[NET]: Make the device list and device lookups per namespace.
[linux-2.6/libata-dev.git] / net / ipv6 / addrconf.c
blobee55be975407b113f72332d9b32c671b759d4666
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 <linux/if_tunnel.h>
78 #include <linux/rtnetlink.h>
80 #ifdef CONFIG_IPV6_PRIVACY
81 #include <linux/random.h>
82 #endif
84 #include <asm/uaccess.h>
85 #include <asm/unaligned.h>
87 #include <linux/proc_fs.h>
88 #include <linux/seq_file.h>
90 /* Set to 3 to get tracing... */
91 #define ACONF_DEBUG 2
93 #if ACONF_DEBUG >= 3
94 #define ADBG(x) printk x
95 #else
96 #define ADBG(x)
97 #endif
99 #define INFINITY_LIFE_TIME 0xFFFFFFFF
100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
102 #ifdef CONFIG_SYSCTL
103 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p);
104 static void addrconf_sysctl_unregister(struct ipv6_devconf *p);
105 #endif
107 #ifdef CONFIG_IPV6_PRIVACY
108 static int __ipv6_regen_rndid(struct inet6_dev *idev);
109 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
110 static void ipv6_regen_rndid(unsigned long data);
112 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
113 #endif
115 static int ipv6_count_addresses(struct inet6_dev *idev);
118 * Configured unicast address hash table
120 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
121 static DEFINE_RWLOCK(addrconf_hash_lock);
123 static void addrconf_verify(unsigned long);
125 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
126 static DEFINE_SPINLOCK(addrconf_verify_lock);
128 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
129 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
131 static int addrconf_ifdown(struct net_device *dev, int how);
133 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
134 static void addrconf_dad_timer(unsigned long data);
135 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
136 static void addrconf_dad_run(struct inet6_dev *idev);
137 static void addrconf_rs_timer(unsigned long data);
138 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
139 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
141 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
142 struct prefix_info *pinfo);
143 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
145 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
147 struct ipv6_devconf ipv6_devconf __read_mostly = {
148 .forwarding = 0,
149 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
150 .mtu6 = IPV6_MIN_MTU,
151 .accept_ra = 1,
152 .accept_redirects = 1,
153 .autoconf = 1,
154 .force_mld_version = 0,
155 .dad_transmits = 1,
156 .rtr_solicits = MAX_RTR_SOLICITATIONS,
157 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
158 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
159 #ifdef CONFIG_IPV6_PRIVACY
160 .use_tempaddr = 0,
161 .temp_valid_lft = TEMP_VALID_LIFETIME,
162 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
163 .regen_max_retry = REGEN_MAX_RETRY,
164 .max_desync_factor = MAX_DESYNC_FACTOR,
165 #endif
166 .max_addresses = IPV6_MAX_ADDRESSES,
167 .accept_ra_defrtr = 1,
168 .accept_ra_pinfo = 1,
169 #ifdef CONFIG_IPV6_ROUTER_PREF
170 .accept_ra_rtr_pref = 1,
171 .rtr_probe_interval = 60 * HZ,
172 #ifdef CONFIG_IPV6_ROUTE_INFO
173 .accept_ra_rt_info_max_plen = 0,
174 #endif
175 #endif
176 .proxy_ndp = 0,
177 .accept_source_route = 0, /* we do not accept RH0 by default. */
180 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
181 .forwarding = 0,
182 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
183 .mtu6 = IPV6_MIN_MTU,
184 .accept_ra = 1,
185 .accept_redirects = 1,
186 .autoconf = 1,
187 .dad_transmits = 1,
188 .rtr_solicits = MAX_RTR_SOLICITATIONS,
189 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
190 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
191 #ifdef CONFIG_IPV6_PRIVACY
192 .use_tempaddr = 0,
193 .temp_valid_lft = TEMP_VALID_LIFETIME,
194 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
195 .regen_max_retry = REGEN_MAX_RETRY,
196 .max_desync_factor = MAX_DESYNC_FACTOR,
197 #endif
198 .max_addresses = IPV6_MAX_ADDRESSES,
199 .accept_ra_defrtr = 1,
200 .accept_ra_pinfo = 1,
201 #ifdef CONFIG_IPV6_ROUTER_PREF
202 .accept_ra_rtr_pref = 1,
203 .rtr_probe_interval = 60 * HZ,
204 #ifdef CONFIG_IPV6_ROUTE_INFO
205 .accept_ra_rt_info_max_plen = 0,
206 #endif
207 #endif
208 .proxy_ndp = 0,
209 .accept_source_route = 0, /* we do not accept RH0 by default. */
212 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
213 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
214 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
216 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
218 if (del_timer(&ifp->timer))
219 __in6_ifa_put(ifp);
222 enum addrconf_timer_t
224 AC_NONE,
225 AC_DAD,
226 AC_RS,
229 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
230 enum addrconf_timer_t what,
231 unsigned long when)
233 if (!del_timer(&ifp->timer))
234 in6_ifa_hold(ifp);
236 switch (what) {
237 case AC_DAD:
238 ifp->timer.function = addrconf_dad_timer;
239 break;
240 case AC_RS:
241 ifp->timer.function = addrconf_rs_timer;
242 break;
243 default:;
245 ifp->timer.expires = jiffies + when;
246 add_timer(&ifp->timer);
249 static int snmp6_alloc_dev(struct inet6_dev *idev)
251 int err = -ENOMEM;
253 if (!idev || !idev->dev)
254 return -EINVAL;
256 if (snmp_mib_init((void **)idev->stats.ipv6,
257 sizeof(struct ipstats_mib),
258 __alignof__(struct ipstats_mib)) < 0)
259 goto err_ip;
260 if (snmp_mib_init((void **)idev->stats.icmpv6,
261 sizeof(struct icmpv6_mib),
262 __alignof__(struct icmpv6_mib)) < 0)
263 goto err_icmp;
265 return 0;
267 err_icmp:
268 snmp_mib_free((void **)idev->stats.ipv6);
269 err_ip:
270 return err;
273 static int snmp6_free_dev(struct inet6_dev *idev)
275 snmp_mib_free((void **)idev->stats.icmpv6);
276 snmp_mib_free((void **)idev->stats.ipv6);
277 return 0;
280 /* Nobody refers to this device, we may destroy it. */
282 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
284 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
285 kfree(idev);
288 void in6_dev_finish_destroy(struct inet6_dev *idev)
290 struct net_device *dev = idev->dev;
291 BUG_TRAP(idev->addr_list==NULL);
292 BUG_TRAP(idev->mc_list==NULL);
293 #ifdef NET_REFCNT_DEBUG
294 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
295 #endif
296 dev_put(dev);
297 if (!idev->dead) {
298 printk("Freeing alive inet6 device %p\n", idev);
299 return;
301 snmp6_free_dev(idev);
302 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
305 EXPORT_SYMBOL(in6_dev_finish_destroy);
307 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
309 struct inet6_dev *ndev;
310 struct in6_addr maddr;
312 ASSERT_RTNL();
314 if (dev->mtu < IPV6_MIN_MTU)
315 return NULL;
317 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
319 if (ndev == NULL)
320 return NULL;
322 rwlock_init(&ndev->lock);
323 ndev->dev = dev;
324 memcpy(&ndev->cnf, &ipv6_devconf_dflt, sizeof(ndev->cnf));
325 ndev->cnf.mtu6 = dev->mtu;
326 ndev->cnf.sysctl = NULL;
327 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
328 if (ndev->nd_parms == NULL) {
329 kfree(ndev);
330 return NULL;
332 /* We refer to the device */
333 dev_hold(dev);
335 if (snmp6_alloc_dev(ndev) < 0) {
336 ADBG((KERN_WARNING
337 "%s(): cannot allocate memory for statistics; dev=%s.\n",
338 __FUNCTION__, dev->name));
339 neigh_parms_release(&nd_tbl, ndev->nd_parms);
340 ndev->dead = 1;
341 in6_dev_finish_destroy(ndev);
342 return NULL;
345 if (snmp6_register_dev(ndev) < 0) {
346 ADBG((KERN_WARNING
347 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
348 __FUNCTION__, dev->name));
349 neigh_parms_release(&nd_tbl, ndev->nd_parms);
350 ndev->dead = 1;
351 in6_dev_finish_destroy(ndev);
352 return NULL;
355 /* One reference from device. We must do this before
356 * we invoke __ipv6_regen_rndid().
358 in6_dev_hold(ndev);
360 #ifdef CONFIG_IPV6_PRIVACY
361 init_timer(&ndev->regen_timer);
362 ndev->regen_timer.function = ipv6_regen_rndid;
363 ndev->regen_timer.data = (unsigned long) ndev;
364 if ((dev->flags&IFF_LOOPBACK) ||
365 dev->type == ARPHRD_TUNNEL ||
366 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
367 dev->type == ARPHRD_SIT ||
368 #endif
369 dev->type == ARPHRD_NONE) {
370 printk(KERN_INFO
371 "%s: Disabled Privacy Extensions\n",
372 dev->name);
373 ndev->cnf.use_tempaddr = -1;
374 } else {
375 in6_dev_hold(ndev);
376 ipv6_regen_rndid((unsigned long) ndev);
378 #endif
380 if (netif_running(dev) && netif_carrier_ok(dev))
381 ndev->if_flags |= IF_READY;
383 ipv6_mc_init_dev(ndev);
384 ndev->tstamp = jiffies;
385 #ifdef CONFIG_SYSCTL
386 neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
387 NET_IPV6_NEIGH, "ipv6",
388 &ndisc_ifinfo_sysctl_change,
389 NULL);
390 addrconf_sysctl_register(ndev, &ndev->cnf);
391 #endif
392 /* protected by rtnl_lock */
393 rcu_assign_pointer(dev->ip6_ptr, ndev);
395 /* Join all-node multicast group */
396 ipv6_addr_all_nodes(&maddr);
397 ipv6_dev_mc_inc(dev, &maddr);
399 return ndev;
402 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
404 struct inet6_dev *idev;
406 ASSERT_RTNL();
408 if ((idev = __in6_dev_get(dev)) == NULL) {
409 if ((idev = ipv6_add_dev(dev)) == NULL)
410 return NULL;
413 if (dev->flags&IFF_UP)
414 ipv6_mc_up(idev);
415 return idev;
418 #ifdef CONFIG_SYSCTL
419 static void dev_forward_change(struct inet6_dev *idev)
421 struct net_device *dev;
422 struct inet6_ifaddr *ifa;
423 struct in6_addr addr;
425 if (!idev)
426 return;
427 dev = idev->dev;
428 if (dev && (dev->flags & IFF_MULTICAST)) {
429 ipv6_addr_all_routers(&addr);
431 if (idev->cnf.forwarding)
432 ipv6_dev_mc_inc(dev, &addr);
433 else
434 ipv6_dev_mc_dec(dev, &addr);
436 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
437 if (ifa->flags&IFA_F_TENTATIVE)
438 continue;
439 if (idev->cnf.forwarding)
440 addrconf_join_anycast(ifa);
441 else
442 addrconf_leave_anycast(ifa);
447 static void addrconf_forward_change(void)
449 struct net_device *dev;
450 struct inet6_dev *idev;
452 read_lock(&dev_base_lock);
453 for_each_netdev(&init_net, dev) {
454 rcu_read_lock();
455 idev = __in6_dev_get(dev);
456 if (idev) {
457 int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
458 idev->cnf.forwarding = ipv6_devconf.forwarding;
459 if (changed)
460 dev_forward_change(idev);
462 rcu_read_unlock();
464 read_unlock(&dev_base_lock);
466 #endif
468 /* Nobody refers to this ifaddr, destroy it */
470 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
472 BUG_TRAP(ifp->if_next==NULL);
473 BUG_TRAP(ifp->lst_next==NULL);
474 #ifdef NET_REFCNT_DEBUG
475 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
476 #endif
478 in6_dev_put(ifp->idev);
480 if (del_timer(&ifp->timer))
481 printk("Timer is still running, when freeing ifa=%p\n", ifp);
483 if (!ifp->dead) {
484 printk("Freeing alive inet6 address %p\n", ifp);
485 return;
487 dst_release(&ifp->rt->u.dst);
489 kfree(ifp);
492 static void
493 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
495 struct inet6_ifaddr *ifa, **ifap;
496 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
499 * Each device address list is sorted in order of scope -
500 * global before linklocal.
