1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPv6 Address [auto]configuration
4 * Linux INET6 implementation
14 * Janos Farkas : delete timer on ifdown
16 * Andi Kleen : kill double kfree on module
18 * Maciej W. Rozycki : FDDI support
19 * sekiya@USAGI : Don't send too many RS
21 * yoshfuji@USAGI : Fixed interval between DAD
23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
24 * address validation timer.
25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
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
32 * YOSHIFUJI Hideaki @USAGI : improved source address
33 * selection; consider scope,
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>
58 #include <linux/sysctl.h>
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>
70 #include <net/6lowpan.h>
71 #include <net/firewire.h>
73 #include <net/protocol.h>
74 #include <net/ndisc.h>
75 #include <net/ip6_route.h>
76 #include <net/addrconf.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>
92 #include <linux/ioam6.h>
94 #define INFINITY_LIFE_TIME 0xFFFFFFFF
96 #define IPV6_MAX_STRLEN \
97 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
99 static inline u32 cstamp_delta(unsigned long cstamp)
101 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
104 static inline s32 rfc3315_s14_backoff_init(s32 irt)
106 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */
107 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt;
108 do_div(tmp, 1000000);
112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
114 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */
115 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt;
116 do_div(tmp, 1000000);
117 if ((s32)tmp > mrt) {
118 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
119 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt;
120 do_div(tmp, 1000000);
126 static int addrconf_sysctl_register(struct inet6_dev *idev);
127 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
129 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
139 static void ipv6_gen_rnd_iid(struct in6_addr *addr);
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,
145 const struct inet6_dev *idev);
147 #define IN6_ADDR_HSIZE_SHIFT 8
148 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT)
150 static void addrconf_verify(struct net *net);
151 static void addrconf_verify_rtnl(struct net *net);
153 static struct workqueue_struct *addrconf_wq;
155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
158 static void addrconf_type_change(struct net_device *dev,
159 unsigned long event);
160 static int addrconf_ifdown(struct net_device *dev, bool unregister);
162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
164 const struct net_device *dev,
165 u32 flags, u32 noflags,
168 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
169 static void addrconf_dad_work(struct work_struct *w);
170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
173 static void addrconf_rs_timer(struct timer_list *t);
174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
177 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
178 struct prefix_info *pinfo);
180 static struct ipv6_devconf ipv6_devconf __read_mostly = {
182 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
183 .mtu6 = IPV6_MIN_MTU,
185 .accept_redirects = 1,
187 .force_mld_version = 0,
188 .mldv1_unsolicited_report_interval = 10 * HZ,
189 .mldv2_unsolicited_report_interval = HZ,
191 .rtr_solicits = MAX_RTR_SOLICITATIONS,
192 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
193 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
194 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
196 .temp_valid_lft = TEMP_VALID_LIFETIME,
197 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
198 .regen_min_advance = REGEN_MIN_ADVANCE,
199 .regen_max_retry = REGEN_MAX_RETRY,
200 .max_desync_factor = MAX_DESYNC_FACTOR,
201 .max_addresses = IPV6_MAX_ADDRESSES,
202 .accept_ra_defrtr = 1,
203 .ra_defrtr_metric = IP6_RT_PRIO_USER,
204 .accept_ra_from_local = 0,
205 .accept_ra_min_hop_limit= 1,
206 .accept_ra_min_lft = 0,
207 .accept_ra_pinfo = 1,
208 #ifdef CONFIG_IPV6_ROUTER_PREF
209 .accept_ra_rtr_pref = 1,
210 .rtr_probe_interval = 60 * HZ,
211 #ifdef CONFIG_IPV6_ROUTE_INFO
212 .accept_ra_rt_info_min_plen = 0,
213 .accept_ra_rt_info_max_plen = 0,
217 .accept_source_route = 0, /* we do not accept RH0 by default. */
220 .suppress_frag_ndisc = 1,
223 .initialized = false,
225 .use_oif_addrs_only = 0,
226 .ignore_routes_with_linkdown = 0,
227 .keep_addr_on_down = 0,
229 #ifdef CONFIG_IPV6_SEG6_HMAC
230 .seg6_require_hmac = 0,
233 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
235 .rpl_seg_enabled = 0,
237 .ioam6_id = IOAM6_DEFAULT_IF_ID,
238 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
239 .ndisc_evict_nocarrier = 1,
240 .ra_honor_pio_life = 0,
243 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
245 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
246 .mtu6 = IPV6_MIN_MTU,
248 .accept_redirects = 1,
250 .force_mld_version = 0,
251 .mldv1_unsolicited_report_interval = 10 * HZ,
252 .mldv2_unsolicited_report_interval = HZ,
254 .rtr_solicits = MAX_RTR_SOLICITATIONS,
255 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
256 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
257 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
259 .temp_valid_lft = TEMP_VALID_LIFETIME,
260 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
261 .regen_min_advance = REGEN_MIN_ADVANCE,
262 .regen_max_retry = REGEN_MAX_RETRY,
263 .max_desync_factor = MAX_DESYNC_FACTOR,
264 .max_addresses = IPV6_MAX_ADDRESSES,
265 .accept_ra_defrtr = 1,
266 .ra_defrtr_metric = IP6_RT_PRIO_USER,
267 .accept_ra_from_local = 0,
268 .accept_ra_min_hop_limit= 1,
269 .accept_ra_min_lft = 0,
270 .accept_ra_pinfo = 1,
271 #ifdef CONFIG_IPV6_ROUTER_PREF
272 .accept_ra_rtr_pref = 1,
273 .rtr_probe_interval = 60 * HZ,
274 #ifdef CONFIG_IPV6_ROUTE_INFO
275 .accept_ra_rt_info_min_plen = 0,
276 .accept_ra_rt_info_max_plen = 0,
280 .accept_source_route = 0, /* we do not accept RH0 by default. */
283 .suppress_frag_ndisc = 1,
286 .initialized = false,
288 .use_oif_addrs_only = 0,
289 .ignore_routes_with_linkdown = 0,
290 .keep_addr_on_down = 0,
292 #ifdef CONFIG_IPV6_SEG6_HMAC
293 .seg6_require_hmac = 0,
296 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64,
298 .rpl_seg_enabled = 0,
300 .ioam6_id = IOAM6_DEFAULT_IF_ID,
301 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE,
302 .ndisc_evict_nocarrier = 1,
303 .ra_honor_pio_life = 0,
306 /* Check if link is ready: is it up and is a valid qdisc available */
307 static inline bool addrconf_link_ready(const struct net_device *dev)
309 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
312 static void addrconf_del_rs_timer(struct inet6_dev *idev)
314 if (del_timer(&idev->rs_timer))
318 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
320 if (cancel_delayed_work(&ifp->dad_work))
324 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
327 if (!mod_timer(&idev->rs_timer, jiffies + when))
331 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
335 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
339 static int snmp6_alloc_dev(struct inet6_dev *idev)
343 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
344 if (!idev->stats.ipv6)
347 for_each_possible_cpu(i) {
348 struct ipstats_mib *addrconf_stats;
349 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
350 u64_stats_init(&addrconf_stats->syncp);
354 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
356 if (!idev->stats.icmpv6dev)
358 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
360 if (!idev->stats.icmpv6msgdev)
366 kfree(idev->stats.icmpv6dev);
368 free_percpu(idev->stats.ipv6);
373 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
375 struct inet6_dev *ndev;
380 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
381 return ERR_PTR(-EINVAL);
383 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
387 rwlock_init(&ndev->lock);
389 INIT_LIST_HEAD(&ndev->addr_list);
390 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
391 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
393 if (ndev->cnf.stable_secret.initialized)
394 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
396 ndev->cnf.mtu6 = dev->mtu;
398 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
399 if (!ndev->nd_parms) {
403 if (ndev->cnf.forwarding)
404 dev_disable_lro(dev);
405 /* We refer to the device */
406 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL);
408 if (snmp6_alloc_dev(ndev) < 0) {
409 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
411 neigh_parms_release(&nd_tbl, ndev->nd_parms);
412 netdev_put(dev, &ndev->dev_tracker);
417 if (dev != blackhole_netdev) {
418 if (snmp6_register_dev(ndev) < 0) {
419 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
420 __func__, dev->name);
424 /* One reference from device. */
425 refcount_set(&ndev->refcnt, 1);
427 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
428 ndev->cnf.accept_dad = -1;
430 #if IS_ENABLED(CONFIG_IPV6_SIT)
431 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
432 pr_info("%s: Disabled Multicast RS\n", dev->name);
433 ndev->cnf.rtr_solicits = 0;
437 INIT_LIST_HEAD(&ndev->tempaddr_list);
438 ndev->desync_factor = U32_MAX;
439 if ((dev->flags&IFF_LOOPBACK) ||
440 dev->type == ARPHRD_TUNNEL ||
441 dev->type == ARPHRD_TUNNEL6 ||
442 dev->type == ARPHRD_SIT ||
443 dev->type == ARPHRD_NONE) {
444 ndev->cnf.use_tempaddr = -1;
447 ndev->token = in6addr_any;
449 if (netif_running(dev) && addrconf_link_ready(dev))
450 ndev->if_flags |= IF_READY;
452 ipv6_mc_init_dev(ndev);
453 ndev->tstamp = jiffies;
454 if (dev != blackhole_netdev) {
455 err = addrconf_sysctl_register(ndev);
457 ipv6_mc_destroy_dev(ndev);
458 snmp6_unregister_dev(ndev);
462 /* protected by rtnl_lock */
463 rcu_assign_pointer(dev->ip6_ptr, ndev);
465 if (dev != blackhole_netdev) {
466 /* Join interface-local all-node multicast group */
467 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
469 /* Join all-node multicast group */
470 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
472 /* Join all-router multicast group if forwarding is set */
473 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
474 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
479 neigh_parms_release(&nd_tbl, ndev->nd_parms);
481 in6_dev_finish_destroy(ndev);
485 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
487 struct inet6_dev *idev;
491 idev = __in6_dev_get(dev);
493 idev = ipv6_add_dev(dev);
498 if (dev->flags&IFF_UP)
503 static int inet6_netconf_msgsize_devconf(int type)
505 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
506 + nla_total_size(4); /* NETCONFA_IFINDEX */
509 if (type == NETCONFA_ALL)
512 if (all || type == NETCONFA_FORWARDING)
513 size += nla_total_size(4);
514 #ifdef CONFIG_IPV6_MROUTE
515 if (all || type == NETCONFA_MC_FORWARDING)
516 size += nla_total_size(4);
518 if (all || type == NETCONFA_PROXY_NEIGH)
519 size += nla_total_size(4);
521 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
522 size += nla_total_size(4);
527 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
528 struct ipv6_devconf *devconf, u32 portid,
529 u32 seq, int event, unsigned int flags,
532 struct nlmsghdr *nlh;
533 struct netconfmsg *ncm;
536 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
541 if (type == NETCONFA_ALL)
544 ncm = nlmsg_data(nlh);
545 ncm->ncm_family = AF_INET6;
547 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
548 goto nla_put_failure;
553 if ((all || type == NETCONFA_FORWARDING) &&
554 nla_put_s32(skb, NETCONFA_FORWARDING,
555 READ_ONCE(devconf->forwarding)) < 0)
556 goto nla_put_failure;
557 #ifdef CONFIG_IPV6_MROUTE
558 if ((all || type == NETCONFA_MC_FORWARDING) &&
559 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
560 atomic_read(&devconf->mc_forwarding)) < 0)
561 goto nla_put_failure;
563 if ((all || type == NETCONFA_PROXY_NEIGH) &&
564 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
565 READ_ONCE(devconf->proxy_ndp)) < 0)
566 goto nla_put_failure;
568 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
569 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
570 READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0)
571 goto nla_put_failure;
578 nlmsg_cancel(skb, nlh);
582 void inet6_netconf_notify_devconf(struct net *net, int event, int type,
583 int ifindex, struct ipv6_devconf *devconf)
588 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
592 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
595 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
596 WARN_ON(err == -EMSGSIZE);
600 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
603 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
606 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
607 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
608 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
609 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
610 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
613 static int inet6_netconf_valid_get_req(struct sk_buff *skb,
614 const struct nlmsghdr *nlh,
616 struct netlink_ext_ack *extack)
620 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
621 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
625 if (!netlink_strict_get_check(skb))
626 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
628 devconf_ipv6_policy, extack);
630 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
632 devconf_ipv6_policy, extack);
636 for (i = 0; i <= NETCONFA_MAX; i++) {
641 case NETCONFA_IFINDEX:
644 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
652 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
653 struct nlmsghdr *nlh,
654 struct netlink_ext_ack *extack)
656 struct net *net = sock_net(in_skb->sk);
657 struct nlattr *tb[NETCONFA_MAX+1];
658 struct inet6_dev *in6_dev = NULL;
659 struct net_device *dev = NULL;
661 struct ipv6_devconf *devconf;
665 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
669 if (!tb[NETCONFA_IFINDEX])
673 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
675 case NETCONFA_IFINDEX_ALL:
676 devconf = net->ipv6.devconf_all;
678 case NETCONFA_IFINDEX_DEFAULT:
679 devconf = net->ipv6.devconf_dflt;
682 dev = dev_get_by_index(net, ifindex);
685 in6_dev = in6_dev_get(dev);
688 devconf = &in6_dev->cnf;
693 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
697 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
698 NETLINK_CB(in_skb).portid,
699 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
702 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
703 WARN_ON(err == -EMSGSIZE);
707 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
710 in6_dev_put(in6_dev);
715 /* Combine dev_addr_genid and dev_base_seq to detect changes.
717 static u32 inet6_base_seq(const struct net *net)
719 u32 res = atomic_read(&net->ipv6.dev_addr_genid) +
720 READ_ONCE(net->dev_base_seq);
722 /* Must not return 0 (see nl_dump_check_consistent()).
723 * Chose a value far away from 0.
730 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
731 struct netlink_callback *cb)
733 const struct nlmsghdr *nlh = cb->nlh;
734 struct net *net = sock_net(skb->sk);
736 unsigned long ifindex;
737 unsigned int all_default;
738 } *ctx = (void *)cb->ctx;
739 struct net_device *dev;
740 struct inet6_dev *idev;
743 if (cb->strict_check) {
744 struct netlink_ext_ack *extack = cb->extack;
745 struct netconfmsg *ncm;
747 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
748 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
752 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
753 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
759 for_each_netdev_dump(net, dev, ctx->ifindex) {
760 idev = __in6_dev_get(dev);
763 err = inet6_netconf_fill_devconf(skb, dev->ifindex,
765 NETLINK_CB(cb->skb).portid,
773 if (ctx->all_default == 0) {
774 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
775 net->ipv6.devconf_all,
776 NETLINK_CB(cb->skb).portid,
778 RTM_NEWNETCONF, NLM_F_MULTI,
784 if (ctx->all_default == 1) {
785 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
786 net->ipv6.devconf_dflt,
787 NETLINK_CB(cb->skb).portid,
789 RTM_NEWNETCONF, NLM_F_MULTI,
801 static void dev_forward_change(struct inet6_dev *idev)
803 struct net_device *dev;
804 struct inet6_ifaddr *ifa;
805 LIST_HEAD(tmp_addr_list);
810 if (idev->cnf.forwarding)
811 dev_disable_lro(dev);
812 if (dev->flags & IFF_MULTICAST) {
813 if (idev->cnf.forwarding) {
814 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
815 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
816 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
818 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
819 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
820 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
824 read_lock_bh(&idev->lock);
825 list_for_each_entry(ifa, &idev->addr_list, if_list) {
826 if (ifa->flags&IFA_F_TENTATIVE)
828 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
830 read_unlock_bh(&idev->lock);
832 while (!list_empty(&tmp_addr_list)) {
833 ifa = list_first_entry(&tmp_addr_list,
834 struct inet6_ifaddr, if_list_aux);
835 list_del(&ifa->if_list_aux);
836 if (idev->cnf.forwarding)
837 addrconf_join_anycast(ifa);
839 addrconf_leave_anycast(ifa);
842 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
844 dev->ifindex, &idev->cnf);
848 static void addrconf_forward_change(struct net *net, __s32 newf)
850 struct net_device *dev;
851 struct inet6_dev *idev;
853 for_each_netdev(net, dev) {
854 idev = __in6_dev_get(dev);
856 int changed = (!idev->cnf.forwarding) ^ (!newf);
858 WRITE_ONCE(idev->cnf.forwarding, newf);
860 dev_forward_change(idev);
865 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
871 return restart_syscall();
873 net = (struct net *)table->extra2;
875 WRITE_ONCE(*p, newf);
877 if (p == &net->ipv6.devconf_dflt->forwarding) {
878 if ((!newf) ^ (!old))
879 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
881 NETCONFA_IFINDEX_DEFAULT,
882 net->ipv6.devconf_dflt);
887 if (p == &net->ipv6.devconf_all->forwarding) {
888 int old_dflt = net->ipv6.devconf_dflt->forwarding;
890 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf);
891 if ((!newf) ^ (!old_dflt))
892 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
894 NETCONFA_IFINDEX_DEFAULT,
895 net->ipv6.devconf_dflt);
897 addrconf_forward_change(net, newf);
898 if ((!newf) ^ (!old))
899 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
901 NETCONFA_IFINDEX_ALL,
902 net->ipv6.devconf_all);
903 } else if ((!newf) ^ (!old))
904 dev_forward_change((struct inet6_dev *)table->extra1);
908 rt6_purge_dflt_routers(net);
912 static void addrconf_linkdown_change(struct net *net, __s32 newf)
914 struct net_device *dev;
915 struct inet6_dev *idev;
917 for_each_netdev(net, dev) {
918 idev = __in6_dev_get(dev);
920 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
922 WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf);
924 inet6_netconf_notify_devconf(dev_net(dev),
926 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
933 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
939 return restart_syscall();
941 net = (struct net *)table->extra2;
943 WRITE_ONCE(*p, newf);
945 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
946 if ((!newf) ^ (!old))
947 inet6_netconf_notify_devconf(net,
949 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
950 NETCONFA_IFINDEX_DEFAULT,
951 net->ipv6.devconf_dflt);
956 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
957 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf);
958 addrconf_linkdown_change(net, newf);
959 if ((!newf) ^ (!old))
960 inet6_netconf_notify_devconf(net,
962 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
963 NETCONFA_IFINDEX_ALL,
964 net->ipv6.devconf_all);
973 /* Nobody refers to this ifaddr, destroy it */
974 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
976 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
978 #ifdef NET_REFCNT_DEBUG
979 pr_debug("%s\n", __func__);
982 in6_dev_put(ifp->idev);
984 if (cancel_delayed_work(&ifp->dad_work))
985 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
988 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
989 pr_warn("Freeing alive inet6 address %p\n", ifp);
997 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
1000 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
1003 * Each device address list is sorted in order of scope -
1004 * global before linklocal.
