Committer: Michael Beasley <mike@snafu.setup>
[mikesnafu-overlay.git] / net / ipv6 / addrconf.c
blobe08955baedffc491a6fafaae378e71adb50d9e23
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 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
375 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
376 printk(KERN_INFO
377 "%s: Disabled Multicast RS\n",
378 dev->name);
379 ndev->cnf.rtr_solicits = 0;
381 #endif
383 #ifdef CONFIG_IPV6_PRIVACY
384 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
385 if ((dev->flags&IFF_LOOPBACK) ||
386 dev->type == ARPHRD_TUNNEL ||
387 dev->type == ARPHRD_TUNNEL6 ||
388 dev->type == ARPHRD_SIT ||
389 dev->type == ARPHRD_NONE) {
390 printk(KERN_INFO
391 "%s: Disabled Privacy Extensions\n",
392 dev->name);
393 ndev->cnf.use_tempaddr = -1;
394 } else {
395 in6_dev_hold(ndev);
396 ipv6_regen_rndid((unsigned long) ndev);
398 #endif
400 if (netif_running(dev) && addrconf_qdisc_ok(dev))
401 ndev->if_flags |= IF_READY;
403 ipv6_mc_init_dev(ndev);
404 ndev->tstamp = jiffies;
405 addrconf_sysctl_register(ndev);
406 /* protected by rtnl_lock */
407 rcu_assign_pointer(dev->ip6_ptr, ndev);
409 /* Join all-node multicast group */
410 ipv6_addr_all_nodes(&maddr);
411 ipv6_dev_mc_inc(dev, &maddr);
413 return ndev;
416 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
418 struct inet6_dev *idev;
420 ASSERT_RTNL();
422 if ((idev = __in6_dev_get(dev)) == NULL) {
423 if ((idev = ipv6_add_dev(dev)) == NULL)
424 return NULL;
427 if (dev->flags&IFF_UP)
428 ipv6_mc_up(idev);
429 return idev;
432 #ifdef CONFIG_SYSCTL
433 static void dev_forward_change(struct inet6_dev *idev)
435 struct net_device *dev;
436 struct inet6_ifaddr *ifa;
437 struct in6_addr addr;
439 if (!idev)
440 return;
441 dev = idev->dev;
442 if (dev && (dev->flags & IFF_MULTICAST)) {
443 ipv6_addr_all_routers(&addr);
445 if (idev->cnf.forwarding)
446 ipv6_dev_mc_inc(dev, &addr);
447 else
448 ipv6_dev_mc_dec(dev, &addr);
450 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
451 if (ifa->flags&IFA_F_TENTATIVE)
452 continue;
453 if (idev->cnf.forwarding)
454 addrconf_join_anycast(ifa);
455 else
456 addrconf_leave_anycast(ifa);
461 static void addrconf_forward_change(struct net *net, __s32 newf)
463 struct net_device *dev;
464 struct inet6_dev *idev;
466 read_lock(&dev_base_lock);
467 for_each_netdev(net, dev) {
468 rcu_read_lock();
469 idev = __in6_dev_get(dev);
470 if (idev) {
471 int changed = (!idev->cnf.forwarding) ^ (!newf);
472 idev->cnf.forwarding = newf;
473 if (changed)
474 dev_forward_change(idev);
476 rcu_read_unlock();
478 read_unlock(&dev_base_lock);
481 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
483 struct net *net;
485 net = (struct net *)table->extra2;
486 if (p == &net->ipv6.devconf_dflt->forwarding)
487 return;
489 if (p == &net->ipv6.devconf_all->forwarding) {
490 __s32 newf = net->ipv6.devconf_all->forwarding;
491 net->ipv6.devconf_dflt->forwarding = newf;
492 addrconf_forward_change(net, newf);
493 } else if ((!*p) ^ (!old))
494 dev_forward_change((struct inet6_dev *)table->extra1);
496 if (*p)
497 rt6_purge_dflt_routers();
499 #endif
501 /* Nobody refers to this ifaddr, destroy it */
503 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
505 BUG_TRAP(ifp->if_next==NULL);
506 BUG_TRAP(ifp->lst_next==NULL);
507 #ifdef NET_REFCNT_DEBUG
508 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
509 #endif
511 in6_dev_put(ifp->idev);
513 if (del_timer(&ifp->timer))
514 printk("Timer is still running, when freeing ifa=%p\n", ifp);
516 if (!ifp->dead) {
517 printk("Freeing alive inet6 address %p\n", ifp);
518 return;
520 dst_release(&ifp->rt->u.dst);
522 kfree(ifp);
525 static void
526 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
528 struct inet6_ifaddr *ifa, **ifap;
529 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
532 * Each device address list is sorted in order of scope -
533 * global before linklocal.
535 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
536 ifap = &ifa->if_next) {
537 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
538 break;
541 ifp->if_next = *ifap;
542 *ifap = ifp;
545 /* On success it returns ifp with increased reference count */
547 static struct inet6_ifaddr *
548 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
549 int scope, u32 flags)
551 struct inet6_ifaddr *ifa = NULL;
552 struct rt6_info *rt;
553 int hash;
554 int err = 0;
556 rcu_read_lock_bh();
557 if (idev->dead) {
558 err = -ENODEV; /*XXX*/
559 goto out2;
562 write_lock(&addrconf_hash_lock);
564 /* Ignore adding duplicate addresses on an interface */
565 if (ipv6_chk_same_addr(&init_net, addr, idev->dev)) {
566 ADBG(("ipv6_add_addr: already assigned\n"));
567 err = -EEXIST;
568 goto out;
571 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
573 if (ifa == NULL) {
574 ADBG(("ipv6_add_addr: malloc failed\n"));
575 err = -ENOBUFS;
576 goto out;
579 rt = addrconf_dst_alloc(idev, addr, 0);
580 if (IS_ERR(rt)) {
581 err = PTR_ERR(rt);
582 goto out;
585 ipv6_addr_copy(&ifa->addr, addr);
587 spin_lock_init(&ifa->lock);
588 init_timer(&ifa->timer);
589 ifa->timer.data = (unsigned long) ifa;
590 ifa->scope = scope;
591 ifa->prefix_len = pfxlen;
592 ifa->flags = flags | IFA_F_TENTATIVE;
593 ifa->cstamp = ifa->tstamp = jiffies;
595 ifa->rt = rt;
598 * part one of RFC 4429, section 3.3
599 * We should not configure an address as
600 * optimistic if we do not yet know the link
601 * layer address of our nexhop router
604 if (rt->rt6i_nexthop == NULL)
605 ifa->flags &= ~IFA_F_OPTIMISTIC;
607 ifa->idev = idev;
608 in6_dev_hold(idev);
609 /* For caller */
610 in6_ifa_hold(ifa);
612 /* Add to big hash table */
613 hash = ipv6_addr_hash(addr);
615 ifa->lst_next = inet6_addr_lst[hash];
616 inet6_addr_lst[hash] = ifa;
617 in6_ifa_hold(ifa);
618 write_unlock(&addrconf_hash_lock);
620 write_lock(&idev->lock);
621 /* Add to inet6_dev unicast addr list. */
622 ipv6_link_dev_addr(idev, ifa);
624 #ifdef CONFIG_IPV6_PRIVACY
625 if (ifa->flags&IFA_F_TEMPORARY) {
626 ifa->tmp_next = idev->tempaddr_list;
627 idev->tempaddr_list = ifa;
628 in6_ifa_hold(ifa);
630 #endif
632 in6_ifa_hold(ifa);
633 write_unlock(&idev->lock);
634 out2:
635 rcu_read_unlock_bh();
637 if (likely(err == 0))
638 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
639 else {
640 kfree(ifa);
641 ifa = ERR_PTR(err);
644 return ifa;
645 out:
646 write_unlock(&addrconf_hash_lock);
647 goto out2;
650 /* This function wants to get referenced ifp and releases it before return */
652 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
654 struct inet6_ifaddr *ifa, **ifap;
655 struct inet6_dev *idev = ifp->idev;
656 int hash;
657 int deleted = 0, onlink = 0;
658 unsigned long expires = jiffies;
660 hash = ipv6_addr_hash(&ifp->addr);
662 ifp->dead = 1;
664 write_lock_bh(&addrconf_hash_lock);
665 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
666 ifap = &ifa->lst_next) {
667 if (ifa == ifp) {
668 *ifap = ifa->lst_next;
669 __in6_ifa_put(ifp);
670 ifa->lst_next = NULL;
671 break;
674 write_unlock_bh(&addrconf_hash_lock);
676 write_lock_bh(&idev->lock);
677 #ifdef CONFIG_IPV6_PRIVACY
678 if (ifp->flags&IFA_F_TEMPORARY) {
679 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
680 ifap = &ifa->tmp_next) {
681 if (ifa == ifp) {
682 *ifap = ifa->tmp_next;
683 if (ifp->ifpub) {
684 in6_ifa_put(ifp->ifpub);
685 ifp->ifpub = NULL;
687 __in6_ifa_put(ifp);
688 ifa->tmp_next = NULL;
689 break;
693 #endif
695 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
696 if (ifa == ifp) {
697 *ifap = ifa->if_next;
698 __in6_ifa_put(ifp);
699 ifa->if_next = NULL;
700 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
701 break;
702 deleted = 1;
703 continue;
704 } else if (ifp->flags & IFA_F_PERMANENT) {
705 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
706 ifp->prefix_len)) {
707 if (ifa->flags & IFA_F_PERMANENT) {
708 onlink = 1;
709 if (deleted)
710 break;
711 } else {
712 unsigned long lifetime;
714 if (!onlink)
715 onlink = -1;
717 spin_lock(&ifa->lock);
718 lifetime = min_t(unsigned long,
719 ifa->valid_lft, 0x7fffffffUL/HZ);
720 if (time_before(expires,
721 ifa->tstamp + lifetime * HZ))
722 expires = ifa->tstamp + lifetime * HZ;
723 spin_unlock(&ifa->lock);
727 ifap = &ifa->if_next;
729 write_unlock_bh(&idev->lock);
731 ipv6_ifa_notify(RTM_DELADDR, ifp);
733 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
735 addrconf_del_timer(ifp);
738 * Purge or update corresponding prefix
740 * 1) we don't purge prefix here if address was not permanent.
741 * prefix is managed by its own lifetime.
742 * 2) if there're no addresses, delete prefix.
743 * 3) if there're still other permanent address(es),
744 * corresponding prefix is still permanent.
745 * 4) otherwise, update prefix lifetime to the
746 * longest valid lifetime among the corresponding
747 * addresses on the device.
748 * Note: subsequent RA will update lifetime.
750 * --yoshfuji
752 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
753 struct in6_addr prefix;
754 struct rt6_info *rt;
756 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
757 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
759 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
760 if (onlink == 0) {
761 ip6_del_rt(rt);
762 rt = NULL;
763 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
764 rt->rt6i_expires = expires;
765 rt->rt6i_flags |= RTF_EXPIRES;
768 dst_release(&rt->u.dst);
771 in6_ifa_put(ifp);
774 #ifdef CONFIG_IPV6_PRIVACY
775 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
777 struct inet6_dev *idev = ifp->idev;
778 struct in6_addr addr, *tmpaddr;
779 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
780 unsigned long regen_advance;
781 int tmp_plen;
782 int ret = 0;
783 int max_addresses;
784 u32 addr_flags;
786 write_lock(&idev->lock);
787 if (ift) {
788 spin_lock_bh(&ift->lock);
789 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
790 spin_unlock_bh(&ift->lock);
791 tmpaddr = &addr;
792 } else {
793 tmpaddr = NULL;
795 retry:
796 in6_dev_hold(idev);
797 if (idev->cnf.use_tempaddr <= 0) {
798 write_unlock(&idev->lock);
799 printk(KERN_INFO
800 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
801 in6_dev_put(idev);
802 ret = -1;
803 goto out;
805 spin_lock_bh(&ifp->lock);
806 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
807 idev->cnf.use_tempaddr = -1; /*XXX*/
808 spin_unlock_bh(&ifp->lock);
809 write_unlock(&idev->lock);
810 printk(KERN_WARNING
811 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
812 in6_dev_put(idev);
813 ret = -1;
814 goto out;
816 in6_ifa_hold(ifp);
817 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
818 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
819 spin_unlock_bh(&ifp->lock);
820 write_unlock(&idev->lock);
821 printk(KERN_WARNING
822 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
823 in6_ifa_put(ifp);
824 in6_dev_put(idev);
825 ret = -1;
826 goto out;
828 memcpy(&addr.s6_addr[8], idev->rndid, 8);
829 tmp_valid_lft = min_t(__u32,
830 ifp->valid_lft,
831 idev->cnf.temp_valid_lft);
832 tmp_prefered_lft = min_t(__u32,
833 ifp->prefered_lft,
834 idev->cnf.temp_prefered_lft - desync_factor / HZ);
835 tmp_plen = ifp->prefix_len;
836 max_addresses = idev->cnf.max_addresses;
837 tmp_cstamp = ifp->cstamp;
838 tmp_tstamp = ifp->tstamp;
839 spin_unlock_bh(&ifp->lock);
841 regen_advance = idev->cnf.regen_max_retry *
842 idev->cnf.dad_transmits *
843 idev->nd_parms->retrans_time / HZ;
844 write_unlock(&idev->lock);
846 /* A temporary address is created only if this calculated Preferred
847 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
848 * an implementation must not create a temporary address with a zero
849 * Preferred Lifetime.
