x86/mm: Improve switch_mm() barrier comments
[linux-2.6/btrfs-unstable.git] / net / ipv6 / addrconf.c
blob1f21087accab258e5be6cc104f8b8ced13fedd6d
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 <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
99 #if ACONF_DEBUG >= 3
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101 #else
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #endif
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 #define IPV6_MAX_STRLEN \
108 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
110 static inline u32 cstamp_delta(unsigned long cstamp)
112 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
115 #ifdef CONFIG_SYSCTL
116 static int addrconf_sysctl_register(struct inet6_dev *idev);
117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118 #else
119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
121 return 0;
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
127 #endif
129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131 static void ipv6_regen_rndid(unsigned long data);
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135 static int ipv6_generate_stable_address(struct in6_addr *addr,
136 u8 dad_count,
137 const struct inet6_dev *idev);
140 * Configured unicast address hash table
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
145 static void addrconf_verify(void);
146 static void addrconf_verify_rtnl(void);
147 static void addrconf_verify_work(struct work_struct *);
149 static struct workqueue_struct *addrconf_wq;
150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
155 static void addrconf_type_change(struct net_device *dev,
156 unsigned long event);
157 static int addrconf_ifdown(struct net_device *dev, int how);
159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
160 int plen,
161 const struct net_device *dev,
162 u32 flags, u32 noflags);
164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165 static void addrconf_dad_work(struct work_struct *w);
166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167 static void addrconf_dad_run(struct inet6_dev *idev);
168 static void addrconf_rs_timer(unsigned long data);
169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173 struct prefix_info *pinfo);
174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175 struct net_device *dev);
177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
178 .forwarding = 0,
179 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
180 .mtu6 = IPV6_MIN_MTU,
181 .accept_ra = 1,
182 .accept_redirects = 1,
183 .autoconf = 1,
184 .force_mld_version = 0,
185 .mldv1_unsolicited_report_interval = 10 * HZ,
186 .mldv2_unsolicited_report_interval = HZ,
187 .dad_transmits = 1,
188 .rtr_solicits = MAX_RTR_SOLICITATIONS,
189 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
190 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
191 .use_tempaddr = 0,
192 .temp_valid_lft = TEMP_VALID_LIFETIME,
193 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
194 .regen_max_retry = REGEN_MAX_RETRY,
195 .max_desync_factor = MAX_DESYNC_FACTOR,
196 .max_addresses = IPV6_MAX_ADDRESSES,
197 .accept_ra_defrtr = 1,
198 .accept_ra_from_local = 0,
199 .accept_ra_min_hop_limit= 1,
200 .accept_ra_pinfo = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202 .accept_ra_rtr_pref = 1,
203 .rtr_probe_interval = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205 .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208 .proxy_ndp = 0,
209 .accept_source_route = 0, /* we do not accept RH0 by default. */
210 .disable_ipv6 = 0,
211 .accept_dad = 1,
212 .suppress_frag_ndisc = 1,
213 .accept_ra_mtu = 1,
214 .stable_secret = {
215 .initialized = false,
217 .use_oif_addrs_only = 0,
218 .ignore_routes_with_linkdown = 0,
221 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
222 .forwarding = 0,
223 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
224 .mtu6 = IPV6_MIN_MTU,
225 .accept_ra = 1,
226 .accept_redirects = 1,
227 .autoconf = 1,
228 .force_mld_version = 0,
229 .mldv1_unsolicited_report_interval = 10 * HZ,
230 .mldv2_unsolicited_report_interval = HZ,
231 .dad_transmits = 1,
232 .rtr_solicits = MAX_RTR_SOLICITATIONS,
233 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
234 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
235 .use_tempaddr = 0,
236 .temp_valid_lft = TEMP_VALID_LIFETIME,
237 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
238 .regen_max_retry = REGEN_MAX_RETRY,
239 .max_desync_factor = MAX_DESYNC_FACTOR,
240 .max_addresses = IPV6_MAX_ADDRESSES,
241 .accept_ra_defrtr = 1,
242 .accept_ra_from_local = 0,
243 .accept_ra_min_hop_limit= 1,
244 .accept_ra_pinfo = 1,
245 #ifdef CONFIG_IPV6_ROUTER_PREF
246 .accept_ra_rtr_pref = 1,
247 .rtr_probe_interval = 60 * HZ,
248 #ifdef CONFIG_IPV6_ROUTE_INFO
249 .accept_ra_rt_info_max_plen = 0,
250 #endif
251 #endif
252 .proxy_ndp = 0,
253 .accept_source_route = 0, /* we do not accept RH0 by default. */
254 .disable_ipv6 = 0,
255 .accept_dad = 1,
256 .suppress_frag_ndisc = 1,
257 .accept_ra_mtu = 1,
258 .stable_secret = {
259 .initialized = false,
261 .use_oif_addrs_only = 0,
262 .ignore_routes_with_linkdown = 0,
265 /* Check if a valid qdisc is available */
266 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
268 return !qdisc_tx_is_noop(dev);
271 static void addrconf_del_rs_timer(struct inet6_dev *idev)
273 if (del_timer(&idev->rs_timer))
274 __in6_dev_put(idev);
277 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
279 if (cancel_delayed_work(&ifp->dad_work))
280 __in6_ifa_put(ifp);
283 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
284 unsigned long when)
286 if (!timer_pending(&idev->rs_timer))
287 in6_dev_hold(idev);
288 mod_timer(&idev->rs_timer, jiffies + when);
291 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
292 unsigned long delay)
294 if (!delayed_work_pending(&ifp->dad_work))
295 in6_ifa_hold(ifp);
296 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
299 static int snmp6_alloc_dev(struct inet6_dev *idev)
301 int i;
303 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
304 if (!idev->stats.ipv6)
305 goto err_ip;
307 for_each_possible_cpu(i) {
308 struct ipstats_mib *addrconf_stats;
309 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
310 u64_stats_init(&addrconf_stats->syncp);
314 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
315 GFP_KERNEL);
316 if (!idev->stats.icmpv6dev)
317 goto err_icmp;
318 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
319 GFP_KERNEL);
320 if (!idev->stats.icmpv6msgdev)
321 goto err_icmpmsg;
323 return 0;
325 err_icmpmsg:
326 kfree(idev->stats.icmpv6dev);
327 err_icmp:
328 free_percpu(idev->stats.ipv6);
329 err_ip:
330 return -ENOMEM;
333 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
335 struct inet6_dev *ndev;
336 int err = -ENOMEM;
338 ASSERT_RTNL();
340 if (dev->mtu < IPV6_MIN_MTU)
341 return ERR_PTR(-EINVAL);
343 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
344 if (!ndev)
345 return ERR_PTR(err);
347 rwlock_init(&ndev->lock);
348 ndev->dev = dev;
349 INIT_LIST_HEAD(&ndev->addr_list);
350 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
351 (unsigned long)ndev);
352 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
354 if (ndev->cnf.stable_secret.initialized)
355 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
356 else
357 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
359 ndev->cnf.mtu6 = dev->mtu;
360 ndev->cnf.sysctl = NULL;
361 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
362 if (!ndev->nd_parms) {
363 kfree(ndev);
364 return ERR_PTR(err);
366 if (ndev->cnf.forwarding)
367 dev_disable_lro(dev);
368 /* We refer to the device */
369 dev_hold(dev);
371 if (snmp6_alloc_dev(ndev) < 0) {
372 ADBG(KERN_WARNING
373 "%s: cannot allocate memory for statistics; dev=%s.\n",
374 __func__, dev->name);
375 neigh_parms_release(&nd_tbl, ndev->nd_parms);
376 dev_put(dev);
377 kfree(ndev);
378 return ERR_PTR(err);
381 if (snmp6_register_dev(ndev) < 0) {
382 ADBG(KERN_WARNING
383 "%s: cannot create /proc/net/dev_snmp6/%s\n",
384 __func__, dev->name);
385 goto err_release;
388 /* One reference from device. We must do this before
389 * we invoke __ipv6_regen_rndid().
391 in6_dev_hold(ndev);
393 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
394 ndev->cnf.accept_dad = -1;
396 #if IS_ENABLED(CONFIG_IPV6_SIT)
397 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
398 pr_info("%s: Disabled Multicast RS\n", dev->name);
399 ndev->cnf.rtr_solicits = 0;
401 #endif
403 INIT_LIST_HEAD(&ndev->tempaddr_list);
404 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
405 if ((dev->flags&IFF_LOOPBACK) ||
406 dev->type == ARPHRD_TUNNEL ||
407 dev->type == ARPHRD_TUNNEL6 ||
408 dev->type == ARPHRD_SIT ||
409 dev->type == ARPHRD_NONE) {
410 ndev->cnf.use_tempaddr = -1;
411 } else {
412 in6_dev_hold(ndev);
413 ipv6_regen_rndid((unsigned long) ndev);
416 ndev->token = in6addr_any;
418 if (netif_running(dev) && addrconf_qdisc_ok(dev))
419 ndev->if_flags |= IF_READY;
421 ipv6_mc_init_dev(ndev);
422 ndev->tstamp = jiffies;
423 err = addrconf_sysctl_register(ndev);
424 if (err) {
425 ipv6_mc_destroy_dev(ndev);
426 del_timer(&ndev->regen_timer);
427 snmp6_unregister_dev(ndev);
428 goto err_release;
430 /* protected by rtnl_lock */
431 rcu_assign_pointer(dev->ip6_ptr, ndev);
433 /* Join interface-local all-node multicast group */
434 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
436 /* Join all-node multicast group */
437 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
439 /* Join all-router multicast group if forwarding is set */
440 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
441 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
443 return ndev;
445 err_release:
446 neigh_parms_release(&nd_tbl, ndev->nd_parms);
447 ndev->dead = 1;
448 in6_dev_finish_destroy(ndev);
449 return ERR_PTR(err);
452 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
454 struct inet6_dev *idev;
456 ASSERT_RTNL();
458 idev = __in6_dev_get(dev);
459 if (!idev) {
460 idev = ipv6_add_dev(dev);
461 if (IS_ERR(idev))
462 return NULL;
465 if (dev->flags&IFF_UP)
466 ipv6_mc_up(idev);
467 return idev;
470 static int inet6_netconf_msgsize_devconf(int type)
472 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
473 + nla_total_size(4); /* NETCONFA_IFINDEX */
475 /* type -1 is used for ALL */
476 if (type == -1 || type == NETCONFA_FORWARDING)
477 size += nla_total_size(4);
478 #ifdef CONFIG_IPV6_MROUTE
479 if (type == -1 || type == NETCONFA_MC_FORWARDING)
480 size += nla_total_size(4);
481 #endif
482 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
483 size += nla_total_size(4);
485 if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
486 size += nla_total_size(4);
488 return size;
491 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
492 struct ipv6_devconf *devconf, u32 portid,
493 u32 seq, int event, unsigned int flags,
494 int type)
496 struct nlmsghdr *nlh;
497 struct netconfmsg *ncm;
499 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
500 flags);
501 if (!nlh)
502 return -EMSGSIZE;
504 ncm = nlmsg_data(nlh);
505 ncm->ncm_family = AF_INET6;
507 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
508 goto nla_put_failure;
510 /* type -1 is used for ALL */
511 if ((type == -1 || type == NETCONFA_FORWARDING) &&
512 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
513 goto nla_put_failure;
514 #ifdef CONFIG_IPV6_MROUTE
515 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
516 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
517 devconf->mc_forwarding) < 0)
518 goto nla_put_failure;
519 #endif
520 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
521 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
522 goto nla_put_failure;
524 if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
525 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
526 devconf->ignore_routes_with_linkdown) < 0)
527 goto nla_put_failure;
529 nlmsg_end(skb, nlh);
530 return 0;
532 nla_put_failure:
533 nlmsg_cancel(skb, nlh);
534 return -EMSGSIZE;
537 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
538 struct ipv6_devconf *devconf)
540 struct sk_buff *skb;
541 int err = -ENOBUFS;
543 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
544 if (!skb)
545 goto errout;
547 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
548 RTM_NEWNETCONF, 0, type);
549 if (err < 0) {
550 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
551 WARN_ON(err == -EMSGSIZE);
552 kfree_skb(skb);
553 goto errout;
555 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
556 return;
557 errout:
558 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
561 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
562 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
563 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
564 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
565 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
568 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
569 struct nlmsghdr *nlh)
571 struct net *net = sock_net(in_skb->sk);
572 struct nlattr *tb[NETCONFA_MAX+1];
573 struct netconfmsg *ncm;
574 struct sk_buff *skb;
575 struct ipv6_devconf *devconf;
576 struct inet6_dev *in6_dev;
577 struct net_device *dev;
578 int ifindex;
579 int err;
581 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
582 devconf_ipv6_policy);
583 if (err < 0)
584 goto errout;
586 err = EINVAL;
587 if (!tb[NETCONFA_IFINDEX])
588 goto errout;
590 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
591 switch (ifindex) {
592 case NETCONFA_IFINDEX_ALL:
593 devconf = net->ipv6.devconf_all;
594 break;
595 case NETCONFA_IFINDEX_DEFAULT:
596 devconf = net->ipv6.devconf_dflt;
597 break;
598 default:
599 dev = __dev_get_by_index(net, ifindex);
600 if (!dev)
601 goto errout;
602 in6_dev = __in6_dev_get(dev);
603 if (!in6_dev)
604 goto errout;
605 devconf = &in6_dev->cnf;
606 break;
609 err = -ENOBUFS;
610 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
611 if (!skb)
612 goto errout;
614 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
615 NETLINK_CB(in_skb).portid,
616 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
617 -1);
618 if (err < 0) {
619 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
620 WARN_ON(err == -EMSGSIZE);
621 kfree_skb(skb);
622 goto errout;
624 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
625 errout:
626 return err;
629 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
630 struct netlink_callback *cb)
632 struct net *net = sock_net(skb->sk);
633 int h, s_h;
634 int idx, s_idx;
635 struct net_device *dev;
636 struct inet6_dev *idev;
637 struct hlist_head *head;
639 s_h = cb->args[0];
640 s_idx = idx = cb->args[1];
642 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
643 idx = 0;
644 head = &net->dev_index_head[h];
645 rcu_read_lock();
646 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
647 net->dev_base_seq;
648 hlist_for_each_entry_rcu(dev, head, index_hlist) {
649 if (idx < s_idx)
650 goto cont;
651 idev = __in6_dev_get(dev);
652 if (!idev)
653 goto cont;
655 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
656 &idev->cnf,
657 NETLINK_CB(cb->skb).portid,
658 cb->nlh->nlmsg_seq,
659 RTM_NEWNETCONF,
660 NLM_F_MULTI,
661 -1) < 0) {
662 rcu_read_unlock();
663 goto done;
665 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
666 cont:
667 idx++;
669 rcu_read_unlock();
671 if (h == NETDEV_HASHENTRIES) {
672 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
673 net->ipv6.devconf_all,
674 NETLINK_CB(cb->skb).portid,
675 cb->nlh->nlmsg_seq,
676 RTM_NEWNETCONF, NLM_F_MULTI,
677 -1) < 0)
678 goto done;
679 else
680 h++;
682 if (h == NETDEV_HASHENTRIES + 1) {
683 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
684 net->ipv6.devconf_dflt,
685 NETLINK_CB(cb->skb).portid,
686 cb->nlh->nlmsg_seq,
687 RTM_NEWNETCONF, NLM_F_MULTI,
688 -1) < 0)
689 goto done;
690 else
691 h++;
693 done:
694 cb->args[0] = h;
695 cb->args[1] = idx;
697 return skb->len;
700 #ifdef CONFIG_SYSCTL
701 static void dev_forward_change(struct inet6_dev *idev)
703 struct net_device *dev;
704 struct inet6_ifaddr *ifa;
706 if (!idev)
707 return;
708 dev = idev->dev;
709 if (idev->cnf.forwarding)
710 dev_disable_lro(dev);
711 if (dev->flags & IFF_MULTICAST) {
712 if (idev->cnf.forwarding) {
713 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
714 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
715 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
716 } else {
717 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
718 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
719 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
723 list_for_each_entry(ifa, &idev->addr_list, if_list) {
724 if (ifa->flags&IFA_F_TENTATIVE)
725 continue;
726 if (idev->cnf.forwarding)
727 addrconf_join_anycast(ifa);
728 else
729 addrconf_leave_anycast(ifa);
731 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
732 dev->ifindex, &idev->cnf);
736 static void addrconf_forward_change(struct net *net, __s32 newf)
738 struct net_device *dev;
739 struct inet6_dev *idev;
741 for_each_netdev(net, dev) {
742 idev = __in6_dev_get(dev);
743 if (idev) {
744 int changed = (!idev->cnf.forwarding) ^ (!newf);
745 idev->cnf.forwarding = newf;
746 if (changed)
747 dev_forward_change(idev);
752 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
754 struct net *net;
755 int old;
757 if (!rtnl_trylock())
758 return restart_syscall();
760 net = (struct net *)table->extra2;
761 old = *p;
762 *p = newf;
764 if (p == &net->ipv6.devconf_dflt->forwarding) {
765 if ((!newf) ^ (!old))
766 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
767 NETCONFA_IFINDEX_DEFAULT,
768 net->ipv6.devconf_dflt);
769 rtnl_unlock();
770 return 0;
773 if (p == &net->ipv6.devconf_all->forwarding) {
774 net->ipv6.devconf_dflt->forwarding = newf;
775 addrconf_forward_change(net, newf);
776 if ((!newf) ^ (!old))
777 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
778 NETCONFA_IFINDEX_ALL,
779 net->ipv6.devconf_all);
780 } else if ((!newf) ^ (!old))
781 dev_forward_change((struct inet6_dev *)table->extra1);
782 rtnl_unlock();
784 if (newf)
785 rt6_purge_dflt_routers(net);
786 return 1;
789 static void addrconf_linkdown_change(struct net *net, __s32 newf)
791 struct net_device *dev;
792 struct inet6_dev *idev;
794 for_each_netdev(net, dev) {
795 idev = __in6_dev_get(dev);
796 if (idev) {
797 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
799 idev->cnf.ignore_routes_with_linkdown = newf;
800 if (changed)
801 inet6_netconf_notify_devconf(dev_net(dev),
802 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
803 dev->ifindex,
804 &idev->cnf);
809 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
811 struct net *net;
812 int old;
814 if (!rtnl_trylock())
815 return restart_syscall();
817 net = (struct net *)table->extra2;
818 old = *p;
819 *p = newf;
821 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
822 if ((!newf) ^ (!old))
823 inet6_netconf_notify_devconf(net,
824 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
825 NETCONFA_IFINDEX_DEFAULT,
826 net->ipv6.devconf_dflt);
827 rtnl_unlock();
828 return 0;
831 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
832 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
833 addrconf_linkdown_change(net, newf);
834 if ((!newf) ^ (!old))
835 inet6_netconf_notify_devconf(net,
836 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
837 NETCONFA_IFINDEX_ALL,
838 net->ipv6.devconf_all);
840 rtnl_unlock();
842 return 1;
845 #endif
847 /* Nobody refers to this ifaddr, destroy it */
848 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
850 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
852 #ifdef NET_REFCNT_DEBUG
853 pr_debug("%s\n", __func__);
854 #endif
856 in6_dev_put(ifp->idev);
858 if (cancel_delayed_work(&ifp->dad_work))
859 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
860 ifp);
862 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
863 pr_warn("Freeing alive inet6 address %p\n", ifp);
864 return;
866 ip6_rt_put(ifp->rt);
868 kfree_rcu(ifp, rcu);
871 static void
872 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
874 struct list_head *p;
875 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
878 * Each device address list is sorted in order of scope -
879 * global before linklocal.
