octeontx2-pf: Fix error return code in otx2_probe()
[linux/fpc-iii.git] / net / ipv6 / addrconf.c
blob2c4f20ec1e2ae12e8e39bac90b36f4c6ef705080
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 */
12 * Changes:
14 * Janos Farkas : delete timer on ifdown
15 * <chexum@bankinf.banki.hu>
16 * Andi Kleen : kill double kfree on module
17 * unload.
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
20 * packets.
21 * yoshfuji@USAGI : Fixed interval between DAD
22 * packets.
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
26 * support.
27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
28 * address on a same interface.
29 * YOSHIFUJI Hideaki @USAGI : ARCnet support
30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
31 * seq_file.
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
34 * status etc.
37 #define pr_fmt(fmt) "IPv6: " fmt
39 #include <linux/errno.h>
40 #include <linux/types.h>
41 #include <linux/kernel.h>
42 #include <linux/sched/signal.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/inet.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
64 #include <linux/hash.h>
66 #include <net/net_namespace.h>
67 #include <net/sock.h>
68 #include <net/snmp.h>
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
72 #include <net/ipv6.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.h>
77 #include <net/tcp.h>
78 #include <net/ip.h>
79 #include <net/netlink.h>
80 #include <net/pkt_sched.h>
81 #include <net/l3mdev.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
85 #include <linux/random.h>
86 #include <linux/uaccess.h>
87 #include <asm/unaligned.h>
89 #include <linux/proc_fs.h>
90 #include <linux/seq_file.h>
91 #include <linux/export.h>
93 #define INFINITY_LIFE_TIME 0xFFFFFFFF
95 #define IPV6_MAX_STRLEN \
96 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
98 static inline u32 cstamp_delta(unsigned long cstamp)
100 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
103 static inline s32 rfc3315_s14_backoff_init(s32 irt)
105 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
106 u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
107 do_div(tmp, 1000000);
108 return (s32)tmp;
111 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
113 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
114 u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
115 do_div(tmp, 1000000);
116 if ((s32)tmp > mrt) {
117 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
118 tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
119 do_div(tmp, 1000000);
121 return (s32)tmp;
124 #ifdef CONFIG_SYSCTL
125 static int addrconf_sysctl_register(struct inet6_dev *idev);
126 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
127 #else
128 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
130 return 0;
133 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
136 #endif
138 static void ipv6_regen_rndid(struct inet6_dev *idev);
139 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
142 static int ipv6_count_addresses(const struct inet6_dev *idev);
143 static int ipv6_generate_stable_address(struct in6_addr *addr,
144 u8 dad_count,
145 const struct inet6_dev *idev);
147 #define IN6_ADDR_HSIZE_SHIFT 8
148 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
150 * Configured unicast address hash table
152 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
153 static DEFINE_SPINLOCK(addrconf_hash_lock);
155 static void addrconf_verify(void);
156 static void addrconf_verify_rtnl(void);
157 static void addrconf_verify_work(struct work_struct *);
159 static struct workqueue_struct *addrconf_wq;
160 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
162 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
163 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
165 static void addrconf_type_change(struct net_device *dev,
166 unsigned long event);
167 static int addrconf_ifdown(struct net_device *dev, int how);
169 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
170 int plen,
171 const struct net_device *dev,
172 u32 flags, u32 noflags,
173 bool no_gw);
175 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
176 static void addrconf_dad_work(struct work_struct *w);
177 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
178 bool send_na);
179 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
180 static void addrconf_rs_timer(struct timer_list *t);
181 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
182 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
184 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
185 struct prefix_info *pinfo);
187 static struct ipv6_devconf ipv6_devconf __read_mostly = {
188 .forwarding = 0,
189 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
190 .mtu6 = IPV6_MIN_MTU,
191 .accept_ra = 1,
192 .accept_redirects = 1,
193 .autoconf = 1,
194 .force_mld_version = 0,
195 .mldv1_unsolicited_report_interval = 10 * HZ,
196 .mldv2_unsolicited_report_interval = HZ,
197 .dad_transmits = 1,
198 .rtr_solicits = MAX_RTR_SOLICITATIONS,
199 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
200 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
201 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
202 .use_tempaddr = 0,
203 .temp_valid_lft = TEMP_VALID_LIFETIME,
204 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
205 .regen_max_retry = REGEN_MAX_RETRY,
206 .max_desync_factor = MAX_DESYNC_FACTOR,
207 .max_addresses = IPV6_MAX_ADDRESSES,
208 .accept_ra_defrtr = 1,
209 .accept_ra_from_local = 0,
210 .accept_ra_min_hop_limit= 1,
211 .accept_ra_pinfo = 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213 .accept_ra_rtr_pref = 1,
214 .rtr_probe_interval = 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216 .accept_ra_rt_info_min_plen = 0,
217 .accept_ra_rt_info_max_plen = 0,
218 #endif
219 #endif
220 .proxy_ndp = 0,
221 .accept_source_route = 0, /* we do not accept RH0 by default. */
222 .disable_ipv6 = 0,
223 .accept_dad = 0,
224 .suppress_frag_ndisc = 1,
225 .accept_ra_mtu = 1,
226 .stable_secret = {
227 .initialized = false,
229 .use_oif_addrs_only = 0,
230 .ignore_routes_with_linkdown = 0,
231 .keep_addr_on_down = 0,
232 .seg6_enabled = 0,
233 #ifdef CONFIG_IPV6_SEG6_HMAC
234 .seg6_require_hmac = 0,
235 #endif
236 .enhanced_dad = 1,
237 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
238 .disable_policy = 0,
239 .rpl_seg_enabled = 0,
242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
243 .forwarding = 0,
244 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
245 .mtu6 = IPV6_MIN_MTU,
246 .accept_ra = 1,
247 .accept_redirects = 1,
248 .autoconf = 1,
249 .force_mld_version = 0,
250 .mldv1_unsolicited_report_interval = 10 * HZ,
251 .mldv2_unsolicited_report_interval = HZ,
252 .dad_transmits = 1,
253 .rtr_solicits = MAX_RTR_SOLICITATIONS,
254 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
255 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
256 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
257 .use_tempaddr = 0,
258 .temp_valid_lft = TEMP_VALID_LIFETIME,
259 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
260 .regen_max_retry = REGEN_MAX_RETRY,
261 .max_desync_factor = MAX_DESYNC_FACTOR,
262 .max_addresses = IPV6_MAX_ADDRESSES,
263 .accept_ra_defrtr = 1,
264 .accept_ra_from_local = 0,
265 .accept_ra_min_hop_limit= 1,
266 .accept_ra_pinfo = 1,
267 #ifdef CONFIG_IPV6_ROUTER_PREF
268 .accept_ra_rtr_pref = 1,
269 .rtr_probe_interval = 60 * HZ,
270 #ifdef CONFIG_IPV6_ROUTE_INFO
271 .accept_ra_rt_info_min_plen = 0,
272 .accept_ra_rt_info_max_plen = 0,
273 #endif
274 #endif
275 .proxy_ndp = 0,
276 .accept_source_route = 0, /* we do not accept RH0 by default. */
277 .disable_ipv6 = 0,
278 .accept_dad = 1,
279 .suppress_frag_ndisc = 1,
280 .accept_ra_mtu = 1,
281 .stable_secret = {
282 .initialized = false,
284 .use_oif_addrs_only = 0,
285 .ignore_routes_with_linkdown = 0,
286 .keep_addr_on_down = 0,
287 .seg6_enabled = 0,
288 #ifdef CONFIG_IPV6_SEG6_HMAC
289 .seg6_require_hmac = 0,
290 #endif
291 .enhanced_dad = 1,
292 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
293 .disable_policy = 0,
294 .rpl_seg_enabled = 0,
297 /* Check if link is ready: is it up and is a valid qdisc available */
298 static inline bool addrconf_link_ready(const struct net_device *dev)
300 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
303 static void addrconf_del_rs_timer(struct inet6_dev *idev)
305 if (del_timer(&idev->rs_timer))
306 __in6_dev_put(idev);
309 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
311 if (cancel_delayed_work(&ifp->dad_work))
312 __in6_ifa_put(ifp);
315 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
316 unsigned long when)
318 if (!timer_pending(&idev->rs_timer))
319 in6_dev_hold(idev);
320 mod_timer(&idev->rs_timer, jiffies + when);
323 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
324 unsigned long delay)
326 in6_ifa_hold(ifp);
327 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
328 in6_ifa_put(ifp);
331 static int snmp6_alloc_dev(struct inet6_dev *idev)
333 int i;
335 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
336 if (!idev->stats.ipv6)
337 goto err_ip;
339 for_each_possible_cpu(i) {
340 struct ipstats_mib *addrconf_stats;
341 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
342 u64_stats_init(&addrconf_stats->syncp);
346 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
347 GFP_KERNEL);
348 if (!idev->stats.icmpv6dev)
349 goto err_icmp;
350 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
351 GFP_KERNEL);
352 if (!idev->stats.icmpv6msgdev)
353 goto err_icmpmsg;
355 return 0;
357 err_icmpmsg:
358 kfree(idev->stats.icmpv6dev);
359 err_icmp:
360 free_percpu(idev->stats.ipv6);
361 err_ip:
362 return -ENOMEM;
365 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
367 struct inet6_dev *ndev;
368 int err = -ENOMEM;
370 ASSERT_RTNL();
372 if (dev->mtu < IPV6_MIN_MTU)
373 return ERR_PTR(-EINVAL);
375 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
376 if (!ndev)
377 return ERR_PTR(err);
379 rwlock_init(&ndev->lock);
380 ndev->dev = dev;
381 INIT_LIST_HEAD(&ndev->addr_list);
382 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
383 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
385 if (ndev->cnf.stable_secret.initialized)
386 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
388 ndev->cnf.mtu6 = dev->mtu;
389 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
390 if (!ndev->nd_parms) {
391 kfree(ndev);
392 return ERR_PTR(err);
394 if (ndev->cnf.forwarding)
395 dev_disable_lro(dev);
396 /* We refer to the device */
397 dev_hold(dev);
399 if (snmp6_alloc_dev(ndev) < 0) {
400 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
401 __func__);
402 neigh_parms_release(&nd_tbl, ndev->nd_parms);
403 dev_put(dev);
404 kfree(ndev);
405 return ERR_PTR(err);
408 if (snmp6_register_dev(ndev) < 0) {
409 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
410 __func__, dev->name);
411 goto err_release;
414 /* One reference from device. */
415 refcount_set(&ndev->refcnt, 1);
417 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
418 ndev->cnf.accept_dad = -1;
420 #if IS_ENABLED(CONFIG_IPV6_SIT)
421 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
422 pr_info("%s: Disabled Multicast RS\n", dev->name);
423 ndev->cnf.rtr_solicits = 0;
425 #endif
427 INIT_LIST_HEAD(&ndev->tempaddr_list);
428 ndev->desync_factor = U32_MAX;
429 if ((dev->flags&IFF_LOOPBACK) ||
430 dev->type == ARPHRD_TUNNEL ||
431 dev->type == ARPHRD_TUNNEL6 ||
432 dev->type == ARPHRD_SIT ||
433 dev->type == ARPHRD_NONE) {
434 ndev->cnf.use_tempaddr = -1;
435 } else
436 ipv6_regen_rndid(ndev);
438 ndev->token = in6addr_any;
440 if (netif_running(dev) && addrconf_link_ready(dev))
441 ndev->if_flags |= IF_READY;
443 ipv6_mc_init_dev(ndev);
444 ndev->tstamp = jiffies;
445 err = addrconf_sysctl_register(ndev);
446 if (err) {
447 ipv6_mc_destroy_dev(ndev);
448 snmp6_unregister_dev(ndev);
449 goto err_release;
451 /* protected by rtnl_lock */
452 rcu_assign_pointer(dev->ip6_ptr, ndev);
454 /* Join interface-local all-node multicast group */
455 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
457 /* Join all-node multicast group */
458 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
460 /* Join all-router multicast group if forwarding is set */
461 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
462 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
464 return ndev;
466 err_release:
467 neigh_parms_release(&nd_tbl, ndev->nd_parms);
468 ndev->dead = 1;
469 in6_dev_finish_destroy(ndev);
470 return ERR_PTR(err);
473 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
475 struct inet6_dev *idev;
477 ASSERT_RTNL();
479 idev = __in6_dev_get(dev);
480 if (!idev) {
481 idev = ipv6_add_dev(dev);
482 if (IS_ERR(idev))
483 return idev;
486 if (dev->flags&IFF_UP)
487 ipv6_mc_up(idev);
488 return idev;
491 static int inet6_netconf_msgsize_devconf(int type)
493 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
494 + nla_total_size(4); /* NETCONFA_IFINDEX */
495 bool all = false;
497 if (type == NETCONFA_ALL)
498 all = true;
500 if (all || type == NETCONFA_FORWARDING)
501 size += nla_total_size(4);
502 #ifdef CONFIG_IPV6_MROUTE
503 if (all || type == NETCONFA_MC_FORWARDING)
504 size += nla_total_size(4);
505 #endif
506 if (all || type == NETCONFA_PROXY_NEIGH)
507 size += nla_total_size(4);
509 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
510 size += nla_total_size(4);
512 return size;
515 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
516 struct ipv6_devconf *devconf, u32 portid,
517 u32 seq, int event, unsigned int flags,
518 int type)
520 struct nlmsghdr *nlh;
521 struct netconfmsg *ncm;
522 bool all = false;
524 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
525 flags);
526 if (!nlh)
527 return -EMSGSIZE;
529 if (type == NETCONFA_ALL)
530 all = true;
532 ncm = nlmsg_data(nlh);
533 ncm->ncm_family = AF_INET6;
535 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
536 goto nla_put_failure;
538 if (!devconf)
539 goto out;
541 if ((all || type == NETCONFA_FORWARDING) &&
542 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
543 goto nla_put_failure;
544 #ifdef CONFIG_IPV6_MROUTE
545 if ((all || type == NETCONFA_MC_FORWARDING) &&
546 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
547 devconf->mc_forwarding) < 0)
548 goto nla_put_failure;
549 #endif
550 if ((all || type == NETCONFA_PROXY_NEIGH) &&
551 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
552 goto nla_put_failure;
554 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
555 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
556 devconf->ignore_routes_with_linkdown) < 0)
557 goto nla_put_failure;
559 out:
560 nlmsg_end(skb, nlh);
561 return 0;
563 nla_put_failure:
564 nlmsg_cancel(skb, nlh);
565 return -EMSGSIZE;
568 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
569 int ifindex, struct ipv6_devconf *devconf)
571 struct sk_buff *skb;
572 int err = -ENOBUFS;
574 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
575 if (!skb)
576 goto errout;
578 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
579 event, 0, type);
580 if (err < 0) {
581 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
582 WARN_ON(err == -EMSGSIZE);
583 kfree_skb(skb);
584 goto errout;
586 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
587 return;
588 errout:
589 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
592 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
593 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
594 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
595 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
596 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
599 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
600 const struct nlmsghdr *nlh,
601 struct nlattr **tb,
602 struct netlink_ext_ack *extack)
604 int i, err;
606 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
607 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
608 return -EINVAL;
611 if (!netlink_strict_get_check(skb))
612 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
613 tb, NETCONFA_MAX,
614 devconf_ipv6_policy, extack);
616 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
617 tb, NETCONFA_MAX,
618 devconf_ipv6_policy, extack);
619 if (err)
620 return err;
622 for (i = 0; i <= NETCONFA_MAX; i++) {
623 if (!tb[i])
624 continue;
626 switch (i) {
627 case NETCONFA_IFINDEX:
628 break;
629 default:
630 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
631 return -EINVAL;
635 return 0;
638 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
639 struct nlmsghdr *nlh,
640 struct netlink_ext_ack *extack)
642 struct net *net = sock_net(in_skb->sk);
643 struct nlattr *tb[NETCONFA_MAX+1];
644 struct inet6_dev *in6_dev = NULL;
645 struct net_device *dev = NULL;
646 struct sk_buff *skb;
647 struct ipv6_devconf *devconf;
648 int ifindex;
649 int err;
651 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
652 if (err < 0)
653 return err;
655 if (!tb[NETCONFA_IFINDEX])
656 return -EINVAL;
658 err = -EINVAL;
659 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
660 switch (ifindex) {
661 case NETCONFA_IFINDEX_ALL:
662 devconf = net->ipv6.devconf_all;
663 break;
664 case NETCONFA_IFINDEX_DEFAULT:
665 devconf = net->ipv6.devconf_dflt;
666 break;
667 default:
668 dev = dev_get_by_index(net, ifindex);
669 if (!dev)
670 return -EINVAL;
671 in6_dev = in6_dev_get(dev);
672 if (!in6_dev)
673 goto errout;
674 devconf = &in6_dev->cnf;
675 break;
678 err = -ENOBUFS;
679 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
680 if (!skb)
681 goto errout;
683 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
684 NETLINK_CB(in_skb).portid,
685 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
686 NETCONFA_ALL);
687 if (err < 0) {
688 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
689 WARN_ON(err == -EMSGSIZE);
690 kfree_skb(skb);
691 goto errout;
693 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
694 errout:
695 if (in6_dev)
696 in6_dev_put(in6_dev);
697 if (dev)
698 dev_put(dev);
699 return err;
702 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
703 struct netlink_callback *cb)
705 const struct nlmsghdr *nlh = cb->nlh;
706 struct net *net = sock_net(skb->sk);
707 int h, s_h;
708 int idx, s_idx;
709 struct net_device *dev;
710 struct inet6_dev *idev;
711 struct hlist_head *head;
713 if (cb->strict_check) {
714 struct netlink_ext_ack *extack = cb->extack;
715 struct netconfmsg *ncm;
717 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
718 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
719 return -EINVAL;
722 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
723 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
724 return -EINVAL;
728 s_h = cb->args[0];
729 s_idx = idx = cb->args[1];
731 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
732 idx = 0;
733 head = &net->dev_index_head[h];
734 rcu_read_lock();
735 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
736 net->dev_base_seq;
737 hlist_for_each_entry_rcu(dev, head, index_hlist) {
738 if (idx < s_idx)
739 goto cont;
740 idev = __in6_dev_get(dev);
741 if (!idev)
742 goto cont;
744 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
745 &idev->cnf,
746 NETLINK_CB(cb->skb).portid,
747 nlh->nlmsg_seq,
748 RTM_NEWNETCONF,
749 NLM_F_MULTI,
750 NETCONFA_ALL) < 0) {
751 rcu_read_unlock();
752 goto done;
754 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
755 cont:
756 idx++;
758 rcu_read_unlock();
760 if (h == NETDEV_HASHENTRIES) {
761 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
762 net->ipv6.devconf_all,
763 NETLINK_CB(cb->skb).portid,
764 nlh->nlmsg_seq,
765 RTM_NEWNETCONF, NLM_F_MULTI,
766 NETCONFA_ALL) < 0)
767 goto done;
768 else
769 h++;
771 if (h == NETDEV_HASHENTRIES + 1) {
772 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
773 net->ipv6.devconf_dflt,
774 NETLINK_CB(cb->skb).portid,
775 nlh->nlmsg_seq,
776 RTM_NEWNETCONF, NLM_F_MULTI,
777 NETCONFA_ALL) < 0)
778 goto done;
779 else
780 h++;
782 done:
783 cb->args[0] = h;
784 cb->args[1] = idx;
786 return skb->len;
789 #ifdef CONFIG_SYSCTL
790 static void dev_forward_change(struct inet6_dev *idev)
792 struct net_device *dev;
793 struct inet6_ifaddr *ifa;
795 if (!idev)
796 return;
797 dev = idev->dev;
798 if (idev->cnf.forwarding)
799 dev_disable_lro(dev);
800 if (dev->flags & IFF_MULTICAST) {
801 if (idev->cnf.forwarding) {
802 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
803 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
804 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
805 } else {
806 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
807 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
808 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
812 list_for_each_entry(ifa, &idev->addr_list, if_list) {
813 if (ifa->flags&IFA_F_TENTATIVE)
814 continue;
815 if (idev->cnf.forwarding)
816 addrconf_join_anycast(ifa);
817 else
818 addrconf_leave_anycast(ifa);
820 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
821 NETCONFA_FORWARDING,
822 dev->ifindex, &idev->cnf);
826 static void addrconf_forward_change(struct net *net, __s32 newf)
828 struct net_device *dev;
829 struct inet6_dev *idev;
831 for_each_netdev(net, dev) {
832 idev = __in6_dev_get(dev);
833 if (idev) {
834 int changed = (!idev->cnf.forwarding) ^ (!newf);
835 idev->cnf.forwarding = newf;
836 if (changed)
837 dev_forward_change(idev);
842 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
844 struct net *net;
845 int old;
847 if (!rtnl_trylock())
848 return restart_syscall();
850 net = (struct net *)table->extra2;
851 old = *p;
852 *p = newf;
854 if (p == &net->ipv6.devconf_dflt->forwarding) {
855 if ((!newf) ^ (!old))
856 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
857 NETCONFA_FORWARDING,
858 NETCONFA_IFINDEX_DEFAULT,
859 net->ipv6.devconf_dflt);
860 rtnl_unlock();
861 return 0;
864 if (p == &net->ipv6.devconf_all->forwarding) {
865 int old_dflt = net->ipv6.devconf_dflt->forwarding;
867 net->ipv6.devconf_dflt->forwarding = newf;
868 if ((!newf) ^ (!old_dflt))
869 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
870 NETCONFA_FORWARDING,
871 NETCONFA_IFINDEX_DEFAULT,
872 net->ipv6.devconf_dflt);
874 addrconf_forward_change(net, newf);
875 if ((!newf) ^ (!old))
876 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
877 NETCONFA_FORWARDING,
878 NETCONFA_IFINDEX_ALL,
879 net->ipv6.devconf_all);
880 } else if ((!newf) ^ (!old))
881 dev_forward_change((struct inet6_dev *)table->extra1);
882 rtnl_unlock();
884 if (newf)
885 rt6_purge_dflt_routers(net);
886 return 1;
889 static void addrconf_linkdown_change(struct net *net, __s32 newf)
891 struct net_device *dev;
892 struct inet6_dev *idev;
894 for_each_netdev(net, dev) {
895 idev = __in6_dev_get(dev);
896 if (idev) {
897 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
899 idev->cnf.ignore_routes_with_linkdown = newf;
900 if (changed)
901 inet6_netconf_notify_devconf(dev_net(dev),
902 RTM_NEWNETCONF,
903 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
904 dev->ifindex,
905 &idev->cnf);
910 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
912 struct net *net;
913 int old;
915 if (!rtnl_trylock())
916 return restart_syscall();
918 net = (struct net *)table->extra2;
919 old = *p;
920 *p = newf;
922 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
923 if ((!newf) ^ (!old))
924 inet6_netconf_notify_devconf(net,
925 RTM_NEWNETCONF,
926 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
927 NETCONFA_IFINDEX_DEFAULT,
928 net->ipv6.devconf_dflt);
929 rtnl_unlock();
930 return 0;
933 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
934 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
935 addrconf_linkdown_change(net, newf);
936 if ((!newf) ^ (!old))
937 inet6_netconf_notify_devconf(net,
938 RTM_NEWNETCONF,
939 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
940 NETCONFA_IFINDEX_ALL,
941 net->ipv6.devconf_all);
943 rtnl_unlock();
945 return 1;
948 #endif
950 /* Nobody refers to this ifaddr, destroy it */
951 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
953 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
955 #ifdef NET_REFCNT_DEBUG
956 pr_debug("%s\n", __func__);
957 #endif
959 in6_dev_put(ifp->idev);
961 if (cancel_delayed_work(&ifp->dad_work))
962 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
963 ifp);
965 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
966 pr_warn("Freeing alive inet6 address %p\n", ifp);
967 return;
970 kfree_rcu(ifp, rcu);
973 static void
974 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
976 struct list_head *p;
977 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
980 * Each device address list is sorted in order of scope -
981 * global before linklocal.
