media: cpia2_usb: drop bogus interface-release call
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blobe1846b97ee695e809cefb856835b59b45e4b34a1
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 bool send_na);
191 static void addrconf_dad_run(struct inet6_dev *idev);
192 static void addrconf_rs_timer(struct timer_list *t);
193 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
194 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
196 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
197 struct prefix_info *pinfo);
199 static struct ipv6_devconf ipv6_devconf __read_mostly = {
200 .forwarding = 0,
201 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
202 .mtu6 = IPV6_MIN_MTU,
203 .accept_ra = 1,
204 .accept_redirects = 1,
205 .autoconf = 1,
206 .force_mld_version = 0,
207 .mldv1_unsolicited_report_interval = 10 * HZ,
208 .mldv2_unsolicited_report_interval = HZ,
209 .dad_transmits = 1,
210 .rtr_solicits = MAX_RTR_SOLICITATIONS,
211 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
212 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
213 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
214 .use_tempaddr = 0,
215 .temp_valid_lft = TEMP_VALID_LIFETIME,
216 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
217 .regen_max_retry = REGEN_MAX_RETRY,
218 .max_desync_factor = MAX_DESYNC_FACTOR,
219 .max_addresses = IPV6_MAX_ADDRESSES,
220 .accept_ra_defrtr = 1,
221 .accept_ra_from_local = 0,
222 .accept_ra_min_hop_limit= 1,
223 .accept_ra_pinfo = 1,
224 #ifdef CONFIG_IPV6_ROUTER_PREF
225 .accept_ra_rtr_pref = 1,
226 .rtr_probe_interval = 60 * HZ,
227 #ifdef CONFIG_IPV6_ROUTE_INFO
228 .accept_ra_rt_info_min_plen = 0,
229 .accept_ra_rt_info_max_plen = 0,
230 #endif
231 #endif
232 .proxy_ndp = 0,
233 .accept_source_route = 0, /* we do not accept RH0 by default. */
234 .disable_ipv6 = 0,
235 .accept_dad = 0,
236 .suppress_frag_ndisc = 1,
237 .accept_ra_mtu = 1,
238 .stable_secret = {
239 .initialized = false,
241 .use_oif_addrs_only = 0,
242 .ignore_routes_with_linkdown = 0,
243 .keep_addr_on_down = 0,
244 .seg6_enabled = 0,
245 #ifdef CONFIG_IPV6_SEG6_HMAC
246 .seg6_require_hmac = 0,
247 #endif
248 .enhanced_dad = 1,
249 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
250 .disable_policy = 0,
253 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
254 .forwarding = 0,
255 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
256 .mtu6 = IPV6_MIN_MTU,
257 .accept_ra = 1,
258 .accept_redirects = 1,
259 .autoconf = 1,
260 .force_mld_version = 0,
261 .mldv1_unsolicited_report_interval = 10 * HZ,
262 .mldv2_unsolicited_report_interval = HZ,
263 .dad_transmits = 1,
264 .rtr_solicits = MAX_RTR_SOLICITATIONS,
265 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
266 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
267 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
268 .use_tempaddr = 0,
269 .temp_valid_lft = TEMP_VALID_LIFETIME,
270 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
271 .regen_max_retry = REGEN_MAX_RETRY,
272 .max_desync_factor = MAX_DESYNC_FACTOR,
273 .max_addresses = IPV6_MAX_ADDRESSES,
274 .accept_ra_defrtr = 1,
275 .accept_ra_from_local = 0,
276 .accept_ra_min_hop_limit= 1,
277 .accept_ra_pinfo = 1,
278 #ifdef CONFIG_IPV6_ROUTER_PREF
279 .accept_ra_rtr_pref = 1,
280 .rtr_probe_interval = 60 * HZ,
281 #ifdef CONFIG_IPV6_ROUTE_INFO
282 .accept_ra_rt_info_min_plen = 0,
283 .accept_ra_rt_info_max_plen = 0,
284 #endif
285 #endif
286 .proxy_ndp = 0,
287 .accept_source_route = 0, /* we do not accept RH0 by default. */
288 .disable_ipv6 = 0,
289 .accept_dad = 1,
290 .suppress_frag_ndisc = 1,
291 .accept_ra_mtu = 1,
292 .stable_secret = {
293 .initialized = false,
295 .use_oif_addrs_only = 0,
296 .ignore_routes_with_linkdown = 0,
297 .keep_addr_on_down = 0,
298 .seg6_enabled = 0,
299 #ifdef CONFIG_IPV6_SEG6_HMAC
300 .seg6_require_hmac = 0,
301 #endif
302 .enhanced_dad = 1,
303 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
304 .disable_policy = 0,
307 /* Check if link is ready: is it up and is a valid qdisc available */
308 static inline bool addrconf_link_ready(const struct net_device *dev)
310 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
313 static void addrconf_del_rs_timer(struct inet6_dev *idev)
315 if (del_timer(&idev->rs_timer))
316 __in6_dev_put(idev);
319 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
321 if (cancel_delayed_work(&ifp->dad_work))
322 __in6_ifa_put(ifp);
325 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
326 unsigned long when)
328 if (!timer_pending(&idev->rs_timer))
329 in6_dev_hold(idev);
330 mod_timer(&idev->rs_timer, jiffies + when);
333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
334 unsigned long delay)
336 in6_ifa_hold(ifp);
337 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
338 in6_ifa_put(ifp);
341 static int snmp6_alloc_dev(struct inet6_dev *idev)
343 int i;
345 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
346 if (!idev->stats.ipv6)
347 goto err_ip;
349 for_each_possible_cpu(i) {
350 struct ipstats_mib *addrconf_stats;
351 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
352 u64_stats_init(&addrconf_stats->syncp);
356 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
357 GFP_KERNEL);
358 if (!idev->stats.icmpv6dev)
359 goto err_icmp;
360 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
361 GFP_KERNEL);
362 if (!idev->stats.icmpv6msgdev)
363 goto err_icmpmsg;
365 return 0;
367 err_icmpmsg:
368 kfree(idev->stats.icmpv6dev);
369 err_icmp:
370 free_percpu(idev->stats.ipv6);
371 err_ip:
372 return -ENOMEM;
375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
377 struct inet6_dev *ndev;
378 int err = -ENOMEM;
380 ASSERT_RTNL();
382 if (dev->mtu < IPV6_MIN_MTU)
383 return ERR_PTR(-EINVAL);
385 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
386 if (!ndev)
387 return ERR_PTR(err);
389 rwlock_init(&ndev->lock);
390 ndev->dev = dev;
391 INIT_LIST_HEAD(&ndev->addr_list);
392 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
393 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
395 if (ndev->cnf.stable_secret.initialized)
396 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
397 else
398 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
400 ndev->cnf.mtu6 = dev->mtu;
401 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
402 if (!ndev->nd_parms) {
403 kfree(ndev);
404 return ERR_PTR(err);
406 if (ndev->cnf.forwarding)
407 dev_disable_lro(dev);
408 /* We refer to the device */
409 dev_hold(dev);
411 if (snmp6_alloc_dev(ndev) < 0) {
412 ADBG(KERN_WARNING
413 "%s: cannot allocate memory for statistics; dev=%s.\n",
414 __func__, dev->name);
415 neigh_parms_release(&nd_tbl, ndev->nd_parms);
416 dev_put(dev);
417 kfree(ndev);
418 return ERR_PTR(err);
421 if (snmp6_register_dev(ndev) < 0) {
422 ADBG(KERN_WARNING
423 "%s: cannot create /proc/net/dev_snmp6/%s\n",
424 __func__, dev->name);
425 goto err_release;
428 /* One reference from device. */
429 refcount_set(&ndev->refcnt, 1);
431 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
432 ndev->cnf.accept_dad = -1;
434 #if IS_ENABLED(CONFIG_IPV6_SIT)
435 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
436 pr_info("%s: Disabled Multicast RS\n", dev->name);
437 ndev->cnf.rtr_solicits = 0;
439 #endif
441 INIT_LIST_HEAD(&ndev->tempaddr_list);
442 ndev->desync_factor = U32_MAX;
443 if ((dev->flags&IFF_LOOPBACK) ||
444 dev->type == ARPHRD_TUNNEL ||
445 dev->type == ARPHRD_TUNNEL6 ||
446 dev->type == ARPHRD_SIT ||
447 dev->type == ARPHRD_NONE) {
448 ndev->cnf.use_tempaddr = -1;
449 } else
450 ipv6_regen_rndid(ndev);
452 ndev->token = in6addr_any;
454 if (netif_running(dev) && addrconf_link_ready(dev))
455 ndev->if_flags |= IF_READY;
457 ipv6_mc_init_dev(ndev);
458 ndev->tstamp = jiffies;
459 err = addrconf_sysctl_register(ndev);
460 if (err) {
461 ipv6_mc_destroy_dev(ndev);
462 snmp6_unregister_dev(ndev);
463 goto err_release;
465 /* protected by rtnl_lock */
466 rcu_assign_pointer(dev->ip6_ptr, ndev);
468 /* Join interface-local all-node multicast group */
469 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
471 /* Join all-node multicast group */
472 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
474 /* Join all-router multicast group if forwarding is set */
475 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
476 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
478 return ndev;
480 err_release:
481 neigh_parms_release(&nd_tbl, ndev->nd_parms);
482 ndev->dead = 1;
483 in6_dev_finish_destroy(ndev);
484 return ERR_PTR(err);
487 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
489 struct inet6_dev *idev;
491 ASSERT_RTNL();
493 idev = __in6_dev_get(dev);
494 if (!idev) {
495 idev = ipv6_add_dev(dev);
496 if (IS_ERR(idev))
497 return NULL;
500 if (dev->flags&IFF_UP)
501 ipv6_mc_up(idev);
502 return idev;
505 static int inet6_netconf_msgsize_devconf(int type)
507 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
508 + nla_total_size(4); /* NETCONFA_IFINDEX */
509 bool all = false;
511 if (type == NETCONFA_ALL)
512 all = true;
514 if (all || type == NETCONFA_FORWARDING)
515 size += nla_total_size(4);
516 #ifdef CONFIG_IPV6_MROUTE
517 if (all || type == NETCONFA_MC_FORWARDING)
518 size += nla_total_size(4);
519 #endif
520 if (all || type == NETCONFA_PROXY_NEIGH)
521 size += nla_total_size(4);
523 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
524 size += nla_total_size(4);
526 return size;
529 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
530 struct ipv6_devconf *devconf, u32 portid,
531 u32 seq, int event, unsigned int flags,
532 int type)
534 struct nlmsghdr *nlh;
535 struct netconfmsg *ncm;
536 bool all = false;
538 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
539 flags);
540 if (!nlh)
541 return -EMSGSIZE;
543 if (type == NETCONFA_ALL)
544 all = true;
546 ncm = nlmsg_data(nlh);
547 ncm->ncm_family = AF_INET6;
549 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
550 goto nla_put_failure;
552 if (!devconf)
553 goto out;
555 if ((all || type == NETCONFA_FORWARDING) &&
556 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
557 goto nla_put_failure;
558 #ifdef CONFIG_IPV6_MROUTE
559 if ((all || type == NETCONFA_MC_FORWARDING) &&
560 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
561 devconf->mc_forwarding) < 0)
562 goto nla_put_failure;
563 #endif
564 if ((all || type == NETCONFA_PROXY_NEIGH) &&
565 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
566 goto nla_put_failure;
568 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
569 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
570 devconf->ignore_routes_with_linkdown) < 0)
571 goto nla_put_failure;
573 out:
574 nlmsg_end(skb, nlh);
575 return 0;
577 nla_put_failure:
578 nlmsg_cancel(skb, nlh);
579 return -EMSGSIZE;
582 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
583 int ifindex, struct ipv6_devconf *devconf)
585 struct sk_buff *skb;
586 int err = -ENOBUFS;
588 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
589 if (!skb)
590 goto errout;
592 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
593 event, 0, type);
594 if (err < 0) {
595 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
596 WARN_ON(err == -EMSGSIZE);
597 kfree_skb(skb);
598 goto errout;
600 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
601 return;
602 errout:
603 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
606 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
607 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
608 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
609 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
610 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
613 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
614 struct nlmsghdr *nlh,
615 struct netlink_ext_ack *extack)
617 struct net *net = sock_net(in_skb->sk);
618 struct nlattr *tb[NETCONFA_MAX+1];
619 struct inet6_dev *in6_dev = NULL;
620 struct net_device *dev = NULL;
621 struct netconfmsg *ncm;
622 struct sk_buff *skb;
623 struct ipv6_devconf *devconf;
624 int ifindex;
625 int err;
627 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
628 devconf_ipv6_policy, extack);
629 if (err < 0)
630 return err;
632 if (!tb[NETCONFA_IFINDEX])
633 return -EINVAL;
635 err = -EINVAL;
636 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
637 switch (ifindex) {
638 case NETCONFA_IFINDEX_ALL:
639 devconf = net->ipv6.devconf_all;
640 break;
641 case NETCONFA_IFINDEX_DEFAULT:
642 devconf = net->ipv6.devconf_dflt;
643 break;
644 default:
645 dev = dev_get_by_index(net, ifindex);
646 if (!dev)
647 return -EINVAL;
648 in6_dev = in6_dev_get(dev);
649 if (!in6_dev)
650 goto errout;
651 devconf = &in6_dev->cnf;
652 break;
655 err = -ENOBUFS;
656 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
657 if (!skb)
658 goto errout;
660 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
661 NETLINK_CB(in_skb).portid,
662 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
663 NETCONFA_ALL);
664 if (err < 0) {
665 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
666 WARN_ON(err == -EMSGSIZE);
667 kfree_skb(skb);
668 goto errout;
670 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
671 errout:
672 if (in6_dev)
673 in6_dev_put(in6_dev);
674 if (dev)
675 dev_put(dev);
676 return err;
679 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
680 struct netlink_callback *cb)
682 struct net *net = sock_net(skb->sk);
683 int h, s_h;
684 int idx, s_idx;
685 struct net_device *dev;
686 struct inet6_dev *idev;
687 struct hlist_head *head;
689 s_h = cb->args[0];
690 s_idx = idx = cb->args[1];
692 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
693 idx = 0;
694 head = &net->dev_index_head[h];
695 rcu_read_lock();
696 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
697 net->dev_base_seq;
698 hlist_for_each_entry_rcu(dev, head, index_hlist) {
699 if (idx < s_idx)
700 goto cont;
701 idev = __in6_dev_get(dev);
702 if (!idev)
703 goto cont;
705 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
706 &idev->cnf,
707 NETLINK_CB(cb->skb).portid,
708 cb->nlh->nlmsg_seq,
709 RTM_NEWNETCONF,
710 NLM_F_MULTI,
711 NETCONFA_ALL) < 0) {
712 rcu_read_unlock();
713 goto done;
715 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
716 cont:
717 idx++;
719 rcu_read_unlock();
721 if (h == NETDEV_HASHENTRIES) {
722 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
723 net->ipv6.devconf_all,
724 NETLINK_CB(cb->skb).portid,
725 cb->nlh->nlmsg_seq,
726 RTM_NEWNETCONF, NLM_F_MULTI,
727 NETCONFA_ALL) < 0)
728 goto done;
729 else
730 h++;
732 if (h == NETDEV_HASHENTRIES + 1) {
733 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
734 net->ipv6.devconf_dflt,
735 NETLINK_CB(cb->skb).portid,
736 cb->nlh->nlmsg_seq,
737 RTM_NEWNETCONF, NLM_F_MULTI,
738 NETCONFA_ALL) < 0)
739 goto done;
740 else
741 h++;
743 done:
744 cb->args[0] = h;
745 cb->args[1] = idx;
747 return skb->len;
750 #ifdef CONFIG_SYSCTL
751 static void dev_forward_change(struct inet6_dev *idev)
753 struct net_device *dev;
754 struct inet6_ifaddr *ifa;
756 if (!idev)
757 return;
758 dev = idev->dev;
759 if (idev->cnf.forwarding)
760 dev_disable_lro(dev);
761 if (dev->flags & IFF_MULTICAST) {
762 if (idev->cnf.forwarding) {
763 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
764 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
765 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
766 } else {
767 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
768 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
769 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
773 list_for_each_entry(ifa, &idev->addr_list, if_list) {
774 if (ifa->flags&IFA_F_TENTATIVE)
775 continue;
776 if (idev->cnf.forwarding)
777 addrconf_join_anycast(ifa);
778 else
779 addrconf_leave_anycast(ifa);
781 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
782 NETCONFA_FORWARDING,
783 dev->ifindex, &idev->cnf);
787 static void addrconf_forward_change(struct net *net, __s32 newf)
789 struct net_device *dev;
790 struct inet6_dev *idev;
792 for_each_netdev(net, dev) {
793 idev = __in6_dev_get(dev);
794 if (idev) {
795 int changed = (!idev->cnf.forwarding) ^ (!newf);
796 idev->cnf.forwarding = newf;
797 if (changed)
798 dev_forward_change(idev);
803 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
805 struct net *net;
806 int old;
808 if (!rtnl_trylock())
809 return restart_syscall();
811 net = (struct net *)table->extra2;
812 old = *p;
813 *p = newf;
815 if (p == &net->ipv6.devconf_dflt->forwarding) {
816 if ((!newf) ^ (!old))
817 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
818 NETCONFA_FORWARDING,
819 NETCONFA_IFINDEX_DEFAULT,
820 net->ipv6.devconf_dflt);
821 rtnl_unlock();
822 return 0;
825 if (p == &net->ipv6.devconf_all->forwarding) {
826 int old_dflt = net->ipv6.devconf_dflt->forwarding;
828 net->ipv6.devconf_dflt->forwarding = newf;
829 if ((!newf) ^ (!old_dflt))
830 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
831 NETCONFA_FORWARDING,
832 NETCONFA_IFINDEX_DEFAULT,
833 net->ipv6.devconf_dflt);
835 addrconf_forward_change(net, newf);
836 if ((!newf) ^ (!old))
837 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
838 NETCONFA_FORWARDING,
839 NETCONFA_IFINDEX_ALL,
840 net->ipv6.devconf_all);
841 } else if ((!newf) ^ (!old))
842 dev_forward_change((struct inet6_dev *)table->extra1);
843 rtnl_unlock();
845 if (newf)
846 rt6_purge_dflt_routers(net);
847 return 1;
850 static void addrconf_linkdown_change(struct net *net, __s32 newf)
852 struct net_device *dev;
853 struct inet6_dev *idev;
855 for_each_netdev(net, dev) {
856 idev = __in6_dev_get(dev);
857 if (idev) {
858 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
860 idev->cnf.ignore_routes_with_linkdown = newf;
861 if (changed)
862 inet6_netconf_notify_devconf(dev_net(dev),
863 RTM_NEWNETCONF,
864 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
865 dev->ifindex,
866 &idev->cnf);
871 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
873 struct net *net;
874 int old;
876 if (!rtnl_trylock())
877 return restart_syscall();
879 net = (struct net *)table->extra2;
880 old = *p;
881 *p = newf;
883 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
884 if ((!newf) ^ (!old))
885 inet6_netconf_notify_devconf(net,
886 RTM_NEWNETCONF,
887 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
888 NETCONFA_IFINDEX_DEFAULT,
889 net->ipv6.devconf_dflt);
890 rtnl_unlock();
891 return 0;
894 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
895 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
896 addrconf_linkdown_change(net, newf);
897 if ((!newf) ^ (!old))
898 inet6_netconf_notify_devconf(net,
899 RTM_NEWNETCONF,
900 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
901 NETCONFA_IFINDEX_ALL,
902 net->ipv6.devconf_all);
904 rtnl_unlock();
906 return 1;
909 #endif
911 /* Nobody refers to this ifaddr, destroy it */
912 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
914 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
916 #ifdef NET_REFCNT_DEBUG
917 pr_debug("%s\n", __func__);
918 #endif
920 in6_dev_put(ifp->idev);
922 if (cancel_delayed_work(&ifp->dad_work))
923 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
924 ifp);
926 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
927 pr_warn("Freeing alive inet6 address %p\n", ifp);
928 return;
930 ip6_rt_put(ifp->rt);
932 kfree_rcu(ifp, rcu);
935 static void
936 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
938 struct list_head *p;
939 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
942 * Each device address list is sorted in order of scope -
943 * global before linklocal.
