net: delete /proc THIS_MODULE references
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blobab99cb641b7cccdda0ad4ae553c09274d7dbc047
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/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
61 #ifdef CONFIG_SYSCTL
62 #include <linux/sysctl.h>
63 #endif
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
70 #include <net/net_namespace.h>
71 #include <net/sock.h>
72 #include <net/snmp.h>
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
76 #include <net/ipv6.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
81 #include <net/tcp.h>
82 #include <net/ip.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
100 #if ACONF_DEBUG >= 3
101 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
102 #else
103 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
104 #endif
106 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108 #define IPV6_MAX_STRLEN \
109 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
111 static inline u32 cstamp_delta(unsigned long cstamp)
113 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
116 static inline s32 rfc3315_s14_backoff_init(s32 irt)
118 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
119 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
120 do_div(tmp, 1000000);
121 return (s32)tmp;
124 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
126 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
127 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
128 do_div(tmp, 1000000);
129 if ((s32)tmp > mrt) {
130 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
131 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
132 do_div(tmp, 1000000);
134 return (s32)tmp;
137 #ifdef CONFIG_SYSCTL
138 static int addrconf_sysctl_register(struct inet6_dev *idev);
139 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
140 #else
141 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
143 return 0;
146 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
149 #endif
151 static void ipv6_regen_rndid(struct inet6_dev *idev);
152 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
155 static int ipv6_count_addresses(const struct inet6_dev *idev);
156 static int ipv6_generate_stable_address(struct in6_addr *addr,
157 u8 dad_count,
158 const struct inet6_dev *idev);
160 #define IN6_ADDR_HSIZE_SHIFT 8
161 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
163 * Configured unicast address hash table
165 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
166 static DEFINE_SPINLOCK(addrconf_hash_lock);
168 static void addrconf_verify(void);
169 static void addrconf_verify_rtnl(void);
170 static void addrconf_verify_work(struct work_struct *);
172 static struct workqueue_struct *addrconf_wq;
173 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
175 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
176 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
178 static void addrconf_type_change(struct net_device *dev,
179 unsigned long event);
180 static int addrconf_ifdown(struct net_device *dev, int how);
182 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
183 int plen,
184 const struct net_device *dev,
185 u32 flags, u32 noflags);
187 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
188 static void addrconf_dad_work(struct work_struct *w);
189 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
190 static void addrconf_dad_run(struct inet6_dev *idev);
191 static void addrconf_rs_timer(struct timer_list *t);
192 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
193 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
195 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
196 struct prefix_info *pinfo);
198 static struct ipv6_devconf ipv6_devconf __read_mostly = {
199 .forwarding = 0,
200 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
201 .mtu6 = IPV6_MIN_MTU,
202 .accept_ra = 1,
203 .accept_redirects = 1,
204 .autoconf = 1,
205 .force_mld_version = 0,
206 .mldv1_unsolicited_report_interval = 10 * HZ,
207 .mldv2_unsolicited_report_interval = HZ,
208 .dad_transmits = 1,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
213 .use_tempaddr = 0,
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
218 .max_addresses = IPV6_MAX_ADDRESSES,
219 .accept_ra_defrtr = 1,
220 .accept_ra_from_local = 0,
221 .accept_ra_min_hop_limit= 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_min_plen = 0,
228 .accept_ra_rt_info_max_plen = 0,
229 #endif
230 #endif
231 .proxy_ndp = 0,
232 .accept_source_route = 0, /* we do not accept RH0 by default. */
233 .disable_ipv6 = 0,
234 .accept_dad = 0,
235 .suppress_frag_ndisc = 1,
236 .accept_ra_mtu = 1,
237 .stable_secret = {
238 .initialized = false,
240 .use_oif_addrs_only = 0,
241 .ignore_routes_with_linkdown = 0,
242 .keep_addr_on_down = 0,
243 .seg6_enabled = 0,
244 #ifdef CONFIG_IPV6_SEG6_HMAC
245 .seg6_require_hmac = 0,
246 #endif
247 .enhanced_dad = 1,
248 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
249 .disable_policy = 0,
252 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
253 .forwarding = 0,
254 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
255 .mtu6 = IPV6_MIN_MTU,
256 .accept_ra = 1,
257 .accept_redirects = 1,
258 .autoconf = 1,
259 .force_mld_version = 0,
260 .mldv1_unsolicited_report_interval = 10 * HZ,
261 .mldv2_unsolicited_report_interval = HZ,
262 .dad_transmits = 1,
263 .rtr_solicits = MAX_RTR_SOLICITATIONS,
264 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
265 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
266 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
267 .use_tempaddr = 0,
268 .temp_valid_lft = TEMP_VALID_LIFETIME,
269 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
270 .regen_max_retry = REGEN_MAX_RETRY,
271 .max_desync_factor = MAX_DESYNC_FACTOR,
272 .max_addresses = IPV6_MAX_ADDRESSES,
273 .accept_ra_defrtr = 1,
274 .accept_ra_from_local = 0,
275 .accept_ra_min_hop_limit= 1,
276 .accept_ra_pinfo = 1,
277 #ifdef CONFIG_IPV6_ROUTER_PREF
278 .accept_ra_rtr_pref = 1,
279 .rtr_probe_interval = 60 * HZ,
280 #ifdef CONFIG_IPV6_ROUTE_INFO
281 .accept_ra_rt_info_min_plen = 0,
282 .accept_ra_rt_info_max_plen = 0,
283 #endif
284 #endif
285 .proxy_ndp = 0,
286 .accept_source_route = 0, /* we do not accept RH0 by default. */
287 .disable_ipv6 = 0,
288 .accept_dad = 1,
289 .suppress_frag_ndisc = 1,
290 .accept_ra_mtu = 1,
291 .stable_secret = {
292 .initialized = false,
294 .use_oif_addrs_only = 0,
295 .ignore_routes_with_linkdown = 0,
296 .keep_addr_on_down = 0,
297 .seg6_enabled = 0,
298 #ifdef CONFIG_IPV6_SEG6_HMAC
299 .seg6_require_hmac = 0,
300 #endif
301 .enhanced_dad = 1,
302 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
303 .disable_policy = 0,
306 /* Check if link is ready: is it up and is a valid qdisc available */
307 static inline bool addrconf_link_ready(const struct net_device *dev)
309 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
312 static void addrconf_del_rs_timer(struct inet6_dev *idev)
314 if (del_timer(&idev->rs_timer))
315 __in6_dev_put(idev);
318 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
320 if (cancel_delayed_work(&ifp->dad_work))
321 __in6_ifa_put(ifp);
324 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
325 unsigned long when)
327 if (!timer_pending(&idev->rs_timer))
328 in6_dev_hold(idev);
329 mod_timer(&idev->rs_timer, jiffies + when);
332 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
333 unsigned long delay)
335 in6_ifa_hold(ifp);
336 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
337 in6_ifa_put(ifp);
340 static int snmp6_alloc_dev(struct inet6_dev *idev)
342 int i;
344 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
345 if (!idev->stats.ipv6)
346 goto err_ip;
348 for_each_possible_cpu(i) {
349 struct ipstats_mib *addrconf_stats;
350 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
351 u64_stats_init(&addrconf_stats->syncp);
355 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
356 GFP_KERNEL);
357 if (!idev->stats.icmpv6dev)
358 goto err_icmp;
359 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
360 GFP_KERNEL);
361 if (!idev->stats.icmpv6msgdev)
362 goto err_icmpmsg;
364 return 0;
366 err_icmpmsg:
367 kfree(idev->stats.icmpv6dev);
368 err_icmp:
369 free_percpu(idev->stats.ipv6);
370 err_ip:
371 return -ENOMEM;
374 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
376 struct inet6_dev *ndev;
377 int err = -ENOMEM;
379 ASSERT_RTNL();
381 if (dev->mtu < IPV6_MIN_MTU)
382 return ERR_PTR(-EINVAL);
384 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
385 if (!ndev)
386 return ERR_PTR(err);
388 rwlock_init(&ndev->lock);
389 ndev->dev = dev;
390 INIT_LIST_HEAD(&ndev->addr_list);
391 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
392 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
394 if (ndev->cnf.stable_secret.initialized)
395 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
396 else
397 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
399 ndev->cnf.mtu6 = dev->mtu;
400 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
401 if (!ndev->nd_parms) {
402 kfree(ndev);
403 return ERR_PTR(err);
405 if (ndev->cnf.forwarding)
406 dev_disable_lro(dev);
407 /* We refer to the device */
408 dev_hold(dev);
410 if (snmp6_alloc_dev(ndev) < 0) {
411 ADBG(KERN_WARNING
412 "%s: cannot allocate memory for statistics; dev=%s.\n",
413 __func__, dev->name);
414 neigh_parms_release(&nd_tbl, ndev->nd_parms);
415 dev_put(dev);
416 kfree(ndev);
417 return ERR_PTR(err);
420 if (snmp6_register_dev(ndev) < 0) {
421 ADBG(KERN_WARNING
422 "%s: cannot create /proc/net/dev_snmp6/%s\n",
423 __func__, dev->name);
424 goto err_release;
427 /* One reference from device. */
428 refcount_set(&ndev->refcnt, 1);
430 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
431 ndev->cnf.accept_dad = -1;
433 #if IS_ENABLED(CONFIG_IPV6_SIT)
434 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
435 pr_info("%s: Disabled Multicast RS\n", dev->name);
436 ndev->cnf.rtr_solicits = 0;
438 #endif
440 INIT_LIST_HEAD(&ndev->tempaddr_list);
441 ndev->desync_factor = U32_MAX;
442 if ((dev->flags&IFF_LOOPBACK) ||
443 dev->type == ARPHRD_TUNNEL ||
444 dev->type == ARPHRD_TUNNEL6 ||
445 dev->type == ARPHRD_SIT ||
446 dev->type == ARPHRD_NONE) {
447 ndev->cnf.use_tempaddr = -1;
448 } else
449 ipv6_regen_rndid(ndev);
451 ndev->token = in6addr_any;
453 if (netif_running(dev) && addrconf_link_ready(dev))
454 ndev->if_flags |= IF_READY;
456 ipv6_mc_init_dev(ndev);
457 ndev->tstamp = jiffies;
458 err = addrconf_sysctl_register(ndev);
459 if (err) {
460 ipv6_mc_destroy_dev(ndev);
461 snmp6_unregister_dev(ndev);
462 goto err_release;
464 /* protected by rtnl_lock */
465 rcu_assign_pointer(dev->ip6_ptr, ndev);
467 /* Join interface-local all-node multicast group */
468 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
470 /* Join all-node multicast group */
471 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
473 /* Join all-router multicast group if forwarding is set */
474 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
475 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
477 return ndev;
479 err_release:
480 neigh_parms_release(&nd_tbl, ndev->nd_parms);
481 ndev->dead = 1;
482 in6_dev_finish_destroy(ndev);
483 return ERR_PTR(err);
486 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
488 struct inet6_dev *idev;
490 ASSERT_RTNL();
492 idev = __in6_dev_get(dev);
493 if (!idev) {
494 idev = ipv6_add_dev(dev);
495 if (IS_ERR(idev))
496 return NULL;
499 if (dev->flags&IFF_UP)
500 ipv6_mc_up(idev);
501 return idev;
504 static int inet6_netconf_msgsize_devconf(int type)
506 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
507 + nla_total_size(4); /* NETCONFA_IFINDEX */
508 bool all = false;
510 if (type == NETCONFA_ALL)
511 all = true;
513 if (all || type == NETCONFA_FORWARDING)
514 size += nla_total_size(4);
515 #ifdef CONFIG_IPV6_MROUTE
516 if (all || type == NETCONFA_MC_FORWARDING)
517 size += nla_total_size(4);
518 #endif
519 if (all || type == NETCONFA_PROXY_NEIGH)
520 size += nla_total_size(4);
522 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
523 size += nla_total_size(4);
525 return size;
528 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
529 struct ipv6_devconf *devconf, u32 portid,
530 u32 seq, int event, unsigned int flags,
531 int type)
533 struct nlmsghdr *nlh;
534 struct netconfmsg *ncm;
535 bool all = false;
537 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
538 flags);
539 if (!nlh)
540 return -EMSGSIZE;
542 if (type == NETCONFA_ALL)
543 all = true;
545 ncm = nlmsg_data(nlh);
546 ncm->ncm_family = AF_INET6;
548 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
549 goto nla_put_failure;
551 if (!devconf)
552 goto out;
554 if ((all || type == NETCONFA_FORWARDING) &&
555 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
556 goto nla_put_failure;
557 #ifdef CONFIG_IPV6_MROUTE
558 if ((all || type == NETCONFA_MC_FORWARDING) &&
559 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
560 devconf->mc_forwarding) < 0)
561 goto nla_put_failure;
562 #endif
563 if ((all || type == NETCONFA_PROXY_NEIGH) &&
564 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
565 goto nla_put_failure;
567 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
568 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
569 devconf->ignore_routes_with_linkdown) < 0)
570 goto nla_put_failure;
572 out:
573 nlmsg_end(skb, nlh);
574 return 0;
576 nla_put_failure:
577 nlmsg_cancel(skb, nlh);
578 return -EMSGSIZE;
581 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
582 int ifindex, struct ipv6_devconf *devconf)
584 struct sk_buff *skb;
585 int err = -ENOBUFS;
587 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
588 if (!skb)
589 goto errout;
591 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
592 event, 0, type);
593 if (err < 0) {
594 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
595 WARN_ON(err == -EMSGSIZE);
596 kfree_skb(skb);
597 goto errout;
599 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
600 return;
601 errout:
602 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
605 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
606 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
607 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
608 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
609 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
612 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
613 struct nlmsghdr *nlh,
614 struct netlink_ext_ack *extack)
616 struct net *net = sock_net(in_skb->sk);
617 struct nlattr *tb[NETCONFA_MAX+1];
618 struct inet6_dev *in6_dev = NULL;
619 struct net_device *dev = NULL;
620 struct netconfmsg *ncm;
621 struct sk_buff *skb;
622 struct ipv6_devconf *devconf;
623 int ifindex;
624 int err;
626 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
627 devconf_ipv6_policy, extack);
628 if (err < 0)
629 return err;
631 if (!tb[NETCONFA_IFINDEX])
632 return -EINVAL;
634 err = -EINVAL;
635 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
636 switch (ifindex) {
637 case NETCONFA_IFINDEX_ALL:
638 devconf = net->ipv6.devconf_all;
639 break;
640 case NETCONFA_IFINDEX_DEFAULT:
641 devconf = net->ipv6.devconf_dflt;
642 break;
643 default:
644 dev = dev_get_by_index(net, ifindex);
645 if (!dev)
646 return -EINVAL;
647 in6_dev = in6_dev_get(dev);
648 if (!in6_dev)
649 goto errout;
650 devconf = &in6_dev->cnf;
651 break;
654 err = -ENOBUFS;
655 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
656 if (!skb)
657 goto errout;
659 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
660 NETLINK_CB(in_skb).portid,
661 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
662 NETCONFA_ALL);
663 if (err < 0) {
664 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
665 WARN_ON(err == -EMSGSIZE);
666 kfree_skb(skb);
667 goto errout;
669 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
670 errout:
671 if (in6_dev)
672 in6_dev_put(in6_dev);
673 if (dev)
674 dev_put(dev);
675 return err;
678 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
679 struct netlink_callback *cb)
681 struct net *net = sock_net(skb->sk);
682 int h, s_h;
683 int idx, s_idx;
684 struct net_device *dev;
685 struct inet6_dev *idev;
686 struct hlist_head *head;
688 s_h = cb->args[0];
689 s_idx = idx = cb->args[1];
691 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
692 idx = 0;
693 head = &net->dev_index_head[h];
694 rcu_read_lock();
695 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
696 net->dev_base_seq;
697 hlist_for_each_entry_rcu(dev, head, index_hlist) {
698 if (idx < s_idx)
699 goto cont;
700 idev = __in6_dev_get(dev);
701 if (!idev)
702 goto cont;
704 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
705 &idev->cnf,
706 NETLINK_CB(cb->skb).portid,
707 cb->nlh->nlmsg_seq,
708 RTM_NEWNETCONF,
709 NLM_F_MULTI,
710 NETCONFA_ALL) < 0) {
711 rcu_read_unlock();
712 goto done;
714 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
715 cont:
716 idx++;
718 rcu_read_unlock();
720 if (h == NETDEV_HASHENTRIES) {
721 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
722 net->ipv6.devconf_all,
723 NETLINK_CB(cb->skb).portid,
724 cb->nlh->nlmsg_seq,
725 RTM_NEWNETCONF, NLM_F_MULTI,
726 NETCONFA_ALL) < 0)
727 goto done;
728 else
729 h++;
731 if (h == NETDEV_HASHENTRIES + 1) {
732 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
733 net->ipv6.devconf_dflt,
734 NETLINK_CB(cb->skb).portid,
735 cb->nlh->nlmsg_seq,
736 RTM_NEWNETCONF, NLM_F_MULTI,
737 NETCONFA_ALL) < 0)
738 goto done;
739 else
740 h++;
742 done:
743 cb->args[0] = h;
744 cb->args[1] = idx;
746 return skb->len;
749 #ifdef CONFIG_SYSCTL
750 static void dev_forward_change(struct inet6_dev *idev)
752 struct net_device *dev;
753 struct inet6_ifaddr *ifa;
755 if (!idev)
756 return;
757 dev = idev->dev;
758 if (idev->cnf.forwarding)
759 dev_disable_lro(dev);
760 if (dev->flags & IFF_MULTICAST) {
761 if (idev->cnf.forwarding) {
762 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
763 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
764 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
765 } else {
766 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
767 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
768 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
772 list_for_each_entry(ifa, &idev->addr_list, if_list) {
773 if (ifa->flags&IFA_F_TENTATIVE)
774 continue;
775 if (idev->cnf.forwarding)
776 addrconf_join_anycast(ifa);
777 else
778 addrconf_leave_anycast(ifa);
780 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
781 NETCONFA_FORWARDING,
782 dev->ifindex, &idev->cnf);
786 static void addrconf_forward_change(struct net *net, __s32 newf)
788 struct net_device *dev;
789 struct inet6_dev *idev;
791 for_each_netdev(net, dev) {
792 idev = __in6_dev_get(dev);
793 if (idev) {
794 int changed = (!idev->cnf.forwarding) ^ (!newf);
795 idev->cnf.forwarding = newf;
796 if (changed)
797 dev_forward_change(idev);
802 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
804 struct net *net;
805 int old;
807 if (!rtnl_trylock())
808 return restart_syscall();
810 net = (struct net *)table->extra2;
811 old = *p;
812 *p = newf;
814 if (p == &net->ipv6.devconf_dflt->forwarding) {
815 if ((!newf) ^ (!old))
816 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
817 NETCONFA_FORWARDING,
818 NETCONFA_IFINDEX_DEFAULT,
819 net->ipv6.devconf_dflt);
820 rtnl_unlock();
821 return 0;
824 if (p == &net->ipv6.devconf_all->forwarding) {
825 int old_dflt = net->ipv6.devconf_dflt->forwarding;
827 net->ipv6.devconf_dflt->forwarding = newf;
828 if ((!newf) ^ (!old_dflt))
829 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
830 NETCONFA_FORWARDING,
831 NETCONFA_IFINDEX_DEFAULT,
832 net->ipv6.devconf_dflt);
834 addrconf_forward_change(net, newf);
835 if ((!newf) ^ (!old))
836 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
837 NETCONFA_FORWARDING,
838 NETCONFA_IFINDEX_ALL,
839 net->ipv6.devconf_all);
840 } else if ((!newf) ^ (!old))
841 dev_forward_change((struct inet6_dev *)table->extra1);
842 rtnl_unlock();
844 if (newf)
845 rt6_purge_dflt_routers(net);
846 return 1;
849 static void addrconf_linkdown_change(struct net *net, __s32 newf)
851 struct net_device *dev;
852 struct inet6_dev *idev;
854 for_each_netdev(net, dev) {
855 idev = __in6_dev_get(dev);
856 if (idev) {
857 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
859 idev->cnf.ignore_routes_with_linkdown = newf;
860 if (changed)
861 inet6_netconf_notify_devconf(dev_net(dev),
862 RTM_NEWNETCONF,
863 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
864 dev->ifindex,
865 &idev->cnf);
870 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
872 struct net *net;
873 int old;
875 if (!rtnl_trylock())
876 return restart_syscall();
878 net = (struct net *)table->extra2;
879 old = *p;
880 *p = newf;
882 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
883 if ((!newf) ^ (!old))
884 inet6_netconf_notify_devconf(net,
885 RTM_NEWNETCONF,
886 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
887 NETCONFA_IFINDEX_DEFAULT,
888 net->ipv6.devconf_dflt);
889 rtnl_unlock();
890 return 0;
893 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
894 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
895 addrconf_linkdown_change(net, newf);
896 if ((!newf) ^ (!old))
897 inet6_netconf_notify_devconf(net,
898 RTM_NEWNETCONF,
899 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
900 NETCONFA_IFINDEX_ALL,
901 net->ipv6.devconf_all);
903 rtnl_unlock();
905 return 1;
908 #endif
910 /* Nobody refers to this ifaddr, destroy it */
911 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
913 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
915 #ifdef NET_REFCNT_DEBUG
916 pr_debug("%s\n", __func__);
917 #endif
919 in6_dev_put(ifp->idev);
921 if (cancel_delayed_work(&ifp->dad_work))
922 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
923 ifp);
925 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
926 pr_warn("Freeing alive inet6 address %p\n", ifp);
927 return;
929 ip6_rt_put(ifp->rt);
931 kfree_rcu(ifp, rcu);
934 static void
935 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
937 struct list_head *p;
938 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
941 * Each device address list is sorted in order of scope -
942 * global before linklocal.
