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(x) printk x
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,
181 .rtr_solicits = MAX_RTR_SOLICITATIONS,
182 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
183 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
184 #ifdef CONFIG_IPV6_PRIVACY
186 .temp_valid_lft = TEMP_VALID_LIFETIME,
187 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
188 .regen_max_retry = REGEN_MAX_RETRY,
189 .max_desync_factor = MAX_DESYNC_FACTOR,
191 .max_addresses = IPV6_MAX_ADDRESSES,
192 .accept_ra_defrtr = 1,
193 .accept_ra_pinfo = 1,
194 #ifdef CONFIG_IPV6_ROUTER_PREF
195 .accept_ra_rtr_pref = 1,
196 .rtr_probe_interval = 60 * HZ,
197 #ifdef CONFIG_IPV6_ROUTE_INFO
198 .accept_ra_rt_info_max_plen = 0,
202 .accept_source_route = 0, /* we do not accept RH0 by default. */
207 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
209 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
210 .mtu6 = IPV6_MIN_MTU,
212 .accept_redirects = 1,
215 .rtr_solicits = MAX_RTR_SOLICITATIONS,
216 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
217 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
218 #ifdef CONFIG_IPV6_PRIVACY
220 .temp_valid_lft = TEMP_VALID_LIFETIME,
221 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
222 .regen_max_retry = REGEN_MAX_RETRY,
223 .max_desync_factor = MAX_DESYNC_FACTOR,
225 .max_addresses = IPV6_MAX_ADDRESSES,
226 .accept_ra_defrtr = 1,
227 .accept_ra_pinfo = 1,
228 #ifdef CONFIG_IPV6_ROUTER_PREF
229 .accept_ra_rtr_pref = 1,
230 .rtr_probe_interval = 60 * HZ,
231 #ifdef CONFIG_IPV6_ROUTE_INFO
232 .accept_ra_rt_info_max_plen = 0,
236 .accept_source_route = 0, /* we do not accept RH0 by default. */
241 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
242 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
243 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
244 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
245 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
246 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
247 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
248 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
250 /* Check if a valid qdisc is available */
251 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
253 return !qdisc_tx_is_noop(dev);
256 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
258 if (del_timer(&ifp->timer))
262 enum addrconf_timer_t {
268 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
269 enum addrconf_timer_t what,
272 if (!del_timer(&ifp->timer))
277 ifp->timer.function = addrconf_dad_timer;
280 ifp->timer.function = addrconf_rs_timer;
285 ifp->timer.expires = jiffies + when;
286 add_timer(&ifp->timer);
289 static int snmp6_alloc_dev(struct inet6_dev *idev)
291 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
292 sizeof(struct ipstats_mib),
293 __alignof__(struct ipstats_mib)) < 0)
295 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
297 if (!idev->stats.icmpv6dev)
299 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
301 if (!idev->stats.icmpv6msgdev)
307 kfree(idev->stats.icmpv6dev);
309 snmp_mib_free((void __percpu **)idev->stats.ipv6);
314 static void snmp6_free_dev(struct inet6_dev *idev)
316 kfree(idev->stats.icmpv6msgdev);
317 kfree(idev->stats.icmpv6dev);
318 snmp_mib_free((void __percpu **)idev->stats.ipv6);
321 /* Nobody refers to this device, we may destroy it. */
323 void in6_dev_finish_destroy(struct inet6_dev *idev)
325 struct net_device *dev = idev->dev;
327 WARN_ON(!list_empty(&idev->addr_list));
328 WARN_ON(idev->mc_list != NULL);
330 #ifdef NET_REFCNT_DEBUG
331 pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
335 pr_warn("Freeing alive inet6 device %p\n", idev);
338 snmp6_free_dev(idev);
339 kfree_rcu(idev, rcu);
341 EXPORT_SYMBOL(in6_dev_finish_destroy);
343 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
345 struct inet6_dev *ndev;
349 if (dev->mtu < IPV6_MIN_MTU)
352 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
357 rwlock_init(&ndev->lock);
359 INIT_LIST_HEAD(&ndev->addr_list);
361 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
362 ndev->cnf.mtu6 = dev->mtu;
363 ndev->cnf.sysctl = NULL;
364 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
365 if (ndev->nd_parms == NULL) {
369 if (ndev->cnf.forwarding)
370 dev_disable_lro(dev);
371 /* We refer to the device */
374 if (snmp6_alloc_dev(ndev) < 0) {
376 "%s: cannot allocate memory for statistics; dev=%s.\n",
377 __func__, dev->name));
378 neigh_parms_release(&nd_tbl, ndev->nd_parms);
384 if (snmp6_register_dev(ndev) < 0) {
386 "%s: cannot create /proc/net/dev_snmp6/%s\n",
387 __func__, dev->name));
388 neigh_parms_release(&nd_tbl, ndev->nd_parms);
390 in6_dev_finish_destroy(ndev);
394 /* One reference from device. We must do this before
395 * we invoke __ipv6_regen_rndid().
399 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
400 ndev->cnf.accept_dad = -1;
402 #if IS_ENABLED(CONFIG_IPV6_SIT)
403 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
404 pr_info("%s: Disabled Multicast RS\n", dev->name);
405 ndev->cnf.rtr_solicits = 0;
409 #ifdef CONFIG_IPV6_PRIVACY
410 INIT_LIST_HEAD(&ndev->tempaddr_list);
411 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
412 if ((dev->flags&IFF_LOOPBACK) ||
413 dev->type == ARPHRD_TUNNEL ||
414 dev->type == ARPHRD_TUNNEL6 ||
415 dev->type == ARPHRD_SIT ||
416 dev->type == ARPHRD_NONE) {
417 ndev->cnf.use_tempaddr = -1;
420 ipv6_regen_rndid((unsigned long) ndev);
423 ndev->token = in6addr_any;
425 if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 ndev->if_flags |= IF_READY;
428 ipv6_mc_init_dev(ndev);
429 ndev->tstamp = jiffies;
430 addrconf_sysctl_register(ndev);
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
434 /* Join interface-local all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
437 /* Join all-node multicast group */
438 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
440 /* Join all-router multicast group if forwarding is set */
441 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
442 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
447 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
449 struct inet6_dev *idev;
453 idev = __in6_dev_get(dev);
455 idev = ipv6_add_dev(dev);
460 if (dev->flags&IFF_UP)
465 static int inet6_netconf_msgsize_devconf(int type)
467 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
468 + nla_total_size(4); /* NETCONFA_IFINDEX */
470 /* type -1 is used for ALL */
471 if (type == -1 || type == NETCONFA_FORWARDING)
472 size += nla_total_size(4);
473 #ifdef CONFIG_IPV6_MROUTE
474 if (type == -1 || type == NETCONFA_MC_FORWARDING)
475 size += nla_total_size(4);
481 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
482 struct ipv6_devconf *devconf, u32 portid,
483 u32 seq, int event, unsigned int flags,
486 struct nlmsghdr *nlh;
487 struct netconfmsg *ncm;
489 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
494 ncm = nlmsg_data(nlh);
495 ncm->ncm_family = AF_INET6;
497 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
498 goto nla_put_failure;
500 /* type -1 is used for ALL */
501 if ((type == -1 || type == NETCONFA_FORWARDING) &&
502 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
503 goto nla_put_failure;
504 #ifdef CONFIG_IPV6_MROUTE
505 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
506 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
507 devconf->mc_forwarding) < 0)
508 goto nla_put_failure;
510 return nlmsg_end(skb, nlh);
513 nlmsg_cancel(skb, nlh);
517 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
518 struct ipv6_devconf *devconf)
523 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
527 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
528 RTM_NEWNETCONF, 0, type);
530 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
531 WARN_ON(err == -EMSGSIZE);
535 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
538 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
541 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
542 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
543 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
546 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
547 struct nlmsghdr *nlh)
549 struct net *net = sock_net(in_skb->sk);
550 struct nlattr *tb[NETCONFA_MAX+1];
551 struct netconfmsg *ncm;
553 struct ipv6_devconf *devconf;
554 struct inet6_dev *in6_dev;
555 struct net_device *dev;
559 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
560 devconf_ipv6_policy);
565 if (!tb[NETCONFA_IFINDEX])
568 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
570 case NETCONFA_IFINDEX_ALL:
571 devconf = net->ipv6.devconf_all;
573 case NETCONFA_IFINDEX_DEFAULT:
574 devconf = net->ipv6.devconf_dflt;
577 dev = __dev_get_by_index(net, ifindex);
580 in6_dev = __in6_dev_get(dev);
583 devconf = &in6_dev->cnf;
588 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
592 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
593 NETLINK_CB(in_skb).portid,
594 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
597 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
598 WARN_ON(err == -EMSGSIZE);
602 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
607 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
608 struct netlink_callback *cb)
610 struct net *net = sock_net(skb->sk);
613 struct net_device *dev;
614 struct inet6_dev *idev;
615 struct hlist_head *head;
618 s_idx = idx = cb->args[1];
620 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
622 head = &net->dev_index_head[h];
624 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
626 hlist_for_each_entry_rcu(dev, head, index_hlist) {
629 idev = __in6_dev_get(dev);
633 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
635 NETLINK_CB(cb->skb).portid,
643 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
649 if (h == NETDEV_HASHENTRIES) {
650 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
651 net->ipv6.devconf_all,
652 NETLINK_CB(cb->skb).portid,
654 RTM_NEWNETCONF, NLM_F_MULTI,
660 if (h == NETDEV_HASHENTRIES + 1) {
661 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
662 net->ipv6.devconf_dflt,
663 NETLINK_CB(cb->skb).portid,
665 RTM_NEWNETCONF, NLM_F_MULTI,
679 static void dev_forward_change(struct inet6_dev *idev)
681 struct net_device *dev;
682 struct inet6_ifaddr *ifa;
687 if (idev->cnf.forwarding)
688 dev_disable_lro(dev);
689 if (dev->flags & IFF_MULTICAST) {
690 if (idev->cnf.forwarding) {
691 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
692 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
693 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
695 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
696 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
697 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
701 list_for_each_entry(ifa, &idev->addr_list, if_list) {
702 if (ifa->flags&IFA_F_TENTATIVE)
704 if (idev->cnf.forwarding)
705 addrconf_join_anycast(ifa);
707 addrconf_leave_anycast(ifa);
709 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
710 dev->ifindex, &idev->cnf);
714 static void addrconf_forward_change(struct net *net, __s32 newf)
716 struct net_device *dev;
717 struct inet6_dev *idev;
719 for_each_netdev(net, dev) {
720 idev = __in6_dev_get(dev);
722 int changed = (!idev->cnf.forwarding) ^ (!newf);
723 idev->cnf.forwarding = newf;
725 dev_forward_change(idev);
730 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
736 return restart_syscall();
738 net = (struct net *)table->extra2;
742 if (p == &net->ipv6.devconf_dflt->forwarding) {
743 if ((!newf) ^ (!old))
744 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
745 NETCONFA_IFINDEX_DEFAULT,
746 net->ipv6.devconf_dflt);
751 if (p == &net->ipv6.devconf_all->forwarding) {
752 net->ipv6.devconf_dflt->forwarding = newf;
753 addrconf_forward_change(net, newf);
754 if ((!newf) ^ (!old))
755 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
756 NETCONFA_IFINDEX_ALL,
757 net->ipv6.devconf_all);
758 } else if ((!newf) ^ (!old))
759 dev_forward_change((struct inet6_dev *)table->extra1);
763 rt6_purge_dflt_routers(net);
768 /* Nobody refers to this ifaddr, destroy it */
769 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
771 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
773 #ifdef NET_REFCNT_DEBUG
774 pr_debug("%s\n", __func__);
777 in6_dev_put(ifp->idev);
779 if (del_timer(&ifp->timer))
780 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
782 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
783 pr_warn("Freeing alive inet6 address %p\n", ifp);
792 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
795 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
798 * Each device address list is sorted in order of scope -
799 * global before linklocal.
