x86: implement gbpages support in change_page_attr()
[wrt350n-kernel.git] / net / ipv6 / addrconf.c
blobe40213db9e4c6355e665c8684e4549feca77614d
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,
1562 ipv6_addr_copy(&cfg.fc_dst, pfx);
1564 /* Prevent useless cloning on PtP SIT.
1565 This thing is done here expecting that the whole
1566 class of non-broadcast devices need not cloning.
1568 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1569 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1570 cfg.fc_flags |= RTF_NONEXTHOP;
1571 #endif
1573 ip6_route_add(&cfg);
1576 /* Create "default" multicast route to the interface */
1578 static void addrconf_add_mroute(struct net_device *dev)
1580 struct fib6_config cfg = {
1581 .fc_table = RT6_TABLE_LOCAL,
1582 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1583 .fc_ifindex = dev->ifindex,
1584 .fc_dst_len = 8,
1585 .fc_flags = RTF_UP,
1588 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1590 ip6_route_add(&cfg);
1593 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1594 static void sit_route_add(struct net_device *dev)
1596 struct fib6_config cfg = {
1597 .fc_table = RT6_TABLE_MAIN,
1598 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1599 .fc_ifindex = dev->ifindex,
1600 .fc_dst_len = 96,
1601 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1604 /* prefix length - 96 bits "::d.d.d.d" */
1605 ip6_route_add(&cfg);
1607 #endif
1609 static void addrconf_add_lroute(struct net_device *dev)
1611 struct in6_addr addr;
1613 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1614 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1617 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1619 struct inet6_dev *idev;
1621 ASSERT_RTNL();
1623 if ((idev = ipv6_find_idev(dev)) == NULL)
1624 return NULL;
1626 /* Add default multicast route */
1627 addrconf_add_mroute(dev);
1629 /* Add link local route */
1630 addrconf_add_lroute(dev);
1631 return idev;
1634 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1636 struct prefix_info *pinfo;
1637 __u32 valid_lft;
1638 __u32 prefered_lft;
1639 int addr_type;
1640 unsigned long rt_expires;
1641 struct inet6_dev *in6_dev;
1643 pinfo = (struct prefix_info *) opt;
1645 if (len < sizeof(struct prefix_info)) {
1646 ADBG(("addrconf: prefix option too short\n"));
1647 return;
1651 * Validation checks ([ADDRCONF], page 19)
1654 addr_type = ipv6_addr_type(&pinfo->prefix);
1656 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1657 return;
1659 valid_lft = ntohl(pinfo->valid);
1660 prefered_lft = ntohl(pinfo->prefered);
1662 if (prefered_lft > valid_lft) {
1663 if (net_ratelimit())
1664 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1665 return;
1668 in6_dev = in6_dev_get(dev);
1670 if (in6_dev == NULL) {
1671 if (net_ratelimit())
1672 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1673 return;
1677 * Two things going on here:
1678 * 1) Add routes for on-link prefixes
1679 * 2) Configure prefixes with the auto flag set
1682 /* Avoid arithmetic overflow. Really, we could
1683 save rt_expires in seconds, likely valid_lft,
1684 but it would require division in fib gc, that it
1685 not good.
1687 if (valid_lft >= 0x7FFFFFFF/HZ)
1688 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1689 else
1690 rt_expires = valid_lft * HZ;
1693 * We convert this (in jiffies) to clock_t later.
1694 * Avoid arithmetic overflow there as well.
1695 * Overflow can happen only if HZ < USER_HZ.
1697 if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1698 rt_expires = 0x7FFFFFFF / USER_HZ;
1700 if (pinfo->onlink) {
1701 struct rt6_info *rt;
1702 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1704 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1705 if (rt->rt6i_flags&RTF_EXPIRES) {
1706 if (valid_lft == 0) {
1707 ip6_del_rt(rt);
1708 rt = NULL;
1709 } else {
1710 rt->rt6i_expires = jiffies + rt_expires;
1713 } else if (valid_lft) {
1714 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1715 dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1717 if (rt)
1718 dst_release(&rt->u.dst);
1721 /* Try to figure out our local address for this prefix */
1723 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1724 struct inet6_ifaddr * ifp;
1725 struct in6_addr addr;
1726 int create = 0, update_lft = 0;
1728 if (pinfo->prefix_len == 64) {
1729 memcpy(&addr, &pinfo->prefix, 8);
1730 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1731 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1732 in6_dev_put(in6_dev);
1733 return;
1735 goto ok;
1737 if (net_ratelimit())
1738 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1739 pinfo->prefix_len);
1740 in6_dev_put(in6_dev);
1741 return;
1745 ifp = ipv6_get_ifaddr(&init_net, &addr, dev, 1);
1747 if (ifp == NULL && valid_lft) {
1748 int max_addresses = in6_dev->cnf.max_addresses;
1749 u32 addr_flags = 0;
1751 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1752 if (in6_dev->cnf.optimistic_dad &&
1753 !ipv6_devconf.forwarding)
1754 addr_flags = IFA_F_OPTIMISTIC;
1755 #endif
1757 /* Do not allow to create too much of autoconfigured
1758 * addresses; this would be too easy way to crash kernel.
1760 if (!max_addresses ||
1761 ipv6_count_addresses(in6_dev) < max_addresses)
1762 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1763 addr_type&IPV6_ADDR_SCOPE_MASK,
1764 addr_flags);
1766 if (!ifp || IS_ERR(ifp)) {
1767 in6_dev_put(in6_dev);
1768 return;
1771 update_lft = create = 1;
1772 ifp->cstamp = jiffies;
1773 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1776 if (ifp) {
1777 int flags;
1778 unsigned long now;
1779 #ifdef CONFIG_IPV6_PRIVACY
1780 struct inet6_ifaddr *ift;
1781 #endif
1782 u32 stored_lft;
1784 /* update lifetime (RFC2462 5.5.3 e) */
1785 spin_lock(&ifp->lock);
1786 now = jiffies;
1787 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1788 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1789 else
1790 stored_lft = 0;
1791 if (!update_lft && stored_lft) {
1792 if (valid_lft > MIN_VALID_LIFETIME ||
1793 valid_lft > stored_lft)
1794 update_lft = 1;
1795 else if (stored_lft <= MIN_VALID_LIFETIME) {
1796 /* valid_lft <= stored_lft is always true */
1797 /* XXX: IPsec */
1798 update_lft = 0;
1799 } else {
1800 valid_lft = MIN_VALID_LIFETIME;
1801 if (valid_lft < prefered_lft)
1802 prefered_lft = valid_lft;
1803 update_lft = 1;
1807 if (update_lft) {
1808 ifp->valid_lft = valid_lft;
1809 ifp->prefered_lft = prefered_lft;
1810 ifp->tstamp = now;
1811 flags = ifp->flags;
1812 ifp->flags &= ~IFA_F_DEPRECATED;
1813 spin_unlock(&ifp->lock);
1815 if (!(flags&IFA_F_TENTATIVE))
1816 ipv6_ifa_notify(0, ifp);
1817 } else
1818 spin_unlock(&ifp->lock);
1820 #ifdef CONFIG_IPV6_PRIVACY
1821 read_lock_bh(&in6_dev->lock);
1822 /* update all temporary addresses in the list */
1823 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1825 * When adjusting the lifetimes of an existing
1826 * temporary address, only lower the lifetimes.
1827 * Implementations must not increase the
1828 * lifetimes of an existing temporary address
1829 * when processing a Prefix Information Option.
1831 spin_lock(&ift->lock);
1832 flags = ift->flags;
1833 if (ift->valid_lft > valid_lft &&
1834 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1835 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1836 if (ift->prefered_lft > prefered_lft &&
1837 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1838 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1839 spin_unlock(&ift->lock);
1840 if (!(flags&IFA_F_TENTATIVE))
1841 ipv6_ifa_notify(0, ift);
1844 if (create && in6_dev->cnf.use_tempaddr > 0) {
1846 * When a new public address is created as described in [ADDRCONF],
1847 * also create a new temporary address.
1849 read_unlock_bh(&in6_dev->lock);
1850 ipv6_create_tempaddr(ifp, NULL);
1851 } else {
1852 read_unlock_bh(&in6_dev->lock);
1854 #endif
1855 in6_ifa_put(ifp);
1856 addrconf_verify(0);
1859 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1860 in6_dev_put(in6_dev);
1864 * Set destination address.
1865 * Special case for SIT interfaces where we create a new "virtual"
1866 * device.