1006 list_for_each(p, &idev->addr_list) {
1007 struct inet6_ifaddr *ifa
1008 = list_entry(p, struct inet6_ifaddr, if_list);
1009 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
1013 list_add_tail_rcu(&ifp->if_list, p);
1016 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
1018 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
1020 return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
1023 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1024 struct net_device *dev, unsigned int hash)
1026 struct inet6_ifaddr *ifp;
1028 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1029 if (ipv6_addr_equal(&ifp->addr, addr)) {
1030 if (!dev || ifp->idev->dev == dev)
1037 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
1039 struct net *net = dev_net(dev);
1040 unsigned int hash = inet6_addr_hash(net, &ifa->addr);
1043 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1045 /* Ignore adding duplicate addresses on an interface */
1046 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
1047 netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
1050 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
1053 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1058 /* On success it returns ifp with increased reference count */
1060 static struct inet6_ifaddr *
1061 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
1062 bool can_block, struct netlink_ext_ack *extack)
1064 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
1065 int addr_type = ipv6_addr_type(cfg->pfx);
1066 struct net *net = dev_net(idev->dev);
1067 struct inet6_ifaddr *ifa = NULL;
1068 struct fib6_info *f6i = NULL;
1071 if (addr_type == IPV6_ADDR_ANY) {
1072 NL_SET_ERR_MSG_MOD(extack, "Invalid address");
1073 return ERR_PTR(-EADDRNOTAVAIL);
1074 } else if (addr_type & IPV6_ADDR_MULTICAST &&
1075 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) {
1076 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag");
1077 return ERR_PTR(-EADDRNOTAVAIL);
1078 } else if (!(idev->dev->flags & IFF_LOOPBACK) &&
1079 !netif_is_l3_master(idev->dev) &&
1080 addr_type & IPV6_ADDR_LOOPBACK) {
1081 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device");
1082 return ERR_PTR(-EADDRNOTAVAIL);
1086 NL_SET_ERR_MSG_MOD(extack, "device is going away");
1091 if (idev->cnf.disable_ipv6) {
1092 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
1097 /* validator notifier needs to be blocking;
1098 * do not call in atomic context
1101 struct in6_validator_info i6vi = {
1102 .i6vi_addr = *cfg->pfx,
1107 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
1108 err = notifier_to_errno(err);
1113 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
1119 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack);
1126 neigh_parms_data_state_setall(idev->nd_parms);
1128 ifa->addr = *cfg->pfx;
1130 ifa->peer_addr = *cfg->peer_pfx;
1132 spin_lock_init(&ifa->lock);
1133 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
1134 INIT_HLIST_NODE(&ifa->addr_lst);
1135 ifa->scope = cfg->scope;
1136 ifa->prefix_len = cfg->plen;
1137 ifa->rt_priority = cfg->rt_priority;
1138 ifa->flags = cfg->ifa_flags;
1139 ifa->ifa_proto = cfg->ifa_proto;
1140 /* No need to add the TENTATIVE flag for addresses with NODAD */
1141 if (!(cfg->ifa_flags & IFA_F_NODAD))
1142 ifa->flags |= IFA_F_TENTATIVE;
1143 ifa->valid_lft = cfg->valid_lft;
1144 ifa->prefered_lft = cfg->preferred_lft;
1145 ifa->cstamp = ifa->tstamp = jiffies;
1146 ifa->tokenized = false;
1154 refcount_set(&ifa->refcnt, 1);
1158 err = ipv6_add_addr_hash(idev->dev, ifa);
1164 write_lock_bh(&idev->lock);
1166 /* Add to inet6_dev unicast addr list. */
1167 ipv6_link_dev_addr(idev, ifa);
1169 if (ifa->flags&IFA_F_TEMPORARY) {
1170 list_add(&ifa->tmp_list, &idev->tempaddr_list);
1175 write_unlock_bh(&idev->lock);
1179 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
1181 if (unlikely(err < 0)) {
1182 fib6_info_release(f6i);
1186 in6_dev_put(ifa->idev);
1195 enum cleanup_prefix_rt_t {
1196 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
1197 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
1198 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
1202 * Check, whether the prefix for ifp would still need a prefix route
1203 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
1206 * 1) we don't purge prefix if address was not permanent.
1207 * prefix is managed by its own lifetime.
1208 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
1209 * 3) if there are no addresses, delete prefix.
1210 * 4) if there are still other permanent address(es),
1211 * corresponding prefix is still permanent.
1212 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
1213 * don't purge the prefix, assume user space is managing it.
1214 * 6) otherwise, update prefix lifetime to the
1215 * longest valid lifetime among the corresponding
1216 * addresses on the device.
1217 * Note: subsequent RA will update lifetime.
1219 static enum cleanup_prefix_rt_t
1220 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
1222 struct inet6_ifaddr *ifa;
1223 struct inet6_dev *idev = ifp->idev;
1224 unsigned long lifetime;
1225 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
1229 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1232 if (ifa->prefix_len != ifp->prefix_len ||
1233 !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
1236 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
1237 return CLEANUP_PREFIX_RT_NOP;
1239 action = CLEANUP_PREFIX_RT_EXPIRE;
1241 spin_lock(&ifa->lock);
1243 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
1245 * Note: Because this address is
1246 * not permanent, lifetime <
1247 * LONG_MAX / HZ here.
1249 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
1250 *expires = ifa->tstamp + lifetime * HZ;
1251 spin_unlock(&ifa->lock);
1258 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
1259 bool del_rt, bool del_peer)
1261 struct fib6_table *table;
1262 struct fib6_info *f6i;
1264 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
1266 ifp->idev->dev, 0, RTF_DEFAULT, true);
1269 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
1271 if (!(f6i->fib6_flags & RTF_EXPIRES)) {
1272 table = f6i->fib6_table;
1273 spin_lock_bh(&table->tb6_lock);
1275 fib6_set_expires(f6i, expires);
1276 fib6_add_gc_list(f6i);
1278 spin_unlock_bh(&table->tb6_lock);
1280 fib6_info_release(f6i);
1286 /* This function wants to get referenced ifp and releases it before return */
1288 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
1290 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
1291 struct net *net = dev_net(ifp->idev->dev);
1292 unsigned long expires;
1297 spin_lock_bh(&ifp->lock);
1299 ifp->state = INET6_IFADDR_STATE_DEAD;
1300 spin_unlock_bh(&ifp->lock);
1302 if (state == INET6_IFADDR_STATE_DEAD)
1305 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
1306 hlist_del_init_rcu(&ifp->addr_lst);
1307 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
1309 write_lock_bh(&ifp->idev->lock);
1311 if (ifp->flags&IFA_F_TEMPORARY) {
1312 list_del(&ifp->tmp_list);
1314 in6_ifa_put(ifp->ifpub);
1320 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1321 action = check_cleanup_prefix_route(ifp, &expires);
1323 list_del_rcu(&ifp->if_list);
1326 write_unlock_bh(&ifp->idev->lock);
1328 addrconf_del_dad_work(ifp);
1330 ipv6_ifa_notify(RTM_DELADDR, ifp);
1332 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1334 if (action != CLEANUP_PREFIX_RT_NOP) {
1335 cleanup_prefix_route(ifp, expires,
1336 action == CLEANUP_PREFIX_RT_DEL, false);
1339 /* clean up prefsrc entries */
1340 rt6_remove_prefsrc(ifp);
1345 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev)
1347 return READ_ONCE(idev->cnf.regen_min_advance) +
1348 READ_ONCE(idev->cnf.regen_max_retry) *
1349 READ_ONCE(idev->cnf.dad_transmits) *
1350 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
1353 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
1355 struct inet6_dev *idev = ifp->idev;
1356 unsigned long tmp_tstamp, age;
1357 unsigned long regen_advance;
1358 unsigned long now = jiffies;
1359 u32 if_public_preferred_lft;
1360 s32 cnf_temp_preferred_lft;
1361 struct inet6_ifaddr *ift;
1362 struct ifa6_config cfg;
1363 long max_desync_factor;
1364 struct in6_addr addr;
1367 write_lock_bh(&idev->lock);
1371 if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) {
1372 write_unlock_bh(&idev->lock);
1373 pr_info("%s: use_tempaddr is disabled\n", __func__);
1378 spin_lock_bh(&ifp->lock);
1379 if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) {
1380 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/
1381 spin_unlock_bh(&ifp->lock);
1382 write_unlock_bh(&idev->lock);
1383 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1390 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1391 ipv6_gen_rnd_iid(&addr);
1393 age = (now - ifp->tstamp) / HZ;
1395 regen_advance = ipv6_get_regen_advance(idev);
1397 /* recalculate max_desync_factor each time and update
1398 * idev->desync_factor if it's larger
1400 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
1401 max_desync_factor = min_t(long,
1402 READ_ONCE(idev->cnf.max_desync_factor),
1403 cnf_temp_preferred_lft - regen_advance);
1405 if (unlikely(idev->desync_factor > max_desync_factor)) {
1406 if (max_desync_factor > 0) {
1407 get_random_bytes(&idev->desync_factor,
1408 sizeof(idev->desync_factor));
1409 idev->desync_factor %= max_desync_factor;
1411 idev->desync_factor = 0;
1415 if_public_preferred_lft = ifp->prefered_lft;
1417 memset(&cfg, 0, sizeof(cfg));
1418 cfg.valid_lft = min_t(__u32, ifp->valid_lft,
1419 READ_ONCE(idev->cnf.temp_valid_lft) + age);
1420 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
1421 cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft);
1422 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft);
1424 cfg.plen = ifp->prefix_len;
1425 tmp_tstamp = ifp->tstamp;
1426 spin_unlock_bh(&ifp->lock);
1428 write_unlock_bh(&idev->lock);
1432 * A temporary address is created only if this calculated Preferred
1433 * Lifetime is greater than REGEN_ADVANCE time units. In
1434 * particular, an implementation must not create a temporary address
1435 * with a zero Preferred Lifetime.
1439 * When creating a temporary address, the lifetime values MUST be
1440 * derived from the corresponding prefix as follows:
1444 * * Its Preferred Lifetime is the lower of the Preferred Lifetime
1445 * of the public address or TEMP_PREFERRED_LIFETIME -
1448 * To comply with the RFC's requirements, clamp the preferred lifetime
1449 * to a minimum of regen_advance, unless that would exceed valid_lft or
1450 * ifp->prefered_lft.
1452 * Use age calculation as in addrconf_verify to avoid unnecessary
1453 * temporary addresses being generated.
1455 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1456 if (cfg.preferred_lft <= regen_advance + age) {
1457 cfg.preferred_lft = regen_advance + age + 1;
1458 if (cfg.preferred_lft > cfg.valid_lft ||
1459 cfg.preferred_lft > if_public_preferred_lft) {
1467 cfg.ifa_flags = IFA_F_TEMPORARY;
1468 /* set in addrconf_prefix_rcv() */
1469 if (ifp->flags & IFA_F_OPTIMISTIC)
1470 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
1473 cfg.scope = ipv6_addr_scope(cfg.pfx);
1475 ift = ipv6_add_addr(idev, &cfg, block, NULL);
1479 pr_info("%s: retry temporary address regeneration\n", __func__);
1480 write_lock_bh(&idev->lock);
1484 spin_lock_bh(&ift->lock);
1487 ift->tstamp = tmp_tstamp;
1488 spin_unlock_bh(&ift->lock);
1490 addrconf_dad_start(ift);
1498 * Choose an appropriate source address (RFC3484)
1501 IPV6_SADDR_RULE_INIT = 0,
1502 IPV6_SADDR_RULE_LOCAL,
1503 IPV6_SADDR_RULE_SCOPE,
1504 IPV6_SADDR_RULE_PREFERRED,
1505 #ifdef CONFIG_IPV6_MIP6
1506 IPV6_SADDR_RULE_HOA,
1508 IPV6_SADDR_RULE_OIF,
1509 IPV6_SADDR_RULE_LABEL,
1510 IPV6_SADDR_RULE_PRIVACY,
1511 IPV6_SADDR_RULE_ORCHID,
1512 IPV6_SADDR_RULE_PREFIX,
1513 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1514 IPV6_SADDR_RULE_NOT_OPTIMISTIC,
1519 struct ipv6_saddr_score {
1522 struct inet6_ifaddr *ifa;
1523 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1528 struct ipv6_saddr_dst {
1529 const struct in6_addr *addr;
1536 static inline int ipv6_saddr_preferred(int type)
1538 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1543 static bool ipv6_use_optimistic_addr(const struct net *net,
1544 const struct inet6_dev *idev)
1546 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1549 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1550 !READ_ONCE(idev->cnf.optimistic_dad))
1552 if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) &&
1553 !READ_ONCE(idev->cnf.use_optimistic))
1562 static bool ipv6_allow_optimistic_dad(const struct net *net,
1563 const struct inet6_dev *idev)
1565 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1568 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) &&
1569 !READ_ONCE(idev->cnf.optimistic_dad))
1578 static int ipv6_get_saddr_eval(struct net *net,
1579 struct ipv6_saddr_score *score,
1580 struct ipv6_saddr_dst *dst,
1585 if (i <= score->rule) {
1587 case IPV6_SADDR_RULE_SCOPE:
1588 ret = score->scopedist;
1590 case IPV6_SADDR_RULE_PREFIX:
1591 ret = score->matchlen;
1594 ret = !!test_bit(i, score->scorebits);
1600 case IPV6_SADDR_RULE_INIT:
1601 /* Rule 0: remember if hiscore is not ready yet */
1604 case IPV6_SADDR_RULE_LOCAL:
1605 /* Rule 1: Prefer same address */
1606 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1608 case IPV6_SADDR_RULE_SCOPE:
1609 /* Rule 2: Prefer appropriate scope
1614 * ---+--+-+---> scope
1616 * | d is scope of the destination.
1618 * | \ <- smaller scope is better if
1619 * B-15 | \ if scope is enough for destination.
1620 * | ret = B - scope (-1 <= scope >= d <= 15).
1622 * |/ <- greater is better
1623 * -C / if scope is not enough for destination.
1624 * /| ret = scope - C (-1 <= d < scope <= 15).
1626 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1627 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1628 * Assume B = 0 and we get C > 29.
1630 ret = __ipv6_addr_src_scope(score->addr_type);
1631 if (ret >= dst->scope)
1634 ret -= 128; /* 30 is enough */
1635 score->scopedist = ret;
1637 case IPV6_SADDR_RULE_PREFERRED:
1639 /* Rule 3: Avoid deprecated and optimistic addresses */
1640 u8 avoid = IFA_F_DEPRECATED;
1642 if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
1643 avoid |= IFA_F_OPTIMISTIC;
1644 ret = ipv6_saddr_preferred(score->addr_type) ||
1645 !(score->ifa->flags & avoid);
1648 #ifdef CONFIG_IPV6_MIP6
1649 case IPV6_SADDR_RULE_HOA:
1651 /* Rule 4: Prefer home address */
1652 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1653 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1657 case IPV6_SADDR_RULE_OIF:
1658 /* Rule 5: Prefer outgoing interface */
1659 ret = (!dst->ifindex ||
1660 dst->ifindex == score->ifa->idev->dev->ifindex);
1662 case IPV6_SADDR_RULE_LABEL:
1663 /* Rule 6: Prefer matching label */
1664 ret = ipv6_addr_label(net,
1665 &score->ifa->addr, score->addr_type,
1666 score->ifa->idev->dev->ifindex) == dst->label;
1668 case IPV6_SADDR_RULE_PRIVACY:
1670 /* Rule 7: Prefer public address
1671 * Note: prefer temporary address if use_tempaddr >= 2
1673 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1674 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1675 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2;
1676 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1679 case IPV6_SADDR_RULE_ORCHID:
1680 /* Rule 8-: Prefer ORCHID vs ORCHID or
1681 * non-ORCHID vs non-ORCHID
1683 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1684 ipv6_addr_orchid(dst->addr));
1686 case IPV6_SADDR_RULE_PREFIX:
1687 /* Rule 8: Use longest matching prefix */
1688 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1689 if (ret > score->ifa->prefix_len)
1690 ret = score->ifa->prefix_len;
1691 score->matchlen = ret;
1693 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1694 case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
1695 /* Optimistic addresses still have lower precedence than other
1696 * preferred addresses.
1698 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
1706 __set_bit(i, score->scorebits);
1712 static int __ipv6_dev_get_saddr(struct net *net,
1713 struct ipv6_saddr_dst *dst,
1714 struct inet6_dev *idev,
1715 struct ipv6_saddr_score *scores,
1718 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
1720 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
1724 * - Tentative Address (RFC2462 section 5.4)
1725 * - A tentative address is not considered
1726 * "assigned to an interface" in the traditional
1727 * sense, unless it is also flagged as optimistic.
1728 * - Candidate Source Address (section 4)
1729 * - In any case, anycast addresses, multicast
1730 * addresses, and the unspecified address MUST
1731 * NOT be included in a candidate set.
1733 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1734 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1737 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1739 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1740 score->addr_type & IPV6_ADDR_MULTICAST)) {
1741 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
1747 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1749 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1750 int minihiscore, miniscore;
1752 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
1753 miniscore = ipv6_get_saddr_eval(net, score, dst, i);
1755 if (minihiscore > miniscore) {
1756 if (i == IPV6_SADDR_RULE_SCOPE &&
1757 score->scopedist > 0) {
1760 * each remaining entry
1761 * has too small (not enough)
1762 * scope, because ifa entries
1763 * are sorted by their scope
1769 } else if (minihiscore < miniscore) {
1770 swap(hiscore, score);
1771 hiscore_idx = 1 - hiscore_idx;
1773 /* restore our iterator */
1774 score->ifa = hiscore->ifa;
1784 static int ipv6_get_saddr_master(struct net *net,
1785 const struct net_device *dst_dev,
1786 const struct net_device *master,
1787 struct ipv6_saddr_dst *dst,
1788 struct ipv6_saddr_score *scores,
1791 struct inet6_dev *idev;
1793 idev = __in6_dev_get(dst_dev);
1795 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1796 scores, hiscore_idx);
1798 idev = __in6_dev_get(master);
1800 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
1801 scores, hiscore_idx);
1806 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1807 const struct in6_addr *daddr, unsigned int prefs,
1808 struct in6_addr *saddr)
1810 struct ipv6_saddr_score scores[2], *hiscore;
1811 struct ipv6_saddr_dst dst;
1812 struct inet6_dev *idev;
1813 struct net_device *dev;
1815 bool use_oif_addr = false;
1816 int hiscore_idx = 0;
1819 dst_type = __ipv6_addr_type(daddr);
1821 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1822 dst.scope = __ipv6_addr_src_scope(dst_type);
1823 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1826 scores[hiscore_idx].rule = -1;
1827 scores[hiscore_idx].ifa = NULL;
1831 /* Candidate Source Address (section 4)
1832 * - multicast and link-local destination address,
1833 * the set of candidate source address MUST only
1834 * include addresses assigned to interfaces
1835 * belonging to the same link as the outgoing
1837 * (- For site-local destination addresses, the
1838 * set of candidate source addresses MUST only
1839 * include addresses assigned to interfaces
1840 * belonging to the same site as the outgoing
1842 * - "It is RECOMMENDED that the candidate source addresses
1843 * be the set of unicast addresses assigned to the
1844 * interface that will be used to send to the destination
1845 * (the 'outgoing' interface)." (RFC 6724)
1848 idev = __in6_dev_get(dst_dev);
1849 if ((dst_type & IPV6_ADDR_MULTICAST) ||
1850 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
1851 (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) {
1852 use_oif_addr = true;
1858 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1860 const struct net_device *master;
1863 /* if dst_dev exists and is enslaved to an L3 device, then
1864 * prefer addresses from dst_dev and then the master over
1865 * any other enslaved devices in the L3 domain.
1867 master = l3mdev_master_dev_rcu(dst_dev);
1869 master_idx = master->ifindex;
1871 hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
1873 scores, hiscore_idx);
1875 if (scores[hiscore_idx].ifa)
1879 for_each_netdev_rcu(net, dev) {
1880 /* only consider addresses on devices in the
1883 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1885 idev = __in6_dev_get(dev);
1888 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
1893 hiscore = &scores[hiscore_idx];
1895 ret = -EADDRNOTAVAIL;
1897 *saddr = hiscore->ifa->addr;
1902 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1904 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1907 struct inet6_ifaddr *ifp;
1908 int err = -EADDRNOTAVAIL;
1910 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1911 if (ifp->scope > IFA_LINK)
1913 if (ifp->scope == IFA_LINK &&
1914 !(ifp->flags & banned_flags)) {
1923 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1926 struct inet6_dev *idev;
1927 int err = -EADDRNOTAVAIL;
1930 idev = __in6_dev_get(dev);
1932 read_lock_bh(&idev->lock);
1933 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1934 read_unlock_bh(&idev->lock);
1940 static int ipv6_count_addresses(const struct inet6_dev *idev)
1942 const struct inet6_ifaddr *ifp;
1946 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
1952 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1953 const struct net_device *dev, int strict)
1955 return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
1956 strict, IFA_F_TENTATIVE);
1958 EXPORT_SYMBOL(ipv6_chk_addr);
1960 /* device argument is used to find the L3 domain of interest. If
1961 * skip_dev_check is set, then the ifp device is not checked against
1962 * the passed in dev argument. So the 2 cases for addresses checks are:
1963 * 1. does the address exist in the L3 domain that dev is part of
1964 * (skip_dev_check = true), or
1966 * 2. does the address exist on the specific device
1967 * (skip_dev_check = false)
1969 static struct net_device *
1970 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
1971 const struct net_device *dev, bool skip_dev_check,
1972 int strict, u32 banned_flags)
1974 unsigned int hash = inet6_addr_hash(net, addr);
1975 struct net_device *l3mdev, *ndev;
1976 struct inet6_ifaddr *ifp;
1981 l3mdev = l3mdev_master_dev_rcu(dev);
1985 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
1986 ndev = ifp->idev->dev;
1988 if (l3mdev_master_dev_rcu(ndev) != l3mdev)
1991 /* Decouple optimistic from tentative for evaluation here.
