driver core: bus: use list_for_each_entry*
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob38eeddedfc21baa843b315d720cc94a28b6af605
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
69 #include <net/net_namespace.h>
70 #include <net/sock.h>
71 #include <net/snmp.h>
73 #include <net/6lowpan.h>
74 #include <net/firewire.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/tcp.h>
81 #include <net/ip.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
99 #if ACONF_DEBUG >= 3
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101 #else
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #endif
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 #define IPV6_MAX_STRLEN \
108 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
110 static inline u32 cstamp_delta(unsigned long cstamp)
112 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
115 #ifdef CONFIG_SYSCTL
116 static int addrconf_sysctl_register(struct inet6_dev *idev);
117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118 #else
119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
121 return 0;
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
127 #endif
129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131 static void ipv6_regen_rndid(unsigned long data);
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135 static int ipv6_generate_stable_address(struct in6_addr *addr,
136 u8 dad_count,
137 const struct inet6_dev *idev);
140 * Configured unicast address hash table
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
145 static void addrconf_verify(void);
146 static void addrconf_verify_rtnl(void);
147 static void addrconf_verify_work(struct work_struct *);
149 static struct workqueue_struct *addrconf_wq;
150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
155 static void addrconf_type_change(struct net_device *dev,
156 unsigned long event);
157 static int addrconf_ifdown(struct net_device *dev, int how);
159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
160 int plen,
161 const struct net_device *dev,
162 u32 flags, u32 noflags);
164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165 static void addrconf_dad_work(struct work_struct *w);
166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167 static void addrconf_dad_run(struct inet6_dev *idev);
168 static void addrconf_rs_timer(unsigned long data);
169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173 struct prefix_info *pinfo);
174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175 struct net_device *dev);
177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
178 .forwarding = 0,
179 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
180 .mtu6 = IPV6_MIN_MTU,
181 .accept_ra = 1,
182 .accept_redirects = 1,
183 .autoconf = 1,
184 .force_mld_version = 0,
185 .mldv1_unsolicited_report_interval = 10 * HZ,
186 .mldv2_unsolicited_report_interval = HZ,
187 .dad_transmits = 1,
188 .rtr_solicits = MAX_RTR_SOLICITATIONS,
189 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
190 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
191 .use_tempaddr = 0,
192 .temp_valid_lft = TEMP_VALID_LIFETIME,
193 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
194 .regen_max_retry = REGEN_MAX_RETRY,
195 .max_desync_factor = MAX_DESYNC_FACTOR,
196 .max_addresses = IPV6_MAX_ADDRESSES,
197 .accept_ra_defrtr = 1,
198 .accept_ra_from_local = 0,
199 .accept_ra_min_hop_limit= 1,
200 .accept_ra_pinfo = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202 .accept_ra_rtr_pref = 1,
203 .rtr_probe_interval = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205 .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208 .proxy_ndp = 0,
209 .accept_source_route = 0, /* we do not accept RH0 by default. */
210 .disable_ipv6 = 0,
211 .accept_dad = 1,
212 .suppress_frag_ndisc = 1,
213 .accept_ra_mtu = 1,
214 .stable_secret = {
215 .initialized = false,
217 .use_oif_addrs_only = 0,
218 .ignore_routes_with_linkdown = 0,
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 (dad_failed)
1776 ifp->flags |= IFA_F_DADFAILED;
1778 if (ifp->flags&IFA_F_PERMANENT) {
1779 spin_lock_bh(&ifp->lock);
1780 addrconf_del_dad_work(ifp);
1781 ifp->flags |= IFA_F_TENTATIVE;
1782 spin_unlock_bh(&ifp->lock);
1783 if (dad_failed)
1784 ipv6_ifa_notify(0, ifp);
1785 in6_ifa_put(ifp);
1786 } else if (ifp->flags&IFA_F_TEMPORARY) {
1787 struct inet6_ifaddr *ifpub;
1788 spin_lock_bh(&ifp->lock);
1789 ifpub = ifp->ifpub;
1790 if (ifpub) {
1791 in6_ifa_hold(ifpub);
1792 spin_unlock_bh(&ifp->lock);
1793 ipv6_create_tempaddr(ifpub, ifp);
1794 in6_ifa_put(ifpub);
1795 } else {
1796 spin_unlock_bh(&ifp->lock);
1798 ipv6_del_addr(ifp);
1799 } else {
1800 ipv6_del_addr(ifp);
1804 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1806 int err = -ENOENT;
1808 spin_lock_bh(&ifp->lock);
1809 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1810 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1811 err = 0;
1813 spin_unlock_bh(&ifp->lock);
1815 return err;
1818 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1820 struct in6_addr addr;
1821 struct inet6_dev *idev = ifp->idev;
1822 struct net *net = dev_net(ifp->idev->dev);
1824 if (addrconf_dad_end(ifp)) {
1825 in6_ifa_put(ifp);
1826 return;
1829 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1830 ifp->idev->dev->name, &ifp->addr);
1832 spin_lock_bh(&ifp->lock);
1834 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1835 int scope = ifp->scope;
1836 u32 flags = ifp->flags;
1837 struct in6_addr new_addr;
1838 struct inet6_ifaddr *ifp2;
1839 u32 valid_lft, preferred_lft;
1840 int pfxlen = ifp->prefix_len;
1841 int retries = ifp->stable_privacy_retry + 1;
1843 if (retries > net->ipv6.sysctl.idgen_retries) {
1844 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1845 ifp->idev->dev->name);
1846 goto errdad;
1849 new_addr = ifp->addr;
1850 if (ipv6_generate_stable_address(&new_addr, retries,
1851 idev))
1852 goto errdad;
1854 valid_lft = ifp->valid_lft;
1855 preferred_lft = ifp->prefered_lft;
1857 spin_unlock_bh(&ifp->lock);
1859 if (idev->cnf.max_addresses &&
1860 ipv6_count_addresses(idev) >=
1861 idev->cnf.max_addresses)
1862 goto lock_errdad;
1864 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1865 ifp->idev->dev->name);
1867 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1868 scope, flags, valid_lft,
1869 preferred_lft);
1870 if (IS_ERR(ifp2))
1871 goto lock_errdad;
1873 spin_lock_bh(&ifp2->lock);
1874 ifp2->stable_privacy_retry = retries;
1875 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1876 spin_unlock_bh(&ifp2->lock);
1878 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1879 in6_ifa_put(ifp2);
1880 lock_errdad:
1881 spin_lock_bh(&ifp->lock);
1882 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1883 addr.s6_addr32[0] = htonl(0xfe800000);
1884 addr.s6_addr32[1] = 0;
1886 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1887 ipv6_addr_equal(&ifp->addr, &addr)) {
1888 /* DAD failed for link-local based on MAC address */
1889 idev->cnf.disable_ipv6 = 1;
1891 pr_info("%s: IPv6 being disabled!\n",
1892 ifp->idev->dev->name);
1896 errdad:
1897 /* transition from _POSTDAD to _ERRDAD */
1898 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1899 spin_unlock_bh(&ifp->lock);
1901 addrconf_mod_dad_work(ifp, 0);
1904 /* Join to solicited addr multicast group.
1905 * caller must hold RTNL */
1906 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1908 struct in6_addr maddr;
1910 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1911 return;
1913 addrconf_addr_solict_mult(addr, &maddr);
1914 ipv6_dev_mc_inc(dev, &maddr);
1917 /* caller must hold RTNL */
1918 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1920 struct in6_addr maddr;
1922 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1923 return;
1925 addrconf_addr_solict_mult(addr, &maddr);
1926 __ipv6_dev_mc_dec(idev, &maddr);
1929 /* caller must hold RTNL */
1930 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1932 struct in6_addr addr;
1934 if (ifp->prefix_len >= 127) /* RFC 6164 */
1935 return;
1936 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1937 if (ipv6_addr_any(&addr))
1938 return;
1939 __ipv6_dev_ac_inc(ifp->idev, &addr);
1942 /* caller must hold RTNL */
1943 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1945 struct in6_addr addr;
1947 if (ifp->prefix_len >= 127) /* RFC 6164 */
1948 return;
1949 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1950 if (ipv6_addr_any(&addr))
1951 return;
1952 __ipv6_dev_ac_dec(ifp->idev, &addr);
1955 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1957 if (dev->addr_len != EUI64_ADDR_LEN)
1958 return -1;
1959 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
1960 eui[0] ^= 2;
1961 return 0;
1964 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1966 union fwnet_hwaddr *ha;
1968 if (dev->addr_len != FWNET_ALEN)
1969 return -1;
1971 ha = (union fwnet_hwaddr *)dev->dev_addr;
1973 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1974 eui[0] ^= 2;
1975 return 0;
1978 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1980 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1981 if (dev->addr_len != ARCNET_ALEN)
1982 return -1;
1983 memset(eui, 0, 7);
1984 eui[7] = *(u8 *)dev->dev_addr;
1985 return 0;
1988 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1990 if (dev->addr_len != INFINIBAND_ALEN)
1991 return -1;
1992 memcpy(eui, dev->dev_addr + 12, 8);
1993 eui[0] |= 2;
1994 return 0;
1997 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1999 if (addr == 0)
2000 return -1;
2001 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2002 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2003 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2004 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2005 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2006 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2007 eui[1] = 0;
2008 eui[2] = 0x5E;
2009 eui[3] = 0xFE;
2010 memcpy(eui + 4, &addr, 4);
2011 return 0;
2014 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2016 if (dev->priv_flags & IFF_ISATAP)
2017 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2018 return -1;
2021 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2023 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2026 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2028 memcpy(eui, dev->perm_addr, 3);
2029 memcpy(eui + 5, dev->perm_addr + 3, 3);
2030 eui[3] = 0xFF;
2031 eui[4] = 0xFE;
2032 eui[0] ^= 2;
2033 return 0;
2036 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2038 switch (dev->type) {
2039 case ARPHRD_ETHER:
2040 case ARPHRD_FDDI:
2041 return addrconf_ifid_eui48(eui, dev);
2042 case ARPHRD_ARCNET:
2043 return addrconf_ifid_arcnet(eui, dev);
2044 case ARPHRD_INFINIBAND:
2045 return addrconf_ifid_infiniband(eui, dev);
2046 case ARPHRD_SIT:
2047 return addrconf_ifid_sit(eui, dev);
2048 case ARPHRD_IPGRE:
2049 return addrconf_ifid_gre(eui, dev);
2050 case ARPHRD_6LOWPAN:
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 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2325 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2326 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2329 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2331 struct prefix_info *pinfo;
2332 __u32 valid_lft;
2333 __u32 prefered_lft;
2334 int addr_type;
2335 u32 addr_flags = 0;
2336 struct inet6_dev *in6_dev;
2337 struct net *net = dev_net(dev);
2339 pinfo = (struct prefix_info *) opt;
2341 if (len < sizeof(struct prefix_info)) {
2342 ADBG("addrconf: prefix option too short\n");
2343 return;
2347 * Validation checks ([ADDRCONF], page 19)
2350 addr_type = ipv6_addr_type(&pinfo->prefix);
2352 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2353 return;
2355 valid_lft = ntohl(pinfo->valid);
2356 prefered_lft = ntohl(pinfo->prefered);
2358 if (prefered_lft > valid_lft) {
2359 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2360 return;
2363 in6_dev = in6_dev_get(dev);
2365 if (!in6_dev) {
2366 net_dbg_ratelimited("addrconf: device %s not configured\n",
2367 dev->name);
2368 return;
2372 * Two things going on here:
2373 * 1) Add routes for on-link prefixes
2374 * 2) Configure prefixes with the auto flag set
2377 if (pinfo->onlink) {
2378 struct rt6_info *rt;
2379 unsigned long rt_expires;
2381 /* Avoid arithmetic overflow. Really, we could
2382 * save rt_expires in seconds, likely valid_lft,
2383 * but it would require division in fib gc, that it
2384 * not good.
2386 if (HZ > USER_HZ)
2387 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2388 else
2389 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2391 if (addrconf_finite_timeout(rt_expires))
2392 rt_expires *= HZ;
2394 rt = addrconf_get_prefix_route(&pinfo->prefix,
2395 pinfo->prefix_len,
2396 dev,
2397 RTF_ADDRCONF | RTF_PREFIX_RT,
2398 RTF_GATEWAY | RTF_DEFAULT);
2400 if (rt) {
2401 /* Autoconf prefix route */
2402 if (valid_lft == 0) {
2403 ip6_del_rt(rt);
2404 rt = NULL;
2405 } else if (addrconf_finite_timeout(rt_expires)) {
2406 /* not infinity */
2407 rt6_set_expires(rt, jiffies + rt_expires);
2408 } else {
2409 rt6_clean_expires(rt);
2411 } else if (valid_lft) {
2412 clock_t expires = 0;
2413 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2414 if (addrconf_finite_timeout(rt_expires)) {
2415 /* not infinity */
2416 flags |= RTF_EXPIRES;
2417 expires = jiffies_to_clock_t(rt_expires);
2419 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2420 dev, expires, flags);
2422 ip6_rt_put(rt);
2425 /* Try to figure out our local address for this prefix */
2427 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2428 struct inet6_ifaddr *ifp;
2429 struct in6_addr addr;
2430 int create = 0, update_lft = 0;
2431 bool tokenized = false;
2433 if (pinfo->prefix_len == 64) {
2434 memcpy(&addr, &pinfo->prefix, 8);
2436 if (!ipv6_addr_any(&in6_dev->token)) {
2437 read_lock_bh(&in6_dev->lock);
2438 memcpy(addr.s6_addr + 8,
2439 in6_dev->token.s6_addr + 8, 8);
2440 read_unlock_bh(&in6_dev->lock);
2441 tokenized = true;
2442 } else if (is_addr_mode_generate_stable(in6_dev) &&
2443 !ipv6_generate_stable_address(&addr, 0,
2444 in6_dev)) {
2445 addr_flags |= IFA_F_STABLE_PRIVACY;
2446 goto ok;
2447 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2448 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2449 in6_dev_put(in6_dev);
2450 return;
2452 goto ok;
2454 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2455 pinfo->prefix_len);
2456 in6_dev_put(in6_dev);
2457 return;
2461 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2463 if (!ifp && valid_lft) {
2464 int max_addresses = in6_dev->cnf.max_addresses;
2466 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2467 if (in6_dev->cnf.optimistic_dad &&
2468 !net->ipv6.devconf_all->forwarding && sllao)
2469 addr_flags |= IFA_F_OPTIMISTIC;
2470 #endif
2472 /* Do not allow to create too much of autoconfigured
2473 * addresses; this would be too easy way to crash kernel.
