xen/netfront: add netconsole support.
[linux-2.6/btrfs-unstable.git] / net / ipv6 / addrconf.c
blobc02280a4d126980540daac44a8b541b92c0b16f0
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
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.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
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
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90 #include <linux/export.h>
92 /* Set to 3 to get tracing... */
93 #define ACONF_DEBUG 2
95 #if ACONF_DEBUG >= 3
96 #define ADBG(x) printk x
97 #else
98 #define ADBG(x)
99 #endif
101 #define INFINITY_LIFE_TIME 0xFFFFFFFF
103 static inline u32 cstamp_delta(unsigned long cstamp)
105 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
108 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
109 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
110 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
112 #ifdef CONFIG_SYSCTL
113 static void addrconf_sysctl_register(struct inet6_dev *idev);
114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 #else
116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
123 #endif
125 #ifdef CONFIG_IPV6_PRIVACY
126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
128 static void ipv6_regen_rndid(unsigned long data);
129 #endif
131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
132 static int ipv6_count_addresses(struct inet6_dev *idev);
135 * Configured unicast address hash table
137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
138 static DEFINE_SPINLOCK(addrconf_hash_lock);
140 static void addrconf_verify(unsigned long);
142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
143 static DEFINE_SPINLOCK(addrconf_verify_lock);
145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
148 static void addrconf_type_change(struct net_device *dev,
149 unsigned long event);
150 static int addrconf_ifdown(struct net_device *dev, int how);
152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
153 static void addrconf_dad_timer(unsigned long data);
154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
155 static void addrconf_dad_run(struct inet6_dev *idev);
156 static void addrconf_rs_timer(unsigned long data);
157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
161 struct prefix_info *pinfo);
162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
163 struct net_device *dev);
165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168 .forwarding = 0,
169 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
170 .mtu6 = IPV6_MIN_MTU,
171 .accept_ra = 1,
172 .accept_redirects = 1,
173 .autoconf = 1,
174 .force_mld_version = 0,
175 .dad_transmits = 1,
176 .rtr_solicits = MAX_RTR_SOLICITATIONS,
177 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
178 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
179 #ifdef CONFIG_IPV6_PRIVACY
180 .use_tempaddr = 0,
181 .temp_valid_lft = TEMP_VALID_LIFETIME,
182 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
183 .regen_max_retry = REGEN_MAX_RETRY,
184 .max_desync_factor = MAX_DESYNC_FACTOR,
185 #endif
186 .max_addresses = IPV6_MAX_ADDRESSES,
187 .accept_ra_defrtr = 1,
188 .accept_ra_pinfo = 1,
189 #ifdef CONFIG_IPV6_ROUTER_PREF
190 .accept_ra_rtr_pref = 1,
191 .rtr_probe_interval = 60 * HZ,
192 #ifdef CONFIG_IPV6_ROUTE_INFO
193 .accept_ra_rt_info_max_plen = 0,
194 #endif
195 #endif
196 .proxy_ndp = 0,
197 .accept_source_route = 0, /* we do not accept RH0 by default. */
198 .disable_ipv6 = 0,
199 .accept_dad = 1,
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 .forwarding = 0,
204 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
205 .mtu6 = IPV6_MIN_MTU,
206 .accept_ra = 1,
207 .accept_redirects = 1,
208 .autoconf = 1,
209 .dad_transmits = 1,
210 .rtr_solicits = MAX_RTR_SOLICITATIONS,
211 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
213 #ifdef CONFIG_IPV6_PRIVACY
214 .use_tempaddr = 0,
215 .temp_valid_lft = TEMP_VALID_LIFETIME,
216 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
217 .regen_max_retry = REGEN_MAX_RETRY,
218 .max_desync_factor = MAX_DESYNC_FACTOR,
219 #endif
220 .max_addresses = IPV6_MAX_ADDRESSES,
221 .accept_ra_defrtr = 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_max_plen = 0,
228 #endif
229 #endif
230 .proxy_ndp = 0,
231 .accept_source_route = 0, /* we do not accept RH0 by default. */
232 .disable_ipv6 = 0,
233 .accept_dad = 1,
236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
245 return !qdisc_tx_is_noop(dev);
248 /* Check if a route is valid prefix route */
249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
251 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
256 if (del_timer(&ifp->timer))
257 __in6_ifa_put(ifp);
260 enum addrconf_timer_t {
261 AC_NONE,
262 AC_DAD,
263 AC_RS,
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267 enum addrconf_timer_t what,
268 unsigned long when)
270 if (!del_timer(&ifp->timer))
271 in6_ifa_hold(ifp);
273 switch (what) {
274 case AC_DAD:
275 ifp->timer.function = addrconf_dad_timer;
276 break;
277 case AC_RS:
278 ifp->timer.function = addrconf_rs_timer;
279 break;
280 default:
281 break;
283 ifp->timer.expires = jiffies + when;
284 add_timer(&ifp->timer);
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
289 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290 sizeof(struct ipstats_mib),
291 __alignof__(struct ipstats_mib)) < 0)
292 goto err_ip;
293 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294 GFP_KERNEL);
295 if (!idev->stats.icmpv6dev)
296 goto err_icmp;
297 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298 GFP_KERNEL);
299 if (!idev->stats.icmpv6msgdev)
300 goto err_icmpmsg;
302 return 0;
304 err_icmpmsg:
305 kfree(idev->stats.icmpv6dev);
306 err_icmp:
307 snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 err_ip:
309 return -ENOMEM;
312 static void snmp6_free_dev(struct inet6_dev *idev)
314 kfree(idev->stats.icmpv6msgdev);
315 kfree(idev->stats.icmpv6dev);
316 snmp_mib_free((void __percpu **)idev->stats.ipv6);
319 /* Nobody refers to this device, we may destroy it. */
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
323 struct net_device *dev = idev->dev;
325 WARN_ON(!list_empty(&idev->addr_list));
326 WARN_ON(idev->mc_list != NULL);
328 #ifdef NET_REFCNT_DEBUG
329 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
330 #endif
331 dev_put(dev);
332 if (!idev->dead) {
333 pr_warning("Freeing alive inet6 device %p\n", idev);
334 return;
336 snmp6_free_dev(idev);
337 kfree_rcu(idev, rcu);
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 struct inet6_dev *ndev;
346 ASSERT_RTNL();
348 if (dev->mtu < IPV6_MIN_MTU)
349 return NULL;
351 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
353 if (ndev == NULL)
354 return NULL;
356 rwlock_init(&ndev->lock);
357 ndev->dev = dev;
358 INIT_LIST_HEAD(&ndev->addr_list);
360 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (ndev->nd_parms == NULL) {
365 kfree(ndev);
366 return NULL;
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
371 dev_hold(dev);
373 if (snmp6_alloc_dev(ndev) < 0) {
374 ADBG((KERN_WARNING
375 "%s(): cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name));
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 dev_put(dev);
379 kfree(ndev);
380 return NULL;
383 if (snmp6_register_dev(ndev) < 0) {
384 ADBG((KERN_WARNING
385 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name));
387 neigh_parms_release(&nd_tbl, ndev->nd_parms);
388 ndev->dead = 1;
389 in6_dev_finish_destroy(ndev);
390 return NULL;
393 /* One reference from device. We must do this before
394 * we invoke __ipv6_regen_rndid().
396 in6_dev_hold(ndev);
398 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399 ndev->cnf.accept_dad = -1;
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
403 printk(KERN_INFO
404 "%s: Disabled Multicast RS\n",
405 dev->name);
406 ndev->cnf.rtr_solicits = 0;
408 #endif
410 #ifdef CONFIG_IPV6_PRIVACY
411 INIT_LIST_HEAD(&ndev->tempaddr_list);
412 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 if ((dev->flags&IFF_LOOPBACK) ||
414 dev->type == ARPHRD_TUNNEL ||
415 dev->type == ARPHRD_TUNNEL6 ||
416 dev->type == ARPHRD_SIT ||
417 dev->type == ARPHRD_NONE) {
418 ndev->cnf.use_tempaddr = -1;
419 } else {
420 in6_dev_hold(ndev);
421 ipv6_regen_rndid((unsigned long) ndev);
423 #endif
425 if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 ndev->if_flags |= IF_READY;
428 ipv6_mc_init_dev(ndev);
429 ndev->tstamp = jiffies;
430 addrconf_sysctl_register(ndev);
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
434 /* Join all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
437 return ndev;
440 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
442 struct inet6_dev *idev;
444 ASSERT_RTNL();
446 idev = __in6_dev_get(dev);
447 if (!idev) {
448 idev = ipv6_add_dev(dev);
449 if (!idev)
450 return NULL;
453 if (dev->flags&IFF_UP)
454 ipv6_mc_up(idev);
455 return idev;
458 #ifdef CONFIG_SYSCTL
459 static void dev_forward_change(struct inet6_dev *idev)
461 struct net_device *dev;
462 struct inet6_ifaddr *ifa;
464 if (!idev)
465 return;
466 dev = idev->dev;
467 if (idev->cnf.forwarding)
468 dev_disable_lro(dev);
469 if (dev && (dev->flags & IFF_MULTICAST)) {
470 if (idev->cnf.forwarding)
471 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
472 else
473 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
476 list_for_each_entry(ifa, &idev->addr_list, if_list) {
477 if (ifa->flags&IFA_F_TENTATIVE)
478 continue;
479 if (idev->cnf.forwarding)
480 addrconf_join_anycast(ifa);
481 else
482 addrconf_leave_anycast(ifa);
487 static void addrconf_forward_change(struct net *net, __s32 newf)
489 struct net_device *dev;
490 struct inet6_dev *idev;
492 rcu_read_lock();
493 for_each_netdev_rcu(net, dev) {
494 idev = __in6_dev_get(dev);
495 if (idev) {
496 int changed = (!idev->cnf.forwarding) ^ (!newf);
497 idev->cnf.forwarding = newf;
498 if (changed)
499 dev_forward_change(idev);
502 rcu_read_unlock();
505 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
507 struct net *net;
508 int old;
510 if (!rtnl_trylock())
511 return restart_syscall();
513 net = (struct net *)table->extra2;
514 old = *p;
515 *p = newf;
517 if (p == &net->ipv6.devconf_dflt->forwarding) {
518 rtnl_unlock();
519 return 0;
522 if (p == &net->ipv6.devconf_all->forwarding) {
523 net->ipv6.devconf_dflt->forwarding = newf;
524 addrconf_forward_change(net, newf);
525 } else if ((!newf) ^ (!old))
526 dev_forward_change((struct inet6_dev *)table->extra1);
527 rtnl_unlock();
529 if (newf)
530 rt6_purge_dflt_routers(net);
531 return 1;
533 #endif
535 /* Nobody refers to this ifaddr, destroy it */
536 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
538 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
540 #ifdef NET_REFCNT_DEBUG
541 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
542 #endif
544 in6_dev_put(ifp->idev);
546 if (del_timer(&ifp->timer))
547 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
549 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
550 pr_warning("Freeing alive inet6 address %p\n", ifp);
551 return;
553 dst_release(&ifp->rt->dst);
555 kfree_rcu(ifp, rcu);
558 static void
559 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
561 struct list_head *p;
562 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
565 * Each device address list is sorted in order of scope -
566 * global before linklocal.
568 list_for_each(p, &idev->addr_list) {
569 struct inet6_ifaddr *ifa
570 = list_entry(p, struct inet6_ifaddr, if_list);
571 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
572 break;
575 list_add_tail(&ifp->if_list, p);
578 static u32 ipv6_addr_hash(const struct in6_addr *addr)
581 * We perform the hash function over the last 64 bits of the address
582 * This will include the IEEE address token on links that support it.
584 return jhash_2words((__force u32)addr->s6_addr32[2],
585 (__force u32)addr->s6_addr32[3], 0)
586 & (IN6_ADDR_HSIZE - 1);
589 /* On success it returns ifp with increased reference count */
591 static struct inet6_ifaddr *
592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
593 int scope, u32 flags)
595 struct inet6_ifaddr *ifa = NULL;
596 struct rt6_info *rt;
597 unsigned int hash;
598 int err = 0;
599 int addr_type = ipv6_addr_type(addr);
601 if (addr_type == IPV6_ADDR_ANY ||
602 addr_type & IPV6_ADDR_MULTICAST ||
603 (!(idev->dev->flags & IFF_LOOPBACK) &&
604 addr_type & IPV6_ADDR_LOOPBACK))
605 return ERR_PTR(-EADDRNOTAVAIL);
607 rcu_read_lock_bh();
608 if (idev->dead) {
609 err = -ENODEV; /*XXX*/
610 goto out2;
613 if (idev->cnf.disable_ipv6) {
614 err = -EACCES;
615 goto out2;
618 spin_lock(&addrconf_hash_lock);
620 /* Ignore adding duplicate addresses on an interface */
621 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
622 ADBG(("ipv6_add_addr: already assigned\n"));
623 err = -EEXIST;
624 goto out;
627 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
629 if (ifa == NULL) {
630 ADBG(("ipv6_add_addr: malloc failed\n"));
631 err = -ENOBUFS;
632 goto out;
635 rt = addrconf_dst_alloc(idev, addr, false);
636 if (IS_ERR(rt)) {
637 err = PTR_ERR(rt);
638 goto out;
641 ifa->addr = *addr;
643 spin_lock_init(&ifa->lock);
644 spin_lock_init(&ifa->state_lock);
645 init_timer(&ifa->timer);
646 INIT_HLIST_NODE(&ifa->addr_lst);
647 ifa->timer.data = (unsigned long) ifa;
648 ifa->scope = scope;
649 ifa->prefix_len = pfxlen;
650 ifa->flags = flags | IFA_F_TENTATIVE;
651 ifa->cstamp = ifa->tstamp = jiffies;
653 ifa->rt = rt;
655 ifa->idev = idev;
656 in6_dev_hold(idev);
657 /* For caller */
658 in6_ifa_hold(ifa);
660 /* Add to big hash table */
661 hash = ipv6_addr_hash(addr);
663 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
664 spin_unlock(&addrconf_hash_lock);
666 write_lock(&idev->lock);
667 /* Add to inet6_dev unicast addr list. */
668 ipv6_link_dev_addr(idev, ifa);
670 #ifdef CONFIG_IPV6_PRIVACY
671 if (ifa->flags&IFA_F_TEMPORARY) {
672 list_add(&ifa->tmp_list, &idev->tempaddr_list);
673 in6_ifa_hold(ifa);
675 #endif
677 in6_ifa_hold(ifa);
678 write_unlock(&idev->lock);
679 out2:
680 rcu_read_unlock_bh();
682 if (likely(err == 0))
683 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
684 else {
685 kfree(ifa);
686 ifa = ERR_PTR(err);
689 return ifa;
690 out:
691 spin_unlock(&addrconf_hash_lock);
692 goto out2;
695 /* This function wants to get referenced ifp and releases it before return */
697 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
699 struct inet6_ifaddr *ifa, *ifn;
700 struct inet6_dev *idev = ifp->idev;
701 int state;
702 int deleted = 0, onlink = 0;
703 unsigned long expires = jiffies;
705 spin_lock_bh(&ifp->state_lock);
706 state = ifp->state;
707 ifp->state = INET6_IFADDR_STATE_DEAD;
708 spin_unlock_bh(&ifp->state_lock);
710 if (state == INET6_IFADDR_STATE_DEAD)
711 goto out;
713 spin_lock_bh(&addrconf_hash_lock);
714 hlist_del_init_rcu(&ifp->addr_lst);
715 spin_unlock_bh(&addrconf_hash_lock);
717 write_lock_bh(&idev->lock);
718 #ifdef CONFIG_IPV6_PRIVACY
719 if (ifp->flags&IFA_F_TEMPORARY) {
720 list_del(&ifp->tmp_list);
721 if (ifp->ifpub) {
722 in6_ifa_put(ifp->ifpub);
723 ifp->ifpub = NULL;
725 __in6_ifa_put(ifp);
727 #endif
729 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
730 if (ifa == ifp) {
731 list_del_init(&ifp->if_list);
732 __in6_ifa_put(ifp);
734 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
735 break;
736 deleted = 1;
737 continue;
738 } else if (ifp->flags & IFA_F_PERMANENT) {
739 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
740 ifp->prefix_len)) {
741 if (ifa->flags & IFA_F_PERMANENT) {
742 onlink = 1;
743 if (deleted)
744 break;
745 } else {
746 unsigned long lifetime;
748 if (!onlink)
749 onlink = -1;
751 spin_lock(&ifa->lock);
753 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
755 * Note: Because this address is
756 * not permanent, lifetime <
757 * LONG_MAX / HZ here.