502 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
503 ifap = &ifa->if_next) {
504 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
505 break;
508 ifp->if_next = *ifap;
509 *ifap = ifp;
512 /* On success it returns ifp with increased reference count */
514 static struct inet6_ifaddr *
515 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
516 int scope, u32 flags)
518 struct inet6_ifaddr *ifa = NULL;
519 struct rt6_info *rt;
520 int hash;
521 int err = 0;
523 rcu_read_lock_bh();
524 if (idev->dead) {
525 err = -ENODEV; /*XXX*/
526 goto out2;
529 write_lock(&addrconf_hash_lock);
531 /* Ignore adding duplicate addresses on an interface */
532 if (ipv6_chk_same_addr(addr, idev->dev)) {
533 ADBG(("ipv6_add_addr: already assigned\n"));
534 err = -EEXIST;
535 goto out;
538 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
540 if (ifa == NULL) {
541 ADBG(("ipv6_add_addr: malloc failed\n"));
542 err = -ENOBUFS;
543 goto out;
546 rt = addrconf_dst_alloc(idev, addr, 0);
547 if (IS_ERR(rt)) {
548 err = PTR_ERR(rt);
549 goto out;
552 ipv6_addr_copy(&ifa->addr, addr);
554 spin_lock_init(&ifa->lock);
555 init_timer(&ifa->timer);
556 ifa->timer.data = (unsigned long) ifa;
557 ifa->scope = scope;
558 ifa->prefix_len = pfxlen;
559 ifa->flags = flags | IFA_F_TENTATIVE;
560 ifa->cstamp = ifa->tstamp = jiffies;
562 ifa->rt = rt;
565 * part one of RFC 4429, section 3.3
566 * We should not configure an address as
567 * optimistic if we do not yet know the link
568 * layer address of our nexhop router
571 if (rt->rt6i_nexthop == NULL)
572 ifa->flags &= ~IFA_F_OPTIMISTIC;
574 ifa->idev = idev;
575 in6_dev_hold(idev);
576 /* For caller */
577 in6_ifa_hold(ifa);
579 /* Add to big hash table */
580 hash = ipv6_addr_hash(addr);
582 ifa->lst_next = inet6_addr_lst[hash];
583 inet6_addr_lst[hash] = ifa;
584 in6_ifa_hold(ifa);
585 write_unlock(&addrconf_hash_lock);
587 write_lock(&idev->lock);
588 /* Add to inet6_dev unicast addr list. */
589 ipv6_link_dev_addr(idev, ifa);
591 #ifdef CONFIG_IPV6_PRIVACY
592 if (ifa->flags&IFA_F_TEMPORARY) {
593 ifa->tmp_next = idev->tempaddr_list;
594 idev->tempaddr_list = ifa;
595 in6_ifa_hold(ifa);
597 #endif
599 in6_ifa_hold(ifa);
600 write_unlock(&idev->lock);
601 out2:
602 rcu_read_unlock_bh();
604 if (likely(err == 0))
605 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
606 else {
607 kfree(ifa);
608 ifa = ERR_PTR(err);
611 return ifa;
612 out:
613 write_unlock(&addrconf_hash_lock);
614 goto out2;
617 /* This function wants to get referenced ifp and releases it before return */
619 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
621 struct inet6_ifaddr *ifa, **ifap;
622 struct inet6_dev *idev = ifp->idev;
623 int hash;
624 int deleted = 0, onlink = 0;
625 unsigned long expires = jiffies;
627 hash = ipv6_addr_hash(&ifp->addr);
629 ifp->dead = 1;
631 write_lock_bh(&addrconf_hash_lock);
632 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
633 ifap = &ifa->lst_next) {
634 if (ifa == ifp) {
635 *ifap = ifa->lst_next;
636 __in6_ifa_put(ifp);
637 ifa->lst_next = NULL;
638 break;
641 write_unlock_bh(&addrconf_hash_lock);
643 write_lock_bh(&idev->lock);
644 #ifdef CONFIG_IPV6_PRIVACY
645 if (ifp->flags&IFA_F_TEMPORARY) {
646 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
647 ifap = &ifa->tmp_next) {
648 if (ifa == ifp) {
649 *ifap = ifa->tmp_next;
650 if (ifp->ifpub) {
651 in6_ifa_put(ifp->ifpub);
652 ifp->ifpub = NULL;
654 __in6_ifa_put(ifp);
655 ifa->tmp_next = NULL;
656 break;
660 #endif
662 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
663 if (ifa == ifp) {
664 *ifap = ifa->if_next;
665 __in6_ifa_put(ifp);
666 ifa->if_next = NULL;
667 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
668 break;
669 deleted = 1;
670 continue;
671 } else if (ifp->flags & IFA_F_PERMANENT) {
672 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
673 ifp->prefix_len)) {
674 if (ifa->flags & IFA_F_PERMANENT) {
675 onlink = 1;
676 if (deleted)
677 break;
678 } else {
679 unsigned long lifetime;
681 if (!onlink)
682 onlink = -1;
684 spin_lock(&ifa->lock);
685 lifetime = min_t(unsigned long,
686 ifa->valid_lft, 0x7fffffffUL/HZ);
687 if (time_before(expires,
688 ifa->tstamp + lifetime * HZ))
689 expires = ifa->tstamp + lifetime * HZ;
690 spin_unlock(&ifa->lock);
694 ifap = &ifa->if_next;
696 write_unlock_bh(&idev->lock);
698 ipv6_ifa_notify(RTM_DELADDR, ifp);
700 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
702 addrconf_del_timer(ifp);
705 * Purge or update corresponding prefix
707 * 1) we don't purge prefix here if address was not permanent.
708 * prefix is managed by its own lifetime.
709 * 2) if there're no addresses, delete prefix.
710 * 3) if there're still other permanent address(es),
711 * corresponding prefix is still permanent.
712 * 4) otherwise, update prefix lifetime to the
713 * longest valid lifetime among the corresponding
714 * addresses on the device.
715 * Note: subsequent RA will update lifetime.
717 * --yoshfuji
719 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
720 struct in6_addr prefix;
721 struct rt6_info *rt;
723 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
724 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
726 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
727 if (onlink == 0) {
728 ip6_del_rt(rt);
729 rt = NULL;
730 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
731 rt->rt6i_expires = expires;
732 rt->rt6i_flags |= RTF_EXPIRES;
735 dst_release(&rt->u.dst);
738 in6_ifa_put(ifp);
741 #ifdef CONFIG_IPV6_PRIVACY
742 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
744 struct inet6_dev *idev = ifp->idev;
745 struct in6_addr addr, *tmpaddr;
746 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
747 int tmp_plen;
748 int ret = 0;
749 int max_addresses;
750 u32 addr_flags;
752 write_lock(&idev->lock);
753 if (ift) {
754 spin_lock_bh(&ift->lock);
755 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
756 spin_unlock_bh(&ift->lock);
757 tmpaddr = &addr;
758 } else {
759 tmpaddr = NULL;
761 retry:
762 in6_dev_hold(idev);
763 if (idev->cnf.use_tempaddr <= 0) {
764 write_unlock(&idev->lock);
765 printk(KERN_INFO
766 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
767 in6_dev_put(idev);
768 ret = -1;
769 goto out;
771 spin_lock_bh(&ifp->lock);
772 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
773 idev->cnf.use_tempaddr = -1; /*XXX*/
774 spin_unlock_bh(&ifp->lock);
775 write_unlock(&idev->lock);
776 printk(KERN_WARNING
777 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
778 in6_dev_put(idev);
779 ret = -1;
780 goto out;
782 in6_ifa_hold(ifp);
783 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
784 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
785 spin_unlock_bh(&ifp->lock);
786 write_unlock(&idev->lock);
787 printk(KERN_WARNING
788 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
789 in6_ifa_put(ifp);
790 in6_dev_put(idev);
791 ret = -1;
792 goto out;
794 memcpy(&addr.s6_addr[8], idev->rndid, 8);
795 tmp_valid_lft = min_t(__u32,
796 ifp->valid_lft,
797 idev->cnf.temp_valid_lft);
798 tmp_prefered_lft = min_t(__u32,
799 ifp->prefered_lft,
800 idev->cnf.temp_prefered_lft - desync_factor / HZ);
801 tmp_plen = ifp->prefix_len;
802 max_addresses = idev->cnf.max_addresses;
803 tmp_cstamp = ifp->cstamp;
804 tmp_tstamp = ifp->tstamp;
805 spin_unlock_bh(&ifp->lock);
807 write_unlock(&idev->lock);
809 addr_flags = IFA_F_TEMPORARY;
810 /* set in addrconf_prefix_rcv() */
811 if (ifp->flags & IFA_F_OPTIMISTIC)
812 addr_flags |= IFA_F_OPTIMISTIC;
814 ift = !max_addresses ||
815 ipv6_count_addresses(idev) < max_addresses ?
816 ipv6_add_addr(idev, &addr, tmp_plen,
817 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
818 addr_flags) : NULL;
819 if (!ift || IS_ERR(ift)) {
820 in6_ifa_put(ifp);
821 in6_dev_put(idev);
822 printk(KERN_INFO
823 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
824 tmpaddr = &addr;
825 write_lock(&idev->lock);
826 goto retry;
829 spin_lock_bh(&ift->lock);
830 ift->ifpub = ifp;
831 ift->valid_lft = tmp_valid_lft;
832 ift->prefered_lft = tmp_prefered_lft;
833 ift->cstamp = tmp_cstamp;
834 ift->tstamp = tmp_tstamp;
835 spin_unlock_bh(&ift->lock);
837 addrconf_dad_start(ift, 0);
838 in6_ifa_put(ift);
839 in6_dev_put(idev);
840 out:
841 return ret;
843 #endif
846 * Choose an appropriate source address (RFC3484)
848 struct ipv6_saddr_score {
849 int addr_type;
850 unsigned int attrs;
851 int matchlen;
852 int scope;
853 unsigned int rule;
856 #define IPV6_SADDR_SCORE_LOCAL 0x0001
857 #define IPV6_SADDR_SCORE_PREFERRED 0x0004
858 #define IPV6_SADDR_SCORE_HOA 0x0008
859 #define IPV6_SADDR_SCORE_OIF 0x0010
860 #define IPV6_SADDR_SCORE_LABEL 0x0020
861 #define IPV6_SADDR_SCORE_PRIVACY 0x0040
863 static inline int ipv6_saddr_preferred(int type)
865 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
866 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
867 return 1;
868 return 0;
871 /* static matching label */
872 static inline int ipv6_saddr_label(const struct in6_addr *addr, int type)
875 * prefix (longest match) label
876 * -----------------------------
877 * ::1/128 0
878 * ::/0 1
879 * 2002::/16 2
880 * ::/96 3
881 * ::ffff:0:0/96 4
882 * fc00::/7 5
883 * 2001::/32 6
885 if (type & IPV6_ADDR_LOOPBACK)
886 return 0;
887 else if (type & IPV6_ADDR_COMPATv4)
888 return 3;
889 else if (type & IPV6_ADDR_MAPPED)
890 return 4;
891 else if (addr->s6_addr32[0] == htonl(0x20010000))
892 return 6;
893 else if (addr->s6_addr16[0] == htons(0x2002))
894 return 2;
895 else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
896 return 5;
897 return 1;
900 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
901 struct in6_addr *daddr, struct in6_addr *saddr)
903 struct ipv6_saddr_score hiscore;
904 struct inet6_ifaddr *ifa_result = NULL;
905 int daddr_type = __ipv6_addr_type(daddr);
906 int daddr_scope = __ipv6_addr_src_scope(daddr_type);
907 u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
908 struct net_device *dev;
910 memset(&hiscore, 0, sizeof(hiscore));
912 read_lock(&dev_base_lock);
913 rcu_read_lock();
915 for_each_netdev(&init_net, dev) {
916 struct inet6_dev *idev;
917 struct inet6_ifaddr *ifa;
919 /* Rule 0: Candidate Source Address (section 4)
920 * - multicast and link-local destination address,
921 * the set of candidate source address MUST only
922 * include addresses assigned to interfaces
923 * belonging to the same link as the outgoing
924 * interface.
925 * (- For site-local destination addresses, the
926 * set of candidate source addresses MUST only
927 * include addresses assigned to interfaces
928 * belonging to the same site as the outgoing
929 * interface.)
931 if ((daddr_type & IPV6_ADDR_MULTICAST ||
932 daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
933 daddr_dev && dev != daddr_dev)
934 continue;
936 idev = __in6_dev_get(dev);
937 if (!idev)
938 continue;
940 read_lock_bh(&idev->lock);
941 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
942 struct ipv6_saddr_score score;
944 score.addr_type = __ipv6_addr_type(&ifa->addr);
946 /* Rule 0:
947 * - Tentative Address (RFC2462 section 5.4)
948 * - A tentative address is not considered
949 * "assigned to an interface" in the traditional
950 * sense, unless it is also flagged as optimistic.
951 * - Candidate Source Address (section 4)
952 * - In any case, anycast addresses, multicast
953 * addresses, and the unspecified address MUST
954 * NOT be included in a candidate set.