881 list_for_each(p, &idev->addr_list) {
882 struct inet6_ifaddr *ifa
883 = list_entry(p, struct inet6_ifaddr, if_list);
884 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
885 break;
888 list_add_tail(&ifp->if_list, p);
891 static u32 inet6_addr_hash(const struct in6_addr *addr)
893 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
896 /* On success it returns ifp with increased reference count */
898 static struct inet6_ifaddr *
899 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
900 const struct in6_addr *peer_addr, int pfxlen,
901 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
903 struct inet6_ifaddr *ifa = NULL;
904 struct rt6_info *rt;
905 unsigned int hash;
906 int err = 0;
907 int addr_type = ipv6_addr_type(addr);
909 if (addr_type == IPV6_ADDR_ANY ||
910 addr_type & IPV6_ADDR_MULTICAST ||
911 (!(idev->dev->flags & IFF_LOOPBACK) &&
912 addr_type & IPV6_ADDR_LOOPBACK))
913 return ERR_PTR(-EADDRNOTAVAIL);
915 rcu_read_lock_bh();
916 if (idev->dead) {
917 err = -ENODEV; /*XXX*/
918 goto out2;
921 if (idev->cnf.disable_ipv6) {
922 err = -EACCES;
923 goto out2;
926 spin_lock(&addrconf_hash_lock);
928 /* Ignore adding duplicate addresses on an interface */
929 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
930 ADBG("ipv6_add_addr: already assigned\n");
931 err = -EEXIST;
932 goto out;
935 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
937 if (!ifa) {
938 ADBG("ipv6_add_addr: malloc failed\n");
939 err = -ENOBUFS;
940 goto out;
943 rt = addrconf_dst_alloc(idev, addr, false);
944 if (IS_ERR(rt)) {
945 err = PTR_ERR(rt);
946 goto out;
949 neigh_parms_data_state_setall(idev->nd_parms);
951 ifa->addr = *addr;
952 if (peer_addr)
953 ifa->peer_addr = *peer_addr;
955 spin_lock_init(&ifa->lock);
956 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
957 INIT_HLIST_NODE(&ifa->addr_lst);
958 ifa->scope = scope;
959 ifa->prefix_len = pfxlen;
960 ifa->flags = flags | IFA_F_TENTATIVE;
961 ifa->valid_lft = valid_lft;
962 ifa->prefered_lft = prefered_lft;
963 ifa->cstamp = ifa->tstamp = jiffies;
964 ifa->tokenized = false;
966 ifa->rt = rt;
968 ifa->idev = idev;
969 in6_dev_hold(idev);
970 /* For caller */
971 in6_ifa_hold(ifa);
973 /* Add to big hash table */
974 hash = inet6_addr_hash(addr);
976 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
977 spin_unlock(&addrconf_hash_lock);
979 write_lock(&idev->lock);
980 /* Add to inet6_dev unicast addr list. */
981 ipv6_link_dev_addr(idev, ifa);
983 if (ifa->flags&IFA_F_TEMPORARY) {
984 list_add(&ifa->tmp_list, &idev->tempaddr_list);
985 in6_ifa_hold(ifa);
988 in6_ifa_hold(ifa);
989 write_unlock(&idev->lock);
990 out2:
991 rcu_read_unlock_bh();
993 if (likely(err == 0))
994 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
995 else {
996 kfree(ifa);
997 ifa = ERR_PTR(err);
1000 return ifa;
1001 out:
1002 spin_unlock(&addrconf_hash_lock);
1003 goto out2;
1006 enum cleanup_prefix_rt_t {
1007 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1008 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1009 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1013 * Check, whether the prefix for ifp would still need a prefix route
1014 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1015 * constants.
1017 * 1) we don't purge prefix if address was not permanent.
1018 * prefix is managed by its own lifetime.
1019 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1020 * 3) if there are no addresses, delete prefix.
1021 * 4) if there are still other permanent address(es),
1022 * corresponding prefix is still permanent.
1023 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1024 * don't purge the prefix, assume user space is managing it.
1025 * 6) otherwise, update prefix lifetime to the
1026 * longest valid lifetime among the corresponding
1027 * addresses on the device.
1028 * Note: subsequent RA will update lifetime.
1030 static enum cleanup_prefix_rt_t
1031 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1033 struct inet6_ifaddr *ifa;
1034 struct inet6_dev *idev = ifp->idev;
1035 unsigned long lifetime;
1036 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1038 *expires = jiffies;
1040 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1041 if (ifa == ifp)
1042 continue;
1043 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1044 ifp->prefix_len))
1045 continue;
1046 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1047 return CLEANUP_PREFIX_RT_NOP;
1049 action = CLEANUP_PREFIX_RT_EXPIRE;
1051 spin_lock(&ifa->lock);
1053 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1055 * Note: Because this address is
1056 * not permanent, lifetime <
1057 * LONG_MAX / HZ here.
1059 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1060 *expires = ifa->tstamp + lifetime * HZ;
1061 spin_unlock(&ifa->lock);
1064 return action;
1067 static void
1068 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1070 struct rt6_info *rt;
1072 rt = addrconf_get_prefix_route(&ifp->addr,
1073 ifp->prefix_len,
1074 ifp->idev->dev,
1075 0, RTF_GATEWAY | RTF_DEFAULT);
1076 if (rt) {
1077 if (del_rt)
1078 ip6_del_rt(rt);
1079 else {
1080 if (!(rt->rt6i_flags & RTF_EXPIRES))
1081 rt6_set_expires(rt, expires);
1082 ip6_rt_put(rt);
1088 /* This function wants to get referenced ifp and releases it before return */
1090 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1092 int state;
1093 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1094 unsigned long expires;
1096 ASSERT_RTNL();
1098 spin_lock_bh(&ifp->lock);
1099 state = ifp->state;
1100 ifp->state = INET6_IFADDR_STATE_DEAD;
1101 spin_unlock_bh(&ifp->lock);
1103 if (state == INET6_IFADDR_STATE_DEAD)
1104 goto out;
1106 spin_lock_bh(&addrconf_hash_lock);
1107 hlist_del_init_rcu(&ifp->addr_lst);
1108 spin_unlock_bh(&addrconf_hash_lock);
1110 write_lock_bh(&ifp->idev->lock);
1112 if (ifp->flags&IFA_F_TEMPORARY) {
1113 list_del(&ifp->tmp_list);
1114 if (ifp->ifpub) {
1115 in6_ifa_put(ifp->ifpub);
1116 ifp->ifpub = NULL;
1118 __in6_ifa_put(ifp);
1121 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1122 action = check_cleanup_prefix_route(ifp, &expires);
1124 list_del_init(&ifp->if_list);
1125 __in6_ifa_put(ifp);
1127 write_unlock_bh(&ifp->idev->lock);
1129 addrconf_del_dad_work(ifp);
1131 ipv6_ifa_notify(RTM_DELADDR, ifp);
1133 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1135 if (action != CLEANUP_PREFIX_RT_NOP) {
1136 cleanup_prefix_route(ifp, expires,
1137 action == CLEANUP_PREFIX_RT_DEL);
1140 /* clean up prefsrc entries */
1141 rt6_remove_prefsrc(ifp);
1142 out:
1143 in6_ifa_put(ifp);
1146 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1148 struct inet6_dev *idev = ifp->idev;
1149 struct in6_addr addr, *tmpaddr;
1150 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1151 unsigned long regen_advance;
1152 int tmp_plen;
1153 int ret = 0;
1154 u32 addr_flags;
1155 unsigned long now = jiffies;
1157 write_lock_bh(&idev->lock);
1158 if (ift) {
1159 spin_lock_bh(&ift->lock);
1160 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1161 spin_unlock_bh(&ift->lock);
1162 tmpaddr = &addr;
1163 } else {
1164 tmpaddr = NULL;
1166 retry:
1167 in6_dev_hold(idev);
1168 if (idev->cnf.use_tempaddr <= 0) {
1169 write_unlock_bh(&idev->lock);
1170 pr_info("%s: use_tempaddr is disabled\n", __func__);
1171 in6_dev_put(idev);
1172 ret = -1;
1173 goto out;
1175 spin_lock_bh(&ifp->lock);
1176 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1177 idev->cnf.use_tempaddr = -1; /*XXX*/
1178 spin_unlock_bh(&ifp->lock);
1179 write_unlock_bh(&idev->lock);
1180 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1181 __func__);
1182 in6_dev_put(idev);
1183 ret = -1;
1184 goto out;
1186 in6_ifa_hold(ifp);
1187 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1188 __ipv6_try_regen_rndid(idev, tmpaddr);
1189 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1190 age = (now - ifp->tstamp) / HZ;
1191 tmp_valid_lft = min_t(__u32,
1192 ifp->valid_lft,
1193 idev->cnf.temp_valid_lft + age);
1194 tmp_prefered_lft = min_t(__u32,
1195 ifp->prefered_lft,
1196 idev->cnf.temp_prefered_lft + age -
1197 idev->cnf.max_desync_factor);
1198 tmp_plen = ifp->prefix_len;
1199 tmp_tstamp = ifp->tstamp;
1200 spin_unlock_bh(&ifp->lock);
1202 regen_advance = idev->cnf.regen_max_retry *
1203 idev->cnf.dad_transmits *
1204 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1205 write_unlock_bh(&idev->lock);
1207 /* A temporary address is created only if this calculated Preferred
1208 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1209 * an implementation must not create a temporary address with a zero
1210 * Preferred Lifetime.
1211 * Use age calculation as in addrconf_verify to avoid unnecessary
1212 * temporary addresses being generated.
1214 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1215 if (tmp_prefered_lft <= regen_advance + age) {
1216 in6_ifa_put(ifp);
1217 in6_dev_put(idev);
1218 ret = -1;
1219 goto out;
1222 addr_flags = IFA_F_TEMPORARY;
1223 /* set in addrconf_prefix_rcv() */
1224 if (ifp->flags & IFA_F_OPTIMISTIC)
1225 addr_flags |= IFA_F_OPTIMISTIC;
1227 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1228 ipv6_addr_scope(&addr), addr_flags,
1229 tmp_valid_lft, tmp_prefered_lft);
1230 if (IS_ERR(ift)) {
1231 in6_ifa_put(ifp);
1232 in6_dev_put(idev);
1233 pr_info("%s: retry temporary address regeneration\n", __func__);
1234 tmpaddr = &addr;
1235 write_lock_bh(&idev->lock);
1236 goto retry;
1239 spin_lock_bh(&ift->lock);
1240 ift->ifpub = ifp;
1241 ift->cstamp = now;
1242 ift->tstamp = tmp_tstamp;
1243 spin_unlock_bh(&ift->lock);
1245 addrconf_dad_start(ift);
1246 in6_ifa_put(ift);
1247 in6_dev_put(idev);
1248 out:
1249 return ret;
1253 * Choose an appropriate source address (RFC3484)
1255 enum {
1256 IPV6_SADDR_RULE_INIT = 0,
1257 IPV6_SADDR_RULE_LOCAL,
1258 IPV6_SADDR_RULE_SCOPE,
1259 IPV6_SADDR_RULE_PREFERRED,
1260 #ifdef CONFIG_IPV6_MIP6
1261 IPV6_SADDR_RULE_HOA,
1262 #endif
1263 IPV6_SADDR_RULE_OIF,
1264 IPV6_SADDR_RULE_LABEL,
1265 IPV6_SADDR_RULE_PRIVACY,
1266 IPV6_SADDR_RULE_ORCHID,
1267 IPV6_SADDR_RULE_PREFIX,
1268 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1269 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1270 #endif
1271 IPV6_SADDR_RULE_MAX
1274 struct ipv6_saddr_score {
1275 int rule;
1276 int addr_type;
1277 struct inet6_ifaddr *ifa;
1278 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1279 int scopedist;
1280 int matchlen;
1283 struct ipv6_saddr_dst {
1284 const struct in6_addr *addr;
1285 int ifindex;
1286 int scope;
1287 int label;
1288 unsigned int prefs;
1291 static inline int ipv6_saddr_preferred(int type)
1293 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1294 return 1;
1295 return 0;
1298 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1300 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1301 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1302 #else
1303 return false;
1304 #endif
1307 static int ipv6_get_saddr_eval(struct net *net,
1308 struct ipv6_saddr_score *score,
1309 struct ipv6_saddr_dst *dst,
1310 int i)
1312 int ret;
1314 if (i <= score->rule) {
1315 switch (i) {
1316 case IPV6_SADDR_RULE_SCOPE:
1317 ret = score->scopedist;
1318 break;
1319 case IPV6_SADDR_RULE_PREFIX:
1320 ret = score->matchlen;
1321 break;
1322 default:
1323 ret = !!test_bit(i, score->scorebits);
1325 goto out;
1328 switch (i) {
1329 case IPV6_SADDR_RULE_INIT:
1330 /* Rule 0: remember if hiscore is not ready yet */
1331 ret = !!score->ifa;
1332 break;
1333 case IPV6_SADDR_RULE_LOCAL:
1334 /* Rule 1: Prefer same address */
1335 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1336 break;
1337 case IPV6_SADDR_RULE_SCOPE:
1338 /* Rule 2: Prefer appropriate scope
1340 * ret
1342 * -1 | d 15
1343 * ---+--+-+---> scope
1345 * | d is scope of the destination.
1346 * B-d | \
1347 * | \ <- smaller scope is better if
1348 * B-15 | \ if scope is enough for destination.
1349 * | ret = B - scope (-1 <= scope >= d <= 15).
1350 * d-C-1 | /
1351 * |/ <- greater is better
1352 * -C / if scope is not enough for destination.
1353 * /| ret = scope - C (-1 <= d < scope <= 15).
1355 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1356 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1357 * Assume B = 0 and we get C > 29.
1359 ret = __ipv6_addr_src_scope(score->addr_type);
1360 if (ret >= dst->scope)
1361 ret = -ret;
1362 else
1363 ret -= 128; /* 30 is enough */
1364 score->scopedist = ret;
1365 break;
1366 case IPV6_SADDR_RULE_PREFERRED:
1368 /* Rule 3: Avoid deprecated and optimistic addresses */
1369 u8 avoid = IFA_F_DEPRECATED;
1371 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1372 avoid |= IFA_F_OPTIMISTIC;
1373 ret = ipv6_saddr_preferred(score->addr_type) ||
1374 !(score->ifa->flags & avoid);
1375 break;
1377 #ifdef CONFIG_IPV6_MIP6
1378 case IPV6_SADDR_RULE_HOA:
1380 /* Rule 4: Prefer home address */
1381 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1382 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1383 break;
1385 #endif
1386 case IPV6_SADDR_RULE_OIF:
1387 /* Rule 5: Prefer outgoing interface */
1388 ret = (!dst->ifindex ||
1389 dst->ifindex == score->ifa->idev->dev->ifindex);
1390 break;
1391 case IPV6_SADDR_RULE_LABEL:
1392 /* Rule 6: Prefer matching label */
1393 ret = ipv6_addr_label(net,
1394 &score->ifa->addr, score->addr_type,
1395 score->ifa->idev->dev->ifindex) == dst->label;
1396 break;
1397 case IPV6_SADDR_RULE_PRIVACY:
1399 /* Rule 7: Prefer public address
1400 * Note: prefer temporary address if use_tempaddr >= 2
1402 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1403 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1404 score->ifa->idev->cnf.use_tempaddr >= 2;
1405 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1406 break;
1408 case IPV6_SADDR_RULE_ORCHID:
1409 /* Rule 8-: Prefer ORCHID vs ORCHID or
1410 * non-ORCHID vs non-ORCHID
1412 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1413 ipv6_addr_orchid(dst->addr));
1414 break;
1415 case IPV6_SADDR_RULE_PREFIX:
1416 /* Rule 8: Use longest matching prefix */
1417 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1418 if (ret > score->ifa->prefix_len)
1419 ret = score->ifa->prefix_len;
1420 score->matchlen = ret;
1421 break;
1422 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1423 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1424 /* Optimistic addresses still have lower precedence than other
1425 * preferred addresses.
1427 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1428 break;
1429 #endif
1430 default:
1431 ret = 0;
1434 if (ret)
1435 __set_bit(i, score->scorebits);
1436 score->rule = i;
1437 out:
1438 return ret;
1441 static int __ipv6_dev_get_saddr(struct net *net,
1442 struct ipv6_saddr_dst *dst,
1443 struct inet6_dev *idev,
1444 struct ipv6_saddr_score *scores,
1445 int hiscore_idx)
1447 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1449 read_lock_bh(&idev->lock);
1450 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1451 int i;
1454 * - Tentative Address (RFC2462 section 5.4)
1455 * - A tentative address is not considered
1456 * "assigned to an interface" in the traditional
1457 * sense, unless it is also flagged as optimistic.
1458 * - Candidate Source Address (section 4)
1459 * - In any case, anycast addresses, multicast
1460 * addresses, and the unspecified address MUST
1461 * NOT be included in a candidate set.
1463 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1464 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1465 continue;
1467 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1469 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1470 score->addr_type & IPV6_ADDR_MULTICAST)) {
1471 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1472 idev->dev->name);
1473 continue;
1476 score->rule = -1;
1477 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1479 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1480 int minihiscore, miniscore;
1482 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1483 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1485 if (minihiscore > miniscore) {
1486 if (i == IPV6_SADDR_RULE_SCOPE &&
1487 score->scopedist > 0) {
1489 * special case:
1490 * each remaining entry
1491 * has too small (not enough)
1492 * scope, because ifa entries
1493 * are sorted by their scope
1494 * values.
1496 goto out;
1498 break;
1499 } else if (minihiscore < miniscore) {
1500 if (hiscore->ifa)
1501 in6_ifa_put(hiscore->ifa);
1503 in6_ifa_hold(score->ifa);
1505 swap(hiscore, score);
1506 hiscore_idx = 1 - hiscore_idx;
1508 /* restore our iterator */
1509 score->ifa = hiscore->ifa;
1511 break;
1515 out:
1516 read_unlock_bh(&idev->lock);
1517 return hiscore_idx;
1520 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1521 const struct in6_addr *daddr, unsigned int prefs,
1522 struct in6_addr *saddr)
1524 struct ipv6_saddr_score scores[2], *hiscore;
1525 struct ipv6_saddr_dst dst;
1526 struct inet6_dev *idev;
1527 struct net_device *dev;
1528 int dst_type;
1529 bool use_oif_addr = false;
1530 int hiscore_idx = 0;
1532 dst_type = __ipv6_addr_type(daddr);
1533 dst.addr = daddr;
1534 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1535 dst.scope = __ipv6_addr_src_scope(dst_type);
1536 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1537 dst.prefs = prefs;
1539 scores[hiscore_idx].rule = -1;
1540 scores[hiscore_idx].ifa = NULL;
1542 rcu_read_lock();
1544 /* Candidate Source Address (section 4)
1545 * - multicast and link-local destination address,
1546 * the set of candidate source address MUST only
1547 * include addresses assigned to interfaces
1548 * belonging to the same link as the outgoing
1549 * interface.