759 if (time_before(expires,
760 ifa->tstamp + lifetime * HZ))
761 expires = ifa->tstamp + lifetime * HZ;
762 spin_unlock(&ifa->lock);
767 write_unlock_bh(&idev->lock);
769 addrconf_del_timer(ifp);
771 ipv6_ifa_notify(RTM_DELADDR, ifp);
773 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
776 * Purge or update corresponding prefix
778 * 1) we don't purge prefix here if address was not permanent.
779 * prefix is managed by its own lifetime.
780 * 2) if there're no addresses, delete prefix.
781 * 3) if there're still other permanent address(es),
782 * corresponding prefix is still permanent.
783 * 4) otherwise, update prefix lifetime to the
784 * longest valid lifetime among the corresponding
785 * addresses on the device.
786 * Note: subsequent RA will update lifetime.
788 * --yoshfuji
790 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
791 struct in6_addr prefix;
792 struct rt6_info *rt;
793 struct net *net = dev_net(ifp->idev->dev);
794 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
795 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
797 if (rt && addrconf_is_prefix_route(rt)) {
798 if (onlink == 0) {
799 ip6_del_rt(rt);
800 rt = NULL;
801 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
802 rt->dst.expires = expires;
803 rt->rt6i_flags |= RTF_EXPIRES;
806 dst_release(&rt->dst);
809 /* clean up prefsrc entries */
810 rt6_remove_prefsrc(ifp);
811 out:
812 in6_ifa_put(ifp);
815 #ifdef CONFIG_IPV6_PRIVACY
816 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
818 struct inet6_dev *idev = ifp->idev;
819 struct in6_addr addr, *tmpaddr;
820 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
821 unsigned long regen_advance;
822 int tmp_plen;
823 int ret = 0;
824 int max_addresses;
825 u32 addr_flags;
826 unsigned long now = jiffies;
828 write_lock(&idev->lock);
829 if (ift) {
830 spin_lock_bh(&ift->lock);
831 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
832 spin_unlock_bh(&ift->lock);
833 tmpaddr = &addr;
834 } else {
835 tmpaddr = NULL;
837 retry:
838 in6_dev_hold(idev);
839 if (idev->cnf.use_tempaddr <= 0) {
840 write_unlock(&idev->lock);
841 printk(KERN_INFO
842 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
843 in6_dev_put(idev);
844 ret = -1;
845 goto out;
847 spin_lock_bh(&ifp->lock);
848 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
849 idev->cnf.use_tempaddr = -1; /*XXX*/
850 spin_unlock_bh(&ifp->lock);
851 write_unlock(&idev->lock);
852 printk(KERN_WARNING
853 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
854 in6_dev_put(idev);
855 ret = -1;
856 goto out;
858 in6_ifa_hold(ifp);
859 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
860 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
861 spin_unlock_bh(&ifp->lock);
862 write_unlock(&idev->lock);
863 printk(KERN_WARNING
864 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
865 in6_ifa_put(ifp);
866 in6_dev_put(idev);
867 ret = -1;
868 goto out;
870 memcpy(&addr.s6_addr[8], idev->rndid, 8);
871 age = (now - ifp->tstamp) / HZ;
872 tmp_valid_lft = min_t(__u32,
873 ifp->valid_lft,
874 idev->cnf.temp_valid_lft + age);
875 tmp_prefered_lft = min_t(__u32,
876 ifp->prefered_lft,
877 idev->cnf.temp_prefered_lft + age -
878 idev->cnf.max_desync_factor);
879 tmp_plen = ifp->prefix_len;
880 max_addresses = idev->cnf.max_addresses;
881 tmp_tstamp = ifp->tstamp;
882 spin_unlock_bh(&ifp->lock);
884 regen_advance = idev->cnf.regen_max_retry *
885 idev->cnf.dad_transmits *
886 idev->nd_parms->retrans_time / HZ;
887 write_unlock(&idev->lock);
889 /* A temporary address is created only if this calculated Preferred
890 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
891 * an implementation must not create a temporary address with a zero
892 * Preferred Lifetime.
894 if (tmp_prefered_lft <= regen_advance) {
895 in6_ifa_put(ifp);
896 in6_dev_put(idev);
897 ret = -1;
898 goto out;
901 addr_flags = IFA_F_TEMPORARY;
902 /* set in addrconf_prefix_rcv() */
903 if (ifp->flags & IFA_F_OPTIMISTIC)
904 addr_flags |= IFA_F_OPTIMISTIC;
906 ift = !max_addresses ||
907 ipv6_count_addresses(idev) < max_addresses ?
908 ipv6_add_addr(idev, &addr, tmp_plen,
909 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
910 addr_flags) : NULL;
911 if (!ift || IS_ERR(ift)) {
912 in6_ifa_put(ifp);
913 in6_dev_put(idev);
914 printk(KERN_INFO
915 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
916 tmpaddr = &addr;
917 write_lock(&idev->lock);
918 goto retry;
921 spin_lock_bh(&ift->lock);
922 ift->ifpub = ifp;
923 ift->valid_lft = tmp_valid_lft;
924 ift->prefered_lft = tmp_prefered_lft;
925 ift->cstamp = now;
926 ift->tstamp = tmp_tstamp;
927 spin_unlock_bh(&ift->lock);
929 addrconf_dad_start(ift, 0);
930 in6_ifa_put(ift);
931 in6_dev_put(idev);
932 out:
933 return ret;
935 #endif
938 * Choose an appropriate source address (RFC3484)
940 enum {
941 IPV6_SADDR_RULE_INIT = 0,
942 IPV6_SADDR_RULE_LOCAL,
943 IPV6_SADDR_RULE_SCOPE,
944 IPV6_SADDR_RULE_PREFERRED,
945 #ifdef CONFIG_IPV6_MIP6
946 IPV6_SADDR_RULE_HOA,
947 #endif
948 IPV6_SADDR_RULE_OIF,
949 IPV6_SADDR_RULE_LABEL,
950 #ifdef CONFIG_IPV6_PRIVACY
951 IPV6_SADDR_RULE_PRIVACY,
952 #endif
953 IPV6_SADDR_RULE_ORCHID,
954 IPV6_SADDR_RULE_PREFIX,
955 IPV6_SADDR_RULE_MAX
958 struct ipv6_saddr_score {
959 int rule;
960 int addr_type;
961 struct inet6_ifaddr *ifa;
962 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
963 int scopedist;
964 int matchlen;
967 struct ipv6_saddr_dst {
968 const struct in6_addr *addr;
969 int ifindex;
970 int scope;
971 int label;
972 unsigned int prefs;
975 static inline int ipv6_saddr_preferred(int type)
977 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
978 return 1;
979 return 0;
982 static int ipv6_get_saddr_eval(struct net *net,
983 struct ipv6_saddr_score *score,
984 struct ipv6_saddr_dst *dst,
985 int i)
987 int ret;
989 if (i <= score->rule) {
990 switch (i) {
991 case IPV6_SADDR_RULE_SCOPE:
992 ret = score->scopedist;
993 break;
994 case IPV6_SADDR_RULE_PREFIX:
995 ret = score->matchlen;
996 break;
997 default:
998 ret = !!test_bit(i, score->scorebits);
1000 goto out;
1003 switch (i) {
1004 case IPV6_SADDR_RULE_INIT:
1005 /* Rule 0: remember if hiscore is not ready yet */
1006 ret = !!score->ifa;
1007 break;
1008 case IPV6_SADDR_RULE_LOCAL:
1009 /* Rule 1: Prefer same address */
1010 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1011 break;
1012 case IPV6_SADDR_RULE_SCOPE:
1013 /* Rule 2: Prefer appropriate scope
1015 * ret
1017 * -1 | d 15
1018 * ---+--+-+---> scope
1020 * | d is scope of the destination.
1021 * B-d | \
1022 * | \ <- smaller scope is better if
1023 * B-15 | \ if scope is enough for destinaion.
1024 * | ret = B - scope (-1 <= scope >= d <= 15).
1025 * d-C-1 | /
1026 * |/ <- greater is better
1027 * -C / if scope is not enough for destination.
1028 * /| ret = scope - C (-1 <= d < scope <= 15).
1030 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1031 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1032 * Assume B = 0 and we get C > 29.
1034 ret = __ipv6_addr_src_scope(score->addr_type);
1035 if (ret >= dst->scope)
1036 ret = -ret;
1037 else
1038 ret -= 128; /* 30 is enough */
1039 score->scopedist = ret;
1040 break;
1041 case IPV6_SADDR_RULE_PREFERRED:
1042 /* Rule 3: Avoid deprecated and optimistic addresses */
1043 ret = ipv6_saddr_preferred(score->addr_type) ||
1044 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1045 break;
1046 #ifdef CONFIG_IPV6_MIP6
1047 case IPV6_SADDR_RULE_HOA:
1049 /* Rule 4: Prefer home address */
1050 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1051 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1052 break;
1054 #endif
1055 case IPV6_SADDR_RULE_OIF:
1056 /* Rule 5: Prefer outgoing interface */
1057 ret = (!dst->ifindex ||
1058 dst->ifindex == score->ifa->idev->dev->ifindex);
1059 break;
1060 case IPV6_SADDR_RULE_LABEL:
1061 /* Rule 6: Prefer matching label */
1062 ret = ipv6_addr_label(net,
1063 &score->ifa->addr, score->addr_type,
1064 score->ifa->idev->dev->ifindex) == dst->label;
1065 break;
1066 #ifdef CONFIG_IPV6_PRIVACY
1067 case IPV6_SADDR_RULE_PRIVACY:
1069 /* Rule 7: Prefer public address
1070 * Note: prefer temporary address if use_tempaddr >= 2
1072 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1073 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1074 score->ifa->idev->cnf.use_tempaddr >= 2;
1075 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1076 break;
1078 #endif
1079 case IPV6_SADDR_RULE_ORCHID:
1080 /* Rule 8-: Prefer ORCHID vs ORCHID or
1081 * non-ORCHID vs non-ORCHID
1083 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1084 ipv6_addr_orchid(dst->addr));
1085 break;
1086 case IPV6_SADDR_RULE_PREFIX:
1087 /* Rule 8: Use longest matching prefix */
1088 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1089 dst->addr);
1090 break;
1091 default:
1092 ret = 0;
1095 if (ret)
1096 __set_bit(i, score->scorebits);
1097 score->rule = i;
1098 out:
1099 return ret;
1102 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1103 const struct in6_addr *daddr, unsigned int prefs,
1104 struct in6_addr *saddr)
1106 struct ipv6_saddr_score scores[2],
1107 *score = &scores[0], *hiscore = &scores[1];
1108 struct ipv6_saddr_dst dst;
1109 struct net_device *dev;
1110 int dst_type;
1112 dst_type = __ipv6_addr_type(daddr);
1113 dst.addr = daddr;
1114 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1115 dst.scope = __ipv6_addr_src_scope(dst_type);
1116 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1117 dst.prefs = prefs;
1119 hiscore->rule = -1;
1120 hiscore->ifa = NULL;
1122 rcu_read_lock();
1124 for_each_netdev_rcu(net, dev) {
1125 struct inet6_dev *idev;
1127 /* Candidate Source Address (section 4)
1128 * - multicast and link-local destination address,
1129 * the set of candidate source address MUST only
1130 * include addresses assigned to interfaces
1131 * belonging to the same link as the outgoing
1132 * interface.
1133 * (- For site-local destination addresses, the
1134 * set of candidate source addresses MUST only
1135 * include addresses assigned to interfaces
1136 * belonging to the same site as the outgoing
1137 * interface.)
1139 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1140 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1141 dst.ifindex && dev->ifindex != dst.ifindex)
1142 continue;
1144 idev = __in6_dev_get(dev);
1145 if (!idev)
1146 continue;
1148 read_lock_bh(&idev->lock);
1149 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1150 int i;
1153 * - Tentative Address (RFC2462 section 5.4)
1154 * - A tentative address is not considered
1155 * "assigned to an interface" in the traditional
1156 * sense, unless it is also flagged as optimistic.
1157 * - Candidate Source Address (section 4)
1158 * - In any case, anycast addresses, multicast
1159 * addresses, and the unspecified address MUST
1160 * NOT be included in a candidate set.
1162 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1163 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1164 continue;
1166 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1168 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1169 score->addr_type & IPV6_ADDR_MULTICAST)) {
1170 LIMIT_NETDEBUG(KERN_DEBUG
1171 "ADDRCONF: unspecified / multicast address "
1172 "assigned as unicast address on %s",
1173 dev->name);
1174 continue;
1177 score->rule = -1;
1178 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1180 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1181 int minihiscore, miniscore;
1183 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1184 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1186 if (minihiscore > miniscore) {
1187 if (i == IPV6_SADDR_RULE_SCOPE &&
1188 score->scopedist > 0) {
1190 * special case:
1191 * each remaining entry
1192 * has too small (not enough)
1193 * scope, because ifa entries
1194 * are sorted by their scope
1195 * values.