801 list_for_each(p, &idev->addr_list) {
802 struct inet6_ifaddr *ifa
803 = list_entry(p, struct inet6_ifaddr, if_list);
804 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
808 list_add_tail(&ifp->if_list, p);
811 static u32 inet6_addr_hash(const struct in6_addr *addr)
813 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
816 /* On success it returns ifp with increased reference count */
818 static struct inet6_ifaddr *
819 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
820 int scope, u32 flags)
822 struct inet6_ifaddr *ifa = NULL;
826 int addr_type = ipv6_addr_type(addr);
828 if (addr_type == IPV6_ADDR_ANY ||
829 addr_type & IPV6_ADDR_MULTICAST ||
830 (!(idev->dev->flags & IFF_LOOPBACK) &&
831 addr_type & IPV6_ADDR_LOOPBACK))
832 return ERR_PTR(-EADDRNOTAVAIL);
836 err = -ENODEV; /*XXX*/
840 if (idev->cnf.disable_ipv6) {
845 spin_lock(&addrconf_hash_lock);
847 /* Ignore adding duplicate addresses on an interface */
848 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
849 ADBG(("ipv6_add_addr: already assigned\n"));
854 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
857 ADBG(("ipv6_add_addr: malloc failed\n"));
862 rt = addrconf_dst_alloc(idev, addr, false);
870 spin_lock_init(&ifa->lock);
871 spin_lock_init(&ifa->state_lock);
872 init_timer(&ifa->timer);
873 INIT_HLIST_NODE(&ifa->addr_lst);
874 ifa->timer.data = (unsigned long) ifa;
876 ifa->prefix_len = pfxlen;
877 ifa->flags = flags | IFA_F_TENTATIVE;
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_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 = !max_addresses ||
1128 ipv6_count_addresses(idev) < max_addresses ?
1129 ipv6_add_addr(idev, &addr, tmp_plen,
1130 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1132 if (IS_ERR_OR_NULL(ift)) {
1135 pr_info("%s: retry temporary address regeneration\n", __func__);
1137 write_lock(&idev->lock);
1141 spin_lock_bh(&ift->lock);
1143 ift->valid_lft = tmp_valid_lft;
1144 ift->prefered_lft = tmp_prefered_lft;
1146 ift->tstamp = tmp_tstamp;
1147 spin_unlock_bh(&ift->lock);
1149 addrconf_dad_start(ift);
1158 * Choose an appropriate source address (RFC3484)
1161 IPV6_SADDR_RULE_INIT = 0,
1162 IPV6_SADDR_RULE_LOCAL,
1163 IPV6_SADDR_RULE_SCOPE,
1164 IPV6_SADDR_RULE_PREFERRED,
1165 #ifdef CONFIG_IPV6_MIP6
1166 IPV6_SADDR_RULE_HOA,
1168 IPV6_SADDR_RULE_OIF,
1169 IPV6_SADDR_RULE_LABEL,
1170 #ifdef CONFIG_IPV6_PRIVACY
1171 IPV6_SADDR_RULE_PRIVACY,
1173 IPV6_SADDR_RULE_ORCHID,
1174 IPV6_SADDR_RULE_PREFIX,
1178 struct ipv6_saddr_score {
1181 struct inet6_ifaddr *ifa;
1182 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1187 struct ipv6_saddr_dst {
1188 const struct in6_addr *addr;
1195 static inline int ipv6_saddr_preferred(int type)
1197 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1202 static int ipv6_get_saddr_eval(struct net *net,
1203 struct ipv6_saddr_score *score,
1204 struct ipv6_saddr_dst *dst,
1209 if (i <= score->rule) {
1211 case IPV6_SADDR_RULE_SCOPE:
1212 ret = score->scopedist;
1214 case IPV6_SADDR_RULE_PREFIX:
1215 ret = score->matchlen;
1218 ret = !!test_bit(i, score->scorebits);
1224 case IPV6_SADDR_RULE_INIT:
1225 /* Rule 0: remember if hiscore is not ready yet */
1228 case IPV6_SADDR_RULE_LOCAL:
1229 /* Rule 1: Prefer same address */
1230 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1232 case IPV6_SADDR_RULE_SCOPE:
1233 /* Rule 2: Prefer appropriate scope
1238 * ---+--+-+---> scope
1240 * | d is scope of the destination.
1242 * | \ <- smaller scope is better if
1243 * B-15 | \ if scope is enough for destinaion.
1244 * | ret = B - scope (-1 <= scope >= d <= 15).
1246 * |/ <- greater is better
1247 * -C / if scope is not enough for destination.
1248 * /| ret = scope - C (-1 <= d < scope <= 15).
1250 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1251 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1252 * Assume B = 0 and we get C > 29.
1254 ret = __ipv6_addr_src_scope(score->addr_type);
1255 if (ret >= dst->scope)
1258 ret -= 128; /* 30 is enough */
1259 score->scopedist = ret;
1261 case IPV6_SADDR_RULE_PREFERRED:
1262 /* Rule 3: Avoid deprecated and optimistic addresses */
1263 ret = ipv6_saddr_preferred(score->addr_type) ||
1264 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1266 #ifdef CONFIG_IPV6_MIP6
1267 case IPV6_SADDR_RULE_HOA:
1269 /* Rule 4: Prefer home address */
1270 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1271 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1275 case IPV6_SADDR_RULE_OIF:
1276 /* Rule 5: Prefer outgoing interface */
1277 ret = (!dst->ifindex ||
1278 dst->ifindex == score->ifa->idev->dev->ifindex);
1280 case IPV6_SADDR_RULE_LABEL:
1281 /* Rule 6: Prefer matching label */
1282 ret = ipv6_addr_label(net,
1283 &score->ifa->addr, score->addr_type,
1284 score->ifa->idev->dev->ifindex) == dst->label;
1286 #ifdef CONFIG_IPV6_PRIVACY
1287 case IPV6_SADDR_RULE_PRIVACY:
1289 /* Rule 7: Prefer public address
1290 * Note: prefer temporary address if use_tempaddr >= 2
1292 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1293 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1294 score->ifa->idev->cnf.use_tempaddr >= 2;
1295 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1299 case IPV6_SADDR_RULE_ORCHID:
1300 /* Rule 8-: Prefer ORCHID vs ORCHID or
1301 * non-ORCHID vs non-ORCHID
1303 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1304 ipv6_addr_orchid(dst->addr));
1306 case IPV6_SADDR_RULE_PREFIX:
1307 /* Rule 8: Use longest matching prefix */
1308 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1309 if (ret > score->ifa->prefix_len)
1310 ret = score->ifa->prefix_len;
1311 score->matchlen = ret;
1318 __set_bit(i, score->scorebits);
1324 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1325 const struct in6_addr *daddr, unsigned int prefs,
1326 struct in6_addr *saddr)
1328 struct ipv6_saddr_score scores[2],
1329 *score = &scores[0], *hiscore = &scores[1];
1330 struct ipv6_saddr_dst dst;
1331 struct net_device *dev;
1334 dst_type = __ipv6_addr_type(daddr);
1336 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1337 dst.scope = __ipv6_addr_src_scope(dst_type);
1338 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1342 hiscore->ifa = NULL;
1346 for_each_netdev_rcu(net, dev) {
1347 struct inet6_dev *idev;
1349 /* Candidate Source Address (section 4)
1350 * - multicast and link-local destination address,
1351 * the set of candidate source address MUST only
1352 * include addresses assigned to interfaces
1353 * belonging to the same link as the outgoing
1355 * (- For site-local destination addresses, the
1356 * set of candidate source addresses MUST only
1357 * include addresses assigned to interfaces
1358 * belonging to the same site as the outgoing
1361 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1362 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1363 dst.ifindex && dev->ifindex != dst.ifindex)
1366 idev = __in6_dev_get(dev);
1370 read_lock_bh(&idev->lock);
1371 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1375 * - Tentative Address (RFC2462 section 5.4)
1376 * - A tentative address is not considered
1377 * "assigned to an interface" in the traditional
1378 * sense, unless it is also flagged as optimistic.
1379 * - Candidate Source Address (section 4)
1380 * - In any case, anycast addresses, multicast
1381 * addresses, and the unspecified address MUST
1382 * NOT be included in a candidate set.
1384 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1385 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1388 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1390 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1391 score->addr_type & IPV6_ADDR_MULTICAST)) {
1392 LIMIT_NETDEBUG(KERN_DEBUG
1393 "ADDRCONF: unspecified / multicast address "
1394 "assigned as unicast address on %s",
1400 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1402 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1403 int minihiscore, miniscore;
1405 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1406 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1408 if (minihiscore > miniscore) {
1409 if (i == IPV6_SADDR_RULE_SCOPE &&
1410 score->scopedist > 0) {
1413 * each remaining entry
1414 * has too small (not enough)
1415 * scope, because ifa entries
1416 * are sorted by their scope
1422 } else if (minihiscore < miniscore) {
1424 in6_ifa_put(hiscore->ifa);
1426 in6_ifa_hold(score->ifa);
1428 swap(hiscore, score);
1430 /* restore our iterator */
1431 score->ifa = hiscore->ifa;
1438 read_unlock_bh(&idev->lock);
1443 return -EADDRNOTAVAIL;
1445 *saddr = hiscore->ifa->addr;
1446 in6_ifa_put(hiscore->ifa);
1449 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1451 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1452 unsigned char banned_flags)
1454 struct inet6_dev *idev;
1455 int err = -EADDRNOTAVAIL;
1458 idev = __in6_dev_get(dev);
1460 struct inet6_ifaddr *ifp;
1462 read_lock_bh(&idev->lock);
1463 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1464 if (ifp->scope == IFA_LINK &&
1465 !(ifp->flags & banned_flags)) {
1471 read_unlock_bh(&idev->lock);
1477 static int ipv6_count_addresses(struct inet6_dev *idev)
1480 struct inet6_ifaddr *ifp;
1482 read_lock_bh(&idev->lock);
1483 list_for_each_entry(ifp, &idev->addr_list, if_list)
1485 read_unlock_bh(&idev->lock);
1489 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1490 struct net_device *dev, int strict)
1492 struct inet6_ifaddr *ifp;
1493 unsigned int hash = inet6_addr_hash(addr);
1496 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1497 if (!net_eq(dev_net(ifp->idev->dev), net))
1499 if (ipv6_addr_equal(&ifp->addr, addr) &&
1500 !(ifp->flags&IFA_F_TENTATIVE) &&
1501 (dev == NULL || ifp->idev->dev == dev ||
1502 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1503 rcu_read_unlock_bh();
1508 rcu_read_unlock_bh();
1511 EXPORT_SYMBOL(ipv6_chk_addr);
1513 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1514 struct net_device *dev)
1516 unsigned int hash = inet6_addr_hash(addr);
1517 struct inet6_ifaddr *ifp;
1519 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1520 if (!net_eq(dev_net(ifp->idev->dev), net))
1522 if (ipv6_addr_equal(&ifp->addr, addr)) {
1523 if (dev == NULL || ifp->idev->dev == dev)
1530 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1532 struct inet6_dev *idev;
1533 struct inet6_ifaddr *ifa;
1538 idev = __in6_dev_get(dev);
1540 read_lock_bh(&idev->lock);
1541 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1542 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1547 read_unlock_bh(&idev->lock);
1552 EXPORT_SYMBOL(ipv6_chk_prefix);
1554 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1555 struct net_device *dev, int strict)
1557 struct inet6_ifaddr *ifp, *result = NULL;
1558 unsigned int hash = inet6_addr_hash(addr);
1561 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1562 if (!net_eq(dev_net(ifp->idev->dev), net))
1564 if (ipv6_addr_equal(&ifp->addr, addr)) {
1565 if (dev == NULL || ifp->idev->dev == dev ||
1566 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1573 rcu_read_unlock_bh();
1578 /* Gets referenced address, destroys ifaddr */
1580 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1582 if (ifp->flags&IFA_F_PERMANENT) {
1583 spin_lock_bh(&ifp->lock);
1584 addrconf_del_timer(ifp);
1585 ifp->flags |= IFA_F_TENTATIVE;
1587 ifp->flags |= IFA_F_DADFAILED;
1588 spin_unlock_bh(&ifp->lock);
1590 ipv6_ifa_notify(0, ifp);
1592 #ifdef CONFIG_IPV6_PRIVACY
1593 } else if (ifp->flags&IFA_F_TEMPORARY) {
1594 struct inet6_ifaddr *ifpub;
1595 spin_lock_bh(&ifp->lock);
1598 in6_ifa_hold(ifpub);
1599 spin_unlock_bh(&ifp->lock);
1600 ipv6_create_tempaddr(ifpub, ifp);
1603 spin_unlock_bh(&ifp->lock);
1611 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1615 spin_lock(&ifp->state_lock);
1616 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1617 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1620 spin_unlock(&ifp->state_lock);
1625 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1627 struct inet6_dev *idev = ifp->idev;
1629 if (addrconf_dad_end(ifp)) {
1634 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1635 ifp->idev->dev->name, &ifp->addr);
1637 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1638 struct in6_addr addr;
1640 addr.s6_addr32[0] = htonl(0xfe800000);
1641 addr.s6_addr32[1] = 0;
1643 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1644 ipv6_addr_equal(&ifp->addr, &addr)) {
1645 /* DAD failed for link-local based on MAC address */
1646 idev->cnf.disable_ipv6 = 1;
1648 pr_info("%s: IPv6 being disabled!\n",
1649 ifp->idev->dev->name);
1653 addrconf_dad_stop(ifp, 1);
1656 /* Join to solicited addr multicast group. */
1658 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1660 struct in6_addr maddr;
1662 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1665 addrconf_addr_solict_mult(addr, &maddr);
1666 ipv6_dev_mc_inc(dev, &maddr);
1669 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1671 struct in6_addr maddr;
1673 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1676 addrconf_addr_solict_mult(addr, &maddr);
1677 __ipv6_dev_mc_dec(idev, &maddr);
1680 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1682 struct in6_addr addr;
1683 if (ifp->prefix_len == 127) /* RFC 6164 */
1685 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1686 if (ipv6_addr_any(&addr))
1688 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1691 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1693 struct in6_addr addr;
1694 if (ifp->prefix_len == 127) /* RFC 6164 */
1696 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1697 if (ipv6_addr_any(&addr))
1699 __ipv6_dev_ac_dec(ifp->idev, &addr);
1702 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1704 if (dev->addr_len != ETH_ALEN)
1706 memcpy(eui, dev->dev_addr, 3);
1707 memcpy(eui + 5, dev->dev_addr + 3, 3);
1710 * The zSeries OSA network cards can be shared among various
1711 * OS instances, but the OSA cards have only one MAC address.