1868 int addrconf_set_dstaddr(void __user *arg)
1870 struct in6_ifreq ireq;
1871 struct net_device *dev;
1872 int err = -EINVAL;
1874 rtnl_lock();
1876 err = -EFAULT;
1877 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1878 goto err_exit;
1880 dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1882 err = -ENODEV;
1883 if (dev == NULL)
1884 goto err_exit;
1886 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1887 if (dev->type == ARPHRD_SIT) {
1888 struct ifreq ifr;
1889 mm_segment_t oldfs;
1890 struct ip_tunnel_parm p;
1892 err = -EADDRNOTAVAIL;
1893 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1894 goto err_exit;
1896 memset(&p, 0, sizeof(p));
1897 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1898 p.iph.saddr = 0;
1899 p.iph.version = 4;
1900 p.iph.ihl = 5;
1901 p.iph.protocol = IPPROTO_IPV6;
1902 p.iph.ttl = 64;
1903 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1905 oldfs = get_fs(); set_fs(KERNEL_DS);
1906 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1907 set_fs(oldfs);
1909 if (err == 0) {
1910 err = -ENOBUFS;
1911 if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1912 goto err_exit;
1913 err = dev_open(dev);
1916 #endif
1918 err_exit:
1919 rtnl_unlock();
1920 return err;
1924 * Manual configuration of address on an interface
1926 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1927 __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1929 struct inet6_ifaddr *ifp;
1930 struct inet6_dev *idev;
1931 struct net_device *dev;
1932 int scope;
1933 u32 flags = RTF_EXPIRES;
1935 ASSERT_RTNL();
1937 /* check the lifetime */
1938 if (!valid_lft || prefered_lft > valid_lft)
1939 return -EINVAL;
1941 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1942 return -ENODEV;
1944 if ((idev = addrconf_add_dev(dev)) == NULL)
1945 return -ENOBUFS;
1947 scope = ipv6_addr_scope(pfx);
1949 if (valid_lft == INFINITY_LIFE_TIME) {
1950 ifa_flags |= IFA_F_PERMANENT;
1951 flags = 0;
1952 } else if (valid_lft >= 0x7FFFFFFF/HZ)
1953 valid_lft = 0x7FFFFFFF/HZ;
1955 if (prefered_lft == 0)
1956 ifa_flags |= IFA_F_DEPRECATED;
1957 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1958 (prefered_lft != INFINITY_LIFE_TIME))
1959 prefered_lft = 0x7FFFFFFF/HZ;
1961 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1963 if (!IS_ERR(ifp)) {
1964 spin_lock_bh(&ifp->lock);
1965 ifp->valid_lft = valid_lft;
1966 ifp->prefered_lft = prefered_lft;
1967 ifp->tstamp = jiffies;
1968 spin_unlock_bh(&ifp->lock);
1970 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1971 jiffies_to_clock_t(valid_lft * HZ), flags);
1973 * Note that section 3.1 of RFC 4429 indicates
1974 * that the Optimistic flag should not be set for
1975 * manually configured addresses
1977 addrconf_dad_start(ifp, 0);
1978 in6_ifa_put(ifp);
1979 addrconf_verify(0);
1980 return 0;
1983 return PTR_ERR(ifp);
1986 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1988 struct inet6_ifaddr *ifp;
1989 struct inet6_dev *idev;
1990 struct net_device *dev;
1992 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1993 return -ENODEV;
1995 if ((idev = __in6_dev_get(dev)) == NULL)
1996 return -ENXIO;
1998 read_lock_bh(&idev->lock);
1999 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2000 if (ifp->prefix_len == plen &&
2001 ipv6_addr_equal(pfx, &ifp->addr)) {
2002 in6_ifa_hold(ifp);
2003 read_unlock_bh(&idev->lock);
2005 ipv6_del_addr(ifp);
2007 /* If the last address is deleted administratively,
2008 disable IPv6 on this interface.
2010 if (idev->addr_list == NULL)
2011 addrconf_ifdown(idev->dev, 1);
2012 return 0;
2015 read_unlock_bh(&idev->lock);
2016 return -EADDRNOTAVAIL;
2020 int addrconf_add_ifaddr(void __user *arg)
2022 struct in6_ifreq ireq;
2023 int err;
2025 if (!capable(CAP_NET_ADMIN))
2026 return -EPERM;
2028 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2029 return -EFAULT;
2031 rtnl_lock();
2032 err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2033 IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2034 rtnl_unlock();
2035 return err;
2038 int addrconf_del_ifaddr(void __user *arg)
2040 struct in6_ifreq ireq;
2041 int err;
2043 if (!capable(CAP_NET_ADMIN))
2044 return -EPERM;
2046 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2047 return -EFAULT;
2049 rtnl_lock();
2050 err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2051 rtnl_unlock();
2052 return err;
2055 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2056 static void sit_add_v4_addrs(struct inet6_dev *idev)
2058 struct inet6_ifaddr * ifp;
2059 struct in6_addr addr;
2060 struct net_device *dev;
2061 int scope;
2063 ASSERT_RTNL();
2065 memset(&addr, 0, sizeof(struct in6_addr));
2066 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2068 if (idev->dev->flags&IFF_POINTOPOINT) {
2069 addr.s6_addr32[0] = htonl(0xfe800000);
2070 scope = IFA_LINK;
2071 } else {
2072 scope = IPV6_ADDR_COMPATv4;
2075 if (addr.s6_addr32[3]) {
2076 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2077 if (!IS_ERR(ifp)) {
2078 spin_lock_bh(&ifp->lock);
2079 ifp->flags &= ~IFA_F_TENTATIVE;
2080 spin_unlock_bh(&ifp->lock);
2081 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2082 in6_ifa_put(ifp);
2084 return;
2087 for_each_netdev(&init_net, dev) {
2088 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2089 if (in_dev && (dev->flags & IFF_UP)) {
2090 struct in_ifaddr * ifa;
2092 int flag = scope;
2094 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2095 int plen;
2097 addr.s6_addr32[3] = ifa->ifa_local;
2099 if (ifa->ifa_scope == RT_SCOPE_LINK)
2100 continue;
2101 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2102 if (idev->dev->flags&IFF_POINTOPOINT)
2103 continue;
2104 flag |= IFA_HOST;
2106 if (idev->dev->flags&IFF_POINTOPOINT)
2107 plen = 64;
2108 else
2109 plen = 96;
2111 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2112 IFA_F_PERMANENT);
2113 if (!IS_ERR(ifp)) {
2114 spin_lock_bh(&ifp->lock);
2115 ifp->flags &= ~IFA_F_TENTATIVE;
2116 spin_unlock_bh(&ifp->lock);
2117 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2118 in6_ifa_put(ifp);
2124 #endif
2126 static void init_loopback(struct net_device *dev)
2128 struct inet6_dev *idev;
2129 struct inet6_ifaddr * ifp;
2131 /* ::1 */
2133 ASSERT_RTNL();
2135 if ((idev = ipv6_find_idev(dev)) == NULL) {
2136 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2137 return;
2140 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2141 if (!IS_ERR(ifp)) {
2142 spin_lock_bh(&ifp->lock);
2143 ifp->flags &= ~IFA_F_TENTATIVE;
2144 spin_unlock_bh(&ifp->lock);
2145 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2146 in6_ifa_put(ifp);
2150 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2152 struct inet6_ifaddr * ifp;
2153 u32 addr_flags = IFA_F_PERMANENT;
2155 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2156 if (idev->cnf.optimistic_dad &&
2157 !ipv6_devconf.forwarding)
2158 addr_flags |= IFA_F_OPTIMISTIC;
2159 #endif
2162 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2163 if (!IS_ERR(ifp)) {
2164 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2165 addrconf_dad_start(ifp, 0);
2166 in6_ifa_put(ifp);
2170 static void addrconf_dev_config(struct net_device *dev)
2172 struct in6_addr addr;
2173 struct inet6_dev * idev;
2175 ASSERT_RTNL();
2177 if ((dev->type != ARPHRD_ETHER) &&
2178 (dev->type != ARPHRD_FDDI) &&
2179 (dev->type != ARPHRD_IEEE802_TR) &&
2180 (dev->type != ARPHRD_ARCNET) &&
2181 (dev->type != ARPHRD_INFINIBAND)) {
2182 /* Alas, we support only Ethernet autoconfiguration. */
2183 return;
2186 idev = addrconf_add_dev(dev);
2187 if (idev == NULL)
2188 return;
2190 memset(&addr, 0, sizeof(struct in6_addr));
2191 addr.s6_addr32[0] = htonl(0xFE800000);
2193 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2194 addrconf_add_linklocal(idev, &addr);
2197 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2198 static void addrconf_sit_config(struct net_device *dev)
2200 struct inet6_dev *idev;
2202 ASSERT_RTNL();
2205 * Configure the tunnel with one of our IPv4
2206 * addresses... we should configure all of
2207 * our v4 addrs in the tunnel
2210 if ((idev = ipv6_find_idev(dev)) == NULL) {
2211 printk(KERN_DEBUG "init sit: add_dev failed\n");
2212 return;
2215 if (dev->priv_flags & IFF_ISATAP) {
2216 struct in6_addr addr;
2218 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2219 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2220 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2221 addrconf_add_linklocal(idev, &addr);
2222 return;
2225 sit_add_v4_addrs(idev);
2227 if (dev->flags&IFF_POINTOPOINT) {
2228 addrconf_add_mroute(dev);
2229 addrconf_add_lroute(dev);
2230 } else
2231 sit_route_add(dev);
2233 #endif
2235 static inline int
2236 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2238 struct in6_addr lladdr;
2240 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2241 addrconf_add_linklocal(idev, &lladdr);
2242 return 0;
2244 return -1;
2247 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2249 struct net_device *link_dev;
2251 /* first try to inherit the link-local address from the link device */
2252 if (idev->dev->iflink &&
2253 (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2254 if (!ipv6_inherit_linklocal(idev, link_dev))
2255 return;
2257 /* then try to inherit it from any device */
2258 for_each_netdev(&init_net, link_dev) {
2259 if (!ipv6_inherit_linklocal(idev, link_dev))
2260 return;
2262 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2266 * Autoconfigure tunnel with a link-local address so routing protocols,
2267 * DHCPv6, MLD etc. can be run over the virtual link
2270 static void addrconf_ip6_tnl_config(struct net_device *dev)
2272 struct inet6_dev *idev;
2274 ASSERT_RTNL();
2276 if ((idev = addrconf_add_dev(dev)) == NULL) {
2277 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2278 return;
2280 ip6_tnl_add_linklocal(idev);
2283 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2284 void * data)
2286 struct net_device *dev = (struct net_device *) data;
2287 struct inet6_dev *idev = __in6_dev_get(dev);
2288 int run_pending = 0;
2289 int err;
2291 if (dev->nd_net != &init_net)
2292 return NOTIFY_DONE;
2294 switch(event) {
2295 case NETDEV_REGISTER:
2296 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2297 idev = ipv6_add_dev(dev);
2298 if (!idev)
2299 return notifier_from_errno(-ENOMEM);
2301 break;
2302 case NETDEV_UP:
2303 case NETDEV_CHANGE:
2304 if (dev->flags & IFF_SLAVE)
2305 break;
2307 if (event == NETDEV_UP) {
2308 if (!addrconf_qdisc_ok(dev)) {
2309 /* device is not ready yet. */
2310 printk(KERN_INFO
2311 "ADDRCONF(NETDEV_UP): %s: "
2312 "link is not ready\n",
2313 dev->name);
2314 break;
2317 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2318 idev = ipv6_add_dev(dev);
2320 if (idev)
2321 idev->if_flags |= IF_READY;
2322 } else {
2323 if (!addrconf_qdisc_ok(dev)) {
2324 /* device is still not ready. */
2325 break;
2328 if (idev) {
2329 if (idev->if_flags & IF_READY) {
2330 /* device is already configured. */
2331 break;
2333 idev->if_flags |= IF_READY;
2336 printk(KERN_INFO
2337 "ADDRCONF(NETDEV_CHANGE): %s: "
2338 "link becomes ready\n",
2339 dev->name);
2341 run_pending = 1;
2344 switch(dev->type) {
2345 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2346 case ARPHRD_SIT:
2347 addrconf_sit_config(dev);
2348 break;
2349 #endif
2350 case ARPHRD_TUNNEL6:
2351 addrconf_ip6_tnl_config(dev);
2352 break;
2353 case ARPHRD_LOOPBACK:
2354 init_loopback(dev);
2355 break;
2357 default:
2358 addrconf_dev_config(dev);
2359 break;
2361 if (idev) {
2362 if (run_pending)
2363 addrconf_dad_run(idev);
2365 /* If the MTU changed during the interface down, when the
2366 interface up, the changed MTU must be reflected in the
2367 idev as well as routers.
