cfg80211: clean up regulatory mess
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / addrconf.c
blob7b6a584b62ddbcbac75ba193fe04cfebc26357da
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/socket.h>
44 #include <linux/sockios.h>
45 #include <linux/net.h>
46 #include <linux/in6.h>
47 #include <linux/netdevice.h>
48 #include <linux/if_addr.h>
49 #include <linux/if_arp.h>
50 #include <linux/if_arcnet.h>
51 #include <linux/if_infiniband.h>
52 #include <linux/route.h>
53 #include <linux/inetdevice.h>
54 #include <linux/init.h>
55 #ifdef CONFIG_SYSCTL
56 #include <linux/sysctl.h>
57 #endif
58 #include <linux/capability.h>
59 #include <linux/delay.h>
60 #include <linux/notifier.h>
61 #include <linux/string.h>
63 #include <net/net_namespace.h>
64 #include <net/sock.h>
65 #include <net/snmp.h>
67 #include <net/ipv6.h>
68 #include <net/protocol.h>
69 #include <net/ndisc.h>
70 #include <net/ip6_route.h>
71 #include <net/addrconf.h>
72 #include <net/tcp.h>
73 #include <net/ip.h>
74 #include <net/netlink.h>
75 #include <net/pkt_sched.h>
76 #include <linux/if_tunnel.h>
77 #include <linux/rtnetlink.h>
79 #ifdef CONFIG_IPV6_PRIVACY
80 #include <linux/random.h>
81 #endif
83 #include <asm/uaccess.h>
84 #include <asm/unaligned.h>
86 #include <linux/proc_fs.h>
87 #include <linux/seq_file.h>
89 /* Set to 3 to get tracing... */
90 #define ACONF_DEBUG 2
92 #if ACONF_DEBUG >= 3
93 #define ADBG(x) printk x
94 #else
95 #define ADBG(x)
96 #endif
98 #define INFINITY_LIFE_TIME 0xFFFFFFFF
99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
101 #ifdef CONFIG_SYSCTL
102 static void addrconf_sysctl_register(struct inet6_dev *idev);
103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
104 #else
105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
112 #endif
114 #ifdef CONFIG_IPV6_PRIVACY
115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
117 static void ipv6_regen_rndid(unsigned long data);
119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
120 #endif
122 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
123 static int ipv6_count_addresses(struct inet6_dev *idev);
126 * Configured unicast address hash table
128 static struct inet6_ifaddr *inet6_addr_lst[IN6_ADDR_HSIZE];
129 static DEFINE_RWLOCK(addrconf_hash_lock);
131 static void addrconf_verify(unsigned long);
133 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
134 static DEFINE_SPINLOCK(addrconf_verify_lock);
136 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
137 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
139 static int addrconf_ifdown(struct net_device *dev, int how);
141 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
142 static void addrconf_dad_timer(unsigned long data);
143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
144 static void addrconf_dad_run(struct inet6_dev *idev);
145 static void addrconf_rs_timer(unsigned long data);
146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
147 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
149 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
150 struct prefix_info *pinfo);
151 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
152 struct net_device *dev);
154 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
156 static struct ipv6_devconf ipv6_devconf __read_mostly = {
157 .forwarding = 0,
158 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
159 .mtu6 = IPV6_MIN_MTU,
160 .accept_ra = 1,
161 .accept_redirects = 1,
162 .autoconf = 1,
163 .force_mld_version = 0,
164 .dad_transmits = 1,
165 .rtr_solicits = MAX_RTR_SOLICITATIONS,
166 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
167 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
168 #ifdef CONFIG_IPV6_PRIVACY
169 .use_tempaddr = 0,
170 .temp_valid_lft = TEMP_VALID_LIFETIME,
171 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
172 .regen_max_retry = REGEN_MAX_RETRY,
173 .max_desync_factor = MAX_DESYNC_FACTOR,
174 #endif
175 .max_addresses = IPV6_MAX_ADDRESSES,
176 .accept_ra_defrtr = 1,
177 .accept_ra_pinfo = 1,
178 #ifdef CONFIG_IPV6_ROUTER_PREF
179 .accept_ra_rtr_pref = 1,
180 .rtr_probe_interval = 60 * HZ,
181 #ifdef CONFIG_IPV6_ROUTE_INFO
182 .accept_ra_rt_info_max_plen = 0,
183 #endif
184 #endif
185 .proxy_ndp = 0,
186 .accept_source_route = 0, /* we do not accept RH0 by default. */
187 .disable_ipv6 = 0,
188 .accept_dad = 1,
191 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
192 .forwarding = 0,
193 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
194 .mtu6 = IPV6_MIN_MTU,
195 .accept_ra = 1,
196 .accept_redirects = 1,
197 .autoconf = 1,
198 .dad_transmits = 1,
199 .rtr_solicits = MAX_RTR_SOLICITATIONS,
200 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
201 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
202 #ifdef CONFIG_IPV6_PRIVACY
203 .use_tempaddr = 0,
204 .temp_valid_lft = TEMP_VALID_LIFETIME,
205 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
206 .regen_max_retry = REGEN_MAX_RETRY,
207 .max_desync_factor = MAX_DESYNC_FACTOR,
208 #endif
209 .max_addresses = IPV6_MAX_ADDRESSES,
210 .accept_ra_defrtr = 1,
211 .accept_ra_pinfo = 1,
212 #ifdef CONFIG_IPV6_ROUTER_PREF
213 .accept_ra_rtr_pref = 1,
214 .rtr_probe_interval = 60 * HZ,
215 #ifdef CONFIG_IPV6_ROUTE_INFO
216 .accept_ra_rt_info_max_plen = 0,
217 #endif
218 #endif
219 .proxy_ndp = 0,
220 .accept_source_route = 0, /* we do not accept RH0 by default. */
221 .disable_ipv6 = 0,
222 .accept_dad = 1,
225 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
226 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
227 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
228 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
229 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
231 /* Check if a valid qdisc is available */
232 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
234 return !qdisc_tx_is_noop(dev);
237 /* Check if a route is valid prefix route */
238 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
240 return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
243 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
245 if (del_timer(&ifp->timer))
246 __in6_ifa_put(ifp);
249 enum addrconf_timer_t
251 AC_NONE,
252 AC_DAD,
253 AC_RS,
256 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
257 enum addrconf_timer_t what,
258 unsigned long when)
260 if (!del_timer(&ifp->timer))
261 in6_ifa_hold(ifp);
263 switch (what) {
264 case AC_DAD:
265 ifp->timer.function = addrconf_dad_timer;
266 break;
267 case AC_RS:
268 ifp->timer.function = addrconf_rs_timer;
269 break;
270 default:;
272 ifp->timer.expires = jiffies + when;
273 add_timer(&ifp->timer);
276 static int snmp6_alloc_dev(struct inet6_dev *idev)
278 if (snmp_mib_init((void **)idev->stats.ipv6,
279 sizeof(struct ipstats_mib)) < 0)
280 goto err_ip;
281 if (snmp_mib_init((void **)idev->stats.icmpv6,
282 sizeof(struct icmpv6_mib)) < 0)
283 goto err_icmp;
284 if (snmp_mib_init((void **)idev->stats.icmpv6msg,
285 sizeof(struct icmpv6msg_mib)) < 0)
286 goto err_icmpmsg;
288 return 0;
290 err_icmpmsg:
291 snmp_mib_free((void **)idev->stats.icmpv6);
292 err_icmp:
293 snmp_mib_free((void **)idev->stats.ipv6);
294 err_ip:
295 return -ENOMEM;
298 static void snmp6_free_dev(struct inet6_dev *idev)
300 snmp_mib_free((void **)idev->stats.icmpv6msg);
301 snmp_mib_free((void **)idev->stats.icmpv6);
302 snmp_mib_free((void **)idev->stats.ipv6);
305 /* Nobody refers to this device, we may destroy it. */
307 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
309 struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
310 kfree(idev);
313 void in6_dev_finish_destroy(struct inet6_dev *idev)
315 struct net_device *dev = idev->dev;
317 WARN_ON(idev->addr_list != NULL);
318 WARN_ON(idev->mc_list != NULL);
320 #ifdef NET_REFCNT_DEBUG
321 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
322 #endif
323 dev_put(dev);
324 if (!idev->dead) {
325 printk("Freeing alive inet6 device %p\n", idev);
326 return;
328 snmp6_free_dev(idev);
329 call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
332 EXPORT_SYMBOL(in6_dev_finish_destroy);
334 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
336 struct inet6_dev *ndev;
338 ASSERT_RTNL();
340 if (dev->mtu < IPV6_MIN_MTU)
341 return NULL;
343 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
345 if (ndev == NULL)
346 return NULL;
348 rwlock_init(&ndev->lock);
349 ndev->dev = dev;
350 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
351 ndev->cnf.mtu6 = dev->mtu;
352 ndev->cnf.sysctl = NULL;
353 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
354 if (ndev->nd_parms == NULL) {
355 kfree(ndev);
356 return NULL;
358 if (ndev->cnf.forwarding)
359 dev_disable_lro(dev);
360 /* We refer to the device */
361 dev_hold(dev);
363 if (snmp6_alloc_dev(ndev) < 0) {
364 ADBG((KERN_WARNING
365 "%s(): cannot allocate memory for statistics; dev=%s.\n",
366 __func__, dev->name));
367 neigh_parms_release(&nd_tbl, ndev->nd_parms);
368 ndev->dead = 1;
369 in6_dev_finish_destroy(ndev);
370 return NULL;
373 if (snmp6_register_dev(ndev) < 0) {
374 ADBG((KERN_WARNING
375 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
376 __func__, dev->name));
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 ndev->dead = 1;
379 in6_dev_finish_destroy(ndev);
380 return NULL;
383 /* One reference from device. We must do this before
384 * we invoke __ipv6_regen_rndid().