956 if ((ifa->flags & IFA_F_TENTATIVE) &&
957 (!(ifa->flags & IFA_F_OPTIMISTIC)))
958 continue;
959 if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
960 score.addr_type & IPV6_ADDR_MULTICAST)) {
961 LIMIT_NETDEBUG(KERN_DEBUG
962 "ADDRCONF: unspecified / multicast address"
963 "assigned as unicast address on %s",
964 dev->name);
965 continue;
968 score.attrs = 0;
969 score.matchlen = 0;
970 score.scope = 0;
971 score.rule = 0;
973 if (ifa_result == NULL) {
974 /* record it if the first available entry */
975 goto record_it;
978 /* Rule 1: Prefer same address */
979 if (hiscore.rule < 1) {
980 if (ipv6_addr_equal(&ifa_result->addr, daddr))
981 hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
982 hiscore.rule++;
984 if (ipv6_addr_equal(&ifa->addr, daddr)) {
985 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
986 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
987 score.rule = 1;
988 goto record_it;
990 } else {
991 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
992 continue;
995 /* Rule 2: Prefer appropriate scope */
996 if (hiscore.rule < 2) {
997 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
998 hiscore.rule++;
1000 score.scope = __ipv6_addr_src_scope(score.addr_type);
1001 if (hiscore.scope < score.scope) {
1002 if (hiscore.scope < daddr_scope) {
1003 score.rule = 2;
1004 goto record_it;
1005 } else
1006 continue;
1007 } else if (score.scope < hiscore.scope) {
1008 if (score.scope < daddr_scope)
1009 break; /* addresses sorted by scope */
1010 else {
1011 score.rule = 2;
1012 goto record_it;
1016 /* Rule 3: Avoid deprecated and optimistic addresses */
1017 if (hiscore.rule < 3) {
1018 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1019 (((ifa_result->flags &
1020 (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0)))
1021 hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1022 hiscore.rule++;
1024 if (ipv6_saddr_preferred(score.addr_type) ||
1025 (((ifa->flags &
1026 (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0))) {
1027 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1028 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1029 score.rule = 3;
1030 goto record_it;
1032 } else {
1033 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1034 continue;
1037 /* Rule 4: Prefer home address */
1038 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
1039 if (hiscore.rule < 4) {
1040 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1041 hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1042 hiscore.rule++;
1044 if (ifa->flags & IFA_F_HOMEADDRESS) {
1045 score.attrs |= IPV6_SADDR_SCORE_HOA;
1046 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1047 score.rule = 4;
1048 goto record_it;
1050 } else {
1051 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1052 continue;
1054 #else
1055 if (hiscore.rule < 4)
1056 hiscore.rule++;
1057 #endif
1059 /* Rule 5: Prefer outgoing interface */
1060 if (hiscore.rule < 5) {
1061 if (daddr_dev == NULL ||
1062 daddr_dev == ifa_result->idev->dev)
1063 hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1064 hiscore.rule++;
1066 if (daddr_dev == NULL ||
1067 daddr_dev == ifa->idev->dev) {
1068 score.attrs |= IPV6_SADDR_SCORE_OIF;
1069 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1070 score.rule = 5;
1071 goto record_it;
1073 } else {
1074 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1075 continue;
1078 /* Rule 6: Prefer matching label */
1079 if (hiscore.rule < 6) {
1080 if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1081 hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1082 hiscore.rule++;
1084 if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1085 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1086 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1087 score.rule = 6;
1088 goto record_it;
1090 } else {
1091 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1092 continue;
1095 #ifdef CONFIG_IPV6_PRIVACY
1096 /* Rule 7: Prefer public address
1097 * Note: prefer temprary address if use_tempaddr >= 2
1099 if (hiscore.rule < 7) {
1100 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1101 (ifa_result->idev->cnf.use_tempaddr >= 2))
1102 hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1103 hiscore.rule++;
1105 if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1106 (ifa->idev->cnf.use_tempaddr >= 2)) {
1107 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1108 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1109 score.rule = 7;
1110 goto record_it;
1112 } else {
1113 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1114 continue;
1116 #else
1117 if (hiscore.rule < 7)
1118 hiscore.rule++;
1119 #endif
1120 /* Rule 8: Use longest matching prefix */
1121 if (hiscore.rule < 8) {
1122 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1123 hiscore.rule++;
1125 score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1126 if (score.matchlen > hiscore.matchlen) {
1127 score.rule = 8;
1128 goto record_it;
1130 #if 0
1131 else if (score.matchlen < hiscore.matchlen)
1132 continue;
1133 #endif
1135 /* Final Rule: choose first available one */
1136 continue;
1137 record_it:
1138 if (ifa_result)
1139 in6_ifa_put(ifa_result);
1140 in6_ifa_hold(ifa);
1141 ifa_result = ifa;
1142 hiscore = score;
1144 read_unlock_bh(&idev->lock);
1146 rcu_read_unlock();
1147 read_unlock(&dev_base_lock);
1149 if (!ifa_result)
1150 return -EADDRNOTAVAIL;
1152 ipv6_addr_copy(saddr, &ifa_result->addr);
1153 in6_ifa_put(ifa_result);
1154 return 0;
1158 int ipv6_get_saddr(struct dst_entry *dst,
1159 struct in6_addr *daddr, struct in6_addr *saddr)
1161 return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1164 EXPORT_SYMBOL(ipv6_get_saddr);
1166 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1167 unsigned char banned_flags)
1169 struct inet6_dev *idev;
1170 int err = -EADDRNOTAVAIL;
1172 rcu_read_lock();
1173 if ((idev = __in6_dev_get(dev)) != NULL) {
1174 struct inet6_ifaddr *ifp;
1176 read_lock_bh(&idev->lock);
1177 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1178 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1179 ipv6_addr_copy(addr, &ifp->addr);
1180 err = 0;
1181 break;
1184 read_unlock_bh(&idev->lock);
1186 rcu_read_unlock();
1187 return err;
1190 static int ipv6_count_addresses(struct inet6_dev *idev)
1192 int cnt = 0;
1193 struct inet6_ifaddr *ifp;
1195 read_lock_bh(&idev->lock);
1196 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1197 cnt++;
1198 read_unlock_bh(&idev->lock);
1199 return cnt;
1202 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1204 struct inet6_ifaddr * ifp;
1205 u8 hash = ipv6_addr_hash(addr);
1207 read_lock_bh(&addrconf_hash_lock);
1208 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1209 if (ipv6_addr_equal(&ifp->addr, addr) &&
1210 !(ifp->flags&IFA_F_TENTATIVE)) {
1211 if (dev == NULL || ifp->idev->dev == dev ||
1212 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1213 break;
1216 read_unlock_bh(&addrconf_hash_lock);
1217 return ifp != NULL;
1220 EXPORT_SYMBOL(ipv6_chk_addr);
1222 static
1223 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1225 struct inet6_ifaddr * ifp;
1226 u8 hash = ipv6_addr_hash(addr);
1228 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1229 if (ipv6_addr_equal(&ifp->addr, addr)) {
1230 if (dev == NULL || ifp->idev->dev == dev)
1231 break;
1234 return ifp != NULL;
1237 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1239 struct inet6_ifaddr * ifp;
1240 u8 hash = ipv6_addr_hash(addr);
1242 read_lock_bh(&addrconf_hash_lock);
1243 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1244 if (ipv6_addr_equal(&ifp->addr, addr)) {
1245 if (dev == NULL || ifp->idev->dev == dev ||
1246 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1247 in6_ifa_hold(ifp);
1248 break;
1252 read_unlock_bh(&addrconf_hash_lock);
1254 return ifp;
1257 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1259 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1260 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1261 __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1262 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1263 int sk_ipv6only = ipv6_only_sock(sk);
1264 int sk2_ipv6only = inet_v6_ipv6only(sk2);
1265 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1266 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1268 if (!sk2_rcv_saddr && !sk_ipv6only)
1269 return 1;
1271 if (addr_type2 == IPV6_ADDR_ANY &&
1272 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1273 return 1;
1275 if (addr_type == IPV6_ADDR_ANY &&
1276 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1277 return 1;
1279 if (sk2_rcv_saddr6 &&
1280 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1281 return 1;
1283 if (addr_type == IPV6_ADDR_MAPPED &&
1284 !sk2_ipv6only &&
1285 (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1286 return 1;
1288 return 0;
1291 /* Gets referenced address, destroys ifaddr */
1293 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1295 if (ifp->flags&IFA_F_PERMANENT) {
1296 spin_lock_bh(&ifp->lock);
1297 addrconf_del_timer(ifp);
1298 ifp->flags |= IFA_F_TENTATIVE;
1299 spin_unlock_bh(&ifp->lock);
1300 in6_ifa_put(ifp);
1301 #ifdef CONFIG_IPV6_PRIVACY
1302 } else if (ifp->flags&IFA_F_TEMPORARY) {
1303 struct inet6_ifaddr *ifpub;
1304 spin_lock_bh(&ifp->lock);
1305 ifpub = ifp->ifpub;
1306 if (ifpub) {
1307 in6_ifa_hold(ifpub);
1308 spin_unlock_bh(&ifp->lock);
1309 ipv6_create_tempaddr(ifpub, ifp);
1310 in6_ifa_put(ifpub);
1311 } else {
1312 spin_unlock_bh(&ifp->lock);
1314 ipv6_del_addr(ifp);
1315 #endif
1316 } else
1317 ipv6_del_addr(ifp);
1320 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1322 if (net_ratelimit())
1323 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1324 addrconf_dad_stop(ifp);
1327 /* Join to solicited addr multicast group. */
1329 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1331 struct in6_addr maddr;
1333 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1334 return;
1336 addrconf_addr_solict_mult(addr, &maddr);
1337 ipv6_dev_mc_inc(dev, &maddr);
1340 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1342 struct in6_addr maddr;
1344 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1345 return;
1347 addrconf_addr_solict_mult(addr, &maddr);
1348 __ipv6_dev_mc_dec(idev, &maddr);
1351 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1353 struct in6_addr addr;
1354 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1355 if (ipv6_addr_any(&addr))
1356 return;
1357 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1360 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1362 struct in6_addr addr;
1363 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1364 if (ipv6_addr_any(&addr))
1365 return;
1366 __ipv6_dev_ac_dec(ifp->idev, &addr);
1369 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1371 if (dev->addr_len != ETH_ALEN)
1372 return -1;
1373 memcpy(eui, dev->dev_addr, 3);
1374 memcpy(eui + 5, dev->dev_addr + 3, 3);
1377 * The zSeries OSA network cards can be shared among various
1378 * OS instances, but the OSA cards have only one MAC address.
1379 * This leads to duplicate address conflicts in conjunction
1380 * with IPv6 if more than one instance uses the same card.
1382 * The driver for these cards can deliver a unique 16-bit
1383 * identifier for each instance sharing the same card. It is
1384 * placed instead of 0xFFFE in the interface identifier. The
1385 * "u" bit of the interface identifier is not inverted in this
1386 * case. Hence the resulting interface identifier has local
1387 * scope according to RFC2373.
1389 if (dev->dev_id) {
1390 eui[3] = (dev->dev_id >> 8) & 0xFF;
1391 eui[4] = dev->dev_id & 0xFF;
1392 } else {
1393 eui[3] = 0xFF;
1394 eui[4] = 0xFE;
1395 eui[0] ^= 2;
1397 return 0;
1400 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1402 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1403 if (dev->addr_len != ARCNET_ALEN)
1404 return -1;
1405 memset(eui, 0, 7);
1406 eui[7] = *(u8*)dev->dev_addr;
1407 return 0;
1410 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1412 if (dev->addr_len != INFINIBAND_ALEN)
1413 return -1;
1414 memcpy(eui, dev->dev_addr + 12, 8);
1415 eui[0] |= 2;
1416 return 0;
1419 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1421 switch (dev->type) {
1422 case ARPHRD_ETHER:
1423 case ARPHRD_FDDI:
1424 case ARPHRD_IEEE802_TR:
1425 return addrconf_ifid_eui48(eui, dev);
1426 case ARPHRD_ARCNET:
1427 return addrconf_ifid_arcnet(eui, dev);
1428 case ARPHRD_INFINIBAND:
1429 return addrconf_ifid_infiniband(eui, dev);
1431 return -1;
1434 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1436 int err = -1;
1437 struct inet6_ifaddr *ifp;
1439 read_lock_bh(&idev->lock);
1440 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1441 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1442 memcpy(eui, ifp->addr.s6_addr+8, 8);
1443 err = 0;
1444 break;
1447 read_unlock_bh(&idev->lock);
1448 return err;
1451 #ifdef CONFIG_IPV6_PRIVACY
1452 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1453 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1455 regen:
1456 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1457 idev->rndid[0] &= ~0x02;
1460 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1461 * check if generated address is not inappropriate
1463 * - Reserved subnet anycast (RFC 2526)
1464 * 11111101 11....11 1xxxxxxx
1465 * - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1466 * 00-00-5E-FE-xx-xx-xx-xx
1467 * - value 0
1468 * - XXX: already assigned to an address on the device
1470 if (idev->rndid[0] == 0xfd &&
1471 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1472 (idev->rndid[7]&0x80))
1473 goto regen;
1474 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1475 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1476 goto regen;
1477 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1478 goto regen;
1481 return 0;
1484 static void ipv6_regen_rndid(unsigned long data)
1486 struct inet6_dev *idev = (struct inet6_dev *) data;
1487 unsigned long expires;
1489 rcu_read_lock_bh();
1490 write_lock_bh(&idev->lock);
1492 if (idev->dead)
1493 goto out;
1495 if (__ipv6_regen_rndid(idev) < 0)
1496 goto out;
1498 expires = jiffies +
1499 idev->cnf.temp_prefered_lft * HZ -
1500 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1501 if (time_before(expires, jiffies)) {
1502 printk(KERN_WARNING
1503 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1504 idev->dev->name);
1505 goto out;
1508 if (!mod_timer(&idev->regen_timer, expires))
1509 in6_dev_hold(idev);
1511 out:
1512 write_unlock_bh(&idev->lock);
1513 rcu_read_unlock_bh();
1514 in6_dev_put(idev);
1517 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1518 int ret = 0;
1520 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1521 ret = __ipv6_regen_rndid(idev);
1522 return ret;
1524 #endif
1527 * Add prefix route.
1530 static void
1531 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1532 unsigned long expires, u32 flags)
1534 struct fib6_config cfg = {
1535 .fc_table = RT6_TABLE_PREFIX,
1536 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1537 .fc_ifindex = dev->ifindex,
1538 .fc_expires = expires,
1539 .fc_dst_len = plen,
1540 .fc_flags = RTF_UP | flags,
1543 ipv6_addr_copy(&cfg.fc_dst, pfx);
1545 /* Prevent useless cloning on PtP SIT.
1546 This thing is done here expecting that the whole
1547 class of non-broadcast devices need not cloning.
1549 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1550 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1551 cfg.fc_flags |= RTF_NONEXTHOP;
1552 #endif
1554 ip6_route_add(&cfg);
1557 /* Create "default" multicast route to the interface */
1559 static void addrconf_add_mroute(struct net_device *dev)
1561 struct fib6_config cfg = {
1562 .fc_table = RT6_TABLE_LOCAL,
1563 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1564 .fc_ifindex = dev->ifindex,
1565 .fc_dst_len = 8,
1566 .fc_flags = RTF_UP,
1569 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1571 ip6_route_add(&cfg);
1574 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1575 static void sit_route_add(struct net_device *dev)
1577 struct fib6_config cfg = {
1578 .fc_table = RT6_TABLE_MAIN,
1579 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1580 .fc_ifindex = dev->ifindex,
1581 .fc_dst_len = 96,
1582 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1585 /* prefix length - 96 bits "::d.d.d.d" */
1586 ip6_route_add(&cfg);
1588 #endif
1590 static void addrconf_add_lroute(struct net_device *dev)
1592 struct in6_addr addr;
1594 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1595 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1598 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1600 struct inet6_dev *idev;
1602 ASSERT_RTNL();
1604 if ((idev = ipv6_find_idev(dev)) == NULL)
1605 return NULL;
1607 /* Add default multicast route */
1608 addrconf_add_mroute(dev);
1610 /* Add link local route */
1611 addrconf_add_lroute(dev);
1612 return idev;
1615 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1617 struct prefix_info *pinfo;
1618 __u32 valid_lft;
1619 __u32 prefered_lft;
1620 int addr_type;
1621 unsigned long rt_expires;
1622 struct inet6_dev *in6_dev;
1624 pinfo = (struct prefix_info *) opt;
1626 if (len < sizeof(struct prefix_info)) {
1627 ADBG(("addrconf: prefix option too short\n"));
1628 return;
1632 * Validation checks ([ADDRCONF], page 19)
1635 addr_type = ipv6_addr_type(&pinfo->prefix);
1637 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1638 return;
1640 valid_lft = ntohl(pinfo->valid);
1641 prefered_lft = ntohl(pinfo->prefered);
1643 if (prefered_lft > valid_lft) {
1644 if (net_ratelimit())
1645 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1646 return;
1649 in6_dev = in6_dev_get(dev);
1651 if (in6_dev == NULL) {
1652 if (net_ratelimit())
1653 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1654 return;
1658 * Two things going on here:
1659 * 1) Add routes for on-link prefixes
1660 * 2) Configure prefixes with the auto flag set
1663 /* Avoid arithmetic overflow. Really, we could
1664 save rt_expires in seconds, likely valid_lft,
1665 but it would require division in fib gc, that it
1666 not good.