1712 * This leads to duplicate address conflicts in conjunction
1713 * with IPv6 if more than one instance uses the same card.
1715 * The driver for these cards can deliver a unique 16-bit
1716 * identifier for each instance sharing the same card. It is
1717 * placed instead of 0xFFFE in the interface identifier. The
1718 * "u" bit of the interface identifier is not inverted in this
1719 * case. Hence the resulting interface identifier has local
1720 * scope according to RFC2373.
1723 eui[3] = (dev->dev_id >> 8) & 0xFF;
1724 eui[4] = dev->dev_id & 0xFF;
1733 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1735 if (dev->addr_len != IEEE802154_ADDR_LEN)
1737 memcpy(eui, dev->dev_addr, 8);
1742 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1744 union fwnet_hwaddr *ha;
1746 if (dev->addr_len != FWNET_ALEN)
1749 ha = (union fwnet_hwaddr *)dev->dev_addr;
1751 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1756 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1758 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1759 if (dev->addr_len != ARCNET_ALEN)
1762 eui[7] = *(u8 *)dev->dev_addr;
1766 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1768 if (dev->addr_len != INFINIBAND_ALEN)
1770 memcpy(eui, dev->dev_addr + 12, 8);
1775 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1779 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1780 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1781 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1782 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1783 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1784 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1788 memcpy(eui + 4, &addr, 4);
1792 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1794 if (dev->priv_flags & IFF_ISATAP)
1795 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1799 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1801 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1804 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1806 switch (dev->type) {
1809 return addrconf_ifid_eui48(eui, dev);
1811 return addrconf_ifid_arcnet(eui, dev);
1812 case ARPHRD_INFINIBAND:
1813 return addrconf_ifid_infiniband(eui, dev);
1815 return addrconf_ifid_sit(eui, dev);
1817 return addrconf_ifid_gre(eui, dev);
1818 case ARPHRD_IEEE802154:
1819 return addrconf_ifid_eui64(eui, dev);
1820 case ARPHRD_IEEE1394:
1821 return addrconf_ifid_ieee1394(eui, dev);
1826 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1829 struct inet6_ifaddr *ifp;
1831 read_lock_bh(&idev->lock);
1832 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1833 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1834 memcpy(eui, ifp->addr.s6_addr+8, 8);
1839 read_unlock_bh(&idev->lock);
1843 #ifdef CONFIG_IPV6_PRIVACY
1844 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1845 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1848 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1849 idev->rndid[0] &= ~0x02;
1852 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1853 * check if generated address is not inappropriate
1855 * - Reserved subnet anycast (RFC 2526)
1856 * 11111101 11....11 1xxxxxxx
1857 * - ISATAP (RFC4214) 6.1
1858 * 00-00-5E-FE-xx-xx-xx-xx
1860 * - XXX: already assigned to an address on the device
1862 if (idev->rndid[0] == 0xfd &&
1863 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1864 (idev->rndid[7]&0x80))
1866 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1867 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1869 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1874 static void ipv6_regen_rndid(unsigned long data)
1876 struct inet6_dev *idev = (struct inet6_dev *) data;
1877 unsigned long expires;
1880 write_lock_bh(&idev->lock);
1885 __ipv6_regen_rndid(idev);
1888 idev->cnf.temp_prefered_lft * HZ -
1889 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1890 idev->cnf.max_desync_factor * HZ;
1891 if (time_before(expires, jiffies)) {
1892 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1893 __func__, idev->dev->name);
1897 if (!mod_timer(&idev->regen_timer, expires))
1901 write_unlock_bh(&idev->lock);
1902 rcu_read_unlock_bh();
1906 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1908 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1909 __ipv6_regen_rndid(idev);
1918 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1919 unsigned long expires, u32 flags)
1921 struct fib6_config cfg = {
1922 .fc_table = RT6_TABLE_PREFIX,
1923 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1924 .fc_ifindex = dev->ifindex,
1925 .fc_expires = expires,
1927 .fc_flags = RTF_UP | flags,
1928 .fc_nlinfo.nl_net = dev_net(dev),
1929 .fc_protocol = RTPROT_KERNEL,
1934 /* Prevent useless cloning on PtP SIT.
1935 This thing is done here expecting that the whole
1936 class of non-broadcast devices need not cloning.
1938 #if IS_ENABLED(CONFIG_IPV6_SIT)
1939 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1940 cfg.fc_flags |= RTF_NONEXTHOP;
1943 ip6_route_add(&cfg);
1947 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1949 const struct net_device *dev,
1950 u32 flags, u32 noflags)
1952 struct fib6_node *fn;
1953 struct rt6_info *rt = NULL;
1954 struct fib6_table *table;
1956 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1960 read_lock_bh(&table->tb6_lock);
1961 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1964 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1965 if (rt->dst.dev->ifindex != dev->ifindex)
1967 if ((rt->rt6i_flags & flags) != flags)
1969 if ((rt->rt6i_flags & noflags) != 0)
1975 read_unlock_bh(&table->tb6_lock);
1980 /* Create "default" multicast route to the interface */
1982 static void addrconf_add_mroute(struct net_device *dev)
1984 struct fib6_config cfg = {
1985 .fc_table = RT6_TABLE_LOCAL,
1986 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1987 .fc_ifindex = dev->ifindex,
1990 .fc_nlinfo.nl_net = dev_net(dev),
1993 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1995 ip6_route_add(&cfg);
1998 #if IS_ENABLED(CONFIG_IPV6_SIT)
1999 static void sit_route_add(struct net_device *dev)
2001 struct fib6_config cfg = {
2002 .fc_table = RT6_TABLE_MAIN,
2003 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2004 .fc_ifindex = dev->ifindex,
2006 .fc_flags = RTF_UP | RTF_NONEXTHOP,
2007 .fc_nlinfo.nl_net = dev_net(dev),
2010 /* prefix length - 96 bits "::d.d.d.d" */
2011 ip6_route_add(&cfg);
2015 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2017 struct inet6_dev *idev;
2021 idev = ipv6_find_idev(dev);
2023 return ERR_PTR(-ENOBUFS);
2025 if (idev->cnf.disable_ipv6)
2026 return ERR_PTR(-EACCES);
2028 /* Add default multicast route */
2029 if (!(dev->flags & IFF_LOOPBACK))
2030 addrconf_add_mroute(dev);
2035 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2037 struct prefix_info *pinfo;
2041 struct inet6_dev *in6_dev;
2042 struct net *net = dev_net(dev);
2044 pinfo = (struct prefix_info *) opt;
2046 if (len < sizeof(struct prefix_info)) {
2047 ADBG(("addrconf: prefix option too short\n"));
2052 * Validation checks ([ADDRCONF], page 19)
2055 addr_type = ipv6_addr_type(&pinfo->prefix);
2057 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2060 valid_lft = ntohl(pinfo->valid);
2061 prefered_lft = ntohl(pinfo->prefered);
2063 if (prefered_lft > valid_lft) {
2064 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2068 in6_dev = in6_dev_get(dev);
2070 if (in6_dev == NULL) {
2071 net_dbg_ratelimited("addrconf: device %s not configured\n",
2077 * Two things going on here:
2078 * 1) Add routes for on-link prefixes
2079 * 2) Configure prefixes with the auto flag set
2082 if (pinfo->onlink) {
2083 struct rt6_info *rt;
2084 unsigned long rt_expires;
2086 /* Avoid arithmetic overflow. Really, we could
2087 * save rt_expires in seconds, likely valid_lft,
2088 * but it would require division in fib gc, that it
2092 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2094 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2096 if (addrconf_finite_timeout(rt_expires))
2099 rt = addrconf_get_prefix_route(&pinfo->prefix,
2102 RTF_ADDRCONF | RTF_PREFIX_RT,
2103 RTF_GATEWAY | RTF_DEFAULT);
2106 /* Autoconf prefix route */
2107 if (valid_lft == 0) {
2110 } else if (addrconf_finite_timeout(rt_expires)) {
2112 rt6_set_expires(rt, jiffies + rt_expires);
2114 rt6_clean_expires(rt);
2116 } else if (valid_lft) {
2117 clock_t expires = 0;
2118 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2119 if (addrconf_finite_timeout(rt_expires)) {
2121 flags |= RTF_EXPIRES;
2122 expires = jiffies_to_clock_t(rt_expires);
2124 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2125 dev, expires, flags);
2130 /* Try to figure out our local address for this prefix */
2132 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2133 struct inet6_ifaddr *ifp;
2134 struct in6_addr addr;
2135 int create = 0, update_lft = 0;
2136 bool tokenized = false;
2138 if (pinfo->prefix_len == 64) {
2139 memcpy(&addr, &pinfo->prefix, 8);
2141 if (!ipv6_addr_any(&in6_dev->token)) {
2142 read_lock_bh(&in6_dev->lock);
2143 memcpy(addr.s6_addr + 8,
2144 in6_dev->token.s6_addr + 8, 8);
2145 read_unlock_bh(&in6_dev->lock);
2147 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2148 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2149 in6_dev_put(in6_dev);
2154 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2156 in6_dev_put(in6_dev);
2161 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2163 if (ifp == NULL && valid_lft) {
2164 int max_addresses = in6_dev->cnf.max_addresses;
2167 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2168 if (in6_dev->cnf.optimistic_dad &&
2169 !net->ipv6.devconf_all->forwarding && sllao)
2170 addr_flags = IFA_F_OPTIMISTIC;
2173 /* Do not allow to create too much of autoconfigured
2174 * addresses; this would be too easy way to crash kernel.
