Linux 4.8-rc8
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob2f1f5d439788820aa2b75be8a26addb757ccf50c
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/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
69 #include <net/net_namespace.h>
70 #include <net/sock.h>
71 #include <net/snmp.h>
73 #include <net/6lowpan.h>
74 #include <net/firewire.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/tcp.h>
81 #include <net/ip.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
99 #if ACONF_DEBUG >= 3
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101 #else
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #endif
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 #define IPV6_MAX_STRLEN \
108 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
110 static inline u32 cstamp_delta(unsigned long cstamp)
112 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
115 #ifdef CONFIG_SYSCTL
116 static int addrconf_sysctl_register(struct inet6_dev *idev);
117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118 #else
119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
121 return 0;
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
127 #endif
129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131 static void ipv6_regen_rndid(unsigned long data);
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135 static int ipv6_generate_stable_address(struct in6_addr *addr,
136 u8 dad_count,
137 const struct inet6_dev *idev);
140 * Configured unicast address hash table
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
145 static void addrconf_verify(void);
146 static void addrconf_verify_rtnl(void);
147 static void addrconf_verify_work(struct work_struct *);
149 static struct workqueue_struct *addrconf_wq;
150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
155 static void addrconf_type_change(struct net_device *dev,
156 unsigned long event);
157 static int addrconf_ifdown(struct net_device *dev, int how);
159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
160 int plen,
161 const struct net_device *dev,
162 u32 flags, u32 noflags);
164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165 static void addrconf_dad_work(struct work_struct *w);
166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167 static void addrconf_dad_run(struct inet6_dev *idev);
168 static void addrconf_rs_timer(unsigned long data);
169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173 struct prefix_info *pinfo);
174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175 struct net_device *dev);
177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
178 .forwarding = 0,
179 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
180 .mtu6 = IPV6_MIN_MTU,
181 .accept_ra = 1,
182 .accept_redirects = 1,
183 .autoconf = 1,
184 .force_mld_version = 0,
185 .mldv1_unsolicited_report_interval = 10 * HZ,
186 .mldv2_unsolicited_report_interval = HZ,
187 .dad_transmits = 1,
188 .rtr_solicits = MAX_RTR_SOLICITATIONS,
189 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
190 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
191 .use_tempaddr = 0,
192 .temp_valid_lft = TEMP_VALID_LIFETIME,
193 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
194 .regen_max_retry = REGEN_MAX_RETRY,
195 .max_desync_factor = MAX_DESYNC_FACTOR,
196 .max_addresses = IPV6_MAX_ADDRESSES,
197 .accept_ra_defrtr = 1,
198 .accept_ra_from_local = 0,
199 .accept_ra_min_hop_limit= 1,
200 .accept_ra_pinfo = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202 .accept_ra_rtr_pref = 1,
203 .rtr_probe_interval = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205 .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208 .proxy_ndp = 0,
209 .accept_source_route = 0, /* we do not accept RH0 by default. */
210 .disable_ipv6 = 0,
211 .accept_dad = 1,
212 .suppress_frag_ndisc = 1,
213 .accept_ra_mtu = 1,
214 .stable_secret = {
215 .initialized = false,
217 .use_oif_addrs_only = 0,
218 .ignore_routes_with_linkdown = 0,
219 .keep_addr_on_down = 0,
222 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
223 .forwarding = 0,
224 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
225 .mtu6 = IPV6_MIN_MTU,
226 .accept_ra = 1,
227 .accept_redirects = 1,
228 .autoconf = 1,
229 .force_mld_version = 0,
230 .mldv1_unsolicited_report_interval = 10 * HZ,
231 .mldv2_unsolicited_report_interval = HZ,
232 .dad_transmits = 1,
233 .rtr_solicits = MAX_RTR_SOLICITATIONS,
234 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
235 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
236 .use_tempaddr = 0,
237 .temp_valid_lft = TEMP_VALID_LIFETIME,
238 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
239 .regen_max_retry = REGEN_MAX_RETRY,
240 .max_desync_factor = MAX_DESYNC_FACTOR,
241 .max_addresses = IPV6_MAX_ADDRESSES,
242 .accept_ra_defrtr = 1,
243 .accept_ra_from_local = 0,
244 .accept_ra_min_hop_limit= 1,
245 .accept_ra_pinfo = 1,
246 #ifdef CONFIG_IPV6_ROUTER_PREF
247 .accept_ra_rtr_pref = 1,
248 .rtr_probe_interval = 60 * HZ,
249 #ifdef CONFIG_IPV6_ROUTE_INFO
250 .accept_ra_rt_info_max_plen = 0,
251 #endif
252 #endif
253 .proxy_ndp = 0,
254 .accept_source_route = 0, /* we do not accept RH0 by default. */
255 .disable_ipv6 = 0,
256 .accept_dad = 1,
257 .suppress_frag_ndisc = 1,
258 .accept_ra_mtu = 1,
259 .stable_secret = {
260 .initialized = false,
262 .use_oif_addrs_only = 0,
263 .ignore_routes_with_linkdown = 0,
264 .keep_addr_on_down = 0,
267 /* Check if a valid qdisc is available */
268 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
270 return !qdisc_tx_is_noop(dev);
273 static void addrconf_del_rs_timer(struct inet6_dev *idev)
275 if (del_timer(&idev->rs_timer))
276 __in6_dev_put(idev);
279 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
281 if (cancel_delayed_work(&ifp->dad_work))
282 __in6_ifa_put(ifp);
285 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
286 unsigned long when)
288 if (!timer_pending(&idev->rs_timer))
289 in6_dev_hold(idev);
290 mod_timer(&idev->rs_timer, jiffies + when);
293 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
294 unsigned long delay)
296 if (!delayed_work_pending(&ifp->dad_work))
297 in6_ifa_hold(ifp);
298 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
301 static int snmp6_alloc_dev(struct inet6_dev *idev)
303 int i;
305 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
306 if (!idev->stats.ipv6)
307 goto err_ip;
309 for_each_possible_cpu(i) {
310 struct ipstats_mib *addrconf_stats;
311 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
312 u64_stats_init(&addrconf_stats->syncp);
316 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
317 GFP_KERNEL);
318 if (!idev->stats.icmpv6dev)
319 goto err_icmp;
320 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
321 GFP_KERNEL);
322 if (!idev->stats.icmpv6msgdev)
323 goto err_icmpmsg;
325 return 0;
327 err_icmpmsg:
328 kfree(idev->stats.icmpv6dev);
329 err_icmp:
330 free_percpu(idev->stats.ipv6);
331 err_ip:
332 return -ENOMEM;
335 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
337 struct inet6_dev *ndev;
338 int err = -ENOMEM;
340 ASSERT_RTNL();
342 if (dev->mtu < IPV6_MIN_MTU)
343 return ERR_PTR(-EINVAL);
345 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
346 if (!ndev)
347 return ERR_PTR(err);
349 rwlock_init(&ndev->lock);
350 ndev->dev = dev;
351 INIT_LIST_HEAD(&ndev->addr_list);
352 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
353 (unsigned long)ndev);
354 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
356 if (ndev->cnf.stable_secret.initialized)
357 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
358 else
359 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
363 if (!ndev->nd_parms) {
364 kfree(ndev);
365 return ERR_PTR(err);
367 if (ndev->cnf.forwarding)
368 dev_disable_lro(dev);
369 /* We refer to the device */
370 dev_hold(dev);
372 if (snmp6_alloc_dev(ndev) < 0) {
373 ADBG(KERN_WARNING
374 "%s: cannot allocate memory for statistics; dev=%s.\n",
375 __func__, dev->name);
376 neigh_parms_release(&nd_tbl, ndev->nd_parms);
377 dev_put(dev);
378 kfree(ndev);
379 return ERR_PTR(err);
382 if (snmp6_register_dev(ndev) < 0) {
383 ADBG(KERN_WARNING
384 "%s: cannot create /proc/net/dev_snmp6/%s\n",
385 __func__, dev->name);
386 goto err_release;
389 /* One reference from device. We must do this before
390 * we invoke __ipv6_regen_rndid().
392 in6_dev_hold(ndev);
394 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
395 ndev->cnf.accept_dad = -1;
397 #if IS_ENABLED(CONFIG_IPV6_SIT)
398 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
399 pr_info("%s: Disabled Multicast RS\n", dev->name);
400 ndev->cnf.rtr_solicits = 0;
402 #endif
404 INIT_LIST_HEAD(&ndev->tempaddr_list);
405 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
406 if ((dev->flags&IFF_LOOPBACK) ||
407 dev->type == ARPHRD_TUNNEL ||
408 dev->type == ARPHRD_TUNNEL6 ||
409 dev->type == ARPHRD_SIT ||
410 dev->type == ARPHRD_NONE) {
411 ndev->cnf.use_tempaddr = -1;
412 } else {
413 in6_dev_hold(ndev);
414 ipv6_regen_rndid((unsigned long) ndev);
417 ndev->token = in6addr_any;
419 if (netif_running(dev) && addrconf_qdisc_ok(dev))
420 ndev->if_flags |= IF_READY;
422 ipv6_mc_init_dev(ndev);
423 ndev->tstamp = jiffies;
424 err = addrconf_sysctl_register(ndev);
425 if (err) {
426 ipv6_mc_destroy_dev(ndev);
427 del_timer(&ndev->regen_timer);
428 snmp6_unregister_dev(ndev);
429 goto err_release;
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
434 /* Join interface-local all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
437 /* Join all-node multicast group */
438 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
440 /* Join all-router multicast group if forwarding is set */
441 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
442 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
444 return ndev;
446 err_release:
447 neigh_parms_release(&nd_tbl, ndev->nd_parms);
448 ndev->dead = 1;
449 in6_dev_finish_destroy(ndev);
450 return ERR_PTR(err);
453 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
455 struct inet6_dev *idev;
457 ASSERT_RTNL();
459 idev = __in6_dev_get(dev);
460 if (!idev) {
461 idev = ipv6_add_dev(dev);
462 if (IS_ERR(idev))
463 return NULL;
466 if (dev->flags&IFF_UP)
467 ipv6_mc_up(idev);
468 return idev;
471 static int inet6_netconf_msgsize_devconf(int type)
473 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
474 + nla_total_size(4); /* NETCONFA_IFINDEX */
475 bool all = false;
477 if (type == NETCONFA_ALL)
478 all = true;
480 if (all || type == NETCONFA_FORWARDING)
481 size += nla_total_size(4);
482 #ifdef CONFIG_IPV6_MROUTE
483 if (all || type == NETCONFA_MC_FORWARDING)
484 size += nla_total_size(4);
485 #endif
486 if (all || type == NETCONFA_PROXY_NEIGH)
487 size += nla_total_size(4);
489 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
490 size += nla_total_size(4);
492 return size;
495 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
496 struct ipv6_devconf *devconf, u32 portid,
497 u32 seq, int event, unsigned int flags,
498 int type)
500 struct nlmsghdr *nlh;
501 struct netconfmsg *ncm;
502 bool all = false;
504 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
505 flags);
506 if (!nlh)
507 return -EMSGSIZE;
509 if (type == NETCONFA_ALL)
510 all = true;
512 ncm = nlmsg_data(nlh);
513 ncm->ncm_family = AF_INET6;
515 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
516 goto nla_put_failure;
518 if ((all || type == NETCONFA_FORWARDING) &&
519 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
520 goto nla_put_failure;
521 #ifdef CONFIG_IPV6_MROUTE
522 if ((all || type == NETCONFA_MC_FORWARDING) &&
523 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
524 devconf->mc_forwarding) < 0)
525 goto nla_put_failure;
526 #endif
527 if ((all || type == NETCONFA_PROXY_NEIGH) &&
528 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
529 goto nla_put_failure;
531 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
532 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
533 devconf->ignore_routes_with_linkdown) < 0)
534 goto nla_put_failure;
536 nlmsg_end(skb, nlh);
537 return 0;
539 nla_put_failure:
540 nlmsg_cancel(skb, nlh);
541 return -EMSGSIZE;
544 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
545 struct ipv6_devconf *devconf)
547 struct sk_buff *skb;
548 int err = -ENOBUFS;
550 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
551 if (!skb)
552 goto errout;
554 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
555 RTM_NEWNETCONF, 0, type);
556 if (err < 0) {
557 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
558 WARN_ON(err == -EMSGSIZE);
559 kfree_skb(skb);
560 goto errout;
562 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
563 return;
564 errout:
565 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
568 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
569 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
570 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
571 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
572 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
575 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
576 struct nlmsghdr *nlh)
578 struct net *net = sock_net(in_skb->sk);
579 struct nlattr *tb[NETCONFA_MAX+1];
580 struct netconfmsg *ncm;
581 struct sk_buff *skb;
582 struct ipv6_devconf *devconf;
583 struct inet6_dev *in6_dev;
584 struct net_device *dev;
585 int ifindex;
586 int err;
588 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
589 devconf_ipv6_policy);
590 if (err < 0)
591 goto errout;
593 err = -EINVAL;
594 if (!tb[NETCONFA_IFINDEX])
595 goto errout;
597 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
598 switch (ifindex) {
599 case NETCONFA_IFINDEX_ALL:
600 devconf = net->ipv6.devconf_all;
601 break;
602 case NETCONFA_IFINDEX_DEFAULT:
603 devconf = net->ipv6.devconf_dflt;
604 break;
605 default:
606 dev = __dev_get_by_index(net, ifindex);
607 if (!dev)
608 goto errout;
609 in6_dev = __in6_dev_get(dev);
610 if (!in6_dev)
611 goto errout;
612 devconf = &in6_dev->cnf;
613 break;
616 err = -ENOBUFS;
617 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
618 if (!skb)
619 goto errout;
621 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
622 NETLINK_CB(in_skb).portid,
623 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
624 NETCONFA_ALL);
625 if (err < 0) {
626 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
627 WARN_ON(err == -EMSGSIZE);
628 kfree_skb(skb);
629 goto errout;
631 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
632 errout:
633 return err;
636 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
637 struct netlink_callback *cb)
639 struct net *net = sock_net(skb->sk);
640 int h, s_h;
641 int idx, s_idx;
642 struct net_device *dev;
643 struct inet6_dev *idev;
644 struct hlist_head *head;
646 s_h = cb->args[0];
647 s_idx = idx = cb->args[1];
649 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
650 idx = 0;
651 head = &net->dev_index_head[h];
652 rcu_read_lock();
653 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
654 net->dev_base_seq;
655 hlist_for_each_entry_rcu(dev, head, index_hlist) {
656 if (idx < s_idx)
657 goto cont;
658 idev = __in6_dev_get(dev);
659 if (!idev)
660 goto cont;
662 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
663 &idev->cnf,
664 NETLINK_CB(cb->skb).portid,
665 cb->nlh->nlmsg_seq,
666 RTM_NEWNETCONF,
667 NLM_F_MULTI,
668 NETCONFA_ALL) < 0) {
669 rcu_read_unlock();
670 goto done;
672 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
673 cont:
674 idx++;
676 rcu_read_unlock();
678 if (h == NETDEV_HASHENTRIES) {
679 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
680 net->ipv6.devconf_all,
681 NETLINK_CB(cb->skb).portid,
682 cb->nlh->nlmsg_seq,
683 RTM_NEWNETCONF, NLM_F_MULTI,
684 NETCONFA_ALL) < 0)
685 goto done;
686 else
687 h++;
689 if (h == NETDEV_HASHENTRIES + 1) {
690 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
691 net->ipv6.devconf_dflt,
692 NETLINK_CB(cb->skb).portid,
693 cb->nlh->nlmsg_seq,
694 RTM_NEWNETCONF, NLM_F_MULTI,
695 NETCONFA_ALL) < 0)
696 goto done;
697 else
698 h++;
700 done:
701 cb->args[0] = h;
702 cb->args[1] = idx;
704 return skb->len;
707 #ifdef CONFIG_SYSCTL
708 static void dev_forward_change(struct inet6_dev *idev)
710 struct net_device *dev;
711 struct inet6_ifaddr *ifa;
713 if (!idev)
714 return;
715 dev = idev->dev;
716 if (idev->cnf.forwarding)
717 dev_disable_lro(dev);
718 if (dev->flags & IFF_MULTICAST) {
719 if (idev->cnf.forwarding) {
720 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
721 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
722 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
723 } else {
724 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
725 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
726 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
730 list_for_each_entry(ifa, &idev->addr_list, if_list) {
731 if (ifa->flags&IFA_F_TENTATIVE)
732 continue;
733 if (idev->cnf.forwarding)
734 addrconf_join_anycast(ifa);
735 else
736 addrconf_leave_anycast(ifa);
738 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
739 dev->ifindex, &idev->cnf);
743 static void addrconf_forward_change(struct net *net, __s32 newf)
745 struct net_device *dev;
746 struct inet6_dev *idev;
748 for_each_netdev(net, dev) {
749 idev = __in6_dev_get(dev);
750 if (idev) {
751 int changed = (!idev->cnf.forwarding) ^ (!newf);
752 idev->cnf.forwarding = newf;
753 if (changed)
754 dev_forward_change(idev);
759 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
761 struct net *net;
762 int old;
764 if (!rtnl_trylock())
765 return restart_syscall();
767 net = (struct net *)table->extra2;
768 old = *p;
769 *p = newf;
771 if (p == &net->ipv6.devconf_dflt->forwarding) {
772 if ((!newf) ^ (!old))
773 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
774 NETCONFA_IFINDEX_DEFAULT,
775 net->ipv6.devconf_dflt);
776 rtnl_unlock();
777 return 0;
780 if (p == &net->ipv6.devconf_all->forwarding) {
781 int old_dflt = net->ipv6.devconf_dflt->forwarding;
783 net->ipv6.devconf_dflt->forwarding = newf;
784 if ((!newf) ^ (!old_dflt))
785 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
786 NETCONFA_IFINDEX_DEFAULT,
787 net->ipv6.devconf_dflt);
789 addrconf_forward_change(net, newf);
790 if ((!newf) ^ (!old))
791 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
792 NETCONFA_IFINDEX_ALL,
793 net->ipv6.devconf_all);
794 } else if ((!newf) ^ (!old))
795 dev_forward_change((struct inet6_dev *)table->extra1);
796 rtnl_unlock();
798 if (newf)
799 rt6_purge_dflt_routers(net);
800 return 1;
803 static void addrconf_linkdown_change(struct net *net, __s32 newf)
805 struct net_device *dev;
806 struct inet6_dev *idev;
808 for_each_netdev(net, dev) {
809 idev = __in6_dev_get(dev);
810 if (idev) {
811 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
813 idev->cnf.ignore_routes_with_linkdown = newf;
814 if (changed)
815 inet6_netconf_notify_devconf(dev_net(dev),
816 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
817 dev->ifindex,
818 &idev->cnf);
823 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
825 struct net *net;
826 int old;
828 if (!rtnl_trylock())
829 return restart_syscall();
831 net = (struct net *)table->extra2;
832 old = *p;
833 *p = newf;
835 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
836 if ((!newf) ^ (!old))
837 inet6_netconf_notify_devconf(net,
838 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
839 NETCONFA_IFINDEX_DEFAULT,
840 net->ipv6.devconf_dflt);
841 rtnl_unlock();
842 return 0;
845 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
846 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
847 addrconf_linkdown_change(net, newf);
848 if ((!newf) ^ (!old))
849 inet6_netconf_notify_devconf(net,
850 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
851 NETCONFA_IFINDEX_ALL,
852 net->ipv6.devconf_all);
854 rtnl_unlock();
856 return 1;
859 #endif
861 /* Nobody refers to this ifaddr, destroy it */
862 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
864 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
866 #ifdef NET_REFCNT_DEBUG
867 pr_debug("%s\n", __func__);
868 #endif
870 in6_dev_put(ifp->idev);
872 if (cancel_delayed_work(&ifp->dad_work))
873 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
874 ifp);
876 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
877 pr_warn("Freeing alive inet6 address %p\n", ifp);
878 return;
880 ip6_rt_put(ifp->rt);
882 kfree_rcu(ifp, rcu);
885 static void
886 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
888 struct list_head *p;
889 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
892 * Each device address list is sorted in order of scope -
893 * global before linklocal.