1197 goto try_nextdev;
1199 break;
1200 } else if (minihiscore < miniscore) {
1201 if (hiscore->ifa)
1202 in6_ifa_put(hiscore->ifa);
1204 in6_ifa_hold(score->ifa);
1206 swap(hiscore, score);
1208 /* restore our iterator */
1209 score->ifa = hiscore->ifa;
1211 break;
1215 try_nextdev:
1216 read_unlock_bh(&idev->lock);
1218 rcu_read_unlock();
1220 if (!hiscore->ifa)
1221 return -EADDRNOTAVAIL;
1223 *saddr = hiscore->ifa->addr;
1224 in6_ifa_put(hiscore->ifa);
1225 return 0;
1227 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1229 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1230 unsigned char banned_flags)
1232 struct inet6_dev *idev;
1233 int err = -EADDRNOTAVAIL;
1235 rcu_read_lock();
1236 idev = __in6_dev_get(dev);
1237 if (idev) {
1238 struct inet6_ifaddr *ifp;
1240 read_lock_bh(&idev->lock);
1241 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1242 if (ifp->scope == IFA_LINK &&
1243 !(ifp->flags & banned_flags)) {
1244 *addr = ifp->addr;
1245 err = 0;
1246 break;
1249 read_unlock_bh(&idev->lock);
1251 rcu_read_unlock();
1252 return err;
1255 static int ipv6_count_addresses(struct inet6_dev *idev)
1257 int cnt = 0;
1258 struct inet6_ifaddr *ifp;
1260 read_lock_bh(&idev->lock);
1261 list_for_each_entry(ifp, &idev->addr_list, if_list)
1262 cnt++;
1263 read_unlock_bh(&idev->lock);
1264 return cnt;
1267 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1268 struct net_device *dev, int strict)
1270 struct inet6_ifaddr *ifp;
1271 struct hlist_node *node;
1272 unsigned int hash = ipv6_addr_hash(addr);
1274 rcu_read_lock_bh();
1275 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1276 if (!net_eq(dev_net(ifp->idev->dev), net))
1277 continue;
1278 if (ipv6_addr_equal(&ifp->addr, addr) &&
1279 !(ifp->flags&IFA_F_TENTATIVE) &&
1280 (dev == NULL || ifp->idev->dev == dev ||
1281 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1282 rcu_read_unlock_bh();
1283 return 1;
1287 rcu_read_unlock_bh();
1288 return 0;
1290 EXPORT_SYMBOL(ipv6_chk_addr);
1292 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1293 struct net_device *dev)
1295 unsigned int hash = ipv6_addr_hash(addr);
1296 struct inet6_ifaddr *ifp;
1297 struct hlist_node *node;
1299 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1300 if (!net_eq(dev_net(ifp->idev->dev), net))
1301 continue;
1302 if (ipv6_addr_equal(&ifp->addr, addr)) {
1303 if (dev == NULL || ifp->idev->dev == dev)
1304 return true;
1307 return false;
1310 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1312 struct inet6_dev *idev;
1313 struct inet6_ifaddr *ifa;
1314 int onlink;
1316 onlink = 0;
1317 rcu_read_lock();
1318 idev = __in6_dev_get(dev);
1319 if (idev) {
1320 read_lock_bh(&idev->lock);
1321 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1322 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1323 ifa->prefix_len);
1324 if (onlink)
1325 break;
1327 read_unlock_bh(&idev->lock);
1329 rcu_read_unlock();
1330 return onlink;
1333 EXPORT_SYMBOL(ipv6_chk_prefix);
1335 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1336 struct net_device *dev, int strict)
1338 struct inet6_ifaddr *ifp, *result = NULL;
1339 unsigned int hash = ipv6_addr_hash(addr);
1340 struct hlist_node *node;
1342 rcu_read_lock_bh();
1343 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1344 if (!net_eq(dev_net(ifp->idev->dev), net))
1345 continue;
1346 if (ipv6_addr_equal(&ifp->addr, addr)) {
1347 if (dev == NULL || ifp->idev->dev == dev ||
1348 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1349 result = ifp;
1350 in6_ifa_hold(ifp);
1351 break;
1355 rcu_read_unlock_bh();
1357 return result;
1360 /* Gets referenced address, destroys ifaddr */
1362 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1364 if (ifp->flags&IFA_F_PERMANENT) {
1365 spin_lock_bh(&ifp->lock);
1366 addrconf_del_timer(ifp);
1367 ifp->flags |= IFA_F_TENTATIVE;
1368 if (dad_failed)
1369 ifp->flags |= IFA_F_DADFAILED;
1370 spin_unlock_bh(&ifp->lock);
1371 if (dad_failed)
1372 ipv6_ifa_notify(0, ifp);
1373 in6_ifa_put(ifp);
1374 #ifdef CONFIG_IPV6_PRIVACY
1375 } else if (ifp->flags&IFA_F_TEMPORARY) {
1376 struct inet6_ifaddr *ifpub;
1377 spin_lock_bh(&ifp->lock);
1378 ifpub = ifp->ifpub;
1379 if (ifpub) {
1380 in6_ifa_hold(ifpub);
1381 spin_unlock_bh(&ifp->lock);
1382 ipv6_create_tempaddr(ifpub, ifp);
1383 in6_ifa_put(ifpub);
1384 } else {
1385 spin_unlock_bh(&ifp->lock);
1387 ipv6_del_addr(ifp);
1388 #endif
1389 } else
1390 ipv6_del_addr(ifp);
1393 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1395 int err = -ENOENT;
1397 spin_lock(&ifp->state_lock);
1398 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1399 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1400 err = 0;
1402 spin_unlock(&ifp->state_lock);
1404 return err;
1407 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1409 struct inet6_dev *idev = ifp->idev;
1411 if (addrconf_dad_end(ifp)) {
1412 in6_ifa_put(ifp);
1413 return;
1416 if (net_ratelimit())
1417 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1418 ifp->idev->dev->name, &ifp->addr);
1420 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1421 struct in6_addr addr;
1423 addr.s6_addr32[0] = htonl(0xfe800000);
1424 addr.s6_addr32[1] = 0;
1426 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1427 ipv6_addr_equal(&ifp->addr, &addr)) {
1428 /* DAD failed for link-local based on MAC address */
1429 idev->cnf.disable_ipv6 = 1;
1431 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1432 ifp->idev->dev->name);
1436 addrconf_dad_stop(ifp, 1);
1439 /* Join to solicited addr multicast group. */
1441 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1443 struct in6_addr maddr;
1445 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1446 return;
1448 addrconf_addr_solict_mult(addr, &maddr);
1449 ipv6_dev_mc_inc(dev, &maddr);
1452 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1454 struct in6_addr maddr;
1456 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1457 return;
1459 addrconf_addr_solict_mult(addr, &maddr);
1460 __ipv6_dev_mc_dec(idev, &maddr);
1463 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1465 struct in6_addr addr;
1466 if (ifp->prefix_len == 127) /* RFC 6164 */
1467 return;
1468 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1469 if (ipv6_addr_any(&addr))
1470 return;
1471 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1474 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1476 struct in6_addr addr;
1477 if (ifp->prefix_len == 127) /* RFC 6164 */
1478 return;
1479 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1480 if (ipv6_addr_any(&addr))
1481 return;
1482 __ipv6_dev_ac_dec(ifp->idev, &addr);
1485 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1487 if (dev->addr_len != ETH_ALEN)
1488 return -1;
1489 memcpy(eui, dev->dev_addr, 3);
1490 memcpy(eui + 5, dev->dev_addr + 3, 3);
1493 * The zSeries OSA network cards can be shared among various
1494 * OS instances, but the OSA cards have only one MAC address.
1495 * This leads to duplicate address conflicts in conjunction
1496 * with IPv6 if more than one instance uses the same card.
1498 * The driver for these cards can deliver a unique 16-bit
1499 * identifier for each instance sharing the same card. It is
1500 * placed instead of 0xFFFE in the interface identifier. The
1501 * "u" bit of the interface identifier is not inverted in this
1502 * case. Hence the resulting interface identifier has local
1503 * scope according to RFC2373.
1505 if (dev->dev_id) {
1506 eui[3] = (dev->dev_id >> 8) & 0xFF;
1507 eui[4] = dev->dev_id & 0xFF;
1508 } else {
1509 eui[3] = 0xFF;
1510 eui[4] = 0xFE;
1511 eui[0] ^= 2;
1513 return 0;
1516 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1518 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1519 if (dev->addr_len != ARCNET_ALEN)
1520 return -1;
1521 memset(eui, 0, 7);
1522 eui[7] = *(u8*)dev->dev_addr;
1523 return 0;
1526 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1528 if (dev->addr_len != INFINIBAND_ALEN)
1529 return -1;
1530 memcpy(eui, dev->dev_addr + 12, 8);
1531 eui[0] |= 2;
1532 return 0;
1535 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1537 if (addr == 0)
1538 return -1;
1539 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1540 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1541 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1542 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1543 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1544 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1545 eui[1] = 0;
1546 eui[2] = 0x5E;
1547 eui[3] = 0xFE;
1548 memcpy(eui + 4, &addr, 4);
1549 return 0;
1552 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1554 if (dev->priv_flags & IFF_ISATAP)
1555 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1556 return -1;
1559 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1561 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1564 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1566 switch (dev->type) {
1567 case ARPHRD_ETHER:
1568 case ARPHRD_FDDI:
1569 case ARPHRD_IEEE802_TR:
1570 return addrconf_ifid_eui48(eui, dev);
1571 case ARPHRD_ARCNET:
1572 return addrconf_ifid_arcnet(eui, dev);
1573 case ARPHRD_INFINIBAND:
1574 return addrconf_ifid_infiniband(eui, dev);
1575 case ARPHRD_SIT:
1576 return addrconf_ifid_sit(eui, dev);
1577 case ARPHRD_IPGRE:
1578 return addrconf_ifid_gre(eui, dev);
1580 return -1;
1583 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1585 int err = -1;
1586 struct inet6_ifaddr *ifp;
1588 read_lock_bh(&idev->lock);
1589 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1590 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1591 memcpy(eui, ifp->addr.s6_addr+8, 8);
1592 err = 0;
1593 break;
1596 read_unlock_bh(&idev->lock);
1597 return err;
1600 #ifdef CONFIG_IPV6_PRIVACY
1601 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1602 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1604 regen:
1605 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1606 idev->rndid[0] &= ~0x02;
1609 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1610 * check if generated address is not inappropriate
1612 * - Reserved subnet anycast (RFC 2526)
1613 * 11111101 11....11 1xxxxxxx
1614 * - ISATAP (RFC4214) 6.1
1615 * 00-00-5E-FE-xx-xx-xx-xx
1616 * - value 0
1617 * - XXX: already assigned to an address on the device
1619 if (idev->rndid[0] == 0xfd &&
1620 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1621 (idev->rndid[7]&0x80))
1622 goto regen;
1623 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1624 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1625 goto regen;
1626 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627 goto regen;
1630 return 0;
1633 static void ipv6_regen_rndid(unsigned long data)
1635 struct inet6_dev *idev = (struct inet6_dev *) data;
1636 unsigned long expires;
1638 rcu_read_lock_bh();
1639 write_lock_bh(&idev->lock);
1641 if (idev->dead)
1642 goto out;
1644 if (__ipv6_regen_rndid(idev) < 0)
1645 goto out;
1647 expires = jiffies +
1648 idev->cnf.temp_prefered_lft * HZ -
1649 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1650 idev->cnf.max_desync_factor * HZ;
1651 if (time_before(expires, jiffies)) {
1652 printk(KERN_WARNING
1653 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1654 idev->dev->name);
1655 goto out;
1658 if (!mod_timer(&idev->regen_timer, expires))
1659 in6_dev_hold(idev);
1661 out:
1662 write_unlock_bh(&idev->lock);
1663 rcu_read_unlock_bh();
1664 in6_dev_put(idev);
1667 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1668 int ret = 0;
1670 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1671 ret = __ipv6_regen_rndid(idev);
1672 return ret;
1674 #endif
1677 * Add prefix route.
1680 static void
1681 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1682 unsigned long expires, u32 flags)
1684 struct fib6_config cfg = {
1685 .fc_table = RT6_TABLE_PREFIX,
1686 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1687 .fc_ifindex = dev->ifindex,
1688 .fc_expires = expires,
1689 .fc_dst_len = plen,
1690 .fc_flags = RTF_UP | flags,
1691 .fc_nlinfo.nl_net = dev_net(dev),
1692 .fc_protocol = RTPROT_KERNEL,
1695 cfg.fc_dst = *pfx;
1697 /* Prevent useless cloning on PtP SIT.
1698 This thing is done here expecting that the whole
1699 class of non-broadcast devices need not cloning.
1701 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1702 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1703 cfg.fc_flags |= RTF_NONEXTHOP;
1704 #endif
1706 ip6_route_add(&cfg);
1710 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1711 int plen,
1712 const struct net_device *dev,
1713 u32 flags, u32 noflags)
1715 struct fib6_node *fn;
1716 struct rt6_info *rt = NULL;
1717 struct fib6_table *table;
1719 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1720 if (table == NULL)
1721 return NULL;
1723 write_lock_bh(&table->tb6_lock);
1724 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1725 if (!fn)
1726 goto out;
1727 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1728 if (rt->dst.dev->ifindex != dev->ifindex)
1729 continue;
1730 if ((rt->rt6i_flags & flags) != flags)
1731 continue;
1732 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1733 continue;
1734 dst_hold(&rt->dst);
1735 break;
1737 out:
1738 write_unlock_bh(&table->tb6_lock);
1739 return rt;
1743 /* Create "default" multicast route to the interface */
1745 static void addrconf_add_mroute(struct net_device *dev)
1747 struct fib6_config cfg = {
1748 .fc_table = RT6_TABLE_LOCAL,
1749 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1750 .fc_ifindex = dev->ifindex,
1751 .fc_dst_len = 8,
1752 .fc_flags = RTF_UP,
1753 .fc_nlinfo.nl_net = dev_net(dev),
1756 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1758 ip6_route_add(&cfg);
1761 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1762 static void sit_route_add(struct net_device *dev)
1764 struct fib6_config cfg = {
1765 .fc_table = RT6_TABLE_MAIN,
1766 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1767 .fc_ifindex = dev->ifindex,
1768 .fc_dst_len = 96,
1769 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1770 .fc_nlinfo.nl_net = dev_net(dev),
1773 /* prefix length - 96 bits "::d.d.d.d" */
1774 ip6_route_add(&cfg);
1776 #endif
1778 static void addrconf_add_lroute(struct net_device *dev)
1780 struct in6_addr addr;
1782 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1783 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1786 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1788 struct inet6_dev *idev;
1790 ASSERT_RTNL();
1792 idev = ipv6_find_idev(dev);
1793 if (!idev)
1794 return ERR_PTR(-ENOBUFS);
1796 if (idev->cnf.disable_ipv6)
1797 return ERR_PTR(-EACCES);
1799 /* Add default multicast route */
1800 if (!(dev->flags & IFF_LOOPBACK))
1801 addrconf_add_mroute(dev);
1803 /* Add link local route */
1804 addrconf_add_lroute(dev);
1805 return idev;
1808 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1810 struct prefix_info *pinfo;
1811 __u32 valid_lft;
1812 __u32 prefered_lft;
1813 int addr_type;
1814 struct inet6_dev *in6_dev;
1815 struct net *net = dev_net(dev);
1817 pinfo = (struct prefix_info *) opt;
1819 if (len < sizeof(struct prefix_info)) {
1820 ADBG(("addrconf: prefix option too short\n"));
1821 return;
1825 * Validation checks ([ADDRCONF], page 19)
1828 addr_type = ipv6_addr_type(&pinfo->prefix);
1830 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1831 return;
1833 valid_lft = ntohl(pinfo->valid);
1834 prefered_lft = ntohl(pinfo->prefered);
1836 if (prefered_lft > valid_lft) {
1837 if (net_ratelimit())
1838 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1839 return;
1842 in6_dev = in6_dev_get(dev);
1844 if (in6_dev == NULL) {
1845 if (net_ratelimit())
1846 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1847 return;
1851 * Two things going on here:
1852 * 1) Add routes for on-link prefixes
1853 * 2) Configure prefixes with the auto flag set
1856 if (pinfo->onlink) {
1857 struct rt6_info *rt;
1858 unsigned long rt_expires;
1860 /* Avoid arithmetic overflow. Really, we could
1861 * save rt_expires in seconds, likely valid_lft,
1862 * but it would require division in fib gc, that it
1863 * not good.
