ipv6: do not abuse GFP_ATOMIC in inet6_netconf_notify_devconf()
[linux-2.6/btrfs-unstable.git] / net / ipv6 / addrconf.c
blob24f1b0898e4013a45ce21cc73090607f9578131a
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 net->ipv6.devconf_dflt->forwarding = newf;
782 addrconf_forward_change(net, newf);
783 if ((!newf) ^ (!old))
784 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
785 NETCONFA_IFINDEX_ALL,
786 net->ipv6.devconf_all);
787 } else if ((!newf) ^ (!old))
788 dev_forward_change((struct inet6_dev *)table->extra1);
789 rtnl_unlock();
791 if (newf)
792 rt6_purge_dflt_routers(net);
793 return 1;
796 static void addrconf_linkdown_change(struct net *net, __s32 newf)
798 struct net_device *dev;
799 struct inet6_dev *idev;
801 for_each_netdev(net, dev) {
802 idev = __in6_dev_get(dev);
803 if (idev) {
804 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
806 idev->cnf.ignore_routes_with_linkdown = newf;
807 if (changed)
808 inet6_netconf_notify_devconf(dev_net(dev),
809 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
810 dev->ifindex,
811 &idev->cnf);
816 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
818 struct net *net;
819 int old;
821 if (!rtnl_trylock())
822 return restart_syscall();
824 net = (struct net *)table->extra2;
825 old = *p;
826 *p = newf;
828 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
829 if ((!newf) ^ (!old))
830 inet6_netconf_notify_devconf(net,
831 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
832 NETCONFA_IFINDEX_DEFAULT,
833 net->ipv6.devconf_dflt);
834 rtnl_unlock();
835 return 0;
838 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
839 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
840 addrconf_linkdown_change(net, newf);
841 if ((!newf) ^ (!old))
842 inet6_netconf_notify_devconf(net,
843 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
844 NETCONFA_IFINDEX_ALL,
845 net->ipv6.devconf_all);
847 rtnl_unlock();
849 return 1;
852 #endif
854 /* Nobody refers to this ifaddr, destroy it */
855 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
857 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
859 #ifdef NET_REFCNT_DEBUG
860 pr_debug("%s\n", __func__);
861 #endif
863 in6_dev_put(ifp->idev);
865 if (cancel_delayed_work(&ifp->dad_work))
866 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
867 ifp);
869 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
870 pr_warn("Freeing alive inet6 address %p\n", ifp);
871 return;
873 ip6_rt_put(ifp->rt);
875 kfree_rcu(ifp, rcu);
878 static void
879 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
881 struct list_head *p;
882 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
885 * Each device address list is sorted in order of scope -
886 * global before linklocal.
888 list_for_each(p, &idev->addr_list) {
889 struct inet6_ifaddr *ifa
890 = list_entry(p, struct inet6_ifaddr, if_list);
891 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
892 break;
895 list_add_tail(&ifp->if_list, p);
898 static u32 inet6_addr_hash(const struct in6_addr *addr)
900 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
903 /* On success it returns ifp with increased reference count */
905 static struct inet6_ifaddr *
906 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
907 const struct in6_addr *peer_addr, int pfxlen,
908 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
910 struct inet6_ifaddr *ifa = NULL;
911 struct rt6_info *rt;
912 unsigned int hash;
913 int err = 0;
914 int addr_type = ipv6_addr_type(addr);
916 if (addr_type == IPV6_ADDR_ANY ||
917 addr_type & IPV6_ADDR_MULTICAST ||
918 (!(idev->dev->flags & IFF_LOOPBACK) &&
919 addr_type & IPV6_ADDR_LOOPBACK))
920 return ERR_PTR(-EADDRNOTAVAIL);
922 rcu_read_lock_bh();
923 if (idev->dead) {
924 err = -ENODEV; /*XXX*/
925 goto out2;
928 if (idev->cnf.disable_ipv6) {
929 err = -EACCES;
930 goto out2;
933 spin_lock(&addrconf_hash_lock);
935 /* Ignore adding duplicate addresses on an interface */
936 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
937 ADBG("ipv6_add_addr: already assigned\n");
938 err = -EEXIST;
939 goto out;
942 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
944 if (!ifa) {
945 ADBG("ipv6_add_addr: malloc failed\n");
946 err = -ENOBUFS;
947 goto out;
950 rt = addrconf_dst_alloc(idev, addr, false);
951 if (IS_ERR(rt)) {
952 err = PTR_ERR(rt);
953 goto out;
956 neigh_parms_data_state_setall(idev->nd_parms);
958 ifa->addr = *addr;
959 if (peer_addr)
960 ifa->peer_addr = *peer_addr;
962 spin_lock_init(&ifa->lock);
963 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
964 INIT_HLIST_NODE(&ifa->addr_lst);
965 ifa->scope = scope;
966 ifa->prefix_len = pfxlen;
967 ifa->flags = flags | IFA_F_TENTATIVE;
968 ifa->valid_lft = valid_lft;
969 ifa->prefered_lft = prefered_lft;
970 ifa->cstamp = ifa->tstamp = jiffies;
971 ifa->tokenized = false;
973 ifa->rt = rt;
975 ifa->idev = idev;
976 in6_dev_hold(idev);
977 /* For caller */
978 in6_ifa_hold(ifa);
980 /* Add to big hash table */
981 hash = inet6_addr_hash(addr);
983 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
984 spin_unlock(&addrconf_hash_lock);
986 write_lock(&idev->lock);
987 /* Add to inet6_dev unicast addr list. */
988 ipv6_link_dev_addr(idev, ifa);
990 if (ifa->flags&IFA_F_TEMPORARY) {
991 list_add(&ifa->tmp_list, &idev->tempaddr_list);
992 in6_ifa_hold(ifa);
995 in6_ifa_hold(ifa);
996 write_unlock(&idev->lock);
997 out2:
998 rcu_read_unlock_bh();
1000 if (likely(err == 0))
1001 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1002 else {
1003 kfree(ifa);
1004 ifa = ERR_PTR(err);
1007 return ifa;
1008 out:
1009 spin_unlock(&addrconf_hash_lock);
1010 goto out2;
1013 enum cleanup_prefix_rt_t {
1014 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1015 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1016 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1020 * Check, whether the prefix for ifp would still need a prefix route
1021 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1022 * constants.
1024 * 1) we don't purge prefix if address was not permanent.
1025 * prefix is managed by its own lifetime.
1026 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1027 * 3) if there are no addresses, delete prefix.
1028 * 4) if there are still other permanent address(es),
1029 * corresponding prefix is still permanent.
1030 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1031 * don't purge the prefix, assume user space is managing it.
1032 * 6) otherwise, update prefix lifetime to the
1033 * longest valid lifetime among the corresponding
1034 * addresses on the device.
1035 * Note: subsequent RA will update lifetime.
1037 static enum cleanup_prefix_rt_t
1038 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1040 struct inet6_ifaddr *ifa;
1041 struct inet6_dev *idev = ifp->idev;
1042 unsigned long lifetime;
1043 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1045 *expires = jiffies;
1047 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1048 if (ifa == ifp)
1049 continue;
1050 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1051 ifp->prefix_len))
1052 continue;
1053 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1054 return CLEANUP_PREFIX_RT_NOP;
1056 action = CLEANUP_PREFIX_RT_EXPIRE;
1058 spin_lock(&ifa->lock);
1060 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1062 * Note: Because this address is
1063 * not permanent, lifetime <
1064 * LONG_MAX / HZ here.
1066 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1067 *expires = ifa->tstamp + lifetime * HZ;
1068 spin_unlock(&ifa->lock);
1071 return action;
1074 static void
1075 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1077 struct rt6_info *rt;
1079 rt = addrconf_get_prefix_route(&ifp->addr,
1080 ifp->prefix_len,
1081 ifp->idev->dev,
1082 0, RTF_GATEWAY | RTF_DEFAULT);
1083 if (rt) {
1084 if (del_rt)
1085 ip6_del_rt(rt);
1086 else {
1087 if (!(rt->rt6i_flags & RTF_EXPIRES))
1088 rt6_set_expires(rt, expires);
1089 ip6_rt_put(rt);
1095 /* This function wants to get referenced ifp and releases it before return */
1097 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1099 int state;
1100 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1101 unsigned long expires;
1103 ASSERT_RTNL();
1105 spin_lock_bh(&ifp->lock);
1106 state = ifp->state;
1107 ifp->state = INET6_IFADDR_STATE_DEAD;
1108 spin_unlock_bh(&ifp->lock);
1110 if (state == INET6_IFADDR_STATE_DEAD)
1111 goto out;
1113 spin_lock_bh(&addrconf_hash_lock);
1114 hlist_del_init_rcu(&ifp->addr_lst);
1115 spin_unlock_bh(&addrconf_hash_lock);
1117 write_lock_bh(&ifp->idev->lock);
1119 if (ifp->flags&IFA_F_TEMPORARY) {
1120 list_del(&ifp->tmp_list);
1121 if (ifp->ifpub) {
1122 in6_ifa_put(ifp->ifpub);
1123 ifp->ifpub = NULL;
1125 __in6_ifa_put(ifp);
1128 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1129 action = check_cleanup_prefix_route(ifp, &expires);
1131 list_del_init(&ifp->if_list);
1132 __in6_ifa_put(ifp);
1134 write_unlock_bh(&ifp->idev->lock);
1136 addrconf_del_dad_work(ifp);
1138 ipv6_ifa_notify(RTM_DELADDR, ifp);
1140 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1142 if (action != CLEANUP_PREFIX_RT_NOP) {
1143 cleanup_prefix_route(ifp, expires,
1144 action == CLEANUP_PREFIX_RT_DEL);
1147 /* clean up prefsrc entries */
1148 rt6_remove_prefsrc(ifp);
1149 out:
1150 in6_ifa_put(ifp);
1153 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1155 struct inet6_dev *idev = ifp->idev;
1156 struct in6_addr addr, *tmpaddr;
1157 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1158 unsigned long regen_advance;
1159 int tmp_plen;
1160 int ret = 0;
1161 u32 addr_flags;
1162 unsigned long now = jiffies;
1164 write_lock_bh(&idev->lock);
1165 if (ift) {
1166 spin_lock_bh(&ift->lock);
1167 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1168 spin_unlock_bh(&ift->lock);
1169 tmpaddr = &addr;
1170 } else {
1171 tmpaddr = NULL;
1173 retry:
1174 in6_dev_hold(idev);
1175 if (idev->cnf.use_tempaddr <= 0) {
1176 write_unlock_bh(&idev->lock);
1177 pr_info("%s: use_tempaddr is disabled\n", __func__);
1178 in6_dev_put(idev);
1179 ret = -1;
1180 goto out;
1182 spin_lock_bh(&ifp->lock);
1183 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1184 idev->cnf.use_tempaddr = -1; /*XXX*/
1185 spin_unlock_bh(&ifp->lock);
1186 write_unlock_bh(&idev->lock);
1187 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1188 __func__);
1189 in6_dev_put(idev);
1190 ret = -1;
1191 goto out;
1193 in6_ifa_hold(ifp);
1194 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1195 __ipv6_try_regen_rndid(idev, tmpaddr);
1196 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1197 age = (now - ifp->tstamp) / HZ;
1198 tmp_valid_lft = min_t(__u32,
1199 ifp->valid_lft,
1200 idev->cnf.temp_valid_lft + age);
1201 tmp_prefered_lft = min_t(__u32,
1202 ifp->prefered_lft,
1203 idev->cnf.temp_prefered_lft + age -
1204 idev->cnf.max_desync_factor);
1205 tmp_plen = ifp->prefix_len;
1206 tmp_tstamp = ifp->tstamp;
1207 spin_unlock_bh(&ifp->lock);
1209 regen_advance = idev->cnf.regen_max_retry *
1210 idev->cnf.dad_transmits *
1211 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1212 write_unlock_bh(&idev->lock);
1214 /* A temporary address is created only if this calculated Preferred
1215 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1216 * an implementation must not create a temporary address with a zero
1217 * Preferred Lifetime.
1218 * Use age calculation as in addrconf_verify to avoid unnecessary
1219 * temporary addresses being generated.
1221 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1222 if (tmp_prefered_lft <= regen_advance + age) {
1223 in6_ifa_put(ifp);
1224 in6_dev_put(idev);
1225 ret = -1;
1226 goto out;
1229 addr_flags = IFA_F_TEMPORARY;
1230 /* set in addrconf_prefix_rcv() */
1231 if (ifp->flags & IFA_F_OPTIMISTIC)
1232 addr_flags |= IFA_F_OPTIMISTIC;
1234 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1235 ipv6_addr_scope(&addr), addr_flags,
1236 tmp_valid_lft, tmp_prefered_lft);
1237 if (IS_ERR(ift)) {
1238 in6_ifa_put(ifp);
1239 in6_dev_put(idev);
1240 pr_info("%s: retry temporary address regeneration\n", __func__);
1241 tmpaddr = &addr;
1242 write_lock_bh(&idev->lock);
1243 goto retry;
1246 spin_lock_bh(&ift->lock);
1247 ift->ifpub = ifp;
1248 ift->cstamp = now;
1249 ift->tstamp = tmp_tstamp;
1250 spin_unlock_bh(&ift->lock);
1252 addrconf_dad_start(ift);
1253 in6_ifa_put(ift);
1254 in6_dev_put(idev);
1255 out:
1256 return ret;
1260 * Choose an appropriate source address (RFC3484)
1262 enum {
1263 IPV6_SADDR_RULE_INIT = 0,
1264 IPV6_SADDR_RULE_LOCAL,
1265 IPV6_SADDR_RULE_SCOPE,
1266 IPV6_SADDR_RULE_PREFERRED,
1267 #ifdef CONFIG_IPV6_MIP6
1268 IPV6_SADDR_RULE_HOA,
1269 #endif
1270 IPV6_SADDR_RULE_OIF,
1271 IPV6_SADDR_RULE_LABEL,
1272 IPV6_SADDR_RULE_PRIVACY,
1273 IPV6_SADDR_RULE_ORCHID,
1274 IPV6_SADDR_RULE_PREFIX,
1275 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1276 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1277 #endif
1278 IPV6_SADDR_RULE_MAX
1281 struct ipv6_saddr_score {
1282 int rule;
1283 int addr_type;
1284 struct inet6_ifaddr *ifa;
1285 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1286 int scopedist;
1287 int matchlen;
1290 struct ipv6_saddr_dst {
1291 const struct in6_addr *addr;
1292 int ifindex;
1293 int scope;
1294 int label;
1295 unsigned int prefs;
1298 static inline int ipv6_saddr_preferred(int type)
1300 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1301 return 1;
1302 return 0;
1305 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1307 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1308 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1309 #else
1310 return false;
1311 #endif
1314 static int ipv6_get_saddr_eval(struct net *net,
1315 struct ipv6_saddr_score *score,
1316 struct ipv6_saddr_dst *dst,
1317 int i)
1319 int ret;
1321 if (i <= score->rule) {
1322 switch (i) {
1323 case IPV6_SADDR_RULE_SCOPE:
1324 ret = score->scopedist;
1325 break;
1326 case IPV6_SADDR_RULE_PREFIX:
1327 ret = score->matchlen;
1328 break;
1329 default:
1330 ret = !!test_bit(i, score->scorebits);
1332 goto out;
1335 switch (i) {
1336 case IPV6_SADDR_RULE_INIT:
1337 /* Rule 0: remember if hiscore is not ready yet */
1338 ret = !!score->ifa;
1339 break;
1340 case IPV6_SADDR_RULE_LOCAL:
1341 /* Rule 1: Prefer same address */
1342 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1343 break;
1344 case IPV6_SADDR_RULE_SCOPE:
1345 /* Rule 2: Prefer appropriate scope
1347 * ret
1349 * -1 | d 15
1350 * ---+--+-+---> scope
1352 * | d is scope of the destination.
1353 * B-d | \
1354 * | \ <- smaller scope is better if
1355 * B-15 | \ if scope is enough for destination.
1356 * | ret = B - scope (-1 <= scope >= d <= 15).
1357 * d-C-1 | /
1358 * |/ <- greater is better
1359 * -C / if scope is not enough for destination.
1360 * /| ret = scope - C (-1 <= d < scope <= 15).
1362 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1363 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1364 * Assume B = 0 and we get C > 29.
1366 ret = __ipv6_addr_src_scope(score->addr_type);
1367 if (ret >= dst->scope)
1368 ret = -ret;
1369 else
1370 ret -= 128; /* 30 is enough */
1371 score->scopedist = ret;
1372 break;
1373 case IPV6_SADDR_RULE_PREFERRED:
1375 /* Rule 3: Avoid deprecated and optimistic addresses */
1376 u8 avoid = IFA_F_DEPRECATED;
1378 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1379 avoid |= IFA_F_OPTIMISTIC;
1380 ret = ipv6_saddr_preferred(score->addr_type) ||
1381 !(score->ifa->flags & avoid);
1382 break;
1384 #ifdef CONFIG_IPV6_MIP6
1385 case IPV6_SADDR_RULE_HOA:
1387 /* Rule 4: Prefer home address */
1388 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1389 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1390 break;
1392 #endif
1393 case IPV6_SADDR_RULE_OIF:
1394 /* Rule 5: Prefer outgoing interface */
1395 ret = (!dst->ifindex ||
1396 dst->ifindex == score->ifa->idev->dev->ifindex);
1397 break;
1398 case IPV6_SADDR_RULE_LABEL:
1399 /* Rule 6: Prefer matching label */
1400 ret = ipv6_addr_label(net,
1401 &score->ifa->addr, score->addr_type,
1402 score->ifa->idev->dev->ifindex) == dst->label;
1403 break;
1404 case IPV6_SADDR_RULE_PRIVACY:
1406 /* Rule 7: Prefer public address
1407 * Note: prefer temporary address if use_tempaddr >= 2
1409 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1410 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1411 score->ifa->idev->cnf.use_tempaddr >= 2;
1412 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1413 break;
1415 case IPV6_SADDR_RULE_ORCHID:
1416 /* Rule 8-: Prefer ORCHID vs ORCHID or
1417 * non-ORCHID vs non-ORCHID
1419 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1420 ipv6_addr_orchid(dst->addr));
1421 break;
1422 case IPV6_SADDR_RULE_PREFIX:
1423 /* Rule 8: Use longest matching prefix */
1424 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1425 if (ret > score->ifa->prefix_len)
1426 ret = score->ifa->prefix_len;
1427 score->matchlen = ret;
1428 break;
1429 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1430 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1431 /* Optimistic addresses still have lower precedence than other
1432 * preferred addresses.