1550 * (- For site-local destination addresses, the
1551 * set of candidate source addresses MUST only
1552 * include addresses assigned to interfaces
1553 * belonging to the same site as the outgoing
1554 * interface.)
1555 * - "It is RECOMMENDED that the candidate source addresses
1556 * be the set of unicast addresses assigned to the
1557 * interface that will be used to send to the destination
1558 * (the 'outgoing' interface)." (RFC 6724)
1560 if (dst_dev) {
1561 idev = __in6_dev_get(dst_dev);
1562 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1563 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1564 (idev && idev->cnf.use_oif_addrs_only)) {
1565 use_oif_addr = true;
1569 if (use_oif_addr) {
1570 if (idev)
1571 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1572 } else {
1573 for_each_netdev_rcu(net, dev) {
1574 idev = __in6_dev_get(dev);
1575 if (!idev)
1576 continue;
1577 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1580 rcu_read_unlock();
1582 hiscore = &scores[hiscore_idx];
1583 if (!hiscore->ifa)
1584 return -EADDRNOTAVAIL;
1586 *saddr = hiscore->ifa->addr;
1587 in6_ifa_put(hiscore->ifa);
1588 return 0;
1590 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1592 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1593 u32 banned_flags)
1595 struct inet6_ifaddr *ifp;
1596 int err = -EADDRNOTAVAIL;
1598 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1599 if (ifp->scope > IFA_LINK)
1600 break;
1601 if (ifp->scope == IFA_LINK &&
1602 !(ifp->flags & banned_flags)) {
1603 *addr = ifp->addr;
1604 err = 0;
1605 break;
1608 return err;
1611 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1612 u32 banned_flags)
1614 struct inet6_dev *idev;
1615 int err = -EADDRNOTAVAIL;
1617 rcu_read_lock();
1618 idev = __in6_dev_get(dev);
1619 if (idev) {
1620 read_lock_bh(&idev->lock);
1621 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1622 read_unlock_bh(&idev->lock);
1624 rcu_read_unlock();
1625 return err;
1628 static int ipv6_count_addresses(struct inet6_dev *idev)
1630 int cnt = 0;
1631 struct inet6_ifaddr *ifp;
1633 read_lock_bh(&idev->lock);
1634 list_for_each_entry(ifp, &idev->addr_list, if_list)
1635 cnt++;
1636 read_unlock_bh(&idev->lock);
1637 return cnt;
1640 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1641 const struct net_device *dev, int strict)
1643 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1645 EXPORT_SYMBOL(ipv6_chk_addr);
1647 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1648 const struct net_device *dev, int strict,
1649 u32 banned_flags)
1651 struct inet6_ifaddr *ifp;
1652 unsigned int hash = inet6_addr_hash(addr);
1653 u32 ifp_flags;
1655 rcu_read_lock_bh();
1656 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1657 if (!net_eq(dev_net(ifp->idev->dev), net))
1658 continue;
1659 /* Decouple optimistic from tentative for evaluation here.
1660 * Ban optimistic addresses explicitly, when required.
1662 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1663 ? (ifp->flags&~IFA_F_TENTATIVE)
1664 : ifp->flags;
1665 if (ipv6_addr_equal(&ifp->addr, addr) &&
1666 !(ifp_flags&banned_flags) &&
1667 (!dev || ifp->idev->dev == dev ||
1668 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1669 rcu_read_unlock_bh();
1670 return 1;
1674 rcu_read_unlock_bh();
1675 return 0;
1677 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1679 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1680 struct net_device *dev)
1682 unsigned int hash = inet6_addr_hash(addr);
1683 struct inet6_ifaddr *ifp;
1685 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1686 if (!net_eq(dev_net(ifp->idev->dev), net))
1687 continue;
1688 if (ipv6_addr_equal(&ifp->addr, addr)) {
1689 if (!dev || ifp->idev->dev == dev)
1690 return true;
1693 return false;
1696 /* Compares an address/prefix_len with addresses on device @dev.
1697 * If one is found it returns true.
1699 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1700 const unsigned int prefix_len, struct net_device *dev)
1702 struct inet6_dev *idev;
1703 struct inet6_ifaddr *ifa;
1704 bool ret = false;
1706 rcu_read_lock();
1707 idev = __in6_dev_get(dev);
1708 if (idev) {
1709 read_lock_bh(&idev->lock);
1710 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1711 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1712 if (ret)
1713 break;
1715 read_unlock_bh(&idev->lock);
1717 rcu_read_unlock();
1719 return ret;
1721 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1723 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1725 struct inet6_dev *idev;
1726 struct inet6_ifaddr *ifa;
1727 int onlink;
1729 onlink = 0;
1730 rcu_read_lock();
1731 idev = __in6_dev_get(dev);
1732 if (idev) {
1733 read_lock_bh(&idev->lock);
1734 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1735 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1736 ifa->prefix_len);
1737 if (onlink)
1738 break;
1740 read_unlock_bh(&idev->lock);
1742 rcu_read_unlock();
1743 return onlink;
1745 EXPORT_SYMBOL(ipv6_chk_prefix);
1747 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1748 struct net_device *dev, int strict)
1750 struct inet6_ifaddr *ifp, *result = NULL;
1751 unsigned int hash = inet6_addr_hash(addr);
1753 rcu_read_lock_bh();
1754 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1755 if (!net_eq(dev_net(ifp->idev->dev), net))
1756 continue;
1757 if (ipv6_addr_equal(&ifp->addr, addr)) {
1758 if (!dev || ifp->idev->dev == dev ||
1759 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1760 result = ifp;
1761 in6_ifa_hold(ifp);
1762 break;
1766 rcu_read_unlock_bh();
1768 return result;
1771 /* Gets referenced address, destroys ifaddr */
1773 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1775 if (ifp->flags&IFA_F_PERMANENT) {
1776 spin_lock_bh(&ifp->lock);
1777 addrconf_del_dad_work(ifp);
1778 ifp->flags |= IFA_F_TENTATIVE;
1779 if (dad_failed)
1780 ifp->flags |= IFA_F_DADFAILED;
1781 spin_unlock_bh(&ifp->lock);
1782 if (dad_failed)
1783 ipv6_ifa_notify(0, ifp);
1784 in6_ifa_put(ifp);
1785 } else if (ifp->flags&IFA_F_TEMPORARY) {
1786 struct inet6_ifaddr *ifpub;
1787 spin_lock_bh(&ifp->lock);
1788 ifpub = ifp->ifpub;
1789 if (ifpub) {
1790 in6_ifa_hold(ifpub);
1791 spin_unlock_bh(&ifp->lock);
1792 ipv6_create_tempaddr(ifpub, ifp);
1793 in6_ifa_put(ifpub);
1794 } else {
1795 spin_unlock_bh(&ifp->lock);
1797 ipv6_del_addr(ifp);
1798 } else {
1799 ipv6_del_addr(ifp);
1803 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1805 int err = -ENOENT;
1807 spin_lock_bh(&ifp->lock);
1808 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1809 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1810 err = 0;
1812 spin_unlock_bh(&ifp->lock);
1814 return err;
1817 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1819 struct in6_addr addr;
1820 struct inet6_dev *idev = ifp->idev;
1821 struct net *net = dev_net(ifp->idev->dev);
1823 if (addrconf_dad_end(ifp)) {
1824 in6_ifa_put(ifp);
1825 return;
1828 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1829 ifp->idev->dev->name, &ifp->addr);
1831 spin_lock_bh(&ifp->lock);
1833 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1834 int scope = ifp->scope;
1835 u32 flags = ifp->flags;
1836 struct in6_addr new_addr;
1837 struct inet6_ifaddr *ifp2;
1838 u32 valid_lft, preferred_lft;
1839 int pfxlen = ifp->prefix_len;
1840 int retries = ifp->stable_privacy_retry + 1;
1842 if (retries > net->ipv6.sysctl.idgen_retries) {
1843 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1844 ifp->idev->dev->name);
1845 goto errdad;
1848 new_addr = ifp->addr;
1849 if (ipv6_generate_stable_address(&new_addr, retries,
1850 idev))
1851 goto errdad;
1853 valid_lft = ifp->valid_lft;
1854 preferred_lft = ifp->prefered_lft;
1856 spin_unlock_bh(&ifp->lock);
1858 if (idev->cnf.max_addresses &&
1859 ipv6_count_addresses(idev) >=
1860 idev->cnf.max_addresses)
1861 goto lock_errdad;
1863 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1864 ifp->idev->dev->name);
1866 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1867 scope, flags, valid_lft,
1868 preferred_lft);
1869 if (IS_ERR(ifp2))
1870 goto lock_errdad;
1872 spin_lock_bh(&ifp2->lock);
1873 ifp2->stable_privacy_retry = retries;
1874 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1875 spin_unlock_bh(&ifp2->lock);
1877 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1878 in6_ifa_put(ifp2);
1879 lock_errdad:
1880 spin_lock_bh(&ifp->lock);
1881 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1882 addr.s6_addr32[0] = htonl(0xfe800000);
1883 addr.s6_addr32[1] = 0;
1885 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1886 ipv6_addr_equal(&ifp->addr, &addr)) {
1887 /* DAD failed for link-local based on MAC address */
1888 idev->cnf.disable_ipv6 = 1;
1890 pr_info("%s: IPv6 being disabled!\n",
1891 ifp->idev->dev->name);
1895 errdad:
1896 /* transition from _POSTDAD to _ERRDAD */
1897 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1898 spin_unlock_bh(&ifp->lock);
1900 addrconf_mod_dad_work(ifp, 0);
1903 /* Join to solicited addr multicast group.
1904 * caller must hold RTNL */
1905 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1907 struct in6_addr maddr;
1909 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1910 return;
1912 addrconf_addr_solict_mult(addr, &maddr);
1913 ipv6_dev_mc_inc(dev, &maddr);
1916 /* caller must hold RTNL */
1917 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1919 struct in6_addr maddr;
1921 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1922 return;
1924 addrconf_addr_solict_mult(addr, &maddr);
1925 __ipv6_dev_mc_dec(idev, &maddr);
1928 /* caller must hold RTNL */
1929 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1931 struct in6_addr addr;
1933 if (ifp->prefix_len >= 127) /* RFC 6164 */
1934 return;
1935 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1936 if (ipv6_addr_any(&addr))
1937 return;
1938 __ipv6_dev_ac_inc(ifp->idev, &addr);
1941 /* caller must hold RTNL */
1942 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1944 struct in6_addr addr;
1946 if (ifp->prefix_len >= 127) /* RFC 6164 */
1947 return;
1948 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1949 if (ipv6_addr_any(&addr))
1950 return;
1951 __ipv6_dev_ac_dec(ifp->idev, &addr);
1954 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1956 if (dev->addr_len != IEEE802154_ADDR_LEN)
1957 return -1;
1958 memcpy(eui, dev->dev_addr, 8);
1959 eui[0] ^= 2;
1960 return 0;
1963 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1965 union fwnet_hwaddr *ha;
1967 if (dev->addr_len != FWNET_ALEN)
1968 return -1;
1970 ha = (union fwnet_hwaddr *)dev->dev_addr;
1972 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1973 eui[0] ^= 2;
1974 return 0;
1977 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1979 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1980 if (dev->addr_len != ARCNET_ALEN)
1981 return -1;
1982 memset(eui, 0, 7);
1983 eui[7] = *(u8 *)dev->dev_addr;
1984 return 0;
1987 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1989 if (dev->addr_len != INFINIBAND_ALEN)
1990 return -1;
1991 memcpy(eui, dev->dev_addr + 12, 8);
1992 eui[0] |= 2;
1993 return 0;
1996 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1998 if (addr == 0)
1999 return -1;
2000 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2001 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2002 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2003 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2004 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2005 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2006 eui[1] = 0;
2007 eui[2] = 0x5E;
2008 eui[3] = 0xFE;
2009 memcpy(eui + 4, &addr, 4);
2010 return 0;
2013 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2015 if (dev->priv_flags & IFF_ISATAP)
2016 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2017 return -1;
2020 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2022 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2025 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2027 memcpy(eui, dev->perm_addr, 3);
2028 memcpy(eui + 5, dev->perm_addr + 3, 3);
2029 eui[3] = 0xFF;
2030 eui[4] = 0xFE;
2031 eui[0] ^= 2;
2032 return 0;
2035 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2037 switch (dev->type) {
2038 case ARPHRD_ETHER:
2039 case ARPHRD_FDDI:
2040 return addrconf_ifid_eui48(eui, dev);
2041 case ARPHRD_ARCNET:
2042 return addrconf_ifid_arcnet(eui, dev);
2043 case ARPHRD_INFINIBAND:
2044 return addrconf_ifid_infiniband(eui, dev);
2045 case ARPHRD_SIT:
2046 return addrconf_ifid_sit(eui, dev);
2047 case ARPHRD_IPGRE:
2048 return addrconf_ifid_gre(eui, dev);
2049 case ARPHRD_6LOWPAN:
2050 case ARPHRD_IEEE802154:
2051 return addrconf_ifid_eui64(eui, dev);
2052 case ARPHRD_IEEE1394:
2053 return addrconf_ifid_ieee1394(eui, dev);
2054 case ARPHRD_TUNNEL6:
2055 return addrconf_ifid_ip6tnl(eui, dev);
2057 return -1;
2060 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2062 int err = -1;
2063 struct inet6_ifaddr *ifp;
2065 read_lock_bh(&idev->lock);
2066 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2067 if (ifp->scope > IFA_LINK)
2068 break;
2069 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2070 memcpy(eui, ifp->addr.s6_addr+8, 8);
2071 err = 0;
2072 break;
2075 read_unlock_bh(&idev->lock);
2076 return err;
2079 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2080 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2082 regen:
2083 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2084 idev->rndid[0] &= ~0x02;
2087 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2088 * check if generated address is not inappropriate
2090 * - Reserved subnet anycast (RFC 2526)
2091 * 11111101 11....11 1xxxxxxx
2092 * - ISATAP (RFC4214) 6.1
2093 * 00-00-5E-FE-xx-xx-xx-xx
2094 * - value 0
2095 * - XXX: already assigned to an address on the device
2097 if (idev->rndid[0] == 0xfd &&
2098 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2099 (idev->rndid[7]&0x80))
2100 goto regen;
2101 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2102 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2103 goto regen;
2104 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2105 goto regen;
2109 static void ipv6_regen_rndid(unsigned long data)
2111 struct inet6_dev *idev = (struct inet6_dev *) data;
2112 unsigned long expires;
2114 rcu_read_lock_bh();
2115 write_lock_bh(&idev->lock);
2117 if (idev->dead)
2118 goto out;
2120 __ipv6_regen_rndid(idev);
2122 expires = jiffies +
2123 idev->cnf.temp_prefered_lft * HZ -
2124 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2125 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2126 idev->cnf.max_desync_factor * HZ;
2127 if (time_before(expires, jiffies)) {
2128 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2129 __func__, idev->dev->name);
2130 goto out;
2133 if (!mod_timer(&idev->regen_timer, expires))
2134 in6_dev_hold(idev);
2136 out:
2137 write_unlock_bh(&idev->lock);
2138 rcu_read_unlock_bh();
2139 in6_dev_put(idev);
2142 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2144 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2145 __ipv6_regen_rndid(idev);
2149 * Add prefix route.
2152 static void
2153 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2154 unsigned long expires, u32 flags)
2156 struct fib6_config cfg = {
2157 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2158 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2159 .fc_ifindex = dev->ifindex,
2160 .fc_expires = expires,
2161 .fc_dst_len = plen,
2162 .fc_flags = RTF_UP | flags,
2163 .fc_nlinfo.nl_net = dev_net(dev),
2164 .fc_protocol = RTPROT_KERNEL,
2167 cfg.fc_dst = *pfx;
2169 /* Prevent useless cloning on PtP SIT.
2170 This thing is done here expecting that the whole
2171 class of non-broadcast devices need not cloning.
2173 #if IS_ENABLED(CONFIG_IPV6_SIT)
2174 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2175 cfg.fc_flags |= RTF_NONEXTHOP;
2176 #endif
2178 ip6_route_add(&cfg);
2182 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2183 int plen,
2184 const struct net_device *dev,
2185 u32 flags, u32 noflags)
2187 struct fib6_node *fn;
2188 struct rt6_info *rt = NULL;
2189 struct fib6_table *table;
2190 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2192 table = fib6_get_table(dev_net(dev), tb_id);
2193 if (!table)
2194 return NULL;
2196 read_lock_bh(&table->tb6_lock);
2197 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2198 if (!fn)
2199 goto out;
2201 noflags |= RTF_CACHE;
2202 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2203 if (rt->dst.dev->ifindex != dev->ifindex)
2204 continue;
2205 if ((rt->rt6i_flags & flags) != flags)
2206 continue;
2207 if ((rt->rt6i_flags & noflags) != 0)
2208 continue;
2209 dst_hold(&rt->dst);
2210 break;
2212 out:
2213 read_unlock_bh(&table->tb6_lock);
2214 return rt;
2218 /* Create "default" multicast route to the interface */
2220 static void addrconf_add_mroute(struct net_device *dev)
2222 struct fib6_config cfg = {
2223 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2224 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2225 .fc_ifindex = dev->ifindex,
2226 .fc_dst_len = 8,
2227 .fc_flags = RTF_UP,
2228 .fc_nlinfo.nl_net = dev_net(dev),
2231 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2233 ip6_route_add(&cfg);
2236 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2238 struct inet6_dev *idev;
2240 ASSERT_RTNL();
2242 idev = ipv6_find_idev(dev);
2243 if (!idev)
2244 return ERR_PTR(-ENOBUFS);
2246 if (idev->cnf.disable_ipv6)
2247 return ERR_PTR(-EACCES);
2249 /* Add default multicast route */
2250 if (!(dev->flags & IFF_LOOPBACK))
2251 addrconf_add_mroute(dev);
2253 return idev;
2256 static void manage_tempaddrs(struct inet6_dev *idev,
2257 struct inet6_ifaddr *ifp,
2258 __u32 valid_lft, __u32 prefered_lft,
2259 bool create, unsigned long now)
2261 u32 flags;
2262 struct inet6_ifaddr *ift;
2264 read_lock_bh(&idev->lock);
2265 /* update all temporary addresses in the list */
2266 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2267 int age, max_valid, max_prefered;
2269 if (ifp != ift->ifpub)
2270 continue;
2272 /* RFC 4941 section 3.3:
2273 * If a received option will extend the lifetime of a public
2274 * address, the lifetimes of temporary addresses should
2275 * be extended, subject to the overall constraint that no
2276 * temporary addresses should ever remain "valid" or "preferred"
2277 * for a time longer than (TEMP_VALID_LIFETIME) or
2278 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2280 age = (now - ift->cstamp) / HZ;
2281 max_valid = idev->cnf.temp_valid_lft - age;
2282 if (max_valid < 0)
2283 max_valid = 0;
2285 max_prefered = idev->cnf.temp_prefered_lft -
2286 idev->cnf.max_desync_factor - age;
2287 if (max_prefered < 0)
2288 max_prefered = 0;
2290 if (valid_lft > max_valid)
2291 valid_lft = max_valid;
2293 if (prefered_lft > max_prefered)
2294 prefered_lft = max_prefered;
2296 spin_lock(&ift->lock);
2297 flags = ift->flags;
2298 ift->valid_lft = valid_lft;
2299 ift->prefered_lft = prefered_lft;
2300 ift->tstamp = now;
2301 if (prefered_lft > 0)
2302 ift->flags &= ~IFA_F_DEPRECATED;
2304 spin_unlock(&ift->lock);
2305 if (!(flags&IFA_F_TENTATIVE))
2306 ipv6_ifa_notify(0, ift);
2309 if ((create || list_empty(&idev->tempaddr_list)) &&
2310 idev->cnf.use_tempaddr > 0) {
2311 /* When a new public address is created as described
2312 * in [ADDRCONF], also create a new temporary address.