2176 if (!max_addresses ||
2177 ipv6_count_addresses(in6_dev) < max_addresses)
2178 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2179 addr_type&IPV6_ADDR_SCOPE_MASK,
2182 if (IS_ERR_OR_NULL(ifp)) {
2183 in6_dev_put(in6_dev);
2187 update_lft = create = 1;
2188 ifp->cstamp = jiffies;
2189 ifp->tokenized = tokenized;
2190 addrconf_dad_start(ifp);
2196 #ifdef CONFIG_IPV6_PRIVACY
2197 struct inet6_ifaddr *ift;
2201 /* update lifetime (RFC2462 5.5.3 e) */
2202 spin_lock(&ifp->lock);
2204 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2205 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2208 if (!update_lft && stored_lft) {
2209 if (valid_lft > MIN_VALID_LIFETIME ||
2210 valid_lft > stored_lft)
2212 else if (stored_lft <= MIN_VALID_LIFETIME) {
2213 /* valid_lft <= stored_lft is always true */
2215 * RFC 4862 Section 5.5.3e:
2216 * "Note that the preferred lifetime of
2217 * the corresponding address is always
2218 * reset to the Preferred Lifetime in
2219 * the received Prefix Information
2220 * option, regardless of whether the
2221 * valid lifetime is also reset or
2224 * So if the preferred lifetime in
2225 * this advertisement is different
2226 * than what we have stored, but the
2227 * valid lifetime is invalid, just
2228 * reset prefered_lft.
2230 * We must set the valid lifetime
2231 * to the stored lifetime since we'll
2232 * be updating the timestamp below,
2233 * else we'll set it back to the
2236 if (prefered_lft != ifp->prefered_lft) {
2237 valid_lft = stored_lft;
2241 valid_lft = MIN_VALID_LIFETIME;
2242 if (valid_lft < prefered_lft)
2243 prefered_lft = valid_lft;
2249 ifp->valid_lft = valid_lft;
2250 ifp->prefered_lft = prefered_lft;
2253 ifp->flags &= ~IFA_F_DEPRECATED;
2254 spin_unlock(&ifp->lock);
2256 if (!(flags&IFA_F_TENTATIVE))
2257 ipv6_ifa_notify(0, ifp);
2259 spin_unlock(&ifp->lock);
2261 #ifdef CONFIG_IPV6_PRIVACY
2262 read_lock_bh(&in6_dev->lock);
2263 /* update all temporary addresses in the list */
2264 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2266 int age, max_valid, max_prefered;
2268 if (ifp != ift->ifpub)
2272 * RFC 4941 section 3.3:
2273 * If a received option will extend the lifetime
2274 * of a public address, the lifetimes of
2275 * temporary addresses should be extended,
2276 * subject to the overall constraint that no
2277 * temporary addresses should ever remain
2278 * "valid" or "preferred" for a time longer than
2279 * (TEMP_VALID_LIFETIME) or
2280 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2283 age = (now - ift->cstamp) / HZ;
2284 max_valid = in6_dev->cnf.temp_valid_lft - age;
2288 max_prefered = in6_dev->cnf.temp_prefered_lft -
2289 in6_dev->cnf.max_desync_factor -
2291 if (max_prefered < 0)
2294 if (valid_lft > max_valid)
2295 valid_lft = max_valid;
2297 if (prefered_lft > max_prefered)
2298 prefered_lft = max_prefered;
2300 spin_lock(&ift->lock);
2302 ift->valid_lft = valid_lft;
2303 ift->prefered_lft = prefered_lft;
2305 if (prefered_lft > 0)
2306 ift->flags &= ~IFA_F_DEPRECATED;
2308 spin_unlock(&ift->lock);
2309 if (!(flags&IFA_F_TENTATIVE))
2310 ipv6_ifa_notify(0, ift);
2313 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2315 * When a new public address is created as
2316 * described in [ADDRCONF], also create a new
2317 * temporary address. Also create a temporary
2318 * address if it's enabled but no temporary
2319 * address currently exists.
2321 read_unlock_bh(&in6_dev->lock);
2322 ipv6_create_tempaddr(ifp, NULL);
2324 read_unlock_bh(&in6_dev->lock);
2331 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2332 in6_dev_put(in6_dev);
2336 * Set destination address.
2337 * Special case for SIT interfaces where we create a new "virtual"
2340 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2342 struct in6_ifreq ireq;
2343 struct net_device *dev;
2349 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2352 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2358 #if IS_ENABLED(CONFIG_IPV6_SIT)
2359 if (dev->type == ARPHRD_SIT) {
2360 const struct net_device_ops *ops = dev->netdev_ops;
2362 struct ip_tunnel_parm p;
2364 err = -EADDRNOTAVAIL;
2365 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2368 memset(&p, 0, sizeof(p));
2369 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2373 p.iph.protocol = IPPROTO_IPV6;
2375 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2377 if (ops->ndo_do_ioctl) {
2378 mm_segment_t oldfs = get_fs();
2381 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2388 dev = __dev_get_by_name(net, p.name);
2391 err = dev_open(dev);
2402 * Manual configuration of address on an interface
2404 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2405 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2408 struct inet6_ifaddr *ifp;
2409 struct inet6_dev *idev;
2410 struct net_device *dev;
2414 unsigned long timeout;
2421 /* check the lifetime */
2422 if (!valid_lft || prefered_lft > valid_lft)
2425 dev = __dev_get_by_index(net, ifindex);
2429 idev = addrconf_add_dev(dev);
2431 return PTR_ERR(idev);
2433 scope = ipv6_addr_scope(pfx);
2435 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2436 if (addrconf_finite_timeout(timeout)) {
2437 expires = jiffies_to_clock_t(timeout * HZ);
2438 valid_lft = timeout;
2439 flags = RTF_EXPIRES;
2443 ifa_flags |= IFA_F_PERMANENT;
2446 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2447 if (addrconf_finite_timeout(timeout)) {
2449 ifa_flags |= IFA_F_DEPRECATED;
2450 prefered_lft = timeout;
2453 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2456 spin_lock_bh(&ifp->lock);
2457 ifp->valid_lft = valid_lft;
2458 ifp->prefered_lft = prefered_lft;
2459 ifp->tstamp = jiffies;
2460 spin_unlock_bh(&ifp->lock);
2462 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2465 * Note that section 3.1 of RFC 4429 indicates
2466 * that the Optimistic flag should not be set for
2467 * manually configured addresses
2469 addrconf_dad_start(ifp);
2475 return PTR_ERR(ifp);
2478 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2481 struct inet6_ifaddr *ifp;
2482 struct inet6_dev *idev;
2483 struct net_device *dev;
2488 dev = __dev_get_by_index(net, ifindex);
2492 if ((idev = __in6_dev_get(dev)) == NULL)
2495 read_lock_bh(&idev->lock);
2496 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2497 if (ifp->prefix_len == plen &&
2498 ipv6_addr_equal(pfx, &ifp->addr)) {
2500 read_unlock_bh(&idev->lock);
2504 /* If the last address is deleted administratively,
2505 disable IPv6 on this interface.
2507 if (list_empty(&idev->addr_list))
2508 addrconf_ifdown(idev->dev, 1);
2512 read_unlock_bh(&idev->lock);
2513 return -EADDRNOTAVAIL;
2517 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2519 struct in6_ifreq ireq;
2522 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2525 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2529 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2530 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2531 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2536 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2538 struct in6_ifreq ireq;
2541 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2544 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2548 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2549 ireq.ifr6_prefixlen);
2554 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2555 int plen, int scope)
2557 struct inet6_ifaddr *ifp;
2559 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2561 spin_lock_bh(&ifp->lock);
2562 ifp->flags &= ~IFA_F_TENTATIVE;
2563 spin_unlock_bh(&ifp->lock);
2564 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2569 #if IS_ENABLED(CONFIG_IPV6_SIT)
2570 static void sit_add_v4_addrs(struct inet6_dev *idev)
2572 struct in6_addr addr;
2573 struct net_device *dev;
2574 struct net *net = dev_net(idev->dev);
2579 memset(&addr, 0, sizeof(struct in6_addr));
2580 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2582 if (idev->dev->flags&IFF_POINTOPOINT) {
2583 addr.s6_addr32[0] = htonl(0xfe800000);
2586 scope = IPV6_ADDR_COMPATv4;
2589 if (addr.s6_addr32[3]) {
2590 add_addr(idev, &addr, 128, scope);
2594 for_each_netdev(net, dev) {
2595 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2596 if (in_dev && (dev->flags & IFF_UP)) {
2597 struct in_ifaddr *ifa;
2601 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2604 addr.s6_addr32[3] = ifa->ifa_local;
2606 if (ifa->ifa_scope == RT_SCOPE_LINK)
2608 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2609 if (idev->dev->flags&IFF_POINTOPOINT)
2613 if (idev->dev->flags&IFF_POINTOPOINT)
2618 add_addr(idev, &addr, plen, flag);
2625 static void init_loopback(struct net_device *dev)
2627 struct inet6_dev *idev;
2628 struct net_device *sp_dev;
2629 struct inet6_ifaddr *sp_ifa;
2630 struct rt6_info *sp_rt;
2636 if ((idev = ipv6_find_idev(dev)) == NULL) {
2637 pr_debug("%s: add_dev failed\n", __func__);
2641 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2643 /* Add routes to other interface's IPv6 addresses */
2644 for_each_netdev(dev_net(dev), sp_dev) {
2645 if (!strcmp(sp_dev->name, dev->name))
2648 idev = __in6_dev_get(sp_dev);
2652 read_lock_bh(&idev->lock);
2653 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2655 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2658 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2660 /* Failure cases are ignored */
2664 read_unlock_bh(&idev->lock);
2668 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2670 struct inet6_ifaddr *ifp;
2671 u32 addr_flags = IFA_F_PERMANENT;
2673 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2674 if (idev->cnf.optimistic_dad &&
2675 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2676 addr_flags |= IFA_F_OPTIMISTIC;
2680 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2682 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2683 addrconf_dad_start(ifp);
2688 static void addrconf_dev_config(struct net_device *dev)
2690 struct in6_addr addr;
2691 struct inet6_dev *idev;
2695 if ((dev->type != ARPHRD_ETHER) &&
2696 (dev->type != ARPHRD_FDDI) &&
2697 (dev->type != ARPHRD_ARCNET) &&
2698 (dev->type != ARPHRD_INFINIBAND) &&
2699 (dev->type != ARPHRD_IEEE802154) &&
2700 (dev->type != ARPHRD_IEEE1394)) {
2701 /* Alas, we support only Ethernet autoconfiguration. */
2705 idev = addrconf_add_dev(dev);
2709 memset(&addr, 0, sizeof(struct in6_addr));
2710 addr.s6_addr32[0] = htonl(0xFE800000);
2712 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2713 addrconf_add_linklocal(idev, &addr);
2716 #if IS_ENABLED(CONFIG_IPV6_SIT)
2717 static void addrconf_sit_config(struct net_device *dev)
2719 struct inet6_dev *idev;
2724 * Configure the tunnel with one of our IPv4
2725 * addresses... we should configure all of
2726 * our v4 addrs in the tunnel
2729 if ((idev = ipv6_find_idev(dev)) == NULL) {
2730 pr_debug("%s: add_dev failed\n", __func__);
2734 if (dev->priv_flags & IFF_ISATAP) {
2735 struct in6_addr addr;
2737 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2738 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2739 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2740 addrconf_add_linklocal(idev, &addr);
2744 sit_add_v4_addrs(idev);
2746 if (dev->flags&IFF_POINTOPOINT)
2747 addrconf_add_mroute(dev);
2753 #if IS_ENABLED(CONFIG_NET_IPGRE)
2754 static void addrconf_gre_config(struct net_device *dev)
2756 struct inet6_dev *idev;
2757 struct in6_addr addr;
2759 pr_info("%s(%s)\n", __func__, dev->name);
2763 if ((idev = ipv6_find_idev(dev)) == NULL) {
2764 pr_debug("%s: add_dev failed\n", __func__);
2768 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2769 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2771 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2772 addrconf_add_linklocal(idev, &addr);
2777 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2779 struct in6_addr lladdr;
2781 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2782 addrconf_add_linklocal(idev, &lladdr);
2788 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2790 struct net_device *link_dev;
2791 struct net *net = dev_net(idev->dev);
2793 /* first try to inherit the link-local address from the link device */
2794 if (idev->dev->iflink &&
2795 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2796 if (!ipv6_inherit_linklocal(idev, link_dev))
2799 /* then try to inherit it from any device */
2800 for_each_netdev(net, link_dev) {
2801 if (!ipv6_inherit_linklocal(idev, link_dev))
2804 pr_debug("init ip6-ip6: add_linklocal failed\n");
2808 * Autoconfigure tunnel with a link-local address so routing protocols,
2809 * DHCPv6, MLD etc. can be run over the virtual link
2812 static void addrconf_ip6_tnl_config(struct net_device *dev)
2814 struct inet6_dev *idev;
2818 idev = addrconf_add_dev(dev);
2820 pr_debug("init ip6-ip6: add_dev failed\n");
2823 ip6_tnl_add_linklocal(idev);
2826 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2829 struct net_device *dev = (struct net_device *) data;
2830 struct inet6_dev *idev = __in6_dev_get(dev);
2831 int run_pending = 0;
2835 case NETDEV_REGISTER:
2836 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2837 idev = ipv6_add_dev(dev);
2839 return notifier_from_errno(-ENOMEM);
2845 if (dev->flags & IFF_SLAVE)
2848 if (event == NETDEV_UP) {
2849 if (!addrconf_qdisc_ok(dev)) {
2850 /* device is not ready yet. */
2851 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2856 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2857 idev = ipv6_add_dev(dev);
2860 idev->if_flags |= IF_READY;
2864 if (!addrconf_qdisc_ok(dev)) {
2865 /* device is still not ready. */
2870 if (idev->if_flags & IF_READY)
2871 /* device is already configured. */
2873 idev->if_flags |= IF_READY;
2876 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2882 switch (dev->type) {
2883 #if IS_ENABLED(CONFIG_IPV6_SIT)
2885 addrconf_sit_config(dev);
2888 #if IS_ENABLED(CONFIG_NET_IPGRE)
2890 addrconf_gre_config(dev);
2893 case ARPHRD_TUNNEL6:
2894 addrconf_ip6_tnl_config(dev);
2896 case ARPHRD_LOOPBACK:
2901 addrconf_dev_config(dev);
2907 addrconf_dad_run(idev);
2910 * If the MTU changed during the interface down,
2911 * when the interface up, the changed MTU must be
2912 * reflected in the idev as well as routers.