1434 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1435 break;
1436 #endif
1437 default:
1438 ret = 0;
1441 if (ret)
1442 __set_bit(i, score->scorebits);
1443 score->rule = i;
1444 out:
1445 return ret;
1448 static int __ipv6_dev_get_saddr(struct net *net,
1449 struct ipv6_saddr_dst *dst,
1450 struct inet6_dev *idev,
1451 struct ipv6_saddr_score *scores,
1452 int hiscore_idx)
1454 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1456 read_lock_bh(&idev->lock);
1457 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1458 int i;
1461 * - Tentative Address (RFC2462 section 5.4)
1462 * - A tentative address is not considered
1463 * "assigned to an interface" in the traditional
1464 * sense, unless it is also flagged as optimistic.
1465 * - Candidate Source Address (section 4)
1466 * - In any case, anycast addresses, multicast
1467 * addresses, and the unspecified address MUST
1468 * NOT be included in a candidate set.
1470 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1471 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1472 continue;
1474 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1476 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1477 score->addr_type & IPV6_ADDR_MULTICAST)) {
1478 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1479 idev->dev->name);
1480 continue;
1483 score->rule = -1;
1484 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1486 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1487 int minihiscore, miniscore;
1489 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1490 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1492 if (minihiscore > miniscore) {
1493 if (i == IPV6_SADDR_RULE_SCOPE &&
1494 score->scopedist > 0) {
1496 * special case:
1497 * each remaining entry
1498 * has too small (not enough)
1499 * scope, because ifa entries
1500 * are sorted by their scope
1501 * values.
1503 goto out;
1505 break;
1506 } else if (minihiscore < miniscore) {
1507 if (hiscore->ifa)
1508 in6_ifa_put(hiscore->ifa);
1510 in6_ifa_hold(score->ifa);
1512 swap(hiscore, score);
1513 hiscore_idx = 1 - hiscore_idx;
1515 /* restore our iterator */
1516 score->ifa = hiscore->ifa;
1518 break;
1522 out:
1523 read_unlock_bh(&idev->lock);
1524 return hiscore_idx;
1527 static int ipv6_get_saddr_master(struct net *net,
1528 const struct net_device *dst_dev,
1529 const struct net_device *master,
1530 struct ipv6_saddr_dst *dst,
1531 struct ipv6_saddr_score *scores,
1532 int hiscore_idx)
1534 struct inet6_dev *idev;
1536 idev = __in6_dev_get(dst_dev);
1537 if (idev)
1538 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1539 scores, hiscore_idx);
1541 idev = __in6_dev_get(master);
1542 if (idev)
1543 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1544 scores, hiscore_idx);
1546 return hiscore_idx;
1549 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1550 const struct in6_addr *daddr, unsigned int prefs,
1551 struct in6_addr *saddr)
1553 struct ipv6_saddr_score scores[2], *hiscore;
1554 struct ipv6_saddr_dst dst;
1555 struct inet6_dev *idev;
1556 struct net_device *dev;
1557 int dst_type;
1558 bool use_oif_addr = false;
1559 int hiscore_idx = 0;
1561 dst_type = __ipv6_addr_type(daddr);
1562 dst.addr = daddr;
1563 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1564 dst.scope = __ipv6_addr_src_scope(dst_type);
1565 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1566 dst.prefs = prefs;
1568 scores[hiscore_idx].rule = -1;
1569 scores[hiscore_idx].ifa = NULL;
1571 rcu_read_lock();
1573 /* Candidate Source Address (section 4)
1574 * - multicast and link-local destination address,
1575 * the set of candidate source address MUST only
1576 * include addresses assigned to interfaces
1577 * belonging to the same link as the outgoing
1578 * interface.
1579 * (- For site-local destination addresses, the
1580 * set of candidate source addresses MUST only
1581 * include addresses assigned to interfaces
1582 * belonging to the same site as the outgoing
1583 * interface.)
1584 * - "It is RECOMMENDED that the candidate source addresses
1585 * be the set of unicast addresses assigned to the
1586 * interface that will be used to send to the destination
1587 * (the 'outgoing' interface)." (RFC 6724)
1589 if (dst_dev) {
1590 idev = __in6_dev_get(dst_dev);
1591 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1592 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1593 (idev && idev->cnf.use_oif_addrs_only)) {
1594 use_oif_addr = true;
1598 if (use_oif_addr) {
1599 if (idev)
1600 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1601 } else {
1602 const struct net_device *master;
1603 int master_idx = 0;
1605 /* if dst_dev exists and is enslaved to an L3 device, then
1606 * prefer addresses from dst_dev and then the master over
1607 * any other enslaved devices in the L3 domain.
1609 master = l3mdev_master_dev_rcu(dst_dev);
1610 if (master) {
1611 master_idx = master->ifindex;
1613 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1614 master, &dst,
1615 scores, hiscore_idx);
1617 if (scores[hiscore_idx].ifa)
1618 goto out;
1621 for_each_netdev_rcu(net, dev) {
1622 /* only consider addresses on devices in the
1623 * same L3 domain
1625 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1626 continue;
1627 idev = __in6_dev_get(dev);
1628 if (!idev)
1629 continue;
1630 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1634 out:
1635 rcu_read_unlock();
1637 hiscore = &scores[hiscore_idx];
1638 if (!hiscore->ifa)
1639 return -EADDRNOTAVAIL;
1641 *saddr = hiscore->ifa->addr;
1642 in6_ifa_put(hiscore->ifa);
1643 return 0;
1645 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1647 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1648 u32 banned_flags)
1650 struct inet6_ifaddr *ifp;
1651 int err = -EADDRNOTAVAIL;
1653 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1654 if (ifp->scope > IFA_LINK)
1655 break;
1656 if (ifp->scope == IFA_LINK &&
1657 !(ifp->flags & banned_flags)) {
1658 *addr = ifp->addr;
1659 err = 0;
1660 break;
1663 return err;
1666 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1667 u32 banned_flags)
1669 struct inet6_dev *idev;
1670 int err = -EADDRNOTAVAIL;
1672 rcu_read_lock();
1673 idev = __in6_dev_get(dev);
1674 if (idev) {
1675 read_lock_bh(&idev->lock);
1676 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1677 read_unlock_bh(&idev->lock);
1679 rcu_read_unlock();
1680 return err;
1683 static int ipv6_count_addresses(struct inet6_dev *idev)
1685 int cnt = 0;
1686 struct inet6_ifaddr *ifp;
1688 read_lock_bh(&idev->lock);
1689 list_for_each_entry(ifp, &idev->addr_list, if_list)
1690 cnt++;
1691 read_unlock_bh(&idev->lock);
1692 return cnt;
1695 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1696 const struct net_device *dev, int strict)
1698 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1700 EXPORT_SYMBOL(ipv6_chk_addr);
1702 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1703 const struct net_device *dev, int strict,
1704 u32 banned_flags)
1706 struct inet6_ifaddr *ifp;
1707 unsigned int hash = inet6_addr_hash(addr);
1708 u32 ifp_flags;
1710 rcu_read_lock_bh();
1711 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1712 if (!net_eq(dev_net(ifp->idev->dev), net))
1713 continue;
1714 /* Decouple optimistic from tentative for evaluation here.
1715 * Ban optimistic addresses explicitly, when required.
1717 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1718 ? (ifp->flags&~IFA_F_TENTATIVE)
1719 : ifp->flags;
1720 if (ipv6_addr_equal(&ifp->addr, addr) &&
1721 !(ifp_flags&banned_flags) &&
1722 (!dev || ifp->idev->dev == dev ||
1723 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1724 rcu_read_unlock_bh();
1725 return 1;
1729 rcu_read_unlock_bh();
1730 return 0;
1732 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1734 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1735 struct net_device *dev)
1737 unsigned int hash = inet6_addr_hash(addr);
1738 struct inet6_ifaddr *ifp;
1740 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1741 if (!net_eq(dev_net(ifp->idev->dev), net))
1742 continue;
1743 if (ipv6_addr_equal(&ifp->addr, addr)) {
1744 if (!dev || ifp->idev->dev == dev)
1745 return true;
1748 return false;
1751 /* Compares an address/prefix_len with addresses on device @dev.
1752 * If one is found it returns true.
1754 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1755 const unsigned int prefix_len, struct net_device *dev)
1757 struct inet6_dev *idev;
1758 struct inet6_ifaddr *ifa;
1759 bool ret = false;
1761 rcu_read_lock();
1762 idev = __in6_dev_get(dev);
1763 if (idev) {
1764 read_lock_bh(&idev->lock);
1765 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1766 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1767 if (ret)
1768 break;
1770 read_unlock_bh(&idev->lock);
1772 rcu_read_unlock();
1774 return ret;
1776 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1778 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1780 struct inet6_dev *idev;
1781 struct inet6_ifaddr *ifa;
1782 int onlink;
1784 onlink = 0;
1785 rcu_read_lock();
1786 idev = __in6_dev_get(dev);
1787 if (idev) {
1788 read_lock_bh(&idev->lock);
1789 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1790 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1791 ifa->prefix_len);
1792 if (onlink)
1793 break;
1795 read_unlock_bh(&idev->lock);
1797 rcu_read_unlock();
1798 return onlink;
1800 EXPORT_SYMBOL(ipv6_chk_prefix);
1802 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1803 struct net_device *dev, int strict)
1805 struct inet6_ifaddr *ifp, *result = NULL;
1806 unsigned int hash = inet6_addr_hash(addr);
1808 rcu_read_lock_bh();
1809 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1810 if (!net_eq(dev_net(ifp->idev->dev), net))
1811 continue;
1812 if (ipv6_addr_equal(&ifp->addr, addr)) {
1813 if (!dev || ifp->idev->dev == dev ||
1814 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1815 result = ifp;
1816 in6_ifa_hold(ifp);
1817 break;
1821 rcu_read_unlock_bh();
1823 return result;
1826 /* Gets referenced address, destroys ifaddr */
1828 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1830 if (dad_failed)
1831 ifp->flags |= IFA_F_DADFAILED;
1833 if (ifp->flags&IFA_F_PERMANENT) {
1834 spin_lock_bh(&ifp->lock);
1835 addrconf_del_dad_work(ifp);
1836 ifp->flags |= IFA_F_TENTATIVE;
1837 spin_unlock_bh(&ifp->lock);
1838 if (dad_failed)
1839 ipv6_ifa_notify(0, ifp);
1840 in6_ifa_put(ifp);
1841 } else if (ifp->flags&IFA_F_TEMPORARY) {
1842 struct inet6_ifaddr *ifpub;
1843 spin_lock_bh(&ifp->lock);
1844 ifpub = ifp->ifpub;
1845 if (ifpub) {
1846 in6_ifa_hold(ifpub);
1847 spin_unlock_bh(&ifp->lock);
1848 ipv6_create_tempaddr(ifpub, ifp);
1849 in6_ifa_put(ifpub);
1850 } else {
1851 spin_unlock_bh(&ifp->lock);
1853 ipv6_del_addr(ifp);
1854 } else {
1855 ipv6_del_addr(ifp);
1859 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1861 int err = -ENOENT;
1863 spin_lock_bh(&ifp->lock);
1864 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1865 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1866 err = 0;
1868 spin_unlock_bh(&ifp->lock);
1870 return err;
1873 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1875 struct in6_addr addr;
1876 struct inet6_dev *idev = ifp->idev;
1877 struct net *net = dev_net(ifp->idev->dev);
1879 if (addrconf_dad_end(ifp)) {
1880 in6_ifa_put(ifp);
1881 return;
1884 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1885 ifp->idev->dev->name, &ifp->addr);
1887 spin_lock_bh(&ifp->lock);
1889 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1890 int scope = ifp->scope;
1891 u32 flags = ifp->flags;
1892 struct in6_addr new_addr;
1893 struct inet6_ifaddr *ifp2;
1894 u32 valid_lft, preferred_lft;
1895 int pfxlen = ifp->prefix_len;
1896 int retries = ifp->stable_privacy_retry + 1;
1898 if (retries > net->ipv6.sysctl.idgen_retries) {
1899 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1900 ifp->idev->dev->name);
1901 goto errdad;
1904 new_addr = ifp->addr;
1905 if (ipv6_generate_stable_address(&new_addr, retries,
1906 idev))
1907 goto errdad;
1909 valid_lft = ifp->valid_lft;
1910 preferred_lft = ifp->prefered_lft;
1912 spin_unlock_bh(&ifp->lock);
1914 if (idev->cnf.max_addresses &&
1915 ipv6_count_addresses(idev) >=
1916 idev->cnf.max_addresses)
1917 goto lock_errdad;
1919 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1920 ifp->idev->dev->name);
1922 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1923 scope, flags, valid_lft,
1924 preferred_lft);
1925 if (IS_ERR(ifp2))
1926 goto lock_errdad;
1928 spin_lock_bh(&ifp2->lock);
1929 ifp2->stable_privacy_retry = retries;
1930 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1931 spin_unlock_bh(&ifp2->lock);
1933 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1934 in6_ifa_put(ifp2);
1935 lock_errdad:
1936 spin_lock_bh(&ifp->lock);
1937 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1938 addr.s6_addr32[0] = htonl(0xfe800000);
1939 addr.s6_addr32[1] = 0;
1941 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1942 ipv6_addr_equal(&ifp->addr, &addr)) {
1943 /* DAD failed for link-local based on MAC address */
1944 idev->cnf.disable_ipv6 = 1;
1946 pr_info("%s: IPv6 being disabled!\n",
1947 ifp->idev->dev->name);
1951 errdad:
1952 /* transition from _POSTDAD to _ERRDAD */
1953 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1954 spin_unlock_bh(&ifp->lock);
1956 addrconf_mod_dad_work(ifp, 0);
1959 /* Join to solicited addr multicast group.
1960 * caller must hold RTNL */
1961 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1963 struct in6_addr maddr;
1965 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1966 return;
1968 addrconf_addr_solict_mult(addr, &maddr);
1969 ipv6_dev_mc_inc(dev, &maddr);
1972 /* caller must hold RTNL */
1973 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1975 struct in6_addr maddr;
1977 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1978 return;
1980 addrconf_addr_solict_mult(addr, &maddr);
1981 __ipv6_dev_mc_dec(idev, &maddr);
1984 /* caller must hold RTNL */
1985 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1987 struct in6_addr addr;
1989 if (ifp->prefix_len >= 127) /* RFC 6164 */
1990 return;
1991 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1992 if (ipv6_addr_any(&addr))
1993 return;
1994 __ipv6_dev_ac_inc(ifp->idev, &addr);
1997 /* caller must hold RTNL */
1998 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2000 struct in6_addr addr;
2002 if (ifp->prefix_len >= 127) /* RFC 6164 */
2003 return;
2004 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2005 if (ipv6_addr_any(&addr))
2006 return;
2007 __ipv6_dev_ac_dec(ifp->idev, &addr);
2010 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2012 if (dev->addr_len != EUI64_ADDR_LEN)
2013 return -1;
2014 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2015 eui[0] ^= 2;
2016 return 0;
2019 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2021 union fwnet_hwaddr *ha;
2023 if (dev->addr_len != FWNET_ALEN)
2024 return -1;
2026 ha = (union fwnet_hwaddr *)dev->dev_addr;
2028 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2029 eui[0] ^= 2;
2030 return 0;
2033 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2035 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2036 if (dev->addr_len != ARCNET_ALEN)
2037 return -1;
2038 memset(eui, 0, 7);
2039 eui[7] = *(u8 *)dev->dev_addr;
2040 return 0;
2043 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2045 if (dev->addr_len != INFINIBAND_ALEN)
2046 return -1;
2047 memcpy(eui, dev->dev_addr + 12, 8);
2048 eui[0] |= 2;
2049 return 0;
2052 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2054 if (addr == 0)
2055 return -1;
2056 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2057 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2058 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2059 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2060 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2061 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2062 eui[1] = 0;
2063 eui[2] = 0x5E;
2064 eui[3] = 0xFE;
2065 memcpy(eui + 4, &addr, 4);
2066 return 0;
2069 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2071 if (dev->priv_flags & IFF_ISATAP)
2072 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2073 return -1;
2076 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2078 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2081 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2083 memcpy(eui, dev->perm_addr, 3);
2084 memcpy(eui + 5, dev->perm_addr + 3, 3);
2085 eui[3] = 0xFF;
2086 eui[4] = 0xFE;
2087 eui[0] ^= 2;
2088 return 0;
2091 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2093 switch (dev->type) {
2094 case ARPHRD_ETHER:
2095 case ARPHRD_FDDI:
2096 return addrconf_ifid_eui48(eui, dev);
2097 case ARPHRD_ARCNET:
2098 return addrconf_ifid_arcnet(eui, dev);
2099 case ARPHRD_INFINIBAND:
2100 return addrconf_ifid_infiniband(eui, dev);
2101 case ARPHRD_SIT:
2102 return addrconf_ifid_sit(eui, dev);
2103 case ARPHRD_IPGRE:
2104 return addrconf_ifid_gre(eui, dev);
2105 case ARPHRD_6LOWPAN:
2106 return addrconf_ifid_eui64(eui, dev);
2107 case ARPHRD_IEEE1394:
2108 return addrconf_ifid_ieee1394(eui, dev);
2109 case ARPHRD_TUNNEL6:
2110 return addrconf_ifid_ip6tnl(eui, dev);
2112 return -1;
2115 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2117 int err = -1;
2118 struct inet6_ifaddr *ifp;
2120 read_lock_bh(&idev->lock);
2121 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2122 if (ifp->scope > IFA_LINK)
2123 break;
2124 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2125 memcpy(eui, ifp->addr.s6_addr+8, 8);
2126 err = 0;
2127 break;
2130 read_unlock_bh(&idev->lock);
2131 return err;
2134 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2135 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2137 regen:
2138 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2139 idev->rndid[0] &= ~0x02;
2142 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2143 * check if generated address is not inappropriate
2145 * - Reserved subnet anycast (RFC 2526)
2146 * 11111101 11....11 1xxxxxxx
2147 * - ISATAP (RFC4214) 6.1
2148 * 00-00-5E-FE-xx-xx-xx-xx
2149 * - value 0
2150 * - XXX: already assigned to an address on the device
2152 if (idev->rndid[0] == 0xfd &&
2153 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2154 (idev->rndid[7]&0x80))
2155 goto regen;
2156 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2157 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2158 goto regen;
2159 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2160 goto regen;
2164 static void ipv6_regen_rndid(unsigned long data)
2166 struct inet6_dev *idev = (struct inet6_dev *) data;
2167 unsigned long expires;
2169 rcu_read_lock_bh();
2170 write_lock_bh(&idev->lock);
2172 if (idev->dead)
2173 goto out;
2175 __ipv6_regen_rndid(idev);
2177 expires = jiffies +
2178 idev->cnf.temp_prefered_lft * HZ -
2179 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2180 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2181 idev->cnf.max_desync_factor * HZ;
2182 if (time_before(expires, jiffies)) {
2183 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2184 __func__, idev->dev->name);
2185 goto out;
2188 if (!mod_timer(&idev->regen_timer, expires))
2189 in6_dev_hold(idev);
2191 out:
2192 write_unlock_bh(&idev->lock);
2193 rcu_read_unlock_bh();
2194 in6_dev_put(idev);
2197 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2199 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2200 __ipv6_regen_rndid(idev);
2204 * Add prefix route.