1992 * Ban optimistic addresses explicitly, when required.
1994 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
1995 ? (ifp->flags&~IFA_F_TENTATIVE)
1997 if (ipv6_addr_equal(&ifp->addr, addr) &&
1998 !(ifp_flags&banned_flags) &&
1999 (!dev || ndev == dev ||
2000 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
2010 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
2011 const struct net_device *dev, bool skip_dev_check,
2012 int strict, u32 banned_flags)
2014 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
2015 strict, banned_flags) ? 1 : 0;
2017 EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
2020 /* Compares an address/prefix_len with addresses on device @dev.
2021 * If one is found it returns true.
2023 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
2024 const unsigned int prefix_len, struct net_device *dev)
2026 const struct inet6_ifaddr *ifa;
2027 const struct inet6_dev *idev;
2031 idev = __in6_dev_get(dev);
2033 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2034 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
2043 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
2045 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
2047 const struct inet6_ifaddr *ifa;
2048 const struct inet6_dev *idev;
2053 idev = __in6_dev_get(dev);
2055 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
2056 onlink = ipv6_prefix_equal(addr, &ifa->addr,
2065 EXPORT_SYMBOL(ipv6_chk_prefix);
2068 * ipv6_dev_find - find the first device with a given source address.
2069 * @net: the net namespace
2070 * @addr: the source address
2071 * @dev: used to find the L3 domain of interest
2073 * The caller should be protected by RCU, or RTNL.
2075 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
2076 struct net_device *dev)
2078 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
2081 EXPORT_SYMBOL(ipv6_dev_find);
2083 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
2084 struct net_device *dev, int strict)
2086 unsigned int hash = inet6_addr_hash(net, addr);
2087 struct inet6_ifaddr *ifp, *result = NULL;
2090 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
2091 if (ipv6_addr_equal(&ifp->addr, addr)) {
2092 if (!dev || ifp->idev->dev == dev ||
2093 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
2105 /* Gets referenced address, destroys ifaddr */
2107 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
2110 ifp->flags |= IFA_F_DADFAILED;
2112 if (ifp->flags&IFA_F_TEMPORARY) {
2113 struct inet6_ifaddr *ifpub;
2114 spin_lock_bh(&ifp->lock);
2117 in6_ifa_hold(ifpub);
2118 spin_unlock_bh(&ifp->lock);
2119 ipv6_create_tempaddr(ifpub, true);
2122 spin_unlock_bh(&ifp->lock);
2125 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
2126 spin_lock_bh(&ifp->lock);
2127 addrconf_del_dad_work(ifp);
2128 ifp->flags |= IFA_F_TENTATIVE;
2130 ifp->flags &= ~IFA_F_OPTIMISTIC;
2131 spin_unlock_bh(&ifp->lock);
2133 ipv6_ifa_notify(0, ifp);
2140 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
2144 spin_lock_bh(&ifp->lock);
2145 if (ifp->state == INET6_IFADDR_STATE_DAD) {
2146 ifp->state = INET6_IFADDR_STATE_POSTDAD;
2149 spin_unlock_bh(&ifp->lock);
2154 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
2156 struct inet6_dev *idev = ifp->idev;
2157 struct net *net = dev_net(idev->dev);
2160 if (addrconf_dad_end(ifp)) {
2165 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
2166 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
2168 spin_lock_bh(&ifp->lock);
2170 if (ifp->flags & IFA_F_STABLE_PRIVACY) {
2171 struct in6_addr new_addr;
2172 struct inet6_ifaddr *ifp2;
2173 int retries = ifp->stable_privacy_retry + 1;
2174 struct ifa6_config cfg = {
2176 .plen = ifp->prefix_len,
2177 .ifa_flags = ifp->flags,
2178 .valid_lft = ifp->valid_lft,
2179 .preferred_lft = ifp->prefered_lft,
2180 .scope = ifp->scope,
2183 if (retries > net->ipv6.sysctl.idgen_retries) {
2184 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
2185 ifp->idev->dev->name);
2189 new_addr = ifp->addr;
2190 if (ipv6_generate_stable_address(&new_addr, retries,
2194 spin_unlock_bh(&ifp->lock);
2196 max_addresses = READ_ONCE(idev->cnf.max_addresses);
2197 if (max_addresses &&
2198 ipv6_count_addresses(idev) >= max_addresses)
2201 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
2202 ifp->idev->dev->name);
2204 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
2208 spin_lock_bh(&ifp2->lock);
2209 ifp2->stable_privacy_retry = retries;
2210 ifp2->state = INET6_IFADDR_STATE_PREDAD;
2211 spin_unlock_bh(&ifp2->lock);
2213 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
2216 spin_lock_bh(&ifp->lock);
2220 /* transition from _POSTDAD to _ERRDAD */
2221 ifp->state = INET6_IFADDR_STATE_ERRDAD;
2222 spin_unlock_bh(&ifp->lock);
2224 addrconf_mod_dad_work(ifp, 0);
2228 /* Join to solicited addr multicast group.
2229 * caller must hold RTNL */
2230 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
2232 struct in6_addr maddr;
2234 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2237 addrconf_addr_solict_mult(addr, &maddr);
2238 ipv6_dev_mc_inc(dev, &maddr);
2241 /* caller must hold RTNL */
2242 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
2244 struct in6_addr maddr;
2246 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
2249 addrconf_addr_solict_mult(addr, &maddr);
2250 __ipv6_dev_mc_dec(idev, &maddr);
2253 /* caller must hold RTNL */
2254 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
2256 struct in6_addr addr;
2258 if (ifp->prefix_len >= 127) /* RFC 6164 */
2260 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2261 if (ipv6_addr_any(&addr))
2263 __ipv6_dev_ac_inc(ifp->idev, &addr);
2266 /* caller must hold RTNL */
2267 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
2269 struct in6_addr addr;
2271 if (ifp->prefix_len >= 127) /* RFC 6164 */
2273 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
2274 if (ipv6_addr_any(&addr))
2276 __ipv6_dev_ac_dec(ifp->idev, &addr);
2279 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
2281 switch (dev->addr_len) {
2283 memcpy(eui, dev->dev_addr, 3);
2286 memcpy(eui + 5, dev->dev_addr + 3, 3);
2288 case EUI64_ADDR_LEN:
2289 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
2299 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
2301 const union fwnet_hwaddr *ha;
2303 if (dev->addr_len != FWNET_ALEN)
2306 ha = (const union fwnet_hwaddr *)dev->dev_addr;
2308 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
2313 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
2315 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
2316 if (dev->addr_len != ARCNET_ALEN)
2319 eui[7] = *(u8 *)dev->dev_addr;
2323 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
2325 if (dev->addr_len != INFINIBAND_ALEN)
2327 memcpy(eui, dev->dev_addr + 12, 8);
2332 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
2336 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
2337 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
2338 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
2339 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
2340 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
2341 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
2345 memcpy(eui + 4, &addr, 4);
2349 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
2351 if (dev->priv_flags & IFF_ISATAP)
2352 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2356 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
2358 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
2361 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
2363 memcpy(eui, dev->perm_addr, 3);
2364 memcpy(eui + 5, dev->perm_addr + 3, 3);
2371 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
2373 switch (dev->type) {
2376 return addrconf_ifid_eui48(eui, dev);
2378 return addrconf_ifid_arcnet(eui, dev);
2379 case ARPHRD_INFINIBAND:
2380 return addrconf_ifid_infiniband(eui, dev);
2382 return addrconf_ifid_sit(eui, dev);
2385 return addrconf_ifid_gre(eui, dev);
2386 case ARPHRD_6LOWPAN:
2387 return addrconf_ifid_6lowpan(eui, dev);
2388 case ARPHRD_IEEE1394:
2389 return addrconf_ifid_ieee1394(eui, dev);
2390 case ARPHRD_TUNNEL6:
2393 return addrconf_ifid_ip6tnl(eui, dev);
2398 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
2401 struct inet6_ifaddr *ifp;
2403 read_lock_bh(&idev->lock);
2404 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
2405 if (ifp->scope > IFA_LINK)
2407 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
2408 memcpy(eui, ifp->addr.s6_addr+8, 8);
2413 read_unlock_bh(&idev->lock);
2417 /* Generation of a randomized Interface Identifier
2418 * draft-ietf-6man-rfc4941bis, Section 3.3.1
2421 static void ipv6_gen_rnd_iid(struct in6_addr *addr)
2424 get_random_bytes(&addr->s6_addr[8], 8);
2426 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
2427 * check if generated address is not inappropriate:
2429 * - Reserved IPv6 Interface Identifiers
2430 * - XXX: already assigned to an address on the device
2433 /* Subnet-router anycast: 0000:0000:0000:0000 */
2434 if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
2437 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
2438 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213
2439 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
2441 if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
2442 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
2445 /* Reserved subnet anycast addresses */
2446 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
2447 ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
2456 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
2457 struct net_device *dev, unsigned long expires,
2458 u32 flags, gfp_t gfp_flags)
2460 struct fib6_config cfg = {
2461 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
2462 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
2463 .fc_ifindex = dev->ifindex,
2464 .fc_expires = expires,
2466 .fc_flags = RTF_UP | flags,
2467 .fc_nlinfo.nl_net = dev_net(dev),
2468 .fc_protocol = RTPROT_KERNEL,
2469 .fc_type = RTN_UNICAST,
2474 /* Prevent useless cloning on PtP SIT.
2475 This thing is done here expecting that the whole
2476 class of non-broadcast devices need not cloning.
2478 #if IS_ENABLED(CONFIG_IPV6_SIT)
2479 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2480 cfg.fc_flags |= RTF_NONEXTHOP;
2483 ip6_route_add(&cfg, gfp_flags, NULL);
2487 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2489 const struct net_device *dev,
2490 u32 flags, u32 noflags,
2493 struct fib6_node *fn;
2494 struct fib6_info *rt = NULL;
2495 struct fib6_table *table;
2496 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
2498 table = fib6_get_table(dev_net(dev), tb_id);
2503 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
2507 for_each_fib6_node_rt_rcu(fn) {
2508 /* prefix routes only use builtin fib6_nh */
2512 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
2514 if (no_gw && rt->fib6_nh->fib_nh_gw_family)
2516 if ((rt->fib6_flags & flags) != flags)
2518 if ((rt->fib6_flags & noflags) != 0)
2520 if (!fib6_info_hold_safe(rt))
2530 /* Create "default" multicast route to the interface */
2532 static void addrconf_add_mroute(struct net_device *dev)
2534 struct fib6_config cfg = {
2535 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
2536 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2537 .fc_ifindex = dev->ifindex,
2540 .fc_type = RTN_MULTICAST,
2541 .fc_nlinfo.nl_net = dev_net(dev),
2542 .fc_protocol = RTPROT_KERNEL,
2545 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2547 ip6_route_add(&cfg, GFP_KERNEL, NULL);
2550 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2552 struct inet6_dev *idev;
2556 idev = ipv6_find_idev(dev);
2560 if (idev->cnf.disable_ipv6)
2561 return ERR_PTR(-EACCES);
2563 /* Add default multicast route */
2564 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
2565 addrconf_add_mroute(dev);
2570 static void manage_tempaddrs(struct inet6_dev *idev,
2571 struct inet6_ifaddr *ifp,
2572 __u32 valid_lft, __u32 prefered_lft,
2573 bool create, unsigned long now)
2576 struct inet6_ifaddr *ift;
2578 read_lock_bh(&idev->lock);
2579 /* update all temporary addresses in the list */
2580 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2581 int age, max_valid, max_prefered;
2583 if (ifp != ift->ifpub)
2586 /* RFC 4941 section 3.3:
2587 * If a received option will extend the lifetime of a public
2588 * address, the lifetimes of temporary addresses should
2589 * be extended, subject to the overall constraint that no
2590 * temporary addresses should ever remain "valid" or "preferred"
2591 * for a time longer than (TEMP_VALID_LIFETIME) or
2592 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2594 age = (now - ift->cstamp) / HZ;
2595 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age;
2599 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) -
2600 idev->desync_factor - age;
2601 if (max_prefered < 0)
2604 if (valid_lft > max_valid)
2605 valid_lft = max_valid;
2607 if (prefered_lft > max_prefered)
2608 prefered_lft = max_prefered;
2610 spin_lock(&ift->lock);
2612 ift->valid_lft = valid_lft;
2613 ift->prefered_lft = prefered_lft;
2615 if (prefered_lft > 0)
2616 ift->flags &= ~IFA_F_DEPRECATED;
2618 spin_unlock(&ift->lock);
2619 if (!(flags&IFA_F_TENTATIVE))
2620 ipv6_ifa_notify(0, ift);
2623 /* Also create a temporary address if it's enabled but no temporary
2624 * address currently exists.
2625 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false
2626 * as part of cleanup (ie. deleting the mngtmpaddr).
2627 * We don't want that to result in creating a new temporary ip address.
2629 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft))
2632 if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) {
2633 /* When a new public address is created as described
2634 * in [ADDRCONF], also create a new temporary address.
2636 read_unlock_bh(&idev->lock);
2637 ipv6_create_tempaddr(ifp, false);
2639 read_unlock_bh(&idev->lock);
2643 static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
2645 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
2646 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
2649 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
2650 const struct prefix_info *pinfo,
2651 struct inet6_dev *in6_dev,
2652 const struct in6_addr *addr, int addr_type,
2653 u32 addr_flags, bool sllao, bool tokenized,
2654 __u32 valid_lft, u32 prefered_lft)
2656 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
2657 int create = 0, update_lft = 0;
2659 if (!ifp && valid_lft) {
2660 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses);
2661 struct ifa6_config cfg = {
2663 .plen = pinfo->prefix_len,
2664 .ifa_flags = addr_flags,
2665 .valid_lft = valid_lft,
2666 .preferred_lft = prefered_lft,
2667 .scope = addr_type & IPV6_ADDR_SCOPE_MASK,
2668 .ifa_proto = IFAPROT_KERNEL_RA
2671 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2672 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) ||
2673 READ_ONCE(in6_dev->cnf.optimistic_dad)) &&
2674 !net->ipv6.devconf_all->forwarding && sllao)
2675 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
2678 /* Do not allow to create too much of autoconfigured
2679 * addresses; this would be too easy way to crash kernel.
2681 if (!max_addresses ||
2682 ipv6_count_addresses(in6_dev) < max_addresses)
2683 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
2685 if (IS_ERR_OR_NULL(ifp))
2689 spin_lock_bh(&ifp->lock);
2690 ifp->flags |= IFA_F_MANAGETEMPADDR;
2691 ifp->cstamp = jiffies;
2692 ifp->tokenized = tokenized;
2693 spin_unlock_bh(&ifp->lock);
2694 addrconf_dad_start(ifp);
2702 /* update lifetime (RFC2462 5.5.3 e) */
2703 spin_lock_bh(&ifp->lock);
2705 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2706 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2710 /* RFC4862 Section 5.5.3e:
2711 * "Note that the preferred lifetime of the
2712 * corresponding address is always reset to
2713 * the Preferred Lifetime in the received
2714 * Prefix Information option, regardless of
2715 * whether the valid lifetime is also reset or
2718 * So we should always update prefered_lft here.
2720 update_lft = !create && stored_lft;
2722 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) {
2723 const u32 minimum_lft = min_t(u32,
2724 stored_lft, MIN_VALID_LIFETIME);
2725 valid_lft = max(valid_lft, minimum_lft);
2729 ifp->valid_lft = valid_lft;
2730 ifp->prefered_lft = prefered_lft;
2731 WRITE_ONCE(ifp->tstamp, now);
2733 ifp->flags &= ~IFA_F_DEPRECATED;
2734 spin_unlock_bh(&ifp->lock);
2736 if (!(flags&IFA_F_TENTATIVE))
2737 ipv6_ifa_notify(0, ifp);
2739 spin_unlock_bh(&ifp->lock);
2741 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2745 addrconf_verify(net);
2750 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
2752 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2754 struct prefix_info *pinfo;
2755 struct fib6_table *table;
2760 struct inet6_dev *in6_dev;
2761 struct net *net = dev_net(dev);
2763 pinfo = (struct prefix_info *) opt;
2765 if (len < sizeof(struct prefix_info)) {
2766 netdev_dbg(dev, "addrconf: prefix option too short\n");
2771 * Validation checks ([ADDRCONF], page 19)
2774 addr_type = ipv6_addr_type(&pinfo->prefix);
2776 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2779 valid_lft = ntohl(pinfo->valid);
2780 prefered_lft = ntohl(pinfo->prefered);
2782 if (prefered_lft > valid_lft) {
2783 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2787 in6_dev = in6_dev_get(dev);
2790 net_dbg_ratelimited("addrconf: device %s not configured\n",
2795 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft)
2799 * Two things going on here:
2800 * 1) Add routes for on-link prefixes
2801 * 2) Configure prefixes with the auto flag set
2804 if (pinfo->onlink) {
2805 struct fib6_info *rt;
2806 unsigned long rt_expires;
2808 /* Avoid arithmetic overflow. Really, we could
2809 * save rt_expires in seconds, likely valid_lft,
2810 * but it would require division in fib gc, that it
2814 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2816 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2818 if (addrconf_finite_timeout(rt_expires))
2821 rt = addrconf_get_prefix_route(&pinfo->prefix,
2824 RTF_ADDRCONF | RTF_PREFIX_RT,
2828 /* Autoconf prefix route */
2829 if (valid_lft == 0) {
2830 ip6_del_rt(net, rt, false);
2833 table = rt->fib6_table;
2834 spin_lock_bh(&table->tb6_lock);
2836 if (addrconf_finite_timeout(rt_expires)) {
2838 fib6_set_expires(rt, jiffies + rt_expires);
2839 fib6_add_gc_list(rt);
2841 fib6_clean_expires(rt);
2842 fib6_remove_gc_list(rt);
2845 spin_unlock_bh(&table->tb6_lock);
2847 } else if (valid_lft) {
2848 clock_t expires = 0;
2849 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2850 if (addrconf_finite_timeout(rt_expires)) {
2852 flags |= RTF_EXPIRES;
2853 expires = jiffies_to_clock_t(rt_expires);
2855 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2856 0, dev, expires, flags,
2859 fib6_info_release(rt);
2862 /* Try to figure out our local address for this prefix */
2864 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2865 struct in6_addr addr;
2866 bool tokenized = false, dev_addr_generated = false;
2868 if (pinfo->prefix_len == 64) {
2869 memcpy(&addr, &pinfo->prefix, 8);
2871 if (!ipv6_addr_any(&in6_dev->token)) {
2872 read_lock_bh(&in6_dev->lock);
2873 memcpy(addr.s6_addr + 8,
2874 in6_dev->token.s6_addr + 8, 8);
2875 read_unlock_bh(&in6_dev->lock);
2877 } else if (is_addr_mode_generate_stable(in6_dev) &&
2878 !ipv6_generate_stable_address(&addr, 0,
2880 addr_flags |= IFA_F_STABLE_PRIVACY;
2882 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2883 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2886 dev_addr_generated = true;
2890 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2895 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
2898 tokenized, valid_lft,
2903 /* Ignore error case here because previous prefix add addr was
2904 * successful which will be notified.
2906 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
2907 addr_type, addr_flags, sllao,
2908 tokenized, valid_lft,
2910 dev_addr_generated);
2912 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2914 in6_dev_put(in6_dev);
2917 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
2918 struct in6_ifreq *ireq)
2920 struct ip_tunnel_parm p = { };
2923 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
2924 return -EADDRNOTAVAIL;
2926 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
2929 p.iph.protocol = IPPROTO_IPV6;
2932 if (!dev->netdev_ops->ndo_tunnel_ctl)
2934 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
2938 dev = __dev_get_by_name(net, p.name);
2941 return dev_open(dev, NULL);
2945 * Set destination address.
