Merge git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched
[wrt350n-kernel.git] / net / ipv6 / addrconf.c
blob101e0e70ba276966a281b99601c17cfaf648f0c5
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 * $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
18 * Changes:
20 * Janos Farkas : delete timer on ifdown
21 * <chexum@bankinf.banki.hu>
22 * Andi Kleen : kill double kfree on module
23 * unload.
24 * Maciej W. Rozycki : FDDI support
25 * sekiya@USAGI : Don't send too many RS
26 * packets.
27 * yoshfuji@USAGI : Fixed interval between DAD
28 * packets.
29 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
30 * address validation timer.
31 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
32 * support.
33 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
34 * address on a same interface.
35 * YOSHIFUJI Hideaki @USAGI : ARCnet support
36 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
37 * seq_file.
38 * YOSHIFUJI Hideaki @USAGI : improved source address
39 * selection; consider scope,
40 * status etc.
43 #include <linux/errno.h>
44 #include <linux/types.h>
45 #include <linux/socket.h>
46 #include <linux/sockios.h>
47 #include <linux/net.h>
48 #include <linux/in6.h>
49 #include <linux/netdevice.h>
50 #include <linux/if_addr.h>
51 #include <linux/if_arp.h>
52 #include <linux/if_arcnet.h>
53 #include <linux/if_infiniband.h>
54 #include <linux/route.h>
55 #include <linux/inetdevice.h>
56 #include <linux/init.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 <asm/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
91 /* Set to 3 to get tracing... */
92 #define ACONF_DEBUG 2
94 #if ACONF_DEBUG >= 3
95 #define ADBG(x) printk x
96 #else
97 #define ADBG(x)
98 #endif
100 #define INFINITY_LIFE_TIME 0xFFFFFFFF
101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
103 #ifdef CONFIG_SYSCTL
104 static void addrconf_sysctl_register(struct inet6_dev *idev);
105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
106 #else
107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
114 #endif
116 #ifdef CONFIG_IPV6_PRIVACY
117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
119 static void ipv6_regen_rndid(unsigned long data);
121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
122 #endif
124 static int ipv6_count_addresses(struct inet6_dev *idev);
127 * Configured unicast address hash table
129 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
130 static DEFINE_RWLOCK(addrconf_hash_lock);
132 static void addrconf_verify(unsigned long);
134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
135 static DEFINE_SPINLOCK(addrconf_verify_lock);
137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
140 static int addrconf_ifdown(struct net_device *dev, int how);
142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
143 static void addrconf_dad_timer(unsigned long data);
144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
145 static void addrconf_dad_run(struct inet6_dev *idev);
146 static void addrconf_rs_timer(unsigned long data);
147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
151 struct prefix_info *pinfo);
152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
153 struct net_device *dev);
155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
157 struct ipv6_devconf ipv6_devconf __read_mostly = {
158 .forwarding = 0,
159 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
160 .mtu6 = IPV6_MIN_MTU,
161 .accept_ra = 1,
162 .accept_redirects = 1,
163 .autoconf = 1,
164 .force_mld_version = 0,
165 .dad_transmits = 1,
166 .rtr_solicits = MAX_RTR_SOLICITATIONS,
167 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
168 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
169 #ifdef CONFIG_IPV6_PRIVACY
170 .use_tempaddr = 0,
171 .temp_valid_lft = TEMP_VALID_LIFETIME,
172 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
173 .regen_max_retry = REGEN_MAX_RETRY,
174 .max_desync_factor = MAX_DESYNC_FACTOR,
175 #endif
176 .max_addresses = IPV6_MAX_ADDRESSES,
177 .accept_ra_defrtr = 1,
178 .accept_ra_pinfo = 1,
179 #ifdef CONFIG_IPV6_ROUTER_PREF
180 .accept_ra_rtr_pref = 1,
181 .rtr_probe_interval = 60 * HZ,
182 #ifdef CONFIG_IPV6_ROUTE_INFO
183 .accept_ra_rt_info_max_plen = 0,
184 #endif
185 #endif
186 .proxy_ndp = 0,
187 .accept_source_route = 0, /* we do not accept RH0 by default. */
190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
191 .forwarding = 0,
192 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
193 .mtu6 = IPV6_MIN_MTU,
194 .accept_ra = 1,
195 .accept_redirects = 1,
196 .autoconf = 1,
197 .dad_transmits = 1,
198 .rtr_solicits = MAX_RTR_SOLICITATIONS,
199 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
200 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
201 #ifdef CONFIG_IPV6_PRIVACY
202 .use_tempaddr = 0,
203 .temp_valid_lft = TEMP_VALID_LIFETIME,
204 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
205 .regen_max_retry = REGEN_MAX_RETRY,
206 .max_desync_factor = MAX_DESYNC_FACTOR,
207 #endif
208 .max_addresses = IPV6_MAX_ADDRESSES,
209 .accept_ra_defrtr = 1,
210 .accept_ra_pinfo = 1,
211 #ifdef CONFIG_IPV6_ROUTER_PREF
212 .accept_ra_rtr_pref = 1,
213 .rtr_probe_interval = 60 * HZ,
214 #ifdef CONFIG_IPV6_ROUTE_INFO
215 .accept_ra_rt_info_max_plen = 0,
216 #endif
217 #endif
218 .proxy_ndp = 0,
219 .accept_source_route = 0, /* we do not accept RH0 by default. */
222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
226 /* Check if a valid qdisc is available */
227 static inline int addrconf_qdisc_ok(struct net_device *dev)
229 return (dev->qdisc != &noop_qdisc);
232 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
234 if (del_timer(&ifp->timer))
235 __in6_ifa_put(ifp);
238 enum addrconf_timer_t
240 AC_NONE,
241 AC_DAD,
242 AC_RS,
245 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
246 enum addrconf_timer_t what,
247 unsigned long when)
249 if (!del_timer(&ifp->timer))
250 in6_ifa_hold(ifp);
252 switch (what) {
253 case AC_DAD:
254 ifp->timer.function = addrconf_dad_timer;
255 break;
256 case AC_RS:
257 ifp->timer.function = addrconf_rs_timer;
258 break;
259 default:;
261 ifp->timer.expires = jiffies + when;
262 add_timer(&ifp->timer);
265 static int snmp6_alloc_dev(struct inet6_dev *idev)
267 if (snmp_mib_init((void **)idev->stats.ipv6,
268 sizeof(struct ipstats_mib)) < 0)
269 goto err_ip;
270 if (snmp_mib_init((void **)idev->stats.icmpv6,
271 sizeof(struct icmpv6_mib)) < 0)
272 goto err_icmp;
273 if (snmp_mib_init((void **)idev->stats.icmpv6msg,
274 sizeof(struct icmpv6msg_mib)) < 0)
275 goto err_icmpmsg;
277 return 0;
279 err_icmpmsg:
280 snmp_mib_free((void **)idev->stats.icmpv6);
281 err_icmp:
282 snmp_mib_free((void **)idev->stats.ipv6);
283 err_ip:
284 return -ENOMEM;
287 static void snmp6_free_dev(struct inet6_dev *idev)
289 snmp_mib_free((void **)idev->stats.icmpv6msg);
290 snmp_mib_free((void **)idev->stats.icmpv6);
291 snmp_mib_free((void **)idev->stats.ipv6);
294 /* Nobody refers to this device, we may destroy it. */
296 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
298 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
299 kfree(idev);
302 void in6_dev_finish_destroy(struct inet6_dev *idev)
304 struct net_device *dev = idev->dev;
305 BUG_TRAP(idev->addr_list==NULL);
306 BUG_TRAP(idev->mc_list==NULL);
307 #ifdef NET_REFCNT_DEBUG
308 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
309 #endif
310 dev_put(dev);
311 if (!idev->dead) {
312 printk("Freeing alive inet6 device %p\n", idev);
313 return;
315 snmp6_free_dev(idev);
316 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
319 EXPORT_SYMBOL(in6_dev_finish_destroy);
321 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
323 struct inet6_dev *ndev;
324 struct in6_addr maddr;
326 ASSERT_RTNL();
328 if (dev->mtu < IPV6_MIN_MTU)
329 return NULL;
331 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
333 if (ndev == NULL)
334 return NULL;
336 rwlock_init(&ndev->lock);
337 ndev->dev = dev;
338 memcpy(&ndev->cnf, dev->nd_net->ipv6.devconf_dflt, sizeof(ndev->cnf));
339 ndev->cnf.mtu6 = dev->mtu;
340 ndev->cnf.sysctl = NULL;
341 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
342 if (ndev->nd_parms == NULL) {
343 kfree(ndev);
344 return NULL;
346 /* We refer to the device */
347 dev_hold(dev);
349 if (snmp6_alloc_dev(ndev) < 0) {
350 ADBG((KERN_WARNING
351 "%s(): cannot allocate memory for statistics; dev=%s.\n",
352 __FUNCTION__, dev->name));
353 neigh_parms_release(&nd_tbl, ndev->nd_parms);
354 ndev->dead = 1;
355 in6_dev_finish_destroy(ndev);
356 return NULL;
359 if (snmp6_register_dev(ndev) < 0) {
360 ADBG((KERN_WARNING
361 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
362 __FUNCTION__, dev->name));
363 neigh_parms_release(&nd_tbl, ndev->nd_parms);
364 ndev->dead = 1;
365 in6_dev_finish_destroy(ndev);
366 return NULL;
369 /* One reference from device. We must do this before
370 * we invoke __ipv6_regen_rndid().
372 in6_dev_hold(ndev);
374 #ifdef CONFIG_IPV6_PRIVACY
375 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
376 if ((dev->flags&IFF_LOOPBACK) ||
377 dev->type == ARPHRD_TUNNEL ||
378 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
379 dev->type == ARPHRD_SIT ||
380 #endif
381 dev->type == ARPHRD_NONE) {
382 printk(KERN_INFO
383 "%s: Disabled Privacy Extensions\n",
384 dev->name);
385 ndev->cnf.use_tempaddr = -1;
387 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
388 printk(KERN_INFO
389 "%s: Disabled Multicast RS\n",
390 dev->name);
391 ndev->cnf.rtr_solicits = 0;
393 } else {
394 in6_dev_hold(ndev);
395 ipv6_regen_rndid((unsigned long) ndev);
397 #endif
399 if (netif_running(dev) && addrconf_qdisc_ok(dev))
400 ndev->if_flags |= IF_READY;
402 ipv6_mc_init_dev(ndev);
403 ndev->tstamp = jiffies;
404 addrconf_sysctl_register(ndev);
405 /* protected by rtnl_lock */
406 rcu_assign_pointer(dev->ip6_ptr, ndev);
408 /* Join all-node multicast group */
409 ipv6_addr_all_nodes(&maddr);
410 ipv6_dev_mc_inc(dev, &maddr);
412 return ndev;
415 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
417 struct inet6_dev *idev;
419 ASSERT_RTNL();
421 if ((idev = __in6_dev_get(dev)) == NULL) {
422 if ((idev = ipv6_add_dev(dev)) == NULL)
423 return NULL;
426 if (dev->flags&IFF_UP)
427 ipv6_mc_up(idev);
428 return idev;
431 #ifdef CONFIG_SYSCTL
432 static void dev_forward_change(struct inet6_dev *idev)
434 struct net_device *dev;
435 struct inet6_ifaddr *ifa;
436 struct in6_addr addr;
438 if (!idev)
439 return;
440 dev = idev->dev;
441 if (dev && (dev->flags & IFF_MULTICAST)) {
442 ipv6_addr_all_routers(&addr);
444 if (idev->cnf.forwarding)
445 ipv6_dev_mc_inc(dev, &addr);
446 else
447 ipv6_dev_mc_dec(dev, &addr);
449 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
450 if (ifa->flags&IFA_F_TENTATIVE)
451 continue;
452 if (idev->cnf.forwarding)
453 addrconf_join_anycast(ifa);
454 else
455 addrconf_leave_anycast(ifa);
460 static void addrconf_forward_change(struct net *net, __s32 newf)
462 struct net_device *dev;
463 struct inet6_dev *idev;
465 read_lock(&dev_base_lock);
466 for_each_netdev(net, dev) {
467 rcu_read_lock();
468 idev = __in6_dev_get(dev);
469 if (idev) {
470 int changed = (!idev->cnf.forwarding) ^ (!newf);
471 idev->cnf.forwarding = newf;
472 if (changed)
473 dev_forward_change(idev);
475 rcu_read_unlock();
477 read_unlock(&dev_base_lock);
480 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
482 struct net *net;
484 net = (struct net *)table->extra2;
485 if (p == &net->ipv6.devconf_dflt->forwarding)
486 return;
488 if (p == &net->ipv6.devconf_all->forwarding) {
489 __s32 newf = net->ipv6.devconf_all->forwarding;
490 net->ipv6.devconf_dflt->forwarding = newf;
491 addrconf_forward_change(net, newf);
492 } else if ((!*p) ^ (!old))
493 dev_forward_change((struct inet6_dev *)table->extra1);
495 if (*p)
496 rt6_purge_dflt_routers();
498 #endif
500 /* Nobody refers to this ifaddr, destroy it */
502 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
504 BUG_TRAP(ifp->if_next==NULL);
505 BUG_TRAP(ifp->lst_next==NULL);
506 #ifdef NET_REFCNT_DEBUG
507 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
508 #endif
510 in6_dev_put(ifp->idev);
512 if (del_timer(&ifp->timer))
513 printk("Timer is still running, when freeing ifa=%p\n", ifp);
515 if (!ifp->dead) {
516 printk("Freeing alive inet6 address %p\n", ifp);
517 return;
519 dst_release(&ifp->rt->u.dst);
521 kfree(ifp);
524 static void
525 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
527 struct inet6_ifaddr *ifa, **ifap;
528 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
531 * Each device address list is sorted in order of scope -
532 * global before linklocal.