2207 static void
2208 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2209 unsigned long expires, u32 flags)
2211 struct fib6_config cfg = {
2212 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2213 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2214 .fc_ifindex = dev->ifindex,
2215 .fc_expires = expires,
2216 .fc_dst_len = plen,
2217 .fc_flags = RTF_UP | flags,
2218 .fc_nlinfo.nl_net = dev_net(dev),
2219 .fc_protocol = RTPROT_KERNEL,
2222 cfg.fc_dst = *pfx;
2224 /* Prevent useless cloning on PtP SIT.
2225 This thing is done here expecting that the whole
2226 class of non-broadcast devices need not cloning.
2228 #if IS_ENABLED(CONFIG_IPV6_SIT)
2229 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2230 cfg.fc_flags |= RTF_NONEXTHOP;
2231 #endif
2233 ip6_route_add(&cfg);
2237 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2238 int plen,
2239 const struct net_device *dev,
2240 u32 flags, u32 noflags)
2242 struct fib6_node *fn;
2243 struct rt6_info *rt = NULL;
2244 struct fib6_table *table;
2245 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2247 table = fib6_get_table(dev_net(dev), tb_id);
2248 if (!table)
2249 return NULL;
2251 read_lock_bh(&table->tb6_lock);
2252 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2253 if (!fn)
2254 goto out;
2256 noflags |= RTF_CACHE;
2257 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2258 if (rt->dst.dev->ifindex != dev->ifindex)
2259 continue;
2260 if ((rt->rt6i_flags & flags) != flags)
2261 continue;
2262 if ((rt->rt6i_flags & noflags) != 0)
2263 continue;
2264 dst_hold(&rt->dst);
2265 break;
2267 out:
2268 read_unlock_bh(&table->tb6_lock);
2269 return rt;
2273 /* Create "default" multicast route to the interface */
2275 static void addrconf_add_mroute(struct net_device *dev)
2277 struct fib6_config cfg = {
2278 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2279 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2280 .fc_ifindex = dev->ifindex,
2281 .fc_dst_len = 8,
2282 .fc_flags = RTF_UP,
2283 .fc_nlinfo.nl_net = dev_net(dev),
2286 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2288 ip6_route_add(&cfg);
2291 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2293 struct inet6_dev *idev;
2295 ASSERT_RTNL();
2297 idev = ipv6_find_idev(dev);
2298 if (!idev)
2299 return ERR_PTR(-ENOBUFS);
2301 if (idev->cnf.disable_ipv6)
2302 return ERR_PTR(-EACCES);
2304 /* Add default multicast route */
2305 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2306 addrconf_add_mroute(dev);
2308 return idev;
2311 static void manage_tempaddrs(struct inet6_dev *idev,
2312 struct inet6_ifaddr *ifp,
2313 __u32 valid_lft, __u32 prefered_lft,
2314 bool create, unsigned long now)
2316 u32 flags;
2317 struct inet6_ifaddr *ift;
2319 read_lock_bh(&idev->lock);
2320 /* update all temporary addresses in the list */
2321 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2322 int age, max_valid, max_prefered;
2324 if (ifp != ift->ifpub)
2325 continue;
2327 /* RFC 4941 section 3.3:
2328 * If a received option will extend the lifetime of a public
2329 * address, the lifetimes of temporary addresses should
2330 * be extended, subject to the overall constraint that no
2331 * temporary addresses should ever remain "valid" or "preferred"
2332 * for a time longer than (TEMP_VALID_LIFETIME) or
2333 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2335 age = (now - ift->cstamp) / HZ;
2336 max_valid = idev->cnf.temp_valid_lft - age;
2337 if (max_valid < 0)
2338 max_valid = 0;
2340 max_prefered = idev->cnf.temp_prefered_lft -
2341 idev->cnf.max_desync_factor - age;
2342 if (max_prefered < 0)
2343 max_prefered = 0;
2345 if (valid_lft > max_valid)
2346 valid_lft = max_valid;
2348 if (prefered_lft > max_prefered)
2349 prefered_lft = max_prefered;
2351 spin_lock(&ift->lock);
2352 flags = ift->flags;
2353 ift->valid_lft = valid_lft;
2354 ift->prefered_lft = prefered_lft;
2355 ift->tstamp = now;
2356 if (prefered_lft > 0)
2357 ift->flags &= ~IFA_F_DEPRECATED;
2359 spin_unlock(&ift->lock);
2360 if (!(flags&IFA_F_TENTATIVE))
2361 ipv6_ifa_notify(0, ift);
2364 if ((create || list_empty(&idev->tempaddr_list)) &&
2365 idev->cnf.use_tempaddr > 0) {
2366 /* When a new public address is created as described
2367 * in [ADDRCONF], also create a new temporary address.
2368 * Also create a temporary address if it's enabled but
2369 * no temporary address currently exists.
2371 read_unlock_bh(&idev->lock);
2372 ipv6_create_tempaddr(ifp, NULL);
2373 } else {
2374 read_unlock_bh(&idev->lock);
2378 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2380 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2381 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2384 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2385 const struct prefix_info *pinfo,
2386 struct inet6_dev *in6_dev,
2387 const struct in6_addr *addr, int addr_type,
2388 u32 addr_flags, bool sllao, bool tokenized,
2389 __u32 valid_lft, u32 prefered_lft)
2391 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2392 int create = 0, update_lft = 0;
2394 if (!ifp && valid_lft) {
2395 int max_addresses = in6_dev->cnf.max_addresses;
2397 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2398 if (in6_dev->cnf.optimistic_dad &&
2399 !net->ipv6.devconf_all->forwarding && sllao)
2400 addr_flags |= IFA_F_OPTIMISTIC;
2401 #endif
2403 /* Do not allow to create too much of autoconfigured
2404 * addresses; this would be too easy way to crash kernel.
2406 if (!max_addresses ||
2407 ipv6_count_addresses(in6_dev) < max_addresses)
2408 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2409 pinfo->prefix_len,
2410 addr_type&IPV6_ADDR_SCOPE_MASK,
2411 addr_flags, valid_lft,
2412 prefered_lft);
2414 if (IS_ERR_OR_NULL(ifp))
2415 return -1;
2417 update_lft = 0;
2418 create = 1;
2419 spin_lock_bh(&ifp->lock);
2420 ifp->flags |= IFA_F_MANAGETEMPADDR;
2421 ifp->cstamp = jiffies;
2422 ifp->tokenized = tokenized;
2423 spin_unlock_bh(&ifp->lock);
2424 addrconf_dad_start(ifp);
2427 if (ifp) {
2428 u32 flags;
2429 unsigned long now;
2430 u32 stored_lft;
2432 /* update lifetime (RFC2462 5.5.3 e) */
2433 spin_lock_bh(&ifp->lock);
2434 now = jiffies;
2435 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2436 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2437 else
2438 stored_lft = 0;
2439 if (!update_lft && !create && stored_lft) {
2440 const u32 minimum_lft = min_t(u32,
2441 stored_lft, MIN_VALID_LIFETIME);
2442 valid_lft = max(valid_lft, minimum_lft);
2444 /* RFC4862 Section 5.5.3e:
2445 * "Note that the preferred lifetime of the
2446 * corresponding address is always reset to
2447 * the Preferred Lifetime in the received
2448 * Prefix Information option, regardless of
2449 * whether the valid lifetime is also reset or
2450 * ignored."
2452 * So we should always update prefered_lft here.
2454 update_lft = 1;
2457 if (update_lft) {
2458 ifp->valid_lft = valid_lft;
2459 ifp->prefered_lft = prefered_lft;
2460 ifp->tstamp = now;
2461 flags = ifp->flags;
2462 ifp->flags &= ~IFA_F_DEPRECATED;
2463 spin_unlock_bh(&ifp->lock);
2465 if (!(flags&IFA_F_TENTATIVE))
2466 ipv6_ifa_notify(0, ifp);
2467 } else
2468 spin_unlock_bh(&ifp->lock);
2470 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2471 create, now);
2473 in6_ifa_put(ifp);
2474 addrconf_verify();
2477 return 0;
2479 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2481 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2483 struct prefix_info *pinfo;
2484 __u32 valid_lft;
2485 __u32 prefered_lft;
2486 int addr_type, err;
2487 u32 addr_flags = 0;
2488 struct inet6_dev *in6_dev;
2489 struct net *net = dev_net(dev);
2491 pinfo = (struct prefix_info *) opt;
2493 if (len < sizeof(struct prefix_info)) {
2494 ADBG("addrconf: prefix option too short\n");
2495 return;
2499 * Validation checks ([ADDRCONF], page 19)
2502 addr_type = ipv6_addr_type(&pinfo->prefix);
2504 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2505 return;
2507 valid_lft = ntohl(pinfo->valid);
2508 prefered_lft = ntohl(pinfo->prefered);
2510 if (prefered_lft > valid_lft) {
2511 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2512 return;
2515 in6_dev = in6_dev_get(dev);
2517 if (!in6_dev) {
2518 net_dbg_ratelimited("addrconf: device %s not configured\n",
2519 dev->name);
2520 return;
2524 * Two things going on here:
2525 * 1) Add routes for on-link prefixes
2526 * 2) Configure prefixes with the auto flag set
2529 if (pinfo->onlink) {
2530 struct rt6_info *rt;
2531 unsigned long rt_expires;
2533 /* Avoid arithmetic overflow. Really, we could
2534 * save rt_expires in seconds, likely valid_lft,
2535 * but it would require division in fib gc, that it
2536 * not good.
2538 if (HZ > USER_HZ)
2539 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2540 else
2541 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2543 if (addrconf_finite_timeout(rt_expires))
2544 rt_expires *= HZ;
2546 rt = addrconf_get_prefix_route(&pinfo->prefix,
2547 pinfo->prefix_len,
2548 dev,
2549 RTF_ADDRCONF | RTF_PREFIX_RT,
2550 RTF_GATEWAY | RTF_DEFAULT);
2552 if (rt) {
2553 /* Autoconf prefix route */
2554 if (valid_lft == 0) {
2555 ip6_del_rt(rt);
2556 rt = NULL;
2557 } else if (addrconf_finite_timeout(rt_expires)) {
2558 /* not infinity */
2559 rt6_set_expires(rt, jiffies + rt_expires);
2560 } else {
2561 rt6_clean_expires(rt);
2563 } else if (valid_lft) {
2564 clock_t expires = 0;
2565 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2566 if (addrconf_finite_timeout(rt_expires)) {
2567 /* not infinity */
2568 flags |= RTF_EXPIRES;
2569 expires = jiffies_to_clock_t(rt_expires);
2571 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2572 dev, expires, flags);
2574 ip6_rt_put(rt);
2577 /* Try to figure out our local address for this prefix */
2579 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2580 struct in6_addr addr;
2581 bool tokenized = false, dev_addr_generated = false;
2583 if (pinfo->prefix_len == 64) {
2584 memcpy(&addr, &pinfo->prefix, 8);
2586 if (!ipv6_addr_any(&in6_dev->token)) {
2587 read_lock_bh(&in6_dev->lock);
2588 memcpy(addr.s6_addr + 8,
2589 in6_dev->token.s6_addr + 8, 8);
2590 read_unlock_bh(&in6_dev->lock);
2591 tokenized = true;
2592 } else if (is_addr_mode_generate_stable(in6_dev) &&
2593 !ipv6_generate_stable_address(&addr, 0,
2594 in6_dev)) {
2595 addr_flags |= IFA_F_STABLE_PRIVACY;
2596 goto ok;
2597 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2598 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2599 goto put;
2600 } else {
2601 dev_addr_generated = true;
2603 goto ok;
2605 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2606 pinfo->prefix_len);
2607 goto put;
2610 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2611 &addr, addr_type,
2612 addr_flags, sllao,
2613 tokenized, valid_lft,
2614 prefered_lft);
2615 if (err)
2616 goto put;
2618 /* Ignore error case here because previous prefix add addr was
2619 * successful which will be notified.
2621 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2622 addr_type, addr_flags, sllao,
2623 tokenized, valid_lft,
2624 prefered_lft,
2625 dev_addr_generated);
2627 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2628 put:
2629 in6_dev_put(in6_dev);
2633 * Set destination address.
2634 * Special case for SIT interfaces where we create a new "virtual"
2635 * device.