2946 * Special case for SIT interfaces where we create a new "virtual"
2949 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2951 struct net_device *dev;
2952 struct in6_ifreq ireq;
2955 if (!IS_ENABLED(CONFIG_IPV6_SIT))
2957 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2961 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2962 if (dev && dev->type == ARPHRD_SIT)
2963 err = addrconf_set_sit_dstaddr(net, dev, &ireq);
2968 static int ipv6_mc_config(struct sock *sk, bool join,
2969 const struct in6_addr *addr, int ifindex)
2977 ret = ipv6_sock_mc_join(sk, ifindex, addr);
2979 ret = ipv6_sock_mc_drop(sk, ifindex, addr);
2986 * Manual configuration of address on an interface
2988 static int inet6_addr_add(struct net *net, int ifindex,
2989 struct ifa6_config *cfg,
2990 struct netlink_ext_ack *extack)
2992 struct inet6_ifaddr *ifp;
2993 struct inet6_dev *idev;
2994 struct net_device *dev;
2995 unsigned long timeout;
3001 if (cfg->plen > 128) {
3002 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3006 /* check the lifetime */
3007 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) {
3008 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid");
3012 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) {
3013 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64");
3017 dev = __dev_get_by_index(net, ifindex);
3021 idev = addrconf_add_dev(dev);
3023 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3024 return PTR_ERR(idev);
3027 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3028 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3029 true, cfg->pfx, ifindex);
3032 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed");
3037 cfg->scope = ipv6_addr_scope(cfg->pfx);
3039 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
3040 if (addrconf_finite_timeout(timeout)) {
3041 expires = jiffies_to_clock_t(timeout * HZ);
3042 cfg->valid_lft = timeout;
3043 flags = RTF_EXPIRES;
3047 cfg->ifa_flags |= IFA_F_PERMANENT;
3050 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
3051 if (addrconf_finite_timeout(timeout)) {
3053 cfg->ifa_flags |= IFA_F_DEPRECATED;
3054 cfg->preferred_lft = timeout;
3057 ifp = ipv6_add_addr(idev, cfg, true, extack);
3059 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
3060 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3061 ifp->rt_priority, dev, expires,
3065 /* Send a netlink notification if DAD is enabled and
3066 * optimistic flag is not set
3068 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
3069 ipv6_ifa_notify(0, ifp);
3071 * Note that section 3.1 of RFC 4429 indicates
3072 * that the Optimistic flag should not be set for
3073 * manually configured addresses
3075 addrconf_dad_start(ifp);
3076 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
3077 manage_tempaddrs(idev, ifp, cfg->valid_lft,
3078 cfg->preferred_lft, true, jiffies);
3080 addrconf_verify_rtnl(net);
3082 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
3083 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
3087 return PTR_ERR(ifp);
3090 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
3091 const struct in6_addr *pfx, unsigned int plen,
3092 struct netlink_ext_ack *extack)
3094 struct inet6_ifaddr *ifp;
3095 struct inet6_dev *idev;
3096 struct net_device *dev;
3099 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length");
3103 dev = __dev_get_by_index(net, ifindex);
3105 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
3109 idev = __in6_dev_get(dev);
3111 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device");
3115 read_lock_bh(&idev->lock);
3116 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3117 if (ifp->prefix_len == plen &&
3118 ipv6_addr_equal(pfx, &ifp->addr)) {
3120 read_unlock_bh(&idev->lock);
3122 if (!(ifp->flags & IFA_F_TEMPORARY) &&
3123 (ifa_flags & IFA_F_MANAGETEMPADDR))
3124 manage_tempaddrs(idev, ifp, 0, 0, false,
3127 addrconf_verify_rtnl(net);
3128 if (ipv6_addr_is_multicast(pfx)) {
3129 ipv6_mc_config(net->ipv6.mc_autojoin_sk,
3130 false, pfx, dev->ifindex);
3135 read_unlock_bh(&idev->lock);
3137 NL_SET_ERR_MSG_MOD(extack, "address not found");
3138 return -EADDRNOTAVAIL;
3142 int addrconf_add_ifaddr(struct net *net, void __user *arg)
3144 struct ifa6_config cfg = {
3145 .ifa_flags = IFA_F_PERMANENT,
3146 .preferred_lft = INFINITY_LIFE_TIME,
3147 .valid_lft = INFINITY_LIFE_TIME,
3149 struct in6_ifreq ireq;
3152 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3155 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3158 cfg.pfx = &ireq.ifr6_addr;
3159 cfg.plen = ireq.ifr6_prefixlen;
3162 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
3167 int addrconf_del_ifaddr(struct net *net, void __user *arg)
3169 struct in6_ifreq ireq;
3172 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3175 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
3179 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
3180 ireq.ifr6_prefixlen, NULL);
3185 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
3186 int plen, int scope, u8 proto)
3188 struct inet6_ifaddr *ifp;
3189 struct ifa6_config cfg = {
3192 .ifa_flags = IFA_F_PERMANENT,
3193 .valid_lft = INFINITY_LIFE_TIME,
3194 .preferred_lft = INFINITY_LIFE_TIME,
3199 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3201 spin_lock_bh(&ifp->lock);
3202 ifp->flags &= ~IFA_F_TENTATIVE;
3203 spin_unlock_bh(&ifp->lock);
3204 rt_genid_bump_ipv6(dev_net(idev->dev));
3205 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3210 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3211 static void add_v4_addrs(struct inet6_dev *idev)
3213 struct in6_addr addr;
3214 struct net_device *dev;
3215 struct net *net = dev_net(idev->dev);
3216 int scope, plen, offset = 0;
3221 memset(&addr, 0, sizeof(struct in6_addr));
3222 /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
3223 if (idev->dev->addr_len == sizeof(struct in6_addr))
3224 offset = sizeof(struct in6_addr) - 4;
3225 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
3227 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) {
3228 scope = IPV6_ADDR_COMPATv4;
3230 pflags |= RTF_NONEXTHOP;
3232 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
3235 addr.s6_addr32[0] = htonl(0xfe800000);
3240 if (addr.s6_addr32[3]) {
3241 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
3242 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
3247 for_each_netdev(net, dev) {
3248 struct in_device *in_dev = __in_dev_get_rtnl(dev);
3249 if (in_dev && (dev->flags & IFF_UP)) {
3250 struct in_ifaddr *ifa;
3253 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
3254 addr.s6_addr32[3] = ifa->ifa_local;
3256 if (ifa->ifa_scope == RT_SCOPE_LINK)
3258 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
3259 if (idev->dev->flags&IFF_POINTOPOINT)
3264 add_addr(idev, &addr, plen, flag,
3266 addrconf_prefix_route(&addr, plen, 0, idev->dev,
3267 0, pflags, GFP_KERNEL);
3274 static void init_loopback(struct net_device *dev)
3276 struct inet6_dev *idev;
3282 idev = ipv6_find_idev(dev);
3284 pr_debug("%s: add_dev failed\n", __func__);
3288 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
3291 void addrconf_add_linklocal(struct inet6_dev *idev,
3292 const struct in6_addr *addr, u32 flags)
3294 struct ifa6_config cfg = {
3297 .ifa_flags = flags | IFA_F_PERMANENT,
3298 .valid_lft = INFINITY_LIFE_TIME,
3299 .preferred_lft = INFINITY_LIFE_TIME,
3301 .ifa_proto = IFAPROT_KERNEL_LL
3303 struct inet6_ifaddr *ifp;
3305 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3306 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) ||
3307 READ_ONCE(idev->cnf.optimistic_dad)) &&
3308 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
3309 cfg.ifa_flags |= IFA_F_OPTIMISTIC;
3312 ifp = ipv6_add_addr(idev, &cfg, true, NULL);
3314 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
3316 addrconf_dad_start(ifp);
3320 EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
3322 static bool ipv6_reserved_interfaceid(struct in6_addr address)
3324 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
3327 if (address.s6_addr32[2] == htonl(0x02005eff) &&
3328 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
3331 if (address.s6_addr32[2] == htonl(0xfdffffff) &&
3332 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
3338 static int ipv6_generate_stable_address(struct in6_addr *address,
3340 const struct inet6_dev *idev)
3342 static DEFINE_SPINLOCK(lock);
3343 static __u32 digest[SHA1_DIGEST_WORDS];
3344 static __u32 workspace[SHA1_WORKSPACE_WORDS];
3347 char __data[SHA1_BLOCK_SIZE];
3349 struct in6_addr secret;
3351 unsigned char hwaddr[MAX_ADDR_LEN];
3356 struct in6_addr secret;
3357 struct in6_addr temp;
3358 struct net *net = dev_net(idev->dev);
3360 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
3362 if (idev->cnf.stable_secret.initialized)
3363 secret = idev->cnf.stable_secret.secret;
3364 else if (net->ipv6.devconf_dflt->stable_secret.initialized)
3365 secret = net->ipv6.devconf_dflt->stable_secret.secret;
3370 spin_lock_bh(&lock);
3373 memset(&data, 0, sizeof(data));
3374 memset(workspace, 0, sizeof(workspace));
3375 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
3376 data.prefix[0] = address->s6_addr32[0];
3377 data.prefix[1] = address->s6_addr32[1];
3378 data.secret = secret;
3379 data.dad_count = dad_count;
3381 sha1_transform(digest, data.__data, workspace);
3384 temp.s6_addr32[2] = (__force __be32)digest[0];
3385 temp.s6_addr32[3] = (__force __be32)digest[1];
3387 spin_unlock_bh(&lock);
3389 if (ipv6_reserved_interfaceid(temp)) {
3391 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
3400 static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
3402 struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
3406 s = &idev->cnf.stable_secret;
3407 get_random_bytes(&s->secret, sizeof(s->secret));
3408 s->initialized = true;
3411 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
3413 struct in6_addr addr;
3415 /* no link local addresses on L3 master devices */
3416 if (netif_is_l3_master(idev->dev))
3419 /* no link local addresses on devices flagged as slaves */
3420 if (idev->dev->priv_flags & IFF_NO_ADDRCONF)
3423 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
3425 switch (idev->cnf.addr_gen_mode) {
3426 case IN6_ADDR_GEN_MODE_RANDOM:
3427 ipv6_gen_mode_random_init(idev);
3429 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
3430 if (!ipv6_generate_stable_address(&addr, 0, idev))
3431 addrconf_add_linklocal(idev, &addr,
3432 IFA_F_STABLE_PRIVACY);
3433 else if (prefix_route)
3434 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3437 case IN6_ADDR_GEN_MODE_EUI64:
3438 /* addrconf_add_linklocal also adds a prefix_route and we
3439 * only need to care about prefix routes if ipv6_generate_eui64
3440 * couldn't generate one.
3442 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
3443 addrconf_add_linklocal(idev, &addr, 0);
3444 else if (prefix_route)
3445 addrconf_prefix_route(&addr, 64, 0, idev->dev,
3448 case IN6_ADDR_GEN_MODE_NONE:
3450 /* will not add any link local address */
3455 static void addrconf_dev_config(struct net_device *dev)
3457 struct inet6_dev *idev;
3461 if ((dev->type != ARPHRD_ETHER) &&
3462 (dev->type != ARPHRD_FDDI) &&
3463 (dev->type != ARPHRD_ARCNET) &&
3464 (dev->type != ARPHRD_INFINIBAND) &&
3465 (dev->type != ARPHRD_IEEE1394) &&
3466 (dev->type != ARPHRD_TUNNEL6) &&
3467 (dev->type != ARPHRD_6LOWPAN) &&
3468 (dev->type != ARPHRD_TUNNEL) &&
3469 (dev->type != ARPHRD_NONE) &&
3470 (dev->type != ARPHRD_RAWIP)) {
3471 /* Alas, we support only Ethernet autoconfiguration. */
3472 idev = __in6_dev_get(dev);
3473 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
3474 dev->flags & IFF_MULTICAST)
3479 idev = addrconf_add_dev(dev);
3483 /* this device type has no EUI support */
3484 if (dev->type == ARPHRD_NONE &&
3485 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
3486 WRITE_ONCE(idev->cnf.addr_gen_mode,
3487 IN6_ADDR_GEN_MODE_RANDOM);
3489 addrconf_addr_gen(idev, false);
3492 #if IS_ENABLED(CONFIG_IPV6_SIT)
3493 static void addrconf_sit_config(struct net_device *dev)
3495 struct inet6_dev *idev;
3500 * Configure the tunnel with one of our IPv4
3501 * addresses... we should configure all of
3502 * our v4 addrs in the tunnel
3505 idev = ipv6_find_idev(dev);
3507 pr_debug("%s: add_dev failed\n", __func__);
3511 if (dev->priv_flags & IFF_ISATAP) {
3512 addrconf_addr_gen(idev, false);
3518 if (dev->flags&IFF_POINTOPOINT)
3519 addrconf_add_mroute(dev);
3523 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3524 static void addrconf_gre_config(struct net_device *dev)
3526 struct inet6_dev *idev;
3530 idev = ipv6_find_idev(dev);
3532 pr_debug("%s: add_dev failed\n", __func__);
3536 if (dev->type == ARPHRD_ETHER) {
3537 addrconf_addr_gen(idev, true);
3543 if (dev->flags & IFF_POINTOPOINT)
3544 addrconf_add_mroute(dev);
3548 static void addrconf_init_auto_addrs(struct net_device *dev)
3550 switch (dev->type) {
3551 #if IS_ENABLED(CONFIG_IPV6_SIT)
3553 addrconf_sit_config(dev);
3556 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
3559 addrconf_gre_config(dev);
3562 case ARPHRD_LOOPBACK:
3567 addrconf_dev_config(dev);
3572 static int fixup_permanent_addr(struct net *net,
3573 struct inet6_dev *idev,
3574 struct inet6_ifaddr *ifp)
3576 /* !fib6_node means the host route was removed from the
3577 * FIB, for example, if 'lo' device is taken down. In that
3578 * case regenerate the host route.
3580 if (!ifp->rt || !ifp->rt->fib6_node) {
3581 struct fib6_info *f6i, *prev;
3583 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
3586 return PTR_ERR(f6i);
3588 /* ifp->rt can be accessed outside of rtnl */
3589 spin_lock(&ifp->lock);
3592 spin_unlock(&ifp->lock);
3594 fib6_info_release(prev);
3597 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
3598 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
3599 ifp->rt_priority, idev->dev, 0, 0,
3603 if (ifp->state == INET6_IFADDR_STATE_PREDAD)
3604 addrconf_dad_start(ifp);
3609 static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
3611 struct inet6_ifaddr *ifp, *tmp;
3612 struct inet6_dev *idev;
3614 idev = __in6_dev_get(dev);
3618 write_lock_bh(&idev->lock);
3620 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
3621 if ((ifp->flags & IFA_F_PERMANENT) &&
3622 fixup_permanent_addr(net, idev, ifp) < 0) {
3623 write_unlock_bh(&idev->lock);
3626 write_lock_bh(&idev->lock);
3628 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
3629 idev->dev->name, &ifp->addr);
3633 write_unlock_bh(&idev->lock);
3636 static int addrconf_notify(struct notifier_block *this, unsigned long event,
3639 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3640 struct netdev_notifier_change_info *change_info;
3641 struct netdev_notifier_changeupper_info *info;
3642 struct inet6_dev *idev = __in6_dev_get(dev);
3643 struct net *net = dev_net(dev);
3644 int run_pending = 0;
3648 case NETDEV_REGISTER:
3649 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
3650 idev = ipv6_add_dev(dev);
3652 return notifier_from_errno(PTR_ERR(idev));
3656 case NETDEV_CHANGEMTU:
3657 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
3658 if (dev->mtu < IPV6_MIN_MTU) {
3659 addrconf_ifdown(dev, dev != net->loopback_dev);
3664 rt6_mtu_change(dev, dev->mtu);
3665 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3669 /* allocate new idev */
3670 idev = ipv6_add_dev(dev);
3674 /* device is still not ready */
3675 if (!(idev->if_flags & IF_READY))
3682 if (idev && idev->cnf.disable_ipv6)
3685 if (dev->priv_flags & IFF_NO_ADDRCONF) {
3686 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) &&
3687 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST)
3692 if (event == NETDEV_UP) {
3693 /* restore routes for permanent addresses */
3694 addrconf_permanent_addr(net, dev);
3696 if (!addrconf_link_ready(dev)) {
3697 /* device is not ready yet. */
3698 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
3703 if (!idev && dev->mtu >= IPV6_MIN_MTU)
3704 idev = ipv6_add_dev(dev);
3706 if (!IS_ERR_OR_NULL(idev)) {
3707 idev->if_flags |= IF_READY;
3710 } else if (event == NETDEV_CHANGE) {
3711 if (!addrconf_link_ready(dev)) {
3712 /* device is still not ready. */
3713 rt6_sync_down_dev(dev, event);
3717 if (!IS_ERR_OR_NULL(idev)) {
3718 if (idev->if_flags & IF_READY) {
3719 /* device is already configured -
3720 * but resend MLD reports, we might
3721 * have roamed and need to update
3722 * multicast snooping switches
3726 if (change_info->flags_changed & IFF_NOARP)
3727 addrconf_dad_run(idev, true);
3728 rt6_sync_up(dev, RTNH_F_LINKDOWN);
3731 idev->if_flags |= IF_READY;
3734 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
3740 addrconf_init_auto_addrs(dev);
3742 if (!IS_ERR_OR_NULL(idev)) {
3744 addrconf_dad_run(idev, false);
3746 /* Device has an address by now */
3747 rt6_sync_up(dev, RTNH_F_DEAD);
3750 * If the MTU changed during the interface down,
3751 * when the interface up, the changed MTU must be
3752 * reflected in the idev as well as routers.
3754 if (idev->cnf.mtu6 != dev->mtu &&
3755 dev->mtu >= IPV6_MIN_MTU) {
3756 rt6_mtu_change(dev, dev->mtu);
3757 WRITE_ONCE(idev->cnf.mtu6, dev->mtu);
3759 WRITE_ONCE(idev->tstamp, jiffies);
3760 inet6_ifinfo_notify(RTM_NEWLINK, idev);
3763 * If the changed mtu during down is lower than
3764 * IPV6_MIN_MTU stop IPv6 on this interface.
3766 if (dev->mtu < IPV6_MIN_MTU)
3767 addrconf_ifdown(dev, dev != net->loopback_dev);
3772 case NETDEV_UNREGISTER:
3774 * Remove all addresses from this interface.
3776 addrconf_ifdown(dev, event != NETDEV_DOWN);
3779 case NETDEV_CHANGENAME:
3781 snmp6_unregister_dev(idev);
3782 addrconf_sysctl_unregister(idev);
3783 err = addrconf_sysctl_register(idev);
3785 return notifier_from_errno(err);
3786 err = snmp6_register_dev(idev);
3788 addrconf_sysctl_unregister(idev);
3789 return notifier_from_errno(err);
3794 case NETDEV_PRE_TYPE_CHANGE:
3795 case NETDEV_POST_TYPE_CHANGE:
3797 addrconf_type_change(dev, event);
3800 case NETDEV_CHANGEUPPER:
3803 /* flush all routes if dev is linked to or unlinked from
3804 * an L3 master device (e.g., VRF)
3806 if (info->upper_dev && netif_is_l3_master(info->upper_dev))
3807 addrconf_ifdown(dev, false);
3814 * addrconf module should be notified of a device going up
3816 static struct notifier_block ipv6_dev_notf = {
3817 .notifier_call = addrconf_notify,
3818 .priority = ADDRCONF_NOTIFY_PRIORITY,
3821 static void addrconf_type_change(struct net_device *dev, unsigned long event)
3823 struct inet6_dev *idev;
3826 idev = __in6_dev_get(dev);
3828 if (event == NETDEV_POST_TYPE_CHANGE)
3829 ipv6_mc_remap(idev);
3830 else if (event == NETDEV_PRE_TYPE_CHANGE)
3831 ipv6_mc_unmap(idev);
3834 static bool addr_is_local(const struct in6_addr *addr)
3836 return ipv6_addr_type(addr) &
3837 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
3840 static int addrconf_ifdown(struct net_device *dev, bool unregister)
3842 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
3843 struct net *net = dev_net(dev);
3844 struct inet6_dev *idev;
3845 struct inet6_ifaddr *ifa;
3846 LIST_HEAD(tmp_addr_list);
3847 bool keep_addr = false;
3853 rt6_disable_ip(dev, event);
3855 idev = __in6_dev_get(dev);
3860 * Step 1: remove reference to ipv6 device from parent device.
