powerpc/powernv: Report size of OPAL memcons log
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blobc1e124bc8e1e3fcab42290acd8888e42d93ccf50
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/inet.h>
50 #include <linux/in6.h>
51 #include <linux/netdevice.h>
52 #include <linux/if_addr.h>
53 #include <linux/if_arp.h>
54 #include <linux/if_arcnet.h>
55 #include <linux/if_infiniband.h>
56 #include <linux/route.h>
57 #include <linux/inetdevice.h>
58 #include <linux/init.h>
59 #include <linux/slab.h>
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63 #include <linux/capability.h>
64 #include <linux/delay.h>
65 #include <linux/notifier.h>
66 #include <linux/string.h>
67 #include <linux/hash.h>
69 #include <net/net_namespace.h>
70 #include <net/sock.h>
71 #include <net/snmp.h>
73 #include <net/6lowpan.h>
74 #include <net/firewire.h>
75 #include <net/ipv6.h>
76 #include <net/protocol.h>
77 #include <net/ndisc.h>
78 #include <net/ip6_route.h>
79 #include <net/addrconf.h>
80 #include <net/tcp.h>
81 #include <net/ip.h>
82 #include <net/netlink.h>
83 #include <net/pkt_sched.h>
84 #include <net/l3mdev.h>
85 #include <linux/if_tunnel.h>
86 #include <linux/rtnetlink.h>
87 #include <linux/netconf.h>
88 #include <linux/random.h>
89 #include <linux/uaccess.h>
90 #include <asm/unaligned.h>
92 #include <linux/proc_fs.h>
93 #include <linux/seq_file.h>
94 #include <linux/export.h>
96 /* Set to 3 to get tracing... */
97 #define ACONF_DEBUG 2
99 #if ACONF_DEBUG >= 3
100 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
101 #else
102 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #endif
105 #define INFINITY_LIFE_TIME 0xFFFFFFFF
107 #define IPV6_MAX_STRLEN \
108 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
110 static inline u32 cstamp_delta(unsigned long cstamp)
112 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
115 static inline s32 rfc3315_s14_backoff_init(s32 irt)
117 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
118 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
119 do_div(tmp, 1000000);
120 return (s32)tmp;
123 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
125 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
126 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
127 do_div(tmp, 1000000);
128 if ((s32)tmp > mrt) {
129 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
130 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
131 do_div(tmp, 1000000);
133 return (s32)tmp;
136 #ifdef CONFIG_SYSCTL
137 static int addrconf_sysctl_register(struct inet6_dev *idev);
138 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
139 #else
140 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
142 return 0;
145 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
148 #endif
150 static void ipv6_regen_rndid(struct inet6_dev *idev);
151 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
153 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
154 static int ipv6_count_addresses(struct inet6_dev *idev);
155 static int ipv6_generate_stable_address(struct in6_addr *addr,
156 u8 dad_count,
157 const struct inet6_dev *idev);
160 * Configured unicast address hash table
162 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
163 static DEFINE_SPINLOCK(addrconf_hash_lock);
165 static void addrconf_verify(void);
166 static void addrconf_verify_rtnl(void);
167 static void addrconf_verify_work(struct work_struct *);
169 static struct workqueue_struct *addrconf_wq;
170 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
172 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
173 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
175 static void addrconf_type_change(struct net_device *dev,
176 unsigned long event);
177 static int addrconf_ifdown(struct net_device *dev, int how);
179 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
180 int plen,
181 const struct net_device *dev,
182 u32 flags, u32 noflags);
184 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
185 static void addrconf_dad_work(struct work_struct *w);
186 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
187 static void addrconf_dad_run(struct inet6_dev *idev);
188 static void addrconf_rs_timer(unsigned long data);
189 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
190 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
192 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
193 struct prefix_info *pinfo);
194 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
195 struct net_device *dev);
197 static struct ipv6_devconf ipv6_devconf __read_mostly = {
198 .forwarding = 0,
199 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
200 .mtu6 = IPV6_MIN_MTU,
201 .accept_ra = 1,
202 .accept_redirects = 1,
203 .autoconf = 1,
204 .force_mld_version = 0,
205 .mldv1_unsolicited_report_interval = 10 * HZ,
206 .mldv2_unsolicited_report_interval = HZ,
207 .dad_transmits = 1,
208 .rtr_solicits = MAX_RTR_SOLICITATIONS,
209 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
210 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
211 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
212 .use_tempaddr = 0,
213 .temp_valid_lft = TEMP_VALID_LIFETIME,
214 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
215 .regen_max_retry = REGEN_MAX_RETRY,
216 .max_desync_factor = MAX_DESYNC_FACTOR,
217 .max_addresses = IPV6_MAX_ADDRESSES,
218 .accept_ra_defrtr = 1,
219 .accept_ra_from_local = 0,
220 .accept_ra_min_hop_limit= 1,
221 .accept_ra_pinfo = 1,
222 #ifdef CONFIG_IPV6_ROUTER_PREF
223 .accept_ra_rtr_pref = 1,
224 .rtr_probe_interval = 60 * HZ,
225 #ifdef CONFIG_IPV6_ROUTE_INFO
226 .accept_ra_rt_info_max_plen = 0,
227 #endif
228 #endif
229 .proxy_ndp = 0,
230 .accept_source_route = 0, /* we do not accept RH0 by default. */
231 .disable_ipv6 = 0,
232 .accept_dad = 1,
233 .suppress_frag_ndisc = 1,
234 .accept_ra_mtu = 1,
235 .stable_secret = {
236 .initialized = false,
238 .use_oif_addrs_only = 0,
239 .ignore_routes_with_linkdown = 0,
240 .keep_addr_on_down = 0,
241 .seg6_enabled = 0,
242 #ifdef CONFIG_IPV6_SEG6_HMAC
243 .seg6_require_hmac = 0,
244 #endif
245 .enhanced_dad = 1,
248 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
249 .forwarding = 0,
250 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
251 .mtu6 = IPV6_MIN_MTU,
252 .accept_ra = 1,
253 .accept_redirects = 1,
254 .autoconf = 1,
255 .force_mld_version = 0,
256 .mldv1_unsolicited_report_interval = 10 * HZ,
257 .mldv2_unsolicited_report_interval = HZ,
258 .dad_transmits = 1,
259 .rtr_solicits = MAX_RTR_SOLICITATIONS,
260 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
261 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
262 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
263 .use_tempaddr = 0,
264 .temp_valid_lft = TEMP_VALID_LIFETIME,
265 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
266 .regen_max_retry = REGEN_MAX_RETRY,
267 .max_desync_factor = MAX_DESYNC_FACTOR,
268 .max_addresses = IPV6_MAX_ADDRESSES,
269 .accept_ra_defrtr = 1,
270 .accept_ra_from_local = 0,
271 .accept_ra_min_hop_limit= 1,
272 .accept_ra_pinfo = 1,
273 #ifdef CONFIG_IPV6_ROUTER_PREF
274 .accept_ra_rtr_pref = 1,
275 .rtr_probe_interval = 60 * HZ,
276 #ifdef CONFIG_IPV6_ROUTE_INFO
277 .accept_ra_rt_info_max_plen = 0,
278 #endif
279 #endif
280 .proxy_ndp = 0,
281 .accept_source_route = 0, /* we do not accept RH0 by default. */
282 .disable_ipv6 = 0,
283 .accept_dad = 1,
284 .suppress_frag_ndisc = 1,
285 .accept_ra_mtu = 1,
286 .stable_secret = {
287 .initialized = false,
289 .use_oif_addrs_only = 0,
290 .ignore_routes_with_linkdown = 0,
291 .keep_addr_on_down = 0,
292 .seg6_enabled = 0,
293 #ifdef CONFIG_IPV6_SEG6_HMAC
294 .seg6_require_hmac = 0,
295 #endif
296 .enhanced_dad = 1,
299 /* Check if a valid qdisc is available */
300 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
302 return !qdisc_tx_is_noop(dev);
305 static void addrconf_del_rs_timer(struct inet6_dev *idev)
307 if (del_timer(&idev->rs_timer))
308 __in6_dev_put(idev);
311 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
313 if (cancel_delayed_work(&ifp->dad_work))
314 __in6_ifa_put(ifp);
317 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
318 unsigned long when)
320 if (!timer_pending(&idev->rs_timer))
321 in6_dev_hold(idev);
322 mod_timer(&idev->rs_timer, jiffies + when);
325 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
326 unsigned long delay)
328 if (!delayed_work_pending(&ifp->dad_work))
329 in6_ifa_hold(ifp);
330 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
333 static int snmp6_alloc_dev(struct inet6_dev *idev)
335 int i;
337 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
338 if (!idev->stats.ipv6)
339 goto err_ip;
341 for_each_possible_cpu(i) {
342 struct ipstats_mib *addrconf_stats;
343 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
344 u64_stats_init(&addrconf_stats->syncp);
348 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
349 GFP_KERNEL);
350 if (!idev->stats.icmpv6dev)
351 goto err_icmp;
352 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
353 GFP_KERNEL);
354 if (!idev->stats.icmpv6msgdev)
355 goto err_icmpmsg;
357 return 0;
359 err_icmpmsg:
360 kfree(idev->stats.icmpv6dev);
361 err_icmp:
362 free_percpu(idev->stats.ipv6);
363 err_ip:
364 return -ENOMEM;
367 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
369 struct inet6_dev *ndev;
370 int err = -ENOMEM;
372 ASSERT_RTNL();
374 if (dev->mtu < IPV6_MIN_MTU)
375 return ERR_PTR(-EINVAL);
377 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
378 if (!ndev)
379 return ERR_PTR(err);
381 rwlock_init(&ndev->lock);
382 ndev->dev = dev;
383 INIT_LIST_HEAD(&ndev->addr_list);
384 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
385 (unsigned long)ndev);
386 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
388 if (ndev->cnf.stable_secret.initialized)
389 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
390 else
391 ndev->addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64;
393 ndev->cnf.mtu6 = dev->mtu;
394 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
395 if (!ndev->nd_parms) {
396 kfree(ndev);
397 return ERR_PTR(err);
399 if (ndev->cnf.forwarding)
400 dev_disable_lro(dev);
401 /* We refer to the device */
402 dev_hold(dev);
404 if (snmp6_alloc_dev(ndev) < 0) {
405 ADBG(KERN_WARNING
406 "%s: cannot allocate memory for statistics; dev=%s.\n",
407 __func__, dev->name);
408 neigh_parms_release(&nd_tbl, ndev->nd_parms);
409 dev_put(dev);
410 kfree(ndev);
411 return ERR_PTR(err);
414 if (snmp6_register_dev(ndev) < 0) {
415 ADBG(KERN_WARNING
416 "%s: cannot create /proc/net/dev_snmp6/%s\n",
417 __func__, dev->name);
418 goto err_release;
421 /* One reference from device. */
422 in6_dev_hold(ndev);
424 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
425 ndev->cnf.accept_dad = -1;
427 #if IS_ENABLED(CONFIG_IPV6_SIT)
428 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
429 pr_info("%s: Disabled Multicast RS\n", dev->name);
430 ndev->cnf.rtr_solicits = 0;
432 #endif
434 INIT_LIST_HEAD(&ndev->tempaddr_list);
435 ndev->desync_factor = U32_MAX;
436 if ((dev->flags&IFF_LOOPBACK) ||
437 dev->type == ARPHRD_TUNNEL ||
438 dev->type == ARPHRD_TUNNEL6 ||
439 dev->type == ARPHRD_SIT ||
440 dev->type == ARPHRD_NONE) {
441 ndev->cnf.use_tempaddr = -1;
442 } else
443 ipv6_regen_rndid(ndev);
445 ndev->token = in6addr_any;
447 if (netif_running(dev) && addrconf_qdisc_ok(dev))
448 ndev->if_flags |= IF_READY;
450 ipv6_mc_init_dev(ndev);
451 ndev->tstamp = jiffies;
452 err = addrconf_sysctl_register(ndev);
453 if (err) {
454 ipv6_mc_destroy_dev(ndev);
455 snmp6_unregister_dev(ndev);
456 goto err_release;
458 /* protected by rtnl_lock */
459 rcu_assign_pointer(dev->ip6_ptr, ndev);
461 /* Join interface-local all-node multicast group */
462 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
464 /* Join all-node multicast group */
465 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
467 /* Join all-router multicast group if forwarding is set */
468 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
469 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
471 return ndev;
473 err_release:
474 neigh_parms_release(&nd_tbl, ndev->nd_parms);
475 ndev->dead = 1;
476 in6_dev_finish_destroy(ndev);
477 return ERR_PTR(err);
480 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
482 struct inet6_dev *idev;
484 ASSERT_RTNL();
486 idev = __in6_dev_get(dev);
487 if (!idev) {
488 idev = ipv6_add_dev(dev);
489 if (IS_ERR(idev))
490 return NULL;
493 if (dev->flags&IFF_UP)
494 ipv6_mc_up(idev);
495 return idev;
498 static int inet6_netconf_msgsize_devconf(int type)
500 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
501 + nla_total_size(4); /* NETCONFA_IFINDEX */
502 bool all = false;
504 if (type == NETCONFA_ALL)
505 all = true;
507 if (all || type == NETCONFA_FORWARDING)
508 size += nla_total_size(4);
509 #ifdef CONFIG_IPV6_MROUTE
510 if (all || type == NETCONFA_MC_FORWARDING)
511 size += nla_total_size(4);
512 #endif
513 if (all || type == NETCONFA_PROXY_NEIGH)
514 size += nla_total_size(4);
516 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
517 size += nla_total_size(4);
519 return size;
522 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
523 struct ipv6_devconf *devconf, u32 portid,
524 u32 seq, int event, unsigned int flags,
525 int type)
527 struct nlmsghdr *nlh;
528 struct netconfmsg *ncm;
529 bool all = false;
531 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
532 flags);
533 if (!nlh)
534 return -EMSGSIZE;
536 if (type == NETCONFA_ALL)
537 all = true;
539 ncm = nlmsg_data(nlh);
540 ncm->ncm_family = AF_INET6;
542 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
543 goto nla_put_failure;
545 if ((all || type == NETCONFA_FORWARDING) &&
546 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
547 goto nla_put_failure;
548 #ifdef CONFIG_IPV6_MROUTE
549 if ((all || type == NETCONFA_MC_FORWARDING) &&
550 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
551 devconf->mc_forwarding) < 0)
552 goto nla_put_failure;
553 #endif
554 if ((all || type == NETCONFA_PROXY_NEIGH) &&
555 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
556 goto nla_put_failure;
558 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
559 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
560 devconf->ignore_routes_with_linkdown) < 0)
561 goto nla_put_failure;
563 nlmsg_end(skb, nlh);
564 return 0;
566 nla_put_failure:
567 nlmsg_cancel(skb, nlh);
568 return -EMSGSIZE;
571 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
572 struct ipv6_devconf *devconf)
574 struct sk_buff *skb;
575 int err = -ENOBUFS;
577 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
578 if (!skb)
579 goto errout;
581 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
582 RTM_NEWNETCONF, 0, type);
583 if (err < 0) {
584 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
585 WARN_ON(err == -EMSGSIZE);
586 kfree_skb(skb);
587 goto errout;
589 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
590 return;
591 errout:
592 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
595 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
596 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
597 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
598 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
599 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
602 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
603 struct nlmsghdr *nlh)
605 struct net *net = sock_net(in_skb->sk);
606 struct nlattr *tb[NETCONFA_MAX+1];
607 struct netconfmsg *ncm;
608 struct sk_buff *skb;
609 struct ipv6_devconf *devconf;
610 struct inet6_dev *in6_dev;
611 struct net_device *dev;
612 int ifindex;
613 int err;
615 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
616 devconf_ipv6_policy);
617 if (err < 0)
618 goto errout;
620 err = -EINVAL;
621 if (!tb[NETCONFA_IFINDEX])
622 goto errout;
624 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
625 switch (ifindex) {
626 case NETCONFA_IFINDEX_ALL:
627 devconf = net->ipv6.devconf_all;
628 break;
629 case NETCONFA_IFINDEX_DEFAULT:
630 devconf = net->ipv6.devconf_dflt;
631 break;
632 default:
633 dev = __dev_get_by_index(net, ifindex);
634 if (!dev)
635 goto errout;
636 in6_dev = __in6_dev_get(dev);
637 if (!in6_dev)
638 goto errout;
639 devconf = &in6_dev->cnf;
640 break;
643 err = -ENOBUFS;
644 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
645 if (!skb)
646 goto errout;
648 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
649 NETLINK_CB(in_skb).portid,
650 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
651 NETCONFA_ALL);
652 if (err < 0) {
653 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
654 WARN_ON(err == -EMSGSIZE);
655 kfree_skb(skb);
656 goto errout;
658 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
659 errout:
660 return err;
663 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
664 struct netlink_callback *cb)
666 struct net *net = sock_net(skb->sk);
667 int h, s_h;
668 int idx, s_idx;
669 struct net_device *dev;
670 struct inet6_dev *idev;
671 struct hlist_head *head;
673 s_h = cb->args[0];
674 s_idx = idx = cb->args[1];
676 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
677 idx = 0;
678 head = &net->dev_index_head[h];
679 rcu_read_lock();
680 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
681 net->dev_base_seq;
682 hlist_for_each_entry_rcu(dev, head, index_hlist) {
683 if (idx < s_idx)
684 goto cont;
685 idev = __in6_dev_get(dev);
686 if (!idev)
687 goto cont;
689 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
690 &idev->cnf,
691 NETLINK_CB(cb->skb).portid,
692 cb->nlh->nlmsg_seq,
693 RTM_NEWNETCONF,
694 NLM_F_MULTI,
695 NETCONFA_ALL) < 0) {
696 rcu_read_unlock();
697 goto done;
699 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
700 cont:
701 idx++;
703 rcu_read_unlock();
705 if (h == NETDEV_HASHENTRIES) {
706 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
707 net->ipv6.devconf_all,
708 NETLINK_CB(cb->skb).portid,
709 cb->nlh->nlmsg_seq,
710 RTM_NEWNETCONF, NLM_F_MULTI,
711 NETCONFA_ALL) < 0)
712 goto done;
713 else
714 h++;
716 if (h == NETDEV_HASHENTRIES + 1) {
717 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
718 net->ipv6.devconf_dflt,
719 NETLINK_CB(cb->skb).portid,
720 cb->nlh->nlmsg_seq,
721 RTM_NEWNETCONF, NLM_F_MULTI,
722 NETCONFA_ALL) < 0)
723 goto done;
724 else
725 h++;
727 done:
728 cb->args[0] = h;
729 cb->args[1] = idx;
731 return skb->len;
734 #ifdef CONFIG_SYSCTL
735 static void dev_forward_change(struct inet6_dev *idev)
737 struct net_device *dev;
738 struct inet6_ifaddr *ifa;
740 if (!idev)
741 return;
742 dev = idev->dev;
743 if (idev->cnf.forwarding)
744 dev_disable_lro(dev);
745 if (dev->flags & IFF_MULTICAST) {
746 if (idev->cnf.forwarding) {
747 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
748 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
749 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
750 } else {
751 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
752 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
753 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
757 list_for_each_entry(ifa, &idev->addr_list, if_list) {
758 if (ifa->flags&IFA_F_TENTATIVE)
759 continue;
760 if (idev->cnf.forwarding)
761 addrconf_join_anycast(ifa);
762 else
763 addrconf_leave_anycast(ifa);
765 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
766 dev->ifindex, &idev->cnf);
770 static void addrconf_forward_change(struct net *net, __s32 newf)
772 struct net_device *dev;
773 struct inet6_dev *idev;
775 for_each_netdev(net, dev) {
776 idev = __in6_dev_get(dev);
777 if (idev) {
778 int changed = (!idev->cnf.forwarding) ^ (!newf);
779 idev->cnf.forwarding = newf;
780 if (changed)
781 dev_forward_change(idev);
786 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
788 struct net *net;
789 int old;
791 if (!rtnl_trylock())
792 return restart_syscall();
794 net = (struct net *)table->extra2;
795 old = *p;
796 *p = newf;
798 if (p == &net->ipv6.devconf_dflt->forwarding) {
799 if ((!newf) ^ (!old))
800 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
801 NETCONFA_IFINDEX_DEFAULT,
802 net->ipv6.devconf_dflt);
803 rtnl_unlock();
804 return 0;
807 if (p == &net->ipv6.devconf_all->forwarding) {
808 int old_dflt = net->ipv6.devconf_dflt->forwarding;
810 net->ipv6.devconf_dflt->forwarding = newf;
811 if ((!newf) ^ (!old_dflt))
812 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
813 NETCONFA_IFINDEX_DEFAULT,
814 net->ipv6.devconf_dflt);
816 addrconf_forward_change(net, newf);
817 if ((!newf) ^ (!old))
818 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
819 NETCONFA_IFINDEX_ALL,
820 net->ipv6.devconf_all);
821 } else if ((!newf) ^ (!old))
822 dev_forward_change((struct inet6_dev *)table->extra1);
823 rtnl_unlock();
825 if (newf)
826 rt6_purge_dflt_routers(net);
827 return 1;
830 static void addrconf_linkdown_change(struct net *net, __s32 newf)
832 struct net_device *dev;
833 struct inet6_dev *idev;
835 for_each_netdev(net, dev) {
836 idev = __in6_dev_get(dev);
837 if (idev) {
838 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
840 idev->cnf.ignore_routes_with_linkdown = newf;
841 if (changed)
842 inet6_netconf_notify_devconf(dev_net(dev),
843 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
844 dev->ifindex,
845 &idev->cnf);
850 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
852 struct net *net;
853 int old;
855 if (!rtnl_trylock())
856 return restart_syscall();
858 net = (struct net *)table->extra2;
859 old = *p;
860 *p = newf;
862 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
863 if ((!newf) ^ (!old))
864 inet6_netconf_notify_devconf(net,
865 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
866 NETCONFA_IFINDEX_DEFAULT,
867 net->ipv6.devconf_dflt);
868 rtnl_unlock();
869 return 0;
872 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
873 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
874 addrconf_linkdown_change(net, newf);
875 if ((!newf) ^ (!old))
876 inet6_netconf_notify_devconf(net,
877 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
878 NETCONFA_IFINDEX_ALL,
879 net->ipv6.devconf_all);
881 rtnl_unlock();
883 return 1;
886 #endif
888 /* Nobody refers to this ifaddr, destroy it */
889 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
891 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
893 #ifdef NET_REFCNT_DEBUG
894 pr_debug("%s\n", __func__);
895 #endif
897 in6_dev_put(ifp->idev);
899 if (cancel_delayed_work(&ifp->dad_work))
900 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
901 ifp);
903 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
904 pr_warn("Freeing alive inet6 address %p\n", ifp);
905 return;
907 ip6_rt_put(ifp->rt);
909 kfree_rcu(ifp, rcu);
912 static void
913 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
915 struct list_head *p;
916 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
919 * Each device address list is sorted in order of scope -
920 * global before linklocal.