1668 if (valid_lft >= 0x7FFFFFFF/HZ)
1669 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1670 else
1671 rt_expires = valid_lft * HZ;
1674 * We convert this (in jiffies) to clock_t later.
1675 * Avoid arithmetic overflow there as well.
1676 * Overflow can happen only if HZ < USER_HZ.
1678 if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1679 rt_expires = 0x7FFFFFFF / USER_HZ;
1681 if (pinfo->onlink) {
1682 struct rt6_info *rt;
1683 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1685 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1686 if (rt->rt6i_flags&RTF_EXPIRES) {
1687 if (valid_lft == 0) {
1688 ip6_del_rt(rt);
1689 rt = NULL;
1690 } else {
1691 rt->rt6i_expires = jiffies + rt_expires;
1694 } else if (valid_lft) {
1695 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1696 dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1698 if (rt)
1699 dst_release(&rt->u.dst);
1702 /* Try to figure out our local address for this prefix */
1704 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1705 struct inet6_ifaddr * ifp;
1706 struct in6_addr addr;
1707 int create = 0, update_lft = 0;
1709 if (pinfo->prefix_len == 64) {
1710 memcpy(&addr, &pinfo->prefix, 8);
1711 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1712 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1713 in6_dev_put(in6_dev);
1714 return;
1716 goto ok;
1718 if (net_ratelimit())
1719 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1720 pinfo->prefix_len);
1721 in6_dev_put(in6_dev);
1722 return;
1726 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1728 if (ifp == NULL && valid_lft) {
1729 int max_addresses = in6_dev->cnf.max_addresses;
1730 u32 addr_flags = 0;
1732 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1733 if (in6_dev->cnf.optimistic_dad &&
1734 !ipv6_devconf.forwarding)
1735 addr_flags = IFA_F_OPTIMISTIC;
1736 #endif
1738 /* Do not allow to create too much of autoconfigured
1739 * addresses; this would be too easy way to crash kernel.
1741 if (!max_addresses ||
1742 ipv6_count_addresses(in6_dev) < max_addresses)
1743 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1744 addr_type&IPV6_ADDR_SCOPE_MASK,
1745 addr_flags);
1747 if (!ifp || IS_ERR(ifp)) {
1748 in6_dev_put(in6_dev);
1749 return;
1752 update_lft = create = 1;
1753 ifp->cstamp = jiffies;
1754 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1757 if (ifp) {
1758 int flags;
1759 unsigned long now;
1760 #ifdef CONFIG_IPV6_PRIVACY
1761 struct inet6_ifaddr *ift;
1762 #endif
1763 u32 stored_lft;
1765 /* update lifetime (RFC2462 5.5.3 e) */
1766 spin_lock(&ifp->lock);
1767 now = jiffies;
1768 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1769 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1770 else
1771 stored_lft = 0;
1772 if (!update_lft && stored_lft) {
1773 if (valid_lft > MIN_VALID_LIFETIME ||
1774 valid_lft > stored_lft)
1775 update_lft = 1;
1776 else if (stored_lft <= MIN_VALID_LIFETIME) {
1777 /* valid_lft <= stored_lft is always true */
1778 /* XXX: IPsec */
1779 update_lft = 0;
1780 } else {
1781 valid_lft = MIN_VALID_LIFETIME;
1782 if (valid_lft < prefered_lft)
1783 prefered_lft = valid_lft;
1784 update_lft = 1;
1788 if (update_lft) {
1789 ifp->valid_lft = valid_lft;
1790 ifp->prefered_lft = prefered_lft;
1791 ifp->tstamp = now;
1792 flags = ifp->flags;
1793 ifp->flags &= ~IFA_F_DEPRECATED;
1794 spin_unlock(&ifp->lock);
1796 if (!(flags&IFA_F_TENTATIVE))
1797 ipv6_ifa_notify(0, ifp);
1798 } else
1799 spin_unlock(&ifp->lock);
1801 #ifdef CONFIG_IPV6_PRIVACY
1802 read_lock_bh(&in6_dev->lock);
1803 /* update all temporary addresses in the list */
1804 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1806 * When adjusting the lifetimes of an existing
1807 * temporary address, only lower the lifetimes.
1808 * Implementations must not increase the
1809 * lifetimes of an existing temporary address
1810 * when processing a Prefix Information Option.
1812 spin_lock(&ift->lock);
1813 flags = ift->flags;
1814 if (ift->valid_lft > valid_lft &&
1815 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1816 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1817 if (ift->prefered_lft > prefered_lft &&
1818 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1819 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1820 spin_unlock(&ift->lock);
1821 if (!(flags&IFA_F_TENTATIVE))
1822 ipv6_ifa_notify(0, ift);
1825 if (create && in6_dev->cnf.use_tempaddr > 0) {
1827 * When a new public address is created as described in [ADDRCONF],
1828 * also create a new temporary address.
1830 read_unlock_bh(&in6_dev->lock);
1831 ipv6_create_tempaddr(ifp, NULL);
1832 } else {
1833 read_unlock_bh(&in6_dev->lock);
1835 #endif
1836 in6_ifa_put(ifp);
1837 addrconf_verify(0);
1840 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1841 in6_dev_put(in6_dev);
1845 * Set destination address.
1846 * Special case for SIT interfaces where we create a new "virtual"
1847 * device.
1849 int addrconf_set_dstaddr(void __user *arg)
1851 struct in6_ifreq ireq;
1852 struct net_device *dev;
1853 int err = -EINVAL;
1855 rtnl_lock();
1857 err = -EFAULT;
1858 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1859 goto err_exit;
1861 dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1863 err = -ENODEV;
1864 if (dev == NULL)
1865 goto err_exit;
1867 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1868 if (dev->type == ARPHRD_SIT) {
1869 struct ifreq ifr;
1870 mm_segment_t oldfs;
1871 struct ip_tunnel_parm p;
1873 err = -EADDRNOTAVAIL;
1874 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1875 goto err_exit;
1877 memset(&p, 0, sizeof(p));
1878 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1879 p.iph.saddr = 0;
1880 p.iph.version = 4;
1881 p.iph.ihl = 5;
1882 p.iph.protocol = IPPROTO_IPV6;
1883 p.iph.ttl = 64;
1884 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1886 oldfs = get_fs(); set_fs(KERNEL_DS);
1887 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1888 set_fs(oldfs);
1890 if (err == 0) {
1891 err = -ENOBUFS;
1892 if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1893 goto err_exit;
1894 err = dev_open(dev);
1897 #endif
1899 err_exit:
1900 rtnl_unlock();
1901 return err;
1905 * Manual configuration of address on an interface
1907 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1908 __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1910 struct inet6_ifaddr *ifp;
1911 struct inet6_dev *idev;
1912 struct net_device *dev;
1913 int scope;
1914 u32 flags = RTF_EXPIRES;
1916 ASSERT_RTNL();
1918 /* check the lifetime */
1919 if (!valid_lft || prefered_lft > valid_lft)
1920 return -EINVAL;
1922 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1923 return -ENODEV;
1925 if ((idev = addrconf_add_dev(dev)) == NULL)
1926 return -ENOBUFS;
1928 scope = ipv6_addr_scope(pfx);
1930 if (valid_lft == INFINITY_LIFE_TIME) {
1931 ifa_flags |= IFA_F_PERMANENT;
1932 flags = 0;
1933 } else if (valid_lft >= 0x7FFFFFFF/HZ)
1934 valid_lft = 0x7FFFFFFF/HZ;
1936 if (prefered_lft == 0)
1937 ifa_flags |= IFA_F_DEPRECATED;
1938 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1939 (prefered_lft != INFINITY_LIFE_TIME))
1940 prefered_lft = 0x7FFFFFFF/HZ;
1942 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1944 if (!IS_ERR(ifp)) {
1945 spin_lock_bh(&ifp->lock);
1946 ifp->valid_lft = valid_lft;
1947 ifp->prefered_lft = prefered_lft;
1948 ifp->tstamp = jiffies;
1949 spin_unlock_bh(&ifp->lock);
1951 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1952 jiffies_to_clock_t(valid_lft * HZ), flags);
1954 * Note that section 3.1 of RFC 4429 indicates
1955 * that the Optimistic flag should not be set for
1956 * manually configured addresses
1958 addrconf_dad_start(ifp, 0);
1959 in6_ifa_put(ifp);
1960 addrconf_verify(0);
1961 return 0;
1964 return PTR_ERR(ifp);
1967 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1969 struct inet6_ifaddr *ifp;
1970 struct inet6_dev *idev;
1971 struct net_device *dev;
1973 if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1974 return -ENODEV;
1976 if ((idev = __in6_dev_get(dev)) == NULL)
1977 return -ENXIO;
1979 read_lock_bh(&idev->lock);
1980 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1981 if (ifp->prefix_len == plen &&
1982 ipv6_addr_equal(pfx, &ifp->addr)) {
1983 in6_ifa_hold(ifp);
1984 read_unlock_bh(&idev->lock);
1986 ipv6_del_addr(ifp);
1988 /* If the last address is deleted administratively,
1989 disable IPv6 on this interface.
1991 if (idev->addr_list == NULL)
1992 addrconf_ifdown(idev->dev, 1);
1993 return 0;
1996 read_unlock_bh(&idev->lock);
1997 return -EADDRNOTAVAIL;
2001 int addrconf_add_ifaddr(void __user *arg)
2003 struct in6_ifreq ireq;
2004 int err;
2006 if (!capable(CAP_NET_ADMIN))
2007 return -EPERM;
2009 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2010 return -EFAULT;
2012 rtnl_lock();
2013 err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2014 IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2015 rtnl_unlock();
2016 return err;
2019 int addrconf_del_ifaddr(void __user *arg)
2021 struct in6_ifreq ireq;
2022 int err;
2024 if (!capable(CAP_NET_ADMIN))
2025 return -EPERM;
2027 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2028 return -EFAULT;
2030 rtnl_lock();
2031 err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2032 rtnl_unlock();
2033 return err;
2036 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2037 static void sit_add_v4_addrs(struct inet6_dev *idev)
2039 struct inet6_ifaddr * ifp;
2040 struct in6_addr addr;
2041 struct net_device *dev;
2042 int scope;
2044 ASSERT_RTNL();
2046 memset(&addr, 0, sizeof(struct in6_addr));
2047 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2049 if (idev->dev->flags&IFF_POINTOPOINT) {
2050 addr.s6_addr32[0] = htonl(0xfe800000);
2051 scope = IFA_LINK;
2052 } else {
2053 scope = IPV6_ADDR_COMPATv4;
2056 if (addr.s6_addr32[3]) {
2057 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2058 if (!IS_ERR(ifp)) {
2059 spin_lock_bh(&ifp->lock);
2060 ifp->flags &= ~IFA_F_TENTATIVE;
2061 spin_unlock_bh(&ifp->lock);
2062 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2063 in6_ifa_put(ifp);
2065 return;
2068 for_each_netdev(&init_net, dev) {
2069 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2070 if (in_dev && (dev->flags & IFF_UP)) {
2071 struct in_ifaddr * ifa;
2073 int flag = scope;
2075 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2076 int plen;
2078 addr.s6_addr32[3] = ifa->ifa_local;
2080 if (ifa->ifa_scope == RT_SCOPE_LINK)
2081 continue;
2082 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2083 if (idev->dev->flags&IFF_POINTOPOINT)
2084 continue;
2085 flag |= IFA_HOST;
2087 if (idev->dev->flags&IFF_POINTOPOINT)
2088 plen = 64;
2089 else
2090 plen = 96;
2092 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2093 IFA_F_PERMANENT);
2094 if (!IS_ERR(ifp)) {
2095 spin_lock_bh(&ifp->lock);
2096 ifp->flags &= ~IFA_F_TENTATIVE;
2097 spin_unlock_bh(&ifp->lock);
2098 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2099 in6_ifa_put(ifp);
2105 #endif
2107 static void init_loopback(struct net_device *dev)
2109 struct inet6_dev *idev;
2110 struct inet6_ifaddr * ifp;
2112 /* ::1 */
2114 ASSERT_RTNL();
2116 if ((idev = ipv6_find_idev(dev)) == NULL) {
2117 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2118 return;
2121 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2122 if (!IS_ERR(ifp)) {
2123 spin_lock_bh(&ifp->lock);
2124 ifp->flags &= ~IFA_F_TENTATIVE;
2125 spin_unlock_bh(&ifp->lock);
2126 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2127 in6_ifa_put(ifp);
2131 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2133 struct inet6_ifaddr * ifp;
2134 u32 addr_flags = IFA_F_PERMANENT;
2136 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2137 if (idev->cnf.optimistic_dad &&
2138 !ipv6_devconf.forwarding)
2139 addr_flags |= IFA_F_OPTIMISTIC;
2140 #endif
2143 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2144 if (!IS_ERR(ifp)) {
2145 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2146 addrconf_dad_start(ifp, 0);
2147 in6_ifa_put(ifp);
2151 static void addrconf_dev_config(struct net_device *dev)
2153 struct in6_addr addr;
2154 struct inet6_dev * idev;
2156 ASSERT_RTNL();