895 list_for_each(p, &idev->addr_list) {
896 struct inet6_ifaddr *ifa
897 = list_entry(p, struct inet6_ifaddr, if_list);
898 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
899 break;
902 list_add_tail(&ifp->if_list, p);
905 static u32 inet6_addr_hash(const struct in6_addr *addr)
907 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
910 /* On success it returns ifp with increased reference count */
912 static struct inet6_ifaddr *
913 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
914 const struct in6_addr *peer_addr, int pfxlen,
915 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
917 struct inet6_ifaddr *ifa = NULL;
918 struct rt6_info *rt;
919 unsigned int hash;
920 int err = 0;
921 int addr_type = ipv6_addr_type(addr);
923 if (addr_type == IPV6_ADDR_ANY ||
924 addr_type & IPV6_ADDR_MULTICAST ||
925 (!(idev->dev->flags & IFF_LOOPBACK) &&
926 addr_type & IPV6_ADDR_LOOPBACK))
927 return ERR_PTR(-EADDRNOTAVAIL);
929 rcu_read_lock_bh();
930 if (idev->dead) {
931 err = -ENODEV; /*XXX*/
932 goto out2;
935 if (idev->cnf.disable_ipv6) {
936 err = -EACCES;
937 goto out2;
940 spin_lock(&addrconf_hash_lock);
942 /* Ignore adding duplicate addresses on an interface */
943 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
944 ADBG("ipv6_add_addr: already assigned\n");
945 err = -EEXIST;
946 goto out;
949 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
951 if (!ifa) {
952 ADBG("ipv6_add_addr: malloc failed\n");
953 err = -ENOBUFS;
954 goto out;
957 rt = addrconf_dst_alloc(idev, addr, false);
958 if (IS_ERR(rt)) {
959 err = PTR_ERR(rt);
960 goto out;
963 neigh_parms_data_state_setall(idev->nd_parms);
965 ifa->addr = *addr;
966 if (peer_addr)
967 ifa->peer_addr = *peer_addr;
969 spin_lock_init(&ifa->lock);
970 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
971 INIT_HLIST_NODE(&ifa->addr_lst);
972 ifa->scope = scope;
973 ifa->prefix_len = pfxlen;
974 ifa->flags = flags | IFA_F_TENTATIVE;
975 ifa->valid_lft = valid_lft;
976 ifa->prefered_lft = prefered_lft;
977 ifa->cstamp = ifa->tstamp = jiffies;
978 ifa->tokenized = false;
980 ifa->rt = rt;
982 ifa->idev = idev;
983 in6_dev_hold(idev);
984 /* For caller */
985 in6_ifa_hold(ifa);
987 /* Add to big hash table */
988 hash = inet6_addr_hash(addr);
990 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
991 spin_unlock(&addrconf_hash_lock);
993 write_lock(&idev->lock);
994 /* Add to inet6_dev unicast addr list. */
995 ipv6_link_dev_addr(idev, ifa);
997 if (ifa->flags&IFA_F_TEMPORARY) {
998 list_add(&ifa->tmp_list, &idev->tempaddr_list);
999 in6_ifa_hold(ifa);
1002 in6_ifa_hold(ifa);
1003 write_unlock(&idev->lock);
1004 out2:
1005 rcu_read_unlock_bh();
1007 if (likely(err == 0))
1008 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1009 else {
1010 kfree(ifa);
1011 ifa = ERR_PTR(err);
1014 return ifa;
1015 out:
1016 spin_unlock(&addrconf_hash_lock);
1017 goto out2;
1020 enum cleanup_prefix_rt_t {
1021 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1022 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1023 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1027 * Check, whether the prefix for ifp would still need a prefix route
1028 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1029 * constants.
1031 * 1) we don't purge prefix if address was not permanent.
1032 * prefix is managed by its own lifetime.
1033 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1034 * 3) if there are no addresses, delete prefix.
1035 * 4) if there are still other permanent address(es),
1036 * corresponding prefix is still permanent.
1037 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1038 * don't purge the prefix, assume user space is managing it.
1039 * 6) otherwise, update prefix lifetime to the
1040 * longest valid lifetime among the corresponding
1041 * addresses on the device.
1042 * Note: subsequent RA will update lifetime.
1044 static enum cleanup_prefix_rt_t
1045 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1047 struct inet6_ifaddr *ifa;
1048 struct inet6_dev *idev = ifp->idev;
1049 unsigned long lifetime;
1050 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1052 *expires = jiffies;
1054 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1055 if (ifa == ifp)
1056 continue;
1057 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1058 ifp->prefix_len))
1059 continue;
1060 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1061 return CLEANUP_PREFIX_RT_NOP;
1063 action = CLEANUP_PREFIX_RT_EXPIRE;
1065 spin_lock(&ifa->lock);
1067 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1069 * Note: Because this address is
1070 * not permanent, lifetime <
1071 * LONG_MAX / HZ here.
1073 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1074 *expires = ifa->tstamp + lifetime * HZ;
1075 spin_unlock(&ifa->lock);
1078 return action;
1081 static void
1082 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1084 struct rt6_info *rt;
1086 rt = addrconf_get_prefix_route(&ifp->addr,
1087 ifp->prefix_len,
1088 ifp->idev->dev,
1089 0, RTF_GATEWAY | RTF_DEFAULT);
1090 if (rt) {
1091 if (del_rt)
1092 ip6_del_rt(rt);
1093 else {
1094 if (!(rt->rt6i_flags & RTF_EXPIRES))
1095 rt6_set_expires(rt, expires);
1096 ip6_rt_put(rt);
1102 /* This function wants to get referenced ifp and releases it before return */
1104 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1106 int state;
1107 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1108 unsigned long expires;
1110 ASSERT_RTNL();
1112 spin_lock_bh(&ifp->lock);
1113 state = ifp->state;
1114 ifp->state = INET6_IFADDR_STATE_DEAD;
1115 spin_unlock_bh(&ifp->lock);
1117 if (state == INET6_IFADDR_STATE_DEAD)
1118 goto out;
1120 spin_lock_bh(&addrconf_hash_lock);
1121 hlist_del_init_rcu(&ifp->addr_lst);
1122 spin_unlock_bh(&addrconf_hash_lock);
1124 write_lock_bh(&ifp->idev->lock);
1126 if (ifp->flags&IFA_F_TEMPORARY) {
1127 list_del(&ifp->tmp_list);
1128 if (ifp->ifpub) {
1129 in6_ifa_put(ifp->ifpub);
1130 ifp->ifpub = NULL;
1132 __in6_ifa_put(ifp);
1135 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1136 action = check_cleanup_prefix_route(ifp, &expires);
1138 list_del_init(&ifp->if_list);
1139 __in6_ifa_put(ifp);
1141 write_unlock_bh(&ifp->idev->lock);
1143 addrconf_del_dad_work(ifp);
1145 ipv6_ifa_notify(RTM_DELADDR, ifp);
1147 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1149 if (action != CLEANUP_PREFIX_RT_NOP) {
1150 cleanup_prefix_route(ifp, expires,
1151 action == CLEANUP_PREFIX_RT_DEL);
1154 /* clean up prefsrc entries */
1155 rt6_remove_prefsrc(ifp);
1156 out:
1157 in6_ifa_put(ifp);
1160 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1162 struct inet6_dev *idev = ifp->idev;
1163 struct in6_addr addr, *tmpaddr;
1164 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1165 unsigned long regen_advance;
1166 int tmp_plen;
1167 int ret = 0;
1168 u32 addr_flags;
1169 unsigned long now = jiffies;
1171 write_lock_bh(&idev->lock);
1172 if (ift) {
1173 spin_lock_bh(&ift->lock);
1174 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1175 spin_unlock_bh(&ift->lock);
1176 tmpaddr = &addr;
1177 } else {
1178 tmpaddr = NULL;
1180 retry:
1181 in6_dev_hold(idev);
1182 if (idev->cnf.use_tempaddr <= 0) {
1183 write_unlock_bh(&idev->lock);
1184 pr_info("%s: use_tempaddr is disabled\n", __func__);
1185 in6_dev_put(idev);
1186 ret = -1;
1187 goto out;
1189 spin_lock_bh(&ifp->lock);
1190 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1191 idev->cnf.use_tempaddr = -1; /*XXX*/
1192 spin_unlock_bh(&ifp->lock);
1193 write_unlock_bh(&idev->lock);
1194 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1195 __func__);
1196 in6_dev_put(idev);
1197 ret = -1;
1198 goto out;
1200 in6_ifa_hold(ifp);
1201 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1202 __ipv6_try_regen_rndid(idev, tmpaddr);
1203 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1204 age = (now - ifp->tstamp) / HZ;
1205 tmp_valid_lft = min_t(__u32,
1206 ifp->valid_lft,
1207 idev->cnf.temp_valid_lft + age);
1208 tmp_prefered_lft = min_t(__u32,
1209 ifp->prefered_lft,
1210 idev->cnf.temp_prefered_lft + age -
1211 idev->cnf.max_desync_factor);
1212 tmp_plen = ifp->prefix_len;
1213 tmp_tstamp = ifp->tstamp;
1214 spin_unlock_bh(&ifp->lock);
1216 regen_advance = idev->cnf.regen_max_retry *
1217 idev->cnf.dad_transmits *
1218 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1219 write_unlock_bh(&idev->lock);
1221 /* A temporary address is created only if this calculated Preferred
1222 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1223 * an implementation must not create a temporary address with a zero
1224 * Preferred Lifetime.
1225 * Use age calculation as in addrconf_verify to avoid unnecessary
1226 * temporary addresses being generated.
1228 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1229 if (tmp_prefered_lft <= regen_advance + age) {
1230 in6_ifa_put(ifp);
1231 in6_dev_put(idev);
1232 ret = -1;
1233 goto out;
1236 addr_flags = IFA_F_TEMPORARY;
1237 /* set in addrconf_prefix_rcv() */
1238 if (ifp->flags & IFA_F_OPTIMISTIC)
1239 addr_flags |= IFA_F_OPTIMISTIC;
1241 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1242 ipv6_addr_scope(&addr), addr_flags,
1243 tmp_valid_lft, tmp_prefered_lft);
1244 if (IS_ERR(ift)) {
1245 in6_ifa_put(ifp);
1246 in6_dev_put(idev);
1247 pr_info("%s: retry temporary address regeneration\n", __func__);
1248 tmpaddr = &addr;
1249 write_lock_bh(&idev->lock);
1250 goto retry;
1253 spin_lock_bh(&ift->lock);
1254 ift->ifpub = ifp;
1255 ift->cstamp = now;
1256 ift->tstamp = tmp_tstamp;
1257 spin_unlock_bh(&ift->lock);
1259 addrconf_dad_start(ift);
1260 in6_ifa_put(ift);
1261 in6_dev_put(idev);
1262 out:
1263 return ret;
1267 * Choose an appropriate source address (RFC3484)
1269 enum {
1270 IPV6_SADDR_RULE_INIT = 0,
1271 IPV6_SADDR_RULE_LOCAL,
1272 IPV6_SADDR_RULE_SCOPE,
1273 IPV6_SADDR_RULE_PREFERRED,
1274 #ifdef CONFIG_IPV6_MIP6
1275 IPV6_SADDR_RULE_HOA,
1276 #endif
1277 IPV6_SADDR_RULE_OIF,
1278 IPV6_SADDR_RULE_LABEL,
1279 IPV6_SADDR_RULE_PRIVACY,
1280 IPV6_SADDR_RULE_ORCHID,
1281 IPV6_SADDR_RULE_PREFIX,
1282 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1283 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1284 #endif
1285 IPV6_SADDR_RULE_MAX
1288 struct ipv6_saddr_score {
1289 int rule;
1290 int addr_type;
1291 struct inet6_ifaddr *ifa;
1292 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1293 int scopedist;
1294 int matchlen;
1297 struct ipv6_saddr_dst {
1298 const struct in6_addr *addr;
1299 int ifindex;
1300 int scope;
1301 int label;
1302 unsigned int prefs;
1305 static inline int ipv6_saddr_preferred(int type)
1307 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1308 return 1;
1309 return 0;
1312 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1314 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1315 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1316 #else
1317 return false;
1318 #endif
1321 static int ipv6_get_saddr_eval(struct net *net,
1322 struct ipv6_saddr_score *score,
1323 struct ipv6_saddr_dst *dst,
1324 int i)
1326 int ret;
1328 if (i <= score->rule) {
1329 switch (i) {
1330 case IPV6_SADDR_RULE_SCOPE:
1331 ret = score->scopedist;
1332 break;
1333 case IPV6_SADDR_RULE_PREFIX:
1334 ret = score->matchlen;
1335 break;
1336 default:
1337 ret = !!test_bit(i, score->scorebits);
1339 goto out;
1342 switch (i) {
1343 case IPV6_SADDR_RULE_INIT:
1344 /* Rule 0: remember if hiscore is not ready yet */
1345 ret = !!score->ifa;
1346 break;
1347 case IPV6_SADDR_RULE_LOCAL:
1348 /* Rule 1: Prefer same address */
1349 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1350 break;
1351 case IPV6_SADDR_RULE_SCOPE:
1352 /* Rule 2: Prefer appropriate scope
1354 * ret
1356 * -1 | d 15
1357 * ---+--+-+---> scope
1359 * | d is scope of the destination.
1360 * B-d | \
1361 * | \ <- smaller scope is better if
1362 * B-15 | \ if scope is enough for destination.
1363 * | ret = B - scope (-1 <= scope >= d <= 15).
1364 * d-C-1 | /
1365 * |/ <- greater is better
1366 * -C / if scope is not enough for destination.
1367 * /| ret = scope - C (-1 <= d < scope <= 15).
1369 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1370 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1371 * Assume B = 0 and we get C > 29.
1373 ret = __ipv6_addr_src_scope(score->addr_type);
1374 if (ret >= dst->scope)
1375 ret = -ret;
1376 else
1377 ret -= 128; /* 30 is enough */
1378 score->scopedist = ret;
1379 break;
1380 case IPV6_SADDR_RULE_PREFERRED:
1382 /* Rule 3: Avoid deprecated and optimistic addresses */
1383 u8 avoid = IFA_F_DEPRECATED;
1385 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1386 avoid |= IFA_F_OPTIMISTIC;
1387 ret = ipv6_saddr_preferred(score->addr_type) ||
1388 !(score->ifa->flags & avoid);
1389 break;
1391 #ifdef CONFIG_IPV6_MIP6
1392 case IPV6_SADDR_RULE_HOA:
1394 /* Rule 4: Prefer home address */
1395 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1396 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1397 break;
1399 #endif
1400 case IPV6_SADDR_RULE_OIF:
1401 /* Rule 5: Prefer outgoing interface */
1402 ret = (!dst->ifindex ||
1403 dst->ifindex == score->ifa->idev->dev->ifindex);
1404 break;
1405 case IPV6_SADDR_RULE_LABEL:
1406 /* Rule 6: Prefer matching label */
1407 ret = ipv6_addr_label(net,
1408 &score->ifa->addr, score->addr_type,
1409 score->ifa->idev->dev->ifindex) == dst->label;
1410 break;
1411 case IPV6_SADDR_RULE_PRIVACY:
1413 /* Rule 7: Prefer public address
1414 * Note: prefer temporary address if use_tempaddr >= 2
1416 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1417 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1418 score->ifa->idev->cnf.use_tempaddr >= 2;
1419 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1420 break;
1422 case IPV6_SADDR_RULE_ORCHID:
1423 /* Rule 8-: Prefer ORCHID vs ORCHID or
1424 * non-ORCHID vs non-ORCHID
1426 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1427 ipv6_addr_orchid(dst->addr));
1428 break;
1429 case IPV6_SADDR_RULE_PREFIX:
1430 /* Rule 8: Use longest matching prefix */
1431 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1432 if (ret > score->ifa->prefix_len)
1433 ret = score->ifa->prefix_len;
1434 score->matchlen = ret;
1435 break;
1436 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1437 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1438 /* Optimistic addresses still have lower precedence than other
1439 * preferred addresses.
1441 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1442 break;
1443 #endif
1444 default:
1445 ret = 0;
1448 if (ret)
1449 __set_bit(i, score->scorebits);
1450 score->rule = i;
1451 out:
1452 return ret;
1455 static int __ipv6_dev_get_saddr(struct net *net,
1456 struct ipv6_saddr_dst *dst,
1457 struct inet6_dev *idev,
1458 struct ipv6_saddr_score *scores,
1459 int hiscore_idx)
1461 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1463 read_lock_bh(&idev->lock);
1464 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1465 int i;
1468 * - Tentative Address (RFC2462 section 5.4)
1469 * - A tentative address is not considered
1470 * "assigned to an interface" in the traditional
1471 * sense, unless it is also flagged as optimistic.
1472 * - Candidate Source Address (section 4)
1473 * - In any case, anycast addresses, multicast
1474 * addresses, and the unspecified address MUST
1475 * NOT be included in a candidate set.
1477 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1478 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1479 continue;
1481 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1483 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1484 score->addr_type & IPV6_ADDR_MULTICAST)) {
1485 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1486 idev->dev->name);
1487 continue;
1490 score->rule = -1;
1491 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1493 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1494 int minihiscore, miniscore;
1496 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1497 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1499 if (minihiscore > miniscore) {
1500 if (i == IPV6_SADDR_RULE_SCOPE &&
1501 score->scopedist > 0) {
1503 * special case:
1504 * each remaining entry
1505 * has too small (not enough)
1506 * scope, because ifa entries
1507 * are sorted by their scope
1508 * values.
1510 goto out;
1512 break;
1513 } else if (minihiscore < miniscore) {
1514 if (hiscore->ifa)
1515 in6_ifa_put(hiscore->ifa);
1517 in6_ifa_hold(score->ifa);
1519 swap(hiscore, score);
1520 hiscore_idx = 1 - hiscore_idx;
1522 /* restore our iterator */
1523 score->ifa = hiscore->ifa;
1525 break;
1529 out:
1530 read_unlock_bh(&idev->lock);
1531 return hiscore_idx;
1534 static int ipv6_get_saddr_master(struct net *net,
1535 const struct net_device *dst_dev,
1536 const struct net_device *master,
1537 struct ipv6_saddr_dst *dst,
1538 struct ipv6_saddr_score *scores,
1539 int hiscore_idx)
1541 struct inet6_dev *idev;
1543 idev = __in6_dev_get(dst_dev);
1544 if (idev)
1545 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1546 scores, hiscore_idx);
1548 idev = __in6_dev_get(master);
1549 if (idev)
1550 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1551 scores, hiscore_idx);
1553 return hiscore_idx;
1556 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1557 const struct in6_addr *daddr, unsigned int prefs,
1558 struct in6_addr *saddr)
1560 struct ipv6_saddr_score scores[2], *hiscore;
1561 struct ipv6_saddr_dst dst;
1562 struct inet6_dev *idev;
1563 struct net_device *dev;
1564 int dst_type;
1565 bool use_oif_addr = false;
1566 int hiscore_idx = 0;
1568 dst_type = __ipv6_addr_type(daddr);
1569 dst.addr = daddr;
1570 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1571 dst.scope = __ipv6_addr_src_scope(dst_type);
1572 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1573 dst.prefs = prefs;
1575 scores[hiscore_idx].rule = -1;
1576 scores[hiscore_idx].ifa = NULL;
1578 rcu_read_lock();
1580 /* Candidate Source Address (section 4)
1581 * - multicast and link-local destination address,
1582 * the set of candidate source address MUST only
1583 * include addresses assigned to interfaces
1584 * belonging to the same link as the outgoing
1585 * interface.
1586 * (- For site-local destination addresses, the
1587 * set of candidate source addresses MUST only
1588 * include addresses assigned to interfaces
1589 * belonging to the same site as the outgoing
1590 * interface.)
1591 * - "It is RECOMMENDED that the candidate source addresses
1592 * be the set of unicast addresses assigned to the
1593 * interface that will be used to send to the destination
1594 * (the 'outgoing' interface)." (RFC 6724)
1596 if (dst_dev) {
1597 idev = __in6_dev_get(dst_dev);
1598 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1599 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1600 (idev && idev->cnf.use_oif_addrs_only)) {
1601 use_oif_addr = true;
1605 if (use_oif_addr) {
1606 if (idev)
1607 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1608 } else {
1609 const struct net_device *master;
1610 int master_idx = 0;
1612 /* if dst_dev exists and is enslaved to an L3 device, then
1613 * prefer addresses from dst_dev and then the master over
1614 * any other enslaved devices in the L3 domain.