534 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
535 ifap = &ifa->if_next) {
536 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
537 break;
540 ifp->if_next = *ifap;
541 *ifap = ifp;
544 /* On success it returns ifp with increased reference count */
546 static struct inet6_ifaddr *
547 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
548 int scope, u32 flags)
550 struct inet6_ifaddr *ifa = NULL;
551 struct rt6_info *rt;
552 int hash;
553 int err = 0;
555 rcu_read_lock_bh();
556 if (idev->dead) {
557 err = -ENODEV; /*XXX*/
558 goto out2;
561 write_lock(&addrconf_hash_lock);
563 /* Ignore adding duplicate addresses on an interface */
564 if (ipv6_chk_same_addr(&init_net, addr, idev->dev)) {
565 ADBG(("ipv6_add_addr: already assigned\n"));
566 err = -EEXIST;
567 goto out;
570 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
572 if (ifa == NULL) {
573 ADBG(("ipv6_add_addr: malloc failed\n"));
574 err = -ENOBUFS;
575 goto out;
578 rt = addrconf_dst_alloc(idev, addr, 0);
579 if (IS_ERR(rt)) {
580 err = PTR_ERR(rt);
581 goto out;
584 ipv6_addr_copy(&ifa->addr, addr);
586 spin_lock_init(&ifa->lock);
587 init_timer(&ifa->timer);
588 ifa->timer.data = (unsigned long) ifa;
589 ifa->scope = scope;
590 ifa->prefix_len = pfxlen;
591 ifa->flags = flags | IFA_F_TENTATIVE;
592 ifa->cstamp = ifa->tstamp = jiffies;
594 ifa->rt = rt;
597 * part one of RFC 4429, section 3.3
598 * We should not configure an address as
599 * optimistic if we do not yet know the link
600 * layer address of our nexhop router
603 if (rt->rt6i_nexthop == NULL)
604 ifa->flags &= ~IFA_F_OPTIMISTIC;
606 ifa->idev = idev;
607 in6_dev_hold(idev);
608 /* For caller */
609 in6_ifa_hold(ifa);
611 /* Add to big hash table */
612 hash = ipv6_addr_hash(addr);
614 ifa->lst_next = inet6_addr_lst[hash];
615 inet6_addr_lst[hash] = ifa;
616 in6_ifa_hold(ifa);
617 write_unlock(&addrconf_hash_lock);
619 write_lock(&idev->lock);
620 /* Add to inet6_dev unicast addr list. */
621 ipv6_link_dev_addr(idev, ifa);
623 #ifdef CONFIG_IPV6_PRIVACY
624 if (ifa->flags&IFA_F_TEMPORARY) {
625 ifa->tmp_next = idev->tempaddr_list;
626 idev->tempaddr_list = ifa;
627 in6_ifa_hold(ifa);
629 #endif
631 in6_ifa_hold(ifa);
632 write_unlock(&idev->lock);
633 out2:
634 rcu_read_unlock_bh();
636 if (likely(err == 0))
637 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
638 else {
639 kfree(ifa);
640 ifa = ERR_PTR(err);
643 return ifa;
644 out:
645 write_unlock(&addrconf_hash_lock);
646 goto out2;
649 /* This function wants to get referenced ifp and releases it before return */
651 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
653 struct inet6_ifaddr *ifa, **ifap;
654 struct inet6_dev *idev = ifp->idev;
655 int hash;
656 int deleted = 0, onlink = 0;
657 unsigned long expires = jiffies;
659 hash = ipv6_addr_hash(&ifp->addr);
661 ifp->dead = 1;
663 write_lock_bh(&addrconf_hash_lock);
664 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
665 ifap = &ifa->lst_next) {
666 if (ifa == ifp) {
667 *ifap = ifa->lst_next;
668 __in6_ifa_put(ifp);
669 ifa->lst_next = NULL;
670 break;
673 write_unlock_bh(&addrconf_hash_lock);
675 write_lock_bh(&idev->lock);
676 #ifdef CONFIG_IPV6_PRIVACY
677 if (ifp->flags&IFA_F_TEMPORARY) {
678 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
679 ifap = &ifa->tmp_next) {
680 if (ifa == ifp) {
681 *ifap = ifa->tmp_next;
682 if (ifp->ifpub) {
683 in6_ifa_put(ifp->ifpub);
684 ifp->ifpub = NULL;
686 __in6_ifa_put(ifp);
687 ifa->tmp_next = NULL;
688 break;
692 #endif
694 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
695 if (ifa == ifp) {
696 *ifap = ifa->if_next;
697 __in6_ifa_put(ifp);
698 ifa->if_next = NULL;
699 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
700 break;
701 deleted = 1;
702 continue;
703 } else if (ifp->flags & IFA_F_PERMANENT) {
704 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
705 ifp->prefix_len)) {
706 if (ifa->flags & IFA_F_PERMANENT) {
707 onlink = 1;
708 if (deleted)
709 break;
710 } else {
711 unsigned long lifetime;
713 if (!onlink)
714 onlink = -1;
716 spin_lock(&ifa->lock);
717 lifetime = min_t(unsigned long,
718 ifa->valid_lft, 0x7fffffffUL/HZ);
719 if (time_before(expires,
720 ifa->tstamp + lifetime * HZ))
721 expires = ifa->tstamp + lifetime * HZ;
722 spin_unlock(&ifa->lock);
726 ifap = &ifa->if_next;
728 write_unlock_bh(&idev->lock);
730 ipv6_ifa_notify(RTM_DELADDR, ifp);
732 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
734 addrconf_del_timer(ifp);
737 * Purge or update corresponding prefix
739 * 1) we don't purge prefix here if address was not permanent.
740 * prefix is managed by its own lifetime.
741 * 2) if there're no addresses, delete prefix.
742 * 3) if there're still other permanent address(es),
743 * corresponding prefix is still permanent.
744 * 4) otherwise, update prefix lifetime to the
745 * longest valid lifetime among the corresponding
746 * addresses on the device.
747 * Note: subsequent RA will update lifetime.
749 * --yoshfuji
751 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
752 struct in6_addr prefix;
753 struct rt6_info *rt;
755 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
756 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
758 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
759 if (onlink == 0) {
760 ip6_del_rt(rt);
761 rt = NULL;
762 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
763 rt->rt6i_expires = expires;
764 rt->rt6i_flags |= RTF_EXPIRES;
767 dst_release(&rt->u.dst);
770 in6_ifa_put(ifp);
773 #ifdef CONFIG_IPV6_PRIVACY
774 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
776 struct inet6_dev *idev = ifp->idev;
777 struct in6_addr addr, *tmpaddr;
778 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
779 int tmp_plen;
780 int ret = 0;
781 int max_addresses;
782 u32 addr_flags;
784 write_lock(&idev->lock);
785 if (ift) {
786 spin_lock_bh(&ift->lock);
787 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
788 spin_unlock_bh(&ift->lock);
789 tmpaddr = &addr;
790 } else {
791 tmpaddr = NULL;
793 retry:
794 in6_dev_hold(idev);
795 if (idev->cnf.use_tempaddr <= 0) {
796 write_unlock(&idev->lock);
797 printk(KERN_INFO
798 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
799 in6_dev_put(idev);
800 ret = -1;
801 goto out;
803 spin_lock_bh(&ifp->lock);
804 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
805 idev->cnf.use_tempaddr = -1; /*XXX*/
806 spin_unlock_bh(&ifp->lock);
807 write_unlock(&idev->lock);
808 printk(KERN_WARNING
809 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
810 in6_dev_put(idev);
811 ret = -1;
812 goto out;
814 in6_ifa_hold(ifp);
815 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
816 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
817 spin_unlock_bh(&ifp->lock);
818 write_unlock(&idev->lock);
819 printk(KERN_WARNING
820 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
821 in6_ifa_put(ifp);
822 in6_dev_put(idev);
823 ret = -1;
824 goto out;
826 memcpy(&addr.s6_addr[8], idev->rndid, 8);
827 tmp_valid_lft = min_t(__u32,
828 ifp->valid_lft,
829 idev->cnf.temp_valid_lft);
830 tmp_prefered_lft = min_t(__u32,
831 ifp->prefered_lft,
832 idev->cnf.temp_prefered_lft - desync_factor / HZ);
833 tmp_plen = ifp->prefix_len;
834 max_addresses = idev->cnf.max_addresses;
835 tmp_cstamp = ifp->cstamp;
836 tmp_tstamp = ifp->tstamp;
837 spin_unlock_bh(&ifp->lock);
839 write_unlock(&idev->lock);
841 addr_flags = IFA_F_TEMPORARY;
842 /* set in addrconf_prefix_rcv() */
843 if (ifp->flags & IFA_F_OPTIMISTIC)
844 addr_flags |= IFA_F_OPTIMISTIC;
846 ift = !max_addresses ||
847 ipv6_count_addresses(idev) < max_addresses ?
848 ipv6_add_addr(idev, &addr, tmp_plen,
849 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
850 addr_flags) : NULL;
851 if (!ift || IS_ERR(ift)) {
852 in6_ifa_put(ifp);
853 in6_dev_put(idev);
854 printk(KERN_INFO
855 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
856 tmpaddr = &addr;
857 write_lock(&idev->lock);
858 goto retry;
861 spin_lock_bh(&ift->lock);
862 ift->ifpub = ifp;
863 ift->valid_lft = tmp_valid_lft;
864 ift->prefered_lft = tmp_prefered_lft;
865 ift->cstamp = tmp_cstamp;
866 ift->tstamp = tmp_tstamp;
867 spin_unlock_bh(&ift->lock);
869 addrconf_dad_start(ift, 0);
870 in6_ifa_put(ift);
871 in6_dev_put(idev);
872 out:
873 return ret;
875 #endif
878 * Choose an appropriate source address (RFC3484)
880 struct ipv6_saddr_score {
881 int addr_type;
882 unsigned int attrs;
883 int matchlen;
884 int scope;
885 unsigned int rule;
888 #define IPV6_SADDR_SCORE_LOCAL 0x0001
889 #define IPV6_SADDR_SCORE_PREFERRED 0x0004
890 #define IPV6_SADDR_SCORE_HOA 0x0008
891 #define IPV6_SADDR_SCORE_OIF 0x0010
892 #define IPV6_SADDR_SCORE_LABEL 0x0020
893 #define IPV6_SADDR_SCORE_PRIVACY 0x0040
895 static inline int ipv6_saddr_preferred(int type)
897 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
898 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
899 return 1;
900 return 0;
903 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
904 struct in6_addr *daddr, struct in6_addr *saddr)
906 struct ipv6_saddr_score hiscore;
907 struct inet6_ifaddr *ifa_result = NULL;
908 int daddr_type = __ipv6_addr_type(daddr);
909 int daddr_scope = __ipv6_addr_src_scope(daddr_type);
910 int daddr_ifindex = daddr_dev ? daddr_dev->ifindex : 0;
911 u32 daddr_label = ipv6_addr_label(daddr, daddr_type, daddr_ifindex);
912 struct net_device *dev;
914 memset(&hiscore, 0, sizeof(hiscore));
916 read_lock(&dev_base_lock);
917 rcu_read_lock();
919 for_each_netdev(&init_net, dev) {
920 struct inet6_dev *idev;
921 struct inet6_ifaddr *ifa;
923 /* Rule 0: Candidate Source Address (section 4)
924 * - multicast and link-local destination address,
925 * the set of candidate source address MUST only
926 * include addresses assigned to interfaces
927 * belonging to the same link as the outgoing
928 * interface.
929 * (- For site-local destination addresses, the
930 * set of candidate source addresses MUST only
931 * include addresses assigned to interfaces
932 * belonging to the same site as the outgoing
933 * interface.)
935 if ((daddr_type & IPV6_ADDR_MULTICAST ||
936 daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
937 daddr_dev && dev != daddr_dev)
938 continue;
940 idev = __in6_dev_get(dev);
941 if (!idev)
942 continue;
944 read_lock_bh(&idev->lock);
945 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
946 struct ipv6_saddr_score score;
948 score.addr_type = __ipv6_addr_type(&ifa->addr);
950 /* Rule 0:
951 * - Tentative Address (RFC2462 section 5.4)
952 * - A tentative address is not considered
953 * "assigned to an interface" in the traditional
954 * sense, unless it is also flagged as optimistic.
955 * - Candidate Source Address (section 4)
956 * - In any case, anycast addresses, multicast
957 * addresses, and the unspecified address MUST
958 * NOT be included in a candidate set.
