2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
60 #include <linux/sysctl.h>
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
68 #include <net/net_namespace.h>
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
87 #ifdef CONFIG_IPV6_PRIVACY
88 #include <linux/random.h>
91 #include <linux/uaccess.h>
92 #include <asm/unaligned.h>
94 #include <linux/proc_fs.h>
95 #include <linux/seq_file.h>
96 #include <linux/export.h>
98 /* Set to 3 to get tracing... */
102 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
104 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
107 #define INFINITY_LIFE_TIME 0xFFFFFFFF
109 static inline u32 cstamp_delta(unsigned long cstamp)
111 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
115 static void addrconf_sysctl_register(struct inet6_dev *idev);
116 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
118 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
122 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
127 #ifdef CONFIG_IPV6_PRIVACY
128 static void __ipv6_regen_rndid(struct inet6_dev *idev);
129 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
130 static void ipv6_regen_rndid(unsigned long data);
133 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
134 static int ipv6_count_addresses(struct inet6_dev *idev);
137 * Configured unicast address hash table
139 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
140 static DEFINE_SPINLOCK(addrconf_hash_lock);
142 static void addrconf_verify(unsigned long);
144 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
145 static DEFINE_SPINLOCK(addrconf_verify_lock);
147 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
148 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
150 static void addrconf_type_change(struct net_device *dev,
151 unsigned long event);
152 static int addrconf_ifdown(struct net_device *dev, int how);
154 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
156 const struct net_device *dev,
157 u32 flags, u32 noflags);
159 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
160 static void addrconf_dad_timer(unsigned long data);
161 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
162 static void addrconf_dad_run(struct inet6_dev *idev);
163 static void addrconf_rs_timer(unsigned long data);
164 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
165 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
167 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
168 struct prefix_info *pinfo);
169 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
170 struct net_device *dev);
172 static struct ipv6_devconf ipv6_devconf __read_mostly = {
174 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
175 .mtu6 = IPV6_MIN_MTU,
177 .accept_redirects = 1,
179 .force_mld_version = 0,
180 .mldv1_unsolicited_report_interval = 10 * HZ,
181 .mldv2_unsolicited_report_interval = HZ,
183 .rtr_solicits = MAX_RTR_SOLICITATIONS,
184 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
185 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
186 #ifdef CONFIG_IPV6_PRIVACY
188 .temp_valid_lft = TEMP_VALID_LIFETIME,
189 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
190 .regen_max_retry = REGEN_MAX_RETRY,
191 .max_desync_factor = MAX_DESYNC_FACTOR,
193 .max_addresses = IPV6_MAX_ADDRESSES,
194 .accept_ra_defrtr = 1,
195 .accept_ra_pinfo = 1,
196 #ifdef CONFIG_IPV6_ROUTER_PREF
197 .accept_ra_rtr_pref = 1,
198 .rtr_probe_interval = 60 * HZ,
199 #ifdef CONFIG_IPV6_ROUTE_INFO
200 .accept_ra_rt_info_max_plen = 0,
204 .accept_source_route = 0, /* we do not accept RH0 by default. */
207 .suppress_frag_ndisc = 1,
210 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
212 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
213 .mtu6 = IPV6_MIN_MTU,
215 .accept_redirects = 1,
217 .force_mld_version = 0,
218 .mldv1_unsolicited_report_interval = 10 * HZ,
219 .mldv2_unsolicited_report_interval = HZ,
221 .rtr_solicits = MAX_RTR_SOLICITATIONS,
222 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
223 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
224 #ifdef CONFIG_IPV6_PRIVACY
226 .temp_valid_lft = TEMP_VALID_LIFETIME,
227 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
228 .regen_max_retry = REGEN_MAX_RETRY,
229 .max_desync_factor = MAX_DESYNC_FACTOR,
231 .max_addresses = IPV6_MAX_ADDRESSES,
232 .accept_ra_defrtr = 1,
233 .accept_ra_pinfo = 1,
234 #ifdef CONFIG_IPV6_ROUTER_PREF
235 .accept_ra_rtr_pref = 1,
236 .rtr_probe_interval = 60 * HZ,
237 #ifdef CONFIG_IPV6_ROUTE_INFO
238 .accept_ra_rt_info_max_plen = 0,
242 .accept_source_route = 0, /* we do not accept RH0 by default. */
245 .suppress_frag_ndisc = 1,
248 /* Check if a valid qdisc is available */
249 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
251 return !qdisc_tx_is_noop(dev);
254 static void addrconf_del_rs_timer(struct inet6_dev *idev)
256 if (del_timer(&idev->rs_timer))
260 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp)
262 if (del_timer(&ifp->dad_timer))
266 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
269 if (!timer_pending(&idev->rs_timer))
271 mod_timer(&idev->rs_timer, jiffies + when);
274 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp,
277 if (!timer_pending(&ifp->dad_timer))
279 mod_timer(&ifp->dad_timer, jiffies + when);
282 static int snmp6_alloc_dev(struct inet6_dev *idev)
284 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
285 sizeof(struct ipstats_mib),
286 __alignof__(struct ipstats_mib)) < 0)
288 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
290 if (!idev->stats.icmpv6dev)
292 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
294 if (!idev->stats.icmpv6msgdev)
300 kfree(idev->stats.icmpv6dev);
302 snmp_mib_free((void __percpu **)idev->stats.ipv6);
307 static void snmp6_free_dev(struct inet6_dev *idev)
309 kfree(idev->stats.icmpv6msgdev);
310 kfree(idev->stats.icmpv6dev);
311 snmp_mib_free((void __percpu **)idev->stats.ipv6);
314 /* Nobody refers to this device, we may destroy it. */
316 void in6_dev_finish_destroy(struct inet6_dev *idev)
318 struct net_device *dev = idev->dev;
320 WARN_ON(!list_empty(&idev->addr_list));
321 WARN_ON(idev->mc_list != NULL);
322 WARN_ON(timer_pending(&idev->rs_timer));
324 #ifdef NET_REFCNT_DEBUG
325 pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
329 pr_warn("Freeing alive inet6 device %p\n", idev);
332 snmp6_free_dev(idev);
333 kfree_rcu(idev, rcu);
335 EXPORT_SYMBOL(in6_dev_finish_destroy);
337 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
339 struct inet6_dev *ndev;
343 if (dev->mtu < IPV6_MIN_MTU)
346 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
351 rwlock_init(&ndev->lock);
353 INIT_LIST_HEAD(&ndev->addr_list);
354 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
355 (unsigned long)ndev);
356 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
357 ndev->cnf.mtu6 = dev->mtu;
358 ndev->cnf.sysctl = NULL;
359 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
360 if (ndev->nd_parms == NULL) {
364 if (ndev->cnf.forwarding)
365 dev_disable_lro(dev);
366 /* We refer to the device */
369 if (snmp6_alloc_dev(ndev) < 0) {
371 "%s: cannot allocate memory for statistics; dev=%s.\n",
372 __func__, dev->name);
373 neigh_parms_release(&nd_tbl, ndev->nd_parms);
379 if (snmp6_register_dev(ndev) < 0) {
381 "%s: cannot create /proc/net/dev_snmp6/%s\n",
382 __func__, dev->name);
383 neigh_parms_release(&nd_tbl, ndev->nd_parms);
385 in6_dev_finish_destroy(ndev);
389 /* One reference from device. We must do this before
390 * we invoke __ipv6_regen_rndid().
394 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
395 ndev->cnf.accept_dad = -1;
397 #if IS_ENABLED(CONFIG_IPV6_SIT)
398 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
399 pr_info("%s: Disabled Multicast RS\n", dev->name);
400 ndev->cnf.rtr_solicits = 0;
404 #ifdef CONFIG_IPV6_PRIVACY
405 INIT_LIST_HEAD(&ndev->tempaddr_list);
406 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
407 if ((dev->flags&IFF_LOOPBACK) ||
408 dev->type == ARPHRD_TUNNEL ||
409 dev->type == ARPHRD_TUNNEL6 ||
410 dev->type == ARPHRD_SIT ||
411 dev->type == ARPHRD_NONE) {
412 ndev->cnf.use_tempaddr = -1;
415 ipv6_regen_rndid((unsigned long) ndev);
418 ndev->token = in6addr_any;
420 if (netif_running(dev) && addrconf_qdisc_ok(dev))
421 ndev->if_flags |= IF_READY;
423 ipv6_mc_init_dev(ndev);
424 ndev->tstamp = jiffies;
425 addrconf_sysctl_register(ndev);
426 /* protected by rtnl_lock */
427 rcu_assign_pointer(dev->ip6_ptr, ndev);
429 /* Join interface-local all-node multicast group */
430 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
432 /* Join all-node multicast group */
433 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
435 /* Join all-router multicast group if forwarding is set */
436 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
437 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
442 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
444 struct inet6_dev *idev;
448 idev = __in6_dev_get(dev);
450 idev = ipv6_add_dev(dev);
455 if (dev->flags&IFF_UP)
460 static int inet6_netconf_msgsize_devconf(int type)
462 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
463 + nla_total_size(4); /* NETCONFA_IFINDEX */
465 /* type -1 is used for ALL */
466 if (type == -1 || type == NETCONFA_FORWARDING)
467 size += nla_total_size(4);
468 #ifdef CONFIG_IPV6_MROUTE
469 if (type == -1 || type == NETCONFA_MC_FORWARDING)
470 size += nla_total_size(4);
476 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
477 struct ipv6_devconf *devconf, u32 portid,
478 u32 seq, int event, unsigned int flags,
481 struct nlmsghdr *nlh;
482 struct netconfmsg *ncm;
484 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
489 ncm = nlmsg_data(nlh);
490 ncm->ncm_family = AF_INET6;
492 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
493 goto nla_put_failure;
495 /* type -1 is used for ALL */
496 if ((type == -1 || type == NETCONFA_FORWARDING) &&
497 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
498 goto nla_put_failure;
499 #ifdef CONFIG_IPV6_MROUTE
500 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
501 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
502 devconf->mc_forwarding) < 0)
503 goto nla_put_failure;
505 return nlmsg_end(skb, nlh);
508 nlmsg_cancel(skb, nlh);
512 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
513 struct ipv6_devconf *devconf)
518 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
522 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
523 RTM_NEWNETCONF, 0, type);
525 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
526 WARN_ON(err == -EMSGSIZE);
530 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
533 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
536 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
537 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
538 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
541 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
542 struct nlmsghdr *nlh)
544 struct net *net = sock_net(in_skb->sk);
545 struct nlattr *tb[NETCONFA_MAX+1];
546 struct netconfmsg *ncm;
548 struct ipv6_devconf *devconf;
549 struct inet6_dev *in6_dev;
550 struct net_device *dev;
554 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
555 devconf_ipv6_policy);
560 if (!tb[NETCONFA_IFINDEX])
563 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
565 case NETCONFA_IFINDEX_ALL:
566 devconf = net->ipv6.devconf_all;
568 case NETCONFA_IFINDEX_DEFAULT:
569 devconf = net->ipv6.devconf_dflt;
572 dev = __dev_get_by_index(net, ifindex);
575 in6_dev = __in6_dev_get(dev);
578 devconf = &in6_dev->cnf;
583 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
587 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
588 NETLINK_CB(in_skb).portid,
589 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
592 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
593 WARN_ON(err == -EMSGSIZE);
597 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
602 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
603 struct netlink_callback *cb)
605 struct net *net = sock_net(skb->sk);
608 struct net_device *dev;
609 struct inet6_dev *idev;
610 struct hlist_head *head;
613 s_idx = idx = cb->args[1];
615 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
617 head = &net->dev_index_head[h];
619 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
621 hlist_for_each_entry_rcu(dev, head, index_hlist) {
624 idev = __in6_dev_get(dev);
628 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
630 NETLINK_CB(cb->skb).portid,
638 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
644 if (h == NETDEV_HASHENTRIES) {
645 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
646 net->ipv6.devconf_all,
647 NETLINK_CB(cb->skb).portid,
649 RTM_NEWNETCONF, NLM_F_MULTI,
655 if (h == NETDEV_HASHENTRIES + 1) {
656 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
657 net->ipv6.devconf_dflt,
658 NETLINK_CB(cb->skb).portid,
660 RTM_NEWNETCONF, NLM_F_MULTI,
674 static void dev_forward_change(struct inet6_dev *idev)
676 struct net_device *dev;
677 struct inet6_ifaddr *ifa;
682 if (idev->cnf.forwarding)
683 dev_disable_lro(dev);
684 if (dev->flags & IFF_MULTICAST) {
685 if (idev->cnf.forwarding) {
686 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
687 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
688 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
690 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
691 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
692 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
696 list_for_each_entry(ifa, &idev->addr_list, if_list) {
697 if (ifa->flags&IFA_F_TENTATIVE)
699 if (idev->cnf.forwarding)
700 addrconf_join_anycast(ifa);
702 addrconf_leave_anycast(ifa);
704 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
705 dev->ifindex, &idev->cnf);
709 static void addrconf_forward_change(struct net *net, __s32 newf)
711 struct net_device *dev;
712 struct inet6_dev *idev;
714 for_each_netdev(net, dev) {
715 idev = __in6_dev_get(dev);
717 int changed = (!idev->cnf.forwarding) ^ (!newf);
718 idev->cnf.forwarding = newf;
720 dev_forward_change(idev);
725 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
731 return restart_syscall();
733 net = (struct net *)table->extra2;
737 if (p == &net->ipv6.devconf_dflt->forwarding) {
738 if ((!newf) ^ (!old))
739 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
740 NETCONFA_IFINDEX_DEFAULT,
741 net->ipv6.devconf_dflt);
746 if (p == &net->ipv6.devconf_all->forwarding) {
747 net->ipv6.devconf_dflt->forwarding = newf;
748 addrconf_forward_change(net, newf);
749 if ((!newf) ^ (!old))
750 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
751 NETCONFA_IFINDEX_ALL,
752 net->ipv6.devconf_all);
753 } else if ((!newf) ^ (!old))
754 dev_forward_change((struct inet6_dev *)table->extra1);
758 rt6_purge_dflt_routers(net);
763 /* Nobody refers to this ifaddr, destroy it */
764 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
766 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
768 #ifdef NET_REFCNT_DEBUG
769 pr_debug("%s\n", __func__);
772 in6_dev_put(ifp->idev);
774 if (del_timer(&ifp->dad_timer))
775 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
777 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
778 pr_warn("Freeing alive inet6 address %p\n", ifp);
787 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
790 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
793 * Each device address list is sorted in order of scope -
794 * global before linklocal.
