Linux 4.3
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob36b85bd05ac8a320b1b910c7d3454da1139ecd3c
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/af_ieee802154.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 <linux/if_tunnel.h>
85 #include <linux/rtnetlink.h>
86 #include <linux/netconf.h>
87 #include <linux/random.h>
88 #include <linux/uaccess.h>
89 #include <asm/unaligned.h>
91 #include <linux/proc_fs.h>
92 #include <linux/seq_file.h>
93 #include <linux/export.h>
95 /* Set to 3 to get tracing... */
96 #define ACONF_DEBUG 2
98 #if ACONF_DEBUG >= 3
99 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
100 #else
101 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
102 #endif
104 #define INFINITY_LIFE_TIME 0xFFFFFFFF
106 #define IPV6_MAX_STRLEN \
107 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
109 static inline u32 cstamp_delta(unsigned long cstamp)
111 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
114 #ifdef CONFIG_SYSCTL
115 static int addrconf_sysctl_register(struct inet6_dev *idev);
116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
117 #else
118 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
120 return 0;
123 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
126 #endif
128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
130 static void ipv6_regen_rndid(unsigned long data);
132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
133 static int ipv6_count_addresses(struct inet6_dev *idev);
134 static int ipv6_generate_stable_address(struct in6_addr *addr,
135 u8 dad_count,
136 const struct inet6_dev *idev);
139 * Configured unicast address hash table
141 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
142 static DEFINE_SPINLOCK(addrconf_hash_lock);
144 static void addrconf_verify(void);
145 static void addrconf_verify_rtnl(void);
146 static void addrconf_verify_work(struct work_struct *);
148 static struct workqueue_struct *addrconf_wq;
149 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
151 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
152 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
154 static void addrconf_type_change(struct net_device *dev,
155 unsigned long event);
156 static int addrconf_ifdown(struct net_device *dev, int how);
158 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
159 int plen,
160 const struct net_device *dev,
161 u32 flags, u32 noflags);
163 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
164 static void addrconf_dad_work(struct work_struct *w);
165 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
166 static void addrconf_dad_run(struct inet6_dev *idev);
167 static void addrconf_rs_timer(unsigned long data);
168 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
169 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
171 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
172 struct prefix_info *pinfo);
173 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
174 struct net_device *dev);
176 static struct ipv6_devconf ipv6_devconf __read_mostly = {
177 .forwarding = 0,
178 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
179 .mtu6 = IPV6_MIN_MTU,
180 .accept_ra = 1,
181 .accept_redirects = 1,
182 .autoconf = 1,
183 .force_mld_version = 0,
184 .mldv1_unsolicited_report_interval = 10 * HZ,
185 .mldv2_unsolicited_report_interval = HZ,
186 .dad_transmits = 1,
187 .rtr_solicits = MAX_RTR_SOLICITATIONS,
188 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
189 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
190 .use_tempaddr = 0,
191 .temp_valid_lft = TEMP_VALID_LIFETIME,
192 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
193 .regen_max_retry = REGEN_MAX_RETRY,
194 .max_desync_factor = MAX_DESYNC_FACTOR,
195 .max_addresses = IPV6_MAX_ADDRESSES,
196 .accept_ra_defrtr = 1,
197 .accept_ra_from_local = 0,
198 .accept_ra_min_hop_limit= 1,
199 .accept_ra_pinfo = 1,
200 #ifdef CONFIG_IPV6_ROUTER_PREF
201 .accept_ra_rtr_pref = 1,
202 .rtr_probe_interval = 60 * HZ,
203 #ifdef CONFIG_IPV6_ROUTE_INFO
204 .accept_ra_rt_info_max_plen = 0,
205 #endif
206 #endif
207 .proxy_ndp = 0,
208 .accept_source_route = 0, /* we do not accept RH0 by default. */
209 .disable_ipv6 = 0,
210 .accept_dad = 1,
211 .suppress_frag_ndisc = 1,
212 .accept_ra_mtu = 1,
213 .stable_secret = {
214 .initialized = false,
216 .use_oif_addrs_only = 0,
217 .ignore_routes_with_linkdown = 0,
220 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
221 .forwarding = 0,
222 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
223 .mtu6 = IPV6_MIN_MTU,
224 .accept_ra = 1,
225 .accept_redirects = 1,
226 .autoconf = 1,
227 .force_mld_version = 0,
228 .mldv1_unsolicited_report_interval = 10 * HZ,
229 .mldv2_unsolicited_report_interval = HZ,
230 .dad_transmits = 1,
231 .rtr_solicits = MAX_RTR_SOLICITATIONS,
232 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
233 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
234 .use_tempaddr = 0,
235 .temp_valid_lft = TEMP_VALID_LIFETIME,
236 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
237 .regen_max_retry = REGEN_MAX_RETRY,
238 .max_desync_factor = MAX_DESYNC_FACTOR,
239 .max_addresses = IPV6_MAX_ADDRESSES,
240 .accept_ra_defrtr = 1,
241 .accept_ra_from_local = 0,
242 .accept_ra_min_hop_limit= 1,
243 .accept_ra_pinfo = 1,
244 #ifdef CONFIG_IPV6_ROUTER_PREF
245 .accept_ra_rtr_pref = 1,
246 .rtr_probe_interval = 60 * HZ,
247 #ifdef CONFIG_IPV6_ROUTE_INFO
248 .accept_ra_rt_info_max_plen = 0,
249 #endif
250 #endif
251 .proxy_ndp = 0,
252 .accept_source_route = 0, /* we do not accept RH0 by default. */
253 .disable_ipv6 = 0,
254 .accept_dad = 1,
255 .suppress_frag_ndisc = 1,
256 .accept_ra_mtu = 1,
257 .stable_secret = {
258 .initialized = false,
260 .use_oif_addrs_only = 0,
261 .ignore_routes_with_linkdown = 0,
264 /* Check if a valid qdisc is available */
265 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
267 return !qdisc_tx_is_noop(dev);
270 static void addrconf_del_rs_timer(struct inet6_dev *idev)
272 if (del_timer(&idev->rs_timer))
273 __in6_dev_put(idev);
276 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
278 if (cancel_delayed_work(&ifp->dad_work))
279 __in6_ifa_put(ifp);
282 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
283 unsigned long when)
285 if (!timer_pending(&idev->rs_timer))
286 in6_dev_hold(idev);
287 mod_timer(&idev->rs_timer, jiffies + when);
290 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
291 unsigned long delay)
293 if (!delayed_work_pending(&ifp->dad_work))
294 in6_ifa_hold(ifp);
295 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
298 static int snmp6_alloc_dev(struct inet6_dev *idev)
300 int i;
302 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
303 if (!idev->stats.ipv6)
304 goto err_ip;
306 for_each_possible_cpu(i) {
307 struct ipstats_mib *addrconf_stats;
308 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
309 u64_stats_init(&addrconf_stats->syncp);
313 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
314 GFP_KERNEL);
315 if (!idev->stats.icmpv6dev)
316 goto err_icmp;
317 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
318 GFP_KERNEL);
319 if (!idev->stats.icmpv6msgdev)
320 goto err_icmpmsg;
322 return 0;
324 err_icmpmsg:
325 kfree(idev->stats.icmpv6dev);
326 err_icmp:
327 free_percpu(idev->stats.ipv6);
328 err_ip:
329 return -ENOMEM;
332 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
334 struct inet6_dev *ndev;
335 int err = -ENOMEM;
337 ASSERT_RTNL();
339 if (dev->mtu < IPV6_MIN_MTU)
340 return ERR_PTR(-EINVAL);
342 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
343 if (!ndev)
344 return ERR_PTR(err);
346 rwlock_init(&ndev->lock);
347 ndev->dev = dev;
348 INIT_LIST_HEAD(&ndev->addr_list);
349 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
350 (unsigned long)ndev);
351 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
352 ndev->cnf.mtu6 = dev->mtu;
353 ndev->cnf.sysctl = NULL;
354 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
355 if (!ndev->nd_parms) {
356 kfree(ndev);
357 return ERR_PTR(err);
359 if (ndev->cnf.forwarding)
360 dev_disable_lro(dev);
361 /* We refer to the device */
362 dev_hold(dev);
364 if (snmp6_alloc_dev(ndev) < 0) {
365 ADBG(KERN_WARNING
366 "%s: cannot allocate memory for statistics; dev=%s.\n",
367 __func__, dev->name);
368 neigh_parms_release(&nd_tbl, ndev->nd_parms);
369 dev_put(dev);
370 kfree(ndev);
371 return ERR_PTR(err);
374 if (snmp6_register_dev(ndev) < 0) {
375 ADBG(KERN_WARNING
376 "%s: cannot create /proc/net/dev_snmp6/%s\n",
377 __func__, dev->name);
378 goto err_release;
381 /* One reference from device. We must do this before
382 * we invoke __ipv6_regen_rndid().
384 in6_dev_hold(ndev);
386 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
387 ndev->cnf.accept_dad = -1;
389 #if IS_ENABLED(CONFIG_IPV6_SIT)
390 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
391 pr_info("%s: Disabled Multicast RS\n", dev->name);
392 ndev->cnf.rtr_solicits = 0;
394 #endif
396 INIT_LIST_HEAD(&ndev->tempaddr_list);
397 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
398 if ((dev->flags&IFF_LOOPBACK) ||
399 dev->type == ARPHRD_TUNNEL ||
400 dev->type == ARPHRD_TUNNEL6 ||
401 dev->type == ARPHRD_SIT ||
402 dev->type == ARPHRD_NONE) {
403 ndev->cnf.use_tempaddr = -1;
404 } else {
405 in6_dev_hold(ndev);
406 ipv6_regen_rndid((unsigned long) ndev);
409 ndev->token = in6addr_any;
411 if (netif_running(dev) && addrconf_qdisc_ok(dev))
412 ndev->if_flags |= IF_READY;
414 ipv6_mc_init_dev(ndev);
415 ndev->tstamp = jiffies;
416 err = addrconf_sysctl_register(ndev);
417 if (err) {
418 ipv6_mc_destroy_dev(ndev);
419 del_timer(&ndev->regen_timer);
420 goto err_release;
422 /* protected by rtnl_lock */
423 rcu_assign_pointer(dev->ip6_ptr, ndev);
425 /* Join interface-local all-node multicast group */
426 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
428 /* Join all-node multicast group */
429 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
431 /* Join all-router multicast group if forwarding is set */
432 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
433 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
435 return ndev;
437 err_release:
438 neigh_parms_release(&nd_tbl, ndev->nd_parms);
439 ndev->dead = 1;
440 in6_dev_finish_destroy(ndev);
441 return ERR_PTR(err);
444 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
446 struct inet6_dev *idev;
448 ASSERT_RTNL();
450 idev = __in6_dev_get(dev);
451 if (!idev) {
452 idev = ipv6_add_dev(dev);
453 if (IS_ERR(idev))
454 return NULL;
457 if (dev->flags&IFF_UP)
458 ipv6_mc_up(idev);
459 return idev;
462 static int inet6_netconf_msgsize_devconf(int type)
464 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
465 + nla_total_size(4); /* NETCONFA_IFINDEX */
467 /* type -1 is used for ALL */
468 if (type == -1 || type == NETCONFA_FORWARDING)
469 size += nla_total_size(4);
470 #ifdef CONFIG_IPV6_MROUTE
471 if (type == -1 || type == NETCONFA_MC_FORWARDING)
472 size += nla_total_size(4);
473 #endif
474 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
475 size += nla_total_size(4);
477 if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
478 size += nla_total_size(4);
480 return size;
483 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
484 struct ipv6_devconf *devconf, u32 portid,
485 u32 seq, int event, unsigned int flags,
486 int type)
488 struct nlmsghdr *nlh;
489 struct netconfmsg *ncm;
491 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
492 flags);
493 if (!nlh)
494 return -EMSGSIZE;
496 ncm = nlmsg_data(nlh);
497 ncm->ncm_family = AF_INET6;
499 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
500 goto nla_put_failure;
502 /* type -1 is used for ALL */
503 if ((type == -1 || type == NETCONFA_FORWARDING) &&
504 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
505 goto nla_put_failure;
506 #ifdef CONFIG_IPV6_MROUTE
507 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
508 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
509 devconf->mc_forwarding) < 0)
510 goto nla_put_failure;
511 #endif
512 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
513 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
514 goto nla_put_failure;
516 if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
517 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
518 devconf->ignore_routes_with_linkdown) < 0)
519 goto nla_put_failure;
521 nlmsg_end(skb, nlh);
522 return 0;
524 nla_put_failure:
525 nlmsg_cancel(skb, nlh);
526 return -EMSGSIZE;
529 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
530 struct ipv6_devconf *devconf)
532 struct sk_buff *skb;
533 int err = -ENOBUFS;
535 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
536 if (!skb)
537 goto errout;
539 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
540 RTM_NEWNETCONF, 0, type);
541 if (err < 0) {
542 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
543 WARN_ON(err == -EMSGSIZE);
544 kfree_skb(skb);
545 goto errout;
547 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
548 return;
549 errout:
550 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
553 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
554 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
555 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
556 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
557 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
560 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
561 struct nlmsghdr *nlh)
563 struct net *net = sock_net(in_skb->sk);
564 struct nlattr *tb[NETCONFA_MAX+1];
565 struct netconfmsg *ncm;
566 struct sk_buff *skb;
567 struct ipv6_devconf *devconf;
568 struct inet6_dev *in6_dev;
569 struct net_device *dev;
570 int ifindex;
571 int err;
573 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
574 devconf_ipv6_policy);
575 if (err < 0)
576 goto errout;
578 err = EINVAL;
579 if (!tb[NETCONFA_IFINDEX])
580 goto errout;
582 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
583 switch (ifindex) {
584 case NETCONFA_IFINDEX_ALL:
585 devconf = net->ipv6.devconf_all;
586 break;
587 case NETCONFA_IFINDEX_DEFAULT:
588 devconf = net->ipv6.devconf_dflt;
589 break;
590 default:
591 dev = __dev_get_by_index(net, ifindex);
592 if (!dev)
593 goto errout;
594 in6_dev = __in6_dev_get(dev);
595 if (!in6_dev)
596 goto errout;
597 devconf = &in6_dev->cnf;
598 break;
601 err = -ENOBUFS;
602 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
603 if (!skb)
604 goto errout;
606 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
607 NETLINK_CB(in_skb).portid,
608 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
609 -1);
610 if (err < 0) {
611 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
612 WARN_ON(err == -EMSGSIZE);
613 kfree_skb(skb);
614 goto errout;
616 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
617 errout:
618 return err;
621 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
622 struct netlink_callback *cb)
624 struct net *net = sock_net(skb->sk);
625 int h, s_h;
626 int idx, s_idx;
627 struct net_device *dev;
628 struct inet6_dev *idev;
629 struct hlist_head *head;
631 s_h = cb->args[0];
632 s_idx = idx = cb->args[1];
634 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
635 idx = 0;
636 head = &net->dev_index_head[h];
637 rcu_read_lock();
638 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
639 net->dev_base_seq;
640 hlist_for_each_entry_rcu(dev, head, index_hlist) {
641 if (idx < s_idx)
642 goto cont;
643 idev = __in6_dev_get(dev);
644 if (!idev)
645 goto cont;
647 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
648 &idev->cnf,
649 NETLINK_CB(cb->skb).portid,
650 cb->nlh->nlmsg_seq,
651 RTM_NEWNETCONF,
652 NLM_F_MULTI,
653 -1) < 0) {
654 rcu_read_unlock();
655 goto done;
657 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
658 cont:
659 idx++;
661 rcu_read_unlock();
663 if (h == NETDEV_HASHENTRIES) {
664 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
665 net->ipv6.devconf_all,
666 NETLINK_CB(cb->skb).portid,
667 cb->nlh->nlmsg_seq,
668 RTM_NEWNETCONF, NLM_F_MULTI,
669 -1) < 0)
670 goto done;
671 else
672 h++;
674 if (h == NETDEV_HASHENTRIES + 1) {
675 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
676 net->ipv6.devconf_dflt,
677 NETLINK_CB(cb->skb).portid,
678 cb->nlh->nlmsg_seq,
679 RTM_NEWNETCONF, NLM_F_MULTI,
680 -1) < 0)
681 goto done;
682 else
683 h++;
685 done:
686 cb->args[0] = h;
687 cb->args[1] = idx;
689 return skb->len;
692 #ifdef CONFIG_SYSCTL
693 static void dev_forward_change(struct inet6_dev *idev)
695 struct net_device *dev;
696 struct inet6_ifaddr *ifa;
698 if (!idev)
699 return;
700 dev = idev->dev;
701 if (idev->cnf.forwarding)
702 dev_disable_lro(dev);
703 if (dev->flags & IFF_MULTICAST) {
704 if (idev->cnf.forwarding) {
705 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
706 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
707 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
708 } else {
709 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
710 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
711 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
715 list_for_each_entry(ifa, &idev->addr_list, if_list) {
716 if (ifa->flags&IFA_F_TENTATIVE)
717 continue;
718 if (idev->cnf.forwarding)
719 addrconf_join_anycast(ifa);
720 else
721 addrconf_leave_anycast(ifa);
723 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
724 dev->ifindex, &idev->cnf);
728 static void addrconf_forward_change(struct net *net, __s32 newf)
730 struct net_device *dev;
731 struct inet6_dev *idev;
733 for_each_netdev(net, dev) {
734 idev = __in6_dev_get(dev);
735 if (idev) {
736 int changed = (!idev->cnf.forwarding) ^ (!newf);
737 idev->cnf.forwarding = newf;
738 if (changed)
739 dev_forward_change(idev);
744 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
746 struct net *net;
747 int old;
749 if (!rtnl_trylock())
750 return restart_syscall();
752 net = (struct net *)table->extra2;
753 old = *p;
754 *p = newf;
756 if (p == &net->ipv6.devconf_dflt->forwarding) {
757 if ((!newf) ^ (!old))
758 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
759 NETCONFA_IFINDEX_DEFAULT,
760 net->ipv6.devconf_dflt);
761 rtnl_unlock();
762 return 0;
765 if (p == &net->ipv6.devconf_all->forwarding) {
766 net->ipv6.devconf_dflt->forwarding = newf;
767 addrconf_forward_change(net, newf);
768 if ((!newf) ^ (!old))
769 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
770 NETCONFA_IFINDEX_ALL,
771 net->ipv6.devconf_all);
772 } else if ((!newf) ^ (!old))
773 dev_forward_change((struct inet6_dev *)table->extra1);
774 rtnl_unlock();
776 if (newf)
777 rt6_purge_dflt_routers(net);
778 return 1;
781 static void addrconf_linkdown_change(struct net *net, __s32 newf)
783 struct net_device *dev;
784 struct inet6_dev *idev;
786 for_each_netdev(net, dev) {
787 idev = __in6_dev_get(dev);
788 if (idev) {
789 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
791 idev->cnf.ignore_routes_with_linkdown = newf;
792 if (changed)
793 inet6_netconf_notify_devconf(dev_net(dev),
794 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
795 dev->ifindex,
796 &idev->cnf);
801 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
803 struct net *net;
804 int old;
806 if (!rtnl_trylock())
807 return restart_syscall();
809 net = (struct net *)table->extra2;
810 old = *p;
811 *p = newf;
813 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
814 if ((!newf) ^ (!old))
815 inet6_netconf_notify_devconf(net,
816 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
817 NETCONFA_IFINDEX_DEFAULT,
818 net->ipv6.devconf_dflt);
819 rtnl_unlock();
820 return 0;
823 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
824 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
825 addrconf_linkdown_change(net, newf);
826 if ((!newf) ^ (!old))
827 inet6_netconf_notify_devconf(net,
828 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
829 NETCONFA_IFINDEX_ALL,
830 net->ipv6.devconf_all);
832 rtnl_unlock();
834 return 1;
837 #endif
839 /* Nobody refers to this ifaddr, destroy it */
840 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
842 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
844 #ifdef NET_REFCNT_DEBUG
845 pr_debug("%s\n", __func__);
846 #endif
848 in6_dev_put(ifp->idev);
850 if (cancel_delayed_work(&ifp->dad_work))
851 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
852 ifp);
854 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
855 pr_warn("Freeing alive inet6 address %p\n", ifp);
856 return;
858 ip6_rt_put(ifp->rt);
860 kfree_rcu(ifp, rcu);
863 static void
864 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
866 struct list_head *p;
867 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
870 * Each device address list is sorted in order of scope -
871 * global before linklocal.