386 in6_dev_hold(ndev);
388 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
389 ndev->cnf.accept_dad = -1;
391 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
392 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
393 printk(KERN_INFO
394 "%s: Disabled Multicast RS\n",
395 dev->name);
396 ndev->cnf.rtr_solicits = 0;
398 #endif
400 #ifdef CONFIG_IPV6_PRIVACY
401 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
402 if ((dev->flags&IFF_LOOPBACK) ||
403 dev->type == ARPHRD_TUNNEL ||
404 dev->type == ARPHRD_TUNNEL6 ||
405 dev->type == ARPHRD_SIT ||
406 dev->type == ARPHRD_NONE) {
407 printk(KERN_INFO
408 "%s: Disabled Privacy Extensions\n",
409 dev->name);
410 ndev->cnf.use_tempaddr = -1;
411 } else {
412 in6_dev_hold(ndev);
413 ipv6_regen_rndid((unsigned long) ndev);
415 #endif
417 if (netif_running(dev) && addrconf_qdisc_ok(dev))
418 ndev->if_flags |= IF_READY;
420 ipv6_mc_init_dev(ndev);
421 ndev->tstamp = jiffies;
422 addrconf_sysctl_register(ndev);
423 /* protected by rtnl_lock */
424 rcu_assign_pointer(dev->ip6_ptr, ndev);
426 /* Join all-node multicast group */
427 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
429 return ndev;
432 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
434 struct inet6_dev *idev;
436 ASSERT_RTNL();
438 if ((idev = __in6_dev_get(dev)) == NULL) {
439 if ((idev = ipv6_add_dev(dev)) == NULL)
440 return NULL;
443 if (dev->flags&IFF_UP)
444 ipv6_mc_up(idev);
445 return idev;
448 #ifdef CONFIG_SYSCTL
449 static void dev_forward_change(struct inet6_dev *idev)
451 struct net_device *dev;
452 struct inet6_ifaddr *ifa;
454 if (!idev)
455 return;
456 dev = idev->dev;
457 if (idev->cnf.forwarding)
458 dev_disable_lro(dev);
459 if (dev && (dev->flags & IFF_MULTICAST)) {
460 if (idev->cnf.forwarding)
461 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
462 else
463 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
465 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
466 if (ifa->flags&IFA_F_TENTATIVE)
467 continue;
468 if (idev->cnf.forwarding)
469 addrconf_join_anycast(ifa);
470 else
471 addrconf_leave_anycast(ifa);
476 static void addrconf_forward_change(struct net *net, __s32 newf)
478 struct net_device *dev;
479 struct inet6_dev *idev;
481 read_lock(&dev_base_lock);
482 for_each_netdev(net, dev) {
483 rcu_read_lock();
484 idev = __in6_dev_get(dev);
485 if (idev) {
486 int changed = (!idev->cnf.forwarding) ^ (!newf);
487 idev->cnf.forwarding = newf;
488 if (changed)
489 dev_forward_change(idev);
491 rcu_read_unlock();
493 read_unlock(&dev_base_lock);
496 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
498 struct net *net;
500 net = (struct net *)table->extra2;
501 if (p == &net->ipv6.devconf_dflt->forwarding)
502 return;
504 rtnl_lock();
505 if (p == &net->ipv6.devconf_all->forwarding) {
506 __s32 newf = net->ipv6.devconf_all->forwarding;
507 net->ipv6.devconf_dflt->forwarding = newf;
508 addrconf_forward_change(net, newf);
509 } else if ((!*p) ^ (!old))
510 dev_forward_change((struct inet6_dev *)table->extra1);
511 rtnl_unlock();
513 if (*p)
514 rt6_purge_dflt_routers(net);
516 #endif
518 /* Nobody refers to this ifaddr, destroy it */
520 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
522 WARN_ON(ifp->if_next != NULL);
523 WARN_ON(ifp->lst_next != NULL);
525 #ifdef NET_REFCNT_DEBUG
526 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
527 #endif
529 in6_dev_put(ifp->idev);
531 if (del_timer(&ifp->timer))
532 printk("Timer is still running, when freeing ifa=%p\n", ifp);
534 if (!ifp->dead) {
535 printk("Freeing alive inet6 address %p\n", ifp);
536 return;
538 dst_release(&ifp->rt->u.dst);
540 kfree(ifp);
543 static void
544 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
546 struct inet6_ifaddr *ifa, **ifap;
547 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
550 * Each device address list is sorted in order of scope -
551 * global before linklocal.
553 for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
554 ifap = &ifa->if_next) {
555 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
556 break;
559 ifp->if_next = *ifap;
560 *ifap = ifp;
564 * Hash function taken from net_alias.c
566 static u8 ipv6_addr_hash(const struct in6_addr *addr)
568 __u32 word;
571 * We perform the hash function over the last 64 bits of the address
572 * This will include the IEEE address token on links that support it.
