dm: fix REQ_OP_ZONE_RESET bio handling
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob6a4fb1e629fb7609048156974ae2eb322cebddae
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/sched/signal.h>
47 #include <linux/socket.h>
48 #include <linux/sockios.h>
49 #include <linux/net.h>
50 #include <linux/inet.h>
51 #include <linux/in6.h>
52 #include <linux/netdevice.h>
53 #include <linux/if_addr.h>
54 #include <linux/if_arp.h>
55 #include <linux/if_arcnet.h>
56 #include <linux/if_infiniband.h>
57 #include <linux/route.h>
58 #include <linux/inetdevice.h>
59 #include <linux/init.h>
60 #include <linux/slab.h>
61 #ifdef CONFIG_SYSCTL
62 #include <linux/sysctl.h>
63 #endif
64 #include <linux/capability.h>
65 #include <linux/delay.h>
66 #include <linux/notifier.h>
67 #include <linux/string.h>
68 #include <linux/hash.h>
70 #include <net/net_namespace.h>
71 #include <net/sock.h>
72 #include <net/snmp.h>
74 #include <net/6lowpan.h>
75 #include <net/firewire.h>
76 #include <net/ipv6.h>
77 #include <net/protocol.h>
78 #include <net/ndisc.h>
79 #include <net/ip6_route.h>
80 #include <net/addrconf.h>
81 #include <net/tcp.h>
82 #include <net/ip.h>
83 #include <net/netlink.h>
84 #include <net/pkt_sched.h>
85 #include <net/l3mdev.h>
86 #include <linux/if_tunnel.h>
87 #include <linux/rtnetlink.h>
88 #include <linux/netconf.h>
89 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 /* Set to 3 to get tracing... */
98 #define ACONF_DEBUG 2
100 #if ACONF_DEBUG >= 3
101 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
102 #else
103 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
104 #endif
106 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108 #define IPV6_MAX_STRLEN \
109 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
111 static inline u32 cstamp_delta(unsigned long cstamp)
113 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
116 static inline s32 rfc3315_s14_backoff_init(s32 irt)
118 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
119 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
120 do_div(tmp, 1000000);
121 return (s32)tmp;
124 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
126 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
127 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
128 do_div(tmp, 1000000);
129 if ((s32)tmp > mrt) {
130 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
131 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
132 do_div(tmp, 1000000);
134 return (s32)tmp;
137 #ifdef CONFIG_SYSCTL
138 static int addrconf_sysctl_register(struct inet6_dev *idev);
139 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
140 #else
141 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
143 return 0;
146 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
149 #endif
151 static void ipv6_regen_rndid(struct inet6_dev *idev);
152 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
154 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
155 static int ipv6_count_addresses(struct inet6_dev *idev);
156 static int ipv6_generate_stable_address(struct in6_addr *addr,
157 u8 dad_count,
158 const struct inet6_dev *idev);
161 * Configured unicast address hash table
163 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
164 static DEFINE_SPINLOCK(addrconf_hash_lock);
166 static void addrconf_verify(void);
167 static void addrconf_verify_rtnl(void);
168 static void addrconf_verify_work(struct work_struct *);
170 static struct workqueue_struct *addrconf_wq;
171 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
173 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
174 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
176 static void addrconf_type_change(struct net_device *dev,
177 unsigned long event);
178 static int addrconf_ifdown(struct net_device *dev, int how);
180 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
181 int plen,
182 const struct net_device *dev,
183 u32 flags, u32 noflags);
185 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
186 static void addrconf_dad_work(struct work_struct *w);
187 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id);
188 static void addrconf_dad_run(struct inet6_dev *idev);
189 static void addrconf_rs_timer(unsigned long data);
190 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
191 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
193 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
194 struct prefix_info *pinfo);
195 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
196 struct net_device *dev);
198 static struct ipv6_devconf ipv6_devconf __read_mostly = {
199 .forwarding = 0,
200 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
201 .mtu6 = IPV6_MIN_MTU,
202 .accept_ra = 1,
203 .accept_redirects = 1,
204 .autoconf = 1,
205 .force_mld_version = 0,
206 .mldv1_unsolicited_report_interval = 10 * HZ,
207 .mldv2_unsolicited_report_interval = HZ,
208 .dad_transmits = 1,
209 .rtr_solicits = MAX_RTR_SOLICITATIONS,
210 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
211 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
213 .use_tempaddr = 0,
214 .temp_valid_lft = TEMP_VALID_LIFETIME,
215 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
216 .regen_max_retry = REGEN_MAX_RETRY,
217 .max_desync_factor = MAX_DESYNC_FACTOR,
218 .max_addresses = IPV6_MAX_ADDRESSES,
219 .accept_ra_defrtr = 1,
220 .accept_ra_from_local = 0,
221 .accept_ra_min_hop_limit= 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_min_plen = 0,
228 .accept_ra_rt_info_max_plen = 0,
229 #endif
230 #endif
231 .proxy_ndp = 0,
232 .accept_source_route = 0, /* we do not accept RH0 by default. */
233 .disable_ipv6 = 0,
234 .accept_dad = 1,
235 .suppress_frag_ndisc = 1,
236 .accept_ra_mtu = 1,
237 .stable_secret = {
238 .initialized = false,
240 .use_oif_addrs_only = 0,
241 .ignore_routes_with_linkdown = 0,
242 .keep_addr_on_down = 0,
243 .seg6_enabled = 0,
244 #ifdef CONFIG_IPV6_SEG6_HMAC
245 .seg6_require_hmac = 0,
246 #endif
247 .enhanced_dad = 1,
248 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
249 .disable_policy = 0,
252 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
253 .forwarding = 0,
254 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
255 .mtu6 = IPV6_MIN_MTU,
256 .accept_ra = 1,
257 .accept_redirects = 1,
258 .autoconf = 1,
259 .force_mld_version = 0,
260 .mldv1_unsolicited_report_interval = 10 * HZ,
261 .mldv2_unsolicited_report_interval = HZ,
262 .dad_transmits = 1,
263 .rtr_solicits = MAX_RTR_SOLICITATIONS,
264 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
265 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
266 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
267 .use_tempaddr = 0,
268 .temp_valid_lft = TEMP_VALID_LIFETIME,
269 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
270 .regen_max_retry = REGEN_MAX_RETRY,
271 .max_desync_factor = MAX_DESYNC_FACTOR,
272 .max_addresses = IPV6_MAX_ADDRESSES,
273 .accept_ra_defrtr = 1,
274 .accept_ra_from_local = 0,
275 .accept_ra_min_hop_limit= 1,
276 .accept_ra_pinfo = 1,
277 #ifdef CONFIG_IPV6_ROUTER_PREF
278 .accept_ra_rtr_pref = 1,
279 .rtr_probe_interval = 60 * HZ,
280 #ifdef CONFIG_IPV6_ROUTE_INFO
281 .accept_ra_rt_info_min_plen = 0,
282 .accept_ra_rt_info_max_plen = 0,
283 #endif
284 #endif
285 .proxy_ndp = 0,
286 .accept_source_route = 0, /* we do not accept RH0 by default. */
287 .disable_ipv6 = 0,
288 .accept_dad = 1,
289 .suppress_frag_ndisc = 1,
290 .accept_ra_mtu = 1,
291 .stable_secret = {
292 .initialized = false,
294 .use_oif_addrs_only = 0,
295 .ignore_routes_with_linkdown = 0,
296 .keep_addr_on_down = 0,
297 .seg6_enabled = 0,
298 #ifdef CONFIG_IPV6_SEG6_HMAC
299 .seg6_require_hmac = 0,
300 #endif
301 .enhanced_dad = 1,
302 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
303 .disable_policy = 0,
306 /* Check if a valid qdisc is available */
307 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
309 return !qdisc_tx_is_noop(dev);
312 static void addrconf_del_rs_timer(struct inet6_dev *idev)
314 if (del_timer(&idev->rs_timer))
315 __in6_dev_put(idev);
318 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
320 if (cancel_delayed_work(&ifp->dad_work))
321 __in6_ifa_put(ifp);
324 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
325 unsigned long when)
327 if (!timer_pending(&idev->rs_timer))
328 in6_dev_hold(idev);
329 mod_timer(&idev->rs_timer, jiffies + when);
332 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
333 unsigned long delay)
335 if (!delayed_work_pending(&ifp->dad_work))
336 in6_ifa_hold(ifp);
337 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
340 static int snmp6_alloc_dev(struct inet6_dev *idev)
342 int i;
344 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
345 if (!idev->stats.ipv6)
346 goto err_ip;
348 for_each_possible_cpu(i) {
349 struct ipstats_mib *addrconf_stats;
350 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
351 u64_stats_init(&addrconf_stats->syncp);
355 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
356 GFP_KERNEL);
357 if (!idev->stats.icmpv6dev)
358 goto err_icmp;
359 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
360 GFP_KERNEL);
361 if (!idev->stats.icmpv6msgdev)
362 goto err_icmpmsg;
364 return 0;
366 err_icmpmsg:
367 kfree(idev->stats.icmpv6dev);
368 err_icmp:
369 free_percpu(idev->stats.ipv6);
370 err_ip:
371 return -ENOMEM;
374 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
376 struct inet6_dev *ndev;
377 int err = -ENOMEM;
379 ASSERT_RTNL();
381 if (dev->mtu < IPV6_MIN_MTU)
382 return ERR_PTR(-EINVAL);
384 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
385 if (!ndev)
386 return ERR_PTR(err);
388 rwlock_init(&ndev->lock);
389 ndev->dev = dev;
390 INIT_LIST_HEAD(&ndev->addr_list);
391 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
392 (unsigned long)ndev);
393 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
395 if (ndev->cnf.stable_secret.initialized)
396 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
397 else
398 ndev->cnf.addr_gen_mode = ipv6_devconf_dflt.addr_gen_mode;
400 ndev->cnf.mtu6 = dev->mtu;
401 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
402 if (!ndev->nd_parms) {
403 kfree(ndev);
404 return ERR_PTR(err);
406 if (ndev->cnf.forwarding)
407 dev_disable_lro(dev);
408 /* We refer to the device */
409 dev_hold(dev);
411 if (snmp6_alloc_dev(ndev) < 0) {
412 ADBG(KERN_WARNING
413 "%s: cannot allocate memory for statistics; dev=%s.\n",
414 __func__, dev->name);
415 neigh_parms_release(&nd_tbl, ndev->nd_parms);
416 dev_put(dev);
417 kfree(ndev);
418 return ERR_PTR(err);
421 if (snmp6_register_dev(ndev) < 0) {
422 ADBG(KERN_WARNING
423 "%s: cannot create /proc/net/dev_snmp6/%s\n",
424 __func__, dev->name);
425 goto err_release;
428 /* One reference from device. */
429 in6_dev_hold(ndev);
431 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
432 ndev->cnf.accept_dad = -1;
434 #if IS_ENABLED(CONFIG_IPV6_SIT)
435 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
436 pr_info("%s: Disabled Multicast RS\n", dev->name);
437 ndev->cnf.rtr_solicits = 0;
439 #endif
441 INIT_LIST_HEAD(&ndev->tempaddr_list);
442 ndev->desync_factor = U32_MAX;
443 if ((dev->flags&IFF_LOOPBACK) ||
444 dev->type == ARPHRD_TUNNEL ||
445 dev->type == ARPHRD_TUNNEL6 ||
446 dev->type == ARPHRD_SIT ||
447 dev->type == ARPHRD_NONE) {
448 ndev->cnf.use_tempaddr = -1;
449 } else
450 ipv6_regen_rndid(ndev);
452 ndev->token = in6addr_any;
454 if (netif_running(dev) && addrconf_qdisc_ok(dev))
455 ndev->if_flags |= IF_READY;
457 ipv6_mc_init_dev(ndev);
458 ndev->tstamp = jiffies;
459 err = addrconf_sysctl_register(ndev);
460 if (err) {
461 ipv6_mc_destroy_dev(ndev);
462 snmp6_unregister_dev(ndev);
463 goto err_release;
465 /* protected by rtnl_lock */
466 rcu_assign_pointer(dev->ip6_ptr, ndev);
468 /* Join interface-local all-node multicast group */
469 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
471 /* Join all-node multicast group */
472 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
474 /* Join all-router multicast group if forwarding is set */
475 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
476 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
478 return ndev;
480 err_release:
481 neigh_parms_release(&nd_tbl, ndev->nd_parms);
482 ndev->dead = 1;
483 in6_dev_finish_destroy(ndev);
484 return ERR_PTR(err);
487 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
489 struct inet6_dev *idev;
491 ASSERT_RTNL();
493 idev = __in6_dev_get(dev);
494 if (!idev) {
495 idev = ipv6_add_dev(dev);
496 if (IS_ERR(idev))
497 return NULL;
500 if (dev->flags&IFF_UP)
501 ipv6_mc_up(idev);
502 return idev;
505 static int inet6_netconf_msgsize_devconf(int type)
507 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
508 + nla_total_size(4); /* NETCONFA_IFINDEX */
509 bool all = false;
511 if (type == NETCONFA_ALL)
512 all = true;
514 if (all || type == NETCONFA_FORWARDING)
515 size += nla_total_size(4);
516 #ifdef CONFIG_IPV6_MROUTE
517 if (all || type == NETCONFA_MC_FORWARDING)
518 size += nla_total_size(4);
519 #endif
520 if (all || type == NETCONFA_PROXY_NEIGH)
521 size += nla_total_size(4);
523 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
524 size += nla_total_size(4);
526 return size;
529 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
530 struct ipv6_devconf *devconf, u32 portid,
531 u32 seq, int event, unsigned int flags,
532 int type)
534 struct nlmsghdr *nlh;
535 struct netconfmsg *ncm;
536 bool all = false;
538 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
539 flags);
540 if (!nlh)
541 return -EMSGSIZE;
543 if (type == NETCONFA_ALL)
544 all = true;
546 ncm = nlmsg_data(nlh);
547 ncm->ncm_family = AF_INET6;
549 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
550 goto nla_put_failure;
552 if (!devconf)
553 goto out;
555 if ((all || type == NETCONFA_FORWARDING) &&
556 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
557 goto nla_put_failure;
558 #ifdef CONFIG_IPV6_MROUTE
559 if ((all || type == NETCONFA_MC_FORWARDING) &&
560 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
561 devconf->mc_forwarding) < 0)
562 goto nla_put_failure;
563 #endif
564 if ((all || type == NETCONFA_PROXY_NEIGH) &&
565 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
566 goto nla_put_failure;
568 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
569 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
570 devconf->ignore_routes_with_linkdown) < 0)
571 goto nla_put_failure;
573 out:
574 nlmsg_end(skb, nlh);
575 return 0;
577 nla_put_failure:
578 nlmsg_cancel(skb, nlh);
579 return -EMSGSIZE;
582 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
583 int ifindex, struct ipv6_devconf *devconf)
585 struct sk_buff *skb;
586 int err = -ENOBUFS;
588 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
589 if (!skb)
590 goto errout;
592 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
593 event, 0, type);
594 if (err < 0) {
595 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
596 WARN_ON(err == -EMSGSIZE);
597 kfree_skb(skb);
598 goto errout;
600 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
601 return;
602 errout:
603 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
606 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
607 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
608 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
609 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
610 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
613 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
614 struct nlmsghdr *nlh,
615 struct netlink_ext_ack *extack)
617 struct net *net = sock_net(in_skb->sk);
618 struct nlattr *tb[NETCONFA_MAX+1];
619 struct netconfmsg *ncm;
620 struct sk_buff *skb;
621 struct ipv6_devconf *devconf;
622 struct inet6_dev *in6_dev;
623 struct net_device *dev;
624 int ifindex;
625 int err;
627 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
628 devconf_ipv6_policy, extack);
629 if (err < 0)
630 goto errout;
632 err = -EINVAL;
633 if (!tb[NETCONFA_IFINDEX])
634 goto errout;
636 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
637 switch (ifindex) {
638 case NETCONFA_IFINDEX_ALL:
639 devconf = net->ipv6.devconf_all;
640 break;
641 case NETCONFA_IFINDEX_DEFAULT:
642 devconf = net->ipv6.devconf_dflt;
643 break;
644 default:
645 dev = __dev_get_by_index(net, ifindex);
646 if (!dev)
647 goto errout;
648 in6_dev = __in6_dev_get(dev);
649 if (!in6_dev)
650 goto errout;
651 devconf = &in6_dev->cnf;
652 break;
655 err = -ENOBUFS;
656 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_ATOMIC);
657 if (!skb)
658 goto errout;
660 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
661 NETLINK_CB(in_skb).portid,
662 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
663 NETCONFA_ALL);
664 if (err < 0) {
665 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
666 WARN_ON(err == -EMSGSIZE);
667 kfree_skb(skb);
668 goto errout;
670 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
671 errout:
672 return err;
675 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
676 struct netlink_callback *cb)
678 struct net *net = sock_net(skb->sk);
679 int h, s_h;
680 int idx, s_idx;
681 struct net_device *dev;
682 struct inet6_dev *idev;
683 struct hlist_head *head;
685 s_h = cb->args[0];
686 s_idx = idx = cb->args[1];
688 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
689 idx = 0;
690 head = &net->dev_index_head[h];
691 rcu_read_lock();
692 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
693 net->dev_base_seq;
694 hlist_for_each_entry_rcu(dev, head, index_hlist) {
695 if (idx < s_idx)
696 goto cont;
697 idev = __in6_dev_get(dev);
698 if (!idev)
699 goto cont;
701 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
702 &idev->cnf,
703 NETLINK_CB(cb->skb).portid,
704 cb->nlh->nlmsg_seq,
705 RTM_NEWNETCONF,
706 NLM_F_MULTI,
707 NETCONFA_ALL) < 0) {
708 rcu_read_unlock();
709 goto done;
711 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
712 cont:
713 idx++;
715 rcu_read_unlock();
717 if (h == NETDEV_HASHENTRIES) {
718 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
719 net->ipv6.devconf_all,
720 NETLINK_CB(cb->skb).portid,
721 cb->nlh->nlmsg_seq,
722 RTM_NEWNETCONF, NLM_F_MULTI,
723 NETCONFA_ALL) < 0)
724 goto done;
725 else
726 h++;
728 if (h == NETDEV_HASHENTRIES + 1) {
729 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
730 net->ipv6.devconf_dflt,
731 NETLINK_CB(cb->skb).portid,
732 cb->nlh->nlmsg_seq,
733 RTM_NEWNETCONF, NLM_F_MULTI,
734 NETCONFA_ALL) < 0)
735 goto done;
736 else
737 h++;
739 done:
740 cb->args[0] = h;
741 cb->args[1] = idx;
743 return skb->len;
746 #ifdef CONFIG_SYSCTL
747 static void dev_forward_change(struct inet6_dev *idev)
749 struct net_device *dev;
750 struct inet6_ifaddr *ifa;
752 if (!idev)
753 return;
754 dev = idev->dev;
755 if (idev->cnf.forwarding)
756 dev_disable_lro(dev);
757 if (dev->flags & IFF_MULTICAST) {
758 if (idev->cnf.forwarding) {
759 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
760 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
761 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
762 } else {
763 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
764 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
765 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
769 list_for_each_entry(ifa, &idev->addr_list, if_list) {
770 if (ifa->flags&IFA_F_TENTATIVE)
771 continue;
772 if (idev->cnf.forwarding)
773 addrconf_join_anycast(ifa);
774 else
775 addrconf_leave_anycast(ifa);
777 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
778 NETCONFA_FORWARDING,
779 dev->ifindex, &idev->cnf);
783 static void addrconf_forward_change(struct net *net, __s32 newf)
785 struct net_device *dev;
786 struct inet6_dev *idev;
788 for_each_netdev(net, dev) {
789 idev = __in6_dev_get(dev);
790 if (idev) {
791 int changed = (!idev->cnf.forwarding) ^ (!newf);
792 idev->cnf.forwarding = newf;
793 if (changed)
794 dev_forward_change(idev);
799 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
801 struct net *net;
802 int old;
804 if (!rtnl_trylock())
805 return restart_syscall();
807 net = (struct net *)table->extra2;
808 old = *p;
809 *p = newf;
811 if (p == &net->ipv6.devconf_dflt->forwarding) {
812 if ((!newf) ^ (!old))
813 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
814 NETCONFA_FORWARDING,
815 NETCONFA_IFINDEX_DEFAULT,
816 net->ipv6.devconf_dflt);
817 rtnl_unlock();
818 return 0;
821 if (p == &net->ipv6.devconf_all->forwarding) {
822 int old_dflt = net->ipv6.devconf_dflt->forwarding;
824 net->ipv6.devconf_dflt->forwarding = newf;
825 if ((!newf) ^ (!old_dflt))
826 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
827 NETCONFA_FORWARDING,
828 NETCONFA_IFINDEX_DEFAULT,
829 net->ipv6.devconf_dflt);
831 addrconf_forward_change(net, newf);
832 if ((!newf) ^ (!old))
833 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
834 NETCONFA_FORWARDING,
835 NETCONFA_IFINDEX_ALL,
836 net->ipv6.devconf_all);
837 } else if ((!newf) ^ (!old))
838 dev_forward_change((struct inet6_dev *)table->extra1);
839 rtnl_unlock();
841 if (newf)
842 rt6_purge_dflt_routers(net);
843 return 1;
846 static void addrconf_linkdown_change(struct net *net, __s32 newf)
848 struct net_device *dev;
849 struct inet6_dev *idev;
851 for_each_netdev(net, dev) {
852 idev = __in6_dev_get(dev);
853 if (idev) {
854 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
856 idev->cnf.ignore_routes_with_linkdown = newf;
857 if (changed)
858 inet6_netconf_notify_devconf(dev_net(dev),
859 RTM_NEWNETCONF,
860 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
861 dev->ifindex,
862 &idev->cnf);
867 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
869 struct net *net;
870 int old;
872 if (!rtnl_trylock())
873 return restart_syscall();
875 net = (struct net *)table->extra2;
876 old = *p;
877 *p = newf;
879 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
880 if ((!newf) ^ (!old))
881 inet6_netconf_notify_devconf(net,
882 RTM_NEWNETCONF,
883 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
884 NETCONFA_IFINDEX_DEFAULT,
885 net->ipv6.devconf_dflt);
886 rtnl_unlock();
887 return 0;
890 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
891 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
892 addrconf_linkdown_change(net, newf);
893 if ((!newf) ^ (!old))
894 inet6_netconf_notify_devconf(net,
895 RTM_NEWNETCONF,
896 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
897 NETCONFA_IFINDEX_ALL,
898 net->ipv6.devconf_all);
900 rtnl_unlock();
902 return 1;
905 #endif
907 /* Nobody refers to this ifaddr, destroy it */
908 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
910 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
912 #ifdef NET_REFCNT_DEBUG
913 pr_debug("%s\n", __func__);
914 #endif
916 in6_dev_put(ifp->idev);
918 if (cancel_delayed_work(&ifp->dad_work))
919 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
920 ifp);
922 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
923 pr_warn("Freeing alive inet6 address %p\n", ifp);
924 return;
926 ip6_rt_put(ifp->rt);
928 kfree_rcu(ifp, rcu);
931 static void
932 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
934 struct list_head *p;
935 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
938 * Each device address list is sorted in order of scope -
939 * global before linklocal.