1616 master = l3mdev_master_dev_rcu(dst_dev);
1617 if (master) {
1618 master_idx = master->ifindex;
1620 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1621 master, &dst,
1622 scores, hiscore_idx);
1624 if (scores[hiscore_idx].ifa)
1625 goto out;
1628 for_each_netdev_rcu(net, dev) {
1629 /* only consider addresses on devices in the
1630 * same L3 domain
1632 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1633 continue;
1634 idev = __in6_dev_get(dev);
1635 if (!idev)
1636 continue;
1637 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1641 out:
1642 rcu_read_unlock();
1644 hiscore = &scores[hiscore_idx];
1645 if (!hiscore->ifa)
1646 return -EADDRNOTAVAIL;
1648 *saddr = hiscore->ifa->addr;
1649 in6_ifa_put(hiscore->ifa);
1650 return 0;
1652 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1654 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1655 u32 banned_flags)
1657 struct inet6_ifaddr *ifp;
1658 int err = -EADDRNOTAVAIL;
1660 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1661 if (ifp->scope > IFA_LINK)
1662 break;
1663 if (ifp->scope == IFA_LINK &&
1664 !(ifp->flags & banned_flags)) {
1665 *addr = ifp->addr;
1666 err = 0;
1667 break;
1670 return err;
1673 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1674 u32 banned_flags)
1676 struct inet6_dev *idev;
1677 int err = -EADDRNOTAVAIL;
1679 rcu_read_lock();
1680 idev = __in6_dev_get(dev);
1681 if (idev) {
1682 read_lock_bh(&idev->lock);
1683 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1684 read_unlock_bh(&idev->lock);
1686 rcu_read_unlock();
1687 return err;
1690 static int ipv6_count_addresses(struct inet6_dev *idev)
1692 int cnt = 0;
1693 struct inet6_ifaddr *ifp;
1695 read_lock_bh(&idev->lock);
1696 list_for_each_entry(ifp, &idev->addr_list, if_list)
1697 cnt++;
1698 read_unlock_bh(&idev->lock);
1699 return cnt;
1702 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1703 const struct net_device *dev, int strict)
1705 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1707 EXPORT_SYMBOL(ipv6_chk_addr);
1709 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1710 const struct net_device *dev, int strict,
1711 u32 banned_flags)
1713 struct inet6_ifaddr *ifp;
1714 unsigned int hash = inet6_addr_hash(addr);
1715 u32 ifp_flags;
1717 rcu_read_lock_bh();
1718 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1719 if (!net_eq(dev_net(ifp->idev->dev), net))
1720 continue;
1721 /* Decouple optimistic from tentative for evaluation here.
1722 * Ban optimistic addresses explicitly, when required.
1724 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1725 ? (ifp->flags&~IFA_F_TENTATIVE)
1726 : ifp->flags;
1727 if (ipv6_addr_equal(&ifp->addr, addr) &&
1728 !(ifp_flags&banned_flags) &&
1729 (!dev || ifp->idev->dev == dev ||
1730 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1731 rcu_read_unlock_bh();
1732 return 1;
1736 rcu_read_unlock_bh();
1737 return 0;
1739 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1741 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1742 struct net_device *dev)
1744 unsigned int hash = inet6_addr_hash(addr);
1745 struct inet6_ifaddr *ifp;
1747 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1748 if (!net_eq(dev_net(ifp->idev->dev), net))
1749 continue;
1750 if (ipv6_addr_equal(&ifp->addr, addr)) {
1751 if (!dev || ifp->idev->dev == dev)
1752 return true;
1755 return false;
1758 /* Compares an address/prefix_len with addresses on device @dev.
1759 * If one is found it returns true.
1761 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1762 const unsigned int prefix_len, struct net_device *dev)
1764 struct inet6_dev *idev;
1765 struct inet6_ifaddr *ifa;
1766 bool ret = false;
1768 rcu_read_lock();
1769 idev = __in6_dev_get(dev);
1770 if (idev) {
1771 read_lock_bh(&idev->lock);
1772 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1773 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1774 if (ret)
1775 break;
1777 read_unlock_bh(&idev->lock);
1779 rcu_read_unlock();
1781 return ret;
1783 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1785 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1787 struct inet6_dev *idev;
1788 struct inet6_ifaddr *ifa;
1789 int onlink;
1791 onlink = 0;
1792 rcu_read_lock();
1793 idev = __in6_dev_get(dev);
1794 if (idev) {
1795 read_lock_bh(&idev->lock);
1796 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1797 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1798 ifa->prefix_len);
1799 if (onlink)
1800 break;
1802 read_unlock_bh(&idev->lock);
1804 rcu_read_unlock();
1805 return onlink;
1807 EXPORT_SYMBOL(ipv6_chk_prefix);
1809 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1810 struct net_device *dev, int strict)
1812 struct inet6_ifaddr *ifp, *result = NULL;
1813 unsigned int hash = inet6_addr_hash(addr);
1815 rcu_read_lock_bh();
1816 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1817 if (!net_eq(dev_net(ifp->idev->dev), net))
1818 continue;
1819 if (ipv6_addr_equal(&ifp->addr, addr)) {
1820 if (!dev || ifp->idev->dev == dev ||
1821 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1822 result = ifp;
1823 in6_ifa_hold(ifp);
1824 break;
1828 rcu_read_unlock_bh();
1830 return result;
1833 /* Gets referenced address, destroys ifaddr */
1835 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1837 if (dad_failed)
1838 ifp->flags |= IFA_F_DADFAILED;
1840 if (ifp->flags&IFA_F_PERMANENT) {
1841 spin_lock_bh(&ifp->lock);
1842 addrconf_del_dad_work(ifp);
1843 ifp->flags |= IFA_F_TENTATIVE;
1844 spin_unlock_bh(&ifp->lock);
1845 if (dad_failed)
1846 ipv6_ifa_notify(0, ifp);
1847 in6_ifa_put(ifp);
1848 } else if (ifp->flags&IFA_F_TEMPORARY) {
1849 struct inet6_ifaddr *ifpub;
1850 spin_lock_bh(&ifp->lock);
1851 ifpub = ifp->ifpub;
1852 if (ifpub) {
1853 in6_ifa_hold(ifpub);
1854 spin_unlock_bh(&ifp->lock);
1855 ipv6_create_tempaddr(ifpub, ifp);
1856 in6_ifa_put(ifpub);
1857 } else {
1858 spin_unlock_bh(&ifp->lock);
1860 ipv6_del_addr(ifp);
1861 } else {
1862 ipv6_del_addr(ifp);
1866 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1868 int err = -ENOENT;
1870 spin_lock_bh(&ifp->lock);
1871 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1872 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1873 err = 0;
1875 spin_unlock_bh(&ifp->lock);
1877 return err;
1880 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1882 struct inet6_dev *idev = ifp->idev;
1883 struct net *net = dev_net(ifp->idev->dev);
1885 if (addrconf_dad_end(ifp)) {
1886 in6_ifa_put(ifp);
1887 return;
1890 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1891 ifp->idev->dev->name, &ifp->addr);
1893 spin_lock_bh(&ifp->lock);
1895 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1896 int scope = ifp->scope;
1897 u32 flags = ifp->flags;
1898 struct in6_addr new_addr;
1899 struct inet6_ifaddr *ifp2;
1900 u32 valid_lft, preferred_lft;
1901 int pfxlen = ifp->prefix_len;
1902 int retries = ifp->stable_privacy_retry + 1;
1904 if (retries > net->ipv6.sysctl.idgen_retries) {
1905 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1906 ifp->idev->dev->name);
1907 goto errdad;
1910 new_addr = ifp->addr;
1911 if (ipv6_generate_stable_address(&new_addr, retries,
1912 idev))
1913 goto errdad;
1915 valid_lft = ifp->valid_lft;
1916 preferred_lft = ifp->prefered_lft;
1918 spin_unlock_bh(&ifp->lock);
1920 if (idev->cnf.max_addresses &&
1921 ipv6_count_addresses(idev) >=
1922 idev->cnf.max_addresses)
1923 goto lock_errdad;
1925 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1926 ifp->idev->dev->name);
1928 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1929 scope, flags, valid_lft,
1930 preferred_lft);
1931 if (IS_ERR(ifp2))
1932 goto lock_errdad;
1934 spin_lock_bh(&ifp2->lock);
1935 ifp2->stable_privacy_retry = retries;
1936 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1937 spin_unlock_bh(&ifp2->lock);
1939 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1940 in6_ifa_put(ifp2);
1941 lock_errdad:
1942 spin_lock_bh(&ifp->lock);
1945 errdad:
1946 /* transition from _POSTDAD to _ERRDAD */
1947 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1948 spin_unlock_bh(&ifp->lock);
1950 addrconf_mod_dad_work(ifp, 0);
1951 in6_ifa_put(ifp);
1954 /* Join to solicited addr multicast group.
1955 * caller must hold RTNL */
1956 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1958 struct in6_addr maddr;
1960 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1961 return;
1963 addrconf_addr_solict_mult(addr, &maddr);
1964 ipv6_dev_mc_inc(dev, &maddr);
1967 /* caller must hold RTNL */
1968 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1970 struct in6_addr maddr;
1972 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1973 return;
1975 addrconf_addr_solict_mult(addr, &maddr);
1976 __ipv6_dev_mc_dec(idev, &maddr);
1979 /* caller must hold RTNL */
1980 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1982 struct in6_addr addr;
1984 if (ifp->prefix_len >= 127) /* RFC 6164 */
1985 return;
1986 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1987 if (ipv6_addr_any(&addr))
1988 return;
1989 __ipv6_dev_ac_inc(ifp->idev, &addr);
1992 /* caller must hold RTNL */
1993 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1995 struct in6_addr addr;
1997 if (ifp->prefix_len >= 127) /* RFC 6164 */
1998 return;
1999 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2000 if (ipv6_addr_any(&addr))
2001 return;
2002 __ipv6_dev_ac_dec(ifp->idev, &addr);
2005 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2007 if (dev->addr_len != EUI64_ADDR_LEN)
2008 return -1;
2009 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2010 eui[0] ^= 2;
2011 return 0;
2014 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2016 union fwnet_hwaddr *ha;
2018 if (dev->addr_len != FWNET_ALEN)
2019 return -1;
2021 ha = (union fwnet_hwaddr *)dev->dev_addr;
2023 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2024 eui[0] ^= 2;
2025 return 0;
2028 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2030 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2031 if (dev->addr_len != ARCNET_ALEN)
2032 return -1;
2033 memset(eui, 0, 7);
2034 eui[7] = *(u8 *)dev->dev_addr;
2035 return 0;
2038 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2040 if (dev->addr_len != INFINIBAND_ALEN)
2041 return -1;
2042 memcpy(eui, dev->dev_addr + 12, 8);
2043 eui[0] |= 2;
2044 return 0;
2047 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2049 if (addr == 0)
2050 return -1;
2051 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2052 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2053 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2054 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2055 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2056 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2057 eui[1] = 0;
2058 eui[2] = 0x5E;
2059 eui[3] = 0xFE;
2060 memcpy(eui + 4, &addr, 4);
2061 return 0;
2064 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2066 if (dev->priv_flags & IFF_ISATAP)
2067 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2068 return -1;
2071 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2073 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2076 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2078 memcpy(eui, dev->perm_addr, 3);
2079 memcpy(eui + 5, dev->perm_addr + 3, 3);
2080 eui[3] = 0xFF;
2081 eui[4] = 0xFE;
2082 eui[0] ^= 2;
2083 return 0;
2086 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2088 switch (dev->type) {
2089 case ARPHRD_ETHER:
2090 case ARPHRD_FDDI:
2091 return addrconf_ifid_eui48(eui, dev);
2092 case ARPHRD_ARCNET:
2093 return addrconf_ifid_arcnet(eui, dev);
2094 case ARPHRD_INFINIBAND:
2095 return addrconf_ifid_infiniband(eui, dev);
2096 case ARPHRD_SIT:
2097 return addrconf_ifid_sit(eui, dev);
2098 case ARPHRD_IPGRE:
2099 return addrconf_ifid_gre(eui, dev);
2100 case ARPHRD_6LOWPAN:
2101 return addrconf_ifid_eui64(eui, dev);
2102 case ARPHRD_IEEE1394:
2103 return addrconf_ifid_ieee1394(eui, dev);
2104 case ARPHRD_TUNNEL6:
2105 return addrconf_ifid_ip6tnl(eui, dev);
2107 return -1;
2110 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2112 int err = -1;
2113 struct inet6_ifaddr *ifp;
2115 read_lock_bh(&idev->lock);
2116 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2117 if (ifp->scope > IFA_LINK)
2118 break;
2119 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2120 memcpy(eui, ifp->addr.s6_addr+8, 8);
2121 err = 0;
2122 break;
2125 read_unlock_bh(&idev->lock);
2126 return err;
2129 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2130 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2132 regen:
2133 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2134 idev->rndid[0] &= ~0x02;
2137 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2138 * check if generated address is not inappropriate
2140 * - Reserved subnet anycast (RFC 2526)
2141 * 11111101 11....11 1xxxxxxx
2142 * - ISATAP (RFC4214) 6.1
2143 * 00-00-5E-FE-xx-xx-xx-xx
2144 * - value 0
2145 * - XXX: already assigned to an address on the device
2147 if (idev->rndid[0] == 0xfd &&
2148 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2149 (idev->rndid[7]&0x80))
2150 goto regen;
2151 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2152 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2153 goto regen;
2154 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2155 goto regen;
2159 static void ipv6_regen_rndid(unsigned long data)
2161 struct inet6_dev *idev = (struct inet6_dev *) data;
2162 unsigned long expires;
2164 rcu_read_lock_bh();
2165 write_lock_bh(&idev->lock);
2167 if (idev->dead)
2168 goto out;
2170 __ipv6_regen_rndid(idev);
2172 expires = jiffies +
2173 idev->cnf.temp_prefered_lft * HZ -
2174 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2175 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2176 idev->cnf.max_desync_factor * HZ;
2177 if (time_before(expires, jiffies)) {
2178 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2179 __func__, idev->dev->name);
2180 goto out;
2183 if (!mod_timer(&idev->regen_timer, expires))
2184 in6_dev_hold(idev);
2186 out:
2187 write_unlock_bh(&idev->lock);
2188 rcu_read_unlock_bh();
2189 in6_dev_put(idev);
2192 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2194 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2195 __ipv6_regen_rndid(idev);
2199 * Add prefix route.
2202 static void
2203 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2204 unsigned long expires, u32 flags)
2206 struct fib6_config cfg = {
2207 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2208 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2209 .fc_ifindex = dev->ifindex,
2210 .fc_expires = expires,
2211 .fc_dst_len = plen,
2212 .fc_flags = RTF_UP | flags,
2213 .fc_nlinfo.nl_net = dev_net(dev),
2214 .fc_protocol = RTPROT_KERNEL,
2217 cfg.fc_dst = *pfx;
2219 /* Prevent useless cloning on PtP SIT.
2220 This thing is done here expecting that the whole
2221 class of non-broadcast devices need not cloning.
2223 #if IS_ENABLED(CONFIG_IPV6_SIT)
2224 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2225 cfg.fc_flags |= RTF_NONEXTHOP;
2226 #endif
2228 ip6_route_add(&cfg);
2232 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2233 int plen,
2234 const struct net_device *dev,
2235 u32 flags, u32 noflags)
2237 struct fib6_node *fn;
2238 struct rt6_info *rt = NULL;
2239 struct fib6_table *table;
2240 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2242 table = fib6_get_table(dev_net(dev), tb_id);
2243 if (!table)
2244 return NULL;
2246 read_lock_bh(&table->tb6_lock);
2247 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2248 if (!fn)
2249 goto out;
2251 noflags |= RTF_CACHE;
2252 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2253 if (rt->dst.dev->ifindex != dev->ifindex)
2254 continue;
2255 if ((rt->rt6i_flags & flags) != flags)
2256 continue;
2257 if ((rt->rt6i_flags & noflags) != 0)
2258 continue;
2259 dst_hold(&rt->dst);
2260 break;
2262 out:
2263 read_unlock_bh(&table->tb6_lock);
2264 return rt;
2268 /* Create "default" multicast route to the interface */
2270 static void addrconf_add_mroute(struct net_device *dev)
2272 struct fib6_config cfg = {
2273 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2274 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2275 .fc_ifindex = dev->ifindex,
2276 .fc_dst_len = 8,
2277 .fc_flags = RTF_UP,
2278 .fc_nlinfo.nl_net = dev_net(dev),
2281 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2283 ip6_route_add(&cfg);
2286 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2288 struct inet6_dev *idev;
2290 ASSERT_RTNL();
2292 idev = ipv6_find_idev(dev);
2293 if (!idev)
2294 return ERR_PTR(-ENOBUFS);
2296 if (idev->cnf.disable_ipv6)
2297 return ERR_PTR(-EACCES);
2299 /* Add default multicast route */
2300 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2301 addrconf_add_mroute(dev);
2303 return idev;
2306 static void manage_tempaddrs(struct inet6_dev *idev,
2307 struct inet6_ifaddr *ifp,
2308 __u32 valid_lft, __u32 prefered_lft,
2309 bool create, unsigned long now)
2311 u32 flags;
2312 struct inet6_ifaddr *ift;
2314 read_lock_bh(&idev->lock);
2315 /* update all temporary addresses in the list */
2316 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2317 int age, max_valid, max_prefered;
2319 if (ifp != ift->ifpub)
2320 continue;
2322 /* RFC 4941 section 3.3:
2323 * If a received option will extend the lifetime of a public
2324 * address, the lifetimes of temporary addresses should
2325 * be extended, subject to the overall constraint that no
2326 * temporary addresses should ever remain "valid" or "preferred"
2327 * for a time longer than (TEMP_VALID_LIFETIME) or
2328 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2330 age = (now - ift->cstamp) / HZ;
2331 max_valid = idev->cnf.temp_valid_lft - age;
2332 if (max_valid < 0)
2333 max_valid = 0;
2335 max_prefered = idev->cnf.temp_prefered_lft -
2336 idev->cnf.max_desync_factor - age;
2337 if (max_prefered < 0)
2338 max_prefered = 0;
2340 if (valid_lft > max_valid)
2341 valid_lft = max_valid;
2343 if (prefered_lft > max_prefered)
2344 prefered_lft = max_prefered;
2346 spin_lock(&ift->lock);
2347 flags = ift->flags;
2348 ift->valid_lft = valid_lft;
2349 ift->prefered_lft = prefered_lft;
2350 ift->tstamp = now;
2351 if (prefered_lft > 0)
2352 ift->flags &= ~IFA_F_DEPRECATED;
2354 spin_unlock(&ift->lock);
2355 if (!(flags&IFA_F_TENTATIVE))
2356 ipv6_ifa_notify(0, ift);
2359 if ((create || list_empty(&idev->tempaddr_list)) &&
2360 idev->cnf.use_tempaddr > 0) {
2361 /* When a new public address is created as described
2362 * in [ADDRCONF], also create a new temporary address.
2363 * Also create a temporary address if it's enabled but
2364 * no temporary address currently exists.
2366 read_unlock_bh(&idev->lock);
2367 ipv6_create_tempaddr(ifp, NULL);
2368 } else {
2369 read_unlock_bh(&idev->lock);
2373 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2375 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2376 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2379 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2380 const struct prefix_info *pinfo,
2381 struct inet6_dev *in6_dev,
2382 const struct in6_addr *addr, int addr_type,
2383 u32 addr_flags, bool sllao, bool tokenized,
2384 __u32 valid_lft, u32 prefered_lft)
2386 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2387 int create = 0, update_lft = 0;
2389 if (!ifp && valid_lft) {
2390 int max_addresses = in6_dev->cnf.max_addresses;
2392 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2393 if (in6_dev->cnf.optimistic_dad &&
2394 !net->ipv6.devconf_all->forwarding && sllao)
2395 addr_flags |= IFA_F_OPTIMISTIC;
2396 #endif
2398 /* Do not allow to create too much of autoconfigured
2399 * addresses; this would be too easy way to crash kernel.