873 list_for_each(p, &idev->addr_list) {
874 struct inet6_ifaddr *ifa
875 = list_entry(p, struct inet6_ifaddr, if_list);
876 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
877 break;
880 list_add_tail(&ifp->if_list, p);
883 static u32 inet6_addr_hash(const struct in6_addr *addr)
885 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
888 /* On success it returns ifp with increased reference count */
890 static struct inet6_ifaddr *
891 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
892 const struct in6_addr *peer_addr, int pfxlen,
893 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
895 struct inet6_ifaddr *ifa = NULL;
896 struct rt6_info *rt;
897 unsigned int hash;
898 int err = 0;
899 int addr_type = ipv6_addr_type(addr);
901 if (addr_type == IPV6_ADDR_ANY ||
902 addr_type & IPV6_ADDR_MULTICAST ||
903 (!(idev->dev->flags & IFF_LOOPBACK) &&
904 addr_type & IPV6_ADDR_LOOPBACK))
905 return ERR_PTR(-EADDRNOTAVAIL);
907 rcu_read_lock_bh();
908 if (idev->dead) {
909 err = -ENODEV; /*XXX*/
910 goto out2;
913 if (idev->cnf.disable_ipv6) {
914 err = -EACCES;
915 goto out2;
918 spin_lock(&addrconf_hash_lock);
920 /* Ignore adding duplicate addresses on an interface */
921 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
922 ADBG("ipv6_add_addr: already assigned\n");
923 err = -EEXIST;
924 goto out;
927 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
929 if (!ifa) {
930 ADBG("ipv6_add_addr: malloc failed\n");
931 err = -ENOBUFS;
932 goto out;
935 rt = addrconf_dst_alloc(idev, addr, false);
936 if (IS_ERR(rt)) {
937 err = PTR_ERR(rt);
938 goto out;
941 neigh_parms_data_state_setall(idev->nd_parms);
943 ifa->addr = *addr;
944 if (peer_addr)
945 ifa->peer_addr = *peer_addr;
947 spin_lock_init(&ifa->lock);
948 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
949 INIT_HLIST_NODE(&ifa->addr_lst);
950 ifa->scope = scope;
951 ifa->prefix_len = pfxlen;
952 ifa->flags = flags | IFA_F_TENTATIVE;
953 ifa->valid_lft = valid_lft;
954 ifa->prefered_lft = prefered_lft;
955 ifa->cstamp = ifa->tstamp = jiffies;
956 ifa->tokenized = false;
958 ifa->rt = rt;
960 ifa->idev = idev;
961 in6_dev_hold(idev);
962 /* For caller */
963 in6_ifa_hold(ifa);
965 /* Add to big hash table */
966 hash = inet6_addr_hash(addr);
968 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
969 spin_unlock(&addrconf_hash_lock);
971 write_lock(&idev->lock);
972 /* Add to inet6_dev unicast addr list. */
973 ipv6_link_dev_addr(idev, ifa);
975 if (ifa->flags&IFA_F_TEMPORARY) {
976 list_add(&ifa->tmp_list, &idev->tempaddr_list);
977 in6_ifa_hold(ifa);
980 in6_ifa_hold(ifa);
981 write_unlock(&idev->lock);
982 out2:
983 rcu_read_unlock_bh();
985 if (likely(err == 0))
986 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
987 else {
988 kfree(ifa);
989 ifa = ERR_PTR(err);
992 return ifa;
993 out:
994 spin_unlock(&addrconf_hash_lock);
995 goto out2;
998 enum cleanup_prefix_rt_t {
999 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1000 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1001 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1005 * Check, whether the prefix for ifp would still need a prefix route
1006 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1007 * constants.
1009 * 1) we don't purge prefix if address was not permanent.
1010 * prefix is managed by its own lifetime.
1011 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1012 * 3) if there are no addresses, delete prefix.
1013 * 4) if there are still other permanent address(es),
1014 * corresponding prefix is still permanent.
1015 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1016 * don't purge the prefix, assume user space is managing it.
1017 * 6) otherwise, update prefix lifetime to the
1018 * longest valid lifetime among the corresponding
1019 * addresses on the device.
1020 * Note: subsequent RA will update lifetime.
1022 static enum cleanup_prefix_rt_t
1023 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1025 struct inet6_ifaddr *ifa;
1026 struct inet6_dev *idev = ifp->idev;
1027 unsigned long lifetime;
1028 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1030 *expires = jiffies;
1032 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1033 if (ifa == ifp)
1034 continue;
1035 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1036 ifp->prefix_len))
1037 continue;
1038 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1039 return CLEANUP_PREFIX_RT_NOP;
1041 action = CLEANUP_PREFIX_RT_EXPIRE;
1043 spin_lock(&ifa->lock);
1045 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1047 * Note: Because this address is
1048 * not permanent, lifetime <
1049 * LONG_MAX / HZ here.
1051 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1052 *expires = ifa->tstamp + lifetime * HZ;
1053 spin_unlock(&ifa->lock);
1056 return action;
1059 static void
1060 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1062 struct rt6_info *rt;
1064 rt = addrconf_get_prefix_route(&ifp->addr,
1065 ifp->prefix_len,
1066 ifp->idev->dev,
1067 0, RTF_GATEWAY | RTF_DEFAULT);
1068 if (rt) {
1069 if (del_rt)
1070 ip6_del_rt(rt);
1071 else {
1072 if (!(rt->rt6i_flags & RTF_EXPIRES))
1073 rt6_set_expires(rt, expires);
1074 ip6_rt_put(rt);
1080 /* This function wants to get referenced ifp and releases it before return */
1082 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1084 int state;
1085 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1086 unsigned long expires;
1088 ASSERT_RTNL();
1090 spin_lock_bh(&ifp->lock);
1091 state = ifp->state;
1092 ifp->state = INET6_IFADDR_STATE_DEAD;
1093 spin_unlock_bh(&ifp->lock);
1095 if (state == INET6_IFADDR_STATE_DEAD)
1096 goto out;
1098 spin_lock_bh(&addrconf_hash_lock);
1099 hlist_del_init_rcu(&ifp->addr_lst);
1100 spin_unlock_bh(&addrconf_hash_lock);
1102 write_lock_bh(&ifp->idev->lock);
1104 if (ifp->flags&IFA_F_TEMPORARY) {
1105 list_del(&ifp->tmp_list);
1106 if (ifp->ifpub) {
1107 in6_ifa_put(ifp->ifpub);
1108 ifp->ifpub = NULL;
1110 __in6_ifa_put(ifp);
1113 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1114 action = check_cleanup_prefix_route(ifp, &expires);
1116 list_del_init(&ifp->if_list);
1117 __in6_ifa_put(ifp);
1119 write_unlock_bh(&ifp->idev->lock);
1121 addrconf_del_dad_work(ifp);
1123 ipv6_ifa_notify(RTM_DELADDR, ifp);
1125 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1127 if (action != CLEANUP_PREFIX_RT_NOP) {
1128 cleanup_prefix_route(ifp, expires,
1129 action == CLEANUP_PREFIX_RT_DEL);
1132 /* clean up prefsrc entries */
1133 rt6_remove_prefsrc(ifp);
1134 out:
1135 in6_ifa_put(ifp);
1138 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1140 struct inet6_dev *idev = ifp->idev;
1141 struct in6_addr addr, *tmpaddr;
1142 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1143 unsigned long regen_advance;
1144 int tmp_plen;
1145 int ret = 0;
1146 u32 addr_flags;
1147 unsigned long now = jiffies;
1149 write_lock_bh(&idev->lock);
1150 if (ift) {
1151 spin_lock_bh(&ift->lock);
1152 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1153 spin_unlock_bh(&ift->lock);
1154 tmpaddr = &addr;
1155 } else {
1156 tmpaddr = NULL;
1158 retry:
1159 in6_dev_hold(idev);
1160 if (idev->cnf.use_tempaddr <= 0) {
1161 write_unlock_bh(&idev->lock);
1162 pr_info("%s: use_tempaddr is disabled\n", __func__);
1163 in6_dev_put(idev);
1164 ret = -1;
1165 goto out;
1167 spin_lock_bh(&ifp->lock);
1168 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1169 idev->cnf.use_tempaddr = -1; /*XXX*/
1170 spin_unlock_bh(&ifp->lock);
1171 write_unlock_bh(&idev->lock);
1172 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1173 __func__);
1174 in6_dev_put(idev);
1175 ret = -1;
1176 goto out;
1178 in6_ifa_hold(ifp);
1179 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1180 __ipv6_try_regen_rndid(idev, tmpaddr);
1181 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1182 age = (now - ifp->tstamp) / HZ;
1183 tmp_valid_lft = min_t(__u32,
1184 ifp->valid_lft,
1185 idev->cnf.temp_valid_lft + age);
1186 tmp_prefered_lft = min_t(__u32,
1187 ifp->prefered_lft,
1188 idev->cnf.temp_prefered_lft + age -
1189 idev->cnf.max_desync_factor);
1190 tmp_plen = ifp->prefix_len;
1191 tmp_tstamp = ifp->tstamp;
1192 spin_unlock_bh(&ifp->lock);
1194 regen_advance = idev->cnf.regen_max_retry *
1195 idev->cnf.dad_transmits *
1196 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1197 write_unlock_bh(&idev->lock);
1199 /* A temporary address is created only if this calculated Preferred
1200 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1201 * an implementation must not create a temporary address with a zero
1202 * Preferred Lifetime.
1203 * Use age calculation as in addrconf_verify to avoid unnecessary
1204 * temporary addresses being generated.
1206 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1207 if (tmp_prefered_lft <= regen_advance + age) {
1208 in6_ifa_put(ifp);
1209 in6_dev_put(idev);
1210 ret = -1;
1211 goto out;
1214 addr_flags = IFA_F_TEMPORARY;
1215 /* set in addrconf_prefix_rcv() */
1216 if (ifp->flags & IFA_F_OPTIMISTIC)
1217 addr_flags |= IFA_F_OPTIMISTIC;
1219 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1220 ipv6_addr_scope(&addr), addr_flags,
1221 tmp_valid_lft, tmp_prefered_lft);
1222 if (IS_ERR(ift)) {
1223 in6_ifa_put(ifp);
1224 in6_dev_put(idev);
1225 pr_info("%s: retry temporary address regeneration\n", __func__);
1226 tmpaddr = &addr;
1227 write_lock_bh(&idev->lock);
1228 goto retry;
1231 spin_lock_bh(&ift->lock);
1232 ift->ifpub = ifp;
1233 ift->cstamp = now;
1234 ift->tstamp = tmp_tstamp;
1235 spin_unlock_bh(&ift->lock);
1237 addrconf_dad_start(ift);
1238 in6_ifa_put(ift);
1239 in6_dev_put(idev);
1240 out:
1241 return ret;
1245 * Choose an appropriate source address (RFC3484)
1247 enum {
1248 IPV6_SADDR_RULE_INIT = 0,
1249 IPV6_SADDR_RULE_LOCAL,
1250 IPV6_SADDR_RULE_SCOPE,
1251 IPV6_SADDR_RULE_PREFERRED,
1252 #ifdef CONFIG_IPV6_MIP6
1253 IPV6_SADDR_RULE_HOA,
1254 #endif
1255 IPV6_SADDR_RULE_OIF,
1256 IPV6_SADDR_RULE_LABEL,
1257 IPV6_SADDR_RULE_PRIVACY,
1258 IPV6_SADDR_RULE_ORCHID,
1259 IPV6_SADDR_RULE_PREFIX,
1260 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1261 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1262 #endif
1263 IPV6_SADDR_RULE_MAX
1266 struct ipv6_saddr_score {
1267 int rule;
1268 int addr_type;
1269 struct inet6_ifaddr *ifa;
1270 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1271 int scopedist;
1272 int matchlen;
1275 struct ipv6_saddr_dst {
1276 const struct in6_addr *addr;
1277 int ifindex;
1278 int scope;
1279 int label;
1280 unsigned int prefs;
1283 static inline int ipv6_saddr_preferred(int type)
1285 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1286 return 1;
1287 return 0;
1290 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1292 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1293 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1294 #else
1295 return false;
1296 #endif
1299 static int ipv6_get_saddr_eval(struct net *net,
1300 struct ipv6_saddr_score *score,
1301 struct ipv6_saddr_dst *dst,
1302 int i)
1304 int ret;
1306 if (i <= score->rule) {
1307 switch (i) {
1308 case IPV6_SADDR_RULE_SCOPE:
1309 ret = score->scopedist;
1310 break;
1311 case IPV6_SADDR_RULE_PREFIX:
1312 ret = score->matchlen;
1313 break;
1314 default:
1315 ret = !!test_bit(i, score->scorebits);
1317 goto out;
1320 switch (i) {
1321 case IPV6_SADDR_RULE_INIT:
1322 /* Rule 0: remember if hiscore is not ready yet */
1323 ret = !!score->ifa;
1324 break;
1325 case IPV6_SADDR_RULE_LOCAL:
1326 /* Rule 1: Prefer same address */
1327 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1328 break;
1329 case IPV6_SADDR_RULE_SCOPE:
1330 /* Rule 2: Prefer appropriate scope
1332 * ret
1334 * -1 | d 15
1335 * ---+--+-+---> scope
1337 * | d is scope of the destination.
1338 * B-d | \
1339 * | \ <- smaller scope is better if
1340 * B-15 | \ if scope is enough for destination.
1341 * | ret = B - scope (-1 <= scope >= d <= 15).
1342 * d-C-1 | /
1343 * |/ <- greater is better
1344 * -C / if scope is not enough for destination.
1345 * /| ret = scope - C (-1 <= d < scope <= 15).
1347 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1348 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1349 * Assume B = 0 and we get C > 29.
1351 ret = __ipv6_addr_src_scope(score->addr_type);
1352 if (ret >= dst->scope)
1353 ret = -ret;
1354 else
1355 ret -= 128; /* 30 is enough */
1356 score->scopedist = ret;
1357 break;
1358 case IPV6_SADDR_RULE_PREFERRED:
1360 /* Rule 3: Avoid deprecated and optimistic addresses */
1361 u8 avoid = IFA_F_DEPRECATED;
1363 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1364 avoid |= IFA_F_OPTIMISTIC;
1365 ret = ipv6_saddr_preferred(score->addr_type) ||
1366 !(score->ifa->flags & avoid);
1367 break;
1369 #ifdef CONFIG_IPV6_MIP6
1370 case IPV6_SADDR_RULE_HOA:
1372 /* Rule 4: Prefer home address */
1373 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1374 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1375 break;
1377 #endif
1378 case IPV6_SADDR_RULE_OIF:
1379 /* Rule 5: Prefer outgoing interface */
1380 ret = (!dst->ifindex ||
1381 dst->ifindex == score->ifa->idev->dev->ifindex);
1382 break;
1383 case IPV6_SADDR_RULE_LABEL:
1384 /* Rule 6: Prefer matching label */
1385 ret = ipv6_addr_label(net,
1386 &score->ifa->addr, score->addr_type,
1387 score->ifa->idev->dev->ifindex) == dst->label;
1388 break;
1389 case IPV6_SADDR_RULE_PRIVACY:
1391 /* Rule 7: Prefer public address
1392 * Note: prefer temporary address if use_tempaddr >= 2
1394 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1395 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1396 score->ifa->idev->cnf.use_tempaddr >= 2;
1397 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1398 break;
1400 case IPV6_SADDR_RULE_ORCHID:
1401 /* Rule 8-: Prefer ORCHID vs ORCHID or
1402 * non-ORCHID vs non-ORCHID
1404 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1405 ipv6_addr_orchid(dst->addr));
1406 break;
1407 case IPV6_SADDR_RULE_PREFIX:
1408 /* Rule 8: Use longest matching prefix */
1409 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1410 if (ret > score->ifa->prefix_len)
1411 ret = score->ifa->prefix_len;
1412 score->matchlen = ret;
1413 break;
1414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1415 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1416 /* Optimistic addresses still have lower precedence than other
1417 * preferred addresses.
1419 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1420 break;
1421 #endif
1422 default:
1423 ret = 0;
1426 if (ret)
1427 __set_bit(i, score->scorebits);
1428 score->rule = i;
1429 out:
1430 return ret;
1433 static int __ipv6_dev_get_saddr(struct net *net,
1434 struct ipv6_saddr_dst *dst,
1435 struct inet6_dev *idev,
1436 struct ipv6_saddr_score *scores,
1437 int hiscore_idx)
1439 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1441 read_lock_bh(&idev->lock);
1442 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1443 int i;
1446 * - Tentative Address (RFC2462 section 5.4)
1447 * - A tentative address is not considered
1448 * "assigned to an interface" in the traditional
1449 * sense, unless it is also flagged as optimistic.
1450 * - Candidate Source Address (section 4)
1451 * - In any case, anycast addresses, multicast
1452 * addresses, and the unspecified address MUST
1453 * NOT be included in a candidate set.
1455 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1456 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1457 continue;
1459 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1461 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1462 score->addr_type & IPV6_ADDR_MULTICAST)) {
1463 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1464 idev->dev->name);
1465 continue;
1468 score->rule = -1;
1469 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1471 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1472 int minihiscore, miniscore;
1474 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1475 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1477 if (minihiscore > miniscore) {
1478 if (i == IPV6_SADDR_RULE_SCOPE &&
1479 score->scopedist > 0) {
1481 * special case:
1482 * each remaining entry
1483 * has too small (not enough)
1484 * scope, because ifa entries
1485 * are sorted by their scope
1486 * values.