1865 if (HZ > USER_HZ)
1866 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1867 else
1868 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1870 if (addrconf_finite_timeout(rt_expires))
1871 rt_expires *= HZ;
1873 rt = addrconf_get_prefix_route(&pinfo->prefix,
1874 pinfo->prefix_len,
1875 dev,
1876 RTF_ADDRCONF | RTF_PREFIX_RT,
1877 RTF_GATEWAY | RTF_DEFAULT);
1879 if (rt) {
1880 /* Autoconf prefix route */
1881 if (valid_lft == 0) {
1882 ip6_del_rt(rt);
1883 rt = NULL;
1884 } else if (addrconf_finite_timeout(rt_expires)) {
1885 /* not infinity */
1886 rt->dst.expires = jiffies + rt_expires;
1887 rt->rt6i_flags |= RTF_EXPIRES;
1888 } else {
1889 rt->rt6i_flags &= ~RTF_EXPIRES;
1890 rt->dst.expires = 0;
1892 } else if (valid_lft) {
1893 clock_t expires = 0;
1894 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1895 if (addrconf_finite_timeout(rt_expires)) {
1896 /* not infinity */
1897 flags |= RTF_EXPIRES;
1898 expires = jiffies_to_clock_t(rt_expires);
1900 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1901 dev, expires, flags);
1903 if (rt)
1904 dst_release(&rt->dst);
1907 /* Try to figure out our local address for this prefix */
1909 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1910 struct inet6_ifaddr * ifp;
1911 struct in6_addr addr;
1912 int create = 0, update_lft = 0;
1914 if (pinfo->prefix_len == 64) {
1915 memcpy(&addr, &pinfo->prefix, 8);
1916 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1917 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1918 in6_dev_put(in6_dev);
1919 return;
1921 goto ok;
1923 if (net_ratelimit())
1924 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1925 pinfo->prefix_len);
1926 in6_dev_put(in6_dev);
1927 return;
1931 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1933 if (ifp == NULL && valid_lft) {
1934 int max_addresses = in6_dev->cnf.max_addresses;
1935 u32 addr_flags = 0;
1937 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1938 if (in6_dev->cnf.optimistic_dad &&
1939 !net->ipv6.devconf_all->forwarding && sllao)
1940 addr_flags = IFA_F_OPTIMISTIC;
1941 #endif
1943 /* Do not allow to create too much of autoconfigured
1944 * addresses; this would be too easy way to crash kernel.
1946 if (!max_addresses ||
1947 ipv6_count_addresses(in6_dev) < max_addresses)
1948 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1949 addr_type&IPV6_ADDR_SCOPE_MASK,
1950 addr_flags);
1952 if (!ifp || IS_ERR(ifp)) {
1953 in6_dev_put(in6_dev);
1954 return;
1957 update_lft = create = 1;
1958 ifp->cstamp = jiffies;
1959 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1962 if (ifp) {
1963 int flags;
1964 unsigned long now;
1965 #ifdef CONFIG_IPV6_PRIVACY
1966 struct inet6_ifaddr *ift;
1967 #endif
1968 u32 stored_lft;
1970 /* update lifetime (RFC2462 5.5.3 e) */
1971 spin_lock(&ifp->lock);
1972 now = jiffies;
1973 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1974 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1975 else
1976 stored_lft = 0;
1977 if (!update_lft && stored_lft) {
1978 if (valid_lft > MIN_VALID_LIFETIME ||
1979 valid_lft > stored_lft)
1980 update_lft = 1;
1981 else if (stored_lft <= MIN_VALID_LIFETIME) {
1982 /* valid_lft <= stored_lft is always true */
1984 * RFC 4862 Section 5.5.3e:
1985 * "Note that the preferred lifetime of
1986 * the corresponding address is always
1987 * reset to the Preferred Lifetime in
1988 * the received Prefix Information
1989 * option, regardless of whether the
1990 * valid lifetime is also reset or
1991 * ignored."
1993 * So if the preferred lifetime in
1994 * this advertisement is different
1995 * than what we have stored, but the
1996 * valid lifetime is invalid, just
1997 * reset prefered_lft.
1999 * We must set the valid lifetime
2000 * to the stored lifetime since we'll
2001 * be updating the timestamp below,
2002 * else we'll set it back to the
2003 * minimum.
2005 if (prefered_lft != ifp->prefered_lft) {
2006 valid_lft = stored_lft;
2007 update_lft = 1;
2009 } else {
2010 valid_lft = MIN_VALID_LIFETIME;
2011 if (valid_lft < prefered_lft)
2012 prefered_lft = valid_lft;
2013 update_lft = 1;
2017 if (update_lft) {
2018 ifp->valid_lft = valid_lft;
2019 ifp->prefered_lft = prefered_lft;
2020 ifp->tstamp = now;
2021 flags = ifp->flags;
2022 ifp->flags &= ~IFA_F_DEPRECATED;
2023 spin_unlock(&ifp->lock);
2025 if (!(flags&IFA_F_TENTATIVE))
2026 ipv6_ifa_notify(0, ifp);
2027 } else
2028 spin_unlock(&ifp->lock);
2030 #ifdef CONFIG_IPV6_PRIVACY
2031 read_lock_bh(&in6_dev->lock);
2032 /* update all temporary addresses in the list */
2033 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2034 tmp_list) {
2035 int age, max_valid, max_prefered;
2037 if (ifp != ift->ifpub)
2038 continue;
2041 * RFC 4941 section 3.3:
2042 * If a received option will extend the lifetime
2043 * of a public address, the lifetimes of
2044 * temporary addresses should be extended,
2045 * subject to the overall constraint that no
2046 * temporary addresses should ever remain
2047 * "valid" or "preferred" for a time longer than
2048 * (TEMP_VALID_LIFETIME) or
2049 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2050 * respectively.
2052 age = (now - ift->cstamp) / HZ;
2053 max_valid = in6_dev->cnf.temp_valid_lft - age;
2054 if (max_valid < 0)
2055 max_valid = 0;
2057 max_prefered = in6_dev->cnf.temp_prefered_lft -
2058 in6_dev->cnf.max_desync_factor -
2059 age;
2060 if (max_prefered < 0)
2061 max_prefered = 0;
2063 if (valid_lft > max_valid)
2064 valid_lft = max_valid;
2066 if (prefered_lft > max_prefered)
2067 prefered_lft = max_prefered;
2069 spin_lock(&ift->lock);
2070 flags = ift->flags;
2071 ift->valid_lft = valid_lft;
2072 ift->prefered_lft = prefered_lft;
2073 ift->tstamp = now;
2074 if (prefered_lft > 0)
2075 ift->flags &= ~IFA_F_DEPRECATED;
2077 spin_unlock(&ift->lock);
2078 if (!(flags&IFA_F_TENTATIVE))
2079 ipv6_ifa_notify(0, ift);
2082 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2084 * When a new public address is created as
2085 * described in [ADDRCONF], also create a new
2086 * temporary address. Also create a temporary
2087 * address if it's enabled but no temporary
2088 * address currently exists.
2090 read_unlock_bh(&in6_dev->lock);
2091 ipv6_create_tempaddr(ifp, NULL);
2092 } else {
2093 read_unlock_bh(&in6_dev->lock);
2095 #endif
2096 in6_ifa_put(ifp);
2097 addrconf_verify(0);
2100 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2101 in6_dev_put(in6_dev);
2105 * Set destination address.
2106 * Special case for SIT interfaces where we create a new "virtual"
2107 * device.
2109 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2111 struct in6_ifreq ireq;
2112 struct net_device *dev;
2113 int err = -EINVAL;
2115 rtnl_lock();
2117 err = -EFAULT;
2118 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2119 goto err_exit;
2121 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2123 err = -ENODEV;
2124 if (dev == NULL)
2125 goto err_exit;
2127 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2128 if (dev->type == ARPHRD_SIT) {
2129 const struct net_device_ops *ops = dev->netdev_ops;
2130 struct ifreq ifr;
2131 struct ip_tunnel_parm p;
2133 err = -EADDRNOTAVAIL;
2134 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2135 goto err_exit;
2137 memset(&p, 0, sizeof(p));
2138 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2139 p.iph.saddr = 0;
2140 p.iph.version = 4;
2141 p.iph.ihl = 5;
2142 p.iph.protocol = IPPROTO_IPV6;
2143 p.iph.ttl = 64;
2144 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2146 if (ops->ndo_do_ioctl) {
2147 mm_segment_t oldfs = get_fs();
2149 set_fs(KERNEL_DS);
2150 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2151 set_fs(oldfs);
2152 } else
2153 err = -EOPNOTSUPP;
2155 if (err == 0) {
2156 err = -ENOBUFS;
2157 dev = __dev_get_by_name(net, p.name);
2158 if (!dev)
2159 goto err_exit;
2160 err = dev_open(dev);
2163 #endif
2165 err_exit:
2166 rtnl_unlock();
2167 return err;
2171 * Manual configuration of address on an interface
2173 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2174 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2175 __u32 valid_lft)
2177 struct inet6_ifaddr *ifp;
2178 struct inet6_dev *idev;
2179 struct net_device *dev;
2180 int scope;
2181 u32 flags;
2182 clock_t expires;
2183 unsigned long timeout;
2185 ASSERT_RTNL();
2187 if (plen > 128)
2188 return -EINVAL;
2190 /* check the lifetime */
2191 if (!valid_lft || prefered_lft > valid_lft)
2192 return -EINVAL;
2194 dev = __dev_get_by_index(net, ifindex);
2195 if (!dev)
2196 return -ENODEV;
2198 idev = addrconf_add_dev(dev);
2199 if (IS_ERR(idev))
2200 return PTR_ERR(idev);
2202 scope = ipv6_addr_scope(pfx);
2204 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2205 if (addrconf_finite_timeout(timeout)) {
2206 expires = jiffies_to_clock_t(timeout * HZ);
2207 valid_lft = timeout;
2208 flags = RTF_EXPIRES;
2209 } else {
2210 expires = 0;
2211 flags = 0;
2212 ifa_flags |= IFA_F_PERMANENT;
2215 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2216 if (addrconf_finite_timeout(timeout)) {
2217 if (timeout == 0)
2218 ifa_flags |= IFA_F_DEPRECATED;
2219 prefered_lft = timeout;
2222 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2224 if (!IS_ERR(ifp)) {
2225 spin_lock_bh(&ifp->lock);
2226 ifp->valid_lft = valid_lft;
2227 ifp->prefered_lft = prefered_lft;
2228 ifp->tstamp = jiffies;
2229 spin_unlock_bh(&ifp->lock);
2231 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2232 expires, flags);
2234 * Note that section 3.1 of RFC 4429 indicates
2235 * that the Optimistic flag should not be set for
2236 * manually configured addresses
2238 addrconf_dad_start(ifp, 0);
2239 in6_ifa_put(ifp);
2240 addrconf_verify(0);
2241 return 0;
2244 return PTR_ERR(ifp);
2247 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2248 unsigned int plen)
2250 struct inet6_ifaddr *ifp;
2251 struct inet6_dev *idev;
2252 struct net_device *dev;
2254 if (plen > 128)
2255 return -EINVAL;
2257 dev = __dev_get_by_index(net, ifindex);
2258 if (!dev)
2259 return -ENODEV;
2261 if ((idev = __in6_dev_get(dev)) == NULL)
2262 return -ENXIO;
2264 read_lock_bh(&idev->lock);
2265 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2266 if (ifp->prefix_len == plen &&
2267 ipv6_addr_equal(pfx, &ifp->addr)) {
2268 in6_ifa_hold(ifp);
2269 read_unlock_bh(&idev->lock);
2271 ipv6_del_addr(ifp);
2273 /* If the last address is deleted administratively,
2274 disable IPv6 on this interface.