945 list_for_each(p, &idev->addr_list) {
946 struct inet6_ifaddr *ifa
947 = list_entry(p, struct inet6_ifaddr, if_list);
948 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
949 break;
952 list_add_tail_rcu(&ifp->if_list, p);
955 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
957 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
959 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
962 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
963 struct net_device *dev, unsigned int hash)
965 struct inet6_ifaddr *ifp;
967 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
968 if (!net_eq(dev_net(ifp->idev->dev), net))
969 continue;
970 if (ipv6_addr_equal(&ifp->addr, addr)) {
971 if (!dev || ifp->idev->dev == dev)
972 return true;
975 return false;
978 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
980 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
981 int err = 0;
983 spin_lock(&addrconf_hash_lock);
985 /* Ignore adding duplicate addresses on an interface */
986 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
987 ADBG("ipv6_add_addr: already assigned\n");
988 err = -EEXIST;
989 } else {
990 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
993 spin_unlock(&addrconf_hash_lock);
995 return err;
998 /* On success it returns ifp with increased reference count */
1000 static struct inet6_ifaddr *
1001 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
1002 const struct in6_addr *peer_addr, int pfxlen,
1003 int scope, u32 flags, u32 valid_lft, u32 prefered_lft,
1004 bool can_block, struct netlink_ext_ack *extack)
1006 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1007 struct net *net = dev_net(idev->dev);
1008 struct inet6_ifaddr *ifa = NULL;
1009 struct rt6_info *rt = NULL;
1010 int err = 0;
1011 int addr_type = ipv6_addr_type(addr);
1013 if (addr_type == IPV6_ADDR_ANY ||
1014 addr_type & IPV6_ADDR_MULTICAST ||
1015 (!(idev->dev->flags & IFF_LOOPBACK) &&
1016 addr_type & IPV6_ADDR_LOOPBACK))
1017 return ERR_PTR(-EADDRNOTAVAIL);
1019 if (idev->dead) {
1020 err = -ENODEV; /*XXX*/
1021 goto out;
1024 if (idev->cnf.disable_ipv6) {
1025 err = -EACCES;
1026 goto out;
1029 /* validator notifier needs to be blocking;
1030 * do not call in atomic context
1032 if (can_block) {
1033 struct in6_validator_info i6vi = {
1034 .i6vi_addr = *addr,
1035 .i6vi_dev = idev,
1036 .extack = extack,
1039 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1040 err = notifier_to_errno(err);
1041 if (err < 0)
1042 goto out;
1045 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1046 if (!ifa) {
1047 ADBG("ipv6_add_addr: malloc failed\n");
1048 err = -ENOBUFS;
1049 goto out;
1052 rt = addrconf_dst_alloc(idev, addr, false);
1053 if (IS_ERR(rt)) {
1054 err = PTR_ERR(rt);
1055 rt = NULL;
1056 goto out;
1059 if (net->ipv6.devconf_all->disable_policy ||
1060 idev->cnf.disable_policy)
1061 rt->dst.flags |= DST_NOPOLICY;
1063 neigh_parms_data_state_setall(idev->nd_parms);
1065 ifa->addr = *addr;
1066 if (peer_addr)
1067 ifa->peer_addr = *peer_addr;
1069 spin_lock_init(&ifa->lock);
1070 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1071 INIT_HLIST_NODE(&ifa->addr_lst);
1072 ifa->scope = scope;
1073 ifa->prefix_len = pfxlen;
1074 ifa->flags = flags;
1075 /* No need to add the TENTATIVE flag for addresses with NODAD */
1076 if (!(flags & IFA_F_NODAD))
1077 ifa->flags |= IFA_F_TENTATIVE;
1078 ifa->valid_lft = valid_lft;
1079 ifa->prefered_lft = prefered_lft;
1080 ifa->cstamp = ifa->tstamp = jiffies;
1081 ifa->tokenized = false;
1083 ifa->rt = rt;
1085 ifa->idev = idev;
1086 in6_dev_hold(idev);
1088 /* For caller */
1089 refcount_set(&ifa->refcnt, 1);
1091 rcu_read_lock_bh();
1093 err = ipv6_add_addr_hash(idev->dev, ifa);
1094 if (err < 0) {
1095 rcu_read_unlock_bh();
1096 goto out;
1099 write_lock(&idev->lock);
1101 /* Add to inet6_dev unicast addr list. */
1102 ipv6_link_dev_addr(idev, ifa);
1104 if (ifa->flags&IFA_F_TEMPORARY) {
1105 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1106 in6_ifa_hold(ifa);
1109 in6_ifa_hold(ifa);
1110 write_unlock(&idev->lock);
1112 rcu_read_unlock_bh();
1114 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1115 out:
1116 if (unlikely(err < 0)) {
1117 if (rt)
1118 ip6_rt_put(rt);
1119 if (ifa) {
1120 if (ifa->idev)
1121 in6_dev_put(ifa->idev);
1122 kfree(ifa);
1124 ifa = ERR_PTR(err);
1127 return ifa;
1130 enum cleanup_prefix_rt_t {
1131 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1132 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1133 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1137 * Check, whether the prefix for ifp would still need a prefix route
1138 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1139 * constants.
1141 * 1) we don't purge prefix if address was not permanent.
1142 * prefix is managed by its own lifetime.
1143 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1144 * 3) if there are no addresses, delete prefix.
1145 * 4) if there are still other permanent address(es),
1146 * corresponding prefix is still permanent.
1147 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1148 * don't purge the prefix, assume user space is managing it.
1149 * 6) otherwise, update prefix lifetime to the
1150 * longest valid lifetime among the corresponding
1151 * addresses on the device.
1152 * Note: subsequent RA will update lifetime.
1154 static enum cleanup_prefix_rt_t
1155 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1157 struct inet6_ifaddr *ifa;
1158 struct inet6_dev *idev = ifp->idev;
1159 unsigned long lifetime;
1160 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1162 *expires = jiffies;
1164 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1165 if (ifa == ifp)
1166 continue;
1167 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1168 ifp->prefix_len))
1169 continue;
1170 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1171 return CLEANUP_PREFIX_RT_NOP;
1173 action = CLEANUP_PREFIX_RT_EXPIRE;
1175 spin_lock(&ifa->lock);
1177 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1179 * Note: Because this address is
1180 * not permanent, lifetime <
1181 * LONG_MAX / HZ here.
1183 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1184 *expires = ifa->tstamp + lifetime * HZ;
1185 spin_unlock(&ifa->lock);
1188 return action;
1191 static void
1192 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1194 struct rt6_info *rt;
1196 rt = addrconf_get_prefix_route(&ifp->addr,
1197 ifp->prefix_len,
1198 ifp->idev->dev,
1199 0, RTF_GATEWAY | RTF_DEFAULT);
1200 if (rt) {
1201 if (del_rt)
1202 ip6_del_rt(rt);
1203 else {
1204 if (!(rt->rt6i_flags & RTF_EXPIRES))
1205 rt6_set_expires(rt, expires);
1206 ip6_rt_put(rt);
1212 /* This function wants to get referenced ifp and releases it before return */
1214 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1216 int state;
1217 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1218 unsigned long expires;
1220 ASSERT_RTNL();
1222 spin_lock_bh(&ifp->lock);
1223 state = ifp->state;
1224 ifp->state = INET6_IFADDR_STATE_DEAD;
1225 spin_unlock_bh(&ifp->lock);
1227 if (state == INET6_IFADDR_STATE_DEAD)
1228 goto out;
1230 spin_lock_bh(&addrconf_hash_lock);
1231 hlist_del_init_rcu(&ifp->addr_lst);
1232 spin_unlock_bh(&addrconf_hash_lock);
1234 write_lock_bh(&ifp->idev->lock);
1236 if (ifp->flags&IFA_F_TEMPORARY) {
1237 list_del(&ifp->tmp_list);
1238 if (ifp->ifpub) {
1239 in6_ifa_put(ifp->ifpub);
1240 ifp->ifpub = NULL;
1242 __in6_ifa_put(ifp);
1245 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1246 action = check_cleanup_prefix_route(ifp, &expires);
1248 list_del_rcu(&ifp->if_list);
1249 __in6_ifa_put(ifp);
1251 write_unlock_bh(&ifp->idev->lock);
1253 addrconf_del_dad_work(ifp);
1255 ipv6_ifa_notify(RTM_DELADDR, ifp);
1257 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1259 if (action != CLEANUP_PREFIX_RT_NOP) {
1260 cleanup_prefix_route(ifp, expires,
1261 action == CLEANUP_PREFIX_RT_DEL);
1264 /* clean up prefsrc entries */
1265 rt6_remove_prefsrc(ifp);
1266 out:
1267 in6_ifa_put(ifp);
1270 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1271 struct inet6_ifaddr *ift,
1272 bool block)
1274 struct inet6_dev *idev = ifp->idev;
1275 struct in6_addr addr, *tmpaddr;
1276 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1277 unsigned long regen_advance;
1278 int tmp_plen;
1279 int ret = 0;
1280 u32 addr_flags;
1281 unsigned long now = jiffies;
1282 long max_desync_factor;
1283 s32 cnf_temp_preferred_lft;
1285 write_lock_bh(&idev->lock);
1286 if (ift) {
1287 spin_lock_bh(&ift->lock);
1288 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1289 spin_unlock_bh(&ift->lock);
1290 tmpaddr = &addr;
1291 } else {
1292 tmpaddr = NULL;
1294 retry:
1295 in6_dev_hold(idev);
1296 if (idev->cnf.use_tempaddr <= 0) {
1297 write_unlock_bh(&idev->lock);
1298 pr_info("%s: use_tempaddr is disabled\n", __func__);
1299 in6_dev_put(idev);
1300 ret = -1;
1301 goto out;
1303 spin_lock_bh(&ifp->lock);
1304 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1305 idev->cnf.use_tempaddr = -1; /*XXX*/
1306 spin_unlock_bh(&ifp->lock);
1307 write_unlock_bh(&idev->lock);
1308 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1309 __func__);
1310 in6_dev_put(idev);
1311 ret = -1;
1312 goto out;
1314 in6_ifa_hold(ifp);
1315 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1316 ipv6_try_regen_rndid(idev, tmpaddr);
1317 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1318 age = (now - ifp->tstamp) / HZ;
1320 regen_advance = idev->cnf.regen_max_retry *
1321 idev->cnf.dad_transmits *
1322 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1324 /* recalculate max_desync_factor each time and update
1325 * idev->desync_factor if it's larger
1327 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1328 max_desync_factor = min_t(__u32,
1329 idev->cnf.max_desync_factor,
1330 cnf_temp_preferred_lft - regen_advance);
1332 if (unlikely(idev->desync_factor > max_desync_factor)) {
1333 if (max_desync_factor > 0) {
1334 get_random_bytes(&idev->desync_factor,
1335 sizeof(idev->desync_factor));
1336 idev->desync_factor %= max_desync_factor;
1337 } else {
1338 idev->desync_factor = 0;
1342 tmp_valid_lft = min_t(__u32,
1343 ifp->valid_lft,
1344 idev->cnf.temp_valid_lft + age);
1345 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1346 idev->desync_factor;
1347 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1348 tmp_plen = ifp->prefix_len;
1349 tmp_tstamp = ifp->tstamp;
1350 spin_unlock_bh(&ifp->lock);
1352 write_unlock_bh(&idev->lock);
1354 /* A temporary address is created only if this calculated Preferred
1355 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1356 * an implementation must not create a temporary address with a zero
1357 * Preferred Lifetime.
1358 * Use age calculation as in addrconf_verify to avoid unnecessary
1359 * temporary addresses being generated.
1361 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1362 if (tmp_prefered_lft <= regen_advance + age) {
1363 in6_ifa_put(ifp);
1364 in6_dev_put(idev);
1365 ret = -1;
1366 goto out;
1369 addr_flags = IFA_F_TEMPORARY;
1370 /* set in addrconf_prefix_rcv() */
1371 if (ifp->flags & IFA_F_OPTIMISTIC)
1372 addr_flags |= IFA_F_OPTIMISTIC;
1374 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1375 ipv6_addr_scope(&addr), addr_flags,
1376 tmp_valid_lft, tmp_prefered_lft, block, NULL);
1377 if (IS_ERR(ift)) {
1378 in6_ifa_put(ifp);
1379 in6_dev_put(idev);
1380 pr_info("%s: retry temporary address regeneration\n", __func__);
1381 tmpaddr = &addr;
1382 write_lock_bh(&idev->lock);
1383 goto retry;
1386 spin_lock_bh(&ift->lock);
1387 ift->ifpub = ifp;
1388 ift->cstamp = now;
1389 ift->tstamp = tmp_tstamp;
1390 spin_unlock_bh(&ift->lock);
1392 addrconf_dad_start(ift);
1393 in6_ifa_put(ift);
1394 in6_dev_put(idev);
1395 out:
1396 return ret;
1400 * Choose an appropriate source address (RFC3484)
1402 enum {
1403 IPV6_SADDR_RULE_INIT = 0,
1404 IPV6_SADDR_RULE_LOCAL,
1405 IPV6_SADDR_RULE_SCOPE,
1406 IPV6_SADDR_RULE_PREFERRED,
1407 #ifdef CONFIG_IPV6_MIP6
1408 IPV6_SADDR_RULE_HOA,
1409 #endif
1410 IPV6_SADDR_RULE_OIF,
1411 IPV6_SADDR_RULE_LABEL,
1412 IPV6_SADDR_RULE_PRIVACY,
1413 IPV6_SADDR_RULE_ORCHID,
1414 IPV6_SADDR_RULE_PREFIX,
1415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1416 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1417 #endif
1418 IPV6_SADDR_RULE_MAX
1421 struct ipv6_saddr_score {
1422 int rule;
1423 int addr_type;
1424 struct inet6_ifaddr *ifa;
1425 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1426 int scopedist;
1427 int matchlen;
1430 struct ipv6_saddr_dst {
1431 const struct in6_addr *addr;
1432 int ifindex;
1433 int scope;
1434 int label;
1435 unsigned int prefs;
1438 static inline int ipv6_saddr_preferred(int type)
1440 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1441 return 1;
1442 return 0;
1445 static bool ipv6_use_optimistic_addr(struct net *net,
1446 struct inet6_dev *idev)
1448 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1449 if (!idev)
1450 return false;
1451 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1452 return false;
1453 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1454 return false;
1456 return true;
1457 #else
1458 return false;
1459 #endif
1462 static int ipv6_get_saddr_eval(struct net *net,
1463 struct ipv6_saddr_score *score,
1464 struct ipv6_saddr_dst *dst,
1465 int i)
1467 int ret;
1469 if (i <= score->rule) {
1470 switch (i) {
1471 case IPV6_SADDR_RULE_SCOPE:
1472 ret = score->scopedist;
1473 break;
1474 case IPV6_SADDR_RULE_PREFIX:
1475 ret = score->matchlen;
1476 break;
1477 default:
1478 ret = !!test_bit(i, score->scorebits);
1480 goto out;
1483 switch (i) {
1484 case IPV6_SADDR_RULE_INIT:
1485 /* Rule 0: remember if hiscore is not ready yet */
1486 ret = !!score->ifa;
1487 break;
1488 case IPV6_SADDR_RULE_LOCAL:
1489 /* Rule 1: Prefer same address */
1490 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1491 break;
1492 case IPV6_SADDR_RULE_SCOPE:
1493 /* Rule 2: Prefer appropriate scope
1495 * ret
1497 * -1 | d 15
1498 * ---+--+-+---> scope
1500 * | d is scope of the destination.
1501 * B-d | \
1502 * | \ <- smaller scope is better if
1503 * B-15 | \ if scope is enough for destination.
1504 * | ret = B - scope (-1 <= scope >= d <= 15).
1505 * d-C-1 | /
1506 * |/ <- greater is better
1507 * -C / if scope is not enough for destination.
1508 * /| ret = scope - C (-1 <= d < scope <= 15).
1510 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1511 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1512 * Assume B = 0 and we get C > 29.
1514 ret = __ipv6_addr_src_scope(score->addr_type);
1515 if (ret >= dst->scope)
1516 ret = -ret;
1517 else
1518 ret -= 128; /* 30 is enough */
1519 score->scopedist = ret;
1520 break;
1521 case IPV6_SADDR_RULE_PREFERRED:
1523 /* Rule 3: Avoid deprecated and optimistic addresses */
1524 u8 avoid = IFA_F_DEPRECATED;
1526 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1527 avoid |= IFA_F_OPTIMISTIC;
1528 ret = ipv6_saddr_preferred(score->addr_type) ||
1529 !(score->ifa->flags & avoid);
1530 break;
1532 #ifdef CONFIG_IPV6_MIP6
1533 case IPV6_SADDR_RULE_HOA:
1535 /* Rule 4: Prefer home address */
1536 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1537 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1538 break;
1540 #endif
1541 case IPV6_SADDR_RULE_OIF:
1542 /* Rule 5: Prefer outgoing interface */
1543 ret = (!dst->ifindex ||
1544 dst->ifindex == score->ifa->idev->dev->ifindex);
1545 break;
1546 case IPV6_SADDR_RULE_LABEL:
1547 /* Rule 6: Prefer matching label */
1548 ret = ipv6_addr_label(net,
1549 &score->ifa->addr, score->addr_type,
1550 score->ifa->idev->dev->ifindex) == dst->label;
1551 break;
1552 case IPV6_SADDR_RULE_PRIVACY:
1554 /* Rule 7: Prefer public address
1555 * Note: prefer temporary address if use_tempaddr >= 2
1557 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1558 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1559 score->ifa->idev->cnf.use_tempaddr >= 2;
1560 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1561 break;
1563 case IPV6_SADDR_RULE_ORCHID:
1564 /* Rule 8-: Prefer ORCHID vs ORCHID or
1565 * non-ORCHID vs non-ORCHID
1567 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1568 ipv6_addr_orchid(dst->addr));
1569 break;
1570 case IPV6_SADDR_RULE_PREFIX:
1571 /* Rule 8: Use longest matching prefix */
1572 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1573 if (ret > score->ifa->prefix_len)
1574 ret = score->ifa->prefix_len;
1575 score->matchlen = ret;
1576 break;
1577 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1578 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1579 /* Optimistic addresses still have lower precedence than other
1580 * preferred addresses.
1582 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1583 break;
1584 #endif
1585 default:
1586 ret = 0;
1589 if (ret)
1590 __set_bit(i, score->scorebits);
1591 score->rule = i;
1592 out:
1593 return ret;
1596 static int __ipv6_dev_get_saddr(struct net *net,
1597 struct ipv6_saddr_dst *dst,
1598 struct inet6_dev *idev,
1599 struct ipv6_saddr_score *scores,
1600 int hiscore_idx)
1602 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1604 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1605 int i;
1608 * - Tentative Address (RFC2462 section 5.4)
1609 * - A tentative address is not considered
1610 * "assigned to an interface" in the traditional
1611 * sense, unless it is also flagged as optimistic.
1612 * - Candidate Source Address (section 4)
1613 * - In any case, anycast addresses, multicast
1614 * addresses, and the unspecified address MUST
1615 * NOT be included in a candidate set.
1617 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1618 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1619 continue;
1621 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1623 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1624 score->addr_type & IPV6_ADDR_MULTICAST)) {
1625 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1626 idev->dev->name);
1627 continue;
1630 score->rule = -1;
1631 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1633 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1634 int minihiscore, miniscore;
1636 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1637 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1639 if (minihiscore > miniscore) {
1640 if (i == IPV6_SADDR_RULE_SCOPE &&
1641 score->scopedist > 0) {
1643 * special case:
1644 * each remaining entry
1645 * has too small (not enough)
1646 * scope, because ifa entries
1647 * are sorted by their scope
1648 * values.
1650 goto out;
1652 break;
1653 } else if (minihiscore < miniscore) {
1654 swap(hiscore, score);
1655 hiscore_idx = 1 - hiscore_idx;
1657 /* restore our iterator */
1658 score->ifa = hiscore->ifa;
1660 break;
1664 out:
1665 return hiscore_idx;
1668 static int ipv6_get_saddr_master(struct net *net,
1669 const struct net_device *dst_dev,
1670 const struct net_device *master,
1671 struct ipv6_saddr_dst *dst,
1672 struct ipv6_saddr_score *scores,
1673 int hiscore_idx)
1675 struct inet6_dev *idev;
1677 idev = __in6_dev_get(dst_dev);
1678 if (idev)
1679 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1680 scores, hiscore_idx);
1682 idev = __in6_dev_get(master);
1683 if (idev)
1684 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1685 scores, hiscore_idx);
1687 return hiscore_idx;
1690 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1691 const struct in6_addr *daddr, unsigned int prefs,
1692 struct in6_addr *saddr)
1694 struct ipv6_saddr_score scores[2], *hiscore;
1695 struct ipv6_saddr_dst dst;
1696 struct inet6_dev *idev;
1697 struct net_device *dev;
1698 int dst_type;
1699 bool use_oif_addr = false;
1700 int hiscore_idx = 0;
1701 int ret = 0;
1703 dst_type = __ipv6_addr_type(daddr);
1704 dst.addr = daddr;
1705 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1706 dst.scope = __ipv6_addr_src_scope(dst_type);
1707 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1708 dst.prefs = prefs;
1710 scores[hiscore_idx].rule = -1;
1711 scores[hiscore_idx].ifa = NULL;
1713 rcu_read_lock();
1715 /* Candidate Source Address (section 4)
1716 * - multicast and link-local destination address,
1717 * the set of candidate source address MUST only
1718 * include addresses assigned to interfaces
1719 * belonging to the same link as the outgoing
1720 * interface.
1721 * (- For site-local destination addresses, the
1722 * set of candidate source addresses MUST only
1723 * include addresses assigned to interfaces
1724 * belonging to the same site as the outgoing
1725 * interface.)
1726 * - "It is RECOMMENDED that the candidate source addresses
1727 * be the set of unicast addresses assigned to the
1728 * interface that will be used to send to the destination
1729 * (the 'outgoing' interface)." (RFC 6724)
1731 if (dst_dev) {
1732 idev = __in6_dev_get(dst_dev);
1733 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1734 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1735 (idev && idev->cnf.use_oif_addrs_only)) {
1736 use_oif_addr = true;
1740 if (use_oif_addr) {
1741 if (idev)
1742 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1743 } else {
1744 const struct net_device *master;
1745 int master_idx = 0;
1747 /* if dst_dev exists and is enslaved to an L3 device, then
1748 * prefer addresses from dst_dev and then the master over
1749 * any other enslaved devices in the L3 domain.