941 list_for_each(p, &idev->addr_list) {
942 struct inet6_ifaddr *ifa
943 = list_entry(p, struct inet6_ifaddr, if_list);
944 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
945 break;
948 list_add_tail(&ifp->if_list, p);
951 static u32 inet6_addr_hash(const struct in6_addr *addr)
953 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
956 /* On success it returns ifp with increased reference count */
958 static struct inet6_ifaddr *
959 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
960 const struct in6_addr *peer_addr, int pfxlen,
961 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
963 struct net *net = dev_net(idev->dev);
964 struct inet6_ifaddr *ifa = NULL;
965 struct rt6_info *rt;
966 unsigned int hash;
967 int err = 0;
968 int addr_type = ipv6_addr_type(addr);
970 if (addr_type == IPV6_ADDR_ANY ||
971 addr_type & IPV6_ADDR_MULTICAST ||
972 (!(idev->dev->flags & IFF_LOOPBACK) &&
973 addr_type & IPV6_ADDR_LOOPBACK))
974 return ERR_PTR(-EADDRNOTAVAIL);
976 rcu_read_lock_bh();
977 if (idev->dead) {
978 err = -ENODEV; /*XXX*/
979 goto out2;
982 if (idev->cnf.disable_ipv6) {
983 err = -EACCES;
984 goto out2;
987 spin_lock(&addrconf_hash_lock);
989 /* Ignore adding duplicate addresses on an interface */
990 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
991 ADBG("ipv6_add_addr: already assigned\n");
992 err = -EEXIST;
993 goto out;
996 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
998 if (!ifa) {
999 ADBG("ipv6_add_addr: malloc failed\n");
1000 err = -ENOBUFS;
1001 goto out;
1004 rt = addrconf_dst_alloc(idev, addr, false);
1005 if (IS_ERR(rt)) {
1006 err = PTR_ERR(rt);
1007 goto out;
1010 if (net->ipv6.devconf_all->disable_policy ||
1011 idev->cnf.disable_policy)
1012 rt->dst.flags |= DST_NOPOLICY;
1014 neigh_parms_data_state_setall(idev->nd_parms);
1016 ifa->addr = *addr;
1017 if (peer_addr)
1018 ifa->peer_addr = *peer_addr;
1020 spin_lock_init(&ifa->lock);
1021 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1022 INIT_HLIST_NODE(&ifa->addr_lst);
1023 ifa->scope = scope;
1024 ifa->prefix_len = pfxlen;
1025 ifa->flags = flags;
1026 /* No need to add the TENTATIVE flag for addresses with NODAD */
1027 if (!(flags & IFA_F_NODAD))
1028 ifa->flags |= IFA_F_TENTATIVE;
1029 ifa->valid_lft = valid_lft;
1030 ifa->prefered_lft = prefered_lft;
1031 ifa->cstamp = ifa->tstamp = jiffies;
1032 ifa->tokenized = false;
1034 ifa->rt = rt;
1036 ifa->idev = idev;
1037 in6_dev_hold(idev);
1038 /* For caller */
1039 in6_ifa_hold(ifa);
1041 /* Add to big hash table */
1042 hash = inet6_addr_hash(addr);
1044 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1045 spin_unlock(&addrconf_hash_lock);
1047 write_lock(&idev->lock);
1048 /* Add to inet6_dev unicast addr list. */
1049 ipv6_link_dev_addr(idev, ifa);
1051 if (ifa->flags&IFA_F_TEMPORARY) {
1052 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1053 in6_ifa_hold(ifa);
1056 in6_ifa_hold(ifa);
1057 write_unlock(&idev->lock);
1058 out2:
1059 rcu_read_unlock_bh();
1061 if (likely(err == 0))
1062 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1063 else {
1064 kfree(ifa);
1065 ifa = ERR_PTR(err);
1068 return ifa;
1069 out:
1070 spin_unlock(&addrconf_hash_lock);
1071 goto out2;
1074 enum cleanup_prefix_rt_t {
1075 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1076 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1077 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1081 * Check, whether the prefix for ifp would still need a prefix route
1082 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1083 * constants.
1085 * 1) we don't purge prefix if address was not permanent.
1086 * prefix is managed by its own lifetime.
1087 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1088 * 3) if there are no addresses, delete prefix.
1089 * 4) if there are still other permanent address(es),
1090 * corresponding prefix is still permanent.
1091 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1092 * don't purge the prefix, assume user space is managing it.
1093 * 6) otherwise, update prefix lifetime to the
1094 * longest valid lifetime among the corresponding
1095 * addresses on the device.
1096 * Note: subsequent RA will update lifetime.
1098 static enum cleanup_prefix_rt_t
1099 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1101 struct inet6_ifaddr *ifa;
1102 struct inet6_dev *idev = ifp->idev;
1103 unsigned long lifetime;
1104 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1106 *expires = jiffies;
1108 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1109 if (ifa == ifp)
1110 continue;
1111 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1112 ifp->prefix_len))
1113 continue;
1114 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1115 return CLEANUP_PREFIX_RT_NOP;
1117 action = CLEANUP_PREFIX_RT_EXPIRE;
1119 spin_lock(&ifa->lock);
1121 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1123 * Note: Because this address is
1124 * not permanent, lifetime <
1125 * LONG_MAX / HZ here.
1127 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1128 *expires = ifa->tstamp + lifetime * HZ;
1129 spin_unlock(&ifa->lock);
1132 return action;
1135 static void
1136 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
1138 struct rt6_info *rt;
1140 rt = addrconf_get_prefix_route(&ifp->addr,
1141 ifp->prefix_len,
1142 ifp->idev->dev,
1143 0, RTF_GATEWAY | RTF_DEFAULT);
1144 if (rt) {
1145 if (del_rt)
1146 ip6_del_rt(rt);
1147 else {
1148 if (!(rt->rt6i_flags & RTF_EXPIRES))
1149 rt6_set_expires(rt, expires);
1150 ip6_rt_put(rt);
1156 /* This function wants to get referenced ifp and releases it before return */
1158 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1160 int state;
1161 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1162 unsigned long expires;
1164 ASSERT_RTNL();
1166 spin_lock_bh(&ifp->lock);
1167 state = ifp->state;
1168 ifp->state = INET6_IFADDR_STATE_DEAD;
1169 spin_unlock_bh(&ifp->lock);
1171 if (state == INET6_IFADDR_STATE_DEAD)
1172 goto out;
1174 spin_lock_bh(&addrconf_hash_lock);
1175 hlist_del_init_rcu(&ifp->addr_lst);
1176 spin_unlock_bh(&addrconf_hash_lock);
1178 write_lock_bh(&ifp->idev->lock);
1180 if (ifp->flags&IFA_F_TEMPORARY) {
1181 list_del(&ifp->tmp_list);
1182 if (ifp->ifpub) {
1183 in6_ifa_put(ifp->ifpub);
1184 ifp->ifpub = NULL;
1186 __in6_ifa_put(ifp);
1189 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1190 action = check_cleanup_prefix_route(ifp, &expires);
1192 list_del_init(&ifp->if_list);
1193 __in6_ifa_put(ifp);
1195 write_unlock_bh(&ifp->idev->lock);
1197 addrconf_del_dad_work(ifp);
1199 ipv6_ifa_notify(RTM_DELADDR, ifp);
1201 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1203 if (action != CLEANUP_PREFIX_RT_NOP) {
1204 cleanup_prefix_route(ifp, expires,
1205 action == CLEANUP_PREFIX_RT_DEL);
1208 /* clean up prefsrc entries */
1209 rt6_remove_prefsrc(ifp);
1210 out:
1211 in6_ifa_put(ifp);
1214 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1216 struct inet6_dev *idev = ifp->idev;
1217 struct in6_addr addr, *tmpaddr;
1218 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1219 unsigned long regen_advance;
1220 int tmp_plen;
1221 int ret = 0;
1222 u32 addr_flags;
1223 unsigned long now = jiffies;
1224 long max_desync_factor;
1225 s32 cnf_temp_preferred_lft;
1227 write_lock_bh(&idev->lock);
1228 if (ift) {
1229 spin_lock_bh(&ift->lock);
1230 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1231 spin_unlock_bh(&ift->lock);
1232 tmpaddr = &addr;
1233 } else {
1234 tmpaddr = NULL;
1236 retry:
1237 in6_dev_hold(idev);
1238 if (idev->cnf.use_tempaddr <= 0) {
1239 write_unlock_bh(&idev->lock);
1240 pr_info("%s: use_tempaddr is disabled\n", __func__);
1241 in6_dev_put(idev);
1242 ret = -1;
1243 goto out;
1245 spin_lock_bh(&ifp->lock);
1246 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1247 idev->cnf.use_tempaddr = -1; /*XXX*/
1248 spin_unlock_bh(&ifp->lock);
1249 write_unlock_bh(&idev->lock);
1250 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1251 __func__);
1252 in6_dev_put(idev);
1253 ret = -1;
1254 goto out;
1256 in6_ifa_hold(ifp);
1257 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1258 ipv6_try_regen_rndid(idev, tmpaddr);
1259 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1260 age = (now - ifp->tstamp) / HZ;
1262 regen_advance = idev->cnf.regen_max_retry *
1263 idev->cnf.dad_transmits *
1264 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1266 /* recalculate max_desync_factor each time and update
1267 * idev->desync_factor if it's larger
1269 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1270 max_desync_factor = min_t(__u32,
1271 idev->cnf.max_desync_factor,
1272 cnf_temp_preferred_lft - regen_advance);
1274 if (unlikely(idev->desync_factor > max_desync_factor)) {
1275 if (max_desync_factor > 0) {
1276 get_random_bytes(&idev->desync_factor,
1277 sizeof(idev->desync_factor));
1278 idev->desync_factor %= max_desync_factor;
1279 } else {
1280 idev->desync_factor = 0;
1284 tmp_valid_lft = min_t(__u32,
1285 ifp->valid_lft,
1286 idev->cnf.temp_valid_lft + age);
1287 tmp_prefered_lft = cnf_temp_preferred_lft + age -
1288 idev->desync_factor;
1289 tmp_prefered_lft = min_t(__u32, ifp->prefered_lft, tmp_prefered_lft);
1290 tmp_plen = ifp->prefix_len;
1291 tmp_tstamp = ifp->tstamp;
1292 spin_unlock_bh(&ifp->lock);
1294 write_unlock_bh(&idev->lock);
1296 /* A temporary address is created only if this calculated Preferred
1297 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1298 * an implementation must not create a temporary address with a zero
1299 * Preferred Lifetime.
1300 * Use age calculation as in addrconf_verify to avoid unnecessary
1301 * temporary addresses being generated.
1303 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1304 if (tmp_prefered_lft <= regen_advance + age) {
1305 in6_ifa_put(ifp);
1306 in6_dev_put(idev);
1307 ret = -1;
1308 goto out;
1311 addr_flags = IFA_F_TEMPORARY;
1312 /* set in addrconf_prefix_rcv() */
1313 if (ifp->flags & IFA_F_OPTIMISTIC)
1314 addr_flags |= IFA_F_OPTIMISTIC;
1316 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1317 ipv6_addr_scope(&addr), addr_flags,
1318 tmp_valid_lft, tmp_prefered_lft);
1319 if (IS_ERR(ift)) {
1320 in6_ifa_put(ifp);
1321 in6_dev_put(idev);
1322 pr_info("%s: retry temporary address regeneration\n", __func__);
1323 tmpaddr = &addr;
1324 write_lock_bh(&idev->lock);
1325 goto retry;
1328 spin_lock_bh(&ift->lock);
1329 ift->ifpub = ifp;
1330 ift->cstamp = now;
1331 ift->tstamp = tmp_tstamp;
1332 spin_unlock_bh(&ift->lock);
1334 addrconf_dad_start(ift);
1335 in6_ifa_put(ift);
1336 in6_dev_put(idev);
1337 out:
1338 return ret;
1342 * Choose an appropriate source address (RFC3484)
1344 enum {
1345 IPV6_SADDR_RULE_INIT = 0,
1346 IPV6_SADDR_RULE_LOCAL,
1347 IPV6_SADDR_RULE_SCOPE,
1348 IPV6_SADDR_RULE_PREFERRED,
1349 #ifdef CONFIG_IPV6_MIP6
1350 IPV6_SADDR_RULE_HOA,
1351 #endif
1352 IPV6_SADDR_RULE_OIF,
1353 IPV6_SADDR_RULE_LABEL,
1354 IPV6_SADDR_RULE_PRIVACY,
1355 IPV6_SADDR_RULE_ORCHID,
1356 IPV6_SADDR_RULE_PREFIX,
1357 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1358 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1359 #endif
1360 IPV6_SADDR_RULE_MAX
1363 struct ipv6_saddr_score {
1364 int rule;
1365 int addr_type;
1366 struct inet6_ifaddr *ifa;
1367 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1368 int scopedist;
1369 int matchlen;
1372 struct ipv6_saddr_dst {
1373 const struct in6_addr *addr;
1374 int ifindex;
1375 int scope;
1376 int label;
1377 unsigned int prefs;
1380 static inline int ipv6_saddr_preferred(int type)
1382 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1383 return 1;
1384 return 0;
1387 static inline bool ipv6_use_optimistic_addr(struct inet6_dev *idev)
1389 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1390 return idev && idev->cnf.optimistic_dad && idev->cnf.use_optimistic;
1391 #else
1392 return false;
1393 #endif
1396 static int ipv6_get_saddr_eval(struct net *net,
1397 struct ipv6_saddr_score *score,
1398 struct ipv6_saddr_dst *dst,
1399 int i)
1401 int ret;
1403 if (i <= score->rule) {
1404 switch (i) {
1405 case IPV6_SADDR_RULE_SCOPE:
1406 ret = score->scopedist;
1407 break;
1408 case IPV6_SADDR_RULE_PREFIX:
1409 ret = score->matchlen;
1410 break;
1411 default:
1412 ret = !!test_bit(i, score->scorebits);
1414 goto out;
1417 switch (i) {
1418 case IPV6_SADDR_RULE_INIT:
1419 /* Rule 0: remember if hiscore is not ready yet */
1420 ret = !!score->ifa;
1421 break;
1422 case IPV6_SADDR_RULE_LOCAL:
1423 /* Rule 1: Prefer same address */
1424 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1425 break;
1426 case IPV6_SADDR_RULE_SCOPE:
1427 /* Rule 2: Prefer appropriate scope
1429 * ret
1431 * -1 | d 15
1432 * ---+--+-+---> scope
1434 * | d is scope of the destination.
1435 * B-d | \
1436 * | \ <- smaller scope is better if
1437 * B-15 | \ if scope is enough for destination.
1438 * | ret = B - scope (-1 <= scope >= d <= 15).
1439 * d-C-1 | /
1440 * |/ <- greater is better
1441 * -C / if scope is not enough for destination.
1442 * /| ret = scope - C (-1 <= d < scope <= 15).
1444 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1445 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1446 * Assume B = 0 and we get C > 29.
1448 ret = __ipv6_addr_src_scope(score->addr_type);
1449 if (ret >= dst->scope)
1450 ret = -ret;
1451 else
1452 ret -= 128; /* 30 is enough */
1453 score->scopedist = ret;
1454 break;
1455 case IPV6_SADDR_RULE_PREFERRED:
1457 /* Rule 3: Avoid deprecated and optimistic addresses */
1458 u8 avoid = IFA_F_DEPRECATED;
1460 if (!ipv6_use_optimistic_addr(score->ifa->idev))
1461 avoid |= IFA_F_OPTIMISTIC;
1462 ret = ipv6_saddr_preferred(score->addr_type) ||
1463 !(score->ifa->flags & avoid);
1464 break;
1466 #ifdef CONFIG_IPV6_MIP6
1467 case IPV6_SADDR_RULE_HOA:
1469 /* Rule 4: Prefer home address */
1470 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1471 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1472 break;
1474 #endif
1475 case IPV6_SADDR_RULE_OIF:
1476 /* Rule 5: Prefer outgoing interface */
1477 ret = (!dst->ifindex ||
1478 dst->ifindex == score->ifa->idev->dev->ifindex);
1479 break;
1480 case IPV6_SADDR_RULE_LABEL:
1481 /* Rule 6: Prefer matching label */
1482 ret = ipv6_addr_label(net,
1483 &score->ifa->addr, score->addr_type,
1484 score->ifa->idev->dev->ifindex) == dst->label;
1485 break;
1486 case IPV6_SADDR_RULE_PRIVACY:
1488 /* Rule 7: Prefer public address
1489 * Note: prefer temporary address if use_tempaddr >= 2
1491 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1492 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1493 score->ifa->idev->cnf.use_tempaddr >= 2;
1494 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1495 break;
1497 case IPV6_SADDR_RULE_ORCHID:
1498 /* Rule 8-: Prefer ORCHID vs ORCHID or
1499 * non-ORCHID vs non-ORCHID
1501 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1502 ipv6_addr_orchid(dst->addr));
1503 break;
1504 case IPV6_SADDR_RULE_PREFIX:
1505 /* Rule 8: Use longest matching prefix */
1506 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1507 if (ret > score->ifa->prefix_len)
1508 ret = score->ifa->prefix_len;
1509 score->matchlen = ret;
1510 break;
1511 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1512 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1513 /* Optimistic addresses still have lower precedence than other
1514 * preferred addresses.
1516 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1517 break;
1518 #endif
1519 default:
1520 ret = 0;
1523 if (ret)
1524 __set_bit(i, score->scorebits);
1525 score->rule = i;
1526 out:
1527 return ret;
1530 static int __ipv6_dev_get_saddr(struct net *net,
1531 struct ipv6_saddr_dst *dst,
1532 struct inet6_dev *idev,
1533 struct ipv6_saddr_score *scores,
1534 int hiscore_idx)
1536 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1538 read_lock_bh(&idev->lock);
1539 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1540 int i;
1543 * - Tentative Address (RFC2462 section 5.4)
1544 * - A tentative address is not considered
1545 * "assigned to an interface" in the traditional
1546 * sense, unless it is also flagged as optimistic.
1547 * - Candidate Source Address (section 4)
1548 * - In any case, anycast addresses, multicast
1549 * addresses, and the unspecified address MUST
1550 * NOT be included in a candidate set.
1552 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1553 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1554 continue;
1556 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1558 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1559 score->addr_type & IPV6_ADDR_MULTICAST)) {
1560 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1561 idev->dev->name);
1562 continue;
1565 score->rule = -1;
1566 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1568 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1569 int minihiscore, miniscore;
1571 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1572 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1574 if (minihiscore > miniscore) {
1575 if (i == IPV6_SADDR_RULE_SCOPE &&
1576 score->scopedist > 0) {
1578 * special case:
1579 * each remaining entry
1580 * has too small (not enough)
1581 * scope, because ifa entries
1582 * are sorted by their scope
1583 * values.
1585 goto out;
1587 break;
1588 } else if (minihiscore < miniscore) {
1589 if (hiscore->ifa)
1590 in6_ifa_put(hiscore->ifa);
1592 in6_ifa_hold(score->ifa);
1594 swap(hiscore, score);
1595 hiscore_idx = 1 - hiscore_idx;
1597 /* restore our iterator */
1598 score->ifa = hiscore->ifa;
1600 break;
1604 out:
1605 read_unlock_bh(&idev->lock);
1606 return hiscore_idx;
1609 static int ipv6_get_saddr_master(struct net *net,
1610 const struct net_device *dst_dev,
1611 const struct net_device *master,
1612 struct ipv6_saddr_dst *dst,
1613 struct ipv6_saddr_score *scores,
1614 int hiscore_idx)
1616 struct inet6_dev *idev;
1618 idev = __in6_dev_get(dst_dev);
1619 if (idev)
1620 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1621 scores, hiscore_idx);
1623 idev = __in6_dev_get(master);
1624 if (idev)
1625 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1626 scores, hiscore_idx);
1628 return hiscore_idx;
1631 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1632 const struct in6_addr *daddr, unsigned int prefs,
1633 struct in6_addr *saddr)
1635 struct ipv6_saddr_score scores[2], *hiscore;
1636 struct ipv6_saddr_dst dst;
1637 struct inet6_dev *idev;
1638 struct net_device *dev;
1639 int dst_type;
1640 bool use_oif_addr = false;
1641 int hiscore_idx = 0;
1643 dst_type = __ipv6_addr_type(daddr);
1644 dst.addr = daddr;
1645 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1646 dst.scope = __ipv6_addr_src_scope(dst_type);
1647 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1648 dst.prefs = prefs;
1650 scores[hiscore_idx].rule = -1;
1651 scores[hiscore_idx].ifa = NULL;
1653 rcu_read_lock();
1655 /* Candidate Source Address (section 4)
1656 * - multicast and link-local destination address,
1657 * the set of candidate source address MUST only
1658 * include addresses assigned to interfaces
1659 * belonging to the same link as the outgoing
1660 * interface.