2401 if (!max_addresses ||
2402 ipv6_count_addresses(in6_dev) < max_addresses)
2403 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2404 pinfo->prefix_len,
2405 addr_type&IPV6_ADDR_SCOPE_MASK,
2406 addr_flags, valid_lft,
2407 prefered_lft);
2409 if (IS_ERR_OR_NULL(ifp))
2410 return -1;
2412 update_lft = 0;
2413 create = 1;
2414 spin_lock_bh(&ifp->lock);
2415 ifp->flags |= IFA_F_MANAGETEMPADDR;
2416 ifp->cstamp = jiffies;
2417 ifp->tokenized = tokenized;
2418 spin_unlock_bh(&ifp->lock);
2419 addrconf_dad_start(ifp);
2422 if (ifp) {
2423 u32 flags;
2424 unsigned long now;
2425 u32 stored_lft;
2427 /* update lifetime (RFC2462 5.5.3 e) */
2428 spin_lock_bh(&ifp->lock);
2429 now = jiffies;
2430 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2431 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2432 else
2433 stored_lft = 0;
2434 if (!update_lft && !create && stored_lft) {
2435 const u32 minimum_lft = min_t(u32,
2436 stored_lft, MIN_VALID_LIFETIME);
2437 valid_lft = max(valid_lft, minimum_lft);
2439 /* RFC4862 Section 5.5.3e:
2440 * "Note that the preferred lifetime of the
2441 * corresponding address is always reset to
2442 * the Preferred Lifetime in the received
2443 * Prefix Information option, regardless of
2444 * whether the valid lifetime is also reset or
2445 * ignored."
2447 * So we should always update prefered_lft here.
2449 update_lft = 1;
2452 if (update_lft) {
2453 ifp->valid_lft = valid_lft;
2454 ifp->prefered_lft = prefered_lft;
2455 ifp->tstamp = now;
2456 flags = ifp->flags;
2457 ifp->flags &= ~IFA_F_DEPRECATED;
2458 spin_unlock_bh(&ifp->lock);
2460 if (!(flags&IFA_F_TENTATIVE))
2461 ipv6_ifa_notify(0, ifp);
2462 } else
2463 spin_unlock_bh(&ifp->lock);
2465 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2466 create, now);
2468 in6_ifa_put(ifp);
2469 addrconf_verify();
2472 return 0;
2474 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2476 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2478 struct prefix_info *pinfo;
2479 __u32 valid_lft;
2480 __u32 prefered_lft;
2481 int addr_type, err;
2482 u32 addr_flags = 0;
2483 struct inet6_dev *in6_dev;
2484 struct net *net = dev_net(dev);
2486 pinfo = (struct prefix_info *) opt;
2488 if (len < sizeof(struct prefix_info)) {
2489 ADBG("addrconf: prefix option too short\n");
2490 return;
2494 * Validation checks ([ADDRCONF], page 19)
2497 addr_type = ipv6_addr_type(&pinfo->prefix);
2499 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2500 return;
2502 valid_lft = ntohl(pinfo->valid);
2503 prefered_lft = ntohl(pinfo->prefered);
2505 if (prefered_lft > valid_lft) {
2506 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2507 return;
2510 in6_dev = in6_dev_get(dev);
2512 if (!in6_dev) {
2513 net_dbg_ratelimited("addrconf: device %s not configured\n",
2514 dev->name);
2515 return;
2519 * Two things going on here:
2520 * 1) Add routes for on-link prefixes
2521 * 2) Configure prefixes with the auto flag set
2524 if (pinfo->onlink) {
2525 struct rt6_info *rt;
2526 unsigned long rt_expires;
2528 /* Avoid arithmetic overflow. Really, we could
2529 * save rt_expires in seconds, likely valid_lft,
2530 * but it would require division in fib gc, that it
2531 * not good.
2533 if (HZ > USER_HZ)
2534 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2535 else
2536 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2538 if (addrconf_finite_timeout(rt_expires))
2539 rt_expires *= HZ;
2541 rt = addrconf_get_prefix_route(&pinfo->prefix,
2542 pinfo->prefix_len,
2543 dev,
2544 RTF_ADDRCONF | RTF_PREFIX_RT,
2545 RTF_GATEWAY | RTF_DEFAULT);
2547 if (rt) {
2548 /* Autoconf prefix route */
2549 if (valid_lft == 0) {
2550 ip6_del_rt(rt);
2551 rt = NULL;
2552 } else if (addrconf_finite_timeout(rt_expires)) {
2553 /* not infinity */
2554 rt6_set_expires(rt, jiffies + rt_expires);
2555 } else {
2556 rt6_clean_expires(rt);
2558 } else if (valid_lft) {
2559 clock_t expires = 0;
2560 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2561 if (addrconf_finite_timeout(rt_expires)) {
2562 /* not infinity */
2563 flags |= RTF_EXPIRES;
2564 expires = jiffies_to_clock_t(rt_expires);
2566 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2567 dev, expires, flags);
2569 ip6_rt_put(rt);
2572 /* Try to figure out our local address for this prefix */
2574 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2575 struct in6_addr addr;
2576 bool tokenized = false, dev_addr_generated = false;
2578 if (pinfo->prefix_len == 64) {
2579 memcpy(&addr, &pinfo->prefix, 8);
2581 if (!ipv6_addr_any(&in6_dev->token)) {
2582 read_lock_bh(&in6_dev->lock);
2583 memcpy(addr.s6_addr + 8,
2584 in6_dev->token.s6_addr + 8, 8);
2585 read_unlock_bh(&in6_dev->lock);
2586 tokenized = true;
2587 } else if (is_addr_mode_generate_stable(in6_dev) &&
2588 !ipv6_generate_stable_address(&addr, 0,
2589 in6_dev)) {
2590 addr_flags |= IFA_F_STABLE_PRIVACY;
2591 goto ok;
2592 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2593 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2594 goto put;
2595 } else {
2596 dev_addr_generated = true;
2598 goto ok;
2600 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2601 pinfo->prefix_len);
2602 goto put;
2605 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2606 &addr, addr_type,
2607 addr_flags, sllao,
2608 tokenized, valid_lft,
2609 prefered_lft);
2610 if (err)
2611 goto put;
2613 /* Ignore error case here because previous prefix add addr was
2614 * successful which will be notified.
2616 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2617 addr_type, addr_flags, sllao,
2618 tokenized, valid_lft,
2619 prefered_lft,
2620 dev_addr_generated);
2622 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2623 put:
2624 in6_dev_put(in6_dev);
2628 * Set destination address.
2629 * Special case for SIT interfaces where we create a new "virtual"
2630 * device.
2632 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2634 struct in6_ifreq ireq;
2635 struct net_device *dev;
2636 int err = -EINVAL;
2638 rtnl_lock();
2640 err = -EFAULT;
2641 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2642 goto err_exit;
2644 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2646 err = -ENODEV;
2647 if (!dev)
2648 goto err_exit;
2650 #if IS_ENABLED(CONFIG_IPV6_SIT)
2651 if (dev->type == ARPHRD_SIT) {
2652 const struct net_device_ops *ops = dev->netdev_ops;
2653 struct ifreq ifr;
2654 struct ip_tunnel_parm p;
2656 err = -EADDRNOTAVAIL;
2657 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2658 goto err_exit;
2660 memset(&p, 0, sizeof(p));
2661 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2662 p.iph.saddr = 0;
2663 p.iph.version = 4;
2664 p.iph.ihl = 5;
2665 p.iph.protocol = IPPROTO_IPV6;
2666 p.iph.ttl = 64;
2667 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2669 if (ops->ndo_do_ioctl) {
2670 mm_segment_t oldfs = get_fs();
2672 set_fs(KERNEL_DS);
2673 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2674 set_fs(oldfs);
2675 } else
2676 err = -EOPNOTSUPP;
2678 if (err == 0) {
2679 err = -ENOBUFS;
2680 dev = __dev_get_by_name(net, p.name);
2681 if (!dev)
2682 goto err_exit;
2683 err = dev_open(dev);
2686 #endif
2688 err_exit:
2689 rtnl_unlock();
2690 return err;
2693 static int ipv6_mc_config(struct sock *sk, bool join,
2694 const struct in6_addr *addr, int ifindex)
2696 int ret;
2698 ASSERT_RTNL();
2700 lock_sock(sk);
2701 if (join)
2702 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2703 else
2704 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2705 release_sock(sk);
2707 return ret;
2711 * Manual configuration of address on an interface
2713 static int inet6_addr_add(struct net *net, int ifindex,
2714 const struct in6_addr *pfx,
2715 const struct in6_addr *peer_pfx,
2716 unsigned int plen, __u32 ifa_flags,
2717 __u32 prefered_lft, __u32 valid_lft)
2719 struct inet6_ifaddr *ifp;
2720 struct inet6_dev *idev;
2721 struct net_device *dev;
2722 unsigned long timeout;
2723 clock_t expires;
2724 int scope;
2725 u32 flags;
2727 ASSERT_RTNL();
2729 if (plen > 128)
2730 return -EINVAL;
2732 /* check the lifetime */
2733 if (!valid_lft || prefered_lft > valid_lft)
2734 return -EINVAL;
2736 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2737 return -EINVAL;
2739 dev = __dev_get_by_index(net, ifindex);
2740 if (!dev)
2741 return -ENODEV;
2743 idev = addrconf_add_dev(dev);
2744 if (IS_ERR(idev))
2745 return PTR_ERR(idev);
2747 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2748 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2749 true, pfx, ifindex);
2751 if (ret < 0)
2752 return ret;
2755 scope = ipv6_addr_scope(pfx);
2757 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2758 if (addrconf_finite_timeout(timeout)) {
2759 expires = jiffies_to_clock_t(timeout * HZ);
2760 valid_lft = timeout;
2761 flags = RTF_EXPIRES;
2762 } else {
2763 expires = 0;
2764 flags = 0;
2765 ifa_flags |= IFA_F_PERMANENT;
2768 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2769 if (addrconf_finite_timeout(timeout)) {
2770 if (timeout == 0)
2771 ifa_flags |= IFA_F_DEPRECATED;
2772 prefered_lft = timeout;
2775 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2776 valid_lft, prefered_lft);
2778 if (!IS_ERR(ifp)) {
2779 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2780 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2781 expires, flags);
2785 * Note that section 3.1 of RFC 4429 indicates
2786 * that the Optimistic flag should not be set for
2787 * manually configured addresses
2789 addrconf_dad_start(ifp);
2790 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2791 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2792 true, jiffies);
2793 in6_ifa_put(ifp);
2794 addrconf_verify_rtnl();
2795 return 0;
2796 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2797 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2798 false, pfx, ifindex);
2801 return PTR_ERR(ifp);
2804 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2805 const struct in6_addr *pfx, unsigned int plen)
2807 struct inet6_ifaddr *ifp;
2808 struct inet6_dev *idev;
2809 struct net_device *dev;
2811 if (plen > 128)
2812 return -EINVAL;
2814 dev = __dev_get_by_index(net, ifindex);
2815 if (!dev)
2816 return -ENODEV;
2818 idev = __in6_dev_get(dev);
2819 if (!idev)
2820 return -ENXIO;
2822 read_lock_bh(&idev->lock);
2823 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2824 if (ifp->prefix_len == plen &&
2825 ipv6_addr_equal(pfx, &ifp->addr)) {
2826 in6_ifa_hold(ifp);
2827 read_unlock_bh(&idev->lock);
2829 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2830 (ifa_flags & IFA_F_MANAGETEMPADDR))
2831 manage_tempaddrs(idev, ifp, 0, 0, false,
2832 jiffies);
2833 ipv6_del_addr(ifp);
2834 addrconf_verify_rtnl();
2835 if (ipv6_addr_is_multicast(pfx)) {
2836 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2837 false, pfx, dev->ifindex);
2839 return 0;
2842 read_unlock_bh(&idev->lock);
2843 return -EADDRNOTAVAIL;
2847 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2849 struct in6_ifreq ireq;
2850 int err;
2852 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2853 return -EPERM;
2855 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2856 return -EFAULT;
2858 rtnl_lock();
2859 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2860 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2861 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2862 rtnl_unlock();
2863 return err;
2866 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2868 struct in6_ifreq ireq;
2869 int err;
2871 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2872 return -EPERM;
2874 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2875 return -EFAULT;
2877 rtnl_lock();
2878 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2879 ireq.ifr6_prefixlen);
2880 rtnl_unlock();
2881 return err;
2884 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2885 int plen, int scope)
2887 struct inet6_ifaddr *ifp;
2889 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2890 scope, IFA_F_PERMANENT,
2891 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2892 if (!IS_ERR(ifp)) {
2893 spin_lock_bh(&ifp->lock);
2894 ifp->flags &= ~IFA_F_TENTATIVE;
2895 spin_unlock_bh(&ifp->lock);
2896 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2897 in6_ifa_put(ifp);
2901 #if IS_ENABLED(CONFIG_IPV6_SIT)
2902 static void sit_add_v4_addrs(struct inet6_dev *idev)
2904 struct in6_addr addr;
2905 struct net_device *dev;
2906 struct net *net = dev_net(idev->dev);
2907 int scope, plen;
2908 u32 pflags = 0;
2910 ASSERT_RTNL();
2912 memset(&addr, 0, sizeof(struct in6_addr));
2913 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2915 if (idev->dev->flags&IFF_POINTOPOINT) {
2916 addr.s6_addr32[0] = htonl(0xfe800000);
2917 scope = IFA_LINK;
2918 plen = 64;
2919 } else {
2920 scope = IPV6_ADDR_COMPATv4;
2921 plen = 96;
2922 pflags |= RTF_NONEXTHOP;
2925 if (addr.s6_addr32[3]) {
2926 add_addr(idev, &addr, plen, scope);
2927 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2928 return;
2931 for_each_netdev(net, dev) {
2932 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2933 if (in_dev && (dev->flags & IFF_UP)) {
2934 struct in_ifaddr *ifa;
2936 int flag = scope;
2938 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2940 addr.s6_addr32[3] = ifa->ifa_local;
2942 if (ifa->ifa_scope == RT_SCOPE_LINK)
2943 continue;
2944 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2945 if (idev->dev->flags&IFF_POINTOPOINT)
2946 continue;
2947 flag |= IFA_HOST;
2950 add_addr(idev, &addr, plen, flag);
2951 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2952 pflags);
2957 #endif
2959 static void init_loopback(struct net_device *dev)
2961 struct inet6_dev *idev;
2962 struct net_device *sp_dev;
2963 struct inet6_ifaddr *sp_ifa;
2964 struct rt6_info *sp_rt;
2966 /* ::1 */
2968 ASSERT_RTNL();
2970 idev = ipv6_find_idev(dev);
2971 if (!idev) {
2972 pr_debug("%s: add_dev failed\n", __func__);
2973 return;
2976 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2978 /* Add routes to other interface's IPv6 addresses */
2979 for_each_netdev(dev_net(dev), sp_dev) {
2980 if (!strcmp(sp_dev->name, dev->name))
2981 continue;
2983 idev = __in6_dev_get(sp_dev);
2984 if (!idev)
2985 continue;
2987 read_lock_bh(&idev->lock);
2988 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2990 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2991 continue;
2993 if (sp_ifa->rt) {
2994 /* This dst has been added to garbage list when
2995 * lo device down, release this obsolete dst and
2996 * reallocate a new router for ifa.
2998 if (sp_ifa->rt->dst.obsolete > 0) {
2999 ip6_rt_put(sp_ifa->rt);
3000 sp_ifa->rt = NULL;
3001 } else {
3002 continue;
3006 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3008 /* Failure cases are ignored */
3009 if (!IS_ERR(sp_rt)) {
3010 sp_ifa->rt = sp_rt;
3011 ip6_ins_rt(sp_rt);
3014 read_unlock_bh(&idev->lock);
3018 void addrconf_add_linklocal(struct inet6_dev *idev,
3019 const struct in6_addr *addr, u32 flags)
3021 struct inet6_ifaddr *ifp;
3022 u32 addr_flags = flags | IFA_F_PERMANENT;
3024 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3025 if (idev->cnf.optimistic_dad &&
3026 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3027 addr_flags |= IFA_F_OPTIMISTIC;
3028 #endif
3030 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3031 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3032 if (!IS_ERR(ifp)) {
3033 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3034 addrconf_dad_start(ifp);
3035 in6_ifa_put(ifp);
3038 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3040 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3042 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3043 return true;
3045 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3046 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3047 return true;
3049 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3050 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3051 return true;
3053 return false;
3056 static int ipv6_generate_stable_address(struct in6_addr *address,
3057 u8 dad_count,
3058 const struct inet6_dev *idev)
3060 static DEFINE_SPINLOCK(lock);
3061 static __u32 digest[SHA_DIGEST_WORDS];
3062 static __u32 workspace[SHA_WORKSPACE_WORDS];
3064 static union {
3065 char __data[SHA_MESSAGE_BYTES];
3066 struct {
3067 struct in6_addr secret;
3068 __be32 prefix[2];
3069 unsigned char hwaddr[MAX_ADDR_LEN];
3070 u8 dad_count;
3071 } __packed;
3072 } data;
3074 struct in6_addr secret;
3075 struct in6_addr temp;
3076 struct net *net = dev_net(idev->dev);
3078 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3080 if (idev->cnf.stable_secret.initialized)
3081 secret = idev->cnf.stable_secret.secret;
3082 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3083 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3084 else
3085 return -1;
3087 retry:
3088 spin_lock_bh(&lock);
3090 sha_init(digest);
3091 memset(&data, 0, sizeof(data));
3092 memset(workspace, 0, sizeof(workspace));
3093 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3094 data.prefix[0] = address->s6_addr32[0];
3095 data.prefix[1] = address->s6_addr32[1];
3096 data.secret = secret;
3097 data.dad_count = dad_count;
3099 sha_transform(digest, data.__data, workspace);
3101 temp = *address;
3102 temp.s6_addr32[2] = (__force __be32)digest[0];
3103 temp.s6_addr32[3] = (__force __be32)digest[1];
3105 spin_unlock_bh(&lock);
3107 if (ipv6_reserved_interfaceid(temp)) {
3108 dad_count++;
3109 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3110 return -1;
3111 goto retry;
3114 *address = temp;
3115 return 0;
3118 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3120 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3122 if (s->initialized)
3123 return;
3124 s = &idev->cnf.stable_secret;
3125 get_random_bytes(&s->secret, sizeof(s->secret));
3126 s->initialized = true;
3129 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3131 struct in6_addr addr;
3133 /* no link local addresses on L3 master devices */
3134 if (netif_is_l3_master(idev->dev))
3135 return;
3137 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3139 switch (idev->addr_gen_mode) {
3140 case IN6_ADDR_GEN_MODE_RANDOM:
3141 ipv6_gen_mode_random_init(idev);
3142 /* fallthrough */
3143 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3144 if (!ipv6_generate_stable_address(&addr, 0, idev))
3145 addrconf_add_linklocal(idev, &addr,
3146 IFA_F_STABLE_PRIVACY);
3147 else if (prefix_route)
3148 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3149 break;
3150 case IN6_ADDR_GEN_MODE_EUI64:
3151 /* addrconf_add_linklocal also adds a prefix_route and we
3152 * only need to care about prefix routes if ipv6_generate_eui64
3153 * couldn't generate one.
