perf stat: Add extra output of counter values with -vv
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob8ec4b3089e206e10dd60d36b2643911e2e23daba
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->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (!ndev->nd_parms) {
365 kfree(ndev);
366 return ERR_PTR(err);
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
371 dev_hold(dev);
373 if (snmp6_alloc_dev(ndev) < 0) {
374 ADBG(KERN_WARNING
375 "%s: cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name);
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 dev_put(dev);
379 kfree(ndev);
380 return ERR_PTR(err);
383 if (snmp6_register_dev(ndev) < 0) {
384 ADBG(KERN_WARNING
385 "%s: cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name);
387 goto err_release;
390 /* One reference from device. We must do this before
391 * we invoke __ipv6_regen_rndid().
393 in6_dev_hold(ndev);
395 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
396 ndev->cnf.accept_dad = -1;
398 #if IS_ENABLED(CONFIG_IPV6_SIT)
399 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
400 pr_info("%s: Disabled Multicast RS\n", dev->name);
401 ndev->cnf.rtr_solicits = 0;
403 #endif
405 INIT_LIST_HEAD(&ndev->tempaddr_list);
406 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
407 if ((dev->flags&IFF_LOOPBACK) ||
408 dev->type == ARPHRD_TUNNEL ||
409 dev->type == ARPHRD_TUNNEL6 ||
410 dev->type == ARPHRD_SIT ||
411 dev->type == ARPHRD_NONE) {
412 ndev->cnf.use_tempaddr = -1;
413 } else {
414 in6_dev_hold(ndev);
415 ipv6_regen_rndid((unsigned long) ndev);
418 ndev->token = in6addr_any;
420 if (netif_running(dev) && addrconf_qdisc_ok(dev))
421 ndev->if_flags |= IF_READY;
423 ipv6_mc_init_dev(ndev);
424 ndev->tstamp = jiffies;
425 err = addrconf_sysctl_register(ndev);
426 if (err) {
427 ipv6_mc_destroy_dev(ndev);
428 del_timer(&ndev->regen_timer);
429 snmp6_unregister_dev(ndev);
430 goto err_release;
432 /* protected by rtnl_lock */
433 rcu_assign_pointer(dev->ip6_ptr, ndev);
435 /* Join interface-local all-node multicast group */
436 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
438 /* Join all-node multicast group */
439 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
441 /* Join all-router multicast group if forwarding is set */
442 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
443 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
445 return ndev;
447 err_release:
448 neigh_parms_release(&nd_tbl, ndev->nd_parms);
449 ndev->dead = 1;
450 in6_dev_finish_destroy(ndev);
451 return ERR_PTR(err);
454 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
456 struct inet6_dev *idev;
458 ASSERT_RTNL();
460 idev = __in6_dev_get(dev);
461 if (!idev) {
462 idev = ipv6_add_dev(dev);
463 if (IS_ERR(idev))
464 return NULL;
467 if (dev->flags&IFF_UP)
468 ipv6_mc_up(idev);
469 return idev;
472 static int inet6_netconf_msgsize_devconf(int type)
474 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
475 + nla_total_size(4); /* NETCONFA_IFINDEX */
476 bool all = false;
478 if (type == NETCONFA_ALL)
479 all = true;
481 if (all || type == NETCONFA_FORWARDING)
482 size += nla_total_size(4);
483 #ifdef CONFIG_IPV6_MROUTE
484 if (all || type == NETCONFA_MC_FORWARDING)
485 size += nla_total_size(4);
486 #endif
487 if (all || type == NETCONFA_PROXY_NEIGH)
488 size += nla_total_size(4);
490 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
491 size += nla_total_size(4);
493 return size;
496 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
497 struct ipv6_devconf *devconf, u32 portid,
498 u32 seq, int event, unsigned int flags,
499 int type)
501 struct nlmsghdr *nlh;
502 struct netconfmsg *ncm;
503 bool all = false;
505 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
506 flags);
507 if (!nlh)
508 return -EMSGSIZE;
510 if (type == NETCONFA_ALL)
511 all = true;
513 ncm = nlmsg_data(nlh);
514 ncm->ncm_family = AF_INET6;
516 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
517 goto nla_put_failure;
519 if ((all || type == NETCONFA_FORWARDING) &&
520 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
521 goto nla_put_failure;
522 #ifdef CONFIG_IPV6_MROUTE
523 if ((all || type == NETCONFA_MC_FORWARDING) &&
524 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
525 devconf->mc_forwarding) < 0)
526 goto nla_put_failure;
527 #endif
528 if ((all || type == NETCONFA_PROXY_NEIGH) &&
529 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
530 goto nla_put_failure;
532 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
533 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
534 devconf->ignore_routes_with_linkdown) < 0)
535 goto nla_put_failure;
537 nlmsg_end(skb, nlh);
538 return 0;
540 nla_put_failure:
541 nlmsg_cancel(skb, nlh);
542 return -EMSGSIZE;
545 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
546 struct ipv6_devconf *devconf)
548 struct sk_buff *skb;
549 int err = -ENOBUFS;
551 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
552 if (!skb)
553 goto errout;
555 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
556 RTM_NEWNETCONF, 0, type);
557 if (err < 0) {
558 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
559 WARN_ON(err == -EMSGSIZE);
560 kfree_skb(skb);
561 goto errout;
563 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
564 return;
565 errout:
566 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
569 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
570 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
571 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
572 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
573 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
576 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
577 struct nlmsghdr *nlh)
579 struct net *net = sock_net(in_skb->sk);
580 struct nlattr *tb[NETCONFA_MAX+1];
581 struct netconfmsg *ncm;
582 struct sk_buff *skb;
583 struct ipv6_devconf *devconf;
584 struct inet6_dev *in6_dev;
585 struct net_device *dev;
586 int ifindex;
587 int err;
589 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
590 devconf_ipv6_policy);
591 if (err < 0)
592 goto errout;
594 err = -EINVAL;
595 if (!tb[NETCONFA_IFINDEX])
596 goto errout;
598 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
599 switch (ifindex) {
600 case NETCONFA_IFINDEX_ALL:
601 devconf = net->ipv6.devconf_all;
602 break;
603 case NETCONFA_IFINDEX_DEFAULT:
604 devconf = net->ipv6.devconf_dflt;
605 break;
606 default:
607 dev = __dev_get_by_index(net, ifindex);
608 if (!dev)
609 goto errout;
610 in6_dev = __in6_dev_get(dev);
611 if (!in6_dev)
612 goto errout;
613 devconf = &in6_dev->cnf;
614 break;
617 err = -ENOBUFS;
618 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
619 if (!skb)
620 goto errout;
622 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
623 NETLINK_CB(in_skb).portid,
624 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
625 NETCONFA_ALL);
626 if (err < 0) {
627 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
628 WARN_ON(err == -EMSGSIZE);
629 kfree_skb(skb);
630 goto errout;
632 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
633 errout:
634 return err;
637 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
638 struct netlink_callback *cb)
640 struct net *net = sock_net(skb->sk);
641 int h, s_h;
642 int idx, s_idx;
643 struct net_device *dev;
644 struct inet6_dev *idev;
645 struct hlist_head *head;
647 s_h = cb->args[0];
648 s_idx = idx = cb->args[1];
650 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
651 idx = 0;
652 head = &net->dev_index_head[h];
653 rcu_read_lock();
654 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
655 net->dev_base_seq;
656 hlist_for_each_entry_rcu(dev, head, index_hlist) {
657 if (idx < s_idx)
658 goto cont;
659 idev = __in6_dev_get(dev);
660 if (!idev)
661 goto cont;
663 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
664 &idev->cnf,
665 NETLINK_CB(cb->skb).portid,
666 cb->nlh->nlmsg_seq,
667 RTM_NEWNETCONF,
668 NLM_F_MULTI,
669 NETCONFA_ALL) < 0) {
670 rcu_read_unlock();
671 goto done;
673 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
674 cont:
675 idx++;
677 rcu_read_unlock();
679 if (h == NETDEV_HASHENTRIES) {
680 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
681 net->ipv6.devconf_all,
682 NETLINK_CB(cb->skb).portid,
683 cb->nlh->nlmsg_seq,
684 RTM_NEWNETCONF, NLM_F_MULTI,
685 NETCONFA_ALL) < 0)
686 goto done;
687 else
688 h++;
690 if (h == NETDEV_HASHENTRIES + 1) {
691 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
692 net->ipv6.devconf_dflt,
693 NETLINK_CB(cb->skb).portid,
694 cb->nlh->nlmsg_seq,
695 RTM_NEWNETCONF, NLM_F_MULTI,
696 NETCONFA_ALL) < 0)
697 goto done;
698 else
699 h++;
701 done:
702 cb->args[0] = h;
703 cb->args[1] = idx;
705 return skb->len;
708 #ifdef CONFIG_SYSCTL
709 static void dev_forward_change(struct inet6_dev *idev)
711 struct net_device *dev;
712 struct inet6_ifaddr *ifa;
714 if (!idev)
715 return;
716 dev = idev->dev;
717 if (idev->cnf.forwarding)
718 dev_disable_lro(dev);
719 if (dev->flags & IFF_MULTICAST) {
720 if (idev->cnf.forwarding) {
721 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
722 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
723 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
724 } else {
725 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
726 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
727 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
731 list_for_each_entry(ifa, &idev->addr_list, if_list) {
732 if (ifa->flags&IFA_F_TENTATIVE)
733 continue;
734 if (idev->cnf.forwarding)
735 addrconf_join_anycast(ifa);
736 else
737 addrconf_leave_anycast(ifa);
739 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
740 dev->ifindex, &idev->cnf);
744 static void addrconf_forward_change(struct net *net, __s32 newf)
746 struct net_device *dev;
747 struct inet6_dev *idev;
749 for_each_netdev(net, dev) {
750 idev = __in6_dev_get(dev);
751 if (idev) {
752 int changed = (!idev->cnf.forwarding) ^ (!newf);
753 idev->cnf.forwarding = newf;
754 if (changed)
755 dev_forward_change(idev);
760 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
762 struct net *net;
763 int old;
765 if (!rtnl_trylock())
766 return restart_syscall();
768 net = (struct net *)table->extra2;
769 old = *p;
770 *p = newf;
772 if (p == &net->ipv6.devconf_dflt->forwarding) {
773 if ((!newf) ^ (!old))
774 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
775 NETCONFA_IFINDEX_DEFAULT,
776 net->ipv6.devconf_dflt);
777 rtnl_unlock();
778 return 0;
781 if (p == &net->ipv6.devconf_all->forwarding) {
782 net->ipv6.devconf_dflt->forwarding = newf;
783 addrconf_forward_change(net, newf);
784 if ((!newf) ^ (!old))
785 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
786 NETCONFA_IFINDEX_ALL,
787 net->ipv6.devconf_all);
788 } else if ((!newf) ^ (!old))
789 dev_forward_change((struct inet6_dev *)table->extra1);
790 rtnl_unlock();
792 if (newf)
793 rt6_purge_dflt_routers(net);
794 return 1;
797 static void addrconf_linkdown_change(struct net *net, __s32 newf)
799 struct net_device *dev;
800 struct inet6_dev *idev;
802 for_each_netdev(net, dev) {
803 idev = __in6_dev_get(dev);
804 if (idev) {
805 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
807 idev->cnf.ignore_routes_with_linkdown = newf;
808 if (changed)
809 inet6_netconf_notify_devconf(dev_net(dev),
810 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
811 dev->ifindex,
812 &idev->cnf);
817 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
819 struct net *net;
820 int old;
822 if (!rtnl_trylock())
823 return restart_syscall();
825 net = (struct net *)table->extra2;
826 old = *p;
827 *p = newf;
829 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
830 if ((!newf) ^ (!old))
831 inet6_netconf_notify_devconf(net,
832 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
833 NETCONFA_IFINDEX_DEFAULT,
834 net->ipv6.devconf_dflt);
835 rtnl_unlock();
836 return 0;
839 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
840 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
841 addrconf_linkdown_change(net, newf);
842 if ((!newf) ^ (!old))
843 inet6_netconf_notify_devconf(net,
844 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
845 NETCONFA_IFINDEX_ALL,
846 net->ipv6.devconf_all);
848 rtnl_unlock();
850 return 1;
853 #endif
855 /* Nobody refers to this ifaddr, destroy it */
856 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
858 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
860 #ifdef NET_REFCNT_DEBUG
861 pr_debug("%s\n", __func__);
862 #endif
864 in6_dev_put(ifp->idev);
866 if (cancel_delayed_work(&ifp->dad_work))
867 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
868 ifp);
870 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
871 pr_warn("Freeing alive inet6 address %p\n", ifp);
872 return;
874 ip6_rt_put(ifp->rt);
876 kfree_rcu(ifp, rcu);
879 static void
880 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
882 struct list_head *p;
883 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
886 * Each device address list is sorted in order of scope -
887 * global before linklocal.
889 list_for_each(p, &idev->addr_list) {
890 struct inet6_ifaddr *ifa
891 = list_entry(p, struct inet6_ifaddr, if_list);
892 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
893 break;
896 list_add_tail(&ifp->if_list, p);
899 static u32 inet6_addr_hash(const struct in6_addr *addr)
901 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
904 /* On success it returns ifp with increased reference count */
906 static struct inet6_ifaddr *
907 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
908 const struct in6_addr *peer_addr, int pfxlen,
909 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
911 struct inet6_ifaddr *ifa = NULL;
912 struct rt6_info *rt;
913 unsigned int hash;
914 int err = 0;
915 int addr_type = ipv6_addr_type(addr);
917 if (addr_type == IPV6_ADDR_ANY ||
918 addr_type & IPV6_ADDR_MULTICAST ||
919 (!(idev->dev->flags & IFF_LOOPBACK) &&
920 addr_type & IPV6_ADDR_LOOPBACK))
921 return ERR_PTR(-EADDRNOTAVAIL);
923 rcu_read_lock_bh();
924 if (idev->dead) {
925 err = -ENODEV; /*XXX*/
926 goto out2;
929 if (idev->cnf.disable_ipv6) {
930 err = -EACCES;
931 goto out2;
934 spin_lock(&addrconf_hash_lock);
936 /* Ignore adding duplicate addresses on an interface */
937 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
938 ADBG("ipv6_add_addr: already assigned\n");
939 err = -EEXIST;
940 goto out;
943 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
945 if (!ifa) {
946 ADBG("ipv6_add_addr: malloc failed\n");
947 err = -ENOBUFS;
948 goto out;
951 rt = addrconf_dst_alloc(idev, addr, false);
952 if (IS_ERR(rt)) {
953 err = PTR_ERR(rt);
954 goto out;
957 neigh_parms_data_state_setall(idev->nd_parms);
959 ifa->addr = *addr;
960 if (peer_addr)
961 ifa->peer_addr = *peer_addr;
963 spin_lock_init(&ifa->lock);
964 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
965 INIT_HLIST_NODE(&ifa->addr_lst);
966 ifa->scope = scope;
967 ifa->prefix_len = pfxlen;
968 ifa->flags = flags | IFA_F_TENTATIVE;
969 ifa->valid_lft = valid_lft;
970 ifa->prefered_lft = prefered_lft;
971 ifa->cstamp = ifa->tstamp = jiffies;
972 ifa->tokenized = false;
974 ifa->rt = rt;
976 ifa->idev = idev;
977 in6_dev_hold(idev);
978 /* For caller */
979 in6_ifa_hold(ifa);
981 /* Add to big hash table */
982 hash = inet6_addr_hash(addr);
984 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
985 spin_unlock(&addrconf_hash_lock);
987 write_lock(&idev->lock);
988 /* Add to inet6_dev unicast addr list. */
989 ipv6_link_dev_addr(idev, ifa);
991 if (ifa->flags&IFA_F_TEMPORARY) {
992 list_add(&ifa->tmp_list, &idev->tempaddr_list);
993 in6_ifa_hold(ifa);
996 in6_ifa_hold(ifa);
997 write_unlock(&idev->lock);
998 out2:
999 rcu_read_unlock_bh();
1001 if (likely(err == 0))
1002 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1003 else {
1004 kfree(ifa);
1005 ifa = ERR_PTR(err);
1008 return ifa;
1009 out:
1010 spin_unlock(&addrconf_hash_lock);
1011 goto out2;
1014 enum cleanup_prefix_rt_t {
1015 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1016 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1017 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1021 * Check, whether the prefix for ifp would still need a prefix route
1022 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1023 * constants.
1025 * 1) we don't purge prefix if address was not permanent.
1026 * prefix is managed by its own lifetime.
1027 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1028 * 3) if there are no addresses, delete prefix.
1029 * 4) if there are still other permanent address(es),
1030 * corresponding prefix is still permanent.
1031 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1032 * don't purge the prefix, assume user space is managing it.
1033 * 6) otherwise, update prefix lifetime to the
1034 * longest valid lifetime among the corresponding
1035 * addresses on the device.
1036 * Note: subsequent RA will update lifetime.
1038 static enum cleanup_prefix_rt_t
1039 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1041 struct inet6_ifaddr *ifa;
1042 struct inet6_dev *idev = ifp->idev;
1043 unsigned long lifetime;
1044 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1046 *expires = jiffies;
1048 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1049 if (ifa == ifp)
1050 continue;
1051 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1052 ifp->prefix_len))
1053 continue;
1054 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1055 return CLEANUP_PREFIX_RT_NOP;
1057 action = CLEANUP_PREFIX_RT_EXPIRE;
1059 spin_lock(&ifa->lock);
1061 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1063 * Note: Because this address is
1064 * not permanent, lifetime <
1065 * LONG_MAX / HZ here.