1488 goto out;
1490 break;
1491 } else if (minihiscore < miniscore) {
1492 if (hiscore->ifa)
1493 in6_ifa_put(hiscore->ifa);
1495 in6_ifa_hold(score->ifa);
1497 swap(hiscore, score);
1498 hiscore_idx = 1 - hiscore_idx;
1500 /* restore our iterator */
1501 score->ifa = hiscore->ifa;
1503 break;
1507 out:
1508 read_unlock_bh(&idev->lock);
1509 return hiscore_idx;
1512 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1513 const struct in6_addr *daddr, unsigned int prefs,
1514 struct in6_addr *saddr)
1516 struct ipv6_saddr_score scores[2], *hiscore;
1517 struct ipv6_saddr_dst dst;
1518 struct inet6_dev *idev;
1519 struct net_device *dev;
1520 int dst_type;
1521 bool use_oif_addr = false;
1522 int hiscore_idx = 0;
1524 dst_type = __ipv6_addr_type(daddr);
1525 dst.addr = daddr;
1526 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1527 dst.scope = __ipv6_addr_src_scope(dst_type);
1528 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1529 dst.prefs = prefs;
1531 scores[hiscore_idx].rule = -1;
1532 scores[hiscore_idx].ifa = NULL;
1534 rcu_read_lock();
1536 /* Candidate Source Address (section 4)
1537 * - multicast and link-local destination address,
1538 * the set of candidate source address MUST only
1539 * include addresses assigned to interfaces
1540 * belonging to the same link as the outgoing
1541 * interface.
1542 * (- For site-local destination addresses, the
1543 * set of candidate source addresses MUST only
1544 * include addresses assigned to interfaces
1545 * belonging to the same site as the outgoing
1546 * interface.)
1547 * - "It is RECOMMENDED that the candidate source addresses
1548 * be the set of unicast addresses assigned to the
1549 * interface that will be used to send to the destination
1550 * (the 'outgoing' interface)." (RFC 6724)
1552 if (dst_dev) {
1553 idev = __in6_dev_get(dst_dev);
1554 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1555 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1556 (idev && idev->cnf.use_oif_addrs_only)) {
1557 use_oif_addr = true;
1561 if (use_oif_addr) {
1562 if (idev)
1563 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1564 } else {
1565 for_each_netdev_rcu(net, dev) {
1566 idev = __in6_dev_get(dev);
1567 if (!idev)
1568 continue;
1569 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1572 rcu_read_unlock();
1574 hiscore = &scores[hiscore_idx];
1575 if (!hiscore->ifa)
1576 return -EADDRNOTAVAIL;
1578 *saddr = hiscore->ifa->addr;
1579 in6_ifa_put(hiscore->ifa);
1580 return 0;
1582 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1584 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1585 u32 banned_flags)
1587 struct inet6_ifaddr *ifp;
1588 int err = -EADDRNOTAVAIL;
1590 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1591 if (ifp->scope > IFA_LINK)
1592 break;
1593 if (ifp->scope == IFA_LINK &&
1594 !(ifp->flags & banned_flags)) {
1595 *addr = ifp->addr;
1596 err = 0;
1597 break;
1600 return err;
1603 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1604 u32 banned_flags)
1606 struct inet6_dev *idev;
1607 int err = -EADDRNOTAVAIL;
1609 rcu_read_lock();
1610 idev = __in6_dev_get(dev);
1611 if (idev) {
1612 read_lock_bh(&idev->lock);
1613 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1614 read_unlock_bh(&idev->lock);
1616 rcu_read_unlock();
1617 return err;
1620 static int ipv6_count_addresses(struct inet6_dev *idev)
1622 int cnt = 0;
1623 struct inet6_ifaddr *ifp;
1625 read_lock_bh(&idev->lock);
1626 list_for_each_entry(ifp, &idev->addr_list, if_list)
1627 cnt++;
1628 read_unlock_bh(&idev->lock);
1629 return cnt;
1632 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1633 const struct net_device *dev, int strict)
1635 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1637 EXPORT_SYMBOL(ipv6_chk_addr);
1639 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1640 const struct net_device *dev, int strict,
1641 u32 banned_flags)
1643 struct inet6_ifaddr *ifp;
1644 unsigned int hash = inet6_addr_hash(addr);
1645 u32 ifp_flags;
1647 rcu_read_lock_bh();
1648 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1649 if (!net_eq(dev_net(ifp->idev->dev), net))
1650 continue;
1651 /* Decouple optimistic from tentative for evaluation here.
1652 * Ban optimistic addresses explicitly, when required.
1654 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1655 ? (ifp->flags&~IFA_F_TENTATIVE)
1656 : ifp->flags;
1657 if (ipv6_addr_equal(&ifp->addr, addr) &&
1658 !(ifp_flags&banned_flags) &&
1659 (!dev || ifp->idev->dev == dev ||
1660 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1661 rcu_read_unlock_bh();
1662 return 1;
1666 rcu_read_unlock_bh();
1667 return 0;
1669 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1671 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1672 struct net_device *dev)
1674 unsigned int hash = inet6_addr_hash(addr);
1675 struct inet6_ifaddr *ifp;
1677 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1678 if (!net_eq(dev_net(ifp->idev->dev), net))
1679 continue;
1680 if (ipv6_addr_equal(&ifp->addr, addr)) {
1681 if (!dev || ifp->idev->dev == dev)
1682 return true;
1685 return false;
1688 /* Compares an address/prefix_len with addresses on device @dev.
1689 * If one is found it returns true.
1691 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1692 const unsigned int prefix_len, struct net_device *dev)
1694 struct inet6_dev *idev;
1695 struct inet6_ifaddr *ifa;
1696 bool ret = false;
1698 rcu_read_lock();
1699 idev = __in6_dev_get(dev);
1700 if (idev) {
1701 read_lock_bh(&idev->lock);
1702 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1703 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1704 if (ret)
1705 break;
1707 read_unlock_bh(&idev->lock);
1709 rcu_read_unlock();
1711 return ret;
1713 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1715 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1717 struct inet6_dev *idev;
1718 struct inet6_ifaddr *ifa;
1719 int onlink;
1721 onlink = 0;
1722 rcu_read_lock();
1723 idev = __in6_dev_get(dev);
1724 if (idev) {
1725 read_lock_bh(&idev->lock);
1726 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1727 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1728 ifa->prefix_len);
1729 if (onlink)
1730 break;
1732 read_unlock_bh(&idev->lock);
1734 rcu_read_unlock();
1735 return onlink;
1737 EXPORT_SYMBOL(ipv6_chk_prefix);
1739 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1740 struct net_device *dev, int strict)
1742 struct inet6_ifaddr *ifp, *result = NULL;
1743 unsigned int hash = inet6_addr_hash(addr);
1745 rcu_read_lock_bh();
1746 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1747 if (!net_eq(dev_net(ifp->idev->dev), net))
1748 continue;
1749 if (ipv6_addr_equal(&ifp->addr, addr)) {
1750 if (!dev || ifp->idev->dev == dev ||
1751 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1752 result = ifp;
1753 in6_ifa_hold(ifp);
1754 break;
1758 rcu_read_unlock_bh();
1760 return result;
1763 /* Gets referenced address, destroys ifaddr */
1765 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1767 if (ifp->flags&IFA_F_PERMANENT) {
1768 spin_lock_bh(&ifp->lock);
1769 addrconf_del_dad_work(ifp);
1770 ifp->flags |= IFA_F_TENTATIVE;
1771 if (dad_failed)
1772 ifp->flags |= IFA_F_DADFAILED;
1773 spin_unlock_bh(&ifp->lock);
1774 if (dad_failed)
1775 ipv6_ifa_notify(0, ifp);
1776 in6_ifa_put(ifp);
1777 } else if (ifp->flags&IFA_F_TEMPORARY) {
1778 struct inet6_ifaddr *ifpub;
1779 spin_lock_bh(&ifp->lock);
1780 ifpub = ifp->ifpub;
1781 if (ifpub) {
1782 in6_ifa_hold(ifpub);
1783 spin_unlock_bh(&ifp->lock);
1784 ipv6_create_tempaddr(ifpub, ifp);
1785 in6_ifa_put(ifpub);
1786 } else {
1787 spin_unlock_bh(&ifp->lock);
1789 ipv6_del_addr(ifp);
1790 } else {
1791 ipv6_del_addr(ifp);
1795 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1797 int err = -ENOENT;
1799 spin_lock_bh(&ifp->lock);
1800 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1801 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1802 err = 0;
1804 spin_unlock_bh(&ifp->lock);
1806 return err;
1809 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1811 struct in6_addr addr;
1812 struct inet6_dev *idev = ifp->idev;
1813 struct net *net = dev_net(ifp->idev->dev);
1815 if (addrconf_dad_end(ifp)) {
1816 in6_ifa_put(ifp);
1817 return;
1820 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1821 ifp->idev->dev->name, &ifp->addr);
1823 spin_lock_bh(&ifp->lock);
1825 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1826 int scope = ifp->scope;
1827 u32 flags = ifp->flags;
1828 struct in6_addr new_addr;
1829 struct inet6_ifaddr *ifp2;
1830 u32 valid_lft, preferred_lft;
1831 int pfxlen = ifp->prefix_len;
1832 int retries = ifp->stable_privacy_retry + 1;
1834 if (retries > net->ipv6.sysctl.idgen_retries) {
1835 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1836 ifp->idev->dev->name);
1837 goto errdad;
1840 new_addr = ifp->addr;
1841 if (ipv6_generate_stable_address(&new_addr, retries,
1842 idev))
1843 goto errdad;
1845 valid_lft = ifp->valid_lft;
1846 preferred_lft = ifp->prefered_lft;
1848 spin_unlock_bh(&ifp->lock);
1850 if (idev->cnf.max_addresses &&
1851 ipv6_count_addresses(idev) >=
1852 idev->cnf.max_addresses)
1853 goto lock_errdad;
1855 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1856 ifp->idev->dev->name);
1858 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1859 scope, flags, valid_lft,
1860 preferred_lft);
1861 if (IS_ERR(ifp2))
1862 goto lock_errdad;
1864 spin_lock_bh(&ifp2->lock);
1865 ifp2->stable_privacy_retry = retries;
1866 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1867 spin_unlock_bh(&ifp2->lock);
1869 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1870 in6_ifa_put(ifp2);
1871 lock_errdad:
1872 spin_lock_bh(&ifp->lock);
1873 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1874 addr.s6_addr32[0] = htonl(0xfe800000);
1875 addr.s6_addr32[1] = 0;
1877 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1878 ipv6_addr_equal(&ifp->addr, &addr)) {
1879 /* DAD failed for link-local based on MAC address */
1880 idev->cnf.disable_ipv6 = 1;
1882 pr_info("%s: IPv6 being disabled!\n",
1883 ifp->idev->dev->name);
1887 errdad:
1888 /* transition from _POSTDAD to _ERRDAD */
1889 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1890 spin_unlock_bh(&ifp->lock);
1892 addrconf_mod_dad_work(ifp, 0);
1895 /* Join to solicited addr multicast group.
1896 * caller must hold RTNL */
1897 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1899 struct in6_addr maddr;
1901 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1902 return;
1904 addrconf_addr_solict_mult(addr, &maddr);
1905 ipv6_dev_mc_inc(dev, &maddr);
1908 /* caller must hold RTNL */
1909 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1911 struct in6_addr maddr;
1913 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1914 return;
1916 addrconf_addr_solict_mult(addr, &maddr);
1917 __ipv6_dev_mc_dec(idev, &maddr);
1920 /* caller must hold RTNL */
1921 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1923 struct in6_addr addr;
1925 if (ifp->prefix_len >= 127) /* RFC 6164 */
1926 return;
1927 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1928 if (ipv6_addr_any(&addr))
1929 return;
1930 __ipv6_dev_ac_inc(ifp->idev, &addr);
1933 /* caller must hold RTNL */
1934 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1936 struct in6_addr addr;
1938 if (ifp->prefix_len >= 127) /* RFC 6164 */
1939 return;
1940 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1941 if (ipv6_addr_any(&addr))
1942 return;
1943 __ipv6_dev_ac_dec(ifp->idev, &addr);
1946 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1948 if (dev->addr_len != IEEE802154_ADDR_LEN)
1949 return -1;
1950 memcpy(eui, dev->dev_addr, 8);
1951 eui[0] ^= 2;
1952 return 0;
1955 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1957 union fwnet_hwaddr *ha;
1959 if (dev->addr_len != FWNET_ALEN)
1960 return -1;
1962 ha = (union fwnet_hwaddr *)dev->dev_addr;
1964 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1965 eui[0] ^= 2;
1966 return 0;
1969 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1971 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1972 if (dev->addr_len != ARCNET_ALEN)
1973 return -1;
1974 memset(eui, 0, 7);
1975 eui[7] = *(u8 *)dev->dev_addr;
1976 return 0;
1979 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1981 if (dev->addr_len != INFINIBAND_ALEN)
1982 return -1;
1983 memcpy(eui, dev->dev_addr + 12, 8);
1984 eui[0] |= 2;
1985 return 0;
1988 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1990 if (addr == 0)
1991 return -1;
1992 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1993 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1994 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1995 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1996 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1997 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1998 eui[1] = 0;
1999 eui[2] = 0x5E;
2000 eui[3] = 0xFE;
2001 memcpy(eui + 4, &addr, 4);
2002 return 0;
2005 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2007 if (dev->priv_flags & IFF_ISATAP)
2008 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2009 return -1;
2012 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2014 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2017 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2019 memcpy(eui, dev->perm_addr, 3);
2020 memcpy(eui + 5, dev->perm_addr + 3, 3);
2021 eui[3] = 0xFF;
2022 eui[4] = 0xFE;
2023 eui[0] ^= 2;
2024 return 0;
2027 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2029 switch (dev->type) {
2030 case ARPHRD_ETHER:
2031 case ARPHRD_FDDI:
2032 return addrconf_ifid_eui48(eui, dev);
2033 case ARPHRD_ARCNET:
2034 return addrconf_ifid_arcnet(eui, dev);
2035 case ARPHRD_INFINIBAND:
2036 return addrconf_ifid_infiniband(eui, dev);
2037 case ARPHRD_SIT:
2038 return addrconf_ifid_sit(eui, dev);
2039 case ARPHRD_IPGRE:
2040 return addrconf_ifid_gre(eui, dev);
2041 case ARPHRD_6LOWPAN:
2042 case ARPHRD_IEEE802154:
2043 return addrconf_ifid_eui64(eui, dev);
2044 case ARPHRD_IEEE1394:
2045 return addrconf_ifid_ieee1394(eui, dev);
2046 case ARPHRD_TUNNEL6:
2047 return addrconf_ifid_ip6tnl(eui, dev);
2049 return -1;
2052 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2054 int err = -1;
2055 struct inet6_ifaddr *ifp;
2057 read_lock_bh(&idev->lock);
2058 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2059 if (ifp->scope > IFA_LINK)
2060 break;
2061 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2062 memcpy(eui, ifp->addr.s6_addr+8, 8);
2063 err = 0;
2064 break;
2067 read_unlock_bh(&idev->lock);
2068 return err;
2071 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2072 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2074 regen:
2075 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2076 idev->rndid[0] &= ~0x02;
2079 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2080 * check if generated address is not inappropriate
2082 * - Reserved subnet anycast (RFC 2526)
2083 * 11111101 11....11 1xxxxxxx
2084 * - ISATAP (RFC4214) 6.1
2085 * 00-00-5E-FE-xx-xx-xx-xx
2086 * - value 0
2087 * - XXX: already assigned to an address on the device
2089 if (idev->rndid[0] == 0xfd &&
2090 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2091 (idev->rndid[7]&0x80))
2092 goto regen;
2093 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2094 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2095 goto regen;
2096 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2097 goto regen;
2101 static void ipv6_regen_rndid(unsigned long data)
2103 struct inet6_dev *idev = (struct inet6_dev *) data;
2104 unsigned long expires;
2106 rcu_read_lock_bh();
2107 write_lock_bh(&idev->lock);
2109 if (idev->dead)
2110 goto out;
2112 __ipv6_regen_rndid(idev);
2114 expires = jiffies +
2115 idev->cnf.temp_prefered_lft * HZ -
2116 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2117 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2118 idev->cnf.max_desync_factor * HZ;
2119 if (time_before(expires, jiffies)) {
2120 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2121 __func__, idev->dev->name);
2122 goto out;
2125 if (!mod_timer(&idev->regen_timer, expires))
2126 in6_dev_hold(idev);
2128 out:
2129 write_unlock_bh(&idev->lock);
2130 rcu_read_unlock_bh();
2131 in6_dev_put(idev);
2134 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2136 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2137 __ipv6_regen_rndid(idev);
2141 * Add prefix route.
2144 static void
2145 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2146 unsigned long expires, u32 flags)
2148 struct fib6_config cfg = {
2149 .fc_table = RT6_TABLE_PREFIX,
2150 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2151 .fc_ifindex = dev->ifindex,
2152 .fc_expires = expires,
2153 .fc_dst_len = plen,
2154 .fc_flags = RTF_UP | flags,
2155 .fc_nlinfo.nl_net = dev_net(dev),
2156 .fc_protocol = RTPROT_KERNEL,
2159 cfg.fc_dst = *pfx;
2161 /* Prevent useless cloning on PtP SIT.
2162 This thing is done here expecting that the whole
2163 class of non-broadcast devices need not cloning.
2165 #if IS_ENABLED(CONFIG_IPV6_SIT)
2166 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2167 cfg.fc_flags |= RTF_NONEXTHOP;
2168 #endif
2170 ip6_route_add(&cfg);
2174 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2175 int plen,
2176 const struct net_device *dev,
2177 u32 flags, u32 noflags)
2179 struct fib6_node *fn;
2180 struct rt6_info *rt = NULL;
2181 struct fib6_table *table;
2183 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2184 if (!table)
2185 return NULL;
2187 read_lock_bh(&table->tb6_lock);
2188 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2189 if (!fn)
2190 goto out;
2192 noflags |= RTF_CACHE;
2193 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2194 if (rt->dst.dev->ifindex != dev->ifindex)
2195 continue;
2196 if ((rt->rt6i_flags & flags) != flags)
2197 continue;
2198 if ((rt->rt6i_flags & noflags) != 0)
2199 continue;
2200 dst_hold(&rt->dst);
2201 break;
2203 out:
2204 read_unlock_bh(&table->tb6_lock);
2205 return rt;
2209 /* Create "default" multicast route to the interface */
2211 static void addrconf_add_mroute(struct net_device *dev)
2213 struct fib6_config cfg = {
2214 .fc_table = RT6_TABLE_LOCAL,
2215 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2216 .fc_ifindex = dev->ifindex,
2217 .fc_dst_len = 8,
2218 .fc_flags = RTF_UP,
2219 .fc_nlinfo.nl_net = dev_net(dev),
2222 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2224 ip6_route_add(&cfg);
2227 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2229 struct inet6_dev *idev;
2231 ASSERT_RTNL();
2233 idev = ipv6_find_idev(dev);
2234 if (!idev)
2235 return ERR_PTR(-ENOBUFS);
2237 if (idev->cnf.disable_ipv6)
2238 return ERR_PTR(-EACCES);
2240 /* Add default multicast route */
2241 if (!(dev->flags & IFF_LOOPBACK))
2242 addrconf_add_mroute(dev);
2244 return idev;
2247 static void manage_tempaddrs(struct inet6_dev *idev,
2248 struct inet6_ifaddr *ifp,
2249 __u32 valid_lft, __u32 prefered_lft,
2250 bool create, unsigned long now)
2252 u32 flags;
2253 struct inet6_ifaddr *ift;
2255 read_lock_bh(&idev->lock);
2256 /* update all temporary addresses in the list */
2257 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2258 int age, max_valid, max_prefered;
2260 if (ifp != ift->ifpub)
2261 continue;
2263 /* RFC 4941 section 3.3:
2264 * If a received option will extend the lifetime of a public
2265 * address, the lifetimes of temporary addresses should
2266 * be extended, subject to the overall constraint that no
2267 * temporary addresses should ever remain "valid" or "preferred"
2268 * for a time longer than (TEMP_VALID_LIFETIME) or
2269 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2271 age = (now - ift->cstamp) / HZ;
2272 max_valid = idev->cnf.temp_valid_lft - age;
2273 if (max_valid < 0)
2274 max_valid = 0;
2276 max_prefered = idev->cnf.temp_prefered_lft -
2277 idev->cnf.max_desync_factor - age;
2278 if (max_prefered < 0)
2279 max_prefered = 0;
2281 if (valid_lft > max_valid)
2282 valid_lft = max_valid;
2284 if (prefered_lft > max_prefered)
2285 prefered_lft = max_prefered;
2287 spin_lock(&ift->lock);
2288 flags = ift->flags;
2289 ift->valid_lft = valid_lft;
2290 ift->prefered_lft = prefered_lft;
2291 ift->tstamp = now;
2292 if (prefered_lft > 0)
2293 ift->flags &= ~IFA_F_DEPRECATED;
2295 spin_unlock(&ift->lock);
2296 if (!(flags&IFA_F_TENTATIVE))
2297 ipv6_ifa_notify(0, ift);
2300 if ((create || list_empty(&idev->tempaddr_list)) &&
2301 idev->cnf.use_tempaddr > 0) {
2302 /* When a new public address is created as described
2303 * in [ADDRCONF], also create a new temporary address.