575 word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
576 word ^= (word >> 16);
577 word ^= (word >> 8);
579 return ((word ^ (word >> 4)) & 0x0f);
582 /* On success it returns ifp with increased reference count */
584 static struct inet6_ifaddr *
585 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
586 int scope, u32 flags)
588 struct inet6_ifaddr *ifa = NULL;
589 struct rt6_info *rt;
590 int hash;
591 int err = 0;
592 int addr_type = ipv6_addr_type(addr);
594 if (addr_type == IPV6_ADDR_ANY ||
595 addr_type & IPV6_ADDR_MULTICAST ||
596 (!(idev->dev->flags & IFF_LOOPBACK) &&
597 addr_type & IPV6_ADDR_LOOPBACK))
598 return ERR_PTR(-EADDRNOTAVAIL);
600 rcu_read_lock_bh();
601 if (idev->dead) {
602 err = -ENODEV; /*XXX*/
603 goto out2;
606 write_lock(&addrconf_hash_lock);
608 /* Ignore adding duplicate addresses on an interface */
609 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
610 ADBG(("ipv6_add_addr: already assigned\n"));
611 err = -EEXIST;
612 goto out;
615 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
617 if (ifa == NULL) {
618 ADBG(("ipv6_add_addr: malloc failed\n"));
619 err = -ENOBUFS;
620 goto out;
623 rt = addrconf_dst_alloc(idev, addr, 0);
624 if (IS_ERR(rt)) {
625 err = PTR_ERR(rt);
626 goto out;
629 ipv6_addr_copy(&ifa->addr, addr);
631 spin_lock_init(&ifa->lock);
632 init_timer(&ifa->timer);
633 ifa->timer.data = (unsigned long) ifa;
634 ifa->scope = scope;
635 ifa->prefix_len = pfxlen;
636 ifa->flags = flags | IFA_F_TENTATIVE;
637 ifa->cstamp = ifa->tstamp = jiffies;
639 ifa->rt = rt;
642 * part one of RFC 4429, section 3.3
643 * We should not configure an address as
644 * optimistic if we do not yet know the link
645 * layer address of our nexhop router
648 if (rt->rt6i_nexthop == NULL)
649 ifa->flags &= ~IFA_F_OPTIMISTIC;
651 ifa->idev = idev;
652 in6_dev_hold(idev);
653 /* For caller */
654 in6_ifa_hold(ifa);
656 /* Add to big hash table */
657 hash = ipv6_addr_hash(addr);
659 ifa->lst_next = inet6_addr_lst[hash];
660 inet6_addr_lst[hash] = ifa;
661 in6_ifa_hold(ifa);
662 write_unlock(&addrconf_hash_lock);
664 write_lock(&idev->lock);
665 /* Add to inet6_dev unicast addr list. */
666 ipv6_link_dev_addr(idev, ifa);
668 #ifdef CONFIG_IPV6_PRIVACY
669 if (ifa->flags&IFA_F_TEMPORARY) {
670 ifa->tmp_next = idev->tempaddr_list;
671 idev->tempaddr_list = ifa;
672 in6_ifa_hold(ifa);
674 #endif
676 in6_ifa_hold(ifa);
677 write_unlock(&idev->lock);
678 out2:
679 rcu_read_unlock_bh();
681 if (likely(err == 0))
682 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
683 else {
684 kfree(ifa);
685 ifa = ERR_PTR(err);
688 return ifa;
689 out:
690 write_unlock(&addrconf_hash_lock);
691 goto out2;
694 /* This function wants to get referenced ifp and releases it before return */
696 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
698 struct inet6_ifaddr *ifa, **ifap;
699 struct inet6_dev *idev = ifp->idev;
700 int hash;
701 int deleted = 0, onlink = 0;
702 unsigned long expires = jiffies;
704 hash = ipv6_addr_hash(&ifp->addr);
706 ifp->dead = 1;
708 write_lock_bh(&addrconf_hash_lock);
709 for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
710 ifap = &ifa->lst_next) {
711 if (ifa == ifp) {
712 *ifap = ifa->lst_next;
713 __in6_ifa_put(ifp);
714 ifa->lst_next = NULL;
715 break;
718 write_unlock_bh(&addrconf_hash_lock);
720 write_lock_bh(&idev->lock);
721 #ifdef CONFIG_IPV6_PRIVACY
722 if (ifp->flags&IFA_F_TEMPORARY) {
723 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
724 ifap = &ifa->tmp_next) {
725 if (ifa == ifp) {
726 *ifap = ifa->tmp_next;
727 if (ifp->ifpub) {
728 in6_ifa_put(ifp->ifpub);
729 ifp->ifpub = NULL;
731 __in6_ifa_put(ifp);
732 ifa->tmp_next = NULL;
733 break;
737 #endif
739 for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
740 if (ifa == ifp) {
741 *ifap = ifa->if_next;
742 __in6_ifa_put(ifp);
743 ifa->if_next = NULL;
744 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
745 break;
746 deleted = 1;
747 continue;
748 } else if (ifp->flags & IFA_F_PERMANENT) {
749 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
750 ifp->prefix_len)) {
751 if (ifa->flags & IFA_F_PERMANENT) {
752 onlink = 1;
753 if (deleted)
754 break;
755 } else {
756 unsigned long lifetime;
758 if (!onlink)
759 onlink = -1;
761 spin_lock(&ifa->lock);
763 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
765 * Note: Because this address is
766 * not permanent, lifetime <
767 * LONG_MAX / HZ here.
769 if (time_before(expires,
770 ifa->tstamp + lifetime * HZ))
771 expires = ifa->tstamp + lifetime * HZ;
772 spin_unlock(&ifa->lock);
776 ifap = &ifa->if_next;
778 write_unlock_bh(&idev->lock);
780 addrconf_del_timer(ifp);
782 ipv6_ifa_notify(RTM_DELADDR, ifp);
784 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
787 * Purge or update corresponding prefix
789 * 1) we don't purge prefix here if address was not permanent.
790 * prefix is managed by its own lifetime.
791 * 2) if there're no addresses, delete prefix.
792 * 3) if there're still other permanent address(es),
793 * corresponding prefix is still permanent.
794 * 4) otherwise, update prefix lifetime to the
795 * longest valid lifetime among the corresponding
796 * addresses on the device.
797 * Note: subsequent RA will update lifetime.
799 * --yoshfuji
801 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
802 struct in6_addr prefix;
803 struct rt6_info *rt;
804 struct net *net = dev_net(ifp->idev->dev);
805 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
806 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
808 if (rt && addrconf_is_prefix_route(rt)) {
809 if (onlink == 0) {
810 ip6_del_rt(rt);
811 rt = NULL;
812 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
813 rt->rt6i_expires = expires;
814 rt->rt6i_flags |= RTF_EXPIRES;
817 dst_release(&rt->u.dst);
820 in6_ifa_put(ifp);
823 #ifdef CONFIG_IPV6_PRIVACY
824 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
826 struct inet6_dev *idev = ifp->idev;
827 struct in6_addr addr, *tmpaddr;
828 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
829 unsigned long regen_advance;
830 int tmp_plen;
831 int ret = 0;
832 int max_addresses;
833 u32 addr_flags;
835 write_lock(&idev->lock);
836 if (ift) {
837 spin_lock_bh(&ift->lock);
838 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
839 spin_unlock_bh(&ift->lock);
840 tmpaddr = &addr;
841 } else {
842 tmpaddr = NULL;
844 retry:
845 in6_dev_hold(idev);
846 if (idev->cnf.use_tempaddr <= 0) {
847 write_unlock(&idev->lock);
848 printk(KERN_INFO
849 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
850 in6_dev_put(idev);
851 ret = -1;
852 goto out;
854 spin_lock_bh(&ifp->lock);
855 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
856 idev->cnf.use_tempaddr = -1; /*XXX*/
857 spin_unlock_bh(&ifp->lock);
858 write_unlock(&idev->lock);
859 printk(KERN_WARNING
860 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
861 in6_dev_put(idev);
862 ret = -1;
863 goto out;
865 in6_ifa_hold(ifp);
866 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
867 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
868 spin_unlock_bh(&ifp->lock);
869 write_unlock(&idev->lock);
870 printk(KERN_WARNING
871 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
872 in6_ifa_put(ifp);
873 in6_dev_put(idev);
874 ret = -1;
875 goto out;
877 memcpy(&addr.s6_addr[8], idev->rndid, 8);
878 tmp_valid_lft = min_t(__u32,
879 ifp->valid_lft,
880 idev->cnf.temp_valid_lft);
881 tmp_prefered_lft = min_t(__u32,
882 ifp->prefered_lft,
883 idev->cnf.temp_prefered_lft - desync_factor / HZ);
884 tmp_plen = ifp->prefix_len;
885 max_addresses = idev->cnf.max_addresses;
886 tmp_cstamp = ifp->cstamp;
887 tmp_tstamp = ifp->tstamp;
888 spin_unlock_bh(&ifp->lock);
890 regen_advance = idev->cnf.regen_max_retry *
891 idev->cnf.dad_transmits *
892 idev->nd_parms->retrans_time / HZ;
893 write_unlock(&idev->lock);
895 /* A temporary address is created only if this calculated Preferred
896 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
897 * an implementation must not create a temporary address with a zero
898 * Preferred Lifetime.