2313 * Also create a temporary address if it's enabled but
2314 * no temporary address currently exists.
2316 read_unlock_bh(&idev->lock);
2317 ipv6_create_tempaddr(ifp, NULL);
2318 } else {
2319 read_unlock_bh(&idev->lock);
2323 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2325 struct prefix_info *pinfo;
2326 __u32 valid_lft;
2327 __u32 prefered_lft;
2328 int addr_type;
2329 u32 addr_flags = 0;
2330 struct inet6_dev *in6_dev;
2331 struct net *net = dev_net(dev);
2333 pinfo = (struct prefix_info *) opt;
2335 if (len < sizeof(struct prefix_info)) {
2336 ADBG("addrconf: prefix option too short\n");
2337 return;
2341 * Validation checks ([ADDRCONF], page 19)
2344 addr_type = ipv6_addr_type(&pinfo->prefix);
2346 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2347 return;
2349 valid_lft = ntohl(pinfo->valid);
2350 prefered_lft = ntohl(pinfo->prefered);
2352 if (prefered_lft > valid_lft) {
2353 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2354 return;
2357 in6_dev = in6_dev_get(dev);
2359 if (!in6_dev) {
2360 net_dbg_ratelimited("addrconf: device %s not configured\n",
2361 dev->name);
2362 return;
2366 * Two things going on here:
2367 * 1) Add routes for on-link prefixes
2368 * 2) Configure prefixes with the auto flag set
2371 if (pinfo->onlink) {
2372 struct rt6_info *rt;
2373 unsigned long rt_expires;
2375 /* Avoid arithmetic overflow. Really, we could
2376 * save rt_expires in seconds, likely valid_lft,
2377 * but it would require division in fib gc, that it
2378 * not good.
2380 if (HZ > USER_HZ)
2381 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2382 else
2383 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2385 if (addrconf_finite_timeout(rt_expires))
2386 rt_expires *= HZ;
2388 rt = addrconf_get_prefix_route(&pinfo->prefix,
2389 pinfo->prefix_len,
2390 dev,
2391 RTF_ADDRCONF | RTF_PREFIX_RT,
2392 RTF_GATEWAY | RTF_DEFAULT);
2394 if (rt) {
2395 /* Autoconf prefix route */
2396 if (valid_lft == 0) {
2397 ip6_del_rt(rt);
2398 rt = NULL;
2399 } else if (addrconf_finite_timeout(rt_expires)) {
2400 /* not infinity */
2401 rt6_set_expires(rt, jiffies + rt_expires);
2402 } else {
2403 rt6_clean_expires(rt);
2405 } else if (valid_lft) {
2406 clock_t expires = 0;
2407 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2408 if (addrconf_finite_timeout(rt_expires)) {
2409 /* not infinity */
2410 flags |= RTF_EXPIRES;
2411 expires = jiffies_to_clock_t(rt_expires);
2413 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2414 dev, expires, flags);
2416 ip6_rt_put(rt);
2419 /* Try to figure out our local address for this prefix */
2421 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2422 struct inet6_ifaddr *ifp;
2423 struct in6_addr addr;
2424 int create = 0, update_lft = 0;
2425 bool tokenized = false;
2427 if (pinfo->prefix_len == 64) {
2428 memcpy(&addr, &pinfo->prefix, 8);
2430 if (!ipv6_addr_any(&in6_dev->token)) {
2431 read_lock_bh(&in6_dev->lock);
2432 memcpy(addr.s6_addr + 8,
2433 in6_dev->token.s6_addr + 8, 8);
2434 read_unlock_bh(&in6_dev->lock);
2435 tokenized = true;
2436 } else if (in6_dev->addr_gen_mode ==
2437 IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2438 !ipv6_generate_stable_address(&addr, 0,
2439 in6_dev)) {
2440 addr_flags |= IFA_F_STABLE_PRIVACY;
2441 goto ok;
2442 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2443 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2444 in6_dev_put(in6_dev);
2445 return;
2447 goto ok;
2449 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2450 pinfo->prefix_len);
2451 in6_dev_put(in6_dev);
2452 return;
2456 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2458 if (!ifp && valid_lft) {
2459 int max_addresses = in6_dev->cnf.max_addresses;
2461 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2462 if (in6_dev->cnf.optimistic_dad &&
2463 !net->ipv6.devconf_all->forwarding && sllao)
2464 addr_flags |= IFA_F_OPTIMISTIC;
2465 #endif
2467 /* Do not allow to create too much of autoconfigured
2468 * addresses; this would be too easy way to crash kernel.
2470 if (!max_addresses ||
2471 ipv6_count_addresses(in6_dev) < max_addresses)
2472 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2473 pinfo->prefix_len,
2474 addr_type&IPV6_ADDR_SCOPE_MASK,
2475 addr_flags, valid_lft,
2476 prefered_lft);
2478 if (IS_ERR_OR_NULL(ifp)) {
2479 in6_dev_put(in6_dev);
2480 return;
2483 update_lft = 0;
2484 create = 1;
2485 spin_lock_bh(&ifp->lock);
2486 ifp->flags |= IFA_F_MANAGETEMPADDR;
2487 ifp->cstamp = jiffies;
2488 ifp->tokenized = tokenized;
2489 spin_unlock_bh(&ifp->lock);
2490 addrconf_dad_start(ifp);
2493 if (ifp) {
2494 u32 flags;
2495 unsigned long now;
2496 u32 stored_lft;
2498 /* update lifetime (RFC2462 5.5.3 e) */
2499 spin_lock_bh(&ifp->lock);
2500 now = jiffies;
2501 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2502 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2503 else
2504 stored_lft = 0;
2505 if (!update_lft && !create && stored_lft) {
2506 const u32 minimum_lft = min_t(u32,
2507 stored_lft, MIN_VALID_LIFETIME);
2508 valid_lft = max(valid_lft, minimum_lft);
2510 /* RFC4862 Section 5.5.3e:
2511 * "Note that the preferred lifetime of the
2512 * corresponding address is always reset to
2513 * the Preferred Lifetime in the received
2514 * Prefix Information option, regardless of
2515 * whether the valid lifetime is also reset or
2516 * ignored."
2518 * So we should always update prefered_lft here.
2520 update_lft = 1;
2523 if (update_lft) {
2524 ifp->valid_lft = valid_lft;
2525 ifp->prefered_lft = prefered_lft;
2526 ifp->tstamp = now;
2527 flags = ifp->flags;
2528 ifp->flags &= ~IFA_F_DEPRECATED;
2529 spin_unlock_bh(&ifp->lock);
2531 if (!(flags&IFA_F_TENTATIVE))
2532 ipv6_ifa_notify(0, ifp);
2533 } else
2534 spin_unlock_bh(&ifp->lock);
2536 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2537 create, now);
2539 in6_ifa_put(ifp);
2540 addrconf_verify();
2543 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2544 in6_dev_put(in6_dev);
2548 * Set destination address.
2549 * Special case for SIT interfaces where we create a new "virtual"
2550 * device.
2552 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2554 struct in6_ifreq ireq;
2555 struct net_device *dev;
2556 int err = -EINVAL;
2558 rtnl_lock();
2560 err = -EFAULT;
2561 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2562 goto err_exit;
2564 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2566 err = -ENODEV;
2567 if (!dev)
2568 goto err_exit;
2570 #if IS_ENABLED(CONFIG_IPV6_SIT)
2571 if (dev->type == ARPHRD_SIT) {
2572 const struct net_device_ops *ops = dev->netdev_ops;
2573 struct ifreq ifr;
2574 struct ip_tunnel_parm p;
2576 err = -EADDRNOTAVAIL;
2577 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2578 goto err_exit;
2580 memset(&p, 0, sizeof(p));
2581 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2582 p.iph.saddr = 0;
2583 p.iph.version = 4;
2584 p.iph.ihl = 5;
2585 p.iph.protocol = IPPROTO_IPV6;
2586 p.iph.ttl = 64;
2587 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2589 if (ops->ndo_do_ioctl) {
2590 mm_segment_t oldfs = get_fs();
2592 set_fs(KERNEL_DS);
2593 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2594 set_fs(oldfs);
2595 } else
2596 err = -EOPNOTSUPP;
2598 if (err == 0) {
2599 err = -ENOBUFS;
2600 dev = __dev_get_by_name(net, p.name);
2601 if (!dev)
2602 goto err_exit;
2603 err = dev_open(dev);
2606 #endif
2608 err_exit:
2609 rtnl_unlock();
2610 return err;
2613 static int ipv6_mc_config(struct sock *sk, bool join,
2614 const struct in6_addr *addr, int ifindex)
2616 int ret;
2618 ASSERT_RTNL();
2620 lock_sock(sk);
2621 if (join)
2622 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2623 else
2624 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2625 release_sock(sk);
2627 return ret;
2631 * Manual configuration of address on an interface
2633 static int inet6_addr_add(struct net *net, int ifindex,
2634 const struct in6_addr *pfx,
2635 const struct in6_addr *peer_pfx,
2636 unsigned int plen, __u32 ifa_flags,
2637 __u32 prefered_lft, __u32 valid_lft)
2639 struct inet6_ifaddr *ifp;
2640 struct inet6_dev *idev;
2641 struct net_device *dev;
2642 unsigned long timeout;
2643 clock_t expires;
2644 int scope;
2645 u32 flags;
2647 ASSERT_RTNL();
2649 if (plen > 128)
2650 return -EINVAL;
2652 /* check the lifetime */
2653 if (!valid_lft || prefered_lft > valid_lft)
2654 return -EINVAL;
2656 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2657 return -EINVAL;
2659 dev = __dev_get_by_index(net, ifindex);
2660 if (!dev)
2661 return -ENODEV;
2663 idev = addrconf_add_dev(dev);
2664 if (IS_ERR(idev))
2665 return PTR_ERR(idev);
2667 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2668 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2669 true, pfx, ifindex);
2671 if (ret < 0)
2672 return ret;
2675 scope = ipv6_addr_scope(pfx);
2677 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2678 if (addrconf_finite_timeout(timeout)) {
2679 expires = jiffies_to_clock_t(timeout * HZ);
2680 valid_lft = timeout;
2681 flags = RTF_EXPIRES;
2682 } else {
2683 expires = 0;
2684 flags = 0;
2685 ifa_flags |= IFA_F_PERMANENT;
2688 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2689 if (addrconf_finite_timeout(timeout)) {
2690 if (timeout == 0)
2691 ifa_flags |= IFA_F_DEPRECATED;
2692 prefered_lft = timeout;
2695 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2696 valid_lft, prefered_lft);
2698 if (!IS_ERR(ifp)) {
2699 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2700 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2701 expires, flags);
2705 * Note that section 3.1 of RFC 4429 indicates
2706 * that the Optimistic flag should not be set for
2707 * manually configured addresses
2709 addrconf_dad_start(ifp);
2710 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2711 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2712 true, jiffies);
2713 in6_ifa_put(ifp);
2714 addrconf_verify_rtnl();
2715 return 0;
2716 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2717 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2718 false, pfx, ifindex);
2721 return PTR_ERR(ifp);
2724 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2725 const struct in6_addr *pfx, unsigned int plen)
2727 struct inet6_ifaddr *ifp;
2728 struct inet6_dev *idev;
2729 struct net_device *dev;
2731 if (plen > 128)
2732 return -EINVAL;
2734 dev = __dev_get_by_index(net, ifindex);
2735 if (!dev)
2736 return -ENODEV;
2738 idev = __in6_dev_get(dev);
2739 if (!idev)
2740 return -ENXIO;
2742 read_lock_bh(&idev->lock);
2743 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2744 if (ifp->prefix_len == plen &&
2745 ipv6_addr_equal(pfx, &ifp->addr)) {
2746 in6_ifa_hold(ifp);
2747 read_unlock_bh(&idev->lock);
2749 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2750 (ifa_flags & IFA_F_MANAGETEMPADDR))
2751 manage_tempaddrs(idev, ifp, 0, 0, false,
2752 jiffies);
2753 ipv6_del_addr(ifp);
2754 addrconf_verify_rtnl();
2755 if (ipv6_addr_is_multicast(pfx)) {
2756 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2757 false, pfx, dev->ifindex);
2759 return 0;
2762 read_unlock_bh(&idev->lock);
2763 return -EADDRNOTAVAIL;
2767 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2769 struct in6_ifreq ireq;
2770 int err;
2772 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2773 return -EPERM;
2775 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2776 return -EFAULT;
2778 rtnl_lock();
2779 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2780 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2781 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2782 rtnl_unlock();
2783 return err;
2786 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2788 struct in6_ifreq ireq;
2789 int err;
2791 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2792 return -EPERM;
2794 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2795 return -EFAULT;
2797 rtnl_lock();
2798 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2799 ireq.ifr6_prefixlen);
2800 rtnl_unlock();
2801 return err;
2804 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2805 int plen, int scope)
2807 struct inet6_ifaddr *ifp;
2809 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2810 scope, IFA_F_PERMANENT,
2811 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2812 if (!IS_ERR(ifp)) {
2813 spin_lock_bh(&ifp->lock);
2814 ifp->flags &= ~IFA_F_TENTATIVE;
2815 spin_unlock_bh(&ifp->lock);
2816 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2817 in6_ifa_put(ifp);
2821 #if IS_ENABLED(CONFIG_IPV6_SIT)
2822 static void sit_add_v4_addrs(struct inet6_dev *idev)
2824 struct in6_addr addr;
2825 struct net_device *dev;
2826 struct net *net = dev_net(idev->dev);
2827 int scope, plen;
2828 u32 pflags = 0;
2830 ASSERT_RTNL();
2832 memset(&addr, 0, sizeof(struct in6_addr));
2833 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2835 if (idev->dev->flags&IFF_POINTOPOINT) {
2836 addr.s6_addr32[0] = htonl(0xfe800000);
2837 scope = IFA_LINK;
2838 plen = 64;
2839 } else {
2840 scope = IPV6_ADDR_COMPATv4;
2841 plen = 96;
2842 pflags |= RTF_NONEXTHOP;
2845 if (addr.s6_addr32[3]) {
2846 add_addr(idev, &addr, plen, scope);
2847 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2848 return;
2851 for_each_netdev(net, dev) {
2852 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2853 if (in_dev && (dev->flags & IFF_UP)) {
2854 struct in_ifaddr *ifa;
2856 int flag = scope;
2858 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2860 addr.s6_addr32[3] = ifa->ifa_local;
2862 if (ifa->ifa_scope == RT_SCOPE_LINK)
2863 continue;
2864 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2865 if (idev->dev->flags&IFF_POINTOPOINT)
2866 continue;
2867 flag |= IFA_HOST;
2870 add_addr(idev, &addr, plen, flag);
2871 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2872 pflags);
2877 #endif
2879 static void init_loopback(struct net_device *dev)
2881 struct inet6_dev *idev;
2882 struct net_device *sp_dev;
2883 struct inet6_ifaddr *sp_ifa;
2884 struct rt6_info *sp_rt;
2886 /* ::1 */
2888 ASSERT_RTNL();
2890 idev = ipv6_find_idev(dev);
2891 if (!idev) {
2892 pr_debug("%s: add_dev failed\n", __func__);
2893 return;
2896 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2898 /* Add routes to other interface's IPv6 addresses */
2899 for_each_netdev(dev_net(dev), sp_dev) {
2900 if (!strcmp(sp_dev->name, dev->name))
2901 continue;
2903 idev = __in6_dev_get(sp_dev);
2904 if (!idev)
2905 continue;
2907 read_lock_bh(&idev->lock);
2908 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2910 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2911 continue;
2913 if (sp_ifa->rt) {
2914 /* This dst has been added to garbage list when
2915 * lo device down, release this obsolete dst and
2916 * reallocate a new router for ifa.
2918 if (sp_ifa->rt->dst.obsolete > 0) {
2919 ip6_rt_put(sp_ifa->rt);
2920 sp_ifa->rt = NULL;
2921 } else {
2922 continue;
2926 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2928 /* Failure cases are ignored */
2929 if (!IS_ERR(sp_rt)) {
2930 sp_ifa->rt = sp_rt;
2931 ip6_ins_rt(sp_rt);
2934 read_unlock_bh(&idev->lock);
2938 static void addrconf_add_linklocal(struct inet6_dev *idev,
2939 const struct in6_addr *addr, u32 flags)
2941 struct inet6_ifaddr *ifp;
2942 u32 addr_flags = flags | IFA_F_PERMANENT;
2944 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2945 if (idev->cnf.optimistic_dad &&
2946 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2947 addr_flags |= IFA_F_OPTIMISTIC;
2948 #endif
2950 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2951 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2952 if (!IS_ERR(ifp)) {
2953 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2954 addrconf_dad_start(ifp);
2955 in6_ifa_put(ifp);
2959 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2961 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2962 return true;
2964 if (address.s6_addr32[2] == htonl(0x02005eff) &&
2965 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2966 return true;
2968 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2969 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2970 return true;
2972 return false;
2975 static int ipv6_generate_stable_address(struct in6_addr *address,
2976 u8 dad_count,
2977 const struct inet6_dev *idev)
2979 static DEFINE_SPINLOCK(lock);
2980 static __u32 digest[SHA_DIGEST_WORDS];
2981 static __u32 workspace[SHA_WORKSPACE_WORDS];
2983 static union {
2984 char __data[SHA_MESSAGE_BYTES];
2985 struct {
2986 struct in6_addr secret;
2987 __be32 prefix[2];
2988 unsigned char hwaddr[MAX_ADDR_LEN];
2989 u8 dad_count;
2990 } __packed;
2991 } data;
2993 struct in6_addr secret;
2994 struct in6_addr temp;
2995 struct net *net = dev_net(idev->dev);
2997 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
2999 if (idev->cnf.stable_secret.initialized)
3000 secret = idev->cnf.stable_secret.secret;
3001 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3002 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3003 else
3004 return -1;
3006 retry:
3007 spin_lock_bh(&lock);
3009 sha_init(digest);
3010 memset(&data, 0, sizeof(data));
3011 memset(workspace, 0, sizeof(workspace));
3012 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3013 data.prefix[0] = address->s6_addr32[0];
3014 data.prefix[1] = address->s6_addr32[1];
3015 data.secret = secret;
3016 data.dad_count = dad_count;
3018 sha_transform(digest, data.__data, workspace);
3020 temp = *address;
3021 temp.s6_addr32[2] = (__force __be32)digest[0];
3022 temp.s6_addr32[3] = (__force __be32)digest[1];
3024 spin_unlock_bh(&lock);
3026 if (ipv6_reserved_interfaceid(temp)) {
3027 dad_count++;
3028 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3029 return -1;
3030 goto retry;
3033 *address = temp;
3034 return 0;
3037 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3039 struct in6_addr addr;
3041 /* no link local addresses on L3 master devices */
3042 if (netif_is_l3_master(idev->dev))
3043 return;
3045 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3047 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3048 if (!ipv6_generate_stable_address(&addr, 0, idev))
3049 addrconf_add_linklocal(idev, &addr,
3050 IFA_F_STABLE_PRIVACY);
3051 else if (prefix_route)
3052 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3053 } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3054 /* addrconf_add_linklocal also adds a prefix_route and we
3055 * only need to care about prefix routes if ipv6_generate_eui64
3056 * couldn't generate one.