2304 * Also create a temporary address if it's enabled but
2305 * no temporary address currently exists.
2307 read_unlock_bh(&idev->lock);
2308 ipv6_create_tempaddr(ifp, NULL);
2309 } else {
2310 read_unlock_bh(&idev->lock);
2314 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2316 struct prefix_info *pinfo;
2317 __u32 valid_lft;
2318 __u32 prefered_lft;
2319 int addr_type;
2320 u32 addr_flags = 0;
2321 struct inet6_dev *in6_dev;
2322 struct net *net = dev_net(dev);
2324 pinfo = (struct prefix_info *) opt;
2326 if (len < sizeof(struct prefix_info)) {
2327 ADBG("addrconf: prefix option too short\n");
2328 return;
2332 * Validation checks ([ADDRCONF], page 19)
2335 addr_type = ipv6_addr_type(&pinfo->prefix);
2337 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2338 return;
2340 valid_lft = ntohl(pinfo->valid);
2341 prefered_lft = ntohl(pinfo->prefered);
2343 if (prefered_lft > valid_lft) {
2344 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2345 return;
2348 in6_dev = in6_dev_get(dev);
2350 if (!in6_dev) {
2351 net_dbg_ratelimited("addrconf: device %s not configured\n",
2352 dev->name);
2353 return;
2357 * Two things going on here:
2358 * 1) Add routes for on-link prefixes
2359 * 2) Configure prefixes with the auto flag set
2362 if (pinfo->onlink) {
2363 struct rt6_info *rt;
2364 unsigned long rt_expires;
2366 /* Avoid arithmetic overflow. Really, we could
2367 * save rt_expires in seconds, likely valid_lft,
2368 * but it would require division in fib gc, that it
2369 * not good.
2371 if (HZ > USER_HZ)
2372 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2373 else
2374 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2376 if (addrconf_finite_timeout(rt_expires))
2377 rt_expires *= HZ;
2379 rt = addrconf_get_prefix_route(&pinfo->prefix,
2380 pinfo->prefix_len,
2381 dev,
2382 RTF_ADDRCONF | RTF_PREFIX_RT,
2383 RTF_GATEWAY | RTF_DEFAULT);
2385 if (rt) {
2386 /* Autoconf prefix route */
2387 if (valid_lft == 0) {
2388 ip6_del_rt(rt);
2389 rt = NULL;
2390 } else if (addrconf_finite_timeout(rt_expires)) {
2391 /* not infinity */
2392 rt6_set_expires(rt, jiffies + rt_expires);
2393 } else {
2394 rt6_clean_expires(rt);
2396 } else if (valid_lft) {
2397 clock_t expires = 0;
2398 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2399 if (addrconf_finite_timeout(rt_expires)) {
2400 /* not infinity */
2401 flags |= RTF_EXPIRES;
2402 expires = jiffies_to_clock_t(rt_expires);
2404 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2405 dev, expires, flags);
2407 ip6_rt_put(rt);
2410 /* Try to figure out our local address for this prefix */
2412 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2413 struct inet6_ifaddr *ifp;
2414 struct in6_addr addr;
2415 int create = 0, update_lft = 0;
2416 bool tokenized = false;
2418 if (pinfo->prefix_len == 64) {
2419 memcpy(&addr, &pinfo->prefix, 8);
2421 if (!ipv6_addr_any(&in6_dev->token)) {
2422 read_lock_bh(&in6_dev->lock);
2423 memcpy(addr.s6_addr + 8,
2424 in6_dev->token.s6_addr + 8, 8);
2425 read_unlock_bh(&in6_dev->lock);
2426 tokenized = true;
2427 } else if (in6_dev->addr_gen_mode ==
2428 IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2429 !ipv6_generate_stable_address(&addr, 0,
2430 in6_dev)) {
2431 addr_flags |= IFA_F_STABLE_PRIVACY;
2432 goto ok;
2433 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2434 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2435 in6_dev_put(in6_dev);
2436 return;
2438 goto ok;
2440 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2441 pinfo->prefix_len);
2442 in6_dev_put(in6_dev);
2443 return;
2447 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2449 if (!ifp && valid_lft) {
2450 int max_addresses = in6_dev->cnf.max_addresses;
2452 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2453 if (in6_dev->cnf.optimistic_dad &&
2454 !net->ipv6.devconf_all->forwarding && sllao)
2455 addr_flags = IFA_F_OPTIMISTIC;
2456 #endif
2458 /* Do not allow to create too much of autoconfigured
2459 * addresses; this would be too easy way to crash kernel.
2461 if (!max_addresses ||
2462 ipv6_count_addresses(in6_dev) < max_addresses)
2463 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2464 pinfo->prefix_len,
2465 addr_type&IPV6_ADDR_SCOPE_MASK,
2466 addr_flags, valid_lft,
2467 prefered_lft);
2469 if (IS_ERR_OR_NULL(ifp)) {
2470 in6_dev_put(in6_dev);
2471 return;
2474 update_lft = 0;
2475 create = 1;
2476 spin_lock_bh(&ifp->lock);
2477 ifp->flags |= IFA_F_MANAGETEMPADDR;
2478 ifp->cstamp = jiffies;
2479 ifp->tokenized = tokenized;
2480 spin_unlock_bh(&ifp->lock);
2481 addrconf_dad_start(ifp);
2484 if (ifp) {
2485 u32 flags;
2486 unsigned long now;
2487 u32 stored_lft;
2489 /* update lifetime (RFC2462 5.5.3 e) */
2490 spin_lock_bh(&ifp->lock);
2491 now = jiffies;
2492 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2493 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2494 else
2495 stored_lft = 0;
2496 if (!update_lft && !create && stored_lft) {
2497 const u32 minimum_lft = min_t(u32,
2498 stored_lft, MIN_VALID_LIFETIME);
2499 valid_lft = max(valid_lft, minimum_lft);
2501 /* RFC4862 Section 5.5.3e:
2502 * "Note that the preferred lifetime of the
2503 * corresponding address is always reset to
2504 * the Preferred Lifetime in the received
2505 * Prefix Information option, regardless of
2506 * whether the valid lifetime is also reset or
2507 * ignored."
2509 * So we should always update prefered_lft here.
2511 update_lft = 1;
2514 if (update_lft) {
2515 ifp->valid_lft = valid_lft;
2516 ifp->prefered_lft = prefered_lft;
2517 ifp->tstamp = now;
2518 flags = ifp->flags;
2519 ifp->flags &= ~IFA_F_DEPRECATED;
2520 spin_unlock_bh(&ifp->lock);
2522 if (!(flags&IFA_F_TENTATIVE))
2523 ipv6_ifa_notify(0, ifp);
2524 } else
2525 spin_unlock_bh(&ifp->lock);
2527 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2528 create, now);
2530 in6_ifa_put(ifp);
2531 addrconf_verify();
2534 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2535 in6_dev_put(in6_dev);
2539 * Set destination address.
2540 * Special case for SIT interfaces where we create a new "virtual"
2541 * device.
2543 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2545 struct in6_ifreq ireq;
2546 struct net_device *dev;
2547 int err = -EINVAL;
2549 rtnl_lock();
2551 err = -EFAULT;
2552 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2553 goto err_exit;
2555 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2557 err = -ENODEV;
2558 if (!dev)
2559 goto err_exit;
2561 #if IS_ENABLED(CONFIG_IPV6_SIT)
2562 if (dev->type == ARPHRD_SIT) {
2563 const struct net_device_ops *ops = dev->netdev_ops;
2564 struct ifreq ifr;
2565 struct ip_tunnel_parm p;
2567 err = -EADDRNOTAVAIL;
2568 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2569 goto err_exit;
2571 memset(&p, 0, sizeof(p));
2572 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2573 p.iph.saddr = 0;
2574 p.iph.version = 4;
2575 p.iph.ihl = 5;
2576 p.iph.protocol = IPPROTO_IPV6;
2577 p.iph.ttl = 64;
2578 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2580 if (ops->ndo_do_ioctl) {
2581 mm_segment_t oldfs = get_fs();
2583 set_fs(KERNEL_DS);
2584 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2585 set_fs(oldfs);
2586 } else
2587 err = -EOPNOTSUPP;
2589 if (err == 0) {
2590 err = -ENOBUFS;
2591 dev = __dev_get_by_name(net, p.name);
2592 if (!dev)
2593 goto err_exit;
2594 err = dev_open(dev);
2597 #endif
2599 err_exit:
2600 rtnl_unlock();
2601 return err;
2604 static int ipv6_mc_config(struct sock *sk, bool join,
2605 const struct in6_addr *addr, int ifindex)
2607 int ret;
2609 ASSERT_RTNL();
2611 lock_sock(sk);
2612 if (join)
2613 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2614 else
2615 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2616 release_sock(sk);
2618 return ret;
2622 * Manual configuration of address on an interface
2624 static int inet6_addr_add(struct net *net, int ifindex,
2625 const struct in6_addr *pfx,
2626 const struct in6_addr *peer_pfx,
2627 unsigned int plen, __u32 ifa_flags,
2628 __u32 prefered_lft, __u32 valid_lft)
2630 struct inet6_ifaddr *ifp;
2631 struct inet6_dev *idev;
2632 struct net_device *dev;
2633 unsigned long timeout;
2634 clock_t expires;
2635 int scope;
2636 u32 flags;
2638 ASSERT_RTNL();
2640 if (plen > 128)
2641 return -EINVAL;
2643 /* check the lifetime */
2644 if (!valid_lft || prefered_lft > valid_lft)
2645 return -EINVAL;
2647 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2648 return -EINVAL;
2650 dev = __dev_get_by_index(net, ifindex);
2651 if (!dev)
2652 return -ENODEV;
2654 idev = addrconf_add_dev(dev);
2655 if (IS_ERR(idev))
2656 return PTR_ERR(idev);
2658 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2659 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2660 true, pfx, ifindex);
2662 if (ret < 0)
2663 return ret;
2666 scope = ipv6_addr_scope(pfx);
2668 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2669 if (addrconf_finite_timeout(timeout)) {
2670 expires = jiffies_to_clock_t(timeout * HZ);
2671 valid_lft = timeout;
2672 flags = RTF_EXPIRES;
2673 } else {
2674 expires = 0;
2675 flags = 0;
2676 ifa_flags |= IFA_F_PERMANENT;
2679 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2680 if (addrconf_finite_timeout(timeout)) {
2681 if (timeout == 0)
2682 ifa_flags |= IFA_F_DEPRECATED;
2683 prefered_lft = timeout;
2686 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2687 valid_lft, prefered_lft);
2689 if (!IS_ERR(ifp)) {
2690 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2691 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2692 expires, flags);
2696 * Note that section 3.1 of RFC 4429 indicates
2697 * that the Optimistic flag should not be set for
2698 * manually configured addresses
2700 addrconf_dad_start(ifp);
2701 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2702 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2703 true, jiffies);
2704 in6_ifa_put(ifp);
2705 addrconf_verify_rtnl();
2706 return 0;
2707 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2708 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2709 false, pfx, ifindex);
2712 return PTR_ERR(ifp);
2715 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2716 const struct in6_addr *pfx, unsigned int plen)
2718 struct inet6_ifaddr *ifp;
2719 struct inet6_dev *idev;
2720 struct net_device *dev;
2722 if (plen > 128)
2723 return -EINVAL;
2725 dev = __dev_get_by_index(net, ifindex);
2726 if (!dev)
2727 return -ENODEV;
2729 idev = __in6_dev_get(dev);
2730 if (!idev)
2731 return -ENXIO;
2733 read_lock_bh(&idev->lock);
2734 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2735 if (ifp->prefix_len == plen &&
2736 ipv6_addr_equal(pfx, &ifp->addr)) {
2737 in6_ifa_hold(ifp);
2738 read_unlock_bh(&idev->lock);
2740 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2741 (ifa_flags & IFA_F_MANAGETEMPADDR))
2742 manage_tempaddrs(idev, ifp, 0, 0, false,
2743 jiffies);
2744 ipv6_del_addr(ifp);
2745 addrconf_verify_rtnl();
2746 if (ipv6_addr_is_multicast(pfx)) {
2747 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2748 false, pfx, dev->ifindex);
2750 return 0;
2753 read_unlock_bh(&idev->lock);
2754 return -EADDRNOTAVAIL;
2758 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2760 struct in6_ifreq ireq;
2761 int err;
2763 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2764 return -EPERM;
2766 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2767 return -EFAULT;
2769 rtnl_lock();
2770 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2771 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2772 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2773 rtnl_unlock();
2774 return err;
2777 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2779 struct in6_ifreq ireq;
2780 int err;
2782 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2783 return -EPERM;
2785 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2786 return -EFAULT;
2788 rtnl_lock();
2789 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2790 ireq.ifr6_prefixlen);
2791 rtnl_unlock();
2792 return err;
2795 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2796 int plen, int scope)
2798 struct inet6_ifaddr *ifp;
2800 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2801 scope, IFA_F_PERMANENT,
2802 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2803 if (!IS_ERR(ifp)) {
2804 spin_lock_bh(&ifp->lock);
2805 ifp->flags &= ~IFA_F_TENTATIVE;
2806 spin_unlock_bh(&ifp->lock);
2807 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2808 in6_ifa_put(ifp);
2812 #if IS_ENABLED(CONFIG_IPV6_SIT)
2813 static void sit_add_v4_addrs(struct inet6_dev *idev)
2815 struct in6_addr addr;
2816 struct net_device *dev;
2817 struct net *net = dev_net(idev->dev);
2818 int scope, plen;
2819 u32 pflags = 0;
2821 ASSERT_RTNL();
2823 memset(&addr, 0, sizeof(struct in6_addr));
2824 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2826 if (idev->dev->flags&IFF_POINTOPOINT) {
2827 addr.s6_addr32[0] = htonl(0xfe800000);
2828 scope = IFA_LINK;
2829 plen = 64;
2830 } else {
2831 scope = IPV6_ADDR_COMPATv4;
2832 plen = 96;
2833 pflags |= RTF_NONEXTHOP;
2836 if (addr.s6_addr32[3]) {
2837 add_addr(idev, &addr, plen, scope);
2838 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2839 return;
2842 for_each_netdev(net, dev) {
2843 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2844 if (in_dev && (dev->flags & IFF_UP)) {
2845 struct in_ifaddr *ifa;
2847 int flag = scope;
2849 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2851 addr.s6_addr32[3] = ifa->ifa_local;
2853 if (ifa->ifa_scope == RT_SCOPE_LINK)
2854 continue;
2855 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2856 if (idev->dev->flags&IFF_POINTOPOINT)
2857 continue;
2858 flag |= IFA_HOST;
2861 add_addr(idev, &addr, plen, flag);
2862 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2863 pflags);
2868 #endif
2870 static void init_loopback(struct net_device *dev)
2872 struct inet6_dev *idev;
2873 struct net_device *sp_dev;
2874 struct inet6_ifaddr *sp_ifa;
2875 struct rt6_info *sp_rt;
2877 /* ::1 */
2879 ASSERT_RTNL();
2881 idev = ipv6_find_idev(dev);
2882 if (!idev) {
2883 pr_debug("%s: add_dev failed\n", __func__);
2884 return;
2887 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2889 /* Add routes to other interface's IPv6 addresses */
2890 for_each_netdev(dev_net(dev), sp_dev) {
2891 if (!strcmp(sp_dev->name, dev->name))
2892 continue;
2894 idev = __in6_dev_get(sp_dev);
2895 if (!idev)
2896 continue;
2898 read_lock_bh(&idev->lock);
2899 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2901 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2902 continue;
2904 if (sp_ifa->rt) {
2905 /* This dst has been added to garbage list when
2906 * lo device down, release this obsolete dst and
2907 * reallocate a new router for ifa.
2909 if (sp_ifa->rt->dst.obsolete > 0) {
2910 ip6_rt_put(sp_ifa->rt);
2911 sp_ifa->rt = NULL;
2912 } else {
2913 continue;
2917 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2919 /* Failure cases are ignored */
2920 if (!IS_ERR(sp_rt)) {
2921 sp_ifa->rt = sp_rt;
2922 ip6_ins_rt(sp_rt);
2925 read_unlock_bh(&idev->lock);
2929 static void addrconf_add_linklocal(struct inet6_dev *idev,
2930 const struct in6_addr *addr, u32 flags)
2932 struct inet6_ifaddr *ifp;
2933 u32 addr_flags = flags | IFA_F_PERMANENT;
2935 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2936 if (idev->cnf.optimistic_dad &&
2937 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2938 addr_flags |= IFA_F_OPTIMISTIC;
2939 #endif
2941 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2942 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2943 if (!IS_ERR(ifp)) {
2944 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2945 addrconf_dad_start(ifp);
2946 in6_ifa_put(ifp);
2950 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2952 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2953 return true;
2955 if (address.s6_addr32[2] == htonl(0x02005eff) &&
2956 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2957 return true;
2959 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2960 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2961 return true;
2963 return false;
2966 static int ipv6_generate_stable_address(struct in6_addr *address,
2967 u8 dad_count,
2968 const struct inet6_dev *idev)
2970 static DEFINE_SPINLOCK(lock);
2971 static __u32 digest[SHA_DIGEST_WORDS];
2972 static __u32 workspace[SHA_WORKSPACE_WORDS];
2974 static union {
2975 char __data[SHA_MESSAGE_BYTES];
2976 struct {
2977 struct in6_addr secret;
2978 __be32 prefix[2];
2979 unsigned char hwaddr[MAX_ADDR_LEN];
2980 u8 dad_count;
2981 } __packed;
2982 } data;
2984 struct in6_addr secret;
2985 struct in6_addr temp;
2986 struct net *net = dev_net(idev->dev);
2988 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2990 if (idev->cnf.stable_secret.initialized)
2991 secret = idev->cnf.stable_secret.secret;
2992 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
2993 secret = net->ipv6.devconf_dflt->stable_secret.secret;
2994 else
2995 return -1;
2997 retry:
2998 spin_lock_bh(&lock);
3000 sha_init(digest);
3001 memset(&data, 0, sizeof(data));
3002 memset(workspace, 0, sizeof(workspace));
3003 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3004 data.prefix[0] = address->s6_addr32[0];
3005 data.prefix[1] = address->s6_addr32[1];
3006 data.secret = secret;
3007 data.dad_count = dad_count;
3009 sha_transform(digest, data.__data, workspace);
3011 temp = *address;
3012 temp.s6_addr32[2] = (__force __be32)digest[0];
3013 temp.s6_addr32[3] = (__force __be32)digest[1];
3015 spin_unlock_bh(&lock);
3017 if (ipv6_reserved_interfaceid(temp)) {
3018 dad_count++;
3019 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3020 return -1;
3021 goto retry;
3024 *address = temp;
3025 return 0;
3028 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3030 struct in6_addr addr;
3032 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3034 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3035 if (!ipv6_generate_stable_address(&addr, 0, idev))
3036 addrconf_add_linklocal(idev, &addr,
3037 IFA_F_STABLE_PRIVACY);
3038 else if (prefix_route)
3039 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3040 } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3041 /* addrconf_add_linklocal also adds a prefix_route and we
3042 * only need to care about prefix routes if ipv6_generate_eui64
3043 * couldn't generate one.
