ipv6: bump genid when the IFA_F_TENTATIVE flag is clear
[linux-2.6/btrfs-unstable.git] / net / ipv6 / addrconf.c
blob4bc5ba3ae452eb99ed3a70c0926f6bb1ed5c001a
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 static inline s32 rfc3315_s14_backoff_init(s32 irt)
117 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
118 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
119 do_div(tmp, 1000000);
120 return (s32)tmp;
123 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
125 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
126 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
127 do_div(tmp, 1000000);
128 if ((s32)tmp > mrt) {
129 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
130 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
131 do_div(tmp, 1000000);
133 return (s32)tmp;
136 #ifdef CONFIG_SYSCTL
137 static int addrconf_sysctl_register(struct inet6_dev *idev);
138 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
139 #else
140 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
142 return 0;
145 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
148 #endif
150 static void ipv6_regen_rndid(struct inet6_dev *idev);
151 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
153 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
154 static int ipv6_count_addresses(struct inet6_dev *idev);
155 static int ipv6_generate_stable_address(struct in6_addr *addr,
156 u8 dad_count,
157 const struct inet6_dev *idev);
160 * Configured unicast address hash table
162 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
163 static DEFINE_SPINLOCK(addrconf_hash_lock);
165 static void addrconf_verify(void);
166 static void addrconf_verify_rtnl(void);
167 static void addrconf_verify_work(struct work_struct *);
169 static struct workqueue_struct *addrconf_wq;
170 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
172 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
173 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
175 static void addrconf_type_change(struct net_device *dev,
176 unsigned long event);
177 static int addrconf_ifdown(struct net_device *dev, int how);
179 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
180 int plen,
181 const struct net_device *dev,
182 u32 flags, u32 noflags);
184 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
185 static void addrconf_dad_work(struct work_struct *w);
186 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
187 static void addrconf_dad_run(struct inet6_dev *idev);
188 static void addrconf_rs_timer(unsigned long data);
189 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
190 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
192 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
193 struct prefix_info *pinfo);
194 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
195 struct net_device *dev);
197 static struct ipv6_devconf ipv6_devconf __read_mostly = {
198 .forwarding = 0,
199 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
200 .mtu6 = IPV6_MIN_MTU,
201 .accept_ra = 1,
202 .accept_redirects = 1,
203 .autoconf = 1,
204 .force_mld_version = 0,
205 .mldv1_unsolicited_report_interval = 10 * HZ,
206 .mldv2_unsolicited_report_interval = HZ,
207 .dad_transmits = 1,
208 .rtr_solicits = MAX_RTR_SOLICITATIONS,
209 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
210 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
211 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
212 .use_tempaddr = 0,
213 .temp_valid_lft = TEMP_VALID_LIFETIME,
214 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
215 .regen_max_retry = REGEN_MAX_RETRY,
216 .max_desync_factor = MAX_DESYNC_FACTOR,
217 .max_addresses = IPV6_MAX_ADDRESSES,
218 .accept_ra_defrtr = 1,
219 .accept_ra_from_local = 0,
220 .accept_ra_min_hop_limit= 1,
221 .accept_ra_pinfo = 1,
222 #ifdef CONFIG_IPV6_ROUTER_PREF
223 .accept_ra_rtr_pref = 1,
224 .rtr_probe_interval = 60 * HZ,
225 #ifdef CONFIG_IPV6_ROUTE_INFO
226 .accept_ra_rt_info_max_plen = 0,
227 #endif
228 #endif
229 .proxy_ndp = 0,
230 .accept_source_route = 0, /* we do not accept RH0 by default. */
231 .disable_ipv6 = 0,
232 .accept_dad = 1,
233 .suppress_frag_ndisc = 1,
234 .accept_ra_mtu = 1,
235 .stable_secret = {
236 .initialized = false,
238 .use_oif_addrs_only = 0,
239 .ignore_routes_with_linkdown = 0,
240 .keep_addr_on_down = 0,
243 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
244 .forwarding = 0,
245 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
246 .mtu6 = IPV6_MIN_MTU,
247 .accept_ra = 1,
248 .accept_redirects = 1,
249 .autoconf = 1,
250 .force_mld_version = 0,
251 .mldv1_unsolicited_report_interval = 10 * HZ,
252 .mldv2_unsolicited_report_interval = HZ,
253 .dad_transmits = 1,
254 .rtr_solicits = MAX_RTR_SOLICITATIONS,
255 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
256 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
257 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
258 .use_tempaddr = 0,
259 .temp_valid_lft = TEMP_VALID_LIFETIME,
260 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
261 .regen_max_retry = REGEN_MAX_RETRY,
262 .max_desync_factor = MAX_DESYNC_FACTOR,
263 .max_addresses = IPV6_MAX_ADDRESSES,
264 .accept_ra_defrtr = 1,
265 .accept_ra_from_local = 0,
266 .accept_ra_min_hop_limit= 1,
267 .accept_ra_pinfo = 1,
268 #ifdef CONFIG_IPV6_ROUTER_PREF
269 .accept_ra_rtr_pref = 1,
270 .rtr_probe_interval = 60 * HZ,
271 #ifdef CONFIG_IPV6_ROUTE_INFO
272 .accept_ra_rt_info_max_plen = 0,
273 #endif
274 #endif
275 .proxy_ndp = 0,
276 .accept_source_route = 0, /* we do not accept RH0 by default. */
277 .disable_ipv6 = 0,
278 .accept_dad = 1,
279 .suppress_frag_ndisc = 1,
280 .accept_ra_mtu = 1,
281 .stable_secret = {
282 .initialized = false,
284 .use_oif_addrs_only = 0,
285 .ignore_routes_with_linkdown = 0,
286 .keep_addr_on_down = 0,
289 /* Check if a valid qdisc is available */
290 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
292 return !qdisc_tx_is_noop(dev);
295 static void addrconf_del_rs_timer(struct inet6_dev *idev)
297 if (del_timer(&idev->rs_timer))
298 __in6_dev_put(idev);
301 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
303 if (cancel_delayed_work(&ifp->dad_work))
304 __in6_ifa_put(ifp);
307 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
308 unsigned long when)
310 if (!timer_pending(&idev->rs_timer))
311 in6_dev_hold(idev);
312 mod_timer(&idev->rs_timer, jiffies + when);
315 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
316 unsigned long delay)
318 if (!delayed_work_pending(&ifp->dad_work))
319 in6_ifa_hold(ifp);
320 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
323 static int snmp6_alloc_dev(struct inet6_dev *idev)
325 int i;
327 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
328 if (!idev->stats.ipv6)
329 goto err_ip;
331 for_each_possible_cpu(i) {
332 struct ipstats_mib *addrconf_stats;
333 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
334 u64_stats_init(&addrconf_stats->syncp);
338 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
339 GFP_KERNEL);
340 if (!idev->stats.icmpv6dev)
341 goto err_icmp;
342 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
343 GFP_KERNEL);
344 if (!idev->stats.icmpv6msgdev)
345 goto err_icmpmsg;
347 return 0;
349 err_icmpmsg:
350 kfree(idev->stats.icmpv6dev);
351 err_icmp:
352 free_percpu(idev->stats.ipv6);
353 err_ip:
354 return -ENOMEM;
357 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
359 struct inet6_dev *ndev;
360 int err = -ENOMEM;
362 ASSERT_RTNL();
364 if (dev->mtu < IPV6_MIN_MTU)
365 return ERR_PTR(-EINVAL);
367 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
368 if (!ndev)
369 return ERR_PTR(err);
371 rwlock_init(&ndev->lock);
372 ndev->dev = dev;
373 INIT_LIST_HEAD(&ndev->addr_list);
374 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
375 (unsigned long)ndev);
376 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
378 if (ndev->cnf.stable_secret.initialized)
379 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
380 else
381 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
383 ndev->cnf.mtu6 = dev->mtu;
384 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
385 if (!ndev->nd_parms) {
386 kfree(ndev);
387 return ERR_PTR(err);
389 if (ndev->cnf.forwarding)
390 dev_disable_lro(dev);
391 /* We refer to the device */
392 dev_hold(dev);
394 if (snmp6_alloc_dev(ndev) < 0) {
395 ADBG(KERN_WARNING
396 "%s: cannot allocate memory for statistics; dev=%s.\n",
397 __func__, dev->name);
398 neigh_parms_release(&nd_tbl, ndev->nd_parms);
399 dev_put(dev);
400 kfree(ndev);
401 return ERR_PTR(err);
404 if (snmp6_register_dev(ndev) < 0) {
405 ADBG(KERN_WARNING
406 "%s: cannot create /proc/net/dev_snmp6/%s\n",
407 __func__, dev->name);
408 goto err_release;
411 /* One reference from device. */
412 in6_dev_hold(ndev);
414 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
415 ndev->cnf.accept_dad = -1;
417 #if IS_ENABLED(CONFIG_IPV6_SIT)
418 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
419 pr_info("%s: Disabled Multicast RS\n", dev->name);
420 ndev->cnf.rtr_solicits = 0;
422 #endif
424 INIT_LIST_HEAD(&ndev->tempaddr_list);
425 ndev->desync_factor = U32_MAX;
426 if ((dev->flags&IFF_LOOPBACK) ||
427 dev->type == ARPHRD_TUNNEL ||
428 dev->type == ARPHRD_TUNNEL6 ||
429 dev->type == ARPHRD_SIT ||
430 dev->type == ARPHRD_NONE) {
431 ndev->cnf.use_tempaddr = -1;
432 } else
433 ipv6_regen_rndid(ndev);
435 ndev->token = in6addr_any;
437 if (netif_running(dev) && addrconf_qdisc_ok(dev))
438 ndev->if_flags |= IF_READY;
440 ipv6_mc_init_dev(ndev);
441 ndev->tstamp = jiffies;
442 err = addrconf_sysctl_register(ndev);
443 if (err) {
444 ipv6_mc_destroy_dev(ndev);
445 snmp6_unregister_dev(ndev);
446 goto err_release;
448 /* protected by rtnl_lock */
449 rcu_assign_pointer(dev->ip6_ptr, ndev);
451 /* Join interface-local all-node multicast group */
452 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
454 /* Join all-node multicast group */
455 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
457 /* Join all-router multicast group if forwarding is set */
458 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
459 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
461 return ndev;
463 err_release:
464 neigh_parms_release(&nd_tbl, ndev->nd_parms);
465 ndev->dead = 1;
466 in6_dev_finish_destroy(ndev);
467 return ERR_PTR(err);
470 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
472 struct inet6_dev *idev;
474 ASSERT_RTNL();
476 idev = __in6_dev_get(dev);
477 if (!idev) {
478 idev = ipv6_add_dev(dev);
479 if (IS_ERR(idev))
480 return NULL;
483 if (dev->flags&IFF_UP)
484 ipv6_mc_up(idev);
485 return idev;
488 static int inet6_netconf_msgsize_devconf(int type)
490 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
491 + nla_total_size(4); /* NETCONFA_IFINDEX */
492 bool all = false;
494 if (type == NETCONFA_ALL)
495 all = true;
497 if (all || type == NETCONFA_FORWARDING)
498 size += nla_total_size(4);
499 #ifdef CONFIG_IPV6_MROUTE
500 if (all || type == NETCONFA_MC_FORWARDING)
501 size += nla_total_size(4);
502 #endif
503 if (all || type == NETCONFA_PROXY_NEIGH)
504 size += nla_total_size(4);
506 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
507 size += nla_total_size(4);
509 return size;
512 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
513 struct ipv6_devconf *devconf, u32 portid,
514 u32 seq, int event, unsigned int flags,
515 int type)
517 struct nlmsghdr *nlh;
518 struct netconfmsg *ncm;
519 bool all = false;
521 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
522 flags);
523 if (!nlh)
524 return -EMSGSIZE;
526 if (type == NETCONFA_ALL)
527 all = true;
529 ncm = nlmsg_data(nlh);
530 ncm->ncm_family = AF_INET6;
532 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
533 goto nla_put_failure;
535 if ((all || type == NETCONFA_FORWARDING) &&
536 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
537 goto nla_put_failure;
538 #ifdef CONFIG_IPV6_MROUTE
539 if ((all || type == NETCONFA_MC_FORWARDING) &&
540 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
541 devconf->mc_forwarding) < 0)
542 goto nla_put_failure;
543 #endif
544 if ((all || type == NETCONFA_PROXY_NEIGH) &&
545 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
546 goto nla_put_failure;
548 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
549 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
550 devconf->ignore_routes_with_linkdown) < 0)
551 goto nla_put_failure;
553 nlmsg_end(skb, nlh);
554 return 0;
556 nla_put_failure:
557 nlmsg_cancel(skb, nlh);
558 return -EMSGSIZE;
561 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
562 struct ipv6_devconf *devconf)
564 struct sk_buff *skb;
565 int err = -ENOBUFS;
567 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
568 if (!skb)
569 goto errout;
571 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
572 RTM_NEWNETCONF, 0, type);
573 if (err < 0) {
574 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
575 WARN_ON(err == -EMSGSIZE);
576 kfree_skb(skb);
577 goto errout;
579 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
580 return;
581 errout:
582 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
585 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
586 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
587 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
588 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
589 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
592 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
593 struct nlmsghdr *nlh)
595 struct net *net = sock_net(in_skb->sk);
596 struct nlattr *tb[NETCONFA_MAX+1];
597 struct netconfmsg *ncm;
598 struct sk_buff *skb;
599 struct ipv6_devconf *devconf;
600 struct inet6_dev *in6_dev;
601 struct net_device *dev;
602 int ifindex;
603 int err;
605 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
606 devconf_ipv6_policy);
607 if (err < 0)
608 goto errout;
610 err = -EINVAL;
611 if (!tb[NETCONFA_IFINDEX])
612 goto errout;
614 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
615 switch (ifindex) {
616 case NETCONFA_IFINDEX_ALL:
617 devconf = net->ipv6.devconf_all;
618 break;
619 case NETCONFA_IFINDEX_DEFAULT:
620 devconf = net->ipv6.devconf_dflt;
621 break;
622 default:
623 dev = __dev_get_by_index(net, ifindex);
624 if (!dev)
625 goto errout;
626 in6_dev = __in6_dev_get(dev);
627 if (!in6_dev)
628 goto errout;
629 devconf = &in6_dev->cnf;
630 break;
633 err = -ENOBUFS;
634 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
635 if (!skb)
636 goto errout;
638 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
639 NETLINK_CB(in_skb).portid,
640 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
641 NETCONFA_ALL);
642 if (err < 0) {
643 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
644 WARN_ON(err == -EMSGSIZE);
645 kfree_skb(skb);
646 goto errout;
648 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
649 errout:
650 return err;
653 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
654 struct netlink_callback *cb)
656 struct net *net = sock_net(skb->sk);
657 int h, s_h;
658 int idx, s_idx;
659 struct net_device *dev;
660 struct inet6_dev *idev;
661 struct hlist_head *head;
663 s_h = cb->args[0];
664 s_idx = idx = cb->args[1];
666 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
667 idx = 0;
668 head = &net->dev_index_head[h];
669 rcu_read_lock();
670 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
671 net->dev_base_seq;
672 hlist_for_each_entry_rcu(dev, head, index_hlist) {
673 if (idx < s_idx)
674 goto cont;
675 idev = __in6_dev_get(dev);
676 if (!idev)
677 goto cont;
679 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
680 &idev->cnf,
681 NETLINK_CB(cb->skb).portid,
682 cb->nlh->nlmsg_seq,
683 RTM_NEWNETCONF,
684 NLM_F_MULTI,
685 NETCONFA_ALL) < 0) {
686 rcu_read_unlock();
687 goto done;
689 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
690 cont:
691 idx++;
693 rcu_read_unlock();
695 if (h == NETDEV_HASHENTRIES) {
696 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
697 net->ipv6.devconf_all,
698 NETLINK_CB(cb->skb).portid,
699 cb->nlh->nlmsg_seq,
700 RTM_NEWNETCONF, NLM_F_MULTI,
701 NETCONFA_ALL) < 0)
702 goto done;
703 else
704 h++;
706 if (h == NETDEV_HASHENTRIES + 1) {
707 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
708 net->ipv6.devconf_dflt,
709 NETLINK_CB(cb->skb).portid,
710 cb->nlh->nlmsg_seq,
711 RTM_NEWNETCONF, NLM_F_MULTI,
712 NETCONFA_ALL) < 0)
713 goto done;
714 else
715 h++;
717 done:
718 cb->args[0] = h;
719 cb->args[1] = idx;
721 return skb->len;
724 #ifdef CONFIG_SYSCTL
725 static void dev_forward_change(struct inet6_dev *idev)
727 struct net_device *dev;
728 struct inet6_ifaddr *ifa;
730 if (!idev)
731 return;
732 dev = idev->dev;
733 if (idev->cnf.forwarding)
734 dev_disable_lro(dev);
735 if (dev->flags & IFF_MULTICAST) {
736 if (idev->cnf.forwarding) {
737 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
738 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
739 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
740 } else {
741 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
742 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
743 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
747 list_for_each_entry(ifa, &idev->addr_list, if_list) {
748 if (ifa->flags&IFA_F_TENTATIVE)
749 continue;
750 if (idev->cnf.forwarding)
751 addrconf_join_anycast(ifa);
752 else
753 addrconf_leave_anycast(ifa);
755 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
756 dev->ifindex, &idev->cnf);
760 static void addrconf_forward_change(struct net *net, __s32 newf)
762 struct net_device *dev;
763 struct inet6_dev *idev;
765 for_each_netdev(net, dev) {
766 idev = __in6_dev_get(dev);
767 if (idev) {
768 int changed = (!idev->cnf.forwarding) ^ (!newf);
769 idev->cnf.forwarding = newf;
770 if (changed)
771 dev_forward_change(idev);
776 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
778 struct net *net;
779 int old;
781 if (!rtnl_trylock())
782 return restart_syscall();
784 net = (struct net *)table->extra2;
785 old = *p;
786 *p = newf;
788 if (p == &net->ipv6.devconf_dflt->forwarding) {
789 if ((!newf) ^ (!old))
790 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
791 NETCONFA_IFINDEX_DEFAULT,
792 net->ipv6.devconf_dflt);
793 rtnl_unlock();
794 return 0;
797 if (p == &net->ipv6.devconf_all->forwarding) {
798 int old_dflt = net->ipv6.devconf_dflt->forwarding;
800 net->ipv6.devconf_dflt->forwarding = newf;
801 if ((!newf) ^ (!old_dflt))
802 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
803 NETCONFA_IFINDEX_DEFAULT,
804 net->ipv6.devconf_dflt);
806 addrconf_forward_change(net, newf);
807 if ((!newf) ^ (!old))
808 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
809 NETCONFA_IFINDEX_ALL,
810 net->ipv6.devconf_all);
811 } else if ((!newf) ^ (!old))
812 dev_forward_change((struct inet6_dev *)table->extra1);
813 rtnl_unlock();
815 if (newf)
816 rt6_purge_dflt_routers(net);
817 return 1;
820 static void addrconf_linkdown_change(struct net *net, __s32 newf)
822 struct net_device *dev;
823 struct inet6_dev *idev;
825 for_each_netdev(net, dev) {
826 idev = __in6_dev_get(dev);
827 if (idev) {
828 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
830 idev->cnf.ignore_routes_with_linkdown = newf;
831 if (changed)
832 inet6_netconf_notify_devconf(dev_net(dev),
833 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
834 dev->ifindex,
835 &idev->cnf);
840 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
842 struct net *net;
843 int old;
845 if (!rtnl_trylock())
846 return restart_syscall();
848 net = (struct net *)table->extra2;
849 old = *p;
850 *p = newf;
852 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
853 if ((!newf) ^ (!old))
854 inet6_netconf_notify_devconf(net,
855 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
856 NETCONFA_IFINDEX_DEFAULT,
857 net->ipv6.devconf_dflt);
858 rtnl_unlock();
859 return 0;
862 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
863 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
864 addrconf_linkdown_change(net, newf);
865 if ((!newf) ^ (!old))
866 inet6_netconf_notify_devconf(net,
867 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
868 NETCONFA_IFINDEX_ALL,
869 net->ipv6.devconf_all);
871 rtnl_unlock();
873 return 1;
876 #endif
878 /* Nobody refers to this ifaddr, destroy it */
879 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
881 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
883 #ifdef NET_REFCNT_DEBUG
884 pr_debug("%s\n", __func__);
885 #endif
887 in6_dev_put(ifp->idev);
889 if (cancel_delayed_work(&ifp->dad_work))
890 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
891 ifp);
893 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
894 pr_warn("Freeing alive inet6 address %p\n", ifp);
895 return;
897 ip6_rt_put(ifp->rt);
899 kfree_rcu(ifp, rcu);
902 static void
903 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
905 struct list_head *p;
906 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
909 * Each device address list is sorted in order of scope -
910 * global before linklocal.