3155 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3156 addrconf_add_linklocal(idev, &addr, 0);
3157 else if (prefix_route)
3158 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3159 break;
3160 case IN6_ADDR_GEN_MODE_NONE:
3161 default:
3162 /* will not add any link local address */
3163 break;
3167 static void addrconf_dev_config(struct net_device *dev)
3169 struct inet6_dev *idev;
3171 ASSERT_RTNL();
3173 if ((dev->type != ARPHRD_ETHER) &&
3174 (dev->type != ARPHRD_FDDI) &&
3175 (dev->type != ARPHRD_ARCNET) &&
3176 (dev->type != ARPHRD_INFINIBAND) &&
3177 (dev->type != ARPHRD_IEEE1394) &&
3178 (dev->type != ARPHRD_TUNNEL6) &&
3179 (dev->type != ARPHRD_6LOWPAN) &&
3180 (dev->type != ARPHRD_NONE)) {
3181 /* Alas, we support only Ethernet autoconfiguration. */
3182 return;
3185 idev = addrconf_add_dev(dev);
3186 if (IS_ERR(idev))
3187 return;
3189 /* this device type has no EUI support */
3190 if (dev->type == ARPHRD_NONE &&
3191 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3192 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3194 addrconf_addr_gen(idev, false);
3197 #if IS_ENABLED(CONFIG_IPV6_SIT)
3198 static void addrconf_sit_config(struct net_device *dev)
3200 struct inet6_dev *idev;
3202 ASSERT_RTNL();
3205 * Configure the tunnel with one of our IPv4
3206 * addresses... we should configure all of
3207 * our v4 addrs in the tunnel
3210 idev = ipv6_find_idev(dev);
3211 if (!idev) {
3212 pr_debug("%s: add_dev failed\n", __func__);
3213 return;
3216 if (dev->priv_flags & IFF_ISATAP) {
3217 addrconf_addr_gen(idev, false);
3218 return;
3221 sit_add_v4_addrs(idev);
3223 if (dev->flags&IFF_POINTOPOINT)
3224 addrconf_add_mroute(dev);
3226 #endif
3228 #if IS_ENABLED(CONFIG_NET_IPGRE)
3229 static void addrconf_gre_config(struct net_device *dev)
3231 struct inet6_dev *idev;
3233 ASSERT_RTNL();
3235 idev = ipv6_find_idev(dev);
3236 if (!idev) {
3237 pr_debug("%s: add_dev failed\n", __func__);
3238 return;
3241 addrconf_addr_gen(idev, true);
3242 if (dev->flags & IFF_POINTOPOINT)
3243 addrconf_add_mroute(dev);
3245 #endif
3247 static int fixup_permanent_addr(struct inet6_dev *idev,
3248 struct inet6_ifaddr *ifp)
3250 if (!ifp->rt) {
3251 struct rt6_info *rt;
3253 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3254 if (unlikely(IS_ERR(rt)))
3255 return PTR_ERR(rt);
3257 ifp->rt = rt;
3260 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3261 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3262 idev->dev, 0, 0);
3265 addrconf_dad_start(ifp);
3267 return 0;
3270 static void addrconf_permanent_addr(struct net_device *dev)
3272 struct inet6_ifaddr *ifp, *tmp;
3273 struct inet6_dev *idev;
3275 idev = __in6_dev_get(dev);
3276 if (!idev)
3277 return;
3279 write_lock_bh(&idev->lock);
3281 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3282 if ((ifp->flags & IFA_F_PERMANENT) &&
3283 fixup_permanent_addr(idev, ifp) < 0) {
3284 write_unlock_bh(&idev->lock);
3285 ipv6_del_addr(ifp);
3286 write_lock_bh(&idev->lock);
3288 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3289 idev->dev->name, &ifp->addr);
3293 write_unlock_bh(&idev->lock);
3296 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3297 void *ptr)
3299 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3300 struct netdev_notifier_changeupper_info *info;
3301 struct inet6_dev *idev = __in6_dev_get(dev);
3302 int run_pending = 0;
3303 int err;
3305 switch (event) {
3306 case NETDEV_REGISTER:
3307 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3308 idev = ipv6_add_dev(dev);
3309 if (IS_ERR(idev))
3310 return notifier_from_errno(PTR_ERR(idev));
3312 break;
3314 case NETDEV_CHANGEMTU:
3315 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3316 if (dev->mtu < IPV6_MIN_MTU) {
3317 addrconf_ifdown(dev, 1);
3318 break;
3321 if (idev) {
3322 rt6_mtu_change(dev, dev->mtu);
3323 idev->cnf.mtu6 = dev->mtu;
3324 break;
3327 /* allocate new idev */
3328 idev = ipv6_add_dev(dev);
3329 if (IS_ERR(idev))
3330 break;
3332 /* device is still not ready */
3333 if (!(idev->if_flags & IF_READY))
3334 break;
3336 run_pending = 1;
3338 /* fall through */
3340 case NETDEV_UP:
3341 case NETDEV_CHANGE:
3342 if (dev->flags & IFF_SLAVE)
3343 break;
3345 if (idev && idev->cnf.disable_ipv6)
3346 break;
3348 if (event == NETDEV_UP) {
3349 /* restore routes for permanent addresses */
3350 addrconf_permanent_addr(dev);
3352 if (!addrconf_qdisc_ok(dev)) {
3353 /* device is not ready yet. */
3354 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3355 dev->name);
3356 break;
3359 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3360 idev = ipv6_add_dev(dev);
3362 if (!IS_ERR_OR_NULL(idev)) {
3363 idev->if_flags |= IF_READY;
3364 run_pending = 1;
3366 } else if (event == NETDEV_CHANGE) {
3367 if (!addrconf_qdisc_ok(dev)) {
3368 /* device is still not ready. */
3369 break;
3372 if (idev) {
3373 if (idev->if_flags & IF_READY)
3374 /* device is already configured. */
3375 break;
3376 idev->if_flags |= IF_READY;
3379 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3380 dev->name);
3382 run_pending = 1;
3385 switch (dev->type) {
3386 #if IS_ENABLED(CONFIG_IPV6_SIT)
3387 case ARPHRD_SIT:
3388 addrconf_sit_config(dev);
3389 break;
3390 #endif
3391 #if IS_ENABLED(CONFIG_NET_IPGRE)
3392 case ARPHRD_IPGRE:
3393 addrconf_gre_config(dev);
3394 break;
3395 #endif
3396 case ARPHRD_LOOPBACK:
3397 init_loopback(dev);
3398 break;
3400 default:
3401 addrconf_dev_config(dev);
3402 break;
3405 if (!IS_ERR_OR_NULL(idev)) {
3406 if (run_pending)
3407 addrconf_dad_run(idev);
3410 * If the MTU changed during the interface down,
3411 * when the interface up, the changed MTU must be
3412 * reflected in the idev as well as routers.
3414 if (idev->cnf.mtu6 != dev->mtu &&
3415 dev->mtu >= IPV6_MIN_MTU) {
3416 rt6_mtu_change(dev, dev->mtu);
3417 idev->cnf.mtu6 = dev->mtu;
3419 idev->tstamp = jiffies;
3420 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3423 * If the changed mtu during down is lower than
3424 * IPV6_MIN_MTU stop IPv6 on this interface.
3426 if (dev->mtu < IPV6_MIN_MTU)
3427 addrconf_ifdown(dev, 1);
3429 break;
3431 case NETDEV_DOWN:
3432 case NETDEV_UNREGISTER:
3434 * Remove all addresses from this interface.
3436 addrconf_ifdown(dev, event != NETDEV_DOWN);
3437 break;
3439 case NETDEV_CHANGENAME:
3440 if (idev) {
3441 snmp6_unregister_dev(idev);
3442 addrconf_sysctl_unregister(idev);
3443 err = addrconf_sysctl_register(idev);
3444 if (err)
3445 return notifier_from_errno(err);
3446 err = snmp6_register_dev(idev);
3447 if (err) {
3448 addrconf_sysctl_unregister(idev);
3449 return notifier_from_errno(err);
3452 break;
3454 case NETDEV_PRE_TYPE_CHANGE:
3455 case NETDEV_POST_TYPE_CHANGE:
3456 if (idev)
3457 addrconf_type_change(dev, event);
3458 break;
3460 case NETDEV_CHANGEUPPER:
3461 info = ptr;
3463 /* flush all routes if dev is linked to or unlinked from
3464 * an L3 master device (e.g., VRF)
3466 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3467 addrconf_ifdown(dev, 0);
3470 return NOTIFY_OK;
3474 * addrconf module should be notified of a device going up
3476 static struct notifier_block ipv6_dev_notf = {
3477 .notifier_call = addrconf_notify,
3480 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3482 struct inet6_dev *idev;
3483 ASSERT_RTNL();
3485 idev = __in6_dev_get(dev);
3487 if (event == NETDEV_POST_TYPE_CHANGE)
3488 ipv6_mc_remap(idev);
3489 else if (event == NETDEV_PRE_TYPE_CHANGE)
3490 ipv6_mc_unmap(idev);
3493 static bool addr_is_local(const struct in6_addr *addr)
3495 return ipv6_addr_type(addr) &
3496 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3499 static int addrconf_ifdown(struct net_device *dev, int how)
3501 struct net *net = dev_net(dev);
3502 struct inet6_dev *idev;
3503 struct inet6_ifaddr *ifa, *tmp;
3504 struct list_head del_list;
3505 int _keep_addr;
3506 bool keep_addr;
3507 int state, i;
3509 ASSERT_RTNL();
3511 rt6_ifdown(net, dev);
3512 neigh_ifdown(&nd_tbl, dev);
3514 idev = __in6_dev_get(dev);
3515 if (!idev)
3516 return -ENODEV;
3519 * Step 1: remove reference to ipv6 device from parent device.
3520 * Do not dev_put!
3522 if (how) {
3523 idev->dead = 1;
3525 /* protected by rtnl_lock */
3526 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3528 /* Step 1.5: remove snmp6 entry */
3529 snmp6_unregister_dev(idev);
3533 /* aggregate the system setting and interface setting */
3534 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3535 if (!_keep_addr)
3536 _keep_addr = idev->cnf.keep_addr_on_down;
3538 /* combine the user config with event to determine if permanent
3539 * addresses are to be removed from address hash table
3541 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3543 /* Step 2: clear hash table */
3544 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3545 struct hlist_head *h = &inet6_addr_lst[i];
3547 spin_lock_bh(&addrconf_hash_lock);
3548 restart:
3549 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3550 if (ifa->idev == idev) {
3551 addrconf_del_dad_work(ifa);
3552 /* combined flag + permanent flag decide if
3553 * address is retained on a down event
3555 if (!keep_addr ||
3556 !(ifa->flags & IFA_F_PERMANENT) ||
3557 addr_is_local(&ifa->addr)) {
3558 hlist_del_init_rcu(&ifa->addr_lst);
3559 goto restart;
3563 spin_unlock_bh(&addrconf_hash_lock);
3566 write_lock_bh(&idev->lock);
3568 addrconf_del_rs_timer(idev);
3570 /* Step 2: clear flags for stateless addrconf */
3571 if (!how)
3572 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3574 if (how && del_timer(&idev->regen_timer))
3575 in6_dev_put(idev);
3577 /* Step 3: clear tempaddr list */
3578 while (!list_empty(&idev->tempaddr_list)) {
3579 ifa = list_first_entry(&idev->tempaddr_list,
3580 struct inet6_ifaddr, tmp_list);
3581 list_del(&ifa->tmp_list);
3582 write_unlock_bh(&idev->lock);
3583 spin_lock_bh(&ifa->lock);
3585 if (ifa->ifpub) {
3586 in6_ifa_put(ifa->ifpub);
3587 ifa->ifpub = NULL;
3589 spin_unlock_bh(&ifa->lock);
3590 in6_ifa_put(ifa);
3591 write_lock_bh(&idev->lock);
3594 /* re-combine the user config with event to determine if permanent
3595 * addresses are to be removed from the interface list
3597 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3599 INIT_LIST_HEAD(&del_list);
3600 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3601 struct rt6_info *rt = NULL;
3603 addrconf_del_dad_work(ifa);
3605 write_unlock_bh(&idev->lock);
3606 spin_lock_bh(&ifa->lock);
3608 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3609 !addr_is_local(&ifa->addr)) {
3610 /* set state to skip the notifier below */
3611 state = INET6_IFADDR_STATE_DEAD;
3612 ifa->state = 0;
3613 if (!(ifa->flags & IFA_F_NODAD))
3614 ifa->flags |= IFA_F_TENTATIVE;
3616 rt = ifa->rt;
3617 ifa->rt = NULL;
3618 } else {
3619 state = ifa->state;
3620 ifa->state = INET6_IFADDR_STATE_DEAD;
3622 list_move(&ifa->if_list, &del_list);
3625 spin_unlock_bh(&ifa->lock);
3627 if (rt)
3628 ip6_del_rt(rt);
3630 if (state != INET6_IFADDR_STATE_DEAD) {
3631 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3632 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3633 } else {
3634 if (idev->cnf.forwarding)
3635 addrconf_leave_anycast(ifa);
3636 addrconf_leave_solict(ifa->idev, &ifa->addr);
3639 write_lock_bh(&idev->lock);
3642 write_unlock_bh(&idev->lock);
3644 /* now clean up addresses to be removed */
3645 while (!list_empty(&del_list)) {
3646 ifa = list_first_entry(&del_list,
3647 struct inet6_ifaddr, if_list);
3648 list_del(&ifa->if_list);
3650 in6_ifa_put(ifa);
3653 /* Step 5: Discard anycast and multicast list */
3654 if (how) {
3655 ipv6_ac_destroy_dev(idev);
3656 ipv6_mc_destroy_dev(idev);
3657 } else {
3658 ipv6_mc_down(idev);
3661 idev->tstamp = jiffies;
3663 /* Last: Shot the device (if unregistered) */
3664 if (how) {
3665 addrconf_sysctl_unregister(idev);
3666 neigh_parms_release(&nd_tbl, idev->nd_parms);
3667 neigh_ifdown(&nd_tbl, dev);
3668 in6_dev_put(idev);
3670 return 0;
3673 static void addrconf_rs_timer(unsigned long data)
3675 struct inet6_dev *idev = (struct inet6_dev *)data;
3676 struct net_device *dev = idev->dev;
3677 struct in6_addr lladdr;
3679 write_lock(&idev->lock);
3680 if (idev->dead || !(idev->if_flags & IF_READY))
3681 goto out;
3683 if (!ipv6_accept_ra(idev))
3684 goto out;
3686 /* Announcement received after solicitation was sent */
3687 if (idev->if_flags & IF_RA_RCVD)
3688 goto out;
3690 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3691 write_unlock(&idev->lock);
3692 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3693 ndisc_send_rs(dev, &lladdr,
3694 &in6addr_linklocal_allrouters);
3695 else
3696 goto put;
3698 write_lock(&idev->lock);
3699 /* The wait after the last probe can be shorter */
3700 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3701 idev->cnf.rtr_solicits) ?
3702 idev->cnf.rtr_solicit_delay :
3703 idev->cnf.rtr_solicit_interval);
3704 } else {
3706 * Note: we do not support deprecated "all on-link"
3707 * assumption any longer.
3709 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3712 out:
3713 write_unlock(&idev->lock);
3714 put:
3715 in6_dev_put(idev);
3719 * Duplicate Address Detection
3721 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3723 unsigned long rand_num;
3724 struct inet6_dev *idev = ifp->idev;
3726 if (ifp->flags & IFA_F_OPTIMISTIC)
3727 rand_num = 0;
3728 else
3729 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3731 ifp->dad_probes = idev->cnf.dad_transmits;
3732 addrconf_mod_dad_work(ifp, rand_num);
3735 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3737 struct inet6_dev *idev = ifp->idev;
3738 struct net_device *dev = idev->dev;
3739 bool notify = false;
3741 addrconf_join_solict(dev, &ifp->addr);
3743 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3745 read_lock_bh(&idev->lock);
3746 spin_lock(&ifp->lock);
3747 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3748 goto out;
3750 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3751 idev->cnf.accept_dad < 1 ||
3752 !(ifp->flags&IFA_F_TENTATIVE) ||
3753 ifp->flags & IFA_F_NODAD) {
3754 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3755 spin_unlock(&ifp->lock);
3756 read_unlock_bh(&idev->lock);
3758 addrconf_dad_completed(ifp);
3759 return;
3762 if (!(idev->if_flags & IF_READY)) {
3763 spin_unlock(&ifp->lock);
3764 read_unlock_bh(&idev->lock);
3766 * If the device is not ready:
3767 * - keep it tentative if it is a permanent address.
3768 * - otherwise, kill it.
3770 in6_ifa_hold(ifp);
3771 addrconf_dad_stop(ifp, 0);
3772 return;
3776 * Optimistic nodes can start receiving
3777 * Frames right away
3779 if (ifp->flags & IFA_F_OPTIMISTIC) {
3780 ip6_ins_rt(ifp->rt);
3781 if (ipv6_use_optimistic_addr(idev)) {
3782 /* Because optimistic nodes can use this address,
3783 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3785 notify = true;
3789 addrconf_dad_kick(ifp);
3790 out:
3791 spin_unlock(&ifp->lock);
3792 read_unlock_bh(&idev->lock);
3793 if (notify)
3794 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3797 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3799 bool begin_dad = false;
3801 spin_lock_bh(&ifp->lock);
3802 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3803 ifp->state = INET6_IFADDR_STATE_PREDAD;
3804 begin_dad = true;
3806 spin_unlock_bh(&ifp->lock);
3808 if (begin_dad)
3809 addrconf_mod_dad_work(ifp, 0);
3812 static void addrconf_dad_work(struct work_struct *w)
3814 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3815 struct inet6_ifaddr,
3816 dad_work);
3817 struct inet6_dev *idev = ifp->idev;
3818 struct in6_addr mcaddr;
3819 bool disable_ipv6 = false;
3821 enum {
3822 DAD_PROCESS,
3823 DAD_BEGIN,
3824 DAD_ABORT,
3825 } action = DAD_PROCESS;
3827 rtnl_lock();
3829 spin_lock_bh(&ifp->lock);
3830 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3831 action = DAD_BEGIN;
3832 ifp->state = INET6_IFADDR_STATE_DAD;
3833 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3834 action = DAD_ABORT;
3835 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3837 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3838 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3839 struct in6_addr addr;
3841 addr.s6_addr32[0] = htonl(0xfe800000);
3842 addr.s6_addr32[1] = 0;
3844 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3845 ipv6_addr_equal(&ifp->addr, &addr)) {
3846 /* DAD failed for link-local based on MAC */
3847 idev->cnf.disable_ipv6 = 1;
3849 pr_info("%s: IPv6 being disabled!\n",
3850 ifp->idev->dev->name);
3851 disable_ipv6 = true;
3855 spin_unlock_bh(&ifp->lock);
3857 if (action == DAD_BEGIN) {
3858 addrconf_dad_begin(ifp);
3859 goto out;
3860 } else if (action == DAD_ABORT) {
3861 in6_ifa_hold(ifp);
3862 addrconf_dad_stop(ifp, 1);
3863 if (disable_ipv6)
3864 addrconf_ifdown(idev->dev, 0);
3865 goto out;
3868 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3869 goto out;
3871 write_lock_bh(&idev->lock);
3872 if (idev->dead || !(idev->if_flags & IF_READY)) {
3873 write_unlock_bh(&idev->lock);
3874 goto out;
3877 spin_lock(&ifp->lock);
3878 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3879 spin_unlock(&ifp->lock);
3880 write_unlock_bh(&idev->lock);
3881 goto out;
3884 if (ifp->dad_probes == 0) {
3886 * DAD was successful
3889 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3890 spin_unlock(&ifp->lock);
3891 write_unlock_bh(&idev->lock);
3893 addrconf_dad_completed(ifp);
3895 goto out;
3898 ifp->dad_probes--;
3899 addrconf_mod_dad_work(ifp,
3900 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3901 spin_unlock(&ifp->lock);
3902 write_unlock_bh(&idev->lock);
3904 /* send a neighbour solicitation for our addr */
3905 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3906 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3907 out:
3908 in6_ifa_put(ifp);
3909 rtnl_unlock();
3912 /* ifp->idev must be at least read locked */
3913 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3915 struct inet6_ifaddr *ifpiter;
3916 struct inet6_dev *idev = ifp->idev;
3918 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3919 if (ifpiter->scope > IFA_LINK)
3920 break;
3921 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3922 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3923 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3924 IFA_F_PERMANENT)
3925 return false;
3927 return true;
3930 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3932 struct net_device *dev = ifp->idev->dev;
3933 struct in6_addr lladdr;
3934 bool send_rs, send_mld;
3936 addrconf_del_dad_work(ifp);
3939 * Configure the address for reception. Now it is valid.