944 list_for_each(p, &idev->addr_list) {
945 struct inet6_ifaddr *ifa
946 = list_entry(p, struct inet6_ifaddr, if_list);
947 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
948 break;
951 list_add_tail_rcu(&ifp->if_list, p);
954 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
956 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
958 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
961 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
962 struct net_device *dev, unsigned int hash)
964 struct inet6_ifaddr *ifp;
966 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
967 if (!net_eq(dev_net(ifp->idev->dev), net))
968 continue;
969 if (ipv6_addr_equal(&ifp->addr, addr)) {
970 if (!dev || ifp->idev->dev == dev)
971 return true;
974 return false;
977 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
979 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
980 int err = 0;
982 spin_lock(&addrconf_hash_lock);
984 /* Ignore adding duplicate addresses on an interface */
985 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
986 ADBG("ipv6_add_addr: already assigned\n");
987 err = -EEXIST;
988 } else {
989 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
992 spin_unlock(&addrconf_hash_lock);
994 return err;
997 /* On success it returns ifp with increased reference count */
999 static struct inet6_ifaddr *
1000 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
1001 const struct in6_addr *peer_addr, int pfxlen,
1002 int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
1003 bool can_block, struct netlink_ext_ack *extack)
1005 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1006 struct net *net = dev_net(idev->dev);
1007 struct inet6_ifaddr *ifa = NULL;
1008 struct rt6_info *rt = NULL;
1009 int err = 0;
1010 int addr_type = ipv6_addr_type(addr);
1012 if (addr_type == IPV6_ADDR_ANY ||
1013 addr_type & IPV6_ADDR_MULTICAST ||
1014 (!(idev->dev->flags & IFF_LOOPBACK) &&
1015 addr_type & IPV6_ADDR_LOOPBACK))
1016 return ERR_PTR(-EADDRNOTAVAIL);
1018 if (idev->dead) {
1019 err = -ENODEV; /*XXX*/
1020 goto out;
1023 if (idev->cnf.disable_ipv6) {
1024 err = -EACCES;
1025 goto out;
1028 /* validator notifier needs to be blocking;
1029 * do not call in atomic context
1031 if (can_block) {
1032 struct in6_validator_info i6vi = {
1033 .i6vi_addr = *addr,
1034 .i6vi_dev = idev,
1035 .extack = extack,
1038 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1039 err = notifier_to_errno(err);
1040 if (err < 0)
1041 goto out;
1044 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1045 if (!ifa) {
1046 ADBG("ipv6_add_addr: malloc failed\n");
1047 err = -ENOBUFS;
1048 goto out;
1051 rt = addrconf_dst_alloc(idev, addr, false);
1052 if (IS_ERR(rt)) {
1053 err = PTR_ERR(rt);
1054 rt = NULL;
1055 goto out;
1058 if (net->ipv6.devconf_all->disable_policy ||
1059 idev->cnf.disable_policy)
1060 rt->dst.flags |= DST_NOPOLICY;
1062 neigh_parms_data_state_setall(idev->nd_parms);
1064 ifa->addr = *addr;
1065 if (peer_addr)
1066 ifa->peer_addr = *peer_addr;
1068 spin_lock_init(&ifa->lock);
1069 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1070 INIT_HLIST_NODE(&ifa->addr_lst);
1071 ifa->scope = scope;
1072 ifa->prefix_len = pfxlen;
1073 ifa->flags = flags;
1074 /* No need to add the TENTATIVE flag for addresses with NODAD */
1075 if (!(flags & IFA_F_NODAD))
1076 ifa->flags |= IFA_F_TENTATIVE;
1077 ifa->valid_lft = valid_lft;
1078 ifa->prefered_lft = prefered_lft;
1079 ifa->cstamp = ifa->tstamp = jiffies;
1080 ifa->tokenized = false;
1082 ifa->rt = rt;
1084 ifa->idev = idev;
1085 in6_dev_hold(idev);
1087 /* For caller */
1088 refcount_set(&ifa->refcnt, 1);
1090 rcu_read_lock_bh();
1092 err = ipv6_add_addr_hash(idev->dev, ifa);
1093 if (err < 0) {
1094 rcu_read_unlock_bh();
1095 goto out;
1098 write_lock(&idev->lock);
1100 /* Add to inet6_dev unicast addr list. */
1101 ipv6_link_dev_addr(idev, ifa);
1103 if (ifa->flags&IFA_F_TEMPORARY) {
1104 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1105 in6_ifa_hold(ifa);
1108 in6_ifa_hold(ifa);
1109 write_unlock(&idev->lock);
1111 rcu_read_unlock_bh();
1113 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1114 out:
1115 if (unlikely(err < 0)) {
1116 if (rt)
1117 ip6_rt_put(rt);
1118 if (ifa) {
1119 if (ifa->idev)
1120 in6_dev_put(ifa->idev);
1121 kfree(ifa);
1123 ifa = ERR_PTR(err);
1126 return ifa;
1129 enum cleanup_prefix_rt_t {
1130 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1131 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1132 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1136 * Check, whether the prefix for ifp would still need a prefix route
1137 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1138 * constants.
1140 * 1) we don't purge prefix if address was not permanent.
1141 * prefix is managed by its own lifetime.
1142 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1143 * 3) if there are no addresses, delete prefix.
1144 * 4) if there are still other permanent address(es),
1145 * corresponding prefix is still permanent.
1146 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1147 * don't purge the prefix, assume user space is managing it.
1148 * 6) otherwise, update prefix lifetime to the
1149 * longest valid lifetime among the corresponding
1150 * addresses on the device.
1151 * Note: subsequent RA will update lifetime.
1153 static enum cleanup_prefix_rt_t
1154 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1156 struct inet6_ifaddr *ifa;
1157 struct inet6_dev *idev = ifp->idev;
1158 unsigned long lifetime;
1159 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1161 *expires = jiffies;
1163 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1164 if (ifa == ifp)
1165 continue;
1166 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1167 ifp->prefix_len))
1168 continue;
1169 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1170 return CLEANUP_PREFIX_RT_NOP;
1172 action = CLEANUP_PREFIX_RT_EXPIRE;
1174 spin_lock(&ifa->lock);
1176 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1178 * Note: Because this address is
1179 * not permanent, lifetime <
1180 * LONG_MAX / HZ here.
1182 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1183 *expires = ifa->tstamp + lifetime * HZ;
1184 spin_unlock(&ifa->lock);
1187 return action;
1190 static void
1191 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1193 struct rt6_info *rt;
1195 rt = addrconf_get_prefix_route(&ifp->addr,
1196 ifp->prefix_len,
1197 ifp->idev->dev,
1198 0, RTF_GATEWAY | RTF_DEFAULT);
1199 if (rt) {
1200 if (del_rt)
1201 ip6_del_rt(rt);
1202 else {
1203 if (!(rt->rt6i_flags & RTF_EXPIRES))
1204 rt6_set_expires(rt, expires);
1205 ip6_rt_put(rt);
1211 /* This function wants to get referenced ifp and releases it before return */
1213 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1215 int state;
1216 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1217 unsigned long expires;
1219 ASSERT_RTNL();
1221 spin_lock_bh(&ifp->lock);
1222 state = ifp->state;
1223 ifp->state = INET6_IFADDR_STATE_DEAD;
1224 spin_unlock_bh(&ifp->lock);
1226 if (state == INET6_IFADDR_STATE_DEAD)
1227 goto out;
1229 spin_lock_bh(&addrconf_hash_lock);
1230 hlist_del_init_rcu(&ifp->addr_lst);
1231 spin_unlock_bh(&addrconf_hash_lock);
1233 write_lock_bh(&ifp->idev->lock);
1235 if (ifp->flags&IFA_F_TEMPORARY) {
1236 list_del(&ifp->tmp_list);
1237 if (ifp->ifpub) {
1238 in6_ifa_put(ifp->ifpub);
1239 ifp->ifpub = NULL;
1241 __in6_ifa_put(ifp);
1244 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1245 action = check_cleanup_prefix_route(ifp, &expires);
1247 list_del_rcu(&ifp->if_list);
1248 __in6_ifa_put(ifp);
1250 write_unlock_bh(&ifp->idev->lock);
1252 addrconf_del_dad_work(ifp);
1254 ipv6_ifa_notify(RTM_DELADDR, ifp);
1256 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1258 if (action != CLEANUP_PREFIX_RT_NOP) {
1259 cleanup_prefix_route(ifp, expires,
1260 action == CLEANUP_PREFIX_RT_DEL);
1263 /* clean up prefsrc entries */
1264 rt6_remove_prefsrc(ifp);
1265 out:
1266 in6_ifa_put(ifp);
1269 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1270 struct inet6_ifaddr *ift,
1271 bool block)
1273 struct inet6_dev *idev = ifp->idev;
1274 struct in6_addr addr, *tmpaddr;
1275 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1276 unsigned long regen_advance;
1277 int tmp_plen;
1278 int ret = 0;
1279 u32 addr_flags;
1280 unsigned long now = jiffies;
1281 long max_desync_factor;
1282 s32 cnf_temp_preferred_lft;
1284 write_lock_bh(&idev->lock);
1285 if (ift) {
1286 spin_lock_bh(&ift->lock);
1287 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1288 spin_unlock_bh(&ift->lock);
1289 tmpaddr = &addr;
1290 } else {
1291 tmpaddr = NULL;
1293 retry:
1294 in6_dev_hold(idev);
1295 if (idev->cnf.use_tempaddr <= 0) {
1296 write_unlock_bh(&idev->lock);
1297 pr_info("%s: use_tempaddr is disabled\n", __func__);
1298 in6_dev_put(idev);
1299 ret = -1;
1300 goto out;
1302 spin_lock_bh(&ifp->lock);
1303 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1304 idev->cnf.use_tempaddr = -1; /*XXX*/
1305 spin_unlock_bh(&ifp->lock);
1306 write_unlock_bh(&idev->lock);
1307 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1308 __func__);
1309 in6_dev_put(idev);
1310 ret = -1;
1311 goto out;
1313 in6_ifa_hold(ifp);
1314 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1315 ipv6_try_regen_rndid(idev, tmpaddr);
1316 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1317 age = (now - ifp->tstamp) / HZ;
1319 regen_advance = idev->cnf.regen_max_retry *
1320 idev->cnf.dad_transmits *
1321 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1323 /* recalculate max_desync_factor each time and update
1324 * idev->desync_factor if it's larger
1326 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1327 max_desync_factor = min_t(__u32,
1328 idev->cnf.max_desync_factor,
1329 cnf_temp_preferred_lft - regen_advance);
1331 if (unlikely(idev->desync_factor > max_desync_factor)) {
1332 if (max_desync_factor > 0) {
1333 get_random_bytes(&idev->desync_factor,
1334 sizeof(idev->desync_factor));
1335 idev->desync_factor %= max_desync_factor;
1336 } else {
1337 idev->desync_factor = 0;
1341 tmp_valid_lft = min_t(__u32,
1342 ifp->valid_lft,
1343 idev->cnf.temp_valid_lft + age);
1344 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1345 idev->desync_factor;
1346 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1347 tmp_plen = ifp->prefix_len;
1348 tmp_tstamp = ifp->tstamp;
1349 spin_unlock_bh(&ifp->lock);
1351 write_unlock_bh(&idev->lock);
1353 /* A temporary address is created only if this calculated Preferred
1354 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1355 * an implementation must not create a temporary address with a zero
1356 * Preferred Lifetime.
1357 * Use age calculation as in addrconf_verify to avoid unnecessary
1358 * temporary addresses being generated.
1360 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1361 if (tmp_prefered_lft <= regen_advance + age) {
1362 in6_ifa_put(ifp);
1363 in6_dev_put(idev);
1364 ret = -1;
1365 goto out;
1368 addr_flags = IFA_F_TEMPORARY;
1369 /* set in addrconf_prefix_rcv() */
1370 if (ifp->flags & IFA_F_OPTIMISTIC)
1371 addr_flags |= IFA_F_OPTIMISTIC;
1373 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1374 ipv6_addr_scope(&addr), addr_flags,
1375 tmp_valid_lft, tmp_prefered_lft, block, NULL);
1376 if (IS_ERR(ift)) {
1377 in6_ifa_put(ifp);
1378 in6_dev_put(idev);
1379 pr_info("%s: retry temporary address regeneration\n", __func__);
1380 tmpaddr = &addr;
1381 write_lock_bh(&idev->lock);
1382 goto retry;
1385 spin_lock_bh(&ift->lock);
1386 ift->ifpub = ifp;
1387 ift->cstamp = now;
1388 ift->tstamp = tmp_tstamp;
1389 spin_unlock_bh(&ift->lock);
1391 addrconf_dad_start(ift);
1392 in6_ifa_put(ift);
1393 in6_dev_put(idev);
1394 out:
1395 return ret;
1399 * Choose an appropriate source address (RFC3484)
1401 enum {
1402 IPV6_SADDR_RULE_INIT = 0,
1403 IPV6_SADDR_RULE_LOCAL,
1404 IPV6_SADDR_RULE_SCOPE,
1405 IPV6_SADDR_RULE_PREFERRED,
1406 #ifdef CONFIG_IPV6_MIP6
1407 IPV6_SADDR_RULE_HOA,
1408 #endif
1409 IPV6_SADDR_RULE_OIF,
1410 IPV6_SADDR_RULE_LABEL,
1411 IPV6_SADDR_RULE_PRIVACY,
1412 IPV6_SADDR_RULE_ORCHID,
1413 IPV6_SADDR_RULE_PREFIX,
1414 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1415 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1416 #endif
1417 IPV6_SADDR_RULE_MAX
1420 struct ipv6_saddr_score {
1421 int rule;
1422 int addr_type;
1423 struct inet6_ifaddr *ifa;
1424 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1425 int scopedist;
1426 int matchlen;
1429 struct ipv6_saddr_dst {
1430 const struct in6_addr *addr;
1431 int ifindex;
1432 int scope;
1433 int label;
1434 unsigned int prefs;
1437 static inline int ipv6_saddr_preferred(int type)
1439 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1440 return 1;
1441 return 0;
1444 static bool ipv6_use_optimistic_addr(struct net *net,
1445 struct inet6_dev *idev)
1447 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1448 if (!idev)
1449 return false;
1450 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1451 return false;
1452 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1453 return false;
1455 return true;
1456 #else
1457 return false;
1458 #endif
1461 static int ipv6_get_saddr_eval(struct net *net,
1462 struct ipv6_saddr_score *score,
1463 struct ipv6_saddr_dst *dst,
1464 int i)
1466 int ret;
1468 if (i <= score->rule) {
1469 switch (i) {
1470 case IPV6_SADDR_RULE_SCOPE:
1471 ret = score->scopedist;
1472 break;
1473 case IPV6_SADDR_RULE_PREFIX:
1474 ret = score->matchlen;
1475 break;
1476 default:
1477 ret = !!test_bit(i, score->scorebits);
1479 goto out;
1482 switch (i) {
1483 case IPV6_SADDR_RULE_INIT:
1484 /* Rule 0: remember if hiscore is not ready yet */
1485 ret = !!score->ifa;
1486 break;
1487 case IPV6_SADDR_RULE_LOCAL:
1488 /* Rule 1: Prefer same address */
1489 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1490 break;
1491 case IPV6_SADDR_RULE_SCOPE:
1492 /* Rule 2: Prefer appropriate scope
1494 * ret
1496 * -1 | d 15
1497 * ---+--+-+---> scope
1499 * | d is scope of the destination.
1500 * B-d | \
1501 * | \ <- smaller scope is better if
1502 * B-15 | \ if scope is enough for destination.
1503 * | ret = B - scope (-1 <= scope >= d <= 15).
1504 * d-C-1 | /
1505 * |/ <- greater is better
1506 * -C / if scope is not enough for destination.
1507 * /| ret = scope - C (-1 <= d < scope <= 15).
1509 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1510 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1511 * Assume B = 0 and we get C > 29.
1513 ret = __ipv6_addr_src_scope(score->addr_type);
1514 if (ret >= dst->scope)
1515 ret = -ret;
1516 else
1517 ret -= 128; /* 30 is enough */
1518 score->scopedist = ret;
1519 break;
1520 case IPV6_SADDR_RULE_PREFERRED:
1522 /* Rule 3: Avoid deprecated and optimistic addresses */
1523 u8 avoid = IFA_F_DEPRECATED;
1525 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1526 avoid |= IFA_F_OPTIMISTIC;
1527 ret = ipv6_saddr_preferred(score->addr_type) ||
1528 !(score->ifa->flags & avoid);
1529 break;
1531 #ifdef CONFIG_IPV6_MIP6
1532 case IPV6_SADDR_RULE_HOA:
1534 /* Rule 4: Prefer home address */
1535 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1536 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1537 break;
1539 #endif
1540 case IPV6_SADDR_RULE_OIF:
1541 /* Rule 5: Prefer outgoing interface */
1542 ret = (!dst->ifindex ||
1543 dst->ifindex == score->ifa->idev->dev->ifindex);
1544 break;
1545 case IPV6_SADDR_RULE_LABEL:
1546 /* Rule 6: Prefer matching label */
1547 ret = ipv6_addr_label(net,
1548 &score->ifa->addr, score->addr_type,
1549 score->ifa->idev->dev->ifindex) == dst->label;
1550 break;
1551 case IPV6_SADDR_RULE_PRIVACY:
1553 /* Rule 7: Prefer public address
1554 * Note: prefer temporary address if use_tempaddr >= 2
1556 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1557 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1558 score->ifa->idev->cnf.use_tempaddr >= 2;
1559 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1560 break;
1562 case IPV6_SADDR_RULE_ORCHID:
1563 /* Rule 8-: Prefer ORCHID vs ORCHID or
1564 * non-ORCHID vs non-ORCHID
1566 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1567 ipv6_addr_orchid(dst->addr));
1568 break;
1569 case IPV6_SADDR_RULE_PREFIX:
1570 /* Rule 8: Use longest matching prefix */
1571 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1572 if (ret > score->ifa->prefix_len)
1573 ret = score->ifa->prefix_len;
1574 score->matchlen = ret;
1575 break;
1576 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1577 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1578 /* Optimistic addresses still have lower precedence than other
1579 * preferred addresses.
1581 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1582 break;
1583 #endif
1584 default:
1585 ret = 0;
1588 if (ret)
1589 __set_bit(i, score->scorebits);
1590 score->rule = i;
1591 out:
1592 return ret;
1595 static int __ipv6_dev_get_saddr(struct net *net,
1596 struct ipv6_saddr_dst *dst,
1597 struct inet6_dev *idev,
1598 struct ipv6_saddr_score *scores,
1599 int hiscore_idx)
1601 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1603 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1604 int i;
1607 * - Tentative Address (RFC2462 section 5.4)
1608 * - A tentative address is not considered
1609 * "assigned to an interface" in the traditional
1610 * sense, unless it is also flagged as optimistic.
1611 * - Candidate Source Address (section 4)
1612 * - In any case, anycast addresses, multicast
1613 * addresses, and the unspecified address MUST
1614 * NOT be included in a candidate set.
1616 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1617 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1618 continue;
1620 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1622 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1623 score->addr_type & IPV6_ADDR_MULTICAST)) {
1624 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1625 idev->dev->name);
1626 continue;
1629 score->rule = -1;
1630 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1632 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1633 int minihiscore, miniscore;
1635 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1636 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1638 if (minihiscore > miniscore) {
1639 if (i == IPV6_SADDR_RULE_SCOPE &&
1640 score->scopedist > 0) {
1642 * special case:
1643 * each remaining entry
1644 * has too small (not enough)
1645 * scope, because ifa entries
1646 * are sorted by their scope
1647 * values.
1649 goto out;
1651 break;
1652 } else if (minihiscore < miniscore) {
1653 swap(hiscore, score);
1654 hiscore_idx = 1 - hiscore_idx;
1656 /* restore our iterator */
1657 score->ifa = hiscore->ifa;
1659 break;
1663 out:
1664 return hiscore_idx;
1667 static int ipv6_get_saddr_master(struct net *net,
1668 const struct net_device *dst_dev,
1669 const struct net_device *master,
1670 struct ipv6_saddr_dst *dst,
1671 struct ipv6_saddr_score *scores,
1672 int hiscore_idx)
1674 struct inet6_dev *idev;
1676 idev = __in6_dev_get(dst_dev);
1677 if (idev)
1678 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1679 scores, hiscore_idx);
1681 idev = __in6_dev_get(master);
1682 if (idev)
1683 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1684 scores, hiscore_idx);
1686 return hiscore_idx;
1689 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1690 const struct in6_addr *daddr, unsigned int prefs,
1691 struct in6_addr *saddr)
1693 struct ipv6_saddr_score scores[2], *hiscore;
1694 struct ipv6_saddr_dst dst;
1695 struct inet6_dev *idev;
1696 struct net_device *dev;
1697 int dst_type;
1698 bool use_oif_addr = false;
1699 int hiscore_idx = 0;
1700 int ret = 0;
1702 dst_type = __ipv6_addr_type(daddr);
1703 dst.addr = daddr;
1704 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1705 dst.scope = __ipv6_addr_src_scope(dst_type);
1706 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1707 dst.prefs = prefs;
1709 scores[hiscore_idx].rule = -1;
1710 scores[hiscore_idx].ifa = NULL;
1712 rcu_read_lock();
1714 /* Candidate Source Address (section 4)
1715 * - multicast and link-local destination address,
1716 * the set of candidate source address MUST only
1717 * include addresses assigned to interfaces
1718 * belonging to the same link as the outgoing
1719 * interface.