851 if (tmp_prefered_lft <= regen_advance) {
852 in6_ifa_put(ifp);
853 in6_dev_put(idev);
854 ret = -1;
855 goto out;
858 addr_flags = IFA_F_TEMPORARY;
859 /* set in addrconf_prefix_rcv() */
860 if (ifp->flags & IFA_F_OPTIMISTIC)
861 addr_flags |= IFA_F_OPTIMISTIC;
863 ift = !max_addresses ||
864 ipv6_count_addresses(idev) < max_addresses ?
865 ipv6_add_addr(idev, &addr, tmp_plen,
866 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
867 addr_flags) : NULL;
868 if (!ift || IS_ERR(ift)) {
869 in6_ifa_put(ifp);
870 in6_dev_put(idev);
871 printk(KERN_INFO
872 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
873 tmpaddr = &addr;
874 write_lock(&idev->lock);
875 goto retry;
878 spin_lock_bh(&ift->lock);
879 ift->ifpub = ifp;
880 ift->valid_lft = tmp_valid_lft;
881 ift->prefered_lft = tmp_prefered_lft;
882 ift->cstamp = tmp_cstamp;
883 ift->tstamp = tmp_tstamp;
884 spin_unlock_bh(&ift->lock);
886 addrconf_dad_start(ift, 0);
887 in6_ifa_put(ift);
888 in6_dev_put(idev);
889 out:
890 return ret;
892 #endif
895 * Choose an appropriate source address (RFC3484)
897 struct ipv6_saddr_score {
898 int addr_type;
899 unsigned int attrs;
900 int matchlen;
901 int scope;
902 unsigned int rule;
905 #define IPV6_SADDR_SCORE_LOCAL 0x0001
906 #define IPV6_SADDR_SCORE_PREFERRED 0x0004
907 #define IPV6_SADDR_SCORE_HOA 0x0008
908 #define IPV6_SADDR_SCORE_OIF 0x0010
909 #define IPV6_SADDR_SCORE_LABEL 0x0020
910 #define IPV6_SADDR_SCORE_PRIVACY 0x0040
912 static inline int ipv6_saddr_preferred(int type)
914 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
915 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
916 return 1;
917 return 0;
920 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
921 struct in6_addr *daddr, struct in6_addr *saddr)
923 struct ipv6_saddr_score hiscore;
924 struct inet6_ifaddr *ifa_result = NULL;
925 int daddr_type = __ipv6_addr_type(daddr);
926 int daddr_scope = __ipv6_addr_src_scope(daddr_type);
927 int daddr_ifindex = daddr_dev ? daddr_dev->ifindex : 0;
928 u32 daddr_label = ipv6_addr_label(daddr, daddr_type, daddr_ifindex);
929 struct net_device *dev;
931 memset(&hiscore, 0, sizeof(hiscore));
933 read_lock(&dev_base_lock);
934 rcu_read_lock();
936 for_each_netdev(&init_net, dev) {
937 struct inet6_dev *idev;
938 struct inet6_ifaddr *ifa;
940 /* Rule 0: Candidate Source Address (section 4)
941 * - multicast and link-local destination address,
942 * the set of candidate source address MUST only
943 * include addresses assigned to interfaces
944 * belonging to the same link as the outgoing
945 * interface.
946 * (- For site-local destination addresses, the
947 * set of candidate source addresses MUST only
948 * include addresses assigned to interfaces
949 * belonging to the same site as the outgoing
950 * interface.)
952 if ((daddr_type & IPV6_ADDR_MULTICAST ||
953 daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
954 daddr_dev && dev != daddr_dev)
955 continue;
957 idev = __in6_dev_get(dev);
958 if (!idev)
959 continue;
961 read_lock_bh(&idev->lock);
962 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
963 struct ipv6_saddr_score score;
965 score.addr_type = __ipv6_addr_type(&ifa->addr);
967 /* Rule 0:
968 * - Tentative Address (RFC2462 section 5.4)
969 * - A tentative address is not considered
970 * "assigned to an interface" in the traditional
971 * sense, unless it is also flagged as optimistic.
972 * - Candidate Source Address (section 4)
973 * - In any case, anycast addresses, multicast
974 * addresses, and the unspecified address MUST
975 * NOT be included in a candidate set.
977 if ((ifa->flags & IFA_F_TENTATIVE) &&
978 (!(ifa->flags & IFA_F_OPTIMISTIC)))
979 continue;
980 if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
981 score.addr_type & IPV6_ADDR_MULTICAST)) {
982 LIMIT_NETDEBUG(KERN_DEBUG
983 "ADDRCONF: unspecified / multicast address "
984 "assigned as unicast address on %s",
985 dev->name);
986 continue;
989 score.attrs = 0;
990 score.matchlen = 0;
991 score.scope = 0;
992 score.rule = 0;
994 if (ifa_result == NULL) {
995 /* record it if the first available entry */
996 goto record_it;
999 /* Rule 1: Prefer same address */
1000 if (hiscore.rule < 1) {
1001 if (ipv6_addr_equal(&ifa_result->addr, daddr))
1002 hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
1003 hiscore.rule++;
1005 if (ipv6_addr_equal(&ifa->addr, daddr)) {
1006 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
1007 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
1008 score.rule = 1;
1009 goto record_it;
1011 } else {
1012 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
1013 continue;
1016 /* Rule 2: Prefer appropriate scope */
1017 if (hiscore.rule < 2) {
1018 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1019 hiscore.rule++;
1021 score.scope = __ipv6_addr_src_scope(score.addr_type);
1022 if (hiscore.scope < score.scope) {
1023 if (hiscore.scope < daddr_scope) {
1024 score.rule = 2;
1025 goto record_it;
1026 } else
1027 continue;
1028 } else if (score.scope < hiscore.scope) {
1029 if (score.scope < daddr_scope)
1030 break; /* addresses sorted by scope */
1031 else {
1032 score.rule = 2;
1033 goto record_it;
1037 /* Rule 3: Avoid deprecated and optimistic addresses */
1038 if (hiscore.rule < 3) {
1039 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1040 (((ifa_result->flags &
1041 (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0)))
1042 hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1043 hiscore.rule++;
1045 if (ipv6_saddr_preferred(score.addr_type) ||
1046 (((ifa->flags &
1047 (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0))) {
1048 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1049 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1050 score.rule = 3;
1051 goto record_it;
1053 } else {
1054 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1055 continue;
1058 /* Rule 4: Prefer home address */
1059 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
1060 if (hiscore.rule < 4) {
1061 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1062 hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1063 hiscore.rule++;
1065 if (ifa->flags & IFA_F_HOMEADDRESS) {
1066 score.attrs |= IPV6_SADDR_SCORE_HOA;
1067 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1068 score.rule = 4;
1069 goto record_it;
1071 } else {
1072 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1073 continue;
1075 #else
1076 if (hiscore.rule < 4)
1077 hiscore.rule++;
1078 #endif
1080 /* Rule 5: Prefer outgoing interface */
1081 if (hiscore.rule < 5) {
1082 if (daddr_dev == NULL ||
1083 daddr_dev == ifa_result->idev->dev)
1084 hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1085 hiscore.rule++;
1087 if (daddr_dev == NULL ||
1088 daddr_dev == ifa->idev->dev) {
1089 score.attrs |= IPV6_SADDR_SCORE_OIF;
1090 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1091 score.rule = 5;
1092 goto record_it;
1094 } else {
1095 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1096 continue;
1099 /* Rule 6: Prefer matching label */
1100 if (hiscore.rule < 6) {
1101 if (ipv6_addr_label(&ifa_result->addr,
1102 hiscore.addr_type,
1103 ifa_result->idev->dev->ifindex) == daddr_label)
1104 hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1105 hiscore.rule++;
1107 if (ipv6_addr_label(&ifa->addr,
1108 score.addr_type,
1109 ifa->idev->dev->ifindex) == daddr_label) {
1110 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1111 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1112 score.rule = 6;
1113 goto record_it;
1115 } else {
1116 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1117 continue;
1120 #ifdef CONFIG_IPV6_PRIVACY
1121 /* Rule 7: Prefer public address
1122 * Note: prefer temprary address if use_tempaddr >= 2
1124 if (hiscore.rule < 7) {
1125 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1126 (ifa_result->idev->cnf.use_tempaddr >= 2))
1127 hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1128 hiscore.rule++;
1130 if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1131 (ifa->idev->cnf.use_tempaddr >= 2)) {
1132 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1133 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1134 score.rule = 7;
1135 goto record_it;
1137 } else {
1138 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1139 continue;
1141 #else
1142 if (hiscore.rule < 7)
1143 hiscore.rule++;
1144 #endif
1145 /* Rule 8: Use longest matching prefix */
1146 if (hiscore.rule < 8) {
1147 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1148 hiscore.rule++;
1150 score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1151 if (score.matchlen > hiscore.matchlen) {
1152 score.rule = 8;
1153 goto record_it;
1155 #if 0
1156 else if (score.matchlen < hiscore.matchlen)
1157 continue;
1158 #endif
1160 /* Final Rule: choose first available one */
1161 continue;
1162 record_it:
1163 if (ifa_result)
1164 in6_ifa_put(ifa_result);
1165 in6_ifa_hold(ifa);
1166 ifa_result = ifa;
1167 hiscore = score;
1169 read_unlock_bh(&idev->lock);
1171 rcu_read_unlock();
1172 read_unlock(&dev_base_lock);
1174 if (!ifa_result)
1175 return -EADDRNOTAVAIL;
1177 ipv6_addr_copy(saddr, &ifa_result->addr);
1178 in6_ifa_put(ifa_result);
1179 return 0;
1183 int ipv6_get_saddr(struct dst_entry *dst,
1184 struct in6_addr *daddr, struct in6_addr *saddr)
1186 return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1189 EXPORT_SYMBOL(ipv6_get_saddr);
1191 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1192 unsigned char banned_flags)
1194 struct inet6_dev *idev;
1195 int err = -EADDRNOTAVAIL;
1197 rcu_read_lock();
1198 if ((idev = __in6_dev_get(dev)) != NULL) {
1199 struct inet6_ifaddr *ifp;
1201 read_lock_bh(&idev->lock);
1202 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1203 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1204 ipv6_addr_copy(addr, &ifp->addr);
1205 err = 0;
1206 break;
1209 read_unlock_bh(&idev->lock);
1211 rcu_read_unlock();
1212 return err;
1215 static int ipv6_count_addresses(struct inet6_dev *idev)
1217 int cnt = 0;
1218 struct inet6_ifaddr *ifp;
1220 read_lock_bh(&idev->lock);
1221 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1222 cnt++;
1223 read_unlock_bh(&idev->lock);
1224 return cnt;
1227 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1228 struct net_device *dev, int strict)
1230 struct inet6_ifaddr * ifp;
1231 u8 hash = ipv6_addr_hash(addr);
1233 read_lock_bh(&addrconf_hash_lock);
1234 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1235 if (ifp->idev->dev->nd_net != net)
1236 continue;
1237 if (ipv6_addr_equal(&ifp->addr, addr) &&
1238 !(ifp->flags&IFA_F_TENTATIVE)) {
1239 if (dev == NULL || ifp->idev->dev == dev ||
1240 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1241 break;
1244 read_unlock_bh(&addrconf_hash_lock);
1245 return ifp != NULL;
1247 EXPORT_SYMBOL(ipv6_chk_addr);
1249 static
1250 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1251 struct net_device *dev)
1253 struct inet6_ifaddr * ifp;
1254 u8 hash = ipv6_addr_hash(addr);
1256 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1257 if (ifp->idev->dev->nd_net != net)
1258 continue;
1259 if (ipv6_addr_equal(&ifp->addr, addr)) {
1260 if (dev == NULL || ifp->idev->dev == dev)
1261 break;
1264 return ifp != NULL;
1267 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1268 struct net_device *dev, int strict)
1270 struct inet6_ifaddr * ifp;
1271 u8 hash = ipv6_addr_hash(addr);
1273 read_lock_bh(&addrconf_hash_lock);
1274 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1275 if (ifp->idev->dev->nd_net != net)
1276 continue;
1277 if (ipv6_addr_equal(&ifp->addr, addr)) {
1278 if (dev == NULL || ifp->idev->dev == dev ||
1279 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1280 in6_ifa_hold(ifp);
1281 break;
1285 read_unlock_bh(&addrconf_hash_lock);
1287 return ifp;
1290 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1292 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1293 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1294 __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1295 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1296 int sk_ipv6only = ipv6_only_sock(sk);
1297 int sk2_ipv6only = inet_v6_ipv6only(sk2);
1298 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1299 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1301 if (!sk2_rcv_saddr && !sk_ipv6only)
1302 return 1;
1304 if (addr_type2 == IPV6_ADDR_ANY &&
1305 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1306 return 1;
1308 if (addr_type == IPV6_ADDR_ANY &&
1309 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1310 return 1;
1312 if (sk2_rcv_saddr6 &&
1313 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1314 return 1;
1316 if (addr_type == IPV6_ADDR_MAPPED &&
1317 !sk2_ipv6only &&
1318 (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1319 return 1;
1321 return 0;
1324 /* Gets referenced address, destroys ifaddr */
1326 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1328 if (ifp->flags&IFA_F_PERMANENT) {
1329 spin_lock_bh(&ifp->lock);
1330 addrconf_del_timer(ifp);
1331 ifp->flags |= IFA_F_TENTATIVE;
1332 spin_unlock_bh(&ifp->lock);
1333 in6_ifa_put(ifp);
1334 #ifdef CONFIG_IPV6_PRIVACY
1335 } else if (ifp->flags&IFA_F_TEMPORARY) {
1336 struct inet6_ifaddr *ifpub;
1337 spin_lock_bh(&ifp->lock);
1338 ifpub = ifp->ifpub;
1339 if (ifpub) {
1340 in6_ifa_hold(ifpub);
1341 spin_unlock_bh(&ifp->lock);
1342 ipv6_create_tempaddr(ifpub, ifp);
1343 in6_ifa_put(ifpub);
1344 } else {
1345 spin_unlock_bh(&ifp->lock);
1347 ipv6_del_addr(ifp);
1348 #endif
1349 } else
1350 ipv6_del_addr(ifp);
1353 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1355 if (net_ratelimit())
1356 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1357 addrconf_dad_stop(ifp);
1360 /* Join to solicited addr multicast group. */
1362 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1364 struct in6_addr maddr;
1366 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1367 return;
1369 addrconf_addr_solict_mult(addr, &maddr);
1370 ipv6_dev_mc_inc(dev, &maddr);
1373 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1375 struct in6_addr maddr;
1377 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1378 return;
1380 addrconf_addr_solict_mult(addr, &maddr);
1381 __ipv6_dev_mc_dec(idev, &maddr);
1384 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1386 struct in6_addr addr;
1387 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1388 if (ipv6_addr_any(&addr))
1389 return;
1390 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1393 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1395 struct in6_addr addr;
1396 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1397 if (ipv6_addr_any(&addr))
1398 return;
1399 __ipv6_dev_ac_dec(ifp->idev, &addr);
1402 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1404 if (dev->addr_len != ETH_ALEN)
1405 return -1;
1406 memcpy(eui, dev->dev_addr, 3);
1407 memcpy(eui + 5, dev->dev_addr + 3, 3);
1410 * The zSeries OSA network cards can be shared among various
1411 * OS instances, but the OSA cards have only one MAC address.