922 list_for_each(p, &idev->addr_list) {
923 struct inet6_ifaddr *ifa
924 = list_entry(p, struct inet6_ifaddr, if_list);
925 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
926 break;
929 list_add_tail(&ifp->if_list, p);
932 static u32 inet6_addr_hash(const struct in6_addr *addr)
934 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
937 /* On success it returns ifp with increased reference count */
939 static struct inet6_ifaddr *
940 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
941 const struct in6_addr *peer_addr, int pfxlen,
942 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
944 struct inet6_ifaddr *ifa = NULL;
945 struct rt6_info *rt;
946 unsigned int hash;
947 int err = 0;
948 int addr_type = ipv6_addr_type(addr);
950 if (addr_type == IPV6_ADDR_ANY ||
951 addr_type & IPV6_ADDR_MULTICAST ||
952 (!(idev->dev->flags & IFF_LOOPBACK) &&
953 addr_type & IPV6_ADDR_LOOPBACK))
954 return ERR_PTR(-EADDRNOTAVAIL);
956 rcu_read_lock_bh();
957 if (idev->dead) {
958 err = -ENODEV; /*XXX*/
959 goto out2;
962 if (idev->cnf.disable_ipv6) {
963 err = -EACCES;
964 goto out2;
967 spin_lock(&addrconf_hash_lock);
969 /* Ignore adding duplicate addresses on an interface */
970 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
971 ADBG("ipv6_add_addr: already assigned\n");
972 err = -EEXIST;
973 goto out;
976 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
978 if (!ifa) {
979 ADBG("ipv6_add_addr: malloc failed\n");
980 err = -ENOBUFS;
981 goto out;
984 rt = addrconf_dst_alloc(idev, addr, false);
985 if (IS_ERR(rt)) {
986 err = PTR_ERR(rt);
987 goto out;
990 neigh_parms_data_state_setall(idev->nd_parms);
992 ifa->addr = *addr;
993 if (peer_addr)
994 ifa->peer_addr = *peer_addr;
996 spin_lock_init(&ifa->lock);
997 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
998 INIT_HLIST_NODE(&ifa->addr_lst);
999 ifa->scope = scope;
1000 ifa->prefix_len = pfxlen;
1001 ifa->flags = flags | IFA_F_TENTATIVE;
1002 ifa->valid_lft = valid_lft;
1003 ifa->prefered_lft = prefered_lft;
1004 ifa->cstamp = ifa->tstamp = jiffies;
1005 ifa->tokenized = false;
1007 ifa->rt = rt;
1009 ifa->idev = idev;
1010 in6_dev_hold(idev);
1011 /* For caller */
1012 in6_ifa_hold(ifa);
1014 /* Add to big hash table */
1015 hash = inet6_addr_hash(addr);
1017 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1018 spin_unlock(&addrconf_hash_lock);
1020 write_lock(&idev->lock);
1021 /* Add to inet6_dev unicast addr list. */
1022 ipv6_link_dev_addr(idev, ifa);
1024 if (ifa->flags&IFA_F_TEMPORARY) {
1025 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1026 in6_ifa_hold(ifa);
1029 in6_ifa_hold(ifa);
1030 write_unlock(&idev->lock);
1031 out2:
1032 rcu_read_unlock_bh();
1034 if (likely(err == 0))
1035 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1036 else {
1037 kfree(ifa);
1038 ifa = ERR_PTR(err);
1041 return ifa;
1042 out:
1043 spin_unlock(&addrconf_hash_lock);
1044 goto out2;
1047 enum cleanup_prefix_rt_t {
1048 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1049 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1050 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1054 * Check, whether the prefix for ifp would still need a prefix route
1055 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1056 * constants.
1058 * 1) we don't purge prefix if address was not permanent.
1059 * prefix is managed by its own lifetime.
1060 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1061 * 3) if there are no addresses, delete prefix.
1062 * 4) if there are still other permanent address(es),
1063 * corresponding prefix is still permanent.
1064 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1065 * don't purge the prefix, assume user space is managing it.
1066 * 6) otherwise, update prefix lifetime to the
1067 * longest valid lifetime among the corresponding
1068 * addresses on the device.
1069 * Note: subsequent RA will update lifetime.
1071 static enum cleanup_prefix_rt_t
1072 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1074 struct inet6_ifaddr *ifa;
1075 struct inet6_dev *idev = ifp->idev;
1076 unsigned long lifetime;
1077 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1079 *expires = jiffies;
1081 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1082 if (ifa == ifp)
1083 continue;
1084 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1085 ifp->prefix_len))
1086 continue;
1087 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1088 return CLEANUP_PREFIX_RT_NOP;
1090 action = CLEANUP_PREFIX_RT_EXPIRE;
1092 spin_lock(&ifa->lock);
1094 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1096 * Note: Because this address is
1097 * not permanent, lifetime <
1098 * LONG_MAX / HZ here.
1100 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1101 *expires = ifa->tstamp + lifetime * HZ;
1102 spin_unlock(&ifa->lock);
1105 return action;
1108 static void
1109 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1111 struct rt6_info *rt;
1113 rt = addrconf_get_prefix_route(&ifp->addr,
1114 ifp->prefix_len,
1115 ifp->idev->dev,
1116 0, RTF_GATEWAY | RTF_DEFAULT);
1117 if (rt) {
1118 if (del_rt)
1119 ip6_del_rt(rt);
1120 else {
1121 if (!(rt->rt6i_flags & RTF_EXPIRES))
1122 rt6_set_expires(rt, expires);
1123 ip6_rt_put(rt);
1129 /* This function wants to get referenced ifp and releases it before return */
1131 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1133 int state;
1134 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1135 unsigned long expires;
1137 ASSERT_RTNL();
1139 spin_lock_bh(&ifp->lock);
1140 state = ifp->state;
1141 ifp->state = INET6_IFADDR_STATE_DEAD;
1142 spin_unlock_bh(&ifp->lock);
1144 if (state == INET6_IFADDR_STATE_DEAD)
1145 goto out;
1147 spin_lock_bh(&addrconf_hash_lock);
1148 hlist_del_init_rcu(&ifp->addr_lst);
1149 spin_unlock_bh(&addrconf_hash_lock);
1151 write_lock_bh(&ifp->idev->lock);
1153 if (ifp->flags&IFA_F_TEMPORARY) {
1154 list_del(&ifp->tmp_list);
1155 if (ifp->ifpub) {
1156 in6_ifa_put(ifp->ifpub);
1157 ifp->ifpub = NULL;
1159 __in6_ifa_put(ifp);
1162 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1163 action = check_cleanup_prefix_route(ifp, &expires);
1165 list_del_init(&ifp->if_list);
1166 __in6_ifa_put(ifp);
1168 write_unlock_bh(&ifp->idev->lock);
1170 addrconf_del_dad_work(ifp);
1172 ipv6_ifa_notify(RTM_DELADDR, ifp);
1174 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1176 if (action != CLEANUP_PREFIX_RT_NOP) {
1177 cleanup_prefix_route(ifp, expires,
1178 action == CLEANUP_PREFIX_RT_DEL);
1181 /* clean up prefsrc entries */
1182 rt6_remove_prefsrc(ifp);
1183 out:
1184 in6_ifa_put(ifp);
1187 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1189 struct inet6_dev *idev = ifp->idev;
1190 struct in6_addr addr, *tmpaddr;
1191 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1192 unsigned long regen_advance;
1193 int tmp_plen;
1194 int ret = 0;
1195 u32 addr_flags;
1196 unsigned long now = jiffies;
1197 long max_desync_factor;
1198 s32 cnf_temp_preferred_lft;
1200 write_lock_bh(&idev->lock);
1201 if (ift) {
1202 spin_lock_bh(&ift->lock);
1203 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1204 spin_unlock_bh(&ift->lock);
1205 tmpaddr = &addr;
1206 } else {
1207 tmpaddr = NULL;
1209 retry:
1210 in6_dev_hold(idev);
1211 if (idev->cnf.use_tempaddr <= 0) {
1212 write_unlock_bh(&idev->lock);
1213 pr_info("%s: use_tempaddr is disabled\n", __func__);
1214 in6_dev_put(idev);
1215 ret = -1;
1216 goto out;
1218 spin_lock_bh(&ifp->lock);
1219 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1220 idev->cnf.use_tempaddr = -1; /*XXX*/
1221 spin_unlock_bh(&ifp->lock);
1222 write_unlock_bh(&idev->lock);
1223 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1224 __func__);
1225 in6_dev_put(idev);
1226 ret = -1;
1227 goto out;
1229 in6_ifa_hold(ifp);
1230 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1231 ipv6_try_regen_rndid(idev, tmpaddr);
1232 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1233 age = (now - ifp->tstamp) / HZ;
1235 regen_advance = idev->cnf.regen_max_retry *
1236 idev->cnf.dad_transmits *
1237 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1239 /* recalculate max_desync_factor each time and update
1240 * idev->desync_factor if it's larger
1242 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1243 max_desync_factor = min_t(__u32,
1244 idev->cnf.max_desync_factor,
1245 cnf_temp_preferred_lft - regen_advance);
1247 if (unlikely(idev->desync_factor > max_desync_factor)) {
1248 if (max_desync_factor > 0) {
1249 get_random_bytes(&idev->desync_factor,
1250 sizeof(idev->desync_factor));
1251 idev->desync_factor %= max_desync_factor;
1252 } else {
1253 idev->desync_factor = 0;
1257 tmp_valid_lft = min_t(__u32,
1258 ifp->valid_lft,
1259 idev->cnf.temp_valid_lft + age);
1260 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1261 idev->desync_factor;
1262 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1263 tmp_plen = ifp->prefix_len;
1264 tmp_tstamp = ifp->tstamp;
1265 spin_unlock_bh(&ifp->lock);
1267 write_unlock_bh(&idev->lock);
1269 /* A temporary address is created only if this calculated Preferred
1270 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1271 * an implementation must not create a temporary address with a zero
1272 * Preferred Lifetime.
1273 * Use age calculation as in addrconf_verify to avoid unnecessary
1274 * temporary addresses being generated.
1276 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1277 if (tmp_prefered_lft <= regen_advance + age) {
1278 in6_ifa_put(ifp);
1279 in6_dev_put(idev);
1280 ret = -1;
1281 goto out;
1284 addr_flags = IFA_F_TEMPORARY;
1285 /* set in addrconf_prefix_rcv() */
1286 if (ifp->flags & IFA_F_OPTIMISTIC)
1287 addr_flags |= IFA_F_OPTIMISTIC;
1289 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1290 ipv6_addr_scope(&addr), addr_flags,
1291 tmp_valid_lft, tmp_prefered_lft);
1292 if (IS_ERR(ift)) {
1293 in6_ifa_put(ifp);
1294 in6_dev_put(idev);
1295 pr_info("%s: retry temporary address regeneration\n", __func__);
1296 tmpaddr = &addr;
1297 write_lock_bh(&idev->lock);
1298 goto retry;
1301 spin_lock_bh(&ift->lock);
1302 ift->ifpub = ifp;
1303 ift->cstamp = now;
1304 ift->tstamp = tmp_tstamp;
1305 spin_unlock_bh(&ift->lock);
1307 addrconf_dad_start(ift);
1308 in6_ifa_put(ift);
1309 in6_dev_put(idev);
1310 out:
1311 return ret;
1315 * Choose an appropriate source address (RFC3484)
1317 enum {
1318 IPV6_SADDR_RULE_INIT = 0,
1319 IPV6_SADDR_RULE_LOCAL,
1320 IPV6_SADDR_RULE_SCOPE,
1321 IPV6_SADDR_RULE_PREFERRED,
1322 #ifdef CONFIG_IPV6_MIP6
1323 IPV6_SADDR_RULE_HOA,
1324 #endif
1325 IPV6_SADDR_RULE_OIF,
1326 IPV6_SADDR_RULE_LABEL,
1327 IPV6_SADDR_RULE_PRIVACY,
1328 IPV6_SADDR_RULE_ORCHID,
1329 IPV6_SADDR_RULE_PREFIX,
1330 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1331 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1332 #endif
1333 IPV6_SADDR_RULE_MAX
1336 struct ipv6_saddr_score {
1337 int rule;
1338 int addr_type;
1339 struct inet6_ifaddr *ifa;
1340 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1341 int scopedist;
1342 int matchlen;
1345 struct ipv6_saddr_dst {
1346 const struct in6_addr *addr;
1347 int ifindex;
1348 int scope;
1349 int label;
1350 unsigned int prefs;
1353 static inline int ipv6_saddr_preferred(int type)
1355 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1356 return 1;
1357 return 0;
1360 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1362 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1363 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1364 #else
1365 return false;
1366 #endif
1369 static int ipv6_get_saddr_eval(struct net *net,
1370 struct ipv6_saddr_score *score,
1371 struct ipv6_saddr_dst *dst,
1372 int i)
1374 int ret;
1376 if (i <= score->rule) {
1377 switch (i) {
1378 case IPV6_SADDR_RULE_SCOPE:
1379 ret = score->scopedist;
1380 break;
1381 case IPV6_SADDR_RULE_PREFIX:
1382 ret = score->matchlen;
1383 break;
1384 default:
1385 ret = !!test_bit(i, score->scorebits);
1387 goto out;
1390 switch (i) {
1391 case IPV6_SADDR_RULE_INIT:
1392 /* Rule 0: remember if hiscore is not ready yet */
1393 ret = !!score->ifa;
1394 break;
1395 case IPV6_SADDR_RULE_LOCAL:
1396 /* Rule 1: Prefer same address */
1397 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1398 break;
1399 case IPV6_SADDR_RULE_SCOPE:
1400 /* Rule 2: Prefer appropriate scope
1402 * ret
1404 * -1 | d 15
1405 * ---+--+-+---> scope
1407 * | d is scope of the destination.
1408 * B-d | \
1409 * | \ <- smaller scope is better if
1410 * B-15 | \ if scope is enough for destination.
1411 * | ret = B - scope (-1 <= scope >= d <= 15).
1412 * d-C-1 | /
1413 * |/ <- greater is better
1414 * -C / if scope is not enough for destination.
1415 * /| ret = scope - C (-1 <= d < scope <= 15).
1417 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1418 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1419 * Assume B = 0 and we get C > 29.
1421 ret = __ipv6_addr_src_scope(score->addr_type);
1422 if (ret >= dst->scope)
1423 ret = -ret;
1424 else
1425 ret -= 128; /* 30 is enough */
1426 score->scopedist = ret;
1427 break;
1428 case IPV6_SADDR_RULE_PREFERRED:
1430 /* Rule 3: Avoid deprecated and optimistic addresses */
1431 u8 avoid = IFA_F_DEPRECATED;
1433 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1434 avoid |= IFA_F_OPTIMISTIC;
1435 ret = ipv6_saddr_preferred(score->addr_type) ||
1436 !(score->ifa->flags & avoid);
1437 break;
1439 #ifdef CONFIG_IPV6_MIP6
1440 case IPV6_SADDR_RULE_HOA:
1442 /* Rule 4: Prefer home address */
1443 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1444 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1445 break;
1447 #endif
1448 case IPV6_SADDR_RULE_OIF:
1449 /* Rule 5: Prefer outgoing interface */
1450 ret = (!dst->ifindex ||
1451 dst->ifindex == score->ifa->idev->dev->ifindex);
1452 break;
1453 case IPV6_SADDR_RULE_LABEL:
1454 /* Rule 6: Prefer matching label */
1455 ret = ipv6_addr_label(net,
1456 &score->ifa->addr, score->addr_type,
1457 score->ifa->idev->dev->ifindex) == dst->label;
1458 break;
1459 case IPV6_SADDR_RULE_PRIVACY:
1461 /* Rule 7: Prefer public address
1462 * Note: prefer temporary address if use_tempaddr >= 2
1464 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1465 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1466 score->ifa->idev->cnf.use_tempaddr >= 2;
1467 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1468 break;
1470 case IPV6_SADDR_RULE_ORCHID:
1471 /* Rule 8-: Prefer ORCHID vs ORCHID or
1472 * non-ORCHID vs non-ORCHID
1474 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1475 ipv6_addr_orchid(dst->addr));
1476 break;
1477 case IPV6_SADDR_RULE_PREFIX:
1478 /* Rule 8: Use longest matching prefix */
1479 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1480 if (ret > score->ifa->prefix_len)
1481 ret = score->ifa->prefix_len;
1482 score->matchlen = ret;
1483 break;
1484 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1485 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1486 /* Optimistic addresses still have lower precedence than other
1487 * preferred addresses.
1489 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1490 break;
1491 #endif
1492 default:
1493 ret = 0;
1496 if (ret)
1497 __set_bit(i, score->scorebits);
1498 score->rule = i;
1499 out:
1500 return ret;
1503 static int __ipv6_dev_get_saddr(struct net *net,
1504 struct ipv6_saddr_dst *dst,
1505 struct inet6_dev *idev,
1506 struct ipv6_saddr_score *scores,
1507 int hiscore_idx)
1509 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1511 read_lock_bh(&idev->lock);
1512 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1513 int i;
1516 * - Tentative Address (RFC2462 section 5.4)
1517 * - A tentative address is not considered
1518 * "assigned to an interface" in the traditional
1519 * sense, unless it is also flagged as optimistic.
1520 * - Candidate Source Address (section 4)
1521 * - In any case, anycast addresses, multicast
1522 * addresses, and the unspecified address MUST
1523 * NOT be included in a candidate set.
1525 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1526 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1527 continue;
1529 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1531 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1532 score->addr_type & IPV6_ADDR_MULTICAST)) {
1533 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1534 idev->dev->name);
1535 continue;
1538 score->rule = -1;
1539 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1541 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1542 int minihiscore, miniscore;
1544 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1545 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1547 if (minihiscore > miniscore) {
1548 if (i == IPV6_SADDR_RULE_SCOPE &&
1549 score->scopedist > 0) {
1551 * special case:
1552 * each remaining entry
1553 * has too small (not enough)
1554 * scope, because ifa entries
1555 * are sorted by their scope
1556 * values.
