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>
74 #include <net/protocol.h>
75 #include <net/ndisc.h>
76 #include <net/ip6_route.h>
77 #include <net/addrconf.h>
80 #include <net/netlink.h>
81 #include <net/pkt_sched.h>
82 #include <linux/if_tunnel.h>
83 #include <linux/rtnetlink.h>
84 #include <linux/netconf.h>
86 #ifdef CONFIG_IPV6_PRIVACY
87 #include <linux/random.h>
90 #include <linux/uaccess.h>
91 #include <asm/unaligned.h>
93 #include <linux/proc_fs.h>
94 #include <linux/seq_file.h>
95 #include <linux/export.h>
97 /* Set to 3 to get tracing... */
101 #define ADBG(x) printk x
106 #define INFINITY_LIFE_TIME 0xFFFFFFFF
108 static inline u32 cstamp_delta(unsigned long cstamp)
110 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
114 static void addrconf_sysctl_register(struct inet6_dev *idev);
115 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
117 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
121 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
126 #ifdef CONFIG_IPV6_PRIVACY
127 static void __ipv6_regen_rndid(struct inet6_dev *idev);
128 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
129 static void ipv6_regen_rndid(unsigned long data);
132 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
133 static int ipv6_count_addresses(struct inet6_dev *idev);
136 * Configured unicast address hash table
138 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
139 static DEFINE_SPINLOCK(addrconf_hash_lock);
141 static void addrconf_verify(unsigned long);
143 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
144 static DEFINE_SPINLOCK(addrconf_verify_lock);
146 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
149 static void addrconf_type_change(struct net_device *dev,
150 unsigned long event);
151 static int addrconf_ifdown(struct net_device *dev, int how);
153 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
155 const struct net_device *dev,
156 u32 flags, u32 noflags);
158 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_timer(unsigned long data);
160 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
161 static void addrconf_dad_run(struct inet6_dev *idev);
162 static void addrconf_rs_timer(unsigned long data);
163 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
166 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
167 struct prefix_info *pinfo);
168 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
169 struct net_device *dev);
171 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
173 static struct ipv6_devconf ipv6_devconf __read_mostly = {
175 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
176 .mtu6 = IPV6_MIN_MTU,
178 .accept_redirects = 1,
180 .force_mld_version = 0,
182 .rtr_solicits = MAX_RTR_SOLICITATIONS,
183 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
184 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
185 #ifdef CONFIG_IPV6_PRIVACY
187 .temp_valid_lft = TEMP_VALID_LIFETIME,
188 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
189 .regen_max_retry = REGEN_MAX_RETRY,
190 .max_desync_factor = MAX_DESYNC_FACTOR,
192 .max_addresses = IPV6_MAX_ADDRESSES,
193 .accept_ra_defrtr = 1,
194 .accept_ra_pinfo = 1,
195 #ifdef CONFIG_IPV6_ROUTER_PREF
196 .accept_ra_rtr_pref = 1,
197 .rtr_probe_interval = 60 * HZ,
198 #ifdef CONFIG_IPV6_ROUTE_INFO
199 .accept_ra_rt_info_max_plen = 0,
203 .accept_source_route = 0, /* we do not accept RH0 by default. */
208 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
210 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
211 .mtu6 = IPV6_MIN_MTU,
213 .accept_redirects = 1,
216 .rtr_solicits = MAX_RTR_SOLICITATIONS,
217 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
218 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
219 #ifdef CONFIG_IPV6_PRIVACY
221 .temp_valid_lft = TEMP_VALID_LIFETIME,
222 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
223 .regen_max_retry = REGEN_MAX_RETRY,
224 .max_desync_factor = MAX_DESYNC_FACTOR,
226 .max_addresses = IPV6_MAX_ADDRESSES,
227 .accept_ra_defrtr = 1,
228 .accept_ra_pinfo = 1,
229 #ifdef CONFIG_IPV6_ROUTER_PREF
230 .accept_ra_rtr_pref = 1,
231 .rtr_probe_interval = 60 * HZ,
232 #ifdef CONFIG_IPV6_ROUTE_INFO
233 .accept_ra_rt_info_max_plen = 0,
237 .accept_source_route = 0, /* we do not accept RH0 by default. */
242 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
243 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
244 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
245 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
246 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
247 const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADDR_INTERFACELOCAL_ALLNODES_INIT;
248 const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADDR_INTERFACELOCAL_ALLROUTERS_INIT;
249 const struct in6_addr in6addr_sitelocal_allrouters = IN6ADDR_SITELOCAL_ALLROUTERS_INIT;
251 /* Check if a valid qdisc is available */
252 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
254 return !qdisc_tx_is_noop(dev);
257 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
259 if (del_timer(&ifp->timer))
263 enum addrconf_timer_t {
269 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
270 enum addrconf_timer_t what,
273 if (!del_timer(&ifp->timer))
278 ifp->timer.function = addrconf_dad_timer;
281 ifp->timer.function = addrconf_rs_timer;
286 ifp->timer.expires = jiffies + when;
287 add_timer(&ifp->timer);
290 static int snmp6_alloc_dev(struct inet6_dev *idev)
292 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
293 sizeof(struct ipstats_mib),
294 __alignof__(struct ipstats_mib)) < 0)
296 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
298 if (!idev->stats.icmpv6dev)
300 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
302 if (!idev->stats.icmpv6msgdev)
308 kfree(idev->stats.icmpv6dev);
310 snmp_mib_free((void __percpu **)idev->stats.ipv6);
315 static void snmp6_free_dev(struct inet6_dev *idev)
317 kfree(idev->stats.icmpv6msgdev);
318 kfree(idev->stats.icmpv6dev);
319 snmp_mib_free((void __percpu **)idev->stats.ipv6);
322 /* Nobody refers to this device, we may destroy it. */
324 void in6_dev_finish_destroy(struct inet6_dev *idev)
326 struct net_device *dev = idev->dev;
328 WARN_ON(!list_empty(&idev->addr_list));
329 WARN_ON(idev->mc_list != NULL);
331 #ifdef NET_REFCNT_DEBUG
332 pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
336 pr_warn("Freeing alive inet6 device %p\n", idev);
339 snmp6_free_dev(idev);
340 kfree_rcu(idev, rcu);
342 EXPORT_SYMBOL(in6_dev_finish_destroy);
344 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
346 struct inet6_dev *ndev;
350 if (dev->mtu < IPV6_MIN_MTU)
353 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
358 rwlock_init(&ndev->lock);
360 INIT_LIST_HEAD(&ndev->addr_list);
362 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
363 ndev->cnf.mtu6 = dev->mtu;
364 ndev->cnf.sysctl = NULL;
365 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
366 if (ndev->nd_parms == NULL) {
370 if (ndev->cnf.forwarding)
371 dev_disable_lro(dev);
372 /* We refer to the device */
375 if (snmp6_alloc_dev(ndev) < 0) {
377 "%s: cannot allocate memory for statistics; dev=%s.\n",
378 __func__, dev->name));
379 neigh_parms_release(&nd_tbl, ndev->nd_parms);
385 if (snmp6_register_dev(ndev) < 0) {
387 "%s: cannot create /proc/net/dev_snmp6/%s\n",
388 __func__, dev->name));
389 neigh_parms_release(&nd_tbl, ndev->nd_parms);
391 in6_dev_finish_destroy(ndev);
395 /* One reference from device. We must do this before
396 * we invoke __ipv6_regen_rndid().
400 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
401 ndev->cnf.accept_dad = -1;
403 #if IS_ENABLED(CONFIG_IPV6_SIT)
404 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
405 pr_info("%s: Disabled Multicast RS\n", dev->name);
406 ndev->cnf.rtr_solicits = 0;
410 #ifdef CONFIG_IPV6_PRIVACY
411 INIT_LIST_HEAD(&ndev->tempaddr_list);
412 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 if ((dev->flags&IFF_LOOPBACK) ||
414 dev->type == ARPHRD_TUNNEL ||
415 dev->type == ARPHRD_TUNNEL6 ||
416 dev->type == ARPHRD_SIT ||
417 dev->type == ARPHRD_NONE) {
418 ndev->cnf.use_tempaddr = -1;
421 ipv6_regen_rndid((unsigned long) ndev);
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,
550 struct net *net = sock_net(in_skb->sk);
551 struct nlattr *tb[NETCONFA_MAX+1];
552 struct netconfmsg *ncm;
554 struct ipv6_devconf *devconf;
555 struct inet6_dev *in6_dev;
556 struct net_device *dev;
560 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
561 devconf_ipv6_policy);
566 if (!tb[NETCONFA_IFINDEX])
569 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
571 case NETCONFA_IFINDEX_ALL:
572 devconf = net->ipv6.devconf_all;
574 case NETCONFA_IFINDEX_DEFAULT:
575 devconf = net->ipv6.devconf_dflt;
578 dev = __dev_get_by_index(net, ifindex);
581 in6_dev = __in6_dev_get(dev);
584 devconf = &in6_dev->cnf;
589 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
593 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
594 NETLINK_CB(in_skb).portid,
595 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
598 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
599 WARN_ON(err == -EMSGSIZE);
603 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
609 static void dev_forward_change(struct inet6_dev *idev)
611 struct net_device *dev;
612 struct inet6_ifaddr *ifa;
617 if (idev->cnf.forwarding)
618 dev_disable_lro(dev);
619 if (dev->flags & IFF_MULTICAST) {
620 if (idev->cnf.forwarding) {
621 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
622 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
623 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
625 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
626 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
627 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
631 list_for_each_entry(ifa, &idev->addr_list, if_list) {
632 if (ifa->flags&IFA_F_TENTATIVE)
634 if (idev->cnf.forwarding)
635 addrconf_join_anycast(ifa);
637 addrconf_leave_anycast(ifa);
639 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
640 dev->ifindex, &idev->cnf);
644 static void addrconf_forward_change(struct net *net, __s32 newf)
646 struct net_device *dev;
647 struct inet6_dev *idev;
649 for_each_netdev(net, dev) {
650 idev = __in6_dev_get(dev);
652 int changed = (!idev->cnf.forwarding) ^ (!newf);
653 idev->cnf.forwarding = newf;
655 dev_forward_change(idev);
660 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
666 return restart_syscall();
668 net = (struct net *)table->extra2;
672 if (p == &net->ipv6.devconf_dflt->forwarding) {
673 if ((!newf) ^ (!old))
674 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
675 NETCONFA_IFINDEX_DEFAULT,
676 net->ipv6.devconf_dflt);
681 if (p == &net->ipv6.devconf_all->forwarding) {
682 net->ipv6.devconf_dflt->forwarding = newf;
683 addrconf_forward_change(net, newf);
684 if ((!newf) ^ (!old))
685 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
686 NETCONFA_IFINDEX_ALL,
687 net->ipv6.devconf_all);
688 } else if ((!newf) ^ (!old))
689 dev_forward_change((struct inet6_dev *)table->extra1);
693 rt6_purge_dflt_routers(net);
698 /* Nobody refers to this ifaddr, destroy it */
699 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
701 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
703 #ifdef NET_REFCNT_DEBUG
704 pr_debug("%s\n", __func__);
707 in6_dev_put(ifp->idev);
709 if (del_timer(&ifp->timer))
710 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
712 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
713 pr_warn("Freeing alive inet6 address %p\n", ifp);
722 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
725 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
728 * Each device address list is sorted in order of scope -
729 * global before linklocal.