2475 if (!max_addresses ||
2476 ipv6_count_addresses(in6_dev) < max_addresses)
2477 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2478 pinfo->prefix_len,
2479 addr_type&IPV6_ADDR_SCOPE_MASK,
2480 addr_flags, valid_lft,
2481 prefered_lft);
2483 if (IS_ERR_OR_NULL(ifp)) {
2484 in6_dev_put(in6_dev);
2485 return;
2488 update_lft = 0;
2489 create = 1;
2490 spin_lock_bh(&ifp->lock);
2491 ifp->flags |= IFA_F_MANAGETEMPADDR;
2492 ifp->cstamp = jiffies;
2493 ifp->tokenized = tokenized;
2494 spin_unlock_bh(&ifp->lock);
2495 addrconf_dad_start(ifp);
2498 if (ifp) {
2499 u32 flags;
2500 unsigned long now;
2501 u32 stored_lft;
2503 /* update lifetime (RFC2462 5.5.3 e) */
2504 spin_lock_bh(&ifp->lock);
2505 now = jiffies;
2506 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2507 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2508 else
2509 stored_lft = 0;
2510 if (!update_lft && !create && stored_lft) {
2511 const u32 minimum_lft = min_t(u32,
2512 stored_lft, MIN_VALID_LIFETIME);
2513 valid_lft = max(valid_lft, minimum_lft);
2515 /* RFC4862 Section 5.5.3e:
2516 * "Note that the preferred lifetime of the
2517 * corresponding address is always reset to
2518 * the Preferred Lifetime in the received
2519 * Prefix Information option, regardless of
2520 * whether the valid lifetime is also reset or
2521 * ignored."
2523 * So we should always update prefered_lft here.
2525 update_lft = 1;
2528 if (update_lft) {
2529 ifp->valid_lft = valid_lft;
2530 ifp->prefered_lft = prefered_lft;
2531 ifp->tstamp = now;
2532 flags = ifp->flags;
2533 ifp->flags &= ~IFA_F_DEPRECATED;
2534 spin_unlock_bh(&ifp->lock);
2536 if (!(flags&IFA_F_TENTATIVE))
2537 ipv6_ifa_notify(0, ifp);
2538 } else
2539 spin_unlock_bh(&ifp->lock);
2541 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2542 create, now);
2544 in6_ifa_put(ifp);
2545 addrconf_verify();
2548 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2549 in6_dev_put(in6_dev);
2553 * Set destination address.
2554 * Special case for SIT interfaces where we create a new "virtual"
2555 * device.
2557 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2559 struct in6_ifreq ireq;
2560 struct net_device *dev;
2561 int err = -EINVAL;
2563 rtnl_lock();
2565 err = -EFAULT;
2566 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2567 goto err_exit;
2569 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2571 err = -ENODEV;
2572 if (!dev)
2573 goto err_exit;
2575 #if IS_ENABLED(CONFIG_IPV6_SIT)
2576 if (dev->type == ARPHRD_SIT) {
2577 const struct net_device_ops *ops = dev->netdev_ops;
2578 struct ifreq ifr;
2579 struct ip_tunnel_parm p;
2581 err = -EADDRNOTAVAIL;
2582 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2583 goto err_exit;
2585 memset(&p, 0, sizeof(p));
2586 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2587 p.iph.saddr = 0;
2588 p.iph.version = 4;
2589 p.iph.ihl = 5;
2590 p.iph.protocol = IPPROTO_IPV6;
2591 p.iph.ttl = 64;
2592 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2594 if (ops->ndo_do_ioctl) {
2595 mm_segment_t oldfs = get_fs();
2597 set_fs(KERNEL_DS);
2598 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2599 set_fs(oldfs);
2600 } else
2601 err = -EOPNOTSUPP;
2603 if (err == 0) {
2604 err = -ENOBUFS;
2605 dev = __dev_get_by_name(net, p.name);
2606 if (!dev)
2607 goto err_exit;
2608 err = dev_open(dev);
2611 #endif
2613 err_exit:
2614 rtnl_unlock();
2615 return err;
2618 static int ipv6_mc_config(struct sock *sk, bool join,
2619 const struct in6_addr *addr, int ifindex)
2621 int ret;
2623 ASSERT_RTNL();
2625 lock_sock(sk);
2626 if (join)
2627 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2628 else
2629 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2630 release_sock(sk);
2632 return ret;
2636 * Manual configuration of address on an interface
2638 static int inet6_addr_add(struct net *net, int ifindex,
2639 const struct in6_addr *pfx,
2640 const struct in6_addr *peer_pfx,
2641 unsigned int plen, __u32 ifa_flags,
2642 __u32 prefered_lft, __u32 valid_lft)
2644 struct inet6_ifaddr *ifp;
2645 struct inet6_dev *idev;
2646 struct net_device *dev;
2647 unsigned long timeout;
2648 clock_t expires;
2649 int scope;
2650 u32 flags;
2652 ASSERT_RTNL();
2654 if (plen > 128)
2655 return -EINVAL;
2657 /* check the lifetime */
2658 if (!valid_lft || prefered_lft > valid_lft)
2659 return -EINVAL;
2661 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2662 return -EINVAL;
2664 dev = __dev_get_by_index(net, ifindex);
2665 if (!dev)
2666 return -ENODEV;
2668 idev = addrconf_add_dev(dev);
2669 if (IS_ERR(idev))
2670 return PTR_ERR(idev);
2672 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2673 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2674 true, pfx, ifindex);
2676 if (ret < 0)
2677 return ret;
2680 scope = ipv6_addr_scope(pfx);
2682 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2683 if (addrconf_finite_timeout(timeout)) {
2684 expires = jiffies_to_clock_t(timeout * HZ);
2685 valid_lft = timeout;
2686 flags = RTF_EXPIRES;
2687 } else {
2688 expires = 0;
2689 flags = 0;
2690 ifa_flags |= IFA_F_PERMANENT;
2693 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2694 if (addrconf_finite_timeout(timeout)) {
2695 if (timeout == 0)
2696 ifa_flags |= IFA_F_DEPRECATED;
2697 prefered_lft = timeout;
2700 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2701 valid_lft, prefered_lft);
2703 if (!IS_ERR(ifp)) {
2704 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2705 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2706 expires, flags);
2710 * Note that section 3.1 of RFC 4429 indicates
2711 * that the Optimistic flag should not be set for
2712 * manually configured addresses
2714 addrconf_dad_start(ifp);
2715 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2716 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2717 true, jiffies);
2718 in6_ifa_put(ifp);
2719 addrconf_verify_rtnl();
2720 return 0;
2721 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2722 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2723 false, pfx, ifindex);
2726 return PTR_ERR(ifp);
2729 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2730 const struct in6_addr *pfx, unsigned int plen)
2732 struct inet6_ifaddr *ifp;
2733 struct inet6_dev *idev;
2734 struct net_device *dev;
2736 if (plen > 128)
2737 return -EINVAL;
2739 dev = __dev_get_by_index(net, ifindex);
2740 if (!dev)
2741 return -ENODEV;
2743 idev = __in6_dev_get(dev);
2744 if (!idev)
2745 return -ENXIO;
2747 read_lock_bh(&idev->lock);
2748 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2749 if (ifp->prefix_len == plen &&
2750 ipv6_addr_equal(pfx, &ifp->addr)) {
2751 in6_ifa_hold(ifp);
2752 read_unlock_bh(&idev->lock);
2754 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2755 (ifa_flags & IFA_F_MANAGETEMPADDR))
2756 manage_tempaddrs(idev, ifp, 0, 0, false,
2757 jiffies);
2758 ipv6_del_addr(ifp);
2759 addrconf_verify_rtnl();
2760 if (ipv6_addr_is_multicast(pfx)) {
2761 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2762 false, pfx, dev->ifindex);
2764 return 0;
2767 read_unlock_bh(&idev->lock);
2768 return -EADDRNOTAVAIL;
2772 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2774 struct in6_ifreq ireq;
2775 int err;
2777 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2778 return -EPERM;
2780 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2781 return -EFAULT;
2783 rtnl_lock();
2784 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2785 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2786 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2787 rtnl_unlock();
2788 return err;
2791 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2793 struct in6_ifreq ireq;
2794 int err;
2796 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2797 return -EPERM;
2799 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2800 return -EFAULT;
2802 rtnl_lock();
2803 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2804 ireq.ifr6_prefixlen);
2805 rtnl_unlock();
2806 return err;
2809 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2810 int plen, int scope)
2812 struct inet6_ifaddr *ifp;
2814 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2815 scope, IFA_F_PERMANENT,
2816 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2817 if (!IS_ERR(ifp)) {
2818 spin_lock_bh(&ifp->lock);
2819 ifp->flags &= ~IFA_F_TENTATIVE;
2820 spin_unlock_bh(&ifp->lock);
2821 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2822 in6_ifa_put(ifp);
2826 #if IS_ENABLED(CONFIG_IPV6_SIT)
2827 static void sit_add_v4_addrs(struct inet6_dev *idev)
2829 struct in6_addr addr;
2830 struct net_device *dev;
2831 struct net *net = dev_net(idev->dev);
2832 int scope, plen;
2833 u32 pflags = 0;
2835 ASSERT_RTNL();
2837 memset(&addr, 0, sizeof(struct in6_addr));
2838 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2840 if (idev->dev->flags&IFF_POINTOPOINT) {
2841 addr.s6_addr32[0] = htonl(0xfe800000);
2842 scope = IFA_LINK;
2843 plen = 64;
2844 } else {
2845 scope = IPV6_ADDR_COMPATv4;
2846 plen = 96;
2847 pflags |= RTF_NONEXTHOP;
2850 if (addr.s6_addr32[3]) {
2851 add_addr(idev, &addr, plen, scope);
2852 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2853 return;
2856 for_each_netdev(net, dev) {
2857 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2858 if (in_dev && (dev->flags & IFF_UP)) {
2859 struct in_ifaddr *ifa;
2861 int flag = scope;
2863 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2865 addr.s6_addr32[3] = ifa->ifa_local;
2867 if (ifa->ifa_scope == RT_SCOPE_LINK)
2868 continue;
2869 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2870 if (idev->dev->flags&IFF_POINTOPOINT)
2871 continue;
2872 flag |= IFA_HOST;
2875 add_addr(idev, &addr, plen, flag);
2876 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2877 pflags);
2882 #endif
2884 static void init_loopback(struct net_device *dev)
2886 struct inet6_dev *idev;
2887 struct net_device *sp_dev;
2888 struct inet6_ifaddr *sp_ifa;
2889 struct rt6_info *sp_rt;
2891 /* ::1 */
2893 ASSERT_RTNL();
2895 idev = ipv6_find_idev(dev);
2896 if (!idev) {
2897 pr_debug("%s: add_dev failed\n", __func__);
2898 return;
2901 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2903 /* Add routes to other interface's IPv6 addresses */
2904 for_each_netdev(dev_net(dev), sp_dev) {
2905 if (!strcmp(sp_dev->name, dev->name))
2906 continue;
2908 idev = __in6_dev_get(sp_dev);
2909 if (!idev)
2910 continue;
2912 read_lock_bh(&idev->lock);
2913 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2915 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2916 continue;
2918 if (sp_ifa->rt) {
2919 /* This dst has been added to garbage list when
2920 * lo device down, release this obsolete dst and
2921 * reallocate a new router for ifa.