1558 goto out;
1560 break;
1561 } else if (minihiscore < miniscore) {
1562 if (hiscore->ifa)
1563 in6_ifa_put(hiscore->ifa);
1565 in6_ifa_hold(score->ifa);
1567 swap(hiscore, score);
1568 hiscore_idx = 1 - hiscore_idx;
1570 /* restore our iterator */
1571 score->ifa = hiscore->ifa;
1573 break;
1577 out:
1578 read_unlock_bh(&idev->lock);
1579 return hiscore_idx;
1582 static int ipv6_get_saddr_master(struct net *net,
1583 const struct net_device *dst_dev,
1584 const struct net_device *master,
1585 struct ipv6_saddr_dst *dst,
1586 struct ipv6_saddr_score *scores,
1587 int hiscore_idx)
1589 struct inet6_dev *idev;
1591 idev = __in6_dev_get(dst_dev);
1592 if (idev)
1593 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1594 scores, hiscore_idx);
1596 idev = __in6_dev_get(master);
1597 if (idev)
1598 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1599 scores, hiscore_idx);
1601 return hiscore_idx;
1604 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1605 const struct in6_addr *daddr, unsigned int prefs,
1606 struct in6_addr *saddr)
1608 struct ipv6_saddr_score scores[2], *hiscore;
1609 struct ipv6_saddr_dst dst;
1610 struct inet6_dev *idev;
1611 struct net_device *dev;
1612 int dst_type;
1613 bool use_oif_addr = false;
1614 int hiscore_idx = 0;
1616 dst_type = __ipv6_addr_type(daddr);
1617 dst.addr = daddr;
1618 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1619 dst.scope = __ipv6_addr_src_scope(dst_type);
1620 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1621 dst.prefs = prefs;
1623 scores[hiscore_idx].rule = -1;
1624 scores[hiscore_idx].ifa = NULL;
1626 rcu_read_lock();
1628 /* Candidate Source Address (section 4)
1629 * - multicast and link-local destination address,
1630 * the set of candidate source address MUST only
1631 * include addresses assigned to interfaces
1632 * belonging to the same link as the outgoing
1633 * interface.
1634 * (- For site-local destination addresses, the
1635 * set of candidate source addresses MUST only
1636 * include addresses assigned to interfaces
1637 * belonging to the same site as the outgoing
1638 * interface.)
1639 * - "It is RECOMMENDED that the candidate source addresses
1640 * be the set of unicast addresses assigned to the
1641 * interface that will be used to send to the destination
1642 * (the 'outgoing' interface)." (RFC 6724)
1644 if (dst_dev) {
1645 idev = __in6_dev_get(dst_dev);
1646 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1647 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1648 (idev && idev->cnf.use_oif_addrs_only)) {
1649 use_oif_addr = true;
1653 if (use_oif_addr) {
1654 if (idev)
1655 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1656 } else {
1657 const struct net_device *master;
1658 int master_idx = 0;
1660 /* if dst_dev exists and is enslaved to an L3 device, then
1661 * prefer addresses from dst_dev and then the master over
1662 * any other enslaved devices in the L3 domain.
1664 master = l3mdev_master_dev_rcu(dst_dev);
1665 if (master) {
1666 master_idx = master->ifindex;
1668 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1669 master, &dst,
1670 scores, hiscore_idx);
1672 if (scores[hiscore_idx].ifa)
1673 goto out;
1676 for_each_netdev_rcu(net, dev) {
1677 /* only consider addresses on devices in the
1678 * same L3 domain
1680 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1681 continue;
1682 idev = __in6_dev_get(dev);
1683 if (!idev)
1684 continue;
1685 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1689 out:
1690 rcu_read_unlock();
1692 hiscore = &scores[hiscore_idx];
1693 if (!hiscore->ifa)
1694 return -EADDRNOTAVAIL;
1696 *saddr = hiscore->ifa->addr;
1697 in6_ifa_put(hiscore->ifa);
1698 return 0;
1700 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1702 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1703 u32 banned_flags)
1705 struct inet6_ifaddr *ifp;
1706 int err = -EADDRNOTAVAIL;
1708 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1709 if (ifp->scope > IFA_LINK)
1710 break;
1711 if (ifp->scope == IFA_LINK &&
1712 !(ifp->flags & banned_flags)) {
1713 *addr = ifp->addr;
1714 err = 0;
1715 break;
1718 return err;
1721 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1722 u32 banned_flags)
1724 struct inet6_dev *idev;
1725 int err = -EADDRNOTAVAIL;
1727 rcu_read_lock();
1728 idev = __in6_dev_get(dev);
1729 if (idev) {
1730 read_lock_bh(&idev->lock);
1731 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1732 read_unlock_bh(&idev->lock);
1734 rcu_read_unlock();
1735 return err;
1738 static int ipv6_count_addresses(struct inet6_dev *idev)
1740 int cnt = 0;
1741 struct inet6_ifaddr *ifp;
1743 read_lock_bh(&idev->lock);
1744 list_for_each_entry(ifp, &idev->addr_list, if_list)
1745 cnt++;
1746 read_unlock_bh(&idev->lock);
1747 return cnt;
1750 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1751 const struct net_device *dev, int strict)
1753 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1755 EXPORT_SYMBOL(ipv6_chk_addr);
1757 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1758 const struct net_device *dev, int strict,
1759 u32 banned_flags)
1761 struct inet6_ifaddr *ifp;
1762 unsigned int hash = inet6_addr_hash(addr);
1763 u32 ifp_flags;
1765 rcu_read_lock_bh();
1766 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1767 if (!net_eq(dev_net(ifp->idev->dev), net))
1768 continue;
1769 /* Decouple optimistic from tentative for evaluation here.
1770 * Ban optimistic addresses explicitly, when required.
1772 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1773 ? (ifp->flags&~IFA_F_TENTATIVE)
1774 : ifp->flags;
1775 if (ipv6_addr_equal(&ifp->addr, addr) &&
1776 !(ifp_flags&banned_flags) &&
1777 (!dev || ifp->idev->dev == dev ||
1778 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1779 rcu_read_unlock_bh();
1780 return 1;
1784 rcu_read_unlock_bh();
1785 return 0;
1787 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1789 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1790 struct net_device *dev)
1792 unsigned int hash = inet6_addr_hash(addr);
1793 struct inet6_ifaddr *ifp;
1795 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1796 if (!net_eq(dev_net(ifp->idev->dev), net))
1797 continue;
1798 if (ipv6_addr_equal(&ifp->addr, addr)) {
1799 if (!dev || ifp->idev->dev == dev)
1800 return true;
1803 return false;
1806 /* Compares an address/prefix_len with addresses on device @dev.
1807 * If one is found it returns true.
1809 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1810 const unsigned int prefix_len, struct net_device *dev)
1812 struct inet6_dev *idev;
1813 struct inet6_ifaddr *ifa;
1814 bool ret = false;
1816 rcu_read_lock();
1817 idev = __in6_dev_get(dev);
1818 if (idev) {
1819 read_lock_bh(&idev->lock);
1820 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1821 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1822 if (ret)
1823 break;
1825 read_unlock_bh(&idev->lock);
1827 rcu_read_unlock();
1829 return ret;
1831 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1833 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1835 struct inet6_dev *idev;
1836 struct inet6_ifaddr *ifa;
1837 int onlink;
1839 onlink = 0;
1840 rcu_read_lock();
1841 idev = __in6_dev_get(dev);
1842 if (idev) {
1843 read_lock_bh(&idev->lock);
1844 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1845 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1846 ifa->prefix_len);
1847 if (onlink)
1848 break;
1850 read_unlock_bh(&idev->lock);
1852 rcu_read_unlock();
1853 return onlink;
1855 EXPORT_SYMBOL(ipv6_chk_prefix);
1857 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1858 struct net_device *dev, int strict)
1860 struct inet6_ifaddr *ifp, *result = NULL;
1861 unsigned int hash = inet6_addr_hash(addr);
1863 rcu_read_lock_bh();
1864 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1865 if (!net_eq(dev_net(ifp->idev->dev), net))
1866 continue;
1867 if (ipv6_addr_equal(&ifp->addr, addr)) {
1868 if (!dev || ifp->idev->dev == dev ||
1869 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1870 result = ifp;
1871 in6_ifa_hold(ifp);
1872 break;
1876 rcu_read_unlock_bh();
1878 return result;
1881 /* Gets referenced address, destroys ifaddr */
1883 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1885 if (dad_failed)
1886 ifp->flags |= IFA_F_DADFAILED;
1888 if (ifp->flags&IFA_F_PERMANENT) {
1889 spin_lock_bh(&ifp->lock);
1890 addrconf_del_dad_work(ifp);
1891 ifp->flags |= IFA_F_TENTATIVE;
1892 spin_unlock_bh(&ifp->lock);
1893 if (dad_failed)
1894 ipv6_ifa_notify(0, ifp);
1895 in6_ifa_put(ifp);
1896 } else if (ifp->flags&IFA_F_TEMPORARY) {
1897 struct inet6_ifaddr *ifpub;
1898 spin_lock_bh(&ifp->lock);
1899 ifpub = ifp->ifpub;
1900 if (ifpub) {
1901 in6_ifa_hold(ifpub);
1902 spin_unlock_bh(&ifp->lock);
1903 ipv6_create_tempaddr(ifpub, ifp);
1904 in6_ifa_put(ifpub);
1905 } else {
1906 spin_unlock_bh(&ifp->lock);
1908 ipv6_del_addr(ifp);
1909 } else {
1910 ipv6_del_addr(ifp);
1914 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1916 int err = -ENOENT;
1918 spin_lock_bh(&ifp->lock);
1919 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1920 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1921 err = 0;
1923 spin_unlock_bh(&ifp->lock);
1925 return err;
1928 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1930 struct inet6_dev *idev = ifp->idev;
1931 struct net *net = dev_net(ifp->idev->dev);
1933 if (addrconf_dad_end(ifp)) {
1934 in6_ifa_put(ifp);
1935 return;
1938 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1939 ifp->idev->dev->name, &ifp->addr);
1941 spin_lock_bh(&ifp->lock);
1943 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1944 int scope = ifp->scope;
1945 u32 flags = ifp->flags;
1946 struct in6_addr new_addr;
1947 struct inet6_ifaddr *ifp2;
1948 u32 valid_lft, preferred_lft;
1949 int pfxlen = ifp->prefix_len;
1950 int retries = ifp->stable_privacy_retry + 1;
1952 if (retries > net->ipv6.sysctl.idgen_retries) {
1953 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1954 ifp->idev->dev->name);
1955 goto errdad;
1958 new_addr = ifp->addr;
1959 if (ipv6_generate_stable_address(&new_addr, retries,
1960 idev))
1961 goto errdad;
1963 valid_lft = ifp->valid_lft;
1964 preferred_lft = ifp->prefered_lft;
1966 spin_unlock_bh(&ifp->lock);
1968 if (idev->cnf.max_addresses &&
1969 ipv6_count_addresses(idev) >=
1970 idev->cnf.max_addresses)
1971 goto lock_errdad;
1973 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
1974 ifp->idev->dev->name);
1976 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
1977 scope, flags, valid_lft,
1978 preferred_lft);
1979 if (IS_ERR(ifp2))
1980 goto lock_errdad;
1982 spin_lock_bh(&ifp2->lock);
1983 ifp2->stable_privacy_retry = retries;
1984 ifp2->state = INET6_IFADDR_STATE_PREDAD;
1985 spin_unlock_bh(&ifp2->lock);
1987 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
1988 in6_ifa_put(ifp2);
1989 lock_errdad:
1990 spin_lock_bh(&ifp->lock);
1993 errdad:
1994 /* transition from _POSTDAD to _ERRDAD */
1995 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1996 spin_unlock_bh(&ifp->lock);
1998 addrconf_mod_dad_work(ifp, 0);
1999 in6_ifa_put(ifp);
2002 /* Join to solicited addr multicast group.
2003 * caller must hold RTNL */
2004 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2006 struct in6_addr maddr;
2008 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2009 return;
2011 addrconf_addr_solict_mult(addr, &maddr);
2012 ipv6_dev_mc_inc(dev, &maddr);
2015 /* caller must hold RTNL */
2016 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2018 struct in6_addr maddr;
2020 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2021 return;
2023 addrconf_addr_solict_mult(addr, &maddr);
2024 __ipv6_dev_mc_dec(idev, &maddr);
2027 /* caller must hold RTNL */
2028 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2030 struct in6_addr addr;
2032 if (ifp->prefix_len >= 127) /* RFC 6164 */
2033 return;
2034 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2035 if (ipv6_addr_any(&addr))
2036 return;
2037 __ipv6_dev_ac_inc(ifp->idev, &addr);
2040 /* caller must hold RTNL */
2041 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2043 struct in6_addr addr;
2045 if (ifp->prefix_len >= 127) /* RFC 6164 */
2046 return;
2047 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2048 if (ipv6_addr_any(&addr))
2049 return;
2050 __ipv6_dev_ac_dec(ifp->idev, &addr);
2053 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
2055 if (dev->addr_len != EUI64_ADDR_LEN)
2056 return -1;
2057 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2058 eui[0] ^= 2;
2059 return 0;
2062 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2064 union fwnet_hwaddr *ha;
2066 if (dev->addr_len != FWNET_ALEN)
2067 return -1;
2069 ha = (union fwnet_hwaddr *)dev->dev_addr;
2071 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2072 eui[0] ^= 2;
2073 return 0;
2076 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2078 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2079 if (dev->addr_len != ARCNET_ALEN)
2080 return -1;
2081 memset(eui, 0, 7);
2082 eui[7] = *(u8 *)dev->dev_addr;
2083 return 0;
2086 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2088 if (dev->addr_len != INFINIBAND_ALEN)
2089 return -1;
2090 memcpy(eui, dev->dev_addr + 12, 8);
2091 eui[0] |= 2;
2092 return 0;
2095 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2097 if (addr == 0)
2098 return -1;
2099 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2100 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2101 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2102 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2103 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2104 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2105 eui[1] = 0;
2106 eui[2] = 0x5E;
2107 eui[3] = 0xFE;
2108 memcpy(eui + 4, &addr, 4);
2109 return 0;
2112 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2114 if (dev->priv_flags & IFF_ISATAP)
2115 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2116 return -1;
2119 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2121 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2124 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2126 memcpy(eui, dev->perm_addr, 3);
2127 memcpy(eui + 5, dev->perm_addr + 3, 3);
2128 eui[3] = 0xFF;
2129 eui[4] = 0xFE;
2130 eui[0] ^= 2;
2131 return 0;
2134 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2136 switch (dev->type) {
2137 case ARPHRD_ETHER:
2138 case ARPHRD_FDDI:
2139 return addrconf_ifid_eui48(eui, dev);
2140 case ARPHRD_ARCNET:
2141 return addrconf_ifid_arcnet(eui, dev);
2142 case ARPHRD_INFINIBAND:
2143 return addrconf_ifid_infiniband(eui, dev);
2144 case ARPHRD_SIT:
2145 return addrconf_ifid_sit(eui, dev);
2146 case ARPHRD_IPGRE:
2147 return addrconf_ifid_gre(eui, dev);
2148 case ARPHRD_6LOWPAN:
2149 return addrconf_ifid_eui64(eui, dev);
2150 case ARPHRD_IEEE1394:
2151 return addrconf_ifid_ieee1394(eui, dev);
2152 case ARPHRD_TUNNEL6:
2153 return addrconf_ifid_ip6tnl(eui, dev);
2155 return -1;
2158 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2160 int err = -1;
2161 struct inet6_ifaddr *ifp;
2163 read_lock_bh(&idev->lock);
2164 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2165 if (ifp->scope > IFA_LINK)
2166 break;
2167 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2168 memcpy(eui, ifp->addr.s6_addr+8, 8);
2169 err = 0;
2170 break;
2173 read_unlock_bh(&idev->lock);
2174 return err;
2177 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2178 static void ipv6_regen_rndid(struct inet6_dev *idev)
2180 regen:
2181 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2182 idev->rndid[0] &= ~0x02;
2185 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2186 * check if generated address is not inappropriate
2188 * - Reserved subnet anycast (RFC 2526)
2189 * 11111101 11....11 1xxxxxxx
2190 * - ISATAP (RFC4214) 6.1
2191 * 00-00-5E-FE-xx-xx-xx-xx
2192 * - value 0
2193 * - XXX: already assigned to an address on the device
2195 if (idev->rndid[0] == 0xfd &&
2196 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2197 (idev->rndid[7]&0x80))
2198 goto regen;
2199 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2200 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2201 goto regen;
2202 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2203 goto regen;
2207 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2209 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2210 ipv6_regen_rndid(idev);
2214 * Add prefix route.
2217 static void
2218 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2219 unsigned long expires, u32 flags)
2221 struct fib6_config cfg = {
2222 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2223 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2224 .fc_ifindex = dev->ifindex,
2225 .fc_expires = expires,
2226 .fc_dst_len = plen,
2227 .fc_flags = RTF_UP | flags,
2228 .fc_nlinfo.nl_net = dev_net(dev),
2229 .fc_protocol = RTPROT_KERNEL,
2232 cfg.fc_dst = *pfx;
2234 /* Prevent useless cloning on PtP SIT.
2235 This thing is done here expecting that the whole
2236 class of non-broadcast devices need not cloning.
2238 #if IS_ENABLED(CONFIG_IPV6_SIT)
2239 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2240 cfg.fc_flags |= RTF_NONEXTHOP;
2241 #endif
2243 ip6_route_add(&cfg);
2247 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2248 int plen,
2249 const struct net_device *dev,
2250 u32 flags, u32 noflags)
2252 struct fib6_node *fn;
2253 struct rt6_info *rt = NULL;
2254 struct fib6_table *table;
2255 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2257 table = fib6_get_table(dev_net(dev), tb_id);
2258 if (!table)
2259 return NULL;
2261 read_lock_bh(&table->tb6_lock);
2262 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2263 if (!fn)
2264 goto out;
2266 noflags |= RTF_CACHE;
2267 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2268 if (rt->dst.dev->ifindex != dev->ifindex)
2269 continue;
2270 if ((rt->rt6i_flags & flags) != flags)
2271 continue;
2272 if ((rt->rt6i_flags & noflags) != 0)
2273 continue;
2274 dst_hold(&rt->dst);
2275 break;
2277 out:
2278 read_unlock_bh(&table->tb6_lock);
2279 return rt;
2283 /* Create "default" multicast route to the interface */
2285 static void addrconf_add_mroute(struct net_device *dev)
2287 struct fib6_config cfg = {
2288 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2289 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2290 .fc_ifindex = dev->ifindex,
2291 .fc_dst_len = 8,
2292 .fc_flags = RTF_UP,
2293 .fc_nlinfo.nl_net = dev_net(dev),
2296 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2298 ip6_route_add(&cfg);
2301 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2303 struct inet6_dev *idev;
2305 ASSERT_RTNL();
2307 idev = ipv6_find_idev(dev);
2308 if (!idev)
2309 return ERR_PTR(-ENOBUFS);
2311 if (idev->cnf.disable_ipv6)
2312 return ERR_PTR(-EACCES);
2314 /* Add default multicast route */
2315 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2316 addrconf_add_mroute(dev);
2318 return idev;
2321 static void manage_tempaddrs(struct inet6_dev *idev,
2322 struct inet6_ifaddr *ifp,
2323 __u32 valid_lft, __u32 prefered_lft,
2324 bool create, unsigned long now)
2326 u32 flags;
2327 struct inet6_ifaddr *ift;
2329 read_lock_bh(&idev->lock);
2330 /* update all temporary addresses in the list */
2331 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2332 int age, max_valid, max_prefered;
2334 if (ifp != ift->ifpub)
2335 continue;
2337 /* RFC 4941 section 3.3:
2338 * If a received option will extend the lifetime of a public
2339 * address, the lifetimes of temporary addresses should
2340 * be extended, subject to the overall constraint that no
2341 * temporary addresses should ever remain "valid" or "preferred"
2342 * for a time longer than (TEMP_VALID_LIFETIME) or
2343 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2345 age = (now - ift->cstamp) / HZ;
2346 max_valid = idev->cnf.temp_valid_lft - age;
2347 if (max_valid < 0)
2348 max_valid = 0;
2350 max_prefered = idev->cnf.temp_prefered_lft -
2351 idev->desync_factor - age;
2352 if (max_prefered < 0)
2353 max_prefered = 0;
2355 if (valid_lft > max_valid)
2356 valid_lft = max_valid;
2358 if (prefered_lft > max_prefered)
2359 prefered_lft = max_prefered;
2361 spin_lock(&ift->lock);
2362 flags = ift->flags;
2363 ift->valid_lft = valid_lft;
2364 ift->prefered_lft = prefered_lft;
2365 ift->tstamp = now;
2366 if (prefered_lft > 0)
2367 ift->flags &= ~IFA_F_DEPRECATED;
2369 spin_unlock(&ift->lock);
2370 if (!(flags&IFA_F_TENTATIVE))
2371 ipv6_ifa_notify(0, ift);
2374 if ((create || list_empty(&idev->tempaddr_list)) &&
2375 idev->cnf.use_tempaddr > 0) {
2376 /* When a new public address is created as described
2377 * in [ADDRCONF], also create a new temporary address.