3866 /* protected by rtnl_lock */
3867 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3869 /* Step 1.5: remove snmp6 entry */
3870 snmp6_unregister_dev(idev);
3874 /* combine the user config with event to determine if permanent
3875 * addresses are to be removed from address hash table
3877 if (!unregister && !idev->cnf.disable_ipv6) {
3878 /* aggregate the system setting and interface setting */
3879 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down);
3882 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down);
3884 keep_addr = (_keep_addr > 0);
3887 /* Step 2: clear hash table */
3888 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3889 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
3891 spin_lock_bh(&net->ipv6.addrconf_hash_lock);
3893 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3894 if (ifa->idev == idev) {
3895 addrconf_del_dad_work(ifa);
3896 /* combined flag + permanent flag decide if
3897 * address is retained on a down event
3900 !(ifa->flags & IFA_F_PERMANENT) ||
3901 addr_is_local(&ifa->addr)) {
3902 hlist_del_init_rcu(&ifa->addr_lst);
3907 spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
3910 write_lock_bh(&idev->lock);
3912 addrconf_del_rs_timer(idev);
3914 /* Step 2: clear flags for stateless addrconf, repeated down
3917 was_ready = idev->if_flags & IF_READY;
3919 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3921 /* Step 3: clear tempaddr list */
3922 while (!list_empty(&idev->tempaddr_list)) {
3923 ifa = list_first_entry(&idev->tempaddr_list,
3924 struct inet6_ifaddr, tmp_list);
3925 list_del(&ifa->tmp_list);
3926 write_unlock_bh(&idev->lock);
3927 spin_lock_bh(&ifa->lock);
3930 in6_ifa_put(ifa->ifpub);
3933 spin_unlock_bh(&ifa->lock);
3935 write_lock_bh(&idev->lock);
3938 list_for_each_entry(ifa, &idev->addr_list, if_list)
3939 list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
3940 write_unlock_bh(&idev->lock);
3942 while (!list_empty(&tmp_addr_list)) {
3943 struct fib6_info *rt = NULL;
3946 ifa = list_first_entry(&tmp_addr_list,
3947 struct inet6_ifaddr, if_list_aux);
3948 list_del(&ifa->if_list_aux);
3950 addrconf_del_dad_work(ifa);
3952 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
3953 !addr_is_local(&ifa->addr);
3955 spin_lock_bh(&ifa->lock);
3958 /* set state to skip the notifier below */
3959 state = INET6_IFADDR_STATE_DEAD;
3960 ifa->state = INET6_IFADDR_STATE_PREDAD;
3961 if (!(ifa->flags & IFA_F_NODAD))
3962 ifa->flags |= IFA_F_TENTATIVE;
3968 ifa->state = INET6_IFADDR_STATE_DEAD;
3971 spin_unlock_bh(&ifa->lock);
3974 ip6_del_rt(net, rt, false);
3976 if (state != INET6_IFADDR_STATE_DEAD) {
3977 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3978 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3980 if (idev->cnf.forwarding)
3981 addrconf_leave_anycast(ifa);
3982 addrconf_leave_solict(ifa->idev, &ifa->addr);
3986 write_lock_bh(&idev->lock);
3987 list_del_rcu(&ifa->if_list);
3988 write_unlock_bh(&idev->lock);
3993 /* Step 5: Discard anycast and multicast list */
3995 ipv6_ac_destroy_dev(idev);
3996 ipv6_mc_destroy_dev(idev);
3997 } else if (was_ready) {
4001 WRITE_ONCE(idev->tstamp, jiffies);
4004 /* Last: Shot the device (if unregistered) */
4006 addrconf_sysctl_unregister(idev);
4007 neigh_parms_release(&nd_tbl, idev->nd_parms);
4008 neigh_ifdown(&nd_tbl, dev);
4014 static void addrconf_rs_timer(struct timer_list *t)
4016 struct inet6_dev *idev = from_timer(idev, t, rs_timer);
4017 struct net_device *dev = idev->dev;
4018 struct in6_addr lladdr;
4021 write_lock(&idev->lock);
4022 if (idev->dead || !(idev->if_flags & IF_READY))
4025 if (!ipv6_accept_ra(idev))
4028 /* Announcement received after solicitation was sent */
4029 if (idev->if_flags & IF_RA_RCVD)
4032 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits);
4034 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) {
4035 write_unlock(&idev->lock);
4036 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4037 ndisc_send_rs(dev, &lladdr,
4038 &in6addr_linklocal_allrouters);
4042 write_lock(&idev->lock);
4043 idev->rs_interval = rfc3315_s14_backoff_update(
4045 READ_ONCE(idev->cnf.rtr_solicit_max_interval));
4046 /* The wait after the last probe can be shorter */
4047 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
4048 READ_ONCE(idev->cnf.rtr_solicits)) ?
4049 READ_ONCE(idev->cnf.rtr_solicit_delay) :
4053 * Note: we do not support deprecated "all on-link"
4054 * assumption any longer.
4056 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
4060 write_unlock(&idev->lock);
4066 * Duplicate Address Detection
4068 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
4070 struct inet6_dev *idev = ifp->idev;
4071 unsigned long rand_num;
4074 if (ifp->flags & IFA_F_OPTIMISTIC)
4077 rand_num = get_random_u32_below(
4078 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1);
4081 if (READ_ONCE(idev->cnf.enhanced_dad) ||
4082 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) {
4084 get_random_bytes(&nonce, 6);
4087 ifp->dad_nonce = nonce;
4088 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits);
4089 addrconf_mod_dad_work(ifp, rand_num);
4092 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
4094 struct inet6_dev *idev = ifp->idev;
4095 struct net_device *dev = idev->dev;
4096 bool bump_id, notify = false;
4099 addrconf_join_solict(dev, &ifp->addr);
4101 read_lock_bh(&idev->lock);
4102 spin_lock(&ifp->lock);
4103 if (ifp->state == INET6_IFADDR_STATE_DEAD)
4107 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
4108 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 &&
4109 READ_ONCE(idev->cnf.accept_dad) < 1) ||
4110 !(ifp->flags&IFA_F_TENTATIVE) ||
4111 ifp->flags & IFA_F_NODAD) {
4112 bool send_na = false;
4114 if (ifp->flags & IFA_F_TENTATIVE &&
4115 !(ifp->flags & IFA_F_OPTIMISTIC))
4117 bump_id = ifp->flags & IFA_F_TENTATIVE;
4118 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4119 spin_unlock(&ifp->lock);
4120 read_unlock_bh(&idev->lock);
4122 addrconf_dad_completed(ifp, bump_id, send_na);
4126 if (!(idev->if_flags & IF_READY)) {
4127 spin_unlock(&ifp->lock);
4128 read_unlock_bh(&idev->lock);
4130 * If the device is not ready:
4131 * - keep it tentative if it is a permanent address.
4132 * - otherwise, kill it.
4135 addrconf_dad_stop(ifp, 0);
4140 * Optimistic nodes can start receiving
4143 if (ifp->flags & IFA_F_OPTIMISTIC) {
4144 ip6_ins_rt(net, ifp->rt);
4145 if (ipv6_use_optimistic_addr(net, idev)) {
4146 /* Because optimistic nodes can use this address,
4147 * notify listeners. If DAD fails, RTM_DELADDR is sent.
4153 addrconf_dad_kick(ifp);
4155 spin_unlock(&ifp->lock);
4156 read_unlock_bh(&idev->lock);
4158 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4161 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
4163 bool begin_dad = false;
4165 spin_lock_bh(&ifp->lock);
4166 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
4167 ifp->state = INET6_IFADDR_STATE_PREDAD;
4170 spin_unlock_bh(&ifp->lock);
4173 addrconf_mod_dad_work(ifp, 0);
4176 static void addrconf_dad_work(struct work_struct *w)
4178 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
4179 struct inet6_ifaddr,
4181 struct inet6_dev *idev = ifp->idev;
4182 bool bump_id, disable_ipv6 = false;
4183 struct in6_addr mcaddr;
4189 } action = DAD_PROCESS;
4193 spin_lock_bh(&ifp->lock);
4194 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
4196 ifp->state = INET6_IFADDR_STATE_DAD;
4197 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
4199 ifp->state = INET6_IFADDR_STATE_POSTDAD;
4201 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->accept_dad) > 1 ||
4202 READ_ONCE(idev->cnf.accept_dad) > 1) &&
4203 !idev->cnf.disable_ipv6 &&
4204 !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
4205 struct in6_addr addr;
4207 addr.s6_addr32[0] = htonl(0xfe800000);
4208 addr.s6_addr32[1] = 0;
4210 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
4211 ipv6_addr_equal(&ifp->addr, &addr)) {
4212 /* DAD failed for link-local based on MAC */
4213 WRITE_ONCE(idev->cnf.disable_ipv6, 1);
4215 pr_info("%s: IPv6 being disabled!\n",
4216 ifp->idev->dev->name);
4217 disable_ipv6 = true;
4221 spin_unlock_bh(&ifp->lock);
4223 if (action == DAD_BEGIN) {
4224 addrconf_dad_begin(ifp);
4226 } else if (action == DAD_ABORT) {
4228 addrconf_dad_stop(ifp, 1);
4230 addrconf_ifdown(idev->dev, false);
4234 if (!ifp->dad_probes && addrconf_dad_end(ifp))
4237 write_lock_bh(&idev->lock);
4238 if (idev->dead || !(idev->if_flags & IF_READY)) {
4239 write_unlock_bh(&idev->lock);
4243 spin_lock(&ifp->lock);
4244 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
4245 spin_unlock(&ifp->lock);
4246 write_unlock_bh(&idev->lock);
4250 if (ifp->dad_probes == 0) {
4251 bool send_na = false;
4254 * DAD was successful
4257 if (ifp->flags & IFA_F_TENTATIVE &&
4258 !(ifp->flags & IFA_F_OPTIMISTIC))
4260 bump_id = ifp->flags & IFA_F_TENTATIVE;
4261 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
4262 spin_unlock(&ifp->lock);
4263 write_unlock_bh(&idev->lock);
4265 addrconf_dad_completed(ifp, bump_id, send_na);
4271 addrconf_mod_dad_work(ifp,
4272 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
4274 spin_unlock(&ifp->lock);
4275 write_unlock_bh(&idev->lock);
4277 /* send a neighbour solicitation for our addr */
4278 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
4279 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
4286 /* ifp->idev must be at least read locked */
4287 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
4289 struct inet6_ifaddr *ifpiter;
4290 struct inet6_dev *idev = ifp->idev;
4292 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
4293 if (ifpiter->scope > IFA_LINK)
4295 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
4296 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
4297 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
4304 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
4307 struct net_device *dev = ifp->idev->dev;
4308 struct in6_addr lladdr;
4309 bool send_rs, send_mld;
4311 addrconf_del_dad_work(ifp);
4314 * Configure the address for reception. Now it is valid.
4317 ipv6_ifa_notify(RTM_NEWADDR, ifp);
4319 /* If added prefix is link local and we are prepared to process
4320 router advertisements, start sending router solicitations.
4323 read_lock_bh(&ifp->idev->lock);
4324 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
4325 send_rs = send_mld &&
4326 ipv6_accept_ra(ifp->idev) &&
4327 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 &&
4328 (dev->flags & IFF_LOOPBACK) == 0 &&
4329 (dev->type != ARPHRD_TUNNEL) &&
4330 !netif_is_team_port(dev);
4331 read_unlock_bh(&ifp->idev->lock);
4333 /* While dad is in progress mld report's source address is in6_addrany.
4334 * Resend with proper ll now.
4337 ipv6_mc_dad_complete(ifp->idev);
4339 /* send unsolicited NA if enabled */
4341 (READ_ONCE(ifp->idev->cnf.ndisc_notify) ||
4342 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) {
4343 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
4344 /*router=*/ !!ifp->idev->cnf.forwarding,
4345 /*solicited=*/ false, /*override=*/ true,
4351 * If a host as already performed a random delay
4352 * [...] as part of DAD [...] there is no need
4353 * to delay again before sending the first RS
4355 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
4357 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
4359 write_lock_bh(&ifp->idev->lock);
4360 spin_lock(&ifp->lock);
4361 ifp->idev->rs_interval = rfc3315_s14_backoff_init(
4362 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval));
4363 ifp->idev->rs_probes = 1;
4364 ifp->idev->if_flags |= IF_RS_SENT;
4365 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
4366 spin_unlock(&ifp->lock);
4367 write_unlock_bh(&ifp->idev->lock);
4371 rt_genid_bump_ipv6(dev_net(dev));
4373 /* Make sure that a new temporary address will be created
4374 * before this temporary address becomes deprecated.
4376 if (ifp->flags & IFA_F_TEMPORARY)
4377 addrconf_verify_rtnl(dev_net(dev));
4380 static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
4382 struct inet6_ifaddr *ifp;
4384 read_lock_bh(&idev->lock);
4385 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4386 spin_lock(&ifp->lock);
4387 if ((ifp->flags & IFA_F_TENTATIVE &&
4388 ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
4390 ifp->state = INET6_IFADDR_STATE_PREDAD;
4391 addrconf_dad_kick(ifp);
4393 spin_unlock(&ifp->lock);
4395 read_unlock_bh(&idev->lock);
4398 #ifdef CONFIG_PROC_FS
4399 struct if6_iter_state {
4400 struct seq_net_private p;
4405 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
4407 struct if6_iter_state *state = seq->private;
4408 struct net *net = seq_file_net(seq);
4409 struct inet6_ifaddr *ifa = NULL;
4412 /* initial bucket if pos is 0 */
4418 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
4419 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
4421 /* sync with offset */
4422 if (p < state->offset) {
4429 /* prepare for next bucket */
4436 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
4437 struct inet6_ifaddr *ifa)
4439 struct if6_iter_state *state = seq->private;
4440 struct net *net = seq_file_net(seq);
4442 hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
4448 while (++state->bucket < IN6_ADDR_HSIZE) {
4449 hlist_for_each_entry_rcu(ifa,
4450 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
4458 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
4462 return if6_get_first(seq, *pos);
4465 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
4467 struct inet6_ifaddr *ifa;
4469 ifa = if6_get_next(seq, v);
4474 static void if6_seq_stop(struct seq_file *seq, void *v)
4480 static int if6_seq_show(struct seq_file *seq, void *v)
4482 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
4483 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
4485 ifp->idev->dev->ifindex,
4489 ifp->idev->dev->name);
4493 static const struct seq_operations if6_seq_ops = {
4494 .start = if6_seq_start,
4495 .next = if6_seq_next,
4496 .show = if6_seq_show,
4497 .stop = if6_seq_stop,
4500 static int __net_init if6_proc_net_init(struct net *net)
4502 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
4503 sizeof(struct if6_iter_state)))
4508 static void __net_exit if6_proc_net_exit(struct net *net)
4510 remove_proc_entry("if_inet6", net->proc_net);
4513 static struct pernet_operations if6_proc_net_ops = {
4514 .init = if6_proc_net_init,
4515 .exit = if6_proc_net_exit,
4518 int __init if6_proc_init(void)
4520 return register_pernet_subsys(&if6_proc_net_ops);
4523 void if6_proc_exit(void)
4525 unregister_pernet_subsys(&if6_proc_net_ops);
4527 #endif /* CONFIG_PROC_FS */
4529 #if IS_ENABLED(CONFIG_IPV6_MIP6)
4530 /* Check if address is a home address configured on any interface. */
4531 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
4533 unsigned int hash = inet6_addr_hash(net, addr);
4534 struct inet6_ifaddr *ifp = NULL;
4538 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4539 if (ipv6_addr_equal(&ifp->addr, addr) &&
4540 (ifp->flags & IFA_F_HOMEADDRESS)) {
4550 /* RFC6554 has some algorithm to avoid loops in segment routing by
4551 * checking if the segments contains any of a local interface address.
4555 * To detect loops in the SRH, a router MUST determine if the SRH
4556 * includes multiple addresses assigned to any interface on that router.
4557 * If such addresses appear more than once and are separated by at least
4558 * one address not assigned to that router.
4560 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
4561 unsigned char nsegs)
4563 const struct in6_addr *addr;
4564 int i, ret = 0, found = 0;
4565 struct inet6_ifaddr *ifp;
4566 bool separated = false;
4571 for (i = 0; i < nsegs; i++) {
4573 hash = inet6_addr_hash(net, addr);
4576 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
4578 if (ipv6_addr_equal(&ifp->addr, addr)) {
4585 if (found > 1 && separated) {
4602 * Periodic address status verification
4605 static void addrconf_verify_rtnl(struct net *net)
4607 unsigned long now, next, next_sec, next_sched;
4608 struct inet6_ifaddr *ifp;
4615 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
4617 cancel_delayed_work(&net->ipv6.addr_chk_work);
4619 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
4621 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
4624 /* When setting preferred_lft to a value not zero or
4625 * infinity, while valid_lft is infinity
4626 * IFA_F_PERMANENT has a non-infinity life time.