1067 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1068 *expires = ifa->tstamp + lifetime * HZ;
1069 spin_unlock(&ifa->lock);
1072 return action;
1075 static void
1076 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1078 struct rt6_info *rt;
1080 rt = addrconf_get_prefix_route(&ifp->addr,
1081 ifp->prefix_len,
1082 ifp->idev->dev,
1083 0, RTF_GATEWAY | RTF_DEFAULT);
1084 if (rt) {
1085 if (del_rt)
1086 ip6_del_rt(rt);
1087 else {
1088 if (!(rt->rt6i_flags & RTF_EXPIRES))
1089 rt6_set_expires(rt, expires);
1090 ip6_rt_put(rt);
1096 /* This function wants to get referenced ifp and releases it before return */
1098 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1100 int state;
1101 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1102 unsigned long expires;
1104 ASSERT_RTNL();
1106 spin_lock_bh(&ifp->lock);
1107 state = ifp->state;
1108 ifp->state = INET6_IFADDR_STATE_DEAD;
1109 spin_unlock_bh(&ifp->lock);
1111 if (state == INET6_IFADDR_STATE_DEAD)
1112 goto out;
1114 spin_lock_bh(&addrconf_hash_lock);
1115 hlist_del_init_rcu(&ifp->addr_lst);
1116 spin_unlock_bh(&addrconf_hash_lock);
1118 write_lock_bh(&ifp->idev->lock);
1120 if (ifp->flags&IFA_F_TEMPORARY) {
1121 list_del(&ifp->tmp_list);
1122 if (ifp->ifpub) {
1123 in6_ifa_put(ifp->ifpub);
1124 ifp->ifpub = NULL;
1126 __in6_ifa_put(ifp);
1129 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1130 action = check_cleanup_prefix_route(ifp, &expires);
1132 list_del_init(&ifp->if_list);
1133 __in6_ifa_put(ifp);
1135 write_unlock_bh(&ifp->idev->lock);
1137 addrconf_del_dad_work(ifp);
1139 ipv6_ifa_notify(RTM_DELADDR, ifp);
1141 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1143 if (action != CLEANUP_PREFIX_RT_NOP) {
1144 cleanup_prefix_route(ifp, expires,
1145 action == CLEANUP_PREFIX_RT_DEL);
1148 /* clean up prefsrc entries */
1149 rt6_remove_prefsrc(ifp);
1150 out:
1151 in6_ifa_put(ifp);
1154 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1156 struct inet6_dev *idev = ifp->idev;
1157 struct in6_addr addr, *tmpaddr;
1158 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1159 unsigned long regen_advance;
1160 int tmp_plen;
1161 int ret = 0;
1162 u32 addr_flags;
1163 unsigned long now = jiffies;
1165 write_lock_bh(&idev->lock);
1166 if (ift) {
1167 spin_lock_bh(&ift->lock);
1168 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1169 spin_unlock_bh(&ift->lock);
1170 tmpaddr = &addr;
1171 } else {
1172 tmpaddr = NULL;
1174 retry:
1175 in6_dev_hold(idev);
1176 if (idev->cnf.use_tempaddr <= 0) {
1177 write_unlock_bh(&idev->lock);
1178 pr_info("%s: use_tempaddr is disabled\n", __func__);
1179 in6_dev_put(idev);
1180 ret = -1;
1181 goto out;
1183 spin_lock_bh(&ifp->lock);
1184 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1185 idev->cnf.use_tempaddr = -1; /*XXX*/
1186 spin_unlock_bh(&ifp->lock);
1187 write_unlock_bh(&idev->lock);
1188 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1189 __func__);
1190 in6_dev_put(idev);
1191 ret = -1;
1192 goto out;
1194 in6_ifa_hold(ifp);
1195 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1196 __ipv6_try_regen_rndid(idev, tmpaddr);
1197 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1198 age = (now - ifp->tstamp) / HZ;
1199 tmp_valid_lft = min_t(__u32,
1200 ifp->valid_lft,
1201 idev->cnf.temp_valid_lft + age);
1202 tmp_prefered_lft = min_t(__u32,
1203 ifp->prefered_lft,
1204 idev->cnf.temp_prefered_lft + age -
1205 idev->cnf.max_desync_factor);
1206 tmp_plen = ifp->prefix_len;
1207 tmp_tstamp = ifp->tstamp;
1208 spin_unlock_bh(&ifp->lock);
1210 regen_advance = idev->cnf.regen_max_retry *
1211 idev->cnf.dad_transmits *
1212 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1213 write_unlock_bh(&idev->lock);
1215 /* A temporary address is created only if this calculated Preferred
1216 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1217 * an implementation must not create a temporary address with a zero
1218 * Preferred Lifetime.
1219 * Use age calculation as in addrconf_verify to avoid unnecessary
1220 * temporary addresses being generated.
1222 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1223 if (tmp_prefered_lft <= regen_advance + age) {
1224 in6_ifa_put(ifp);
1225 in6_dev_put(idev);
1226 ret = -1;
1227 goto out;
1230 addr_flags = IFA_F_TEMPORARY;
1231 /* set in addrconf_prefix_rcv() */
1232 if (ifp->flags & IFA_F_OPTIMISTIC)
1233 addr_flags |= IFA_F_OPTIMISTIC;
1235 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1236 ipv6_addr_scope(&addr), addr_flags,
1237 tmp_valid_lft, tmp_prefered_lft);
1238 if (IS_ERR(ift)) {
1239 in6_ifa_put(ifp);
1240 in6_dev_put(idev);
1241 pr_info("%s: retry temporary address regeneration\n", __func__);
1242 tmpaddr = &addr;
1243 write_lock_bh(&idev->lock);
1244 goto retry;
1247 spin_lock_bh(&ift->lock);
1248 ift->ifpub = ifp;
1249 ift->cstamp = now;
1250 ift->tstamp = tmp_tstamp;
1251 spin_unlock_bh(&ift->lock);
1253 addrconf_dad_start(ift);
1254 in6_ifa_put(ift);
1255 in6_dev_put(idev);
1256 out:
1257 return ret;
1261 * Choose an appropriate source address (RFC3484)
1263 enum {
1264 IPV6_SADDR_RULE_INIT = 0,
1265 IPV6_SADDR_RULE_LOCAL,
1266 IPV6_SADDR_RULE_SCOPE,
1267 IPV6_SADDR_RULE_PREFERRED,
1268 #ifdef CONFIG_IPV6_MIP6
1269 IPV6_SADDR_RULE_HOA,
1270 #endif
1271 IPV6_SADDR_RULE_OIF,
1272 IPV6_SADDR_RULE_LABEL,
1273 IPV6_SADDR_RULE_PRIVACY,
1274 IPV6_SADDR_RULE_ORCHID,
1275 IPV6_SADDR_RULE_PREFIX,
1276 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1277 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1278 #endif
1279 IPV6_SADDR_RULE_MAX
1282 struct ipv6_saddr_score {
1283 int rule;
1284 int addr_type;
1285 struct inet6_ifaddr *ifa;
1286 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1287 int scopedist;
1288 int matchlen;
1291 struct ipv6_saddr_dst {
1292 const struct in6_addr *addr;
1293 int ifindex;
1294 int scope;
1295 int label;
1296 unsigned int prefs;
1299 static inline int ipv6_saddr_preferred(int type)
1301 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1302 return 1;
1303 return 0;
1306 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1308 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1309 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1310 #else
1311 return false;
1312 #endif
1315 static int ipv6_get_saddr_eval(struct net *net,
1316 struct ipv6_saddr_score *score,
1317 struct ipv6_saddr_dst *dst,
1318 int i)
1320 int ret;
1322 if (i <= score->rule) {
1323 switch (i) {
1324 case IPV6_SADDR_RULE_SCOPE:
1325 ret = score->scopedist;
1326 break;
1327 case IPV6_SADDR_RULE_PREFIX:
1328 ret = score->matchlen;
1329 break;
1330 default:
1331 ret = !!test_bit(i, score->scorebits);
1333 goto out;
1336 switch (i) {
1337 case IPV6_SADDR_RULE_INIT:
1338 /* Rule 0: remember if hiscore is not ready yet */
1339 ret = !!score->ifa;
1340 break;
1341 case IPV6_SADDR_RULE_LOCAL:
1342 /* Rule 1: Prefer same address */
1343 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1344 break;
1345 case IPV6_SADDR_RULE_SCOPE:
1346 /* Rule 2: Prefer appropriate scope
1348 * ret
1350 * -1 | d 15
1351 * ---+--+-+---> scope
1353 * | d is scope of the destination.
1354 * B-d | \
1355 * | \ <- smaller scope is better if
1356 * B-15 | \ if scope is enough for destination.
1357 * | ret = B - scope (-1 <= scope >= d <= 15).
1358 * d-C-1 | /
1359 * |/ <- greater is better
1360 * -C / if scope is not enough for destination.
1361 * /| ret = scope - C (-1 <= d < scope <= 15).
1363 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1364 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1365 * Assume B = 0 and we get C > 29.
1367 ret = __ipv6_addr_src_scope(score->addr_type);
1368 if (ret >= dst->scope)
1369 ret = -ret;
1370 else
1371 ret -= 128; /* 30 is enough */
1372 score->scopedist = ret;
1373 break;
1374 case IPV6_SADDR_RULE_PREFERRED:
1376 /* Rule 3: Avoid deprecated and optimistic addresses */
1377 u8 avoid = IFA_F_DEPRECATED;
1379 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1380 avoid |= IFA_F_OPTIMISTIC;
1381 ret = ipv6_saddr_preferred(score->addr_type) ||
1382 !(score->ifa->flags & avoid);
1383 break;
1385 #ifdef CONFIG_IPV6_MIP6
1386 case IPV6_SADDR_RULE_HOA:
1388 /* Rule 4: Prefer home address */
1389 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1390 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1391 break;
1393 #endif
1394 case IPV6_SADDR_RULE_OIF:
1395 /* Rule 5: Prefer outgoing interface */
1396 ret = (!dst->ifindex ||
1397 dst->ifindex == score->ifa->idev->dev->ifindex);
1398 break;
1399 case IPV6_SADDR_RULE_LABEL:
1400 /* Rule 6: Prefer matching label */
1401 ret = ipv6_addr_label(net,
1402 &score->ifa->addr, score->addr_type,
1403 score->ifa->idev->dev->ifindex) == dst->label;
1404 break;
1405 case IPV6_SADDR_RULE_PRIVACY:
1407 /* Rule 7: Prefer public address
1408 * Note: prefer temporary address if use_tempaddr >= 2
1410 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1411 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1412 score->ifa->idev->cnf.use_tempaddr >= 2;
1413 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1414 break;
1416 case IPV6_SADDR_RULE_ORCHID:
1417 /* Rule 8-: Prefer ORCHID vs ORCHID or
1418 * non-ORCHID vs non-ORCHID
1420 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1421 ipv6_addr_orchid(dst->addr));
1422 break;
1423 case IPV6_SADDR_RULE_PREFIX:
1424 /* Rule 8: Use longest matching prefix */
1425 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1426 if (ret > score->ifa->prefix_len)
1427 ret = score->ifa->prefix_len;
1428 score->matchlen = ret;
1429 break;
1430 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1431 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1432 /* Optimistic addresses still have lower precedence than other
1433 * preferred addresses.
1435 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1436 break;
1437 #endif
1438 default:
1439 ret = 0;
1442 if (ret)
1443 __set_bit(i, score->scorebits);
1444 score->rule = i;
1445 out:
1446 return ret;
1449 static int __ipv6_dev_get_saddr(struct net *net,
1450 struct ipv6_saddr_dst *dst,
1451 struct inet6_dev *idev,
1452 struct ipv6_saddr_score *scores,
1453 int hiscore_idx)
1455 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1457 read_lock_bh(&idev->lock);
1458 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1459 int i;
1462 * - Tentative Address (RFC2462 section 5.4)
1463 * - A tentative address is not considered
1464 * "assigned to an interface" in the traditional
1465 * sense, unless it is also flagged as optimistic.
1466 * - Candidate Source Address (section 4)
1467 * - In any case, anycast addresses, multicast
1468 * addresses, and the unspecified address MUST
1469 * NOT be included in a candidate set.
1471 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1472 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1473 continue;
1475 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1477 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1478 score->addr_type & IPV6_ADDR_MULTICAST)) {
1479 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1480 idev->dev->name);
1481 continue;
1484 score->rule = -1;
1485 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1487 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1488 int minihiscore, miniscore;
1490 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1491 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1493 if (minihiscore > miniscore) {
1494 if (i == IPV6_SADDR_RULE_SCOPE &&
1495 score->scopedist > 0) {
1497 * special case:
1498 * each remaining entry
1499 * has too small (not enough)
1500 * scope, because ifa entries
1501 * are sorted by their scope
1502 * values.
1504 goto out;
1506 break;
1507 } else if (minihiscore < miniscore) {
1508 if (hiscore->ifa)
1509 in6_ifa_put(hiscore->ifa);
1511 in6_ifa_hold(score->ifa);
1513 swap(hiscore, score);
1514 hiscore_idx = 1 - hiscore_idx;
1516 /* restore our iterator */
1517 score->ifa = hiscore->ifa;
1519 break;
1523 out:
1524 read_unlock_bh(&idev->lock);
1525 return hiscore_idx;
1528 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1529 const struct in6_addr *daddr, unsigned int prefs,
1530 struct in6_addr *saddr)
1532 struct ipv6_saddr_score scores[2], *hiscore;
1533 struct ipv6_saddr_dst dst;
1534 struct inet6_dev *idev;
1535 struct net_device *dev;
1536 int dst_type;
1537 bool use_oif_addr = false;
1538 int hiscore_idx = 0;
1540 dst_type = __ipv6_addr_type(daddr);
1541 dst.addr = daddr;
1542 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1543 dst.scope = __ipv6_addr_src_scope(dst_type);
1544 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1545 dst.prefs = prefs;
1547 scores[hiscore_idx].rule = -1;
1548 scores[hiscore_idx].ifa = NULL;
1550 rcu_read_lock();
1552 /* Candidate Source Address (section 4)
1553 * - multicast and link-local destination address,
1554 * the set of candidate source address MUST only
1555 * include addresses assigned to interfaces
1556 * belonging to the same link as the outgoing
1557 * interface.
1558 * (- For site-local destination addresses, the
1559 * set of candidate source addresses MUST only
1560 * include addresses assigned to interfaces
1561 * belonging to the same site as the outgoing
1562 * interface.)
1563 * - "It is RECOMMENDED that the candidate source addresses
1564 * be the set of unicast addresses assigned to the
1565 * interface that will be used to send to the destination
1566 * (the 'outgoing' interface)." (RFC 6724)
1568 if (dst_dev) {
1569 idev = __in6_dev_get(dst_dev);
1570 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1571 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1572 (idev && idev->cnf.use_oif_addrs_only)) {
1573 use_oif_addr = true;
1577 if (use_oif_addr) {
1578 if (idev)
1579 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1580 } else {
1581 for_each_netdev_rcu(net, dev) {
1582 idev = __in6_dev_get(dev);
1583 if (!idev)
1584 continue;
1585 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1588 rcu_read_unlock();
1590 hiscore = &scores[hiscore_idx];
1591 if (!hiscore->ifa)
1592 return -EADDRNOTAVAIL;
1594 *saddr = hiscore->ifa->addr;
1595 in6_ifa_put(hiscore->ifa);
1596 return 0;
1598 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1600 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1601 u32 banned_flags)
1603 struct inet6_ifaddr *ifp;
1604 int err = -EADDRNOTAVAIL;
1606 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1607 if (ifp->scope > IFA_LINK)
1608 break;
1609 if (ifp->scope == IFA_LINK &&
1610 !(ifp->flags & banned_flags)) {
1611 *addr = ifp->addr;
1612 err = 0;
1613 break;
1616 return err;
1619 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1620 u32 banned_flags)
1622 struct inet6_dev *idev;
1623 int err = -EADDRNOTAVAIL;
1625 rcu_read_lock();
1626 idev = __in6_dev_get(dev);
1627 if (idev) {
1628 read_lock_bh(&idev->lock);
1629 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1630 read_unlock_bh(&idev->lock);
1632 rcu_read_unlock();
1633 return err;
1636 static int ipv6_count_addresses(struct inet6_dev *idev)
1638 int cnt = 0;
1639 struct inet6_ifaddr *ifp;
1641 read_lock_bh(&idev->lock);
1642 list_for_each_entry(ifp, &idev->addr_list, if_list)
1643 cnt++;
1644 read_unlock_bh(&idev->lock);
1645 return cnt;
1648 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1649 const struct net_device *dev, int strict)
1651 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1653 EXPORT_SYMBOL(ipv6_chk_addr);
1655 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1656 const struct net_device *dev, int strict,
1657 u32 banned_flags)
1659 struct inet6_ifaddr *ifp;
1660 unsigned int hash = inet6_addr_hash(addr);
1661 u32 ifp_flags;
1663 rcu_read_lock_bh();
1664 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1665 if (!net_eq(dev_net(ifp->idev->dev), net))
1666 continue;
1667 /* Decouple optimistic from tentative for evaluation here.
1668 * Ban optimistic addresses explicitly, when required.
1670 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1671 ? (ifp->flags&~IFA_F_TENTATIVE)
1672 : ifp->flags;
1673 if (ipv6_addr_equal(&ifp->addr, addr) &&
1674 !(ifp_flags&banned_flags) &&
1675 (!dev || ifp->idev->dev == dev ||
1676 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1677 rcu_read_unlock_bh();
1678 return 1;
1682 rcu_read_unlock_bh();
1683 return 0;
1685 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1687 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1688 struct net_device *dev)
1690 unsigned int hash = inet6_addr_hash(addr);
1691 struct inet6_ifaddr *ifp;
1693 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1694 if (!net_eq(dev_net(ifp->idev->dev), net))
1695 continue;
1696 if (ipv6_addr_equal(&ifp->addr, addr)) {
1697 if (!dev || ifp->idev->dev == dev)
1698 return true;
1701 return false;
1704 /* Compares an address/prefix_len with addresses on device @dev.
1705 * If one is found it returns true.