796 list_for_each(p, &idev->addr_list) {
797 struct inet6_ifaddr *ifa
798 = list_entry(p, struct inet6_ifaddr, if_list);
799 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
803 list_add_tail(&ifp->if_list, p);
806 static u32 inet6_addr_hash(const struct in6_addr *addr)
808 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
811 /* On success it returns ifp with increased reference count */
813 static struct inet6_ifaddr *
814 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
815 const struct in6_addr *peer_addr, int pfxlen,
816 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
818 struct inet6_ifaddr *ifa = NULL;
822 int addr_type = ipv6_addr_type(addr);
824 if (addr_type == IPV6_ADDR_ANY ||
825 addr_type & IPV6_ADDR_MULTICAST ||
826 (!(idev->dev->flags & IFF_LOOPBACK) &&
827 addr_type & IPV6_ADDR_LOOPBACK))
828 return ERR_PTR(-EADDRNOTAVAIL);
832 err = -ENODEV; /*XXX*/
836 if (idev->cnf.disable_ipv6) {
841 spin_lock(&addrconf_hash_lock);
843 /* Ignore adding duplicate addresses on an interface */
844 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
845 ADBG("ipv6_add_addr: already assigned\n");
850 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
853 ADBG("ipv6_add_addr: malloc failed\n");
858 rt = addrconf_dst_alloc(idev, addr, false);
866 ifa->peer_addr = *peer_addr;
868 spin_lock_init(&ifa->lock);
869 spin_lock_init(&ifa->state_lock);
870 setup_timer(&ifa->dad_timer, addrconf_dad_timer,
872 INIT_HLIST_NODE(&ifa->addr_lst);
874 ifa->prefix_len = pfxlen;
875 ifa->flags = flags | IFA_F_TENTATIVE;
876 ifa->valid_lft = valid_lft;
877 ifa->prefered_lft = prefered_lft;
878 ifa->cstamp = ifa->tstamp = jiffies;
879 ifa->tokenized = false;
888 /* Add to big hash table */
889 hash = inet6_addr_hash(addr);
891 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
892 spin_unlock(&addrconf_hash_lock);
894 write_lock(&idev->lock);
895 /* Add to inet6_dev unicast addr list. */
896 ipv6_link_dev_addr(idev, ifa);
898 #ifdef CONFIG_IPV6_PRIVACY
899 if (ifa->flags&IFA_F_TEMPORARY) {
900 list_add(&ifa->tmp_list, &idev->tempaddr_list);
906 write_unlock(&idev->lock);
908 rcu_read_unlock_bh();
910 if (likely(err == 0))
911 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
919 spin_unlock(&addrconf_hash_lock);
923 /* This function wants to get referenced ifp and releases it before return */
925 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
927 struct inet6_ifaddr *ifa, *ifn;
928 struct inet6_dev *idev = ifp->idev;
930 int deleted = 0, onlink = 0;
931 unsigned long expires = jiffies;
933 spin_lock_bh(&ifp->state_lock);
935 ifp->state = INET6_IFADDR_STATE_DEAD;
936 spin_unlock_bh(&ifp->state_lock);
938 if (state == INET6_IFADDR_STATE_DEAD)
941 spin_lock_bh(&addrconf_hash_lock);
942 hlist_del_init_rcu(&ifp->addr_lst);
943 spin_unlock_bh(&addrconf_hash_lock);
945 write_lock_bh(&idev->lock);
946 #ifdef CONFIG_IPV6_PRIVACY
947 if (ifp->flags&IFA_F_TEMPORARY) {
948 list_del(&ifp->tmp_list);
950 in6_ifa_put(ifp->ifpub);
957 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
959 list_del_init(&ifp->if_list);
962 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
966 } else if (ifp->flags & IFA_F_PERMANENT) {
967 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
969 if (ifa->flags & IFA_F_PERMANENT) {
974 unsigned long lifetime;
979 spin_lock(&ifa->lock);
981 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
983 * Note: Because this address is
984 * not permanent, lifetime <
985 * LONG_MAX / HZ here.
987 if (time_before(expires,
988 ifa->tstamp + lifetime * HZ))
989 expires = ifa->tstamp + lifetime * HZ;
990 spin_unlock(&ifa->lock);
995 write_unlock_bh(&idev->lock);
997 addrconf_del_dad_timer(ifp);
999 ipv6_ifa_notify(RTM_DELADDR, ifp);
1001 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1004 * Purge or update corresponding prefix
1006 * 1) we don't purge prefix here if address was not permanent.
1007 * prefix is managed by its own lifetime.
1008 * 2) if there're no addresses, delete prefix.
1009 * 3) if there're still other permanent address(es),
1010 * corresponding prefix is still permanent.
1011 * 4) otherwise, update prefix lifetime to the
1012 * longest valid lifetime among the corresponding
1013 * addresses on the device.
1014 * Note: subsequent RA will update lifetime.
1018 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
1019 struct in6_addr prefix;
1020 struct rt6_info *rt;
1022 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
1024 rt = addrconf_get_prefix_route(&prefix,
1027 0, RTF_GATEWAY | RTF_DEFAULT);
1033 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1034 rt6_set_expires(rt, expires);
1040 /* clean up prefsrc entries */
1041 rt6_remove_prefsrc(ifp);
1046 #ifdef CONFIG_IPV6_PRIVACY
1047 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1049 struct inet6_dev *idev = ifp->idev;
1050 struct in6_addr addr, *tmpaddr;
1051 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1052 unsigned long regen_advance;
1057 unsigned long now = jiffies;
1059 write_lock(&idev->lock);
1061 spin_lock_bh(&ift->lock);
1062 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1063 spin_unlock_bh(&ift->lock);
1070 if (idev->cnf.use_tempaddr <= 0) {
1071 write_unlock(&idev->lock);
1072 pr_info("%s: use_tempaddr is disabled\n", __func__);
1077 spin_lock_bh(&ifp->lock);
1078 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1079 idev->cnf.use_tempaddr = -1; /*XXX*/
1080 spin_unlock_bh(&ifp->lock);
1081 write_unlock(&idev->lock);
1082 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1089 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1090 __ipv6_try_regen_rndid(idev, tmpaddr);
1091 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1092 age = (now - ifp->tstamp) / HZ;
1093 tmp_valid_lft = min_t(__u32,
1095 idev->cnf.temp_valid_lft + age);
1096 tmp_prefered_lft = min_t(__u32,
1098 idev->cnf.temp_prefered_lft + age -
1099 idev->cnf.max_desync_factor);
1100 tmp_plen = ifp->prefix_len;
1101 max_addresses = idev->cnf.max_addresses;
1102 tmp_tstamp = ifp->tstamp;
1103 spin_unlock_bh(&ifp->lock);
1105 regen_advance = idev->cnf.regen_max_retry *
1106 idev->cnf.dad_transmits *
1107 idev->nd_parms->retrans_time / HZ;
1108 write_unlock(&idev->lock);
1110 /* A temporary address is created only if this calculated Preferred
1111 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1112 * an implementation must not create a temporary address with a zero
1113 * Preferred Lifetime.
1115 if (tmp_prefered_lft <= regen_advance) {
1122 addr_flags = IFA_F_TEMPORARY;
1123 /* set in addrconf_prefix_rcv() */
1124 if (ifp->flags & IFA_F_OPTIMISTIC)
1125 addr_flags |= IFA_F_OPTIMISTIC;
1127 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1128 ipv6_addr_scope(&addr), addr_flags,
1129 tmp_valid_lft, tmp_prefered_lft);
1133 pr_info("%s: retry temporary address regeneration\n", __func__);
1135 write_lock(&idev->lock);
1139 spin_lock_bh(&ift->lock);
1142 ift->tstamp = tmp_tstamp;
1143 spin_unlock_bh(&ift->lock);
1145 addrconf_dad_start(ift);
1154 * Choose an appropriate source address (RFC3484)
1157 IPV6_SADDR_RULE_INIT = 0,
1158 IPV6_SADDR_RULE_LOCAL,
1159 IPV6_SADDR_RULE_SCOPE,
1160 IPV6_SADDR_RULE_PREFERRED,
1161 #ifdef CONFIG_IPV6_MIP6
1162 IPV6_SADDR_RULE_HOA,
1164 IPV6_SADDR_RULE_OIF,
1165 IPV6_SADDR_RULE_LABEL,
1166 #ifdef CONFIG_IPV6_PRIVACY
1167 IPV6_SADDR_RULE_PRIVACY,
1169 IPV6_SADDR_RULE_ORCHID,
1170 IPV6_SADDR_RULE_PREFIX,
1174 struct ipv6_saddr_score {
1177 struct inet6_ifaddr *ifa;
1178 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1183 struct ipv6_saddr_dst {
1184 const struct in6_addr *addr;
1191 static inline int ipv6_saddr_preferred(int type)
1193 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1198 static int ipv6_get_saddr_eval(struct net *net,
1199 struct ipv6_saddr_score *score,
1200 struct ipv6_saddr_dst *dst,
1205 if (i <= score->rule) {
1207 case IPV6_SADDR_RULE_SCOPE:
1208 ret = score->scopedist;
1210 case IPV6_SADDR_RULE_PREFIX:
1211 ret = score->matchlen;
1214 ret = !!test_bit(i, score->scorebits);
1220 case IPV6_SADDR_RULE_INIT:
1221 /* Rule 0: remember if hiscore is not ready yet */
1224 case IPV6_SADDR_RULE_LOCAL:
1225 /* Rule 1: Prefer same address */
1226 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1228 case IPV6_SADDR_RULE_SCOPE:
1229 /* Rule 2: Prefer appropriate scope
1234 * ---+--+-+---> scope
1236 * | d is scope of the destination.
1238 * | \ <- smaller scope is better if
1239 * B-15 | \ if scope is enough for destinaion.
1240 * | ret = B - scope (-1 <= scope >= d <= 15).
1242 * |/ <- greater is better
1243 * -C / if scope is not enough for destination.
1244 * /| ret = scope - C (-1 <= d < scope <= 15).
1246 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1247 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1248 * Assume B = 0 and we get C > 29.
1250 ret = __ipv6_addr_src_scope(score->addr_type);
1251 if (ret >= dst->scope)
1254 ret -= 128; /* 30 is enough */
1255 score->scopedist = ret;
1257 case IPV6_SADDR_RULE_PREFERRED:
1258 /* Rule 3: Avoid deprecated and optimistic addresses */
1259 ret = ipv6_saddr_preferred(score->addr_type) ||
1260 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1262 #ifdef CONFIG_IPV6_MIP6
1263 case IPV6_SADDR_RULE_HOA:
1265 /* Rule 4: Prefer home address */
1266 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1267 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1271 case IPV6_SADDR_RULE_OIF:
1272 /* Rule 5: Prefer outgoing interface */
1273 ret = (!dst->ifindex ||
1274 dst->ifindex == score->ifa->idev->dev->ifindex);
1276 case IPV6_SADDR_RULE_LABEL:
1277 /* Rule 6: Prefer matching label */
1278 ret = ipv6_addr_label(net,
1279 &score->ifa->addr, score->addr_type,
1280 score->ifa->idev->dev->ifindex) == dst->label;
1282 #ifdef CONFIG_IPV6_PRIVACY
1283 case IPV6_SADDR_RULE_PRIVACY:
1285 /* Rule 7: Prefer public address
1286 * Note: prefer temporary address if use_tempaddr >= 2
1288 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1289 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1290 score->ifa->idev->cnf.use_tempaddr >= 2;
1291 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1295 case IPV6_SADDR_RULE_ORCHID:
1296 /* Rule 8-: Prefer ORCHID vs ORCHID or
1297 * non-ORCHID vs non-ORCHID
1299 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1300 ipv6_addr_orchid(dst->addr));
1302 case IPV6_SADDR_RULE_PREFIX:
1303 /* Rule 8: Use longest matching prefix */
1304 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1305 if (ret > score->ifa->prefix_len)
1306 ret = score->ifa->prefix_len;
1307 score->matchlen = ret;
1314 __set_bit(i, score->scorebits);
1320 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1321 const struct in6_addr *daddr, unsigned int prefs,
1322 struct in6_addr *saddr)
1324 struct ipv6_saddr_score scores[2],
1325 *score = &scores[0], *hiscore = &scores[1];
1326 struct ipv6_saddr_dst dst;
1327 struct net_device *dev;
1330 dst_type = __ipv6_addr_type(daddr);
1332 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1333 dst.scope = __ipv6_addr_src_scope(dst_type);
1334 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1338 hiscore->ifa = NULL;
1342 for_each_netdev_rcu(net, dev) {
1343 struct inet6_dev *idev;
1345 /* Candidate Source Address (section 4)
1346 * - multicast and link-local destination address,
1347 * the set of candidate source address MUST only
1348 * include addresses assigned to interfaces
1349 * belonging to the same link as the outgoing
1351 * (- For site-local destination addresses, the
1352 * set of candidate source addresses MUST only
1353 * include addresses assigned to interfaces
1354 * belonging to the same site as the outgoing
1357 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1358 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1359 dst.ifindex && dev->ifindex != dst.ifindex)
1362 idev = __in6_dev_get(dev);
1366 read_lock_bh(&idev->lock);
1367 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1371 * - Tentative Address (RFC2462 section 5.4)
1372 * - A tentative address is not considered
1373 * "assigned to an interface" in the traditional
1374 * sense, unless it is also flagged as optimistic.