1412 * This leads to duplicate address conflicts in conjunction
1413 * with IPv6 if more than one instance uses the same card.
1415 * The driver for these cards can deliver a unique 16-bit
1416 * identifier for each instance sharing the same card. It is
1417 * placed instead of 0xFFFE in the interface identifier. The
1418 * "u" bit of the interface identifier is not inverted in this
1419 * case. Hence the resulting interface identifier has local
1420 * scope according to RFC2373.
1422 if (dev->dev_id) {
1423 eui[3] = (dev->dev_id >> 8) & 0xFF;
1424 eui[4] = dev->dev_id & 0xFF;
1425 } else {
1426 eui[3] = 0xFF;
1427 eui[4] = 0xFE;
1428 eui[0] ^= 2;
1430 return 0;
1433 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1435 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1436 if (dev->addr_len != ARCNET_ALEN)
1437 return -1;
1438 memset(eui, 0, 7);
1439 eui[7] = *(u8*)dev->dev_addr;
1440 return 0;
1443 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1445 if (dev->addr_len != INFINIBAND_ALEN)
1446 return -1;
1447 memcpy(eui, dev->dev_addr + 12, 8);
1448 eui[0] |= 2;
1449 return 0;
1452 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1454 switch (dev->type) {
1455 case ARPHRD_ETHER:
1456 case ARPHRD_FDDI:
1457 case ARPHRD_IEEE802_TR:
1458 return addrconf_ifid_eui48(eui, dev);
1459 case ARPHRD_ARCNET:
1460 return addrconf_ifid_arcnet(eui, dev);
1461 case ARPHRD_INFINIBAND:
1462 return addrconf_ifid_infiniband(eui, dev);
1463 case ARPHRD_SIT:
1464 if (dev->priv_flags & IFF_ISATAP)
1465 return ipv6_isatap_eui64(eui, *(__be32 *)dev->dev_addr);
1467 return -1;
1470 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1472 int err = -1;
1473 struct inet6_ifaddr *ifp;
1475 read_lock_bh(&idev->lock);
1476 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1477 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1478 memcpy(eui, ifp->addr.s6_addr+8, 8);
1479 err = 0;
1480 break;
1483 read_unlock_bh(&idev->lock);
1484 return err;
1487 #ifdef CONFIG_IPV6_PRIVACY
1488 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1489 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1491 regen:
1492 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1493 idev->rndid[0] &= ~0x02;
1496 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1497 * check if generated address is not inappropriate
1499 * - Reserved subnet anycast (RFC 2526)
1500 * 11111101 11....11 1xxxxxxx
1501 * - ISATAP (RFC4214) 6.1
1502 * 00-00-5E-FE-xx-xx-xx-xx
1503 * - value 0
1504 * - XXX: already assigned to an address on the device
1506 if (idev->rndid[0] == 0xfd &&
1507 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1508 (idev->rndid[7]&0x80))
1509 goto regen;
1510 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1511 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1512 goto regen;
1513 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1514 goto regen;
1517 return 0;
1520 static void ipv6_regen_rndid(unsigned long data)
1522 struct inet6_dev *idev = (struct inet6_dev *) data;
1523 unsigned long expires;
1525 rcu_read_lock_bh();
1526 write_lock_bh(&idev->lock);
1528 if (idev->dead)
1529 goto out;
1531 if (__ipv6_regen_rndid(idev) < 0)
1532 goto out;
1534 expires = jiffies +
1535 idev->cnf.temp_prefered_lft * HZ -
1536 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1537 if (time_before(expires, jiffies)) {
1538 printk(KERN_WARNING
1539 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1540 idev->dev->name);
1541 goto out;
1544 if (!mod_timer(&idev->regen_timer, expires))
1545 in6_dev_hold(idev);
1547 out:
1548 write_unlock_bh(&idev->lock);
1549 rcu_read_unlock_bh();
1550 in6_dev_put(idev);
1553 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1554 int ret = 0;
1556 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1557 ret = __ipv6_regen_rndid(idev);
1558 return ret;
1560 #endif
1563 * Add prefix route.
1566 static void
1567 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1568 unsigned long expires, u32 flags)
1570 struct fib6_config cfg = {
1571 .fc_table = RT6_TABLE_PREFIX,
1572 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1573 .fc_ifindex = dev->ifindex,
1574 .fc_expires = expires,
1575 .fc_dst_len = plen,
1576 .fc_flags = RTF_UP | flags,
1577 .fc_nlinfo.nl_net = &init_net,
1580 ipv6_addr_copy(&cfg.fc_dst, pfx);
1582 /* Prevent useless cloning on PtP SIT.
1583 This thing is done here expecting that the whole
1584 class of non-broadcast devices need not cloning.
1586 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1587 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1588 cfg.fc_flags |= RTF_NONEXTHOP;
1589 #endif
1591 ip6_route_add(&cfg);
1594 /* Create "default" multicast route to the interface */
1596 static void addrconf_add_mroute(struct net_device *dev)
1598 struct fib6_config cfg = {
1599 .fc_table = RT6_TABLE_LOCAL,
1600 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1601 .fc_ifindex = dev->ifindex,
1602 .fc_dst_len = 8,
1603 .fc_flags = RTF_UP,
1604 .fc_nlinfo.nl_net = &init_net,
1607 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1609 ip6_route_add(&cfg);
1612 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1613 static void sit_route_add(struct net_device *dev)
1615 struct fib6_config cfg = {
1616 .fc_table = RT6_TABLE_MAIN,
1617 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1618 .fc_ifindex = dev->ifindex,
1619 .fc_dst_len = 96,
1620 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1621 .fc_nlinfo.nl_net = &init_net,
1624 /* prefix length - 96 bits "::d.d.d.d" */
1625 ip6_route_add(&cfg);
1627 #endif
1629 static void addrconf_add_lroute(struct net_device *dev)
1631 struct in6_addr addr;
1633 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1634 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1637 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1639 struct inet6_dev *idev;
1641 ASSERT_RTNL();
1643 if ((idev = ipv6_find_idev(dev)) == NULL)
1644 return NULL;
1646 /* Add default multicast route */
1647 addrconf_add_mroute(dev);
1649 /* Add link local route */
1650 addrconf_add_lroute(dev);
1651 return idev;
1654 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1656 struct prefix_info *pinfo;
1657 __u32 valid_lft;
1658 __u32 prefered_lft;
1659 int addr_type;
1660 unsigned long rt_expires;
1661 struct inet6_dev *in6_dev;
1663 pinfo = (struct prefix_info *) opt;
1665 if (len < sizeof(struct prefix_info)) {
1666 ADBG(("addrconf: prefix option too short\n"));
1667 return;
1671 * Validation checks ([ADDRCONF], page 19)
1674 addr_type = ipv6_addr_type(&pinfo->prefix);
1676 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1677 return;
1679 valid_lft = ntohl(pinfo->valid);
1680 prefered_lft = ntohl(pinfo->prefered);
1682 if (prefered_lft > valid_lft) {
1683 if (net_ratelimit())
1684 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1685 return;
1688 in6_dev = in6_dev_get(dev);
1690 if (in6_dev == NULL) {
1691 if (net_ratelimit())
1692 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1693 return;
1697 * Two things going on here:
1698 * 1) Add routes for on-link prefixes
1699 * 2) Configure prefixes with the auto flag set
1702 /* Avoid arithmetic overflow. Really, we could
1703 save rt_expires in seconds, likely valid_lft,
1704 but it would require division in fib gc, that it
1705 not good.
1707 if (valid_lft >= 0x7FFFFFFF/HZ)
1708 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1709 else
1710 rt_expires = valid_lft * HZ;
1713 * We convert this (in jiffies) to clock_t later.
1714 * Avoid arithmetic overflow there as well.
1715 * Overflow can happen only if HZ < USER_HZ.
1717 if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1718 rt_expires = 0x7FFFFFFF / USER_HZ;
1720 if (pinfo->onlink) {
1721 struct rt6_info *rt;
1722 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1724 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1725 if (rt->rt6i_flags&RTF_EXPIRES) {
1726 if (valid_lft == 0) {
1727 ip6_del_rt(rt);
1728 rt = NULL;
1729 } else {
1730 rt->rt6i_expires = jiffies + rt_expires;
1733 } else if (valid_lft) {
1734 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1735 dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1737 if (rt)
1738 dst_release(&rt->u.dst);
1741 /* Try to figure out our local address for this prefix */
1743 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1744 struct inet6_ifaddr * ifp;
1745 struct in6_addr addr;
1746 int create = 0, update_lft = 0;
1748 if (pinfo->prefix_len == 64) {
1749 memcpy(&addr, &pinfo->prefix, 8);
1750 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1751 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1752 in6_dev_put(in6_dev);
1753 return;
1755 goto ok;
1757 if (net_ratelimit())
1758 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1759 pinfo->prefix_len);
1760 in6_dev_put(in6_dev);
1761 return;
1765 ifp = ipv6_get_ifaddr(&init_net, &addr, dev, 1);
1767 if (ifp == NULL && valid_lft) {
1768 int max_addresses = in6_dev->cnf.max_addresses;
1769 u32 addr_flags = 0;
1771 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1772 if (in6_dev->cnf.optimistic_dad &&
1773 !ipv6_devconf.forwarding)
1774 addr_flags = IFA_F_OPTIMISTIC;
1775 #endif
1777 /* Do not allow to create too much of autoconfigured
1778 * addresses; this would be too easy way to crash kernel.
1780 if (!max_addresses ||
1781 ipv6_count_addresses(in6_dev) < max_addresses)
1782 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1783 addr_type&IPV6_ADDR_SCOPE_MASK,
1784 addr_flags);
1786 if (!ifp || IS_ERR(ifp)) {
1787 in6_dev_put(in6_dev);
1788 return;
1791 update_lft = create = 1;
1792 ifp->cstamp = jiffies;
1793 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1796 if (ifp) {
1797 int flags;
1798 unsigned long now;
1799 #ifdef CONFIG_IPV6_PRIVACY
1800 struct inet6_ifaddr *ift;
1801 #endif
1802 u32 stored_lft;
1804 /* update lifetime (RFC2462 5.5.3 e) */
1805 spin_lock(&ifp->lock);
1806 now = jiffies;
1807 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1808 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1809 else
1810 stored_lft = 0;
1811 if (!update_lft && stored_lft) {
1812 if (valid_lft > MIN_VALID_LIFETIME ||
1813 valid_lft > stored_lft)
1814 update_lft = 1;
1815 else if (stored_lft <= MIN_VALID_LIFETIME) {
1816 /* valid_lft <= stored_lft is always true */
1817 /* XXX: IPsec */
1818 update_lft = 0;
1819 } else {
1820 valid_lft = MIN_VALID_LIFETIME;
1821 if (valid_lft < prefered_lft)
1822 prefered_lft = valid_lft;
1823 update_lft = 1;
1827 if (update_lft) {
1828 ifp->valid_lft = valid_lft;
1829 ifp->prefered_lft = prefered_lft;
1830 ifp->tstamp = now;
1831 flags = ifp->flags;
1832 ifp->flags &= ~IFA_F_DEPRECATED;
1833 spin_unlock(&ifp->lock);
1835 if (!(flags&IFA_F_TENTATIVE))
1836 ipv6_ifa_notify(0, ifp);
1837 } else
1838 spin_unlock(&ifp->lock);
1840 #ifdef CONFIG_IPV6_PRIVACY
1841 read_lock_bh(&in6_dev->lock);
1842 /* update all temporary addresses in the list */
1843 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1845 * When adjusting the lifetimes of an existing
1846 * temporary address, only lower the lifetimes.