1751 master = l3mdev_master_dev_rcu(dst_dev);
1752 if (master) {
1753 master_idx = master->ifindex;
1755 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1756 master, &dst,
1757 scores, hiscore_idx);
1759 if (scores[hiscore_idx].ifa)
1760 goto out;
1763 for_each_netdev_rcu(net, dev) {
1764 /* only consider addresses on devices in the
1765 * same L3 domain
1767 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1768 continue;
1769 idev = __in6_dev_get(dev);
1770 if (!idev)
1771 continue;
1772 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1776 out:
1777 hiscore = &scores[hiscore_idx];
1778 if (!hiscore->ifa)
1779 ret = -EADDRNOTAVAIL;
1780 else
1781 *saddr = hiscore->ifa->addr;
1783 rcu_read_unlock();
1784 return ret;
1786 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1788 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1789 u32 banned_flags)
1791 struct inet6_ifaddr *ifp;
1792 int err = -EADDRNOTAVAIL;
1794 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1795 if (ifp->scope > IFA_LINK)
1796 break;
1797 if (ifp->scope == IFA_LINK &&
1798 !(ifp->flags & banned_flags)) {
1799 *addr = ifp->addr;
1800 err = 0;
1801 break;
1804 return err;
1807 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1808 u32 banned_flags)
1810 struct inet6_dev *idev;
1811 int err = -EADDRNOTAVAIL;
1813 rcu_read_lock();
1814 idev = __in6_dev_get(dev);
1815 if (idev) {
1816 read_lock_bh(&idev->lock);
1817 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1818 read_unlock_bh(&idev->lock);
1820 rcu_read_unlock();
1821 return err;
1824 static int ipv6_count_addresses(const struct inet6_dev *idev)
1826 const struct inet6_ifaddr *ifp;
1827 int cnt = 0;
1829 rcu_read_lock();
1830 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1831 cnt++;
1832 rcu_read_unlock();
1833 return cnt;
1836 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1837 const struct net_device *dev, int strict)
1839 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1841 EXPORT_SYMBOL(ipv6_chk_addr);
1843 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1844 const struct net_device *dev, int strict,
1845 u32 banned_flags)
1847 unsigned int hash = inet6_addr_hash(net, addr);
1848 struct inet6_ifaddr *ifp;
1849 u32 ifp_flags;
1851 rcu_read_lock();
1852 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1853 if (!net_eq(dev_net(ifp->idev->dev), net))
1854 continue;
1855 /* Decouple optimistic from tentative for evaluation here.
1856 * Ban optimistic addresses explicitly, when required.
1858 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1859 ? (ifp->flags&~IFA_F_TENTATIVE)
1860 : ifp->flags;
1861 if (ipv6_addr_equal(&ifp->addr, addr) &&
1862 !(ifp_flags&banned_flags) &&
1863 (!dev || ifp->idev->dev == dev ||
1864 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1865 rcu_read_unlock();
1866 return 1;
1870 rcu_read_unlock();
1871 return 0;
1873 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1876 /* Compares an address/prefix_len with addresses on device @dev.
1877 * If one is found it returns true.
1879 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1880 const unsigned int prefix_len, struct net_device *dev)
1882 const struct inet6_ifaddr *ifa;
1883 const struct inet6_dev *idev;
1884 bool ret = false;
1886 rcu_read_lock();
1887 idev = __in6_dev_get(dev);
1888 if (idev) {
1889 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1890 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1891 if (ret)
1892 break;
1895 rcu_read_unlock();
1897 return ret;
1899 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1901 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1903 const struct inet6_ifaddr *ifa;
1904 const struct inet6_dev *idev;
1905 int onlink;
1907 onlink = 0;
1908 rcu_read_lock();
1909 idev = __in6_dev_get(dev);
1910 if (idev) {
1911 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1912 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1913 ifa->prefix_len);
1914 if (onlink)
1915 break;
1918 rcu_read_unlock();
1919 return onlink;
1921 EXPORT_SYMBOL(ipv6_chk_prefix);
1923 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1924 struct net_device *dev, int strict)
1926 unsigned int hash = inet6_addr_hash(net, addr);
1927 struct inet6_ifaddr *ifp, *result = NULL;
1929 rcu_read_lock();
1930 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1931 if (!net_eq(dev_net(ifp->idev->dev), net))
1932 continue;
1933 if (ipv6_addr_equal(&ifp->addr, addr)) {
1934 if (!dev || ifp->idev->dev == dev ||
1935 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1936 result = ifp;
1937 in6_ifa_hold(ifp);
1938 break;
1942 rcu_read_unlock();
1944 return result;
1947 /* Gets referenced address, destroys ifaddr */
1949 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1951 if (dad_failed)
1952 ifp->flags |= IFA_F_DADFAILED;
1954 if (ifp->flags&IFA_F_TEMPORARY) {
1955 struct inet6_ifaddr *ifpub;
1956 spin_lock_bh(&ifp->lock);
1957 ifpub = ifp->ifpub;
1958 if (ifpub) {
1959 in6_ifa_hold(ifpub);
1960 spin_unlock_bh(&ifp->lock);
1961 ipv6_create_tempaddr(ifpub, ifp, true);
1962 in6_ifa_put(ifpub);
1963 } else {
1964 spin_unlock_bh(&ifp->lock);
1966 ipv6_del_addr(ifp);
1967 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
1968 spin_lock_bh(&ifp->lock);
1969 addrconf_del_dad_work(ifp);
1970 ifp->flags |= IFA_F_TENTATIVE;
1971 spin_unlock_bh(&ifp->lock);
1972 if (dad_failed)
1973 ipv6_ifa_notify(0, ifp);
1974 in6_ifa_put(ifp);
1975 } else {
1976 ipv6_del_addr(ifp);
1980 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1982 int err = -ENOENT;
1984 spin_lock_bh(&ifp->lock);
1985 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1986 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1987 err = 0;
1989 spin_unlock_bh(&ifp->lock);
1991 return err;
1994 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
1996 struct inet6_dev *idev = ifp->idev;
1997 struct net *net = dev_net(ifp->idev->dev);
1999 if (addrconf_dad_end(ifp)) {
2000 in6_ifa_put(ifp);
2001 return;
2004 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2005 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2007 spin_lock_bh(&ifp->lock);
2009 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2010 int scope = ifp->scope;
2011 u32 flags = ifp->flags;
2012 struct in6_addr new_addr;
2013 struct inet6_ifaddr *ifp2;
2014 u32 valid_lft, preferred_lft;
2015 int pfxlen = ifp->prefix_len;
2016 int retries = ifp->stable_privacy_retry + 1;
2018 if (retries > net->ipv6.sysctl.idgen_retries) {
2019 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2020 ifp->idev->dev->name);
2021 goto errdad;
2024 new_addr = ifp->addr;
2025 if (ipv6_generate_stable_address(&new_addr, retries,
2026 idev))
2027 goto errdad;
2029 valid_lft = ifp->valid_lft;
2030 preferred_lft = ifp->prefered_lft;
2032 spin_unlock_bh(&ifp->lock);
2034 if (idev->cnf.max_addresses &&
2035 ipv6_count_addresses(idev) >=
2036 idev->cnf.max_addresses)
2037 goto lock_errdad;
2039 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2040 ifp->idev->dev->name);
2042 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2043 scope, flags, valid_lft,
2044 preferred_lft, false, NULL);
2045 if (IS_ERR(ifp2))
2046 goto lock_errdad;
2048 spin_lock_bh(&ifp2->lock);
2049 ifp2->stable_privacy_retry = retries;
2050 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2051 spin_unlock_bh(&ifp2->lock);
2053 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2054 in6_ifa_put(ifp2);
2055 lock_errdad:
2056 spin_lock_bh(&ifp->lock);
2059 errdad:
2060 /* transition from _POSTDAD to _ERRDAD */
2061 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2062 spin_unlock_bh(&ifp->lock);
2064 addrconf_mod_dad_work(ifp, 0);
2065 in6_ifa_put(ifp);
2068 /* Join to solicited addr multicast group.
2069 * caller must hold RTNL */
2070 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2072 struct in6_addr maddr;
2074 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2075 return;
2077 addrconf_addr_solict_mult(addr, &maddr);
2078 ipv6_dev_mc_inc(dev, &maddr);
2081 /* caller must hold RTNL */
2082 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2084 struct in6_addr maddr;
2086 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2087 return;
2089 addrconf_addr_solict_mult(addr, &maddr);
2090 __ipv6_dev_mc_dec(idev, &maddr);
2093 /* caller must hold RTNL */
2094 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2096 struct in6_addr addr;
2098 if (ifp->prefix_len >= 127) /* RFC 6164 */
2099 return;
2100 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2101 if (ipv6_addr_any(&addr))
2102 return;
2103 __ipv6_dev_ac_inc(ifp->idev, &addr);
2106 /* caller must hold RTNL */
2107 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2109 struct in6_addr addr;
2111 if (ifp->prefix_len >= 127) /* RFC 6164 */
2112 return;
2113 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2114 if (ipv6_addr_any(&addr))
2115 return;
2116 __ipv6_dev_ac_dec(ifp->idev, &addr);
2119 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2121 switch (dev->addr_len) {
2122 case ETH_ALEN:
2123 memcpy(eui, dev->dev_addr, 3);
2124 eui[3] = 0xFF;
2125 eui[4] = 0xFE;
2126 memcpy(eui + 5, dev->dev_addr + 3, 3);
2127 break;
2128 case EUI64_ADDR_LEN:
2129 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2130 eui[0] ^= 2;
2131 break;
2132 default:
2133 return -1;
2136 return 0;
2139 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2141 union fwnet_hwaddr *ha;
2143 if (dev->addr_len != FWNET_ALEN)
2144 return -1;
2146 ha = (union fwnet_hwaddr *)dev->dev_addr;
2148 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2149 eui[0] ^= 2;
2150 return 0;
2153 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2155 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2156 if (dev->addr_len != ARCNET_ALEN)
2157 return -1;
2158 memset(eui, 0, 7);
2159 eui[7] = *(u8 *)dev->dev_addr;
2160 return 0;
2163 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2165 if (dev->addr_len != INFINIBAND_ALEN)
2166 return -1;
2167 memcpy(eui, dev->dev_addr + 12, 8);
2168 eui[0] |= 2;
2169 return 0;
2172 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2174 if (addr == 0)
2175 return -1;
2176 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2177 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2178 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2179 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2180 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2181 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2182 eui[1] = 0;
2183 eui[2] = 0x5E;
2184 eui[3] = 0xFE;
2185 memcpy(eui + 4, &addr, 4);
2186 return 0;
2189 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2191 if (dev->priv_flags & IFF_ISATAP)
2192 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2193 return -1;
2196 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2198 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2201 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2203 memcpy(eui, dev->perm_addr, 3);
2204 memcpy(eui + 5, dev->perm_addr + 3, 3);
2205 eui[3] = 0xFF;
2206 eui[4] = 0xFE;
2207 eui[0] ^= 2;
2208 return 0;
2211 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2213 switch (dev->type) {
2214 case ARPHRD_ETHER:
2215 case ARPHRD_FDDI:
2216 return addrconf_ifid_eui48(eui, dev);
2217 case ARPHRD_ARCNET:
2218 return addrconf_ifid_arcnet(eui, dev);
2219 case ARPHRD_INFINIBAND:
2220 return addrconf_ifid_infiniband(eui, dev);
2221 case ARPHRD_SIT:
2222 return addrconf_ifid_sit(eui, dev);
2223 case ARPHRD_IPGRE:
2224 case ARPHRD_TUNNEL:
2225 return addrconf_ifid_gre(eui, dev);
2226 case ARPHRD_6LOWPAN:
2227 return addrconf_ifid_6lowpan(eui, dev);
2228 case ARPHRD_IEEE1394:
2229 return addrconf_ifid_ieee1394(eui, dev);
2230 case ARPHRD_TUNNEL6:
2231 case ARPHRD_IP6GRE:
2232 return addrconf_ifid_ip6tnl(eui, dev);
2234 return -1;
2237 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2239 int err = -1;
2240 struct inet6_ifaddr *ifp;
2242 read_lock_bh(&idev->lock);
2243 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2244 if (ifp->scope > IFA_LINK)
2245 break;
2246 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2247 memcpy(eui, ifp->addr.s6_addr+8, 8);
2248 err = 0;
2249 break;
2252 read_unlock_bh(&idev->lock);
2253 return err;
2256 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2257 static void ipv6_regen_rndid(struct inet6_dev *idev)
2259 regen:
2260 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2261 idev->rndid[0] &= ~0x02;
2264 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2265 * check if generated address is not inappropriate
2267 * - Reserved subnet anycast (RFC 2526)
2268 * 11111101 11....11 1xxxxxxx
2269 * - ISATAP (RFC4214) 6.1
2270 * 00-00-5E-FE-xx-xx-xx-xx
2271 * - value 0
2272 * - XXX: already assigned to an address on the device
2274 if (idev->rndid[0] == 0xfd &&
2275 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2276 (idev->rndid[7]&0x80))
2277 goto regen;
2278 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2279 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2280 goto regen;
2281 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2282 goto regen;
2286 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2288 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2289 ipv6_regen_rndid(idev);
2293 * Add prefix route.
2296 static void
2297 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2298 unsigned long expires, u32 flags)
2300 struct fib6_config cfg = {
2301 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2302 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2303 .fc_ifindex = dev->ifindex,
2304 .fc_expires = expires,
2305 .fc_dst_len = plen,
2306 .fc_flags = RTF_UP | flags,
2307 .fc_nlinfo.nl_net = dev_net(dev),
2308 .fc_protocol = RTPROT_KERNEL,
2311 cfg.fc_dst = *pfx;
2313 /* Prevent useless cloning on PtP SIT.
2314 This thing is done here expecting that the whole
2315 class of non-broadcast devices need not cloning.
2317 #if IS_ENABLED(CONFIG_IPV6_SIT)
2318 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2319 cfg.fc_flags |= RTF_NONEXTHOP;
2320 #endif
2322 ip6_route_add(&cfg, NULL);
2326 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2327 int plen,
2328 const struct net_device *dev,
2329 u32 flags, u32 noflags)
2331 struct fib6_node *fn;
2332 struct rt6_info *rt = NULL;
2333 struct fib6_table *table;
2334 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2336 table = fib6_get_table(dev_net(dev), tb_id);
2337 if (!table)
2338 return NULL;
2340 rcu_read_lock();
2341 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2342 if (!fn)
2343 goto out;
2345 for_each_fib6_node_rt_rcu(fn) {
2346 if (rt->dst.dev->ifindex != dev->ifindex)
2347 continue;
2348 if ((rt->rt6i_flags & flags) != flags)
2349 continue;
2350 if ((rt->rt6i_flags & noflags) != 0)
2351 continue;
2352 if (!dst_hold_safe(&rt->dst))
2353 rt = NULL;
2354 break;
2356 out:
2357 rcu_read_unlock();
2358 return rt;
2362 /* Create "default" multicast route to the interface */
2364 static void addrconf_add_mroute(struct net_device *dev)
2366 struct fib6_config cfg = {
2367 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2368 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2369 .fc_ifindex = dev->ifindex,
2370 .fc_dst_len = 8,
2371 .fc_flags = RTF_UP,
2372 .fc_nlinfo.nl_net = dev_net(dev),
2375 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2377 ip6_route_add(&cfg, NULL);
2380 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2382 struct inet6_dev *idev;
2384 ASSERT_RTNL();
2386 idev = ipv6_find_idev(dev);
2387 if (!idev)
2388 return ERR_PTR(-ENOBUFS);
2390 if (idev->cnf.disable_ipv6)
2391 return ERR_PTR(-EACCES);
2393 /* Add default multicast route */
2394 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2395 addrconf_add_mroute(dev);
2397 return idev;
2400 static void manage_tempaddrs(struct inet6_dev *idev,
2401 struct inet6_ifaddr *ifp,
2402 __u32 valid_lft, __u32 prefered_lft,
2403 bool create, unsigned long now)
2405 u32 flags;
2406 struct inet6_ifaddr *ift;
2408 read_lock_bh(&idev->lock);
2409 /* update all temporary addresses in the list */
2410 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2411 int age, max_valid, max_prefered;
2413 if (ifp != ift->ifpub)
2414 continue;
2416 /* RFC 4941 section 3.3:
2417 * If a received option will extend the lifetime of a public
2418 * address, the lifetimes of temporary addresses should
2419 * be extended, subject to the overall constraint that no
2420 * temporary addresses should ever remain "valid" or "preferred"
2421 * for a time longer than (TEMP_VALID_LIFETIME) or
2422 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2424 age = (now - ift->cstamp) / HZ;
2425 max_valid = idev->cnf.temp_valid_lft - age;
2426 if (max_valid < 0)
2427 max_valid = 0;
2429 max_prefered = idev->cnf.temp_prefered_lft -
2430 idev->desync_factor - age;
2431 if (max_prefered < 0)
2432 max_prefered = 0;
2434 if (valid_lft > max_valid)
2435 valid_lft = max_valid;
2437 if (prefered_lft > max_prefered)
2438 prefered_lft = max_prefered;
2440 spin_lock(&ift->lock);
2441 flags = ift->flags;
2442 ift->valid_lft = valid_lft;
2443 ift->prefered_lft = prefered_lft;
2444 ift->tstamp = now;
2445 if (prefered_lft > 0)
2446 ift->flags &= ~IFA_F_DEPRECATED;
2448 spin_unlock(&ift->lock);
2449 if (!(flags&IFA_F_TENTATIVE))
2450 ipv6_ifa_notify(0, ift);
2453 if ((create || list_empty(&idev->tempaddr_list)) &&
2454 idev->cnf.use_tempaddr > 0) {
2455 /* When a new public address is created as described
2456 * in [ADDRCONF], also create a new temporary address.
2457 * Also create a temporary address if it's enabled but
2458 * no temporary address currently exists.
2460 read_unlock_bh(&idev->lock);
2461 ipv6_create_tempaddr(ifp, NULL, false);
2462 } else {
2463 read_unlock_bh(&idev->lock);
2467 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2469 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2470 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2473 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2474 const struct prefix_info *pinfo,
2475 struct inet6_dev *in6_dev,
2476 const struct in6_addr *addr, int addr_type,
2477 u32 addr_flags, bool sllao, bool tokenized,
2478 __u32 valid_lft, u32 prefered_lft)
2480 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2481 int create = 0, update_lft = 0;
2483 if (!ifp && valid_lft) {
2484 int max_addresses = in6_dev->cnf.max_addresses;
2486 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2487 if ((net->ipv6.devconf_all->optimistic_dad ||
2488 in6_dev->cnf.optimistic_dad) &&
2489 !net->ipv6.devconf_all->forwarding && sllao)
2490 addr_flags |= IFA_F_OPTIMISTIC;
2491 #endif
2493 /* Do not allow to create too much of autoconfigured
2494 * addresses; this would be too easy way to crash kernel.
2496 if (!max_addresses ||
2497 ipv6_count_addresses(in6_dev) < max_addresses)
2498 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2499 pinfo->prefix_len,
2500 addr_type&IPV6_ADDR_SCOPE_MASK,
2501 addr_flags, valid_lft,
2502 prefered_lft, false, NULL);
2504 if (IS_ERR_OR_NULL(ifp))
2505 return -1;
2507 update_lft = 0;
2508 create = 1;
2509 spin_lock_bh(&ifp->lock);
2510 ifp->flags |= IFA_F_MANAGETEMPADDR;
2511 ifp->cstamp = jiffies;
2512 ifp->tokenized = tokenized;
2513 spin_unlock_bh(&ifp->lock);
2514 addrconf_dad_start(ifp);
2517 if (ifp) {
2518 u32 flags;
2519 unsigned long now;
2520 u32 stored_lft;
2522 /* update lifetime (RFC2462 5.5.3 e) */
2523 spin_lock_bh(&ifp->lock);
2524 now = jiffies;
2525 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2526 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2527 else
2528 stored_lft = 0;
2529 if (!update_lft && !create && stored_lft) {
2530 const u32 minimum_lft = min_t(u32,
2531 stored_lft, MIN_VALID_LIFETIME);
2532 valid_lft = max(valid_lft, minimum_lft);
2534 /* RFC4862 Section 5.5.3e:
2535 * "Note that the preferred lifetime of the
2536 * corresponding address is always reset to
2537 * the Preferred Lifetime in the received
2538 * Prefix Information option, regardless of
2539 * whether the valid lifetime is also reset or
2540 * ignored."
2542 * So we should always update prefered_lft here.