960 if ((ifa->flags & IFA_F_TENTATIVE) &&
961 (!(ifa->flags & IFA_F_OPTIMISTIC)))
962 continue;
963 if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
964 score.addr_type & IPV6_ADDR_MULTICAST)) {
965 LIMIT_NETDEBUG(KERN_DEBUG
966 "ADDRCONF: unspecified / multicast address "
967 "assigned as unicast address on %s",
968 dev->name);
969 continue;
972 score.attrs = 0;
973 score.matchlen = 0;
974 score.scope = 0;
975 score.rule = 0;
977 if (ifa_result == NULL) {
978 /* record it if the first available entry */
979 goto record_it;
982 /* Rule 1: Prefer same address */
983 if (hiscore.rule < 1) {
984 if (ipv6_addr_equal(&ifa_result->addr, daddr))
985 hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
986 hiscore.rule++;
988 if (ipv6_addr_equal(&ifa->addr, daddr)) {
989 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
990 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
991 score.rule = 1;
992 goto record_it;
994 } else {
995 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
996 continue;
999 /* Rule 2: Prefer appropriate scope */
1000 if (hiscore.rule < 2) {
1001 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1002 hiscore.rule++;
1004 score.scope = __ipv6_addr_src_scope(score.addr_type);
1005 if (hiscore.scope < score.scope) {
1006 if (hiscore.scope < daddr_scope) {
1007 score.rule = 2;
1008 goto record_it;
1009 } else
1010 continue;
1011 } else if (score.scope < hiscore.scope) {
1012 if (score.scope < daddr_scope)
1013 break; /* addresses sorted by scope */
1014 else {
1015 score.rule = 2;
1016 goto record_it;
1020 /* Rule 3: Avoid deprecated and optimistic addresses */
1021 if (hiscore.rule < 3) {
1022 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1023 (((ifa_result->flags &
1024 (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0)))
1025 hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1026 hiscore.rule++;
1028 if (ipv6_saddr_preferred(score.addr_type) ||
1029 (((ifa->flags &
1030 (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0))) {
1031 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1032 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1033 score.rule = 3;
1034 goto record_it;
1036 } else {
1037 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1038 continue;
1041 /* Rule 4: Prefer home address */
1042 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
1043 if (hiscore.rule < 4) {
1044 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1045 hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1046 hiscore.rule++;
1048 if (ifa->flags & IFA_F_HOMEADDRESS) {
1049 score.attrs |= IPV6_SADDR_SCORE_HOA;
1050 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1051 score.rule = 4;
1052 goto record_it;
1054 } else {
1055 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1056 continue;
1058 #else
1059 if (hiscore.rule < 4)
1060 hiscore.rule++;
1061 #endif
1063 /* Rule 5: Prefer outgoing interface */
1064 if (hiscore.rule < 5) {
1065 if (daddr_dev == NULL ||
1066 daddr_dev == ifa_result->idev->dev)
1067 hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1068 hiscore.rule++;
1070 if (daddr_dev == NULL ||
1071 daddr_dev == ifa->idev->dev) {
1072 score.attrs |= IPV6_SADDR_SCORE_OIF;
1073 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1074 score.rule = 5;
1075 goto record_it;
1077 } else {
1078 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1079 continue;
1082 /* Rule 6: Prefer matching label */
1083 if (hiscore.rule < 6) {
1084 if (ipv6_addr_label(&ifa_result->addr,
1085 hiscore.addr_type,
1086 ifa_result->idev->dev->ifindex) == daddr_label)
1087 hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1088 hiscore.rule++;
1090 if (ipv6_addr_label(&ifa->addr,
1091 score.addr_type,
1092 ifa->idev->dev->ifindex) == daddr_label) {
1093 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1094 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1095 score.rule = 6;
1096 goto record_it;
1098 } else {
1099 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1100 continue;
1103 #ifdef CONFIG_IPV6_PRIVACY
1104 /* Rule 7: Prefer public address
1105 * Note: prefer temprary address if use_tempaddr >= 2
1107 if (hiscore.rule < 7) {
1108 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1109 (ifa_result->idev->cnf.use_tempaddr >= 2))
1110 hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1111 hiscore.rule++;
1113 if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1114 (ifa->idev->cnf.use_tempaddr >= 2)) {
1115 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1116 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1117 score.rule = 7;
1118 goto record_it;
1120 } else {
1121 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1122 continue;
1124 #else
1125 if (hiscore.rule < 7)
1126 hiscore.rule++;
1127 #endif
1128 /* Rule 8: Use longest matching prefix */
1129 if (hiscore.rule < 8) {
1130 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1131 hiscore.rule++;
1133 score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1134 if (score.matchlen > hiscore.matchlen) {
1135 score.rule = 8;
1136 goto record_it;
1138 #if 0
1139 else if (score.matchlen < hiscore.matchlen)
1140 continue;
1141 #endif
1143 /* Final Rule: choose first available one */
1144 continue;
1145 record_it:
1146 if (ifa_result)
1147 in6_ifa_put(ifa_result);
1148 in6_ifa_hold(ifa);
1149 ifa_result = ifa;
1150 hiscore = score;
1152 read_unlock_bh(&idev->lock);
1154 rcu_read_unlock();
1155 read_unlock(&dev_base_lock);
1157 if (!ifa_result)
1158 return -EADDRNOTAVAIL;
1160 ipv6_addr_copy(saddr, &ifa_result->addr);
1161 in6_ifa_put(ifa_result);
1162 return 0;
1166 int ipv6_get_saddr(struct dst_entry *dst,
1167 struct in6_addr *daddr, struct in6_addr *saddr)
1169 return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1172 EXPORT_SYMBOL(ipv6_get_saddr);
1174 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1175 unsigned char banned_flags)
1177 struct inet6_dev *idev;
1178 int err = -EADDRNOTAVAIL;
1180 rcu_read_lock();
1181 if ((idev = __in6_dev_get(dev)) != NULL) {
1182 struct inet6_ifaddr *ifp;
1184 read_lock_bh(&idev->lock);
1185 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1186 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1187 ipv6_addr_copy(addr, &ifp->addr);
1188 err = 0;
1189 break;
1192 read_unlock_bh(&idev->lock);
1194 rcu_read_unlock();
1195 return err;
1198 static int ipv6_count_addresses(struct inet6_dev *idev)
1200 int cnt = 0;
1201 struct inet6_ifaddr *ifp;
1203 read_lock_bh(&idev->lock);
1204 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1205 cnt++;
1206 read_unlock_bh(&idev->lock);
1207 return cnt;
1210 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1211 struct net_device *dev, int strict)
1213 struct inet6_ifaddr * ifp;
1214 u8 hash = ipv6_addr_hash(addr);
1216 read_lock_bh(&addrconf_hash_lock);
1217 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1218 if (ifp->idev->dev->nd_net != net)
1219 continue;
1220 if (ipv6_addr_equal(&ifp->addr, addr) &&
1221 !(ifp->flags&IFA_F_TENTATIVE)) {
1222 if (dev == NULL || ifp->idev->dev == dev ||
1223 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1224 break;
1227 read_unlock_bh(&addrconf_hash_lock);
1228 return ifp != NULL;
1230 EXPORT_SYMBOL(ipv6_chk_addr);
1232 static
1233 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1234 struct net_device *dev)
1236 struct inet6_ifaddr * ifp;
1237 u8 hash = ipv6_addr_hash(addr);
1239 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1240 if (ifp->idev->dev->nd_net != net)
1241 continue;
1242 if (ipv6_addr_equal(&ifp->addr, addr)) {
1243 if (dev == NULL || ifp->idev->dev == dev)
1244 break;
1247 return ifp != NULL;
1250 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1251 struct net_device *dev, int strict)
1253 struct inet6_ifaddr * ifp;
1254 u8 hash = ipv6_addr_hash(addr);
1256 read_lock_bh(&addrconf_hash_lock);
1257 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1258 if (ifp->idev->dev->nd_net != net)
1259 continue;
1260 if (ipv6_addr_equal(&ifp->addr, addr)) {
1261 if (dev == NULL || ifp->idev->dev == dev ||
1262 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1263 in6_ifa_hold(ifp);
1264 break;
1268 read_unlock_bh(&addrconf_hash_lock);
1270 return ifp;
1273 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1275 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1276 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1277 __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1278 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1279 int sk_ipv6only = ipv6_only_sock(sk);
1280 int sk2_ipv6only = inet_v6_ipv6only(sk2);
1281 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1282 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1284 if (!sk2_rcv_saddr && !sk_ipv6only)
1285 return 1;
1287 if (addr_type2 == IPV6_ADDR_ANY &&
1288 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1289 return 1;
1291 if (addr_type == IPV6_ADDR_ANY &&
1292 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1293 return 1;
1295 if (sk2_rcv_saddr6 &&
1296 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1297 return 1;
1299 if (addr_type == IPV6_ADDR_MAPPED &&
1300 !sk2_ipv6only &&
1301 (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1302 return 1;
1304 return 0;
1307 /* Gets referenced address, destroys ifaddr */
1309 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1311 if (ifp->flags&IFA_F_PERMANENT) {
1312 spin_lock_bh(&ifp->lock);
1313 addrconf_del_timer(ifp);
1314 ifp->flags |= IFA_F_TENTATIVE;
1315 spin_unlock_bh(&ifp->lock);
1316 in6_ifa_put(ifp);
1317 #ifdef CONFIG_IPV6_PRIVACY
1318 } else if (ifp->flags&IFA_F_TEMPORARY) {
1319 struct inet6_ifaddr *ifpub;
1320 spin_lock_bh(&ifp->lock);
1321 ifpub = ifp->ifpub;
1322 if (ifpub) {
1323 in6_ifa_hold(ifpub);
1324 spin_unlock_bh(&ifp->lock);
1325 ipv6_create_tempaddr(ifpub, ifp);
1326 in6_ifa_put(ifpub);
1327 } else {
1328 spin_unlock_bh(&ifp->lock);
1330 ipv6_del_addr(ifp);
1331 #endif
1332 } else
1333 ipv6_del_addr(ifp);
1336 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1338 if (net_ratelimit())
1339 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1340 addrconf_dad_stop(ifp);
1343 /* Join to solicited addr multicast group. */
1345 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1347 struct in6_addr maddr;
1349 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1350 return;
1352 addrconf_addr_solict_mult(addr, &maddr);
1353 ipv6_dev_mc_inc(dev, &maddr);
1356 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1358 struct in6_addr maddr;
1360 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1361 return;
1363 addrconf_addr_solict_mult(addr, &maddr);
1364 __ipv6_dev_mc_dec(idev, &maddr);
1367 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1369 struct in6_addr addr;
1370 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1371 if (ipv6_addr_any(&addr))
1372 return;
1373 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1376 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1378 struct in6_addr addr;
1379 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1380 if (ipv6_addr_any(&addr))
1381 return;
1382 __ipv6_dev_ac_dec(ifp->idev, &addr);
1385 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1387 if (dev->addr_len != ETH_ALEN)
1388 return -1;
1389 memcpy(eui, dev->dev_addr, 3);
1390 memcpy(eui + 5, dev->dev_addr + 3, 3);
1393 * The zSeries OSA network cards can be shared among various
1394 * OS instances, but the OSA cards have only one MAC address.
1395 * This leads to duplicate address conflicts in conjunction
1396 * with IPv6 if more than one instance uses the same card.
1398 * The driver for these cards can deliver a unique 16-bit
1399 * identifier for each instance sharing the same card. It is
1400 * placed instead of 0xFFFE in the interface identifier. The
1401 * "u" bit of the interface identifier is not inverted in this
1402 * case. Hence the resulting interface identifier has local
1403 * scope according to RFC2373.
1405 if (dev->dev_id) {
1406 eui[3] = (dev->dev_id >> 8) & 0xFF;
1407 eui[4] = dev->dev_id & 0xFF;
1408 } else {
1409 eui[3] = 0xFF;
1410 eui[4] = 0xFE;
1411 eui[0] ^= 2;
1413 return 0;
1416 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1418 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1419 if (dev->addr_len != ARCNET_ALEN)
1420 return -1;
1421 memset(eui, 0, 7);
1422 eui[7] = *(u8*)dev->dev_addr;
1423 return 0;
1426 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1428 if (dev->addr_len != INFINIBAND_ALEN)
1429 return -1;
1430 memcpy(eui, dev->dev_addr + 12, 8);
1431 eui[0] |= 2;
1432 return 0;
1435 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1437 switch (dev->type) {
1438 case ARPHRD_ETHER:
1439 case ARPHRD_FDDI:
1440 case ARPHRD_IEEE802_TR:
1441 return addrconf_ifid_eui48(eui, dev);
1442 case ARPHRD_ARCNET:
1443 return addrconf_ifid_arcnet(eui, dev);
1444 case ARPHRD_INFINIBAND:
1445 return addrconf_ifid_infiniband(eui, dev);
1446 case ARPHRD_SIT:
1447 if (dev->priv_flags & IFF_ISATAP)
1448 return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1450 return -1;
1453 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1455 int err = -1;
1456 struct inet6_ifaddr *ifp;
1458 read_lock_bh(&idev->lock);
1459 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1460 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1461 memcpy(eui, ifp->addr.s6_addr+8, 8);
1462 err = 0;
1463 break;
1466 read_unlock_bh(&idev->lock);
1467 return err;
1470 #ifdef CONFIG_IPV6_PRIVACY
1471 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1472 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1474 regen:
1475 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1476 idev->rndid[0] &= ~0x02;
1479 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1480 * check if generated address is not inappropriate
1482 * - Reserved subnet anycast (RFC 2526)
1483 * 11111101 11....11 1xxxxxxx
1484 * - ISATAP (RFC4214) 6.1
1485 * 00-00-5E-FE-xx-xx-xx-xx
1486 * - value 0
1487 * - XXX: already assigned to an address on the device
1489 if (idev->rndid[0] == 0xfd &&
1490 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1491 (idev->rndid[7]&0x80))
1492 goto regen;
1493 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1494 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1495 goto regen;
1496 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1497 goto regen;
1500 return 0;
1503 static void ipv6_regen_rndid(unsigned long data)
1505 struct inet6_dev *idev = (struct inet6_dev *) data;
1506 unsigned long expires;
1508 rcu_read_lock_bh();
1509 write_lock_bh(&idev->lock);
1511 if (idev->dead)
1512 goto out;
1514 if (__ipv6_regen_rndid(idev) < 0)
1515 goto out;
1517 expires = jiffies +
1518 idev->cnf.temp_prefered_lft * HZ -
1519 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1520 if (time_before(expires, jiffies)) {
1521 printk(KERN_WARNING
1522 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1523 idev->dev->name);
1524 goto out;
1527 if (!mod_timer(&idev->regen_timer, expires))
1528 in6_dev_hold(idev);
1530 out:
1531 write_unlock_bh(&idev->lock);
1532 rcu_read_unlock_bh();
1533 in6_dev_put(idev);
1536 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1537 int ret = 0;
1539 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1540 ret = __ipv6_regen_rndid(idev);
1541 return ret;
1543 #endif
1546 * Add prefix route.
1549 static void
1550 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1551 unsigned long expires, u32 flags)
1553 struct fib6_config cfg = {
1554 .fc_table = RT6_TABLE_PREFIX,
1555 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1556 .fc_ifindex = dev->ifindex,
1557 .fc_expires = expires,
1558 .fc_dst_len = plen,
1559 .fc_flags = RTF_UP | flags,
1560 .fc_nlinfo.nl_net = &init_net,
1563 ipv6_addr_copy(&cfg.fc_dst, pfx);
1565 /* Prevent useless cloning on PtP SIT.
1566 This thing is done here expecting that the whole
1567 class of non-broadcast devices need not cloning.