1847 * Implementations must not increase the
1848 * lifetimes of an existing temporary address
1849 * when processing a Prefix Information Option.
1851 if (ifp != ift->ifpub)
1852 continue;
1854 spin_lock(&ift->lock);
1855 flags = ift->flags;
1856 if (ift->valid_lft > valid_lft &&
1857 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1858 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1859 if (ift->prefered_lft > prefered_lft &&
1860 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1861 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1862 spin_unlock(&ift->lock);
1863 if (!(flags&IFA_F_TENTATIVE))
1864 ipv6_ifa_notify(0, ift);
1867 if (create && in6_dev->cnf.use_tempaddr > 0) {
1869 * When a new public address is created as described in [ADDRCONF],
1870 * also create a new temporary address.
1872 read_unlock_bh(&in6_dev->lock);
1873 ipv6_create_tempaddr(ifp, NULL);
1874 } else {
1875 read_unlock_bh(&in6_dev->lock);
1877 #endif
1878 in6_ifa_put(ifp);
1879 addrconf_verify(0);
1882 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1883 in6_dev_put(in6_dev);
1887 * Set destination address.
1888 * Special case for SIT interfaces where we create a new "virtual"
1889 * device.
1891 int addrconf_set_dstaddr(void __user *arg)
1893 struct in6_ifreq ireq;
1894 struct net_device *dev;
1895 int err = -EINVAL;
1897 rtnl_lock();
1899 err = -EFAULT;
1900 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1901 goto err_exit;
1903 dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1905 err = -ENODEV;
1906 if (dev == NULL)
1907 goto err_exit;
1909 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1910 if (dev->type == ARPHRD_SIT) {
1911 struct ifreq ifr;
1912 mm_segment_t oldfs;
1913 struct ip_tunnel_parm p;
1915 err = -EADDRNOTAVAIL;
1916 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1917 goto err_exit;
1919 memset(&p, 0, sizeof(p));
1920 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1921 p.iph.saddr = 0;
1922 p.iph.version = 4;
1923 p.iph.ihl = 5;
1924 p.iph.protocol = IPPROTO_IPV6;
1925 p.iph.ttl = 64;
1926 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1928 oldfs = get_fs(); set_fs(KERNEL_DS);
1929 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1930 set_fs(oldfs);
1932 if (err == 0) {
1933 err = -ENOBUFS;
1934 if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1935 goto err_exit;
1936 err = dev_open(dev);
1939 #endif
1941 err_exit:
1942 rtnl_unlock();
1943 return err;
1947 * Manual configuration of address on an interface
1949 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1950 __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1952 struct inet6_ifaddr *ifp;
1953 struct inet6_dev *idev;
1954 struct net_device *dev;
1955 int scope;
1956 u32 flags = RTF_EXPIRES;
1958 ASSERT_RTNL();
1960 /* check the lifetime */
1961 if (!valid_lft || prefered_lft > valid_lft)
1962 return -EINVAL;
1964 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1965 return -ENODEV;
1967 if ((idev = addrconf_add_dev(dev)) == NULL)
1968 return -ENOBUFS;
1970 scope = ipv6_addr_scope(pfx);
1972 if (valid_lft == INFINITY_LIFE_TIME) {
1973 ifa_flags |= IFA_F_PERMANENT;
1974 flags = 0;
1975 } else if (valid_lft >= 0x7FFFFFFF/HZ)
1976 valid_lft = 0x7FFFFFFF/HZ;
1978 if (prefered_lft == 0)
1979 ifa_flags |= IFA_F_DEPRECATED;
1980 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1981 (prefered_lft != INFINITY_LIFE_TIME))
1982 prefered_lft = 0x7FFFFFFF/HZ;
1984 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1986 if (!IS_ERR(ifp)) {
1987 spin_lock_bh(&ifp->lock);
1988 ifp->valid_lft = valid_lft;
1989 ifp->prefered_lft = prefered_lft;
1990 ifp->tstamp = jiffies;
1991 spin_unlock_bh(&ifp->lock);
1993 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1994 jiffies_to_clock_t(valid_lft * HZ), flags);
1996 * Note that section 3.1 of RFC 4429 indicates
1997 * that the Optimistic flag should not be set for
1998 * manually configured addresses
2000 addrconf_dad_start(ifp, 0);
2001 in6_ifa_put(ifp);
2002 addrconf_verify(0);
2003 return 0;
2006 return PTR_ERR(ifp);
2009 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
2011 struct inet6_ifaddr *ifp;
2012 struct inet6_dev *idev;
2013 struct net_device *dev;
2015 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
2016 return -ENODEV;
2018 if ((idev = __in6_dev_get(dev)) == NULL)
2019 return -ENXIO;
2021 read_lock_bh(&idev->lock);
2022 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2023 if (ifp->prefix_len == plen &&
2024 ipv6_addr_equal(pfx, &ifp->addr)) {
2025 in6_ifa_hold(ifp);
2026 read_unlock_bh(&idev->lock);
2028 ipv6_del_addr(ifp);
2030 /* If the last address is deleted administratively,
2031 disable IPv6 on this interface.
2033 if (idev->addr_list == NULL)
2034 addrconf_ifdown(idev->dev, 1);
2035 return 0;
2038 read_unlock_bh(&idev->lock);
2039 return -EADDRNOTAVAIL;
2043 int addrconf_add_ifaddr(void __user *arg)
2045 struct in6_ifreq ireq;
2046 int err;
2048 if (!capable(CAP_NET_ADMIN))
2049 return -EPERM;
2051 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2052 return -EFAULT;
2054 rtnl_lock();
2055 err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2056 IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2057 rtnl_unlock();
2058 return err;
2061 int addrconf_del_ifaddr(void __user *arg)
2063 struct in6_ifreq ireq;
2064 int err;
2066 if (!capable(CAP_NET_ADMIN))
2067 return -EPERM;
2069 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2070 return -EFAULT;
2072 rtnl_lock();
2073 err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2074 rtnl_unlock();
2075 return err;
2078 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2079 static void sit_add_v4_addrs(struct inet6_dev *idev)
2081 struct inet6_ifaddr * ifp;
2082 struct in6_addr addr;
2083 struct net_device *dev;
2084 int scope;
2086 ASSERT_RTNL();
2088 memset(&addr, 0, sizeof(struct in6_addr));
2089 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2091 if (idev->dev->flags&IFF_POINTOPOINT) {
2092 addr.s6_addr32[0] = htonl(0xfe800000);
2093 scope = IFA_LINK;
2094 } else {
2095 scope = IPV6_ADDR_COMPATv4;
2098 if (addr.s6_addr32[3]) {
2099 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2100 if (!IS_ERR(ifp)) {
2101 spin_lock_bh(&ifp->lock);
2102 ifp->flags &= ~IFA_F_TENTATIVE;
2103 spin_unlock_bh(&ifp->lock);
2104 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2105 in6_ifa_put(ifp);
2107 return;
2110 for_each_netdev(&init_net, dev) {
2111 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2112 if (in_dev && (dev->flags & IFF_UP)) {
2113 struct in_ifaddr * ifa;
2115 int flag = scope;
2117 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2118 int plen;
2120 addr.s6_addr32[3] = ifa->ifa_local;
2122 if (ifa->ifa_scope == RT_SCOPE_LINK)
2123 continue;
2124 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2125 if (idev->dev->flags&IFF_POINTOPOINT)
2126 continue;
2127 flag |= IFA_HOST;
2129 if (idev->dev->flags&IFF_POINTOPOINT)
2130 plen = 64;
2131 else
2132 plen = 96;
2134 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2135 IFA_F_PERMANENT);
2136 if (!IS_ERR(ifp)) {
2137 spin_lock_bh(&ifp->lock);
2138 ifp->flags &= ~IFA_F_TENTATIVE;
2139 spin_unlock_bh(&ifp->lock);
2140 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2141 in6_ifa_put(ifp);
2147 #endif
2149 static void init_loopback(struct net_device *dev)
2151 struct inet6_dev *idev;
2152 struct inet6_ifaddr * ifp;
2154 /* ::1 */
2156 ASSERT_RTNL();
2158 if ((idev = ipv6_find_idev(dev)) == NULL) {
2159 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2160 return;
2163 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2164 if (!IS_ERR(ifp)) {
2165 spin_lock_bh(&ifp->lock);
2166 ifp->flags &= ~IFA_F_TENTATIVE;
2167 spin_unlock_bh(&ifp->lock);
2168 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2169 in6_ifa_put(ifp);
2173 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2175 struct inet6_ifaddr * ifp;
2176 u32 addr_flags = IFA_F_PERMANENT;
2178 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2179 if (idev->cnf.optimistic_dad &&
2180 !ipv6_devconf.forwarding)
2181 addr_flags |= IFA_F_OPTIMISTIC;
2182 #endif
2185 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2186 if (!IS_ERR(ifp)) {
2187 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2188 addrconf_dad_start(ifp, 0);
2189 in6_ifa_put(ifp);
2193 static void addrconf_dev_config(struct net_device *dev)
2195 struct in6_addr addr;
2196 struct inet6_dev * idev;
2198 ASSERT_RTNL();
2200 if ((dev->type != ARPHRD_ETHER) &&
2201 (dev->type != ARPHRD_FDDI) &&
2202 (dev->type != ARPHRD_IEEE802_TR) &&
2203 (dev->type != ARPHRD_ARCNET) &&
2204 (dev->type != ARPHRD_INFINIBAND)) {
2205 /* Alas, we support only Ethernet autoconfiguration. */
2206 return;
2209 idev = addrconf_add_dev(dev);
2210 if (idev == NULL)
2211 return;
2213 memset(&addr, 0, sizeof(struct in6_addr));
2214 addr.s6_addr32[0] = htonl(0xFE800000);
2216 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2217 addrconf_add_linklocal(idev, &addr);
2220 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2221 static void addrconf_sit_config(struct net_device *dev)
2223 struct inet6_dev *idev;
2225 ASSERT_RTNL();
2228 * Configure the tunnel with one of our IPv4
2229 * addresses... we should configure all of
2230 * our v4 addrs in the tunnel
2233 if ((idev = ipv6_find_idev(dev)) == NULL) {
2234 printk(KERN_DEBUG "init sit: add_dev failed\n");
2235 return;
2238 if (dev->priv_flags & IFF_ISATAP) {
2239 struct in6_addr addr;
2241 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2242 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2243 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2244 addrconf_add_linklocal(idev, &addr);
2245 return;
2248 sit_add_v4_addrs(idev);
2250 if (dev->flags&IFF_POINTOPOINT) {
2251 addrconf_add_mroute(dev);
2252 addrconf_add_lroute(dev);
2253 } else
2254 sit_route_add(dev);
2256 #endif
2258 static inline int
2259 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2261 struct in6_addr lladdr;
2263 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2264 addrconf_add_linklocal(idev, &lladdr);
2265 return 0;
2267 return -1;
2270 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2272 struct net_device *link_dev;
2274 /* first try to inherit the link-local address from the link device */
2275 if (idev->dev->iflink &&
2276 (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2277 if (!ipv6_inherit_linklocal(idev, link_dev))
2278 return;
2280 /* then try to inherit it from any device */
2281 for_each_netdev(&init_net, link_dev) {
2282 if (!ipv6_inherit_linklocal(idev, link_dev))
2283 return;
2285 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2289 * Autoconfigure tunnel with a link-local address so routing protocols,
2290 * DHCPv6, MLD etc. can be run over the virtual link
2293 static void addrconf_ip6_tnl_config(struct net_device *dev)
2295 struct inet6_dev *idev;
2297 ASSERT_RTNL();
2299 if ((idev = addrconf_add_dev(dev)) == NULL) {
2300 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2301 return;
2303 ip6_tnl_add_linklocal(idev);
2306 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2307 void * data)
2309 struct net_device *dev = (struct net_device *) data;
2310 struct inet6_dev *idev = __in6_dev_get(dev);
2311 int run_pending = 0;
2312 int err;
2314 if (dev->nd_net != &init_net)
2315 return NOTIFY_DONE;
2317 switch(event) {
2318 case NETDEV_REGISTER:
2319 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2320 idev = ipv6_add_dev(dev);
2321 if (!idev)
2322 return notifier_from_errno(-ENOMEM);
2324 break;
2325 case NETDEV_UP:
2326 case NETDEV_CHANGE:
2327 if (dev->flags & IFF_SLAVE)
2328 break;
2330 if (event == NETDEV_UP) {
2331 if (!addrconf_qdisc_ok(dev)) {
2332 /* device is not ready yet. */
2333 printk(KERN_INFO
2334 "ADDRCONF(NETDEV_UP): %s: "
2335 "link is not ready\n",
2336 dev->name);
2337 break;
2340 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2341 idev = ipv6_add_dev(dev);
2343 if (idev)
2344 idev->if_flags |= IF_READY;
2345 } else {
2346 if (!addrconf_qdisc_ok(dev)) {
2347 /* device is still not ready. */
2348 break;
2351 if (idev) {
2352 if (idev->if_flags & IF_READY) {
2353 /* device is already configured. */
2354 break;
2356 idev->if_flags |= IF_READY;
2359 printk(KERN_INFO
2360 "ADDRCONF(NETDEV_CHANGE): %s: "
2361 "link becomes ready\n",
2362 dev->name);
2364 run_pending = 1;
2367 switch(dev->type) {
2368 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2369 case ARPHRD_SIT:
2370 addrconf_sit_config(dev);
2371 break;
2372 #endif
2373 case ARPHRD_TUNNEL6:
2374 addrconf_ip6_tnl_config(dev);
2375 break;
2376 case ARPHRD_LOOPBACK:
2377 init_loopback(dev);
2378 break;
2380 default:
2381 addrconf_dev_config(dev);
2382 break;
2384 if (idev) {
2385 if (run_pending)
2386 addrconf_dad_run(idev);
2388 /* If the MTU changed during the interface down, when the
2389 interface up, the changed MTU must be reflected in the
2390 idev as well as routers.
