ALSA: usb-audio: Fix an out-of-bound read in create_composite_quirks
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob4dde1e0e7d37968b060ca68a1dd161ced0f6cbe9
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
69 #include <net/net_namespace.h>
70 #include <net/sock.h>
71 #include <net/snmp.h>
73 #include <net/af_ieee802154.h>
74 #include <net/firewire.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/tcp.h>
81 #include <net/ip.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
99 #if ACONF_DEBUG >= 3
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101 #else
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #endif
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 #define IPV6_MAX_STRLEN \
108 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
110 static inline u32 cstamp_delta(unsigned long cstamp)
112 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
115 #ifdef CONFIG_SYSCTL
116 static int addrconf_sysctl_register(struct inet6_dev *idev);
117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118 #else
119 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
121 return 0;
124 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
127 #endif
129 static void __ipv6_regen_rndid(struct inet6_dev *idev);
130 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
131 static void ipv6_regen_rndid(unsigned long data);
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
135 static int ipv6_generate_stable_address(struct in6_addr *addr,
136 u8 dad_count,
137 const struct inet6_dev *idev);
140 * Configured unicast address hash table
142 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
143 static DEFINE_SPINLOCK(addrconf_hash_lock);
145 static void addrconf_verify(void);
146 static void addrconf_verify_rtnl(void);
147 static void addrconf_verify_work(struct work_struct *);
149 static struct workqueue_struct *addrconf_wq;
150 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
152 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
153 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
155 static void addrconf_type_change(struct net_device *dev,
156 unsigned long event);
157 static int addrconf_ifdown(struct net_device *dev, int how);
159 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
160 int plen,
161 const struct net_device *dev,
162 u32 flags, u32 noflags);
164 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
165 static void addrconf_dad_work(struct work_struct *w);
166 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
167 static void addrconf_dad_run(struct inet6_dev *idev);
168 static void addrconf_rs_timer(unsigned long data);
169 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
170 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
172 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
173 struct prefix_info *pinfo);
174 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
175 struct net_device *dev);
177 static struct ipv6_devconf ipv6_devconf __read_mostly = {
178 .forwarding = 0,
179 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
180 .mtu6 = IPV6_MIN_MTU,
181 .accept_ra = 1,
182 .accept_redirects = 1,
183 .autoconf = 1,
184 .force_mld_version = 0,
185 .mldv1_unsolicited_report_interval = 10 * HZ,
186 .mldv2_unsolicited_report_interval = HZ,
187 .dad_transmits = 1,
188 .rtr_solicits = MAX_RTR_SOLICITATIONS,
189 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
190 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
191 .use_tempaddr = 0,
192 .temp_valid_lft = TEMP_VALID_LIFETIME,
193 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
194 .regen_max_retry = REGEN_MAX_RETRY,
195 .max_desync_factor = MAX_DESYNC_FACTOR,
196 .max_addresses = IPV6_MAX_ADDRESSES,
197 .accept_ra_defrtr = 1,
198 .accept_ra_from_local = 0,
199 .accept_ra_min_hop_limit= 1,
200 .accept_ra_pinfo = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202 .accept_ra_rtr_pref = 1,
203 .rtr_probe_interval = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205 .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208 .proxy_ndp = 0,
209 .accept_source_route = 0, /* we do not accept RH0 by default. */
210 .disable_ipv6 = 0,
211 .accept_dad = 1,
212 .suppress_frag_ndisc = 1,
213 .accept_ra_mtu = 1,
214 .stable_secret = {
215 .initialized = false,
217 .use_oif_addrs_only = 0,
218 .ignore_routes_with_linkdown = 0,
221 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
222 .forwarding = 0,
223 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
224 .mtu6 = IPV6_MIN_MTU,
225 .accept_ra = 1,
226 .accept_redirects = 1,
227 .autoconf = 1,
228 .force_mld_version = 0,
229 .mldv1_unsolicited_report_interval = 10 * HZ,
230 .mldv2_unsolicited_report_interval = HZ,
231 .dad_transmits = 1,
232 .rtr_solicits = MAX_RTR_SOLICITATIONS,
233 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
234 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
235 .use_tempaddr = 0,
236 .temp_valid_lft = TEMP_VALID_LIFETIME,
237 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
238 .regen_max_retry = REGEN_MAX_RETRY,
239 .max_desync_factor = MAX_DESYNC_FACTOR,
240 .max_addresses = IPV6_MAX_ADDRESSES,
241 .accept_ra_defrtr = 1,
242 .accept_ra_from_local = 0,
243 .accept_ra_min_hop_limit= 1,
244 .accept_ra_pinfo = 1,
245 #ifdef CONFIG_IPV6_ROUTER_PREF
246 .accept_ra_rtr_pref = 1,
247 .rtr_probe_interval = 60 * HZ,
248 #ifdef CONFIG_IPV6_ROUTE_INFO
249 .accept_ra_rt_info_max_plen = 0,
250 #endif
251 #endif
252 .proxy_ndp = 0,
253 .accept_source_route = 0, /* we do not accept RH0 by default. */
254 .disable_ipv6 = 0,
255 .accept_dad = 1,
256 .suppress_frag_ndisc = 1,
257 .accept_ra_mtu = 1,
258 .stable_secret = {
259 .initialized = false,
261 .use_oif_addrs_only = 0,
262 .ignore_routes_with_linkdown = 0,
265 /* Check if a valid qdisc is available */
266 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
268 return !qdisc_tx_is_noop(dev);
271 static void addrconf_del_rs_timer(struct inet6_dev *idev)
273 if (del_timer(&idev->rs_timer))
274 __in6_dev_put(idev);
277 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
279 if (cancel_delayed_work(&ifp->dad_work))
280 __in6_ifa_put(ifp);
283 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
284 unsigned long when)
286 if (!timer_pending(&idev->rs_timer))
287 in6_dev_hold(idev);
288 mod_timer(&idev->rs_timer, jiffies + when);
291 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
292 unsigned long delay)
294 in6_ifa_hold(ifp);
295 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
296 in6_ifa_put(ifp);
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;
961 /* No need to add the TENTATIVE flag for addresses with NODAD */
962 if (!(flags & IFA_F_NODAD))
963 ifa->flags |= IFA_F_TENTATIVE;
964 ifa->valid_lft = valid_lft;
965 ifa->prefered_lft = prefered_lft;
966 ifa->cstamp = ifa->tstamp = jiffies;
967 ifa->tokenized = false;
969 ifa->rt = rt;
971 ifa->idev = idev;
972 in6_dev_hold(idev);
973 /* For caller */
974 in6_ifa_hold(ifa);
976 /* Add to big hash table */
977 hash = inet6_addr_hash(addr);
979 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
980 spin_unlock(&addrconf_hash_lock);
982 write_lock(&idev->lock);
983 /* Add to inet6_dev unicast addr list. */
984 ipv6_link_dev_addr(idev, ifa);
986 if (ifa->flags&IFA_F_TEMPORARY) {
987 list_add(&ifa->tmp_list, &idev->tempaddr_list);
988 in6_ifa_hold(ifa);
991 in6_ifa_hold(ifa);
992 write_unlock(&idev->lock);
993 out2:
994 rcu_read_unlock_bh();
996 if (likely(err == 0))
997 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
998 else {
999 kfree(ifa);
1000 ifa = ERR_PTR(err);
1003 return ifa;
1004 out:
1005 spin_unlock(&addrconf_hash_lock);
1006 goto out2;
1009 enum cleanup_prefix_rt_t {
1010 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1011 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1012 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1016 * Check, whether the prefix for ifp would still need a prefix route
1017 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1018 * constants.
1020 * 1) we don't purge prefix if address was not permanent.
1021 * prefix is managed by its own lifetime.
1022 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1023 * 3) if there are no addresses, delete prefix.
1024 * 4) if there are still other permanent address(es),
1025 * corresponding prefix is still permanent.
1026 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1027 * don't purge the prefix, assume user space is managing it.
1028 * 6) otherwise, update prefix lifetime to the
1029 * longest valid lifetime among the corresponding
1030 * addresses on the device.
1031 * Note: subsequent RA will update lifetime.
1033 static enum cleanup_prefix_rt_t
1034 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1036 struct inet6_ifaddr *ifa;
1037 struct inet6_dev *idev = ifp->idev;
1038 unsigned long lifetime;
1039 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1041 *expires = jiffies;
1043 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1044 if (ifa == ifp)
1045 continue;
1046 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1047 ifp->prefix_len))
1048 continue;
1049 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1050 return CLEANUP_PREFIX_RT_NOP;
1052 action = CLEANUP_PREFIX_RT_EXPIRE;
1054 spin_lock(&ifa->lock);
1056 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1058 * Note: Because this address is
1059 * not permanent, lifetime <
1060 * LONG_MAX / HZ here.
1062 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1063 *expires = ifa->tstamp + lifetime * HZ;
1064 spin_unlock(&ifa->lock);
1067 return action;
1070 static void
1071 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1073 struct rt6_info *rt;
1075 rt = addrconf_get_prefix_route(&ifp->addr,
1076 ifp->prefix_len,
1077 ifp->idev->dev,
1078 0, RTF_GATEWAY | RTF_DEFAULT);
1079 if (rt) {
1080 if (del_rt)
1081 ip6_del_rt(rt);
1082 else {
1083 if (!(rt->rt6i_flags & RTF_EXPIRES))
1084 rt6_set_expires(rt, expires);
1085 ip6_rt_put(rt);
1091 /* This function wants to get referenced ifp and releases it before return */
1093 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1095 int state;
1096 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1097 unsigned long expires;
1099 ASSERT_RTNL();
1101 spin_lock_bh(&ifp->lock);
1102 state = ifp->state;
1103 ifp->state = INET6_IFADDR_STATE_DEAD;
1104 spin_unlock_bh(&ifp->lock);
1106 if (state == INET6_IFADDR_STATE_DEAD)
1107 goto out;
1109 spin_lock_bh(&addrconf_hash_lock);
1110 hlist_del_init_rcu(&ifp->addr_lst);
1111 spin_unlock_bh(&addrconf_hash_lock);
1113 write_lock_bh(&ifp->idev->lock);
1115 if (ifp->flags&IFA_F_TEMPORARY) {
1116 list_del(&ifp->tmp_list);
1117 if (ifp->ifpub) {
1118 in6_ifa_put(ifp->ifpub);
1119 ifp->ifpub = NULL;
1121 __in6_ifa_put(ifp);
1124 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1125 action = check_cleanup_prefix_route(ifp, &expires);
1127 list_del_init(&ifp->if_list);
1128 __in6_ifa_put(ifp);
1130 write_unlock_bh(&ifp->idev->lock);
1132 addrconf_del_dad_work(ifp);
1134 ipv6_ifa_notify(RTM_DELADDR, ifp);
1136 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1138 if (action != CLEANUP_PREFIX_RT_NOP) {
1139 cleanup_prefix_route(ifp, expires,
1140 action == CLEANUP_PREFIX_RT_DEL);
1143 /* clean up prefsrc entries */
1144 rt6_remove_prefsrc(ifp);
1145 out:
1146 in6_ifa_put(ifp);
1149 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1151 struct inet6_dev *idev = ifp->idev;
1152 struct in6_addr addr, *tmpaddr;
1153 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1154 unsigned long regen_advance;
1155 int tmp_plen;
1156 int ret = 0;
1157 u32 addr_flags;
1158 unsigned long now = jiffies;
1160 write_lock_bh(&idev->lock);
1161 if (ift) {
1162 spin_lock_bh(&ift->lock);
1163 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1164 spin_unlock_bh(&ift->lock);
1165 tmpaddr = &addr;
1166 } else {
1167 tmpaddr = NULL;
1169 retry:
1170 in6_dev_hold(idev);
1171 if (idev->cnf.use_tempaddr <= 0) {
1172 write_unlock_bh(&idev->lock);
1173 pr_info("%s: use_tempaddr is disabled\n", __func__);
1174 in6_dev_put(idev);
1175 ret = -1;
1176 goto out;
1178 spin_lock_bh(&ifp->lock);
1179 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1180 idev->cnf.use_tempaddr = -1; /*XXX*/
1181 spin_unlock_bh(&ifp->lock);
1182 write_unlock_bh(&idev->lock);
1183 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1184 __func__);
1185 in6_dev_put(idev);
1186 ret = -1;
1187 goto out;
1189 in6_ifa_hold(ifp);
1190 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1191 __ipv6_try_regen_rndid(idev, tmpaddr);
1192 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1193 age = (now - ifp->tstamp) / HZ;
1194 tmp_valid_lft = min_t(__u32,
1195 ifp->valid_lft,
1196 idev->cnf.temp_valid_lft + age);
1197 tmp_prefered_lft = min_t(__u32,
1198 ifp->prefered_lft,
1199 idev->cnf.temp_prefered_lft + age -
1200 idev->cnf.max_desync_factor);
1201 tmp_plen = ifp->prefix_len;
1202 tmp_tstamp = ifp->tstamp;
1203 spin_unlock_bh(&ifp->lock);
1205 regen_advance = idev->cnf.regen_max_retry *
1206 idev->cnf.dad_transmits *
1207 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1208 write_unlock_bh(&idev->lock);
1210 /* A temporary address is created only if this calculated Preferred
1211 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1212 * an implementation must not create a temporary address with a zero
1213 * Preferred Lifetime.
1214 * Use age calculation as in addrconf_verify to avoid unnecessary
1215 * temporary addresses being generated.
1217 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1218 if (tmp_prefered_lft <= regen_advance + age) {
1219 in6_ifa_put(ifp);
1220 in6_dev_put(idev);
1221 ret = -1;
1222 goto out;
1225 addr_flags = IFA_F_TEMPORARY;
1226 /* set in addrconf_prefix_rcv() */
1227 if (ifp->flags & IFA_F_OPTIMISTIC)
1228 addr_flags |= IFA_F_OPTIMISTIC;
1230 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1231 ipv6_addr_scope(&addr), addr_flags,
1232 tmp_valid_lft, tmp_prefered_lft);
1233 if (IS_ERR(ift)) {
1234 in6_ifa_put(ifp);
1235 in6_dev_put(idev);
1236 pr_info("%s: retry temporary address regeneration\n", __func__);
1237 tmpaddr = &addr;
1238 write_lock_bh(&idev->lock);
1239 goto retry;
1242 spin_lock_bh(&ift->lock);
1243 ift->ifpub = ifp;
1244 ift->cstamp = now;
1245 ift->tstamp = tmp_tstamp;
1246 spin_unlock_bh(&ift->lock);
1248 addrconf_dad_start(ift);
1249 in6_ifa_put(ift);
1250 in6_dev_put(idev);
1251 out:
1252 return ret;
1256 * Choose an appropriate source address (RFC3484)
1258 enum {
1259 IPV6_SADDR_RULE_INIT = 0,
1260 IPV6_SADDR_RULE_LOCAL,
1261 IPV6_SADDR_RULE_SCOPE,
1262 IPV6_SADDR_RULE_PREFERRED,
1263 #ifdef CONFIG_IPV6_MIP6
1264 IPV6_SADDR_RULE_HOA,
1265 #endif
1266 IPV6_SADDR_RULE_OIF,
1267 IPV6_SADDR_RULE_LABEL,
1268 IPV6_SADDR_RULE_PRIVACY,
1269 IPV6_SADDR_RULE_ORCHID,
1270 IPV6_SADDR_RULE_PREFIX,
1271 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1272 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1273 #endif
1274 IPV6_SADDR_RULE_MAX
1277 struct ipv6_saddr_score {
1278 int rule;
1279 int addr_type;
1280 struct inet6_ifaddr *ifa;
1281 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1282 int scopedist;
1283 int matchlen;
1286 struct ipv6_saddr_dst {
1287 const struct in6_addr *addr;
1288 int ifindex;
1289 int scope;
1290 int label;
1291 unsigned int prefs;
1294 static inline int ipv6_saddr_preferred(int type)
1296 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1297 return 1;
1298 return 0;
1301 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1303 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1304 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1305 #else
1306 return false;
1307 #endif
1310 static int ipv6_get_saddr_eval(struct net *net,
1311 struct ipv6_saddr_score *score,
1312 struct ipv6_saddr_dst *dst,
1313 int i)
1315 int ret;
1317 if (i <= score->rule) {
1318 switch (i) {
1319 case IPV6_SADDR_RULE_SCOPE:
1320 ret = score->scopedist;
1321 break;
1322 case IPV6_SADDR_RULE_PREFIX:
1323 ret = score->matchlen;
1324 break;
1325 default:
1326 ret = !!test_bit(i, score->scorebits);
1328 goto out;
1331 switch (i) {
1332 case IPV6_SADDR_RULE_INIT:
1333 /* Rule 0: remember if hiscore is not ready yet */
1334 ret = !!score->ifa;
1335 break;
1336 case IPV6_SADDR_RULE_LOCAL:
1337 /* Rule 1: Prefer same address */
1338 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1339 break;
1340 case IPV6_SADDR_RULE_SCOPE:
1341 /* Rule 2: Prefer appropriate scope
1343 * ret
1345 * -1 | d 15
1346 * ---+--+-+---> scope
1348 * | d is scope of the destination.
1349 * B-d | \
1350 * | \ <- smaller scope is better if
1351 * B-15 | \ if scope is enough for destination.
1352 * | ret = B - scope (-1 <= scope >= d <= 15).