731 list_for_each(p, &idev->addr_list) {
732 struct inet6_ifaddr *ifa
733 = list_entry(p, struct inet6_ifaddr, if_list);
734 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
738 list_add_tail(&ifp->if_list, p);
741 static u32 inet6_addr_hash(const struct in6_addr *addr)
743 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
746 /* On success it returns ifp with increased reference count */
748 static struct inet6_ifaddr *
749 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
750 int scope, u32 flags)
752 struct inet6_ifaddr *ifa = NULL;
756 int addr_type = ipv6_addr_type(addr);
758 if (addr_type == IPV6_ADDR_ANY ||
759 addr_type & IPV6_ADDR_MULTICAST ||
760 (!(idev->dev->flags & IFF_LOOPBACK) &&
761 addr_type & IPV6_ADDR_LOOPBACK))
762 return ERR_PTR(-EADDRNOTAVAIL);
766 err = -ENODEV; /*XXX*/
770 if (idev->cnf.disable_ipv6) {
775 spin_lock(&addrconf_hash_lock);
777 /* Ignore adding duplicate addresses on an interface */
778 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
779 ADBG(("ipv6_add_addr: already assigned\n"));
784 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
787 ADBG(("ipv6_add_addr: malloc failed\n"));
792 rt = addrconf_dst_alloc(idev, addr, false);
800 spin_lock_init(&ifa->lock);
801 spin_lock_init(&ifa->state_lock);
802 init_timer(&ifa->timer);
803 INIT_HLIST_NODE(&ifa->addr_lst);
804 ifa->timer.data = (unsigned long) ifa;
806 ifa->prefix_len = pfxlen;
807 ifa->flags = flags | IFA_F_TENTATIVE;
808 ifa->cstamp = ifa->tstamp = jiffies;
817 /* Add to big hash table */
818 hash = inet6_addr_hash(addr);
820 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
821 spin_unlock(&addrconf_hash_lock);
823 write_lock(&idev->lock);
824 /* Add to inet6_dev unicast addr list. */
825 ipv6_link_dev_addr(idev, ifa);
827 #ifdef CONFIG_IPV6_PRIVACY
828 if (ifa->flags&IFA_F_TEMPORARY) {
829 list_add(&ifa->tmp_list, &idev->tempaddr_list);
835 write_unlock(&idev->lock);
837 rcu_read_unlock_bh();
839 if (likely(err == 0))
840 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
848 spin_unlock(&addrconf_hash_lock);
852 /* This function wants to get referenced ifp and releases it before return */
854 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
856 struct inet6_ifaddr *ifa, *ifn;
857 struct inet6_dev *idev = ifp->idev;
859 int deleted = 0, onlink = 0;
860 unsigned long expires = jiffies;
862 spin_lock_bh(&ifp->state_lock);
864 ifp->state = INET6_IFADDR_STATE_DEAD;
865 spin_unlock_bh(&ifp->state_lock);
867 if (state == INET6_IFADDR_STATE_DEAD)
870 spin_lock_bh(&addrconf_hash_lock);
871 hlist_del_init_rcu(&ifp->addr_lst);
872 spin_unlock_bh(&addrconf_hash_lock);
874 write_lock_bh(&idev->lock);
875 #ifdef CONFIG_IPV6_PRIVACY
876 if (ifp->flags&IFA_F_TEMPORARY) {
877 list_del(&ifp->tmp_list);
879 in6_ifa_put(ifp->ifpub);
886 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
888 list_del_init(&ifp->if_list);
891 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
895 } else if (ifp->flags & IFA_F_PERMANENT) {
896 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
898 if (ifa->flags & IFA_F_PERMANENT) {
903 unsigned long lifetime;
908 spin_lock(&ifa->lock);
910 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
912 * Note: Because this address is
913 * not permanent, lifetime <
914 * LONG_MAX / HZ here.
916 if (time_before(expires,
917 ifa->tstamp + lifetime * HZ))
918 expires = ifa->tstamp + lifetime * HZ;
919 spin_unlock(&ifa->lock);
924 write_unlock_bh(&idev->lock);
926 addrconf_del_timer(ifp);
928 ipv6_ifa_notify(RTM_DELADDR, ifp);
930 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
933 * Purge or update corresponding prefix
935 * 1) we don't purge prefix here if address was not permanent.
936 * prefix is managed by its own lifetime.
937 * 2) if there're no addresses, delete prefix.
938 * 3) if there're still other permanent address(es),
939 * corresponding prefix is still permanent.
940 * 4) otherwise, update prefix lifetime to the
941 * longest valid lifetime among the corresponding
942 * addresses on the device.
943 * Note: subsequent RA will update lifetime.
947 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
948 struct in6_addr prefix;
951 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
953 rt = addrconf_get_prefix_route(&prefix,
956 0, RTF_GATEWAY | RTF_DEFAULT);
962 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
963 rt6_set_expires(rt, expires);
969 /* clean up prefsrc entries */
970 rt6_remove_prefsrc(ifp);
975 #ifdef CONFIG_IPV6_PRIVACY
976 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
978 struct inet6_dev *idev = ifp->idev;
979 struct in6_addr addr, *tmpaddr;
980 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
981 unsigned long regen_advance;
986 unsigned long now = jiffies;
988 write_lock(&idev->lock);
990 spin_lock_bh(&ift->lock);
991 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
992 spin_unlock_bh(&ift->lock);
999 if (idev->cnf.use_tempaddr <= 0) {
1000 write_unlock(&idev->lock);
1001 pr_info("%s: use_tempaddr is disabled\n", __func__);
1006 spin_lock_bh(&ifp->lock);
1007 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1008 idev->cnf.use_tempaddr = -1; /*XXX*/
1009 spin_unlock_bh(&ifp->lock);
1010 write_unlock(&idev->lock);
1011 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1018 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1019 __ipv6_try_regen_rndid(idev, tmpaddr);
1020 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1021 age = (now - ifp->tstamp) / HZ;
1022 tmp_valid_lft = min_t(__u32,
1024 idev->cnf.temp_valid_lft + age);
1025 tmp_prefered_lft = min_t(__u32,
1027 idev->cnf.temp_prefered_lft + age -
1028 idev->cnf.max_desync_factor);
1029 tmp_plen = ifp->prefix_len;
1030 max_addresses = idev->cnf.max_addresses;
1031 tmp_tstamp = ifp->tstamp;
1032 spin_unlock_bh(&ifp->lock);
1034 regen_advance = idev->cnf.regen_max_retry *
1035 idev->cnf.dad_transmits *
1036 idev->nd_parms->retrans_time / HZ;
1037 write_unlock(&idev->lock);
1039 /* A temporary address is created only if this calculated Preferred
1040 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1041 * an implementation must not create a temporary address with a zero
1042 * Preferred Lifetime.
1044 if (tmp_prefered_lft <= regen_advance) {
1051 addr_flags = IFA_F_TEMPORARY;
1052 /* set in addrconf_prefix_rcv() */
1053 if (ifp->flags & IFA_F_OPTIMISTIC)
1054 addr_flags |= IFA_F_OPTIMISTIC;
1056 ift = !max_addresses ||
1057 ipv6_count_addresses(idev) < max_addresses ?
1058 ipv6_add_addr(idev, &addr, tmp_plen,
1059 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
1061 if (IS_ERR_OR_NULL(ift)) {
1064 pr_info("%s: retry temporary address regeneration\n", __func__);
1066 write_lock(&idev->lock);
1070 spin_lock_bh(&ift->lock);
1072 ift->valid_lft = tmp_valid_lft;
1073 ift->prefered_lft = tmp_prefered_lft;
1075 ift->tstamp = tmp_tstamp;
1076 spin_unlock_bh(&ift->lock);
1078 addrconf_dad_start(ift);
1087 * Choose an appropriate source address (RFC3484)
1090 IPV6_SADDR_RULE_INIT = 0,
1091 IPV6_SADDR_RULE_LOCAL,
1092 IPV6_SADDR_RULE_SCOPE,
1093 IPV6_SADDR_RULE_PREFERRED,
1094 #ifdef CONFIG_IPV6_MIP6
1095 IPV6_SADDR_RULE_HOA,
1097 IPV6_SADDR_RULE_OIF,
1098 IPV6_SADDR_RULE_LABEL,
1099 #ifdef CONFIG_IPV6_PRIVACY
1100 IPV6_SADDR_RULE_PRIVACY,
1102 IPV6_SADDR_RULE_ORCHID,
1103 IPV6_SADDR_RULE_PREFIX,
1107 struct ipv6_saddr_score {
1110 struct inet6_ifaddr *ifa;
1111 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1116 struct ipv6_saddr_dst {
1117 const struct in6_addr *addr;
1124 static inline int ipv6_saddr_preferred(int type)
1126 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1131 static int ipv6_get_saddr_eval(struct net *net,
1132 struct ipv6_saddr_score *score,
1133 struct ipv6_saddr_dst *dst,
1138 if (i <= score->rule) {
1140 case IPV6_SADDR_RULE_SCOPE:
1141 ret = score->scopedist;
1143 case IPV6_SADDR_RULE_PREFIX:
1144 ret = score->matchlen;
1147 ret = !!test_bit(i, score->scorebits);
1153 case IPV6_SADDR_RULE_INIT:
1154 /* Rule 0: remember if hiscore is not ready yet */
1157 case IPV6_SADDR_RULE_LOCAL:
1158 /* Rule 1: Prefer same address */
1159 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1161 case IPV6_SADDR_RULE_SCOPE:
1162 /* Rule 2: Prefer appropriate scope
1167 * ---+--+-+---> scope
1169 * | d is scope of the destination.
1171 * | \ <- smaller scope is better if
1172 * B-15 | \ if scope is enough for destinaion.
1173 * | ret = B - scope (-1 <= scope >= d <= 15).
1175 * |/ <- greater is better
1176 * -C / if scope is not enough for destination.
1177 * /| ret = scope - C (-1 <= d < scope <= 15).
1179 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1180 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1181 * Assume B = 0 and we get C > 29.
1183 ret = __ipv6_addr_src_scope(score->addr_type);
1184 if (ret >= dst->scope)
1187 ret -= 128; /* 30 is enough */
1188 score->scopedist = ret;
1190 case IPV6_SADDR_RULE_PREFERRED:
1191 /* Rule 3: Avoid deprecated and optimistic addresses */
1192 ret = ipv6_saddr_preferred(score->addr_type) ||
1193 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1195 #ifdef CONFIG_IPV6_MIP6
1196 case IPV6_SADDR_RULE_HOA:
1198 /* Rule 4: Prefer home address */
1199 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1200 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1204 case IPV6_SADDR_RULE_OIF:
1205 /* Rule 5: Prefer outgoing interface */
1206 ret = (!dst->ifindex ||
1207 dst->ifindex == score->ifa->idev->dev->ifindex);
1209 case IPV6_SADDR_RULE_LABEL:
1210 /* Rule 6: Prefer matching label */
1211 ret = ipv6_addr_label(net,
1212 &score->ifa->addr, score->addr_type,
1213 score->ifa->idev->dev->ifindex) == dst->label;
1215 #ifdef CONFIG_IPV6_PRIVACY
1216 case IPV6_SADDR_RULE_PRIVACY:
1218 /* Rule 7: Prefer public address
1219 * Note: prefer temporary address if use_tempaddr >= 2
1221 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1222 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1223 score->ifa->idev->cnf.use_tempaddr >= 2;
1224 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1228 case IPV6_SADDR_RULE_ORCHID:
1229 /* Rule 8-: Prefer ORCHID vs ORCHID or
1230 * non-ORCHID vs non-ORCHID
1232 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1233 ipv6_addr_orchid(dst->addr));
1235 case IPV6_SADDR_RULE_PREFIX:
1236 /* Rule 8: Use longest matching prefix */
1237 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1238 if (ret > score->ifa->prefix_len)
1239 ret = score->ifa->prefix_len;
1240 score->matchlen = ret;
1247 __set_bit(i, score->scorebits);
1253 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1254 const struct in6_addr *daddr, unsigned int prefs,
1255 struct in6_addr *saddr)
1257 struct ipv6_saddr_score scores[2],
1258 *score = &scores[0], *hiscore = &scores[1];
1259 struct ipv6_saddr_dst dst;
1260 struct net_device *dev;
1263 dst_type = __ipv6_addr_type(daddr);
1265 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1266 dst.scope = __ipv6_addr_src_scope(dst_type);
1267 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1271 hiscore->ifa = NULL;
1275 for_each_netdev_rcu(net, dev) {
1276 struct inet6_dev *idev;
1278 /* Candidate Source Address (section 4)
1279 * - multicast and link-local destination address,
1280 * the set of candidate source address MUST only
1281 * include addresses assigned to interfaces
1282 * belonging to the same link as the outgoing
1284 * (- For site-local destination addresses, the
1285 * set of candidate source addresses MUST only
1286 * include addresses assigned to interfaces
1287 * belonging to the same site as the outgoing
1290 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1291 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1292 dst.ifindex && dev->ifindex != dst.ifindex)
1295 idev = __in6_dev_get(dev);
1299 read_lock_bh(&idev->lock);
1300 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1304 * - Tentative Address (RFC2462 section 5.4)
1305 * - A tentative address is not considered
1306 * "assigned to an interface" in the traditional
1307 * sense, unless it is also flagged as optimistic.
1308 * - Candidate Source Address (section 4)
1309 * - In any case, anycast addresses, multicast
1310 * addresses, and the unspecified address MUST
1311 * NOT be included in a candidate set.