3058 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3059 addrconf_add_linklocal(idev, &addr, 0);
3060 else if (prefix_route)
3061 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3065 static void addrconf_dev_config(struct net_device *dev)
3067 struct inet6_dev *idev;
3069 ASSERT_RTNL();
3071 if ((dev->type != ARPHRD_ETHER) &&
3072 (dev->type != ARPHRD_FDDI) &&
3073 (dev->type != ARPHRD_ARCNET) &&
3074 (dev->type != ARPHRD_INFINIBAND) &&
3075 (dev->type != ARPHRD_IEEE802154) &&
3076 (dev->type != ARPHRD_IEEE1394) &&
3077 (dev->type != ARPHRD_TUNNEL6) &&
3078 (dev->type != ARPHRD_6LOWPAN)) {
3079 /* Alas, we support only Ethernet autoconfiguration. */
3080 return;
3083 idev = addrconf_add_dev(dev);
3084 if (IS_ERR(idev))
3085 return;
3087 addrconf_addr_gen(idev, false);
3090 #if IS_ENABLED(CONFIG_IPV6_SIT)
3091 static void addrconf_sit_config(struct net_device *dev)
3093 struct inet6_dev *idev;
3095 ASSERT_RTNL();
3098 * Configure the tunnel with one of our IPv4
3099 * addresses... we should configure all of
3100 * our v4 addrs in the tunnel
3103 idev = ipv6_find_idev(dev);
3104 if (!idev) {
3105 pr_debug("%s: add_dev failed\n", __func__);
3106 return;
3109 if (dev->priv_flags & IFF_ISATAP) {
3110 addrconf_addr_gen(idev, false);
3111 return;
3114 sit_add_v4_addrs(idev);
3116 if (dev->flags&IFF_POINTOPOINT)
3117 addrconf_add_mroute(dev);
3119 #endif
3121 #if IS_ENABLED(CONFIG_NET_IPGRE)
3122 static void addrconf_gre_config(struct net_device *dev)
3124 struct inet6_dev *idev;
3126 ASSERT_RTNL();
3128 idev = ipv6_find_idev(dev);
3129 if (!idev) {
3130 pr_debug("%s: add_dev failed\n", __func__);
3131 return;
3134 addrconf_addr_gen(idev, true);
3135 if (dev->flags & IFF_POINTOPOINT)
3136 addrconf_add_mroute(dev);
3138 #endif
3140 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3141 void *ptr)
3143 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3144 struct inet6_dev *idev = __in6_dev_get(dev);
3145 int run_pending = 0;
3146 int err;
3148 switch (event) {
3149 case NETDEV_REGISTER:
3150 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3151 idev = ipv6_add_dev(dev);
3152 if (IS_ERR(idev))
3153 return notifier_from_errno(PTR_ERR(idev));
3155 break;
3157 case NETDEV_CHANGEMTU:
3158 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3159 if (dev->mtu < IPV6_MIN_MTU) {
3160 addrconf_ifdown(dev, 1);
3161 break;
3164 if (idev) {
3165 rt6_mtu_change(dev, dev->mtu);
3166 idev->cnf.mtu6 = dev->mtu;
3167 break;
3170 /* allocate new idev */
3171 idev = ipv6_add_dev(dev);
3172 if (IS_ERR(idev))
3173 break;
3175 /* device is still not ready */
3176 if (!(idev->if_flags & IF_READY))
3177 break;
3179 run_pending = 1;
3181 /* fall through */
3183 case NETDEV_UP:
3184 case NETDEV_CHANGE:
3185 if (dev->flags & IFF_SLAVE)
3186 break;
3188 if (idev && idev->cnf.disable_ipv6)
3189 break;
3191 if (event == NETDEV_UP) {
3192 if (!addrconf_qdisc_ok(dev)) {
3193 /* device is not ready yet. */
3194 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3195 dev->name);
3196 break;
3199 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3200 idev = ipv6_add_dev(dev);
3202 if (!IS_ERR_OR_NULL(idev)) {
3203 idev->if_flags |= IF_READY;
3204 run_pending = 1;
3206 } else if (event == NETDEV_CHANGE) {
3207 if (!addrconf_qdisc_ok(dev)) {
3208 /* device is still not ready. */
3209 break;
3212 if (idev) {
3213 if (idev->if_flags & IF_READY)
3214 /* device is already configured. */
3215 break;
3216 idev->if_flags |= IF_READY;
3219 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3220 dev->name);
3222 run_pending = 1;
3225 switch (dev->type) {
3226 #if IS_ENABLED(CONFIG_IPV6_SIT)
3227 case ARPHRD_SIT:
3228 addrconf_sit_config(dev);
3229 break;
3230 #endif
3231 #if IS_ENABLED(CONFIG_NET_IPGRE)
3232 case ARPHRD_IPGRE:
3233 addrconf_gre_config(dev);
3234 break;
3235 #endif
3236 case ARPHRD_LOOPBACK:
3237 init_loopback(dev);
3238 break;
3240 default:
3241 addrconf_dev_config(dev);
3242 break;
3245 if (!IS_ERR_OR_NULL(idev)) {
3246 if (run_pending)
3247 addrconf_dad_run(idev);
3250 * If the MTU changed during the interface down,
3251 * when the interface up, the changed MTU must be
3252 * reflected in the idev as well as routers.
3254 if (idev->cnf.mtu6 != dev->mtu &&
3255 dev->mtu >= IPV6_MIN_MTU) {
3256 rt6_mtu_change(dev, dev->mtu);
3257 idev->cnf.mtu6 = dev->mtu;
3259 idev->tstamp = jiffies;
3260 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3263 * If the changed mtu during down is lower than
3264 * IPV6_MIN_MTU stop IPv6 on this interface.
3266 if (dev->mtu < IPV6_MIN_MTU)
3267 addrconf_ifdown(dev, 1);
3269 break;
3271 case NETDEV_DOWN:
3272 case NETDEV_UNREGISTER:
3274 * Remove all addresses from this interface.
3276 addrconf_ifdown(dev, event != NETDEV_DOWN);
3277 break;
3279 case NETDEV_CHANGENAME:
3280 if (idev) {
3281 snmp6_unregister_dev(idev);
3282 addrconf_sysctl_unregister(idev);
3283 err = addrconf_sysctl_register(idev);
3284 if (err)
3285 return notifier_from_errno(err);
3286 err = snmp6_register_dev(idev);
3287 if (err) {
3288 addrconf_sysctl_unregister(idev);
3289 return notifier_from_errno(err);
3292 break;
3294 case NETDEV_PRE_TYPE_CHANGE:
3295 case NETDEV_POST_TYPE_CHANGE:
3296 addrconf_type_change(dev, event);
3297 break;
3300 return NOTIFY_OK;
3304 * addrconf module should be notified of a device going up
3306 static struct notifier_block ipv6_dev_notf = {
3307 .notifier_call = addrconf_notify,
3310 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3312 struct inet6_dev *idev;
3313 ASSERT_RTNL();
3315 idev = __in6_dev_get(dev);
3317 if (event == NETDEV_POST_TYPE_CHANGE)
3318 ipv6_mc_remap(idev);
3319 else if (event == NETDEV_PRE_TYPE_CHANGE)
3320 ipv6_mc_unmap(idev);
3323 static int addrconf_ifdown(struct net_device *dev, int how)
3325 struct net *net = dev_net(dev);
3326 struct inet6_dev *idev;
3327 struct inet6_ifaddr *ifa;
3328 int state, i;
3330 ASSERT_RTNL();
3332 rt6_ifdown(net, dev);
3333 neigh_ifdown(&nd_tbl, dev);
3335 idev = __in6_dev_get(dev);
3336 if (!idev)
3337 return -ENODEV;
3340 * Step 1: remove reference to ipv6 device from parent device.
3341 * Do not dev_put!
3343 if (how) {
3344 idev->dead = 1;
3346 /* protected by rtnl_lock */
3347 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3349 /* Step 1.5: remove snmp6 entry */
3350 snmp6_unregister_dev(idev);
3354 /* Step 2: clear hash table */
3355 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3356 struct hlist_head *h = &inet6_addr_lst[i];
3358 spin_lock_bh(&addrconf_hash_lock);
3359 restart:
3360 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3361 if (ifa->idev == idev) {
3362 hlist_del_init_rcu(&ifa->addr_lst);
3363 addrconf_del_dad_work(ifa);
3364 goto restart;
3367 spin_unlock_bh(&addrconf_hash_lock);
3370 write_lock_bh(&idev->lock);
3372 addrconf_del_rs_timer(idev);
3374 /* Step 2: clear flags for stateless addrconf */
3375 if (!how)
3376 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3378 if (how && del_timer(&idev->regen_timer))
3379 in6_dev_put(idev);
3381 /* Step 3: clear tempaddr list */
3382 while (!list_empty(&idev->tempaddr_list)) {
3383 ifa = list_first_entry(&idev->tempaddr_list,
3384 struct inet6_ifaddr, tmp_list);
3385 list_del(&ifa->tmp_list);
3386 write_unlock_bh(&idev->lock);
3387 spin_lock_bh(&ifa->lock);
3389 if (ifa->ifpub) {
3390 in6_ifa_put(ifa->ifpub);
3391 ifa->ifpub = NULL;
3393 spin_unlock_bh(&ifa->lock);
3394 in6_ifa_put(ifa);
3395 write_lock_bh(&idev->lock);
3398 while (!list_empty(&idev->addr_list)) {
3399 ifa = list_first_entry(&idev->addr_list,
3400 struct inet6_ifaddr, if_list);
3401 addrconf_del_dad_work(ifa);
3403 list_del(&ifa->if_list);
3405 write_unlock_bh(&idev->lock);
3407 spin_lock_bh(&ifa->lock);
3408 state = ifa->state;
3409 ifa->state = INET6_IFADDR_STATE_DEAD;
3410 spin_unlock_bh(&ifa->lock);
3412 if (state != INET6_IFADDR_STATE_DEAD) {
3413 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3414 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3416 in6_ifa_put(ifa);
3418 write_lock_bh(&idev->lock);
3421 write_unlock_bh(&idev->lock);
3423 /* Step 5: Discard anycast and multicast list */
3424 if (how) {
3425 ipv6_ac_destroy_dev(idev);
3426 ipv6_mc_destroy_dev(idev);
3427 } else {
3428 ipv6_mc_down(idev);
3431 idev->tstamp = jiffies;
3433 /* Last: Shot the device (if unregistered) */
3434 if (how) {
3435 addrconf_sysctl_unregister(idev);
3436 neigh_parms_release(&nd_tbl, idev->nd_parms);
3437 neigh_ifdown(&nd_tbl, dev);
3438 in6_dev_put(idev);
3440 return 0;
3443 static void addrconf_rs_timer(unsigned long data)
3445 struct inet6_dev *idev = (struct inet6_dev *)data;
3446 struct net_device *dev = idev->dev;
3447 struct in6_addr lladdr;
3449 write_lock(&idev->lock);
3450 if (idev->dead || !(idev->if_flags & IF_READY))
3451 goto out;
3453 if (!ipv6_accept_ra(idev))
3454 goto out;
3456 /* Announcement received after solicitation was sent */
3457 if (idev->if_flags & IF_RA_RCVD)
3458 goto out;
3460 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3461 write_unlock(&idev->lock);
3462 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3463 ndisc_send_rs(dev, &lladdr,
3464 &in6addr_linklocal_allrouters);
3465 else
3466 goto put;
3468 write_lock(&idev->lock);
3469 /* The wait after the last probe can be shorter */
3470 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3471 idev->cnf.rtr_solicits) ?
3472 idev->cnf.rtr_solicit_delay :
3473 idev->cnf.rtr_solicit_interval);
3474 } else {
3476 * Note: we do not support deprecated "all on-link"
3477 * assumption any longer.
3479 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3482 out:
3483 write_unlock(&idev->lock);
3484 put:
3485 in6_dev_put(idev);
3489 * Duplicate Address Detection
3491 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3493 unsigned long rand_num;
3494 struct inet6_dev *idev = ifp->idev;
3496 if (ifp->flags & IFA_F_OPTIMISTIC)
3497 rand_num = 0;
3498 else
3499 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3501 ifp->dad_probes = idev->cnf.dad_transmits;
3502 addrconf_mod_dad_work(ifp, rand_num);
3505 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3507 struct inet6_dev *idev = ifp->idev;
3508 struct net_device *dev = idev->dev;
3510 addrconf_join_solict(dev, &ifp->addr);
3512 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3514 read_lock_bh(&idev->lock);
3515 spin_lock(&ifp->lock);
3516 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3517 goto out;
3519 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3520 idev->cnf.accept_dad < 1 ||
3521 !(ifp->flags&IFA_F_TENTATIVE) ||
3522 ifp->flags & IFA_F_NODAD) {
3523 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3524 spin_unlock(&ifp->lock);
3525 read_unlock_bh(&idev->lock);
3527 addrconf_dad_completed(ifp);
3528 return;
3531 if (!(idev->if_flags & IF_READY)) {
3532 spin_unlock(&ifp->lock);
3533 read_unlock_bh(&idev->lock);
3535 * If the device is not ready:
3536 * - keep it tentative if it is a permanent address.
3537 * - otherwise, kill it.
3539 in6_ifa_hold(ifp);
3540 addrconf_dad_stop(ifp, 0);
3541 return;
3545 * Optimistic nodes can start receiving
3546 * Frames right away
3548 if (ifp->flags & IFA_F_OPTIMISTIC) {
3549 ip6_ins_rt(ifp->rt);
3550 if (ipv6_use_optimistic_addr(idev)) {
3551 /* Because optimistic nodes can use this address,
3552 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3554 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3558 addrconf_dad_kick(ifp);
3559 out:
3560 spin_unlock(&ifp->lock);
3561 read_unlock_bh(&idev->lock);
3564 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3566 bool begin_dad = false;
3568 spin_lock_bh(&ifp->lock);
3569 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3570 ifp->state = INET6_IFADDR_STATE_PREDAD;
3571 begin_dad = true;
3573 spin_unlock_bh(&ifp->lock);
3575 if (begin_dad)
3576 addrconf_mod_dad_work(ifp, 0);
3579 static void addrconf_dad_work(struct work_struct *w)
3581 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3582 struct inet6_ifaddr,
3583 dad_work);
3584 struct inet6_dev *idev = ifp->idev;
3585 struct in6_addr mcaddr;
3587 enum {
3588 DAD_PROCESS,
3589 DAD_BEGIN,
3590 DAD_ABORT,
3591 } action = DAD_PROCESS;
3593 rtnl_lock();
3595 spin_lock_bh(&ifp->lock);
3596 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3597 action = DAD_BEGIN;
3598 ifp->state = INET6_IFADDR_STATE_DAD;
3599 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3600 action = DAD_ABORT;
3601 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3603 spin_unlock_bh(&ifp->lock);
3605 if (action == DAD_BEGIN) {
3606 addrconf_dad_begin(ifp);
3607 goto out;
3608 } else if (action == DAD_ABORT) {
3609 addrconf_dad_stop(ifp, 1);
3610 goto out;
3613 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3614 goto out;
3616 write_lock_bh(&idev->lock);
3617 if (idev->dead || !(idev->if_flags & IF_READY)) {
3618 write_unlock_bh(&idev->lock);
3619 goto out;
3622 spin_lock(&ifp->lock);
3623 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3624 spin_unlock(&ifp->lock);
3625 write_unlock_bh(&idev->lock);
3626 goto out;
3629 if (ifp->dad_probes == 0) {
3631 * DAD was successful
3634 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3635 spin_unlock(&ifp->lock);
3636 write_unlock_bh(&idev->lock);
3638 addrconf_dad_completed(ifp);
3640 goto out;
3643 ifp->dad_probes--;
3644 addrconf_mod_dad_work(ifp,
3645 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3646 spin_unlock(&ifp->lock);
3647 write_unlock_bh(&idev->lock);
3649 /* send a neighbour solicitation for our addr */
3650 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3651 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3652 out:
3653 in6_ifa_put(ifp);
3654 rtnl_unlock();
3657 /* ifp->idev must be at least read locked */
3658 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3660 struct inet6_ifaddr *ifpiter;
3661 struct inet6_dev *idev = ifp->idev;
3663 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3664 if (ifpiter->scope > IFA_LINK)
3665 break;
3666 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3667 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3668 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3669 IFA_F_PERMANENT)
3670 return false;
3672 return true;
3675 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3677 struct net_device *dev = ifp->idev->dev;
3678 struct in6_addr lladdr;
3679 bool send_rs, send_mld;
3681 addrconf_del_dad_work(ifp);
3684 * Configure the address for reception. Now it is valid.
3687 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3689 /* If added prefix is link local and we are prepared to process
3690 router advertisements, start sending router solicitations.
3693 read_lock_bh(&ifp->idev->lock);
3694 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3695 send_rs = send_mld &&
3696 ipv6_accept_ra(ifp->idev) &&
3697 ifp->idev->cnf.rtr_solicits > 0 &&
3698 (dev->flags&IFF_LOOPBACK) == 0;
3699 read_unlock_bh(&ifp->idev->lock);
3701 /* While dad is in progress mld report's source address is in6_addrany.
3702 * Resend with proper ll now.