900 if (tmp_prefered_lft <= regen_advance) {
901 in6_ifa_put(ifp);
902 in6_dev_put(idev);
903 ret = -1;
904 goto out;
907 addr_flags = IFA_F_TEMPORARY;
908 /* set in addrconf_prefix_rcv() */
909 if (ifp->flags & IFA_F_OPTIMISTIC)
910 addr_flags |= IFA_F_OPTIMISTIC;
912 ift = !max_addresses ||
913 ipv6_count_addresses(idev) < max_addresses ?
914 ipv6_add_addr(idev, &addr, tmp_plen,
915 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
916 addr_flags) : NULL;
917 if (!ift || IS_ERR(ift)) {
918 in6_ifa_put(ifp);
919 in6_dev_put(idev);
920 printk(KERN_INFO
921 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
922 tmpaddr = &addr;
923 write_lock(&idev->lock);
924 goto retry;
927 spin_lock_bh(&ift->lock);
928 ift->ifpub = ifp;
929 ift->valid_lft = tmp_valid_lft;
930 ift->prefered_lft = tmp_prefered_lft;
931 ift->cstamp = tmp_cstamp;
932 ift->tstamp = tmp_tstamp;
933 spin_unlock_bh(&ift->lock);
935 addrconf_dad_start(ift, 0);
936 in6_ifa_put(ift);
937 in6_dev_put(idev);
938 out:
939 return ret;
941 #endif
944 * Choose an appropriate source address (RFC3484)
946 enum {
947 IPV6_SADDR_RULE_INIT = 0,
948 IPV6_SADDR_RULE_LOCAL,
949 IPV6_SADDR_RULE_SCOPE,
950 IPV6_SADDR_RULE_PREFERRED,
951 #ifdef CONFIG_IPV6_MIP6
952 IPV6_SADDR_RULE_HOA,
953 #endif
954 IPV6_SADDR_RULE_OIF,
955 IPV6_SADDR_RULE_LABEL,
956 #ifdef CONFIG_IPV6_PRIVACY
957 IPV6_SADDR_RULE_PRIVACY,
958 #endif
959 IPV6_SADDR_RULE_ORCHID,
960 IPV6_SADDR_RULE_PREFIX,
961 IPV6_SADDR_RULE_MAX
964 struct ipv6_saddr_score {
965 int rule;
966 int addr_type;
967 struct inet6_ifaddr *ifa;
968 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
969 int scopedist;
970 int matchlen;
973 struct ipv6_saddr_dst {
974 const struct in6_addr *addr;
975 int ifindex;
976 int scope;
977 int label;
978 unsigned int prefs;
981 static inline int ipv6_saddr_preferred(int type)
983 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
984 IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
985 return 1;
986 return 0;
989 static int ipv6_get_saddr_eval(struct net *net,
990 struct ipv6_saddr_score *score,
991 struct ipv6_saddr_dst *dst,
992 int i)
994 int ret;
996 if (i <= score->rule) {
997 switch (i) {
998 case IPV6_SADDR_RULE_SCOPE:
999 ret = score->scopedist;
1000 break;
1001 case IPV6_SADDR_RULE_PREFIX:
1002 ret = score->matchlen;
1003 break;
1004 default:
1005 ret = !!test_bit(i, score->scorebits);
1007 goto out;
1010 switch (i) {
1011 case IPV6_SADDR_RULE_INIT:
1012 /* Rule 0: remember if hiscore is not ready yet */
1013 ret = !!score->ifa;
1014 break;
1015 case IPV6_SADDR_RULE_LOCAL:
1016 /* Rule 1: Prefer same address */
1017 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018 break;
1019 case IPV6_SADDR_RULE_SCOPE:
1020 /* Rule 2: Prefer appropriate scope
1022 * ret
1024 * -1 | d 15
1025 * ---+--+-+---> scope
1027 * | d is scope of the destination.
1028 * B-d | \
1029 * | \ <- smaller scope is better if
1030 * B-15 | \ if scope is enough for destinaion.
1031 * | ret = B - scope (-1 <= scope >= d <= 15).
1032 * d-C-1 | /
1033 * |/ <- greater is better
1034 * -C / if scope is not enough for destination.
1035 * /| ret = scope - C (-1 <= d < scope <= 15).
1037 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039 * Assume B = 0 and we get C > 29.
1041 ret = __ipv6_addr_src_scope(score->addr_type);
1042 if (ret >= dst->scope)
1043 ret = -ret;
1044 else
1045 ret -= 128; /* 30 is enough */
1046 score->scopedist = ret;
1047 break;
1048 case IPV6_SADDR_RULE_PREFERRED:
1049 /* Rule 3: Avoid deprecated and optimistic addresses */
1050 ret = ipv6_saddr_preferred(score->addr_type) ||
1051 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052 break;
1053 #ifdef CONFIG_IPV6_MIP6
1054 case IPV6_SADDR_RULE_HOA:
1056 /* Rule 4: Prefer home address */
1057 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059 break;
1061 #endif
1062 case IPV6_SADDR_RULE_OIF:
1063 /* Rule 5: Prefer outgoing interface */
1064 ret = (!dst->ifindex ||
1065 dst->ifindex == score->ifa->idev->dev->ifindex);
1066 break;
1067 case IPV6_SADDR_RULE_LABEL:
1068 /* Rule 6: Prefer matching label */
1069 ret = ipv6_addr_label(net,
1070 &score->ifa->addr, score->addr_type,
1071 score->ifa->idev->dev->ifindex) == dst->label;
1072 break;
1073 #ifdef CONFIG_IPV6_PRIVACY
1074 case IPV6_SADDR_RULE_PRIVACY:
1076 /* Rule 7: Prefer public address
1077 * Note: prefer temprary address if use_tempaddr >= 2
1079 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081 score->ifa->idev->cnf.use_tempaddr >= 2;
1082 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083 break;
1085 #endif
1086 case IPV6_SADDR_RULE_ORCHID:
1087 /* Rule 8-: Prefer ORCHID vs ORCHID or
1088 * non-ORCHID vs non-ORCHID
1090 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091 ipv6_addr_orchid(dst->addr));
1092 break;
1093 case IPV6_SADDR_RULE_PREFIX:
1094 /* Rule 8: Use longest matching prefix */
1095 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096 dst->addr);
1097 break;
1098 default:
1099 ret = 0;
1102 if (ret)
1103 __set_bit(i, score->scorebits);
1104 score->rule = i;
1105 out:
1106 return ret;
1109 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110 const struct in6_addr *daddr, unsigned int prefs,
1111 struct in6_addr *saddr)
1113 struct ipv6_saddr_score scores[2],
1114 *score = &scores[0], *hiscore = &scores[1];
1115 struct ipv6_saddr_dst dst;
1116 struct net_device *dev;
1117 int dst_type;
1119 dst_type = __ipv6_addr_type(daddr);
1120 dst.addr = daddr;
1121 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122 dst.scope = __ipv6_addr_src_scope(dst_type);
1123 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124 dst.prefs = prefs;
1126 hiscore->rule = -1;
1127 hiscore->ifa = NULL;
1129 read_lock(&dev_base_lock);
1130 rcu_read_lock();
1132 for_each_netdev(net, dev) {
1133 struct inet6_dev *idev;
1135 /* Candidate Source Address (section 4)
1136 * - multicast and link-local destination address,
1137 * the set of candidate source address MUST only
1138 * include addresses assigned to interfaces
1139 * belonging to the same link as the outgoing
1140 * interface.