1313 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1314 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1317 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1319 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1320 score->addr_type & IPV6_ADDR_MULTICAST)) {
1321 LIMIT_NETDEBUG(KERN_DEBUG
1322 "ADDRCONF: unspecified / multicast address "
1323 "assigned as unicast address on %s",
1329 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1331 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1332 int minihiscore, miniscore;
1334 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1335 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1337 if (minihiscore > miniscore) {
1338 if (i == IPV6_SADDR_RULE_SCOPE &&
1339 score->scopedist > 0) {
1342 * each remaining entry
1343 * has too small (not enough)
1344 * scope, because ifa entries
1345 * are sorted by their scope
1351 } else if (minihiscore < miniscore) {
1353 in6_ifa_put(hiscore->ifa);
1355 in6_ifa_hold(score->ifa);
1357 swap(hiscore, score);
1359 /* restore our iterator */
1360 score->ifa = hiscore->ifa;
1367 read_unlock_bh(&idev->lock);
1372 return -EADDRNOTAVAIL;
1374 *saddr = hiscore->ifa->addr;
1375 in6_ifa_put(hiscore->ifa);
1378 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1380 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1381 unsigned char banned_flags)
1383 struct inet6_dev *idev;
1384 int err = -EADDRNOTAVAIL;
1387 idev = __in6_dev_get(dev);
1389 struct inet6_ifaddr *ifp;
1391 read_lock_bh(&idev->lock);
1392 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1393 if (ifp->scope == IFA_LINK &&
1394 !(ifp->flags & banned_flags)) {
1400 read_unlock_bh(&idev->lock);
1406 static int ipv6_count_addresses(struct inet6_dev *idev)
1409 struct inet6_ifaddr *ifp;
1411 read_lock_bh(&idev->lock);
1412 list_for_each_entry(ifp, &idev->addr_list, if_list)
1414 read_unlock_bh(&idev->lock);
1418 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1419 struct net_device *dev, int strict)
1421 struct inet6_ifaddr *ifp;
1422 struct hlist_node *node;
1423 unsigned int hash = inet6_addr_hash(addr);
1426 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1427 if (!net_eq(dev_net(ifp->idev->dev), net))
1429 if (ipv6_addr_equal(&ifp->addr, addr) &&
1430 !(ifp->flags&IFA_F_TENTATIVE) &&
1431 (dev == NULL || ifp->idev->dev == dev ||
1432 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1433 rcu_read_unlock_bh();
1438 rcu_read_unlock_bh();
1441 EXPORT_SYMBOL(ipv6_chk_addr);
1443 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1444 struct net_device *dev)
1446 unsigned int hash = inet6_addr_hash(addr);
1447 struct inet6_ifaddr *ifp;
1448 struct hlist_node *node;
1450 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1451 if (!net_eq(dev_net(ifp->idev->dev), net))
1453 if (ipv6_addr_equal(&ifp->addr, addr)) {
1454 if (dev == NULL || ifp->idev->dev == dev)
1461 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1463 struct inet6_dev *idev;
1464 struct inet6_ifaddr *ifa;
1469 idev = __in6_dev_get(dev);
1471 read_lock_bh(&idev->lock);
1472 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1473 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1478 read_unlock_bh(&idev->lock);
1483 EXPORT_SYMBOL(ipv6_chk_prefix);
1485 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1486 struct net_device *dev, int strict)
1488 struct inet6_ifaddr *ifp, *result = NULL;
1489 unsigned int hash = inet6_addr_hash(addr);
1490 struct hlist_node *node;
1493 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1494 if (!net_eq(dev_net(ifp->idev->dev), net))
1496 if (ipv6_addr_equal(&ifp->addr, addr)) {
1497 if (dev == NULL || ifp->idev->dev == dev ||
1498 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1505 rcu_read_unlock_bh();
1510 /* Gets referenced address, destroys ifaddr */
1512 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1514 if (ifp->flags&IFA_F_PERMANENT) {
1515 spin_lock_bh(&ifp->lock);
1516 addrconf_del_timer(ifp);
1517 ifp->flags |= IFA_F_TENTATIVE;
1519 ifp->flags |= IFA_F_DADFAILED;
1520 spin_unlock_bh(&ifp->lock);
1522 ipv6_ifa_notify(0, ifp);
1524 #ifdef CONFIG_IPV6_PRIVACY
1525 } else if (ifp->flags&IFA_F_TEMPORARY) {
1526 struct inet6_ifaddr *ifpub;
1527 spin_lock_bh(&ifp->lock);
1530 in6_ifa_hold(ifpub);
1531 spin_unlock_bh(&ifp->lock);
1532 ipv6_create_tempaddr(ifpub, ifp);
1535 spin_unlock_bh(&ifp->lock);
1543 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1547 spin_lock(&ifp->state_lock);
1548 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1549 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1552 spin_unlock(&ifp->state_lock);
1557 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1559 struct inet6_dev *idev = ifp->idev;
1561 if (addrconf_dad_end(ifp)) {
1566 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1567 ifp->idev->dev->name, &ifp->addr);
1569 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1570 struct in6_addr addr;
1572 addr.s6_addr32[0] = htonl(0xfe800000);
1573 addr.s6_addr32[1] = 0;
1575 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1576 ipv6_addr_equal(&ifp->addr, &addr)) {
1577 /* DAD failed for link-local based on MAC address */
1578 idev->cnf.disable_ipv6 = 1;
1580 pr_info("%s: IPv6 being disabled!\n",
1581 ifp->idev->dev->name);
1585 addrconf_dad_stop(ifp, 1);
1588 /* Join to solicited addr multicast group. */
1590 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1592 struct in6_addr maddr;
1594 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1597 addrconf_addr_solict_mult(addr, &maddr);
1598 ipv6_dev_mc_inc(dev, &maddr);
1601 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1603 struct in6_addr maddr;
1605 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1608 addrconf_addr_solict_mult(addr, &maddr);
1609 __ipv6_dev_mc_dec(idev, &maddr);
1612 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1614 struct in6_addr addr;
1615 if (ifp->prefix_len == 127) /* RFC 6164 */
1617 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1618 if (ipv6_addr_any(&addr))
1620 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1623 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1625 struct in6_addr addr;
1626 if (ifp->prefix_len == 127) /* RFC 6164 */
1628 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1629 if (ipv6_addr_any(&addr))
1631 __ipv6_dev_ac_dec(ifp->idev, &addr);
1634 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1636 if (dev->addr_len != ETH_ALEN)
1638 memcpy(eui, dev->dev_addr, 3);
1639 memcpy(eui + 5, dev->dev_addr + 3, 3);
1642 * The zSeries OSA network cards can be shared among various
1643 * OS instances, but the OSA cards have only one MAC address.
1644 * This leads to duplicate address conflicts in conjunction
1645 * with IPv6 if more than one instance uses the same card.
1647 * The driver for these cards can deliver a unique 16-bit
1648 * identifier for each instance sharing the same card. It is
1649 * placed instead of 0xFFFE in the interface identifier. The
1650 * "u" bit of the interface identifier is not inverted in this
1651 * case. Hence the resulting interface identifier has local
1652 * scope according to RFC2373.
1655 eui[3] = (dev->dev_id >> 8) & 0xFF;
1656 eui[4] = dev->dev_id & 0xFF;
1665 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1667 if (dev->addr_len != IEEE802154_ADDR_LEN)
1669 memcpy(eui, dev->dev_addr, 8);
1674 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1676 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1677 if (dev->addr_len != ARCNET_ALEN)
1680 eui[7] = *(u8 *)dev->dev_addr;
1684 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1686 if (dev->addr_len != INFINIBAND_ALEN)
1688 memcpy(eui, dev->dev_addr + 12, 8);
1693 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1697 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1698 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1699 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1700 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1701 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1702 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1706 memcpy(eui + 4, &addr, 4);
1710 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1712 if (dev->priv_flags & IFF_ISATAP)
1713 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1717 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1719 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1722 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1724 switch (dev->type) {
1727 return addrconf_ifid_eui48(eui, dev);
1729 return addrconf_ifid_arcnet(eui, dev);
1730 case ARPHRD_INFINIBAND:
1731 return addrconf_ifid_infiniband(eui, dev);
1733 return addrconf_ifid_sit(eui, dev);
1735 return addrconf_ifid_gre(eui, dev);
1736 case ARPHRD_IEEE802154:
1737 return addrconf_ifid_eui64(eui, dev);
1742 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1745 struct inet6_ifaddr *ifp;
1747 read_lock_bh(&idev->lock);
1748 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1749 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1750 memcpy(eui, ifp->addr.s6_addr+8, 8);
1755 read_unlock_bh(&idev->lock);
1759 #ifdef CONFIG_IPV6_PRIVACY
1760 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1761 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1764 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1765 idev->rndid[0] &= ~0x02;
1768 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1769 * check if generated address is not inappropriate
1771 * - Reserved subnet anycast (RFC 2526)
1772 * 11111101 11....11 1xxxxxxx
1773 * - ISATAP (RFC4214) 6.1
1774 * 00-00-5E-FE-xx-xx-xx-xx
1776 * - XXX: already assigned to an address on the device
1778 if (idev->rndid[0] == 0xfd &&
1779 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1780 (idev->rndid[7]&0x80))
1782 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1783 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1785 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1790 static void ipv6_regen_rndid(unsigned long data)
1792 struct inet6_dev *idev = (struct inet6_dev *) data;
1793 unsigned long expires;
1796 write_lock_bh(&idev->lock);
1801 __ipv6_regen_rndid(idev);
1804 idev->cnf.temp_prefered_lft * HZ -
1805 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1806 idev->cnf.max_desync_factor * HZ;
1807 if (time_before(expires, jiffies)) {
1808 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1809 __func__, idev->dev->name);
1813 if (!mod_timer(&idev->regen_timer, expires))
1817 write_unlock_bh(&idev->lock);
1818 rcu_read_unlock_bh();
1822 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1824 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1825 __ipv6_regen_rndid(idev);
1834 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1835 unsigned long expires, u32 flags)
1837 struct fib6_config cfg = {
1838 .fc_table = RT6_TABLE_PREFIX,
1839 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1840 .fc_ifindex = dev->ifindex,
1841 .fc_expires = expires,
1843 .fc_flags = RTF_UP | flags,
1844 .fc_nlinfo.nl_net = dev_net(dev),
1845 .fc_protocol = RTPROT_KERNEL,
1850 /* Prevent useless cloning on PtP SIT.
1851 This thing is done here expecting that the whole
1852 class of non-broadcast devices need not cloning.