983 list_for_each(p, &idev->addr_list) {
984 struct inet6_ifaddr *ifa
985 = list_entry(p, struct inet6_ifaddr, if_list);
986 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
987 break;
990 list_add_tail_rcu(&ifp->if_list, p);
993 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
995 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
997 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1000 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1001 struct net_device *dev, unsigned int hash)
1003 struct inet6_ifaddr *ifp;
1005 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1006 if (!net_eq(dev_net(ifp->idev->dev), net))
1007 continue;
1008 if (ipv6_addr_equal(&ifp->addr, addr)) {
1009 if (!dev || ifp->idev->dev == dev)
1010 return true;
1013 return false;
1016 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1018 unsigned int hash = inet6_addr_hash(dev_net(dev), &ifa->addr);
1019 int err = 0;
1021 spin_lock(&addrconf_hash_lock);
1023 /* Ignore adding duplicate addresses on an interface */
1024 if (ipv6_chk_same_addr(dev_net(dev), &ifa->addr, dev, hash)) {
1025 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1026 err = -EEXIST;
1027 } else {
1028 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
1031 spin_unlock(&addrconf_hash_lock);
1033 return err;
1036 /* On success it returns ifp with increased reference count */
1038 static struct inet6_ifaddr *
1039 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1040 bool can_block, struct netlink_ext_ack *extack)
1042 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1043 int addr_type = ipv6_addr_type(cfg->pfx);
1044 struct net *net = dev_net(idev->dev);
1045 struct inet6_ifaddr *ifa = NULL;
1046 struct fib6_info *f6i = NULL;
1047 int err = 0;
1049 if (addr_type == IPV6_ADDR_ANY ||
1050 (addr_type & IPV6_ADDR_MULTICAST &&
1051 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
1052 (!(idev->dev->flags & IFF_LOOPBACK) &&
1053 !netif_is_l3_master(idev->dev) &&
1054 addr_type & IPV6_ADDR_LOOPBACK))
1055 return ERR_PTR(-EADDRNOTAVAIL);
1057 if (idev->dead) {
1058 err = -ENODEV; /*XXX*/
1059 goto out;
1062 if (idev->cnf.disable_ipv6) {
1063 err = -EACCES;
1064 goto out;
1067 /* validator notifier needs to be blocking;
1068 * do not call in atomic context
1070 if (can_block) {
1071 struct in6_validator_info i6vi = {
1072 .i6vi_addr = *cfg->pfx,
1073 .i6vi_dev = idev,
1074 .extack = extack,
1077 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1078 err = notifier_to_errno(err);
1079 if (err < 0)
1080 goto out;
1083 ifa = kzalloc(sizeof(*ifa), gfp_flags);
1084 if (!ifa) {
1085 err = -ENOBUFS;
1086 goto out;
1089 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
1090 if (IS_ERR(f6i)) {
1091 err = PTR_ERR(f6i);
1092 f6i = NULL;
1093 goto out;
1096 if (net->ipv6.devconf_all->disable_policy ||
1097 idev->cnf.disable_policy)
1098 f6i->dst_nopolicy = true;
1100 neigh_parms_data_state_setall(idev->nd_parms);
1102 ifa->addr = *cfg->pfx;
1103 if (cfg->peer_pfx)
1104 ifa->peer_addr = *cfg->peer_pfx;
1106 spin_lock_init(&ifa->lock);
1107 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1108 INIT_HLIST_NODE(&ifa->addr_lst);
1109 ifa->scope = cfg->scope;
1110 ifa->prefix_len = cfg->plen;
1111 ifa->rt_priority = cfg->rt_priority;
1112 ifa->flags = cfg->ifa_flags;
1113 /* No need to add the TENTATIVE flag for addresses with NODAD */
1114 if (!(cfg->ifa_flags & IFA_F_NODAD))
1115 ifa->flags |= IFA_F_TENTATIVE;
1116 ifa->valid_lft = cfg->valid_lft;
1117 ifa->prefered_lft = cfg->preferred_lft;
1118 ifa->cstamp = ifa->tstamp = jiffies;
1119 ifa->tokenized = false;
1121 ifa->rt = f6i;
1123 ifa->idev = idev;
1124 in6_dev_hold(idev);
1126 /* For caller */
1127 refcount_set(&ifa->refcnt, 1);
1129 rcu_read_lock_bh();
1131 err = ipv6_add_addr_hash(idev->dev, ifa);
1132 if (err < 0) {
1133 rcu_read_unlock_bh();
1134 goto out;
1137 write_lock(&idev->lock);
1139 /* Add to inet6_dev unicast addr list. */
1140 ipv6_link_dev_addr(idev, ifa);
1142 if (ifa->flags&IFA_F_TEMPORARY) {
1143 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1144 in6_ifa_hold(ifa);
1147 in6_ifa_hold(ifa);
1148 write_unlock(&idev->lock);
1150 rcu_read_unlock_bh();
1152 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1153 out:
1154 if (unlikely(err < 0)) {
1155 fib6_info_release(f6i);
1157 if (ifa) {
1158 if (ifa->idev)
1159 in6_dev_put(ifa->idev);
1160 kfree(ifa);
1162 ifa = ERR_PTR(err);
1165 return ifa;
1168 enum cleanup_prefix_rt_t {
1169 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1170 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1171 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1175 * Check, whether the prefix for ifp would still need a prefix route
1176 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1177 * constants.
1179 * 1) we don't purge prefix if address was not permanent.
1180 * prefix is managed by its own lifetime.
1181 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1182 * 3) if there are no addresses, delete prefix.
1183 * 4) if there are still other permanent address(es),
1184 * corresponding prefix is still permanent.
1185 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1186 * don't purge the prefix, assume user space is managing it.
1187 * 6) otherwise, update prefix lifetime to the
1188 * longest valid lifetime among the corresponding
1189 * addresses on the device.
1190 * Note: subsequent RA will update lifetime.
1192 static enum cleanup_prefix_rt_t
1193 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1195 struct inet6_ifaddr *ifa;
1196 struct inet6_dev *idev = ifp->idev;
1197 unsigned long lifetime;
1198 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1200 *expires = jiffies;
1202 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1203 if (ifa == ifp)
1204 continue;
1205 if (ifa->prefix_len != ifp->prefix_len ||
1206 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1207 ifp->prefix_len))
1208 continue;
1209 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1210 return CLEANUP_PREFIX_RT_NOP;
1212 action = CLEANUP_PREFIX_RT_EXPIRE;
1214 spin_lock(&ifa->lock);
1216 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1218 * Note: Because this address is
1219 * not permanent, lifetime <
1220 * LONG_MAX / HZ here.
1222 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1223 *expires = ifa->tstamp + lifetime * HZ;
1224 spin_unlock(&ifa->lock);
1227 return action;
1230 static void
1231 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1232 bool del_rt, bool del_peer)
1234 struct fib6_info *f6i;
1236 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1237 ifp->prefix_len,
1238 ifp->idev->dev, 0, RTF_DEFAULT, true);
1239 if (f6i) {
1240 if (del_rt)
1241 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1242 else {
1243 if (!(f6i->fib6_flags & RTF_EXPIRES))
1244 fib6_set_expires(f6i, expires);
1245 fib6_info_release(f6i);
1251 /* This function wants to get referenced ifp and releases it before return */
1253 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1255 int state;
1256 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1257 unsigned long expires;
1259 ASSERT_RTNL();
1261 spin_lock_bh(&ifp->lock);
1262 state = ifp->state;
1263 ifp->state = INET6_IFADDR_STATE_DEAD;
1264 spin_unlock_bh(&ifp->lock);
1266 if (state == INET6_IFADDR_STATE_DEAD)
1267 goto out;
1269 spin_lock_bh(&addrconf_hash_lock);
1270 hlist_del_init_rcu(&ifp->addr_lst);
1271 spin_unlock_bh(&addrconf_hash_lock);
1273 write_lock_bh(&ifp->idev->lock);
1275 if (ifp->flags&IFA_F_TEMPORARY) {
1276 list_del(&ifp->tmp_list);
1277 if (ifp->ifpub) {
1278 in6_ifa_put(ifp->ifpub);
1279 ifp->ifpub = NULL;
1281 __in6_ifa_put(ifp);
1284 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1285 action = check_cleanup_prefix_route(ifp, &expires);
1287 list_del_rcu(&ifp->if_list);
1288 __in6_ifa_put(ifp);
1290 write_unlock_bh(&ifp->idev->lock);
1292 addrconf_del_dad_work(ifp);
1294 ipv6_ifa_notify(RTM_DELADDR, ifp);
1296 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1298 if (action != CLEANUP_PREFIX_RT_NOP) {
1299 cleanup_prefix_route(ifp, expires,
1300 action == CLEANUP_PREFIX_RT_DEL, false);
1303 /* clean up prefsrc entries */
1304 rt6_remove_prefsrc(ifp);
1305 out:
1306 in6_ifa_put(ifp);
1309 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp,
1310 struct inet6_ifaddr *ift,
1311 bool block)
1313 struct inet6_dev *idev = ifp->idev;
1314 struct in6_addr addr, *tmpaddr;
1315 unsigned long tmp_tstamp, age;
1316 unsigned long regen_advance;
1317 struct ifa6_config cfg;
1318 int ret = 0;
1319 unsigned long now = jiffies;
1320 long max_desync_factor;
1321 s32 cnf_temp_preferred_lft;
1323 write_lock_bh(&idev->lock);
1324 if (ift) {
1325 spin_lock_bh(&ift->lock);
1326 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1327 spin_unlock_bh(&ift->lock);
1328 tmpaddr = &addr;
1329 } else {
1330 tmpaddr = NULL;
1332 retry:
1333 in6_dev_hold(idev);
1334 if (idev->cnf.use_tempaddr <= 0) {
1335 write_unlock_bh(&idev->lock);
1336 pr_info("%s: use_tempaddr is disabled\n", __func__);
1337 in6_dev_put(idev);
1338 ret = -1;
1339 goto out;
1341 spin_lock_bh(&ifp->lock);
1342 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1343 idev->cnf.use_tempaddr = -1; /*XXX*/
1344 spin_unlock_bh(&ifp->lock);
1345 write_unlock_bh(&idev->lock);
1346 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1347 __func__);
1348 in6_dev_put(idev);
1349 ret = -1;
1350 goto out;
1352 in6_ifa_hold(ifp);
1353 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1354 ipv6_try_regen_rndid(idev, tmpaddr);
1355 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1356 age = (now - ifp->tstamp) / HZ;
1358 regen_advance = idev->cnf.regen_max_retry *
1359 idev->cnf.dad_transmits *
1360 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1362 /* recalculate max_desync_factor each time and update
1363 * idev->desync_factor if it's larger
1365 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1366 max_desync_factor = min_t(__u32,
1367 idev->cnf.max_desync_factor,
1368 cnf_temp_preferred_lft - regen_advance);
1370 if (unlikely(idev->desync_factor > max_desync_factor)) {
1371 if (max_desync_factor > 0) {
1372 get_random_bytes(&idev->desync_factor,
1373 sizeof(idev->desync_factor));
1374 idev->desync_factor %= max_desync_factor;
1375 } else {
1376 idev->desync_factor = 0;
1380 memset(&cfg, 0, sizeof(cfg));
1381 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1382 idev->cnf.temp_valid_lft + age);
1383 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1384 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
1386 cfg.plen = ifp->prefix_len;
1387 tmp_tstamp = ifp->tstamp;
1388 spin_unlock_bh(&ifp->lock);
1390 write_unlock_bh(&idev->lock);
1392 /* A temporary address is created only if this calculated Preferred
1393 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1394 * an implementation must not create a temporary address with a zero
1395 * Preferred Lifetime.
1396 * Use age calculation as in addrconf_verify to avoid unnecessary
1397 * temporary addresses being generated.
1399 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1400 if (cfg.preferred_lft <= regen_advance + age) {
1401 in6_ifa_put(ifp);
1402 in6_dev_put(idev);
1403 ret = -1;
1404 goto out;
1407 cfg.ifa_flags = IFA_F_TEMPORARY;
1408 /* set in addrconf_prefix_rcv() */
1409 if (ifp->flags & IFA_F_OPTIMISTIC)
1410 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1412 cfg.pfx = &addr;
1413 cfg.scope = ipv6_addr_scope(cfg.pfx);
1415 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1416 if (IS_ERR(ift)) {
1417 in6_ifa_put(ifp);
1418 in6_dev_put(idev);
1419 pr_info("%s: retry temporary address regeneration\n", __func__);
1420 tmpaddr = &addr;
1421 write_lock_bh(&idev->lock);
1422 goto retry;
1425 spin_lock_bh(&ift->lock);
1426 ift->ifpub = ifp;
1427 ift->cstamp = now;
1428 ift->tstamp = tmp_tstamp;
1429 spin_unlock_bh(&ift->lock);
1431 addrconf_dad_start(ift);
1432 in6_ifa_put(ift);
1433 in6_dev_put(idev);
1434 out:
1435 return ret;
1439 * Choose an appropriate source address (RFC3484)
1441 enum {
1442 IPV6_SADDR_RULE_INIT = 0,
1443 IPV6_SADDR_RULE_LOCAL,
1444 IPV6_SADDR_RULE_SCOPE,
1445 IPV6_SADDR_RULE_PREFERRED,
1446 #ifdef CONFIG_IPV6_MIP6
1447 IPV6_SADDR_RULE_HOA,
1448 #endif
1449 IPV6_SADDR_RULE_OIF,
1450 IPV6_SADDR_RULE_LABEL,
1451 IPV6_SADDR_RULE_PRIVACY,
1452 IPV6_SADDR_RULE_ORCHID,
1453 IPV6_SADDR_RULE_PREFIX,
1454 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1455 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1456 #endif
1457 IPV6_SADDR_RULE_MAX
1460 struct ipv6_saddr_score {
1461 int rule;
1462 int addr_type;
1463 struct inet6_ifaddr *ifa;
1464 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1465 int scopedist;
1466 int matchlen;
1469 struct ipv6_saddr_dst {
1470 const struct in6_addr *addr;
1471 int ifindex;
1472 int scope;
1473 int label;
1474 unsigned int prefs;
1477 static inline int ipv6_saddr_preferred(int type)
1479 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1480 return 1;
1481 return 0;
1484 static bool ipv6_use_optimistic_addr(struct net *net,
1485 struct inet6_dev *idev)
1487 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1488 if (!idev)
1489 return false;
1490 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1491 return false;
1492 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
1493 return false;
1495 return true;
1496 #else
1497 return false;
1498 #endif
1501 static bool ipv6_allow_optimistic_dad(struct net *net,
1502 struct inet6_dev *idev)
1504 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1505 if (!idev)
1506 return false;
1507 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
1508 return false;
1510 return true;
1511 #else
1512 return false;
1513 #endif
1516 static int ipv6_get_saddr_eval(struct net *net,
1517 struct ipv6_saddr_score *score,
1518 struct ipv6_saddr_dst *dst,
1519 int i)
1521 int ret;
1523 if (i <= score->rule) {
1524 switch (i) {
1525 case IPV6_SADDR_RULE_SCOPE:
1526 ret = score->scopedist;
1527 break;
1528 case IPV6_SADDR_RULE_PREFIX:
1529 ret = score->matchlen;
1530 break;
1531 default:
1532 ret = !!test_bit(i, score->scorebits);
1534 goto out;
1537 switch (i) {
1538 case IPV6_SADDR_RULE_INIT:
1539 /* Rule 0: remember if hiscore is not ready yet */
1540 ret = !!score->ifa;
1541 break;
1542 case IPV6_SADDR_RULE_LOCAL:
1543 /* Rule 1: Prefer same address */
1544 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1545 break;
1546 case IPV6_SADDR_RULE_SCOPE:
1547 /* Rule 2: Prefer appropriate scope
1549 * ret
1551 * -1 | d 15
1552 * ---+--+-+---> scope
1554 * | d is scope of the destination.
1555 * B-d | \
1556 * | \ <- smaller scope is better if
1557 * B-15 | \ if scope is enough for destination.
1558 * | ret = B - scope (-1 <= scope >= d <= 15).
1559 * d-C-1 | /
1560 * |/ <- greater is better
1561 * -C / if scope is not enough for destination.
1562 * /| ret = scope - C (-1 <= d < scope <= 15).
1564 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1565 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1566 * Assume B = 0 and we get C > 29.
1568 ret = __ipv6_addr_src_scope(score->addr_type);
1569 if (ret >= dst->scope)
1570 ret = -ret;
1571 else
1572 ret -= 128; /* 30 is enough */
1573 score->scopedist = ret;
1574 break;
1575 case IPV6_SADDR_RULE_PREFERRED:
1577 /* Rule 3: Avoid deprecated and optimistic addresses */
1578 u8 avoid = IFA_F_DEPRECATED;
1580 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1581 avoid |= IFA_F_OPTIMISTIC;
1582 ret = ipv6_saddr_preferred(score->addr_type) ||
1583 !(score->ifa->flags & avoid);
1584 break;
1586 #ifdef CONFIG_IPV6_MIP6
1587 case IPV6_SADDR_RULE_HOA:
1589 /* Rule 4: Prefer home address */
1590 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1591 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1592 break;
1594 #endif
1595 case IPV6_SADDR_RULE_OIF:
1596 /* Rule 5: Prefer outgoing interface */
1597 ret = (!dst->ifindex ||
1598 dst->ifindex == score->ifa->idev->dev->ifindex);
1599 break;
1600 case IPV6_SADDR_RULE_LABEL:
1601 /* Rule 6: Prefer matching label */
1602 ret = ipv6_addr_label(net,
1603 &score->ifa->addr, score->addr_type,
1604 score->ifa->idev->dev->ifindex) == dst->label;
1605 break;
1606 case IPV6_SADDR_RULE_PRIVACY:
1608 /* Rule 7: Prefer public address
1609 * Note: prefer temporary address if use_tempaddr >= 2
1611 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1612 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1613 score->ifa->idev->cnf.use_tempaddr >= 2;
1614 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1615 break;
1617 case IPV6_SADDR_RULE_ORCHID:
1618 /* Rule 8-: Prefer ORCHID vs ORCHID or
1619 * non-ORCHID vs non-ORCHID
1621 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1622 ipv6_addr_orchid(dst->addr));
1623 break;
1624 case IPV6_SADDR_RULE_PREFIX:
1625 /* Rule 8: Use longest matching prefix */
1626 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1627 if (ret > score->ifa->prefix_len)
1628 ret = score->ifa->prefix_len;
1629 score->matchlen = ret;
1630 break;
1631 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1632 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1633 /* Optimistic addresses still have lower precedence than other
1634 * preferred addresses.
1636 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1637 break;
1638 #endif
1639 default:
1640 ret = 0;
1643 if (ret)
1644 __set_bit(i, score->scorebits);
1645 score->rule = i;
1646 out:
1647 return ret;
1650 static int __ipv6_dev_get_saddr(struct net *net,
1651 struct ipv6_saddr_dst *dst,
1652 struct inet6_dev *idev,
1653 struct ipv6_saddr_score *scores,
1654 int hiscore_idx)
1656 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1658 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1659 int i;
1662 * - Tentative Address (RFC2462 section 5.4)
1663 * - A tentative address is not considered
1664 * "assigned to an interface" in the traditional
1665 * sense, unless it is also flagged as optimistic.
1666 * - Candidate Source Address (section 4)
1667 * - In any case, anycast addresses, multicast
1668 * addresses, and the unspecified address MUST
1669 * NOT be included in a candidate set.
1671 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1672 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1673 continue;
1675 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1677 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1678 score->addr_type & IPV6_ADDR_MULTICAST)) {
1679 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1680 idev->dev->name);
1681 continue;
1684 score->rule = -1;
1685 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1687 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1688 int minihiscore, miniscore;
1690 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1691 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1693 if (minihiscore > miniscore) {
1694 if (i == IPV6_SADDR_RULE_SCOPE &&
1695 score->scopedist > 0) {
1697 * special case:
1698 * each remaining entry
1699 * has too small (not enough)
1700 * scope, because ifa entries
1701 * are sorted by their scope
1702 * values.
1704 goto out;
1706 break;
1707 } else if (minihiscore < miniscore) {
1708 swap(hiscore, score);
1709 hiscore_idx = 1 - hiscore_idx;
1711 /* restore our iterator */
1712 score->ifa = hiscore->ifa;
1714 break;
1718 out:
1719 return hiscore_idx;
1722 static int ipv6_get_saddr_master(struct net *net,
1723 const struct net_device *dst_dev,
1724 const struct net_device *master,
1725 struct ipv6_saddr_dst *dst,
1726 struct ipv6_saddr_score *scores,
1727 int hiscore_idx)
1729 struct inet6_dev *idev;
1731 idev = __in6_dev_get(dst_dev);
1732 if (idev)
1733 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1734 scores, hiscore_idx);
1736 idev = __in6_dev_get(master);
1737 if (idev)
1738 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1739 scores, hiscore_idx);
1741 return hiscore_idx;
1744 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1745 const struct in6_addr *daddr, unsigned int prefs,
1746 struct in6_addr *saddr)
1748 struct ipv6_saddr_score scores[2], *hiscore;
1749 struct ipv6_saddr_dst dst;
1750 struct inet6_dev *idev;
1751 struct net_device *dev;
1752 int dst_type;
1753 bool use_oif_addr = false;
1754 int hiscore_idx = 0;
1755 int ret = 0;
1757 dst_type = __ipv6_addr_type(daddr);
1758 dst.addr = daddr;
1759 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1760 dst.scope = __ipv6_addr_src_scope(dst_type);
1761 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1762 dst.prefs = prefs;
1764 scores[hiscore_idx].rule = -1;
1765 scores[hiscore_idx].ifa = NULL;
1767 rcu_read_lock();
1769 /* Candidate Source Address (section 4)
1770 * - multicast and link-local destination address,
1771 * the set of candidate source address MUST only
1772 * include addresses assigned to interfaces
1773 * belonging to the same link as the outgoing
1774 * interface.
1775 * (- For site-local destination addresses, the
1776 * set of candidate source addresses MUST only
1777 * include addresses assigned to interfaces
1778 * belonging to the same site as the outgoing
1779 * interface.)
1780 * - "It is RECOMMENDED that the candidate source addresses
1781 * be the set of unicast addresses assigned to the
1782 * interface that will be used to send to the destination
1783 * (the 'outgoing' interface)." (RFC 6724)
1785 if (dst_dev) {
1786 idev = __in6_dev_get(dst_dev);
1787 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1788 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1789 (idev && idev->cnf.use_oif_addrs_only)) {
1790 use_oif_addr = true;
1794 if (use_oif_addr) {
1795 if (idev)
1796 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1797 } else {
1798 const struct net_device *master;
1799 int master_idx = 0;
1801 /* if dst_dev exists and is enslaved to an L3 device, then
1802 * prefer addresses from dst_dev and then the master over
1803 * any other enslaved devices in the L3 domain.