2923 if (sp_ifa->rt->dst.obsolete > 0) {
2924 ip6_rt_put(sp_ifa->rt);
2925 sp_ifa->rt = NULL;
2926 } else {
2927 continue;
2931 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2933 /* Failure cases are ignored */
2934 if (!IS_ERR(sp_rt)) {
2935 sp_ifa->rt = sp_rt;
2936 ip6_ins_rt(sp_rt);
2939 read_unlock_bh(&idev->lock);
2943 static void addrconf_add_linklocal(struct inet6_dev *idev,
2944 const struct in6_addr *addr, u32 flags)
2946 struct inet6_ifaddr *ifp;
2947 u32 addr_flags = flags | IFA_F_PERMANENT;
2949 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2950 if (idev->cnf.optimistic_dad &&
2951 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2952 addr_flags |= IFA_F_OPTIMISTIC;
2953 #endif
2955 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2956 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2957 if (!IS_ERR(ifp)) {
2958 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2959 addrconf_dad_start(ifp);
2960 in6_ifa_put(ifp);
2964 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2966 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2967 return true;
2969 if (address.s6_addr32[2] == htonl(0x02005eff) &&
2970 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2971 return true;
2973 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2974 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2975 return true;
2977 return false;
2980 static int ipv6_generate_stable_address(struct in6_addr *address,
2981 u8 dad_count,
2982 const struct inet6_dev *idev)
2984 static DEFINE_SPINLOCK(lock);
2985 static __u32 digest[SHA_DIGEST_WORDS];
2986 static __u32 workspace[SHA_WORKSPACE_WORDS];
2988 static union {
2989 char __data[SHA_MESSAGE_BYTES];
2990 struct {
2991 struct in6_addr secret;
2992 __be32 prefix[2];
2993 unsigned char hwaddr[MAX_ADDR_LEN];
2994 u8 dad_count;
2995 } __packed;
2996 } data;
2998 struct in6_addr secret;
2999 struct in6_addr temp;
3000 struct net *net = dev_net(idev->dev);
3002 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3004 if (idev->cnf.stable_secret.initialized)
3005 secret = idev->cnf.stable_secret.secret;
3006 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3007 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3008 else
3009 return -1;
3011 retry:
3012 spin_lock_bh(&lock);
3014 sha_init(digest);
3015 memset(&data, 0, sizeof(data));
3016 memset(workspace, 0, sizeof(workspace));
3017 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3018 data.prefix[0] = address->s6_addr32[0];
3019 data.prefix[1] = address->s6_addr32[1];
3020 data.secret = secret;
3021 data.dad_count = dad_count;
3023 sha_transform(digest, data.__data, workspace);
3025 temp = *address;
3026 temp.s6_addr32[2] = (__force __be32)digest[0];
3027 temp.s6_addr32[3] = (__force __be32)digest[1];
3029 spin_unlock_bh(&lock);
3031 if (ipv6_reserved_interfaceid(temp)) {
3032 dad_count++;
3033 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3034 return -1;
3035 goto retry;
3038 *address = temp;
3039 return 0;
3042 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3044 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3046 if (s->initialized)
3047 return;
3048 s = &idev->cnf.stable_secret;
3049 get_random_bytes(&s->secret, sizeof(s->secret));
3050 s->initialized = true;
3053 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3055 struct in6_addr addr;
3057 /* no link local addresses on L3 master devices */
3058 if (netif_is_l3_master(idev->dev))
3059 return;
3061 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3063 switch (idev->addr_gen_mode) {
3064 case IN6_ADDR_GEN_MODE_RANDOM:
3065 ipv6_gen_mode_random_init(idev);
3066 /* fallthrough */
3067 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3068 if (!ipv6_generate_stable_address(&addr, 0, idev))
3069 addrconf_add_linklocal(idev, &addr,
3070 IFA_F_STABLE_PRIVACY);
3071 else if (prefix_route)
3072 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3073 break;
3074 case IN6_ADDR_GEN_MODE_EUI64:
3075 /* addrconf_add_linklocal also adds a prefix_route and we
3076 * only need to care about prefix routes if ipv6_generate_eui64
3077 * couldn't generate one.
3079 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3080 addrconf_add_linklocal(idev, &addr, 0);
3081 else if (prefix_route)
3082 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3083 break;
3084 case IN6_ADDR_GEN_MODE_NONE:
3085 default:
3086 /* will not add any link local address */
3087 break;
3091 static void addrconf_dev_config(struct net_device *dev)
3093 struct inet6_dev *idev;
3095 ASSERT_RTNL();
3097 if ((dev->type != ARPHRD_ETHER) &&
3098 (dev->type != ARPHRD_FDDI) &&
3099 (dev->type != ARPHRD_ARCNET) &&
3100 (dev->type != ARPHRD_INFINIBAND) &&
3101 (dev->type != ARPHRD_IEEE1394) &&
3102 (dev->type != ARPHRD_TUNNEL6) &&
3103 (dev->type != ARPHRD_6LOWPAN) &&
3104 (dev->type != ARPHRD_NONE)) {
3105 /* Alas, we support only Ethernet autoconfiguration. */
3106 return;
3109 idev = addrconf_add_dev(dev);
3110 if (IS_ERR(idev))
3111 return;
3113 /* this device type has no EUI support */
3114 if (dev->type == ARPHRD_NONE &&
3115 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3116 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3118 addrconf_addr_gen(idev, false);
3121 #if IS_ENABLED(CONFIG_IPV6_SIT)
3122 static void addrconf_sit_config(struct net_device *dev)
3124 struct inet6_dev *idev;
3126 ASSERT_RTNL();
3129 * Configure the tunnel with one of our IPv4
3130 * addresses... we should configure all of
3131 * our v4 addrs in the tunnel
3134 idev = ipv6_find_idev(dev);
3135 if (!idev) {
3136 pr_debug("%s: add_dev failed\n", __func__);
3137 return;
3140 if (dev->priv_flags & IFF_ISATAP) {
3141 addrconf_addr_gen(idev, false);
3142 return;
3145 sit_add_v4_addrs(idev);
3147 if (dev->flags&IFF_POINTOPOINT)
3148 addrconf_add_mroute(dev);
3150 #endif
3152 #if IS_ENABLED(CONFIG_NET_IPGRE)
3153 static void addrconf_gre_config(struct net_device *dev)
3155 struct inet6_dev *idev;
3157 ASSERT_RTNL();
3159 idev = ipv6_find_idev(dev);
3160 if (!idev) {
3161 pr_debug("%s: add_dev failed\n", __func__);
3162 return;
3165 addrconf_addr_gen(idev, true);
3166 if (dev->flags & IFF_POINTOPOINT)
3167 addrconf_add_mroute(dev);
3169 #endif
3171 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3172 void *ptr)
3174 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3175 struct inet6_dev *idev = __in6_dev_get(dev);
3176 int run_pending = 0;
3177 int err;
3179 switch (event) {
3180 case NETDEV_REGISTER:
3181 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3182 idev = ipv6_add_dev(dev);
3183 if (IS_ERR(idev))
3184 return notifier_from_errno(PTR_ERR(idev));
3186 break;
3188 case NETDEV_CHANGEMTU:
3189 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3190 if (dev->mtu < IPV6_MIN_MTU) {
3191 addrconf_ifdown(dev, 1);
3192 break;
3195 if (idev) {
3196 rt6_mtu_change(dev, dev->mtu);
3197 idev->cnf.mtu6 = dev->mtu;
3198 break;
3201 /* allocate new idev */
3202 idev = ipv6_add_dev(dev);
3203 if (IS_ERR(idev))
3204 break;
3206 /* device is still not ready */
3207 if (!(idev->if_flags & IF_READY))
3208 break;
3210 run_pending = 1;
3212 /* fall through */
3214 case NETDEV_UP:
3215 case NETDEV_CHANGE:
3216 if (dev->flags & IFF_SLAVE)
3217 break;
3219 if (idev && idev->cnf.disable_ipv6)
3220 break;
3222 if (event == NETDEV_UP) {
3223 if (!addrconf_qdisc_ok(dev)) {
3224 /* device is not ready yet. */
3225 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3226 dev->name);
3227 break;
3230 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3231 idev = ipv6_add_dev(dev);
3233 if (!IS_ERR_OR_NULL(idev)) {
3234 idev->if_flags |= IF_READY;
3235 run_pending = 1;
3237 } else if (event == NETDEV_CHANGE) {
3238 if (!addrconf_qdisc_ok(dev)) {
3239 /* device is still not ready. */
3240 break;
3243 if (idev) {
3244 if (idev->if_flags & IF_READY)
3245 /* device is already configured. */
3246 break;
3247 idev->if_flags |= IF_READY;
3250 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3251 dev->name);
3253 run_pending = 1;
3256 switch (dev->type) {
3257 #if IS_ENABLED(CONFIG_IPV6_SIT)
3258 case ARPHRD_SIT:
3259 addrconf_sit_config(dev);
3260 break;
3261 #endif
3262 #if IS_ENABLED(CONFIG_NET_IPGRE)
3263 case ARPHRD_IPGRE:
3264 addrconf_gre_config(dev);
3265 break;
3266 #endif
3267 case ARPHRD_LOOPBACK:
3268 init_loopback(dev);
3269 break;
3271 default:
3272 addrconf_dev_config(dev);
3273 break;
3276 if (!IS_ERR_OR_NULL(idev)) {
3277 if (run_pending)
3278 addrconf_dad_run(idev);
3281 * If the MTU changed during the interface down,
3282 * when the interface up, the changed MTU must be
3283 * reflected in the idev as well as routers.
3285 if (idev->cnf.mtu6 != dev->mtu &&
3286 dev->mtu >= IPV6_MIN_MTU) {
3287 rt6_mtu_change(dev, dev->mtu);
3288 idev->cnf.mtu6 = dev->mtu;
3290 idev->tstamp = jiffies;
3291 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3294 * If the changed mtu during down is lower than
3295 * IPV6_MIN_MTU stop IPv6 on this interface.
3297 if (dev->mtu < IPV6_MIN_MTU)
3298 addrconf_ifdown(dev, 1);
3300 break;
3302 case NETDEV_DOWN:
3303 case NETDEV_UNREGISTER:
3305 * Remove all addresses from this interface.
3307 addrconf_ifdown(dev, event != NETDEV_DOWN);
3308 break;
3310 case NETDEV_CHANGENAME:
3311 if (idev) {
3312 snmp6_unregister_dev(idev);
3313 addrconf_sysctl_unregister(idev);
3314 err = addrconf_sysctl_register(idev);
3315 if (err)
3316 return notifier_from_errno(err);
3317 err = snmp6_register_dev(idev);
3318 if (err) {
3319 addrconf_sysctl_unregister(idev);
3320 return notifier_from_errno(err);
3323 break;
3325 case NETDEV_PRE_TYPE_CHANGE:
3326 case NETDEV_POST_TYPE_CHANGE:
3327 if (idev)
3328 addrconf_type_change(dev, event);
3329 break;
3332 return NOTIFY_OK;
3336 * addrconf module should be notified of a device going up
3338 static struct notifier_block ipv6_dev_notf = {
3339 .notifier_call = addrconf_notify,
3342 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3344 struct inet6_dev *idev;
3345 ASSERT_RTNL();
3347 idev = __in6_dev_get(dev);
3349 if (event == NETDEV_POST_TYPE_CHANGE)
3350 ipv6_mc_remap(idev);
3351 else if (event == NETDEV_PRE_TYPE_CHANGE)
3352 ipv6_mc_unmap(idev);
3355 static int addrconf_ifdown(struct net_device *dev, int how)
3357 struct net *net = dev_net(dev);
3358 struct inet6_dev *idev;
3359 struct inet6_ifaddr *ifa;
3360 int state, i;
3362 ASSERT_RTNL();
3364 rt6_ifdown(net, dev);
3365 neigh_ifdown(&nd_tbl, dev);
3367 idev = __in6_dev_get(dev);
3368 if (!idev)
3369 return -ENODEV;
3372 * Step 1: remove reference to ipv6 device from parent device.
3373 * Do not dev_put!
3375 if (how) {
3376 idev->dead = 1;
3378 /* protected by rtnl_lock */
3379 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3381 /* Step 1.5: remove snmp6 entry */
3382 snmp6_unregister_dev(idev);
3386 /* Step 2: clear hash table */
3387 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3388 struct hlist_head *h = &inet6_addr_lst[i];
3390 spin_lock_bh(&addrconf_hash_lock);
3391 restart:
3392 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3393 if (ifa->idev == idev) {
3394 hlist_del_init_rcu(&ifa->addr_lst);
3395 addrconf_del_dad_work(ifa);
3396 goto restart;
3399 spin_unlock_bh(&addrconf_hash_lock);
3402 write_lock_bh(&idev->lock);
3404 addrconf_del_rs_timer(idev);
3406 /* Step 2: clear flags for stateless addrconf */
3407 if (!how)
3408 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3410 if (how && del_timer(&idev->regen_timer))
3411 in6_dev_put(idev);
3413 /* Step 3: clear tempaddr list */
3414 while (!list_empty(&idev->tempaddr_list)) {
3415 ifa = list_first_entry(&idev->tempaddr_list,
3416 struct inet6_ifaddr, tmp_list);
3417 list_del(&ifa->tmp_list);
3418 write_unlock_bh(&idev->lock);
3419 spin_lock_bh(&ifa->lock);
3421 if (ifa->ifpub) {
3422 in6_ifa_put(ifa->ifpub);
3423 ifa->ifpub = NULL;
3425 spin_unlock_bh(&ifa->lock);
3426 in6_ifa_put(ifa);
3427 write_lock_bh(&idev->lock);
3430 while (!list_empty(&idev->addr_list)) {
3431 ifa = list_first_entry(&idev->addr_list,
3432 struct inet6_ifaddr, if_list);
3433 addrconf_del_dad_work(ifa);
3435 list_del(&ifa->if_list);
3437 write_unlock_bh(&idev->lock);
3439 spin_lock_bh(&ifa->lock);
3440 state = ifa->state;
3441 ifa->state = INET6_IFADDR_STATE_DEAD;
3442 spin_unlock_bh(&ifa->lock);
3444 if (state != INET6_IFADDR_STATE_DEAD) {
3445 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3446 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3448 in6_ifa_put(ifa);
3450 write_lock_bh(&idev->lock);
3453 write_unlock_bh(&idev->lock);
3455 /* Step 5: Discard anycast and multicast list */
3456 if (how) {
3457 ipv6_ac_destroy_dev(idev);
3458 ipv6_mc_destroy_dev(idev);
3459 } else {
3460 ipv6_mc_down(idev);
3463 idev->tstamp = jiffies;
3465 /* Last: Shot the device (if unregistered) */
3466 if (how) {
3467 addrconf_sysctl_unregister(idev);
3468 neigh_parms_release(&nd_tbl, idev->nd_parms);
3469 neigh_ifdown(&nd_tbl, dev);
3470 in6_dev_put(idev);
3472 return 0;
3475 static void addrconf_rs_timer(unsigned long data)
3477 struct inet6_dev *idev = (struct inet6_dev *)data;
3478 struct net_device *dev = idev->dev;
3479 struct in6_addr lladdr;
3481 write_lock(&idev->lock);
3482 if (idev->dead || !(idev->if_flags & IF_READY))
3483 goto out;
3485 if (!ipv6_accept_ra(idev))
3486 goto out;
3488 /* Announcement received after solicitation was sent */
3489 if (idev->if_flags & IF_RA_RCVD)
3490 goto out;
3492 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3493 write_unlock(&idev->lock);
3494 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3495 ndisc_send_rs(dev, &lladdr,
3496 &in6addr_linklocal_allrouters);
3497 else
3498 goto put;
3500 write_lock(&idev->lock);
3501 /* The wait after the last probe can be shorter */
3502 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3503 idev->cnf.rtr_solicits) ?