912 list_for_each(p, &idev->addr_list) {
913 struct inet6_ifaddr *ifa
914 = list_entry(p, struct inet6_ifaddr, if_list);
915 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
916 break;
919 list_add_tail(&ifp->if_list, p);
922 static u32 inet6_addr_hash(const struct in6_addr *addr)
924 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
927 /* On success it returns ifp with increased reference count */
929 static struct inet6_ifaddr *
930 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
931 const struct in6_addr *peer_addr, int pfxlen,
932 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
934 struct inet6_ifaddr *ifa = NULL;
935 struct rt6_info *rt;
936 unsigned int hash;
937 int err = 0;
938 int addr_type = ipv6_addr_type(addr);
940 if (addr_type == IPV6_ADDR_ANY ||
941 addr_type & IPV6_ADDR_MULTICAST ||
942 (!(idev->dev->flags & IFF_LOOPBACK) &&
943 addr_type & IPV6_ADDR_LOOPBACK))
944 return ERR_PTR(-EADDRNOTAVAIL);
946 rcu_read_lock_bh();
947 if (idev->dead) {
948 err = -ENODEV; /*XXX*/
949 goto out2;
952 if (idev->cnf.disable_ipv6) {
953 err = -EACCES;
954 goto out2;
957 spin_lock(&addrconf_hash_lock);
959 /* Ignore adding duplicate addresses on an interface */
960 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
961 ADBG("ipv6_add_addr: already assigned\n");
962 err = -EEXIST;
963 goto out;
966 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
968 if (!ifa) {
969 ADBG("ipv6_add_addr: malloc failed\n");
970 err = -ENOBUFS;
971 goto out;
974 rt = addrconf_dst_alloc(idev, addr, false);
975 if (IS_ERR(rt)) {
976 err = PTR_ERR(rt);
977 goto out;
980 neigh_parms_data_state_setall(idev->nd_parms);
982 ifa->addr = *addr;
983 if (peer_addr)
984 ifa->peer_addr = *peer_addr;
986 spin_lock_init(&ifa->lock);
987 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
988 INIT_HLIST_NODE(&ifa->addr_lst);
989 ifa->scope = scope;
990 ifa->prefix_len = pfxlen;
991 ifa->flags = flags | IFA_F_TENTATIVE;
992 ifa->valid_lft = valid_lft;
993 ifa->prefered_lft = prefered_lft;
994 ifa->cstamp = ifa->tstamp = jiffies;
995 ifa->tokenized = false;
997 ifa->rt = rt;
999 ifa->idev = idev;
1000 in6_dev_hold(idev);
1001 /* For caller */
1002 in6_ifa_hold(ifa);
1004 /* Add to big hash table */
1005 hash = inet6_addr_hash(addr);
1007 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1008 spin_unlock(&addrconf_hash_lock);
1010 write_lock(&idev->lock);
1011 /* Add to inet6_dev unicast addr list. */
1012 ipv6_link_dev_addr(idev, ifa);
1014 if (ifa->flags&IFA_F_TEMPORARY) {
1015 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1016 in6_ifa_hold(ifa);
1019 in6_ifa_hold(ifa);
1020 write_unlock(&idev->lock);
1021 out2:
1022 rcu_read_unlock_bh();
1024 if (likely(err == 0))
1025 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1026 else {
1027 kfree(ifa);
1028 ifa = ERR_PTR(err);
1031 return ifa;
1032 out:
1033 spin_unlock(&addrconf_hash_lock);
1034 goto out2;
1037 enum cleanup_prefix_rt_t {
1038 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1039 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1040 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1044 * Check, whether the prefix for ifp would still need a prefix route
1045 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1046 * constants.
1048 * 1) we don't purge prefix if address was not permanent.
1049 * prefix is managed by its own lifetime.
1050 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1051 * 3) if there are no addresses, delete prefix.
1052 * 4) if there are still other permanent address(es),
1053 * corresponding prefix is still permanent.
1054 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1055 * don't purge the prefix, assume user space is managing it.
1056 * 6) otherwise, update prefix lifetime to the
1057 * longest valid lifetime among the corresponding
1058 * addresses on the device.
1059 * Note: subsequent RA will update lifetime.
1061 static enum cleanup_prefix_rt_t
1062 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1064 struct inet6_ifaddr *ifa;
1065 struct inet6_dev *idev = ifp->idev;
1066 unsigned long lifetime;
1067 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1069 *expires = jiffies;
1071 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1072 if (ifa == ifp)
1073 continue;
1074 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1075 ifp->prefix_len))
1076 continue;
1077 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1078 return CLEANUP_PREFIX_RT_NOP;
1080 action = CLEANUP_PREFIX_RT_EXPIRE;
1082 spin_lock(&ifa->lock);
1084 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1086 * Note: Because this address is
1087 * not permanent, lifetime <
1088 * LONG_MAX / HZ here.
1090 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1091 *expires = ifa->tstamp + lifetime * HZ;
1092 spin_unlock(&ifa->lock);
1095 return action;
1098 static void
1099 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1101 struct rt6_info *rt;
1103 rt = addrconf_get_prefix_route(&ifp->addr,
1104 ifp->prefix_len,
1105 ifp->idev->dev,
1106 0, RTF_GATEWAY | RTF_DEFAULT);
1107 if (rt) {
1108 if (del_rt)
1109 ip6_del_rt(rt);
1110 else {
1111 if (!(rt->rt6i_flags & RTF_EXPIRES))
1112 rt6_set_expires(rt, expires);
1113 ip6_rt_put(rt);
1119 /* This function wants to get referenced ifp and releases it before return */
1121 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1123 int state;
1124 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1125 unsigned long expires;
1127 ASSERT_RTNL();
1129 spin_lock_bh(&ifp->lock);
1130 state = ifp->state;
1131 ifp->state = INET6_IFADDR_STATE_DEAD;
1132 spin_unlock_bh(&ifp->lock);
1134 if (state == INET6_IFADDR_STATE_DEAD)
1135 goto out;
1137 spin_lock_bh(&addrconf_hash_lock);
1138 hlist_del_init_rcu(&ifp->addr_lst);
1139 spin_unlock_bh(&addrconf_hash_lock);
1141 write_lock_bh(&ifp->idev->lock);
1143 if (ifp->flags&IFA_F_TEMPORARY) {
1144 list_del(&ifp->tmp_list);
1145 if (ifp->ifpub) {
1146 in6_ifa_put(ifp->ifpub);
1147 ifp->ifpub = NULL;
1149 __in6_ifa_put(ifp);
1152 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1153 action = check_cleanup_prefix_route(ifp, &expires);
1155 list_del_init(&ifp->if_list);
1156 __in6_ifa_put(ifp);
1158 write_unlock_bh(&ifp->idev->lock);
1160 addrconf_del_dad_work(ifp);
1162 ipv6_ifa_notify(RTM_DELADDR, ifp);
1164 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1166 if (action != CLEANUP_PREFIX_RT_NOP) {
1167 cleanup_prefix_route(ifp, expires,
1168 action == CLEANUP_PREFIX_RT_DEL);
1171 /* clean up prefsrc entries */
1172 rt6_remove_prefsrc(ifp);
1173 out:
1174 in6_ifa_put(ifp);
1177 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1179 struct inet6_dev *idev = ifp->idev;
1180 struct in6_addr addr, *tmpaddr;
1181 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1182 unsigned long regen_advance;
1183 int tmp_plen;
1184 int ret = 0;
1185 u32 addr_flags;
1186 unsigned long now = jiffies;
1187 long max_desync_factor;
1188 s32 cnf_temp_preferred_lft;
1190 write_lock_bh(&idev->lock);
1191 if (ift) {
1192 spin_lock_bh(&ift->lock);
1193 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1194 spin_unlock_bh(&ift->lock);
1195 tmpaddr = &addr;
1196 } else {
1197 tmpaddr = NULL;
1199 retry:
1200 in6_dev_hold(idev);
1201 if (idev->cnf.use_tempaddr <= 0) {
1202 write_unlock_bh(&idev->lock);
1203 pr_info("%s: use_tempaddr is disabled\n", __func__);
1204 in6_dev_put(idev);
1205 ret = -1;
1206 goto out;
1208 spin_lock_bh(&ifp->lock);
1209 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1210 idev->cnf.use_tempaddr = -1; /*XXX*/
1211 spin_unlock_bh(&ifp->lock);
1212 write_unlock_bh(&idev->lock);
1213 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1214 __func__);
1215 in6_dev_put(idev);
1216 ret = -1;
1217 goto out;
1219 in6_ifa_hold(ifp);
1220 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1221 ipv6_try_regen_rndid(idev, tmpaddr);
1222 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1223 age = (now - ifp->tstamp) / HZ;
1225 regen_advance = idev->cnf.regen_max_retry *
1226 idev->cnf.dad_transmits *
1227 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1229 /* recalculate max_desync_factor each time and update
1230 * idev->desync_factor if it's larger
1232 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1233 max_desync_factor = min_t(__u32,
1234 idev->cnf.max_desync_factor,
1235 cnf_temp_preferred_lft - regen_advance);
1237 if (unlikely(idev->desync_factor > max_desync_factor)) {
1238 if (max_desync_factor > 0) {
1239 get_random_bytes(&idev->desync_factor,
1240 sizeof(idev->desync_factor));
1241 idev->desync_factor %= max_desync_factor;
1242 } else {
1243 idev->desync_factor = 0;
1247 tmp_valid_lft = min_t(__u32,
1248 ifp->valid_lft,
1249 idev->cnf.temp_valid_lft + age);
1250 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1251 idev->desync_factor;
1252 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1253 tmp_plen = ifp->prefix_len;
1254 tmp_tstamp = ifp->tstamp;
1255 spin_unlock_bh(&ifp->lock);
1257 write_unlock_bh(&idev->lock);
1259 /* A temporary address is created only if this calculated Preferred
1260 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1261 * an implementation must not create a temporary address with a zero
1262 * Preferred Lifetime.
1263 * Use age calculation as in addrconf_verify to avoid unnecessary
1264 * temporary addresses being generated.
1266 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1267 if (tmp_prefered_lft <= regen_advance + age) {
1268 in6_ifa_put(ifp);
1269 in6_dev_put(idev);
1270 ret = -1;
1271 goto out;
1274 addr_flags = IFA_F_TEMPORARY;
1275 /* set in addrconf_prefix_rcv() */
1276 if (ifp->flags & IFA_F_OPTIMISTIC)
1277 addr_flags |= IFA_F_OPTIMISTIC;
1279 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1280 ipv6_addr_scope(&addr), addr_flags,
1281 tmp_valid_lft, tmp_prefered_lft);
1282 if (IS_ERR(ift)) {
1283 in6_ifa_put(ifp);
1284 in6_dev_put(idev);
1285 pr_info("%s: retry temporary address regeneration\n", __func__);
1286 tmpaddr = &addr;
1287 write_lock_bh(&idev->lock);
1288 goto retry;
1291 spin_lock_bh(&ift->lock);
1292 ift->ifpub = ifp;
1293 ift->cstamp = now;
1294 ift->tstamp = tmp_tstamp;
1295 spin_unlock_bh(&ift->lock);
1297 addrconf_dad_start(ift);
1298 in6_ifa_put(ift);
1299 in6_dev_put(idev);
1300 out:
1301 return ret;
1305 * Choose an appropriate source address (RFC3484)
1307 enum {
1308 IPV6_SADDR_RULE_INIT = 0,
1309 IPV6_SADDR_RULE_LOCAL,
1310 IPV6_SADDR_RULE_SCOPE,
1311 IPV6_SADDR_RULE_PREFERRED,
1312 #ifdef CONFIG_IPV6_MIP6
1313 IPV6_SADDR_RULE_HOA,
1314 #endif
1315 IPV6_SADDR_RULE_OIF,
1316 IPV6_SADDR_RULE_LABEL,
1317 IPV6_SADDR_RULE_PRIVACY,
1318 IPV6_SADDR_RULE_ORCHID,
1319 IPV6_SADDR_RULE_PREFIX,
1320 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1321 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1322 #endif
1323 IPV6_SADDR_RULE_MAX
1326 struct ipv6_saddr_score {
1327 int rule;
1328 int addr_type;
1329 struct inet6_ifaddr *ifa;
1330 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1331 int scopedist;
1332 int matchlen;
1335 struct ipv6_saddr_dst {
1336 const struct in6_addr *addr;
1337 int ifindex;
1338 int scope;
1339 int label;
1340 unsigned int prefs;
1343 static inline int ipv6_saddr_preferred(int type)
1345 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1346 return 1;
1347 return 0;
1350 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1352 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1353 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1354 #else
1355 return false;
1356 #endif
1359 static int ipv6_get_saddr_eval(struct net *net,
1360 struct ipv6_saddr_score *score,
1361 struct ipv6_saddr_dst *dst,
1362 int i)
1364 int ret;
1366 if (i <= score->rule) {
1367 switch (i) {
1368 case IPV6_SADDR_RULE_SCOPE:
1369 ret = score->scopedist;
1370 break;
1371 case IPV6_SADDR_RULE_PREFIX:
1372 ret = score->matchlen;
1373 break;
1374 default:
1375 ret = !!test_bit(i, score->scorebits);
1377 goto out;
1380 switch (i) {
1381 case IPV6_SADDR_RULE_INIT:
1382 /* Rule 0: remember if hiscore is not ready yet */
1383 ret = !!score->ifa;
1384 break;
1385 case IPV6_SADDR_RULE_LOCAL:
1386 /* Rule 1: Prefer same address */
1387 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1388 break;
1389 case IPV6_SADDR_RULE_SCOPE:
1390 /* Rule 2: Prefer appropriate scope
1392 * ret
1394 * -1 | d 15
1395 * ---+--+-+---> scope
1397 * | d is scope of the destination.
1398 * B-d | \
1399 * | \ <- smaller scope is better if
1400 * B-15 | \ if scope is enough for destination.
1401 * | ret = B - scope (-1 <= scope >= d <= 15).
1402 * d-C-1 | /
1403 * |/ <- greater is better
1404 * -C / if scope is not enough for destination.
1405 * /| ret = scope - C (-1 <= d < scope <= 15).
1407 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1408 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1409 * Assume B = 0 and we get C > 29.
1411 ret = __ipv6_addr_src_scope(score->addr_type);
1412 if (ret >= dst->scope)
1413 ret = -ret;
1414 else
1415 ret -= 128; /* 30 is enough */
1416 score->scopedist = ret;
1417 break;
1418 case IPV6_SADDR_RULE_PREFERRED:
1420 /* Rule 3: Avoid deprecated and optimistic addresses */
1421 u8 avoid = IFA_F_DEPRECATED;
1423 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1424 avoid |= IFA_F_OPTIMISTIC;
1425 ret = ipv6_saddr_preferred(score->addr_type) ||
1426 !(score->ifa->flags & avoid);
1427 break;
1429 #ifdef CONFIG_IPV6_MIP6
1430 case IPV6_SADDR_RULE_HOA:
1432 /* Rule 4: Prefer home address */
1433 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1434 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1435 break;
1437 #endif
1438 case IPV6_SADDR_RULE_OIF:
1439 /* Rule 5: Prefer outgoing interface */
1440 ret = (!dst->ifindex ||
1441 dst->ifindex == score->ifa->idev->dev->ifindex);
1442 break;
1443 case IPV6_SADDR_RULE_LABEL:
1444 /* Rule 6: Prefer matching label */
1445 ret = ipv6_addr_label(net,
1446 &score->ifa->addr, score->addr_type,
1447 score->ifa->idev->dev->ifindex) == dst->label;
1448 break;
1449 case IPV6_SADDR_RULE_PRIVACY:
1451 /* Rule 7: Prefer public address
1452 * Note: prefer temporary address if use_tempaddr >= 2
1454 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1455 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1456 score->ifa->idev->cnf.use_tempaddr >= 2;
1457 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1458 break;
1460 case IPV6_SADDR_RULE_ORCHID:
1461 /* Rule 8-: Prefer ORCHID vs ORCHID or
1462 * non-ORCHID vs non-ORCHID
1464 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1465 ipv6_addr_orchid(dst->addr));
1466 break;
1467 case IPV6_SADDR_RULE_PREFIX:
1468 /* Rule 8: Use longest matching prefix */
1469 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1470 if (ret > score->ifa->prefix_len)
1471 ret = score->ifa->prefix_len;
1472 score->matchlen = ret;
1473 break;
1474 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1475 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1476 /* Optimistic addresses still have lower precedence than other
1477 * preferred addresses.
1479 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1480 break;
1481 #endif
1482 default:
1483 ret = 0;
1486 if (ret)
1487 __set_bit(i, score->scorebits);
1488 score->rule = i;
1489 out:
1490 return ret;
1493 static int __ipv6_dev_get_saddr(struct net *net,
1494 struct ipv6_saddr_dst *dst,
1495 struct inet6_dev *idev,
1496 struct ipv6_saddr_score *scores,
1497 int hiscore_idx)
1499 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1501 read_lock_bh(&idev->lock);
1502 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1503 int i;
1506 * - Tentative Address (RFC2462 section 5.4)
1507 * - A tentative address is not considered
1508 * "assigned to an interface" in the traditional
1509 * sense, unless it is also flagged as optimistic.
1510 * - Candidate Source Address (section 4)
1511 * - In any case, anycast addresses, multicast
1512 * addresses, and the unspecified address MUST
1513 * NOT be included in a candidate set.
1515 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1516 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1517 continue;
1519 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1521 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1522 score->addr_type & IPV6_ADDR_MULTICAST)) {
1523 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1524 idev->dev->name);
1525 continue;
1528 score->rule = -1;
1529 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1531 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1532 int minihiscore, miniscore;
1534 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1535 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1537 if (minihiscore > miniscore) {
1538 if (i == IPV6_SADDR_RULE_SCOPE &&
1539 score->scopedist > 0) {
1541 * special case:
1542 * each remaining entry
1543 * has too small (not enough)
1544 * scope, because ifa entries
1545 * are sorted by their scope
1546 * values.
1548 goto out;
1550 break;
1551 } else if (minihiscore < miniscore) {
1552 if (hiscore->ifa)
1553 in6_ifa_put(hiscore->ifa);
1555 in6_ifa_hold(score->ifa);
1557 swap(hiscore, score);
1558 hiscore_idx = 1 - hiscore_idx;
1560 /* restore our iterator */
1561 score->ifa = hiscore->ifa;
1563 break;
1567 out:
1568 read_unlock_bh(&idev->lock);
1569 return hiscore_idx;
1572 static int ipv6_get_saddr_master(struct net *net,
1573 const struct net_device *dst_dev,
1574 const struct net_device *master,
1575 struct ipv6_saddr_dst *dst,
1576 struct ipv6_saddr_score *scores,
1577 int hiscore_idx)
1579 struct inet6_dev *idev;
1581 idev = __in6_dev_get(dst_dev);
1582 if (idev)
1583 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1584 scores, hiscore_idx);
1586 idev = __in6_dev_get(master);
1587 if (idev)
1588 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1589 scores, hiscore_idx);
1591 return hiscore_idx;
1594 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1595 const struct in6_addr *daddr, unsigned int prefs,
1596 struct in6_addr *saddr)
1598 struct ipv6_saddr_score scores[2], *hiscore;
1599 struct ipv6_saddr_dst dst;
1600 struct inet6_dev *idev;
1601 struct net_device *dev;
1602 int dst_type;
1603 bool use_oif_addr = false;
1604 int hiscore_idx = 0;
1606 dst_type = __ipv6_addr_type(daddr);
1607 dst.addr = daddr;
1608 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1609 dst.scope = __ipv6_addr_src_scope(dst_type);
1610 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1611 dst.prefs = prefs;
1613 scores[hiscore_idx].rule = -1;
1614 scores[hiscore_idx].ifa = NULL;
1616 rcu_read_lock();
1618 /* Candidate Source Address (section 4)
1619 * - multicast and link-local destination address,
1620 * the set of candidate source address MUST only
1621 * include addresses assigned to interfaces
1622 * belonging to the same link as the outgoing
1623 * interface.