1805 master = l3mdev_master_dev_rcu(dst_dev);
1806 if (master) {
1807 master_idx = master->ifindex;
1809 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1810 master, &dst,
1811 scores, hiscore_idx);
1813 if (scores[hiscore_idx].ifa)
1814 goto out;
1817 for_each_netdev_rcu(net, dev) {
1818 /* only consider addresses on devices in the
1819 * same L3 domain
1821 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1822 continue;
1823 idev = __in6_dev_get(dev);
1824 if (!idev)
1825 continue;
1826 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1830 out:
1831 hiscore = &scores[hiscore_idx];
1832 if (!hiscore->ifa)
1833 ret = -EADDRNOTAVAIL;
1834 else
1835 *saddr = hiscore->ifa->addr;
1837 rcu_read_unlock();
1838 return ret;
1840 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1842 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1843 u32 banned_flags)
1845 struct inet6_ifaddr *ifp;
1846 int err = -EADDRNOTAVAIL;
1848 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1849 if (ifp->scope > IFA_LINK)
1850 break;
1851 if (ifp->scope == IFA_LINK &&
1852 !(ifp->flags & banned_flags)) {
1853 *addr = ifp->addr;
1854 err = 0;
1855 break;
1858 return err;
1861 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1862 u32 banned_flags)
1864 struct inet6_dev *idev;
1865 int err = -EADDRNOTAVAIL;
1867 rcu_read_lock();
1868 idev = __in6_dev_get(dev);
1869 if (idev) {
1870 read_lock_bh(&idev->lock);
1871 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1872 read_unlock_bh(&idev->lock);
1874 rcu_read_unlock();
1875 return err;
1878 static int ipv6_count_addresses(const struct inet6_dev *idev)
1880 const struct inet6_ifaddr *ifp;
1881 int cnt = 0;
1883 rcu_read_lock();
1884 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1885 cnt++;
1886 rcu_read_unlock();
1887 return cnt;
1890 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1891 const struct net_device *dev, int strict)
1893 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1894 strict, IFA_F_TENTATIVE);
1896 EXPORT_SYMBOL(ipv6_chk_addr);
1898 /* device argument is used to find the L3 domain of interest. If
1899 * skip_dev_check is set, then the ifp device is not checked against
1900 * the passed in dev argument. So the 2 cases for addresses checks are:
1901 * 1. does the address exist in the L3 domain that dev is part of
1902 * (skip_dev_check = true), or
1904 * 2. does the address exist on the specific device
1905 * (skip_dev_check = false)
1907 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1908 const struct net_device *dev, bool skip_dev_check,
1909 int strict, u32 banned_flags)
1911 unsigned int hash = inet6_addr_hash(net, addr);
1912 const struct net_device *l3mdev;
1913 struct inet6_ifaddr *ifp;
1914 u32 ifp_flags;
1916 rcu_read_lock();
1918 l3mdev = l3mdev_master_dev_rcu(dev);
1919 if (skip_dev_check)
1920 dev = NULL;
1922 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1923 if (!net_eq(dev_net(ifp->idev->dev), net))
1924 continue;
1926 if (l3mdev_master_dev_rcu(ifp->idev->dev) != l3mdev)
1927 continue;
1929 /* Decouple optimistic from tentative for evaluation here.
1930 * Ban optimistic addresses explicitly, when required.
1932 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1933 ? (ifp->flags&~IFA_F_TENTATIVE)
1934 : ifp->flags;
1935 if (ipv6_addr_equal(&ifp->addr, addr) &&
1936 !(ifp_flags&banned_flags) &&
1937 (!dev || ifp->idev->dev == dev ||
1938 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1939 rcu_read_unlock();
1940 return 1;
1944 rcu_read_unlock();
1945 return 0;
1947 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
1950 /* Compares an address/prefix_len with addresses on device @dev.
1951 * If one is found it returns true.
1953 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1954 const unsigned int prefix_len, struct net_device *dev)
1956 const struct inet6_ifaddr *ifa;
1957 const struct inet6_dev *idev;
1958 bool ret = false;
1960 rcu_read_lock();
1961 idev = __in6_dev_get(dev);
1962 if (idev) {
1963 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1964 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1965 if (ret)
1966 break;
1969 rcu_read_unlock();
1971 return ret;
1973 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1975 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1977 const struct inet6_ifaddr *ifa;
1978 const struct inet6_dev *idev;
1979 int onlink;
1981 onlink = 0;
1982 rcu_read_lock();
1983 idev = __in6_dev_get(dev);
1984 if (idev) {
1985 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
1986 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1987 ifa->prefix_len);
1988 if (onlink)
1989 break;
1992 rcu_read_unlock();
1993 return onlink;
1995 EXPORT_SYMBOL(ipv6_chk_prefix);
1997 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1998 struct net_device *dev, int strict)
2000 unsigned int hash = inet6_addr_hash(net, addr);
2001 struct inet6_ifaddr *ifp, *result = NULL;
2003 rcu_read_lock();
2004 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
2005 if (!net_eq(dev_net(ifp->idev->dev), net))
2006 continue;
2007 if (ipv6_addr_equal(&ifp->addr, addr)) {
2008 if (!dev || ifp->idev->dev == dev ||
2009 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2010 result = ifp;
2011 in6_ifa_hold(ifp);
2012 break;
2016 rcu_read_unlock();
2018 return result;
2021 /* Gets referenced address, destroys ifaddr */
2023 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2025 if (dad_failed)
2026 ifp->flags |= IFA_F_DADFAILED;
2028 if (ifp->flags&IFA_F_TEMPORARY) {
2029 struct inet6_ifaddr *ifpub;
2030 spin_lock_bh(&ifp->lock);
2031 ifpub = ifp->ifpub;
2032 if (ifpub) {
2033 in6_ifa_hold(ifpub);
2034 spin_unlock_bh(&ifp->lock);
2035 ipv6_create_tempaddr(ifpub, ifp, true);
2036 in6_ifa_put(ifpub);
2037 } else {
2038 spin_unlock_bh(&ifp->lock);
2040 ipv6_del_addr(ifp);
2041 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2042 spin_lock_bh(&ifp->lock);
2043 addrconf_del_dad_work(ifp);
2044 ifp->flags |= IFA_F_TENTATIVE;
2045 if (dad_failed)
2046 ifp->flags &= ~IFA_F_OPTIMISTIC;
2047 spin_unlock_bh(&ifp->lock);
2048 if (dad_failed)
2049 ipv6_ifa_notify(0, ifp);
2050 in6_ifa_put(ifp);
2051 } else {
2052 ipv6_del_addr(ifp);
2056 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2058 int err = -ENOENT;
2060 spin_lock_bh(&ifp->lock);
2061 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2062 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2063 err = 0;
2065 spin_unlock_bh(&ifp->lock);
2067 return err;
2070 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2072 struct inet6_dev *idev = ifp->idev;
2073 struct net *net = dev_net(ifp->idev->dev);
2075 if (addrconf_dad_end(ifp)) {
2076 in6_ifa_put(ifp);
2077 return;
2080 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2081 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2083 spin_lock_bh(&ifp->lock);
2085 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2086 struct in6_addr new_addr;
2087 struct inet6_ifaddr *ifp2;
2088 int retries = ifp->stable_privacy_retry + 1;
2089 struct ifa6_config cfg = {
2090 .pfx = &new_addr,
2091 .plen = ifp->prefix_len,
2092 .ifa_flags = ifp->flags,
2093 .valid_lft = ifp->valid_lft,
2094 .preferred_lft = ifp->prefered_lft,
2095 .scope = ifp->scope,
2098 if (retries > net->ipv6.sysctl.idgen_retries) {
2099 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2100 ifp->idev->dev->name);
2101 goto errdad;
2104 new_addr = ifp->addr;
2105 if (ipv6_generate_stable_address(&new_addr, retries,
2106 idev))
2107 goto errdad;
2109 spin_unlock_bh(&ifp->lock);
2111 if (idev->cnf.max_addresses &&
2112 ipv6_count_addresses(idev) >=
2113 idev->cnf.max_addresses)
2114 goto lock_errdad;
2116 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2117 ifp->idev->dev->name);
2119 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2120 if (IS_ERR(ifp2))
2121 goto lock_errdad;
2123 spin_lock_bh(&ifp2->lock);
2124 ifp2->stable_privacy_retry = retries;
2125 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2126 spin_unlock_bh(&ifp2->lock);
2128 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2129 in6_ifa_put(ifp2);
2130 lock_errdad:
2131 spin_lock_bh(&ifp->lock);
2134 errdad:
2135 /* transition from _POSTDAD to _ERRDAD */
2136 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2137 spin_unlock_bh(&ifp->lock);
2139 addrconf_mod_dad_work(ifp, 0);
2140 in6_ifa_put(ifp);
2143 /* Join to solicited addr multicast group.
2144 * caller must hold RTNL */
2145 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2147 struct in6_addr maddr;
2149 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2150 return;
2152 addrconf_addr_solict_mult(addr, &maddr);
2153 ipv6_dev_mc_inc(dev, &maddr);
2156 /* caller must hold RTNL */
2157 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2159 struct in6_addr maddr;
2161 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2162 return;
2164 addrconf_addr_solict_mult(addr, &maddr);
2165 __ipv6_dev_mc_dec(idev, &maddr);
2168 /* caller must hold RTNL */
2169 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2171 struct in6_addr addr;
2173 if (ifp->prefix_len >= 127) /* RFC 6164 */
2174 return;
2175 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2176 if (ipv6_addr_any(&addr))
2177 return;
2178 __ipv6_dev_ac_inc(ifp->idev, &addr);
2181 /* caller must hold RTNL */
2182 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2184 struct in6_addr addr;
2186 if (ifp->prefix_len >= 127) /* RFC 6164 */
2187 return;
2188 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2189 if (ipv6_addr_any(&addr))
2190 return;
2191 __ipv6_dev_ac_dec(ifp->idev, &addr);
2194 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2196 switch (dev->addr_len) {
2197 case ETH_ALEN:
2198 memcpy(eui, dev->dev_addr, 3);
2199 eui[3] = 0xFF;
2200 eui[4] = 0xFE;
2201 memcpy(eui + 5, dev->dev_addr + 3, 3);
2202 break;
2203 case EUI64_ADDR_LEN:
2204 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2205 eui[0] ^= 2;
2206 break;
2207 default:
2208 return -1;
2211 return 0;
2214 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2216 union fwnet_hwaddr *ha;
2218 if (dev->addr_len != FWNET_ALEN)
2219 return -1;
2221 ha = (union fwnet_hwaddr *)dev->dev_addr;
2223 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2224 eui[0] ^= 2;
2225 return 0;
2228 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2230 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2231 if (dev->addr_len != ARCNET_ALEN)
2232 return -1;
2233 memset(eui, 0, 7);
2234 eui[7] = *(u8 *)dev->dev_addr;
2235 return 0;
2238 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2240 if (dev->addr_len != INFINIBAND_ALEN)
2241 return -1;
2242 memcpy(eui, dev->dev_addr + 12, 8);
2243 eui[0] |= 2;
2244 return 0;
2247 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2249 if (addr == 0)
2250 return -1;
2251 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2252 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2253 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2254 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2255 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2256 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2257 eui[1] = 0;
2258 eui[2] = 0x5E;
2259 eui[3] = 0xFE;
2260 memcpy(eui + 4, &addr, 4);
2261 return 0;
2264 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2266 if (dev->priv_flags & IFF_ISATAP)
2267 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2268 return -1;
2271 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2273 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2276 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2278 memcpy(eui, dev->perm_addr, 3);
2279 memcpy(eui + 5, dev->perm_addr + 3, 3);
2280 eui[3] = 0xFF;
2281 eui[4] = 0xFE;
2282 eui[0] ^= 2;
2283 return 0;
2286 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2288 switch (dev->type) {
2289 case ARPHRD_ETHER:
2290 case ARPHRD_FDDI:
2291 return addrconf_ifid_eui48(eui, dev);
2292 case ARPHRD_ARCNET:
2293 return addrconf_ifid_arcnet(eui, dev);
2294 case ARPHRD_INFINIBAND:
2295 return addrconf_ifid_infiniband(eui, dev);
2296 case ARPHRD_SIT:
2297 return addrconf_ifid_sit(eui, dev);
2298 case ARPHRD_IPGRE:
2299 case ARPHRD_TUNNEL:
2300 return addrconf_ifid_gre(eui, dev);
2301 case ARPHRD_6LOWPAN:
2302 return addrconf_ifid_6lowpan(eui, dev);
2303 case ARPHRD_IEEE1394:
2304 return addrconf_ifid_ieee1394(eui, dev);
2305 case ARPHRD_TUNNEL6:
2306 case ARPHRD_IP6GRE:
2307 case ARPHRD_RAWIP:
2308 return addrconf_ifid_ip6tnl(eui, dev);
2310 return -1;
2313 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2315 int err = -1;
2316 struct inet6_ifaddr *ifp;
2318 read_lock_bh(&idev->lock);
2319 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2320 if (ifp->scope > IFA_LINK)
2321 break;
2322 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2323 memcpy(eui, ifp->addr.s6_addr+8, 8);
2324 err = 0;
2325 break;
2328 read_unlock_bh(&idev->lock);
2329 return err;
2332 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
2333 static void ipv6_regen_rndid(struct inet6_dev *idev)
2335 regen:
2336 get_random_bytes(idev->rndid, sizeof(idev->rndid));
2337 idev->rndid[0] &= ~0x02;
2340 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
2341 * check if generated address is not inappropriate
2343 * - Reserved subnet anycast (RFC 2526)
2344 * 11111101 11....11 1xxxxxxx
2345 * - ISATAP (RFC4214) 6.1
2346 * 00-00-5E-FE-xx-xx-xx-xx
2347 * - value 0
2348 * - XXX: already assigned to an address on the device
2350 if (idev->rndid[0] == 0xfd &&
2351 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
2352 (idev->rndid[7]&0x80))
2353 goto regen;
2354 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
2355 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
2356 goto regen;
2357 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
2358 goto regen;
2362 static void ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
2364 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
2365 ipv6_regen_rndid(idev);
2369 * Add prefix route.
2372 static void
2373 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2374 struct net_device *dev, unsigned long expires,
2375 u32 flags, gfp_t gfp_flags)
2377 struct fib6_config cfg = {
2378 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2379 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2380 .fc_ifindex = dev->ifindex,
2381 .fc_expires = expires,
2382 .fc_dst_len = plen,
2383 .fc_flags = RTF_UP | flags,
2384 .fc_nlinfo.nl_net = dev_net(dev),
2385 .fc_protocol = RTPROT_KERNEL,
2386 .fc_type = RTN_UNICAST,
2389 cfg.fc_dst = *pfx;
2391 /* Prevent useless cloning on PtP SIT.
2392 This thing is done here expecting that the whole
2393 class of non-broadcast devices need not cloning.
2395 #if IS_ENABLED(CONFIG_IPV6_SIT)
2396 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2397 cfg.fc_flags |= RTF_NONEXTHOP;
2398 #endif
2400 ip6_route_add(&cfg, gfp_flags, NULL);
2404 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2405 int plen,
2406 const struct net_device *dev,
2407 u32 flags, u32 noflags,
2408 bool no_gw)
2410 struct fib6_node *fn;
2411 struct fib6_info *rt = NULL;
2412 struct fib6_table *table;
2413 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2415 table = fib6_get_table(dev_net(dev), tb_id);
2416 if (!table)
2417 return NULL;
2419 rcu_read_lock();
2420 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2421 if (!fn)
2422 goto out;
2424 for_each_fib6_node_rt_rcu(fn) {
2425 /* prefix routes only use builtin fib6_nh */
2426 if (rt->nh)
2427 continue;
2429 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2430 continue;
2431 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2432 continue;
2433 if ((rt->fib6_flags & flags) != flags)
2434 continue;
2435 if ((rt->fib6_flags & noflags) != 0)
2436 continue;
2437 if (!fib6_info_hold_safe(rt))
2438 continue;
2439 break;
2441 out:
2442 rcu_read_unlock();
2443 return rt;
2447 /* Create "default" multicast route to the interface */
2449 static void addrconf_add_mroute(struct net_device *dev)
2451 struct fib6_config cfg = {
2452 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2453 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2454 .fc_ifindex = dev->ifindex,
2455 .fc_dst_len = 8,
2456 .fc_flags = RTF_UP,
2457 .fc_type = RTN_UNICAST,
2458 .fc_nlinfo.nl_net = dev_net(dev),
2461 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2463 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2466 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2468 struct inet6_dev *idev;
2470 ASSERT_RTNL();
2472 idev = ipv6_find_idev(dev);
2473 if (IS_ERR(idev))
2474 return idev;
2476 if (idev->cnf.disable_ipv6)
2477 return ERR_PTR(-EACCES);
2479 /* Add default multicast route */
2480 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2481 addrconf_add_mroute(dev);
2483 return idev;
2486 static void manage_tempaddrs(struct inet6_dev *idev,
2487 struct inet6_ifaddr *ifp,
2488 __u32 valid_lft, __u32 prefered_lft,
2489 bool create, unsigned long now)
2491 u32 flags;
2492 struct inet6_ifaddr *ift;
2494 read_lock_bh(&idev->lock);
2495 /* update all temporary addresses in the list */
2496 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2497 int age, max_valid, max_prefered;
2499 if (ifp != ift->ifpub)
2500 continue;
2502 /* RFC 4941 section 3.3:
2503 * If a received option will extend the lifetime of a public
2504 * address, the lifetimes of temporary addresses should
2505 * be extended, subject to the overall constraint that no
2506 * temporary addresses should ever remain "valid" or "preferred"
2507 * for a time longer than (TEMP_VALID_LIFETIME) or
2508 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2510 age = (now - ift->cstamp) / HZ;
2511 max_valid = idev->cnf.temp_valid_lft - age;
2512 if (max_valid < 0)
2513 max_valid = 0;
2515 max_prefered = idev->cnf.temp_prefered_lft -
2516 idev->desync_factor - age;
2517 if (max_prefered < 0)
2518 max_prefered = 0;
2520 if (valid_lft > max_valid)
2521 valid_lft = max_valid;
2523 if (prefered_lft > max_prefered)
2524 prefered_lft = max_prefered;
2526 spin_lock(&ift->lock);
2527 flags = ift->flags;
2528 ift->valid_lft = valid_lft;
2529 ift->prefered_lft = prefered_lft;
2530 ift->tstamp = now;
2531 if (prefered_lft > 0)
2532 ift->flags &= ~IFA_F_DEPRECATED;
2534 spin_unlock(&ift->lock);
2535 if (!(flags&IFA_F_TENTATIVE))
2536 ipv6_ifa_notify(0, ift);
2539 if ((create || list_empty(&idev->tempaddr_list)) &&
2540 idev->cnf.use_tempaddr > 0) {
2541 /* When a new public address is created as described
2542 * in [ADDRCONF], also create a new temporary address.
2543 * Also create a temporary address if it's enabled but
2544 * no temporary address currently exists.
2546 read_unlock_bh(&idev->lock);
2547 ipv6_create_tempaddr(ifp, NULL, false);
2548 } else {
2549 read_unlock_bh(&idev->lock);
2553 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2555 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2556 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2559 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2560 const struct prefix_info *pinfo,
2561 struct inet6_dev *in6_dev,
2562 const struct in6_addr *addr, int addr_type,
2563 u32 addr_flags, bool sllao, bool tokenized,
2564 __u32 valid_lft, u32 prefered_lft)
2566 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2567 int create = 0;
2569 if (!ifp && valid_lft) {
2570 int max_addresses = in6_dev->cnf.max_addresses;
2571 struct ifa6_config cfg = {
2572 .pfx = addr,
2573 .plen = pinfo->prefix_len,
2574 .ifa_flags = addr_flags,
2575 .valid_lft = valid_lft,
2576 .preferred_lft = prefered_lft,
2577 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2580 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2581 if ((net->ipv6.devconf_all->optimistic_dad ||
2582 in6_dev->cnf.optimistic_dad) &&
2583 !net->ipv6.devconf_all->forwarding && sllao)
2584 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2585 #endif
2587 /* Do not allow to create too much of autoconfigured
2588 * addresses; this would be too easy way to crash kernel.
2590 if (!max_addresses ||
2591 ipv6_count_addresses(in6_dev) < max_addresses)
2592 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2594 if (IS_ERR_OR_NULL(ifp))
2595 return -1;
2597 create = 1;
2598 spin_lock_bh(&ifp->lock);
2599 ifp->flags |= IFA_F_MANAGETEMPADDR;
2600 ifp->cstamp = jiffies;
2601 ifp->tokenized = tokenized;
2602 spin_unlock_bh(&ifp->lock);
2603 addrconf_dad_start(ifp);
2606 if (ifp) {
2607 u32 flags;
2608 unsigned long now;
2609 u32 stored_lft;
2611 /* Update lifetime (RFC4862 5.5.3 e)
2612 * We deviate from RFC4862 by honoring all Valid Lifetimes to
2613 * improve the reaction of SLAAC to renumbering events
2614 * (draft-gont-6man-slaac-renum-06, Section 4.2)
2616 spin_lock_bh(&ifp->lock);
2617 now = jiffies;
2618 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2619 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2620 else
2621 stored_lft = 0;
2623 if (!create && stored_lft) {
2624 ifp->valid_lft = valid_lft;
2625 ifp->prefered_lft = prefered_lft;
2626 ifp->tstamp = now;
2627 flags = ifp->flags;
2628 ifp->flags &= ~IFA_F_DEPRECATED;
2629 spin_unlock_bh(&ifp->lock);
2631 if (!(flags&IFA_F_TENTATIVE))
2632 ipv6_ifa_notify(0, ifp);
2633 } else
2634 spin_unlock_bh(&ifp->lock);
2636 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2637 create, now);
2639 in6_ifa_put(ifp);
2640 addrconf_verify();
2643 return 0;
2645 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2647 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2649 struct prefix_info *pinfo;
2650 __u32 valid_lft;
2651 __u32 prefered_lft;
2652 int addr_type, err;
2653 u32 addr_flags = 0;
2654 struct inet6_dev *in6_dev;
2655 struct net *net = dev_net(dev);
2657 pinfo = (struct prefix_info *) opt;
2659 if (len < sizeof(struct prefix_info)) {
2660 netdev_dbg(dev, "addrconf: prefix option too short\n");
2661 return;
2665 * Validation checks ([ADDRCONF], page 19)
2668 addr_type = ipv6_addr_type(&pinfo->prefix);
2670 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2671 return;
2673 valid_lft = ntohl(pinfo->valid);
2674 prefered_lft = ntohl(pinfo->prefered);
2676 if (prefered_lft > valid_lft) {
2677 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2678 return;
2681 in6_dev = in6_dev_get(dev);
2683 if (!in6_dev) {
2684 net_dbg_ratelimited("addrconf: device %s not configured\n",
2685 dev->name);
2686 return;
2690 * Two things going on here:
2691 * 1) Add routes for on-link prefixes
2692 * 2) Configure prefixes with the auto flag set
2695 if (pinfo->onlink) {
2696 struct fib6_info *rt;
2697 unsigned long rt_expires;
2699 /* Avoid arithmetic overflow. Really, we could
2700 * save rt_expires in seconds, likely valid_lft,
2701 * but it would require division in fib gc, that it
2702 * not good.