1569 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1570 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1571 cfg.fc_flags |= RTF_NONEXTHOP;
1572 #endif
1574 ip6_route_add(&cfg);
1577 /* Create "default" multicast route to the interface */
1579 static void addrconf_add_mroute(struct net_device *dev)
1581 struct fib6_config cfg = {
1582 .fc_table = RT6_TABLE_LOCAL,
1583 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1584 .fc_ifindex = dev->ifindex,
1585 .fc_dst_len = 8,
1586 .fc_flags = RTF_UP,
1587 .fc_nlinfo.nl_net = &init_net,
1590 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1592 ip6_route_add(&cfg);
1595 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1596 static void sit_route_add(struct net_device *dev)
1598 struct fib6_config cfg = {
1599 .fc_table = RT6_TABLE_MAIN,
1600 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1601 .fc_ifindex = dev->ifindex,
1602 .fc_dst_len = 96,
1603 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1604 .fc_nlinfo.nl_net = &init_net,
1607 /* prefix length - 96 bits "::d.d.d.d" */
1608 ip6_route_add(&cfg);
1610 #endif
1612 static void addrconf_add_lroute(struct net_device *dev)
1614 struct in6_addr addr;
1616 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1617 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1620 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1622 struct inet6_dev *idev;
1624 ASSERT_RTNL();
1626 if ((idev = ipv6_find_idev(dev)) == NULL)
1627 return NULL;
1629 /* Add default multicast route */
1630 addrconf_add_mroute(dev);
1632 /* Add link local route */
1633 addrconf_add_lroute(dev);
1634 return idev;
1637 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1639 struct prefix_info *pinfo;
1640 __u32 valid_lft;
1641 __u32 prefered_lft;
1642 int addr_type;
1643 unsigned long rt_expires;
1644 struct inet6_dev *in6_dev;
1646 pinfo = (struct prefix_info *) opt;
1648 if (len < sizeof(struct prefix_info)) {
1649 ADBG(("addrconf: prefix option too short\n"));
1650 return;
1654 * Validation checks ([ADDRCONF], page 19)
1657 addr_type = ipv6_addr_type(&pinfo->prefix);
1659 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1660 return;
1662 valid_lft = ntohl(pinfo->valid);
1663 prefered_lft = ntohl(pinfo->prefered);
1665 if (prefered_lft > valid_lft) {
1666 if (net_ratelimit())
1667 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1668 return;
1671 in6_dev = in6_dev_get(dev);
1673 if (in6_dev == NULL) {
1674 if (net_ratelimit())
1675 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1676 return;
1680 * Two things going on here:
1681 * 1) Add routes for on-link prefixes
1682 * 2) Configure prefixes with the auto flag set
1685 /* Avoid arithmetic overflow. Really, we could
1686 save rt_expires in seconds, likely valid_lft,
1687 but it would require division in fib gc, that it
1688 not good.
1690 if (valid_lft >= 0x7FFFFFFF/HZ)
1691 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1692 else
1693 rt_expires = valid_lft * HZ;
1696 * We convert this (in jiffies) to clock_t later.
1697 * Avoid arithmetic overflow there as well.
1698 * Overflow can happen only if HZ < USER_HZ.
1700 if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1701 rt_expires = 0x7FFFFFFF / USER_HZ;
1703 if (pinfo->onlink) {
1704 struct rt6_info *rt;
1705 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1707 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1708 if (rt->rt6i_flags&RTF_EXPIRES) {
1709 if (valid_lft == 0) {
1710 ip6_del_rt(rt);
1711 rt = NULL;
1712 } else {
1713 rt->rt6i_expires = jiffies + rt_expires;
1716 } else if (valid_lft) {
1717 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1718 dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1720 if (rt)
1721 dst_release(&rt->u.dst);
1724 /* Try to figure out our local address for this prefix */
1726 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1727 struct inet6_ifaddr * ifp;
1728 struct in6_addr addr;
1729 int create = 0, update_lft = 0;
1731 if (pinfo->prefix_len == 64) {
1732 memcpy(&addr, &pinfo->prefix, 8);
1733 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1734 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1735 in6_dev_put(in6_dev);
1736 return;
1738 goto ok;
1740 if (net_ratelimit())
1741 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1742 pinfo->prefix_len);
1743 in6_dev_put(in6_dev);
1744 return;
1748 ifp = ipv6_get_ifaddr(&init_net, &addr, dev, 1);
1750 if (ifp == NULL && valid_lft) {
1751 int max_addresses = in6_dev->cnf.max_addresses;
1752 u32 addr_flags = 0;
1754 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1755 if (in6_dev->cnf.optimistic_dad &&
1756 !ipv6_devconf.forwarding)
1757 addr_flags = IFA_F_OPTIMISTIC;
1758 #endif
1760 /* Do not allow to create too much of autoconfigured
1761 * addresses; this would be too easy way to crash kernel.
1763 if (!max_addresses ||
1764 ipv6_count_addresses(in6_dev) < max_addresses)
1765 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1766 addr_type&IPV6_ADDR_SCOPE_MASK,
1767 addr_flags);
1769 if (!ifp || IS_ERR(ifp)) {
1770 in6_dev_put(in6_dev);
1771 return;
1774 update_lft = create = 1;
1775 ifp->cstamp = jiffies;
1776 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1779 if (ifp) {
1780 int flags;
1781 unsigned long now;
1782 #ifdef CONFIG_IPV6_PRIVACY
1783 struct inet6_ifaddr *ift;
1784 #endif
1785 u32 stored_lft;
1787 /* update lifetime (RFC2462 5.5.3 e) */
1788 spin_lock(&ifp->lock);
1789 now = jiffies;
1790 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1791 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1792 else
1793 stored_lft = 0;
1794 if (!update_lft && stored_lft) {
1795 if (valid_lft > MIN_VALID_LIFETIME ||
1796 valid_lft > stored_lft)
1797 update_lft = 1;
1798 else if (stored_lft <= MIN_VALID_LIFETIME) {
1799 /* valid_lft <= stored_lft is always true */
1800 /* XXX: IPsec */
1801 update_lft = 0;
1802 } else {
1803 valid_lft = MIN_VALID_LIFETIME;
1804 if (valid_lft < prefered_lft)
1805 prefered_lft = valid_lft;
1806 update_lft = 1;
1810 if (update_lft) {
1811 ifp->valid_lft = valid_lft;
1812 ifp->prefered_lft = prefered_lft;
1813 ifp->tstamp = now;
1814 flags = ifp->flags;
1815 ifp->flags &= ~IFA_F_DEPRECATED;
1816 spin_unlock(&ifp->lock);
1818 if (!(flags&IFA_F_TENTATIVE))
1819 ipv6_ifa_notify(0, ifp);
1820 } else
1821 spin_unlock(&ifp->lock);
1823 #ifdef CONFIG_IPV6_PRIVACY
1824 read_lock_bh(&in6_dev->lock);
1825 /* update all temporary addresses in the list */
1826 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1828 * When adjusting the lifetimes of an existing
1829 * temporary address, only lower the lifetimes.
1830 * Implementations must not increase the
1831 * lifetimes of an existing temporary address
1832 * when processing a Prefix Information Option.
1834 spin_lock(&ift->lock);
1835 flags = ift->flags;
1836 if (ift->valid_lft > valid_lft &&
1837 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1838 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1839 if (ift->prefered_lft > prefered_lft &&
1840 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1841 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1842 spin_unlock(&ift->lock);
1843 if (!(flags&IFA_F_TENTATIVE))
1844 ipv6_ifa_notify(0, ift);
1847 if (create && in6_dev->cnf.use_tempaddr > 0) {
1849 * When a new public address is created as described in [ADDRCONF],
1850 * also create a new temporary address.
1852 read_unlock_bh(&in6_dev->lock);
1853 ipv6_create_tempaddr(ifp, NULL);
1854 } else {
1855 read_unlock_bh(&in6_dev->lock);
1857 #endif
1858 in6_ifa_put(ifp);
1859 addrconf_verify(0);
1862 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1863 in6_dev_put(in6_dev);
1867 * Set destination address.
1868 * Special case for SIT interfaces where we create a new "virtual"
1869 * device.
1871 int addrconf_set_dstaddr(void __user *arg)
1873 struct in6_ifreq ireq;
1874 struct net_device *dev;
1875 int err = -EINVAL;
1877 rtnl_lock();
1879 err = -EFAULT;
1880 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1881 goto err_exit;
1883 dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1885 err = -ENODEV;
1886 if (dev == NULL)
1887 goto err_exit;
1889 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1890 if (dev->type == ARPHRD_SIT) {
1891 struct ifreq ifr;
1892 mm_segment_t oldfs;
1893 struct ip_tunnel_parm p;
1895 err = -EADDRNOTAVAIL;
1896 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1897 goto err_exit;
1899 memset(&p, 0, sizeof(p));
1900 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1901 p.iph.saddr = 0;
1902 p.iph.version = 4;
1903 p.iph.ihl = 5;
1904 p.iph.protocol = IPPROTO_IPV6;
1905 p.iph.ttl = 64;
1906 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1908 oldfs = get_fs(); set_fs(KERNEL_DS);
1909 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1910 set_fs(oldfs);
1912 if (err == 0) {
1913 err = -ENOBUFS;
1914 if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1915 goto err_exit;
1916 err = dev_open(dev);
1919 #endif
1921 err_exit:
1922 rtnl_unlock();
1923 return err;
1927 * Manual configuration of address on an interface
1929 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1930 __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1932 struct inet6_ifaddr *ifp;
1933 struct inet6_dev *idev;
1934 struct net_device *dev;
1935 int scope;
1936 u32 flags = RTF_EXPIRES;
1938 ASSERT_RTNL();
1940 /* check the lifetime */
1941 if (!valid_lft || prefered_lft > valid_lft)
1942 return -EINVAL;
1944 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1945 return -ENODEV;
1947 if ((idev = addrconf_add_dev(dev)) == NULL)
1948 return -ENOBUFS;
1950 scope = ipv6_addr_scope(pfx);
1952 if (valid_lft == INFINITY_LIFE_TIME) {
1953 ifa_flags |= IFA_F_PERMANENT;
1954 flags = 0;
1955 } else if (valid_lft >= 0x7FFFFFFF/HZ)
1956 valid_lft = 0x7FFFFFFF/HZ;
1958 if (prefered_lft == 0)
1959 ifa_flags |= IFA_F_DEPRECATED;
1960 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1961 (prefered_lft != INFINITY_LIFE_TIME))
1962 prefered_lft = 0x7FFFFFFF/HZ;
1964 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1966 if (!IS_ERR(ifp)) {
1967 spin_lock_bh(&ifp->lock);
1968 ifp->valid_lft = valid_lft;
1969 ifp->prefered_lft = prefered_lft;
1970 ifp->tstamp = jiffies;
1971 spin_unlock_bh(&ifp->lock);
1973 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1974 jiffies_to_clock_t(valid_lft * HZ), flags);
1976 * Note that section 3.1 of RFC 4429 indicates
1977 * that the Optimistic flag should not be set for
1978 * manually configured addresses
1980 addrconf_dad_start(ifp, 0);
1981 in6_ifa_put(ifp);
1982 addrconf_verify(0);
1983 return 0;
1986 return PTR_ERR(ifp);
1989 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1991 struct inet6_ifaddr *ifp;
1992 struct inet6_dev *idev;
1993 struct net_device *dev;
1995 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1996 return -ENODEV;
1998 if ((idev = __in6_dev_get(dev)) == NULL)
1999 return -ENXIO;
2001 read_lock_bh(&idev->lock);
2002 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2003 if (ifp->prefix_len == plen &&
2004 ipv6_addr_equal(pfx, &ifp->addr)) {
2005 in6_ifa_hold(ifp);
2006 read_unlock_bh(&idev->lock);
2008 ipv6_del_addr(ifp);
2010 /* If the last address is deleted administratively,
2011 disable IPv6 on this interface.