3504 idev->cnf.rtr_solicit_delay :
3505 idev->cnf.rtr_solicit_interval);
3506 } else {
3508 * Note: we do not support deprecated "all on-link"
3509 * assumption any longer.
3511 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3514 out:
3515 write_unlock(&idev->lock);
3516 put:
3517 in6_dev_put(idev);
3521 * Duplicate Address Detection
3523 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3525 unsigned long rand_num;
3526 struct inet6_dev *idev = ifp->idev;
3528 if (ifp->flags & IFA_F_OPTIMISTIC)
3529 rand_num = 0;
3530 else
3531 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3533 ifp->dad_probes = idev->cnf.dad_transmits;
3534 addrconf_mod_dad_work(ifp, rand_num);
3537 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3539 struct inet6_dev *idev = ifp->idev;
3540 struct net_device *dev = idev->dev;
3542 addrconf_join_solict(dev, &ifp->addr);
3544 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3546 read_lock_bh(&idev->lock);
3547 spin_lock(&ifp->lock);
3548 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3549 goto out;
3551 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3552 idev->cnf.accept_dad < 1 ||
3553 !(ifp->flags&IFA_F_TENTATIVE) ||
3554 ifp->flags & IFA_F_NODAD) {
3555 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3556 spin_unlock(&ifp->lock);
3557 read_unlock_bh(&idev->lock);
3559 addrconf_dad_completed(ifp);
3560 return;
3563 if (!(idev->if_flags & IF_READY)) {
3564 spin_unlock(&ifp->lock);
3565 read_unlock_bh(&idev->lock);
3567 * If the device is not ready:
3568 * - keep it tentative if it is a permanent address.
3569 * - otherwise, kill it.
3571 in6_ifa_hold(ifp);
3572 addrconf_dad_stop(ifp, 0);
3573 return;
3577 * Optimistic nodes can start receiving
3578 * Frames right away
3580 if (ifp->flags & IFA_F_OPTIMISTIC) {
3581 ip6_ins_rt(ifp->rt);
3582 if (ipv6_use_optimistic_addr(idev)) {
3583 /* Because optimistic nodes can use this address,
3584 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3586 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3590 addrconf_dad_kick(ifp);
3591 out:
3592 spin_unlock(&ifp->lock);
3593 read_unlock_bh(&idev->lock);
3596 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3598 bool begin_dad = false;
3600 spin_lock_bh(&ifp->lock);
3601 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3602 ifp->state = INET6_IFADDR_STATE_PREDAD;
3603 begin_dad = true;
3605 spin_unlock_bh(&ifp->lock);
3607 if (begin_dad)
3608 addrconf_mod_dad_work(ifp, 0);
3611 static void addrconf_dad_work(struct work_struct *w)
3613 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3614 struct inet6_ifaddr,
3615 dad_work);
3616 struct inet6_dev *idev = ifp->idev;
3617 struct in6_addr mcaddr;
3619 enum {
3620 DAD_PROCESS,
3621 DAD_BEGIN,
3622 DAD_ABORT,
3623 } action = DAD_PROCESS;
3625 rtnl_lock();
3627 spin_lock_bh(&ifp->lock);
3628 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3629 action = DAD_BEGIN;
3630 ifp->state = INET6_IFADDR_STATE_DAD;
3631 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3632 action = DAD_ABORT;
3633 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3635 spin_unlock_bh(&ifp->lock);
3637 if (action == DAD_BEGIN) {
3638 addrconf_dad_begin(ifp);
3639 goto out;
3640 } else if (action == DAD_ABORT) {
3641 addrconf_dad_stop(ifp, 1);
3642 goto out;
3645 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3646 goto out;
3648 write_lock_bh(&idev->lock);
3649 if (idev->dead || !(idev->if_flags & IF_READY)) {
3650 write_unlock_bh(&idev->lock);
3651 goto out;
3654 spin_lock(&ifp->lock);
3655 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3656 spin_unlock(&ifp->lock);
3657 write_unlock_bh(&idev->lock);
3658 goto out;
3661 if (ifp->dad_probes == 0) {
3663 * DAD was successful
3666 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3667 spin_unlock(&ifp->lock);
3668 write_unlock_bh(&idev->lock);
3670 addrconf_dad_completed(ifp);
3672 goto out;
3675 ifp->dad_probes--;
3676 addrconf_mod_dad_work(ifp,
3677 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3678 spin_unlock(&ifp->lock);
3679 write_unlock_bh(&idev->lock);
3681 /* send a neighbour solicitation for our addr */
3682 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3683 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3684 out:
3685 in6_ifa_put(ifp);
3686 rtnl_unlock();
3689 /* ifp->idev must be at least read locked */
3690 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3692 struct inet6_ifaddr *ifpiter;
3693 struct inet6_dev *idev = ifp->idev;
3695 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3696 if (ifpiter->scope > IFA_LINK)
3697 break;
3698 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3699 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3700 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3701 IFA_F_PERMANENT)
3702 return false;
3704 return true;
3707 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3709 struct net_device *dev = ifp->idev->dev;
3710 struct in6_addr lladdr;
3711 bool send_rs, send_mld;
3713 addrconf_del_dad_work(ifp);
3716 * Configure the address for reception. Now it is valid.
3719 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3721 /* If added prefix is link local and we are prepared to process
3722 router advertisements, start sending router solicitations.
3725 read_lock_bh(&ifp->idev->lock);
3726 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3727 send_rs = send_mld &&
3728 ipv6_accept_ra(ifp->idev) &&
3729 ifp->idev->cnf.rtr_solicits > 0 &&
3730 (dev->flags&IFF_LOOPBACK) == 0;
3731 read_unlock_bh(&ifp->idev->lock);
3733 /* While dad is in progress mld report's source address is in6_addrany.
3734 * Resend with proper ll now.
3736 if (send_mld)
3737 ipv6_mc_dad_complete(ifp->idev);
3739 if (send_rs) {
3741 * If a host as already performed a random delay
3742 * [...] as part of DAD [...] there is no need
3743 * to delay again before sending the first RS
3745 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3746 return;
3747 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3749 write_lock_bh(&ifp->idev->lock);
3750 spin_lock(&ifp->lock);
3751 ifp->idev->rs_probes = 1;
3752 ifp->idev->if_flags |= IF_RS_SENT;
3753 addrconf_mod_rs_timer(ifp->idev,
3754 ifp->idev->cnf.rtr_solicit_interval);
3755 spin_unlock(&ifp->lock);
3756 write_unlock_bh(&ifp->idev->lock);
3760 static void addrconf_dad_run(struct inet6_dev *idev)
3762 struct inet6_ifaddr *ifp;
3764 read_lock_bh(&idev->lock);
3765 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3766 spin_lock(&ifp->lock);
3767 if (ifp->flags & IFA_F_TENTATIVE &&
3768 ifp->state == INET6_IFADDR_STATE_DAD)
3769 addrconf_dad_kick(ifp);
3770 spin_unlock(&ifp->lock);
3772 read_unlock_bh(&idev->lock);
3775 #ifdef CONFIG_PROC_FS
3776 struct if6_iter_state {
3777 struct seq_net_private p;
3778 int bucket;
3779 int offset;
3782 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3784 struct inet6_ifaddr *ifa = NULL;
3785 struct if6_iter_state *state = seq->private;
3786 struct net *net = seq_file_net(seq);
3787 int p = 0;
3789 /* initial bucket if pos is 0 */
3790 if (pos == 0) {
3791 state->bucket = 0;
3792 state->offset = 0;
3795 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3796 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3797 addr_lst) {
3798 if (!net_eq(dev_net(ifa->idev->dev), net))
3799 continue;
3800 /* sync with offset */
3801 if (p < state->offset) {
3802 p++;
3803 continue;
3805 state->offset++;
3806 return ifa;
3809 /* prepare for next bucket */
3810 state->offset = 0;
3811 p = 0;
3813 return NULL;
3816 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3817 struct inet6_ifaddr *ifa)
3819 struct if6_iter_state *state = seq->private;
3820 struct net *net = seq_file_net(seq);
3822 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3823 if (!net_eq(dev_net(ifa->idev->dev), net))
3824 continue;
3825 state->offset++;
3826 return ifa;
3829 while (++state->bucket < IN6_ADDR_HSIZE) {
3830 state->offset = 0;
3831 hlist_for_each_entry_rcu_bh(ifa,
3832 &inet6_addr_lst[state->bucket], addr_lst) {
3833 if (!net_eq(dev_net(ifa->idev->dev), net))
3834 continue;
3835 state->offset++;
3836 return ifa;
3840 return NULL;
3843 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3844 __acquires(rcu_bh)
3846 rcu_read_lock_bh();
3847 return if6_get_first(seq, *pos);
3850 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3852 struct inet6_ifaddr *ifa;
3854 ifa = if6_get_next(seq, v);
3855 ++*pos;
3856 return ifa;
3859 static void if6_seq_stop(struct seq_file *seq, void *v)
3860 __releases(rcu_bh)
3862 rcu_read_unlock_bh();
3865 static int if6_seq_show(struct seq_file *seq, void *v)
3867 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3868 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3869 &ifp->addr,
3870 ifp->idev->dev->ifindex,
3871 ifp->prefix_len,
3872 ifp->scope,
3873 (u8) ifp->flags,
3874 ifp->idev->dev->name);
3875 return 0;
3878 static const struct seq_operations if6_seq_ops = {
3879 .start = if6_seq_start,
3880 .next = if6_seq_next,
3881 .show = if6_seq_show,
3882 .stop = if6_seq_stop,
3885 static int if6_seq_open(struct inode *inode, struct file *file)
3887 return seq_open_net(inode, file, &if6_seq_ops,
3888 sizeof(struct if6_iter_state));
3891 static const struct file_operations if6_fops = {
3892 .owner = THIS_MODULE,
3893 .open = if6_seq_open,
3894 .read = seq_read,
3895 .llseek = seq_lseek,
3896 .release = seq_release_net,
3899 static int __net_init if6_proc_net_init(struct net *net)
3901 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3902 return -ENOMEM;
3903 return 0;
3906 static void __net_exit if6_proc_net_exit(struct net *net)
3908 remove_proc_entry("if_inet6", net->proc_net);
3911 static struct pernet_operations if6_proc_net_ops = {
3912 .init = if6_proc_net_init,
3913 .exit = if6_proc_net_exit,
3916 int __init if6_proc_init(void)
3918 return register_pernet_subsys(&if6_proc_net_ops);
3921 void if6_proc_exit(void)
3923 unregister_pernet_subsys(&if6_proc_net_ops);
3925 #endif /* CONFIG_PROC_FS */
3927 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3928 /* Check if address is a home address configured on any interface. */
3929 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3931 int ret = 0;
3932 struct inet6_ifaddr *ifp = NULL;
3933 unsigned int hash = inet6_addr_hash(addr);
3935 rcu_read_lock_bh();
3936 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3937 if (!net_eq(dev_net(ifp->idev->dev), net))
3938 continue;
3939 if (ipv6_addr_equal(&ifp->addr, addr) &&
3940 (ifp->flags & IFA_F_HOMEADDRESS)) {
3941 ret = 1;
3942 break;
3945 rcu_read_unlock_bh();
3946 return ret;
3948 #endif
3951 * Periodic address status verification
3954 static void addrconf_verify_rtnl(void)
3956 unsigned long now, next, next_sec, next_sched;
3957 struct inet6_ifaddr *ifp;
3958 int i;
3960 ASSERT_RTNL();
3962 rcu_read_lock_bh();
3963 now = jiffies;
3964 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3966 cancel_delayed_work(&addr_chk_work);
3968 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3969 restart:
3970 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3971 unsigned long age;
3973 /* When setting preferred_lft to a value not zero or
3974 * infinity, while valid_lft is infinity
3975 * IFA_F_PERMANENT has a non-infinity life time.