1353 * d-C-1 | /
1354 * |/ <- greater is better
1355 * -C / if scope is not enough for destination.
1356 * /| ret = scope - C (-1 <= d < scope <= 15).
1358 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1359 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1360 * Assume B = 0 and we get C > 29.
1362 ret = __ipv6_addr_src_scope(score->addr_type);
1363 if (ret >= dst->scope)
1364 ret = -ret;
1365 else
1366 ret -= 128; /* 30 is enough */
1367 score->scopedist = ret;
1368 break;
1369 case IPV6_SADDR_RULE_PREFERRED:
1371 /* Rule 3: Avoid deprecated and optimistic addresses */
1372 u8 avoid = IFA_F_DEPRECATED;
1374 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1375 avoid |= IFA_F_OPTIMISTIC;
1376 ret = ipv6_saddr_preferred(score->addr_type) ||
1377 !(score->ifa->flags & avoid);
1378 break;
1380 #ifdef CONFIG_IPV6_MIP6
1381 case IPV6_SADDR_RULE_HOA:
1383 /* Rule 4: Prefer home address */
1384 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1385 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1386 break;
1388 #endif
1389 case IPV6_SADDR_RULE_OIF:
1390 /* Rule 5: Prefer outgoing interface */
1391 ret = (!dst->ifindex ||
1392 dst->ifindex == score->ifa->idev->dev->ifindex);
1393 break;
1394 case IPV6_SADDR_RULE_LABEL:
1395 /* Rule 6: Prefer matching label */
1396 ret = ipv6_addr_label(net,
1397 &score->ifa->addr, score->addr_type,
1398 score->ifa->idev->dev->ifindex) == dst->label;
1399 break;
1400 case IPV6_SADDR_RULE_PRIVACY:
1402 /* Rule 7: Prefer public address
1403 * Note: prefer temporary address if use_tempaddr >= 2
1405 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1406 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1407 score->ifa->idev->cnf.use_tempaddr >= 2;
1408 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1409 break;
1411 case IPV6_SADDR_RULE_ORCHID:
1412 /* Rule 8-: Prefer ORCHID vs ORCHID or
1413 * non-ORCHID vs non-ORCHID
1415 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1416 ipv6_addr_orchid(dst->addr));
1417 break;
1418 case IPV6_SADDR_RULE_PREFIX:
1419 /* Rule 8: Use longest matching prefix */
1420 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1421 if (ret > score->ifa->prefix_len)
1422 ret = score->ifa->prefix_len;
1423 score->matchlen = ret;
1424 break;
1425 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1426 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1427 /* Optimistic addresses still have lower precedence than other
1428 * preferred addresses.
1430 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1431 break;
1432 #endif
1433 default:
1434 ret = 0;
1437 if (ret)
1438 __set_bit(i, score->scorebits);
1439 score->rule = i;
1440 out:
1441 return ret;
1444 static int __ipv6_dev_get_saddr(struct net *net,
1445 struct ipv6_saddr_dst *dst,
1446 struct inet6_dev *idev,
1447 struct ipv6_saddr_score *scores,
1448 int hiscore_idx)
1450 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1452 read_lock_bh(&idev->lock);
1453 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1454 int i;
1457 * - Tentative Address (RFC2462 section 5.4)
1458 * - A tentative address is not considered
1459 * "assigned to an interface" in the traditional
1460 * sense, unless it is also flagged as optimistic.
1461 * - Candidate Source Address (section 4)
1462 * - In any case, anycast addresses, multicast
1463 * addresses, and the unspecified address MUST
1464 * NOT be included in a candidate set.
1466 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1467 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1468 continue;
1470 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1472 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1473 score->addr_type & IPV6_ADDR_MULTICAST)) {
1474 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1475 idev->dev->name);
1476 continue;
1479 score->rule = -1;
1480 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1482 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1483 int minihiscore, miniscore;
1485 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1486 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1488 if (minihiscore > miniscore) {
1489 if (i == IPV6_SADDR_RULE_SCOPE &&
1490 score->scopedist > 0) {
1492 * special case:
1493 * each remaining entry
1494 * has too small (not enough)
1495 * scope, because ifa entries
1496 * are sorted by their scope
1497 * values.
1499 goto out;
1501 break;
1502 } else if (minihiscore < miniscore) {
1503 if (hiscore->ifa)
1504 in6_ifa_put(hiscore->ifa);
1506 in6_ifa_hold(score->ifa);
1508 swap(hiscore, score);
1509 hiscore_idx = 1 - hiscore_idx;
1511 /* restore our iterator */
1512 score->ifa = hiscore->ifa;
1514 break;
1518 out:
1519 read_unlock_bh(&idev->lock);
1520 return hiscore_idx;
1523 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1524 const struct in6_addr *daddr, unsigned int prefs,
1525 struct in6_addr *saddr)
1527 struct ipv6_saddr_score scores[2], *hiscore;
1528 struct ipv6_saddr_dst dst;
1529 struct inet6_dev *idev;
1530 struct net_device *dev;
1531 int dst_type;
1532 bool use_oif_addr = false;
1533 int hiscore_idx = 0;
1535 dst_type = __ipv6_addr_type(daddr);
1536 dst.addr = daddr;
1537 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1538 dst.scope = __ipv6_addr_src_scope(dst_type);
1539 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1540 dst.prefs = prefs;
1542 scores[hiscore_idx].rule = -1;
1543 scores[hiscore_idx].ifa = NULL;
1545 rcu_read_lock();
1547 /* Candidate Source Address (section 4)
1548 * - multicast and link-local destination address,
1549 * the set of candidate source address MUST only
1550 * include addresses assigned to interfaces
1551 * belonging to the same link as the outgoing
1552 * interface.
1553 * (- For site-local destination addresses, the
1554 * set of candidate source addresses MUST only
1555 * include addresses assigned to interfaces
1556 * belonging to the same site as the outgoing
1557 * interface.)
1558 * - "It is RECOMMENDED that the candidate source addresses
1559 * be the set of unicast addresses assigned to the
1560 * interface that will be used to send to the destination
1561 * (the 'outgoing' interface)." (RFC 6724)
1563 if (dst_dev) {
1564 idev = __in6_dev_get(dst_dev);
1565 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1566 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1567 (idev && idev->cnf.use_oif_addrs_only)) {
1568 use_oif_addr = true;
1572 if (use_oif_addr) {
1573 if (idev)
1574 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1575 } else {
1576 for_each_netdev_rcu(net, dev) {
1577 idev = __in6_dev_get(dev);
1578 if (!idev)
1579 continue;
1580 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1583 rcu_read_unlock();
1585 hiscore = &scores[hiscore_idx];
1586 if (!hiscore->ifa)
1587 return -EADDRNOTAVAIL;
1589 *saddr = hiscore->ifa->addr;
1590 in6_ifa_put(hiscore->ifa);
1591 return 0;
1593 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1595 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1596 u32 banned_flags)
1598 struct inet6_ifaddr *ifp;
1599 int err = -EADDRNOTAVAIL;
1601 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1602 if (ifp->scope > IFA_LINK)
1603 break;
1604 if (ifp->scope == IFA_LINK &&
1605 !(ifp->flags & banned_flags)) {
1606 *addr = ifp->addr;
1607 err = 0;
1608 break;
1611 return err;
1614 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1615 u32 banned_flags)
1617 struct inet6_dev *idev;
1618 int err = -EADDRNOTAVAIL;
1620 rcu_read_lock();
1621 idev = __in6_dev_get(dev);
1622 if (idev) {
1623 read_lock_bh(&idev->lock);
1624 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1625 read_unlock_bh(&idev->lock);
1627 rcu_read_unlock();
1628 return err;
1631 static int ipv6_count_addresses(struct inet6_dev *idev)
1633 int cnt = 0;
1634 struct inet6_ifaddr *ifp;
1636 read_lock_bh(&idev->lock);
1637 list_for_each_entry(ifp, &idev->addr_list, if_list)
1638 cnt++;
1639 read_unlock_bh(&idev->lock);
1640 return cnt;
1643 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1644 const struct net_device *dev, int strict)
1646 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1648 EXPORT_SYMBOL(ipv6_chk_addr);
1650 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1651 const struct net_device *dev, int strict,
1652 u32 banned_flags)
1654 struct inet6_ifaddr *ifp;
1655 unsigned int hash = inet6_addr_hash(addr);
1656 u32 ifp_flags;
1658 rcu_read_lock_bh();
1659 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1660 if (!net_eq(dev_net(ifp->idev->dev), net))
1661 continue;
1662 /* Decouple optimistic from tentative for evaluation here.
1663 * Ban optimistic addresses explicitly, when required.
1665 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1666 ? (ifp->flags&~IFA_F_TENTATIVE)
1667 : ifp->flags;
1668 if (ipv6_addr_equal(&ifp->addr, addr) &&
1669 !(ifp_flags&banned_flags) &&
1670 (!dev || ifp->idev->dev == dev ||
1671 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1672 rcu_read_unlock_bh();
1673 return 1;
1677 rcu_read_unlock_bh();
1678 return 0;
1680 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1682 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1683 struct net_device *dev)
1685 unsigned int hash = inet6_addr_hash(addr);
1686 struct inet6_ifaddr *ifp;
1688 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1689 if (!net_eq(dev_net(ifp->idev->dev), net))
1690 continue;
1691 if (ipv6_addr_equal(&ifp->addr, addr)) {
1692 if (!dev || ifp->idev->dev == dev)
1693 return true;
1696 return false;
1699 /* Compares an address/prefix_len with addresses on device @dev.
1700 * If one is found it returns true.
1702 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1703 const unsigned int prefix_len, struct net_device *dev)
1705 struct inet6_dev *idev;
1706 struct inet6_ifaddr *ifa;
1707 bool ret = false;
1709 rcu_read_lock();
1710 idev = __in6_dev_get(dev);
1711 if (idev) {
1712 read_lock_bh(&idev->lock);
1713 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1714 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1715 if (ret)
1716 break;
1718 read_unlock_bh(&idev->lock);
1720 rcu_read_unlock();
1722 return ret;
1724 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1726 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1728 struct inet6_dev *idev;
1729 struct inet6_ifaddr *ifa;
1730 int onlink;
1732 onlink = 0;
1733 rcu_read_lock();
1734 idev = __in6_dev_get(dev);
1735 if (idev) {
1736 read_lock_bh(&idev->lock);
1737 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1738 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1739 ifa->prefix_len);
1740 if (onlink)
1741 break;
1743 read_unlock_bh(&idev->lock);
1745 rcu_read_unlock();
1746 return onlink;
1748 EXPORT_SYMBOL(ipv6_chk_prefix);
1750 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1751 struct net_device *dev, int strict)
1753 struct inet6_ifaddr *ifp, *result = NULL;
1754 unsigned int hash = inet6_addr_hash(addr);
1756 rcu_read_lock_bh();
1757 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1758 if (!net_eq(dev_net(ifp->idev->dev), net))
1759 continue;
1760 if (ipv6_addr_equal(&ifp->addr, addr)) {
1761 if (!dev || ifp->idev->dev == dev ||
1762 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1763 result = ifp;
1764 in6_ifa_hold(ifp);
1765 break;
1769 rcu_read_unlock_bh();
1771 return result;
1774 /* Gets referenced address, destroys ifaddr */
1776 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1778 if (ifp->flags&IFA_F_TEMPORARY) {
1779 struct inet6_ifaddr *ifpub;
1780 spin_lock_bh(&ifp->lock);
1781 ifpub = ifp->ifpub;
1782 if (ifpub) {
1783 in6_ifa_hold(ifpub);
1784 spin_unlock_bh(&ifp->lock);
1785 ipv6_create_tempaddr(ifpub, ifp);
1786 in6_ifa_put(ifpub);
1787 } else {
1788 spin_unlock_bh(&ifp->lock);
1790 ipv6_del_addr(ifp);
1791 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1792 spin_lock_bh(&ifp->lock);
1793 addrconf_del_dad_work(ifp);
1794 ifp->flags |= IFA_F_TENTATIVE;
1795 if (dad_failed)
1796 ifp->flags |= IFA_F_DADFAILED;
1797 spin_unlock_bh(&ifp->lock);
1798 if (dad_failed)
1799 ipv6_ifa_notify(0, ifp);
1800 in6_ifa_put(ifp);
1801 } else {
1802 ipv6_del_addr(ifp);
1806 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1808 int err = -ENOENT;
1810 spin_lock_bh(&ifp->lock);
1811 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1812 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1813 err = 0;
1815 spin_unlock_bh(&ifp->lock);
1817 return err;
1820 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1822 struct in6_addr addr;
1823 struct inet6_dev *idev = ifp->idev;
1824 struct net *net = dev_net(ifp->idev->dev);
1826 if (addrconf_dad_end(ifp)) {
1827 in6_ifa_put(ifp);
1828 return;
1831 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1832 ifp->idev->dev->name, &ifp->addr);
1834 spin_lock_bh(&ifp->lock);
1836 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1837 int scope = ifp->scope;
1838 u32 flags = ifp->flags;
1839 struct in6_addr new_addr;
1840 struct inet6_ifaddr *ifp2;
1841 u32 valid_lft, preferred_lft;
1842 int pfxlen = ifp->prefix_len;
1843 int retries = ifp->stable_privacy_retry + 1;
1845 if (retries > net->ipv6.sysctl.idgen_retries) {
1846 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1847 ifp->idev->dev->name);
1848 goto errdad;
1851 new_addr = ifp->addr;
1852 if (ipv6_generate_stable_address(&new_addr, retries,
1853 idev))
1854 goto errdad;
1856 valid_lft = ifp->valid_lft;
1857 preferred_lft = ifp->prefered_lft;
1859 spin_unlock_bh(&ifp->lock);
1861 if (idev->cnf.max_addresses &&
1862 ipv6_count_addresses(idev) >=
1863 idev->cnf.max_addresses)
1864 goto lock_errdad;
1866 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1867 ifp->idev->dev->name);
1869 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1870 scope, flags, valid_lft,
1871 preferred_lft);
1872 if (IS_ERR(ifp2))
1873 goto lock_errdad;
1875 spin_lock_bh(&ifp2->lock);
1876 ifp2->stable_privacy_retry = retries;
1877 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1878 spin_unlock_bh(&ifp2->lock);
1880 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1881 in6_ifa_put(ifp2);
1882 lock_errdad:
1883 spin_lock_bh(&ifp->lock);
1884 } else if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1885 addr.s6_addr32[0] = htonl(0xfe800000);
1886 addr.s6_addr32[1] = 0;
1888 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1889 ipv6_addr_equal(&ifp->addr, &addr)) {
1890 /* DAD failed for link-local based on MAC address */
1891 idev->cnf.disable_ipv6 = 1;
1893 pr_info("%s: IPv6 being disabled!\n",
1894 ifp->idev->dev->name);
1898 errdad:
1899 /* transition from _POSTDAD to _ERRDAD */
1900 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1901 spin_unlock_bh(&ifp->lock);
1903 addrconf_mod_dad_work(ifp, 0);
1904 in6_ifa_put(ifp);
1907 /* Join to solicited addr multicast group.