1661 * (- For site-local destination addresses, the
1662 * set of candidate source addresses MUST only
1663 * include addresses assigned to interfaces
1664 * belonging to the same site as the outgoing
1665 * interface.)
1666 * - "It is RECOMMENDED that the candidate source addresses
1667 * be the set of unicast addresses assigned to the
1668 * interface that will be used to send to the destination
1669 * (the 'outgoing' interface)." (RFC 6724)
1671 if (dst_dev) {
1672 idev = __in6_dev_get(dst_dev);
1673 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1674 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1675 (idev && idev->cnf.use_oif_addrs_only)) {
1676 use_oif_addr = true;
1680 if (use_oif_addr) {
1681 if (idev)
1682 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1683 } else {
1684 const struct net_device *master;
1685 int master_idx = 0;
1687 /* if dst_dev exists and is enslaved to an L3 device, then
1688 * prefer addresses from dst_dev and then the master over
1689 * any other enslaved devices in the L3 domain.
1691 master = l3mdev_master_dev_rcu(dst_dev);
1692 if (master) {
1693 master_idx = master->ifindex;
1695 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1696 master, &dst,
1697 scores, hiscore_idx);
1699 if (scores[hiscore_idx].ifa)
1700 goto out;
1703 for_each_netdev_rcu(net, dev) {
1704 /* only consider addresses on devices in the
1705 * same L3 domain
1707 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1708 continue;
1709 idev = __in6_dev_get(dev);
1710 if (!idev)
1711 continue;
1712 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1716 out:
1717 rcu_read_unlock();
1719 hiscore = &scores[hiscore_idx];
1720 if (!hiscore->ifa)
1721 return -EADDRNOTAVAIL;
1723 *saddr = hiscore->ifa->addr;
1724 in6_ifa_put(hiscore->ifa);
1725 return 0;
1727 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1729 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1730 u32 banned_flags)
1732 struct inet6_ifaddr *ifp;
1733 int err = -EADDRNOTAVAIL;
1735 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1736 if (ifp->scope > IFA_LINK)
1737 break;
1738 if (ifp->scope == IFA_LINK &&
1739 !(ifp->flags & banned_flags)) {
1740 *addr = ifp->addr;
1741 err = 0;
1742 break;
1745 return err;
1748 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1749 u32 banned_flags)
1751 struct inet6_dev *idev;
1752 int err = -EADDRNOTAVAIL;
1754 rcu_read_lock();
1755 idev = __in6_dev_get(dev);
1756 if (idev) {
1757 read_lock_bh(&idev->lock);
1758 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1759 read_unlock_bh(&idev->lock);
1761 rcu_read_unlock();
1762 return err;
1765 static int ipv6_count_addresses(struct inet6_dev *idev)
1767 int cnt = 0;
1768 struct inet6_ifaddr *ifp;
1770 read_lock_bh(&idev->lock);
1771 list_for_each_entry(ifp, &idev->addr_list, if_list)
1772 cnt++;
1773 read_unlock_bh(&idev->lock);
1774 return cnt;
1777 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1778 const struct net_device *dev, int strict)
1780 return ipv6_chk_addr_and_flags(net, addr, dev, strict, IFA_F_TENTATIVE);
1782 EXPORT_SYMBOL(ipv6_chk_addr);
1784 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1785 const struct net_device *dev, int strict,
1786 u32 banned_flags)
1788 struct inet6_ifaddr *ifp;
1789 unsigned int hash = inet6_addr_hash(addr);
1790 u32 ifp_flags;
1792 rcu_read_lock_bh();
1793 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1794 if (!net_eq(dev_net(ifp->idev->dev), net))
1795 continue;
1796 /* Decouple optimistic from tentative for evaluation here.
1797 * Ban optimistic addresses explicitly, when required.
1799 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1800 ? (ifp->flags&~IFA_F_TENTATIVE)
1801 : ifp->flags;
1802 if (ipv6_addr_equal(&ifp->addr, addr) &&
1803 !(ifp_flags&banned_flags) &&
1804 (!dev || ifp->idev->dev == dev ||
1805 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1806 rcu_read_unlock_bh();
1807 return 1;
1811 rcu_read_unlock_bh();
1812 return 0;
1814 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1816 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1817 struct net_device *dev)
1819 unsigned int hash = inet6_addr_hash(addr);
1820 struct inet6_ifaddr *ifp;
1822 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1823 if (!net_eq(dev_net(ifp->idev->dev), net))
1824 continue;
1825 if (ipv6_addr_equal(&ifp->addr, addr)) {
1826 if (!dev || ifp->idev->dev == dev)
1827 return true;
1830 return false;
1833 /* Compares an address/prefix_len with addresses on device @dev.
1834 * If one is found it returns true.
1836 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1837 const unsigned int prefix_len, struct net_device *dev)
1839 struct inet6_dev *idev;
1840 struct inet6_ifaddr *ifa;
1841 bool ret = false;
1843 rcu_read_lock();
1844 idev = __in6_dev_get(dev);
1845 if (idev) {
1846 read_lock_bh(&idev->lock);
1847 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1848 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1849 if (ret)
1850 break;
1852 read_unlock_bh(&idev->lock);
1854 rcu_read_unlock();
1856 return ret;
1858 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1860 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1862 struct inet6_dev *idev;
1863 struct inet6_ifaddr *ifa;
1864 int onlink;
1866 onlink = 0;
1867 rcu_read_lock();
1868 idev = __in6_dev_get(dev);
1869 if (idev) {
1870 read_lock_bh(&idev->lock);
1871 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1872 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1873 ifa->prefix_len);
1874 if (onlink)
1875 break;
1877 read_unlock_bh(&idev->lock);
1879 rcu_read_unlock();
1880 return onlink;
1882 EXPORT_SYMBOL(ipv6_chk_prefix);
1884 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1885 struct net_device *dev, int strict)
1887 struct inet6_ifaddr *ifp, *result = NULL;
1888 unsigned int hash = inet6_addr_hash(addr);
1890 rcu_read_lock_bh();
1891 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1892 if (!net_eq(dev_net(ifp->idev->dev), net))
1893 continue;
1894 if (ipv6_addr_equal(&ifp->addr, addr)) {
1895 if (!dev || ifp->idev->dev == dev ||
1896 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1897 result = ifp;
1898 in6_ifa_hold(ifp);
1899 break;
1903 rcu_read_unlock_bh();
1905 return result;
1908 /* Gets referenced address, destroys ifaddr */
1910 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1912 if (dad_failed)
1913 ifp->flags |= IFA_F_DADFAILED;
1915 if (ifp->flags&IFA_F_PERMANENT) {
1916 spin_lock_bh(&ifp->lock);
1917 addrconf_del_dad_work(ifp);
1918 ifp->flags |= IFA_F_TENTATIVE;
1919 spin_unlock_bh(&ifp->lock);
1920 if (dad_failed)
1921 ipv6_ifa_notify(0, ifp);
1922 in6_ifa_put(ifp);
1923 } else if (ifp->flags&IFA_F_TEMPORARY) {
1924 struct inet6_ifaddr *ifpub;
1925 spin_lock_bh(&ifp->lock);
1926 ifpub = ifp->ifpub;
1927 if (ifpub) {
1928 in6_ifa_hold(ifpub);
1929 spin_unlock_bh(&ifp->lock);
1930 ipv6_create_tempaddr(ifpub, ifp);
1931 in6_ifa_put(ifpub);
1932 } else {
1933 spin_unlock_bh(&ifp->lock);
1935 ipv6_del_addr(ifp);
1936 } else {
1937 ipv6_del_addr(ifp);
1941 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1943 int err = -ENOENT;
1945 spin_lock_bh(&ifp->lock);
1946 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1947 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1948 err = 0;
1950 spin_unlock_bh(&ifp->lock);
1952 return err;
1955 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1957 struct inet6_dev *idev = ifp->idev;
1958 struct net *net = dev_net(ifp->idev->dev);
1960 if (addrconf_dad_end(ifp)) {
1961 in6_ifa_put(ifp);
1962 return;
1965 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1966 ifp->idev->dev->name, &ifp->addr);
1968 spin_lock_bh(&ifp->lock);
1970 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
1971 int scope = ifp->scope;
1972 u32 flags = ifp->flags;
1973 struct in6_addr new_addr;
1974 struct inet6_ifaddr *ifp2;
1975 u32 valid_lft, preferred_lft;
1976 int pfxlen = ifp->prefix_len;
1977 int retries = ifp->stable_privacy_retry + 1;
1979 if (retries > net->ipv6.sysctl.idgen_retries) {
1980 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
1981 ifp->idev->dev->name);
1982 goto errdad;
1985 new_addr = ifp->addr;
1986 if (ipv6_generate_stable_address(&new_addr, retries,
1987 idev))
1988 goto errdad;
1990 valid_lft = ifp->valid_lft;
1991 preferred_lft = ifp->prefered_lft;
1993 spin_unlock_bh(&ifp->lock);
1995 if (idev->cnf.max_addresses &&
1996 ipv6_count_addresses(idev) >=
1997 idev->cnf.max_addresses)
1998 goto lock_errdad;
2000 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2001 ifp->idev->dev->name);
2003 ifp2 = ipv6_add_addr(idev, &new_addr, NULL, pfxlen,
2004 scope, flags, valid_lft,
2005 preferred_lft);
2006 if (IS_ERR(ifp2))
2007 goto lock_errdad;
2009 spin_lock_bh(&ifp2->lock);
2010 ifp2->stable_privacy_retry = retries;
2011 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2012 spin_unlock_bh(&ifp2->lock);
2014 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2015 in6_ifa_put(ifp2);
2016 lock_errdad:
2017 spin_lock_bh(&ifp->lock);
2020 errdad:
2021 /* transition from _POSTDAD to _ERRDAD */
2022 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2023 spin_unlock_bh(&ifp->lock);
2025 addrconf_mod_dad_work(ifp, 0);
2026 in6_ifa_put(ifp);
2029 /* Join to solicited addr multicast group.
2030 * caller must hold RTNL */
2031 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2033 struct in6_addr maddr;
2035 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2036 return;
2038 addrconf_addr_solict_mult(addr, &maddr);
2039 ipv6_dev_mc_inc(dev, &maddr);
2042 /* caller must hold RTNL */
2043 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2045 struct in6_addr maddr;
2047 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2048 return;
2050 addrconf_addr_solict_mult(addr, &maddr);
2051 __ipv6_dev_mc_dec(idev, &maddr);
2054 /* caller must hold RTNL */
2055 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2057 struct in6_addr addr;
2059 if (ifp->prefix_len >= 127) /* RFC 6164 */
2060 return;
2061 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2062 if (ipv6_addr_any(&addr))
2063 return;
2064 __ipv6_dev_ac_inc(ifp->idev, &addr);
2067 /* caller must hold RTNL */
2068 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2070 struct in6_addr addr;
2072 if (ifp->prefix_len >= 127) /* RFC 6164 */
2073 return;
2074 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2075 if (ipv6_addr_any(&addr))
2076 return;
2077 __ipv6_dev_ac_dec(ifp->idev, &addr);
2080 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2082 switch (dev->addr_len) {
2083 case ETH_ALEN:
2084 memcpy(eui, dev->dev_addr, 3);
2085 eui[3] = 0xFF;
2086 eui[4] = 0xFE;
2087 memcpy(eui + 5, dev->dev_addr + 3, 3);
2088 break;
2089 case EUI64_ADDR_LEN:
2090 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2091 eui[0] ^= 2;
2092 break;
2093 default:
2094 return -1;
2097 return 0;
2100 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2102 union fwnet_hwaddr *ha;
2104 if (dev->addr_len != FWNET_ALEN)
2105 return -1;
2107 ha = (union fwnet_hwaddr *)dev->dev_addr;
2109 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2110 eui[0] ^= 2;
2111 return 0;
2114 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2116 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2117 if (dev->addr_len != ARCNET_ALEN)
2118 return -1;
2119 memset(eui, 0, 7);
2120 eui[7] = *(u8 *)dev->dev_addr;
2121 return 0;
2124 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2126 if (dev->addr_len != INFINIBAND_ALEN)
2127 return -1;
2128 memcpy(eui, dev->dev_addr + 12, 8);
2129 eui[0] |= 2;
2130 return 0;
2133 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2135 if (addr == 0)
2136 return -1;
2137 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2138 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2139 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2140 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2141 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2142 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2143 eui[1] = 0;
2144 eui[2] = 0x5E;
2145 eui[3] = 0xFE;
2146 memcpy(eui + 4, &addr, 4);
2147 return 0;
2150 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2152 if (dev->priv_flags & IFF_ISATAP)
2153 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2154 return -1;
2157 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2159 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2162 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2164 memcpy(eui, dev->perm_addr, 3);
2165 memcpy(eui + 5, dev->perm_addr + 3, 3);
2166 eui[3] = 0xFF;
2167 eui[4] = 0xFE;
2168 eui[0] ^= 2;
2169 return 0;
2172 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2174 switch (dev->type) {
2175 case ARPHRD_ETHER:
2176 case ARPHRD_FDDI:
2177 return addrconf_ifid_eui48(eui, dev);
2178 case ARPHRD_ARCNET:
2179 return addrconf_ifid_arcnet(eui, dev);
2180 case ARPHRD_INFINIBAND:
2181 return addrconf_ifid_infiniband(eui, dev);
2182 case ARPHRD_SIT:
2183 return addrconf_ifid_sit(eui, dev);
2184 case ARPHRD_IPGRE:
2185 case ARPHRD_TUNNEL:
2186 return addrconf_ifid_gre(eui, dev);
2187 case ARPHRD_6LOWPAN:
2188 return addrconf_ifid_6lowpan(eui, dev);
2189 case ARPHRD_IEEE1394:
2190 return addrconf_ifid_ieee1394(eui, dev);
2191 case ARPHRD_TUNNEL6:
2192 case ARPHRD_IP6GRE:
2193 return addrconf_ifid_ip6tnl(eui, dev);
2195 return -1;
2198 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2200 int err = -1;
2201 struct inet6_ifaddr *ifp;
2203 read_lock_bh(&idev->lock);
2204 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2205 if (ifp->scope > IFA_LINK)
2206 break;
2207 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2208 memcpy(eui, ifp->addr.s6_addr+8, 8);
2209 err = 0;
2210 break;
2213 read_unlock_bh(&idev->lock);
2214 return err;
2217 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2218 static void ipv6_regen_rndid(struct inet6_dev *idev)
2220 regen:
2221 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2222 idev->rndid[0] &= ~0x02;
2225 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2226 * check if generated address is not inappropriate
2228 * - Reserved subnet anycast (RFC 2526)
2229 * 11111101 11....11 1xxxxxxx
2230 * - ISATAP (RFC4214) 6.1
2231 * 00-00-5E-FE-xx-xx-xx-xx
2232 * - value 0
2233 * - XXX: already assigned to an address on the device
2235 if (idev->rndid[0] == 0xfd &&
2236 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2237 (idev->rndid[7]&0x80))
2238 goto regen;
2239 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2240 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2241 goto regen;
2242 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2243 goto regen;
2247 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2249 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2250 ipv6_regen_rndid(idev);
2254 * Add prefix route.
2257 static void
2258 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
2259 unsigned long expires, u32 flags)
2261 struct fib6_config cfg = {
2262 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2263 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2264 .fc_ifindex = dev->ifindex,
2265 .fc_expires = expires,
2266 .fc_dst_len = plen,
2267 .fc_flags = RTF_UP | flags,
2268 .fc_nlinfo.nl_net = dev_net(dev),
2269 .fc_protocol = RTPROT_KERNEL,
2272 cfg.fc_dst = *pfx;
2274 /* Prevent useless cloning on PtP SIT.
2275 This thing is done here expecting that the whole
2276 class of non-broadcast devices need not cloning.
2278 #if IS_ENABLED(CONFIG_IPV6_SIT)
2279 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2280 cfg.fc_flags |= RTF_NONEXTHOP;
2281 #endif
2283 ip6_route_add(&cfg);
2287 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2288 int plen,
2289 const struct net_device *dev,
2290 u32 flags, u32 noflags)
2292 struct fib6_node *fn;
2293 struct rt6_info *rt = NULL;
2294 struct fib6_table *table;
2295 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2297 table = fib6_get_table(dev_net(dev), tb_id);
2298 if (!table)
2299 return NULL;
2301 read_lock_bh(&table->tb6_lock);
2302 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2303 if (!fn)
2304 goto out;
2306 noflags |= RTF_CACHE;
2307 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2308 if (rt->dst.dev->ifindex != dev->ifindex)
2309 continue;
2310 if ((rt->rt6i_flags & flags) != flags)
2311 continue;
2312 if ((rt->rt6i_flags & noflags) != 0)
2313 continue;
2314 dst_hold(&rt->dst);
2315 break;
2317 out:
2318 read_unlock_bh(&table->tb6_lock);
2319 return rt;
2323 /* Create "default" multicast route to the interface */
2325 static void addrconf_add_mroute(struct net_device *dev)
2327 struct fib6_config cfg = {
2328 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2329 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2330 .fc_ifindex = dev->ifindex,
2331 .fc_dst_len = 8,
2332 .fc_flags = RTF_UP,
2333 .fc_nlinfo.nl_net = dev_net(dev),
2336 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2338 ip6_route_add(&cfg);
2341 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2343 struct inet6_dev *idev;
2345 ASSERT_RTNL();
2347 idev = ipv6_find_idev(dev);
2348 if (!idev)
2349 return ERR_PTR(-ENOBUFS);
2351 if (idev->cnf.disable_ipv6)
2352 return ERR_PTR(-EACCES);
2354 /* Add default multicast route */
2355 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2356 addrconf_add_mroute(dev);
2358 return idev;
2361 static void manage_tempaddrs(struct inet6_dev *idev,
2362 struct inet6_ifaddr *ifp,
2363 __u32 valid_lft, __u32 prefered_lft,
2364 bool create, unsigned long now)
2366 u32 flags;
2367 struct inet6_ifaddr *ift;
2369 read_lock_bh(&idev->lock);
2370 /* update all temporary addresses in the list */
2371 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2372 int age, max_valid, max_prefered;
2374 if (ifp != ift->ifpub)
2375 continue;
2377 /* RFC 4941 section 3.3:
2378 * If a received option will extend the lifetime of a public
2379 * address, the lifetimes of temporary addresses should
2380 * be extended, subject to the overall constraint that no
2381 * temporary addresses should ever remain "valid" or "preferred"
2382 * for a time longer than (TEMP_VALID_LIFETIME) or
2383 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2385 age = (now - ift->cstamp) / HZ;
2386 max_valid = idev->cnf.temp_valid_lft - age;
2387 if (max_valid < 0)
2388 max_valid = 0;
2390 max_prefered = idev->cnf.temp_prefered_lft -
2391 idev->desync_factor - age;
2392 if (max_prefered < 0)
2393 max_prefered = 0;
2395 if (valid_lft > max_valid)
2396 valid_lft = max_valid;
2398 if (prefered_lft > max_prefered)
2399 prefered_lft = max_prefered;
2401 spin_lock(&ift->lock);
2402 flags = ift->flags;
2403 ift->valid_lft = valid_lft;
2404 ift->prefered_lft = prefered_lft;
2405 ift->tstamp = now;
2406 if (prefered_lft > 0)
2407 ift->flags &= ~IFA_F_DEPRECATED;
2409 spin_unlock(&ift->lock);
2410 if (!(flags&IFA_F_TENTATIVE))
2411 ipv6_ifa_notify(0, ift);
2414 if ((create || list_empty(&idev->tempaddr_list)) &&
2415 idev->cnf.use_tempaddr > 0) {
2416 /* When a new public address is created as described
2417 * in [ADDRCONF], also create a new temporary address.
2418 * Also create a temporary address if it's enabled but
2419 * no temporary address currently exists.
2421 read_unlock_bh(&idev->lock);
2422 ipv6_create_tempaddr(ifp, NULL);
2423 } else {
2424 read_unlock_bh(&idev->lock);
2428 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2430 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2431 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2434 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2435 const struct prefix_info *pinfo,
2436 struct inet6_dev *in6_dev,
2437 const struct in6_addr *addr, int addr_type,
2438 u32 addr_flags, bool sllao, bool tokenized,
2439 __u32 valid_lft, u32 prefered_lft)
2441 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2442 int create = 0, update_lft = 0;
2444 if (!ifp && valid_lft) {
2445 int max_addresses = in6_dev->cnf.max_addresses;
2447 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2448 if (in6_dev->cnf.optimistic_dad &&
2449 !net->ipv6.devconf_all->forwarding && sllao)
2450 addr_flags |= IFA_F_OPTIMISTIC;
2451 #endif
2453 /* Do not allow to create too much of autoconfigured
2454 * addresses; this would be too easy way to crash kernel.