2704 if (HZ > USER_HZ)
2705 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2706 else
2707 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2709 if (addrconf_finite_timeout(rt_expires))
2710 rt_expires *= HZ;
2712 rt = addrconf_get_prefix_route(&pinfo->prefix,
2713 pinfo->prefix_len,
2714 dev,
2715 RTF_ADDRCONF | RTF_PREFIX_RT,
2716 RTF_DEFAULT, true);
2718 if (rt) {
2719 /* Autoconf prefix route */
2720 if (valid_lft == 0) {
2721 ip6_del_rt(net, rt, false);
2722 rt = NULL;
2723 } else if (addrconf_finite_timeout(rt_expires)) {
2724 /* not infinity */
2725 fib6_set_expires(rt, jiffies + rt_expires);
2726 } else {
2727 fib6_clean_expires(rt);
2729 } else if (valid_lft) {
2730 clock_t expires = 0;
2731 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2732 if (addrconf_finite_timeout(rt_expires)) {
2733 /* not infinity */
2734 flags |= RTF_EXPIRES;
2735 expires = jiffies_to_clock_t(rt_expires);
2737 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2738 0, dev, expires, flags,
2739 GFP_ATOMIC);
2741 fib6_info_release(rt);
2744 /* Try to figure out our local address for this prefix */
2746 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2747 struct in6_addr addr;
2748 bool tokenized = false, dev_addr_generated = false;
2750 if (pinfo->prefix_len == 64) {
2751 memcpy(&addr, &pinfo->prefix, 8);
2753 if (!ipv6_addr_any(&in6_dev->token)) {
2754 read_lock_bh(&in6_dev->lock);
2755 memcpy(addr.s6_addr + 8,
2756 in6_dev->token.s6_addr + 8, 8);
2757 read_unlock_bh(&in6_dev->lock);
2758 tokenized = true;
2759 } else if (is_addr_mode_generate_stable(in6_dev) &&
2760 !ipv6_generate_stable_address(&addr, 0,
2761 in6_dev)) {
2762 addr_flags |= IFA_F_STABLE_PRIVACY;
2763 goto ok;
2764 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2765 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2766 goto put;
2767 } else {
2768 dev_addr_generated = true;
2770 goto ok;
2772 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2773 pinfo->prefix_len);
2774 goto put;
2777 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2778 &addr, addr_type,
2779 addr_flags, sllao,
2780 tokenized, valid_lft,
2781 prefered_lft);
2782 if (err)
2783 goto put;
2785 /* Ignore error case here because previous prefix add addr was
2786 * successful which will be notified.
2788 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2789 addr_type, addr_flags, sllao,
2790 tokenized, valid_lft,
2791 prefered_lft,
2792 dev_addr_generated);
2794 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2795 put:
2796 in6_dev_put(in6_dev);
2800 * Set destination address.
2801 * Special case for SIT interfaces where we create a new "virtual"
2802 * device.
2804 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2806 struct in6_ifreq ireq;
2807 struct net_device *dev;
2808 int err = -EINVAL;
2810 rtnl_lock();
2812 err = -EFAULT;
2813 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2814 goto err_exit;
2816 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2818 err = -ENODEV;
2819 if (!dev)
2820 goto err_exit;
2822 #if IS_ENABLED(CONFIG_IPV6_SIT)
2823 if (dev->type == ARPHRD_SIT) {
2824 const struct net_device_ops *ops = dev->netdev_ops;
2825 struct ifreq ifr;
2826 struct ip_tunnel_parm p;
2828 err = -EADDRNOTAVAIL;
2829 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2830 goto err_exit;
2832 memset(&p, 0, sizeof(p));
2833 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2834 p.iph.saddr = 0;
2835 p.iph.version = 4;
2836 p.iph.ihl = 5;
2837 p.iph.protocol = IPPROTO_IPV6;
2838 p.iph.ttl = 64;
2839 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2841 if (ops->ndo_do_ioctl) {
2842 mm_segment_t oldfs = get_fs();
2844 set_fs(KERNEL_DS);
2845 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2846 set_fs(oldfs);
2847 } else
2848 err = -EOPNOTSUPP;
2850 if (err == 0) {
2851 err = -ENOBUFS;
2852 dev = __dev_get_by_name(net, p.name);
2853 if (!dev)
2854 goto err_exit;
2855 err = dev_open(dev, NULL);
2858 #endif
2860 err_exit:
2861 rtnl_unlock();
2862 return err;
2865 static int ipv6_mc_config(struct sock *sk, bool join,
2866 const struct in6_addr *addr, int ifindex)
2868 int ret;
2870 ASSERT_RTNL();
2872 lock_sock(sk);
2873 if (join)
2874 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2875 else
2876 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2877 release_sock(sk);
2879 return ret;
2883 * Manual configuration of address on an interface
2885 static int inet6_addr_add(struct net *net, int ifindex,
2886 struct ifa6_config *cfg,
2887 struct netlink_ext_ack *extack)
2889 struct inet6_ifaddr *ifp;
2890 struct inet6_dev *idev;
2891 struct net_device *dev;
2892 unsigned long timeout;
2893 clock_t expires;
2894 u32 flags;
2896 ASSERT_RTNL();
2898 if (cfg->plen > 128)
2899 return -EINVAL;
2901 /* check the lifetime */
2902 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
2903 return -EINVAL;
2905 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
2906 return -EINVAL;
2908 dev = __dev_get_by_index(net, ifindex);
2909 if (!dev)
2910 return -ENODEV;
2912 idev = addrconf_add_dev(dev);
2913 if (IS_ERR(idev))
2914 return PTR_ERR(idev);
2916 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2917 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
2918 true, cfg->pfx, ifindex);
2920 if (ret < 0)
2921 return ret;
2924 cfg->scope = ipv6_addr_scope(cfg->pfx);
2926 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
2927 if (addrconf_finite_timeout(timeout)) {
2928 expires = jiffies_to_clock_t(timeout * HZ);
2929 cfg->valid_lft = timeout;
2930 flags = RTF_EXPIRES;
2931 } else {
2932 expires = 0;
2933 flags = 0;
2934 cfg->ifa_flags |= IFA_F_PERMANENT;
2937 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
2938 if (addrconf_finite_timeout(timeout)) {
2939 if (timeout == 0)
2940 cfg->ifa_flags |= IFA_F_DEPRECATED;
2941 cfg->preferred_lft = timeout;
2944 ifp = ipv6_add_addr(idev, cfg, true, extack);
2945 if (!IS_ERR(ifp)) {
2946 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
2947 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
2948 ifp->rt_priority, dev, expires,
2949 flags, GFP_KERNEL);
2952 /* Send a netlink notification if DAD is enabled and
2953 * optimistic flag is not set
2955 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
2956 ipv6_ifa_notify(0, ifp);
2958 * Note that section 3.1 of RFC 4429 indicates
2959 * that the Optimistic flag should not be set for
2960 * manually configured addresses
2962 addrconf_dad_start(ifp);
2963 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
2964 manage_tempaddrs(idev, ifp, cfg->valid_lft,
2965 cfg->preferred_lft, true, jiffies);
2966 in6_ifa_put(ifp);
2967 addrconf_verify_rtnl();
2968 return 0;
2969 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
2970 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
2971 cfg->pfx, ifindex);
2974 return PTR_ERR(ifp);
2977 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2978 const struct in6_addr *pfx, unsigned int plen)
2980 struct inet6_ifaddr *ifp;
2981 struct inet6_dev *idev;
2982 struct net_device *dev;
2984 if (plen > 128)
2985 return -EINVAL;
2987 dev = __dev_get_by_index(net, ifindex);
2988 if (!dev)
2989 return -ENODEV;
2991 idev = __in6_dev_get(dev);
2992 if (!idev)
2993 return -ENXIO;
2995 read_lock_bh(&idev->lock);
2996 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2997 if (ifp->prefix_len == plen &&
2998 ipv6_addr_equal(pfx, &ifp->addr)) {
2999 in6_ifa_hold(ifp);
3000 read_unlock_bh(&idev->lock);
3002 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3003 (ifa_flags & IFA_F_MANAGETEMPADDR))
3004 manage_tempaddrs(idev, ifp, 0, 0, false,
3005 jiffies);
3006 ipv6_del_addr(ifp);
3007 addrconf_verify_rtnl();
3008 if (ipv6_addr_is_multicast(pfx)) {
3009 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3010 false, pfx, dev->ifindex);
3012 return 0;
3015 read_unlock_bh(&idev->lock);
3016 return -EADDRNOTAVAIL;
3020 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3022 struct ifa6_config cfg = {
3023 .ifa_flags = IFA_F_PERMANENT,
3024 .preferred_lft = INFINITY_LIFE_TIME,
3025 .valid_lft = INFINITY_LIFE_TIME,
3027 struct in6_ifreq ireq;
3028 int err;
3030 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3031 return -EPERM;
3033 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3034 return -EFAULT;
3036 cfg.pfx = &ireq.ifr6_addr;
3037 cfg.plen = ireq.ifr6_prefixlen;
3039 rtnl_lock();
3040 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3041 rtnl_unlock();
3042 return err;
3045 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3047 struct in6_ifreq ireq;
3048 int err;
3050 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3051 return -EPERM;
3053 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3054 return -EFAULT;
3056 rtnl_lock();
3057 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3058 ireq.ifr6_prefixlen);
3059 rtnl_unlock();
3060 return err;
3063 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3064 int plen, int scope)
3066 struct inet6_ifaddr *ifp;
3067 struct ifa6_config cfg = {
3068 .pfx = addr,
3069 .plen = plen,
3070 .ifa_flags = IFA_F_PERMANENT,
3071 .valid_lft = INFINITY_LIFE_TIME,
3072 .preferred_lft = INFINITY_LIFE_TIME,
3073 .scope = scope
3076 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3077 if (!IS_ERR(ifp)) {
3078 spin_lock_bh(&ifp->lock);
3079 ifp->flags &= ~IFA_F_TENTATIVE;
3080 spin_unlock_bh(&ifp->lock);
3081 rt_genid_bump_ipv6(dev_net(idev->dev));
3082 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3083 in6_ifa_put(ifp);
3087 #if IS_ENABLED(CONFIG_IPV6_SIT)
3088 static void sit_add_v4_addrs(struct inet6_dev *idev)
3090 struct in6_addr addr;
3091 struct net_device *dev;
3092 struct net *net = dev_net(idev->dev);
3093 int scope, plen;
3094 u32 pflags = 0;
3096 ASSERT_RTNL();
3098 memset(&addr, 0, sizeof(struct in6_addr));
3099 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
3101 if (idev->dev->flags&IFF_POINTOPOINT) {
3102 addr.s6_addr32[0] = htonl(0xfe800000);
3103 scope = IFA_LINK;
3104 plen = 64;
3105 } else {
3106 scope = IPV6_ADDR_COMPATv4;
3107 plen = 96;
3108 pflags |= RTF_NONEXTHOP;
3111 if (addr.s6_addr32[3]) {
3112 add_addr(idev, &addr, plen, scope);
3113 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3114 GFP_KERNEL);
3115 return;
3118 for_each_netdev(net, dev) {
3119 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3120 if (in_dev && (dev->flags & IFF_UP)) {
3121 struct in_ifaddr *ifa;
3122 int flag = scope;
3124 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3125 addr.s6_addr32[3] = ifa->ifa_local;
3127 if (ifa->ifa_scope == RT_SCOPE_LINK)
3128 continue;
3129 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3130 if (idev->dev->flags&IFF_POINTOPOINT)
3131 continue;
3132 flag |= IFA_HOST;
3135 add_addr(idev, &addr, plen, flag);
3136 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3137 0, pflags, GFP_KERNEL);
3142 #endif
3144 static void init_loopback(struct net_device *dev)
3146 struct inet6_dev *idev;
3148 /* ::1 */
3150 ASSERT_RTNL();
3152 idev = ipv6_find_idev(dev);
3153 if (IS_ERR(idev)) {
3154 pr_debug("%s: add_dev failed\n", __func__);
3155 return;
3158 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
3161 void addrconf_add_linklocal(struct inet6_dev *idev,
3162 const struct in6_addr *addr, u32 flags)
3164 struct ifa6_config cfg = {
3165 .pfx = addr,
3166 .plen = 64,
3167 .ifa_flags = flags | IFA_F_PERMANENT,
3168 .valid_lft = INFINITY_LIFE_TIME,
3169 .preferred_lft = INFINITY_LIFE_TIME,
3170 .scope = IFA_LINK
3172 struct inet6_ifaddr *ifp;
3174 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3175 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
3176 idev->cnf.optimistic_dad) &&
3177 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3178 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3179 #endif
3181 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3182 if (!IS_ERR(ifp)) {
3183 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3184 0, 0, GFP_ATOMIC);
3185 addrconf_dad_start(ifp);
3186 in6_ifa_put(ifp);
3189 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3191 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3193 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3194 return true;
3196 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3197 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3198 return true;
3200 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3201 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3202 return true;
3204 return false;
3207 static int ipv6_generate_stable_address(struct in6_addr *address,
3208 u8 dad_count,
3209 const struct inet6_dev *idev)
3211 static DEFINE_SPINLOCK(lock);
3212 static __u32 digest[SHA_DIGEST_WORDS];
3213 static __u32 workspace[SHA_WORKSPACE_WORDS];
3215 static union {
3216 char __data[SHA_MESSAGE_BYTES];
3217 struct {
3218 struct in6_addr secret;
3219 __be32 prefix[2];
3220 unsigned char hwaddr[MAX_ADDR_LEN];
3221 u8 dad_count;
3222 } __packed;
3223 } data;
3225 struct in6_addr secret;
3226 struct in6_addr temp;
3227 struct net *net = dev_net(idev->dev);
3229 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3231 if (idev->cnf.stable_secret.initialized)
3232 secret = idev->cnf.stable_secret.secret;
3233 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3234 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3235 else
3236 return -1;
3238 retry:
3239 spin_lock_bh(&lock);
3241 sha_init(digest);
3242 memset(&data, 0, sizeof(data));
3243 memset(workspace, 0, sizeof(workspace));
3244 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3245 data.prefix[0] = address->s6_addr32[0];
3246 data.prefix[1] = address->s6_addr32[1];
3247 data.secret = secret;
3248 data.dad_count = dad_count;
3250 sha_transform(digest, data.__data, workspace);
3252 temp = *address;
3253 temp.s6_addr32[2] = (__force __be32)digest[0];
3254 temp.s6_addr32[3] = (__force __be32)digest[1];
3256 spin_unlock_bh(&lock);
3258 if (ipv6_reserved_interfaceid(temp)) {
3259 dad_count++;
3260 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3261 return -1;
3262 goto retry;
3265 *address = temp;
3266 return 0;
3269 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3271 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3273 if (s->initialized)
3274 return;
3275 s = &idev->cnf.stable_secret;
3276 get_random_bytes(&s->secret, sizeof(s->secret));
3277 s->initialized = true;
3280 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3282 struct in6_addr addr;
3284 /* no link local addresses on L3 master devices */
3285 if (netif_is_l3_master(idev->dev))
3286 return;
3288 /* no link local addresses on devices flagged as slaves */
3289 if (idev->dev->flags & IFF_SLAVE)
3290 return;
3292 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3294 switch (idev->cnf.addr_gen_mode) {
3295 case IN6_ADDR_GEN_MODE_RANDOM:
3296 ipv6_gen_mode_random_init(idev);
3297 fallthrough;
3298 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3299 if (!ipv6_generate_stable_address(&addr, 0, idev))
3300 addrconf_add_linklocal(idev, &addr,
3301 IFA_F_STABLE_PRIVACY);
3302 else if (prefix_route)
3303 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3304 0, 0, GFP_KERNEL);
3305 break;
3306 case IN6_ADDR_GEN_MODE_EUI64:
3307 /* addrconf_add_linklocal also adds a prefix_route and we
3308 * only need to care about prefix routes if ipv6_generate_eui64
3309 * couldn't generate one.
3311 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3312 addrconf_add_linklocal(idev, &addr, 0);
3313 else if (prefix_route)
3314 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3315 0, 0, GFP_KERNEL);
3316 break;
3317 case IN6_ADDR_GEN_MODE_NONE:
3318 default:
3319 /* will not add any link local address */
3320 break;
3324 static void addrconf_dev_config(struct net_device *dev)
3326 struct inet6_dev *idev;
3328 ASSERT_RTNL();
3330 if ((dev->type != ARPHRD_ETHER) &&
3331 (dev->type != ARPHRD_FDDI) &&
3332 (dev->type != ARPHRD_ARCNET) &&
3333 (dev->type != ARPHRD_INFINIBAND) &&
3334 (dev->type != ARPHRD_IEEE1394) &&
3335 (dev->type != ARPHRD_TUNNEL6) &&
3336 (dev->type != ARPHRD_6LOWPAN) &&
3337 (dev->type != ARPHRD_IP6GRE) &&
3338 (dev->type != ARPHRD_IPGRE) &&
3339 (dev->type != ARPHRD_TUNNEL) &&
3340 (dev->type != ARPHRD_NONE) &&
3341 (dev->type != ARPHRD_RAWIP)) {
3342 /* Alas, we support only Ethernet autoconfiguration. */
3343 idev = __in6_dev_get(dev);
3344 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3345 dev->flags & IFF_MULTICAST)
3346 ipv6_mc_up(idev);
3347 return;
3350 idev = addrconf_add_dev(dev);
3351 if (IS_ERR(idev))
3352 return;
3354 /* this device type has no EUI support */
3355 if (dev->type == ARPHRD_NONE &&
3356 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3357 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
3359 addrconf_addr_gen(idev, false);
3362 #if IS_ENABLED(CONFIG_IPV6_SIT)
3363 static void addrconf_sit_config(struct net_device *dev)
3365 struct inet6_dev *idev;
3367 ASSERT_RTNL();
3370 * Configure the tunnel with one of our IPv4
3371 * addresses... we should configure all of
3372 * our v4 addrs in the tunnel
3375 idev = ipv6_find_idev(dev);
3376 if (IS_ERR(idev)) {
3377 pr_debug("%s: add_dev failed\n", __func__);
3378 return;
3381 if (dev->priv_flags & IFF_ISATAP) {
3382 addrconf_addr_gen(idev, false);
3383 return;
3386 sit_add_v4_addrs(idev);
3388 if (dev->flags&IFF_POINTOPOINT)
3389 addrconf_add_mroute(dev);
3391 #endif
3393 #if IS_ENABLED(CONFIG_NET_IPGRE)
3394 static void addrconf_gre_config(struct net_device *dev)
3396 struct inet6_dev *idev;
3398 ASSERT_RTNL();
3400 idev = ipv6_find_idev(dev);
3401 if (IS_ERR(idev)) {
3402 pr_debug("%s: add_dev failed\n", __func__);
3403 return;
3406 addrconf_addr_gen(idev, true);
3407 if (dev->flags & IFF_POINTOPOINT)
3408 addrconf_add_mroute(dev);
3410 #endif
3412 static int fixup_permanent_addr(struct net *net,
3413 struct inet6_dev *idev,
3414 struct inet6_ifaddr *ifp)
3416 /* !fib6_node means the host route was removed from the
3417 * FIB, for example, if 'lo' device is taken down. In that
3418 * case regenerate the host route.
3420 if (!ifp->rt || !ifp->rt->fib6_node) {
3421 struct fib6_info *f6i, *prev;
3423 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3424 GFP_ATOMIC);
3425 if (IS_ERR(f6i))
3426 return PTR_ERR(f6i);
3428 /* ifp->rt can be accessed outside of rtnl */
3429 spin_lock(&ifp->lock);
3430 prev = ifp->rt;
3431 ifp->rt = f6i;
3432 spin_unlock(&ifp->lock);
3434 fib6_info_release(prev);
3437 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3438 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3439 ifp->rt_priority, idev->dev, 0, 0,
3440 GFP_ATOMIC);
3443 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3444 addrconf_dad_start(ifp);
3446 return 0;
3449 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3451 struct inet6_ifaddr *ifp, *tmp;
3452 struct inet6_dev *idev;
3454 idev = __in6_dev_get(dev);
3455 if (!idev)
3456 return;
3458 write_lock_bh(&idev->lock);
3460 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3461 if ((ifp->flags & IFA_F_PERMANENT) &&
3462 fixup_permanent_addr(net, idev, ifp) < 0) {
3463 write_unlock_bh(&idev->lock);
3464 in6_ifa_hold(ifp);
3465 ipv6_del_addr(ifp);
3466 write_lock_bh(&idev->lock);
3468 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3469 idev->dev->name, &ifp->addr);
3473 write_unlock_bh(&idev->lock);
3476 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3477 void *ptr)
3479 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3480 struct netdev_notifier_change_info *change_info;
3481 struct netdev_notifier_changeupper_info *info;
3482 struct inet6_dev *idev = __in6_dev_get(dev);
3483 struct net *net = dev_net(dev);
3484 int run_pending = 0;
3485 int err;
3487 switch (event) {
3488 case NETDEV_REGISTER:
3489 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3490 idev = ipv6_add_dev(dev);
3491 if (IS_ERR(idev))
3492 return notifier_from_errno(PTR_ERR(idev));
3494 break;
3496 case NETDEV_CHANGEMTU:
3497 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3498 if (dev->mtu < IPV6_MIN_MTU) {
3499 addrconf_ifdown(dev, dev != net->loopback_dev);
3500 break;
3503 if (idev) {
3504 rt6_mtu_change(dev, dev->mtu);
3505 idev->cnf.mtu6 = dev->mtu;
3506 break;
3509 /* allocate new idev */
3510 idev = ipv6_add_dev(dev);
3511 if (IS_ERR(idev))
3512 break;
3514 /* device is still not ready */
3515 if (!(idev->if_flags & IF_READY))
3516 break;
3518 run_pending = 1;
3519 fallthrough;
3520 case NETDEV_UP:
3521 case NETDEV_CHANGE:
3522 if (dev->flags & IFF_SLAVE)
3523 break;
3525 if (idev && idev->cnf.disable_ipv6)
3526 break;
3528 if (event == NETDEV_UP) {
3529 /* restore routes for permanent addresses */
3530 addrconf_permanent_addr(net, dev);
3532 if (!addrconf_link_ready(dev)) {
3533 /* device is not ready yet. */
3534 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3535 dev->name);
3536 break;
3539 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3540 idev = ipv6_add_dev(dev);
3542 if (!IS_ERR_OR_NULL(idev)) {
3543 idev->if_flags |= IF_READY;
3544 run_pending = 1;
3546 } else if (event == NETDEV_CHANGE) {
3547 if (!addrconf_link_ready(dev)) {
3548 /* device is still not ready. */
3549 rt6_sync_down_dev(dev, event);
3550 break;
3553 if (!IS_ERR_OR_NULL(idev)) {
3554 if (idev->if_flags & IF_READY) {
3555 /* device is already configured -
3556 * but resend MLD reports, we might
3557 * have roamed and need to update
3558 * multicast snooping switches
3560 ipv6_mc_up(idev);
3561 change_info = ptr;
3562 if (change_info->flags_changed & IFF_NOARP)
3563 addrconf_dad_run(idev, true);
3564 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3565 break;
3567 idev->if_flags |= IF_READY;
3570 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3571 dev->name);
3573 run_pending = 1;
3576 switch (dev->type) {
3577 #if IS_ENABLED(CONFIG_IPV6_SIT)
3578 case ARPHRD_SIT:
3579 addrconf_sit_config(dev);
3580 break;
3581 #endif
3582 #if IS_ENABLED(CONFIG_NET_IPGRE)
3583 case ARPHRD_IPGRE:
3584 addrconf_gre_config(dev);
3585 break;
3586 #endif
3587 case ARPHRD_LOOPBACK:
3588 init_loopback(dev);
3589 break;
3591 default:
3592 addrconf_dev_config(dev);
3593 break;
3596 if (!IS_ERR_OR_NULL(idev)) {
3597 if (run_pending)
3598 addrconf_dad_run(idev, false);
3600 /* Device has an address by now */
3601 rt6_sync_up(dev, RTNH_F_DEAD);
3604 * If the MTU changed during the interface down,
3605 * when the interface up, the changed MTU must be
3606 * reflected in the idev as well as routers.