1854 #if IS_ENABLED(CONFIG_IPV6_SIT)
1855 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1856 cfg.fc_flags |= RTF_NONEXTHOP;
1859 ip6_route_add(&cfg);
1863 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1865 const struct net_device *dev,
1866 u32 flags, u32 noflags)
1868 struct fib6_node *fn;
1869 struct rt6_info *rt = NULL;
1870 struct fib6_table *table;
1872 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1876 read_lock_bh(&table->tb6_lock);
1877 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1880 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1881 if (rt->dst.dev->ifindex != dev->ifindex)
1883 if ((rt->rt6i_flags & flags) != flags)
1885 if ((rt->rt6i_flags & noflags) != 0)
1891 read_unlock_bh(&table->tb6_lock);
1896 /* Create "default" multicast route to the interface */
1898 static void addrconf_add_mroute(struct net_device *dev)
1900 struct fib6_config cfg = {
1901 .fc_table = RT6_TABLE_LOCAL,
1902 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1903 .fc_ifindex = dev->ifindex,
1906 .fc_nlinfo.nl_net = dev_net(dev),
1909 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1911 ip6_route_add(&cfg);
1914 #if IS_ENABLED(CONFIG_IPV6_SIT)
1915 static void sit_route_add(struct net_device *dev)
1917 struct fib6_config cfg = {
1918 .fc_table = RT6_TABLE_MAIN,
1919 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1920 .fc_ifindex = dev->ifindex,
1922 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1923 .fc_nlinfo.nl_net = dev_net(dev),
1926 /* prefix length - 96 bits "::d.d.d.d" */
1927 ip6_route_add(&cfg);
1931 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1933 struct inet6_dev *idev;
1937 idev = ipv6_find_idev(dev);
1939 return ERR_PTR(-ENOBUFS);
1941 if (idev->cnf.disable_ipv6)
1942 return ERR_PTR(-EACCES);
1944 /* Add default multicast route */
1945 if (!(dev->flags & IFF_LOOPBACK))
1946 addrconf_add_mroute(dev);
1951 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1953 struct prefix_info *pinfo;
1957 struct inet6_dev *in6_dev;
1958 struct net *net = dev_net(dev);
1960 pinfo = (struct prefix_info *) opt;
1962 if (len < sizeof(struct prefix_info)) {
1963 ADBG(("addrconf: prefix option too short\n"));
1968 * Validation checks ([ADDRCONF], page 19)
1971 addr_type = ipv6_addr_type(&pinfo->prefix);
1973 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1976 valid_lft = ntohl(pinfo->valid);
1977 prefered_lft = ntohl(pinfo->prefered);
1979 if (prefered_lft > valid_lft) {
1980 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1984 in6_dev = in6_dev_get(dev);
1986 if (in6_dev == NULL) {
1987 net_dbg_ratelimited("addrconf: device %s not configured\n",
1993 * Two things going on here:
1994 * 1) Add routes for on-link prefixes
1995 * 2) Configure prefixes with the auto flag set
1998 if (pinfo->onlink) {
1999 struct rt6_info *rt;
2000 unsigned long rt_expires;
2002 /* Avoid arithmetic overflow. Really, we could
2003 * save rt_expires in seconds, likely valid_lft,
2004 * but it would require division in fib gc, that it
2008 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2010 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2012 if (addrconf_finite_timeout(rt_expires))
2015 rt = addrconf_get_prefix_route(&pinfo->prefix,
2018 RTF_ADDRCONF | RTF_PREFIX_RT,
2019 RTF_GATEWAY | RTF_DEFAULT);
2022 /* Autoconf prefix route */
2023 if (valid_lft == 0) {
2026 } else if (addrconf_finite_timeout(rt_expires)) {
2028 rt6_set_expires(rt, jiffies + rt_expires);
2030 rt6_clean_expires(rt);
2032 } else if (valid_lft) {
2033 clock_t expires = 0;
2034 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2035 if (addrconf_finite_timeout(rt_expires)) {
2037 flags |= RTF_EXPIRES;
2038 expires = jiffies_to_clock_t(rt_expires);
2040 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2041 dev, expires, flags);
2046 /* Try to figure out our local address for this prefix */
2048 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2049 struct inet6_ifaddr *ifp;
2050 struct in6_addr addr;
2051 int create = 0, update_lft = 0;
2053 if (pinfo->prefix_len == 64) {
2054 memcpy(&addr, &pinfo->prefix, 8);
2055 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2056 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2057 in6_dev_put(in6_dev);
2062 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2064 in6_dev_put(in6_dev);
2069 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2071 if (ifp == NULL && valid_lft) {
2072 int max_addresses = in6_dev->cnf.max_addresses;
2075 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2076 if (in6_dev->cnf.optimistic_dad &&
2077 !net->ipv6.devconf_all->forwarding && sllao)
2078 addr_flags = IFA_F_OPTIMISTIC;
2081 /* Do not allow to create too much of autoconfigured
2082 * addresses; this would be too easy way to crash kernel.
2084 if (!max_addresses ||
2085 ipv6_count_addresses(in6_dev) < max_addresses)
2086 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
2087 addr_type&IPV6_ADDR_SCOPE_MASK,
2090 if (IS_ERR_OR_NULL(ifp)) {
2091 in6_dev_put(in6_dev);
2095 update_lft = create = 1;
2096 ifp->cstamp = jiffies;
2097 addrconf_dad_start(ifp);
2103 #ifdef CONFIG_IPV6_PRIVACY
2104 struct inet6_ifaddr *ift;
2108 /* update lifetime (RFC2462 5.5.3 e) */
2109 spin_lock(&ifp->lock);
2111 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2112 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2115 if (!update_lft && stored_lft) {
2116 if (valid_lft > MIN_VALID_LIFETIME ||
2117 valid_lft > stored_lft)
2119 else if (stored_lft <= MIN_VALID_LIFETIME) {
2120 /* valid_lft <= stored_lft is always true */
2122 * RFC 4862 Section 5.5.3e:
2123 * "Note that the preferred lifetime of
2124 * the corresponding address is always
2125 * reset to the Preferred Lifetime in
2126 * the received Prefix Information
2127 * option, regardless of whether the
2128 * valid lifetime is also reset or
2131 * So if the preferred lifetime in
2132 * this advertisement is different
2133 * than what we have stored, but the
2134 * valid lifetime is invalid, just
2135 * reset prefered_lft.
2137 * We must set the valid lifetime
2138 * to the stored lifetime since we'll
2139 * be updating the timestamp below,
2140 * else we'll set it back to the
2143 if (prefered_lft != ifp->prefered_lft) {
2144 valid_lft = stored_lft;
2148 valid_lft = MIN_VALID_LIFETIME;
2149 if (valid_lft < prefered_lft)
2150 prefered_lft = valid_lft;
2156 ifp->valid_lft = valid_lft;
2157 ifp->prefered_lft = prefered_lft;
2160 ifp->flags &= ~IFA_F_DEPRECATED;
2161 spin_unlock(&ifp->lock);
2163 if (!(flags&IFA_F_TENTATIVE))
2164 ipv6_ifa_notify(0, ifp);
2166 spin_unlock(&ifp->lock);
2168 #ifdef CONFIG_IPV6_PRIVACY
2169 read_lock_bh(&in6_dev->lock);
2170 /* update all temporary addresses in the list */
2171 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2173 int age, max_valid, max_prefered;
2175 if (ifp != ift->ifpub)
2179 * RFC 4941 section 3.3:
2180 * If a received option will extend the lifetime
2181 * of a public address, the lifetimes of
2182 * temporary addresses should be extended,
2183 * subject to the overall constraint that no
2184 * temporary addresses should ever remain
2185 * "valid" or "preferred" for a time longer than
2186 * (TEMP_VALID_LIFETIME) or
2187 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2190 age = (now - ift->cstamp) / HZ;
2191 max_valid = in6_dev->cnf.temp_valid_lft - age;
2195 max_prefered = in6_dev->cnf.temp_prefered_lft -
2196 in6_dev->cnf.max_desync_factor -
2198 if (max_prefered < 0)
2201 if (valid_lft > max_valid)
2202 valid_lft = max_valid;
2204 if (prefered_lft > max_prefered)
2205 prefered_lft = max_prefered;
2207 spin_lock(&ift->lock);
2209 ift->valid_lft = valid_lft;
2210 ift->prefered_lft = prefered_lft;
2212 if (prefered_lft > 0)
2213 ift->flags &= ~IFA_F_DEPRECATED;
2215 spin_unlock(&ift->lock);
2216 if (!(flags&IFA_F_TENTATIVE))
2217 ipv6_ifa_notify(0, ift);
2220 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2222 * When a new public address is created as
2223 * described in [ADDRCONF], also create a new
2224 * temporary address. Also create a temporary
2225 * address if it's enabled but no temporary
2226 * address currently exists.
2228 read_unlock_bh(&in6_dev->lock);
2229 ipv6_create_tempaddr(ifp, NULL);
2231 read_unlock_bh(&in6_dev->lock);
2238 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2239 in6_dev_put(in6_dev);
2243 * Set destination address.
2244 * Special case for SIT interfaces where we create a new "virtual"
2247 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2249 struct in6_ifreq ireq;
2250 struct net_device *dev;
2256 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2259 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2265 #if IS_ENABLED(CONFIG_IPV6_SIT)
2266 if (dev->type == ARPHRD_SIT) {
2267 const struct net_device_ops *ops = dev->netdev_ops;
2269 struct ip_tunnel_parm p;
2271 err = -EADDRNOTAVAIL;
2272 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2275 memset(&p, 0, sizeof(p));
2276 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2280 p.iph.protocol = IPPROTO_IPV6;
2282 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2284 if (ops->ndo_do_ioctl) {
2285 mm_segment_t oldfs = get_fs();
2288 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2295 dev = __dev_get_by_name(net, p.name);
2298 err = dev_open(dev);
2309 * Manual configuration of address on an interface
2311 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2312 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2315 struct inet6_ifaddr *ifp;
2316 struct inet6_dev *idev;
2317 struct net_device *dev;
2321 unsigned long timeout;
2328 /* check the lifetime */
2329 if (!valid_lft || prefered_lft > valid_lft)
2332 dev = __dev_get_by_index(net, ifindex);
2336 idev = addrconf_add_dev(dev);
2338 return PTR_ERR(idev);
2340 scope = ipv6_addr_scope(pfx);
2342 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2343 if (addrconf_finite_timeout(timeout)) {
2344 expires = jiffies_to_clock_t(timeout * HZ);
2345 valid_lft = timeout;
2346 flags = RTF_EXPIRES;
2350 ifa_flags |= IFA_F_PERMANENT;
2353 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2354 if (addrconf_finite_timeout(timeout)) {
2356 ifa_flags |= IFA_F_DEPRECATED;
2357 prefered_lft = timeout;
2360 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2363 spin_lock_bh(&ifp->lock);
2364 ifp->valid_lft = valid_lft;
2365 ifp->prefered_lft = prefered_lft;
2366 ifp->tstamp = jiffies;
2367 spin_unlock_bh(&ifp->lock);
2369 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2372 * Note that section 3.1 of RFC 4429 indicates
2373 * that the Optimistic flag should not be set for
2374 * manually configured addresses
2376 addrconf_dad_start(ifp);
2382 return PTR_ERR(ifp);
2385 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2388 struct inet6_ifaddr *ifp;
2389 struct inet6_dev *idev;
2390 struct net_device *dev;
2395 dev = __dev_get_by_index(net, ifindex);
2399 if ((idev = __in6_dev_get(dev)) == NULL)
2402 read_lock_bh(&idev->lock);
2403 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2404 if (ifp->prefix_len == plen &&
2405 ipv6_addr_equal(pfx, &ifp->addr)) {
2407 read_unlock_bh(&idev->lock);
2411 /* If the last address is deleted administratively,
2412 disable IPv6 on this interface.
