2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
25 * yoshfuji@USAGI : Fixed interval between DAD
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
41 #define pr_fmt(fmt) "IPv6: " fmt
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/kernel.h>
46 #include <linux/socket.h>
47 #include <linux/sockios.h>
48 #include <linux/net.h>
49 #include <linux/in6.h>
50 #include <linux/netdevice.h>
51 #include <linux/if_addr.h>
52 #include <linux/if_arp.h>
53 #include <linux/if_arcnet.h>
54 #include <linux/if_infiniband.h>
55 #include <linux/route.h>
56 #include <linux/inetdevice.h>
57 #include <linux/init.h>
58 #include <linux/slab.h>
60 #include <linux/sysctl.h>
62 #include <linux/capability.h>
63 #include <linux/delay.h>
64 #include <linux/notifier.h>
65 #include <linux/string.h>
66 #include <linux/hash.h>
68 #include <net/net_namespace.h>
72 #include <net/af_ieee802154.h>
73 #include <net/firewire.h>
75 #include <net/protocol.h>
76 #include <net/ndisc.h>
77 #include <net/ip6_route.h>
78 #include <net/addrconf.h>
81 #include <net/netlink.h>
82 #include <net/pkt_sched.h>
83 #include <linux/if_tunnel.h>
84 #include <linux/rtnetlink.h>
85 #include <linux/netconf.h>
86 #include <linux/random.h>
87 #include <linux/uaccess.h>
88 #include <asm/unaligned.h>
90 #include <linux/proc_fs.h>
91 #include <linux/seq_file.h>
92 #include <linux/export.h>
94 /* Set to 3 to get tracing... */
98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__)
100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0)
103 #define INFINITY_LIFE_TIME 0xFFFFFFFF
105 static inline u32 cstamp_delta(unsigned long cstamp)
107 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
111 static int addrconf_sysctl_register(struct inet6_dev *idev);
112 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
114 static inline int addrconf_sysctl_register(struct inet6_dev *idev)
119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
124 static void __ipv6_regen_rndid(struct inet6_dev *idev);
125 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
126 static void ipv6_regen_rndid(unsigned long data);
128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
129 static int ipv6_count_addresses(struct inet6_dev *idev);
132 * Configured unicast address hash table
134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
135 static DEFINE_SPINLOCK(addrconf_hash_lock);
137 static void addrconf_verify(void);
138 static void addrconf_verify_rtnl(void);
139 static void addrconf_verify_work(struct work_struct *);
141 static struct workqueue_struct *addrconf_wq;
142 static DECLARE_DELAYED_WORK(addr_chk_work, addrconf_verify_work);
144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
147 static void addrconf_type_change(struct net_device *dev,
148 unsigned long event);
149 static int addrconf_ifdown(struct net_device *dev, int how);
151 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
153 const struct net_device *dev,
154 u32 flags, u32 noflags);
156 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
157 static void addrconf_dad_work(struct work_struct *w);
158 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
159 static void addrconf_dad_run(struct inet6_dev *idev);
160 static void addrconf_rs_timer(unsigned long data);
161 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
162 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
164 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
165 struct prefix_info *pinfo);
166 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
167 struct net_device *dev);
169 static struct ipv6_devconf ipv6_devconf __read_mostly = {
171 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
172 .mtu6 = IPV6_MIN_MTU,
174 .accept_redirects = 1,
176 .force_mld_version = 0,
177 .mldv1_unsolicited_report_interval = 10 * HZ,
178 .mldv2_unsolicited_report_interval = HZ,
180 .rtr_solicits = MAX_RTR_SOLICITATIONS,
181 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
182 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
184 .temp_valid_lft = TEMP_VALID_LIFETIME,
185 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
186 .regen_max_retry = REGEN_MAX_RETRY,
187 .max_desync_factor = MAX_DESYNC_FACTOR,
188 .max_addresses = IPV6_MAX_ADDRESSES,
189 .accept_ra_defrtr = 1,
190 .accept_ra_from_local = 0,
191 .accept_ra_pinfo = 1,
192 #ifdef CONFIG_IPV6_ROUTER_PREF
193 .accept_ra_rtr_pref = 1,
194 .rtr_probe_interval = 60 * HZ,
195 #ifdef CONFIG_IPV6_ROUTE_INFO
196 .accept_ra_rt_info_max_plen = 0,
200 .accept_source_route = 0, /* we do not accept RH0 by default. */
203 .suppress_frag_ndisc = 1,
206 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
208 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
209 .mtu6 = IPV6_MIN_MTU,
211 .accept_redirects = 1,
213 .force_mld_version = 0,
214 .mldv1_unsolicited_report_interval = 10 * HZ,
215 .mldv2_unsolicited_report_interval = HZ,
217 .rtr_solicits = MAX_RTR_SOLICITATIONS,
218 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
219 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
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,
225 .max_addresses = IPV6_MAX_ADDRESSES,
226 .accept_ra_defrtr = 1,
227 .accept_ra_from_local = 0,
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. */
240 .suppress_frag_ndisc = 1,
243 /* Check if a valid qdisc is available */
244 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
246 return !qdisc_tx_is_noop(dev);
249 static void addrconf_del_rs_timer(struct inet6_dev *idev)
251 if (del_timer(&idev->rs_timer))
255 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
257 if (cancel_delayed_work(&ifp->dad_work))
261 static void addrconf_mod_rs_timer(struct inet6_dev *idev,
264 if (!timer_pending(&idev->rs_timer))
266 mod_timer(&idev->rs_timer, jiffies + when);
269 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
272 if (!delayed_work_pending(&ifp->dad_work))
274 mod_delayed_work(addrconf_wq, &ifp->dad_work, delay);
277 static int snmp6_alloc_dev(struct inet6_dev *idev)
281 idev->stats.ipv6 = alloc_percpu(struct ipstats_mib);
282 if (!idev->stats.ipv6)
285 for_each_possible_cpu(i) {
286 struct ipstats_mib *addrconf_stats;
287 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
288 u64_stats_init(&addrconf_stats->syncp);
292 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294 if (!idev->stats.icmpv6dev)
296 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298 if (!idev->stats.icmpv6msgdev)
304 kfree(idev->stats.icmpv6dev);
306 free_percpu(idev->stats.ipv6);
311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
313 struct inet6_dev *ndev;
318 if (dev->mtu < IPV6_MIN_MTU)
319 return ERR_PTR(-EINVAL);
321 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
325 rwlock_init(&ndev->lock);
327 INIT_LIST_HEAD(&ndev->addr_list);
328 setup_timer(&ndev->rs_timer, addrconf_rs_timer,
329 (unsigned long)ndev);
330 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
331 ndev->cnf.mtu6 = dev->mtu;
332 ndev->cnf.sysctl = NULL;
333 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
334 if (ndev->nd_parms == NULL) {
338 if (ndev->cnf.forwarding)
339 dev_disable_lro(dev);
340 /* We refer to the device */
343 if (snmp6_alloc_dev(ndev) < 0) {
345 "%s: cannot allocate memory for statistics; dev=%s.\n",
346 __func__, dev->name);
347 neigh_parms_release(&nd_tbl, ndev->nd_parms);
353 if (snmp6_register_dev(ndev) < 0) {
355 "%s: cannot create /proc/net/dev_snmp6/%s\n",
356 __func__, dev->name);
360 /* One reference from device. We must do this before
361 * we invoke __ipv6_regen_rndid().
365 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
366 ndev->cnf.accept_dad = -1;
368 #if IS_ENABLED(CONFIG_IPV6_SIT)
369 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
370 pr_info("%s: Disabled Multicast RS\n", dev->name);
371 ndev->cnf.rtr_solicits = 0;
375 INIT_LIST_HEAD(&ndev->tempaddr_list);
376 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
377 if ((dev->flags&IFF_LOOPBACK) ||
378 dev->type == ARPHRD_TUNNEL ||
379 dev->type == ARPHRD_TUNNEL6 ||
380 dev->type == ARPHRD_SIT ||
381 dev->type == ARPHRD_NONE) {
382 ndev->cnf.use_tempaddr = -1;
385 ipv6_regen_rndid((unsigned long) ndev);
388 ndev->token = in6addr_any;
390 if (netif_running(dev) && addrconf_qdisc_ok(dev))
391 ndev->if_flags |= IF_READY;
393 ipv6_mc_init_dev(ndev);
394 ndev->tstamp = jiffies;
395 err = addrconf_sysctl_register(ndev);
397 ipv6_mc_destroy_dev(ndev);
398 del_timer(&ndev->regen_timer);
401 /* protected by rtnl_lock */
402 rcu_assign_pointer(dev->ip6_ptr, ndev);
404 /* Join interface-local all-node multicast group */
405 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
407 /* Join all-node multicast group */
408 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
410 /* Join all-router multicast group if forwarding is set */
411 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
412 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
417 neigh_parms_release(&nd_tbl, ndev->nd_parms);
419 in6_dev_finish_destroy(ndev);
423 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
425 struct inet6_dev *idev;
429 idev = __in6_dev_get(dev);
431 idev = ipv6_add_dev(dev);
436 if (dev->flags&IFF_UP)
441 static int inet6_netconf_msgsize_devconf(int type)
443 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
444 + nla_total_size(4); /* NETCONFA_IFINDEX */
446 /* type -1 is used for ALL */
447 if (type == -1 || type == NETCONFA_FORWARDING)
448 size += nla_total_size(4);
449 #ifdef CONFIG_IPV6_MROUTE
450 if (type == -1 || type == NETCONFA_MC_FORWARDING)
451 size += nla_total_size(4);
453 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
454 size += nla_total_size(4);
459 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
460 struct ipv6_devconf *devconf, u32 portid,
461 u32 seq, int event, unsigned int flags,
464 struct nlmsghdr *nlh;
465 struct netconfmsg *ncm;
467 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
472 ncm = nlmsg_data(nlh);
473 ncm->ncm_family = AF_INET6;
475 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
476 goto nla_put_failure;
478 /* type -1 is used for ALL */
479 if ((type == -1 || type == NETCONFA_FORWARDING) &&
480 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
481 goto nla_put_failure;
482 #ifdef CONFIG_IPV6_MROUTE
483 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
484 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
485 devconf->mc_forwarding) < 0)
486 goto nla_put_failure;
488 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
489 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
490 goto nla_put_failure;
492 return nlmsg_end(skb, nlh);
495 nlmsg_cancel(skb, nlh);
499 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex,
500 struct ipv6_devconf *devconf)
505 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC);
509 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
510 RTM_NEWNETCONF, 0, type);
512 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
513 WARN_ON(err == -EMSGSIZE);
517 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC);
520 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
523 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
524 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
525 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
526 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
529 static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
530 struct nlmsghdr *nlh)
532 struct net *net = sock_net(in_skb->sk);
533 struct nlattr *tb[NETCONFA_MAX+1];
534 struct netconfmsg *ncm;
536 struct ipv6_devconf *devconf;
537 struct inet6_dev *in6_dev;
538 struct net_device *dev;
542 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
543 devconf_ipv6_policy);
548 if (!tb[NETCONFA_IFINDEX])
551 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
553 case NETCONFA_IFINDEX_ALL:
554 devconf = net->ipv6.devconf_all;
556 case NETCONFA_IFINDEX_DEFAULT:
557 devconf = net->ipv6.devconf_dflt;
560 dev = __dev_get_by_index(net, ifindex);
563 in6_dev = __in6_dev_get(dev);
566 devconf = &in6_dev->cnf;
571 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC);
575 err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
576 NETLINK_CB(in_skb).portid,
577 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
580 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
581 WARN_ON(err == -EMSGSIZE);
585 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
590 static int inet6_netconf_dump_devconf(struct sk_buff *skb,
591 struct netlink_callback *cb)
593 struct net *net = sock_net(skb->sk);
596 struct net_device *dev;
597 struct inet6_dev *idev;
598 struct hlist_head *head;
601 s_idx = idx = cb->args[1];
603 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
605 head = &net->dev_index_head[h];
607 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
609 hlist_for_each_entry_rcu(dev, head, index_hlist) {
612 idev = __in6_dev_get(dev);
616 if (inet6_netconf_fill_devconf(skb, dev->ifindex,
618 NETLINK_CB(cb->skb).portid,
626 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
632 if (h == NETDEV_HASHENTRIES) {
633 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
634 net->ipv6.devconf_all,
635 NETLINK_CB(cb->skb).portid,
637 RTM_NEWNETCONF, NLM_F_MULTI,
643 if (h == NETDEV_HASHENTRIES + 1) {
644 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
645 net->ipv6.devconf_dflt,
646 NETLINK_CB(cb->skb).portid,
648 RTM_NEWNETCONF, NLM_F_MULTI,
662 static void dev_forward_change(struct inet6_dev *idev)
664 struct net_device *dev;
665 struct inet6_ifaddr *ifa;
670 if (idev->cnf.forwarding)
671 dev_disable_lro(dev);
672 if (dev->flags & IFF_MULTICAST) {
673 if (idev->cnf.forwarding) {
674 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
675 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
676 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
678 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
679 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
680 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
684 list_for_each_entry(ifa, &idev->addr_list, if_list) {
685 if (ifa->flags&IFA_F_TENTATIVE)
687 if (idev->cnf.forwarding)
688 addrconf_join_anycast(ifa);
690 addrconf_leave_anycast(ifa);
692 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING,
693 dev->ifindex, &idev->cnf);
697 static void addrconf_forward_change(struct net *net, __s32 newf)
699 struct net_device *dev;
700 struct inet6_dev *idev;
702 for_each_netdev(net, dev) {
703 idev = __in6_dev_get(dev);
705 int changed = (!idev->cnf.forwarding) ^ (!newf);
706 idev->cnf.forwarding = newf;
708 dev_forward_change(idev);
713 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
719 return restart_syscall();
721 net = (struct net *)table->extra2;
725 if (p == &net->ipv6.devconf_dflt->forwarding) {
726 if ((!newf) ^ (!old))
727 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
728 NETCONFA_IFINDEX_DEFAULT,
729 net->ipv6.devconf_dflt);
734 if (p == &net->ipv6.devconf_all->forwarding) {
735 net->ipv6.devconf_dflt->forwarding = newf;
736 addrconf_forward_change(net, newf);
737 if ((!newf) ^ (!old))
738 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING,
739 NETCONFA_IFINDEX_ALL,
740 net->ipv6.devconf_all);
741 } else if ((!newf) ^ (!old))
742 dev_forward_change((struct inet6_dev *)table->extra1);
746 rt6_purge_dflt_routers(net);
751 /* Nobody refers to this ifaddr, destroy it */
752 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
754 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
756 #ifdef NET_REFCNT_DEBUG
757 pr_debug("%s\n", __func__);
760 in6_dev_put(ifp->idev);
762 if (cancel_delayed_work(&ifp->dad_work))
763 pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
766 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
767 pr_warn("Freeing alive inet6 address %p\n", ifp);
776 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
779 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
782 * Each device address list is sorted in order of scope -
783 * global before linklocal.