2637 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2639 struct in6_ifreq ireq;
2640 struct net_device *dev;
2641 int err = -EINVAL;
2643 rtnl_lock();
2645 err = -EFAULT;
2646 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2647 goto err_exit;
2649 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2651 err = -ENODEV;
2652 if (!dev)
2653 goto err_exit;
2655 #if IS_ENABLED(CONFIG_IPV6_SIT)
2656 if (dev->type == ARPHRD_SIT) {
2657 const struct net_device_ops *ops = dev->netdev_ops;
2658 struct ifreq ifr;
2659 struct ip_tunnel_parm p;
2661 err = -EADDRNOTAVAIL;
2662 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2663 goto err_exit;
2665 memset(&p, 0, sizeof(p));
2666 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2667 p.iph.saddr = 0;
2668 p.iph.version = 4;
2669 p.iph.ihl = 5;
2670 p.iph.protocol = IPPROTO_IPV6;
2671 p.iph.ttl = 64;
2672 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2674 if (ops->ndo_do_ioctl) {
2675 mm_segment_t oldfs = get_fs();
2677 set_fs(KERNEL_DS);
2678 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2679 set_fs(oldfs);
2680 } else
2681 err = -EOPNOTSUPP;
2683 if (err == 0) {
2684 err = -ENOBUFS;
2685 dev = __dev_get_by_name(net, p.name);
2686 if (!dev)
2687 goto err_exit;
2688 err = dev_open(dev);
2691 #endif
2693 err_exit:
2694 rtnl_unlock();
2695 return err;
2698 static int ipv6_mc_config(struct sock *sk, bool join,
2699 const struct in6_addr *addr, int ifindex)
2701 int ret;
2703 ASSERT_RTNL();
2705 lock_sock(sk);
2706 if (join)
2707 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2708 else
2709 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2710 release_sock(sk);
2712 return ret;
2716 * Manual configuration of address on an interface
2718 static int inet6_addr_add(struct net *net, int ifindex,
2719 const struct in6_addr *pfx,
2720 const struct in6_addr *peer_pfx,
2721 unsigned int plen, __u32 ifa_flags,
2722 __u32 prefered_lft, __u32 valid_lft)
2724 struct inet6_ifaddr *ifp;
2725 struct inet6_dev *idev;
2726 struct net_device *dev;
2727 unsigned long timeout;
2728 clock_t expires;
2729 int scope;
2730 u32 flags;
2732 ASSERT_RTNL();
2734 if (plen > 128)
2735 return -EINVAL;
2737 /* check the lifetime */
2738 if (!valid_lft || prefered_lft > valid_lft)
2739 return -EINVAL;
2741 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2742 return -EINVAL;
2744 dev = __dev_get_by_index(net, ifindex);
2745 if (!dev)
2746 return -ENODEV;
2748 idev = addrconf_add_dev(dev);
2749 if (IS_ERR(idev))
2750 return PTR_ERR(idev);
2752 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2753 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2754 true, pfx, ifindex);
2756 if (ret < 0)
2757 return ret;
2760 scope = ipv6_addr_scope(pfx);
2762 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2763 if (addrconf_finite_timeout(timeout)) {
2764 expires = jiffies_to_clock_t(timeout * HZ);
2765 valid_lft = timeout;
2766 flags = RTF_EXPIRES;
2767 } else {
2768 expires = 0;
2769 flags = 0;
2770 ifa_flags |= IFA_F_PERMANENT;
2773 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2774 if (addrconf_finite_timeout(timeout)) {
2775 if (timeout == 0)
2776 ifa_flags |= IFA_F_DEPRECATED;
2777 prefered_lft = timeout;
2780 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2781 valid_lft, prefered_lft);
2783 if (!IS_ERR(ifp)) {
2784 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2785 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2786 expires, flags);
2790 * Note that section 3.1 of RFC 4429 indicates
2791 * that the Optimistic flag should not be set for
2792 * manually configured addresses
2794 addrconf_dad_start(ifp);
2795 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2796 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2797 true, jiffies);
2798 in6_ifa_put(ifp);
2799 addrconf_verify_rtnl();
2800 return 0;
2801 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2802 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2803 false, pfx, ifindex);
2806 return PTR_ERR(ifp);
2809 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2810 const struct in6_addr *pfx, unsigned int plen)
2812 struct inet6_ifaddr *ifp;
2813 struct inet6_dev *idev;
2814 struct net_device *dev;
2816 if (plen > 128)
2817 return -EINVAL;
2819 dev = __dev_get_by_index(net, ifindex);
2820 if (!dev)
2821 return -ENODEV;
2823 idev = __in6_dev_get(dev);
2824 if (!idev)
2825 return -ENXIO;
2827 read_lock_bh(&idev->lock);
2828 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2829 if (ifp->prefix_len == plen &&
2830 ipv6_addr_equal(pfx, &ifp->addr)) {
2831 in6_ifa_hold(ifp);
2832 read_unlock_bh(&idev->lock);
2834 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2835 (ifa_flags & IFA_F_MANAGETEMPADDR))
2836 manage_tempaddrs(idev, ifp, 0, 0, false,
2837 jiffies);
2838 ipv6_del_addr(ifp);
2839 addrconf_verify_rtnl();
2840 if (ipv6_addr_is_multicast(pfx)) {
2841 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2842 false, pfx, dev->ifindex);
2844 return 0;
2847 read_unlock_bh(&idev->lock);
2848 return -EADDRNOTAVAIL;
2852 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2854 struct in6_ifreq ireq;
2855 int err;
2857 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2858 return -EPERM;
2860 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2861 return -EFAULT;
2863 rtnl_lock();
2864 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2865 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2866 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2867 rtnl_unlock();
2868 return err;
2871 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2873 struct in6_ifreq ireq;
2874 int err;
2876 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2877 return -EPERM;
2879 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2880 return -EFAULT;
2882 rtnl_lock();
2883 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2884 ireq.ifr6_prefixlen);
2885 rtnl_unlock();
2886 return err;
2889 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2890 int plen, int scope)
2892 struct inet6_ifaddr *ifp;
2894 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2895 scope, IFA_F_PERMANENT,
2896 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2897 if (!IS_ERR(ifp)) {
2898 spin_lock_bh(&ifp->lock);
2899 ifp->flags &= ~IFA_F_TENTATIVE;
2900 spin_unlock_bh(&ifp->lock);
2901 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2902 in6_ifa_put(ifp);
2906 #if IS_ENABLED(CONFIG_IPV6_SIT)
2907 static void sit_add_v4_addrs(struct inet6_dev *idev)
2909 struct in6_addr addr;
2910 struct net_device *dev;
2911 struct net *net = dev_net(idev->dev);
2912 int scope, plen;
2913 u32 pflags = 0;
2915 ASSERT_RTNL();
2917 memset(&addr, 0, sizeof(struct in6_addr));
2918 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2920 if (idev->dev->flags&IFF_POINTOPOINT) {
2921 addr.s6_addr32[0] = htonl(0xfe800000);
2922 scope = IFA_LINK;
2923 plen = 64;
2924 } else {
2925 scope = IPV6_ADDR_COMPATv4;
2926 plen = 96;
2927 pflags |= RTF_NONEXTHOP;
2930 if (addr.s6_addr32[3]) {
2931 add_addr(idev, &addr, plen, scope);
2932 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2933 return;
2936 for_each_netdev(net, dev) {
2937 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2938 if (in_dev && (dev->flags & IFF_UP)) {
2939 struct in_ifaddr *ifa;
2941 int flag = scope;
2943 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2945 addr.s6_addr32[3] = ifa->ifa_local;
2947 if (ifa->ifa_scope == RT_SCOPE_LINK)
2948 continue;
2949 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2950 if (idev->dev->flags&IFF_POINTOPOINT)
2951 continue;
2952 flag |= IFA_HOST;
2955 add_addr(idev, &addr, plen, flag);
2956 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2957 pflags);
2962 #endif
2964 static void init_loopback(struct net_device *dev)
2966 struct inet6_dev *idev;
2967 struct net_device *sp_dev;
2968 struct inet6_ifaddr *sp_ifa;
2969 struct rt6_info *sp_rt;
2971 /* ::1 */
2973 ASSERT_RTNL();
2975 idev = ipv6_find_idev(dev);
2976 if (!idev) {
2977 pr_debug("%s: add_dev failed\n", __func__);
2978 return;
2981 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2983 /* Add routes to other interface's IPv6 addresses */
2984 for_each_netdev(dev_net(dev), sp_dev) {
2985 if (!strcmp(sp_dev->name, dev->name))
2986 continue;
2988 idev = __in6_dev_get(sp_dev);
2989 if (!idev)
2990 continue;
2992 read_lock_bh(&idev->lock);
2993 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2995 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2996 continue;
2998 if (sp_ifa->rt) {
2999 /* This dst has been added to garbage list when
3000 * lo device down, release this obsolete dst and
3001 * reallocate a new router for ifa.
3003 if (sp_ifa->rt->dst.obsolete > 0) {
3004 ip6_rt_put(sp_ifa->rt);
3005 sp_ifa->rt = NULL;
3006 } else {
3007 continue;
3011 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3013 /* Failure cases are ignored */
3014 if (!IS_ERR(sp_rt)) {
3015 sp_ifa->rt = sp_rt;
3016 ip6_ins_rt(sp_rt);
3019 read_unlock_bh(&idev->lock);
3023 void addrconf_add_linklocal(struct inet6_dev *idev,
3024 const struct in6_addr *addr, u32 flags)
3026 struct inet6_ifaddr *ifp;
3027 u32 addr_flags = flags | IFA_F_PERMANENT;
3029 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3030 if (idev->cnf.optimistic_dad &&
3031 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3032 addr_flags |= IFA_F_OPTIMISTIC;
3033 #endif
3035 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3036 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3037 if (!IS_ERR(ifp)) {
3038 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3039 addrconf_dad_start(ifp);
3040 in6_ifa_put(ifp);
3043 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3045 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3047 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3048 return true;
3050 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3051 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3052 return true;
3054 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3055 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3056 return true;
3058 return false;
3061 static int ipv6_generate_stable_address(struct in6_addr *address,
3062 u8 dad_count,
3063 const struct inet6_dev *idev)
3065 static DEFINE_SPINLOCK(lock);
3066 static __u32 digest[SHA_DIGEST_WORDS];
3067 static __u32 workspace[SHA_WORKSPACE_WORDS];
3069 static union {
3070 char __data[SHA_MESSAGE_BYTES];
3071 struct {
3072 struct in6_addr secret;
3073 __be32 prefix[2];
3074 unsigned char hwaddr[MAX_ADDR_LEN];
3075 u8 dad_count;
3076 } __packed;
3077 } data;
3079 struct in6_addr secret;
3080 struct in6_addr temp;
3081 struct net *net = dev_net(idev->dev);
3083 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3085 if (idev->cnf.stable_secret.initialized)
3086 secret = idev->cnf.stable_secret.secret;
3087 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3088 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3089 else
3090 return -1;
3092 retry:
3093 spin_lock_bh(&lock);
3095 sha_init(digest);
3096 memset(&data, 0, sizeof(data));
3097 memset(workspace, 0, sizeof(workspace));
3098 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3099 data.prefix[0] = address->s6_addr32[0];
3100 data.prefix[1] = address->s6_addr32[1];
3101 data.secret = secret;
3102 data.dad_count = dad_count;
3104 sha_transform(digest, data.__data, workspace);
3106 temp = *address;
3107 temp.s6_addr32[2] = (__force __be32)digest[0];
3108 temp.s6_addr32[3] = (__force __be32)digest[1];
3110 spin_unlock_bh(&lock);
3112 if (ipv6_reserved_interfaceid(temp)) {
3113 dad_count++;
3114 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3115 return -1;
3116 goto retry;
3119 *address = temp;
3120 return 0;
3123 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3125 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3127 if (s->initialized)
3128 return;
3129 s = &idev->cnf.stable_secret;
3130 get_random_bytes(&s->secret, sizeof(s->secret));
3131 s->initialized = true;
3134 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3136 struct in6_addr addr;
3138 /* no link local addresses on L3 master devices */
3139 if (netif_is_l3_master(idev->dev))
3140 return;
3142 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3144 switch (idev->addr_gen_mode) {
3145 case IN6_ADDR_GEN_MODE_RANDOM:
3146 ipv6_gen_mode_random_init(idev);
3147 /* fallthrough */
3148 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3149 if (!ipv6_generate_stable_address(&addr, 0, idev))
3150 addrconf_add_linklocal(idev, &addr,
3151 IFA_F_STABLE_PRIVACY);
3152 else if (prefix_route)
3153 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3154 break;
3155 case IN6_ADDR_GEN_MODE_EUI64:
3156 /* addrconf_add_linklocal also adds a prefix_route and we
3157 * only need to care about prefix routes if ipv6_generate_eui64
3158 * couldn't generate one.
3160 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3161 addrconf_add_linklocal(idev, &addr, 0);
3162 else if (prefix_route)
3163 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3164 break;
3165 case IN6_ADDR_GEN_MODE_NONE:
3166 default:
3167 /* will not add any link local address */
3168 break;
3172 static void addrconf_dev_config(struct net_device *dev)
3174 struct inet6_dev *idev;
3176 ASSERT_RTNL();
3178 if ((dev->type != ARPHRD_ETHER) &&
3179 (dev->type != ARPHRD_FDDI) &&
3180 (dev->type != ARPHRD_ARCNET) &&
3181 (dev->type != ARPHRD_INFINIBAND) &&
3182 (dev->type != ARPHRD_IEEE1394) &&
3183 (dev->type != ARPHRD_TUNNEL6) &&
3184 (dev->type != ARPHRD_6LOWPAN) &&
3185 (dev->type != ARPHRD_NONE)) {
3186 /* Alas, we support only Ethernet autoconfiguration. */
3187 return;
3190 idev = addrconf_add_dev(dev);
3191 if (IS_ERR(idev))
3192 return;
3194 /* this device type has no EUI support */
3195 if (dev->type == ARPHRD_NONE &&
3196 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3197 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3199 addrconf_addr_gen(idev, false);
3202 #if IS_ENABLED(CONFIG_IPV6_SIT)
3203 static void addrconf_sit_config(struct net_device *dev)
3205 struct inet6_dev *idev;
3207 ASSERT_RTNL();
3210 * Configure the tunnel with one of our IPv4
3211 * addresses... we should configure all of
3212 * our v4 addrs in the tunnel
3215 idev = ipv6_find_idev(dev);
3216 if (!idev) {
3217 pr_debug("%s: add_dev failed\n", __func__);
3218 return;
3221 if (dev->priv_flags & IFF_ISATAP) {
3222 addrconf_addr_gen(idev, false);
3223 return;
3226 sit_add_v4_addrs(idev);
3228 if (dev->flags&IFF_POINTOPOINT)
3229 addrconf_add_mroute(dev);
3231 #endif
3233 #if IS_ENABLED(CONFIG_NET_IPGRE)
3234 static void addrconf_gre_config(struct net_device *dev)
3236 struct inet6_dev *idev;
3238 ASSERT_RTNL();
3240 idev = ipv6_find_idev(dev);
3241 if (!idev) {
3242 pr_debug("%s: add_dev failed\n", __func__);
3243 return;
3246 addrconf_addr_gen(idev, true);
3247 if (dev->flags & IFF_POINTOPOINT)
3248 addrconf_add_mroute(dev);
3250 #endif
3252 static int fixup_permanent_addr(struct inet6_dev *idev,
3253 struct inet6_ifaddr *ifp)
3255 if (!ifp->rt) {
3256 struct rt6_info *rt;
3258 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3259 if (unlikely(IS_ERR(rt)))
3260 return PTR_ERR(rt);
3262 ifp->rt = rt;
3265 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3266 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3267 idev->dev, 0, 0);
3270 addrconf_dad_start(ifp);
3272 return 0;
3275 static void addrconf_permanent_addr(struct net_device *dev)
3277 struct inet6_ifaddr *ifp, *tmp;
3278 struct inet6_dev *idev;
3280 idev = __in6_dev_get(dev);
3281 if (!idev)
3282 return;
3284 write_lock_bh(&idev->lock);
3286 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3287 if ((ifp->flags & IFA_F_PERMANENT) &&
3288 fixup_permanent_addr(idev, ifp) < 0) {
3289 write_unlock_bh(&idev->lock);
3290 ipv6_del_addr(ifp);
3291 write_lock_bh(&idev->lock);
3293 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3294 idev->dev->name, &ifp->addr);
3298 write_unlock_bh(&idev->lock);
3301 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3302 void *ptr)
3304 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3305 struct netdev_notifier_changeupper_info *info;
3306 struct inet6_dev *idev = __in6_dev_get(dev);
3307 int run_pending = 0;
3308 int err;
3310 switch (event) {
3311 case NETDEV_REGISTER:
3312 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3313 idev = ipv6_add_dev(dev);
3314 if (IS_ERR(idev))
3315 return notifier_from_errno(PTR_ERR(idev));
3317 break;
3319 case NETDEV_CHANGEMTU:
3320 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3321 if (dev->mtu < IPV6_MIN_MTU) {
3322 addrconf_ifdown(dev, 1);
3323 break;
3326 if (idev) {
3327 rt6_mtu_change(dev, dev->mtu);
3328 idev->cnf.mtu6 = dev->mtu;
3329 break;
3332 /* allocate new idev */
3333 idev = ipv6_add_dev(dev);
3334 if (IS_ERR(idev))
3335 break;
3337 /* device is still not ready */
3338 if (!(idev->if_flags & IF_READY))
3339 break;
3341 run_pending = 1;
3343 /* fall through */
3345 case NETDEV_UP:
3346 case NETDEV_CHANGE:
3347 if (dev->flags & IFF_SLAVE)
3348 break;
3350 if (idev && idev->cnf.disable_ipv6)
3351 break;
3353 if (event == NETDEV_UP) {
3354 /* restore routes for permanent addresses */
3355 addrconf_permanent_addr(dev);
3357 if (!addrconf_qdisc_ok(dev)) {
3358 /* device is not ready yet. */
3359 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3360 dev->name);
3361 break;
3364 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3365 idev = ipv6_add_dev(dev);
3367 if (!IS_ERR_OR_NULL(idev)) {
3368 idev->if_flags |= IF_READY;
3369 run_pending = 1;
3371 } else if (event == NETDEV_CHANGE) {
3372 if (!addrconf_qdisc_ok(dev)) {
3373 /* device is still not ready. */
3374 break;
3377 if (idev) {
3378 if (idev->if_flags & IF_READY)
3379 /* device is already configured. */
3380 break;
3381 idev->if_flags |= IF_READY;
3384 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3385 dev->name);
3387 run_pending = 1;
3390 switch (dev->type) {
3391 #if IS_ENABLED(CONFIG_IPV6_SIT)
3392 case ARPHRD_SIT:
3393 addrconf_sit_config(dev);
3394 break;
3395 #endif
3396 #if IS_ENABLED(CONFIG_NET_IPGRE)
3397 case ARPHRD_IPGRE:
3398 addrconf_gre_config(dev);
3399 break;
3400 #endif
3401 case ARPHRD_LOOPBACK:
3402 init_loopback(dev);
3403 break;
3405 default:
3406 addrconf_dev_config(dev);
3407 break;
3410 if (!IS_ERR_OR_NULL(idev)) {
3411 if (run_pending)
3412 addrconf_dad_run(idev);
3415 * If the MTU changed during the interface down,
3416 * when the interface up, the changed MTU must be
3417 * reflected in the idev as well as routers.
3419 if (idev->cnf.mtu6 != dev->mtu &&
3420 dev->mtu >= IPV6_MIN_MTU) {
3421 rt6_mtu_change(dev, dev->mtu);
3422 idev->cnf.mtu6 = dev->mtu;
3424 idev->tstamp = jiffies;
3425 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3428 * If the changed mtu during down is lower than
3429 * IPV6_MIN_MTU stop IPv6 on this interface.
3431 if (dev->mtu < IPV6_MIN_MTU)
3432 addrconf_ifdown(dev, 1);
3434 break;
3436 case NETDEV_DOWN:
3437 case NETDEV_UNREGISTER:
3439 * Remove all addresses from this interface.
3441 addrconf_ifdown(dev, event != NETDEV_DOWN);
3442 break;
3444 case NETDEV_CHANGENAME:
3445 if (idev) {
3446 snmp6_unregister_dev(idev);
3447 addrconf_sysctl_unregister(idev);
3448 err = addrconf_sysctl_register(idev);
3449 if (err)
3450 return notifier_from_errno(err);
3451 err = snmp6_register_dev(idev);
3452 if (err) {
3453 addrconf_sysctl_unregister(idev);
3454 return notifier_from_errno(err);
3457 break;
3459 case NETDEV_PRE_TYPE_CHANGE:
3460 case NETDEV_POST_TYPE_CHANGE:
3461 if (idev)
3462 addrconf_type_change(dev, event);
3463 break;
3465 case NETDEV_CHANGEUPPER:
3466 info = ptr;
3468 /* flush all routes if dev is linked to or unlinked from
3469 * an L3 master device (e.g., VRF)
3471 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3472 addrconf_ifdown(dev, 0);
3475 return NOTIFY_OK;
3479 * addrconf module should be notified of a device going up
3481 static struct notifier_block ipv6_dev_notf = {
3482 .notifier_call = addrconf_notify,
3485 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3487 struct inet6_dev *idev;
3488 ASSERT_RTNL();
3490 idev = __in6_dev_get(dev);
3492 if (event == NETDEV_POST_TYPE_CHANGE)
3493 ipv6_mc_remap(idev);
3494 else if (event == NETDEV_PRE_TYPE_CHANGE)
3495 ipv6_mc_unmap(idev);
3498 static bool addr_is_local(const struct in6_addr *addr)
3500 return ipv6_addr_type(addr) &
3501 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3504 static int addrconf_ifdown(struct net_device *dev, int how)
3506 struct net *net = dev_net(dev);
3507 struct inet6_dev *idev;
3508 struct inet6_ifaddr *ifa, *tmp;
3509 struct list_head del_list;
3510 int _keep_addr;
3511 bool keep_addr;
3512 int state, i;
3514 ASSERT_RTNL();
3516 rt6_ifdown(net, dev);
3517 neigh_ifdown(&nd_tbl, dev);
3519 idev = __in6_dev_get(dev);
3520 if (!idev)
3521 return -ENODEV;
3524 * Step 1: remove reference to ipv6 device from parent device.