2414 if (list_empty(&idev->addr_list))
2415 addrconf_ifdown(idev->dev, 1);
2419 read_unlock_bh(&idev->lock);
2420 return -EADDRNOTAVAIL;
2424 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2426 struct in6_ifreq ireq;
2429 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2432 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2436 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2437 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2438 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2443 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2445 struct in6_ifreq ireq;
2448 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2451 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2455 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2456 ireq.ifr6_prefixlen);
2461 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2462 int plen, int scope)
2464 struct inet6_ifaddr *ifp;
2466 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2468 spin_lock_bh(&ifp->lock);
2469 ifp->flags &= ~IFA_F_TENTATIVE;
2470 spin_unlock_bh(&ifp->lock);
2471 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2476 #if IS_ENABLED(CONFIG_IPV6_SIT)
2477 static void sit_add_v4_addrs(struct inet6_dev *idev)
2479 struct in6_addr addr;
2480 struct net_device *dev;
2481 struct net *net = dev_net(idev->dev);
2486 memset(&addr, 0, sizeof(struct in6_addr));
2487 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2489 if (idev->dev->flags&IFF_POINTOPOINT) {
2490 addr.s6_addr32[0] = htonl(0xfe800000);
2493 scope = IPV6_ADDR_COMPATv4;
2496 if (addr.s6_addr32[3]) {
2497 add_addr(idev, &addr, 128, scope);
2501 for_each_netdev(net, dev) {
2502 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2503 if (in_dev && (dev->flags & IFF_UP)) {
2504 struct in_ifaddr *ifa;
2508 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2511 addr.s6_addr32[3] = ifa->ifa_local;
2513 if (ifa->ifa_scope == RT_SCOPE_LINK)
2515 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2516 if (idev->dev->flags&IFF_POINTOPOINT)
2520 if (idev->dev->flags&IFF_POINTOPOINT)
2525 add_addr(idev, &addr, plen, flag);
2532 static void init_loopback(struct net_device *dev)
2534 struct inet6_dev *idev;
2540 if ((idev = ipv6_find_idev(dev)) == NULL) {
2541 pr_debug("%s: add_dev failed\n", __func__);
2545 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2548 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2550 struct inet6_ifaddr *ifp;
2551 u32 addr_flags = IFA_F_PERMANENT;
2553 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2554 if (idev->cnf.optimistic_dad &&
2555 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2556 addr_flags |= IFA_F_OPTIMISTIC;
2560 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2562 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2563 addrconf_dad_start(ifp);
2568 static void addrconf_dev_config(struct net_device *dev)
2570 struct in6_addr addr;
2571 struct inet6_dev *idev;
2575 if ((dev->type != ARPHRD_ETHER) &&
2576 (dev->type != ARPHRD_FDDI) &&
2577 (dev->type != ARPHRD_ARCNET) &&
2578 (dev->type != ARPHRD_INFINIBAND) &&
2579 (dev->type != ARPHRD_IEEE802154)) {
2580 /* Alas, we support only Ethernet autoconfiguration. */
2584 idev = addrconf_add_dev(dev);
2588 memset(&addr, 0, sizeof(struct in6_addr));
2589 addr.s6_addr32[0] = htonl(0xFE800000);
2591 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2592 addrconf_add_linklocal(idev, &addr);
2595 #if IS_ENABLED(CONFIG_IPV6_SIT)
2596 static void addrconf_sit_config(struct net_device *dev)
2598 struct inet6_dev *idev;
2603 * Configure the tunnel with one of our IPv4
2604 * addresses... we should configure all of
2605 * our v4 addrs in the tunnel
2608 if ((idev = ipv6_find_idev(dev)) == NULL) {
2609 pr_debug("%s: add_dev failed\n", __func__);
2613 if (dev->priv_flags & IFF_ISATAP) {
2614 struct in6_addr addr;
2616 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2617 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2618 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2619 addrconf_add_linklocal(idev, &addr);
2623 sit_add_v4_addrs(idev);
2625 if (dev->flags&IFF_POINTOPOINT)
2626 addrconf_add_mroute(dev);
2632 #if IS_ENABLED(CONFIG_NET_IPGRE)
2633 static void addrconf_gre_config(struct net_device *dev)
2635 struct inet6_dev *idev;
2636 struct in6_addr addr;
2638 pr_info("%s(%s)\n", __func__, dev->name);
2642 if ((idev = ipv6_find_idev(dev)) == NULL) {
2643 pr_debug("%s: add_dev failed\n", __func__);
2647 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2648 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2650 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2651 addrconf_add_linklocal(idev, &addr);
2656 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2658 struct in6_addr lladdr;
2660 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2661 addrconf_add_linklocal(idev, &lladdr);
2667 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2669 struct net_device *link_dev;
2670 struct net *net = dev_net(idev->dev);
2672 /* first try to inherit the link-local address from the link device */
2673 if (idev->dev->iflink &&
2674 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2675 if (!ipv6_inherit_linklocal(idev, link_dev))
2678 /* then try to inherit it from any device */
2679 for_each_netdev(net, link_dev) {
2680 if (!ipv6_inherit_linklocal(idev, link_dev))
2683 pr_debug("init ip6-ip6: add_linklocal failed\n");
2687 * Autoconfigure tunnel with a link-local address so routing protocols,
2688 * DHCPv6, MLD etc. can be run over the virtual link
2691 static void addrconf_ip6_tnl_config(struct net_device *dev)
2693 struct inet6_dev *idev;
2697 idev = addrconf_add_dev(dev);
2699 pr_debug("init ip6-ip6: add_dev failed\n");
2702 ip6_tnl_add_linklocal(idev);
2705 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2708 struct net_device *dev = (struct net_device *) data;
2709 struct inet6_dev *idev = __in6_dev_get(dev);
2710 int run_pending = 0;
2714 case NETDEV_REGISTER:
2715 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2716 idev = ipv6_add_dev(dev);
2718 return notifier_from_errno(-ENOMEM);
2724 if (dev->flags & IFF_SLAVE)
2727 if (event == NETDEV_UP) {
2728 if (!addrconf_qdisc_ok(dev)) {
2729 /* device is not ready yet. */
2730 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2735 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2736 idev = ipv6_add_dev(dev);
2739 idev->if_flags |= IF_READY;
2743 if (!addrconf_qdisc_ok(dev)) {
2744 /* device is still not ready. */
2749 if (idev->if_flags & IF_READY)
2750 /* device is already configured. */
2752 idev->if_flags |= IF_READY;
2755 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2761 switch (dev->type) {
2762 #if IS_ENABLED(CONFIG_IPV6_SIT)
2764 addrconf_sit_config(dev);
2767 #if IS_ENABLED(CONFIG_NET_IPGRE)
2769 addrconf_gre_config(dev);
2772 case ARPHRD_TUNNEL6:
2773 addrconf_ip6_tnl_config(dev);
2775 case ARPHRD_LOOPBACK:
2780 addrconf_dev_config(dev);
2786 addrconf_dad_run(idev);
2789 * If the MTU changed during the interface down,
2790 * when the interface up, the changed MTU must be
2791 * reflected in the idev as well as routers.
2793 if (idev->cnf.mtu6 != dev->mtu &&
2794 dev->mtu >= IPV6_MIN_MTU) {
2795 rt6_mtu_change(dev, dev->mtu);
2796 idev->cnf.mtu6 = dev->mtu;
2798 idev->tstamp = jiffies;
2799 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2802 * If the changed mtu during down is lower than
2803 * IPV6_MIN_MTU stop IPv6 on this interface.
2805 if (dev->mtu < IPV6_MIN_MTU)
2806 addrconf_ifdown(dev, 1);
2810 case NETDEV_CHANGEMTU:
2811 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2812 rt6_mtu_change(dev, dev->mtu);
2813 idev->cnf.mtu6 = dev->mtu;
2817 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2818 idev = ipv6_add_dev(dev);
2824 * MTU falled under IPV6_MIN_MTU.
2825 * Stop IPv6 on this interface.
2829 case NETDEV_UNREGISTER:
2831 * Remove all addresses from this interface.
2833 addrconf_ifdown(dev, event != NETDEV_DOWN);
2836 case NETDEV_CHANGENAME:
2838 snmp6_unregister_dev(idev);
2839 addrconf_sysctl_unregister(idev);
2840 addrconf_sysctl_register(idev);
2841 err = snmp6_register_dev(idev);
2843 return notifier_from_errno(err);
2847 case NETDEV_PRE_TYPE_CHANGE:
2848 case NETDEV_POST_TYPE_CHANGE:
2849 addrconf_type_change(dev, event);
2857 * addrconf module should be notified of a device going up
2859 static struct notifier_block ipv6_dev_notf = {
2860 .notifier_call = addrconf_notify,
2863 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2865 struct inet6_dev *idev;
2868 idev = __in6_dev_get(dev);
2870 if (event == NETDEV_POST_TYPE_CHANGE)
2871 ipv6_mc_remap(idev);
2872 else if (event == NETDEV_PRE_TYPE_CHANGE)
2873 ipv6_mc_unmap(idev);
2876 static int addrconf_ifdown(struct net_device *dev, int how)
2878 struct net *net = dev_net(dev);
2879 struct inet6_dev *idev;
2880 struct inet6_ifaddr *ifa;
2885 rt6_ifdown(net, dev);
2886 neigh_ifdown(&nd_tbl, dev);
2888 idev = __in6_dev_get(dev);
2893 * Step 1: remove reference to ipv6 device from parent device.
2899 /* protected by rtnl_lock */
2900 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2902 /* Step 1.5: remove snmp6 entry */
2903 snmp6_unregister_dev(idev);
2907 /* Step 2: clear hash table */
2908 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2909 struct hlist_head *h = &inet6_addr_lst[i];
2910 struct hlist_node *n;
2912 spin_lock_bh(&addrconf_hash_lock);
2914 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2915 if (ifa->idev == idev) {
2916 hlist_del_init_rcu(&ifa->addr_lst);
2917 addrconf_del_timer(ifa);
2921 spin_unlock_bh(&addrconf_hash_lock);
2924 write_lock_bh(&idev->lock);
2926 /* Step 2: clear flags for stateless addrconf */
2928 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2930 #ifdef CONFIG_IPV6_PRIVACY
2931 if (how && del_timer(&idev->regen_timer))
2934 /* Step 3: clear tempaddr list */
2935 while (!list_empty(&idev->tempaddr_list)) {
2936 ifa = list_first_entry(&idev->tempaddr_list,
2937 struct inet6_ifaddr, tmp_list);
2938 list_del(&ifa->tmp_list);
2939 write_unlock_bh(&idev->lock);
2940 spin_lock_bh(&ifa->lock);
2943 in6_ifa_put(ifa->ifpub);
2946 spin_unlock_bh(&ifa->lock);
2948 write_lock_bh(&idev->lock);
2952 while (!list_empty(&idev->addr_list)) {
2953 ifa = list_first_entry(&idev->addr_list,
2954 struct inet6_ifaddr, if_list);
2955 addrconf_del_timer(ifa);
2957 list_del(&ifa->if_list);
2959 write_unlock_bh(&idev->lock);
2961 spin_lock_bh(&ifa->state_lock);
2963 ifa->state = INET6_IFADDR_STATE_DEAD;
2964 spin_unlock_bh(&ifa->state_lock);
2966 if (state != INET6_IFADDR_STATE_DEAD) {
2967 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2968 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2972 write_lock_bh(&idev->lock);
2975 write_unlock_bh(&idev->lock);
2977 /* Step 5: Discard multicast list */
2979 ipv6_mc_destroy_dev(idev);
2983 idev->tstamp = jiffies;
2985 /* Last: Shot the device (if unregistered) */
2987 addrconf_sysctl_unregister(idev);
2988 neigh_parms_release(&nd_tbl, idev->nd_parms);
2989 neigh_ifdown(&nd_tbl, dev);
2995 static void addrconf_rs_timer(unsigned long data)
2997 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2998 struct inet6_dev *idev = ifp->idev;
3000 read_lock(&idev->lock);
3001 if (idev->dead || !(idev->if_flags & IF_READY))
3004 if (!ipv6_accept_ra(idev))
3007 /* Announcement received after solicitation was sent */
3008 if (idev->if_flags & IF_RA_RCVD)
3011 spin_lock(&ifp->lock);
3012 if (ifp->probes++ < idev->cnf.rtr_solicits) {
3013 /* The wait after the last probe can be shorter */
3014 addrconf_mod_timer(ifp, AC_RS,
3015 (ifp->probes == idev->cnf.rtr_solicits) ?
3016 idev->cnf.rtr_solicit_delay :
3017 idev->cnf.rtr_solicit_interval);
3018 spin_unlock(&ifp->lock);
3020 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3022 spin_unlock(&ifp->lock);
3024 * Note: we do not support deprecated "all on-link"
3025 * assumption any longer.
3027 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3031 read_unlock(&idev->lock);
3036 * Duplicate Address Detection
3038 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3040 unsigned long rand_num;
3041 struct inet6_dev *idev = ifp->idev;
3043 if (ifp->flags & IFA_F_OPTIMISTIC)
3046 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
3048 ifp->probes = idev->cnf.dad_transmits;
3049 addrconf_mod_timer(ifp, AC_DAD, rand_num);
3052 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3054 struct inet6_dev *idev = ifp->idev;
3055 struct net_device *dev = idev->dev;
3057 addrconf_join_solict(dev, &ifp->addr);
3059 net_srandom(ifp->addr.s6_addr32[3]);
3061 read_lock_bh(&idev->lock);
3062 spin_lock(&ifp->lock);
3063 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3066 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3067 idev->cnf.accept_dad < 1 ||
3068 !(ifp->flags&IFA_F_TENTATIVE) ||
3069 ifp->flags & IFA_F_NODAD) {
3070 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3071 spin_unlock(&ifp->lock);
3072 read_unlock_bh(&idev->lock);
3074 addrconf_dad_completed(ifp);
3078 if (!(idev->if_flags & IF_READY)) {
3079 spin_unlock(&ifp->lock);
3080 read_unlock_bh(&idev->lock);
3082 * If the device is not ready:
3083 * - keep it tentative if it is a permanent address.
3084 * - otherwise, kill it.