1707 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1708 const unsigned int prefix_len, struct net_device *dev)
1710 struct inet6_dev *idev;
1711 struct inet6_ifaddr *ifa;
1712 bool ret = false;
1714 rcu_read_lock();
1715 idev = __in6_dev_get(dev);
1716 if (idev) {
1717 read_lock_bh(&idev->lock);
1718 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1719 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1720 if (ret)
1721 break;
1723 read_unlock_bh(&idev->lock);
1725 rcu_read_unlock();
1727 return ret;
1729 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1731 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1733 struct inet6_dev *idev;
1734 struct inet6_ifaddr *ifa;
1735 int onlink;
1737 onlink = 0;
1738 rcu_read_lock();
1739 idev = __in6_dev_get(dev);
1740 if (idev) {
1741 read_lock_bh(&idev->lock);
1742 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1743 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1744 ifa->prefix_len);
1745 if (onlink)
1746 break;
1748 read_unlock_bh(&idev->lock);
1750 rcu_read_unlock();
1751 return onlink;
1753 EXPORT_SYMBOL(ipv6_chk_prefix);
1755 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1756 struct net_device *dev, int strict)
1758 struct inet6_ifaddr *ifp, *result = NULL;
1759 unsigned int hash = inet6_addr_hash(addr);
1761 rcu_read_lock_bh();
1762 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1763 if (!net_eq(dev_net(ifp->idev->dev), net))
1764 continue;
1765 if (ipv6_addr_equal(&ifp->addr, addr)) {
1766 if (!dev || ifp->idev->dev == dev ||
1767 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1768 result = ifp;
1769 in6_ifa_hold(ifp);
1770 break;
1774 rcu_read_unlock_bh();
1776 return result;
1779 /* Gets referenced address, destroys ifaddr */
1781 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1783 if (dad_failed)
1784 ifp->flags |= IFA_F_DADFAILED;
1786 if (ifp->flags&IFA_F_PERMANENT) {
1787 spin_lock_bh(&ifp->lock);
1788 addrconf_del_dad_work(ifp);
1789 ifp->flags |= IFA_F_TENTATIVE;
1790 spin_unlock_bh(&ifp->lock);
1791 if (dad_failed)
1792 ipv6_ifa_notify(0, ifp);
1793 in6_ifa_put(ifp);
1794 } else if (ifp->flags&IFA_F_TEMPORARY) {
1795 struct inet6_ifaddr *ifpub;
1796 spin_lock_bh(&ifp->lock);
1797 ifpub = ifp->ifpub;
1798 if (ifpub) {
1799 in6_ifa_hold(ifpub);
1800 spin_unlock_bh(&ifp->lock);
1801 ipv6_create_tempaddr(ifpub, ifp);
1802 in6_ifa_put(ifpub);
1803 } else {
1804 spin_unlock_bh(&ifp->lock);
1806 ipv6_del_addr(ifp);
1807 } else {
1808 ipv6_del_addr(ifp);
1812 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1814 int err = -ENOENT;
1816 spin_lock_bh(&ifp->lock);
1817 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1818 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1819 err = 0;
1821 spin_unlock_bh(&ifp->lock);
1823 return err;
1826 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1828 struct in6_addr addr;
1829 struct inet6_dev *idev = ifp->idev;
1830 struct net *net = dev_net(ifp->idev->dev);
1832 if (addrconf_dad_end(ifp)) {
1833 in6_ifa_put(ifp);
1834 return;
1837 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1838 ifp->idev->dev->name, &ifp->addr);
1840 spin_lock_bh(&ifp->lock);
1842 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1843 int scope = ifp->scope;
1844 u32 flags = ifp->flags;
1845 struct in6_addr new_addr;
1846 struct inet6_ifaddr *ifp2;
1847 u32 valid_lft, preferred_lft;
1848 int pfxlen = ifp->prefix_len;
1849 int retries = ifp->stable_privacy_retry + 1;
1851 if (retries > net->ipv6.sysctl.idgen_retries) {
1852 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1853 ifp->idev->dev->name);
1854 goto errdad;
1857 new_addr = ifp->addr;
1858 if (ipv6_generate_stable_address(&new_addr, retries,
1859 idev))
1860 goto errdad;
1862 valid_lft = ifp->valid_lft;
1863 preferred_lft = ifp->prefered_lft;
1865 spin_unlock_bh(&ifp->lock);
1867 if (idev->cnf.max_addresses &&
1868 ipv6_count_addresses(idev) >=
1869 idev->cnf.max_addresses)
1870 goto lock_errdad;
1872 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1873 ifp->idev->dev->name);
1875 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1876 scope, flags, valid_lft,
1877 preferred_lft);
1878 if (IS_ERR(ifp2))
1879 goto lock_errdad;
1881 spin_lock_bh(&ifp2->lock);
1882 ifp2->stable_privacy_retry = retries;
1883 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1884 spin_unlock_bh(&ifp2->lock);
1886 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1887 in6_ifa_put(ifp2);
1888 lock_errdad:
1889 spin_lock_bh(&ifp->lock);
1890 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1891 addr.s6_addr32[0] = htonl(0xfe800000);
1892 addr.s6_addr32[1] = 0;
1894 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1895 ipv6_addr_equal(&ifp->addr, &addr)) {
1896 /* DAD failed for link-local based on MAC address */
1897 idev->cnf.disable_ipv6 = 1;
1899 pr_info("%s: IPv6 being disabled!\n",
1900 ifp->idev->dev->name);
1904 errdad:
1905 /* transition from _POSTDAD to _ERRDAD */
1906 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1907 spin_unlock_bh(&ifp->lock);
1909 addrconf_mod_dad_work(ifp, 0);
1912 /* Join to solicited addr multicast group.
1913 * caller must hold RTNL */
1914 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1916 struct in6_addr maddr;
1918 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1919 return;
1921 addrconf_addr_solict_mult(addr, &maddr);
1922 ipv6_dev_mc_inc(dev, &maddr);
1925 /* caller must hold RTNL */
1926 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1928 struct in6_addr maddr;
1930 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1931 return;
1933 addrconf_addr_solict_mult(addr, &maddr);
1934 __ipv6_dev_mc_dec(idev, &maddr);
1937 /* caller must hold RTNL */
1938 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1940 struct in6_addr addr;
1942 if (ifp->prefix_len >= 127) /* RFC 6164 */
1943 return;
1944 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1945 if (ipv6_addr_any(&addr))
1946 return;
1947 __ipv6_dev_ac_inc(ifp->idev, &addr);
1950 /* caller must hold RTNL */
1951 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1953 struct in6_addr addr;
1955 if (ifp->prefix_len >= 127) /* RFC 6164 */
1956 return;
1957 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1958 if (ipv6_addr_any(&addr))
1959 return;
1960 __ipv6_dev_ac_dec(ifp->idev, &addr);
1963 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1965 if (dev->addr_len != EUI64_ADDR_LEN)
1966 return -1;
1967 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
1968 eui[0] ^= 2;
1969 return 0;
1972 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1974 union fwnet_hwaddr *ha;
1976 if (dev->addr_len != FWNET_ALEN)
1977 return -1;
1979 ha = (union fwnet_hwaddr *)dev->dev_addr;
1981 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1982 eui[0] ^= 2;
1983 return 0;
1986 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1988 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1989 if (dev->addr_len != ARCNET_ALEN)
1990 return -1;
1991 memset(eui, 0, 7);
1992 eui[7] = *(u8 *)dev->dev_addr;
1993 return 0;
1996 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1998 if (dev->addr_len != INFINIBAND_ALEN)
1999 return -1;
2000 memcpy(eui, dev->dev_addr + 12, 8);
2001 eui[0] |= 2;
2002 return 0;
2005 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2007 if (addr == 0)
2008 return -1;
2009 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2010 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2011 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2012 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2013 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2014 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2015 eui[1] = 0;
2016 eui[2] = 0x5E;
2017 eui[3] = 0xFE;
2018 memcpy(eui + 4, &addr, 4);
2019 return 0;
2022 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2024 if (dev->priv_flags & IFF_ISATAP)
2025 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2026 return -1;
2029 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2031 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2034 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2036 memcpy(eui, dev->perm_addr, 3);
2037 memcpy(eui + 5, dev->perm_addr + 3, 3);
2038 eui[3] = 0xFF;
2039 eui[4] = 0xFE;
2040 eui[0] ^= 2;
2041 return 0;
2044 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2046 switch (dev->type) {
2047 case ARPHRD_ETHER:
2048 case ARPHRD_FDDI:
2049 return addrconf_ifid_eui48(eui, dev);
2050 case ARPHRD_ARCNET:
2051 return addrconf_ifid_arcnet(eui, dev);
2052 case ARPHRD_INFINIBAND:
2053 return addrconf_ifid_infiniband(eui, dev);
2054 case ARPHRD_SIT:
2055 return addrconf_ifid_sit(eui, dev);
2056 case ARPHRD_IPGRE:
2057 return addrconf_ifid_gre(eui, dev);
2058 case ARPHRD_6LOWPAN:
2059 return addrconf_ifid_eui64(eui, dev);
2060 case ARPHRD_IEEE1394:
2061 return addrconf_ifid_ieee1394(eui, dev);
2062 case ARPHRD_TUNNEL6:
2063 return addrconf_ifid_ip6tnl(eui, dev);
2065 return -1;
2068 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2070 int err = -1;
2071 struct inet6_ifaddr *ifp;
2073 read_lock_bh(&idev->lock);
2074 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2075 if (ifp->scope > IFA_LINK)
2076 break;
2077 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2078 memcpy(eui, ifp->addr.s6_addr+8, 8);
2079 err = 0;
2080 break;
2083 read_unlock_bh(&idev->lock);
2084 return err;
2087 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2088 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2090 regen:
2091 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2092 idev->rndid[0] &= ~0x02;
2095 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2096 * check if generated address is not inappropriate
2098 * - Reserved subnet anycast (RFC 2526)
2099 * 11111101 11....11 1xxxxxxx
2100 * - ISATAP (RFC4214) 6.1
2101 * 00-00-5E-FE-xx-xx-xx-xx
2102 * - value 0
2103 * - XXX: already assigned to an address on the device
2105 if (idev->rndid[0] == 0xfd &&
2106 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2107 (idev->rndid[7]&0x80))
2108 goto regen;
2109 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2110 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2111 goto regen;
2112 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2113 goto regen;
2117 static void ipv6_regen_rndid(unsigned long data)
2119 struct inet6_dev *idev = (struct inet6_dev *) data;
2120 unsigned long expires;
2122 rcu_read_lock_bh();
2123 write_lock_bh(&idev->lock);
2125 if (idev->dead)
2126 goto out;
2128 __ipv6_regen_rndid(idev);
2130 expires = jiffies +
2131 idev->cnf.temp_prefered_lft * HZ -
2132 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2133 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2134 idev->cnf.max_desync_factor * HZ;
2135 if (time_before(expires, jiffies)) {
2136 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2137 __func__, idev->dev->name);
2138 goto out;
2141 if (!mod_timer(&idev->regen_timer, expires))
2142 in6_dev_hold(idev);
2144 out:
2145 write_unlock_bh(&idev->lock);
2146 rcu_read_unlock_bh();
2147 in6_dev_put(idev);
2150 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2152 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2153 __ipv6_regen_rndid(idev);
2157 * Add prefix route.
2160 static void
2161 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2162 unsigned long expires, u32 flags)
2164 struct fib6_config cfg = {
2165 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2166 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2167 .fc_ifindex = dev->ifindex,
2168 .fc_expires = expires,
2169 .fc_dst_len = plen,
2170 .fc_flags = RTF_UP | flags,
2171 .fc_nlinfo.nl_net = dev_net(dev),
2172 .fc_protocol = RTPROT_KERNEL,
2175 cfg.fc_dst = *pfx;
2177 /* Prevent useless cloning on PtP SIT.
2178 This thing is done here expecting that the whole
2179 class of non-broadcast devices need not cloning.
2181 #if IS_ENABLED(CONFIG_IPV6_SIT)
2182 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2183 cfg.fc_flags |= RTF_NONEXTHOP;
2184 #endif
2186 ip6_route_add(&cfg);
2190 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2191 int plen,
2192 const struct net_device *dev,
2193 u32 flags, u32 noflags)
2195 struct fib6_node *fn;
2196 struct rt6_info *rt = NULL;
2197 struct fib6_table *table;
2198 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2200 table = fib6_get_table(dev_net(dev), tb_id);
2201 if (!table)
2202 return NULL;
2204 read_lock_bh(&table->tb6_lock);
2205 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2206 if (!fn)
2207 goto out;
2209 noflags |= RTF_CACHE;
2210 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2211 if (rt->dst.dev->ifindex != dev->ifindex)
2212 continue;
2213 if ((rt->rt6i_flags & flags) != flags)
2214 continue;
2215 if ((rt->rt6i_flags & noflags) != 0)
2216 continue;
2217 dst_hold(&rt->dst);
2218 break;
2220 out:
2221 read_unlock_bh(&table->tb6_lock);
2222 return rt;
2226 /* Create "default" multicast route to the interface */
2228 static void addrconf_add_mroute(struct net_device *dev)
2230 struct fib6_config cfg = {
2231 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2232 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2233 .fc_ifindex = dev->ifindex,
2234 .fc_dst_len = 8,
2235 .fc_flags = RTF_UP,
2236 .fc_nlinfo.nl_net = dev_net(dev),
2239 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2241 ip6_route_add(&cfg);
2244 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2246 struct inet6_dev *idev;
2248 ASSERT_RTNL();
2250 idev = ipv6_find_idev(dev);
2251 if (!idev)
2252 return ERR_PTR(-ENOBUFS);
2254 if (idev->cnf.disable_ipv6)
2255 return ERR_PTR(-EACCES);
2257 /* Add default multicast route */
2258 if (!(dev->flags & IFF_LOOPBACK))
2259 addrconf_add_mroute(dev);
2261 return idev;
2264 static void manage_tempaddrs(struct inet6_dev *idev,
2265 struct inet6_ifaddr *ifp,
2266 __u32 valid_lft, __u32 prefered_lft,
2267 bool create, unsigned long now)
2269 u32 flags;
2270 struct inet6_ifaddr *ift;
2272 read_lock_bh(&idev->lock);
2273 /* update all temporary addresses in the list */
2274 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2275 int age, max_valid, max_prefered;
2277 if (ifp != ift->ifpub)
2278 continue;
2280 /* RFC 4941 section 3.3:
2281 * If a received option will extend the lifetime of a public
2282 * address, the lifetimes of temporary addresses should
2283 * be extended, subject to the overall constraint that no
2284 * temporary addresses should ever remain "valid" or "preferred"
2285 * for a time longer than (TEMP_VALID_LIFETIME) or
2286 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2288 age = (now - ift->cstamp) / HZ;
2289 max_valid = idev->cnf.temp_valid_lft - age;
2290 if (max_valid < 0)
2291 max_valid = 0;
2293 max_prefered = idev->cnf.temp_prefered_lft -
2294 idev->cnf.max_desync_factor - age;
2295 if (max_prefered < 0)
2296 max_prefered = 0;
2298 if (valid_lft > max_valid)
2299 valid_lft = max_valid;
2301 if (prefered_lft > max_prefered)
2302 prefered_lft = max_prefered;
2304 spin_lock(&ift->lock);
2305 flags = ift->flags;
2306 ift->valid_lft = valid_lft;
2307 ift->prefered_lft = prefered_lft;
2308 ift->tstamp = now;
2309 if (prefered_lft > 0)
2310 ift->flags &= ~IFA_F_DEPRECATED;
2312 spin_unlock(&ift->lock);
2313 if (!(flags&IFA_F_TENTATIVE))
2314 ipv6_ifa_notify(0, ift);
2317 if ((create || list_empty(&idev->tempaddr_list)) &&
2318 idev->cnf.use_tempaddr > 0) {
2319 /* When a new public address is created as described
2320 * in [ADDRCONF], also create a new temporary address.
2321 * Also create a temporary address if it's enabled but
2322 * no temporary address currently exists.
2324 read_unlock_bh(&idev->lock);
2325 ipv6_create_tempaddr(ifp, NULL);
2326 } else {
2327 read_unlock_bh(&idev->lock);
2331 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2333 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2334 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2337 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2339 struct prefix_info *pinfo;
2340 __u32 valid_lft;
2341 __u32 prefered_lft;
2342 int addr_type;
2343 u32 addr_flags = 0;
2344 struct inet6_dev *in6_dev;
2345 struct net *net = dev_net(dev);
2347 pinfo = (struct prefix_info *) opt;
2349 if (len < sizeof(struct prefix_info)) {
2350 ADBG("addrconf: prefix option too short\n");
2351 return;
2355 * Validation checks ([ADDRCONF], page 19)
2358 addr_type = ipv6_addr_type(&pinfo->prefix);
2360 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2361 return;
2363 valid_lft = ntohl(pinfo->valid);
2364 prefered_lft = ntohl(pinfo->prefered);
2366 if (prefered_lft > valid_lft) {
2367 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2368 return;
2371 in6_dev = in6_dev_get(dev);
2373 if (!in6_dev) {
2374 net_dbg_ratelimited("addrconf: device %s not configured\n",
2375 dev->name);
2376 return;
2380 * Two things going on here:
2381 * 1) Add routes for on-link prefixes
2382 * 2) Configure prefixes with the auto flag set
2385 if (pinfo->onlink) {
2386 struct rt6_info *rt;
2387 unsigned long rt_expires;
2389 /* Avoid arithmetic overflow. Really, we could
2390 * save rt_expires in seconds, likely valid_lft,
2391 * but it would require division in fib gc, that it
2392 * not good.
2394 if (HZ > USER_HZ)
2395 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2396 else
2397 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2399 if (addrconf_finite_timeout(rt_expires))
2400 rt_expires *= HZ;
2402 rt = addrconf_get_prefix_route(&pinfo->prefix,
2403 pinfo->prefix_len,
2404 dev,
2405 RTF_ADDRCONF | RTF_PREFIX_RT,
2406 RTF_GATEWAY | RTF_DEFAULT);
2408 if (rt) {
2409 /* Autoconf prefix route */
2410 if (valid_lft == 0) {
2411 ip6_del_rt(rt);
2412 rt = NULL;
2413 } else if (addrconf_finite_timeout(rt_expires)) {
2414 /* not infinity */
2415 rt6_set_expires(rt, jiffies + rt_expires);
2416 } else {
2417 rt6_clean_expires(rt);
2419 } else if (valid_lft) {
2420 clock_t expires = 0;
2421 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2422 if (addrconf_finite_timeout(rt_expires)) {
2423 /* not infinity */
2424 flags |= RTF_EXPIRES;
2425 expires = jiffies_to_clock_t(rt_expires);
2427 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2428 dev, expires, flags);
2430 ip6_rt_put(rt);
2433 /* Try to figure out our local address for this prefix */
2435 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2436 struct inet6_ifaddr *ifp;
2437 struct in6_addr addr;
2438 int create = 0, update_lft = 0;
2439 bool tokenized = false;
2441 if (pinfo->prefix_len == 64) {
2442 memcpy(&addr, &pinfo->prefix, 8);
2444 if (!ipv6_addr_any(&in6_dev->token)) {
2445 read_lock_bh(&in6_dev->lock);
2446 memcpy(addr.s6_addr + 8,
2447 in6_dev->token.s6_addr + 8, 8);
2448 read_unlock_bh(&in6_dev->lock);
2449 tokenized = true;
2450 } else if (is_addr_mode_generate_stable(in6_dev) &&
2451 !ipv6_generate_stable_address(&addr, 0,
2452 in6_dev)) {
2453 addr_flags |= IFA_F_STABLE_PRIVACY;
2454 goto ok;
2455 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2456 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2457 in6_dev_put(in6_dev);
2458 return;
2460 goto ok;
2462 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2463 pinfo->prefix_len);
2464 in6_dev_put(in6_dev);
2465 return;
2469 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2471 if (!ifp && valid_lft) {
2472 int max_addresses = in6_dev->cnf.max_addresses;
2474 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2475 if (in6_dev->cnf.optimistic_dad &&
2476 !net->ipv6.devconf_all->forwarding && sllao)
2477 addr_flags |= IFA_F_OPTIMISTIC;
2478 #endif
2480 /* Do not allow to create too much of autoconfigured
2481 * addresses; this would be too easy way to crash kernel.
2483 if (!max_addresses ||
2484 ipv6_count_addresses(in6_dev) < max_addresses)
2485 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2486 pinfo->prefix_len,
2487 addr_type&IPV6_ADDR_SCOPE_MASK,
2488 addr_flags, valid_lft,
2489 prefered_lft);
2491 if (IS_ERR_OR_NULL(ifp)) {
2492 in6_dev_put(in6_dev);
2493 return;
2496 update_lft = 0;
2497 create = 1;
2498 spin_lock_bh(&ifp->lock);
2499 ifp->flags |= IFA_F_MANAGETEMPADDR;
2500 ifp->cstamp = jiffies;
2501 ifp->tokenized = tokenized;
2502 spin_unlock_bh(&ifp->lock);
2503 addrconf_dad_start(ifp);
2506 if (ifp) {
2507 u32 flags;
2508 unsigned long now;
2509 u32 stored_lft;
2511 /* update lifetime (RFC2462 5.5.3 e) */
2512 spin_lock_bh(&ifp->lock);
2513 now = jiffies;
2514 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2515 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2516 else
2517 stored_lft = 0;
2518 if (!update_lft && !create && stored_lft) {
2519 const u32 minimum_lft = min_t(u32,
2520 stored_lft, MIN_VALID_LIFETIME);
2521 valid_lft = max(valid_lft, minimum_lft);
2523 /* RFC4862 Section 5.5.3e:
2524 * "Note that the preferred lifetime of the
2525 * corresponding address is always reset to
2526 * the Preferred Lifetime in the received
2527 * Prefix Information option, regardless of
2528 * whether the valid lifetime is also reset or
2529 * ignored."