1624 * (- For site-local destination addresses, the
1625 * set of candidate source addresses MUST only
1626 * include addresses assigned to interfaces
1627 * belonging to the same site as the outgoing
1628 * interface.)
1629 * - "It is RECOMMENDED that the candidate source addresses
1630 * be the set of unicast addresses assigned to the
1631 * interface that will be used to send to the destination
1632 * (the 'outgoing' interface)." (RFC 6724)
1634 if (dst_dev) {
1635 idev = __in6_dev_get(dst_dev);
1636 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1637 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1638 (idev && idev->cnf.use_oif_addrs_only)) {
1639 use_oif_addr = true;
1643 if (use_oif_addr) {
1644 if (idev)
1645 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1646 } else {
1647 const struct net_device *master;
1648 int master_idx = 0;
1650 /* if dst_dev exists and is enslaved to an L3 device, then
1651 * prefer addresses from dst_dev and then the master over
1652 * any other enslaved devices in the L3 domain.
1654 master = l3mdev_master_dev_rcu(dst_dev);
1655 if (master) {
1656 master_idx = master->ifindex;
1658 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1659 master, &dst,
1660 scores, hiscore_idx);
1662 if (scores[hiscore_idx].ifa)
1663 goto out;
1666 for_each_netdev_rcu(net, dev) {
1667 /* only consider addresses on devices in the
1668 * same L3 domain
1670 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1671 continue;
1672 idev = __in6_dev_get(dev);
1673 if (!idev)
1674 continue;
1675 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1679 out:
1680 rcu_read_unlock();
1682 hiscore = &scores[hiscore_idx];
1683 if (!hiscore->ifa)
1684 return -EADDRNOTAVAIL;
1686 *saddr = hiscore->ifa->addr;
1687 in6_ifa_put(hiscore->ifa);
1688 return 0;
1690 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1692 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1693 u32 banned_flags)
1695 struct inet6_ifaddr *ifp;
1696 int err = -EADDRNOTAVAIL;
1698 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1699 if (ifp->scope > IFA_LINK)
1700 break;
1701 if (ifp->scope == IFA_LINK &&
1702 !(ifp->flags & banned_flags)) {
1703 *addr = ifp->addr;
1704 err = 0;
1705 break;
1708 return err;
1711 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1712 u32 banned_flags)
1714 struct inet6_dev *idev;
1715 int err = -EADDRNOTAVAIL;
1717 rcu_read_lock();
1718 idev = __in6_dev_get(dev);
1719 if (idev) {
1720 read_lock_bh(&idev->lock);
1721 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1722 read_unlock_bh(&idev->lock);
1724 rcu_read_unlock();
1725 return err;
1728 static int ipv6_count_addresses(struct inet6_dev *idev)
1730 int cnt = 0;
1731 struct inet6_ifaddr *ifp;
1733 read_lock_bh(&idev->lock);
1734 list_for_each_entry(ifp, &idev->addr_list, if_list)
1735 cnt++;
1736 read_unlock_bh(&idev->lock);
1737 return cnt;
1740 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1741 const struct net_device *dev, int strict)
1743 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1745 EXPORT_SYMBOL(ipv6_chk_addr);
1747 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1748 const struct net_device *dev, int strict,
1749 u32 banned_flags)
1751 struct inet6_ifaddr *ifp;
1752 unsigned int hash = inet6_addr_hash(addr);
1753 u32 ifp_flags;
1755 rcu_read_lock_bh();
1756 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1757 if (!net_eq(dev_net(ifp->idev->dev), net))
1758 continue;
1759 /* Decouple optimistic from tentative for evaluation here.
1760 * Ban optimistic addresses explicitly, when required.
1762 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1763 ? (ifp->flags&~IFA_F_TENTATIVE)
1764 : ifp->flags;
1765 if (ipv6_addr_equal(&ifp->addr, addr) &&
1766 !(ifp_flags&banned_flags) &&
1767 (!dev || ifp->idev->dev == dev ||
1768 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1769 rcu_read_unlock_bh();
1770 return 1;
1774 rcu_read_unlock_bh();
1775 return 0;
1777 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1779 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1780 struct net_device *dev)
1782 unsigned int hash = inet6_addr_hash(addr);
1783 struct inet6_ifaddr *ifp;
1785 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1786 if (!net_eq(dev_net(ifp->idev->dev), net))
1787 continue;
1788 if (ipv6_addr_equal(&ifp->addr, addr)) {
1789 if (!dev || ifp->idev->dev == dev)
1790 return true;
1793 return false;
1796 /* Compares an address/prefix_len with addresses on device @dev.
1797 * If one is found it returns true.
1799 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1800 const unsigned int prefix_len, struct net_device *dev)
1802 struct inet6_dev *idev;
1803 struct inet6_ifaddr *ifa;
1804 bool ret = false;
1806 rcu_read_lock();
1807 idev = __in6_dev_get(dev);
1808 if (idev) {
1809 read_lock_bh(&idev->lock);
1810 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1811 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1812 if (ret)
1813 break;
1815 read_unlock_bh(&idev->lock);
1817 rcu_read_unlock();
1819 return ret;
1821 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1823 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1825 struct inet6_dev *idev;
1826 struct inet6_ifaddr *ifa;
1827 int onlink;
1829 onlink = 0;
1830 rcu_read_lock();
1831 idev = __in6_dev_get(dev);
1832 if (idev) {
1833 read_lock_bh(&idev->lock);
1834 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1835 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1836 ifa->prefix_len);
1837 if (onlink)
1838 break;
1840 read_unlock_bh(&idev->lock);
1842 rcu_read_unlock();
1843 return onlink;
1845 EXPORT_SYMBOL(ipv6_chk_prefix);
1847 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1848 struct net_device *dev, int strict)
1850 struct inet6_ifaddr *ifp, *result = NULL;
1851 unsigned int hash = inet6_addr_hash(addr);
1853 rcu_read_lock_bh();
1854 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1855 if (!net_eq(dev_net(ifp->idev->dev), net))
1856 continue;
1857 if (ipv6_addr_equal(&ifp->addr, addr)) {
1858 if (!dev || ifp->idev->dev == dev ||
1859 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1860 result = ifp;
1861 in6_ifa_hold(ifp);
1862 break;
1866 rcu_read_unlock_bh();
1868 return result;
1871 /* Gets referenced address, destroys ifaddr */
1873 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1875 if (dad_failed)
1876 ifp->flags |= IFA_F_DADFAILED;
1878 if (ifp->flags&IFA_F_PERMANENT) {
1879 spin_lock_bh(&ifp->lock);
1880 addrconf_del_dad_work(ifp);
1881 ifp->flags |= IFA_F_TENTATIVE;
1882 spin_unlock_bh(&ifp->lock);
1883 if (dad_failed)
1884 ipv6_ifa_notify(0, ifp);
1885 in6_ifa_put(ifp);
1886 } else if (ifp->flags&IFA_F_TEMPORARY) {
1887 struct inet6_ifaddr *ifpub;
1888 spin_lock_bh(&ifp->lock);
1889 ifpub = ifp->ifpub;
1890 if (ifpub) {
1891 in6_ifa_hold(ifpub);
1892 spin_unlock_bh(&ifp->lock);
1893 ipv6_create_tempaddr(ifpub, ifp);
1894 in6_ifa_put(ifpub);
1895 } else {
1896 spin_unlock_bh(&ifp->lock);
1898 ipv6_del_addr(ifp);
1899 } else {
1900 ipv6_del_addr(ifp);
1904 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1906 int err = -ENOENT;
1908 spin_lock_bh(&ifp->lock);
1909 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1910 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1911 err = 0;
1913 spin_unlock_bh(&ifp->lock);
1915 return err;
1918 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1920 struct inet6_dev *idev = ifp->idev;
1921 struct net *net = dev_net(ifp->idev->dev);
1923 if (addrconf_dad_end(ifp)) {
1924 in6_ifa_put(ifp);
1925 return;
1928 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1929 ifp->idev->dev->name, &ifp->addr);
1931 spin_lock_bh(&ifp->lock);
1933 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1934 int scope = ifp->scope;
1935 u32 flags = ifp->flags;
1936 struct in6_addr new_addr;
1937 struct inet6_ifaddr *ifp2;
1938 u32 valid_lft, preferred_lft;
1939 int pfxlen = ifp->prefix_len;
1940 int retries = ifp->stable_privacy_retry + 1;
1942 if (retries > net->ipv6.sysctl.idgen_retries) {
1943 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1944 ifp->idev->dev->name);
1945 goto errdad;
1948 new_addr = ifp->addr;
1949 if (ipv6_generate_stable_address(&new_addr, retries,
1950 idev))
1951 goto errdad;
1953 valid_lft = ifp->valid_lft;
1954 preferred_lft = ifp->prefered_lft;
1956 spin_unlock_bh(&ifp->lock);
1958 if (idev->cnf.max_addresses &&
1959 ipv6_count_addresses(idev) >=
1960 idev->cnf.max_addresses)
1961 goto lock_errdad;
1963 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1964 ifp->idev->dev->name);
1966 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1967 scope, flags, valid_lft,
1968 preferred_lft);
1969 if (IS_ERR(ifp2))
1970 goto lock_errdad;
1972 spin_lock_bh(&ifp2->lock);
1973 ifp2->stable_privacy_retry = retries;
1974 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1975 spin_unlock_bh(&ifp2->lock);
1977 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1978 in6_ifa_put(ifp2);
1979 lock_errdad:
1980 spin_lock_bh(&ifp->lock);
1983 errdad:
1984 /* transition from _POSTDAD to _ERRDAD */
1985 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1986 spin_unlock_bh(&ifp->lock);
1988 addrconf_mod_dad_work(ifp, 0);
1989 in6_ifa_put(ifp);
1992 /* Join to solicited addr multicast group.
1993 * caller must hold RTNL */
1994 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1996 struct in6_addr maddr;
1998 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1999 return;
2001 addrconf_addr_solict_mult(addr, &maddr);
2002 ipv6_dev_mc_inc(dev, &maddr);
2005 /* caller must hold RTNL */
2006 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2008 struct in6_addr maddr;
2010 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2011 return;
2013 addrconf_addr_solict_mult(addr, &maddr);
2014 __ipv6_dev_mc_dec(idev, &maddr);
2017 /* caller must hold RTNL */
2018 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2020 struct in6_addr addr;
2022 if (ifp->prefix_len >= 127) /* RFC 6164 */
2023 return;
2024 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2025 if (ipv6_addr_any(&addr))
2026 return;
2027 __ipv6_dev_ac_inc(ifp->idev, &addr);
2030 /* caller must hold RTNL */
2031 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2033 struct in6_addr addr;
2035 if (ifp->prefix_len >= 127) /* RFC 6164 */
2036 return;
2037 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2038 if (ipv6_addr_any(&addr))
2039 return;
2040 __ipv6_dev_ac_dec(ifp->idev, &addr);
2043 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2045 if (dev->addr_len != EUI64_ADDR_LEN)
2046 return -1;
2047 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2048 eui[0] ^= 2;
2049 return 0;
2052 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2054 union fwnet_hwaddr *ha;
2056 if (dev->addr_len != FWNET_ALEN)
2057 return -1;
2059 ha = (union fwnet_hwaddr *)dev->dev_addr;
2061 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2062 eui[0] ^= 2;
2063 return 0;
2066 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2068 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2069 if (dev->addr_len != ARCNET_ALEN)
2070 return -1;
2071 memset(eui, 0, 7);
2072 eui[7] = *(u8 *)dev->dev_addr;
2073 return 0;
2076 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2078 if (dev->addr_len != INFINIBAND_ALEN)
2079 return -1;
2080 memcpy(eui, dev->dev_addr + 12, 8);
2081 eui[0] |= 2;
2082 return 0;
2085 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2087 if (addr == 0)
2088 return -1;
2089 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2090 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2091 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2092 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2093 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2094 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2095 eui[1] = 0;
2096 eui[2] = 0x5E;
2097 eui[3] = 0xFE;
2098 memcpy(eui + 4, &addr, 4);
2099 return 0;
2102 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2104 if (dev->priv_flags & IFF_ISATAP)
2105 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2106 return -1;
2109 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2111 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2114 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2116 memcpy(eui, dev->perm_addr, 3);
2117 memcpy(eui + 5, dev->perm_addr + 3, 3);
2118 eui[3] = 0xFF;
2119 eui[4] = 0xFE;
2120 eui[0] ^= 2;
2121 return 0;
2124 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2126 switch (dev->type) {
2127 case ARPHRD_ETHER:
2128 case ARPHRD_FDDI:
2129 return addrconf_ifid_eui48(eui, dev);
2130 case ARPHRD_ARCNET:
2131 return addrconf_ifid_arcnet(eui, dev);
2132 case ARPHRD_INFINIBAND:
2133 return addrconf_ifid_infiniband(eui, dev);
2134 case ARPHRD_SIT:
2135 return addrconf_ifid_sit(eui, dev);
2136 case ARPHRD_IPGRE:
2137 return addrconf_ifid_gre(eui, dev);
2138 case ARPHRD_6LOWPAN:
2139 return addrconf_ifid_eui64(eui, dev);
2140 case ARPHRD_IEEE1394:
2141 return addrconf_ifid_ieee1394(eui, dev);
2142 case ARPHRD_TUNNEL6:
2143 return addrconf_ifid_ip6tnl(eui, dev);
2145 return -1;
2148 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2150 int err = -1;
2151 struct inet6_ifaddr *ifp;
2153 read_lock_bh(&idev->lock);
2154 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2155 if (ifp->scope > IFA_LINK)
2156 break;
2157 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2158 memcpy(eui, ifp->addr.s6_addr+8, 8);
2159 err = 0;
2160 break;
2163 read_unlock_bh(&idev->lock);
2164 return err;
2167 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2168 static void ipv6_regen_rndid(struct inet6_dev *idev)
2170 regen:
2171 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2172 idev->rndid[0] &= ~0x02;
2175 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2176 * check if generated address is not inappropriate
2178 * - Reserved subnet anycast (RFC 2526)
2179 * 11111101 11....11 1xxxxxxx
2180 * - ISATAP (RFC4214) 6.1
2181 * 00-00-5E-FE-xx-xx-xx-xx
2182 * - value 0
2183 * - XXX: already assigned to an address on the device
2185 if (idev->rndid[0] == 0xfd &&
2186 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2187 (idev->rndid[7]&0x80))
2188 goto regen;
2189 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2190 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2191 goto regen;
2192 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2193 goto regen;
2197 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2199 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2200 ipv6_regen_rndid(idev);
2204 * Add prefix route.
2207 static void
2208 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2209 unsigned long expires, u32 flags)
2211 struct fib6_config cfg = {
2212 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2213 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2214 .fc_ifindex = dev->ifindex,
2215 .fc_expires = expires,
2216 .fc_dst_len = plen,
2217 .fc_flags = RTF_UP | flags,
2218 .fc_nlinfo.nl_net = dev_net(dev),
2219 .fc_protocol = RTPROT_KERNEL,
2222 cfg.fc_dst = *pfx;
2224 /* Prevent useless cloning on PtP SIT.
2225 This thing is done here expecting that the whole
2226 class of non-broadcast devices need not cloning.
2228 #if IS_ENABLED(CONFIG_IPV6_SIT)
2229 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2230 cfg.fc_flags |= RTF_NONEXTHOP;
2231 #endif
2233 ip6_route_add(&cfg);
2237 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2238 int plen,
2239 const struct net_device *dev,
2240 u32 flags, u32 noflags)
2242 struct fib6_node *fn;
2243 struct rt6_info *rt = NULL;
2244 struct fib6_table *table;
2245 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2247 table = fib6_get_table(dev_net(dev), tb_id);
2248 if (!table)
2249 return NULL;
2251 read_lock_bh(&table->tb6_lock);
2252 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2253 if (!fn)
2254 goto out;
2256 noflags |= RTF_CACHE;
2257 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2258 if (rt->dst.dev->ifindex != dev->ifindex)
2259 continue;
2260 if ((rt->rt6i_flags & flags) != flags)
2261 continue;
2262 if ((rt->rt6i_flags & noflags) != 0)
2263 continue;
2264 dst_hold(&rt->dst);
2265 break;
2267 out:
2268 read_unlock_bh(&table->tb6_lock);
2269 return rt;
2273 /* Create "default" multicast route to the interface */
2275 static void addrconf_add_mroute(struct net_device *dev)
2277 struct fib6_config cfg = {
2278 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2279 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2280 .fc_ifindex = dev->ifindex,
2281 .fc_dst_len = 8,
2282 .fc_flags = RTF_UP,
2283 .fc_nlinfo.nl_net = dev_net(dev),
2286 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2288 ip6_route_add(&cfg);
2291 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2293 struct inet6_dev *idev;
2295 ASSERT_RTNL();
2297 idev = ipv6_find_idev(dev);
2298 if (!idev)
2299 return ERR_PTR(-ENOBUFS);
2301 if (idev->cnf.disable_ipv6)
2302 return ERR_PTR(-EACCES);
2304 /* Add default multicast route */
2305 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2306 addrconf_add_mroute(dev);
2308 return idev;
2311 static void manage_tempaddrs(struct inet6_dev *idev,
2312 struct inet6_ifaddr *ifp,
2313 __u32 valid_lft, __u32 prefered_lft,
2314 bool create, unsigned long now)
2316 u32 flags;
2317 struct inet6_ifaddr *ift;
2319 read_lock_bh(&idev->lock);
2320 /* update all temporary addresses in the list */
2321 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2322 int age, max_valid, max_prefered;
2324 if (ifp != ift->ifpub)
2325 continue;
2327 /* RFC 4941 section 3.3:
2328 * If a received option will extend the lifetime of a public
2329 * address, the lifetimes of temporary addresses should
2330 * be extended, subject to the overall constraint that no
2331 * temporary addresses should ever remain "valid" or "preferred"
2332 * for a time longer than (TEMP_VALID_LIFETIME) or
2333 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2335 age = (now - ift->cstamp) / HZ;
2336 max_valid = idev->cnf.temp_valid_lft - age;
2337 if (max_valid < 0)
2338 max_valid = 0;
2340 max_prefered = idev->cnf.temp_prefered_lft -
2341 idev->desync_factor - age;
2342 if (max_prefered < 0)
2343 max_prefered = 0;
2345 if (valid_lft > max_valid)
2346 valid_lft = max_valid;
2348 if (prefered_lft > max_prefered)
2349 prefered_lft = max_prefered;
2351 spin_lock(&ift->lock);
2352 flags = ift->flags;
2353 ift->valid_lft = valid_lft;
2354 ift->prefered_lft = prefered_lft;
2355 ift->tstamp = now;
2356 if (prefered_lft > 0)
2357 ift->flags &= ~IFA_F_DEPRECATED;
2359 spin_unlock(&ift->lock);
2360 if (!(flags&IFA_F_TENTATIVE))
2361 ipv6_ifa_notify(0, ift);
2364 if ((create || list_empty(&idev->tempaddr_list)) &&
2365 idev->cnf.use_tempaddr > 0) {
2366 /* When a new public address is created as described
2367 * in [ADDRCONF], also create a new temporary address.
2368 * Also create a temporary address if it's enabled but
2369 * no temporary address currently exists.
2371 read_unlock_bh(&idev->lock);
2372 ipv6_create_tempaddr(ifp, NULL);
2373 } else {
2374 read_unlock_bh(&idev->lock);
2378 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2380 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2381 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2384 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2385 const struct prefix_info *pinfo,
2386 struct inet6_dev *in6_dev,
2387 const struct in6_addr *addr, int addr_type,
2388 u32 addr_flags, bool sllao, bool tokenized,
2389 __u32 valid_lft, u32 prefered_lft)
2391 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2392 int create = 0, update_lft = 0;
2394 if (!ifp && valid_lft) {
2395 int max_addresses = in6_dev->cnf.max_addresses;
2397 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2398 if (in6_dev->cnf.optimistic_dad &&
2399 !net->ipv6.devconf_all->forwarding && sllao)
2400 addr_flags |= IFA_F_OPTIMISTIC;
2401 #endif
2403 /* Do not allow to create too much of autoconfigured
2404 * addresses; this would be too easy way to crash kernel.