785 list_for_each(p, &idev->addr_list) {
786 struct inet6_ifaddr *ifa
787 = list_entry(p, struct inet6_ifaddr, if_list);
788 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
792 list_add_tail(&ifp->if_list, p);
795 static u32 inet6_addr_hash(const struct in6_addr *addr)
797 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT);
800 /* On success it returns ifp with increased reference count */
802 static struct inet6_ifaddr *
803 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
804 const struct in6_addr *peer_addr, int pfxlen,
805 int scope, u32 flags, u32 valid_lft, u32 prefered_lft)
807 struct inet6_ifaddr *ifa = NULL;
811 int addr_type = ipv6_addr_type(addr);
813 if (addr_type == IPV6_ADDR_ANY ||
814 addr_type & IPV6_ADDR_MULTICAST ||
815 (!(idev->dev->flags & IFF_LOOPBACK) &&
816 addr_type & IPV6_ADDR_LOOPBACK))
817 return ERR_PTR(-EADDRNOTAVAIL);
821 err = -ENODEV; /*XXX*/
825 if (idev->cnf.disable_ipv6) {
830 spin_lock(&addrconf_hash_lock);
832 /* Ignore adding duplicate addresses on an interface */
833 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
834 ADBG("ipv6_add_addr: already assigned\n");
839 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
842 ADBG("ipv6_add_addr: malloc failed\n");
847 rt = addrconf_dst_alloc(idev, addr, false);
853 neigh_parms_data_state_setall(idev->nd_parms);
857 ifa->peer_addr = *peer_addr;
859 spin_lock_init(&ifa->lock);
860 spin_lock_init(&ifa->state_lock);
861 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
862 INIT_HLIST_NODE(&ifa->addr_lst);
864 ifa->prefix_len = pfxlen;
865 ifa->flags = flags | IFA_F_TENTATIVE;
866 ifa->valid_lft = valid_lft;
867 ifa->prefered_lft = prefered_lft;
868 ifa->cstamp = ifa->tstamp = jiffies;
869 ifa->tokenized = false;
878 /* Add to big hash table */
879 hash = inet6_addr_hash(addr);
881 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
882 spin_unlock(&addrconf_hash_lock);
884 write_lock(&idev->lock);
885 /* Add to inet6_dev unicast addr list. */
886 ipv6_link_dev_addr(idev, ifa);
888 if (ifa->flags&IFA_F_TEMPORARY) {
889 list_add(&ifa->tmp_list, &idev->tempaddr_list);
894 write_unlock(&idev->lock);
896 rcu_read_unlock_bh();
898 if (likely(err == 0))
899 inet6addr_notifier_call_chain(NETDEV_UP, ifa);
907 spin_unlock(&addrconf_hash_lock);
911 enum cleanup_prefix_rt_t {
912 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */
913 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */
914 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
918 * Check, whether the prefix for ifp would still need a prefix route
919 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
922 * 1) we don't purge prefix if address was not permanent.
923 * prefix is managed by its own lifetime.
924 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
925 * 3) if there are no addresses, delete prefix.
926 * 4) if there are still other permanent address(es),
927 * corresponding prefix is still permanent.
928 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
929 * don't purge the prefix, assume user space is managing it.
930 * 6) otherwise, update prefix lifetime to the
931 * longest valid lifetime among the corresponding
932 * addresses on the device.
933 * Note: subsequent RA will update lifetime.
935 static enum cleanup_prefix_rt_t
936 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
938 struct inet6_ifaddr *ifa;
939 struct inet6_dev *idev = ifp->idev;
940 unsigned long lifetime;
941 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
945 list_for_each_entry(ifa, &idev->addr_list, if_list) {
948 if (!ipv6_prefix_equal(&ifa->addr, &ifp->addr,
951 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
952 return CLEANUP_PREFIX_RT_NOP;
954 action = CLEANUP_PREFIX_RT_EXPIRE;
956 spin_lock(&ifa->lock);
958 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
960 * Note: Because this address is
961 * not permanent, lifetime <
962 * LONG_MAX / HZ here.
964 if (time_before(*expires, ifa->tstamp + lifetime * HZ))
965 *expires = ifa->tstamp + lifetime * HZ;
966 spin_unlock(&ifa->lock);
973 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, bool del_rt)
977 rt = addrconf_get_prefix_route(&ifp->addr,
980 0, RTF_GATEWAY | RTF_DEFAULT);
985 if (!(rt->rt6i_flags & RTF_EXPIRES))
986 rt6_set_expires(rt, expires);
993 /* This function wants to get referenced ifp and releases it before return */
995 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
998 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
999 unsigned long expires;
1003 spin_lock_bh(&ifp->state_lock);
1005 ifp->state = INET6_IFADDR_STATE_DEAD;
1006 spin_unlock_bh(&ifp->state_lock);
1008 if (state == INET6_IFADDR_STATE_DEAD)
1011 spin_lock_bh(&addrconf_hash_lock);
1012 hlist_del_init_rcu(&ifp->addr_lst);
1013 spin_unlock_bh(&addrconf_hash_lock);
1015 write_lock_bh(&ifp->idev->lock);
1017 if (ifp->flags&IFA_F_TEMPORARY) {
1018 list_del(&ifp->tmp_list);
1020 in6_ifa_put(ifp->ifpub);
1026 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
1027 action = check_cleanup_prefix_route(ifp, &expires);
1029 list_del_init(&ifp->if_list);
1032 write_unlock_bh(&ifp->idev->lock);
1034 addrconf_del_dad_work(ifp);
1036 ipv6_ifa_notify(RTM_DELADDR, ifp);
1038 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
1040 if (action != CLEANUP_PREFIX_RT_NOP) {
1041 cleanup_prefix_route(ifp, expires,
1042 action == CLEANUP_PREFIX_RT_DEL);
1045 /* clean up prefsrc entries */
1046 rt6_remove_prefsrc(ifp);
1051 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
1053 struct inet6_dev *idev = ifp->idev;
1054 struct in6_addr addr, *tmpaddr;
1055 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
1056 unsigned long regen_advance;
1060 unsigned long now = jiffies;
1062 write_lock_bh(&idev->lock);
1064 spin_lock_bh(&ift->lock);
1065 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
1066 spin_unlock_bh(&ift->lock);
1073 if (idev->cnf.use_tempaddr <= 0) {
1074 write_unlock_bh(&idev->lock);
1075 pr_info("%s: use_tempaddr is disabled\n", __func__);
1080 spin_lock_bh(&ifp->lock);
1081 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
1082 idev->cnf.use_tempaddr = -1; /*XXX*/
1083 spin_unlock_bh(&ifp->lock);
1084 write_unlock_bh(&idev->lock);
1085 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
1092 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
1093 __ipv6_try_regen_rndid(idev, tmpaddr);
1094 memcpy(&addr.s6_addr[8], idev->rndid, 8);
1095 age = (now - ifp->tstamp) / HZ;
1096 tmp_valid_lft = min_t(__u32,
1098 idev->cnf.temp_valid_lft + age);
1099 tmp_prefered_lft = min_t(__u32,
1101 idev->cnf.temp_prefered_lft + age -
1102 idev->cnf.max_desync_factor);
1103 tmp_plen = ifp->prefix_len;
1104 tmp_tstamp = ifp->tstamp;
1105 spin_unlock_bh(&ifp->lock);
1107 regen_advance = idev->cnf.regen_max_retry *
1108 idev->cnf.dad_transmits *
1109 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ;
1110 write_unlock_bh(&idev->lock);
1112 /* A temporary address is created only if this calculated Preferred
1113 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
1114 * an implementation must not create a temporary address with a zero
1115 * Preferred Lifetime.
1116 * Use age calculation as in addrconf_verify to avoid unnecessary
1117 * temporary addresses being generated.
1119 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
1120 if (tmp_prefered_lft <= regen_advance + age) {
1127 addr_flags = IFA_F_TEMPORARY;
1128 /* set in addrconf_prefix_rcv() */
1129 if (ifp->flags & IFA_F_OPTIMISTIC)
1130 addr_flags |= IFA_F_OPTIMISTIC;
1132 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen,
1133 ipv6_addr_scope(&addr), addr_flags,
1134 tmp_valid_lft, tmp_prefered_lft);
1138 pr_info("%s: retry temporary address regeneration\n", __func__);
1140 write_lock_bh(&idev->lock);
1144 spin_lock_bh(&ift->lock);
1147 ift->tstamp = tmp_tstamp;
1148 spin_unlock_bh(&ift->lock);
1150 addrconf_dad_start(ift);
1158 * Choose an appropriate source address (RFC3484)
1161 IPV6_SADDR_RULE_INIT = 0,
1162 IPV6_SADDR_RULE_LOCAL,
1163 IPV6_SADDR_RULE_SCOPE,
1164 IPV6_SADDR_RULE_PREFERRED,
1165 #ifdef CONFIG_IPV6_MIP6
1166 IPV6_SADDR_RULE_HOA,
1168 IPV6_SADDR_RULE_OIF,
1169 IPV6_SADDR_RULE_LABEL,
1170 IPV6_SADDR_RULE_PRIVACY,
1171 IPV6_SADDR_RULE_ORCHID,
1172 IPV6_SADDR_RULE_PREFIX,
1176 struct ipv6_saddr_score {
1179 struct inet6_ifaddr *ifa;
1180 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
1185 struct ipv6_saddr_dst {
1186 const struct in6_addr *addr;
1193 static inline int ipv6_saddr_preferred(int type)
1195 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1200 static int ipv6_get_saddr_eval(struct net *net,
1201 struct ipv6_saddr_score *score,
1202 struct ipv6_saddr_dst *dst,
1207 if (i <= score->rule) {
1209 case IPV6_SADDR_RULE_SCOPE:
1210 ret = score->scopedist;
1212 case IPV6_SADDR_RULE_PREFIX:
1213 ret = score->matchlen;
1216 ret = !!test_bit(i, score->scorebits);
1222 case IPV6_SADDR_RULE_INIT:
1223 /* Rule 0: remember if hiscore is not ready yet */
1226 case IPV6_SADDR_RULE_LOCAL:
1227 /* Rule 1: Prefer same address */
1228 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1230 case IPV6_SADDR_RULE_SCOPE:
1231 /* Rule 2: Prefer appropriate scope
1236 * ---+--+-+---> scope
1238 * | d is scope of the destination.
1240 * | \ <- smaller scope is better if
1241 * B-15 | \ if scope is enough for destination.
1242 * | ret = B - scope (-1 <= scope >= d <= 15).
1244 * |/ <- greater is better
1245 * -C / if scope is not enough for destination.
1246 * /| ret = scope - C (-1 <= d < scope <= 15).
1248 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1249 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1250 * Assume B = 0 and we get C > 29.
1252 ret = __ipv6_addr_src_scope(score->addr_type);
1253 if (ret >= dst->scope)
1256 ret -= 128; /* 30 is enough */
1257 score->scopedist = ret;
1259 case IPV6_SADDR_RULE_PREFERRED:
1260 /* Rule 3: Avoid deprecated and optimistic addresses */
1261 ret = ipv6_saddr_preferred(score->addr_type) ||
1262 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1264 #ifdef CONFIG_IPV6_MIP6
1265 case IPV6_SADDR_RULE_HOA:
1267 /* Rule 4: Prefer home address */
1268 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1269 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1273 case IPV6_SADDR_RULE_OIF:
1274 /* Rule 5: Prefer outgoing interface */
1275 ret = (!dst->ifindex ||
1276 dst->ifindex == score->ifa->idev->dev->ifindex);
1278 case IPV6_SADDR_RULE_LABEL:
1279 /* Rule 6: Prefer matching label */
1280 ret = ipv6_addr_label(net,
1281 &score->ifa->addr, score->addr_type,
1282 score->ifa->idev->dev->ifindex) == dst->label;
1284 case IPV6_SADDR_RULE_PRIVACY:
1286 /* Rule 7: Prefer public address
1287 * Note: prefer temporary address if use_tempaddr >= 2
1289 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1290 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1291 score->ifa->idev->cnf.use_tempaddr >= 2;
1292 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1295 case IPV6_SADDR_RULE_ORCHID:
1296 /* Rule 8-: Prefer ORCHID vs ORCHID or
1297 * non-ORCHID vs non-ORCHID
1299 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1300 ipv6_addr_orchid(dst->addr));
1302 case IPV6_SADDR_RULE_PREFIX:
1303 /* Rule 8: Use longest matching prefix */
1304 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
1305 if (ret > score->ifa->prefix_len)
1306 ret = score->ifa->prefix_len;
1307 score->matchlen = ret;
1314 __set_bit(i, score->scorebits);
1320 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
1321 const struct in6_addr *daddr, unsigned int prefs,
1322 struct in6_addr *saddr)
1324 struct ipv6_saddr_score scores[2],
1325 *score = &scores[0], *hiscore = &scores[1];
1326 struct ipv6_saddr_dst dst;
1327 struct net_device *dev;
1330 dst_type = __ipv6_addr_type(daddr);
1332 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1333 dst.scope = __ipv6_addr_src_scope(dst_type);
1334 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1338 hiscore->ifa = NULL;
1342 for_each_netdev_rcu(net, dev) {
1343 struct inet6_dev *idev;
1345 /* Candidate Source Address (section 4)
1346 * - multicast and link-local destination address,
1347 * the set of candidate source address MUST only
1348 * include addresses assigned to interfaces
1349 * belonging to the same link as the outgoing
1351 * (- For site-local destination addresses, the
1352 * set of candidate source addresses MUST only
1353 * include addresses assigned to interfaces
1354 * belonging to the same site as the outgoing
1357 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1358 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1359 dst.ifindex && dev->ifindex != dst.ifindex)
1362 idev = __in6_dev_get(dev);
1366 read_lock_bh(&idev->lock);
1367 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1371 * - Tentative Address (RFC2462 section 5.4)
1372 * - A tentative address is not considered
1373 * "assigned to an interface" in the traditional
1374 * sense, unless it is also flagged as optimistic.