4628 if ((ifp->flags & IFA_F_PERMANENT) &&
4629 (ifp->prefered_lft == INFINITY_LIFE_TIME))
4632 spin_lock(&ifp->lock);
4633 /* We try to batch several events at once. */
4634 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
4636 if ((ifp->flags&IFA_F_TEMPORARY) &&
4637 !(ifp->flags&IFA_F_TENTATIVE) &&
4638 ifp->prefered_lft != INFINITY_LIFE_TIME &&
4639 !ifp->regen_count && ifp->ifpub) {
4640 /* This is a non-regenerated temporary addr. */
4642 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev);
4644 if (age + regen_advance >= ifp->prefered_lft) {
4645 struct inet6_ifaddr *ifpub = ifp->ifpub;
4646 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4647 next = ifp->tstamp + ifp->prefered_lft * HZ;
4651 in6_ifa_hold(ifpub);
4652 spin_unlock(&ifp->lock);
4654 spin_lock(&ifpub->lock);
4655 ifpub->regen_count = 0;
4656 spin_unlock(&ifpub->lock);
4657 rcu_read_unlock_bh();
4658 ipv6_create_tempaddr(ifpub, true);
4663 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
4664 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
4667 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
4668 age >= ifp->valid_lft) {
4669 spin_unlock(&ifp->lock);
4671 rcu_read_unlock_bh();
4675 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
4676 spin_unlock(&ifp->lock);
4678 } else if (age >= ifp->prefered_lft) {
4679 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
4682 if (!(ifp->flags&IFA_F_DEPRECATED)) {
4684 ifp->flags |= IFA_F_DEPRECATED;
4687 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
4688 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
4689 next = ifp->tstamp + ifp->valid_lft * HZ;
4691 spin_unlock(&ifp->lock);
4696 ipv6_ifa_notify(0, ifp);
4701 /* ifp->prefered_lft <= ifp->valid_lft */
4702 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
4703 next = ifp->tstamp + ifp->prefered_lft * HZ;
4704 spin_unlock(&ifp->lock);
4709 next_sec = round_jiffies_up(next);
4712 /* If rounded timeout is accurate enough, accept it. */
4713 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
4714 next_sched = next_sec;
4716 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
4717 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
4718 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
4720 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
4721 now, next, next_sec, next_sched);
4722 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
4723 rcu_read_unlock_bh();
4726 static void addrconf_verify_work(struct work_struct *w)
4728 struct net *net = container_of(to_delayed_work(w), struct net,
4729 ipv6.addr_chk_work);
4732 addrconf_verify_rtnl(net);
4736 static void addrconf_verify(struct net *net)
4738 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
4741 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
4742 struct in6_addr **peer_pfx)
4744 struct in6_addr *pfx = NULL;
4749 pfx = nla_data(addr);
4752 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
4754 pfx = nla_data(local);
4760 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
4761 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
4762 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
4763 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
4764 [IFA_FLAGS] = { .len = sizeof(u32) },
4765 [IFA_RT_PRIORITY] = { .len = sizeof(u32) },
4766 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
4767 [IFA_PROTO] = { .type = NLA_U8 },
4771 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
4772 struct netlink_ext_ack *extack)
4774 struct net *net = sock_net(skb->sk);
4775 struct ifaddrmsg *ifm;
4776 struct nlattr *tb[IFA_MAX+1];
4777 struct in6_addr *pfx, *peer_pfx;
4781 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4782 ifa_ipv6_policy, extack);
4786 ifm = nlmsg_data(nlh);
4787 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4791 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
4793 /* We ignore other flags so far. */
4794 ifa_flags &= IFA_F_MANAGETEMPADDR;
4796 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
4797 ifm->ifa_prefixlen, extack);
4800 static int modify_prefix_route(struct inet6_ifaddr *ifp,
4801 unsigned long expires, u32 flags,
4804 struct fib6_table *table;
4805 struct fib6_info *f6i;
4808 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4810 ifp->idev->dev, 0, RTF_DEFAULT, true);
4814 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
4815 if (f6i->fib6_metric != prio) {
4816 /* delete old one */
4817 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
4820 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
4822 ifp->rt_priority, ifp->idev->dev,
4823 expires, flags, GFP_KERNEL);
4825 table = f6i->fib6_table;
4826 spin_lock_bh(&table->tb6_lock);
4828 if (!(flags & RTF_EXPIRES)) {
4829 fib6_clean_expires(f6i);
4830 fib6_remove_gc_list(f6i);
4832 fib6_set_expires(f6i, expires);
4833 fib6_add_gc_list(f6i);
4836 spin_unlock_bh(&table->tb6_lock);
4838 fib6_info_release(f6i);
4844 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
4845 struct ifa6_config *cfg)
4849 unsigned long timeout;
4850 bool was_managetempaddr;
4851 bool had_prefixroute;
4852 bool new_peer = false;
4856 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
4859 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
4860 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
4863 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
4864 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
4866 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
4867 if (addrconf_finite_timeout(timeout)) {
4868 expires = jiffies_to_clock_t(timeout * HZ);
4869 cfg->valid_lft = timeout;
4870 flags = RTF_EXPIRES;
4874 cfg->ifa_flags |= IFA_F_PERMANENT;
4877 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
4878 if (addrconf_finite_timeout(timeout)) {
4880 cfg->ifa_flags |= IFA_F_DEPRECATED;
4881 cfg->preferred_lft = timeout;
4884 if (cfg->peer_pfx &&
4885 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
4886 if (!ipv6_addr_any(&ifp->peer_addr))
4887 cleanup_prefix_route(ifp, expires, true, true);
4891 spin_lock_bh(&ifp->lock);
4892 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
4893 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
4894 !(ifp->flags & IFA_F_NOPREFIXROUTE);
4895 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
4896 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
4897 IFA_F_NOPREFIXROUTE);
4898 ifp->flags |= cfg->ifa_flags;
4899 WRITE_ONCE(ifp->tstamp, jiffies);
4900 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft);
4901 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft);
4902 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto);
4904 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
4905 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority);
4908 ifp->peer_addr = *cfg->peer_pfx;
4910 spin_unlock_bh(&ifp->lock);
4911 if (!(ifp->flags&IFA_F_TENTATIVE))
4912 ipv6_ifa_notify(0, ifp);
4914 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
4917 if (had_prefixroute)
4918 rc = modify_prefix_route(ifp, expires, flags, false);
4920 /* prefix route could have been deleted; if so restore it */
4921 if (rc == -ENOENT) {
4922 addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
4923 ifp->rt_priority, ifp->idev->dev,
4924 expires, flags, GFP_KERNEL);
4927 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
4928 rc = modify_prefix_route(ifp, expires, flags, true);
4930 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
4931 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
4932 ifp->rt_priority, ifp->idev->dev,
4933 expires, flags, GFP_KERNEL);
4935 } else if (had_prefixroute) {
4936 enum cleanup_prefix_rt_t action;
4937 unsigned long rt_expires;
4939 write_lock_bh(&ifp->idev->lock);
4940 action = check_cleanup_prefix_route(ifp, &rt_expires);
4941 write_unlock_bh(&ifp->idev->lock);
4943 if (action != CLEANUP_PREFIX_RT_NOP) {
4944 cleanup_prefix_route(ifp, rt_expires,
4945 action == CLEANUP_PREFIX_RT_DEL, false);
4949 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
4950 if (was_managetempaddr &&
4951 !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
4953 cfg->preferred_lft = 0;
4955 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
4956 cfg->preferred_lft, !was_managetempaddr,
4960 addrconf_verify_rtnl(net);
4966 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
4967 struct netlink_ext_ack *extack)
4969 struct net *net = sock_net(skb->sk);
4970 struct ifaddrmsg *ifm;
4971 struct nlattr *tb[IFA_MAX+1];
4972 struct in6_addr *peer_pfx;
4973 struct inet6_ifaddr *ifa;
4974 struct net_device *dev;
4975 struct inet6_dev *idev;
4976 struct ifa6_config cfg;
4979 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
4980 ifa_ipv6_policy, extack);
4984 memset(&cfg, 0, sizeof(cfg));
4986 ifm = nlmsg_data(nlh);
4987 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
4991 cfg.peer_pfx = peer_pfx;
4992 cfg.plen = ifm->ifa_prefixlen;
4993 if (tb[IFA_RT_PRIORITY])
4994 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
4997 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
4999 cfg.valid_lft = INFINITY_LIFE_TIME;
5000 cfg.preferred_lft = INFINITY_LIFE_TIME;
5002 if (tb[IFA_CACHEINFO]) {
5003 struct ifa_cacheinfo *ci;
5005 ci = nla_data(tb[IFA_CACHEINFO]);
5006 cfg.valid_lft = ci->ifa_valid;
5007 cfg.preferred_lft = ci->ifa_prefered;
5010 dev = __dev_get_by_index(net, ifm->ifa_index);
5012 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface");
5017 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
5019 cfg.ifa_flags = ifm->ifa_flags;
5021 /* We ignore other flags so far. */
5022 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
5023 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
5024 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
5026 idev = ipv6_find_idev(dev);
5028 return PTR_ERR(idev);
5030 if (!ipv6_allow_optimistic_dad(net, idev))
5031 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
5033 if (cfg.ifa_flags & IFA_F_NODAD &&
5034 cfg.ifa_flags & IFA_F_OPTIMISTIC) {
5035 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
5039 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
5042 * It would be best to check for !NLM_F_CREATE here but
5043 * userspace already relies on not having to provide this.
5045 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
5048 if (nlh->nlmsg_flags & NLM_F_EXCL ||
5049 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
5050 NL_SET_ERR_MSG_MOD(extack, "address already assigned");
5053 err = inet6_addr_modify(net, ifa, &cfg);
5061 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
5062 u8 scope, int ifindex)
5064 struct ifaddrmsg *ifm;
5066 ifm = nlmsg_data(nlh);
5067 ifm->ifa_family = AF_INET6;
5068 ifm->ifa_prefixlen = prefixlen;
5069 ifm->ifa_flags = flags;
5070 ifm->ifa_scope = scope;
5071 ifm->ifa_index = ifindex;
5074 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
5075 unsigned long tstamp, u32 preferred, u32 valid)
5077 struct ifa_cacheinfo ci;
5079 ci.cstamp = cstamp_delta(cstamp);
5080 ci.tstamp = cstamp_delta(tstamp);
5081 ci.ifa_prefered = preferred;
5082 ci.ifa_valid = valid;
5084 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
5087 static inline int rt_scope(int ifa_scope)
5089 if (ifa_scope & IFA_HOST)
5090 return RT_SCOPE_HOST;
5091 else if (ifa_scope & IFA_LINK)
5092 return RT_SCOPE_LINK;
5093 else if (ifa_scope & IFA_SITE)
5094 return RT_SCOPE_SITE;
5096 return RT_SCOPE_UNIVERSE;
5099 static inline int inet6_ifaddr_msgsize(void)
5101 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
5102 + nla_total_size(16) /* IFA_LOCAL */
5103 + nla_total_size(16) /* IFA_ADDRESS */
5104 + nla_total_size(sizeof(struct ifa_cacheinfo))
5105 + nla_total_size(4) /* IFA_FLAGS */
5106 + nla_total_size(1) /* IFA_PROTO */
5107 + nla_total_size(4) /* IFA_RT_PRIORITY */;
5116 struct inet6_fill_args {
5123 enum addr_type_t type;
5126 static int inet6_fill_ifaddr(struct sk_buff *skb,
5127 const struct inet6_ifaddr *ifa,
5128 struct inet6_fill_args *args)
5130 struct nlmsghdr *nlh;
5131 u32 preferred, valid;
5132 u32 flags, priority;
5135 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5136 sizeof(struct ifaddrmsg), args->flags);
5140 flags = READ_ONCE(ifa->flags);
5141 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
5142 ifa->idev->dev->ifindex);
5144 if (args->netnsid >= 0 &&
5145 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
5148 preferred = READ_ONCE(ifa->prefered_lft);
5149 valid = READ_ONCE(ifa->valid_lft);
5151 if (!((flags & IFA_F_PERMANENT) &&
5152 (preferred == INFINITY_LIFE_TIME))) {
5153 if (preferred != INFINITY_LIFE_TIME) {
5154 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ;
5156 if (preferred > tval)
5160 if (valid != INFINITY_LIFE_TIME) {
5168 preferred = INFINITY_LIFE_TIME;
5169 valid = INFINITY_LIFE_TIME;
5172 if (!ipv6_addr_any(&ifa->peer_addr)) {
5173 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
5174 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
5177 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
5181 priority = READ_ONCE(ifa->rt_priority);
5182 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority))
5185 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp),
5186 preferred, valid) < 0)
5189 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0)
5192 proto = READ_ONCE(ifa->ifa_proto);
5193 if (proto && nla_put_u8(skb, IFA_PROTO, proto))
5196 nlmsg_end(skb, nlh);
5200 nlmsg_cancel(skb, nlh);
5204 static int inet6_fill_ifmcaddr(struct sk_buff *skb,
5205 const struct ifmcaddr6 *ifmca,
5206 struct inet6_fill_args *args)
5208 int ifindex = ifmca->idev->dev->ifindex;
5209 u8 scope = RT_SCOPE_UNIVERSE;
5210 struct nlmsghdr *nlh;
5212 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
5213 scope = RT_SCOPE_SITE;
5215 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5216 sizeof(struct ifaddrmsg), args->flags);
5220 if (args->netnsid >= 0 &&
5221 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5222 nlmsg_cancel(skb, nlh);
5226 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5227 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
5228 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp),
5229 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5230 nlmsg_cancel(skb, nlh);
5234 nlmsg_end(skb, nlh);
5238 static int inet6_fill_ifacaddr(struct sk_buff *skb,
5239 const struct ifacaddr6 *ifaca,
5240 struct inet6_fill_args *args)
5242 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
5243 int ifindex = dev ? dev->ifindex : 1;
5244 u8 scope = RT_SCOPE_UNIVERSE;
5245 struct nlmsghdr *nlh;
5247 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
5248 scope = RT_SCOPE_SITE;
5250 nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
5251 sizeof(struct ifaddrmsg), args->flags);
5255 if (args->netnsid >= 0 &&
5256 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
5257 nlmsg_cancel(skb, nlh);
5261 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
5262 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
5263 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp),
5264 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
5265 nlmsg_cancel(skb, nlh);
5269 nlmsg_end(skb, nlh);
5273 /* called with rcu_read_lock() */
5274 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb,
5275 struct netlink_callback *cb, int *s_ip_idx,
5276 struct inet6_fill_args *fillargs)
5278 const struct ifmcaddr6 *ifmca;
5279 const struct ifacaddr6 *ifaca;
5283 switch (fillargs->type) {
5284 case UNICAST_ADDR: {
5285 const struct inet6_ifaddr *ifa;
5286 fillargs->event = RTM_NEWADDR;
5288 /* unicast address incl. temp addr */
5289 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
5290 if (ip_idx < *s_ip_idx)
5292 err = inet6_fill_ifaddr(skb, ifa, fillargs);
5295 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5301 case MULTICAST_ADDR:
5302 fillargs->event = RTM_GETMULTICAST;
5304 /* multicast address */
5305 for (ifmca = rcu_dereference(idev->mc_list);
5307 ifmca = rcu_dereference(ifmca->next), ip_idx++) {
5308 if (ip_idx < *s_ip_idx)
5310 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
5316 fillargs->event = RTM_GETANYCAST;
5317 /* anycast address */
5318 for (ifaca = rcu_dereference(idev->ac_list); ifaca;
5319 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) {
5320 if (ip_idx < *s_ip_idx)
5322 err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
5330 *s_ip_idx = err ? ip_idx : 0;
5334 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
5335 struct inet6_fill_args *fillargs,
5336 struct net **tgt_net, struct sock *sk,
5337 struct netlink_callback *cb)
5339 struct netlink_ext_ack *extack = cb->extack;
5340 struct nlattr *tb[IFA_MAX+1];
5341 struct ifaddrmsg *ifm;
5344 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5345 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
5349 ifm = nlmsg_data(nlh);
5350 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5351 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
5355 fillargs->ifindex = ifm->ifa_index;
5356 if (fillargs->ifindex) {
5357 cb->answer_flags |= NLM_F_DUMP_FILTERED;
5358 fillargs->flags |= NLM_F_DUMP_FILTERED;
5361 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5362 ifa_ipv6_policy, extack);
5366 for (i = 0; i <= IFA_MAX; ++i) {
5370 if (i == IFA_TARGET_NETNSID) {
5373 fillargs->netnsid = nla_get_s32(tb[i]);
5374 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
5376 fillargs->netnsid = -1;
5377 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
5378 return PTR_ERR(net);
5382 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
5390 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
5391 enum addr_type_t type)
5393 struct net *tgt_net = sock_net(skb->sk);
5394 const struct nlmsghdr *nlh = cb->nlh;
5395 struct inet6_fill_args fillargs = {
5396 .portid = NETLINK_CB(cb->skb).portid,
5397 .seq = cb->nlh->nlmsg_seq,
5398 .flags = NLM_F_MULTI,
5403 unsigned long ifindex;
5405 } *ctx = (void *)cb->ctx;
5406 struct net_device *dev;
5407 struct inet6_dev *idev;
5411 if (cb->strict_check) {
5412 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
5418 if (fillargs.ifindex) {
5420 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
5423 idev = __in6_dev_get(dev);
5425 err = in6_dump_addrs(idev, skb, cb,
5432 cb->seq = inet6_base_seq(tgt_net);
5433 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
5434 idev = __in6_dev_get(dev);
5437 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx,
5444 if (fillargs.netnsid >= 0)
5450 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
5452 enum addr_type_t type = UNICAST_ADDR;
5454 return inet6_dump_addr(skb, cb, type);
5457 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
5459 enum addr_type_t type = MULTICAST_ADDR;
5461 return inet6_dump_addr(skb, cb, type);
5465 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
5467 enum addr_type_t type = ANYCAST_ADDR;
5469 return inet6_dump_addr(skb, cb, type);
5472 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
5473 const struct nlmsghdr *nlh,
5475 struct netlink_ext_ack *extack)
5477 struct ifaddrmsg *ifm;
5480 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5481 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
5485 if (!netlink_strict_get_check(skb))
5486 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
5487 ifa_ipv6_policy, extack);
5489 ifm = nlmsg_data(nlh);
5490 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
5491 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
5495 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
5496 ifa_ipv6_policy, extack);
5500 for (i = 0; i <= IFA_MAX; i++) {
5505 case IFA_TARGET_NETNSID:
5510 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
5518 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5519 struct netlink_ext_ack *extack)
5521 struct net *tgt_net = sock_net(in_skb->sk);
5522 struct inet6_fill_args fillargs = {
5523 .portid = NETLINK_CB(in_skb).portid,
5524 .seq = nlh->nlmsg_seq,
5525 .event = RTM_NEWADDR,
5529 struct ifaddrmsg *ifm;
5530 struct nlattr *tb[IFA_MAX+1];
5531 struct in6_addr *addr = NULL, *peer;
5532 struct net_device *dev = NULL;
5533 struct inet6_ifaddr *ifa;
5534 struct sk_buff *skb;
5537 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
5541 if (tb[IFA_TARGET_NETNSID]) {
5542 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
5544 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
5546 if (IS_ERR(tgt_net))
5547 return PTR_ERR(tgt_net);
5550 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
5555 ifm = nlmsg_data(nlh);
5557 dev = dev_get_by_index(tgt_net, ifm->ifa_index);
5559 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
5561 err = -EADDRNOTAVAIL;
5565 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
5571 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5573 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5574 WARN_ON(err == -EMSGSIZE);
5578 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
5583 if (fillargs.netnsid >= 0)
5589 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
5591 struct sk_buff *skb;
5592 struct net *net = dev_net(ifa->idev->dev);
5593 struct inet6_fill_args fillargs = {
5602 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
5606 err = inet6_fill_ifaddr(skb, ifa, &fillargs);
5608 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
5609 WARN_ON(err == -EMSGSIZE);
5613 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
5617 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
5620 static void ipv6_store_devconf(const struct ipv6_devconf *cnf,
5621 __s32 *array, int bytes)
5623 BUG_ON(bytes < (DEVCONF_MAX * 4));
5625 memset(array, 0, bytes);
5626 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding);
5627 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit);
5628 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6);
5629 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra);
5630 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects);
5631 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf);
5632 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits);
5633 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits);
5634 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
5635 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval));
5636 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
5637 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval));
5638 array[DEVCONF_RTR_SOLICIT_DELAY] =
5639 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay));
5640 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version);
5641 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
5642 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval));
5643 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
5644 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval));
5645 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr);
5646 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft);
5647 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft);
5648 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry);
5649 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor);
5650 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses);
5651 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr);
5652 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric);
5653 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] =
5654 READ_ONCE(cnf->accept_ra_min_hop_limit);
5655 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo);
5656 #ifdef CONFIG_IPV6_ROUTER_PREF
5657 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref);
5658 array[DEVCONF_RTR_PROBE_INTERVAL] =
5659 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval));
5660 #ifdef CONFIG_IPV6_ROUTE_INFO
5661 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] =
5662 READ_ONCE(cnf->accept_ra_rt_info_min_plen);
5663 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] =
5664 READ_ONCE(cnf->accept_ra_rt_info_max_plen);
5667 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp);
5668 array[DEVCONF_ACCEPT_SOURCE_ROUTE] =
5669 READ_ONCE(cnf->accept_source_route);
5670 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5671 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad);
5672 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic);
5674 #ifdef CONFIG_IPV6_MROUTE
5675 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
5677 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6);
5678 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad);
5679 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao);
5680 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify);
5681 array[DEVCONF_SUPPRESS_FRAG_NDISC] =
5682 READ_ONCE(cnf->suppress_frag_ndisc);
5683 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] =
5684 READ_ONCE(cnf->accept_ra_from_local);
5685 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu);
5686 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] =
5687 READ_ONCE(cnf->ignore_routes_with_linkdown);
5688 /* we omit DEVCONF_STABLE_SECRET for now */
5689 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only);
5690 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] =
5691 READ_ONCE(cnf->drop_unicast_in_l2_multicast);
5692 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na);
5693 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down);
5694 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled);
5695 #ifdef CONFIG_IPV6_SEG6_HMAC
5696 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac);
5698 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad);
5699 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode);
5700 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy);
5701 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass);
5702 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled);
5703 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled);
5704 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id);
5705 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide);
5706 array[DEVCONF_NDISC_EVICT_NOCARRIER] =
5707 READ_ONCE(cnf->ndisc_evict_nocarrier);
5708 array[DEVCONF_ACCEPT_UNTRACKED_NA] =
5709 READ_ONCE(cnf->accept_untracked_na);
5710 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft);
5713 static inline size_t inet6_ifla6_size(void)
5715 return nla_total_size(4) /* IFLA_INET6_FLAGS */
5716 + nla_total_size(sizeof(struct ifla_cacheinfo))
5717 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
5718 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
5719 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
5720 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
5721 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
5722 + nla_total_size(4) /* IFLA_INET6_RA_MTU */
5726 static inline size_t inet6_if_nlmsg_size(void)
5728 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5729 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
5730 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
5731 + nla_total_size(4) /* IFLA_MTU */
5732 + nla_total_size(4) /* IFLA_LINK */
5733 + nla_total_size(1) /* IFLA_OPERSTATE */
5734 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
5737 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
5741 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
5744 /* Use put_unaligned() because stats may not be aligned for u64. */
5745 put_unaligned(ICMP6_MIB_MAX, &stats[0]);
5746 for (i = 1; i < ICMP6_MIB_MAX; i++)
5747 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
5749 memset(&stats[ICMP6_MIB_MAX], 0, pad);
5752 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
5753 int bytes, size_t syncpoff)
5756 u64 buff[IPSTATS_MIB_MAX];
5757 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
5761 memset(buff, 0, sizeof(buff));
5762 buff[0] = IPSTATS_MIB_MAX;
5764 for_each_possible_cpu(c) {
5765 for (i = 1; i < IPSTATS_MIB_MAX; i++)
5766 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
5769 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
5770 memset(&stats[IPSTATS_MIB_MAX], 0, pad);
5773 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
5777 case IFLA_INET6_STATS:
5778 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
5779 offsetof(struct ipstats_mib, syncp));
5781 case IFLA_INET6_ICMP6STATS:
5782 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
5787 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
5788 u32 ext_filter_mask)
5790 struct ifla_cacheinfo ci;
5794 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags)))
5795 goto nla_put_failure;
5796 ci.max_reasm_len = IPV6_MAXPLEN;
5797 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp));
5798 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
5799 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
5800 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
5801 goto nla_put_failure;
5802 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
5804 goto nla_put_failure;
5805 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
5807 /* XXX - MC not implemented */
5809 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
5812 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
5814 goto nla_put_failure;
5815 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
5817 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
5819 goto nla_put_failure;
5820 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
5822 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
5824 goto nla_put_failure;
5825 read_lock_bh(&idev->lock);
5826 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
5827 read_unlock_bh(&idev->lock);
5829 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE,
5830 READ_ONCE(idev->cnf.addr_gen_mode)))
5831 goto nla_put_failure;
5833 ra_mtu = READ_ONCE(idev->ra_mtu);
5834 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu))
5835 goto nla_put_failure;
5843 static size_t inet6_get_link_af_size(const struct net_device *dev,
5844 u32 ext_filter_mask)
5846 if (!__in6_dev_get(dev))
5849 return inet6_ifla6_size();
5852 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
5853 u32 ext_filter_mask)
5855 struct inet6_dev *idev = __in6_dev_get(dev);
5860 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
5866 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
5867 struct netlink_ext_ack *extack)
5869 struct inet6_ifaddr *ifp;
5870 struct net_device *dev = idev->dev;
5871 bool clear_token, update_rs = false;
5872 struct in6_addr ll_addr;
5879 if (dev->flags & IFF_LOOPBACK) {
5880 NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
5884 if (dev->flags & IFF_NOARP) {
5885 NL_SET_ERR_MSG_MOD(extack,
5886 "Device does not do neighbour discovery");
5890 if (!ipv6_accept_ra(idev)) {
5891 NL_SET_ERR_MSG_MOD(extack,
5892 "Router advertisement is disabled on device");
5896 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) {
5897 NL_SET_ERR_MSG(extack,
5898 "Router solicitation is disabled on device");
5902 write_lock_bh(&idev->lock);
5904 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
5905 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
5907 write_unlock_bh(&idev->lock);
5909 clear_token = ipv6_addr_any(token);
5913 if (!idev->dead && (idev->if_flags & IF_READY) &&
5914 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
5915 IFA_F_OPTIMISTIC)) {
5916 /* If we're not ready, then normal ifup will take care
5917 * of this. Otherwise, we need to request our rs here.