2914 if (idev->cnf.mtu6 != dev->mtu &&
2915 dev->mtu >= IPV6_MIN_MTU) {
2916 rt6_mtu_change(dev, dev->mtu);
2917 idev->cnf.mtu6 = dev->mtu;
2919 idev->tstamp = jiffies;
2920 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2923 * If the changed mtu during down is lower than
2924 * IPV6_MIN_MTU stop IPv6 on this interface.
2926 if (dev->mtu < IPV6_MIN_MTU)
2927 addrconf_ifdown(dev, 1);
2931 case NETDEV_CHANGEMTU:
2932 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2933 rt6_mtu_change(dev, dev->mtu);
2934 idev->cnf.mtu6 = dev->mtu;
2938 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2939 idev = ipv6_add_dev(dev);
2945 * MTU falled under IPV6_MIN_MTU.
2946 * Stop IPv6 on this interface.
2950 case NETDEV_UNREGISTER:
2952 * Remove all addresses from this interface.
2954 addrconf_ifdown(dev, event != NETDEV_DOWN);
2957 case NETDEV_CHANGENAME:
2959 snmp6_unregister_dev(idev);
2960 addrconf_sysctl_unregister(idev);
2961 addrconf_sysctl_register(idev);
2962 err = snmp6_register_dev(idev);
2964 return notifier_from_errno(err);
2968 case NETDEV_PRE_TYPE_CHANGE:
2969 case NETDEV_POST_TYPE_CHANGE:
2970 addrconf_type_change(dev, event);
2978 * addrconf module should be notified of a device going up
2980 static struct notifier_block ipv6_dev_notf = {
2981 .notifier_call = addrconf_notify,
2984 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2986 struct inet6_dev *idev;
2989 idev = __in6_dev_get(dev);
2991 if (event == NETDEV_POST_TYPE_CHANGE)
2992 ipv6_mc_remap(idev);
2993 else if (event == NETDEV_PRE_TYPE_CHANGE)
2994 ipv6_mc_unmap(idev);
2997 static int addrconf_ifdown(struct net_device *dev, int how)
2999 struct net *net = dev_net(dev);
3000 struct inet6_dev *idev;
3001 struct inet6_ifaddr *ifa;
3006 rt6_ifdown(net, dev);
3007 neigh_ifdown(&nd_tbl, dev);
3009 idev = __in6_dev_get(dev);
3014 * Step 1: remove reference to ipv6 device from parent device.
3020 /* protected by rtnl_lock */
3021 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3023 /* Step 1.5: remove snmp6 entry */
3024 snmp6_unregister_dev(idev);
3028 /* Step 2: clear hash table */
3029 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3030 struct hlist_head *h = &inet6_addr_lst[i];
3032 spin_lock_bh(&addrconf_hash_lock);
3034 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3035 if (ifa->idev == idev) {
3036 hlist_del_init_rcu(&ifa->addr_lst);
3037 addrconf_del_timer(ifa);
3041 spin_unlock_bh(&addrconf_hash_lock);
3044 write_lock_bh(&idev->lock);
3046 /* Step 2: clear flags for stateless addrconf */
3048 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3050 #ifdef CONFIG_IPV6_PRIVACY
3051 if (how && del_timer(&idev->regen_timer))
3054 /* Step 3: clear tempaddr list */
3055 while (!list_empty(&idev->tempaddr_list)) {
3056 ifa = list_first_entry(&idev->tempaddr_list,
3057 struct inet6_ifaddr, tmp_list);
3058 list_del(&ifa->tmp_list);
3059 write_unlock_bh(&idev->lock);
3060 spin_lock_bh(&ifa->lock);
3063 in6_ifa_put(ifa->ifpub);
3066 spin_unlock_bh(&ifa->lock);
3068 write_lock_bh(&idev->lock);
3072 while (!list_empty(&idev->addr_list)) {
3073 ifa = list_first_entry(&idev->addr_list,
3074 struct inet6_ifaddr, if_list);
3075 addrconf_del_timer(ifa);
3077 list_del(&ifa->if_list);
3079 write_unlock_bh(&idev->lock);
3081 spin_lock_bh(&ifa->state_lock);
3083 ifa->state = INET6_IFADDR_STATE_DEAD;
3084 spin_unlock_bh(&ifa->state_lock);
3086 if (state != INET6_IFADDR_STATE_DEAD) {
3087 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3088 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3092 write_lock_bh(&idev->lock);
3095 write_unlock_bh(&idev->lock);
3097 /* Step 5: Discard multicast list */
3099 ipv6_mc_destroy_dev(idev);
3103 idev->tstamp = jiffies;
3105 /* Last: Shot the device (if unregistered) */
3107 addrconf_sysctl_unregister(idev);
3108 neigh_parms_release(&nd_tbl, idev->nd_parms);
3109 neigh_ifdown(&nd_tbl, dev);
3115 static void addrconf_rs_timer(unsigned long data)
3117 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3118 struct inet6_dev *idev = ifp->idev;
3120 read_lock(&idev->lock);
3121 if (idev->dead || !(idev->if_flags & IF_READY))
3124 if (!ipv6_accept_ra(idev))
3127 /* Announcement received after solicitation was sent */
3128 if (idev->if_flags & IF_RA_RCVD)
3131 spin_lock(&ifp->lock);
3132 if (ifp->probes++ < idev->cnf.rtr_solicits) {
3133 /* The wait after the last probe can be shorter */
3134 addrconf_mod_timer(ifp, AC_RS,
3135 (ifp->probes == idev->cnf.rtr_solicits) ?
3136 idev->cnf.rtr_solicit_delay :
3137 idev->cnf.rtr_solicit_interval);
3138 spin_unlock(&ifp->lock);
3140 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3142 spin_unlock(&ifp->lock);
3144 * Note: we do not support deprecated "all on-link"
3145 * assumption any longer.
3147 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3151 read_unlock(&idev->lock);
3156 * Duplicate Address Detection
3158 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3160 unsigned long rand_num;
3161 struct inet6_dev *idev = ifp->idev;
3163 if (ifp->flags & IFA_F_OPTIMISTIC)
3166 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3168 ifp->probes = idev->cnf.dad_transmits;
3169 addrconf_mod_timer(ifp, AC_DAD, rand_num);
3172 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3174 struct inet6_dev *idev = ifp->idev;
3175 struct net_device *dev = idev->dev;
3177 addrconf_join_solict(dev, &ifp->addr);
3179 net_srandom(ifp->addr.s6_addr32[3]);
3181 read_lock_bh(&idev->lock);
3182 spin_lock(&ifp->lock);
3183 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3186 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3187 idev->cnf.accept_dad < 1 ||
3188 !(ifp->flags&IFA_F_TENTATIVE) ||
3189 ifp->flags & IFA_F_NODAD) {
3190 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3191 spin_unlock(&ifp->lock);
3192 read_unlock_bh(&idev->lock);
3194 addrconf_dad_completed(ifp);
3198 if (!(idev->if_flags & IF_READY)) {
3199 spin_unlock(&ifp->lock);
3200 read_unlock_bh(&idev->lock);
3202 * If the device is not ready:
3203 * - keep it tentative if it is a permanent address.
3204 * - otherwise, kill it.
3207 addrconf_dad_stop(ifp, 0);
3212 * Optimistic nodes can start receiving
3215 if (ifp->flags & IFA_F_OPTIMISTIC)
3216 ip6_ins_rt(ifp->rt);
3218 addrconf_dad_kick(ifp);
3220 spin_unlock(&ifp->lock);
3221 read_unlock_bh(&idev->lock);
3224 static void addrconf_dad_timer(unsigned long data)
3226 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3227 struct inet6_dev *idev = ifp->idev;
3228 struct in6_addr mcaddr;
3230 if (!ifp->probes && addrconf_dad_end(ifp))
3233 read_lock(&idev->lock);
3234 if (idev->dead || !(idev->if_flags & IF_READY)) {
3235 read_unlock(&idev->lock);
3239 spin_lock(&ifp->lock);
3240 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3241 spin_unlock(&ifp->lock);
3242 read_unlock(&idev->lock);
3246 if (ifp->probes == 0) {
3248 * DAD was successful
3251 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3252 spin_unlock(&ifp->lock);
3253 read_unlock(&idev->lock);
3255 addrconf_dad_completed(ifp);
3261 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3262 spin_unlock(&ifp->lock);
3263 read_unlock(&idev->lock);
3265 /* send a neighbour solicitation for our addr */
3266 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3267 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3272 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3274 struct net_device *dev = ifp->idev->dev;
3277 * Configure the address for reception. Now it is valid.
3280 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3282 /* If added prefix is link local and we are prepared to process
3283 router advertisements, start sending router solicitations.