2531 * So we should always update prefered_lft here.
2533 update_lft = 1;
2536 if (update_lft) {
2537 ifp->valid_lft = valid_lft;
2538 ifp->prefered_lft = prefered_lft;
2539 ifp->tstamp = now;
2540 flags = ifp->flags;
2541 ifp->flags &= ~IFA_F_DEPRECATED;
2542 spin_unlock_bh(&ifp->lock);
2544 if (!(flags&IFA_F_TENTATIVE))
2545 ipv6_ifa_notify(0, ifp);
2546 } else
2547 spin_unlock_bh(&ifp->lock);
2549 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2550 create, now);
2552 in6_ifa_put(ifp);
2553 addrconf_verify();
2556 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2557 in6_dev_put(in6_dev);
2561 * Set destination address.
2562 * Special case for SIT interfaces where we create a new "virtual"
2563 * device.
2565 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2567 struct in6_ifreq ireq;
2568 struct net_device *dev;
2569 int err = -EINVAL;
2571 rtnl_lock();
2573 err = -EFAULT;
2574 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2575 goto err_exit;
2577 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2579 err = -ENODEV;
2580 if (!dev)
2581 goto err_exit;
2583 #if IS_ENABLED(CONFIG_IPV6_SIT)
2584 if (dev->type == ARPHRD_SIT) {
2585 const struct net_device_ops *ops = dev->netdev_ops;
2586 struct ifreq ifr;
2587 struct ip_tunnel_parm p;
2589 err = -EADDRNOTAVAIL;
2590 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2591 goto err_exit;
2593 memset(&p, 0, sizeof(p));
2594 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2595 p.iph.saddr = 0;
2596 p.iph.version = 4;
2597 p.iph.ihl = 5;
2598 p.iph.protocol = IPPROTO_IPV6;
2599 p.iph.ttl = 64;
2600 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2602 if (ops->ndo_do_ioctl) {
2603 mm_segment_t oldfs = get_fs();
2605 set_fs(KERNEL_DS);
2606 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2607 set_fs(oldfs);
2608 } else
2609 err = -EOPNOTSUPP;
2611 if (err == 0) {
2612 err = -ENOBUFS;
2613 dev = __dev_get_by_name(net, p.name);
2614 if (!dev)
2615 goto err_exit;
2616 err = dev_open(dev);
2619 #endif
2621 err_exit:
2622 rtnl_unlock();
2623 return err;
2626 static int ipv6_mc_config(struct sock *sk, bool join,
2627 const struct in6_addr *addr, int ifindex)
2629 int ret;
2631 ASSERT_RTNL();
2633 lock_sock(sk);
2634 if (join)
2635 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2636 else
2637 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2638 release_sock(sk);
2640 return ret;
2644 * Manual configuration of address on an interface
2646 static int inet6_addr_add(struct net *net, int ifindex,
2647 const struct in6_addr *pfx,
2648 const struct in6_addr *peer_pfx,
2649 unsigned int plen, __u32 ifa_flags,
2650 __u32 prefered_lft, __u32 valid_lft)
2652 struct inet6_ifaddr *ifp;
2653 struct inet6_dev *idev;
2654 struct net_device *dev;
2655 unsigned long timeout;
2656 clock_t expires;
2657 int scope;
2658 u32 flags;
2660 ASSERT_RTNL();
2662 if (plen > 128)
2663 return -EINVAL;
2665 /* check the lifetime */
2666 if (!valid_lft || prefered_lft > valid_lft)
2667 return -EINVAL;
2669 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2670 return -EINVAL;
2672 dev = __dev_get_by_index(net, ifindex);
2673 if (!dev)
2674 return -ENODEV;
2676 idev = addrconf_add_dev(dev);
2677 if (IS_ERR(idev))
2678 return PTR_ERR(idev);
2680 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2681 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2682 true, pfx, ifindex);
2684 if (ret < 0)
2685 return ret;
2688 scope = ipv6_addr_scope(pfx);
2690 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2691 if (addrconf_finite_timeout(timeout)) {
2692 expires = jiffies_to_clock_t(timeout * HZ);
2693 valid_lft = timeout;
2694 flags = RTF_EXPIRES;
2695 } else {
2696 expires = 0;
2697 flags = 0;
2698 ifa_flags |= IFA_F_PERMANENT;
2701 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2702 if (addrconf_finite_timeout(timeout)) {
2703 if (timeout == 0)
2704 ifa_flags |= IFA_F_DEPRECATED;
2705 prefered_lft = timeout;
2708 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2709 valid_lft, prefered_lft);
2711 if (!IS_ERR(ifp)) {
2712 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2713 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2714 expires, flags);
2718 * Note that section 3.1 of RFC 4429 indicates
2719 * that the Optimistic flag should not be set for
2720 * manually configured addresses
2722 addrconf_dad_start(ifp);
2723 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2724 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2725 true, jiffies);
2726 in6_ifa_put(ifp);
2727 addrconf_verify_rtnl();
2728 return 0;
2729 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2730 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2731 false, pfx, ifindex);
2734 return PTR_ERR(ifp);
2737 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2738 const struct in6_addr *pfx, unsigned int plen)
2740 struct inet6_ifaddr *ifp;
2741 struct inet6_dev *idev;
2742 struct net_device *dev;
2744 if (plen > 128)
2745 return -EINVAL;
2747 dev = __dev_get_by_index(net, ifindex);
2748 if (!dev)
2749 return -ENODEV;
2751 idev = __in6_dev_get(dev);
2752 if (!idev)
2753 return -ENXIO;
2755 read_lock_bh(&idev->lock);
2756 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2757 if (ifp->prefix_len == plen &&
2758 ipv6_addr_equal(pfx, &ifp->addr)) {
2759 in6_ifa_hold(ifp);
2760 read_unlock_bh(&idev->lock);
2762 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2763 (ifa_flags & IFA_F_MANAGETEMPADDR))
2764 manage_tempaddrs(idev, ifp, 0, 0, false,
2765 jiffies);
2766 ipv6_del_addr(ifp);
2767 addrconf_verify_rtnl();
2768 if (ipv6_addr_is_multicast(pfx)) {
2769 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2770 false, pfx, dev->ifindex);
2772 return 0;
2775 read_unlock_bh(&idev->lock);
2776 return -EADDRNOTAVAIL;
2780 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2782 struct in6_ifreq ireq;
2783 int err;
2785 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2786 return -EPERM;
2788 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2789 return -EFAULT;
2791 rtnl_lock();
2792 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2793 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2794 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2795 rtnl_unlock();
2796 return err;
2799 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2801 struct in6_ifreq ireq;
2802 int err;
2804 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2805 return -EPERM;
2807 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2808 return -EFAULT;
2810 rtnl_lock();
2811 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2812 ireq.ifr6_prefixlen);
2813 rtnl_unlock();
2814 return err;
2817 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2818 int plen, int scope)
2820 struct inet6_ifaddr *ifp;
2822 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2823 scope, IFA_F_PERMANENT,
2824 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2825 if (!IS_ERR(ifp)) {
2826 spin_lock_bh(&ifp->lock);
2827 ifp->flags &= ~IFA_F_TENTATIVE;
2828 spin_unlock_bh(&ifp->lock);
2829 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2830 in6_ifa_put(ifp);
2834 #if IS_ENABLED(CONFIG_IPV6_SIT)
2835 static void sit_add_v4_addrs(struct inet6_dev *idev)
2837 struct in6_addr addr;
2838 struct net_device *dev;
2839 struct net *net = dev_net(idev->dev);
2840 int scope, plen;
2841 u32 pflags = 0;
2843 ASSERT_RTNL();
2845 memset(&addr, 0, sizeof(struct in6_addr));
2846 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2848 if (idev->dev->flags&IFF_POINTOPOINT) {
2849 addr.s6_addr32[0] = htonl(0xfe800000);
2850 scope = IFA_LINK;
2851 plen = 64;
2852 } else {
2853 scope = IPV6_ADDR_COMPATv4;
2854 plen = 96;
2855 pflags |= RTF_NONEXTHOP;
2858 if (addr.s6_addr32[3]) {
2859 add_addr(idev, &addr, plen, scope);
2860 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2861 return;
2864 for_each_netdev(net, dev) {
2865 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2866 if (in_dev && (dev->flags & IFF_UP)) {
2867 struct in_ifaddr *ifa;
2869 int flag = scope;
2871 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2873 addr.s6_addr32[3] = ifa->ifa_local;
2875 if (ifa->ifa_scope == RT_SCOPE_LINK)
2876 continue;
2877 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2878 if (idev->dev->flags&IFF_POINTOPOINT)
2879 continue;
2880 flag |= IFA_HOST;
2883 add_addr(idev, &addr, plen, flag);
2884 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2885 pflags);
2890 #endif
2892 static void init_loopback(struct net_device *dev)
2894 struct inet6_dev *idev;
2895 struct net_device *sp_dev;
2896 struct inet6_ifaddr *sp_ifa;
2897 struct rt6_info *sp_rt;
2899 /* ::1 */
2901 ASSERT_RTNL();
2903 idev = ipv6_find_idev(dev);
2904 if (!idev) {
2905 pr_debug("%s: add_dev failed\n", __func__);
2906 return;
2909 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2911 /* Add routes to other interface's IPv6 addresses */
2912 for_each_netdev(dev_net(dev), sp_dev) {
2913 if (!strcmp(sp_dev->name, dev->name))
2914 continue;
2916 idev = __in6_dev_get(sp_dev);
2917 if (!idev)
2918 continue;
2920 read_lock_bh(&idev->lock);
2921 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2923 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2924 continue;
2926 if (sp_ifa->rt) {
2927 /* This dst has been added to garbage list when
2928 * lo device down, release this obsolete dst and
2929 * reallocate a new router for ifa.
2931 if (sp_ifa->rt->dst.obsolete > 0) {
2932 ip6_rt_put(sp_ifa->rt);
2933 sp_ifa->rt = NULL;
2934 } else {
2935 continue;
2939 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2941 /* Failure cases are ignored */
2942 if (!IS_ERR(sp_rt)) {
2943 sp_ifa->rt = sp_rt;
2944 ip6_ins_rt(sp_rt);
2947 read_unlock_bh(&idev->lock);
2951 static void addrconf_add_linklocal(struct inet6_dev *idev,
2952 const struct in6_addr *addr, u32 flags)
2954 struct inet6_ifaddr *ifp;
2955 u32 addr_flags = flags | IFA_F_PERMANENT;
2957 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2958 if (idev->cnf.optimistic_dad &&
2959 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2960 addr_flags |= IFA_F_OPTIMISTIC;
2961 #endif
2963 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2964 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2965 if (!IS_ERR(ifp)) {
2966 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2967 addrconf_dad_start(ifp);
2968 in6_ifa_put(ifp);
2972 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2974 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2975 return true;
2977 if (address.s6_addr32[2] == htonl(0x02005eff) &&
2978 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2979 return true;
2981 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2982 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2983 return true;
2985 return false;
2988 static int ipv6_generate_stable_address(struct in6_addr *address,
2989 u8 dad_count,
2990 const struct inet6_dev *idev)
2992 static DEFINE_SPINLOCK(lock);
2993 static __u32 digest[SHA_DIGEST_WORDS];
2994 static __u32 workspace[SHA_WORKSPACE_WORDS];
2996 static union {
2997 char __data[SHA_MESSAGE_BYTES];
2998 struct {
2999 struct in6_addr secret;
3000 __be32 prefix[2];
3001 unsigned char hwaddr[MAX_ADDR_LEN];
3002 u8 dad_count;
3003 } __packed;
3004 } data;
3006 struct in6_addr secret;
3007 struct in6_addr temp;
3008 struct net *net = dev_net(idev->dev);
3010 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3012 if (idev->cnf.stable_secret.initialized)
3013 secret = idev->cnf.stable_secret.secret;
3014 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3015 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3016 else
3017 return -1;
3019 retry:
3020 spin_lock_bh(&lock);
3022 sha_init(digest);
3023 memset(&data, 0, sizeof(data));
3024 memset(workspace, 0, sizeof(workspace));
3025 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3026 data.prefix[0] = address->s6_addr32[0];
3027 data.prefix[1] = address->s6_addr32[1];
3028 data.secret = secret;
3029 data.dad_count = dad_count;
3031 sha_transform(digest, data.__data, workspace);
3033 temp = *address;
3034 temp.s6_addr32[2] = (__force __be32)digest[0];
3035 temp.s6_addr32[3] = (__force __be32)digest[1];
3037 spin_unlock_bh(&lock);
3039 if (ipv6_reserved_interfaceid(temp)) {
3040 dad_count++;
3041 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3042 return -1;
3043 goto retry;
3046 *address = temp;
3047 return 0;
3050 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3052 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3054 if (s->initialized)
3055 return;
3056 s = &idev->cnf.stable_secret;
3057 get_random_bytes(&s->secret, sizeof(s->secret));
3058 s->initialized = true;
3061 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3063 struct in6_addr addr;
3065 /* no link local addresses on L3 master devices */
3066 if (netif_is_l3_master(idev->dev))
3067 return;
3069 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3071 switch (idev->addr_gen_mode) {
3072 case IN6_ADDR_GEN_MODE_RANDOM:
3073 ipv6_gen_mode_random_init(idev);
3074 /* fallthrough */
3075 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3076 if (!ipv6_generate_stable_address(&addr, 0, idev))
3077 addrconf_add_linklocal(idev, &addr,
3078 IFA_F_STABLE_PRIVACY);
3079 else if (prefix_route)
3080 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3081 break;
3082 case IN6_ADDR_GEN_MODE_EUI64:
3083 /* addrconf_add_linklocal also adds a prefix_route and we
3084 * only need to care about prefix routes if ipv6_generate_eui64
3085 * couldn't generate one.
3087 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3088 addrconf_add_linklocal(idev, &addr, 0);
3089 else if (prefix_route)
3090 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3091 break;
3092 case IN6_ADDR_GEN_MODE_NONE:
3093 default:
3094 /* will not add any link local address */
3095 break;
3099 static void addrconf_dev_config(struct net_device *dev)
3101 struct inet6_dev *idev;
3103 ASSERT_RTNL();
3105 if ((dev->type != ARPHRD_ETHER) &&
3106 (dev->type != ARPHRD_FDDI) &&
3107 (dev->type != ARPHRD_ARCNET) &&
3108 (dev->type != ARPHRD_INFINIBAND) &&
3109 (dev->type != ARPHRD_IEEE1394) &&
3110 (dev->type != ARPHRD_TUNNEL6) &&
3111 (dev->type != ARPHRD_6LOWPAN) &&
3112 (dev->type != ARPHRD_NONE)) {
3113 /* Alas, we support only Ethernet autoconfiguration. */
3114 return;
3117 idev = addrconf_add_dev(dev);
3118 if (IS_ERR(idev))
3119 return;
3121 /* this device type has no EUI support */
3122 if (dev->type == ARPHRD_NONE &&
3123 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3124 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3126 addrconf_addr_gen(idev, false);
3129 #if IS_ENABLED(CONFIG_IPV6_SIT)
3130 static void addrconf_sit_config(struct net_device *dev)
3132 struct inet6_dev *idev;
3134 ASSERT_RTNL();
3137 * Configure the tunnel with one of our IPv4
3138 * addresses... we should configure all of
3139 * our v4 addrs in the tunnel
3142 idev = ipv6_find_idev(dev);
3143 if (!idev) {
3144 pr_debug("%s: add_dev failed\n", __func__);
3145 return;
3148 if (dev->priv_flags & IFF_ISATAP) {
3149 addrconf_addr_gen(idev, false);
3150 return;
3153 sit_add_v4_addrs(idev);
3155 if (dev->flags&IFF_POINTOPOINT)
3156 addrconf_add_mroute(dev);
3158 #endif
3160 #if IS_ENABLED(CONFIG_NET_IPGRE)
3161 static void addrconf_gre_config(struct net_device *dev)
3163 struct inet6_dev *idev;
3165 ASSERT_RTNL();
3167 idev = ipv6_find_idev(dev);
3168 if (!idev) {
3169 pr_debug("%s: add_dev failed\n", __func__);
3170 return;
3173 addrconf_addr_gen(idev, true);
3174 if (dev->flags & IFF_POINTOPOINT)
3175 addrconf_add_mroute(dev);
3177 #endif
3179 static int fixup_permanent_addr(struct inet6_dev *idev,
3180 struct inet6_ifaddr *ifp)
3182 if (!ifp->rt) {
3183 struct rt6_info *rt;
3185 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3186 if (unlikely(IS_ERR(rt)))
3187 return PTR_ERR(rt);
3189 ifp->rt = rt;
3192 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3193 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3194 idev->dev, 0, 0);
3197 addrconf_dad_start(ifp);
3199 return 0;
3202 static void addrconf_permanent_addr(struct net_device *dev)
3204 struct inet6_ifaddr *ifp, *tmp;
3205 struct inet6_dev *idev;
3207 idev = __in6_dev_get(dev);
3208 if (!idev)
3209 return;
3211 write_lock_bh(&idev->lock);
3213 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3214 if ((ifp->flags & IFA_F_PERMANENT) &&
3215 fixup_permanent_addr(idev, ifp) < 0) {
3216 write_unlock_bh(&idev->lock);
3217 ipv6_del_addr(ifp);
3218 write_lock_bh(&idev->lock);
3220 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3221 idev->dev->name, &ifp->addr);
3225 write_unlock_bh(&idev->lock);
3228 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3229 void *ptr)
3231 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3232 struct netdev_notifier_changeupper_info *info;
3233 struct inet6_dev *idev = __in6_dev_get(dev);
3234 int run_pending = 0;
3235 int err;
3237 switch (event) {
3238 case NETDEV_REGISTER:
3239 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3240 idev = ipv6_add_dev(dev);
3241 if (IS_ERR(idev))
3242 return notifier_from_errno(PTR_ERR(idev));
3244 break;
3246 case NETDEV_CHANGEMTU:
3247 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3248 if (dev->mtu < IPV6_MIN_MTU) {
3249 addrconf_ifdown(dev, 1);
3250 break;
3253 if (idev) {
3254 rt6_mtu_change(dev, dev->mtu);
3255 idev->cnf.mtu6 = dev->mtu;
3256 break;
3259 /* allocate new idev */
3260 idev = ipv6_add_dev(dev);
3261 if (IS_ERR(idev))
3262 break;
3264 /* device is still not ready */
3265 if (!(idev->if_flags & IF_READY))
3266 break;
3268 run_pending = 1;
3270 /* fall through */
3272 case NETDEV_UP:
3273 case NETDEV_CHANGE:
3274 if (dev->flags & IFF_SLAVE)
3275 break;
3277 if (idev && idev->cnf.disable_ipv6)
3278 break;
3280 if (event == NETDEV_UP) {
3281 /* restore routes for permanent addresses */
3282 addrconf_permanent_addr(dev);
3284 if (!addrconf_qdisc_ok(dev)) {
3285 /* device is not ready yet. */
3286 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3287 dev->name);
3288 break;
3291 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3292 idev = ipv6_add_dev(dev);
3294 if (!IS_ERR_OR_NULL(idev)) {
3295 idev->if_flags |= IF_READY;
3296 run_pending = 1;
3298 } else if (event == NETDEV_CHANGE) {
3299 if (!addrconf_qdisc_ok(dev)) {
3300 /* device is still not ready. */
3301 break;
3304 if (idev) {
3305 if (idev->if_flags & IF_READY)
3306 /* device is already configured. */
3307 break;
3308 idev->if_flags |= IF_READY;
3311 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3312 dev->name);
3314 run_pending = 1;
3317 switch (dev->type) {
3318 #if IS_ENABLED(CONFIG_IPV6_SIT)
3319 case ARPHRD_SIT:
3320 addrconf_sit_config(dev);
3321 break;
3322 #endif
3323 #if IS_ENABLED(CONFIG_NET_IPGRE)
3324 case ARPHRD_IPGRE:
3325 addrconf_gre_config(dev);
3326 break;
3327 #endif
3328 case ARPHRD_LOOPBACK:
3329 init_loopback(dev);
3330 break;
3332 default:
3333 addrconf_dev_config(dev);
3334 break;
3337 if (!IS_ERR_OR_NULL(idev)) {
3338 if (run_pending)
3339 addrconf_dad_run(idev);
3342 * If the MTU changed during the interface down,
3343 * when the interface up, the changed MTU must be
3344 * reflected in the idev as well as routers.