2378 * Also create a temporary address if it's enabled but
2379 * no temporary address currently exists.
2381 read_unlock_bh(&idev->lock);
2382 ipv6_create_tempaddr(ifp, NULL);
2383 } else {
2384 read_unlock_bh(&idev->lock);
2388 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2390 return idev->addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2391 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2394 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2395 const struct prefix_info *pinfo,
2396 struct inet6_dev *in6_dev,
2397 const struct in6_addr *addr, int addr_type,
2398 u32 addr_flags, bool sllao, bool tokenized,
2399 __u32 valid_lft, u32 prefered_lft)
2401 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2402 int create = 0, update_lft = 0;
2404 if (!ifp && valid_lft) {
2405 int max_addresses = in6_dev->cnf.max_addresses;
2407 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2408 if (in6_dev->cnf.optimistic_dad &&
2409 !net->ipv6.devconf_all->forwarding && sllao)
2410 addr_flags |= IFA_F_OPTIMISTIC;
2411 #endif
2413 /* Do not allow to create too much of autoconfigured
2414 * addresses; this would be too easy way to crash kernel.
2416 if (!max_addresses ||
2417 ipv6_count_addresses(in6_dev) < max_addresses)
2418 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2419 pinfo->prefix_len,
2420 addr_type&IPV6_ADDR_SCOPE_MASK,
2421 addr_flags, valid_lft,
2422 prefered_lft);
2424 if (IS_ERR_OR_NULL(ifp))
2425 return -1;
2427 update_lft = 0;
2428 create = 1;
2429 spin_lock_bh(&ifp->lock);
2430 ifp->flags |= IFA_F_MANAGETEMPADDR;
2431 ifp->cstamp = jiffies;
2432 ifp->tokenized = tokenized;
2433 spin_unlock_bh(&ifp->lock);
2434 addrconf_dad_start(ifp);
2437 if (ifp) {
2438 u32 flags;
2439 unsigned long now;
2440 u32 stored_lft;
2442 /* update lifetime (RFC2462 5.5.3 e) */
2443 spin_lock_bh(&ifp->lock);
2444 now = jiffies;
2445 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2446 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2447 else
2448 stored_lft = 0;
2449 if (!update_lft && !create && stored_lft) {
2450 const u32 minimum_lft = min_t(u32,
2451 stored_lft, MIN_VALID_LIFETIME);
2452 valid_lft = max(valid_lft, minimum_lft);
2454 /* RFC4862 Section 5.5.3e:
2455 * "Note that the preferred lifetime of the
2456 * corresponding address is always reset to
2457 * the Preferred Lifetime in the received
2458 * Prefix Information option, regardless of
2459 * whether the valid lifetime is also reset or
2460 * ignored."
2462 * So we should always update prefered_lft here.
2464 update_lft = 1;
2467 if (update_lft) {
2468 ifp->valid_lft = valid_lft;
2469 ifp->prefered_lft = prefered_lft;
2470 ifp->tstamp = now;
2471 flags = ifp->flags;
2472 ifp->flags &= ~IFA_F_DEPRECATED;
2473 spin_unlock_bh(&ifp->lock);
2475 if (!(flags&IFA_F_TENTATIVE))
2476 ipv6_ifa_notify(0, ifp);
2477 } else
2478 spin_unlock_bh(&ifp->lock);
2480 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2481 create, now);
2483 in6_ifa_put(ifp);
2484 addrconf_verify();
2487 return 0;
2489 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2491 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2493 struct prefix_info *pinfo;
2494 __u32 valid_lft;
2495 __u32 prefered_lft;
2496 int addr_type, err;
2497 u32 addr_flags = 0;
2498 struct inet6_dev *in6_dev;
2499 struct net *net = dev_net(dev);
2501 pinfo = (struct prefix_info *) opt;
2503 if (len < sizeof(struct prefix_info)) {
2504 ADBG("addrconf: prefix option too short\n");
2505 return;
2509 * Validation checks ([ADDRCONF], page 19)
2512 addr_type = ipv6_addr_type(&pinfo->prefix);
2514 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2515 return;
2517 valid_lft = ntohl(pinfo->valid);
2518 prefered_lft = ntohl(pinfo->prefered);
2520 if (prefered_lft > valid_lft) {
2521 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2522 return;
2525 in6_dev = in6_dev_get(dev);
2527 if (!in6_dev) {
2528 net_dbg_ratelimited("addrconf: device %s not configured\n",
2529 dev->name);
2530 return;
2534 * Two things going on here:
2535 * 1) Add routes for on-link prefixes
2536 * 2) Configure prefixes with the auto flag set
2539 if (pinfo->onlink) {
2540 struct rt6_info *rt;
2541 unsigned long rt_expires;
2543 /* Avoid arithmetic overflow. Really, we could
2544 * save rt_expires in seconds, likely valid_lft,
2545 * but it would require division in fib gc, that it
2546 * not good.
2548 if (HZ > USER_HZ)
2549 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2550 else
2551 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2553 if (addrconf_finite_timeout(rt_expires))
2554 rt_expires *= HZ;
2556 rt = addrconf_get_prefix_route(&pinfo->prefix,
2557 pinfo->prefix_len,
2558 dev,
2559 RTF_ADDRCONF | RTF_PREFIX_RT,
2560 RTF_GATEWAY | RTF_DEFAULT);
2562 if (rt) {
2563 /* Autoconf prefix route */
2564 if (valid_lft == 0) {
2565 ip6_del_rt(rt);
2566 rt = NULL;
2567 } else if (addrconf_finite_timeout(rt_expires)) {
2568 /* not infinity */
2569 rt6_set_expires(rt, jiffies + rt_expires);
2570 } else {
2571 rt6_clean_expires(rt);
2573 } else if (valid_lft) {
2574 clock_t expires = 0;
2575 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2576 if (addrconf_finite_timeout(rt_expires)) {
2577 /* not infinity */
2578 flags |= RTF_EXPIRES;
2579 expires = jiffies_to_clock_t(rt_expires);
2581 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2582 dev, expires, flags);
2584 ip6_rt_put(rt);
2587 /* Try to figure out our local address for this prefix */
2589 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2590 struct in6_addr addr;
2591 bool tokenized = false, dev_addr_generated = false;
2593 if (pinfo->prefix_len == 64) {
2594 memcpy(&addr, &pinfo->prefix, 8);
2596 if (!ipv6_addr_any(&in6_dev->token)) {
2597 read_lock_bh(&in6_dev->lock);
2598 memcpy(addr.s6_addr + 8,
2599 in6_dev->token.s6_addr + 8, 8);
2600 read_unlock_bh(&in6_dev->lock);
2601 tokenized = true;
2602 } else if (is_addr_mode_generate_stable(in6_dev) &&
2603 !ipv6_generate_stable_address(&addr, 0,
2604 in6_dev)) {
2605 addr_flags |= IFA_F_STABLE_PRIVACY;
2606 goto ok;
2607 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2608 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2609 goto put;
2610 } else {
2611 dev_addr_generated = true;
2613 goto ok;
2615 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2616 pinfo->prefix_len);
2617 goto put;
2620 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2621 &addr, addr_type,
2622 addr_flags, sllao,
2623 tokenized, valid_lft,
2624 prefered_lft);
2625 if (err)
2626 goto put;
2628 /* Ignore error case here because previous prefix add addr was
2629 * successful which will be notified.
2631 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2632 addr_type, addr_flags, sllao,
2633 tokenized, valid_lft,
2634 prefered_lft,
2635 dev_addr_generated);
2637 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2638 put:
2639 in6_dev_put(in6_dev);
2643 * Set destination address.
2644 * Special case for SIT interfaces where we create a new "virtual"
2645 * device.
2647 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2649 struct in6_ifreq ireq;
2650 struct net_device *dev;
2651 int err = -EINVAL;
2653 rtnl_lock();
2655 err = -EFAULT;
2656 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2657 goto err_exit;
2659 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2661 err = -ENODEV;
2662 if (!dev)
2663 goto err_exit;
2665 #if IS_ENABLED(CONFIG_IPV6_SIT)
2666 if (dev->type == ARPHRD_SIT) {
2667 const struct net_device_ops *ops = dev->netdev_ops;
2668 struct ifreq ifr;
2669 struct ip_tunnel_parm p;
2671 err = -EADDRNOTAVAIL;
2672 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2673 goto err_exit;
2675 memset(&p, 0, sizeof(p));
2676 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2677 p.iph.saddr = 0;
2678 p.iph.version = 4;
2679 p.iph.ihl = 5;
2680 p.iph.protocol = IPPROTO_IPV6;
2681 p.iph.ttl = 64;
2682 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2684 if (ops->ndo_do_ioctl) {
2685 mm_segment_t oldfs = get_fs();
2687 set_fs(KERNEL_DS);
2688 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2689 set_fs(oldfs);
2690 } else
2691 err = -EOPNOTSUPP;
2693 if (err == 0) {
2694 err = -ENOBUFS;
2695 dev = __dev_get_by_name(net, p.name);
2696 if (!dev)
2697 goto err_exit;
2698 err = dev_open(dev);
2701 #endif
2703 err_exit:
2704 rtnl_unlock();
2705 return err;
2708 static int ipv6_mc_config(struct sock *sk, bool join,
2709 const struct in6_addr *addr, int ifindex)
2711 int ret;
2713 ASSERT_RTNL();
2715 lock_sock(sk);
2716 if (join)
2717 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2718 else
2719 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2720 release_sock(sk);
2722 return ret;
2726 * Manual configuration of address on an interface
2728 static int inet6_addr_add(struct net *net, int ifindex,
2729 const struct in6_addr *pfx,
2730 const struct in6_addr *peer_pfx,
2731 unsigned int plen, __u32 ifa_flags,
2732 __u32 prefered_lft, __u32 valid_lft)
2734 struct inet6_ifaddr *ifp;
2735 struct inet6_dev *idev;
2736 struct net_device *dev;
2737 unsigned long timeout;
2738 clock_t expires;
2739 int scope;
2740 u32 flags;
2742 ASSERT_RTNL();
2744 if (plen > 128)
2745 return -EINVAL;
2747 /* check the lifetime */
2748 if (!valid_lft || prefered_lft > valid_lft)
2749 return -EINVAL;
2751 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2752 return -EINVAL;
2754 dev = __dev_get_by_index(net, ifindex);
2755 if (!dev)
2756 return -ENODEV;
2758 idev = addrconf_add_dev(dev);
2759 if (IS_ERR(idev))
2760 return PTR_ERR(idev);
2762 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2763 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2764 true, pfx, ifindex);
2766 if (ret < 0)
2767 return ret;
2770 scope = ipv6_addr_scope(pfx);
2772 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2773 if (addrconf_finite_timeout(timeout)) {
2774 expires = jiffies_to_clock_t(timeout * HZ);
2775 valid_lft = timeout;
2776 flags = RTF_EXPIRES;
2777 } else {
2778 expires = 0;
2779 flags = 0;
2780 ifa_flags |= IFA_F_PERMANENT;
2783 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2784 if (addrconf_finite_timeout(timeout)) {
2785 if (timeout == 0)
2786 ifa_flags |= IFA_F_DEPRECATED;
2787 prefered_lft = timeout;
2790 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2791 valid_lft, prefered_lft);
2793 if (!IS_ERR(ifp)) {
2794 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2795 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2796 expires, flags);
2800 * Note that section 3.1 of RFC 4429 indicates
2801 * that the Optimistic flag should not be set for
2802 * manually configured addresses
2804 addrconf_dad_start(ifp);
2805 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2806 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2807 true, jiffies);
2808 in6_ifa_put(ifp);
2809 addrconf_verify_rtnl();
2810 return 0;
2811 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2812 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2813 false, pfx, ifindex);
2816 return PTR_ERR(ifp);
2819 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2820 const struct in6_addr *pfx, unsigned int plen)
2822 struct inet6_ifaddr *ifp;
2823 struct inet6_dev *idev;
2824 struct net_device *dev;
2826 if (plen > 128)
2827 return -EINVAL;
2829 dev = __dev_get_by_index(net, ifindex);
2830 if (!dev)
2831 return -ENODEV;
2833 idev = __in6_dev_get(dev);
2834 if (!idev)
2835 return -ENXIO;
2837 read_lock_bh(&idev->lock);
2838 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2839 if (ifp->prefix_len == plen &&
2840 ipv6_addr_equal(pfx, &ifp->addr)) {
2841 in6_ifa_hold(ifp);
2842 read_unlock_bh(&idev->lock);
2844 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2845 (ifa_flags & IFA_F_MANAGETEMPADDR))
2846 manage_tempaddrs(idev, ifp, 0, 0, false,
2847 jiffies);
2848 ipv6_del_addr(ifp);
2849 addrconf_verify_rtnl();
2850 if (ipv6_addr_is_multicast(pfx)) {
2851 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2852 false, pfx, dev->ifindex);
2854 return 0;
2857 read_unlock_bh(&idev->lock);
2858 return -EADDRNOTAVAIL;
2862 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2864 struct in6_ifreq ireq;
2865 int err;
2867 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2868 return -EPERM;
2870 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2871 return -EFAULT;
2873 rtnl_lock();
2874 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2875 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2876 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2877 rtnl_unlock();
2878 return err;
2881 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2883 struct in6_ifreq ireq;
2884 int err;
2886 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2887 return -EPERM;
2889 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2890 return -EFAULT;
2892 rtnl_lock();
2893 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2894 ireq.ifr6_prefixlen);
2895 rtnl_unlock();
2896 return err;
2899 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2900 int plen, int scope)
2902 struct inet6_ifaddr *ifp;
2904 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2905 scope, IFA_F_PERMANENT,
2906 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2907 if (!IS_ERR(ifp)) {
2908 spin_lock_bh(&ifp->lock);
2909 ifp->flags &= ~IFA_F_TENTATIVE;
2910 spin_unlock_bh(&ifp->lock);
2911 rt_genid_bump_ipv6(dev_net(idev->dev));
2912 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2913 in6_ifa_put(ifp);
2917 #if IS_ENABLED(CONFIG_IPV6_SIT)
2918 static void sit_add_v4_addrs(struct inet6_dev *idev)
2920 struct in6_addr addr;
2921 struct net_device *dev;
2922 struct net *net = dev_net(idev->dev);
2923 int scope, plen;
2924 u32 pflags = 0;
2926 ASSERT_RTNL();
2928 memset(&addr, 0, sizeof(struct in6_addr));
2929 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2931 if (idev->dev->flags&IFF_POINTOPOINT) {
2932 addr.s6_addr32[0] = htonl(0xfe800000);
2933 scope = IFA_LINK;
2934 plen = 64;
2935 } else {
2936 scope = IPV6_ADDR_COMPATv4;
2937 plen = 96;
2938 pflags |= RTF_NONEXTHOP;
2941 if (addr.s6_addr32[3]) {
2942 add_addr(idev, &addr, plen, scope);
2943 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2944 return;
2947 for_each_netdev(net, dev) {
2948 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2949 if (in_dev && (dev->flags & IFF_UP)) {
2950 struct in_ifaddr *ifa;
2952 int flag = scope;
2954 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2956 addr.s6_addr32[3] = ifa->ifa_local;
2958 if (ifa->ifa_scope == RT_SCOPE_LINK)
2959 continue;
2960 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2961 if (idev->dev->flags&IFF_POINTOPOINT)
2962 continue;
2963 flag |= IFA_HOST;
2966 add_addr(idev, &addr, plen, flag);
2967 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2968 pflags);
2973 #endif
2975 static void init_loopback(struct net_device *dev)
2977 struct inet6_dev *idev;
2978 struct net_device *sp_dev;
2979 struct inet6_ifaddr *sp_ifa;
2980 struct rt6_info *sp_rt;
2982 /* ::1 */
2984 ASSERT_RTNL();
2986 idev = ipv6_find_idev(dev);
2987 if (!idev) {
2988 pr_debug("%s: add_dev failed\n", __func__);
2989 return;
2992 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2994 /* Add routes to other interface's IPv6 addresses */
2995 for_each_netdev(dev_net(dev), sp_dev) {
2996 if (!strcmp(sp_dev->name, dev->name))
2997 continue;
2999 idev = __in6_dev_get(sp_dev);
3000 if (!idev)
3001 continue;
3003 read_lock_bh(&idev->lock);
3004 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
3006 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
3007 continue;
3009 if (sp_ifa->rt) {
3010 /* This dst has been added to garbage list when
3011 * lo device down, release this obsolete dst and
3012 * reallocate a new router for ifa.
3014 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3015 ip6_rt_put(sp_ifa->rt);
3016 sp_ifa->rt = NULL;
3017 } else {
3018 continue;
3022 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3024 /* Failure cases are ignored */
3025 if (!IS_ERR(sp_rt)) {
3026 sp_ifa->rt = sp_rt;
3027 ip6_ins_rt(sp_rt);
3030 read_unlock_bh(&idev->lock);
3034 void addrconf_add_linklocal(struct inet6_dev *idev,
3035 const struct in6_addr *addr, u32 flags)
3037 struct inet6_ifaddr *ifp;
3038 u32 addr_flags = flags | IFA_F_PERMANENT;
3040 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3041 if (idev->cnf.optimistic_dad &&
3042 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3043 addr_flags |= IFA_F_OPTIMISTIC;
3044 #endif
3046 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3047 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3048 if (!IS_ERR(ifp)) {
3049 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3050 addrconf_dad_start(ifp);
3051 in6_ifa_put(ifp);
3054 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3056 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3058 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3059 return true;
3061 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3062 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3063 return true;
3065 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3066 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3067 return true;
3069 return false;
3072 static int ipv6_generate_stable_address(struct in6_addr *address,
3073 u8 dad_count,
3074 const struct inet6_dev *idev)
3076 static DEFINE_SPINLOCK(lock);
3077 static __u32 digest[SHA_DIGEST_WORDS];
3078 static __u32 workspace[SHA_WORKSPACE_WORDS];
3080 static union {
3081 char __data[SHA_MESSAGE_BYTES];
3082 struct {
3083 struct in6_addr secret;
3084 __be32 prefix[2];
3085 unsigned char hwaddr[MAX_ADDR_LEN];
3086 u8 dad_count;
3087 } __packed;
3088 } data;
3090 struct in6_addr secret;
3091 struct in6_addr temp;
3092 struct net *net = dev_net(idev->dev);
3094 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3096 if (idev->cnf.stable_secret.initialized)
3097 secret = idev->cnf.stable_secret.secret;
3098 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3099 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3100 else
3101 return -1;
3103 retry:
3104 spin_lock_bh(&lock);
3106 sha_init(digest);
3107 memset(&data, 0, sizeof(data));
3108 memset(workspace, 0, sizeof(workspace));
3109 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3110 data.prefix[0] = address->s6_addr32[0];
3111 data.prefix[1] = address->s6_addr32[1];
3112 data.secret = secret;
3113 data.dad_count = dad_count;
3115 sha_transform(digest, data.__data, workspace);
3117 temp = *address;
3118 temp.s6_addr32[2] = (__force __be32)digest[0];
3119 temp.s6_addr32[3] = (__force __be32)digest[1];
3121 spin_unlock_bh(&lock);
3123 if (ipv6_reserved_interfaceid(temp)) {
3124 dad_count++;
3125 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3126 return -1;
3127 goto retry;
3130 *address = temp;
3131 return 0;
3134 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3136 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3138 if (s->initialized)
3139 return;
3140 s = &idev->cnf.stable_secret;
3141 get_random_bytes(&s->secret, sizeof(s->secret));
3142 s->initialized = true;
3145 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3147 struct in6_addr addr;
3149 /* no link local addresses on L3 master devices */
3150 if (netif_is_l3_master(idev->dev))
3151 return;
3153 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3155 switch (idev->addr_gen_mode) {
3156 case IN6_ADDR_GEN_MODE_RANDOM:
3157 ipv6_gen_mode_random_init(idev);
3158 /* fallthrough */
3159 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3160 if (!ipv6_generate_stable_address(&addr, 0, idev))
3161 addrconf_add_linklocal(idev, &addr,
3162 IFA_F_STABLE_PRIVACY);
3163 else if (prefix_route)
3164 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3165 break;
3166 case IN6_ADDR_GEN_MODE_EUI64:
3167 /* addrconf_add_linklocal also adds a prefix_route and we
3168 * only need to care about prefix routes if ipv6_generate_eui64
3169 * couldn't generate one.