1375 * - Candidate Source Address (section 4)
1376 * - In any case, anycast addresses, multicast
1377 * addresses, and the unspecified address MUST
1378 * NOT be included in a candidate set.
1380 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1381 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1384 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1386 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1387 score->addr_type & IPV6_ADDR_MULTICAST)) {
1388 LIMIT_NETDEBUG(KERN_DEBUG
1389 "ADDRCONF: unspecified / multicast address "
1390 "assigned as unicast address on %s",
1396 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1398 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1399 int minihiscore, miniscore;
1401 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1402 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1404 if (minihiscore > miniscore) {
1405 if (i == IPV6_SADDR_RULE_SCOPE &&
1406 score->scopedist > 0) {
1409 * each remaining entry
1410 * has too small (not enough)
1411 * scope, because ifa entries
1412 * are sorted by their scope
1418 } else if (minihiscore < miniscore) {
1420 in6_ifa_put(hiscore->ifa);
1422 in6_ifa_hold(score->ifa);
1424 swap(hiscore, score);
1426 /* restore our iterator */
1427 score->ifa = hiscore->ifa;
1434 read_unlock_bh(&idev->lock);
1439 return -EADDRNOTAVAIL;
1441 *saddr = hiscore->ifa->addr;
1442 in6_ifa_put(hiscore->ifa);
1445 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1447 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1448 unsigned char banned_flags)
1450 struct inet6_ifaddr *ifp;
1451 int err = -EADDRNOTAVAIL;
1453 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1454 if (ifp->scope == IFA_LINK &&
1455 !(ifp->flags & banned_flags)) {
1464 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1465 unsigned char banned_flags)
1467 struct inet6_dev *idev;
1468 int err = -EADDRNOTAVAIL;
1471 idev = __in6_dev_get(dev);
1473 read_lock_bh(&idev->lock);
1474 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1475 read_unlock_bh(&idev->lock);
1481 static int ipv6_count_addresses(struct inet6_dev *idev)
1484 struct inet6_ifaddr *ifp;
1486 read_lock_bh(&idev->lock);
1487 list_for_each_entry(ifp, &idev->addr_list, if_list)
1489 read_unlock_bh(&idev->lock);
1493 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1494 const struct net_device *dev, int strict)
1496 struct inet6_ifaddr *ifp;
1497 unsigned int hash = inet6_addr_hash(addr);
1500 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1501 if (!net_eq(dev_net(ifp->idev->dev), net))
1503 if (ipv6_addr_equal(&ifp->addr, addr) &&
1504 !(ifp->flags&IFA_F_TENTATIVE) &&
1505 (dev == NULL || ifp->idev->dev == dev ||
1506 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1507 rcu_read_unlock_bh();
1512 rcu_read_unlock_bh();
1515 EXPORT_SYMBOL(ipv6_chk_addr);
1517 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1518 struct net_device *dev)
1520 unsigned int hash = inet6_addr_hash(addr);
1521 struct inet6_ifaddr *ifp;
1523 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1524 if (!net_eq(dev_net(ifp->idev->dev), net))
1526 if (ipv6_addr_equal(&ifp->addr, addr)) {
1527 if (dev == NULL || ifp->idev->dev == dev)
1534 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1536 struct inet6_dev *idev;
1537 struct inet6_ifaddr *ifa;
1542 idev = __in6_dev_get(dev);
1544 read_lock_bh(&idev->lock);
1545 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1546 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1551 read_unlock_bh(&idev->lock);
1556 EXPORT_SYMBOL(ipv6_chk_prefix);
1558 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1559 struct net_device *dev, int strict)
1561 struct inet6_ifaddr *ifp, *result = NULL;
1562 unsigned int hash = inet6_addr_hash(addr);
1565 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1566 if (!net_eq(dev_net(ifp->idev->dev), net))
1568 if (ipv6_addr_equal(&ifp->addr, addr)) {
1569 if (dev == NULL || ifp->idev->dev == dev ||
1570 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1577 rcu_read_unlock_bh();
1582 /* Gets referenced address, destroys ifaddr */
1584 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1586 if (ifp->flags&IFA_F_PERMANENT) {
1587 spin_lock_bh(&ifp->lock);
1588 addrconf_del_dad_timer(ifp);
1589 ifp->flags |= IFA_F_TENTATIVE;
1591 ifp->flags |= IFA_F_DADFAILED;
1592 spin_unlock_bh(&ifp->lock);
1594 ipv6_ifa_notify(0, ifp);
1596 #ifdef CONFIG_IPV6_PRIVACY
1597 } else if (ifp->flags&IFA_F_TEMPORARY) {
1598 struct inet6_ifaddr *ifpub;
1599 spin_lock_bh(&ifp->lock);
1602 in6_ifa_hold(ifpub);
1603 spin_unlock_bh(&ifp->lock);
1604 ipv6_create_tempaddr(ifpub, ifp);
1607 spin_unlock_bh(&ifp->lock);
1615 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1619 spin_lock(&ifp->state_lock);
1620 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1621 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1624 spin_unlock(&ifp->state_lock);
1629 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1631 struct inet6_dev *idev = ifp->idev;
1633 if (addrconf_dad_end(ifp)) {
1638 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1639 ifp->idev->dev->name, &ifp->addr);
1641 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1642 struct in6_addr addr;
1644 addr.s6_addr32[0] = htonl(0xfe800000);
1645 addr.s6_addr32[1] = 0;
1647 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1648 ipv6_addr_equal(&ifp->addr, &addr)) {
1649 /* DAD failed for link-local based on MAC address */
1650 idev->cnf.disable_ipv6 = 1;
1652 pr_info("%s: IPv6 being disabled!\n",
1653 ifp->idev->dev->name);
1657 addrconf_dad_stop(ifp, 1);
1660 /* Join to solicited addr multicast group. */
1662 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1664 struct in6_addr maddr;
1666 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1669 addrconf_addr_solict_mult(addr, &maddr);
1670 ipv6_dev_mc_inc(dev, &maddr);
1673 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1675 struct in6_addr maddr;
1677 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1680 addrconf_addr_solict_mult(addr, &maddr);
1681 __ipv6_dev_mc_dec(idev, &maddr);
1684 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1686 struct in6_addr addr;
1687 if (ifp->prefix_len == 127) /* RFC 6164 */
1689 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1690 if (ipv6_addr_any(&addr))
1692 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1695 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1697 struct in6_addr addr;
1698 if (ifp->prefix_len == 127) /* RFC 6164 */
1700 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1701 if (ipv6_addr_any(&addr))
1703 __ipv6_dev_ac_dec(ifp->idev, &addr);
1706 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1708 if (dev->addr_len != ETH_ALEN)
1710 memcpy(eui, dev->dev_addr, 3);
1711 memcpy(eui + 5, dev->dev_addr + 3, 3);
1714 * The zSeries OSA network cards can be shared among various
1715 * OS instances, but the OSA cards have only one MAC address.
1716 * This leads to duplicate address conflicts in conjunction
1717 * with IPv6 if more than one instance uses the same card.
1719 * The driver for these cards can deliver a unique 16-bit
1720 * identifier for each instance sharing the same card. It is
1721 * placed instead of 0xFFFE in the interface identifier. The
1722 * "u" bit of the interface identifier is not inverted in this
1723 * case. Hence the resulting interface identifier has local
1724 * scope according to RFC2373.
1727 eui[3] = (dev->dev_id >> 8) & 0xFF;
1728 eui[4] = dev->dev_id & 0xFF;
1737 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1739 if (dev->addr_len != IEEE802154_ADDR_LEN)
1741 memcpy(eui, dev->dev_addr, 8);
1746 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1748 union fwnet_hwaddr *ha;
1750 if (dev->addr_len != FWNET_ALEN)
1753 ha = (union fwnet_hwaddr *)dev->dev_addr;
1755 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1760 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1762 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1763 if (dev->addr_len != ARCNET_ALEN)
1766 eui[7] = *(u8 *)dev->dev_addr;
1770 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1772 if (dev->addr_len != INFINIBAND_ALEN)
1774 memcpy(eui, dev->dev_addr + 12, 8);
1779 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1783 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1784 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1785 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1786 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1787 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1788 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1792 memcpy(eui + 4, &addr, 4);
1796 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1798 if (dev->priv_flags & IFF_ISATAP)
1799 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1803 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1805 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1808 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1810 memcpy(eui, dev->perm_addr, 3);
1811 memcpy(eui + 5, dev->perm_addr + 3, 3);
1818 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1820 switch (dev->type) {
1823 return addrconf_ifid_eui48(eui, dev);
1825 return addrconf_ifid_arcnet(eui, dev);
1826 case ARPHRD_INFINIBAND:
1827 return addrconf_ifid_infiniband(eui, dev);
1829 return addrconf_ifid_sit(eui, dev);
1831 return addrconf_ifid_gre(eui, dev);
1832 case ARPHRD_IEEE802154:
1833 return addrconf_ifid_eui64(eui, dev);
1834 case ARPHRD_IEEE1394:
1835 return addrconf_ifid_ieee1394(eui, dev);
1836 case ARPHRD_TUNNEL6:
1837 return addrconf_ifid_ip6tnl(eui, dev);
1842 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1845 struct inet6_ifaddr *ifp;
1847 read_lock_bh(&idev->lock);
1848 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1849 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1850 memcpy(eui, ifp->addr.s6_addr+8, 8);
1855 read_unlock_bh(&idev->lock);
1859 #ifdef CONFIG_IPV6_PRIVACY
1860 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1861 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1864 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1865 idev->rndid[0] &= ~0x02;
1868 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1869 * check if generated address is not inappropriate
1871 * - Reserved subnet anycast (RFC 2526)
1872 * 11111101 11....11 1xxxxxxx
1873 * - ISATAP (RFC4214) 6.1
1874 * 00-00-5E-FE-xx-xx-xx-xx
1876 * - XXX: already assigned to an address on the device
1878 if (idev->rndid[0] == 0xfd &&
1879 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1880 (idev->rndid[7]&0x80))
1882 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1883 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1885 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1890 static void ipv6_regen_rndid(unsigned long data)
1892 struct inet6_dev *idev = (struct inet6_dev *) data;
1893 unsigned long expires;
1896 write_lock_bh(&idev->lock);
1901 __ipv6_regen_rndid(idev);
1904 idev->cnf.temp_prefered_lft * HZ -
1905 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1906 idev->cnf.max_desync_factor * HZ;
1907 if (time_before(expires, jiffies)) {
1908 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1909 __func__, idev->dev->name);
1913 if (!mod_timer(&idev->regen_timer, expires))
1917 write_unlock_bh(&idev->lock);
1918 rcu_read_unlock_bh();
1922 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1924 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1925 __ipv6_regen_rndid(idev);
1934 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1935 unsigned long expires, u32 flags)
1937 struct fib6_config cfg = {
1938 .fc_table = RT6_TABLE_PREFIX,
1939 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1940 .fc_ifindex = dev->ifindex,
1941 .fc_expires = expires,
1943 .fc_flags = RTF_UP | flags,
1944 .fc_nlinfo.nl_net = dev_net(dev),
1945 .fc_protocol = RTPROT_KERNEL,
1950 /* Prevent useless cloning on PtP SIT.
1951 This thing is done here expecting that the whole
1952 class of non-broadcast devices need not cloning.