3087 addrconf_dad_stop(ifp, 0);
3092 * Optimistic nodes can start receiving
3095 if (ifp->flags & IFA_F_OPTIMISTIC)
3096 ip6_ins_rt(ifp->rt);
3098 addrconf_dad_kick(ifp);
3100 spin_unlock(&ifp->lock);
3101 read_unlock_bh(&idev->lock);
3104 static void addrconf_dad_timer(unsigned long data)
3106 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3107 struct inet6_dev *idev = ifp->idev;
3108 struct in6_addr mcaddr;
3110 if (!ifp->probes && addrconf_dad_end(ifp))
3113 read_lock(&idev->lock);
3114 if (idev->dead || !(idev->if_flags & IF_READY)) {
3115 read_unlock(&idev->lock);
3119 spin_lock(&ifp->lock);
3120 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3121 spin_unlock(&ifp->lock);
3122 read_unlock(&idev->lock);
3126 if (ifp->probes == 0) {
3128 * DAD was successful
3131 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3132 spin_unlock(&ifp->lock);
3133 read_unlock(&idev->lock);
3135 addrconf_dad_completed(ifp);
3141 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3142 spin_unlock(&ifp->lock);
3143 read_unlock(&idev->lock);
3145 /* send a neighbour solicitation for our addr */
3146 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3147 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3152 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3154 struct net_device *dev = ifp->idev->dev;
3157 * Configure the address for reception. Now it is valid.
3160 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3162 /* If added prefix is link local and we are prepared to process
3163 router advertisements, start sending router solicitations.
3166 if (ipv6_accept_ra(ifp->idev) &&
3167 ifp->idev->cnf.rtr_solicits > 0 &&
3168 (dev->flags&IFF_LOOPBACK) == 0 &&
3169 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3171 * If a host as already performed a random delay
3172 * [...] as part of DAD [...] there is no need
3173 * to delay again before sending the first RS
3175 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3177 spin_lock_bh(&ifp->lock);
3179 ifp->idev->if_flags |= IF_RS_SENT;
3180 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3181 spin_unlock_bh(&ifp->lock);
3185 static void addrconf_dad_run(struct inet6_dev *idev)
3187 struct inet6_ifaddr *ifp;
3189 read_lock_bh(&idev->lock);
3190 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3191 spin_lock(&ifp->lock);
3192 if (ifp->flags & IFA_F_TENTATIVE &&
3193 ifp->state == INET6_IFADDR_STATE_DAD)
3194 addrconf_dad_kick(ifp);
3195 spin_unlock(&ifp->lock);
3197 read_unlock_bh(&idev->lock);
3200 #ifdef CONFIG_PROC_FS
3201 struct if6_iter_state {
3202 struct seq_net_private p;
3207 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3209 struct inet6_ifaddr *ifa = NULL;
3210 struct if6_iter_state *state = seq->private;
3211 struct net *net = seq_file_net(seq);
3214 /* initial bucket if pos is 0 */
3220 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3221 struct hlist_node *n;
3222 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3224 if (!net_eq(dev_net(ifa->idev->dev), net))
3226 /* sync with offset */
3227 if (p < state->offset) {
3235 /* prepare for next bucket */
3242 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3243 struct inet6_ifaddr *ifa)
3245 struct if6_iter_state *state = seq->private;
3246 struct net *net = seq_file_net(seq);
3247 struct hlist_node *n = &ifa->addr_lst;
3249 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3250 if (!net_eq(dev_net(ifa->idev->dev), net))
3256 while (++state->bucket < IN6_ADDR_HSIZE) {
3258 hlist_for_each_entry_rcu_bh(ifa, n,
3259 &inet6_addr_lst[state->bucket], addr_lst) {
3260 if (!net_eq(dev_net(ifa->idev->dev), net))
3270 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3274 return if6_get_first(seq, *pos);
3277 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3279 struct inet6_ifaddr *ifa;
3281 ifa = if6_get_next(seq, v);
3286 static void if6_seq_stop(struct seq_file *seq, void *v)
3289 rcu_read_unlock_bh();
3292 static int if6_seq_show(struct seq_file *seq, void *v)
3294 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3295 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3297 ifp->idev->dev->ifindex,
3301 ifp->idev->dev->name);
3305 static const struct seq_operations if6_seq_ops = {
3306 .start = if6_seq_start,
3307 .next = if6_seq_next,
3308 .show = if6_seq_show,
3309 .stop = if6_seq_stop,
3312 static int if6_seq_open(struct inode *inode, struct file *file)
3314 return seq_open_net(inode, file, &if6_seq_ops,
3315 sizeof(struct if6_iter_state));
3318 static const struct file_operations if6_fops = {
3319 .owner = THIS_MODULE,
3320 .open = if6_seq_open,
3322 .llseek = seq_lseek,
3323 .release = seq_release_net,
3326 static int __net_init if6_proc_net_init(struct net *net)
3328 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3333 static void __net_exit if6_proc_net_exit(struct net *net)
3335 proc_net_remove(net, "if_inet6");
3338 static struct pernet_operations if6_proc_net_ops = {
3339 .init = if6_proc_net_init,
3340 .exit = if6_proc_net_exit,
3343 int __init if6_proc_init(void)
3345 return register_pernet_subsys(&if6_proc_net_ops);
3348 void if6_proc_exit(void)
3350 unregister_pernet_subsys(&if6_proc_net_ops);
3352 #endif /* CONFIG_PROC_FS */
3354 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3355 /* Check if address is a home address configured on any interface. */
3356 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3359 struct inet6_ifaddr *ifp = NULL;
3360 struct hlist_node *n;
3361 unsigned int hash = inet6_addr_hash(addr);
3364 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3365 if (!net_eq(dev_net(ifp->idev->dev), net))
3367 if (ipv6_addr_equal(&ifp->addr, addr) &&
3368 (ifp->flags & IFA_F_HOMEADDRESS)) {
3373 rcu_read_unlock_bh();
3379 * Periodic address status verification
3382 static void addrconf_verify(unsigned long foo)
3384 unsigned long now, next, next_sec, next_sched;
3385 struct inet6_ifaddr *ifp;
3386 struct hlist_node *node;
3390 spin_lock(&addrconf_verify_lock);
3392 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3394 del_timer(&addr_chk_timer);
3396 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3398 hlist_for_each_entry_rcu_bh(ifp, node,
3399 &inet6_addr_lst[i], addr_lst) {
3402 if (ifp->flags & IFA_F_PERMANENT)
3405 spin_lock(&ifp->lock);
3406 /* We try to batch several events at once. */
3407 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3409 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3410 age >= ifp->valid_lft) {
3411 spin_unlock(&ifp->lock);
3415 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3416 spin_unlock(&ifp->lock);
3418 } else if (age >= ifp->prefered_lft) {
3419 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3422 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3424 ifp->flags |= IFA_F_DEPRECATED;
3427 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3428 next = ifp->tstamp + ifp->valid_lft * HZ;
3430 spin_unlock(&ifp->lock);
3435 ipv6_ifa_notify(0, ifp);
3439 #ifdef CONFIG_IPV6_PRIVACY
3440 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3441 !(ifp->flags&IFA_F_TENTATIVE)) {
3442 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3443 ifp->idev->cnf.dad_transmits *
3444 ifp->idev->nd_parms->retrans_time / HZ;
3446 if (age >= ifp->prefered_lft - regen_advance) {
3447 struct inet6_ifaddr *ifpub = ifp->ifpub;
3448 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3449 next = ifp->tstamp + ifp->prefered_lft * HZ;
3450 if (!ifp->regen_count && ifpub) {
3453 in6_ifa_hold(ifpub);
3454 spin_unlock(&ifp->lock);
3456 spin_lock(&ifpub->lock);
3457 ifpub->regen_count = 0;
3458 spin_unlock(&ifpub->lock);
3459 ipv6_create_tempaddr(ifpub, ifp);
3464 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3465 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3466 spin_unlock(&ifp->lock);
3469 /* ifp->prefered_lft <= ifp->valid_lft */
3470 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3471 next = ifp->tstamp + ifp->prefered_lft * HZ;
3472 spin_unlock(&ifp->lock);
3477 next_sec = round_jiffies_up(next);
3480 /* If rounded timeout is accurate enough, accept it. */
3481 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3482 next_sched = next_sec;
3484 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3485 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3486 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3488 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3489 now, next, next_sec, next_sched));
3491 addr_chk_timer.expires = next_sched;
3492 add_timer(&addr_chk_timer);
3493 spin_unlock(&addrconf_verify_lock);
3494 rcu_read_unlock_bh();
3497 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3499 struct in6_addr *pfx = NULL;
3502 pfx = nla_data(addr);
3505 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3508 pfx = nla_data(local);
3514 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3515 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3516 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3517 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3521 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3523 struct net *net = sock_net(skb->sk);
3524 struct ifaddrmsg *ifm;
3525 struct nlattr *tb[IFA_MAX+1];
3526 struct in6_addr *pfx;
3529 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3533 ifm = nlmsg_data(nlh);
3534 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3538 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3541 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3542 u32 prefered_lft, u32 valid_lft)
3546 unsigned long timeout;
3548 if (!valid_lft || (prefered_lft > valid_lft))
3551 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3552 if (addrconf_finite_timeout(timeout)) {
3553 expires = jiffies_to_clock_t(timeout * HZ);
3554 valid_lft = timeout;
3555 flags = RTF_EXPIRES;
3559 ifa_flags |= IFA_F_PERMANENT;
3562 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3563 if (addrconf_finite_timeout(timeout)) {
3565 ifa_flags |= IFA_F_DEPRECATED;
3566 prefered_lft = timeout;
3569 spin_lock_bh(&ifp->lock);
3570 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3571 ifp->tstamp = jiffies;
3572 ifp->valid_lft = valid_lft;
3573 ifp->prefered_lft = prefered_lft;
3575 spin_unlock_bh(&ifp->lock);
3576 if (!(ifp->flags&IFA_F_TENTATIVE))
3577 ipv6_ifa_notify(0, ifp);
3579 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3587 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3589 struct net *net = sock_net(skb->sk);
3590 struct ifaddrmsg *ifm;
3591 struct nlattr *tb[IFA_MAX+1];
3592 struct in6_addr *pfx;
3593 struct inet6_ifaddr *ifa;
3594 struct net_device *dev;
3595 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3599 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3603 ifm = nlmsg_data(nlh);
3604 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3608 if (tb[IFA_CACHEINFO]) {
3609 struct ifa_cacheinfo *ci;
3611 ci = nla_data(tb[IFA_CACHEINFO]);
3612 valid_lft = ci->ifa_valid;
3613 preferred_lft = ci->ifa_prefered;
3615 preferred_lft = INFINITY_LIFE_TIME;
3616 valid_lft = INFINITY_LIFE_TIME;
3619 dev = __dev_get_by_index(net, ifm->ifa_index);
3623 /* We ignore other flags so far. */
3624 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3626 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3629 * It would be best to check for !NLM_F_CREATE here but
3630 * userspace alreay relies on not having to provide this.