2369 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2370 rt6_mtu_change(dev, dev->mtu);
2371 idev->cnf.mtu6 = dev->mtu;
2373 idev->tstamp = jiffies;
2374 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2375 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2376 stop IPv6 on this interface.
2378 if (dev->mtu < IPV6_MIN_MTU)
2379 addrconf_ifdown(dev, event != NETDEV_DOWN);
2381 break;
2383 case NETDEV_CHANGEMTU:
2384 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2385 rt6_mtu_change(dev, dev->mtu);
2386 idev->cnf.mtu6 = dev->mtu;
2387 break;
2390 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2391 idev = ipv6_add_dev(dev);
2392 if (idev)
2393 break;
2396 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2398 case NETDEV_DOWN:
2399 case NETDEV_UNREGISTER:
2401 * Remove all addresses from this interface.
2403 addrconf_ifdown(dev, event != NETDEV_DOWN);
2404 break;
2406 case NETDEV_CHANGENAME:
2407 if (idev) {
2408 snmp6_unregister_dev(idev);
2409 addrconf_sysctl_unregister(idev);
2410 addrconf_sysctl_register(idev);
2411 err = snmp6_register_dev(idev);
2412 if (err)
2413 return notifier_from_errno(err);
2415 break;
2418 return NOTIFY_OK;
2422 * addrconf module should be notified of a device going up
2424 static struct notifier_block ipv6_dev_notf = {
2425 .notifier_call = addrconf_notify,
2426 .priority = 0
2429 static int addrconf_ifdown(struct net_device *dev, int how)
2431 struct inet6_dev *idev;
2432 struct inet6_ifaddr *ifa, **bifa;
2433 int i;
2435 ASSERT_RTNL();
2437 if (dev == init_net.loopback_dev && how == 1)
2438 how = 0;
2440 rt6_ifdown(dev);
2441 neigh_ifdown(&nd_tbl, dev);
2443 idev = __in6_dev_get(dev);
2444 if (idev == NULL)
2445 return -ENODEV;
2447 /* Step 1: remove reference to ipv6 device from parent device.
2448 Do not dev_put!
2450 if (how == 1) {
2451 idev->dead = 1;
2453 /* protected by rtnl_lock */
2454 rcu_assign_pointer(dev->ip6_ptr, NULL);
2456 /* Step 1.5: remove snmp6 entry */
2457 snmp6_unregister_dev(idev);
2461 /* Step 2: clear hash table */
2462 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2463 bifa = &inet6_addr_lst[i];
2465 write_lock_bh(&addrconf_hash_lock);
2466 while ((ifa = *bifa) != NULL) {
2467 if (ifa->idev == idev) {
2468 *bifa = ifa->lst_next;
2469 ifa->lst_next = NULL;
2470 addrconf_del_timer(ifa);
2471 in6_ifa_put(ifa);
2472 continue;
2474 bifa = &ifa->lst_next;
2476 write_unlock_bh(&addrconf_hash_lock);
2479 write_lock_bh(&idev->lock);
2481 /* Step 3: clear flags for stateless addrconf */
2482 if (how != 1)
2483 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2485 /* Step 4: clear address list */
2486 #ifdef CONFIG_IPV6_PRIVACY
2487 if (how == 1 && del_timer(&idev->regen_timer))
2488 in6_dev_put(idev);
2490 /* clear tempaddr list */
2491 while ((ifa = idev->tempaddr_list) != NULL) {
2492 idev->tempaddr_list = ifa->tmp_next;
2493 ifa->tmp_next = NULL;
2494 ifa->dead = 1;
2495 write_unlock_bh(&idev->lock);
2496 spin_lock_bh(&ifa->lock);
2498 if (ifa->ifpub) {
2499 in6_ifa_put(ifa->ifpub);
2500 ifa->ifpub = NULL;
2502 spin_unlock_bh(&ifa->lock);
2503 in6_ifa_put(ifa);
2504 write_lock_bh(&idev->lock);
2506 #endif
2507 while ((ifa = idev->addr_list) != NULL) {
2508 idev->addr_list = ifa->if_next;
2509 ifa->if_next = NULL;
2510 ifa->dead = 1;
2511 addrconf_del_timer(ifa);
2512 write_unlock_bh(&idev->lock);
2514 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2515 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2516 in6_ifa_put(ifa);
2518 write_lock_bh(&idev->lock);
2520 write_unlock_bh(&idev->lock);
2522 /* Step 5: Discard multicast list */
2524 if (how == 1)
2525 ipv6_mc_destroy_dev(idev);
2526 else
2527 ipv6_mc_down(idev);
2529 idev->tstamp = jiffies;
2531 /* Shot the device (if unregistered) */
2533 if (how == 1) {
2534 addrconf_sysctl_unregister(idev);
2535 neigh_parms_release(&nd_tbl, idev->nd_parms);
2536 neigh_ifdown(&nd_tbl, dev);
2537 in6_dev_put(idev);
2539 return 0;
2542 static void addrconf_rs_timer(unsigned long data)
2544 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2546 if (ifp->idev->cnf.forwarding)
2547 goto out;
2549 if (ifp->idev->if_flags & IF_RA_RCVD) {
2551 * Announcement received after solicitation
2552 * was sent
2554 goto out;
2557 spin_lock(&ifp->lock);
2558 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2559 struct in6_addr all_routers;
2561 /* The wait after the last probe can be shorter */
2562 addrconf_mod_timer(ifp, AC_RS,
2563 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2564 ifp->idev->cnf.rtr_solicit_delay :
2565 ifp->idev->cnf.rtr_solicit_interval);
2566 spin_unlock(&ifp->lock);
2568 ipv6_addr_all_routers(&all_routers);
2570 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2571 } else {
2572 spin_unlock(&ifp->lock);
2574 * Note: we do not support deprecated "all on-link"
2575 * assumption any longer.
2577 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2578 ifp->idev->dev->name);
2581 out:
2582 in6_ifa_put(ifp);
2586 * Duplicate Address Detection
2588 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2590 unsigned long rand_num;
2591 struct inet6_dev *idev = ifp->idev;
2593 if (ifp->flags & IFA_F_OPTIMISTIC)
2594 rand_num = 0;
2595 else
2596 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2598 ifp->probes = idev->cnf.dad_transmits;
2599 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2602 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2604 struct inet6_dev *idev = ifp->idev;
2605 struct net_device *dev = idev->dev;
2607 addrconf_join_solict(dev, &ifp->addr);
2609 net_srandom(ifp->addr.s6_addr32[3]);
2611 read_lock_bh(&idev->lock);
2612 if (ifp->dead)
2613 goto out;
2614 spin_lock_bh(&ifp->lock);
2616 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2617 !(ifp->flags&IFA_F_TENTATIVE) ||
2618 ifp->flags & IFA_F_NODAD) {
2619 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2620 spin_unlock_bh(&ifp->lock);
2621 read_unlock_bh(&idev->lock);
2623 addrconf_dad_completed(ifp);
2624 return;
2627 if (!(idev->if_flags & IF_READY)) {
2628 spin_unlock_bh(&ifp->lock);
2629 read_unlock_bh(&idev->lock);
2631 * If the defice is not ready:
2632 * - keep it tentative if it is a permanent address.