3171 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3172 addrconf_add_linklocal(idev, &addr, 0);
3173 else if (prefix_route)
3174 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3175 break;
3176 case IN6_ADDR_GEN_MODE_NONE:
3177 default:
3178 /* will not add any link local address */
3179 break;
3183 static void addrconf_dev_config(struct net_device *dev)
3185 struct inet6_dev *idev;
3187 ASSERT_RTNL();
3189 if ((dev->type != ARPHRD_ETHER) &&
3190 (dev->type != ARPHRD_FDDI) &&
3191 (dev->type != ARPHRD_ARCNET) &&
3192 (dev->type != ARPHRD_INFINIBAND) &&
3193 (dev->type != ARPHRD_IEEE1394) &&
3194 (dev->type != ARPHRD_TUNNEL6) &&
3195 (dev->type != ARPHRD_6LOWPAN) &&
3196 (dev->type != ARPHRD_NONE)) {
3197 /* Alas, we support only Ethernet autoconfiguration. */
3198 return;
3201 idev = addrconf_add_dev(dev);
3202 if (IS_ERR(idev))
3203 return;
3205 /* this device type has no EUI support */
3206 if (dev->type == ARPHRD_NONE &&
3207 idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3208 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3210 addrconf_addr_gen(idev, false);
3213 #if IS_ENABLED(CONFIG_IPV6_SIT)
3214 static void addrconf_sit_config(struct net_device *dev)
3216 struct inet6_dev *idev;
3218 ASSERT_RTNL();
3221 * Configure the tunnel with one of our IPv4
3222 * addresses... we should configure all of
3223 * our v4 addrs in the tunnel
3226 idev = ipv6_find_idev(dev);
3227 if (!idev) {
3228 pr_debug("%s: add_dev failed\n", __func__);
3229 return;
3232 if (dev->priv_flags & IFF_ISATAP) {
3233 addrconf_addr_gen(idev, false);
3234 return;
3237 sit_add_v4_addrs(idev);
3239 if (dev->flags&IFF_POINTOPOINT)
3240 addrconf_add_mroute(dev);
3242 #endif
3244 #if IS_ENABLED(CONFIG_NET_IPGRE)
3245 static void addrconf_gre_config(struct net_device *dev)
3247 struct inet6_dev *idev;
3249 ASSERT_RTNL();
3251 idev = ipv6_find_idev(dev);
3252 if (!idev) {
3253 pr_debug("%s: add_dev failed\n", __func__);
3254 return;
3257 addrconf_addr_gen(idev, true);
3258 if (dev->flags & IFF_POINTOPOINT)
3259 addrconf_add_mroute(dev);
3261 #endif
3263 static int fixup_permanent_addr(struct inet6_dev *idev,
3264 struct inet6_ifaddr *ifp)
3266 if (!ifp->rt) {
3267 struct rt6_info *rt;
3269 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3270 if (unlikely(IS_ERR(rt)))
3271 return PTR_ERR(rt);
3273 ifp->rt = rt;
3276 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3277 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3278 idev->dev, 0, 0);
3281 addrconf_dad_start(ifp);
3283 return 0;
3286 static void addrconf_permanent_addr(struct net_device *dev)
3288 struct inet6_ifaddr *ifp, *tmp;
3289 struct inet6_dev *idev;
3291 idev = __in6_dev_get(dev);
3292 if (!idev)
3293 return;
3295 write_lock_bh(&idev->lock);
3297 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3298 if ((ifp->flags & IFA_F_PERMANENT) &&
3299 fixup_permanent_addr(idev, ifp) < 0) {
3300 write_unlock_bh(&idev->lock);
3301 ipv6_del_addr(ifp);
3302 write_lock_bh(&idev->lock);
3304 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3305 idev->dev->name, &ifp->addr);
3309 write_unlock_bh(&idev->lock);
3312 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3313 void *ptr)
3315 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3316 struct netdev_notifier_changeupper_info *info;
3317 struct inet6_dev *idev = __in6_dev_get(dev);
3318 int run_pending = 0;
3319 int err;
3321 switch (event) {
3322 case NETDEV_REGISTER:
3323 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3324 idev = ipv6_add_dev(dev);
3325 if (IS_ERR(idev))
3326 return notifier_from_errno(PTR_ERR(idev));
3328 break;
3330 case NETDEV_CHANGEMTU:
3331 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3332 if (dev->mtu < IPV6_MIN_MTU) {
3333 addrconf_ifdown(dev, 1);
3334 break;
3337 if (idev) {
3338 rt6_mtu_change(dev, dev->mtu);
3339 idev->cnf.mtu6 = dev->mtu;
3340 break;
3343 /* allocate new idev */
3344 idev = ipv6_add_dev(dev);
3345 if (IS_ERR(idev))
3346 break;
3348 /* device is still not ready */
3349 if (!(idev->if_flags & IF_READY))
3350 break;
3352 run_pending = 1;
3354 /* fall through */
3356 case NETDEV_UP:
3357 case NETDEV_CHANGE:
3358 if (dev->flags & IFF_SLAVE)
3359 break;
3361 if (idev && idev->cnf.disable_ipv6)
3362 break;
3364 if (event == NETDEV_UP) {
3365 /* restore routes for permanent addresses */
3366 addrconf_permanent_addr(dev);
3368 if (!addrconf_qdisc_ok(dev)) {
3369 /* device is not ready yet. */
3370 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3371 dev->name);
3372 break;
3375 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3376 idev = ipv6_add_dev(dev);
3378 if (!IS_ERR_OR_NULL(idev)) {
3379 idev->if_flags |= IF_READY;
3380 run_pending = 1;
3382 } else if (event == NETDEV_CHANGE) {
3383 if (!addrconf_qdisc_ok(dev)) {
3384 /* device is still not ready. */
3385 break;
3388 if (idev) {
3389 if (idev->if_flags & IF_READY)
3390 /* device is already configured. */
3391 break;
3392 idev->if_flags |= IF_READY;
3395 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3396 dev->name);
3398 run_pending = 1;
3401 switch (dev->type) {
3402 #if IS_ENABLED(CONFIG_IPV6_SIT)
3403 case ARPHRD_SIT:
3404 addrconf_sit_config(dev);
3405 break;
3406 #endif
3407 #if IS_ENABLED(CONFIG_NET_IPGRE)
3408 case ARPHRD_IPGRE:
3409 addrconf_gre_config(dev);
3410 break;
3411 #endif
3412 case ARPHRD_LOOPBACK:
3413 init_loopback(dev);
3414 break;
3416 default:
3417 addrconf_dev_config(dev);
3418 break;
3421 if (!IS_ERR_OR_NULL(idev)) {
3422 if (run_pending)
3423 addrconf_dad_run(idev);
3426 * If the MTU changed during the interface down,
3427 * when the interface up, the changed MTU must be
3428 * reflected in the idev as well as routers.
3430 if (idev->cnf.mtu6 != dev->mtu &&
3431 dev->mtu >= IPV6_MIN_MTU) {
3432 rt6_mtu_change(dev, dev->mtu);
3433 idev->cnf.mtu6 = dev->mtu;
3435 idev->tstamp = jiffies;
3436 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3439 * If the changed mtu during down is lower than
3440 * IPV6_MIN_MTU stop IPv6 on this interface.
3442 if (dev->mtu < IPV6_MIN_MTU)
3443 addrconf_ifdown(dev, 1);
3445 break;
3447 case NETDEV_DOWN:
3448 case NETDEV_UNREGISTER:
3450 * Remove all addresses from this interface.
3452 addrconf_ifdown(dev, event != NETDEV_DOWN);
3453 break;
3455 case NETDEV_CHANGENAME:
3456 if (idev) {
3457 snmp6_unregister_dev(idev);
3458 addrconf_sysctl_unregister(idev);
3459 err = addrconf_sysctl_register(idev);
3460 if (err)
3461 return notifier_from_errno(err);
3462 err = snmp6_register_dev(idev);
3463 if (err) {
3464 addrconf_sysctl_unregister(idev);
3465 return notifier_from_errno(err);
3468 break;
3470 case NETDEV_PRE_TYPE_CHANGE:
3471 case NETDEV_POST_TYPE_CHANGE:
3472 if (idev)
3473 addrconf_type_change(dev, event);
3474 break;
3476 case NETDEV_CHANGEUPPER:
3477 info = ptr;
3479 /* flush all routes if dev is linked to or unlinked from
3480 * an L3 master device (e.g., VRF)
3482 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3483 addrconf_ifdown(dev, 0);
3486 return NOTIFY_OK;
3490 * addrconf module should be notified of a device going up
3492 static struct notifier_block ipv6_dev_notf = {
3493 .notifier_call = addrconf_notify,
3496 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3498 struct inet6_dev *idev;
3499 ASSERT_RTNL();
3501 idev = __in6_dev_get(dev);
3503 if (event == NETDEV_POST_TYPE_CHANGE)
3504 ipv6_mc_remap(idev);
3505 else if (event == NETDEV_PRE_TYPE_CHANGE)
3506 ipv6_mc_unmap(idev);
3509 static bool addr_is_local(const struct in6_addr *addr)
3511 return ipv6_addr_type(addr) &
3512 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3515 static int addrconf_ifdown(struct net_device *dev, int how)
3517 struct net *net = dev_net(dev);
3518 struct inet6_dev *idev;
3519 struct inet6_ifaddr *ifa, *tmp;
3520 struct list_head del_list;
3521 int _keep_addr;
3522 bool keep_addr;
3523 int state, i;
3525 ASSERT_RTNL();
3527 rt6_ifdown(net, dev);
3528 neigh_ifdown(&nd_tbl, dev);
3530 idev = __in6_dev_get(dev);
3531 if (!idev)
3532 return -ENODEV;
3535 * Step 1: remove reference to ipv6 device from parent device.
3536 * Do not dev_put!
3538 if (how) {
3539 idev->dead = 1;
3541 /* protected by rtnl_lock */
3542 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3544 /* Step 1.5: remove snmp6 entry */
3545 snmp6_unregister_dev(idev);
3549 /* aggregate the system setting and interface setting */
3550 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3551 if (!_keep_addr)
3552 _keep_addr = idev->cnf.keep_addr_on_down;
3554 /* combine the user config with event to determine if permanent
3555 * addresses are to be removed from address hash table
3557 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3559 /* Step 2: clear hash table */
3560 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3561 struct hlist_head *h = &inet6_addr_lst[i];
3563 spin_lock_bh(&addrconf_hash_lock);
3564 restart:
3565 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3566 if (ifa->idev == idev) {
3567 addrconf_del_dad_work(ifa);
3568 /* combined flag + permanent flag decide if
3569 * address is retained on a down event
3571 if (!keep_addr ||
3572 !(ifa->flags & IFA_F_PERMANENT) ||
3573 addr_is_local(&ifa->addr)) {
3574 hlist_del_init_rcu(&ifa->addr_lst);
3575 goto restart;
3579 spin_unlock_bh(&addrconf_hash_lock);
3582 write_lock_bh(&idev->lock);
3584 addrconf_del_rs_timer(idev);
3586 /* Step 2: clear flags for stateless addrconf */
3587 if (!how)
3588 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3590 /* Step 3: clear tempaddr list */
3591 while (!list_empty(&idev->tempaddr_list)) {
3592 ifa = list_first_entry(&idev->tempaddr_list,
3593 struct inet6_ifaddr, tmp_list);
3594 list_del(&ifa->tmp_list);
3595 write_unlock_bh(&idev->lock);
3596 spin_lock_bh(&ifa->lock);
3598 if (ifa->ifpub) {
3599 in6_ifa_put(ifa->ifpub);
3600 ifa->ifpub = NULL;
3602 spin_unlock_bh(&ifa->lock);
3603 in6_ifa_put(ifa);
3604 write_lock_bh(&idev->lock);
3607 /* re-combine the user config with event to determine if permanent
3608 * addresses are to be removed from the interface list
3610 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3612 INIT_LIST_HEAD(&del_list);
3613 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3614 struct rt6_info *rt = NULL;
3616 addrconf_del_dad_work(ifa);
3618 write_unlock_bh(&idev->lock);
3619 spin_lock_bh(&ifa->lock);
3621 if (keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3622 !addr_is_local(&ifa->addr)) {
3623 /* set state to skip the notifier below */
3624 state = INET6_IFADDR_STATE_DEAD;
3625 ifa->state = 0;
3626 if (!(ifa->flags & IFA_F_NODAD))
3627 ifa->flags |= IFA_F_TENTATIVE;
3629 rt = ifa->rt;
3630 ifa->rt = NULL;
3631 } else {
3632 state = ifa->state;
3633 ifa->state = INET6_IFADDR_STATE_DEAD;
3635 list_move(&ifa->if_list, &del_list);
3638 spin_unlock_bh(&ifa->lock);
3640 if (rt)
3641 ip6_del_rt(rt);
3643 if (state != INET6_IFADDR_STATE_DEAD) {
3644 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3645 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3646 } else {
3647 if (idev->cnf.forwarding)
3648 addrconf_leave_anycast(ifa);
3649 addrconf_leave_solict(ifa->idev, &ifa->addr);
3652 write_lock_bh(&idev->lock);
3655 write_unlock_bh(&idev->lock);
3657 /* now clean up addresses to be removed */
3658 while (!list_empty(&del_list)) {
3659 ifa = list_first_entry(&del_list,
3660 struct inet6_ifaddr, if_list);
3661 list_del(&ifa->if_list);
3663 in6_ifa_put(ifa);
3666 /* Step 5: Discard anycast and multicast list */
3667 if (how) {
3668 ipv6_ac_destroy_dev(idev);
3669 ipv6_mc_destroy_dev(idev);
3670 } else {
3671 ipv6_mc_down(idev);
3674 idev->tstamp = jiffies;
3676 /* Last: Shot the device (if unregistered) */
3677 if (how) {
3678 addrconf_sysctl_unregister(idev);
3679 neigh_parms_release(&nd_tbl, idev->nd_parms);
3680 neigh_ifdown(&nd_tbl, dev);
3681 in6_dev_put(idev);
3683 return 0;
3686 static void addrconf_rs_timer(unsigned long data)
3688 struct inet6_dev *idev = (struct inet6_dev *)data;
3689 struct net_device *dev = idev->dev;
3690 struct in6_addr lladdr;
3692 write_lock(&idev->lock);
3693 if (idev->dead || !(idev->if_flags & IF_READY))
3694 goto out;
3696 if (!ipv6_accept_ra(idev))
3697 goto out;
3699 /* Announcement received after solicitation was sent */
3700 if (idev->if_flags & IF_RA_RCVD)
3701 goto out;
3703 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3704 write_unlock(&idev->lock);
3705 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3706 ndisc_send_rs(dev, &lladdr,
3707 &in6addr_linklocal_allrouters);
3708 else
3709 goto put;
3711 write_lock(&idev->lock);
3712 idev->rs_interval = rfc3315_s14_backoff_update(
3713 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3714 /* The wait after the last probe can be shorter */
3715 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3716 idev->cnf.rtr_solicits) ?
3717 idev->cnf.rtr_solicit_delay :
3718 idev->rs_interval);
3719 } else {
3721 * Note: we do not support deprecated "all on-link"
3722 * assumption any longer.
3724 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3727 out:
3728 write_unlock(&idev->lock);
3729 put:
3730 in6_dev_put(idev);
3734 * Duplicate Address Detection
3736 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3738 unsigned long rand_num;
3739 struct inet6_dev *idev = ifp->idev;
3740 u64 nonce;
3742 if (ifp->flags & IFA_F_OPTIMISTIC)
3743 rand_num = 0;
3744 else
3745 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3747 nonce = 0;
3748 if (idev->cnf.enhanced_dad ||
3749 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3751 get_random_bytes(&nonce, 6);
3752 while (nonce == 0);
3754 ifp->dad_nonce = nonce;
3755 ifp->dad_probes = idev->cnf.dad_transmits;
3756 addrconf_mod_dad_work(ifp, rand_num);
3759 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3761 struct inet6_dev *idev = ifp->idev;
3762 struct net_device *dev = idev->dev;
3763 bool bump_id, notify = false;
3765 addrconf_join_solict(dev, &ifp->addr);
3767 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3769 read_lock_bh(&idev->lock);
3770 spin_lock(&ifp->lock);
3771 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3772 goto out;
3774 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3775 idev->cnf.accept_dad < 1 ||
3776 !(ifp->flags&IFA_F_TENTATIVE) ||
3777 ifp->flags & IFA_F_NODAD) {
3778 bump_id = ifp->flags & IFA_F_TENTATIVE;
3779 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3780 spin_unlock(&ifp->lock);
3781 read_unlock_bh(&idev->lock);
3783 addrconf_dad_completed(ifp, bump_id);
3784 return;
3787 if (!(idev->if_flags & IF_READY)) {
3788 spin_unlock(&ifp->lock);
3789 read_unlock_bh(&idev->lock);
3791 * If the device is not ready:
3792 * - keep it tentative if it is a permanent address.
3793 * - otherwise, kill it.
3795 in6_ifa_hold(ifp);
3796 addrconf_dad_stop(ifp, 0);
3797 return;
3801 * Optimistic nodes can start receiving
3802 * Frames right away
3804 if (ifp->flags & IFA_F_OPTIMISTIC) {
3805 ip6_ins_rt(ifp->rt);
3806 if (ipv6_use_optimistic_addr(idev)) {
3807 /* Because optimistic nodes can use this address,
3808 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3810 notify = true;
3814 addrconf_dad_kick(ifp);
3815 out:
3816 spin_unlock(&ifp->lock);
3817 read_unlock_bh(&idev->lock);
3818 if (notify)
3819 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3822 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3824 bool begin_dad = false;
3826 spin_lock_bh(&ifp->lock);
3827 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3828 ifp->state = INET6_IFADDR_STATE_PREDAD;
3829 begin_dad = true;
3831 spin_unlock_bh(&ifp->lock);
3833 if (begin_dad)
3834 addrconf_mod_dad_work(ifp, 0);
3837 static void addrconf_dad_work(struct work_struct *w)
3839 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3840 struct inet6_ifaddr,
3841 dad_work);
3842 struct inet6_dev *idev = ifp->idev;
3843 bool bump_id, disable_ipv6 = false;
3844 struct in6_addr mcaddr;
3846 enum {
3847 DAD_PROCESS,
3848 DAD_BEGIN,
3849 DAD_ABORT,
3850 } action = DAD_PROCESS;
3852 rtnl_lock();
3854 spin_lock_bh(&ifp->lock);
3855 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3856 action = DAD_BEGIN;
3857 ifp->state = INET6_IFADDR_STATE_DAD;
3858 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3859 action = DAD_ABORT;
3860 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3862 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3863 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3864 struct in6_addr addr;
3866 addr.s6_addr32[0] = htonl(0xfe800000);
3867 addr.s6_addr32[1] = 0;
3869 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3870 ipv6_addr_equal(&ifp->addr, &addr)) {
3871 /* DAD failed for link-local based on MAC */
3872 idev->cnf.disable_ipv6 = 1;
3874 pr_info("%s: IPv6 being disabled!\n",
3875 ifp->idev->dev->name);
3876 disable_ipv6 = true;
3880 spin_unlock_bh(&ifp->lock);
3882 if (action == DAD_BEGIN) {
3883 addrconf_dad_begin(ifp);
3884 goto out;
3885 } else if (action == DAD_ABORT) {
3886 in6_ifa_hold(ifp);
3887 addrconf_dad_stop(ifp, 1);
3888 if (disable_ipv6)
3889 addrconf_ifdown(idev->dev, 0);
3890 goto out;
3893 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3894 goto out;
3896 write_lock_bh(&idev->lock);
3897 if (idev->dead || !(idev->if_flags & IF_READY)) {
3898 write_unlock_bh(&idev->lock);
3899 goto out;
3902 spin_lock(&ifp->lock);
3903 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3904 spin_unlock(&ifp->lock);
3905 write_unlock_bh(&idev->lock);
3906 goto out;
3909 if (ifp->dad_probes == 0) {
3911 * DAD was successful
3914 bump_id = ifp->flags & IFA_F_TENTATIVE;
3915 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3916 spin_unlock(&ifp->lock);
3917 write_unlock_bh(&idev->lock);
3919 addrconf_dad_completed(ifp, bump_id);
3921 goto out;
3924 ifp->dad_probes--;
3925 addrconf_mod_dad_work(ifp,
3926 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3927 spin_unlock(&ifp->lock);
3928 write_unlock_bh(&idev->lock);
3930 /* send a neighbour solicitation for our addr */
3931 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3932 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3933 ifp->dad_nonce);
3934 out:
3935 in6_ifa_put(ifp);
3936 rtnl_unlock();
3939 /* ifp->idev must be at least read locked */
3940 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3942 struct inet6_ifaddr *ifpiter;
3943 struct inet6_dev *idev = ifp->idev;
3945 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3946 if (ifpiter->scope > IFA_LINK)
3947 break;
3948 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3949 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3950 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3951 IFA_F_PERMANENT)
3952 return false;
3954 return true;
3957 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
3959 struct net_device *dev = ifp->idev->dev;
3960 struct in6_addr lladdr;
3961 bool send_rs, send_mld;
3963 addrconf_del_dad_work(ifp);
3966 * Configure the address for reception. Now it is valid.