3704 if (send_mld)
3705 ipv6_mc_dad_complete(ifp->idev);
3707 if (send_rs) {
3709 * If a host as already performed a random delay
3710 * [...] as part of DAD [...] there is no need
3711 * to delay again before sending the first RS
3713 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3714 return;
3715 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3717 write_lock_bh(&ifp->idev->lock);
3718 spin_lock(&ifp->lock);
3719 ifp->idev->rs_probes = 1;
3720 ifp->idev->if_flags |= IF_RS_SENT;
3721 addrconf_mod_rs_timer(ifp->idev,
3722 ifp->idev->cnf.rtr_solicit_interval);
3723 spin_unlock(&ifp->lock);
3724 write_unlock_bh(&ifp->idev->lock);
3728 static void addrconf_dad_run(struct inet6_dev *idev)
3730 struct inet6_ifaddr *ifp;
3732 read_lock_bh(&idev->lock);
3733 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3734 spin_lock(&ifp->lock);
3735 if (ifp->flags & IFA_F_TENTATIVE &&
3736 ifp->state == INET6_IFADDR_STATE_DAD)
3737 addrconf_dad_kick(ifp);
3738 spin_unlock(&ifp->lock);
3740 read_unlock_bh(&idev->lock);
3743 #ifdef CONFIG_PROC_FS
3744 struct if6_iter_state {
3745 struct seq_net_private p;
3746 int bucket;
3747 int offset;
3750 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3752 struct inet6_ifaddr *ifa = NULL;
3753 struct if6_iter_state *state = seq->private;
3754 struct net *net = seq_file_net(seq);
3755 int p = 0;
3757 /* initial bucket if pos is 0 */
3758 if (pos == 0) {
3759 state->bucket = 0;
3760 state->offset = 0;
3763 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3764 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3765 addr_lst) {
3766 if (!net_eq(dev_net(ifa->idev->dev), net))
3767 continue;
3768 /* sync with offset */
3769 if (p < state->offset) {
3770 p++;
3771 continue;
3773 state->offset++;
3774 return ifa;
3777 /* prepare for next bucket */
3778 state->offset = 0;
3779 p = 0;
3781 return NULL;
3784 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3785 struct inet6_ifaddr *ifa)
3787 struct if6_iter_state *state = seq->private;
3788 struct net *net = seq_file_net(seq);
3790 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3791 if (!net_eq(dev_net(ifa->idev->dev), net))
3792 continue;
3793 state->offset++;
3794 return ifa;
3797 while (++state->bucket < IN6_ADDR_HSIZE) {
3798 state->offset = 0;
3799 hlist_for_each_entry_rcu_bh(ifa,
3800 &inet6_addr_lst[state->bucket], addr_lst) {
3801 if (!net_eq(dev_net(ifa->idev->dev), net))
3802 continue;
3803 state->offset++;
3804 return ifa;
3808 return NULL;
3811 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3812 __acquires(rcu_bh)
3814 rcu_read_lock_bh();
3815 return if6_get_first(seq, *pos);
3818 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3820 struct inet6_ifaddr *ifa;
3822 ifa = if6_get_next(seq, v);
3823 ++*pos;
3824 return ifa;
3827 static void if6_seq_stop(struct seq_file *seq, void *v)
3828 __releases(rcu_bh)
3830 rcu_read_unlock_bh();
3833 static int if6_seq_show(struct seq_file *seq, void *v)
3835 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3836 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3837 &ifp->addr,
3838 ifp->idev->dev->ifindex,
3839 ifp->prefix_len,
3840 ifp->scope,
3841 (u8) ifp->flags,
3842 ifp->idev->dev->name);
3843 return 0;
3846 static const struct seq_operations if6_seq_ops = {
3847 .start = if6_seq_start,
3848 .next = if6_seq_next,
3849 .show = if6_seq_show,
3850 .stop = if6_seq_stop,
3853 static int if6_seq_open(struct inode *inode, struct file *file)
3855 return seq_open_net(inode, file, &if6_seq_ops,
3856 sizeof(struct if6_iter_state));
3859 static const struct file_operations if6_fops = {
3860 .owner = THIS_MODULE,
3861 .open = if6_seq_open,
3862 .read = seq_read,
3863 .llseek = seq_lseek,
3864 .release = seq_release_net,
3867 static int __net_init if6_proc_net_init(struct net *net)
3869 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3870 return -ENOMEM;
3871 return 0;
3874 static void __net_exit if6_proc_net_exit(struct net *net)
3876 remove_proc_entry("if_inet6", net->proc_net);
3879 static struct pernet_operations if6_proc_net_ops = {
3880 .init = if6_proc_net_init,
3881 .exit = if6_proc_net_exit,
3884 int __init if6_proc_init(void)
3886 return register_pernet_subsys(&if6_proc_net_ops);
3889 void if6_proc_exit(void)
3891 unregister_pernet_subsys(&if6_proc_net_ops);
3893 #endif /* CONFIG_PROC_FS */
3895 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3896 /* Check if address is a home address configured on any interface. */
3897 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3899 int ret = 0;
3900 struct inet6_ifaddr *ifp = NULL;
3901 unsigned int hash = inet6_addr_hash(addr);
3903 rcu_read_lock_bh();
3904 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3905 if (!net_eq(dev_net(ifp->idev->dev), net))
3906 continue;
3907 if (ipv6_addr_equal(&ifp->addr, addr) &&
3908 (ifp->flags & IFA_F_HOMEADDRESS)) {
3909 ret = 1;
3910 break;
3913 rcu_read_unlock_bh();
3914 return ret;
3916 #endif
3919 * Periodic address status verification
3922 static void addrconf_verify_rtnl(void)
3924 unsigned long now, next, next_sec, next_sched;
3925 struct inet6_ifaddr *ifp;
3926 int i;
3928 ASSERT_RTNL();
3930 rcu_read_lock_bh();
3931 now = jiffies;
3932 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3934 cancel_delayed_work(&addr_chk_work);
3936 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3937 restart:
3938 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3939 unsigned long age;
3941 /* When setting preferred_lft to a value not zero or
3942 * infinity, while valid_lft is infinity
3943 * IFA_F_PERMANENT has a non-infinity life time.
3945 if ((ifp->flags & IFA_F_PERMANENT) &&
3946 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3947 continue;
3949 spin_lock(&ifp->lock);
3950 /* We try to batch several events at once. */
3951 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3953 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3954 age >= ifp->valid_lft) {
3955 spin_unlock(&ifp->lock);
3956 in6_ifa_hold(ifp);
3957 ipv6_del_addr(ifp);
3958 goto restart;
3959 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3960 spin_unlock(&ifp->lock);
3961 continue;
3962 } else if (age >= ifp->prefered_lft) {
3963 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3964 int deprecate = 0;
3966 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3967 deprecate = 1;
3968 ifp->flags |= IFA_F_DEPRECATED;
3971 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3972 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3973 next = ifp->tstamp + ifp->valid_lft * HZ;
3975 spin_unlock(&ifp->lock);
3977 if (deprecate) {
3978 in6_ifa_hold(ifp);
3980 ipv6_ifa_notify(0, ifp);
3981 in6_ifa_put(ifp);
3982 goto restart;
3984 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3985 !(ifp->flags&IFA_F_TENTATIVE)) {
3986 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3987 ifp->idev->cnf.dad_transmits *
3988 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3990 if (age >= ifp->prefered_lft - regen_advance) {
3991 struct inet6_ifaddr *ifpub = ifp->ifpub;
3992 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3993 next = ifp->tstamp + ifp->prefered_lft * HZ;
3994 if (!ifp->regen_count && ifpub) {
3995 ifp->regen_count++;
3996 in6_ifa_hold(ifp);
3997 in6_ifa_hold(ifpub);
3998 spin_unlock(&ifp->lock);
4000 spin_lock(&ifpub->lock);
4001 ifpub->regen_count = 0;
4002 spin_unlock(&ifpub->lock);
4003 ipv6_create_tempaddr(ifpub, ifp);
4004 in6_ifa_put(ifpub);
4005 in6_ifa_put(ifp);
4006 goto restart;
4008 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4009 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4010 spin_unlock(&ifp->lock);
4011 } else {
4012 /* ifp->prefered_lft <= ifp->valid_lft */
4013 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4014 next = ifp->tstamp + ifp->prefered_lft * HZ;
4015 spin_unlock(&ifp->lock);
4020 next_sec = round_jiffies_up(next);
4021 next_sched = next;
4023 /* If rounded timeout is accurate enough, accept it. */
4024 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4025 next_sched = next_sec;
4027 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4028 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4029 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4031 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4032 now, next, next_sec, next_sched);
4033 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4034 rcu_read_unlock_bh();
4037 static void addrconf_verify_work(struct work_struct *w)
4039 rtnl_lock();
4040 addrconf_verify_rtnl();
4041 rtnl_unlock();
4044 static void addrconf_verify(void)
4046 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4049 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4050 struct in6_addr **peer_pfx)
4052 struct in6_addr *pfx = NULL;
4054 *peer_pfx = NULL;
4056 if (addr)
4057 pfx = nla_data(addr);
4059 if (local) {
4060 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4061 *peer_pfx = pfx;
4062 pfx = nla_data(local);
4065 return pfx;
4068 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4069 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4070 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4071 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4072 [IFA_FLAGS] = { .len = sizeof(u32) },
4075 static int
4076 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4078 struct net *net = sock_net(skb->sk);
4079 struct ifaddrmsg *ifm;
4080 struct nlattr *tb[IFA_MAX+1];
4081 struct in6_addr *pfx, *peer_pfx;
4082 u32 ifa_flags;
4083 int err;
4085 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4086 if (err < 0)
4087 return err;
4089 ifm = nlmsg_data(nlh);
4090 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4091 if (!pfx)
4092 return -EINVAL;
4094 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4096 /* We ignore other flags so far. */
4097 ifa_flags &= IFA_F_MANAGETEMPADDR;
4099 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4100 ifm->ifa_prefixlen);
4103 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4104 u32 prefered_lft, u32 valid_lft)
4106 u32 flags;
4107 clock_t expires;
4108 unsigned long timeout;
4109 bool was_managetempaddr;
4110 bool had_prefixroute;
4112 ASSERT_RTNL();
4114 if (!valid_lft || (prefered_lft > valid_lft))
4115 return -EINVAL;
4117 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4118 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4119 return -EINVAL;
4121 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4122 if (addrconf_finite_timeout(timeout)) {
4123 expires = jiffies_to_clock_t(timeout * HZ);
4124 valid_lft = timeout;
4125 flags = RTF_EXPIRES;
4126 } else {
4127 expires = 0;
4128 flags = 0;
4129 ifa_flags |= IFA_F_PERMANENT;
4132 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4133 if (addrconf_finite_timeout(timeout)) {
4134 if (timeout == 0)
4135 ifa_flags |= IFA_F_DEPRECATED;
4136 prefered_lft = timeout;
4139 spin_lock_bh(&ifp->lock);
4140 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4141 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4142 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4143 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4144 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4145 IFA_F_NOPREFIXROUTE);
4146 ifp->flags |= ifa_flags;
4147 ifp->tstamp = jiffies;
4148 ifp->valid_lft = valid_lft;
4149 ifp->prefered_lft = prefered_lft;
4151 spin_unlock_bh(&ifp->lock);
4152 if (!(ifp->flags&IFA_F_TENTATIVE))
4153 ipv6_ifa_notify(0, ifp);
4155 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4156 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4157 expires, flags);
4158 } else if (had_prefixroute) {
4159 enum cleanup_prefix_rt_t action;
4160 unsigned long rt_expires;
4162 write_lock_bh(&ifp->idev->lock);
4163 action = check_cleanup_prefix_route(ifp, &rt_expires);
4164 write_unlock_bh(&ifp->idev->lock);
4166 if (action != CLEANUP_PREFIX_RT_NOP) {
4167 cleanup_prefix_route(ifp, rt_expires,
4168 action == CLEANUP_PREFIX_RT_DEL);
4172 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4173 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4174 valid_lft = prefered_lft = 0;
4175 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4176 !was_managetempaddr, jiffies);
4179 addrconf_verify_rtnl();
4181 return 0;
4184 static int
4185 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4187 struct net *net = sock_net(skb->sk);
4188 struct ifaddrmsg *ifm;
4189 struct nlattr *tb[IFA_MAX+1];
4190 struct in6_addr *pfx, *peer_pfx;
4191 struct inet6_ifaddr *ifa;
4192 struct net_device *dev;
4193 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4194 u32 ifa_flags;
4195 int err;
4197 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4198 if (err < 0)
4199 return err;
4201 ifm = nlmsg_data(nlh);
4202 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4203 if (!pfx)
4204 return -EINVAL;
4206 if (tb[IFA_CACHEINFO]) {
4207 struct ifa_cacheinfo *ci;
4209 ci = nla_data(tb[IFA_CACHEINFO]);
4210 valid_lft = ci->ifa_valid;
4211 preferred_lft = ci->ifa_prefered;
4212 } else {
4213 preferred_lft = INFINITY_LIFE_TIME;
4214 valid_lft = INFINITY_LIFE_TIME;
4217 dev = __dev_get_by_index(net, ifm->ifa_index);
4218 if (!dev)
4219 return -ENODEV;
4221 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4223 /* We ignore other flags so far. */
4224 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4225 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4227 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4228 if (!ifa) {
4230 * It would be best to check for !NLM_F_CREATE here but
4231 * userspace already relies on not having to provide this.
4233 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4234 ifm->ifa_prefixlen, ifa_flags,
4235 preferred_lft, valid_lft);
4238 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4239 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4240 err = -EEXIST;
4241 else
4242 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4244 in6_ifa_put(ifa);
4246 return err;
4249 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4250 u8 scope, int ifindex)
4252 struct ifaddrmsg *ifm;
4254 ifm = nlmsg_data(nlh);
4255 ifm->ifa_family = AF_INET6;
4256 ifm->ifa_prefixlen = prefixlen;
4257 ifm->ifa_flags = flags;
4258 ifm->ifa_scope = scope;
4259 ifm->ifa_index = ifindex;
4262 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4263 unsigned long tstamp, u32 preferred, u32 valid)
4265 struct ifa_cacheinfo ci;
4267 ci.cstamp = cstamp_delta(cstamp);
4268 ci.tstamp = cstamp_delta(tstamp);
4269 ci.ifa_prefered = preferred;
4270 ci.ifa_valid = valid;
4272 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4275 static inline int rt_scope(int ifa_scope)
4277 if (ifa_scope & IFA_HOST)
4278 return RT_SCOPE_HOST;
4279 else if (ifa_scope & IFA_LINK)
4280 return RT_SCOPE_LINK;
4281 else if (ifa_scope & IFA_SITE)
4282 return RT_SCOPE_SITE;
4283 else
4284 return RT_SCOPE_UNIVERSE;
4287 static inline int inet6_ifaddr_msgsize(void)
4289 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4290 + nla_total_size(16) /* IFA_LOCAL */
4291 + nla_total_size(16) /* IFA_ADDRESS */
4292 + nla_total_size(sizeof(struct ifa_cacheinfo))
4293 + nla_total_size(4) /* IFA_FLAGS */;
4296 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4297 u32 portid, u32 seq, int event, unsigned int flags)
4299 struct nlmsghdr *nlh;
4300 u32 preferred, valid;
4302 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4303 if (!nlh)
4304 return -EMSGSIZE;
4306 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4307 ifa->idev->dev->ifindex);
4309 if (!((ifa->flags&IFA_F_PERMANENT) &&
4310 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4311 preferred = ifa->prefered_lft;
4312 valid = ifa->valid_lft;
4313 if (preferred != INFINITY_LIFE_TIME) {
4314 long tval = (jiffies - ifa->tstamp)/HZ;
4315 if (preferred > tval)
4316 preferred -= tval;
4317 else
4318 preferred = 0;
4319 if (valid != INFINITY_LIFE_TIME) {
4320 if (valid > tval)
4321 valid -= tval;
4322 else
4323 valid = 0;
4326 } else {
4327 preferred = INFINITY_LIFE_TIME;
4328 valid = INFINITY_LIFE_TIME;
4331 if (!ipv6_addr_any(&ifa->peer_addr)) {
4332 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4333 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4334 goto error;
4335 } else
4336 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4337 goto error;
4339 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4340 goto error;
4342 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4343 goto error;
4345 nlmsg_end(skb, nlh);
4346 return 0;
4348 error:
4349 nlmsg_cancel(skb, nlh);
4350 return -EMSGSIZE;
4353 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4354 u32 portid, u32 seq, int event, u16 flags)
4356 struct nlmsghdr *nlh;
4357 u8 scope = RT_SCOPE_UNIVERSE;
4358 int ifindex = ifmca->idev->dev->ifindex;
4360 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4361 scope = RT_SCOPE_SITE;
4363 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4364 if (!nlh)
4365 return -EMSGSIZE;
4367 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4368 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4369 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4370 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4371 nlmsg_cancel(skb, nlh);
4372 return -EMSGSIZE;
4375 nlmsg_end(skb, nlh);
4376 return 0;
4379 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4380 u32 portid, u32 seq, int event, unsigned int flags)
4382 struct nlmsghdr *nlh;
4383 u8 scope = RT_SCOPE_UNIVERSE;
4384 int ifindex = ifaca->aca_idev->dev->ifindex;
4386 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4387 scope = RT_SCOPE_SITE;
4389 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4390 if (!nlh)
4391 return -EMSGSIZE;
4393 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4394 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4395 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4396 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4397 nlmsg_cancel(skb, nlh);
4398 return -EMSGSIZE;
4401 nlmsg_end(skb, nlh);
4402 return 0;
4405 enum addr_type_t {
4406 UNICAST_ADDR,
4407 MULTICAST_ADDR,
4408 ANYCAST_ADDR,
4411 /* called with rcu_read_lock() */
4412 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4413 struct netlink_callback *cb, enum addr_type_t type,
4414 int s_ip_idx, int *p_ip_idx)
4416 struct ifmcaddr6 *ifmca;
4417 struct ifacaddr6 *ifaca;
4418 int err = 1;
4419 int ip_idx = *p_ip_idx;
4421 read_lock_bh(&idev->lock);
4422 switch (type) {
4423 case UNICAST_ADDR: {
4424 struct inet6_ifaddr *ifa;
4426 /* unicast address incl. temp addr */
4427 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4428 if (++ip_idx < s_ip_idx)
4429 continue;
4430 err = inet6_fill_ifaddr(skb, ifa,
4431 NETLINK_CB(cb->skb).portid,
4432 cb->nlh->nlmsg_seq,
4433 RTM_NEWADDR,
4434 NLM_F_MULTI);
4435 if (err < 0)
4436 break;
4437 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4439 break;
4441 case MULTICAST_ADDR:
4442 /* multicast address */
4443 for (ifmca = idev->mc_list; ifmca;
4444 ifmca = ifmca->next, ip_idx++) {
4445 if (ip_idx < s_ip_idx)
4446 continue;
4447 err = inet6_fill_ifmcaddr(skb, ifmca,
4448 NETLINK_CB(cb->skb).portid,
4449 cb->nlh->nlmsg_seq,
4450 RTM_GETMULTICAST,
4451 NLM_F_MULTI);
4452 if (err < 0)
4453 break;
4455 break;
4456 case ANYCAST_ADDR:
4457 /* anycast address */
4458 for (ifaca = idev->ac_list; ifaca;
4459 ifaca = ifaca->aca_next, ip_idx++) {
4460 if (ip_idx < s_ip_idx)
4461 continue;
4462 err = inet6_fill_ifacaddr(skb, ifaca,
4463 NETLINK_CB(cb->skb).portid,
4464 cb->nlh->nlmsg_seq,
4465 RTM_GETANYCAST,
4466 NLM_F_MULTI);
4467 if (err < 0)
4468 break;
4470 break;
4471 default:
4472 break;
4474 read_unlock_bh(&idev->lock);
4475 *p_ip_idx = ip_idx;
4476 return err;
4479 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4480 enum addr_type_t type)
4482 struct net *net = sock_net(skb->sk);
4483 int h, s_h;
4484 int idx, ip_idx;
4485 int s_idx, s_ip_idx;
4486 struct net_device *dev;
4487 struct inet6_dev *idev;
4488 struct hlist_head *head;
4490 s_h = cb->args[0];
4491 s_idx = idx = cb->args[1];
4492 s_ip_idx = ip_idx = cb->args[2];
4494 rcu_read_lock();
4495 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4496 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4497 idx = 0;
4498 head = &net->dev_index_head[h];
4499 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4500 if (idx < s_idx)
4501 goto cont;
4502 if (h > s_h || idx > s_idx)
4503 s_ip_idx = 0;
4504 ip_idx = 0;
4505 idev = __in6_dev_get(dev);
4506 if (!idev)
4507 goto cont;
4509 if (in6_dump_addrs(idev, skb, cb, type,
4510 s_ip_idx, &ip_idx) < 0)
4511 goto done;
4512 cont:
4513 idx++;
4516 done:
4517 rcu_read_unlock();
4518 cb->args[0] = h;
4519 cb->args[1] = idx;
4520 cb->args[2] = ip_idx;
4522 return skb->len;
4525 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4527 enum addr_type_t type = UNICAST_ADDR;