1720 * (- For site-local destination addresses, the
1721 * set of candidate source addresses MUST only
1722 * include addresses assigned to interfaces
1723 * belonging to the same site as the outgoing
1724 * interface.)
1725 * - "It is RECOMMENDED that the candidate source addresses
1726 * be the set of unicast addresses assigned to the
1727 * interface that will be used to send to the destination
1728 * (the 'outgoing' interface)." (RFC 6724)
1730 if (dst_dev) {
1731 idev = __in6_dev_get(dst_dev);
1732 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1733 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1734 (idev && idev->cnf.use_oif_addrs_only)) {
1735 use_oif_addr = true;
1739 if (use_oif_addr) {
1740 if (idev)
1741 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1742 } else {
1743 const struct net_device *master;
1744 int master_idx = 0;
1746 /* if dst_dev exists and is enslaved to an L3 device, then
1747 * prefer addresses from dst_dev and then the master over
1748 * any other enslaved devices in the L3 domain.
1750 master = l3mdev_master_dev_rcu(dst_dev);
1751 if (master) {
1752 master_idx = master->ifindex;
1754 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1755 master, &dst,
1756 scores, hiscore_idx);
1758 if (scores[hiscore_idx].ifa)
1759 goto out;
1762 for_each_netdev_rcu(net, dev) {
1763 /* only consider addresses on devices in the
1764 * same L3 domain
1766 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1767 continue;
1768 idev = __in6_dev_get(dev);
1769 if (!idev)
1770 continue;
1771 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1775 out:
1776 hiscore = &scores[hiscore_idx];
1777 if (!hiscore->ifa)
1778 ret = -EADDRNOTAVAIL;
1779 else
1780 *saddr = hiscore->ifa->addr;
1782 rcu_read_unlock();
1783 return ret;
1785 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1787 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1788 u32 banned_flags)
1790 struct inet6_ifaddr *ifp;
1791 int err = -EADDRNOTAVAIL;
1793 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1794 if (ifp->scope > IFA_LINK)
1795 break;
1796 if (ifp->scope == IFA_LINK &&
1797 !(ifp->flags & banned_flags)) {
1798 *addr = ifp->addr;
1799 err = 0;
1800 break;
1803 return err;
1806 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1807 u32 banned_flags)
1809 struct inet6_dev *idev;
1810 int err = -EADDRNOTAVAIL;
1812 rcu_read_lock();
1813 idev = __in6_dev_get(dev);
1814 if (idev) {
1815 read_lock_bh(&idev->lock);
1816 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1817 read_unlock_bh(&idev->lock);
1819 rcu_read_unlock();
1820 return err;
1823 static int ipv6_count_addresses(const struct inet6_dev *idev)
1825 const struct inet6_ifaddr *ifp;
1826 int cnt = 0;
1828 rcu_read_lock();
1829 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1830 cnt++;
1831 rcu_read_unlock();
1832 return cnt;
1835 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1836 const struct net_device *dev, int strict)
1838 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1840 EXPORT_SYMBOL(ipv6_chk_addr);
1842 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1843 const struct net_device *dev, int strict,
1844 u32 banned_flags)
1846 unsigned int hash = inet6_addr_hash(net, addr);
1847 struct inet6_ifaddr *ifp;
1848 u32 ifp_flags;
1850 rcu_read_lock();
1851 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1852 if (!net_eq(dev_net(ifp->idev->dev), net))
1853 continue;
1854 /* Decouple optimistic from tentative for evaluation here.
1855 * Ban optimistic addresses explicitly, when required.
1857 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1858 ? (ifp->flags&~IFA_F_TENTATIVE)
1859 : ifp->flags;
1860 if (ipv6_addr_equal(&ifp->addr, addr) &&
1861 !(ifp_flags&banned_flags) &&
1862 (!dev || ifp->idev->dev == dev ||
1863 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1864 rcu_read_unlock();
1865 return 1;
1869 rcu_read_unlock();
1870 return 0;
1872 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1875 /* Compares an address/prefix_len with addresses on device @dev.
1876 * If one is found it returns true.
1878 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1879 const unsigned int prefix_len, struct net_device *dev)
1881 const struct inet6_ifaddr *ifa;
1882 const struct inet6_dev *idev;
1883 bool ret = false;
1885 rcu_read_lock();
1886 idev = __in6_dev_get(dev);
1887 if (idev) {
1888 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1889 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1890 if (ret)
1891 break;
1894 rcu_read_unlock();
1896 return ret;
1898 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1900 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1902 const struct inet6_ifaddr *ifa;
1903 const struct inet6_dev *idev;
1904 int onlink;
1906 onlink = 0;
1907 rcu_read_lock();
1908 idev = __in6_dev_get(dev);
1909 if (idev) {
1910 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1911 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1912 ifa->prefix_len);
1913 if (onlink)
1914 break;
1917 rcu_read_unlock();
1918 return onlink;
1920 EXPORT_SYMBOL(ipv6_chk_prefix);
1922 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1923 struct net_device *dev, int strict)
1925 unsigned int hash = inet6_addr_hash(net, addr);
1926 struct inet6_ifaddr *ifp, *result = NULL;
1928 rcu_read_lock();
1929 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1930 if (!net_eq(dev_net(ifp->idev->dev), net))
1931 continue;
1932 if (ipv6_addr_equal(&ifp->addr, addr)) {
1933 if (!dev || ifp->idev->dev == dev ||
1934 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1935 result = ifp;
1936 in6_ifa_hold(ifp);
1937 break;
1941 rcu_read_unlock();
1943 return result;
1946 /* Gets referenced address, destroys ifaddr */
1948 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1950 if (dad_failed)
1951 ifp->flags |= IFA_F_DADFAILED;
1953 if (ifp->flags&IFA_F_TEMPORARY) {
1954 struct inet6_ifaddr *ifpub;
1955 spin_lock_bh(&ifp->lock);
1956 ifpub = ifp->ifpub;
1957 if (ifpub) {
1958 in6_ifa_hold(ifpub);
1959 spin_unlock_bh(&ifp->lock);
1960 ipv6_create_tempaddr(ifpub, ifp, true);
1961 in6_ifa_put(ifpub);
1962 } else {
1963 spin_unlock_bh(&ifp->lock);
1965 ipv6_del_addr(ifp);
1966 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1967 spin_lock_bh(&ifp->lock);
1968 addrconf_del_dad_work(ifp);
1969 ifp->flags |= IFA_F_TENTATIVE;
1970 spin_unlock_bh(&ifp->lock);
1971 if (dad_failed)
1972 ipv6_ifa_notify(0, ifp);
1973 in6_ifa_put(ifp);
1974 } else {
1975 ipv6_del_addr(ifp);
1979 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1981 int err = -ENOENT;
1983 spin_lock_bh(&ifp->lock);
1984 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1985 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1986 err = 0;
1988 spin_unlock_bh(&ifp->lock);
1990 return err;
1993 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
1995 struct inet6_dev *idev = ifp->idev;
1996 struct net *net = dev_net(ifp->idev->dev);
1998 if (addrconf_dad_end(ifp)) {
1999 in6_ifa_put(ifp);
2000 return;
2003 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2004 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2006 spin_lock_bh(&ifp->lock);
2008 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2009 int scope = ifp->scope;
2010 u32 flags = ifp->flags;
2011 struct in6_addr new_addr;
2012 struct inet6_ifaddr *ifp2;
2013 u32 valid_lft, preferred_lft;
2014 int pfxlen = ifp->prefix_len;
2015 int retries = ifp->stable_privacy_retry + 1;
2017 if (retries > net->ipv6.sysctl.idgen_retries) {
2018 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2019 ifp->idev->dev->name);
2020 goto errdad;
2023 new_addr = ifp->addr;
2024 if (ipv6_generate_stable_address(&new_addr, retries,
2025 idev))
2026 goto errdad;
2028 valid_lft = ifp->valid_lft;
2029 preferred_lft = ifp->prefered_lft;
2031 spin_unlock_bh(&ifp->lock);
2033 if (idev->cnf.max_addresses &&
2034 ipv6_count_addresses(idev) >=
2035 idev->cnf.max_addresses)
2036 goto lock_errdad;
2038 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2039 ifp->idev->dev->name);
2041 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2042 scope, flags, valid_lft,
2043 preferred_lft, false, NULL);
2044 if (IS_ERR(ifp2))
2045 goto lock_errdad;
2047 spin_lock_bh(&ifp2->lock);
2048 ifp2->stable_privacy_retry = retries;
2049 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2050 spin_unlock_bh(&ifp2->lock);
2052 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2053 in6_ifa_put(ifp2);
2054 lock_errdad:
2055 spin_lock_bh(&ifp->lock);
2058 errdad:
2059 /* transition from _POSTDAD to _ERRDAD */
2060 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2061 spin_unlock_bh(&ifp->lock);
2063 addrconf_mod_dad_work(ifp, 0);
2064 in6_ifa_put(ifp);
2067 /* Join to solicited addr multicast group.
2068 * caller must hold RTNL */
2069 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2071 struct in6_addr maddr;
2073 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2074 return;
2076 addrconf_addr_solict_mult(addr, &maddr);
2077 ipv6_dev_mc_inc(dev, &maddr);
2080 /* caller must hold RTNL */
2081 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2083 struct in6_addr maddr;
2085 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2086 return;
2088 addrconf_addr_solict_mult(addr, &maddr);
2089 __ipv6_dev_mc_dec(idev, &maddr);
2092 /* caller must hold RTNL */
2093 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2095 struct in6_addr addr;
2097 if (ifp->prefix_len >= 127) /* RFC 6164 */
2098 return;
2099 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2100 if (ipv6_addr_any(&addr))
2101 return;
2102 __ipv6_dev_ac_inc(ifp->idev, &addr);
2105 /* caller must hold RTNL */
2106 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2108 struct in6_addr addr;
2110 if (ifp->prefix_len >= 127) /* RFC 6164 */
2111 return;
2112 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2113 if (ipv6_addr_any(&addr))
2114 return;
2115 __ipv6_dev_ac_dec(ifp->idev, &addr);
2118 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2120 switch (dev->addr_len) {
2121 case ETH_ALEN:
2122 memcpy(eui, dev->dev_addr, 3);
2123 eui[3] = 0xFF;
2124 eui[4] = 0xFE;
2125 memcpy(eui + 5, dev->dev_addr + 3, 3);
2126 break;
2127 case EUI64_ADDR_LEN:
2128 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2129 eui[0] ^= 2;
2130 break;
2131 default:
2132 return -1;
2135 return 0;
2138 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2140 union fwnet_hwaddr *ha;
2142 if (dev->addr_len != FWNET_ALEN)
2143 return -1;
2145 ha = (union fwnet_hwaddr *)dev->dev_addr;
2147 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2148 eui[0] ^= 2;
2149 return 0;
2152 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2154 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2155 if (dev->addr_len != ARCNET_ALEN)
2156 return -1;
2157 memset(eui, 0, 7);
2158 eui[7] = *(u8 *)dev->dev_addr;
2159 return 0;
2162 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2164 if (dev->addr_len != INFINIBAND_ALEN)
2165 return -1;
2166 memcpy(eui, dev->dev_addr + 12, 8);
2167 eui[0] |= 2;
2168 return 0;
2171 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2173 if (addr == 0)
2174 return -1;
2175 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2176 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2177 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2178 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2179 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2180 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2181 eui[1] = 0;
2182 eui[2] = 0x5E;
2183 eui[3] = 0xFE;
2184 memcpy(eui + 4, &addr, 4);
2185 return 0;
2188 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2190 if (dev->priv_flags & IFF_ISATAP)
2191 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2192 return -1;
2195 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2197 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2200 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2202 memcpy(eui, dev->perm_addr, 3);
2203 memcpy(eui + 5, dev->perm_addr + 3, 3);
2204 eui[3] = 0xFF;
2205 eui[4] = 0xFE;
2206 eui[0] ^= 2;
2207 return 0;
2210 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2212 switch (dev->type) {
2213 case ARPHRD_ETHER:
2214 case ARPHRD_FDDI:
2215 return addrconf_ifid_eui48(eui, dev);
2216 case ARPHRD_ARCNET:
2217 return addrconf_ifid_arcnet(eui, dev);
2218 case ARPHRD_INFINIBAND:
2219 return addrconf_ifid_infiniband(eui, dev);
2220 case ARPHRD_SIT:
2221 return addrconf_ifid_sit(eui, dev);
2222 case ARPHRD_IPGRE:
2223 case ARPHRD_TUNNEL:
2224 return addrconf_ifid_gre(eui, dev);
2225 case ARPHRD_6LOWPAN:
2226 return addrconf_ifid_6lowpan(eui, dev);
2227 case ARPHRD_IEEE1394:
2228 return addrconf_ifid_ieee1394(eui, dev);
2229 case ARPHRD_TUNNEL6:
2230 case ARPHRD_IP6GRE:
2231 return addrconf_ifid_ip6tnl(eui, dev);
2233 return -1;
2236 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2238 int err = -1;
2239 struct inet6_ifaddr *ifp;
2241 read_lock_bh(&idev->lock);
2242 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2243 if (ifp->scope > IFA_LINK)
2244 break;
2245 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2246 memcpy(eui, ifp->addr.s6_addr+8, 8);
2247 err = 0;
2248 break;
2251 read_unlock_bh(&idev->lock);
2252 return err;
2255 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2256 static void ipv6_regen_rndid(struct inet6_dev *idev)
2258 regen:
2259 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2260 idev->rndid[0] &= ~0x02;
2263 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2264 * check if generated address is not inappropriate
2266 * - Reserved subnet anycast (RFC 2526)
2267 * 11111101 11....11 1xxxxxxx
2268 * - ISATAP (RFC4214) 6.1
2269 * 00-00-5E-FE-xx-xx-xx-xx
2270 * - value 0
2271 * - XXX: already assigned to an address on the device
2273 if (idev->rndid[0] == 0xfd &&
2274 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2275 (idev->rndid[7]&0x80))
2276 goto regen;
2277 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2278 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2279 goto regen;
2280 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2281 goto regen;
2285 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2287 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2288 ipv6_regen_rndid(idev);
2292 * Add prefix route.
2295 static void
2296 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2297 unsigned long expires, u32 flags)
2299 struct fib6_config cfg = {
2300 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2301 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2302 .fc_ifindex = dev->ifindex,
2303 .fc_expires = expires,
2304 .fc_dst_len = plen,
2305 .fc_flags = RTF_UP | flags,
2306 .fc_nlinfo.nl_net = dev_net(dev),
2307 .fc_protocol = RTPROT_KERNEL,
2310 cfg.fc_dst = *pfx;
2312 /* Prevent useless cloning on PtP SIT.
2313 This thing is done here expecting that the whole
2314 class of non-broadcast devices need not cloning.
2316 #if IS_ENABLED(CONFIG_IPV6_SIT)
2317 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2318 cfg.fc_flags |= RTF_NONEXTHOP;
2319 #endif
2321 ip6_route_add(&cfg, NULL);
2325 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2326 int plen,
2327 const struct net_device *dev,
2328 u32 flags, u32 noflags)
2330 struct fib6_node *fn;
2331 struct rt6_info *rt = NULL;
2332 struct fib6_table *table;
2333 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2335 table = fib6_get_table(dev_net(dev), tb_id);
2336 if (!table)
2337 return NULL;
2339 rcu_read_lock();
2340 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2341 if (!fn)
2342 goto out;
2344 for_each_fib6_node_rt_rcu(fn) {
2345 if (rt->dst.dev->ifindex != dev->ifindex)
2346 continue;
2347 if ((rt->rt6i_flags & flags) != flags)
2348 continue;
2349 if ((rt->rt6i_flags & noflags) != 0)
2350 continue;
2351 if (!dst_hold_safe(&rt->dst))
2352 rt = NULL;
2353 break;
2355 out:
2356 rcu_read_unlock();
2357 return rt;
2361 /* Create "default" multicast route to the interface */
2363 static void addrconf_add_mroute(struct net_device *dev)
2365 struct fib6_config cfg = {
2366 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2367 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2368 .fc_ifindex = dev->ifindex,
2369 .fc_dst_len = 8,
2370 .fc_flags = RTF_UP,
2371 .fc_nlinfo.nl_net = dev_net(dev),
2374 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2376 ip6_route_add(&cfg, NULL);
2379 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2381 struct inet6_dev *idev;
2383 ASSERT_RTNL();
2385 idev = ipv6_find_idev(dev);
2386 if (!idev)
2387 return ERR_PTR(-ENOBUFS);
2389 if (idev->cnf.disable_ipv6)
2390 return ERR_PTR(-EACCES);
2392 /* Add default multicast route */
2393 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2394 addrconf_add_mroute(dev);
2396 return idev;
2399 static void manage_tempaddrs(struct inet6_dev *idev,
2400 struct inet6_ifaddr *ifp,
2401 __u32 valid_lft, __u32 prefered_lft,
2402 bool create, unsigned long now)
2404 u32 flags;
2405 struct inet6_ifaddr *ift;
2407 read_lock_bh(&idev->lock);
2408 /* update all temporary addresses in the list */
2409 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2410 int age, max_valid, max_prefered;
2412 if (ifp != ift->ifpub)
2413 continue;
2415 /* RFC 4941 section 3.3:
2416 * If a received option will extend the lifetime of a public
2417 * address, the lifetimes of temporary addresses should
2418 * be extended, subject to the overall constraint that no
2419 * temporary addresses should ever remain "valid" or "preferred"
2420 * for a time longer than (TEMP_VALID_LIFETIME) or
2421 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2423 age = (now - ift->cstamp) / HZ;
2424 max_valid = idev->cnf.temp_valid_lft - age;
2425 if (max_valid < 0)
2426 max_valid = 0;
2428 max_prefered = idev->cnf.temp_prefered_lft -
2429 idev->desync_factor - age;
2430 if (max_prefered < 0)
2431 max_prefered = 0;
2433 if (valid_lft > max_valid)
2434 valid_lft = max_valid;
2436 if (prefered_lft > max_prefered)
2437 prefered_lft = max_prefered;
2439 spin_lock(&ift->lock);
2440 flags = ift->flags;
2441 ift->valid_lft = valid_lft;
2442 ift->prefered_lft = prefered_lft;
2443 ift->tstamp = now;
2444 if (prefered_lft > 0)
2445 ift->flags &= ~IFA_F_DEPRECATED;
2447 spin_unlock(&ift->lock);
2448 if (!(flags&IFA_F_TENTATIVE))
2449 ipv6_ifa_notify(0, ift);
2452 if ((create || list_empty(&idev->tempaddr_list)) &&
2453 idev->cnf.use_tempaddr > 0) {
2454 /* When a new public address is created as described
2455 * in [ADDRCONF], also create a new temporary address.
2456 * Also create a temporary address if it's enabled but
2457 * no temporary address currently exists.
2459 read_unlock_bh(&idev->lock);
2460 ipv6_create_tempaddr(ifp, NULL, false);
2461 } else {
2462 read_unlock_bh(&idev->lock);
2466 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2468 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2469 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2472 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2473 const struct prefix_info *pinfo,
2474 struct inet6_dev *in6_dev,
2475 const struct in6_addr *addr, int addr_type,
2476 u32 addr_flags, bool sllao, bool tokenized,
2477 __u32 valid_lft, u32 prefered_lft)
2479 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2480 int create = 0, update_lft = 0;
2482 if (!ifp && valid_lft) {
2483 int max_addresses = in6_dev->cnf.max_addresses;
2485 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2486 if ((net->ipv6.devconf_all->optimistic_dad ||
2487 in6_dev->cnf.optimistic_dad) &&
2488 !net->ipv6.devconf_all->forwarding && sllao)
2489 addr_flags |= IFA_F_OPTIMISTIC;
2490 #endif
2492 /* Do not allow to create too much of autoconfigured
2493 * addresses; this would be too easy way to crash kernel.
2495 if (!max_addresses ||
2496 ipv6_count_addresses(in6_dev) < max_addresses)
2497 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2498 pinfo->prefix_len,
2499 addr_type&IPV6_ADDR_SCOPE_MASK,
2500 addr_flags, valid_lft,
2501 prefered_lft, false, NULL);
2503 if (IS_ERR_OR_NULL(ifp))
2504 return -1;
2506 update_lft = 0;
2507 create = 1;
2508 spin_lock_bh(&ifp->lock);
2509 ifp->flags |= IFA_F_MANAGETEMPADDR;
2510 ifp->cstamp = jiffies;
2511 ifp->tokenized = tokenized;
2512 spin_unlock_bh(&ifp->lock);
2513 addrconf_dad_start(ifp);
2516 if (ifp) {
2517 u32 flags;
2518 unsigned long now;
2519 u32 stored_lft;
2521 /* update lifetime (RFC2462 5.5.3 e) */
2522 spin_lock_bh(&ifp->lock);
2523 now = jiffies;
2524 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2525 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2526 else
2527 stored_lft = 0;
2528 if (!update_lft && !create && stored_lft) {
2529 const u32 minimum_lft = min_t(u32,
2530 stored_lft, MIN_VALID_LIFETIME);
2531 valid_lft = max(valid_lft, minimum_lft);
2533 /* RFC4862 Section 5.5.3e:
2534 * "Note that the preferred lifetime of the
2535 * corresponding address is always reset to
2536 * the Preferred Lifetime in the received
2537 * Prefix Information option, regardless of
2538 * whether the valid lifetime is also reset or
2539 * ignored."