2276 if (list_empty(&idev->addr_list))
2277 addrconf_ifdown(idev->dev, 1);
2278 return 0;
2281 read_unlock_bh(&idev->lock);
2282 return -EADDRNOTAVAIL;
2286 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2288 struct in6_ifreq ireq;
2289 int err;
2291 if (!capable(CAP_NET_ADMIN))
2292 return -EPERM;
2294 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2295 return -EFAULT;
2297 rtnl_lock();
2298 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2299 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2300 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2301 rtnl_unlock();
2302 return err;
2305 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2307 struct in6_ifreq ireq;
2308 int err;
2310 if (!capable(CAP_NET_ADMIN))
2311 return -EPERM;
2313 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2314 return -EFAULT;
2316 rtnl_lock();
2317 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2318 ireq.ifr6_prefixlen);
2319 rtnl_unlock();
2320 return err;
2323 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2324 int plen, int scope)
2326 struct inet6_ifaddr *ifp;
2328 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2329 if (!IS_ERR(ifp)) {
2330 spin_lock_bh(&ifp->lock);
2331 ifp->flags &= ~IFA_F_TENTATIVE;
2332 spin_unlock_bh(&ifp->lock);
2333 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2334 in6_ifa_put(ifp);
2338 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2339 static void sit_add_v4_addrs(struct inet6_dev *idev)
2341 struct in6_addr addr;
2342 struct net_device *dev;
2343 struct net *net = dev_net(idev->dev);
2344 int scope;
2346 ASSERT_RTNL();
2348 memset(&addr, 0, sizeof(struct in6_addr));
2349 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2351 if (idev->dev->flags&IFF_POINTOPOINT) {
2352 addr.s6_addr32[0] = htonl(0xfe800000);
2353 scope = IFA_LINK;
2354 } else {
2355 scope = IPV6_ADDR_COMPATv4;
2358 if (addr.s6_addr32[3]) {
2359 add_addr(idev, &addr, 128, scope);
2360 return;
2363 for_each_netdev(net, dev) {
2364 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2365 if (in_dev && (dev->flags & IFF_UP)) {
2366 struct in_ifaddr * ifa;
2368 int flag = scope;
2370 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2371 int plen;
2373 addr.s6_addr32[3] = ifa->ifa_local;
2375 if (ifa->ifa_scope == RT_SCOPE_LINK)
2376 continue;
2377 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2378 if (idev->dev->flags&IFF_POINTOPOINT)
2379 continue;
2380 flag |= IFA_HOST;
2382 if (idev->dev->flags&IFF_POINTOPOINT)
2383 plen = 64;
2384 else
2385 plen = 96;
2387 add_addr(idev, &addr, plen, flag);
2392 #endif
2394 static void init_loopback(struct net_device *dev)
2396 struct inet6_dev *idev;
2398 /* ::1 */
2400 ASSERT_RTNL();
2402 if ((idev = ipv6_find_idev(dev)) == NULL) {
2403 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2404 return;
2407 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2410 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2412 struct inet6_ifaddr * ifp;
2413 u32 addr_flags = IFA_F_PERMANENT;
2415 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2416 if (idev->cnf.optimistic_dad &&
2417 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2418 addr_flags |= IFA_F_OPTIMISTIC;
2419 #endif
2422 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2423 if (!IS_ERR(ifp)) {
2424 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2425 addrconf_dad_start(ifp, 0);
2426 in6_ifa_put(ifp);
2430 static void addrconf_dev_config(struct net_device *dev)
2432 struct in6_addr addr;
2433 struct inet6_dev * idev;
2435 ASSERT_RTNL();
2437 if ((dev->type != ARPHRD_ETHER) &&
2438 (dev->type != ARPHRD_FDDI) &&
2439 (dev->type != ARPHRD_IEEE802_TR) &&
2440 (dev->type != ARPHRD_ARCNET) &&
2441 (dev->type != ARPHRD_INFINIBAND)) {
2442 /* Alas, we support only Ethernet autoconfiguration. */
2443 return;
2446 idev = addrconf_add_dev(dev);
2447 if (IS_ERR(idev))
2448 return;
2450 memset(&addr, 0, sizeof(struct in6_addr));
2451 addr.s6_addr32[0] = htonl(0xFE800000);
2453 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2454 addrconf_add_linklocal(idev, &addr);
2457 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2458 static void addrconf_sit_config(struct net_device *dev)
2460 struct inet6_dev *idev;
2462 ASSERT_RTNL();
2465 * Configure the tunnel with one of our IPv4
2466 * addresses... we should configure all of
2467 * our v4 addrs in the tunnel
2470 if ((idev = ipv6_find_idev(dev)) == NULL) {
2471 printk(KERN_DEBUG "init sit: add_dev failed\n");
2472 return;
2475 if (dev->priv_flags & IFF_ISATAP) {
2476 struct in6_addr addr;
2478 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2479 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2480 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2481 addrconf_add_linklocal(idev, &addr);
2482 return;
2485 sit_add_v4_addrs(idev);
2487 if (dev->flags&IFF_POINTOPOINT) {
2488 addrconf_add_mroute(dev);
2489 addrconf_add_lroute(dev);
2490 } else
2491 sit_route_add(dev);
2493 #endif
2495 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2496 static void addrconf_gre_config(struct net_device *dev)
2498 struct inet6_dev *idev;
2499 struct in6_addr addr;
2501 pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2503 ASSERT_RTNL();
2505 if ((idev = ipv6_find_idev(dev)) == NULL) {
2506 printk(KERN_DEBUG "init gre: add_dev failed\n");
2507 return;
2510 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2511 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2513 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2514 addrconf_add_linklocal(idev, &addr);
2516 #endif
2518 static inline int
2519 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2521 struct in6_addr lladdr;
2523 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2524 addrconf_add_linklocal(idev, &lladdr);
2525 return 0;
2527 return -1;
2530 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2532 struct net_device *link_dev;
2533 struct net *net = dev_net(idev->dev);
2535 /* first try to inherit the link-local address from the link device */
2536 if (idev->dev->iflink &&
2537 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2538 if (!ipv6_inherit_linklocal(idev, link_dev))
2539 return;
2541 /* then try to inherit it from any device */
2542 for_each_netdev(net, link_dev) {
2543 if (!ipv6_inherit_linklocal(idev, link_dev))
2544 return;
2546 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2550 * Autoconfigure tunnel with a link-local address so routing protocols,
2551 * DHCPv6, MLD etc. can be run over the virtual link
2554 static void addrconf_ip6_tnl_config(struct net_device *dev)
2556 struct inet6_dev *idev;
2558 ASSERT_RTNL();
2560 idev = addrconf_add_dev(dev);
2561 if (IS_ERR(idev)) {
2562 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2563 return;
2565 ip6_tnl_add_linklocal(idev);
2568 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2569 void * data)
2571 struct net_device *dev = (struct net_device *) data;
2572 struct inet6_dev *idev = __in6_dev_get(dev);
2573 int run_pending = 0;
2574 int err;
2576 switch (event) {
2577 case NETDEV_REGISTER:
2578 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2579 idev = ipv6_add_dev(dev);
2580 if (!idev)
2581 return notifier_from_errno(-ENOMEM);
2583 break;
2585 case NETDEV_UP:
2586 case NETDEV_CHANGE:
2587 if (dev->flags & IFF_SLAVE)
2588 break;
2590 if (event == NETDEV_UP) {
2591 if (!addrconf_qdisc_ok(dev)) {
2592 /* device is not ready yet. */
2593 printk(KERN_INFO
2594 "ADDRCONF(NETDEV_UP): %s: "
2595 "link is not ready\n",
2596 dev->name);
2597 break;
2600 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2601 idev = ipv6_add_dev(dev);
2603 if (idev) {
2604 idev->if_flags |= IF_READY;
2605 run_pending = 1;
2607 } else {
2608 if (!addrconf_qdisc_ok(dev)) {
2609 /* device is still not ready. */
2610 break;
2613 if (idev) {
2614 if (idev->if_flags & IF_READY)
2615 /* device is already configured. */
2616 break;
2617 idev->if_flags |= IF_READY;
2620 printk(KERN_INFO
2621 "ADDRCONF(NETDEV_CHANGE): %s: "
2622 "link becomes ready\n",
2623 dev->name);
2625 run_pending = 1;
2628 switch (dev->type) {
2629 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2630 case ARPHRD_SIT:
2631 addrconf_sit_config(dev);
2632 break;
2633 #endif
2634 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2635 case ARPHRD_IPGRE:
2636 addrconf_gre_config(dev);
2637 break;
2638 #endif
2639 case ARPHRD_TUNNEL6:
2640 addrconf_ip6_tnl_config(dev);
2641 break;
2642 case ARPHRD_LOOPBACK:
2643 init_loopback(dev);
2644 break;
2646 default:
2647 addrconf_dev_config(dev);
2648 break;
2651 if (idev) {
2652 if (run_pending)
2653 addrconf_dad_run(idev);
2656 * If the MTU changed during the interface down,
2657 * when the interface up, the changed MTU must be
2658 * reflected in the idev as well as routers.
2660 if (idev->cnf.mtu6 != dev->mtu &&
2661 dev->mtu >= IPV6_MIN_MTU) {
2662 rt6_mtu_change(dev, dev->mtu);
2663 idev->cnf.mtu6 = dev->mtu;
2665 idev->tstamp = jiffies;
2666 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2669 * If the changed mtu during down is lower than
2670 * IPV6_MIN_MTU stop IPv6 on this interface.
2672 if (dev->mtu < IPV6_MIN_MTU)
2673 addrconf_ifdown(dev, 1);
2675 break;
2677 case NETDEV_CHANGEMTU:
2678 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2679 rt6_mtu_change(dev, dev->mtu);
2680 idev->cnf.mtu6 = dev->mtu;
2681 break;
2684 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2685 idev = ipv6_add_dev(dev);
2686 if (idev)
2687 break;
2691 * MTU falled under IPV6_MIN_MTU.
2692 * Stop IPv6 on this interface.
2695 case NETDEV_DOWN:
2696 case NETDEV_UNREGISTER:
2698 * Remove all addresses from this interface.
2700 addrconf_ifdown(dev, event != NETDEV_DOWN);
2701 break;
2703 case NETDEV_CHANGENAME:
2704 if (idev) {
2705 snmp6_unregister_dev(idev);
2706 addrconf_sysctl_unregister(idev);
2707 addrconf_sysctl_register(idev);
2708 err = snmp6_register_dev(idev);
2709 if (err)
2710 return notifier_from_errno(err);
2712 break;
2714 case NETDEV_PRE_TYPE_CHANGE:
2715 case NETDEV_POST_TYPE_CHANGE:
2716 addrconf_type_change(dev, event);
2717 break;
2720 return NOTIFY_OK;
2724 * addrconf module should be notified of a device going up
2726 static struct notifier_block ipv6_dev_notf = {
2727 .notifier_call = addrconf_notify,
2730 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2732 struct inet6_dev *idev;
2733 ASSERT_RTNL();
2735 idev = __in6_dev_get(dev);
2737 if (event == NETDEV_POST_TYPE_CHANGE)
2738 ipv6_mc_remap(idev);
2739 else if (event == NETDEV_PRE_TYPE_CHANGE)
2740 ipv6_mc_unmap(idev);
2743 static int addrconf_ifdown(struct net_device *dev, int how)
2745 struct net *net = dev_net(dev);
2746 struct inet6_dev *idev;
2747 struct inet6_ifaddr *ifa;
2748 int state, i;
2750 ASSERT_RTNL();
2752 rt6_ifdown(net, dev);
2753 neigh_ifdown(&nd_tbl, dev);
2755 idev = __in6_dev_get(dev);
2756 if (idev == NULL)
2757 return -ENODEV;
2760 * Step 1: remove reference to ipv6 device from parent device.
2761 * Do not dev_put!
2763 if (how) {
2764 idev->dead = 1;
2766 /* protected by rtnl_lock */
2767 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2769 /* Step 1.5: remove snmp6 entry */
2770 snmp6_unregister_dev(idev);
2774 /* Step 2: clear hash table */
2775 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2776 struct hlist_head *h = &inet6_addr_lst[i];
2777 struct hlist_node *n;
2779 spin_lock_bh(&addrconf_hash_lock);
2780 restart:
2781 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2782 if (ifa->idev == idev) {
2783 hlist_del_init_rcu(&ifa->addr_lst);
2784 addrconf_del_timer(ifa);
2785 goto restart;
2788 spin_unlock_bh(&addrconf_hash_lock);
2791 write_lock_bh(&idev->lock);
2793 /* Step 2: clear flags for stateless addrconf */
2794 if (!how)
2795 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2797 #ifdef CONFIG_IPV6_PRIVACY
2798 if (how && del_timer(&idev->regen_timer))
2799 in6_dev_put(idev);
2801 /* Step 3: clear tempaddr list */
2802 while (!list_empty(&idev->tempaddr_list)) {
2803 ifa = list_first_entry(&idev->tempaddr_list,
2804 struct inet6_ifaddr, tmp_list);
2805 list_del(&ifa->tmp_list);
2806 write_unlock_bh(&idev->lock);
2807 spin_lock_bh(&ifa->lock);
2809 if (ifa->ifpub) {
2810 in6_ifa_put(ifa->ifpub);
2811 ifa->ifpub = NULL;
2813 spin_unlock_bh(&ifa->lock);
2814 in6_ifa_put(ifa);
2815 write_lock_bh(&idev->lock);
2817 #endif
2819 while (!list_empty(&idev->addr_list)) {
2820 ifa = list_first_entry(&idev->addr_list,
2821 struct inet6_ifaddr, if_list);
2822 addrconf_del_timer(ifa);
2824 list_del(&ifa->if_list);
2826 write_unlock_bh(&idev->lock);
2828 spin_lock_bh(&ifa->state_lock);
2829 state = ifa->state;
2830 ifa->state = INET6_IFADDR_STATE_DEAD;
2831 spin_unlock_bh(&ifa->state_lock);
2833 if (state != INET6_IFADDR_STATE_DEAD) {
2834 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2835 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2837 in6_ifa_put(ifa);
2839 write_lock_bh(&idev->lock);
2842 write_unlock_bh(&idev->lock);
2844 /* Step 5: Discard multicast list */
2845 if (how)
2846 ipv6_mc_destroy_dev(idev);
2847 else
2848 ipv6_mc_down(idev);
2850 idev->tstamp = jiffies;
2852 /* Last: Shot the device (if unregistered) */
2853 if (how) {
2854 addrconf_sysctl_unregister(idev);
2855 neigh_parms_release(&nd_tbl, idev->nd_parms);
2856 neigh_ifdown(&nd_tbl, dev);
2857 in6_dev_put(idev);
2859 return 0;
2862 static void addrconf_rs_timer(unsigned long data)
2864 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2865 struct inet6_dev *idev = ifp->idev;
2867 read_lock(&idev->lock);
2868 if (idev->dead || !(idev->if_flags & IF_READY))
2869 goto out;
2871 if (idev->cnf.forwarding)
2872 goto out;
2874 /* Announcement received after solicitation was sent */
2875 if (idev->if_flags & IF_RA_RCVD)
2876 goto out;
2878 spin_lock(&ifp->lock);
2879 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2880 /* The wait after the last probe can be shorter */
2881 addrconf_mod_timer(ifp, AC_RS,
2882 (ifp->probes == idev->cnf.rtr_solicits) ?
2883 idev->cnf.rtr_solicit_delay :
2884 idev->cnf.rtr_solicit_interval);
2885 spin_unlock(&ifp->lock);
2887 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2888 } else {
2889 spin_unlock(&ifp->lock);
2891 * Note: we do not support deprecated "all on-link"
2892 * assumption any longer.
2894 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2895 idev->dev->name);
2898 out:
2899 read_unlock(&idev->lock);
2900 in6_ifa_put(ifp);
2904 * Duplicate Address Detection
2906 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2908 unsigned long rand_num;
2909 struct inet6_dev *idev = ifp->idev;
2911 if (ifp->flags & IFA_F_OPTIMISTIC)
2912 rand_num = 0;
2913 else
2914 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2916 ifp->probes = idev->cnf.dad_transmits;
2917 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2920 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2922 struct inet6_dev *idev = ifp->idev;
2923 struct net_device *dev = idev->dev;
2925 addrconf_join_solict(dev, &ifp->addr);
2927 net_srandom(ifp->addr.s6_addr32[3]);
2929 read_lock_bh(&idev->lock);
2930 spin_lock(&ifp->lock);
2931 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2932 goto out;
2934 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2935 idev->cnf.accept_dad < 1 ||
2936 !(ifp->flags&IFA_F_TENTATIVE) ||
2937 ifp->flags & IFA_F_NODAD) {
2938 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2939 spin_unlock(&ifp->lock);
2940 read_unlock_bh(&idev->lock);
2942 addrconf_dad_completed(ifp);
2943 return;
2946 if (!(idev->if_flags & IF_READY)) {
2947 spin_unlock(&ifp->lock);
2948 read_unlock_bh(&idev->lock);
2950 * If the device is not ready:
2951 * - keep it tentative if it is a permanent address.
2952 * - otherwise, kill it.
2954 in6_ifa_hold(ifp);
2955 addrconf_dad_stop(ifp, 0);
2956 return;
2960 * Optimistic nodes can start receiving
2961 * Frames right away
2963 if (ifp->flags & IFA_F_OPTIMISTIC)
2964 ip6_ins_rt(ifp->rt);
2966 addrconf_dad_kick(ifp);
2967 out:
2968 spin_unlock(&ifp->lock);
2969 read_unlock_bh(&idev->lock);
2972 static void addrconf_dad_timer(unsigned long data)
2974 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2975 struct inet6_dev *idev = ifp->idev;
2976 struct in6_addr mcaddr;
2978 if (!ifp->probes && addrconf_dad_end(ifp))
2979 goto out;
2981 read_lock(&idev->lock);
2982 if (idev->dead || !(idev->if_flags & IF_READY)) {
2983 read_unlock(&idev->lock);
2984 goto out;
2987 spin_lock(&ifp->lock);
2988 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2989 spin_unlock(&ifp->lock);
2990 read_unlock(&idev->lock);
2991 goto out;
2994 if (ifp->probes == 0) {
2996 * DAD was successful
2999 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3000 spin_unlock(&ifp->lock);
3001 read_unlock(&idev->lock);
3003 addrconf_dad_completed(ifp);
3005 goto out;
3008 ifp->probes--;
3009 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3010 spin_unlock(&ifp->lock);
3011 read_unlock(&idev->lock);
3013 /* send a neighbour solicitation for our addr */
3014 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3015 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3016 out:
3017 in6_ifa_put(ifp);
3020 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3022 struct net_device *dev = ifp->idev->dev;
3025 * Configure the address for reception. Now it is valid.
3028 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3030 /* If added prefix is link local and we are prepared to process
3031 router advertisements, start sending router solicitations.