2456 if (!max_addresses ||
2457 ipv6_count_addresses(in6_dev) < max_addresses)
2458 ifp = ipv6_add_addr(in6_dev, addr, NULL,
2459 pinfo->prefix_len,
2460 addr_type&IPV6_ADDR_SCOPE_MASK,
2461 addr_flags, valid_lft,
2462 prefered_lft);
2464 if (IS_ERR_OR_NULL(ifp))
2465 return -1;
2467 update_lft = 0;
2468 create = 1;
2469 spin_lock_bh(&ifp->lock);
2470 ifp->flags |= IFA_F_MANAGETEMPADDR;
2471 ifp->cstamp = jiffies;
2472 ifp->tokenized = tokenized;
2473 spin_unlock_bh(&ifp->lock);
2474 addrconf_dad_start(ifp);
2477 if (ifp) {
2478 u32 flags;
2479 unsigned long now;
2480 u32 stored_lft;
2482 /* update lifetime (RFC2462 5.5.3 e) */
2483 spin_lock_bh(&ifp->lock);
2484 now = jiffies;
2485 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2486 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2487 else
2488 stored_lft = 0;
2489 if (!update_lft && !create && stored_lft) {
2490 const u32 minimum_lft = min_t(u32,
2491 stored_lft, MIN_VALID_LIFETIME);
2492 valid_lft = max(valid_lft, minimum_lft);
2494 /* RFC4862 Section 5.5.3e:
2495 * "Note that the preferred lifetime of the
2496 * corresponding address is always reset to
2497 * the Preferred Lifetime in the received
2498 * Prefix Information option, regardless of
2499 * whether the valid lifetime is also reset or
2500 * ignored."
2502 * So we should always update prefered_lft here.
2504 update_lft = 1;
2507 if (update_lft) {
2508 ifp->valid_lft = valid_lft;
2509 ifp->prefered_lft = prefered_lft;
2510 ifp->tstamp = now;
2511 flags = ifp->flags;
2512 ifp->flags &= ~IFA_F_DEPRECATED;
2513 spin_unlock_bh(&ifp->lock);
2515 if (!(flags&IFA_F_TENTATIVE))
2516 ipv6_ifa_notify(0, ifp);
2517 } else
2518 spin_unlock_bh(&ifp->lock);
2520 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2521 create, now);
2523 in6_ifa_put(ifp);
2524 addrconf_verify();
2527 return 0;
2529 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2531 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2533 struct prefix_info *pinfo;
2534 __u32 valid_lft;
2535 __u32 prefered_lft;
2536 int addr_type, err;
2537 u32 addr_flags = 0;
2538 struct inet6_dev *in6_dev;
2539 struct net *net = dev_net(dev);
2541 pinfo = (struct prefix_info *) opt;
2543 if (len < sizeof(struct prefix_info)) {
2544 ADBG("addrconf: prefix option too short\n");
2545 return;
2549 * Validation checks ([ADDRCONF], page 19)
2552 addr_type = ipv6_addr_type(&pinfo->prefix);
2554 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2555 return;
2557 valid_lft = ntohl(pinfo->valid);
2558 prefered_lft = ntohl(pinfo->prefered);
2560 if (prefered_lft > valid_lft) {
2561 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2562 return;
2565 in6_dev = in6_dev_get(dev);
2567 if (!in6_dev) {
2568 net_dbg_ratelimited("addrconf: device %s not configured\n",
2569 dev->name);
2570 return;
2574 * Two things going on here:
2575 * 1) Add routes for on-link prefixes
2576 * 2) Configure prefixes with the auto flag set
2579 if (pinfo->onlink) {
2580 struct rt6_info *rt;
2581 unsigned long rt_expires;
2583 /* Avoid arithmetic overflow. Really, we could
2584 * save rt_expires in seconds, likely valid_lft,
2585 * but it would require division in fib gc, that it
2586 * not good.
2588 if (HZ > USER_HZ)
2589 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2590 else
2591 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2593 if (addrconf_finite_timeout(rt_expires))
2594 rt_expires *= HZ;
2596 rt = addrconf_get_prefix_route(&pinfo->prefix,
2597 pinfo->prefix_len,
2598 dev,
2599 RTF_ADDRCONF | RTF_PREFIX_RT,
2600 RTF_GATEWAY | RTF_DEFAULT);
2602 if (rt) {
2603 /* Autoconf prefix route */
2604 if (valid_lft == 0) {
2605 ip6_del_rt(rt);
2606 rt = NULL;
2607 } else if (addrconf_finite_timeout(rt_expires)) {
2608 /* not infinity */
2609 rt6_set_expires(rt, jiffies + rt_expires);
2610 } else {
2611 rt6_clean_expires(rt);
2613 } else if (valid_lft) {
2614 clock_t expires = 0;
2615 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2616 if (addrconf_finite_timeout(rt_expires)) {
2617 /* not infinity */
2618 flags |= RTF_EXPIRES;
2619 expires = jiffies_to_clock_t(rt_expires);
2621 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2622 dev, expires, flags);
2624 ip6_rt_put(rt);
2627 /* Try to figure out our local address for this prefix */
2629 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2630 struct in6_addr addr;
2631 bool tokenized = false, dev_addr_generated = false;
2633 if (pinfo->prefix_len == 64) {
2634 memcpy(&addr, &pinfo->prefix, 8);
2636 if (!ipv6_addr_any(&in6_dev->token)) {
2637 read_lock_bh(&in6_dev->lock);
2638 memcpy(addr.s6_addr + 8,
2639 in6_dev->token.s6_addr + 8, 8);
2640 read_unlock_bh(&in6_dev->lock);
2641 tokenized = true;
2642 } else if (is_addr_mode_generate_stable(in6_dev) &&
2643 !ipv6_generate_stable_address(&addr, 0,
2644 in6_dev)) {
2645 addr_flags |= IFA_F_STABLE_PRIVACY;
2646 goto ok;
2647 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2648 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2649 goto put;
2650 } else {
2651 dev_addr_generated = true;
2653 goto ok;
2655 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2656 pinfo->prefix_len);
2657 goto put;
2660 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2661 &addr, addr_type,
2662 addr_flags, sllao,
2663 tokenized, valid_lft,
2664 prefered_lft);
2665 if (err)
2666 goto put;
2668 /* Ignore error case here because previous prefix add addr was
2669 * successful which will be notified.
2671 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2672 addr_type, addr_flags, sllao,
2673 tokenized, valid_lft,
2674 prefered_lft,
2675 dev_addr_generated);
2677 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2678 put:
2679 in6_dev_put(in6_dev);
2683 * Set destination address.
2684 * Special case for SIT interfaces where we create a new "virtual"
2685 * device.
2687 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2689 struct in6_ifreq ireq;
2690 struct net_device *dev;
2691 int err = -EINVAL;
2693 rtnl_lock();
2695 err = -EFAULT;
2696 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2697 goto err_exit;
2699 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2701 err = -ENODEV;
2702 if (!dev)
2703 goto err_exit;
2705 #if IS_ENABLED(CONFIG_IPV6_SIT)
2706 if (dev->type == ARPHRD_SIT) {
2707 const struct net_device_ops *ops = dev->netdev_ops;
2708 struct ifreq ifr;
2709 struct ip_tunnel_parm p;
2711 err = -EADDRNOTAVAIL;
2712 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2713 goto err_exit;
2715 memset(&p, 0, sizeof(p));
2716 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2717 p.iph.saddr = 0;
2718 p.iph.version = 4;
2719 p.iph.ihl = 5;
2720 p.iph.protocol = IPPROTO_IPV6;
2721 p.iph.ttl = 64;
2722 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2724 if (ops->ndo_do_ioctl) {
2725 mm_segment_t oldfs = get_fs();
2727 set_fs(KERNEL_DS);
2728 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2729 set_fs(oldfs);
2730 } else
2731 err = -EOPNOTSUPP;
2733 if (err == 0) {
2734 err = -ENOBUFS;
2735 dev = __dev_get_by_name(net, p.name);
2736 if (!dev)
2737 goto err_exit;
2738 err = dev_open(dev);
2741 #endif
2743 err_exit:
2744 rtnl_unlock();
2745 return err;
2748 static int ipv6_mc_config(struct sock *sk, bool join,
2749 const struct in6_addr *addr, int ifindex)
2751 int ret;
2753 ASSERT_RTNL();
2755 lock_sock(sk);
2756 if (join)
2757 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2758 else
2759 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2760 release_sock(sk);
2762 return ret;
2766 * Manual configuration of address on an interface
2768 static int inet6_addr_add(struct net *net, int ifindex,
2769 const struct in6_addr *pfx,
2770 const struct in6_addr *peer_pfx,
2771 unsigned int plen, __u32 ifa_flags,
2772 __u32 prefered_lft, __u32 valid_lft)
2774 struct inet6_ifaddr *ifp;
2775 struct inet6_dev *idev;
2776 struct net_device *dev;
2777 unsigned long timeout;
2778 clock_t expires;
2779 int scope;
2780 u32 flags;
2782 ASSERT_RTNL();
2784 if (plen > 128)
2785 return -EINVAL;
2787 /* check the lifetime */
2788 if (!valid_lft || prefered_lft > valid_lft)
2789 return -EINVAL;
2791 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2792 return -EINVAL;
2794 dev = __dev_get_by_index(net, ifindex);
2795 if (!dev)
2796 return -ENODEV;
2798 idev = addrconf_add_dev(dev);
2799 if (IS_ERR(idev))
2800 return PTR_ERR(idev);
2802 if (ifa_flags & IFA_F_MCAUTOJOIN) {
2803 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2804 true, pfx, ifindex);
2806 if (ret < 0)
2807 return ret;
2810 scope = ipv6_addr_scope(pfx);
2812 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2813 if (addrconf_finite_timeout(timeout)) {
2814 expires = jiffies_to_clock_t(timeout * HZ);
2815 valid_lft = timeout;
2816 flags = RTF_EXPIRES;
2817 } else {
2818 expires = 0;
2819 flags = 0;
2820 ifa_flags |= IFA_F_PERMANENT;
2823 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2824 if (addrconf_finite_timeout(timeout)) {
2825 if (timeout == 0)
2826 ifa_flags |= IFA_F_DEPRECATED;
2827 prefered_lft = timeout;
2830 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2831 valid_lft, prefered_lft);
2833 if (!IS_ERR(ifp)) {
2834 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2835 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2836 expires, flags);
2840 * Note that section 3.1 of RFC 4429 indicates
2841 * that the Optimistic flag should not be set for
2842 * manually configured addresses
2844 addrconf_dad_start(ifp);
2845 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2846 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2847 true, jiffies);
2848 in6_ifa_put(ifp);
2849 addrconf_verify_rtnl();
2850 return 0;
2851 } else if (ifa_flags & IFA_F_MCAUTOJOIN) {
2852 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2853 false, pfx, ifindex);
2856 return PTR_ERR(ifp);
2859 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2860 const struct in6_addr *pfx, unsigned int plen)
2862 struct inet6_ifaddr *ifp;
2863 struct inet6_dev *idev;
2864 struct net_device *dev;
2866 if (plen > 128)
2867 return -EINVAL;
2869 dev = __dev_get_by_index(net, ifindex);
2870 if (!dev)
2871 return -ENODEV;
2873 idev = __in6_dev_get(dev);
2874 if (!idev)
2875 return -ENXIO;
2877 read_lock_bh(&idev->lock);
2878 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2879 if (ifp->prefix_len == plen &&
2880 ipv6_addr_equal(pfx, &ifp->addr)) {
2881 in6_ifa_hold(ifp);
2882 read_unlock_bh(&idev->lock);
2884 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2885 (ifa_flags & IFA_F_MANAGETEMPADDR))
2886 manage_tempaddrs(idev, ifp, 0, 0, false,
2887 jiffies);
2888 ipv6_del_addr(ifp);
2889 addrconf_verify_rtnl();
2890 if (ipv6_addr_is_multicast(pfx)) {
2891 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2892 false, pfx, dev->ifindex);
2894 return 0;
2897 read_unlock_bh(&idev->lock);
2898 return -EADDRNOTAVAIL;
2902 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2904 struct in6_ifreq ireq;
2905 int err;
2907 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2908 return -EPERM;
2910 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2911 return -EFAULT;
2913 rtnl_lock();
2914 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2915 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2916 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2917 rtnl_unlock();
2918 return err;
2921 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2923 struct in6_ifreq ireq;
2924 int err;
2926 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2927 return -EPERM;
2929 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2930 return -EFAULT;
2932 rtnl_lock();
2933 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2934 ireq.ifr6_prefixlen);
2935 rtnl_unlock();
2936 return err;
2939 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2940 int plen, int scope)
2942 struct inet6_ifaddr *ifp;
2944 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2945 scope, IFA_F_PERMANENT,
2946 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2947 if (!IS_ERR(ifp)) {
2948 spin_lock_bh(&ifp->lock);
2949 ifp->flags &= ~IFA_F_TENTATIVE;
2950 spin_unlock_bh(&ifp->lock);
2951 rt_genid_bump_ipv6(dev_net(idev->dev));
2952 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2953 in6_ifa_put(ifp);
2957 #if IS_ENABLED(CONFIG_IPV6_SIT)
2958 static void sit_add_v4_addrs(struct inet6_dev *idev)
2960 struct in6_addr addr;
2961 struct net_device *dev;
2962 struct net *net = dev_net(idev->dev);
2963 int scope, plen;
2964 u32 pflags = 0;
2966 ASSERT_RTNL();
2968 memset(&addr, 0, sizeof(struct in6_addr));
2969 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2971 if (idev->dev->flags&IFF_POINTOPOINT) {
2972 addr.s6_addr32[0] = htonl(0xfe800000);
2973 scope = IFA_LINK;
2974 plen = 64;
2975 } else {
2976 scope = IPV6_ADDR_COMPATv4;
2977 plen = 96;
2978 pflags |= RTF_NONEXTHOP;
2981 if (addr.s6_addr32[3]) {
2982 add_addr(idev, &addr, plen, scope);
2983 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2984 return;
2987 for_each_netdev(net, dev) {
2988 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2989 if (in_dev && (dev->flags & IFF_UP)) {
2990 struct in_ifaddr *ifa;
2992 int flag = scope;
2994 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2996 addr.s6_addr32[3] = ifa->ifa_local;
2998 if (ifa->ifa_scope == RT_SCOPE_LINK)
2999 continue;
3000 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3001 if (idev->dev->flags&IFF_POINTOPOINT)
3002 continue;
3003 flag |= IFA_HOST;
3006 add_addr(idev, &addr, plen, flag);
3007 addrconf_prefix_route(&addr, plen, idev->dev, 0,
3008 pflags);
3013 #endif
3015 static void init_loopback(struct net_device *dev)
3017 struct inet6_dev *idev;
3018 struct net_device *sp_dev;
3019 struct inet6_ifaddr *sp_ifa;
3020 struct rt6_info *sp_rt;
3022 /* ::1 */
3024 ASSERT_RTNL();
3026 idev = ipv6_find_idev(dev);
3027 if (!idev) {
3028 pr_debug("%s: add_dev failed\n", __func__);
3029 return;
3032 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3034 /* Add routes to other interface's IPv6 addresses */
3035 for_each_netdev(dev_net(dev), sp_dev) {
3036 if (!strcmp(sp_dev->name, dev->name))
3037 continue;
3039 idev = __in6_dev_get(sp_dev);
3040 if (!idev)
3041 continue;
3043 read_lock_bh(&idev->lock);
3044 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
3046 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
3047 continue;
3049 if (sp_ifa->rt) {
3050 /* This dst has been added to garbage list when
3051 * lo device down, release this obsolete dst and
3052 * reallocate a new router for ifa.
3054 if (!atomic_read(&sp_ifa->rt->rt6i_ref)) {
3055 ip6_rt_put(sp_ifa->rt);
3056 sp_ifa->rt = NULL;
3057 } else {
3058 continue;
3062 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
3064 /* Failure cases are ignored */
3065 if (!IS_ERR(sp_rt)) {
3066 sp_ifa->rt = sp_rt;
3067 ip6_ins_rt(sp_rt);
3070 read_unlock_bh(&idev->lock);
3074 void addrconf_add_linklocal(struct inet6_dev *idev,
3075 const struct in6_addr *addr, u32 flags)
3077 struct inet6_ifaddr *ifp;
3078 u32 addr_flags = flags | IFA_F_PERMANENT;
3080 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3081 if (idev->cnf.optimistic_dad &&
3082 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3083 addr_flags |= IFA_F_OPTIMISTIC;
3084 #endif
3086 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
3087 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
3088 if (!IS_ERR(ifp)) {
3089 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
3090 addrconf_dad_start(ifp);
3091 in6_ifa_put(ifp);
3094 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3096 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3098 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3099 return true;
3101 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3102 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3103 return true;
3105 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3106 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3107 return true;
3109 return false;
3112 static int ipv6_generate_stable_address(struct in6_addr *address,
3113 u8 dad_count,
3114 const struct inet6_dev *idev)
3116 static DEFINE_SPINLOCK(lock);
3117 static __u32 digest[SHA_DIGEST_WORDS];
3118 static __u32 workspace[SHA_WORKSPACE_WORDS];
3120 static union {
3121 char __data[SHA_MESSAGE_BYTES];
3122 struct {
3123 struct in6_addr secret;
3124 __be32 prefix[2];
3125 unsigned char hwaddr[MAX_ADDR_LEN];
3126 u8 dad_count;
3127 } __packed;
3128 } data;
3130 struct in6_addr secret;
3131 struct in6_addr temp;
3132 struct net *net = dev_net(idev->dev);
3134 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3136 if (idev->cnf.stable_secret.initialized)
3137 secret = idev->cnf.stable_secret.secret;
3138 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3139 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3140 else
3141 return -1;
3143 retry:
3144 spin_lock_bh(&lock);
3146 sha_init(digest);
3147 memset(&data, 0, sizeof(data));
3148 memset(workspace, 0, sizeof(workspace));
3149 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3150 data.prefix[0] = address->s6_addr32[0];
3151 data.prefix[1] = address->s6_addr32[1];
3152 data.secret = secret;
3153 data.dad_count = dad_count;
3155 sha_transform(digest, data.__data, workspace);
3157 temp = *address;
3158 temp.s6_addr32[2] = (__force __be32)digest[0];
3159 temp.s6_addr32[3] = (__force __be32)digest[1];
3161 spin_unlock_bh(&lock);
3163 if (ipv6_reserved_interfaceid(temp)) {
3164 dad_count++;
3165 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3166 return -1;
3167 goto retry;
3170 *address = temp;
3171 return 0;
3174 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3176 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3178 if (s->initialized)
3179 return;
3180 s = &idev->cnf.stable_secret;
3181 get_random_bytes(&s->secret, sizeof(s->secret));
3182 s->initialized = true;
3185 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3187 struct in6_addr addr;
3189 /* no link local addresses on L3 master devices */
3190 if (netif_is_l3_master(idev->dev))
3191 return;
3193 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3195 switch (idev->cnf.addr_gen_mode) {
3196 case IN6_ADDR_GEN_MODE_RANDOM:
3197 ipv6_gen_mode_random_init(idev);
3198 /* fallthrough */
3199 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3200 if (!ipv6_generate_stable_address(&addr, 0, idev))
3201 addrconf_add_linklocal(idev, &addr,
3202 IFA_F_STABLE_PRIVACY);
3203 else if (prefix_route)
3204 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3205 break;
3206 case IN6_ADDR_GEN_MODE_EUI64:
3207 /* addrconf_add_linklocal also adds a prefix_route and we
3208 * only need to care about prefix routes if ipv6_generate_eui64
3209 * couldn't generate one.
3211 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3212 addrconf_add_linklocal(idev, &addr, 0);
3213 else if (prefix_route)
3214 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
3215 break;
3216 case IN6_ADDR_GEN_MODE_NONE:
3217 default:
3218 /* will not add any link local address */
3219 break;
3223 static void addrconf_dev_config(struct net_device *dev)
3225 struct inet6_dev *idev;
3227 ASSERT_RTNL();
3229 if ((dev->type != ARPHRD_ETHER) &&
3230 (dev->type != ARPHRD_FDDI) &&
3231 (dev->type != ARPHRD_ARCNET) &&
3232 (dev->type != ARPHRD_INFINIBAND) &&
3233 (dev->type != ARPHRD_IEEE1394) &&
3234 (dev->type != ARPHRD_TUNNEL6) &&
3235 (dev->type != ARPHRD_6LOWPAN) &&
3236 (dev->type != ARPHRD_IP6GRE) &&
3237 (dev->type != ARPHRD_IPGRE) &&
3238 (dev->type != ARPHRD_TUNNEL) &&
3239 (dev->type != ARPHRD_NONE)) {
3240 /* Alas, we support only Ethernet autoconfiguration. */
3241 return;
3244 idev = addrconf_add_dev(dev);
3245 if (IS_ERR(idev))
3246 return;
3248 /* this device type has no EUI support */
3249 if (dev->type == ARPHRD_NONE &&
3250 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3251 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3253 addrconf_addr_gen(idev, false);
3256 #if IS_ENABLED(CONFIG_IPV6_SIT)
3257 static void addrconf_sit_config(struct net_device *dev)
3259 struct inet6_dev *idev;
3261 ASSERT_RTNL();
3264 * Configure the tunnel with one of our IPv4
3265 * addresses... we should configure all of
3266 * our v4 addrs in the tunnel
3269 idev = ipv6_find_idev(dev);
3270 if (!idev) {
3271 pr_debug("%s: add_dev failed\n", __func__);
3272 return;
3275 if (dev->priv_flags & IFF_ISATAP) {
3276 addrconf_addr_gen(idev, false);
3277 return;
3280 sit_add_v4_addrs(idev);
3282 if (dev->flags&IFF_POINTOPOINT)
3283 addrconf_add_mroute(dev);
3285 #endif
3287 #if IS_ENABLED(CONFIG_NET_IPGRE)
3288 static void addrconf_gre_config(struct net_device *dev)
3290 struct inet6_dev *idev;
3292 ASSERT_RTNL();
3294 idev = ipv6_find_idev(dev);
3295 if (!idev) {
3296 pr_debug("%s: add_dev failed\n", __func__);
3297 return;
3300 addrconf_addr_gen(idev, true);
3301 if (dev->flags & IFF_POINTOPOINT)
3302 addrconf_add_mroute(dev);
3304 #endif
3306 static int fixup_permanent_addr(struct inet6_dev *idev,
3307 struct inet6_ifaddr *ifp)
3309 /* rt6i_ref == 0 means the host route was removed from the
3310 * FIB, for example, if 'lo' device is taken down. In that
3311 * case regenerate the host route.