2541 * So we should always update prefered_lft here.
2543 update_lft = 1;
2546 if (update_lft) {
2547 ifp->valid_lft = valid_lft;
2548 ifp->prefered_lft = prefered_lft;
2549 ifp->tstamp = now;
2550 flags = ifp->flags;
2551 ifp->flags &= ~IFA_F_DEPRECATED;
2552 spin_unlock_bh(&ifp->lock);
2554 if (!(flags&IFA_F_TENTATIVE))
2555 ipv6_ifa_notify(0, ifp);
2556 } else
2557 spin_unlock_bh(&ifp->lock);
2559 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2560 create, now);
2562 in6_ifa_put(ifp);
2563 addrconf_verify();
2566 return 0;
2568 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2570 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2572 struct prefix_info *pinfo;
2573 __u32 valid_lft;
2574 __u32 prefered_lft;
2575 int addr_type, err;
2576 u32 addr_flags = 0;
2577 struct inet6_dev *in6_dev;
2578 struct net *net = dev_net(dev);
2580 pinfo = (struct prefix_info *) opt;
2582 if (len < sizeof(struct prefix_info)) {
2583 ADBG("addrconf: prefix option too short\n");
2584 return;
2588 * Validation checks ([ADDRCONF], page 19)
2591 addr_type = ipv6_addr_type(&pinfo->prefix);
2593 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2594 return;
2596 valid_lft = ntohl(pinfo->valid);
2597 prefered_lft = ntohl(pinfo->prefered);
2599 if (prefered_lft > valid_lft) {
2600 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2601 return;
2604 in6_dev = in6_dev_get(dev);
2606 if (!in6_dev) {
2607 net_dbg_ratelimited("addrconf: device %s not configured\n",
2608 dev->name);
2609 return;
2613 * Two things going on here:
2614 * 1) Add routes for on-link prefixes
2615 * 2) Configure prefixes with the auto flag set
2618 if (pinfo->onlink) {
2619 struct rt6_info *rt;
2620 unsigned long rt_expires;
2622 /* Avoid arithmetic overflow. Really, we could
2623 * save rt_expires in seconds, likely valid_lft,
2624 * but it would require division in fib gc, that it
2625 * not good.
2627 if (HZ > USER_HZ)
2628 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2629 else
2630 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2632 if (addrconf_finite_timeout(rt_expires))
2633 rt_expires *= HZ;
2635 rt = addrconf_get_prefix_route(&pinfo->prefix,
2636 pinfo->prefix_len,
2637 dev,
2638 RTF_ADDRCONF | RTF_PREFIX_RT,
2639 RTF_GATEWAY | RTF_DEFAULT);
2641 if (rt) {
2642 /* Autoconf prefix route */
2643 if (valid_lft == 0) {
2644 ip6_del_rt(rt);
2645 rt = NULL;
2646 } else if (addrconf_finite_timeout(rt_expires)) {
2647 /* not infinity */
2648 rt6_set_expires(rt, jiffies + rt_expires);
2649 } else {
2650 rt6_clean_expires(rt);
2652 } else if (valid_lft) {
2653 clock_t expires = 0;
2654 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2655 if (addrconf_finite_timeout(rt_expires)) {
2656 /* not infinity */
2657 flags |= RTF_EXPIRES;
2658 expires = jiffies_to_clock_t(rt_expires);
2660 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2661 dev, expires, flags);
2663 ip6_rt_put(rt);
2666 /* Try to figure out our local address for this prefix */
2668 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2669 struct in6_addr addr;
2670 bool tokenized = false, dev_addr_generated = false;
2672 if (pinfo->prefix_len == 64) {
2673 memcpy(&addr, &pinfo->prefix, 8);
2675 if (!ipv6_addr_any(&in6_dev->token)) {
2676 read_lock_bh(&in6_dev->lock);
2677 memcpy(addr.s6_addr + 8,
2678 in6_dev->token.s6_addr + 8, 8);
2679 read_unlock_bh(&in6_dev->lock);
2680 tokenized = true;
2681 } else if (is_addr_mode_generate_stable(in6_dev) &&
2682 !ipv6_generate_stable_address(&addr, 0,
2683 in6_dev)) {
2684 addr_flags |= IFA_F_STABLE_PRIVACY;
2685 goto ok;
2686 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2687 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2688 goto put;
2689 } else {
2690 dev_addr_generated = true;
2692 goto ok;
2694 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2695 pinfo->prefix_len);
2696 goto put;
2699 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2700 &addr, addr_type,
2701 addr_flags, sllao,
2702 tokenized, valid_lft,
2703 prefered_lft);
2704 if (err)
2705 goto put;
2707 /* Ignore error case here because previous prefix add addr was
2708 * successful which will be notified.
2710 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2711 addr_type, addr_flags, sllao,
2712 tokenized, valid_lft,
2713 prefered_lft,
2714 dev_addr_generated);
2716 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2717 put:
2718 in6_dev_put(in6_dev);
2722 * Set destination address.
2723 * Special case for SIT interfaces where we create a new "virtual"
2724 * device.
2726 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2728 struct in6_ifreq ireq;
2729 struct net_device *dev;
2730 int err = -EINVAL;
2732 rtnl_lock();
2734 err = -EFAULT;
2735 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2736 goto err_exit;
2738 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2740 err = -ENODEV;
2741 if (!dev)
2742 goto err_exit;
2744 #if IS_ENABLED(CONFIG_IPV6_SIT)
2745 if (dev->type == ARPHRD_SIT) {
2746 const struct net_device_ops *ops = dev->netdev_ops;
2747 struct ifreq ifr;
2748 struct ip_tunnel_parm p;
2750 err = -EADDRNOTAVAIL;
2751 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2752 goto err_exit;
2754 memset(&p, 0, sizeof(p));
2755 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2756 p.iph.saddr = 0;
2757 p.iph.version = 4;
2758 p.iph.ihl = 5;
2759 p.iph.protocol = IPPROTO_IPV6;
2760 p.iph.ttl = 64;
2761 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2763 if (ops->ndo_do_ioctl) {
2764 mm_segment_t oldfs = get_fs();
2766 set_fs(KERNEL_DS);
2767 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2768 set_fs(oldfs);
2769 } else
2770 err = -EOPNOTSUPP;
2772 if (err == 0) {
2773 err = -ENOBUFS;
2774 dev = __dev_get_by_name(net, p.name);
2775 if (!dev)
2776 goto err_exit;
2777 err = dev_open(dev);
2780 #endif
2782 err_exit:
2783 rtnl_unlock();
2784 return err;
2787 static int ipv6_mc_config(struct sock *sk, bool join,
2788 const struct in6_addr *addr, int ifindex)
2790 int ret;
2792 ASSERT_RTNL();
2794 lock_sock(sk);
2795 if (join)
2796 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2797 else
2798 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2799 release_sock(sk);
2801 return ret;
2805 * Manual configuration of address on an interface
2807 static int inet6_addr_add(struct net *net, int ifindex,
2808 const struct in6_addr *pfx,
2809 const struct in6_addr *peer_pfx,
2810 unsigned int plen, __u32 ifa_flags,
2811 __u32 prefered_lft, __u32 valid_lft,
2812 struct netlink_ext_ack *extack)
2814 struct inet6_ifaddr *ifp;
2815 struct inet6_dev *idev;
2816 struct net_device *dev;
2817 unsigned long timeout;
2818 clock_t expires;
2819 int scope;
2820 u32 flags;
2822 ASSERT_RTNL();
2824 if (plen > 128)
2825 return -EINVAL;
2827 /* check the lifetime */
2828 if (!valid_lft || prefered_lft > valid_lft)
2829 return -EINVAL;
2831 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2832 return -EINVAL;
2834 dev = __dev_get_by_index(net, ifindex);
2835 if (!dev)
2836 return -ENODEV;
2838 idev = addrconf_add_dev(dev);
2839 if (IS_ERR(idev))
2840 return PTR_ERR(idev);
2842 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2843 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2844 true, pfx, ifindex);
2846 if (ret < 0)
2847 return ret;
2850 scope = ipv6_addr_scope(pfx);
2852 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2853 if (addrconf_finite_timeout(timeout)) {
2854 expires = jiffies_to_clock_t(timeout * HZ);
2855 valid_lft = timeout;
2856 flags = RTF_EXPIRES;
2857 } else {
2858 expires = 0;
2859 flags = 0;
2860 ifa_flags |= IFA_F_PERMANENT;
2863 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2864 if (addrconf_finite_timeout(timeout)) {
2865 if (timeout == 0)
2866 ifa_flags |= IFA_F_DEPRECATED;
2867 prefered_lft = timeout;
2870 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2871 valid_lft, prefered_lft, true, extack);
2873 if (!IS_ERR(ifp)) {
2874 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2875 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2876 expires, flags);
2880 * Note that section 3.1 of RFC 4429 indicates
2881 * that the Optimistic flag should not be set for
2882 * manually configured addresses
2884 addrconf_dad_start(ifp);
2885 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2886 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2887 true, jiffies);
2888 in6_ifa_put(ifp);
2889 addrconf_verify_rtnl();
2890 return 0;
2891 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2892 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2893 false, pfx, ifindex);
2896 return PTR_ERR(ifp);
2899 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2900 const struct in6_addr *pfx, unsigned int plen)
2902 struct inet6_ifaddr *ifp;
2903 struct inet6_dev *idev;
2904 struct net_device *dev;
2906 if (plen > 128)
2907 return -EINVAL;
2909 dev = __dev_get_by_index(net, ifindex);
2910 if (!dev)
2911 return -ENODEV;
2913 idev = __in6_dev_get(dev);
2914 if (!idev)
2915 return -ENXIO;
2917 read_lock_bh(&idev->lock);
2918 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2919 if (ifp->prefix_len == plen &&
2920 ipv6_addr_equal(pfx, &ifp->addr)) {
2921 in6_ifa_hold(ifp);
2922 read_unlock_bh(&idev->lock);
2924 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2925 (ifa_flags & IFA_F_MANAGETEMPADDR))
2926 manage_tempaddrs(idev, ifp, 0, 0, false,
2927 jiffies);
2928 ipv6_del_addr(ifp);
2929 addrconf_verify_rtnl();
2930 if (ipv6_addr_is_multicast(pfx)) {
2931 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2932 false, pfx, dev->ifindex);
2934 return 0;
2937 read_unlock_bh(&idev->lock);
2938 return -EADDRNOTAVAIL;
2942 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2944 struct in6_ifreq ireq;
2945 int err;
2947 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2948 return -EPERM;
2950 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2951 return -EFAULT;
2953 rtnl_lock();
2954 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2955 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2956 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2957 rtnl_unlock();
2958 return err;
2961 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2963 struct in6_ifreq ireq;
2964 int err;
2966 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2967 return -EPERM;
2969 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2970 return -EFAULT;
2972 rtnl_lock();
2973 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2974 ireq.ifr6_prefixlen);
2975 rtnl_unlock();
2976 return err;
2979 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2980 int plen, int scope)
2982 struct inet6_ifaddr *ifp;
2984 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2985 scope, IFA_F_PERMANENT,
2986 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
2987 true, NULL);
2988 if (!IS_ERR(ifp)) {
2989 spin_lock_bh(&ifp->lock);
2990 ifp->flags &= ~IFA_F_TENTATIVE;
2991 spin_unlock_bh(&ifp->lock);
2992 rt_genid_bump_ipv6(dev_net(idev->dev));
2993 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2994 in6_ifa_put(ifp);
2998 #if IS_ENABLED(CONFIG_IPV6_SIT)
2999 static void sit_add_v4_addrs(struct inet6_dev *idev)
3001 struct in6_addr addr;
3002 struct net_device *dev;
3003 struct net *net = dev_net(idev->dev);
3004 int scope, plen;
3005 u32 pflags = 0;
3007 ASSERT_RTNL();
3009 memset(&addr, 0, sizeof(struct in6_addr));
3010 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3012 if (idev->dev->flags&IFF_POINTOPOINT) {
3013 addr.s6_addr32[0] = htonl(0xfe800000);
3014 scope = IFA_LINK;
3015 plen = 64;
3016 } else {
3017 scope = IPV6_ADDR_COMPATv4;
3018 plen = 96;
3019 pflags |= RTF_NONEXTHOP;
3022 if (addr.s6_addr32[3]) {
3023 add_addr(idev, &addr, plen, scope);
3024 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3025 return;
3028 for_each_netdev(net, dev) {
3029 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3030 if (in_dev && (dev->flags & IFF_UP)) {
3031 struct in_ifaddr *ifa;
3033 int flag = scope;
3035 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3037 addr.s6_addr32[3] = ifa->ifa_local;
3039 if (ifa->ifa_scope == RT_SCOPE_LINK)
3040 continue;
3041 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3042 if (idev->dev->flags&IFF_POINTOPOINT)
3043 continue;
3044 flag |= IFA_HOST;
3047 add_addr(idev, &addr, plen, flag);
3048 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3049 pflags);
3054 #endif
3056 static void init_loopback(struct net_device *dev)
3058 struct inet6_dev *idev;
3060 /* ::1 */
3062 ASSERT_RTNL();
3064 idev = ipv6_find_idev(dev);
3065 if (!idev) {
3066 pr_debug("%s: add_dev failed\n", __func__);
3067 return;
3070 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3073 void addrconf_add_linklocal(struct inet6_dev *idev,
3074 const struct in6_addr *addr, u32 flags)
3076 struct inet6_ifaddr *ifp;
3077 u32 addr_flags = flags | IFA_F_PERMANENT;
3079 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3080 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3081 idev->cnf.optimistic_dad) &&
3082 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3083 addr_flags |= IFA_F_OPTIMISTIC;
3084 #endif
3086 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3087 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3088 if (!IS_ERR(ifp)) {
3089 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3090 addrconf_dad_start(ifp);
3091 in6_ifa_put(ifp);
3094 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3096 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3098 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3099 return true;
3101 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3102 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3103 return true;
3105 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3106 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3107 return true;
3109 return false;
3112 static int ipv6_generate_stable_address(struct in6_addr *address,
3113 u8 dad_count,
3114 const struct inet6_dev *idev)
3116 static DEFINE_SPINLOCK(lock);
3117 static __u32 digest[SHA_DIGEST_WORDS];
3118 static __u32 workspace[SHA_WORKSPACE_WORDS];
3120 static union {
3121 char __data[SHA_MESSAGE_BYTES];
3122 struct {
3123 struct in6_addr secret;
3124 __be32 prefix[2];
3125 unsigned char hwaddr[MAX_ADDR_LEN];
3126 u8 dad_count;
3127 } __packed;
3128 } data;
3130 struct in6_addr secret;
3131 struct in6_addr temp;
3132 struct net *net = dev_net(idev->dev);
3134 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3136 if (idev->cnf.stable_secret.initialized)
3137 secret = idev->cnf.stable_secret.secret;
3138 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3139 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3140 else
3141 return -1;
3143 retry:
3144 spin_lock_bh(&lock);
3146 sha_init(digest);
3147 memset(&data, 0, sizeof(data));
3148 memset(workspace, 0, sizeof(workspace));
3149 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3150 data.prefix[0] = address->s6_addr32[0];
3151 data.prefix[1] = address->s6_addr32[1];
3152 data.secret = secret;
3153 data.dad_count = dad_count;
3155 sha_transform(digest, data.__data, workspace);
3157 temp = *address;
3158 temp.s6_addr32[2] = (__force __be32)digest[0];
3159 temp.s6_addr32[3] = (__force __be32)digest[1];
3161 spin_unlock_bh(&lock);
3163 if (ipv6_reserved_interfaceid(temp)) {
3164 dad_count++;
3165 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3166 return -1;
3167 goto retry;
3170 *address = temp;
3171 return 0;
3174 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3176 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3178 if (s->initialized)
3179 return;
3180 s = &idev->cnf.stable_secret;
3181 get_random_bytes(&s->secret, sizeof(s->secret));
3182 s->initialized = true;
3185 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3187 struct in6_addr addr;
3189 /* no link local addresses on L3 master devices */
3190 if (netif_is_l3_master(idev->dev))
3191 return;
3193 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3195 switch (idev->cnf.addr_gen_mode) {
3196 case IN6_ADDR_GEN_MODE_RANDOM:
3197 ipv6_gen_mode_random_init(idev);
3198 /* fallthrough */
3199 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3200 if (!ipv6_generate_stable_address(&addr, 0, idev))
3201 addrconf_add_linklocal(idev, &addr,
3202 IFA_F_STABLE_PRIVACY);
3203 else if (prefix_route)
3204 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3205 break;
3206 case IN6_ADDR_GEN_MODE_EUI64:
3207 /* addrconf_add_linklocal also adds a prefix_route and we
3208 * only need to care about prefix routes if ipv6_generate_eui64
3209 * couldn't generate one.
3211 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3212 addrconf_add_linklocal(idev, &addr, 0);
3213 else if (prefix_route)
3214 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3215 break;
3216 case IN6_ADDR_GEN_MODE_NONE:
3217 default:
3218 /* will not add any link local address */
3219 break;
3223 static void addrconf_dev_config(struct net_device *dev)
3225 struct inet6_dev *idev;
3227 ASSERT_RTNL();
3229 if ((dev->type != ARPHRD_ETHER) &&
3230 (dev->type != ARPHRD_FDDI) &&
3231 (dev->type != ARPHRD_ARCNET) &&
3232 (dev->type != ARPHRD_INFINIBAND) &&
3233 (dev->type != ARPHRD_IEEE1394) &&
3234 (dev->type != ARPHRD_TUNNEL6) &&
3235 (dev->type != ARPHRD_6LOWPAN) &&
3236 (dev->type != ARPHRD_IP6GRE) &&
3237 (dev->type != ARPHRD_IPGRE) &&
3238 (dev->type != ARPHRD_TUNNEL) &&
3239 (dev->type != ARPHRD_NONE)) {
3240 /* Alas, we support only Ethernet autoconfiguration. */
3241 return;
3244 idev = addrconf_add_dev(dev);
3245 if (IS_ERR(idev))
3246 return;
3248 /* this device type has no EUI support */
3249 if (dev->type == ARPHRD_NONE &&
3250 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3251 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3253 addrconf_addr_gen(idev, false);
3256 #if IS_ENABLED(CONFIG_IPV6_SIT)
3257 static void addrconf_sit_config(struct net_device *dev)
3259 struct inet6_dev *idev;
3261 ASSERT_RTNL();
3264 * Configure the tunnel with one of our IPv4
3265 * addresses... we should configure all of
3266 * our v4 addrs in the tunnel
3269 idev = ipv6_find_idev(dev);
3270 if (!idev) {
3271 pr_debug("%s: add_dev failed\n", __func__);
3272 return;
3275 if (dev->priv_flags & IFF_ISATAP) {
3276 addrconf_addr_gen(idev, false);
3277 return;
3280 sit_add_v4_addrs(idev);
3282 if (dev->flags&IFF_POINTOPOINT)
3283 addrconf_add_mroute(dev);
3285 #endif
3287 #if IS_ENABLED(CONFIG_NET_IPGRE)
3288 static void addrconf_gre_config(struct net_device *dev)
3290 struct inet6_dev *idev;
3292 ASSERT_RTNL();
3294 idev = ipv6_find_idev(dev);
3295 if (!idev) {
3296 pr_debug("%s: add_dev failed\n", __func__);
3297 return;
3300 addrconf_addr_gen(idev, true);
3301 if (dev->flags & IFF_POINTOPOINT)
3302 addrconf_add_mroute(dev);
3304 #endif
3306 static int fixup_permanent_addr(struct inet6_dev *idev,
3307 struct inet6_ifaddr *ifp)
3309 /* !rt6i_node means the host route was removed from the
3310 * FIB, for example, if 'lo' device is taken down. In that
3311 * case regenerate the host route.