3045 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3046 addrconf_add_linklocal(idev, &addr, 0);
3047 else if (prefix_route)
3048 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3052 static void addrconf_dev_config(struct net_device *dev)
3054 struct inet6_dev *idev;
3056 ASSERT_RTNL();
3058 if ((dev->type != ARPHRD_ETHER) &&
3059 (dev->type != ARPHRD_FDDI) &&
3060 (dev->type != ARPHRD_ARCNET) &&
3061 (dev->type != ARPHRD_INFINIBAND) &&
3062 (dev->type != ARPHRD_IEEE802154) &&
3063 (dev->type != ARPHRD_IEEE1394) &&
3064 (dev->type != ARPHRD_TUNNEL6) &&
3065 (dev->type != ARPHRD_6LOWPAN)) {
3066 /* Alas, we support only Ethernet autoconfiguration. */
3067 return;
3070 idev = addrconf_add_dev(dev);
3071 if (IS_ERR(idev))
3072 return;
3074 addrconf_addr_gen(idev, false);
3077 #if IS_ENABLED(CONFIG_IPV6_SIT)
3078 static void addrconf_sit_config(struct net_device *dev)
3080 struct inet6_dev *idev;
3082 ASSERT_RTNL();
3085 * Configure the tunnel with one of our IPv4
3086 * addresses... we should configure all of
3087 * our v4 addrs in the tunnel
3090 idev = ipv6_find_idev(dev);
3091 if (!idev) {
3092 pr_debug("%s: add_dev failed\n", __func__);
3093 return;
3096 if (dev->priv_flags & IFF_ISATAP) {
3097 addrconf_addr_gen(idev, false);
3098 return;
3101 sit_add_v4_addrs(idev);
3103 if (dev->flags&IFF_POINTOPOINT)
3104 addrconf_add_mroute(dev);
3106 #endif
3108 #if IS_ENABLED(CONFIG_NET_IPGRE)
3109 static void addrconf_gre_config(struct net_device *dev)
3111 struct inet6_dev *idev;
3113 ASSERT_RTNL();
3115 idev = ipv6_find_idev(dev);
3116 if (!idev) {
3117 pr_debug("%s: add_dev failed\n", __func__);
3118 return;
3121 addrconf_addr_gen(idev, true);
3122 if (dev->flags & IFF_POINTOPOINT)
3123 addrconf_add_mroute(dev);
3125 #endif
3127 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3128 void *ptr)
3130 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3131 struct inet6_dev *idev = __in6_dev_get(dev);
3132 int run_pending = 0;
3133 int err;
3135 switch (event) {
3136 case NETDEV_REGISTER:
3137 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3138 idev = ipv6_add_dev(dev);
3139 if (IS_ERR(idev))
3140 return notifier_from_errno(PTR_ERR(idev));
3142 break;
3144 case NETDEV_UP:
3145 case NETDEV_CHANGE:
3146 if (dev->flags & IFF_SLAVE)
3147 break;
3149 if (idev && idev->cnf.disable_ipv6)
3150 break;
3152 if (event == NETDEV_UP) {
3153 if (!addrconf_qdisc_ok(dev)) {
3154 /* device is not ready yet. */
3155 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3156 dev->name);
3157 break;
3160 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3161 idev = ipv6_add_dev(dev);
3163 if (!IS_ERR_OR_NULL(idev)) {
3164 idev->if_flags |= IF_READY;
3165 run_pending = 1;
3167 } else {
3168 if (!addrconf_qdisc_ok(dev)) {
3169 /* device is still not ready. */
3170 break;
3173 if (idev) {
3174 if (idev->if_flags & IF_READY)
3175 /* device is already configured. */
3176 break;
3177 idev->if_flags |= IF_READY;
3180 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3181 dev->name);
3183 run_pending = 1;
3186 switch (dev->type) {
3187 #if IS_ENABLED(CONFIG_IPV6_SIT)
3188 case ARPHRD_SIT:
3189 addrconf_sit_config(dev);
3190 break;
3191 #endif
3192 #if IS_ENABLED(CONFIG_NET_IPGRE)
3193 case ARPHRD_IPGRE:
3194 addrconf_gre_config(dev);
3195 break;
3196 #endif
3197 case ARPHRD_LOOPBACK:
3198 init_loopback(dev);
3199 break;
3201 default:
3202 addrconf_dev_config(dev);
3203 break;
3206 if (!IS_ERR_OR_NULL(idev)) {
3207 if (run_pending)
3208 addrconf_dad_run(idev);
3211 * If the MTU changed during the interface down,
3212 * when the interface up, the changed MTU must be
3213 * reflected in the idev as well as routers.
3215 if (idev->cnf.mtu6 != dev->mtu &&
3216 dev->mtu >= IPV6_MIN_MTU) {
3217 rt6_mtu_change(dev, dev->mtu);
3218 idev->cnf.mtu6 = dev->mtu;
3220 idev->tstamp = jiffies;
3221 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3224 * If the changed mtu during down is lower than
3225 * IPV6_MIN_MTU stop IPv6 on this interface.
3227 if (dev->mtu < IPV6_MIN_MTU)
3228 addrconf_ifdown(dev, 1);
3230 break;
3232 case NETDEV_CHANGEMTU:
3233 if (idev && dev->mtu >= IPV6_MIN_MTU) {
3234 rt6_mtu_change(dev, dev->mtu);
3235 idev->cnf.mtu6 = dev->mtu;
3236 break;
3239 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3240 idev = ipv6_add_dev(dev);
3241 if (!IS_ERR(idev))
3242 break;
3246 * if MTU under IPV6_MIN_MTU.
3247 * Stop IPv6 on this interface.
3250 case NETDEV_DOWN:
3251 case NETDEV_UNREGISTER:
3253 * Remove all addresses from this interface.
3255 addrconf_ifdown(dev, event != NETDEV_DOWN);
3256 break;
3258 case NETDEV_CHANGENAME:
3259 if (idev) {
3260 snmp6_unregister_dev(idev);
3261 addrconf_sysctl_unregister(idev);
3262 err = addrconf_sysctl_register(idev);
3263 if (err)
3264 return notifier_from_errno(err);
3265 err = snmp6_register_dev(idev);
3266 if (err) {
3267 addrconf_sysctl_unregister(idev);
3268 return notifier_from_errno(err);
3271 break;
3273 case NETDEV_PRE_TYPE_CHANGE:
3274 case NETDEV_POST_TYPE_CHANGE:
3275 addrconf_type_change(dev, event);
3276 break;
3279 return NOTIFY_OK;
3283 * addrconf module should be notified of a device going up
3285 static struct notifier_block ipv6_dev_notf = {
3286 .notifier_call = addrconf_notify,
3289 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3291 struct inet6_dev *idev;
3292 ASSERT_RTNL();
3294 idev = __in6_dev_get(dev);
3296 if (event == NETDEV_POST_TYPE_CHANGE)
3297 ipv6_mc_remap(idev);
3298 else if (event == NETDEV_PRE_TYPE_CHANGE)
3299 ipv6_mc_unmap(idev);
3302 static int addrconf_ifdown(struct net_device *dev, int how)
3304 struct net *net = dev_net(dev);
3305 struct inet6_dev *idev;
3306 struct inet6_ifaddr *ifa;
3307 int state, i;
3309 ASSERT_RTNL();
3311 rt6_ifdown(net, dev);
3312 neigh_ifdown(&nd_tbl, dev);
3314 idev = __in6_dev_get(dev);
3315 if (!idev)
3316 return -ENODEV;
3319 * Step 1: remove reference to ipv6 device from parent device.
3320 * Do not dev_put!
3322 if (how) {
3323 idev->dead = 1;
3325 /* protected by rtnl_lock */
3326 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3328 /* Step 1.5: remove snmp6 entry */
3329 snmp6_unregister_dev(idev);
3333 /* Step 2: clear hash table */
3334 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3335 struct hlist_head *h = &inet6_addr_lst[i];
3337 spin_lock_bh(&addrconf_hash_lock);
3338 restart:
3339 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3340 if (ifa->idev == idev) {
3341 hlist_del_init_rcu(&ifa->addr_lst);
3342 addrconf_del_dad_work(ifa);
3343 goto restart;
3346 spin_unlock_bh(&addrconf_hash_lock);
3349 write_lock_bh(&idev->lock);
3351 addrconf_del_rs_timer(idev);
3353 /* Step 2: clear flags for stateless addrconf */
3354 if (!how)
3355 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3357 if (how && del_timer(&idev->regen_timer))
3358 in6_dev_put(idev);
3360 /* Step 3: clear tempaddr list */
3361 while (!list_empty(&idev->tempaddr_list)) {
3362 ifa = list_first_entry(&idev->tempaddr_list,
3363 struct inet6_ifaddr, tmp_list);
3364 list_del(&ifa->tmp_list);
3365 write_unlock_bh(&idev->lock);
3366 spin_lock_bh(&ifa->lock);
3368 if (ifa->ifpub) {
3369 in6_ifa_put(ifa->ifpub);
3370 ifa->ifpub = NULL;
3372 spin_unlock_bh(&ifa->lock);
3373 in6_ifa_put(ifa);
3374 write_lock_bh(&idev->lock);
3377 while (!list_empty(&idev->addr_list)) {
3378 ifa = list_first_entry(&idev->addr_list,
3379 struct inet6_ifaddr, if_list);
3380 addrconf_del_dad_work(ifa);
3382 list_del(&ifa->if_list);
3384 write_unlock_bh(&idev->lock);
3386 spin_lock_bh(&ifa->lock);
3387 state = ifa->state;
3388 ifa->state = INET6_IFADDR_STATE_DEAD;
3389 spin_unlock_bh(&ifa->lock);
3391 if (state != INET6_IFADDR_STATE_DEAD) {
3392 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3393 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3395 in6_ifa_put(ifa);
3397 write_lock_bh(&idev->lock);
3400 write_unlock_bh(&idev->lock);
3402 /* Step 5: Discard anycast and multicast list */
3403 if (how) {
3404 ipv6_ac_destroy_dev(idev);
3405 ipv6_mc_destroy_dev(idev);
3406 } else {
3407 ipv6_mc_down(idev);
3410 idev->tstamp = jiffies;
3412 /* Last: Shot the device (if unregistered) */
3413 if (how) {
3414 addrconf_sysctl_unregister(idev);
3415 neigh_parms_release(&nd_tbl, idev->nd_parms);
3416 neigh_ifdown(&nd_tbl, dev);
3417 in6_dev_put(idev);
3419 return 0;
3422 static void addrconf_rs_timer(unsigned long data)
3424 struct inet6_dev *idev = (struct inet6_dev *)data;
3425 struct net_device *dev = idev->dev;
3426 struct in6_addr lladdr;
3428 write_lock(&idev->lock);
3429 if (idev->dead || !(idev->if_flags & IF_READY))
3430 goto out;
3432 if (!ipv6_accept_ra(idev))
3433 goto out;
3435 /* Announcement received after solicitation was sent */
3436 if (idev->if_flags & IF_RA_RCVD)
3437 goto out;
3439 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3440 write_unlock(&idev->lock);
3441 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3442 ndisc_send_rs(dev, &lladdr,
3443 &in6addr_linklocal_allrouters);
3444 else
3445 goto put;
3447 write_lock(&idev->lock);
3448 /* The wait after the last probe can be shorter */
3449 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3450 idev->cnf.rtr_solicits) ?
3451 idev->cnf.rtr_solicit_delay :
3452 idev->cnf.rtr_solicit_interval);
3453 } else {
3455 * Note: we do not support deprecated "all on-link"
3456 * assumption any longer.
3458 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3461 out:
3462 write_unlock(&idev->lock);
3463 put:
3464 in6_dev_put(idev);
3468 * Duplicate Address Detection
3470 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3472 unsigned long rand_num;
3473 struct inet6_dev *idev = ifp->idev;
3475 if (ifp->flags & IFA_F_OPTIMISTIC)
3476 rand_num = 0;
3477 else
3478 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3480 ifp->dad_probes = idev->cnf.dad_transmits;
3481 addrconf_mod_dad_work(ifp, rand_num);
3484 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3486 struct inet6_dev *idev = ifp->idev;
3487 struct net_device *dev = idev->dev;
3489 addrconf_join_solict(dev, &ifp->addr);
3491 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3493 read_lock_bh(&idev->lock);
3494 spin_lock(&ifp->lock);
3495 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3496 goto out;
3498 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3499 idev->cnf.accept_dad < 1 ||
3500 !(ifp->flags&IFA_F_TENTATIVE) ||
3501 ifp->flags & IFA_F_NODAD) {
3502 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3503 spin_unlock(&ifp->lock);
3504 read_unlock_bh(&idev->lock);
3506 addrconf_dad_completed(ifp);
3507 return;
3510 if (!(idev->if_flags & IF_READY)) {
3511 spin_unlock(&ifp->lock);
3512 read_unlock_bh(&idev->lock);
3514 * If the device is not ready:
3515 * - keep it tentative if it is a permanent address.
3516 * - otherwise, kill it.
3518 in6_ifa_hold(ifp);
3519 addrconf_dad_stop(ifp, 0);
3520 return;
3524 * Optimistic nodes can start receiving
3525 * Frames right away
3527 if (ifp->flags & IFA_F_OPTIMISTIC) {
3528 ip6_ins_rt(ifp->rt);
3529 if (ipv6_use_optimistic_addr(idev)) {
3530 /* Because optimistic nodes can use this address,
3531 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3533 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3537 addrconf_dad_kick(ifp);
3538 out:
3539 spin_unlock(&ifp->lock);
3540 read_unlock_bh(&idev->lock);
3543 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3545 bool begin_dad = false;
3547 spin_lock_bh(&ifp->lock);
3548 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3549 ifp->state = INET6_IFADDR_STATE_PREDAD;
3550 begin_dad = true;
3552 spin_unlock_bh(&ifp->lock);
3554 if (begin_dad)
3555 addrconf_mod_dad_work(ifp, 0);
3558 static void addrconf_dad_work(struct work_struct *w)
3560 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3561 struct inet6_ifaddr,
3562 dad_work);
3563 struct inet6_dev *idev = ifp->idev;
3564 struct in6_addr mcaddr;
3566 enum {
3567 DAD_PROCESS,
3568 DAD_BEGIN,
3569 DAD_ABORT,
3570 } action = DAD_PROCESS;
3572 rtnl_lock();
3574 spin_lock_bh(&ifp->lock);
3575 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3576 action = DAD_BEGIN;
3577 ifp->state = INET6_IFADDR_STATE_DAD;
3578 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3579 action = DAD_ABORT;
3580 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3582 spin_unlock_bh(&ifp->lock);
3584 if (action == DAD_BEGIN) {
3585 addrconf_dad_begin(ifp);
3586 goto out;
3587 } else if (action == DAD_ABORT) {
3588 addrconf_dad_stop(ifp, 1);
3589 goto out;
3592 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3593 goto out;
3595 write_lock_bh(&idev->lock);
3596 if (idev->dead || !(idev->if_flags & IF_READY)) {
3597 write_unlock_bh(&idev->lock);
3598 goto out;
3601 spin_lock(&ifp->lock);
3602 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3603 spin_unlock(&ifp->lock);
3604 write_unlock_bh(&idev->lock);
3605 goto out;
3608 if (ifp->dad_probes == 0) {
3610 * DAD was successful
3613 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3614 spin_unlock(&ifp->lock);
3615 write_unlock_bh(&idev->lock);
3617 addrconf_dad_completed(ifp);
3619 goto out;
3622 ifp->dad_probes--;
3623 addrconf_mod_dad_work(ifp,
3624 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3625 spin_unlock(&ifp->lock);
3626 write_unlock_bh(&idev->lock);
3628 /* send a neighbour solicitation for our addr */
3629 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3630 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any, NULL);
3631 out:
3632 in6_ifa_put(ifp);
3633 rtnl_unlock();
3636 /* ifp->idev must be at least read locked */
3637 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3639 struct inet6_ifaddr *ifpiter;
3640 struct inet6_dev *idev = ifp->idev;
3642 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3643 if (ifpiter->scope > IFA_LINK)
3644 break;
3645 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3646 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3647 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3648 IFA_F_PERMANENT)
3649 return false;
3651 return true;
3654 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3656 struct net_device *dev = ifp->idev->dev;
3657 struct in6_addr lladdr;
3658 bool send_rs, send_mld;
3660 addrconf_del_dad_work(ifp);
3663 * Configure the address for reception. Now it is valid.
3666 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3668 /* If added prefix is link local and we are prepared to process
3669 router advertisements, start sending router solicitations.
3672 read_lock_bh(&ifp->idev->lock);
3673 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3674 send_rs = send_mld &&
3675 ipv6_accept_ra(ifp->idev) &&
3676 ifp->idev->cnf.rtr_solicits > 0 &&
3677 (dev->flags&IFF_LOOPBACK) == 0;
3678 read_unlock_bh(&ifp->idev->lock);
3680 /* While dad is in progress mld report's source address is in6_addrany.
3681 * Resend with proper ll now.