2544 update_lft = 1;
2547 if (update_lft) {
2548 ifp->valid_lft = valid_lft;
2549 ifp->prefered_lft = prefered_lft;
2550 ifp->tstamp = now;
2551 flags = ifp->flags;
2552 ifp->flags &= ~IFA_F_DEPRECATED;
2553 spin_unlock_bh(&ifp->lock);
2555 if (!(flags&IFA_F_TENTATIVE))
2556 ipv6_ifa_notify(0, ifp);
2557 } else
2558 spin_unlock_bh(&ifp->lock);
2560 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2561 create, now);
2563 in6_ifa_put(ifp);
2564 addrconf_verify();
2567 return 0;
2569 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2571 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2573 struct prefix_info *pinfo;
2574 __u32 valid_lft;
2575 __u32 prefered_lft;
2576 int addr_type, err;
2577 u32 addr_flags = 0;
2578 struct inet6_dev *in6_dev;
2579 struct net *net = dev_net(dev);
2581 pinfo = (struct prefix_info *) opt;
2583 if (len < sizeof(struct prefix_info)) {
2584 ADBG("addrconf: prefix option too short\n");
2585 return;
2589 * Validation checks ([ADDRCONF], page 19)
2592 addr_type = ipv6_addr_type(&pinfo->prefix);
2594 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2595 return;
2597 valid_lft = ntohl(pinfo->valid);
2598 prefered_lft = ntohl(pinfo->prefered);
2600 if (prefered_lft > valid_lft) {
2601 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2602 return;
2605 in6_dev = in6_dev_get(dev);
2607 if (!in6_dev) {
2608 net_dbg_ratelimited("addrconf: device %s not configured\n",
2609 dev->name);
2610 return;
2614 * Two things going on here:
2615 * 1) Add routes for on-link prefixes
2616 * 2) Configure prefixes with the auto flag set
2619 if (pinfo->onlink) {
2620 struct rt6_info *rt;
2621 unsigned long rt_expires;
2623 /* Avoid arithmetic overflow. Really, we could
2624 * save rt_expires in seconds, likely valid_lft,
2625 * but it would require division in fib gc, that it
2626 * not good.
2628 if (HZ > USER_HZ)
2629 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2630 else
2631 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2633 if (addrconf_finite_timeout(rt_expires))
2634 rt_expires *= HZ;
2636 rt = addrconf_get_prefix_route(&pinfo->prefix,
2637 pinfo->prefix_len,
2638 dev,
2639 RTF_ADDRCONF | RTF_PREFIX_RT,
2640 RTF_GATEWAY | RTF_DEFAULT);
2642 if (rt) {
2643 /* Autoconf prefix route */
2644 if (valid_lft == 0) {
2645 ip6_del_rt(rt);
2646 rt = NULL;
2647 } else if (addrconf_finite_timeout(rt_expires)) {
2648 /* not infinity */
2649 rt6_set_expires(rt, jiffies + rt_expires);
2650 } else {
2651 rt6_clean_expires(rt);
2653 } else if (valid_lft) {
2654 clock_t expires = 0;
2655 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2656 if (addrconf_finite_timeout(rt_expires)) {
2657 /* not infinity */
2658 flags |= RTF_EXPIRES;
2659 expires = jiffies_to_clock_t(rt_expires);
2661 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2662 dev, expires, flags);
2664 ip6_rt_put(rt);
2667 /* Try to figure out our local address for this prefix */
2669 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2670 struct in6_addr addr;
2671 bool tokenized = false, dev_addr_generated = false;
2673 if (pinfo->prefix_len == 64) {
2674 memcpy(&addr, &pinfo->prefix, 8);
2676 if (!ipv6_addr_any(&in6_dev->token)) {
2677 read_lock_bh(&in6_dev->lock);
2678 memcpy(addr.s6_addr + 8,
2679 in6_dev->token.s6_addr + 8, 8);
2680 read_unlock_bh(&in6_dev->lock);
2681 tokenized = true;
2682 } else if (is_addr_mode_generate_stable(in6_dev) &&
2683 !ipv6_generate_stable_address(&addr, 0,
2684 in6_dev)) {
2685 addr_flags |= IFA_F_STABLE_PRIVACY;
2686 goto ok;
2687 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2688 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2689 goto put;
2690 } else {
2691 dev_addr_generated = true;
2693 goto ok;
2695 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2696 pinfo->prefix_len);
2697 goto put;
2700 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2701 &addr, addr_type,
2702 addr_flags, sllao,
2703 tokenized, valid_lft,
2704 prefered_lft);
2705 if (err)
2706 goto put;
2708 /* Ignore error case here because previous prefix add addr was
2709 * successful which will be notified.
2711 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2712 addr_type, addr_flags, sllao,
2713 tokenized, valid_lft,
2714 prefered_lft,
2715 dev_addr_generated);
2717 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2718 put:
2719 in6_dev_put(in6_dev);
2723 * Set destination address.
2724 * Special case for SIT interfaces where we create a new "virtual"
2725 * device.
2727 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2729 struct in6_ifreq ireq;
2730 struct net_device *dev;
2731 int err = -EINVAL;
2733 rtnl_lock();
2735 err = -EFAULT;
2736 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2737 goto err_exit;
2739 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2741 err = -ENODEV;
2742 if (!dev)
2743 goto err_exit;
2745 #if IS_ENABLED(CONFIG_IPV6_SIT)
2746 if (dev->type == ARPHRD_SIT) {
2747 const struct net_device_ops *ops = dev->netdev_ops;
2748 struct ifreq ifr;
2749 struct ip_tunnel_parm p;
2751 err = -EADDRNOTAVAIL;
2752 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2753 goto err_exit;
2755 memset(&p, 0, sizeof(p));
2756 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2757 p.iph.saddr = 0;
2758 p.iph.version = 4;
2759 p.iph.ihl = 5;
2760 p.iph.protocol = IPPROTO_IPV6;
2761 p.iph.ttl = 64;
2762 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2764 if (ops->ndo_do_ioctl) {
2765 mm_segment_t oldfs = get_fs();
2767 set_fs(KERNEL_DS);
2768 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2769 set_fs(oldfs);
2770 } else
2771 err = -EOPNOTSUPP;
2773 if (err == 0) {
2774 err = -ENOBUFS;
2775 dev = __dev_get_by_name(net, p.name);
2776 if (!dev)
2777 goto err_exit;
2778 err = dev_open(dev);
2781 #endif
2783 err_exit:
2784 rtnl_unlock();
2785 return err;
2788 static int ipv6_mc_config(struct sock *sk, bool join,
2789 const struct in6_addr *addr, int ifindex)
2791 int ret;
2793 ASSERT_RTNL();
2795 lock_sock(sk);
2796 if (join)
2797 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2798 else
2799 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2800 release_sock(sk);
2802 return ret;
2806 * Manual configuration of address on an interface
2808 static int inet6_addr_add(struct net *net, int ifindex,
2809 const struct in6_addr *pfx,
2810 const struct in6_addr *peer_pfx,
2811 unsigned int plen, __u32 ifa_flags,
2812 __u32 prefered_lft, __u32 valid_lft,
2813 struct netlink_ext_ack *extack)
2815 struct inet6_ifaddr *ifp;
2816 struct inet6_dev *idev;
2817 struct net_device *dev;
2818 unsigned long timeout;
2819 clock_t expires;
2820 int scope;
2821 u32 flags;
2823 ASSERT_RTNL();
2825 if (plen > 128)
2826 return -EINVAL;
2828 /* check the lifetime */
2829 if (!valid_lft || prefered_lft > valid_lft)
2830 return -EINVAL;
2832 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2833 return -EINVAL;
2835 dev = __dev_get_by_index(net, ifindex);
2836 if (!dev)
2837 return -ENODEV;
2839 idev = addrconf_add_dev(dev);
2840 if (IS_ERR(idev))
2841 return PTR_ERR(idev);
2843 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2844 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2845 true, pfx, ifindex);
2847 if (ret < 0)
2848 return ret;
2851 scope = ipv6_addr_scope(pfx);
2853 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2854 if (addrconf_finite_timeout(timeout)) {
2855 expires = jiffies_to_clock_t(timeout * HZ);
2856 valid_lft = timeout;
2857 flags = RTF_EXPIRES;
2858 } else {
2859 expires = 0;
2860 flags = 0;
2861 ifa_flags |= IFA_F_PERMANENT;
2864 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2865 if (addrconf_finite_timeout(timeout)) {
2866 if (timeout == 0)
2867 ifa_flags |= IFA_F_DEPRECATED;
2868 prefered_lft = timeout;
2871 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2872 valid_lft, prefered_lft, true, extack);
2874 if (!IS_ERR(ifp)) {
2875 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2876 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2877 expires, flags);
2881 * Note that section 3.1 of RFC 4429 indicates
2882 * that the Optimistic flag should not be set for
2883 * manually configured addresses
2885 addrconf_dad_start(ifp);
2886 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2887 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2888 true, jiffies);
2889 in6_ifa_put(ifp);
2890 addrconf_verify_rtnl();
2891 return 0;
2892 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2893 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2894 false, pfx, ifindex);
2897 return PTR_ERR(ifp);
2900 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2901 const struct in6_addr *pfx, unsigned int plen)
2903 struct inet6_ifaddr *ifp;
2904 struct inet6_dev *idev;
2905 struct net_device *dev;
2907 if (plen > 128)
2908 return -EINVAL;
2910 dev = __dev_get_by_index(net, ifindex);
2911 if (!dev)
2912 return -ENODEV;
2914 idev = __in6_dev_get(dev);
2915 if (!idev)
2916 return -ENXIO;
2918 read_lock_bh(&idev->lock);
2919 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2920 if (ifp->prefix_len == plen &&
2921 ipv6_addr_equal(pfx, &ifp->addr)) {
2922 in6_ifa_hold(ifp);
2923 read_unlock_bh(&idev->lock);
2925 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2926 (ifa_flags & IFA_F_MANAGETEMPADDR))
2927 manage_tempaddrs(idev, ifp, 0, 0, false,
2928 jiffies);
2929 ipv6_del_addr(ifp);
2930 addrconf_verify_rtnl();
2931 if (ipv6_addr_is_multicast(pfx)) {
2932 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2933 false, pfx, dev->ifindex);
2935 return 0;
2938 read_unlock_bh(&idev->lock);
2939 return -EADDRNOTAVAIL;
2943 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2945 struct in6_ifreq ireq;
2946 int err;
2948 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2949 return -EPERM;
2951 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2952 return -EFAULT;
2954 rtnl_lock();
2955 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2956 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2957 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, NULL);
2958 rtnl_unlock();
2959 return err;
2962 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2964 struct in6_ifreq ireq;
2965 int err;
2967 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2968 return -EPERM;
2970 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2971 return -EFAULT;
2973 rtnl_lock();
2974 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2975 ireq.ifr6_prefixlen);
2976 rtnl_unlock();
2977 return err;
2980 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2981 int plen, int scope)
2983 struct inet6_ifaddr *ifp;
2985 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2986 scope, IFA_F_PERMANENT,
2987 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME,
2988 true, NULL);
2989 if (!IS_ERR(ifp)) {
2990 spin_lock_bh(&ifp->lock);
2991 ifp->flags &= ~IFA_F_TENTATIVE;
2992 spin_unlock_bh(&ifp->lock);
2993 rt_genid_bump_ipv6(dev_net(idev->dev));
2994 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2995 in6_ifa_put(ifp);
2999 #if IS_ENABLED(CONFIG_IPV6_SIT)
3000 static void sit_add_v4_addrs(struct inet6_dev *idev)
3002 struct in6_addr addr;
3003 struct net_device *dev;
3004 struct net *net = dev_net(idev->dev);
3005 int scope, plen;
3006 u32 pflags = 0;
3008 ASSERT_RTNL();
3010 memset(&addr, 0, sizeof(struct in6_addr));
3011 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3013 if (idev->dev->flags&IFF_POINTOPOINT) {
3014 addr.s6_addr32[0] = htonl(0xfe800000);
3015 scope = IFA_LINK;
3016 plen = 64;
3017 } else {
3018 scope = IPV6_ADDR_COMPATv4;
3019 plen = 96;
3020 pflags |= RTF_NONEXTHOP;
3023 if (addr.s6_addr32[3]) {
3024 add_addr(idev, &addr, plen, scope);
3025 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
3026 return;
3029 for_each_netdev(net, dev) {
3030 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3031 if (in_dev && (dev->flags & IFF_UP)) {
3032 struct in_ifaddr *ifa;
3034 int flag = scope;
3036 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
3038 addr.s6_addr32[3] = ifa->ifa_local;
3040 if (ifa->ifa_scope == RT_SCOPE_LINK)
3041 continue;
3042 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3043 if (idev->dev->flags&IFF_POINTOPOINT)
3044 continue;
3045 flag |= IFA_HOST;
3048 add_addr(idev, &addr, plen, flag);
3049 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3050 pflags);
3055 #endif
3057 static void init_loopback(struct net_device *dev)
3059 struct inet6_dev *idev;
3061 /* ::1 */
3063 ASSERT_RTNL();
3065 idev = ipv6_find_idev(dev);
3066 if (!idev) {
3067 pr_debug("%s: add_dev failed\n", __func__);
3068 return;
3071 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3074 void addrconf_add_linklocal(struct inet6_dev *idev,
3075 const struct in6_addr *addr, u32 flags)
3077 struct inet6_ifaddr *ifp;
3078 u32 addr_flags = flags | IFA_F_PERMANENT;
3080 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3081 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3082 idev->cnf.optimistic_dad) &&
3083 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3084 addr_flags |= IFA_F_OPTIMISTIC;
3085 #endif
3087 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3088 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME, true, NULL);
3089 if (!IS_ERR(ifp)) {
3090 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3091 addrconf_dad_start(ifp);
3092 in6_ifa_put(ifp);
3095 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3097 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3099 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3100 return true;
3102 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3103 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3104 return true;
3106 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3107 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3108 return true;
3110 return false;
3113 static int ipv6_generate_stable_address(struct in6_addr *address,
3114 u8 dad_count,
3115 const struct inet6_dev *idev)
3117 static DEFINE_SPINLOCK(lock);
3118 static __u32 digest[SHA_DIGEST_WORDS];
3119 static __u32 workspace[SHA_WORKSPACE_WORDS];
3121 static union {
3122 char __data[SHA_MESSAGE_BYTES];
3123 struct {
3124 struct in6_addr secret;
3125 __be32 prefix[2];
3126 unsigned char hwaddr[MAX_ADDR_LEN];
3127 u8 dad_count;
3128 } __packed;
3129 } data;
3131 struct in6_addr secret;
3132 struct in6_addr temp;
3133 struct net *net = dev_net(idev->dev);
3135 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3137 if (idev->cnf.stable_secret.initialized)
3138 secret = idev->cnf.stable_secret.secret;
3139 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3140 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3141 else
3142 return -1;
3144 retry:
3145 spin_lock_bh(&lock);
3147 sha_init(digest);
3148 memset(&data, 0, sizeof(data));
3149 memset(workspace, 0, sizeof(workspace));
3150 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3151 data.prefix[0] = address->s6_addr32[0];
3152 data.prefix[1] = address->s6_addr32[1];
3153 data.secret = secret;
3154 data.dad_count = dad_count;
3156 sha_transform(digest, data.__data, workspace);
3158 temp = *address;
3159 temp.s6_addr32[2] = (__force __be32)digest[0];
3160 temp.s6_addr32[3] = (__force __be32)digest[1];
3162 spin_unlock_bh(&lock);
3164 if (ipv6_reserved_interfaceid(temp)) {
3165 dad_count++;
3166 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3167 return -1;
3168 goto retry;
3171 *address = temp;
3172 return 0;
3175 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3177 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3179 if (s->initialized)
3180 return;
3181 s = &idev->cnf.stable_secret;
3182 get_random_bytes(&s->secret, sizeof(s->secret));
3183 s->initialized = true;
3186 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3188 struct in6_addr addr;
3190 /* no link local addresses on L3 master devices */
3191 if (netif_is_l3_master(idev->dev))
3192 return;
3194 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3196 switch (idev->cnf.addr_gen_mode) {
3197 case IN6_ADDR_GEN_MODE_RANDOM:
3198 ipv6_gen_mode_random_init(idev);
3199 /* fallthrough */
3200 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3201 if (!ipv6_generate_stable_address(&addr, 0, idev))
3202 addrconf_add_linklocal(idev, &addr,
3203 IFA_F_STABLE_PRIVACY);
3204 else if (prefix_route)
3205 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3206 break;
3207 case IN6_ADDR_GEN_MODE_EUI64:
3208 /* addrconf_add_linklocal also adds a prefix_route and we
3209 * only need to care about prefix routes if ipv6_generate_eui64
3210 * couldn't generate one.
3212 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3213 addrconf_add_linklocal(idev, &addr, 0);
3214 else if (prefix_route)
3215 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3216 break;
3217 case IN6_ADDR_GEN_MODE_NONE:
3218 default:
3219 /* will not add any link local address */
3220 break;
3224 static void addrconf_dev_config(struct net_device *dev)
3226 struct inet6_dev *idev;
3228 ASSERT_RTNL();
3230 if ((dev->type != ARPHRD_ETHER) &&
3231 (dev->type != ARPHRD_FDDI) &&
3232 (dev->type != ARPHRD_ARCNET) &&
3233 (dev->type != ARPHRD_INFINIBAND) &&
3234 (dev->type != ARPHRD_IEEE1394) &&
3235 (dev->type != ARPHRD_TUNNEL6) &&
3236 (dev->type != ARPHRD_6LOWPAN) &&
3237 (dev->type != ARPHRD_IP6GRE) &&
3238 (dev->type != ARPHRD_IPGRE) &&
3239 (dev->type != ARPHRD_TUNNEL) &&
3240 (dev->type != ARPHRD_NONE)) {
3241 /* Alas, we support only Ethernet autoconfiguration. */
3242 return;
3245 idev = addrconf_add_dev(dev);
3246 if (IS_ERR(idev))
3247 return;
3249 /* this device type has no EUI support */
3250 if (dev->type == ARPHRD_NONE &&
3251 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3252 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3254 addrconf_addr_gen(idev, false);
3257 #if IS_ENABLED(CONFIG_IPV6_SIT)
3258 static void addrconf_sit_config(struct net_device *dev)
3260 struct inet6_dev *idev;
3262 ASSERT_RTNL();
3265 * Configure the tunnel with one of our IPv4
3266 * addresses... we should configure all of
3267 * our v4 addrs in the tunnel
3270 idev = ipv6_find_idev(dev);
3271 if (!idev) {
3272 pr_debug("%s: add_dev failed\n", __func__);
3273 return;
3276 if (dev->priv_flags & IFF_ISATAP) {
3277 addrconf_addr_gen(idev, false);
3278 return;
3281 sit_add_v4_addrs(idev);
3283 if (dev->flags&IFF_POINTOPOINT)
3284 addrconf_add_mroute(dev);
3286 #endif
3288 #if IS_ENABLED(CONFIG_NET_IPGRE)
3289 static void addrconf_gre_config(struct net_device *dev)
3291 struct inet6_dev *idev;
3293 ASSERT_RTNL();
3295 idev = ipv6_find_idev(dev);
3296 if (!idev) {
3297 pr_debug("%s: add_dev failed\n", __func__);
3298 return;
3301 addrconf_addr_gen(idev, true);
3302 if (dev->flags & IFF_POINTOPOINT)
3303 addrconf_add_mroute(dev);
3305 #endif
3307 static int fixup_permanent_addr(struct inet6_dev *idev,
3308 struct inet6_ifaddr *ifp)
3310 /* !rt6i_node means the host route was removed from the
3311 * FIB, for example, if 'lo' device is taken down. In that
3312 * case regenerate the host route.
3314 if (!ifp->rt || !ifp->rt->rt6i_node) {
3315 struct rt6_info *rt, *prev;
3317 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3318 if (IS_ERR(rt))
3319 return PTR_ERR(rt);
3321 /* ifp->rt can be accessed outside of rtnl */
3322 spin_lock(&ifp->lock);
3323 prev = ifp->rt;
3324 ifp->rt = rt;
3325 spin_unlock(&ifp->lock);
3327 ip6_rt_put(prev);
3330 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3331 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3332 idev->dev, 0, 0);
3335 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3336 addrconf_dad_start(ifp);
3338 return 0;
3341 static void addrconf_permanent_addr(struct net_device *dev)
3343 struct inet6_ifaddr *ifp, *tmp;
3344 struct inet6_dev *idev;
3346 idev = __in6_dev_get(dev);
3347 if (!idev)
3348 return;
3350 write_lock_bh(&idev->lock);
3352 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3353 if ((ifp->flags & IFA_F_PERMANENT) &&
3354 fixup_permanent_addr(idev, ifp) < 0) {
3355 write_unlock_bh(&idev->lock);
3356 in6_ifa_hold(ifp);
3357 ipv6_del_addr(ifp);
3358 write_lock_bh(&idev->lock);
3360 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3361 idev->dev->name, &ifp->addr);
3365 write_unlock_bh(&idev->lock);
3368 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3369 void *ptr)
3371 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3372 struct netdev_notifier_changeupper_info *info;
3373 struct inet6_dev *idev = __in6_dev_get(dev);
3374 struct net *net = dev_net(dev);
3375 int run_pending = 0;
3376 int err;
3378 switch (event) {
3379 case NETDEV_REGISTER:
3380 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3381 idev = ipv6_add_dev(dev);
3382 if (IS_ERR(idev))
3383 return notifier_from_errno(PTR_ERR(idev));
3385 break;
3387 case NETDEV_CHANGEMTU:
3388 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3389 if (dev->mtu < IPV6_MIN_MTU) {
3390 addrconf_ifdown(dev, dev != net->loopback_dev);
3391 break;
3394 if (idev) {
3395 rt6_mtu_change(dev, dev->mtu);
3396 idev->cnf.mtu6 = dev->mtu;
3397 break;
3400 /* allocate new idev */
3401 idev = ipv6_add_dev(dev);
3402 if (IS_ERR(idev))
3403 break;
3405 /* device is still not ready */
3406 if (!(idev->if_flags & IF_READY))
3407 break;
3409 run_pending = 1;
3411 /* fall through */
3413 case NETDEV_UP:
3414 case NETDEV_CHANGE:
3415 if (dev->flags & IFF_SLAVE)
3416 break;
3418 if (idev && idev->cnf.disable_ipv6)
3419 break;
3421 if (event == NETDEV_UP) {
3422 /* restore routes for permanent addresses */
3423 addrconf_permanent_addr(dev);
3425 if (!addrconf_link_ready(dev)) {
3426 /* device is not ready yet. */
3427 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3428 dev->name);
3429 break;
3432 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3433 idev = ipv6_add_dev(dev);
3435 if (!IS_ERR_OR_NULL(idev)) {
3436 idev->if_flags |= IF_READY;
3437 run_pending = 1;
3439 } else if (event == NETDEV_CHANGE) {
3440 if (!addrconf_link_ready(dev)) {
3441 /* device is still not ready. */
3442 rt6_sync_down_dev(dev, event);
3443 break;
3446 if (idev) {
3447 if (idev->if_flags & IF_READY) {
3448 /* device is already configured -
3449 * but resend MLD reports, we might
3450 * have roamed and need to update
3451 * multicast snooping switches
3453 ipv6_mc_up(idev);
3454 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3455 break;
3457 idev->if_flags |= IF_READY;
3460 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3461 dev->name);
3463 run_pending = 1;
3466 switch (dev->type) {
3467 #if IS_ENABLED(CONFIG_IPV6_SIT)
3468 case ARPHRD_SIT:
3469 addrconf_sit_config(dev);
3470 break;
3471 #endif
3472 #if IS_ENABLED(CONFIG_NET_IPGRE)
3473 case ARPHRD_IPGRE:
3474 addrconf_gre_config(dev);
3475 break;
3476 #endif
3477 case ARPHRD_LOOPBACK:
3478 init_loopback(dev);
3479 break;
3481 default:
3482 addrconf_dev_config(dev);
3483 break;
3486 if (!IS_ERR_OR_NULL(idev)) {
3487 if (run_pending)
3488 addrconf_dad_run(idev);
3490 /* Device has an address by now */
3491 rt6_sync_up(dev, RTNH_F_DEAD);
3494 * If the MTU changed during the interface down,
3495 * when the interface up, the changed MTU must be
3496 * reflected in the idev as well as routers.