1954 #if IS_ENABLED(CONFIG_IPV6_SIT)
1955 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1956 cfg.fc_flags |= RTF_NONEXTHOP;
1959 ip6_route_add(&cfg);
1963 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1965 const struct net_device *dev,
1966 u32 flags, u32 noflags)
1968 struct fib6_node *fn;
1969 struct rt6_info *rt = NULL;
1970 struct fib6_table *table;
1972 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1976 read_lock_bh(&table->tb6_lock);
1977 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1980 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1981 if (rt->dst.dev->ifindex != dev->ifindex)
1983 if ((rt->rt6i_flags & flags) != flags)
1985 if ((rt->rt6i_flags & noflags) != 0)
1991 read_unlock_bh(&table->tb6_lock);
1996 /* Create "default" multicast route to the interface */
1998 static void addrconf_add_mroute(struct net_device *dev)
2000 struct fib6_config cfg = {
2001 .fc_table = RT6_TABLE_LOCAL,
2002 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2003 .fc_ifindex = dev->ifindex,
2006 .fc_nlinfo.nl_net = dev_net(dev),
2009 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2011 ip6_route_add(&cfg);
2014 #if IS_ENABLED(CONFIG_IPV6_SIT)
2015 static void sit_route_add(struct net_device *dev)
2017 struct fib6_config cfg = {
2018 .fc_table = RT6_TABLE_MAIN,
2019 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2020 .fc_ifindex = dev->ifindex,
2022 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2023 .fc_nlinfo.nl_net = dev_net(dev),
2026 /* prefix length - 96 bits "::d.d.d.d" */
2027 ip6_route_add(&cfg);
2031 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2033 struct inet6_dev *idev;
2037 idev = ipv6_find_idev(dev);
2039 return ERR_PTR(-ENOBUFS);
2041 if (idev->cnf.disable_ipv6)
2042 return ERR_PTR(-EACCES);
2044 /* Add default multicast route */
2045 if (!(dev->flags & IFF_LOOPBACK))
2046 addrconf_add_mroute(dev);
2051 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2053 struct prefix_info *pinfo;
2057 struct inet6_dev *in6_dev;
2058 struct net *net = dev_net(dev);
2060 pinfo = (struct prefix_info *) opt;
2062 if (len < sizeof(struct prefix_info)) {
2063 ADBG("addrconf: prefix option too short\n");
2068 * Validation checks ([ADDRCONF], page 19)
2071 addr_type = ipv6_addr_type(&pinfo->prefix);
2073 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2076 valid_lft = ntohl(pinfo->valid);
2077 prefered_lft = ntohl(pinfo->prefered);
2079 if (prefered_lft > valid_lft) {
2080 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2084 in6_dev = in6_dev_get(dev);
2086 if (in6_dev == NULL) {
2087 net_dbg_ratelimited("addrconf: device %s not configured\n",
2093 * Two things going on here:
2094 * 1) Add routes for on-link prefixes
2095 * 2) Configure prefixes with the auto flag set
2098 if (pinfo->onlink) {
2099 struct rt6_info *rt;
2100 unsigned long rt_expires;
2102 /* Avoid arithmetic overflow. Really, we could
2103 * save rt_expires in seconds, likely valid_lft,
2104 * but it would require division in fib gc, that it
2108 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2110 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2112 if (addrconf_finite_timeout(rt_expires))
2115 rt = addrconf_get_prefix_route(&pinfo->prefix,
2118 RTF_ADDRCONF | RTF_PREFIX_RT,
2119 RTF_GATEWAY | RTF_DEFAULT);
2122 /* Autoconf prefix route */
2123 if (valid_lft == 0) {
2126 } else if (addrconf_finite_timeout(rt_expires)) {
2128 rt6_set_expires(rt, jiffies + rt_expires);
2130 rt6_clean_expires(rt);
2132 } else if (valid_lft) {
2133 clock_t expires = 0;
2134 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2135 if (addrconf_finite_timeout(rt_expires)) {
2137 flags |= RTF_EXPIRES;
2138 expires = jiffies_to_clock_t(rt_expires);
2140 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2141 dev, expires, flags);
2146 /* Try to figure out our local address for this prefix */
2148 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2149 struct inet6_ifaddr *ifp;
2150 struct in6_addr addr;
2151 int create = 0, update_lft = 0;
2152 bool tokenized = false;
2154 if (pinfo->prefix_len == 64) {
2155 memcpy(&addr, &pinfo->prefix, 8);
2157 if (!ipv6_addr_any(&in6_dev->token)) {
2158 read_lock_bh(&in6_dev->lock);
2159 memcpy(addr.s6_addr + 8,
2160 in6_dev->token.s6_addr + 8, 8);
2161 read_unlock_bh(&in6_dev->lock);
2163 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2164 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2165 in6_dev_put(in6_dev);
2170 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2172 in6_dev_put(in6_dev);
2177 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2179 if (ifp == NULL && valid_lft) {
2180 int max_addresses = in6_dev->cnf.max_addresses;
2183 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2184 if (in6_dev->cnf.optimistic_dad &&
2185 !net->ipv6.devconf_all->forwarding && sllao)
2186 addr_flags = IFA_F_OPTIMISTIC;
2189 /* Do not allow to create too much of autoconfigured
2190 * addresses; this would be too easy way to crash kernel.
2192 if (!max_addresses ||
2193 ipv6_count_addresses(in6_dev) < max_addresses)
2194 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2196 addr_type&IPV6_ADDR_SCOPE_MASK,
2197 addr_flags, valid_lft,
2200 if (IS_ERR_OR_NULL(ifp)) {
2201 in6_dev_put(in6_dev);
2207 ifp->cstamp = jiffies;
2208 ifp->tokenized = tokenized;
2209 addrconf_dad_start(ifp);
2215 #ifdef CONFIG_IPV6_PRIVACY
2216 struct inet6_ifaddr *ift;
2220 /* update lifetime (RFC2462 5.5.3 e) */
2221 spin_lock(&ifp->lock);
2223 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2224 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2227 if (!update_lft && !create && stored_lft) {
2228 if (valid_lft > MIN_VALID_LIFETIME ||
2229 valid_lft > stored_lft)
2231 else if (stored_lft <= MIN_VALID_LIFETIME) {
2232 /* valid_lft <= stored_lft is always true */
2234 * RFC 4862 Section 5.5.3e:
2235 * "Note that the preferred lifetime of
2236 * the corresponding address is always
2237 * reset to the Preferred Lifetime in
2238 * the received Prefix Information
2239 * option, regardless of whether the
2240 * valid lifetime is also reset or
2243 * So if the preferred lifetime in
2244 * this advertisement is different
2245 * than what we have stored, but the
2246 * valid lifetime is invalid, just
2247 * reset prefered_lft.
2249 * We must set the valid lifetime
2250 * to the stored lifetime since we'll
2251 * be updating the timestamp below,
2252 * else we'll set it back to the
2255 if (prefered_lft != ifp->prefered_lft) {
2256 valid_lft = stored_lft;
2260 valid_lft = MIN_VALID_LIFETIME;
2261 if (valid_lft < prefered_lft)
2262 prefered_lft = valid_lft;
2268 ifp->valid_lft = valid_lft;
2269 ifp->prefered_lft = prefered_lft;
2272 ifp->flags &= ~IFA_F_DEPRECATED;
2273 spin_unlock(&ifp->lock);
2275 if (!(flags&IFA_F_TENTATIVE))
2276 ipv6_ifa_notify(0, ifp);
2278 spin_unlock(&ifp->lock);
2280 #ifdef CONFIG_IPV6_PRIVACY
2281 read_lock_bh(&in6_dev->lock);
2282 /* update all temporary addresses in the list */
2283 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2285 int age, max_valid, max_prefered;
2287 if (ifp != ift->ifpub)
2291 * RFC 4941 section 3.3:
2292 * If a received option will extend the lifetime
2293 * of a public address, the lifetimes of
2294 * temporary addresses should be extended,
2295 * subject to the overall constraint that no
2296 * temporary addresses should ever remain
2297 * "valid" or "preferred" for a time longer than
2298 * (TEMP_VALID_LIFETIME) or
2299 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2302 age = (now - ift->cstamp) / HZ;
2303 max_valid = in6_dev->cnf.temp_valid_lft - age;
2307 max_prefered = in6_dev->cnf.temp_prefered_lft -
2308 in6_dev->cnf.max_desync_factor -
2310 if (max_prefered < 0)
2313 if (valid_lft > max_valid)
2314 valid_lft = max_valid;
2316 if (prefered_lft > max_prefered)
2317 prefered_lft = max_prefered;
2319 spin_lock(&ift->lock);
2321 ift->valid_lft = valid_lft;
2322 ift->prefered_lft = prefered_lft;
2324 if (prefered_lft > 0)
2325 ift->flags &= ~IFA_F_DEPRECATED;
2327 spin_unlock(&ift->lock);
2328 if (!(flags&IFA_F_TENTATIVE))
2329 ipv6_ifa_notify(0, ift);
2332 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2334 * When a new public address is created as
2335 * described in [ADDRCONF], also create a new
2336 * temporary address. Also create a temporary
2337 * address if it's enabled but no temporary
2338 * address currently exists.
2340 read_unlock_bh(&in6_dev->lock);
2341 ipv6_create_tempaddr(ifp, NULL);
2343 read_unlock_bh(&in6_dev->lock);
2350 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2351 in6_dev_put(in6_dev);
2355 * Set destination address.
2356 * Special case for SIT interfaces where we create a new "virtual"
2359 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2361 struct in6_ifreq ireq;
2362 struct net_device *dev;
2368 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2371 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2377 #if IS_ENABLED(CONFIG_IPV6_SIT)
2378 if (dev->type == ARPHRD_SIT) {
2379 const struct net_device_ops *ops = dev->netdev_ops;
2381 struct ip_tunnel_parm p;
2383 err = -EADDRNOTAVAIL;
2384 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2387 memset(&p, 0, sizeof(p));
2388 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2392 p.iph.protocol = IPPROTO_IPV6;
2394 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2396 if (ops->ndo_do_ioctl) {
2397 mm_segment_t oldfs = get_fs();
2400 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2407 dev = __dev_get_by_name(net, p.name);
2410 err = dev_open(dev);
2421 * Manual configuration of address on an interface
2423 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2424 const struct in6_addr *peer_pfx,
2425 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2428 struct inet6_ifaddr *ifp;
2429 struct inet6_dev *idev;
2430 struct net_device *dev;
2434 unsigned long timeout;
2441 /* check the lifetime */
2442 if (!valid_lft || prefered_lft > valid_lft)
2445 dev = __dev_get_by_index(net, ifindex);
2449 idev = addrconf_add_dev(dev);
2451 return PTR_ERR(idev);
2453 scope = ipv6_addr_scope(pfx);
2455 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2456 if (addrconf_finite_timeout(timeout)) {
2457 expires = jiffies_to_clock_t(timeout * HZ);
2458 valid_lft = timeout;
2459 flags = RTF_EXPIRES;
2463 ifa_flags |= IFA_F_PERMANENT;
2466 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2467 if (addrconf_finite_timeout(timeout)) {
2469 ifa_flags |= IFA_F_DEPRECATED;
2470 prefered_lft = timeout;
2473 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2474 valid_lft, prefered_lft);
2477 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2480 * Note that section 3.1 of RFC 4429 indicates
2481 * that the Optimistic flag should not be set for
2482 * manually configured addresses
2484 addrconf_dad_start(ifp);
2490 return PTR_ERR(ifp);
2493 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2496 struct inet6_ifaddr *ifp;
2497 struct inet6_dev *idev;
2498 struct net_device *dev;
2503 dev = __dev_get_by_index(net, ifindex);
2507 if ((idev = __in6_dev_get(dev)) == NULL)
2510 read_lock_bh(&idev->lock);
2511 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2512 if (ifp->prefix_len == plen &&
2513 ipv6_addr_equal(pfx, &ifp->addr)) {
2515 read_unlock_bh(&idev->lock);
2521 read_unlock_bh(&idev->lock);
2522 return -EADDRNOTAVAIL;
2526 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2528 struct in6_ifreq ireq;
2531 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2534 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2538 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2539 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2540 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2545 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2547 struct in6_ifreq ireq;
2550 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2553 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2557 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2558 ireq.ifr6_prefixlen);
2563 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2564 int plen, int scope)
2566 struct inet6_ifaddr *ifp;
2568 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2569 scope, IFA_F_PERMANENT, 0, 0);
2571 spin_lock_bh(&ifp->lock);
2572 ifp->flags &= ~IFA_F_TENTATIVE;
2573 spin_unlock_bh(&ifp->lock);
2574 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2579 #if IS_ENABLED(CONFIG_IPV6_SIT)
2580 static void sit_add_v4_addrs(struct inet6_dev *idev)
2582 struct in6_addr addr;
2583 struct net_device *dev;
2584 struct net *net = dev_net(idev->dev);
2589 memset(&addr, 0, sizeof(struct in6_addr));
2590 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2592 if (idev->dev->flags&IFF_POINTOPOINT) {
2593 addr.s6_addr32[0] = htonl(0xfe800000);
2596 scope = IPV6_ADDR_COMPATv4;
2599 if (addr.s6_addr32[3]) {
2600 add_addr(idev, &addr, 128, scope);
2604 for_each_netdev(net, dev) {