3977 if ((ifp->flags & IFA_F_PERMANENT) &&
3978 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3979 continue;
3981 spin_lock(&ifp->lock);
3982 /* We try to batch several events at once. */
3983 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3985 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3986 age >= ifp->valid_lft) {
3987 spin_unlock(&ifp->lock);
3988 in6_ifa_hold(ifp);
3989 ipv6_del_addr(ifp);
3990 goto restart;
3991 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3992 spin_unlock(&ifp->lock);
3993 continue;
3994 } else if (age >= ifp->prefered_lft) {
3995 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3996 int deprecate = 0;
3998 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3999 deprecate = 1;
4000 ifp->flags |= IFA_F_DEPRECATED;
4003 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4004 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4005 next = ifp->tstamp + ifp->valid_lft * HZ;
4007 spin_unlock(&ifp->lock);
4009 if (deprecate) {
4010 in6_ifa_hold(ifp);
4012 ipv6_ifa_notify(0, ifp);
4013 in6_ifa_put(ifp);
4014 goto restart;
4016 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4017 !(ifp->flags&IFA_F_TENTATIVE)) {
4018 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4019 ifp->idev->cnf.dad_transmits *
4020 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4022 if (age >= ifp->prefered_lft - regen_advance) {
4023 struct inet6_ifaddr *ifpub = ifp->ifpub;
4024 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4025 next = ifp->tstamp + ifp->prefered_lft * HZ;
4026 if (!ifp->regen_count && ifpub) {
4027 ifp->regen_count++;
4028 in6_ifa_hold(ifp);
4029 in6_ifa_hold(ifpub);
4030 spin_unlock(&ifp->lock);
4032 spin_lock(&ifpub->lock);
4033 ifpub->regen_count = 0;
4034 spin_unlock(&ifpub->lock);
4035 ipv6_create_tempaddr(ifpub, ifp);
4036 in6_ifa_put(ifpub);
4037 in6_ifa_put(ifp);
4038 goto restart;
4040 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4041 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4042 spin_unlock(&ifp->lock);
4043 } else {
4044 /* ifp->prefered_lft <= ifp->valid_lft */
4045 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4046 next = ifp->tstamp + ifp->prefered_lft * HZ;
4047 spin_unlock(&ifp->lock);
4052 next_sec = round_jiffies_up(next);
4053 next_sched = next;
4055 /* If rounded timeout is accurate enough, accept it. */
4056 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4057 next_sched = next_sec;
4059 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4060 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4061 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4063 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4064 now, next, next_sec, next_sched);
4065 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4066 rcu_read_unlock_bh();
4069 static void addrconf_verify_work(struct work_struct *w)
4071 rtnl_lock();
4072 addrconf_verify_rtnl();
4073 rtnl_unlock();
4076 static void addrconf_verify(void)
4078 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4081 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4082 struct in6_addr **peer_pfx)
4084 struct in6_addr *pfx = NULL;
4086 *peer_pfx = NULL;
4088 if (addr)
4089 pfx = nla_data(addr);
4091 if (local) {
4092 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4093 *peer_pfx = pfx;
4094 pfx = nla_data(local);
4097 return pfx;
4100 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4101 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4102 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4103 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4104 [IFA_FLAGS] = { .len = sizeof(u32) },
4107 static int
4108 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4110 struct net *net = sock_net(skb->sk);
4111 struct ifaddrmsg *ifm;
4112 struct nlattr *tb[IFA_MAX+1];
4113 struct in6_addr *pfx, *peer_pfx;
4114 u32 ifa_flags;
4115 int err;
4117 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4118 if (err < 0)
4119 return err;
4121 ifm = nlmsg_data(nlh);
4122 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4123 if (!pfx)
4124 return -EINVAL;
4126 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4128 /* We ignore other flags so far. */
4129 ifa_flags &= IFA_F_MANAGETEMPADDR;
4131 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4132 ifm->ifa_prefixlen);
4135 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4136 u32 prefered_lft, u32 valid_lft)
4138 u32 flags;
4139 clock_t expires;
4140 unsigned long timeout;
4141 bool was_managetempaddr;
4142 bool had_prefixroute;
4144 ASSERT_RTNL();
4146 if (!valid_lft || (prefered_lft > valid_lft))
4147 return -EINVAL;
4149 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4150 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4151 return -EINVAL;
4153 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4154 if (addrconf_finite_timeout(timeout)) {
4155 expires = jiffies_to_clock_t(timeout * HZ);
4156 valid_lft = timeout;
4157 flags = RTF_EXPIRES;
4158 } else {
4159 expires = 0;
4160 flags = 0;
4161 ifa_flags |= IFA_F_PERMANENT;
4164 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4165 if (addrconf_finite_timeout(timeout)) {
4166 if (timeout == 0)
4167 ifa_flags |= IFA_F_DEPRECATED;
4168 prefered_lft = timeout;
4171 spin_lock_bh(&ifp->lock);
4172 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4173 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4174 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4175 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4176 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4177 IFA_F_NOPREFIXROUTE);
4178 ifp->flags |= ifa_flags;
4179 ifp->tstamp = jiffies;
4180 ifp->valid_lft = valid_lft;
4181 ifp->prefered_lft = prefered_lft;
4183 spin_unlock_bh(&ifp->lock);
4184 if (!(ifp->flags&IFA_F_TENTATIVE))
4185 ipv6_ifa_notify(0, ifp);
4187 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4188 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4189 expires, flags);
4190 } else if (had_prefixroute) {
4191 enum cleanup_prefix_rt_t action;
4192 unsigned long rt_expires;
4194 write_lock_bh(&ifp->idev->lock);
4195 action = check_cleanup_prefix_route(ifp, &rt_expires);
4196 write_unlock_bh(&ifp->idev->lock);
4198 if (action != CLEANUP_PREFIX_RT_NOP) {
4199 cleanup_prefix_route(ifp, rt_expires,
4200 action == CLEANUP_PREFIX_RT_DEL);
4204 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4205 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4206 valid_lft = prefered_lft = 0;
4207 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4208 !was_managetempaddr, jiffies);
4211 addrconf_verify_rtnl();
4213 return 0;
4216 static int
4217 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4219 struct net *net = sock_net(skb->sk);
4220 struct ifaddrmsg *ifm;
4221 struct nlattr *tb[IFA_MAX+1];
4222 struct in6_addr *pfx, *peer_pfx;
4223 struct inet6_ifaddr *ifa;
4224 struct net_device *dev;
4225 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4226 u32 ifa_flags;
4227 int err;
4229 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4230 if (err < 0)
4231 return err;
4233 ifm = nlmsg_data(nlh);
4234 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4235 if (!pfx)
4236 return -EINVAL;
4238 if (tb[IFA_CACHEINFO]) {
4239 struct ifa_cacheinfo *ci;
4241 ci = nla_data(tb[IFA_CACHEINFO]);
4242 valid_lft = ci->ifa_valid;
4243 preferred_lft = ci->ifa_prefered;
4244 } else {
4245 preferred_lft = INFINITY_LIFE_TIME;
4246 valid_lft = INFINITY_LIFE_TIME;
4249 dev = __dev_get_by_index(net, ifm->ifa_index);
4250 if (!dev)
4251 return -ENODEV;
4253 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4255 /* We ignore other flags so far. */
4256 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4257 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4259 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4260 if (!ifa) {
4262 * It would be best to check for !NLM_F_CREATE here but
4263 * userspace already relies on not having to provide this.
4265 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4266 ifm->ifa_prefixlen, ifa_flags,
4267 preferred_lft, valid_lft);
4270 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4271 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4272 err = -EEXIST;
4273 else
4274 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4276 in6_ifa_put(ifa);
4278 return err;
4281 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4282 u8 scope, int ifindex)
4284 struct ifaddrmsg *ifm;
4286 ifm = nlmsg_data(nlh);
4287 ifm->ifa_family = AF_INET6;
4288 ifm->ifa_prefixlen = prefixlen;
4289 ifm->ifa_flags = flags;
4290 ifm->ifa_scope = scope;
4291 ifm->ifa_index = ifindex;
4294 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4295 unsigned long tstamp, u32 preferred, u32 valid)
4297 struct ifa_cacheinfo ci;
4299 ci.cstamp = cstamp_delta(cstamp);
4300 ci.tstamp = cstamp_delta(tstamp);
4301 ci.ifa_prefered = preferred;
4302 ci.ifa_valid = valid;
4304 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4307 static inline int rt_scope(int ifa_scope)
4309 if (ifa_scope & IFA_HOST)
4310 return RT_SCOPE_HOST;
4311 else if (ifa_scope & IFA_LINK)
4312 return RT_SCOPE_LINK;
4313 else if (ifa_scope & IFA_SITE)
4314 return RT_SCOPE_SITE;
4315 else
4316 return RT_SCOPE_UNIVERSE;
4319 static inline int inet6_ifaddr_msgsize(void)
4321 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4322 + nla_total_size(16) /* IFA_LOCAL */
4323 + nla_total_size(16) /* IFA_ADDRESS */
4324 + nla_total_size(sizeof(struct ifa_cacheinfo))
4325 + nla_total_size(4) /* IFA_FLAGS */;
4328 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4329 u32 portid, u32 seq, int event, unsigned int flags)
4331 struct nlmsghdr *nlh;
4332 u32 preferred, valid;
4334 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4335 if (!nlh)
4336 return -EMSGSIZE;
4338 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4339 ifa->idev->dev->ifindex);
4341 if (!((ifa->flags&IFA_F_PERMANENT) &&
4342 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4343 preferred = ifa->prefered_lft;
4344 valid = ifa->valid_lft;
4345 if (preferred != INFINITY_LIFE_TIME) {
4346 long tval = (jiffies - ifa->tstamp)/HZ;
4347 if (preferred > tval)
4348 preferred -= tval;
4349 else
4350 preferred = 0;
4351 if (valid != INFINITY_LIFE_TIME) {
4352 if (valid > tval)
4353 valid -= tval;
4354 else
4355 valid = 0;
4358 } else {
4359 preferred = INFINITY_LIFE_TIME;
4360 valid = INFINITY_LIFE_TIME;
4363 if (!ipv6_addr_any(&ifa->peer_addr)) {
4364 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4365 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4366 goto error;
4367 } else
4368 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4369 goto error;
4371 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4372 goto error;
4374 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4375 goto error;
4377 nlmsg_end(skb, nlh);
4378 return 0;
4380 error:
4381 nlmsg_cancel(skb, nlh);
4382 return -EMSGSIZE;
4385 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4386 u32 portid, u32 seq, int event, u16 flags)
4388 struct nlmsghdr *nlh;
4389 u8 scope = RT_SCOPE_UNIVERSE;
4390 int ifindex = ifmca->idev->dev->ifindex;
4392 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4393 scope = RT_SCOPE_SITE;
4395 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4396 if (!nlh)
4397 return -EMSGSIZE;
4399 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4400 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4401 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4402 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4403 nlmsg_cancel(skb, nlh);
4404 return -EMSGSIZE;
4407 nlmsg_end(skb, nlh);
4408 return 0;
4411 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4412 u32 portid, u32 seq, int event, unsigned int flags)
4414 struct nlmsghdr *nlh;
4415 u8 scope = RT_SCOPE_UNIVERSE;
4416 int ifindex = ifaca->aca_idev->dev->ifindex;
4418 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4419 scope = RT_SCOPE_SITE;
4421 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4422 if (!nlh)
4423 return -EMSGSIZE;
4425 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4426 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4427 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4428 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4429 nlmsg_cancel(skb, nlh);
4430 return -EMSGSIZE;
4433 nlmsg_end(skb, nlh);
4434 return 0;
4437 enum addr_type_t {
4438 UNICAST_ADDR,
4439 MULTICAST_ADDR,
4440 ANYCAST_ADDR,
4443 /* called with rcu_read_lock() */
4444 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4445 struct netlink_callback *cb, enum addr_type_t type,
4446 int s_ip_idx, int *p_ip_idx)
4448 struct ifmcaddr6 *ifmca;
4449 struct ifacaddr6 *ifaca;
4450 int err = 1;
4451 int ip_idx = *p_ip_idx;
4453 read_lock_bh(&idev->lock);
4454 switch (type) {
4455 case UNICAST_ADDR: {
4456 struct inet6_ifaddr *ifa;
4458 /* unicast address incl. temp addr */
4459 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4460 if (++ip_idx < s_ip_idx)
4461 continue;
4462 err = inet6_fill_ifaddr(skb, ifa,
4463 NETLINK_CB(cb->skb).portid,
