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5132 inet6_netconf_dump_devconf, NULL);
5134 ipv6_addr_label_rtnl_register();
5136 return 0;
5137 errout:
5138 rtnl_af_unregister(&inet6_ops);
5139 errout_af:
5140 unregister_netdevice_notifier(&ipv6_dev_notf);
5141 errlo:
5142 unregister_pernet_subsys(&addrconf_ops);
5143 out_addrlabel:
5144 ipv6_addr_label_cleanup();
5145 out:
5146 return err;
5149 void addrconf_cleanup(void)
5151 struct net_device *dev;
5152 int i;
5154 unregister_netdevice_notifier(&ipv6_dev_notf);
5155 unregister_pernet_subsys(&addrconf_ops);
5156 ipv6_addr_label_cleanup();
5158 rtnl_lock();
5160 __rtnl_af_unregister(&inet6_ops);
5162 /* clean dev list */
5163 for_each_netdev(&init_net, dev) {
5164 if (__in6_dev_get(dev) == NULL)
5165 continue;
5166 addrconf_ifdown(dev, 1);
5168 addrconf_ifdown(init_net.loopback_dev, 2);
5171 * Check hash table.
5173 spin_lock_bh(&addrconf_hash_lock);
5174 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5175 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5176 spin_unlock_bh(&addrconf_hash_lock);
5178 del_timer(&addr_chk_timer);
5179 rtnl_unlock();