2013 if (idev->addr_list == NULL)
2014 addrconf_ifdown(idev->dev, 1);
2015 return 0;
2018 read_unlock_bh(&idev->lock);
2019 return -EADDRNOTAVAIL;
2023 int addrconf_add_ifaddr(void __user *arg)
2025 struct in6_ifreq ireq;
2026 int err;
2028 if (!capable(CAP_NET_ADMIN))
2029 return -EPERM;
2031 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2032 return -EFAULT;
2034 rtnl_lock();
2035 err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2036 IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2037 rtnl_unlock();
2038 return err;
2041 int addrconf_del_ifaddr(void __user *arg)
2043 struct in6_ifreq ireq;
2044 int err;
2046 if (!capable(CAP_NET_ADMIN))
2047 return -EPERM;
2049 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2050 return -EFAULT;
2052 rtnl_lock();
2053 err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2054 rtnl_unlock();
2055 return err;
2058 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2059 static void sit_add_v4_addrs(struct inet6_dev *idev)
2061 struct inet6_ifaddr * ifp;
2062 struct in6_addr addr;
2063 struct net_device *dev;
2064 int scope;
2066 ASSERT_RTNL();
2068 memset(&addr, 0, sizeof(struct in6_addr));
2069 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2071 if (idev->dev->flags&IFF_POINTOPOINT) {
2072 addr.s6_addr32[0] = htonl(0xfe800000);
2073 scope = IFA_LINK;
2074 } else {
2075 scope = IPV6_ADDR_COMPATv4;
2078 if (addr.s6_addr32[3]) {
2079 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2080 if (!IS_ERR(ifp)) {
2081 spin_lock_bh(&ifp->lock);
2082 ifp->flags &= ~IFA_F_TENTATIVE;
2083 spin_unlock_bh(&ifp->lock);
2084 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2085 in6_ifa_put(ifp);
2087 return;
2090 for_each_netdev(&init_net, dev) {
2091 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2092 if (in_dev && (dev->flags & IFF_UP)) {
2093 struct in_ifaddr * ifa;
2095 int flag = scope;
2097 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2098 int plen;
2100 addr.s6_addr32[3] = ifa->ifa_local;
2102 if (ifa->ifa_scope == RT_SCOPE_LINK)
2103 continue;
2104 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2105 if (idev->dev->flags&IFF_POINTOPOINT)
2106 continue;
2107 flag |= IFA_HOST;
2109 if (idev->dev->flags&IFF_POINTOPOINT)
2110 plen = 64;
2111 else
2112 plen = 96;
2114 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2115 IFA_F_PERMANENT);
2116 if (!IS_ERR(ifp)) {
2117 spin_lock_bh(&ifp->lock);
2118 ifp->flags &= ~IFA_F_TENTATIVE;
2119 spin_unlock_bh(&ifp->lock);
2120 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2121 in6_ifa_put(ifp);
2127 #endif
2129 static void init_loopback(struct net_device *dev)
2131 struct inet6_dev *idev;
2132 struct inet6_ifaddr * ifp;
2134 /* ::1 */
2136 ASSERT_RTNL();
2138 if ((idev = ipv6_find_idev(dev)) == NULL) {
2139 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2140 return;
2143 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2144 if (!IS_ERR(ifp)) {
2145 spin_lock_bh(&ifp->lock);
2146 ifp->flags &= ~IFA_F_TENTATIVE;
2147 spin_unlock_bh(&ifp->lock);
2148 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2149 in6_ifa_put(ifp);
2153 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2155 struct inet6_ifaddr * ifp;
2156 u32 addr_flags = IFA_F_PERMANENT;
2158 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2159 if (idev->cnf.optimistic_dad &&
2160 !ipv6_devconf.forwarding)
2161 addr_flags |= IFA_F_OPTIMISTIC;
2162 #endif
2165 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2166 if (!IS_ERR(ifp)) {
2167 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2168 addrconf_dad_start(ifp, 0);
2169 in6_ifa_put(ifp);
2173 static void addrconf_dev_config(struct net_device *dev)
2175 struct in6_addr addr;
2176 struct inet6_dev * idev;
2178 ASSERT_RTNL();
2180 if ((dev->type != ARPHRD_ETHER) &&
2181 (dev->type != ARPHRD_FDDI) &&
2182 (dev->type != ARPHRD_IEEE802_TR) &&
2183 (dev->type != ARPHRD_ARCNET) &&
2184 (dev->type != ARPHRD_INFINIBAND)) {
2185 /* Alas, we support only Ethernet autoconfiguration. */
2186 return;
2189 idev = addrconf_add_dev(dev);
2190 if (idev == NULL)
2191 return;
2193 memset(&addr, 0, sizeof(struct in6_addr));
2194 addr.s6_addr32[0] = htonl(0xFE800000);
2196 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2197 addrconf_add_linklocal(idev, &addr);
2200 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2201 static void addrconf_sit_config(struct net_device *dev)
2203 struct inet6_dev *idev;
2205 ASSERT_RTNL();
2208 * Configure the tunnel with one of our IPv4
2209 * addresses... we should configure all of
2210 * our v4 addrs in the tunnel
2213 if ((idev = ipv6_find_idev(dev)) == NULL) {
2214 printk(KERN_DEBUG "init sit: add_dev failed\n");
2215 return;
2218 if (dev->priv_flags & IFF_ISATAP) {
2219 struct in6_addr addr;
2221 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2222 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2223 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2224 addrconf_add_linklocal(idev, &addr);
2225 return;
2228 sit_add_v4_addrs(idev);
2230 if (dev->flags&IFF_POINTOPOINT) {
2231 addrconf_add_mroute(dev);
2232 addrconf_add_lroute(dev);
2233 } else
2234 sit_route_add(dev);
2236 #endif
2238 static inline int
2239 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2241 struct in6_addr lladdr;
2243 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2244 addrconf_add_linklocal(idev, &lladdr);
2245 return 0;
2247 return -1;
2250 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2252 struct net_device *link_dev;
2254 /* first try to inherit the link-local address from the link device */
2255 if (idev->dev->iflink &&
2256 (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2257 if (!ipv6_inherit_linklocal(idev, link_dev))
2258 return;
2260 /* then try to inherit it from any device */
2261 for_each_netdev(&init_net, link_dev) {
2262 if (!ipv6_inherit_linklocal(idev, link_dev))
2263 return;
2265 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2269 * Autoconfigure tunnel with a link-local address so routing protocols,
2270 * DHCPv6, MLD etc. can be run over the virtual link
2273 static void addrconf_ip6_tnl_config(struct net_device *dev)
2275 struct inet6_dev *idev;
2277 ASSERT_RTNL();
2279 if ((idev = addrconf_add_dev(dev)) == NULL) {
2280 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2281 return;
2283 ip6_tnl_add_linklocal(idev);
2286 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2287 void * data)
2289 struct net_device *dev = (struct net_device *) data;
2290 struct inet6_dev *idev = __in6_dev_get(dev);
2291 int run_pending = 0;
2292 int err;
2294 if (dev->nd_net != &init_net)
2295 return NOTIFY_DONE;
2297 switch(event) {
2298 case NETDEV_REGISTER:
2299 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2300 idev = ipv6_add_dev(dev);
2301 if (!idev)
2302 return notifier_from_errno(-ENOMEM);
2304 break;
2305 case NETDEV_UP:
2306 case NETDEV_CHANGE:
2307 if (dev->flags & IFF_SLAVE)
2308 break;
2310 if (event == NETDEV_UP) {
2311 if (!addrconf_qdisc_ok(dev)) {
2312 /* device is not ready yet. */
2313 printk(KERN_INFO
2314 "ADDRCONF(NETDEV_UP): %s: "
2315 "link is not ready\n",
2316 dev->name);
2317 break;
2320 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2321 idev = ipv6_add_dev(dev);
2323 if (idev)
2324 idev->if_flags |= IF_READY;
2325 } else {
2326 if (!addrconf_qdisc_ok(dev)) {
2327 /* device is still not ready. */
2328 break;
2331 if (idev) {
2332 if (idev->if_flags & IF_READY) {
2333 /* device is already configured. */
2334 break;
2336 idev->if_flags |= IF_READY;
2339 printk(KERN_INFO
2340 "ADDRCONF(NETDEV_CHANGE): %s: "
2341 "link becomes ready\n",
2342 dev->name);
2344 run_pending = 1;
2347 switch(dev->type) {
2348 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2349 case ARPHRD_SIT:
2350 addrconf_sit_config(dev);
2351 break;
2352 #endif
2353 case ARPHRD_TUNNEL6:
2354 addrconf_ip6_tnl_config(dev);
2355 break;
2356 case ARPHRD_LOOPBACK:
2357 init_loopback(dev);
2358 break;
2360 default:
2361 addrconf_dev_config(dev);
2362 break;
2364 if (idev) {
2365 if (run_pending)
2366 addrconf_dad_run(idev);
2368 /* If the MTU changed during the interface down, when the
2369 interface up, the changed MTU must be reflected in the
2370 idev as well as routers.
2372 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2373 rt6_mtu_change(dev, dev->mtu);
2374 idev->cnf.mtu6 = dev->mtu;
2376 idev->tstamp = jiffies;
2377 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2378 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2379 stop IPv6 on this interface.
2381 if (dev->mtu < IPV6_MIN_MTU)
2382 addrconf_ifdown(dev, event != NETDEV_DOWN);
2384 break;
2386 case NETDEV_CHANGEMTU:
2387 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2388 rt6_mtu_change(dev, dev->mtu);
2389 idev->cnf.mtu6 = dev->mtu;
2390 break;
2393 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2394 idev = ipv6_add_dev(dev);
2395 if (idev)
2396 break;
2399 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2401 case NETDEV_DOWN:
2402 case NETDEV_UNREGISTER:
2404 * Remove all addresses from this interface.
2406 addrconf_ifdown(dev, event != NETDEV_DOWN);
2407 break;
2409 case NETDEV_CHANGENAME:
2410 if (idev) {
2411 snmp6_unregister_dev(idev);
2412 addrconf_sysctl_unregister(idev);
2413 addrconf_sysctl_register(idev);
2414 err = snmp6_register_dev(idev);
2415 if (err)
2416 return notifier_from_errno(err);
2418 break;
2421 return NOTIFY_OK;
2425 * addrconf module should be notified of a device going up
2427 static struct notifier_block ipv6_dev_notf = {
2428 .notifier_call = addrconf_notify,
2429 .priority = 0
2432 static int addrconf_ifdown(struct net_device *dev, int how)
2434 struct inet6_dev *idev;
2435 struct inet6_ifaddr *ifa, **bifa;
2436 int i;
2438 ASSERT_RTNL();
2440 if (dev == init_net.loopback_dev && how == 1)
2441 how = 0;
2443 rt6_ifdown(dev);
2444 neigh_ifdown(&nd_tbl, dev);
2446 idev = __in6_dev_get(dev);
2447 if (idev == NULL)
2448 return -ENODEV;
2450 /* Step 1: remove reference to ipv6 device from parent device.
2451 Do not dev_put!
2453 if (how == 1) {
2454 idev->dead = 1;
2456 /* protected by rtnl_lock */
2457 rcu_assign_pointer(dev->ip6_ptr, NULL);
2459 /* Step 1.5: remove snmp6 entry */
2460 snmp6_unregister_dev(idev);
2464 /* Step 2: clear hash table */
2465 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2466 bifa = &inet6_addr_lst[i];
2468 write_lock_bh(&addrconf_hash_lock);
2469 while ((ifa = *bifa) != NULL) {
2470 if (ifa->idev == idev) {
2471 *bifa = ifa->lst_next;
2472 ifa->lst_next = NULL;
2473 addrconf_del_timer(ifa);
2474 in6_ifa_put(ifa);
2475 continue;
2477 bifa = &ifa->lst_next;
2479 write_unlock_bh(&addrconf_hash_lock);
2482 write_lock_bh(&idev->lock);
2484 /* Step 3: clear flags for stateless addrconf */
2485 if (how != 1)
2486 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2488 /* Step 4: clear address list */
2489 #ifdef CONFIG_IPV6_PRIVACY
2490 if (how == 1 && del_timer(&idev->regen_timer))
2491 in6_dev_put(idev);
2493 /* clear tempaddr list */
2494 while ((ifa = idev->tempaddr_list) != NULL) {
2495 idev->tempaddr_list = ifa->tmp_next;
2496 ifa->tmp_next = NULL;
2497 ifa->dead = 1;
2498 write_unlock_bh(&idev->lock);
2499 spin_lock_bh(&ifa->lock);
2501 if (ifa->ifpub) {
2502 in6_ifa_put(ifa->ifpub);
2503 ifa->ifpub = NULL;
2505 spin_unlock_bh(&ifa->lock);
2506 in6_ifa_put(ifa);
2507 write_lock_bh(&idev->lock);
2509 #endif
2510 while ((ifa = idev->addr_list) != NULL) {
2511 idev->addr_list = ifa->if_next;
2512 ifa->if_next = NULL;
2513 ifa->dead = 1;
2514 addrconf_del_timer(ifa);
2515 write_unlock_bh(&idev->lock);
2517 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2518 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2519 in6_ifa_put(ifa);
2521 write_lock_bh(&idev->lock);
2523 write_unlock_bh(&idev->lock);
2525 /* Step 5: Discard multicast list */
2527 if (how == 1)
2528 ipv6_mc_destroy_dev(idev);
2529 else
2530 ipv6_mc_down(idev);
2532 idev->tstamp = jiffies;
2534 /* Shot the device (if unregistered) */
2536 if (how == 1) {
2537 addrconf_sysctl_unregister(idev);
2538 neigh_parms_release(&nd_tbl, idev->nd_parms);
2539 neigh_ifdown(&nd_tbl, dev);
2540 in6_dev_put(idev);
2542 return 0;
2545 static void addrconf_rs_timer(unsigned long data)
2547 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2549 if (ifp->idev->cnf.forwarding)
2550 goto out;
2552 if (ifp->idev->if_flags & IF_RA_RCVD) {
2554 * Announcement received after solicitation
2555 * was sent
2557 goto out;
2560 spin_lock(&ifp->lock);
2561 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2562 struct in6_addr all_routers;
2564 /* The wait after the last probe can be shorter */
2565 addrconf_mod_timer(ifp, AC_RS,
2566 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2567 ifp->idev->cnf.rtr_solicit_delay :
2568 ifp->idev->cnf.rtr_solicit_interval);
2569 spin_unlock(&ifp->lock);
2571 ipv6_addr_all_routers(&all_routers);
2573 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2574 } else {
2575 spin_unlock(&ifp->lock);
2577 * Note: we do not support deprecated "all on-link"
2578 * assumption any longer.
2580 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2581 ifp->idev->dev->name);
2584 out:
2585 in6_ifa_put(ifp);
2589 * Duplicate Address Detection
2591 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2593 unsigned long rand_num;
2594 struct inet6_dev *idev = ifp->idev;
2596 if (ifp->flags & IFA_F_OPTIMISTIC)
2597 rand_num = 0;
2598 else
2599 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2601 ifp->probes = idev->cnf.dad_transmits;
2602 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2605 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2607 struct inet6_dev *idev = ifp->idev;
2608 struct net_device *dev = idev->dev;
2610 addrconf_join_solict(dev, &ifp->addr);
2612 net_srandom(ifp->addr.s6_addr32[3]);
2614 read_lock_bh(&idev->lock);
2615 if (ifp->dead)
2616 goto out;
2617 spin_lock_bh(&ifp->lock);
2619 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2620 !(ifp->flags&IFA_F_TENTATIVE) ||
2621 ifp->flags & IFA_F_NODAD) {
2622 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2623 spin_unlock_bh(&ifp->lock);
2624 read_unlock_bh(&idev->lock);
2626 addrconf_dad_completed(ifp);
2627 return;
2630 if (!(idev->if_flags & IF_READY)) {
2631 spin_unlock_bh(&ifp->lock);
2632 read_unlock_bh(&idev->lock);
2634 * If the defice is not ready:
2635 * - keep it tentative if it is a permanent address.
2636 * - otherwise, kill it.
2638 in6_ifa_hold(ifp);
2639 addrconf_dad_stop(ifp);
2640 return;
2644 * Optimistic nodes can start receiving
2645 * Frames right away
2647 if(ifp->flags & IFA_F_OPTIMISTIC)
2648 ip6_ins_rt(ifp->rt);
2650 addrconf_dad_kick(ifp);
2651 spin_unlock_bh(&ifp->lock);
2652 out:
2653 read_unlock_bh(&idev->lock);
2656 static void addrconf_dad_timer(unsigned long data)
2658 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2659 struct inet6_dev *idev = ifp->idev;
2660 struct in6_addr unspec;
2661 struct in6_addr mcaddr;
2663 read_lock_bh(&idev->lock);
2664 if (idev->dead) {
2665 read_unlock_bh(&idev->lock);
2666 goto out;
2668 spin_lock_bh(&ifp->lock);
2669 if (ifp->probes == 0) {
2671 * DAD was successful
2674 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2675 spin_unlock_bh(&ifp->lock);
2676 read_unlock_bh(&idev->lock);
2678 addrconf_dad_completed(ifp);
2680 goto out;
2683 ifp->probes--;
2684 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2685 spin_unlock_bh(&ifp->lock);
2686 read_unlock_bh(&idev->lock);
2688 /* send a neighbour solicitation for our addr */
2689 memset(&unspec, 0, sizeof(unspec));
2690 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2691 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2692 out:
2693 in6_ifa_put(ifp);
2696 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2698 struct net_device * dev = ifp->idev->dev;
2701 * Configure the address for reception. Now it is valid.
2704 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2706 /* If added prefix is link local and forwarding is off,
2707 start sending router solicitations.