3313 if (!ifp->rt || !atomic_read(&ifp->rt->rt6i_ref)) {
3314 struct rt6_info *rt, *prev;
3316 rt = addrconf_dst_alloc(idev, &ifp->addr, false);
3317 if (unlikely(IS_ERR(rt)))
3318 return PTR_ERR(rt);
3320 /* ifp->rt can be accessed outside of rtnl */
3321 spin_lock(&ifp->lock);
3322 prev = ifp->rt;
3323 ifp->rt = rt;
3324 spin_unlock(&ifp->lock);
3326 ip6_rt_put(prev);
3329 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3330 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3331 idev->dev, 0, 0);
3334 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3335 addrconf_dad_start(ifp);
3337 return 0;
3340 static void addrconf_permanent_addr(struct net_device *dev)
3342 struct inet6_ifaddr *ifp, *tmp;
3343 struct inet6_dev *idev;
3345 idev = __in6_dev_get(dev);
3346 if (!idev)
3347 return;
3349 write_lock_bh(&idev->lock);
3351 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3352 if ((ifp->flags & IFA_F_PERMANENT) &&
3353 fixup_permanent_addr(idev, ifp) < 0) {
3354 write_unlock_bh(&idev->lock);
3355 ipv6_del_addr(ifp);
3356 write_lock_bh(&idev->lock);
3358 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3359 idev->dev->name, &ifp->addr);
3363 write_unlock_bh(&idev->lock);
3366 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3367 void *ptr)
3369 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3370 struct netdev_notifier_changeupper_info *info;
3371 struct inet6_dev *idev = __in6_dev_get(dev);
3372 int run_pending = 0;
3373 int err;
3375 switch (event) {
3376 case NETDEV_REGISTER:
3377 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3378 idev = ipv6_add_dev(dev);
3379 if (IS_ERR(idev))
3380 return notifier_from_errno(PTR_ERR(idev));
3382 break;
3384 case NETDEV_CHANGEMTU:
3385 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3386 if (dev->mtu < IPV6_MIN_MTU) {
3387 addrconf_ifdown(dev, 1);
3388 break;
3391 if (idev) {
3392 rt6_mtu_change(dev, dev->mtu);
3393 idev->cnf.mtu6 = dev->mtu;
3394 break;
3397 /* allocate new idev */
3398 idev = ipv6_add_dev(dev);
3399 if (IS_ERR(idev))
3400 break;
3402 /* device is still not ready */
3403 if (!(idev->if_flags & IF_READY))
3404 break;
3406 run_pending = 1;
3408 /* fall through */
3410 case NETDEV_UP:
3411 case NETDEV_CHANGE:
3412 if (dev->flags & IFF_SLAVE)
3413 break;
3415 if (idev && idev->cnf.disable_ipv6)
3416 break;
3418 if (event == NETDEV_UP) {
3419 /* restore routes for permanent addresses */
3420 addrconf_permanent_addr(dev);
3422 if (!addrconf_qdisc_ok(dev)) {
3423 /* device is not ready yet. */
3424 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3425 dev->name);
3426 break;
3429 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3430 idev = ipv6_add_dev(dev);
3432 if (!IS_ERR_OR_NULL(idev)) {
3433 idev->if_flags |= IF_READY;
3434 run_pending = 1;
3436 } else if (event == NETDEV_CHANGE) {
3437 if (!addrconf_qdisc_ok(dev)) {
3438 /* device is still not ready. */
3439 break;
3442 if (idev) {
3443 if (idev->if_flags & IF_READY) {
3444 /* device is already configured -
3445 * but resend MLD reports, we might
3446 * have roamed and need to update
3447 * multicast snooping switches
3449 ipv6_mc_up(idev);
3450 break;
3452 idev->if_flags |= IF_READY;
3455 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3456 dev->name);
3458 run_pending = 1;
3461 switch (dev->type) {
3462 #if IS_ENABLED(CONFIG_IPV6_SIT)
3463 case ARPHRD_SIT:
3464 addrconf_sit_config(dev);
3465 break;
3466 #endif
3467 #if IS_ENABLED(CONFIG_NET_IPGRE)
3468 case ARPHRD_IPGRE:
3469 addrconf_gre_config(dev);
3470 break;
3471 #endif
3472 case ARPHRD_LOOPBACK:
3473 init_loopback(dev);
3474 break;
3476 default:
3477 addrconf_dev_config(dev);
3478 break;
3481 if (!IS_ERR_OR_NULL(idev)) {
3482 if (run_pending)
3483 addrconf_dad_run(idev);
3486 * If the MTU changed during the interface down,
3487 * when the interface up, the changed MTU must be
3488 * reflected in the idev as well as routers.
3490 if (idev->cnf.mtu6 != dev->mtu &&
3491 dev->mtu >= IPV6_MIN_MTU) {
3492 rt6_mtu_change(dev, dev->mtu);
3493 idev->cnf.mtu6 = dev->mtu;
3495 idev->tstamp = jiffies;
3496 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3499 * If the changed mtu during down is lower than
3500 * IPV6_MIN_MTU stop IPv6 on this interface.
3502 if (dev->mtu < IPV6_MIN_MTU)
3503 addrconf_ifdown(dev, 1);
3505 break;
3507 case NETDEV_DOWN:
3508 case NETDEV_UNREGISTER:
3510 * Remove all addresses from this interface.
3512 addrconf_ifdown(dev, event != NETDEV_DOWN);
3513 break;
3515 case NETDEV_CHANGENAME:
3516 if (idev) {
3517 snmp6_unregister_dev(idev);
3518 addrconf_sysctl_unregister(idev);
3519 err = addrconf_sysctl_register(idev);
3520 if (err)
3521 return notifier_from_errno(err);
3522 err = snmp6_register_dev(idev);
3523 if (err) {
3524 addrconf_sysctl_unregister(idev);
3525 return notifier_from_errno(err);
3528 break;
3530 case NETDEV_PRE_TYPE_CHANGE:
3531 case NETDEV_POST_TYPE_CHANGE:
3532 if (idev)
3533 addrconf_type_change(dev, event);
3534 break;
3536 case NETDEV_CHANGEUPPER:
3537 info = ptr;
3539 /* flush all routes if dev is linked to or unlinked from
3540 * an L3 master device (e.g., VRF)
3542 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3543 addrconf_ifdown(dev, 0);
3546 return NOTIFY_OK;
3550 * addrconf module should be notified of a device going up
3552 static struct notifier_block ipv6_dev_notf = {
3553 .notifier_call = addrconf_notify,
3554 .priority = ADDRCONF_NOTIFY_PRIORITY,
3557 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3559 struct inet6_dev *idev;
3560 ASSERT_RTNL();
3562 idev = __in6_dev_get(dev);
3564 if (event == NETDEV_POST_TYPE_CHANGE)
3565 ipv6_mc_remap(idev);
3566 else if (event == NETDEV_PRE_TYPE_CHANGE)
3567 ipv6_mc_unmap(idev);
3570 static bool addr_is_local(const struct in6_addr *addr)
3572 return ipv6_addr_type(addr) &
3573 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3576 static int addrconf_ifdown(struct net_device *dev, int how)
3578 struct net *net = dev_net(dev);
3579 struct inet6_dev *idev;
3580 struct inet6_ifaddr *ifa, *tmp;
3581 struct list_head del_list;
3582 int _keep_addr;
3583 bool keep_addr;
3584 int state, i;
3586 ASSERT_RTNL();
3588 rt6_ifdown(net, dev);
3589 neigh_ifdown(&nd_tbl, dev);
3591 idev = __in6_dev_get(dev);
3592 if (!idev)
3593 return -ENODEV;
3596 * Step 1: remove reference to ipv6 device from parent device.
3597 * Do not dev_put!
3599 if (how) {
3600 idev->dead = 1;
3602 /* protected by rtnl_lock */
3603 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3605 /* Step 1.5: remove snmp6 entry */
3606 snmp6_unregister_dev(idev);
3610 /* aggregate the system setting and interface setting */
3611 _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3612 if (!_keep_addr)
3613 _keep_addr = idev->cnf.keep_addr_on_down;
3615 /* combine the user config with event to determine if permanent
3616 * addresses are to be removed from address hash table
3618 keep_addr = !(how || _keep_addr <= 0 || idev->cnf.disable_ipv6);
3620 /* Step 2: clear hash table */
3621 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3622 struct hlist_head *h = &inet6_addr_lst[i];
3624 spin_lock_bh(&addrconf_hash_lock);
3625 restart:
3626 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3627 if (ifa->idev == idev) {
3628 addrconf_del_dad_work(ifa);
3629 /* combined flag + permanent flag decide if
3630 * address is retained on a down event
3632 if (!keep_addr ||
3633 !(ifa->flags & IFA_F_PERMANENT) ||
3634 addr_is_local(&ifa->addr)) {
3635 hlist_del_init_rcu(&ifa->addr_lst);
3636 goto restart;
3640 spin_unlock_bh(&addrconf_hash_lock);
3643 write_lock_bh(&idev->lock);
3645 addrconf_del_rs_timer(idev);
3647 /* Step 2: clear flags for stateless addrconf */
3648 if (!how)
3649 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3651 /* Step 3: clear tempaddr list */
3652 while (!list_empty(&idev->tempaddr_list)) {
3653 ifa = list_first_entry(&idev->tempaddr_list,
3654 struct inet6_ifaddr, tmp_list);
3655 list_del(&ifa->tmp_list);
3656 write_unlock_bh(&idev->lock);
3657 spin_lock_bh(&ifa->lock);
3659 if (ifa->ifpub) {
3660 in6_ifa_put(ifa->ifpub);
3661 ifa->ifpub = NULL;
3663 spin_unlock_bh(&ifa->lock);
3664 in6_ifa_put(ifa);
3665 write_lock_bh(&idev->lock);
3668 /* re-combine the user config with event to determine if permanent
3669 * addresses are to be removed from the interface list
3671 keep_addr = (!how && _keep_addr > 0 && !idev->cnf.disable_ipv6);
3673 INIT_LIST_HEAD(&del_list);
3674 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3675 struct rt6_info *rt = NULL;
3676 bool keep;
3678 addrconf_del_dad_work(ifa);
3680 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3681 !addr_is_local(&ifa->addr);
3682 if (!keep)
3683 list_move(&ifa->if_list, &del_list);
3685 write_unlock_bh(&idev->lock);
3686 spin_lock_bh(&ifa->lock);
3688 if (keep) {
3689 /* set state to skip the notifier below */
3690 state = INET6_IFADDR_STATE_DEAD;
3691 ifa->state = INET6_IFADDR_STATE_PREDAD;
3692 if (!(ifa->flags & IFA_F_NODAD))
3693 ifa->flags |= IFA_F_TENTATIVE;
3695 rt = ifa->rt;
3696 ifa->rt = NULL;
3697 } else {
3698 state = ifa->state;
3699 ifa->state = INET6_IFADDR_STATE_DEAD;
3702 spin_unlock_bh(&ifa->lock);
3704 if (rt)
3705 ip6_del_rt(rt);
3707 if (state != INET6_IFADDR_STATE_DEAD) {
3708 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3709 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3710 } else {
3711 if (idev->cnf.forwarding)
3712 addrconf_leave_anycast(ifa);
3713 addrconf_leave_solict(ifa->idev, &ifa->addr);
3716 write_lock_bh(&idev->lock);
3719 write_unlock_bh(&idev->lock);
3721 /* now clean up addresses to be removed */
3722 while (!list_empty(&del_list)) {
3723 ifa = list_first_entry(&del_list,
3724 struct inet6_ifaddr, if_list);
3725 list_del(&ifa->if_list);
3727 in6_ifa_put(ifa);
3730 /* Step 5: Discard anycast and multicast list */
3731 if (how) {
3732 ipv6_ac_destroy_dev(idev);
3733 ipv6_mc_destroy_dev(idev);
3734 } else {
3735 ipv6_mc_down(idev);
3738 idev->tstamp = jiffies;
3740 /* Last: Shot the device (if unregistered) */
3741 if (how) {
3742 addrconf_sysctl_unregister(idev);
3743 neigh_parms_release(&nd_tbl, idev->nd_parms);
3744 neigh_ifdown(&nd_tbl, dev);
3745 in6_dev_put(idev);
3747 return 0;
3750 static void addrconf_rs_timer(unsigned long data)
3752 struct inet6_dev *idev = (struct inet6_dev *)data;
3753 struct net_device *dev = idev->dev;
3754 struct in6_addr lladdr;
3756 write_lock(&idev->lock);
3757 if (idev->dead || !(idev->if_flags & IF_READY))
3758 goto out;
3760 if (!ipv6_accept_ra(idev))
3761 goto out;
3763 /* Announcement received after solicitation was sent */
3764 if (idev->if_flags & IF_RA_RCVD)
3765 goto out;
3767 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3768 write_unlock(&idev->lock);
3769 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3770 ndisc_send_rs(dev, &lladdr,
3771 &in6addr_linklocal_allrouters);
3772 else
3773 goto put;
3775 write_lock(&idev->lock);
3776 idev->rs_interval = rfc3315_s14_backoff_update(
3777 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3778 /* The wait after the last probe can be shorter */
3779 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3780 idev->cnf.rtr_solicits) ?
3781 idev->cnf.rtr_solicit_delay :
3782 idev->rs_interval);
3783 } else {
3785 * Note: we do not support deprecated "all on-link"
3786 * assumption any longer.
3788 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3791 out:
3792 write_unlock(&idev->lock);
3793 put:
3794 in6_dev_put(idev);
3798 * Duplicate Address Detection
3800 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3802 unsigned long rand_num;
3803 struct inet6_dev *idev = ifp->idev;
3804 u64 nonce;
3806 if (ifp->flags & IFA_F_OPTIMISTIC)
3807 rand_num = 0;
3808 else
3809 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3811 nonce = 0;
3812 if (idev->cnf.enhanced_dad ||
3813 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3815 get_random_bytes(&nonce, 6);
3816 while (nonce == 0);
3818 ifp->dad_nonce = nonce;
3819 ifp->dad_probes = idev->cnf.dad_transmits;
3820 addrconf_mod_dad_work(ifp, rand_num);
3823 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3825 struct inet6_dev *idev = ifp->idev;
3826 struct net_device *dev = idev->dev;
3827 bool bump_id, notify = false;
3829 addrconf_join_solict(dev, &ifp->addr);
3831 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3833 read_lock_bh(&idev->lock);
3834 spin_lock(&ifp->lock);
3835 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3836 goto out;
3838 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3839 idev->cnf.accept_dad < 1 ||
3840 !(ifp->flags&IFA_F_TENTATIVE) ||
3841 ifp->flags & IFA_F_NODAD) {
3842 bump_id = ifp->flags & IFA_F_TENTATIVE;
3843 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3844 spin_unlock(&ifp->lock);
3845 read_unlock_bh(&idev->lock);
3847 addrconf_dad_completed(ifp, bump_id);
3848 return;
3851 if (!(idev->if_flags & IF_READY)) {
3852 spin_unlock(&ifp->lock);
3853 read_unlock_bh(&idev->lock);
3855 * If the device is not ready:
3856 * - keep it tentative if it is a permanent address.
3857 * - otherwise, kill it.
3859 in6_ifa_hold(ifp);
3860 addrconf_dad_stop(ifp, 0);
3861 return;
3865 * Optimistic nodes can start receiving
3866 * Frames right away
3868 if (ifp->flags & IFA_F_OPTIMISTIC) {
3869 ip6_ins_rt(ifp->rt);
3870 if (ipv6_use_optimistic_addr(idev)) {
3871 /* Because optimistic nodes can use this address,
3872 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3874 notify = true;
3878 addrconf_dad_kick(ifp);
3879 out:
3880 spin_unlock(&ifp->lock);
3881 read_unlock_bh(&idev->lock);
3882 if (notify)
3883 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3886 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3888 bool begin_dad = false;
3890 spin_lock_bh(&ifp->lock);
3891 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3892 ifp->state = INET6_IFADDR_STATE_PREDAD;
3893 begin_dad = true;
3895 spin_unlock_bh(&ifp->lock);
3897 if (begin_dad)
3898 addrconf_mod_dad_work(ifp, 0);
3901 static void addrconf_dad_work(struct work_struct *w)
3903 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3904 struct inet6_ifaddr,
3905 dad_work);
3906 struct inet6_dev *idev = ifp->idev;
3907 bool bump_id, disable_ipv6 = false;
3908 struct in6_addr mcaddr;
3910 enum {
3911 DAD_PROCESS,
3912 DAD_BEGIN,
3913 DAD_ABORT,
3914 } action = DAD_PROCESS;
3916 rtnl_lock();
3918 spin_lock_bh(&ifp->lock);
3919 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3920 action = DAD_BEGIN;
3921 ifp->state = INET6_IFADDR_STATE_DAD;
3922 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3923 action = DAD_ABORT;
3924 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3926 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6 &&
3927 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
3928 struct in6_addr addr;
3930 addr.s6_addr32[0] = htonl(0xfe800000);
3931 addr.s6_addr32[1] = 0;
3933 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
3934 ipv6_addr_equal(&ifp->addr, &addr)) {
3935 /* DAD failed for link-local based on MAC */
3936 idev->cnf.disable_ipv6 = 1;
3938 pr_info("%s: IPv6 being disabled!\n",
3939 ifp->idev->dev->name);
3940 disable_ipv6 = true;
3944 spin_unlock_bh(&ifp->lock);
3946 if (action == DAD_BEGIN) {
3947 addrconf_dad_begin(ifp);
3948 goto out;
3949 } else if (action == DAD_ABORT) {
3950 in6_ifa_hold(ifp);
3951 addrconf_dad_stop(ifp, 1);
3952 if (disable_ipv6)
3953 addrconf_ifdown(idev->dev, 0);
3954 goto out;
3957 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3958 goto out;
3960 write_lock_bh(&idev->lock);
3961 if (idev->dead || !(idev->if_flags & IF_READY)) {
3962 write_unlock_bh(&idev->lock);
3963 goto out;
3966 spin_lock(&ifp->lock);
3967 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3968 spin_unlock(&ifp->lock);
3969 write_unlock_bh(&idev->lock);
3970 goto out;
3973 if (ifp->dad_probes == 0) {
3975 * DAD was successful
3978 bump_id = ifp->flags & IFA_F_TENTATIVE;
3979 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3980 spin_unlock(&ifp->lock);
3981 write_unlock_bh(&idev->lock);
3983 addrconf_dad_completed(ifp, bump_id);
3985 goto out;
3988 ifp->dad_probes--;
3989 addrconf_mod_dad_work(ifp,
3990 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3991 spin_unlock(&ifp->lock);
3992 write_unlock_bh(&idev->lock);
3994 /* send a neighbour solicitation for our addr */
3995 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3996 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
3997 ifp->dad_nonce);
3998 out:
3999 in6_ifa_put(ifp);
4000 rtnl_unlock();
4003 /* ifp->idev must be at least read locked */
4004 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4006 struct inet6_ifaddr *ifpiter;
4007 struct inet6_dev *idev = ifp->idev;
4009 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4010 if (ifpiter->scope > IFA_LINK)
4011 break;
4012 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4013 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4014 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4015 IFA_F_PERMANENT)
4016 return false;
4018 return true;
4021 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id)
4023 struct net_device *dev = ifp->idev->dev;
4024 struct in6_addr lladdr;
4025 bool send_rs, send_mld;
4027 addrconf_del_dad_work(ifp);
4030 * Configure the address for reception. Now it is valid.
4033 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4035 /* If added prefix is link local and we are prepared to process
4036 router advertisements, start sending router solicitations.
4039 read_lock_bh(&ifp->idev->lock);
4040 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4041 send_rs = send_mld &&
4042 ipv6_accept_ra(ifp->idev) &&
4043 ifp->idev->cnf.rtr_solicits != 0 &&
4044 (dev->flags&IFF_LOOPBACK) == 0;
4045 read_unlock_bh(&ifp->idev->lock);
4047 /* While dad is in progress mld report's source address is in6_addrany.
4048 * Resend with proper ll now.
4050 if (send_mld)
4051 ipv6_mc_dad_complete(ifp->idev);
4053 if (send_rs) {
4055 * If a host as already performed a random delay
4056 * [...] as part of DAD [...] there is no need
4057 * to delay again before sending the first RS
4059 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4060 return;
4061 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4063 write_lock_bh(&ifp->idev->lock);
4064 spin_lock(&ifp->lock);
4065 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4066 ifp->idev->cnf.rtr_solicit_interval);
4067 ifp->idev->rs_probes = 1;
4068 ifp->idev->if_flags |= IF_RS_SENT;
4069 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4070 spin_unlock(&ifp->lock);
4071 write_unlock_bh(&ifp->idev->lock);
4074 if (bump_id)
4075 rt_genid_bump_ipv6(dev_net(dev));
4077 /* Make sure that a new temporary address will be created
4078 * before this temporary address becomes deprecated.