1908 * caller must hold RTNL */
1909 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1911 struct in6_addr maddr;
1913 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1914 return;
1916 addrconf_addr_solict_mult(addr, &maddr);
1917 ipv6_dev_mc_inc(dev, &maddr);
1920 /* caller must hold RTNL */
1921 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1923 struct in6_addr maddr;
1925 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1926 return;
1928 addrconf_addr_solict_mult(addr, &maddr);
1929 __ipv6_dev_mc_dec(idev, &maddr);
1932 /* caller must hold RTNL */
1933 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1935 struct in6_addr addr;
1937 if (ifp->prefix_len >= 127) /* RFC 6164 */
1938 return;
1939 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1940 if (ipv6_addr_any(&addr))
1941 return;
1942 __ipv6_dev_ac_inc(ifp->idev, &addr);
1945 /* caller must hold RTNL */
1946 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1948 struct in6_addr addr;
1950 if (ifp->prefix_len >= 127) /* RFC 6164 */
1951 return;
1952 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1953 if (ipv6_addr_any(&addr))
1954 return;
1955 __ipv6_dev_ac_dec(ifp->idev, &addr);
1958 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1960 if (dev->addr_len != IEEE802154_ADDR_LEN)
1961 return -1;
1962 memcpy(eui, dev->dev_addr, 8);
1963 eui[0] ^= 2;
1964 return 0;
1967 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1969 union fwnet_hwaddr *ha;
1971 if (dev->addr_len != FWNET_ALEN)
1972 return -1;
1974 ha = (union fwnet_hwaddr *)dev->dev_addr;
1976 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1977 eui[0] ^= 2;
1978 return 0;
1981 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1983 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1984 if (dev->addr_len != ARCNET_ALEN)
1985 return -1;
1986 memset(eui, 0, 7);
1987 eui[7] = *(u8 *)dev->dev_addr;
1988 return 0;
1991 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1993 if (dev->addr_len != INFINIBAND_ALEN)
1994 return -1;
1995 memcpy(eui, dev->dev_addr + 12, 8);
1996 eui[0] |= 2;
1997 return 0;
2000 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2002 if (addr == 0)
2003 return -1;
2004 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2005 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2006 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2007 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2008 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2009 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2010 eui[1] = 0;
2011 eui[2] = 0x5E;
2012 eui[3] = 0xFE;
2013 memcpy(eui + 4, &addr, 4);
2014 return 0;
2017 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2019 if (dev->priv_flags & IFF_ISATAP)
2020 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2021 return -1;
2024 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2026 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2029 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2031 memcpy(eui, dev->perm_addr, 3);
2032 memcpy(eui + 5, dev->perm_addr + 3, 3);
2033 eui[3] = 0xFF;
2034 eui[4] = 0xFE;
2035 eui[0] ^= 2;
2036 return 0;
2039 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2041 switch (dev->type) {
2042 case ARPHRD_ETHER:
2043 case ARPHRD_FDDI:
2044 return addrconf_ifid_eui48(eui, dev);
2045 case ARPHRD_ARCNET:
2046 return addrconf_ifid_arcnet(eui, dev);
2047 case ARPHRD_INFINIBAND:
2048 return addrconf_ifid_infiniband(eui, dev);
2049 case ARPHRD_SIT:
2050 return addrconf_ifid_sit(eui, dev);
2051 case ARPHRD_IPGRE:
2052 return addrconf_ifid_gre(eui, dev);
2053 case ARPHRD_6LOWPAN:
2054 case ARPHRD_IEEE802154:
2055 return addrconf_ifid_eui64(eui, dev);
2056 case ARPHRD_IEEE1394:
2057 return addrconf_ifid_ieee1394(eui, dev);
2058 case ARPHRD_TUNNEL6:
2059 return addrconf_ifid_ip6tnl(eui, dev);
2061 return -1;
2064 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2066 int err = -1;
2067 struct inet6_ifaddr *ifp;
2069 read_lock_bh(&idev->lock);
2070 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2071 if (ifp->scope > IFA_LINK)
2072 break;
2073 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2074 memcpy(eui, ifp->addr.s6_addr+8, 8);
2075 err = 0;
2076 break;
2079 read_unlock_bh(&idev->lock);
2080 return err;
2083 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2084 static void __ipv6_regen_rndid(struct inet6_dev *idev)
2086 regen:
2087 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2088 idev->rndid[0] &= ~0x02;
2091 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2092 * check if generated address is not inappropriate
2094 * - Reserved subnet anycast (RFC 2526)
2095 * 11111101 11....11 1xxxxxxx
2096 * - ISATAP (RFC4214) 6.1
2097 * 00-00-5E-FE-xx-xx-xx-xx
2098 * - value 0
2099 * - XXX: already assigned to an address on the device
2101 if (idev->rndid[0] == 0xfd &&
2102 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2103 (idev->rndid[7]&0x80))
2104 goto regen;
2105 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2106 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2107 goto regen;
2108 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2109 goto regen;
2113 static void ipv6_regen_rndid(unsigned long data)
2115 struct inet6_dev *idev = (struct inet6_dev *) data;
2116 unsigned long expires;
2118 rcu_read_lock_bh();
2119 write_lock_bh(&idev->lock);
2121 if (idev->dead)
2122 goto out;
2124 __ipv6_regen_rndid(idev);
2126 expires = jiffies +
2127 idev->cnf.temp_prefered_lft * HZ -
2128 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
2129 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
2130 idev->cnf.max_desync_factor * HZ;
2131 if (time_before(expires, jiffies)) {
2132 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
2133 __func__, idev->dev->name);
2134 goto out;
2137 if (!mod_timer(&idev->regen_timer, expires))
2138 in6_dev_hold(idev);
2140 out:
2141 write_unlock_bh(&idev->lock);
2142 rcu_read_unlock_bh();
2143 in6_dev_put(idev);
2146 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2148 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2149 __ipv6_regen_rndid(idev);
2153 * Add prefix route.
2156 static void
2157 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2158 unsigned long expires, u32 flags)
2160 struct fib6_config cfg = {
2161 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2162 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2163 .fc_ifindex = dev->ifindex,
2164 .fc_expires = expires,
2165 .fc_dst_len = plen,
2166 .fc_flags = RTF_UP | flags,
2167 .fc_nlinfo.nl_net = dev_net(dev),
2168 .fc_protocol = RTPROT_KERNEL,
2171 cfg.fc_dst = *pfx;
2173 /* Prevent useless cloning on PtP SIT.
2174 This thing is done here expecting that the whole
2175 class of non-broadcast devices need not cloning.
2177 #if IS_ENABLED(CONFIG_IPV6_SIT)
2178 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2179 cfg.fc_flags |= RTF_NONEXTHOP;
2180 #endif
2182 ip6_route_add(&cfg);
2186 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2187 int plen,
2188 const struct net_device *dev,
2189 u32 flags, u32 noflags)
2191 struct fib6_node *fn;
2192 struct rt6_info *rt = NULL;
2193 struct fib6_table *table;
2194 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2196 table = fib6_get_table(dev_net(dev), tb_id);
2197 if (!table)
2198 return NULL;
2200 read_lock_bh(&table->tb6_lock);
2201 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2202 if (!fn)
2203 goto out;
2205 noflags |= RTF_CACHE;
2206 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2207 if (rt->dst.dev->ifindex != dev->ifindex)
2208 continue;
2209 if ((rt->rt6i_flags & flags) != flags)
2210 continue;
2211 if ((rt->rt6i_flags & noflags) != 0)
2212 continue;
2213 dst_hold(&rt->dst);
2214 break;
2216 out:
2217 read_unlock_bh(&table->tb6_lock);
2218 return rt;
2222 /* Create "default" multicast route to the interface */
2224 static void addrconf_add_mroute(struct net_device *dev)
2226 struct fib6_config cfg = {
2227 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2228 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2229 .fc_ifindex = dev->ifindex,
2230 .fc_dst_len = 8,
2231 .fc_flags = RTF_UP,
2232 .fc_nlinfo.nl_net = dev_net(dev),
2235 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2237 ip6_route_add(&cfg);
2240 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2242 struct inet6_dev *idev;
2244 ASSERT_RTNL();
2246 idev = ipv6_find_idev(dev);
2247 if (!idev)
2248 return ERR_PTR(-ENOBUFS);
2250 if (idev->cnf.disable_ipv6)
2251 return ERR_PTR(-EACCES);
2253 /* Add default multicast route */
2254 if (!(dev->flags & IFF_LOOPBACK))
2255 addrconf_add_mroute(dev);
2257 return idev;
2260 static void manage_tempaddrs(struct inet6_dev *idev,
2261 struct inet6_ifaddr *ifp,
2262 __u32 valid_lft, __u32 prefered_lft,
2263 bool create, unsigned long now)
2265 u32 flags;
2266 struct inet6_ifaddr *ift;
2268 read_lock_bh(&idev->lock);
2269 /* update all temporary addresses in the list */
2270 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2271 int age, max_valid, max_prefered;
2273 if (ifp != ift->ifpub)
2274 continue;
2276 /* RFC 4941 section 3.3:
2277 * If a received option will extend the lifetime of a public
2278 * address, the lifetimes of temporary addresses should
2279 * be extended, subject to the overall constraint that no
2280 * temporary addresses should ever remain "valid" or "preferred"
2281 * for a time longer than (TEMP_VALID_LIFETIME) or
2282 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2284 age = (now - ift->cstamp) / HZ;
2285 max_valid = idev->cnf.temp_valid_lft - age;
2286 if (max_valid < 0)
2287 max_valid = 0;
2289 max_prefered = idev->cnf.temp_prefered_lft -
2290 idev->cnf.max_desync_factor - age;
2291 if (max_prefered < 0)
2292 max_prefered = 0;
2294 if (valid_lft > max_valid)
2295 valid_lft = max_valid;
2297 if (prefered_lft > max_prefered)
2298 prefered_lft = max_prefered;
2300 spin_lock(&ift->lock);
2301 flags = ift->flags;
2302 ift->valid_lft = valid_lft;
2303 ift->prefered_lft = prefered_lft;
2304 ift->tstamp = now;
2305 if (prefered_lft > 0)
2306 ift->flags &= ~IFA_F_DEPRECATED;
2308 spin_unlock(&ift->lock);
2309 if (!(flags&IFA_F_TENTATIVE))
2310 ipv6_ifa_notify(0, ift);
2313 if ((create || list_empty(&idev->tempaddr_list)) &&
2314 idev->cnf.use_tempaddr > 0) {
2315 /* When a new public address is created as described
2316 * in [ADDRCONF], also create a new temporary address.
2317 * Also create a temporary address if it's enabled but
2318 * no temporary address currently exists.
2320 read_unlock_bh(&idev->lock);
2321 ipv6_create_tempaddr(ifp, NULL);
2322 } else {
2323 read_unlock_bh(&idev->lock);
2327 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2329 struct prefix_info *pinfo;
2330 __u32 valid_lft;
2331 __u32 prefered_lft;
2332 int addr_type;
2333 u32 addr_flags = 0;
2334 struct inet6_dev *in6_dev;
2335 struct net *net = dev_net(dev);
2337 pinfo = (struct prefix_info *) opt;
2339 if (len < sizeof(struct prefix_info)) {
2340 ADBG("addrconf: prefix option too short\n");
2341 return;
2345 * Validation checks ([ADDRCONF], page 19)
2348 addr_type = ipv6_addr_type(&pinfo->prefix);
2350 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2351 return;
2353 valid_lft = ntohl(pinfo->valid);
2354 prefered_lft = ntohl(pinfo->prefered);
2356 if (prefered_lft > valid_lft) {
2357 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2358 return;
2361 in6_dev = in6_dev_get(dev);
2363 if (!in6_dev) {
2364 net_dbg_ratelimited("addrconf: device %s not configured\n",
2365 dev->name);
2366 return;
2370 * Two things going on here:
2371 * 1) Add routes for on-link prefixes
2372 * 2) Configure prefixes with the auto flag set
2375 if (pinfo->onlink) {
2376 struct rt6_info *rt;
2377 unsigned long rt_expires;
2379 /* Avoid arithmetic overflow. Really, we could
2380 * save rt_expires in seconds, likely valid_lft,
2381 * but it would require division in fib gc, that it
2382 * not good.
2384 if (HZ > USER_HZ)
2385 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2386 else
2387 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2389 if (addrconf_finite_timeout(rt_expires))
2390 rt_expires *= HZ;
2392 rt = addrconf_get_prefix_route(&pinfo->prefix,
2393 pinfo->prefix_len,
2394 dev,
2395 RTF_ADDRCONF | RTF_PREFIX_RT,
2396 RTF_GATEWAY | RTF_DEFAULT);
2398 if (rt) {
2399 /* Autoconf prefix route */
2400 if (valid_lft == 0) {
2401 ip6_del_rt(rt);
2402 rt = NULL;
2403 } else if (addrconf_finite_timeout(rt_expires)) {
2404 /* not infinity */
2405 rt6_set_expires(rt, jiffies + rt_expires);
2406 } else {
2407 rt6_clean_expires(rt);
2409 } else if (valid_lft) {
2410 clock_t expires = 0;
2411 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2412 if (addrconf_finite_timeout(rt_expires)) {
2413 /* not infinity */
2414 flags |= RTF_EXPIRES;
2415 expires = jiffies_to_clock_t(rt_expires);
2417 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2418 dev, expires, flags);
2420 ip6_rt_put(rt);
2423 /* Try to figure out our local address for this prefix */
2425 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2426 struct inet6_ifaddr *ifp;
2427 struct in6_addr addr;
2428 int create = 0, update_lft = 0;
2429 bool tokenized = false;
2431 if (pinfo->prefix_len == 64) {
2432 memcpy(&addr, &pinfo->prefix, 8);
2434 if (!ipv6_addr_any(&in6_dev->token)) {
2435 read_lock_bh(&in6_dev->lock);
2436 memcpy(addr.s6_addr + 8,
2437 in6_dev->token.s6_addr + 8, 8);
2438 read_unlock_bh(&in6_dev->lock);
2439 tokenized = true;
2440 } else if (in6_dev->addr_gen_mode ==
2441 IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
2442 !ipv6_generate_stable_address(&addr, 0,
2443 in6_dev)) {
2444 addr_flags |= IFA_F_STABLE_PRIVACY;
2445 goto ok;
2446 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2447 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2448 in6_dev_put(in6_dev);
2449 return;
2451 goto ok;
2453 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2454 pinfo->prefix_len);
2455 in6_dev_put(in6_dev);
2456 return;
2460 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2462 if (!ifp && valid_lft) {
2463 int max_addresses = in6_dev->cnf.max_addresses;
2465 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2466 if (in6_dev->cnf.optimistic_dad &&
2467 !net->ipv6.devconf_all->forwarding && sllao)
2468 addr_flags |= IFA_F_OPTIMISTIC;
2469 #endif
2471 /* Do not allow to create too much of autoconfigured
2472 * addresses; this would be too easy way to crash kernel.
2474 if (!max_addresses ||
2475 ipv6_count_addresses(in6_dev) < max_addresses)
2476 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2477 pinfo->prefix_len,
2478 addr_type&IPV6_ADDR_SCOPE_MASK,
2479 addr_flags, valid_lft,
2480 prefered_lft);
2482 if (IS_ERR_OR_NULL(ifp)) {
2483 in6_dev_put(in6_dev);
2484 return;
2487 update_lft = 0;
2488 create = 1;
2489 spin_lock_bh(&ifp->lock);
2490 ifp->flags |= IFA_F_MANAGETEMPADDR;
2491 ifp->cstamp = jiffies;
2492 ifp->tokenized = tokenized;
2493 spin_unlock_bh(&ifp->lock);
2494 addrconf_dad_start(ifp);
2497 if (ifp) {
2498 u32 flags;
2499 unsigned long now;
2500 u32 stored_lft;
2502 /* update lifetime (RFC2462 5.5.3 e) */
2503 spin_lock_bh(&ifp->lock);
2504 now = jiffies;
2505 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2506 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2507 else
2508 stored_lft = 0;
2509 if (!update_lft && !create && stored_lft) {
2510 const u32 minimum_lft = min_t(u32,
2511 stored_lft, MIN_VALID_LIFETIME);
2512 valid_lft = max(valid_lft, minimum_lft);
2514 /* RFC4862 Section 5.5.3e:
2515 * "Note that the preferred lifetime of the
2516 * corresponding address is always reset to
2517 * the Preferred Lifetime in the received
2518 * Prefix Information option, regardless of
2519 * whether the valid lifetime is also reset or
2520 * ignored."
2522 * So we should always update prefered_lft here.
2524 update_lft = 1;
2527 if (update_lft) {
2528 ifp->valid_lft = valid_lft;
2529 ifp->prefered_lft = prefered_lft;
2530 ifp->tstamp = now;
2531 flags = ifp->flags;
2532 ifp->flags &= ~IFA_F_DEPRECATED;
2533 spin_unlock_bh(&ifp->lock);
2535 if (!(flags&IFA_F_TENTATIVE))
2536 ipv6_ifa_notify(0, ifp);
2537 } else
2538 spin_unlock_bh(&ifp->lock);
2540 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2541 create, now);
2543 in6_ifa_put(ifp);
2544 addrconf_verify();
2547 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2548 in6_dev_put(in6_dev);
2552 * Set destination address.
2553 * Special case for SIT interfaces where we create a new "virtual"
2554 * device.