3969 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3971 /* If added prefix is link local and we are prepared to process
3972 router advertisements, start sending router solicitations.
3975 read_lock_bh(&ifp->idev->lock);
3976 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3977 send_rs = send_mld &&
3978 ipv6_accept_ra(ifp->idev) &&
3979 ifp->idev->cnf.rtr_solicits != 0 &&
3980 (dev->flags&IFF_LOOPBACK) == 0;
3981 read_unlock_bh(&ifp->idev->lock);
3983 /* While dad is in progress mld report's source address is in6_addrany.
3984 * Resend with proper ll now.
3986 if (send_mld)
3987 ipv6_mc_dad_complete(ifp->idev);
3989 if (send_rs) {
3991 * If a host as already performed a random delay
3992 * [...] as part of DAD [...] there is no need
3993 * to delay again before sending the first RS
3995 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3996 return;
3997 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3999 write_lock_bh(&ifp->idev->lock);
4000 spin_lock(&ifp->lock);
4001 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4002 ifp->idev->cnf.rtr_solicit_interval);
4003 ifp->idev->rs_probes = 1;
4004 ifp->idev->if_flags |= IF_RS_SENT;
4005 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4006 spin_unlock(&ifp->lock);
4007 write_unlock_bh(&ifp->idev->lock);
4010 if (bump_id)
4011 rt_genid_bump_ipv6(dev_net(dev));
4014 static void addrconf_dad_run(struct inet6_dev *idev)
4016 struct inet6_ifaddr *ifp;
4018 read_lock_bh(&idev->lock);
4019 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4020 spin_lock(&ifp->lock);
4021 if (ifp->flags & IFA_F_TENTATIVE &&
4022 ifp->state == INET6_IFADDR_STATE_DAD)
4023 addrconf_dad_kick(ifp);
4024 spin_unlock(&ifp->lock);
4026 read_unlock_bh(&idev->lock);
4029 #ifdef CONFIG_PROC_FS
4030 struct if6_iter_state {
4031 struct seq_net_private p;
4032 int bucket;
4033 int offset;
4036 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4038 struct inet6_ifaddr *ifa = NULL;
4039 struct if6_iter_state *state = seq->private;
4040 struct net *net = seq_file_net(seq);
4041 int p = 0;
4043 /* initial bucket if pos is 0 */
4044 if (pos == 0) {
4045 state->bucket = 0;
4046 state->offset = 0;
4049 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4050 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4051 addr_lst) {
4052 if (!net_eq(dev_net(ifa->idev->dev), net))
4053 continue;
4054 /* sync with offset */
4055 if (p < state->offset) {
4056 p++;
4057 continue;
4059 state->offset++;
4060 return ifa;
4063 /* prepare for next bucket */
4064 state->offset = 0;
4065 p = 0;
4067 return NULL;
4070 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4071 struct inet6_ifaddr *ifa)
4073 struct if6_iter_state *state = seq->private;
4074 struct net *net = seq_file_net(seq);
4076 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4077 if (!net_eq(dev_net(ifa->idev->dev), net))
4078 continue;
4079 state->offset++;
4080 return ifa;
4083 while (++state->bucket < IN6_ADDR_HSIZE) {
4084 state->offset = 0;
4085 hlist_for_each_entry_rcu_bh(ifa,
4086 &inet6_addr_lst[state->bucket], addr_lst) {
4087 if (!net_eq(dev_net(ifa->idev->dev), net))
4088 continue;
4089 state->offset++;
4090 return ifa;
4094 return NULL;
4097 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4098 __acquires(rcu_bh)
4100 rcu_read_lock_bh();
4101 return if6_get_first(seq, *pos);
4104 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4106 struct inet6_ifaddr *ifa;
4108 ifa = if6_get_next(seq, v);
4109 ++*pos;
4110 return ifa;
4113 static void if6_seq_stop(struct seq_file *seq, void *v)
4114 __releases(rcu_bh)
4116 rcu_read_unlock_bh();
4119 static int if6_seq_show(struct seq_file *seq, void *v)
4121 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4122 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4123 &ifp->addr,
4124 ifp->idev->dev->ifindex,
4125 ifp->prefix_len,
4126 ifp->scope,
4127 (u8) ifp->flags,
4128 ifp->idev->dev->name);
4129 return 0;
4132 static const struct seq_operations if6_seq_ops = {
4133 .start = if6_seq_start,
4134 .next = if6_seq_next,
4135 .show = if6_seq_show,
4136 .stop = if6_seq_stop,
4139 static int if6_seq_open(struct inode *inode, struct file *file)
4141 return seq_open_net(inode, file, &if6_seq_ops,
4142 sizeof(struct if6_iter_state));
4145 static const struct file_operations if6_fops = {
4146 .owner = THIS_MODULE,
4147 .open = if6_seq_open,
4148 .read = seq_read,
4149 .llseek = seq_lseek,
4150 .release = seq_release_net,
4153 static int __net_init if6_proc_net_init(struct net *net)
4155 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4156 return -ENOMEM;
4157 return 0;
4160 static void __net_exit if6_proc_net_exit(struct net *net)
4162 remove_proc_entry("if_inet6", net->proc_net);
4165 static struct pernet_operations if6_proc_net_ops = {
4166 .init = if6_proc_net_init,
4167 .exit = if6_proc_net_exit,
4170 int __init if6_proc_init(void)
4172 return register_pernet_subsys(&if6_proc_net_ops);
4175 void if6_proc_exit(void)
4177 unregister_pernet_subsys(&if6_proc_net_ops);
4179 #endif /* CONFIG_PROC_FS */
4181 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4182 /* Check if address is a home address configured on any interface. */
4183 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4185 int ret = 0;
4186 struct inet6_ifaddr *ifp = NULL;
4187 unsigned int hash = inet6_addr_hash(addr);
4189 rcu_read_lock_bh();
4190 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4191 if (!net_eq(dev_net(ifp->idev->dev), net))
4192 continue;
4193 if (ipv6_addr_equal(&ifp->addr, addr) &&
4194 (ifp->flags & IFA_F_HOMEADDRESS)) {
4195 ret = 1;
4196 break;
4199 rcu_read_unlock_bh();
4200 return ret;
4202 #endif
4205 * Periodic address status verification
4208 static void addrconf_verify_rtnl(void)
4210 unsigned long now, next, next_sec, next_sched;
4211 struct inet6_ifaddr *ifp;
4212 int i;
4214 ASSERT_RTNL();
4216 rcu_read_lock_bh();
4217 now = jiffies;
4218 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4220 cancel_delayed_work(&addr_chk_work);
4222 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4223 restart:
4224 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4225 unsigned long age;
4227 /* When setting preferred_lft to a value not zero or
4228 * infinity, while valid_lft is infinity
4229 * IFA_F_PERMANENT has a non-infinity life time.
4231 if ((ifp->flags & IFA_F_PERMANENT) &&
4232 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4233 continue;
4235 spin_lock(&ifp->lock);
4236 /* We try to batch several events at once. */
4237 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4239 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4240 age >= ifp->valid_lft) {
4241 spin_unlock(&ifp->lock);
4242 in6_ifa_hold(ifp);
4243 ipv6_del_addr(ifp);
4244 goto restart;
4245 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4246 spin_unlock(&ifp->lock);
4247 continue;
4248 } else if (age >= ifp->prefered_lft) {
4249 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4250 int deprecate = 0;
4252 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4253 deprecate = 1;
4254 ifp->flags |= IFA_F_DEPRECATED;
4257 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4258 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4259 next = ifp->tstamp + ifp->valid_lft * HZ;
4261 spin_unlock(&ifp->lock);
4263 if (deprecate) {
4264 in6_ifa_hold(ifp);
4266 ipv6_ifa_notify(0, ifp);
4267 in6_ifa_put(ifp);
4268 goto restart;
4270 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4271 !(ifp->flags&IFA_F_TENTATIVE)) {
4272 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4273 ifp->idev->cnf.dad_transmits *
4274 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4276 if (age >= ifp->prefered_lft - regen_advance) {
4277 struct inet6_ifaddr *ifpub = ifp->ifpub;
4278 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4279 next = ifp->tstamp + ifp->prefered_lft * HZ;
4280 if (!ifp->regen_count && ifpub) {
4281 ifp->regen_count++;
4282 in6_ifa_hold(ifp);
4283 in6_ifa_hold(ifpub);
4284 spin_unlock(&ifp->lock);
4286 spin_lock(&ifpub->lock);
4287 ifpub->regen_count = 0;
4288 spin_unlock(&ifpub->lock);
4289 ipv6_create_tempaddr(ifpub, ifp);
4290 in6_ifa_put(ifpub);
4291 in6_ifa_put(ifp);
4292 goto restart;
4294 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4295 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4296 spin_unlock(&ifp->lock);
4297 } else {
4298 /* ifp->prefered_lft <= ifp->valid_lft */
4299 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4300 next = ifp->tstamp + ifp->prefered_lft * HZ;
4301 spin_unlock(&ifp->lock);
4306 next_sec = round_jiffies_up(next);
4307 next_sched = next;
4309 /* If rounded timeout is accurate enough, accept it. */
4310 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4311 next_sched = next_sec;
4313 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4314 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4315 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4317 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4318 now, next, next_sec, next_sched);
4319 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4320 rcu_read_unlock_bh();
4323 static void addrconf_verify_work(struct work_struct *w)
4325 rtnl_lock();
4326 addrconf_verify_rtnl();
4327 rtnl_unlock();
4330 static void addrconf_verify(void)
4332 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4335 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4336 struct in6_addr **peer_pfx)
4338 struct in6_addr *pfx = NULL;
4340 *peer_pfx = NULL;
4342 if (addr)
4343 pfx = nla_data(addr);
4345 if (local) {
4346 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4347 *peer_pfx = pfx;
4348 pfx = nla_data(local);
4351 return pfx;
4354 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4355 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4356 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4357 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4358 [IFA_FLAGS] = { .len = sizeof(u32) },
4361 static int
4362 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
4364 struct net *net = sock_net(skb->sk);
4365 struct ifaddrmsg *ifm;
4366 struct nlattr *tb[IFA_MAX+1];
4367 struct in6_addr *pfx, *peer_pfx;
4368 u32 ifa_flags;
4369 int err;
4371 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4372 if (err < 0)
4373 return err;
4375 ifm = nlmsg_data(nlh);
4376 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4377 if (!pfx)
4378 return -EINVAL;
4380 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4382 /* We ignore other flags so far. */
4383 ifa_flags &= IFA_F_MANAGETEMPADDR;
4385 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4386 ifm->ifa_prefixlen);
4389 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4390 u32 prefered_lft, u32 valid_lft)
4392 u32 flags;
4393 clock_t expires;
4394 unsigned long timeout;
4395 bool was_managetempaddr;
4396 bool had_prefixroute;
4398 ASSERT_RTNL();
4400 if (!valid_lft || (prefered_lft > valid_lft))
4401 return -EINVAL;
4403 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4404 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4405 return -EINVAL;
4407 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4408 if (addrconf_finite_timeout(timeout)) {
4409 expires = jiffies_to_clock_t(timeout * HZ);
4410 valid_lft = timeout;
4411 flags = RTF_EXPIRES;
4412 } else {
4413 expires = 0;
4414 flags = 0;
4415 ifa_flags |= IFA_F_PERMANENT;
4418 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4419 if (addrconf_finite_timeout(timeout)) {
4420 if (timeout == 0)
4421 ifa_flags |= IFA_F_DEPRECATED;
4422 prefered_lft = timeout;
4425 spin_lock_bh(&ifp->lock);
4426 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4427 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4428 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4429 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4430 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4431 IFA_F_NOPREFIXROUTE);
4432 ifp->flags |= ifa_flags;
4433 ifp->tstamp = jiffies;
4434 ifp->valid_lft = valid_lft;
4435 ifp->prefered_lft = prefered_lft;
4437 spin_unlock_bh(&ifp->lock);
4438 if (!(ifp->flags&IFA_F_TENTATIVE))
4439 ipv6_ifa_notify(0, ifp);
4441 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4442 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4443 expires, flags);
4444 } else if (had_prefixroute) {
4445 enum cleanup_prefix_rt_t action;
4446 unsigned long rt_expires;
4448 write_lock_bh(&ifp->idev->lock);
4449 action = check_cleanup_prefix_route(ifp, &rt_expires);
4450 write_unlock_bh(&ifp->idev->lock);
4452 if (action != CLEANUP_PREFIX_RT_NOP) {
4453 cleanup_prefix_route(ifp, rt_expires,
4454 action == CLEANUP_PREFIX_RT_DEL);
4458 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4459 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4460 valid_lft = prefered_lft = 0;
4461 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4462 !was_managetempaddr, jiffies);
4465 addrconf_verify_rtnl();
4467 return 0;
4470 static int
4471 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
4473 struct net *net = sock_net(skb->sk);
4474 struct ifaddrmsg *ifm;
4475 struct nlattr *tb[IFA_MAX+1];
4476 struct in6_addr *pfx, *peer_pfx;
4477 struct inet6_ifaddr *ifa;
4478 struct net_device *dev;
4479 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4480 u32 ifa_flags;
4481 int err;
4483 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4484 if (err < 0)
4485 return err;
4487 ifm = nlmsg_data(nlh);
4488 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4489 if (!pfx)
4490 return -EINVAL;
4492 if (tb[IFA_CACHEINFO]) {
4493 struct ifa_cacheinfo *ci;
4495 ci = nla_data(tb[IFA_CACHEINFO]);
4496 valid_lft = ci->ifa_valid;
4497 preferred_lft = ci->ifa_prefered;
4498 } else {
4499 preferred_lft = INFINITY_LIFE_TIME;
4500 valid_lft = INFINITY_LIFE_TIME;
4503 dev = __dev_get_by_index(net, ifm->ifa_index);
4504 if (!dev)
4505 return -ENODEV;
4507 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4509 /* We ignore other flags so far. */
4510 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4511 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4513 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4514 if (!ifa) {
4516 * It would be best to check for !NLM_F_CREATE here but
4517 * userspace already relies on not having to provide this.
4519 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4520 ifm->ifa_prefixlen, ifa_flags,
4521 preferred_lft, valid_lft);
4524 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4525 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4526 err = -EEXIST;
4527 else
4528 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4530 in6_ifa_put(ifa);
4532 return err;
4535 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4536 u8 scope, int ifindex)
4538 struct ifaddrmsg *ifm;
4540 ifm = nlmsg_data(nlh);
4541 ifm->ifa_family = AF_INET6;
4542 ifm->ifa_prefixlen = prefixlen;
4543 ifm->ifa_flags = flags;
4544 ifm->ifa_scope = scope;
4545 ifm->ifa_index = ifindex;
4548 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4549 unsigned long tstamp, u32 preferred, u32 valid)
4551 struct ifa_cacheinfo ci;
4553 ci.cstamp = cstamp_delta(cstamp);
4554 ci.tstamp = cstamp_delta(tstamp);
4555 ci.ifa_prefered = preferred;
4556 ci.ifa_valid = valid;
4558 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4561 static inline int rt_scope(int ifa_scope)
4563 if (ifa_scope & IFA_HOST)
4564 return RT_SCOPE_HOST;
4565 else if (ifa_scope & IFA_LINK)
4566 return RT_SCOPE_LINK;
4567 else if (ifa_scope & IFA_SITE)
4568 return RT_SCOPE_SITE;
4569 else
4570 return RT_SCOPE_UNIVERSE;
4573 static inline int inet6_ifaddr_msgsize(void)
4575 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4576 + nla_total_size(16) /* IFA_LOCAL */
4577 + nla_total_size(16) /* IFA_ADDRESS */
4578 + nla_total_size(sizeof(struct ifa_cacheinfo))
4579 + nla_total_size(4) /* IFA_FLAGS */;
4582 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4583 u32 portid, u32 seq, int event, unsigned int flags)
4585 struct nlmsghdr *nlh;
4586 u32 preferred, valid;
4588 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4589 if (!nlh)
4590 return -EMSGSIZE;
4592 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4593 ifa->idev->dev->ifindex);
4595 if (!((ifa->flags&IFA_F_PERMANENT) &&
4596 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4597 preferred = ifa->prefered_lft;
4598 valid = ifa->valid_lft;
4599 if (preferred != INFINITY_LIFE_TIME) {
4600 long tval = (jiffies - ifa->tstamp)/HZ;
4601 if (preferred > tval)
4602 preferred -= tval;
4603 else
4604 preferred = 0;
4605 if (valid != INFINITY_LIFE_TIME) {
4606 if (valid > tval)
4607 valid -= tval;
4608 else
4609 valid = 0;
4612 } else {
4613 preferred = INFINITY_LIFE_TIME;
4614 valid = INFINITY_LIFE_TIME;
4617 if (!ipv6_addr_any(&ifa->peer_addr)) {
4618 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4619 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4620 goto error;
4621 } else
4622 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4623 goto error;
4625 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4626 goto error;
4628 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4629 goto error;
4631 nlmsg_end(skb, nlh);
4632 return 0;
4634 error:
4635 nlmsg_cancel(skb, nlh);
4636 return -EMSGSIZE;
4639 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4640 u32 portid, u32 seq, int event, u16 flags)
4642 struct nlmsghdr *nlh;
4643 u8 scope = RT_SCOPE_UNIVERSE;
4644 int ifindex = ifmca->idev->dev->ifindex;
4646 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4647 scope = RT_SCOPE_SITE;
4649 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4650 if (!nlh)
4651 return -EMSGSIZE;
4653 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4654 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4655 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4656 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4657 nlmsg_cancel(skb, nlh);
4658 return -EMSGSIZE;
4661 nlmsg_end(skb, nlh);
4662 return 0;
4665 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4666 u32 portid, u32 seq, int event, unsigned int flags)
4668 struct nlmsghdr *nlh;
4669 u8 scope = RT_SCOPE_UNIVERSE;
4670 int ifindex = ifaca->aca_idev->dev->ifindex;
4672 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4673 scope = RT_SCOPE_SITE;
4675 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4676 if (!nlh)
4677 return -EMSGSIZE;
4679 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4680 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4681 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4682 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4683 nlmsg_cancel(skb, nlh);
4684 return -EMSGSIZE;
4687 nlmsg_end(skb, nlh);
4688 return 0;
4691 enum addr_type_t {
4692 UNICAST_ADDR,
4693 MULTICAST_ADDR,
4694 ANYCAST_ADDR,
4697 /* called with rcu_read_lock() */
4698 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4699 struct netlink_callback *cb, enum addr_type_t type,
4700 int s_ip_idx, int *p_ip_idx)
4702 struct ifmcaddr6 *ifmca;
4703 struct ifacaddr6 *ifaca;
4704 int err = 1;
4705 int ip_idx = *p_ip_idx;
4707 read_lock_bh(&idev->lock);
4708 switch (type) {
4709 case UNICAST_ADDR: {
4710 struct inet6_ifaddr *ifa;
4712 /* unicast address incl. temp addr */
4713 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4714 if (++ip_idx < s_ip_idx)
4715 continue;
4716 err = inet6_fill_ifaddr(skb, ifa,
4717 NETLINK_CB(cb->skb).portid,
4718 cb->nlh->nlmsg_seq,
4719 RTM_NEWADDR,
4720 NLM_F_MULTI);
4721 if (err < 0)
4722 break;
4723 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4725 break;
4727 case MULTICAST_ADDR:
4728 /* multicast address */
4729 for (ifmca = idev->mc_list; ifmca;
4730 ifmca = ifmca->next, ip_idx++) {
4731 if (ip_idx < s_ip_idx)
4732 continue;
4733 err = inet6_fill_ifmcaddr(skb, ifmca,
4734 NETLINK_CB(cb->skb).portid,
4735 cb->nlh->nlmsg_seq,
4736 RTM_GETMULTICAST,
4737 NLM_F_MULTI);
4738 if (err < 0)
4739 break;
4741 break;
4742 case ANYCAST_ADDR:
4743 /* anycast address */
4744 for (ifaca = idev->ac_list; ifaca;
4745 ifaca = ifaca->aca_next, ip_idx++) {
4746 if (ip_idx < s_ip_idx)
4747 continue;
4748 err = inet6_fill_ifacaddr(skb, ifaca,
4749 NETLINK_CB(cb->skb).portid,
4750 cb->nlh->nlmsg_seq,
4751 RTM_GETANYCAST,
4752 NLM_F_MULTI);
4753 if (err < 0)
4754 break;
4756 break;
4757 default:
4758 break;
4760 read_unlock_bh(&idev->lock);
4761 *p_ip_idx = ip_idx;
4762 return err;
4765 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4766 enum addr_type_t type)
4768 struct net *net = sock_net(skb->sk);
4769 int h, s_h;
4770 int idx, ip_idx;
4771 int s_idx, s_ip_idx;
4772 struct net_device *dev;
4773 struct inet6_dev *idev;
4774 struct hlist_head *head;
4776 s_h = cb->args[0];
4777 s_idx = idx = cb->args[1];
4778 s_ip_idx = ip_idx = cb->args[2];
4780 rcu_read_lock();
4781 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4782 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4783 idx = 0;
4784 head = &net->dev_index_head[h];
4785 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4786 if (idx < s_idx)
4787 goto cont;
4788 if (h > s_h || idx > s_idx)
4789 s_ip_idx = 0;
4790 ip_idx = 0;
4791 idev = __in6_dev_get(dev);
4792 if (!idev)
4793 goto cont;
4795 if (in6_dump_addrs(idev, skb, cb, type,
4796 s_ip_idx, &ip_idx) < 0)
4797 goto done;
4798 cont:
4799 idx++;
4802 done:
4803 rcu_read_unlock();
4804 cb->args[0] = h;
4805 cb->args[1] = idx;
4806 cb->args[2] = ip_idx;
4808 return skb->len;
4811 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4813 enum addr_type_t type = UNICAST_ADDR;
4815 return inet6_dump_addr(skb, cb, type);