4464 cb->nlh->nlmsg_seq,
4465 RTM_NEWADDR,
4466 NLM_F_MULTI);
4467 if (err < 0)
4468 break;
4469 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4471 break;
4473 case MULTICAST_ADDR:
4474 /* multicast address */
4475 for (ifmca = idev->mc_list; ifmca;
4476 ifmca = ifmca->next, ip_idx++) {
4477 if (ip_idx < s_ip_idx)
4478 continue;
4479 err = inet6_fill_ifmcaddr(skb, ifmca,
4480 NETLINK_CB(cb->skb).portid,
4481 cb->nlh->nlmsg_seq,
4482 RTM_GETMULTICAST,
4483 NLM_F_MULTI);
4484 if (err < 0)
4485 break;
4487 break;
4488 case ANYCAST_ADDR:
4489 /* anycast address */
4490 for (ifaca = idev->ac_list; ifaca;
4491 ifaca = ifaca->aca_next, ip_idx++) {
4492 if (ip_idx < s_ip_idx)
4493 continue;
4494 err = inet6_fill_ifacaddr(skb, ifaca,
4495 NETLINK_CB(cb->skb).portid,
4496 cb->nlh->nlmsg_seq,
4497 RTM_GETANYCAST,
4498 NLM_F_MULTI);
4499 if (err < 0)
4500 break;
4502 break;
4503 default:
4504 break;
4506 read_unlock_bh(&idev->lock);
4507 *p_ip_idx = ip_idx;
4508 return err;
4511 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4512 enum addr_type_t type)
4514 struct net *net = sock_net(skb->sk);
4515 int h, s_h;
4516 int idx, ip_idx;
4517 int s_idx, s_ip_idx;
4518 struct net_device *dev;
4519 struct inet6_dev *idev;
4520 struct hlist_head *head;
4522 s_h = cb->args[0];
4523 s_idx = idx = cb->args[1];
4524 s_ip_idx = ip_idx = cb->args[2];
4526 rcu_read_lock();
4527 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4528 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4529 idx = 0;
4530 head = &net->dev_index_head[h];
4531 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4532 if (idx < s_idx)
4533 goto cont;
4534 if (h > s_h || idx > s_idx)
4535 s_ip_idx = 0;
4536 ip_idx = 0;
4537 idev = __in6_dev_get(dev);
4538 if (!idev)
4539 goto cont;
4541 if (in6_dump_addrs(idev, skb, cb, type,
4542 s_ip_idx, &ip_idx) < 0)
4543 goto done;
4544 cont:
4545 idx++;
4548 done:
4549 rcu_read_unlock();
4550 cb->args[0] = h;
4551 cb->args[1] = idx;
4552 cb->args[2] = ip_idx;
4554 return skb->len;
4557 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4559 enum addr_type_t type = UNICAST_ADDR;
4561 return inet6_dump_addr(skb, cb, type);
4564 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4566 enum addr_type_t type = MULTICAST_ADDR;
4568 return inet6_dump_addr(skb, cb, type);
4572 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4574 enum addr_type_t type = ANYCAST_ADDR;
4576 return inet6_dump_addr(skb, cb, type);
4579 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4581 struct net *net = sock_net(in_skb->sk);
4582 struct ifaddrmsg *ifm;
4583 struct nlattr *tb[IFA_MAX+1];
4584 struct in6_addr *addr = NULL, *peer;
4585 struct net_device *dev = NULL;
4586 struct inet6_ifaddr *ifa;
4587 struct sk_buff *skb;
4588 int err;
4590 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4591 if (err < 0)
4592 goto errout;
4594 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4595 if (!addr) {
4596 err = -EINVAL;
4597 goto errout;
4600 ifm = nlmsg_data(nlh);
4601 if (ifm->ifa_index)
4602 dev = __dev_get_by_index(net, ifm->ifa_index);
4604 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4605 if (!ifa) {
4606 err = -EADDRNOTAVAIL;
4607 goto errout;
4610 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4611 if (!skb) {
4612 err = -ENOBUFS;
4613 goto errout_ifa;
4616 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4617 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4618 if (err < 0) {
4619 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4620 WARN_ON(err == -EMSGSIZE);
4621 kfree_skb(skb);
4622 goto errout_ifa;
4624 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4625 errout_ifa:
4626 in6_ifa_put(ifa);
4627 errout:
4628 return err;
4631 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4633 struct sk_buff *skb;
4634 struct net *net = dev_net(ifa->idev->dev);
4635 int err = -ENOBUFS;
4637 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4638 if (!skb)
4639 goto errout;
4641 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4642 if (err < 0) {
4643 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4644 WARN_ON(err == -EMSGSIZE);
4645 kfree_skb(skb);
4646 goto errout;
4648 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4649 return;
4650 errout:
4651 if (err < 0)
4652 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4655 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4656 __s32 *array, int bytes)
4658 BUG_ON(bytes < (DEVCONF_MAX * 4));
4660 memset(array, 0, bytes);
4661 array[DEVCONF_FORWARDING] = cnf->forwarding;
4662 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4663 array[DEVCONF_MTU6] = cnf->mtu6;
4664 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4665 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4666 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4667 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4668 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4669 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4670 jiffies_to_msecs(cnf->rtr_solicit_interval);
4671 array[DEVCONF_RTR_SOLICIT_DELAY] =
4672 jiffies_to_msecs(cnf->rtr_solicit_delay);
4673 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4674 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4675 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4676 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4677 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4678 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4679 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4680 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4681 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4682 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4683 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4684 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4685 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4686 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4687 #ifdef CONFIG_IPV6_ROUTER_PREF
4688 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4689 array[DEVCONF_RTR_PROBE_INTERVAL] =
4690 jiffies_to_msecs(cnf->rtr_probe_interval);
4691 #ifdef CONFIG_IPV6_ROUTE_INFO
4692 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4693 #endif
4694 #endif
4695 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4696 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4697 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4698 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4699 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4700 #endif
4701 #ifdef CONFIG_IPV6_MROUTE
4702 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4703 #endif
4704 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4705 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4706 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4707 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4708 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4709 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4710 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4711 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4712 /* we omit DEVCONF_STABLE_SECRET for now */
4713 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4716 static inline size_t inet6_ifla6_size(void)
4718 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4719 + nla_total_size(sizeof(struct ifla_cacheinfo))
4720 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4721 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4722 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4723 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4726 static inline size_t inet6_if_nlmsg_size(void)
4728 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4729 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4730 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4731 + nla_total_size(4) /* IFLA_MTU */
4732 + nla_total_size(4) /* IFLA_LINK */
4733 + nla_total_size(1) /* IFLA_OPERSTATE */
4734 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4737 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4738 int items, int bytes)
4740 int i;
4741 int pad = bytes - sizeof(u64) * items;
4742 BUG_ON(pad < 0);
4744 /* Use put_unaligned() because stats may not be aligned for u64. */
4745 put_unaligned(items, &stats[0]);
4746 for (i = 1; i < items; i++)
4747 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4749 memset(&stats[items], 0, pad);
4752 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4753 int bytes, size_t syncpoff)
4755 int i, c;
4756 u64 buff[IPSTATS_MIB_MAX];
4757 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4759 BUG_ON(pad < 0);
4761 memset(buff, 0, sizeof(buff));
4762 buff[0] = IPSTATS_MIB_MAX;
4764 for_each_possible_cpu(c) {
4765 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4766 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4769 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4770 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4773 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4774 int bytes)
4776 switch (attrtype) {
4777 case IFLA_INET6_STATS:
4778 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4779 offsetof(struct ipstats_mib, syncp));
4780 break;
4781 case IFLA_INET6_ICMP6STATS:
4782 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4783 break;
4787 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4788 u32 ext_filter_mask)
4790 struct nlattr *nla;
4791 struct ifla_cacheinfo ci;
4793 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4794 goto nla_put_failure;
4795 ci.max_reasm_len = IPV6_MAXPLEN;
4796 ci.tstamp = cstamp_delta(idev->tstamp);
4797 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4798 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4799 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4800 goto nla_put_failure;
4801 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4802 if (!nla)
4803 goto nla_put_failure;
4804 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4806 /* XXX - MC not implemented */
4808 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4809 return 0;
4811 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4812 if (!nla)
4813 goto nla_put_failure;
4814 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4816 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4817 if (!nla)
4818 goto nla_put_failure;
4819 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4821 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4822 if (!nla)
4823 goto nla_put_failure;
4825 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4826 goto nla_put_failure;
4828 read_lock_bh(&idev->lock);
4829 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4830 read_unlock_bh(&idev->lock);
4832 return 0;
4834 nla_put_failure:
4835 return -EMSGSIZE;
4838 static size_t inet6_get_link_af_size(const struct net_device *dev,
4839 u32 ext_filter_mask)
4841 if (!__in6_dev_get(dev))
4842 return 0;
4844 return inet6_ifla6_size();
4847 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4848 u32 ext_filter_mask)
4850 struct inet6_dev *idev = __in6_dev_get(dev);
4852 if (!idev)
4853 return -ENODATA;
4855 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4856 return -EMSGSIZE;
4858 return 0;
4861 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4863 struct inet6_ifaddr *ifp;
4864 struct net_device *dev = idev->dev;
4865 bool update_rs = false;
4866 struct in6_addr ll_addr;
4868 ASSERT_RTNL();
4870 if (!token)
4871 return -EINVAL;
4872 if (ipv6_addr_any(token))
4873 return -EINVAL;
4874 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4875 return -EINVAL;
4876 if (!ipv6_accept_ra(idev))
4877 return -EINVAL;
4878 if (idev->cnf.rtr_solicits <= 0)
4879 return -EINVAL;
4881 write_lock_bh(&idev->lock);
4883 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4884 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4886 write_unlock_bh(&idev->lock);
4888 if (!idev->dead && (idev->if_flags & IF_READY) &&
4889 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4890 IFA_F_OPTIMISTIC)) {
4892 /* If we're not ready, then normal ifup will take care
4893 * of this. Otherwise, we need to request our rs here.
4895 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4896 update_rs = true;
4899 write_lock_bh(&idev->lock);
4901 if (update_rs) {
4902 idev->if_flags |= IF_RS_SENT;
4903 idev->rs_probes = 1;
4904 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4907 /* Well, that's kinda nasty ... */
4908 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4909 spin_lock(&ifp->lock);
4910 if (ifp->tokenized) {
4911 ifp->valid_lft = 0;
4912 ifp->prefered_lft = 0;
4914 spin_unlock(&ifp->lock);
4917 write_unlock_bh(&idev->lock);
4918 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4919 addrconf_verify_rtnl();
4920 return 0;
4923 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4924 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
4925 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
4928 static int inet6_validate_link_af(const struct net_device *dev,
4929 const struct nlattr *nla)
4931 struct nlattr *tb[IFLA_INET6_MAX + 1];