1375 * - Candidate Source Address (section 4)
1376 * - In any case, anycast addresses, multicast
1377 * addresses, and the unspecified address MUST
1378 * NOT be included in a candidate set.
1380 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1381 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1384 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1386 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1387 score->addr_type & IPV6_ADDR_MULTICAST)) {
1388 LIMIT_NETDEBUG(KERN_DEBUG
1389 "ADDRCONF: unspecified / multicast address "
1390 "assigned as unicast address on %s",
1396 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1398 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1399 int minihiscore, miniscore;
1401 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1402 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1404 if (minihiscore > miniscore) {
1405 if (i == IPV6_SADDR_RULE_SCOPE &&
1406 score->scopedist > 0) {
1409 * each remaining entry
1410 * has too small (not enough)
1411 * scope, because ifa entries
1412 * are sorted by their scope
1418 } else if (minihiscore < miniscore) {
1420 in6_ifa_put(hiscore->ifa);
1422 in6_ifa_hold(score->ifa);
1424 swap(hiscore, score);
1426 /* restore our iterator */
1427 score->ifa = hiscore->ifa;
1434 read_unlock_bh(&idev->lock);
1439 return -EADDRNOTAVAIL;
1441 *saddr = hiscore->ifa->addr;
1442 in6_ifa_put(hiscore->ifa);
1445 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1447 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1450 struct inet6_ifaddr *ifp;
1451 int err = -EADDRNOTAVAIL;
1453 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1454 if (ifp->scope > IFA_LINK)
1456 if (ifp->scope == IFA_LINK &&
1457 !(ifp->flags & banned_flags)) {
1466 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1469 struct inet6_dev *idev;
1470 int err = -EADDRNOTAVAIL;
1473 idev = __in6_dev_get(dev);
1475 read_lock_bh(&idev->lock);
1476 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1477 read_unlock_bh(&idev->lock);
1483 static int ipv6_count_addresses(struct inet6_dev *idev)
1486 struct inet6_ifaddr *ifp;
1488 read_lock_bh(&idev->lock);
1489 list_for_each_entry(ifp, &idev->addr_list, if_list)
1491 read_unlock_bh(&idev->lock);
1495 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1496 const struct net_device *dev, int strict)
1498 struct inet6_ifaddr *ifp;
1499 unsigned int hash = inet6_addr_hash(addr);
1502 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) {
1503 if (!net_eq(dev_net(ifp->idev->dev), net))
1505 if (ipv6_addr_equal(&ifp->addr, addr) &&
1506 !(ifp->flags&IFA_F_TENTATIVE) &&
1507 (dev == NULL || ifp->idev->dev == dev ||
1508 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1509 rcu_read_unlock_bh();
1514 rcu_read_unlock_bh();
1517 EXPORT_SYMBOL(ipv6_chk_addr);
1519 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1520 struct net_device *dev)
1522 unsigned int hash = inet6_addr_hash(addr);
1523 struct inet6_ifaddr *ifp;
1525 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) {
1526 if (!net_eq(dev_net(ifp->idev->dev), net))
1528 if (ipv6_addr_equal(&ifp->addr, addr)) {
1529 if (dev == NULL || ifp->idev->dev == dev)
1536 /* Compares an address/prefix_len with addresses on device @dev.
1537 * If one is found it returns true.
1539 bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
1540 const unsigned int prefix_len, struct net_device *dev)
1542 struct inet6_dev *idev;
1543 struct inet6_ifaddr *ifa;
1547 idev = __in6_dev_get(dev);
1549 read_lock_bh(&idev->lock);
1550 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1551 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
1555 read_unlock_bh(&idev->lock);
1561 EXPORT_SYMBOL(ipv6_chk_custom_prefix);
1563 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1565 struct inet6_dev *idev;
1566 struct inet6_ifaddr *ifa;
1571 idev = __in6_dev_get(dev);
1573 read_lock_bh(&idev->lock);
1574 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1575 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1580 read_unlock_bh(&idev->lock);
1585 EXPORT_SYMBOL(ipv6_chk_prefix);
1587 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1588 struct net_device *dev, int strict)
1590 struct inet6_ifaddr *ifp, *result = NULL;
1591 unsigned int hash = inet6_addr_hash(addr);
1594 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
1595 if (!net_eq(dev_net(ifp->idev->dev), net))
1597 if (ipv6_addr_equal(&ifp->addr, addr)) {
1598 if (dev == NULL || ifp->idev->dev == dev ||
1599 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1606 rcu_read_unlock_bh();
1611 /* Gets referenced address, destroys ifaddr */
1613 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1615 if (ifp->flags&IFA_F_PERMANENT) {
1616 spin_lock_bh(&ifp->lock);
1617 addrconf_del_dad_work(ifp);
1618 ifp->flags |= IFA_F_TENTATIVE;
1620 ifp->flags |= IFA_F_DADFAILED;
1621 spin_unlock_bh(&ifp->lock);
1623 ipv6_ifa_notify(0, ifp);
1625 } else if (ifp->flags&IFA_F_TEMPORARY) {
1626 struct inet6_ifaddr *ifpub;
1627 spin_lock_bh(&ifp->lock);
1630 in6_ifa_hold(ifpub);
1631 spin_unlock_bh(&ifp->lock);
1632 ipv6_create_tempaddr(ifpub, ifp);
1635 spin_unlock_bh(&ifp->lock);
1643 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1647 spin_lock_bh(&ifp->state_lock);
1648 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1649 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1652 spin_unlock_bh(&ifp->state_lock);
1657 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1659 struct inet6_dev *idev = ifp->idev;
1661 if (addrconf_dad_end(ifp)) {
1666 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1667 ifp->idev->dev->name, &ifp->addr);
1669 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1670 struct in6_addr addr;
1672 addr.s6_addr32[0] = htonl(0xfe800000);
1673 addr.s6_addr32[1] = 0;
1675 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1676 ipv6_addr_equal(&ifp->addr, &addr)) {
1677 /* DAD failed for link-local based on MAC address */
1678 idev->cnf.disable_ipv6 = 1;
1680 pr_info("%s: IPv6 being disabled!\n",
1681 ifp->idev->dev->name);
1685 spin_lock_bh(&ifp->state_lock);
1686 /* transition from _POSTDAD to _ERRDAD */
1687 ifp->state = INET6_IFADDR_STATE_ERRDAD;
1688 spin_unlock_bh(&ifp->state_lock);
1690 addrconf_mod_dad_work(ifp, 0);
1693 /* Join to solicited addr multicast group. */
1695 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1697 struct in6_addr maddr;
1701 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1704 addrconf_addr_solict_mult(addr, &maddr);
1705 ipv6_dev_mc_inc(dev, &maddr);
1708 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1710 struct in6_addr maddr;
1714 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1717 addrconf_addr_solict_mult(addr, &maddr);
1718 __ipv6_dev_mc_dec(idev, &maddr);
1721 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1723 struct in6_addr addr;
1727 if (ifp->prefix_len >= 127) /* RFC 6164 */
1729 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1730 if (ipv6_addr_any(&addr))
1732 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1735 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1737 struct in6_addr addr;
1741 if (ifp->prefix_len >= 127) /* RFC 6164 */
1743 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1744 if (ipv6_addr_any(&addr))
1746 __ipv6_dev_ac_dec(ifp->idev, &addr);
1749 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1751 if (dev->addr_len != ETH_ALEN)
1753 memcpy(eui, dev->dev_addr, 3);
1754 memcpy(eui + 5, dev->dev_addr + 3, 3);
1757 * The zSeries OSA network cards can be shared among various
1758 * OS instances, but the OSA cards have only one MAC address.
1759 * This leads to duplicate address conflicts in conjunction
1760 * with IPv6 if more than one instance uses the same card.
1762 * The driver for these cards can deliver a unique 16-bit
1763 * identifier for each instance sharing the same card. It is
1764 * placed instead of 0xFFFE in the interface identifier. The
1765 * "u" bit of the interface identifier is not inverted in this
1766 * case. Hence the resulting interface identifier has local
1767 * scope according to RFC2373.
1770 eui[3] = (dev->dev_id >> 8) & 0xFF;
1771 eui[4] = dev->dev_id & 0xFF;
1780 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1782 if (dev->addr_len != IEEE802154_ADDR_LEN)
1784 memcpy(eui, dev->dev_addr, 8);
1789 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
1791 union fwnet_hwaddr *ha;
1793 if (dev->addr_len != FWNET_ALEN)
1796 ha = (union fwnet_hwaddr *)dev->dev_addr;
1798 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
1803 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1805 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1806 if (dev->addr_len != ARCNET_ALEN)
1809 eui[7] = *(u8 *)dev->dev_addr;
1813 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1815 if (dev->addr_len != INFINIBAND_ALEN)
1817 memcpy(eui, dev->dev_addr + 12, 8);
1822 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1826 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1827 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1828 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1829 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1830 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1831 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1835 memcpy(eui + 4, &addr, 4);
1839 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1841 if (dev->priv_flags & IFF_ISATAP)
1842 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1846 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1848 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1851 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
1853 memcpy(eui, dev->perm_addr, 3);
1854 memcpy(eui + 5, dev->perm_addr + 3, 3);
1861 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1863 switch (dev->type) {
1866 return addrconf_ifid_eui48(eui, dev);
1868 return addrconf_ifid_arcnet(eui, dev);
1869 case ARPHRD_INFINIBAND:
1870 return addrconf_ifid_infiniband(eui, dev);
1872 return addrconf_ifid_sit(eui, dev);
1874 return addrconf_ifid_gre(eui, dev);
1875 case ARPHRD_6LOWPAN:
1876 case ARPHRD_IEEE802154:
1877 return addrconf_ifid_eui64(eui, dev);
1878 case ARPHRD_IEEE1394:
1879 return addrconf_ifid_ieee1394(eui, dev);
1880 case ARPHRD_TUNNEL6:
1881 return addrconf_ifid_ip6tnl(eui, dev);
1886 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1889 struct inet6_ifaddr *ifp;
1891 read_lock_bh(&idev->lock);
1892 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
1893 if (ifp->scope > IFA_LINK)
1895 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1896 memcpy(eui, ifp->addr.s6_addr+8, 8);
1901 read_unlock_bh(&idev->lock);
1905 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1906 static void __ipv6_regen_rndid(struct inet6_dev *idev)
1909 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1910 idev->rndid[0] &= ~0x02;
1913 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1914 * check if generated address is not inappropriate
1916 * - Reserved subnet anycast (RFC 2526)
1917 * 11111101 11....11 1xxxxxxx
1918 * - ISATAP (RFC4214) 6.1
1919 * 00-00-5E-FE-xx-xx-xx-xx
1921 * - XXX: already assigned to an address on the device
1923 if (idev->rndid[0] == 0xfd &&
1924 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1925 (idev->rndid[7]&0x80))
1927 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1928 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1930 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1935 static void ipv6_regen_rndid(unsigned long data)
1937 struct inet6_dev *idev = (struct inet6_dev *) data;
1938 unsigned long expires;
1941 write_lock_bh(&idev->lock);
1946 __ipv6_regen_rndid(idev);
1949 idev->cnf.temp_prefered_lft * HZ -
1950 idev->cnf.regen_max_retry * idev->cnf.dad_transmits *
1951 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) -
1952 idev->cnf.max_desync_factor * HZ;
1953 if (time_before(expires, jiffies)) {
1954 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1955 __func__, idev->dev->name);
1959 if (!mod_timer(&idev->regen_timer, expires))
1963 write_unlock_bh(&idev->lock);
1964 rcu_read_unlock_bh();
1968 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1970 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1971 __ipv6_regen_rndid(idev);
1979 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1980 unsigned long expires, u32 flags)
1982 struct fib6_config cfg = {
1983 .fc_table = RT6_TABLE_PREFIX,
1984 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1985 .fc_ifindex = dev->ifindex,
1986 .fc_expires = expires,
1988 .fc_flags = RTF_UP | flags,
1989 .fc_nlinfo.nl_net = dev_net(dev),
1990 .fc_protocol = RTPROT_KERNEL,
1995 /* Prevent useless cloning on PtP SIT.
1996 This thing is done here expecting that the whole
1997 class of non-broadcast devices need not cloning.
1999 #if IS_ENABLED(CONFIG_IPV6_SIT)
2000 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
2001 cfg.fc_flags |= RTF_NONEXTHOP;
2004 ip6_route_add(&cfg);
2008 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
2010 const struct net_device *dev,
2011 u32 flags, u32 noflags)
2013 struct fib6_node *fn;
2014 struct rt6_info *rt = NULL;
2015 struct fib6_table *table;
2017 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
2021 read_lock_bh(&table->tb6_lock);
2022 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
2025 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2026 if (rt->dst.dev->ifindex != dev->ifindex)
2028 if ((rt->rt6i_flags & flags) != flags)
2030 if ((rt->rt6i_flags & noflags) != 0)
2036 read_unlock_bh(&table->tb6_lock);
2041 /* Create "default" multicast route to the interface */
2043 static void addrconf_add_mroute(struct net_device *dev)
2045 struct fib6_config cfg = {
2046 .fc_table = RT6_TABLE_LOCAL,
2047 .fc_metric = IP6_RT_PRIO_ADDRCONF,
2048 .fc_ifindex = dev->ifindex,
2051 .fc_nlinfo.nl_net = dev_net(dev),
2054 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
2056 ip6_route_add(&cfg);
2059 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
2061 struct inet6_dev *idev;
2065 idev = ipv6_find_idev(dev);
2067 return ERR_PTR(-ENOBUFS);
2069 if (idev->cnf.disable_ipv6)
2070 return ERR_PTR(-EACCES);
2072 /* Add default multicast route */
2073 if (!(dev->flags & IFF_LOOPBACK))
2074 addrconf_add_mroute(dev);
2079 static void manage_tempaddrs(struct inet6_dev *idev,
2080 struct inet6_ifaddr *ifp,
2081 __u32 valid_lft, __u32 prefered_lft,
2082 bool create, unsigned long now)
2085 struct inet6_ifaddr *ift;
2087 read_lock_bh(&idev->lock);
2088 /* update all temporary addresses in the list */
2089 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
2090 int age, max_valid, max_prefered;
2092 if (ifp != ift->ifpub)
2095 /* RFC 4941 section 3.3:
2096 * If a received option will extend the lifetime of a public
2097 * address, the lifetimes of temporary addresses should
2098 * be extended, subject to the overall constraint that no
2099 * temporary addresses should ever remain "valid" or "preferred"
2100 * for a time longer than (TEMP_VALID_LIFETIME) or
2101 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
2103 age = (now - ift->cstamp) / HZ;
2104 max_valid = idev->cnf.temp_valid_lft - age;
2108 max_prefered = idev->cnf.temp_prefered_lft -
2109 idev->cnf.max_desync_factor - age;
2110 if (max_prefered < 0)
2113 if (valid_lft > max_valid)
2114 valid_lft = max_valid;
2116 if (prefered_lft > max_prefered)
2117 prefered_lft = max_prefered;
2119 spin_lock(&ift->lock);
2121 ift->valid_lft = valid_lft;
2122 ift->prefered_lft = prefered_lft;
2124 if (prefered_lft > 0)
2125 ift->flags &= ~IFA_F_DEPRECATED;
2127 spin_unlock(&ift->lock);
2128 if (!(flags&IFA_F_TENTATIVE))
2129 ipv6_ifa_notify(0, ift);
2132 if ((create || list_empty(&idev->tempaddr_list)) &&
2133 idev->cnf.use_tempaddr > 0) {
2134 /* When a new public address is created as described
2135 * in [ADDRCONF], also create a new temporary address.