3286 if (ipv6_accept_ra(ifp->idev) &&
3287 ifp->idev->cnf.rtr_solicits > 0 &&
3288 (dev->flags&IFF_LOOPBACK) == 0 &&
3289 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3291 * If a host as already performed a random delay
3292 * [...] as part of DAD [...] there is no need
3293 * to delay again before sending the first RS
3295 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3297 spin_lock_bh(&ifp->lock);
3299 ifp->idev->if_flags |= IF_RS_SENT;
3300 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3301 spin_unlock_bh(&ifp->lock);
3305 static void addrconf_dad_run(struct inet6_dev *idev)
3307 struct inet6_ifaddr *ifp;
3309 read_lock_bh(&idev->lock);
3310 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3311 spin_lock(&ifp->lock);
3312 if (ifp->flags & IFA_F_TENTATIVE &&
3313 ifp->state == INET6_IFADDR_STATE_DAD)
3314 addrconf_dad_kick(ifp);
3315 spin_unlock(&ifp->lock);
3317 read_unlock_bh(&idev->lock);
3320 #ifdef CONFIG_PROC_FS
3321 struct if6_iter_state {
3322 struct seq_net_private p;
3327 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3329 struct inet6_ifaddr *ifa = NULL;
3330 struct if6_iter_state *state = seq->private;
3331 struct net *net = seq_file_net(seq);
3334 /* initial bucket if pos is 0 */
3340 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3341 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3343 if (!net_eq(dev_net(ifa->idev->dev), net))
3345 /* sync with offset */
3346 if (p < state->offset) {
3354 /* prepare for next bucket */
3361 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3362 struct inet6_ifaddr *ifa)
3364 struct if6_iter_state *state = seq->private;
3365 struct net *net = seq_file_net(seq);
3367 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3368 if (!net_eq(dev_net(ifa->idev->dev), net))
3374 while (++state->bucket < IN6_ADDR_HSIZE) {
3376 hlist_for_each_entry_rcu_bh(ifa,
3377 &inet6_addr_lst[state->bucket], addr_lst) {
3378 if (!net_eq(dev_net(ifa->idev->dev), net))
3388 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3392 return if6_get_first(seq, *pos);
3395 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3397 struct inet6_ifaddr *ifa;
3399 ifa = if6_get_next(seq, v);
3404 static void if6_seq_stop(struct seq_file *seq, void *v)
3407 rcu_read_unlock_bh();
3410 static int if6_seq_show(struct seq_file *seq, void *v)
3412 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3413 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3415 ifp->idev->dev->ifindex,
3419 ifp->idev->dev->name);
3423 static const struct seq_operations if6_seq_ops = {
3424 .start = if6_seq_start,
3425 .next = if6_seq_next,
3426 .show = if6_seq_show,
3427 .stop = if6_seq_stop,
3430 static int if6_seq_open(struct inode *inode, struct file *file)
3432 return seq_open_net(inode, file, &if6_seq_ops,
3433 sizeof(struct if6_iter_state));
3436 static const struct file_operations if6_fops = {
3437 .owner = THIS_MODULE,
3438 .open = if6_seq_open,
3440 .llseek = seq_lseek,
3441 .release = seq_release_net,
3444 static int __net_init if6_proc_net_init(struct net *net)
3446 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3451 static void __net_exit if6_proc_net_exit(struct net *net)
3453 remove_proc_entry("if_inet6", net->proc_net);
3456 static struct pernet_operations if6_proc_net_ops = {
3457 .init = if6_proc_net_init,
3458 .exit = if6_proc_net_exit,
3461 int __init if6_proc_init(void)
3463 return register_pernet_subsys(&if6_proc_net_ops);
3466 void if6_proc_exit(void)
3468 unregister_pernet_subsys(&if6_proc_net_ops);
3470 #endif /* CONFIG_PROC_FS */
3472 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3473 /* Check if address is a home address configured on any interface. */
3474 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3477 struct inet6_ifaddr *ifp = NULL;
3478 unsigned int hash = inet6_addr_hash(addr);
3481 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3482 if (!net_eq(dev_net(ifp->idev->dev), net))
3484 if (ipv6_addr_equal(&ifp->addr, addr) &&
3485 (ifp->flags & IFA_F_HOMEADDRESS)) {
3490 rcu_read_unlock_bh();
3496 * Periodic address status verification
3499 static void addrconf_verify(unsigned long foo)
3501 unsigned long now, next, next_sec, next_sched;
3502 struct inet6_ifaddr *ifp;
3506 spin_lock(&addrconf_verify_lock);
3508 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3510 del_timer(&addr_chk_timer);
3512 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3514 hlist_for_each_entry_rcu_bh(ifp,
3515 &inet6_addr_lst[i], addr_lst) {
3518 if (ifp->flags & IFA_F_PERMANENT)
3521 spin_lock(&ifp->lock);
3522 /* We try to batch several events at once. */
3523 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3525 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3526 age >= ifp->valid_lft) {
3527 spin_unlock(&ifp->lock);
3531 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3532 spin_unlock(&ifp->lock);
3534 } else if (age >= ifp->prefered_lft) {
3535 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3538 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3540 ifp->flags |= IFA_F_DEPRECATED;
3543 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3544 next = ifp->tstamp + ifp->valid_lft * HZ;
3546 spin_unlock(&ifp->lock);
3551 ipv6_ifa_notify(0, ifp);
3555 #ifdef CONFIG_IPV6_PRIVACY
3556 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3557 !(ifp->flags&IFA_F_TENTATIVE)) {
3558 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3559 ifp->idev->cnf.dad_transmits *
3560 ifp->idev->nd_parms->retrans_time / HZ;
3562 if (age >= ifp->prefered_lft - regen_advance) {
3563 struct inet6_ifaddr *ifpub = ifp->ifpub;
3564 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3565 next = ifp->tstamp + ifp->prefered_lft * HZ;
3566 if (!ifp->regen_count && ifpub) {
3569 in6_ifa_hold(ifpub);
3570 spin_unlock(&ifp->lock);
3572 spin_lock(&ifpub->lock);
3573 ifpub->regen_count = 0;
3574 spin_unlock(&ifpub->lock);
3575 ipv6_create_tempaddr(ifpub, ifp);
3580 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3581 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3582 spin_unlock(&ifp->lock);
3585 /* ifp->prefered_lft <= ifp->valid_lft */
3586 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3587 next = ifp->tstamp + ifp->prefered_lft * HZ;
3588 spin_unlock(&ifp->lock);
3593 next_sec = round_jiffies_up(next);
3596 /* If rounded timeout is accurate enough, accept it. */
3597 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3598 next_sched = next_sec;
3600 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3601 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3602 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3604 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3605 now, next, next_sec, next_sched));
3607 addr_chk_timer.expires = next_sched;
3608 add_timer(&addr_chk_timer);
3609 spin_unlock(&addrconf_verify_lock);
3610 rcu_read_unlock_bh();
3613 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3615 struct in6_addr *pfx = NULL;
3618 pfx = nla_data(addr);
3621 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3624 pfx = nla_data(local);
3630 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3631 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3632 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3633 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3637 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3639 struct net *net = sock_net(skb->sk);
3640 struct ifaddrmsg *ifm;
3641 struct nlattr *tb[IFA_MAX+1];
3642 struct in6_addr *pfx;
3645 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3649 ifm = nlmsg_data(nlh);
3650 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3654 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3657 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3658 u32 prefered_lft, u32 valid_lft)
3662 unsigned long timeout;
3664 if (!valid_lft || (prefered_lft > valid_lft))
3667 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3668 if (addrconf_finite_timeout(timeout)) {
3669 expires = jiffies_to_clock_t(timeout * HZ);
3670 valid_lft = timeout;
3671 flags = RTF_EXPIRES;
3675 ifa_flags |= IFA_F_PERMANENT;
3678 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3679 if (addrconf_finite_timeout(timeout)) {
3681 ifa_flags |= IFA_F_DEPRECATED;
3682 prefered_lft = timeout;
3685 spin_lock_bh(&ifp->lock);
3686 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3687 ifp->tstamp = jiffies;
3688 ifp->valid_lft = valid_lft;
3689 ifp->prefered_lft = prefered_lft;
3691 spin_unlock_bh(&ifp->lock);
3692 if (!(ifp->flags&IFA_F_TENTATIVE))
3693 ipv6_ifa_notify(0, ifp);
3695 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3703 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3705 struct net *net = sock_net(skb->sk);
3706 struct ifaddrmsg *ifm;
3707 struct nlattr *tb[IFA_MAX+1];
3708 struct in6_addr *pfx;
3709 struct inet6_ifaddr *ifa;
3710 struct net_device *dev;
3711 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3715 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3719 ifm = nlmsg_data(nlh);
3720 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3724 if (tb[IFA_CACHEINFO]) {
3725 struct ifa_cacheinfo *ci;
3727 ci = nla_data(tb[IFA_CACHEINFO]);
3728 valid_lft = ci->ifa_valid;
3729 preferred_lft = ci->ifa_prefered;
3731 preferred_lft = INFINITY_LIFE_TIME;
3732 valid_lft = INFINITY_LIFE_TIME;
3735 dev = __dev_get_by_index(net, ifm->ifa_index);
3739 /* We ignore other flags so far. */
3740 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3742 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3745 * It would be best to check for !NLM_F_CREATE here but
3746 * userspace alreay relies on not having to provide this.
3748 return inet6_addr_add(net, ifm->ifa_index, pfx,
3749 ifm->ifa_prefixlen, ifa_flags,
3750 preferred_lft, valid_lft);
3753 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3754 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3757 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3764 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3765 u8 scope, int ifindex)
3767 struct ifaddrmsg *ifm;
3769 ifm = nlmsg_data(nlh);
3770 ifm->ifa_family = AF_INET6;
3771 ifm->ifa_prefixlen = prefixlen;
3772 ifm->ifa_flags = flags;
3773 ifm->ifa_scope = scope;
3774 ifm->ifa_index = ifindex;
3777 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3778 unsigned long tstamp, u32 preferred, u32 valid)
3780 struct ifa_cacheinfo ci;
3782 ci.cstamp = cstamp_delta(cstamp);
3783 ci.tstamp = cstamp_delta(tstamp);
3784 ci.ifa_prefered = preferred;
3785 ci.ifa_valid = valid;
3787 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3790 static inline int rt_scope(int ifa_scope)
3792 if (ifa_scope & IFA_HOST)
3793 return RT_SCOPE_HOST;
3794 else if (ifa_scope & IFA_LINK)
3795 return RT_SCOPE_LINK;
3796 else if (ifa_scope & IFA_SITE)
3797 return RT_SCOPE_SITE;
3799 return RT_SCOPE_UNIVERSE;
3802 static inline int inet6_ifaddr_msgsize(void)
3804 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3805 + nla_total_size(16) /* IFA_ADDRESS */
3806 + nla_total_size(sizeof(struct ifa_cacheinfo));
3809 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3810 u32 portid, u32 seq, int event, unsigned int flags)
3812 struct nlmsghdr *nlh;
3813 u32 preferred, valid;
3815 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3819 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3820 ifa->idev->dev->ifindex);
3822 if (!(ifa->flags&IFA_F_PERMANENT)) {
3823 preferred = ifa->prefered_lft;
3824 valid = ifa->valid_lft;
3825 if (preferred != INFINITY_LIFE_TIME) {
3826 long tval = (jiffies - ifa->tstamp)/HZ;
3827 if (preferred > tval)
3831 if (valid != INFINITY_LIFE_TIME) {
3839 preferred = INFINITY_LIFE_TIME;
3840 valid = INFINITY_LIFE_TIME;
3843 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3844 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3845 nlmsg_cancel(skb, nlh);
3849 return nlmsg_end(skb, nlh);
3852 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3853 u32 portid, u32 seq, int event, u16 flags)
3855 struct nlmsghdr *nlh;
3856 u8 scope = RT_SCOPE_UNIVERSE;
3857 int ifindex = ifmca->idev->dev->ifindex;
3859 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3860 scope = RT_SCOPE_SITE;
3862 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3866 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3867 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3868 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3869 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3870 nlmsg_cancel(skb, nlh);
3874 return nlmsg_end(skb, nlh);
3877 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3878 u32 portid, u32 seq, int event, unsigned int flags)
3880 struct nlmsghdr *nlh;
3881 u8 scope = RT_SCOPE_UNIVERSE;
3882 int ifindex = ifaca->aca_idev->dev->ifindex;
3884 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3885 scope = RT_SCOPE_SITE;
3887 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3891 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3892 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3893 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3894 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3895 nlmsg_cancel(skb, nlh);
3899 return nlmsg_end(skb, nlh);
3908 /* called with rcu_read_lock() */
3909 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3910 struct netlink_callback *cb, enum addr_type_t type,
3911 int s_ip_idx, int *p_ip_idx)
3913 struct ifmcaddr6 *ifmca;
3914 struct ifacaddr6 *ifaca;
3916 int ip_idx = *p_ip_idx;
3918 read_lock_bh(&idev->lock);
3920 case UNICAST_ADDR: {
3921 struct inet6_ifaddr *ifa;
3923 /* unicast address incl. temp addr */
3924 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3925 if (++ip_idx < s_ip_idx)
3927 err = inet6_fill_ifaddr(skb, ifa,
3928 NETLINK_CB(cb->skb).portid,
3934 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
3938 case MULTICAST_ADDR:
3939 /* multicast address */
3940 for (ifmca = idev->mc_list; ifmca;
3941 ifmca = ifmca->next, ip_idx++) {
3942 if (ip_idx < s_ip_idx)
3944 err = inet6_fill_ifmcaddr(skb, ifmca,
3945 NETLINK_CB(cb->skb).portid,
3954 /* anycast address */
3955 for (ifaca = idev->ac_list; ifaca;
3956 ifaca = ifaca->aca_next, ip_idx++) {
3957 if (ip_idx < s_ip_idx)
3959 err = inet6_fill_ifacaddr(skb, ifaca,
3960 NETLINK_CB(cb->skb).portid,