2605 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2606 if (in_dev && (dev->flags & IFF_UP)) {
2607 struct in_ifaddr *ifa;
2611 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2614 addr.s6_addr32[3] = ifa->ifa_local;
2616 if (ifa->ifa_scope == RT_SCOPE_LINK)
2618 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2619 if (idev->dev->flags&IFF_POINTOPOINT)
2623 if (idev->dev->flags&IFF_POINTOPOINT)
2628 add_addr(idev, &addr, plen, flag);
2635 static void init_loopback(struct net_device *dev)
2637 struct inet6_dev *idev;
2638 struct net_device *sp_dev;
2639 struct inet6_ifaddr *sp_ifa;
2640 struct rt6_info *sp_rt;
2646 if ((idev = ipv6_find_idev(dev)) == NULL) {
2647 pr_debug("%s: add_dev failed\n", __func__);
2651 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2653 /* Add routes to other interface's IPv6 addresses */
2654 for_each_netdev(dev_net(dev), sp_dev) {
2655 if (!strcmp(sp_dev->name, dev->name))
2658 idev = __in6_dev_get(sp_dev);
2662 read_lock_bh(&idev->lock);
2663 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2665 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2671 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2673 /* Failure cases are ignored */
2674 if (!IS_ERR(sp_rt)) {
2679 read_unlock_bh(&idev->lock);
2683 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2685 struct inet6_ifaddr *ifp;
2686 u32 addr_flags = IFA_F_PERMANENT;
2688 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2689 if (idev->cnf.optimistic_dad &&
2690 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2691 addr_flags |= IFA_F_OPTIMISTIC;
2695 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0);
2697 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2698 addrconf_dad_start(ifp);
2703 static void addrconf_dev_config(struct net_device *dev)
2705 struct in6_addr addr;
2706 struct inet6_dev *idev;
2710 if ((dev->type != ARPHRD_ETHER) &&
2711 (dev->type != ARPHRD_FDDI) &&
2712 (dev->type != ARPHRD_ARCNET) &&
2713 (dev->type != ARPHRD_INFINIBAND) &&
2714 (dev->type != ARPHRD_IEEE802154) &&
2715 (dev->type != ARPHRD_IEEE1394) &&
2716 (dev->type != ARPHRD_TUNNEL6)) {
2717 /* Alas, we support only Ethernet autoconfiguration. */
2721 idev = addrconf_add_dev(dev);
2725 memset(&addr, 0, sizeof(struct in6_addr));
2726 addr.s6_addr32[0] = htonl(0xFE800000);
2728 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2729 addrconf_add_linklocal(idev, &addr);
2732 #if IS_ENABLED(CONFIG_IPV6_SIT)
2733 static void addrconf_sit_config(struct net_device *dev)
2735 struct inet6_dev *idev;
2740 * Configure the tunnel with one of our IPv4
2741 * addresses... we should configure all of
2742 * our v4 addrs in the tunnel
2745 if ((idev = ipv6_find_idev(dev)) == NULL) {
2746 pr_debug("%s: add_dev failed\n", __func__);
2750 if (dev->priv_flags & IFF_ISATAP) {
2751 struct in6_addr addr;
2753 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2754 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2755 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2756 addrconf_add_linklocal(idev, &addr);
2760 sit_add_v4_addrs(idev);
2762 if (dev->flags&IFF_POINTOPOINT)
2763 addrconf_add_mroute(dev);
2769 #if IS_ENABLED(CONFIG_NET_IPGRE)
2770 static void addrconf_gre_config(struct net_device *dev)
2772 struct inet6_dev *idev;
2773 struct in6_addr addr;
2777 if ((idev = ipv6_find_idev(dev)) == NULL) {
2778 pr_debug("%s: add_dev failed\n", __func__);
2782 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2783 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2785 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2786 addrconf_add_linklocal(idev, &addr);
2791 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2793 struct in6_addr lladdr;
2795 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2796 addrconf_add_linklocal(idev, &lladdr);
2802 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2805 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2806 struct inet6_dev *idev = __in6_dev_get(dev);
2807 int run_pending = 0;
2811 case NETDEV_REGISTER:
2812 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2813 idev = ipv6_add_dev(dev);
2815 return notifier_from_errno(-ENOMEM);
2821 if (dev->flags & IFF_SLAVE)
2824 if (event == NETDEV_UP) {
2825 if (!addrconf_qdisc_ok(dev)) {
2826 /* device is not ready yet. */
2827 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2832 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2833 idev = ipv6_add_dev(dev);
2836 idev->if_flags |= IF_READY;
2840 if (!addrconf_qdisc_ok(dev)) {
2841 /* device is still not ready. */
2846 if (idev->if_flags & IF_READY)
2847 /* device is already configured. */
2849 idev->if_flags |= IF_READY;
2852 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2858 switch (dev->type) {
2859 #if IS_ENABLED(CONFIG_IPV6_SIT)
2861 addrconf_sit_config(dev);
2864 #if IS_ENABLED(CONFIG_NET_IPGRE)
2866 addrconf_gre_config(dev);
2869 case ARPHRD_LOOPBACK:
2874 addrconf_dev_config(dev);
2880 addrconf_dad_run(idev);
2883 * If the MTU changed during the interface down,
2884 * when the interface up, the changed MTU must be
2885 * reflected in the idev as well as routers.
2887 if (idev->cnf.mtu6 != dev->mtu &&
2888 dev->mtu >= IPV6_MIN_MTU) {
2889 rt6_mtu_change(dev, dev->mtu);
2890 idev->cnf.mtu6 = dev->mtu;
2892 idev->tstamp = jiffies;
2893 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2896 * If the changed mtu during down is lower than
2897 * IPV6_MIN_MTU stop IPv6 on this interface.
2899 if (dev->mtu < IPV6_MIN_MTU)
2900 addrconf_ifdown(dev, 1);
2904 case NETDEV_CHANGEMTU:
2905 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2906 rt6_mtu_change(dev, dev->mtu);
2907 idev->cnf.mtu6 = dev->mtu;
2911 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2912 idev = ipv6_add_dev(dev);
2918 * MTU falled under IPV6_MIN_MTU.
2919 * Stop IPv6 on this interface.
2923 case NETDEV_UNREGISTER:
2925 * Remove all addresses from this interface.
2927 addrconf_ifdown(dev, event != NETDEV_DOWN);
2930 case NETDEV_CHANGENAME:
2932 snmp6_unregister_dev(idev);
2933 addrconf_sysctl_unregister(idev);
2934 addrconf_sysctl_register(idev);
2935 err = snmp6_register_dev(idev);
2937 return notifier_from_errno(err);
2941 case NETDEV_PRE_TYPE_CHANGE:
2942 case NETDEV_POST_TYPE_CHANGE:
2943 addrconf_type_change(dev, event);
2951 * addrconf module should be notified of a device going up
2953 static struct notifier_block ipv6_dev_notf = {
2954 .notifier_call = addrconf_notify,
2957 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2959 struct inet6_dev *idev;
2962 idev = __in6_dev_get(dev);
2964 if (event == NETDEV_POST_TYPE_CHANGE)
2965 ipv6_mc_remap(idev);
2966 else if (event == NETDEV_PRE_TYPE_CHANGE)
2967 ipv6_mc_unmap(idev);
2970 static int addrconf_ifdown(struct net_device *dev, int how)
2972 struct net *net = dev_net(dev);
2973 struct inet6_dev *idev;
2974 struct inet6_ifaddr *ifa;
2979 rt6_ifdown(net, dev);
2980 neigh_ifdown(&nd_tbl, dev);
2982 idev = __in6_dev_get(dev);
2987 * Step 1: remove reference to ipv6 device from parent device.
2993 /* protected by rtnl_lock */
2994 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2996 /* Step 1.5: remove snmp6 entry */
2997 snmp6_unregister_dev(idev);
3001 /* Step 2: clear hash table */
3002 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3003 struct hlist_head *h = &inet6_addr_lst[i];
3005 spin_lock_bh(&addrconf_hash_lock);
3007 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3008 if (ifa->idev == idev) {
3009 hlist_del_init_rcu(&ifa->addr_lst);
3010 addrconf_del_dad_timer(ifa);
3014 spin_unlock_bh(&addrconf_hash_lock);
3017 write_lock_bh(&idev->lock);
3019 addrconf_del_rs_timer(idev);
3021 /* Step 2: clear flags for stateless addrconf */
3023 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3025 #ifdef CONFIG_IPV6_PRIVACY
3026 if (how && del_timer(&idev->regen_timer))
3029 /* Step 3: clear tempaddr list */
3030 while (!list_empty(&idev->tempaddr_list)) {
3031 ifa = list_first_entry(&idev->tempaddr_list,
3032 struct inet6_ifaddr, tmp_list);
3033 list_del(&ifa->tmp_list);
3034 write_unlock_bh(&idev->lock);
3035 spin_lock_bh(&ifa->lock);
3038 in6_ifa_put(ifa->ifpub);
3041 spin_unlock_bh(&ifa->lock);
3043 write_lock_bh(&idev->lock);
3047 while (!list_empty(&idev->addr_list)) {
3048 ifa = list_first_entry(&idev->addr_list,
3049 struct inet6_ifaddr, if_list);
3050 addrconf_del_dad_timer(ifa);
3052 list_del(&ifa->if_list);
3054 write_unlock_bh(&idev->lock);
3056 spin_lock_bh(&ifa->state_lock);
3058 ifa->state = INET6_IFADDR_STATE_DEAD;
3059 spin_unlock_bh(&ifa->state_lock);
3061 if (state != INET6_IFADDR_STATE_DEAD) {
3062 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3063 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3067 write_lock_bh(&idev->lock);
3070 write_unlock_bh(&idev->lock);
3072 /* Step 5: Discard multicast list */
3074 ipv6_mc_destroy_dev(idev);
3078 idev->tstamp = jiffies;
3080 /* Last: Shot the device (if unregistered) */
3082 addrconf_sysctl_unregister(idev);
3083 neigh_parms_release(&nd_tbl, idev->nd_parms);
3084 neigh_ifdown(&nd_tbl, dev);
3090 static void addrconf_rs_timer(unsigned long data)
3092 struct inet6_dev *idev = (struct inet6_dev *)data;
3093 struct in6_addr lladdr;
3095 write_lock(&idev->lock);
3096 if (idev->dead || !(idev->if_flags & IF_READY))
3099 if (!ipv6_accept_ra(idev))
3102 /* Announcement received after solicitation was sent */
3103 if (idev->if_flags & IF_RA_RCVD)
3106 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3107 if (!__ipv6_get_lladdr(idev, &lladdr, IFA_F_TENTATIVE))
3108 ndisc_send_rs(idev->dev, &lladdr,
3109 &in6addr_linklocal_allrouters);
3113 /* The wait after the last probe can be shorter */
3114 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3115 idev->cnf.rtr_solicits) ?
3116 idev->cnf.rtr_solicit_delay :
3117 idev->cnf.rtr_solicit_interval);
3120 * Note: we do not support deprecated "all on-link"
3121 * assumption any longer.
3123 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3127 write_unlock(&idev->lock);
3132 * Duplicate Address Detection
3134 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3136 unsigned long rand_num;
3137 struct inet6_dev *idev = ifp->idev;
3139 if (ifp->flags & IFA_F_OPTIMISTIC)
3142 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3144 ifp->dad_probes = idev->cnf.dad_transmits;
3145 addrconf_mod_dad_timer(ifp, rand_num);
3148 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3150 struct inet6_dev *idev = ifp->idev;
3151 struct net_device *dev = idev->dev;
3153 addrconf_join_solict(dev, &ifp->addr);
3155 net_srandom(ifp->addr.s6_addr32[3]);
3157 read_lock_bh(&idev->lock);
3158 spin_lock(&ifp->lock);
3159 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3162 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3163 idev->cnf.accept_dad < 1 ||
3164 !(ifp->flags&IFA_F_TENTATIVE) ||
3165 ifp->flags & IFA_F_NODAD) {
3166 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3167 spin_unlock(&ifp->lock);
3168 read_unlock_bh(&idev->lock);
3170 addrconf_dad_completed(ifp);
3174 if (!(idev->if_flags & IF_READY)) {
3175 spin_unlock(&ifp->lock);
3176 read_unlock_bh(&idev->lock);
3178 * If the device is not ready:
3179 * - keep it tentative if it is a permanent address.
3180 * - otherwise, kill it.
3183 addrconf_dad_stop(ifp, 0);
3188 * Optimistic nodes can start receiving
3191 if (ifp->flags & IFA_F_OPTIMISTIC)
3192 ip6_ins_rt(ifp->rt);
3194 addrconf_dad_kick(ifp);
3196 spin_unlock(&ifp->lock);
3197 read_unlock_bh(&idev->lock);
3200 static void addrconf_dad_timer(unsigned long data)
3202 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3203 struct inet6_dev *idev = ifp->idev;
3204 struct in6_addr mcaddr;
3206 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3209 write_lock(&idev->lock);
3210 if (idev->dead || !(idev->if_flags & IF_READY)) {
3211 write_unlock(&idev->lock);
3215 spin_lock(&ifp->lock);
3216 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3217 spin_unlock(&ifp->lock);
3218 write_unlock(&idev->lock);
3222 if (ifp->dad_probes == 0) {
3224 * DAD was successful
3227 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3228 spin_unlock(&ifp->lock);
3229 write_unlock(&idev->lock);
3231 addrconf_dad_completed(ifp);
3237 addrconf_mod_dad_timer(ifp, ifp->idev->nd_parms->retrans_time);
3238 spin_unlock(&ifp->lock);
3239 write_unlock(&idev->lock);
3241 /* send a neighbour solicitation for our addr */
3242 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3243 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3248 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3250 struct net_device *dev = ifp->idev->dev;
3251 struct in6_addr lladdr;
3252 bool send_rs, send_mld;
3254 addrconf_del_dad_timer(ifp);
3257 * Configure the address for reception. Now it is valid.
3260 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3262 /* If added prefix is link local and we are prepared to process
3263 router advertisements, start sending router solicitations.