3313 if (!ifp->rt || !ifp->rt->rt6i_node) {
3314 struct rt6_info *rt, *prev;
3316 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3317 if (IS_ERR(rt))
3318 return PTR_ERR(rt);
3320 /* ifp->rt can be accessed outside of rtnl */
3321 spin_lock(&ifp->lock);
3322 prev = ifp->rt;
3323 ifp->rt = rt;
3324 spin_unlock(&ifp->lock);
3326 ip6_rt_put(prev);
3329 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3330 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3331 idev->dev, 0, 0);
3334 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3335 addrconf_dad_start(ifp);
3337 return 0;
3340 static void addrconf_permanent_addr(struct net_device *dev)
3342 struct inet6_ifaddr *ifp, *tmp;
3343 struct inet6_dev *idev;
3345 idev = __in6_dev_get(dev);
3346 if (!idev)
3347 return;
3349 write_lock_bh(&idev->lock);
3351 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3352 if ((ifp->flags & IFA_F_PERMANENT) &&
3353 fixup_permanent_addr(idev, ifp) < 0) {
3354 write_unlock_bh(&idev->lock);
3355 in6_ifa_hold(ifp);
3356 ipv6_del_addr(ifp);
3357 write_lock_bh(&idev->lock);
3359 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3360 idev->dev->name, &ifp->addr);
3364 write_unlock_bh(&idev->lock);
3367 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3368 void *ptr)
3370 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3371 struct netdev_notifier_changeupper_info *info;
3372 struct inet6_dev *idev = __in6_dev_get(dev);
3373 struct net *net = dev_net(dev);
3374 int run_pending = 0;
3375 int err;
3377 switch (event) {
3378 case NETDEV_REGISTER:
3379 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3380 idev = ipv6_add_dev(dev);
3381 if (IS_ERR(idev))
3382 return notifier_from_errno(PTR_ERR(idev));
3384 break;
3386 case NETDEV_CHANGEMTU:
3387 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3388 if (dev->mtu < IPV6_MIN_MTU) {
3389 addrconf_ifdown(dev, dev != net->loopback_dev);
3390 break;
3393 if (idev) {
3394 rt6_mtu_change(dev, dev->mtu);
3395 idev->cnf.mtu6 = dev->mtu;
3396 break;
3399 /* allocate new idev */
3400 idev = ipv6_add_dev(dev);
3401 if (IS_ERR(idev))
3402 break;
3404 /* device is still not ready */
3405 if (!(idev->if_flags & IF_READY))
3406 break;
3408 run_pending = 1;
3410 /* fall through */
3412 case NETDEV_UP:
3413 case NETDEV_CHANGE:
3414 if (dev->flags & IFF_SLAVE)
3415 break;
3417 if (idev && idev->cnf.disable_ipv6)
3418 break;
3420 if (event == NETDEV_UP) {
3421 /* restore routes for permanent addresses */
3422 addrconf_permanent_addr(dev);
3424 if (!addrconf_link_ready(dev)) {
3425 /* device is not ready yet. */
3426 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3427 dev->name);
3428 break;
3431 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3432 idev = ipv6_add_dev(dev);
3434 if (!IS_ERR_OR_NULL(idev)) {
3435 idev->if_flags |= IF_READY;
3436 run_pending = 1;
3438 } else if (event == NETDEV_CHANGE) {
3439 if (!addrconf_link_ready(dev)) {
3440 /* device is still not ready. */
3441 rt6_sync_down_dev(dev, event);
3442 break;
3445 if (idev) {
3446 if (idev->if_flags & IF_READY) {
3447 /* device is already configured -
3448 * but resend MLD reports, we might
3449 * have roamed and need to update
3450 * multicast snooping switches
3452 ipv6_mc_up(idev);
3453 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3454 break;
3456 idev->if_flags |= IF_READY;
3459 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3460 dev->name);
3462 run_pending = 1;
3465 switch (dev->type) {
3466 #if IS_ENABLED(CONFIG_IPV6_SIT)
3467 case ARPHRD_SIT:
3468 addrconf_sit_config(dev);
3469 break;
3470 #endif
3471 #if IS_ENABLED(CONFIG_NET_IPGRE)
3472 case ARPHRD_IPGRE:
3473 addrconf_gre_config(dev);
3474 break;
3475 #endif
3476 case ARPHRD_LOOPBACK:
3477 init_loopback(dev);
3478 break;
3480 default:
3481 addrconf_dev_config(dev);
3482 break;
3485 if (!IS_ERR_OR_NULL(idev)) {
3486 if (run_pending)
3487 addrconf_dad_run(idev);
3489 /* Device has an address by now */
3490 rt6_sync_up(dev, RTNH_F_DEAD);
3493 * If the MTU changed during the interface down,
3494 * when the interface up, the changed MTU must be
3495 * reflected in the idev as well as routers.
3497 if (idev->cnf.mtu6 != dev->mtu &&
3498 dev->mtu >= IPV6_MIN_MTU) {
3499 rt6_mtu_change(dev, dev->mtu);
3500 idev->cnf.mtu6 = dev->mtu;
3502 idev->tstamp = jiffies;
3503 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3506 * If the changed mtu during down is lower than
3507 * IPV6_MIN_MTU stop IPv6 on this interface.
3509 if (dev->mtu < IPV6_MIN_MTU)
3510 addrconf_ifdown(dev, dev != net->loopback_dev);
3512 break;
3514 case NETDEV_DOWN:
3515 case NETDEV_UNREGISTER:
3517 * Remove all addresses from this interface.
3519 addrconf_ifdown(dev, event != NETDEV_DOWN);
3520 break;
3522 case NETDEV_CHANGENAME:
3523 if (idev) {
3524 snmp6_unregister_dev(idev);
3525 addrconf_sysctl_unregister(idev);
3526 err = addrconf_sysctl_register(idev);
3527 if (err)
3528 return notifier_from_errno(err);
3529 err = snmp6_register_dev(idev);
3530 if (err) {
3531 addrconf_sysctl_unregister(idev);
3532 return notifier_from_errno(err);
3535 break;
3537 case NETDEV_PRE_TYPE_CHANGE:
3538 case NETDEV_POST_TYPE_CHANGE:
3539 if (idev)
3540 addrconf_type_change(dev, event);
3541 break;
3543 case NETDEV_CHANGEUPPER:
3544 info = ptr;
3546 /* flush all routes if dev is linked to or unlinked from
3547 * an L3 master device (e.g., VRF)
3549 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3550 addrconf_ifdown(dev, 0);
3553 return NOTIFY_OK;
3557 * addrconf module should be notified of a device going up
3559 static struct notifier_block ipv6_dev_notf = {
3560 .notifier_call = addrconf_notify,
3561 .priority = ADDRCONF_NOTIFY_PRIORITY,
3564 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3566 struct inet6_dev *idev;
3567 ASSERT_RTNL();
3569 idev = __in6_dev_get(dev);
3571 if (event == NETDEV_POST_TYPE_CHANGE)
3572 ipv6_mc_remap(idev);
3573 else if (event == NETDEV_PRE_TYPE_CHANGE)
3574 ipv6_mc_unmap(idev);
3577 static bool addr_is_local(const struct in6_addr *addr)
3579 return ipv6_addr_type(addr) &
3580 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3583 static int addrconf_ifdown(struct net_device *dev, int how)
3585 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3586 struct net *net = dev_net(dev);
3587 struct inet6_dev *idev;
3588 struct inet6_ifaddr *ifa, *tmp;
3589 int _keep_addr;
3590 bool keep_addr;
3591 int state, i;
3593 ASSERT_RTNL();
3595 rt6_disable_ip(dev, event);
3597 idev = __in6_dev_get(dev);
3598 if (!idev)
3599 return -ENODEV;
3602 * Step 1: remove reference to ipv6 device from parent device.
3603 * Do not dev_put!
3605 if (how) {
3606 idev->dead = 1;
3608 /* protected by rtnl_lock */
3609 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3611 /* Step 1.5: remove snmp6 entry */
3612 snmp6_unregister_dev(idev);
3616 /* aggregate the system setting and interface setting */
3617 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3618 if (!_keep_addr)
3619 _keep_addr = idev->cnf.keep_addr_on_down;
3621 /* combine the user config with event to determine if permanent
3622 * addresses are to be removed from address hash table
3624 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3626 /* Step 2: clear hash table */
3627 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3628 struct hlist_head *h = &inet6_addr_lst[i];
3630 spin_lock_bh(&addrconf_hash_lock);
3631 restart:
3632 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3633 if (ifa->idev == idev) {
3634 addrconf_del_dad_work(ifa);
3635 /* combined flag + permanent flag decide if
3636 * address is retained on a down event
3638 if (!keep_addr ||
3639 !(ifa->flags & IFA_F_PERMANENT) ||
3640 addr_is_local(&ifa->addr)) {
3641 hlist_del_init_rcu(&ifa->addr_lst);
3642 goto restart;
3646 spin_unlock_bh(&addrconf_hash_lock);
3649 write_lock_bh(&idev->lock);
3651 addrconf_del_rs_timer(idev);
3653 /* Step 2: clear flags for stateless addrconf */
3654 if (!how)
3655 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3657 /* Step 3: clear tempaddr list */
3658 while (!list_empty(&idev->tempaddr_list)) {
3659 ifa = list_first_entry(&idev->tempaddr_list,
3660 struct inet6_ifaddr, tmp_list);
3661 list_del(&ifa->tmp_list);
3662 write_unlock_bh(&idev->lock);
3663 spin_lock_bh(&ifa->lock);
3665 if (ifa->ifpub) {
3666 in6_ifa_put(ifa->ifpub);
3667 ifa->ifpub = NULL;
3669 spin_unlock_bh(&ifa->lock);
3670 in6_ifa_put(ifa);
3671 write_lock_bh(&idev->lock);
3674 /* re-combine the user config with event to determine if permanent
3675 * addresses are to be removed from the interface list
3677 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3679 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3680 struct rt6_info *rt = NULL;
3681 bool keep;
3683 addrconf_del_dad_work(ifa);
3685 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3686 !addr_is_local(&ifa->addr);
3688 write_unlock_bh(&idev->lock);
3689 spin_lock_bh(&ifa->lock);
3691 if (keep) {
3692 /* set state to skip the notifier below */
3693 state = INET6_IFADDR_STATE_DEAD;
3694 ifa->state = INET6_IFADDR_STATE_PREDAD;
3695 if (!(ifa->flags & IFA_F_NODAD))
3696 ifa->flags |= IFA_F_TENTATIVE;
3698 rt = ifa->rt;
3699 ifa->rt = NULL;
3700 } else {
3701 state = ifa->state;
3702 ifa->state = INET6_IFADDR_STATE_DEAD;
3705 spin_unlock_bh(&ifa->lock);
3707 if (rt)
3708 ip6_del_rt(rt);
3710 if (state != INET6_IFADDR_STATE_DEAD) {
3711 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3712 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3713 } else {
3714 if (idev->cnf.forwarding)
3715 addrconf_leave_anycast(ifa);
3716 addrconf_leave_solict(ifa->idev, &ifa->addr);
3719 write_lock_bh(&idev->lock);
3720 if (!keep) {
3721 list_del_rcu(&ifa->if_list);
3722 in6_ifa_put(ifa);
3726 write_unlock_bh(&idev->lock);
3728 /* Step 5: Discard anycast and multicast list */
3729 if (how) {
3730 ipv6_ac_destroy_dev(idev);
3731 ipv6_mc_destroy_dev(idev);
3732 } else {
3733 ipv6_mc_down(idev);
3736 idev->tstamp = jiffies;
3738 /* Last: Shot the device (if unregistered) */
3739 if (how) {
3740 addrconf_sysctl_unregister(idev);
3741 neigh_parms_release(&nd_tbl, idev->nd_parms);
3742 neigh_ifdown(&nd_tbl, dev);
3743 in6_dev_put(idev);
3745 return 0;
3748 static void addrconf_rs_timer(struct timer_list *t)
3750 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3751 struct net_device *dev = idev->dev;
3752 struct in6_addr lladdr;
3754 write_lock(&idev->lock);
3755 if (idev->dead || !(idev->if_flags & IF_READY))
3756 goto out;
3758 if (!ipv6_accept_ra(idev))
3759 goto out;
3761 /* Announcement received after solicitation was sent */
3762 if (idev->if_flags & IF_RA_RCVD)
3763 goto out;
3765 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3766 write_unlock(&idev->lock);
3767 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3768 ndisc_send_rs(dev, &lladdr,
3769 &in6addr_linklocal_allrouters);
3770 else
3771 goto put;
3773 write_lock(&idev->lock);
3774 idev->rs_interval = rfc3315_s14_backoff_update(
3775 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3776 /* The wait after the last probe can be shorter */
3777 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3778 idev->cnf.rtr_solicits) ?
3779 idev->cnf.rtr_solicit_delay :
3780 idev->rs_interval);
3781 } else {
3783 * Note: we do not support deprecated "all on-link"
3784 * assumption any longer.
3786 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3789 out:
3790 write_unlock(&idev->lock);
3791 put:
3792 in6_dev_put(idev);
3796 * Duplicate Address Detection
3798 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3800 unsigned long rand_num;
3801 struct inet6_dev *idev = ifp->idev;
3802 u64 nonce;
3804 if (ifp->flags & IFA_F_OPTIMISTIC)
3805 rand_num = 0;
3806 else
3807 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3809 nonce = 0;
3810 if (idev->cnf.enhanced_dad ||
3811 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3813 get_random_bytes(&nonce, 6);
3814 while (nonce == 0);
3816 ifp->dad_nonce = nonce;
3817 ifp->dad_probes = idev->cnf.dad_transmits;
3818 addrconf_mod_dad_work(ifp, rand_num);
3821 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3823 struct inet6_dev *idev = ifp->idev;
3824 struct net_device *dev = idev->dev;
3825 bool bump_id, notify = false;
3827 addrconf_join_solict(dev, &ifp->addr);
3829 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3831 read_lock_bh(&idev->lock);
3832 spin_lock(&ifp->lock);
3833 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3834 goto out;
3836 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3837 (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3838 idev->cnf.accept_dad < 1) ||
3839 !(ifp->flags&IFA_F_TENTATIVE) ||
3840 ifp->flags & IFA_F_NODAD) {
3841 bump_id = ifp->flags & IFA_F_TENTATIVE;
3842 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3843 spin_unlock(&ifp->lock);
3844 read_unlock_bh(&idev->lock);
3846 addrconf_dad_completed(ifp, bump_id);
3847 return;
3850 if (!(idev->if_flags & IF_READY)) {
3851 spin_unlock(&ifp->lock);
3852 read_unlock_bh(&idev->lock);
3854 * If the device is not ready:
3855 * - keep it tentative if it is a permanent address.
3856 * - otherwise, kill it.
3858 in6_ifa_hold(ifp);
3859 addrconf_dad_stop(ifp, 0);
3860 return;
3864 * Optimistic nodes can start receiving
3865 * Frames right away
3867 if (ifp->flags & IFA_F_OPTIMISTIC) {
3868 ip6_ins_rt(ifp->rt);
3869 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3870 /* Because optimistic nodes can use this address,
3871 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3873 notify = true;
3877 addrconf_dad_kick(ifp);
3878 out:
3879 spin_unlock(&ifp->lock);
3880 read_unlock_bh(&idev->lock);
3881 if (notify)
3882 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3885 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3887 bool begin_dad = false;
3889 spin_lock_bh(&ifp->lock);
3890 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3891 ifp->state = INET6_IFADDR_STATE_PREDAD;
3892 begin_dad = true;
3894 spin_unlock_bh(&ifp->lock);
3896 if (begin_dad)
3897 addrconf_mod_dad_work(ifp, 0);
3900 static void addrconf_dad_work(struct work_struct *w)
3902 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3903 struct inet6_ifaddr,
3904 dad_work);
3905 struct inet6_dev *idev = ifp->idev;
3906 bool bump_id, disable_ipv6 = false;
3907 struct in6_addr mcaddr;
3909 enum {
3910 DAD_PROCESS,
3911 DAD_BEGIN,
3912 DAD_ABORT,
3913 } action = DAD_PROCESS;
3915 rtnl_lock();
3917 spin_lock_bh(&ifp->lock);
3918 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3919 action = DAD_BEGIN;
3920 ifp->state = INET6_IFADDR_STATE_DAD;
3921 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3922 action = DAD_ABORT;
3923 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3925 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3926 idev->cnf.accept_dad > 1) &&
3927 !idev->cnf.disable_ipv6 &&
3928 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3929 struct in6_addr addr;
3931 addr.s6_addr32[0] = htonl(0xfe800000);
3932 addr.s6_addr32[1] = 0;
3934 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3935 ipv6_addr_equal(&ifp->addr, &addr)) {
3936 /* DAD failed for link-local based on MAC */
3937 idev->cnf.disable_ipv6 = 1;
3939 pr_info("%s: IPv6 being disabled!\n",
3940 ifp->idev->dev->name);
3941 disable_ipv6 = true;
3945 spin_unlock_bh(&ifp->lock);
3947 if (action == DAD_BEGIN) {
3948 addrconf_dad_begin(ifp);
3949 goto out;
3950 } else if (action == DAD_ABORT) {
3951 in6_ifa_hold(ifp);
3952 addrconf_dad_stop(ifp, 1);
3953 if (disable_ipv6)
3954 addrconf_ifdown(idev->dev, 0);
3955 goto out;
3958 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3959 goto out;
3961 write_lock_bh(&idev->lock);
3962 if (idev->dead || !(idev->if_flags & IF_READY)) {
3963 write_unlock_bh(&idev->lock);
3964 goto out;
3967 spin_lock(&ifp->lock);
3968 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3969 spin_unlock(&ifp->lock);
3970 write_unlock_bh(&idev->lock);
3971 goto out;
3974 if (ifp->dad_probes == 0) {
3976 * DAD was successful
3979 bump_id = ifp->flags & IFA_F_TENTATIVE;
3980 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3981 spin_unlock(&ifp->lock);
3982 write_unlock_bh(&idev->lock);
3984 addrconf_dad_completed(ifp, bump_id);
3986 goto out;
3989 ifp->dad_probes--;
3990 addrconf_mod_dad_work(ifp,
3991 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3992 spin_unlock(&ifp->lock);
3993 write_unlock_bh(&idev->lock);
3995 /* send a neighbour solicitation for our addr */
3996 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3997 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3998 ifp->dad_nonce);
3999 out:
4000 in6_ifa_put(ifp);
4001 rtnl_unlock();
4004 /* ifp->idev must be at least read locked */
4005 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4007 struct inet6_ifaddr *ifpiter;
4008 struct inet6_dev *idev = ifp->idev;
4010 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4011 if (ifpiter->scope > IFA_LINK)
4012 break;
4013 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4014 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4015 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4016 IFA_F_PERMANENT)
4017 return false;
4019 return true;
4022 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
4024 struct net_device *dev = ifp->idev->dev;
4025 struct in6_addr lladdr;
4026 bool send_rs, send_mld;
4028 addrconf_del_dad_work(ifp);
4031 * Configure the address for reception. Now it is valid.
4034 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4036 /* If added prefix is link local and we are prepared to process
4037 router advertisements, start sending router solicitations.
4040 read_lock_bh(&ifp->idev->lock);
4041 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4042 send_rs = send_mld &&
4043 ipv6_accept_ra(ifp->idev) &&
4044 ifp->idev->cnf.rtr_solicits != 0 &&
4045 (dev->flags&IFF_LOOPBACK) == 0;
4046 read_unlock_bh(&ifp->idev->lock);
4048 /* While dad is in progress mld report's source address is in6_addrany.
4049 * Resend with proper ll now.
4051 if (send_mld)
4052 ipv6_mc_dad_complete(ifp->idev);
4054 if (send_rs) {
4056 * If a host as already performed a random delay
4057 * [...] as part of DAD [...] there is no need
4058 * to delay again before sending the first RS
4060 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4061 return;
4062 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4064 write_lock_bh(&ifp->idev->lock);
4065 spin_lock(&ifp->lock);
4066 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4067 ifp->idev->cnf.rtr_solicit_interval);
4068 ifp->idev->rs_probes = 1;
4069 ifp->idev->if_flags |= IF_RS_SENT;
4070 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4071 spin_unlock(&ifp->lock);
4072 write_unlock_bh(&ifp->idev->lock);
4075 if (bump_id)
4076 rt_genid_bump_ipv6(dev_net(dev));
4078 /* Make sure that a new temporary address will be created
4079 * before this temporary address becomes deprecated.