2158 if ((dev->type != ARPHRD_ETHER) &&
2159 (dev->type != ARPHRD_FDDI) &&
2160 (dev->type != ARPHRD_IEEE802_TR) &&
2161 (dev->type != ARPHRD_ARCNET) &&
2162 (dev->type != ARPHRD_INFINIBAND)) {
2163 /* Alas, we support only Ethernet autoconfiguration. */
2164 return;
2167 idev = addrconf_add_dev(dev);
2168 if (idev == NULL)
2169 return;
2171 memset(&addr, 0, sizeof(struct in6_addr));
2172 addr.s6_addr32[0] = htonl(0xFE800000);
2174 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2175 addrconf_add_linklocal(idev, &addr);
2178 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2179 static void addrconf_sit_config(struct net_device *dev)
2181 struct inet6_dev *idev;
2183 ASSERT_RTNL();
2186 * Configure the tunnel with one of our IPv4
2187 * addresses... we should configure all of
2188 * our v4 addrs in the tunnel
2191 if ((idev = ipv6_find_idev(dev)) == NULL) {
2192 printk(KERN_DEBUG "init sit: add_dev failed\n");
2193 return;
2196 sit_add_v4_addrs(idev);
2198 if (dev->flags&IFF_POINTOPOINT) {
2199 addrconf_add_mroute(dev);
2200 addrconf_add_lroute(dev);
2201 } else
2202 sit_route_add(dev);
2204 #endif
2206 static inline int
2207 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2209 struct in6_addr lladdr;
2211 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2212 addrconf_add_linklocal(idev, &lladdr);
2213 return 0;
2215 return -1;
2218 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2220 struct net_device *link_dev;
2222 /* first try to inherit the link-local address from the link device */
2223 if (idev->dev->iflink &&
2224 (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2225 if (!ipv6_inherit_linklocal(idev, link_dev))
2226 return;
2228 /* then try to inherit it from any device */
2229 for_each_netdev(&init_net, link_dev) {
2230 if (!ipv6_inherit_linklocal(idev, link_dev))
2231 return;
2233 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2237 * Autoconfigure tunnel with a link-local address so routing protocols,
2238 * DHCPv6, MLD etc. can be run over the virtual link
2241 static void addrconf_ip6_tnl_config(struct net_device *dev)
2243 struct inet6_dev *idev;
2245 ASSERT_RTNL();
2247 if ((idev = addrconf_add_dev(dev)) == NULL) {
2248 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2249 return;
2251 ip6_tnl_add_linklocal(idev);
2254 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2255 void * data)
2257 struct net_device *dev = (struct net_device *) data;
2258 struct inet6_dev *idev = __in6_dev_get(dev);
2259 int run_pending = 0;
2260 int err;
2262 if (dev->nd_net != &init_net)
2263 return NOTIFY_DONE;
2265 switch(event) {
2266 case NETDEV_REGISTER:
2267 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2268 idev = ipv6_add_dev(dev);
2269 if (!idev)
2270 return notifier_from_errno(-ENOMEM);
2272 break;
2273 case NETDEV_UP:
2274 case NETDEV_CHANGE:
2275 if (dev->flags & IFF_SLAVE)
2276 break;
2278 if (event == NETDEV_UP) {
2279 if (!netif_carrier_ok(dev)) {
2280 /* device is not ready yet. */
2281 printk(KERN_INFO
2282 "ADDRCONF(NETDEV_UP): %s: "
2283 "link is not ready\n",
2284 dev->name);
2285 break;
2288 if (idev)
2289 idev->if_flags |= IF_READY;
2290 } else {
2291 if (!netif_carrier_ok(dev)) {
2292 /* device is still not ready. */
2293 break;
2296 if (idev) {
2297 if (idev->if_flags & IF_READY) {
2298 /* device is already configured. */
2299 break;
2301 idev->if_flags |= IF_READY;
2304 printk(KERN_INFO
2305 "ADDRCONF(NETDEV_CHANGE): %s: "
2306 "link becomes ready\n",
2307 dev->name);
2309 run_pending = 1;
2312 switch(dev->type) {
2313 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2314 case ARPHRD_SIT:
2315 addrconf_sit_config(dev);
2316 break;
2317 #endif
2318 case ARPHRD_TUNNEL6:
2319 addrconf_ip6_tnl_config(dev);
2320 break;
2321 case ARPHRD_LOOPBACK:
2322 init_loopback(dev);
2323 break;
2325 default:
2326 addrconf_dev_config(dev);
2327 break;
2329 if (idev) {
2330 if (run_pending)
2331 addrconf_dad_run(idev);
2333 /* If the MTU changed during the interface down, when the
2334 interface up, the changed MTU must be reflected in the
2335 idev as well as routers.
2337 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2338 rt6_mtu_change(dev, dev->mtu);
2339 idev->cnf.mtu6 = dev->mtu;
2341 idev->tstamp = jiffies;
2342 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2343 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2344 stop IPv6 on this interface.
2346 if (dev->mtu < IPV6_MIN_MTU)
2347 addrconf_ifdown(dev, event != NETDEV_DOWN);
2349 break;
2351 case NETDEV_CHANGEMTU:
2352 if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2353 rt6_mtu_change(dev, dev->mtu);
2354 idev->cnf.mtu6 = dev->mtu;
2355 break;
2358 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2360 case NETDEV_DOWN:
2361 case NETDEV_UNREGISTER:
2363 * Remove all addresses from this interface.
2365 addrconf_ifdown(dev, event != NETDEV_DOWN);
2366 break;
2368 case NETDEV_CHANGENAME:
2369 if (idev) {
2370 snmp6_unregister_dev(idev);
2371 #ifdef CONFIG_SYSCTL
2372 addrconf_sysctl_unregister(&idev->cnf);
2373 neigh_sysctl_unregister(idev->nd_parms);
2374 neigh_sysctl_register(dev, idev->nd_parms,
2375 NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2376 &ndisc_ifinfo_sysctl_change,
2377 NULL);
2378 addrconf_sysctl_register(idev, &idev->cnf);
2379 #endif
2380 err = snmp6_register_dev(idev);
2381 if (err)
2382 return notifier_from_errno(err);
2384 break;
2387 return NOTIFY_OK;
2391 * addrconf module should be notified of a device going up
2393 static struct notifier_block ipv6_dev_notf = {
2394 .notifier_call = addrconf_notify,
2395 .priority = 0
2398 static int addrconf_ifdown(struct net_device *dev, int how)
2400 struct inet6_dev *idev;
2401 struct inet6_ifaddr *ifa, **bifa;
2402 int i;
2404 ASSERT_RTNL();
2406 if (dev == &loopback_dev && how == 1)
2407 how = 0;
2409 rt6_ifdown(dev);
2410 neigh_ifdown(&nd_tbl, dev);
2412 idev = __in6_dev_get(dev);
2413 if (idev == NULL)
2414 return -ENODEV;
2416 /* Step 1: remove reference to ipv6 device from parent device.
2417 Do not dev_put!
2419 if (how == 1) {
2420 idev->dead = 1;
2422 /* protected by rtnl_lock */
2423 rcu_assign_pointer(dev->ip6_ptr, NULL);
2425 /* Step 1.5: remove snmp6 entry */
2426 snmp6_unregister_dev(idev);
2430 /* Step 2: clear hash table */
2431 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2432 bifa = &inet6_addr_lst[i];
2434 write_lock_bh(&addrconf_hash_lock);
2435 while ((ifa = *bifa) != NULL) {
2436 if (ifa->idev == idev) {
2437 *bifa = ifa->lst_next;
2438 ifa->lst_next = NULL;
2439 addrconf_del_timer(ifa);
2440 in6_ifa_put(ifa);
2441 continue;
2443 bifa = &ifa->lst_next;
2445 write_unlock_bh(&addrconf_hash_lock);
2448 write_lock_bh(&idev->lock);
2450 /* Step 3: clear flags for stateless addrconf */
2451 if (how != 1)
2452 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2454 /* Step 4: clear address list */
2455 #ifdef CONFIG_IPV6_PRIVACY
2456 if (how == 1 && del_timer(&idev->regen_timer))
2457 in6_dev_put(idev);
2459 /* clear tempaddr list */
2460 while ((ifa = idev->tempaddr_list) != NULL) {
2461 idev->tempaddr_list = ifa->tmp_next;
2462 ifa->tmp_next = NULL;
2463 ifa->dead = 1;
2464 write_unlock_bh(&idev->lock);
2465 spin_lock_bh(&ifa->lock);
2467 if (ifa->ifpub) {
2468 in6_ifa_put(ifa->ifpub);
2469 ifa->ifpub = NULL;
2471 spin_unlock_bh(&ifa->lock);
2472 in6_ifa_put(ifa);
2473 write_lock_bh(&idev->lock);
2475 #endif
2476 while ((ifa = idev->addr_list) != NULL) {
2477 idev->addr_list = ifa->if_next;
2478 ifa->if_next = NULL;
2479 ifa->dead = 1;
2480 addrconf_del_timer(ifa);
2481 write_unlock_bh(&idev->lock);
2483 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2484 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2485 in6_ifa_put(ifa);
2487 write_lock_bh(&idev->lock);
2489 write_unlock_bh(&idev->lock);
2491 /* Step 5: Discard multicast list */
2493 if (how == 1)
2494 ipv6_mc_destroy_dev(idev);
2495 else
2496 ipv6_mc_down(idev);
2498 /* Step 5: netlink notification of this interface */
2499 idev->tstamp = jiffies;
2500 inet6_ifinfo_notify(RTM_DELLINK, idev);
2502 /* Shot the device (if unregistered) */
2504 if (how == 1) {
2505 #ifdef CONFIG_SYSCTL
2506 addrconf_sysctl_unregister(&idev->cnf);
2507 neigh_sysctl_unregister(idev->nd_parms);
2508 #endif
2509 neigh_parms_release(&nd_tbl, idev->nd_parms);
2510 neigh_ifdown(&nd_tbl, dev);
2511 in6_dev_put(idev);
2513 return 0;
2516 static void addrconf_rs_timer(unsigned long data)
2518 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2520 if (ifp->idev->cnf.forwarding)
2521 goto out;
2523 if (ifp->idev->if_flags & IF_RA_RCVD) {
2525 * Announcement received after solicitation
2526 * was sent
2528 goto out;
2531 spin_lock(&ifp->lock);
2532 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2533 struct in6_addr all_routers;
2535 /* The wait after the last probe can be shorter */
2536 addrconf_mod_timer(ifp, AC_RS,
2537 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2538 ifp->idev->cnf.rtr_solicit_delay :
2539 ifp->idev->cnf.rtr_solicit_interval);
2540 spin_unlock(&ifp->lock);
2542 ipv6_addr_all_routers(&all_routers);
2544 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2545 } else {
2546 spin_unlock(&ifp->lock);
2548 * Note: we do not support deprecated "all on-link"
2549 * assumption any longer.
2551 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2552 ifp->idev->dev->name);
2555 out:
2556 in6_ifa_put(ifp);
2560 * Duplicate Address Detection
2562 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2564 unsigned long rand_num;
2565 struct inet6_dev *idev = ifp->idev;
2567 if (ifp->flags & IFA_F_OPTIMISTIC)
2568 rand_num = 0;
2569 else
2570 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2572 ifp->probes = idev->cnf.dad_transmits;
2573 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2576 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2578 struct inet6_dev *idev = ifp->idev;
2579 struct net_device *dev = idev->dev;
2581 addrconf_join_solict(dev, &ifp->addr);
2583 net_srandom(ifp->addr.s6_addr32[3]);
2585 read_lock_bh(&idev->lock);
2586 if (ifp->dead)
2587 goto out;
2588 spin_lock_bh(&ifp->lock);
2590 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2591 !(ifp->flags&IFA_F_TENTATIVE) ||
2592 ifp->flags & IFA_F_NODAD) {
2593 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2594 spin_unlock_bh(&ifp->lock);
2595 read_unlock_bh(&idev->lock);
2597 addrconf_dad_completed(ifp);
2598 return;
2601 if (!(idev->if_flags & IF_READY)) {
2602 spin_unlock_bh(&ifp->lock);
2603 read_unlock_bh(&idev->lock);
2605 * If the defice is not ready:
2606 * - keep it tentative if it is a permanent address.
2607 * - otherwise, kill it.
2609 in6_ifa_hold(ifp);
2610 addrconf_dad_stop(ifp);
2611 return;
2615 * Optimistic nodes can start receiving
2616 * Frames right away
2618 if(ifp->flags & IFA_F_OPTIMISTIC)
2619 ip6_ins_rt(ifp->rt);
2621 addrconf_dad_kick(ifp);
2622 spin_unlock_bh(&ifp->lock);
2623 out:
2624 read_unlock_bh(&idev->lock);
2627 static void addrconf_dad_timer(unsigned long data)
2629 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2630 struct inet6_dev *idev = ifp->idev;
2631 struct in6_addr unspec;
2632 struct in6_addr mcaddr;
2634 read_lock_bh(&idev->lock);
2635 if (idev->dead) {
2636 read_unlock_bh(&idev->lock);
2637 goto out;
2639 spin_lock_bh(&ifp->lock);
2640 if (ifp->probes == 0) {
2642 * DAD was successful
2645 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2646 spin_unlock_bh(&ifp->lock);
2647 read_unlock_bh(&idev->lock);
2649 addrconf_dad_completed(ifp);
2651 goto out;
2654 ifp->probes--;
2655 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2656 spin_unlock_bh(&ifp->lock);
2657 read_unlock_bh(&idev->lock);
2659 /* send a neighbour solicitation for our addr */
2660 memset(&unspec, 0, sizeof(unspec));
2661 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2662 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2663 out:
2664 in6_ifa_put(ifp);
2667 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2669 struct net_device * dev = ifp->idev->dev;
2672 * Configure the address for reception. Now it is valid.
2675 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2677 /* If added prefix is link local and forwarding is off,
2678 start sending router solicitations.