2406 if (!max_addresses ||
2407 ipv6_count_addresses(in6_dev) < max_addresses)
2408 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2409 pinfo->prefix_len,
2410 addr_type&IPV6_ADDR_SCOPE_MASK,
2411 addr_flags, valid_lft,
2412 prefered_lft);
2414 if (IS_ERR_OR_NULL(ifp))
2415 return -1;
2417 update_lft = 0;
2418 create = 1;
2419 spin_lock_bh(&ifp->lock);
2420 ifp->flags |= IFA_F_MANAGETEMPADDR;
2421 ifp->cstamp = jiffies;
2422 ifp->tokenized = tokenized;
2423 spin_unlock_bh(&ifp->lock);
2424 addrconf_dad_start(ifp);
2427 if (ifp) {
2428 u32 flags;
2429 unsigned long now;
2430 u32 stored_lft;
2432 /* update lifetime (RFC2462 5.5.3 e) */
2433 spin_lock_bh(&ifp->lock);
2434 now = jiffies;
2435 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2436 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2437 else
2438 stored_lft = 0;
2439 if (!update_lft && !create && stored_lft) {
2440 const u32 minimum_lft = min_t(u32,
2441 stored_lft, MIN_VALID_LIFETIME);
2442 valid_lft = max(valid_lft, minimum_lft);
2444 /* RFC4862 Section 5.5.3e:
2445 * "Note that the preferred lifetime of the
2446 * corresponding address is always reset to
2447 * the Preferred Lifetime in the received
2448 * Prefix Information option, regardless of
2449 * whether the valid lifetime is also reset or
2450 * ignored."
2452 * So we should always update prefered_lft here.
2454 update_lft = 1;
2457 if (update_lft) {
2458 ifp->valid_lft = valid_lft;
2459 ifp->prefered_lft = prefered_lft;
2460 ifp->tstamp = now;
2461 flags = ifp->flags;
2462 ifp->flags &= ~IFA_F_DEPRECATED;
2463 spin_unlock_bh(&ifp->lock);
2465 if (!(flags&IFA_F_TENTATIVE))
2466 ipv6_ifa_notify(0, ifp);
2467 } else
2468 spin_unlock_bh(&ifp->lock);
2470 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2471 create, now);
2473 in6_ifa_put(ifp);
2474 addrconf_verify();
2477 return 0;
2479 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2481 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2483 struct prefix_info *pinfo;
2484 __u32 valid_lft;
2485 __u32 prefered_lft;
2486 int addr_type, err;
2487 u32 addr_flags = 0;
2488 struct inet6_dev *in6_dev;
2489 struct net *net = dev_net(dev);
2491 pinfo = (struct prefix_info *) opt;
2493 if (len < sizeof(struct prefix_info)) {
2494 ADBG("addrconf: prefix option too short\n");
2495 return;
2499 * Validation checks ([ADDRCONF], page 19)
2502 addr_type = ipv6_addr_type(&pinfo->prefix);
2504 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2505 return;
2507 valid_lft = ntohl(pinfo->valid);
2508 prefered_lft = ntohl(pinfo->prefered);
2510 if (prefered_lft > valid_lft) {
2511 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2512 return;
2515 in6_dev = in6_dev_get(dev);
2517 if (!in6_dev) {
2518 net_dbg_ratelimited("addrconf: device %s not configured\n",
2519 dev->name);
2520 return;
2524 * Two things going on here:
2525 * 1) Add routes for on-link prefixes
2526 * 2) Configure prefixes with the auto flag set
2529 if (pinfo->onlink) {
2530 struct rt6_info *rt;
2531 unsigned long rt_expires;
2533 /* Avoid arithmetic overflow. Really, we could
2534 * save rt_expires in seconds, likely valid_lft,
2535 * but it would require division in fib gc, that it
2536 * not good.
2538 if (HZ > USER_HZ)
2539 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2540 else
2541 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2543 if (addrconf_finite_timeout(rt_expires))
2544 rt_expires *= HZ;
2546 rt = addrconf_get_prefix_route(&pinfo->prefix,
2547 pinfo->prefix_len,
2548 dev,
2549 RTF_ADDRCONF | RTF_PREFIX_RT,
2550 RTF_GATEWAY | RTF_DEFAULT);
2552 if (rt) {
2553 /* Autoconf prefix route */
2554 if (valid_lft == 0) {
2555 ip6_del_rt(rt);
2556 rt = NULL;
2557 } else if (addrconf_finite_timeout(rt_expires)) {
2558 /* not infinity */
2559 rt6_set_expires(rt, jiffies + rt_expires);
2560 } else {
2561 rt6_clean_expires(rt);
2563 } else if (valid_lft) {
2564 clock_t expires = 0;
2565 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2566 if (addrconf_finite_timeout(rt_expires)) {
2567 /* not infinity */
2568 flags |= RTF_EXPIRES;
2569 expires = jiffies_to_clock_t(rt_expires);
2571 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2572 dev, expires, flags);
2574 ip6_rt_put(rt);
2577 /* Try to figure out our local address for this prefix */
2579 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2580 struct in6_addr addr;
2581 bool tokenized = false, dev_addr_generated = false;
2583 if (pinfo->prefix_len == 64) {
2584 memcpy(&addr, &pinfo->prefix, 8);
2586 if (!ipv6_addr_any(&in6_dev->token)) {
2587 read_lock_bh(&in6_dev->lock);
2588 memcpy(addr.s6_addr + 8,
2589 in6_dev->token.s6_addr + 8, 8);
2590 read_unlock_bh(&in6_dev->lock);
2591 tokenized = true;
2592 } else if (is_addr_mode_generate_stable(in6_dev) &&
2593 !ipv6_generate_stable_address(&addr, 0,
2594 in6_dev)) {
2595 addr_flags |= IFA_F_STABLE_PRIVACY;
2596 goto ok;
2597 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2598 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2599 goto put;
2600 } else {
2601 dev_addr_generated = true;
2603 goto ok;
2605 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2606 pinfo->prefix_len);
2607 goto put;
2610 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2611 &addr, addr_type,
2612 addr_flags, sllao,
2613 tokenized, valid_lft,
2614 prefered_lft);
2615 if (err)
2616 goto put;
2618 /* Ignore error case here because previous prefix add addr was
2619 * successful which will be notified.
2621 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2622 addr_type, addr_flags, sllao,
2623 tokenized, valid_lft,
2624 prefered_lft,
2625 dev_addr_generated);
2627 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2628 put:
2629 in6_dev_put(in6_dev);
2633 * Set destination address.
2634 * Special case for SIT interfaces where we create a new "virtual"
2635 * device.
2637 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2639 struct in6_ifreq ireq;
2640 struct net_device *dev;
2641 int err = -EINVAL;
2643 rtnl_lock();
2645 err = -EFAULT;
2646 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2647 goto err_exit;
2649 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2651 err = -ENODEV;
2652 if (!dev)
2653 goto err_exit;
2655 #if IS_ENABLED(CONFIG_IPV6_SIT)
2656 if (dev->type == ARPHRD_SIT) {
2657 const struct net_device_ops *ops = dev->netdev_ops;
2658 struct ifreq ifr;
2659 struct ip_tunnel_parm p;
2661 err = -EADDRNOTAVAIL;
2662 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2663 goto err_exit;
2665 memset(&p, 0, sizeof(p));
2666 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2667 p.iph.saddr = 0;
2668 p.iph.version = 4;
2669 p.iph.ihl = 5;
2670 p.iph.protocol = IPPROTO_IPV6;
2671 p.iph.ttl = 64;
2672 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2674 if (ops->ndo_do_ioctl) {
2675 mm_segment_t oldfs = get_fs();
2677 set_fs(KERNEL_DS);
2678 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2679 set_fs(oldfs);
2680 } else
2681 err = -EOPNOTSUPP;
2683 if (err == 0) {
2684 err = -ENOBUFS;
2685 dev = __dev_get_by_name(net, p.name);
2686 if (!dev)
2687 goto err_exit;
2688 err = dev_open(dev);
2691 #endif
2693 err_exit:
2694 rtnl_unlock();
2695 return err;
2698 static int ipv6_mc_config(struct sock *sk, bool join,
2699 const struct in6_addr *addr, int ifindex)
2701 int ret;
2703 ASSERT_RTNL();
2705 lock_sock(sk);
2706 if (join)
2707 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2708 else
2709 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2710 release_sock(sk);
2712 return ret;
2716 * Manual configuration of address on an interface
2718 static int inet6_addr_add(struct net *net, int ifindex,
2719 const struct in6_addr *pfx,
2720 const struct in6_addr *peer_pfx,
2721 unsigned int plen, __u32 ifa_flags,
2722 __u32 prefered_lft, __u32 valid_lft)
2724 struct inet6_ifaddr *ifp;
2725 struct inet6_dev *idev;
2726 struct net_device *dev;
2727 unsigned long timeout;
2728 clock_t expires;
2729 int scope;
2730 u32 flags;
2732 ASSERT_RTNL();
2734 if (plen > 128)
2735 return -EINVAL;
2737 /* check the lifetime */
2738 if (!valid_lft || prefered_lft > valid_lft)
2739 return -EINVAL;
2741 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2742 return -EINVAL;
2744 dev = __dev_get_by_index(net, ifindex);
2745 if (!dev)
2746 return -ENODEV;
2748 idev = addrconf_add_dev(dev);
2749 if (IS_ERR(idev))
2750 return PTR_ERR(idev);
2752 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2753 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2754 true, pfx, ifindex);
2756 if (ret < 0)
2757 return ret;
2760 scope = ipv6_addr_scope(pfx);
2762 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2763 if (addrconf_finite_timeout(timeout)) {
2764 expires = jiffies_to_clock_t(timeout * HZ);
2765 valid_lft = timeout;
2766 flags = RTF_EXPIRES;
2767 } else {
2768 expires = 0;
2769 flags = 0;
2770 ifa_flags |= IFA_F_PERMANENT;
2773 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2774 if (addrconf_finite_timeout(timeout)) {
2775 if (timeout == 0)
2776 ifa_flags |= IFA_F_DEPRECATED;
2777 prefered_lft = timeout;
2780 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2781 valid_lft, prefered_lft);
2783 if (!IS_ERR(ifp)) {
2784 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2785 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2786 expires, flags);
2790 * Note that section 3.1 of RFC 4429 indicates
2791 * that the Optimistic flag should not be set for
2792 * manually configured addresses
2794 addrconf_dad_start(ifp);
2795 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2796 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2797 true, jiffies);
2798 in6_ifa_put(ifp);
2799 addrconf_verify_rtnl();
2800 return 0;
2801 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2802 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2803 false, pfx, ifindex);
2806 return PTR_ERR(ifp);
2809 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2810 const struct in6_addr *pfx, unsigned int plen)
2812 struct inet6_ifaddr *ifp;
2813 struct inet6_dev *idev;
2814 struct net_device *dev;
2816 if (plen > 128)
2817 return -EINVAL;
2819 dev = __dev_get_by_index(net, ifindex);
2820 if (!dev)
2821 return -ENODEV;
2823 idev = __in6_dev_get(dev);
2824 if (!idev)
2825 return -ENXIO;
2827 read_lock_bh(&idev->lock);
2828 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2829 if (ifp->prefix_len == plen &&
2830 ipv6_addr_equal(pfx, &ifp->addr)) {
2831 in6_ifa_hold(ifp);
2832 read_unlock_bh(&idev->lock);
2834 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2835 (ifa_flags & IFA_F_MANAGETEMPADDR))
2836 manage_tempaddrs(idev, ifp, 0, 0, false,
2837 jiffies);
2838 ipv6_del_addr(ifp);
2839 addrconf_verify_rtnl();
2840 if (ipv6_addr_is_multicast(pfx)) {
2841 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2842 false, pfx, dev->ifindex);
2844 return 0;
2847 read_unlock_bh(&idev->lock);
2848 return -EADDRNOTAVAIL;
2852 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2854 struct in6_ifreq ireq;
2855 int err;
2857 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2858 return -EPERM;
2860 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2861 return -EFAULT;
2863 rtnl_lock();
2864 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2865 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2866 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2867 rtnl_unlock();
2868 return err;
2871 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2873 struct in6_ifreq ireq;
2874 int err;
2876 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2877 return -EPERM;
2879 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2880 return -EFAULT;
2882 rtnl_lock();
2883 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2884 ireq.ifr6_prefixlen);
2885 rtnl_unlock();
2886 return err;
2889 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2890 int plen, int scope)
2892 struct inet6_ifaddr *ifp;
2894 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2895 scope, IFA_F_PERMANENT,
2896 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2897 if (!IS_ERR(ifp)) {
2898 spin_lock_bh(&ifp->lock);
2899 ifp->flags &= ~IFA_F_TENTATIVE;
2900 spin_unlock_bh(&ifp->lock);
2901 rt_genid_bump_ipv6(dev_net(idev->dev));
2902 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2903 in6_ifa_put(ifp);
2907 #if IS_ENABLED(CONFIG_IPV6_SIT)
2908 static void sit_add_v4_addrs(struct inet6_dev *idev)
2910 struct in6_addr addr;
2911 struct net_device *dev;
2912 struct net *net = dev_net(idev->dev);
2913 int scope, plen;
2914 u32 pflags = 0;
2916 ASSERT_RTNL();
2918 memset(&addr, 0, sizeof(struct in6_addr));
2919 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2921 if (idev->dev->flags&IFF_POINTOPOINT) {
2922 addr.s6_addr32[0] = htonl(0xfe800000);
2923 scope = IFA_LINK;
2924 plen = 64;
2925 } else {
2926 scope = IPV6_ADDR_COMPATv4;
2927 plen = 96;
2928 pflags |= RTF_NONEXTHOP;
2931 if (addr.s6_addr32[3]) {
2932 add_addr(idev, &addr, plen, scope);
2933 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2934 return;
2937 for_each_netdev(net, dev) {
2938 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2939 if (in_dev && (dev->flags & IFF_UP)) {
2940 struct in_ifaddr *ifa;
2942 int flag = scope;
2944 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2946 addr.s6_addr32[3] = ifa->ifa_local;
2948 if (ifa->ifa_scope == RT_SCOPE_LINK)
2949 continue;
2950 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2951 if (idev->dev->flags&IFF_POINTOPOINT)
2952 continue;
2953 flag |= IFA_HOST;
2956 add_addr(idev, &addr, plen, flag);
2957 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2958 pflags);
2963 #endif
2965 static void init_loopback(struct net_device *dev)
2967 struct inet6_dev *idev;
2968 struct net_device *sp_dev;
2969 struct inet6_ifaddr *sp_ifa;
2970 struct rt6_info *sp_rt;
2972 /* ::1 */
2974 ASSERT_RTNL();
2976 idev = ipv6_find_idev(dev);
2977 if (!idev) {
2978 pr_debug("%s: add_dev failed\n", __func__);
2979 return;
2982 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2984 /* Add routes to other interface's IPv6 addresses */
2985 for_each_netdev(dev_net(dev), sp_dev) {
2986 if (!strcmp(sp_dev->name, dev->name))
2987 continue;
2989 idev = __in6_dev_get(sp_dev);
2990 if (!idev)
2991 continue;
2993 read_lock_bh(&idev->lock);
2994 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2996 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2997 continue;
2999 if (sp_ifa->rt) {
3000 /* This dst has been added to garbage list when
3001 * lo device down, release this obsolete dst and
3002 * reallocate a new router for ifa.
3004 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3005 ip6_rt_put(sp_ifa->rt);
3006 sp_ifa->rt = NULL;
3007 } else {
3008 continue;
3012 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3014 /* Failure cases are ignored */
3015 if (!IS_ERR(sp_rt)) {
3016 sp_ifa->rt = sp_rt;
3017 ip6_ins_rt(sp_rt);
3020 read_unlock_bh(&idev->lock);
3024 void addrconf_add_linklocal(struct inet6_dev *idev,
3025 const struct in6_addr *addr, u32 flags)
3027 struct inet6_ifaddr *ifp;
3028 u32 addr_flags = flags | IFA_F_PERMANENT;
3030 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3031 if (idev->cnf.optimistic_dad &&
3032 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3033 addr_flags |= IFA_F_OPTIMISTIC;
3034 #endif
3036 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3037 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3038 if (!IS_ERR(ifp)) {
3039 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3040 addrconf_dad_start(ifp);
3041 in6_ifa_put(ifp);
3044 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3046 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3048 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3049 return true;
3051 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3052 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3053 return true;
3055 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3056 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3057 return true;
3059 return false;
3062 static int ipv6_generate_stable_address(struct in6_addr *address,
3063 u8 dad_count,
3064 const struct inet6_dev *idev)
3066 static DEFINE_SPINLOCK(lock);
3067 static __u32 digest[SHA_DIGEST_WORDS];
3068 static __u32 workspace[SHA_WORKSPACE_WORDS];
3070 static union {
3071 char __data[SHA_MESSAGE_BYTES];
3072 struct {
3073 struct in6_addr secret;
3074 __be32 prefix[2];
3075 unsigned char hwaddr[MAX_ADDR_LEN];
3076 u8 dad_count;
3077 } __packed;
3078 } data;
3080 struct in6_addr secret;
3081 struct in6_addr temp;
3082 struct net *net = dev_net(idev->dev);
3084 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3086 if (idev->cnf.stable_secret.initialized)
3087 secret = idev->cnf.stable_secret.secret;
3088 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3089 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3090 else
3091 return -1;
3093 retry:
3094 spin_lock_bh(&lock);
3096 sha_init(digest);
3097 memset(&data, 0, sizeof(data));
3098 memset(workspace, 0, sizeof(workspace));
3099 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3100 data.prefix[0] = address->s6_addr32[0];
3101 data.prefix[1] = address->s6_addr32[1];
3102 data.secret = secret;
3103 data.dad_count = dad_count;
3105 sha_transform(digest, data.__data, workspace);
3107 temp = *address;
3108 temp.s6_addr32[2] = (__force __be32)digest[0];
3109 temp.s6_addr32[3] = (__force __be32)digest[1];
3111 spin_unlock_bh(&lock);
3113 if (ipv6_reserved_interfaceid(temp)) {
3114 dad_count++;
3115 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3116 return -1;
3117 goto retry;
3120 *address = temp;
3121 return 0;
3124 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3126 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3128 if (s->initialized)
3129 return;
3130 s = &idev->cnf.stable_secret;
3131 get_random_bytes(&s->secret, sizeof(s->secret));
3132 s->initialized = true;
3135 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3137 struct in6_addr addr;
3139 /* no link local addresses on L3 master devices */
3140 if (netif_is_l3_master(idev->dev))
3141 return;
3143 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3145 switch (idev->addr_gen_mode) {
3146 case IN6_ADDR_GEN_MODE_RANDOM:
3147 ipv6_gen_mode_random_init(idev);
3148 /* fallthrough */
3149 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3150 if (!ipv6_generate_stable_address(&addr, 0, idev))
3151 addrconf_add_linklocal(idev, &addr,
3152 IFA_F_STABLE_PRIVACY);
3153 else if (prefix_route)
3154 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3155 break;
3156 case IN6_ADDR_GEN_MODE_EUI64:
3157 /* addrconf_add_linklocal also adds a prefix_route and we
3158 * only need to care about prefix routes if ipv6_generate_eui64
3159 * couldn't generate one.