3346 if (idev->cnf.mtu6 != dev->mtu &&
3347 dev->mtu >= IPV6_MIN_MTU) {
3348 rt6_mtu_change(dev, dev->mtu);
3349 idev->cnf.mtu6 = dev->mtu;
3351 idev->tstamp = jiffies;
3352 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3355 * If the changed mtu during down is lower than
3356 * IPV6_MIN_MTU stop IPv6 on this interface.
3358 if (dev->mtu < IPV6_MIN_MTU)
3359 addrconf_ifdown(dev, 1);
3361 break;
3363 case NETDEV_DOWN:
3364 case NETDEV_UNREGISTER:
3366 * Remove all addresses from this interface.
3368 addrconf_ifdown(dev, event != NETDEV_DOWN);
3369 break;
3371 case NETDEV_CHANGENAME:
3372 if (idev) {
3373 snmp6_unregister_dev(idev);
3374 addrconf_sysctl_unregister(idev);
3375 err = addrconf_sysctl_register(idev);
3376 if (err)
3377 return notifier_from_errno(err);
3378 err = snmp6_register_dev(idev);
3379 if (err) {
3380 addrconf_sysctl_unregister(idev);
3381 return notifier_from_errno(err);
3384 break;
3386 case NETDEV_PRE_TYPE_CHANGE:
3387 case NETDEV_POST_TYPE_CHANGE:
3388 if (idev)
3389 addrconf_type_change(dev, event);
3390 break;
3392 case NETDEV_CHANGEUPPER:
3393 info = ptr;
3395 /* flush all routes if dev is linked to or unlinked from
3396 * an L3 master device (e.g., VRF)
3398 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3399 addrconf_ifdown(dev, 0);
3402 return NOTIFY_OK;
3406 * addrconf module should be notified of a device going up
3408 static struct notifier_block ipv6_dev_notf = {
3409 .notifier_call = addrconf_notify,
3412 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3414 struct inet6_dev *idev;
3415 ASSERT_RTNL();
3417 idev = __in6_dev_get(dev);
3419 if (event == NETDEV_POST_TYPE_CHANGE)
3420 ipv6_mc_remap(idev);
3421 else if (event == NETDEV_PRE_TYPE_CHANGE)
3422 ipv6_mc_unmap(idev);
3425 static bool addr_is_local(const struct in6_addr *addr)
3427 return ipv6_addr_type(addr) &
3428 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3431 static int addrconf_ifdown(struct net_device *dev, int how)
3433 struct net *net = dev_net(dev);
3434 struct inet6_dev *idev;
3435 struct inet6_ifaddr *ifa, *tmp;
3436 struct list_head del_list;
3437 int _keep_addr;
3438 bool keep_addr;
3439 int state, i;
3441 ASSERT_RTNL();
3443 rt6_ifdown(net, dev);
3444 neigh_ifdown(&nd_tbl, dev);
3446 idev = __in6_dev_get(dev);
3447 if (!idev)
3448 return -ENODEV;
3451 * Step 1: remove reference to ipv6 device from parent device.
3452 * Do not dev_put!
3454 if (how) {
3455 idev->dead = 1;
3457 /* protected by rtnl_lock */
3458 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3460 /* Step 1.5: remove snmp6 entry */
3461 snmp6_unregister_dev(idev);
3465 /* aggregate the system setting and interface setting */
3466 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3467 if (!_keep_addr)
3468 _keep_addr = idev->cnf.keep_addr_on_down;
3470 /* combine the user config with event to determine if permanent
3471 * addresses are to be removed from address hash table
3473 keep_addr = !(how || _keep_addr <= 0);
3475 /* Step 2: clear hash table */
3476 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3477 struct hlist_head *h = &inet6_addr_lst[i];
3479 spin_lock_bh(&addrconf_hash_lock);
3480 restart:
3481 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3482 if (ifa->idev == idev) {
3483 addrconf_del_dad_work(ifa);
3484 /* combined flag + permanent flag decide if
3485 * address is retained on a down event
3487 if (!keep_addr ||
3488 !(ifa->flags & IFA_F_PERMANENT) ||
3489 addr_is_local(&ifa->addr)) {
3490 hlist_del_init_rcu(&ifa->addr_lst);
3491 goto restart;
3495 spin_unlock_bh(&addrconf_hash_lock);
3498 write_lock_bh(&idev->lock);
3500 addrconf_del_rs_timer(idev);
3502 /* Step 2: clear flags for stateless addrconf */
3503 if (!how)
3504 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3506 if (how && del_timer(&idev->regen_timer))
3507 in6_dev_put(idev);
3509 /* Step 3: clear tempaddr list */
3510 while (!list_empty(&idev->tempaddr_list)) {
3511 ifa = list_first_entry(&idev->tempaddr_list,
3512 struct inet6_ifaddr, tmp_list);
3513 list_del(&ifa->tmp_list);
3514 write_unlock_bh(&idev->lock);
3515 spin_lock_bh(&ifa->lock);
3517 if (ifa->ifpub) {
3518 in6_ifa_put(ifa->ifpub);
3519 ifa->ifpub = NULL;
3521 spin_unlock_bh(&ifa->lock);
3522 in6_ifa_put(ifa);
3523 write_lock_bh(&idev->lock);
3526 /* re-combine the user config with event to determine if permanent
3527 * addresses are to be removed from the interface list
3529 keep_addr = (!how && _keep_addr > 0);
3531 INIT_LIST_HEAD(&del_list);
3532 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3533 struct rt6_info *rt = NULL;
3535 addrconf_del_dad_work(ifa);
3537 write_unlock_bh(&idev->lock);
3538 spin_lock_bh(&ifa->lock);
3540 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3541 !addr_is_local(&ifa->addr)) {
3542 /* set state to skip the notifier below */
3543 state = INET6_IFADDR_STATE_DEAD;
3544 ifa->state = 0;
3545 if (!(ifa->flags & IFA_F_NODAD))
3546 ifa->flags |= IFA_F_TENTATIVE;
3548 rt = ifa->rt;
3549 ifa->rt = NULL;
3550 } else {
3551 state = ifa->state;
3552 ifa->state = INET6_IFADDR_STATE_DEAD;
3554 list_del(&ifa->if_list);
3555 list_add(&ifa->if_list, &del_list);
3558 spin_unlock_bh(&ifa->lock);
3560 if (rt)
3561 ip6_del_rt(rt);
3563 if (state != INET6_IFADDR_STATE_DEAD) {
3564 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3565 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3568 write_lock_bh(&idev->lock);
3571 write_unlock_bh(&idev->lock);
3573 /* now clean up addresses to be removed */
3574 while (!list_empty(&del_list)) {
3575 ifa = list_first_entry(&del_list,
3576 struct inet6_ifaddr, if_list);
3577 list_del(&ifa->if_list);
3579 in6_ifa_put(ifa);
3582 /* Step 5: Discard anycast and multicast list */
3583 if (how) {
3584 ipv6_ac_destroy_dev(idev);
3585 ipv6_mc_destroy_dev(idev);
3586 } else {
3587 ipv6_mc_down(idev);
3590 idev->tstamp = jiffies;
3592 /* Last: Shot the device (if unregistered) */
3593 if (how) {
3594 addrconf_sysctl_unregister(idev);
3595 neigh_parms_release(&nd_tbl, idev->nd_parms);
3596 neigh_ifdown(&nd_tbl, dev);
3597 in6_dev_put(idev);
3599 return 0;
3602 static void addrconf_rs_timer(unsigned long data)
3604 struct inet6_dev *idev = (struct inet6_dev *)data;
3605 struct net_device *dev = idev->dev;
3606 struct in6_addr lladdr;
3608 write_lock(&idev->lock);
3609 if (idev->dead || !(idev->if_flags & IF_READY))
3610 goto out;
3612 if (!ipv6_accept_ra(idev))
3613 goto out;
3615 /* Announcement received after solicitation was sent */
3616 if (idev->if_flags & IF_RA_RCVD)
3617 goto out;
3619 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3620 write_unlock(&idev->lock);
3621 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3622 ndisc_send_rs(dev, &lladdr,
3623 &in6addr_linklocal_allrouters);
3624 else
3625 goto put;
3627 write_lock(&idev->lock);
3628 /* The wait after the last probe can be shorter */
3629 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3630 idev->cnf.rtr_solicits) ?
3631 idev->cnf.rtr_solicit_delay :
3632 idev->cnf.rtr_solicit_interval);
3633 } else {
3635 * Note: we do not support deprecated "all on-link"
3636 * assumption any longer.
3638 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3641 out:
3642 write_unlock(&idev->lock);
3643 put:
3644 in6_dev_put(idev);
3648 * Duplicate Address Detection
3650 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3652 unsigned long rand_num;
3653 struct inet6_dev *idev = ifp->idev;
3655 if (ifp->flags & IFA_F_OPTIMISTIC)
3656 rand_num = 0;
3657 else
3658 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3660 ifp->dad_probes = idev->cnf.dad_transmits;
3661 addrconf_mod_dad_work(ifp, rand_num);
3664 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3666 struct inet6_dev *idev = ifp->idev;
3667 struct net_device *dev = idev->dev;
3668 bool notify = false;
3670 addrconf_join_solict(dev, &ifp->addr);
3672 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3674 read_lock_bh(&idev->lock);
3675 spin_lock(&ifp->lock);
3676 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3677 goto out;
3679 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3680 idev->cnf.accept_dad < 1 ||
3681 !(ifp->flags&IFA_F_TENTATIVE) ||
3682 ifp->flags & IFA_F_NODAD) {
3683 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3684 spin_unlock(&ifp->lock);
3685 read_unlock_bh(&idev->lock);
3687 addrconf_dad_completed(ifp);
3688 return;
3691 if (!(idev->if_flags & IF_READY)) {
3692 spin_unlock(&ifp->lock);
3693 read_unlock_bh(&idev->lock);
3695 * If the device is not ready:
3696 * - keep it tentative if it is a permanent address.
3697 * - otherwise, kill it.
3699 in6_ifa_hold(ifp);
3700 addrconf_dad_stop(ifp, 0);
3701 return;
3705 * Optimistic nodes can start receiving
3706 * Frames right away
3708 if (ifp->flags & IFA_F_OPTIMISTIC) {
3709 ip6_ins_rt(ifp->rt);
3710 if (ipv6_use_optimistic_addr(idev)) {
3711 /* Because optimistic nodes can use this address,
3712 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3714 notify = true;
3718 addrconf_dad_kick(ifp);
3719 out:
3720 spin_unlock(&ifp->lock);
3721 read_unlock_bh(&idev->lock);
3722 if (notify)
3723 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3726 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3728 bool begin_dad = false;
3730 spin_lock_bh(&ifp->lock);
3731 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3732 ifp->state = INET6_IFADDR_STATE_PREDAD;
3733 begin_dad = true;
3735 spin_unlock_bh(&ifp->lock);
3737 if (begin_dad)
3738 addrconf_mod_dad_work(ifp, 0);
3741 static void addrconf_dad_work(struct work_struct *w)
3743 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3744 struct inet6_ifaddr,
3745 dad_work);
3746 struct inet6_dev *idev = ifp->idev;
3747 struct in6_addr mcaddr;
3749 enum {
3750 DAD_PROCESS,
3751 DAD_BEGIN,
3752 DAD_ABORT,
3753 } action = DAD_PROCESS;
3755 rtnl_lock();
3757 spin_lock_bh(&ifp->lock);
3758 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3759 action = DAD_BEGIN;
3760 ifp->state = INET6_IFADDR_STATE_DAD;
3761 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3762 action = DAD_ABORT;
3763 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3765 spin_unlock_bh(&ifp->lock);
3767 if (action == DAD_BEGIN) {
3768 addrconf_dad_begin(ifp);
3769 goto out;
3770 } else if (action == DAD_ABORT) {
3771 addrconf_dad_stop(ifp, 1);
3772 goto out;
3775 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3776 goto out;
3778 write_lock_bh(&idev->lock);
3779 if (idev->dead || !(idev->if_flags & IF_READY)) {
3780 write_unlock_bh(&idev->lock);
3781 goto out;
3784 spin_lock(&ifp->lock);
3785 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3786 spin_unlock(&ifp->lock);
3787 write_unlock_bh(&idev->lock);
3788 goto out;
3791 if (ifp->dad_probes == 0) {
3793 * DAD was successful
3796 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3797 spin_unlock(&ifp->lock);
3798 write_unlock_bh(&idev->lock);
3800 addrconf_dad_completed(ifp);
3802 goto out;
3805 ifp->dad_probes--;
3806 addrconf_mod_dad_work(ifp,
3807 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3808 spin_unlock(&ifp->lock);
3809 write_unlock_bh(&idev->lock);
3811 /* send a neighbour solicitation for our addr */
3812 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3813 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3814 out:
3815 in6_ifa_put(ifp);
3816 rtnl_unlock();
3819 /* ifp->idev must be at least read locked */
3820 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3822 struct inet6_ifaddr *ifpiter;
3823 struct inet6_dev *idev = ifp->idev;
3825 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3826 if (ifpiter->scope > IFA_LINK)
3827 break;
3828 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3829 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3830 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3831 IFA_F_PERMANENT)
3832 return false;
3834 return true;
3837 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3839 struct net_device *dev = ifp->idev->dev;
3840 struct in6_addr lladdr;
3841 bool send_rs, send_mld;
3843 addrconf_del_dad_work(ifp);
3846 * Configure the address for reception. Now it is valid.
3849 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3851 /* If added prefix is link local and we are prepared to process
3852 router advertisements, start sending router solicitations.
3855 read_lock_bh(&ifp->idev->lock);
3856 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3857 send_rs = send_mld &&
3858 ipv6_accept_ra(ifp->idev) &&
3859 ifp->idev->cnf.rtr_solicits > 0 &&
3860 (dev->flags&IFF_LOOPBACK) == 0;
3861 read_unlock_bh(&ifp->idev->lock);
3863 /* While dad is in progress mld report's source address is in6_addrany.
3864 * Resend with proper ll now.