3632 return inet6_addr_add(net, ifm->ifa_index, pfx,
3633 ifm->ifa_prefixlen, ifa_flags,
3634 preferred_lft, valid_lft);
3637 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3638 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3641 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3648 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3649 u8 scope, int ifindex)
3651 struct ifaddrmsg *ifm;
3653 ifm = nlmsg_data(nlh);
3654 ifm->ifa_family = AF_INET6;
3655 ifm->ifa_prefixlen = prefixlen;
3656 ifm->ifa_flags = flags;
3657 ifm->ifa_scope = scope;
3658 ifm->ifa_index = ifindex;
3661 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3662 unsigned long tstamp, u32 preferred, u32 valid)
3664 struct ifa_cacheinfo ci;
3666 ci.cstamp = cstamp_delta(cstamp);
3667 ci.tstamp = cstamp_delta(tstamp);
3668 ci.ifa_prefered = preferred;
3669 ci.ifa_valid = valid;
3671 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3674 static inline int rt_scope(int ifa_scope)
3676 if (ifa_scope & IFA_HOST)
3677 return RT_SCOPE_HOST;
3678 else if (ifa_scope & IFA_LINK)
3679 return RT_SCOPE_LINK;
3680 else if (ifa_scope & IFA_SITE)
3681 return RT_SCOPE_SITE;
3683 return RT_SCOPE_UNIVERSE;
3686 static inline int inet6_ifaddr_msgsize(void)
3688 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3689 + nla_total_size(16) /* IFA_ADDRESS */
3690 + nla_total_size(sizeof(struct ifa_cacheinfo));
3693 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3694 u32 portid, u32 seq, int event, unsigned int flags)
3696 struct nlmsghdr *nlh;
3697 u32 preferred, valid;
3699 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3703 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3704 ifa->idev->dev->ifindex);
3706 if (!(ifa->flags&IFA_F_PERMANENT)) {
3707 preferred = ifa->prefered_lft;
3708 valid = ifa->valid_lft;
3709 if (preferred != INFINITY_LIFE_TIME) {
3710 long tval = (jiffies - ifa->tstamp)/HZ;
3711 if (preferred > tval)
3715 if (valid != INFINITY_LIFE_TIME) {
3723 preferred = INFINITY_LIFE_TIME;
3724 valid = INFINITY_LIFE_TIME;
3727 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3728 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3729 nlmsg_cancel(skb, nlh);
3733 return nlmsg_end(skb, nlh);
3736 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3737 u32 portid, u32 seq, int event, u16 flags)
3739 struct nlmsghdr *nlh;
3740 u8 scope = RT_SCOPE_UNIVERSE;
3741 int ifindex = ifmca->idev->dev->ifindex;
3743 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3744 scope = RT_SCOPE_SITE;
3746 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3750 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3751 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3752 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3753 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3754 nlmsg_cancel(skb, nlh);
3758 return nlmsg_end(skb, nlh);
3761 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3762 u32 portid, u32 seq, int event, unsigned int flags)
3764 struct nlmsghdr *nlh;
3765 u8 scope = RT_SCOPE_UNIVERSE;
3766 int ifindex = ifaca->aca_idev->dev->ifindex;
3768 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3769 scope = RT_SCOPE_SITE;
3771 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3775 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3776 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3777 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3778 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3779 nlmsg_cancel(skb, nlh);
3783 return nlmsg_end(skb, nlh);
3792 /* called with rcu_read_lock() */
3793 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3794 struct netlink_callback *cb, enum addr_type_t type,
3795 int s_ip_idx, int *p_ip_idx)
3797 struct ifmcaddr6 *ifmca;
3798 struct ifacaddr6 *ifaca;
3800 int ip_idx = *p_ip_idx;
3802 read_lock_bh(&idev->lock);
3804 case UNICAST_ADDR: {
3805 struct inet6_ifaddr *ifa;
3807 /* unicast address incl. temp addr */
3808 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3809 if (++ip_idx < s_ip_idx)
3811 err = inet6_fill_ifaddr(skb, ifa,
3812 NETLINK_CB(cb->skb).portid,
3821 case MULTICAST_ADDR:
3822 /* multicast address */
3823 for (ifmca = idev->mc_list; ifmca;
3824 ifmca = ifmca->next, ip_idx++) {
3825 if (ip_idx < s_ip_idx)
3827 err = inet6_fill_ifmcaddr(skb, ifmca,
3828 NETLINK_CB(cb->skb).portid,
3837 /* anycast address */
3838 for (ifaca = idev->ac_list; ifaca;
3839 ifaca = ifaca->aca_next, ip_idx++) {
3840 if (ip_idx < s_ip_idx)
3842 err = inet6_fill_ifacaddr(skb, ifaca,
3843 NETLINK_CB(cb->skb).portid,
3854 read_unlock_bh(&idev->lock);
3859 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3860 enum addr_type_t type)
3862 struct net *net = sock_net(skb->sk);
3865 int s_idx, s_ip_idx;
3866 struct net_device *dev;
3867 struct inet6_dev *idev;
3868 struct hlist_head *head;
3869 struct hlist_node *node;
3872 s_idx = idx = cb->args[1];
3873 s_ip_idx = ip_idx = cb->args[2];
3876 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3878 head = &net->dev_index_head[h];
3879 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3882 if (h > s_h || idx > s_idx)
3885 idev = __in6_dev_get(dev);
3889 if (in6_dump_addrs(idev, skb, cb, type,
3890 s_ip_idx, &ip_idx) <= 0)
3900 cb->args[2] = ip_idx;
3905 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3907 enum addr_type_t type = UNICAST_ADDR;
3909 return inet6_dump_addr(skb, cb, type);
3912 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3914 enum addr_type_t type = MULTICAST_ADDR;
3916 return inet6_dump_addr(skb, cb, type);
3920 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3922 enum addr_type_t type = ANYCAST_ADDR;
3924 return inet6_dump_addr(skb, cb, type);
3927 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3930 struct net *net = sock_net(in_skb->sk);
3931 struct ifaddrmsg *ifm;
3932 struct nlattr *tb[IFA_MAX+1];
3933 struct in6_addr *addr = NULL;
3934 struct net_device *dev = NULL;
3935 struct inet6_ifaddr *ifa;
3936 struct sk_buff *skb;
3939 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3943 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3949 ifm = nlmsg_data(nlh);
3951 dev = __dev_get_by_index(net, ifm->ifa_index);
3953 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3955 err = -EADDRNOTAVAIL;
3959 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3965 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
3966 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3968 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3969 WARN_ON(err == -EMSGSIZE);
3973 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3980 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3982 struct sk_buff *skb;
3983 struct net *net = dev_net(ifa->idev->dev);
3986 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3990 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3992 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3993 WARN_ON(err == -EMSGSIZE);
3997 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4001 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4004 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4005 __s32 *array, int bytes)
4007 BUG_ON(bytes < (DEVCONF_MAX * 4));
4009 memset(array, 0, bytes);
4010 array[DEVCONF_FORWARDING] = cnf->forwarding;
4011 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4012 array[DEVCONF_MTU6] = cnf->mtu6;
4013 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4014 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4015 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4016 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4017 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4018 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4019 jiffies_to_msecs(cnf->rtr_solicit_interval);
4020 array[DEVCONF_RTR_SOLICIT_DELAY] =
4021 jiffies_to_msecs(cnf->rtr_solicit_delay);
4022 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4023 #ifdef CONFIG_IPV6_PRIVACY
4024 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4025 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4026 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4027 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4028 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4030 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4031 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4032 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4033 #ifdef CONFIG_IPV6_ROUTER_PREF
4034 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4035 array[DEVCONF_RTR_PROBE_INTERVAL] =
4036 jiffies_to_msecs(cnf->rtr_probe_interval);
4037 #ifdef CONFIG_IPV6_ROUTE_INFO
4038 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4041 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4042 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4043 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4044 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4046 #ifdef CONFIG_IPV6_MROUTE
4047 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4049 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4050 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4051 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4052 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4055 static inline size_t inet6_ifla6_size(void)
4057 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4058 + nla_total_size(sizeof(struct ifla_cacheinfo))
4059 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4060 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4061 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
4064 static inline size_t inet6_if_nlmsg_size(void)
4066 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4067 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4068 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4069 + nla_total_size(4) /* IFLA_MTU */
4070 + nla_total_size(4) /* IFLA_LINK */
4071 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4074 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4075 int items, int bytes)
4078 int pad = bytes - sizeof(u64) * items;
4081 /* Use put_unaligned() because stats may not be aligned for u64. */
4082 put_unaligned(items, &stats[0]);
4083 for (i = 1; i < items; i++)
4084 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4086 memset(&stats[items], 0, pad);
4089 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4090 int items, int bytes, size_t syncpoff)
4093 int pad = bytes - sizeof(u64) * items;
4096 /* Use put_unaligned() because stats may not be aligned for u64. */
4097 put_unaligned(items, &stats[0]);
4098 for (i = 1; i < items; i++)
4099 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4101 memset(&stats[items], 0, pad);
4104 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4108 case IFLA_INET6_STATS:
4109 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4110 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4112 case IFLA_INET6_ICMP6STATS:
4113 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4118 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4121 struct ifla_cacheinfo ci;
4123 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4124 goto nla_put_failure;
4125 ci.max_reasm_len = IPV6_MAXPLEN;
4126 ci.tstamp = cstamp_delta(idev->tstamp);
4127 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4128 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4129 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4130 goto nla_put_failure;
4131 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4133 goto nla_put_failure;
4134 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4136 /* XXX - MC not implemented */
4138 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4140 goto nla_put_failure;
4141 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4143 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4145 goto nla_put_failure;
4146 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4154 static size_t inet6_get_link_af_size(const struct net_device *dev)
4156 if (!__in6_dev_get(dev))
4159 return inet6_ifla6_size();
4162 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4164 struct inet6_dev *idev = __in6_dev_get(dev);
4169 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4175 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4176 u32 portid, u32 seq, int event, unsigned int flags)
4178 struct net_device *dev = idev->dev;
4179 struct ifinfomsg *hdr;
4180 struct nlmsghdr *nlh;
4183 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4187 hdr = nlmsg_data(nlh);
4188 hdr->ifi_family = AF_INET6;
4190 hdr->ifi_type = dev->type;
4191 hdr->ifi_index = dev->ifindex;
4192 hdr->ifi_flags = dev_get_flags(dev);
4193 hdr->ifi_change = 0;
4195 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4197 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4198 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4199 (dev->ifindex != dev->iflink &&
4200 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4201 goto nla_put_failure;
4202 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4203 if (protoinfo == NULL)
4204 goto nla_put_failure;
4206 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4207 goto nla_put_failure;
4209 nla_nest_end(skb, protoinfo);
4210 return nlmsg_end(skb, nlh);
4213 nlmsg_cancel(skb, nlh);
4217 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4219 struct net *net = sock_net(skb->sk);
4222 struct net_device *dev;
4223 struct inet6_dev *idev;
4224 struct hlist_head *head;
4225 struct hlist_node *node;
4228 s_idx = cb->args[1];
4231 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4233 head = &net->dev_index_head[h];
4234 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4237 idev = __in6_dev_get(dev);
4240 if (inet6_fill_ifinfo(skb, idev,
4241 NETLINK_CB(cb->skb).portid,
4243 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4257 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4259 struct sk_buff *skb;
4260 struct net *net = dev_net(idev->dev);
4263 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4267 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4269 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4270 WARN_ON(err == -EMSGSIZE);
4274 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4278 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4281 static inline size_t inet6_prefix_nlmsg_size(void)
4283 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4284 + nla_total_size(sizeof(struct in6_addr))
4285 + nla_total_size(sizeof(struct prefix_cacheinfo));
4288 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4289 struct prefix_info *pinfo, u32 portid, u32 seq,
4290 int event, unsigned int flags)
4292 struct prefixmsg *pmsg;
4293 struct nlmsghdr *nlh;
4294 struct prefix_cacheinfo ci;
4296 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4300 pmsg = nlmsg_data(nlh);
4301 pmsg->prefix_family = AF_INET6;
4302 pmsg->prefix_pad1 = 0;
4303 pmsg->prefix_pad2 = 0;
4304 pmsg->prefix_ifindex = idev->dev->ifindex;
4305 pmsg->prefix_len = pinfo->prefix_len;
4306 pmsg->prefix_type = pinfo->type;
4307 pmsg->prefix_pad3 = 0;
4308 pmsg->prefix_flags = 0;
4310 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4311 if (pinfo->autoconf)
4312 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4314 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4315 goto nla_put_failure;
4316 ci.preferred_time = ntohl(pinfo->prefered);
4317 ci.valid_time = ntohl(pinfo->valid);
4318 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4319 goto nla_put_failure;
4320 return nlmsg_end(skb, nlh);
4323 nlmsg_cancel(skb, nlh);
4327 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4328 struct prefix_info *pinfo)
4330 struct sk_buff *skb;
4331 struct net *net = dev_net(idev->dev);
4334 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4338 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4340 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4341 WARN_ON(err == -EMSGSIZE);
4345 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4349 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4352 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4354 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4359 * If the address was optimistic
4360 * we inserted the route at the start of
4361 * our DAD process, so we don't need
4364 if (!(ifp->rt->rt6i_node))
4365 ip6_ins_rt(ifp->rt);
4366 if (ifp->idev->cnf.forwarding)
4367 addrconf_join_anycast(ifp);
4370 if (ifp->idev->cnf.forwarding)
4371 addrconf_leave_anycast(ifp);
4372 addrconf_leave_solict(ifp->idev, &ifp->addr);
4373 dst_hold(&ifp->rt->dst);
4375 if (ip6_del_rt(ifp->rt))
4376 dst_free(&ifp->rt->dst);
4381 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4384 if (likely(ifp->idev->dead == 0))
4385 __ipv6_ifa_notify(event, ifp);
4386 rcu_read_unlock_bh();
4389 #ifdef CONFIG_SYSCTL
4392 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4393 void __user *buffer, size_t *lenp, loff_t *ppos)
4395 int *valp = ctl->data;
4402 * ctl->data points to idev->cnf.forwarding, we should
4403 * not modify it until we get the rtnl lock.