2136 * Also create a temporary address if it's enabled but
2137 * no temporary address currently exists.
2139 read_unlock_bh(&idev->lock);
2140 ipv6_create_tempaddr(ifp, NULL);
2142 read_unlock_bh(&idev->lock);
2146 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
2148 struct prefix_info *pinfo;
2152 struct inet6_dev *in6_dev;
2153 struct net *net = dev_net(dev);
2155 pinfo = (struct prefix_info *) opt;
2157 if (len < sizeof(struct prefix_info)) {
2158 ADBG("addrconf: prefix option too short\n");
2163 * Validation checks ([ADDRCONF], page 19)
2166 addr_type = ipv6_addr_type(&pinfo->prefix);
2168 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
2171 valid_lft = ntohl(pinfo->valid);
2172 prefered_lft = ntohl(pinfo->prefered);
2174 if (prefered_lft > valid_lft) {
2175 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
2179 in6_dev = in6_dev_get(dev);
2181 if (in6_dev == NULL) {
2182 net_dbg_ratelimited("addrconf: device %s not configured\n",
2188 * Two things going on here:
2189 * 1) Add routes for on-link prefixes
2190 * 2) Configure prefixes with the auto flag set
2193 if (pinfo->onlink) {
2194 struct rt6_info *rt;
2195 unsigned long rt_expires;
2197 /* Avoid arithmetic overflow. Really, we could
2198 * save rt_expires in seconds, likely valid_lft,
2199 * but it would require division in fib gc, that it
2203 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
2205 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
2207 if (addrconf_finite_timeout(rt_expires))
2210 rt = addrconf_get_prefix_route(&pinfo->prefix,
2213 RTF_ADDRCONF | RTF_PREFIX_RT,
2214 RTF_GATEWAY | RTF_DEFAULT);
2217 /* Autoconf prefix route */
2218 if (valid_lft == 0) {
2221 } else if (addrconf_finite_timeout(rt_expires)) {
2223 rt6_set_expires(rt, jiffies + rt_expires);
2225 rt6_clean_expires(rt);
2227 } else if (valid_lft) {
2228 clock_t expires = 0;
2229 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
2230 if (addrconf_finite_timeout(rt_expires)) {
2232 flags |= RTF_EXPIRES;
2233 expires = jiffies_to_clock_t(rt_expires);
2235 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
2236 dev, expires, flags);
2241 /* Try to figure out our local address for this prefix */
2243 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
2244 struct inet6_ifaddr *ifp;
2245 struct in6_addr addr;
2246 int create = 0, update_lft = 0;
2247 bool tokenized = false;
2249 if (pinfo->prefix_len == 64) {
2250 memcpy(&addr, &pinfo->prefix, 8);
2252 if (!ipv6_addr_any(&in6_dev->token)) {
2253 read_lock_bh(&in6_dev->lock);
2254 memcpy(addr.s6_addr + 8,
2255 in6_dev->token.s6_addr + 8, 8);
2256 read_unlock_bh(&in6_dev->lock);
2258 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
2259 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
2260 in6_dev_put(in6_dev);
2265 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
2267 in6_dev_put(in6_dev);
2272 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
2274 if (ifp == NULL && valid_lft) {
2275 int max_addresses = in6_dev->cnf.max_addresses;
2278 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2279 if (in6_dev->cnf.optimistic_dad &&
2280 !net->ipv6.devconf_all->forwarding && sllao)
2281 addr_flags = IFA_F_OPTIMISTIC;
2284 /* Do not allow to create too much of autoconfigured
2285 * addresses; this would be too easy way to crash kernel.
2287 if (!max_addresses ||
2288 ipv6_count_addresses(in6_dev) < max_addresses)
2289 ifp = ipv6_add_addr(in6_dev, &addr, NULL,
2291 addr_type&IPV6_ADDR_SCOPE_MASK,
2292 addr_flags, valid_lft,
2295 if (IS_ERR_OR_NULL(ifp)) {
2296 in6_dev_put(in6_dev);
2302 spin_lock_bh(&ifp->lock);
2303 ifp->flags |= IFA_F_MANAGETEMPADDR;
2304 ifp->cstamp = jiffies;
2305 ifp->tokenized = tokenized;
2306 spin_unlock_bh(&ifp->lock);
2307 addrconf_dad_start(ifp);
2315 /* update lifetime (RFC2462 5.5.3 e) */
2316 spin_lock(&ifp->lock);
2318 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
2319 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
2322 if (!update_lft && !create && stored_lft) {
2323 const u32 minimum_lft = min(
2324 stored_lft, (u32)MIN_VALID_LIFETIME);
2325 valid_lft = max(valid_lft, minimum_lft);
2327 /* RFC4862 Section 5.5.3e:
2328 * "Note that the preferred lifetime of the
2329 * corresponding address is always reset to
2330 * the Preferred Lifetime in the received
2331 * Prefix Information option, regardless of
2332 * whether the valid lifetime is also reset or
2335 * So we should always update prefered_lft here.
2341 ifp->valid_lft = valid_lft;
2342 ifp->prefered_lft = prefered_lft;
2345 ifp->flags &= ~IFA_F_DEPRECATED;
2346 spin_unlock(&ifp->lock);
2348 if (!(flags&IFA_F_TENTATIVE))
2349 ipv6_ifa_notify(0, ifp);
2351 spin_unlock(&ifp->lock);
2353 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
2360 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2361 in6_dev_put(in6_dev);
2365 * Set destination address.
2366 * Special case for SIT interfaces where we create a new "virtual"
2369 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2371 struct in6_ifreq ireq;
2372 struct net_device *dev;
2378 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2381 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2387 #if IS_ENABLED(CONFIG_IPV6_SIT)
2388 if (dev->type == ARPHRD_SIT) {
2389 const struct net_device_ops *ops = dev->netdev_ops;
2391 struct ip_tunnel_parm p;
2393 err = -EADDRNOTAVAIL;
2394 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2397 memset(&p, 0, sizeof(p));
2398 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2402 p.iph.protocol = IPPROTO_IPV6;
2404 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2406 if (ops->ndo_do_ioctl) {
2407 mm_segment_t oldfs = get_fs();
2410 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2417 dev = __dev_get_by_name(net, p.name);
2420 err = dev_open(dev);
2431 * Manual configuration of address on an interface
2433 static int inet6_addr_add(struct net *net, int ifindex,
2434 const struct in6_addr *pfx,
2435 const struct in6_addr *peer_pfx,
2436 unsigned int plen, __u32 ifa_flags,
2437 __u32 prefered_lft, __u32 valid_lft)
2439 struct inet6_ifaddr *ifp;
2440 struct inet6_dev *idev;
2441 struct net_device *dev;
2445 unsigned long timeout;
2452 /* check the lifetime */
2453 if (!valid_lft || prefered_lft > valid_lft)
2456 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64)
2459 dev = __dev_get_by_index(net, ifindex);
2463 idev = addrconf_add_dev(dev);
2465 return PTR_ERR(idev);
2467 scope = ipv6_addr_scope(pfx);
2469 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2470 if (addrconf_finite_timeout(timeout)) {
2471 expires = jiffies_to_clock_t(timeout * HZ);
2472 valid_lft = timeout;
2473 flags = RTF_EXPIRES;
2477 ifa_flags |= IFA_F_PERMANENT;
2480 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2481 if (addrconf_finite_timeout(timeout)) {
2483 ifa_flags |= IFA_F_DEPRECATED;
2484 prefered_lft = timeout;
2487 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags,
2488 valid_lft, prefered_lft);
2491 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
2492 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2497 * Note that section 3.1 of RFC 4429 indicates
2498 * that the Optimistic flag should not be set for
2499 * manually configured addresses
2501 addrconf_dad_start(ifp);
2502 if (ifa_flags & IFA_F_MANAGETEMPADDR)
2503 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft,
2506 addrconf_verify_rtnl();
2510 return PTR_ERR(ifp);
2513 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
2514 const struct in6_addr *pfx, unsigned int plen)
2516 struct inet6_ifaddr *ifp;
2517 struct inet6_dev *idev;
2518 struct net_device *dev;
2523 dev = __dev_get_by_index(net, ifindex);
2527 if ((idev = __in6_dev_get(dev)) == NULL)
2530 read_lock_bh(&idev->lock);
2531 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2532 if (ifp->prefix_len == plen &&
2533 ipv6_addr_equal(pfx, &ifp->addr)) {
2535 read_unlock_bh(&idev->lock);
2537 if (!(ifp->flags & IFA_F_TEMPORARY) &&
2538 (ifa_flags & IFA_F_MANAGETEMPADDR))
2539 manage_tempaddrs(idev, ifp, 0, 0, false,
2542 addrconf_verify_rtnl();
2546 read_unlock_bh(&idev->lock);
2547 return -EADDRNOTAVAIL;
2551 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2553 struct in6_ifreq ireq;
2556 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2559 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2563 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL,
2564 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2565 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2570 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2572 struct in6_ifreq ireq;
2575 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2578 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2582 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
2583 ireq.ifr6_prefixlen);
2588 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2589 int plen, int scope)
2591 struct inet6_ifaddr *ifp;
2593 ifp = ipv6_add_addr(idev, addr, NULL, plen,
2594 scope, IFA_F_PERMANENT,
2595 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2597 spin_lock_bh(&ifp->lock);
2598 ifp->flags &= ~IFA_F_TENTATIVE;
2599 spin_unlock_bh(&ifp->lock);
2600 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2605 #if IS_ENABLED(CONFIG_IPV6_SIT)
2606 static void sit_add_v4_addrs(struct inet6_dev *idev)
2608 struct in6_addr addr;
2609 struct net_device *dev;
2610 struct net *net = dev_net(idev->dev);
2616 memset(&addr, 0, sizeof(struct in6_addr));
2617 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2619 if (idev->dev->flags&IFF_POINTOPOINT) {
2620 addr.s6_addr32[0] = htonl(0xfe800000);
2624 scope = IPV6_ADDR_COMPATv4;
2626 pflags |= RTF_NONEXTHOP;
2629 if (addr.s6_addr32[3]) {
2630 add_addr(idev, &addr, plen, scope);
2631 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags);
2635 for_each_netdev(net, dev) {
2636 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2637 if (in_dev && (dev->flags & IFF_UP)) {
2638 struct in_ifaddr *ifa;
2642 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2644 addr.s6_addr32[3] = ifa->ifa_local;
2646 if (ifa->ifa_scope == RT_SCOPE_LINK)
2648 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2649 if (idev->dev->flags&IFF_POINTOPOINT)
2654 add_addr(idev, &addr, plen, flag);
2655 addrconf_prefix_route(&addr, plen, idev->dev, 0,
2663 static void init_loopback(struct net_device *dev)
2665 struct inet6_dev *idev;
2666 struct net_device *sp_dev;
2667 struct inet6_ifaddr *sp_ifa;
2668 struct rt6_info *sp_rt;
2674 if ((idev = ipv6_find_idev(dev)) == NULL) {
2675 pr_debug("%s: add_dev failed\n", __func__);
2679 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2681 /* Add routes to other interface's IPv6 addresses */
2682 for_each_netdev(dev_net(dev), sp_dev) {
2683 if (!strcmp(sp_dev->name, dev->name))
2686 idev = __in6_dev_get(sp_dev);
2690 read_lock_bh(&idev->lock);
2691 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2693 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2697 /* This dst has been added to garbage list when
2698 * lo device down, release this obsolete dst and
2699 * reallocate a new router for ifa.
2701 if (sp_ifa->rt->dst.obsolete > 0) {
2702 ip6_rt_put(sp_ifa->rt);
2709 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false);
2711 /* Failure cases are ignored */
2712 if (!IS_ERR(sp_rt)) {
2717 read_unlock_bh(&idev->lock);
2721 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2723 struct inet6_ifaddr *ifp;
2724 u32 addr_flags = IFA_F_PERMANENT;
2726 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2727 if (idev->cnf.optimistic_dad &&
2728 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2729 addr_flags |= IFA_F_OPTIMISTIC;
2733 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags,
2734 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2736 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2737 addrconf_dad_start(ifp);
2742 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
2744 if (idev->addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) {
2745 struct in6_addr addr;
2747 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2748 /* addrconf_add_linklocal also adds a prefix_route and we
2749 * only need to care about prefix routes if ipv6_generate_eui64
2750 * couldn't generate one.