1141 * (- For site-local destination addresses, the
1142 * set of candidate source addresses MUST only
1143 * include addresses assigned to interfaces
1144 * belonging to the same site as the outgoing
1145 * interface.)
1147 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1148 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1149 dst.ifindex && dev->ifindex != dst.ifindex)
1150 continue;
1152 idev = __in6_dev_get(dev);
1153 if (!idev)
1154 continue;
1156 read_lock_bh(&idev->lock);
1157 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1158 int i;
1161 * - Tentative Address (RFC2462 section 5.4)
1162 * - A tentative address is not considered
1163 * "assigned to an interface" in the traditional
1164 * sense, unless it is also flagged as optimistic.
1165 * - Candidate Source Address (section 4)
1166 * - In any case, anycast addresses, multicast
1167 * addresses, and the unspecified address MUST
1168 * NOT be included in a candidate set.
1170 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1171 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1172 continue;
1174 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1176 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1177 score->addr_type & IPV6_ADDR_MULTICAST)) {
1178 LIMIT_NETDEBUG(KERN_DEBUG
1179 "ADDRCONF: unspecified / multicast address "
1180 "assigned as unicast address on %s",
1181 dev->name);
1182 continue;
1185 score->rule = -1;
1186 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1188 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1189 int minihiscore, miniscore;
1191 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1192 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1194 if (minihiscore > miniscore) {
1195 if (i == IPV6_SADDR_RULE_SCOPE &&
1196 score->scopedist > 0) {
1198 * special case:
1199 * each remaining entry
1200 * has too small (not enough)
1201 * scope, because ifa entries
1202 * are sorted by their scope
1203 * values.
1205 goto try_nextdev;
1207 break;
1208 } else if (minihiscore < miniscore) {
1209 struct ipv6_saddr_score *tmp;
1211 if (hiscore->ifa)
1212 in6_ifa_put(hiscore->ifa);
1214 in6_ifa_hold(score->ifa);
1216 tmp = hiscore;
1217 hiscore = score;
1218 score = tmp;
1220 /* restore our iterator */
1221 score->ifa = hiscore->ifa;
1223 break;
1227 try_nextdev:
1228 read_unlock_bh(&idev->lock);
1230 rcu_read_unlock();
1231 read_unlock(&dev_base_lock);
1233 if (!hiscore->ifa)
1234 return -EADDRNOTAVAIL;
1236 ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1237 in6_ifa_put(hiscore->ifa);
1238 return 0;
1241 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1243 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1244 unsigned char banned_flags)
1246 struct inet6_dev *idev;
1247 int err = -EADDRNOTAVAIL;
1249 rcu_read_lock();
1250 if ((idev = __in6_dev_get(dev)) != NULL) {
1251 struct inet6_ifaddr *ifp;
1253 read_lock_bh(&idev->lock);
1254 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1255 if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1256 ipv6_addr_copy(addr, &ifp->addr);
1257 err = 0;
1258 break;
1261 read_unlock_bh(&idev->lock);
1263 rcu_read_unlock();
1264 return err;
1267 static int ipv6_count_addresses(struct inet6_dev *idev)
1269 int cnt = 0;
1270 struct inet6_ifaddr *ifp;
1272 read_lock_bh(&idev->lock);
1273 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1274 cnt++;
1275 read_unlock_bh(&idev->lock);
1276 return cnt;
1279 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1280 struct net_device *dev, int strict)
1282 struct inet6_ifaddr * ifp;
1283 u8 hash = ipv6_addr_hash(addr);
1285 read_lock_bh(&addrconf_hash_lock);
1286 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1287 if (!net_eq(dev_net(ifp->idev->dev), net))
1288 continue;
1289 if (ipv6_addr_equal(&ifp->addr, addr) &&
1290 !(ifp->flags&IFA_F_TENTATIVE)) {
1291 if (dev == NULL || ifp->idev->dev == dev ||
1292 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1293 break;
1296 read_unlock_bh(&addrconf_hash_lock);
1297 return ifp != NULL;
1299 EXPORT_SYMBOL(ipv6_chk_addr);
1301 static
1302 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1303 struct net_device *dev)
1305 struct inet6_ifaddr * ifp;
1306 u8 hash = ipv6_addr_hash(addr);
1308 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1309 if (!net_eq(dev_net(ifp->idev->dev), net))
1310 continue;
1311 if (ipv6_addr_equal(&ifp->addr, addr)) {
1312 if (dev == NULL || ifp->idev->dev == dev)
1313 break;
1316 return ifp != NULL;
1319 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1321 struct inet6_dev *idev;
1322 struct inet6_ifaddr *ifa;
1323 int onlink;
1325 onlink = 0;
1326 rcu_read_lock();
1327 idev = __in6_dev_get(dev);
1328 if (idev) {
1329 read_lock_bh(&idev->lock);
1330 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1331 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1332 ifa->prefix_len);
1333 if (onlink)
1334 break;
1336 read_unlock_bh(&idev->lock);
1338 rcu_read_unlock();
1339 return onlink;
1342 EXPORT_SYMBOL(ipv6_chk_prefix);
1344 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1345 struct net_device *dev, int strict)
1347 struct inet6_ifaddr * ifp;
1348 u8 hash = ipv6_addr_hash(addr);
1350 read_lock_bh(&addrconf_hash_lock);
1351 for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1352 if (!net_eq(dev_net(ifp->idev->dev), net))
1353 continue;
1354 if (ipv6_addr_equal(&ifp->addr, addr)) {
1355 if (dev == NULL || ifp->idev->dev == dev ||
1356 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1357 in6_ifa_hold(ifp);
1358 break;
1362 read_unlock_bh(&addrconf_hash_lock);
1364 return ifp;
1367 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1369 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1370 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1371 __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1372 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1373 int sk_ipv6only = ipv6_only_sock(sk);
1374 int sk2_ipv6only = inet_v6_ipv6only(sk2);
1375 int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1376 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1378 if (!sk2_rcv_saddr && !sk_ipv6only)
1379 return 1;
1381 if (addr_type2 == IPV6_ADDR_ANY &&
1382 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1383 return 1;
1385 if (addr_type == IPV6_ADDR_ANY &&
1386 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1387 return 1;
1389 if (sk2_rcv_saddr6 &&
1390 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1391 return 1;
1393 if (addr_type == IPV6_ADDR_MAPPED &&
1394 !sk2_ipv6only &&
1395 (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1396 return 1;
1398 return 0;
1401 /* Gets referenced address, destroys ifaddr */
1403 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1405 if (ifp->flags&IFA_F_PERMANENT) {
1406 spin_lock_bh(&ifp->lock);
1407 addrconf_del_timer(ifp);
1408 ifp->flags |= IFA_F_TENTATIVE;
1409 spin_unlock_bh(&ifp->lock);
1410 in6_ifa_put(ifp);
1411 #ifdef CONFIG_IPV6_PRIVACY
1412 } else if (ifp->flags&IFA_F_TEMPORARY) {
1413 struct inet6_ifaddr *ifpub;
1414 spin_lock_bh(&ifp->lock);
1415 ifpub = ifp->ifpub;
1416 if (ifpub) {
1417 in6_ifa_hold(ifpub);
1418 spin_unlock_bh(&ifp->lock);
1419 ipv6_create_tempaddr(ifpub, ifp);
1420 in6_ifa_put(ifpub);
1421 } else {
1422 spin_unlock_bh(&ifp->lock);
1424 ipv6_del_addr(ifp);
1425 #endif
1426 } else
1427 ipv6_del_addr(ifp);
1430 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1432 struct inet6_dev *idev = ifp->idev;
1433 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1434 struct in6_addr addr;
1436 addr.s6_addr32[0] = htonl(0xfe800000);
1437 addr.s6_addr32[1] = 0;
1439 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1440 ipv6_addr_equal(&ifp->addr, &addr)) {
1441 /* DAD failed for link-local based on MAC address */
1442 idev->cnf.disable_ipv6 = 1;
1446 if (net_ratelimit())
1447 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1448 addrconf_dad_stop(ifp);
1451 /* Join to solicited addr multicast group. */
1453 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1455 struct in6_addr maddr;
1457 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1458 return;
1460 addrconf_addr_solict_mult(addr, &maddr);
1461 ipv6_dev_mc_inc(dev, &maddr);
1464 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1466 struct in6_addr maddr;
1468 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1469 return;
1471 addrconf_addr_solict_mult(addr, &maddr);
1472 __ipv6_dev_mc_dec(idev, &maddr);
1475 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1477 struct in6_addr addr;
1478 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1479 if (ipv6_addr_any(&addr))
1480 return;
1481 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1484 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1486 struct in6_addr addr;
1487 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1488 if (ipv6_addr_any(&addr))
1489 return;
1490 __ipv6_dev_ac_dec(ifp->idev, &addr);
1493 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1495 if (dev->addr_len != ETH_ALEN)
1496 return -1;
1497 memcpy(eui, dev->dev_addr, 3);
1498 memcpy(eui + 5, dev->dev_addr + 3, 3);
1501 * The zSeries OSA network cards can be shared among various
1502 * OS instances, but the OSA cards have only one MAC address.