3034 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3035 ifp->idev->cnf.accept_ra == 2) &&
3036 ifp->idev->cnf.rtr_solicits > 0 &&
3037 (dev->flags&IFF_LOOPBACK) == 0 &&
3038 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3040 * If a host as already performed a random delay
3041 * [...] as part of DAD [...] there is no need
3042 * to delay again before sending the first RS
3044 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3046 spin_lock_bh(&ifp->lock);
3047 ifp->probes = 1;
3048 ifp->idev->if_flags |= IF_RS_SENT;
3049 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3050 spin_unlock_bh(&ifp->lock);
3054 static void addrconf_dad_run(struct inet6_dev *idev)
3056 struct inet6_ifaddr *ifp;
3058 read_lock_bh(&idev->lock);
3059 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3060 spin_lock(&ifp->lock);
3061 if (ifp->flags & IFA_F_TENTATIVE &&
3062 ifp->state == INET6_IFADDR_STATE_DAD)
3063 addrconf_dad_kick(ifp);
3064 spin_unlock(&ifp->lock);
3066 read_unlock_bh(&idev->lock);
3069 #ifdef CONFIG_PROC_FS
3070 struct if6_iter_state {
3071 struct seq_net_private p;
3072 int bucket;
3073 int offset;
3076 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3078 struct inet6_ifaddr *ifa = NULL;
3079 struct if6_iter_state *state = seq->private;
3080 struct net *net = seq_file_net(seq);
3081 int p = 0;
3083 /* initial bucket if pos is 0 */
3084 if (pos == 0) {
3085 state->bucket = 0;
3086 state->offset = 0;
3089 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3090 struct hlist_node *n;
3091 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3092 addr_lst) {
3093 /* sync with offset */
3094 if (p < state->offset) {
3095 p++;
3096 continue;
3098 state->offset++;
3099 if (net_eq(dev_net(ifa->idev->dev), net))
3100 return ifa;
3103 /* prepare for next bucket */
3104 state->offset = 0;
3105 p = 0;
3107 return NULL;
3110 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3111 struct inet6_ifaddr *ifa)
3113 struct if6_iter_state *state = seq->private;
3114 struct net *net = seq_file_net(seq);
3115 struct hlist_node *n = &ifa->addr_lst;
3117 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3118 state->offset++;
3119 if (net_eq(dev_net(ifa->idev->dev), net))
3120 return ifa;
3123 while (++state->bucket < IN6_ADDR_HSIZE) {
3124 state->offset = 0;
3125 hlist_for_each_entry_rcu_bh(ifa, n,
3126 &inet6_addr_lst[state->bucket], addr_lst) {
3127 state->offset++;
3128 if (net_eq(dev_net(ifa->idev->dev), net))
3129 return ifa;
3133 return NULL;
3136 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3137 __acquires(rcu_bh)
3139 rcu_read_lock_bh();
3140 return if6_get_first(seq, *pos);
3143 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3145 struct inet6_ifaddr *ifa;
3147 ifa = if6_get_next(seq, v);
3148 ++*pos;
3149 return ifa;
3152 static void if6_seq_stop(struct seq_file *seq, void *v)
3153 __releases(rcu_bh)
3155 rcu_read_unlock_bh();
3158 static int if6_seq_show(struct seq_file *seq, void *v)
3160 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3161 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3162 &ifp->addr,
3163 ifp->idev->dev->ifindex,
3164 ifp->prefix_len,
3165 ifp->scope,
3166 ifp->flags,
3167 ifp->idev->dev->name);
3168 return 0;
3171 static const struct seq_operations if6_seq_ops = {
3172 .start = if6_seq_start,
3173 .next = if6_seq_next,
3174 .show = if6_seq_show,
3175 .stop = if6_seq_stop,
3178 static int if6_seq_open(struct inode *inode, struct file *file)
3180 return seq_open_net(inode, file, &if6_seq_ops,
3181 sizeof(struct if6_iter_state));
3184 static const struct file_operations if6_fops = {
3185 .owner = THIS_MODULE,
3186 .open = if6_seq_open,
3187 .read = seq_read,
3188 .llseek = seq_lseek,
3189 .release = seq_release_net,
3192 static int __net_init if6_proc_net_init(struct net *net)
3194 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3195 return -ENOMEM;
3196 return 0;
3199 static void __net_exit if6_proc_net_exit(struct net *net)
3201 proc_net_remove(net, "if_inet6");
3204 static struct pernet_operations if6_proc_net_ops = {
3205 .init = if6_proc_net_init,
3206 .exit = if6_proc_net_exit,
3209 int __init if6_proc_init(void)
3211 return register_pernet_subsys(&if6_proc_net_ops);
3214 void if6_proc_exit(void)
3216 unregister_pernet_subsys(&if6_proc_net_ops);
3218 #endif /* CONFIG_PROC_FS */
3220 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3221 /* Check if address is a home address configured on any interface. */
3222 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3224 int ret = 0;
3225 struct inet6_ifaddr *ifp = NULL;
3226 struct hlist_node *n;
3227 unsigned int hash = ipv6_addr_hash(addr);
3229 rcu_read_lock_bh();
3230 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3231 if (!net_eq(dev_net(ifp->idev->dev), net))
3232 continue;
3233 if (ipv6_addr_equal(&ifp->addr, addr) &&
3234 (ifp->flags & IFA_F_HOMEADDRESS)) {
3235 ret = 1;
3236 break;
3239 rcu_read_unlock_bh();
3240 return ret;
3242 #endif
3245 * Periodic address status verification
3248 static void addrconf_verify(unsigned long foo)
3250 unsigned long now, next, next_sec, next_sched;
3251 struct inet6_ifaddr *ifp;
3252 struct hlist_node *node;
3253 int i;
3255 rcu_read_lock_bh();
3256 spin_lock(&addrconf_verify_lock);
3257 now = jiffies;
3258 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3260 del_timer(&addr_chk_timer);
3262 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3263 restart:
3264 hlist_for_each_entry_rcu_bh(ifp, node,
3265 &inet6_addr_lst[i], addr_lst) {
3266 unsigned long age;
3268 if (ifp->flags & IFA_F_PERMANENT)
3269 continue;
3271 spin_lock(&ifp->lock);
3272 /* We try to batch several events at once. */
3273 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3275 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3276 age >= ifp->valid_lft) {
3277 spin_unlock(&ifp->lock);
3278 in6_ifa_hold(ifp);
3279 ipv6_del_addr(ifp);
3280 goto restart;
3281 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3282 spin_unlock(&ifp->lock);
3283 continue;
3284 } else if (age >= ifp->prefered_lft) {
3285 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3286 int deprecate = 0;
3288 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3289 deprecate = 1;
3290 ifp->flags |= IFA_F_DEPRECATED;
3293 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3294 next = ifp->tstamp + ifp->valid_lft * HZ;
3296 spin_unlock(&ifp->lock);
3298 if (deprecate) {
3299 in6_ifa_hold(ifp);
3301 ipv6_ifa_notify(0, ifp);
3302 in6_ifa_put(ifp);
3303 goto restart;
3305 #ifdef CONFIG_IPV6_PRIVACY
3306 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3307 !(ifp->flags&IFA_F_TENTATIVE)) {
3308 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3309 ifp->idev->cnf.dad_transmits *
3310 ifp->idev->nd_parms->retrans_time / HZ;
3312 if (age >= ifp->prefered_lft - regen_advance) {
3313 struct inet6_ifaddr *ifpub = ifp->ifpub;
3314 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3315 next = ifp->tstamp + ifp->prefered_lft * HZ;
3316 if (!ifp->regen_count && ifpub) {
3317 ifp->regen_count++;
3318 in6_ifa_hold(ifp);
3319 in6_ifa_hold(ifpub);
3320 spin_unlock(&ifp->lock);
3322 spin_lock(&ifpub->lock);
3323 ifpub->regen_count = 0;
3324 spin_unlock(&ifpub->lock);
3325 ipv6_create_tempaddr(ifpub, ifp);
3326 in6_ifa_put(ifpub);
3327 in6_ifa_put(ifp);
3328 goto restart;
3330 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3331 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3332 spin_unlock(&ifp->lock);
3333 #endif
3334 } else {
3335 /* ifp->prefered_lft <= ifp->valid_lft */
3336 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3337 next = ifp->tstamp + ifp->prefered_lft * HZ;
3338 spin_unlock(&ifp->lock);
3343 next_sec = round_jiffies_up(next);
3344 next_sched = next;
3346 /* If rounded timeout is accurate enough, accept it. */
3347 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3348 next_sched = next_sec;
3350 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3351 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3352 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3354 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3355 now, next, next_sec, next_sched));
3357 addr_chk_timer.expires = next_sched;
3358 add_timer(&addr_chk_timer);
3359 spin_unlock(&addrconf_verify_lock);
3360 rcu_read_unlock_bh();
3363 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3365 struct in6_addr *pfx = NULL;
3367 if (addr)
3368 pfx = nla_data(addr);
3370 if (local) {
3371 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3372 pfx = NULL;
3373 else
3374 pfx = nla_data(local);
3377 return pfx;
3380 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3381 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3382 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3383 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3386 static int
3387 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3389 struct net *net = sock_net(skb->sk);
3390 struct ifaddrmsg *ifm;
3391 struct nlattr *tb[IFA_MAX+1];
3392 struct in6_addr *pfx;
3393 int err;
3395 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3396 if (err < 0)
3397 return err;
3399 ifm = nlmsg_data(nlh);
3400 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3401 if (pfx == NULL)
3402 return -EINVAL;
3404 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3407 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3408 u32 prefered_lft, u32 valid_lft)
3410 u32 flags;
3411 clock_t expires;
3412 unsigned long timeout;
3414 if (!valid_lft || (prefered_lft > valid_lft))
3415 return -EINVAL;
3417 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3418 if (addrconf_finite_timeout(timeout)) {
3419 expires = jiffies_to_clock_t(timeout * HZ);
3420 valid_lft = timeout;
3421 flags = RTF_EXPIRES;
3422 } else {
3423 expires = 0;
3424 flags = 0;
3425 ifa_flags |= IFA_F_PERMANENT;
3428 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3429 if (addrconf_finite_timeout(timeout)) {
3430 if (timeout == 0)
3431 ifa_flags |= IFA_F_DEPRECATED;
3432 prefered_lft = timeout;
3435 spin_lock_bh(&ifp->lock);
3436 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3437 ifp->tstamp = jiffies;
3438 ifp->valid_lft = valid_lft;
3439 ifp->prefered_lft = prefered_lft;
3441 spin_unlock_bh(&ifp->lock);
3442 if (!(ifp->flags&IFA_F_TENTATIVE))
3443 ipv6_ifa_notify(0, ifp);
3445 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3446 expires, flags);
3447 addrconf_verify(0);
3449 return 0;
3452 static int
3453 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3455 struct net *net = sock_net(skb->sk);
3456 struct ifaddrmsg *ifm;
3457 struct nlattr *tb[IFA_MAX+1];
3458 struct in6_addr *pfx;
3459 struct inet6_ifaddr *ifa;
3460 struct net_device *dev;
3461 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3462 u8 ifa_flags;
3463 int err;
3465 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3466 if (err < 0)
3467 return err;
3469 ifm = nlmsg_data(nlh);
3470 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3471 if (pfx == NULL)
3472 return -EINVAL;
3474 if (tb[IFA_CACHEINFO]) {
3475 struct ifa_cacheinfo *ci;
3477 ci = nla_data(tb[IFA_CACHEINFO]);
3478 valid_lft = ci->ifa_valid;
3479 preferred_lft = ci->ifa_prefered;
3480 } else {
3481 preferred_lft = INFINITY_LIFE_TIME;
3482 valid_lft = INFINITY_LIFE_TIME;
3485 dev = __dev_get_by_index(net, ifm->ifa_index);
3486 if (dev == NULL)
3487 return -ENODEV;
3489 /* We ignore other flags so far. */
3490 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3492 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3493 if (ifa == NULL) {
3495 * It would be best to check for !NLM_F_CREATE here but
3496 * userspace alreay relies on not having to provide this.