2752 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
2753 addrconf_add_linklocal(idev, &addr);
2754 else if (prefix_route)
2755 addrconf_prefix_route(&addr, 64, idev->dev, 0, 0);
2759 static void addrconf_dev_config(struct net_device *dev)
2761 struct inet6_dev *idev;
2765 if ((dev->type != ARPHRD_ETHER) &&
2766 (dev->type != ARPHRD_FDDI) &&
2767 (dev->type != ARPHRD_ARCNET) &&
2768 (dev->type != ARPHRD_INFINIBAND) &&
2769 (dev->type != ARPHRD_IEEE802154) &&
2770 (dev->type != ARPHRD_IEEE1394) &&
2771 (dev->type != ARPHRD_TUNNEL6) &&
2772 (dev->type != ARPHRD_6LOWPAN)) {
2773 /* Alas, we support only Ethernet autoconfiguration. */
2777 idev = addrconf_add_dev(dev);
2781 addrconf_addr_gen(idev, false);
2784 #if IS_ENABLED(CONFIG_IPV6_SIT)
2785 static void addrconf_sit_config(struct net_device *dev)
2787 struct inet6_dev *idev;
2792 * Configure the tunnel with one of our IPv4
2793 * addresses... we should configure all of
2794 * our v4 addrs in the tunnel
2797 if ((idev = ipv6_find_idev(dev)) == NULL) {
2798 pr_debug("%s: add_dev failed\n", __func__);
2802 if (dev->priv_flags & IFF_ISATAP) {
2803 addrconf_addr_gen(idev, false);
2807 sit_add_v4_addrs(idev);
2809 if (dev->flags&IFF_POINTOPOINT)
2810 addrconf_add_mroute(dev);
2814 #if IS_ENABLED(CONFIG_NET_IPGRE)
2815 static void addrconf_gre_config(struct net_device *dev)
2817 struct inet6_dev *idev;
2821 if ((idev = ipv6_find_idev(dev)) == NULL) {
2822 pr_debug("%s: add_dev failed\n", __func__);
2826 addrconf_addr_gen(idev, true);
2830 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2833 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2834 struct inet6_dev *idev = __in6_dev_get(dev);
2835 int run_pending = 0;
2839 case NETDEV_REGISTER:
2840 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2841 idev = ipv6_add_dev(dev);
2843 return notifier_from_errno(PTR_ERR(idev));
2849 if (dev->flags & IFF_SLAVE)
2852 if (event == NETDEV_UP) {
2853 if (!addrconf_qdisc_ok(dev)) {
2854 /* device is not ready yet. */
2855 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2860 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2861 idev = ipv6_add_dev(dev);
2863 if (!IS_ERR_OR_NULL(idev)) {
2864 idev->if_flags |= IF_READY;
2868 if (!addrconf_qdisc_ok(dev)) {
2869 /* device is still not ready. */
2874 if (idev->if_flags & IF_READY)
2875 /* device is already configured. */
2877 idev->if_flags |= IF_READY;
2880 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2886 switch (dev->type) {
2887 #if IS_ENABLED(CONFIG_IPV6_SIT)
2889 addrconf_sit_config(dev);
2892 #if IS_ENABLED(CONFIG_NET_IPGRE)
2894 addrconf_gre_config(dev);
2897 case ARPHRD_LOOPBACK:
2902 addrconf_dev_config(dev);
2906 if (!IS_ERR_OR_NULL(idev)) {
2908 addrconf_dad_run(idev);
2911 * If the MTU changed during the interface down,
2912 * when the interface up, the changed MTU must be
2913 * reflected in the idev as well as routers.
2915 if (idev->cnf.mtu6 != dev->mtu &&
2916 dev->mtu >= IPV6_MIN_MTU) {
2917 rt6_mtu_change(dev, dev->mtu);
2918 idev->cnf.mtu6 = dev->mtu;
2920 idev->tstamp = jiffies;
2921 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2924 * If the changed mtu during down is lower than
2925 * IPV6_MIN_MTU stop IPv6 on this interface.
2927 if (dev->mtu < IPV6_MIN_MTU)
2928 addrconf_ifdown(dev, 1);
2932 case NETDEV_CHANGEMTU:
2933 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2934 rt6_mtu_change(dev, dev->mtu);
2935 idev->cnf.mtu6 = dev->mtu;
2939 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2940 idev = ipv6_add_dev(dev);
2946 * if MTU under IPV6_MIN_MTU.
2947 * Stop IPv6 on this interface.
2951 case NETDEV_UNREGISTER:
2953 * Remove all addresses from this interface.
2955 addrconf_ifdown(dev, event != NETDEV_DOWN);
2958 case NETDEV_CHANGENAME:
2960 snmp6_unregister_dev(idev);
2961 addrconf_sysctl_unregister(idev);
2962 err = addrconf_sysctl_register(idev);
2964 return notifier_from_errno(err);
2965 err = snmp6_register_dev(idev);
2967 addrconf_sysctl_unregister(idev);
2968 return notifier_from_errno(err);
2973 case NETDEV_PRE_TYPE_CHANGE:
2974 case NETDEV_POST_TYPE_CHANGE:
2975 addrconf_type_change(dev, event);
2983 * addrconf module should be notified of a device going up
2985 static struct notifier_block ipv6_dev_notf = {
2986 .notifier_call = addrconf_notify,
2989 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2991 struct inet6_dev *idev;
2994 idev = __in6_dev_get(dev);
2996 if (event == NETDEV_POST_TYPE_CHANGE)
2997 ipv6_mc_remap(idev);
2998 else if (event == NETDEV_PRE_TYPE_CHANGE)
2999 ipv6_mc_unmap(idev);
3002 static int addrconf_ifdown(struct net_device *dev, int how)
3004 struct net *net = dev_net(dev);
3005 struct inet6_dev *idev;
3006 struct inet6_ifaddr *ifa;
3011 rt6_ifdown(net, dev);
3012 neigh_ifdown(&nd_tbl, dev);
3014 idev = __in6_dev_get(dev);
3019 * Step 1: remove reference to ipv6 device from parent device.
3025 /* protected by rtnl_lock */
3026 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
3028 /* Step 1.5: remove snmp6 entry */
3029 snmp6_unregister_dev(idev);
3033 /* Step 2: clear hash table */
3034 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3035 struct hlist_head *h = &inet6_addr_lst[i];
3037 spin_lock_bh(&addrconf_hash_lock);
3039 hlist_for_each_entry_rcu(ifa, h, addr_lst) {
3040 if (ifa->idev == idev) {
3041 hlist_del_init_rcu(&ifa->addr_lst);
3042 addrconf_del_dad_work(ifa);
3046 spin_unlock_bh(&addrconf_hash_lock);
3049 write_lock_bh(&idev->lock);
3051 addrconf_del_rs_timer(idev);
3053 /* Step 2: clear flags for stateless addrconf */
3055 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
3057 if (how && del_timer(&idev->regen_timer))
3060 /* Step 3: clear tempaddr list */
3061 while (!list_empty(&idev->tempaddr_list)) {
3062 ifa = list_first_entry(&idev->tempaddr_list,
3063 struct inet6_ifaddr, tmp_list);
3064 list_del(&ifa->tmp_list);
3065 write_unlock_bh(&idev->lock);
3066 spin_lock_bh(&ifa->lock);
3069 in6_ifa_put(ifa->ifpub);
3072 spin_unlock_bh(&ifa->lock);
3074 write_lock_bh(&idev->lock);
3077 while (!list_empty(&idev->addr_list)) {
3078 ifa = list_first_entry(&idev->addr_list,
3079 struct inet6_ifaddr, if_list);
3080 addrconf_del_dad_work(ifa);
3082 list_del(&ifa->if_list);
3084 write_unlock_bh(&idev->lock);
3086 spin_lock_bh(&ifa->state_lock);
3088 ifa->state = INET6_IFADDR_STATE_DEAD;
3089 spin_unlock_bh(&ifa->state_lock);
3091 if (state != INET6_IFADDR_STATE_DEAD) {
3092 __ipv6_ifa_notify(RTM_DELADDR, ifa);
3093 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
3097 write_lock_bh(&idev->lock);
3100 write_unlock_bh(&idev->lock);
3102 /* Step 5: Discard multicast list */
3104 ipv6_mc_destroy_dev(idev);
3108 idev->tstamp = jiffies;
3110 /* Last: Shot the device (if unregistered) */
3112 addrconf_sysctl_unregister(idev);
3113 neigh_parms_release(&nd_tbl, idev->nd_parms);
3114 neigh_ifdown(&nd_tbl, dev);
3120 static void addrconf_rs_timer(unsigned long data)
3122 struct inet6_dev *idev = (struct inet6_dev *)data;
3123 struct net_device *dev = idev->dev;
3124 struct in6_addr lladdr;
3126 write_lock(&idev->lock);
3127 if (idev->dead || !(idev->if_flags & IF_READY))
3130 if (!ipv6_accept_ra(idev))
3133 /* Announcement received after solicitation was sent */
3134 if (idev->if_flags & IF_RA_RCVD)
3137 if (idev->rs_probes++ < idev->cnf.rtr_solicits) {
3138 write_unlock(&idev->lock);
3139 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3140 ndisc_send_rs(dev, &lladdr,
3141 &in6addr_linklocal_allrouters);
3145 write_lock(&idev->lock);
3146 /* The wait after the last probe can be shorter */
3147 addrconf_mod_rs_timer(idev, (idev->rs_probes ==
3148 idev->cnf.rtr_solicits) ?
3149 idev->cnf.rtr_solicit_delay :
3150 idev->cnf.rtr_solicit_interval);
3153 * Note: we do not support deprecated "all on-link"
3154 * assumption any longer.
3156 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
3160 write_unlock(&idev->lock);
3166 * Duplicate Address Detection
3168 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
3170 unsigned long rand_num;
3171 struct inet6_dev *idev = ifp->idev;
3173 if (ifp->flags & IFA_F_OPTIMISTIC)
3176 rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
3178 ifp->dad_probes = idev->cnf.dad_transmits;
3179 addrconf_mod_dad_work(ifp, rand_num);
3182 static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
3184 struct inet6_dev *idev = ifp->idev;
3185 struct net_device *dev = idev->dev;
3187 addrconf_join_solict(dev, &ifp->addr);
3189 prandom_seed((__force u32) ifp->addr.s6_addr32[3]);
3191 read_lock_bh(&idev->lock);
3192 spin_lock(&ifp->lock);
3193 if (ifp->state == INET6_IFADDR_STATE_DEAD)
3196 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
3197 idev->cnf.accept_dad < 1 ||
3198 !(ifp->flags&IFA_F_TENTATIVE) ||
3199 ifp->flags & IFA_F_NODAD) {
3200 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3201 spin_unlock(&ifp->lock);
3202 read_unlock_bh(&idev->lock);
3204 addrconf_dad_completed(ifp);
3208 if (!(idev->if_flags & IF_READY)) {
3209 spin_unlock(&ifp->lock);
3210 read_unlock_bh(&idev->lock);
3212 * If the device is not ready:
3213 * - keep it tentative if it is a permanent address.
3214 * - otherwise, kill it.
3217 addrconf_dad_stop(ifp, 0);
3222 * Optimistic nodes can start receiving
3225 if (ifp->flags & IFA_F_OPTIMISTIC)
3226 ip6_ins_rt(ifp->rt);
3228 addrconf_dad_kick(ifp);
3230 spin_unlock(&ifp->lock);
3231 read_unlock_bh(&idev->lock);
3234 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
3236 bool begin_dad = false;
3238 spin_lock_bh(&ifp->state_lock);
3239 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
3240 ifp->state = INET6_IFADDR_STATE_PREDAD;
3243 spin_unlock_bh(&ifp->state_lock);
3246 addrconf_mod_dad_work(ifp, 0);
3249 static void addrconf_dad_work(struct work_struct *w)
3251 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
3252 struct inet6_ifaddr,
3254 struct inet6_dev *idev = ifp->idev;
3255 struct in6_addr mcaddr;
3261 } action = DAD_PROCESS;
3265 spin_lock_bh(&ifp->state_lock);
3266 if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
3268 ifp->state = INET6_IFADDR_STATE_DAD;
3269 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
3271 ifp->state = INET6_IFADDR_STATE_POSTDAD;
3273 spin_unlock_bh(&ifp->state_lock);
3275 if (action == DAD_BEGIN) {
3276 addrconf_dad_begin(ifp);
3278 } else if (action == DAD_ABORT) {
3279 addrconf_dad_stop(ifp, 1);
3283 if (!ifp->dad_probes && addrconf_dad_end(ifp))
3286 write_lock_bh(&idev->lock);
3287 if (idev->dead || !(idev->if_flags & IF_READY)) {
3288 write_unlock_bh(&idev->lock);
3292 spin_lock(&ifp->lock);
3293 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3294 spin_unlock(&ifp->lock);
3295 write_unlock_bh(&idev->lock);
3299 if (ifp->dad_probes == 0) {
3301 * DAD was successful
3304 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3305 spin_unlock(&ifp->lock);
3306 write_unlock_bh(&idev->lock);
3308 addrconf_dad_completed(ifp);
3314 addrconf_mod_dad_work(ifp,
3315 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME));
3316 spin_unlock(&ifp->lock);
3317 write_unlock_bh(&idev->lock);
3319 /* send a neighbour solicitation for our addr */
3320 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3321 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3327 /* ifp->idev must be at least read locked */
3328 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
3330 struct inet6_ifaddr *ifpiter;
3331 struct inet6_dev *idev = ifp->idev;
3333 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
3334 if (ifpiter->scope > IFA_LINK)
3336 if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
3337 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
3338 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
3345 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3347 struct net_device *dev = ifp->idev->dev;
3348 struct in6_addr lladdr;
3349 bool send_rs, send_mld;
3351 addrconf_del_dad_work(ifp);
3354 * Configure the address for reception. Now it is valid.
3357 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3359 /* If added prefix is link local and we are prepared to process
3360 router advertisements, start sending router solicitations.
3363 read_lock_bh(&ifp->idev->lock);
3364 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
3365 send_rs = send_mld &&
3366 ipv6_accept_ra(ifp->idev) &&
3367 ifp->idev->cnf.rtr_solicits > 0 &&
3368 (dev->flags&IFF_LOOPBACK) == 0;
3369 read_unlock_bh(&ifp->idev->lock);
3371 /* While dad is in progress mld report's source address is in6_addrany.
3372 * Resend with proper ll now.