3608 if (idev->cnf.mtu6 != dev->mtu &&
3609 dev->mtu >= IPV6_MIN_MTU) {
3610 rt6_mtu_change(dev, dev->mtu);
3611 idev->cnf.mtu6 = dev->mtu;
3613 idev->tstamp = jiffies;
3614 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3617 * If the changed mtu during down is lower than
3618 * IPV6_MIN_MTU stop IPv6 on this interface.
3620 if (dev->mtu < IPV6_MIN_MTU)
3621 addrconf_ifdown(dev, dev != net->loopback_dev);
3623 break;
3625 case NETDEV_DOWN:
3626 case NETDEV_UNREGISTER:
3628 * Remove all addresses from this interface.
3630 addrconf_ifdown(dev, event != NETDEV_DOWN);
3631 break;
3633 case NETDEV_CHANGENAME:
3634 if (idev) {
3635 snmp6_unregister_dev(idev);
3636 addrconf_sysctl_unregister(idev);
3637 err = addrconf_sysctl_register(idev);
3638 if (err)
3639 return notifier_from_errno(err);
3640 err = snmp6_register_dev(idev);
3641 if (err) {
3642 addrconf_sysctl_unregister(idev);
3643 return notifier_from_errno(err);
3646 break;
3648 case NETDEV_PRE_TYPE_CHANGE:
3649 case NETDEV_POST_TYPE_CHANGE:
3650 if (idev)
3651 addrconf_type_change(dev, event);
3652 break;
3654 case NETDEV_CHANGEUPPER:
3655 info = ptr;
3657 /* flush all routes if dev is linked to or unlinked from
3658 * an L3 master device (e.g., VRF)
3660 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3661 addrconf_ifdown(dev, 0);
3664 return NOTIFY_OK;
3668 * addrconf module should be notified of a device going up
3670 static struct notifier_block ipv6_dev_notf = {
3671 .notifier_call = addrconf_notify,
3672 .priority = ADDRCONF_NOTIFY_PRIORITY,
3675 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3677 struct inet6_dev *idev;
3678 ASSERT_RTNL();
3680 idev = __in6_dev_get(dev);
3682 if (event == NETDEV_POST_TYPE_CHANGE)
3683 ipv6_mc_remap(idev);
3684 else if (event == NETDEV_PRE_TYPE_CHANGE)
3685 ipv6_mc_unmap(idev);
3688 static bool addr_is_local(const struct in6_addr *addr)
3690 return ipv6_addr_type(addr) &
3691 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3694 static int addrconf_ifdown(struct net_device *dev, int how)
3696 unsigned long event = how ? NETDEV_UNREGISTER : NETDEV_DOWN;
3697 struct net *net = dev_net(dev);
3698 struct inet6_dev *idev;
3699 struct inet6_ifaddr *ifa, *tmp;
3700 bool keep_addr = false;
3701 int state, i;
3703 ASSERT_RTNL();
3705 rt6_disable_ip(dev, event);
3707 idev = __in6_dev_get(dev);
3708 if (!idev)
3709 return -ENODEV;
3712 * Step 1: remove reference to ipv6 device from parent device.
3713 * Do not dev_put!
3715 if (how) {
3716 idev->dead = 1;
3718 /* protected by rtnl_lock */
3719 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3721 /* Step 1.5: remove snmp6 entry */
3722 snmp6_unregister_dev(idev);
3726 /* combine the user config with event to determine if permanent
3727 * addresses are to be removed from address hash table
3729 if (!how && !idev->cnf.disable_ipv6) {
3730 /* aggregate the system setting and interface setting */
3731 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
3733 if (!_keep_addr)
3734 _keep_addr = idev->cnf.keep_addr_on_down;
3736 keep_addr = (_keep_addr > 0);
3739 /* Step 2: clear hash table */
3740 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3741 struct hlist_head *h = &inet6_addr_lst[i];
3743 spin_lock_bh(&addrconf_hash_lock);
3744 restart:
3745 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3746 if (ifa->idev == idev) {
3747 addrconf_del_dad_work(ifa);
3748 /* combined flag + permanent flag decide if
3749 * address is retained on a down event
3751 if (!keep_addr ||
3752 !(ifa->flags & IFA_F_PERMANENT) ||
3753 addr_is_local(&ifa->addr)) {
3754 hlist_del_init_rcu(&ifa->addr_lst);
3755 goto restart;
3759 spin_unlock_bh(&addrconf_hash_lock);
3762 write_lock_bh(&idev->lock);
3764 addrconf_del_rs_timer(idev);
3766 /* Step 2: clear flags for stateless addrconf */
3767 if (!how)
3768 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3770 /* Step 3: clear tempaddr list */
3771 while (!list_empty(&idev->tempaddr_list)) {
3772 ifa = list_first_entry(&idev->tempaddr_list,
3773 struct inet6_ifaddr, tmp_list);
3774 list_del(&ifa->tmp_list);
3775 write_unlock_bh(&idev->lock);
3776 spin_lock_bh(&ifa->lock);
3778 if (ifa->ifpub) {
3779 in6_ifa_put(ifa->ifpub);
3780 ifa->ifpub = NULL;
3782 spin_unlock_bh(&ifa->lock);
3783 in6_ifa_put(ifa);
3784 write_lock_bh(&idev->lock);
3787 list_for_each_entry_safe(ifa, tmp, &idev->addr_list, if_list) {
3788 struct fib6_info *rt = NULL;
3789 bool keep;
3791 addrconf_del_dad_work(ifa);
3793 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3794 !addr_is_local(&ifa->addr);
3796 write_unlock_bh(&idev->lock);
3797 spin_lock_bh(&ifa->lock);
3799 if (keep) {
3800 /* set state to skip the notifier below */
3801 state = INET6_IFADDR_STATE_DEAD;
3802 ifa->state = INET6_IFADDR_STATE_PREDAD;
3803 if (!(ifa->flags & IFA_F_NODAD))
3804 ifa->flags |= IFA_F_TENTATIVE;
3806 rt = ifa->rt;
3807 ifa->rt = NULL;
3808 } else {
3809 state = ifa->state;
3810 ifa->state = INET6_IFADDR_STATE_DEAD;
3813 spin_unlock_bh(&ifa->lock);
3815 if (rt)
3816 ip6_del_rt(net, rt, false);
3818 if (state != INET6_IFADDR_STATE_DEAD) {
3819 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3820 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3821 } else {
3822 if (idev->cnf.forwarding)
3823 addrconf_leave_anycast(ifa);
3824 addrconf_leave_solict(ifa->idev, &ifa->addr);
3827 write_lock_bh(&idev->lock);
3828 if (!keep) {
3829 list_del_rcu(&ifa->if_list);
3830 in6_ifa_put(ifa);
3834 write_unlock_bh(&idev->lock);
3836 /* Step 5: Discard anycast and multicast list */
3837 if (how) {
3838 ipv6_ac_destroy_dev(idev);
3839 ipv6_mc_destroy_dev(idev);
3840 } else {
3841 ipv6_mc_down(idev);
3844 idev->tstamp = jiffies;
3846 /* Last: Shot the device (if unregistered) */
3847 if (how) {
3848 addrconf_sysctl_unregister(idev);
3849 neigh_parms_release(&nd_tbl, idev->nd_parms);
3850 neigh_ifdown(&nd_tbl, dev);
3851 in6_dev_put(idev);
3853 return 0;
3856 static void addrconf_rs_timer(struct timer_list *t)
3858 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
3859 struct net_device *dev = idev->dev;
3860 struct in6_addr lladdr;
3862 write_lock(&idev->lock);
3863 if (idev->dead || !(idev->if_flags & IF_READY))
3864 goto out;
3866 if (!ipv6_accept_ra(idev))
3867 goto out;
3869 /* Announcement received after solicitation was sent */
3870 if (idev->if_flags & IF_RA_RCVD)
3871 goto out;
3873 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
3874 write_unlock(&idev->lock);
3875 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3876 ndisc_send_rs(dev, &lladdr,
3877 &in6addr_linklocal_allrouters);
3878 else
3879 goto put;
3881 write_lock(&idev->lock);
3882 idev->rs_interval = rfc3315_s14_backoff_update(
3883 idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
3884 /* The wait after the last probe can be shorter */
3885 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3886 idev->cnf.rtr_solicits) ?
3887 idev->cnf.rtr_solicit_delay :
3888 idev->rs_interval);
3889 } else {
3891 * Note: we do not support deprecated "all on-link"
3892 * assumption any longer.
3894 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3897 out:
3898 write_unlock(&idev->lock);
3899 put:
3900 in6_dev_put(idev);
3904 * Duplicate Address Detection
3906 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3908 unsigned long rand_num;
3909 struct inet6_dev *idev = ifp->idev;
3910 u64 nonce;
3912 if (ifp->flags & IFA_F_OPTIMISTIC)
3913 rand_num = 0;
3914 else
3915 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3917 nonce = 0;
3918 if (idev->cnf.enhanced_dad ||
3919 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
3921 get_random_bytes(&nonce, 6);
3922 while (nonce == 0);
3924 ifp->dad_nonce = nonce;
3925 ifp->dad_probes = idev->cnf.dad_transmits;
3926 addrconf_mod_dad_work(ifp, rand_num);
3929 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3931 struct inet6_dev *idev = ifp->idev;
3932 struct net_device *dev = idev->dev;
3933 bool bump_id, notify = false;
3934 struct net *net;
3936 addrconf_join_solict(dev, &ifp->addr);
3938 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3940 read_lock_bh(&idev->lock);
3941 spin_lock(&ifp->lock);
3942 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3943 goto out;
3945 net = dev_net(dev);
3946 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3947 (net->ipv6.devconf_all->accept_dad < 1 &&
3948 idev->cnf.accept_dad < 1) ||
3949 !(ifp->flags&IFA_F_TENTATIVE) ||
3950 ifp->flags & IFA_F_NODAD) {
3951 bool send_na = false;
3953 if (ifp->flags & IFA_F_TENTATIVE &&
3954 !(ifp->flags & IFA_F_OPTIMISTIC))
3955 send_na = true;
3956 bump_id = ifp->flags & IFA_F_TENTATIVE;
3957 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3958 spin_unlock(&ifp->lock);
3959 read_unlock_bh(&idev->lock);
3961 addrconf_dad_completed(ifp, bump_id, send_na);
3962 return;
3965 if (!(idev->if_flags & IF_READY)) {
3966 spin_unlock(&ifp->lock);
3967 read_unlock_bh(&idev->lock);
3969 * If the device is not ready:
3970 * - keep it tentative if it is a permanent address.
3971 * - otherwise, kill it.
3973 in6_ifa_hold(ifp);
3974 addrconf_dad_stop(ifp, 0);
3975 return;
3979 * Optimistic nodes can start receiving
3980 * Frames right away
3982 if (ifp->flags & IFA_F_OPTIMISTIC) {
3983 ip6_ins_rt(net, ifp->rt);
3984 if (ipv6_use_optimistic_addr(net, idev)) {
3985 /* Because optimistic nodes can use this address,
3986 * notify listeners. If DAD fails, RTM_DELADDR is sent.
3988 notify = true;
3992 addrconf_dad_kick(ifp);
3993 out:
3994 spin_unlock(&ifp->lock);
3995 read_unlock_bh(&idev->lock);
3996 if (notify)
3997 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4000 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4002 bool begin_dad = false;
4004 spin_lock_bh(&ifp->lock);
4005 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4006 ifp->state = INET6_IFADDR_STATE_PREDAD;
4007 begin_dad = true;
4009 spin_unlock_bh(&ifp->lock);
4011 if (begin_dad)
4012 addrconf_mod_dad_work(ifp, 0);
4015 static void addrconf_dad_work(struct work_struct *w)
4017 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4018 struct inet6_ifaddr,
4019 dad_work);
4020 struct inet6_dev *idev = ifp->idev;
4021 bool bump_id, disable_ipv6 = false;
4022 struct in6_addr mcaddr;
4024 enum {
4025 DAD_PROCESS,
4026 DAD_BEGIN,
4027 DAD_ABORT,
4028 } action = DAD_PROCESS;
4030 rtnl_lock();
4032 spin_lock_bh(&ifp->lock);
4033 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4034 action = DAD_BEGIN;
4035 ifp->state = INET6_IFADDR_STATE_DAD;
4036 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4037 action = DAD_ABORT;
4038 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4040 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
4041 idev->cnf.accept_dad > 1) &&
4042 !idev->cnf.disable_ipv6 &&
4043 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4044 struct in6_addr addr;
4046 addr.s6_addr32[0] = htonl(0xfe800000);
4047 addr.s6_addr32[1] = 0;
4049 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4050 ipv6_addr_equal(&ifp->addr, &addr)) {
4051 /* DAD failed for link-local based on MAC */
4052 idev->cnf.disable_ipv6 = 1;
4054 pr_info("%s: IPv6 being disabled!\n",
4055 ifp->idev->dev->name);
4056 disable_ipv6 = true;
4060 spin_unlock_bh(&ifp->lock);
4062 if (action == DAD_BEGIN) {
4063 addrconf_dad_begin(ifp);
4064 goto out;
4065 } else if (action == DAD_ABORT) {
4066 in6_ifa_hold(ifp);
4067 addrconf_dad_stop(ifp, 1);
4068 if (disable_ipv6)
4069 addrconf_ifdown(idev->dev, 0);
4070 goto out;
4073 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4074 goto out;
4076 write_lock_bh(&idev->lock);
4077 if (idev->dead || !(idev->if_flags & IF_READY)) {
4078 write_unlock_bh(&idev->lock);
4079 goto out;
4082 spin_lock(&ifp->lock);
4083 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4084 spin_unlock(&ifp->lock);
4085 write_unlock_bh(&idev->lock);
4086 goto out;
4089 if (ifp->dad_probes == 0) {
4090 bool send_na = false;
4093 * DAD was successful
4096 if (ifp->flags & IFA_F_TENTATIVE &&
4097 !(ifp->flags & IFA_F_OPTIMISTIC))
4098 send_na = true;
4099 bump_id = ifp->flags & IFA_F_TENTATIVE;
4100 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4101 spin_unlock(&ifp->lock);
4102 write_unlock_bh(&idev->lock);
4104 addrconf_dad_completed(ifp, bump_id, send_na);
4106 goto out;
4109 ifp->dad_probes--;
4110 addrconf_mod_dad_work(ifp,
4111 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4112 HZ/100));
4113 spin_unlock(&ifp->lock);
4114 write_unlock_bh(&idev->lock);
4116 /* send a neighbour solicitation for our addr */
4117 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4118 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4119 ifp->dad_nonce);
4120 out:
4121 in6_ifa_put(ifp);
4122 rtnl_unlock();
4125 /* ifp->idev must be at least read locked */
4126 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4128 struct inet6_ifaddr *ifpiter;
4129 struct inet6_dev *idev = ifp->idev;
4131 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4132 if (ifpiter->scope > IFA_LINK)
4133 break;
4134 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4135 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4136 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4137 IFA_F_PERMANENT)
4138 return false;
4140 return true;
4143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4144 bool send_na)
4146 struct net_device *dev = ifp->idev->dev;
4147 struct in6_addr lladdr;
4148 bool send_rs, send_mld;
4150 addrconf_del_dad_work(ifp);
4153 * Configure the address for reception. Now it is valid.
4156 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4158 /* If added prefix is link local and we are prepared to process
4159 router advertisements, start sending router solicitations.
4162 read_lock_bh(&ifp->idev->lock);
4163 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4164 send_rs = send_mld &&
4165 ipv6_accept_ra(ifp->idev) &&
4166 ifp->idev->cnf.rtr_solicits != 0 &&
4167 (dev->flags&IFF_LOOPBACK) == 0;
4168 read_unlock_bh(&ifp->idev->lock);
4170 /* While dad is in progress mld report's source address is in6_addrany.
4171 * Resend with proper ll now.
4173 if (send_mld)
4174 ipv6_mc_dad_complete(ifp->idev);
4176 /* send unsolicited NA if enabled */
4177 if (send_na &&
4178 (ifp->idev->cnf.ndisc_notify ||
4179 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
4180 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4181 /*router=*/ !!ifp->idev->cnf.forwarding,
4182 /*solicited=*/ false, /*override=*/ true,
4183 /*inc_opt=*/ true);
4186 if (send_rs) {
4188 * If a host as already performed a random delay
4189 * [...] as part of DAD [...] there is no need
4190 * to delay again before sending the first RS
4192 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4193 return;
4194 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4196 write_lock_bh(&ifp->idev->lock);
4197 spin_lock(&ifp->lock);
4198 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4199 ifp->idev->cnf.rtr_solicit_interval);
4200 ifp->idev->rs_probes = 1;
4201 ifp->idev->if_flags |= IF_RS_SENT;
4202 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4203 spin_unlock(&ifp->lock);
4204 write_unlock_bh(&ifp->idev->lock);
4207 if (bump_id)
4208 rt_genid_bump_ipv6(dev_net(dev));
4210 /* Make sure that a new temporary address will be created
4211 * before this temporary address becomes deprecated.
4213 if (ifp->flags & IFA_F_TEMPORARY)
4214 addrconf_verify_rtnl();
4217 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4219 struct inet6_ifaddr *ifp;
4221 read_lock_bh(&idev->lock);
4222 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4223 spin_lock(&ifp->lock);
4224 if ((ifp->flags & IFA_F_TENTATIVE &&
4225 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4226 if (restart)
4227 ifp->state = INET6_IFADDR_STATE_PREDAD;
4228 addrconf_dad_kick(ifp);
4230 spin_unlock(&ifp->lock);
4232 read_unlock_bh(&idev->lock);
4235 #ifdef CONFIG_PROC_FS
4236 struct if6_iter_state {
4237 struct seq_net_private p;
4238 int bucket;
4239 int offset;
4242 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4244 struct if6_iter_state *state = seq->private;
4245 struct net *net = seq_file_net(seq);
4246 struct inet6_ifaddr *ifa = NULL;
4247 int p = 0;
4249 /* initial bucket if pos is 0 */
4250 if (pos == 0) {
4251 state->bucket = 0;
4252 state->offset = 0;
4255 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4256 hlist_for_each_entry_rcu(ifa, &inet6_addr_lst[state->bucket],
4257 addr_lst) {
4258 if (!net_eq(dev_net(ifa->idev->dev), net))
4259 continue;
4260 /* sync with offset */
4261 if (p < state->offset) {
4262 p++;
4263 continue;
4265 return ifa;
4268 /* prepare for next bucket */
4269 state->offset = 0;
4270 p = 0;
4272 return NULL;
4275 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4276 struct inet6_ifaddr *ifa)
4278 struct if6_iter_state *state = seq->private;
4279 struct net *net = seq_file_net(seq);
4281 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4282 if (!net_eq(dev_net(ifa->idev->dev), net))
4283 continue;
4284 state->offset++;
4285 return ifa;
4288 state->offset = 0;
4289 while (++state->bucket < IN6_ADDR_HSIZE) {
4290 hlist_for_each_entry_rcu(ifa,
4291 &inet6_addr_lst[state->bucket], addr_lst) {
4292 if (!net_eq(dev_net(ifa->idev->dev), net))
4293 continue;
4294 return ifa;
4298 return NULL;
4301 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4302 __acquires(rcu)
4304 rcu_read_lock();
4305 return if6_get_first(seq, *pos);
4308 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4310 struct inet6_ifaddr *ifa;
4312 ifa = if6_get_next(seq, v);
4313 ++*pos;
4314 return ifa;
4317 static void if6_seq_stop(struct seq_file *seq, void *v)
4318 __releases(rcu)
4320 rcu_read_unlock();
4323 static int if6_seq_show(struct seq_file *seq, void *v)
4325 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4326 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4327 &ifp->addr,
4328 ifp->idev->dev->ifindex,
4329 ifp->prefix_len,
4330 ifp->scope,
4331 (u8) ifp->flags,
4332 ifp->idev->dev->name);
4333 return 0;
4336 static const struct seq_operations if6_seq_ops = {
4337 .start = if6_seq_start,
4338 .next = if6_seq_next,
4339 .show = if6_seq_show,
4340 .stop = if6_seq_stop,
4343 static int __net_init if6_proc_net_init(struct net *net)
4345 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4346 sizeof(struct if6_iter_state)))
4347 return -ENOMEM;
4348 return 0;
4351 static void __net_exit if6_proc_net_exit(struct net *net)
4353 remove_proc_entry("if_inet6", net->proc_net);
4356 static struct pernet_operations if6_proc_net_ops = {
4357 .init = if6_proc_net_init,
4358 .exit = if6_proc_net_exit,
4361 int __init if6_proc_init(void)
4363 return register_pernet_subsys(&if6_proc_net_ops);
4366 void if6_proc_exit(void)
4368 unregister_pernet_subsys(&if6_proc_net_ops);
4370 #endif /* CONFIG_PROC_FS */
4372 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4373 /* Check if address is a home address configured on any interface. */
4374 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4376 unsigned int hash = inet6_addr_hash(net, addr);
4377 struct inet6_ifaddr *ifp = NULL;
4378 int ret = 0;
4380 rcu_read_lock();
4381 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4382 if (!net_eq(dev_net(ifp->idev->dev), net))
4383 continue;
4384 if (ipv6_addr_equal(&ifp->addr, addr) &&
4385 (ifp->flags & IFA_F_HOMEADDRESS)) {
4386 ret = 1;
4387 break;
4390 rcu_read_unlock();
4391 return ret;
4393 #endif
4395 /* RFC6554 has some algorithm to avoid loops in segment routing by
4396 * checking if the segments contains any of a local interface address.
4398 * Quote:
4400 * To detect loops in the SRH, a router MUST determine if the SRH
4401 * includes multiple addresses assigned to any interface on that router.
4402 * If such addresses appear more than once and are separated by at least
4403 * one address not assigned to that router.
4405 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4406 unsigned char nsegs)
4408 const struct in6_addr *addr;
4409 int i, ret = 0, found = 0;
4410 struct inet6_ifaddr *ifp;
4411 bool separated = false;
4412 unsigned int hash;
4413 bool hash_found;
4415 rcu_read_lock();
4416 for (i = 0; i < nsegs; i++) {
4417 addr = &segs[i];
4418 hash = inet6_addr_hash(net, addr);
4420 hash_found = false;
4421 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
4422 if (!net_eq(dev_net(ifp->idev->dev), net))
4423 continue;
4425 if (ipv6_addr_equal(&ifp->addr, addr)) {
4426 hash_found = true;
4427 break;
4431 if (hash_found) {
4432 if (found > 1 && separated) {
4433 ret = 1;
4434 break;
4437 separated = false;
4438 found++;
4439 } else {
4440 separated = true;
4443 rcu_read_unlock();
4445 return ret;
4449 * Periodic address status verification
4452 static void addrconf_verify_rtnl(void)
4454 unsigned long now, next, next_sec, next_sched;
4455 struct inet6_ifaddr *ifp;
4456 int i;
4458 ASSERT_RTNL();
4460 rcu_read_lock_bh();
4461 now = jiffies;
4462 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4464 cancel_delayed_work(&addr_chk_work);
4466 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4467 restart:
4468 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
4469 unsigned long age;
4471 /* When setting preferred_lft to a value not zero or
4472 * infinity, while valid_lft is infinity
4473 * IFA_F_PERMANENT has a non-infinity life time.