3498 return inet6_addr_add(net, ifm->ifa_index, pfx,
3499 ifm->ifa_prefixlen, ifa_flags,
3500 preferred_lft, valid_lft);
3503 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3504 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3505 err = -EEXIST;
3506 else
3507 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3509 in6_ifa_put(ifa);
3511 return err;
3514 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3515 u8 scope, int ifindex)
3517 struct ifaddrmsg *ifm;
3519 ifm = nlmsg_data(nlh);
3520 ifm->ifa_family = AF_INET6;
3521 ifm->ifa_prefixlen = prefixlen;
3522 ifm->ifa_flags = flags;
3523 ifm->ifa_scope = scope;
3524 ifm->ifa_index = ifindex;
3527 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3528 unsigned long tstamp, u32 preferred, u32 valid)
3530 struct ifa_cacheinfo ci;
3532 ci.cstamp = cstamp_delta(cstamp);
3533 ci.tstamp = cstamp_delta(tstamp);
3534 ci.ifa_prefered = preferred;
3535 ci.ifa_valid = valid;
3537 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3540 static inline int rt_scope(int ifa_scope)
3542 if (ifa_scope & IFA_HOST)
3543 return RT_SCOPE_HOST;
3544 else if (ifa_scope & IFA_LINK)
3545 return RT_SCOPE_LINK;
3546 else if (ifa_scope & IFA_SITE)
3547 return RT_SCOPE_SITE;
3548 else
3549 return RT_SCOPE_UNIVERSE;
3552 static inline int inet6_ifaddr_msgsize(void)
3554 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3555 + nla_total_size(16) /* IFA_ADDRESS */
3556 + nla_total_size(sizeof(struct ifa_cacheinfo));
3559 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3560 u32 pid, u32 seq, int event, unsigned int flags)
3562 struct nlmsghdr *nlh;
3563 u32 preferred, valid;
3565 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3566 if (nlh == NULL)
3567 return -EMSGSIZE;
3569 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3570 ifa->idev->dev->ifindex);
3572 if (!(ifa->flags&IFA_F_PERMANENT)) {
3573 preferred = ifa->prefered_lft;
3574 valid = ifa->valid_lft;
3575 if (preferred != INFINITY_LIFE_TIME) {
3576 long tval = (jiffies - ifa->tstamp)/HZ;
3577 if (preferred > tval)
3578 preferred -= tval;
3579 else
3580 preferred = 0;
3581 if (valid != INFINITY_LIFE_TIME) {
3582 if (valid > tval)
3583 valid -= tval;
3584 else
3585 valid = 0;
3588 } else {
3589 preferred = INFINITY_LIFE_TIME;
3590 valid = INFINITY_LIFE_TIME;
3593 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3594 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3595 nlmsg_cancel(skb, nlh);
3596 return -EMSGSIZE;
3599 return nlmsg_end(skb, nlh);
3602 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3603 u32 pid, u32 seq, int event, u16 flags)
3605 struct nlmsghdr *nlh;
3606 u8 scope = RT_SCOPE_UNIVERSE;
3607 int ifindex = ifmca->idev->dev->ifindex;
3609 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3610 scope = RT_SCOPE_SITE;
3612 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3613 if (nlh == NULL)
3614 return -EMSGSIZE;
3616 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3617 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3618 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3619 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3620 nlmsg_cancel(skb, nlh);
3621 return -EMSGSIZE;
3624 return nlmsg_end(skb, nlh);
3627 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3628 u32 pid, u32 seq, int event, unsigned int flags)
3630 struct nlmsghdr *nlh;
3631 u8 scope = RT_SCOPE_UNIVERSE;
3632 int ifindex = ifaca->aca_idev->dev->ifindex;
3634 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3635 scope = RT_SCOPE_SITE;
3637 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3638 if (nlh == NULL)
3639 return -EMSGSIZE;
3641 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3642 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3643 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3644 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3645 nlmsg_cancel(skb, nlh);
3646 return -EMSGSIZE;
3649 return nlmsg_end(skb, nlh);
3652 enum addr_type_t {
3653 UNICAST_ADDR,
3654 MULTICAST_ADDR,
3655 ANYCAST_ADDR,
3658 /* called with rcu_read_lock() */
3659 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3660 struct netlink_callback *cb, enum addr_type_t type,
3661 int s_ip_idx, int *p_ip_idx)
3663 struct ifmcaddr6 *ifmca;
3664 struct ifacaddr6 *ifaca;
3665 int err = 1;
3666 int ip_idx = *p_ip_idx;
3668 read_lock_bh(&idev->lock);
3669 switch (type) {
3670 case UNICAST_ADDR: {
3671 struct inet6_ifaddr *ifa;
3673 /* unicast address incl. temp addr */
3674 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3675 if (++ip_idx < s_ip_idx)
3676 continue;
3677 err = inet6_fill_ifaddr(skb, ifa,
3678 NETLINK_CB(cb->skb).pid,
3679 cb->nlh->nlmsg_seq,
3680 RTM_NEWADDR,
3681 NLM_F_MULTI);
3682 if (err <= 0)
3683 break;
3685 break;
3687 case MULTICAST_ADDR:
3688 /* multicast address */
3689 for (ifmca = idev->mc_list; ifmca;
3690 ifmca = ifmca->next, ip_idx++) {
3691 if (ip_idx < s_ip_idx)
3692 continue;
3693 err = inet6_fill_ifmcaddr(skb, ifmca,
3694 NETLINK_CB(cb->skb).pid,
3695 cb->nlh->nlmsg_seq,
3696 RTM_GETMULTICAST,
3697 NLM_F_MULTI);
3698 if (err <= 0)
3699 break;
3701 break;
3702 case ANYCAST_ADDR:
3703 /* anycast address */
3704 for (ifaca = idev->ac_list; ifaca;
3705 ifaca = ifaca->aca_next, ip_idx++) {
3706 if (ip_idx < s_ip_idx)
3707 continue;
3708 err = inet6_fill_ifacaddr(skb, ifaca,
3709 NETLINK_CB(cb->skb).pid,
3710 cb->nlh->nlmsg_seq,
3711 RTM_GETANYCAST,
3712 NLM_F_MULTI);
3713 if (err <= 0)
3714 break;
3716 break;
3717 default:
3718 break;
3720 read_unlock_bh(&idev->lock);
3721 *p_ip_idx = ip_idx;
3722 return err;
3725 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3726 enum addr_type_t type)
3728 struct net *net = sock_net(skb->sk);
3729 int h, s_h;
3730 int idx, ip_idx;
3731 int s_idx, s_ip_idx;
3732 struct net_device *dev;
3733 struct inet6_dev *idev;
3734 struct hlist_head *head;
3735 struct hlist_node *node;
3737 s_h = cb->args[0];
3738 s_idx = idx = cb->args[1];
3739 s_ip_idx = ip_idx = cb->args[2];
3741 rcu_read_lock();
3742 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3743 idx = 0;
3744 head = &net->dev_index_head[h];
3745 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3746 if (idx < s_idx)
3747 goto cont;
3748 if (h > s_h || idx > s_idx)
3749 s_ip_idx = 0;
3750 ip_idx = 0;
3751 idev = __in6_dev_get(dev);
3752 if (!idev)
3753 goto cont;
3755 if (in6_dump_addrs(idev, skb, cb, type,
3756 s_ip_idx, &ip_idx) <= 0)
3757 goto done;
3758 cont:
3759 idx++;
3762 done:
3763 rcu_read_unlock();
3764 cb->args[0] = h;
3765 cb->args[1] = idx;
3766 cb->args[2] = ip_idx;
3768 return skb->len;
3771 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3773 enum addr_type_t type = UNICAST_ADDR;
3775 return inet6_dump_addr(skb, cb, type);
3778 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3780 enum addr_type_t type = MULTICAST_ADDR;
3782 return inet6_dump_addr(skb, cb, type);
3786 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3788 enum addr_type_t type = ANYCAST_ADDR;
3790 return inet6_dump_addr(skb, cb, type);
3793 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3794 void *arg)
3796 struct net *net = sock_net(in_skb->sk);
3797 struct ifaddrmsg *ifm;
3798 struct nlattr *tb[IFA_MAX+1];
3799 struct in6_addr *addr = NULL;
3800 struct net_device *dev = NULL;
3801 struct inet6_ifaddr *ifa;
3802 struct sk_buff *skb;
3803 int err;
3805 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3806 if (err < 0)
3807 goto errout;
3809 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3810 if (addr == NULL) {
3811 err = -EINVAL;
3812 goto errout;
3815 ifm = nlmsg_data(nlh);
3816 if (ifm->ifa_index)
3817 dev = __dev_get_by_index(net, ifm->ifa_index);
3819 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3820 if (!ifa) {
3821 err = -EADDRNOTAVAIL;
3822 goto errout;
3825 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3826 if (!skb) {
3827 err = -ENOBUFS;
3828 goto errout_ifa;
3831 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3832 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3833 if (err < 0) {
3834 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3835 WARN_ON(err == -EMSGSIZE);
3836 kfree_skb(skb);
3837 goto errout_ifa;
3839 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3840 errout_ifa:
3841 in6_ifa_put(ifa);
3842 errout:
3843 return err;
3846 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3848 struct sk_buff *skb;
3849 struct net *net = dev_net(ifa->idev->dev);
3850 int err = -ENOBUFS;
3852 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3853 if (skb == NULL)
3854 goto errout;
3856 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3857 if (err < 0) {
3858 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3859 WARN_ON(err == -EMSGSIZE);
3860 kfree_skb(skb);
3861 goto errout;
3863 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3864 return;
3865 errout:
3866 if (err < 0)
3867 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3870 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3871 __s32 *array, int bytes)
3873 BUG_ON(bytes < (DEVCONF_MAX * 4));
3875 memset(array, 0, bytes);
3876 array[DEVCONF_FORWARDING] = cnf->forwarding;
3877 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3878 array[DEVCONF_MTU6] = cnf->mtu6;
3879 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3880 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3881 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3882 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3883 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3884 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3885 jiffies_to_msecs(cnf->rtr_solicit_interval);
3886 array[DEVCONF_RTR_SOLICIT_DELAY] =
3887 jiffies_to_msecs(cnf->rtr_solicit_delay);
3888 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3889 #ifdef CONFIG_IPV6_PRIVACY
3890 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3891 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3892 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3893 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3894 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3895 #endif
3896 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3897 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3898 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3899 #ifdef CONFIG_IPV6_ROUTER_PREF
3900 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3901 array[DEVCONF_RTR_PROBE_INTERVAL] =
3902 jiffies_to_msecs(cnf->rtr_probe_interval);
3903 #ifdef CONFIG_IPV6_ROUTE_INFO
3904 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3905 #endif
3906 #endif
3907 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3908 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3909 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3910 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3911 #endif
3912 #ifdef CONFIG_IPV6_MROUTE
3913 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3914 #endif
3915 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3916 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3917 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3920 static inline size_t inet6_ifla6_size(void)
3922 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3923 + nla_total_size(sizeof(struct ifla_cacheinfo))
3924 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3925 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3926 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3929 static inline size_t inet6_if_nlmsg_size(void)
3931 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3932 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3933 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3934 + nla_total_size(4) /* IFLA_MTU */
3935 + nla_total_size(4) /* IFLA_LINK */
3936 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3939 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3940 int items, int bytes)
3942 int i;
3943 int pad = bytes - sizeof(u64) * items;
3944 BUG_ON(pad < 0);
3946 /* Use put_unaligned() because stats may not be aligned for u64. */
3947 put_unaligned(items, &stats[0]);
3948 for (i = 1; i < items; i++)
3949 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3951 memset(&stats[items], 0, pad);
3954 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3955 int items, int bytes, size_t syncpoff)
3957 int i;
3958 int pad = bytes - sizeof(u64) * items;
3959 BUG_ON(pad < 0);
3961 /* Use put_unaligned() because stats may not be aligned for u64. */
3962 put_unaligned(items, &stats[0]);
3963 for (i = 1; i < items; i++)
3964 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3966 memset(&stats[items], 0, pad);
3969 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3970 int bytes)
3972 switch (attrtype) {
3973 case IFLA_INET6_STATS:
3974 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3975 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3976 break;
3977 case IFLA_INET6_ICMP6STATS:
3978 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3979 break;
3983 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3985 struct nlattr *nla;
3986 struct ifla_cacheinfo ci;
3988 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3990 ci.max_reasm_len = IPV6_MAXPLEN;
3991 ci.tstamp = cstamp_delta(idev->tstamp);
3992 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3993 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3994 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3996 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3997 if (nla == NULL)
3998 goto nla_put_failure;
3999 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4001 /* XXX - MC not implemented */
4003 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4004 if (nla == NULL)
4005 goto nla_put_failure;
4006 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4008 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4009 if (nla == NULL)
4010 goto nla_put_failure;
4011 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4013 return 0;
4015 nla_put_failure:
4016 return -EMSGSIZE;
4019 static size_t inet6_get_link_af_size(const struct net_device *dev)
4021 if (!__in6_dev_get(dev))
4022 return 0;
4024 return inet6_ifla6_size();
4027 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4029 struct inet6_dev *idev = __in6_dev_get(dev);
4031 if (!idev)
4032 return -ENODATA;
4034 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4035 return -EMSGSIZE;
4037 return 0;
4040 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4041 u32 pid, u32 seq, int event, unsigned int flags)
4043 struct net_device *dev = idev->dev;
4044 struct ifinfomsg *hdr;
4045 struct nlmsghdr *nlh;
4046 void *protoinfo;
4048 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4049 if (nlh == NULL)
4050 return -EMSGSIZE;
4052 hdr = nlmsg_data(nlh);
4053 hdr->ifi_family = AF_INET6;
4054 hdr->__ifi_pad = 0;
4055 hdr->ifi_type = dev->type;
4056 hdr->ifi_index = dev->ifindex;
4057 hdr->ifi_flags = dev_get_flags(dev);
4058 hdr->ifi_change = 0;
4060 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4062 if (dev->addr_len)
4063 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4065 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4066 if (dev->ifindex != dev->iflink)
4067 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4069 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4070 if (protoinfo == NULL)
4071 goto nla_put_failure;
4073 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4074 goto nla_put_failure;
4076 nla_nest_end(skb, protoinfo);
4077 return nlmsg_end(skb, nlh);
4079 nla_put_failure:
4080 nlmsg_cancel(skb, nlh);
4081 return -EMSGSIZE;
4084 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4086 struct net *net = sock_net(skb->sk);
4087 int h, s_h;
4088 int idx = 0, s_idx;
4089 struct net_device *dev;
4090 struct inet6_dev *idev;
4091 struct hlist_head *head;
4092 struct hlist_node *node;
4094 s_h = cb->args[0];
4095 s_idx = cb->args[1];
4097 rcu_read_lock();
4098 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4099 idx = 0;
4100 head = &net->dev_index_head[h];
4101 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4102 if (idx < s_idx)
4103 goto cont;
4104 idev = __in6_dev_get(dev);
4105 if (!idev)
4106 goto cont;
4107 if (inet6_fill_ifinfo(skb, idev,
4108 NETLINK_CB(cb->skb).pid,
4109 cb->nlh->nlmsg_seq,
4110 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4111 goto out;
4112 cont:
4113 idx++;
4116 out:
4117 rcu_read_unlock();
4118 cb->args[1] = idx;
4119 cb->args[0] = h;
4121 return skb->len;
4124 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4126 struct sk_buff *skb;
4127 struct net *net = dev_net(idev->dev);
4128 int err = -ENOBUFS;
4130 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4131 if (skb == NULL)
4132 goto errout;
4134 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4135 if (err < 0) {
4136 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4137 WARN_ON(err == -EMSGSIZE);
4138 kfree_skb(skb);
4139 goto errout;
4141 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4142 return;
4143 errout:
4144 if (err < 0)
4145 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4148 static inline size_t inet6_prefix_nlmsg_size(void)
4150 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4151 + nla_total_size(sizeof(struct in6_addr))
4152 + nla_total_size(sizeof(struct prefix_cacheinfo));
4155 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4156 struct prefix_info *pinfo, u32 pid, u32 seq,
4157 int event, unsigned int flags)
4159 struct prefixmsg *pmsg;
4160 struct nlmsghdr *nlh;
4161 struct prefix_cacheinfo ci;
4163 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4164 if (nlh == NULL)
4165 return -EMSGSIZE;
4167 pmsg = nlmsg_data(nlh);
4168 pmsg->prefix_family = AF_INET6;
4169 pmsg->prefix_pad1 = 0;
4170 pmsg->prefix_pad2 = 0;
4171 pmsg->prefix_ifindex = idev->dev->ifindex;
4172 pmsg->prefix_len = pinfo->prefix_len;
4173 pmsg->prefix_type = pinfo->type;
4174 pmsg->prefix_pad3 = 0;
4175 pmsg->prefix_flags = 0;
4176 if (pinfo->onlink)
4177 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4178 if (pinfo->autoconf)
4179 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4181 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4183 ci.preferred_time = ntohl(pinfo->prefered);
4184 ci.valid_time = ntohl(pinfo->valid);
4185 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4187 return nlmsg_end(skb, nlh);
4189 nla_put_failure:
4190 nlmsg_cancel(skb, nlh);
4191 return -EMSGSIZE;
4194 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4195 struct prefix_info *pinfo)
4197 struct sk_buff *skb;
4198 struct net *net = dev_net(idev->dev);
4199 int err = -ENOBUFS;
4201 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4202 if (skb == NULL)
4203 goto errout;
4205 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4206 if (err < 0) {
4207 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4208 WARN_ON(err == -EMSGSIZE);
4209 kfree_skb(skb);
4210 goto errout;
4212 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4213 return;
4214 errout:
4215 if (err < 0)
4216 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4219 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4221 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4223 switch (event) {
4224 case RTM_NEWADDR:
4226 * If the address was optimistic
4227 * we inserted the route at the start of
4228 * our DAD process, so we don't need
4229 * to do it again
4231 if (!(ifp->rt->rt6i_node))
4232 ip6_ins_rt(ifp->rt);
4233 if (ifp->idev->cnf.forwarding)
4234 addrconf_join_anycast(ifp);
4235 break;
4236 case RTM_DELADDR:
4237 if (ifp->idev->cnf.forwarding)
4238 addrconf_leave_anycast(ifp);
4239 addrconf_leave_solict(ifp->idev, &ifp->addr);
4240 dst_hold(&ifp->rt->dst);
4242 if (ip6_del_rt(ifp->rt))
4243 dst_free(&ifp->rt->dst);
4244 break;
4248 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4250 rcu_read_lock_bh();
4251 if (likely(ifp->idev->dead == 0))
4252 __ipv6_ifa_notify(event, ifp);
4253 rcu_read_unlock_bh();
4256 #ifdef CONFIG_SYSCTL
4258 static
4259 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4260 void __user *buffer, size_t *lenp, loff_t *ppos)
4262 int *valp = ctl->data;
4263 int val = *valp;
4264 loff_t pos = *ppos;
4265 ctl_table lctl;
4266 int ret;
4269 * ctl->data points to idev->cnf.forwarding, we should
4270 * not modify it until we get the rtnl lock.
4272 lctl = *ctl;
4273 lctl.data = &val;
4275 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4277 if (write)
4278 ret = addrconf_fixup_forwarding(ctl, valp, val);
4279 if (ret)
4280 *ppos = pos;
4281 return ret;
4284 static void dev_disable_change(struct inet6_dev *idev)
4286 if (!idev || !idev->dev)
4287 return;
4289 if (idev->cnf.disable_ipv6)
4290 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4291 else
4292 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4295 static void addrconf_disable_change(struct net *net, __s32 newf)
4297 struct net_device *dev;
4298 struct inet6_dev *idev;
4300 rcu_read_lock();
4301 for_each_netdev_rcu(net, dev) {
4302 idev = __in6_dev_get(dev);
4303 if (idev) {
4304 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4305 idev->cnf.disable_ipv6 = newf;
4306 if (changed)
4307 dev_disable_change(idev);
4310 rcu_read_unlock();
4313 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4315 struct net *net;
4316 int old;
4318 if (!rtnl_trylock())
4319 return restart_syscall();
4321 net = (struct net *)table->extra2;
4322 old = *p;
4323 *p = newf;
4325 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4326 rtnl_unlock();
4327 return 0;
4330 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4331 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4332 addrconf_disable_change(net, newf);
4333 } else if ((!newf) ^ (!old))
4334 dev_disable_change((struct inet6_dev *)table->extra1);
4336 rtnl_unlock();
4337 return 0;
4340 static
4341 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4342 void __user *buffer, size_t *lenp, loff_t *ppos)
4344 int *valp = ctl->data;
4345 int val = *valp;
4346 loff_t pos = *ppos;
4347 ctl_table lctl;
4348 int ret;
4351 * ctl->data points to idev->cnf.disable_ipv6, we should
4352 * not modify it until we get the rtnl lock.