4408 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4411 ret = addrconf_fixup_forwarding(ctl, valp, val);
4417 static void dev_disable_change(struct inet6_dev *idev)
4419 if (!idev || !idev->dev)
4422 if (idev->cnf.disable_ipv6)
4423 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4425 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4428 static void addrconf_disable_change(struct net *net, __s32 newf)
4430 struct net_device *dev;
4431 struct inet6_dev *idev;
4434 for_each_netdev_rcu(net, dev) {
4435 idev = __in6_dev_get(dev);
4437 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4438 idev->cnf.disable_ipv6 = newf;
4440 dev_disable_change(idev);
4446 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4451 if (!rtnl_trylock())
4452 return restart_syscall();
4454 net = (struct net *)table->extra2;
4458 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4463 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4464 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4465 addrconf_disable_change(net, newf);
4466 } else if ((!newf) ^ (!old))
4467 dev_disable_change((struct inet6_dev *)table->extra1);
4474 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4475 void __user *buffer, size_t *lenp, loff_t *ppos)
4477 int *valp = ctl->data;
4484 * ctl->data points to idev->cnf.disable_ipv6, we should
4485 * not modify it until we get the rtnl lock.
4490 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4493 ret = addrconf_disable_ipv6(ctl, valp, val);
4499 static struct addrconf_sysctl_table
4501 struct ctl_table_header *sysctl_header;
4502 ctl_table addrconf_vars[DEVCONF_MAX+1];
4503 } addrconf_sysctl __read_mostly = {
4504 .sysctl_header = NULL,
4507 .procname = "forwarding",
4508 .data = &ipv6_devconf.forwarding,
4509 .maxlen = sizeof(int),
4511 .proc_handler = addrconf_sysctl_forward,
4514 .procname = "hop_limit",
4515 .data = &ipv6_devconf.hop_limit,
4516 .maxlen = sizeof(int),
4518 .proc_handler = proc_dointvec,
4522 .data = &ipv6_devconf.mtu6,
4523 .maxlen = sizeof(int),
4525 .proc_handler = proc_dointvec,
4528 .procname = "accept_ra",
4529 .data = &ipv6_devconf.accept_ra,
4530 .maxlen = sizeof(int),
4532 .proc_handler = proc_dointvec,
4535 .procname = "accept_redirects",
4536 .data = &ipv6_devconf.accept_redirects,
4537 .maxlen = sizeof(int),
4539 .proc_handler = proc_dointvec,
4542 .procname = "autoconf",
4543 .data = &ipv6_devconf.autoconf,
4544 .maxlen = sizeof(int),
4546 .proc_handler = proc_dointvec,
4549 .procname = "dad_transmits",
4550 .data = &ipv6_devconf.dad_transmits,
4551 .maxlen = sizeof(int),
4553 .proc_handler = proc_dointvec,
4556 .procname = "router_solicitations",
4557 .data = &ipv6_devconf.rtr_solicits,
4558 .maxlen = sizeof(int),
4560 .proc_handler = proc_dointvec,
4563 .procname = "router_solicitation_interval",
4564 .data = &ipv6_devconf.rtr_solicit_interval,
4565 .maxlen = sizeof(int),
4567 .proc_handler = proc_dointvec_jiffies,
4570 .procname = "router_solicitation_delay",
4571 .data = &ipv6_devconf.rtr_solicit_delay,
4572 .maxlen = sizeof(int),
4574 .proc_handler = proc_dointvec_jiffies,
4577 .procname = "force_mld_version",
4578 .data = &ipv6_devconf.force_mld_version,
4579 .maxlen = sizeof(int),
4581 .proc_handler = proc_dointvec,
4583 #ifdef CONFIG_IPV6_PRIVACY
4585 .procname = "use_tempaddr",
4586 .data = &ipv6_devconf.use_tempaddr,
4587 .maxlen = sizeof(int),
4589 .proc_handler = proc_dointvec,
4592 .procname = "temp_valid_lft",
4593 .data = &ipv6_devconf.temp_valid_lft,
4594 .maxlen = sizeof(int),
4596 .proc_handler = proc_dointvec,
4599 .procname = "temp_prefered_lft",
4600 .data = &ipv6_devconf.temp_prefered_lft,
4601 .maxlen = sizeof(int),
4603 .proc_handler = proc_dointvec,
4606 .procname = "regen_max_retry",
4607 .data = &ipv6_devconf.regen_max_retry,
4608 .maxlen = sizeof(int),
4610 .proc_handler = proc_dointvec,
4613 .procname = "max_desync_factor",
4614 .data = &ipv6_devconf.max_desync_factor,
4615 .maxlen = sizeof(int),
4617 .proc_handler = proc_dointvec,
4621 .procname = "max_addresses",
4622 .data = &ipv6_devconf.max_addresses,
4623 .maxlen = sizeof(int),
4625 .proc_handler = proc_dointvec,
4628 .procname = "accept_ra_defrtr",
4629 .data = &ipv6_devconf.accept_ra_defrtr,
4630 .maxlen = sizeof(int),
4632 .proc_handler = proc_dointvec,
4635 .procname = "accept_ra_pinfo",
4636 .data = &ipv6_devconf.accept_ra_pinfo,
4637 .maxlen = sizeof(int),
4639 .proc_handler = proc_dointvec,
4641 #ifdef CONFIG_IPV6_ROUTER_PREF
4643 .procname = "accept_ra_rtr_pref",
4644 .data = &ipv6_devconf.accept_ra_rtr_pref,
4645 .maxlen = sizeof(int),
4647 .proc_handler = proc_dointvec,
4650 .procname = "router_probe_interval",
4651 .data = &ipv6_devconf.rtr_probe_interval,
4652 .maxlen = sizeof(int),
4654 .proc_handler = proc_dointvec_jiffies,
4656 #ifdef CONFIG_IPV6_ROUTE_INFO
4658 .procname = "accept_ra_rt_info_max_plen",
4659 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4660 .maxlen = sizeof(int),
4662 .proc_handler = proc_dointvec,
4667 .procname = "proxy_ndp",
4668 .data = &ipv6_devconf.proxy_ndp,
4669 .maxlen = sizeof(int),
4671 .proc_handler = proc_dointvec,
4674 .procname = "accept_source_route",
4675 .data = &ipv6_devconf.accept_source_route,
4676 .maxlen = sizeof(int),
4678 .proc_handler = proc_dointvec,
4680 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4682 .procname = "optimistic_dad",
4683 .data = &ipv6_devconf.optimistic_dad,
4684 .maxlen = sizeof(int),
4686 .proc_handler = proc_dointvec,
4690 #ifdef CONFIG_IPV6_MROUTE
4692 .procname = "mc_forwarding",
4693 .data = &ipv6_devconf.mc_forwarding,
4694 .maxlen = sizeof(int),
4696 .proc_handler = proc_dointvec,
4700 .procname = "disable_ipv6",
4701 .data = &ipv6_devconf.disable_ipv6,
4702 .maxlen = sizeof(int),
4704 .proc_handler = addrconf_sysctl_disable,
4707 .procname = "accept_dad",
4708 .data = &ipv6_devconf.accept_dad,
4709 .maxlen = sizeof(int),
4711 .proc_handler = proc_dointvec,
4714 .procname = "force_tllao",
4715 .data = &ipv6_devconf.force_tllao,
4716 .maxlen = sizeof(int),
4718 .proc_handler = proc_dointvec
4721 .procname = "ndisc_notify",
4722 .data = &ipv6_devconf.ndisc_notify,
4723 .maxlen = sizeof(int),
4725 .proc_handler = proc_dointvec
4733 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4734 struct inet6_dev *idev, struct ipv6_devconf *p)
4737 struct addrconf_sysctl_table *t;
4738 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4740 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4744 for (i = 0; t->addrconf_vars[i].data; i++) {
4745 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4746 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4747 t->addrconf_vars[i].extra2 = net;
4750 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4752 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4753 if (t->sysctl_header == NULL)
4765 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4767 struct addrconf_sysctl_table *t;
4769 if (p->sysctl == NULL)
4774 unregister_net_sysctl_table(t->sysctl_header);
4778 static void addrconf_sysctl_register(struct inet6_dev *idev)
4780 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4781 &ndisc_ifinfo_sysctl_change);
4782 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4786 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4788 __addrconf_sysctl_unregister(&idev->cnf);
4789 neigh_sysctl_unregister(idev->nd_parms);
4795 static int __net_init addrconf_init_net(struct net *net)
4798 struct ipv6_devconf *all, *dflt;
4801 all = &ipv6_devconf;
4802 dflt = &ipv6_devconf_dflt;
4804 if (!net_eq(net, &init_net)) {
4805 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4809 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4811 goto err_alloc_dflt;
4813 /* these will be inherited by all namespaces */
4814 dflt->autoconf = ipv6_defaults.autoconf;
4815 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4818 net->ipv6.devconf_all = all;
4819 net->ipv6.devconf_dflt = dflt;
4821 #ifdef CONFIG_SYSCTL
4822 err = __addrconf_sysctl_register(net, "all", NULL, all);
4826 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4832 #ifdef CONFIG_SYSCTL
4834 __addrconf_sysctl_unregister(all);
4844 static void __net_exit addrconf_exit_net(struct net *net)
4846 #ifdef CONFIG_SYSCTL
4847 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4848 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4850 if (!net_eq(net, &init_net)) {
4851 kfree(net->ipv6.devconf_dflt);
4852 kfree(net->ipv6.devconf_all);
4856 static struct pernet_operations addrconf_ops = {
4857 .init = addrconf_init_net,
4858 .exit = addrconf_exit_net,
4865 int register_inet6addr_notifier(struct notifier_block *nb)
4867 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4869 EXPORT_SYMBOL(register_inet6addr_notifier);
4871 int unregister_inet6addr_notifier(struct notifier_block *nb)
4873 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4875 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4877 static struct rtnl_af_ops inet6_ops = {
4879 .fill_link_af = inet6_fill_link_af,
4880 .get_link_af_size = inet6_get_link_af_size,
4884 * Init / cleanup code
4887 int __init addrconf_init(void)
4891 err = ipv6_addr_label_init();
4893 pr_crit("%s: cannot initialize default policy table: %d\n",
4898 err = register_pernet_subsys(&addrconf_ops);
4902 /* The addrconf netdev notifier requires that loopback_dev
4903 * has it's ipv6 private information allocated and setup
4904 * before it can bring up and give link-local addresses
4905 * to other devices which are up.
4907 * Unfortunately, loopback_dev is not necessarily the first
4908 * entry in the global dev_base list of net devices. In fact,
4909 * it is likely to be the very last entry on that list.
4910 * So this causes the notifier registry below to try and
4911 * give link-local addresses to all devices besides loopback_dev
4912 * first, then loopback_dev, which cases all the non-loopback_dev
4913 * devices to fail to get a link-local address.
4915 * So, as a temporary fix, allocate the ipv6 structure for
4916 * loopback_dev first by hand.
4917 * Longer term, all of the dependencies ipv6 has upon the loopback
4918 * device and it being up should be removed.
4921 if (!ipv6_add_dev(init_net.loopback_dev))
4927 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4928 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4930 register_netdevice_notifier(&ipv6_dev_notf);
4934 err = rtnl_af_register(&inet6_ops);
4938 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4943 /* Only the first call to __rtnl_register can fail */
4944 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4945 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4946 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4947 inet6_dump_ifaddr, NULL);
4948 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4949 inet6_dump_ifmcaddr, NULL);
4950 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4951 inet6_dump_ifacaddr, NULL);
4952 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
4955 ipv6_addr_label_rtnl_register();
4959 rtnl_af_unregister(&inet6_ops);
4961 unregister_netdevice_notifier(&ipv6_dev_notf);
4963 unregister_pernet_subsys(&addrconf_ops);
4965 ipv6_addr_label_cleanup();
4970 void addrconf_cleanup(void)
4972 struct net_device *dev;
4975 unregister_netdevice_notifier(&ipv6_dev_notf);
4976 unregister_pernet_subsys(&addrconf_ops);
4977 ipv6_addr_label_cleanup();
4981 __rtnl_af_unregister(&inet6_ops);
4983 /* clean dev list */
4984 for_each_netdev(&init_net, dev) {
4985 if (__in6_dev_get(dev) == NULL)
4987 addrconf_ifdown(dev, 1);
4989 addrconf_ifdown(init_net.loopback_dev, 2);
4994 spin_lock_bh(&addrconf_hash_lock);
4995 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4996 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4997 spin_unlock_bh(&addrconf_hash_lock);
4999 del_timer(&addr_chk_timer);