2392 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2393 rt6_mtu_change(dev, dev->mtu);
2394 idev->cnf.mtu6 = dev->mtu;
2396 idev->tstamp = jiffies;
2397 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2398 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2399 stop IPv6 on this interface.
2401 if (dev->mtu < IPV6_MIN_MTU)
2402 addrconf_ifdown(dev, event != NETDEV_DOWN);
2404 break;
2406 case NETDEV_CHANGEMTU:
2407 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2408 rt6_mtu_change(dev, dev->mtu);
2409 idev->cnf.mtu6 = dev->mtu;
2410 break;
2413 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2414 idev = ipv6_add_dev(dev);
2415 if (idev)
2416 break;
2419 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2421 case NETDEV_DOWN:
2422 case NETDEV_UNREGISTER:
2424 * Remove all addresses from this interface.
2426 addrconf_ifdown(dev, event != NETDEV_DOWN);
2427 break;
2429 case NETDEV_CHANGENAME:
2430 if (idev) {
2431 snmp6_unregister_dev(idev);
2432 addrconf_sysctl_unregister(idev);
2433 addrconf_sysctl_register(idev);
2434 err = snmp6_register_dev(idev);
2435 if (err)
2436 return notifier_from_errno(err);
2438 break;
2441 return NOTIFY_OK;
2445 * addrconf module should be notified of a device going up
2447 static struct notifier_block ipv6_dev_notf = {
2448 .notifier_call = addrconf_notify,
2449 .priority = 0
2452 static int addrconf_ifdown(struct net_device *dev, int how)
2454 struct inet6_dev *idev;
2455 struct inet6_ifaddr *ifa, **bifa;
2456 int i;
2458 ASSERT_RTNL();
2460 if ((dev->flags & IFF_LOOPBACK) && how == 1)
2461 how = 0;
2463 rt6_ifdown(dev);
2464 neigh_ifdown(&nd_tbl, dev);
2466 idev = __in6_dev_get(dev);
2467 if (idev == NULL)
2468 return -ENODEV;
2470 /* Step 1: remove reference to ipv6 device from parent device.
2471 Do not dev_put!
2473 if (how) {
2474 idev->dead = 1;
2476 /* protected by rtnl_lock */
2477 rcu_assign_pointer(dev->ip6_ptr, NULL);
2479 /* Step 1.5: remove snmp6 entry */
2480 snmp6_unregister_dev(idev);
2484 /* Step 2: clear hash table */
2485 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2486 bifa = &inet6_addr_lst[i];
2488 write_lock_bh(&addrconf_hash_lock);
2489 while ((ifa = *bifa) != NULL) {
2490 if (ifa->idev == idev) {
2491 *bifa = ifa->lst_next;
2492 ifa->lst_next = NULL;
2493 addrconf_del_timer(ifa);
2494 in6_ifa_put(ifa);
2495 continue;
2497 bifa = &ifa->lst_next;
2499 write_unlock_bh(&addrconf_hash_lock);
2502 write_lock_bh(&idev->lock);
2504 /* Step 3: clear flags for stateless addrconf */
2505 if (!how)
2506 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2508 /* Step 4: clear address list */
2509 #ifdef CONFIG_IPV6_PRIVACY
2510 if (how && del_timer(&idev->regen_timer))
2511 in6_dev_put(idev);
2513 /* clear tempaddr list */
2514 while ((ifa = idev->tempaddr_list) != NULL) {
2515 idev->tempaddr_list = ifa->tmp_next;
2516 ifa->tmp_next = NULL;
2517 ifa->dead = 1;
2518 write_unlock_bh(&idev->lock);
2519 spin_lock_bh(&ifa->lock);
2521 if (ifa->ifpub) {
2522 in6_ifa_put(ifa->ifpub);
2523 ifa->ifpub = NULL;
2525 spin_unlock_bh(&ifa->lock);
2526 in6_ifa_put(ifa);
2527 write_lock_bh(&idev->lock);
2529 #endif
2530 while ((ifa = idev->addr_list) != NULL) {
2531 idev->addr_list = ifa->if_next;
2532 ifa->if_next = NULL;
2533 ifa->dead = 1;
2534 addrconf_del_timer(ifa);
2535 write_unlock_bh(&idev->lock);
2537 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2538 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2539 in6_ifa_put(ifa);
2541 write_lock_bh(&idev->lock);
2543 write_unlock_bh(&idev->lock);
2545 /* Step 5: Discard multicast list */
2547 if (how)
2548 ipv6_mc_destroy_dev(idev);
2549 else
2550 ipv6_mc_down(idev);
2552 idev->tstamp = jiffies;
2554 /* Shot the device (if unregistered) */
2556 if (how) {
2557 addrconf_sysctl_unregister(idev);
2558 neigh_parms_release(&nd_tbl, idev->nd_parms);
2559 neigh_ifdown(&nd_tbl, dev);
2560 in6_dev_put(idev);
2562 return 0;
2565 static void addrconf_rs_timer(unsigned long data)
2567 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2569 if (ifp->idev->cnf.forwarding)
2570 goto out;
2572 if (ifp->idev->if_flags & IF_RA_RCVD) {
2574 * Announcement received after solicitation
2575 * was sent
2577 goto out;
2580 spin_lock(&ifp->lock);
2581 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2582 struct in6_addr all_routers;
2584 /* The wait after the last probe can be shorter */
2585 addrconf_mod_timer(ifp, AC_RS,
2586 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2587 ifp->idev->cnf.rtr_solicit_delay :
2588 ifp->idev->cnf.rtr_solicit_interval);
2589 spin_unlock(&ifp->lock);
2591 ipv6_addr_all_routers(&all_routers);
2593 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2594 } else {
2595 spin_unlock(&ifp->lock);
2597 * Note: we do not support deprecated "all on-link"
2598 * assumption any longer.
2600 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2601 ifp->idev->dev->name);
2604 out:
2605 in6_ifa_put(ifp);
2609 * Duplicate Address Detection
2611 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2613 unsigned long rand_num;
2614 struct inet6_dev *idev = ifp->idev;
2616 if (ifp->flags & IFA_F_OPTIMISTIC)
2617 rand_num = 0;
2618 else
2619 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2621 ifp->probes = idev->cnf.dad_transmits;
2622 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2625 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2627 struct inet6_dev *idev = ifp->idev;
2628 struct net_device *dev = idev->dev;
2630 addrconf_join_solict(dev, &ifp->addr);
2632 net_srandom(ifp->addr.s6_addr32[3]);
2634 read_lock_bh(&idev->lock);
2635 if (ifp->dead)
2636 goto out;
2637 spin_lock_bh(&ifp->lock);
2639 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2640 !(ifp->flags&IFA_F_TENTATIVE) ||
2641 ifp->flags & IFA_F_NODAD) {
2642 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2643 spin_unlock_bh(&ifp->lock);
2644 read_unlock_bh(&idev->lock);
2646 addrconf_dad_completed(ifp);
2647 return;
2650 if (!(idev->if_flags & IF_READY)) {
2651 spin_unlock_bh(&ifp->lock);
2652 read_unlock_bh(&idev->lock);
2654 * If the defice is not ready:
2655 * - keep it tentative if it is a permanent address.
2656 * - otherwise, kill it.
2658 in6_ifa_hold(ifp);
2659 addrconf_dad_stop(ifp);
2660 return;
2664 * Optimistic nodes can start receiving
2665 * Frames right away
2667 if(ifp->flags & IFA_F_OPTIMISTIC)
2668 ip6_ins_rt(ifp->rt);
2670 addrconf_dad_kick(ifp);
2671 spin_unlock_bh(&ifp->lock);
2672 out:
2673 read_unlock_bh(&idev->lock);
2676 static void addrconf_dad_timer(unsigned long data)
2678 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2679 struct inet6_dev *idev = ifp->idev;
2680 struct in6_addr unspec;
2681 struct in6_addr mcaddr;
2683 read_lock_bh(&idev->lock);
2684 if (idev->dead) {
2685 read_unlock_bh(&idev->lock);
2686 goto out;
2688 spin_lock_bh(&ifp->lock);
2689 if (ifp->probes == 0) {
2691 * DAD was successful
2694 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2695 spin_unlock_bh(&ifp->lock);
2696 read_unlock_bh(&idev->lock);
2698 addrconf_dad_completed(ifp);
2700 goto out;
2703 ifp->probes--;
2704 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2705 spin_unlock_bh(&ifp->lock);
2706 read_unlock_bh(&idev->lock);
2708 /* send a neighbour solicitation for our addr */
2709 memset(&unspec, 0, sizeof(unspec));
2710 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2711 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2712 out:
2713 in6_ifa_put(ifp);
2716 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2718 struct net_device * dev = ifp->idev->dev;
2721 * Configure the address for reception. Now it is valid.
2724 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2726 /* If added prefix is link local and forwarding is off,
2727 start sending router solicitations.