3525 * Do not dev_put!
3527 if (how) {
3528 idev->dead = 1;
3530 /* protected by rtnl_lock */
3531 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3533 /* Step 1.5: remove snmp6 entry */
3534 snmp6_unregister_dev(idev);
3538 /* aggregate the system setting and interface setting */
3539 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3540 if (!_keep_addr)
3541 _keep_addr = idev->cnf.keep_addr_on_down;
3543 /* combine the user config with event to determine if permanent
3544 * addresses are to be removed from address hash table
3546 keep_addr = !(how || _keep_addr <= 0);
3548 /* Step 2: clear hash table */
3549 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3550 struct hlist_head *h = &inet6_addr_lst[i];
3552 spin_lock_bh(&addrconf_hash_lock);
3553 restart:
3554 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3555 if (ifa->idev == idev) {
3556 addrconf_del_dad_work(ifa);
3557 /* combined flag + permanent flag decide if
3558 * address is retained on a down event
3560 if (!keep_addr ||
3561 !(ifa->flags & IFA_F_PERMANENT) ||
3562 addr_is_local(&ifa->addr)) {
3563 hlist_del_init_rcu(&ifa->addr_lst);
3564 goto restart;
3568 spin_unlock_bh(&addrconf_hash_lock);
3571 write_lock_bh(&idev->lock);
3573 addrconf_del_rs_timer(idev);
3575 /* Step 2: clear flags for stateless addrconf */
3576 if (!how)
3577 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3579 if (how && del_timer(&idev->regen_timer))
3580 in6_dev_put(idev);
3582 /* Step 3: clear tempaddr list */
3583 while (!list_empty(&idev->tempaddr_list)) {
3584 ifa = list_first_entry(&idev->tempaddr_list,
3585 struct inet6_ifaddr, tmp_list);
3586 list_del(&ifa->tmp_list);
3587 write_unlock_bh(&idev->lock);
3588 spin_lock_bh(&ifa->lock);
3590 if (ifa->ifpub) {
3591 in6_ifa_put(ifa->ifpub);
3592 ifa->ifpub = NULL;
3594 spin_unlock_bh(&ifa->lock);
3595 in6_ifa_put(ifa);
3596 write_lock_bh(&idev->lock);
3599 /* re-combine the user config with event to determine if permanent
3600 * addresses are to be removed from the interface list
3602 keep_addr = (!how && _keep_addr > 0);
3604 INIT_LIST_HEAD(&del_list);
3605 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3606 struct rt6_info *rt = NULL;
3608 addrconf_del_dad_work(ifa);
3610 write_unlock_bh(&idev->lock);
3611 spin_lock_bh(&ifa->lock);
3613 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3614 !addr_is_local(&ifa->addr)) {
3615 /* set state to skip the notifier below */
3616 state = INET6_IFADDR_STATE_DEAD;
3617 ifa->state = 0;
3618 if (!(ifa->flags & IFA_F_NODAD))
3619 ifa->flags |= IFA_F_TENTATIVE;
3621 rt = ifa->rt;
3622 ifa->rt = NULL;
3623 } else {
3624 state = ifa->state;
3625 ifa->state = INET6_IFADDR_STATE_DEAD;
3627 list_del(&ifa->if_list);
3628 list_add(&ifa->if_list, &del_list);
3631 spin_unlock_bh(&ifa->lock);
3633 if (rt)
3634 ip6_del_rt(rt);
3636 if (state != INET6_IFADDR_STATE_DEAD) {
3637 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3638 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3641 write_lock_bh(&idev->lock);
3644 write_unlock_bh(&idev->lock);
3646 /* now clean up addresses to be removed */
3647 while (!list_empty(&del_list)) {
3648 ifa = list_first_entry(&del_list,
3649 struct inet6_ifaddr, if_list);
3650 list_del(&ifa->if_list);
3652 in6_ifa_put(ifa);
3655 /* Step 5: Discard anycast and multicast list */
3656 if (how) {
3657 ipv6_ac_destroy_dev(idev);
3658 ipv6_mc_destroy_dev(idev);
3659 } else {
3660 ipv6_mc_down(idev);
3663 idev->tstamp = jiffies;
3665 /* Last: Shot the device (if unregistered) */
3666 if (how) {
3667 addrconf_sysctl_unregister(idev);
3668 neigh_parms_release(&nd_tbl, idev->nd_parms);
3669 neigh_ifdown(&nd_tbl, dev);
3670 in6_dev_put(idev);
3672 return 0;
3675 static void addrconf_rs_timer(unsigned long data)
3677 struct inet6_dev *idev = (struct inet6_dev *)data;
3678 struct net_device *dev = idev->dev;
3679 struct in6_addr lladdr;
3681 write_lock(&idev->lock);
3682 if (idev->dead || !(idev->if_flags & IF_READY))
3683 goto out;
3685 if (!ipv6_accept_ra(idev))
3686 goto out;
3688 /* Announcement received after solicitation was sent */
3689 if (idev->if_flags & IF_RA_RCVD)
3690 goto out;
3692 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3693 write_unlock(&idev->lock);
3694 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3695 ndisc_send_rs(dev, &lladdr,
3696 &in6addr_linklocal_allrouters);
3697 else
3698 goto put;
3700 write_lock(&idev->lock);
3701 /* The wait after the last probe can be shorter */
3702 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3703 idev->cnf.rtr_solicits) ?
3704 idev->cnf.rtr_solicit_delay :
3705 idev->cnf.rtr_solicit_interval);
3706 } else {
3708 * Note: we do not support deprecated "all on-link"
3709 * assumption any longer.
3711 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3714 out:
3715 write_unlock(&idev->lock);
3716 put:
3717 in6_dev_put(idev);
3721 * Duplicate Address Detection
3723 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3725 unsigned long rand_num;
3726 struct inet6_dev *idev = ifp->idev;
3728 if (ifp->flags & IFA_F_OPTIMISTIC)
3729 rand_num = 0;
3730 else
3731 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3733 ifp->dad_probes = idev->cnf.dad_transmits;
3734 addrconf_mod_dad_work(ifp, rand_num);
3737 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3739 struct inet6_dev *idev = ifp->idev;
3740 struct net_device *dev = idev->dev;
3741 bool notify = false;
3743 addrconf_join_solict(dev, &ifp->addr);
3745 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3747 read_lock_bh(&idev->lock);
3748 spin_lock(&ifp->lock);
3749 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3750 goto out;
3752 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3753 idev->cnf.accept_dad < 1 ||
3754 !(ifp->flags&IFA_F_TENTATIVE) ||
3755 ifp->flags & IFA_F_NODAD) {
3756 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3757 spin_unlock(&ifp->lock);
3758 read_unlock_bh(&idev->lock);
3760 addrconf_dad_completed(ifp);
3761 return;
3764 if (!(idev->if_flags & IF_READY)) {
3765 spin_unlock(&ifp->lock);
3766 read_unlock_bh(&idev->lock);
3768 * If the device is not ready:
3769 * - keep it tentative if it is a permanent address.
3770 * - otherwise, kill it.
3772 in6_ifa_hold(ifp);
3773 addrconf_dad_stop(ifp, 0);
3774 return;
3778 * Optimistic nodes can start receiving
3779 * Frames right away
3781 if (ifp->flags & IFA_F_OPTIMISTIC) {
3782 ip6_ins_rt(ifp->rt);
3783 if (ipv6_use_optimistic_addr(idev)) {
3784 /* Because optimistic nodes can use this address,
3785 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3787 notify = true;
3791 addrconf_dad_kick(ifp);
3792 out:
3793 spin_unlock(&ifp->lock);
3794 read_unlock_bh(&idev->lock);
3795 if (notify)
3796 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3799 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3801 bool begin_dad = false;
3803 spin_lock_bh(&ifp->lock);
3804 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3805 ifp->state = INET6_IFADDR_STATE_PREDAD;
3806 begin_dad = true;
3808 spin_unlock_bh(&ifp->lock);
3810 if (begin_dad)
3811 addrconf_mod_dad_work(ifp, 0);
3814 static void addrconf_dad_work(struct work_struct *w)
3816 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3817 struct inet6_ifaddr,
3818 dad_work);
3819 struct inet6_dev *idev = ifp->idev;
3820 struct in6_addr mcaddr;
3822 enum {
3823 DAD_PROCESS,
3824 DAD_BEGIN,
3825 DAD_ABORT,
3826 } action = DAD_PROCESS;
3828 rtnl_lock();
3830 spin_lock_bh(&ifp->lock);
3831 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3832 action = DAD_BEGIN;
3833 ifp->state = INET6_IFADDR_STATE_DAD;
3834 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3835 action = DAD_ABORT;
3836 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3838 spin_unlock_bh(&ifp->lock);
3840 if (action == DAD_BEGIN) {
3841 addrconf_dad_begin(ifp);
3842 goto out;
3843 } else if (action == DAD_ABORT) {
3844 addrconf_dad_stop(ifp, 1);
3845 goto out;
3848 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3849 goto out;
3851 write_lock_bh(&idev->lock);
3852 if (idev->dead || !(idev->if_flags & IF_READY)) {
3853 write_unlock_bh(&idev->lock);
3854 goto out;
3857 spin_lock(&ifp->lock);
3858 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3859 spin_unlock(&ifp->lock);
3860 write_unlock_bh(&idev->lock);
3861 goto out;
3864 if (ifp->dad_probes == 0) {
3866 * DAD was successful
3869 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3870 spin_unlock(&ifp->lock);
3871 write_unlock_bh(&idev->lock);
3873 addrconf_dad_completed(ifp);
3875 goto out;
3878 ifp->dad_probes--;
3879 addrconf_mod_dad_work(ifp,
3880 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3881 spin_unlock(&ifp->lock);
3882 write_unlock_bh(&idev->lock);
3884 /* send a neighbour solicitation for our addr */
3885 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3886 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3887 out:
3888 in6_ifa_put(ifp);
3889 rtnl_unlock();
3892 /* ifp->idev must be at least read locked */
3893 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3895 struct inet6_ifaddr *ifpiter;
3896 struct inet6_dev *idev = ifp->idev;
3898 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3899 if (ifpiter->scope > IFA_LINK)
3900 break;
3901 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3902 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3903 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3904 IFA_F_PERMANENT)
3905 return false;
3907 return true;
3910 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3912 struct net_device *dev = ifp->idev->dev;
3913 struct in6_addr lladdr;
3914 bool send_rs, send_mld;
3916 addrconf_del_dad_work(ifp);
3919 * Configure the address for reception. Now it is valid.
3922 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3924 /* If added prefix is link local and we are prepared to process
3925 router advertisements, start sending router solicitations.
3928 read_lock_bh(&ifp->idev->lock);
3929 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3930 send_rs = send_mld &&
3931 ipv6_accept_ra(ifp->idev) &&
3932 ifp->idev->cnf.rtr_solicits > 0 &&
3933 (dev->flags&IFF_LOOPBACK) == 0;
3934 read_unlock_bh(&ifp->idev->lock);
3936 /* While dad is in progress mld report's source address is in6_addrany.
3937 * Resend with proper ll now.
3939 if (send_mld)
3940 ipv6_mc_dad_complete(ifp->idev);
3942 if (send_rs) {
3944 * If a host as already performed a random delay
3945 * [...] as part of DAD [...] there is no need
3946 * to delay again before sending the first RS
3948 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3949 return;
3950 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3952 write_lock_bh(&ifp->idev->lock);
3953 spin_lock(&ifp->lock);
3954 ifp->idev->rs_probes = 1;
3955 ifp->idev->if_flags |= IF_RS_SENT;
3956 addrconf_mod_rs_timer(ifp->idev,
3957 ifp->idev->cnf.rtr_solicit_interval);
3958 spin_unlock(&ifp->lock);
3959 write_unlock_bh(&ifp->idev->lock);
3963 static void addrconf_dad_run(struct inet6_dev *idev)
3965 struct inet6_ifaddr *ifp;
3967 read_lock_bh(&idev->lock);
3968 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3969 spin_lock(&ifp->lock);
3970 if (ifp->flags & IFA_F_TENTATIVE &&
3971 ifp->state == INET6_IFADDR_STATE_DAD)
3972 addrconf_dad_kick(ifp);
3973 spin_unlock(&ifp->lock);
3975 read_unlock_bh(&idev->lock);
3978 #ifdef CONFIG_PROC_FS
3979 struct if6_iter_state {
3980 struct seq_net_private p;
3981 int bucket;
3982 int offset;
3985 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3987 struct inet6_ifaddr *ifa = NULL;
3988 struct if6_iter_state *state = seq->private;
3989 struct net *net = seq_file_net(seq);
3990 int p = 0;
3992 /* initial bucket if pos is 0 */
3993 if (pos == 0) {
3994 state->bucket = 0;
3995 state->offset = 0;
3998 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3999 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4000 addr_lst) {
4001 if (!net_eq(dev_net(ifa->idev->dev), net))
4002 continue;
4003 /* sync with offset */
4004 if (p < state->offset) {
4005 p++;
4006 continue;
4008 state->offset++;
4009 return ifa;
4012 /* prepare for next bucket */
4013 state->offset = 0;
4014 p = 0;
4016 return NULL;
4019 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4020 struct inet6_ifaddr *ifa)
4022 struct if6_iter_state *state = seq->private;
4023 struct net *net = seq_file_net(seq);
4025 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4026 if (!net_eq(dev_net(ifa->idev->dev), net))
4027 continue;
4028 state->offset++;
4029 return ifa;
4032 while (++state->bucket < IN6_ADDR_HSIZE) {
4033 state->offset = 0;
4034 hlist_for_each_entry_rcu_bh(ifa,
4035 &inet6_addr_lst[state->bucket], addr_lst) {
4036 if (!net_eq(dev_net(ifa->idev->dev), net))
4037 continue;
4038 state->offset++;
4039 return ifa;
4043 return NULL;
4046 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4047 __acquires(rcu_bh)
4049 rcu_read_lock_bh();
4050 return if6_get_first(seq, *pos);
4053 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4055 struct inet6_ifaddr *ifa;
4057 ifa = if6_get_next(seq, v);
4058 ++*pos;
4059 return ifa;
4062 static void if6_seq_stop(struct seq_file *seq, void *v)
4063 __releases(rcu_bh)
4065 rcu_read_unlock_bh();
4068 static int if6_seq_show(struct seq_file *seq, void *v)
4070 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4071 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4072 &ifp->addr,
4073 ifp->idev->dev->ifindex,
4074 ifp->prefix_len,
4075 ifp->scope,
4076 (u8) ifp->flags,
4077 ifp->idev->dev->name);
4078 return 0;
4081 static const struct seq_operations if6_seq_ops = {
4082 .start = if6_seq_start,
4083 .next = if6_seq_next,
4084 .show = if6_seq_show,
4085 .stop = if6_seq_stop,
4088 static int if6_seq_open(struct inode *inode, struct file *file)
4090 return seq_open_net(inode, file, &if6_seq_ops,
4091 sizeof(struct if6_iter_state));
4094 static const struct file_operations if6_fops = {
4095 .owner = THIS_MODULE,
4096 .open = if6_seq_open,
4097 .read = seq_read,
4098 .llseek = seq_lseek,
4099 .release = seq_release_net,
4102 static int __net_init if6_proc_net_init(struct net *net)
4104 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4105 return -ENOMEM;
4106 return 0;
4109 static void __net_exit if6_proc_net_exit(struct net *net)
4111 remove_proc_entry("if_inet6", net->proc_net);
4114 static struct pernet_operations if6_proc_net_ops = {
4115 .init = if6_proc_net_init,
4116 .exit = if6_proc_net_exit,
4119 int __init if6_proc_init(void)
4121 return register_pernet_subsys(&if6_proc_net_ops);
4124 void if6_proc_exit(void)
4126 unregister_pernet_subsys(&if6_proc_net_ops);
4128 #endif /* CONFIG_PROC_FS */
4130 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4131 /* Check if address is a home address configured on any interface. */
4132 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4134 int ret = 0;
4135 struct inet6_ifaddr *ifp = NULL;
4136 unsigned int hash = inet6_addr_hash(addr);
4138 rcu_read_lock_bh();
4139 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4140 if (!net_eq(dev_net(ifp->idev->dev), net))
4141 continue;
4142 if (ipv6_addr_equal(&ifp->addr, addr) &&
4143 (ifp->flags & IFA_F_HOMEADDRESS)) {
4144 ret = 1;
4145 break;
4148 rcu_read_unlock_bh();
4149 return ret;
4151 #endif
4154 * Periodic address status verification
4157 static void addrconf_verify_rtnl(void)
4159 unsigned long now, next, next_sec, next_sched;
4160 struct inet6_ifaddr *ifp;
4161 int i;
4163 ASSERT_RTNL();
4165 rcu_read_lock_bh();
4166 now = jiffies;
4167 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4169 cancel_delayed_work(&addr_chk_work);
4171 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4172 restart:
4173 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4174 unsigned long age;
4176 /* When setting preferred_lft to a value not zero or
4177 * infinity, while valid_lft is infinity
4178 * IFA_F_PERMANENT has a non-infinity life time.