1503 * This leads to duplicate address conflicts in conjunction
1504 * with IPv6 if more than one instance uses the same card.
1506 * The driver for these cards can deliver a unique 16-bit
1507 * identifier for each instance sharing the same card. It is
1508 * placed instead of 0xFFFE in the interface identifier. The
1509 * "u" bit of the interface identifier is not inverted in this
1510 * case. Hence the resulting interface identifier has local
1511 * scope according to RFC2373.
1513 if (dev->dev_id) {
1514 eui[3] = (dev->dev_id >> 8) & 0xFF;
1515 eui[4] = dev->dev_id & 0xFF;
1516 } else {
1517 eui[3] = 0xFF;
1518 eui[4] = 0xFE;
1519 eui[0] ^= 2;
1521 return 0;
1524 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1526 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1527 if (dev->addr_len != ARCNET_ALEN)
1528 return -1;
1529 memset(eui, 0, 7);
1530 eui[7] = *(u8*)dev->dev_addr;
1531 return 0;
1534 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1536 if (dev->addr_len != INFINIBAND_ALEN)
1537 return -1;
1538 memcpy(eui, dev->dev_addr + 12, 8);
1539 eui[0] |= 2;
1540 return 0;
1543 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1545 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1546 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1547 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1548 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1549 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1550 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1551 eui[1] = 0;
1552 eui[2] = 0x5E;
1553 eui[3] = 0xFE;
1554 memcpy(eui + 4, &addr, 4);
1555 return 0;
1557 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1559 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1561 if (dev->priv_flags & IFF_ISATAP)
1562 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563 return -1;
1566 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1568 switch (dev->type) {
1569 case ARPHRD_ETHER:
1570 case ARPHRD_FDDI:
1571 case ARPHRD_IEEE802_TR:
1572 return addrconf_ifid_eui48(eui, dev);
1573 case ARPHRD_ARCNET:
1574 return addrconf_ifid_arcnet(eui, dev);
1575 case ARPHRD_INFINIBAND:
1576 return addrconf_ifid_infiniband(eui, dev);
1577 case ARPHRD_SIT:
1578 return addrconf_ifid_sit(eui, dev);
1580 return -1;
1583 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1585 int err = -1;
1586 struct inet6_ifaddr *ifp;
1588 read_lock_bh(&idev->lock);
1589 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1590 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1591 memcpy(eui, ifp->addr.s6_addr+8, 8);
1592 err = 0;
1593 break;
1596 read_unlock_bh(&idev->lock);
1597 return err;
1600 #ifdef CONFIG_IPV6_PRIVACY
1601 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1602 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1604 regen:
1605 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1606 idev->rndid[0] &= ~0x02;
1609 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1610 * check if generated address is not inappropriate
1612 * - Reserved subnet anycast (RFC 2526)
1613 * 11111101 11....11 1xxxxxxx
1614 * - ISATAP (RFC4214) 6.1
1615 * 00-00-5E-FE-xx-xx-xx-xx
1616 * - value 0
1617 * - XXX: already assigned to an address on the device
1619 if (idev->rndid[0] == 0xfd &&
1620 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1621 (idev->rndid[7]&0x80))
1622 goto regen;
1623 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1624 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1625 goto regen;
1626 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627 goto regen;
1630 return 0;
1633 static void ipv6_regen_rndid(unsigned long data)
1635 struct inet6_dev *idev = (struct inet6_dev *) data;
1636 unsigned long expires;
1638 rcu_read_lock_bh();
1639 write_lock_bh(&idev->lock);
1641 if (idev->dead)
1642 goto out;
1644 if (__ipv6_regen_rndid(idev) < 0)
1645 goto out;
1647 expires = jiffies +
1648 idev->cnf.temp_prefered_lft * HZ -
1649 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1650 if (time_before(expires, jiffies)) {
1651 printk(KERN_WARNING
1652 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1653 idev->dev->name);
1654 goto out;
1657 if (!mod_timer(&idev->regen_timer, expires))
1658 in6_dev_hold(idev);
1660 out:
1661 write_unlock_bh(&idev->lock);
1662 rcu_read_unlock_bh();
1663 in6_dev_put(idev);
1666 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1667 int ret = 0;
1669 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1670 ret = __ipv6_regen_rndid(idev);
1671 return ret;
1673 #endif
1676 * Add prefix route.
1679 static void
1680 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1681 unsigned long expires, u32 flags)
1683 struct fib6_config cfg = {
1684 .fc_table = RT6_TABLE_PREFIX,
1685 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1686 .fc_ifindex = dev->ifindex,
1687 .fc_expires = expires,
1688 .fc_dst_len = plen,
1689 .fc_flags = RTF_UP | flags,
1690 .fc_nlinfo.nl_net = dev_net(dev),
1691 .fc_protocol = RTPROT_KERNEL,
1694 ipv6_addr_copy(&cfg.fc_dst, pfx);
1696 /* Prevent useless cloning on PtP SIT.
1697 This thing is done here expecting that the whole
1698 class of non-broadcast devices need not cloning.