2556 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2558 struct in6_ifreq ireq;
2559 struct net_device *dev;
2560 int err = -EINVAL;
2562 rtnl_lock();
2564 err = -EFAULT;
2565 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2566 goto err_exit;
2568 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2570 err = -ENODEV;
2571 if (!dev)
2572 goto err_exit;
2574 #if IS_ENABLED(CONFIG_IPV6_SIT)
2575 if (dev->type == ARPHRD_SIT) {
2576 const struct net_device_ops *ops = dev->netdev_ops;
2577 struct ifreq ifr;
2578 struct ip_tunnel_parm p;
2580 err = -EADDRNOTAVAIL;
2581 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2582 goto err_exit;
2584 memset(&p, 0, sizeof(p));
2585 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2586 p.iph.saddr = 0;
2587 p.iph.version = 4;
2588 p.iph.ihl = 5;
2589 p.iph.protocol = IPPROTO_IPV6;
2590 p.iph.ttl = 64;
2591 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2593 if (ops->ndo_do_ioctl) {
2594 mm_segment_t oldfs = get_fs();
2596 set_fs(KERNEL_DS);
2597 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2598 set_fs(oldfs);
2599 } else
2600 err = -EOPNOTSUPP;
2602 if (err == 0) {
2603 err = -ENOBUFS;
2604 dev = __dev_get_by_name(net, p.name);
2605 if (!dev)
2606 goto err_exit;
2607 err = dev_open(dev);
2610 #endif
2612 err_exit:
2613 rtnl_unlock();
2614 return err;
2617 static int ipv6_mc_config(struct sock *sk, bool join,
2618 const struct in6_addr *addr, int ifindex)
2620 int ret;
2622 ASSERT_RTNL();
2624 lock_sock(sk);
2625 if (join)
2626 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2627 else
2628 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2629 release_sock(sk);
2631 return ret;
2635 * Manual configuration of address on an interface
2637 static int inet6_addr_add(struct net *net, int ifindex,
2638 const struct in6_addr *pfx,
2639 const struct in6_addr *peer_pfx,
2640 unsigned int plen, __u32 ifa_flags,
2641 __u32 prefered_lft, __u32 valid_lft)
2643 struct inet6_ifaddr *ifp;
2644 struct inet6_dev *idev;
2645 struct net_device *dev;
2646 unsigned long timeout;
2647 clock_t expires;
2648 int scope;
2649 u32 flags;
2651 ASSERT_RTNL();
2653 if (plen > 128)
2654 return -EINVAL;
2656 /* check the lifetime */
2657 if (!valid_lft || prefered_lft > valid_lft)
2658 return -EINVAL;
2660 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2661 return -EINVAL;
2663 dev = __dev_get_by_index(net, ifindex);
2664 if (!dev)
2665 return -ENODEV;
2667 idev = addrconf_add_dev(dev);
2668 if (IS_ERR(idev))
2669 return PTR_ERR(idev);
2671 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2672 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2673 true, pfx, ifindex);
2675 if (ret < 0)
2676 return ret;
2679 scope = ipv6_addr_scope(pfx);
2681 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2682 if (addrconf_finite_timeout(timeout)) {
2683 expires = jiffies_to_clock_t(timeout * HZ);
2684 valid_lft = timeout;
2685 flags = RTF_EXPIRES;
2686 } else {
2687 expires = 0;
2688 flags = 0;
2689 ifa_flags |= IFA_F_PERMANENT;
2692 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2693 if (addrconf_finite_timeout(timeout)) {
2694 if (timeout == 0)
2695 ifa_flags |= IFA_F_DEPRECATED;
2696 prefered_lft = timeout;
2699 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2700 valid_lft, prefered_lft);
2702 if (!IS_ERR(ifp)) {
2703 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2704 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2705 expires, flags);
2709 * Note that section 3.1 of RFC 4429 indicates
2710 * that the Optimistic flag should not be set for
2711 * manually configured addresses
2713 addrconf_dad_start(ifp);
2714 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2715 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2716 true, jiffies);
2717 in6_ifa_put(ifp);
2718 addrconf_verify_rtnl();
2719 return 0;
2720 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2721 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2722 false, pfx, ifindex);
2725 return PTR_ERR(ifp);
2728 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2729 const struct in6_addr *pfx, unsigned int plen)
2731 struct inet6_ifaddr *ifp;
2732 struct inet6_dev *idev;
2733 struct net_device *dev;
2735 if (plen > 128)
2736 return -EINVAL;
2738 dev = __dev_get_by_index(net, ifindex);
2739 if (!dev)
2740 return -ENODEV;
2742 idev = __in6_dev_get(dev);
2743 if (!idev)
2744 return -ENXIO;
2746 read_lock_bh(&idev->lock);
2747 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2748 if (ifp->prefix_len == plen &&
2749 ipv6_addr_equal(pfx, &ifp->addr)) {
2750 in6_ifa_hold(ifp);
2751 read_unlock_bh(&idev->lock);
2753 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2754 (ifa_flags & IFA_F_MANAGETEMPADDR))
2755 manage_tempaddrs(idev, ifp, 0, 0, false,
2756 jiffies);
2757 ipv6_del_addr(ifp);
2758 addrconf_verify_rtnl();
2759 if (ipv6_addr_is_multicast(pfx)) {
2760 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2761 false, pfx, dev->ifindex);
2763 return 0;
2766 read_unlock_bh(&idev->lock);
2767 return -EADDRNOTAVAIL;
2771 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2773 struct in6_ifreq ireq;
2774 int err;
2776 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2777 return -EPERM;
2779 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2780 return -EFAULT;
2782 rtnl_lock();
2783 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2784 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2785 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2786 rtnl_unlock();
2787 return err;
2790 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2792 struct in6_ifreq ireq;
2793 int err;
2795 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2796 return -EPERM;
2798 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2799 return -EFAULT;
2801 rtnl_lock();
2802 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2803 ireq.ifr6_prefixlen);
2804 rtnl_unlock();
2805 return err;
2808 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2809 int plen, int scope)
2811 struct inet6_ifaddr *ifp;
2813 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2814 scope, IFA_F_PERMANENT,
2815 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2816 if (!IS_ERR(ifp)) {
2817 spin_lock_bh(&ifp->lock);
2818 ifp->flags &= ~IFA_F_TENTATIVE;
2819 spin_unlock_bh(&ifp->lock);
2820 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2821 in6_ifa_put(ifp);
2825 #if IS_ENABLED(CONFIG_IPV6_SIT)
2826 static void sit_add_v4_addrs(struct inet6_dev *idev)
2828 struct in6_addr addr;
2829 struct net_device *dev;
2830 struct net *net = dev_net(idev->dev);
2831 int scope, plen;
2832 u32 pflags = 0;
2834 ASSERT_RTNL();
2836 memset(&addr, 0, sizeof(struct in6_addr));
2837 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2839 if (idev->dev->flags&IFF_POINTOPOINT) {
2840 addr.s6_addr32[0] = htonl(0xfe800000);
2841 scope = IFA_LINK;
2842 plen = 64;
2843 } else {
2844 scope = IPV6_ADDR_COMPATv4;
2845 plen = 96;
2846 pflags |= RTF_NONEXTHOP;
2849 if (addr.s6_addr32[3]) {
2850 add_addr(idev, &addr, plen, scope);
2851 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2852 return;
2855 for_each_netdev(net, dev) {
2856 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2857 if (in_dev && (dev->flags & IFF_UP)) {
2858 struct in_ifaddr *ifa;
2860 int flag = scope;
2862 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2864 addr.s6_addr32[3] = ifa->ifa_local;
2866 if (ifa->ifa_scope == RT_SCOPE_LINK)
2867 continue;
2868 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2869 if (idev->dev->flags&IFF_POINTOPOINT)
2870 continue;
2871 flag |= IFA_HOST;
2874 add_addr(idev, &addr, plen, flag);
2875 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2876 pflags);
2881 #endif
2883 static void init_loopback(struct net_device *dev)
2885 struct inet6_dev *idev;
2886 struct net_device *sp_dev;
2887 struct inet6_ifaddr *sp_ifa;
2888 struct rt6_info *sp_rt;
2890 /* ::1 */
2892 ASSERT_RTNL();
2894 idev = ipv6_find_idev(dev);
2895 if (!idev) {
2896 pr_debug("%s: add_dev failed\n", __func__);
2897 return;
2900 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2902 /* Add routes to other interface's IPv6 addresses */
2903 for_each_netdev(dev_net(dev), sp_dev) {
2904 if (!strcmp(sp_dev->name, dev->name))
2905 continue;
2907 idev = __in6_dev_get(sp_dev);
2908 if (!idev)
2909 continue;
2911 read_lock_bh(&idev->lock);
2912 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2914 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2915 continue;
2917 if (sp_ifa->rt) {
2918 /* This dst has been added to garbage list when
2919 * lo device down, release this obsolete dst and
2920 * reallocate a new router for ifa.
2922 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
2923 ip6_rt_put(sp_ifa->rt);
2924 sp_ifa->rt = NULL;
2925 } else {
2926 continue;
2930 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2932 /* Failure cases are ignored */
2933 if (!IS_ERR(sp_rt)) {
2934 sp_ifa->rt = sp_rt;
2935 ip6_ins_rt(sp_rt);
2938 read_unlock_bh(&idev->lock);
2942 static void addrconf_add_linklocal(struct inet6_dev *idev,
2943 const struct in6_addr *addr, u32 flags)
2945 struct inet6_ifaddr *ifp;
2946 u32 addr_flags = flags | IFA_F_PERMANENT;
2948 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2949 if (idev->cnf.optimistic_dad &&
2950 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2951 addr_flags |= IFA_F_OPTIMISTIC;
2952 #endif
2954 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2955 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2956 if (!IS_ERR(ifp)) {
2957 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2958 addrconf_dad_start(ifp);
2959 in6_ifa_put(ifp);
2963 static bool ipv6_reserved_interfaceid(struct in6_addr address)
2965 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
2966 return true;
2968 if (address.s6_addr32[2] == htonl(0x02005eff) &&
2969 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
2970 return true;
2972 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
2973 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
2974 return true;
2976 return false;
2979 static int ipv6_generate_stable_address(struct in6_addr *address,
2980 u8 dad_count,
2981 const struct inet6_dev *idev)
2983 static DEFINE_SPINLOCK(lock);
2984 static __u32 digest[SHA_DIGEST_WORDS];
2985 static __u32 workspace[SHA_WORKSPACE_WORDS];
2987 static union {
2988 char __data[SHA_MESSAGE_BYTES];
2989 struct {
2990 struct in6_addr secret;
2991 __be32 prefix[2];
2992 unsigned char hwaddr[MAX_ADDR_LEN];
2993 u8 dad_count;
2994 } __packed;
2995 } data;
2997 struct in6_addr secret;
2998 struct in6_addr temp;
2999 struct net *net = dev_net(idev->dev);
3001 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3003 if (idev->cnf.stable_secret.initialized)
3004 secret = idev->cnf.stable_secret.secret;
3005 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3006 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3007 else
3008 return -1;
3010 retry:
3011 spin_lock_bh(&lock);
3013 sha_init(digest);
3014 memset(&data, 0, sizeof(data));
3015 memset(workspace, 0, sizeof(workspace));
3016 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3017 data.prefix[0] = address->s6_addr32[0];
3018 data.prefix[1] = address->s6_addr32[1];
3019 data.secret = secret;
3020 data.dad_count = dad_count;
3022 sha_transform(digest, data.__data, workspace);
3024 temp = *address;
3025 temp.s6_addr32[2] = (__force __be32)digest[0];
3026 temp.s6_addr32[3] = (__force __be32)digest[1];
3028 spin_unlock_bh(&lock);
3030 if (ipv6_reserved_interfaceid(temp)) {
3031 dad_count++;
3032 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3033 return -1;
3034 goto retry;
3037 *address = temp;
3038 return 0;
3041 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3043 struct in6_addr addr;
3045 /* no link local addresses on L3 master devices */
3046 if (netif_is_l3_master(idev->dev))
3047 return;
3049 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3051 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY) {
3052 if (!ipv6_generate_stable_address(&addr, 0, idev))
3053 addrconf_add_linklocal(idev, &addr,
3054 IFA_F_STABLE_PRIVACY);
3055 else if (prefix_route)
3056 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3057 } else if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
3058 /* addrconf_add_linklocal also adds a prefix_route and we
3059 * only need to care about prefix routes if ipv6_generate_eui64
3060 * couldn't generate one.
3062 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3063 addrconf_add_linklocal(idev, &addr, 0);
3064 else if (prefix_route)
3065 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3069 static void addrconf_dev_config(struct net_device *dev)
3071 struct inet6_dev *idev;
3073 ASSERT_RTNL();
3075 if ((dev->type != ARPHRD_ETHER) &&
3076 (dev->type != ARPHRD_FDDI) &&
3077 (dev->type != ARPHRD_ARCNET) &&
3078 (dev->type != ARPHRD_INFINIBAND) &&
3079 (dev->type != ARPHRD_IEEE802154) &&
3080 (dev->type != ARPHRD_IEEE1394) &&
3081 (dev->type != ARPHRD_TUNNEL6) &&
3082 (dev->type != ARPHRD_6LOWPAN)) {
3083 /* Alas, we support only Ethernet autoconfiguration. */
3084 return;
3087 idev = addrconf_add_dev(dev);
3088 if (IS_ERR(idev))
3089 return;
3091 addrconf_addr_gen(idev, false);
3094 #if IS_ENABLED(CONFIG_IPV6_SIT)
3095 static void addrconf_sit_config(struct net_device *dev)
3097 struct inet6_dev *idev;
3099 ASSERT_RTNL();
3102 * Configure the tunnel with one of our IPv4
3103 * addresses... we should configure all of
3104 * our v4 addrs in the tunnel
3107 idev = ipv6_find_idev(dev);
3108 if (!idev) {
3109 pr_debug("%s: add_dev failed\n", __func__);
3110 return;
3113 if (dev->priv_flags & IFF_ISATAP) {
3114 addrconf_addr_gen(idev, false);
3115 return;
3118 sit_add_v4_addrs(idev);
3120 if (dev->flags&IFF_POINTOPOINT)
3121 addrconf_add_mroute(dev);
3123 #endif
3125 #if IS_ENABLED(CONFIG_NET_IPGRE)
3126 static void addrconf_gre_config(struct net_device *dev)
3128 struct inet6_dev *idev;
3130 ASSERT_RTNL();
3132 idev = ipv6_find_idev(dev);
3133 if (!idev) {
3134 pr_debug("%s: add_dev failed\n", __func__);
3135 return;
3138 addrconf_addr_gen(idev, true);
3139 if (dev->flags & IFF_POINTOPOINT)
3140 addrconf_add_mroute(dev);
3142 #endif
3144 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3145 void *ptr)
3147 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3148 struct inet6_dev *idev = __in6_dev_get(dev);
3149 struct net *net = dev_net(dev);
3150 int run_pending = 0;
3151 int err;
3153 switch (event) {
3154 case NETDEV_REGISTER:
3155 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3156 idev = ipv6_add_dev(dev);
3157 if (IS_ERR(idev))
3158 return notifier_from_errno(PTR_ERR(idev));
3160 break;
3162 case NETDEV_CHANGEMTU:
3163 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3164 if (dev->mtu < IPV6_MIN_MTU) {
3165 addrconf_ifdown(dev, dev != net->loopback_dev);
3166 break;
3169 if (idev) {
3170 rt6_mtu_change(dev, dev->mtu);
3171 idev->cnf.mtu6 = dev->mtu;
3172 break;
3175 /* allocate new idev */
3176 idev = ipv6_add_dev(dev);
3177 if (IS_ERR(idev))
3178 break;
3180 /* device is still not ready */
3181 if (!(idev->if_flags & IF_READY))
3182 break;
3184 run_pending = 1;
3186 /* fall through */
3188 case NETDEV_UP:
3189 case NETDEV_CHANGE:
3190 if (dev->flags & IFF_SLAVE)
3191 break;
3193 if (idev && idev->cnf.disable_ipv6)
3194 break;
3196 if (event == NETDEV_UP) {
3197 if (!addrconf_qdisc_ok(dev)) {
3198 /* device is not ready yet. */
3199 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3200 dev->name);
3201 break;
3204 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3205 idev = ipv6_add_dev(dev);
3207 if (!IS_ERR_OR_NULL(idev)) {
3208 idev->if_flags |= IF_READY;
3209 run_pending = 1;
3211 } else if (event == NETDEV_CHANGE) {
3212 if (!addrconf_qdisc_ok(dev)) {
3213 /* device is still not ready. */
3214 break;
3217 if (idev) {
3218 if (idev->if_flags & IF_READY) {
3219 /* device is already configured -
3220 * but resend MLD reports, we might
3221 * have roamed and need to update
3222 * multicast snooping switches
3224 ipv6_mc_up(idev);
3225 break;
3227 idev->if_flags |= IF_READY;
3230 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3231 dev->name);
3233 run_pending = 1;
3236 switch (dev->type) {
3237 #if IS_ENABLED(CONFIG_IPV6_SIT)
3238 case ARPHRD_SIT:
3239 addrconf_sit_config(dev);
3240 break;
3241 #endif
3242 #if IS_ENABLED(CONFIG_NET_IPGRE)
3243 case ARPHRD_IPGRE:
3244 addrconf_gre_config(dev);
3245 break;
3246 #endif
3247 case ARPHRD_LOOPBACK:
3248 init_loopback(dev);
3249 break;
3251 default:
3252 addrconf_dev_config(dev);
3253 break;
3256 if (!IS_ERR_OR_NULL(idev)) {
3257 if (run_pending)
3258 addrconf_dad_run(idev);
3261 * If the MTU changed during the interface down,
3262 * when the interface up, the changed MTU must be
3263 * reflected in the idev as well as routers.
3265 if (idev->cnf.mtu6 != dev->mtu &&
3266 dev->mtu >= IPV6_MIN_MTU) {
3267 rt6_mtu_change(dev, dev->mtu);
3268 idev->cnf.mtu6 = dev->mtu;
3270 idev->tstamp = jiffies;
3271 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3274 * If the changed mtu during down is lower than
3275 * IPV6_MIN_MTU stop IPv6 on this interface.
3277 if (dev->mtu < IPV6_MIN_MTU)
3278 addrconf_ifdown(dev, dev != net->loopback_dev);
3280 break;
3282 case NETDEV_DOWN:
3283 case NETDEV_UNREGISTER:
3285 * Remove all addresses from this interface.
3287 addrconf_ifdown(dev, event != NETDEV_DOWN);
3288 break;
3290 case NETDEV_CHANGENAME:
3291 if (idev) {
3292 snmp6_unregister_dev(idev);
3293 addrconf_sysctl_unregister(idev);
3294 err = addrconf_sysctl_register(idev);
3295 if (err)
3296 return notifier_from_errno(err);
3297 err = snmp6_register_dev(idev);
3298 if (err) {
3299 addrconf_sysctl_unregister(idev);
3300 return notifier_from_errno(err);
3303 break;
3305 case NETDEV_PRE_TYPE_CHANGE:
3306 case NETDEV_POST_TYPE_CHANGE:
3307 addrconf_type_change(dev, event);
3308 break;
3311 return NOTIFY_OK;
3315 * addrconf module should be notified of a device going up
3317 static struct notifier_block ipv6_dev_notf = {
3318 .notifier_call = addrconf_notify,
3319 .priority = ADDRCONF_NOTIFY_PRIORITY,
3322 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3324 struct inet6_dev *idev;
3325 ASSERT_RTNL();
3327 idev = __in6_dev_get(dev);
3329 if (event == NETDEV_POST_TYPE_CHANGE)
3330 ipv6_mc_remap(idev);
3331 else if (event == NETDEV_PRE_TYPE_CHANGE)
3332 ipv6_mc_unmap(idev);
3335 static int addrconf_ifdown(struct net_device *dev, int how)
3337 struct net *net = dev_net(dev);
3338 struct inet6_dev *idev;
3339 struct inet6_ifaddr *ifa;
3340 int state, i;
3342 ASSERT_RTNL();
3344 rt6_ifdown(net, dev);
3345 neigh_ifdown(&nd_tbl, dev);
3347 idev = __in6_dev_get(dev);
3348 if (!idev)
3349 return -ENODEV;
3352 * Step 1: remove reference to ipv6 device from parent device.