4180 if ((ifp->flags & IFA_F_PERMANENT) &&
4181 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4182 continue;
4184 spin_lock(&ifp->lock);
4185 /* We try to batch several events at once. */
4186 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4188 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4189 age >= ifp->valid_lft) {
4190 spin_unlock(&ifp->lock);
4191 in6_ifa_hold(ifp);
4192 ipv6_del_addr(ifp);
4193 goto restart;
4194 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4195 spin_unlock(&ifp->lock);
4196 continue;
4197 } else if (age >= ifp->prefered_lft) {
4198 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4199 int deprecate = 0;
4201 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4202 deprecate = 1;
4203 ifp->flags |= IFA_F_DEPRECATED;
4206 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4207 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4208 next = ifp->tstamp + ifp->valid_lft * HZ;
4210 spin_unlock(&ifp->lock);
4212 if (deprecate) {
4213 in6_ifa_hold(ifp);
4215 ipv6_ifa_notify(0, ifp);
4216 in6_ifa_put(ifp);
4217 goto restart;
4219 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4220 !(ifp->flags&IFA_F_TENTATIVE)) {
4221 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4222 ifp->idev->cnf.dad_transmits *
4223 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4225 if (age >= ifp->prefered_lft - regen_advance) {
4226 struct inet6_ifaddr *ifpub = ifp->ifpub;
4227 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4228 next = ifp->tstamp + ifp->prefered_lft * HZ;
4229 if (!ifp->regen_count && ifpub) {
4230 ifp->regen_count++;
4231 in6_ifa_hold(ifp);
4232 in6_ifa_hold(ifpub);
4233 spin_unlock(&ifp->lock);
4235 spin_lock(&ifpub->lock);
4236 ifpub->regen_count = 0;
4237 spin_unlock(&ifpub->lock);
4238 ipv6_create_tempaddr(ifpub, ifp);
4239 in6_ifa_put(ifpub);
4240 in6_ifa_put(ifp);
4241 goto restart;
4243 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4244 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4245 spin_unlock(&ifp->lock);
4246 } else {
4247 /* ifp->prefered_lft <= ifp->valid_lft */
4248 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4249 next = ifp->tstamp + ifp->prefered_lft * HZ;
4250 spin_unlock(&ifp->lock);
4255 next_sec = round_jiffies_up(next);
4256 next_sched = next;
4258 /* If rounded timeout is accurate enough, accept it. */
4259 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4260 next_sched = next_sec;
4262 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4263 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4264 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4266 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4267 now, next, next_sec, next_sched);
4268 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4269 rcu_read_unlock_bh();
4272 static void addrconf_verify_work(struct work_struct *w)
4274 rtnl_lock();
4275 addrconf_verify_rtnl();
4276 rtnl_unlock();
4279 static void addrconf_verify(void)
4281 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4284 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4285 struct in6_addr **peer_pfx)
4287 struct in6_addr *pfx = NULL;
4289 *peer_pfx = NULL;
4291 if (addr)
4292 pfx = nla_data(addr);
4294 if (local) {
4295 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4296 *peer_pfx = pfx;
4297 pfx = nla_data(local);
4300 return pfx;
4303 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4304 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4305 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4306 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4307 [IFA_FLAGS] = { .len = sizeof(u32) },
4310 static int
4311 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4313 struct net *net = sock_net(skb->sk);
4314 struct ifaddrmsg *ifm;
4315 struct nlattr *tb[IFA_MAX+1];
4316 struct in6_addr *pfx, *peer_pfx;
4317 u32 ifa_flags;
4318 int err;
4320 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4321 if (err < 0)
4322 return err;
4324 ifm = nlmsg_data(nlh);
4325 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4326 if (!pfx)
4327 return -EINVAL;
4329 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4331 /* We ignore other flags so far. */
4332 ifa_flags &= IFA_F_MANAGETEMPADDR;
4334 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4335 ifm->ifa_prefixlen);
4338 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4339 u32 prefered_lft, u32 valid_lft)
4341 u32 flags;
4342 clock_t expires;
4343 unsigned long timeout;
4344 bool was_managetempaddr;
4345 bool had_prefixroute;
4347 ASSERT_RTNL();
4349 if (!valid_lft || (prefered_lft > valid_lft))
4350 return -EINVAL;
4352 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4353 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4354 return -EINVAL;
4356 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4357 if (addrconf_finite_timeout(timeout)) {
4358 expires = jiffies_to_clock_t(timeout * HZ);
4359 valid_lft = timeout;
4360 flags = RTF_EXPIRES;
4361 } else {
4362 expires = 0;
4363 flags = 0;
4364 ifa_flags |= IFA_F_PERMANENT;
4367 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4368 if (addrconf_finite_timeout(timeout)) {
4369 if (timeout == 0)
4370 ifa_flags |= IFA_F_DEPRECATED;
4371 prefered_lft = timeout;
4374 spin_lock_bh(&ifp->lock);
4375 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4376 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4377 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4378 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4379 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4380 IFA_F_NOPREFIXROUTE);
4381 ifp->flags |= ifa_flags;
4382 ifp->tstamp = jiffies;
4383 ifp->valid_lft = valid_lft;
4384 ifp->prefered_lft = prefered_lft;
4386 spin_unlock_bh(&ifp->lock);
4387 if (!(ifp->flags&IFA_F_TENTATIVE))
4388 ipv6_ifa_notify(0, ifp);
4390 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4391 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4392 expires, flags);
4393 } else if (had_prefixroute) {
4394 enum cleanup_prefix_rt_t action;
4395 unsigned long rt_expires;
4397 write_lock_bh(&ifp->idev->lock);
4398 action = check_cleanup_prefix_route(ifp, &rt_expires);
4399 write_unlock_bh(&ifp->idev->lock);
4401 if (action != CLEANUP_PREFIX_RT_NOP) {
4402 cleanup_prefix_route(ifp, rt_expires,
4403 action == CLEANUP_PREFIX_RT_DEL);
4407 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4408 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4409 valid_lft = prefered_lft = 0;
4410 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4411 !was_managetempaddr, jiffies);
4414 addrconf_verify_rtnl();
4416 return 0;
4419 static int
4420 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4422 struct net *net = sock_net(skb->sk);
4423 struct ifaddrmsg *ifm;
4424 struct nlattr *tb[IFA_MAX+1];
4425 struct in6_addr *pfx, *peer_pfx;
4426 struct inet6_ifaddr *ifa;
4427 struct net_device *dev;
4428 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4429 u32 ifa_flags;
4430 int err;
4432 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4433 if (err < 0)
4434 return err;
4436 ifm = nlmsg_data(nlh);
4437 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4438 if (!pfx)
4439 return -EINVAL;
4441 if (tb[IFA_CACHEINFO]) {
4442 struct ifa_cacheinfo *ci;
4444 ci = nla_data(tb[IFA_CACHEINFO]);
4445 valid_lft = ci->ifa_valid;
4446 preferred_lft = ci->ifa_prefered;
4447 } else {
4448 preferred_lft = INFINITY_LIFE_TIME;
4449 valid_lft = INFINITY_LIFE_TIME;
4452 dev = __dev_get_by_index(net, ifm->ifa_index);
4453 if (!dev)
4454 return -ENODEV;
4456 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4458 /* We ignore other flags so far. */
4459 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4460 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4462 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4463 if (!ifa) {
4465 * It would be best to check for !NLM_F_CREATE here but
4466 * userspace already relies on not having to provide this.
4468 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4469 ifm->ifa_prefixlen, ifa_flags,
4470 preferred_lft, valid_lft);
4473 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4474 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4475 err = -EEXIST;
4476 else
4477 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4479 in6_ifa_put(ifa);
4481 return err;
4484 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4485 u8 scope, int ifindex)
4487 struct ifaddrmsg *ifm;
4489 ifm = nlmsg_data(nlh);
4490 ifm->ifa_family = AF_INET6;
4491 ifm->ifa_prefixlen = prefixlen;
4492 ifm->ifa_flags = flags;
4493 ifm->ifa_scope = scope;
4494 ifm->ifa_index = ifindex;
4497 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4498 unsigned long tstamp, u32 preferred, u32 valid)
4500 struct ifa_cacheinfo ci;
4502 ci.cstamp = cstamp_delta(cstamp);
4503 ci.tstamp = cstamp_delta(tstamp);
4504 ci.ifa_prefered = preferred;
4505 ci.ifa_valid = valid;
4507 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4510 static inline int rt_scope(int ifa_scope)
4512 if (ifa_scope & IFA_HOST)
4513 return RT_SCOPE_HOST;
4514 else if (ifa_scope & IFA_LINK)
4515 return RT_SCOPE_LINK;
4516 else if (ifa_scope & IFA_SITE)
4517 return RT_SCOPE_SITE;
4518 else
4519 return RT_SCOPE_UNIVERSE;
4522 static inline int inet6_ifaddr_msgsize(void)
4524 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4525 + nla_total_size(16) /* IFA_LOCAL */
4526 + nla_total_size(16) /* IFA_ADDRESS */
4527 + nla_total_size(sizeof(struct ifa_cacheinfo))
4528 + nla_total_size(4) /* IFA_FLAGS */;
4531 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4532 u32 portid, u32 seq, int event, unsigned int flags)
4534 struct nlmsghdr *nlh;
4535 u32 preferred, valid;
4537 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4538 if (!nlh)
4539 return -EMSGSIZE;
4541 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4542 ifa->idev->dev->ifindex);
4544 if (!((ifa->flags&IFA_F_PERMANENT) &&
4545 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4546 preferred = ifa->prefered_lft;
4547 valid = ifa->valid_lft;
4548 if (preferred != INFINITY_LIFE_TIME) {
4549 long tval = (jiffies - ifa->tstamp)/HZ;
4550 if (preferred > tval)
4551 preferred -= tval;
4552 else
4553 preferred = 0;
4554 if (valid != INFINITY_LIFE_TIME) {
4555 if (valid > tval)
4556 valid -= tval;
4557 else
4558 valid = 0;
4561 } else {
4562 preferred = INFINITY_LIFE_TIME;
4563 valid = INFINITY_LIFE_TIME;
4566 if (!ipv6_addr_any(&ifa->peer_addr)) {
4567 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4568 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4569 goto error;
4570 } else
4571 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4572 goto error;
4574 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4575 goto error;
4577 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4578 goto error;
4580 nlmsg_end(skb, nlh);
4581 return 0;
4583 error:
4584 nlmsg_cancel(skb, nlh);
4585 return -EMSGSIZE;
4588 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4589 u32 portid, u32 seq, int event, u16 flags)
4591 struct nlmsghdr *nlh;
4592 u8 scope = RT_SCOPE_UNIVERSE;
4593 int ifindex = ifmca->idev->dev->ifindex;
4595 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4596 scope = RT_SCOPE_SITE;
4598 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4599 if (!nlh)
4600 return -EMSGSIZE;
4602 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4603 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4604 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4605 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4606 nlmsg_cancel(skb, nlh);
4607 return -EMSGSIZE;
4610 nlmsg_end(skb, nlh);
4611 return 0;
4614 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4615 u32 portid, u32 seq, int event, unsigned int flags)
4617 struct nlmsghdr *nlh;
4618 u8 scope = RT_SCOPE_UNIVERSE;
4619 int ifindex = ifaca->aca_idev->dev->ifindex;
4621 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4622 scope = RT_SCOPE_SITE;
4624 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4625 if (!nlh)
4626 return -EMSGSIZE;
4628 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4629 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4630 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4631 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4632 nlmsg_cancel(skb, nlh);
4633 return -EMSGSIZE;
4636 nlmsg_end(skb, nlh);
4637 return 0;
4640 enum addr_type_t {
4641 UNICAST_ADDR,
4642 MULTICAST_ADDR,
4643 ANYCAST_ADDR,
4646 /* called with rcu_read_lock() */
4647 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4648 struct netlink_callback *cb, enum addr_type_t type,
4649 int s_ip_idx, int *p_ip_idx)
4651 struct ifmcaddr6 *ifmca;
4652 struct ifacaddr6 *ifaca;
4653 int err = 1;
4654 int ip_idx = *p_ip_idx;
4656 read_lock_bh(&idev->lock);
4657 switch (type) {
4658 case UNICAST_ADDR: {
4659 struct inet6_ifaddr *ifa;
4661 /* unicast address incl. temp addr */
4662 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4663 if (++ip_idx < s_ip_idx)
4664 continue;
4665 err = inet6_fill_ifaddr(skb, ifa,
4666 NETLINK_CB(cb->skb).portid,
4667 cb->nlh->nlmsg_seq,
4668 RTM_NEWADDR,
4669 NLM_F_MULTI);
4670 if (err < 0)
4671 break;
4672 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4674 break;
4676 case MULTICAST_ADDR:
4677 /* multicast address */
4678 for (ifmca = idev->mc_list; ifmca;
4679 ifmca = ifmca->next, ip_idx++) {
4680 if (ip_idx < s_ip_idx)
4681 continue;
4682 err = inet6_fill_ifmcaddr(skb, ifmca,
4683 NETLINK_CB(cb->skb).portid,
4684 cb->nlh->nlmsg_seq,
4685 RTM_GETMULTICAST,
4686 NLM_F_MULTI);
4687 if (err < 0)
4688 break;
4690 break;
4691 case ANYCAST_ADDR:
4692 /* anycast address */
4693 for (ifaca = idev->ac_list; ifaca;
4694 ifaca = ifaca->aca_next, ip_idx++) {
4695 if (ip_idx < s_ip_idx)
4696 continue;
4697 err = inet6_fill_ifacaddr(skb, ifaca,
4698 NETLINK_CB(cb->skb).portid,
4699 cb->nlh->nlmsg_seq,
4700 RTM_GETANYCAST,
4701 NLM_F_MULTI);
4702 if (err < 0)
4703 break;
4705 break;
4706 default:
4707 break;
4709 read_unlock_bh(&idev->lock);
4710 *p_ip_idx = ip_idx;
4711 return err;
4714 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4715 enum addr_type_t type)
4717 struct net *net = sock_net(skb->sk);
4718 int h, s_h;
4719 int idx, ip_idx;
4720 int s_idx, s_ip_idx;
4721 struct net_device *dev;
4722 struct inet6_dev *idev;
4723 struct hlist_head *head;
4725 s_h = cb->args[0];
4726 s_idx = idx = cb->args[1];
4727 s_ip_idx = ip_idx = cb->args[2];
4729 rcu_read_lock();
4730 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4731 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4732 idx = 0;
4733 head = &net->dev_index_head[h];
4734 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4735 if (idx < s_idx)
4736 goto cont;
4737 if (h > s_h || idx > s_idx)
4738 s_ip_idx = 0;
4739 ip_idx = 0;
4740 idev = __in6_dev_get(dev);
4741 if (!idev)
4742 goto cont;
4744 if (in6_dump_addrs(idev, skb, cb, type,
4745 s_ip_idx, &ip_idx) < 0)
4746 goto done;
4747 cont:
4748 idx++;
4751 done:
4752 rcu_read_unlock();
4753 cb->args[0] = h;
4754 cb->args[1] = idx;
4755 cb->args[2] = ip_idx;
4757 return skb->len;
4760 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4762 enum addr_type_t type = UNICAST_ADDR;
4764 return inet6_dump_addr(skb, cb, type);
4767 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4769 enum addr_type_t type = MULTICAST_ADDR;
4771 return inet6_dump_addr(skb, cb, type);
4775 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4777 enum addr_type_t type = ANYCAST_ADDR;
4779 return inet6_dump_addr(skb, cb, type);
4782 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4784 struct net *net = sock_net(in_skb->sk);
4785 struct ifaddrmsg *ifm;
4786 struct nlattr *tb[IFA_MAX+1];
4787 struct in6_addr *addr = NULL, *peer;
4788 struct net_device *dev = NULL;
4789 struct inet6_ifaddr *ifa;
4790 struct sk_buff *skb;
4791 int err;
4793 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4794 if (err < 0)
4795 goto errout;
4797 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4798 if (!addr) {
4799 err = -EINVAL;
4800 goto errout;
4803 ifm = nlmsg_data(nlh);
4804 if (ifm->ifa_index)
4805 dev = __dev_get_by_index(net, ifm->ifa_index);
4807 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4808 if (!ifa) {
4809 err = -EADDRNOTAVAIL;
4810 goto errout;
4813 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4814 if (!skb) {
4815 err = -ENOBUFS;
4816 goto errout_ifa;
4819 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4820 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4821 if (err < 0) {
4822 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4823 WARN_ON(err == -EMSGSIZE);
4824 kfree_skb(skb);
4825 goto errout_ifa;
4827 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4828 errout_ifa:
4829 in6_ifa_put(ifa);
4830 errout:
4831 return err;
4834 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4836 struct sk_buff *skb;
4837 struct net *net = dev_net(ifa->idev->dev);
4838 int err = -ENOBUFS;
4840 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4841 if (!skb)
4842 goto errout;
4844 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4845 if (err < 0) {
4846 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4847 WARN_ON(err == -EMSGSIZE);
4848 kfree_skb(skb);
4849 goto errout;
4851 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4852 return;
4853 errout:
4854 if (err < 0)
4855 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4858 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4859 __s32 *array, int bytes)
4861 BUG_ON(bytes < (DEVCONF_MAX * 4));
4863 memset(array, 0, bytes);
4864 array[DEVCONF_FORWARDING] = cnf->forwarding;
4865 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4866 array[DEVCONF_MTU6] = cnf->mtu6;
4867 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4868 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4869 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4870 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4871 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4872 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4873 jiffies_to_msecs(cnf->rtr_solicit_interval);
4874 array[DEVCONF_RTR_SOLICIT_DELAY] =
4875 jiffies_to_msecs(cnf->rtr_solicit_delay);
4876 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4877 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4878 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4879 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4880 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4881 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4882 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4883 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4884 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4885 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4886 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4887 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4888 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4889 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4890 #ifdef CONFIG_IPV6_ROUTER_PREF
4891 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4892 array[DEVCONF_RTR_PROBE_INTERVAL] =
4893 jiffies_to_msecs(cnf->rtr_probe_interval);
4894 #ifdef CONFIG_IPV6_ROUTE_INFO
4895 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4896 #endif
4897 #endif
4898 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4899 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4900 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4901 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4902 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4903 #endif
4904 #ifdef CONFIG_IPV6_MROUTE
4905 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4906 #endif
4907 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4908 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4909 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4910 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4911 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4912 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4913 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4914 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4915 /* we omit DEVCONF_STABLE_SECRET for now */
4916 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4917 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4918 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4919 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4922 static inline size_t inet6_ifla6_size(void)
4924 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4925 + nla_total_size(sizeof(struct ifla_cacheinfo))
4926 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4927 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4928 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4929 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4932 static inline size_t inet6_if_nlmsg_size(void)
4934 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4935 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4936 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4937 + nla_total_size(4) /* IFLA_MTU */
4938 + nla_total_size(4) /* IFLA_LINK */
4939 + nla_total_size(1) /* IFLA_OPERSTATE */
4940 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4943 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4944 int items, int bytes)
4946 int i;
4947 int pad = bytes - sizeof(u64) * items;
4948 BUG_ON(pad < 0);
4950 /* Use put_unaligned() because stats may not be aligned for u64. */
4951 put_unaligned(items, &stats[0]);
4952 for (i = 1; i < items; i++)
4953 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4955 memset(&stats[items], 0, pad);
4958 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4959 int bytes, size_t syncpoff)
4961 int i, c;
4962 u64 buff[IPSTATS_MIB_MAX];
4963 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4965 BUG_ON(pad < 0);
4967 memset(buff, 0, sizeof(buff));
4968 buff[0] = IPSTATS_MIB_MAX;
4970 for_each_possible_cpu(c) {
4971 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4972 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4975 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4976 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4979 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4980 int bytes)
4982 switch (attrtype) {
4983 case IFLA_INET6_STATS:
4984 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4985 offsetof(struct ipstats_mib, syncp));
4986 break;
4987 case IFLA_INET6_ICMP6STATS:
4988 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4989 break;
4993 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4994 u32 ext_filter_mask)
4996 struct nlattr *nla;
4997 struct ifla_cacheinfo ci;
4999 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5000 goto nla_put_failure;
5001 ci.max_reasm_len = IPV6_MAXPLEN;
5002 ci.tstamp = cstamp_delta(idev->tstamp);
5003 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5004 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5005 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5006 goto nla_put_failure;
5007 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5008 if (!nla)
5009 goto nla_put_failure;
5010 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5012 /* XXX - MC not implemented */
5014 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5015 return 0;
5017 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5018 if (!nla)
5019 goto nla_put_failure;
5020 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5022 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5023 if (!nla)
5024 goto nla_put_failure;
5025 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5027 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5028 if (!nla)
5029 goto nla_put_failure;
5031 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5032 goto nla_put_failure;
5034 read_lock_bh(&idev->lock);
5035 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5036 read_unlock_bh(&idev->lock);
5038 return 0;
5040 nla_put_failure:
5041 return -EMSGSIZE;
5044 static size_t inet6_get_link_af_size(const struct net_device *dev,
5045 u32 ext_filter_mask)
5047 if (!__in6_dev_get(dev))
5048 return 0;
5050 return inet6_ifla6_size();
5053 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5054 u32 ext_filter_mask)
5056 struct inet6_dev *idev = __in6_dev_get(dev);
5058 if (!idev)
5059 return -ENODATA;
5061 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5062 return -EMSGSIZE;
5064 return 0;
5067 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5069 struct inet6_ifaddr *ifp;
5070 struct net_device *dev = idev->dev;
5071 bool clear_token, update_rs = false;
5072 struct in6_addr ll_addr;
5074 ASSERT_RTNL();
5076 if (!token)
5077 return -EINVAL;
5078 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5079 return -EINVAL;
5080 if (!ipv6_accept_ra(idev))
5081 return -EINVAL;
5082 if (idev->cnf.rtr_solicits <= 0)
5083 return -EINVAL;
5085 write_lock_bh(&idev->lock);
5087 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5088 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5090 write_unlock_bh(&idev->lock);
5092 clear_token = ipv6_addr_any(token);
5093 if (clear_token)
5094 goto update_lft;
5096 if (!idev->dead && (idev->if_flags & IF_READY) &&
5097 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5098 IFA_F_OPTIMISTIC)) {
5099 /* If we're not ready, then normal ifup will take care
5100 * of this. Otherwise, we need to request our rs here.