1700 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1701 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1702 cfg.fc_flags |= RTF_NONEXTHOP;
1703 #endif
1705 ip6_route_add(&cfg);
1708 /* Create "default" multicast route to the interface */
1710 static void addrconf_add_mroute(struct net_device *dev)
1712 struct fib6_config cfg = {
1713 .fc_table = RT6_TABLE_LOCAL,
1714 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1715 .fc_ifindex = dev->ifindex,
1716 .fc_dst_len = 8,
1717 .fc_flags = RTF_UP,
1718 .fc_nlinfo.nl_net = dev_net(dev),
1721 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1723 ip6_route_add(&cfg);
1726 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1727 static void sit_route_add(struct net_device *dev)
1729 struct fib6_config cfg = {
1730 .fc_table = RT6_TABLE_MAIN,
1731 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1732 .fc_ifindex = dev->ifindex,
1733 .fc_dst_len = 96,
1734 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1735 .fc_nlinfo.nl_net = dev_net(dev),
1738 /* prefix length - 96 bits "::d.d.d.d" */
1739 ip6_route_add(&cfg);
1741 #endif
1743 static void addrconf_add_lroute(struct net_device *dev)
1745 struct in6_addr addr;
1747 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1748 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1751 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1753 struct inet6_dev *idev;
1755 ASSERT_RTNL();
1757 if ((idev = ipv6_find_idev(dev)) == NULL)
1758 return NULL;
1760 /* Add default multicast route */
1761 addrconf_add_mroute(dev);
1763 /* Add link local route */
1764 addrconf_add_lroute(dev);
1765 return idev;
1768 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1770 struct prefix_info *pinfo;
1771 __u32 valid_lft;
1772 __u32 prefered_lft;
1773 int addr_type;
1774 struct inet6_dev *in6_dev;
1776 pinfo = (struct prefix_info *) opt;
1778 if (len < sizeof(struct prefix_info)) {
1779 ADBG(("addrconf: prefix option too short\n"));
1780 return;
1784 * Validation checks ([ADDRCONF], page 19)
1787 addr_type = ipv6_addr_type(&pinfo->prefix);
1789 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1790 return;
1792 valid_lft = ntohl(pinfo->valid);
1793 prefered_lft = ntohl(pinfo->prefered);
1795 if (prefered_lft > valid_lft) {
1796 if (net_ratelimit())
1797 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1798 return;
1801 in6_dev = in6_dev_get(dev);
1803 if (in6_dev == NULL) {
1804 if (net_ratelimit())
1805 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1806 return;
1810 * Two things going on here:
1811 * 1) Add routes for on-link prefixes
1812 * 2) Configure prefixes with the auto flag set
1815 if (pinfo->onlink) {
1816 struct rt6_info *rt;
1817 unsigned long rt_expires;
1819 /* Avoid arithmetic overflow. Really, we could
1820 * save rt_expires in seconds, likely valid_lft,
1821 * but it would require division in fib gc, that it
1822 * not good.
1824 if (HZ > USER_HZ)
1825 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1826 else
1827 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1829 if (addrconf_finite_timeout(rt_expires))
1830 rt_expires *= HZ;
1832 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1833 dev->ifindex, 1);
1835 if (rt && addrconf_is_prefix_route(rt)) {
1836 /* Autoconf prefix route */
1837 if (valid_lft == 0) {
1838 ip6_del_rt(rt);
1839 rt = NULL;
1840 } else if (addrconf_finite_timeout(rt_expires)) {
1841 /* not infinity */
1842 rt->rt6i_expires = jiffies + rt_expires;
1843 rt->rt6i_flags |= RTF_EXPIRES;
1844 } else {
1845 rt->rt6i_flags &= ~RTF_EXPIRES;
1846 rt->rt6i_expires = 0;
1848 } else if (valid_lft) {
1849 clock_t expires = 0;
1850 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1851 if (addrconf_finite_timeout(rt_expires)) {
1852 /* not infinity */
1853 flags |= RTF_EXPIRES;
1854 expires = jiffies_to_clock_t(rt_expires);
1856 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1857 dev, expires, flags);
1859 if (rt)
1860 dst_release(&rt->u.dst);
1863 /* Try to figure out our local address for this prefix */
1865 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1866 struct inet6_ifaddr * ifp;
1867 struct in6_addr addr;
1868 int create = 0, update_lft = 0;
1869 struct net *net = dev_net(dev);
1871 if (pinfo->prefix_len == 64) {
1872 memcpy(&addr, &pinfo->prefix, 8);
1873 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1874 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1875 in6_dev_put(in6_dev);
1876 return;
1878 goto ok;
1880 if (net_ratelimit())
1881 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1882 pinfo->prefix_len);
1883 in6_dev_put(in6_dev);
1884 return;
1888 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1890 if (ifp == NULL && valid_lft) {
1891 int max_addresses = in6_dev->cnf.max_addresses;
1892 u32 addr_flags = 0;
1894 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1895 if (in6_dev->cnf.optimistic_dad &&
1896 !net->ipv6.devconf_all->forwarding)
1897 addr_flags = IFA_F_OPTIMISTIC;
1898 #endif
1900 /* Do not allow to create too much of autoconfigured
1901 * addresses; this would be too easy way to crash kernel.
1903 if (!max_addresses ||
1904 ipv6_count_addresses(in6_dev) < max_addresses)
1905 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1906 addr_type&IPV6_ADDR_SCOPE_MASK,
1907 addr_flags);
1909 if (!ifp || IS_ERR(ifp)) {
1910 in6_dev_put(in6_dev);
1911 return;
1914 update_lft = create = 1;
1915 ifp->cstamp = jiffies;
1916 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1919 if (ifp) {
1920 int flags;
1921 unsigned long now;
1922 #ifdef CONFIG_IPV6_PRIVACY
1923 struct inet6_ifaddr *ift;
1924 #endif
1925 u32 stored_lft;
1927 /* update lifetime (RFC2462 5.5.3 e) */
1928 spin_lock(&ifp->lock);
1929 now = jiffies;
1930 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1931 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1932 else
1933 stored_lft = 0;
1934 if (!update_lft && stored_lft) {
1935 if (valid_lft > MIN_VALID_LIFETIME ||
1936 valid_lft > stored_lft)
1937 update_lft = 1;
1938 else if (stored_lft <= MIN_VALID_LIFETIME) {
1939 /* valid_lft <= stored_lft is always true */
1940 /* XXX: IPsec */
1941 update_lft = 0;
1942 } else {
1943 valid_lft = MIN_VALID_LIFETIME;
1944 if (valid_lft < prefered_lft)
1945 prefered_lft = valid_lft;
1946 update_lft = 1;
1950 if (update_lft) {
1951 ifp->valid_lft = valid_lft;
1952 ifp->prefered_lft = prefered_lft;
1953 ifp->tstamp = now;
1954 flags = ifp->flags;
1955 ifp->flags &= ~IFA_F_DEPRECATED;
1956 spin_unlock(&ifp->lock);
1958 if (!(flags&IFA_F_TENTATIVE))
1959 ipv6_ifa_notify(0, ifp);
1960 } else
1961 spin_unlock(&ifp->lock);
1963 #ifdef CONFIG_IPV6_PRIVACY
1964 read_lock_bh(&in6_dev->lock);
1965 /* update all temporary addresses in the list */
1966 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1968 * When adjusting the lifetimes of an existing
1969 * temporary address, only lower the lifetimes.
1970 * Implementations must not increase the
1971 * lifetimes of an existing temporary address
1972 * when processing a Prefix Information Option.
1974 if (ifp != ift->ifpub)
1975 continue;
1977 spin_lock(&ift->lock);
1978 flags = ift->flags;
1979 if (ift->valid_lft > valid_lft &&
1980 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1981 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1982 if (ift->prefered_lft > prefered_lft &&
1983 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1984 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1985 spin_unlock(&ift->lock);
1986 if (!(flags&IFA_F_TENTATIVE))
1987 ipv6_ifa_notify(0, ift);
1990 if (create && in6_dev->cnf.use_tempaddr > 0) {
1992 * When a new public address is created as described in [ADDRCONF],
1993 * also create a new temporary address.
1995 read_unlock_bh(&in6_dev->lock);
1996 ipv6_create_tempaddr(ifp, NULL);
1997 } else {
1998 read_unlock_bh(&in6_dev->lock);
2000 #endif
2001 in6_ifa_put(ifp);
2002 addrconf_verify(0);
2005 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2006 in6_dev_put(in6_dev);
2010 * Set destination address.
2011 * Special case for SIT interfaces where we create a new "virtual"
2012 * device.