5919 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
5924 write_lock_bh(&idev->lock);
5927 idev->if_flags |= IF_RS_SENT;
5928 idev->rs_interval = rfc3315_s14_backoff_init(
5929 READ_ONCE(idev->cnf.rtr_solicit_interval));
5930 idev->rs_probes = 1;
5931 addrconf_mod_rs_timer(idev, idev->rs_interval);
5934 /* Well, that's kinda nasty ... */
5935 list_for_each_entry(ifp, &idev->addr_list, if_list) {
5936 spin_lock(&ifp->lock);
5937 if (ifp->tokenized) {
5939 ifp->prefered_lft = 0;
5941 spin_unlock(&ifp->lock);
5944 write_unlock_bh(&idev->lock);
5945 inet6_ifinfo_notify(RTM_NEWLINK, idev);
5946 addrconf_verify_rtnl(dev_net(dev));
5950 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
5951 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 },
5952 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) },
5953 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT,
5955 "IFLA_INET6_RA_MTU can not be set" },
5958 static int check_addr_gen_mode(int mode)
5960 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
5961 mode != IN6_ADDR_GEN_MODE_NONE &&
5962 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5963 mode != IN6_ADDR_GEN_MODE_RANDOM)
5968 static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
5971 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
5972 !idev->cnf.stable_secret.initialized &&
5973 !net->ipv6.devconf_dflt->stable_secret.initialized)
5978 static int inet6_validate_link_af(const struct net_device *dev,
5979 const struct nlattr *nla,
5980 struct netlink_ext_ack *extack)
5982 struct nlattr *tb[IFLA_INET6_MAX + 1];
5983 struct inet6_dev *idev = NULL;
5987 idev = __in6_dev_get(dev);
5989 return -EAFNOSUPPORT;
5992 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
5993 inet6_af_policy, extack);
5997 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
6000 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6001 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6003 if (check_addr_gen_mode(mode) < 0)
6005 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
6012 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
6013 struct netlink_ext_ack *extack)
6015 struct inet6_dev *idev = __in6_dev_get(dev);
6016 struct nlattr *tb[IFLA_INET6_MAX + 1];
6020 return -EAFNOSUPPORT;
6022 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
6025 if (tb[IFLA_INET6_TOKEN]) {
6026 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
6032 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
6033 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
6035 WRITE_ONCE(idev->cnf.addr_gen_mode, mode);
6041 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
6042 u32 portid, u32 seq, int event, unsigned int flags)
6044 struct net_device *dev = idev->dev;
6045 struct ifinfomsg *hdr;
6046 struct nlmsghdr *nlh;
6047 int ifindex, iflink;
6050 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
6054 hdr = nlmsg_data(nlh);
6055 hdr->ifi_family = AF_INET6;
6057 hdr->ifi_type = dev->type;
6058 ifindex = READ_ONCE(dev->ifindex);
6059 hdr->ifi_index = ifindex;
6060 hdr->ifi_flags = dev_get_flags(dev);
6061 hdr->ifi_change = 0;
6063 iflink = dev_get_iflink(dev);
6064 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
6066 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
6067 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
6068 (ifindex != iflink &&
6069 nla_put_u32(skb, IFLA_LINK, iflink)) ||
6070 nla_put_u8(skb, IFLA_OPERSTATE,
6071 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN))
6072 goto nla_put_failure;
6073 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
6075 goto nla_put_failure;
6077 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
6078 goto nla_put_failure;
6080 nla_nest_end(skb, protoinfo);
6081 nlmsg_end(skb, nlh);
6085 nlmsg_cancel(skb, nlh);
6089 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
6090 struct netlink_ext_ack *extack)
6092 struct ifinfomsg *ifm;
6094 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
6095 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
6099 if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
6100 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
6104 ifm = nlmsg_data(nlh);
6105 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
6106 ifm->ifi_change || ifm->ifi_index) {
6107 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
6114 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
6116 struct net *net = sock_net(skb->sk);
6118 unsigned long ifindex;
6119 } *ctx = (void *)cb->ctx;
6120 struct net_device *dev;
6121 struct inet6_dev *idev;
6124 /* only requests using strict checking can pass data to
6125 * influence the dump
6127 if (cb->strict_check) {
6128 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
6136 for_each_netdev_dump(net, dev, ctx->ifindex) {
6137 idev = __in6_dev_get(dev);
6140 err = inet6_fill_ifinfo(skb, idev,
6141 NETLINK_CB(cb->skb).portid,
6143 RTM_NEWLINK, NLM_F_MULTI);
6152 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
6154 struct sk_buff *skb;
6155 struct net *net = dev_net(idev->dev);
6158 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
6162 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
6164 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
6165 WARN_ON(err == -EMSGSIZE);
6169 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
6173 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
6176 static inline size_t inet6_prefix_nlmsg_size(void)
6178 return NLMSG_ALIGN(sizeof(struct prefixmsg))
6179 + nla_total_size(sizeof(struct in6_addr))
6180 + nla_total_size(sizeof(struct prefix_cacheinfo));
6183 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
6184 struct prefix_info *pinfo, u32 portid, u32 seq,
6185 int event, unsigned int flags)
6187 struct prefixmsg *pmsg;
6188 struct nlmsghdr *nlh;
6189 struct prefix_cacheinfo ci;
6191 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
6195 pmsg = nlmsg_data(nlh);
6196 pmsg->prefix_family = AF_INET6;
6197 pmsg->prefix_pad1 = 0;
6198 pmsg->prefix_pad2 = 0;
6199 pmsg->prefix_ifindex = idev->dev->ifindex;
6200 pmsg->prefix_len = pinfo->prefix_len;
6201 pmsg->prefix_type = pinfo->type;
6202 pmsg->prefix_pad3 = 0;
6203 pmsg->prefix_flags = pinfo->flags;
6205 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
6206 goto nla_put_failure;
6207 ci.preferred_time = ntohl(pinfo->prefered);
6208 ci.valid_time = ntohl(pinfo->valid);
6209 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
6210 goto nla_put_failure;
6211 nlmsg_end(skb, nlh);
6215 nlmsg_cancel(skb, nlh);
6219 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
6220 struct prefix_info *pinfo)
6222 struct sk_buff *skb;
6223 struct net *net = dev_net(idev->dev);
6226 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
6230 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
6232 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
6233 WARN_ON(err == -EMSGSIZE);
6237 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
6241 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
6244 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6246 struct net *net = dev_net(ifp->idev->dev);
6251 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
6256 * If the address was optimistic we inserted the route at the
6257 * start of our DAD process, so we don't need to do it again.
6258 * If the device was taken down in the middle of the DAD
6259 * cycle there is a race where we could get here without a
6260 * host route, so nothing to insert. That will be fixed when
6261 * the device is brought up.
6263 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
6264 ip6_ins_rt(net, ifp->rt);
6265 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
6266 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
6267 &ifp->addr, ifp->idev->dev->name);
6270 if (ifp->idev->cnf.forwarding)
6271 addrconf_join_anycast(ifp);
6272 if (!ipv6_addr_any(&ifp->peer_addr))
6273 addrconf_prefix_route(&ifp->peer_addr, 128,
6274 ifp->rt_priority, ifp->idev->dev,
6278 if (ifp->idev->cnf.forwarding)
6279 addrconf_leave_anycast(ifp);
6280 addrconf_leave_solict(ifp->idev, &ifp->addr);
6281 if (!ipv6_addr_any(&ifp->peer_addr)) {
6282 struct fib6_info *rt;
6284 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
6285 ifp->idev->dev, 0, 0,
6288 ip6_del_rt(net, rt, false);
6291 ip6_del_rt(net, ifp->rt, false);
6294 rt_genid_bump_ipv6(net);
6297 atomic_inc(&net->ipv6.dev_addr_genid);
6300 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
6302 if (likely(ifp->idev->dead == 0))
6303 __ipv6_ifa_notify(event, ifp);
6306 #ifdef CONFIG_SYSCTL
6308 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
6309 void *buffer, size_t *lenp, loff_t *ppos)
6311 int *valp = ctl->data;
6314 struct ctl_table lctl;
6318 * ctl->data points to idev->cnf.forwarding, we should
6319 * not modify it until we get the rtnl lock.
6324 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6327 ret = addrconf_fixup_forwarding(ctl, valp, val);
6333 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
6334 void *buffer, size_t *lenp, loff_t *ppos)
6336 struct inet6_dev *idev = ctl->extra1;
6337 int min_mtu = IPV6_MIN_MTU;
6338 struct ctl_table lctl;
6341 lctl.extra1 = &min_mtu;
6342 lctl.extra2 = idev ? &idev->dev->mtu : NULL;
6344 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
6347 static void dev_disable_change(struct inet6_dev *idev)
6349 struct netdev_notifier_info info;
6351 if (!idev || !idev->dev)
6354 netdev_notifier_info_init(&info, idev->dev);
6355 if (idev->cnf.disable_ipv6)
6356 addrconf_notify(NULL, NETDEV_DOWN, &info);
6358 addrconf_notify(NULL, NETDEV_UP, &info);
6361 static void addrconf_disable_change(struct net *net, __s32 newf)
6363 struct net_device *dev;
6364 struct inet6_dev *idev;
6366 for_each_netdev(net, dev) {
6367 idev = __in6_dev_get(dev);
6369 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
6371 WRITE_ONCE(idev->cnf.disable_ipv6, newf);
6373 dev_disable_change(idev);
6378 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
6380 struct net *net = (struct net *)table->extra2;
6383 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
6384 WRITE_ONCE(*p, newf);
6388 if (!rtnl_trylock())
6389 return restart_syscall();
6392 WRITE_ONCE(*p, newf);
6394 if (p == &net->ipv6.devconf_all->disable_ipv6) {
6395 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf);
6396 addrconf_disable_change(net, newf);
6397 } else if ((!newf) ^ (!old))
6398 dev_disable_change((struct inet6_dev *)table->extra1);
6404 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
6405 void *buffer, size_t *lenp, loff_t *ppos)
6407 int *valp = ctl->data;
6410 struct ctl_table lctl;
6414 * ctl->data points to idev->cnf.disable_ipv6, we should
6415 * not modify it until we get the rtnl lock.
6420 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6423 ret = addrconf_disable_ipv6(ctl, valp, val);
6429 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
6430 void *buffer, size_t *lenp, loff_t *ppos)
6432 int *valp = ctl->data;
6437 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6440 if (write && old != new) {
6441 struct net *net = ctl->extra2;
6443 if (!rtnl_trylock())
6444 return restart_syscall();
6446 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
6447 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6448 NETCONFA_PROXY_NEIGH,
6449 NETCONFA_IFINDEX_DEFAULT,
6450 net->ipv6.devconf_dflt);
6451 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
6452 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6453 NETCONFA_PROXY_NEIGH,
6454 NETCONFA_IFINDEX_ALL,
6455 net->ipv6.devconf_all);
6457 struct inet6_dev *idev = ctl->extra1;
6459 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
6460 NETCONFA_PROXY_NEIGH,
6470 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
6471 void *buffer, size_t *lenp,
6476 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
6477 struct net *net = (struct net *)ctl->extra2;
6478 struct ctl_table tmp = {
6480 .maxlen = sizeof(new_val),
6484 if (!rtnl_trylock())
6485 return restart_syscall();
6487 new_val = *((u32 *)ctl->data);
6489 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
6494 if (check_addr_gen_mode(new_val) < 0) {
6500 if (check_stable_privacy(idev, net, new_val) < 0) {
6505 if (idev->cnf.addr_gen_mode != new_val) {
6506 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val);
6507 addrconf_init_auto_addrs(idev->dev);
6509 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
6510 struct net_device *dev;
6512 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val);
6513 for_each_netdev(net, dev) {
6514 idev = __in6_dev_get(dev);
6516 idev->cnf.addr_gen_mode != new_val) {
6517 WRITE_ONCE(idev->cnf.addr_gen_mode,
6519 addrconf_init_auto_addrs(idev->dev);
6524 WRITE_ONCE(*((u32 *)ctl->data), new_val);
6533 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
6534 void *buffer, size_t *lenp,
6538 struct in6_addr addr;
6539 char str[IPV6_MAX_STRLEN];
6540 struct ctl_table lctl = *ctl;
6541 struct net *net = ctl->extra2;
6542 struct ipv6_stable_secret *secret = ctl->data;
6544 if (&net->ipv6.devconf_all->stable_secret == ctl->data)
6547 lctl.maxlen = IPV6_MAX_STRLEN;
6550 if (!rtnl_trylock())
6551 return restart_syscall();
6553 if (!write && !secret->initialized) {
6558 err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
6559 if (err >= sizeof(str)) {
6564 err = proc_dostring(&lctl, write, buffer, lenp, ppos);
6568 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
6573 secret->initialized = true;
6574 secret->secret = addr;
6576 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
6577 struct net_device *dev;
6579 for_each_netdev(net, dev) {
6580 struct inet6_dev *idev = __in6_dev_get(dev);
6583 WRITE_ONCE(idev->cnf.addr_gen_mode,
6584 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6588 struct inet6_dev *idev = ctl->extra1;
6590 WRITE_ONCE(idev->cnf.addr_gen_mode,
6591 IN6_ADDR_GEN_MODE_STABLE_PRIVACY);
6601 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
6602 int write, void *buffer,
6606 int *valp = ctl->data;
6609 struct ctl_table lctl;
6612 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown
6613 * we should not modify it until we get the rtnl lock.