3498 if (idev->cnf.mtu6 != dev->mtu &&
3499 dev->mtu >= IPV6_MIN_MTU) {
3500 rt6_mtu_change(dev, dev->mtu);
3501 idev->cnf.mtu6 = dev->mtu;
3503 idev->tstamp = jiffies;
3504 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3507 * If the changed mtu during down is lower than
3508 * IPV6_MIN_MTU stop IPv6 on this interface.
3510 if (dev->mtu < IPV6_MIN_MTU)
3511 addrconf_ifdown(dev, dev != net->loopback_dev);
3513 break;
3515 case NETDEV_DOWN:
3516 case NETDEV_UNREGISTER:
3518 * Remove all addresses from this interface.
3520 addrconf_ifdown(dev, event != NETDEV_DOWN);
3521 break;
3523 case NETDEV_CHANGENAME:
3524 if (idev) {
3525 snmp6_unregister_dev(idev);
3526 addrconf_sysctl_unregister(idev);
3527 err = addrconf_sysctl_register(idev);
3528 if (err)
3529 return notifier_from_errno(err);
3530 err = snmp6_register_dev(idev);
3531 if (err) {
3532 addrconf_sysctl_unregister(idev);
3533 return notifier_from_errno(err);
3536 break;
3538 case NETDEV_PRE_TYPE_CHANGE:
3539 case NETDEV_POST_TYPE_CHANGE:
3540 if (idev)
3541 addrconf_type_change(dev, event);
3542 break;
3544 case NETDEV_CHANGEUPPER:
3545 info = ptr;
3547 /* flush all routes if dev is linked to or unlinked from
3548 * an L3 master device (e.g., VRF)
3550 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3551 addrconf_ifdown(dev, 0);
3554 return NOTIFY_OK;
3558 * addrconf module should be notified of a device going up
3560 static struct notifier_block ipv6_dev_notf = {
3561 .notifier_call = addrconf_notify,
3562 .priority = ADDRCONF_NOTIFY_PRIORITY,
3565 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3567 struct inet6_dev *idev;
3568 ASSERT_RTNL();
3570 idev = __in6_dev_get(dev);
3572 if (event == NETDEV_POST_TYPE_CHANGE)
3573 ipv6_mc_remap(idev);
3574 else if (event == NETDEV_PRE_TYPE_CHANGE)
3575 ipv6_mc_unmap(idev);
3578 static bool addr_is_local(const struct in6_addr *addr)
3580 return ipv6_addr_type(addr) &
3581 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3584 static int addrconf_ifdown(struct net_device *dev, int how)
3586 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3587 struct net *net = dev_net(dev);
3588 struct inet6_dev *idev;
3589 struct inet6_ifaddr *ifa, *tmp;
3590 int _keep_addr;
3591 bool keep_addr;
3592 int state, i;
3594 ASSERT_RTNL();
3596 rt6_disable_ip(dev, event);
3598 idev = __in6_dev_get(dev);
3599 if (!idev)
3600 return -ENODEV;
3603 * Step 1: remove reference to ipv6 device from parent device.
3604 * Do not dev_put!
3606 if (how) {
3607 idev->dead = 1;
3609 /* protected by rtnl_lock */
3610 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3612 /* Step 1.5: remove snmp6 entry */
3613 snmp6_unregister_dev(idev);
3617 /* aggregate the system setting and interface setting */
3618 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3619 if (!_keep_addr)
3620 _keep_addr = idev->cnf.keep_addr_on_down;
3622 /* combine the user config with event to determine if permanent
3623 * addresses are to be removed from address hash table
3625 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3627 /* Step 2: clear hash table */
3628 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3629 struct hlist_head *h = &inet6_addr_lst[i];
3631 spin_lock_bh(&addrconf_hash_lock);
3632 restart:
3633 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3634 if (ifa->idev == idev) {
3635 addrconf_del_dad_work(ifa);
3636 /* combined flag + permanent flag decide if
3637 * address is retained on a down event
3639 if (!keep_addr ||
3640 !(ifa->flags & IFA_F_PERMANENT) ||
3641 addr_is_local(&ifa->addr)) {
3642 hlist_del_init_rcu(&ifa->addr_lst);
3643 goto restart;
3647 spin_unlock_bh(&addrconf_hash_lock);
3650 write_lock_bh(&idev->lock);
3652 addrconf_del_rs_timer(idev);
3654 /* Step 2: clear flags for stateless addrconf */
3655 if (!how)
3656 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3658 /* Step 3: clear tempaddr list */
3659 while (!list_empty(&idev->tempaddr_list)) {
3660 ifa = list_first_entry(&idev->tempaddr_list,
3661 struct inet6_ifaddr, tmp_list);
3662 list_del(&ifa->tmp_list);
3663 write_unlock_bh(&idev->lock);
3664 spin_lock_bh(&ifa->lock);
3666 if (ifa->ifpub) {
3667 in6_ifa_put(ifa->ifpub);
3668 ifa->ifpub = NULL;
3670 spin_unlock_bh(&ifa->lock);
3671 in6_ifa_put(ifa);
3672 write_lock_bh(&idev->lock);
3675 /* re-combine the user config with event to determine if permanent
3676 * addresses are to be removed from the interface list
3678 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3680 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3681 struct rt6_info *rt = NULL;
3682 bool keep;
3684 addrconf_del_dad_work(ifa);
3686 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3687 !addr_is_local(&ifa->addr);
3689 write_unlock_bh(&idev->lock);
3690 spin_lock_bh(&ifa->lock);
3692 if (keep) {
3693 /* set state to skip the notifier below */
3694 state = INET6_IFADDR_STATE_DEAD;
3695 ifa->state = INET6_IFADDR_STATE_PREDAD;
3696 if (!(ifa->flags & IFA_F_NODAD))
3697 ifa->flags |= IFA_F_TENTATIVE;
3699 rt = ifa->rt;
3700 ifa->rt = NULL;
3701 } else {
3702 state = ifa->state;
3703 ifa->state = INET6_IFADDR_STATE_DEAD;
3706 spin_unlock_bh(&ifa->lock);
3708 if (rt)
3709 ip6_del_rt(rt);
3711 if (state != INET6_IFADDR_STATE_DEAD) {
3712 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3713 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3714 } else {
3715 if (idev->cnf.forwarding)
3716 addrconf_leave_anycast(ifa);
3717 addrconf_leave_solict(ifa->idev, &ifa->addr);
3720 write_lock_bh(&idev->lock);
3721 if (!keep) {
3722 list_del_rcu(&ifa->if_list);
3723 in6_ifa_put(ifa);
3727 write_unlock_bh(&idev->lock);
3729 /* Step 5: Discard anycast and multicast list */
3730 if (how) {
3731 ipv6_ac_destroy_dev(idev);
3732 ipv6_mc_destroy_dev(idev);
3733 } else {
3734 ipv6_mc_down(idev);
3737 idev->tstamp = jiffies;
3739 /* Last: Shot the device (if unregistered) */
3740 if (how) {
3741 addrconf_sysctl_unregister(idev);
3742 neigh_parms_release(&nd_tbl, idev->nd_parms);
3743 neigh_ifdown(&nd_tbl, dev);
3744 in6_dev_put(idev);
3746 return 0;
3749 static void addrconf_rs_timer(struct timer_list *t)
3751 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3752 struct net_device *dev = idev->dev;
3753 struct in6_addr lladdr;
3755 write_lock(&idev->lock);
3756 if (idev->dead || !(idev->if_flags & IF_READY))
3757 goto out;
3759 if (!ipv6_accept_ra(idev))
3760 goto out;
3762 /* Announcement received after solicitation was sent */
3763 if (idev->if_flags & IF_RA_RCVD)
3764 goto out;
3766 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3767 write_unlock(&idev->lock);
3768 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3769 ndisc_send_rs(dev, &lladdr,
3770 &in6addr_linklocal_allrouters);
3771 else
3772 goto put;
3774 write_lock(&idev->lock);
3775 idev->rs_interval = rfc3315_s14_backoff_update(
3776 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3777 /* The wait after the last probe can be shorter */
3778 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3779 idev->cnf.rtr_solicits) ?
3780 idev->cnf.rtr_solicit_delay :
3781 idev->rs_interval);
3782 } else {
3784 * Note: we do not support deprecated "all on-link"
3785 * assumption any longer.
3787 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3790 out:
3791 write_unlock(&idev->lock);
3792 put:
3793 in6_dev_put(idev);
3797 * Duplicate Address Detection
3799 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3801 unsigned long rand_num;
3802 struct inet6_dev *idev = ifp->idev;
3803 u64 nonce;
3805 if (ifp->flags & IFA_F_OPTIMISTIC)
3806 rand_num = 0;
3807 else
3808 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3810 nonce = 0;
3811 if (idev->cnf.enhanced_dad ||
3812 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3814 get_random_bytes(&nonce, 6);
3815 while (nonce == 0);
3817 ifp->dad_nonce = nonce;
3818 ifp->dad_probes = idev->cnf.dad_transmits;
3819 addrconf_mod_dad_work(ifp, rand_num);
3822 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3824 struct inet6_dev *idev = ifp->idev;
3825 struct net_device *dev = idev->dev;
3826 bool bump_id, notify = false;
3828 addrconf_join_solict(dev, &ifp->addr);
3830 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3832 read_lock_bh(&idev->lock);
3833 spin_lock(&ifp->lock);
3834 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3835 goto out;
3837 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3838 (dev_net(dev)->ipv6.devconf_all->accept_dad < 1 &&
3839 idev->cnf.accept_dad < 1) ||
3840 !(ifp->flags&IFA_F_TENTATIVE) ||
3841 ifp->flags & IFA_F_NODAD) {
3842 bool send_na = false;
3844 if (ifp->flags & IFA_F_TENTATIVE &&
3845 !(ifp->flags & IFA_F_OPTIMISTIC))
3846 send_na = true;
3847 bump_id = ifp->flags & IFA_F_TENTATIVE;
3848 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3849 spin_unlock(&ifp->lock);
3850 read_unlock_bh(&idev->lock);
3852 addrconf_dad_completed(ifp, bump_id, send_na);
3853 return;
3856 if (!(idev->if_flags & IF_READY)) {
3857 spin_unlock(&ifp->lock);
3858 read_unlock_bh(&idev->lock);
3860 * If the device is not ready:
3861 * - keep it tentative if it is a permanent address.
3862 * - otherwise, kill it.
3864 in6_ifa_hold(ifp);
3865 addrconf_dad_stop(ifp, 0);
3866 return;
3870 * Optimistic nodes can start receiving
3871 * Frames right away
3873 if (ifp->flags & IFA_F_OPTIMISTIC) {
3874 ip6_ins_rt(ifp->rt);
3875 if (ipv6_use_optimistic_addr(dev_net(dev), idev)) {
3876 /* Because optimistic nodes can use this address,
3877 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3879 notify = true;
3883 addrconf_dad_kick(ifp);
3884 out:
3885 spin_unlock(&ifp->lock);
3886 read_unlock_bh(&idev->lock);
3887 if (notify)
3888 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3891 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3893 bool begin_dad = false;
3895 spin_lock_bh(&ifp->lock);
3896 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3897 ifp->state = INET6_IFADDR_STATE_PREDAD;
3898 begin_dad = true;
3900 spin_unlock_bh(&ifp->lock);
3902 if (begin_dad)
3903 addrconf_mod_dad_work(ifp, 0);
3906 static void addrconf_dad_work(struct work_struct *w)
3908 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3909 struct inet6_ifaddr,
3910 dad_work);
3911 struct inet6_dev *idev = ifp->idev;
3912 bool bump_id, disable_ipv6 = false;
3913 struct in6_addr mcaddr;
3915 enum {
3916 DAD_PROCESS,
3917 DAD_BEGIN,
3918 DAD_ABORT,
3919 } action = DAD_PROCESS;
3921 rtnl_lock();
3923 spin_lock_bh(&ifp->lock);
3924 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3925 action = DAD_BEGIN;
3926 ifp->state = INET6_IFADDR_STATE_DAD;
3927 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3928 action = DAD_ABORT;
3929 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3931 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
3932 idev->cnf.accept_dad > 1) &&
3933 !idev->cnf.disable_ipv6 &&
3934 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3935 struct in6_addr addr;
3937 addr.s6_addr32[0] = htonl(0xfe800000);
3938 addr.s6_addr32[1] = 0;
3940 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3941 ipv6_addr_equal(&ifp->addr, &addr)) {
3942 /* DAD failed for link-local based on MAC */
3943 idev->cnf.disable_ipv6 = 1;
3945 pr_info("%s: IPv6 being disabled!\n",
3946 ifp->idev->dev->name);
3947 disable_ipv6 = true;
3951 spin_unlock_bh(&ifp->lock);
3953 if (action == DAD_BEGIN) {
3954 addrconf_dad_begin(ifp);
3955 goto out;
3956 } else if (action == DAD_ABORT) {
3957 in6_ifa_hold(ifp);
3958 addrconf_dad_stop(ifp, 1);
3959 if (disable_ipv6)
3960 addrconf_ifdown(idev->dev, 0);
3961 goto out;
3964 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3965 goto out;
3967 write_lock_bh(&idev->lock);
3968 if (idev->dead || !(idev->if_flags & IF_READY)) {
3969 write_unlock_bh(&idev->lock);
3970 goto out;
3973 spin_lock(&ifp->lock);
3974 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3975 spin_unlock(&ifp->lock);
3976 write_unlock_bh(&idev->lock);
3977 goto out;
3980 if (ifp->dad_probes == 0) {
3981 bool send_na = false;
3984 * DAD was successful
3987 if (ifp->flags & IFA_F_TENTATIVE &&
3988 !(ifp->flags & IFA_F_OPTIMISTIC))
3989 send_na = true;
3990 bump_id = ifp->flags & IFA_F_TENTATIVE;
3991 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3992 spin_unlock(&ifp->lock);
3993 write_unlock_bh(&idev->lock);
3995 addrconf_dad_completed(ifp, bump_id, send_na);
3997 goto out;
4000 ifp->dad_probes--;
4001 addrconf_mod_dad_work(ifp,
4002 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
4003 spin_unlock(&ifp->lock);
4004 write_unlock_bh(&idev->lock);
4006 /* send a neighbour solicitation for our addr */
4007 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4008 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4009 ifp->dad_nonce);
4010 out:
4011 in6_ifa_put(ifp);
4012 rtnl_unlock();
4015 /* ifp->idev must be at least read locked */
4016 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4018 struct inet6_ifaddr *ifpiter;
4019 struct inet6_dev *idev = ifp->idev;
4021 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4022 if (ifpiter->scope > IFA_LINK)
4023 break;
4024 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4025 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4026 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4027 IFA_F_PERMANENT)
4028 return false;
4030 return true;
4033 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4034 bool send_na)
4036 struct net_device *dev = ifp->idev->dev;
4037 struct in6_addr lladdr;
4038 bool send_rs, send_mld;
4040 addrconf_del_dad_work(ifp);
4043 * Configure the address for reception. Now it is valid.
4046 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4048 /* If added prefix is link local and we are prepared to process
4049 router advertisements, start sending router solicitations.
4052 read_lock_bh(&ifp->idev->lock);
4053 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4054 send_rs = send_mld &&
4055 ipv6_accept_ra(ifp->idev) &&
4056 ifp->idev->cnf.rtr_solicits != 0 &&
4057 (dev->flags&IFF_LOOPBACK) == 0;
4058 read_unlock_bh(&ifp->idev->lock);
4060 /* While dad is in progress mld report's source address is in6_addrany.
4061 * Resend with proper ll now.
4063 if (send_mld)
4064 ipv6_mc_dad_complete(ifp->idev);
4066 /* send unsolicited NA if enabled */
4067 if (send_na &&
4068 (ifp->idev->cnf.ndisc_notify ||
4069 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4070 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4071 /*router=*/ !!ifp->idev->cnf.forwarding,
4072 /*solicited=*/ false, /*override=*/ true,
4073 /*inc_opt=*/ true);
4076 if (send_rs) {
4078 * If a host as already performed a random delay
4079 * [...] as part of DAD [...] there is no need
4080 * to delay again before sending the first RS
4082 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4083 return;
4084 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4086 write_lock_bh(&ifp->idev->lock);
4087 spin_lock(&ifp->lock);
4088 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4089 ifp->idev->cnf.rtr_solicit_interval);
4090 ifp->idev->rs_probes = 1;
4091 ifp->idev->if_flags |= IF_RS_SENT;
4092 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4093 spin_unlock(&ifp->lock);
4094 write_unlock_bh(&ifp->idev->lock);
4097 if (bump_id)
4098 rt_genid_bump_ipv6(dev_net(dev));
4100 /* Make sure that a new temporary address will be created
4101 * before this temporary address becomes deprecated.