3942 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3944 /* If added prefix is link local and we are prepared to process
3945 router advertisements, start sending router solicitations.
3948 read_lock_bh(&ifp->idev->lock);
3949 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3950 send_rs = send_mld &&
3951 ipv6_accept_ra(ifp->idev) &&
3952 ifp->idev->cnf.rtr_solicits > 0 &&
3953 (dev->flags&IFF_LOOPBACK) == 0;
3954 read_unlock_bh(&ifp->idev->lock);
3956 /* While dad is in progress mld report's source address is in6_addrany.
3957 * Resend with proper ll now.
3959 if (send_mld)
3960 ipv6_mc_dad_complete(ifp->idev);
3962 if (send_rs) {
3964 * If a host as already performed a random delay
3965 * [...] as part of DAD [...] there is no need
3966 * to delay again before sending the first RS
3968 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3969 return;
3970 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3972 write_lock_bh(&ifp->idev->lock);
3973 spin_lock(&ifp->lock);
3974 ifp->idev->rs_probes = 1;
3975 ifp->idev->if_flags |= IF_RS_SENT;
3976 addrconf_mod_rs_timer(ifp->idev,
3977 ifp->idev->cnf.rtr_solicit_interval);
3978 spin_unlock(&ifp->lock);
3979 write_unlock_bh(&ifp->idev->lock);
3983 static void addrconf_dad_run(struct inet6_dev *idev)
3985 struct inet6_ifaddr *ifp;
3987 read_lock_bh(&idev->lock);
3988 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3989 spin_lock(&ifp->lock);
3990 if (ifp->flags & IFA_F_TENTATIVE &&
3991 ifp->state == INET6_IFADDR_STATE_DAD)
3992 addrconf_dad_kick(ifp);
3993 spin_unlock(&ifp->lock);
3995 read_unlock_bh(&idev->lock);
3998 #ifdef CONFIG_PROC_FS
3999 struct if6_iter_state {
4000 struct seq_net_private p;
4001 int bucket;
4002 int offset;
4005 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4007 struct inet6_ifaddr *ifa = NULL;
4008 struct if6_iter_state *state = seq->private;
4009 struct net *net = seq_file_net(seq);
4010 int p = 0;
4012 /* initial bucket if pos is 0 */
4013 if (pos == 0) {
4014 state->bucket = 0;
4015 state->offset = 0;
4018 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4019 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4020 addr_lst) {
4021 if (!net_eq(dev_net(ifa->idev->dev), net))
4022 continue;
4023 /* sync with offset */
4024 if (p < state->offset) {
4025 p++;
4026 continue;
4028 state->offset++;
4029 return ifa;
4032 /* prepare for next bucket */
4033 state->offset = 0;
4034 p = 0;
4036 return NULL;
4039 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4040 struct inet6_ifaddr *ifa)
4042 struct if6_iter_state *state = seq->private;
4043 struct net *net = seq_file_net(seq);
4045 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4046 if (!net_eq(dev_net(ifa->idev->dev), net))
4047 continue;
4048 state->offset++;
4049 return ifa;
4052 while (++state->bucket < IN6_ADDR_HSIZE) {
4053 state->offset = 0;
4054 hlist_for_each_entry_rcu_bh(ifa,
4055 &inet6_addr_lst[state->bucket], addr_lst) {
4056 if (!net_eq(dev_net(ifa->idev->dev), net))
4057 continue;
4058 state->offset++;
4059 return ifa;
4063 return NULL;
4066 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4067 __acquires(rcu_bh)
4069 rcu_read_lock_bh();
4070 return if6_get_first(seq, *pos);
4073 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4075 struct inet6_ifaddr *ifa;
4077 ifa = if6_get_next(seq, v);
4078 ++*pos;
4079 return ifa;
4082 static void if6_seq_stop(struct seq_file *seq, void *v)
4083 __releases(rcu_bh)
4085 rcu_read_unlock_bh();
4088 static int if6_seq_show(struct seq_file *seq, void *v)
4090 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4091 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4092 &ifp->addr,
4093 ifp->idev->dev->ifindex,
4094 ifp->prefix_len,
4095 ifp->scope,
4096 (u8) ifp->flags,
4097 ifp->idev->dev->name);
4098 return 0;
4101 static const struct seq_operations if6_seq_ops = {
4102 .start = if6_seq_start,
4103 .next = if6_seq_next,
4104 .show = if6_seq_show,
4105 .stop = if6_seq_stop,
4108 static int if6_seq_open(struct inode *inode, struct file *file)
4110 return seq_open_net(inode, file, &if6_seq_ops,
4111 sizeof(struct if6_iter_state));
4114 static const struct file_operations if6_fops = {
4115 .owner = THIS_MODULE,
4116 .open = if6_seq_open,
4117 .read = seq_read,
4118 .llseek = seq_lseek,
4119 .release = seq_release_net,
4122 static int __net_init if6_proc_net_init(struct net *net)
4124 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4125 return -ENOMEM;
4126 return 0;
4129 static void __net_exit if6_proc_net_exit(struct net *net)
4131 remove_proc_entry("if_inet6", net->proc_net);
4134 static struct pernet_operations if6_proc_net_ops = {
4135 .init = if6_proc_net_init,
4136 .exit = if6_proc_net_exit,
4139 int __init if6_proc_init(void)
4141 return register_pernet_subsys(&if6_proc_net_ops);
4144 void if6_proc_exit(void)
4146 unregister_pernet_subsys(&if6_proc_net_ops);
4148 #endif /* CONFIG_PROC_FS */
4150 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4151 /* Check if address is a home address configured on any interface. */
4152 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4154 int ret = 0;
4155 struct inet6_ifaddr *ifp = NULL;
4156 unsigned int hash = inet6_addr_hash(addr);
4158 rcu_read_lock_bh();
4159 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4160 if (!net_eq(dev_net(ifp->idev->dev), net))
4161 continue;
4162 if (ipv6_addr_equal(&ifp->addr, addr) &&
4163 (ifp->flags & IFA_F_HOMEADDRESS)) {
4164 ret = 1;
4165 break;
4168 rcu_read_unlock_bh();
4169 return ret;
4171 #endif
4174 * Periodic address status verification
4177 static void addrconf_verify_rtnl(void)
4179 unsigned long now, next, next_sec, next_sched;
4180 struct inet6_ifaddr *ifp;
4181 int i;
4183 ASSERT_RTNL();
4185 rcu_read_lock_bh();
4186 now = jiffies;
4187 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4189 cancel_delayed_work(&addr_chk_work);
4191 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4192 restart:
4193 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4194 unsigned long age;
4196 /* When setting preferred_lft to a value not zero or
4197 * infinity, while valid_lft is infinity
4198 * IFA_F_PERMANENT has a non-infinity life time.
4200 if ((ifp->flags & IFA_F_PERMANENT) &&
4201 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4202 continue;
4204 spin_lock(&ifp->lock);
4205 /* We try to batch several events at once. */
4206 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4208 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4209 age >= ifp->valid_lft) {
4210 spin_unlock(&ifp->lock);
4211 in6_ifa_hold(ifp);
4212 ipv6_del_addr(ifp);
4213 goto restart;
4214 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4215 spin_unlock(&ifp->lock);
4216 continue;
4217 } else if (age >= ifp->prefered_lft) {
4218 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4219 int deprecate = 0;
4221 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4222 deprecate = 1;
4223 ifp->flags |= IFA_F_DEPRECATED;
4226 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4227 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4228 next = ifp->tstamp + ifp->valid_lft * HZ;
4230 spin_unlock(&ifp->lock);
4232 if (deprecate) {
4233 in6_ifa_hold(ifp);
4235 ipv6_ifa_notify(0, ifp);
4236 in6_ifa_put(ifp);
4237 goto restart;
4239 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4240 !(ifp->flags&IFA_F_TENTATIVE)) {
4241 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4242 ifp->idev->cnf.dad_transmits *
4243 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4245 if (age >= ifp->prefered_lft - regen_advance) {
4246 struct inet6_ifaddr *ifpub = ifp->ifpub;
4247 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4248 next = ifp->tstamp + ifp->prefered_lft * HZ;
4249 if (!ifp->regen_count && ifpub) {
4250 ifp->regen_count++;
4251 in6_ifa_hold(ifp);
4252 in6_ifa_hold(ifpub);
4253 spin_unlock(&ifp->lock);
4255 spin_lock(&ifpub->lock);
4256 ifpub->regen_count = 0;
4257 spin_unlock(&ifpub->lock);
4258 ipv6_create_tempaddr(ifpub, ifp);
4259 in6_ifa_put(ifpub);
4260 in6_ifa_put(ifp);
4261 goto restart;
4263 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4264 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4265 spin_unlock(&ifp->lock);
4266 } else {
4267 /* ifp->prefered_lft <= ifp->valid_lft */
4268 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4269 next = ifp->tstamp + ifp->prefered_lft * HZ;
4270 spin_unlock(&ifp->lock);
4275 next_sec = round_jiffies_up(next);
4276 next_sched = next;
4278 /* If rounded timeout is accurate enough, accept it. */
4279 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4280 next_sched = next_sec;
4282 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4283 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4284 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4286 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4287 now, next, next_sec, next_sched);
4288 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4289 rcu_read_unlock_bh();
4292 static void addrconf_verify_work(struct work_struct *w)
4294 rtnl_lock();
4295 addrconf_verify_rtnl();
4296 rtnl_unlock();
4299 static void addrconf_verify(void)
4301 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4304 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4305 struct in6_addr **peer_pfx)
4307 struct in6_addr *pfx = NULL;
4309 *peer_pfx = NULL;
4311 if (addr)
4312 pfx = nla_data(addr);
4314 if (local) {
4315 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4316 *peer_pfx = pfx;
4317 pfx = nla_data(local);
4320 return pfx;
4323 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4324 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4325 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4326 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4327 [IFA_FLAGS] = { .len = sizeof(u32) },
4330 static int
4331 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4333 struct net *net = sock_net(skb->sk);
4334 struct ifaddrmsg *ifm;
4335 struct nlattr *tb[IFA_MAX+1];
4336 struct in6_addr *pfx, *peer_pfx;
4337 u32 ifa_flags;
4338 int err;
4340 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4341 if (err < 0)
4342 return err;
4344 ifm = nlmsg_data(nlh);
4345 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4346 if (!pfx)
4347 return -EINVAL;
4349 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4351 /* We ignore other flags so far. */
4352 ifa_flags &= IFA_F_MANAGETEMPADDR;
4354 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4355 ifm->ifa_prefixlen);
4358 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4359 u32 prefered_lft, u32 valid_lft)
4361 u32 flags;
4362 clock_t expires;
4363 unsigned long timeout;
4364 bool was_managetempaddr;
4365 bool had_prefixroute;
4367 ASSERT_RTNL();
4369 if (!valid_lft || (prefered_lft > valid_lft))
4370 return -EINVAL;
4372 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4373 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4374 return -EINVAL;
4376 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4377 if (addrconf_finite_timeout(timeout)) {
4378 expires = jiffies_to_clock_t(timeout * HZ);
4379 valid_lft = timeout;
4380 flags = RTF_EXPIRES;
4381 } else {
4382 expires = 0;
4383 flags = 0;
4384 ifa_flags |= IFA_F_PERMANENT;
4387 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4388 if (addrconf_finite_timeout(timeout)) {
4389 if (timeout == 0)
4390 ifa_flags |= IFA_F_DEPRECATED;
4391 prefered_lft = timeout;
4394 spin_lock_bh(&ifp->lock);
4395 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4396 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4397 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4398 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4399 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4400 IFA_F_NOPREFIXROUTE);
4401 ifp->flags |= ifa_flags;
4402 ifp->tstamp = jiffies;
4403 ifp->valid_lft = valid_lft;
4404 ifp->prefered_lft = prefered_lft;
4406 spin_unlock_bh(&ifp->lock);
4407 if (!(ifp->flags&IFA_F_TENTATIVE))
4408 ipv6_ifa_notify(0, ifp);
4410 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4411 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4412 expires, flags);
4413 } else if (had_prefixroute) {
4414 enum cleanup_prefix_rt_t action;
4415 unsigned long rt_expires;
4417 write_lock_bh(&ifp->idev->lock);
4418 action = check_cleanup_prefix_route(ifp, &rt_expires);
4419 write_unlock_bh(&ifp->idev->lock);
4421 if (action != CLEANUP_PREFIX_RT_NOP) {
4422 cleanup_prefix_route(ifp, rt_expires,
4423 action == CLEANUP_PREFIX_RT_DEL);
4427 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4428 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4429 valid_lft = prefered_lft = 0;
4430 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4431 !was_managetempaddr, jiffies);
4434 addrconf_verify_rtnl();
4436 return 0;
4439 static int
4440 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4442 struct net *net = sock_net(skb->sk);
4443 struct ifaddrmsg *ifm;
4444 struct nlattr *tb[IFA_MAX+1];
4445 struct in6_addr *pfx, *peer_pfx;
4446 struct inet6_ifaddr *ifa;
4447 struct net_device *dev;
4448 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4449 u32 ifa_flags;
4450 int err;
4452 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4453 if (err < 0)
4454 return err;
4456 ifm = nlmsg_data(nlh);
4457 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4458 if (!pfx)
4459 return -EINVAL;
4461 if (tb[IFA_CACHEINFO]) {
4462 struct ifa_cacheinfo *ci;
4464 ci = nla_data(tb[IFA_CACHEINFO]);
4465 valid_lft = ci->ifa_valid;
4466 preferred_lft = ci->ifa_prefered;
4467 } else {
4468 preferred_lft = INFINITY_LIFE_TIME;
4469 valid_lft = INFINITY_LIFE_TIME;
4472 dev = __dev_get_by_index(net, ifm->ifa_index);
4473 if (!dev)
4474 return -ENODEV;
4476 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4478 /* We ignore other flags so far. */
4479 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4480 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4482 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4483 if (!ifa) {
4485 * It would be best to check for !NLM_F_CREATE here but
4486 * userspace already relies on not having to provide this.
4488 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4489 ifm->ifa_prefixlen, ifa_flags,
4490 preferred_lft, valid_lft);
4493 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4494 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4495 err = -EEXIST;
4496 else
4497 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4499 in6_ifa_put(ifa);
4501 return err;
4504 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4505 u8 scope, int ifindex)
4507 struct ifaddrmsg *ifm;
4509 ifm = nlmsg_data(nlh);
4510 ifm->ifa_family = AF_INET6;
4511 ifm->ifa_prefixlen = prefixlen;
4512 ifm->ifa_flags = flags;
4513 ifm->ifa_scope = scope;
4514 ifm->ifa_index = ifindex;
4517 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4518 unsigned long tstamp, u32 preferred, u32 valid)
4520 struct ifa_cacheinfo ci;
4522 ci.cstamp = cstamp_delta(cstamp);
4523 ci.tstamp = cstamp_delta(tstamp);
4524 ci.ifa_prefered = preferred;
4525 ci.ifa_valid = valid;
4527 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4530 static inline int rt_scope(int ifa_scope)
4532 if (ifa_scope & IFA_HOST)
4533 return RT_SCOPE_HOST;
4534 else if (ifa_scope & IFA_LINK)
4535 return RT_SCOPE_LINK;
4536 else if (ifa_scope & IFA_SITE)
4537 return RT_SCOPE_SITE;
4538 else
4539 return RT_SCOPE_UNIVERSE;
4542 static inline int inet6_ifaddr_msgsize(void)
4544 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4545 + nla_total_size(16) /* IFA_LOCAL */
4546 + nla_total_size(16) /* IFA_ADDRESS */
4547 + nla_total_size(sizeof(struct ifa_cacheinfo))
4548 + nla_total_size(4) /* IFA_FLAGS */;
4551 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4552 u32 portid, u32 seq, int event, unsigned int flags)
4554 struct nlmsghdr *nlh;
4555 u32 preferred, valid;
4557 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4558 if (!nlh)
4559 return -EMSGSIZE;
4561 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4562 ifa->idev->dev->ifindex);
4564 if (!((ifa->flags&IFA_F_PERMANENT) &&
4565 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4566 preferred = ifa->prefered_lft;
4567 valid = ifa->valid_lft;
4568 if (preferred != INFINITY_LIFE_TIME) {
4569 long tval = (jiffies - ifa->tstamp)/HZ;
4570 if (preferred > tval)
4571 preferred -= tval;
4572 else
4573 preferred = 0;
4574 if (valid != INFINITY_LIFE_TIME) {
4575 if (valid > tval)
4576 valid -= tval;
4577 else
4578 valid = 0;
4581 } else {
4582 preferred = INFINITY_LIFE_TIME;
4583 valid = INFINITY_LIFE_TIME;
4586 if (!ipv6_addr_any(&ifa->peer_addr)) {
4587 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4588 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4589 goto error;
4590 } else
4591 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4592 goto error;
4594 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4595 goto error;
4597 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4598 goto error;
4600 nlmsg_end(skb, nlh);
4601 return 0;
4603 error:
4604 nlmsg_cancel(skb, nlh);
4605 return -EMSGSIZE;
4608 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4609 u32 portid, u32 seq, int event, u16 flags)
4611 struct nlmsghdr *nlh;
4612 u8 scope = RT_SCOPE_UNIVERSE;
4613 int ifindex = ifmca->idev->dev->ifindex;
4615 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4616 scope = RT_SCOPE_SITE;
4618 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4619 if (!nlh)
4620 return -EMSGSIZE;
4622 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4623 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4624 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4625 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4626 nlmsg_cancel(skb, nlh);
4627 return -EMSGSIZE;
4630 nlmsg_end(skb, nlh);
4631 return 0;
4634 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4635 u32 portid, u32 seq, int event, unsigned int flags)
4637 struct nlmsghdr *nlh;
4638 u8 scope = RT_SCOPE_UNIVERSE;
4639 int ifindex = ifaca->aca_idev->dev->ifindex;
4641 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4642 scope = RT_SCOPE_SITE;
4644 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4645 if (!nlh)
4646 return -EMSGSIZE;
4648 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4649 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4650 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4651 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4652 nlmsg_cancel(skb, nlh);
4653 return -EMSGSIZE;
4656 nlmsg_end(skb, nlh);
4657 return 0;
4660 enum addr_type_t {
4661 UNICAST_ADDR,
4662 MULTICAST_ADDR,
4663 ANYCAST_ADDR,
4666 /* called with rcu_read_lock() */
4667 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4668 struct netlink_callback *cb, enum addr_type_t type,
4669 int s_ip_idx, int *p_ip_idx)
4671 struct ifmcaddr6 *ifmca;
4672 struct ifacaddr6 *ifaca;
4673 int err = 1;
4674 int ip_idx = *p_ip_idx;
4676 read_lock_bh(&idev->lock);
4677 switch (type) {
4678 case UNICAST_ADDR: {
4679 struct inet6_ifaddr *ifa;
4681 /* unicast address incl. temp addr */
4682 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4683 if (++ip_idx < s_ip_idx)
4684 continue;
4685 err = inet6_fill_ifaddr(skb, ifa,
4686 NETLINK_CB(cb->skb).portid,
4687 cb->nlh->nlmsg_seq,
4688 RTM_NEWADDR,
4689 NLM_F_MULTI);
4690 if (err < 0)
4691 break;
4692 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4694 break;
4696 case MULTICAST_ADDR:
4697 /* multicast address */
4698 for (ifmca = idev->mc_list; ifmca;
4699 ifmca = ifmca->next, ip_idx++) {
4700 if (ip_idx < s_ip_idx)
4701 continue;
4702 err = inet6_fill_ifmcaddr(skb, ifmca,
4703 NETLINK_CB(cb->skb).portid,
4704 cb->nlh->nlmsg_seq,
4705 RTM_GETMULTICAST,
4706 NLM_F_MULTI);
4707 if (err < 0)
4708 break;
4710 break;
4711 case ANYCAST_ADDR:
4712 /* anycast address */
4713 for (ifaca = idev->ac_list; ifaca;
4714 ifaca = ifaca->aca_next, ip_idx++) {
4715 if (ip_idx < s_ip_idx)
4716 continue;
4717 err = inet6_fill_ifacaddr(skb, ifaca,
4718 NETLINK_CB(cb->skb).portid,
4719 cb->nlh->nlmsg_seq,
4720 RTM_GETANYCAST,
4721 NLM_F_MULTI);
4722 if (err < 0)
4723 break;
4725 break;
4726 default:
4727 break;
4729 read_unlock_bh(&idev->lock);
4730 *p_ip_idx = ip_idx;
4731 return err;
4734 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4735 enum addr_type_t type)
4737 struct net *net = sock_net(skb->sk);
4738 int h, s_h;
4739 int idx, ip_idx;
4740 int s_idx, s_ip_idx;
4741 struct net_device *dev;
4742 struct inet6_dev *idev;
4743 struct hlist_head *head;
4745 s_h = cb->args[0];
4746 s_idx = idx = cb->args[1];
4747 s_ip_idx = ip_idx = cb->args[2];
4749 rcu_read_lock();
4750 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4751 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4752 idx = 0;
4753 head = &net->dev_index_head[h];
4754 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4755 if (idx < s_idx)
4756 goto cont;
4757 if (h > s_h || idx > s_idx)
4758 s_ip_idx = 0;
4759 ip_idx = 0;
4760 idev = __in6_dev_get(dev);
4761 if (!idev)
4762 goto cont;
4764 if (in6_dump_addrs(idev, skb, cb, type,
4765 s_ip_idx, &ip_idx) < 0)
4766 goto done;
4767 cont:
4768 idx++;
4771 done:
4772 rcu_read_unlock();
4773 cb->args[0] = h;
4774 cb->args[1] = idx;
4775 cb->args[2] = ip_idx;
4777 return skb->len;
4780 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4782 enum addr_type_t type = UNICAST_ADDR;