3161 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3162 addrconf_add_linklocal(idev, &addr, 0);
3163 else if (prefix_route)
3164 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3165 break;
3166 case IN6_ADDR_GEN_MODE_NONE:
3167 default:
3168 /* will not add any link local address */
3169 break;
3173 static void addrconf_dev_config(struct net_device *dev)
3175 struct inet6_dev *idev;
3177 ASSERT_RTNL();
3179 if ((dev->type != ARPHRD_ETHER) &&
3180 (dev->type != ARPHRD_FDDI) &&
3181 (dev->type != ARPHRD_ARCNET) &&
3182 (dev->type != ARPHRD_INFINIBAND) &&
3183 (dev->type != ARPHRD_IEEE1394) &&
3184 (dev->type != ARPHRD_TUNNEL6) &&
3185 (dev->type != ARPHRD_6LOWPAN) &&
3186 (dev->type != ARPHRD_NONE)) {
3187 /* Alas, we support only Ethernet autoconfiguration. */
3188 return;
3191 idev = addrconf_add_dev(dev);
3192 if (IS_ERR(idev))
3193 return;
3195 /* this device type has no EUI support */
3196 if (dev->type == ARPHRD_NONE &&
3197 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3198 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3200 addrconf_addr_gen(idev, false);
3203 #if IS_ENABLED(CONFIG_IPV6_SIT)
3204 static void addrconf_sit_config(struct net_device *dev)
3206 struct inet6_dev *idev;
3208 ASSERT_RTNL();
3211 * Configure the tunnel with one of our IPv4
3212 * addresses... we should configure all of
3213 * our v4 addrs in the tunnel
3216 idev = ipv6_find_idev(dev);
3217 if (!idev) {
3218 pr_debug("%s: add_dev failed\n", __func__);
3219 return;
3222 if (dev->priv_flags & IFF_ISATAP) {
3223 addrconf_addr_gen(idev, false);
3224 return;
3227 sit_add_v4_addrs(idev);
3229 if (dev->flags&IFF_POINTOPOINT)
3230 addrconf_add_mroute(dev);
3232 #endif
3234 #if IS_ENABLED(CONFIG_NET_IPGRE)
3235 static void addrconf_gre_config(struct net_device *dev)
3237 struct inet6_dev *idev;
3239 ASSERT_RTNL();
3241 idev = ipv6_find_idev(dev);
3242 if (!idev) {
3243 pr_debug("%s: add_dev failed\n", __func__);
3244 return;
3247 addrconf_addr_gen(idev, true);
3248 if (dev->flags & IFF_POINTOPOINT)
3249 addrconf_add_mroute(dev);
3251 #endif
3253 static int fixup_permanent_addr(struct inet6_dev *idev,
3254 struct inet6_ifaddr *ifp)
3256 if (!ifp->rt) {
3257 struct rt6_info *rt;
3259 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3260 if (unlikely(IS_ERR(rt)))
3261 return PTR_ERR(rt);
3263 ifp->rt = rt;
3266 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3267 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3268 idev->dev, 0, 0);
3271 addrconf_dad_start(ifp);
3273 return 0;
3276 static void addrconf_permanent_addr(struct net_device *dev)
3278 struct inet6_ifaddr *ifp, *tmp;
3279 struct inet6_dev *idev;
3281 idev = __in6_dev_get(dev);
3282 if (!idev)
3283 return;
3285 write_lock_bh(&idev->lock);
3287 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3288 if ((ifp->flags & IFA_F_PERMANENT) &&
3289 fixup_permanent_addr(idev, ifp) < 0) {
3290 write_unlock_bh(&idev->lock);
3291 ipv6_del_addr(ifp);
3292 write_lock_bh(&idev->lock);
3294 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3295 idev->dev->name, &ifp->addr);
3299 write_unlock_bh(&idev->lock);
3302 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3303 void *ptr)
3305 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3306 struct netdev_notifier_changeupper_info *info;
3307 struct inet6_dev *idev = __in6_dev_get(dev);
3308 int run_pending = 0;
3309 int err;
3311 switch (event) {
3312 case NETDEV_REGISTER:
3313 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3314 idev = ipv6_add_dev(dev);
3315 if (IS_ERR(idev))
3316 return notifier_from_errno(PTR_ERR(idev));
3318 break;
3320 case NETDEV_CHANGEMTU:
3321 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3322 if (dev->mtu < IPV6_MIN_MTU) {
3323 addrconf_ifdown(dev, 1);
3324 break;
3327 if (idev) {
3328 rt6_mtu_change(dev, dev->mtu);
3329 idev->cnf.mtu6 = dev->mtu;
3330 break;
3333 /* allocate new idev */
3334 idev = ipv6_add_dev(dev);
3335 if (IS_ERR(idev))
3336 break;
3338 /* device is still not ready */
3339 if (!(idev->if_flags & IF_READY))
3340 break;
3342 run_pending = 1;
3344 /* fall through */
3346 case NETDEV_UP:
3347 case NETDEV_CHANGE:
3348 if (dev->flags & IFF_SLAVE)
3349 break;
3351 if (idev && idev->cnf.disable_ipv6)
3352 break;
3354 if (event == NETDEV_UP) {
3355 /* restore routes for permanent addresses */
3356 addrconf_permanent_addr(dev);
3358 if (!addrconf_qdisc_ok(dev)) {
3359 /* device is not ready yet. */
3360 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3361 dev->name);
3362 break;
3365 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3366 idev = ipv6_add_dev(dev);
3368 if (!IS_ERR_OR_NULL(idev)) {
3369 idev->if_flags |= IF_READY;
3370 run_pending = 1;
3372 } else if (event == NETDEV_CHANGE) {
3373 if (!addrconf_qdisc_ok(dev)) {
3374 /* device is still not ready. */
3375 break;
3378 if (idev) {
3379 if (idev->if_flags & IF_READY)
3380 /* device is already configured. */
3381 break;
3382 idev->if_flags |= IF_READY;
3385 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3386 dev->name);
3388 run_pending = 1;
3391 switch (dev->type) {
3392 #if IS_ENABLED(CONFIG_IPV6_SIT)
3393 case ARPHRD_SIT:
3394 addrconf_sit_config(dev);
3395 break;
3396 #endif
3397 #if IS_ENABLED(CONFIG_NET_IPGRE)
3398 case ARPHRD_IPGRE:
3399 addrconf_gre_config(dev);
3400 break;
3401 #endif
3402 case ARPHRD_LOOPBACK:
3403 init_loopback(dev);
3404 break;
3406 default:
3407 addrconf_dev_config(dev);
3408 break;
3411 if (!IS_ERR_OR_NULL(idev)) {
3412 if (run_pending)
3413 addrconf_dad_run(idev);
3416 * If the MTU changed during the interface down,
3417 * when the interface up, the changed MTU must be
3418 * reflected in the idev as well as routers.
3420 if (idev->cnf.mtu6 != dev->mtu &&
3421 dev->mtu >= IPV6_MIN_MTU) {
3422 rt6_mtu_change(dev, dev->mtu);
3423 idev->cnf.mtu6 = dev->mtu;
3425 idev->tstamp = jiffies;
3426 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3429 * If the changed mtu during down is lower than
3430 * IPV6_MIN_MTU stop IPv6 on this interface.
3432 if (dev->mtu < IPV6_MIN_MTU)
3433 addrconf_ifdown(dev, 1);
3435 break;
3437 case NETDEV_DOWN:
3438 case NETDEV_UNREGISTER:
3440 * Remove all addresses from this interface.
3442 addrconf_ifdown(dev, event != NETDEV_DOWN);
3443 break;
3445 case NETDEV_CHANGENAME:
3446 if (idev) {
3447 snmp6_unregister_dev(idev);
3448 addrconf_sysctl_unregister(idev);
3449 err = addrconf_sysctl_register(idev);
3450 if (err)
3451 return notifier_from_errno(err);
3452 err = snmp6_register_dev(idev);
3453 if (err) {
3454 addrconf_sysctl_unregister(idev);
3455 return notifier_from_errno(err);
3458 break;
3460 case NETDEV_PRE_TYPE_CHANGE:
3461 case NETDEV_POST_TYPE_CHANGE:
3462 if (idev)
3463 addrconf_type_change(dev, event);
3464 break;
3466 case NETDEV_CHANGEUPPER:
3467 info = ptr;
3469 /* flush all routes if dev is linked to or unlinked from
3470 * an L3 master device (e.g., VRF)
3472 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3473 addrconf_ifdown(dev, 0);
3476 return NOTIFY_OK;
3480 * addrconf module should be notified of a device going up
3482 static struct notifier_block ipv6_dev_notf = {
3483 .notifier_call = addrconf_notify,
3486 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3488 struct inet6_dev *idev;
3489 ASSERT_RTNL();
3491 idev = __in6_dev_get(dev);
3493 if (event == NETDEV_POST_TYPE_CHANGE)
3494 ipv6_mc_remap(idev);
3495 else if (event == NETDEV_PRE_TYPE_CHANGE)
3496 ipv6_mc_unmap(idev);
3499 static bool addr_is_local(const struct in6_addr *addr)
3501 return ipv6_addr_type(addr) &
3502 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3505 static int addrconf_ifdown(struct net_device *dev, int how)
3507 struct net *net = dev_net(dev);
3508 struct inet6_dev *idev;
3509 struct inet6_ifaddr *ifa, *tmp;
3510 struct list_head del_list;
3511 int _keep_addr;
3512 bool keep_addr;
3513 int state, i;
3515 ASSERT_RTNL();
3517 rt6_ifdown(net, dev);
3518 neigh_ifdown(&nd_tbl, dev);
3520 idev = __in6_dev_get(dev);
3521 if (!idev)
3522 return -ENODEV;
3525 * Step 1: remove reference to ipv6 device from parent device.
3526 * Do not dev_put!
3528 if (how) {
3529 idev->dead = 1;
3531 /* protected by rtnl_lock */
3532 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3534 /* Step 1.5: remove snmp6 entry */
3535 snmp6_unregister_dev(idev);
3539 /* aggregate the system setting and interface setting */
3540 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3541 if (!_keep_addr)
3542 _keep_addr = idev->cnf.keep_addr_on_down;
3544 /* combine the user config with event to determine if permanent
3545 * addresses are to be removed from address hash table
3547 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3549 /* Step 2: clear hash table */
3550 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3551 struct hlist_head *h = &inet6_addr_lst[i];
3553 spin_lock_bh(&addrconf_hash_lock);
3554 restart:
3555 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3556 if (ifa->idev == idev) {
3557 addrconf_del_dad_work(ifa);
3558 /* combined flag + permanent flag decide if
3559 * address is retained on a down event
3561 if (!keep_addr ||
3562 !(ifa->flags & IFA_F_PERMANENT) ||
3563 addr_is_local(&ifa->addr)) {
3564 hlist_del_init_rcu(&ifa->addr_lst);
3565 goto restart;
3569 spin_unlock_bh(&addrconf_hash_lock);
3572 write_lock_bh(&idev->lock);
3574 addrconf_del_rs_timer(idev);
3576 /* Step 2: clear flags for stateless addrconf */
3577 if (!how)
3578 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3580 /* Step 3: clear tempaddr list */
3581 while (!list_empty(&idev->tempaddr_list)) {
3582 ifa = list_first_entry(&idev->tempaddr_list,
3583 struct inet6_ifaddr, tmp_list);
3584 list_del(&ifa->tmp_list);
3585 write_unlock_bh(&idev->lock);
3586 spin_lock_bh(&ifa->lock);
3588 if (ifa->ifpub) {
3589 in6_ifa_put(ifa->ifpub);
3590 ifa->ifpub = NULL;
3592 spin_unlock_bh(&ifa->lock);
3593 in6_ifa_put(ifa);
3594 write_lock_bh(&idev->lock);
3597 /* re-combine the user config with event to determine if permanent
3598 * addresses are to be removed from the interface list
3600 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3602 INIT_LIST_HEAD(&del_list);
3603 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3604 struct rt6_info *rt = NULL;
3606 addrconf_del_dad_work(ifa);
3608 write_unlock_bh(&idev->lock);
3609 spin_lock_bh(&ifa->lock);
3611 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3612 !addr_is_local(&ifa->addr)) {
3613 /* set state to skip the notifier below */
3614 state = INET6_IFADDR_STATE_DEAD;
3615 ifa->state = 0;
3616 if (!(ifa->flags & IFA_F_NODAD))
3617 ifa->flags |= IFA_F_TENTATIVE;
3619 rt = ifa->rt;
3620 ifa->rt = NULL;
3621 } else {
3622 state = ifa->state;
3623 ifa->state = INET6_IFADDR_STATE_DEAD;
3625 list_move(&ifa->if_list, &del_list);
3628 spin_unlock_bh(&ifa->lock);
3630 if (rt)
3631 ip6_del_rt(rt);
3633 if (state != INET6_IFADDR_STATE_DEAD) {
3634 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3635 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3636 } else {
3637 if (idev->cnf.forwarding)
3638 addrconf_leave_anycast(ifa);
3639 addrconf_leave_solict(ifa->idev, &ifa->addr);
3642 write_lock_bh(&idev->lock);
3645 write_unlock_bh(&idev->lock);
3647 /* now clean up addresses to be removed */
3648 while (!list_empty(&del_list)) {
3649 ifa = list_first_entry(&del_list,
3650 struct inet6_ifaddr, if_list);
3651 list_del(&ifa->if_list);
3653 in6_ifa_put(ifa);
3656 /* Step 5: Discard anycast and multicast list */
3657 if (how) {
3658 ipv6_ac_destroy_dev(idev);
3659 ipv6_mc_destroy_dev(idev);
3660 } else {
3661 ipv6_mc_down(idev);
3664 idev->tstamp = jiffies;
3666 /* Last: Shot the device (if unregistered) */
3667 if (how) {
3668 addrconf_sysctl_unregister(idev);
3669 neigh_parms_release(&nd_tbl, idev->nd_parms);
3670 neigh_ifdown(&nd_tbl, dev);
3671 in6_dev_put(idev);
3673 return 0;
3676 static void addrconf_rs_timer(unsigned long data)
3678 struct inet6_dev *idev = (struct inet6_dev *)data;
3679 struct net_device *dev = idev->dev;
3680 struct in6_addr lladdr;
3682 write_lock(&idev->lock);
3683 if (idev->dead || !(idev->if_flags & IF_READY))
3684 goto out;
3686 if (!ipv6_accept_ra(idev))
3687 goto out;
3689 /* Announcement received after solicitation was sent */
3690 if (idev->if_flags & IF_RA_RCVD)
3691 goto out;
3693 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3694 write_unlock(&idev->lock);
3695 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3696 ndisc_send_rs(dev, &lladdr,
3697 &in6addr_linklocal_allrouters);
3698 else
3699 goto put;
3701 write_lock(&idev->lock);
3702 idev->rs_interval = rfc3315_s14_backoff_update(
3703 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3704 /* The wait after the last probe can be shorter */
3705 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3706 idev->cnf.rtr_solicits) ?
3707 idev->cnf.rtr_solicit_delay :
3708 idev->rs_interval);
3709 } else {
3711 * Note: we do not support deprecated "all on-link"
3712 * assumption any longer.
3714 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3717 out:
3718 write_unlock(&idev->lock);
3719 put:
3720 in6_dev_put(idev);
3724 * Duplicate Address Detection
3726 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3728 unsigned long rand_num;
3729 struct inet6_dev *idev = ifp->idev;
3731 if (ifp->flags & IFA_F_OPTIMISTIC)
3732 rand_num = 0;
3733 else
3734 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3736 ifp->dad_probes = idev->cnf.dad_transmits;
3737 addrconf_mod_dad_work(ifp, rand_num);
3740 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3742 struct inet6_dev *idev = ifp->idev;
3743 struct net_device *dev = idev->dev;
3744 bool bump_id, notify = false;
3746 addrconf_join_solict(dev, &ifp->addr);
3748 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3750 read_lock_bh(&idev->lock);
3751 spin_lock(&ifp->lock);
3752 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3753 goto out;
3755 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3756 idev->cnf.accept_dad < 1 ||
3757 !(ifp->flags&IFA_F_TENTATIVE) ||
3758 ifp->flags & IFA_F_NODAD) {
3759 bump_id = ifp->flags & IFA_F_TENTATIVE;
3760 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3761 spin_unlock(&ifp->lock);
3762 read_unlock_bh(&idev->lock);
3764 addrconf_dad_completed(ifp, bump_id);
3765 return;
3768 if (!(idev->if_flags & IF_READY)) {
3769 spin_unlock(&ifp->lock);
3770 read_unlock_bh(&idev->lock);
3772 * If the device is not ready:
3773 * - keep it tentative if it is a permanent address.
3774 * - otherwise, kill it.
3776 in6_ifa_hold(ifp);
3777 addrconf_dad_stop(ifp, 0);
3778 return;
3782 * Optimistic nodes can start receiving
3783 * Frames right away
3785 if (ifp->flags & IFA_F_OPTIMISTIC) {
3786 ip6_ins_rt(ifp->rt);
3787 if (ipv6_use_optimistic_addr(idev)) {
3788 /* Because optimistic nodes can use this address,
3789 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3791 notify = true;
3795 addrconf_dad_kick(ifp);
3796 out:
3797 spin_unlock(&ifp->lock);
3798 read_unlock_bh(&idev->lock);
3799 if (notify)
3800 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3803 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3805 bool begin_dad = false;
3807 spin_lock_bh(&ifp->lock);
3808 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3809 ifp->state = INET6_IFADDR_STATE_PREDAD;
3810 begin_dad = true;
3812 spin_unlock_bh(&ifp->lock);
3814 if (begin_dad)
3815 addrconf_mod_dad_work(ifp, 0);
3818 static void addrconf_dad_work(struct work_struct *w)
3820 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3821 struct inet6_ifaddr,
3822 dad_work);
3823 struct inet6_dev *idev = ifp->idev;
3824 bool bump_id, disable_ipv6 = false;
3825 struct in6_addr mcaddr;
3827 enum {
3828 DAD_PROCESS,
3829 DAD_BEGIN,
3830 DAD_ABORT,
3831 } action = DAD_PROCESS;
3833 rtnl_lock();
3835 spin_lock_bh(&ifp->lock);
3836 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3837 action = DAD_BEGIN;
3838 ifp->state = INET6_IFADDR_STATE_DAD;
3839 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3840 action = DAD_ABORT;
3841 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3843 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3844 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3845 struct in6_addr addr;
3847 addr.s6_addr32[0] = htonl(0xfe800000);
3848 addr.s6_addr32[1] = 0;
3850 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3851 ipv6_addr_equal(&ifp->addr, &addr)) {
3852 /* DAD failed for link-local based on MAC */
3853 idev->cnf.disable_ipv6 = 1;
3855 pr_info("%s: IPv6 being disabled!\n",
3856 ifp->idev->dev->name);
3857 disable_ipv6 = true;
3861 spin_unlock_bh(&ifp->lock);
3863 if (action == DAD_BEGIN) {
3864 addrconf_dad_begin(ifp);
3865 goto out;
3866 } else if (action == DAD_ABORT) {
3867 in6_ifa_hold(ifp);
3868 addrconf_dad_stop(ifp, 1);
3869 if (disable_ipv6)
3870 addrconf_ifdown(idev->dev, 0);
3871 goto out;
3874 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3875 goto out;
3877 write_lock_bh(&idev->lock);
3878 if (idev->dead || !(idev->if_flags & IF_READY)) {
3879 write_unlock_bh(&idev->lock);
3880 goto out;
3883 spin_lock(&ifp->lock);
3884 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3885 spin_unlock(&ifp->lock);
3886 write_unlock_bh(&idev->lock);
3887 goto out;
3890 if (ifp->dad_probes == 0) {
3892 * DAD was successful
3895 bump_id = ifp->flags & IFA_F_TENTATIVE;
3896 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3897 spin_unlock(&ifp->lock);
3898 write_unlock_bh(&idev->lock);
3900 addrconf_dad_completed(ifp, bump_id);
3902 goto out;
3905 ifp->dad_probes--;
3906 addrconf_mod_dad_work(ifp,
3907 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3908 spin_unlock(&ifp->lock);
3909 write_unlock_bh(&idev->lock);
3911 /* send a neighbour solicitation for our addr */
3912 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3913 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3914 out:
3915 in6_ifa_put(ifp);
3916 rtnl_unlock();
3919 /* ifp->idev must be at least read locked */
3920 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3922 struct inet6_ifaddr *ifpiter;
3923 struct inet6_dev *idev = ifp->idev;
3925 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3926 if (ifpiter->scope > IFA_LINK)
3927 break;
3928 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3929 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3930 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3931 IFA_F_PERMANENT)
3932 return false;
3934 return true;
3937 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
3939 struct net_device *dev = ifp->idev->dev;
3940 struct in6_addr lladdr;
3941 bool send_rs, send_mld;
3943 addrconf_del_dad_work(ifp);
3946 * Configure the address for reception. Now it is valid.
3949 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3951 /* If added prefix is link local and we are prepared to process
3952 router advertisements, start sending router solicitations.