2710 if (ifp->idev->cnf.forwarding == 0 &&
2711 ifp->idev->cnf.rtr_solicits > 0 &&
2712 (dev->flags&IFF_LOOPBACK) == 0 &&
2713 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2714 struct in6_addr all_routers;
2716 ipv6_addr_all_routers(&all_routers);
2719 * If a host as already performed a random delay
2720 * [...] as part of DAD [...] there is no need
2721 * to delay again before sending the first RS
2723 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2725 spin_lock_bh(&ifp->lock);
2726 ifp->probes = 1;
2727 ifp->idev->if_flags |= IF_RS_SENT;
2728 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2729 spin_unlock_bh(&ifp->lock);
2733 static void addrconf_dad_run(struct inet6_dev *idev) {
2734 struct inet6_ifaddr *ifp;
2736 read_lock_bh(&idev->lock);
2737 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2738 spin_lock_bh(&ifp->lock);
2739 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2740 spin_unlock_bh(&ifp->lock);
2741 continue;
2743 spin_unlock_bh(&ifp->lock);
2744 addrconf_dad_kick(ifp);
2746 read_unlock_bh(&idev->lock);
2749 #ifdef CONFIG_PROC_FS
2750 struct if6_iter_state {
2751 struct seq_net_private p;
2752 int bucket;
2755 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2757 struct inet6_ifaddr *ifa = NULL;
2758 struct if6_iter_state *state = seq->private;
2759 struct net *net = state->p.net;
2761 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2762 ifa = inet6_addr_lst[state->bucket];
2764 while (ifa && ifa->idev->dev->nd_net != net)
2765 ifa = ifa->lst_next;
2766 if (ifa)
2767 break;
2769 return ifa;
2772 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2774 struct if6_iter_state *state = seq->private;
2775 struct net *net = state->p.net;
2777 ifa = ifa->lst_next;
2778 try_again:
2779 if (ifa) {
2780 if (ifa->idev->dev->nd_net != net) {
2781 ifa = ifa->lst_next;
2782 goto try_again;
2786 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2787 ifa = inet6_addr_lst[state->bucket];
2788 goto try_again;
2791 return ifa;
2794 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2796 struct inet6_ifaddr *ifa = if6_get_first(seq);
2798 if (ifa)
2799 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2800 --pos;
2801 return pos ? NULL : ifa;
2804 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2805 __acquires(addrconf_hash_lock)
2807 read_lock_bh(&addrconf_hash_lock);
2808 return if6_get_idx(seq, *pos);
2811 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2813 struct inet6_ifaddr *ifa;
2815 ifa = if6_get_next(seq, v);
2816 ++*pos;
2817 return ifa;
2820 static void if6_seq_stop(struct seq_file *seq, void *v)
2821 __releases(addrconf_hash_lock)
2823 read_unlock_bh(&addrconf_hash_lock);
2826 static int if6_seq_show(struct seq_file *seq, void *v)
2828 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2829 seq_printf(seq,
2830 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2831 NIP6(ifp->addr),
2832 ifp->idev->dev->ifindex,
2833 ifp->prefix_len,
2834 ifp->scope,
2835 ifp->flags,
2836 ifp->idev->dev->name);
2837 return 0;
2840 static const struct seq_operations if6_seq_ops = {
2841 .start = if6_seq_start,
2842 .next = if6_seq_next,
2843 .show = if6_seq_show,
2844 .stop = if6_seq_stop,
2847 static int if6_seq_open(struct inode *inode, struct file *file)
2849 return seq_open_net(inode, file, &if6_seq_ops,
2850 sizeof(struct if6_iter_state));
2853 static const struct file_operations if6_fops = {
2854 .owner = THIS_MODULE,
2855 .open = if6_seq_open,
2856 .read = seq_read,
2857 .llseek = seq_lseek,
2858 .release = seq_release_net,
2861 static int if6_proc_net_init(struct net *net)
2863 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2864 return -ENOMEM;
2865 return 0;
2868 static void if6_proc_net_exit(struct net *net)
2870 proc_net_remove(net, "if_inet6");
2873 static struct pernet_operations if6_proc_net_ops = {
2874 .init = if6_proc_net_init,
2875 .exit = if6_proc_net_exit,
2878 int __init if6_proc_init(void)
2880 return register_pernet_subsys(&if6_proc_net_ops);
2883 void if6_proc_exit(void)
2885 unregister_pernet_subsys(&if6_proc_net_ops);
2887 #endif /* CONFIG_PROC_FS */
2889 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2890 /* Check if address is a home address configured on any interface. */
2891 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2893 int ret = 0;
2894 struct inet6_ifaddr * ifp;
2895 u8 hash = ipv6_addr_hash(addr);
2896 read_lock_bh(&addrconf_hash_lock);
2897 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2898 if (ifp->idev->dev->nd_net != net)
2899 continue;
2900 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2901 (ifp->flags & IFA_F_HOMEADDRESS)) {
2902 ret = 1;
2903 break;
2906 read_unlock_bh(&addrconf_hash_lock);
2907 return ret;
2909 #endif
2912 * Periodic address status verification
2915 static void addrconf_verify(unsigned long foo)
2917 struct inet6_ifaddr *ifp;
2918 unsigned long now, next;
2919 int i;
2921 spin_lock_bh(&addrconf_verify_lock);
2922 now = jiffies;
2923 next = now + ADDR_CHECK_FREQUENCY;
2925 del_timer(&addr_chk_timer);
2927 for (i=0; i < IN6_ADDR_HSIZE; i++) {
2929 restart:
2930 read_lock(&addrconf_hash_lock);
2931 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2932 unsigned long age;
2933 #ifdef CONFIG_IPV6_PRIVACY
2934 unsigned long regen_advance;
2935 #endif
2937 if (ifp->flags & IFA_F_PERMANENT)
2938 continue;
2940 spin_lock(&ifp->lock);
2941 age = (now - ifp->tstamp) / HZ;
2943 #ifdef CONFIG_IPV6_PRIVACY
2944 regen_advance = ifp->idev->cnf.regen_max_retry *
2945 ifp->idev->cnf.dad_transmits *
2946 ifp->idev->nd_parms->retrans_time / HZ;
2947 #endif
2949 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2950 age >= ifp->valid_lft) {
2951 spin_unlock(&ifp->lock);
2952 in6_ifa_hold(ifp);
2953 read_unlock(&addrconf_hash_lock);
2954 ipv6_del_addr(ifp);
2955 goto restart;
2956 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2957 spin_unlock(&ifp->lock);
2958 continue;
2959 } else if (age >= ifp->prefered_lft) {
2960 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2961 int deprecate = 0;
2963 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2964 deprecate = 1;
2965 ifp->flags |= IFA_F_DEPRECATED;
2968 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2969 next = ifp->tstamp + ifp->valid_lft * HZ;
2971 spin_unlock(&ifp->lock);
2973 if (deprecate) {
2974 in6_ifa_hold(ifp);
2975 read_unlock(&addrconf_hash_lock);
2977 ipv6_ifa_notify(0, ifp);
2978 in6_ifa_put(ifp);
2979 goto restart;
2981 #ifdef CONFIG_IPV6_PRIVACY
2982 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2983 !(ifp->flags&IFA_F_TENTATIVE)) {
2984 if (age >= ifp->prefered_lft - regen_advance) {
2985 struct inet6_ifaddr *ifpub = ifp->ifpub;
2986 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2987 next = ifp->tstamp + ifp->prefered_lft * HZ;
2988 if (!ifp->regen_count && ifpub) {
2989 ifp->regen_count++;
2990 in6_ifa_hold(ifp);
2991 in6_ifa_hold(ifpub);
2992 spin_unlock(&ifp->lock);
2993 read_unlock(&addrconf_hash_lock);
2994 spin_lock(&ifpub->lock);
2995 ifpub->regen_count = 0;
2996 spin_unlock(&ifpub->lock);
2997 ipv6_create_tempaddr(ifpub, ifp);
2998 in6_ifa_put(ifpub);
2999 in6_ifa_put(ifp);
3000 goto restart;
3002 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3003 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3004 spin_unlock(&ifp->lock);
3005 #endif
3006 } else {
3007 /* ifp->prefered_lft <= ifp->valid_lft */
3008 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3009 next = ifp->tstamp + ifp->prefered_lft * HZ;
3010 spin_unlock(&ifp->lock);
3013 read_unlock(&addrconf_hash_lock);
3016 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3017 add_timer(&addr_chk_timer);
3018 spin_unlock_bh(&addrconf_verify_lock);
3021 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3023 struct in6_addr *pfx = NULL;
3025 if (addr)
3026 pfx = nla_data(addr);
3028 if (local) {
3029 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3030 pfx = NULL;
3031 else
3032 pfx = nla_data(local);
3035 return pfx;
3038 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3039 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3040 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3041 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3044 static int
3045 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3047 struct net *net = skb->sk->sk_net;
3048 struct ifaddrmsg *ifm;
3049 struct nlattr *tb[IFA_MAX+1];
3050 struct in6_addr *pfx;
3051 int err;
3053 if (net != &init_net)
3054 return -EINVAL;
3056 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3057 if (err < 0)
3058 return err;
3060 ifm = nlmsg_data(nlh);
3061 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3062 if (pfx == NULL)
3063 return -EINVAL;
3065 return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3068 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3069 u32 prefered_lft, u32 valid_lft)
3071 u32 flags = RTF_EXPIRES;
3073 if (!valid_lft || (prefered_lft > valid_lft))
3074 return -EINVAL;
3076 if (valid_lft == INFINITY_LIFE_TIME) {
3077 ifa_flags |= IFA_F_PERMANENT;
3078 flags = 0;
3079 } else if (valid_lft >= 0x7FFFFFFF/HZ)
3080 valid_lft = 0x7FFFFFFF/HZ;
3082 if (prefered_lft == 0)
3083 ifa_flags |= IFA_F_DEPRECATED;
3084 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3085 (prefered_lft != INFINITY_LIFE_TIME))
3086 prefered_lft = 0x7FFFFFFF/HZ;
3088 spin_lock_bh(&ifp->lock);
3089 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3090 ifp->tstamp = jiffies;
3091 ifp->valid_lft = valid_lft;
3092 ifp->prefered_lft = prefered_lft;
3094 spin_unlock_bh(&ifp->lock);
3095 if (!(ifp->flags&IFA_F_TENTATIVE))
3096 ipv6_ifa_notify(0, ifp);
3098 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3099 jiffies_to_clock_t(valid_lft * HZ), flags);
3100 addrconf_verify(0);
3102 return 0;
3105 static int
3106 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3108 struct net *net = skb->sk->sk_net;
3109 struct ifaddrmsg *ifm;
3110 struct nlattr *tb[IFA_MAX+1];
3111 struct in6_addr *pfx;
3112 struct inet6_ifaddr *ifa;
3113 struct net_device *dev;
3114 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3115 u8 ifa_flags;
3116 int err;
3118 if (net != &init_net)
3119 return -EINVAL;
3121 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3122 if (err < 0)
3123 return err;
3125 ifm = nlmsg_data(nlh);
3126 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3127 if (pfx == NULL)
3128 return -EINVAL;
3130 if (tb[IFA_CACHEINFO]) {
3131 struct ifa_cacheinfo *ci;
3133 ci = nla_data(tb[IFA_CACHEINFO]);
3134 valid_lft = ci->ifa_valid;
3135 preferred_lft = ci->ifa_prefered;
3136 } else {
3137 preferred_lft = INFINITY_LIFE_TIME;
3138 valid_lft = INFINITY_LIFE_TIME;
3141 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3142 if (dev == NULL)
3143 return -ENODEV;
3145 /* We ignore other flags so far. */
3146 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3148 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3149 if (ifa == NULL) {
3151 * It would be best to check for !NLM_F_CREATE here but
3152 * userspace alreay relies on not having to provide this.