3353 * Do not dev_put!
3355 if (how) {
3356 idev->dead = 1;
3358 /* protected by rtnl_lock */
3359 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3361 /* Step 1.5: remove snmp6 entry */
3362 snmp6_unregister_dev(idev);
3366 /* Step 2: clear hash table */
3367 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3368 struct hlist_head *h = &inet6_addr_lst[i];
3370 spin_lock_bh(&addrconf_hash_lock);
3371 restart:
3372 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3373 if (ifa->idev == idev) {
3374 hlist_del_init_rcu(&ifa->addr_lst);
3375 addrconf_del_dad_work(ifa);
3376 goto restart;
3379 spin_unlock_bh(&addrconf_hash_lock);
3382 write_lock_bh(&idev->lock);
3384 addrconf_del_rs_timer(idev);
3386 /* Step 2: clear flags for stateless addrconf */
3387 if (!how)
3388 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3390 if (how && del_timer(&idev->regen_timer))
3391 in6_dev_put(idev);
3393 /* Step 3: clear tempaddr list */
3394 while (!list_empty(&idev->tempaddr_list)) {
3395 ifa = list_first_entry(&idev->tempaddr_list,
3396 struct inet6_ifaddr, tmp_list);
3397 list_del(&ifa->tmp_list);
3398 write_unlock_bh(&idev->lock);
3399 spin_lock_bh(&ifa->lock);
3401 if (ifa->ifpub) {
3402 in6_ifa_put(ifa->ifpub);
3403 ifa->ifpub = NULL;
3405 spin_unlock_bh(&ifa->lock);
3406 in6_ifa_put(ifa);
3407 write_lock_bh(&idev->lock);
3410 while (!list_empty(&idev->addr_list)) {
3411 ifa = list_first_entry(&idev->addr_list,
3412 struct inet6_ifaddr, if_list);
3413 addrconf_del_dad_work(ifa);
3415 list_del(&ifa->if_list);
3417 write_unlock_bh(&idev->lock);
3419 spin_lock_bh(&ifa->lock);
3420 state = ifa->state;
3421 ifa->state = INET6_IFADDR_STATE_DEAD;
3422 spin_unlock_bh(&ifa->lock);
3424 if (state != INET6_IFADDR_STATE_DEAD) {
3425 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3426 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3428 in6_ifa_put(ifa);
3430 write_lock_bh(&idev->lock);
3433 write_unlock_bh(&idev->lock);
3435 /* Step 5: Discard anycast and multicast list */
3436 if (how) {
3437 ipv6_ac_destroy_dev(idev);
3438 ipv6_mc_destroy_dev(idev);
3439 } else {
3440 ipv6_mc_down(idev);
3443 idev->tstamp = jiffies;
3445 /* Last: Shot the device (if unregistered) */
3446 if (how) {
3447 addrconf_sysctl_unregister(idev);
3448 neigh_parms_release(&nd_tbl, idev->nd_parms);
3449 neigh_ifdown(&nd_tbl, dev);
3450 in6_dev_put(idev);
3452 return 0;
3455 static void addrconf_rs_timer(unsigned long data)
3457 struct inet6_dev *idev = (struct inet6_dev *)data;
3458 struct net_device *dev = idev->dev;
3459 struct in6_addr lladdr;
3461 write_lock(&idev->lock);
3462 if (idev->dead || !(idev->if_flags & IF_READY))
3463 goto out;
3465 if (!ipv6_accept_ra(idev))
3466 goto out;
3468 /* Announcement received after solicitation was sent */
3469 if (idev->if_flags & IF_RA_RCVD)
3470 goto out;
3472 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3473 write_unlock(&idev->lock);
3474 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3475 ndisc_send_rs(dev, &lladdr,
3476 &in6addr_linklocal_allrouters);
3477 else
3478 goto put;
3480 write_lock(&idev->lock);
3481 /* The wait after the last probe can be shorter */
3482 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3483 idev->cnf.rtr_solicits) ?
3484 idev->cnf.rtr_solicit_delay :
3485 idev->cnf.rtr_solicit_interval);
3486 } else {
3488 * Note: we do not support deprecated "all on-link"
3489 * assumption any longer.
3491 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3494 out:
3495 write_unlock(&idev->lock);
3496 put:
3497 in6_dev_put(idev);
3501 * Duplicate Address Detection
3503 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3505 unsigned long rand_num;
3506 struct inet6_dev *idev = ifp->idev;
3508 if (ifp->flags & IFA_F_OPTIMISTIC)
3509 rand_num = 0;
3510 else
3511 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3513 ifp->dad_probes = idev->cnf.dad_transmits;
3514 addrconf_mod_dad_work(ifp, rand_num);
3517 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3519 struct inet6_dev *idev = ifp->idev;
3520 struct net_device *dev = idev->dev;
3521 bool notify = false;
3523 addrconf_join_solict(dev, &ifp->addr);
3525 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3527 read_lock_bh(&idev->lock);
3528 spin_lock(&ifp->lock);
3529 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3530 goto out;
3532 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3533 idev->cnf.accept_dad < 1 ||
3534 !(ifp->flags&IFA_F_TENTATIVE) ||
3535 ifp->flags & IFA_F_NODAD) {
3536 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3537 spin_unlock(&ifp->lock);
3538 read_unlock_bh(&idev->lock);
3540 addrconf_dad_completed(ifp);
3541 return;
3544 if (!(idev->if_flags & IF_READY)) {
3545 spin_unlock(&ifp->lock);
3546 read_unlock_bh(&idev->lock);
3548 * If the device is not ready:
3549 * - keep it tentative if it is a permanent address.
3550 * - otherwise, kill it.
3552 in6_ifa_hold(ifp);
3553 addrconf_dad_stop(ifp, 0);
3554 return;
3558 * Optimistic nodes can start receiving
3559 * Frames right away
3561 if (ifp->flags & IFA_F_OPTIMISTIC) {
3562 ip6_ins_rt(ifp->rt);
3563 if (ipv6_use_optimistic_addr(idev)) {
3564 /* Because optimistic nodes can use this address,
3565 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3567 notify = true;
3571 addrconf_dad_kick(ifp);
3572 out:
3573 spin_unlock(&ifp->lock);
3574 read_unlock_bh(&idev->lock);
3575 if (notify)
3576 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3579 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3581 bool begin_dad = false;
3583 spin_lock_bh(&ifp->lock);
3584 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3585 ifp->state = INET6_IFADDR_STATE_PREDAD;
3586 begin_dad = true;
3588 spin_unlock_bh(&ifp->lock);
3590 if (begin_dad)
3591 addrconf_mod_dad_work(ifp, 0);
3594 static void addrconf_dad_work(struct work_struct *w)
3596 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3597 struct inet6_ifaddr,
3598 dad_work);
3599 struct inet6_dev *idev = ifp->idev;
3600 struct in6_addr mcaddr;
3602 enum {
3603 DAD_PROCESS,
3604 DAD_BEGIN,
3605 DAD_ABORT,
3606 } action = DAD_PROCESS;
3608 rtnl_lock();
3610 spin_lock_bh(&ifp->lock);
3611 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3612 action = DAD_BEGIN;
3613 ifp->state = INET6_IFADDR_STATE_DAD;
3614 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3615 action = DAD_ABORT;
3616 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3618 spin_unlock_bh(&ifp->lock);
3620 if (action == DAD_BEGIN) {
3621 addrconf_dad_begin(ifp);
3622 goto out;
3623 } else if (action == DAD_ABORT) {
3624 in6_ifa_hold(ifp);
3625 addrconf_dad_stop(ifp, 1);
3626 goto out;
3629 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3630 goto out;
3632 write_lock_bh(&idev->lock);
3633 if (idev->dead || !(idev->if_flags & IF_READY)) {
3634 write_unlock_bh(&idev->lock);
3635 goto out;
3638 spin_lock(&ifp->lock);
3639 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3640 spin_unlock(&ifp->lock);
3641 write_unlock_bh(&idev->lock);
3642 goto out;
3645 if (ifp->dad_probes == 0) {
3647 * DAD was successful
3650 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3651 spin_unlock(&ifp->lock);
3652 write_unlock_bh(&idev->lock);
3654 addrconf_dad_completed(ifp);
3656 goto out;
3659 ifp->dad_probes--;
3660 addrconf_mod_dad_work(ifp,
3661 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3662 spin_unlock(&ifp->lock);
3663 write_unlock_bh(&idev->lock);
3665 /* send a neighbour solicitation for our addr */
3666 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3667 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any);
3668 out:
3669 in6_ifa_put(ifp);
3670 rtnl_unlock();
3673 /* ifp->idev must be at least read locked */
3674 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3676 struct inet6_ifaddr *ifpiter;
3677 struct inet6_dev *idev = ifp->idev;
3679 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3680 if (ifpiter->scope > IFA_LINK)
3681 break;
3682 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3683 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3684 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3685 IFA_F_PERMANENT)
3686 return false;
3688 return true;
3691 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3693 struct net_device *dev = ifp->idev->dev;
3694 struct in6_addr lladdr;
3695 bool send_rs, send_mld;
3697 addrconf_del_dad_work(ifp);
3700 * Configure the address for reception. Now it is valid.
3703 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3705 /* If added prefix is link local and we are prepared to process
3706 router advertisements, start sending router solicitations.
3709 read_lock_bh(&ifp->idev->lock);
3710 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3711 send_rs = send_mld &&
3712 ipv6_accept_ra(ifp->idev) &&
3713 ifp->idev->cnf.rtr_solicits > 0 &&
3714 (dev->flags&IFF_LOOPBACK) == 0;
3715 read_unlock_bh(&ifp->idev->lock);
3717 /* While dad is in progress mld report's source address is in6_addrany.
3718 * Resend with proper ll now.
3720 if (send_mld)
3721 ipv6_mc_dad_complete(ifp->idev);
3723 if (send_rs) {
3725 * If a host as already performed a random delay
3726 * [...] as part of DAD [...] there is no need
3727 * to delay again before sending the first RS
3729 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3730 return;
3731 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3733 write_lock_bh(&ifp->idev->lock);
3734 spin_lock(&ifp->lock);
3735 ifp->idev->rs_probes = 1;
3736 ifp->idev->if_flags |= IF_RS_SENT;
3737 addrconf_mod_rs_timer(ifp->idev,
3738 ifp->idev->cnf.rtr_solicit_interval);
3739 spin_unlock(&ifp->lock);
3740 write_unlock_bh(&ifp->idev->lock);
3744 static void addrconf_dad_run(struct inet6_dev *idev)
3746 struct inet6_ifaddr *ifp;
3748 read_lock_bh(&idev->lock);
3749 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3750 spin_lock(&ifp->lock);
3751 if (ifp->flags & IFA_F_TENTATIVE &&
3752 ifp->state == INET6_IFADDR_STATE_DAD)
3753 addrconf_dad_kick(ifp);
3754 spin_unlock(&ifp->lock);
3756 read_unlock_bh(&idev->lock);
3759 #ifdef CONFIG_PROC_FS
3760 struct if6_iter_state {
3761 struct seq_net_private p;
3762 int bucket;
3763 int offset;
3766 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3768 struct inet6_ifaddr *ifa = NULL;
3769 struct if6_iter_state *state = seq->private;
3770 struct net *net = seq_file_net(seq);
3771 int p = 0;
3773 /* initial bucket if pos is 0 */
3774 if (pos == 0) {
3775 state->bucket = 0;
3776 state->offset = 0;
3779 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3780 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3781 addr_lst) {
3782 if (!net_eq(dev_net(ifa->idev->dev), net))
3783 continue;
3784 /* sync with offset */
3785 if (p < state->offset) {
3786 p++;
3787 continue;
3789 return ifa;
3792 /* prepare for next bucket */
3793 state->offset = 0;
3794 p = 0;
3796 return NULL;
3799 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3800 struct inet6_ifaddr *ifa)
3802 struct if6_iter_state *state = seq->private;
3803 struct net *net = seq_file_net(seq);
3805 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3806 if (!net_eq(dev_net(ifa->idev->dev), net))
3807 continue;
3808 state->offset++;
3809 return ifa;
3812 state->offset = 0;
3813 while (++state->bucket < IN6_ADDR_HSIZE) {
3814 hlist_for_each_entry_rcu_bh(ifa,
3815 &inet6_addr_lst[state->bucket], addr_lst) {
3816 if (!net_eq(dev_net(ifa->idev->dev), net))
3817 continue;
3818 return ifa;
3822 return NULL;
3825 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3826 __acquires(rcu_bh)
3828 rcu_read_lock_bh();
3829 return if6_get_first(seq, *pos);
3832 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3834 struct inet6_ifaddr *ifa;
3836 ifa = if6_get_next(seq, v);
3837 ++*pos;
3838 return ifa;
3841 static void if6_seq_stop(struct seq_file *seq, void *v)
3842 __releases(rcu_bh)
3844 rcu_read_unlock_bh();
3847 static int if6_seq_show(struct seq_file *seq, void *v)
3849 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3850 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3851 &ifp->addr,
3852 ifp->idev->dev->ifindex,
3853 ifp->prefix_len,
3854 ifp->scope,
3855 (u8) ifp->flags,
3856 ifp->idev->dev->name);
3857 return 0;
3860 static const struct seq_operations if6_seq_ops = {
3861 .start = if6_seq_start,
3862 .next = if6_seq_next,
3863 .show = if6_seq_show,
3864 .stop = if6_seq_stop,
3867 static int if6_seq_open(struct inode *inode, struct file *file)
3869 return seq_open_net(inode, file, &if6_seq_ops,
3870 sizeof(struct if6_iter_state));
3873 static const struct file_operations if6_fops = {
3874 .owner = THIS_MODULE,
3875 .open = if6_seq_open,
3876 .read = seq_read,
3877 .llseek = seq_lseek,
3878 .release = seq_release_net,
3881 static int __net_init if6_proc_net_init(struct net *net)
3883 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3884 return -ENOMEM;
3885 return 0;
3888 static void __net_exit if6_proc_net_exit(struct net *net)
3890 remove_proc_entry("if_inet6", net->proc_net);
3893 static struct pernet_operations if6_proc_net_ops = {
3894 .init = if6_proc_net_init,
3895 .exit = if6_proc_net_exit,
3898 int __init if6_proc_init(void)
3900 return register_pernet_subsys(&if6_proc_net_ops);
3903 void if6_proc_exit(void)
3905 unregister_pernet_subsys(&if6_proc_net_ops);
3907 #endif /* CONFIG_PROC_FS */
3909 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3910 /* Check if address is a home address configured on any interface. */
3911 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3913 int ret = 0;
3914 struct inet6_ifaddr *ifp = NULL;
3915 unsigned int hash = inet6_addr_hash(addr);
3917 rcu_read_lock_bh();
3918 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3919 if (!net_eq(dev_net(ifp->idev->dev), net))
3920 continue;
3921 if (ipv6_addr_equal(&ifp->addr, addr) &&
3922 (ifp->flags & IFA_F_HOMEADDRESS)) {
3923 ret = 1;
3924 break;
3927 rcu_read_unlock_bh();
3928 return ret;
3930 #endif
3933 * Periodic address status verification
3936 static void addrconf_verify_rtnl(void)
3938 unsigned long now, next, next_sec, next_sched;
3939 struct inet6_ifaddr *ifp;
3940 int i;
3942 ASSERT_RTNL();
3944 rcu_read_lock_bh();
3945 now = jiffies;
3946 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3948 cancel_delayed_work(&addr_chk_work);
3950 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3951 restart:
3952 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3953 unsigned long age;
3955 /* When setting preferred_lft to a value not zero or
3956 * infinity, while valid_lft is infinity
3957 * IFA_F_PERMANENT has a non-infinity life time.