3866 if (send_mld)
3867 ipv6_mc_dad_complete(ifp->idev);
3869 if (send_rs) {
3871 * If a host as already performed a random delay
3872 * [...] as part of DAD [...] there is no need
3873 * to delay again before sending the first RS
3875 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3876 return;
3877 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3879 write_lock_bh(&ifp->idev->lock);
3880 spin_lock(&ifp->lock);
3881 ifp->idev->rs_probes = 1;
3882 ifp->idev->if_flags |= IF_RS_SENT;
3883 addrconf_mod_rs_timer(ifp->idev,
3884 ifp->idev->cnf.rtr_solicit_interval);
3885 spin_unlock(&ifp->lock);
3886 write_unlock_bh(&ifp->idev->lock);
3890 static void addrconf_dad_run(struct inet6_dev *idev)
3892 struct inet6_ifaddr *ifp;
3894 read_lock_bh(&idev->lock);
3895 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3896 spin_lock(&ifp->lock);
3897 if (ifp->flags & IFA_F_TENTATIVE &&
3898 ifp->state == INET6_IFADDR_STATE_DAD)
3899 addrconf_dad_kick(ifp);
3900 spin_unlock(&ifp->lock);
3902 read_unlock_bh(&idev->lock);
3905 #ifdef CONFIG_PROC_FS
3906 struct if6_iter_state {
3907 struct seq_net_private p;
3908 int bucket;
3909 int offset;
3912 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3914 struct inet6_ifaddr *ifa = NULL;
3915 struct if6_iter_state *state = seq->private;
3916 struct net *net = seq_file_net(seq);
3917 int p = 0;
3919 /* initial bucket if pos is 0 */
3920 if (pos == 0) {
3921 state->bucket = 0;
3922 state->offset = 0;
3925 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3926 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3927 addr_lst) {
3928 if (!net_eq(dev_net(ifa->idev->dev), net))
3929 continue;
3930 /* sync with offset */
3931 if (p < state->offset) {
3932 p++;
3933 continue;
3935 state->offset++;
3936 return ifa;
3939 /* prepare for next bucket */
3940 state->offset = 0;
3941 p = 0;
3943 return NULL;
3946 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3947 struct inet6_ifaddr *ifa)
3949 struct if6_iter_state *state = seq->private;
3950 struct net *net = seq_file_net(seq);
3952 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3953 if (!net_eq(dev_net(ifa->idev->dev), net))
3954 continue;
3955 state->offset++;
3956 return ifa;
3959 while (++state->bucket < IN6_ADDR_HSIZE) {
3960 state->offset = 0;
3961 hlist_for_each_entry_rcu_bh(ifa,
3962 &inet6_addr_lst[state->bucket], addr_lst) {
3963 if (!net_eq(dev_net(ifa->idev->dev), net))
3964 continue;
3965 state->offset++;
3966 return ifa;
3970 return NULL;
3973 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3974 __acquires(rcu_bh)
3976 rcu_read_lock_bh();
3977 return if6_get_first(seq, *pos);
3980 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3982 struct inet6_ifaddr *ifa;
3984 ifa = if6_get_next(seq, v);
3985 ++*pos;
3986 return ifa;
3989 static void if6_seq_stop(struct seq_file *seq, void *v)
3990 __releases(rcu_bh)
3992 rcu_read_unlock_bh();
3995 static int if6_seq_show(struct seq_file *seq, void *v)
3997 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3998 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3999 &ifp->addr,
4000 ifp->idev->dev->ifindex,
4001 ifp->prefix_len,
4002 ifp->scope,
4003 (u8) ifp->flags,
4004 ifp->idev->dev->name);
4005 return 0;
4008 static const struct seq_operations if6_seq_ops = {
4009 .start = if6_seq_start,
4010 .next = if6_seq_next,
4011 .show = if6_seq_show,
4012 .stop = if6_seq_stop,
4015 static int if6_seq_open(struct inode *inode, struct file *file)
4017 return seq_open_net(inode, file, &if6_seq_ops,
4018 sizeof(struct if6_iter_state));
4021 static const struct file_operations if6_fops = {
4022 .owner = THIS_MODULE,
4023 .open = if6_seq_open,
4024 .read = seq_read,
4025 .llseek = seq_lseek,
4026 .release = seq_release_net,
4029 static int __net_init if6_proc_net_init(struct net *net)
4031 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4032 return -ENOMEM;
4033 return 0;
4036 static void __net_exit if6_proc_net_exit(struct net *net)
4038 remove_proc_entry("if_inet6", net->proc_net);
4041 static struct pernet_operations if6_proc_net_ops = {
4042 .init = if6_proc_net_init,
4043 .exit = if6_proc_net_exit,
4046 int __init if6_proc_init(void)
4048 return register_pernet_subsys(&if6_proc_net_ops);
4051 void if6_proc_exit(void)
4053 unregister_pernet_subsys(&if6_proc_net_ops);
4055 #endif /* CONFIG_PROC_FS */
4057 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4058 /* Check if address is a home address configured on any interface. */
4059 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4061 int ret = 0;
4062 struct inet6_ifaddr *ifp = NULL;
4063 unsigned int hash = inet6_addr_hash(addr);
4065 rcu_read_lock_bh();
4066 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4067 if (!net_eq(dev_net(ifp->idev->dev), net))
4068 continue;
4069 if (ipv6_addr_equal(&ifp->addr, addr) &&
4070 (ifp->flags & IFA_F_HOMEADDRESS)) {
4071 ret = 1;
4072 break;
4075 rcu_read_unlock_bh();
4076 return ret;
4078 #endif
4081 * Periodic address status verification
4084 static void addrconf_verify_rtnl(void)
4086 unsigned long now, next, next_sec, next_sched;
4087 struct inet6_ifaddr *ifp;
4088 int i;
4090 ASSERT_RTNL();
4092 rcu_read_lock_bh();
4093 now = jiffies;
4094 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4096 cancel_delayed_work(&addr_chk_work);
4098 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4099 restart:
4100 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4101 unsigned long age;
4103 /* When setting preferred_lft to a value not zero or
4104 * infinity, while valid_lft is infinity
4105 * IFA_F_PERMANENT has a non-infinity life time.
4107 if ((ifp->flags & IFA_F_PERMANENT) &&
4108 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4109 continue;
4111 spin_lock(&ifp->lock);
4112 /* We try to batch several events at once. */
4113 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4115 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4116 age >= ifp->valid_lft) {
4117 spin_unlock(&ifp->lock);
4118 in6_ifa_hold(ifp);
4119 ipv6_del_addr(ifp);
4120 goto restart;
4121 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4122 spin_unlock(&ifp->lock);
4123 continue;
4124 } else if (age >= ifp->prefered_lft) {
4125 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4126 int deprecate = 0;
4128 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4129 deprecate = 1;
4130 ifp->flags |= IFA_F_DEPRECATED;
4133 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4134 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4135 next = ifp->tstamp + ifp->valid_lft * HZ;
4137 spin_unlock(&ifp->lock);
4139 if (deprecate) {
4140 in6_ifa_hold(ifp);
4142 ipv6_ifa_notify(0, ifp);
4143 in6_ifa_put(ifp);
4144 goto restart;
4146 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4147 !(ifp->flags&IFA_F_TENTATIVE)) {
4148 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4149 ifp->idev->cnf.dad_transmits *
4150 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4152 if (age >= ifp->prefered_lft - regen_advance) {
4153 struct inet6_ifaddr *ifpub = ifp->ifpub;
4154 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4155 next = ifp->tstamp + ifp->prefered_lft * HZ;
4156 if (!ifp->regen_count && ifpub) {
4157 ifp->regen_count++;
4158 in6_ifa_hold(ifp);
4159 in6_ifa_hold(ifpub);
4160 spin_unlock(&ifp->lock);
4162 spin_lock(&ifpub->lock);
4163 ifpub->regen_count = 0;
4164 spin_unlock(&ifpub->lock);
4165 ipv6_create_tempaddr(ifpub, ifp);
4166 in6_ifa_put(ifpub);
4167 in6_ifa_put(ifp);
4168 goto restart;
4170 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4171 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4172 spin_unlock(&ifp->lock);
4173 } else {
4174 /* ifp->prefered_lft <= ifp->valid_lft */
4175 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4176 next = ifp->tstamp + ifp->prefered_lft * HZ;
4177 spin_unlock(&ifp->lock);
4182 next_sec = round_jiffies_up(next);
4183 next_sched = next;
4185 /* If rounded timeout is accurate enough, accept it. */
4186 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4187 next_sched = next_sec;
4189 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4190 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4191 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4193 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4194 now, next, next_sec, next_sched);
4195 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4196 rcu_read_unlock_bh();
4199 static void addrconf_verify_work(struct work_struct *w)
4201 rtnl_lock();
4202 addrconf_verify_rtnl();
4203 rtnl_unlock();
4206 static void addrconf_verify(void)
4208 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4211 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4212 struct in6_addr **peer_pfx)
4214 struct in6_addr *pfx = NULL;
4216 *peer_pfx = NULL;
4218 if (addr)
4219 pfx = nla_data(addr);
4221 if (local) {
4222 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4223 *peer_pfx = pfx;
4224 pfx = nla_data(local);
4227 return pfx;
4230 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4231 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4232 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4233 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4234 [IFA_FLAGS] = { .len = sizeof(u32) },
4237 static int
4238 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4240 struct net *net = sock_net(skb->sk);
4241 struct ifaddrmsg *ifm;
4242 struct nlattr *tb[IFA_MAX+1];
4243 struct in6_addr *pfx, *peer_pfx;
4244 u32 ifa_flags;
4245 int err;
4247 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4248 if (err < 0)
4249 return err;
4251 ifm = nlmsg_data(nlh);
4252 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4253 if (!pfx)
4254 return -EINVAL;
4256 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4258 /* We ignore other flags so far. */
4259 ifa_flags &= IFA_F_MANAGETEMPADDR;
4261 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4262 ifm->ifa_prefixlen);
4265 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4266 u32 prefered_lft, u32 valid_lft)
4268 u32 flags;
4269 clock_t expires;
4270 unsigned long timeout;
4271 bool was_managetempaddr;
4272 bool had_prefixroute;
4274 ASSERT_RTNL();
4276 if (!valid_lft || (prefered_lft > valid_lft))
4277 return -EINVAL;
4279 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4280 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4281 return -EINVAL;
4283 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4284 if (addrconf_finite_timeout(timeout)) {
4285 expires = jiffies_to_clock_t(timeout * HZ);
4286 valid_lft = timeout;
4287 flags = RTF_EXPIRES;
4288 } else {
4289 expires = 0;
4290 flags = 0;
4291 ifa_flags |= IFA_F_PERMANENT;
4294 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4295 if (addrconf_finite_timeout(timeout)) {
4296 if (timeout == 0)
4297 ifa_flags |= IFA_F_DEPRECATED;
4298 prefered_lft = timeout;
4301 spin_lock_bh(&ifp->lock);
4302 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4303 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4304 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4305 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4306 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4307 IFA_F_NOPREFIXROUTE);
4308 ifp->flags |= ifa_flags;
4309 ifp->tstamp = jiffies;
4310 ifp->valid_lft = valid_lft;
4311 ifp->prefered_lft = prefered_lft;
4313 spin_unlock_bh(&ifp->lock);
4314 if (!(ifp->flags&IFA_F_TENTATIVE))
4315 ipv6_ifa_notify(0, ifp);
4317 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4318 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4319 expires, flags);
4320 } else if (had_prefixroute) {
4321 enum cleanup_prefix_rt_t action;
4322 unsigned long rt_expires;
4324 write_lock_bh(&ifp->idev->lock);
4325 action = check_cleanup_prefix_route(ifp, &rt_expires);
4326 write_unlock_bh(&ifp->idev->lock);
4328 if (action != CLEANUP_PREFIX_RT_NOP) {
4329 cleanup_prefix_route(ifp, rt_expires,
4330 action == CLEANUP_PREFIX_RT_DEL);
4334 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4335 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4336 valid_lft = prefered_lft = 0;
4337 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4338 !was_managetempaddr, jiffies);
4341 addrconf_verify_rtnl();
4343 return 0;
4346 static int
4347 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4349 struct net *net = sock_net(skb->sk);
4350 struct ifaddrmsg *ifm;
4351 struct nlattr *tb[IFA_MAX+1];
4352 struct in6_addr *pfx, *peer_pfx;
4353 struct inet6_ifaddr *ifa;
4354 struct net_device *dev;
4355 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4356 u32 ifa_flags;
4357 int err;
4359 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4360 if (err < 0)
4361 return err;
4363 ifm = nlmsg_data(nlh);
4364 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4365 if (!pfx)
4366 return -EINVAL;
4368 if (tb[IFA_CACHEINFO]) {
4369 struct ifa_cacheinfo *ci;
4371 ci = nla_data(tb[IFA_CACHEINFO]);
4372 valid_lft = ci->ifa_valid;
4373 preferred_lft = ci->ifa_prefered;
4374 } else {
4375 preferred_lft = INFINITY_LIFE_TIME;
4376 valid_lft = INFINITY_LIFE_TIME;
4379 dev = __dev_get_by_index(net, ifm->ifa_index);
4380 if (!dev)
4381 return -ENODEV;
4383 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4385 /* We ignore other flags so far. */
4386 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4387 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4389 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4390 if (!ifa) {
4392 * It would be best to check for !NLM_F_CREATE here but
4393 * userspace already relies on not having to provide this.
4395 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4396 ifm->ifa_prefixlen, ifa_flags,
4397 preferred_lft, valid_lft);
4400 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4401 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4402 err = -EEXIST;
4403 else
4404 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4406 in6_ifa_put(ifa);
4408 return err;
4411 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4412 u8 scope, int ifindex)
4414 struct ifaddrmsg *ifm;
4416 ifm = nlmsg_data(nlh);
4417 ifm->ifa_family = AF_INET6;
4418 ifm->ifa_prefixlen = prefixlen;
4419 ifm->ifa_flags = flags;
4420 ifm->ifa_scope = scope;
4421 ifm->ifa_index = ifindex;
4424 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4425 unsigned long tstamp, u32 preferred, u32 valid)
4427 struct ifa_cacheinfo ci;
4429 ci.cstamp = cstamp_delta(cstamp);
4430 ci.tstamp = cstamp_delta(tstamp);
4431 ci.ifa_prefered = preferred;
4432 ci.ifa_valid = valid;
4434 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4437 static inline int rt_scope(int ifa_scope)
4439 if (ifa_scope & IFA_HOST)
4440 return RT_SCOPE_HOST;
4441 else if (ifa_scope & IFA_LINK)
4442 return RT_SCOPE_LINK;
4443 else if (ifa_scope & IFA_SITE)
4444 return RT_SCOPE_SITE;
4445 else
4446 return RT_SCOPE_UNIVERSE;
4449 static inline int inet6_ifaddr_msgsize(void)
4451 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4452 + nla_total_size(16) /* IFA_LOCAL */
4453 + nla_total_size(16) /* IFA_ADDRESS */
4454 + nla_total_size(sizeof(struct ifa_cacheinfo))
4455 + nla_total_size(4) /* IFA_FLAGS */;
4458 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4459 u32 portid, u32 seq, int event, unsigned int flags)
4461 struct nlmsghdr *nlh;
4462 u32 preferred, valid;
4464 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4465 if (!nlh)
4466 return -EMSGSIZE;
4468 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4469 ifa->idev->dev->ifindex);
4471 if (!((ifa->flags&IFA_F_PERMANENT) &&
4472 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4473 preferred = ifa->prefered_lft;
4474 valid = ifa->valid_lft;
4475 if (preferred != INFINITY_LIFE_TIME) {
4476 long tval = (jiffies - ifa->tstamp)/HZ;
4477 if (preferred > tval)
4478 preferred -= tval;
4479 else
4480 preferred = 0;
4481 if (valid != INFINITY_LIFE_TIME) {
4482 if (valid > tval)
4483 valid -= tval;
4484 else
4485 valid = 0;
4488 } else {
4489 preferred = INFINITY_LIFE_TIME;
4490 valid = INFINITY_LIFE_TIME;
4493 if (!ipv6_addr_any(&ifa->peer_addr)) {
4494 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4495 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4496 goto error;
4497 } else
4498 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4499 goto error;
4501 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4502 goto error;
4504 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4505 goto error;
4507 nlmsg_end(skb, nlh);
4508 return 0;
4510 error:
4511 nlmsg_cancel(skb, nlh);
4512 return -EMSGSIZE;
4515 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4516 u32 portid, u32 seq, int event, u16 flags)
4518 struct nlmsghdr *nlh;
4519 u8 scope = RT_SCOPE_UNIVERSE;
4520 int ifindex = ifmca->idev->dev->ifindex;
4522 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4523 scope = RT_SCOPE_SITE;
4525 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4526 if (!nlh)
4527 return -EMSGSIZE;
4529 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4530 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4531 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4532 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4533 nlmsg_cancel(skb, nlh);
4534 return -EMSGSIZE;
4537 nlmsg_end(skb, nlh);
4538 return 0;
4541 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4542 u32 portid, u32 seq, int event, unsigned int flags)
4544 struct nlmsghdr *nlh;
4545 u8 scope = RT_SCOPE_UNIVERSE;
4546 int ifindex = ifaca->aca_idev->dev->ifindex;
4548 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4549 scope = RT_SCOPE_SITE;
4551 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4552 if (!nlh)
4553 return -EMSGSIZE;
4555 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4556 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4557 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4558 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4559 nlmsg_cancel(skb, nlh);
4560 return -EMSGSIZE;
4563 nlmsg_end(skb, nlh);
4564 return 0;
4567 enum addr_type_t {
4568 UNICAST_ADDR,
4569 MULTICAST_ADDR,
4570 ANYCAST_ADDR,
4573 /* called with rcu_read_lock() */
4574 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4575 struct netlink_callback *cb, enum addr_type_t type,
4576 int s_ip_idx, int *p_ip_idx)
4578 struct ifmcaddr6 *ifmca;
4579 struct ifacaddr6 *ifaca;
4580 int err = 1;
4581 int ip_idx = *p_ip_idx;
4583 read_lock_bh(&idev->lock);
4584 switch (type) {
4585 case UNICAST_ADDR: {
4586 struct inet6_ifaddr *ifa;
4588 /* unicast address incl. temp addr */
4589 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4590 if (++ip_idx < s_ip_idx)
4591 continue;
4592 err = inet6_fill_ifaddr(skb, ifa,
4593 NETLINK_CB(cb->skb).portid,
4594 cb->nlh->nlmsg_seq,
4595 RTM_NEWADDR,
4596 NLM_F_MULTI);
4597 if (err < 0)
4598 break;
4599 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4601 break;
4603 case MULTICAST_ADDR:
4604 /* multicast address */
4605 for (ifmca = idev->mc_list; ifmca;
4606 ifmca = ifmca->next, ip_idx++) {
4607 if (ip_idx < s_ip_idx)
4608 continue;
4609 err = inet6_fill_ifmcaddr(skb, ifmca,
4610 NETLINK_CB(cb->skb).portid,
4611 cb->nlh->nlmsg_seq,
4612 RTM_GETMULTICAST,
4613 NLM_F_MULTI);
4614 if (err < 0)
4615 break;
4617 break;
4618 case ANYCAST_ADDR:
4619 /* anycast address */
4620 for (ifaca = idev->ac_list; ifaca;
4621 ifaca = ifaca->aca_next, ip_idx++) {
4622 if (ip_idx < s_ip_idx)
4623 continue;
4624 err = inet6_fill_ifacaddr(skb, ifaca,
4625 NETLINK_CB(cb->skb).portid,
4626 cb->nlh->nlmsg_seq,
4627 RTM_GETANYCAST,
4628 NLM_F_MULTI);
4629 if (err < 0)
4630 break;
4632 break;
4633 default:
4634 break;
4636 read_unlock_bh(&idev->lock);
4637 *p_ip_idx = ip_idx;
4638 return err;
4641 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4642 enum addr_type_t type)
4644 struct net *net = sock_net(skb->sk);
4645 int h, s_h;
4646 int idx, ip_idx;
4647 int s_idx, s_ip_idx;
4648 struct net_device *dev;
4649 struct inet6_dev *idev;
4650 struct hlist_head *head;
4652 s_h = cb->args[0];
4653 s_idx = idx = cb->args[1];
4654 s_ip_idx = ip_idx = cb->args[2];
4656 rcu_read_lock();
4657 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4658 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4659 idx = 0;
4660 head = &net->dev_index_head[h];
4661 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4662 if (idx < s_idx)
4663 goto cont;
4664 if (h > s_h || idx > s_idx)
4665 s_ip_idx = 0;
4666 ip_idx = 0;
4667 idev = __in6_dev_get(dev);
4668 if (!idev)
4669 goto cont;
4671 if (in6_dump_addrs(idev, skb, cb, type,
4672 s_ip_idx, &ip_idx) < 0)
4673 goto done;
4674 cont:
4675 idx++;
4678 done:
4679 rcu_read_unlock();
4680 cb->args[0] = h;
4681 cb->args[1] = idx;
4682 cb->args[2] = ip_idx;
4684 return skb->len;
4687 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4689 enum addr_type_t type = UNICAST_ADDR;
4691 return inet6_dump_addr(skb, cb, type);
4694 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4696 enum addr_type_t type = MULTICAST_ADDR;
4698 return inet6_dump_addr(skb, cb, type);
4702 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4704 enum addr_type_t type = ANYCAST_ADDR;