3266 read_lock_bh(&ifp->idev->lock);
3267 spin_lock(&ifp->lock);
3268 send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL &&
3269 ifp->idev->valid_ll_addr_cnt == 1;
3270 send_rs = send_mld &&
3271 ipv6_accept_ra(ifp->idev) &&
3272 ifp->idev->cnf.rtr_solicits > 0 &&
3273 (dev->flags&IFF_LOOPBACK) == 0;
3274 spin_unlock(&ifp->lock);
3275 read_unlock_bh(&ifp->idev->lock);
3277 /* While dad is in progress mld report's source address is in6_addrany.
3278 * Resend with proper ll now.
3281 ipv6_mc_dad_complete(ifp->idev);
3285 * If a host as already performed a random delay
3286 * [...] as part of DAD [...] there is no need
3287 * to delay again before sending the first RS
3289 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3291 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3293 write_lock_bh(&ifp->idev->lock);
3294 spin_lock(&ifp->lock);
3295 ifp->idev->rs_probes = 1;
3296 ifp->idev->if_flags |= IF_RS_SENT;
3297 addrconf_mod_rs_timer(ifp->idev,
3298 ifp->idev->cnf.rtr_solicit_interval);
3299 spin_unlock(&ifp->lock);
3300 write_unlock_bh(&ifp->idev->lock);
3304 static void addrconf_dad_run(struct inet6_dev *idev)
3306 struct inet6_ifaddr *ifp;
3308 read_lock_bh(&idev->lock);
3309 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3310 spin_lock(&ifp->lock);
3311 if (ifp->flags & IFA_F_TENTATIVE &&
3312 ifp->state == INET6_IFADDR_STATE_DAD)
3313 addrconf_dad_kick(ifp);
3314 spin_unlock(&ifp->lock);
3316 read_unlock_bh(&idev->lock);
3319 #ifdef CONFIG_PROC_FS
3320 struct if6_iter_state {
3321 struct seq_net_private p;
3326 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3328 struct inet6_ifaddr *ifa = NULL;
3329 struct if6_iter_state *state = seq->private;
3330 struct net *net = seq_file_net(seq);
3333 /* initial bucket if pos is 0 */
3339 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3340 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3342 if (!net_eq(dev_net(ifa->idev->dev), net))
3344 /* sync with offset */
3345 if (p < state->offset) {
3353 /* prepare for next bucket */
3360 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3361 struct inet6_ifaddr *ifa)
3363 struct if6_iter_state *state = seq->private;
3364 struct net *net = seq_file_net(seq);
3366 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3367 if (!net_eq(dev_net(ifa->idev->dev), net))
3373 while (++state->bucket < IN6_ADDR_HSIZE) {
3375 hlist_for_each_entry_rcu_bh(ifa,
3376 &inet6_addr_lst[state->bucket], addr_lst) {
3377 if (!net_eq(dev_net(ifa->idev->dev), net))
3387 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3391 return if6_get_first(seq, *pos);
3394 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3396 struct inet6_ifaddr *ifa;
3398 ifa = if6_get_next(seq, v);
3403 static void if6_seq_stop(struct seq_file *seq, void *v)
3406 rcu_read_unlock_bh();
3409 static int if6_seq_show(struct seq_file *seq, void *v)
3411 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3412 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3414 ifp->idev->dev->ifindex,
3418 ifp->idev->dev->name);
3422 static const struct seq_operations if6_seq_ops = {
3423 .start = if6_seq_start,
3424 .next = if6_seq_next,
3425 .show = if6_seq_show,
3426 .stop = if6_seq_stop,
3429 static int if6_seq_open(struct inode *inode, struct file *file)
3431 return seq_open_net(inode, file, &if6_seq_ops,
3432 sizeof(struct if6_iter_state));
3435 static const struct file_operations if6_fops = {
3436 .owner = THIS_MODULE,
3437 .open = if6_seq_open,
3439 .llseek = seq_lseek,
3440 .release = seq_release_net,
3443 static int __net_init if6_proc_net_init(struct net *net)
3445 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3450 static void __net_exit if6_proc_net_exit(struct net *net)
3452 remove_proc_entry("if_inet6", net->proc_net);
3455 static struct pernet_operations if6_proc_net_ops = {
3456 .init = if6_proc_net_init,
3457 .exit = if6_proc_net_exit,
3460 int __init if6_proc_init(void)
3462 return register_pernet_subsys(&if6_proc_net_ops);
3465 void if6_proc_exit(void)
3467 unregister_pernet_subsys(&if6_proc_net_ops);
3469 #endif /* CONFIG_PROC_FS */
3471 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3472 /* Check if address is a home address configured on any interface. */
3473 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3476 struct inet6_ifaddr *ifp = NULL;
3477 unsigned int hash = inet6_addr_hash(addr);
3480 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3481 if (!net_eq(dev_net(ifp->idev->dev), net))
3483 if (ipv6_addr_equal(&ifp->addr, addr) &&
3484 (ifp->flags & IFA_F_HOMEADDRESS)) {
3489 rcu_read_unlock_bh();
3495 * Periodic address status verification
3498 static void addrconf_verify(unsigned long foo)
3500 unsigned long now, next, next_sec, next_sched;
3501 struct inet6_ifaddr *ifp;
3505 spin_lock(&addrconf_verify_lock);
3507 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3509 del_timer(&addr_chk_timer);
3511 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3513 hlist_for_each_entry_rcu_bh(ifp,
3514 &inet6_addr_lst[i], addr_lst) {
3517 if (ifp->flags & IFA_F_PERMANENT)
3520 spin_lock(&ifp->lock);
3521 /* We try to batch several events at once. */
3522 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3524 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3525 age >= ifp->valid_lft) {
3526 spin_unlock(&ifp->lock);
3530 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3531 spin_unlock(&ifp->lock);
3533 } else if (age >= ifp->prefered_lft) {
3534 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3537 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3539 ifp->flags |= IFA_F_DEPRECATED;
3542 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3543 next = ifp->tstamp + ifp->valid_lft * HZ;
3545 spin_unlock(&ifp->lock);
3550 ipv6_ifa_notify(0, ifp);
3554 #ifdef CONFIG_IPV6_PRIVACY
3555 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3556 !(ifp->flags&IFA_F_TENTATIVE)) {
3557 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3558 ifp->idev->cnf.dad_transmits *
3559 ifp->idev->nd_parms->retrans_time / HZ;
3561 if (age >= ifp->prefered_lft - regen_advance) {
3562 struct inet6_ifaddr *ifpub = ifp->ifpub;
3563 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3564 next = ifp->tstamp + ifp->prefered_lft * HZ;
3565 if (!ifp->regen_count && ifpub) {
3568 in6_ifa_hold(ifpub);
3569 spin_unlock(&ifp->lock);
3571 spin_lock(&ifpub->lock);
3572 ifpub->regen_count = 0;
3573 spin_unlock(&ifpub->lock);
3574 ipv6_create_tempaddr(ifpub, ifp);
3579 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3580 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3581 spin_unlock(&ifp->lock);
3584 /* ifp->prefered_lft <= ifp->valid_lft */
3585 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3586 next = ifp->tstamp + ifp->prefered_lft * HZ;
3587 spin_unlock(&ifp->lock);
3592 next_sec = round_jiffies_up(next);
3595 /* If rounded timeout is accurate enough, accept it. */
3596 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3597 next_sched = next_sec;
3599 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3600 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3601 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3603 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3604 now, next, next_sec, next_sched);
3606 addr_chk_timer.expires = next_sched;
3607 add_timer(&addr_chk_timer);
3608 spin_unlock(&addrconf_verify_lock);
3609 rcu_read_unlock_bh();
3612 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3613 struct in6_addr **peer_pfx)
3615 struct in6_addr *pfx = NULL;
3620 pfx = nla_data(addr);
3623 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3625 pfx = nla_data(local);
3631 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3632 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3633 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3634 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3638 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3640 struct net *net = sock_net(skb->sk);
3641 struct ifaddrmsg *ifm;
3642 struct nlattr *tb[IFA_MAX+1];
3643 struct in6_addr *pfx, *peer_pfx;
3646 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3650 ifm = nlmsg_data(nlh);
3651 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3655 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3658 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3659 u32 prefered_lft, u32 valid_lft)
3663 unsigned long timeout;
3665 if (!valid_lft || (prefered_lft > valid_lft))
3668 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3669 if (addrconf_finite_timeout(timeout)) {
3670 expires = jiffies_to_clock_t(timeout * HZ);
3671 valid_lft = timeout;
3672 flags = RTF_EXPIRES;
3676 ifa_flags |= IFA_F_PERMANENT;
3679 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3680 if (addrconf_finite_timeout(timeout)) {
3682 ifa_flags |= IFA_F_DEPRECATED;
3683 prefered_lft = timeout;
3686 spin_lock_bh(&ifp->lock);
3687 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3688 ifp->tstamp = jiffies;
3689 ifp->valid_lft = valid_lft;
3690 ifp->prefered_lft = prefered_lft;
3692 spin_unlock_bh(&ifp->lock);
3693 if (!(ifp->flags&IFA_F_TENTATIVE))
3694 ipv6_ifa_notify(0, ifp);
3696 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3704 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3706 struct net *net = sock_net(skb->sk);
3707 struct ifaddrmsg *ifm;
3708 struct nlattr *tb[IFA_MAX+1];
3709 struct in6_addr *pfx, *peer_pfx;
3710 struct inet6_ifaddr *ifa;
3711 struct net_device *dev;
3712 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3716 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3720 ifm = nlmsg_data(nlh);
3721 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3725 if (tb[IFA_CACHEINFO]) {
3726 struct ifa_cacheinfo *ci;
3728 ci = nla_data(tb[IFA_CACHEINFO]);
3729 valid_lft = ci->ifa_valid;
3730 preferred_lft = ci->ifa_prefered;
3732 preferred_lft = INFINITY_LIFE_TIME;
3733 valid_lft = INFINITY_LIFE_TIME;
3736 dev = __dev_get_by_index(net, ifm->ifa_index);
3740 /* We ignore other flags so far. */
3741 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3743 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3746 * It would be best to check for !NLM_F_CREATE here but
3747 * userspace alreay relies on not having to provide this.