3971 read_unlock_bh(&idev->lock);
3976 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3977 enum addr_type_t type)
3979 struct net *net = sock_net(skb->sk);
3982 int s_idx, s_ip_idx;
3983 struct net_device *dev;
3984 struct inet6_dev *idev;
3985 struct hlist_head *head;
3988 s_idx = idx = cb->args[1];
3989 s_ip_idx = ip_idx = cb->args[2];
3992 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
3993 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3995 head = &net->dev_index_head[h];
3996 hlist_for_each_entry_rcu(dev, head, index_hlist) {
3999 if (h > s_h || idx > s_idx)
4002 idev = __in6_dev_get(dev);
4006 if (in6_dump_addrs(idev, skb, cb, type,
4007 s_ip_idx, &ip_idx) <= 0)
4017 cb->args[2] = ip_idx;
4022 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4024 enum addr_type_t type = UNICAST_ADDR;
4026 return inet6_dump_addr(skb, cb, type);
4029 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4031 enum addr_type_t type = MULTICAST_ADDR;
4033 return inet6_dump_addr(skb, cb, type);
4037 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4039 enum addr_type_t type = ANYCAST_ADDR;
4041 return inet6_dump_addr(skb, cb, type);
4044 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4046 struct net *net = sock_net(in_skb->sk);
4047 struct ifaddrmsg *ifm;
4048 struct nlattr *tb[IFA_MAX+1];
4049 struct in6_addr *addr = NULL;
4050 struct net_device *dev = NULL;
4051 struct inet6_ifaddr *ifa;
4052 struct sk_buff *skb;
4055 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4059 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
4065 ifm = nlmsg_data(nlh);
4067 dev = __dev_get_by_index(net, ifm->ifa_index);
4069 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4071 err = -EADDRNOTAVAIL;
4075 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4081 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4082 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4084 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4085 WARN_ON(err == -EMSGSIZE);
4089 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4096 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4098 struct sk_buff *skb;
4099 struct net *net = dev_net(ifa->idev->dev);
4102 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4106 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4108 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4109 WARN_ON(err == -EMSGSIZE);
4113 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4117 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4120 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4121 __s32 *array, int bytes)
4123 BUG_ON(bytes < (DEVCONF_MAX * 4));
4125 memset(array, 0, bytes);
4126 array[DEVCONF_FORWARDING] = cnf->forwarding;
4127 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4128 array[DEVCONF_MTU6] = cnf->mtu6;
4129 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4130 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4131 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4132 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4133 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4134 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4135 jiffies_to_msecs(cnf->rtr_solicit_interval);
4136 array[DEVCONF_RTR_SOLICIT_DELAY] =
4137 jiffies_to_msecs(cnf->rtr_solicit_delay);
4138 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4139 #ifdef CONFIG_IPV6_PRIVACY
4140 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4141 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4142 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4143 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4144 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4146 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4147 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4148 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4149 #ifdef CONFIG_IPV6_ROUTER_PREF
4150 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4151 array[DEVCONF_RTR_PROBE_INTERVAL] =
4152 jiffies_to_msecs(cnf->rtr_probe_interval);
4153 #ifdef CONFIG_IPV6_ROUTE_INFO
4154 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4157 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4158 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4159 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4160 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4162 #ifdef CONFIG_IPV6_MROUTE
4163 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4165 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4166 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4167 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4168 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4171 static inline size_t inet6_ifla6_size(void)
4173 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4174 + nla_total_size(sizeof(struct ifla_cacheinfo))
4175 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4176 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4177 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4178 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4181 static inline size_t inet6_if_nlmsg_size(void)
4183 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4184 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4185 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4186 + nla_total_size(4) /* IFLA_MTU */
4187 + nla_total_size(4) /* IFLA_LINK */
4188 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4191 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4192 int items, int bytes)
4195 int pad = bytes - sizeof(u64) * items;
4198 /* Use put_unaligned() because stats may not be aligned for u64. */
4199 put_unaligned(items, &stats[0]);
4200 for (i = 1; i < items; i++)
4201 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4203 memset(&stats[items], 0, pad);
4206 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4207 int items, int bytes, size_t syncpoff)
4210 int pad = bytes - sizeof(u64) * items;
4213 /* Use put_unaligned() because stats may not be aligned for u64. */
4214 put_unaligned(items, &stats[0]);
4215 for (i = 1; i < items; i++)
4216 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4218 memset(&stats[items], 0, pad);
4221 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4225 case IFLA_INET6_STATS:
4226 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4227 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4229 case IFLA_INET6_ICMP6STATS:
4230 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4235 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4238 struct ifla_cacheinfo ci;
4240 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4241 goto nla_put_failure;
4242 ci.max_reasm_len = IPV6_MAXPLEN;
4243 ci.tstamp = cstamp_delta(idev->tstamp);
4244 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4245 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4246 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4247 goto nla_put_failure;
4248 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4250 goto nla_put_failure;
4251 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4253 /* XXX - MC not implemented */
4255 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4257 goto nla_put_failure;
4258 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4260 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4262 goto nla_put_failure;
4263 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4265 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4267 goto nla_put_failure;
4268 read_lock_bh(&idev->lock);
4269 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4270 read_unlock_bh(&idev->lock);
4278 static size_t inet6_get_link_af_size(const struct net_device *dev)
4280 if (!__in6_dev_get(dev))
4283 return inet6_ifla6_size();
4286 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4288 struct inet6_dev *idev = __in6_dev_get(dev);
4293 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4299 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4301 struct inet6_ifaddr *ifp;
4302 struct net_device *dev = idev->dev;
4303 bool update_rs = false;
4307 if (ipv6_addr_any(token))
4309 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4311 if (!ipv6_accept_ra(idev))
4313 if (idev->cnf.rtr_solicits <= 0)
4316 write_lock_bh(&idev->lock);
4318 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4319 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4321 write_unlock_bh(&idev->lock);
4323 if (!idev->dead && (idev->if_flags & IF_READY)) {
4324 struct in6_addr ll_addr;
4326 ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4329 /* If we're not ready, then normal ifup will take care
4330 * of this. Otherwise, we need to request our rs here.
4332 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4336 write_lock_bh(&idev->lock);
4339 idev->if_flags |= IF_RS_SENT;
4341 /* Well, that's kinda nasty ... */
4342 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4343 spin_lock(&ifp->lock);
4344 if (ifp->tokenized) {
4346 ifp->prefered_lft = 0;
4348 spin_unlock(&ifp->lock);
4351 write_unlock_bh(&idev->lock);
4355 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4358 struct inet6_dev *idev = __in6_dev_get(dev);
4359 struct nlattr *tb[IFLA_INET6_MAX + 1];
4362 return -EAFNOSUPPORT;
4364 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4367 if (tb[IFLA_INET6_TOKEN])
4368 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4373 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4374 u32 portid, u32 seq, int event, unsigned int flags)
4376 struct net_device *dev = idev->dev;
4377 struct ifinfomsg *hdr;
4378 struct nlmsghdr *nlh;
4381 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4385 hdr = nlmsg_data(nlh);
4386 hdr->ifi_family = AF_INET6;
4388 hdr->ifi_type = dev->type;
4389 hdr->ifi_index = dev->ifindex;
4390 hdr->ifi_flags = dev_get_flags(dev);
4391 hdr->ifi_change = 0;
4393 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4395 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4396 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4397 (dev->ifindex != dev->iflink &&
4398 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4399 goto nla_put_failure;
4400 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4401 if (protoinfo == NULL)
4402 goto nla_put_failure;
4404 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4405 goto nla_put_failure;
4407 nla_nest_end(skb, protoinfo);
4408 return nlmsg_end(skb, nlh);
4411 nlmsg_cancel(skb, nlh);
4415 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4417 struct net *net = sock_net(skb->sk);
4420 struct net_device *dev;
4421 struct inet6_dev *idev;
4422 struct hlist_head *head;
4425 s_idx = cb->args[1];
4428 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4430 head = &net->dev_index_head[h];
4431 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4434 idev = __in6_dev_get(dev);
4437 if (inet6_fill_ifinfo(skb, idev,
4438 NETLINK_CB(cb->skb).portid,
4440 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4454 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4456 struct sk_buff *skb;
4457 struct net *net = dev_net(idev->dev);
4460 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4464 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4466 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4467 WARN_ON(err == -EMSGSIZE);
4471 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4475 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4478 static inline size_t inet6_prefix_nlmsg_size(void)
4480 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4481 + nla_total_size(sizeof(struct in6_addr))
4482 + nla_total_size(sizeof(struct prefix_cacheinfo));
4485 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4486 struct prefix_info *pinfo, u32 portid, u32 seq,
4487 int event, unsigned int flags)
4489 struct prefixmsg *pmsg;
4490 struct nlmsghdr *nlh;
4491 struct prefix_cacheinfo ci;
4493 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4497 pmsg = nlmsg_data(nlh);
4498 pmsg->prefix_family = AF_INET6;
4499 pmsg->prefix_pad1 = 0;
4500 pmsg->prefix_pad2 = 0;
4501 pmsg->prefix_ifindex = idev->dev->ifindex;
4502 pmsg->prefix_len = pinfo->prefix_len;
4503 pmsg->prefix_type = pinfo->type;
4504 pmsg->prefix_pad3 = 0;
4505 pmsg->prefix_flags = 0;
4507 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4508 if (pinfo->autoconf)
4509 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4511 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4512 goto nla_put_failure;
4513 ci.preferred_time = ntohl(pinfo->prefered);
4514 ci.valid_time = ntohl(pinfo->valid);
4515 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4516 goto nla_put_failure;
4517 return nlmsg_end(skb, nlh);
4520 nlmsg_cancel(skb, nlh);
4524 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4525 struct prefix_info *pinfo)
4527 struct sk_buff *skb;
4528 struct net *net = dev_net(idev->dev);
4531 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4535 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4537 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4538 WARN_ON(err == -EMSGSIZE);
4542 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4546 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4549 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4551 struct net *net = dev_net(ifp->idev->dev);
4553 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4558 * If the address was optimistic
4559 * we inserted the route at the start of
4560 * our DAD process, so we don't need
4563 if (!(ifp->rt->rt6i_node))
4564 ip6_ins_rt(ifp->rt);
4565 if (ifp->idev->cnf.forwarding)
4566 addrconf_join_anycast(ifp);
4569 if (ifp->idev->cnf.forwarding)
4570 addrconf_leave_anycast(ifp);
4571 addrconf_leave_solict(ifp->idev, &ifp->addr);
4572 dst_hold(&ifp->rt->dst);
4574 if (ip6_del_rt(ifp->rt))
4575 dst_free(&ifp->rt->dst);
4578 atomic_inc(&net->ipv6.dev_addr_genid);
4581 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4584 if (likely(ifp->idev->dead == 0))
4585 __ipv6_ifa_notify(event, ifp);
4586 rcu_read_unlock_bh();
4589 #ifdef CONFIG_SYSCTL
4592 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4593 void __user *buffer, size_t *lenp, loff_t *ppos)
4595 int *valp = ctl->data;
4602 * ctl->data points to idev->cnf.forwarding, we should
4603 * not modify it until we get the rtnl lock.