4529 return inet6_dump_addr(skb, cb, type);
4532 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4534 enum addr_type_t type = MULTICAST_ADDR;
4536 return inet6_dump_addr(skb, cb, type);
4540 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4542 enum addr_type_t type = ANYCAST_ADDR;
4544 return inet6_dump_addr(skb, cb, type);
4547 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4549 struct net *net = sock_net(in_skb->sk);
4550 struct ifaddrmsg *ifm;
4551 struct nlattr *tb[IFA_MAX+1];
4552 struct in6_addr *addr = NULL, *peer;
4553 struct net_device *dev = NULL;
4554 struct inet6_ifaddr *ifa;
4555 struct sk_buff *skb;
4556 int err;
4558 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4559 if (err < 0)
4560 goto errout;
4562 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4563 if (!addr) {
4564 err = -EINVAL;
4565 goto errout;
4568 ifm = nlmsg_data(nlh);
4569 if (ifm->ifa_index)
4570 dev = __dev_get_by_index(net, ifm->ifa_index);
4572 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4573 if (!ifa) {
4574 err = -EADDRNOTAVAIL;
4575 goto errout;
4578 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4579 if (!skb) {
4580 err = -ENOBUFS;
4581 goto errout_ifa;
4584 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4585 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4586 if (err < 0) {
4587 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4588 WARN_ON(err == -EMSGSIZE);
4589 kfree_skb(skb);
4590 goto errout_ifa;
4592 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4593 errout_ifa:
4594 in6_ifa_put(ifa);
4595 errout:
4596 return err;
4599 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4601 struct sk_buff *skb;
4602 struct net *net = dev_net(ifa->idev->dev);
4603 int err = -ENOBUFS;
4605 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4606 if (!skb)
4607 goto errout;
4609 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4610 if (err < 0) {
4611 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4612 WARN_ON(err == -EMSGSIZE);
4613 kfree_skb(skb);
4614 goto errout;
4616 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4617 return;
4618 errout:
4619 if (err < 0)
4620 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4623 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4624 __s32 *array, int bytes)
4626 BUG_ON(bytes < (DEVCONF_MAX * 4));
4628 memset(array, 0, bytes);
4629 array[DEVCONF_FORWARDING] = cnf->forwarding;
4630 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4631 array[DEVCONF_MTU6] = cnf->mtu6;
4632 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4633 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4634 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4635 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4636 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4637 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4638 jiffies_to_msecs(cnf->rtr_solicit_interval);
4639 array[DEVCONF_RTR_SOLICIT_DELAY] =
4640 jiffies_to_msecs(cnf->rtr_solicit_delay);
4641 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4642 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4643 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4644 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4645 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4646 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4647 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4648 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4649 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4650 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4651 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4652 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4653 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4654 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4655 #ifdef CONFIG_IPV6_ROUTER_PREF
4656 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4657 array[DEVCONF_RTR_PROBE_INTERVAL] =
4658 jiffies_to_msecs(cnf->rtr_probe_interval);
4659 #ifdef CONFIG_IPV6_ROUTE_INFO
4660 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4661 #endif
4662 #endif
4663 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4664 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4665 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4666 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4667 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4668 #endif
4669 #ifdef CONFIG_IPV6_MROUTE
4670 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4671 #endif
4672 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4673 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4674 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4675 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4676 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4677 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4678 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4679 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4680 /* we omit DEVCONF_STABLE_SECRET for now */
4681 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4684 static inline size_t inet6_ifla6_size(void)
4686 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4687 + nla_total_size(sizeof(struct ifla_cacheinfo))
4688 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4689 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4690 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4691 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4694 static inline size_t inet6_if_nlmsg_size(void)
4696 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4697 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4698 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4699 + nla_total_size(4) /* IFLA_MTU */
4700 + nla_total_size(4) /* IFLA_LINK */
4701 + nla_total_size(1) /* IFLA_OPERSTATE */
4702 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4705 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4706 int items, int bytes)
4708 int i;
4709 int pad = bytes - sizeof(u64) * items;
4710 BUG_ON(pad < 0);
4712 /* Use put_unaligned() because stats may not be aligned for u64. */
4713 put_unaligned(items, &stats[0]);
4714 for (i = 1; i < items; i++)
4715 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4717 memset(&stats[items], 0, pad);
4720 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4721 int bytes, size_t syncpoff)
4723 int i, c;
4724 u64 buff[IPSTATS_MIB_MAX];
4725 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4727 BUG_ON(pad < 0);
4729 memset(buff, 0, sizeof(buff));
4730 buff[0] = IPSTATS_MIB_MAX;
4732 for_each_possible_cpu(c) {
4733 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4734 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4737 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4738 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4741 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4742 int bytes)
4744 switch (attrtype) {
4745 case IFLA_INET6_STATS:
4746 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4747 offsetof(struct ipstats_mib, syncp));
4748 break;
4749 case IFLA_INET6_ICMP6STATS:
4750 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4751 break;
4755 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4756 u32 ext_filter_mask)
4758 struct nlattr *nla;
4759 struct ifla_cacheinfo ci;
4761 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4762 goto nla_put_failure;
4763 ci.max_reasm_len = IPV6_MAXPLEN;
4764 ci.tstamp = cstamp_delta(idev->tstamp);
4765 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4766 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4767 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4768 goto nla_put_failure;
4769 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4770 if (!nla)
4771 goto nla_put_failure;
4772 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4774 /* XXX - MC not implemented */
4776 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4777 return 0;
4779 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4780 if (!nla)
4781 goto nla_put_failure;
4782 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4784 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4785 if (!nla)
4786 goto nla_put_failure;
4787 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4789 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4790 if (!nla)
4791 goto nla_put_failure;
4793 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4794 goto nla_put_failure;
4796 read_lock_bh(&idev->lock);
4797 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4798 read_unlock_bh(&idev->lock);
4800 return 0;
4802 nla_put_failure:
4803 return -EMSGSIZE;
4806 static size_t inet6_get_link_af_size(const struct net_device *dev,
4807 u32 ext_filter_mask)
4809 if (!__in6_dev_get(dev))
4810 return 0;
4812 return inet6_ifla6_size();
4815 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4816 u32 ext_filter_mask)
4818 struct inet6_dev *idev = __in6_dev_get(dev);
4820 if (!idev)
4821 return -ENODATA;
4823 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4824 return -EMSGSIZE;
4826 return 0;
4829 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4831 struct inet6_ifaddr *ifp;
4832 struct net_device *dev = idev->dev;
4833 bool update_rs = false;
4834 struct in6_addr ll_addr;
4836 ASSERT_RTNL();
4838 if (!token)
4839 return -EINVAL;
4840 if (ipv6_addr_any(token))
4841 return -EINVAL;
4842 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4843 return -EINVAL;
4844 if (!ipv6_accept_ra(idev))
4845 return -EINVAL;
4846 if (idev->cnf.rtr_solicits <= 0)
4847 return -EINVAL;
4849 write_lock_bh(&idev->lock);
4851 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4852 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4854 write_unlock_bh(&idev->lock);
4856 if (!idev->dead && (idev->if_flags & IF_READY) &&
4857 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4858 IFA_F_OPTIMISTIC)) {
4860 /* If we're not ready, then normal ifup will take care
4861 * of this. Otherwise, we need to request our rs here.
4863 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4864 update_rs = true;
4867 write_lock_bh(&idev->lock);
4869 if (update_rs) {
4870 idev->if_flags |= IF_RS_SENT;
4871 idev->rs_probes = 1;
4872 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4875 /* Well, that's kinda nasty ... */
4876 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4877 spin_lock(&ifp->lock);
4878 if (ifp->tokenized) {
4879 ifp->valid_lft = 0;
4880 ifp->prefered_lft = 0;
4882 spin_unlock(&ifp->lock);
4885 write_unlock_bh(&idev->lock);
4886 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4887 addrconf_verify_rtnl();
4888 return 0;
4891 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4892 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
4893 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
4896 static int inet6_validate_link_af(const struct net_device *dev,
4897 const struct nlattr *nla)
4899 struct nlattr *tb[IFLA_INET6_MAX + 1];
4901 if (dev && !__in6_dev_get(dev))
4902 return -EAFNOSUPPORT;
4904 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4907 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4909 int err = -EINVAL;
4910 struct inet6_dev *idev = __in6_dev_get(dev);
4911 struct nlattr *tb[IFLA_INET6_MAX + 1];
4913 if (!idev)
4914 return -EAFNOSUPPORT;
4916 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4917 BUG();
4919 if (tb[IFLA_INET6_TOKEN]) {
4920 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4921 if (err)
4922 return err;
4925 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4926 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4928 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4929 mode != IN6_ADDR_GEN_MODE_NONE &&
4930 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4931 return -EINVAL;
4933 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4934 !idev->cnf.stable_secret.initialized &&
4935 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4936 return -EINVAL;
4938 idev->addr_gen_mode = mode;
4939 err = 0;
4942 return err;
4945 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4946 u32 portid, u32 seq, int event, unsigned int flags)
4948 struct net_device *dev = idev->dev;
4949 struct ifinfomsg *hdr;
4950 struct nlmsghdr *nlh;
4951 void *protoinfo;
4953 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4954 if (!nlh)
4955 return -EMSGSIZE;
4957 hdr = nlmsg_data(nlh);
4958 hdr->ifi_family = AF_INET6;
4959 hdr->__ifi_pad = 0;
4960 hdr->ifi_type = dev->type;
4961 hdr->ifi_index = dev->ifindex;
4962 hdr->ifi_flags = dev_get_flags(dev);
4963 hdr->ifi_change = 0;
4965 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4966 (dev->addr_len &&
4967 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4968 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4969 (dev->ifindex != dev_get_iflink(dev) &&
4970 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
4971 nla_put_u8(skb, IFLA_OPERSTATE,
4972 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
4973 goto nla_put_failure;
4974 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4975 if (!protoinfo)
4976 goto nla_put_failure;
4978 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
4979 goto nla_put_failure;
4981 nla_nest_end(skb, protoinfo);
4982 nlmsg_end(skb, nlh);
4983 return 0;
4985 nla_put_failure:
4986 nlmsg_cancel(skb, nlh);
4987 return -EMSGSIZE;
4990 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4992 struct net *net = sock_net(skb->sk);
4993 int h, s_h;
4994 int idx = 0, s_idx;
4995 struct net_device *dev;
4996 struct inet6_dev *idev;
4997 struct hlist_head *head;
4999 s_h = cb->args[0];
5000 s_idx = cb->args[1];
5002 rcu_read_lock();
5003 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5004 idx = 0;
5005 head = &net->dev_index_head[h];
5006 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5007 if (idx < s_idx)
5008 goto cont;
5009 idev = __in6_dev_get(dev);
5010 if (!idev)
5011 goto cont;
5012 if (inet6_fill_ifinfo(skb, idev,
5013 NETLINK_CB(cb->skb).portid,
5014 cb->nlh->nlmsg_seq,
5015 RTM_NEWLINK, NLM_F_MULTI) < 0)
5016 goto out;
5017 cont:
5018 idx++;
5021 out:
5022 rcu_read_unlock();
5023 cb->args[1] = idx;
5024 cb->args[0] = h;
5026 return skb->len;
5029 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5031 struct sk_buff *skb;
5032 struct net *net = dev_net(idev->dev);
5033 int err = -ENOBUFS;
5035 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5036 if (!skb)
5037 goto errout;
5039 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5040 if (err < 0) {
5041 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5042 WARN_ON(err == -EMSGSIZE);
5043 kfree_skb(skb);
5044 goto errout;
5046 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5047 return;
5048 errout:
5049 if (err < 0)
5050 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5053 static inline size_t inet6_prefix_nlmsg_size(void)
5055 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5056 + nla_total_size(sizeof(struct in6_addr))
5057 + nla_total_size(sizeof(struct prefix_cacheinfo));
5060 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5061 struct prefix_info *pinfo, u32 portid, u32 seq,
5062 int event, unsigned int flags)
5064 struct prefixmsg *pmsg;
5065 struct nlmsghdr *nlh;
5066 struct prefix_cacheinfo ci;
5068 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5069 if (!nlh)
5070 return -EMSGSIZE;
5072 pmsg = nlmsg_data(nlh);
5073 pmsg->prefix_family = AF_INET6;
5074 pmsg->prefix_pad1 = 0;
5075 pmsg->prefix_pad2 = 0;
5076 pmsg->prefix_ifindex = idev->dev->ifindex;
5077 pmsg->prefix_len = pinfo->prefix_len;
5078 pmsg->prefix_type = pinfo->type;
5079 pmsg->prefix_pad3 = 0;
5080 pmsg->prefix_flags = 0;
5081 if (pinfo->onlink)
5082 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5083 if (pinfo->autoconf)
5084 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5086 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5087 goto nla_put_failure;
5088 ci.preferred_time = ntohl(pinfo->prefered);
5089 ci.valid_time = ntohl(pinfo->valid);
5090 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5091 goto nla_put_failure;
5092 nlmsg_end(skb, nlh);
5093 return 0;
5095 nla_put_failure:
5096 nlmsg_cancel(skb, nlh);
5097 return -EMSGSIZE;
5100 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5101 struct prefix_info *pinfo)
5103 struct sk_buff *skb;
5104 struct net *net = dev_net(idev->dev);
5105 int err = -ENOBUFS;
5107 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5108 if (!skb)
5109 goto errout;
5111 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5112 if (err < 0) {
5113 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5114 WARN_ON(err == -EMSGSIZE);
5115 kfree_skb(skb);
5116 goto errout;
5118 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5119 return;
5120 errout:
5121 if (err < 0)
5122 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5125 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5127 struct net *net = dev_net(ifp->idev->dev);
5129 if (event)
5130 ASSERT_RTNL();
5132 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5134 switch (event) {
5135 case RTM_NEWADDR:
5137 * If the address was optimistic
5138 * we inserted the route at the start of
5139 * our DAD process, so we don't need
5140 * to do it again
5142 if (!(ifp->rt->rt6i_node))
5143 ip6_ins_rt(ifp->rt);
5144 if (ifp->idev->cnf.forwarding)
5145 addrconf_join_anycast(ifp);
5146 if (!ipv6_addr_any(&ifp->peer_addr))
5147 addrconf_prefix_route(&ifp->peer_addr, 128,
5148 ifp->idev->dev, 0, 0);
5149 break;
5150 case RTM_DELADDR:
5151 if (ifp->idev->cnf.forwarding)
5152 addrconf_leave_anycast(ifp);
5153 addrconf_leave_solict(ifp->idev, &ifp->addr);
5154 if (!ipv6_addr_any(&ifp->peer_addr)) {
5155 struct rt6_info *rt;
5157 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5158 ifp->idev->dev, 0, 0);
5159 if (rt)
5160 ip6_del_rt(rt);
5162 dst_hold(&ifp->rt->dst);
5164 ip6_del_rt(ifp->rt);
5166 rt_genid_bump_ipv6(net);
5167 break;
5169 atomic_inc(&net->ipv6.dev_addr_genid);
5172 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5174 rcu_read_lock_bh();
5175 if (likely(ifp->idev->dead == 0))
5176 __ipv6_ifa_notify(event, ifp);
5177 rcu_read_unlock_bh();
5180 #ifdef CONFIG_SYSCTL
5182 static
5183 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5184 void __user *buffer, size_t *lenp, loff_t *ppos)
5186 int *valp = ctl->data;
5187 int val = *valp;
5188 loff_t pos = *ppos;
5189 struct ctl_table lctl;
5190 int ret;
5193 * ctl->data points to idev->cnf.forwarding, we should
5194 * not modify it until we get the rtnl lock.
5196 lctl = *ctl;
5197 lctl.data = &val;
5199 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5201 if (write)
5202 ret = addrconf_fixup_forwarding(ctl, valp, val);
5203 if (ret)
5204 *ppos = pos;
5205 return ret;
5208 static
5209 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5210 void __user *buffer, size_t *lenp, loff_t *ppos)
5212 struct inet6_dev *idev = ctl->extra1;
5213 int min_mtu = IPV6_MIN_MTU;
5214 struct ctl_table lctl;
5216 lctl = *ctl;
5217 lctl.extra1 = &min_mtu;
5218 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5220 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5223 static void dev_disable_change(struct inet6_dev *idev)
5225 struct netdev_notifier_info info;
5227 if (!idev || !idev->dev)
5228 return;
5230 netdev_notifier_info_init(&info, idev->dev);
5231 if (idev->cnf.disable_ipv6)
5232 addrconf_notify(NULL, NETDEV_DOWN, &info);
5233 else
5234 addrconf_notify(NULL, NETDEV_UP, &info);
5237 static void addrconf_disable_change(struct net *net, __s32 newf)
5239 struct net_device *dev;
5240 struct inet6_dev *idev;
5242 rcu_read_lock();
5243 for_each_netdev_rcu(net, dev) {
5244 idev = __in6_dev_get(dev);
5245 if (idev) {
5246 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5247 idev->cnf.disable_ipv6 = newf;
5248 if (changed)
5249 dev_disable_change(idev);
5252 rcu_read_unlock();
5255 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5257 struct net *net;
5258 int old;
5260 if (!rtnl_trylock())
5261 return restart_syscall();
5263 net = (struct net *)table->extra2;
5264 old = *p;
5265 *p = newf;
5267 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5268 rtnl_unlock();
5269 return 0;
5272 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5273 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5274 addrconf_disable_change(net, newf);
5275 } else if ((!newf) ^ (!old))
5276 dev_disable_change((struct inet6_dev *)table->extra1);
5278 rtnl_unlock();
5279 return 0;
5282 static
5283 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5284 void __user *buffer, size_t *lenp, loff_t *ppos)
5286 int *valp = ctl->data;
5287 int val = *valp;
5288 loff_t pos = *ppos;
5289 struct ctl_table lctl;
5290 int ret;
5293 * ctl->data points to idev->cnf.disable_ipv6, we should
5294 * not modify it until we get the rtnl lock.