3683 if (send_mld)
3684 ipv6_mc_dad_complete(ifp->idev);
3686 if (send_rs) {
3688 * If a host as already performed a random delay
3689 * [...] as part of DAD [...] there is no need
3690 * to delay again before sending the first RS
3692 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3693 return;
3694 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3696 write_lock_bh(&ifp->idev->lock);
3697 spin_lock(&ifp->lock);
3698 ifp->idev->rs_probes = 1;
3699 ifp->idev->if_flags |= IF_RS_SENT;
3700 addrconf_mod_rs_timer(ifp->idev,
3701 ifp->idev->cnf.rtr_solicit_interval);
3702 spin_unlock(&ifp->lock);
3703 write_unlock_bh(&ifp->idev->lock);
3707 static void addrconf_dad_run(struct inet6_dev *idev)
3709 struct inet6_ifaddr *ifp;
3711 read_lock_bh(&idev->lock);
3712 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3713 spin_lock(&ifp->lock);
3714 if (ifp->flags & IFA_F_TENTATIVE &&
3715 ifp->state == INET6_IFADDR_STATE_DAD)
3716 addrconf_dad_kick(ifp);
3717 spin_unlock(&ifp->lock);
3719 read_unlock_bh(&idev->lock);
3722 #ifdef CONFIG_PROC_FS
3723 struct if6_iter_state {
3724 struct seq_net_private p;
3725 int bucket;
3726 int offset;
3729 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3731 struct inet6_ifaddr *ifa = NULL;
3732 struct if6_iter_state *state = seq->private;
3733 struct net *net = seq_file_net(seq);
3734 int p = 0;
3736 /* initial bucket if pos is 0 */
3737 if (pos == 0) {
3738 state->bucket = 0;
3739 state->offset = 0;
3742 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3743 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3744 addr_lst) {
3745 if (!net_eq(dev_net(ifa->idev->dev), net))
3746 continue;
3747 /* sync with offset */
3748 if (p < state->offset) {
3749 p++;
3750 continue;
3752 state->offset++;
3753 return ifa;
3756 /* prepare for next bucket */
3757 state->offset = 0;
3758 p = 0;
3760 return NULL;
3763 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3764 struct inet6_ifaddr *ifa)
3766 struct if6_iter_state *state = seq->private;
3767 struct net *net = seq_file_net(seq);
3769 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3770 if (!net_eq(dev_net(ifa->idev->dev), net))
3771 continue;
3772 state->offset++;
3773 return ifa;
3776 while (++state->bucket < IN6_ADDR_HSIZE) {
3777 state->offset = 0;
3778 hlist_for_each_entry_rcu_bh(ifa,
3779 &inet6_addr_lst[state->bucket], addr_lst) {
3780 if (!net_eq(dev_net(ifa->idev->dev), net))
3781 continue;
3782 state->offset++;
3783 return ifa;
3787 return NULL;
3790 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3791 __acquires(rcu_bh)
3793 rcu_read_lock_bh();
3794 return if6_get_first(seq, *pos);
3797 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3799 struct inet6_ifaddr *ifa;
3801 ifa = if6_get_next(seq, v);
3802 ++*pos;
3803 return ifa;
3806 static void if6_seq_stop(struct seq_file *seq, void *v)
3807 __releases(rcu_bh)
3809 rcu_read_unlock_bh();
3812 static int if6_seq_show(struct seq_file *seq, void *v)
3814 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3815 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3816 &ifp->addr,
3817 ifp->idev->dev->ifindex,
3818 ifp->prefix_len,
3819 ifp->scope,
3820 (u8) ifp->flags,
3821 ifp->idev->dev->name);
3822 return 0;
3825 static const struct seq_operations if6_seq_ops = {
3826 .start = if6_seq_start,
3827 .next = if6_seq_next,
3828 .show = if6_seq_show,
3829 .stop = if6_seq_stop,
3832 static int if6_seq_open(struct inode *inode, struct file *file)
3834 return seq_open_net(inode, file, &if6_seq_ops,
3835 sizeof(struct if6_iter_state));
3838 static const struct file_operations if6_fops = {
3839 .owner = THIS_MODULE,
3840 .open = if6_seq_open,
3841 .read = seq_read,
3842 .llseek = seq_lseek,
3843 .release = seq_release_net,
3846 static int __net_init if6_proc_net_init(struct net *net)
3848 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3849 return -ENOMEM;
3850 return 0;
3853 static void __net_exit if6_proc_net_exit(struct net *net)
3855 remove_proc_entry("if_inet6", net->proc_net);
3858 static struct pernet_operations if6_proc_net_ops = {
3859 .init = if6_proc_net_init,
3860 .exit = if6_proc_net_exit,
3863 int __init if6_proc_init(void)
3865 return register_pernet_subsys(&if6_proc_net_ops);
3868 void if6_proc_exit(void)
3870 unregister_pernet_subsys(&if6_proc_net_ops);
3872 #endif /* CONFIG_PROC_FS */
3874 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3875 /* Check if address is a home address configured on any interface. */
3876 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3878 int ret = 0;
3879 struct inet6_ifaddr *ifp = NULL;
3880 unsigned int hash = inet6_addr_hash(addr);
3882 rcu_read_lock_bh();
3883 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3884 if (!net_eq(dev_net(ifp->idev->dev), net))
3885 continue;
3886 if (ipv6_addr_equal(&ifp->addr, addr) &&
3887 (ifp->flags & IFA_F_HOMEADDRESS)) {
3888 ret = 1;
3889 break;
3892 rcu_read_unlock_bh();
3893 return ret;
3895 #endif
3898 * Periodic address status verification
3901 static void addrconf_verify_rtnl(void)
3903 unsigned long now, next, next_sec, next_sched;
3904 struct inet6_ifaddr *ifp;
3905 int i;
3907 ASSERT_RTNL();
3909 rcu_read_lock_bh();
3910 now = jiffies;
3911 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3913 cancel_delayed_work(&addr_chk_work);
3915 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3916 restart:
3917 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3918 unsigned long age;
3920 /* When setting preferred_lft to a value not zero or
3921 * infinity, while valid_lft is infinity
3922 * IFA_F_PERMANENT has a non-infinity life time.
3924 if ((ifp->flags & IFA_F_PERMANENT) &&
3925 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3926 continue;
3928 spin_lock(&ifp->lock);
3929 /* We try to batch several events at once. */
3930 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3932 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3933 age >= ifp->valid_lft) {
3934 spin_unlock(&ifp->lock);
3935 in6_ifa_hold(ifp);
3936 ipv6_del_addr(ifp);
3937 goto restart;
3938 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3939 spin_unlock(&ifp->lock);
3940 continue;
3941 } else if (age >= ifp->prefered_lft) {
3942 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3943 int deprecate = 0;
3945 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3946 deprecate = 1;
3947 ifp->flags |= IFA_F_DEPRECATED;
3950 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3951 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3952 next = ifp->tstamp + ifp->valid_lft * HZ;
3954 spin_unlock(&ifp->lock);
3956 if (deprecate) {
3957 in6_ifa_hold(ifp);
3959 ipv6_ifa_notify(0, ifp);
3960 in6_ifa_put(ifp);
3961 goto restart;
3963 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3964 !(ifp->flags&IFA_F_TENTATIVE)) {
3965 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3966 ifp->idev->cnf.dad_transmits *
3967 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3969 if (age >= ifp->prefered_lft - regen_advance) {
3970 struct inet6_ifaddr *ifpub = ifp->ifpub;
3971 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3972 next = ifp->tstamp + ifp->prefered_lft * HZ;
3973 if (!ifp->regen_count && ifpub) {
3974 ifp->regen_count++;
3975 in6_ifa_hold(ifp);
3976 in6_ifa_hold(ifpub);
3977 spin_unlock(&ifp->lock);
3979 spin_lock(&ifpub->lock);
3980 ifpub->regen_count = 0;
3981 spin_unlock(&ifpub->lock);
3982 ipv6_create_tempaddr(ifpub, ifp);
3983 in6_ifa_put(ifpub);
3984 in6_ifa_put(ifp);
3985 goto restart;
3987 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3988 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3989 spin_unlock(&ifp->lock);
3990 } else {
3991 /* ifp->prefered_lft <= ifp->valid_lft */
3992 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3993 next = ifp->tstamp + ifp->prefered_lft * HZ;
3994 spin_unlock(&ifp->lock);
3999 next_sec = round_jiffies_up(next);
4000 next_sched = next;
4002 /* If rounded timeout is accurate enough, accept it. */
4003 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4004 next_sched = next_sec;
4006 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4007 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4008 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4010 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4011 now, next, next_sec, next_sched);
4012 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4013 rcu_read_unlock_bh();
4016 static void addrconf_verify_work(struct work_struct *w)
4018 rtnl_lock();
4019 addrconf_verify_rtnl();
4020 rtnl_unlock();
4023 static void addrconf_verify(void)
4025 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4028 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4029 struct in6_addr **peer_pfx)
4031 struct in6_addr *pfx = NULL;
4033 *peer_pfx = NULL;
4035 if (addr)
4036 pfx = nla_data(addr);
4038 if (local) {
4039 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4040 *peer_pfx = pfx;
4041 pfx = nla_data(local);
4044 return pfx;
4047 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4048 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4049 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4050 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4051 [IFA_FLAGS] = { .len = sizeof(u32) },
4054 static int
4055 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4057 struct net *net = sock_net(skb->sk);
4058 struct ifaddrmsg *ifm;
4059 struct nlattr *tb[IFA_MAX+1];
4060 struct in6_addr *pfx, *peer_pfx;
4061 u32 ifa_flags;
4062 int err;
4064 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4065 if (err < 0)
4066 return err;
4068 ifm = nlmsg_data(nlh);
4069 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4070 if (!pfx)
4071 return -EINVAL;
4073 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4075 /* We ignore other flags so far. */
4076 ifa_flags &= IFA_F_MANAGETEMPADDR;
4078 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4079 ifm->ifa_prefixlen);
4082 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4083 u32 prefered_lft, u32 valid_lft)
4085 u32 flags;
4086 clock_t expires;
4087 unsigned long timeout;
4088 bool was_managetempaddr;
4089 bool had_prefixroute;
4091 ASSERT_RTNL();
4093 if (!valid_lft || (prefered_lft > valid_lft))
4094 return -EINVAL;
4096 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4097 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4098 return -EINVAL;
4100 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4101 if (addrconf_finite_timeout(timeout)) {
4102 expires = jiffies_to_clock_t(timeout * HZ);
4103 valid_lft = timeout;
4104 flags = RTF_EXPIRES;
4105 } else {
4106 expires = 0;
4107 flags = 0;
4108 ifa_flags |= IFA_F_PERMANENT;
4111 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4112 if (addrconf_finite_timeout(timeout)) {
4113 if (timeout == 0)
4114 ifa_flags |= IFA_F_DEPRECATED;
4115 prefered_lft = timeout;
4118 spin_lock_bh(&ifp->lock);
4119 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4120 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4121 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4122 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4123 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4124 IFA_F_NOPREFIXROUTE);
4125 ifp->flags |= ifa_flags;
4126 ifp->tstamp = jiffies;
4127 ifp->valid_lft = valid_lft;
4128 ifp->prefered_lft = prefered_lft;
4130 spin_unlock_bh(&ifp->lock);
4131 if (!(ifp->flags&IFA_F_TENTATIVE))
4132 ipv6_ifa_notify(0, ifp);
4134 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4135 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4136 expires, flags);
4137 } else if (had_prefixroute) {
4138 enum cleanup_prefix_rt_t action;
4139 unsigned long rt_expires;
4141 write_lock_bh(&ifp->idev->lock);
4142 action = check_cleanup_prefix_route(ifp, &rt_expires);
4143 write_unlock_bh(&ifp->idev->lock);
4145 if (action != CLEANUP_PREFIX_RT_NOP) {
4146 cleanup_prefix_route(ifp, rt_expires,
4147 action == CLEANUP_PREFIX_RT_DEL);
4151 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4152 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4153 valid_lft = prefered_lft = 0;
4154 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4155 !was_managetempaddr, jiffies);
4158 addrconf_verify_rtnl();
4160 return 0;
4163 static int
4164 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4166 struct net *net = sock_net(skb->sk);
4167 struct ifaddrmsg *ifm;
4168 struct nlattr *tb[IFA_MAX+1];
4169 struct in6_addr *pfx, *peer_pfx;
4170 struct inet6_ifaddr *ifa;
4171 struct net_device *dev;
4172 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4173 u32 ifa_flags;
4174 int err;
4176 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4177 if (err < 0)
4178 return err;
4180 ifm = nlmsg_data(nlh);
4181 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4182 if (!pfx)
4183 return -EINVAL;
4185 if (tb[IFA_CACHEINFO]) {
4186 struct ifa_cacheinfo *ci;
4188 ci = nla_data(tb[IFA_CACHEINFO]);
4189 valid_lft = ci->ifa_valid;
4190 preferred_lft = ci->ifa_prefered;
4191 } else {
4192 preferred_lft = INFINITY_LIFE_TIME;
4193 valid_lft = INFINITY_LIFE_TIME;
4196 dev = __dev_get_by_index(net, ifm->ifa_index);
4197 if (!dev)
4198 return -ENODEV;
4200 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4202 /* We ignore other flags so far. */
4203 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4204 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4206 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4207 if (!ifa) {
4209 * It would be best to check for !NLM_F_CREATE here but
4210 * userspace already relies on not having to provide this.
4212 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4213 ifm->ifa_prefixlen, ifa_flags,
4214 preferred_lft, valid_lft);
4217 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4218 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4219 err = -EEXIST;
4220 else
4221 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4223 in6_ifa_put(ifa);
4225 return err;
4228 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4229 u8 scope, int ifindex)
4231 struct ifaddrmsg *ifm;
4233 ifm = nlmsg_data(nlh);
4234 ifm->ifa_family = AF_INET6;
4235 ifm->ifa_prefixlen = prefixlen;
4236 ifm->ifa_flags = flags;
4237 ifm->ifa_scope = scope;
4238 ifm->ifa_index = ifindex;
4241 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4242 unsigned long tstamp, u32 preferred, u32 valid)
4244 struct ifa_cacheinfo ci;
4246 ci.cstamp = cstamp_delta(cstamp);
4247 ci.tstamp = cstamp_delta(tstamp);
4248 ci.ifa_prefered = preferred;
4249 ci.ifa_valid = valid;
4251 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4254 static inline int rt_scope(int ifa_scope)
4256 if (ifa_scope & IFA_HOST)
4257 return RT_SCOPE_HOST;
4258 else if (ifa_scope & IFA_LINK)
4259 return RT_SCOPE_LINK;
4260 else if (ifa_scope & IFA_SITE)
4261 return RT_SCOPE_SITE;
4262 else
4263 return RT_SCOPE_UNIVERSE;
4266 static inline int inet6_ifaddr_msgsize(void)
4268 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4269 + nla_total_size(16) /* IFA_LOCAL */
4270 + nla_total_size(16) /* IFA_ADDRESS */
4271 + nla_total_size(sizeof(struct ifa_cacheinfo))
4272 + nla_total_size(4) /* IFA_FLAGS */;
4275 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4276 u32 portid, u32 seq, int event, unsigned int flags)
4278 struct nlmsghdr *nlh;
4279 u32 preferred, valid;
4281 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4282 if (!nlh)
4283 return -EMSGSIZE;
4285 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4286 ifa->idev->dev->ifindex);
4288 if (!((ifa->flags&IFA_F_PERMANENT) &&
4289 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4290 preferred = ifa->prefered_lft;
4291 valid = ifa->valid_lft;
4292 if (preferred != INFINITY_LIFE_TIME) {
4293 long tval = (jiffies - ifa->tstamp)/HZ;
4294 if (preferred > tval)
4295 preferred -= tval;
4296 else
4297 preferred = 0;
4298 if (valid != INFINITY_LIFE_TIME) {
4299 if (valid > tval)
4300 valid -= tval;
4301 else
4302 valid = 0;
4305 } else {
4306 preferred = INFINITY_LIFE_TIME;
4307 valid = INFINITY_LIFE_TIME;
4310 if (!ipv6_addr_any(&ifa->peer_addr)) {
4311 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4312 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4313 goto error;
4314 } else
4315 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4316 goto error;
4318 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4319 goto error;
4321 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4322 goto error;
4324 nlmsg_end(skb, nlh);
4325 return 0;
4327 error:
4328 nlmsg_cancel(skb, nlh);
4329 return -EMSGSIZE;
4332 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4333 u32 portid, u32 seq, int event, u16 flags)
4335 struct nlmsghdr *nlh;
4336 u8 scope = RT_SCOPE_UNIVERSE;
4337 int ifindex = ifmca->idev->dev->ifindex;
4339 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4340 scope = RT_SCOPE_SITE;
4342 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4343 if (!nlh)
4344 return -EMSGSIZE;
4346 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4347 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4348 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4349 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4350 nlmsg_cancel(skb, nlh);
4351 return -EMSGSIZE;
4354 nlmsg_end(skb, nlh);
4355 return 0;
4358 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4359 u32 portid, u32 seq, int event, unsigned int flags)
4361 struct nlmsghdr *nlh;
4362 u8 scope = RT_SCOPE_UNIVERSE;
4363 int ifindex = ifaca->aca_idev->dev->ifindex;
4365 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4366 scope = RT_SCOPE_SITE;
4368 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4369 if (!nlh)
4370 return -EMSGSIZE;
4372 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4373 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4374 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4375 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4376 nlmsg_cancel(skb, nlh);
4377 return -EMSGSIZE;
4380 nlmsg_end(skb, nlh);
4381 return 0;
4384 enum addr_type_t {
4385 UNICAST_ADDR,
4386 MULTICAST_ADDR,
4387 ANYCAST_ADDR,
4390 /* called with rcu_read_lock() */
4391 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4392 struct netlink_callback *cb, enum addr_type_t type,
4393 int s_ip_idx, int *p_ip_idx)
4395 struct ifmcaddr6 *ifmca;
4396 struct ifacaddr6 *ifaca;
4397 int err = 1;
4398 int ip_idx = *p_ip_idx;
4400 read_lock_bh(&idev->lock);
4401 switch (type) {
4402 case UNICAST_ADDR: {
4403 struct inet6_ifaddr *ifa;
4405 /* unicast address incl. temp addr */
4406 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4407 if (++ip_idx < s_ip_idx)
4408 continue;
4409 err = inet6_fill_ifaddr(skb, ifa,
4410 NETLINK_CB(cb->skb).portid,
4411 cb->nlh->nlmsg_seq,
4412 RTM_NEWADDR,
4413 NLM_F_MULTI);
4414 if (err < 0)
4415 break;
4416 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4418 break;
4420 case MULTICAST_ADDR:
4421 /* multicast address */
4422 for (ifmca = idev->mc_list; ifmca;
4423 ifmca = ifmca->next, ip_idx++) {
4424 if (ip_idx < s_ip_idx)
4425 continue;
4426 err = inet6_fill_ifmcaddr(skb, ifmca,
4427 NETLINK_CB(cb->skb).portid,
4428 cb->nlh->nlmsg_seq,
4429 RTM_GETMULTICAST,
4430 NLM_F_MULTI);
4431 if (err < 0)
4432 break;
4434 break;
4435 case ANYCAST_ADDR:
4436 /* anycast address */
4437 for (ifaca = idev->ac_list; ifaca;
4438 ifaca = ifaca->aca_next, ip_idx++) {
4439 if (ip_idx < s_ip_idx)
4440 continue;
4441 err = inet6_fill_ifacaddr(skb, ifaca,
4442 NETLINK_CB(cb->skb).portid,
4443 cb->nlh->nlmsg_seq,
4444 RTM_GETANYCAST,
4445 NLM_F_MULTI);
4446 if (err < 0)
4447 break;
4449 break;
4450 default:
4451 break;
4453 read_unlock_bh(&idev->lock);
4454 *p_ip_idx = ip_idx;
4455 return err;
4458 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4459 enum addr_type_t type)
4461 struct net *net = sock_net(skb->sk);
4462 int h, s_h;
4463 int idx, ip_idx;
4464 int s_idx, s_ip_idx;
4465 struct net_device *dev;
4466 struct inet6_dev *idev;
4467 struct hlist_head *head;
4469 s_h = cb->args[0];
4470 s_idx = idx = cb->args[1];
4471 s_ip_idx = ip_idx = cb->args[2];
4473 rcu_read_lock();
4474 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4475 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4476 idx = 0;
4477 head = &net->dev_index_head[h];
4478 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4479 if (idx < s_idx)
4480 goto cont;
4481 if (h > s_h || idx > s_idx)
4482 s_ip_idx = 0;
4483 ip_idx = 0;
4484 idev = __in6_dev_get(dev);
4485 if (!idev)
4486 goto cont;
4488 if (in6_dump_addrs(idev, skb, cb, type,
4489 s_ip_idx, &ip_idx) < 0)
4490 goto done;
4491 cont:
4492 idx++;
4495 done:
4496 rcu_read_unlock();
4497 cb->args[0] = h;
4498 cb->args[1] = idx;
4499 cb->args[2] = ip_idx;
4501 return skb->len;