3375 ipv6_mc_dad_complete(ifp->idev);
3379 * If a host as already performed a random delay
3380 * [...] as part of DAD [...] there is no need
3381 * to delay again before sending the first RS
3383 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
3385 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
3387 write_lock_bh(&ifp->idev->lock);
3388 spin_lock(&ifp->lock);
3389 ifp->idev->rs_probes = 1;
3390 ifp->idev->if_flags |= IF_RS_SENT;
3391 addrconf_mod_rs_timer(ifp->idev,
3392 ifp->idev->cnf.rtr_solicit_interval);
3393 spin_unlock(&ifp->lock);
3394 write_unlock_bh(&ifp->idev->lock);
3398 static void addrconf_dad_run(struct inet6_dev *idev)
3400 struct inet6_ifaddr *ifp;
3402 read_lock_bh(&idev->lock);
3403 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3404 spin_lock(&ifp->lock);
3405 if (ifp->flags & IFA_F_TENTATIVE &&
3406 ifp->state == INET6_IFADDR_STATE_DAD)
3407 addrconf_dad_kick(ifp);
3408 spin_unlock(&ifp->lock);
3410 read_unlock_bh(&idev->lock);
3413 #ifdef CONFIG_PROC_FS
3414 struct if6_iter_state {
3415 struct seq_net_private p;
3420 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3422 struct inet6_ifaddr *ifa = NULL;
3423 struct if6_iter_state *state = seq->private;
3424 struct net *net = seq_file_net(seq);
3427 /* initial bucket if pos is 0 */
3433 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3434 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket],
3436 if (!net_eq(dev_net(ifa->idev->dev), net))
3438 /* sync with offset */
3439 if (p < state->offset) {
3447 /* prepare for next bucket */
3454 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3455 struct inet6_ifaddr *ifa)
3457 struct if6_iter_state *state = seq->private;
3458 struct net *net = seq_file_net(seq);
3460 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) {
3461 if (!net_eq(dev_net(ifa->idev->dev), net))
3467 while (++state->bucket < IN6_ADDR_HSIZE) {
3469 hlist_for_each_entry_rcu_bh(ifa,
3470 &inet6_addr_lst[state->bucket], addr_lst) {
3471 if (!net_eq(dev_net(ifa->idev->dev), net))
3481 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3485 return if6_get_first(seq, *pos);
3488 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3490 struct inet6_ifaddr *ifa;
3492 ifa = if6_get_next(seq, v);
3497 static void if6_seq_stop(struct seq_file *seq, void *v)
3500 rcu_read_unlock_bh();
3503 static int if6_seq_show(struct seq_file *seq, void *v)
3505 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3506 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3508 ifp->idev->dev->ifindex,
3512 ifp->idev->dev->name);
3516 static const struct seq_operations if6_seq_ops = {
3517 .start = if6_seq_start,
3518 .next = if6_seq_next,
3519 .show = if6_seq_show,
3520 .stop = if6_seq_stop,
3523 static int if6_seq_open(struct inode *inode, struct file *file)
3525 return seq_open_net(inode, file, &if6_seq_ops,
3526 sizeof(struct if6_iter_state));
3529 static const struct file_operations if6_fops = {
3530 .owner = THIS_MODULE,
3531 .open = if6_seq_open,
3533 .llseek = seq_lseek,
3534 .release = seq_release_net,
3537 static int __net_init if6_proc_net_init(struct net *net)
3539 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops))
3544 static void __net_exit if6_proc_net_exit(struct net *net)
3546 remove_proc_entry("if_inet6", net->proc_net);
3549 static struct pernet_operations if6_proc_net_ops = {
3550 .init = if6_proc_net_init,
3551 .exit = if6_proc_net_exit,
3554 int __init if6_proc_init(void)
3556 return register_pernet_subsys(&if6_proc_net_ops);
3559 void if6_proc_exit(void)
3561 unregister_pernet_subsys(&if6_proc_net_ops);
3563 #endif /* CONFIG_PROC_FS */
3565 #if IS_ENABLED(CONFIG_IPV6_MIP6)
3566 /* Check if address is a home address configured on any interface. */
3567 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3570 struct inet6_ifaddr *ifp = NULL;
3571 unsigned int hash = inet6_addr_hash(addr);
3574 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) {
3575 if (!net_eq(dev_net(ifp->idev->dev), net))
3577 if (ipv6_addr_equal(&ifp->addr, addr) &&
3578 (ifp->flags & IFA_F_HOMEADDRESS)) {
3583 rcu_read_unlock_bh();
3589 * Periodic address status verification
3592 static void addrconf_verify_rtnl(void)
3594 unsigned long now, next, next_sec, next_sched;
3595 struct inet6_ifaddr *ifp;
3602 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3604 cancel_delayed_work(&addr_chk_work);
3606 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3608 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[i], addr_lst) {
3611 /* When setting preferred_lft to a value not zero or
3612 * infinity, while valid_lft is infinity
3613 * IFA_F_PERMANENT has a non-infinity life time.
3615 if ((ifp->flags & IFA_F_PERMANENT) &&
3616 (ifp->prefered_lft == INFINITY_LIFE_TIME))
3619 spin_lock(&ifp->lock);
3620 /* We try to batch several events at once. */
3621 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3623 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3624 age >= ifp->valid_lft) {
3625 spin_unlock(&ifp->lock);
3629 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3630 spin_unlock(&ifp->lock);
3632 } else if (age >= ifp->prefered_lft) {
3633 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3636 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3638 ifp->flags |= IFA_F_DEPRECATED;
3641 if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
3642 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
3643 next = ifp->tstamp + ifp->valid_lft * HZ;
3645 spin_unlock(&ifp->lock);
3650 ipv6_ifa_notify(0, ifp);
3654 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3655 !(ifp->flags&IFA_F_TENTATIVE)) {
3656 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3657 ifp->idev->cnf.dad_transmits *
3658 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ;
3660 if (age >= ifp->prefered_lft - regen_advance) {
3661 struct inet6_ifaddr *ifpub = ifp->ifpub;
3662 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3663 next = ifp->tstamp + ifp->prefered_lft * HZ;
3664 if (!ifp->regen_count && ifpub) {
3667 in6_ifa_hold(ifpub);
3668 spin_unlock(&ifp->lock);
3670 spin_lock(&ifpub->lock);
3671 ifpub->regen_count = 0;
3672 spin_unlock(&ifpub->lock);
3673 ipv6_create_tempaddr(ifpub, ifp);
3678 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3679 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3680 spin_unlock(&ifp->lock);
3682 /* ifp->prefered_lft <= ifp->valid_lft */
3683 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3684 next = ifp->tstamp + ifp->prefered_lft * HZ;
3685 spin_unlock(&ifp->lock);
3690 next_sec = round_jiffies_up(next);
3693 /* If rounded timeout is accurate enough, accept it. */
3694 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3695 next_sched = next_sec;
3697 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3698 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3699 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3701 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3702 now, next, next_sec, next_sched);
3703 mod_delayed_work(addrconf_wq, &addr_chk_work, next_sched - now);
3704 rcu_read_unlock_bh();
3707 static void addrconf_verify_work(struct work_struct *w)
3710 addrconf_verify_rtnl();
3714 static void addrconf_verify(void)
3716 mod_delayed_work(addrconf_wq, &addr_chk_work, 0);
3719 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
3720 struct in6_addr **peer_pfx)
3722 struct in6_addr *pfx = NULL;
3727 pfx = nla_data(addr);
3730 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3732 pfx = nla_data(local);
3738 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3739 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3740 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3741 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3742 [IFA_FLAGS] = { .len = sizeof(u32) },
3746 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
3748 struct net *net = sock_net(skb->sk);
3749 struct ifaddrmsg *ifm;
3750 struct nlattr *tb[IFA_MAX+1];
3751 struct in6_addr *pfx, *peer_pfx;
3755 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3759 ifm = nlmsg_data(nlh);
3760 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3764 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3766 /* We ignore other flags so far. */
3767 ifa_flags &= IFA_F_MANAGETEMPADDR;
3769 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
3770 ifm->ifa_prefixlen);
3773 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags,
3774 u32 prefered_lft, u32 valid_lft)
3778 unsigned long timeout;
3779 bool was_managetempaddr;
3780 bool had_prefixroute;
3784 if (!valid_lft || (prefered_lft > valid_lft))
3787 if (ifa_flags & IFA_F_MANAGETEMPADDR &&
3788 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
3791 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3792 if (addrconf_finite_timeout(timeout)) {
3793 expires = jiffies_to_clock_t(timeout * HZ);
3794 valid_lft = timeout;
3795 flags = RTF_EXPIRES;
3799 ifa_flags |= IFA_F_PERMANENT;
3802 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3803 if (addrconf_finite_timeout(timeout)) {
3805 ifa_flags |= IFA_F_DEPRECATED;
3806 prefered_lft = timeout;
3809 spin_lock_bh(&ifp->lock);
3810 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
3811 had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
3812 !(ifp->flags & IFA_F_NOPREFIXROUTE);
3813 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
3814 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3815 IFA_F_NOPREFIXROUTE);
3816 ifp->flags |= ifa_flags;
3817 ifp->tstamp = jiffies;
3818 ifp->valid_lft = valid_lft;
3819 ifp->prefered_lft = prefered_lft;
3821 spin_unlock_bh(&ifp->lock);
3822 if (!(ifp->flags&IFA_F_TENTATIVE))
3823 ipv6_ifa_notify(0, ifp);
3825 if (!(ifa_flags & IFA_F_NOPREFIXROUTE)) {
3826 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3828 } else if (had_prefixroute) {
3829 enum cleanup_prefix_rt_t action;
3830 unsigned long rt_expires;
3832 write_lock_bh(&ifp->idev->lock);
3833 action = check_cleanup_prefix_route(ifp, &rt_expires);
3834 write_unlock_bh(&ifp->idev->lock);
3836 if (action != CLEANUP_PREFIX_RT_NOP) {
3837 cleanup_prefix_route(ifp, rt_expires,
3838 action == CLEANUP_PREFIX_RT_DEL);
3842 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
3843 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR))
3844 valid_lft = prefered_lft = 0;
3845 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft,
3846 !was_managetempaddr, jiffies);
3849 addrconf_verify_rtnl();
3855 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
3857 struct net *net = sock_net(skb->sk);
3858 struct ifaddrmsg *ifm;
3859 struct nlattr *tb[IFA_MAX+1];
3860 struct in6_addr *pfx, *peer_pfx;
3861 struct inet6_ifaddr *ifa;
3862 struct net_device *dev;
3863 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3867 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3871 ifm = nlmsg_data(nlh);
3872 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
3876 if (tb[IFA_CACHEINFO]) {
3877 struct ifa_cacheinfo *ci;
3879 ci = nla_data(tb[IFA_CACHEINFO]);
3880 valid_lft = ci->ifa_valid;
3881 preferred_lft = ci->ifa_prefered;
3883 preferred_lft = INFINITY_LIFE_TIME;
3884 valid_lft = INFINITY_LIFE_TIME;
3887 dev = __dev_get_by_index(net, ifm->ifa_index);
3891 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
3893 /* We ignore other flags so far. */
3894 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
3895 IFA_F_NOPREFIXROUTE;
3897 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3900 * It would be best to check for !NLM_F_CREATE here but
3901 * userspace already relies on not having to provide this.