5296 lctl = *ctl;
5297 lctl.data = &val;
5299 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5301 if (write)
5302 ret = addrconf_disable_ipv6(ctl, valp, val);
5303 if (ret)
5304 *ppos = pos;
5305 return ret;
5308 static
5309 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5310 void __user *buffer, size_t *lenp, loff_t *ppos)
5312 int *valp = ctl->data;
5313 int ret;
5314 int old, new;
5316 old = *valp;
5317 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5318 new = *valp;
5320 if (write && old != new) {
5321 struct net *net = ctl->extra2;
5323 if (!rtnl_trylock())
5324 return restart_syscall();
5326 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5327 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5328 NETCONFA_IFINDEX_DEFAULT,
5329 net->ipv6.devconf_dflt);
5330 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5331 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5332 NETCONFA_IFINDEX_ALL,
5333 net->ipv6.devconf_all);
5334 else {
5335 struct inet6_dev *idev = ctl->extra1;
5337 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5338 idev->dev->ifindex,
5339 &idev->cnf);
5341 rtnl_unlock();
5344 return ret;
5347 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5348 void __user *buffer, size_t *lenp,
5349 loff_t *ppos)
5351 int err;
5352 struct in6_addr addr;
5353 char str[IPV6_MAX_STRLEN];
5354 struct ctl_table lctl = *ctl;
5355 struct net *net = ctl->extra2;
5356 struct ipv6_stable_secret *secret = ctl->data;
5358 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5359 return -EIO;
5361 lctl.maxlen = IPV6_MAX_STRLEN;
5362 lctl.data = str;
5364 if (!rtnl_trylock())
5365 return restart_syscall();
5367 if (!write && !secret->initialized) {
5368 err = -EIO;
5369 goto out;
5372 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5373 if (err >= sizeof(str)) {
5374 err = -EIO;
5375 goto out;
5378 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5379 if (err || !write)
5380 goto out;
5382 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5383 err = -EIO;
5384 goto out;
5387 secret->initialized = true;
5388 secret->secret = addr;
5390 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5391 struct net_device *dev;
5393 for_each_netdev(net, dev) {
5394 struct inet6_dev *idev = __in6_dev_get(dev);
5396 if (idev) {
5397 idev->addr_gen_mode =
5398 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5401 } else {
5402 struct inet6_dev *idev = ctl->extra1;
5404 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5407 out:
5408 rtnl_unlock();
5410 return err;
5413 static
5414 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5415 int write,
5416 void __user *buffer,
5417 size_t *lenp,
5418 loff_t *ppos)
5420 int *valp = ctl->data;
5421 int val = *valp;
5422 loff_t pos = *ppos;
5423 struct ctl_table lctl;
5424 int ret;
5426 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5427 * we should not modify it until we get the rtnl lock.
5429 lctl = *ctl;
5430 lctl.data = &val;
5432 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5434 if (write)
5435 ret = addrconf_fixup_linkdown(ctl, valp, val);
5436 if (ret)
5437 *ppos = pos;
5438 return ret;
5441 static struct addrconf_sysctl_table
5443 struct ctl_table_header *sysctl_header;
5444 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5445 } addrconf_sysctl __read_mostly = {
5446 .sysctl_header = NULL,
5447 .addrconf_vars = {
5449 .procname = "forwarding",
5450 .data = &ipv6_devconf.forwarding,
5451 .maxlen = sizeof(int),
5452 .mode = 0644,
5453 .proc_handler = addrconf_sysctl_forward,
5456 .procname = "hop_limit",
5457 .data = &ipv6_devconf.hop_limit,
5458 .maxlen = sizeof(int),
5459 .mode = 0644,
5460 .proc_handler = proc_dointvec,
5463 .procname = "mtu",
5464 .data = &ipv6_devconf.mtu6,
5465 .maxlen = sizeof(int),
5466 .mode = 0644,
5467 .proc_handler = addrconf_sysctl_mtu,
5470 .procname = "accept_ra",
5471 .data = &ipv6_devconf.accept_ra,
5472 .maxlen = sizeof(int),
5473 .mode = 0644,
5474 .proc_handler = proc_dointvec,
5477 .procname = "accept_redirects",
5478 .data = &ipv6_devconf.accept_redirects,
5479 .maxlen = sizeof(int),
5480 .mode = 0644,
5481 .proc_handler = proc_dointvec,
5484 .procname = "autoconf",
5485 .data = &ipv6_devconf.autoconf,
5486 .maxlen = sizeof(int),
5487 .mode = 0644,
5488 .proc_handler = proc_dointvec,
5491 .procname = "dad_transmits",
5492 .data = &ipv6_devconf.dad_transmits,
5493 .maxlen = sizeof(int),
5494 .mode = 0644,
5495 .proc_handler = proc_dointvec,
5498 .procname = "router_solicitations",
5499 .data = &ipv6_devconf.rtr_solicits,
5500 .maxlen = sizeof(int),
5501 .mode = 0644,
5502 .proc_handler = proc_dointvec,
5505 .procname = "router_solicitation_interval",
5506 .data = &ipv6_devconf.rtr_solicit_interval,
5507 .maxlen = sizeof(int),
5508 .mode = 0644,
5509 .proc_handler = proc_dointvec_jiffies,
5512 .procname = "router_solicitation_delay",
5513 .data = &ipv6_devconf.rtr_solicit_delay,
5514 .maxlen = sizeof(int),
5515 .mode = 0644,
5516 .proc_handler = proc_dointvec_jiffies,
5519 .procname = "force_mld_version",
5520 .data = &ipv6_devconf.force_mld_version,
5521 .maxlen = sizeof(int),
5522 .mode = 0644,
5523 .proc_handler = proc_dointvec,
5526 .procname = "mldv1_unsolicited_report_interval",
5527 .data =
5528 &ipv6_devconf.mldv1_unsolicited_report_interval,
5529 .maxlen = sizeof(int),
5530 .mode = 0644,
5531 .proc_handler = proc_dointvec_ms_jiffies,
5534 .procname = "mldv2_unsolicited_report_interval",
5535 .data =
5536 &ipv6_devconf.mldv2_unsolicited_report_interval,
5537 .maxlen = sizeof(int),
5538 .mode = 0644,
5539 .proc_handler = proc_dointvec_ms_jiffies,
5542 .procname = "use_tempaddr",
5543 .data = &ipv6_devconf.use_tempaddr,
5544 .maxlen = sizeof(int),
5545 .mode = 0644,
5546 .proc_handler = proc_dointvec,
5549 .procname = "temp_valid_lft",
5550 .data = &ipv6_devconf.temp_valid_lft,
5551 .maxlen = sizeof(int),
5552 .mode = 0644,
5553 .proc_handler = proc_dointvec,
5556 .procname = "temp_prefered_lft",
5557 .data = &ipv6_devconf.temp_prefered_lft,
5558 .maxlen = sizeof(int),
5559 .mode = 0644,
5560 .proc_handler = proc_dointvec,
5563 .procname = "regen_max_retry",
5564 .data = &ipv6_devconf.regen_max_retry,
5565 .maxlen = sizeof(int),
5566 .mode = 0644,
5567 .proc_handler = proc_dointvec,
5570 .procname = "max_desync_factor",
5571 .data = &ipv6_devconf.max_desync_factor,
5572 .maxlen = sizeof(int),
5573 .mode = 0644,
5574 .proc_handler = proc_dointvec,
5577 .procname = "max_addresses",
5578 .data = &ipv6_devconf.max_addresses,
5579 .maxlen = sizeof(int),
5580 .mode = 0644,
5581 .proc_handler = proc_dointvec,
5584 .procname = "accept_ra_defrtr",
5585 .data = &ipv6_devconf.accept_ra_defrtr,
5586 .maxlen = sizeof(int),
5587 .mode = 0644,
5588 .proc_handler = proc_dointvec,
5591 .procname = "accept_ra_min_hop_limit",
5592 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5593 .maxlen = sizeof(int),
5594 .mode = 0644,
5595 .proc_handler = proc_dointvec,
5598 .procname = "accept_ra_pinfo",
5599 .data = &ipv6_devconf.accept_ra_pinfo,
5600 .maxlen = sizeof(int),
5601 .mode = 0644,
5602 .proc_handler = proc_dointvec,
5604 #ifdef CONFIG_IPV6_ROUTER_PREF
5606 .procname = "accept_ra_rtr_pref",
5607 .data = &ipv6_devconf.accept_ra_rtr_pref,
5608 .maxlen = sizeof(int),
5609 .mode = 0644,
5610 .proc_handler = proc_dointvec,
5613 .procname = "router_probe_interval",
5614 .data = &ipv6_devconf.rtr_probe_interval,
5615 .maxlen = sizeof(int),
5616 .mode = 0644,
5617 .proc_handler = proc_dointvec_jiffies,
5619 #ifdef CONFIG_IPV6_ROUTE_INFO
5621 .procname = "accept_ra_rt_info_max_plen",
5622 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5623 .maxlen = sizeof(int),
5624 .mode = 0644,
5625 .proc_handler = proc_dointvec,
5627 #endif
5628 #endif
5630 .procname = "proxy_ndp",
5631 .data = &ipv6_devconf.proxy_ndp,
5632 .maxlen = sizeof(int),
5633 .mode = 0644,
5634 .proc_handler = addrconf_sysctl_proxy_ndp,
5637 .procname = "accept_source_route",
5638 .data = &ipv6_devconf.accept_source_route,
5639 .maxlen = sizeof(int),
5640 .mode = 0644,
5641 .proc_handler = proc_dointvec,
5643 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5645 .procname = "optimistic_dad",
5646 .data = &ipv6_devconf.optimistic_dad,
5647 .maxlen = sizeof(int),
5648 .mode = 0644,
5649 .proc_handler = proc_dointvec,
5653 .procname = "use_optimistic",
5654 .data = &ipv6_devconf.use_optimistic,
5655 .maxlen = sizeof(int),
5656 .mode = 0644,
5657 .proc_handler = proc_dointvec,
5660 #endif
5661 #ifdef CONFIG_IPV6_MROUTE
5663 .procname = "mc_forwarding",
5664 .data = &ipv6_devconf.mc_forwarding,
5665 .maxlen = sizeof(int),
5666 .mode = 0444,
5667 .proc_handler = proc_dointvec,
5669 #endif
5671 .procname = "disable_ipv6",
5672 .data = &ipv6_devconf.disable_ipv6,
5673 .maxlen = sizeof(int),
5674 .mode = 0644,
5675 .proc_handler = addrconf_sysctl_disable,
5678 .procname = "accept_dad",
5679 .data = &ipv6_devconf.accept_dad,
5680 .maxlen = sizeof(int),
5681 .mode = 0644,
5682 .proc_handler = proc_dointvec,
5685 .procname = "force_tllao",
5686 .data = &ipv6_devconf.force_tllao,
5687 .maxlen = sizeof(int),
5688 .mode = 0644,
5689 .proc_handler = proc_dointvec
5692 .procname = "ndisc_notify",
5693 .data = &ipv6_devconf.ndisc_notify,
5694 .maxlen = sizeof(int),
5695 .mode = 0644,
5696 .proc_handler = proc_dointvec
5699 .procname = "suppress_frag_ndisc",
5700 .data = &ipv6_devconf.suppress_frag_ndisc,
5701 .maxlen = sizeof(int),
5702 .mode = 0644,
5703 .proc_handler = proc_dointvec
5706 .procname = "accept_ra_from_local",
5707 .data = &ipv6_devconf.accept_ra_from_local,
5708 .maxlen = sizeof(int),
5709 .mode = 0644,
5710 .proc_handler = proc_dointvec,
5713 .procname = "accept_ra_mtu",
5714 .data = &ipv6_devconf.accept_ra_mtu,
5715 .maxlen = sizeof(int),
5716 .mode = 0644,
5717 .proc_handler = proc_dointvec,
5720 .procname = "stable_secret",
5721 .data = &ipv6_devconf.stable_secret,
5722 .maxlen = IPV6_MAX_STRLEN,
5723 .mode = 0600,
5724 .proc_handler = addrconf_sysctl_stable_secret,
5727 .procname = "use_oif_addrs_only",
5728 .data = &ipv6_devconf.use_oif_addrs_only,
5729 .maxlen = sizeof(int),
5730 .mode = 0644,
5731 .proc_handler = proc_dointvec,
5734 .procname = "ignore_routes_with_linkdown",
5735 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5736 .maxlen = sizeof(int),
5737 .mode = 0644,
5738 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5741 /* sentinel */
5746 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5747 struct inet6_dev *idev, struct ipv6_devconf *p)
5749 int i;
5750 struct addrconf_sysctl_table *t;
5751 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5753 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5754 if (!t)
5755 goto out;
5757 for (i = 0; t->addrconf_vars[i].data; i++) {
5758 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5759 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5760 t->addrconf_vars[i].extra2 = net;
5763 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5765 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5766 if (!t->sysctl_header)
5767 goto free;
5769 p->sysctl = t;
5770 return 0;
5772 free:
5773 kfree(t);
5774 out:
5775 return -ENOBUFS;
5778 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5780 struct addrconf_sysctl_table *t;
5782 if (!p->sysctl)
5783 return;
5785 t = p->sysctl;
5786 p->sysctl = NULL;
5787 unregister_net_sysctl_table(t->sysctl_header);
5788 kfree(t);
5791 static int addrconf_sysctl_register(struct inet6_dev *idev)
5793 int err;
5795 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5796 return -EINVAL;
5798 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5799 &ndisc_ifinfo_sysctl_change);
5800 if (err)
5801 return err;
5802 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5803 idev, &idev->cnf);
5804 if (err)
5805 neigh_sysctl_unregister(idev->nd_parms);
5807 return err;
5810 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5812 __addrconf_sysctl_unregister(&idev->cnf);
5813 neigh_sysctl_unregister(idev->nd_parms);
5817 #endif
5819 static int __net_init addrconf_init_net(struct net *net)
5821 int err = -ENOMEM;
5822 struct ipv6_devconf *all, *dflt;
5824 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5825 if (!all)
5826 goto err_alloc_all;
5828 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5829 if (!dflt)
5830 goto err_alloc_dflt;
5832 /* these will be inherited by all namespaces */
5833 dflt->autoconf = ipv6_defaults.autoconf;
5834 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5836 dflt->stable_secret.initialized = false;
5837 all->stable_secret.initialized = false;
5839 net->ipv6.devconf_all = all;
5840 net->ipv6.devconf_dflt = dflt;
5842 #ifdef CONFIG_SYSCTL
5843 err = __addrconf_sysctl_register(net, "all", NULL, all);
5844 if (err < 0)
5845 goto err_reg_all;
5847 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5848 if (err < 0)
5849 goto err_reg_dflt;
5850 #endif
5851 return 0;
5853 #ifdef CONFIG_SYSCTL
5854 err_reg_dflt:
5855 __addrconf_sysctl_unregister(all);
5856 err_reg_all:
5857 kfree(dflt);
5858 #endif
5859 err_alloc_dflt:
5860 kfree(all);
5861 err_alloc_all:
5862 return err;
5865 static void __net_exit addrconf_exit_net(struct net *net)
5867 #ifdef CONFIG_SYSCTL
5868 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5869 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5870 #endif
5871 kfree(net->ipv6.devconf_dflt);
5872 kfree(net->ipv6.devconf_all);
5875 static struct pernet_operations addrconf_ops = {
5876 .init = addrconf_init_net,
5877 .exit = addrconf_exit_net,
5880 static struct rtnl_af_ops inet6_ops __read_mostly = {
5881 .family = AF_INET6,
5882 .fill_link_af = inet6_fill_link_af,
5883 .get_link_af_size = inet6_get_link_af_size,
5884 .validate_link_af = inet6_validate_link_af,
5885 .set_link_af = inet6_set_link_af,
5889 * Init / cleanup code
5892 int __init addrconf_init(void)
5894 struct inet6_dev *idev;
5895 int i, err;
5897 err = ipv6_addr_label_init();
5898 if (err < 0) {
5899 pr_crit("%s: cannot initialize default policy table: %d\n",
5900 __func__, err);
5901 goto out;
5904 err = register_pernet_subsys(&addrconf_ops);
5905 if (err < 0)
5906 goto out_addrlabel;
5908 addrconf_wq = create_workqueue("ipv6_addrconf");
5909 if (!addrconf_wq) {
5910 err = -ENOMEM;
5911 goto out_nowq;
5914 /* The addrconf netdev notifier requires that loopback_dev
5915 * has it's ipv6 private information allocated and setup
5916 * before it can bring up and give link-local addresses
5917 * to other devices which are up.
5919 * Unfortunately, loopback_dev is not necessarily the first
5920 * entry in the global dev_base list of net devices. In fact,
5921 * it is likely to be the very last entry on that list.
5922 * So this causes the notifier registry below to try and
5923 * give link-local addresses to all devices besides loopback_dev
5924 * first, then loopback_dev, which cases all the non-loopback_dev
5925 * devices to fail to get a link-local address.
5927 * So, as a temporary fix, allocate the ipv6 structure for
5928 * loopback_dev first by hand.
5929 * Longer term, all of the dependencies ipv6 has upon the loopback
5930 * device and it being up should be removed.
5932 rtnl_lock();
5933 idev = ipv6_add_dev(init_net.loopback_dev);
5934 rtnl_unlock();
5935 if (IS_ERR(idev)) {
5936 err = PTR_ERR(idev);
5937 goto errlo;
5940 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5941 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5943 register_netdevice_notifier(&ipv6_dev_notf);
5945 addrconf_verify();
5947 rtnl_af_register(&inet6_ops);
5949 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5950 NULL);
5951 if (err < 0)
5952 goto errout;
5954 /* Only the first call to __rtnl_register can fail */
5955 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5956 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5957 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5958 inet6_dump_ifaddr, NULL);
5959 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5960 inet6_dump_ifmcaddr, NULL);
5961 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5962 inet6_dump_ifacaddr, NULL);
5963 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5964 inet6_netconf_dump_devconf, NULL);
5966 ipv6_addr_label_rtnl_register();
5968 return 0;
5969 errout:
5970 rtnl_af_unregister(&inet6_ops);
5971 unregister_netdevice_notifier(&ipv6_dev_notf);
5972 errlo:
5973 destroy_workqueue(addrconf_wq);
5974 out_nowq:
5975 unregister_pernet_subsys(&addrconf_ops);
5976 out_addrlabel:
5977 ipv6_addr_label_cleanup();
5978 out:
5979 return err;
5982 void addrconf_cleanup(void)
5984 struct net_device *dev;
5985 int i;
5987 unregister_netdevice_notifier(&ipv6_dev_notf);
5988 unregister_pernet_subsys(&addrconf_ops);
5989 ipv6_addr_label_cleanup();
5991 rtnl_lock();
5993 __rtnl_af_unregister(&inet6_ops);
5995 /* clean dev list */
5996 for_each_netdev(&init_net, dev) {
5997 if (__in6_dev_get(dev) == NULL)
5998 continue;
5999 addrconf_ifdown(dev, 1);
6001 addrconf_ifdown(init_net.loopback_dev, 2);
6004 * Check hash table.
6006 spin_lock_bh(&addrconf_hash_lock);
6007 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6008 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6009 spin_unlock_bh(&addrconf_hash_lock);
6010 cancel_delayed_work(&addr_chk_work);
6011 rtnl_unlock();
6013 destroy_workqueue(addrconf_wq);