4475 if ((ifp->flags & IFA_F_PERMANENT) &&
4476 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4477 continue;
4479 spin_lock(&ifp->lock);
4480 /* We try to batch several events at once. */
4481 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4483 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4484 age >= ifp->valid_lft) {
4485 spin_unlock(&ifp->lock);
4486 in6_ifa_hold(ifp);
4487 ipv6_del_addr(ifp);
4488 goto restart;
4489 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4490 spin_unlock(&ifp->lock);
4491 continue;
4492 } else if (age >= ifp->prefered_lft) {
4493 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4494 int deprecate = 0;
4496 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4497 deprecate = 1;
4498 ifp->flags |= IFA_F_DEPRECATED;
4501 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4502 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4503 next = ifp->tstamp + ifp->valid_lft * HZ;
4505 spin_unlock(&ifp->lock);
4507 if (deprecate) {
4508 in6_ifa_hold(ifp);
4510 ipv6_ifa_notify(0, ifp);
4511 in6_ifa_put(ifp);
4512 goto restart;
4514 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
4515 !(ifp->flags&IFA_F_TENTATIVE)) {
4516 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
4517 ifp->idev->cnf.dad_transmits *
4518 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
4520 if (age >= ifp->prefered_lft - regen_advance) {
4521 struct inet6_ifaddr *ifpub = ifp->ifpub;
4522 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4523 next = ifp->tstamp + ifp->prefered_lft * HZ;
4524 if (!ifp->regen_count && ifpub) {
4525 ifp->regen_count++;
4526 in6_ifa_hold(ifp);
4527 in6_ifa_hold(ifpub);
4528 spin_unlock(&ifp->lock);
4530 spin_lock(&ifpub->lock);
4531 ifpub->regen_count = 0;
4532 spin_unlock(&ifpub->lock);
4533 rcu_read_unlock_bh();
4534 ipv6_create_tempaddr(ifpub, ifp, true);
4535 in6_ifa_put(ifpub);
4536 in6_ifa_put(ifp);
4537 rcu_read_lock_bh();
4538 goto restart;
4540 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4541 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4542 spin_unlock(&ifp->lock);
4543 } else {
4544 /* ifp->prefered_lft <= ifp->valid_lft */
4545 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4546 next = ifp->tstamp + ifp->prefered_lft * HZ;
4547 spin_unlock(&ifp->lock);
4552 next_sec = round_jiffies_up(next);
4553 next_sched = next;
4555 /* If rounded timeout is accurate enough, accept it. */
4556 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4557 next_sched = next_sec;
4559 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4560 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4561 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4563 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4564 now, next, next_sec, next_sched);
4565 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
4566 rcu_read_unlock_bh();
4569 static void addrconf_verify_work(struct work_struct *w)
4571 rtnl_lock();
4572 addrconf_verify_rtnl();
4573 rtnl_unlock();
4576 static void addrconf_verify(void)
4578 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
4581 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4582 struct in6_addr **peer_pfx)
4584 struct in6_addr *pfx = NULL;
4586 *peer_pfx = NULL;
4588 if (addr)
4589 pfx = nla_data(addr);
4591 if (local) {
4592 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4593 *peer_pfx = pfx;
4594 pfx = nla_data(local);
4597 return pfx;
4600 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4601 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4602 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4603 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4604 [IFA_FLAGS] = { .len = sizeof(u32) },
4605 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4606 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4609 static int
4610 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4611 struct netlink_ext_ack *extack)
4613 struct net *net = sock_net(skb->sk);
4614 struct ifaddrmsg *ifm;
4615 struct nlattr *tb[IFA_MAX+1];
4616 struct in6_addr *pfx, *peer_pfx;
4617 u32 ifa_flags;
4618 int err;
4620 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4621 ifa_ipv6_policy, extack);
4622 if (err < 0)
4623 return err;
4625 ifm = nlmsg_data(nlh);
4626 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4627 if (!pfx)
4628 return -EINVAL;
4630 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4632 /* We ignore other flags so far. */
4633 ifa_flags &= IFA_F_MANAGETEMPADDR;
4635 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4636 ifm->ifa_prefixlen);
4639 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4640 unsigned long expires, u32 flags,
4641 bool modify_peer)
4643 struct fib6_info *f6i;
4644 u32 prio;
4646 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4647 ifp->prefix_len,
4648 ifp->idev->dev, 0, RTF_DEFAULT, true);
4649 if (!f6i)
4650 return -ENOENT;
4652 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4653 if (f6i->fib6_metric != prio) {
4654 /* delete old one */
4655 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4657 /* add new one */
4658 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4659 ifp->prefix_len,
4660 ifp->rt_priority, ifp->idev->dev,
4661 expires, flags, GFP_KERNEL);
4662 } else {
4663 if (!expires)
4664 fib6_clean_expires(f6i);
4665 else
4666 fib6_set_expires(f6i, expires);
4668 fib6_info_release(f6i);
4671 return 0;
4674 static int inet6_addr_modify(struct inet6_ifaddr *ifp, struct ifa6_config *cfg)
4676 u32 flags;
4677 clock_t expires;
4678 unsigned long timeout;
4679 bool was_managetempaddr;
4680 bool had_prefixroute;
4681 bool new_peer = false;
4683 ASSERT_RTNL();
4685 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4686 return -EINVAL;
4688 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4689 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4690 return -EINVAL;
4692 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4693 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4695 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4696 if (addrconf_finite_timeout(timeout)) {
4697 expires = jiffies_to_clock_t(timeout * HZ);
4698 cfg->valid_lft = timeout;
4699 flags = RTF_EXPIRES;
4700 } else {
4701 expires = 0;
4702 flags = 0;
4703 cfg->ifa_flags |= IFA_F_PERMANENT;
4706 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4707 if (addrconf_finite_timeout(timeout)) {
4708 if (timeout == 0)
4709 cfg->ifa_flags |= IFA_F_DEPRECATED;
4710 cfg->preferred_lft = timeout;
4713 if (cfg->peer_pfx &&
4714 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4715 if (!ipv6_addr_any(&ifp->peer_addr))
4716 cleanup_prefix_route(ifp, expires, true, true);
4717 new_peer = true;
4720 spin_lock_bh(&ifp->lock);
4721 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4722 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4723 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4724 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4725 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4726 IFA_F_NOPREFIXROUTE);
4727 ifp->flags |= cfg->ifa_flags;
4728 ifp->tstamp = jiffies;
4729 ifp->valid_lft = cfg->valid_lft;
4730 ifp->prefered_lft = cfg->preferred_lft;
4732 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4733 ifp->rt_priority = cfg->rt_priority;
4735 if (new_peer)
4736 ifp->peer_addr = *cfg->peer_pfx;
4738 spin_unlock_bh(&ifp->lock);
4739 if (!(ifp->flags&IFA_F_TENTATIVE))
4740 ipv6_ifa_notify(0, ifp);
4742 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4743 int rc = -ENOENT;
4745 if (had_prefixroute)
4746 rc = modify_prefix_route(ifp, expires, flags, false);
4748 /* prefix route could have been deleted; if so restore it */
4749 if (rc == -ENOENT) {
4750 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4751 ifp->rt_priority, ifp->idev->dev,
4752 expires, flags, GFP_KERNEL);
4755 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4756 rc = modify_prefix_route(ifp, expires, flags, true);
4758 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4759 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4760 ifp->rt_priority, ifp->idev->dev,
4761 expires, flags, GFP_KERNEL);
4763 } else if (had_prefixroute) {
4764 enum cleanup_prefix_rt_t action;
4765 unsigned long rt_expires;
4767 write_lock_bh(&ifp->idev->lock);
4768 action = check_cleanup_prefix_route(ifp, &rt_expires);
4769 write_unlock_bh(&ifp->idev->lock);
4771 if (action != CLEANUP_PREFIX_RT_NOP) {
4772 cleanup_prefix_route(ifp, rt_expires,
4773 action == CLEANUP_PREFIX_RT_DEL, false);
4777 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4778 if (was_managetempaddr &&
4779 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4780 cfg->valid_lft = 0;
4781 cfg->preferred_lft = 0;
4783 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4784 cfg->preferred_lft, !was_managetempaddr,
4785 jiffies);
4788 addrconf_verify_rtnl();
4790 return 0;
4793 static int
4794 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4795 struct netlink_ext_ack *extack)
4797 struct net *net = sock_net(skb->sk);
4798 struct ifaddrmsg *ifm;
4799 struct nlattr *tb[IFA_MAX+1];
4800 struct in6_addr *peer_pfx;
4801 struct inet6_ifaddr *ifa;
4802 struct net_device *dev;
4803 struct inet6_dev *idev;
4804 struct ifa6_config cfg;
4805 int err;
4807 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4808 ifa_ipv6_policy, extack);
4809 if (err < 0)
4810 return err;
4812 memset(&cfg, 0, sizeof(cfg));
4814 ifm = nlmsg_data(nlh);
4815 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4816 if (!cfg.pfx)
4817 return -EINVAL;
4819 cfg.peer_pfx = peer_pfx;
4820 cfg.plen = ifm->ifa_prefixlen;
4821 if (tb[IFA_RT_PRIORITY])
4822 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4824 cfg.valid_lft = INFINITY_LIFE_TIME;
4825 cfg.preferred_lft = INFINITY_LIFE_TIME;
4827 if (tb[IFA_CACHEINFO]) {
4828 struct ifa_cacheinfo *ci;
4830 ci = nla_data(tb[IFA_CACHEINFO]);
4831 cfg.valid_lft = ci->ifa_valid;
4832 cfg.preferred_lft = ci->ifa_prefered;
4835 dev = __dev_get_by_index(net, ifm->ifa_index);
4836 if (!dev)
4837 return -ENODEV;
4839 if (tb[IFA_FLAGS])
4840 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
4841 else
4842 cfg.ifa_flags = ifm->ifa_flags;
4844 /* We ignore other flags so far. */
4845 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
4846 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
4847 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
4849 idev = ipv6_find_idev(dev);
4850 if (IS_ERR(idev))
4851 return PTR_ERR(idev);
4853 if (!ipv6_allow_optimistic_dad(net, idev))
4854 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
4856 if (cfg.ifa_flags & IFA_F_NODAD &&
4857 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
4858 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
4859 return -EINVAL;
4862 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
4863 if (!ifa) {
4865 * It would be best to check for !NLM_F_CREATE here but
4866 * userspace already relies on not having to provide this.
4868 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
4871 if (nlh->nlmsg_flags & NLM_F_EXCL ||
4872 !(nlh->nlmsg_flags & NLM_F_REPLACE))
4873 err = -EEXIST;
4874 else
4875 err = inet6_addr_modify(ifa, &cfg);
4877 in6_ifa_put(ifa);
4879 return err;
4882 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
4883 u8 scope, int ifindex)
4885 struct ifaddrmsg *ifm;
4887 ifm = nlmsg_data(nlh);
4888 ifm->ifa_family = AF_INET6;
4889 ifm->ifa_prefixlen = prefixlen;
4890 ifm->ifa_flags = flags;
4891 ifm->ifa_scope = scope;
4892 ifm->ifa_index = ifindex;
4895 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
4896 unsigned long tstamp, u32 preferred, u32 valid)
4898 struct ifa_cacheinfo ci;
4900 ci.cstamp = cstamp_delta(cstamp);
4901 ci.tstamp = cstamp_delta(tstamp);
4902 ci.ifa_prefered = preferred;
4903 ci.ifa_valid = valid;
4905 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
4908 static inline int rt_scope(int ifa_scope)
4910 if (ifa_scope & IFA_HOST)
4911 return RT_SCOPE_HOST;
4912 else if (ifa_scope & IFA_LINK)
4913 return RT_SCOPE_LINK;
4914 else if (ifa_scope & IFA_SITE)
4915 return RT_SCOPE_SITE;
4916 else
4917 return RT_SCOPE_UNIVERSE;
4920 static inline int inet6_ifaddr_msgsize(void)
4922 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
4923 + nla_total_size(16) /* IFA_LOCAL */
4924 + nla_total_size(16) /* IFA_ADDRESS */
4925 + nla_total_size(sizeof(struct ifa_cacheinfo))
4926 + nla_total_size(4) /* IFA_FLAGS */
4927 + nla_total_size(4) /* IFA_RT_PRIORITY */;
4930 enum addr_type_t {
4931 UNICAST_ADDR,
4932 MULTICAST_ADDR,
4933 ANYCAST_ADDR,
4936 struct inet6_fill_args {
4937 u32 portid;
4938 u32 seq;
4939 int event;
4940 unsigned int flags;
4941 int netnsid;
4942 int ifindex;
4943 enum addr_type_t type;
4946 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
4947 struct inet6_fill_args *args)
4949 struct nlmsghdr *nlh;
4950 u32 preferred, valid;
4952 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
4953 sizeof(struct ifaddrmsg), args->flags);
4954 if (!nlh)
4955 return -EMSGSIZE;
4957 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
4958 ifa->idev->dev->ifindex);
4960 if (args->netnsid >= 0 &&
4961 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
4962 goto error;
4964 if (!((ifa->flags&IFA_F_PERMANENT) &&
4965 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
4966 preferred = ifa->prefered_lft;
4967 valid = ifa->valid_lft;
4968 if (preferred != INFINITY_LIFE_TIME) {
4969 long tval = (jiffies - ifa->tstamp)/HZ;
4970 if (preferred > tval)
4971 preferred -= tval;
4972 else
4973 preferred = 0;
4974 if (valid != INFINITY_LIFE_TIME) {
4975 if (valid > tval)
4976 valid -= tval;
4977 else
4978 valid = 0;
4981 } else {
4982 preferred = INFINITY_LIFE_TIME;
4983 valid = INFINITY_LIFE_TIME;
4986 if (!ipv6_addr_any(&ifa->peer_addr)) {
4987 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
4988 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
4989 goto error;
4990 } else
4991 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
4992 goto error;
4994 if (ifa->rt_priority &&
4995 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
4996 goto error;
4998 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4999 goto error;
5001 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
5002 goto error;
5004 nlmsg_end(skb, nlh);
5005 return 0;
5007 error:
5008 nlmsg_cancel(skb, nlh);
5009 return -EMSGSIZE;
5012 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
5013 struct inet6_fill_args *args)
5015 struct nlmsghdr *nlh;
5016 u8 scope = RT_SCOPE_UNIVERSE;
5017 int ifindex = ifmca->idev->dev->ifindex;
5019 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5020 scope = RT_SCOPE_SITE;
5022 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5023 sizeof(struct ifaddrmsg), args->flags);
5024 if (!nlh)
5025 return -EMSGSIZE;
5027 if (args->netnsid >= 0 &&
5028 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5029 return -EMSGSIZE;
5031 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5032 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5033 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
5034 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5035 nlmsg_cancel(skb, nlh);
5036 return -EMSGSIZE;
5039 nlmsg_end(skb, nlh);
5040 return 0;
5043 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
5044 struct inet6_fill_args *args)
5046 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5047 int ifindex = dev ? dev->ifindex : 1;
5048 struct nlmsghdr *nlh;
5049 u8 scope = RT_SCOPE_UNIVERSE;
5051 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5052 scope = RT_SCOPE_SITE;
5054 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5055 sizeof(struct ifaddrmsg), args->flags);
5056 if (!nlh)
5057 return -EMSGSIZE;
5059 if (args->netnsid >= 0 &&
5060 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5061 return -EMSGSIZE;
5063 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5064 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5065 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
5066 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5067 nlmsg_cancel(skb, nlh);
5068 return -EMSGSIZE;
5071 nlmsg_end(skb, nlh);
5072 return 0;
5075 /* called with rcu_read_lock() */
5076 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
5077 struct netlink_callback *cb, int s_ip_idx,
5078 struct inet6_fill_args *fillargs)
5080 struct ifmcaddr6 *ifmca;
5081 struct ifacaddr6 *ifaca;
5082 int ip_idx = 0;
5083 int err = 1;
5085 read_lock_bh(&idev->lock);
5086 switch (fillargs->type) {
5087 case UNICAST_ADDR: {
5088 struct inet6_ifaddr *ifa;
5089 fillargs->event = RTM_NEWADDR;
5091 /* unicast address incl. temp addr */
5092 list_for_each_entry(ifa, &idev->addr_list, if_list) {
5093 if (ip_idx < s_ip_idx)
5094 goto next;
5095 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5096 if (err < 0)
5097 break;
5098 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5099 next:
5100 ip_idx++;
5102 break;
5104 case MULTICAST_ADDR:
5105 fillargs->event = RTM_GETMULTICAST;
5107 /* multicast address */
5108 for (ifmca = idev->mc_list; ifmca;
5109 ifmca = ifmca->next, ip_idx++) {
5110 if (ip_idx < s_ip_idx)
5111 continue;
5112 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5113 if (err < 0)
5114 break;
5116 break;
5117 case ANYCAST_ADDR:
5118 fillargs->event = RTM_GETANYCAST;
5119 /* anycast address */
5120 for (ifaca = idev->ac_list; ifaca;
5121 ifaca = ifaca->aca_next, ip_idx++) {
5122 if (ip_idx < s_ip_idx)
5123 continue;
5124 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5125 if (err < 0)
5126 break;
5128 break;
5129 default:
5130 break;
5132 read_unlock_bh(&idev->lock);
5133 cb->args[2] = ip_idx;
5134 return err;
5137 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5138 struct inet6_fill_args *fillargs,
5139 struct net **tgt_net, struct sock *sk,
5140 struct netlink_callback *cb)
5142 struct netlink_ext_ack *extack = cb->extack;
5143 struct nlattr *tb[IFA_MAX+1];
5144 struct ifaddrmsg *ifm;
5145 int err, i;
5147 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5148 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5149 return -EINVAL;
5152 ifm = nlmsg_data(nlh);
5153 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5154 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5155 return -EINVAL;
5158 fillargs->ifindex = ifm->ifa_index;
5159 if (fillargs->ifindex) {
5160 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5161 fillargs->flags |= NLM_F_DUMP_FILTERED;
5164 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5165 ifa_ipv6_policy, extack);
5166 if (err < 0)
5167 return err;
5169 for (i = 0; i <= IFA_MAX; ++i) {
5170 if (!tb[i])
5171 continue;
5173 if (i == IFA_TARGET_NETNSID) {
5174 struct net *net;
5176 fillargs->netnsid = nla_get_s32(tb[i]);
5177 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5178 if (IS_ERR(net)) {
5179 fillargs->netnsid = -1;
5180 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5181 return PTR_ERR(net);
5183 *tgt_net = net;
5184 } else {
5185 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5186 return -EINVAL;
5190 return 0;
5193 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5194 enum addr_type_t type)
5196 const struct nlmsghdr *nlh = cb->nlh;
5197 struct inet6_fill_args fillargs = {
5198 .portid = NETLINK_CB(cb->skb).portid,
5199 .seq = cb->nlh->nlmsg_seq,
5200 .flags = NLM_F_MULTI,
5201 .netnsid = -1,
5202 .type = type,
5204 struct net *net = sock_net(skb->sk);
5205 struct net *tgt_net = net;
5206 int idx, s_idx, s_ip_idx;
5207 int h, s_h;
5208 struct net_device *dev;
5209 struct inet6_dev *idev;
5210 struct hlist_head *head;
5211 int err = 0;
5213 s_h = cb->args[0];
5214 s_idx = idx = cb->args[1];
5215 s_ip_idx = cb->args[2];
5217 if (cb->strict_check) {
5218 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5219 skb->sk, cb);
5220 if (err < 0)
5221 goto put_tgt_net;
5223 err = 0;
5224 if (fillargs.ifindex) {
5225 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
5226 if (!dev) {
5227 err = -ENODEV;
5228 goto put_tgt_net;
5230 idev = __in6_dev_get(dev);
5231 if (idev) {
5232 err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
5233 &fillargs);
5234 if (err > 0)
5235 err = 0;
5237 goto put_tgt_net;
5241 rcu_read_lock();
5242 cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
5243 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5244 idx = 0;
5245 head = &tgt_net->dev_index_head[h];
5246 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5247 if (idx < s_idx)
5248 goto cont;
5249 if (h > s_h || idx > s_idx)
5250 s_ip_idx = 0;
5251 idev = __in6_dev_get(dev);
5252 if (!idev)
5253 goto cont;
5255 if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
5256 &fillargs) < 0)
5257 goto done;
5258 cont:
5259 idx++;
5262 done:
5263 rcu_read_unlock();
5264 cb->args[0] = h;
5265 cb->args[1] = idx;
5266 put_tgt_net:
5267 if (fillargs.netnsid >= 0)
5268 put_net(tgt_net);
5270 return skb->len ? : err;
5273 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5275 enum addr_type_t type = UNICAST_ADDR;
5277 return inet6_dump_addr(skb, cb, type);
5280 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5282 enum addr_type_t type = MULTICAST_ADDR;
5284 return inet6_dump_addr(skb, cb, type);
5288 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5290 enum addr_type_t type = ANYCAST_ADDR;
5292 return inet6_dump_addr(skb, cb, type);
5295 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5296 const struct nlmsghdr *nlh,
5297 struct nlattr **tb,
5298 struct netlink_ext_ack *extack)
5300 struct ifaddrmsg *ifm;
5301 int i, err;
5303 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5304 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5305 return -EINVAL;
5308 if (!netlink_strict_get_check(skb))
5309 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5310 ifa_ipv6_policy, extack);
5312 ifm = nlmsg_data(nlh);
5313 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5314 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5315 return -EINVAL;
5318 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5319 ifa_ipv6_policy, extack);
5320 if (err)
5321 return err;
5323 for (i = 0; i <= IFA_MAX; i++) {
5324 if (!tb[i])
5325 continue;
5327 switch (i) {
5328 case IFA_TARGET_NETNSID:
5329 case IFA_ADDRESS:
5330 case IFA_LOCAL:
5331 break;
5332 default:
5333 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5334 return -EINVAL;
5338 return 0;
5341 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5342 struct netlink_ext_ack *extack)
5344 struct net *net = sock_net(in_skb->sk);
5345 struct inet6_fill_args fillargs = {
5346 .portid = NETLINK_CB(in_skb).portid,
5347 .seq = nlh->nlmsg_seq,
5348 .event = RTM_NEWADDR,
5349 .flags = 0,
5350 .netnsid = -1,
5352 struct net *tgt_net = net;
5353 struct ifaddrmsg *ifm;
5354 struct nlattr *tb[IFA_MAX+1];
5355 struct in6_addr *addr = NULL, *peer;
5356 struct net_device *dev = NULL;
5357 struct inet6_ifaddr *ifa;
5358 struct sk_buff *skb;
5359 int err;
5361 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5362 if (err < 0)
5363 return err;
5365 if (tb[IFA_TARGET_NETNSID]) {
5366 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5368 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5369 fillargs.netnsid);
5370 if (IS_ERR(tgt_net))
5371 return PTR_ERR(tgt_net);
5374 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5375 if (!addr)
5376 return -EINVAL;
5378 ifm = nlmsg_data(nlh);
5379 if (ifm->ifa_index)
5380 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5382 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5383 if (!ifa) {
5384 err = -EADDRNOTAVAIL;
5385 goto errout;
5388 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5389 if (!skb) {
5390 err = -ENOBUFS;
5391 goto errout_ifa;
5394 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5395 if (err < 0) {
5396 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5397 WARN_ON(err == -EMSGSIZE);
5398 kfree_skb(skb);
5399 goto errout_ifa;
5401 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5402 errout_ifa:
5403 in6_ifa_put(ifa);
5404 errout:
5405 if (dev)
5406 dev_put(dev);
5407 if (fillargs.netnsid >= 0)
5408 put_net(tgt_net);
5410 return err;
5413 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5415 struct sk_buff *skb;
5416 struct net *net = dev_net(ifa->idev->dev);
5417 struct inet6_fill_args fillargs = {
5418 .portid = 0,
5419 .seq = 0,
5420 .event = event,
5421 .flags = 0,
5422 .netnsid = -1,
5424 int err = -ENOBUFS;
5426 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5427 if (!skb)
5428 goto errout;
5430 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5431 if (err < 0) {
5432 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5433 WARN_ON(err == -EMSGSIZE);
5434 kfree_skb(skb);
5435 goto errout;
5437 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5438 return;
5439 errout:
5440 if (err < 0)
5441 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5444 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
5445 __s32 *array, int bytes)
5447 BUG_ON(bytes < (DEVCONF_MAX * 4));
5449 memset(array, 0, bytes);
5450 array[DEVCONF_FORWARDING] = cnf->forwarding;
5451 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
5452 array[DEVCONF_MTU6] = cnf->mtu6;
5453 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
5454 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
5455 array[DEVCONF_AUTOCONF] = cnf->autoconf;
5456 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
5457 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
5458 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5459 jiffies_to_msecs(cnf->rtr_solicit_interval);
5460 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5461 jiffies_to_msecs(cnf->rtr_solicit_max_interval);
5462 array[DEVCONF_RTR_SOLICIT_DELAY] =
5463 jiffies_to_msecs(cnf->rtr_solicit_delay);
5464 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
5465 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5466 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
5467 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5468 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
5469 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
5470 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
5471 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
5472 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
5473 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
5474 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
5475 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
5476 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
5477 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
5478 #ifdef CONFIG_IPV6_ROUTER_PREF
5479 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
5480 array[DEVCONF_RTR_PROBE_INTERVAL] =
5481 jiffies_to_msecs(cnf->rtr_probe_interval);
5482 #ifdef CONFIG_IPV6_ROUTE_INFO
5483 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
5484 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
5485 #endif
5486 #endif
5487 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
5488 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
5489 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5490 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
5491 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
5492 #endif
5493 #ifdef CONFIG_IPV6_MROUTE
5494 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
5495 #endif
5496 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
5497 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
5498 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
5499 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
5500 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
5501 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
5502 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
5503 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
5504 /* we omit DEVCONF_STABLE_SECRET for now */
5505 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
5506 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
5507 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
5508 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
5509 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
5510 #ifdef CONFIG_IPV6_SEG6_HMAC
5511 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
5512 #endif
5513 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
5514 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
5515 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
5516 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
5517 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
5520 static inline size_t inet6_ifla6_size(void)
5522 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5523 + nla_total_size(sizeof(struct ifla_cacheinfo))
5524 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5525 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5526 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5527 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5528 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5529 + 0;
5532 static inline size_t inet6_if_nlmsg_size(void)
5534 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5535 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5536 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5537 + nla_total_size(4) /* IFLA_MTU */
5538 + nla_total_size(4) /* IFLA_LINK */
5539 + nla_total_size(1) /* IFLA_OPERSTATE */
5540 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5543 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5544 int bytes)
5546 int i;
5547 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5548 BUG_ON(pad < 0);
5550 /* Use put_unaligned() because stats may not be aligned for u64. */
5551 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5552 for (i = 1; i < ICMP6_MIB_MAX; i++)
5553 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5555 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5558 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5559 int bytes, size_t syncpoff)
5561 int i, c;
5562 u64 buff[IPSTATS_MIB_MAX];
5563 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5565 BUG_ON(pad < 0);
5567 memset(buff, 0, sizeof(buff));
5568 buff[0] = IPSTATS_MIB_MAX;
5570 for_each_possible_cpu(c) {
5571 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5572 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5575 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5576 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5579 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5580 int bytes)
5582 switch (attrtype) {
5583 case IFLA_INET6_STATS:
5584 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5585 offsetof(struct ipstats_mib, syncp));
5586 break;
5587 case IFLA_INET6_ICMP6STATS:
5588 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5589 break;
5593 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5594 u32 ext_filter_mask)
5596 struct nlattr *nla;
5597 struct ifla_cacheinfo ci;
5599 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
5600 goto nla_put_failure;
5601 ci.max_reasm_len = IPV6_MAXPLEN;
5602 ci.tstamp = cstamp_delta(idev->tstamp);
5603 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5604 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5605 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5606 goto nla_put_failure;
5607 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5608 if (!nla)
5609 goto nla_put_failure;
5610 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5612 /* XXX - MC not implemented */
5614 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5615 return 0;
5617 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5618 if (!nla)
5619 goto nla_put_failure;
5620 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5622 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5623 if (!nla)
5624 goto nla_put_failure;
5625 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5627 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5628 if (!nla)
5629 goto nla_put_failure;
5630 read_lock_bh(&idev->lock);
5631 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5632 read_unlock_bh(&idev->lock);
5634 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
5635 goto nla_put_failure;
5637 return 0;
5639 nla_put_failure:
5640 return -EMSGSIZE;
5643 static size_t inet6_get_link_af_size(const struct net_device *dev,
5644 u32 ext_filter_mask)
5646 if (!__in6_dev_get(dev))
5647 return 0;
5649 return inet6_ifla6_size();
5652 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5653 u32 ext_filter_mask)
5655 struct inet6_dev *idev = __in6_dev_get(dev);
5657 if (!idev)
5658 return -ENODATA;
5660 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5661 return -EMSGSIZE;
5663 return 0;
5666 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
5668 struct inet6_ifaddr *ifp;
5669 struct net_device *dev = idev->dev;
5670 bool clear_token, update_rs = false;
5671 struct in6_addr ll_addr;
5673 ASSERT_RTNL();
5675 if (!token)
5676 return -EINVAL;
5677 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
5678 return -EINVAL;
5679 if (!ipv6_accept_ra(idev))
5680 return -EINVAL;
5681 if (idev->cnf.rtr_solicits == 0)
5682 return -EINVAL;
5684 write_lock_bh(&idev->lock);
5686 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5687 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5689 write_unlock_bh(&idev->lock);
5691 clear_token = ipv6_addr_any(token);
5692 if (clear_token)
5693 goto update_lft;
5695 if (!idev->dead && (idev->if_flags & IF_READY) &&
5696 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5697 IFA_F_OPTIMISTIC)) {
5698 /* If we're not ready, then normal ifup will take care
5699 * of this. Otherwise, we need to request our rs here.