4080 if (ifp->flags & IFA_F_TEMPORARY)
4081 addrconf_verify_rtnl();
4084 static void addrconf_dad_run(struct inet6_dev *idev)
4086 struct inet6_ifaddr *ifp;
4088 read_lock_bh(&idev->lock);
4089 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4090 spin_lock(&ifp->lock);
4091 if (ifp->flags & IFA_F_TENTATIVE &&
4092 ifp->state == INET6_IFADDR_STATE_DAD)
4093 addrconf_dad_kick(ifp);
4094 spin_unlock(&ifp->lock);
4096 read_unlock_bh(&idev->lock);
4099 #ifdef CONFIG_PROC_FS
4100 struct if6_iter_state {
4101 struct seq_net_private p;
4102 int bucket;
4103 int offset;
4106 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4108 struct inet6_ifaddr *ifa = NULL;
4109 struct if6_iter_state *state = seq->private;
4110 struct net *net = seq_file_net(seq);
4111 int p = 0;
4113 /* initial bucket if pos is 0 */
4114 if (pos == 0) {
4115 state->bucket = 0;
4116 state->offset = 0;
4119 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4120 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
4121 addr_lst) {
4122 if (!net_eq(dev_net(ifa->idev->dev), net))
4123 continue;
4124 /* sync with offset */
4125 if (p < state->offset) {
4126 p++;
4127 continue;
4129 state->offset++;
4130 return ifa;
4133 /* prepare for next bucket */
4134 state->offset = 0;
4135 p = 0;
4137 return NULL;
4140 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4141 struct inet6_ifaddr *ifa)
4143 struct if6_iter_state *state = seq->private;
4144 struct net *net = seq_file_net(seq);
4146 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
4147 if (!net_eq(dev_net(ifa->idev->dev), net))
4148 continue;
4149 state->offset++;
4150 return ifa;
4153 while (++state->bucket < IN6_ADDR_HSIZE) {
4154 state->offset = 0;
4155 hlist_for_each_entry_rcu_bh(ifa,
4156 &inet6_addr_lst[state->bucket], addr_lst) {
4157 if (!net_eq(dev_net(ifa->idev->dev), net))
4158 continue;
4159 state->offset++;
4160 return ifa;
4164 return NULL;
4167 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4168 __acquires(rcu_bh)
4170 rcu_read_lock_bh();
4171 return if6_get_first(seq, *pos);
4174 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4176 struct inet6_ifaddr *ifa;
4178 ifa = if6_get_next(seq, v);
4179 ++*pos;
4180 return ifa;
4183 static void if6_seq_stop(struct seq_file *seq, void *v)
4184 __releases(rcu_bh)
4186 rcu_read_unlock_bh();
4189 static int if6_seq_show(struct seq_file *seq, void *v)
4191 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4192 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4193 &ifp->addr,
4194 ifp->idev->dev->ifindex,
4195 ifp->prefix_len,
4196 ifp->scope,
4197 (u8) ifp->flags,
4198 ifp->idev->dev->name);
4199 return 0;
4202 static const struct seq_operations if6_seq_ops = {
4203 .start = if6_seq_start,
4204 .next = if6_seq_next,
4205 .show = if6_seq_show,
4206 .stop = if6_seq_stop,
4209 static int if6_seq_open(struct inode *inode, struct file *file)
4211 return seq_open_net(inode, file, &if6_seq_ops,
4212 sizeof(struct if6_iter_state));
4215 static const struct file_operations if6_fops = {
4216 .owner = THIS_MODULE,
4217 .open = if6_seq_open,
4218 .read = seq_read,
4219 .llseek = seq_lseek,
4220 .release = seq_release_net,
4223 static int __net_init if6_proc_net_init(struct net *net)
4225 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
4226 return -ENOMEM;
4227 return 0;
4230 static void __net_exit if6_proc_net_exit(struct net *net)
4232 remove_proc_entry("if_inet6", net->proc_net);
4235 static struct pernet_operations if6_proc_net_ops = {
4236 .init = if6_proc_net_init,
4237 .exit = if6_proc_net_exit,
4240 int __init if6_proc_init(void)
4242 return register_pernet_subsys(&if6_proc_net_ops);
4245 void if6_proc_exit(void)
4247 unregister_pernet_subsys(&if6_proc_net_ops);
4249 #endif /* CONFIG_PROC_FS */
4251 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4252 /* Check if address is a home address configured on any interface. */
4253 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4255 int ret = 0;
4256 struct inet6_ifaddr *ifp = NULL;
4257 unsigned int hash = inet6_addr_hash(addr);
4259 rcu_read_lock_bh();
4260 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
4261 if (!net_eq(dev_net(ifp->idev->dev), net))
4262 continue;
4263 if (ipv6_addr_equal(&ifp->addr, addr) &&
4264 (ifp->flags & IFA_F_HOMEADDRESS)) {
4265 ret = 1;
4266 break;
4269 rcu_read_unlock_bh();
4270 return ret;
4272 #endif
4275 * Periodic address status verification
4278 static void addrconf_verify_rtnl(void)
4280 unsigned long now, next, next_sec, next_sched;
4281 struct inet6_ifaddr *ifp;
4282 int i;
4284 ASSERT_RTNL();
4286 rcu_read_lock_bh();
4287 now = jiffies;
4288 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4290 cancel_delayed_work(&addr_chk_work);
4292 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4293 restart:
4294 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4295 unsigned long age;
4297 /* When setting preferred_lft to a value not zero or
4298 * infinity, while valid_lft is infinity
4299 * IFA_F_PERMANENT has a non-infinity life time.
4301 if ((ifp->flags & IFA_F_PERMANENT) &&
4302 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4303 continue;
4305 spin_lock(&ifp->lock);
4306 /* We try to batch several events at once. */
4307 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4309 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4310 age >= ifp->valid_lft) {
4311 spin_unlock(&ifp->lock);
4312 in6_ifa_hold(ifp);
4313 ipv6_del_addr(ifp);
4314 goto restart;
4315 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4316 spin_unlock(&ifp->lock);
4317 continue;
4318 } else if (age >= ifp->prefered_lft) {
4319 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4320 int deprecate = 0;
4322 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4323 deprecate = 1;
4324 ifp->flags |= IFA_F_DEPRECATED;
4327 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4328 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4329 next = ifp->tstamp + ifp->valid_lft * HZ;
4331 spin_unlock(&ifp->lock);
4333 if (deprecate) {
4334 in6_ifa_hold(ifp);
4336 ipv6_ifa_notify(0, ifp);
4337 in6_ifa_put(ifp);
4338 goto restart;
4340 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4341 !(ifp->flags&IFA_F_TENTATIVE)) {
4342 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4343 ifp->idev->cnf.dad_transmits *
4344 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
4346 if (age >= ifp->prefered_lft - regen_advance) {
4347 struct inet6_ifaddr *ifpub = ifp->ifpub;
4348 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4349 next = ifp->tstamp + ifp->prefered_lft * HZ;
4350 if (!ifp->regen_count && ifpub) {
4351 ifp->regen_count++;
4352 in6_ifa_hold(ifp);
4353 in6_ifa_hold(ifpub);
4354 spin_unlock(&ifp->lock);
4356 spin_lock(&ifpub->lock);
4357 ifpub->regen_count = 0;
4358 spin_unlock(&ifpub->lock);
4359 ipv6_create_tempaddr(ifpub, ifp);
4360 in6_ifa_put(ifpub);
4361 in6_ifa_put(ifp);
4362 goto restart;
4364 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4365 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4366 spin_unlock(&ifp->lock);
4367 } else {
4368 /* ifp->prefered_lft <= ifp->valid_lft */
4369 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4370 next = ifp->tstamp + ifp->prefered_lft * HZ;
4371 spin_unlock(&ifp->lock);
4376 next_sec = round_jiffies_up(next);
4377 next_sched = next;
4379 /* If rounded timeout is accurate enough, accept it. */
4380 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4381 next_sched = next_sec;
4383 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4384 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4385 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4387 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4388 now, next, next_sec, next_sched);
4389 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4390 rcu_read_unlock_bh();
4393 static void addrconf_verify_work(struct work_struct *w)
4395 rtnl_lock();
4396 addrconf_verify_rtnl();
4397 rtnl_unlock();
4400 static void addrconf_verify(void)
4402 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4405 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4406 struct in6_addr **peer_pfx)
4408 struct in6_addr *pfx = NULL;
4410 *peer_pfx = NULL;
4412 if (addr)
4413 pfx = nla_data(addr);
4415 if (local) {
4416 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4417 *peer_pfx = pfx;
4418 pfx = nla_data(local);
4421 return pfx;
4424 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4425 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4426 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4427 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4428 [IFA_FLAGS] = { .len = sizeof(u32) },
4431 static int
4432 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4433 struct netlink_ext_ack *extack)
4435 struct net *net = sock_net(skb->sk);
4436 struct ifaddrmsg *ifm;
4437 struct nlattr *tb[IFA_MAX+1];
4438 struct in6_addr *pfx, *peer_pfx;
4439 u32 ifa_flags;
4440 int err;
4442 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4443 extack);
4444 if (err < 0)
4445 return err;
4447 ifm = nlmsg_data(nlh);
4448 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4449 if (!pfx)
4450 return -EINVAL;
4452 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4454 /* We ignore other flags so far. */
4455 ifa_flags &= IFA_F_MANAGETEMPADDR;
4457 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4458 ifm->ifa_prefixlen);
4461 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
4462 u32 prefered_lft, u32 valid_lft)
4464 u32 flags;
4465 clock_t expires;
4466 unsigned long timeout;
4467 bool was_managetempaddr;
4468 bool had_prefixroute;
4470 ASSERT_RTNL();
4472 if (!valid_lft || (prefered_lft > valid_lft))
4473 return -EINVAL;
4475 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
4476 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4477 return -EINVAL;
4479 timeout = addrconf_timeout_fixup(valid_lft, HZ);
4480 if (addrconf_finite_timeout(timeout)) {
4481 expires = jiffies_to_clock_t(timeout * HZ);
4482 valid_lft = timeout;
4483 flags = RTF_EXPIRES;
4484 } else {
4485 expires = 0;
4486 flags = 0;
4487 ifa_flags |= IFA_F_PERMANENT;
4490 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
4491 if (addrconf_finite_timeout(timeout)) {
4492 if (timeout == 0)
4493 ifa_flags |= IFA_F_DEPRECATED;
4494 prefered_lft = timeout;
4497 spin_lock_bh(&ifp->lock);
4498 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4499 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4500 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4501 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4502 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4503 IFA_F_NOPREFIXROUTE);
4504 ifp->flags |= ifa_flags;
4505 ifp->tstamp = jiffies;
4506 ifp->valid_lft = valid_lft;
4507 ifp->prefered_lft = prefered_lft;
4509 spin_unlock_bh(&ifp->lock);
4510 if (!(ifp->flags&IFA_F_TENTATIVE))
4511 ipv6_ifa_notify(0, ifp);
4513 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
4514 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
4515 expires, flags);
4516 } else if (had_prefixroute) {
4517 enum cleanup_prefix_rt_t action;
4518 unsigned long rt_expires;
4520 write_lock_bh(&ifp->idev->lock);
4521 action = check_cleanup_prefix_route(ifp, &rt_expires);
4522 write_unlock_bh(&ifp->idev->lock);
4524 if (action != CLEANUP_PREFIX_RT_NOP) {
4525 cleanup_prefix_route(ifp, rt_expires,
4526 action == CLEANUP_PREFIX_RT_DEL);
4530 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4531 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
4532 valid_lft = prefered_lft = 0;
4533 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
4534 !was_managetempaddr, jiffies);
4537 addrconf_verify_rtnl();
4539 return 0;
4542 static int
4543 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4544 struct netlink_ext_ack *extack)
4546 struct net *net = sock_net(skb->sk);
4547 struct ifaddrmsg *ifm;
4548 struct nlattr *tb[IFA_MAX+1];
4549 struct in6_addr *pfx, *peer_pfx;
4550 struct inet6_ifaddr *ifa;
4551 struct net_device *dev;
4552 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
4553 u32 ifa_flags;
4554 int err;
4556 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4557 extack);
4558 if (err < 0)
4559 return err;
4561 ifm = nlmsg_data(nlh);
4562 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4563 if (!pfx)
4564 return -EINVAL;
4566 if (tb[IFA_CACHEINFO]) {
4567 struct ifa_cacheinfo *ci;
4569 ci = nla_data(tb[IFA_CACHEINFO]);
4570 valid_lft = ci->ifa_valid;
4571 preferred_lft = ci->ifa_prefered;
4572 } else {
4573 preferred_lft = INFINITY_LIFE_TIME;
4574 valid_lft = INFINITY_LIFE_TIME;
4577 dev = __dev_get_by_index(net, ifm->ifa_index);
4578 if (!dev)
4579 return -ENODEV;
4581 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4583 /* We ignore other flags so far. */
4584 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4585 IFA_F_NOPREFIXROUTE | IFA_F_MCAUTOJOIN;
4587 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
4588 if (!ifa) {
4590 * It would be best to check for !NLM_F_CREATE here but
4591 * userspace already relies on not having to provide this.
4593 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
4594 ifm->ifa_prefixlen, ifa_flags,
4595 preferred_lft, valid_lft);
4598 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4599 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4600 err = -EEXIST;
4601 else
4602 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
4604 in6_ifa_put(ifa);
4606 return err;
4609 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4610 u8 scope, int ifindex)
4612 struct ifaddrmsg *ifm;
4614 ifm = nlmsg_data(nlh);
4615 ifm->ifa_family = AF_INET6;
4616 ifm->ifa_prefixlen = prefixlen;
4617 ifm->ifa_flags = flags;
4618 ifm->ifa_scope = scope;
4619 ifm->ifa_index = ifindex;
4622 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4623 unsigned long tstamp, u32 preferred, u32 valid)
4625 struct ifa_cacheinfo ci;
4627 ci.cstamp = cstamp_delta(cstamp);
4628 ci.tstamp = cstamp_delta(tstamp);
4629 ci.ifa_prefered = preferred;
4630 ci.ifa_valid = valid;
4632 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4635 static inline int rt_scope(int ifa_scope)
4637 if (ifa_scope & IFA_HOST)
4638 return RT_SCOPE_HOST;
4639 else if (ifa_scope & IFA_LINK)
4640 return RT_SCOPE_LINK;
4641 else if (ifa_scope & IFA_SITE)
4642 return RT_SCOPE_SITE;
4643 else
4644 return RT_SCOPE_UNIVERSE;
4647 static inline int inet6_ifaddr_msgsize(void)
4649 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4650 + nla_total_size(16) /* IFA_LOCAL */
4651 + nla_total_size(16) /* IFA_ADDRESS */
4652 + nla_total_size(sizeof(struct ifa_cacheinfo))
4653 + nla_total_size(4) /* IFA_FLAGS */;
4656 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4657 u32 portid, u32 seq, int event, unsigned int flags)
4659 struct nlmsghdr *nlh;
4660 u32 preferred, valid;
4662 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4663 if (!nlh)
4664 return -EMSGSIZE;
4666 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4667 ifa->idev->dev->ifindex);
4669 if (!((ifa->flags&IFA_F_PERMANENT) &&
4670 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4671 preferred = ifa->prefered_lft;
4672 valid = ifa->valid_lft;
4673 if (preferred != INFINITY_LIFE_TIME) {
4674 long tval = (jiffies - ifa->tstamp)/HZ;
4675 if (preferred > tval)
4676 preferred -= tval;
4677 else
4678 preferred = 0;
4679 if (valid != INFINITY_LIFE_TIME) {
4680 if (valid > tval)
4681 valid -= tval;
4682 else
4683 valid = 0;
4686 } else {
4687 preferred = INFINITY_LIFE_TIME;
4688 valid = INFINITY_LIFE_TIME;
4691 if (!ipv6_addr_any(&ifa->peer_addr)) {
4692 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4693 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4694 goto error;
4695 } else
4696 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4697 goto error;
4699 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4700 goto error;
4702 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4703 goto error;
4705 nlmsg_end(skb, nlh);
4706 return 0;
4708 error:
4709 nlmsg_cancel(skb, nlh);
4710 return -EMSGSIZE;
4713 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4714 u32 portid, u32 seq, int event, u16 flags)
4716 struct nlmsghdr *nlh;
4717 u8 scope = RT_SCOPE_UNIVERSE;
4718 int ifindex = ifmca->idev->dev->ifindex;
4720 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4721 scope = RT_SCOPE_SITE;
4723 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4724 if (!nlh)
4725 return -EMSGSIZE;
4727 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4728 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
4729 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4730 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4731 nlmsg_cancel(skb, nlh);
4732 return -EMSGSIZE;
4735 nlmsg_end(skb, nlh);
4736 return 0;
4739 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4740 u32 portid, u32 seq, int event, unsigned int flags)
4742 struct nlmsghdr *nlh;
4743 u8 scope = RT_SCOPE_UNIVERSE;
4744 int ifindex = ifaca->aca_idev->dev->ifindex;
4746 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4747 scope = RT_SCOPE_SITE;
4749 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4750 if (!nlh)
4751 return -EMSGSIZE;
4753 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4754 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
4755 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4756 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4757 nlmsg_cancel(skb, nlh);
4758 return -EMSGSIZE;
4761 nlmsg_end(skb, nlh);
4762 return 0;
4765 enum addr_type_t {
4766 UNICAST_ADDR,
4767 MULTICAST_ADDR,
4768 ANYCAST_ADDR,
4771 /* called with rcu_read_lock() */
4772 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4773 struct netlink_callback *cb, enum addr_type_t type,
4774 int s_ip_idx, int *p_ip_idx)
4776 struct ifmcaddr6 *ifmca;
4777 struct ifacaddr6 *ifaca;
4778 int err = 1;
4779 int ip_idx = *p_ip_idx;
4781 read_lock_bh(&idev->lock);
4782 switch (type) {
4783 case UNICAST_ADDR: {
4784 struct inet6_ifaddr *ifa;
4786 /* unicast address incl. temp addr */
4787 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4788 if (++ip_idx < s_ip_idx)
4789 continue;
4790 err = inet6_fill_ifaddr(skb, ifa,
4791 NETLINK_CB(cb->skb).portid,
4792 cb->nlh->nlmsg_seq,
4793 RTM_NEWADDR,
4794 NLM_F_MULTI);
4795 if (err < 0)
4796 break;
4797 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4799 break;
4801 case MULTICAST_ADDR:
4802 /* multicast address */
4803 for (ifmca = idev->mc_list; ifmca;
4804 ifmca = ifmca->next, ip_idx++) {
4805 if (ip_idx < s_ip_idx)
4806 continue;
4807 err = inet6_fill_ifmcaddr(skb, ifmca,
4808 NETLINK_CB(cb->skb).portid,
4809 cb->nlh->nlmsg_seq,
4810 RTM_GETMULTICAST,
4811 NLM_F_MULTI);
4812 if (err < 0)
4813 break;
4815 break;
4816 case ANYCAST_ADDR:
4817 /* anycast address */
4818 for (ifaca = idev->ac_list; ifaca;
4819 ifaca = ifaca->aca_next, ip_idx++) {
4820 if (ip_idx < s_ip_idx)
4821 continue;
4822 err = inet6_fill_ifacaddr(skb, ifaca,
4823 NETLINK_CB(cb->skb).portid,
4824 cb->nlh->nlmsg_seq,
4825 RTM_GETANYCAST,
4826 NLM_F_MULTI);
4827 if (err < 0)
4828 break;
4830 break;
4831 default:
4832 break;
4834 read_unlock_bh(&idev->lock);
4835 *p_ip_idx = ip_idx;
4836 return err;
4839 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4840 enum addr_type_t type)
4842 struct net *net = sock_net(skb->sk);
4843 int h, s_h;
4844 int idx, ip_idx;
4845 int s_idx, s_ip_idx;
4846 struct net_device *dev;
4847 struct inet6_dev *idev;
4848 struct hlist_head *head;
4850 s_h = cb->args[0];
4851 s_idx = idx = cb->args[1];
4852 s_ip_idx = ip_idx = cb->args[2];
4854 rcu_read_lock();
4855 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4856 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4857 idx = 0;
4858 head = &net->dev_index_head[h];
4859 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4860 if (idx < s_idx)
4861 goto cont;
4862 if (h > s_h || idx > s_idx)
4863 s_ip_idx = 0;
4864 ip_idx = 0;
4865 idev = __in6_dev_get(dev);
4866 if (!idev)
4867 goto cont;
4869 if (in6_dump_addrs(idev, skb, cb, type,
4870 s_ip_idx, &ip_idx) < 0)
4871 goto done;
4872 cont:
4873 idx++;
4876 done:
4877 rcu_read_unlock();
4878 cb->args[0] = h;
4879 cb->args[1] = idx;
4880 cb->args[2] = ip_idx;
4882 return skb->len;
4885 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4887 enum addr_type_t type = UNICAST_ADDR;
4889 return inet6_dump_addr(skb, cb, type);
4892 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4894 enum addr_type_t type = MULTICAST_ADDR;
4896 return inet6_dump_addr(skb, cb, type);
4900 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4902 enum addr_type_t type = ANYCAST_ADDR;
4904 return inet6_dump_addr(skb, cb, type);
4907 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4908 struct netlink_ext_ack *extack)
4910 struct net *net = sock_net(in_skb->sk);
4911 struct ifaddrmsg *ifm;
4912 struct nlattr *tb[IFA_MAX+1];
4913 struct in6_addr *addr = NULL, *peer;
4914 struct net_device *dev = NULL;
4915 struct inet6_ifaddr *ifa;
4916 struct sk_buff *skb;
4917 int err;
4919 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy,
4920 extack);
4921 if (err < 0)
4922 goto errout;
4924 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4925 if (!addr) {
4926 err = -EINVAL;
4927 goto errout;
4930 ifm = nlmsg_data(nlh);
4931 if (ifm->ifa_index)
4932 dev = __dev_get_by_index(net, ifm->ifa_index);
4934 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4935 if (!ifa) {
4936 err = -EADDRNOTAVAIL;
4937 goto errout;
4940 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4941 if (!skb) {
4942 err = -ENOBUFS;
4943 goto errout_ifa;
4946 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4947 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4948 if (err < 0) {
4949 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4950 WARN_ON(err == -EMSGSIZE);
4951 kfree_skb(skb);
4952 goto errout_ifa;
4954 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4955 errout_ifa:
4956 in6_ifa_put(ifa);
4957 errout:
4958 return err;
4961 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4963 struct sk_buff *skb;
4964 struct net *net = dev_net(ifa->idev->dev);
4965 int err = -ENOBUFS;
4967 /* Don't send DELADDR notification for TENTATIVE address,
4968 * since NEWADDR notification is sent only after removing
4969 * TENTATIVE flag.