5102 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5103 update_rs = true;
5106 update_lft:
5107 write_lock_bh(&idev->lock);
5109 if (update_rs) {
5110 idev->if_flags |= IF_RS_SENT;
5111 idev->rs_probes = 1;
5112 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
5115 /* Well, that's kinda nasty ... */
5116 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5117 spin_lock(&ifp->lock);
5118 if (ifp->tokenized) {
5119 ifp->valid_lft = 0;
5120 ifp->prefered_lft = 0;
5122 spin_unlock(&ifp->lock);
5125 write_unlock_bh(&idev->lock);
5126 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5127 addrconf_verify_rtnl();
5128 return 0;
5131 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5132 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5133 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5136 static int inet6_validate_link_af(const struct net_device *dev,
5137 const struct nlattr *nla)
5139 struct nlattr *tb[IFLA_INET6_MAX + 1];
5141 if (dev && !__in6_dev_get(dev))
5142 return -EAFNOSUPPORT;
5144 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5147 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5149 int err = -EINVAL;
5150 struct inet6_dev *idev = __in6_dev_get(dev);
5151 struct nlattr *tb[IFLA_INET6_MAX + 1];
5153 if (!idev)
5154 return -EAFNOSUPPORT;
5156 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5157 BUG();
5159 if (tb[IFLA_INET6_TOKEN]) {
5160 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5161 if (err)
5162 return err;
5165 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5166 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5168 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5169 mode != IN6_ADDR_GEN_MODE_NONE &&
5170 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5171 mode != IN6_ADDR_GEN_MODE_RANDOM)
5172 return -EINVAL;
5174 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5175 !idev->cnf.stable_secret.initialized &&
5176 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5177 return -EINVAL;
5179 idev->addr_gen_mode = mode;
5180 err = 0;
5183 return err;
5186 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5187 u32 portid, u32 seq, int event, unsigned int flags)
5189 struct net_device *dev = idev->dev;
5190 struct ifinfomsg *hdr;
5191 struct nlmsghdr *nlh;
5192 void *protoinfo;
5194 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5195 if (!nlh)
5196 return -EMSGSIZE;
5198 hdr = nlmsg_data(nlh);
5199 hdr->ifi_family = AF_INET6;
5200 hdr->__ifi_pad = 0;
5201 hdr->ifi_type = dev->type;
5202 hdr->ifi_index = dev->ifindex;
5203 hdr->ifi_flags = dev_get_flags(dev);
5204 hdr->ifi_change = 0;
5206 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5207 (dev->addr_len &&
5208 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5209 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5210 (dev->ifindex != dev_get_iflink(dev) &&
5211 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5212 nla_put_u8(skb, IFLA_OPERSTATE,
5213 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5214 goto nla_put_failure;
5215 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5216 if (!protoinfo)
5217 goto nla_put_failure;
5219 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5220 goto nla_put_failure;
5222 nla_nest_end(skb, protoinfo);
5223 nlmsg_end(skb, nlh);
5224 return 0;
5226 nla_put_failure:
5227 nlmsg_cancel(skb, nlh);
5228 return -EMSGSIZE;
5231 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5233 struct net *net = sock_net(skb->sk);
5234 int h, s_h;
5235 int idx = 0, s_idx;
5236 struct net_device *dev;
5237 struct inet6_dev *idev;
5238 struct hlist_head *head;
5240 s_h = cb->args[0];
5241 s_idx = cb->args[1];
5243 rcu_read_lock();
5244 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5245 idx = 0;
5246 head = &net->dev_index_head[h];
5247 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5248 if (idx < s_idx)
5249 goto cont;
5250 idev = __in6_dev_get(dev);
5251 if (!idev)
5252 goto cont;
5253 if (inet6_fill_ifinfo(skb, idev,
5254 NETLINK_CB(cb->skb).portid,
5255 cb->nlh->nlmsg_seq,
5256 RTM_NEWLINK, NLM_F_MULTI) < 0)
5257 goto out;
5258 cont:
5259 idx++;
5262 out:
5263 rcu_read_unlock();
5264 cb->args[1] = idx;
5265 cb->args[0] = h;
5267 return skb->len;
5270 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5272 struct sk_buff *skb;
5273 struct net *net = dev_net(idev->dev);
5274 int err = -ENOBUFS;
5276 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5277 if (!skb)
5278 goto errout;
5280 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5281 if (err < 0) {
5282 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5283 WARN_ON(err == -EMSGSIZE);
5284 kfree_skb(skb);
5285 goto errout;
5287 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5288 return;
5289 errout:
5290 if (err < 0)
5291 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5294 static inline size_t inet6_prefix_nlmsg_size(void)
5296 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5297 + nla_total_size(sizeof(struct in6_addr))
5298 + nla_total_size(sizeof(struct prefix_cacheinfo));
5301 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5302 struct prefix_info *pinfo, u32 portid, u32 seq,
5303 int event, unsigned int flags)
5305 struct prefixmsg *pmsg;
5306 struct nlmsghdr *nlh;
5307 struct prefix_cacheinfo ci;
5309 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5310 if (!nlh)
5311 return -EMSGSIZE;
5313 pmsg = nlmsg_data(nlh);
5314 pmsg->prefix_family = AF_INET6;
5315 pmsg->prefix_pad1 = 0;
5316 pmsg->prefix_pad2 = 0;
5317 pmsg->prefix_ifindex = idev->dev->ifindex;
5318 pmsg->prefix_len = pinfo->prefix_len;
5319 pmsg->prefix_type = pinfo->type;
5320 pmsg->prefix_pad3 = 0;
5321 pmsg->prefix_flags = 0;
5322 if (pinfo->onlink)
5323 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5324 if (pinfo->autoconf)
5325 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5327 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5328 goto nla_put_failure;
5329 ci.preferred_time = ntohl(pinfo->prefered);
5330 ci.valid_time = ntohl(pinfo->valid);
5331 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5332 goto nla_put_failure;
5333 nlmsg_end(skb, nlh);
5334 return 0;
5336 nla_put_failure:
5337 nlmsg_cancel(skb, nlh);
5338 return -EMSGSIZE;
5341 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5342 struct prefix_info *pinfo)
5344 struct sk_buff *skb;
5345 struct net *net = dev_net(idev->dev);
5346 int err = -ENOBUFS;
5348 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5349 if (!skb)
5350 goto errout;
5352 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5353 if (err < 0) {
5354 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5355 WARN_ON(err == -EMSGSIZE);
5356 kfree_skb(skb);
5357 goto errout;
5359 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5360 return;
5361 errout:
5362 if (err < 0)
5363 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5366 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5368 struct net *net = dev_net(ifp->idev->dev);
5370 if (event)
5371 ASSERT_RTNL();
5373 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5375 switch (event) {
5376 case RTM_NEWADDR:
5378 * If the address was optimistic
5379 * we inserted the route at the start of
5380 * our DAD process, so we don't need
5381 * to do it again
5383 if (!(ifp->rt->rt6i_node))
5384 ip6_ins_rt(ifp->rt);
5385 if (ifp->idev->cnf.forwarding)
5386 addrconf_join_anycast(ifp);
5387 if (!ipv6_addr_any(&ifp->peer_addr))
5388 addrconf_prefix_route(&ifp->peer_addr, 128,
5389 ifp->idev->dev, 0, 0);
5390 break;
5391 case RTM_DELADDR:
5392 if (ifp->idev->cnf.forwarding)
5393 addrconf_leave_anycast(ifp);
5394 addrconf_leave_solict(ifp->idev, &ifp->addr);
5395 if (!ipv6_addr_any(&ifp->peer_addr)) {
5396 struct rt6_info *rt;
5398 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5399 ifp->idev->dev, 0, 0);
5400 if (rt)
5401 ip6_del_rt(rt);
5403 if (ifp->rt) {
5404 dst_hold(&ifp->rt->dst);
5405 ip6_del_rt(ifp->rt);
5407 rt_genid_bump_ipv6(net);
5408 break;
5410 atomic_inc(&net->ipv6.dev_addr_genid);
5413 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5415 rcu_read_lock_bh();
5416 if (likely(ifp->idev->dead == 0))
5417 __ipv6_ifa_notify(event, ifp);
5418 rcu_read_unlock_bh();
5421 #ifdef CONFIG_SYSCTL
5423 static
5424 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5425 void __user *buffer, size_t *lenp, loff_t *ppos)
5427 int *valp = ctl->data;
5428 int val = *valp;
5429 loff_t pos = *ppos;
5430 struct ctl_table lctl;
5431 int ret;
5434 * ctl->data points to idev->cnf.forwarding, we should
5435 * not modify it until we get the rtnl lock.
5437 lctl = *ctl;
5438 lctl.data = &val;
5440 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5442 if (write)
5443 ret = addrconf_fixup_forwarding(ctl, valp, val);
5444 if (ret)
5445 *ppos = pos;
5446 return ret;
5449 static
5450 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5451 void __user *buffer, size_t *lenp, loff_t *ppos)
5453 struct ctl_table lctl;
5454 int min_hl = 1, max_hl = 255;
5456 lctl = *ctl;
5457 lctl.extra1 = &min_hl;
5458 lctl.extra2 = &max_hl;
5460 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5463 static
5464 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5465 void __user *buffer, size_t *lenp, loff_t *ppos)
5467 struct inet6_dev *idev = ctl->extra1;
5468 int min_mtu = IPV6_MIN_MTU;
5469 struct ctl_table lctl;
5471 lctl = *ctl;
5472 lctl.extra1 = &min_mtu;
5473 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5475 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5478 static void dev_disable_change(struct inet6_dev *idev)
5480 struct netdev_notifier_info info;
5482 if (!idev || !idev->dev)
5483 return;
5485 netdev_notifier_info_init(&info, idev->dev);
5486 if (idev->cnf.disable_ipv6)
5487 addrconf_notify(NULL, NETDEV_DOWN, &info);
5488 else
5489 addrconf_notify(NULL, NETDEV_UP, &info);
5492 static void addrconf_disable_change(struct net *net, __s32 newf)
5494 struct net_device *dev;
5495 struct inet6_dev *idev;
5497 rcu_read_lock();
5498 for_each_netdev_rcu(net, dev) {
5499 idev = __in6_dev_get(dev);
5500 if (idev) {
5501 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5502 idev->cnf.disable_ipv6 = newf;
5503 if (changed)
5504 dev_disable_change(idev);
5507 rcu_read_unlock();
5510 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5512 struct net *net;
5513 int old;
5515 if (!rtnl_trylock())
5516 return restart_syscall();
5518 net = (struct net *)table->extra2;
5519 old = *p;
5520 *p = newf;
5522 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5523 rtnl_unlock();
5524 return 0;
5527 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5528 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5529 addrconf_disable_change(net, newf);
5530 } else if ((!newf) ^ (!old))
5531 dev_disable_change((struct inet6_dev *)table->extra1);
5533 rtnl_unlock();
5534 return 0;
5537 static
5538 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5539 void __user *buffer, size_t *lenp, loff_t *ppos)
5541 int *valp = ctl->data;
5542 int val = *valp;
5543 loff_t pos = *ppos;
5544 struct ctl_table lctl;
5545 int ret;
5548 * ctl->data points to idev->cnf.disable_ipv6, we should
5549 * not modify it until we get the rtnl lock.
5551 lctl = *ctl;
5552 lctl.data = &val;
5554 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5556 if (write)
5557 ret = addrconf_disable_ipv6(ctl, valp, val);
5558 if (ret)
5559 *ppos = pos;
5560 return ret;
5563 static
5564 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5565 void __user *buffer, size_t *lenp, loff_t *ppos)
5567 int *valp = ctl->data;
5568 int ret;
5569 int old, new;
5571 old = *valp;
5572 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5573 new = *valp;
5575 if (write && old != new) {
5576 struct net *net = ctl->extra2;
5578 if (!rtnl_trylock())
5579 return restart_syscall();
5581 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5582 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5583 NETCONFA_IFINDEX_DEFAULT,
5584 net->ipv6.devconf_dflt);
5585 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5586 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5587 NETCONFA_IFINDEX_ALL,
5588 net->ipv6.devconf_all);
5589 else {
5590 struct inet6_dev *idev = ctl->extra1;
5592 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5593 idev->dev->ifindex,
5594 &idev->cnf);
5596 rtnl_unlock();
5599 return ret;
5602 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5603 void __user *buffer, size_t *lenp,
5604 loff_t *ppos)
5606 int err;
5607 struct in6_addr addr;
5608 char str[IPV6_MAX_STRLEN];
5609 struct ctl_table lctl = *ctl;
5610 struct net *net = ctl->extra2;
5611 struct ipv6_stable_secret *secret = ctl->data;
5613 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5614 return -EIO;
5616 lctl.maxlen = IPV6_MAX_STRLEN;
5617 lctl.data = str;
5619 if (!rtnl_trylock())
5620 return restart_syscall();
5622 if (!write && !secret->initialized) {
5623 err = -EIO;
5624 goto out;
5627 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5628 if (err >= sizeof(str)) {
5629 err = -EIO;
5630 goto out;
5633 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5634 if (err || !write)
5635 goto out;
5637 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5638 err = -EIO;
5639 goto out;
5642 secret->initialized = true;
5643 secret->secret = addr;
5645 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5646 struct net_device *dev;
5648 for_each_netdev(net, dev) {
5649 struct inet6_dev *idev = __in6_dev_get(dev);
5651 if (idev) {
5652 idev->addr_gen_mode =
5653 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5656 } else {
5657 struct inet6_dev *idev = ctl->extra1;
5659 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5662 out:
5663 rtnl_unlock();
5665 return err;
5668 static
5669 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5670 int write,
5671 void __user *buffer,
5672 size_t *lenp,
5673 loff_t *ppos)
5675 int *valp = ctl->data;
5676 int val = *valp;
5677 loff_t pos = *ppos;
5678 struct ctl_table lctl;
5679 int ret;
5681 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5682 * we should not modify it until we get the rtnl lock.