4103 if (ifp->flags & IFA_F_TEMPORARY)
4104 addrconf_verify_rtnl();
4107 static void addrconf_dad_run(struct inet6_dev *idev)
4109 struct inet6_ifaddr *ifp;
4111 read_lock_bh(&idev->lock);
4112 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4113 spin_lock(&ifp->lock);
4114 if (ifp->flags & IFA_F_TENTATIVE &&
4115 ifp->state == INET6_IFADDR_STATE_DAD)
4116 addrconf_dad_kick(ifp);
4117 spin_unlock(&ifp->lock);
4119 read_unlock_bh(&idev->lock);
4122 #ifdef CONFIG_PROC_FS
4123 struct if6_iter_state {
4124 struct seq_net_private p;
4125 int bucket;
4126 int offset;
4129 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4131 struct if6_iter_state *state = seq->private;
4132 struct net *net = seq_file_net(seq);
4133 struct inet6_ifaddr *ifa = NULL;
4134 int p = 0;
4136 /* initial bucket if pos is 0 */
4137 if (pos == 0) {
4138 state->bucket = 0;
4139 state->offset = 0;
4142 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4143 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4144 addr_lst) {
4145 if (!net_eq(dev_net(ifa->idev->dev), net))
4146 continue;
4147 /* sync with offset */
4148 if (p < state->offset) {
4149 p++;
4150 continue;
4152 state->offset++;
4153 return ifa;
4156 /* prepare for next bucket */
4157 state->offset = 0;
4158 p = 0;
4160 return NULL;
4163 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4164 struct inet6_ifaddr *ifa)
4166 struct if6_iter_state *state = seq->private;
4167 struct net *net = seq_file_net(seq);
4169 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4170 if (!net_eq(dev_net(ifa->idev->dev), net))
4171 continue;
4172 state->offset++;
4173 return ifa;
4176 while (++state->bucket < IN6_ADDR_HSIZE) {
4177 state->offset = 0;
4178 hlist_for_each_entry_rcu(ifa,
4179 &inet6_addr_lst[state->bucket], addr_lst) {
4180 if (!net_eq(dev_net(ifa->idev->dev), net))
4181 continue;
4182 state->offset++;
4183 return ifa;
4187 return NULL;
4190 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4191 __acquires(rcu)
4193 rcu_read_lock();
4194 return if6_get_first(seq, *pos);
4197 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4199 struct inet6_ifaddr *ifa;
4201 ifa = if6_get_next(seq, v);
4202 ++*pos;
4203 return ifa;
4206 static void if6_seq_stop(struct seq_file *seq, void *v)
4207 __releases(rcu)
4209 rcu_read_unlock();
4212 static int if6_seq_show(struct seq_file *seq, void *v)
4214 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4215 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4216 &ifp->addr,
4217 ifp->idev->dev->ifindex,
4218 ifp->prefix_len,
4219 ifp->scope,
4220 (u8) ifp->flags,
4221 ifp->idev->dev->name);
4222 return 0;
4225 static const struct seq_operations if6_seq_ops = {
4226 .start = if6_seq_start,
4227 .next = if6_seq_next,
4228 .show = if6_seq_show,
4229 .stop = if6_seq_stop,
4232 static int if6_seq_open(struct inode *inode, struct file *file)
4234 return seq_open_net(inode, file, &if6_seq_ops,
4235 sizeof(struct if6_iter_state));
4238 static const struct file_operations if6_fops = {
4239 .open = if6_seq_open,
4240 .read = seq_read,
4241 .llseek = seq_lseek,
4242 .release = seq_release_net,
4245 static int __net_init if6_proc_net_init(struct net *net)
4247 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4248 return -ENOMEM;
4249 return 0;
4252 static void __net_exit if6_proc_net_exit(struct net *net)
4254 remove_proc_entry("if_inet6", net->proc_net);
4257 static struct pernet_operations if6_proc_net_ops = {
4258 .init = if6_proc_net_init,
4259 .exit = if6_proc_net_exit,
4262 int __init if6_proc_init(void)
4264 return register_pernet_subsys(&if6_proc_net_ops);
4267 void if6_proc_exit(void)
4269 unregister_pernet_subsys(&if6_proc_net_ops);
4271 #endif /* CONFIG_PROC_FS */
4273 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4274 /* Check if address is a home address configured on any interface. */
4275 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4277 unsigned int hash = inet6_addr_hash(net, addr);
4278 struct inet6_ifaddr *ifp = NULL;
4279 int ret = 0;
4281 rcu_read_lock();
4282 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4283 if (!net_eq(dev_net(ifp->idev->dev), net))
4284 continue;
4285 if (ipv6_addr_equal(&ifp->addr, addr) &&
4286 (ifp->flags & IFA_F_HOMEADDRESS)) {
4287 ret = 1;
4288 break;
4291 rcu_read_unlock();
4292 return ret;
4294 #endif
4297 * Periodic address status verification
4300 static void addrconf_verify_rtnl(void)
4302 unsigned long now, next, next_sec, next_sched;
4303 struct inet6_ifaddr *ifp;
4304 int i;
4306 ASSERT_RTNL();
4308 rcu_read_lock_bh();
4309 now = jiffies;
4310 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4312 cancel_delayed_work(&addr_chk_work);
4314 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4315 restart:
4316 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4317 unsigned long age;
4319 /* When setting preferred_lft to a value not zero or
4320 * infinity, while valid_lft is infinity
4321 * IFA_F_PERMANENT has a non-infinity life time.
4323 if ((ifp->flags & IFA_F_PERMANENT) &&
4324 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4325 continue;
4327 spin_lock(&ifp->lock);
4328 /* We try to batch several events at once. */
4329 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4331 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4332 age >= ifp->valid_lft) {
4333 spin_unlock(&ifp->lock);
4334 in6_ifa_hold(ifp);
4335 ipv6_del_addr(ifp);
4336 goto restart;
4337 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4338 spin_unlock(&ifp->lock);
4339 continue;
4340 } else if (age >= ifp->prefered_lft) {
4341 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4342 int deprecate = 0;
4344 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4345 deprecate = 1;
4346 ifp->flags |= IFA_F_DEPRECATED;
4349 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4350 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4351 next = ifp->tstamp + ifp->valid_lft * HZ;
4353 spin_unlock(&ifp->lock);
4355 if (deprecate) {
4356 in6_ifa_hold(ifp);
4358 ipv6_ifa_notify(0, ifp);
4359 in6_ifa_put(ifp);
4360 goto restart;
4362 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4363 !(ifp->flags&IFA_F_TENTATIVE)) {
4364 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4365 ifp->idev->cnf.dad_transmits *
4366 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4368 if (age >= ifp->prefered_lft - regen_advance) {
4369 struct inet6_ifaddr *ifpub = ifp->ifpub;
4370 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4371 next = ifp->tstamp + ifp->prefered_lft * HZ;
4372 if (!ifp->regen_count && ifpub) {
4373 ifp->regen_count++;
4374 in6_ifa_hold(ifp);
4375 in6_ifa_hold(ifpub);
4376 spin_unlock(&ifp->lock);
4378 spin_lock(&ifpub->lock);
4379 ifpub->regen_count = 0;
4380 spin_unlock(&ifpub->lock);
4381 rcu_read_unlock_bh();
4382 ipv6_create_tempaddr(ifpub, ifp, true);
4383 in6_ifa_put(ifpub);
4384 in6_ifa_put(ifp);
4385 rcu_read_lock_bh();
4386 goto restart;
4388 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4389 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4390 spin_unlock(&ifp->lock);
4391 } else {
4392 /* ifp->prefered_lft <= ifp->valid_lft */
4393 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4394 next = ifp->tstamp + ifp->prefered_lft * HZ;
4395 spin_unlock(&ifp->lock);
4400 next_sec = round_jiffies_up(next);
4401 next_sched = next;
4403 /* If rounded timeout is accurate enough, accept it. */
4404 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4405 next_sched = next_sec;
4407 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4408 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4409 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4411 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4412 now, next, next_sec, next_sched);
4413 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4414 rcu_read_unlock_bh();
4417 static void addrconf_verify_work(struct work_struct *w)
4419 rtnl_lock();
4420 addrconf_verify_rtnl();
4421 rtnl_unlock();
4424 static void addrconf_verify(void)
4426 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4429 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4430 struct in6_addr **peer_pfx)
4432 struct in6_addr *pfx = NULL;
4434 *peer_pfx = NULL;
4436 if (addr)
4437 pfx = nla_data(addr);
4439 if (local) {
4440 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4441 *peer_pfx = pfx;
4442 pfx = nla_data(local);
4445 return pfx;
4448 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4449 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4450 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4451 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4452 [IFA_FLAGS] = { .len = sizeof(u32) },
4455 static int
4456 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4457 struct netlink_ext_ack *extack)
4459 struct net *net = sock_net(skb->sk);
4460 struct ifaddrmsg *ifm;
4461 struct nlattr *tb[IFA_MAX+1];
4462 struct in6_addr *pfx, *peer_pfx;
4463 u32 ifa_flags;
4464 int err;
4466 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4467 extack);
4468 if (err < 0)
4469 return err;
4471 ifm = nlmsg_data(nlh);
4472 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4473 if (!pfx)
4474 return -EINVAL;
4476 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4478 /* We ignore other flags so far. */
4479 ifa_flags &= IFA_F_MANAGETEMPADDR;
4481 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4482 ifm->ifa_prefixlen);
4485 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4486 u32 prefered_lft, u32 valid_lft)
4488 u32 flags;
4489 clock_t expires;
4490 unsigned long timeout;
4491 bool was_managetempaddr;
4492 bool had_prefixroute;
4494 ASSERT_RTNL();
4496 if (!valid_lft || (prefered_lft > valid_lft))
4497 return -EINVAL;
4499 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4500 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4501 return -EINVAL;
4503 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4504 if (addrconf_finite_timeout(timeout)) {
4505 expires = jiffies_to_clock_t(timeout * HZ);
4506 valid_lft = timeout;
4507 flags = RTF_EXPIRES;
4508 } else {
4509 expires = 0;
4510 flags = 0;
4511 ifa_flags |= IFA_F_PERMANENT;
4514 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4515 if (addrconf_finite_timeout(timeout)) {
4516 if (timeout == 0)
4517 ifa_flags |= IFA_F_DEPRECATED;
4518 prefered_lft = timeout;
4521 spin_lock_bh(&ifp->lock);
4522 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4523 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4524 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4525 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4526 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4527 IFA_F_NOPREFIXROUTE);
4528 ifp->flags |= ifa_flags;
4529 ifp->tstamp = jiffies;
4530 ifp->valid_lft = valid_lft;
4531 ifp->prefered_lft = prefered_lft;
4533 spin_unlock_bh(&ifp->lock);
4534 if (!(ifp->flags&IFA_F_TENTATIVE))
4535 ipv6_ifa_notify(0, ifp);
4537 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4538 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4539 expires, flags);
4540 } else if (had_prefixroute) {
4541 enum cleanup_prefix_rt_t action;
4542 unsigned long rt_expires;
4544 write_lock_bh(&ifp->idev->lock);
4545 action = check_cleanup_prefix_route(ifp, &rt_expires);
4546 write_unlock_bh(&ifp->idev->lock);
4548 if (action != CLEANUP_PREFIX_RT_NOP) {
4549 cleanup_prefix_route(ifp, rt_expires,
4550 action == CLEANUP_PREFIX_RT_DEL);
4554 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4555 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4556 valid_lft = prefered_lft = 0;
4557 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4558 !was_managetempaddr, jiffies);
4561 addrconf_verify_rtnl();
4563 return 0;
4566 static int
4567 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4568 struct netlink_ext_ack *extack)
4570 struct net *net = sock_net(skb->sk);
4571 struct ifaddrmsg *ifm;
4572 struct nlattr *tb[IFA_MAX+1];
4573 struct in6_addr *pfx, *peer_pfx;
4574 struct inet6_ifaddr *ifa;
4575 struct net_device *dev;
4576 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4577 u32 ifa_flags;
4578 int err;
4580 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4581 extack);
4582 if (err < 0)
4583 return err;
4585 ifm = nlmsg_data(nlh);
4586 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4587 if (!pfx)
4588 return -EINVAL;
4590 if (tb[IFA_CACHEINFO]) {
4591 struct ifa_cacheinfo *ci;
4593 ci = nla_data(tb[IFA_CACHEINFO]);
4594 valid_lft = ci->ifa_valid;
4595 preferred_lft = ci->ifa_prefered;
4596 } else {
4597 preferred_lft = INFINITY_LIFE_TIME;
4598 valid_lft = INFINITY_LIFE_TIME;
4601 dev = __dev_get_by_index(net, ifm->ifa_index);
4602 if (!dev)
4603 return -ENODEV;
4605 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4607 /* We ignore other flags so far. */
4608 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4609 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4611 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4612 if (!ifa) {
4614 * It would be best to check for !NLM_F_CREATE here but
4615 * userspace already relies on not having to provide this.
4617 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4618 ifm->ifa_prefixlen, ifa_flags,
4619 preferred_lft, valid_lft, extack);
4622 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4623 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4624 err = -EEXIST;
4625 else
4626 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4628 in6_ifa_put(ifa);
4630 return err;
4633 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4634 u8 scope, int ifindex)
4636 struct ifaddrmsg *ifm;
4638 ifm = nlmsg_data(nlh);
4639 ifm->ifa_family = AF_INET6;
4640 ifm->ifa_prefixlen = prefixlen;
4641 ifm->ifa_flags = flags;
4642 ifm->ifa_scope = scope;
4643 ifm->ifa_index = ifindex;
4646 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4647 unsigned long tstamp, u32 preferred, u32 valid)
4649 struct ifa_cacheinfo ci;
4651 ci.cstamp = cstamp_delta(cstamp);
4652 ci.tstamp = cstamp_delta(tstamp);
4653 ci.ifa_prefered = preferred;
4654 ci.ifa_valid = valid;
4656 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4659 static inline int rt_scope(int ifa_scope)
4661 if (ifa_scope & IFA_HOST)
4662 return RT_SCOPE_HOST;
4663 else if (ifa_scope & IFA_LINK)
4664 return RT_SCOPE_LINK;
4665 else if (ifa_scope & IFA_SITE)
4666 return RT_SCOPE_SITE;
4667 else
4668 return RT_SCOPE_UNIVERSE;
4671 static inline int inet6_ifaddr_msgsize(void)
4673 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4674 + nla_total_size(16) /* IFA_LOCAL */
4675 + nla_total_size(16) /* IFA_ADDRESS */
4676 + nla_total_size(sizeof(struct ifa_cacheinfo))
4677 + nla_total_size(4) /* IFA_FLAGS */;
4680 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4681 u32 portid, u32 seq, int event, unsigned int flags)
4683 struct nlmsghdr *nlh;
4684 u32 preferred, valid;
4686 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4687 if (!nlh)
4688 return -EMSGSIZE;
4690 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4691 ifa->idev->dev->ifindex);
4693 if (!((ifa->flags&IFA_F_PERMANENT) &&
4694 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4695 preferred = ifa->prefered_lft;
4696 valid = ifa->valid_lft;
4697 if (preferred != INFINITY_LIFE_TIME) {
4698 long tval = (jiffies - ifa->tstamp)/HZ;
4699 if (preferred > tval)
4700 preferred -= tval;
4701 else
4702 preferred = 0;
4703 if (valid != INFINITY_LIFE_TIME) {
4704 if (valid > tval)
4705 valid -= tval;
4706 else
4707 valid = 0;
4710 } else {
4711 preferred = INFINITY_LIFE_TIME;
4712 valid = INFINITY_LIFE_TIME;
4715 if (!ipv6_addr_any(&ifa->peer_addr)) {
4716 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4717 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4718 goto error;
4719 } else
4720 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4721 goto error;
4723 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4724 goto error;
4726 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4727 goto error;
4729 nlmsg_end(skb, nlh);
4730 return 0;
4732 error:
4733 nlmsg_cancel(skb, nlh);
4734 return -EMSGSIZE;
4737 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4738 u32 portid, u32 seq, int event, u16 flags)
4740 struct nlmsghdr *nlh;
4741 u8 scope = RT_SCOPE_UNIVERSE;
4742 int ifindex = ifmca->idev->dev->ifindex;
4744 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4745 scope = RT_SCOPE_SITE;
4747 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4748 if (!nlh)
4749 return -EMSGSIZE;
4751 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4752 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4753 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4754 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4755 nlmsg_cancel(skb, nlh);
4756 return -EMSGSIZE;
4759 nlmsg_end(skb, nlh);
4760 return 0;
4763 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4764 u32 portid, u32 seq, int event, unsigned int flags)
4766 struct nlmsghdr *nlh;
4767 u8 scope = RT_SCOPE_UNIVERSE;
4768 int ifindex = ifaca->aca_idev->dev->ifindex;
4770 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4771 scope = RT_SCOPE_SITE;
4773 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4774 if (!nlh)
4775 return -EMSGSIZE;
4777 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4778 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4779 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4780 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4781 nlmsg_cancel(skb, nlh);
4782 return -EMSGSIZE;
4785 nlmsg_end(skb, nlh);
4786 return 0;
4789 enum addr_type_t {
4790 UNICAST_ADDR,
4791 MULTICAST_ADDR,
4792 ANYCAST_ADDR,
4795 /* called with rcu_read_lock() */
4796 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4797 struct netlink_callback *cb, enum addr_type_t type,
4798 int s_ip_idx, int *p_ip_idx)
4800 struct ifmcaddr6 *ifmca;
4801 struct ifacaddr6 *ifaca;
4802 int err = 1;
4803 int ip_idx = *p_ip_idx;
4805 read_lock_bh(&idev->lock);
4806 switch (type) {
4807 case UNICAST_ADDR: {
4808 struct inet6_ifaddr *ifa;
4810 /* unicast address incl. temp addr */
4811 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4812 if (++ip_idx < s_ip_idx)
4813 continue;
4814 err = inet6_fill_ifaddr(skb, ifa,
4815 NETLINK_CB(cb->skb).portid,
4816 cb->nlh->nlmsg_seq,
4817 RTM_NEWADDR,
4818 NLM_F_MULTI);
4819 if (err < 0)
4820 break;
4821 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4823 break;
4825 case MULTICAST_ADDR:
4826 /* multicast address */
4827 for (ifmca = idev->mc_list; ifmca;
4828 ifmca = ifmca->next, ip_idx++) {
4829 if (ip_idx < s_ip_idx)
4830 continue;
4831 err = inet6_fill_ifmcaddr(skb, ifmca,
4832 NETLINK_CB(cb->skb).portid,
4833 cb->nlh->nlmsg_seq,
4834 RTM_GETMULTICAST,
4835 NLM_F_MULTI);
4836 if (err < 0)
4837 break;
4839 break;
4840 case ANYCAST_ADDR:
4841 /* anycast address */
4842 for (ifaca = idev->ac_list; ifaca;
4843 ifaca = ifaca->aca_next, ip_idx++) {
4844 if (ip_idx < s_ip_idx)
4845 continue;
4846 err = inet6_fill_ifacaddr(skb, ifaca,
4847 NETLINK_CB(cb->skb).portid,
4848 cb->nlh->nlmsg_seq,
4849 RTM_GETANYCAST,
4850 NLM_F_MULTI);
4851 if (err < 0)
4852 break;
4854 break;
4855 default:
4856 break;
4858 read_unlock_bh(&idev->lock);
4859 *p_ip_idx = ip_idx;
4860 return err;
4863 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4864 enum addr_type_t type)
4866 struct net *net = sock_net(skb->sk);
4867 int h, s_h;
4868 int idx, ip_idx;
4869 int s_idx, s_ip_idx;
4870 struct net_device *dev;
4871 struct inet6_dev *idev;
4872 struct hlist_head *head;
4874 s_h = cb->args[0];
4875 s_idx = idx = cb->args[1];
4876 s_ip_idx = ip_idx = cb->args[2];
4878 rcu_read_lock();
4879 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4880 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4881 idx = 0;
4882 head = &net->dev_index_head[h];
4883 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4884 if (idx < s_idx)
4885 goto cont;
4886 if (h > s_h || idx > s_idx)
4887 s_ip_idx = 0;
4888 ip_idx = 0;
4889 idev = __in6_dev_get(dev);
4890 if (!idev)
4891 goto cont;
4893 if (in6_dump_addrs(idev, skb, cb, type,
4894 s_ip_idx, &ip_idx) < 0)
4895 goto done;
4896 cont:
4897 idx++;
4900 done:
4901 rcu_read_unlock();
4902 cb->args[0] = h;
4903 cb->args[1] = idx;
4904 cb->args[2] = ip_idx;
4906 return skb->len;
4909 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4911 enum addr_type_t type = UNICAST_ADDR;
4913 return inet6_dump_addr(skb, cb, type);
4916 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4918 enum addr_type_t type = MULTICAST_ADDR;
4920 return inet6_dump_addr(skb, cb, type);
4924 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4926 enum addr_type_t type = ANYCAST_ADDR;
4928 return inet6_dump_addr(skb, cb, type);
4931 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4932 struct netlink_ext_ack *extack)
4934 struct net *net = sock_net(in_skb->sk);
4935 struct ifaddrmsg *ifm;
4936 struct nlattr *tb[IFA_MAX+1];
4937 struct in6_addr *addr = NULL, *peer;
4938 struct net_device *dev = NULL;
4939 struct inet6_ifaddr *ifa;
4940 struct sk_buff *skb;
4941 int err;
4943 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4944 extack);
4945 if (err < 0)
4946 return err;
4948 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4949 if (!addr)
4950 return -EINVAL;
4952 ifm = nlmsg_data(nlh);
4953 if (ifm->ifa_index)
4954 dev = dev_get_by_index(net, ifm->ifa_index);
4956 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4957 if (!ifa) {
4958 err = -EADDRNOTAVAIL;
4959 goto errout;
4962 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4963 if (!skb) {
4964 err = -ENOBUFS;
4965 goto errout_ifa;
4968 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4969 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4970 if (err < 0) {
4971 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4972 WARN_ON(err == -EMSGSIZE);
4973 kfree_skb(skb);
4974 goto errout_ifa;
4976 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4977 errout_ifa:
4978 in6_ifa_put(ifa);
4979 errout:
4980 if (dev)
4981 dev_put(dev);
4982 return err;
4985 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4987 struct sk_buff *skb;
4988 struct net *net = dev_net(ifa->idev->dev);
4989 int err = -ENOBUFS;
4991 /* Don't send DELADDR notification for TENTATIVE address,
4992 * since NEWADDR notification is sent only after removing
4993 * TENTATIVE flag, if DAD has not failed.