4354 lctl = *ctl;
4355 lctl.data = &val;
4357 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4359 if (write)
4360 ret = addrconf_disable_ipv6(ctl, valp, val);
4361 if (ret)
4362 *ppos = pos;
4363 return ret;
4366 static struct addrconf_sysctl_table
4368 struct ctl_table_header *sysctl_header;
4369 ctl_table addrconf_vars[DEVCONF_MAX+1];
4370 char *dev_name;
4371 } addrconf_sysctl __read_mostly = {
4372 .sysctl_header = NULL,
4373 .addrconf_vars = {
4375 .procname = "forwarding",
4376 .data = &ipv6_devconf.forwarding,
4377 .maxlen = sizeof(int),
4378 .mode = 0644,
4379 .proc_handler = addrconf_sysctl_forward,
4382 .procname = "hop_limit",
4383 .data = &ipv6_devconf.hop_limit,
4384 .maxlen = sizeof(int),
4385 .mode = 0644,
4386 .proc_handler = proc_dointvec,
4389 .procname = "mtu",
4390 .data = &ipv6_devconf.mtu6,
4391 .maxlen = sizeof(int),
4392 .mode = 0644,
4393 .proc_handler = proc_dointvec,
4396 .procname = "accept_ra",
4397 .data = &ipv6_devconf.accept_ra,
4398 .maxlen = sizeof(int),
4399 .mode = 0644,
4400 .proc_handler = proc_dointvec,
4403 .procname = "accept_redirects",
4404 .data = &ipv6_devconf.accept_redirects,
4405 .maxlen = sizeof(int),
4406 .mode = 0644,
4407 .proc_handler = proc_dointvec,
4410 .procname = "autoconf",
4411 .data = &ipv6_devconf.autoconf,
4412 .maxlen = sizeof(int),
4413 .mode = 0644,
4414 .proc_handler = proc_dointvec,
4417 .procname = "dad_transmits",
4418 .data = &ipv6_devconf.dad_transmits,
4419 .maxlen = sizeof(int),
4420 .mode = 0644,
4421 .proc_handler = proc_dointvec,
4424 .procname = "router_solicitations",
4425 .data = &ipv6_devconf.rtr_solicits,
4426 .maxlen = sizeof(int),
4427 .mode = 0644,
4428 .proc_handler = proc_dointvec,
4431 .procname = "router_solicitation_interval",
4432 .data = &ipv6_devconf.rtr_solicit_interval,
4433 .maxlen = sizeof(int),
4434 .mode = 0644,
4435 .proc_handler = proc_dointvec_jiffies,
4438 .procname = "router_solicitation_delay",
4439 .data = &ipv6_devconf.rtr_solicit_delay,
4440 .maxlen = sizeof(int),
4441 .mode = 0644,
4442 .proc_handler = proc_dointvec_jiffies,
4445 .procname = "force_mld_version",
4446 .data = &ipv6_devconf.force_mld_version,
4447 .maxlen = sizeof(int),
4448 .mode = 0644,
4449 .proc_handler = proc_dointvec,
4451 #ifdef CONFIG_IPV6_PRIVACY
4453 .procname = "use_tempaddr",
4454 .data = &ipv6_devconf.use_tempaddr,
4455 .maxlen = sizeof(int),
4456 .mode = 0644,
4457 .proc_handler = proc_dointvec,
4460 .procname = "temp_valid_lft",
4461 .data = &ipv6_devconf.temp_valid_lft,
4462 .maxlen = sizeof(int),
4463 .mode = 0644,
4464 .proc_handler = proc_dointvec,
4467 .procname = "temp_prefered_lft",
4468 .data = &ipv6_devconf.temp_prefered_lft,
4469 .maxlen = sizeof(int),
4470 .mode = 0644,
4471 .proc_handler = proc_dointvec,
4474 .procname = "regen_max_retry",
4475 .data = &ipv6_devconf.regen_max_retry,
4476 .maxlen = sizeof(int),
4477 .mode = 0644,
4478 .proc_handler = proc_dointvec,
4481 .procname = "max_desync_factor",
4482 .data = &ipv6_devconf.max_desync_factor,
4483 .maxlen = sizeof(int),
4484 .mode = 0644,
4485 .proc_handler = proc_dointvec,
4487 #endif
4489 .procname = "max_addresses",
4490 .data = &ipv6_devconf.max_addresses,
4491 .maxlen = sizeof(int),
4492 .mode = 0644,
4493 .proc_handler = proc_dointvec,
4496 .procname = "accept_ra_defrtr",
4497 .data = &ipv6_devconf.accept_ra_defrtr,
4498 .maxlen = sizeof(int),
4499 .mode = 0644,
4500 .proc_handler = proc_dointvec,
4503 .procname = "accept_ra_pinfo",
4504 .data = &ipv6_devconf.accept_ra_pinfo,
4505 .maxlen = sizeof(int),
4506 .mode = 0644,
4507 .proc_handler = proc_dointvec,
4509 #ifdef CONFIG_IPV6_ROUTER_PREF
4511 .procname = "accept_ra_rtr_pref",
4512 .data = &ipv6_devconf.accept_ra_rtr_pref,
4513 .maxlen = sizeof(int),
4514 .mode = 0644,
4515 .proc_handler = proc_dointvec,
4518 .procname = "router_probe_interval",
4519 .data = &ipv6_devconf.rtr_probe_interval,
4520 .maxlen = sizeof(int),
4521 .mode = 0644,
4522 .proc_handler = proc_dointvec_jiffies,
4524 #ifdef CONFIG_IPV6_ROUTE_INFO
4526 .procname = "accept_ra_rt_info_max_plen",
4527 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4528 .maxlen = sizeof(int),
4529 .mode = 0644,
4530 .proc_handler = proc_dointvec,
4532 #endif
4533 #endif
4535 .procname = "proxy_ndp",
4536 .data = &ipv6_devconf.proxy_ndp,
4537 .maxlen = sizeof(int),
4538 .mode = 0644,
4539 .proc_handler = proc_dointvec,
4542 .procname = "accept_source_route",
4543 .data = &ipv6_devconf.accept_source_route,
4544 .maxlen = sizeof(int),
4545 .mode = 0644,
4546 .proc_handler = proc_dointvec,
4548 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4550 .procname = "optimistic_dad",
4551 .data = &ipv6_devconf.optimistic_dad,
4552 .maxlen = sizeof(int),
4553 .mode = 0644,
4554 .proc_handler = proc_dointvec,
4557 #endif
4558 #ifdef CONFIG_IPV6_MROUTE
4560 .procname = "mc_forwarding",
4561 .data = &ipv6_devconf.mc_forwarding,
4562 .maxlen = sizeof(int),
4563 .mode = 0444,
4564 .proc_handler = proc_dointvec,
4566 #endif
4568 .procname = "disable_ipv6",
4569 .data = &ipv6_devconf.disable_ipv6,
4570 .maxlen = sizeof(int),
4571 .mode = 0644,
4572 .proc_handler = addrconf_sysctl_disable,
4575 .procname = "accept_dad",
4576 .data = &ipv6_devconf.accept_dad,
4577 .maxlen = sizeof(int),
4578 .mode = 0644,
4579 .proc_handler = proc_dointvec,
4582 .procname = "force_tllao",
4583 .data = &ipv6_devconf.force_tllao,
4584 .maxlen = sizeof(int),
4585 .mode = 0644,
4586 .proc_handler = proc_dointvec
4589 /* sentinel */
4594 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4595 struct inet6_dev *idev, struct ipv6_devconf *p)
4597 int i;
4598 struct addrconf_sysctl_table *t;
4600 #define ADDRCONF_CTL_PATH_DEV 3
4602 struct ctl_path addrconf_ctl_path[] = {
4603 { .procname = "net", },
4604 { .procname = "ipv6", },
4605 { .procname = "conf", },
4606 { /* to be set */ },
4607 { },
4611 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4612 if (t == NULL)
4613 goto out;
4615 for (i = 0; t->addrconf_vars[i].data; i++) {
4616 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4617 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4618 t->addrconf_vars[i].extra2 = net;
4622 * Make a copy of dev_name, because '.procname' is regarded as const
4623 * by sysctl and we wouldn't want anyone to change it under our feet
4624 * (see SIOCSIFNAME).
4626 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4627 if (!t->dev_name)
4628 goto free;
4630 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4632 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4633 t->addrconf_vars);
4634 if (t->sysctl_header == NULL)
4635 goto free_procname;
4637 p->sysctl = t;
4638 return 0;
4640 free_procname:
4641 kfree(t->dev_name);
4642 free:
4643 kfree(t);
4644 out:
4645 return -ENOBUFS;
4648 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4650 struct addrconf_sysctl_table *t;
4652 if (p->sysctl == NULL)
4653 return;
4655 t = p->sysctl;
4656 p->sysctl = NULL;
4657 unregister_net_sysctl_table(t->sysctl_header);
4658 kfree(t->dev_name);
4659 kfree(t);
4662 static void addrconf_sysctl_register(struct inet6_dev *idev)
4664 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4665 &ndisc_ifinfo_sysctl_change);
4666 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4667 idev, &idev->cnf);
4670 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4672 __addrconf_sysctl_unregister(&idev->cnf);
4673 neigh_sysctl_unregister(idev->nd_parms);
4677 #endif
4679 static int __net_init addrconf_init_net(struct net *net)
4681 int err;
4682 struct ipv6_devconf *all, *dflt;
4684 err = -ENOMEM;
4685 all = &ipv6_devconf;
4686 dflt = &ipv6_devconf_dflt;
4688 if (!net_eq(net, &init_net)) {
4689 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4690 if (all == NULL)
4691 goto err_alloc_all;
4693 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4694 if (dflt == NULL)
4695 goto err_alloc_dflt;
4696 } else {
4697 /* these will be inherited by all namespaces */
4698 dflt->autoconf = ipv6_defaults.autoconf;
4699 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4702 net->ipv6.devconf_all = all;
4703 net->ipv6.devconf_dflt = dflt;
4705 #ifdef CONFIG_SYSCTL
4706 err = __addrconf_sysctl_register(net, "all", NULL, all);
4707 if (err < 0)
4708 goto err_reg_all;
4710 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4711 if (err < 0)
4712 goto err_reg_dflt;
4713 #endif
4714 return 0;
4716 #ifdef CONFIG_SYSCTL
4717 err_reg_dflt:
4718 __addrconf_sysctl_unregister(all);
4719 err_reg_all:
4720 kfree(dflt);
4721 #endif
4722 err_alloc_dflt:
4723 kfree(all);
4724 err_alloc_all:
4725 return err;
4728 static void __net_exit addrconf_exit_net(struct net *net)
4730 #ifdef CONFIG_SYSCTL
4731 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4732 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4733 #endif
4734 if (!net_eq(net, &init_net)) {
4735 kfree(net->ipv6.devconf_dflt);
4736 kfree(net->ipv6.devconf_all);
4740 static struct pernet_operations addrconf_ops = {
4741 .init = addrconf_init_net,
4742 .exit = addrconf_exit_net,
4746 * Device notifier
4749 int register_inet6addr_notifier(struct notifier_block *nb)
4751 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4753 EXPORT_SYMBOL(register_inet6addr_notifier);
4755 int unregister_inet6addr_notifier(struct notifier_block *nb)
4757 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4759 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4761 static struct rtnl_af_ops inet6_ops = {
4762 .family = AF_INET6,
4763 .fill_link_af = inet6_fill_link_af,
4764 .get_link_af_size = inet6_get_link_af_size,
4768 * Init / cleanup code
4771 int __init addrconf_init(void)
4773 int i, err;
4775 err = ipv6_addr_label_init();
4776 if (err < 0) {
4777 printk(KERN_CRIT "IPv6 Addrconf:"
4778 " cannot initialize default policy table: %d.\n", err);
4779 goto out;
4782 err = register_pernet_subsys(&addrconf_ops);
4783 if (err < 0)
4784 goto out_addrlabel;
4786 /* The addrconf netdev notifier requires that loopback_dev
4787 * has it's ipv6 private information allocated and setup
4788 * before it can bring up and give link-local addresses
4789 * to other devices which are up.
4791 * Unfortunately, loopback_dev is not necessarily the first
4792 * entry in the global dev_base list of net devices. In fact,
4793 * it is likely to be the very last entry on that list.
4794 * So this causes the notifier registry below to try and
4795 * give link-local addresses to all devices besides loopback_dev
4796 * first, then loopback_dev, which cases all the non-loopback_dev
4797 * devices to fail to get a link-local address.
4799 * So, as a temporary fix, allocate the ipv6 structure for
4800 * loopback_dev first by hand.
4801 * Longer term, all of the dependencies ipv6 has upon the loopback
4802 * device and it being up should be removed.
4804 rtnl_lock();
4805 if (!ipv6_add_dev(init_net.loopback_dev))
4806 err = -ENOMEM;
4807 rtnl_unlock();
4808 if (err)
4809 goto errlo;
4811 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4812 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4814 register_netdevice_notifier(&ipv6_dev_notf);
4816 addrconf_verify(0);
4818 err = rtnl_af_register(&inet6_ops);
4819 if (err < 0)
4820 goto errout_af;
4822 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4823 NULL);
4824 if (err < 0)
4825 goto errout;
4827 /* Only the first call to __rtnl_register can fail */
4828 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4829 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4830 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4831 inet6_dump_ifaddr, NULL);
4832 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4833 inet6_dump_ifmcaddr, NULL);
4834 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4835 inet6_dump_ifacaddr, NULL);
4837 ipv6_addr_label_rtnl_register();
4839 return 0;
4840 errout:
4841 rtnl_af_unregister(&inet6_ops);
4842 errout_af:
4843 unregister_netdevice_notifier(&ipv6_dev_notf);
4844 errlo:
4845 unregister_pernet_subsys(&addrconf_ops);
4846 out_addrlabel:
4847 ipv6_addr_label_cleanup();
4848 out:
4849 return err;
4852 void addrconf_cleanup(void)
4854 struct net_device *dev;
4855 int i;
4857 unregister_netdevice_notifier(&ipv6_dev_notf);
4858 unregister_pernet_subsys(&addrconf_ops);
4859 ipv6_addr_label_cleanup();
4861 rtnl_lock();
4863 __rtnl_af_unregister(&inet6_ops);
4865 /* clean dev list */
4866 for_each_netdev(&init_net, dev) {
4867 if (__in6_dev_get(dev) == NULL)
4868 continue;
4869 addrconf_ifdown(dev, 1);
4871 addrconf_ifdown(init_net.loopback_dev, 2);
4874 * Check hash table.
4876 spin_lock_bh(&addrconf_hash_lock);
4877 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4878 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4879 spin_unlock_bh(&addrconf_hash_lock);
4881 del_timer(&addr_chk_timer);
4882 rtnl_unlock();