3955 read_lock_bh(&ifp->idev->lock);
3956 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3957 send_rs = send_mld &&
3958 ipv6_accept_ra(ifp->idev) &&
3959 ifp->idev->cnf.rtr_solicits != 0 &&
3960 (dev->flags&IFF_LOOPBACK) == 0;
3961 read_unlock_bh(&ifp->idev->lock);
3963 /* While dad is in progress mld report's source address is in6_addrany.
3964 * Resend with proper ll now.
3966 if (send_mld)
3967 ipv6_mc_dad_complete(ifp->idev);
3969 if (send_rs) {
3971 * If a host as already performed a random delay
3972 * [...] as part of DAD [...] there is no need
3973 * to delay again before sending the first RS
3975 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3976 return;
3977 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3979 write_lock_bh(&ifp->idev->lock);
3980 spin_lock(&ifp->lock);
3981 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
3982 ifp->idev->cnf.rtr_solicit_interval);
3983 ifp->idev->rs_probes = 1;
3984 ifp->idev->if_flags |= IF_RS_SENT;
3985 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
3986 spin_unlock(&ifp->lock);
3987 write_unlock_bh(&ifp->idev->lock);
3990 if (bump_id)
3991 rt_genid_bump_ipv6(dev_net(dev));
3994 static void addrconf_dad_run(struct inet6_dev *idev)
3996 struct inet6_ifaddr *ifp;
3998 read_lock_bh(&idev->lock);
3999 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4000 spin_lock(&ifp->lock);
4001 if (ifp->flags & IFA_F_TENTATIVE &&
4002 ifp->state == INET6_IFADDR_STATE_DAD)
4003 addrconf_dad_kick(ifp);
4004 spin_unlock(&ifp->lock);
4006 read_unlock_bh(&idev->lock);
4009 #ifdef CONFIG_PROC_FS
4010 struct if6_iter_state {
4011 struct seq_net_private p;
4012 int bucket;
4013 int offset;
4016 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4018 struct inet6_ifaddr *ifa = NULL;
4019 struct if6_iter_state *state = seq->private;
4020 struct net *net = seq_file_net(seq);
4021 int p = 0;
4023 /* initial bucket if pos is 0 */
4024 if (pos == 0) {
4025 state->bucket = 0;
4026 state->offset = 0;
4029 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4030 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4031 addr_lst) {
4032 if (!net_eq(dev_net(ifa->idev->dev), net))
4033 continue;
4034 /* sync with offset */
4035 if (p < state->offset) {
4036 p++;
4037 continue;
4039 state->offset++;
4040 return ifa;
4043 /* prepare for next bucket */
4044 state->offset = 0;
4045 p = 0;
4047 return NULL;
4050 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4051 struct inet6_ifaddr *ifa)
4053 struct if6_iter_state *state = seq->private;
4054 struct net *net = seq_file_net(seq);
4056 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4057 if (!net_eq(dev_net(ifa->idev->dev), net))
4058 continue;
4059 state->offset++;
4060 return ifa;
4063 while (++state->bucket < IN6_ADDR_HSIZE) {
4064 state->offset = 0;
4065 hlist_for_each_entry_rcu_bh(ifa,
4066 &inet6_addr_lst[state->bucket], addr_lst) {
4067 if (!net_eq(dev_net(ifa->idev->dev), net))
4068 continue;
4069 state->offset++;
4070 return ifa;
4074 return NULL;
4077 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4078 __acquires(rcu_bh)
4080 rcu_read_lock_bh();
4081 return if6_get_first(seq, *pos);
4084 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4086 struct inet6_ifaddr *ifa;
4088 ifa = if6_get_next(seq, v);
4089 ++*pos;
4090 return ifa;
4093 static void if6_seq_stop(struct seq_file *seq, void *v)
4094 __releases(rcu_bh)
4096 rcu_read_unlock_bh();
4099 static int if6_seq_show(struct seq_file *seq, void *v)
4101 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4102 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4103 &ifp->addr,
4104 ifp->idev->dev->ifindex,
4105 ifp->prefix_len,
4106 ifp->scope,
4107 (u8) ifp->flags,
4108 ifp->idev->dev->name);
4109 return 0;
4112 static const struct seq_operations if6_seq_ops = {
4113 .start = if6_seq_start,
4114 .next = if6_seq_next,
4115 .show = if6_seq_show,
4116 .stop = if6_seq_stop,
4119 static int if6_seq_open(struct inode *inode, struct file *file)
4121 return seq_open_net(inode, file, &if6_seq_ops,
4122 sizeof(struct if6_iter_state));
4125 static const struct file_operations if6_fops = {
4126 .owner = THIS_MODULE,
4127 .open = if6_seq_open,
4128 .read = seq_read,
4129 .llseek = seq_lseek,
4130 .release = seq_release_net,
4133 static int __net_init if6_proc_net_init(struct net *net)
4135 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4136 return -ENOMEM;
4137 return 0;
4140 static void __net_exit if6_proc_net_exit(struct net *net)
4142 remove_proc_entry("if_inet6", net->proc_net);
4145 static struct pernet_operations if6_proc_net_ops = {
4146 .init = if6_proc_net_init,
4147 .exit = if6_proc_net_exit,
4150 int __init if6_proc_init(void)
4152 return register_pernet_subsys(&if6_proc_net_ops);
4155 void if6_proc_exit(void)
4157 unregister_pernet_subsys(&if6_proc_net_ops);
4159 #endif /* CONFIG_PROC_FS */
4161 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4162 /* Check if address is a home address configured on any interface. */
4163 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4165 int ret = 0;
4166 struct inet6_ifaddr *ifp = NULL;
4167 unsigned int hash = inet6_addr_hash(addr);
4169 rcu_read_lock_bh();
4170 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4171 if (!net_eq(dev_net(ifp->idev->dev), net))
4172 continue;
4173 if (ipv6_addr_equal(&ifp->addr, addr) &&
4174 (ifp->flags & IFA_F_HOMEADDRESS)) {
4175 ret = 1;
4176 break;
4179 rcu_read_unlock_bh();
4180 return ret;
4182 #endif
4185 * Periodic address status verification
4188 static void addrconf_verify_rtnl(void)
4190 unsigned long now, next, next_sec, next_sched;
4191 struct inet6_ifaddr *ifp;
4192 int i;
4194 ASSERT_RTNL();
4196 rcu_read_lock_bh();
4197 now = jiffies;
4198 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4200 cancel_delayed_work(&addr_chk_work);
4202 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4203 restart:
4204 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4205 unsigned long age;
4207 /* When setting preferred_lft to a value not zero or
4208 * infinity, while valid_lft is infinity
4209 * IFA_F_PERMANENT has a non-infinity life time.
4211 if ((ifp->flags & IFA_F_PERMANENT) &&
4212 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4213 continue;
4215 spin_lock(&ifp->lock);
4216 /* We try to batch several events at once. */
4217 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4219 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4220 age >= ifp->valid_lft) {
4221 spin_unlock(&ifp->lock);
4222 in6_ifa_hold(ifp);
4223 ipv6_del_addr(ifp);
4224 goto restart;
4225 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4226 spin_unlock(&ifp->lock);
4227 continue;
4228 } else if (age >= ifp->prefered_lft) {
4229 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4230 int deprecate = 0;
4232 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4233 deprecate = 1;
4234 ifp->flags |= IFA_F_DEPRECATED;
4237 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4238 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4239 next = ifp->tstamp + ifp->valid_lft * HZ;
4241 spin_unlock(&ifp->lock);
4243 if (deprecate) {
4244 in6_ifa_hold(ifp);
4246 ipv6_ifa_notify(0, ifp);
4247 in6_ifa_put(ifp);
4248 goto restart;
4250 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4251 !(ifp->flags&IFA_F_TENTATIVE)) {
4252 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4253 ifp->idev->cnf.dad_transmits *
4254 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4256 if (age >= ifp->prefered_lft - regen_advance) {
4257 struct inet6_ifaddr *ifpub = ifp->ifpub;
4258 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4259 next = ifp->tstamp + ifp->prefered_lft * HZ;
4260 if (!ifp->regen_count && ifpub) {
4261 ifp->regen_count++;
4262 in6_ifa_hold(ifp);
4263 in6_ifa_hold(ifpub);
4264 spin_unlock(&ifp->lock);
4266 spin_lock(&ifpub->lock);
4267 ifpub->regen_count = 0;
4268 spin_unlock(&ifpub->lock);
4269 ipv6_create_tempaddr(ifpub, ifp);
4270 in6_ifa_put(ifpub);
4271 in6_ifa_put(ifp);
4272 goto restart;
4274 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4275 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4276 spin_unlock(&ifp->lock);
4277 } else {
4278 /* ifp->prefered_lft <= ifp->valid_lft */
4279 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4280 next = ifp->tstamp + ifp->prefered_lft * HZ;
4281 spin_unlock(&ifp->lock);
4286 next_sec = round_jiffies_up(next);
4287 next_sched = next;
4289 /* If rounded timeout is accurate enough, accept it. */
4290 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4291 next_sched = next_sec;
4293 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4294 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4295 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4297 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4298 now, next, next_sec, next_sched);
4299 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4300 rcu_read_unlock_bh();
4303 static void addrconf_verify_work(struct work_struct *w)
4305 rtnl_lock();
4306 addrconf_verify_rtnl();
4307 rtnl_unlock();
4310 static void addrconf_verify(void)
4312 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4315 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4316 struct in6_addr **peer_pfx)
4318 struct in6_addr *pfx = NULL;
4320 *peer_pfx = NULL;
4322 if (addr)
4323 pfx = nla_data(addr);
4325 if (local) {
4326 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4327 *peer_pfx = pfx;
4328 pfx = nla_data(local);
4331 return pfx;
4334 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4335 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4336 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4337 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4338 [IFA_FLAGS] = { .len = sizeof(u32) },
4341 static int
4342 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4344 struct net *net = sock_net(skb->sk);
4345 struct ifaddrmsg *ifm;
4346 struct nlattr *tb[IFA_MAX+1];
4347 struct in6_addr *pfx, *peer_pfx;
4348 u32 ifa_flags;
4349 int err;
4351 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4352 if (err < 0)
4353 return err;
4355 ifm = nlmsg_data(nlh);
4356 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4357 if (!pfx)
4358 return -EINVAL;
4360 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4362 /* We ignore other flags so far. */
4363 ifa_flags &= IFA_F_MANAGETEMPADDR;
4365 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4366 ifm->ifa_prefixlen);
4369 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4370 u32 prefered_lft, u32 valid_lft)
4372 u32 flags;
4373 clock_t expires;
4374 unsigned long timeout;
4375 bool was_managetempaddr;
4376 bool had_prefixroute;
4378 ASSERT_RTNL();
4380 if (!valid_lft || (prefered_lft > valid_lft))
4381 return -EINVAL;
4383 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4384 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4385 return -EINVAL;
4387 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4388 if (addrconf_finite_timeout(timeout)) {
4389 expires = jiffies_to_clock_t(timeout * HZ);
4390 valid_lft = timeout;
4391 flags = RTF_EXPIRES;
4392 } else {
4393 expires = 0;
4394 flags = 0;
4395 ifa_flags |= IFA_F_PERMANENT;
4398 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4399 if (addrconf_finite_timeout(timeout)) {
4400 if (timeout == 0)
4401 ifa_flags |= IFA_F_DEPRECATED;
4402 prefered_lft = timeout;
4405 spin_lock_bh(&ifp->lock);
4406 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4407 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4408 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4409 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4410 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4411 IFA_F_NOPREFIXROUTE);
4412 ifp->flags |= ifa_flags;
4413 ifp->tstamp = jiffies;
4414 ifp->valid_lft = valid_lft;
4415 ifp->prefered_lft = prefered_lft;
4417 spin_unlock_bh(&ifp->lock);
4418 if (!(ifp->flags&IFA_F_TENTATIVE))
4419 ipv6_ifa_notify(0, ifp);
4421 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4422 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4423 expires, flags);
4424 } else if (had_prefixroute) {
4425 enum cleanup_prefix_rt_t action;
4426 unsigned long rt_expires;
4428 write_lock_bh(&ifp->idev->lock);
4429 action = check_cleanup_prefix_route(ifp, &rt_expires);
4430 write_unlock_bh(&ifp->idev->lock);
4432 if (action != CLEANUP_PREFIX_RT_NOP) {
4433 cleanup_prefix_route(ifp, rt_expires,
4434 action == CLEANUP_PREFIX_RT_DEL);
4438 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4439 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4440 valid_lft = prefered_lft = 0;
4441 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4442 !was_managetempaddr, jiffies);
4445 addrconf_verify_rtnl();
4447 return 0;
4450 static int
4451 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4453 struct net *net = sock_net(skb->sk);
4454 struct ifaddrmsg *ifm;
4455 struct nlattr *tb[IFA_MAX+1];
4456 struct in6_addr *pfx, *peer_pfx;
4457 struct inet6_ifaddr *ifa;
4458 struct net_device *dev;
4459 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4460 u32 ifa_flags;
4461 int err;
4463 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4464 if (err < 0)
4465 return err;
4467 ifm = nlmsg_data(nlh);
4468 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4469 if (!pfx)
4470 return -EINVAL;
4472 if (tb[IFA_CACHEINFO]) {
4473 struct ifa_cacheinfo *ci;
4475 ci = nla_data(tb[IFA_CACHEINFO]);
4476 valid_lft = ci->ifa_valid;
4477 preferred_lft = ci->ifa_prefered;
4478 } else {
4479 preferred_lft = INFINITY_LIFE_TIME;
4480 valid_lft = INFINITY_LIFE_TIME;
4483 dev = __dev_get_by_index(net, ifm->ifa_index);
4484 if (!dev)
4485 return -ENODEV;
4487 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4489 /* We ignore other flags so far. */
4490 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4491 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4493 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4494 if (!ifa) {
4496 * It would be best to check for !NLM_F_CREATE here but
4497 * userspace already relies on not having to provide this.
4499 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4500 ifm->ifa_prefixlen, ifa_flags,
4501 preferred_lft, valid_lft);
4504 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4505 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4506 err = -EEXIST;
4507 else
4508 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4510 in6_ifa_put(ifa);
4512 return err;
4515 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4516 u8 scope, int ifindex)
4518 struct ifaddrmsg *ifm;
4520 ifm = nlmsg_data(nlh);
4521 ifm->ifa_family = AF_INET6;
4522 ifm->ifa_prefixlen = prefixlen;
4523 ifm->ifa_flags = flags;
4524 ifm->ifa_scope = scope;
4525 ifm->ifa_index = ifindex;
4528 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4529 unsigned long tstamp, u32 preferred, u32 valid)
4531 struct ifa_cacheinfo ci;
4533 ci.cstamp = cstamp_delta(cstamp);
4534 ci.tstamp = cstamp_delta(tstamp);
4535 ci.ifa_prefered = preferred;
4536 ci.ifa_valid = valid;
4538 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4541 static inline int rt_scope(int ifa_scope)
4543 if (ifa_scope & IFA_HOST)
4544 return RT_SCOPE_HOST;
4545 else if (ifa_scope & IFA_LINK)
4546 return RT_SCOPE_LINK;
4547 else if (ifa_scope & IFA_SITE)
4548 return RT_SCOPE_SITE;
4549 else
4550 return RT_SCOPE_UNIVERSE;
4553 static inline int inet6_ifaddr_msgsize(void)
4555 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4556 + nla_total_size(16) /* IFA_LOCAL */
4557 + nla_total_size(16) /* IFA_ADDRESS */
4558 + nla_total_size(sizeof(struct ifa_cacheinfo))
4559 + nla_total_size(4) /* IFA_FLAGS */;
4562 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4563 u32 portid, u32 seq, int event, unsigned int flags)
4565 struct nlmsghdr *nlh;
4566 u32 preferred, valid;
4568 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4569 if (!nlh)
4570 return -EMSGSIZE;
4572 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4573 ifa->idev->dev->ifindex);
4575 if (!((ifa->flags&IFA_F_PERMANENT) &&
4576 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4577 preferred = ifa->prefered_lft;
4578 valid = ifa->valid_lft;
4579 if (preferred != INFINITY_LIFE_TIME) {
4580 long tval = (jiffies - ifa->tstamp)/HZ;
4581 if (preferred > tval)
4582 preferred -= tval;
4583 else
4584 preferred = 0;
4585 if (valid != INFINITY_LIFE_TIME) {
4586 if (valid > tval)
4587 valid -= tval;
4588 else
4589 valid = 0;
4592 } else {
4593 preferred = INFINITY_LIFE_TIME;
4594 valid = INFINITY_LIFE_TIME;
4597 if (!ipv6_addr_any(&ifa->peer_addr)) {
4598 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4599 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4600 goto error;
4601 } else
4602 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4603 goto error;
4605 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4606 goto error;
4608 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4609 goto error;
4611 nlmsg_end(skb, nlh);
4612 return 0;
4614 error:
4615 nlmsg_cancel(skb, nlh);
4616 return -EMSGSIZE;
4619 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4620 u32 portid, u32 seq, int event, u16 flags)
4622 struct nlmsghdr *nlh;
4623 u8 scope = RT_SCOPE_UNIVERSE;
4624 int ifindex = ifmca->idev->dev->ifindex;
4626 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4627 scope = RT_SCOPE_SITE;
4629 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4630 if (!nlh)
4631 return -EMSGSIZE;
4633 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4634 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4635 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4636 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4637 nlmsg_cancel(skb, nlh);
4638 return -EMSGSIZE;
4641 nlmsg_end(skb, nlh);
4642 return 0;
4645 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4646 u32 portid, u32 seq, int event, unsigned int flags)
4648 struct nlmsghdr *nlh;
4649 u8 scope = RT_SCOPE_UNIVERSE;
4650 int ifindex = ifaca->aca_idev->dev->ifindex;
4652 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4653 scope = RT_SCOPE_SITE;
4655 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4656 if (!nlh)
4657 return -EMSGSIZE;
4659 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4660 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4661 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4662 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4663 nlmsg_cancel(skb, nlh);
4664 return -EMSGSIZE;
4667 nlmsg_end(skb, nlh);
4668 return 0;
4671 enum addr_type_t {
4672 UNICAST_ADDR,
4673 MULTICAST_ADDR,
4674 ANYCAST_ADDR,
4677 /* called with rcu_read_lock() */
4678 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4679 struct netlink_callback *cb, enum addr_type_t type,
4680 int s_ip_idx, int *p_ip_idx)
4682 struct ifmcaddr6 *ifmca;
4683 struct ifacaddr6 *ifaca;
4684 int err = 1;
4685 int ip_idx = *p_ip_idx;
4687 read_lock_bh(&idev->lock);
4688 switch (type) {
4689 case UNICAST_ADDR: {
4690 struct inet6_ifaddr *ifa;
4692 /* unicast address incl. temp addr */
4693 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4694 if (++ip_idx < s_ip_idx)
4695 continue;
4696 err = inet6_fill_ifaddr(skb, ifa,
4697 NETLINK_CB(cb->skb).portid,
4698 cb->nlh->nlmsg_seq,
4699 RTM_NEWADDR,
4700 NLM_F_MULTI);
4701 if (err < 0)
4702 break;
4703 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4705 break;
4707 case MULTICAST_ADDR:
4708 /* multicast address */
4709 for (ifmca = idev->mc_list; ifmca;
4710 ifmca = ifmca->next, ip_idx++) {
4711 if (ip_idx < s_ip_idx)
4712 continue;
4713 err = inet6_fill_ifmcaddr(skb, ifmca,
4714 NETLINK_CB(cb->skb).portid,
4715 cb->nlh->nlmsg_seq,
4716 RTM_GETMULTICAST,
4717 NLM_F_MULTI);
4718 if (err < 0)
4719 break;
4721 break;
4722 case ANYCAST_ADDR:
4723 /* anycast address */
4724 for (ifaca = idev->ac_list; ifaca;
4725 ifaca = ifaca->aca_next, ip_idx++) {
4726 if (ip_idx < s_ip_idx)
4727 continue;
4728 err = inet6_fill_ifacaddr(skb, ifaca,
4729 NETLINK_CB(cb->skb).portid,
4730 cb->nlh->nlmsg_seq,
4731 RTM_GETANYCAST,
4732 NLM_F_MULTI);
4733 if (err < 0)
4734 break;
4736 break;
4737 default:
4738 break;
4740 read_unlock_bh(&idev->lock);
4741 *p_ip_idx = ip_idx;
4742 return err;
4745 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4746 enum addr_type_t type)
4748 struct net *net = sock_net(skb->sk);
4749 int h, s_h;
4750 int idx, ip_idx;
4751 int s_idx, s_ip_idx;
4752 struct net_device *dev;
4753 struct inet6_dev *idev;
4754 struct hlist_head *head;
4756 s_h = cb->args[0];
4757 s_idx = idx = cb->args[1];
4758 s_ip_idx = ip_idx = cb->args[2];
4760 rcu_read_lock();
4761 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4762 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4763 idx = 0;
4764 head = &net->dev_index_head[h];
4765 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4766 if (idx < s_idx)
4767 goto cont;
4768 if (h > s_h || idx > s_idx)
4769 s_ip_idx = 0;
4770 ip_idx = 0;
4771 idev = __in6_dev_get(dev);
4772 if (!idev)
4773 goto cont;
4775 if (in6_dump_addrs(idev, skb, cb, type,
4776 s_ip_idx, &ip_idx) < 0)
4777 goto done;
4778 cont:
4779 idx++;
4782 done:
4783 rcu_read_unlock();
4784 cb->args[0] = h;
4785 cb->args[1] = idx;
4786 cb->args[2] = ip_idx;
4788 return skb->len;
4791 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4793 enum addr_type_t type = UNICAST_ADDR;