2730 if (ifp->idev->cnf.forwarding == 0 &&
2731 ifp->idev->cnf.rtr_solicits > 0 &&
2732 (dev->flags&IFF_LOOPBACK) == 0 &&
2733 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2734 struct in6_addr all_routers;
2736 ipv6_addr_all_routers(&all_routers);
2739 * If a host as already performed a random delay
2740 * [...] as part of DAD [...] there is no need
2741 * to delay again before sending the first RS
2743 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2745 spin_lock_bh(&ifp->lock);
2746 ifp->probes = 1;
2747 ifp->idev->if_flags |= IF_RS_SENT;
2748 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2749 spin_unlock_bh(&ifp->lock);
2753 static void addrconf_dad_run(struct inet6_dev *idev) {
2754 struct inet6_ifaddr *ifp;
2756 read_lock_bh(&idev->lock);
2757 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2758 spin_lock_bh(&ifp->lock);
2759 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2760 spin_unlock_bh(&ifp->lock);
2761 continue;
2763 spin_unlock_bh(&ifp->lock);
2764 addrconf_dad_kick(ifp);
2766 read_unlock_bh(&idev->lock);
2769 #ifdef CONFIG_PROC_FS
2770 struct if6_iter_state {
2771 struct seq_net_private p;
2772 int bucket;
2775 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2777 struct inet6_ifaddr *ifa = NULL;
2778 struct if6_iter_state *state = seq->private;
2779 struct net *net = state->p.net;
2781 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2782 ifa = inet6_addr_lst[state->bucket];
2784 while (ifa && ifa->idev->dev->nd_net != net)
2785 ifa = ifa->lst_next;
2786 if (ifa)
2787 break;
2789 return ifa;
2792 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2794 struct if6_iter_state *state = seq->private;
2795 struct net *net = state->p.net;
2797 ifa = ifa->lst_next;
2798 try_again:
2799 if (ifa) {
2800 if (ifa->idev->dev->nd_net != net) {
2801 ifa = ifa->lst_next;
2802 goto try_again;
2806 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2807 ifa = inet6_addr_lst[state->bucket];
2808 goto try_again;
2811 return ifa;
2814 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2816 struct inet6_ifaddr *ifa = if6_get_first(seq);
2818 if (ifa)
2819 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2820 --pos;
2821 return pos ? NULL : ifa;
2824 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2825 __acquires(addrconf_hash_lock)
2827 read_lock_bh(&addrconf_hash_lock);
2828 return if6_get_idx(seq, *pos);
2831 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2833 struct inet6_ifaddr *ifa;
2835 ifa = if6_get_next(seq, v);
2836 ++*pos;
2837 return ifa;
2840 static void if6_seq_stop(struct seq_file *seq, void *v)
2841 __releases(addrconf_hash_lock)
2843 read_unlock_bh(&addrconf_hash_lock);
2846 static int if6_seq_show(struct seq_file *seq, void *v)
2848 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2849 seq_printf(seq,
2850 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2851 NIP6(ifp->addr),
2852 ifp->idev->dev->ifindex,
2853 ifp->prefix_len,
2854 ifp->scope,
2855 ifp->flags,
2856 ifp->idev->dev->name);
2857 return 0;
2860 static const struct seq_operations if6_seq_ops = {
2861 .start = if6_seq_start,
2862 .next = if6_seq_next,
2863 .show = if6_seq_show,
2864 .stop = if6_seq_stop,
2867 static int if6_seq_open(struct inode *inode, struct file *file)
2869 return seq_open_net(inode, file, &if6_seq_ops,
2870 sizeof(struct if6_iter_state));
2873 static const struct file_operations if6_fops = {
2874 .owner = THIS_MODULE,
2875 .open = if6_seq_open,
2876 .read = seq_read,
2877 .llseek = seq_lseek,
2878 .release = seq_release_net,
2881 static int if6_proc_net_init(struct net *net)
2883 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2884 return -ENOMEM;
2885 return 0;
2888 static void if6_proc_net_exit(struct net *net)
2890 proc_net_remove(net, "if_inet6");
2893 static struct pernet_operations if6_proc_net_ops = {
2894 .init = if6_proc_net_init,
2895 .exit = if6_proc_net_exit,
2898 int __init if6_proc_init(void)
2900 return register_pernet_subsys(&if6_proc_net_ops);
2903 void if6_proc_exit(void)
2905 unregister_pernet_subsys(&if6_proc_net_ops);
2907 #endif /* CONFIG_PROC_FS */
2909 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2910 /* Check if address is a home address configured on any interface. */
2911 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2913 int ret = 0;
2914 struct inet6_ifaddr * ifp;
2915 u8 hash = ipv6_addr_hash(addr);
2916 read_lock_bh(&addrconf_hash_lock);
2917 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2918 if (ifp->idev->dev->nd_net != net)
2919 continue;
2920 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2921 (ifp->flags & IFA_F_HOMEADDRESS)) {
2922 ret = 1;
2923 break;
2926 read_unlock_bh(&addrconf_hash_lock);
2927 return ret;
2929 #endif
2932 * Periodic address status verification
2935 static void addrconf_verify(unsigned long foo)
2937 struct inet6_ifaddr *ifp;
2938 unsigned long now, next;
2939 int i;
2941 spin_lock_bh(&addrconf_verify_lock);
2942 now = jiffies;
2943 next = now + ADDR_CHECK_FREQUENCY;
2945 del_timer(&addr_chk_timer);
2947 for (i=0; i < IN6_ADDR_HSIZE; i++) {
2949 restart:
2950 read_lock(&addrconf_hash_lock);
2951 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2952 unsigned long age;
2953 #ifdef CONFIG_IPV6_PRIVACY
2954 unsigned long regen_advance;
2955 #endif
2957 if (ifp->flags & IFA_F_PERMANENT)
2958 continue;
2960 spin_lock(&ifp->lock);
2961 age = (now - ifp->tstamp) / HZ;
2963 #ifdef CONFIG_IPV6_PRIVACY
2964 regen_advance = ifp->idev->cnf.regen_max_retry *
2965 ifp->idev->cnf.dad_transmits *
2966 ifp->idev->nd_parms->retrans_time / HZ;
2967 #endif
2969 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2970 age >= ifp->valid_lft) {
2971 spin_unlock(&ifp->lock);
2972 in6_ifa_hold(ifp);
2973 read_unlock(&addrconf_hash_lock);
2974 ipv6_del_addr(ifp);
2975 goto restart;
2976 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2977 spin_unlock(&ifp->lock);
2978 continue;
2979 } else if (age >= ifp->prefered_lft) {
2980 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2981 int deprecate = 0;
2983 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2984 deprecate = 1;
2985 ifp->flags |= IFA_F_DEPRECATED;
2988 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2989 next = ifp->tstamp + ifp->valid_lft * HZ;
2991 spin_unlock(&ifp->lock);
2993 if (deprecate) {
2994 in6_ifa_hold(ifp);
2995 read_unlock(&addrconf_hash_lock);
2997 ipv6_ifa_notify(0, ifp);
2998 in6_ifa_put(ifp);
2999 goto restart;
3001 #ifdef CONFIG_IPV6_PRIVACY
3002 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3003 !(ifp->flags&IFA_F_TENTATIVE)) {
3004 if (age >= ifp->prefered_lft - regen_advance) {
3005 struct inet6_ifaddr *ifpub = ifp->ifpub;
3006 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3007 next = ifp->tstamp + ifp->prefered_lft * HZ;
3008 if (!ifp->regen_count && ifpub) {
3009 ifp->regen_count++;
3010 in6_ifa_hold(ifp);
3011 in6_ifa_hold(ifpub);
3012 spin_unlock(&ifp->lock);
3013 read_unlock(&addrconf_hash_lock);
3014 spin_lock(&ifpub->lock);
3015 ifpub->regen_count = 0;
3016 spin_unlock(&ifpub->lock);
3017 ipv6_create_tempaddr(ifpub, ifp);
3018 in6_ifa_put(ifpub);
3019 in6_ifa_put(ifp);
3020 goto restart;
3022 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3023 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3024 spin_unlock(&ifp->lock);
3025 #endif
3026 } else {
3027 /* ifp->prefered_lft <= ifp->valid_lft */
3028 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3029 next = ifp->tstamp + ifp->prefered_lft * HZ;
3030 spin_unlock(&ifp->lock);
3033 read_unlock(&addrconf_hash_lock);
3036 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3037 add_timer(&addr_chk_timer);
3038 spin_unlock_bh(&addrconf_verify_lock);
3041 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3043 struct in6_addr *pfx = NULL;
3045 if (addr)
3046 pfx = nla_data(addr);
3048 if (local) {
3049 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3050 pfx = NULL;
3051 else
3052 pfx = nla_data(local);
3055 return pfx;
3058 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3059 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3060 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3061 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3064 static int
3065 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3067 struct net *net = skb->sk->sk_net;
3068 struct ifaddrmsg *ifm;
3069 struct nlattr *tb[IFA_MAX+1];
3070 struct in6_addr *pfx;
3071 int err;
3073 if (net != &init_net)
3074 return -EINVAL;
3076 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3077 if (err < 0)
3078 return err;
3080 ifm = nlmsg_data(nlh);
3081 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3082 if (pfx == NULL)
3083 return -EINVAL;
3085 return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3088 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3089 u32 prefered_lft, u32 valid_lft)
3091 u32 flags = RTF_EXPIRES;
3093 if (!valid_lft || (prefered_lft > valid_lft))
3094 return -EINVAL;
3096 if (valid_lft == INFINITY_LIFE_TIME) {
3097 ifa_flags |= IFA_F_PERMANENT;
3098 flags = 0;
3099 } else if (valid_lft >= 0x7FFFFFFF/HZ)
3100 valid_lft = 0x7FFFFFFF/HZ;
3102 if (prefered_lft == 0)
3103 ifa_flags |= IFA_F_DEPRECATED;
3104 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3105 (prefered_lft != INFINITY_LIFE_TIME))
3106 prefered_lft = 0x7FFFFFFF/HZ;
3108 spin_lock_bh(&ifp->lock);
3109 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3110 ifp->tstamp = jiffies;
3111 ifp->valid_lft = valid_lft;
3112 ifp->prefered_lft = prefered_lft;
3114 spin_unlock_bh(&ifp->lock);
3115 if (!(ifp->flags&IFA_F_TENTATIVE))
3116 ipv6_ifa_notify(0, ifp);
3118 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3119 jiffies_to_clock_t(valid_lft * HZ), flags);
3120 addrconf_verify(0);
3122 return 0;
3125 static int
3126 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3128 struct net *net = skb->sk->sk_net;
3129 struct ifaddrmsg *ifm;
3130 struct nlattr *tb[IFA_MAX+1];
3131 struct in6_addr *pfx;
3132 struct inet6_ifaddr *ifa;
3133 struct net_device *dev;
3134 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3135 u8 ifa_flags;
3136 int err;
3138 if (net != &init_net)
3139 return -EINVAL;
3141 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3142 if (err < 0)
3143 return err;
3145 ifm = nlmsg_data(nlh);
3146 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3147 if (pfx == NULL)
3148 return -EINVAL;
3150 if (tb[IFA_CACHEINFO]) {
3151 struct ifa_cacheinfo *ci;
3153 ci = nla_data(tb[IFA_CACHEINFO]);
3154 valid_lft = ci->ifa_valid;
3155 preferred_lft = ci->ifa_prefered;
3156 } else {
3157 preferred_lft = INFINITY_LIFE_TIME;
3158 valid_lft = INFINITY_LIFE_TIME;
3161 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3162 if (dev == NULL)
3163 return -ENODEV;
3165 /* We ignore other flags so far. */
3166 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3168 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3169 if (ifa == NULL) {
3171 * It would be best to check for !NLM_F_CREATE here but
3172 * userspace alreay relies on not having to provide this.
3174 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3175 ifa_flags, preferred_lft, valid_lft);
3178 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3179 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3180 err = -EEXIST;
3181 else
3182 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3184 in6_ifa_put(ifa);
3186 return err;
3189 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3190 u8 scope, int ifindex)
3192 struct ifaddrmsg *ifm;
3194 ifm = nlmsg_data(nlh);
3195 ifm->ifa_family = AF_INET6;
3196 ifm->ifa_prefixlen = prefixlen;
3197 ifm->ifa_flags = flags;
3198 ifm->ifa_scope = scope;
3199 ifm->ifa_index = ifindex;
3202 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3203 unsigned long tstamp, u32 preferred, u32 valid)
3205 struct ifa_cacheinfo ci;
3207 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3208 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3209 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3210 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3211 ci.ifa_prefered = preferred;
3212 ci.ifa_valid = valid;
3214 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3217 static inline int rt_scope(int ifa_scope)
3219 if (ifa_scope & IFA_HOST)
3220 return RT_SCOPE_HOST;
3221 else if (ifa_scope & IFA_LINK)
3222 return RT_SCOPE_LINK;
3223 else if (ifa_scope & IFA_SITE)
3224 return RT_SCOPE_SITE;
3225 else
3226 return RT_SCOPE_UNIVERSE;
3229 static inline int inet6_ifaddr_msgsize(void)
3231 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3232 + nla_total_size(16) /* IFA_ADDRESS */
3233 + nla_total_size(sizeof(struct ifa_cacheinfo));
3236 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3237 u32 pid, u32 seq, int event, unsigned int flags)
3239 struct nlmsghdr *nlh;
3240 u32 preferred, valid;
3242 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3243 if (nlh == NULL)
3244 return -EMSGSIZE;
3246 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3247 ifa->idev->dev->ifindex);
3249 if (!(ifa->flags&IFA_F_PERMANENT)) {
3250 preferred = ifa->prefered_lft;
3251 valid = ifa->valid_lft;
3252 if (preferred != INFINITY_LIFE_TIME) {
3253 long tval = (jiffies - ifa->tstamp)/HZ;
3254 preferred -= tval;
3255 if (valid != INFINITY_LIFE_TIME)
3256 valid -= tval;
3258 } else {
3259 preferred = INFINITY_LIFE_TIME;
3260 valid = INFINITY_LIFE_TIME;
3263 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3264 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3265 nlmsg_cancel(skb, nlh);
3266 return -EMSGSIZE;
3269 return nlmsg_end(skb, nlh);
3272 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3273 u32 pid, u32 seq, int event, u16 flags)
3275 struct nlmsghdr *nlh;
3276 u8 scope = RT_SCOPE_UNIVERSE;
3277 int ifindex = ifmca->idev->dev->ifindex;
3279 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3280 scope = RT_SCOPE_SITE;
3282 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3283 if (nlh == NULL)
3284 return -EMSGSIZE;
3286 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3287 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3288 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3289 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3290 nlmsg_cancel(skb, nlh);
3291 return -EMSGSIZE;
3294 return nlmsg_end(skb, nlh);
3297 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3298 u32 pid, u32 seq, int event, unsigned int flags)