4995 if (ifa->flags & IFA_F_TENTATIVE && !(ifa->flags & IFA_F_DADFAILED) &&
4996 event == RTM_DELADDR)
4997 return;
4999 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5000 if (!skb)
5001 goto errout;
5003 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
5004 if (err < 0) {
5005 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5006 WARN_ON(err == -EMSGSIZE);
5007 kfree_skb(skb);
5008 goto errout;
5010 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5011 return;
5012 errout:
5013 if (err < 0)
5014 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5017 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5018 __s32 *array, int bytes)
5020 BUG_ON(bytes < (DEVCONF_MAX * 4));
5022 memset(array, 0, bytes);
5023 array[DEVCONF_FORWARDING] = cnf->forwarding;
5024 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5025 array[DEVCONF_MTU6] = cnf->mtu6;
5026 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5027 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5028 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5029 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5030 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5031 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5032 jiffies_to_msecs(cnf->rtr_solicit_interval);
5033 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5034 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5035 array[DEVCONF_RTR_SOLICIT_DELAY] =
5036 jiffies_to_msecs(cnf->rtr_solicit_delay);
5037 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5038 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5039 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5040 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5041 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5042 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5043 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5044 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5045 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5046 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5047 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5048 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5049 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5050 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5051 #ifdef CONFIG_IPV6_ROUTER_PREF
5052 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5053 array[DEVCONF_RTR_PROBE_INTERVAL] =
5054 jiffies_to_msecs(cnf->rtr_probe_interval);
5055 #ifdef CONFIG_IPV6_ROUTE_INFO
5056 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5057 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5058 #endif
5059 #endif
5060 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5061 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5062 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5063 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5064 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5065 #endif
5066 #ifdef CONFIG_IPV6_MROUTE
5067 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5068 #endif
5069 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5070 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5071 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5072 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5073 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5074 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5075 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5076 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5077 /* we omit DEVCONF_STABLE_SECRET for now */
5078 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5079 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5080 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5081 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5082 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5083 #ifdef CONFIG_IPV6_SEG6_HMAC
5084 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5085 #endif
5086 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5087 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5088 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5089 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5092 static inline size_t inet6_ifla6_size(void)
5094 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5095 + nla_total_size(sizeof(struct ifla_cacheinfo))
5096 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5097 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5098 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5099 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5102 static inline size_t inet6_if_nlmsg_size(void)
5104 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5105 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5106 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5107 + nla_total_size(4) /* IFLA_MTU */
5108 + nla_total_size(4) /* IFLA_LINK */
5109 + nla_total_size(1) /* IFLA_OPERSTATE */
5110 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5113 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5114 int bytes)
5116 int i;
5117 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5118 BUG_ON(pad < 0);
5120 /* Use put_unaligned() because stats may not be aligned for u64. */
5121 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5122 for (i = 1; i < ICMP6_MIB_MAX; i++)
5123 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5125 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5128 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5129 int bytes, size_t syncpoff)
5131 int i, c;
5132 u64 buff[IPSTATS_MIB_MAX];
5133 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5135 BUG_ON(pad < 0);
5137 memset(buff, 0, sizeof(buff));
5138 buff[0] = IPSTATS_MIB_MAX;
5140 for_each_possible_cpu(c) {
5141 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5142 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5145 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5146 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5149 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5150 int bytes)
5152 switch (attrtype) {
5153 case IFLA_INET6_STATS:
5154 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5155 offsetof(struct ipstats_mib, syncp));
5156 break;
5157 case IFLA_INET6_ICMP6STATS:
5158 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5159 break;
5163 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5164 u32 ext_filter_mask)
5166 struct nlattr *nla;
5167 struct ifla_cacheinfo ci;
5169 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5170 goto nla_put_failure;
5171 ci.max_reasm_len = IPV6_MAXPLEN;
5172 ci.tstamp = cstamp_delta(idev->tstamp);
5173 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5174 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5175 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5176 goto nla_put_failure;
5177 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5178 if (!nla)
5179 goto nla_put_failure;
5180 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5182 /* XXX - MC not implemented */
5184 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5185 return 0;
5187 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5188 if (!nla)
5189 goto nla_put_failure;
5190 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5192 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5193 if (!nla)
5194 goto nla_put_failure;
5195 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5197 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5198 if (!nla)
5199 goto nla_put_failure;
5201 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5202 goto nla_put_failure;
5204 read_lock_bh(&idev->lock);
5205 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5206 read_unlock_bh(&idev->lock);
5208 return 0;
5210 nla_put_failure:
5211 return -EMSGSIZE;
5214 static size_t inet6_get_link_af_size(const struct net_device *dev,
5215 u32 ext_filter_mask)
5217 if (!__in6_dev_get(dev))
5218 return 0;
5220 return inet6_ifla6_size();
5223 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5224 u32 ext_filter_mask)
5226 struct inet6_dev *idev = __in6_dev_get(dev);
5228 if (!idev)
5229 return -ENODATA;
5231 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5232 return -EMSGSIZE;
5234 return 0;
5237 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5239 struct inet6_ifaddr *ifp;
5240 struct net_device *dev = idev->dev;
5241 bool clear_token, update_rs = false;
5242 struct in6_addr ll_addr;
5244 ASSERT_RTNL();
5246 if (!token)
5247 return -EINVAL;
5248 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5249 return -EINVAL;
5250 if (!ipv6_accept_ra(idev))
5251 return -EINVAL;
5252 if (idev->cnf.rtr_solicits == 0)
5253 return -EINVAL;
5255 write_lock_bh(&idev->lock);
5257 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5258 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5260 write_unlock_bh(&idev->lock);
5262 clear_token = ipv6_addr_any(token);
5263 if (clear_token)
5264 goto update_lft;
5266 if (!idev->dead && (idev->if_flags & IF_READY) &&
5267 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5268 IFA_F_OPTIMISTIC)) {
5269 /* If we're not ready, then normal ifup will take care
5270 * of this. Otherwise, we need to request our rs here.
5272 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5273 update_rs = true;
5276 update_lft:
5277 write_lock_bh(&idev->lock);
5279 if (update_rs) {
5280 idev->if_flags |= IF_RS_SENT;
5281 idev->rs_interval = rfc3315_s14_backoff_init(
5282 idev->cnf.rtr_solicit_interval);
5283 idev->rs_probes = 1;
5284 addrconf_mod_rs_timer(idev, idev->rs_interval);
5287 /* Well, that's kinda nasty ... */
5288 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5289 spin_lock(&ifp->lock);
5290 if (ifp->tokenized) {
5291 ifp->valid_lft = 0;
5292 ifp->prefered_lft = 0;
5294 spin_unlock(&ifp->lock);
5297 write_unlock_bh(&idev->lock);
5298 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5299 addrconf_verify_rtnl();
5300 return 0;
5303 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5304 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5305 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5308 static int inet6_validate_link_af(const struct net_device *dev,
5309 const struct nlattr *nla)
5311 struct nlattr *tb[IFLA_INET6_MAX + 1];
5313 if (dev && !__in6_dev_get(dev))
5314 return -EAFNOSUPPORT;
5316 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5317 NULL);
5320 static int check_addr_gen_mode(int mode)
5322 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5323 mode != IN6_ADDR_GEN_MODE_NONE &&
5324 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5325 mode != IN6_ADDR_GEN_MODE_RANDOM)
5326 return -EINVAL;
5327 return 1;
5330 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5331 int mode)
5333 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5334 !idev->cnf.stable_secret.initialized &&
5335 !net->ipv6.devconf_dflt->stable_secret.initialized)
5336 return -EINVAL;
5337 return 1;
5340 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5342 int err = -EINVAL;
5343 struct inet6_dev *idev = __in6_dev_get(dev);
5344 struct nlattr *tb[IFLA_INET6_MAX + 1];
5346 if (!idev)
5347 return -EAFNOSUPPORT;
5349 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5350 BUG();
5352 if (tb[IFLA_INET6_TOKEN]) {
5353 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5354 if (err)
5355 return err;
5358 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5359 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5361 if (check_addr_gen_mode(mode) < 0 ||
5362 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5363 return -EINVAL;
5365 idev->cnf.addr_gen_mode = mode;
5366 err = 0;
5369 return err;
5372 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5373 u32 portid, u32 seq, int event, unsigned int flags)
5375 struct net_device *dev = idev->dev;
5376 struct ifinfomsg *hdr;
5377 struct nlmsghdr *nlh;
5378 void *protoinfo;
5380 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5381 if (!nlh)
5382 return -EMSGSIZE;
5384 hdr = nlmsg_data(nlh);
5385 hdr->ifi_family = AF_INET6;
5386 hdr->__ifi_pad = 0;
5387 hdr->ifi_type = dev->type;
5388 hdr->ifi_index = dev->ifindex;
5389 hdr->ifi_flags = dev_get_flags(dev);
5390 hdr->ifi_change = 0;
5392 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5393 (dev->addr_len &&
5394 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5395 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5396 (dev->ifindex != dev_get_iflink(dev) &&
5397 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5398 nla_put_u8(skb, IFLA_OPERSTATE,
5399 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5400 goto nla_put_failure;
5401 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5402 if (!protoinfo)
5403 goto nla_put_failure;
5405 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5406 goto nla_put_failure;
5408 nla_nest_end(skb, protoinfo);
5409 nlmsg_end(skb, nlh);
5410 return 0;
5412 nla_put_failure:
5413 nlmsg_cancel(skb, nlh);
5414 return -EMSGSIZE;
5417 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5419 struct net *net = sock_net(skb->sk);
5420 int h, s_h;
5421 int idx = 0, s_idx;
5422 struct net_device *dev;
5423 struct inet6_dev *idev;
5424 struct hlist_head *head;
5426 s_h = cb->args[0];
5427 s_idx = cb->args[1];
5429 rcu_read_lock();
5430 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5431 idx = 0;
5432 head = &net->dev_index_head[h];
5433 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5434 if (idx < s_idx)
5435 goto cont;
5436 idev = __in6_dev_get(dev);
5437 if (!idev)
5438 goto cont;
5439 if (inet6_fill_ifinfo(skb, idev,
5440 NETLINK_CB(cb->skb).portid,
5441 cb->nlh->nlmsg_seq,
5442 RTM_NEWLINK, NLM_F_MULTI) < 0)
5443 goto out;
5444 cont:
5445 idx++;
5448 out:
5449 rcu_read_unlock();
5450 cb->args[1] = idx;
5451 cb->args[0] = h;
5453 return skb->len;
5456 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5458 struct sk_buff *skb;
5459 struct net *net = dev_net(idev->dev);
5460 int err = -ENOBUFS;
5462 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5463 if (!skb)
5464 goto errout;
5466 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5467 if (err < 0) {
5468 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5469 WARN_ON(err == -EMSGSIZE);
5470 kfree_skb(skb);
5471 goto errout;
5473 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5474 return;
5475 errout:
5476 if (err < 0)
5477 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5480 static inline size_t inet6_prefix_nlmsg_size(void)
5482 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5483 + nla_total_size(sizeof(struct in6_addr))
5484 + nla_total_size(sizeof(struct prefix_cacheinfo));
5487 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5488 struct prefix_info *pinfo, u32 portid, u32 seq,
5489 int event, unsigned int flags)
5491 struct prefixmsg *pmsg;
5492 struct nlmsghdr *nlh;
5493 struct prefix_cacheinfo ci;
5495 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5496 if (!nlh)
5497 return -EMSGSIZE;
5499 pmsg = nlmsg_data(nlh);
5500 pmsg->prefix_family = AF_INET6;
5501 pmsg->prefix_pad1 = 0;
5502 pmsg->prefix_pad2 = 0;
5503 pmsg->prefix_ifindex = idev->dev->ifindex;
5504 pmsg->prefix_len = pinfo->prefix_len;
5505 pmsg->prefix_type = pinfo->type;
5506 pmsg->prefix_pad3 = 0;
5507 pmsg->prefix_flags = 0;
5508 if (pinfo->onlink)
5509 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5510 if (pinfo->autoconf)
5511 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5513 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5514 goto nla_put_failure;
5515 ci.preferred_time = ntohl(pinfo->prefered);
5516 ci.valid_time = ntohl(pinfo->valid);
5517 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5518 goto nla_put_failure;
5519 nlmsg_end(skb, nlh);
5520 return 0;
5522 nla_put_failure:
5523 nlmsg_cancel(skb, nlh);
5524 return -EMSGSIZE;
5527 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5528 struct prefix_info *pinfo)
5530 struct sk_buff *skb;
5531 struct net *net = dev_net(idev->dev);
5532 int err = -ENOBUFS;
5534 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5535 if (!skb)
5536 goto errout;
5538 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5539 if (err < 0) {
5540 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5541 WARN_ON(err == -EMSGSIZE);
5542 kfree_skb(skb);
5543 goto errout;
5545 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5546 return;
5547 errout:
5548 if (err < 0)
5549 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5552 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5554 struct net *net = dev_net(ifp->idev->dev);
5556 if (event)
5557 ASSERT_RTNL();
5559 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5561 switch (event) {
5562 case RTM_NEWADDR:
5564 * If the address was optimistic
5565 * we inserted the route at the start of
5566 * our DAD process, so we don't need
5567 * to do it again
5569 if (!rcu_access_pointer(ifp->rt->rt6i_node))
5570 ip6_ins_rt(ifp->rt);
5571 if (ifp->idev->cnf.forwarding)
5572 addrconf_join_anycast(ifp);
5573 if (!ipv6_addr_any(&ifp->peer_addr))
5574 addrconf_prefix_route(&ifp->peer_addr, 128,
5575 ifp->idev->dev, 0, 0);
5576 break;
5577 case RTM_DELADDR:
5578 if (ifp->idev->cnf.forwarding)
5579 addrconf_leave_anycast(ifp);
5580 addrconf_leave_solict(ifp->idev, &ifp->addr);
5581 if (!ipv6_addr_any(&ifp->peer_addr)) {
5582 struct rt6_info *rt;
5584 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5585 ifp->idev->dev, 0, 0);
5586 if (rt)
5587 ip6_del_rt(rt);
5589 if (ifp->rt) {
5590 if (dst_hold_safe(&ifp->rt->dst))
5591 ip6_del_rt(ifp->rt);
5593 rt_genid_bump_ipv6(net);
5594 break;
5596 atomic_inc(&net->ipv6.dev_addr_genid);
5599 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5601 rcu_read_lock_bh();
5602 if (likely(ifp->idev->dead == 0))
5603 __ipv6_ifa_notify(event, ifp);
5604 rcu_read_unlock_bh();
5607 #ifdef CONFIG_SYSCTL
5609 static
5610 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5611 void __user *buffer, size_t *lenp, loff_t *ppos)
5613 int *valp = ctl->data;
5614 int val = *valp;
5615 loff_t pos = *ppos;
5616 struct ctl_table lctl;
5617 int ret;
5620 * ctl->data points to idev->cnf.forwarding, we should
5621 * not modify it until we get the rtnl lock.
5623 lctl = *ctl;
5624 lctl.data = &val;
5626 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5628 if (write)
5629 ret = addrconf_fixup_forwarding(ctl, valp, val);
5630 if (ret)
5631 *ppos = pos;
5632 return ret;
5635 static
5636 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5637 void __user *buffer, size_t *lenp, loff_t *ppos)
5639 struct inet6_dev *idev = ctl->extra1;
5640 int min_mtu = IPV6_MIN_MTU;
5641 struct ctl_table lctl;
5643 lctl = *ctl;
5644 lctl.extra1 = &min_mtu;
5645 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5647 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5650 static void dev_disable_change(struct inet6_dev *idev)
5652 struct netdev_notifier_info info;
5654 if (!idev || !idev->dev)
5655 return;
5657 netdev_notifier_info_init(&info, idev->dev);
5658 if (idev->cnf.disable_ipv6)
5659 addrconf_notify(NULL, NETDEV_DOWN, &info);
5660 else
5661 addrconf_notify(NULL, NETDEV_UP, &info);
5664 static void addrconf_disable_change(struct net *net, __s32 newf)
5666 struct net_device *dev;
5667 struct inet6_dev *idev;
5669 for_each_netdev(net, dev) {
5670 idev = __in6_dev_get(dev);
5671 if (idev) {
5672 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5673 idev->cnf.disable_ipv6 = newf;
5674 if (changed)
5675 dev_disable_change(idev);
5680 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5682 struct net *net;
5683 int old;
5685 if (!rtnl_trylock())
5686 return restart_syscall();
5688 net = (struct net *)table->extra2;
5689 old = *p;
5690 *p = newf;
5692 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5693 rtnl_unlock();
5694 return 0;
5697 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5698 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5699 addrconf_disable_change(net, newf);
5700 } else if ((!newf) ^ (!old))
5701 dev_disable_change((struct inet6_dev *)table->extra1);
5703 rtnl_unlock();
5704 return 0;
5707 static
5708 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5709 void __user *buffer, size_t *lenp, loff_t *ppos)
5711 int *valp = ctl->data;
5712 int val = *valp;
5713 loff_t pos = *ppos;
5714 struct ctl_table lctl;
5715 int ret;
5718 * ctl->data points to idev->cnf.disable_ipv6, we should
5719 * not modify it until we get the rtnl lock.
5721 lctl = *ctl;
5722 lctl.data = &val;
5724 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5726 if (write)
5727 ret = addrconf_disable_ipv6(ctl, valp, val);
5728 if (ret)
5729 *ppos = pos;
5730 return ret;
5733 static
5734 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5735 void __user *buffer, size_t *lenp, loff_t *ppos)
5737 int *valp = ctl->data;
5738 int ret;
5739 int old, new;
5741 old = *valp;
5742 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5743 new = *valp;
5745 if (write && old != new) {
5746 struct net *net = ctl->extra2;
5748 if (!rtnl_trylock())
5749 return restart_syscall();
5751 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5752 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5753 NETCONFA_PROXY_NEIGH,
5754 NETCONFA_IFINDEX_DEFAULT,
5755 net->ipv6.devconf_dflt);
5756 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5757 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5758 NETCONFA_PROXY_NEIGH,
5759 NETCONFA_IFINDEX_ALL,
5760 net->ipv6.devconf_all);
5761 else {
5762 struct inet6_dev *idev = ctl->extra1;
5764 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5765 NETCONFA_PROXY_NEIGH,
5766 idev->dev->ifindex,
5767 &idev->cnf);
5769 rtnl_unlock();
5772 return ret;
5775 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5776 void __user *buffer, size_t *lenp,
5777 loff_t *ppos)
5779 int ret = 0;
5780 int new_val;
5781 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5782 struct net *net = (struct net *)ctl->extra2;
5784 if (!rtnl_trylock())
5785 return restart_syscall();
5787 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5789 if (write) {
5790 new_val = *((int *)ctl->data);
5792 if (check_addr_gen_mode(new_val) < 0) {
5793 ret = -EINVAL;
5794 goto out;
5797 /* request for default */
5798 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5799 ipv6_devconf_dflt.addr_gen_mode = new_val;
5801 /* request for individual net device */
5802 } else {
5803 if (!idev)
5804 goto out;
5806 if (check_stable_privacy(idev, net, new_val) < 0) {
5807 ret = -EINVAL;
5808 goto out;
5811 if (idev->cnf.addr_gen_mode != new_val) {
5812 idev->cnf.addr_gen_mode = new_val;
5813 addrconf_dev_config(idev->dev);
5818 out:
5819 rtnl_unlock();
5821 return ret;
5824 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5825 void __user *buffer, size_t *lenp,
5826 loff_t *ppos)
5828 int err;
5829 struct in6_addr addr;
5830 char str[IPV6_MAX_STRLEN];
5831 struct ctl_table lctl = *ctl;
5832 struct net *net = ctl->extra2;
5833 struct ipv6_stable_secret *secret = ctl->data;
5835 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5836 return -EIO;
5838 lctl.maxlen = IPV6_MAX_STRLEN;
5839 lctl.data = str;
5841 if (!rtnl_trylock())
5842 return restart_syscall();
5844 if (!write && !secret->initialized) {
5845 err = -EIO;
5846 goto out;
5849 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5850 if (err >= sizeof(str)) {
5851 err = -EIO;
5852 goto out;
5855 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5856 if (err || !write)
5857 goto out;
5859 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5860 err = -EIO;
5861 goto out;
5864 secret->initialized = true;
5865 secret->secret = addr;
5867 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5868 struct net_device *dev;
5870 for_each_netdev(net, dev) {
5871 struct inet6_dev *idev = __in6_dev_get(dev);
5873 if (idev) {
5874 idev->cnf.addr_gen_mode =
5875 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5878 } else {
5879 struct inet6_dev *idev = ctl->extra1;
5881 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5884 out:
5885 rtnl_unlock();
5887 return err;
5890 static
5891 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5892 int write,
5893 void __user *buffer,
5894 size_t *lenp,
5895 loff_t *ppos)
5897 int *valp = ctl->data;
5898 int val = *valp;
5899 loff_t pos = *ppos;
5900 struct ctl_table lctl;
5901 int ret;
5903 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5904 * we should not modify it until we get the rtnl lock.