3154 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3155 ifa_flags, preferred_lft, valid_lft);
3158 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3159 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3160 err = -EEXIST;
3161 else
3162 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3164 in6_ifa_put(ifa);
3166 return err;
3169 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3170 u8 scope, int ifindex)
3172 struct ifaddrmsg *ifm;
3174 ifm = nlmsg_data(nlh);
3175 ifm->ifa_family = AF_INET6;
3176 ifm->ifa_prefixlen = prefixlen;
3177 ifm->ifa_flags = flags;
3178 ifm->ifa_scope = scope;
3179 ifm->ifa_index = ifindex;
3182 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3183 unsigned long tstamp, u32 preferred, u32 valid)
3185 struct ifa_cacheinfo ci;
3187 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3188 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3189 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3190 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3191 ci.ifa_prefered = preferred;
3192 ci.ifa_valid = valid;
3194 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3197 static inline int rt_scope(int ifa_scope)
3199 if (ifa_scope & IFA_HOST)
3200 return RT_SCOPE_HOST;
3201 else if (ifa_scope & IFA_LINK)
3202 return RT_SCOPE_LINK;
3203 else if (ifa_scope & IFA_SITE)
3204 return RT_SCOPE_SITE;
3205 else
3206 return RT_SCOPE_UNIVERSE;
3209 static inline int inet6_ifaddr_msgsize(void)
3211 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3212 + nla_total_size(16) /* IFA_ADDRESS */
3213 + nla_total_size(sizeof(struct ifa_cacheinfo));
3216 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3217 u32 pid, u32 seq, int event, unsigned int flags)
3219 struct nlmsghdr *nlh;
3220 u32 preferred, valid;
3222 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3223 if (nlh == NULL)
3224 return -EMSGSIZE;
3226 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3227 ifa->idev->dev->ifindex);
3229 if (!(ifa->flags&IFA_F_PERMANENT)) {
3230 preferred = ifa->prefered_lft;
3231 valid = ifa->valid_lft;
3232 if (preferred != INFINITY_LIFE_TIME) {
3233 long tval = (jiffies - ifa->tstamp)/HZ;
3234 preferred -= tval;
3235 if (valid != INFINITY_LIFE_TIME)
3236 valid -= tval;
3238 } else {
3239 preferred = INFINITY_LIFE_TIME;
3240 valid = INFINITY_LIFE_TIME;
3243 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3244 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3245 nlmsg_cancel(skb, nlh);
3246 return -EMSGSIZE;
3249 return nlmsg_end(skb, nlh);
3252 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3253 u32 pid, u32 seq, int event, u16 flags)
3255 struct nlmsghdr *nlh;
3256 u8 scope = RT_SCOPE_UNIVERSE;
3257 int ifindex = ifmca->idev->dev->ifindex;
3259 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3260 scope = RT_SCOPE_SITE;
3262 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3263 if (nlh == NULL)
3264 return -EMSGSIZE;
3266 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3267 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3268 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3269 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3270 nlmsg_cancel(skb, nlh);
3271 return -EMSGSIZE;
3274 return nlmsg_end(skb, nlh);
3277 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3278 u32 pid, u32 seq, int event, unsigned int flags)
3280 struct nlmsghdr *nlh;
3281 u8 scope = RT_SCOPE_UNIVERSE;
3282 int ifindex = ifaca->aca_idev->dev->ifindex;
3284 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3285 scope = RT_SCOPE_SITE;
3287 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3288 if (nlh == NULL)
3289 return -EMSGSIZE;
3291 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3292 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3293 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3294 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3295 nlmsg_cancel(skb, nlh);
3296 return -EMSGSIZE;
3299 return nlmsg_end(skb, nlh);
3302 enum addr_type_t
3304 UNICAST_ADDR,
3305 MULTICAST_ADDR,
3306 ANYCAST_ADDR,
3309 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3310 enum addr_type_t type)
3312 int idx, ip_idx;
3313 int s_idx, s_ip_idx;
3314 int err = 1;
3315 struct net_device *dev;
3316 struct inet6_dev *idev = NULL;
3317 struct inet6_ifaddr *ifa;
3318 struct ifmcaddr6 *ifmca;
3319 struct ifacaddr6 *ifaca;
3321 s_idx = cb->args[0];
3322 s_ip_idx = ip_idx = cb->args[1];
3324 idx = 0;
3325 for_each_netdev(&init_net, dev) {
3326 if (idx < s_idx)
3327 goto cont;
3328 if (idx > s_idx)
3329 s_ip_idx = 0;
3330 ip_idx = 0;
3331 if ((idev = in6_dev_get(dev)) == NULL)
3332 goto cont;
3333 read_lock_bh(&idev->lock);
3334 switch (type) {
3335 case UNICAST_ADDR:
3336 /* unicast address incl. temp addr */
3337 for (ifa = idev->addr_list; ifa;
3338 ifa = ifa->if_next, ip_idx++) {
3339 if (ip_idx < s_ip_idx)
3340 continue;
3341 err = inet6_fill_ifaddr(skb, ifa,
3342 NETLINK_CB(cb->skb).pid,
3343 cb->nlh->nlmsg_seq,
3344 RTM_NEWADDR,
3345 NLM_F_MULTI);
3347 break;
3348 case MULTICAST_ADDR:
3349 /* multicast address */
3350 for (ifmca = idev->mc_list; ifmca;
3351 ifmca = ifmca->next, ip_idx++) {
3352 if (ip_idx < s_ip_idx)
3353 continue;
3354 err = inet6_fill_ifmcaddr(skb, ifmca,
3355 NETLINK_CB(cb->skb).pid,
3356 cb->nlh->nlmsg_seq,
3357 RTM_GETMULTICAST,
3358 NLM_F_MULTI);
3360 break;
3361 case ANYCAST_ADDR:
3362 /* anycast address */
3363 for (ifaca = idev->ac_list; ifaca;
3364 ifaca = ifaca->aca_next, ip_idx++) {
3365 if (ip_idx < s_ip_idx)
3366 continue;
3367 err = inet6_fill_ifacaddr(skb, ifaca,
3368 NETLINK_CB(cb->skb).pid,
3369 cb->nlh->nlmsg_seq,
3370 RTM_GETANYCAST,
3371 NLM_F_MULTI);
3373 break;
3374 default:
3375 break;
3377 read_unlock_bh(&idev->lock);
3378 in6_dev_put(idev);
3380 if (err <= 0)
3381 break;
3382 cont:
3383 idx++;
3385 cb->args[0] = idx;
3386 cb->args[1] = ip_idx;
3387 return skb->len;
3390 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3392 struct net *net = skb->sk->sk_net;
3393 enum addr_type_t type = UNICAST_ADDR;
3395 if (net != &init_net)
3396 return 0;
3398 return inet6_dump_addr(skb, cb, type);
3401 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3403 struct net *net = skb->sk->sk_net;
3404 enum addr_type_t type = MULTICAST_ADDR;
3406 if (net != &init_net)
3407 return 0;
3409 return inet6_dump_addr(skb, cb, type);
3413 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3415 struct net *net = skb->sk->sk_net;
3416 enum addr_type_t type = ANYCAST_ADDR;
3418 if (net != &init_net)
3419 return 0;
3421 return inet6_dump_addr(skb, cb, type);
3424 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3425 void *arg)
3427 struct net *net = in_skb->sk->sk_net;
3428 struct ifaddrmsg *ifm;
3429 struct nlattr *tb[IFA_MAX+1];
3430 struct in6_addr *addr = NULL;
3431 struct net_device *dev = NULL;
3432 struct inet6_ifaddr *ifa;
3433 struct sk_buff *skb;
3434 int err;
3436 if (net != &init_net)
3437 return -EINVAL;
3439 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3440 if (err < 0)
3441 goto errout;
3443 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3444 if (addr == NULL) {
3445 err = -EINVAL;
3446 goto errout;
3449 ifm = nlmsg_data(nlh);
3450 if (ifm->ifa_index)
3451 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3453 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3454 err = -EADDRNOTAVAIL;
3455 goto errout;
3458 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3459 err = -ENOBUFS;
3460 goto errout_ifa;
3463 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3464 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3465 if (err < 0) {
3466 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3467 WARN_ON(err == -EMSGSIZE);
3468 kfree_skb(skb);
3469 goto errout_ifa;
3471 err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
3472 errout_ifa:
3473 in6_ifa_put(ifa);
3474 errout:
3475 return err;
3478 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3480 struct sk_buff *skb;
3481 int err = -ENOBUFS;
3483 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3484 if (skb == NULL)
3485 goto errout;
3487 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3488 if (err < 0) {
3489 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3490 WARN_ON(err == -EMSGSIZE);
3491 kfree_skb(skb);
3492 goto errout;
3494 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3495 errout:
3496 if (err < 0)
3497 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3500 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3501 __s32 *array, int bytes)
3503 BUG_ON(bytes < (DEVCONF_MAX * 4));
3505 memset(array, 0, bytes);
3506 array[DEVCONF_FORWARDING] = cnf->forwarding;
3507 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3508 array[DEVCONF_MTU6] = cnf->mtu6;
3509 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3510 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3511 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3512 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3513 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3514 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3515 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3516 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3517 #ifdef CONFIG_IPV6_PRIVACY
3518 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3519 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3520 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3521 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3522 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3523 #endif
3524 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3525 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3526 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3527 #ifdef CONFIG_IPV6_ROUTER_PREF
3528 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3529 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3530 #ifdef CONFIG_IPV6_ROUTE_INFO
3531 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3532 #endif
3533 #endif
3534 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3535 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3536 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3537 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3538 #endif
3541 static inline size_t inet6_if_nlmsg_size(void)
3543 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3544 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3545 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3546 + nla_total_size(4) /* IFLA_MTU */
3547 + nla_total_size(4) /* IFLA_LINK */
3548 + nla_total_size( /* IFLA_PROTINFO */
3549 nla_total_size(4) /* IFLA_INET6_FLAGS */
3550 + nla_total_size(sizeof(struct ifla_cacheinfo))
3551 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3552 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3553 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3557 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3558 int bytes)
3560 int i;
3561 int pad = bytes - sizeof(u64) * items;
3562 BUG_ON(pad < 0);
3564 /* Use put_unaligned() because stats may not be aligned for u64. */
3565 put_unaligned(items, &stats[0]);
3566 for (i = 1; i < items; i++)
3567 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3569 memset(&stats[items], 0, pad);
3572 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3573 int bytes)
3575 switch(attrtype) {
3576 case IFLA_INET6_STATS:
3577 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3578 break;
3579 case IFLA_INET6_ICMP6STATS:
3580 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3581 break;
3585 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3586 u32 pid, u32 seq, int event, unsigned int flags)
3588 struct net_device *dev = idev->dev;
3589 struct nlattr *nla;
3590 struct ifinfomsg *hdr;
3591 struct nlmsghdr *nlh;
3592 void *protoinfo;
3593 struct ifla_cacheinfo ci;
3595 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3596 if (nlh == NULL)
3597 return -EMSGSIZE;
3599 hdr = nlmsg_data(nlh);
3600 hdr->ifi_family = AF_INET6;
3601 hdr->__ifi_pad = 0;
3602 hdr->ifi_type = dev->type;
3603 hdr->ifi_index = dev->ifindex;
3604 hdr->ifi_flags = dev_get_flags(dev);
3605 hdr->ifi_change = 0;
3607 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3609 if (dev->addr_len)
3610 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3612 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3613 if (dev->ifindex != dev->iflink)
3614 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3616 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3617 if (protoinfo == NULL)
3618 goto nla_put_failure;
3620 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3622 ci.max_reasm_len = IPV6_MAXPLEN;
3623 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3624 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3625 ci.reachable_time = idev->nd_parms->reachable_time;
3626 ci.retrans_time = idev->nd_parms->retrans_time;
3627 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3629 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3630 if (nla == NULL)
3631 goto nla_put_failure;
3632 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3634 /* XXX - MC not implemented */
3636 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3637 if (nla == NULL)
3638 goto nla_put_failure;
3639 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3641 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3642 if (nla == NULL)
3643 goto nla_put_failure;
3644 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3646 nla_nest_end(skb, protoinfo);
3647 return nlmsg_end(skb, nlh);
3649 nla_put_failure:
3650 nlmsg_cancel(skb, nlh);
3651 return -EMSGSIZE;
3654 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3656 struct net *net = skb->sk->sk_net;
3657 int idx, err;
3658 int s_idx = cb->args[0];
3659 struct net_device *dev;
3660 struct inet6_dev *idev;
3662 if (net != &init_net)
3663 return 0;
3665 read_lock(&dev_base_lock);
3666 idx = 0;
3667 for_each_netdev(&init_net, dev) {
3668 if (idx < s_idx)
3669 goto cont;
3670 if ((idev = in6_dev_get(dev)) == NULL)
3671 goto cont;
3672 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3673 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3674 in6_dev_put(idev);
3675 if (err <= 0)
3676 break;
3677 cont:
3678 idx++;
3680 read_unlock(&dev_base_lock);
3681 cb->args[0] = idx;
3683 return skb->len;
3686 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3688 struct sk_buff *skb;
3689 int err = -ENOBUFS;
3691 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3692 if (skb == NULL)
3693 goto errout;
3695 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3696 if (err < 0) {
3697 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3698 WARN_ON(err == -EMSGSIZE);
3699 kfree_skb(skb);
3700 goto errout;
3702 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3703 errout:
3704 if (err < 0)
3705 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3708 static inline size_t inet6_prefix_nlmsg_size(void)
3710 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3711 + nla_total_size(sizeof(struct in6_addr))
3712 + nla_total_size(sizeof(struct prefix_cacheinfo));
3715 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3716 struct prefix_info *pinfo, u32 pid, u32 seq,
3717 int event, unsigned int flags)
3719 struct prefixmsg *pmsg;
3720 struct nlmsghdr *nlh;
3721 struct prefix_cacheinfo ci;
3723 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3724 if (nlh == NULL)
3725 return -EMSGSIZE;
3727 pmsg = nlmsg_data(nlh);
3728 pmsg->prefix_family = AF_INET6;
3729 pmsg->prefix_pad1 = 0;
3730 pmsg->prefix_pad2 = 0;
3731 pmsg->prefix_ifindex = idev->dev->ifindex;
3732 pmsg->prefix_len = pinfo->prefix_len;
3733 pmsg->prefix_type = pinfo->type;
3734 pmsg->prefix_pad3 = 0;
3735 pmsg->prefix_flags = 0;
3736 if (pinfo->onlink)
3737 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3738 if (pinfo->autoconf)
3739 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3741 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3743 ci.preferred_time = ntohl(pinfo->prefered);
3744 ci.valid_time = ntohl(pinfo->valid);