3300 struct nlmsghdr *nlh;
3301 u8 scope = RT_SCOPE_UNIVERSE;
3302 int ifindex = ifaca->aca_idev->dev->ifindex;
3304 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3305 scope = RT_SCOPE_SITE;
3307 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3308 if (nlh == NULL)
3309 return -EMSGSIZE;
3311 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3312 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3313 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3314 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3315 nlmsg_cancel(skb, nlh);
3316 return -EMSGSIZE;
3319 return nlmsg_end(skb, nlh);
3322 enum addr_type_t
3324 UNICAST_ADDR,
3325 MULTICAST_ADDR,
3326 ANYCAST_ADDR,
3329 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3330 enum addr_type_t type)
3332 int idx, ip_idx;
3333 int s_idx, s_ip_idx;
3334 int err = 1;
3335 struct net_device *dev;
3336 struct inet6_dev *idev = NULL;
3337 struct inet6_ifaddr *ifa;
3338 struct ifmcaddr6 *ifmca;
3339 struct ifacaddr6 *ifaca;
3341 s_idx = cb->args[0];
3342 s_ip_idx = ip_idx = cb->args[1];
3344 idx = 0;
3345 for_each_netdev(&init_net, dev) {
3346 if (idx < s_idx)
3347 goto cont;
3348 if (idx > s_idx)
3349 s_ip_idx = 0;
3350 ip_idx = 0;
3351 if ((idev = in6_dev_get(dev)) == NULL)
3352 goto cont;
3353 read_lock_bh(&idev->lock);
3354 switch (type) {
3355 case UNICAST_ADDR:
3356 /* unicast address incl. temp addr */
3357 for (ifa = idev->addr_list; ifa;
3358 ifa = ifa->if_next, ip_idx++) {
3359 if (ip_idx < s_ip_idx)
3360 continue;
3361 err = inet6_fill_ifaddr(skb, ifa,
3362 NETLINK_CB(cb->skb).pid,
3363 cb->nlh->nlmsg_seq,
3364 RTM_NEWADDR,
3365 NLM_F_MULTI);
3367 break;
3368 case MULTICAST_ADDR:
3369 /* multicast address */
3370 for (ifmca = idev->mc_list; ifmca;
3371 ifmca = ifmca->next, ip_idx++) {
3372 if (ip_idx < s_ip_idx)
3373 continue;
3374 err = inet6_fill_ifmcaddr(skb, ifmca,
3375 NETLINK_CB(cb->skb).pid,
3376 cb->nlh->nlmsg_seq,
3377 RTM_GETMULTICAST,
3378 NLM_F_MULTI);
3380 break;
3381 case ANYCAST_ADDR:
3382 /* anycast address */
3383 for (ifaca = idev->ac_list; ifaca;
3384 ifaca = ifaca->aca_next, ip_idx++) {
3385 if (ip_idx < s_ip_idx)
3386 continue;
3387 err = inet6_fill_ifacaddr(skb, ifaca,
3388 NETLINK_CB(cb->skb).pid,
3389 cb->nlh->nlmsg_seq,
3390 RTM_GETANYCAST,
3391 NLM_F_MULTI);
3393 break;
3394 default:
3395 break;
3397 read_unlock_bh(&idev->lock);
3398 in6_dev_put(idev);
3400 if (err <= 0)
3401 break;
3402 cont:
3403 idx++;
3405 cb->args[0] = idx;
3406 cb->args[1] = ip_idx;
3407 return skb->len;
3410 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3412 struct net *net = skb->sk->sk_net;
3413 enum addr_type_t type = UNICAST_ADDR;
3415 if (net != &init_net)
3416 return 0;
3418 return inet6_dump_addr(skb, cb, type);
3421 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3423 struct net *net = skb->sk->sk_net;
3424 enum addr_type_t type = MULTICAST_ADDR;
3426 if (net != &init_net)
3427 return 0;
3429 return inet6_dump_addr(skb, cb, type);
3433 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3435 struct net *net = skb->sk->sk_net;
3436 enum addr_type_t type = ANYCAST_ADDR;
3438 if (net != &init_net)
3439 return 0;
3441 return inet6_dump_addr(skb, cb, type);
3444 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3445 void *arg)
3447 struct net *net = in_skb->sk->sk_net;
3448 struct ifaddrmsg *ifm;
3449 struct nlattr *tb[IFA_MAX+1];
3450 struct in6_addr *addr = NULL;
3451 struct net_device *dev = NULL;
3452 struct inet6_ifaddr *ifa;
3453 struct sk_buff *skb;
3454 int err;
3456 if (net != &init_net)
3457 return -EINVAL;
3459 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3460 if (err < 0)
3461 goto errout;
3463 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3464 if (addr == NULL) {
3465 err = -EINVAL;
3466 goto errout;
3469 ifm = nlmsg_data(nlh);
3470 if (ifm->ifa_index)
3471 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3473 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3474 err = -EADDRNOTAVAIL;
3475 goto errout;
3478 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3479 err = -ENOBUFS;
3480 goto errout_ifa;
3483 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3484 nlh->nlmsg_seq, RTM_NEWADDR, 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_ifa;
3491 err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
3492 errout_ifa:
3493 in6_ifa_put(ifa);
3494 errout:
3495 return err;
3498 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3500 struct sk_buff *skb;
3501 int err = -ENOBUFS;
3503 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3504 if (skb == NULL)
3505 goto errout;
3507 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3508 if (err < 0) {
3509 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3510 WARN_ON(err == -EMSGSIZE);
3511 kfree_skb(skb);
3512 goto errout;
3514 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3515 errout:
3516 if (err < 0)
3517 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3520 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3521 __s32 *array, int bytes)
3523 BUG_ON(bytes < (DEVCONF_MAX * 4));
3525 memset(array, 0, bytes);
3526 array[DEVCONF_FORWARDING] = cnf->forwarding;
3527 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3528 array[DEVCONF_MTU6] = cnf->mtu6;
3529 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3530 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3531 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3532 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3533 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3534 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3535 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3536 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3537 #ifdef CONFIG_IPV6_PRIVACY
3538 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3539 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3540 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3541 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3542 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3543 #endif
3544 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3545 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3546 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3547 #ifdef CONFIG_IPV6_ROUTER_PREF
3548 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3549 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3550 #ifdef CONFIG_IPV6_ROUTE_INFO
3551 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3552 #endif
3553 #endif
3554 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3555 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3556 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3557 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3558 #endif
3561 static inline size_t inet6_if_nlmsg_size(void)
3563 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3564 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3565 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3566 + nla_total_size(4) /* IFLA_MTU */
3567 + nla_total_size(4) /* IFLA_LINK */
3568 + nla_total_size( /* IFLA_PROTINFO */
3569 nla_total_size(4) /* IFLA_INET6_FLAGS */
3570 + nla_total_size(sizeof(struct ifla_cacheinfo))
3571 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3572 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3573 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3577 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3578 int bytes)
3580 int i;
3581 int pad = bytes - sizeof(u64) * items;
3582 BUG_ON(pad < 0);
3584 /* Use put_unaligned() because stats may not be aligned for u64. */
3585 put_unaligned(items, &stats[0]);
3586 for (i = 1; i < items; i++)
3587 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3589 memset(&stats[items], 0, pad);
3592 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3593 int bytes)
3595 switch(attrtype) {
3596 case IFLA_INET6_STATS:
3597 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3598 break;
3599 case IFLA_INET6_ICMP6STATS:
3600 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3601 break;
3605 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3606 u32 pid, u32 seq, int event, unsigned int flags)
3608 struct net_device *dev = idev->dev;
3609 struct nlattr *nla;
3610 struct ifinfomsg *hdr;
3611 struct nlmsghdr *nlh;
3612 void *protoinfo;
3613 struct ifla_cacheinfo ci;
3615 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3616 if (nlh == NULL)
3617 return -EMSGSIZE;
3619 hdr = nlmsg_data(nlh);
3620 hdr->ifi_family = AF_INET6;
3621 hdr->__ifi_pad = 0;
3622 hdr->ifi_type = dev->type;
3623 hdr->ifi_index = dev->ifindex;
3624 hdr->ifi_flags = dev_get_flags(dev);
3625 hdr->ifi_change = 0;
3627 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3629 if (dev->addr_len)
3630 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3632 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3633 if (dev->ifindex != dev->iflink)
3634 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3636 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3637 if (protoinfo == NULL)
3638 goto nla_put_failure;
3640 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3642 ci.max_reasm_len = IPV6_MAXPLEN;
3643 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3644 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3645 ci.reachable_time = idev->nd_parms->reachable_time;
3646 ci.retrans_time = idev->nd_parms->retrans_time;
3647 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3649 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3650 if (nla == NULL)
3651 goto nla_put_failure;
3652 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3654 /* XXX - MC not implemented */
3656 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3657 if (nla == NULL)
3658 goto nla_put_failure;
3659 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3661 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3662 if (nla == NULL)
3663 goto nla_put_failure;
3664 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3666 nla_nest_end(skb, protoinfo);
3667 return nlmsg_end(skb, nlh);
3669 nla_put_failure:
3670 nlmsg_cancel(skb, nlh);
3671 return -EMSGSIZE;
3674 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3676 struct net *net = skb->sk->sk_net;
3677 int idx, err;
3678 int s_idx = cb->args[0];
3679 struct net_device *dev;
3680 struct inet6_dev *idev;
3682 if (net != &init_net)
3683 return 0;
3685 read_lock(&dev_base_lock);
3686 idx = 0;
3687 for_each_netdev(&init_net, dev) {
3688 if (idx < s_idx)
3689 goto cont;
3690 if ((idev = in6_dev_get(dev)) == NULL)
3691 goto cont;
3692 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3693 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3694 in6_dev_put(idev);
3695 if (err <= 0)
3696 break;
3697 cont:
3698 idx++;
3700 read_unlock(&dev_base_lock);
3701 cb->args[0] = idx;
3703 return skb->len;
3706 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3708 struct sk_buff *skb;
3709 int err = -ENOBUFS;
3711 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3712 if (skb == NULL)
3713 goto errout;
3715 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3716 if (err < 0) {
3717 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3718 WARN_ON(err == -EMSGSIZE);
3719 kfree_skb(skb);
3720 goto errout;
3722 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3723 errout:
3724 if (err < 0)
3725 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_IFADDR, err);
3728 static inline size_t inet6_prefix_nlmsg_size(void)
3730 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3731 + nla_total_size(sizeof(struct in6_addr))
3732 + nla_total_size(sizeof(struct prefix_cacheinfo));
3735 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3736 struct prefix_info *pinfo, u32 pid, u32 seq,
3737 int event, unsigned int flags)
3739 struct prefixmsg *pmsg;
3740 struct nlmsghdr *nlh;
3741 struct prefix_cacheinfo ci;
3743 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3744 if (nlh == NULL)
3745 return -EMSGSIZE;
3747 pmsg = nlmsg_data(nlh);
3748 pmsg->prefix_family = AF_INET6;
3749 pmsg->prefix_pad1 = 0;
3750 pmsg->prefix_pad2 = 0;
3751 pmsg->prefix_ifindex = idev->dev->ifindex;
3752 pmsg->prefix_len = pinfo->prefix_len;
3753 pmsg->prefix_type = pinfo->type;
3754 pmsg->prefix_pad3 = 0;
3755 pmsg->prefix_flags = 0;
3756 if (pinfo->onlink)
3757 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3758 if (pinfo->autoconf)
3759 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3761 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3763 ci.preferred_time = ntohl(pinfo->prefered);
3764 ci.valid_time = ntohl(pinfo->valid);
3765 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3767 return nlmsg_end(skb, nlh);
3769 nla_put_failure:
3770 nlmsg_cancel(skb, nlh);
3771 return -EMSGSIZE;
3774 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3775 struct prefix_info *pinfo)
3777 struct sk_buff *skb;
3778 int err = -ENOBUFS;
3780 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3781 if (skb == NULL)
3782 goto errout;
3784 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3785 if (err < 0) {
3786 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3787 WARN_ON(err == -EMSGSIZE);
3788 kfree_skb(skb);
3789 goto errout;
3791 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3792 errout:
3793 if (err < 0)
3794 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_PREFIX, err);
3797 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3799 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3801 switch (event) {
3802 case RTM_NEWADDR:
3804 * If the address was optimistic
3805 * we inserted the route at the start of
3806 * our DAD process, so we don't need
3807 * to do it again
3809 if (!(ifp->rt->rt6i_node))
3810 ip6_ins_rt(ifp->rt);
3811 if (ifp->idev->cnf.forwarding)
3812 addrconf_join_anycast(ifp);
3813 break;
3814 case RTM_DELADDR:
3815 if (ifp->idev->cnf.forwarding)
3816 addrconf_leave_anycast(ifp);
3817 addrconf_leave_solict(ifp->idev, &ifp->addr);
3818 dst_hold(&ifp->rt->u.dst);
3819 if (ip6_del_rt(ifp->rt))
3820 dst_free(&ifp->rt->u.dst);
3821 break;
3825 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3827 rcu_read_lock_bh();
3828 if (likely(ifp->idev->dead == 0))
3829 __ipv6_ifa_notify(event, ifp);
3830 rcu_read_unlock_bh();
3833 #ifdef CONFIG_SYSCTL
3835 static