3959 if ((ifp->flags & IFA_F_PERMANENT) &&
3960 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3961 continue;
3963 spin_lock(&ifp->lock);
3964 /* We try to batch several events at once. */
3965 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3967 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3968 age >= ifp->valid_lft) {
3969 spin_unlock(&ifp->lock);
3970 in6_ifa_hold(ifp);
3971 ipv6_del_addr(ifp);
3972 goto restart;
3973 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3974 spin_unlock(&ifp->lock);
3975 continue;
3976 } else if (age >= ifp->prefered_lft) {
3977 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3978 int deprecate = 0;
3980 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3981 deprecate = 1;
3982 ifp->flags |= IFA_F_DEPRECATED;
3985 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3986 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3987 next = ifp->tstamp + ifp->valid_lft * HZ;
3989 spin_unlock(&ifp->lock);
3991 if (deprecate) {
3992 in6_ifa_hold(ifp);
3994 ipv6_ifa_notify(0, ifp);
3995 in6_ifa_put(ifp);
3996 goto restart;
3998 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3999 !(ifp->flags&IFA_F_TENTATIVE)) {
4000 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4001 ifp->idev->cnf.dad_transmits *
4002 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4004 if (age >= ifp->prefered_lft - regen_advance) {
4005 struct inet6_ifaddr *ifpub = ifp->ifpub;
4006 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4007 next = ifp->tstamp + ifp->prefered_lft * HZ;
4008 if (!ifp->regen_count && ifpub) {
4009 ifp->regen_count++;
4010 in6_ifa_hold(ifp);
4011 in6_ifa_hold(ifpub);
4012 spin_unlock(&ifp->lock);
4014 spin_lock(&ifpub->lock);
4015 ifpub->regen_count = 0;
4016 spin_unlock(&ifpub->lock);
4017 ipv6_create_tempaddr(ifpub, ifp);
4018 in6_ifa_put(ifpub);
4019 in6_ifa_put(ifp);
4020 goto restart;
4022 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4023 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4024 spin_unlock(&ifp->lock);
4025 } else {
4026 /* ifp->prefered_lft <= ifp->valid_lft */
4027 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4028 next = ifp->tstamp + ifp->prefered_lft * HZ;
4029 spin_unlock(&ifp->lock);
4034 next_sec = round_jiffies_up(next);
4035 next_sched = next;
4037 /* If rounded timeout is accurate enough, accept it. */
4038 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4039 next_sched = next_sec;
4041 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4042 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4043 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4045 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4046 now, next, next_sec, next_sched);
4047 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4048 rcu_read_unlock_bh();
4051 static void addrconf_verify_work(struct work_struct *w)
4053 rtnl_lock();
4054 addrconf_verify_rtnl();
4055 rtnl_unlock();
4058 static void addrconf_verify(void)
4060 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4063 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4064 struct in6_addr **peer_pfx)
4066 struct in6_addr *pfx = NULL;
4068 *peer_pfx = NULL;
4070 if (addr)
4071 pfx = nla_data(addr);
4073 if (local) {
4074 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4075 *peer_pfx = pfx;
4076 pfx = nla_data(local);
4079 return pfx;
4082 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4083 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4084 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4085 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4086 [IFA_FLAGS] = { .len = sizeof(u32) },
4089 static int
4090 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4092 struct net *net = sock_net(skb->sk);
4093 struct ifaddrmsg *ifm;
4094 struct nlattr *tb[IFA_MAX+1];
4095 struct in6_addr *pfx, *peer_pfx;
4096 u32 ifa_flags;
4097 int err;
4099 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4100 if (err < 0)
4101 return err;
4103 ifm = nlmsg_data(nlh);
4104 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4105 if (!pfx)
4106 return -EINVAL;
4108 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4110 /* We ignore other flags so far. */
4111 ifa_flags &= IFA_F_MANAGETEMPADDR;
4113 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4114 ifm->ifa_prefixlen);
4117 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4118 u32 prefered_lft, u32 valid_lft)
4120 u32 flags;
4121 clock_t expires;
4122 unsigned long timeout;
4123 bool was_managetempaddr;
4124 bool had_prefixroute;
4126 ASSERT_RTNL();
4128 if (!valid_lft || (prefered_lft > valid_lft))
4129 return -EINVAL;
4131 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4132 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4133 return -EINVAL;
4135 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4136 if (addrconf_finite_timeout(timeout)) {
4137 expires = jiffies_to_clock_t(timeout * HZ);
4138 valid_lft = timeout;
4139 flags = RTF_EXPIRES;
4140 } else {
4141 expires = 0;
4142 flags = 0;
4143 ifa_flags |= IFA_F_PERMANENT;
4146 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4147 if (addrconf_finite_timeout(timeout)) {
4148 if (timeout == 0)
4149 ifa_flags |= IFA_F_DEPRECATED;
4150 prefered_lft = timeout;
4153 spin_lock_bh(&ifp->lock);
4154 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4155 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4156 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4157 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4158 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4159 IFA_F_NOPREFIXROUTE);
4160 ifp->flags |= ifa_flags;
4161 ifp->tstamp = jiffies;
4162 ifp->valid_lft = valid_lft;
4163 ifp->prefered_lft = prefered_lft;
4165 spin_unlock_bh(&ifp->lock);
4166 if (!(ifp->flags&IFA_F_TENTATIVE))
4167 ipv6_ifa_notify(0, ifp);
4169 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4170 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4171 expires, flags);
4172 } else if (had_prefixroute) {
4173 enum cleanup_prefix_rt_t action;
4174 unsigned long rt_expires;
4176 write_lock_bh(&ifp->idev->lock);
4177 action = check_cleanup_prefix_route(ifp, &rt_expires);
4178 write_unlock_bh(&ifp->idev->lock);
4180 if (action != CLEANUP_PREFIX_RT_NOP) {
4181 cleanup_prefix_route(ifp, rt_expires,
4182 action == CLEANUP_PREFIX_RT_DEL);
4186 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4187 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4188 valid_lft = prefered_lft = 0;
4189 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4190 !was_managetempaddr, jiffies);
4193 addrconf_verify_rtnl();
4195 return 0;
4198 static int
4199 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4201 struct net *net = sock_net(skb->sk);
4202 struct ifaddrmsg *ifm;
4203 struct nlattr *tb[IFA_MAX+1];
4204 struct in6_addr *pfx, *peer_pfx;
4205 struct inet6_ifaddr *ifa;
4206 struct net_device *dev;
4207 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4208 u32 ifa_flags;
4209 int err;
4211 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4212 if (err < 0)
4213 return err;
4215 ifm = nlmsg_data(nlh);
4216 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4217 if (!pfx)
4218 return -EINVAL;
4220 if (tb[IFA_CACHEINFO]) {
4221 struct ifa_cacheinfo *ci;
4223 ci = nla_data(tb[IFA_CACHEINFO]);
4224 valid_lft = ci->ifa_valid;
4225 preferred_lft = ci->ifa_prefered;
4226 } else {
4227 preferred_lft = INFINITY_LIFE_TIME;
4228 valid_lft = INFINITY_LIFE_TIME;
4231 dev = __dev_get_by_index(net, ifm->ifa_index);
4232 if (!dev)
4233 return -ENODEV;
4235 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4237 /* We ignore other flags so far. */
4238 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4239 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4241 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4242 if (!ifa) {
4244 * It would be best to check for !NLM_F_CREATE here but
4245 * userspace already relies on not having to provide this.
4247 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4248 ifm->ifa_prefixlen, ifa_flags,
4249 preferred_lft, valid_lft);
4252 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4253 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4254 err = -EEXIST;
4255 else
4256 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4258 in6_ifa_put(ifa);
4260 return err;
4263 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4264 u8 scope, int ifindex)
4266 struct ifaddrmsg *ifm;
4268 ifm = nlmsg_data(nlh);
4269 ifm->ifa_family = AF_INET6;
4270 ifm->ifa_prefixlen = prefixlen;
4271 ifm->ifa_flags = flags;
4272 ifm->ifa_scope = scope;
4273 ifm->ifa_index = ifindex;
4276 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4277 unsigned long tstamp, u32 preferred, u32 valid)
4279 struct ifa_cacheinfo ci;
4281 ci.cstamp = cstamp_delta(cstamp);
4282 ci.tstamp = cstamp_delta(tstamp);
4283 ci.ifa_prefered = preferred;
4284 ci.ifa_valid = valid;
4286 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4289 static inline int rt_scope(int ifa_scope)
4291 if (ifa_scope & IFA_HOST)
4292 return RT_SCOPE_HOST;
4293 else if (ifa_scope & IFA_LINK)
4294 return RT_SCOPE_LINK;
4295 else if (ifa_scope & IFA_SITE)
4296 return RT_SCOPE_SITE;
4297 else
4298 return RT_SCOPE_UNIVERSE;
4301 static inline int inet6_ifaddr_msgsize(void)
4303 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4304 + nla_total_size(16) /* IFA_LOCAL */
4305 + nla_total_size(16) /* IFA_ADDRESS */
4306 + nla_total_size(sizeof(struct ifa_cacheinfo))
4307 + nla_total_size(4) /* IFA_FLAGS */;
4310 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4311 u32 portid, u32 seq, int event, unsigned int flags)
4313 struct nlmsghdr *nlh;
4314 u32 preferred, valid;
4316 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4317 if (!nlh)
4318 return -EMSGSIZE;
4320 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4321 ifa->idev->dev->ifindex);
4323 if (!((ifa->flags&IFA_F_PERMANENT) &&
4324 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4325 preferred = ifa->prefered_lft;
4326 valid = ifa->valid_lft;
4327 if (preferred != INFINITY_LIFE_TIME) {
4328 long tval = (jiffies - ifa->tstamp)/HZ;
4329 if (preferred > tval)
4330 preferred -= tval;
4331 else
4332 preferred = 0;
4333 if (valid != INFINITY_LIFE_TIME) {
4334 if (valid > tval)
4335 valid -= tval;
4336 else
4337 valid = 0;
4340 } else {
4341 preferred = INFINITY_LIFE_TIME;
4342 valid = INFINITY_LIFE_TIME;
4345 if (!ipv6_addr_any(&ifa->peer_addr)) {
4346 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4347 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4348 goto error;
4349 } else
4350 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4351 goto error;
4353 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4354 goto error;
4356 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4357 goto error;
4359 nlmsg_end(skb, nlh);
4360 return 0;
4362 error:
4363 nlmsg_cancel(skb, nlh);
4364 return -EMSGSIZE;
4367 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4368 u32 portid, u32 seq, int event, u16 flags)
4370 struct nlmsghdr *nlh;
4371 u8 scope = RT_SCOPE_UNIVERSE;
4372 int ifindex = ifmca->idev->dev->ifindex;
4374 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4375 scope = RT_SCOPE_SITE;
4377 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4378 if (!nlh)
4379 return -EMSGSIZE;
4381 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4382 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4383 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4384 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4385 nlmsg_cancel(skb, nlh);
4386 return -EMSGSIZE;
4389 nlmsg_end(skb, nlh);
4390 return 0;
4393 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4394 u32 portid, u32 seq, int event, unsigned int flags)
4396 struct nlmsghdr *nlh;
4397 u8 scope = RT_SCOPE_UNIVERSE;
4398 int ifindex = ifaca->aca_idev->dev->ifindex;
4400 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4401 scope = RT_SCOPE_SITE;
4403 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4404 if (!nlh)
4405 return -EMSGSIZE;
4407 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4408 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4409 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4410 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4411 nlmsg_cancel(skb, nlh);
4412 return -EMSGSIZE;
4415 nlmsg_end(skb, nlh);
4416 return 0;
4419 enum addr_type_t {
4420 UNICAST_ADDR,
4421 MULTICAST_ADDR,
4422 ANYCAST_ADDR,
4425 /* called with rcu_read_lock() */
4426 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4427 struct netlink_callback *cb, enum addr_type_t type,
4428 int s_ip_idx, int *p_ip_idx)
4430 struct ifmcaddr6 *ifmca;
4431 struct ifacaddr6 *ifaca;
4432 int err = 1;
4433 int ip_idx = *p_ip_idx;
4435 read_lock_bh(&idev->lock);
4436 switch (type) {
4437 case UNICAST_ADDR: {
4438 struct inet6_ifaddr *ifa;
4440 /* unicast address incl. temp addr */
4441 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4442 if (ip_idx < s_ip_idx)
4443 goto next;
4444 err = inet6_fill_ifaddr(skb, ifa,
4445 NETLINK_CB(cb->skb).portid,
4446 cb->nlh->nlmsg_seq,
4447 RTM_NEWADDR,
4448 NLM_F_MULTI);
4449 if (err < 0)
4450 break;
4451 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4452 next:
4453 ip_idx++;
4455 break;
4457 case MULTICAST_ADDR:
4458 /* multicast address */
4459 for (ifmca = idev->mc_list; ifmca;
4460 ifmca = ifmca->next, ip_idx++) {
4461 if (ip_idx < s_ip_idx)
4462 continue;
4463 err = inet6_fill_ifmcaddr(skb, ifmca,
4464 NETLINK_CB(cb->skb).portid,
4465 cb->nlh->nlmsg_seq,
4466 RTM_GETMULTICAST,
4467 NLM_F_MULTI);
4468 if (err < 0)
4469 break;
4471 break;
4472 case ANYCAST_ADDR:
4473 /* anycast address */
4474 for (ifaca = idev->ac_list; ifaca;
4475 ifaca = ifaca->aca_next, ip_idx++) {
4476 if (ip_idx < s_ip_idx)
4477 continue;
4478 err = inet6_fill_ifacaddr(skb, ifaca,
4479 NETLINK_CB(cb->skb).portid,
4480 cb->nlh->nlmsg_seq,
4481 RTM_GETANYCAST,
4482 NLM_F_MULTI);
4483 if (err < 0)
4484 break;
4486 break;
4487 default:
4488 break;
4490 read_unlock_bh(&idev->lock);
4491 *p_ip_idx = ip_idx;
4492 return err;
4495 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4496 enum addr_type_t type)
4498 struct net *net = sock_net(skb->sk);
4499 int h, s_h;
4500 int idx, ip_idx;
4501 int s_idx, s_ip_idx;
4502 struct net_device *dev;
4503 struct inet6_dev *idev;
4504 struct hlist_head *head;
4506 s_h = cb->args[0];
4507 s_idx = idx = cb->args[1];
4508 s_ip_idx = ip_idx = cb->args[2];
4510 rcu_read_lock();
4511 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4512 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4513 idx = 0;
4514 head = &net->dev_index_head[h];
4515 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4516 if (idx < s_idx)
4517 goto cont;
4518 if (h > s_h || idx > s_idx)
4519 s_ip_idx = 0;
4520 ip_idx = 0;
4521 idev = __in6_dev_get(dev);
4522 if (!idev)
4523 goto cont;
4525 if (in6_dump_addrs(idev, skb, cb, type,
4526 s_ip_idx, &ip_idx) < 0)
4527 goto done;
4528 cont:
4529 idx++;
4532 done:
4533 rcu_read_unlock();
4534 cb->args[0] = h;
4535 cb->args[1] = idx;
4536 cb->args[2] = ip_idx;
4538 return skb->len;
4541 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4543 enum addr_type_t type = UNICAST_ADDR;
4545 return inet6_dump_addr(skb, cb, type);
4548 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4550 enum addr_type_t type = MULTICAST_ADDR;
4552 return inet6_dump_addr(skb, cb, type);
4556 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4558 enum addr_type_t type = ANYCAST_ADDR;
4560 return inet6_dump_addr(skb, cb, type);
4563 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4565 struct net *net = sock_net(in_skb->sk);
4566 struct ifaddrmsg *ifm;
4567 struct nlattr *tb[IFA_MAX+1];
4568 struct in6_addr *addr = NULL, *peer;
4569 struct net_device *dev = NULL;
4570 struct inet6_ifaddr *ifa;
4571 struct sk_buff *skb;
4572 int err;
4574 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4575 if (err < 0)
4576 goto errout;
4578 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4579 if (!addr) {
4580 err = -EINVAL;
4581 goto errout;
4584 ifm = nlmsg_data(nlh);
4585 if (ifm->ifa_index)
4586 dev = __dev_get_by_index(net, ifm->ifa_index);
4588 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4589 if (!ifa) {
4590 err = -EADDRNOTAVAIL;
4591 goto errout;
4594 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4595 if (!skb) {
4596 err = -ENOBUFS;
4597 goto errout_ifa;
4600 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4601 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4602 if (err < 0) {
4603 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4604 WARN_ON(err == -EMSGSIZE);
4605 kfree_skb(skb);
4606 goto errout_ifa;
4608 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4609 errout_ifa:
4610 in6_ifa_put(ifa);
4611 errout:
4612 return err;
4615 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4617 struct sk_buff *skb;
4618 struct net *net = dev_net(ifa->idev->dev);
4619 int err = -ENOBUFS;
4621 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4622 if (!skb)
4623 goto errout;
4625 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4626 if (err < 0) {
4627 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4628 WARN_ON(err == -EMSGSIZE);
4629 kfree_skb(skb);
4630 goto errout;
4632 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4633 return;
4634 errout:
4635 if (err < 0)
4636 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4639 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4640 __s32 *array, int bytes)
4642 BUG_ON(bytes < (DEVCONF_MAX * 4));
4644 memset(array, 0, bytes);
4645 array[DEVCONF_FORWARDING] = cnf->forwarding;
4646 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4647 array[DEVCONF_MTU6] = cnf->mtu6;
4648 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4649 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4650 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4651 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4652 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4653 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4654 jiffies_to_msecs(cnf->rtr_solicit_interval);
4655 array[DEVCONF_RTR_SOLICIT_DELAY] =
4656 jiffies_to_msecs(cnf->rtr_solicit_delay);
4657 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4658 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4659 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4660 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4661 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4662 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4663 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4664 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4665 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4666 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4667 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4668 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4669 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4670 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4671 #ifdef CONFIG_IPV6_ROUTER_PREF
4672 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4673 array[DEVCONF_RTR_PROBE_INTERVAL] =
4674 jiffies_to_msecs(cnf->rtr_probe_interval);
4675 #ifdef CONFIG_IPV6_ROUTE_INFO
4676 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4677 #endif
4678 #endif
4679 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4680 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4681 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4682 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4683 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4684 #endif
4685 #ifdef CONFIG_IPV6_MROUTE
4686 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4687 #endif
4688 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4689 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4690 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4691 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4692 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4693 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4694 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4695 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4696 /* we omit DEVCONF_STABLE_SECRET for now */
4697 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4700 static inline size_t inet6_ifla6_size(void)
4702 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4703 + nla_total_size(sizeof(struct ifla_cacheinfo))
4704 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4705 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4706 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4707 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4710 static inline size_t inet6_if_nlmsg_size(void)
4712 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4713 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4714 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4715 + nla_total_size(4) /* IFLA_MTU */
4716 + nla_total_size(4) /* IFLA_LINK */
4717 + nla_total_size(1) /* IFLA_OPERSTATE */
4718 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4721 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4722 int items, int bytes)
4724 int i;
4725 int pad = bytes - sizeof(u64) * items;
4726 BUG_ON(pad < 0);
4728 /* Use put_unaligned() because stats may not be aligned for u64. */
4729 put_unaligned(items, &stats[0]);
4730 for (i = 1; i < items; i++)
4731 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4733 memset(&stats[items], 0, pad);
4736 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4737 int bytes, size_t syncpoff)
4739 int i, c;
4740 u64 buff[IPSTATS_MIB_MAX];
4741 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
4743 BUG_ON(pad < 0);
4745 memset(buff, 0, sizeof(buff));
4746 buff[0] = IPSTATS_MIB_MAX;
4748 for_each_possible_cpu(c) {
4749 for (i = 1; i < IPSTATS_MIB_MAX; i++)
4750 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
4753 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
4754 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
4757 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4758 int bytes)
4760 switch (attrtype) {
4761 case IFLA_INET6_STATS:
4762 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
4763 offsetof(struct ipstats_mib, syncp));
4764 break;
4765 case IFLA_INET6_ICMP6STATS:
4766 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4767 break;
4771 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
4772 u32 ext_filter_mask)
4774 struct nlattr *nla;
4775 struct ifla_cacheinfo ci;
4777 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4778 goto nla_put_failure;
4779 ci.max_reasm_len = IPV6_MAXPLEN;
4780 ci.tstamp = cstamp_delta(idev->tstamp);
4781 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4782 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4783 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4784 goto nla_put_failure;
4785 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4786 if (!nla)
4787 goto nla_put_failure;
4788 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4790 /* XXX - MC not implemented */
4792 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
4793 return 0;
4795 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4796 if (!nla)
4797 goto nla_put_failure;
4798 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4800 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4801 if (!nla)
4802 goto nla_put_failure;
4803 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4805 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4806 if (!nla)
4807 goto nla_put_failure;
4809 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4810 goto nla_put_failure;
4812 read_lock_bh(&idev->lock);
4813 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4814 read_unlock_bh(&idev->lock);
4816 return 0;
4818 nla_put_failure:
4819 return -EMSGSIZE;
4822 static size_t inet6_get_link_af_size(const struct net_device *dev,
4823 u32 ext_filter_mask)
4825 if (!__in6_dev_get(dev))
4826 return 0;
4828 return inet6_ifla6_size();
4831 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
4832 u32 ext_filter_mask)
4834 struct inet6_dev *idev = __in6_dev_get(dev);
4836 if (!idev)
4837 return -ENODATA;
4839 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
4840 return -EMSGSIZE;
4842 return 0;
4845 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4847 struct inet6_ifaddr *ifp;
4848 struct net_device *dev = idev->dev;
4849 bool update_rs = false;
4850 struct in6_addr ll_addr;
4852 ASSERT_RTNL();
4854 if (!token)
4855 return -EINVAL;
4856 if (ipv6_addr_any(token))
4857 return -EINVAL;
4858 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4859 return -EINVAL;
4860 if (!ipv6_accept_ra(idev))
4861 return -EINVAL;
4862 if (idev->cnf.rtr_solicits <= 0)
4863 return -EINVAL;
4865 write_lock_bh(&idev->lock);
4867 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4868 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4870 write_unlock_bh(&idev->lock);
4872 if (!idev->dead && (idev->if_flags & IF_READY) &&
4873 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4874 IFA_F_OPTIMISTIC)) {
4876 /* If we're not ready, then normal ifup will take care
4877 * of this. Otherwise, we need to request our rs here.