4784 return inet6_dump_addr(skb, cb, type);
4787 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4789 enum addr_type_t type = MULTICAST_ADDR;
4791 return inet6_dump_addr(skb, cb, type);
4795 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4797 enum addr_type_t type = ANYCAST_ADDR;
4799 return inet6_dump_addr(skb, cb, type);
4802 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4804 struct net *net = sock_net(in_skb->sk);
4805 struct ifaddrmsg *ifm;
4806 struct nlattr *tb[IFA_MAX+1];
4807 struct in6_addr *addr = NULL, *peer;
4808 struct net_device *dev = NULL;
4809 struct inet6_ifaddr *ifa;
4810 struct sk_buff *skb;
4811 int err;
4813 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4814 if (err < 0)
4815 goto errout;
4817 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4818 if (!addr) {
4819 err = -EINVAL;
4820 goto errout;
4823 ifm = nlmsg_data(nlh);
4824 if (ifm->ifa_index)
4825 dev = __dev_get_by_index(net, ifm->ifa_index);
4827 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4828 if (!ifa) {
4829 err = -EADDRNOTAVAIL;
4830 goto errout;
4833 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4834 if (!skb) {
4835 err = -ENOBUFS;
4836 goto errout_ifa;
4839 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4840 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4841 if (err < 0) {
4842 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4843 WARN_ON(err == -EMSGSIZE);
4844 kfree_skb(skb);
4845 goto errout_ifa;
4847 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4848 errout_ifa:
4849 in6_ifa_put(ifa);
4850 errout:
4851 return err;
4854 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4856 struct sk_buff *skb;
4857 struct net *net = dev_net(ifa->idev->dev);
4858 int err = -ENOBUFS;
4860 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4861 if (!skb)
4862 goto errout;
4864 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4865 if (err < 0) {
4866 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4867 WARN_ON(err == -EMSGSIZE);
4868 kfree_skb(skb);
4869 goto errout;
4871 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4872 return;
4873 errout:
4874 if (err < 0)
4875 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4878 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4879 __s32 *array, int bytes)
4881 BUG_ON(bytes < (DEVCONF_MAX * 4));
4883 memset(array, 0, bytes);
4884 array[DEVCONF_FORWARDING] = cnf->forwarding;
4885 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4886 array[DEVCONF_MTU6] = cnf->mtu6;
4887 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4888 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4889 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4890 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4891 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4892 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4893 jiffies_to_msecs(cnf->rtr_solicit_interval);
4894 array[DEVCONF_RTR_SOLICIT_DELAY] =
4895 jiffies_to_msecs(cnf->rtr_solicit_delay);
4896 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4897 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4898 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4899 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4900 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4901 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4902 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4903 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4904 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4905 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4906 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4907 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4908 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4909 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4910 #ifdef CONFIG_IPV6_ROUTER_PREF
4911 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4912 array[DEVCONF_RTR_PROBE_INTERVAL] =
4913 jiffies_to_msecs(cnf->rtr_probe_interval);
4914 #ifdef CONFIG_IPV6_ROUTE_INFO
4915 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4916 #endif
4917 #endif
4918 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4919 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4920 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4921 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4922 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4923 #endif
4924 #ifdef CONFIG_IPV6_MROUTE
4925 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4926 #endif
4927 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4928 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4929 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4930 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4931 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4932 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4933 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4934 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4935 /* we omit DEVCONF_STABLE_SECRET for now */
4936 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4937 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4938 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4939 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4942 static inline size_t inet6_ifla6_size(void)
4944 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4945 + nla_total_size(sizeof(struct ifla_cacheinfo))
4946 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4947 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4948 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4949 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4952 static inline size_t inet6_if_nlmsg_size(void)
4954 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4955 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4956 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4957 + nla_total_size(4) /* IFLA_MTU */
4958 + nla_total_size(4) /* IFLA_LINK */
4959 + nla_total_size(1) /* IFLA_OPERSTATE */
4960 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4963 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4964 int items, int bytes)
4966 int i;
4967 int pad = bytes - sizeof(u64) * items;
4968 BUG_ON(pad < 0);
4970 /* Use put_unaligned() because stats may not be aligned for u64. */
4971 put_unaligned(items, &stats[0]);
4972 for (i = 1; i < items; i++)
4973 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4975 memset(&stats[items], 0, pad);
4978 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4979 int bytes, size_t syncpoff)
4981 int i, c;
4982 u64 buff[IPSTATS_MIB_MAX];
4983 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4985 BUG_ON(pad < 0);
4987 memset(buff, 0, sizeof(buff));
4988 buff[0] = IPSTATS_MIB_MAX;
4990 for_each_possible_cpu(c) {
4991 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4992 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4995 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4996 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4999 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5000 int bytes)
5002 switch (attrtype) {
5003 case IFLA_INET6_STATS:
5004 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5005 offsetof(struct ipstats_mib, syncp));
5006 break;
5007 case IFLA_INET6_ICMP6STATS:
5008 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
5009 break;
5013 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5014 u32 ext_filter_mask)
5016 struct nlattr *nla;
5017 struct ifla_cacheinfo ci;
5019 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5020 goto nla_put_failure;
5021 ci.max_reasm_len = IPV6_MAXPLEN;
5022 ci.tstamp = cstamp_delta(idev->tstamp);
5023 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5024 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5025 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5026 goto nla_put_failure;
5027 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5028 if (!nla)
5029 goto nla_put_failure;
5030 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5032 /* XXX - MC not implemented */
5034 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5035 return 0;
5037 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5038 if (!nla)
5039 goto nla_put_failure;
5040 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5042 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5043 if (!nla)
5044 goto nla_put_failure;
5045 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5047 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5048 if (!nla)
5049 goto nla_put_failure;
5051 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5052 goto nla_put_failure;
5054 read_lock_bh(&idev->lock);
5055 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5056 read_unlock_bh(&idev->lock);
5058 return 0;
5060 nla_put_failure:
5061 return -EMSGSIZE;
5064 static size_t inet6_get_link_af_size(const struct net_device *dev,
5065 u32 ext_filter_mask)
5067 if (!__in6_dev_get(dev))
5068 return 0;
5070 return inet6_ifla6_size();
5073 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5074 u32 ext_filter_mask)
5076 struct inet6_dev *idev = __in6_dev_get(dev);
5078 if (!idev)
5079 return -ENODATA;
5081 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5082 return -EMSGSIZE;
5084 return 0;
5087 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5089 struct inet6_ifaddr *ifp;
5090 struct net_device *dev = idev->dev;
5091 bool clear_token, update_rs = false;
5092 struct in6_addr ll_addr;
5094 ASSERT_RTNL();
5096 if (!token)
5097 return -EINVAL;
5098 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5099 return -EINVAL;
5100 if (!ipv6_accept_ra(idev))
5101 return -EINVAL;
5102 if (idev->cnf.rtr_solicits <= 0)
5103 return -EINVAL;
5105 write_lock_bh(&idev->lock);
5107 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5108 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5110 write_unlock_bh(&idev->lock);
5112 clear_token = ipv6_addr_any(token);
5113 if (clear_token)
5114 goto update_lft;
5116 if (!idev->dead && (idev->if_flags & IF_READY) &&
5117 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5118 IFA_F_OPTIMISTIC)) {
5119 /* If we're not ready, then normal ifup will take care
5120 * of this. Otherwise, we need to request our rs here.
5122 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5123 update_rs = true;
5126 update_lft:
5127 write_lock_bh(&idev->lock);
5129 if (update_rs) {
5130 idev->if_flags |= IF_RS_SENT;
5131 idev->rs_probes = 1;
5132 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
5135 /* Well, that's kinda nasty ... */
5136 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5137 spin_lock(&ifp->lock);
5138 if (ifp->tokenized) {
5139 ifp->valid_lft = 0;
5140 ifp->prefered_lft = 0;
5142 spin_unlock(&ifp->lock);
5145 write_unlock_bh(&idev->lock);
5146 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5147 addrconf_verify_rtnl();
5148 return 0;
5151 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5152 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5153 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5156 static int inet6_validate_link_af(const struct net_device *dev,
5157 const struct nlattr *nla)
5159 struct nlattr *tb[IFLA_INET6_MAX + 1];
5161 if (dev && !__in6_dev_get(dev))
5162 return -EAFNOSUPPORT;
5164 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5167 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5169 int err = -EINVAL;
5170 struct inet6_dev *idev = __in6_dev_get(dev);
5171 struct nlattr *tb[IFLA_INET6_MAX + 1];
5173 if (!idev)
5174 return -EAFNOSUPPORT;
5176 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5177 BUG();
5179 if (tb[IFLA_INET6_TOKEN]) {
5180 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5181 if (err)
5182 return err;
5185 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5186 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5188 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5189 mode != IN6_ADDR_GEN_MODE_NONE &&
5190 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5191 mode != IN6_ADDR_GEN_MODE_RANDOM)
5192 return -EINVAL;
5194 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5195 !idev->cnf.stable_secret.initialized &&
5196 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5197 return -EINVAL;
5199 idev->addr_gen_mode = mode;
5200 err = 0;
5203 return err;
5206 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5207 u32 portid, u32 seq, int event, unsigned int flags)
5209 struct net_device *dev = idev->dev;
5210 struct ifinfomsg *hdr;
5211 struct nlmsghdr *nlh;
5212 void *protoinfo;
5214 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5215 if (!nlh)
5216 return -EMSGSIZE;
5218 hdr = nlmsg_data(nlh);
5219 hdr->ifi_family = AF_INET6;
5220 hdr->__ifi_pad = 0;
5221 hdr->ifi_type = dev->type;
5222 hdr->ifi_index = dev->ifindex;
5223 hdr->ifi_flags = dev_get_flags(dev);
5224 hdr->ifi_change = 0;
5226 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5227 (dev->addr_len &&
5228 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5229 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5230 (dev->ifindex != dev_get_iflink(dev) &&
5231 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5232 nla_put_u8(skb, IFLA_OPERSTATE,
5233 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5234 goto nla_put_failure;
5235 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5236 if (!protoinfo)
5237 goto nla_put_failure;
5239 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5240 goto nla_put_failure;
5242 nla_nest_end(skb, protoinfo);
5243 nlmsg_end(skb, nlh);
5244 return 0;
5246 nla_put_failure:
5247 nlmsg_cancel(skb, nlh);
5248 return -EMSGSIZE;
5251 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5253 struct net *net = sock_net(skb->sk);
5254 int h, s_h;
5255 int idx = 0, s_idx;
5256 struct net_device *dev;
5257 struct inet6_dev *idev;
5258 struct hlist_head *head;
5260 s_h = cb->args[0];
5261 s_idx = cb->args[1];
5263 rcu_read_lock();
5264 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5265 idx = 0;
5266 head = &net->dev_index_head[h];
5267 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5268 if (idx < s_idx)
5269 goto cont;
5270 idev = __in6_dev_get(dev);
5271 if (!idev)
5272 goto cont;
5273 if (inet6_fill_ifinfo(skb, idev,
5274 NETLINK_CB(cb->skb).portid,
5275 cb->nlh->nlmsg_seq,
5276 RTM_NEWLINK, NLM_F_MULTI) < 0)
5277 goto out;
5278 cont:
5279 idx++;
5282 out:
5283 rcu_read_unlock();
5284 cb->args[1] = idx;
5285 cb->args[0] = h;
5287 return skb->len;
5290 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5292 struct sk_buff *skb;
5293 struct net *net = dev_net(idev->dev);
5294 int err = -ENOBUFS;
5296 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5297 if (!skb)
5298 goto errout;
5300 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5301 if (err < 0) {
5302 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5303 WARN_ON(err == -EMSGSIZE);
5304 kfree_skb(skb);
5305 goto errout;
5307 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5308 return;
5309 errout:
5310 if (err < 0)
5311 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5314 static inline size_t inet6_prefix_nlmsg_size(void)
5316 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5317 + nla_total_size(sizeof(struct in6_addr))
5318 + nla_total_size(sizeof(struct prefix_cacheinfo));
5321 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5322 struct prefix_info *pinfo, u32 portid, u32 seq,
5323 int event, unsigned int flags)
5325 struct prefixmsg *pmsg;
5326 struct nlmsghdr *nlh;
5327 struct prefix_cacheinfo ci;
5329 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5330 if (!nlh)
5331 return -EMSGSIZE;
5333 pmsg = nlmsg_data(nlh);
5334 pmsg->prefix_family = AF_INET6;
5335 pmsg->prefix_pad1 = 0;
5336 pmsg->prefix_pad2 = 0;
5337 pmsg->prefix_ifindex = idev->dev->ifindex;
5338 pmsg->prefix_len = pinfo->prefix_len;
5339 pmsg->prefix_type = pinfo->type;
5340 pmsg->prefix_pad3 = 0;
5341 pmsg->prefix_flags = 0;
5342 if (pinfo->onlink)
5343 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5344 if (pinfo->autoconf)
5345 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5347 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5348 goto nla_put_failure;
5349 ci.preferred_time = ntohl(pinfo->prefered);
5350 ci.valid_time = ntohl(pinfo->valid);
5351 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5352 goto nla_put_failure;
5353 nlmsg_end(skb, nlh);
5354 return 0;
5356 nla_put_failure:
5357 nlmsg_cancel(skb, nlh);
5358 return -EMSGSIZE;
5361 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5362 struct prefix_info *pinfo)
5364 struct sk_buff *skb;
5365 struct net *net = dev_net(idev->dev);
5366 int err = -ENOBUFS;
5368 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5369 if (!skb)
5370 goto errout;
5372 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5373 if (err < 0) {
5374 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5375 WARN_ON(err == -EMSGSIZE);
5376 kfree_skb(skb);
5377 goto errout;
5379 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5380 return;
5381 errout:
5382 if (err < 0)
5383 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5386 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5388 struct net *net = dev_net(ifp->idev->dev);
5390 if (event)
5391 ASSERT_RTNL();
5393 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5395 switch (event) {
5396 case RTM_NEWADDR:
5398 * If the address was optimistic
5399 * we inserted the route at the start of
5400 * our DAD process, so we don't need
5401 * to do it again
5403 if (!(ifp->rt->rt6i_node))
5404 ip6_ins_rt(ifp->rt);
5405 if (ifp->idev->cnf.forwarding)
5406 addrconf_join_anycast(ifp);
5407 if (!ipv6_addr_any(&ifp->peer_addr))
5408 addrconf_prefix_route(&ifp->peer_addr, 128,
5409 ifp->idev->dev, 0, 0);
5410 break;
5411 case RTM_DELADDR:
5412 if (ifp->idev->cnf.forwarding)
5413 addrconf_leave_anycast(ifp);
5414 addrconf_leave_solict(ifp->idev, &ifp->addr);
5415 if (!ipv6_addr_any(&ifp->peer_addr)) {
5416 struct rt6_info *rt;
5418 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5419 ifp->idev->dev, 0, 0);
5420 if (rt)
5421 ip6_del_rt(rt);
5423 if (ifp->rt) {
5424 dst_hold(&ifp->rt->dst);
5425 ip6_del_rt(ifp->rt);
5427 rt_genid_bump_ipv6(net);
5428 break;
5430 atomic_inc(&net->ipv6.dev_addr_genid);
5433 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5435 rcu_read_lock_bh();
5436 if (likely(ifp->idev->dead == 0))
5437 __ipv6_ifa_notify(event, ifp);
5438 rcu_read_unlock_bh();
5441 #ifdef CONFIG_SYSCTL
5443 static
5444 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5445 void __user *buffer, size_t *lenp, loff_t *ppos)
5447 int *valp = ctl->data;
5448 int val = *valp;
5449 loff_t pos = *ppos;
5450 struct ctl_table lctl;
5451 int ret;
5454 * ctl->data points to idev->cnf.forwarding, we should
5455 * not modify it until we get the rtnl lock.
5457 lctl = *ctl;
5458 lctl.data = &val;
5460 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5462 if (write)
5463 ret = addrconf_fixup_forwarding(ctl, valp, val);
5464 if (ret)
5465 *ppos = pos;
5466 return ret;
5469 static
5470 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5471 void __user *buffer, size_t *lenp, loff_t *ppos)
5473 struct ctl_table lctl;
5474 int min_hl = 1, max_hl = 255;
5476 lctl = *ctl;
5477 lctl.extra1 = &min_hl;
5478 lctl.extra2 = &max_hl;
5480 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5483 static
5484 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5485 void __user *buffer, size_t *lenp, loff_t *ppos)
5487 struct inet6_dev *idev = ctl->extra1;
5488 int min_mtu = IPV6_MIN_MTU;
5489 struct ctl_table lctl;
5491 lctl = *ctl;
5492 lctl.extra1 = &min_mtu;
5493 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5495 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5498 static void dev_disable_change(struct inet6_dev *idev)
5500 struct netdev_notifier_info info;
5502 if (!idev || !idev->dev)
5503 return;
5505 netdev_notifier_info_init(&info, idev->dev);
5506 if (idev->cnf.disable_ipv6)
5507 addrconf_notify(NULL, NETDEV_DOWN, &info);
5508 else
5509 addrconf_notify(NULL, NETDEV_UP, &info);
5512 static void addrconf_disable_change(struct net *net, __s32 newf)
5514 struct net_device *dev;
5515 struct inet6_dev *idev;
5517 rcu_read_lock();
5518 for_each_netdev_rcu(net, dev) {
5519 idev = __in6_dev_get(dev);
5520 if (idev) {
5521 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5522 idev->cnf.disable_ipv6 = newf;
5523 if (changed)
5524 dev_disable_change(idev);
5527 rcu_read_unlock();
5530 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5532 struct net *net;
5533 int old;
5535 if (!rtnl_trylock())
5536 return restart_syscall();
5538 net = (struct net *)table->extra2;
5539 old = *p;
5540 *p = newf;
5542 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5543 rtnl_unlock();
5544 return 0;
5547 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5548 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5549 addrconf_disable_change(net, newf);
5550 } else if ((!newf) ^ (!old))
5551 dev_disable_change((struct inet6_dev *)table->extra1);
5553 rtnl_unlock();
5554 return 0;
5557 static
5558 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5559 void __user *buffer, size_t *lenp, loff_t *ppos)
5561 int *valp = ctl->data;
5562 int val = *valp;
5563 loff_t pos = *ppos;
5564 struct ctl_table lctl;
5565 int ret;
5568 * ctl->data points to idev->cnf.disable_ipv6, we should
5569 * not modify it until we get the rtnl lock.