3836 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3837 void __user *buffer, size_t *lenp, loff_t *ppos)
3839 int *valp = ctl->data;
3840 int val = *valp;
3841 int ret;
3843 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3845 if (write)
3846 addrconf_fixup_forwarding(ctl, valp, val);
3847 return ret;
3850 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3851 int __user *name, int nlen,
3852 void __user *oldval,
3853 size_t __user *oldlenp,
3854 void __user *newval, size_t newlen)
3856 int *valp = table->data;
3857 int val = *valp;
3858 int new;
3860 if (!newval || !newlen)
3861 return 0;
3862 if (newlen != sizeof(int))
3863 return -EINVAL;
3864 if (get_user(new, (int __user *)newval))
3865 return -EFAULT;
3866 if (new == *valp)
3867 return 0;
3868 if (oldval && oldlenp) {
3869 size_t len;
3870 if (get_user(len, oldlenp))
3871 return -EFAULT;
3872 if (len) {
3873 if (len > table->maxlen)
3874 len = table->maxlen;
3875 if (copy_to_user(oldval, valp, len))
3876 return -EFAULT;
3877 if (put_user(len, oldlenp))
3878 return -EFAULT;
3882 *valp = new;
3883 addrconf_fixup_forwarding(table, valp, val);
3884 return 1;
3887 static struct addrconf_sysctl_table
3889 struct ctl_table_header *sysctl_header;
3890 ctl_table addrconf_vars[__NET_IPV6_MAX];
3891 char *dev_name;
3892 } addrconf_sysctl __read_mostly = {
3893 .sysctl_header = NULL,
3894 .addrconf_vars = {
3896 .ctl_name = NET_IPV6_FORWARDING,
3897 .procname = "forwarding",
3898 .data = &ipv6_devconf.forwarding,
3899 .maxlen = sizeof(int),
3900 .mode = 0644,
3901 .proc_handler = &addrconf_sysctl_forward,
3902 .strategy = &addrconf_sysctl_forward_strategy,
3905 .ctl_name = NET_IPV6_HOP_LIMIT,
3906 .procname = "hop_limit",
3907 .data = &ipv6_devconf.hop_limit,
3908 .maxlen = sizeof(int),
3909 .mode = 0644,
3910 .proc_handler = proc_dointvec,
3913 .ctl_name = NET_IPV6_MTU,
3914 .procname = "mtu",
3915 .data = &ipv6_devconf.mtu6,
3916 .maxlen = sizeof(int),
3917 .mode = 0644,
3918 .proc_handler = &proc_dointvec,
3921 .ctl_name = NET_IPV6_ACCEPT_RA,
3922 .procname = "accept_ra",
3923 .data = &ipv6_devconf.accept_ra,
3924 .maxlen = sizeof(int),
3925 .mode = 0644,
3926 .proc_handler = &proc_dointvec,
3929 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
3930 .procname = "accept_redirects",
3931 .data = &ipv6_devconf.accept_redirects,
3932 .maxlen = sizeof(int),
3933 .mode = 0644,
3934 .proc_handler = &proc_dointvec,
3937 .ctl_name = NET_IPV6_AUTOCONF,
3938 .procname = "autoconf",
3939 .data = &ipv6_devconf.autoconf,
3940 .maxlen = sizeof(int),
3941 .mode = 0644,
3942 .proc_handler = &proc_dointvec,
3945 .ctl_name = NET_IPV6_DAD_TRANSMITS,
3946 .procname = "dad_transmits",
3947 .data = &ipv6_devconf.dad_transmits,
3948 .maxlen = sizeof(int),
3949 .mode = 0644,
3950 .proc_handler = &proc_dointvec,
3953 .ctl_name = NET_IPV6_RTR_SOLICITS,
3954 .procname = "router_solicitations",
3955 .data = &ipv6_devconf.rtr_solicits,
3956 .maxlen = sizeof(int),
3957 .mode = 0644,
3958 .proc_handler = &proc_dointvec,
3961 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
3962 .procname = "router_solicitation_interval",
3963 .data = &ipv6_devconf.rtr_solicit_interval,
3964 .maxlen = sizeof(int),
3965 .mode = 0644,
3966 .proc_handler = &proc_dointvec_jiffies,
3967 .strategy = &sysctl_jiffies,
3970 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
3971 .procname = "router_solicitation_delay",
3972 .data = &ipv6_devconf.rtr_solicit_delay,
3973 .maxlen = sizeof(int),
3974 .mode = 0644,
3975 .proc_handler = &proc_dointvec_jiffies,
3976 .strategy = &sysctl_jiffies,
3979 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
3980 .procname = "force_mld_version",
3981 .data = &ipv6_devconf.force_mld_version,
3982 .maxlen = sizeof(int),
3983 .mode = 0644,
3984 .proc_handler = &proc_dointvec,
3986 #ifdef CONFIG_IPV6_PRIVACY
3988 .ctl_name = NET_IPV6_USE_TEMPADDR,
3989 .procname = "use_tempaddr",
3990 .data = &ipv6_devconf.use_tempaddr,
3991 .maxlen = sizeof(int),
3992 .mode = 0644,
3993 .proc_handler = &proc_dointvec,
3996 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
3997 .procname = "temp_valid_lft",
3998 .data = &ipv6_devconf.temp_valid_lft,
3999 .maxlen = sizeof(int),
4000 .mode = 0644,
4001 .proc_handler = &proc_dointvec,
4004 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4005 .procname = "temp_prefered_lft",
4006 .data = &ipv6_devconf.temp_prefered_lft,
4007 .maxlen = sizeof(int),
4008 .mode = 0644,
4009 .proc_handler = &proc_dointvec,
4012 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4013 .procname = "regen_max_retry",
4014 .data = &ipv6_devconf.regen_max_retry,
4015 .maxlen = sizeof(int),
4016 .mode = 0644,
4017 .proc_handler = &proc_dointvec,
4020 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4021 .procname = "max_desync_factor",
4022 .data = &ipv6_devconf.max_desync_factor,
4023 .maxlen = sizeof(int),
4024 .mode = 0644,
4025 .proc_handler = &proc_dointvec,
4027 #endif
4029 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4030 .procname = "max_addresses",
4031 .data = &ipv6_devconf.max_addresses,
4032 .maxlen = sizeof(int),
4033 .mode = 0644,
4034 .proc_handler = &proc_dointvec,
4037 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4038 .procname = "accept_ra_defrtr",
4039 .data = &ipv6_devconf.accept_ra_defrtr,
4040 .maxlen = sizeof(int),
4041 .mode = 0644,
4042 .proc_handler = &proc_dointvec,
4045 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4046 .procname = "accept_ra_pinfo",
4047 .data = &ipv6_devconf.accept_ra_pinfo,
4048 .maxlen = sizeof(int),
4049 .mode = 0644,
4050 .proc_handler = &proc_dointvec,
4052 #ifdef CONFIG_IPV6_ROUTER_PREF
4054 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4055 .procname = "accept_ra_rtr_pref",
4056 .data = &ipv6_devconf.accept_ra_rtr_pref,
4057 .maxlen = sizeof(int),
4058 .mode = 0644,
4059 .proc_handler = &proc_dointvec,
4062 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4063 .procname = "router_probe_interval",
4064 .data = &ipv6_devconf.rtr_probe_interval,
4065 .maxlen = sizeof(int),
4066 .mode = 0644,
4067 .proc_handler = &proc_dointvec_jiffies,
4068 .strategy = &sysctl_jiffies,
4070 #ifdef CONFIG_IPV6_ROUTE_INFO
4072 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4073 .procname = "accept_ra_rt_info_max_plen",
4074 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4075 .maxlen = sizeof(int),
4076 .mode = 0644,
4077 .proc_handler = &proc_dointvec,
4079 #endif
4080 #endif
4082 .ctl_name = NET_IPV6_PROXY_NDP,
4083 .procname = "proxy_ndp",
4084 .data = &ipv6_devconf.proxy_ndp,
4085 .maxlen = sizeof(int),
4086 .mode = 0644,
4087 .proc_handler = &proc_dointvec,
4090 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4091 .procname = "accept_source_route",
4092 .data = &ipv6_devconf.accept_source_route,
4093 .maxlen = sizeof(int),
4094 .mode = 0644,
4095 .proc_handler = &proc_dointvec,
4097 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4099 .ctl_name = CTL_UNNUMBERED,
4100 .procname = "optimistic_dad",
4101 .data = &ipv6_devconf.optimistic_dad,
4102 .maxlen = sizeof(int),
4103 .mode = 0644,
4104 .proc_handler = &proc_dointvec,
4107 #endif
4109 .ctl_name = 0, /* sentinel */
4114 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4115 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4117 int i;
4118 struct addrconf_sysctl_table *t;
4120 #define ADDRCONF_CTL_PATH_DEV 3
4122 struct ctl_path addrconf_ctl_path[] = {
4123 { .procname = "net", .ctl_name = CTL_NET, },
4124 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4125 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4126 { /* to be set */ },
4127 { },
4131 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4132 if (t == NULL)
4133 goto out;
4135 for (i=0; t->addrconf_vars[i].data; i++) {
4136 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4137 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4138 t->addrconf_vars[i].extra2 = net;
4142 * Make a copy of dev_name, because '.procname' is regarded as const
4143 * by sysctl and we wouldn't want anyone to change it under our feet
4144 * (see SIOCSIFNAME).
4146 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4147 if (!t->dev_name)
4148 goto free;
4150 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4151 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4153 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4154 t->addrconf_vars);
4155 if (t->sysctl_header == NULL)
4156 goto free_procname;
4158 p->sysctl = t;
4159 return 0;
4161 free_procname:
4162 kfree(t->dev_name);
4163 free:
4164 kfree(t);
4165 out:
4166 return -ENOBUFS;
4169 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4171 struct addrconf_sysctl_table *t;
4173 if (p->sysctl == NULL)
4174 return;
4176 t = p->sysctl;
4177 p->sysctl = NULL;
4178 unregister_sysctl_table(t->sysctl_header);
4179 kfree(t->dev_name);
4180 kfree(t);
4183 static void addrconf_sysctl_register(struct inet6_dev *idev)
4185 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4186 NET_IPV6_NEIGH, "ipv6",
4187 &ndisc_ifinfo_sysctl_change,
4188 NULL);
4189 __addrconf_sysctl_register(idev->dev->nd_net, idev->dev->name,
4190 idev->dev->ifindex, idev, &idev->cnf);
4193 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4195 __addrconf_sysctl_unregister(&idev->cnf);
4196 neigh_sysctl_unregister(idev->nd_parms);
4200 #endif
4202 static int addrconf_init_net(struct net *net)
4204 int err;
4205 struct ipv6_devconf *all, *dflt;
4207 err = -ENOMEM;
4208 all = &ipv6_devconf;
4209 dflt = &ipv6_devconf_dflt;
4211 if (net != &init_net) {
4212 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4213 if (all == NULL)
4214 goto err_alloc_all;
4216 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4217 if (dflt == NULL)
4218 goto err_alloc_dflt;
4221 net->ipv6.devconf_all = all;
4222 net->ipv6.devconf_dflt = dflt;
4224 #ifdef CONFIG_SYSCTL
4225 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4226 NULL, all);
4227 if (err < 0)
4228 goto err_reg_all;
4230 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4231 NULL, dflt);
4232 if (err < 0)
4233 goto err_reg_dflt;
4234 #endif
4235 return 0;
4237 #ifdef CONFIG_SYSCTL
4238 err_reg_dflt:
4239 __addrconf_sysctl_unregister(all);
4240 err_reg_all:
4241 kfree(dflt);
4242 #endif
4243 err_alloc_dflt:
4244 kfree(all);
4245 err_alloc_all:
4246 return err;
4249 static void addrconf_exit_net(struct net *net)
4251 #ifdef CONFIG_SYSCTL
4252 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4253 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4254 #endif
4255 if (net != &init_net) {
4256 kfree(net->ipv6.devconf_dflt);
4257 kfree(net->ipv6.devconf_all);
4261 static struct pernet_operations addrconf_ops = {
4262 .init = addrconf_init_net,
4263 .exit = addrconf_exit_net,
4267 * Device notifier
4270 int register_inet6addr_notifier(struct notifier_block *nb)
4272 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4275 EXPORT_SYMBOL(register_inet6addr_notifier);
4277 int unregister_inet6addr_notifier(struct notifier_block *nb)
4279 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4282 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4285 * Init / cleanup code
4288 int __init addrconf_init(void)
4290 int err;
4292 if ((err = ipv6_addr_label_init()) < 0) {
4293 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4294 err);
4295 return err;
4298 register_pernet_subsys(&addrconf_ops);
4300 /* The addrconf netdev notifier requires that loopback_dev
4301 * has it's ipv6 private information allocated and setup
4302 * before it can bring up and give link-local addresses
4303 * to other devices which are up.
4305 * Unfortunately, loopback_dev is not necessarily the first
4306 * entry in the global dev_base list of net devices. In fact,
4307 * it is likely to be the very last entry on that list.
4308 * So this causes the notifier registry below to try and
4309 * give link-local addresses to all devices besides loopback_dev
4310 * first, then loopback_dev, which cases all the non-loopback_dev
4311 * devices to fail to get a link-local address.
4313 * So, as a temporary fix, allocate the ipv6 structure for
4314 * loopback_dev first by hand.
4315 * Longer term, all of the dependencies ipv6 has upon the loopback
4316 * device and it being up should be removed.
4318 rtnl_lock();
4319 if (!ipv6_add_dev(init_net.loopback_dev))
4320 err = -ENOMEM;
4321 rtnl_unlock();
4322 if (err)
4323 goto errlo;
4325 ip6_null_entry.u.dst.dev = init_net.loopback_dev;
4326 ip6_null_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4327 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4328 ip6_prohibit_entry.u.dst.dev = init_net.loopback_dev;
4329 ip6_prohibit_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4330 ip6_blk_hole_entry.u.dst.dev = init_net.loopback_dev;
4331 ip6_blk_hole_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4332 #endif
4334 register_netdevice_notifier(&ipv6_dev_notf);
4336 addrconf_verify(0);
4338 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4339 if (err < 0)
4340 goto errout;
4342 /* Only the first call to __rtnl_register can fail */
4343 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4344 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4345 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4346 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4347 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4349 ipv6_addr_label_rtnl_register();
4351 return 0;
4352 errout:
4353 unregister_netdevice_notifier(&ipv6_dev_notf);
4354 errlo:
4355 unregister_pernet_subsys(&addrconf_ops);
4357 return err;
4360 void addrconf_cleanup(void)
4362 struct net_device *dev;
4363 struct inet6_ifaddr *ifa;
4364 int i;
4366 unregister_netdevice_notifier(&ipv6_dev_notf);
4368 unregister_pernet_subsys(&addrconf_ops);
4370 rtnl_lock();
4373 * clean dev list.
4376 for_each_netdev(&init_net, dev) {
4377 if (__in6_dev_get(dev) == NULL)
4378 continue;
4379 addrconf_ifdown(dev, 1);
4381 addrconf_ifdown(init_net.loopback_dev, 2);
4384 * Check hash table.
4387 write_lock_bh(&addrconf_hash_lock);
4388 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4389 for (ifa=inet6_addr_lst[i]; ifa; ) {
4390 struct inet6_ifaddr *bifa;
4392 bifa = ifa;
4393 ifa = ifa->lst_next;
4394 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4395 /* Do not free it; something is wrong.
4396 Now we can investigate it with debugger.
4400 write_unlock_bh(&addrconf_hash_lock);
4402 del_timer(&addr_chk_timer);
4404 rtnl_unlock();