3749 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3750 ifm->ifa_prefixlen, ifa_flags,
3751 preferred_lft, valid_lft);
3754 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3755 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3758 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3765 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3766 u8 scope, int ifindex)
3768 struct ifaddrmsg *ifm;
3770 ifm = nlmsg_data(nlh);
3771 ifm->ifa_family = AF_INET6;
3772 ifm->ifa_prefixlen = prefixlen;
3773 ifm->ifa_flags = flags;
3774 ifm->ifa_scope = scope;
3775 ifm->ifa_index = ifindex;
3778 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3779 unsigned long tstamp, u32 preferred, u32 valid)
3781 struct ifa_cacheinfo ci;
3783 ci.cstamp = cstamp_delta(cstamp);
3784 ci.tstamp = cstamp_delta(tstamp);
3785 ci.ifa_prefered = preferred;
3786 ci.ifa_valid = valid;
3788 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3791 static inline int rt_scope(int ifa_scope)
3793 if (ifa_scope & IFA_HOST)
3794 return RT_SCOPE_HOST;
3795 else if (ifa_scope & IFA_LINK)
3796 return RT_SCOPE_LINK;
3797 else if (ifa_scope & IFA_SITE)
3798 return RT_SCOPE_SITE;
3800 return RT_SCOPE_UNIVERSE;
3803 static inline int inet6_ifaddr_msgsize(void)
3805 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3806 + nla_total_size(16) /* IFA_LOCAL */
3807 + nla_total_size(16) /* IFA_ADDRESS */
3808 + nla_total_size(sizeof(struct ifa_cacheinfo));
3811 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3812 u32 portid, u32 seq, int event, unsigned int flags)
3814 struct nlmsghdr *nlh;
3815 u32 preferred, valid;
3817 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3821 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3822 ifa->idev->dev->ifindex);
3824 if (!(ifa->flags&IFA_F_PERMANENT)) {
3825 preferred = ifa->prefered_lft;
3826 valid = ifa->valid_lft;
3827 if (preferred != INFINITY_LIFE_TIME) {
3828 long tval = (jiffies - ifa->tstamp)/HZ;
3829 if (preferred > tval)
3833 if (valid != INFINITY_LIFE_TIME) {
3841 preferred = INFINITY_LIFE_TIME;
3842 valid = INFINITY_LIFE_TIME;
3845 if (!ipv6_addr_any(&ifa->peer_addr)) {
3846 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
3847 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
3850 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
3853 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3856 return nlmsg_end(skb, nlh);
3859 nlmsg_cancel(skb, nlh);
3863 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3864 u32 portid, u32 seq, int event, u16 flags)
3866 struct nlmsghdr *nlh;
3867 u8 scope = RT_SCOPE_UNIVERSE;
3868 int ifindex = ifmca->idev->dev->ifindex;
3870 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3871 scope = RT_SCOPE_SITE;
3873 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3877 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3878 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3879 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3880 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3881 nlmsg_cancel(skb, nlh);
3885 return nlmsg_end(skb, nlh);
3888 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3889 u32 portid, u32 seq, int event, unsigned int flags)
3891 struct nlmsghdr *nlh;
3892 u8 scope = RT_SCOPE_UNIVERSE;
3893 int ifindex = ifaca->aca_idev->dev->ifindex;
3895 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3896 scope = RT_SCOPE_SITE;
3898 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3902 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3903 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3904 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3905 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3906 nlmsg_cancel(skb, nlh);
3910 return nlmsg_end(skb, nlh);
3919 /* called with rcu_read_lock() */
3920 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3921 struct netlink_callback *cb, enum addr_type_t type,
3922 int s_ip_idx, int *p_ip_idx)
3924 struct ifmcaddr6 *ifmca;
3925 struct ifacaddr6 *ifaca;
3927 int ip_idx = *p_ip_idx;
3929 read_lock_bh(&idev->lock);
3931 case UNICAST_ADDR: {
3932 struct inet6_ifaddr *ifa;
3934 /* unicast address incl. temp addr */
3935 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3936 if (++ip_idx < s_ip_idx)
3938 err = inet6_fill_ifaddr(skb, ifa,
3939 NETLINK_CB(cb->skb).portid,
3945 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3949 case MULTICAST_ADDR:
3950 /* multicast address */
3951 for (ifmca = idev->mc_list; ifmca;
3952 ifmca = ifmca->next, ip_idx++) {
3953 if (ip_idx < s_ip_idx)
3955 err = inet6_fill_ifmcaddr(skb, ifmca,
3956 NETLINK_CB(cb->skb).portid,
3965 /* anycast address */
3966 for (ifaca = idev->ac_list; ifaca;
3967 ifaca = ifaca->aca_next, ip_idx++) {
3968 if (ip_idx < s_ip_idx)
3970 err = inet6_fill_ifacaddr(skb, ifaca,
3971 NETLINK_CB(cb->skb).portid,
3982 read_unlock_bh(&idev->lock);
3987 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3988 enum addr_type_t type)
3990 struct net *net = sock_net(skb->sk);
3993 int s_idx, s_ip_idx;
3994 struct net_device *dev;
3995 struct inet6_dev *idev;
3996 struct hlist_head *head;
3999 s_idx = idx = cb->args[1];
4000 s_ip_idx = ip_idx = cb->args[2];
4003 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4004 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4006 head = &net->dev_index_head[h];
4007 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4010 if (h > s_h || idx > s_idx)
4013 idev = __in6_dev_get(dev);
4017 if (in6_dump_addrs(idev, skb, cb, type,
4018 s_ip_idx, &ip_idx) <= 0)
4028 cb->args[2] = ip_idx;
4033 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4035 enum addr_type_t type = UNICAST_ADDR;
4037 return inet6_dump_addr(skb, cb, type);
4040 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4042 enum addr_type_t type = MULTICAST_ADDR;
4044 return inet6_dump_addr(skb, cb, type);
4048 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4050 enum addr_type_t type = ANYCAST_ADDR;
4052 return inet6_dump_addr(skb, cb, type);
4055 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4057 struct net *net = sock_net(in_skb->sk);
4058 struct ifaddrmsg *ifm;
4059 struct nlattr *tb[IFA_MAX+1];
4060 struct in6_addr *addr = NULL, *peer;
4061 struct net_device *dev = NULL;
4062 struct inet6_ifaddr *ifa;
4063 struct sk_buff *skb;
4066 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4070 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4076 ifm = nlmsg_data(nlh);
4078 dev = __dev_get_by_index(net, ifm->ifa_index);
4080 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4082 err = -EADDRNOTAVAIL;
4086 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4092 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4093 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4095 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4096 WARN_ON(err == -EMSGSIZE);
4100 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4107 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4109 struct sk_buff *skb;
4110 struct net *net = dev_net(ifa->idev->dev);
4113 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4117 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4119 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4120 WARN_ON(err == -EMSGSIZE);
4124 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4128 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4131 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4132 __s32 *array, int bytes)
4134 BUG_ON(bytes < (DEVCONF_MAX * 4));
4136 memset(array, 0, bytes);
4137 array[DEVCONF_FORWARDING] = cnf->forwarding;
4138 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4139 array[DEVCONF_MTU6] = cnf->mtu6;
4140 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4141 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4142 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4143 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4144 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4145 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4146 jiffies_to_msecs(cnf->rtr_solicit_interval);
4147 array[DEVCONF_RTR_SOLICIT_DELAY] =
4148 jiffies_to_msecs(cnf->rtr_solicit_delay);
4149 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4150 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4151 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4152 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4153 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4154 #ifdef CONFIG_IPV6_PRIVACY
4155 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4156 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4157 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4158 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4159 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4161 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4162 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4163 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4164 #ifdef CONFIG_IPV6_ROUTER_PREF
4165 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4166 array[DEVCONF_RTR_PROBE_INTERVAL] =
4167 jiffies_to_msecs(cnf->rtr_probe_interval);
4168 #ifdef CONFIG_IPV6_ROUTE_INFO
4169 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4172 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4173 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4174 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4175 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4177 #ifdef CONFIG_IPV6_MROUTE
4178 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4180 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4181 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4182 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4183 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4184 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4187 static inline size_t inet6_ifla6_size(void)
4189 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4190 + nla_total_size(sizeof(struct ifla_cacheinfo))
4191 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4192 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4193 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4194 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4197 static inline size_t inet6_if_nlmsg_size(void)
4199 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4200 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4201 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4202 + nla_total_size(4) /* IFLA_MTU */
4203 + nla_total_size(4) /* IFLA_LINK */
4204 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4207 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4208 int items, int bytes)
4211 int pad = bytes - sizeof(u64) * items;
4214 /* Use put_unaligned() because stats may not be aligned for u64. */
4215 put_unaligned(items, &stats[0]);
4216 for (i = 1; i < items; i++)
4217 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4219 memset(&stats[items], 0, pad);
4222 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4223 int items, int bytes, size_t syncpoff)
4226 int pad = bytes - sizeof(u64) * items;
4229 /* Use put_unaligned() because stats may not be aligned for u64. */
4230 put_unaligned(items, &stats[0]);
4231 for (i = 1; i < items; i++)
4232 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4234 memset(&stats[items], 0, pad);
4237 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4241 case IFLA_INET6_STATS:
4242 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4243 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4245 case IFLA_INET6_ICMP6STATS:
4246 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4251 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4254 struct ifla_cacheinfo ci;
4256 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4257 goto nla_put_failure;
4258 ci.max_reasm_len = IPV6_MAXPLEN;
4259 ci.tstamp = cstamp_delta(idev->tstamp);
4260 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4261 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4262 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4263 goto nla_put_failure;
4264 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4266 goto nla_put_failure;
4267 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4269 /* XXX - MC not implemented */
4271 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4273 goto nla_put_failure;
4274 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4276 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4278 goto nla_put_failure;
4279 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4281 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4283 goto nla_put_failure;
4284 read_lock_bh(&idev->lock);
4285 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4286 read_unlock_bh(&idev->lock);
4294 static size_t inet6_get_link_af_size(const struct net_device *dev)
4296 if (!__in6_dev_get(dev))
4299 return inet6_ifla6_size();
4302 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4304 struct inet6_dev *idev = __in6_dev_get(dev);
4309 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4315 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4317 struct inet6_ifaddr *ifp;
4318 struct net_device *dev = idev->dev;
4319 bool update_rs = false;
4320 struct in6_addr ll_addr;
4324 if (ipv6_addr_any(token))
4326 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4328 if (!ipv6_accept_ra(idev))
4330 if (idev->cnf.rtr_solicits <= 0)
4333 write_lock_bh(&idev->lock);
4335 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4336 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4338 write_unlock_bh(&idev->lock);
4340 if (!idev->dead && (idev->if_flags & IF_READY) &&
4341 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4342 IFA_F_OPTIMISTIC)) {
4344 /* If we're not ready, then normal ifup will take care
4345 * of this. Otherwise, we need to request our rs here.
4347 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4351 write_lock_bh(&idev->lock);
4354 idev->if_flags |= IF_RS_SENT;
4355 idev->rs_probes = 1;
4356 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4359 /* Well, that's kinda nasty ... */
4360 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4361 spin_lock(&ifp->lock);
4362 if (ifp->tokenized) {
4364 ifp->prefered_lft = 0;
4366 spin_unlock(&ifp->lock);
4369 write_unlock_bh(&idev->lock);
4374 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4377 struct inet6_dev *idev = __in6_dev_get(dev);
4378 struct nlattr *tb[IFLA_INET6_MAX + 1];
4381 return -EAFNOSUPPORT;
4383 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4386 if (tb[IFLA_INET6_TOKEN])
4387 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4392 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4393 u32 portid, u32 seq, int event, unsigned int flags)
4395 struct net_device *dev = idev->dev;
4396 struct ifinfomsg *hdr;
4397 struct nlmsghdr *nlh;
4400 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4404 hdr = nlmsg_data(nlh);
4405 hdr->ifi_family = AF_INET6;
4407 hdr->ifi_type = dev->type;
4408 hdr->ifi_index = dev->ifindex;
4409 hdr->ifi_flags = dev_get_flags(dev);
4410 hdr->ifi_change = 0;
4412 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4414 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4415 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4416 (dev->ifindex != dev->iflink &&
4417 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4418 goto nla_put_failure;
4419 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4420 if (protoinfo == NULL)
4421 goto nla_put_failure;
4423 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4424 goto nla_put_failure;
4426 nla_nest_end(skb, protoinfo);
4427 return nlmsg_end(skb, nlh);
4430 nlmsg_cancel(skb, nlh);
4434 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4436 struct net *net = sock_net(skb->sk);
4439 struct net_device *dev;
4440 struct inet6_dev *idev;
4441 struct hlist_head *head;
4444 s_idx = cb->args[1];
4447 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4449 head = &net->dev_index_head[h];
4450 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4453 idev = __in6_dev_get(dev);
4456 if (inet6_fill_ifinfo(skb, idev,
4457 NETLINK_CB(cb->skb).portid,
4459 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4473 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4475 struct sk_buff *skb;
4476 struct net *net = dev_net(idev->dev);
4479 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4483 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4485 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4486 WARN_ON(err == -EMSGSIZE);
4490 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4494 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4497 static inline size_t inet6_prefix_nlmsg_size(void)
4499 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4500 + nla_total_size(sizeof(struct in6_addr))
4501 + nla_total_size(sizeof(struct prefix_cacheinfo));
4504 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4505 struct prefix_info *pinfo, u32 portid, u32 seq,
4506 int event, unsigned int flags)
4508 struct prefixmsg *pmsg;
4509 struct nlmsghdr *nlh;
4510 struct prefix_cacheinfo ci;
4512 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4516 pmsg = nlmsg_data(nlh);
4517 pmsg->prefix_family = AF_INET6;
4518 pmsg->prefix_pad1 = 0;
4519 pmsg->prefix_pad2 = 0;
4520 pmsg->prefix_ifindex = idev->dev->ifindex;
4521 pmsg->prefix_len = pinfo->prefix_len;
4522 pmsg->prefix_type = pinfo->type;
4523 pmsg->prefix_pad3 = 0;
4524 pmsg->prefix_flags = 0;
4526 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4527 if (pinfo->autoconf)
4528 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4530 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4531 goto nla_put_failure;
4532 ci.preferred_time = ntohl(pinfo->prefered);
4533 ci.valid_time = ntohl(pinfo->valid);
4534 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4535 goto nla_put_failure;
4536 return nlmsg_end(skb, nlh);
4539 nlmsg_cancel(skb, nlh);
4543 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4544 struct prefix_info *pinfo)
4546 struct sk_buff *skb;
4547 struct net *net = dev_net(idev->dev);
4550 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4554 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4556 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4557 WARN_ON(err == -EMSGSIZE);
4561 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4565 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4568 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count)
4570 write_lock_bh(&ifp->idev->lock);
4571 spin_lock(&ifp->lock);
4572 if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|
4573 IFA_F_DADFAILED)) == IFA_F_PERMANENT) &&
4574 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL))
4575 ifp->idev->valid_ll_addr_cnt += count;
4576 WARN_ON(ifp->idev->valid_ll_addr_cnt < 0);
4577 spin_unlock(&ifp->lock);
4578 write_unlock_bh(&ifp->idev->lock);
4581 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4583 struct net *net = dev_net(ifp->idev->dev);
4585 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4589 update_valid_ll_addr_cnt(ifp, 1);
4592 * If the address was optimistic
4593 * we inserted the route at the start of
4594 * our DAD process, so we don't need
4597 if (!(ifp->rt->rt6i_node))
4598 ip6_ins_rt(ifp->rt);
4599 if (ifp->idev->cnf.forwarding)
4600 addrconf_join_anycast(ifp);
4601 if (!ipv6_addr_any(&ifp->peer_addr))
4602 addrconf_prefix_route(&ifp->peer_addr, 128,
4603 ifp->idev->dev, 0, 0);
4606 update_valid_ll_addr_cnt(ifp, -1);
4608 if (ifp->idev->cnf.forwarding)
4609 addrconf_leave_anycast(ifp);
4610 addrconf_leave_solict(ifp->idev, &ifp->addr);
4611 if (!ipv6_addr_any(&ifp->peer_addr)) {
4612 struct rt6_info *rt;
4613 struct net_device *dev = ifp->idev->dev;
4615 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4623 dst_hold(&ifp->rt->dst);
4625 if (ip6_del_rt(ifp->rt))
4626 dst_free(&ifp->rt->dst);
4629 atomic_inc(&net->ipv6.dev_addr_genid);
4630 rt_genid_bump_ipv6(net);
4633 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4636 if (likely(ifp->idev->dead == 0))
4637 __ipv6_ifa_notify(event, ifp);
4638 rcu_read_unlock_bh();
4641 #ifdef CONFIG_SYSCTL
4644 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4645 void __user *buffer, size_t *lenp, loff_t *ppos)
4647 int *valp = ctl->data;
4650 struct ctl_table lctl;
4654 * ctl->data points to idev->cnf.forwarding, we should
4655 * not modify it until we get the rtnl lock.