2681 if (ifp->idev->cnf.forwarding == 0 &&
2682 ifp->idev->cnf.rtr_solicits > 0 &&
2683 (dev->flags&IFF_LOOPBACK) == 0 &&
2684 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2685 struct in6_addr all_routers;
2687 ipv6_addr_all_routers(&all_routers);
2690 * If a host as already performed a random delay
2691 * [...] as part of DAD [...] there is no need
2692 * to delay again before sending the first RS
2694 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2696 spin_lock_bh(&ifp->lock);
2697 ifp->probes = 1;
2698 ifp->idev->if_flags |= IF_RS_SENT;
2699 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2700 spin_unlock_bh(&ifp->lock);
2704 static void addrconf_dad_run(struct inet6_dev *idev) {
2705 struct inet6_ifaddr *ifp;
2707 read_lock_bh(&idev->lock);
2708 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2709 spin_lock_bh(&ifp->lock);
2710 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2711 spin_unlock_bh(&ifp->lock);
2712 continue;
2714 spin_unlock_bh(&ifp->lock);
2715 addrconf_dad_kick(ifp);
2717 read_unlock_bh(&idev->lock);
2720 #ifdef CONFIG_PROC_FS
2721 struct if6_iter_state {
2722 int bucket;
2725 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2727 struct inet6_ifaddr *ifa = NULL;
2728 struct if6_iter_state *state = seq->private;
2730 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2731 ifa = inet6_addr_lst[state->bucket];
2732 if (ifa)
2733 break;
2735 return ifa;
2738 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2740 struct if6_iter_state *state = seq->private;
2742 ifa = ifa->lst_next;
2743 try_again:
2744 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2745 ifa = inet6_addr_lst[state->bucket];
2746 goto try_again;
2748 return ifa;
2751 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2753 struct inet6_ifaddr *ifa = if6_get_first(seq);
2755 if (ifa)
2756 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2757 --pos;
2758 return pos ? NULL : ifa;
2761 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2763 read_lock_bh(&addrconf_hash_lock);
2764 return if6_get_idx(seq, *pos);
2767 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2769 struct inet6_ifaddr *ifa;
2771 ifa = if6_get_next(seq, v);
2772 ++*pos;
2773 return ifa;
2776 static void if6_seq_stop(struct seq_file *seq, void *v)
2778 read_unlock_bh(&addrconf_hash_lock);
2781 static int if6_seq_show(struct seq_file *seq, void *v)
2783 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2784 seq_printf(seq,
2785 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2786 NIP6(ifp->addr),
2787 ifp->idev->dev->ifindex,
2788 ifp->prefix_len,
2789 ifp->scope,
2790 ifp->flags,
2791 ifp->idev->dev->name);
2792 return 0;
2795 static const struct seq_operations if6_seq_ops = {
2796 .start = if6_seq_start,
2797 .next = if6_seq_next,
2798 .show = if6_seq_show,
2799 .stop = if6_seq_stop,
2802 static int if6_seq_open(struct inode *inode, struct file *file)
2804 struct seq_file *seq;
2805 int rc = -ENOMEM;
2806 struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2808 if (!s)
2809 goto out;
2811 rc = seq_open(file, &if6_seq_ops);
2812 if (rc)
2813 goto out_kfree;
2815 seq = file->private_data;
2816 seq->private = s;
2817 out:
2818 return rc;
2819 out_kfree:
2820 kfree(s);
2821 goto out;
2824 static const struct file_operations if6_fops = {
2825 .owner = THIS_MODULE,
2826 .open = if6_seq_open,
2827 .read = seq_read,
2828 .llseek = seq_lseek,
2829 .release = seq_release_private,
2832 int __init if6_proc_init(void)
2834 if (!proc_net_fops_create(&init_net, "if_inet6", S_IRUGO, &if6_fops))
2835 return -ENOMEM;
2836 return 0;
2839 void if6_proc_exit(void)
2841 proc_net_remove(&init_net, "if_inet6");
2843 #endif /* CONFIG_PROC_FS */
2845 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2846 /* Check if address is a home address configured on any interface. */
2847 int ipv6_chk_home_addr(struct in6_addr *addr)
2849 int ret = 0;
2850 struct inet6_ifaddr * ifp;
2851 u8 hash = ipv6_addr_hash(addr);
2852 read_lock_bh(&addrconf_hash_lock);
2853 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2854 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2855 (ifp->flags & IFA_F_HOMEADDRESS)) {
2856 ret = 1;
2857 break;
2860 read_unlock_bh(&addrconf_hash_lock);
2861 return ret;
2863 #endif
2866 * Periodic address status verification
2869 static void addrconf_verify(unsigned long foo)
2871 struct inet6_ifaddr *ifp;
2872 unsigned long now, next;
2873 int i;
2875 spin_lock_bh(&addrconf_verify_lock);
2876 now = jiffies;
2877 next = now + ADDR_CHECK_FREQUENCY;
2879 del_timer(&addr_chk_timer);
2881 for (i=0; i < IN6_ADDR_HSIZE; i++) {
2883 restart:
2884 read_lock(&addrconf_hash_lock);
2885 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2886 unsigned long age;
2887 #ifdef CONFIG_IPV6_PRIVACY
2888 unsigned long regen_advance;
2889 #endif
2891 if (ifp->flags & IFA_F_PERMANENT)
2892 continue;
2894 spin_lock(&ifp->lock);
2895 age = (now - ifp->tstamp) / HZ;
2897 #ifdef CONFIG_IPV6_PRIVACY
2898 regen_advance = ifp->idev->cnf.regen_max_retry *
2899 ifp->idev->cnf.dad_transmits *
2900 ifp->idev->nd_parms->retrans_time / HZ;
2901 #endif
2903 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2904 age >= ifp->valid_lft) {
2905 spin_unlock(&ifp->lock);
2906 in6_ifa_hold(ifp);
2907 read_unlock(&addrconf_hash_lock);
2908 ipv6_del_addr(ifp);
2909 goto restart;
2910 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2911 spin_unlock(&ifp->lock);
2912 continue;
2913 } else if (age >= ifp->prefered_lft) {
2914 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2915 int deprecate = 0;
2917 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2918 deprecate = 1;
2919 ifp->flags |= IFA_F_DEPRECATED;
2922 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2923 next = ifp->tstamp + ifp->valid_lft * HZ;
2925 spin_unlock(&ifp->lock);
2927 if (deprecate) {
2928 in6_ifa_hold(ifp);
2929 read_unlock(&addrconf_hash_lock);
2931 ipv6_ifa_notify(0, ifp);
2932 in6_ifa_put(ifp);
2933 goto restart;
2935 #ifdef CONFIG_IPV6_PRIVACY
2936 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2937 !(ifp->flags&IFA_F_TENTATIVE)) {
2938 if (age >= ifp->prefered_lft - regen_advance) {
2939 struct inet6_ifaddr *ifpub = ifp->ifpub;
2940 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2941 next = ifp->tstamp + ifp->prefered_lft * HZ;
2942 if (!ifp->regen_count && ifpub) {
2943 ifp->regen_count++;
2944 in6_ifa_hold(ifp);
2945 in6_ifa_hold(ifpub);
2946 spin_unlock(&ifp->lock);
2947 read_unlock(&addrconf_hash_lock);
2948 spin_lock(&ifpub->lock);
2949 ifpub->regen_count = 0;
2950 spin_unlock(&ifpub->lock);
2951 ipv6_create_tempaddr(ifpub, ifp);
2952 in6_ifa_put(ifpub);
2953 in6_ifa_put(ifp);
2954 goto restart;
2956 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2957 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2958 spin_unlock(&ifp->lock);
2959 #endif
2960 } else {
2961 /* ifp->prefered_lft <= ifp->valid_lft */
2962 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2963 next = ifp->tstamp + ifp->prefered_lft * HZ;
2964 spin_unlock(&ifp->lock);
2967 read_unlock(&addrconf_hash_lock);
2970 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2971 add_timer(&addr_chk_timer);
2972 spin_unlock_bh(&addrconf_verify_lock);
2975 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2977 struct in6_addr *pfx = NULL;
2979 if (addr)
2980 pfx = nla_data(addr);
2982 if (local) {
2983 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2984 pfx = NULL;
2985 else
2986 pfx = nla_data(local);
2989 return pfx;
2992 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
2993 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
2994 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
2995 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
2998 static int
2999 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3001 struct ifaddrmsg *ifm;
3002 struct nlattr *tb[IFA_MAX+1];
3003 struct in6_addr *pfx;
3004 int err;
3006 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3007 if (err < 0)
3008 return err;
3010 ifm = nlmsg_data(nlh);
3011 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3012 if (pfx == NULL)
3013 return -EINVAL;
3015 return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3018 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3019 u32 prefered_lft, u32 valid_lft)
3021 u32 flags = RTF_EXPIRES;
3023 if (!valid_lft || (prefered_lft > valid_lft))
3024 return -EINVAL;
3026 if (valid_lft == INFINITY_LIFE_TIME) {
3027 ifa_flags |= IFA_F_PERMANENT;
3028 flags = 0;
3029 } else if (valid_lft >= 0x7FFFFFFF/HZ)
3030 valid_lft = 0x7FFFFFFF/HZ;
3032 if (prefered_lft == 0)
3033 ifa_flags |= IFA_F_DEPRECATED;
3034 else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3035 (prefered_lft != INFINITY_LIFE_TIME))
3036 prefered_lft = 0x7FFFFFFF/HZ;
3038 spin_lock_bh(&ifp->lock);
3039 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3040 ifp->tstamp = jiffies;
3041 ifp->valid_lft = valid_lft;
3042 ifp->prefered_lft = prefered_lft;
3044 spin_unlock_bh(&ifp->lock);
3045 if (!(ifp->flags&IFA_F_TENTATIVE))
3046 ipv6_ifa_notify(0, ifp);
3048 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3049 jiffies_to_clock_t(valid_lft * HZ), flags);
3050 addrconf_verify(0);
3052 return 0;
3055 static int
3056 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3058 struct ifaddrmsg *ifm;
3059 struct nlattr *tb[IFA_MAX+1];
3060 struct in6_addr *pfx;
3061 struct inet6_ifaddr *ifa;
3062 struct net_device *dev;
3063 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3064 u8 ifa_flags;
3065 int err;
3067 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3068 if (err < 0)
3069 return err;
3071 ifm = nlmsg_data(nlh);
3072 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3073 if (pfx == NULL)
3074 return -EINVAL;
3076 if (tb[IFA_CACHEINFO]) {
3077 struct ifa_cacheinfo *ci;
3079 ci = nla_data(tb[IFA_CACHEINFO]);
3080 valid_lft = ci->ifa_valid;
3081 preferred_lft = ci->ifa_prefered;
3082 } else {
3083 preferred_lft = INFINITY_LIFE_TIME;
3084 valid_lft = INFINITY_LIFE_TIME;
3087 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3088 if (dev == NULL)
3089 return -ENODEV;
3091 /* We ignore other flags so far. */
3092 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3094 ifa = ipv6_get_ifaddr(pfx, dev, 1);
3095 if (ifa == NULL) {
3097 * It would be best to check for !NLM_F_CREATE here but
3098 * userspace alreay relies on not having to provide this.