3903 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx,
3904 ifm->ifa_prefixlen, ifa_flags,
3905 preferred_lft, valid_lft);
3908 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3909 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3912 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3919 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
3920 u8 scope, int ifindex)
3922 struct ifaddrmsg *ifm;
3924 ifm = nlmsg_data(nlh);
3925 ifm->ifa_family = AF_INET6;
3926 ifm->ifa_prefixlen = prefixlen;
3927 ifm->ifa_flags = flags;
3928 ifm->ifa_scope = scope;
3929 ifm->ifa_index = ifindex;
3932 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3933 unsigned long tstamp, u32 preferred, u32 valid)
3935 struct ifa_cacheinfo ci;
3937 ci.cstamp = cstamp_delta(cstamp);
3938 ci.tstamp = cstamp_delta(tstamp);
3939 ci.ifa_prefered = preferred;
3940 ci.ifa_valid = valid;
3942 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3945 static inline int rt_scope(int ifa_scope)
3947 if (ifa_scope & IFA_HOST)
3948 return RT_SCOPE_HOST;
3949 else if (ifa_scope & IFA_LINK)
3950 return RT_SCOPE_LINK;
3951 else if (ifa_scope & IFA_SITE)
3952 return RT_SCOPE_SITE;
3954 return RT_SCOPE_UNIVERSE;
3957 static inline int inet6_ifaddr_msgsize(void)
3959 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3960 + nla_total_size(16) /* IFA_LOCAL */
3961 + nla_total_size(16) /* IFA_ADDRESS */
3962 + nla_total_size(sizeof(struct ifa_cacheinfo))
3963 + nla_total_size(4) /* IFA_FLAGS */;
3966 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3967 u32 portid, u32 seq, int event, unsigned int flags)
3969 struct nlmsghdr *nlh;
3970 u32 preferred, valid;
3972 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
3976 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3977 ifa->idev->dev->ifindex);
3979 if (!((ifa->flags&IFA_F_PERMANENT) &&
3980 (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
3981 preferred = ifa->prefered_lft;
3982 valid = ifa->valid_lft;
3983 if (preferred != INFINITY_LIFE_TIME) {
3984 long tval = (jiffies - ifa->tstamp)/HZ;
3985 if (preferred > tval)
3989 if (valid != INFINITY_LIFE_TIME) {
3997 preferred = INFINITY_LIFE_TIME;
3998 valid = INFINITY_LIFE_TIME;
4001 if (!ipv6_addr_any(&ifa->peer_addr)) {
4002 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 ||
4003 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0)
4006 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0)
4009 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
4012 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
4015 return nlmsg_end(skb, nlh);
4018 nlmsg_cancel(skb, nlh);
4022 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
4023 u32 portid, u32 seq, int event, u16 flags)
4025 struct nlmsghdr *nlh;
4026 u8 scope = RT_SCOPE_UNIVERSE;
4027 int ifindex = ifmca->idev->dev->ifindex;
4029 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
4030 scope = RT_SCOPE_SITE;
4032 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4036 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4037 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
4038 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
4039 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4040 nlmsg_cancel(skb, nlh);
4044 return nlmsg_end(skb, nlh);
4047 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
4048 u32 portid, u32 seq, int event, unsigned int flags)
4050 struct nlmsghdr *nlh;
4051 u8 scope = RT_SCOPE_UNIVERSE;
4052 int ifindex = ifaca->aca_idev->dev->ifindex;
4054 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
4055 scope = RT_SCOPE_SITE;
4057 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags);
4061 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
4062 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
4063 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
4064 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
4065 nlmsg_cancel(skb, nlh);
4069 return nlmsg_end(skb, nlh);
4078 /* called with rcu_read_lock() */
4079 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
4080 struct netlink_callback *cb, enum addr_type_t type,
4081 int s_ip_idx, int *p_ip_idx)
4083 struct ifmcaddr6 *ifmca;
4084 struct ifacaddr6 *ifaca;
4086 int ip_idx = *p_ip_idx;
4088 read_lock_bh(&idev->lock);
4090 case UNICAST_ADDR: {
4091 struct inet6_ifaddr *ifa;
4093 /* unicast address incl. temp addr */
4094 list_for_each_entry(ifa, &idev->addr_list, if_list) {
4095 if (++ip_idx < s_ip_idx)
4097 err = inet6_fill_ifaddr(skb, ifa,
4098 NETLINK_CB(cb->skb).portid,
4104 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4108 case MULTICAST_ADDR:
4109 /* multicast address */
4110 for (ifmca = idev->mc_list; ifmca;
4111 ifmca = ifmca->next, ip_idx++) {
4112 if (ip_idx < s_ip_idx)
4114 err = inet6_fill_ifmcaddr(skb, ifmca,
4115 NETLINK_CB(cb->skb).portid,
4124 /* anycast address */
4125 for (ifaca = idev->ac_list; ifaca;
4126 ifaca = ifaca->aca_next, ip_idx++) {
4127 if (ip_idx < s_ip_idx)
4129 err = inet6_fill_ifacaddr(skb, ifaca,
4130 NETLINK_CB(cb->skb).portid,
4141 read_unlock_bh(&idev->lock);
4146 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
4147 enum addr_type_t type)
4149 struct net *net = sock_net(skb->sk);
4152 int s_idx, s_ip_idx;
4153 struct net_device *dev;
4154 struct inet6_dev *idev;
4155 struct hlist_head *head;
4158 s_idx = idx = cb->args[1];
4159 s_ip_idx = ip_idx = cb->args[2];
4162 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq;
4163 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4165 head = &net->dev_index_head[h];
4166 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4169 if (h > s_h || idx > s_idx)
4172 idev = __in6_dev_get(dev);
4176 if (in6_dump_addrs(idev, skb, cb, type,
4177 s_ip_idx, &ip_idx) <= 0)
4187 cb->args[2] = ip_idx;
4192 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
4194 enum addr_type_t type = UNICAST_ADDR;
4196 return inet6_dump_addr(skb, cb, type);
4199 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
4201 enum addr_type_t type = MULTICAST_ADDR;
4203 return inet6_dump_addr(skb, cb, type);
4207 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
4209 enum addr_type_t type = ANYCAST_ADDR;
4211 return inet6_dump_addr(skb, cb, type);
4214 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh)
4216 struct net *net = sock_net(in_skb->sk);
4217 struct ifaddrmsg *ifm;
4218 struct nlattr *tb[IFA_MAX+1];
4219 struct in6_addr *addr = NULL, *peer;
4220 struct net_device *dev = NULL;
4221 struct inet6_ifaddr *ifa;
4222 struct sk_buff *skb;
4225 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
4229 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
4235 ifm = nlmsg_data(nlh);
4237 dev = __dev_get_by_index(net, ifm->ifa_index);
4239 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
4241 err = -EADDRNOTAVAIL;
4245 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
4251 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid,
4252 nlh->nlmsg_seq, RTM_NEWADDR, 0);
4254 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4255 WARN_ON(err == -EMSGSIZE);
4259 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4266 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
4268 struct sk_buff *skb;
4269 struct net *net = dev_net(ifa->idev->dev);
4272 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
4276 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
4278 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
4279 WARN_ON(err == -EMSGSIZE);
4283 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4287 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4290 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
4291 __s32 *array, int bytes)
4293 BUG_ON(bytes < (DEVCONF_MAX * 4));
4295 memset(array, 0, bytes);
4296 array[DEVCONF_FORWARDING] = cnf->forwarding;
4297 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
4298 array[DEVCONF_MTU6] = cnf->mtu6;
4299 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
4300 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
4301 array[DEVCONF_AUTOCONF] = cnf->autoconf;
4302 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
4303 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
4304 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
4305 jiffies_to_msecs(cnf->rtr_solicit_interval);
4306 array[DEVCONF_RTR_SOLICIT_DELAY] =
4307 jiffies_to_msecs(cnf->rtr_solicit_delay);
4308 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
4309 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
4310 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
4311 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
4312 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
4313 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
4314 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
4315 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
4316 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
4317 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
4318 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
4319 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
4320 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
4321 #ifdef CONFIG_IPV6_ROUTER_PREF
4322 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
4323 array[DEVCONF_RTR_PROBE_INTERVAL] =
4324 jiffies_to_msecs(cnf->rtr_probe_interval);
4325 #ifdef CONFIG_IPV6_ROUTE_INFO
4326 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
4329 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
4330 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
4331 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4332 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
4334 #ifdef CONFIG_IPV6_MROUTE
4335 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
4337 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
4338 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
4339 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
4340 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
4341 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
4342 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
4345 static inline size_t inet6_ifla6_size(void)
4347 return nla_total_size(4) /* IFLA_INET6_FLAGS */
4348 + nla_total_size(sizeof(struct ifla_cacheinfo))
4349 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
4350 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
4351 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
4352 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */
4355 static inline size_t inet6_if_nlmsg_size(void)
4357 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4358 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4359 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4360 + nla_total_size(4) /* IFLA_MTU */
4361 + nla_total_size(4) /* IFLA_LINK */
4362 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
4365 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
4366 int items, int bytes)
4369 int pad = bytes - sizeof(u64) * items;
4372 /* Use put_unaligned() because stats may not be aligned for u64. */
4373 put_unaligned(items, &stats[0]);
4374 for (i = 1; i < items; i++)
4375 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4377 memset(&stats[items], 0, pad);
4380 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
4381 int items, int bytes, size_t syncpoff)
4384 int pad = bytes - sizeof(u64) * items;
4387 /* Use put_unaligned() because stats may not be aligned for u64. */
4388 put_unaligned(items, &stats[0]);
4389 for (i = 1; i < items; i++)
4390 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4392 memset(&stats[items], 0, pad);
4395 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4399 case IFLA_INET6_STATS:
4400 __snmp6_fill_stats64(stats, idev->stats.ipv6,
4401 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4403 case IFLA_INET6_ICMP6STATS:
4404 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4409 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4412 struct ifla_cacheinfo ci;
4414 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
4415 goto nla_put_failure;
4416 ci.max_reasm_len = IPV6_MAXPLEN;
4417 ci.tstamp = cstamp_delta(idev->tstamp);
4418 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4419 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
4420 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
4421 goto nla_put_failure;
4422 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4424 goto nla_put_failure;
4425 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4427 /* XXX - MC not implemented */
4429 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4431 goto nla_put_failure;
4432 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4434 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4436 goto nla_put_failure;
4437 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4439 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
4441 goto nla_put_failure;
4443 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->addr_gen_mode))
4444 goto nla_put_failure;
4446 read_lock_bh(&idev->lock);
4447 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
4448 read_unlock_bh(&idev->lock);
4456 static size_t inet6_get_link_af_size(const struct net_device *dev)
4458 if (!__in6_dev_get(dev))
4461 return inet6_ifla6_size();
4464 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4466 struct inet6_dev *idev = __in6_dev_get(dev);
4471 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4477 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token)
4479 struct inet6_ifaddr *ifp;
4480 struct net_device *dev = idev->dev;
4481 bool update_rs = false;
4482 struct in6_addr ll_addr;
4488 if (ipv6_addr_any(token))
4490 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP))
4492 if (!ipv6_accept_ra(idev))
4494 if (idev->cnf.rtr_solicits <= 0)
4497 write_lock_bh(&idev->lock);
4499 BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
4500 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
4502 write_unlock_bh(&idev->lock);
4504 if (!idev->dead && (idev->if_flags & IF_READY) &&
4505 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
4506 IFA_F_OPTIMISTIC)) {
4508 /* If we're not ready, then normal ifup will take care
4509 * of this. Otherwise, we need to request our rs here.
4511 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
4515 write_lock_bh(&idev->lock);
4518 idev->if_flags |= IF_RS_SENT;
4519 idev->rs_probes = 1;
4520 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval);
4523 /* Well, that's kinda nasty ... */
4524 list_for_each_entry(ifp, &idev->addr_list, if_list) {
4525 spin_lock(&ifp->lock);
4526 if (ifp->tokenized) {
4528 ifp->prefered_lft = 0;
4530 spin_unlock(&ifp->lock);
4533 write_unlock_bh(&idev->lock);
4534 addrconf_verify_rtnl();
4538 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla)
4541 struct inet6_dev *idev = __in6_dev_get(dev);
4542 struct nlattr *tb[IFLA_INET6_MAX + 1];
4545 return -EAFNOSUPPORT;
4547 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0)
4550 if (tb[IFLA_INET6_TOKEN]) {
4551 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]));
4556 if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
4557 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
4559 if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
4560 mode != IN6_ADDR_GEN_MODE_NONE)
4562 idev->addr_gen_mode = mode;
4569 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4570 u32 portid, u32 seq, int event, unsigned int flags)
4572 struct net_device *dev = idev->dev;
4573 struct ifinfomsg *hdr;
4574 struct nlmsghdr *nlh;
4577 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
4581 hdr = nlmsg_data(nlh);
4582 hdr->ifi_family = AF_INET6;
4584 hdr->ifi_type = dev->type;
4585 hdr->ifi_index = dev->ifindex;
4586 hdr->ifi_flags = dev_get_flags(dev);
4587 hdr->ifi_change = 0;
4589 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4591 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4592 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4593 (dev->ifindex != dev->iflink &&
4594 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4595 goto nla_put_failure;
4596 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4597 if (protoinfo == NULL)
4598 goto nla_put_failure;
4600 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4601 goto nla_put_failure;
4603 nla_nest_end(skb, protoinfo);
4604 return nlmsg_end(skb, nlh);
4607 nlmsg_cancel(skb, nlh);
4611 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4613 struct net *net = sock_net(skb->sk);
4616 struct net_device *dev;
4617 struct inet6_dev *idev;
4618 struct hlist_head *head;
4621 s_idx = cb->args[1];
4624 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4626 head = &net->dev_index_head[h];
4627 hlist_for_each_entry_rcu(dev, head, index_hlist) {
4630 idev = __in6_dev_get(dev);
4633 if (inet6_fill_ifinfo(skb, idev,
4634 NETLINK_CB(cb->skb).portid,
4636 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4650 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4652 struct sk_buff *skb;
4653 struct net *net = dev_net(idev->dev);
4656 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4660 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4662 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4663 WARN_ON(err == -EMSGSIZE);
4667 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4671 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4674 static inline size_t inet6_prefix_nlmsg_size(void)
4676 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4677 + nla_total_size(sizeof(struct in6_addr))
4678 + nla_total_size(sizeof(struct prefix_cacheinfo));
4681 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4682 struct prefix_info *pinfo, u32 portid, u32 seq,
4683 int event, unsigned int flags)
4685 struct prefixmsg *pmsg;
4686 struct nlmsghdr *nlh;
4687 struct prefix_cacheinfo ci;
4689 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
4693 pmsg = nlmsg_data(nlh);
4694 pmsg->prefix_family = AF_INET6;
4695 pmsg->prefix_pad1 = 0;
4696 pmsg->prefix_pad2 = 0;
4697 pmsg->prefix_ifindex = idev->dev->ifindex;
4698 pmsg->prefix_len = pinfo->prefix_len;
4699 pmsg->prefix_type = pinfo->type;
4700 pmsg->prefix_pad3 = 0;
4701 pmsg->prefix_flags = 0;
4703 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4704 if (pinfo->autoconf)
4705 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4707 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4708 goto nla_put_failure;
4709 ci.preferred_time = ntohl(pinfo->prefered);
4710 ci.valid_time = ntohl(pinfo->valid);
4711 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4712 goto nla_put_failure;
4713 return nlmsg_end(skb, nlh);
4716 nlmsg_cancel(skb, nlh);
4720 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4721 struct prefix_info *pinfo)
4723 struct sk_buff *skb;
4724 struct net *net = dev_net(idev->dev);
4727 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4731 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4733 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4734 WARN_ON(err == -EMSGSIZE);
4738 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4742 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4745 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4747 struct net *net = dev_net(ifp->idev->dev);
4752 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4757 * If the address was optimistic
4758 * we inserted the route at the start of
4759 * our DAD process, so we don't need
4762 if (!(ifp->rt->rt6i_node))
4763 ip6_ins_rt(ifp->rt);
4764 if (ifp->idev->cnf.forwarding)
4765 addrconf_join_anycast(ifp);
4766 if (!ipv6_addr_any(&ifp->peer_addr))
4767 addrconf_prefix_route(&ifp->peer_addr, 128,
4768 ifp->idev->dev, 0, 0);
4771 if (ifp->idev->cnf.forwarding)
4772 addrconf_leave_anycast(ifp);
4773 addrconf_leave_solict(ifp->idev, &ifp->addr);
4774 if (!ipv6_addr_any(&ifp->peer_addr)) {
4775 struct rt6_info *rt;
4776 struct net_device *dev = ifp->idev->dev;
4778 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL,
4786 dst_hold(&ifp->rt->dst);
4788 if (ip6_del_rt(ifp->rt))
4789 dst_free(&ifp->rt->dst);
4792 atomic_inc(&net->ipv6.dev_addr_genid);
4793 rt_genid_bump_ipv6(net);
4796 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4799 if (likely(ifp->idev->dead == 0))
4800 __ipv6_ifa_notify(event, ifp);
4801 rcu_read_unlock_bh();
4804 #ifdef CONFIG_SYSCTL
4807 int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
4808 void __user *buffer, size_t *lenp, loff_t *ppos)
4810 int *valp = ctl->data;
4813 struct ctl_table lctl;
4817 * ctl->data points to idev->cnf.forwarding, we should
4818 * not modify it until we get the rtnl lock.