4879 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4880 update_rs = true;
4883 write_lock_bh(&idev->lock);
4885 if (update_rs) {
4886 idev->if_flags |= IF_RS_SENT;
4887 idev->rs_probes = 1;
4888 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4891 /* Well, that's kinda nasty ... */
4892 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4893 spin_lock(&ifp->lock);
4894 if (ifp->tokenized) {
4895 ifp->valid_lft = 0;
4896 ifp->prefered_lft = 0;
4898 spin_unlock(&ifp->lock);
4901 write_unlock_bh(&idev->lock);
4902 inet6_ifinfo_notify(RTM_NEWLINK, idev);
4903 addrconf_verify_rtnl();
4904 return 0;
4907 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
4908 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
4909 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
4912 static int inet6_validate_link_af(const struct net_device *dev,
4913 const struct nlattr *nla)
4915 struct nlattr *tb[IFLA_INET6_MAX + 1];
4917 if (dev && !__in6_dev_get(dev))
4918 return -EAFNOSUPPORT;
4920 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
4923 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4925 int err = -EINVAL;
4926 struct inet6_dev *idev = __in6_dev_get(dev);
4927 struct nlattr *tb[IFLA_INET6_MAX + 1];
4929 if (!idev)
4930 return -EAFNOSUPPORT;
4932 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4933 BUG();
4935 if (tb[IFLA_INET6_TOKEN]) {
4936 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4937 if (err)
4938 return err;
4941 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4942 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4944 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4945 mode != IN6_ADDR_GEN_MODE_NONE &&
4946 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY)
4947 return -EINVAL;
4949 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
4950 !idev->cnf.stable_secret.initialized &&
4951 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
4952 return -EINVAL;
4954 idev->addr_gen_mode = mode;
4955 err = 0;
4958 return err;
4961 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4962 u32 portid, u32 seq, int event, unsigned int flags)
4964 struct net_device *dev = idev->dev;
4965 struct ifinfomsg *hdr;
4966 struct nlmsghdr *nlh;
4967 void *protoinfo;
4969 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4970 if (!nlh)
4971 return -EMSGSIZE;
4973 hdr = nlmsg_data(nlh);
4974 hdr->ifi_family = AF_INET6;
4975 hdr->__ifi_pad = 0;
4976 hdr->ifi_type = dev->type;
4977 hdr->ifi_index = dev->ifindex;
4978 hdr->ifi_flags = dev_get_flags(dev);
4979 hdr->ifi_change = 0;
4981 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4982 (dev->addr_len &&
4983 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4984 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4985 (dev->ifindex != dev_get_iflink(dev) &&
4986 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
4987 nla_put_u8(skb, IFLA_OPERSTATE,
4988 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
4989 goto nla_put_failure;
4990 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4991 if (!protoinfo)
4992 goto nla_put_failure;
4994 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
4995 goto nla_put_failure;
4997 nla_nest_end(skb, protoinfo);
4998 nlmsg_end(skb, nlh);
4999 return 0;
5001 nla_put_failure:
5002 nlmsg_cancel(skb, nlh);
5003 return -EMSGSIZE;
5006 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5008 struct net *net = sock_net(skb->sk);
5009 int h, s_h;
5010 int idx = 0, s_idx;
5011 struct net_device *dev;
5012 struct inet6_dev *idev;
5013 struct hlist_head *head;
5015 s_h = cb->args[0];
5016 s_idx = cb->args[1];
5018 rcu_read_lock();
5019 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5020 idx = 0;
5021 head = &net->dev_index_head[h];
5022 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5023 if (idx < s_idx)
5024 goto cont;
5025 idev = __in6_dev_get(dev);
5026 if (!idev)
5027 goto cont;
5028 if (inet6_fill_ifinfo(skb, idev,
5029 NETLINK_CB(cb->skb).portid,
5030 cb->nlh->nlmsg_seq,
5031 RTM_NEWLINK, NLM_F_MULTI) < 0)
5032 goto out;
5033 cont:
5034 idx++;
5037 out:
5038 rcu_read_unlock();
5039 cb->args[1] = idx;
5040 cb->args[0] = h;
5042 return skb->len;
5045 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5047 struct sk_buff *skb;
5048 struct net *net = dev_net(idev->dev);
5049 int err = -ENOBUFS;
5051 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5052 if (!skb)
5053 goto errout;
5055 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5056 if (err < 0) {
5057 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5058 WARN_ON(err == -EMSGSIZE);
5059 kfree_skb(skb);
5060 goto errout;
5062 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5063 return;
5064 errout:
5065 if (err < 0)
5066 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5069 static inline size_t inet6_prefix_nlmsg_size(void)
5071 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5072 + nla_total_size(sizeof(struct in6_addr))
5073 + nla_total_size(sizeof(struct prefix_cacheinfo));
5076 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5077 struct prefix_info *pinfo, u32 portid, u32 seq,
5078 int event, unsigned int flags)
5080 struct prefixmsg *pmsg;
5081 struct nlmsghdr *nlh;
5082 struct prefix_cacheinfo ci;
5084 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5085 if (!nlh)
5086 return -EMSGSIZE;
5088 pmsg = nlmsg_data(nlh);
5089 pmsg->prefix_family = AF_INET6;
5090 pmsg->prefix_pad1 = 0;
5091 pmsg->prefix_pad2 = 0;
5092 pmsg->prefix_ifindex = idev->dev->ifindex;
5093 pmsg->prefix_len = pinfo->prefix_len;
5094 pmsg->prefix_type = pinfo->type;
5095 pmsg->prefix_pad3 = 0;
5096 pmsg->prefix_flags = 0;
5097 if (pinfo->onlink)
5098 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5099 if (pinfo->autoconf)
5100 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5102 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5103 goto nla_put_failure;
5104 ci.preferred_time = ntohl(pinfo->prefered);
5105 ci.valid_time = ntohl(pinfo->valid);
5106 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5107 goto nla_put_failure;
5108 nlmsg_end(skb, nlh);
5109 return 0;
5111 nla_put_failure:
5112 nlmsg_cancel(skb, nlh);
5113 return -EMSGSIZE;
5116 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5117 struct prefix_info *pinfo)
5119 struct sk_buff *skb;
5120 struct net *net = dev_net(idev->dev);
5121 int err = -ENOBUFS;
5123 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5124 if (!skb)
5125 goto errout;
5127 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5128 if (err < 0) {
5129 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5130 WARN_ON(err == -EMSGSIZE);
5131 kfree_skb(skb);
5132 goto errout;
5134 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5135 return;
5136 errout:
5137 if (err < 0)
5138 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5141 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5143 struct net *net = dev_net(ifp->idev->dev);
5145 if (event)
5146 ASSERT_RTNL();
5148 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5150 switch (event) {
5151 case RTM_NEWADDR:
5153 * If the address was optimistic
5154 * we inserted the route at the start of
5155 * our DAD process, so we don't need
5156 * to do it again
5158 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5159 ip6_ins_rt(ifp->rt);
5160 if (ifp->idev->cnf.forwarding)
5161 addrconf_join_anycast(ifp);
5162 if (!ipv6_addr_any(&ifp->peer_addr))
5163 addrconf_prefix_route(&ifp->peer_addr, 128,
5164 ifp->idev->dev, 0, 0);
5165 break;
5166 case RTM_DELADDR:
5167 if (ifp->idev->cnf.forwarding)
5168 addrconf_leave_anycast(ifp);
5169 addrconf_leave_solict(ifp->idev, &ifp->addr);
5170 if (!ipv6_addr_any(&ifp->peer_addr)) {
5171 struct rt6_info *rt;
5173 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5174 ifp->idev->dev, 0, 0);
5175 if (rt)
5176 ip6_del_rt(rt);
5178 dst_hold(&ifp->rt->dst);
5180 ip6_del_rt(ifp->rt);
5182 rt_genid_bump_ipv6(net);
5183 break;
5185 atomic_inc(&net->ipv6.dev_addr_genid);
5188 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5190 rcu_read_lock_bh();
5191 if (likely(ifp->idev->dead == 0))
5192 __ipv6_ifa_notify(event, ifp);
5193 rcu_read_unlock_bh();
5196 #ifdef CONFIG_SYSCTL
5198 static
5199 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5200 void __user *buffer, size_t *lenp, loff_t *ppos)
5202 int *valp = ctl->data;
5203 int val = *valp;
5204 loff_t pos = *ppos;
5205 struct ctl_table lctl;
5206 int ret;
5209 * ctl->data points to idev->cnf.forwarding, we should
5210 * not modify it until we get the rtnl lock.
5212 lctl = *ctl;
5213 lctl.data = &val;
5215 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5217 if (write)
5218 ret = addrconf_fixup_forwarding(ctl, valp, val);
5219 if (ret)
5220 *ppos = pos;
5221 return ret;
5224 static
5225 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5226 void __user *buffer, size_t *lenp, loff_t *ppos)
5228 struct inet6_dev *idev = ctl->extra1;
5229 int min_mtu = IPV6_MIN_MTU;
5230 struct ctl_table lctl;
5232 lctl = *ctl;
5233 lctl.extra1 = &min_mtu;
5234 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5236 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5239 static void dev_disable_change(struct inet6_dev *idev)
5241 struct netdev_notifier_info info;
5243 if (!idev || !idev->dev)
5244 return;
5246 netdev_notifier_info_init(&info, idev->dev);
5247 if (idev->cnf.disable_ipv6)
5248 addrconf_notify(NULL, NETDEV_DOWN, &info);
5249 else
5250 addrconf_notify(NULL, NETDEV_UP, &info);
5253 static void addrconf_disable_change(struct net *net, __s32 newf)
5255 struct net_device *dev;
5256 struct inet6_dev *idev;
5258 for_each_netdev(net, dev) {
5259 idev = __in6_dev_get(dev);
5260 if (idev) {
5261 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5262 idev->cnf.disable_ipv6 = newf;
5263 if (changed)
5264 dev_disable_change(idev);
5269 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5271 struct net *net;
5272 int old;
5274 if (!rtnl_trylock())
5275 return restart_syscall();
5277 net = (struct net *)table->extra2;
5278 old = *p;
5279 *p = newf;
5281 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5282 rtnl_unlock();
5283 return 0;
5286 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5287 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5288 addrconf_disable_change(net, newf);
5289 } else if ((!newf) ^ (!old))
5290 dev_disable_change((struct inet6_dev *)table->extra1);
5292 rtnl_unlock();
5293 return 0;
5296 static
5297 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5298 void __user *buffer, size_t *lenp, loff_t *ppos)
5300 int *valp = ctl->data;
5301 int val = *valp;
5302 loff_t pos = *ppos;
5303 struct ctl_table lctl;
5304 int ret;
5307 * ctl->data points to idev->cnf.disable_ipv6, we should
5308 * not modify it until we get the rtnl lock.
5310 lctl = *ctl;
5311 lctl.data = &val;
5313 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5315 if (write)
5316 ret = addrconf_disable_ipv6(ctl, valp, val);
5317 if (ret)
5318 *ppos = pos;
5319 return ret;
5322 static
5323 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5324 void __user *buffer, size_t *lenp, loff_t *ppos)
5326 int *valp = ctl->data;
5327 int ret;
5328 int old, new;
5330 old = *valp;
5331 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5332 new = *valp;
5334 if (write && old != new) {
5335 struct net *net = ctl->extra2;
5337 if (!rtnl_trylock())
5338 return restart_syscall();
5340 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5341 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5342 NETCONFA_IFINDEX_DEFAULT,
5343 net->ipv6.devconf_dflt);
5344 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5345 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5346 NETCONFA_IFINDEX_ALL,
5347 net->ipv6.devconf_all);
5348 else {
5349 struct inet6_dev *idev = ctl->extra1;
5351 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5352 idev->dev->ifindex,
5353 &idev->cnf);
5355 rtnl_unlock();
5358 return ret;
5361 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5362 void __user *buffer, size_t *lenp,
5363 loff_t *ppos)
5365 int err;
5366 struct in6_addr addr;
5367 char str[IPV6_MAX_STRLEN];
5368 struct ctl_table lctl = *ctl;
5369 struct net *net = ctl->extra2;
5370 struct ipv6_stable_secret *secret = ctl->data;
5372 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5373 return -EIO;
5375 lctl.maxlen = IPV6_MAX_STRLEN;
5376 lctl.data = str;
5378 if (!rtnl_trylock())
5379 return restart_syscall();
5381 if (!write && !secret->initialized) {
5382 err = -EIO;
5383 goto out;
5386 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5387 if (err >= sizeof(str)) {
5388 err = -EIO;
5389 goto out;
5392 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5393 if (err || !write)
5394 goto out;
5396 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5397 err = -EIO;
5398 goto out;
5401 secret->initialized = true;
5402 secret->secret = addr;
5404 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5405 struct net_device *dev;
5407 for_each_netdev(net, dev) {
5408 struct inet6_dev *idev = __in6_dev_get(dev);
5410 if (idev) {
5411 idev->addr_gen_mode =
5412 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5415 } else {
5416 struct inet6_dev *idev = ctl->extra1;
5418 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5421 out:
5422 rtnl_unlock();
5424 return err;
5427 static
5428 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5429 int write,
5430 void __user *buffer,
5431 size_t *lenp,
5432 loff_t *ppos)
5434 int *valp = ctl->data;
5435 int val = *valp;
5436 loff_t pos = *ppos;
5437 struct ctl_table lctl;
5438 int ret;
5440 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5441 * we should not modify it until we get the rtnl lock.