4818 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4820 enum addr_type_t type = MULTICAST_ADDR;
4822 return inet6_dump_addr(skb, cb, type);
4826 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4828 enum addr_type_t type = ANYCAST_ADDR;
4830 return inet6_dump_addr(skb, cb, type);
4833 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4835 struct net *net = sock_net(in_skb->sk);
4836 struct ifaddrmsg *ifm;
4837 struct nlattr *tb[IFA_MAX+1];
4838 struct in6_addr *addr = NULL, *peer;
4839 struct net_device *dev = NULL;
4840 struct inet6_ifaddr *ifa;
4841 struct sk_buff *skb;
4842 int err;
4844 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4845 if (err < 0)
4846 goto errout;
4848 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4849 if (!addr) {
4850 err = -EINVAL;
4851 goto errout;
4854 ifm = nlmsg_data(nlh);
4855 if (ifm->ifa_index)
4856 dev = __dev_get_by_index(net, ifm->ifa_index);
4858 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4859 if (!ifa) {
4860 err = -EADDRNOTAVAIL;
4861 goto errout;
4864 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4865 if (!skb) {
4866 err = -ENOBUFS;
4867 goto errout_ifa;
4870 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4871 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4872 if (err < 0) {
4873 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4874 WARN_ON(err == -EMSGSIZE);
4875 kfree_skb(skb);
4876 goto errout_ifa;
4878 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4879 errout_ifa:
4880 in6_ifa_put(ifa);
4881 errout:
4882 return err;
4885 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4887 struct sk_buff *skb;
4888 struct net *net = dev_net(ifa->idev->dev);
4889 int err = -ENOBUFS;
4891 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4892 if (!skb)
4893 goto errout;
4895 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4896 if (err < 0) {
4897 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4898 WARN_ON(err == -EMSGSIZE);
4899 kfree_skb(skb);
4900 goto errout;
4902 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4903 return;
4904 errout:
4905 if (err < 0)
4906 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4909 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4910 __s32 *array, int bytes)
4912 BUG_ON(bytes < (DEVCONF_MAX * 4));
4914 memset(array, 0, bytes);
4915 array[DEVCONF_FORWARDING] = cnf->forwarding;
4916 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4917 array[DEVCONF_MTU6] = cnf->mtu6;
4918 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4919 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4920 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4921 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4922 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4923 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4924 jiffies_to_msecs(cnf->rtr_solicit_interval);
4925 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
4926 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
4927 array[DEVCONF_RTR_SOLICIT_DELAY] =
4928 jiffies_to_msecs(cnf->rtr_solicit_delay);
4929 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4930 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4931 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4932 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4933 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4934 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4935 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4936 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4937 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4938 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4939 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4940 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4941 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
4942 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4943 #ifdef CONFIG_IPV6_ROUTER_PREF
4944 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4945 array[DEVCONF_RTR_PROBE_INTERVAL] =
4946 jiffies_to_msecs(cnf->rtr_probe_interval);
4947 #ifdef CONFIG_IPV6_ROUTE_INFO
4948 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4949 #endif
4950 #endif
4951 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4952 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4953 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4954 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4955 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
4956 #endif
4957 #ifdef CONFIG_IPV6_MROUTE
4958 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4959 #endif
4960 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4961 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4962 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4963 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4964 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4965 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4966 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
4967 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
4968 /* we omit DEVCONF_STABLE_SECRET for now */
4969 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
4970 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
4971 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
4972 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
4973 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
4974 #ifdef CONFIG_IPV6_SEG6_HMAC
4975 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
4976 #endif
4977 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
4980 static inline size_t inet6_ifla6_size(void)
4982 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4983 + nla_total_size(sizeof(struct ifla_cacheinfo))
4984 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4985 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4986 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4987 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4990 static inline size_t inet6_if_nlmsg_size(void)
4992 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4993 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4994 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4995 + nla_total_size(4) /* IFLA_MTU */
4996 + nla_total_size(4) /* IFLA_LINK */
4997 + nla_total_size(1) /* IFLA_OPERSTATE */
4998 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5001 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5002 int bytes)
5004 int i;
5005 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5006 BUG_ON(pad < 0);
5008 /* Use put_unaligned() because stats may not be aligned for u64. */
5009 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5010 for (i = 1; i < ICMP6_MIB_MAX; i++)
5011 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5013 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5016 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5017 int bytes, size_t syncpoff)
5019 int i, c;
5020 u64 buff[IPSTATS_MIB_MAX];
5021 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5023 BUG_ON(pad < 0);
5025 memset(buff, 0, sizeof(buff));
5026 buff[0] = IPSTATS_MIB_MAX;
5028 for_each_possible_cpu(c) {
5029 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5030 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5033 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5034 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5037 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5038 int bytes)
5040 switch (attrtype) {
5041 case IFLA_INET6_STATS:
5042 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5043 offsetof(struct ipstats_mib, syncp));
5044 break;
5045 case IFLA_INET6_ICMP6STATS:
5046 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5047 break;
5051 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5052 u32 ext_filter_mask)
5054 struct nlattr *nla;
5055 struct ifla_cacheinfo ci;
5057 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5058 goto nla_put_failure;
5059 ci.max_reasm_len = IPV6_MAXPLEN;
5060 ci.tstamp = cstamp_delta(idev->tstamp);
5061 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5062 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5063 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5064 goto nla_put_failure;
5065 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5066 if (!nla)
5067 goto nla_put_failure;
5068 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5070 /* XXX - MC not implemented */
5072 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5073 return 0;
5075 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5076 if (!nla)
5077 goto nla_put_failure;
5078 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5080 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5081 if (!nla)
5082 goto nla_put_failure;
5083 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5085 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5086 if (!nla)
5087 goto nla_put_failure;
5089 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
5090 goto nla_put_failure;
5092 read_lock_bh(&idev->lock);
5093 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5094 read_unlock_bh(&idev->lock);
5096 return 0;
5098 nla_put_failure:
5099 return -EMSGSIZE;
5102 static size_t inet6_get_link_af_size(const struct net_device *dev,
5103 u32 ext_filter_mask)
5105 if (!__in6_dev_get(dev))
5106 return 0;
5108 return inet6_ifla6_size();
5111 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5112 u32 ext_filter_mask)
5114 struct inet6_dev *idev = __in6_dev_get(dev);
5116 if (!idev)
5117 return -ENODATA;
5119 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5120 return -EMSGSIZE;
5122 return 0;
5125 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5127 struct inet6_ifaddr *ifp;
5128 struct net_device *dev = idev->dev;
5129 bool clear_token, update_rs = false;
5130 struct in6_addr ll_addr;
5132 ASSERT_RTNL();
5134 if (!token)
5135 return -EINVAL;
5136 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5137 return -EINVAL;
5138 if (!ipv6_accept_ra(idev))
5139 return -EINVAL;
5140 if (idev->cnf.rtr_solicits == 0)
5141 return -EINVAL;
5143 write_lock_bh(&idev->lock);
5145 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5146 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5148 write_unlock_bh(&idev->lock);
5150 clear_token = ipv6_addr_any(token);
5151 if (clear_token)
5152 goto update_lft;
5154 if (!idev->dead && (idev->if_flags & IF_READY) &&
5155 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5156 IFA_F_OPTIMISTIC)) {
5157 /* If we're not ready, then normal ifup will take care
5158 * of this. Otherwise, we need to request our rs here.
5160 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5161 update_rs = true;
5164 update_lft:
5165 write_lock_bh(&idev->lock);
5167 if (update_rs) {
5168 idev->if_flags |= IF_RS_SENT;
5169 idev->rs_interval = rfc3315_s14_backoff_init(
5170 idev->cnf.rtr_solicit_interval);
5171 idev->rs_probes = 1;
5172 addrconf_mod_rs_timer(idev, idev->rs_interval);
5175 /* Well, that's kinda nasty ... */
5176 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5177 spin_lock(&ifp->lock);
5178 if (ifp->tokenized) {
5179 ifp->valid_lft = 0;
5180 ifp->prefered_lft = 0;
5182 spin_unlock(&ifp->lock);
5185 write_unlock_bh(&idev->lock);
5186 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5187 addrconf_verify_rtnl();
5188 return 0;
5191 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5192 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5193 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5196 static int inet6_validate_link_af(const struct net_device *dev,
5197 const struct nlattr *nla)
5199 struct nlattr *tb[IFLA_INET6_MAX + 1];
5201 if (dev && !__in6_dev_get(dev))
5202 return -EAFNOSUPPORT;
5204 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy);
5207 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5209 int err = -EINVAL;
5210 struct inet6_dev *idev = __in6_dev_get(dev);
5211 struct nlattr *tb[IFLA_INET6_MAX + 1];
5213 if (!idev)
5214 return -EAFNOSUPPORT;
5216 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
5217 BUG();
5219 if (tb[IFLA_INET6_TOKEN]) {
5220 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5221 if (err)
5222 return err;
5225 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5226 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5228 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5229 mode != IN6_ADDR_GEN_MODE_NONE &&
5230 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5231 mode != IN6_ADDR_GEN_MODE_RANDOM)
5232 return -EINVAL;
5234 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5235 !idev->cnf.stable_secret.initialized &&
5236 !dev_net(dev)->ipv6.devconf_dflt->stable_secret.initialized)
5237 return -EINVAL;
5239 idev->addr_gen_mode = mode;
5240 err = 0;
5243 return err;
5246 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5247 u32 portid, u32 seq, int event, unsigned int flags)
5249 struct net_device *dev = idev->dev;
5250 struct ifinfomsg *hdr;
5251 struct nlmsghdr *nlh;
5252 void *protoinfo;
5254 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5255 if (!nlh)
5256 return -EMSGSIZE;
5258 hdr = nlmsg_data(nlh);
5259 hdr->ifi_family = AF_INET6;
5260 hdr->__ifi_pad = 0;
5261 hdr->ifi_type = dev->type;
5262 hdr->ifi_index = dev->ifindex;
5263 hdr->ifi_flags = dev_get_flags(dev);
5264 hdr->ifi_change = 0;
5266 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5267 (dev->addr_len &&
5268 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5269 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5270 (dev->ifindex != dev_get_iflink(dev) &&
5271 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5272 nla_put_u8(skb, IFLA_OPERSTATE,
5273 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5274 goto nla_put_failure;
5275 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5276 if (!protoinfo)
5277 goto nla_put_failure;
5279 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5280 goto nla_put_failure;
5282 nla_nest_end(skb, protoinfo);
5283 nlmsg_end(skb, nlh);
5284 return 0;
5286 nla_put_failure:
5287 nlmsg_cancel(skb, nlh);
5288 return -EMSGSIZE;
5291 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5293 struct net *net = sock_net(skb->sk);
5294 int h, s_h;
5295 int idx = 0, s_idx;
5296 struct net_device *dev;
5297 struct inet6_dev *idev;
5298 struct hlist_head *head;
5300 s_h = cb->args[0];
5301 s_idx = cb->args[1];
5303 rcu_read_lock();
5304 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5305 idx = 0;
5306 head = &net->dev_index_head[h];
5307 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5308 if (idx < s_idx)
5309 goto cont;
5310 idev = __in6_dev_get(dev);
5311 if (!idev)
5312 goto cont;
5313 if (inet6_fill_ifinfo(skb, idev,
5314 NETLINK_CB(cb->skb).portid,
5315 cb->nlh->nlmsg_seq,
5316 RTM_NEWLINK, NLM_F_MULTI) < 0)
5317 goto out;
5318 cont:
5319 idx++;
5322 out:
5323 rcu_read_unlock();
5324 cb->args[1] = idx;
5325 cb->args[0] = h;
5327 return skb->len;
5330 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5332 struct sk_buff *skb;
5333 struct net *net = dev_net(idev->dev);
5334 int err = -ENOBUFS;
5336 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5337 if (!skb)
5338 goto errout;
5340 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5341 if (err < 0) {
5342 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5343 WARN_ON(err == -EMSGSIZE);
5344 kfree_skb(skb);
5345 goto errout;
5347 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5348 return;
5349 errout:
5350 if (err < 0)
5351 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5354 static inline size_t inet6_prefix_nlmsg_size(void)
5356 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5357 + nla_total_size(sizeof(struct in6_addr))
5358 + nla_total_size(sizeof(struct prefix_cacheinfo));
5361 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5362 struct prefix_info *pinfo, u32 portid, u32 seq,
5363 int event, unsigned int flags)
5365 struct prefixmsg *pmsg;
5366 struct nlmsghdr *nlh;
5367 struct prefix_cacheinfo ci;
5369 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5370 if (!nlh)
5371 return -EMSGSIZE;
5373 pmsg = nlmsg_data(nlh);
5374 pmsg->prefix_family = AF_INET6;
5375 pmsg->prefix_pad1 = 0;
5376 pmsg->prefix_pad2 = 0;
5377 pmsg->prefix_ifindex = idev->dev->ifindex;
5378 pmsg->prefix_len = pinfo->prefix_len;
5379 pmsg->prefix_type = pinfo->type;
5380 pmsg->prefix_pad3 = 0;
5381 pmsg->prefix_flags = 0;
5382 if (pinfo->onlink)
5383 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5384 if (pinfo->autoconf)
5385 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5387 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5388 goto nla_put_failure;
5389 ci.preferred_time = ntohl(pinfo->prefered);
5390 ci.valid_time = ntohl(pinfo->valid);
5391 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5392 goto nla_put_failure;
5393 nlmsg_end(skb, nlh);
5394 return 0;
5396 nla_put_failure:
5397 nlmsg_cancel(skb, nlh);
5398 return -EMSGSIZE;
5401 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5402 struct prefix_info *pinfo)
5404 struct sk_buff *skb;
5405 struct net *net = dev_net(idev->dev);
5406 int err = -ENOBUFS;
5408 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5409 if (!skb)
5410 goto errout;
5412 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5413 if (err < 0) {
5414 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5415 WARN_ON(err == -EMSGSIZE);
5416 kfree_skb(skb);
5417 goto errout;
5419 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5420 return;
5421 errout:
5422 if (err < 0)
5423 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5426 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5428 struct net *net = dev_net(ifp->idev->dev);
5430 if (event)
5431 ASSERT_RTNL();
5433 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5435 switch (event) {
5436 case RTM_NEWADDR:
5438 * If the address was optimistic
5439 * we inserted the route at the start of
5440 * our DAD process, so we don't need
5441 * to do it again
5443 if (!(ifp->rt->rt6i_node))
5444 ip6_ins_rt(ifp->rt);
5445 if (ifp->idev->cnf.forwarding)
5446 addrconf_join_anycast(ifp);
5447 if (!ipv6_addr_any(&ifp->peer_addr))
5448 addrconf_prefix_route(&ifp->peer_addr, 128,
5449 ifp->idev->dev, 0, 0);
5450 break;
5451 case RTM_DELADDR:
5452 if (ifp->idev->cnf.forwarding)
5453 addrconf_leave_anycast(ifp);
5454 addrconf_leave_solict(ifp->idev, &ifp->addr);
5455 if (!ipv6_addr_any(&ifp->peer_addr)) {
5456 struct rt6_info *rt;
5458 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5459 ifp->idev->dev, 0, 0);
5460 if (rt)
5461 ip6_del_rt(rt);
5463 if (ifp->rt) {
5464 dst_hold(&ifp->rt->dst);
5465 ip6_del_rt(ifp->rt);
5467 rt_genid_bump_ipv6(net);
5468 break;
5470 atomic_inc(&net->ipv6.dev_addr_genid);
5473 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5475 rcu_read_lock_bh();
5476 if (likely(ifp->idev->dead == 0))
5477 __ipv6_ifa_notify(event, ifp);
5478 rcu_read_unlock_bh();
5481 #ifdef CONFIG_SYSCTL
5483 static
5484 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
5485 void __user *buffer, size_t *lenp, loff_t *ppos)
5487 int *valp = ctl->data;
5488 int val = *valp;
5489 loff_t pos = *ppos;
5490 struct ctl_table lctl;
5491 int ret;
5494 * ctl->data points to idev->cnf.forwarding, we should
5495 * not modify it until we get the rtnl lock.
5497 lctl = *ctl;
5498 lctl.data = &val;
5500 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5502 if (write)
5503 ret = addrconf_fixup_forwarding(ctl, valp, val);
5504 if (ret)
5505 *ppos = pos;
5506 return ret;
5509 static
5510 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5511 void __user *buffer, size_t *lenp, loff_t *ppos)
5513 struct inet6_dev *idev = ctl->extra1;
5514 int min_mtu = IPV6_MIN_MTU;
5515 struct ctl_table lctl;
5517 lctl = *ctl;
5518 lctl.extra1 = &min_mtu;
5519 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5521 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5524 static void dev_disable_change(struct inet6_dev *idev)
5526 struct netdev_notifier_info info;
5528 if (!idev || !idev->dev)
5529 return;
5531 netdev_notifier_info_init(&info, idev->dev);
5532 if (idev->cnf.disable_ipv6)
5533 addrconf_notify(NULL, NETDEV_DOWN, &info);
5534 else
5535 addrconf_notify(NULL, NETDEV_UP, &info);
5538 static void addrconf_disable_change(struct net *net, __s32 newf)
5540 struct net_device *dev;
5541 struct inet6_dev *idev;
5543 rcu_read_lock();
5544 for_each_netdev_rcu(net, dev) {
5545 idev = __in6_dev_get(dev);
5546 if (idev) {
5547 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5548 idev->cnf.disable_ipv6 = newf;
5549 if (changed)
5550 dev_disable_change(idev);
5553 rcu_read_unlock();
5556 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5558 struct net *net;
5559 int old;
5561 if (!rtnl_trylock())
5562 return restart_syscall();
5564 net = (struct net *)table->extra2;
5565 old = *p;
5566 *p = newf;
5568 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5569 rtnl_unlock();
5570 return 0;
5573 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5574 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5575 addrconf_disable_change(net, newf);
5576 } else if ((!newf) ^ (!old))
5577 dev_disable_change((struct inet6_dev *)table->extra1);
5579 rtnl_unlock();
5580 return 0;
5583 static
5584 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5585 void __user *buffer, size_t *lenp, loff_t *ppos)
5587 int *valp = ctl->data;
5588 int val = *valp;
5589 loff_t pos = *ppos;
5590 struct ctl_table lctl;
5591 int ret;
5594 * ctl->data points to idev->cnf.disable_ipv6, we should
5595 * not modify it until we get the rtnl lock.