4504 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4506 enum addr_type_t type = UNICAST_ADDR;
4508 return inet6_dump_addr(skb, cb, type);
4511 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4513 enum addr_type_t type = MULTICAST_ADDR;
4515 return inet6_dump_addr(skb, cb, type);
4519 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4521 enum addr_type_t type = ANYCAST_ADDR;
4523 return inet6_dump_addr(skb, cb, type);
4526 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4528 struct net *net = sock_net(in_skb->sk);
4529 struct ifaddrmsg *ifm;
4530 struct nlattr *tb[IFA_MAX+1];
4531 struct in6_addr *addr = NULL, *peer;
4532 struct net_device *dev = NULL;
4533 struct inet6_ifaddr *ifa;
4534 struct sk_buff *skb;
4535 int err;
4537 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4538 if (err < 0)
4539 goto errout;
4541 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4542 if (!addr) {
4543 err = -EINVAL;
4544 goto errout;
4547 ifm = nlmsg_data(nlh);
4548 if (ifm->ifa_index)
4549 dev = __dev_get_by_index(net, ifm->ifa_index);
4551 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4552 if (!ifa) {
4553 err = -EADDRNOTAVAIL;
4554 goto errout;
4557 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4558 if (!skb) {
4559 err = -ENOBUFS;
4560 goto errout_ifa;
4563 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4564 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4565 if (err < 0) {
4566 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4567 WARN_ON(err == -EMSGSIZE);
4568 kfree_skb(skb);
4569 goto errout_ifa;
4571 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4572 errout_ifa:
4573 in6_ifa_put(ifa);
4574 errout:
4575 return err;
4578 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4580 struct sk_buff *skb;
4581 struct net *net = dev_net(ifa->idev->dev);
4582 int err = -ENOBUFS;
4584 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4585 if (!skb)
4586 goto errout;
4588 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4589 if (err < 0) {
4590 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4591 WARN_ON(err == -EMSGSIZE);
4592 kfree_skb(skb);
4593 goto errout;
4595 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4596 return;
4597 errout:
4598 if (err < 0)
4599 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4602 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4603 __s32 *array, int bytes)
4605 BUG_ON(bytes < (DEVCONF_MAX * 4));
4607 memset(array, 0, bytes);
4608 array[DEVCONF_FORWARDING] = cnf->forwarding;
4609 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4610 array[DEVCONF_MTU6] = cnf->mtu6;
4611 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4612 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4613 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4614 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4615 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4616 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4617 jiffies_to_msecs(cnf->rtr_solicit_interval);
4618 array[DEVCONF_RTR_SOLICIT_DELAY] =
4619 jiffies_to_msecs(cnf->rtr_solicit_delay);
4620 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4621 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4622 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4623 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4624 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4625 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4626 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4627 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4628 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4629 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4630 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4631 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4632 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4633 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4634 #ifdef CONFIG_IPV6_ROUTER_PREF
4635 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4636 array[DEVCONF_RTR_PROBE_INTERVAL] =
4637 jiffies_to_msecs(cnf->rtr_probe_interval);
4638 #ifdef CONFIG_IPV6_ROUTE_INFO
4639 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4640 #endif
4641 #endif
4642 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4643 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4644 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4645 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4646 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4647 #endif
4648 #ifdef CONFIG_IPV6_MROUTE
4649 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4650 #endif
4651 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4652 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4653 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4654 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4655 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4656 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4657 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4658 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4659 /* we omit DEVCONF_STABLE_SECRET for now */
4660 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4663 static inline size_t inet6_ifla6_size(void)
4665 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4666 + nla_total_size(sizeof(struct ifla_cacheinfo))
4667 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4668 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4669 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4670 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4673 static inline size_t inet6_if_nlmsg_size(void)
4675 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4676 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4677 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4678 + nla_total_size(4) /* IFLA_MTU */
4679 + nla_total_size(4) /* IFLA_LINK */
4680 + nla_total_size(1) /* IFLA_OPERSTATE */
4681 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4684 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4685 int items, int bytes)
4687 int i;
4688 int pad = bytes - sizeof(u64) * items;
4689 BUG_ON(pad < 0);
4691 /* Use put_unaligned() because stats may not be aligned for u64. */
4692 put_unaligned(items, &stats[0]);
4693 for (i = 1; i < items; i++)
4694 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4696 memset(&stats[items], 0, pad);
4699 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4700 int bytes, size_t syncpoff)
4702 int i, c;
4703 u64 buff[IPSTATS_MIB_MAX];
4704 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4706 BUG_ON(pad < 0);
4708 memset(buff, 0, sizeof(buff));
4709 buff[0] = IPSTATS_MIB_MAX;
4711 for_each_possible_cpu(c) {
4712 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4713 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4716 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4717 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4720 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4721 int bytes)
4723 switch (attrtype) {
4724 case IFLA_INET6_STATS:
4725 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4726 offsetof(struct ipstats_mib, syncp));
4727 break;
4728 case IFLA_INET6_ICMP6STATS:
4729 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4730 break;
4734 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4736 struct nlattr *nla;
4737 struct ifla_cacheinfo ci;
4739 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4740 goto nla_put_failure;
4741 ci.max_reasm_len = IPV6_MAXPLEN;
4742 ci.tstamp = cstamp_delta(idev->tstamp);
4743 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4744 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4745 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4746 goto nla_put_failure;
4747 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4748 if (!nla)
4749 goto nla_put_failure;
4750 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4752 /* XXX - MC not implemented */
4754 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4755 if (!nla)
4756 goto nla_put_failure;
4757 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4759 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4760 if (!nla)
4761 goto nla_put_failure;
4762 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4764 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4765 if (!nla)
4766 goto nla_put_failure;
4768 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4769 goto nla_put_failure;
4771 read_lock_bh(&idev->lock);
4772 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4773 read_unlock_bh(&idev->lock);
4775 return 0;
4777 nla_put_failure:
4778 return -EMSGSIZE;
4781 static size_t inet6_get_link_af_size(const struct net_device *dev)
4783 if (!__in6_dev_get(dev))
4784 return 0;
4786 return inet6_ifla6_size();
4789 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4791 struct inet6_dev *idev = __in6_dev_get(dev);
4793 if (!idev)
4794 return -ENODATA;
4796 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4797 return -EMSGSIZE;
4799 return 0;
4802 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4804 struct inet6_ifaddr *ifp;
4805 struct net_device *dev = idev->dev;
4806 bool update_rs = false;
4807 struct in6_addr ll_addr;
4809 ASSERT_RTNL();
4811 if (!token)
4812 return -EINVAL;
4813 if (ipv6_addr_any(token))
4814 return -EINVAL;
4815 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4816 return -EINVAL;
4817 if (!ipv6_accept_ra(idev))
4818 return -EINVAL;
4819 if (idev->cnf.rtr_solicits <= 0)
4820 return -EINVAL;
4822 write_lock_bh(&idev->lock);
4824 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4825 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4827 write_unlock_bh(&idev->lock);
4829 if (!idev->dead && (idev->if_flags & IF_READY) &&
4830 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4831 IFA_F_OPTIMISTIC)) {
4833 /* If we're not ready, then normal ifup will take care
4834 * of this. Otherwise, we need to request our rs here.
4836 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4837 update_rs = true;
4840 write_lock_bh(&idev->lock);
4842 if (update_rs) {
4843 idev->if_flags |= IF_RS_SENT;
4844 idev->rs_probes = 1;
4845 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4848 /* Well, that's kinda nasty ... */
4849 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4850 spin_lock(&ifp->lock);
4851 if (ifp->tokenized) {
4852 ifp->valid_lft = 0;
4853 ifp->prefered_lft = 0;
4855 spin_unlock(&ifp->lock);
4858 write_unlock_bh(&idev->lock);
4859 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4860 addrconf_verify_rtnl();
4861 return 0;
4864 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4865 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
4866 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
4869 static int inet6_validate_link_af(const struct net_device *dev,
4870 const struct nlattr *nla)
4872 struct nlattr *tb[IFLA_INET6_MAX + 1];
4874 if (dev && !__in6_dev_get(dev))
4875 return -EAFNOSUPPORT;
4877 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4880 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4882 int err = -EINVAL;
4883 struct inet6_dev *idev = __in6_dev_get(dev);
4884 struct nlattr *tb[IFLA_INET6_MAX + 1];
4886 if (!idev)
4887 return -EAFNOSUPPORT;
4889 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4890 BUG();
4892 if (tb[IFLA_INET6_TOKEN]) {
4893 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4894 if (err)
4895 return err;
4898 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4899 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4901 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4902 mode != IN6_ADDR_GEN_MODE_NONE &&
4903 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4904 return -EINVAL;
4906 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4907 !idev->cnf.stable_secret.initialized &&
4908 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4909 return -EINVAL;
4911 idev->addr_gen_mode = mode;
4912 err = 0;
4915 return err;
4918 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4919 u32 portid, u32 seq, int event, unsigned int flags)
4921 struct net_device *dev = idev->dev;
4922 struct ifinfomsg *hdr;
4923 struct nlmsghdr *nlh;
4924 void *protoinfo;
4926 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4927 if (!nlh)
4928 return -EMSGSIZE;
4930 hdr = nlmsg_data(nlh);
4931 hdr->ifi_family = AF_INET6;
4932 hdr->__ifi_pad = 0;
4933 hdr->ifi_type = dev->type;
4934 hdr->ifi_index = dev->ifindex;
4935 hdr->ifi_flags = dev_get_flags(dev);
4936 hdr->ifi_change = 0;
4938 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4939 (dev->addr_len &&
4940 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4941 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4942 (dev->ifindex != dev_get_iflink(dev) &&
4943 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
4944 nla_put_u8(skb, IFLA_OPERSTATE,
4945 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
4946 goto nla_put_failure;
4947 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4948 if (!protoinfo)
4949 goto nla_put_failure;
4951 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4952 goto nla_put_failure;
4954 nla_nest_end(skb, protoinfo);
4955 nlmsg_end(skb, nlh);
4956 return 0;
4958 nla_put_failure:
4959 nlmsg_cancel(skb, nlh);
4960 return -EMSGSIZE;
4963 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4965 struct net *net = sock_net(skb->sk);
4966 int h, s_h;
4967 int idx = 0, s_idx;
4968 struct net_device *dev;
4969 struct inet6_dev *idev;
4970 struct hlist_head *head;
4972 s_h = cb->args[0];
4973 s_idx = cb->args[1];
4975 rcu_read_lock();
4976 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4977 idx = 0;
4978 head = &net->dev_index_head[h];
4979 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4980 if (idx < s_idx)
4981 goto cont;
4982 idev = __in6_dev_get(dev);
4983 if (!idev)
4984 goto cont;
4985 if (inet6_fill_ifinfo(skb, idev,
4986 NETLINK_CB(cb->skb).portid,
4987 cb->nlh->nlmsg_seq,
4988 RTM_NEWLINK, NLM_F_MULTI) < 0)
4989 goto out;
4990 cont:
4991 idx++;
4994 out:
4995 rcu_read_unlock();
4996 cb->args[1] = idx;
4997 cb->args[0] = h;
4999 return skb->len;
5002 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5004 struct sk_buff *skb;
5005 struct net *net = dev_net(idev->dev);
5006 int err = -ENOBUFS;
5008 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5009 if (!skb)
5010 goto errout;
5012 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5013 if (err < 0) {
5014 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5015 WARN_ON(err == -EMSGSIZE);
5016 kfree_skb(skb);
5017 goto errout;
5019 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5020 return;
5021 errout:
5022 if (err < 0)
5023 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5026 static inline size_t inet6_prefix_nlmsg_size(void)
5028 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5029 + nla_total_size(sizeof(struct in6_addr))
5030 + nla_total_size(sizeof(struct prefix_cacheinfo));
5033 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5034 struct prefix_info *pinfo, u32 portid, u32 seq,
5035 int event, unsigned int flags)
5037 struct prefixmsg *pmsg;
5038 struct nlmsghdr *nlh;
5039 struct prefix_cacheinfo ci;
5041 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5042 if (!nlh)
5043 return -EMSGSIZE;
5045 pmsg = nlmsg_data(nlh);
5046 pmsg->prefix_family = AF_INET6;
5047 pmsg->prefix_pad1 = 0;
5048 pmsg->prefix_pad2 = 0;
5049 pmsg->prefix_ifindex = idev->dev->ifindex;
5050 pmsg->prefix_len = pinfo->prefix_len;
5051 pmsg->prefix_type = pinfo->type;
5052 pmsg->prefix_pad3 = 0;
5053 pmsg->prefix_flags = 0;
5054 if (pinfo->onlink)
5055 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5056 if (pinfo->autoconf)
5057 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5059 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5060 goto nla_put_failure;
5061 ci.preferred_time = ntohl(pinfo->prefered);
5062 ci.valid_time = ntohl(pinfo->valid);
5063 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5064 goto nla_put_failure;
5065 nlmsg_end(skb, nlh);
5066 return 0;
5068 nla_put_failure:
5069 nlmsg_cancel(skb, nlh);
5070 return -EMSGSIZE;
5073 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5074 struct prefix_info *pinfo)
5076 struct sk_buff *skb;
5077 struct net *net = dev_net(idev->dev);
5078 int err = -ENOBUFS;
5080 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5081 if (!skb)
5082 goto errout;
5084 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5085 if (err < 0) {
5086 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5087 WARN_ON(err == -EMSGSIZE);
5088 kfree_skb(skb);
5089 goto errout;
5091 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5092 return;
5093 errout:
5094 if (err < 0)
5095 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5098 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5100 struct net *net = dev_net(ifp->idev->dev);
5102 if (event)
5103 ASSERT_RTNL();
5105 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5107 switch (event) {
5108 case RTM_NEWADDR:
5110 * If the address was optimistic
5111 * we inserted the route at the start of
5112 * our DAD process, so we don't need
5113 * to do it again
5115 if (!(ifp->rt->rt6i_node))
5116 ip6_ins_rt(ifp->rt);
5117 if (ifp->idev->cnf.forwarding)
5118 addrconf_join_anycast(ifp);
5119 if (!ipv6_addr_any(&ifp->peer_addr))
5120 addrconf_prefix_route(&ifp->peer_addr, 128,
5121 ifp->idev->dev, 0, 0);
5122 break;
5123 case RTM_DELADDR:
5124 if (ifp->idev->cnf.forwarding)
5125 addrconf_leave_anycast(ifp);
5126 addrconf_leave_solict(ifp->idev, &ifp->addr);
5127 if (!ipv6_addr_any(&ifp->peer_addr)) {
5128 struct rt6_info *rt;
5130 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5131 ifp->idev->dev, 0, 0);
5132 if (rt)
5133 ip6_del_rt(rt);
5135 dst_hold(&ifp->rt->dst);
5137 ip6_del_rt(ifp->rt);
5139 rt_genid_bump_ipv6(net);
5140 break;
5142 atomic_inc(&net->ipv6.dev_addr_genid);
5145 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5147 rcu_read_lock_bh();
5148 if (likely(ifp->idev->dead == 0))
5149 __ipv6_ifa_notify(event, ifp);
5150 rcu_read_unlock_bh();
5153 #ifdef CONFIG_SYSCTL
5155 static
5156 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5157 void __user *buffer, size_t *lenp, loff_t *ppos)
5159 int *valp = ctl->data;
5160 int val = *valp;
5161 loff_t pos = *ppos;
5162 struct ctl_table lctl;
5163 int ret;
5166 * ctl->data points to idev->cnf.forwarding, we should
5167 * not modify it until we get the rtnl lock.
5169 lctl = *ctl;
5170 lctl.data = &val;
5172 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5174 if (write)
5175 ret = addrconf_fixup_forwarding(ctl, valp, val);
5176 if (ret)
5177 *ppos = pos;
5178 return ret;
5181 static
5182 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5183 void __user *buffer, size_t *lenp, loff_t *ppos)
5185 struct inet6_dev *idev = ctl->extra1;
5186 int min_mtu = IPV6_MIN_MTU;
5187 struct ctl_table lctl;
5189 lctl = *ctl;
5190 lctl.extra1 = &min_mtu;
5191 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5193 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5196 static void dev_disable_change(struct inet6_dev *idev)
5198 struct netdev_notifier_info info;
5200 if (!idev || !idev->dev)
5201 return;
5203 netdev_notifier_info_init(&info, idev->dev);
5204 if (idev->cnf.disable_ipv6)
5205 addrconf_notify(NULL, NETDEV_DOWN, &info);
5206 else
5207 addrconf_notify(NULL, NETDEV_UP, &info);
5210 static void addrconf_disable_change(struct net *net, __s32 newf)
5212 struct net_device *dev;
5213 struct inet6_dev *idev;
5215 rcu_read_lock();
5216 for_each_netdev_rcu(net, dev) {
5217 idev = __in6_dev_get(dev);
5218 if (idev) {
5219 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5220 idev->cnf.disable_ipv6 = newf;
5221 if (changed)
5222 dev_disable_change(idev);
5225 rcu_read_unlock();
5228 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5230 struct net *net;
5231 int old;
5233 if (!rtnl_trylock())
5234 return restart_syscall();
5236 net = (struct net *)table->extra2;
5237 old = *p;
5238 *p = newf;
5240 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5241 rtnl_unlock();
5242 return 0;
5245 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5246 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5247 addrconf_disable_change(net, newf);
5248 } else if ((!newf) ^ (!old))
5249 dev_disable_change((struct inet6_dev *)table->extra1);
5251 rtnl_unlock();
5252 return 0;
5255 static
5256 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5257 void __user *buffer, size_t *lenp, loff_t *ppos)
5259 int *valp = ctl->data;
5260 int val = *valp;
5261 loff_t pos = *ppos;
5262 struct ctl_table lctl;
5263 int ret;
5266 * ctl->data points to idev->cnf.disable_ipv6, we should
5267 * not modify it until we get the rtnl lock.