5701 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5702 update_rs = true;
5705 update_lft:
5706 write_lock_bh(&idev->lock);
5708 if (update_rs) {
5709 idev->if_flags |= IF_RS_SENT;
5710 idev->rs_interval = rfc3315_s14_backoff_init(
5711 idev->cnf.rtr_solicit_interval);
5712 idev->rs_probes = 1;
5713 addrconf_mod_rs_timer(idev, idev->rs_interval);
5716 /* Well, that's kinda nasty ... */
5717 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5718 spin_lock(&ifp->lock);
5719 if (ifp->tokenized) {
5720 ifp->valid_lft = 0;
5721 ifp->prefered_lft = 0;
5723 spin_unlock(&ifp->lock);
5726 write_unlock_bh(&idev->lock);
5727 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5728 addrconf_verify_rtnl();
5729 return 0;
5732 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5733 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5734 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5737 static int check_addr_gen_mode(int mode)
5739 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5740 mode != IN6_ADDR_GEN_MODE_NONE &&
5741 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5742 mode != IN6_ADDR_GEN_MODE_RANDOM)
5743 return -EINVAL;
5744 return 1;
5747 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5748 int mode)
5750 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5751 !idev->cnf.stable_secret.initialized &&
5752 !net->ipv6.devconf_dflt->stable_secret.initialized)
5753 return -EINVAL;
5754 return 1;
5757 static int inet6_validate_link_af(const struct net_device *dev,
5758 const struct nlattr *nla)
5760 struct nlattr *tb[IFLA_INET6_MAX + 1];
5761 struct inet6_dev *idev = NULL;
5762 int err;
5764 if (dev) {
5765 idev = __in6_dev_get(dev);
5766 if (!idev)
5767 return -EAFNOSUPPORT;
5770 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5771 inet6_af_policy, NULL);
5772 if (err)
5773 return err;
5775 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
5776 return -EINVAL;
5778 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5779 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5781 if (check_addr_gen_mode(mode) < 0)
5782 return -EINVAL;
5783 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
5784 return -EINVAL;
5787 return 0;
5790 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
5792 struct inet6_dev *idev = __in6_dev_get(dev);
5793 struct nlattr *tb[IFLA_INET6_MAX + 1];
5794 int err;
5796 if (!idev)
5797 return -EAFNOSUPPORT;
5799 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
5800 BUG();
5802 if (tb[IFLA_INET6_TOKEN]) {
5803 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
5804 if (err)
5805 return err;
5808 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
5809 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
5811 idev->cnf.addr_gen_mode = mode;
5814 return 0;
5817 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
5818 u32 portid, u32 seq, int event, unsigned int flags)
5820 struct net_device *dev = idev->dev;
5821 struct ifinfomsg *hdr;
5822 struct nlmsghdr *nlh;
5823 void *protoinfo;
5825 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
5826 if (!nlh)
5827 return -EMSGSIZE;
5829 hdr = nlmsg_data(nlh);
5830 hdr->ifi_family = AF_INET6;
5831 hdr->__ifi_pad = 0;
5832 hdr->ifi_type = dev->type;
5833 hdr->ifi_index = dev->ifindex;
5834 hdr->ifi_flags = dev_get_flags(dev);
5835 hdr->ifi_change = 0;
5837 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
5838 (dev->addr_len &&
5839 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
5840 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
5841 (dev->ifindex != dev_get_iflink(dev) &&
5842 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
5843 nla_put_u8(skb, IFLA_OPERSTATE,
5844 netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
5845 goto nla_put_failure;
5846 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
5847 if (!protoinfo)
5848 goto nla_put_failure;
5850 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
5851 goto nla_put_failure;
5853 nla_nest_end(skb, protoinfo);
5854 nlmsg_end(skb, nlh);
5855 return 0;
5857 nla_put_failure:
5858 nlmsg_cancel(skb, nlh);
5859 return -EMSGSIZE;
5862 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
5863 struct netlink_ext_ack *extack)
5865 struct ifinfomsg *ifm;
5867 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5868 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
5869 return -EINVAL;
5872 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
5873 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
5874 return -EINVAL;
5877 ifm = nlmsg_data(nlh);
5878 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5879 ifm->ifi_change || ifm->ifi_index) {
5880 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
5881 return -EINVAL;
5884 return 0;
5887 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
5889 struct net *net = sock_net(skb->sk);
5890 int h, s_h;
5891 int idx = 0, s_idx;
5892 struct net_device *dev;
5893 struct inet6_dev *idev;
5894 struct hlist_head *head;
5896 /* only requests using strict checking can pass data to
5897 * influence the dump
5899 if (cb->strict_check) {
5900 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
5902 if (err < 0)
5903 return err;
5906 s_h = cb->args[0];
5907 s_idx = cb->args[1];
5909 rcu_read_lock();
5910 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5911 idx = 0;
5912 head = &net->dev_index_head[h];
5913 hlist_for_each_entry_rcu(dev, head, index_hlist) {
5914 if (idx < s_idx)
5915 goto cont;
5916 idev = __in6_dev_get(dev);
5917 if (!idev)
5918 goto cont;
5919 if (inet6_fill_ifinfo(skb, idev,
5920 NETLINK_CB(cb->skb).portid,
5921 cb->nlh->nlmsg_seq,
5922 RTM_NEWLINK, NLM_F_MULTI) < 0)
5923 goto out;
5924 cont:
5925 idx++;
5928 out:
5929 rcu_read_unlock();
5930 cb->args[1] = idx;
5931 cb->args[0] = h;
5933 return skb->len;
5936 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
5938 struct sk_buff *skb;
5939 struct net *net = dev_net(idev->dev);
5940 int err = -ENOBUFS;
5942 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
5943 if (!skb)
5944 goto errout;
5946 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
5947 if (err < 0) {
5948 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
5949 WARN_ON(err == -EMSGSIZE);
5950 kfree_skb(skb);
5951 goto errout;
5953 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
5954 return;
5955 errout:
5956 if (err < 0)
5957 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
5960 static inline size_t inet6_prefix_nlmsg_size(void)
5962 return NLMSG_ALIGN(sizeof(struct prefixmsg))
5963 + nla_total_size(sizeof(struct in6_addr))
5964 + nla_total_size(sizeof(struct prefix_cacheinfo));
5967 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
5968 struct prefix_info *pinfo, u32 portid, u32 seq,
5969 int event, unsigned int flags)
5971 struct prefixmsg *pmsg;
5972 struct nlmsghdr *nlh;
5973 struct prefix_cacheinfo ci;
5975 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
5976 if (!nlh)
5977 return -EMSGSIZE;
5979 pmsg = nlmsg_data(nlh);
5980 pmsg->prefix_family = AF_INET6;
5981 pmsg->prefix_pad1 = 0;
5982 pmsg->prefix_pad2 = 0;
5983 pmsg->prefix_ifindex = idev->dev->ifindex;
5984 pmsg->prefix_len = pinfo->prefix_len;
5985 pmsg->prefix_type = pinfo->type;
5986 pmsg->prefix_pad3 = 0;
5987 pmsg->prefix_flags = 0;
5988 if (pinfo->onlink)
5989 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
5990 if (pinfo->autoconf)
5991 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
5993 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
5994 goto nla_put_failure;
5995 ci.preferred_time = ntohl(pinfo->prefered);
5996 ci.valid_time = ntohl(pinfo->valid);
5997 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
5998 goto nla_put_failure;
5999 nlmsg_end(skb, nlh);
6000 return 0;
6002 nla_put_failure:
6003 nlmsg_cancel(skb, nlh);
6004 return -EMSGSIZE;
6007 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6008 struct prefix_info *pinfo)
6010 struct sk_buff *skb;
6011 struct net *net = dev_net(idev->dev);
6012 int err = -ENOBUFS;
6014 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6015 if (!skb)
6016 goto errout;
6018 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6019 if (err < 0) {
6020 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6021 WARN_ON(err == -EMSGSIZE);
6022 kfree_skb(skb);
6023 goto errout;
6025 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6026 return;
6027 errout:
6028 if (err < 0)
6029 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6032 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6034 struct net *net = dev_net(ifp->idev->dev);
6036 if (event)
6037 ASSERT_RTNL();
6039 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6041 switch (event) {
6042 case RTM_NEWADDR:
6044 * If the address was optimistic we inserted the route at the
6045 * start of our DAD process, so we don't need to do it again.
6046 * If the device was taken down in the middle of the DAD
6047 * cycle there is a race where we could get here without a
6048 * host route, so nothing to insert. That will be fixed when
6049 * the device is brought up.
6051 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6052 ip6_ins_rt(net, ifp->rt);
6053 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6054 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6055 &ifp->addr, ifp->idev->dev->name);
6058 if (ifp->idev->cnf.forwarding)
6059 addrconf_join_anycast(ifp);
6060 if (!ipv6_addr_any(&ifp->peer_addr))
6061 addrconf_prefix_route(&ifp->peer_addr, 128,
6062 ifp->rt_priority, ifp->idev->dev,
6063 0, 0, GFP_ATOMIC);
6064 break;
6065 case RTM_DELADDR:
6066 if (ifp->idev->cnf.forwarding)
6067 addrconf_leave_anycast(ifp);
6068 addrconf_leave_solict(ifp->idev, &ifp->addr);
6069 if (!ipv6_addr_any(&ifp->peer_addr)) {
6070 struct fib6_info *rt;
6072 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6073 ifp->idev->dev, 0, 0,
6074 false);
6075 if (rt)
6076 ip6_del_rt(net, rt, false);
6078 if (ifp->rt) {
6079 ip6_del_rt(net, ifp->rt, false);
6080 ifp->rt = NULL;
6082 rt_genid_bump_ipv6(net);
6083 break;
6085 atomic_inc(&net->ipv6.dev_addr_genid);
6088 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6090 rcu_read_lock_bh();
6091 if (likely(ifp->idev->dead == 0))
6092 __ipv6_ifa_notify(event, ifp);
6093 rcu_read_unlock_bh();
6096 #ifdef CONFIG_SYSCTL
6098 static
6099 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6100 void __user *buffer, size_t *lenp, loff_t *ppos)
6102 int *valp = ctl->data;
6103 int val = *valp;
6104 loff_t pos = *ppos;
6105 struct ctl_table lctl;
6106 int ret;
6109 * ctl->data points to idev->cnf.forwarding, we should
6110 * not modify it until we get the rtnl lock.
6112 lctl = *ctl;
6113 lctl.data = &val;
6115 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6117 if (write)
6118 ret = addrconf_fixup_forwarding(ctl, valp, val);
6119 if (ret)
6120 *ppos = pos;
6121 return ret;
6124 static
6125 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6126 void __user *buffer, size_t *lenp, loff_t *ppos)
6128 struct inet6_dev *idev = ctl->extra1;
6129 int min_mtu = IPV6_MIN_MTU;
6130 struct ctl_table lctl;
6132 lctl = *ctl;
6133 lctl.extra1 = &min_mtu;
6134 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6136 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6139 static void dev_disable_change(struct inet6_dev *idev)
6141 struct netdev_notifier_info info;
6143 if (!idev || !idev->dev)
6144 return;
6146 netdev_notifier_info_init(&info, idev->dev);
6147 if (idev->cnf.disable_ipv6)
6148 addrconf_notify(NULL, NETDEV_DOWN, &info);
6149 else
6150 addrconf_notify(NULL, NETDEV_UP, &info);
6153 static void addrconf_disable_change(struct net *net, __s32 newf)
6155 struct net_device *dev;
6156 struct inet6_dev *idev;
6158 for_each_netdev(net, dev) {
6159 idev = __in6_dev_get(dev);
6160 if (idev) {
6161 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6162 idev->cnf.disable_ipv6 = newf;
6163 if (changed)
6164 dev_disable_change(idev);
6169 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6171 struct net *net;
6172 int old;
6174 if (!rtnl_trylock())
6175 return restart_syscall();
6177 net = (struct net *)table->extra2;
6178 old = *p;
6179 *p = newf;
6181 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6182 rtnl_unlock();
6183 return 0;
6186 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6187 net->ipv6.devconf_dflt->disable_ipv6 = newf;
6188 addrconf_disable_change(net, newf);
6189 } else if ((!newf) ^ (!old))
6190 dev_disable_change((struct inet6_dev *)table->extra1);
6192 rtnl_unlock();
6193 return 0;
6196 static
6197 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6198 void __user *buffer, size_t *lenp, loff_t *ppos)
6200 int *valp = ctl->data;
6201 int val = *valp;
6202 loff_t pos = *ppos;
6203 struct ctl_table lctl;
6204 int ret;
6207 * ctl->data points to idev->cnf.disable_ipv6, we should
6208 * not modify it until we get the rtnl lock.
6210 lctl = *ctl;
6211 lctl.data = &val;
6213 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6215 if (write)
6216 ret = addrconf_disable_ipv6(ctl, valp, val);
6217 if (ret)
6218 *ppos = pos;
6219 return ret;
6222 static
6223 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6224 void __user *buffer, size_t *lenp, loff_t *ppos)
6226 int *valp = ctl->data;
6227 int ret;
6228 int old, new;
6230 old = *valp;
6231 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6232 new = *valp;
6234 if (write && old != new) {
6235 struct net *net = ctl->extra2;
6237 if (!rtnl_trylock())
6238 return restart_syscall();
6240 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6241 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6242 NETCONFA_PROXY_NEIGH,
6243 NETCONFA_IFINDEX_DEFAULT,
6244 net->ipv6.devconf_dflt);
6245 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6246 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6247 NETCONFA_PROXY_NEIGH,
6248 NETCONFA_IFINDEX_ALL,
6249 net->ipv6.devconf_all);
6250 else {
6251 struct inet6_dev *idev = ctl->extra1;
6253 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6254 NETCONFA_PROXY_NEIGH,
6255 idev->dev->ifindex,
6256 &idev->cnf);
6258 rtnl_unlock();
6261 return ret;
6264 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6265 void __user *buffer, size_t *lenp,
6266 loff_t *ppos)
6268 int ret = 0;
6269 u32 new_val;
6270 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6271 struct net *net = (struct net *)ctl->extra2;
6272 struct ctl_table tmp = {
6273 .data = &new_val,
6274 .maxlen = sizeof(new_val),
6275 .mode = ctl->mode,
6278 if (!rtnl_trylock())
6279 return restart_syscall();
6281 new_val = *((u32 *)ctl->data);
6283 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6284 if (ret != 0)
6285 goto out;
6287 if (write) {
6288 if (check_addr_gen_mode(new_val) < 0) {
6289 ret = -EINVAL;
6290 goto out;
6293 if (idev) {
6294 if (check_stable_privacy(idev, net, new_val) < 0) {
6295 ret = -EINVAL;
6296 goto out;
6299 if (idev->cnf.addr_gen_mode != new_val) {
6300 idev->cnf.addr_gen_mode = new_val;
6301 addrconf_dev_config(idev->dev);
6303 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6304 struct net_device *dev;
6306 net->ipv6.devconf_dflt->addr_gen_mode = new_val;
6307 for_each_netdev(net, dev) {
6308 idev = __in6_dev_get(dev);
6309 if (idev &&
6310 idev->cnf.addr_gen_mode != new_val) {
6311 idev->cnf.addr_gen_mode = new_val;
6312 addrconf_dev_config(idev->dev);
6317 *((u32 *)ctl->data) = new_val;
6320 out:
6321 rtnl_unlock();
6323 return ret;
6326 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6327 void __user *buffer, size_t *lenp,
6328 loff_t *ppos)
6330 int err;
6331 struct in6_addr addr;
6332 char str[IPV6_MAX_STRLEN];
6333 struct ctl_table lctl = *ctl;
6334 struct net *net = ctl->extra2;
6335 struct ipv6_stable_secret *secret = ctl->data;
6337 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6338 return -EIO;
6340 lctl.maxlen = IPV6_MAX_STRLEN;
6341 lctl.data = str;
6343 if (!rtnl_trylock())
6344 return restart_syscall();
6346 if (!write && !secret->initialized) {
6347 err = -EIO;
6348 goto out;
6351 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6352 if (err >= sizeof(str)) {
6353 err = -EIO;
6354 goto out;
6357 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6358 if (err || !write)
6359 goto out;
6361 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6362 err = -EIO;
6363 goto out;
6366 secret->initialized = true;
6367 secret->secret = addr;
6369 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6370 struct net_device *dev;
6372 for_each_netdev(net, dev) {
6373 struct inet6_dev *idev = __in6_dev_get(dev);
6375 if (idev) {
6376 idev->cnf.addr_gen_mode =
6377 IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6380 } else {
6381 struct inet6_dev *idev = ctl->extra1;
6383 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
6386 out:
6387 rtnl_unlock();
6389 return err;
6392 static
6393 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6394 int write,
6395 void __user *buffer,
6396 size_t *lenp,
6397 loff_t *ppos)
6399 int *valp = ctl->data;
6400 int val = *valp;
6401 loff_t pos = *ppos;
6402 struct ctl_table lctl;
6403 int ret;
6405 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6406 * we should not modify it until we get the rtnl lock.