4933 if (dev && !__in6_dev_get(dev))
4934 return -EAFNOSUPPORT;
4936 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4939 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4941 int err = -EINVAL;
4942 struct inet6_dev *idev = __in6_dev_get(dev);
4943 struct nlattr *tb[IFLA_INET6_MAX + 1];
4945 if (!idev)
4946 return -EAFNOSUPPORT;
4948 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4949 BUG();
4951 if (tb[IFLA_INET6_TOKEN]) {
4952 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4953 if (err)
4954 return err;
4957 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4958 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4960 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4961 mode != IN6_ADDR_GEN_MODE_NONE &&
4962 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4963 mode != IN6_ADDR_GEN_MODE_RANDOM)
4964 return -EINVAL;
4966 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4967 !idev->cnf.stable_secret.initialized &&
4968 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4969 return -EINVAL;
4971 idev->addr_gen_mode = mode;
4972 err = 0;
4975 return err;
4978 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4979 u32 portid, u32 seq, int event, unsigned int flags)
4981 struct net_device *dev = idev->dev;
4982 struct ifinfomsg *hdr;
4983 struct nlmsghdr *nlh;
4984 void *protoinfo;
4986 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4987 if (!nlh)
4988 return -EMSGSIZE;
4990 hdr = nlmsg_data(nlh);
4991 hdr->ifi_family = AF_INET6;
4992 hdr->__ifi_pad = 0;
4993 hdr->ifi_type = dev->type;
4994 hdr->ifi_index = dev->ifindex;
4995 hdr->ifi_flags = dev_get_flags(dev);
4996 hdr->ifi_change = 0;
4998 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4999 (dev->addr_len &&
5000 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5001 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5002 (dev->ifindex != dev_get_iflink(dev) &&
5003 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5004 nla_put_u8(skb, IFLA_OPERSTATE,
5005 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5006 goto nla_put_failure;
5007 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5008 if (!protoinfo)
5009 goto nla_put_failure;
5011 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5012 goto nla_put_failure;
5014 nla_nest_end(skb, protoinfo);
5015 nlmsg_end(skb, nlh);
5016 return 0;
5018 nla_put_failure:
5019 nlmsg_cancel(skb, nlh);
5020 return -EMSGSIZE;
5023 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5025 struct net *net = sock_net(skb->sk);
5026 int h, s_h;
5027 int idx = 0, s_idx;
5028 struct net_device *dev;
5029 struct inet6_dev *idev;
5030 struct hlist_head *head;
5032 s_h = cb->args[0];
5033 s_idx = cb->args[1];
5035 rcu_read_lock();
5036 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5037 idx = 0;
5038 head = &net->dev_index_head[h];
5039 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5040 if (idx < s_idx)
5041 goto cont;
5042 idev = __in6_dev_get(dev);
5043 if (!idev)
5044 goto cont;
5045 if (inet6_fill_ifinfo(skb, idev,
5046 NETLINK_CB(cb->skb).portid,
5047 cb->nlh->nlmsg_seq,
5048 RTM_NEWLINK, NLM_F_MULTI) < 0)
5049 goto out;
5050 cont:
5051 idx++;
5054 out:
5055 rcu_read_unlock();
5056 cb->args[1] = idx;
5057 cb->args[0] = h;
5059 return skb->len;
5062 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5064 struct sk_buff *skb;
5065 struct net *net = dev_net(idev->dev);
5066 int err = -ENOBUFS;
5068 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5069 if (!skb)
5070 goto errout;
5072 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5073 if (err < 0) {
5074 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5075 WARN_ON(err == -EMSGSIZE);
5076 kfree_skb(skb);
5077 goto errout;
5079 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5080 return;
5081 errout:
5082 if (err < 0)
5083 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5086 static inline size_t inet6_prefix_nlmsg_size(void)
5088 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5089 + nla_total_size(sizeof(struct in6_addr))
5090 + nla_total_size(sizeof(struct prefix_cacheinfo));
5093 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5094 struct prefix_info *pinfo, u32 portid, u32 seq,
5095 int event, unsigned int flags)
5097 struct prefixmsg *pmsg;
5098 struct nlmsghdr *nlh;
5099 struct prefix_cacheinfo ci;
5101 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5102 if (!nlh)
5103 return -EMSGSIZE;
5105 pmsg = nlmsg_data(nlh);
5106 pmsg->prefix_family = AF_INET6;
5107 pmsg->prefix_pad1 = 0;
5108 pmsg->prefix_pad2 = 0;
5109 pmsg->prefix_ifindex = idev->dev->ifindex;
5110 pmsg->prefix_len = pinfo->prefix_len;
5111 pmsg->prefix_type = pinfo->type;
5112 pmsg->prefix_pad3 = 0;
5113 pmsg->prefix_flags = 0;
5114 if (pinfo->onlink)
5115 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5116 if (pinfo->autoconf)
5117 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5119 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5120 goto nla_put_failure;
5121 ci.preferred_time = ntohl(pinfo->prefered);
5122 ci.valid_time = ntohl(pinfo->valid);
5123 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5124 goto nla_put_failure;
5125 nlmsg_end(skb, nlh);
5126 return 0;
5128 nla_put_failure:
5129 nlmsg_cancel(skb, nlh);
5130 return -EMSGSIZE;
5133 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5134 struct prefix_info *pinfo)
5136 struct sk_buff *skb;
5137 struct net *net = dev_net(idev->dev);
5138 int err = -ENOBUFS;
5140 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5141 if (!skb)
5142 goto errout;
5144 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5145 if (err < 0) {
5146 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5147 WARN_ON(err == -EMSGSIZE);
5148 kfree_skb(skb);
5149 goto errout;
5151 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5152 return;
5153 errout:
5154 if (err < 0)
5155 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5160 struct net *net = dev_net(ifp->idev->dev);
5162 if (event)
5163 ASSERT_RTNL();
5165 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5167 switch (event) {
5168 case RTM_NEWADDR:
5170 * If the address was optimistic
5171 * we inserted the route at the start of
5172 * our DAD process, so we don't need
5173 * to do it again
5175 if (!(ifp->rt->rt6i_node))
5176 ip6_ins_rt(ifp->rt);
5177 if (ifp->idev->cnf.forwarding)
5178 addrconf_join_anycast(ifp);
5179 if (!ipv6_addr_any(&ifp->peer_addr))
5180 addrconf_prefix_route(&ifp->peer_addr, 128,
5181 ifp->idev->dev, 0, 0);
5182 break;
5183 case RTM_DELADDR:
5184 if (ifp->idev->cnf.forwarding)
5185 addrconf_leave_anycast(ifp);
5186 addrconf_leave_solict(ifp->idev, &ifp->addr);
5187 if (!ipv6_addr_any(&ifp->peer_addr)) {
5188 struct rt6_info *rt;
5190 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5191 ifp->idev->dev, 0, 0);
5192 if (rt)
5193 ip6_del_rt(rt);
5195 dst_hold(&ifp->rt->dst);
5197 ip6_del_rt(ifp->rt);
5199 rt_genid_bump_ipv6(net);
5200 break;
5202 atomic_inc(&net->ipv6.dev_addr_genid);
5205 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5207 rcu_read_lock_bh();
5208 if (likely(ifp->idev->dead == 0))
5209 __ipv6_ifa_notify(event, ifp);
5210 rcu_read_unlock_bh();
5213 #ifdef CONFIG_SYSCTL
5215 static
5216 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5217 void __user *buffer, size_t *lenp, loff_t *ppos)
5219 int *valp = ctl->data;
5220 int val = *valp;
5221 loff_t pos = *ppos;
5222 struct ctl_table lctl;
5223 int ret;
5226 * ctl->data points to idev->cnf.forwarding, we should
5227 * not modify it until we get the rtnl lock.
5229 lctl = *ctl;
5230 lctl.data = &val;
5232 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5234 if (write)
5235 ret = addrconf_fixup_forwarding(ctl, valp, val);
5236 if (ret)
5237 *ppos = pos;
5238 return ret;
5241 static
5242 int addrconf_sysctl_hop_limit(struct ctl_table *ctl, int write,
5243 void __user *buffer, size_t *lenp, loff_t *ppos)
5245 struct ctl_table lctl;
5246 int min_hl = 1, max_hl = 255;
5248 lctl = *ctl;
5249 lctl.extra1 = &min_hl;
5250 lctl.extra2 = &max_hl;
5252 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5255 static
5256 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5257 void __user *buffer, size_t *lenp, loff_t *ppos)
5259 struct inet6_dev *idev = ctl->extra1;
5260 int min_mtu = IPV6_MIN_MTU;
5261 struct ctl_table lctl;
5263 lctl = *ctl;
5264 lctl.extra1 = &min_mtu;
5265 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5267 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5270 static void dev_disable_change(struct inet6_dev *idev)
5272 struct netdev_notifier_info info;
5274 if (!idev || !idev->dev)
5275 return;
5277 netdev_notifier_info_init(&info, idev->dev);
5278 if (idev->cnf.disable_ipv6)
5279 addrconf_notify(NULL, NETDEV_DOWN, &info);
5280 else
5281 addrconf_notify(NULL, NETDEV_UP, &info);
5284 static void addrconf_disable_change(struct net *net, __s32 newf)
5286 struct net_device *dev;
5287 struct inet6_dev *idev;
5289 rcu_read_lock();
5290 for_each_netdev_rcu(net, dev) {
5291 idev = __in6_dev_get(dev);
5292 if (idev) {
5293 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5294 idev->cnf.disable_ipv6 = newf;
5295 if (changed)
5296 dev_disable_change(idev);
5299 rcu_read_unlock();
5302 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5304 struct net *net;
5305 int old;
5307 if (!rtnl_trylock())
5308 return restart_syscall();
5310 net = (struct net *)table->extra2;
5311 old = *p;
5312 *p = newf;
5314 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5315 rtnl_unlock();
5316 return 0;
5319 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5320 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5321 addrconf_disable_change(net, newf);
5322 } else if ((!newf) ^ (!old))
5323 dev_disable_change((struct inet6_dev *)table->extra1);
5325 rtnl_unlock();
5326 return 0;
5329 static
5330 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5331 void __user *buffer, size_t *lenp, loff_t *ppos)
5333 int *valp = ctl->data;
5334 int val = *valp;
5335 loff_t pos = *ppos;
5336 struct ctl_table lctl;
5337 int ret;
5340 * ctl->data points to idev->cnf.disable_ipv6, we should
5341 * not modify it until we get the rtnl lock.
5343 lctl = *ctl;
5344 lctl.data = &val;
5346 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5348 if (write)
5349 ret = addrconf_disable_ipv6(ctl, valp, val);
5350 if (ret)
5351 *ppos = pos;
5352 return ret;
5355 static
5356 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5357 void __user *buffer, size_t *lenp, loff_t *ppos)
5359 int *valp = ctl->data;
5360 int ret;
5361 int old, new;
5363 old = *valp;
5364 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5365 new = *valp;
5367 if (write && old != new) {
5368 struct net *net = ctl->extra2;
5370 if (!rtnl_trylock())
5371 return restart_syscall();
5373 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5374 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5375 NETCONFA_IFINDEX_DEFAULT,
5376 net->ipv6.devconf_dflt);
5377 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5378 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5379 NETCONFA_IFINDEX_ALL,
5380 net->ipv6.devconf_all);
5381 else {
5382 struct inet6_dev *idev = ctl->extra1;
5384 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5385 idev->dev->ifindex,
5386 &idev->cnf);
5388 rtnl_unlock();
5391 return ret;
5394 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5395 void __user *buffer, size_t *lenp,
5396 loff_t *ppos)
5398 int err;
5399 struct in6_addr addr;
5400 char str[IPV6_MAX_STRLEN];
5401 struct ctl_table lctl = *ctl;
5402 struct net *net = ctl->extra2;
5403 struct ipv6_stable_secret *secret = ctl->data;
5405 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5406 return -EIO;
5408 lctl.maxlen = IPV6_MAX_STRLEN;
5409 lctl.data = str;
5411 if (!rtnl_trylock())
5412 return restart_syscall();
5414 if (!write && !secret->initialized) {
5415 err = -EIO;
5416 goto out;
5419 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5420 if (err >= sizeof(str)) {
5421 err = -EIO;
5422 goto out;
5425 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5426 if (err || !write)
5427 goto out;
5429 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5430 err = -EIO;
5431 goto out;
5434 secret->initialized = true;
5435 secret->secret = addr;
5437 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5438 struct net_device *dev;
5440 for_each_netdev(net, dev) {
5441 struct inet6_dev *idev = __in6_dev_get(dev);
5443 if (idev) {
5444 idev->addr_gen_mode =
5445 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5448 } else {
5449 struct inet6_dev *idev = ctl->extra1;
5451 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5454 out:
5455 rtnl_unlock();
5457 return err;
5460 static
5461 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5462 int write,
5463 void __user *buffer,
5464 size_t *lenp,
5465 loff_t *ppos)
5467 int *valp = ctl->data;
5468 int val = *valp;
5469 loff_t pos = *ppos;
5470 struct ctl_table lctl;
5471 int ret;
5473 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5474 * we should not modify it until we get the rtnl lock.