4081 if (ifp->flags & IFA_F_TEMPORARY)
4082 addrconf_verify_rtnl();
4085 static void addrconf_dad_run(struct inet6_dev *idev)
4087 struct inet6_ifaddr *ifp;
4089 read_lock_bh(&idev->lock);
4090 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4091 spin_lock(&ifp->lock);
4092 if (ifp->flags & IFA_F_TENTATIVE &&
4093 ifp->state == INET6_IFADDR_STATE_DAD)
4094 addrconf_dad_kick(ifp);
4095 spin_unlock(&ifp->lock);
4097 read_unlock_bh(&idev->lock);
4100 #ifdef CONFIG_PROC_FS
4101 struct if6_iter_state {
4102 struct seq_net_private p;
4103 int bucket;
4104 int offset;
4107 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4109 struct if6_iter_state *state = seq->private;
4110 struct net *net = seq_file_net(seq);
4111 struct inet6_ifaddr *ifa = NULL;
4112 int p = 0;
4114 /* initial bucket if pos is 0 */
4115 if (pos == 0) {
4116 state->bucket = 0;
4117 state->offset = 0;
4120 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4121 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4122 addr_lst) {
4123 if (!net_eq(dev_net(ifa->idev->dev), net))
4124 continue;
4125 /* sync with offset */
4126 if (p < state->offset) {
4127 p++;
4128 continue;
4130 state->offset++;
4131 return ifa;
4134 /* prepare for next bucket */
4135 state->offset = 0;
4136 p = 0;
4138 return NULL;
4141 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4142 struct inet6_ifaddr *ifa)
4144 struct if6_iter_state *state = seq->private;
4145 struct net *net = seq_file_net(seq);
4147 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4148 if (!net_eq(dev_net(ifa->idev->dev), net))
4149 continue;
4150 state->offset++;
4151 return ifa;
4154 while (++state->bucket < IN6_ADDR_HSIZE) {
4155 state->offset = 0;
4156 hlist_for_each_entry_rcu(ifa,
4157 &inet6_addr_lst[state->bucket], addr_lst) {
4158 if (!net_eq(dev_net(ifa->idev->dev), net))
4159 continue;
4160 state->offset++;
4161 return ifa;
4165 return NULL;
4168 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4169 __acquires(rcu)
4171 rcu_read_lock();
4172 return if6_get_first(seq, *pos);
4175 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4177 struct inet6_ifaddr *ifa;
4179 ifa = if6_get_next(seq, v);
4180 ++*pos;
4181 return ifa;
4184 static void if6_seq_stop(struct seq_file *seq, void *v)
4185 __releases(rcu)
4187 rcu_read_unlock();
4190 static int if6_seq_show(struct seq_file *seq, void *v)
4192 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4193 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4194 &ifp->addr,
4195 ifp->idev->dev->ifindex,
4196 ifp->prefix_len,
4197 ifp->scope,
4198 (u8) ifp->flags,
4199 ifp->idev->dev->name);
4200 return 0;
4203 static const struct seq_operations if6_seq_ops = {
4204 .start = if6_seq_start,
4205 .next = if6_seq_next,
4206 .show = if6_seq_show,
4207 .stop = if6_seq_stop,
4210 static int if6_seq_open(struct inode *inode, struct file *file)
4212 return seq_open_net(inode, file, &if6_seq_ops,
4213 sizeof(struct if6_iter_state));
4216 static const struct file_operations if6_fops = {
4217 .open = if6_seq_open,
4218 .read = seq_read,
4219 .llseek = seq_lseek,
4220 .release = seq_release_net,
4223 static int __net_init if6_proc_net_init(struct net *net)
4225 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4226 return -ENOMEM;
4227 return 0;
4230 static void __net_exit if6_proc_net_exit(struct net *net)
4232 remove_proc_entry("if_inet6", net->proc_net);
4235 static struct pernet_operations if6_proc_net_ops = {
4236 .init = if6_proc_net_init,
4237 .exit = if6_proc_net_exit,
4240 int __init if6_proc_init(void)
4242 return register_pernet_subsys(&if6_proc_net_ops);
4245 void if6_proc_exit(void)
4247 unregister_pernet_subsys(&if6_proc_net_ops);
4249 #endif /* CONFIG_PROC_FS */
4251 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4252 /* Check if address is a home address configured on any interface. */
4253 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4255 unsigned int hash = inet6_addr_hash(net, addr);
4256 struct inet6_ifaddr *ifp = NULL;
4257 int ret = 0;
4259 rcu_read_lock();
4260 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4261 if (!net_eq(dev_net(ifp->idev->dev), net))
4262 continue;
4263 if (ipv6_addr_equal(&ifp->addr, addr) &&
4264 (ifp->flags & IFA_F_HOMEADDRESS)) {
4265 ret = 1;
4266 break;
4269 rcu_read_unlock();
4270 return ret;
4272 #endif
4275 * Periodic address status verification
4278 static void addrconf_verify_rtnl(void)
4280 unsigned long now, next, next_sec, next_sched;
4281 struct inet6_ifaddr *ifp;
4282 int i;
4284 ASSERT_RTNL();
4286 rcu_read_lock_bh();
4287 now = jiffies;
4288 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4290 cancel_delayed_work(&addr_chk_work);
4292 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4293 restart:
4294 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4295 unsigned long age;
4297 /* When setting preferred_lft to a value not zero or
4298 * infinity, while valid_lft is infinity
4299 * IFA_F_PERMANENT has a non-infinity life time.
4301 if ((ifp->flags & IFA_F_PERMANENT) &&
4302 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4303 continue;
4305 spin_lock(&ifp->lock);
4306 /* We try to batch several events at once. */
4307 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4309 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4310 age >= ifp->valid_lft) {
4311 spin_unlock(&ifp->lock);
4312 in6_ifa_hold(ifp);
4313 ipv6_del_addr(ifp);
4314 goto restart;
4315 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4316 spin_unlock(&ifp->lock);
4317 continue;
4318 } else if (age >= ifp->prefered_lft) {
4319 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4320 int deprecate = 0;
4322 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4323 deprecate = 1;
4324 ifp->flags |= IFA_F_DEPRECATED;
4327 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4328 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4329 next = ifp->tstamp + ifp->valid_lft * HZ;
4331 spin_unlock(&ifp->lock);
4333 if (deprecate) {
4334 in6_ifa_hold(ifp);
4336 ipv6_ifa_notify(0, ifp);
4337 in6_ifa_put(ifp);
4338 goto restart;
4340 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4341 !(ifp->flags&IFA_F_TENTATIVE)) {
4342 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4343 ifp->idev->cnf.dad_transmits *
4344 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4346 if (age >= ifp->prefered_lft - regen_advance) {
4347 struct inet6_ifaddr *ifpub = ifp->ifpub;
4348 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4349 next = ifp->tstamp + ifp->prefered_lft * HZ;
4350 if (!ifp->regen_count && ifpub) {
4351 ifp->regen_count++;
4352 in6_ifa_hold(ifp);
4353 in6_ifa_hold(ifpub);
4354 spin_unlock(&ifp->lock);
4356 spin_lock(&ifpub->lock);
4357 ifpub->regen_count = 0;
4358 spin_unlock(&ifpub->lock);
4359 ipv6_create_tempaddr(ifpub, ifp, true);
4360 in6_ifa_put(ifpub);
4361 in6_ifa_put(ifp);
4362 goto restart;
4364 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4365 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4366 spin_unlock(&ifp->lock);
4367 } else {
4368 /* ifp->prefered_lft <= ifp->valid_lft */
4369 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4370 next = ifp->tstamp + ifp->prefered_lft * HZ;
4371 spin_unlock(&ifp->lock);
4376 next_sec = round_jiffies_up(next);
4377 next_sched = next;
4379 /* If rounded timeout is accurate enough, accept it. */
4380 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4381 next_sched = next_sec;
4383 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4384 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4385 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4387 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4388 now, next, next_sec, next_sched);
4389 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4390 rcu_read_unlock_bh();
4393 static void addrconf_verify_work(struct work_struct *w)
4395 rtnl_lock();
4396 addrconf_verify_rtnl();
4397 rtnl_unlock();
4400 static void addrconf_verify(void)
4402 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4405 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4406 struct in6_addr **peer_pfx)
4408 struct in6_addr *pfx = NULL;
4410 *peer_pfx = NULL;
4412 if (addr)
4413 pfx = nla_data(addr);
4415 if (local) {
4416 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4417 *peer_pfx = pfx;
4418 pfx = nla_data(local);
4421 return pfx;
4424 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4425 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4426 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4427 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4428 [IFA_FLAGS] = { .len = sizeof(u32) },
4431 static int
4432 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4433 struct netlink_ext_ack *extack)
4435 struct net *net = sock_net(skb->sk);
4436 struct ifaddrmsg *ifm;
4437 struct nlattr *tb[IFA_MAX+1];
4438 struct in6_addr *pfx, *peer_pfx;
4439 u32 ifa_flags;
4440 int err;
4442 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4443 extack);
4444 if (err < 0)
4445 return err;
4447 ifm = nlmsg_data(nlh);
4448 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4449 if (!pfx)
4450 return -EINVAL;
4452 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4454 /* We ignore other flags so far. */
4455 ifa_flags &= IFA_F_MANAGETEMPADDR;
4457 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4458 ifm->ifa_prefixlen);
4461 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4462 u32 prefered_lft, u32 valid_lft)
4464 u32 flags;
4465 clock_t expires;
4466 unsigned long timeout;
4467 bool was_managetempaddr;
4468 bool had_prefixroute;
4470 ASSERT_RTNL();
4472 if (!valid_lft || (prefered_lft > valid_lft))
4473 return -EINVAL;
4475 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4476 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4477 return -EINVAL;
4479 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4480 if (addrconf_finite_timeout(timeout)) {
4481 expires = jiffies_to_clock_t(timeout * HZ);
4482 valid_lft = timeout;
4483 flags = RTF_EXPIRES;
4484 } else {
4485 expires = 0;
4486 flags = 0;
4487 ifa_flags |= IFA_F_PERMANENT;
4490 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4491 if (addrconf_finite_timeout(timeout)) {
4492 if (timeout == 0)
4493 ifa_flags |= IFA_F_DEPRECATED;
4494 prefered_lft = timeout;
4497 spin_lock_bh(&ifp->lock);
4498 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4499 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4500 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4501 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4502 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4503 IFA_F_NOPREFIXROUTE);
4504 ifp->flags |= ifa_flags;
4505 ifp->tstamp = jiffies;
4506 ifp->valid_lft = valid_lft;
4507 ifp->prefered_lft = prefered_lft;
4509 spin_unlock_bh(&ifp->lock);
4510 if (!(ifp->flags&IFA_F_TENTATIVE))
4511 ipv6_ifa_notify(0, ifp);
4513 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4514 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4515 expires, flags);
4516 } else if (had_prefixroute) {
4517 enum cleanup_prefix_rt_t action;
4518 unsigned long rt_expires;
4520 write_lock_bh(&ifp->idev->lock);
4521 action = check_cleanup_prefix_route(ifp, &rt_expires);
4522 write_unlock_bh(&ifp->idev->lock);
4524 if (action != CLEANUP_PREFIX_RT_NOP) {
4525 cleanup_prefix_route(ifp, rt_expires,
4526 action == CLEANUP_PREFIX_RT_DEL);
4530 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4531 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4532 valid_lft = prefered_lft = 0;
4533 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4534 !was_managetempaddr, jiffies);
4537 addrconf_verify_rtnl();
4539 return 0;
4542 static int
4543 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4544 struct netlink_ext_ack *extack)
4546 struct net *net = sock_net(skb->sk);
4547 struct ifaddrmsg *ifm;
4548 struct nlattr *tb[IFA_MAX+1];
4549 struct in6_addr *pfx, *peer_pfx;
4550 struct inet6_ifaddr *ifa;
4551 struct net_device *dev;
4552 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4553 u32 ifa_flags;
4554 int err;
4556 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4557 extack);
4558 if (err < 0)
4559 return err;
4561 ifm = nlmsg_data(nlh);
4562 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4563 if (!pfx)
4564 return -EINVAL;
4566 if (tb[IFA_CACHEINFO]) {
4567 struct ifa_cacheinfo *ci;
4569 ci = nla_data(tb[IFA_CACHEINFO]);
4570 valid_lft = ci->ifa_valid;
4571 preferred_lft = ci->ifa_prefered;
4572 } else {
4573 preferred_lft = INFINITY_LIFE_TIME;
4574 valid_lft = INFINITY_LIFE_TIME;
4577 dev = __dev_get_by_index(net, ifm->ifa_index);
4578 if (!dev)
4579 return -ENODEV;
4581 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4583 /* We ignore other flags so far. */
4584 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4585 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4587 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4588 if (!ifa) {
4590 * It would be best to check for !NLM_F_CREATE here but
4591 * userspace already relies on not having to provide this.
4593 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4594 ifm->ifa_prefixlen, ifa_flags,
4595 preferred_lft, valid_lft, extack);
4598 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4599 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4600 err = -EEXIST;
4601 else
4602 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4604 in6_ifa_put(ifa);
4606 return err;
4609 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4610 u8 scope, int ifindex)
4612 struct ifaddrmsg *ifm;
4614 ifm = nlmsg_data(nlh);
4615 ifm->ifa_family = AF_INET6;
4616 ifm->ifa_prefixlen = prefixlen;
4617 ifm->ifa_flags = flags;
4618 ifm->ifa_scope = scope;
4619 ifm->ifa_index = ifindex;
4622 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4623 unsigned long tstamp, u32 preferred, u32 valid)
4625 struct ifa_cacheinfo ci;
4627 ci.cstamp = cstamp_delta(cstamp);
4628 ci.tstamp = cstamp_delta(tstamp);
4629 ci.ifa_prefered = preferred;
4630 ci.ifa_valid = valid;
4632 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4635 static inline int rt_scope(int ifa_scope)
4637 if (ifa_scope & IFA_HOST)
4638 return RT_SCOPE_HOST;
4639 else if (ifa_scope & IFA_LINK)
4640 return RT_SCOPE_LINK;
4641 else if (ifa_scope & IFA_SITE)
4642 return RT_SCOPE_SITE;
4643 else
4644 return RT_SCOPE_UNIVERSE;
4647 static inline int inet6_ifaddr_msgsize(void)
4649 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4650 + nla_total_size(16) /* IFA_LOCAL */
4651 + nla_total_size(16) /* IFA_ADDRESS */
4652 + nla_total_size(sizeof(struct ifa_cacheinfo))
4653 + nla_total_size(4) /* IFA_FLAGS */;
4656 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4657 u32 portid, u32 seq, int event, unsigned int flags)
4659 struct nlmsghdr *nlh;
4660 u32 preferred, valid;
4662 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4663 if (!nlh)
4664 return -EMSGSIZE;
4666 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4667 ifa->idev->dev->ifindex);
4669 if (!((ifa->flags&IFA_F_PERMANENT) &&
4670 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4671 preferred = ifa->prefered_lft;
4672 valid = ifa->valid_lft;
4673 if (preferred != INFINITY_LIFE_TIME) {
4674 long tval = (jiffies - ifa->tstamp)/HZ;
4675 if (preferred > tval)
4676 preferred -= tval;
4677 else
4678 preferred = 0;
4679 if (valid != INFINITY_LIFE_TIME) {
4680 if (valid > tval)
4681 valid -= tval;
4682 else
4683 valid = 0;
4686 } else {
4687 preferred = INFINITY_LIFE_TIME;
4688 valid = INFINITY_LIFE_TIME;
4691 if (!ipv6_addr_any(&ifa->peer_addr)) {
4692 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4693 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4694 goto error;
4695 } else
4696 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4697 goto error;
4699 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4700 goto error;
4702 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4703 goto error;
4705 nlmsg_end(skb, nlh);
4706 return 0;
4708 error:
4709 nlmsg_cancel(skb, nlh);
4710 return -EMSGSIZE;
4713 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4714 u32 portid, u32 seq, int event, u16 flags)
4716 struct nlmsghdr *nlh;
4717 u8 scope = RT_SCOPE_UNIVERSE;
4718 int ifindex = ifmca->idev->dev->ifindex;
4720 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4721 scope = RT_SCOPE_SITE;
4723 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4724 if (!nlh)
4725 return -EMSGSIZE;
4727 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4728 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4729 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4730 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4731 nlmsg_cancel(skb, nlh);
4732 return -EMSGSIZE;
4735 nlmsg_end(skb, nlh);
4736 return 0;
4739 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4740 u32 portid, u32 seq, int event, unsigned int flags)
4742 struct nlmsghdr *nlh;
4743 u8 scope = RT_SCOPE_UNIVERSE;
4744 int ifindex = ifaca->aca_idev->dev->ifindex;
4746 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4747 scope = RT_SCOPE_SITE;
4749 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4750 if (!nlh)
4751 return -EMSGSIZE;
4753 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4754 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4755 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4756 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4757 nlmsg_cancel(skb, nlh);
4758 return -EMSGSIZE;
4761 nlmsg_end(skb, nlh);
4762 return 0;
4765 enum addr_type_t {
4766 UNICAST_ADDR,
4767 MULTICAST_ADDR,
4768 ANYCAST_ADDR,
4771 /* called with rcu_read_lock() */
4772 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4773 struct netlink_callback *cb, enum addr_type_t type,
4774 int s_ip_idx, int *p_ip_idx)
4776 struct ifmcaddr6 *ifmca;
4777 struct ifacaddr6 *ifaca;
4778 int err = 1;
4779 int ip_idx = *p_ip_idx;
4781 read_lock_bh(&idev->lock);
4782 switch (type) {
4783 case UNICAST_ADDR: {
4784 struct inet6_ifaddr *ifa;
4786 /* unicast address incl. temp addr */
4787 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4788 if (++ip_idx < s_ip_idx)
4789 continue;
4790 err = inet6_fill_ifaddr(skb, ifa,
4791 NETLINK_CB(cb->skb).portid,
4792 cb->nlh->nlmsg_seq,
4793 RTM_NEWADDR,
4794 NLM_F_MULTI);
4795 if (err < 0)
4796 break;
4797 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4799 break;
4801 case MULTICAST_ADDR:
4802 /* multicast address */
4803 for (ifmca = idev->mc_list; ifmca;
4804 ifmca = ifmca->next, ip_idx++) {
4805 if (ip_idx < s_ip_idx)
4806 continue;
4807 err = inet6_fill_ifmcaddr(skb, ifmca,
4808 NETLINK_CB(cb->skb).portid,
4809 cb->nlh->nlmsg_seq,
4810 RTM_GETMULTICAST,
4811 NLM_F_MULTI);
4812 if (err < 0)
4813 break;
4815 break;
4816 case ANYCAST_ADDR:
4817 /* anycast address */
4818 for (ifaca = idev->ac_list; ifaca;
4819 ifaca = ifaca->aca_next, ip_idx++) {
4820 if (ip_idx < s_ip_idx)
4821 continue;
4822 err = inet6_fill_ifacaddr(skb, ifaca,
4823 NETLINK_CB(cb->skb).portid,
4824 cb->nlh->nlmsg_seq,
4825 RTM_GETANYCAST,
4826 NLM_F_MULTI);
4827 if (err < 0)
4828 break;
4830 break;
4831 default:
4832 break;
4834 read_unlock_bh(&idev->lock);
4835 *p_ip_idx = ip_idx;
4836 return err;
4839 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4840 enum addr_type_t type)
4842 struct net *net = sock_net(skb->sk);
4843 int h, s_h;
4844 int idx, ip_idx;
4845 int s_idx, s_ip_idx;
4846 struct net_device *dev;
4847 struct inet6_dev *idev;
4848 struct hlist_head *head;
4850 s_h = cb->args[0];
4851 s_idx = idx = cb->args[1];
4852 s_ip_idx = ip_idx = cb->args[2];
4854 rcu_read_lock();
4855 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4856 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4857 idx = 0;
4858 head = &net->dev_index_head[h];
4859 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4860 if (idx < s_idx)
4861 goto cont;
4862 if (h > s_h || idx > s_idx)
4863 s_ip_idx = 0;
4864 ip_idx = 0;
4865 idev = __in6_dev_get(dev);
4866 if (!idev)
4867 goto cont;
4869 if (in6_dump_addrs(idev, skb, cb, type,
4870 s_ip_idx, &ip_idx) < 0)
4871 goto done;
4872 cont:
4873 idx++;
4876 done:
4877 rcu_read_unlock();
4878 cb->args[0] = h;
4879 cb->args[1] = idx;
4880 cb->args[2] = ip_idx;
4882 return skb->len;
4885 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4887 enum addr_type_t type = UNICAST_ADDR;
4889 return inet6_dump_addr(skb, cb, type);
4892 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4894 enum addr_type_t type = MULTICAST_ADDR;
4896 return inet6_dump_addr(skb, cb, type);
4900 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4902 enum addr_type_t type = ANYCAST_ADDR;
4904 return inet6_dump_addr(skb, cb, type);
4907 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4908 struct netlink_ext_ack *extack)
4910 struct net *net = sock_net(in_skb->sk);
4911 struct ifaddrmsg *ifm;
4912 struct nlattr *tb[IFA_MAX+1];
4913 struct in6_addr *addr = NULL, *peer;
4914 struct net_device *dev = NULL;
4915 struct inet6_ifaddr *ifa;
4916 struct sk_buff *skb;
4917 int err;
4919 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4920 extack);
4921 if (err < 0)
4922 return err;
4924 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4925 if (!addr)
4926 return -EINVAL;
4928 ifm = nlmsg_data(nlh);
4929 if (ifm->ifa_index)
4930 dev = dev_get_by_index(net, ifm->ifa_index);
4932 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4933 if (!ifa) {
4934 err = -EADDRNOTAVAIL;
4935 goto errout;
4938 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4939 if (!skb) {
4940 err = -ENOBUFS;
4941 goto errout_ifa;
4944 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4945 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4946 if (err < 0) {
4947 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4948 WARN_ON(err == -EMSGSIZE);
4949 kfree_skb(skb);
4950 goto errout_ifa;
4952 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4953 errout_ifa:
4954 in6_ifa_put(ifa);
4955 errout:
4956 if (dev)
4957 dev_put(dev);
4958 return err;
4961 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4963 struct sk_buff *skb;
4964 struct net *net = dev_net(ifa->idev->dev);
4965 int err = -ENOBUFS;
4967 /* Don't send DELADDR notification for TENTATIVE address,
4968 * since NEWADDR notification is sent only after removing
4969 * TENTATIVE flag, if DAD has not failed.