6408 lctl = *ctl;
6409 lctl.data = &val;
6411 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6413 if (write)
6414 ret = addrconf_fixup_linkdown(ctl, valp, val);
6415 if (ret)
6416 *ppos = pos;
6417 return ret;
6420 static
6421 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6423 if (rt) {
6424 if (action)
6425 rt->dst.flags |= DST_NOPOLICY;
6426 else
6427 rt->dst.flags &= ~DST_NOPOLICY;
6431 static
6432 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6434 struct inet6_ifaddr *ifa;
6436 read_lock_bh(&idev->lock);
6437 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6438 spin_lock(&ifa->lock);
6439 if (ifa->rt) {
6440 /* host routes only use builtin fib6_nh */
6441 struct fib6_nh *nh = ifa->rt->fib6_nh;
6442 int cpu;
6444 rcu_read_lock();
6445 ifa->rt->dst_nopolicy = val ? true : false;
6446 if (nh->rt6i_pcpu) {
6447 for_each_possible_cpu(cpu) {
6448 struct rt6_info **rtp;
6450 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6451 addrconf_set_nopolicy(*rtp, val);
6454 rcu_read_unlock();
6456 spin_unlock(&ifa->lock);
6458 read_unlock_bh(&idev->lock);
6461 static
6462 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6464 struct inet6_dev *idev;
6465 struct net *net;
6467 if (!rtnl_trylock())
6468 return restart_syscall();
6470 *valp = val;
6472 net = (struct net *)ctl->extra2;
6473 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6474 rtnl_unlock();
6475 return 0;
6478 if (valp == &net->ipv6.devconf_all->disable_policy) {
6479 struct net_device *dev;
6481 for_each_netdev(net, dev) {
6482 idev = __in6_dev_get(dev);
6483 if (idev)
6484 addrconf_disable_policy_idev(idev, val);
6486 } else {
6487 idev = (struct inet6_dev *)ctl->extra1;
6488 addrconf_disable_policy_idev(idev, val);
6491 rtnl_unlock();
6492 return 0;
6495 static
6496 int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6497 void __user *buffer, size_t *lenp,
6498 loff_t *ppos)
6500 int *valp = ctl->data;
6501 int val = *valp;
6502 loff_t pos = *ppos;
6503 struct ctl_table lctl;
6504 int ret;
6506 lctl = *ctl;
6507 lctl.data = &val;
6508 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6510 if (write && (*valp != val))
6511 ret = addrconf_disable_policy(ctl, valp, val);
6513 if (ret)
6514 *ppos = pos;
6516 return ret;
6519 static int minus_one = -1;
6520 static const int two_five_five = 255;
6522 static const struct ctl_table addrconf_sysctl[] = {
6524 .procname = "forwarding",
6525 .data = &ipv6_devconf.forwarding,
6526 .maxlen = sizeof(int),
6527 .mode = 0644,
6528 .proc_handler = addrconf_sysctl_forward,
6531 .procname = "hop_limit",
6532 .data = &ipv6_devconf.hop_limit,
6533 .maxlen = sizeof(int),
6534 .mode = 0644,
6535 .proc_handler = proc_dointvec_minmax,
6536 .extra1 = (void *)SYSCTL_ONE,
6537 .extra2 = (void *)&two_five_five,
6540 .procname = "mtu",
6541 .data = &ipv6_devconf.mtu6,
6542 .maxlen = sizeof(int),
6543 .mode = 0644,
6544 .proc_handler = addrconf_sysctl_mtu,
6547 .procname = "accept_ra",
6548 .data = &ipv6_devconf.accept_ra,
6549 .maxlen = sizeof(int),
6550 .mode = 0644,
6551 .proc_handler = proc_dointvec,
6554 .procname = "accept_redirects",
6555 .data = &ipv6_devconf.accept_redirects,
6556 .maxlen = sizeof(int),
6557 .mode = 0644,
6558 .proc_handler = proc_dointvec,
6561 .procname = "autoconf",
6562 .data = &ipv6_devconf.autoconf,
6563 .maxlen = sizeof(int),
6564 .mode = 0644,
6565 .proc_handler = proc_dointvec,
6568 .procname = "dad_transmits",
6569 .data = &ipv6_devconf.dad_transmits,
6570 .maxlen = sizeof(int),
6571 .mode = 0644,
6572 .proc_handler = proc_dointvec,
6575 .procname = "router_solicitations",
6576 .data = &ipv6_devconf.rtr_solicits,
6577 .maxlen = sizeof(int),
6578 .mode = 0644,
6579 .proc_handler = proc_dointvec_minmax,
6580 .extra1 = &minus_one,
6583 .procname = "router_solicitation_interval",
6584 .data = &ipv6_devconf.rtr_solicit_interval,
6585 .maxlen = sizeof(int),
6586 .mode = 0644,
6587 .proc_handler = proc_dointvec_jiffies,
6590 .procname = "router_solicitation_max_interval",
6591 .data = &ipv6_devconf.rtr_solicit_max_interval,
6592 .maxlen = sizeof(int),
6593 .mode = 0644,
6594 .proc_handler = proc_dointvec_jiffies,
6597 .procname = "router_solicitation_delay",
6598 .data = &ipv6_devconf.rtr_solicit_delay,
6599 .maxlen = sizeof(int),
6600 .mode = 0644,
6601 .proc_handler = proc_dointvec_jiffies,
6604 .procname = "force_mld_version",
6605 .data = &ipv6_devconf.force_mld_version,
6606 .maxlen = sizeof(int),
6607 .mode = 0644,
6608 .proc_handler = proc_dointvec,
6611 .procname = "mldv1_unsolicited_report_interval",
6612 .data =
6613 &ipv6_devconf.mldv1_unsolicited_report_interval,
6614 .maxlen = sizeof(int),
6615 .mode = 0644,
6616 .proc_handler = proc_dointvec_ms_jiffies,
6619 .procname = "mldv2_unsolicited_report_interval",
6620 .data =
6621 &ipv6_devconf.mldv2_unsolicited_report_interval,
6622 .maxlen = sizeof(int),
6623 .mode = 0644,
6624 .proc_handler = proc_dointvec_ms_jiffies,
6627 .procname = "use_tempaddr",
6628 .data = &ipv6_devconf.use_tempaddr,
6629 .maxlen = sizeof(int),
6630 .mode = 0644,
6631 .proc_handler = proc_dointvec,
6634 .procname = "temp_valid_lft",
6635 .data = &ipv6_devconf.temp_valid_lft,
6636 .maxlen = sizeof(int),
6637 .mode = 0644,
6638 .proc_handler = proc_dointvec,
6641 .procname = "temp_prefered_lft",
6642 .data = &ipv6_devconf.temp_prefered_lft,
6643 .maxlen = sizeof(int),
6644 .mode = 0644,
6645 .proc_handler = proc_dointvec,
6648 .procname = "regen_max_retry",
6649 .data = &ipv6_devconf.regen_max_retry,
6650 .maxlen = sizeof(int),
6651 .mode = 0644,
6652 .proc_handler = proc_dointvec,
6655 .procname = "max_desync_factor",
6656 .data = &ipv6_devconf.max_desync_factor,
6657 .maxlen = sizeof(int),
6658 .mode = 0644,
6659 .proc_handler = proc_dointvec,
6662 .procname = "max_addresses",
6663 .data = &ipv6_devconf.max_addresses,
6664 .maxlen = sizeof(int),
6665 .mode = 0644,
6666 .proc_handler = proc_dointvec,
6669 .procname = "accept_ra_defrtr",
6670 .data = &ipv6_devconf.accept_ra_defrtr,
6671 .maxlen = sizeof(int),
6672 .mode = 0644,
6673 .proc_handler = proc_dointvec,
6676 .procname = "accept_ra_min_hop_limit",
6677 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6678 .maxlen = sizeof(int),
6679 .mode = 0644,
6680 .proc_handler = proc_dointvec,
6683 .procname = "accept_ra_pinfo",
6684 .data = &ipv6_devconf.accept_ra_pinfo,
6685 .maxlen = sizeof(int),
6686 .mode = 0644,
6687 .proc_handler = proc_dointvec,
6689 #ifdef CONFIG_IPV6_ROUTER_PREF
6691 .procname = "accept_ra_rtr_pref",
6692 .data = &ipv6_devconf.accept_ra_rtr_pref,
6693 .maxlen = sizeof(int),
6694 .mode = 0644,
6695 .proc_handler = proc_dointvec,
6698 .procname = "router_probe_interval",
6699 .data = &ipv6_devconf.rtr_probe_interval,
6700 .maxlen = sizeof(int),
6701 .mode = 0644,
6702 .proc_handler = proc_dointvec_jiffies,
6704 #ifdef CONFIG_IPV6_ROUTE_INFO
6706 .procname = "accept_ra_rt_info_min_plen",
6707 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6708 .maxlen = sizeof(int),
6709 .mode = 0644,
6710 .proc_handler = proc_dointvec,
6713 .procname = "accept_ra_rt_info_max_plen",
6714 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6715 .maxlen = sizeof(int),
6716 .mode = 0644,
6717 .proc_handler = proc_dointvec,
6719 #endif
6720 #endif
6722 .procname = "proxy_ndp",
6723 .data = &ipv6_devconf.proxy_ndp,
6724 .maxlen = sizeof(int),
6725 .mode = 0644,
6726 .proc_handler = addrconf_sysctl_proxy_ndp,
6729 .procname = "accept_source_route",
6730 .data = &ipv6_devconf.accept_source_route,
6731 .maxlen = sizeof(int),
6732 .mode = 0644,
6733 .proc_handler = proc_dointvec,
6735 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6737 .procname = "optimistic_dad",
6738 .data = &ipv6_devconf.optimistic_dad,
6739 .maxlen = sizeof(int),
6740 .mode = 0644,
6741 .proc_handler = proc_dointvec,
6744 .procname = "use_optimistic",
6745 .data = &ipv6_devconf.use_optimistic,
6746 .maxlen = sizeof(int),
6747 .mode = 0644,
6748 .proc_handler = proc_dointvec,
6750 #endif
6751 #ifdef CONFIG_IPV6_MROUTE
6753 .procname = "mc_forwarding",
6754 .data = &ipv6_devconf.mc_forwarding,
6755 .maxlen = sizeof(int),
6756 .mode = 0444,
6757 .proc_handler = proc_dointvec,
6759 #endif
6761 .procname = "disable_ipv6",
6762 .data = &ipv6_devconf.disable_ipv6,
6763 .maxlen = sizeof(int),
6764 .mode = 0644,
6765 .proc_handler = addrconf_sysctl_disable,
6768 .procname = "accept_dad",
6769 .data = &ipv6_devconf.accept_dad,
6770 .maxlen = sizeof(int),
6771 .mode = 0644,
6772 .proc_handler = proc_dointvec,
6775 .procname = "force_tllao",
6776 .data = &ipv6_devconf.force_tllao,
6777 .maxlen = sizeof(int),
6778 .mode = 0644,
6779 .proc_handler = proc_dointvec
6782 .procname = "ndisc_notify",
6783 .data = &ipv6_devconf.ndisc_notify,
6784 .maxlen = sizeof(int),
6785 .mode = 0644,
6786 .proc_handler = proc_dointvec
6789 .procname = "suppress_frag_ndisc",
6790 .data = &ipv6_devconf.suppress_frag_ndisc,
6791 .maxlen = sizeof(int),
6792 .mode = 0644,
6793 .proc_handler = proc_dointvec
6796 .procname = "accept_ra_from_local",
6797 .data = &ipv6_devconf.accept_ra_from_local,
6798 .maxlen = sizeof(int),
6799 .mode = 0644,
6800 .proc_handler = proc_dointvec,
6803 .procname = "accept_ra_mtu",
6804 .data = &ipv6_devconf.accept_ra_mtu,
6805 .maxlen = sizeof(int),
6806 .mode = 0644,
6807 .proc_handler = proc_dointvec,
6810 .procname = "stable_secret",
6811 .data = &ipv6_devconf.stable_secret,
6812 .maxlen = IPV6_MAX_STRLEN,
6813 .mode = 0600,
6814 .proc_handler = addrconf_sysctl_stable_secret,
6817 .procname = "use_oif_addrs_only",
6818 .data = &ipv6_devconf.use_oif_addrs_only,
6819 .maxlen = sizeof(int),
6820 .mode = 0644,
6821 .proc_handler = proc_dointvec,
6824 .procname = "ignore_routes_with_linkdown",
6825 .data = &ipv6_devconf.ignore_routes_with_linkdown,
6826 .maxlen = sizeof(int),
6827 .mode = 0644,
6828 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
6831 .procname = "drop_unicast_in_l2_multicast",
6832 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
6833 .maxlen = sizeof(int),
6834 .mode = 0644,
6835 .proc_handler = proc_dointvec,
6838 .procname = "drop_unsolicited_na",
6839 .data = &ipv6_devconf.drop_unsolicited_na,
6840 .maxlen = sizeof(int),
6841 .mode = 0644,
6842 .proc_handler = proc_dointvec,
6845 .procname = "keep_addr_on_down",
6846 .data = &ipv6_devconf.keep_addr_on_down,
6847 .maxlen = sizeof(int),
6848 .mode = 0644,
6849 .proc_handler = proc_dointvec,
6853 .procname = "seg6_enabled",
6854 .data = &ipv6_devconf.seg6_enabled,
6855 .maxlen = sizeof(int),
6856 .mode = 0644,
6857 .proc_handler = proc_dointvec,
6859 #ifdef CONFIG_IPV6_SEG6_HMAC
6861 .procname = "seg6_require_hmac",
6862 .data = &ipv6_devconf.seg6_require_hmac,
6863 .maxlen = sizeof(int),
6864 .mode = 0644,
6865 .proc_handler = proc_dointvec,
6867 #endif
6869 .procname = "enhanced_dad",
6870 .data = &ipv6_devconf.enhanced_dad,
6871 .maxlen = sizeof(int),
6872 .mode = 0644,
6873 .proc_handler = proc_dointvec,
6876 .procname = "addr_gen_mode",
6877 .data = &ipv6_devconf.addr_gen_mode,
6878 .maxlen = sizeof(int),
6879 .mode = 0644,
6880 .proc_handler = addrconf_sysctl_addr_gen_mode,
6883 .procname = "disable_policy",
6884 .data = &ipv6_devconf.disable_policy,
6885 .maxlen = sizeof(int),
6886 .mode = 0644,
6887 .proc_handler = addrconf_sysctl_disable_policy,
6890 .procname = "ndisc_tclass",
6891 .data = &ipv6_devconf.ndisc_tclass,
6892 .maxlen = sizeof(int),
6893 .mode = 0644,
6894 .proc_handler = proc_dointvec_minmax,
6895 .extra1 = (void *)SYSCTL_ZERO,
6896 .extra2 = (void *)&two_five_five,
6899 .procname = "rpl_seg_enabled",
6900 .data = &ipv6_devconf.rpl_seg_enabled,
6901 .maxlen = sizeof(int),
6902 .mode = 0644,
6903 .proc_handler = proc_dointvec,
6906 /* sentinel */
6910 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
6911 struct inet6_dev *idev, struct ipv6_devconf *p)
6913 int i, ifindex;
6914 struct ctl_table *table;
6915 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
6917 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL);
6918 if (!table)
6919 goto out;
6921 for (i = 0; table[i].data; i++) {
6922 table[i].data += (char *)p - (char *)&ipv6_devconf;
6923 /* If one of these is already set, then it is not safe to
6924 * overwrite either of them: this makes proc_dointvec_minmax
6925 * usable.
6927 if (!table[i].extra1 && !table[i].extra2) {
6928 table[i].extra1 = idev; /* embedded; no ref */
6929 table[i].extra2 = net;
6933 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
6935 p->sysctl_header = register_net_sysctl(net, path, table);
6936 if (!p->sysctl_header)
6937 goto free;
6939 if (!strcmp(dev_name, "all"))
6940 ifindex = NETCONFA_IFINDEX_ALL;
6941 else if (!strcmp(dev_name, "default"))
6942 ifindex = NETCONFA_IFINDEX_DEFAULT;
6943 else
6944 ifindex = idev->dev->ifindex;
6945 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
6946 ifindex, p);
6947 return 0;
6949 free:
6950 kfree(table);
6951 out:
6952 return -ENOBUFS;
6955 static void __addrconf_sysctl_unregister(struct net *net,
6956 struct ipv6_devconf *p, int ifindex)
6958 struct ctl_table *table;
6960 if (!p->sysctl_header)
6961 return;
6963 table = p->sysctl_header->ctl_table_arg;
6964 unregister_net_sysctl_table(p->sysctl_header);
6965 p->sysctl_header = NULL;
6966 kfree(table);
6968 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
6971 static int addrconf_sysctl_register(struct inet6_dev *idev)
6973 int err;
6975 if (!sysctl_dev_name_is_allowed(idev->dev->name))
6976 return -EINVAL;
6978 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
6979 &ndisc_ifinfo_sysctl_change);
6980 if (err)
6981 return err;
6982 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
6983 idev, &idev->cnf);
6984 if (err)
6985 neigh_sysctl_unregister(idev->nd_parms);
6987 return err;
6990 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
6992 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
6993 idev->dev->ifindex);
6994 neigh_sysctl_unregister(idev->nd_parms);
6998 #endif
7000 static int __net_init addrconf_init_net(struct net *net)
7002 int err = -ENOMEM;
7003 struct ipv6_devconf *all, *dflt;
7005 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7006 if (!all)
7007 goto err_alloc_all;
7009 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7010 if (!dflt)
7011 goto err_alloc_dflt;
7013 if (IS_ENABLED(CONFIG_SYSCTL) &&
7014 sysctl_devconf_inherit_init_net == 1 && !net_eq(net, &init_net)) {
7015 memcpy(all, init_net.ipv6.devconf_all, sizeof(ipv6_devconf));
7016 memcpy(dflt, init_net.ipv6.devconf_dflt, sizeof(ipv6_devconf_dflt));
7019 /* these will be inherited by all namespaces */
7020 dflt->autoconf = ipv6_defaults.autoconf;
7021 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7023 dflt->stable_secret.initialized = false;
7024 all->stable_secret.initialized = false;
7026 net->ipv6.devconf_all = all;
7027 net->ipv6.devconf_dflt = dflt;
7029 #ifdef CONFIG_SYSCTL
7030 err = __addrconf_sysctl_register(net, "all", NULL, all);
7031 if (err < 0)
7032 goto err_reg_all;
7034 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7035 if (err < 0)
7036 goto err_reg_dflt;
7037 #endif
7038 return 0;
7040 #ifdef CONFIG_SYSCTL
7041 err_reg_dflt:
7042 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7043 err_reg_all:
7044 kfree(dflt);
7045 #endif
7046 err_alloc_dflt:
7047 kfree(all);
7048 err_alloc_all:
7049 return err;
7052 static void __net_exit addrconf_exit_net(struct net *net)
7054 #ifdef CONFIG_SYSCTL
7055 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7056 NETCONFA_IFINDEX_DEFAULT);
7057 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7058 NETCONFA_IFINDEX_ALL);
7059 #endif
7060 kfree(net->ipv6.devconf_dflt);
7061 kfree(net->ipv6.devconf_all);
7064 static struct pernet_operations addrconf_ops = {
7065 .init = addrconf_init_net,
7066 .exit = addrconf_exit_net,
7069 static struct rtnl_af_ops inet6_ops __read_mostly = {
7070 .family = AF_INET6,
7071 .fill_link_af = inet6_fill_link_af,
7072 .get_link_af_size = inet6_get_link_af_size,
7073 .validate_link_af = inet6_validate_link_af,
7074 .set_link_af = inet6_set_link_af,
7078 * Init / cleanup code
7081 int __init addrconf_init(void)
7083 struct inet6_dev *idev;
7084 int i, err;
7086 err = ipv6_addr_label_init();
7087 if (err < 0) {
7088 pr_crit("%s: cannot initialize default policy table: %d\n",
7089 __func__, err);
7090 goto out;
7093 err = register_pernet_subsys(&addrconf_ops);
7094 if (err < 0)
7095 goto out_addrlabel;
7097 addrconf_wq = create_workqueue("ipv6_addrconf");
7098 if (!addrconf_wq) {
7099 err = -ENOMEM;
7100 goto out_nowq;
7103 /* The addrconf netdev notifier requires that loopback_dev
7104 * has it's ipv6 private information allocated and setup
7105 * before it can bring up and give link-local addresses
7106 * to other devices which are up.
7108 * Unfortunately, loopback_dev is not necessarily the first
7109 * entry in the global dev_base list of net devices. In fact,
7110 * it is likely to be the very last entry on that list.
7111 * So this causes the notifier registry below to try and
7112 * give link-local addresses to all devices besides loopback_dev
7113 * first, then loopback_dev, which cases all the non-loopback_dev
7114 * devices to fail to get a link-local address.
7116 * So, as a temporary fix, allocate the ipv6 structure for
7117 * loopback_dev first by hand.
7118 * Longer term, all of the dependencies ipv6 has upon the loopback
7119 * device and it being up should be removed.
7121 rtnl_lock();
7122 idev = ipv6_add_dev(init_net.loopback_dev);
7123 rtnl_unlock();
7124 if (IS_ERR(idev)) {
7125 err = PTR_ERR(idev);
7126 goto errlo;
7129 ip6_route_init_special_entries();
7131 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7132 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
7134 register_netdevice_notifier(&ipv6_dev_notf);
7136 addrconf_verify();
7138 rtnl_af_register(&inet6_ops);
7140 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7141 NULL, inet6_dump_ifinfo, 0);
7142 if (err < 0)
7143 goto errout;
7145 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7146 inet6_rtm_newaddr, NULL, 0);
7147 if (err < 0)
7148 goto errout;
7149 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7150 inet6_rtm_deladdr, NULL, 0);
7151 if (err < 0)
7152 goto errout;
7153 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7154 inet6_rtm_getaddr, inet6_dump_ifaddr,
7155 RTNL_FLAG_DOIT_UNLOCKED);
7156 if (err < 0)
7157 goto errout;
7158 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7159 NULL, inet6_dump_ifmcaddr, 0);
7160 if (err < 0)
7161 goto errout;
7162 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7163 NULL, inet6_dump_ifacaddr, 0);
7164 if (err < 0)
7165 goto errout;
7166 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7167 inet6_netconf_get_devconf,
7168 inet6_netconf_dump_devconf,
7169 RTNL_FLAG_DOIT_UNLOCKED);
7170 if (err < 0)
7171 goto errout;
7172 err = ipv6_addr_label_rtnl_register();
7173 if (err < 0)
7174 goto errout;
7176 return 0;
7177 errout:
7178 rtnl_unregister_all(PF_INET6);
7179 rtnl_af_unregister(&inet6_ops);
7180 unregister_netdevice_notifier(&ipv6_dev_notf);
7181 errlo:
7182 destroy_workqueue(addrconf_wq);
7183 out_nowq:
7184 unregister_pernet_subsys(&addrconf_ops);
7185 out_addrlabel:
7186 ipv6_addr_label_cleanup();
7187 out:
7188 return err;
7191 void addrconf_cleanup(void)
7193 struct net_device *dev;
7194 int i;
7196 unregister_netdevice_notifier(&ipv6_dev_notf);
7197 unregister_pernet_subsys(&addrconf_ops);
7198 ipv6_addr_label_cleanup();
7200 rtnl_af_unregister(&inet6_ops);
7202 rtnl_lock();
7204 /* clean dev list */
7205 for_each_netdev(&init_net, dev) {
7206 if (__in6_dev_get(dev) == NULL)
7207 continue;
7208 addrconf_ifdown(dev, 1);
7210 addrconf_ifdown(init_net.loopback_dev, 2);
7213 * Check hash table.
7215 spin_lock_bh(&addrconf_hash_lock);
7216 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7217 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
7218 spin_unlock_bh(&addrconf_hash_lock);
7219 cancel_delayed_work(&addr_chk_work);
7220 rtnl_unlock();
7222 destroy_workqueue(addrconf_wq);