3745 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3747 return nlmsg_end(skb, nlh);
3749 nla_put_failure:
3750 nlmsg_cancel(skb, nlh);
3751 return -EMSGSIZE;
3754 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3755 struct prefix_info *pinfo)
3757 struct sk_buff *skb;
3758 int err = -ENOBUFS;
3760 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3761 if (skb == NULL)
3762 goto errout;
3764 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3765 if (err < 0) {
3766 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3767 WARN_ON(err == -EMSGSIZE);
3768 kfree_skb(skb);
3769 goto errout;
3771 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3772 errout:
3773 if (err < 0)
3774 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_PREFIX, err);
3777 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3779 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3781 switch (event) {
3782 case RTM_NEWADDR:
3784 * If the address was optimistic
3785 * we inserted the route at the start of
3786 * our DAD process, so we don't need
3787 * to do it again
3789 if (!(ifp->rt->rt6i_node))
3790 ip6_ins_rt(ifp->rt);
3791 if (ifp->idev->cnf.forwarding)
3792 addrconf_join_anycast(ifp);
3793 break;
3794 case RTM_DELADDR:
3795 if (ifp->idev->cnf.forwarding)
3796 addrconf_leave_anycast(ifp);
3797 addrconf_leave_solict(ifp->idev, &ifp->addr);
3798 dst_hold(&ifp->rt->u.dst);
3799 if (ip6_del_rt(ifp->rt))
3800 dst_free(&ifp->rt->u.dst);
3801 break;
3805 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3807 rcu_read_lock_bh();
3808 if (likely(ifp->idev->dead == 0))
3809 __ipv6_ifa_notify(event, ifp);
3810 rcu_read_unlock_bh();
3813 #ifdef CONFIG_SYSCTL
3815 static
3816 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3817 void __user *buffer, size_t *lenp, loff_t *ppos)
3819 int *valp = ctl->data;
3820 int val = *valp;
3821 int ret;
3823 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3825 if (write)
3826 addrconf_fixup_forwarding(ctl, valp, val);
3827 return ret;
3830 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3831 int __user *name, int nlen,
3832 void __user *oldval,
3833 size_t __user *oldlenp,
3834 void __user *newval, size_t newlen)
3836 int *valp = table->data;
3837 int val = *valp;
3838 int new;
3840 if (!newval || !newlen)
3841 return 0;
3842 if (newlen != sizeof(int))
3843 return -EINVAL;
3844 if (get_user(new, (int __user *)newval))
3845 return -EFAULT;
3846 if (new == *valp)
3847 return 0;
3848 if (oldval && oldlenp) {
3849 size_t len;
3850 if (get_user(len, oldlenp))
3851 return -EFAULT;
3852 if (len) {
3853 if (len > table->maxlen)
3854 len = table->maxlen;
3855 if (copy_to_user(oldval, valp, len))
3856 return -EFAULT;
3857 if (put_user(len, oldlenp))
3858 return -EFAULT;
3862 *valp = new;
3863 addrconf_fixup_forwarding(table, valp, val);
3864 return 1;
3867 static struct addrconf_sysctl_table
3869 struct ctl_table_header *sysctl_header;
3870 ctl_table addrconf_vars[__NET_IPV6_MAX];
3871 char *dev_name;
3872 } addrconf_sysctl __read_mostly = {
3873 .sysctl_header = NULL,
3874 .addrconf_vars = {
3876 .ctl_name = NET_IPV6_FORWARDING,
3877 .procname = "forwarding",
3878 .data = &ipv6_devconf.forwarding,
3879 .maxlen = sizeof(int),
3880 .mode = 0644,
3881 .proc_handler = &addrconf_sysctl_forward,
3882 .strategy = &addrconf_sysctl_forward_strategy,
3885 .ctl_name = NET_IPV6_HOP_LIMIT,
3886 .procname = "hop_limit",
3887 .data = &ipv6_devconf.hop_limit,
3888 .maxlen = sizeof(int),
3889 .mode = 0644,
3890 .proc_handler = proc_dointvec,
3893 .ctl_name = NET_IPV6_MTU,
3894 .procname = "mtu",
3895 .data = &ipv6_devconf.mtu6,
3896 .maxlen = sizeof(int),
3897 .mode = 0644,
3898 .proc_handler = &proc_dointvec,
3901 .ctl_name = NET_IPV6_ACCEPT_RA,
3902 .procname = "accept_ra",
3903 .data = &ipv6_devconf.accept_ra,
3904 .maxlen = sizeof(int),
3905 .mode = 0644,
3906 .proc_handler = &proc_dointvec,
3909 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
3910 .procname = "accept_redirects",
3911 .data = &ipv6_devconf.accept_redirects,
3912 .maxlen = sizeof(int),
3913 .mode = 0644,
3914 .proc_handler = &proc_dointvec,
3917 .ctl_name = NET_IPV6_AUTOCONF,
3918 .procname = "autoconf",
3919 .data = &ipv6_devconf.autoconf,
3920 .maxlen = sizeof(int),
3921 .mode = 0644,
3922 .proc_handler = &proc_dointvec,
3925 .ctl_name = NET_IPV6_DAD_TRANSMITS,
3926 .procname = "dad_transmits",
3927 .data = &ipv6_devconf.dad_transmits,
3928 .maxlen = sizeof(int),
3929 .mode = 0644,
3930 .proc_handler = &proc_dointvec,
3933 .ctl_name = NET_IPV6_RTR_SOLICITS,
3934 .procname = "router_solicitations",
3935 .data = &ipv6_devconf.rtr_solicits,
3936 .maxlen = sizeof(int),
3937 .mode = 0644,
3938 .proc_handler = &proc_dointvec,
3941 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
3942 .procname = "router_solicitation_interval",
3943 .data = &ipv6_devconf.rtr_solicit_interval,
3944 .maxlen = sizeof(int),
3945 .mode = 0644,
3946 .proc_handler = &proc_dointvec_jiffies,
3947 .strategy = &sysctl_jiffies,
3950 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
3951 .procname = "router_solicitation_delay",
3952 .data = &ipv6_devconf.rtr_solicit_delay,
3953 .maxlen = sizeof(int),
3954 .mode = 0644,
3955 .proc_handler = &proc_dointvec_jiffies,
3956 .strategy = &sysctl_jiffies,
3959 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
3960 .procname = "force_mld_version",
3961 .data = &ipv6_devconf.force_mld_version,
3962 .maxlen = sizeof(int),
3963 .mode = 0644,
3964 .proc_handler = &proc_dointvec,
3966 #ifdef CONFIG_IPV6_PRIVACY
3968 .ctl_name = NET_IPV6_USE_TEMPADDR,
3969 .procname = "use_tempaddr",
3970 .data = &ipv6_devconf.use_tempaddr,
3971 .maxlen = sizeof(int),
3972 .mode = 0644,
3973 .proc_handler = &proc_dointvec,
3976 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
3977 .procname = "temp_valid_lft",
3978 .data = &ipv6_devconf.temp_valid_lft,
3979 .maxlen = sizeof(int),
3980 .mode = 0644,
3981 .proc_handler = &proc_dointvec,
3984 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
3985 .procname = "temp_prefered_lft",
3986 .data = &ipv6_devconf.temp_prefered_lft,
3987 .maxlen = sizeof(int),
3988 .mode = 0644,
3989 .proc_handler = &proc_dointvec,
3992 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
3993 .procname = "regen_max_retry",
3994 .data = &ipv6_devconf.regen_max_retry,
3995 .maxlen = sizeof(int),
3996 .mode = 0644,
3997 .proc_handler = &proc_dointvec,
4000 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4001 .procname = "max_desync_factor",
4002 .data = &ipv6_devconf.max_desync_factor,
4003 .maxlen = sizeof(int),
4004 .mode = 0644,
4005 .proc_handler = &proc_dointvec,
4007 #endif
4009 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4010 .procname = "max_addresses",
4011 .data = &ipv6_devconf.max_addresses,
4012 .maxlen = sizeof(int),
4013 .mode = 0644,
4014 .proc_handler = &proc_dointvec,
4017 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4018 .procname = "accept_ra_defrtr",
4019 .data = &ipv6_devconf.accept_ra_defrtr,
4020 .maxlen = sizeof(int),
4021 .mode = 0644,
4022 .proc_handler = &proc_dointvec,
4025 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4026 .procname = "accept_ra_pinfo",
4027 .data = &ipv6_devconf.accept_ra_pinfo,
4028 .maxlen = sizeof(int),
4029 .mode = 0644,
4030 .proc_handler = &proc_dointvec,
4032 #ifdef CONFIG_IPV6_ROUTER_PREF
4034 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4035 .procname = "accept_ra_rtr_pref",
4036 .data = &ipv6_devconf.accept_ra_rtr_pref,
4037 .maxlen = sizeof(int),
4038 .mode = 0644,
4039 .proc_handler = &proc_dointvec,
4042 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4043 .procname = "router_probe_interval",
4044 .data = &ipv6_devconf.rtr_probe_interval,
4045 .maxlen = sizeof(int),
4046 .mode = 0644,
4047 .proc_handler = &proc_dointvec_jiffies,
4048 .strategy = &sysctl_jiffies,
4050 #ifdef CONFIG_IPV6_ROUTE_INFO
4052 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4053 .procname = "accept_ra_rt_info_max_plen",
4054 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4055 .maxlen = sizeof(int),
4056 .mode = 0644,
4057 .proc_handler = &proc_dointvec,
4059 #endif
4060 #endif
4062 .ctl_name = NET_IPV6_PROXY_NDP,
4063 .procname = "proxy_ndp",
4064 .data = &ipv6_devconf.proxy_ndp,
4065 .maxlen = sizeof(int),
4066 .mode = 0644,
4067 .proc_handler = &proc_dointvec,
4070 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4071 .procname = "accept_source_route",
4072 .data = &ipv6_devconf.accept_source_route,
4073 .maxlen = sizeof(int),
4074 .mode = 0644,
4075 .proc_handler = &proc_dointvec,
4077 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4079 .ctl_name = CTL_UNNUMBERED,
4080 .procname = "optimistic_dad",
4081 .data = &ipv6_devconf.optimistic_dad,
4082 .maxlen = sizeof(int),
4083 .mode = 0644,
4084 .proc_handler = &proc_dointvec,
4087 #endif
4089 .ctl_name = 0, /* sentinel */
4094 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4095 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4097 int i;
4098 struct addrconf_sysctl_table *t;
4100 #define ADDRCONF_CTL_PATH_DEV 3
4102 struct ctl_path addrconf_ctl_path[] = {
4103 { .procname = "net", .ctl_name = CTL_NET, },
4104 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4105 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4106 { /* to be set */ },
4107 { },
4111 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4112 if (t == NULL)
4113 goto out;
4115 for (i=0; t->addrconf_vars[i].data; i++) {
4116 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4117 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4118 t->addrconf_vars[i].extra2 = net;
4122 * Make a copy of dev_name, because '.procname' is regarded as const
4123 * by sysctl and we wouldn't want anyone to change it under our feet
4124 * (see SIOCSIFNAME).
4126 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4127 if (!t->dev_name)
4128 goto free;
4130 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4131 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4133 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4134 t->addrconf_vars);
4135 if (t->sysctl_header == NULL)
4136 goto free_procname;
4138 p->sysctl = t;
4139 return 0;
4141 free_procname:
4142 kfree(t->dev_name);
4143 free:
4144 kfree(t);
4145 out:
4146 return -ENOBUFS;
4149 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4151 struct addrconf_sysctl_table *t;
4153 if (p->sysctl == NULL)
4154 return;
4156 t = p->sysctl;
4157 p->sysctl = NULL;
4158 unregister_sysctl_table(t->sysctl_header);
4159 kfree(t->dev_name);
4160 kfree(t);
4163 static void addrconf_sysctl_register(struct inet6_dev *idev)
4165 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4166 NET_IPV6_NEIGH, "ipv6",
4167 &ndisc_ifinfo_sysctl_change,
4168 NULL);
4169 __addrconf_sysctl_register(idev->dev->nd_net, idev->dev->name,
4170 idev->dev->ifindex, idev, &idev->cnf);
4173 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4175 __addrconf_sysctl_unregister(&idev->cnf);
4176 neigh_sysctl_unregister(idev->nd_parms);
4180 #endif
4182 static int addrconf_init_net(struct net *net)
4184 int err;
4185 struct ipv6_devconf *all, *dflt;
4187 err = -ENOMEM;
4188 all = &ipv6_devconf;
4189 dflt = &ipv6_devconf_dflt;
4191 if (net != &init_net) {
4192 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4193 if (all == NULL)
4194 goto err_alloc_all;
4196 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4197 if (dflt == NULL)
4198 goto err_alloc_dflt;
4201 net->ipv6.devconf_all = all;
4202 net->ipv6.devconf_dflt = dflt;
4204 #ifdef CONFIG_SYSCTL
4205 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4206 NULL, all);
4207 if (err < 0)
4208 goto err_reg_all;
4210 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4211 NULL, dflt);
4212 if (err < 0)
4213 goto err_reg_dflt;
4214 #endif
4215 return 0;
4217 #ifdef CONFIG_SYSCTL
4218 err_reg_dflt:
4219 __addrconf_sysctl_unregister(all);
4220 err_reg_all:
4221 kfree(dflt);
4222 #endif
4223 err_alloc_dflt:
4224 kfree(all);
4225 err_alloc_all:
4226 return err;
4229 static void addrconf_exit_net(struct net *net)
4231 #ifdef CONFIG_SYSCTL
4232 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4233 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4234 #endif
4235 if (net != &init_net) {
4236 kfree(net->ipv6.devconf_dflt);
4237 kfree(net->ipv6.devconf_all);
4241 static struct pernet_operations addrconf_ops = {
4242 .init = addrconf_init_net,
4243 .exit = addrconf_exit_net,
4247 * Device notifier
4250 int register_inet6addr_notifier(struct notifier_block *nb)
4252 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4255 EXPORT_SYMBOL(register_inet6addr_notifier);
4257 int unregister_inet6addr_notifier(struct notifier_block *nb)
4259 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4262 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4265 * Init / cleanup code
4268 int __init addrconf_init(void)
4270 int err;
4272 if ((err = ipv6_addr_label_init()) < 0) {
4273 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4274 err);
4275 return err;
4278 register_pernet_subsys(&addrconf_ops);
4280 /* The addrconf netdev notifier requires that loopback_dev
4281 * has it's ipv6 private information allocated and setup
4282 * before it can bring up and give link-local addresses
4283 * to other devices which are up.
4285 * Unfortunately, loopback_dev is not necessarily the first
4286 * entry in the global dev_base list of net devices. In fact,
4287 * it is likely to be the very last entry on that list.
4288 * So this causes the notifier registry below to try and
4289 * give link-local addresses to all devices besides loopback_dev
4290 * first, then loopback_dev, which cases all the non-loopback_dev
4291 * devices to fail to get a link-local address.
4293 * So, as a temporary fix, allocate the ipv6 structure for
4294 * loopback_dev first by hand.
4295 * Longer term, all of the dependencies ipv6 has upon the loopback
4296 * device and it being up should be removed.
4298 rtnl_lock();
4299 if (!ipv6_add_dev(init_net.loopback_dev))
4300 err = -ENOMEM;
4301 rtnl_unlock();
4302 if (err)
4303 goto errlo;
4305 ip6_null_entry.u.dst.dev = init_net.loopback_dev;
4306 ip6_null_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4307 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4308 ip6_prohibit_entry.u.dst.dev = init_net.loopback_dev;
4309 ip6_prohibit_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4310 ip6_blk_hole_entry.u.dst.dev = init_net.loopback_dev;
4311 ip6_blk_hole_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4312 #endif
4314 register_netdevice_notifier(&ipv6_dev_notf);
4316 addrconf_verify(0);
4318 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4319 if (err < 0)
4320 goto errout;
4322 /* Only the first call to __rtnl_register can fail */
4323 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4324 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4325 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4326 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4327 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4329 ipv6_addr_label_rtnl_register();
4331 return 0;
4332 errout:
4333 unregister_netdevice_notifier(&ipv6_dev_notf);
4334 errlo:
4335 unregister_pernet_subsys(&addrconf_ops);
4337 return err;
4340 void addrconf_cleanup(void)
4342 struct net_device *dev;
4343 struct inet6_ifaddr *ifa;
4344 int i;
4346 unregister_netdevice_notifier(&ipv6_dev_notf);
4348 unregister_pernet_subsys(&addrconf_ops);
4350 rtnl_lock();
4353 * clean dev list.
4356 for_each_netdev(&init_net, dev) {
4357 if (__in6_dev_get(dev) == NULL)
4358 continue;
4359 addrconf_ifdown(dev, 1);
4361 addrconf_ifdown(init_net.loopback_dev, 2);
4364 * Check hash table.
4367 write_lock_bh(&addrconf_hash_lock);
4368 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4369 for (ifa=inet6_addr_lst[i]; ifa; ) {
4370 struct inet6_ifaddr *bifa;
4372 bifa = ifa;
4373 ifa = ifa->lst_next;
4374 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4375 /* Do not free it; something is wrong.
4376 Now we can investigate it with debugger.
4380 write_unlock_bh(&addrconf_hash_lock);
4382 del_timer(&addr_chk_timer);
4384 rtnl_unlock();