2633 * - otherwise, kill it.
2635 in6_ifa_hold(ifp);
2636 addrconf_dad_stop(ifp);
2637 return;
2641 * Optimistic nodes can start receiving
2642 * Frames right away
2644 if(ifp->flags & IFA_F_OPTIMISTIC)
2645 ip6_ins_rt(ifp->rt);
2647 addrconf_dad_kick(ifp);
2648 spin_unlock_bh(&ifp->lock);
2649 out:
2650 read_unlock_bh(&idev->lock);
2653 static void addrconf_dad_timer(unsigned long data)
2655 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2656 struct inet6_dev *idev = ifp->idev;
2657 struct in6_addr unspec;
2658 struct in6_addr mcaddr;
2660 read_lock_bh(&idev->lock);
2661 if (idev->dead) {
2662 read_unlock_bh(&idev->lock);
2663 goto out;
2665 spin_lock_bh(&ifp->lock);
2666 if (ifp->probes == 0) {
2668 * DAD was successful
2671 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2672 spin_unlock_bh(&ifp->lock);
2673 read_unlock_bh(&idev->lock);
2675 addrconf_dad_completed(ifp);
2677 goto out;
2680 ifp->probes--;
2681 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2682 spin_unlock_bh(&ifp->lock);
2683 read_unlock_bh(&idev->lock);
2685 /* send a neighbour solicitation for our addr */
2686 memset(&unspec, 0, sizeof(unspec));
2687 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2688 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2689 out:
2690 in6_ifa_put(ifp);
2693 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2695 struct net_device * dev = ifp->idev->dev;
2698 * Configure the address for reception. Now it is valid.
2701 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2703 /* If added prefix is link local and forwarding is off,
2704 start sending router solicitations.
2707 if (ifp->idev->cnf.forwarding == 0 &&
2708 ifp->idev->cnf.rtr_solicits > 0 &&
2709 (dev->flags&IFF_LOOPBACK) == 0 &&
2710 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2711 struct in6_addr all_routers;
2713 ipv6_addr_all_routers(&all_routers);
2716 * If a host as already performed a random delay
2717 * [...] as part of DAD [...] there is no need
2718 * to delay again before sending the first RS
2720 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2722 spin_lock_bh(&ifp->lock);
2723 ifp->probes = 1;
2724 ifp->idev->if_flags |= IF_RS_SENT;
2725 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2726 spin_unlock_bh(&ifp->lock);
2730 static void addrconf_dad_run(struct inet6_dev *idev) {
2731 struct inet6_ifaddr *ifp;
2733 read_lock_bh(&idev->lock);
2734 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2735 spin_lock_bh(&ifp->lock);
2736 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2737 spin_unlock_bh(&ifp->lock);
2738 continue;
2740 spin_unlock_bh(&ifp->lock);
2741 addrconf_dad_kick(ifp);
2743 read_unlock_bh(&idev->lock);
2746 #ifdef CONFIG_PROC_FS
2747 struct if6_iter_state {
2748 struct seq_net_private p;
2749 int bucket;
2752 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2754 struct inet6_ifaddr *ifa = NULL;
2755 struct if6_iter_state *state = seq->private;
2756 struct net *net = state->p.net;
2758 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2759 ifa = inet6_addr_lst[state->bucket];
2761 while (ifa && ifa->idev->dev->nd_net != net)
2762 ifa = ifa->lst_next;
2763 if (ifa)
2764 break;
2766 return ifa;
2769 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2771 struct if6_iter_state *state = seq->private;
2772 struct net *net = state->p.net;
2774 ifa = ifa->lst_next;
2775 try_again:
2776 if (ifa) {
2777 if (ifa->idev->dev->nd_net != net) {
2778 ifa = ifa->lst_next;
2779 goto try_again;
2783 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2784 ifa = inet6_addr_lst[state->bucket];
2785 goto try_again;
2788 return ifa;
2791 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2793 struct inet6_ifaddr *ifa = if6_get_first(seq);
2795 if (ifa)
2796 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2797 --pos;
2798 return pos ? NULL : ifa;
2801 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2802 __acquires(addrconf_hash_lock)
2804 read_lock_bh(&addrconf_hash_lock);
2805 return if6_get_idx(seq, *pos);
2808 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2810 struct inet6_ifaddr *ifa;
2812 ifa = if6_get_next(seq, v);
2813 ++*pos;
2814 return ifa;
2817 static void if6_seq_stop(struct seq_file *seq, void *v)
2818 __releases(addrconf_hash_lock)
2820 read_unlock_bh(&addrconf_hash_lock);
2823 static int if6_seq_show(struct seq_file *seq, void *v)
2825 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2826 seq_printf(seq,
2827 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2828 NIP6(ifp->addr),
2829 ifp->idev->dev->ifindex,
2830 ifp->prefix_len,
2831 ifp->scope,
2832 ifp->flags,
2833 ifp->idev->dev->name);
2834 return 0;
2837 static const struct seq_operations if6_seq_ops = {
2838 .start = if6_seq_start,
2839 .next = if6_seq_next,
2840 .show = if6_seq_show,
2841 .stop = if6_seq_stop,
2844 static int if6_seq_open(struct inode *inode, struct file *file)
2846 return seq_open_net(inode, file, &if6_seq_ops,
2847 sizeof(struct if6_iter_state));
2850 static const struct file_operations if6_fops = {
2851 .owner = THIS_MODULE,
2852 .open = if6_seq_open,
2853 .read = seq_read,
2854 .llseek = seq_lseek,
2855 .release = seq_release_net,
2858 static int if6_proc_net_init(struct net *net)
2860 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2861 return -ENOMEM;
2862 return 0;
2865 static void if6_proc_net_exit(struct net *net)
2867 proc_net_remove(net, "if_inet6");
2870 static struct pernet_operations if6_proc_net_ops = {
2871 .init = if6_proc_net_init,
2872 .exit = if6_proc_net_exit,
2875 int __init if6_proc_init(void)
2877 return register_pernet_subsys(&if6_proc_net_ops);
2880 void if6_proc_exit(void)
2882 unregister_pernet_subsys(&if6_proc_net_ops);
2884 #endif /* CONFIG_PROC_FS */
2886 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2887 /* Check if address is a home address configured on any interface. */
2888 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2890 int ret = 0;
2891 struct inet6_ifaddr * ifp;
2892 u8 hash = ipv6_addr_hash(addr);
2893 read_lock_bh(&addrconf_hash_lock);
2894 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2895 if (ifp->idev->dev->nd_net != net)
2896 continue;
2897 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2898 (ifp->flags & IFA_F_HOMEADDRESS)) {
2899 ret = 1;
2900 break;
2903 read_unlock_bh(&addrconf_hash_lock);
2904 return ret;
2906 #endif
2909 * Periodic address status verification
2912 static void addrconf_verify(unsigned long foo)
2914 struct inet6_ifaddr *ifp;
2915 unsigned long now, next;
2916 int i;
2918 spin_lock_bh(&addrconf_verify_lock);
2919 now = jiffies;
2920 next = now + ADDR_CHECK_FREQUENCY;
2922 del_timer(&addr_chk_timer);
2924 for (i=0; i < IN6_ADDR_HSIZE; i++) {
2926 restart:
2927 read_lock(&addrconf_hash_lock);
2928 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2929 unsigned long age;
2930 #ifdef CONFIG_IPV6_PRIVACY
2931 unsigned long regen_advance;
2932 #endif
2934 if (ifp->flags & IFA_F_PERMANENT)
2935 continue;
2937 spin_lock(&ifp->lock);
2938 age = (now - ifp->tstamp) / HZ;
2940 #ifdef CONFIG_IPV6_PRIVACY
2941 regen_advance = ifp->idev->cnf.regen_max_retry *
2942 ifp->idev->cnf.dad_transmits *
2943 ifp->idev->nd_parms->retrans_time / HZ;
2944 #endif
2946 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2947 age >= ifp->valid_lft) {
2948 spin_unlock(&ifp->lock);
2949 in6_ifa_hold(ifp);
2950 read_unlock(&addrconf_hash_lock);
2951 ipv6_del_addr(ifp);
2952 goto restart;
2953 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2954 spin_unlock(&ifp->lock);
2955 continue;
2956 } else if (age >= ifp->prefered_lft) {
2957 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2958 int deprecate = 0;
2960 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2961 deprecate = 1;
2962 ifp->flags |= IFA_F_DEPRECATED;
2965 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2966 next = ifp->tstamp + ifp->valid_lft * HZ;
2968 spin_unlock(&ifp->lock);
2970 if (deprecate) {
2971 in6_ifa_hold(ifp);
2972 read_unlock(&addrconf_hash_lock);
2974 ipv6_ifa_notify(0, ifp);
2975 in6_ifa_put(ifp);
2976 goto restart;
2978 #ifdef CONFIG_IPV6_PRIVACY
2979 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2980 !(ifp->flags&IFA_F_TENTATIVE)) {
2981 if (age >= ifp->prefered_lft - regen_advance) {
2982 struct inet6_ifaddr *ifpub = ifp->ifpub;
2983 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2984 next = ifp->tstamp + ifp->prefered_lft * HZ;
2985 if (!ifp->regen_count && ifpub) {
2986 ifp->regen_count++;
2987 in6_ifa_hold(ifp);
2988 in6_ifa_hold(ifpub);
2989 spin_unlock(&ifp->lock);
2990 read_unlock(&addrconf_hash_lock);
2991 spin_lock(&ifpub->lock);
2992 ifpub->regen_count = 0;
2993 spin_unlock(&ifpub->lock);
2994 ipv6_create_tempaddr(ifpub, ifp);
2995 in6_ifa_put(ifpub);
2996 in6_ifa_put(ifp);
2997 goto restart;
2999 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3000 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3001 spin_unlock(&ifp->lock);
3002 #endif
3003 } else {