4823 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4826 ret = addrconf_fixup_forwarding(ctl, valp, val);
4832 static void dev_disable_change(struct inet6_dev *idev)
4834 struct netdev_notifier_info info;
4836 if (!idev || !idev->dev)
4839 netdev_notifier_info_init(&info, idev->dev);
4840 if (idev->cnf.disable_ipv6)
4841 addrconf_notify(NULL, NETDEV_DOWN, &info);
4843 addrconf_notify(NULL, NETDEV_UP, &info);
4846 static void addrconf_disable_change(struct net *net, __s32 newf)
4848 struct net_device *dev;
4849 struct inet6_dev *idev;
4852 for_each_netdev_rcu(net, dev) {
4853 idev = __in6_dev_get(dev);
4855 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4856 idev->cnf.disable_ipv6 = newf;
4858 dev_disable_change(idev);
4864 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4869 if (!rtnl_trylock())
4870 return restart_syscall();
4872 net = (struct net *)table->extra2;
4876 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4881 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4882 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4883 addrconf_disable_change(net, newf);
4884 } else if ((!newf) ^ (!old))
4885 dev_disable_change((struct inet6_dev *)table->extra1);
4892 int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
4893 void __user *buffer, size_t *lenp, loff_t *ppos)
4895 int *valp = ctl->data;
4898 struct ctl_table lctl;
4902 * ctl->data points to idev->cnf.disable_ipv6, we should
4903 * not modify it until we get the rtnl lock.
4908 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4911 ret = addrconf_disable_ipv6(ctl, valp, val);
4918 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
4919 void __user *buffer, size_t *lenp, loff_t *ppos)
4921 int *valp = ctl->data;
4926 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4929 if (write && old != new) {
4930 struct net *net = ctl->extra2;
4932 if (!rtnl_trylock())
4933 return restart_syscall();
4935 if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
4936 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4937 NETCONFA_IFINDEX_DEFAULT,
4938 net->ipv6.devconf_dflt);
4939 else if (valp == &net->ipv6.devconf_all->proxy_ndp)
4940 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4941 NETCONFA_IFINDEX_ALL,
4942 net->ipv6.devconf_all);
4944 struct inet6_dev *idev = ctl->extra1;
4946 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
4957 static struct addrconf_sysctl_table
4959 struct ctl_table_header *sysctl_header;
4960 struct ctl_table addrconf_vars[DEVCONF_MAX+1];
4961 } addrconf_sysctl __read_mostly = {
4962 .sysctl_header = NULL,
4965 .procname = "forwarding",
4966 .data = &ipv6_devconf.forwarding,
4967 .maxlen = sizeof(int),
4969 .proc_handler = addrconf_sysctl_forward,
4972 .procname = "hop_limit",
4973 .data = &ipv6_devconf.hop_limit,
4974 .maxlen = sizeof(int),
4976 .proc_handler = proc_dointvec,
4980 .data = &ipv6_devconf.mtu6,
4981 .maxlen = sizeof(int),
4983 .proc_handler = proc_dointvec,
4986 .procname = "accept_ra",
4987 .data = &ipv6_devconf.accept_ra,
4988 .maxlen = sizeof(int),
4990 .proc_handler = proc_dointvec,
4993 .procname = "accept_redirects",
4994 .data = &ipv6_devconf.accept_redirects,
4995 .maxlen = sizeof(int),
4997 .proc_handler = proc_dointvec,
5000 .procname = "autoconf",
5001 .data = &ipv6_devconf.autoconf,
5002 .maxlen = sizeof(int),
5004 .proc_handler = proc_dointvec,
5007 .procname = "dad_transmits",
5008 .data = &ipv6_devconf.dad_transmits,
5009 .maxlen = sizeof(int),
5011 .proc_handler = proc_dointvec,
5014 .procname = "router_solicitations",
5015 .data = &ipv6_devconf.rtr_solicits,
5016 .maxlen = sizeof(int),
5018 .proc_handler = proc_dointvec,
5021 .procname = "router_solicitation_interval",
5022 .data = &ipv6_devconf.rtr_solicit_interval,
5023 .maxlen = sizeof(int),
5025 .proc_handler = proc_dointvec_jiffies,
5028 .procname = "router_solicitation_delay",
5029 .data = &ipv6_devconf.rtr_solicit_delay,
5030 .maxlen = sizeof(int),
5032 .proc_handler = proc_dointvec_jiffies,
5035 .procname = "force_mld_version",
5036 .data = &ipv6_devconf.force_mld_version,
5037 .maxlen = sizeof(int),
5039 .proc_handler = proc_dointvec,
5042 .procname = "mldv1_unsolicited_report_interval",
5044 &ipv6_devconf.mldv1_unsolicited_report_interval,
5045 .maxlen = sizeof(int),
5047 .proc_handler = proc_dointvec_ms_jiffies,
5050 .procname = "mldv2_unsolicited_report_interval",
5052 &ipv6_devconf.mldv2_unsolicited_report_interval,
5053 .maxlen = sizeof(int),
5055 .proc_handler = proc_dointvec_ms_jiffies,
5058 .procname = "use_tempaddr",
5059 .data = &ipv6_devconf.use_tempaddr,
5060 .maxlen = sizeof(int),
5062 .proc_handler = proc_dointvec,
5065 .procname = "temp_valid_lft",
5066 .data = &ipv6_devconf.temp_valid_lft,
5067 .maxlen = sizeof(int),
5069 .proc_handler = proc_dointvec,
5072 .procname = "temp_prefered_lft",
5073 .data = &ipv6_devconf.temp_prefered_lft,
5074 .maxlen = sizeof(int),
5076 .proc_handler = proc_dointvec,
5079 .procname = "regen_max_retry",
5080 .data = &ipv6_devconf.regen_max_retry,
5081 .maxlen = sizeof(int),
5083 .proc_handler = proc_dointvec,
5086 .procname = "max_desync_factor",
5087 .data = &ipv6_devconf.max_desync_factor,
5088 .maxlen = sizeof(int),
5090 .proc_handler = proc_dointvec,
5093 .procname = "max_addresses",
5094 .data = &ipv6_devconf.max_addresses,
5095 .maxlen = sizeof(int),
5097 .proc_handler = proc_dointvec,
5100 .procname = "accept_ra_defrtr",
5101 .data = &ipv6_devconf.accept_ra_defrtr,
5102 .maxlen = sizeof(int),
5104 .proc_handler = proc_dointvec,
5107 .procname = "accept_ra_pinfo",
5108 .data = &ipv6_devconf.accept_ra_pinfo,
5109 .maxlen = sizeof(int),
5111 .proc_handler = proc_dointvec,
5113 #ifdef CONFIG_IPV6_ROUTER_PREF
5115 .procname = "accept_ra_rtr_pref",
5116 .data = &ipv6_devconf.accept_ra_rtr_pref,
5117 .maxlen = sizeof(int),
5119 .proc_handler = proc_dointvec,
5122 .procname = "router_probe_interval",
5123 .data = &ipv6_devconf.rtr_probe_interval,
5124 .maxlen = sizeof(int),
5126 .proc_handler = proc_dointvec_jiffies,
5128 #ifdef CONFIG_IPV6_ROUTE_INFO
5130 .procname = "accept_ra_rt_info_max_plen",
5131 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
5132 .maxlen = sizeof(int),
5134 .proc_handler = proc_dointvec,
5139 .procname = "proxy_ndp",
5140 .data = &ipv6_devconf.proxy_ndp,
5141 .maxlen = sizeof(int),
5143 .proc_handler = addrconf_sysctl_proxy_ndp,
5146 .procname = "accept_source_route",
5147 .data = &ipv6_devconf.accept_source_route,
5148 .maxlen = sizeof(int),
5150 .proc_handler = proc_dointvec,
5152 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
5154 .procname = "optimistic_dad",
5155 .data = &ipv6_devconf.optimistic_dad,
5156 .maxlen = sizeof(int),
5158 .proc_handler = proc_dointvec,
5162 #ifdef CONFIG_IPV6_MROUTE
5164 .procname = "mc_forwarding",
5165 .data = &ipv6_devconf.mc_forwarding,
5166 .maxlen = sizeof(int),
5168 .proc_handler = proc_dointvec,
5172 .procname = "disable_ipv6",
5173 .data = &ipv6_devconf.disable_ipv6,
5174 .maxlen = sizeof(int),
5176 .proc_handler = addrconf_sysctl_disable,
5179 .procname = "accept_dad",
5180 .data = &ipv6_devconf.accept_dad,
5181 .maxlen = sizeof(int),
5183 .proc_handler = proc_dointvec,
5186 .procname = "force_tllao",
5187 .data = &ipv6_devconf.force_tllao,
5188 .maxlen = sizeof(int),
5190 .proc_handler = proc_dointvec
5193 .procname = "ndisc_notify",
5194 .data = &ipv6_devconf.ndisc_notify,
5195 .maxlen = sizeof(int),
5197 .proc_handler = proc_dointvec
5200 .procname = "suppress_frag_ndisc",
5201 .data = &ipv6_devconf.suppress_frag_ndisc,
5202 .maxlen = sizeof(int),
5204 .proc_handler = proc_dointvec
5207 .procname = "accept_ra_from_local",
5208 .data = &ipv6_devconf.accept_ra_from_local,
5209 .maxlen = sizeof(int),
5211 .proc_handler = proc_dointvec,
5219 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
5220 struct inet6_dev *idev, struct ipv6_devconf *p)
5223 struct addrconf_sysctl_table *t;
5224 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
5226 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
5230 for (i = 0; t->addrconf_vars[i].data; i++) {
5231 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
5232 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
5233 t->addrconf_vars[i].extra2 = net;
5236 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
5238 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
5239 if (t->sysctl_header == NULL)
5251 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
5253 struct addrconf_sysctl_table *t;
5255 if (p->sysctl == NULL)
5260 unregister_net_sysctl_table(t->sysctl_header);
5264 static int addrconf_sysctl_register(struct inet6_dev *idev)
5268 if (!sysctl_dev_name_is_allowed(idev->dev->name))
5271 err = neigh_sysctl_register(idev->dev, idev->nd_parms,
5272 &ndisc_ifinfo_sysctl_change);
5275 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
5278 neigh_sysctl_unregister(idev->nd_parms);
5283 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
5285 __addrconf_sysctl_unregister(&idev->cnf);
5286 neigh_sysctl_unregister(idev->nd_parms);
5292 static int __net_init addrconf_init_net(struct net *net)
5295 struct ipv6_devconf *all, *dflt;
5297 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
5301 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
5303 goto err_alloc_dflt;
5305 /* these will be inherited by all namespaces */
5306 dflt->autoconf = ipv6_defaults.autoconf;
5307 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
5309 net->ipv6.devconf_all = all;
5310 net->ipv6.devconf_dflt = dflt;
5312 #ifdef CONFIG_SYSCTL
5313 err = __addrconf_sysctl_register(net, "all", NULL, all);
5317 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
5323 #ifdef CONFIG_SYSCTL
5325 __addrconf_sysctl_unregister(all);
5335 static void __net_exit addrconf_exit_net(struct net *net)
5337 #ifdef CONFIG_SYSCTL
5338 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
5339 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
5341 if (!net_eq(net, &init_net)) {
5342 kfree(net->ipv6.devconf_dflt);
5343 kfree(net->ipv6.devconf_all);
5347 static struct pernet_operations addrconf_ops = {
5348 .init = addrconf_init_net,
5349 .exit = addrconf_exit_net,
5352 static struct rtnl_af_ops inet6_ops = {
5354 .fill_link_af = inet6_fill_link_af,
5355 .get_link_af_size = inet6_get_link_af_size,
5356 .set_link_af = inet6_set_link_af,
5360 * Init / cleanup code
5363 int __init addrconf_init(void)
5365 struct inet6_dev *idev;
5368 err = ipv6_addr_label_init();
5370 pr_crit("%s: cannot initialize default policy table: %d\n",
5375 err = register_pernet_subsys(&addrconf_ops);
5379 addrconf_wq = create_workqueue("ipv6_addrconf");
5385 /* The addrconf netdev notifier requires that loopback_dev
5386 * has it's ipv6 private information allocated and setup
5387 * before it can bring up and give link-local addresses
5388 * to other devices which are up.
5390 * Unfortunately, loopback_dev is not necessarily the first
5391 * entry in the global dev_base list of net devices. In fact,
5392 * it is likely to be the very last entry on that list.
5393 * So this causes the notifier registry below to try and
5394 * give link-local addresses to all devices besides loopback_dev
5395 * first, then loopback_dev, which cases all the non-loopback_dev
5396 * devices to fail to get a link-local address.
5398 * So, as a temporary fix, allocate the ipv6 structure for
5399 * loopback_dev first by hand.
5400 * Longer term, all of the dependencies ipv6 has upon the loopback
5401 * device and it being up should be removed.
5404 idev = ipv6_add_dev(init_net.loopback_dev);
5407 err = PTR_ERR(idev);
5411 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5412 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
5414 register_netdevice_notifier(&ipv6_dev_notf);
5418 rtnl_af_register(&inet6_ops);
5420 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
5425 /* Only the first call to __rtnl_register can fail */
5426 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
5427 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
5428 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
5429 inet6_dump_ifaddr, NULL);
5430 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
5431 inet6_dump_ifmcaddr, NULL);
5432 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
5433 inet6_dump_ifacaddr, NULL);
5434 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf,
5435 inet6_netconf_dump_devconf, NULL);
5437 ipv6_addr_label_rtnl_register();
5441 rtnl_af_unregister(&inet6_ops);
5442 unregister_netdevice_notifier(&ipv6_dev_notf);
5444 destroy_workqueue(addrconf_wq);
5446 unregister_pernet_subsys(&addrconf_ops);
5448 ipv6_addr_label_cleanup();
5453 void addrconf_cleanup(void)
5455 struct net_device *dev;
5458 unregister_netdevice_notifier(&ipv6_dev_notf);
5459 unregister_pernet_subsys(&addrconf_ops);
5460 ipv6_addr_label_cleanup();
5464 __rtnl_af_unregister(&inet6_ops);
5466 /* clean dev list */
5467 for_each_netdev(&init_net, dev) {
5468 if (__in6_dev_get(dev) == NULL)
5470 addrconf_ifdown(dev, 1);
5472 addrconf_ifdown(init_net.loopback_dev, 2);
5477 spin_lock_bh(&addrconf_hash_lock);
5478 for (i = 0; i < IN6_ADDR_HSIZE; i++)
5479 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
5480 spin_unlock_bh(&addrconf_hash_lock);
5481 cancel_delayed_work(&addr_chk_work);
5484 destroy_workqueue(addrconf_wq);