4660 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4663 ret = addrconf_fixup_forwarding(ctl, valp, val);
4669 static void dev_disable_change(struct inet6_dev *idev)
4671 struct netdev_notifier_info info;
4673 if (!idev || !idev->dev)
4676 netdev_notifier_info_init(&info, idev->dev);
4677 if (idev->cnf.disable_ipv6)
4678 addrconf_notify(NULL, NETDEV_DOWN, &info);
4680 addrconf_notify(NULL, NETDEV_UP, &info);
4683 static void addrconf_disable_change(struct net *net, __s32 newf)
4685 struct net_device *dev;
4686 struct inet6_dev *idev;
4689 for_each_netdev_rcu(net, dev) {
4690 idev = __in6_dev_get(dev);
4692 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4693 idev->cnf.disable_ipv6 = newf;
4695 dev_disable_change(idev);
4701 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4706 if (!rtnl_trylock())
4707 return restart_syscall();
4709 net = (struct net *)table->extra2;
4713 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4718 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4719 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4720 addrconf_disable_change(net, newf);
4721 } else if ((!newf) ^ (!old))
4722 dev_disable_change((struct inet6_dev *)table->extra1);
4729 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4730 void __user *buffer, size_t *lenp, loff_t *ppos)
4732 int *valp = ctl->data;
4735 struct ctl_table lctl;
4739 * ctl->data points to idev->cnf.disable_ipv6, we should
4740 * not modify it until we get the rtnl lock.
4745 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4748 ret = addrconf_disable_ipv6(ctl, valp, val);
4754 static struct addrconf_sysctl_table
4756 struct ctl_table_header *sysctl_header;
4757 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4758 } addrconf_sysctl __read_mostly = {
4759 .sysctl_header = NULL,
4762 .procname = "forwarding",
4763 .data = &ipv6_devconf.forwarding,
4764 .maxlen = sizeof(int),
4766 .proc_handler = addrconf_sysctl_forward,
4769 .procname = "hop_limit",
4770 .data = &ipv6_devconf.hop_limit,
4771 .maxlen = sizeof(int),
4773 .proc_handler = proc_dointvec,
4777 .data = &ipv6_devconf.mtu6,
4778 .maxlen = sizeof(int),
4780 .proc_handler = proc_dointvec,
4783 .procname = "accept_ra",
4784 .data = &ipv6_devconf.accept_ra,
4785 .maxlen = sizeof(int),
4787 .proc_handler = proc_dointvec,
4790 .procname = "accept_redirects",
4791 .data = &ipv6_devconf.accept_redirects,
4792 .maxlen = sizeof(int),
4794 .proc_handler = proc_dointvec,
4797 .procname = "autoconf",
4798 .data = &ipv6_devconf.autoconf,
4799 .maxlen = sizeof(int),
4801 .proc_handler = proc_dointvec,
4804 .procname = "dad_transmits",
4805 .data = &ipv6_devconf.dad_transmits,
4806 .maxlen = sizeof(int),
4808 .proc_handler = proc_dointvec,
4811 .procname = "router_solicitations",
4812 .data = &ipv6_devconf.rtr_solicits,
4813 .maxlen = sizeof(int),
4815 .proc_handler = proc_dointvec,
4818 .procname = "router_solicitation_interval",
4819 .data = &ipv6_devconf.rtr_solicit_interval,
4820 .maxlen = sizeof(int),
4822 .proc_handler = proc_dointvec_jiffies,
4825 .procname = "router_solicitation_delay",
4826 .data = &ipv6_devconf.rtr_solicit_delay,
4827 .maxlen = sizeof(int),
4829 .proc_handler = proc_dointvec_jiffies,
4832 .procname = "force_mld_version",
4833 .data = &ipv6_devconf.force_mld_version,
4834 .maxlen = sizeof(int),
4836 .proc_handler = proc_dointvec,
4839 .procname = "mldv1_unsolicited_report_interval",
4841 &ipv6_devconf.mldv1_unsolicited_report_interval,
4842 .maxlen = sizeof(int),
4844 .proc_handler = proc_dointvec_ms_jiffies,
4847 .procname = "mldv2_unsolicited_report_interval",
4849 &ipv6_devconf.mldv2_unsolicited_report_interval,
4850 .maxlen = sizeof(int),
4852 .proc_handler = proc_dointvec_ms_jiffies,
4854 #ifdef CONFIG_IPV6_PRIVACY
4856 .procname = "use_tempaddr",
4857 .data = &ipv6_devconf.use_tempaddr,
4858 .maxlen = sizeof(int),
4860 .proc_handler = proc_dointvec,
4863 .procname = "temp_valid_lft",
4864 .data = &ipv6_devconf.temp_valid_lft,
4865 .maxlen = sizeof(int),
4867 .proc_handler = proc_dointvec,
4870 .procname = "temp_prefered_lft",
4871 .data = &ipv6_devconf.temp_prefered_lft,
4872 .maxlen = sizeof(int),
4874 .proc_handler = proc_dointvec,
4877 .procname = "regen_max_retry",
4878 .data = &ipv6_devconf.regen_max_retry,
4879 .maxlen = sizeof(int),
4881 .proc_handler = proc_dointvec,
4884 .procname = "max_desync_factor",
4885 .data = &ipv6_devconf.max_desync_factor,
4886 .maxlen = sizeof(int),
4888 .proc_handler = proc_dointvec,
4892 .procname = "max_addresses",
4893 .data = &ipv6_devconf.max_addresses,
4894 .maxlen = sizeof(int),
4896 .proc_handler = proc_dointvec,
4899 .procname = "accept_ra_defrtr",
4900 .data = &ipv6_devconf.accept_ra_defrtr,
4901 .maxlen = sizeof(int),
4903 .proc_handler = proc_dointvec,
4906 .procname = "accept_ra_pinfo",
4907 .data = &ipv6_devconf.accept_ra_pinfo,
4908 .maxlen = sizeof(int),
4910 .proc_handler = proc_dointvec,
4912 #ifdef CONFIG_IPV6_ROUTER_PREF
4914 .procname = "accept_ra_rtr_pref",
4915 .data = &ipv6_devconf.accept_ra_rtr_pref,
4916 .maxlen = sizeof(int),
4918 .proc_handler = proc_dointvec,
4921 .procname = "router_probe_interval",
4922 .data = &ipv6_devconf.rtr_probe_interval,
4923 .maxlen = sizeof(int),
4925 .proc_handler = proc_dointvec_jiffies,
4927 #ifdef CONFIG_IPV6_ROUTE_INFO
4929 .procname = "accept_ra_rt_info_max_plen",
4930 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4931 .maxlen = sizeof(int),
4933 .proc_handler = proc_dointvec,
4938 .procname = "proxy_ndp",
4939 .data = &ipv6_devconf.proxy_ndp,
4940 .maxlen = sizeof(int),
4942 .proc_handler = proc_dointvec,
4945 .procname = "accept_source_route",
4946 .data = &ipv6_devconf.accept_source_route,
4947 .maxlen = sizeof(int),
4949 .proc_handler = proc_dointvec,
4951 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4953 .procname = "optimistic_dad",
4954 .data = &ipv6_devconf.optimistic_dad,
4955 .maxlen = sizeof(int),
4957 .proc_handler = proc_dointvec,
4961 #ifdef CONFIG_IPV6_MROUTE
4963 .procname = "mc_forwarding",
4964 .data = &ipv6_devconf.mc_forwarding,
4965 .maxlen = sizeof(int),
4967 .proc_handler = proc_dointvec,
4971 .procname = "disable_ipv6",
4972 .data = &ipv6_devconf.disable_ipv6,
4973 .maxlen = sizeof(int),
4975 .proc_handler = addrconf_sysctl_disable,
4978 .procname = "accept_dad",
4979 .data = &ipv6_devconf.accept_dad,
4980 .maxlen = sizeof(int),
4982 .proc_handler = proc_dointvec,
4985 .procname = "force_tllao",
4986 .data = &ipv6_devconf.force_tllao,
4987 .maxlen = sizeof(int),
4989 .proc_handler = proc_dointvec
4992 .procname = "ndisc_notify",
4993 .data = &ipv6_devconf.ndisc_notify,
4994 .maxlen = sizeof(int),
4996 .proc_handler = proc_dointvec
4999 .procname = "suppress_frag_ndisc",
5000 .data = &ipv6_devconf.suppress_frag_ndisc,
5001 .maxlen = sizeof(int),
5003 .proc_handler = proc_dointvec
5011 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5012 struct inet6_dev *idev, struct ipv6_devconf *p)
5015 struct addrconf_sysctl_table *t;
5016 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5018 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5022 for (i = 0; t->addrconf_vars[i].data; i++) {
5023 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5024 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5025 t->addrconf_vars[i].extra2 = net;
5028 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5030 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5031 if (t->sysctl_header == NULL)
5043 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5045 struct addrconf_sysctl_table *t;
5047 if (p->sysctl == NULL)
5052 unregister_net_sysctl_table(t->sysctl_header);
5056 static void addrconf_sysctl_register(struct inet6_dev *idev)
5058 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
5059 &ndisc_ifinfo_sysctl_change);
5060 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5064 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5066 __addrconf_sysctl_unregister(&idev->cnf);
5067 neigh_sysctl_unregister(idev->nd_parms);
5073 static int __net_init addrconf_init_net(struct net *net)
5076 struct ipv6_devconf *all, *dflt;
5078 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5082 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5084 goto err_alloc_dflt;
5086 /* these will be inherited by all namespaces */
5087 dflt->autoconf = ipv6_defaults.autoconf;
5088 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5090 net->ipv6.devconf_all = all;
5091 net->ipv6.devconf_dflt = dflt;
5093 #ifdef CONFIG_SYSCTL
5094 err = __addrconf_sysctl_register(net, "all", NULL, all);
5098 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5104 #ifdef CONFIG_SYSCTL
5106 __addrconf_sysctl_unregister(all);
5116 static void __net_exit addrconf_exit_net(struct net *net)
5118 #ifdef CONFIG_SYSCTL
5119 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5120 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5122 if (!net_eq(net, &init_net)) {
5123 kfree(net->ipv6.devconf_dflt);
5124 kfree(net->ipv6.devconf_all);
5128 static struct pernet_operations addrconf_ops = {
5129 .init = addrconf_init_net,
5130 .exit = addrconf_exit_net,
5133 static struct rtnl_af_ops inet6_ops = {
5135 .fill_link_af = inet6_fill_link_af,
5136 .get_link_af_size = inet6_get_link_af_size,
5137 .set_link_af = inet6_set_link_af,
5141 * Init / cleanup code
5144 int __init addrconf_init(void)
5148 err = ipv6_addr_label_init();
5150 pr_crit("%s: cannot initialize default policy table: %d\n",
5155 err = register_pernet_subsys(&addrconf_ops);
5159 /* The addrconf netdev notifier requires that loopback_dev
5160 * has it's ipv6 private information allocated and setup
5161 * before it can bring up and give link-local addresses
5162 * to other devices which are up.
5164 * Unfortunately, loopback_dev is not necessarily the first
5165 * entry in the global dev_base list of net devices. In fact,
5166 * it is likely to be the very last entry on that list.
5167 * So this causes the notifier registry below to try and
5168 * give link-local addresses to all devices besides loopback_dev
5169 * first, then loopback_dev, which cases all the non-loopback_dev
5170 * devices to fail to get a link-local address.
5172 * So, as a temporary fix, allocate the ipv6 structure for
5173 * loopback_dev first by hand.
5174 * Longer term, all of the dependencies ipv6 has upon the loopback
5175 * device and it being up should be removed.
5178 if (!ipv6_add_dev(init_net.loopback_dev))
5184 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5185 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5187 register_netdevice_notifier(&ipv6_dev_notf);
5191 err = rtnl_af_register(&inet6_ops);
5195 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5200 /* Only the first call to __rtnl_register can fail */
5201 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5202 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5203 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5204 inet6_dump_ifaddr, NULL);
5205 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5206 inet6_dump_ifmcaddr, NULL);
5207 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5208 inet6_dump_ifacaddr, NULL);
5209 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5210 inet6_netconf_dump_devconf, NULL);
5212 ipv6_addr_label_rtnl_register();
5216 rtnl_af_unregister(&inet6_ops);
5218 unregister_netdevice_notifier(&ipv6_dev_notf);
5220 unregister_pernet_subsys(&addrconf_ops);
5222 ipv6_addr_label_cleanup();
5227 void addrconf_cleanup(void)
5229 struct net_device *dev;
5232 unregister_netdevice_notifier(&ipv6_dev_notf);
5233 unregister_pernet_subsys(&addrconf_ops);
5234 ipv6_addr_label_cleanup();
5238 __rtnl_af_unregister(&inet6_ops);
5240 /* clean dev list */
5241 for_each_netdev(&init_net, dev) {
5242 if (__in6_dev_get(dev) == NULL)
5244 addrconf_ifdown(dev, 1);
5246 addrconf_ifdown(init_net.loopback_dev, 2);
5251 spin_lock_bh(&addrconf_hash_lock);
5252 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5253 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5254 spin_unlock_bh(&addrconf_hash_lock);
5256 del_timer(&addr_chk_timer);