3100 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3101 ifa_flags, preferred_lft, valid_lft);
3104 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3105 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3106 err = -EEXIST;
3107 else
3108 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3110 in6_ifa_put(ifa);
3112 return err;
3115 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3116 u8 scope, int ifindex)
3118 struct ifaddrmsg *ifm;
3120 ifm = nlmsg_data(nlh);
3121 ifm->ifa_family = AF_INET6;
3122 ifm->ifa_prefixlen = prefixlen;
3123 ifm->ifa_flags = flags;
3124 ifm->ifa_scope = scope;
3125 ifm->ifa_index = ifindex;
3128 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3129 unsigned long tstamp, u32 preferred, u32 valid)
3131 struct ifa_cacheinfo ci;
3133 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3134 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3135 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3136 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3137 ci.ifa_prefered = preferred;
3138 ci.ifa_valid = valid;
3140 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3143 static inline int rt_scope(int ifa_scope)
3145 if (ifa_scope & IFA_HOST)
3146 return RT_SCOPE_HOST;
3147 else if (ifa_scope & IFA_LINK)
3148 return RT_SCOPE_LINK;
3149 else if (ifa_scope & IFA_SITE)
3150 return RT_SCOPE_SITE;
3151 else
3152 return RT_SCOPE_UNIVERSE;
3155 static inline int inet6_ifaddr_msgsize(void)
3157 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3158 + nla_total_size(16) /* IFA_ADDRESS */
3159 + nla_total_size(sizeof(struct ifa_cacheinfo));
3162 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3163 u32 pid, u32 seq, int event, unsigned int flags)
3165 struct nlmsghdr *nlh;
3166 u32 preferred, valid;
3168 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3169 if (nlh == NULL)
3170 return -EMSGSIZE;
3172 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3173 ifa->idev->dev->ifindex);
3175 if (!(ifa->flags&IFA_F_PERMANENT)) {
3176 preferred = ifa->prefered_lft;
3177 valid = ifa->valid_lft;
3178 if (preferred != INFINITY_LIFE_TIME) {
3179 long tval = (jiffies - ifa->tstamp)/HZ;
3180 preferred -= tval;
3181 if (valid != INFINITY_LIFE_TIME)
3182 valid -= tval;
3184 } else {
3185 preferred = INFINITY_LIFE_TIME;
3186 valid = INFINITY_LIFE_TIME;
3189 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3190 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3191 nlmsg_cancel(skb, nlh);
3192 return -EMSGSIZE;
3195 return nlmsg_end(skb, nlh);
3198 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3199 u32 pid, u32 seq, int event, u16 flags)
3201 struct nlmsghdr *nlh;
3202 u8 scope = RT_SCOPE_UNIVERSE;
3203 int ifindex = ifmca->idev->dev->ifindex;
3205 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3206 scope = RT_SCOPE_SITE;
3208 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3209 if (nlh == NULL)
3210 return -EMSGSIZE;
3212 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3213 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3214 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3215 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3216 nlmsg_cancel(skb, nlh);
3217 return -EMSGSIZE;
3220 return nlmsg_end(skb, nlh);
3223 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3224 u32 pid, u32 seq, int event, unsigned int flags)
3226 struct nlmsghdr *nlh;
3227 u8 scope = RT_SCOPE_UNIVERSE;
3228 int ifindex = ifaca->aca_idev->dev->ifindex;
3230 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3231 scope = RT_SCOPE_SITE;
3233 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3234 if (nlh == NULL)
3235 return -EMSGSIZE;
3237 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3238 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3239 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3240 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3241 nlmsg_cancel(skb, nlh);
3242 return -EMSGSIZE;
3245 return nlmsg_end(skb, nlh);
3248 enum addr_type_t
3250 UNICAST_ADDR,
3251 MULTICAST_ADDR,
3252 ANYCAST_ADDR,
3255 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3256 enum addr_type_t type)
3258 int idx, ip_idx;
3259 int s_idx, s_ip_idx;
3260 int err = 1;
3261 struct net_device *dev;
3262 struct inet6_dev *idev = NULL;
3263 struct inet6_ifaddr *ifa;
3264 struct ifmcaddr6 *ifmca;
3265 struct ifacaddr6 *ifaca;
3267 s_idx = cb->args[0];
3268 s_ip_idx = ip_idx = cb->args[1];
3270 idx = 0;
3271 for_each_netdev(&init_net, dev) {
3272 if (idx < s_idx)
3273 goto cont;
3274 if (idx > s_idx)
3275 s_ip_idx = 0;
3276 ip_idx = 0;
3277 if ((idev = in6_dev_get(dev)) == NULL)
3278 goto cont;
3279 read_lock_bh(&idev->lock);
3280 switch (type) {
3281 case UNICAST_ADDR:
3282 /* unicast address incl. temp addr */
3283 for (ifa = idev->addr_list; ifa;
3284 ifa = ifa->if_next, ip_idx++) {
3285 if (ip_idx < s_ip_idx)
3286 continue;
3287 if ((err = inet6_fill_ifaddr(skb, ifa,
3288 NETLINK_CB(cb->skb).pid,
3289 cb->nlh->nlmsg_seq, RTM_NEWADDR,
3290 NLM_F_MULTI)) <= 0)
3291 goto done;
3293 break;
3294 case MULTICAST_ADDR:
3295 /* multicast address */
3296 for (ifmca = idev->mc_list; ifmca;
3297 ifmca = ifmca->next, ip_idx++) {
3298 if (ip_idx < s_ip_idx)
3299 continue;
3300 if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3301 NETLINK_CB(cb->skb).pid,
3302 cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3303 NLM_F_MULTI)) <= 0)
3304 goto done;
3306 break;
3307 case ANYCAST_ADDR:
3308 /* anycast address */
3309 for (ifaca = idev->ac_list; ifaca;
3310 ifaca = ifaca->aca_next, ip_idx++) {
3311 if (ip_idx < s_ip_idx)
3312 continue;
3313 if ((err = inet6_fill_ifacaddr(skb, ifaca,
3314 NETLINK_CB(cb->skb).pid,
3315 cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3316 NLM_F_MULTI)) <= 0)
3317 goto done;
3319 break;
3320 default:
3321 break;
3323 read_unlock_bh(&idev->lock);
3324 in6_dev_put(idev);
3325 cont:
3326 idx++;
3328 done:
3329 if (err <= 0) {
3330 read_unlock_bh(&idev->lock);
3331 in6_dev_put(idev);
3333 cb->args[0] = idx;
3334 cb->args[1] = ip_idx;
3335 return skb->len;
3338 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3340 enum addr_type_t type = UNICAST_ADDR;
3341 return inet6_dump_addr(skb, cb, type);
3344 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3346 enum addr_type_t type = MULTICAST_ADDR;
3347 return inet6_dump_addr(skb, cb, type);
3351 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3353 enum addr_type_t type = ANYCAST_ADDR;
3354 return inet6_dump_addr(skb, cb, type);
3357 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3358 void *arg)
3360 struct ifaddrmsg *ifm;
3361 struct nlattr *tb[IFA_MAX+1];
3362 struct in6_addr *addr = NULL;
3363 struct net_device *dev = NULL;
3364 struct inet6_ifaddr *ifa;
3365 struct sk_buff *skb;
3366 int err;
3368 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3369 if (err < 0)
3370 goto errout;
3372 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3373 if (addr == NULL) {
3374 err = -EINVAL;
3375 goto errout;
3378 ifm = nlmsg_data(nlh);
3379 if (ifm->ifa_index)
3380 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3382 if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3383 err = -EADDRNOTAVAIL;
3384 goto errout;
3387 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3388 err = -ENOBUFS;
3389 goto errout_ifa;
3392 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3393 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3394 if (err < 0) {
3395 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3396 WARN_ON(err == -EMSGSIZE);
3397 kfree_skb(skb);
3398 goto errout_ifa;
3400 err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3401 errout_ifa:
3402 in6_ifa_put(ifa);
3403 errout:
3404 return err;
3407 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3409 struct sk_buff *skb;
3410 int err = -ENOBUFS;
3412 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3413 if (skb == NULL)
3414 goto errout;
3416 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3417 if (err < 0) {
3418 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3419 WARN_ON(err == -EMSGSIZE);
3420 kfree_skb(skb);
3421 goto errout;
3423 err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3424 errout:
3425 if (err < 0)
3426 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3429 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3430 __s32 *array, int bytes)
3432 BUG_ON(bytes < (DEVCONF_MAX * 4));
3434 memset(array, 0, bytes);
3435 array[DEVCONF_FORWARDING] = cnf->forwarding;
3436 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3437 array[DEVCONF_MTU6] = cnf->mtu6;
3438 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3439 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3440 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3441 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3442 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3443 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3444 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3445 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3446 #ifdef CONFIG_IPV6_PRIVACY
3447 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3448 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3449 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3450 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3451 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3452 #endif
3453 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3454 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3455 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3456 #ifdef CONFIG_IPV6_ROUTER_PREF
3457 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3458 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3459 #ifdef CONFIG_IPV6_ROUTE_INFO
3460 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3461 #endif
3462 #endif
3463 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3464 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3465 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3466 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3467 #endif
3470 static inline size_t inet6_if_nlmsg_size(void)
3472 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3473 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3474 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3475 + nla_total_size(4) /* IFLA_MTU */
3476 + nla_total_size(4) /* IFLA_LINK */
3477 + nla_total_size( /* IFLA_PROTINFO */
3478 nla_total_size(4) /* IFLA_INET6_FLAGS */
3479 + nla_total_size(sizeof(struct ifla_cacheinfo))
3480 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3481 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3482 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3486 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3487 int bytes)
3489 int i;
3490 int pad = bytes - sizeof(u64) * items;
3491 BUG_ON(pad < 0);
3493 /* Use put_unaligned() because stats may not be aligned for u64. */
3494 put_unaligned(items, &stats[0]);
3495 for (i = 1; i < items; i++)
3496 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3498 memset(&stats[items], 0, pad);
3501 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3502 int bytes)
3504 switch(attrtype) {
3505 case IFLA_INET6_STATS:
3506 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3507 break;
3508 case IFLA_INET6_ICMP6STATS:
3509 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3510 break;
3514 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3515 u32 pid, u32 seq, int event, unsigned int flags)
3517 struct net_device *dev = idev->dev;
3518 struct nlattr *nla;
3519 struct ifinfomsg *hdr;
3520 struct nlmsghdr *nlh;
3521 void *protoinfo;
3522 struct ifla_cacheinfo ci;
3524 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3525 if (nlh == NULL)
3526 return -EMSGSIZE;
3528 hdr = nlmsg_data(nlh);
3529 hdr->ifi_family = AF_INET6;
3530 hdr->__ifi_pad = 0;
3531 hdr->ifi_type = dev->type;
3532 hdr->ifi_index = dev->ifindex;
3533 hdr->ifi_flags = dev_get_flags(dev);
3534 hdr->ifi_change = 0;
3536 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3538 if (dev->addr_len)
3539 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3541 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3542 if (dev->ifindex != dev->iflink)
3543 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3545 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3546 if (protoinfo == NULL)
3547 goto nla_put_failure;
3549 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3551 ci.max_reasm_len = IPV6_MAXPLEN;
3552 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3553 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3554 ci.reachable_time = idev->nd_parms->reachable_time;
3555 ci.retrans_time = idev->nd_parms->retrans_time;
3556 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3558 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3559 if (nla == NULL)
3560 goto nla_put_failure;
3561 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3563 /* XXX - MC not implemented */
3565 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3566 if (nla == NULL)
3567 goto nla_put_failure;
3568 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3570 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3571 if (nla == NULL)
3572 goto nla_put_failure;
3573 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3575 nla_nest_end(skb, protoinfo);
3576 return nlmsg_end(skb, nlh);
3578 nla_put_failure:
3579 nlmsg_cancel(skb, nlh);
3580 return -EMSGSIZE;
3583 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3585 int idx, err;
3586 int s_idx = cb->args[0];
3587 struct net_device *dev;
3588 struct inet6_dev *idev;
3590 read_lock(&dev_base_lock);
3591 idx = 0;
3592 for_each_netdev(&init_net, dev) {
3593 if (idx < s_idx)
3594 goto cont;
3595 if ((idev = in6_dev_get(dev)) == NULL)
3596 goto cont;
3597 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3598 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3599 in6_dev_put(idev);
3600 if (err <= 0)
3601 break;
3602 cont:
3603 idx++;
3605 read_unlock(&dev_base_lock);
3606 cb->args[0] = idx;
3608 return skb->len;
3611 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3613 struct sk_buff *skb;
3614 int err = -ENOBUFS;
3616 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3617 if (skb == NULL)
3618 goto errout;
3620 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3621 if (err < 0) {
3622 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3623 WARN_ON(err == -EMSGSIZE);
3624 kfree_skb(skb);
3625 goto errout;
3627 err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3628 errout:
3629 if (err < 0)
3630 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3633 static inline size_t inet6_prefix_nlmsg_size(void)
3635 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3636 + nla_total_size(sizeof(struct in6_addr))
3637 + nla_total_size(sizeof(struct prefix_cacheinfo));
3640 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3641 struct prefix_info *pinfo, u32 pid, u32 seq,
3642 int event, unsigned int flags)
3644 struct prefixmsg *pmsg;
3645 struct nlmsghdr *nlh;
3646 struct prefix_cacheinfo ci;
3648 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3649 if (nlh == NULL)
3650 return -EMSGSIZE;
3652 pmsg = nlmsg_data(nlh);
3653 pmsg->prefix_family = AF_INET6;
3654 pmsg->prefix_pad1 = 0;
3655 pmsg->prefix_pad2 = 0;
3656 pmsg->prefix_ifindex = idev->dev->ifindex;
3657 pmsg->prefix_len = pinfo->prefix_len;
3658 pmsg->prefix_type = pinfo->type;
3659 pmsg->prefix_pad3 = 0;
3660 pmsg->prefix_flags = 0;
3661 if (pinfo->onlink)
3662 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3663 if (pinfo->autoconf)
3664 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3666 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3668 ci.preferred_time = ntohl(pinfo->prefered);