5906 lctl = *ctl;
5907 lctl.data = &val;
5909 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5911 if (write)
5912 ret = addrconf_fixup_linkdown(ctl, valp, val);
5913 if (ret)
5914 *ppos = pos;
5915 return ret;
5918 static
5919 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5921 if (rt) {
5922 if (action)
5923 rt->dst.flags |= DST_NOPOLICY;
5924 else
5925 rt->dst.flags &= ~DST_NOPOLICY;
5929 static
5930 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5932 struct inet6_ifaddr *ifa;
5934 read_lock_bh(&idev->lock);
5935 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5936 spin_lock(&ifa->lock);
5937 if (ifa->rt) {
5938 struct rt6_info *rt = ifa->rt;
5939 int cpu;
5941 rcu_read_lock();
5942 addrconf_set_nopolicy(ifa->rt, val);
5943 if (rt->rt6i_pcpu) {
5944 for_each_possible_cpu(cpu) {
5945 struct rt6_info **rtp;
5947 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5948 addrconf_set_nopolicy(*rtp, val);
5951 rcu_read_unlock();
5953 spin_unlock(&ifa->lock);
5955 read_unlock_bh(&idev->lock);
5958 static
5959 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5961 struct inet6_dev *idev;
5962 struct net *net;
5964 if (!rtnl_trylock())
5965 return restart_syscall();
5967 *valp = val;
5969 net = (struct net *)ctl->extra2;
5970 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5971 rtnl_unlock();
5972 return 0;
5975 if (valp == &net->ipv6.devconf_all->disable_policy) {
5976 struct net_device *dev;
5978 for_each_netdev(net, dev) {
5979 idev = __in6_dev_get(dev);
5980 if (idev)
5981 addrconf_disable_policy_idev(idev, val);
5983 } else {
5984 idev = (struct inet6_dev *)ctl->extra1;
5985 addrconf_disable_policy_idev(idev, val);
5988 rtnl_unlock();
5989 return 0;
5992 static
5993 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5994 void __user *buffer, size_t *lenp,
5995 loff_t *ppos)
5997 int *valp = ctl->data;
5998 int val = *valp;
5999 loff_t pos = *ppos;
6000 struct ctl_table lctl;
6001 int ret;
6003 lctl = *ctl;
6004 lctl.data = &val;
6005 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6007 if (write && (*valp != val))
6008 ret = addrconf_disable_policy(ctl, valp, val);
6010 if (ret)
6011 *ppos = pos;
6013 return ret;
6016 static int minus_one = -1;
6017 static const int zero = 0;
6018 static const int one = 1;
6019 static const int two_five_five = 255;
6021 static const struct ctl_table addrconf_sysctl[] = {
6023 .procname = "forwarding",
6024 .data = &ipv6_devconf.forwarding,
6025 .maxlen = sizeof(int),
6026 .mode = 0644,
6027 .proc_handler = addrconf_sysctl_forward,
6030 .procname = "hop_limit",
6031 .data = &ipv6_devconf.hop_limit,
6032 .maxlen = sizeof(int),
6033 .mode = 0644,
6034 .proc_handler = proc_dointvec_minmax,
6035 .extra1 = (void *)&one,
6036 .extra2 = (void *)&two_five_five,
6039 .procname = "mtu",
6040 .data = &ipv6_devconf.mtu6,
6041 .maxlen = sizeof(int),
6042 .mode = 0644,
6043 .proc_handler = addrconf_sysctl_mtu,
6046 .procname = "accept_ra",
6047 .data = &ipv6_devconf.accept_ra,
6048 .maxlen = sizeof(int),
6049 .mode = 0644,
6050 .proc_handler = proc_dointvec,
6053 .procname = "accept_redirects",
6054 .data = &ipv6_devconf.accept_redirects,
6055 .maxlen = sizeof(int),
6056 .mode = 0644,
6057 .proc_handler = proc_dointvec,
6060 .procname = "autoconf",
6061 .data = &ipv6_devconf.autoconf,
6062 .maxlen = sizeof(int),
6063 .mode = 0644,
6064 .proc_handler = proc_dointvec,
6067 .procname = "dad_transmits",
6068 .data = &ipv6_devconf.dad_transmits,
6069 .maxlen = sizeof(int),
6070 .mode = 0644,
6071 .proc_handler = proc_dointvec,
6074 .procname = "router_solicitations",
6075 .data = &ipv6_devconf.rtr_solicits,
6076 .maxlen = sizeof(int),
6077 .mode = 0644,
6078 .proc_handler = proc_dointvec_minmax,
6079 .extra1 = &minus_one,
6082 .procname = "router_solicitation_interval",
6083 .data = &ipv6_devconf.rtr_solicit_interval,
6084 .maxlen = sizeof(int),
6085 .mode = 0644,
6086 .proc_handler = proc_dointvec_jiffies,
6089 .procname = "router_solicitation_max_interval",
6090 .data = &ipv6_devconf.rtr_solicit_max_interval,
6091 .maxlen = sizeof(int),
6092 .mode = 0644,
6093 .proc_handler = proc_dointvec_jiffies,
6096 .procname = "router_solicitation_delay",
6097 .data = &ipv6_devconf.rtr_solicit_delay,
6098 .maxlen = sizeof(int),
6099 .mode = 0644,
6100 .proc_handler = proc_dointvec_jiffies,
6103 .procname = "force_mld_version",
6104 .data = &ipv6_devconf.force_mld_version,
6105 .maxlen = sizeof(int),
6106 .mode = 0644,
6107 .proc_handler = proc_dointvec,
6110 .procname = "mldv1_unsolicited_report_interval",
6111 .data =
6112 &ipv6_devconf.mldv1_unsolicited_report_interval,
6113 .maxlen = sizeof(int),
6114 .mode = 0644,
6115 .proc_handler = proc_dointvec_ms_jiffies,
6118 .procname = "mldv2_unsolicited_report_interval",
6119 .data =
6120 &ipv6_devconf.mldv2_unsolicited_report_interval,
6121 .maxlen = sizeof(int),
6122 .mode = 0644,
6123 .proc_handler = proc_dointvec_ms_jiffies,
6126 .procname = "use_tempaddr",
6127 .data = &ipv6_devconf.use_tempaddr,
6128 .maxlen = sizeof(int),
6129 .mode = 0644,
6130 .proc_handler = proc_dointvec,
6133 .procname = "temp_valid_lft",
6134 .data = &ipv6_devconf.temp_valid_lft,
6135 .maxlen = sizeof(int),
6136 .mode = 0644,
6137 .proc_handler = proc_dointvec,
6140 .procname = "temp_prefered_lft",
6141 .data = &ipv6_devconf.temp_prefered_lft,
6142 .maxlen = sizeof(int),
6143 .mode = 0644,
6144 .proc_handler = proc_dointvec,
6147 .procname = "regen_max_retry",
6148 .data = &ipv6_devconf.regen_max_retry,
6149 .maxlen = sizeof(int),
6150 .mode = 0644,
6151 .proc_handler = proc_dointvec,
6154 .procname = "max_desync_factor",
6155 .data = &ipv6_devconf.max_desync_factor,
6156 .maxlen = sizeof(int),
6157 .mode = 0644,
6158 .proc_handler = proc_dointvec,
6161 .procname = "max_addresses",
6162 .data = &ipv6_devconf.max_addresses,
6163 .maxlen = sizeof(int),
6164 .mode = 0644,
6165 .proc_handler = proc_dointvec,
6168 .procname = "accept_ra_defrtr",
6169 .data = &ipv6_devconf.accept_ra_defrtr,
6170 .maxlen = sizeof(int),
6171 .mode = 0644,
6172 .proc_handler = proc_dointvec,
6175 .procname = "accept_ra_min_hop_limit",
6176 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6177 .maxlen = sizeof(int),
6178 .mode = 0644,
6179 .proc_handler = proc_dointvec,
6182 .procname = "accept_ra_pinfo",
6183 .data = &ipv6_devconf.accept_ra_pinfo,
6184 .maxlen = sizeof(int),
6185 .mode = 0644,
6186 .proc_handler = proc_dointvec,
6188 #ifdef CONFIG_IPV6_ROUTER_PREF
6190 .procname = "accept_ra_rtr_pref",
6191 .data = &ipv6_devconf.accept_ra_rtr_pref,
6192 .maxlen = sizeof(int),
6193 .mode = 0644,
6194 .proc_handler = proc_dointvec,
6197 .procname = "router_probe_interval",
6198 .data = &ipv6_devconf.rtr_probe_interval,
6199 .maxlen = sizeof(int),
6200 .mode = 0644,
6201 .proc_handler = proc_dointvec_jiffies,
6203 #ifdef CONFIG_IPV6_ROUTE_INFO
6205 .procname = "accept_ra_rt_info_min_plen",
6206 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6207 .maxlen = sizeof(int),
6208 .mode = 0644,
6209 .proc_handler = proc_dointvec,
6212 .procname = "accept_ra_rt_info_max_plen",
6213 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6214 .maxlen = sizeof(int),
6215 .mode = 0644,
6216 .proc_handler = proc_dointvec,
6218 #endif
6219 #endif
6221 .procname = "proxy_ndp",
6222 .data = &ipv6_devconf.proxy_ndp,
6223 .maxlen = sizeof(int),
6224 .mode = 0644,
6225 .proc_handler = addrconf_sysctl_proxy_ndp,
6228 .procname = "accept_source_route",
6229 .data = &ipv6_devconf.accept_source_route,
6230 .maxlen = sizeof(int),
6231 .mode = 0644,
6232 .proc_handler = proc_dointvec,
6234 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6236 .procname = "optimistic_dad",
6237 .data = &ipv6_devconf.optimistic_dad,
6238 .maxlen = sizeof(int),
6239 .mode = 0644,
6240 .proc_handler = proc_dointvec,
6243 .procname = "use_optimistic",
6244 .data = &ipv6_devconf.use_optimistic,
6245 .maxlen = sizeof(int),
6246 .mode = 0644,
6247 .proc_handler = proc_dointvec,
6249 #endif
6250 #ifdef CONFIG_IPV6_MROUTE
6252 .procname = "mc_forwarding",
6253 .data = &ipv6_devconf.mc_forwarding,
6254 .maxlen = sizeof(int),
6255 .mode = 0444,
6256 .proc_handler = proc_dointvec,
6258 #endif
6260 .procname = "disable_ipv6",
6261 .data = &ipv6_devconf.disable_ipv6,
6262 .maxlen = sizeof(int),
6263 .mode = 0644,
6264 .proc_handler = addrconf_sysctl_disable,
6267 .procname = "accept_dad",
6268 .data = &ipv6_devconf.accept_dad,
6269 .maxlen = sizeof(int),
6270 .mode = 0644,
6271 .proc_handler = proc_dointvec,
6274 .procname = "force_tllao",
6275 .data = &ipv6_devconf.force_tllao,
6276 .maxlen = sizeof(int),
6277 .mode = 0644,
6278 .proc_handler = proc_dointvec
6281 .procname = "ndisc_notify",
6282 .data = &ipv6_devconf.ndisc_notify,
6283 .maxlen = sizeof(int),
6284 .mode = 0644,
6285 .proc_handler = proc_dointvec
6288 .procname = "suppress_frag_ndisc",
6289 .data = &ipv6_devconf.suppress_frag_ndisc,
6290 .maxlen = sizeof(int),
6291 .mode = 0644,
6292 .proc_handler = proc_dointvec
6295 .procname = "accept_ra_from_local",
6296 .data = &ipv6_devconf.accept_ra_from_local,
6297 .maxlen = sizeof(int),
6298 .mode = 0644,
6299 .proc_handler = proc_dointvec,
6302 .procname = "accept_ra_mtu",
6303 .data = &ipv6_devconf.accept_ra_mtu,
6304 .maxlen = sizeof(int),
6305 .mode = 0644,
6306 .proc_handler = proc_dointvec,
6309 .procname = "stable_secret",
6310 .data = &ipv6_devconf.stable_secret,
6311 .maxlen = IPV6_MAX_STRLEN,
6312 .mode = 0600,
6313 .proc_handler = addrconf_sysctl_stable_secret,
6316 .procname = "use_oif_addrs_only",
6317 .data = &ipv6_devconf.use_oif_addrs_only,
6318 .maxlen = sizeof(int),
6319 .mode = 0644,
6320 .proc_handler = proc_dointvec,
6323 .procname = "ignore_routes_with_linkdown",
6324 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6325 .maxlen = sizeof(int),
6326 .mode = 0644,
6327 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6330 .procname = "drop_unicast_in_l2_multicast",
6331 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6332 .maxlen = sizeof(int),
6333 .mode = 0644,
6334 .proc_handler = proc_dointvec,
6337 .procname = "drop_unsolicited_na",
6338 .data = &ipv6_devconf.drop_unsolicited_na,
6339 .maxlen = sizeof(int),
6340 .mode = 0644,
6341 .proc_handler = proc_dointvec,
6344 .procname = "keep_addr_on_down",
6345 .data = &ipv6_devconf.keep_addr_on_down,
6346 .maxlen = sizeof(int),
6347 .mode = 0644,
6348 .proc_handler = proc_dointvec,
6352 .procname = "seg6_enabled",
6353 .data = &ipv6_devconf.seg6_enabled,
6354 .maxlen = sizeof(int),
6355 .mode = 0644,
6356 .proc_handler = proc_dointvec,
6358 #ifdef CONFIG_IPV6_SEG6_HMAC
6360 .procname = "seg6_require_hmac",
6361 .data = &ipv6_devconf.seg6_require_hmac,
6362 .maxlen = sizeof(int),
6363 .mode = 0644,
6364 .proc_handler = proc_dointvec,
6366 #endif
6368 .procname = "enhanced_dad",
6369 .data = &ipv6_devconf.enhanced_dad,
6370 .maxlen = sizeof(int),
6371 .mode = 0644,
6372 .proc_handler = proc_dointvec,
6375 .procname = "addr_gen_mode",
6376 .data = &ipv6_devconf.addr_gen_mode,
6377 .maxlen = sizeof(int),
6378 .mode = 0644,
6379 .proc_handler = addrconf_sysctl_addr_gen_mode,
6382 .procname = "disable_policy",
6383 .data = &ipv6_devconf.disable_policy,
6384 .maxlen = sizeof(int),
6385 .mode = 0644,
6386 .proc_handler = addrconf_sysctl_disable_policy,
6389 .procname = "ndisc_tclass",
6390 .data = &ipv6_devconf.ndisc_tclass,
6391 .maxlen = sizeof(int),
6392 .mode = 0644,
6393 .proc_handler = proc_dointvec_minmax,
6394 .extra1 = (void *)&zero,
6395 .extra2 = (void *)&two_five_five,
6398 /* sentinel */
6402 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6403 struct inet6_dev *idev, struct ipv6_devconf *p)
6405 int i, ifindex;
6406 struct ctl_table *table;
6407 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6409 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6410 if (!table)
6411 goto out;
6413 for (i = 0; table[i].data; i++) {
6414 table[i].data += (char *)p - (char *)&ipv6_devconf;
6415 /* If one of these is already set, then it is not safe to
6416 * overwrite either of them: this makes proc_dointvec_minmax
6417 * usable.
6419 if (!table[i].extra1 && !table[i].extra2) {
6420 table[i].extra1 = idev; /* embedded; no ref */
6421 table[i].extra2 = net;
6425 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6427 p->sysctl_header = register_net_sysctl(net, path, table);
6428 if (!p->sysctl_header)
6429 goto free;
6431 if (!strcmp(dev_name, "all"))
6432 ifindex = NETCONFA_IFINDEX_ALL;
6433 else if (!strcmp(dev_name, "default"))
6434 ifindex = NETCONFA_IFINDEX_DEFAULT;
6435 else
6436 ifindex = idev->dev->ifindex;
6437 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6438 ifindex, p);
6439 return 0;
6441 free:
6442 kfree(table);
6443 out:
6444 return -ENOBUFS;
6447 static void __addrconf_sysctl_unregister(struct net *net,
6448 struct ipv6_devconf *p, int ifindex)
6450 struct ctl_table *table;
6452 if (!p->sysctl_header)
6453 return;
6455 table = p->sysctl_header->ctl_table_arg;
6456 unregister_net_sysctl_table(p->sysctl_header);
6457 p->sysctl_header = NULL;
6458 kfree(table);
6460 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6463 static int addrconf_sysctl_register(struct inet6_dev *idev)
6465 int err;
6467 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6468 return -EINVAL;
6470 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6471 &ndisc_ifinfo_sysctl_change);
6472 if (err)
6473 return err;
6474 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6475 idev, &idev->cnf);
6476 if (err)
6477 neigh_sysctl_unregister(idev->nd_parms);
6479 return err;
6482 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6484 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6485 idev->dev->ifindex);
6486 neigh_sysctl_unregister(idev->nd_parms);
6490 #endif
6492 static int __net_init addrconf_init_net(struct net *net)
6494 int err = -ENOMEM;
6495 struct ipv6_devconf *all, *dflt;
6497 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6498 if (!all)
6499 goto err_alloc_all;
6501 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6502 if (!dflt)
6503 goto err_alloc_dflt;
6505 /* these will be inherited by all namespaces */
6506 dflt->autoconf = ipv6_defaults.autoconf;
6507 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6509 dflt->stable_secret.initialized = false;
6510 all->stable_secret.initialized = false;
6512 net->ipv6.devconf_all = all;
6513 net->ipv6.devconf_dflt = dflt;
6515 #ifdef CONFIG_SYSCTL
6516 err = __addrconf_sysctl_register(net, "all", NULL, all);
6517 if (err < 0)
6518 goto err_reg_all;
6520 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6521 if (err < 0)
6522 goto err_reg_dflt;
6523 #endif
6524 return 0;
6526 #ifdef CONFIG_SYSCTL
6527 err_reg_dflt:
6528 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6529 err_reg_all:
6530 kfree(dflt);
6531 #endif
6532 err_alloc_dflt:
6533 kfree(all);
6534 err_alloc_all:
6535 return err;
6538 static void __net_exit addrconf_exit_net(struct net *net)
6540 #ifdef CONFIG_SYSCTL
6541 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6542 NETCONFA_IFINDEX_DEFAULT);
6543 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6544 NETCONFA_IFINDEX_ALL);
6545 #endif
6546 kfree(net->ipv6.devconf_dflt);
6547 kfree(net->ipv6.devconf_all);
6550 static struct pernet_operations addrconf_ops = {
6551 .init = addrconf_init_net,
6552 .exit = addrconf_exit_net,
6555 static struct rtnl_af_ops inet6_ops __read_mostly = {
6556 .family = AF_INET6,
6557 .fill_link_af = inet6_fill_link_af,
6558 .get_link_af_size = inet6_get_link_af_size,
6559 .validate_link_af = inet6_validate_link_af,
6560 .set_link_af = inet6_set_link_af,
6564 * Init / cleanup code
6567 int __init addrconf_init(void)
6569 struct inet6_dev *idev;
6570 int i, err;
6572 err = ipv6_addr_label_init();
6573 if (err < 0) {
6574 pr_crit("%s: cannot initialize default policy table: %d\n",
6575 __func__, err);
6576 goto out;
6579 err = register_pernet_subsys(&addrconf_ops);
6580 if (err < 0)
6581 goto out_addrlabel;
6583 addrconf_wq = create_workqueue("ipv6_addrconf");
6584 if (!addrconf_wq) {
6585 err = -ENOMEM;
6586 goto out_nowq;
6589 /* The addrconf netdev notifier requires that loopback_dev
6590 * has it's ipv6 private information allocated and setup
6591 * before it can bring up and give link-local addresses
6592 * to other devices which are up.
6594 * Unfortunately, loopback_dev is not necessarily the first
6595 * entry in the global dev_base list of net devices. In fact,
6596 * it is likely to be the very last entry on that list.
6597 * So this causes the notifier registry below to try and
6598 * give link-local addresses to all devices besides loopback_dev
6599 * first, then loopback_dev, which cases all the non-loopback_dev
6600 * devices to fail to get a link-local address.
6602 * So, as a temporary fix, allocate the ipv6 structure for
6603 * loopback_dev first by hand.
6604 * Longer term, all of the dependencies ipv6 has upon the loopback
6605 * device and it being up should be removed.
6607 rtnl_lock();
6608 idev = ipv6_add_dev(init_net.loopback_dev);
6609 rtnl_unlock();
6610 if (IS_ERR(idev)) {
6611 err = PTR_ERR(idev);
6612 goto errlo;
6615 ip6_route_init_special_entries();
6617 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6618 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6620 register_netdevice_notifier(&ipv6_dev_notf);
6622 addrconf_verify();
6624 rtnl_af_register(&inet6_ops);
6626 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
6627 NULL, inet6_dump_ifinfo, 0);
6628 if (err < 0)
6629 goto errout;
6631 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
6632 inet6_rtm_newaddr, NULL, 0);
6633 if (err < 0)
6634 goto errout;
6635 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
6636 inet6_rtm_deladdr, NULL, 0);
6637 if (err < 0)
6638 goto errout;
6639 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
6640 inet6_rtm_getaddr, inet6_dump_ifaddr,
6641 RTNL_FLAG_DOIT_UNLOCKED);
6642 if (err < 0)
6643 goto errout;
6644 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
6645 NULL, inet6_dump_ifmcaddr, 0);
6646 if (err < 0)
6647 goto errout;
6648 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
6649 NULL, inet6_dump_ifacaddr, 0);
6650 if (err < 0)
6651 goto errout;
6652 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
6653 inet6_netconf_get_devconf,
6654 inet6_netconf_dump_devconf,
6655 RTNL_FLAG_DOIT_UNLOCKED);
6656 if (err < 0)
6657 goto errout;
6658 err = ipv6_addr_label_rtnl_register();
6659 if (err < 0)
6660 goto errout;
6662 return 0;
6663 errout:
6664 rtnl_unregister_all(PF_INET6);
6665 rtnl_af_unregister(&inet6_ops);
6666 unregister_netdevice_notifier(&ipv6_dev_notf);
6667 errlo:
6668 destroy_workqueue(addrconf_wq);
6669 out_nowq:
6670 unregister_pernet_subsys(&addrconf_ops);
6671 out_addrlabel:
6672 ipv6_addr_label_cleanup();
6673 out:
6674 return err;
6677 void addrconf_cleanup(void)
6679 struct net_device *dev;
6680 int i;
6682 unregister_netdevice_notifier(&ipv6_dev_notf);
6683 unregister_pernet_subsys(&addrconf_ops);
6684 ipv6_addr_label_cleanup();
6686 rtnl_af_unregister(&inet6_ops);
6688 rtnl_lock();
6690 /* clean dev list */
6691 for_each_netdev(&init_net, dev) {
6692 if (__in6_dev_get(dev) == NULL)
6693 continue;
6694 addrconf_ifdown(dev, 1);
6696 addrconf_ifdown(init_net.loopback_dev, 2);
6699 * Check hash table.
6701 spin_lock_bh(&addrconf_hash_lock);
6702 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6703 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6704 spin_unlock_bh(&addrconf_hash_lock);
6705 cancel_delayed_work(&addr_chk_work);
6706 rtnl_unlock();
6708 destroy_workqueue(addrconf_wq);