3004 /* ifp->prefered_lft <= ifp->valid_lft */
3005 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3006 next = ifp->tstamp + ifp->prefered_lft * HZ;
3007 spin_unlock(&ifp->lock);
3010 read_unlock(&addrconf_hash_lock);
3013 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3014 add_timer(&addr_chk_timer);
3015 spin_unlock_bh(&addrconf_verify_lock);
3018 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3020 struct in6_addr *pfx = NULL;
3022 if (addr)
3023 pfx = nla_data(addr);
3025 if (local) {
3026 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3027 pfx = NULL;
3028 else
3029 pfx = nla_data(local);
3032 return pfx;
3035 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3036 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3037 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3038 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3041 static int
3042 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3044 struct net *net = skb->sk->sk_net;
3045 struct ifaddrmsg *ifm;
3046 struct nlattr *tb[IFA_MAX+1];
3047 struct in6_addr *pfx;
3048 int err;
3050 if (net != &init_net)
3051 return -EINVAL;
3053 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3054 if (err < 0)
3055 return err;
3057 ifm = nlmsg_data(nlh);
3058 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3059 if (pfx == NULL)
3060 return -EINVAL;
3062 return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3065 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3066 u32 prefered_lft, u32 valid_lft)
3068 u32 flags = RTF_EXPIRES;
3070 if (!valid_lft || (prefered_lft > valid_lft))
3071 return -EINVAL;
3073 if (valid_lft == INFINITY_LIFE_TIME) {
3074 ifa_flags |= IFA_F_PERMANENT;
3075 flags = 0;
3076 } else if (valid_lft >= 0x7FFFFFFF/HZ)
3077 valid_lft = 0x7FFFFFFF/HZ;
3079 if (prefered_lft == 0)
3080 ifa_flags |= IFA_F_DEPRECATED;
3081 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3082 (prefered_lft != INFINITY_LIFE_TIME))
3083 prefered_lft = 0x7FFFFFFF/HZ;
3085 spin_lock_bh(&ifp->lock);
3086 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3087 ifp->tstamp = jiffies;
3088 ifp->valid_lft = valid_lft;
3089 ifp->prefered_lft = prefered_lft;
3091 spin_unlock_bh(&ifp->lock);
3092 if (!(ifp->flags&IFA_F_TENTATIVE))
3093 ipv6_ifa_notify(0, ifp);
3095 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3096 jiffies_to_clock_t(valid_lft * HZ), flags);
3097 addrconf_verify(0);
3099 return 0;
3102 static int
3103 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3105 struct net *net = skb->sk->sk_net;
3106 struct ifaddrmsg *ifm;
3107 struct nlattr *tb[IFA_MAX+1];
3108 struct in6_addr *pfx;
3109 struct inet6_ifaddr *ifa;
3110 struct net_device *dev;
3111 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3112 u8 ifa_flags;
3113 int err;
3115 if (net != &init_net)
3116 return -EINVAL;
3118 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3119 if (err < 0)
3120 return err;
3122 ifm = nlmsg_data(nlh);
3123 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3124 if (pfx == NULL)
3125 return -EINVAL;
3127 if (tb[IFA_CACHEINFO]) {
3128 struct ifa_cacheinfo *ci;
3130 ci = nla_data(tb[IFA_CACHEINFO]);
3131 valid_lft = ci->ifa_valid;
3132 preferred_lft = ci->ifa_prefered;
3133 } else {
3134 preferred_lft = INFINITY_LIFE_TIME;
3135 valid_lft = INFINITY_LIFE_TIME;
3138 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3139 if (dev == NULL)
3140 return -ENODEV;
3142 /* We ignore other flags so far. */
3143 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3145 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3146 if (ifa == NULL) {
3148 * It would be best to check for !NLM_F_CREATE here but
3149 * userspace alreay relies on not having to provide this.
3151 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3152 ifa_flags, preferred_lft, valid_lft);
3155 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3156 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3157 err = -EEXIST;
3158 else
3159 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3161 in6_ifa_put(ifa);
3163 return err;
3166 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3167 u8 scope, int ifindex)
3169 struct ifaddrmsg *ifm;
3171 ifm = nlmsg_data(nlh);
3172 ifm->ifa_family = AF_INET6;
3173 ifm->ifa_prefixlen = prefixlen;
3174 ifm->ifa_flags = flags;
3175 ifm->ifa_scope = scope;
3176 ifm->ifa_index = ifindex;
3179 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3180 unsigned long tstamp, u32 preferred, u32 valid)
3182 struct ifa_cacheinfo ci;
3184 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3185 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3186 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3187 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3188 ci.ifa_prefered = preferred;
3189 ci.ifa_valid = valid;
3191 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3194 static inline int rt_scope(int ifa_scope)
3196 if (ifa_scope & IFA_HOST)
3197 return RT_SCOPE_HOST;
3198 else if (ifa_scope & IFA_LINK)
3199 return RT_SCOPE_LINK;
3200 else if (ifa_scope & IFA_SITE)
3201 return RT_SCOPE_SITE;
3202 else
3203 return RT_SCOPE_UNIVERSE;
3206 static inline int inet6_ifaddr_msgsize(void)
3208 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3209 + nla_total_size(16) /* IFA_ADDRESS */
3210 + nla_total_size(sizeof(struct ifa_cacheinfo));
3213 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3214 u32 pid, u32 seq, int event, unsigned int flags)
3216 struct nlmsghdr *nlh;
3217 u32 preferred, valid;
3219 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3220 if (nlh == NULL)
3221 return -EMSGSIZE;
3223 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3224 ifa->idev->dev->ifindex);
3226 if (!(ifa->flags&IFA_F_PERMANENT)) {
3227 preferred = ifa->prefered_lft;
3228 valid = ifa->valid_lft;
3229 if (preferred != INFINITY_LIFE_TIME) {
3230 long tval = (jiffies - ifa->tstamp)/HZ;
3231 preferred -= tval;
3232 if (valid != INFINITY_LIFE_TIME)
3233 valid -= tval;
3235 } else {
3236 preferred = INFINITY_LIFE_TIME;
3237 valid = INFINITY_LIFE_TIME;
3240 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3241 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3242 nlmsg_cancel(skb, nlh);
3243 return -EMSGSIZE;
3246 return nlmsg_end(skb, nlh);
3249 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3250 u32 pid, u32 seq, int event, u16 flags)
3252 struct nlmsghdr *nlh;
3253 u8 scope = RT_SCOPE_UNIVERSE;
3254 int ifindex = ifmca->idev->dev->ifindex;
3256 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3257 scope = RT_SCOPE_SITE;
3259 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3260 if (nlh == NULL)
3261 return -EMSGSIZE;
3263 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3264 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3265 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3266 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3267 nlmsg_cancel(skb, nlh);
3268 return -EMSGSIZE;
3271 return nlmsg_end(skb, nlh);
3274 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3275 u32 pid, u32 seq, int event, unsigned int flags)
3277 struct nlmsghdr *nlh;
3278 u8 scope = RT_SCOPE_UNIVERSE;
3279 int ifindex = ifaca->aca_idev->dev->ifindex;
3281 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3282 scope = RT_SCOPE_SITE;
3284 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3285 if (nlh == NULL)
3286 return -EMSGSIZE;
3288 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3289 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3290 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3291 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3292 nlmsg_cancel(skb, nlh);
3293 return -EMSGSIZE;
3296 return nlmsg_end(skb, nlh);
3299 enum addr_type_t
3301 UNICAST_ADDR,
3302 MULTICAST_ADDR,
3303 ANYCAST_ADDR,
3306 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3307 enum addr_type_t type)
3309 int idx, ip_idx;
3310 int s_idx, s_ip_idx;
3311 int err = 1;
3312 struct net_device *dev;
3313 struct inet6_dev *idev = NULL;
3314 struct inet6_ifaddr *ifa;
3315 struct ifmcaddr6 *ifmca;
3316 struct ifacaddr6 *ifaca;
3318 s_idx = cb->args[0];
3319 s_ip_idx = ip_idx = cb->args[1];
3321 idx = 0;
3322 for_each_netdev(&init_net, dev) {
3323 if (idx < s_idx)
3324 goto cont;
3325 if (idx > s_idx)
3326 s_ip_idx = 0;
3327 ip_idx = 0;
3328 if ((idev = in6_dev_get(dev)) == NULL)
3329 goto cont;
3330 read_lock_bh(&idev->lock);
3331 switch (type) {
3332 case UNICAST_ADDR:
3333 /* unicast address incl. temp addr */
3334 for (ifa = idev->addr_list; ifa;
3335 ifa = ifa->if_next, ip_idx++) {
3336 if (ip_idx < s_ip_idx)
3337 continue;
3338 err = inet6_fill_ifaddr(skb, ifa,
3339 NETLINK_CB(cb->skb).pid,
3340 cb->nlh->nlmsg_seq,
3341 RTM_NEWADDR,
3342 NLM_F_MULTI);
3344 break;
3345 case MULTICAST_ADDR:
3346 /* multicast address */
3347 for (ifmca = idev->mc_list; ifmca;
3348 ifmca = ifmca->next, ip_idx++) {
3349 if (ip_idx < s_ip_idx)
3350 continue;
3351 err = inet6_fill_ifmcaddr(skb, ifmca,