3669 ci.valid_time = ntohl(pinfo->valid);
3670 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3672 return nlmsg_end(skb, nlh);
3674 nla_put_failure:
3675 nlmsg_cancel(skb, nlh);
3676 return -EMSGSIZE;
3679 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3680 struct prefix_info *pinfo)
3682 struct sk_buff *skb;
3683 int err = -ENOBUFS;
3685 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3686 if (skb == NULL)
3687 goto errout;
3689 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3690 if (err < 0) {
3691 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3692 WARN_ON(err == -EMSGSIZE);
3693 kfree_skb(skb);
3694 goto errout;
3696 err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3697 errout:
3698 if (err < 0)
3699 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3702 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3704 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3706 switch (event) {
3707 case RTM_NEWADDR:
3709 * If the address was optimistic
3710 * we inserted the route at the start of
3711 * our DAD process, so we don't need
3712 * to do it again
3714 if (!(ifp->rt->rt6i_node))
3715 ip6_ins_rt(ifp->rt);
3716 if (ifp->idev->cnf.forwarding)
3717 addrconf_join_anycast(ifp);
3718 break;
3719 case RTM_DELADDR:
3720 if (ifp->idev->cnf.forwarding)
3721 addrconf_leave_anycast(ifp);
3722 addrconf_leave_solict(ifp->idev, &ifp->addr);
3723 dst_hold(&ifp->rt->u.dst);
3724 if (ip6_del_rt(ifp->rt))
3725 dst_free(&ifp->rt->u.dst);
3726 break;
3730 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3732 rcu_read_lock_bh();
3733 if (likely(ifp->idev->dead == 0))
3734 __ipv6_ifa_notify(event, ifp);
3735 rcu_read_unlock_bh();
3738 #ifdef CONFIG_SYSCTL
3740 static
3741 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3742 void __user *buffer, size_t *lenp, loff_t *ppos)
3744 int *valp = ctl->data;
3745 int val = *valp;
3746 int ret;
3748 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3750 if (write && valp != &ipv6_devconf_dflt.forwarding) {
3751 if (valp != &ipv6_devconf.forwarding) {
3752 if ((!*valp) ^ (!val)) {
3753 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3754 if (idev == NULL)
3755 return ret;
3756 dev_forward_change(idev);
3758 } else {
3759 ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3760 addrconf_forward_change();
3762 if (*valp)
3763 rt6_purge_dflt_routers();
3766 return ret;
3769 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3770 int __user *name, int nlen,
3771 void __user *oldval,
3772 size_t __user *oldlenp,
3773 void __user *newval, size_t newlen)
3775 int *valp = table->data;
3776 int new;
3778 if (!newval || !newlen)
3779 return 0;
3780 if (newlen != sizeof(int))
3781 return -EINVAL;
3782 if (get_user(new, (int __user *)newval))
3783 return -EFAULT;
3784 if (new == *valp)
3785 return 0;
3786 if (oldval && oldlenp) {
3787 size_t len;
3788 if (get_user(len, oldlenp))
3789 return -EFAULT;
3790 if (len) {
3791 if (len > table->maxlen)
3792 len = table->maxlen;
3793 if (copy_to_user(oldval, valp, len))
3794 return -EFAULT;
3795 if (put_user(len, oldlenp))
3796 return -EFAULT;
3800 if (valp != &ipv6_devconf_dflt.forwarding) {
3801 if (valp != &ipv6_devconf.forwarding) {
3802 struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3803 int changed;
3804 if (unlikely(idev == NULL))
3805 return -ENODEV;
3806 changed = (!*valp) ^ (!new);
3807 *valp = new;
3808 if (changed)
3809 dev_forward_change(idev);
3810 } else {
3811 *valp = new;
3812 addrconf_forward_change();
3815 if (*valp)
3816 rt6_purge_dflt_routers();
3817 } else
3818 *valp = new;
3820 return 1;
3823 static struct addrconf_sysctl_table
3825 struct ctl_table_header *sysctl_header;
3826 ctl_table addrconf_vars[__NET_IPV6_MAX];
3827 ctl_table addrconf_dev[2];
3828 ctl_table addrconf_conf_dir[2];
3829 ctl_table addrconf_proto_dir[2];
3830 ctl_table addrconf_root_dir[2];
3831 } addrconf_sysctl __read_mostly = {
3832 .sysctl_header = NULL,
3833 .addrconf_vars = {
3835 .ctl_name = NET_IPV6_FORWARDING,
3836 .procname = "forwarding",
3837 .data = &ipv6_devconf.forwarding,
3838 .maxlen = sizeof(int),
3839 .mode = 0644,
3840 .proc_handler = &addrconf_sysctl_forward,
3841 .strategy = &addrconf_sysctl_forward_strategy,
3844 .ctl_name = NET_IPV6_HOP_LIMIT,
3845 .procname = "hop_limit",
3846 .data = &ipv6_devconf.hop_limit,
3847 .maxlen = sizeof(int),
3848 .mode = 0644,
3849 .proc_handler = proc_dointvec,
3852 .ctl_name = NET_IPV6_MTU,
3853 .procname = "mtu",
3854 .data = &ipv6_devconf.mtu6,
3855 .maxlen = sizeof(int),
3856 .mode = 0644,
3857 .proc_handler = &proc_dointvec,
3860 .ctl_name = NET_IPV6_ACCEPT_RA,
3861 .procname = "accept_ra",
3862 .data = &ipv6_devconf.accept_ra,
3863 .maxlen = sizeof(int),
3864 .mode = 0644,
3865 .proc_handler = &proc_dointvec,
3868 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
3869 .procname = "accept_redirects",
3870 .data = &ipv6_devconf.accept_redirects,
3871 .maxlen = sizeof(int),
3872 .mode = 0644,
3873 .proc_handler = &proc_dointvec,
3876 .ctl_name = NET_IPV6_AUTOCONF,
3877 .procname = "autoconf",
3878 .data = &ipv6_devconf.autoconf,
3879 .maxlen = sizeof(int),
3880 .mode = 0644,
3881 .proc_handler = &proc_dointvec,
3884 .ctl_name = NET_IPV6_DAD_TRANSMITS,
3885 .procname = "dad_transmits",
3886 .data = &ipv6_devconf.dad_transmits,
3887 .maxlen = sizeof(int),
3888 .mode = 0644,
3889 .proc_handler = &proc_dointvec,
3892 .ctl_name = NET_IPV6_RTR_SOLICITS,
3893 .procname = "router_solicitations",
3894 .data = &ipv6_devconf.rtr_solicits,
3895 .maxlen = sizeof(int),
3896 .mode = 0644,
3897 .proc_handler = &proc_dointvec,
3900 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
3901 .procname = "router_solicitation_interval",
3902 .data = &ipv6_devconf.rtr_solicit_interval,
3903 .maxlen = sizeof(int),
3904 .mode = 0644,
3905 .proc_handler = &proc_dointvec_jiffies,
3906 .strategy = &sysctl_jiffies,
3909 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
3910 .procname = "router_solicitation_delay",
3911 .data = &ipv6_devconf.rtr_solicit_delay,
3912 .maxlen = sizeof(int),
3913 .mode = 0644,
3914 .proc_handler = &proc_dointvec_jiffies,
3915 .strategy = &sysctl_jiffies,
3918 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
3919 .procname = "force_mld_version",
3920 .data = &ipv6_devconf.force_mld_version,
3921 .maxlen = sizeof(int),
3922 .mode = 0644,
3923 .proc_handler = &proc_dointvec,
3925 #ifdef CONFIG_IPV6_PRIVACY
3927 .ctl_name = NET_IPV6_USE_TEMPADDR,
3928 .procname = "use_tempaddr",
3929 .data = &ipv6_devconf.use_tempaddr,
3930 .maxlen = sizeof(int),
3931 .mode = 0644,
3932 .proc_handler = &proc_dointvec,
3935 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
3936 .procname = "temp_valid_lft",
3937 .data = &ipv6_devconf.temp_valid_lft,
3938 .maxlen = sizeof(int),
3939 .mode = 0644,
3940 .proc_handler = &proc_dointvec,
3943 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
3944 .procname = "temp_prefered_lft",
3945 .data = &ipv6_devconf.temp_prefered_lft,
3946 .maxlen = sizeof(int),
3947 .mode = 0644,
3948 .proc_handler = &proc_dointvec,
3951 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
3952 .procname = "regen_max_retry",
3953 .data = &ipv6_devconf.regen_max_retry,
3954 .maxlen = sizeof(int),
3955 .mode = 0644,
3956 .proc_handler = &proc_dointvec,
3959 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
3960 .procname = "max_desync_factor",
3961 .data = &ipv6_devconf.max_desync_factor,
3962 .maxlen = sizeof(int),
3963 .mode = 0644,
3964 .proc_handler = &proc_dointvec,
3966 #endif
3968 .ctl_name = NET_IPV6_MAX_ADDRESSES,
3969 .procname = "max_addresses",
3970 .data = &ipv6_devconf.max_addresses,
3971 .maxlen = sizeof(int),
3972 .mode = 0644,
3973 .proc_handler = &proc_dointvec,
3976 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
3977 .procname = "accept_ra_defrtr",
3978 .data = &ipv6_devconf.accept_ra_defrtr,
3979 .maxlen = sizeof(int),
3980 .mode = 0644,
3981 .proc_handler = &proc_dointvec,
3984 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
3985 .procname = "accept_ra_pinfo",
3986 .data = &ipv6_devconf.accept_ra_pinfo,
3987 .maxlen = sizeof(int),
3988 .mode = 0644,
3989 .proc_handler = &proc_dointvec,
3991 #ifdef CONFIG_IPV6_ROUTER_PREF
3993 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
3994 .procname = "accept_ra_rtr_pref",
3995 .data = &ipv6_devconf.accept_ra_rtr_pref,
3996 .maxlen = sizeof(int),
3997 .mode = 0644,
3998 .proc_handler = &proc_dointvec,
4001 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4002 .procname = "router_probe_interval",
4003 .data = &ipv6_devconf.rtr_probe_interval,
4004 .maxlen = sizeof(int),
4005 .mode = 0644,
4006 .proc_handler = &proc_dointvec_jiffies,
4007 .strategy = &sysctl_jiffies,
4009 #ifdef CONFIG_IPV6_ROUTE_INFO
4011 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4012 .procname = "accept_ra_rt_info_max_plen",
4013 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4014 .maxlen = sizeof(int),
4015 .mode = 0644,
4016 .proc_handler = &proc_dointvec,
4018 #endif
4019 #endif
4021 .ctl_name = NET_IPV6_PROXY_NDP,
4022 .procname = "proxy_ndp",
4023 .data = &ipv6_devconf.proxy_ndp,
4024 .maxlen = sizeof(int),
4025 .mode = 0644,
4026 .proc_handler = &proc_dointvec,
4029 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4030 .procname = "accept_source_route",
4031 .data = &ipv6_devconf.accept_source_route,
4032 .maxlen = sizeof(int),
4033 .mode = 0644,
4034 .proc_handler = &proc_dointvec,
4036 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4038 .ctl_name = CTL_UNNUMBERED,
4039 .procname = "optimistic_dad",
4040 .data = &ipv6_devconf.optimistic_dad,
4041 .maxlen = sizeof(int),
4042 .mode = 0644,
4043 .proc_handler = &proc_dointvec,
4046 #endif
4048 .ctl_name = 0, /* sentinel */
4051 .addrconf_dev = {
4053 .ctl_name = NET_PROTO_CONF_ALL,
4054 .procname = "all",
4055 .mode = 0555,
4056 .child = addrconf_sysctl.addrconf_vars,
4059 .ctl_name = 0, /* sentinel */
4062 .addrconf_conf_dir = {
4064 .ctl_name = NET_IPV6_CONF,
4065 .procname = "conf",
4066 .mode = 0555,
4067 .child = addrconf_sysctl.addrconf_dev,
4070 .ctl_name = 0, /* sentinel */
4073 .addrconf_proto_dir = {
4075 .ctl_name = NET_IPV6,
4076 .procname = "ipv6",
4077 .mode = 0555,
4078 .child = addrconf_sysctl.addrconf_conf_dir,
4081 .ctl_name = 0, /* sentinel */
4084 .addrconf_root_dir = {
4086 .ctl_name = CTL_NET,
4087 .procname = "net",
4088 .mode = 0555,
4089 .child = addrconf_sysctl.addrconf_proto_dir,
4092 .ctl_name = 0, /* sentinel */
4097 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
4099 int i;
4100 struct net_device *dev = idev ? idev->dev : NULL;
4101 struct addrconf_sysctl_table *t;
4102 char *dev_name = NULL;
4104 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4105 if (t == NULL)
4106 return;
4107 for (i=0; t->addrconf_vars[i].data; i++) {
4108 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4109 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4111 if (dev) {
4112 dev_name = dev->name;
4113 t->addrconf_dev[0].ctl_name = dev->ifindex;
4114 } else {
4115 dev_name = "default";
4116 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
4120 * Make a copy of dev_name, because '.procname' is regarded as const
4121 * by sysctl and we wouldn't want anyone to change it under our feet
4122 * (see SIOCSIFNAME).
4124 dev_name = kstrdup(dev_name, GFP_KERNEL);
4125 if (!dev_name)
4126 goto free;
4128 t->addrconf_dev[0].procname = dev_name;
4130 t->addrconf_dev[0].child = t->addrconf_vars;
4131 t->addrconf_conf_dir[0].child = t->addrconf_dev;
4132 t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
4133 t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
4135 t->sysctl_header = register_sysctl_table(t->addrconf_root_dir);
4136 if (t->sysctl_header == NULL)
4137 goto free_procname;
4138 else
4139 p->sysctl = t;
4140 return;
4142 /* error path */
4143 free_procname:
4144 kfree(dev_name);
4145 free:
4146 kfree(t);
4148 return;
4151 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
4153 if (p->sysctl) {
4154 struct addrconf_sysctl_table *t = p->sysctl;
4155 p->sysctl = NULL;
4156 unregister_sysctl_table(t->sysctl_header);
4157 kfree(t->addrconf_dev[0].procname);
4158 kfree(t);
4163 #endif
4166 * Device notifier
4169 int register_inet6addr_notifier(struct notifier_block *nb)
4171 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4174 EXPORT_SYMBOL(register_inet6addr_notifier);
4176 int unregister_inet6addr_notifier(struct notifier_block *nb)
4178 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4181 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4184 * Init / cleanup code
4187 int __init addrconf_init(void)
4189 int err = 0;
4191 /* The addrconf netdev notifier requires that loopback_dev
4192 * has it's ipv6 private information allocated and setup
4193 * before it can bring up and give link-local addresses
4194 * to other devices which are up.
4196 * Unfortunately, loopback_dev is not necessarily the first
4197 * entry in the global dev_base list of net devices. In fact,
4198 * it is likely to be the very last entry on that list.
4199 * So this causes the notifier registry below to try and
4200 * give link-local addresses to all devices besides loopback_dev
4201 * first, then loopback_dev, which cases all the non-loopback_dev
4202 * devices to fail to get a link-local address.
4204 * So, as a temporary fix, allocate the ipv6 structure for
4205 * loopback_dev first by hand.
4206 * Longer term, all of the dependencies ipv6 has upon the loopback
4207 * device and it being up should be removed.
4209 rtnl_lock();
4210 if (!ipv6_add_dev(&loopback_dev))
4211 err = -ENOMEM;
4212 rtnl_unlock();
4213 if (err)
4214 return err;
4216 ip6_null_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4217 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4218 ip6_prohibit_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4219 ip6_blk_hole_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4220 #endif
4222 register_netdevice_notifier(&ipv6_dev_notf);
4224 addrconf_verify(0);
4226 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4227 if (err < 0)
4228 goto errout;
4230 /* Only the first call to __rtnl_register can fail */
4231 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4232 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4233 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4234 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4235 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4237 #ifdef CONFIG_SYSCTL
4238 addrconf_sysctl.sysctl_header =
4239 register_sysctl_table(addrconf_sysctl.addrconf_root_dir);
4240 addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4241 #endif
4243 return 0;
4244 errout:
4245 unregister_netdevice_notifier(&ipv6_dev_notf);
4247 return err;
4250 void __exit addrconf_cleanup(void)
4252 struct net_device *dev;
4253 struct inet6_ifaddr *ifa;
4254 int i;
4256 unregister_netdevice_notifier(&ipv6_dev_notf);
4258 #ifdef CONFIG_SYSCTL
4259 addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4260 addrconf_sysctl_unregister(&ipv6_devconf);
4261 #endif
4263 rtnl_lock();
4266 * clean dev list.
4269 for_each_netdev(&init_net, dev) {
4270 if (__in6_dev_get(dev) == NULL)
4271 continue;
4272 addrconf_ifdown(dev, 1);
4274 addrconf_ifdown(&loopback_dev, 2);
4277 * Check hash table.
4280 write_lock_bh(&addrconf_hash_lock);
4281 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4282 for (ifa=inet6_addr_lst[i]; ifa; ) {
4283 struct inet6_ifaddr *bifa;
4285 bifa = ifa;
4286 ifa = ifa->lst_next;
4287 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4288 /* Do not free it; something is wrong.
4289 Now we can investigate it with debugger.
4293 write_unlock_bh(&addrconf_hash_lock);
4295 del_timer(&addr_chk_timer);
4297 rtnl_unlock();
4299 #ifdef CONFIG_PROC_FS
4300 proc_net_remove(&init_net, "if_inet6");
4301 #endif