5597 lctl = *ctl;
5598 lctl.data = &val;
5600 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5602 if (write)
5603 ret = addrconf_disable_ipv6(ctl, valp, val);
5604 if (ret)
5605 *ppos = pos;
5606 return ret;
5609 static
5610 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5611 void __user *buffer, size_t *lenp, loff_t *ppos)
5613 int *valp = ctl->data;
5614 int ret;
5615 int old, new;
5617 old = *valp;
5618 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5619 new = *valp;
5621 if (write && old != new) {
5622 struct net *net = ctl->extra2;
5624 if (!rtnl_trylock())
5625 return restart_syscall();
5627 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5628 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5629 NETCONFA_IFINDEX_DEFAULT,
5630 net->ipv6.devconf_dflt);
5631 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5632 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5633 NETCONFA_IFINDEX_ALL,
5634 net->ipv6.devconf_all);
5635 else {
5636 struct inet6_dev *idev = ctl->extra1;
5638 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
5639 idev->dev->ifindex,
5640 &idev->cnf);
5642 rtnl_unlock();
5645 return ret;
5648 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5649 void __user *buffer, size_t *lenp,
5650 loff_t *ppos)
5652 int err;
5653 struct in6_addr addr;
5654 char str[IPV6_MAX_STRLEN];
5655 struct ctl_table lctl = *ctl;
5656 struct net *net = ctl->extra2;
5657 struct ipv6_stable_secret *secret = ctl->data;
5659 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5660 return -EIO;
5662 lctl.maxlen = IPV6_MAX_STRLEN;
5663 lctl.data = str;
5665 if (!rtnl_trylock())
5666 return restart_syscall();
5668 if (!write && !secret->initialized) {
5669 err = -EIO;
5670 goto out;
5673 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5674 if (err >= sizeof(str)) {
5675 err = -EIO;
5676 goto out;
5679 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5680 if (err || !write)
5681 goto out;
5683 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5684 err = -EIO;
5685 goto out;
5688 secret->initialized = true;
5689 secret->secret = addr;
5691 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5692 struct net_device *dev;
5694 for_each_netdev(net, dev) {
5695 struct inet6_dev *idev = __in6_dev_get(dev);
5697 if (idev) {
5698 idev->addr_gen_mode =
5699 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5702 } else {
5703 struct inet6_dev *idev = ctl->extra1;
5705 idev->addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5708 out:
5709 rtnl_unlock();
5711 return err;
5714 static
5715 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5716 int write,
5717 void __user *buffer,
5718 size_t *lenp,
5719 loff_t *ppos)
5721 int *valp = ctl->data;
5722 int val = *valp;
5723 loff_t pos = *ppos;
5724 struct ctl_table lctl;
5725 int ret;
5727 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5728 * we should not modify it until we get the rtnl lock.
5730 lctl = *ctl;
5731 lctl.data = &val;
5733 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5735 if (write)
5736 ret = addrconf_fixup_linkdown(ctl, valp, val);
5737 if (ret)
5738 *ppos = pos;
5739 return ret;
5742 static int minus_one = -1;
5743 static const int one = 1;
5744 static const int two_five_five = 255;
5746 static const struct ctl_table addrconf_sysctl[] = {
5748 .procname = "forwarding",
5749 .data = &ipv6_devconf.forwarding,
5750 .maxlen = sizeof(int),
5751 .mode = 0644,
5752 .proc_handler = addrconf_sysctl_forward,
5755 .procname = "hop_limit",
5756 .data = &ipv6_devconf.hop_limit,
5757 .maxlen = sizeof(int),
5758 .mode = 0644,
5759 .proc_handler = proc_dointvec_minmax,
5760 .extra1 = (void *)&one,
5761 .extra2 = (void *)&two_five_five,
5764 .procname = "mtu",
5765 .data = &ipv6_devconf.mtu6,
5766 .maxlen = sizeof(int),
5767 .mode = 0644,
5768 .proc_handler = addrconf_sysctl_mtu,
5771 .procname = "accept_ra",
5772 .data = &ipv6_devconf.accept_ra,
5773 .maxlen = sizeof(int),
5774 .mode = 0644,
5775 .proc_handler = proc_dointvec,
5778 .procname = "accept_redirects",
5779 .data = &ipv6_devconf.accept_redirects,
5780 .maxlen = sizeof(int),
5781 .mode = 0644,
5782 .proc_handler = proc_dointvec,
5785 .procname = "autoconf",
5786 .data = &ipv6_devconf.autoconf,
5787 .maxlen = sizeof(int),
5788 .mode = 0644,
5789 .proc_handler = proc_dointvec,
5792 .procname = "dad_transmits",
5793 .data = &ipv6_devconf.dad_transmits,
5794 .maxlen = sizeof(int),
5795 .mode = 0644,
5796 .proc_handler = proc_dointvec,
5799 .procname = "router_solicitations",
5800 .data = &ipv6_devconf.rtr_solicits,
5801 .maxlen = sizeof(int),
5802 .mode = 0644,
5803 .proc_handler = proc_dointvec_minmax,
5804 .extra1 = &minus_one,
5807 .procname = "router_solicitation_interval",
5808 .data = &ipv6_devconf.rtr_solicit_interval,
5809 .maxlen = sizeof(int),
5810 .mode = 0644,
5811 .proc_handler = proc_dointvec_jiffies,
5814 .procname = "router_solicitation_max_interval",
5815 .data = &ipv6_devconf.rtr_solicit_max_interval,
5816 .maxlen = sizeof(int),
5817 .mode = 0644,
5818 .proc_handler = proc_dointvec_jiffies,
5821 .procname = "router_solicitation_delay",
5822 .data = &ipv6_devconf.rtr_solicit_delay,
5823 .maxlen = sizeof(int),
5824 .mode = 0644,
5825 .proc_handler = proc_dointvec_jiffies,
5828 .procname = "force_mld_version",
5829 .data = &ipv6_devconf.force_mld_version,
5830 .maxlen = sizeof(int),
5831 .mode = 0644,
5832 .proc_handler = proc_dointvec,
5835 .procname = "mldv1_unsolicited_report_interval",
5836 .data =
5837 &ipv6_devconf.mldv1_unsolicited_report_interval,
5838 .maxlen = sizeof(int),
5839 .mode = 0644,
5840 .proc_handler = proc_dointvec_ms_jiffies,
5843 .procname = "mldv2_unsolicited_report_interval",
5844 .data =
5845 &ipv6_devconf.mldv2_unsolicited_report_interval,
5846 .maxlen = sizeof(int),
5847 .mode = 0644,
5848 .proc_handler = proc_dointvec_ms_jiffies,
5851 .procname = "use_tempaddr",
5852 .data = &ipv6_devconf.use_tempaddr,
5853 .maxlen = sizeof(int),
5854 .mode = 0644,
5855 .proc_handler = proc_dointvec,
5858 .procname = "temp_valid_lft",
5859 .data = &ipv6_devconf.temp_valid_lft,
5860 .maxlen = sizeof(int),
5861 .mode = 0644,
5862 .proc_handler = proc_dointvec,
5865 .procname = "temp_prefered_lft",
5866 .data = &ipv6_devconf.temp_prefered_lft,
5867 .maxlen = sizeof(int),
5868 .mode = 0644,
5869 .proc_handler = proc_dointvec,
5872 .procname = "regen_max_retry",
5873 .data = &ipv6_devconf.regen_max_retry,
5874 .maxlen = sizeof(int),
5875 .mode = 0644,
5876 .proc_handler = proc_dointvec,
5879 .procname = "max_desync_factor",
5880 .data = &ipv6_devconf.max_desync_factor,
5881 .maxlen = sizeof(int),
5882 .mode = 0644,
5883 .proc_handler = proc_dointvec,
5886 .procname = "max_addresses",
5887 .data = &ipv6_devconf.max_addresses,
5888 .maxlen = sizeof(int),
5889 .mode = 0644,
5890 .proc_handler = proc_dointvec,
5893 .procname = "accept_ra_defrtr",
5894 .data = &ipv6_devconf.accept_ra_defrtr,
5895 .maxlen = sizeof(int),
5896 .mode = 0644,
5897 .proc_handler = proc_dointvec,
5900 .procname = "accept_ra_min_hop_limit",
5901 .data = &ipv6_devconf.accept_ra_min_hop_limit,
5902 .maxlen = sizeof(int),
5903 .mode = 0644,
5904 .proc_handler = proc_dointvec,
5907 .procname = "accept_ra_pinfo",
5908 .data = &ipv6_devconf.accept_ra_pinfo,
5909 .maxlen = sizeof(int),
5910 .mode = 0644,
5911 .proc_handler = proc_dointvec,
5913 #ifdef CONFIG_IPV6_ROUTER_PREF
5915 .procname = "accept_ra_rtr_pref",
5916 .data = &ipv6_devconf.accept_ra_rtr_pref,
5917 .maxlen = sizeof(int),
5918 .mode = 0644,
5919 .proc_handler = proc_dointvec,
5922 .procname = "router_probe_interval",
5923 .data = &ipv6_devconf.rtr_probe_interval,
5924 .maxlen = sizeof(int),
5925 .mode = 0644,
5926 .proc_handler = proc_dointvec_jiffies,
5928 #ifdef CONFIG_IPV6_ROUTE_INFO
5930 .procname = "accept_ra_rt_info_max_plen",
5931 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5932 .maxlen = sizeof(int),
5933 .mode = 0644,
5934 .proc_handler = proc_dointvec,
5936 #endif
5937 #endif
5939 .procname = "proxy_ndp",
5940 .data = &ipv6_devconf.proxy_ndp,
5941 .maxlen = sizeof(int),
5942 .mode = 0644,
5943 .proc_handler = addrconf_sysctl_proxy_ndp,
5946 .procname = "accept_source_route",
5947 .data = &ipv6_devconf.accept_source_route,
5948 .maxlen = sizeof(int),
5949 .mode = 0644,
5950 .proc_handler = proc_dointvec,
5952 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5954 .procname = "optimistic_dad",
5955 .data = &ipv6_devconf.optimistic_dad,
5956 .maxlen = sizeof(int),
5957 .mode = 0644,
5958 .proc_handler = proc_dointvec,
5961 .procname = "use_optimistic",
5962 .data = &ipv6_devconf.use_optimistic,
5963 .maxlen = sizeof(int),
5964 .mode = 0644,
5965 .proc_handler = proc_dointvec,
5967 #endif
5968 #ifdef CONFIG_IPV6_MROUTE
5970 .procname = "mc_forwarding",
5971 .data = &ipv6_devconf.mc_forwarding,
5972 .maxlen = sizeof(int),
5973 .mode = 0444,
5974 .proc_handler = proc_dointvec,
5976 #endif
5978 .procname = "disable_ipv6",
5979 .data = &ipv6_devconf.disable_ipv6,
5980 .maxlen = sizeof(int),
5981 .mode = 0644,
5982 .proc_handler = addrconf_sysctl_disable,
5985 .procname = "accept_dad",
5986 .data = &ipv6_devconf.accept_dad,
5987 .maxlen = sizeof(int),
5988 .mode = 0644,
5989 .proc_handler = proc_dointvec,
5992 .procname = "force_tllao",
5993 .data = &ipv6_devconf.force_tllao,
5994 .maxlen = sizeof(int),
5995 .mode = 0644,
5996 .proc_handler = proc_dointvec
5999 .procname = "ndisc_notify",
6000 .data = &ipv6_devconf.ndisc_notify,
6001 .maxlen = sizeof(int),
6002 .mode = 0644,
6003 .proc_handler = proc_dointvec
6006 .procname = "suppress_frag_ndisc",
6007 .data = &ipv6_devconf.suppress_frag_ndisc,
6008 .maxlen = sizeof(int),
6009 .mode = 0644,
6010 .proc_handler = proc_dointvec
6013 .procname = "accept_ra_from_local",
6014 .data = &ipv6_devconf.accept_ra_from_local,
6015 .maxlen = sizeof(int),
6016 .mode = 0644,
6017 .proc_handler = proc_dointvec,
6020 .procname = "accept_ra_mtu",
6021 .data = &ipv6_devconf.accept_ra_mtu,
6022 .maxlen = sizeof(int),
6023 .mode = 0644,
6024 .proc_handler = proc_dointvec,
6027 .procname = "stable_secret",
6028 .data = &ipv6_devconf.stable_secret,
6029 .maxlen = IPV6_MAX_STRLEN,
6030 .mode = 0600,
6031 .proc_handler = addrconf_sysctl_stable_secret,
6034 .procname = "use_oif_addrs_only",
6035 .data = &ipv6_devconf.use_oif_addrs_only,
6036 .maxlen = sizeof(int),
6037 .mode = 0644,
6038 .proc_handler = proc_dointvec,
6041 .procname = "ignore_routes_with_linkdown",
6042 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6043 .maxlen = sizeof(int),
6044 .mode = 0644,
6045 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6048 .procname = "drop_unicast_in_l2_multicast",
6049 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6050 .maxlen = sizeof(int),
6051 .mode = 0644,
6052 .proc_handler = proc_dointvec,
6055 .procname = "drop_unsolicited_na",
6056 .data = &ipv6_devconf.drop_unsolicited_na,
6057 .maxlen = sizeof(int),
6058 .mode = 0644,
6059 .proc_handler = proc_dointvec,
6062 .procname = "keep_addr_on_down",
6063 .data = &ipv6_devconf.keep_addr_on_down,
6064 .maxlen = sizeof(int),
6065 .mode = 0644,
6066 .proc_handler = proc_dointvec,
6070 .procname = "seg6_enabled",
6071 .data = &ipv6_devconf.seg6_enabled,
6072 .maxlen = sizeof(int),
6073 .mode = 0644,
6074 .proc_handler = proc_dointvec,
6076 #ifdef CONFIG_IPV6_SEG6_HMAC
6078 .procname = "seg6_require_hmac",
6079 .data = &ipv6_devconf.seg6_require_hmac,
6080 .maxlen = sizeof(int),
6081 .mode = 0644,
6082 .proc_handler = proc_dointvec,
6084 #endif
6086 .procname = "enhanced_dad",
6087 .data = &ipv6_devconf.enhanced_dad,
6088 .maxlen = sizeof(int),
6089 .mode = 0644,
6090 .proc_handler = proc_dointvec,
6093 /* sentinel */
6097 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6098 struct inet6_dev *idev, struct ipv6_devconf *p)
6100 int i, ifindex;
6101 struct ctl_table *table;
6102 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6104 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6105 if (!table)
6106 goto out;
6108 for (i = 0; table[i].data; i++) {
6109 table[i].data += (char *)p - (char *)&ipv6_devconf;
6110 /* If one of these is already set, then it is not safe to
6111 * overwrite either of them: this makes proc_dointvec_minmax
6112 * usable.
6114 if (!table[i].extra1 && !table[i].extra2) {
6115 table[i].extra1 = idev; /* embedded; no ref */
6116 table[i].extra2 = net;
6120 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6122 p->sysctl_header = register_net_sysctl(net, path, table);
6123 if (!p->sysctl_header)
6124 goto free;
6126 if (!strcmp(dev_name, "all"))
6127 ifindex = NETCONFA_IFINDEX_ALL;
6128 else if (!strcmp(dev_name, "default"))
6129 ifindex = NETCONFA_IFINDEX_DEFAULT;
6130 else
6131 ifindex = idev->dev->ifindex;
6132 inet6_netconf_notify_devconf(net, NETCONFA_ALL, ifindex, p);
6133 return 0;
6135 free:
6136 kfree(table);
6137 out:
6138 return -ENOBUFS;
6141 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
6143 struct ctl_table *table;
6145 if (!p->sysctl_header)
6146 return;
6148 table = p->sysctl_header->ctl_table_arg;
6149 unregister_net_sysctl_table(p->sysctl_header);
6150 p->sysctl_header = NULL;
6151 kfree(table);
6154 static int addrconf_sysctl_register(struct inet6_dev *idev)
6156 int err;
6158 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6159 return -EINVAL;
6161 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6162 &ndisc_ifinfo_sysctl_change);
6163 if (err)
6164 return err;
6165 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6166 idev, &idev->cnf);
6167 if (err)
6168 neigh_sysctl_unregister(idev->nd_parms);
6170 return err;
6173 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6175 __addrconf_sysctl_unregister(&idev->cnf);
6176 neigh_sysctl_unregister(idev->nd_parms);
6180 #endif
6182 static int __net_init addrconf_init_net(struct net *net)
6184 int err = -ENOMEM;
6185 struct ipv6_devconf *all, *dflt;
6187 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6188 if (!all)
6189 goto err_alloc_all;
6191 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6192 if (!dflt)
6193 goto err_alloc_dflt;
6195 /* these will be inherited by all namespaces */
6196 dflt->autoconf = ipv6_defaults.autoconf;
6197 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6199 dflt->stable_secret.initialized = false;
6200 all->stable_secret.initialized = false;
6202 net->ipv6.devconf_all = all;
6203 net->ipv6.devconf_dflt = dflt;
6205 #ifdef CONFIG_SYSCTL
6206 err = __addrconf_sysctl_register(net, "all", NULL, all);
6207 if (err < 0)
6208 goto err_reg_all;
6210 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6211 if (err < 0)
6212 goto err_reg_dflt;
6213 #endif
6214 return 0;
6216 #ifdef CONFIG_SYSCTL
6217 err_reg_dflt:
6218 __addrconf_sysctl_unregister(all);
6219 err_reg_all:
6220 kfree(dflt);
6221 #endif
6222 err_alloc_dflt:
6223 kfree(all);
6224 err_alloc_all:
6225 return err;
6228 static void __net_exit addrconf_exit_net(struct net *net)
6230 #ifdef CONFIG_SYSCTL
6231 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
6232 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
6233 #endif
6234 kfree(net->ipv6.devconf_dflt);
6235 kfree(net->ipv6.devconf_all);
6238 static struct pernet_operations addrconf_ops = {
6239 .init = addrconf_init_net,
6240 .exit = addrconf_exit_net,
6243 static struct rtnl_af_ops inet6_ops __read_mostly = {
6244 .family = AF_INET6,
6245 .fill_link_af = inet6_fill_link_af,
6246 .get_link_af_size = inet6_get_link_af_size,
6247 .validate_link_af = inet6_validate_link_af,
6248 .set_link_af = inet6_set_link_af,
6252 * Init / cleanup code
6255 int __init addrconf_init(void)
6257 struct inet6_dev *idev;
6258 int i, err;
6260 err = ipv6_addr_label_init();
6261 if (err < 0) {
6262 pr_crit("%s: cannot initialize default policy table: %d\n",
6263 __func__, err);
6264 goto out;
6267 err = register_pernet_subsys(&addrconf_ops);
6268 if (err < 0)
6269 goto out_addrlabel;
6271 addrconf_wq = create_workqueue("ipv6_addrconf");
6272 if (!addrconf_wq) {
6273 err = -ENOMEM;
6274 goto out_nowq;
6277 /* The addrconf netdev notifier requires that loopback_dev
6278 * has it's ipv6 private information allocated and setup
6279 * before it can bring up and give link-local addresses
6280 * to other devices which are up.
6282 * Unfortunately, loopback_dev is not necessarily the first
6283 * entry in the global dev_base list of net devices. In fact,
6284 * it is likely to be the very last entry on that list.
6285 * So this causes the notifier registry below to try and
6286 * give link-local addresses to all devices besides loopback_dev
6287 * first, then loopback_dev, which cases all the non-loopback_dev
6288 * devices to fail to get a link-local address.
6290 * So, as a temporary fix, allocate the ipv6 structure for
6291 * loopback_dev first by hand.
6292 * Longer term, all of the dependencies ipv6 has upon the loopback
6293 * device and it being up should be removed.
6295 rtnl_lock();
6296 idev = ipv6_add_dev(init_net.loopback_dev);
6297 rtnl_unlock();
6298 if (IS_ERR(idev)) {
6299 err = PTR_ERR(idev);
6300 goto errlo;
6303 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6304 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6306 register_netdevice_notifier(&ipv6_dev_notf);
6308 addrconf_verify();
6310 rtnl_af_register(&inet6_ops);
6312 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6313 NULL);
6314 if (err < 0)
6315 goto errout;
6317 /* Only the first call to __rtnl_register can fail */
6318 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6319 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6320 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6321 inet6_dump_ifaddr, NULL);
6322 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6323 inet6_dump_ifmcaddr, NULL);
6324 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6325 inet6_dump_ifacaddr, NULL);
6326 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6327 inet6_netconf_dump_devconf, NULL);
6329 ipv6_addr_label_rtnl_register();
6331 return 0;
6332 errout:
6333 rtnl_af_unregister(&inet6_ops);
6334 unregister_netdevice_notifier(&ipv6_dev_notf);
6335 errlo:
6336 destroy_workqueue(addrconf_wq);
6337 out_nowq:
6338 unregister_pernet_subsys(&addrconf_ops);
6339 out_addrlabel:
6340 ipv6_addr_label_cleanup();
6341 out:
6342 return err;
6345 void addrconf_cleanup(void)
6347 struct net_device *dev;
6348 int i;
6350 unregister_netdevice_notifier(&ipv6_dev_notf);
6351 unregister_pernet_subsys(&addrconf_ops);
6352 ipv6_addr_label_cleanup();
6354 rtnl_lock();
6356 __rtnl_af_unregister(&inet6_ops);
6358 /* clean dev list */
6359 for_each_netdev(&init_net, dev) {
6360 if (__in6_dev_get(dev) == NULL)
6361 continue;
6362 addrconf_ifdown(dev, 1);
6364 addrconf_ifdown(init_net.loopback_dev, 2);
6367 * Check hash table.
6369 spin_lock_bh(&addrconf_hash_lock);
6370 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6371 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6372 spin_unlock_bh(&addrconf_hash_lock);
6373 cancel_delayed_work(&addr_chk_work);
6374 rtnl_unlock();
6376 destroy_workqueue(addrconf_wq);