4706 return inet6_dump_addr(skb, cb, type);
4709 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4711 struct net *net = sock_net(in_skb->sk);
4712 struct ifaddrmsg *ifm;
4713 struct nlattr *tb[IFA_MAX+1];
4714 struct in6_addr *addr = NULL, *peer;
4715 struct net_device *dev = NULL;
4716 struct inet6_ifaddr *ifa;
4717 struct sk_buff *skb;
4718 int err;
4720 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4721 if (err < 0)
4722 goto errout;
4724 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4725 if (!addr) {
4726 err = -EINVAL;
4727 goto errout;
4730 ifm = nlmsg_data(nlh);
4731 if (ifm->ifa_index)
4732 dev = __dev_get_by_index(net, ifm->ifa_index);
4734 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4735 if (!ifa) {
4736 err = -EADDRNOTAVAIL;
4737 goto errout;
4740 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4741 if (!skb) {
4742 err = -ENOBUFS;
4743 goto errout_ifa;
4746 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4747 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4748 if (err < 0) {
4749 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4750 WARN_ON(err == -EMSGSIZE);
4751 kfree_skb(skb);
4752 goto errout_ifa;
4754 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4755 errout_ifa:
4756 in6_ifa_put(ifa);
4757 errout:
4758 return err;
4761 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4763 struct sk_buff *skb;
4764 struct net *net = dev_net(ifa->idev->dev);
4765 int err = -ENOBUFS;
4767 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4768 if (!skb)
4769 goto errout;
4771 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4772 if (err < 0) {
4773 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4774 WARN_ON(err == -EMSGSIZE);
4775 kfree_skb(skb);
4776 goto errout;
4778 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4779 return;
4780 errout:
4781 if (err < 0)
4782 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4785 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4786 __s32 *array, int bytes)
4788 BUG_ON(bytes < (DEVCONF_MAX * 4));
4790 memset(array, 0, bytes);
4791 array[DEVCONF_FORWARDING] = cnf->forwarding;
4792 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4793 array[DEVCONF_MTU6] = cnf->mtu6;
4794 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4795 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4796 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4797 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4798 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4799 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4800 jiffies_to_msecs(cnf->rtr_solicit_interval);
4801 array[DEVCONF_RTR_SOLICIT_DELAY] =
4802 jiffies_to_msecs(cnf->rtr_solicit_delay);
4803 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4804 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4805 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4806 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4807 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4808 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4809 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4810 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4811 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4812 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4813 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4814 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4815 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4816 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4817 #ifdef CONFIG_IPV6_ROUTER_PREF
4818 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4819 array[DEVCONF_RTR_PROBE_INTERVAL] =
4820 jiffies_to_msecs(cnf->rtr_probe_interval);
4821 #ifdef CONFIG_IPV6_ROUTE_INFO
4822 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4823 #endif
4824 #endif
4825 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4826 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4827 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4828 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4829 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4830 #endif
4831 #ifdef CONFIG_IPV6_MROUTE
4832 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4833 #endif
4834 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4835 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4836 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4837 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4838 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4839 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4840 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4841 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4842 /* we omit DEVCONF_STABLE_SECRET for now */
4843 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4844 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4845 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4846 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4849 static inline size_t inet6_ifla6_size(void)
4851 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4852 + nla_total_size(sizeof(struct ifla_cacheinfo))
4853 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4854 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4855 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4856 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4859 static inline size_t inet6_if_nlmsg_size(void)
4861 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4862 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4863 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4864 + nla_total_size(4) /* IFLA_MTU */
4865 + nla_total_size(4) /* IFLA_LINK */
4866 + nla_total_size(1) /* IFLA_OPERSTATE */
4867 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4870 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4871 int items, int bytes)
4873 int i;
4874 int pad = bytes - sizeof(u64) * items;
4875 BUG_ON(pad < 0);
4877 /* Use put_unaligned() because stats may not be aligned for u64. */
4878 put_unaligned(items, &stats[0]);
4879 for (i = 1; i < items; i++)
4880 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4882 memset(&stats[items], 0, pad);
4885 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4886 int bytes, size_t syncpoff)
4888 int i, c;
4889 u64 buff[IPSTATS_MIB_MAX];
4890 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4892 BUG_ON(pad < 0);
4894 memset(buff, 0, sizeof(buff));
4895 buff[0] = IPSTATS_MIB_MAX;
4897 for_each_possible_cpu(c) {
4898 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4899 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4902 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4903 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4906 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4907 int bytes)
4909 switch (attrtype) {
4910 case IFLA_INET6_STATS:
4911 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4912 offsetof(struct ipstats_mib, syncp));
4913 break;
4914 case IFLA_INET6_ICMP6STATS:
4915 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4916 break;
4920 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4921 u32 ext_filter_mask)
4923 struct nlattr *nla;
4924 struct ifla_cacheinfo ci;
4926 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4927 goto nla_put_failure;
4928 ci.max_reasm_len = IPV6_MAXPLEN;
4929 ci.tstamp = cstamp_delta(idev->tstamp);
4930 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4931 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4932 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4933 goto nla_put_failure;
4934 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4935 if (!nla)
4936 goto nla_put_failure;
4937 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4939 /* XXX - MC not implemented */
4941 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4942 return 0;
4944 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4945 if (!nla)
4946 goto nla_put_failure;
4947 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4949 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4950 if (!nla)
4951 goto nla_put_failure;
4952 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4954 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4955 if (!nla)
4956 goto nla_put_failure;
4958 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4959 goto nla_put_failure;
4961 read_lock_bh(&idev->lock);
4962 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4963 read_unlock_bh(&idev->lock);
4965 return 0;
4967 nla_put_failure:
4968 return -EMSGSIZE;
4971 static size_t inet6_get_link_af_size(const struct net_device *dev,
4972 u32 ext_filter_mask)
4974 if (!__in6_dev_get(dev))
4975 return 0;
4977 return inet6_ifla6_size();
4980 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4981 u32 ext_filter_mask)
4983 struct inet6_dev *idev = __in6_dev_get(dev);
4985 if (!idev)
4986 return -ENODATA;
4988 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4989 return -EMSGSIZE;
4991 return 0;
4994 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4996 struct inet6_ifaddr *ifp;
4997 struct net_device *dev = idev->dev;
4998 bool update_rs = false;
4999 struct in6_addr ll_addr;
5001 ASSERT_RTNL();
5003 if (!token)
5004 return -EINVAL;
5005 if (ipv6_addr_any(token))
5006 return -EINVAL;
5007 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5008 return -EINVAL;
5009 if (!ipv6_accept_ra(idev))
5010 return -EINVAL;
5011 if (idev->cnf.rtr_solicits <= 0)
5012 return -EINVAL;
5014 write_lock_bh(&idev->lock);
5016 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5017 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5019 write_unlock_bh(&idev->lock);
5021 if (!idev->dead && (idev->if_flags & IF_READY) &&
5022 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5023 IFA_F_OPTIMISTIC)) {
5025 /* If we're not ready, then normal ifup will take care
5026 * of this. Otherwise, we need to request our rs here.
5028 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5029 update_rs = true;
5032 write_lock_bh(&idev->lock);
5034 if (update_rs) {
5035 idev->if_flags |= IF_RS_SENT;
5036 idev->rs_probes = 1;
5037 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
5040 /* Well, that's kinda nasty ... */
5041 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5042 spin_lock(&ifp->lock);
5043 if (ifp->tokenized) {
5044 ifp->valid_lft = 0;
5045 ifp->prefered_lft = 0;
5047 spin_unlock(&ifp->lock);
5050 write_unlock_bh(&idev->lock);
5051 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5052 addrconf_verify_rtnl();
5053 return 0;
5056 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5057 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5058 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5061 static int inet6_validate_link_af(const struct net_device *dev,
5062 const struct nlattr *nla)
5064 struct nlattr *tb[IFLA_INET6_MAX + 1];
5066 if (dev && !__in6_dev_get(dev))
5067 return -EAFNOSUPPORT;
5069 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5072 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5074 int err = -EINVAL;
5075 struct inet6_dev *idev = __in6_dev_get(dev);
5076 struct nlattr *tb[IFLA_INET6_MAX + 1];
5078 if (!idev)
5079 return -EAFNOSUPPORT;
5081 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5082 BUG();
5084 if (tb[IFLA_INET6_TOKEN]) {
5085 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5086 if (err)
5087 return err;
5090 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5091 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5093 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5094 mode != IN6_ADDR_GEN_MODE_NONE &&
5095 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5096 mode != IN6_ADDR_GEN_MODE_RANDOM)
5097 return -EINVAL;
5099 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5100 !idev->cnf.stable_secret.initialized &&
5101 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5102 return -EINVAL;
5104 idev->addr_gen_mode = mode;
5105 err = 0;
5108 return err;
5111 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5112 u32 portid, u32 seq, int event, unsigned int flags)
5114 struct net_device *dev = idev->dev;
5115 struct ifinfomsg *hdr;
5116 struct nlmsghdr *nlh;
5117 void *protoinfo;
5119 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5120 if (!nlh)
5121 return -EMSGSIZE;
5123 hdr = nlmsg_data(nlh);
5124 hdr->ifi_family = AF_INET6;
5125 hdr->__ifi_pad = 0;
5126 hdr->ifi_type = dev->type;
5127 hdr->ifi_index = dev->ifindex;
5128 hdr->ifi_flags = dev_get_flags(dev);
5129 hdr->ifi_change = 0;
5131 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5132 (dev->addr_len &&
5133 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5134 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5135 (dev->ifindex != dev_get_iflink(dev) &&
5136 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5137 nla_put_u8(skb, IFLA_OPERSTATE,
5138 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5139 goto nla_put_failure;
5140 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5141 if (!protoinfo)
5142 goto nla_put_failure;
5144 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5145 goto nla_put_failure;
5147 nla_nest_end(skb, protoinfo);
5148 nlmsg_end(skb, nlh);
5149 return 0;
5151 nla_put_failure:
5152 nlmsg_cancel(skb, nlh);
5153 return -EMSGSIZE;
5156 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5158 struct net *net = sock_net(skb->sk);
5159 int h, s_h;
5160 int idx = 0, s_idx;
5161 struct net_device *dev;
5162 struct inet6_dev *idev;
5163 struct hlist_head *head;
5165 s_h = cb->args[0];
5166 s_idx = cb->args[1];
5168 rcu_read_lock();
5169 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5170 idx = 0;
5171 head = &net->dev_index_head[h];
5172 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5173 if (idx < s_idx)
5174 goto cont;
5175 idev = __in6_dev_get(dev);
5176 if (!idev)
5177 goto cont;
5178 if (inet6_fill_ifinfo(skb, idev,
5179 NETLINK_CB(cb->skb).portid,
5180 cb->nlh->nlmsg_seq,
5181 RTM_NEWLINK, NLM_F_MULTI) < 0)
5182 goto out;
5183 cont:
5184 idx++;
5187 out:
5188 rcu_read_unlock();
5189 cb->args[1] = idx;
5190 cb->args[0] = h;
5192 return skb->len;
5195 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5197 struct sk_buff *skb;
5198 struct net *net = dev_net(idev->dev);
5199 int err = -ENOBUFS;
5201 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5202 if (!skb)
5203 goto errout;
5205 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5206 if (err < 0) {
5207 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5208 WARN_ON(err == -EMSGSIZE);
5209 kfree_skb(skb);
5210 goto errout;
5212 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5213 return;
5214 errout:
5215 if (err < 0)
5216 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5219 static inline size_t inet6_prefix_nlmsg_size(void)
5221 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5222 + nla_total_size(sizeof(struct in6_addr))
5223 + nla_total_size(sizeof(struct prefix_cacheinfo));
5226 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5227 struct prefix_info *pinfo, u32 portid, u32 seq,
5228 int event, unsigned int flags)
5230 struct prefixmsg *pmsg;
5231 struct nlmsghdr *nlh;
5232 struct prefix_cacheinfo ci;
5234 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5235 if (!nlh)
5236 return -EMSGSIZE;
5238 pmsg = nlmsg_data(nlh);
5239 pmsg->prefix_family = AF_INET6;
5240 pmsg->prefix_pad1 = 0;
5241 pmsg->prefix_pad2 = 0;
5242 pmsg->prefix_ifindex = idev->dev->ifindex;
5243 pmsg->prefix_len = pinfo->prefix_len;
5244 pmsg->prefix_type = pinfo->type;
5245 pmsg->prefix_pad3 = 0;
5246 pmsg->prefix_flags = 0;
5247 if (pinfo->onlink)
5248 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5249 if (pinfo->autoconf)
5250 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5252 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5253 goto nla_put_failure;
5254 ci.preferred_time = ntohl(pinfo->prefered);
5255 ci.valid_time = ntohl(pinfo->valid);
5256 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5257 goto nla_put_failure;
5258 nlmsg_end(skb, nlh);
5259 return 0;
5261 nla_put_failure:
5262 nlmsg_cancel(skb, nlh);
5263 return -EMSGSIZE;
5266 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5267 struct prefix_info *pinfo)
5269 struct sk_buff *skb;
5270 struct net *net = dev_net(idev->dev);
5271 int err = -ENOBUFS;
5273 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5274 if (!skb)
5275 goto errout;
5277 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5278 if (err < 0) {
5279 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5280 WARN_ON(err == -EMSGSIZE);
5281 kfree_skb(skb);
5282 goto errout;
5284 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5285 return;
5286 errout:
5287 if (err < 0)
5288 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5291 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5293 struct net *net = dev_net(ifp->idev->dev);
5295 if (event)
5296 ASSERT_RTNL();
5298 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5300 switch (event) {
5301 case RTM_NEWADDR:
5303 * If the address was optimistic
5304 * we inserted the route at the start of
5305 * our DAD process, so we don't need
5306 * to do it again
5308 if (!(ifp->rt->rt6i_node))
5309 ip6_ins_rt(ifp->rt);
5310 if (ifp->idev->cnf.forwarding)
5311 addrconf_join_anycast(ifp);
5312 if (!ipv6_addr_any(&ifp->peer_addr))
5313 addrconf_prefix_route(&ifp->peer_addr, 128,
5314 ifp->idev->dev, 0, 0);
5315 break;
5316 case RTM_DELADDR:
5317 if (ifp->idev->cnf.forwarding)
5318 addrconf_leave_anycast(ifp);
5319 addrconf_leave_solict(ifp->idev, &ifp->addr);
5320 if (!ipv6_addr_any(&ifp->peer_addr)) {
5321 struct rt6_info *rt;
5323 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5324 ifp->idev->dev, 0, 0);
5325 if (rt)
5326 ip6_del_rt(rt);
5328 if (ifp->rt) {
5329 dst_hold(&ifp->rt->dst);
5330 ip6_del_rt(ifp->rt);
5332 rt_genid_bump_ipv6(net);
5333 break;
5335 atomic_inc(&net->ipv6.dev_addr_genid);
5338 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5340 rcu_read_lock_bh();
5341 if (likely(ifp->idev->dead == 0))
5342 __ipv6_ifa_notify(event, ifp);
5343 rcu_read_unlock_bh();
5346 #ifdef CONFIG_SYSCTL
5348 static
5349 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5350 void __user *buffer, size_t *lenp, loff_t *ppos)
5352 int *valp = ctl->data;
5353 int val = *valp;
5354 loff_t pos = *ppos;
5355 struct ctl_table lctl;
5356 int ret;
5359 * ctl->data points to idev->cnf.forwarding, we should
5360 * not modify it until we get the rtnl lock.
5362 lctl = *ctl;
5363 lctl.data = &val;
5365 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5367 if (write)
5368 ret = addrconf_fixup_forwarding(ctl, valp, val);
5369 if (ret)
5370 *ppos = pos;
5371 return ret;
5374 static
5375 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5376 void __user *buffer, size_t *lenp, loff_t *ppos)
5378 struct ctl_table lctl;
5379 int min_hl = 1, max_hl = 255;
5381 lctl = *ctl;
5382 lctl.extra1 = &min_hl;
5383 lctl.extra2 = &max_hl;
5385 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5388 static
5389 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5390 void __user *buffer, size_t *lenp, loff_t *ppos)
5392 struct inet6_dev *idev = ctl->extra1;
5393 int min_mtu = IPV6_MIN_MTU;
5394 struct ctl_table lctl;
5396 lctl = *ctl;
5397 lctl.extra1 = &min_mtu;
5398 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5400 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5403 static void dev_disable_change(struct inet6_dev *idev)
5405 struct netdev_notifier_info info;
5407 if (!idev || !idev->dev)
5408 return;
5410 netdev_notifier_info_init(&info, idev->dev);
5411 if (idev->cnf.disable_ipv6)
5412 addrconf_notify(NULL, NETDEV_DOWN, &info);
5413 else
5414 addrconf_notify(NULL, NETDEV_UP, &info);
5417 static void addrconf_disable_change(struct net *net, __s32 newf)
5419 struct net_device *dev;
5420 struct inet6_dev *idev;
5422 rcu_read_lock();
5423 for_each_netdev_rcu(net, dev) {
5424 idev = __in6_dev_get(dev);
5425 if (idev) {
5426 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5427 idev->cnf.disable_ipv6 = newf;
5428 if (changed)
5429 dev_disable_change(idev);
5432 rcu_read_unlock();
5435 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5437 struct net *net;
5438 int old;
5440 if (!rtnl_trylock())
5441 return restart_syscall();
5443 net = (struct net *)table->extra2;
5444 old = *p;
5445 *p = newf;
5447 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5448 rtnl_unlock();
5449 return 0;
5452 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5453 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5454 addrconf_disable_change(net, newf);
5455 } else if ((!newf) ^ (!old))
5456 dev_disable_change((struct inet6_dev *)table->extra1);
5458 rtnl_unlock();
5459 return 0;
5462 static
5463 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5464 void __user *buffer, size_t *lenp, loff_t *ppos)
5466 int *valp = ctl->data;
5467 int val = *valp;
5468 loff_t pos = *ppos;
5469 struct ctl_table lctl;
5470 int ret;
5473 * ctl->data points to idev->cnf.disable_ipv6, we should
5474 * not modify it until we get the rtnl lock.