5269 lctl = *ctl;
5270 lctl.data = &val;
5272 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5274 if (write)
5275 ret = addrconf_disable_ipv6(ctl, valp, val);
5276 if (ret)
5277 *ppos = pos;
5278 return ret;
5281 static
5282 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5283 void __user *buffer, size_t *lenp, loff_t *ppos)
5285 int *valp = ctl->data;
5286 int ret;
5287 int old, new;
5289 old = *valp;
5290 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5291 new = *valp;
5293 if (write && old != new) {
5294 struct net *net = ctl->extra2;
5296 if (!rtnl_trylock())
5297 return restart_syscall();
5299 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5300 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5301 NETCONFA_IFINDEX_DEFAULT,
5302 net->ipv6.devconf_dflt);
5303 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5304 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5305 NETCONFA_IFINDEX_ALL,
5306 net->ipv6.devconf_all);
5307 else {
5308 struct inet6_dev *idev = ctl->extra1;
5310 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5311 idev->dev->ifindex,
5312 &idev->cnf);
5314 rtnl_unlock();
5317 return ret;
5320 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5321 void __user *buffer, size_t *lenp,
5322 loff_t *ppos)
5324 int err;
5325 struct in6_addr addr;
5326 char str[IPV6_MAX_STRLEN];
5327 struct ctl_table lctl = *ctl;
5328 struct net *net = ctl->extra2;
5329 struct ipv6_stable_secret *secret = ctl->data;
5331 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5332 return -EIO;
5334 lctl.maxlen = IPV6_MAX_STRLEN;
5335 lctl.data = str;
5337 if (!rtnl_trylock())
5338 return restart_syscall();
5340 if (!write && !secret->initialized) {
5341 err = -EIO;
5342 goto out;
5345 if (!write) {
5346 err = snprintf(str, sizeof(str), "%pI6",
5347 &secret->secret);
5348 if (err >= sizeof(str)) {
5349 err = -EIO;
5350 goto out;
5354 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5355 if (err || !write)
5356 goto out;
5358 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5359 err = -EIO;
5360 goto out;
5363 secret->initialized = true;
5364 secret->secret = addr;
5366 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5367 struct net_device *dev;
5369 for_each_netdev(net, dev) {
5370 struct inet6_dev *idev = __in6_dev_get(dev);
5372 if (idev) {
5373 idev->addr_gen_mode =
5374 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5377 } else {
5378 struct inet6_dev *idev = ctl->extra1;
5380 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5383 out:
5384 rtnl_unlock();
5386 return err;
5389 static
5390 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5391 int write,
5392 void __user *buffer,
5393 size_t *lenp,
5394 loff_t *ppos)
5396 int *valp = ctl->data;
5397 int val = *valp;
5398 loff_t pos = *ppos;
5399 struct ctl_table lctl;
5400 int ret;
5402 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5403 * we should not modify it until we get the rtnl lock.
5405 lctl = *ctl;
5406 lctl.data = &val;
5408 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5410 if (write)
5411 ret = addrconf_fixup_linkdown(ctl, valp, val);
5412 if (ret)
5413 *ppos = pos;
5414 return ret;
5417 static struct addrconf_sysctl_table
5419 struct ctl_table_header *sysctl_header;
5420 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5421 } addrconf_sysctl __read_mostly = {
5422 .sysctl_header = NULL,
5423 .addrconf_vars = {
5425 .procname = "forwarding",
5426 .data = &ipv6_devconf.forwarding,
5427 .maxlen = sizeof(int),
5428 .mode = 0644,
5429 .proc_handler = addrconf_sysctl_forward,
5432 .procname = "hop_limit",
5433 .data = &ipv6_devconf.hop_limit,
5434 .maxlen = sizeof(int),
5435 .mode = 0644,
5436 .proc_handler = proc_dointvec,
5439 .procname = "mtu",
5440 .data = &ipv6_devconf.mtu6,
5441 .maxlen = sizeof(int),
5442 .mode = 0644,
5443 .proc_handler = addrconf_sysctl_mtu,
5446 .procname = "accept_ra",
5447 .data = &ipv6_devconf.accept_ra,
5448 .maxlen = sizeof(int),
5449 .mode = 0644,
5450 .proc_handler = proc_dointvec,
5453 .procname = "accept_redirects",
5454 .data = &ipv6_devconf.accept_redirects,
5455 .maxlen = sizeof(int),
5456 .mode = 0644,
5457 .proc_handler = proc_dointvec,
5460 .procname = "autoconf",
5461 .data = &ipv6_devconf.autoconf,
5462 .maxlen = sizeof(int),
5463 .mode = 0644,
5464 .proc_handler = proc_dointvec,
5467 .procname = "dad_transmits",
5468 .data = &ipv6_devconf.dad_transmits,
5469 .maxlen = sizeof(int),
5470 .mode = 0644,
5471 .proc_handler = proc_dointvec,
5474 .procname = "router_solicitations",
5475 .data = &ipv6_devconf.rtr_solicits,
5476 .maxlen = sizeof(int),
5477 .mode = 0644,
5478 .proc_handler = proc_dointvec,
5481 .procname = "router_solicitation_interval",
5482 .data = &ipv6_devconf.rtr_solicit_interval,
5483 .maxlen = sizeof(int),
5484 .mode = 0644,
5485 .proc_handler = proc_dointvec_jiffies,
5488 .procname = "router_solicitation_delay",
5489 .data = &ipv6_devconf.rtr_solicit_delay,
5490 .maxlen = sizeof(int),
5491 .mode = 0644,
5492 .proc_handler = proc_dointvec_jiffies,
5495 .procname = "force_mld_version",
5496 .data = &ipv6_devconf.force_mld_version,
5497 .maxlen = sizeof(int),
5498 .mode = 0644,
5499 .proc_handler = proc_dointvec,
5502 .procname = "mldv1_unsolicited_report_interval",
5503 .data =
5504 &ipv6_devconf.mldv1_unsolicited_report_interval,
5505 .maxlen = sizeof(int),
5506 .mode = 0644,
5507 .proc_handler = proc_dointvec_ms_jiffies,
5510 .procname = "mldv2_unsolicited_report_interval",
5511 .data =
5512 &ipv6_devconf.mldv2_unsolicited_report_interval,
5513 .maxlen = sizeof(int),
5514 .mode = 0644,
5515 .proc_handler = proc_dointvec_ms_jiffies,
5518 .procname = "use_tempaddr",
5519 .data = &ipv6_devconf.use_tempaddr,
5520 .maxlen = sizeof(int),
5521 .mode = 0644,
5522 .proc_handler = proc_dointvec,
5525 .procname = "temp_valid_lft",
5526 .data = &ipv6_devconf.temp_valid_lft,
5527 .maxlen = sizeof(int),
5528 .mode = 0644,
5529 .proc_handler = proc_dointvec,
5532 .procname = "temp_prefered_lft",
5533 .data = &ipv6_devconf.temp_prefered_lft,
5534 .maxlen = sizeof(int),
5535 .mode = 0644,
5536 .proc_handler = proc_dointvec,
5539 .procname = "regen_max_retry",
5540 .data = &ipv6_devconf.regen_max_retry,
5541 .maxlen = sizeof(int),
5542 .mode = 0644,
5543 .proc_handler = proc_dointvec,
5546 .procname = "max_desync_factor",
5547 .data = &ipv6_devconf.max_desync_factor,
5548 .maxlen = sizeof(int),
5549 .mode = 0644,
5550 .proc_handler = proc_dointvec,
5553 .procname = "max_addresses",
5554 .data = &ipv6_devconf.max_addresses,
5555 .maxlen = sizeof(int),
5556 .mode = 0644,
5557 .proc_handler = proc_dointvec,
5560 .procname = "accept_ra_defrtr",
5561 .data = &ipv6_devconf.accept_ra_defrtr,
5562 .maxlen = sizeof(int),
5563 .mode = 0644,
5564 .proc_handler = proc_dointvec,
5567 .procname = "accept_ra_min_hop_limit",
5568 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5569 .maxlen = sizeof(int),
5570 .mode = 0644,
5571 .proc_handler = proc_dointvec,
5574 .procname = "accept_ra_pinfo",
5575 .data = &ipv6_devconf.accept_ra_pinfo,
5576 .maxlen = sizeof(int),
5577 .mode = 0644,
5578 .proc_handler = proc_dointvec,
5580 #ifdef CONFIG_IPV6_ROUTER_PREF
5582 .procname = "accept_ra_rtr_pref",
5583 .data = &ipv6_devconf.accept_ra_rtr_pref,
5584 .maxlen = sizeof(int),
5585 .mode = 0644,
5586 .proc_handler = proc_dointvec,
5589 .procname = "router_probe_interval",
5590 .data = &ipv6_devconf.rtr_probe_interval,
5591 .maxlen = sizeof(int),
5592 .mode = 0644,
5593 .proc_handler = proc_dointvec_jiffies,
5595 #ifdef CONFIG_IPV6_ROUTE_INFO
5597 .procname = "accept_ra_rt_info_max_plen",
5598 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5599 .maxlen = sizeof(int),
5600 .mode = 0644,
5601 .proc_handler = proc_dointvec,
5603 #endif
5604 #endif
5606 .procname = "proxy_ndp",
5607 .data = &ipv6_devconf.proxy_ndp,
5608 .maxlen = sizeof(int),
5609 .mode = 0644,
5610 .proc_handler = addrconf_sysctl_proxy_ndp,
5613 .procname = "accept_source_route",
5614 .data = &ipv6_devconf.accept_source_route,
5615 .maxlen = sizeof(int),
5616 .mode = 0644,
5617 .proc_handler = proc_dointvec,
5619 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5621 .procname = "optimistic_dad",
5622 .data = &ipv6_devconf.optimistic_dad,
5623 .maxlen = sizeof(int),
5624 .mode = 0644,
5625 .proc_handler = proc_dointvec,
5629 .procname = "use_optimistic",
5630 .data = &ipv6_devconf.use_optimistic,
5631 .maxlen = sizeof(int),
5632 .mode = 0644,
5633 .proc_handler = proc_dointvec,
5636 #endif
5637 #ifdef CONFIG_IPV6_MROUTE
5639 .procname = "mc_forwarding",
5640 .data = &ipv6_devconf.mc_forwarding,
5641 .maxlen = sizeof(int),
5642 .mode = 0444,
5643 .proc_handler = proc_dointvec,
5645 #endif
5647 .procname = "disable_ipv6",
5648 .data = &ipv6_devconf.disable_ipv6,
5649 .maxlen = sizeof(int),
5650 .mode = 0644,
5651 .proc_handler = addrconf_sysctl_disable,
5654 .procname = "accept_dad",
5655 .data = &ipv6_devconf.accept_dad,
5656 .maxlen = sizeof(int),
5657 .mode = 0644,
5658 .proc_handler = proc_dointvec,
5661 .procname = "force_tllao",
5662 .data = &ipv6_devconf.force_tllao,
5663 .maxlen = sizeof(int),
5664 .mode = 0644,
5665 .proc_handler = proc_dointvec
5668 .procname = "ndisc_notify",
5669 .data = &ipv6_devconf.ndisc_notify,
5670 .maxlen = sizeof(int),
5671 .mode = 0644,
5672 .proc_handler = proc_dointvec
5675 .procname = "suppress_frag_ndisc",
5676 .data = &ipv6_devconf.suppress_frag_ndisc,
5677 .maxlen = sizeof(int),
5678 .mode = 0644,
5679 .proc_handler = proc_dointvec
5682 .procname = "accept_ra_from_local",
5683 .data = &ipv6_devconf.accept_ra_from_local,
5684 .maxlen = sizeof(int),
5685 .mode = 0644,
5686 .proc_handler = proc_dointvec,
5689 .procname = "accept_ra_mtu",
5690 .data = &ipv6_devconf.accept_ra_mtu,
5691 .maxlen = sizeof(int),
5692 .mode = 0644,
5693 .proc_handler = proc_dointvec,
5696 .procname = "stable_secret",
5697 .data = &ipv6_devconf.stable_secret,
5698 .maxlen = IPV6_MAX_STRLEN,
5699 .mode = 0600,
5700 .proc_handler = addrconf_sysctl_stable_secret,
5703 .procname = "use_oif_addrs_only",
5704 .data = &ipv6_devconf.use_oif_addrs_only,
5705 .maxlen = sizeof(int),
5706 .mode = 0644,
5707 .proc_handler = proc_dointvec,
5710 .procname = "ignore_routes_with_linkdown",
5711 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5712 .maxlen = sizeof(int),
5713 .mode = 0644,
5714 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5717 /* sentinel */
5722 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5723 struct inet6_dev *idev, struct ipv6_devconf *p)
5725 int i;
5726 struct addrconf_sysctl_table *t;
5727 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5729 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5730 if (!t)
5731 goto out;
5733 for (i = 0; t->addrconf_vars[i].data; i++) {
5734 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5735 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5736 t->addrconf_vars[i].extra2 = net;
5739 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5741 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5742 if (!t->sysctl_header)
5743 goto free;
5745 p->sysctl = t;
5746 return 0;
5748 free:
5749 kfree(t);
5750 out:
5751 return -ENOBUFS;
5754 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5756 struct addrconf_sysctl_table *t;
5758 if (!p->sysctl)
5759 return;
5761 t = p->sysctl;
5762 p->sysctl = NULL;
5763 unregister_net_sysctl_table(t->sysctl_header);
5764 kfree(t);
5767 static int addrconf_sysctl_register(struct inet6_dev *idev)
5769 int err;
5771 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5772 return -EINVAL;
5774 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5775 &ndisc_ifinfo_sysctl_change);
5776 if (err)
5777 return err;
5778 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5779 idev, &idev->cnf);
5780 if (err)
5781 neigh_sysctl_unregister(idev->nd_parms);
5783 return err;
5786 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5788 __addrconf_sysctl_unregister(&idev->cnf);
5789 neigh_sysctl_unregister(idev->nd_parms);
5793 #endif
5795 static int __net_init addrconf_init_net(struct net *net)
5797 int err = -ENOMEM;
5798 struct ipv6_devconf *all, *dflt;
5800 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5801 if (!all)
5802 goto err_alloc_all;
5804 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5805 if (!dflt)
5806 goto err_alloc_dflt;
5808 /* these will be inherited by all namespaces */
5809 dflt->autoconf = ipv6_defaults.autoconf;
5810 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5812 dflt->stable_secret.initialized = false;
5813 all->stable_secret.initialized = false;
5815 net->ipv6.devconf_all = all;
5816 net->ipv6.devconf_dflt = dflt;
5818 #ifdef CONFIG_SYSCTL
5819 err = __addrconf_sysctl_register(net, "all", NULL, all);
5820 if (err < 0)
5821 goto err_reg_all;
5823 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5824 if (err < 0)
5825 goto err_reg_dflt;
5826 #endif
5827 return 0;
5829 #ifdef CONFIG_SYSCTL
5830 err_reg_dflt:
5831 __addrconf_sysctl_unregister(all);
5832 err_reg_all:
5833 kfree(dflt);
5834 #endif
5835 err_alloc_dflt:
5836 kfree(all);
5837 err_alloc_all:
5838 return err;
5841 static void __net_exit addrconf_exit_net(struct net *net)
5843 #ifdef CONFIG_SYSCTL
5844 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5845 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5846 #endif
5847 kfree(net->ipv6.devconf_dflt);
5848 kfree(net->ipv6.devconf_all);
5851 static struct pernet_operations addrconf_ops = {
5852 .init = addrconf_init_net,
5853 .exit = addrconf_exit_net,
5856 static struct rtnl_af_ops inet6_ops __read_mostly = {
5857 .family = AF_INET6,
5858 .fill_link_af = inet6_fill_link_af,
5859 .get_link_af_size = inet6_get_link_af_size,
5860 .validate_link_af = inet6_validate_link_af,
5861 .set_link_af = inet6_set_link_af,
5865 * Init / cleanup code
5868 int __init addrconf_init(void)
5870 struct inet6_dev *idev;
5871 int i, err;
5873 err = ipv6_addr_label_init();
5874 if (err < 0) {
5875 pr_crit("%s: cannot initialize default policy table: %d\n",
5876 __func__, err);
5877 goto out;
5880 err = register_pernet_subsys(&addrconf_ops);
5881 if (err < 0)
5882 goto out_addrlabel;
5884 addrconf_wq = create_workqueue("ipv6_addrconf");
5885 if (!addrconf_wq) {
5886 err = -ENOMEM;
5887 goto out_nowq;
5890 /* The addrconf netdev notifier requires that loopback_dev
5891 * has it's ipv6 private information allocated and setup
5892 * before it can bring up and give link-local addresses
5893 * to other devices which are up.
5895 * Unfortunately, loopback_dev is not necessarily the first
5896 * entry in the global dev_base list of net devices. In fact,
5897 * it is likely to be the very last entry on that list.
5898 * So this causes the notifier registry below to try and
5899 * give link-local addresses to all devices besides loopback_dev
5900 * first, then loopback_dev, which cases all the non-loopback_dev
5901 * devices to fail to get a link-local address.
5903 * So, as a temporary fix, allocate the ipv6 structure for
5904 * loopback_dev first by hand.
5905 * Longer term, all of the dependencies ipv6 has upon the loopback
5906 * device and it being up should be removed.
5908 rtnl_lock();
5909 idev = ipv6_add_dev(init_net.loopback_dev);
5910 rtnl_unlock();
5911 if (IS_ERR(idev)) {
5912 err = PTR_ERR(idev);
5913 goto errlo;
5916 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5917 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5919 register_netdevice_notifier(&ipv6_dev_notf);
5921 addrconf_verify();
5923 rtnl_af_register(&inet6_ops);
5925 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5926 NULL);
5927 if (err < 0)
5928 goto errout;
5930 /* Only the first call to __rtnl_register can fail */
5931 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5932 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5933 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5934 inet6_dump_ifaddr, NULL);
5935 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5936 inet6_dump_ifmcaddr, NULL);
5937 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5938 inet6_dump_ifacaddr, NULL);
5939 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5940 inet6_netconf_dump_devconf, NULL);
5942 ipv6_addr_label_rtnl_register();
5944 return 0;
5945 errout:
5946 rtnl_af_unregister(&inet6_ops);
5947 unregister_netdevice_notifier(&ipv6_dev_notf);
5948 errlo:
5949 destroy_workqueue(addrconf_wq);
5950 out_nowq:
5951 unregister_pernet_subsys(&addrconf_ops);
5952 out_addrlabel:
5953 ipv6_addr_label_cleanup();
5954 out:
5955 return err;
5958 void addrconf_cleanup(void)
5960 struct net_device *dev;
5961 int i;
5963 unregister_netdevice_notifier(&ipv6_dev_notf);
5964 unregister_pernet_subsys(&addrconf_ops);
5965 ipv6_addr_label_cleanup();
5967 rtnl_lock();
5969 __rtnl_af_unregister(&inet6_ops);
5971 /* clean dev list */
5972 for_each_netdev(&init_net, dev) {
5973 if (__in6_dev_get(dev) == NULL)
5974 continue;
5975 addrconf_ifdown(dev, 1);
5977 addrconf_ifdown(init_net.loopback_dev, 2);
5980 * Check hash table.
5982 spin_lock_bh(&addrconf_hash_lock);
5983 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5984 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5985 spin_unlock_bh(&addrconf_hash_lock);
5986 cancel_delayed_work(&addr_chk_work);
5987 rtnl_unlock();
5989 destroy_workqueue(addrconf_wq);