6618 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6621 ret = addrconf_fixup_linkdown(ctl, valp, val);
6628 void addrconf_set_nopolicy(struct rt6_info *rt, int action)
6632 rt->dst.flags |= DST_NOPOLICY;
6634 rt->dst.flags &= ~DST_NOPOLICY;
6639 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
6641 struct inet6_ifaddr *ifa;
6643 read_lock_bh(&idev->lock);
6644 list_for_each_entry(ifa, &idev->addr_list, if_list) {
6645 spin_lock(&ifa->lock);
6647 /* host routes only use builtin fib6_nh */
6648 struct fib6_nh *nh = ifa->rt->fib6_nh;
6652 ifa->rt->dst_nopolicy = val ? true : false;
6653 if (nh->rt6i_pcpu) {
6654 for_each_possible_cpu(cpu) {
6655 struct rt6_info **rtp;
6657 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
6658 addrconf_set_nopolicy(*rtp, val);
6663 spin_unlock(&ifa->lock);
6665 read_unlock_bh(&idev->lock);
6669 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
6671 struct net *net = (struct net *)ctl->extra2;
6672 struct inet6_dev *idev;
6674 if (valp == &net->ipv6.devconf_dflt->disable_policy) {
6675 WRITE_ONCE(*valp, val);
6679 if (!rtnl_trylock())
6680 return restart_syscall();
6682 WRITE_ONCE(*valp, val);
6684 if (valp == &net->ipv6.devconf_all->disable_policy) {
6685 struct net_device *dev;
6687 for_each_netdev(net, dev) {
6688 idev = __in6_dev_get(dev);
6690 addrconf_disable_policy_idev(idev, val);
6693 idev = (struct inet6_dev *)ctl->extra1;
6694 addrconf_disable_policy_idev(idev, val);
6701 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
6702 void *buffer, size_t *lenp, loff_t *ppos)
6704 int *valp = ctl->data;
6707 struct ctl_table lctl;
6712 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
6714 if (write && (*valp != val))
6715 ret = addrconf_disable_policy(ctl, valp, val);
6723 static int minus_one = -1;
6724 static const int two_five_five = 255;
6725 static u32 ioam6_if_id_max = U16_MAX;
6727 static const struct ctl_table addrconf_sysctl[] = {
6729 .procname = "forwarding",
6730 .data = &ipv6_devconf.forwarding,
6731 .maxlen = sizeof(int),
6733 .proc_handler = addrconf_sysctl_forward,
6736 .procname = "hop_limit",
6737 .data = &ipv6_devconf.hop_limit,
6738 .maxlen = sizeof(int),
6740 .proc_handler = proc_dointvec_minmax,
6741 .extra1 = (void *)SYSCTL_ONE,
6742 .extra2 = (void *)&two_five_five,
6746 .data = &ipv6_devconf.mtu6,
6747 .maxlen = sizeof(int),
6749 .proc_handler = addrconf_sysctl_mtu,
6752 .procname = "accept_ra",
6753 .data = &ipv6_devconf.accept_ra,
6754 .maxlen = sizeof(int),
6756 .proc_handler = proc_dointvec,
6759 .procname = "accept_redirects",
6760 .data = &ipv6_devconf.accept_redirects,
6761 .maxlen = sizeof(int),
6763 .proc_handler = proc_dointvec,
6766 .procname = "autoconf",
6767 .data = &ipv6_devconf.autoconf,
6768 .maxlen = sizeof(int),
6770 .proc_handler = proc_dointvec,
6773 .procname = "dad_transmits",
6774 .data = &ipv6_devconf.dad_transmits,
6775 .maxlen = sizeof(int),
6777 .proc_handler = proc_dointvec,
6780 .procname = "router_solicitations",
6781 .data = &ipv6_devconf.rtr_solicits,
6782 .maxlen = sizeof(int),
6784 .proc_handler = proc_dointvec_minmax,
6785 .extra1 = &minus_one,
6788 .procname = "router_solicitation_interval",
6789 .data = &ipv6_devconf.rtr_solicit_interval,
6790 .maxlen = sizeof(int),
6792 .proc_handler = proc_dointvec_jiffies,
6795 .procname = "router_solicitation_max_interval",
6796 .data = &ipv6_devconf.rtr_solicit_max_interval,
6797 .maxlen = sizeof(int),
6799 .proc_handler = proc_dointvec_jiffies,
6802 .procname = "router_solicitation_delay",
6803 .data = &ipv6_devconf.rtr_solicit_delay,
6804 .maxlen = sizeof(int),
6806 .proc_handler = proc_dointvec_jiffies,
6809 .procname = "force_mld_version",
6810 .data = &ipv6_devconf.force_mld_version,
6811 .maxlen = sizeof(int),
6813 .proc_handler = proc_dointvec,
6816 .procname = "mldv1_unsolicited_report_interval",
6818 &ipv6_devconf.mldv1_unsolicited_report_interval,
6819 .maxlen = sizeof(int),
6821 .proc_handler = proc_dointvec_ms_jiffies,
6824 .procname = "mldv2_unsolicited_report_interval",
6826 &ipv6_devconf.mldv2_unsolicited_report_interval,
6827 .maxlen = sizeof(int),
6829 .proc_handler = proc_dointvec_ms_jiffies,
6832 .procname = "use_tempaddr",
6833 .data = &ipv6_devconf.use_tempaddr,
6834 .maxlen = sizeof(int),
6836 .proc_handler = proc_dointvec,
6839 .procname = "temp_valid_lft",
6840 .data = &ipv6_devconf.temp_valid_lft,
6841 .maxlen = sizeof(int),
6843 .proc_handler = proc_dointvec,
6846 .procname = "temp_prefered_lft",
6847 .data = &ipv6_devconf.temp_prefered_lft,
6848 .maxlen = sizeof(int),
6850 .proc_handler = proc_dointvec,
6853 .procname = "regen_min_advance",
6854 .data = &ipv6_devconf.regen_min_advance,
6855 .maxlen = sizeof(int),
6857 .proc_handler = proc_dointvec,
6860 .procname = "regen_max_retry",
6861 .data = &ipv6_devconf.regen_max_retry,
6862 .maxlen = sizeof(int),
6864 .proc_handler = proc_dointvec,
6867 .procname = "max_desync_factor",
6868 .data = &ipv6_devconf.max_desync_factor,
6869 .maxlen = sizeof(int),
6871 .proc_handler = proc_dointvec,
6874 .procname = "max_addresses",
6875 .data = &ipv6_devconf.max_addresses,
6876 .maxlen = sizeof(int),
6878 .proc_handler = proc_dointvec,
6881 .procname = "accept_ra_defrtr",
6882 .data = &ipv6_devconf.accept_ra_defrtr,
6883 .maxlen = sizeof(int),
6885 .proc_handler = proc_dointvec,
6888 .procname = "ra_defrtr_metric",
6889 .data = &ipv6_devconf.ra_defrtr_metric,
6890 .maxlen = sizeof(u32),
6892 .proc_handler = proc_douintvec_minmax,
6893 .extra1 = (void *)SYSCTL_ONE,
6896 .procname = "accept_ra_min_hop_limit",
6897 .data = &ipv6_devconf.accept_ra_min_hop_limit,
6898 .maxlen = sizeof(int),
6900 .proc_handler = proc_dointvec,
6903 .procname = "accept_ra_min_lft",
6904 .data = &ipv6_devconf.accept_ra_min_lft,
6905 .maxlen = sizeof(int),
6907 .proc_handler = proc_dointvec,
6910 .procname = "accept_ra_pinfo",
6911 .data = &ipv6_devconf.accept_ra_pinfo,
6912 .maxlen = sizeof(int),
6914 .proc_handler = proc_dointvec,
6917 .procname = "ra_honor_pio_life",
6918 .data = &ipv6_devconf.ra_honor_pio_life,
6919 .maxlen = sizeof(u8),
6921 .proc_handler = proc_dou8vec_minmax,
6922 .extra1 = SYSCTL_ZERO,
6923 .extra2 = SYSCTL_ONE,
6925 #ifdef CONFIG_IPV6_ROUTER_PREF
6927 .procname = "accept_ra_rtr_pref",
6928 .data = &ipv6_devconf.accept_ra_rtr_pref,
6929 .maxlen = sizeof(int),
6931 .proc_handler = proc_dointvec,
6934 .procname = "router_probe_interval",
6935 .data = &ipv6_devconf.rtr_probe_interval,
6936 .maxlen = sizeof(int),
6938 .proc_handler = proc_dointvec_jiffies,
6940 #ifdef CONFIG_IPV6_ROUTE_INFO
6942 .procname = "accept_ra_rt_info_min_plen",
6943 .data = &ipv6_devconf.accept_ra_rt_info_min_plen,
6944 .maxlen = sizeof(int),
6946 .proc_handler = proc_dointvec,
6949 .procname = "accept_ra_rt_info_max_plen",
6950 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
6951 .maxlen = sizeof(int),
6953 .proc_handler = proc_dointvec,
6958 .procname = "proxy_ndp",
6959 .data = &ipv6_devconf.proxy_ndp,
6960 .maxlen = sizeof(int),
6962 .proc_handler = addrconf_sysctl_proxy_ndp,
6965 .procname = "accept_source_route",
6966 .data = &ipv6_devconf.accept_source_route,
6967 .maxlen = sizeof(int),
6969 .proc_handler = proc_dointvec,
6971 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
6973 .procname = "optimistic_dad",
6974 .data = &ipv6_devconf.optimistic_dad,
6975 .maxlen = sizeof(int),
6977 .proc_handler = proc_dointvec,
6980 .procname = "use_optimistic",
6981 .data = &ipv6_devconf.use_optimistic,
6982 .maxlen = sizeof(int),
6984 .proc_handler = proc_dointvec,
6987 #ifdef CONFIG_IPV6_MROUTE
6989 .procname = "mc_forwarding",
6990 .data = &ipv6_devconf.mc_forwarding,
6991 .maxlen = sizeof(int),
6993 .proc_handler = proc_dointvec,
6997 .procname = "disable_ipv6",
6998 .data = &ipv6_devconf.disable_ipv6,
6999 .maxlen = sizeof(int),
7001 .proc_handler = addrconf_sysctl_disable,
7004 .procname = "accept_dad",
7005 .data = &ipv6_devconf.accept_dad,
7006 .maxlen = sizeof(int),
7008 .proc_handler = proc_dointvec,
7011 .procname = "force_tllao",
7012 .data = &ipv6_devconf.force_tllao,
7013 .maxlen = sizeof(int),
7015 .proc_handler = proc_dointvec
7018 .procname = "ndisc_notify",
7019 .data = &ipv6_devconf.ndisc_notify,
7020 .maxlen = sizeof(int),
7022 .proc_handler = proc_dointvec
7025 .procname = "suppress_frag_ndisc",
7026 .data = &ipv6_devconf.suppress_frag_ndisc,
7027 .maxlen = sizeof(int),
7029 .proc_handler = proc_dointvec
7032 .procname = "accept_ra_from_local",
7033 .data = &ipv6_devconf.accept_ra_from_local,
7034 .maxlen = sizeof(int),
7036 .proc_handler = proc_dointvec,
7039 .procname = "accept_ra_mtu",
7040 .data = &ipv6_devconf.accept_ra_mtu,
7041 .maxlen = sizeof(int),
7043 .proc_handler = proc_dointvec,
7046 .procname = "stable_secret",
7047 .data = &ipv6_devconf.stable_secret,
7048 .maxlen = IPV6_MAX_STRLEN,
7050 .proc_handler = addrconf_sysctl_stable_secret,
7053 .procname = "use_oif_addrs_only",
7054 .data = &ipv6_devconf.use_oif_addrs_only,
7055 .maxlen = sizeof(int),
7057 .proc_handler = proc_dointvec,
7060 .procname = "ignore_routes_with_linkdown",
7061 .data = &ipv6_devconf.ignore_routes_with_linkdown,
7062 .maxlen = sizeof(int),
7064 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown,
7067 .procname = "drop_unicast_in_l2_multicast",
7068 .data = &ipv6_devconf.drop_unicast_in_l2_multicast,
7069 .maxlen = sizeof(int),
7071 .proc_handler = proc_dointvec,
7074 .procname = "drop_unsolicited_na",
7075 .data = &ipv6_devconf.drop_unsolicited_na,
7076 .maxlen = sizeof(int),
7078 .proc_handler = proc_dointvec,
7081 .procname = "keep_addr_on_down",
7082 .data = &ipv6_devconf.keep_addr_on_down,
7083 .maxlen = sizeof(int),
7085 .proc_handler = proc_dointvec,
7089 .procname = "seg6_enabled",
7090 .data = &ipv6_devconf.seg6_enabled,
7091 .maxlen = sizeof(int),
7093 .proc_handler = proc_dointvec,
7095 #ifdef CONFIG_IPV6_SEG6_HMAC
7097 .procname = "seg6_require_hmac",
7098 .data = &ipv6_devconf.seg6_require_hmac,
7099 .maxlen = sizeof(int),
7101 .proc_handler = proc_dointvec,
7105 .procname = "enhanced_dad",
7106 .data = &ipv6_devconf.enhanced_dad,
7107 .maxlen = sizeof(int),
7109 .proc_handler = proc_dointvec,
7112 .procname = "addr_gen_mode",
7113 .data = &ipv6_devconf.addr_gen_mode,
7114 .maxlen = sizeof(int),
7116 .proc_handler = addrconf_sysctl_addr_gen_mode,
7119 .procname = "disable_policy",
7120 .data = &ipv6_devconf.disable_policy,
7121 .maxlen = sizeof(int),
7123 .proc_handler = addrconf_sysctl_disable_policy,
7126 .procname = "ndisc_tclass",
7127 .data = &ipv6_devconf.ndisc_tclass,
7128 .maxlen = sizeof(int),
7130 .proc_handler = proc_dointvec_minmax,
7131 .extra1 = (void *)SYSCTL_ZERO,
7132 .extra2 = (void *)&two_five_five,
7135 .procname = "rpl_seg_enabled",
7136 .data = &ipv6_devconf.rpl_seg_enabled,
7137 .maxlen = sizeof(int),
7139 .proc_handler = proc_dointvec,
7142 .procname = "ioam6_enabled",
7143 .data = &ipv6_devconf.ioam6_enabled,
7144 .maxlen = sizeof(u8),
7146 .proc_handler = proc_dou8vec_minmax,
7147 .extra1 = (void *)SYSCTL_ZERO,
7148 .extra2 = (void *)SYSCTL_ONE,
7151 .procname = "ioam6_id",
7152 .data = &ipv6_devconf.ioam6_id,
7153 .maxlen = sizeof(u32),
7155 .proc_handler = proc_douintvec_minmax,
7156 .extra1 = (void *)SYSCTL_ZERO,
7157 .extra2 = (void *)&ioam6_if_id_max,
7160 .procname = "ioam6_id_wide",
7161 .data = &ipv6_devconf.ioam6_id_wide,
7162 .maxlen = sizeof(u32),
7164 .proc_handler = proc_douintvec,
7167 .procname = "ndisc_evict_nocarrier",
7168 .data = &ipv6_devconf.ndisc_evict_nocarrier,
7169 .maxlen = sizeof(u8),
7171 .proc_handler = proc_dou8vec_minmax,
7172 .extra1 = (void *)SYSCTL_ZERO,
7173 .extra2 = (void *)SYSCTL_ONE,
7176 .procname = "accept_untracked_na",
7177 .data = &ipv6_devconf.accept_untracked_na,
7178 .maxlen = sizeof(int),
7180 .proc_handler = proc_dointvec_minmax,
7181 .extra1 = SYSCTL_ZERO,
7182 .extra2 = SYSCTL_TWO,
7189 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
7190 struct inet6_dev *idev, struct ipv6_devconf *p)
7193 struct ctl_table *table;
7194 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
7196 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
7200 for (i = 0; table[i].data; i++) {
7201 table[i].data += (char *)p - (char *)&ipv6_devconf;
7202 /* If one of these is already set, then it is not safe to
7203 * overwrite either of them: this makes proc_dointvec_minmax
7206 if (!table[i].extra1 && !table[i].extra2) {
7207 table[i].extra1 = idev; /* embedded; no ref */
7208 table[i].extra2 = net;
7212 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
7214 p->sysctl_header = register_net_sysctl_sz(net, path, table,
7215 ARRAY_SIZE(addrconf_sysctl));
7216 if (!p->sysctl_header)
7219 if (!strcmp(dev_name, "all"))
7220 ifindex = NETCONFA_IFINDEX_ALL;
7221 else if (!strcmp(dev_name, "default"))
7222 ifindex = NETCONFA_IFINDEX_DEFAULT;
7224 ifindex = idev->dev->ifindex;
7225 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
7235 static void __addrconf_sysctl_unregister(struct net *net,
7236 struct ipv6_devconf *p, int ifindex)
7238 struct ctl_table *table;
7240 if (!p->sysctl_header)
7243 table = p->sysctl_header->ctl_table_arg;
7244 unregister_net_sysctl_table(p->sysctl_header);
7245 p->sysctl_header = NULL;
7248 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
7251 static int addrconf_sysctl_register(struct inet6_dev *idev)
7255 if (!sysctl_dev_name_is_allowed(idev->dev->name))
7258 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
7259 &ndisc_ifinfo_sysctl_change);
7262 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
7265 neigh_sysctl_unregister(idev->nd_parms);
7270 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
7272 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
7273 idev->dev->ifindex);
7274 neigh_sysctl_unregister(idev->nd_parms);
7280 static int __net_init addrconf_init_net(struct net *net)
7283 struct ipv6_devconf *all, *dflt;
7285 spin_lock_init(&net->ipv6.addrconf_hash_lock);
7286 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
7287 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
7288 sizeof(struct hlist_head),
7290 if (!net->ipv6.inet6_addr_lst)
7291 goto err_alloc_addr;
7293 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
7297 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
7299 goto err_alloc_dflt;
7301 if (!net_eq(net, &init_net)) {
7302 switch (net_inherit_devconf()) {
7303 case 1: /* copy from init_net */
7304 memcpy(all, init_net.ipv6.devconf_all,
7305 sizeof(ipv6_devconf));
7306 memcpy(dflt, init_net.ipv6.devconf_dflt,
7307 sizeof(ipv6_devconf_dflt));
7309 case 3: /* copy from the current netns */
7310 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
7311 sizeof(ipv6_devconf));
7313 current->nsproxy->net_ns->ipv6.devconf_dflt,
7314 sizeof(ipv6_devconf_dflt));
7318 /* use compiled values */
7323 /* these will be inherited by all namespaces */
7324 dflt->autoconf = ipv6_defaults.autoconf;
7325 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
7327 dflt->stable_secret.initialized = false;
7328 all->stable_secret.initialized = false;
7330 net->ipv6.devconf_all = all;
7331 net->ipv6.devconf_dflt = dflt;
7333 #ifdef CONFIG_SYSCTL
7334 err = __addrconf_sysctl_register(net, "all", NULL, all);
7338 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
7344 #ifdef CONFIG_SYSCTL
7346 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
7349 net->ipv6.devconf_dflt = NULL;
7353 net->ipv6.devconf_all = NULL;
7355 kfree(net->ipv6.inet6_addr_lst);
7360 static void __net_exit addrconf_exit_net(struct net *net)
7364 #ifdef CONFIG_SYSCTL
7365 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
7366 NETCONFA_IFINDEX_DEFAULT);
7367 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
7368 NETCONFA_IFINDEX_ALL);
7370 kfree(net->ipv6.devconf_dflt);
7371 net->ipv6.devconf_dflt = NULL;
7372 kfree(net->ipv6.devconf_all);
7373 net->ipv6.devconf_all = NULL;
7375 cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
7377 * Check hash table, then free it.
7379 for (i = 0; i < IN6_ADDR_HSIZE; i++)
7380 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
7382 kfree(net->ipv6.inet6_addr_lst);
7383 net->ipv6.inet6_addr_lst = NULL;
7386 static struct pernet_operations addrconf_ops = {
7387 .init = addrconf_init_net,
7388 .exit = addrconf_exit_net,
7391 static struct rtnl_af_ops inet6_ops __read_mostly = {
7393 .fill_link_af = inet6_fill_link_af,
7394 .get_link_af_size = inet6_get_link_af_size,
7395 .validate_link_af = inet6_validate_link_af,
7396 .set_link_af = inet6_set_link_af,
7400 * Init / cleanup code
7403 int __init addrconf_init(void)
7405 struct inet6_dev *idev;
7408 err = ipv6_addr_label_init();
7410 pr_crit("%s: cannot initialize default policy table: %d\n",
7415 err = register_pernet_subsys(&addrconf_ops);
7419 /* All works using addrconf_wq need to lock rtnl. */
7420 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf");
7427 idev = ipv6_add_dev(blackhole_netdev);
7430 err = PTR_ERR(idev);
7434 ip6_route_init_special_entries();
7436 register_netdevice_notifier(&ipv6_dev_notf);
7438 addrconf_verify(&init_net);
7440 rtnl_af_register(&inet6_ops);
7442 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
7443 NULL, inet6_dump_ifinfo, RTNL_FLAG_DUMP_UNLOCKED);
7447 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
7448 inet6_rtm_newaddr, NULL, 0);
7451 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
7452 inet6_rtm_deladdr, NULL, 0);
7455 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
7456 inet6_rtm_getaddr, inet6_dump_ifaddr,
7457 RTNL_FLAG_DOIT_UNLOCKED |
7458 RTNL_FLAG_DUMP_UNLOCKED);
7461 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
7462 NULL, inet6_dump_ifmcaddr,
7463 RTNL_FLAG_DUMP_UNLOCKED);
7466 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
7467 NULL, inet6_dump_ifacaddr,
7468 RTNL_FLAG_DUMP_UNLOCKED);
7471 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
7472 inet6_netconf_get_devconf,
7473 inet6_netconf_dump_devconf,
7474 RTNL_FLAG_DOIT_UNLOCKED |
7475 RTNL_FLAG_DUMP_UNLOCKED);
7478 err = ipv6_addr_label_rtnl_register();
7484 rtnl_unregister_all(PF_INET6);
7485 rtnl_af_unregister(&inet6_ops);
7486 unregister_netdevice_notifier(&ipv6_dev_notf);
7488 destroy_workqueue(addrconf_wq);
7490 unregister_pernet_subsys(&addrconf_ops);
7492 ipv6_addr_label_cleanup();
7497 void addrconf_cleanup(void)
7499 struct net_device *dev;
7501 unregister_netdevice_notifier(&ipv6_dev_notf);
7502 unregister_pernet_subsys(&addrconf_ops);
7503 ipv6_addr_label_cleanup();
7505 rtnl_af_unregister(&inet6_ops);
7509 /* clean dev list */
7510 for_each_netdev(&init_net, dev) {
7511 if (__in6_dev_get(dev) == NULL)
7513 addrconf_ifdown(dev, true);
7515 addrconf_ifdown(init_net.loopback_dev, true);
7519 destroy_workqueue(addrconf_wq);