3352 NETLINK_CB(cb->skb).pid,
3353 cb->nlh->nlmsg_seq,
3354 RTM_GETMULTICAST,
3355 NLM_F_MULTI);
3357 break;
3358 case ANYCAST_ADDR:
3359 /* anycast address */
3360 for (ifaca = idev->ac_list; ifaca;
3361 ifaca = ifaca->aca_next, ip_idx++) {
3362 if (ip_idx < s_ip_idx)
3363 continue;
3364 err = inet6_fill_ifacaddr(skb, ifaca,
3365 NETLINK_CB(cb->skb).pid,
3366 cb->nlh->nlmsg_seq,
3367 RTM_GETANYCAST,
3368 NLM_F_MULTI);
3370 break;
3371 default:
3372 break;
3374 read_unlock_bh(&idev->lock);
3375 in6_dev_put(idev);
3377 if (err <= 0)
3378 break;
3379 cont:
3380 idx++;
3382 cb->args[0] = idx;
3383 cb->args[1] = ip_idx;
3384 return skb->len;
3387 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3389 struct net *net = skb->sk->sk_net;
3390 enum addr_type_t type = UNICAST_ADDR;
3392 if (net != &init_net)
3393 return 0;
3395 return inet6_dump_addr(skb, cb, type);
3398 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3400 struct net *net = skb->sk->sk_net;
3401 enum addr_type_t type = MULTICAST_ADDR;
3403 if (net != &init_net)
3404 return 0;
3406 return inet6_dump_addr(skb, cb, type);
3410 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3412 struct net *net = skb->sk->sk_net;
3413 enum addr_type_t type = ANYCAST_ADDR;
3415 if (net != &init_net)
3416 return 0;
3418 return inet6_dump_addr(skb, cb, type);
3421 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3422 void *arg)
3424 struct net *net = in_skb->sk->sk_net;
3425 struct ifaddrmsg *ifm;
3426 struct nlattr *tb[IFA_MAX+1];
3427 struct in6_addr *addr = NULL;
3428 struct net_device *dev = NULL;
3429 struct inet6_ifaddr *ifa;
3430 struct sk_buff *skb;
3431 int err;
3433 if (net != &init_net)
3434 return -EINVAL;
3436 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3437 if (err < 0)
3438 goto errout;
3440 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3441 if (addr == NULL) {
3442 err = -EINVAL;
3443 goto errout;
3446 ifm = nlmsg_data(nlh);
3447 if (ifm->ifa_index)
3448 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3450 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3451 err = -EADDRNOTAVAIL;
3452 goto errout;
3455 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3456 err = -ENOBUFS;
3457 goto errout_ifa;
3460 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3461 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3462 if (err < 0) {
3463 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3464 WARN_ON(err == -EMSGSIZE);
3465 kfree_skb(skb);
3466 goto errout_ifa;
3468 err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
3469 errout_ifa:
3470 in6_ifa_put(ifa);
3471 errout:
3472 return err;
3475 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3477 struct sk_buff *skb;
3478 int err = -ENOBUFS;
3480 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3481 if (skb == NULL)
3482 goto errout;
3484 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3485 if (err < 0) {
3486 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3487 WARN_ON(err == -EMSGSIZE);
3488 kfree_skb(skb);
3489 goto errout;
3491 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3492 errout:
3493 if (err < 0)
3494 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3497 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3498 __s32 *array, int bytes)
3500 BUG_ON(bytes < (DEVCONF_MAX * 4));
3502 memset(array, 0, bytes);
3503 array[DEVCONF_FORWARDING] = cnf->forwarding;
3504 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3505 array[DEVCONF_MTU6] = cnf->mtu6;
3506 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3507 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3508 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3509 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3510 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3511 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3512 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3513 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3514 #ifdef CONFIG_IPV6_PRIVACY
3515 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3516 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3517 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3518 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3519 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3520 #endif
3521 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3522 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3523 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3524 #ifdef CONFIG_IPV6_ROUTER_PREF
3525 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3526 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3527 #ifdef CONFIG_IPV6_ROUTE_INFO
3528 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3529 #endif
3530 #endif
3531 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3532 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3533 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3534 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3535 #endif
3538 static inline size_t inet6_if_nlmsg_size(void)
3540 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3541 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3542 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3543 + nla_total_size(4) /* IFLA_MTU */
3544 + nla_total_size(4) /* IFLA_LINK */
3545 + nla_total_size( /* IFLA_PROTINFO */
3546 nla_total_size(4) /* IFLA_INET6_FLAGS */
3547 + nla_total_size(sizeof(struct ifla_cacheinfo))
3548 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3549 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3550 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3554 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3555 int bytes)
3557 int i;
3558 int pad = bytes - sizeof(u64) * items;
3559 BUG_ON(pad < 0);
3561 /* Use put_unaligned() because stats may not be aligned for u64. */
3562 put_unaligned(items, &stats[0]);
3563 for (i = 1; i < items; i++)
3564 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3566 memset(&stats[items], 0, pad);
3569 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3570 int bytes)
3572 switch(attrtype) {
3573 case IFLA_INET6_STATS:
3574 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3575 break;
3576 case IFLA_INET6_ICMP6STATS:
3577 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3578 break;
3582 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3583 u32 pid, u32 seq, int event, unsigned int flags)
3585 struct net_device *dev = idev->dev;
3586 struct nlattr *nla;
3587 struct ifinfomsg *hdr;
3588 struct nlmsghdr *nlh;
3589 void *protoinfo;
3590 struct ifla_cacheinfo ci;
3592 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3593 if (nlh == NULL)
3594 return -EMSGSIZE;
3596 hdr = nlmsg_data(nlh);
3597 hdr->ifi_family = AF_INET6;
3598 hdr->__ifi_pad = 0;
3599 hdr->ifi_type = dev->type;
3600 hdr->ifi_index = dev->ifindex;
3601 hdr->ifi_flags = dev_get_flags(dev);
3602 hdr->ifi_change = 0;
3604 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3606 if (dev->addr_len)
3607 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3609 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3610 if (dev->ifindex != dev->iflink)
3611 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3613 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3614 if (protoinfo == NULL)
3615 goto nla_put_failure;
3617 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3619 ci.max_reasm_len = IPV6_MAXPLEN;
3620 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3621 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3622 ci.reachable_time = idev->nd_parms->reachable_time;
3623 ci.retrans_time = idev->nd_parms->retrans_time;
3624 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3626 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3627 if (nla == NULL)
3628 goto nla_put_failure;
3629 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3631 /* XXX - MC not implemented */
3633 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3634 if (nla == NULL)
3635 goto nla_put_failure;
3636 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3638 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3639 if (nla == NULL)
3640 goto nla_put_failure;
3641 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3643 nla_nest_end(skb, protoinfo);
3644 return nlmsg_end(skb, nlh);
3646 nla_put_failure:
3647 nlmsg_cancel(skb, nlh);
3648 return -EMSGSIZE;
3651 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3653 struct net *net = skb->sk->sk_net;
3654 int idx, err;
3655 int s_idx = cb->args[0];
3656 struct net_device *dev;
3657 struct inet6_dev *idev;
3659 if (net != &init_net)
3660 return 0;
3662 read_lock(&dev_base_lock);
3663 idx = 0;
3664 for_each_netdev(&init_net, dev) {
3665 if (idx < s_idx)
3666 goto cont;
3667 if ((idev = in6_dev_get(dev)) == NULL)
3668 goto cont;
3669 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3670 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3671 in6_dev_put(idev);
3672 if (err <= 0)
3673 break;
3674 cont:
3675 idx++;