4795 return inet6_dump_addr(skb, cb, type);
4798 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4800 enum addr_type_t type = MULTICAST_ADDR;
4802 return inet6_dump_addr(skb, cb, type);
4806 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4808 enum addr_type_t type = ANYCAST_ADDR;
4810 return inet6_dump_addr(skb, cb, type);
4813 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4815 struct net *net = sock_net(in_skb->sk);
4816 struct ifaddrmsg *ifm;
4817 struct nlattr *tb[IFA_MAX+1];
4818 struct in6_addr *addr = NULL, *peer;
4819 struct net_device *dev = NULL;
4820 struct inet6_ifaddr *ifa;
4821 struct sk_buff *skb;
4822 int err;
4824 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4825 if (err < 0)
4826 goto errout;
4828 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4829 if (!addr) {
4830 err = -EINVAL;
4831 goto errout;
4834 ifm = nlmsg_data(nlh);
4835 if (ifm->ifa_index)
4836 dev = __dev_get_by_index(net, ifm->ifa_index);
4838 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4839 if (!ifa) {
4840 err = -EADDRNOTAVAIL;
4841 goto errout;
4844 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4845 if (!skb) {
4846 err = -ENOBUFS;
4847 goto errout_ifa;
4850 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4851 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4852 if (err < 0) {
4853 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4854 WARN_ON(err == -EMSGSIZE);
4855 kfree_skb(skb);
4856 goto errout_ifa;
4858 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4859 errout_ifa:
4860 in6_ifa_put(ifa);
4861 errout:
4862 return err;
4865 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4867 struct sk_buff *skb;
4868 struct net *net = dev_net(ifa->idev->dev);
4869 int err = -ENOBUFS;
4871 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4872 if (!skb)
4873 goto errout;
4875 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4876 if (err < 0) {
4877 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4878 WARN_ON(err == -EMSGSIZE);
4879 kfree_skb(skb);
4880 goto errout;
4882 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4883 return;
4884 errout:
4885 if (err < 0)
4886 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4889 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4890 __s32 *array, int bytes)
4892 BUG_ON(bytes < (DEVCONF_MAX * 4));
4894 memset(array, 0, bytes);
4895 array[DEVCONF_FORWARDING] = cnf->forwarding;
4896 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4897 array[DEVCONF_MTU6] = cnf->mtu6;
4898 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4899 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4900 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4901 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4902 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4903 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4904 jiffies_to_msecs(cnf->rtr_solicit_interval);
4905 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4906 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4907 array[DEVCONF_RTR_SOLICIT_DELAY] =
4908 jiffies_to_msecs(cnf->rtr_solicit_delay);
4909 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4910 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4911 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4912 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4913 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4914 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4915 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4916 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4917 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4918 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4919 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4920 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4921 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4922 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4923 #ifdef CONFIG_IPV6_ROUTER_PREF
4924 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4925 array[DEVCONF_RTR_PROBE_INTERVAL] =
4926 jiffies_to_msecs(cnf->rtr_probe_interval);
4927 #ifdef CONFIG_IPV6_ROUTE_INFO
4928 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4929 #endif
4930 #endif
4931 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4932 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4933 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4934 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4935 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4936 #endif
4937 #ifdef CONFIG_IPV6_MROUTE
4938 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4939 #endif
4940 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4941 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4942 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4943 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4944 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4945 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4946 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4947 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4948 /* we omit DEVCONF_STABLE_SECRET for now */
4949 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4950 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4951 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4952 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4955 static inline size_t inet6_ifla6_size(void)
4957 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4958 + nla_total_size(sizeof(struct ifla_cacheinfo))
4959 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4960 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4961 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4962 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4965 static inline size_t inet6_if_nlmsg_size(void)
4967 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4968 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4969 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4970 + nla_total_size(4) /* IFLA_MTU */
4971 + nla_total_size(4) /* IFLA_LINK */
4972 + nla_total_size(1) /* IFLA_OPERSTATE */
4973 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4976 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4977 int bytes)
4979 int i;
4980 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
4981 BUG_ON(pad < 0);
4983 /* Use put_unaligned() because stats may not be aligned for u64. */
4984 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
4985 for (i = 1; i < ICMP6_MIB_MAX; i++)
4986 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4988 memset(&stats[ICMP6_MIB_MAX], 0, pad);
4991 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4992 int bytes, size_t syncpoff)
4994 int i, c;
4995 u64 buff[IPSTATS_MIB_MAX];
4996 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4998 BUG_ON(pad < 0);
5000 memset(buff, 0, sizeof(buff));
5001 buff[0] = IPSTATS_MIB_MAX;
5003 for_each_possible_cpu(c) {
5004 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5005 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5008 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5009 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5012 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5013 int bytes)
5015 switch (attrtype) {
5016 case IFLA_INET6_STATS:
5017 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5018 offsetof(struct ipstats_mib, syncp));
5019 break;
5020 case IFLA_INET6_ICMP6STATS:
5021 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5022 break;
5026 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5027 u32 ext_filter_mask)
5029 struct nlattr *nla;
5030 struct ifla_cacheinfo ci;
5032 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5033 goto nla_put_failure;
5034 ci.max_reasm_len = IPV6_MAXPLEN;
5035 ci.tstamp = cstamp_delta(idev->tstamp);
5036 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5037 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5038 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5039 goto nla_put_failure;
5040 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5041 if (!nla)
5042 goto nla_put_failure;
5043 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5045 /* XXX - MC not implemented */
5047 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5048 return 0;
5050 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5051 if (!nla)
5052 goto nla_put_failure;
5053 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5055 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5056 if (!nla)
5057 goto nla_put_failure;
5058 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5060 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5061 if (!nla)
5062 goto nla_put_failure;
5064 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5065 goto nla_put_failure;
5067 read_lock_bh(&idev->lock);
5068 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5069 read_unlock_bh(&idev->lock);
5071 return 0;
5073 nla_put_failure:
5074 return -EMSGSIZE;
5077 static size_t inet6_get_link_af_size(const struct net_device *dev,
5078 u32 ext_filter_mask)
5080 if (!__in6_dev_get(dev))
5081 return 0;
5083 return inet6_ifla6_size();
5086 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5087 u32 ext_filter_mask)
5089 struct inet6_dev *idev = __in6_dev_get(dev);
5091 if (!idev)
5092 return -ENODATA;
5094 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5095 return -EMSGSIZE;
5097 return 0;
5100 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5102 struct inet6_ifaddr *ifp;
5103 struct net_device *dev = idev->dev;
5104 bool clear_token, update_rs = false;
5105 struct in6_addr ll_addr;
5107 ASSERT_RTNL();
5109 if (!token)
5110 return -EINVAL;
5111 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5112 return -EINVAL;
5113 if (!ipv6_accept_ra(idev))
5114 return -EINVAL;
5115 if (idev->cnf.rtr_solicits == 0)
5116 return -EINVAL;
5118 write_lock_bh(&idev->lock);
5120 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5121 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5123 write_unlock_bh(&idev->lock);
5125 clear_token = ipv6_addr_any(token);
5126 if (clear_token)
5127 goto update_lft;
5129 if (!idev->dead && (idev->if_flags & IF_READY) &&
5130 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5131 IFA_F_OPTIMISTIC)) {
5132 /* If we're not ready, then normal ifup will take care
5133 * of this. Otherwise, we need to request our rs here.
5135 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5136 update_rs = true;
5139 update_lft:
5140 write_lock_bh(&idev->lock);
5142 if (update_rs) {
5143 idev->if_flags |= IF_RS_SENT;
5144 idev->rs_interval = rfc3315_s14_backoff_init(
5145 idev->cnf.rtr_solicit_interval);
5146 idev->rs_probes = 1;
5147 addrconf_mod_rs_timer(idev, idev->rs_interval);
5150 /* Well, that's kinda nasty ... */
5151 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5152 spin_lock(&ifp->lock);
5153 if (ifp->tokenized) {
5154 ifp->valid_lft = 0;
5155 ifp->prefered_lft = 0;
5157 spin_unlock(&ifp->lock);
5160 write_unlock_bh(&idev->lock);
5161 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5162 addrconf_verify_rtnl();
5163 return 0;
5166 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5167 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5168 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5171 static int inet6_validate_link_af(const struct net_device *dev,
5172 const struct nlattr *nla)
5174 struct nlattr *tb[IFLA_INET6_MAX + 1];
5176 if (dev && !__in6_dev_get(dev))
5177 return -EAFNOSUPPORT;
5179 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5182 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5184 int err = -EINVAL;
5185 struct inet6_dev *idev = __in6_dev_get(dev);
5186 struct nlattr *tb[IFLA_INET6_MAX + 1];
5188 if (!idev)
5189 return -EAFNOSUPPORT;
5191 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5192 BUG();
5194 if (tb[IFLA_INET6_TOKEN]) {
5195 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5196 if (err)
5197 return err;
5200 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5201 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5203 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5204 mode != IN6_ADDR_GEN_MODE_NONE &&
5205 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5206 mode != IN6_ADDR_GEN_MODE_RANDOM)
5207 return -EINVAL;
5209 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5210 !idev->cnf.stable_secret.initialized &&
5211 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5212 return -EINVAL;
5214 idev->addr_gen_mode = mode;
5215 err = 0;
5218 return err;
5221 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5222 u32 portid, u32 seq, int event, unsigned int flags)
5224 struct net_device *dev = idev->dev;
5225 struct ifinfomsg *hdr;
5226 struct nlmsghdr *nlh;
5227 void *protoinfo;
5229 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5230 if (!nlh)
5231 return -EMSGSIZE;
5233 hdr = nlmsg_data(nlh);
5234 hdr->ifi_family = AF_INET6;
5235 hdr->__ifi_pad = 0;
5236 hdr->ifi_type = dev->type;
5237 hdr->ifi_index = dev->ifindex;
5238 hdr->ifi_flags = dev_get_flags(dev);
5239 hdr->ifi_change = 0;
5241 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5242 (dev->addr_len &&
5243 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5244 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5245 (dev->ifindex != dev_get_iflink(dev) &&
5246 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5247 nla_put_u8(skb, IFLA_OPERSTATE,
5248 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5249 goto nla_put_failure;
5250 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5251 if (!protoinfo)
5252 goto nla_put_failure;
5254 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5255 goto nla_put_failure;
5257 nla_nest_end(skb, protoinfo);
5258 nlmsg_end(skb, nlh);
5259 return 0;
5261 nla_put_failure:
5262 nlmsg_cancel(skb, nlh);
5263 return -EMSGSIZE;
5266 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5268 struct net *net = sock_net(skb->sk);
5269 int h, s_h;
5270 int idx = 0, s_idx;
5271 struct net_device *dev;
5272 struct inet6_dev *idev;
5273 struct hlist_head *head;
5275 s_h = cb->args[0];
5276 s_idx = cb->args[1];
5278 rcu_read_lock();
5279 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5280 idx = 0;
5281 head = &net->dev_index_head[h];
5282 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5283 if (idx < s_idx)
5284 goto cont;
5285 idev = __in6_dev_get(dev);
5286 if (!idev)
5287 goto cont;
5288 if (inet6_fill_ifinfo(skb, idev,
5289 NETLINK_CB(cb->skb).portid,
5290 cb->nlh->nlmsg_seq,
5291 RTM_NEWLINK, NLM_F_MULTI) < 0)
5292 goto out;
5293 cont:
5294 idx++;
5297 out:
5298 rcu_read_unlock();
5299 cb->args[1] = idx;
5300 cb->args[0] = h;
5302 return skb->len;
5305 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5307 struct sk_buff *skb;
5308 struct net *net = dev_net(idev->dev);
5309 int err = -ENOBUFS;
5311 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5312 if (!skb)
5313 goto errout;
5315 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5316 if (err < 0) {
5317 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5318 WARN_ON(err == -EMSGSIZE);
5319 kfree_skb(skb);
5320 goto errout;
5322 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5323 return;
5324 errout:
5325 if (err < 0)
5326 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5329 static inline size_t inet6_prefix_nlmsg_size(void)
5331 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5332 + nla_total_size(sizeof(struct in6_addr))
5333 + nla_total_size(sizeof(struct prefix_cacheinfo));
5336 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5337 struct prefix_info *pinfo, u32 portid, u32 seq,
5338 int event, unsigned int flags)
5340 struct prefixmsg *pmsg;
5341 struct nlmsghdr *nlh;
5342 struct prefix_cacheinfo ci;
5344 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5345 if (!nlh)
5346 return -EMSGSIZE;
5348 pmsg = nlmsg_data(nlh);
5349 pmsg->prefix_family = AF_INET6;
5350 pmsg->prefix_pad1 = 0;
5351 pmsg->prefix_pad2 = 0;
5352 pmsg->prefix_ifindex = idev->dev->ifindex;
5353 pmsg->prefix_len = pinfo->prefix_len;
5354 pmsg->prefix_type = pinfo->type;
5355 pmsg->prefix_pad3 = 0;
5356 pmsg->prefix_flags = 0;
5357 if (pinfo->onlink)
5358 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5359 if (pinfo->autoconf)
5360 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5362 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5363 goto nla_put_failure;
5364 ci.preferred_time = ntohl(pinfo->prefered);
5365 ci.valid_time = ntohl(pinfo->valid);
5366 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5367 goto nla_put_failure;
5368 nlmsg_end(skb, nlh);
5369 return 0;
5371 nla_put_failure:
5372 nlmsg_cancel(skb, nlh);
5373 return -EMSGSIZE;
5376 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5377 struct prefix_info *pinfo)
5379 struct sk_buff *skb;
5380 struct net *net = dev_net(idev->dev);
5381 int err = -ENOBUFS;
5383 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5384 if (!skb)
5385 goto errout;
5387 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5388 if (err < 0) {
5389 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5390 WARN_ON(err == -EMSGSIZE);
5391 kfree_skb(skb);
5392 goto errout;
5394 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5395 return;
5396 errout:
5397 if (err < 0)
5398 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5401 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5403 struct net *net = dev_net(ifp->idev->dev);
5405 if (event)
5406 ASSERT_RTNL();
5408 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5410 switch (event) {
5411 case RTM_NEWADDR:
5413 * If the address was optimistic
5414 * we inserted the route at the start of
5415 * our DAD process, so we don't need
5416 * to do it again
5418 if (!(ifp->rt->rt6i_node))
5419 ip6_ins_rt(ifp->rt);
5420 if (ifp->idev->cnf.forwarding)
5421 addrconf_join_anycast(ifp);
5422 if (!ipv6_addr_any(&ifp->peer_addr))
5423 addrconf_prefix_route(&ifp->peer_addr, 128,
5424 ifp->idev->dev, 0, 0);
5425 break;
5426 case RTM_DELADDR:
5427 if (ifp->idev->cnf.forwarding)
5428 addrconf_leave_anycast(ifp);
5429 addrconf_leave_solict(ifp->idev, &ifp->addr);
5430 if (!ipv6_addr_any(&ifp->peer_addr)) {
5431 struct rt6_info *rt;
5433 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5434 ifp->idev->dev, 0, 0);
5435 if (rt)
5436 ip6_del_rt(rt);
5438 if (ifp->rt) {
5439 dst_hold(&ifp->rt->dst);
5440 ip6_del_rt(ifp->rt);
5442 rt_genid_bump_ipv6(net);
5443 break;
5445 atomic_inc(&net->ipv6.dev_addr_genid);
5448 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5450 rcu_read_lock_bh();
5451 if (likely(ifp->idev->dead == 0))
5452 __ipv6_ifa_notify(event, ifp);
5453 rcu_read_unlock_bh();
5456 #ifdef CONFIG_SYSCTL
5458 static
5459 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5460 void __user *buffer, size_t *lenp, loff_t *ppos)
5462 int *valp = ctl->data;
5463 int val = *valp;
5464 loff_t pos = *ppos;
5465 struct ctl_table lctl;
5466 int ret;
5469 * ctl->data points to idev->cnf.forwarding, we should
5470 * not modify it until we get the rtnl lock.
5472 lctl = *ctl;
5473 lctl.data = &val;
5475 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5477 if (write)
5478 ret = addrconf_fixup_forwarding(ctl, valp, val);
5479 if (ret)
5480 *ppos = pos;
5481 return ret;
5484 static
5485 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5486 void __user *buffer, size_t *lenp, loff_t *ppos)
5488 struct inet6_dev *idev = ctl->extra1;
5489 int min_mtu = IPV6_MIN_MTU;
5490 struct ctl_table lctl;
5492 lctl = *ctl;
5493 lctl.extra1 = &min_mtu;
5494 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5496 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5499 static void dev_disable_change(struct inet6_dev *idev)
5501 struct netdev_notifier_info info;
5503 if (!idev || !idev->dev)
5504 return;
5506 netdev_notifier_info_init(&info, idev->dev);
5507 if (idev->cnf.disable_ipv6)
5508 addrconf_notify(NULL, NETDEV_DOWN, &info);
5509 else
5510 addrconf_notify(NULL, NETDEV_UP, &info);
5513 static void addrconf_disable_change(struct net *net, __s32 newf)
5515 struct net_device *dev;
5516 struct inet6_dev *idev;
5518 rcu_read_lock();
5519 for_each_netdev_rcu(net, dev) {
5520 idev = __in6_dev_get(dev);
5521 if (idev) {
5522 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5523 idev->cnf.disable_ipv6 = newf;
5524 if (changed)
5525 dev_disable_change(idev);
5528 rcu_read_unlock();
5531 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5533 struct net *net;
5534 int old;
5536 if (!rtnl_trylock())
5537 return restart_syscall();
5539 net = (struct net *)table->extra2;
5540 old = *p;
5541 *p = newf;
5543 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5544 rtnl_unlock();
5545 return 0;
5548 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5549 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5550 addrconf_disable_change(net, newf);
5551 } else if ((!newf) ^ (!old))
5552 dev_disable_change((struct inet6_dev *)table->extra1);
5554 rtnl_unlock();
5555 return 0;
5558 static
5559 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5560 void __user *buffer, size_t *lenp, loff_t *ppos)
5562 int *valp = ctl->data;
5563 int val = *valp;
5564 loff_t pos = *ppos;
5565 struct ctl_table lctl;
5566 int ret;
5569 * ctl->data points to idev->cnf.disable_ipv6, we should
5570 * not modify it until we get the rtnl lock.