5476 lctl = *ctl;
5477 lctl.data = &val;
5479 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5481 if (write)
5482 ret = addrconf_fixup_linkdown(ctl, valp, val);
5483 if (ret)
5484 *ppos = pos;
5485 return ret;
5488 static struct addrconf_sysctl_table
5490 struct ctl_table_header *sysctl_header;
5491 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5492 } addrconf_sysctl __read_mostly = {
5493 .sysctl_header = NULL,
5494 .addrconf_vars = {
5496 .procname = "forwarding",
5497 .data = &ipv6_devconf.forwarding,
5498 .maxlen = sizeof(int),
5499 .mode = 0644,
5500 .proc_handler = addrconf_sysctl_forward,
5503 .procname = "hop_limit",
5504 .data = &ipv6_devconf.hop_limit,
5505 .maxlen = sizeof(int),
5506 .mode = 0644,
5507 .proc_handler = addrconf_sysctl_hop_limit,
5510 .procname = "mtu",
5511 .data = &ipv6_devconf.mtu6,
5512 .maxlen = sizeof(int),
5513 .mode = 0644,
5514 .proc_handler = addrconf_sysctl_mtu,
5517 .procname = "accept_ra",
5518 .data = &ipv6_devconf.accept_ra,
5519 .maxlen = sizeof(int),
5520 .mode = 0644,
5521 .proc_handler = proc_dointvec,
5524 .procname = "accept_redirects",
5525 .data = &ipv6_devconf.accept_redirects,
5526 .maxlen = sizeof(int),
5527 .mode = 0644,
5528 .proc_handler = proc_dointvec,
5531 .procname = "autoconf",
5532 .data = &ipv6_devconf.autoconf,
5533 .maxlen = sizeof(int),
5534 .mode = 0644,
5535 .proc_handler = proc_dointvec,
5538 .procname = "dad_transmits",
5539 .data = &ipv6_devconf.dad_transmits,
5540 .maxlen = sizeof(int),
5541 .mode = 0644,
5542 .proc_handler = proc_dointvec,
5545 .procname = "router_solicitations",
5546 .data = &ipv6_devconf.rtr_solicits,
5547 .maxlen = sizeof(int),
5548 .mode = 0644,
5549 .proc_handler = proc_dointvec,
5552 .procname = "router_solicitation_interval",
5553 .data = &ipv6_devconf.rtr_solicit_interval,
5554 .maxlen = sizeof(int),
5555 .mode = 0644,
5556 .proc_handler = proc_dointvec_jiffies,
5559 .procname = "router_solicitation_delay",
5560 .data = &ipv6_devconf.rtr_solicit_delay,
5561 .maxlen = sizeof(int),
5562 .mode = 0644,
5563 .proc_handler = proc_dointvec_jiffies,
5566 .procname = "force_mld_version",
5567 .data = &ipv6_devconf.force_mld_version,
5568 .maxlen = sizeof(int),
5569 .mode = 0644,
5570 .proc_handler = proc_dointvec,
5573 .procname = "mldv1_unsolicited_report_interval",
5574 .data =
5575 &ipv6_devconf.mldv1_unsolicited_report_interval,
5576 .maxlen = sizeof(int),
5577 .mode = 0644,
5578 .proc_handler = proc_dointvec_ms_jiffies,
5581 .procname = "mldv2_unsolicited_report_interval",
5582 .data =
5583 &ipv6_devconf.mldv2_unsolicited_report_interval,
5584 .maxlen = sizeof(int),
5585 .mode = 0644,
5586 .proc_handler = proc_dointvec_ms_jiffies,
5589 .procname = "use_tempaddr",
5590 .data = &ipv6_devconf.use_tempaddr,
5591 .maxlen = sizeof(int),
5592 .mode = 0644,
5593 .proc_handler = proc_dointvec,
5596 .procname = "temp_valid_lft",
5597 .data = &ipv6_devconf.temp_valid_lft,
5598 .maxlen = sizeof(int),
5599 .mode = 0644,
5600 .proc_handler = proc_dointvec,
5603 .procname = "temp_prefered_lft",
5604 .data = &ipv6_devconf.temp_prefered_lft,
5605 .maxlen = sizeof(int),
5606 .mode = 0644,
5607 .proc_handler = proc_dointvec,
5610 .procname = "regen_max_retry",
5611 .data = &ipv6_devconf.regen_max_retry,
5612 .maxlen = sizeof(int),
5613 .mode = 0644,
5614 .proc_handler = proc_dointvec,
5617 .procname = "max_desync_factor",
5618 .data = &ipv6_devconf.max_desync_factor,
5619 .maxlen = sizeof(int),
5620 .mode = 0644,
5621 .proc_handler = proc_dointvec,
5624 .procname = "max_addresses",
5625 .data = &ipv6_devconf.max_addresses,
5626 .maxlen = sizeof(int),
5627 .mode = 0644,
5628 .proc_handler = proc_dointvec,
5631 .procname = "accept_ra_defrtr",
5632 .data = &ipv6_devconf.accept_ra_defrtr,
5633 .maxlen = sizeof(int),
5634 .mode = 0644,
5635 .proc_handler = proc_dointvec,
5638 .procname = "accept_ra_min_hop_limit",
5639 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5640 .maxlen = sizeof(int),
5641 .mode = 0644,
5642 .proc_handler = proc_dointvec,
5645 .procname = "accept_ra_pinfo",
5646 .data = &ipv6_devconf.accept_ra_pinfo,
5647 .maxlen = sizeof(int),
5648 .mode = 0644,
5649 .proc_handler = proc_dointvec,
5651 #ifdef CONFIG_IPV6_ROUTER_PREF
5653 .procname = "accept_ra_rtr_pref",
5654 .data = &ipv6_devconf.accept_ra_rtr_pref,
5655 .maxlen = sizeof(int),
5656 .mode = 0644,
5657 .proc_handler = proc_dointvec,
5660 .procname = "router_probe_interval",
5661 .data = &ipv6_devconf.rtr_probe_interval,
5662 .maxlen = sizeof(int),
5663 .mode = 0644,
5664 .proc_handler = proc_dointvec_jiffies,
5666 #ifdef CONFIG_IPV6_ROUTE_INFO
5668 .procname = "accept_ra_rt_info_max_plen",
5669 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5670 .maxlen = sizeof(int),
5671 .mode = 0644,
5672 .proc_handler = proc_dointvec,
5674 #endif
5675 #endif
5677 .procname = "proxy_ndp",
5678 .data = &ipv6_devconf.proxy_ndp,
5679 .maxlen = sizeof(int),
5680 .mode = 0644,
5681 .proc_handler = addrconf_sysctl_proxy_ndp,
5684 .procname = "accept_source_route",
5685 .data = &ipv6_devconf.accept_source_route,
5686 .maxlen = sizeof(int),
5687 .mode = 0644,
5688 .proc_handler = proc_dointvec,
5690 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5692 .procname = "optimistic_dad",
5693 .data = &ipv6_devconf.optimistic_dad,
5694 .maxlen = sizeof(int),
5695 .mode = 0644,
5696 .proc_handler = proc_dointvec,
5700 .procname = "use_optimistic",
5701 .data = &ipv6_devconf.use_optimistic,
5702 .maxlen = sizeof(int),
5703 .mode = 0644,
5704 .proc_handler = proc_dointvec,
5707 #endif
5708 #ifdef CONFIG_IPV6_MROUTE
5710 .procname = "mc_forwarding",
5711 .data = &ipv6_devconf.mc_forwarding,
5712 .maxlen = sizeof(int),
5713 .mode = 0444,
5714 .proc_handler = proc_dointvec,
5716 #endif
5718 .procname = "disable_ipv6",
5719 .data = &ipv6_devconf.disable_ipv6,
5720 .maxlen = sizeof(int),
5721 .mode = 0644,
5722 .proc_handler = addrconf_sysctl_disable,
5725 .procname = "accept_dad",
5726 .data = &ipv6_devconf.accept_dad,
5727 .maxlen = sizeof(int),
5728 .mode = 0644,
5729 .proc_handler = proc_dointvec,
5732 .procname = "force_tllao",
5733 .data = &ipv6_devconf.force_tllao,
5734 .maxlen = sizeof(int),
5735 .mode = 0644,
5736 .proc_handler = proc_dointvec
5739 .procname = "ndisc_notify",
5740 .data = &ipv6_devconf.ndisc_notify,
5741 .maxlen = sizeof(int),
5742 .mode = 0644,
5743 .proc_handler = proc_dointvec
5746 .procname = "suppress_frag_ndisc",
5747 .data = &ipv6_devconf.suppress_frag_ndisc,
5748 .maxlen = sizeof(int),
5749 .mode = 0644,
5750 .proc_handler = proc_dointvec
5753 .procname = "accept_ra_from_local",
5754 .data = &ipv6_devconf.accept_ra_from_local,
5755 .maxlen = sizeof(int),
5756 .mode = 0644,
5757 .proc_handler = proc_dointvec,
5760 .procname = "accept_ra_mtu",
5761 .data = &ipv6_devconf.accept_ra_mtu,
5762 .maxlen = sizeof(int),
5763 .mode = 0644,
5764 .proc_handler = proc_dointvec,
5767 .procname = "stable_secret",
5768 .data = &ipv6_devconf.stable_secret,
5769 .maxlen = IPV6_MAX_STRLEN,
5770 .mode = 0600,
5771 .proc_handler = addrconf_sysctl_stable_secret,
5774 .procname = "use_oif_addrs_only",
5775 .data = &ipv6_devconf.use_oif_addrs_only,
5776 .maxlen = sizeof(int),
5777 .mode = 0644,
5778 .proc_handler = proc_dointvec,
5781 .procname = "ignore_routes_with_linkdown",
5782 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5783 .maxlen = sizeof(int),
5784 .mode = 0644,
5785 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5788 /* sentinel */
5793 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5794 struct inet6_dev *idev, struct ipv6_devconf *p)
5796 int i;
5797 struct addrconf_sysctl_table *t;
5798 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5800 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5801 if (!t)
5802 goto out;
5804 for (i = 0; t->addrconf_vars[i].data; i++) {
5805 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5806 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5807 t->addrconf_vars[i].extra2 = net;
5810 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5812 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5813 if (!t->sysctl_header)
5814 goto free;
5816 p->sysctl = t;
5817 return 0;
5819 free:
5820 kfree(t);
5821 out:
5822 return -ENOBUFS;
5825 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5827 struct addrconf_sysctl_table *t;
5829 if (!p->sysctl)
5830 return;
5832 t = p->sysctl;
5833 p->sysctl = NULL;
5834 unregister_net_sysctl_table(t->sysctl_header);
5835 kfree(t);
5838 static int addrconf_sysctl_register(struct inet6_dev *idev)
5840 int err;
5842 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5843 return -EINVAL;
5845 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5846 &ndisc_ifinfo_sysctl_change);
5847 if (err)
5848 return err;
5849 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5850 idev, &idev->cnf);
5851 if (err)
5852 neigh_sysctl_unregister(idev->nd_parms);
5854 return err;
5857 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5859 __addrconf_sysctl_unregister(&idev->cnf);
5860 neigh_sysctl_unregister(idev->nd_parms);
5864 #endif
5866 static int __net_init addrconf_init_net(struct net *net)
5868 int err = -ENOMEM;
5869 struct ipv6_devconf *all, *dflt;
5871 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5872 if (!all)
5873 goto err_alloc_all;
5875 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5876 if (!dflt)
5877 goto err_alloc_dflt;
5879 /* these will be inherited by all namespaces */
5880 dflt->autoconf = ipv6_defaults.autoconf;
5881 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5883 dflt->stable_secret.initialized = false;
5884 all->stable_secret.initialized = false;
5886 net->ipv6.devconf_all = all;
5887 net->ipv6.devconf_dflt = dflt;
5889 #ifdef CONFIG_SYSCTL
5890 err = __addrconf_sysctl_register(net, "all", NULL, all);
5891 if (err < 0)
5892 goto err_reg_all;
5894 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5895 if (err < 0)
5896 goto err_reg_dflt;
5897 #endif
5898 return 0;
5900 #ifdef CONFIG_SYSCTL
5901 err_reg_dflt:
5902 __addrconf_sysctl_unregister(all);
5903 err_reg_all:
5904 kfree(dflt);
5905 #endif
5906 err_alloc_dflt:
5907 kfree(all);
5908 err_alloc_all:
5909 return err;
5912 static void __net_exit addrconf_exit_net(struct net *net)
5914 #ifdef CONFIG_SYSCTL
5915 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5916 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5917 #endif
5918 kfree(net->ipv6.devconf_dflt);
5919 kfree(net->ipv6.devconf_all);
5922 static struct pernet_operations addrconf_ops = {
5923 .init = addrconf_init_net,
5924 .exit = addrconf_exit_net,
5927 static struct rtnl_af_ops inet6_ops __read_mostly = {
5928 .family = AF_INET6,
5929 .fill_link_af = inet6_fill_link_af,
5930 .get_link_af_size = inet6_get_link_af_size,
5931 .validate_link_af = inet6_validate_link_af,
5932 .set_link_af = inet6_set_link_af,
5936 * Init / cleanup code
5939 int __init addrconf_init(void)
5941 struct inet6_dev *idev;
5942 int i, err;
5944 err = ipv6_addr_label_init();
5945 if (err < 0) {
5946 pr_crit("%s: cannot initialize default policy table: %d\n",
5947 __func__, err);
5948 goto out;
5951 err = register_pernet_subsys(&addrconf_ops);
5952 if (err < 0)
5953 goto out_addrlabel;
5955 addrconf_wq = create_workqueue("ipv6_addrconf");
5956 if (!addrconf_wq) {
5957 err = -ENOMEM;
5958 goto out_nowq;
5961 /* The addrconf netdev notifier requires that loopback_dev
5962 * has it's ipv6 private information allocated and setup
5963 * before it can bring up and give link-local addresses
5964 * to other devices which are up.
5966 * Unfortunately, loopback_dev is not necessarily the first
5967 * entry in the global dev_base list of net devices. In fact,
5968 * it is likely to be the very last entry on that list.
5969 * So this causes the notifier registry below to try and
5970 * give link-local addresses to all devices besides loopback_dev
5971 * first, then loopback_dev, which cases all the non-loopback_dev
5972 * devices to fail to get a link-local address.
5974 * So, as a temporary fix, allocate the ipv6 structure for
5975 * loopback_dev first by hand.
5976 * Longer term, all of the dependencies ipv6 has upon the loopback
5977 * device and it being up should be removed.
5979 rtnl_lock();
5980 idev = ipv6_add_dev(init_net.loopback_dev);
5981 rtnl_unlock();
5982 if (IS_ERR(idev)) {
5983 err = PTR_ERR(idev);
5984 goto errlo;
5987 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5988 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5990 register_netdevice_notifier(&ipv6_dev_notf);
5992 addrconf_verify();
5994 rtnl_af_register(&inet6_ops);
5996 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5997 NULL);
5998 if (err < 0)
5999 goto errout;
6001 /* Only the first call to __rtnl_register can fail */
6002 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6003 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6004 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6005 inet6_dump_ifaddr, NULL);
6006 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6007 inet6_dump_ifmcaddr, NULL);
6008 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6009 inet6_dump_ifacaddr, NULL);
6010 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6011 inet6_netconf_dump_devconf, NULL);
6013 ipv6_addr_label_rtnl_register();
6015 return 0;
6016 errout:
6017 rtnl_af_unregister(&inet6_ops);
6018 unregister_netdevice_notifier(&ipv6_dev_notf);
6019 errlo:
6020 destroy_workqueue(addrconf_wq);
6021 out_nowq:
6022 unregister_pernet_subsys(&addrconf_ops);
6023 out_addrlabel:
6024 ipv6_addr_label_cleanup();
6025 out:
6026 return err;
6029 void addrconf_cleanup(void)
6031 struct net_device *dev;
6032 int i;
6034 unregister_netdevice_notifier(&ipv6_dev_notf);
6035 unregister_pernet_subsys(&addrconf_ops);
6036 ipv6_addr_label_cleanup();
6038 rtnl_lock();
6040 __rtnl_af_unregister(&inet6_ops);
6042 /* clean dev list */
6043 for_each_netdev(&init_net, dev) {
6044 if (__in6_dev_get(dev) == NULL)
6045 continue;
6046 addrconf_ifdown(dev, 1);
6048 addrconf_ifdown(init_net.loopback_dev, 2);
6051 * Check hash table.
6053 spin_lock_bh(&addrconf_hash_lock);
6054 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6055 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6056 spin_unlock_bh(&addrconf_hash_lock);
6057 cancel_delayed_work(&addr_chk_work);
6058 rtnl_unlock();
6060 destroy_workqueue(addrconf_wq);