4608 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4611 ret = addrconf_fixup_forwarding(ctl, valp, val);
4617 static void dev_disable_change(struct inet6_dev *idev)
4619 if (!idev || !idev->dev)
4622 if (idev->cnf.disable_ipv6)
4623 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4625 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4628 static void addrconf_disable_change(struct net *net, __s32 newf)
4630 struct net_device *dev;
4631 struct inet6_dev *idev;
4634 for_each_netdev_rcu(net, dev) {
4635 idev = __in6_dev_get(dev);
4637 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4638 idev->cnf.disable_ipv6 = newf;
4640 dev_disable_change(idev);
4646 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4651 if (!rtnl_trylock())
4652 return restart_syscall();
4654 net = (struct net *)table->extra2;
4658 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4663 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4664 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4665 addrconf_disable_change(net, newf);
4666 } else if ((!newf) ^ (!old))
4667 dev_disable_change((struct inet6_dev *)table->extra1);
4674 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4675 void __user *buffer, size_t *lenp, loff_t *ppos)
4677 int *valp = ctl->data;
4684 * ctl->data points to idev->cnf.disable_ipv6, we should
4685 * not modify it until we get the rtnl lock.
4690 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4693 ret = addrconf_disable_ipv6(ctl, valp, val);
4699 static struct addrconf_sysctl_table
4701 struct ctl_table_header *sysctl_header;
4702 ctl_table addrconf_vars[DEVCONF_MAX+1];
4703 } addrconf_sysctl __read_mostly = {
4704 .sysctl_header = NULL,
4707 .procname = "forwarding",
4708 .data = &ipv6_devconf.forwarding,
4709 .maxlen = sizeof(int),
4711 .proc_handler = addrconf_sysctl_forward,
4714 .procname = "hop_limit",
4715 .data = &ipv6_devconf.hop_limit,
4716 .maxlen = sizeof(int),
4718 .proc_handler = proc_dointvec,
4722 .data = &ipv6_devconf.mtu6,
4723 .maxlen = sizeof(int),
4725 .proc_handler = proc_dointvec,
4728 .procname = "accept_ra",
4729 .data = &ipv6_devconf.accept_ra,
4730 .maxlen = sizeof(int),
4732 .proc_handler = proc_dointvec,
4735 .procname = "accept_redirects",
4736 .data = &ipv6_devconf.accept_redirects,
4737 .maxlen = sizeof(int),
4739 .proc_handler = proc_dointvec,
4742 .procname = "autoconf",
4743 .data = &ipv6_devconf.autoconf,
4744 .maxlen = sizeof(int),
4746 .proc_handler = proc_dointvec,
4749 .procname = "dad_transmits",
4750 .data = &ipv6_devconf.dad_transmits,
4751 .maxlen = sizeof(int),
4753 .proc_handler = proc_dointvec,
4756 .procname = "router_solicitations",
4757 .data = &ipv6_devconf.rtr_solicits,
4758 .maxlen = sizeof(int),
4760 .proc_handler = proc_dointvec,
4763 .procname = "router_solicitation_interval",
4764 .data = &ipv6_devconf.rtr_solicit_interval,
4765 .maxlen = sizeof(int),
4767 .proc_handler = proc_dointvec_jiffies,
4770 .procname = "router_solicitation_delay",
4771 .data = &ipv6_devconf.rtr_solicit_delay,
4772 .maxlen = sizeof(int),
4774 .proc_handler = proc_dointvec_jiffies,
4777 .procname = "force_mld_version",
4778 .data = &ipv6_devconf.force_mld_version,
4779 .maxlen = sizeof(int),
4781 .proc_handler = proc_dointvec,
4783 #ifdef CONFIG_IPV6_PRIVACY
4785 .procname = "use_tempaddr",
4786 .data = &ipv6_devconf.use_tempaddr,
4787 .maxlen = sizeof(int),
4789 .proc_handler = proc_dointvec,
4792 .procname = "temp_valid_lft",
4793 .data = &ipv6_devconf.temp_valid_lft,
4794 .maxlen = sizeof(int),
4796 .proc_handler = proc_dointvec,
4799 .procname = "temp_prefered_lft",
4800 .data = &ipv6_devconf.temp_prefered_lft,
4801 .maxlen = sizeof(int),
4803 .proc_handler = proc_dointvec,
4806 .procname = "regen_max_retry",
4807 .data = &ipv6_devconf.regen_max_retry,
4808 .maxlen = sizeof(int),
4810 .proc_handler = proc_dointvec,
4813 .procname = "max_desync_factor",
4814 .data = &ipv6_devconf.max_desync_factor,
4815 .maxlen = sizeof(int),
4817 .proc_handler = proc_dointvec,
4821 .procname = "max_addresses",
4822 .data = &ipv6_devconf.max_addresses,
4823 .maxlen = sizeof(int),
4825 .proc_handler = proc_dointvec,
4828 .procname = "accept_ra_defrtr",
4829 .data = &ipv6_devconf.accept_ra_defrtr,
4830 .maxlen = sizeof(int),
4832 .proc_handler = proc_dointvec,
4835 .procname = "accept_ra_pinfo",
4836 .data = &ipv6_devconf.accept_ra_pinfo,
4837 .maxlen = sizeof(int),
4839 .proc_handler = proc_dointvec,
4841 #ifdef CONFIG_IPV6_ROUTER_PREF
4843 .procname = "accept_ra_rtr_pref",
4844 .data = &ipv6_devconf.accept_ra_rtr_pref,
4845 .maxlen = sizeof(int),
4847 .proc_handler = proc_dointvec,
4850 .procname = "router_probe_interval",
4851 .data = &ipv6_devconf.rtr_probe_interval,
4852 .maxlen = sizeof(int),
4854 .proc_handler = proc_dointvec_jiffies,
4856 #ifdef CONFIG_IPV6_ROUTE_INFO
4858 .procname = "accept_ra_rt_info_max_plen",
4859 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4860 .maxlen = sizeof(int),
4862 .proc_handler = proc_dointvec,
4867 .procname = "proxy_ndp",
4868 .data = &ipv6_devconf.proxy_ndp,
4869 .maxlen = sizeof(int),
4871 .proc_handler = proc_dointvec,
4874 .procname = "accept_source_route",
4875 .data = &ipv6_devconf.accept_source_route,
4876 .maxlen = sizeof(int),
4878 .proc_handler = proc_dointvec,
4880 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4882 .procname = "optimistic_dad",
4883 .data = &ipv6_devconf.optimistic_dad,
4884 .maxlen = sizeof(int),
4886 .proc_handler = proc_dointvec,
4890 #ifdef CONFIG_IPV6_MROUTE
4892 .procname = "mc_forwarding",
4893 .data = &ipv6_devconf.mc_forwarding,
4894 .maxlen = sizeof(int),
4896 .proc_handler = proc_dointvec,
4900 .procname = "disable_ipv6",
4901 .data = &ipv6_devconf.disable_ipv6,
4902 .maxlen = sizeof(int),
4904 .proc_handler = addrconf_sysctl_disable,
4907 .procname = "accept_dad",
4908 .data = &ipv6_devconf.accept_dad,
4909 .maxlen = sizeof(int),
4911 .proc_handler = proc_dointvec,
4914 .procname = "force_tllao",
4915 .data = &ipv6_devconf.force_tllao,
4916 .maxlen = sizeof(int),
4918 .proc_handler = proc_dointvec
4921 .procname = "ndisc_notify",
4922 .data = &ipv6_devconf.ndisc_notify,
4923 .maxlen = sizeof(int),
4925 .proc_handler = proc_dointvec
4933 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4934 struct inet6_dev *idev, struct ipv6_devconf *p)
4937 struct addrconf_sysctl_table *t;
4938 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4940 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4944 for (i = 0; t->addrconf_vars[i].data; i++) {
4945 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4946 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4947 t->addrconf_vars[i].extra2 = net;
4950 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4952 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4953 if (t->sysctl_header == NULL)
4965 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4967 struct addrconf_sysctl_table *t;
4969 if (p->sysctl == NULL)
4974 unregister_net_sysctl_table(t->sysctl_header);
4978 static void addrconf_sysctl_register(struct inet6_dev *idev)
4980 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4981 &ndisc_ifinfo_sysctl_change);
4982 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4986 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4988 __addrconf_sysctl_unregister(&idev->cnf);
4989 neigh_sysctl_unregister(idev->nd_parms);
4995 static int __net_init addrconf_init_net(struct net *net)
4998 struct ipv6_devconf *all, *dflt;
5000 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5004 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5006 goto err_alloc_dflt;
5008 /* these will be inherited by all namespaces */
5009 dflt->autoconf = ipv6_defaults.autoconf;
5010 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5012 net->ipv6.devconf_all = all;
5013 net->ipv6.devconf_dflt = dflt;
5015 #ifdef CONFIG_SYSCTL
5016 err = __addrconf_sysctl_register(net, "all", NULL, all);
5020 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5026 #ifdef CONFIG_SYSCTL
5028 __addrconf_sysctl_unregister(all);
5038 static void __net_exit addrconf_exit_net(struct net *net)
5040 #ifdef CONFIG_SYSCTL
5041 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5042 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5044 if (!net_eq(net, &init_net)) {
5045 kfree(net->ipv6.devconf_dflt);
5046 kfree(net->ipv6.devconf_all);
5050 static struct pernet_operations addrconf_ops = {
5051 .init = addrconf_init_net,
5052 .exit = addrconf_exit_net,
5055 static struct rtnl_af_ops inet6_ops = {
5057 .fill_link_af = inet6_fill_link_af,
5058 .get_link_af_size = inet6_get_link_af_size,
5059 .set_link_af = inet6_set_link_af,
5063 * Init / cleanup code
5066 int __init addrconf_init(void)
5070 err = ipv6_addr_label_init();
5072 pr_crit("%s: cannot initialize default policy table: %d\n",
5077 err = register_pernet_subsys(&addrconf_ops);
5081 /* The addrconf netdev notifier requires that loopback_dev
5082 * has it's ipv6 private information allocated and setup
5083 * before it can bring up and give link-local addresses
5084 * to other devices which are up.
5086 * Unfortunately, loopback_dev is not necessarily the first
5087 * entry in the global dev_base list of net devices. In fact,
5088 * it is likely to be the very last entry on that list.
5089 * So this causes the notifier registry below to try and
5090 * give link-local addresses to all devices besides loopback_dev
5091 * first, then loopback_dev, which cases all the non-loopback_dev
5092 * devices to fail to get a link-local address.
5094 * So, as a temporary fix, allocate the ipv6 structure for
5095 * loopback_dev first by hand.
5096 * Longer term, all of the dependencies ipv6 has upon the loopback
5097 * device and it being up should be removed.
5100 if (!ipv6_add_dev(init_net.loopback_dev))
5106 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5107 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5109 register_netdevice_notifier(&ipv6_dev_notf);
5113 err = rtnl_af_register(&inet6_ops);
5117 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5122 /* Only the first call to __rtnl_register can fail */
5123 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5124 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5125 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5126 inet6_dump_ifaddr, NULL);
5127 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5128 inet6_dump_ifmcaddr, NULL);
5129 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5130 inet6_dump_ifacaddr, NULL);
5131 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5132 inet6_netconf_dump_devconf, NULL);
5134 ipv6_addr_label_rtnl_register();
5138 rtnl_af_unregister(&inet6_ops);
5140 unregister_netdevice_notifier(&ipv6_dev_notf);
5142 unregister_pernet_subsys(&addrconf_ops);
5144 ipv6_addr_label_cleanup();
5149 void addrconf_cleanup(void)
5151 struct net_device *dev;
5154 unregister_netdevice_notifier(&ipv6_dev_notf);
5155 unregister_pernet_subsys(&addrconf_ops);
5156 ipv6_addr_label_cleanup();
5160 __rtnl_af_unregister(&inet6_ops);
5162 /* clean dev list */
5163 for_each_netdev(&init_net, dev) {
5164 if (__in6_dev_get(dev) == NULL)
5166 addrconf_ifdown(dev, 1);
5168 addrconf_ifdown(init_net.loopback_dev, 2);
5173 spin_lock_bh(&addrconf_hash_lock);
5174 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5175 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5176 spin_unlock_bh(&addrconf_hash_lock);
5178 del_timer(&addr_chk_timer);