4971 if (ifa->flags & IFA_F_TENTATIVE && event == RTM_DELADDR)
4972 return;
4974 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4975 if (!skb)
4976 goto errout;
4978 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4979 if (err < 0) {
4980 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4981 WARN_ON(err == -EMSGSIZE);
4982 kfree_skb(skb);
4983 goto errout;
4985 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4986 return;
4987 errout:
4988 if (err < 0)
4989 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4992 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4993 __s32 *array, int bytes)
4995 BUG_ON(bytes < (DEVCONF_MAX * 4));
4997 memset(array, 0, bytes);
4998 array[DEVCONF_FORWARDING] = cnf->forwarding;
4999 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5000 array[DEVCONF_MTU6] = cnf->mtu6;
5001 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5002 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5003 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5004 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5005 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5006 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5007 jiffies_to_msecs(cnf->rtr_solicit_interval);
5008 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5009 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5010 array[DEVCONF_RTR_SOLICIT_DELAY] =
5011 jiffies_to_msecs(cnf->rtr_solicit_delay);
5012 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5013 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5014 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5015 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5016 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5017 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5018 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5019 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5020 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5021 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5022 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5023 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5024 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5025 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5026 #ifdef CONFIG_IPV6_ROUTER_PREF
5027 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5028 array[DEVCONF_RTR_PROBE_INTERVAL] =
5029 jiffies_to_msecs(cnf->rtr_probe_interval);
5030 #ifdef CONFIG_IPV6_ROUTE_INFO
5031 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5032 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5033 #endif
5034 #endif
5035 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5036 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5037 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5038 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5039 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5040 #endif
5041 #ifdef CONFIG_IPV6_MROUTE
5042 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5043 #endif
5044 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5045 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5046 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5047 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5048 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5049 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5050 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5051 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5052 /* we omit DEVCONF_STABLE_SECRET for now */
5053 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5054 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5055 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5056 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5057 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5058 #ifdef CONFIG_IPV6_SEG6_HMAC
5059 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5060 #endif
5061 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5062 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5063 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5066 static inline size_t inet6_ifla6_size(void)
5068 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5069 + nla_total_size(sizeof(struct ifla_cacheinfo))
5070 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5071 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5072 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5073 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
5076 static inline size_t inet6_if_nlmsg_size(void)
5078 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5079 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5080 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5081 + nla_total_size(4) /* IFLA_MTU */
5082 + nla_total_size(4) /* IFLA_LINK */
5083 + nla_total_size(1) /* IFLA_OPERSTATE */
5084 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5087 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5088 int bytes)
5090 int i;
5091 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5092 BUG_ON(pad < 0);
5094 /* Use put_unaligned() because stats may not be aligned for u64. */
5095 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5096 for (i = 1; i < ICMP6_MIB_MAX; i++)
5097 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5099 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5102 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5103 int bytes, size_t syncpoff)
5105 int i, c;
5106 u64 buff[IPSTATS_MIB_MAX];
5107 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5109 BUG_ON(pad < 0);
5111 memset(buff, 0, sizeof(buff));
5112 buff[0] = IPSTATS_MIB_MAX;
5114 for_each_possible_cpu(c) {
5115 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5116 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5119 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5120 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5123 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5124 int bytes)
5126 switch (attrtype) {
5127 case IFLA_INET6_STATS:
5128 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5129 offsetof(struct ipstats_mib, syncp));
5130 break;
5131 case IFLA_INET6_ICMP6STATS:
5132 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5133 break;
5137 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5138 u32 ext_filter_mask)
5140 struct nlattr *nla;
5141 struct ifla_cacheinfo ci;
5143 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5144 goto nla_put_failure;
5145 ci.max_reasm_len = IPV6_MAXPLEN;
5146 ci.tstamp = cstamp_delta(idev->tstamp);
5147 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5148 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5149 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5150 goto nla_put_failure;
5151 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5152 if (!nla)
5153 goto nla_put_failure;
5154 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5156 /* XXX - MC not implemented */
5158 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5159 return 0;
5161 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5162 if (!nla)
5163 goto nla_put_failure;
5164 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5166 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5167 if (!nla)
5168 goto nla_put_failure;
5169 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5171 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5172 if (!nla)
5173 goto nla_put_failure;
5175 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5176 goto nla_put_failure;
5178 read_lock_bh(&idev->lock);
5179 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5180 read_unlock_bh(&idev->lock);
5182 return 0;
5184 nla_put_failure:
5185 return -EMSGSIZE;
5188 static size_t inet6_get_link_af_size(const struct net_device *dev,
5189 u32 ext_filter_mask)
5191 if (!__in6_dev_get(dev))
5192 return 0;
5194 return inet6_ifla6_size();
5197 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5198 u32 ext_filter_mask)
5200 struct inet6_dev *idev = __in6_dev_get(dev);
5202 if (!idev)
5203 return -ENODATA;
5205 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5206 return -EMSGSIZE;
5208 return 0;
5211 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5213 struct inet6_ifaddr *ifp;
5214 struct net_device *dev = idev->dev;
5215 bool clear_token, update_rs = false;
5216 struct in6_addr ll_addr;
5218 ASSERT_RTNL();
5220 if (!token)
5221 return -EINVAL;
5222 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5223 return -EINVAL;
5224 if (!ipv6_accept_ra(idev))
5225 return -EINVAL;
5226 if (idev->cnf.rtr_solicits == 0)
5227 return -EINVAL;
5229 write_lock_bh(&idev->lock);
5231 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5232 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5234 write_unlock_bh(&idev->lock);
5236 clear_token = ipv6_addr_any(token);
5237 if (clear_token)
5238 goto update_lft;
5240 if (!idev->dead && (idev->if_flags & IF_READY) &&
5241 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5242 IFA_F_OPTIMISTIC)) {
5243 /* If we're not ready, then normal ifup will take care
5244 * of this. Otherwise, we need to request our rs here.
5246 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5247 update_rs = true;
5250 update_lft:
5251 write_lock_bh(&idev->lock);
5253 if (update_rs) {
5254 idev->if_flags |= IF_RS_SENT;
5255 idev->rs_interval = rfc3315_s14_backoff_init(
5256 idev->cnf.rtr_solicit_interval);
5257 idev->rs_probes = 1;
5258 addrconf_mod_rs_timer(idev, idev->rs_interval);
5261 /* Well, that's kinda nasty ... */
5262 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5263 spin_lock(&ifp->lock);
5264 if (ifp->tokenized) {
5265 ifp->valid_lft = 0;
5266 ifp->prefered_lft = 0;
5268 spin_unlock(&ifp->lock);
5271 write_unlock_bh(&idev->lock);
5272 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5273 addrconf_verify_rtnl();
5274 return 0;
5277 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5278 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5279 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5282 static int inet6_validate_link_af(const struct net_device *dev,
5283 const struct nlattr *nla)
5285 struct nlattr *tb[IFLA_INET6_MAX + 1];
5287 if (dev && !__in6_dev_get(dev))
5288 return -EAFNOSUPPORT;
5290 return nla_parse_nested(tb, IFLA_INET6_MAX, nla, inet6_af_policy,
5291 NULL);
5294 static int check_addr_gen_mode(int mode)
5296 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5297 mode != IN6_ADDR_GEN_MODE_NONE &&
5298 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5299 mode != IN6_ADDR_GEN_MODE_RANDOM)
5300 return -EINVAL;
5301 return 1;
5304 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5305 int mode)
5307 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5308 !idev->cnf.stable_secret.initialized &&
5309 !net->ipv6.devconf_dflt->stable_secret.initialized)
5310 return -EINVAL;
5311 return 1;
5314 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5316 int err = -EINVAL;
5317 struct inet6_dev *idev = __in6_dev_get(dev);
5318 struct nlattr *tb[IFLA_INET6_MAX + 1];
5320 if (!idev)
5321 return -EAFNOSUPPORT;
5323 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5324 BUG();
5326 if (tb[IFLA_INET6_TOKEN]) {
5327 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5328 if (err)
5329 return err;
5332 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5333 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5335 if (check_addr_gen_mode(mode) < 0 ||
5336 check_stable_privacy(idev, dev_net(dev), mode) < 0)
5337 return -EINVAL;
5339 idev->cnf.addr_gen_mode = mode;
5340 err = 0;
5343 return err;
5346 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5347 u32 portid, u32 seq, int event, unsigned int flags)
5349 struct net_device *dev = idev->dev;
5350 struct ifinfomsg *hdr;
5351 struct nlmsghdr *nlh;
5352 void *protoinfo;
5354 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5355 if (!nlh)
5356 return -EMSGSIZE;
5358 hdr = nlmsg_data(nlh);
5359 hdr->ifi_family = AF_INET6;
5360 hdr->__ifi_pad = 0;
5361 hdr->ifi_type = dev->type;
5362 hdr->ifi_index = dev->ifindex;
5363 hdr->ifi_flags = dev_get_flags(dev);
5364 hdr->ifi_change = 0;
5366 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5367 (dev->addr_len &&
5368 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5369 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5370 (dev->ifindex != dev_get_iflink(dev) &&
5371 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5372 nla_put_u8(skb, IFLA_OPERSTATE,
5373 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5374 goto nla_put_failure;
5375 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
5376 if (!protoinfo)
5377 goto nla_put_failure;
5379 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5380 goto nla_put_failure;
5382 nla_nest_end(skb, protoinfo);
5383 nlmsg_end(skb, nlh);
5384 return 0;
5386 nla_put_failure:
5387 nlmsg_cancel(skb, nlh);
5388 return -EMSGSIZE;
5391 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5393 struct net *net = sock_net(skb->sk);
5394 int h, s_h;
5395 int idx = 0, s_idx;
5396 struct net_device *dev;
5397 struct inet6_dev *idev;
5398 struct hlist_head *head;
5400 s_h = cb->args[0];
5401 s_idx = cb->args[1];
5403 rcu_read_lock();
5404 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5405 idx = 0;
5406 head = &net->dev_index_head[h];
5407 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5408 if (idx < s_idx)
5409 goto cont;
5410 idev = __in6_dev_get(dev);
5411 if (!idev)
5412 goto cont;
5413 if (inet6_fill_ifinfo(skb, idev,
5414 NETLINK_CB(cb->skb).portid,
5415 cb->nlh->nlmsg_seq,
5416 RTM_NEWLINK, NLM_F_MULTI) < 0)
5417 goto out;
5418 cont:
5419 idx++;
5422 out:
5423 rcu_read_unlock();
5424 cb->args[1] = idx;
5425 cb->args[0] = h;
5427 return skb->len;
5430 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5432 struct sk_buff *skb;
5433 struct net *net = dev_net(idev->dev);
5434 int err = -ENOBUFS;
5436 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5437 if (!skb)
5438 goto errout;
5440 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5441 if (err < 0) {
5442 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5443 WARN_ON(err == -EMSGSIZE);
5444 kfree_skb(skb);
5445 goto errout;
5447 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5448 return;
5449 errout:
5450 if (err < 0)
5451 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5454 static inline size_t inet6_prefix_nlmsg_size(void)
5456 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5457 + nla_total_size(sizeof(struct in6_addr))
5458 + nla_total_size(sizeof(struct prefix_cacheinfo));
5461 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5462 struct prefix_info *pinfo, u32 portid, u32 seq,
5463 int event, unsigned int flags)
5465 struct prefixmsg *pmsg;
5466 struct nlmsghdr *nlh;
5467 struct prefix_cacheinfo ci;
5469 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5470 if (!nlh)
5471 return -EMSGSIZE;
5473 pmsg = nlmsg_data(nlh);
5474 pmsg->prefix_family = AF_INET6;
5475 pmsg->prefix_pad1 = 0;
5476 pmsg->prefix_pad2 = 0;
5477 pmsg->prefix_ifindex = idev->dev->ifindex;
5478 pmsg->prefix_len = pinfo->prefix_len;
5479 pmsg->prefix_type = pinfo->type;
5480 pmsg->prefix_pad3 = 0;
5481 pmsg->prefix_flags = 0;
5482 if (pinfo->onlink)
5483 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5484 if (pinfo->autoconf)
5485 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5487 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5488 goto nla_put_failure;
5489 ci.preferred_time = ntohl(pinfo->prefered);
5490 ci.valid_time = ntohl(pinfo->valid);
5491 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5492 goto nla_put_failure;
5493 nlmsg_end(skb, nlh);
5494 return 0;
5496 nla_put_failure:
5497 nlmsg_cancel(skb, nlh);
5498 return -EMSGSIZE;
5501 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
5502 struct prefix_info *pinfo)
5504 struct sk_buff *skb;
5505 struct net *net = dev_net(idev->dev);
5506 int err = -ENOBUFS;
5508 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
5509 if (!skb)
5510 goto errout;
5512 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
5513 if (err < 0) {
5514 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
5515 WARN_ON(err == -EMSGSIZE);
5516 kfree_skb(skb);
5517 goto errout;
5519 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
5520 return;
5521 errout:
5522 if (err < 0)
5523 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
5526 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5528 struct net *net = dev_net(ifp->idev->dev);
5530 if (event)
5531 ASSERT_RTNL();
5533 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
5535 switch (event) {
5536 case RTM_NEWADDR:
5538 * If the address was optimistic
5539 * we inserted the route at the start of
5540 * our DAD process, so we don't need
5541 * to do it again
5543 if (!(ifp->rt->rt6i_node))
5544 ip6_ins_rt(ifp->rt);
5545 if (ifp->idev->cnf.forwarding)
5546 addrconf_join_anycast(ifp);
5547 if (!ipv6_addr_any(&ifp->peer_addr))
5548 addrconf_prefix_route(&ifp->peer_addr, 128,
5549 ifp->idev->dev, 0, 0);
5550 break;
5551 case RTM_DELADDR:
5552 if (ifp->idev->cnf.forwarding)
5553 addrconf_leave_anycast(ifp);
5554 addrconf_leave_solict(ifp->idev, &ifp->addr);
5555 if (!ipv6_addr_any(&ifp->peer_addr)) {
5556 struct rt6_info *rt;
5558 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
5559 ifp->idev->dev, 0, 0);
5560 if (rt)
5561 ip6_del_rt(rt);
5563 if (ifp->rt) {
5564 dst_hold(&ifp->rt->dst);
5565 ip6_del_rt(ifp->rt);
5567 rt_genid_bump_ipv6(net);
5568 break;
5570 atomic_inc(&net->ipv6.dev_addr_genid);
5573 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
5575 rcu_read_lock_bh();
5576 if (likely(ifp->idev->dead == 0))
5577 __ipv6_ifa_notify(event, ifp);
5578 rcu_read_unlock_bh();
5581 #ifdef CONFIG_SYSCTL
5583 static
5584 int addrconf_sysctl_forward(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.forwarding, 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_fixup_forwarding(ctl, valp, val);
5604 if (ret)
5605 *ppos = pos;
5606 return ret;
5609 static
5610 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
5611 void __user *buffer, size_t *lenp, loff_t *ppos)
5613 struct inet6_dev *idev = ctl->extra1;
5614 int min_mtu = IPV6_MIN_MTU;
5615 struct ctl_table lctl;
5617 lctl = *ctl;
5618 lctl.extra1 = &min_mtu;
5619 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
5621 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
5624 static void dev_disable_change(struct inet6_dev *idev)
5626 struct netdev_notifier_info info;
5628 if (!idev || !idev->dev)
5629 return;
5631 netdev_notifier_info_init(&info, idev->dev);
5632 if (idev->cnf.disable_ipv6)
5633 addrconf_notify(NULL, NETDEV_DOWN, &info);
5634 else
5635 addrconf_notify(NULL, NETDEV_UP, &info);
5638 static void addrconf_disable_change(struct net *net, __s32 newf)
5640 struct net_device *dev;
5641 struct inet6_dev *idev;
5643 for_each_netdev(net, dev) {
5644 idev = __in6_dev_get(dev);
5645 if (idev) {
5646 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
5647 idev->cnf.disable_ipv6 = newf;
5648 if (changed)
5649 dev_disable_change(idev);
5654 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
5656 struct net *net;
5657 int old;
5659 if (!rtnl_trylock())
5660 return restart_syscall();
5662 net = (struct net *)table->extra2;
5663 old = *p;
5664 *p = newf;
5666 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
5667 rtnl_unlock();
5668 return 0;
5671 if (p == &net->ipv6.devconf_all->disable_ipv6) {
5672 net->ipv6.devconf_dflt->disable_ipv6 = newf;
5673 addrconf_disable_change(net, newf);
5674 } else if ((!newf) ^ (!old))
5675 dev_disable_change((struct inet6_dev *)table->extra1);
5677 rtnl_unlock();
5678 return 0;
5681 static
5682 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
5683 void __user *buffer, size_t *lenp, loff_t *ppos)
5685 int *valp = ctl->data;
5686 int val = *valp;
5687 loff_t pos = *ppos;
5688 struct ctl_table lctl;
5689 int ret;
5692 * ctl->data points to idev->cnf.disable_ipv6, we should
5693 * not modify it until we get the rtnl lock.
5695 lctl = *ctl;
5696 lctl.data = &val;
5698 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5700 if (write)
5701 ret = addrconf_disable_ipv6(ctl, valp, val);
5702 if (ret)
5703 *ppos = pos;
5704 return ret;
5707 static
5708 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
5709 void __user *buffer, size_t *lenp, loff_t *ppos)
5711 int *valp = ctl->data;
5712 int ret;
5713 int old, new;
5715 old = *valp;
5716 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5717 new = *valp;
5719 if (write && old != new) {
5720 struct net *net = ctl->extra2;
5722 if (!rtnl_trylock())
5723 return restart_syscall();
5725 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
5726 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5727 NETCONFA_PROXY_NEIGH,
5728 NETCONFA_IFINDEX_DEFAULT,
5729 net->ipv6.devconf_dflt);
5730 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
5731 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5732 NETCONFA_PROXY_NEIGH,
5733 NETCONFA_IFINDEX_ALL,
5734 net->ipv6.devconf_all);
5735 else {
5736 struct inet6_dev *idev = ctl->extra1;
5738 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
5739 NETCONFA_PROXY_NEIGH,
5740 idev->dev->ifindex,
5741 &idev->cnf);
5743 rtnl_unlock();
5746 return ret;
5749 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
5750 void __user *buffer, size_t *lenp,
5751 loff_t *ppos)
5753 int ret = 0;
5754 int new_val;
5755 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
5756 struct net *net = (struct net *)ctl->extra2;
5758 if (!rtnl_trylock())
5759 return restart_syscall();
5761 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5763 if (write) {
5764 new_val = *((int *)ctl->data);
5766 if (check_addr_gen_mode(new_val) < 0) {
5767 ret = -EINVAL;
5768 goto out;
5771 /* request for default */
5772 if (&net->ipv6.devconf_dflt->addr_gen_mode == ctl->data) {
5773 ipv6_devconf_dflt.addr_gen_mode = new_val;
5775 /* request for individual net device */
5776 } else {
5777 if (!idev)
5778 goto out;
5780 if (check_stable_privacy(idev, net, new_val) < 0) {
5781 ret = -EINVAL;
5782 goto out;
5785 if (idev->cnf.addr_gen_mode != new_val) {
5786 idev->cnf.addr_gen_mode = new_val;
5787 addrconf_dev_config(idev->dev);
5792 out:
5793 rtnl_unlock();
5795 return ret;
5798 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
5799 void __user *buffer, size_t *lenp,
5800 loff_t *ppos)
5802 int err;
5803 struct in6_addr addr;
5804 char str[IPV6_MAX_STRLEN];
5805 struct ctl_table lctl = *ctl;
5806 struct net *net = ctl->extra2;
5807 struct ipv6_stable_secret *secret = ctl->data;
5809 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
5810 return -EIO;
5812 lctl.maxlen = IPV6_MAX_STRLEN;
5813 lctl.data = str;
5815 if (!rtnl_trylock())
5816 return restart_syscall();
5818 if (!write && !secret->initialized) {
5819 err = -EIO;
5820 goto out;
5823 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
5824 if (err >= sizeof(str)) {
5825 err = -EIO;
5826 goto out;
5829 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
5830 if (err || !write)
5831 goto out;
5833 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
5834 err = -EIO;
5835 goto out;
5838 secret->initialized = true;
5839 secret->secret = addr;
5841 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
5842 struct net_device *dev;
5844 for_each_netdev(net, dev) {
5845 struct inet6_dev *idev = __in6_dev_get(dev);
5847 if (idev) {
5848 idev->cnf.addr_gen_mode =
5849 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5852 } else {
5853 struct inet6_dev *idev = ctl->extra1;
5855 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
5858 out:
5859 rtnl_unlock();
5861 return err;
5864 static
5865 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
5866 int write,
5867 void __user *buffer,
5868 size_t *lenp,
5869 loff_t *ppos)
5871 int *valp = ctl->data;
5872 int val = *valp;
5873 loff_t pos = *ppos;
5874 struct ctl_table lctl;
5875 int ret;
5877 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
5878 * we should not modify it until we get the rtnl lock.