5443 lctl = *ctl;
5444 lctl.data = &val;
5446 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5448 if (write)
5449 ret = addrconf_fixup_linkdown(ctl, valp, val);
5450 if (ret)
5451 *ppos = pos;
5452 return ret;
5455 static struct addrconf_sysctl_table
5457 struct ctl_table_header *sysctl_header;
5458 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
5459 } addrconf_sysctl __read_mostly = {
5460 .sysctl_header = NULL,
5461 .addrconf_vars = {
5463 .procname = "forwarding",
5464 .data = &ipv6_devconf.forwarding,
5465 .maxlen = sizeof(int),
5466 .mode = 0644,
5467 .proc_handler = addrconf_sysctl_forward,
5470 .procname = "hop_limit",
5471 .data = &ipv6_devconf.hop_limit,
5472 .maxlen = sizeof(int),
5473 .mode = 0644,
5474 .proc_handler = proc_dointvec,
5477 .procname = "mtu",
5478 .data = &ipv6_devconf.mtu6,
5479 .maxlen = sizeof(int),
5480 .mode = 0644,
5481 .proc_handler = addrconf_sysctl_mtu,
5484 .procname = "accept_ra",
5485 .data = &ipv6_devconf.accept_ra,
5486 .maxlen = sizeof(int),
5487 .mode = 0644,
5488 .proc_handler = proc_dointvec,
5491 .procname = "accept_redirects",
5492 .data = &ipv6_devconf.accept_redirects,
5493 .maxlen = sizeof(int),
5494 .mode = 0644,
5495 .proc_handler = proc_dointvec,
5498 .procname = "autoconf",
5499 .data = &ipv6_devconf.autoconf,
5500 .maxlen = sizeof(int),
5501 .mode = 0644,
5502 .proc_handler = proc_dointvec,
5505 .procname = "dad_transmits",
5506 .data = &ipv6_devconf.dad_transmits,
5507 .maxlen = sizeof(int),
5508 .mode = 0644,
5509 .proc_handler = proc_dointvec,
5512 .procname = "router_solicitations",
5513 .data = &ipv6_devconf.rtr_solicits,
5514 .maxlen = sizeof(int),
5515 .mode = 0644,
5516 .proc_handler = proc_dointvec,
5519 .procname = "router_solicitation_interval",
5520 .data = &ipv6_devconf.rtr_solicit_interval,
5521 .maxlen = sizeof(int),
5522 .mode = 0644,
5523 .proc_handler = proc_dointvec_jiffies,
5526 .procname = "router_solicitation_delay",
5527 .data = &ipv6_devconf.rtr_solicit_delay,
5528 .maxlen = sizeof(int),
5529 .mode = 0644,
5530 .proc_handler = proc_dointvec_jiffies,
5533 .procname = "force_mld_version",
5534 .data = &ipv6_devconf.force_mld_version,
5535 .maxlen = sizeof(int),
5536 .mode = 0644,
5537 .proc_handler = proc_dointvec,
5540 .procname = "mldv1_unsolicited_report_interval",
5541 .data =
5542 &ipv6_devconf.mldv1_unsolicited_report_interval,
5543 .maxlen = sizeof(int),
5544 .mode = 0644,
5545 .proc_handler = proc_dointvec_ms_jiffies,
5548 .procname = "mldv2_unsolicited_report_interval",
5549 .data =
5550 &ipv6_devconf.mldv2_unsolicited_report_interval,
5551 .maxlen = sizeof(int),
5552 .mode = 0644,
5553 .proc_handler = proc_dointvec_ms_jiffies,
5556 .procname = "use_tempaddr",
5557 .data = &ipv6_devconf.use_tempaddr,
5558 .maxlen = sizeof(int),
5559 .mode = 0644,
5560 .proc_handler = proc_dointvec,
5563 .procname = "temp_valid_lft",
5564 .data = &ipv6_devconf.temp_valid_lft,
5565 .maxlen = sizeof(int),
5566 .mode = 0644,
5567 .proc_handler = proc_dointvec,
5570 .procname = "temp_prefered_lft",
5571 .data = &ipv6_devconf.temp_prefered_lft,
5572 .maxlen = sizeof(int),
5573 .mode = 0644,
5574 .proc_handler = proc_dointvec,
5577 .procname = "regen_max_retry",
5578 .data = &ipv6_devconf.regen_max_retry,
5579 .maxlen = sizeof(int),
5580 .mode = 0644,
5581 .proc_handler = proc_dointvec,
5584 .procname = "max_desync_factor",
5585 .data = &ipv6_devconf.max_desync_factor,
5586 .maxlen = sizeof(int),
5587 .mode = 0644,
5588 .proc_handler = proc_dointvec,
5591 .procname = "max_addresses",
5592 .data = &ipv6_devconf.max_addresses,
5593 .maxlen = sizeof(int),
5594 .mode = 0644,
5595 .proc_handler = proc_dointvec,
5598 .procname = "accept_ra_defrtr",
5599 .data = &ipv6_devconf.accept_ra_defrtr,
5600 .maxlen = sizeof(int),
5601 .mode = 0644,
5602 .proc_handler = proc_dointvec,
5605 .procname = "accept_ra_min_hop_limit",
5606 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5607 .maxlen = sizeof(int),
5608 .mode = 0644,
5609 .proc_handler = proc_dointvec,
5612 .procname = "accept_ra_pinfo",
5613 .data = &ipv6_devconf.accept_ra_pinfo,
5614 .maxlen = sizeof(int),
5615 .mode = 0644,
5616 .proc_handler = proc_dointvec,
5618 #ifdef CONFIG_IPV6_ROUTER_PREF
5620 .procname = "accept_ra_rtr_pref",
5621 .data = &ipv6_devconf.accept_ra_rtr_pref,
5622 .maxlen = sizeof(int),
5623 .mode = 0644,
5624 .proc_handler = proc_dointvec,
5627 .procname = "router_probe_interval",
5628 .data = &ipv6_devconf.rtr_probe_interval,
5629 .maxlen = sizeof(int),
5630 .mode = 0644,
5631 .proc_handler = proc_dointvec_jiffies,
5633 #ifdef CONFIG_IPV6_ROUTE_INFO
5635 .procname = "accept_ra_rt_info_max_plen",
5636 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5637 .maxlen = sizeof(int),
5638 .mode = 0644,
5639 .proc_handler = proc_dointvec,
5641 #endif
5642 #endif
5644 .procname = "proxy_ndp",
5645 .data = &ipv6_devconf.proxy_ndp,
5646 .maxlen = sizeof(int),
5647 .mode = 0644,
5648 .proc_handler = addrconf_sysctl_proxy_ndp,
5651 .procname = "accept_source_route",
5652 .data = &ipv6_devconf.accept_source_route,
5653 .maxlen = sizeof(int),
5654 .mode = 0644,
5655 .proc_handler = proc_dointvec,
5657 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5659 .procname = "optimistic_dad",
5660 .data = &ipv6_devconf.optimistic_dad,
5661 .maxlen = sizeof(int),
5662 .mode = 0644,
5663 .proc_handler = proc_dointvec,
5667 .procname = "use_optimistic",
5668 .data = &ipv6_devconf.use_optimistic,
5669 .maxlen = sizeof(int),
5670 .mode = 0644,
5671 .proc_handler = proc_dointvec,
5674 #endif
5675 #ifdef CONFIG_IPV6_MROUTE
5677 .procname = "mc_forwarding",
5678 .data = &ipv6_devconf.mc_forwarding,
5679 .maxlen = sizeof(int),
5680 .mode = 0444,
5681 .proc_handler = proc_dointvec,
5683 #endif
5685 .procname = "disable_ipv6",
5686 .data = &ipv6_devconf.disable_ipv6,
5687 .maxlen = sizeof(int),
5688 .mode = 0644,
5689 .proc_handler = addrconf_sysctl_disable,
5692 .procname = "accept_dad",
5693 .data = &ipv6_devconf.accept_dad,
5694 .maxlen = sizeof(int),
5695 .mode = 0644,
5696 .proc_handler = proc_dointvec,
5699 .procname = "force_tllao",
5700 .data = &ipv6_devconf.force_tllao,
5701 .maxlen = sizeof(int),
5702 .mode = 0644,
5703 .proc_handler = proc_dointvec
5706 .procname = "ndisc_notify",
5707 .data = &ipv6_devconf.ndisc_notify,
5708 .maxlen = sizeof(int),
5709 .mode = 0644,
5710 .proc_handler = proc_dointvec
5713 .procname = "suppress_frag_ndisc",
5714 .data = &ipv6_devconf.suppress_frag_ndisc,
5715 .maxlen = sizeof(int),
5716 .mode = 0644,
5717 .proc_handler = proc_dointvec
5720 .procname = "accept_ra_from_local",
5721 .data = &ipv6_devconf.accept_ra_from_local,
5722 .maxlen = sizeof(int),
5723 .mode = 0644,
5724 .proc_handler = proc_dointvec,
5727 .procname = "accept_ra_mtu",
5728 .data = &ipv6_devconf.accept_ra_mtu,
5729 .maxlen = sizeof(int),
5730 .mode = 0644,
5731 .proc_handler = proc_dointvec,
5734 .procname = "stable_secret",
5735 .data = &ipv6_devconf.stable_secret,
5736 .maxlen = IPV6_MAX_STRLEN,
5737 .mode = 0600,
5738 .proc_handler = addrconf_sysctl_stable_secret,
5741 .procname = "use_oif_addrs_only",
5742 .data = &ipv6_devconf.use_oif_addrs_only,
5743 .maxlen = sizeof(int),
5744 .mode = 0644,
5745 .proc_handler = proc_dointvec,
5748 .procname = "ignore_routes_with_linkdown",
5749 .data = &ipv6_devconf.ignore_routes_with_linkdown,
5750 .maxlen = sizeof(int),
5751 .mode = 0644,
5752 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
5755 /* sentinel */
5760 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5761 struct inet6_dev *idev, struct ipv6_devconf *p)
5763 int i;
5764 struct addrconf_sysctl_table *t;
5765 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5767 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5768 if (!t)
5769 goto out;
5771 for (i = 0; t->addrconf_vars[i].data; i++) {
5772 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5773 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5774 t->addrconf_vars[i].extra2 = net;
5777 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5779 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5780 if (!t->sysctl_header)
5781 goto free;
5783 p->sysctl = t;
5784 return 0;
5786 free:
5787 kfree(t);
5788 out:
5789 return -ENOBUFS;
5792 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5794 struct addrconf_sysctl_table *t;
5796 if (!p->sysctl)
5797 return;
5799 t = p->sysctl;
5800 p->sysctl = NULL;
5801 unregister_net_sysctl_table(t->sysctl_header);
5802 kfree(t);
5805 static int addrconf_sysctl_register(struct inet6_dev *idev)
5807 int err;
5809 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5810 return -EINVAL;
5812 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5813 &ndisc_ifinfo_sysctl_change);
5814 if (err)
5815 return err;
5816 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5817 idev, &idev->cnf);
5818 if (err)
5819 neigh_sysctl_unregister(idev->nd_parms);
5821 return err;
5824 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5826 __addrconf_sysctl_unregister(&idev->cnf);
5827 neigh_sysctl_unregister(idev->nd_parms);
5831 #endif
5833 static int __net_init addrconf_init_net(struct net *net)
5835 int err = -ENOMEM;
5836 struct ipv6_devconf *all, *dflt;
5838 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5839 if (!all)
5840 goto err_alloc_all;
5842 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5843 if (!dflt)
5844 goto err_alloc_dflt;
5846 /* these will be inherited by all namespaces */
5847 dflt->autoconf = ipv6_defaults.autoconf;
5848 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5850 dflt->stable_secret.initialized = false;
5851 all->stable_secret.initialized = false;
5853 net->ipv6.devconf_all = all;
5854 net->ipv6.devconf_dflt = dflt;
5856 #ifdef CONFIG_SYSCTL
5857 err = __addrconf_sysctl_register(net, "all", NULL, all);
5858 if (err < 0)
5859 goto err_reg_all;
5861 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5862 if (err < 0)
5863 goto err_reg_dflt;
5864 #endif
5865 return 0;
5867 #ifdef CONFIG_SYSCTL
5868 err_reg_dflt:
5869 __addrconf_sysctl_unregister(all);
5870 err_reg_all:
5871 kfree(dflt);
5872 #endif
5873 err_alloc_dflt:
5874 kfree(all);
5875 err_alloc_all:
5876 return err;
5879 static void __net_exit addrconf_exit_net(struct net *net)
5881 #ifdef CONFIG_SYSCTL
5882 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5883 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5884 #endif
5885 kfree(net->ipv6.devconf_dflt);
5886 kfree(net->ipv6.devconf_all);
5889 static struct pernet_operations addrconf_ops = {
5890 .init = addrconf_init_net,
5891 .exit = addrconf_exit_net,
5894 static struct rtnl_af_ops inet6_ops __read_mostly = {
5895 .family = AF_INET6,
5896 .fill_link_af = inet6_fill_link_af,
5897 .get_link_af_size = inet6_get_link_af_size,
5898 .validate_link_af = inet6_validate_link_af,
5899 .set_link_af = inet6_set_link_af,
5903 * Init / cleanup code
5906 int __init addrconf_init(void)
5908 struct inet6_dev *idev;
5909 int i, err;
5911 err = ipv6_addr_label_init();
5912 if (err < 0) {
5913 pr_crit("%s: cannot initialize default policy table: %d\n",
5914 __func__, err);
5915 goto out;
5918 err = register_pernet_subsys(&addrconf_ops);
5919 if (err < 0)
5920 goto out_addrlabel;
5922 addrconf_wq = create_workqueue("ipv6_addrconf");
5923 if (!addrconf_wq) {
5924 err = -ENOMEM;
5925 goto out_nowq;
5928 /* The addrconf netdev notifier requires that loopback_dev
5929 * has it's ipv6 private information allocated and setup
5930 * before it can bring up and give link-local addresses
5931 * to other devices which are up.
5933 * Unfortunately, loopback_dev is not necessarily the first
5934 * entry in the global dev_base list of net devices. In fact,
5935 * it is likely to be the very last entry on that list.
5936 * So this causes the notifier registry below to try and
5937 * give link-local addresses to all devices besides loopback_dev
5938 * first, then loopback_dev, which cases all the non-loopback_dev
5939 * devices to fail to get a link-local address.
5941 * So, as a temporary fix, allocate the ipv6 structure for
5942 * loopback_dev first by hand.
5943 * Longer term, all of the dependencies ipv6 has upon the loopback
5944 * device and it being up should be removed.
5946 rtnl_lock();
5947 idev = ipv6_add_dev(init_net.loopback_dev);
5948 rtnl_unlock();
5949 if (IS_ERR(idev)) {
5950 err = PTR_ERR(idev);
5951 goto errlo;
5954 ip6_route_init_special_entries();
5956 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5957 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5959 register_netdevice_notifier(&ipv6_dev_notf);
5961 addrconf_verify();
5963 rtnl_af_register(&inet6_ops);
5965 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5966 NULL);
5967 if (err < 0)
5968 goto errout;
5970 /* Only the first call to __rtnl_register can fail */
5971 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5972 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5973 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5974 inet6_dump_ifaddr, NULL);
5975 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5976 inet6_dump_ifmcaddr, NULL);
5977 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5978 inet6_dump_ifacaddr, NULL);
5979 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5980 inet6_netconf_dump_devconf, NULL);
5982 ipv6_addr_label_rtnl_register();
5984 return 0;
5985 errout:
5986 rtnl_af_unregister(&inet6_ops);
5987 unregister_netdevice_notifier(&ipv6_dev_notf);
5988 errlo:
5989 destroy_workqueue(addrconf_wq);
5990 out_nowq:
5991 unregister_pernet_subsys(&addrconf_ops);
5992 out_addrlabel:
5993 ipv6_addr_label_cleanup();
5994 out:
5995 return err;
5998 void addrconf_cleanup(void)
6000 struct net_device *dev;
6001 int i;
6003 unregister_netdevice_notifier(&ipv6_dev_notf);
6004 unregister_pernet_subsys(&addrconf_ops);
6005 ipv6_addr_label_cleanup();
6007 rtnl_lock();
6009 __rtnl_af_unregister(&inet6_ops);
6011 /* clean dev list */
6012 for_each_netdev(&init_net, dev) {
6013 if (__in6_dev_get(dev) == NULL)
6014 continue;
6015 addrconf_ifdown(dev, 1);
6017 addrconf_ifdown(init_net.loopback_dev, 2);
6020 * Check hash table.
6022 spin_lock_bh(&addrconf_hash_lock);
6023 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6024 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6025 spin_unlock_bh(&addrconf_hash_lock);
6026 cancel_delayed_work(&addr_chk_work);
6027 rtnl_unlock();
6029 destroy_workqueue(addrconf_wq);