4971 if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
4972 event == RTM_DELADDR)
4973 return;
4975 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4976 if (!skb)
4977 goto errout;
4979 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4980 if (err < 0) {
4981 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4982 WARN_ON(err == -EMSGSIZE);
4983 kfree_skb(skb);
4984 goto errout;
4986 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4987 return;
4988 errout:
4989 if (err < 0)
4990 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4993 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4994 __s32 *array, int bytes)
4996 BUG_ON(bytes < (DEVCONF_MAX * 4));
4998 memset(array, 0, bytes);
4999 array[DEVCONF_FORWARDING] = cnf->forwarding;
5000 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5001 array[DEVCONF_MTU6] = cnf->mtu6;
5002 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5003 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5004 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5005 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5006 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5007 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5008 jiffies_to_msecs(cnf->rtr_solicit_interval);
5009 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5010 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5011 array[DEVCONF_RTR_SOLICIT_DELAY] =
5012 jiffies_to_msecs(cnf->rtr_solicit_delay);
5013 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5014 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5015 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5016 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5017 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5018 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5019 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5020 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5021 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5022 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5023 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5024 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5025 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5026 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5027 #ifdef CONFIG_IPV6_ROUTER_PREF
5028 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5029 array[DEVCONF_RTR_PROBE_INTERVAL] =
5030 jiffies_to_msecs(cnf->rtr_probe_interval);
5031 #ifdef CONFIG_IPV6_ROUTE_INFO
5032 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5033 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5034 #endif
5035 #endif
5036 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5037 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5038 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5039 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5040 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5041 #endif
5042 #ifdef CONFIG_IPV6_MROUTE
5043 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5044 #endif
5045 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5046 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5047 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5048 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5049 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5050 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5051 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5052 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5053 /* we omit DEVCONF_STABLE_SECRET for now */
5054 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5055 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5056 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5057 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5058 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5059 #ifdef CONFIG_IPV6_SEG6_HMAC
5060 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5061 #endif
5062 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5063 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5064 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5065 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5068 static inline size_t inet6_ifla6_size(void)
5070 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5071 + nla_total_size(sizeof(struct ifla_cacheinfo))
5072 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5073 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5074 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5075 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5078 static inline size_t inet6_if_nlmsg_size(void)
5080 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5081 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5082 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5083 + nla_total_size(4) /* IFLA_MTU */
5084 + nla_total_size(4) /* IFLA_LINK */
5085 + nla_total_size(1) /* IFLA_OPERSTATE */
5086 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5089 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5090 int bytes)
5092 int i;
5093 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5094 BUG_ON(pad < 0);
5096 /* Use put_unaligned() because stats may not be aligned for u64. */
5097 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5098 for (i = 1; i < ICMP6_MIB_MAX; i++)
5099 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5101 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5104 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5105 int bytes, size_t syncpoff)
5107 int i, c;
5108 u64 buff[IPSTATS_MIB_MAX];
5109 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5111 BUG_ON(pad < 0);
5113 memset(buff, 0, sizeof(buff));
5114 buff[0] = IPSTATS_MIB_MAX;
5116 for_each_possible_cpu(c) {
5117 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5118 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5121 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5122 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5125 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5126 int bytes)
5128 switch (attrtype) {
5129 case IFLA_INET6_STATS:
5130 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5131 offsetof(struct ipstats_mib, syncp));
5132 break;
5133 case IFLA_INET6_ICMP6STATS:
5134 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5135 break;
5139 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5140 u32 ext_filter_mask)
5142 struct nlattr *nla;
5143 struct ifla_cacheinfo ci;
5145 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5146 goto nla_put_failure;
5147 ci.max_reasm_len = IPV6_MAXPLEN;
5148 ci.tstamp = cstamp_delta(idev->tstamp);
5149 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5150 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5151 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5152 goto nla_put_failure;
5153 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5154 if (!nla)
5155 goto nla_put_failure;
5156 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5158 /* XXX - MC not implemented */
5160 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5161 return 0;
5163 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5164 if (!nla)
5165 goto nla_put_failure;
5166 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5168 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5169 if (!nla)
5170 goto nla_put_failure;
5171 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5173 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5174 if (!nla)
5175 goto nla_put_failure;
5177 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5178 goto nla_put_failure;
5180 read_lock_bh(&idev->lock);
5181 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5182 read_unlock_bh(&idev->lock);
5184 return 0;
5186 nla_put_failure:
5187 return -EMSGSIZE;
5190 static size_t inet6_get_link_af_size(const struct net_device *dev,
5191 u32 ext_filter_mask)
5193 if (!__in6_dev_get(dev))
5194 return 0;
5196 return inet6_ifla6_size();
5199 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5200 u32 ext_filter_mask)
5202 struct inet6_dev *idev = __in6_dev_get(dev);
5204 if (!idev)
5205 return -ENODATA;
5207 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5208 return -EMSGSIZE;
5210 return 0;
5213 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5215 struct inet6_ifaddr *ifp;
5216 struct net_device *dev = idev->dev;
5217 bool clear_token, update_rs = false;
5218 struct in6_addr ll_addr;
5220 ASSERT_RTNL();
5222 if (!token)
5223 return -EINVAL;
5224 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5225 return -EINVAL;
5226 if (!ipv6_accept_ra(idev))
5227 return -EINVAL;
5228 if (idev->cnf.rtr_solicits == 0)
5229 return -EINVAL;
5231 write_lock_bh(&idev->lock);
5233 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5234 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5236 write_unlock_bh(&idev->lock);
5238 clear_token = ipv6_addr_any(token);
5239 if (clear_token)
5240 goto update_lft;
5242 if (!idev->dead && (idev->if_flags & IF_READY) &&
5243 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5244 IFA_F_OPTIMISTIC)) {
5245 /* If we're not ready, then normal ifup will take care
5246 * of this. Otherwise, we need to request our rs here.
5248 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5249 update_rs = true;
5252 update_lft:
5253 write_lock_bh(&idev->lock);
5255 if (update_rs) {
5256 idev->if_flags |= IF_RS_SENT;
5257 idev->rs_interval = rfc3315_s14_backoff_init(
5258 idev->cnf.rtr_solicit_interval);
5259 idev->rs_probes = 1;
5260 addrconf_mod_rs_timer(idev, idev->rs_interval);
5263 /* Well, that's kinda nasty ... */
5264 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5265 spin_lock(&ifp->lock);
5266 if (ifp->tokenized) {
5267 ifp->valid_lft = 0;
5268 ifp->prefered_lft = 0;
5270 spin_unlock(&ifp->lock);
5273 write_unlock_bh(&idev->lock);
5274 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5275 addrconf_verify_rtnl();
5276 return 0;
5279 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5280 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5281 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5284 static int inet6_validate_link_af(const struct net_device *dev,
5285 const struct nlattr *nla)
5287 struct nlattr *tb[IFLA_INET6_MAX + 1];
5289 if (dev && !__in6_dev_get(dev))
5290 return -EAFNOSUPPORT;
5292 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5293 NULL);
5296 static int check_addr_gen_mode(int mode)
5298 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5299 mode != IN6_ADDR_GEN_MODE_NONE &&
5300 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5301 mode != IN6_ADDR_GEN_MODE_RANDOM)
5302 return -EINVAL;
5303 return 1;
5306 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5307 int mode)
5309 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5310 !idev->cnf.stable_secret.initialized &&
5311 !net->ipv6.devconf_dflt->stable_secret.initialized)
5312 return -EINVAL;
5313 return 1;
5316 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5318 int err = -EINVAL;
5319 struct inet6_dev *idev = __in6_dev_get(dev);
5320 struct nlattr *tb[IFLA_INET6_MAX + 1];
5322 if (!idev)
5323 return -EAFNOSUPPORT;
5325 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5326 BUG();
5328 if (tb[IFLA_INET6_TOKEN]) {
5329 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5330 if (err)
5331 return err;
5334 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5335 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5337 if (check_addr_gen_mode(mode) < 0 ||
5338 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5339 return -EINVAL;
5341 idev->cnf.addr_gen_mode = mode;
5342 err = 0;
5345 return err;
5348 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5349 u32 portid, u32 seq, int event, unsigned int flags)
5351 struct net_device *dev = idev->dev;
5352 struct ifinfomsg *hdr;
5353 struct nlmsghdr *nlh;
5354 void *protoinfo;
5356 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5357 if (!nlh)
5358 return -EMSGSIZE;
5360 hdr = nlmsg_data(nlh);
5361 hdr->ifi_family = AF_INET6;
5362 hdr->__ifi_pad = 0;
5363 hdr->ifi_type = dev->type;
5364 hdr->ifi_index = dev->ifindex;
5365 hdr->ifi_flags = dev_get_flags(dev);
5366 hdr->ifi_change = 0;
5368 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5369 (dev->addr_len &&
5370 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5371 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5372 (dev->ifindex != dev_get_iflink(dev) &&
5373 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5374 nla_put_u8(skb, IFLA_OPERSTATE,
5375 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5376 goto nla_put_failure;
5377 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5378 if (!protoinfo)
5379 goto nla_put_failure;
5381 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5382 goto nla_put_failure;
5384 nla_nest_end(skb, protoinfo);
5385 nlmsg_end(skb, nlh);
5386 return 0;
5388 nla_put_failure:
5389 nlmsg_cancel(skb, nlh);
5390 return -EMSGSIZE;
5393 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5395 struct net *net = sock_net(skb->sk);
5396 int h, s_h;
5397 int idx = 0, s_idx;
5398 struct net_device *dev;
5399 struct inet6_dev *idev;
5400 struct hlist_head *head;
5402 s_h = cb->args[0];
5403 s_idx = cb->args[1];
5405 rcu_read_lock();
5406 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5407 idx = 0;
5408 head = &net->dev_index_head[h];
5409 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5410 if (idx < s_idx)
5411 goto cont;
5412 idev = __in6_dev_get(dev);
5413 if (!idev)
5414 goto cont;
5415 if (inet6_fill_ifinfo(skb, idev,
5416 NETLINK_CB(cb->skb).portid,
5417 cb->nlh->nlmsg_seq,
5418 RTM_NEWLINK, NLM_F_MULTI) < 0)
5419 goto out;
5420 cont:
5421 idx++;
5424 out:
5425 rcu_read_unlock();
5426 cb->args[1] = idx;
5427 cb->args[0] = h;
5429 return skb->len;
5432 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5434 struct sk_buff *skb;
5435 struct net *net = dev_net(idev->dev);
5436 int err = -ENOBUFS;
5438 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5439 if (!skb)
5440 goto errout;
5442 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5443 if (err < 0) {
5444 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5445 WARN_ON(err == -EMSGSIZE);
5446 kfree_skb(skb);
5447 goto errout;
5449 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5450 return;
5451 errout:
5452 if (err < 0)
5453 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5456 static inline size_t inet6_prefix_nlmsg_size(void)
5458 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5459 + nla_total_size(sizeof(struct in6_addr))
5460 + nla_total_size(sizeof(struct prefix_cacheinfo));
5463 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5464 struct prefix_info *pinfo, u32 portid, u32 seq,
5465 int event, unsigned int flags)
5467 struct prefixmsg *pmsg;
5468 struct nlmsghdr *nlh;
5469 struct prefix_cacheinfo ci;
5471 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5472 if (!nlh)
5473 return -EMSGSIZE;
5475 pmsg = nlmsg_data(nlh);
5476 pmsg->prefix_family = AF_INET6;
5477 pmsg->prefix_pad1 = 0;
5478 pmsg->prefix_pad2 = 0;
5479 pmsg->prefix_ifindex = idev->dev->ifindex;
5480 pmsg->prefix_len = pinfo->prefix_len;
5481 pmsg->prefix_type = pinfo->type;
5482 pmsg->prefix_pad3 = 0;
5483 pmsg->prefix_flags = 0;
5484 if (pinfo->onlink)
5485 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5486 if (pinfo->autoconf)
5487 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5489 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5490 goto nla_put_failure;
5491 ci.preferred_time = ntohl(pinfo->prefered);
5492 ci.valid_time = ntohl(pinfo->valid);
5493 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5494 goto nla_put_failure;
5495 nlmsg_end(skb, nlh);
5496 return 0;
5498 nla_put_failure:
5499 nlmsg_cancel(skb, nlh);
5500 return -EMSGSIZE;
5503 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5504 struct prefix_info *pinfo)
5506 struct sk_buff *skb;
5507 struct net *net = dev_net(idev->dev);
5508 int err = -ENOBUFS;
5510 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5511 if (!skb)
5512 goto errout;
5514 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5515 if (err < 0) {
5516 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5517 WARN_ON(err == -EMSGSIZE);
5518 kfree_skb(skb);
5519 goto errout;
5521 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5522 return;
5523 errout:
5524 if (err < 0)
5525 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5528 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5530 struct net *net = dev_net(ifp->idev->dev);
5532 if (event)
5533 ASSERT_RTNL();
5535 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5537 switch (event) {
5538 case RTM_NEWADDR:
5540 * If the address was optimistic
5541 * we inserted the route at the start of
5542 * our DAD process, so we don't need
5543 * to do it again
5545 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5546 ip6_ins_rt(ifp->rt);
5547 if (ifp->idev->cnf.forwarding)
5548 addrconf_join_anycast(ifp);
5549 if (!ipv6_addr_any(&ifp->peer_addr))
5550 addrconf_prefix_route(&ifp->peer_addr, 128,
5551 ifp->idev->dev, 0, 0);
5552 break;
5553 case RTM_DELADDR:
5554 if (ifp->idev->cnf.forwarding)
5555 addrconf_leave_anycast(ifp);
5556 addrconf_leave_solict(ifp->idev, &ifp->addr);
5557 if (!ipv6_addr_any(&ifp->peer_addr)) {
5558 struct rt6_info *rt;
5560 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5561 ifp->idev->dev, 0, 0);
5562 if (rt)
5563 ip6_del_rt(rt);
5565 if (ifp->rt) {
5566 if (dst_hold_safe(&ifp->rt->dst))
5567 ip6_del_rt(ifp->rt);
5569 rt_genid_bump_ipv6(net);
5570 break;
5572 atomic_inc(&net->ipv6.dev_addr_genid);
5575 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5577 rcu_read_lock_bh();
5578 if (likely(ifp->idev->dead == 0))
5579 __ipv6_ifa_notify(event, ifp);
5580 rcu_read_unlock_bh();
5583 #ifdef CONFIG_SYSCTL
5585 static
5586 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5587 void __user *buffer, size_t *lenp, loff_t *ppos)
5589 int *valp = ctl->data;
5590 int val = *valp;
5591 loff_t pos = *ppos;
5592 struct ctl_table lctl;
5593 int ret;
5596 * ctl->data points to idev->cnf.forwarding, we should
5597 * not modify it until we get the rtnl lock.
5599 lctl = *ctl;
5600 lctl.data = &val;
5602 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5604 if (write)
5605 ret = addrconf_fixup_forwarding(ctl, valp, val);
5606 if (ret)
5607 *ppos = pos;
5608 return ret;
5611 static
5612 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5613 void __user *buffer, size_t *lenp, loff_t *ppos)
5615 struct inet6_dev *idev = ctl->extra1;
5616 int min_mtu = IPV6_MIN_MTU;
5617 struct ctl_table lctl;
5619 lctl = *ctl;
5620 lctl.extra1 = &min_mtu;
5621 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5623 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5626 static void dev_disable_change(struct inet6_dev *idev)
5628 struct netdev_notifier_info info;
5630 if (!idev || !idev->dev)
5631 return;
5633 netdev_notifier_info_init(&info, idev->dev);
5634 if (idev->cnf.disable_ipv6)
5635 addrconf_notify(NULL, NETDEV_DOWN, &info);
5636 else
5637 addrconf_notify(NULL, NETDEV_UP, &info);
5640 static void addrconf_disable_change(struct net *net, __s32 newf)
5642 struct net_device *dev;
5643 struct inet6_dev *idev;
5645 for_each_netdev(net, dev) {
5646 idev = __in6_dev_get(dev);
5647 if (idev) {
5648 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5649 idev->cnf.disable_ipv6 = newf;
5650 if (changed)
5651 dev_disable_change(idev);
5656 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5658 struct net *net;
5659 int old;
5661 if (!rtnl_trylock())
5662 return restart_syscall();
5664 net = (struct net *)table->extra2;
5665 old = *p;
5666 *p = newf;
5668 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5669 rtnl_unlock();
5670 return 0;
5673 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5674 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5675 addrconf_disable_change(net, newf);
5676 } else if ((!newf) ^ (!old))
5677 dev_disable_change((struct inet6_dev *)table->extra1);
5679 rtnl_unlock();
5680 return 0;
5683 static
5684 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5685 void __user *buffer, size_t *lenp, loff_t *ppos)
5687 int *valp = ctl->data;
5688 int val = *valp;
5689 loff_t pos = *ppos;
5690 struct ctl_table lctl;
5691 int ret;
5694 * ctl->data points to idev->cnf.disable_ipv6, we should
5695 * not modify it until we get the rtnl lock.
5697 lctl = *ctl;
5698 lctl.data = &val;
5700 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5702 if (write)
5703 ret = addrconf_disable_ipv6(ctl, valp, val);
5704 if (ret)
5705 *ppos = pos;
5706 return ret;
5709 static
5710 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5711 void __user *buffer, size_t *lenp, loff_t *ppos)
5713 int *valp = ctl->data;
5714 int ret;
5715 int old, new;
5717 old = *valp;
5718 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5719 new = *valp;
5721 if (write && old != new) {
5722 struct net *net = ctl->extra2;
5724 if (!rtnl_trylock())
5725 return restart_syscall();
5727 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5728 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5729 NETCONFA_PROXY_NEIGH,
5730 NETCONFA_IFINDEX_DEFAULT,
5731 net->ipv6.devconf_dflt);
5732 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5733 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5734 NETCONFA_PROXY_NEIGH,
5735 NETCONFA_IFINDEX_ALL,
5736 net->ipv6.devconf_all);
5737 else {
5738 struct inet6_dev *idev = ctl->extra1;
5740 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5741 NETCONFA_PROXY_NEIGH,
5742 idev->dev->ifindex,
5743 &idev->cnf);
5745 rtnl_unlock();
5748 return ret;
5751 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5752 void __user *buffer, size_t *lenp,
5753 loff_t *ppos)
5755 int ret = 0;
5756 int new_val;
5757 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5758 struct net *net = (struct net *)ctl->extra2;
5760 if (!rtnl_trylock())
5761 return restart_syscall();
5763 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5765 if (write) {
5766 new_val = *((int *)ctl->data);
5768 if (check_addr_gen_mode(new_val) < 0) {
5769 ret = -EINVAL;
5770 goto out;
5773 /* request for default */
5774 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5775 ipv6_devconf_dflt.addr_gen_mode = new_val;
5777 /* request for individual net device */
5778 } else {
5779 if (!idev)
5780 goto out;
5782 if (check_stable_privacy(idev, net, new_val) < 0) {
5783 ret = -EINVAL;
5784 goto out;
5787 if (idev->cnf.addr_gen_mode != new_val) {
5788 idev->cnf.addr_gen_mode = new_val;
5789 addrconf_dev_config(idev->dev);
5794 out:
5795 rtnl_unlock();
5797 return ret;
5800 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5801 void __user *buffer, size_t *lenp,
5802 loff_t *ppos)
5804 int err;
5805 struct in6_addr addr;
5806 char str[IPV6_MAX_STRLEN];
5807 struct ctl_table lctl = *ctl;
5808 struct net *net = ctl->extra2;
5809 struct ipv6_stable_secret *secret = ctl->data;
5811 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5812 return -EIO;
5814 lctl.maxlen = IPV6_MAX_STRLEN;
5815 lctl.data = str;
5817 if (!rtnl_trylock())
5818 return restart_syscall();
5820 if (!write && !secret->initialized) {
5821 err = -EIO;
5822 goto out;
5825 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5826 if (err >= sizeof(str)) {
5827 err = -EIO;
5828 goto out;
5831 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5832 if (err || !write)
5833 goto out;
5835 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5836 err = -EIO;
5837 goto out;
5840 secret->initialized = true;
5841 secret->secret = addr;
5843 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5844 struct net_device *dev;
5846 for_each_netdev(net, dev) {
5847 struct inet6_dev *idev = __in6_dev_get(dev);
5849 if (idev) {
5850 idev->cnf.addr_gen_mode =
5851 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5854 } else {
5855 struct inet6_dev *idev = ctl->extra1;
5857 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5860 out:
5861 rtnl_unlock();
5863 return err;
5866 static
5867 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5868 int write,
5869 void __user *buffer,
5870 size_t *lenp,
5871 loff_t *ppos)
5873 int *valp = ctl->data;
5874 int val = *valp;
5875 loff_t pos = *ppos;
5876 struct ctl_table lctl;
5877 int ret;
5879 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5880 * we should not modify it until we get the rtnl lock.