5572 lctl = *ctl;
5573 lctl.data = &val;
5575 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5577 if (write)
5578 ret = addrconf_disable_ipv6(ctl, valp, val);
5579 if (ret)
5580 *ppos = pos;
5581 return ret;
5584 static
5585 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5586 void __user *buffer, size_t *lenp, loff_t *ppos)
5588 int *valp = ctl->data;
5589 int ret;
5590 int old, new;
5592 old = *valp;
5593 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5594 new = *valp;
5596 if (write && old != new) {
5597 struct net *net = ctl->extra2;
5599 if (!rtnl_trylock())
5600 return restart_syscall();
5602 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5603 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5604 NETCONFA_IFINDEX_DEFAULT,
5605 net->ipv6.devconf_dflt);
5606 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5607 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5608 NETCONFA_IFINDEX_ALL,
5609 net->ipv6.devconf_all);
5610 else {
5611 struct inet6_dev *idev = ctl->extra1;
5613 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5614 idev->dev->ifindex,
5615 &idev->cnf);
5617 rtnl_unlock();
5620 return ret;
5623 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5624 void __user *buffer, size_t *lenp,
5625 loff_t *ppos)
5627 int err;
5628 struct in6_addr addr;
5629 char str[IPV6_MAX_STRLEN];
5630 struct ctl_table lctl = *ctl;
5631 struct net *net = ctl->extra2;
5632 struct ipv6_stable_secret *secret = ctl->data;
5634 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5635 return -EIO;
5637 lctl.maxlen = IPV6_MAX_STRLEN;
5638 lctl.data = str;
5640 if (!rtnl_trylock())
5641 return restart_syscall();
5643 if (!write && !secret->initialized) {
5644 err = -EIO;
5645 goto out;
5648 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5649 if (err >= sizeof(str)) {
5650 err = -EIO;
5651 goto out;
5654 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5655 if (err || !write)
5656 goto out;
5658 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5659 err = -EIO;
5660 goto out;
5663 secret->initialized = true;
5664 secret->secret = addr;
5666 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5667 struct net_device *dev;
5669 for_each_netdev(net, dev) {
5670 struct inet6_dev *idev = __in6_dev_get(dev);
5672 if (idev) {
5673 idev->addr_gen_mode =
5674 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5677 } else {
5678 struct inet6_dev *idev = ctl->extra1;
5680 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5683 out:
5684 rtnl_unlock();
5686 return err;
5689 static
5690 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5691 int write,
5692 void __user *buffer,
5693 size_t *lenp,
5694 loff_t *ppos)
5696 int *valp = ctl->data;
5697 int val = *valp;
5698 loff_t pos = *ppos;
5699 struct ctl_table lctl;
5700 int ret;
5702 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5703 * we should not modify it until we get the rtnl lock.
5705 lctl = *ctl;
5706 lctl.data = &val;
5708 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5710 if (write)
5711 ret = addrconf_fixup_linkdown(ctl, valp, val);
5712 if (ret)
5713 *ppos = pos;
5714 return ret;
5717 static int minus_one = -1;
5718 static const int one = 1;
5719 static const int two_five_five = 255;
5721 static const struct ctl_table addrconf_sysctl[] = {
5723 .procname = "forwarding",
5724 .data = &ipv6_devconf.forwarding,
5725 .maxlen = sizeof(int),
5726 .mode = 0644,
5727 .proc_handler = addrconf_sysctl_forward,
5730 .procname = "hop_limit",
5731 .data = &ipv6_devconf.hop_limit,
5732 .maxlen = sizeof(int),
5733 .mode = 0644,
5734 .proc_handler = proc_dointvec_minmax,
5735 .extra1 = (void *)&one,
5736 .extra2 = (void *)&two_five_five,
5739 .procname = "mtu",
5740 .data = &ipv6_devconf.mtu6,
5741 .maxlen = sizeof(int),
5742 .mode = 0644,
5743 .proc_handler = addrconf_sysctl_mtu,
5746 .procname = "accept_ra",
5747 .data = &ipv6_devconf.accept_ra,
5748 .maxlen = sizeof(int),
5749 .mode = 0644,
5750 .proc_handler = proc_dointvec,
5753 .procname = "accept_redirects",
5754 .data = &ipv6_devconf.accept_redirects,
5755 .maxlen = sizeof(int),
5756 .mode = 0644,
5757 .proc_handler = proc_dointvec,
5760 .procname = "autoconf",
5761 .data = &ipv6_devconf.autoconf,
5762 .maxlen = sizeof(int),
5763 .mode = 0644,
5764 .proc_handler = proc_dointvec,
5767 .procname = "dad_transmits",
5768 .data = &ipv6_devconf.dad_transmits,
5769 .maxlen = sizeof(int),
5770 .mode = 0644,
5771 .proc_handler = proc_dointvec,
5774 .procname = "router_solicitations",
5775 .data = &ipv6_devconf.rtr_solicits,
5776 .maxlen = sizeof(int),
5777 .mode = 0644,
5778 .proc_handler = proc_dointvec_minmax,
5779 .extra1 = &minus_one,
5782 .procname = "router_solicitation_interval",
5783 .data = &ipv6_devconf.rtr_solicit_interval,
5784 .maxlen = sizeof(int),
5785 .mode = 0644,
5786 .proc_handler = proc_dointvec_jiffies,
5789 .procname = "router_solicitation_max_interval",
5790 .data = &ipv6_devconf.rtr_solicit_max_interval,
5791 .maxlen = sizeof(int),
5792 .mode = 0644,
5793 .proc_handler = proc_dointvec_jiffies,
5796 .procname = "router_solicitation_delay",
5797 .data = &ipv6_devconf.rtr_solicit_delay,
5798 .maxlen = sizeof(int),
5799 .mode = 0644,
5800 .proc_handler = proc_dointvec_jiffies,
5803 .procname = "force_mld_version",
5804 .data = &ipv6_devconf.force_mld_version,
5805 .maxlen = sizeof(int),
5806 .mode = 0644,
5807 .proc_handler = proc_dointvec,
5810 .procname = "mldv1_unsolicited_report_interval",
5811 .data =
5812 &ipv6_devconf.mldv1_unsolicited_report_interval,
5813 .maxlen = sizeof(int),
5814 .mode = 0644,
5815 .proc_handler = proc_dointvec_ms_jiffies,
5818 .procname = "mldv2_unsolicited_report_interval",
5819 .data =
5820 &ipv6_devconf.mldv2_unsolicited_report_interval,
5821 .maxlen = sizeof(int),
5822 .mode = 0644,
5823 .proc_handler = proc_dointvec_ms_jiffies,
5826 .procname = "use_tempaddr",
5827 .data = &ipv6_devconf.use_tempaddr,
5828 .maxlen = sizeof(int),
5829 .mode = 0644,
5830 .proc_handler = proc_dointvec,
5833 .procname = "temp_valid_lft",
5834 .data = &ipv6_devconf.temp_valid_lft,
5835 .maxlen = sizeof(int),
5836 .mode = 0644,
5837 .proc_handler = proc_dointvec,
5840 .procname = "temp_prefered_lft",
5841 .data = &ipv6_devconf.temp_prefered_lft,
5842 .maxlen = sizeof(int),
5843 .mode = 0644,
5844 .proc_handler = proc_dointvec,
5847 .procname = "regen_max_retry",
5848 .data = &ipv6_devconf.regen_max_retry,
5849 .maxlen = sizeof(int),
5850 .mode = 0644,
5851 .proc_handler = proc_dointvec,
5854 .procname = "max_desync_factor",
5855 .data = &ipv6_devconf.max_desync_factor,
5856 .maxlen = sizeof(int),
5857 .mode = 0644,
5858 .proc_handler = proc_dointvec,
5861 .procname = "max_addresses",
5862 .data = &ipv6_devconf.max_addresses,
5863 .maxlen = sizeof(int),
5864 .mode = 0644,
5865 .proc_handler = proc_dointvec,
5868 .procname = "accept_ra_defrtr",
5869 .data = &ipv6_devconf.accept_ra_defrtr,
5870 .maxlen = sizeof(int),
5871 .mode = 0644,
5872 .proc_handler = proc_dointvec,
5875 .procname = "accept_ra_min_hop_limit",
5876 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5877 .maxlen = sizeof(int),
5878 .mode = 0644,
5879 .proc_handler = proc_dointvec,
5882 .procname = "accept_ra_pinfo",
5883 .data = &ipv6_devconf.accept_ra_pinfo,
5884 .maxlen = sizeof(int),
5885 .mode = 0644,
5886 .proc_handler = proc_dointvec,
5888 #ifdef CONFIG_IPV6_ROUTER_PREF
5890 .procname = "accept_ra_rtr_pref",
5891 .data = &ipv6_devconf.accept_ra_rtr_pref,
5892 .maxlen = sizeof(int),
5893 .mode = 0644,
5894 .proc_handler = proc_dointvec,
5897 .procname = "router_probe_interval",
5898 .data = &ipv6_devconf.rtr_probe_interval,
5899 .maxlen = sizeof(int),
5900 .mode = 0644,
5901 .proc_handler = proc_dointvec_jiffies,
5903 #ifdef CONFIG_IPV6_ROUTE_INFO
5905 .procname = "accept_ra_rt_info_max_plen",
5906 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5907 .maxlen = sizeof(int),
5908 .mode = 0644,
5909 .proc_handler = proc_dointvec,
5911 #endif
5912 #endif
5914 .procname = "proxy_ndp",
5915 .data = &ipv6_devconf.proxy_ndp,
5916 .maxlen = sizeof(int),
5917 .mode = 0644,
5918 .proc_handler = addrconf_sysctl_proxy_ndp,
5921 .procname = "accept_source_route",
5922 .data = &ipv6_devconf.accept_source_route,
5923 .maxlen = sizeof(int),
5924 .mode = 0644,
5925 .proc_handler = proc_dointvec,
5927 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5929 .procname = "optimistic_dad",
5930 .data = &ipv6_devconf.optimistic_dad,
5931 .maxlen = sizeof(int),
5932 .mode = 0644,
5933 .proc_handler = proc_dointvec,
5936 .procname = "use_optimistic",
5937 .data = &ipv6_devconf.use_optimistic,
5938 .maxlen = sizeof(int),
5939 .mode = 0644,
5940 .proc_handler = proc_dointvec,
5942 #endif
5943 #ifdef CONFIG_IPV6_MROUTE
5945 .procname = "mc_forwarding",
5946 .data = &ipv6_devconf.mc_forwarding,
5947 .maxlen = sizeof(int),
5948 .mode = 0444,
5949 .proc_handler = proc_dointvec,
5951 #endif
5953 .procname = "disable_ipv6",
5954 .data = &ipv6_devconf.disable_ipv6,
5955 .maxlen = sizeof(int),
5956 .mode = 0644,
5957 .proc_handler = addrconf_sysctl_disable,
5960 .procname = "accept_dad",
5961 .data = &ipv6_devconf.accept_dad,
5962 .maxlen = sizeof(int),
5963 .mode = 0644,
5964 .proc_handler = proc_dointvec,
5967 .procname = "force_tllao",
5968 .data = &ipv6_devconf.force_tllao,
5969 .maxlen = sizeof(int),
5970 .mode = 0644,
5971 .proc_handler = proc_dointvec
5974 .procname = "ndisc_notify",
5975 .data = &ipv6_devconf.ndisc_notify,
5976 .maxlen = sizeof(int),
5977 .mode = 0644,
5978 .proc_handler = proc_dointvec
5981 .procname = "suppress_frag_ndisc",
5982 .data = &ipv6_devconf.suppress_frag_ndisc,
5983 .maxlen = sizeof(int),
5984 .mode = 0644,
5985 .proc_handler = proc_dointvec
5988 .procname = "accept_ra_from_local",
5989 .data = &ipv6_devconf.accept_ra_from_local,
5990 .maxlen = sizeof(int),
5991 .mode = 0644,
5992 .proc_handler = proc_dointvec,
5995 .procname = "accept_ra_mtu",
5996 .data = &ipv6_devconf.accept_ra_mtu,
5997 .maxlen = sizeof(int),
5998 .mode = 0644,
5999 .proc_handler = proc_dointvec,
6002 .procname = "stable_secret",
6003 .data = &ipv6_devconf.stable_secret,
6004 .maxlen = IPV6_MAX_STRLEN,
6005 .mode = 0600,
6006 .proc_handler = addrconf_sysctl_stable_secret,
6009 .procname = "use_oif_addrs_only",
6010 .data = &ipv6_devconf.use_oif_addrs_only,
6011 .maxlen = sizeof(int),
6012 .mode = 0644,
6013 .proc_handler = proc_dointvec,
6016 .procname = "ignore_routes_with_linkdown",
6017 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6018 .maxlen = sizeof(int),
6019 .mode = 0644,
6020 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6023 .procname = "drop_unicast_in_l2_multicast",
6024 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6025 .maxlen = sizeof(int),
6026 .mode = 0644,
6027 .proc_handler = proc_dointvec,
6030 .procname = "drop_unsolicited_na",
6031 .data = &ipv6_devconf.drop_unsolicited_na,
6032 .maxlen = sizeof(int),
6033 .mode = 0644,
6034 .proc_handler = proc_dointvec,
6037 .procname = "keep_addr_on_down",
6038 .data = &ipv6_devconf.keep_addr_on_down,
6039 .maxlen = sizeof(int),
6040 .mode = 0644,
6041 .proc_handler = proc_dointvec,
6045 /* sentinel */
6049 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6050 struct inet6_dev *idev, struct ipv6_devconf *p)
6052 int i, ifindex;
6053 struct ctl_table *table;
6054 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6056 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6057 if (!table)
6058 goto out;
6060 for (i = 0; table[i].data; i++) {
6061 table[i].data += (char *)p - (char *)&ipv6_devconf;
6062 /* If one of these is already set, then it is not safe to
6063 * overwrite either of them: this makes proc_dointvec_minmax
6064 * usable.
6066 if (!table[i].extra1 && !table[i].extra2) {
6067 table[i].extra1 = idev; /* embedded; no ref */
6068 table[i].extra2 = net;
6072 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6074 p->sysctl_header = register_net_sysctl(net, path, table);
6075 if (!p->sysctl_header)
6076 goto free;
6078 if (!strcmp(dev_name, "all"))
6079 ifindex = NETCONFA_IFINDEX_ALL;
6080 else if (!strcmp(dev_name, "default"))
6081 ifindex = NETCONFA_IFINDEX_DEFAULT;
6082 else
6083 ifindex = idev->dev->ifindex;
6084 inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6085 return 0;
6087 free:
6088 kfree(table);
6089 out:
6090 return -ENOBUFS;
6093 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6095 struct ctl_table *table;
6097 if (!p->sysctl_header)
6098 return;
6100 table = p->sysctl_header->ctl_table_arg;
6101 unregister_net_sysctl_table(p->sysctl_header);
6102 p->sysctl_header = NULL;
6103 kfree(table);
6106 static int addrconf_sysctl_register(struct inet6_dev *idev)
6108 int err;
6110 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6111 return -EINVAL;
6113 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6114 &ndisc_ifinfo_sysctl_change);
6115 if (err)
6116 return err;
6117 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6118 idev, &idev->cnf);
6119 if (err)
6120 neigh_sysctl_unregister(idev->nd_parms);
6122 return err;
6125 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6127 __addrconf_sysctl_unregister(&idev->cnf);
6128 neigh_sysctl_unregister(idev->nd_parms);
6132 #endif
6134 static int __net_init addrconf_init_net(struct net *net)
6136 int err = -ENOMEM;
6137 struct ipv6_devconf *all, *dflt;
6139 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6140 if (!all)
6141 goto err_alloc_all;
6143 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6144 if (!dflt)
6145 goto err_alloc_dflt;
6147 /* these will be inherited by all namespaces */
6148 dflt->autoconf = ipv6_defaults.autoconf;
6149 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6151 dflt->stable_secret.initialized = false;
6152 all->stable_secret.initialized = false;
6154 net->ipv6.devconf_all = all;
6155 net->ipv6.devconf_dflt = dflt;
6157 #ifdef CONFIG_SYSCTL
6158 err = __addrconf_sysctl_register(net, "all", NULL, all);
6159 if (err < 0)
6160 goto err_reg_all;
6162 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6163 if (err < 0)
6164 goto err_reg_dflt;
6165 #endif
6166 return 0;
6168 #ifdef CONFIG_SYSCTL
6169 err_reg_dflt:
6170 __addrconf_sysctl_unregister(all);
6171 err_reg_all:
6172 kfree(dflt);
6173 #endif
6174 err_alloc_dflt:
6175 kfree(all);
6176 err_alloc_all:
6177 return err;
6180 static void __net_exit addrconf_exit_net(struct net *net)
6182 #ifdef CONFIG_SYSCTL
6183 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6184 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6185 #endif
6186 kfree(net->ipv6.devconf_dflt);
6187 kfree(net->ipv6.devconf_all);
6190 static struct pernet_operations addrconf_ops = {
6191 .init = addrconf_init_net,
6192 .exit = addrconf_exit_net,
6195 static struct rtnl_af_ops inet6_ops __read_mostly = {
6196 .family = AF_INET6,
6197 .fill_link_af = inet6_fill_link_af,
6198 .get_link_af_size = inet6_get_link_af_size,
6199 .validate_link_af = inet6_validate_link_af,
6200 .set_link_af = inet6_set_link_af,
6204 * Init / cleanup code
6207 int __init addrconf_init(void)
6209 struct inet6_dev *idev;
6210 int i, err;
6212 err = ipv6_addr_label_init();
6213 if (err < 0) {
6214 pr_crit("%s: cannot initialize default policy table: %d\n",
6215 __func__, err);
6216 goto out;
6219 err = register_pernet_subsys(&addrconf_ops);
6220 if (err < 0)
6221 goto out_addrlabel;
6223 addrconf_wq = create_workqueue("ipv6_addrconf");
6224 if (!addrconf_wq) {
6225 err = -ENOMEM;
6226 goto out_nowq;
6229 /* The addrconf netdev notifier requires that loopback_dev
6230 * has it's ipv6 private information allocated and setup
6231 * before it can bring up and give link-local addresses
6232 * to other devices which are up.
6234 * Unfortunately, loopback_dev is not necessarily the first
6235 * entry in the global dev_base list of net devices. In fact,
6236 * it is likely to be the very last entry on that list.
6237 * So this causes the notifier registry below to try and
6238 * give link-local addresses to all devices besides loopback_dev
6239 * first, then loopback_dev, which cases all the non-loopback_dev
6240 * devices to fail to get a link-local address.
6242 * So, as a temporary fix, allocate the ipv6 structure for
6243 * loopback_dev first by hand.
6244 * Longer term, all of the dependencies ipv6 has upon the loopback
6245 * device and it being up should be removed.
6247 rtnl_lock();
6248 idev = ipv6_add_dev(init_net.loopback_dev);
6249 rtnl_unlock();
6250 if (IS_ERR(idev)) {
6251 err = PTR_ERR(idev);
6252 goto errlo;
6255 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6256 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6258 register_netdevice_notifier(&ipv6_dev_notf);
6260 addrconf_verify();
6262 rtnl_af_register(&inet6_ops);
6264 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6265 NULL);
6266 if (err < 0)
6267 goto errout;
6269 /* Only the first call to __rtnl_register can fail */
6270 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6271 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6272 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6273 inet6_dump_ifaddr, NULL);
6274 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6275 inet6_dump_ifmcaddr, NULL);
6276 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6277 inet6_dump_ifacaddr, NULL);
6278 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6279 inet6_netconf_dump_devconf, NULL);
6281 ipv6_addr_label_rtnl_register();
6283 return 0;
6284 errout:
6285 rtnl_af_unregister(&inet6_ops);
6286 unregister_netdevice_notifier(&ipv6_dev_notf);
6287 errlo:
6288 destroy_workqueue(addrconf_wq);
6289 out_nowq:
6290 unregister_pernet_subsys(&addrconf_ops);
6291 out_addrlabel:
6292 ipv6_addr_label_cleanup();
6293 out:
6294 return err;
6297 void addrconf_cleanup(void)
6299 struct net_device *dev;
6300 int i;
6302 unregister_netdevice_notifier(&ipv6_dev_notf);
6303 unregister_pernet_subsys(&addrconf_ops);
6304 ipv6_addr_label_cleanup();
6306 rtnl_lock();
6308 __rtnl_af_unregister(&inet6_ops);
6310 /* clean dev list */
6311 for_each_netdev(&init_net, dev) {
6312 if (__in6_dev_get(dev) == NULL)
6313 continue;
6314 addrconf_ifdown(dev, 1);
6316 addrconf_ifdown(init_net.loopback_dev, 2);
6319 * Check hash table.
6321 spin_lock_bh(&addrconf_hash_lock);
6322 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6323 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6324 spin_unlock_bh(&addrconf_hash_lock);
6325 cancel_delayed_work(&addr_chk_work);
6326 rtnl_unlock();
6328 destroy_workqueue(addrconf_wq);