3677 read_unlock(&dev_base_lock);
3678 cb->args[0] = idx;
3680 return skb->len;
3683 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3685 struct sk_buff *skb;
3686 int err = -ENOBUFS;
3688 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3689 if (skb == NULL)
3690 goto errout;
3692 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3693 if (err < 0) {
3694 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3695 WARN_ON(err == -EMSGSIZE);
3696 kfree_skb(skb);
3697 goto errout;
3699 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3700 errout:
3701 if (err < 0)
3702 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3705 static inline size_t inet6_prefix_nlmsg_size(void)
3707 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3708 + nla_total_size(sizeof(struct in6_addr))
3709 + nla_total_size(sizeof(struct prefix_cacheinfo));
3712 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3713 struct prefix_info *pinfo, u32 pid, u32 seq,
3714 int event, unsigned int flags)
3716 struct prefixmsg *pmsg;
3717 struct nlmsghdr *nlh;
3718 struct prefix_cacheinfo ci;
3720 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3721 if (nlh == NULL)
3722 return -EMSGSIZE;
3724 pmsg = nlmsg_data(nlh);
3725 pmsg->prefix_family = AF_INET6;
3726 pmsg->prefix_pad1 = 0;
3727 pmsg->prefix_pad2 = 0;
3728 pmsg->prefix_ifindex = idev->dev->ifindex;
3729 pmsg->prefix_len = pinfo->prefix_len;
3730 pmsg->prefix_type = pinfo->type;
3731 pmsg->prefix_pad3 = 0;
3732 pmsg->prefix_flags = 0;
3733 if (pinfo->onlink)
3734 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3735 if (pinfo->autoconf)
3736 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3738 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3740 ci.preferred_time = ntohl(pinfo->prefered);
3741 ci.valid_time = ntohl(pinfo->valid);
3742 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3744 return nlmsg_end(skb, nlh);
3746 nla_put_failure:
3747 nlmsg_cancel(skb, nlh);
3748 return -EMSGSIZE;
3751 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3752 struct prefix_info *pinfo)
3754 struct sk_buff *skb;
3755 int err = -ENOBUFS;
3757 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3758 if (skb == NULL)
3759 goto errout;
3761 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3762 if (err < 0) {
3763 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3764 WARN_ON(err == -EMSGSIZE);
3765 kfree_skb(skb);
3766 goto errout;
3768 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3769 errout:
3770 if (err < 0)
3771 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_PREFIX, err);
3774 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3776 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3778 switch (event) {
3779 case RTM_NEWADDR:
3781 * If the address was optimistic
3782 * we inserted the route at the start of
3783 * our DAD process, so we don't need
3784 * to do it again
3786 if (!(ifp->rt->rt6i_node))
3787 ip6_ins_rt(ifp->rt);
3788 if (ifp->idev->cnf.forwarding)
3789 addrconf_join_anycast(ifp);
3790 break;
3791 case RTM_DELADDR:
3792 if (ifp->idev->cnf.forwarding)
3793 addrconf_leave_anycast(ifp);
3794 addrconf_leave_solict(ifp->idev, &ifp->addr);
3795 dst_hold(&ifp->rt->u.dst);
3796 if (ip6_del_rt(ifp->rt))
3797 dst_free(&ifp->rt->u.dst);
3798 break;
3802 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3804 rcu_read_lock_bh();
3805 if (likely(ifp->idev->dead == 0))
3806 __ipv6_ifa_notify(event, ifp);
3807 rcu_read_unlock_bh();
3810 #ifdef CONFIG_SYSCTL
3812 static
3813 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3814 void __user *buffer, size_t *lenp, loff_t *ppos)
3816 int *valp = ctl->data;
3817 int val = *valp;
3818 int ret;
3820 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3822 if (write)
3823 addrconf_fixup_forwarding(ctl, valp, val);
3824 return ret;
3827 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3828 int __user *name, int nlen,
3829 void __user *oldval,
3830 size_t __user *oldlenp,
3831 void __user *newval, size_t newlen)
3833 int *valp = table->data;
3834 int val = *valp;
3835 int new;
3837 if (!newval || !newlen)
3838 return 0;
3839 if (newlen != sizeof(int))
3840 return -EINVAL;
3841 if (get_user(new, (int __user *)newval))
3842 return -EFAULT;
3843 if (new == *valp)
3844 return 0;
3845 if (oldval && oldlenp) {
3846 size_t len;
3847 if (get_user(len, oldlenp))
3848 return -EFAULT;
3849 if (len) {
3850 if (len > table->maxlen)
3851 len = table->maxlen;
3852 if (copy_to_user(oldval, valp, len))
3853 return -EFAULT;
3854 if (put_user(len, oldlenp))
3855 return -EFAULT;
3859 *valp = new;
3860 addrconf_fixup_forwarding(table, valp, val);
3861 return 1;
3864 static struct addrconf_sysctl_table
3866 struct ctl_table_header *sysctl_header;
3867 ctl_table addrconf_vars[__NET_IPV6_MAX];
3868 char *dev_name;
3869 } addrconf_sysctl __read_mostly = {
3870 .sysctl_header = NULL,
3871 .addrconf_vars = {
3873 .ctl_name = NET_IPV6_FORWARDING,
3874 .procname = "forwarding",
3875 .data = &ipv6_devconf.forwarding,
3876 .maxlen = sizeof(int),
3877 .mode = 0644,
3878 .proc_handler = &addrconf_sysctl_forward,
3879 .strategy = &addrconf_sysctl_forward_strategy,
3882 .ctl_name = NET_IPV6_HOP_LIMIT,
3883 .procname = "hop_limit",
3884 .data = &ipv6_devconf.hop_limit,
3885 .maxlen = sizeof(int),
3886 .mode = 0644,
3887 .proc_handler = proc_dointvec,
3890 .ctl_name = NET_IPV6_MTU,
3891 .procname = "mtu",
3892 .data = &ipv6_devconf.mtu6,
3893 .maxlen = sizeof(int),
3894 .mode = 0644,
3895 .proc_handler = &proc_dointvec,
3898 .ctl_name = NET_IPV6_ACCEPT_RA,
3899 .procname = "accept_ra",
3900 .data = &ipv6_devconf.accept_ra,
3901 .maxlen = sizeof(int),
3902 .mode = 0644,
3903 .proc_handler = &proc_dointvec,
3906 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
3907 .procname = "accept_redirects",
3908 .data = &ipv6_devconf.accept_redirects,
3909 .maxlen = sizeof(int),
3910 .mode = 0644,
3911 .proc_handler = &proc_dointvec,
3914 .ctl_name = NET_IPV6_AUTOCONF,
3915 .procname = "autoconf",
3916 .data = &ipv6_devconf.autoconf,
3917 .maxlen = sizeof(int),
3918 .mode = 0644,
3919 .proc_handler = &proc_dointvec,
3922 .ctl_name = NET_IPV6_DAD_TRANSMITS,
3923 .procname = "dad_transmits",
3924 .data = &ipv6_devconf.dad_transmits,
3925 .maxlen = sizeof(int),
3926 .mode = 0644,
3927 .proc_handler = &proc_dointvec,
3930 .ctl_name = NET_IPV6_RTR_SOLICITS,
3931 .procname = "router_solicitations",
3932 .data = &ipv6_devconf.rtr_solicits,
3933 .maxlen = sizeof(int),
3934 .mode = 0644,
3935 .proc_handler = &proc_dointvec,
3938 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
3939 .procname = "router_solicitation_interval",
3940 .data = &ipv6_devconf.rtr_solicit_interval,
3941 .maxlen = sizeof(int),
3942 .mode = 0644,
3943 .proc_handler = &proc_dointvec_jiffies,
3944 .strategy = &sysctl_jiffies,
3947 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
3948 .procname = "router_solicitation_delay",
3949 .data = &ipv6_devconf.rtr_solicit_delay,
3950 .maxlen = sizeof(int),
3951 .mode = 0644,
3952 .proc_handler = &proc_dointvec_jiffies,
3953 .strategy = &sysctl_jiffies,
3956 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
3957 .procname = "force_mld_version",
3958 .data = &ipv6_devconf.force_mld_version,
3959 .maxlen = sizeof(int),
3960 .mode = 0644,
3961 .proc_handler = &proc_dointvec,
3963 #ifdef CONFIG_IPV6_PRIVACY
3965 .ctl_name = NET_IPV6_USE_TEMPADDR,
3966 .procname = "use_tempaddr",
3967 .data = &ipv6_devconf.use_tempaddr,
3968 .maxlen = sizeof(int),
3969 .mode = 0644,
3970 .proc_handler = &proc_dointvec,
3973 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
3974 .procname = "temp_valid_lft",
3975 .data = &ipv6_devconf.temp_valid_lft,
3976 .maxlen = sizeof(int),
3977 .mode = 0644,
3978 .proc_handler = &proc_dointvec,
3981 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
3982 .procname = "temp_prefered_lft",
3983 .data = &ipv6_devconf.temp_prefered_lft,
3984 .maxlen = sizeof(int),
3985 .mode = 0644,
3986 .proc_handler = &proc_dointvec,
3989 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
3990 .procname = "regen_max_retry",
3991 .data = &ipv6_devconf.regen_max_retry,
3992 .maxlen = sizeof(int),
3993 .mode = 0644,
3994 .proc_handler = &proc_dointvec,
3997 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
3998 .procname = "max_desync_factor",
3999 .data = &ipv6_devconf.max_desync_factor,
4000 .maxlen = sizeof(int),
4001 .mode = 0644,
4002 .proc_handler = &proc_dointvec,
4004 #endif
4006 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4007 .procname = "max_addresses",
4008 .data = &ipv6_devconf.max_addresses,
4009 .maxlen = sizeof(int),
4010 .mode = 0644,
4011 .proc_handler = &proc_dointvec,
4014 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4015 .procname = "accept_ra_defrtr",
4016 .data = &ipv6_devconf.accept_ra_defrtr,
4017 .maxlen = sizeof(int),
4018 .mode = 0644,
4019 .proc_handler = &proc_dointvec,
4022 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4023 .procname = "accept_ra_pinfo",