5882 lctl = *ctl;
5883 lctl.data = &val;
5885 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5887 if (write)
5888 ret = addrconf_fixup_linkdown(ctl, valp, val);
5889 if (ret)
5890 *ppos = pos;
5891 return ret;
5894 static
5895 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5897 if (rt) {
5898 if (action)
5899 rt->dst.flags |= DST_NOPOLICY;
5900 else
5901 rt->dst.flags &= ~DST_NOPOLICY;
5905 static
5906 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5908 struct inet6_ifaddr *ifa;
5910 read_lock_bh(&idev->lock);
5911 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5912 spin_lock(&ifa->lock);
5913 if (ifa->rt) {
5914 struct rt6_info *rt = ifa->rt;
5915 int cpu;
5917 rcu_read_lock();
5918 addrconf_set_nopolicy(ifa->rt, val);
5919 if (rt->rt6i_pcpu) {
5920 for_each_possible_cpu(cpu) {
5921 struct rt6_info **rtp;
5923 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5924 addrconf_set_nopolicy(*rtp, val);
5927 rcu_read_unlock();
5929 spin_unlock(&ifa->lock);
5931 read_unlock_bh(&idev->lock);
5934 static
5935 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5937 struct inet6_dev *idev;
5938 struct net *net;
5940 if (!rtnl_trylock())
5941 return restart_syscall();
5943 *valp = val;
5945 net = (struct net *)ctl->extra2;
5946 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5947 rtnl_unlock();
5948 return 0;
5951 if (valp == &net->ipv6.devconf_all->disable_policy) {
5952 struct net_device *dev;
5954 for_each_netdev(net, dev) {
5955 idev = __in6_dev_get(dev);
5956 if (idev)
5957 addrconf_disable_policy_idev(idev, val);
5959 } else {
5960 idev = (struct inet6_dev *)ctl->extra1;
5961 addrconf_disable_policy_idev(idev, val);
5964 rtnl_unlock();
5965 return 0;
5968 static
5969 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5970 void __user *buffer, size_t *lenp,
5971 loff_t *ppos)
5973 int *valp = ctl->data;
5974 int val = *valp;
5975 loff_t pos = *ppos;
5976 struct ctl_table lctl;
5977 int ret;
5979 lctl = *ctl;
5980 lctl.data = &val;
5981 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5983 if (write && (*valp != val))
5984 ret = addrconf_disable_policy(ctl, valp, val);
5986 if (ret)
5987 *ppos = pos;
5989 return ret;
5992 static int minus_one = -1;
5993 static const int zero = 0;
5994 static const int one = 1;
5995 static const int two_five_five = 255;
5997 static const struct ctl_table addrconf_sysctl[] = {
5999 .procname = "forwarding",
6000 .data = &ipv6_devconf.forwarding,
6001 .maxlen = sizeof(int),
6002 .mode = 0644,
6003 .proc_handler = addrconf_sysctl_forward,
6006 .procname = "hop_limit",
6007 .data = &ipv6_devconf.hop_limit,
6008 .maxlen = sizeof(int),
6009 .mode = 0644,
6010 .proc_handler = proc_dointvec_minmax,
6011 .extra1 = (void *)&one,
6012 .extra2 = (void *)&two_five_five,
6015 .procname = "mtu",
6016 .data = &ipv6_devconf.mtu6,
6017 .maxlen = sizeof(int),
6018 .mode = 0644,
6019 .proc_handler = addrconf_sysctl_mtu,
6022 .procname = "accept_ra",
6023 .data = &ipv6_devconf.accept_ra,
6024 .maxlen = sizeof(int),
6025 .mode = 0644,
6026 .proc_handler = proc_dointvec,
6029 .procname = "accept_redirects",
6030 .data = &ipv6_devconf.accept_redirects,
6031 .maxlen = sizeof(int),
6032 .mode = 0644,
6033 .proc_handler = proc_dointvec,
6036 .procname = "autoconf",
6037 .data = &ipv6_devconf.autoconf,
6038 .maxlen = sizeof(int),
6039 .mode = 0644,
6040 .proc_handler = proc_dointvec,
6043 .procname = "dad_transmits",
6044 .data = &ipv6_devconf.dad_transmits,
6045 .maxlen = sizeof(int),
6046 .mode = 0644,
6047 .proc_handler = proc_dointvec,
6050 .procname = "router_solicitations",
6051 .data = &ipv6_devconf.rtr_solicits,
6052 .maxlen = sizeof(int),
6053 .mode = 0644,
6054 .proc_handler = proc_dointvec_minmax,
6055 .extra1 = &minus_one,
6058 .procname = "router_solicitation_interval",
6059 .data = &ipv6_devconf.rtr_solicit_interval,
6060 .maxlen = sizeof(int),
6061 .mode = 0644,
6062 .proc_handler = proc_dointvec_jiffies,
6065 .procname = "router_solicitation_max_interval",
6066 .data = &ipv6_devconf.rtr_solicit_max_interval,
6067 .maxlen = sizeof(int),
6068 .mode = 0644,
6069 .proc_handler = proc_dointvec_jiffies,
6072 .procname = "router_solicitation_delay",
6073 .data = &ipv6_devconf.rtr_solicit_delay,
6074 .maxlen = sizeof(int),
6075 .mode = 0644,
6076 .proc_handler = proc_dointvec_jiffies,
6079 .procname = "force_mld_version",
6080 .data = &ipv6_devconf.force_mld_version,
6081 .maxlen = sizeof(int),
6082 .mode = 0644,
6083 .proc_handler = proc_dointvec,
6086 .procname = "mldv1_unsolicited_report_interval",
6087 .data =
6088 &ipv6_devconf.mldv1_unsolicited_report_interval,
6089 .maxlen = sizeof(int),
6090 .mode = 0644,
6091 .proc_handler = proc_dointvec_ms_jiffies,
6094 .procname = "mldv2_unsolicited_report_interval",
6095 .data =
6096 &ipv6_devconf.mldv2_unsolicited_report_interval,
6097 .maxlen = sizeof(int),
6098 .mode = 0644,
6099 .proc_handler = proc_dointvec_ms_jiffies,
6102 .procname = "use_tempaddr",
6103 .data = &ipv6_devconf.use_tempaddr,
6104 .maxlen = sizeof(int),
6105 .mode = 0644,
6106 .proc_handler = proc_dointvec,
6109 .procname = "temp_valid_lft",
6110 .data = &ipv6_devconf.temp_valid_lft,
6111 .maxlen = sizeof(int),
6112 .mode = 0644,
6113 .proc_handler = proc_dointvec,
6116 .procname = "temp_prefered_lft",
6117 .data = &ipv6_devconf.temp_prefered_lft,
6118 .maxlen = sizeof(int),
6119 .mode = 0644,
6120 .proc_handler = proc_dointvec,
6123 .procname = "regen_max_retry",
6124 .data = &ipv6_devconf.regen_max_retry,
6125 .maxlen = sizeof(int),
6126 .mode = 0644,
6127 .proc_handler = proc_dointvec,
6130 .procname = "max_desync_factor",
6131 .data = &ipv6_devconf.max_desync_factor,
6132 .maxlen = sizeof(int),
6133 .mode = 0644,
6134 .proc_handler = proc_dointvec,
6137 .procname = "max_addresses",
6138 .data = &ipv6_devconf.max_addresses,
6139 .maxlen = sizeof(int),
6140 .mode = 0644,
6141 .proc_handler = proc_dointvec,
6144 .procname = "accept_ra_defrtr",
6145 .data = &ipv6_devconf.accept_ra_defrtr,
6146 .maxlen = sizeof(int),
6147 .mode = 0644,
6148 .proc_handler = proc_dointvec,
6151 .procname = "accept_ra_min_hop_limit",
6152 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6153 .maxlen = sizeof(int),
6154 .mode = 0644,
6155 .proc_handler = proc_dointvec,
6158 .procname = "accept_ra_pinfo",
6159 .data = &ipv6_devconf.accept_ra_pinfo,
6160 .maxlen = sizeof(int),
6161 .mode = 0644,
6162 .proc_handler = proc_dointvec,
6164 #ifdef CONFIG_IPV6_ROUTER_PREF
6166 .procname = "accept_ra_rtr_pref",
6167 .data = &ipv6_devconf.accept_ra_rtr_pref,
6168 .maxlen = sizeof(int),
6169 .mode = 0644,
6170 .proc_handler = proc_dointvec,
6173 .procname = "router_probe_interval",
6174 .data = &ipv6_devconf.rtr_probe_interval,
6175 .maxlen = sizeof(int),
6176 .mode = 0644,
6177 .proc_handler = proc_dointvec_jiffies,
6179 #ifdef CONFIG_IPV6_ROUTE_INFO
6181 .procname = "accept_ra_rt_info_min_plen",
6182 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6183 .maxlen = sizeof(int),
6184 .mode = 0644,
6185 .proc_handler = proc_dointvec,
6188 .procname = "accept_ra_rt_info_max_plen",
6189 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6190 .maxlen = sizeof(int),
6191 .mode = 0644,
6192 .proc_handler = proc_dointvec,
6194 #endif
6195 #endif
6197 .procname = "proxy_ndp",
6198 .data = &ipv6_devconf.proxy_ndp,
6199 .maxlen = sizeof(int),
6200 .mode = 0644,
6201 .proc_handler = addrconf_sysctl_proxy_ndp,
6204 .procname = "accept_source_route",
6205 .data = &ipv6_devconf.accept_source_route,
6206 .maxlen = sizeof(int),
6207 .mode = 0644,
6208 .proc_handler = proc_dointvec,
6210 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6212 .procname = "optimistic_dad",
6213 .data = &ipv6_devconf.optimistic_dad,
6214 .maxlen = sizeof(int),
6215 .mode = 0644,
6216 .proc_handler = proc_dointvec,
6219 .procname = "use_optimistic",
6220 .data = &ipv6_devconf.use_optimistic,
6221 .maxlen = sizeof(int),
6222 .mode = 0644,
6223 .proc_handler = proc_dointvec,
6225 #endif
6226 #ifdef CONFIG_IPV6_MROUTE
6228 .procname = "mc_forwarding",
6229 .data = &ipv6_devconf.mc_forwarding,
6230 .maxlen = sizeof(int),
6231 .mode = 0444,
6232 .proc_handler = proc_dointvec,
6234 #endif
6236 .procname = "disable_ipv6",
6237 .data = &ipv6_devconf.disable_ipv6,
6238 .maxlen = sizeof(int),
6239 .mode = 0644,
6240 .proc_handler = addrconf_sysctl_disable,
6243 .procname = "accept_dad",
6244 .data = &ipv6_devconf.accept_dad,
6245 .maxlen = sizeof(int),
6246 .mode = 0644,
6247 .proc_handler = proc_dointvec,
6250 .procname = "force_tllao",
6251 .data = &ipv6_devconf.force_tllao,
6252 .maxlen = sizeof(int),
6253 .mode = 0644,
6254 .proc_handler = proc_dointvec
6257 .procname = "ndisc_notify",
6258 .data = &ipv6_devconf.ndisc_notify,
6259 .maxlen = sizeof(int),
6260 .mode = 0644,
6261 .proc_handler = proc_dointvec
6264 .procname = "suppress_frag_ndisc",
6265 .data = &ipv6_devconf.suppress_frag_ndisc,
6266 .maxlen = sizeof(int),
6267 .mode = 0644,
6268 .proc_handler = proc_dointvec
6271 .procname = "accept_ra_from_local",
6272 .data = &ipv6_devconf.accept_ra_from_local,
6273 .maxlen = sizeof(int),
6274 .mode = 0644,
6275 .proc_handler = proc_dointvec,
6278 .procname = "accept_ra_mtu",
6279 .data = &ipv6_devconf.accept_ra_mtu,
6280 .maxlen = sizeof(int),
6281 .mode = 0644,
6282 .proc_handler = proc_dointvec,
6285 .procname = "stable_secret",
6286 .data = &ipv6_devconf.stable_secret,
6287 .maxlen = IPV6_MAX_STRLEN,
6288 .mode = 0600,
6289 .proc_handler = addrconf_sysctl_stable_secret,
6292 .procname = "use_oif_addrs_only",
6293 .data = &ipv6_devconf.use_oif_addrs_only,
6294 .maxlen = sizeof(int),
6295 .mode = 0644,
6296 .proc_handler = proc_dointvec,
6299 .procname = "ignore_routes_with_linkdown",
6300 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6301 .maxlen = sizeof(int),
6302 .mode = 0644,
6303 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6306 .procname = "drop_unicast_in_l2_multicast",
6307 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6308 .maxlen = sizeof(int),
6309 .mode = 0644,
6310 .proc_handler = proc_dointvec,
6313 .procname = "drop_unsolicited_na",
6314 .data = &ipv6_devconf.drop_unsolicited_na,
6315 .maxlen = sizeof(int),
6316 .mode = 0644,
6317 .proc_handler = proc_dointvec,
6320 .procname = "keep_addr_on_down",
6321 .data = &ipv6_devconf.keep_addr_on_down,
6322 .maxlen = sizeof(int),
6323 .mode = 0644,
6324 .proc_handler = proc_dointvec,
6328 .procname = "seg6_enabled",
6329 .data = &ipv6_devconf.seg6_enabled,
6330 .maxlen = sizeof(int),
6331 .mode = 0644,
6332 .proc_handler = proc_dointvec,
6334 #ifdef CONFIG_IPV6_SEG6_HMAC
6336 .procname = "seg6_require_hmac",
6337 .data = &ipv6_devconf.seg6_require_hmac,
6338 .maxlen = sizeof(int),
6339 .mode = 0644,
6340 .proc_handler = proc_dointvec,
6342 #endif
6344 .procname = "enhanced_dad",
6345 .data = &ipv6_devconf.enhanced_dad,
6346 .maxlen = sizeof(int),
6347 .mode = 0644,
6348 .proc_handler = proc_dointvec,
6351 .procname = "addr_gen_mode",
6352 .data = &ipv6_devconf.addr_gen_mode,
6353 .maxlen = sizeof(int),
6354 .mode = 0644,
6355 .proc_handler = addrconf_sysctl_addr_gen_mode,
6358 .procname = "disable_policy",
6359 .data = &ipv6_devconf.disable_policy,
6360 .maxlen = sizeof(int),
6361 .mode = 0644,
6362 .proc_handler = addrconf_sysctl_disable_policy,
6365 .procname = "ndisc_tclass",
6366 .data = &ipv6_devconf.ndisc_tclass,
6367 .maxlen = sizeof(int),
6368 .mode = 0644,
6369 .proc_handler = proc_dointvec_minmax,
6370 .extra1 = (void *)&zero,
6371 .extra2 = (void *)&two_five_five,
6374 /* sentinel */
6378 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6379 struct inet6_dev *idev, struct ipv6_devconf *p)
6381 int i, ifindex;
6382 struct ctl_table *table;
6383 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6385 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6386 if (!table)
6387 goto out;
6389 for (i = 0; table[i].data; i++) {
6390 table[i].data += (char *)p - (char *)&ipv6_devconf;
6391 /* If one of these is already set, then it is not safe to
6392 * overwrite either of them: this makes proc_dointvec_minmax
6393 * usable.
6395 if (!table[i].extra1 && !table[i].extra2) {
6396 table[i].extra1 = idev; /* embedded; no ref */
6397 table[i].extra2 = net;
6401 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6403 p->sysctl_header = register_net_sysctl(net, path, table);
6404 if (!p->sysctl_header)
6405 goto free;
6407 if (!strcmp(dev_name, "all"))
6408 ifindex = NETCONFA_IFINDEX_ALL;
6409 else if (!strcmp(dev_name, "default"))
6410 ifindex = NETCONFA_IFINDEX_DEFAULT;
6411 else
6412 ifindex = idev->dev->ifindex;
6413 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6414 ifindex, p);
6415 return 0;
6417 free:
6418 kfree(table);
6419 out:
6420 return -ENOBUFS;
6423 static void __addrconf_sysctl_unregister(struct net *net,
6424 struct ipv6_devconf *p, int ifindex)
6426 struct ctl_table *table;
6428 if (!p->sysctl_header)
6429 return;
6431 table = p->sysctl_header->ctl_table_arg;
6432 unregister_net_sysctl_table(p->sysctl_header);
6433 p->sysctl_header = NULL;
6434 kfree(table);
6436 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6439 static int addrconf_sysctl_register(struct inet6_dev *idev)
6441 int err;
6443 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6444 return -EINVAL;
6446 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6447 &ndisc_ifinfo_sysctl_change);
6448 if (err)
6449 return err;
6450 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6451 idev, &idev->cnf);
6452 if (err)
6453 neigh_sysctl_unregister(idev->nd_parms);
6455 return err;
6458 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6460 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6461 idev->dev->ifindex);
6462 neigh_sysctl_unregister(idev->nd_parms);
6466 #endif
6468 static int __net_init addrconf_init_net(struct net *net)
6470 int err = -ENOMEM;
6471 struct ipv6_devconf *all, *dflt;
6473 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6474 if (!all)
6475 goto err_alloc_all;
6477 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6478 if (!dflt)
6479 goto err_alloc_dflt;
6481 /* these will be inherited by all namespaces */
6482 dflt->autoconf = ipv6_defaults.autoconf;
6483 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6485 dflt->stable_secret.initialized = false;
6486 all->stable_secret.initialized = false;
6488 net->ipv6.devconf_all = all;
6489 net->ipv6.devconf_dflt = dflt;
6491 #ifdef CONFIG_SYSCTL
6492 err = __addrconf_sysctl_register(net, "all", NULL, all);
6493 if (err < 0)
6494 goto err_reg_all;
6496 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6497 if (err < 0)
6498 goto err_reg_dflt;
6499 #endif
6500 return 0;
6502 #ifdef CONFIG_SYSCTL
6503 err_reg_dflt:
6504 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6505 err_reg_all:
6506 kfree(dflt);
6507 #endif
6508 err_alloc_dflt:
6509 kfree(all);
6510 err_alloc_all:
6511 return err;
6514 static void __net_exit addrconf_exit_net(struct net *net)
6516 #ifdef CONFIG_SYSCTL
6517 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6518 NETCONFA_IFINDEX_DEFAULT);
6519 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6520 NETCONFA_IFINDEX_ALL);
6521 #endif
6522 kfree(net->ipv6.devconf_dflt);
6523 kfree(net->ipv6.devconf_all);
6526 static struct pernet_operations addrconf_ops = {
6527 .init = addrconf_init_net,
6528 .exit = addrconf_exit_net,
6531 static struct rtnl_af_ops inet6_ops __read_mostly = {
6532 .family = AF_INET6,
6533 .fill_link_af = inet6_fill_link_af,
6534 .get_link_af_size = inet6_get_link_af_size,
6535 .validate_link_af = inet6_validate_link_af,
6536 .set_link_af = inet6_set_link_af,
6540 * Init / cleanup code
6543 int __init addrconf_init(void)
6545 struct inet6_dev *idev;
6546 int i, err;
6548 err = ipv6_addr_label_init();
6549 if (err < 0) {
6550 pr_crit("%s: cannot initialize default policy table: %d\n",
6551 __func__, err);
6552 goto out;
6555 err = register_pernet_subsys(&addrconf_ops);
6556 if (err < 0)
6557 goto out_addrlabel;
6559 addrconf_wq = create_workqueue("ipv6_addrconf");
6560 if (!addrconf_wq) {
6561 err = -ENOMEM;
6562 goto out_nowq;
6565 /* The addrconf netdev notifier requires that loopback_dev
6566 * has it's ipv6 private information allocated and setup
6567 * before it can bring up and give link-local addresses
6568 * to other devices which are up.
6570 * Unfortunately, loopback_dev is not necessarily the first
6571 * entry in the global dev_base list of net devices. In fact,
6572 * it is likely to be the very last entry on that list.
6573 * So this causes the notifier registry below to try and
6574 * give link-local addresses to all devices besides loopback_dev
6575 * first, then loopback_dev, which cases all the non-loopback_dev
6576 * devices to fail to get a link-local address.
6578 * So, as a temporary fix, allocate the ipv6 structure for
6579 * loopback_dev first by hand.
6580 * Longer term, all of the dependencies ipv6 has upon the loopback
6581 * device and it being up should be removed.
6583 rtnl_lock();
6584 idev = ipv6_add_dev(init_net.loopback_dev);
6585 rtnl_unlock();
6586 if (IS_ERR(idev)) {
6587 err = PTR_ERR(idev);
6588 goto errlo;
6591 ip6_route_init_special_entries();
6593 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6594 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6596 register_netdevice_notifier(&ipv6_dev_notf);
6598 addrconf_verify();
6600 rtnl_af_register(&inet6_ops);
6602 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6603 NULL, inet6_dump_ifinfo, 0);
6604 if (err < 0)
6605 goto errout;
6607 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6608 inet6_rtm_newaddr, NULL, 0);
6609 if (err < 0)
6610 goto errout;
6611 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6612 inet6_rtm_deladdr, NULL, 0);
6613 if (err < 0)
6614 goto errout;
6615 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6616 inet6_rtm_getaddr, inet6_dump_ifaddr,
6617 RTNL_FLAG_DOIT_UNLOCKED);
6618 if (err < 0)
6619 goto errout;
6620 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6621 NULL, inet6_dump_ifmcaddr, 0);
6622 if (err < 0)
6623 goto errout;
6624 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6625 NULL, inet6_dump_ifacaddr, 0);
6626 if (err < 0)
6627 goto errout;
6628 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6629 inet6_netconf_get_devconf,
6630 inet6_netconf_dump_devconf,
6631 RTNL_FLAG_DOIT_UNLOCKED);
6632 if (err < 0)
6633 goto errout;
6634 err = ipv6_addr_label_rtnl_register();
6635 if (err < 0)
6636 goto errout;
6638 return 0;
6639 errout:
6640 rtnl_unregister_all(PF_INET6);
6641 rtnl_af_unregister(&inet6_ops);
6642 unregister_netdevice_notifier(&ipv6_dev_notf);
6643 errlo:
6644 destroy_workqueue(addrconf_wq);
6645 out_nowq:
6646 unregister_pernet_subsys(&addrconf_ops);
6647 out_addrlabel:
6648 ipv6_addr_label_cleanup();
6649 out:
6650 return err;
6653 void addrconf_cleanup(void)
6655 struct net_device *dev;
6656 int i;
6658 unregister_netdevice_notifier(&ipv6_dev_notf);
6659 unregister_pernet_subsys(&addrconf_ops);
6660 ipv6_addr_label_cleanup();
6662 rtnl_af_unregister(&inet6_ops);
6664 rtnl_lock();
6666 /* clean dev list */
6667 for_each_netdev(&init_net, dev) {
6668 if (__in6_dev_get(dev) == NULL)
6669 continue;
6670 addrconf_ifdown(dev, 1);
6672 addrconf_ifdown(init_net.loopback_dev, 2);
6675 * Check hash table.
6677 spin_lock_bh(&addrconf_hash_lock);
6678 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6679 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6680 spin_unlock_bh(&addrconf_hash_lock);
6681 cancel_delayed_work(&addr_chk_work);
6682 rtnl_unlock();
6684 destroy_workqueue(addrconf_wq);