5880 lctl = *ctl;
5881 lctl.data = &val;
5883 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5885 if (write)
5886 ret = addrconf_fixup_linkdown(ctl, valp, val);
5887 if (ret)
5888 *ppos = pos;
5889 return ret;
5892 static
5893 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
5895 if (rt) {
5896 if (action)
5897 rt->dst.flags |= DST_NOPOLICY;
5898 else
5899 rt->dst.flags &= ~DST_NOPOLICY;
5903 static
5904 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
5906 struct inet6_ifaddr *ifa;
5908 read_lock_bh(&idev->lock);
5909 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5910 spin_lock(&ifa->lock);
5911 if (ifa->rt) {
5912 struct rt6_info *rt = ifa->rt;
5913 struct fib6_table *table = rt->rt6i_table;
5914 int cpu;
5916 read_lock(&table->tb6_lock);
5917 addrconf_set_nopolicy(ifa->rt, val);
5918 if (rt->rt6i_pcpu) {
5919 for_each_possible_cpu(cpu) {
5920 struct rt6_info **rtp;
5922 rtp = per_cpu_ptr(rt->rt6i_pcpu, cpu);
5923 addrconf_set_nopolicy(*rtp, val);
5926 read_unlock(&table->tb6_lock);
5928 spin_unlock(&ifa->lock);
5930 read_unlock_bh(&idev->lock);
5933 static
5934 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
5936 struct inet6_dev *idev;
5937 struct net *net;
5939 if (!rtnl_trylock())
5940 return restart_syscall();
5942 *valp = val;
5944 net = (struct net *)ctl->extra2;
5945 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
5946 rtnl_unlock();
5947 return 0;
5950 if (valp == &net->ipv6.devconf_all->disable_policy) {
5951 struct net_device *dev;
5953 for_each_netdev(net, dev) {
5954 idev = __in6_dev_get(dev);
5955 if (idev)
5956 addrconf_disable_policy_idev(idev, val);
5958 } else {
5959 idev = (struct inet6_dev *)ctl->extra1;
5960 addrconf_disable_policy_idev(idev, val);
5963 rtnl_unlock();
5964 return 0;
5967 static
5968 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
5969 void __user *buffer, size_t *lenp,
5970 loff_t *ppos)
5972 int *valp = ctl->data;
5973 int val = *valp;
5974 loff_t pos = *ppos;
5975 struct ctl_table lctl;
5976 int ret;
5978 lctl = *ctl;
5979 lctl.data = &val;
5980 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
5982 if (write && (*valp != val))
5983 ret = addrconf_disable_policy(ctl, valp, val);
5985 if (ret)
5986 *ppos = pos;
5988 return ret;
5991 static int minus_one = -1;
5992 static const int one = 1;
5993 static const int two_five_five = 255;
5995 static const struct ctl_table addrconf_sysctl[] = {
5997 .procname = "forwarding",
5998 .data = &ipv6_devconf.forwarding,
5999 .maxlen = sizeof(int),
6000 .mode = 0644,
6001 .proc_handler = addrconf_sysctl_forward,
6004 .procname = "hop_limit",
6005 .data = &ipv6_devconf.hop_limit,
6006 .maxlen = sizeof(int),
6007 .mode = 0644,
6008 .proc_handler = proc_dointvec_minmax,
6009 .extra1 = (void *)&one,
6010 .extra2 = (void *)&two_five_five,
6013 .procname = "mtu",
6014 .data = &ipv6_devconf.mtu6,
6015 .maxlen = sizeof(int),
6016 .mode = 0644,
6017 .proc_handler = addrconf_sysctl_mtu,
6020 .procname = "accept_ra",
6021 .data = &ipv6_devconf.accept_ra,
6022 .maxlen = sizeof(int),
6023 .mode = 0644,
6024 .proc_handler = proc_dointvec,
6027 .procname = "accept_redirects",
6028 .data = &ipv6_devconf.accept_redirects,
6029 .maxlen = sizeof(int),
6030 .mode = 0644,
6031 .proc_handler = proc_dointvec,
6034 .procname = "autoconf",
6035 .data = &ipv6_devconf.autoconf,
6036 .maxlen = sizeof(int),
6037 .mode = 0644,
6038 .proc_handler = proc_dointvec,
6041 .procname = "dad_transmits",
6042 .data = &ipv6_devconf.dad_transmits,
6043 .maxlen = sizeof(int),
6044 .mode = 0644,
6045 .proc_handler = proc_dointvec,
6048 .procname = "router_solicitations",
6049 .data = &ipv6_devconf.rtr_solicits,
6050 .maxlen = sizeof(int),
6051 .mode = 0644,
6052 .proc_handler = proc_dointvec_minmax,
6053 .extra1 = &minus_one,
6056 .procname = "router_solicitation_interval",
6057 .data = &ipv6_devconf.rtr_solicit_interval,
6058 .maxlen = sizeof(int),
6059 .mode = 0644,
6060 .proc_handler = proc_dointvec_jiffies,
6063 .procname = "router_solicitation_max_interval",
6064 .data = &ipv6_devconf.rtr_solicit_max_interval,
6065 .maxlen = sizeof(int),
6066 .mode = 0644,
6067 .proc_handler = proc_dointvec_jiffies,
6070 .procname = "router_solicitation_delay",
6071 .data = &ipv6_devconf.rtr_solicit_delay,
6072 .maxlen = sizeof(int),
6073 .mode = 0644,
6074 .proc_handler = proc_dointvec_jiffies,
6077 .procname = "force_mld_version",
6078 .data = &ipv6_devconf.force_mld_version,
6079 .maxlen = sizeof(int),
6080 .mode = 0644,
6081 .proc_handler = proc_dointvec,
6084 .procname = "mldv1_unsolicited_report_interval",
6085 .data =
6086 &ipv6_devconf.mldv1_unsolicited_report_interval,
6087 .maxlen = sizeof(int),
6088 .mode = 0644,
6089 .proc_handler = proc_dointvec_ms_jiffies,
6092 .procname = "mldv2_unsolicited_report_interval",
6093 .data =
6094 &ipv6_devconf.mldv2_unsolicited_report_interval,
6095 .maxlen = sizeof(int),
6096 .mode = 0644,
6097 .proc_handler = proc_dointvec_ms_jiffies,
6100 .procname = "use_tempaddr",
6101 .data = &ipv6_devconf.use_tempaddr,
6102 .maxlen = sizeof(int),
6103 .mode = 0644,
6104 .proc_handler = proc_dointvec,
6107 .procname = "temp_valid_lft",
6108 .data = &ipv6_devconf.temp_valid_lft,
6109 .maxlen = sizeof(int),
6110 .mode = 0644,
6111 .proc_handler = proc_dointvec,
6114 .procname = "temp_prefered_lft",
6115 .data = &ipv6_devconf.temp_prefered_lft,
6116 .maxlen = sizeof(int),
6117 .mode = 0644,
6118 .proc_handler = proc_dointvec,
6121 .procname = "regen_max_retry",
6122 .data = &ipv6_devconf.regen_max_retry,
6123 .maxlen = sizeof(int),
6124 .mode = 0644,
6125 .proc_handler = proc_dointvec,
6128 .procname = "max_desync_factor",
6129 .data = &ipv6_devconf.max_desync_factor,
6130 .maxlen = sizeof(int),
6131 .mode = 0644,
6132 .proc_handler = proc_dointvec,
6135 .procname = "max_addresses",
6136 .data = &ipv6_devconf.max_addresses,
6137 .maxlen = sizeof(int),
6138 .mode = 0644,
6139 .proc_handler = proc_dointvec,
6142 .procname = "accept_ra_defrtr",
6143 .data = &ipv6_devconf.accept_ra_defrtr,
6144 .maxlen = sizeof(int),
6145 .mode = 0644,
6146 .proc_handler = proc_dointvec,
6149 .procname = "accept_ra_min_hop_limit",
6150 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6151 .maxlen = sizeof(int),
6152 .mode = 0644,
6153 .proc_handler = proc_dointvec,
6156 .procname = "accept_ra_pinfo",
6157 .data = &ipv6_devconf.accept_ra_pinfo,
6158 .maxlen = sizeof(int),
6159 .mode = 0644,
6160 .proc_handler = proc_dointvec,
6162 #ifdef CONFIG_IPV6_ROUTER_PREF
6164 .procname = "accept_ra_rtr_pref",
6165 .data = &ipv6_devconf.accept_ra_rtr_pref,
6166 .maxlen = sizeof(int),
6167 .mode = 0644,
6168 .proc_handler = proc_dointvec,
6171 .procname = "router_probe_interval",
6172 .data = &ipv6_devconf.rtr_probe_interval,
6173 .maxlen = sizeof(int),
6174 .mode = 0644,
6175 .proc_handler = proc_dointvec_jiffies,
6177 #ifdef CONFIG_IPV6_ROUTE_INFO
6179 .procname = "accept_ra_rt_info_min_plen",
6180 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6181 .maxlen = sizeof(int),
6182 .mode = 0644,
6183 .proc_handler = proc_dointvec,
6186 .procname = "accept_ra_rt_info_max_plen",
6187 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6188 .maxlen = sizeof(int),
6189 .mode = 0644,
6190 .proc_handler = proc_dointvec,
6192 #endif
6193 #endif
6195 .procname = "proxy_ndp",
6196 .data = &ipv6_devconf.proxy_ndp,
6197 .maxlen = sizeof(int),
6198 .mode = 0644,
6199 .proc_handler = addrconf_sysctl_proxy_ndp,
6202 .procname = "accept_source_route",
6203 .data = &ipv6_devconf.accept_source_route,
6204 .maxlen = sizeof(int),
6205 .mode = 0644,
6206 .proc_handler = proc_dointvec,
6208 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6210 .procname = "optimistic_dad",
6211 .data = &ipv6_devconf.optimistic_dad,
6212 .maxlen = sizeof(int),
6213 .mode = 0644,
6214 .proc_handler = proc_dointvec,
6217 .procname = "use_optimistic",
6218 .data = &ipv6_devconf.use_optimistic,
6219 .maxlen = sizeof(int),
6220 .mode = 0644,
6221 .proc_handler = proc_dointvec,
6223 #endif
6224 #ifdef CONFIG_IPV6_MROUTE
6226 .procname = "mc_forwarding",
6227 .data = &ipv6_devconf.mc_forwarding,
6228 .maxlen = sizeof(int),
6229 .mode = 0444,
6230 .proc_handler = proc_dointvec,
6232 #endif
6234 .procname = "disable_ipv6",
6235 .data = &ipv6_devconf.disable_ipv6,
6236 .maxlen = sizeof(int),
6237 .mode = 0644,
6238 .proc_handler = addrconf_sysctl_disable,
6241 .procname = "accept_dad",
6242 .data = &ipv6_devconf.accept_dad,
6243 .maxlen = sizeof(int),
6244 .mode = 0644,
6245 .proc_handler = proc_dointvec,
6248 .procname = "force_tllao",
6249 .data = &ipv6_devconf.force_tllao,
6250 .maxlen = sizeof(int),
6251 .mode = 0644,
6252 .proc_handler = proc_dointvec
6255 .procname = "ndisc_notify",
6256 .data = &ipv6_devconf.ndisc_notify,
6257 .maxlen = sizeof(int),
6258 .mode = 0644,
6259 .proc_handler = proc_dointvec
6262 .procname = "suppress_frag_ndisc",
6263 .data = &ipv6_devconf.suppress_frag_ndisc,
6264 .maxlen = sizeof(int),
6265 .mode = 0644,
6266 .proc_handler = proc_dointvec
6269 .procname = "accept_ra_from_local",
6270 .data = &ipv6_devconf.accept_ra_from_local,
6271 .maxlen = sizeof(int),
6272 .mode = 0644,
6273 .proc_handler = proc_dointvec,
6276 .procname = "accept_ra_mtu",
6277 .data = &ipv6_devconf.accept_ra_mtu,
6278 .maxlen = sizeof(int),
6279 .mode = 0644,
6280 .proc_handler = proc_dointvec,
6283 .procname = "stable_secret",
6284 .data = &ipv6_devconf.stable_secret,
6285 .maxlen = IPV6_MAX_STRLEN,
6286 .mode = 0600,
6287 .proc_handler = addrconf_sysctl_stable_secret,
6290 .procname = "use_oif_addrs_only",
6291 .data = &ipv6_devconf.use_oif_addrs_only,
6292 .maxlen = sizeof(int),
6293 .mode = 0644,
6294 .proc_handler = proc_dointvec,
6297 .procname = "ignore_routes_with_linkdown",
6298 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6299 .maxlen = sizeof(int),
6300 .mode = 0644,
6301 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6304 .procname = "drop_unicast_in_l2_multicast",
6305 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6306 .maxlen = sizeof(int),
6307 .mode = 0644,
6308 .proc_handler = proc_dointvec,
6311 .procname = "drop_unsolicited_na",
6312 .data = &ipv6_devconf.drop_unsolicited_na,
6313 .maxlen = sizeof(int),
6314 .mode = 0644,
6315 .proc_handler = proc_dointvec,
6318 .procname = "keep_addr_on_down",
6319 .data = &ipv6_devconf.keep_addr_on_down,
6320 .maxlen = sizeof(int),
6321 .mode = 0644,
6322 .proc_handler = proc_dointvec,
6326 .procname = "seg6_enabled",
6327 .data = &ipv6_devconf.seg6_enabled,
6328 .maxlen = sizeof(int),
6329 .mode = 0644,
6330 .proc_handler = proc_dointvec,
6332 #ifdef CONFIG_IPV6_SEG6_HMAC
6334 .procname = "seg6_require_hmac",
6335 .data = &ipv6_devconf.seg6_require_hmac,
6336 .maxlen = sizeof(int),
6337 .mode = 0644,
6338 .proc_handler = proc_dointvec,
6340 #endif
6342 .procname = "enhanced_dad",
6343 .data = &ipv6_devconf.enhanced_dad,
6344 .maxlen = sizeof(int),
6345 .mode = 0644,
6346 .proc_handler = proc_dointvec,
6349 .procname = "addr_gen_mode",
6350 .data = &ipv6_devconf.addr_gen_mode,
6351 .maxlen = sizeof(int),
6352 .mode = 0644,
6353 .proc_handler = addrconf_sysctl_addr_gen_mode,
6356 .procname = "disable_policy",
6357 .data = &ipv6_devconf.disable_policy,
6358 .maxlen = sizeof(int),
6359 .mode = 0644,
6360 .proc_handler = addrconf_sysctl_disable_policy,
6363 /* sentinel */
6367 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6368 struct inet6_dev *idev, struct ipv6_devconf *p)
6370 int i, ifindex;
6371 struct ctl_table *table;
6372 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6374 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6375 if (!table)
6376 goto out;
6378 for (i = 0; table[i].data; i++) {
6379 table[i].data += (char *)p - (char *)&ipv6_devconf;
6380 /* If one of these is already set, then it is not safe to
6381 * overwrite either of them: this makes proc_dointvec_minmax
6382 * usable.
6384 if (!table[i].extra1 && !table[i].extra2) {
6385 table[i].extra1 = idev; /* embedded; no ref */
6386 table[i].extra2 = net;
6390 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6392 p->sysctl_header = register_net_sysctl(net, path, table);
6393 if (!p->sysctl_header)
6394 goto free;
6396 if (!strcmp(dev_name, "all"))
6397 ifindex = NETCONFA_IFINDEX_ALL;
6398 else if (!strcmp(dev_name, "default"))
6399 ifindex = NETCONFA_IFINDEX_DEFAULT;
6400 else
6401 ifindex = idev->dev->ifindex;
6402 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6403 ifindex, p);
6404 return 0;
6406 free:
6407 kfree(table);
6408 out:
6409 return -ENOBUFS;
6412 static void __addrconf_sysctl_unregister(struct net *net,
6413 struct ipv6_devconf *p, int ifindex)
6415 struct ctl_table *table;
6417 if (!p->sysctl_header)
6418 return;
6420 table = p->sysctl_header->ctl_table_arg;
6421 unregister_net_sysctl_table(p->sysctl_header);
6422 p->sysctl_header = NULL;
6423 kfree(table);
6425 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6428 static int addrconf_sysctl_register(struct inet6_dev *idev)
6430 int err;
6432 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6433 return -EINVAL;
6435 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6436 &ndisc_ifinfo_sysctl_change);
6437 if (err)
6438 return err;
6439 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6440 idev, &idev->cnf);
6441 if (err)
6442 neigh_sysctl_unregister(idev->nd_parms);
6444 return err;
6447 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6449 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6450 idev->dev->ifindex);
6451 neigh_sysctl_unregister(idev->nd_parms);
6455 #endif
6457 static int __net_init addrconf_init_net(struct net *net)
6459 int err = -ENOMEM;
6460 struct ipv6_devconf *all, *dflt;
6462 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
6463 if (!all)
6464 goto err_alloc_all;
6466 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
6467 if (!dflt)
6468 goto err_alloc_dflt;
6470 /* these will be inherited by all namespaces */
6471 dflt->autoconf = ipv6_defaults.autoconf;
6472 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
6474 dflt->stable_secret.initialized = false;
6475 all->stable_secret.initialized = false;
6477 net->ipv6.devconf_all = all;
6478 net->ipv6.devconf_dflt = dflt;
6480 #ifdef CONFIG_SYSCTL
6481 err = __addrconf_sysctl_register(net, "all", NULL, all);
6482 if (err < 0)
6483 goto err_reg_all;
6485 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
6486 if (err < 0)
6487 goto err_reg_dflt;
6488 #endif
6489 return 0;
6491 #ifdef CONFIG_SYSCTL
6492 err_reg_dflt:
6493 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
6494 err_reg_all:
6495 kfree(dflt);
6496 #endif
6497 err_alloc_dflt:
6498 kfree(all);
6499 err_alloc_all:
6500 return err;
6503 static void __net_exit addrconf_exit_net(struct net *net)
6505 #ifdef CONFIG_SYSCTL
6506 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
6507 NETCONFA_IFINDEX_DEFAULT);
6508 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
6509 NETCONFA_IFINDEX_ALL);
6510 #endif
6511 kfree(net->ipv6.devconf_dflt);
6512 kfree(net->ipv6.devconf_all);
6515 static struct pernet_operations addrconf_ops = {
6516 .init = addrconf_init_net,
6517 .exit = addrconf_exit_net,
6520 static struct rtnl_af_ops inet6_ops __read_mostly = {
6521 .family = AF_INET6,
6522 .fill_link_af = inet6_fill_link_af,
6523 .get_link_af_size = inet6_get_link_af_size,
6524 .validate_link_af = inet6_validate_link_af,
6525 .set_link_af = inet6_set_link_af,
6529 * Init / cleanup code
6532 int __init addrconf_init(void)
6534 struct inet6_dev *idev;
6535 int i, err;
6537 err = ipv6_addr_label_init();
6538 if (err < 0) {
6539 pr_crit("%s: cannot initialize default policy table: %d\n",
6540 __func__, err);
6541 goto out;
6544 err = register_pernet_subsys(&addrconf_ops);
6545 if (err < 0)
6546 goto out_addrlabel;
6548 addrconf_wq = create_workqueue("ipv6_addrconf");
6549 if (!addrconf_wq) {
6550 err = -ENOMEM;
6551 goto out_nowq;
6554 /* The addrconf netdev notifier requires that loopback_dev
6555 * has it's ipv6 private information allocated and setup
6556 * before it can bring up and give link-local addresses
6557 * to other devices which are up.
6559 * Unfortunately, loopback_dev is not necessarily the first
6560 * entry in the global dev_base list of net devices. In fact,
6561 * it is likely to be the very last entry on that list.
6562 * So this causes the notifier registry below to try and
6563 * give link-local addresses to all devices besides loopback_dev
6564 * first, then loopback_dev, which cases all the non-loopback_dev
6565 * devices to fail to get a link-local address.
6567 * So, as a temporary fix, allocate the ipv6 structure for
6568 * loopback_dev first by hand.
6569 * Longer term, all of the dependencies ipv6 has upon the loopback
6570 * device and it being up should be removed.
6572 rtnl_lock();
6573 idev = ipv6_add_dev(init_net.loopback_dev);
6574 rtnl_unlock();
6575 if (IS_ERR(idev)) {
6576 err = PTR_ERR(idev);
6577 goto errlo;
6580 ip6_route_init_special_entries();
6582 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6583 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
6585 register_netdevice_notifier(&ipv6_dev_notf);
6587 addrconf_verify();
6589 rtnl_af_register(&inet6_ops);
6591 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
6592 NULL);
6593 if (err < 0)
6594 goto errout;
6596 /* Only the first call to __rtnl_register can fail */
6597 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
6598 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
6599 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
6600 inet6_dump_ifaddr, NULL);
6601 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
6602 inet6_dump_ifmcaddr, NULL);
6603 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
6604 inet6_dump_ifacaddr, NULL);
6605 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
6606 inet6_netconf_dump_devconf, NULL);
6608 ipv6_addr_label_rtnl_register();
6610 return 0;
6611 errout:
6612 rtnl_af_unregister(&inet6_ops);
6613 unregister_netdevice_notifier(&ipv6_dev_notf);
6614 errlo:
6615 destroy_workqueue(addrconf_wq);
6616 out_nowq:
6617 unregister_pernet_subsys(&addrconf_ops);
6618 out_addrlabel:
6619 ipv6_addr_label_cleanup();
6620 out:
6621 return err;
6624 void addrconf_cleanup(void)
6626 struct net_device *dev;
6627 int i;
6629 unregister_netdevice_notifier(&ipv6_dev_notf);
6630 unregister_pernet_subsys(&addrconf_ops);
6631 ipv6_addr_label_cleanup();
6633 rtnl_lock();
6635 __rtnl_af_unregister(&inet6_ops);
6637 /* clean dev list */
6638 for_each_netdev(&init_net, dev) {
6639 if (__in6_dev_get(dev) == NULL)
6640 continue;
6641 addrconf_ifdown(dev, 1);
6643 addrconf_ifdown(init_net.loopback_dev, 2);
6646 * Check hash table.
6648 spin_lock_bh(&addrconf_hash_lock);
6649 for (i = 0; i < IN6_ADDR_HSIZE; i++)
6650 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
6651 spin_unlock_bh(&addrconf_hash_lock);
6652 cancel_delayed_work(&addr_chk_work);
6653 rtnl_unlock();
6655 destroy_workqueue(addrconf_wq);