4024 .data = &ipv6_devconf.accept_ra_pinfo,
4025 .maxlen = sizeof(int),
4026 .mode = 0644,
4027 .proc_handler = &proc_dointvec,
4029 #ifdef CONFIG_IPV6_ROUTER_PREF
4031 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4032 .procname = "accept_ra_rtr_pref",
4033 .data = &ipv6_devconf.accept_ra_rtr_pref,
4034 .maxlen = sizeof(int),
4035 .mode = 0644,
4036 .proc_handler = &proc_dointvec,
4039 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4040 .procname = "router_probe_interval",
4041 .data = &ipv6_devconf.rtr_probe_interval,
4042 .maxlen = sizeof(int),
4043 .mode = 0644,
4044 .proc_handler = &proc_dointvec_jiffies,
4045 .strategy = &sysctl_jiffies,
4047 #ifdef CONFIG_IPV6_ROUTE_INFO
4049 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4050 .procname = "accept_ra_rt_info_max_plen",
4051 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4052 .maxlen = sizeof(int),
4053 .mode = 0644,
4054 .proc_handler = &proc_dointvec,
4056 #endif
4057 #endif
4059 .ctl_name = NET_IPV6_PROXY_NDP,
4060 .procname = "proxy_ndp",
4061 .data = &ipv6_devconf.proxy_ndp,
4062 .maxlen = sizeof(int),
4063 .mode = 0644,
4064 .proc_handler = &proc_dointvec,
4067 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4068 .procname = "accept_source_route",
4069 .data = &ipv6_devconf.accept_source_route,
4070 .maxlen = sizeof(int),
4071 .mode = 0644,
4072 .proc_handler = &proc_dointvec,
4074 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4076 .ctl_name = CTL_UNNUMBERED,
4077 .procname = "optimistic_dad",
4078 .data = &ipv6_devconf.optimistic_dad,
4079 .maxlen = sizeof(int),
4080 .mode = 0644,
4081 .proc_handler = &proc_dointvec,
4084 #endif
4086 .ctl_name = 0, /* sentinel */
4091 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4092 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4094 int i;
4095 struct addrconf_sysctl_table *t;
4097 #define ADDRCONF_CTL_PATH_DEV 3
4099 struct ctl_path addrconf_ctl_path[] = {
4100 { .procname = "net", .ctl_name = CTL_NET, },
4101 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4102 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4103 { /* to be set */ },
4104 { },
4108 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4109 if (t == NULL)
4110 goto out;
4112 for (i=0; t->addrconf_vars[i].data; i++) {
4113 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4114 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4115 t->addrconf_vars[i].extra2 = net;
4119 * Make a copy of dev_name, because '.procname' is regarded as const
4120 * by sysctl and we wouldn't want anyone to change it under our feet
4121 * (see SIOCSIFNAME).
4123 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4124 if (!t->dev_name)
4125 goto free;
4127 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4128 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4130 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4131 t->addrconf_vars);
4132 if (t->sysctl_header == NULL)
4133 goto free_procname;
4135 p->sysctl = t;
4136 return 0;
4138 free_procname:
4139 kfree(t->dev_name);
4140 free:
4141 kfree(t);
4142 out:
4143 return -ENOBUFS;
4146 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4148 struct addrconf_sysctl_table *t;
4150 if (p->sysctl == NULL)
4151 return;
4153 t = p->sysctl;
4154 p->sysctl = NULL;
4155 unregister_sysctl_table(t->sysctl_header);
4156 kfree(t->dev_name);
4157 kfree(t);
4160 static void addrconf_sysctl_register(struct inet6_dev *idev)
4162 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4163 NET_IPV6_NEIGH, "ipv6",
4164 &ndisc_ifinfo_sysctl_change,
4165 NULL);
4166 __addrconf_sysctl_register(idev->dev->nd_net, idev->dev->name,
4167 idev->dev->ifindex, idev, &idev->cnf);
4170 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4172 __addrconf_sysctl_unregister(&idev->cnf);
4173 neigh_sysctl_unregister(idev->nd_parms);
4177 #endif
4179 static int addrconf_init_net(struct net *net)
4181 int err;
4182 struct ipv6_devconf *all, *dflt;
4184 err = -ENOMEM;
4185 all = &ipv6_devconf;
4186 dflt = &ipv6_devconf_dflt;
4188 if (net != &init_net) {
4189 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4190 if (all == NULL)
4191 goto err_alloc_all;
4193 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4194 if (dflt == NULL)
4195 goto err_alloc_dflt;
4198 net->ipv6.devconf_all = all;
4199 net->ipv6.devconf_dflt = dflt;
4201 #ifdef CONFIG_SYSCTL
4202 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4203 NULL, all);
4204 if (err < 0)
4205 goto err_reg_all;
4207 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4208 NULL, dflt);
4209 if (err < 0)
4210 goto err_reg_dflt;
4211 #endif
4212 return 0;
4214 #ifdef CONFIG_SYSCTL
4215 err_reg_dflt:
4216 __addrconf_sysctl_unregister(all);
4217 err_reg_all:
4218 kfree(dflt);
4219 #endif
4220 err_alloc_dflt:
4221 kfree(all);
4222 err_alloc_all:
4223 return err;
4226 static void addrconf_exit_net(struct net *net)
4228 #ifdef CONFIG_SYSCTL
4229 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4230 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4231 #endif
4232 if (net != &init_net) {
4233 kfree(net->ipv6.devconf_dflt);
4234 kfree(net->ipv6.devconf_all);
4238 static struct pernet_operations addrconf_ops = {
4239 .init = addrconf_init_net,
4240 .exit = addrconf_exit_net,
4244 * Device notifier
4247 int register_inet6addr_notifier(struct notifier_block *nb)
4249 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4252 EXPORT_SYMBOL(register_inet6addr_notifier);
4254 int unregister_inet6addr_notifier(struct notifier_block *nb)
4256 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4259 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4262 * Init / cleanup code
4265 int __init addrconf_init(void)
4267 int err;
4269 if ((err = ipv6_addr_label_init()) < 0) {
4270 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4271 err);
4272 return err;
4275 register_pernet_subsys(&addrconf_ops);
4277 /* The addrconf netdev notifier requires that loopback_dev
4278 * has it's ipv6 private information allocated and setup
4279 * before it can bring up and give link-local addresses
4280 * to other devices which are up.
4282 * Unfortunately, loopback_dev is not necessarily the first
4283 * entry in the global dev_base list of net devices. In fact,
4284 * it is likely to be the very last entry on that list.
4285 * So this causes the notifier registry below to try and
4286 * give link-local addresses to all devices besides loopback_dev
4287 * first, then loopback_dev, which cases all the non-loopback_dev
4288 * devices to fail to get a link-local address.
4290 * So, as a temporary fix, allocate the ipv6 structure for
4291 * loopback_dev first by hand.
4292 * Longer term, all of the dependencies ipv6 has upon the loopback
4293 * device and it being up should be removed.
4295 rtnl_lock();
4296 if (!ipv6_add_dev(init_net.loopback_dev))
4297 err = -ENOMEM;
4298 rtnl_unlock();
4299 if (err)
4300 goto errlo;
4302 ip6_null_entry.u.dst.dev = init_net.loopback_dev;
4303 ip6_null_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4304 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4305 ip6_prohibit_entry.u.dst.dev = init_net.loopback_dev;
4306 ip6_prohibit_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4307 ip6_blk_hole_entry.u.dst.dev = init_net.loopback_dev;
4308 ip6_blk_hole_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4309 #endif
4311 register_netdevice_notifier(&ipv6_dev_notf);
4313 addrconf_verify(0);
4315 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4316 if (err < 0)
4317 goto errout;
4319 /* Only the first call to __rtnl_register can fail */
4320 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4321 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4322 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4323 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4324 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4326 ipv6_addr_label_rtnl_register();
4328 return 0;
4329 errout:
4330 unregister_netdevice_notifier(&ipv6_dev_notf);
4331 errlo:
4332 unregister_pernet_subsys(&addrconf_ops);
4334 return err;
4337 void addrconf_cleanup(void)
4339 struct net_device *dev;
4340 struct inet6_ifaddr *ifa;
4341 int i;
4343 unregister_netdevice_notifier(&ipv6_dev_notf);
4345 unregister_pernet_subsys(&addrconf_ops);
4347 rtnl_lock();
4350 * clean dev list.
4353 for_each_netdev(&init_net, dev) {
4354 if (__in6_dev_get(dev) == NULL)
4355 continue;
4356 addrconf_ifdown(dev, 1);
4358 addrconf_ifdown(init_net.loopback_dev, 2);
4361 * Check hash table.
4364 write_lock_bh(&addrconf_hash_lock);
4365 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4366 for (ifa=inet6_addr_lst[i]; ifa; ) {
4367 struct inet6_ifaddr *bifa;
4369 bifa = ifa;
4370 ifa = ifa->lst_next;
4371 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4372 /* Do not free it; something is wrong.
4373 Now we can investigate it with debugger.
4377 write_unlock_bh(&addrconf_hash_lock);
4379 del_timer(&addr_chk_timer);
4381 rtnl_unlock();