5571 lctl = *ctl;
5572 lctl.data = &val;
5574 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5576 if (write)
5577 ret = addrconf_disable_ipv6(ctl, valp, val);
5578 if (ret)
5579 *ppos = pos;
5580 return ret;
5583 static
5584 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5585 void __user *buffer, size_t *lenp, loff_t *ppos)
5587 int *valp = ctl->data;
5588 int ret;
5589 int old, new;
5591 old = *valp;
5592 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5593 new = *valp;
5595 if (write && old != new) {
5596 struct net *net = ctl->extra2;
5598 if (!rtnl_trylock())
5599 return restart_syscall();
5601 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5602 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5603 NETCONFA_IFINDEX_DEFAULT,
5604 net->ipv6.devconf_dflt);
5605 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5606 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5607 NETCONFA_IFINDEX_ALL,
5608 net->ipv6.devconf_all);
5609 else {
5610 struct inet6_dev *idev = ctl->extra1;
5612 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5613 idev->dev->ifindex,
5614 &idev->cnf);
5616 rtnl_unlock();
5619 return ret;
5622 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5623 void __user *buffer, size_t *lenp,
5624 loff_t *ppos)
5626 int err;
5627 struct in6_addr addr;
5628 char str[IPV6_MAX_STRLEN];
5629 struct ctl_table lctl = *ctl;
5630 struct net *net = ctl->extra2;
5631 struct ipv6_stable_secret *secret = ctl->data;
5633 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5634 return -EIO;
5636 lctl.maxlen = IPV6_MAX_STRLEN;
5637 lctl.data = str;
5639 if (!rtnl_trylock())
5640 return restart_syscall();
5642 if (!write && !secret->initialized) {
5643 err = -EIO;
5644 goto out;
5647 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5648 if (err >= sizeof(str)) {
5649 err = -EIO;
5650 goto out;
5653 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5654 if (err || !write)
5655 goto out;
5657 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5658 err = -EIO;
5659 goto out;
5662 secret->initialized = true;
5663 secret->secret = addr;
5665 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5666 struct net_device *dev;
5668 for_each_netdev(net, dev) {
5669 struct inet6_dev *idev = __in6_dev_get(dev);
5671 if (idev) {
5672 idev->addr_gen_mode =
5673 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5676 } else {
5677 struct inet6_dev *idev = ctl->extra1;
5679 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5682 out:
5683 rtnl_unlock();
5685 return err;
5688 static
5689 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5690 int write,
5691 void __user *buffer,
5692 size_t *lenp,
5693 loff_t *ppos)
5695 int *valp = ctl->data;
5696 int val = *valp;
5697 loff_t pos = *ppos;
5698 struct ctl_table lctl;
5699 int ret;
5701 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5702 * we should not modify it until we get the rtnl lock.
5704 lctl = *ctl;
5705 lctl.data = &val;
5707 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5709 if (write)
5710 ret = addrconf_fixup_linkdown(ctl, valp, val);
5711 if (ret)
5712 *ppos = pos;
5713 return ret;
5716 static const struct ctl_table addrconf_sysctl[] = {
5718 .procname = "forwarding",
5719 .data = &ipv6_devconf.forwarding,
5720 .maxlen = sizeof(int),
5721 .mode = 0644,
5722 .proc_handler = addrconf_sysctl_forward,
5725 .procname = "hop_limit",
5726 .data = &ipv6_devconf.hop_limit,
5727 .maxlen = sizeof(int),
5728 .mode = 0644,
5729 .proc_handler = addrconf_sysctl_hop_limit,
5732 .procname = "mtu",
5733 .data = &ipv6_devconf.mtu6,
5734 .maxlen = sizeof(int),
5735 .mode = 0644,
5736 .proc_handler = addrconf_sysctl_mtu,
5739 .procname = "accept_ra",
5740 .data = &ipv6_devconf.accept_ra,
5741 .maxlen = sizeof(int),
5742 .mode = 0644,
5743 .proc_handler = proc_dointvec,
5746 .procname = "accept_redirects",
5747 .data = &ipv6_devconf.accept_redirects,
5748 .maxlen = sizeof(int),
5749 .mode = 0644,
5750 .proc_handler = proc_dointvec,
5753 .procname = "autoconf",
5754 .data = &ipv6_devconf.autoconf,
5755 .maxlen = sizeof(int),
5756 .mode = 0644,
5757 .proc_handler = proc_dointvec,
5760 .procname = "dad_transmits",
5761 .data = &ipv6_devconf.dad_transmits,
5762 .maxlen = sizeof(int),
5763 .mode = 0644,
5764 .proc_handler = proc_dointvec,
5767 .procname = "router_solicitations",
5768 .data = &ipv6_devconf.rtr_solicits,
5769 .maxlen = sizeof(int),
5770 .mode = 0644,
5771 .proc_handler = proc_dointvec,
5774 .procname = "router_solicitation_interval",
5775 .data = &ipv6_devconf.rtr_solicit_interval,
5776 .maxlen = sizeof(int),
5777 .mode = 0644,
5778 .proc_handler = proc_dointvec_jiffies,
5781 .procname = "router_solicitation_delay",
5782 .data = &ipv6_devconf.rtr_solicit_delay,
5783 .maxlen = sizeof(int),
5784 .mode = 0644,
5785 .proc_handler = proc_dointvec_jiffies,
5788 .procname = "force_mld_version",
5789 .data = &ipv6_devconf.force_mld_version,
5790 .maxlen = sizeof(int),
5791 .mode = 0644,
5792 .proc_handler = proc_dointvec,
5795 .procname = "mldv1_unsolicited_report_interval",
5796 .data =
5797 &ipv6_devconf.mldv1_unsolicited_report_interval,
5798 .maxlen = sizeof(int),
5799 .mode = 0644,
5800 .proc_handler = proc_dointvec_ms_jiffies,
5803 .procname = "mldv2_unsolicited_report_interval",
5804 .data =
5805 &ipv6_devconf.mldv2_unsolicited_report_interval,
5806 .maxlen = sizeof(int),
5807 .mode = 0644,
5808 .proc_handler = proc_dointvec_ms_jiffies,
5811 .procname = "use_tempaddr",
5812 .data = &ipv6_devconf.use_tempaddr,
5813 .maxlen = sizeof(int),
5814 .mode = 0644,
5815 .proc_handler = proc_dointvec,
5818 .procname = "temp_valid_lft",
5819 .data = &ipv6_devconf.temp_valid_lft,
5820 .maxlen = sizeof(int),
5821 .mode = 0644,
5822 .proc_handler = proc_dointvec,
5825 .procname = "temp_prefered_lft",
5826 .data = &ipv6_devconf.temp_prefered_lft,
5827 .maxlen = sizeof(int),
5828 .mode = 0644,
5829 .proc_handler = proc_dointvec,
5832 .procname = "regen_max_retry",
5833 .data = &ipv6_devconf.regen_max_retry,
5834 .maxlen = sizeof(int),
5835 .mode = 0644,
5836 .proc_handler = proc_dointvec,
5839 .procname = "max_desync_factor",
5840 .data = &ipv6_devconf.max_desync_factor,
5841 .maxlen = sizeof(int),
5842 .mode = 0644,
5843 .proc_handler = proc_dointvec,
5846 .procname = "max_addresses",
5847 .data = &ipv6_devconf.max_addresses,
5848 .maxlen = sizeof(int),
5849 .mode = 0644,
5850 .proc_handler = proc_dointvec,
5853 .procname = "accept_ra_defrtr",
5854 .data = &ipv6_devconf.accept_ra_defrtr,
5855 .maxlen = sizeof(int),
5856 .mode = 0644,
5857 .proc_handler = proc_dointvec,
5860 .procname = "accept_ra_min_hop_limit",
5861 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5862 .maxlen = sizeof(int),
5863 .mode = 0644,
5864 .proc_handler = proc_dointvec,
5867 .procname = "accept_ra_pinfo",
5868 .data = &ipv6_devconf.accept_ra_pinfo,
5869 .maxlen = sizeof(int),
5870 .mode = 0644,
5871 .proc_handler = proc_dointvec,
5873 #ifdef CONFIG_IPV6_ROUTER_PREF
5875 .procname = "accept_ra_rtr_pref",
5876 .data = &ipv6_devconf.accept_ra_rtr_pref,
5877 .maxlen = sizeof(int),
5878 .mode = 0644,
5879 .proc_handler = proc_dointvec,
5882 .procname = "router_probe_interval",
5883 .data = &ipv6_devconf.rtr_probe_interval,
5884 .maxlen = sizeof(int),
5885 .mode = 0644,
5886 .proc_handler = proc_dointvec_jiffies,
5888 #ifdef CONFIG_IPV6_ROUTE_INFO
5890 .procname = "accept_ra_rt_info_max_plen",
5891 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5892 .maxlen = sizeof(int),
5893 .mode = 0644,
5894 .proc_handler = proc_dointvec,
5896 #endif
5897 #endif
5899 .procname = "proxy_ndp",
5900 .data = &ipv6_devconf.proxy_ndp,
5901 .maxlen = sizeof(int),
5902 .mode = 0644,
5903 .proc_handler = addrconf_sysctl_proxy_ndp,
5906 .procname = "accept_source_route",
5907 .data = &ipv6_devconf.accept_source_route,
5908 .maxlen = sizeof(int),
5909 .mode = 0644,
5910 .proc_handler = proc_dointvec,
5912 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5914 .procname = "optimistic_dad",
5915 .data = &ipv6_devconf.optimistic_dad,
5916 .maxlen = sizeof(int),
5917 .mode = 0644,
5918 .proc_handler = proc_dointvec,
5921 .procname = "use_optimistic",
5922 .data = &ipv6_devconf.use_optimistic,
5923 .maxlen = sizeof(int),
5924 .mode = 0644,
5925 .proc_handler = proc_dointvec,
5927 #endif
5928 #ifdef CONFIG_IPV6_MROUTE
5930 .procname = "mc_forwarding",
5931 .data = &ipv6_devconf.mc_forwarding,
5932 .maxlen = sizeof(int),
5933 .mode = 0444,
5934 .proc_handler = proc_dointvec,
5936 #endif
5938 .procname = "disable_ipv6",
5939 .data = &ipv6_devconf.disable_ipv6,
5940 .maxlen = sizeof(int),
5941 .mode = 0644,
5942 .proc_handler = addrconf_sysctl_disable,
5945 .procname = "accept_dad",
5946 .data = &ipv6_devconf.accept_dad,
5947 .maxlen = sizeof(int),
5948 .mode = 0644,
5949 .proc_handler = proc_dointvec,
5952 .procname = "force_tllao",
5953 .data = &ipv6_devconf.force_tllao,
5954 .maxlen = sizeof(int),
5955 .mode = 0644,
5956 .proc_handler = proc_dointvec
5959 .procname = "ndisc_notify",
5960 .data = &ipv6_devconf.ndisc_notify,
5961 .maxlen = sizeof(int),
5962 .mode = 0644,
5963 .proc_handler = proc_dointvec
5966 .procname = "suppress_frag_ndisc",
5967 .data = &ipv6_devconf.suppress_frag_ndisc,
5968 .maxlen = sizeof(int),
5969 .mode = 0644,
5970 .proc_handler = proc_dointvec
5973 .procname = "accept_ra_from_local",
5974 .data = &ipv6_devconf.accept_ra_from_local,
5975 .maxlen = sizeof(int),
5976 .mode = 0644,
5977 .proc_handler = proc_dointvec,
5980 .procname = "accept_ra_mtu",
5981 .data = &ipv6_devconf.accept_ra_mtu,
5982 .maxlen = sizeof(int),
5983 .mode = 0644,
5984 .proc_handler = proc_dointvec,
5987 .procname = "stable_secret",
5988 .data = &ipv6_devconf.stable_secret,
5989 .maxlen = IPV6_MAX_STRLEN,
5990 .mode = 0600,
5991 .proc_handler = addrconf_sysctl_stable_secret,
5994 .procname = "use_oif_addrs_only",
5995 .data = &ipv6_devconf.use_oif_addrs_only,
5996 .maxlen = sizeof(int),
5997 .mode = 0644,
5998 .proc_handler = proc_dointvec,
6001 .procname = "ignore_routes_with_linkdown",
6002 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6003 .maxlen = sizeof(int),
6004 .mode = 0644,
6005 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6008 .procname = "drop_unicast_in_l2_multicast",
6009 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6010 .maxlen = sizeof(int),
6011 .mode = 0644,
6012 .proc_handler = proc_dointvec,
6015 .procname = "drop_unsolicited_na",
6016 .data = &ipv6_devconf.drop_unsolicited_na,
6017 .maxlen = sizeof(int),
6018 .mode = 0644,
6019 .proc_handler = proc_dointvec,
6022 .procname = "keep_addr_on_down",
6023 .data = &ipv6_devconf.keep_addr_on_down,
6024 .maxlen = sizeof(int),
6025 .mode = 0644,
6026 .proc_handler = proc_dointvec,
6030 /* sentinel */
6034 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6035 struct inet6_dev *idev, struct ipv6_devconf *p)
6037 int i, ifindex;
6038 struct ctl_table *table;
6039 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6041 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6042 if (!table)
6043 goto out;
6045 for (i = 0; table[i].data; i++) {
6046 table[i].data += (char *)p - (char *)&ipv6_devconf;
6047 table[i].extra1 = idev; /* embedded; no ref */
6048 table[i].extra2 = net;
6051 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6053 p->sysctl_header = register_net_sysctl(net, path, table);
6054 if (!p->sysctl_header)
6055 goto free;
6057 if (!strcmp(dev_name, "all"))
6058 ifindex = NETCONFA_IFINDEX_ALL;
6059 else if (!strcmp(dev_name, "default"))
6060 ifindex = NETCONFA_IFINDEX_DEFAULT;
6061 else
6062 ifindex = idev->dev->ifindex;
6063 inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6064 return 0;
6066 free:
6067 kfree(table);
6068 out:
6069 return -ENOBUFS;
6072 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6074 struct ctl_table *table;
6076 if (!p->sysctl_header)
6077 return;
6079 table = p->sysctl_header->ctl_table_arg;
6080 unregister_net_sysctl_table(p->sysctl_header);
6081 p->sysctl_header = NULL;
6082 kfree(table);
6085 static int addrconf_sysctl_register(struct inet6_dev *idev)
6087 int err;
6089 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6090 return -EINVAL;
6092 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6093 &ndisc_ifinfo_sysctl_change);
6094 if (err)
6095 return err;
6096 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6097 idev, &idev->cnf);
6098 if (err)
6099 neigh_sysctl_unregister(idev->nd_parms);
6101 return err;
6104 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6106 __addrconf_sysctl_unregister(&idev->cnf);
6107 neigh_sysctl_unregister(idev->nd_parms);
6111 #endif
6113 static int __net_init addrconf_init_net(struct net *net)
6115 int err = -ENOMEM;
6116 struct ipv6_devconf *all, *dflt;
6118 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6119 if (!all)
6120 goto err_alloc_all;
6122 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6123 if (!dflt)
6124 goto err_alloc_dflt;
6126 /* these will be inherited by all namespaces */
6127 dflt->autoconf = ipv6_defaults.autoconf;
6128 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6130 dflt->stable_secret.initialized = false;
6131 all->stable_secret.initialized = false;
6133 net->ipv6.devconf_all = all;
6134 net->ipv6.devconf_dflt = dflt;
6136 #ifdef CONFIG_SYSCTL
6137 err = __addrconf_sysctl_register(net, "all", NULL, all);
6138 if (err < 0)
6139 goto err_reg_all;
6141 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6142 if (err < 0)
6143 goto err_reg_dflt;
6144 #endif
6145 return 0;
6147 #ifdef CONFIG_SYSCTL
6148 err_reg_dflt:
6149 __addrconf_sysctl_unregister(all);
6150 err_reg_all:
6151 kfree(dflt);
6152 #endif
6153 err_alloc_dflt:
6154 kfree(all);
6155 err_alloc_all:
6156 return err;
6159 static void __net_exit addrconf_exit_net(struct net *net)
6161 #ifdef CONFIG_SYSCTL
6162 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6163 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6164 #endif
6165 kfree(net->ipv6.devconf_dflt);
6166 kfree(net->ipv6.devconf_all);
6169 static struct pernet_operations addrconf_ops = {
6170 .init = addrconf_init_net,
6171 .exit = addrconf_exit_net,
6174 static struct rtnl_af_ops inet6_ops __read_mostly = {
6175 .family = AF_INET6,
6176 .fill_link_af = inet6_fill_link_af,
6177 .get_link_af_size = inet6_get_link_af_size,
6178 .validate_link_af = inet6_validate_link_af,
6179 .set_link_af = inet6_set_link_af,
6183 * Init / cleanup code
6186 int __init addrconf_init(void)
6188 struct inet6_dev *idev;
6189 int i, err;
6191 err = ipv6_addr_label_init();
6192 if (err < 0) {
6193 pr_crit("%s: cannot initialize default policy table: %d\n",
6194 __func__, err);
6195 goto out;
6198 err = register_pernet_subsys(&addrconf_ops);
6199 if (err < 0)
6200 goto out_addrlabel;
6202 addrconf_wq = create_workqueue("ipv6_addrconf");
6203 if (!addrconf_wq) {
6204 err = -ENOMEM;
6205 goto out_nowq;
6208 /* The addrconf netdev notifier requires that loopback_dev
6209 * has it's ipv6 private information allocated and setup
6210 * before it can bring up and give link-local addresses
6211 * to other devices which are up.
6213 * Unfortunately, loopback_dev is not necessarily the first
6214 * entry in the global dev_base list of net devices. In fact,
6215 * it is likely to be the very last entry on that list.
6216 * So this causes the notifier registry below to try and
6217 * give link-local addresses to all devices besides loopback_dev
6218 * first, then loopback_dev, which cases all the non-loopback_dev
6219 * devices to fail to get a link-local address.
6221 * So, as a temporary fix, allocate the ipv6 structure for
6222 * loopback_dev first by hand.
6223 * Longer term, all of the dependencies ipv6 has upon the loopback
6224 * device and it being up should be removed.
6226 rtnl_lock();
6227 idev = ipv6_add_dev(init_net.loopback_dev);
6228 rtnl_unlock();
6229 if (IS_ERR(idev)) {
6230 err = PTR_ERR(idev);
6231 goto errlo;
6234 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6235 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6237 register_netdevice_notifier(&ipv6_dev_notf);
6239 addrconf_verify();
6241 rtnl_af_register(&inet6_ops);
6243 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6244 NULL);
6245 if (err < 0)
6246 goto errout;
6248 /* Only the first call to __rtnl_register can fail */
6249 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6250 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6251 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6252 inet6_dump_ifaddr, NULL);
6253 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6254 inet6_dump_ifmcaddr, NULL);
6255 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6256 inet6_dump_ifacaddr, NULL);
6257 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6258 inet6_netconf_dump_devconf, NULL);
6260 ipv6_addr_label_rtnl_register();
6262 return 0;
6263 errout:
6264 rtnl_af_unregister(&inet6_ops);
6265 unregister_netdevice_notifier(&ipv6_dev_notf);
6266 errlo:
6267 destroy_workqueue(addrconf_wq);
6268 out_nowq:
6269 unregister_pernet_subsys(&addrconf_ops);
6270 out_addrlabel:
6271 ipv6_addr_label_cleanup();
6272 out:
6273 return err;
6276 void addrconf_cleanup(void)
6278 struct net_device *dev;
6279 int i;
6281 unregister_netdevice_notifier(&ipv6_dev_notf);
6282 unregister_pernet_subsys(&addrconf_ops);
6283 ipv6_addr_label_cleanup();
6285 rtnl_lock();
6287 __rtnl_af_unregister(&inet6_ops);
6289 /* clean dev list */
6290 for_each_netdev(&init_net, dev) {
6291 if (__in6_dev_get(dev) == NULL)
6292 continue;
6293 addrconf_ifdown(dev, 1);
6295 addrconf_ifdown(init_net.loopback_dev, 2);
6298 * Check hash table.
6300 spin_lock_bh(&addrconf_hash_lock);
6301 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6302 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6303 spin_unlock_bh(&addrconf_hash_lock);
6304 cancel_delayed_work(&addr_chk_work);
6305 rtnl_unlock();
6307 destroy_workqueue(addrconf_wq);