5684 lctl = *ctl;
5685 lctl.data = &val;
5687 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5689 if (write)
5690 ret = addrconf_fixup_linkdown(ctl, valp, val);
5691 if (ret)
5692 *ppos = pos;
5693 return ret;
5696 static const struct ctl_table addrconf_sysctl[] = {
5698 .procname = "forwarding",
5699 .data = &ipv6_devconf.forwarding,
5700 .maxlen = sizeof(int),
5701 .mode = 0644,
5702 .proc_handler = addrconf_sysctl_forward,
5705 .procname = "hop_limit",
5706 .data = &ipv6_devconf.hop_limit,
5707 .maxlen = sizeof(int),
5708 .mode = 0644,
5709 .proc_handler = addrconf_sysctl_hop_limit,
5712 .procname = "mtu",
5713 .data = &ipv6_devconf.mtu6,
5714 .maxlen = sizeof(int),
5715 .mode = 0644,
5716 .proc_handler = addrconf_sysctl_mtu,
5719 .procname = "accept_ra",
5720 .data = &ipv6_devconf.accept_ra,
5721 .maxlen = sizeof(int),
5722 .mode = 0644,
5723 .proc_handler = proc_dointvec,
5726 .procname = "accept_redirects",
5727 .data = &ipv6_devconf.accept_redirects,
5728 .maxlen = sizeof(int),
5729 .mode = 0644,
5730 .proc_handler = proc_dointvec,
5733 .procname = "autoconf",
5734 .data = &ipv6_devconf.autoconf,
5735 .maxlen = sizeof(int),
5736 .mode = 0644,
5737 .proc_handler = proc_dointvec,
5740 .procname = "dad_transmits",
5741 .data = &ipv6_devconf.dad_transmits,
5742 .maxlen = sizeof(int),
5743 .mode = 0644,
5744 .proc_handler = proc_dointvec,
5747 .procname = "router_solicitations",
5748 .data = &ipv6_devconf.rtr_solicits,
5749 .maxlen = sizeof(int),
5750 .mode = 0644,
5751 .proc_handler = proc_dointvec,
5754 .procname = "router_solicitation_interval",
5755 .data = &ipv6_devconf.rtr_solicit_interval,
5756 .maxlen = sizeof(int),
5757 .mode = 0644,
5758 .proc_handler = proc_dointvec_jiffies,
5761 .procname = "router_solicitation_delay",
5762 .data = &ipv6_devconf.rtr_solicit_delay,
5763 .maxlen = sizeof(int),
5764 .mode = 0644,
5765 .proc_handler = proc_dointvec_jiffies,
5768 .procname = "force_mld_version",
5769 .data = &ipv6_devconf.force_mld_version,
5770 .maxlen = sizeof(int),
5771 .mode = 0644,
5772 .proc_handler = proc_dointvec,
5775 .procname = "mldv1_unsolicited_report_interval",
5776 .data =
5777 &ipv6_devconf.mldv1_unsolicited_report_interval,
5778 .maxlen = sizeof(int),
5779 .mode = 0644,
5780 .proc_handler = proc_dointvec_ms_jiffies,
5783 .procname = "mldv2_unsolicited_report_interval",
5784 .data =
5785 &ipv6_devconf.mldv2_unsolicited_report_interval,
5786 .maxlen = sizeof(int),
5787 .mode = 0644,
5788 .proc_handler = proc_dointvec_ms_jiffies,
5791 .procname = "use_tempaddr",
5792 .data = &ipv6_devconf.use_tempaddr,
5793 .maxlen = sizeof(int),
5794 .mode = 0644,
5795 .proc_handler = proc_dointvec,
5798 .procname = "temp_valid_lft",
5799 .data = &ipv6_devconf.temp_valid_lft,
5800 .maxlen = sizeof(int),
5801 .mode = 0644,
5802 .proc_handler = proc_dointvec,
5805 .procname = "temp_prefered_lft",
5806 .data = &ipv6_devconf.temp_prefered_lft,
5807 .maxlen = sizeof(int),
5808 .mode = 0644,
5809 .proc_handler = proc_dointvec,
5812 .procname = "regen_max_retry",
5813 .data = &ipv6_devconf.regen_max_retry,
5814 .maxlen = sizeof(int),
5815 .mode = 0644,
5816 .proc_handler = proc_dointvec,
5819 .procname = "max_desync_factor",
5820 .data = &ipv6_devconf.max_desync_factor,
5821 .maxlen = sizeof(int),
5822 .mode = 0644,
5823 .proc_handler = proc_dointvec,
5826 .procname = "max_addresses",
5827 .data = &ipv6_devconf.max_addresses,
5828 .maxlen = sizeof(int),
5829 .mode = 0644,
5830 .proc_handler = proc_dointvec,
5833 .procname = "accept_ra_defrtr",
5834 .data = &ipv6_devconf.accept_ra_defrtr,
5835 .maxlen = sizeof(int),
5836 .mode = 0644,
5837 .proc_handler = proc_dointvec,
5840 .procname = "accept_ra_min_hop_limit",
5841 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5842 .maxlen = sizeof(int),
5843 .mode = 0644,
5844 .proc_handler = proc_dointvec,
5847 .procname = "accept_ra_pinfo",
5848 .data = &ipv6_devconf.accept_ra_pinfo,
5849 .maxlen = sizeof(int),
5850 .mode = 0644,
5851 .proc_handler = proc_dointvec,
5853 #ifdef CONFIG_IPV6_ROUTER_PREF
5855 .procname = "accept_ra_rtr_pref",
5856 .data = &ipv6_devconf.accept_ra_rtr_pref,
5857 .maxlen = sizeof(int),
5858 .mode = 0644,
5859 .proc_handler = proc_dointvec,
5862 .procname = "router_probe_interval",
5863 .data = &ipv6_devconf.rtr_probe_interval,
5864 .maxlen = sizeof(int),
5865 .mode = 0644,
5866 .proc_handler = proc_dointvec_jiffies,
5868 #ifdef CONFIG_IPV6_ROUTE_INFO
5870 .procname = "accept_ra_rt_info_max_plen",
5871 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5872 .maxlen = sizeof(int),
5873 .mode = 0644,
5874 .proc_handler = proc_dointvec,
5876 #endif
5877 #endif
5879 .procname = "proxy_ndp",
5880 .data = &ipv6_devconf.proxy_ndp,
5881 .maxlen = sizeof(int),
5882 .mode = 0644,
5883 .proc_handler = addrconf_sysctl_proxy_ndp,
5886 .procname = "accept_source_route",
5887 .data = &ipv6_devconf.accept_source_route,
5888 .maxlen = sizeof(int),
5889 .mode = 0644,
5890 .proc_handler = proc_dointvec,
5892 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5894 .procname = "optimistic_dad",
5895 .data = &ipv6_devconf.optimistic_dad,
5896 .maxlen = sizeof(int),
5897 .mode = 0644,
5898 .proc_handler = proc_dointvec,
5901 .procname = "use_optimistic",
5902 .data = &ipv6_devconf.use_optimistic,
5903 .maxlen = sizeof(int),
5904 .mode = 0644,
5905 .proc_handler = proc_dointvec,
5907 #endif
5908 #ifdef CONFIG_IPV6_MROUTE
5910 .procname = "mc_forwarding",
5911 .data = &ipv6_devconf.mc_forwarding,
5912 .maxlen = sizeof(int),
5913 .mode = 0444,
5914 .proc_handler = proc_dointvec,
5916 #endif
5918 .procname = "disable_ipv6",
5919 .data = &ipv6_devconf.disable_ipv6,
5920 .maxlen = sizeof(int),
5921 .mode = 0644,
5922 .proc_handler = addrconf_sysctl_disable,
5925 .procname = "accept_dad",
5926 .data = &ipv6_devconf.accept_dad,
5927 .maxlen = sizeof(int),
5928 .mode = 0644,
5929 .proc_handler = proc_dointvec,
5932 .procname = "force_tllao",
5933 .data = &ipv6_devconf.force_tllao,
5934 .maxlen = sizeof(int),
5935 .mode = 0644,
5936 .proc_handler = proc_dointvec
5939 .procname = "ndisc_notify",
5940 .data = &ipv6_devconf.ndisc_notify,
5941 .maxlen = sizeof(int),
5942 .mode = 0644,
5943 .proc_handler = proc_dointvec
5946 .procname = "suppress_frag_ndisc",
5947 .data = &ipv6_devconf.suppress_frag_ndisc,
5948 .maxlen = sizeof(int),
5949 .mode = 0644,
5950 .proc_handler = proc_dointvec
5953 .procname = "accept_ra_from_local",
5954 .data = &ipv6_devconf.accept_ra_from_local,
5955 .maxlen = sizeof(int),
5956 .mode = 0644,
5957 .proc_handler = proc_dointvec,
5960 .procname = "accept_ra_mtu",
5961 .data = &ipv6_devconf.accept_ra_mtu,
5962 .maxlen = sizeof(int),
5963 .mode = 0644,
5964 .proc_handler = proc_dointvec,
5967 .procname = "stable_secret",
5968 .data = &ipv6_devconf.stable_secret,
5969 .maxlen = IPV6_MAX_STRLEN,
5970 .mode = 0600,
5971 .proc_handler = addrconf_sysctl_stable_secret,
5974 .procname = "use_oif_addrs_only",
5975 .data = &ipv6_devconf.use_oif_addrs_only,
5976 .maxlen = sizeof(int),
5977 .mode = 0644,
5978 .proc_handler = proc_dointvec,
5981 .procname = "ignore_routes_with_linkdown",
5982 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5983 .maxlen = sizeof(int),
5984 .mode = 0644,
5985 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5988 .procname = "drop_unicast_in_l2_multicast",
5989 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
5990 .maxlen = sizeof(int),
5991 .mode = 0644,
5992 .proc_handler = proc_dointvec,
5995 .procname = "drop_unsolicited_na",
5996 .data = &ipv6_devconf.drop_unsolicited_na,
5997 .maxlen = sizeof(int),
5998 .mode = 0644,
5999 .proc_handler = proc_dointvec,
6002 .procname = "keep_addr_on_down",
6003 .data = &ipv6_devconf.keep_addr_on_down,
6004 .maxlen = sizeof(int),
6005 .mode = 0644,
6006 .proc_handler = proc_dointvec,
6010 /* sentinel */
6014 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6015 struct inet6_dev *idev, struct ipv6_devconf *p)
6017 int i;
6018 struct ctl_table *table;
6019 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6021 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6022 if (!table)
6023 goto out;
6025 for (i = 0; table[i].data; i++) {
6026 table[i].data += (char *)p - (char *)&ipv6_devconf;
6027 table[i].extra1 = idev; /* embedded; no ref */
6028 table[i].extra2 = net;
6031 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6033 p->sysctl_header = register_net_sysctl(net, path, table);
6034 if (!p->sysctl_header)
6035 goto free;
6037 return 0;
6039 free:
6040 kfree(table);
6041 out:
6042 return -ENOBUFS;
6045 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6047 struct ctl_table *table;
6049 if (!p->sysctl_header)
6050 return;
6052 table = p->sysctl_header->ctl_table_arg;
6053 unregister_net_sysctl_table(p->sysctl_header);
6054 p->sysctl_header = NULL;
6055 kfree(table);
6058 static int addrconf_sysctl_register(struct inet6_dev *idev)
6060 int err;
6062 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6063 return -EINVAL;
6065 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6066 &ndisc_ifinfo_sysctl_change);
6067 if (err)
6068 return err;
6069 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6070 idev, &idev->cnf);
6071 if (err)
6072 neigh_sysctl_unregister(idev->nd_parms);
6074 return err;
6077 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6079 __addrconf_sysctl_unregister(&idev->cnf);
6080 neigh_sysctl_unregister(idev->nd_parms);
6084 #endif
6086 static int __net_init addrconf_init_net(struct net *net)
6088 int err = -ENOMEM;
6089 struct ipv6_devconf *all, *dflt;
6091 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6092 if (!all)
6093 goto err_alloc_all;
6095 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6096 if (!dflt)
6097 goto err_alloc_dflt;
6099 /* these will be inherited by all namespaces */
6100 dflt->autoconf = ipv6_defaults.autoconf;
6101 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6103 dflt->stable_secret.initialized = false;
6104 all->stable_secret.initialized = false;
6106 net->ipv6.devconf_all = all;
6107 net->ipv6.devconf_dflt = dflt;
6109 #ifdef CONFIG_SYSCTL
6110 err = __addrconf_sysctl_register(net, "all", NULL, all);
6111 if (err < 0)
6112 goto err_reg_all;
6114 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6115 if (err < 0)
6116 goto err_reg_dflt;
6117 #endif
6118 return 0;
6120 #ifdef CONFIG_SYSCTL
6121 err_reg_dflt:
6122 __addrconf_sysctl_unregister(all);
6123 err_reg_all:
6124 kfree(dflt);
6125 #endif
6126 err_alloc_dflt:
6127 kfree(all);
6128 err_alloc_all:
6129 return err;
6132 static void __net_exit addrconf_exit_net(struct net *net)
6134 #ifdef CONFIG_SYSCTL
6135 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6136 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6137 #endif
6138 kfree(net->ipv6.devconf_dflt);
6139 kfree(net->ipv6.devconf_all);
6142 static struct pernet_operations addrconf_ops = {
6143 .init = addrconf_init_net,
6144 .exit = addrconf_exit_net,
6147 static struct rtnl_af_ops inet6_ops __read_mostly = {
6148 .family = AF_INET6,
6149 .fill_link_af = inet6_fill_link_af,
6150 .get_link_af_size = inet6_get_link_af_size,
6151 .validate_link_af = inet6_validate_link_af,
6152 .set_link_af = inet6_set_link_af,
6156 * Init / cleanup code
6159 int __init addrconf_init(void)
6161 struct inet6_dev *idev;
6162 int i, err;
6164 err = ipv6_addr_label_init();
6165 if (err < 0) {
6166 pr_crit("%s: cannot initialize default policy table: %d\n",
6167 __func__, err);
6168 goto out;
6171 err = register_pernet_subsys(&addrconf_ops);
6172 if (err < 0)
6173 goto out_addrlabel;
6175 addrconf_wq = create_workqueue("ipv6_addrconf");
6176 if (!addrconf_wq) {
6177 err = -ENOMEM;
6178 goto out_nowq;
6181 /* The addrconf netdev notifier requires that loopback_dev
6182 * has it's ipv6 private information allocated and setup
6183 * before it can bring up and give link-local addresses
6184 * to other devices which are up.
6186 * Unfortunately, loopback_dev is not necessarily the first
6187 * entry in the global dev_base list of net devices. In fact,
6188 * it is likely to be the very last entry on that list.
6189 * So this causes the notifier registry below to try and
6190 * give link-local addresses to all devices besides loopback_dev
6191 * first, then loopback_dev, which cases all the non-loopback_dev
6192 * devices to fail to get a link-local address.
6194 * So, as a temporary fix, allocate the ipv6 structure for
6195 * loopback_dev first by hand.
6196 * Longer term, all of the dependencies ipv6 has upon the loopback
6197 * device and it being up should be removed.
6199 rtnl_lock();
6200 idev = ipv6_add_dev(init_net.loopback_dev);
6201 rtnl_unlock();
6202 if (IS_ERR(idev)) {
6203 err = PTR_ERR(idev);
6204 goto errlo;
6207 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6208 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6210 register_netdevice_notifier(&ipv6_dev_notf);
6212 addrconf_verify();
6214 rtnl_af_register(&inet6_ops);
6216 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6217 NULL);
6218 if (err < 0)
6219 goto errout;
6221 /* Only the first call to __rtnl_register can fail */
6222 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6223 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6224 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6225 inet6_dump_ifaddr, NULL);
6226 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6227 inet6_dump_ifmcaddr, NULL);
6228 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6229 inet6_dump_ifacaddr, NULL);
6230 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6231 inet6_netconf_dump_devconf, NULL);
6233 ipv6_addr_label_rtnl_register();
6235 return 0;
6236 errout:
6237 rtnl_af_unregister(&inet6_ops);
6238 unregister_netdevice_notifier(&ipv6_dev_notf);
6239 errlo:
6240 destroy_workqueue(addrconf_wq);
6241 out_nowq:
6242 unregister_pernet_subsys(&addrconf_ops);
6243 out_addrlabel:
6244 ipv6_addr_label_cleanup();
6245 out:
6246 return err;
6249 void addrconf_cleanup(void)
6251 struct net_device *dev;
6252 int i;
6254 unregister_netdevice_notifier(&ipv6_dev_notf);
6255 unregister_pernet_subsys(&addrconf_ops);
6256 ipv6_addr_label_cleanup();
6258 rtnl_lock();
6260 __rtnl_af_unregister(&inet6_ops);
6262 /* clean dev list */
6263 for_each_netdev(&init_net, dev) {
6264 if (__in6_dev_get(dev) == NULL)
6265 continue;
6266 addrconf_ifdown(dev, 1);
6268 addrconf_ifdown(init_net.loopback_dev, 2);
6271 * Check hash table.
6273 spin_lock_bh(&addrconf_hash_lock);
6274 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6275 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6276 spin_unlock_bh(&addrconf_hash_lock);
6277 cancel_delayed_work(&addr_chk_work);
6278 rtnl_unlock();
6280 destroy_workqueue(addrconf_wq);