5476 lctl = *ctl;
5477 lctl.data = &val;
5479 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5481 if (write)
5482 ret = addrconf_disable_ipv6(ctl, valp, val);
5483 if (ret)
5484 *ppos = pos;
5485 return ret;
5488 static
5489 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5490 void __user *buffer, size_t *lenp, loff_t *ppos)
5492 int *valp = ctl->data;
5493 int ret;
5494 int old, new;
5496 old = *valp;
5497 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5498 new = *valp;
5500 if (write && old != new) {
5501 struct net *net = ctl->extra2;
5503 if (!rtnl_trylock())
5504 return restart_syscall();
5506 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5507 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5508 NETCONFA_IFINDEX_DEFAULT,
5509 net->ipv6.devconf_dflt);
5510 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5511 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5512 NETCONFA_IFINDEX_ALL,
5513 net->ipv6.devconf_all);
5514 else {
5515 struct inet6_dev *idev = ctl->extra1;
5517 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5518 idev->dev->ifindex,
5519 &idev->cnf);
5521 rtnl_unlock();
5524 return ret;
5527 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5528 void __user *buffer, size_t *lenp,
5529 loff_t *ppos)
5531 int err;
5532 struct in6_addr addr;
5533 char str[IPV6_MAX_STRLEN];
5534 struct ctl_table lctl = *ctl;
5535 struct net *net = ctl->extra2;
5536 struct ipv6_stable_secret *secret = ctl->data;
5538 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5539 return -EIO;
5541 lctl.maxlen = IPV6_MAX_STRLEN;
5542 lctl.data = str;
5544 if (!rtnl_trylock())
5545 return restart_syscall();
5547 if (!write && !secret->initialized) {
5548 err = -EIO;
5549 goto out;
5552 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5553 if (err >= sizeof(str)) {
5554 err = -EIO;
5555 goto out;
5558 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5559 if (err || !write)
5560 goto out;
5562 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5563 err = -EIO;
5564 goto out;
5567 secret->initialized = true;
5568 secret->secret = addr;
5570 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5571 struct net_device *dev;
5573 for_each_netdev(net, dev) {
5574 struct inet6_dev *idev = __in6_dev_get(dev);
5576 if (idev) {
5577 idev->addr_gen_mode =
5578 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5581 } else {
5582 struct inet6_dev *idev = ctl->extra1;
5584 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5587 out:
5588 rtnl_unlock();
5590 return err;
5593 static
5594 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5595 int write,
5596 void __user *buffer,
5597 size_t *lenp,
5598 loff_t *ppos)
5600 int *valp = ctl->data;
5601 int val = *valp;
5602 loff_t pos = *ppos;
5603 struct ctl_table lctl;
5604 int ret;
5606 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5607 * we should not modify it until we get the rtnl lock.
5609 lctl = *ctl;
5610 lctl.data = &val;
5612 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5614 if (write)
5615 ret = addrconf_fixup_linkdown(ctl, valp, val);
5616 if (ret)
5617 *ppos = pos;
5618 return ret;
5621 static struct addrconf_sysctl_table
5623 struct ctl_table_header *sysctl_header;
5624 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5625 } addrconf_sysctl __read_mostly = {
5626 .sysctl_header = NULL,
5627 .addrconf_vars = {
5629 .procname = "forwarding",
5630 .data = &ipv6_devconf.forwarding,
5631 .maxlen = sizeof(int),
5632 .mode = 0644,
5633 .proc_handler = addrconf_sysctl_forward,
5636 .procname = "hop_limit",
5637 .data = &ipv6_devconf.hop_limit,
5638 .maxlen = sizeof(int),
5639 .mode = 0644,
5640 .proc_handler = addrconf_sysctl_hop_limit,
5643 .procname = "mtu",
5644 .data = &ipv6_devconf.mtu6,
5645 .maxlen = sizeof(int),
5646 .mode = 0644,
5647 .proc_handler = addrconf_sysctl_mtu,
5650 .procname = "accept_ra",
5651 .data = &ipv6_devconf.accept_ra,
5652 .maxlen = sizeof(int),
5653 .mode = 0644,
5654 .proc_handler = proc_dointvec,
5657 .procname = "accept_redirects",
5658 .data = &ipv6_devconf.accept_redirects,
5659 .maxlen = sizeof(int),
5660 .mode = 0644,
5661 .proc_handler = proc_dointvec,
5664 .procname = "autoconf",
5665 .data = &ipv6_devconf.autoconf,
5666 .maxlen = sizeof(int),
5667 .mode = 0644,
5668 .proc_handler = proc_dointvec,
5671 .procname = "dad_transmits",
5672 .data = &ipv6_devconf.dad_transmits,
5673 .maxlen = sizeof(int),
5674 .mode = 0644,
5675 .proc_handler = proc_dointvec,
5678 .procname = "router_solicitations",
5679 .data = &ipv6_devconf.rtr_solicits,
5680 .maxlen = sizeof(int),
5681 .mode = 0644,
5682 .proc_handler = proc_dointvec,
5685 .procname = "router_solicitation_interval",
5686 .data = &ipv6_devconf.rtr_solicit_interval,
5687 .maxlen = sizeof(int),
5688 .mode = 0644,
5689 .proc_handler = proc_dointvec_jiffies,
5692 .procname = "router_solicitation_delay",
5693 .data = &ipv6_devconf.rtr_solicit_delay,
5694 .maxlen = sizeof(int),
5695 .mode = 0644,
5696 .proc_handler = proc_dointvec_jiffies,
5699 .procname = "force_mld_version",
5700 .data = &ipv6_devconf.force_mld_version,
5701 .maxlen = sizeof(int),
5702 .mode = 0644,
5703 .proc_handler = proc_dointvec,
5706 .procname = "mldv1_unsolicited_report_interval",
5707 .data =
5708 &ipv6_devconf.mldv1_unsolicited_report_interval,
5709 .maxlen = sizeof(int),
5710 .mode = 0644,
5711 .proc_handler = proc_dointvec_ms_jiffies,
5714 .procname = "mldv2_unsolicited_report_interval",
5715 .data =
5716 &ipv6_devconf.mldv2_unsolicited_report_interval,
5717 .maxlen = sizeof(int),
5718 .mode = 0644,
5719 .proc_handler = proc_dointvec_ms_jiffies,
5722 .procname = "use_tempaddr",
5723 .data = &ipv6_devconf.use_tempaddr,
5724 .maxlen = sizeof(int),
5725 .mode = 0644,
5726 .proc_handler = proc_dointvec,
5729 .procname = "temp_valid_lft",
5730 .data = &ipv6_devconf.temp_valid_lft,
5731 .maxlen = sizeof(int),
5732 .mode = 0644,
5733 .proc_handler = proc_dointvec,
5736 .procname = "temp_prefered_lft",
5737 .data = &ipv6_devconf.temp_prefered_lft,
5738 .maxlen = sizeof(int),
5739 .mode = 0644,
5740 .proc_handler = proc_dointvec,
5743 .procname = "regen_max_retry",
5744 .data = &ipv6_devconf.regen_max_retry,
5745 .maxlen = sizeof(int),
5746 .mode = 0644,
5747 .proc_handler = proc_dointvec,
5750 .procname = "max_desync_factor",
5751 .data = &ipv6_devconf.max_desync_factor,
5752 .maxlen = sizeof(int),
5753 .mode = 0644,
5754 .proc_handler = proc_dointvec,
5757 .procname = "max_addresses",
5758 .data = &ipv6_devconf.max_addresses,
5759 .maxlen = sizeof(int),
5760 .mode = 0644,
5761 .proc_handler = proc_dointvec,
5764 .procname = "accept_ra_defrtr",
5765 .data = &ipv6_devconf.accept_ra_defrtr,
5766 .maxlen = sizeof(int),
5767 .mode = 0644,
5768 .proc_handler = proc_dointvec,
5771 .procname = "accept_ra_min_hop_limit",
5772 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5773 .maxlen = sizeof(int),
5774 .mode = 0644,
5775 .proc_handler = proc_dointvec,
5778 .procname = "accept_ra_pinfo",
5779 .data = &ipv6_devconf.accept_ra_pinfo,
5780 .maxlen = sizeof(int),
5781 .mode = 0644,
5782 .proc_handler = proc_dointvec,
5784 #ifdef CONFIG_IPV6_ROUTER_PREF
5786 .procname = "accept_ra_rtr_pref",
5787 .data = &ipv6_devconf.accept_ra_rtr_pref,
5788 .maxlen = sizeof(int),
5789 .mode = 0644,
5790 .proc_handler = proc_dointvec,
5793 .procname = "router_probe_interval",
5794 .data = &ipv6_devconf.rtr_probe_interval,
5795 .maxlen = sizeof(int),
5796 .mode = 0644,
5797 .proc_handler = proc_dointvec_jiffies,
5799 #ifdef CONFIG_IPV6_ROUTE_INFO
5801 .procname = "accept_ra_rt_info_max_plen",
5802 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5803 .maxlen = sizeof(int),
5804 .mode = 0644,
5805 .proc_handler = proc_dointvec,
5807 #endif
5808 #endif
5810 .procname = "proxy_ndp",
5811 .data = &ipv6_devconf.proxy_ndp,
5812 .maxlen = sizeof(int),
5813 .mode = 0644,
5814 .proc_handler = addrconf_sysctl_proxy_ndp,
5817 .procname = "accept_source_route",
5818 .data = &ipv6_devconf.accept_source_route,
5819 .maxlen = sizeof(int),
5820 .mode = 0644,
5821 .proc_handler = proc_dointvec,
5823 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5825 .procname = "optimistic_dad",
5826 .data = &ipv6_devconf.optimistic_dad,
5827 .maxlen = sizeof(int),
5828 .mode = 0644,
5829 .proc_handler = proc_dointvec,
5833 .procname = "use_optimistic",
5834 .data = &ipv6_devconf.use_optimistic,
5835 .maxlen = sizeof(int),
5836 .mode = 0644,
5837 .proc_handler = proc_dointvec,
5840 #endif
5841 #ifdef CONFIG_IPV6_MROUTE
5843 .procname = "mc_forwarding",
5844 .data = &ipv6_devconf.mc_forwarding,
5845 .maxlen = sizeof(int),
5846 .mode = 0444,
5847 .proc_handler = proc_dointvec,
5849 #endif
5851 .procname = "disable_ipv6",
5852 .data = &ipv6_devconf.disable_ipv6,
5853 .maxlen = sizeof(int),
5854 .mode = 0644,
5855 .proc_handler = addrconf_sysctl_disable,
5858 .procname = "accept_dad",
5859 .data = &ipv6_devconf.accept_dad,
5860 .maxlen = sizeof(int),
5861 .mode = 0644,
5862 .proc_handler = proc_dointvec,
5865 .procname = "force_tllao",
5866 .data = &ipv6_devconf.force_tllao,
5867 .maxlen = sizeof(int),
5868 .mode = 0644,
5869 .proc_handler = proc_dointvec
5872 .procname = "ndisc_notify",
5873 .data = &ipv6_devconf.ndisc_notify,
5874 .maxlen = sizeof(int),
5875 .mode = 0644,
5876 .proc_handler = proc_dointvec
5879 .procname = "suppress_frag_ndisc",
5880 .data = &ipv6_devconf.suppress_frag_ndisc,
5881 .maxlen = sizeof(int),
5882 .mode = 0644,
5883 .proc_handler = proc_dointvec
5886 .procname = "accept_ra_from_local",
5887 .data = &ipv6_devconf.accept_ra_from_local,
5888 .maxlen = sizeof(int),
5889 .mode = 0644,
5890 .proc_handler = proc_dointvec,
5893 .procname = "accept_ra_mtu",
5894 .data = &ipv6_devconf.accept_ra_mtu,
5895 .maxlen = sizeof(int),
5896 .mode = 0644,
5897 .proc_handler = proc_dointvec,
5900 .procname = "stable_secret",
5901 .data = &ipv6_devconf.stable_secret,
5902 .maxlen = IPV6_MAX_STRLEN,
5903 .mode = 0600,
5904 .proc_handler = addrconf_sysctl_stable_secret,
5907 .procname = "use_oif_addrs_only",
5908 .data = &ipv6_devconf.use_oif_addrs_only,
5909 .maxlen = sizeof(int),
5910 .mode = 0644,
5911 .proc_handler = proc_dointvec,
5914 .procname = "ignore_routes_with_linkdown",
5915 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5916 .maxlen = sizeof(int),
5917 .mode = 0644,
5918 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5921 .procname = "drop_unicast_in_l2_multicast",
5922 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
5923 .maxlen = sizeof(int),
5924 .mode = 0644,
5925 .proc_handler = proc_dointvec,
5928 .procname = "drop_unsolicited_na",
5929 .data = &ipv6_devconf.drop_unsolicited_na,
5930 .maxlen = sizeof(int),
5931 .mode = 0644,
5932 .proc_handler = proc_dointvec,
5935 .procname = "keep_addr_on_down",
5936 .data = &ipv6_devconf.keep_addr_on_down,
5937 .maxlen = sizeof(int),
5938 .mode = 0644,
5939 .proc_handler = proc_dointvec,
5943 /* sentinel */
5948 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5949 struct inet6_dev *idev, struct ipv6_devconf *p)
5951 int i;
5952 struct addrconf_sysctl_table *t;
5953 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5955 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5956 if (!t)
5957 goto out;
5959 for (i = 0; t->addrconf_vars[i].data; i++) {
5960 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5961 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5962 t->addrconf_vars[i].extra2 = net;
5965 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5967 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5968 if (!t->sysctl_header)
5969 goto free;
5971 p->sysctl = t;
5972 return 0;
5974 free:
5975 kfree(t);
5976 out:
5977 return -ENOBUFS;
5980 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5982 struct addrconf_sysctl_table *t;
5984 if (!p->sysctl)
5985 return;
5987 t = p->sysctl;
5988 p->sysctl = NULL;
5989 unregister_net_sysctl_table(t->sysctl_header);
5990 kfree(t);
5993 static int addrconf_sysctl_register(struct inet6_dev *idev)
5995 int err;
5997 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5998 return -EINVAL;
6000 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6001 &ndisc_ifinfo_sysctl_change);
6002 if (err)
6003 return err;
6004 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6005 idev, &idev->cnf);
6006 if (err)
6007 neigh_sysctl_unregister(idev->nd_parms);
6009 return err;
6012 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6014 __addrconf_sysctl_unregister(&idev->cnf);
6015 neigh_sysctl_unregister(idev->nd_parms);
6019 #endif
6021 static int __net_init addrconf_init_net(struct net *net)
6023 int err = -ENOMEM;
6024 struct ipv6_devconf *all, *dflt;
6026 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6027 if (!all)
6028 goto err_alloc_all;
6030 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6031 if (!dflt)
6032 goto err_alloc_dflt;
6034 /* these will be inherited by all namespaces */
6035 dflt->autoconf = ipv6_defaults.autoconf;
6036 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6038 dflt->stable_secret.initialized = false;
6039 all->stable_secret.initialized = false;
6041 net->ipv6.devconf_all = all;
6042 net->ipv6.devconf_dflt = dflt;
6044 #ifdef CONFIG_SYSCTL
6045 err = __addrconf_sysctl_register(net, "all", NULL, all);
6046 if (err < 0)
6047 goto err_reg_all;
6049 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6050 if (err < 0)
6051 goto err_reg_dflt;
6052 #endif
6053 return 0;
6055 #ifdef CONFIG_SYSCTL
6056 err_reg_dflt:
6057 __addrconf_sysctl_unregister(all);
6058 err_reg_all:
6059 kfree(dflt);
6060 #endif
6061 err_alloc_dflt:
6062 kfree(all);
6063 err_alloc_all:
6064 return err;
6067 static void __net_exit addrconf_exit_net(struct net *net)
6069 #ifdef CONFIG_SYSCTL
6070 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6071 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6072 #endif
6073 kfree(net->ipv6.devconf_dflt);
6074 kfree(net->ipv6.devconf_all);
6077 static struct pernet_operations addrconf_ops = {
6078 .init = addrconf_init_net,
6079 .exit = addrconf_exit_net,
6082 static struct rtnl_af_ops inet6_ops __read_mostly = {
6083 .family = AF_INET6,
6084 .fill_link_af = inet6_fill_link_af,
6085 .get_link_af_size = inet6_get_link_af_size,
6086 .validate_link_af = inet6_validate_link_af,
6087 .set_link_af = inet6_set_link_af,
6091 * Init / cleanup code
6094 int __init addrconf_init(void)
6096 struct inet6_dev *idev;
6097 int i, err;
6099 err = ipv6_addr_label_init();
6100 if (err < 0) {
6101 pr_crit("%s: cannot initialize default policy table: %d\n",
6102 __func__, err);
6103 goto out;
6106 err = register_pernet_subsys(&addrconf_ops);
6107 if (err < 0)
6108 goto out_addrlabel;
6110 addrconf_wq = create_workqueue("ipv6_addrconf");
6111 if (!addrconf_wq) {
6112 err = -ENOMEM;
6113 goto out_nowq;
6116 /* The addrconf netdev notifier requires that loopback_dev
6117 * has it's ipv6 private information allocated and setup
6118 * before it can bring up and give link-local addresses
6119 * to other devices which are up.
6121 * Unfortunately, loopback_dev is not necessarily the first
6122 * entry in the global dev_base list of net devices. In fact,
6123 * it is likely to be the very last entry on that list.
6124 * So this causes the notifier registry below to try and
6125 * give link-local addresses to all devices besides loopback_dev
6126 * first, then loopback_dev, which cases all the non-loopback_dev
6127 * devices to fail to get a link-local address.
6129 * So, as a temporary fix, allocate the ipv6 structure for
6130 * loopback_dev first by hand.
6131 * Longer term, all of the dependencies ipv6 has upon the loopback
6132 * device and it being up should be removed.
6134 rtnl_lock();
6135 idev = ipv6_add_dev(init_net.loopback_dev);
6136 rtnl_unlock();
6137 if (IS_ERR(idev)) {
6138 err = PTR_ERR(idev);
6139 goto errlo;
6142 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6143 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6145 register_netdevice_notifier(&ipv6_dev_notf);
6147 addrconf_verify();
6149 rtnl_af_register(&inet6_ops);
6151 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6152 NULL);
6153 if (err < 0)
6154 goto errout;
6156 /* Only the first call to __rtnl_register can fail */
6157 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6158 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6159 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6160 inet6_dump_ifaddr, NULL);
6161 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6162 inet6_dump_ifmcaddr, NULL);
6163 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6164 inet6_dump_ifacaddr, NULL);
6165 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6166 inet6_netconf_dump_devconf, NULL);
6168 ipv6_addr_label_rtnl_register();
6170 return 0;
6171 errout:
6172 rtnl_af_unregister(&inet6_ops);
6173 unregister_netdevice_notifier(&ipv6_dev_notf);
6174 errlo:
6175 destroy_workqueue(addrconf_wq);
6176 out_nowq:
6177 unregister_pernet_subsys(&addrconf_ops);
6178 out_addrlabel:
6179 ipv6_addr_label_cleanup();
6180 out:
6181 return err;
6184 void addrconf_cleanup(void)
6186 struct net_device *dev;
6187 int i;
6189 unregister_netdevice_notifier(&ipv6_dev_notf);
6190 unregister_pernet_subsys(&addrconf_ops);
6191 ipv6_addr_label_cleanup();
6193 rtnl_lock();
6195 __rtnl_af_unregister(&inet6_ops);
6197 /* clean dev list */
6198 for_each_netdev(&init_net, dev) {
6199 if (__in6_dev_get(dev) == NULL)
6200 continue;
6201 addrconf_ifdown(dev, 1);
6203 addrconf_ifdown(init_net.loopback_dev, 2);
6206 * Check hash table.
6208 spin_lock_bh(&addrconf_hash_lock);
6209 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6210 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6211 spin_unlock_bh(&addrconf_hash_lock);
6212 cancel_delayed_work(&addr_chk_work);
6213 rtnl_unlock();
6215 destroy_workqueue(addrconf_wq);