2014 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2016 struct in6_ifreq ireq;
2017 struct net_device *dev;
2018 int err = -EINVAL;
2020 rtnl_lock();
2022 err = -EFAULT;
2023 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2024 goto err_exit;
2026 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2028 err = -ENODEV;
2029 if (dev == NULL)
2030 goto err_exit;
2032 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2033 if (dev->type == ARPHRD_SIT) {
2034 struct ifreq ifr;
2035 mm_segment_t oldfs;
2036 struct ip_tunnel_parm p;
2038 err = -EADDRNOTAVAIL;
2039 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2040 goto err_exit;
2042 memset(&p, 0, sizeof(p));
2043 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2044 p.iph.saddr = 0;
2045 p.iph.version = 4;
2046 p.iph.ihl = 5;
2047 p.iph.protocol = IPPROTO_IPV6;
2048 p.iph.ttl = 64;
2049 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2051 oldfs = get_fs(); set_fs(KERNEL_DS);
2052 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2053 set_fs(oldfs);
2055 if (err == 0) {
2056 err = -ENOBUFS;
2057 dev = __dev_get_by_name(net, p.name);
2058 if (!dev)
2059 goto err_exit;
2060 err = dev_open(dev);
2063 #endif
2065 err_exit:
2066 rtnl_unlock();
2067 return err;
2071 * Manual configuration of address on an interface
2073 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2074 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2075 __u32 valid_lft)
2077 struct inet6_ifaddr *ifp;
2078 struct inet6_dev *idev;
2079 struct net_device *dev;
2080 int scope;
2081 u32 flags;
2082 clock_t expires;
2083 unsigned long timeout;
2085 ASSERT_RTNL();
2087 if (plen > 128)
2088 return -EINVAL;
2090 /* check the lifetime */
2091 if (!valid_lft || prefered_lft > valid_lft)
2092 return -EINVAL;
2094 dev = __dev_get_by_index(net, ifindex);
2095 if (!dev)
2096 return -ENODEV;
2098 if ((idev = addrconf_add_dev(dev)) == NULL)
2099 return -ENOBUFS;
2101 scope = ipv6_addr_scope(pfx);
2103 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2104 if (addrconf_finite_timeout(timeout)) {
2105 expires = jiffies_to_clock_t(timeout * HZ);
2106 valid_lft = timeout;
2107 flags = RTF_EXPIRES;
2108 } else {
2109 expires = 0;
2110 flags = 0;
2111 ifa_flags |= IFA_F_PERMANENT;
2114 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2115 if (addrconf_finite_timeout(timeout)) {
2116 if (timeout == 0)
2117 ifa_flags |= IFA_F_DEPRECATED;
2118 prefered_lft = timeout;
2121 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2123 if (!IS_ERR(ifp)) {
2124 spin_lock_bh(&ifp->lock);
2125 ifp->valid_lft = valid_lft;
2126 ifp->prefered_lft = prefered_lft;
2127 ifp->tstamp = jiffies;
2128 spin_unlock_bh(&ifp->lock);
2130 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2131 expires, flags);
2133 * Note that section 3.1 of RFC 4429 indicates
2134 * that the Optimistic flag should not be set for
2135 * manually configured addresses
2137 addrconf_dad_start(ifp, 0);
2138 in6_ifa_put(ifp);
2139 addrconf_verify(0);
2140 return 0;
2143 return PTR_ERR(ifp);
2146 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2147 unsigned int plen)
2149 struct inet6_ifaddr *ifp;
2150 struct inet6_dev *idev;
2151 struct net_device *dev;
2153 if (plen > 128)
2154 return -EINVAL;
2156 dev = __dev_get_by_index(net, ifindex);
2157 if (!dev)
2158 return -ENODEV;
2160 if ((idev = __in6_dev_get(dev)) == NULL)
2161 return -ENXIO;
2163 read_lock_bh(&idev->lock);
2164 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2165 if (ifp->prefix_len == plen &&
2166 ipv6_addr_equal(pfx, &ifp->addr)) {
2167 in6_ifa_hold(ifp);
2168 read_unlock_bh(&idev->lock);
2170 ipv6_del_addr(ifp);
2172 /* If the last address is deleted administratively,
2173 disable IPv6 on this interface.
2175 if (idev->addr_list == NULL)
2176 addrconf_ifdown(idev->dev, 1);
2177 return 0;
2180 read_unlock_bh(&idev->lock);
2181 return -EADDRNOTAVAIL;
2185 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2187 struct in6_ifreq ireq;
2188 int err;
2190 if (!capable(CAP_NET_ADMIN))
2191 return -EPERM;
2193 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2194 return -EFAULT;
2196 rtnl_lock();
2197 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2198 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2199 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2200 rtnl_unlock();
2201 return err;
2204 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2206 struct in6_ifreq ireq;
2207 int err;
2209 if (!capable(CAP_NET_ADMIN))
2210 return -EPERM;
2212 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2213 return -EFAULT;
2215 rtnl_lock();
2216 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2217 ireq.ifr6_prefixlen);
2218 rtnl_unlock();
2219 return err;
2222 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2223 static void sit_add_v4_addrs(struct inet6_dev *idev)
2225 struct inet6_ifaddr * ifp;
2226 struct in6_addr addr;
2227 struct net_device *dev;
2228 struct net *net = dev_net(idev->dev);
2229 int scope;
2231 ASSERT_RTNL();
2233 memset(&addr, 0, sizeof(struct in6_addr));
2234 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2236 if (idev->dev->flags&IFF_POINTOPOINT) {
2237 addr.s6_addr32[0] = htonl(0xfe800000);
2238 scope = IFA_LINK;
2239 } else {
2240 scope = IPV6_ADDR_COMPATv4;
2243 if (addr.s6_addr32[3]) {
2244 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2245 if (!IS_ERR(ifp)) {
2246 spin_lock_bh(&ifp->lock);
2247 ifp->flags &= ~IFA_F_TENTATIVE;
2248 spin_unlock_bh(&ifp->lock);
2249 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2250 in6_ifa_put(ifp);
2252 return;
2255 for_each_netdev(net, dev) {
2256 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2257 if (in_dev && (dev->flags & IFF_UP)) {
2258 struct in_ifaddr * ifa;
2260 int flag = scope;
2262 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2263 int plen;
2265 addr.s6_addr32[3] = ifa->ifa_local;
2267 if (ifa->ifa_scope == RT_SCOPE_LINK)
2268 continue;
2269 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2270 if (idev->dev->flags&IFF_POINTOPOINT)
2271 continue;
2272 flag |= IFA_HOST;
2274 if (idev->dev->flags&IFF_POINTOPOINT)
2275 plen = 64;
2276 else
2277 plen = 96;
2279 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2280 IFA_F_PERMANENT);
2281 if (!IS_ERR(ifp)) {
2282 spin_lock_bh(&ifp->lock);
2283 ifp->flags &= ~IFA_F_TENTATIVE;
2284 spin_unlock_bh(&ifp->lock);
2285 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2286 in6_ifa_put(ifp);
2292 #endif
2294 static void init_loopback(struct net_device *dev)
2296 struct inet6_dev *idev;
2297 struct inet6_ifaddr * ifp;
2299 /* ::1 */
2301 ASSERT_RTNL();
2303 if ((idev = ipv6_find_idev(dev)) == NULL) {
2304 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2305 return;
2308 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2309 if (!IS_ERR(ifp)) {
2310 spin_lock_bh(&ifp->lock);
2311 ifp->flags &= ~IFA_F_TENTATIVE;
2312 spin_unlock_bh(&ifp->lock);
2313 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2314 in6_ifa_put(ifp);
2318 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2320 struct inet6_ifaddr * ifp;
2321 u32 addr_flags = IFA_F_PERMANENT;
2323 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2324 if (idev->cnf.optimistic_dad &&
2325 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2326 addr_flags |= IFA_F_OPTIMISTIC;
2327 #endif
2330 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2331 if (!IS_ERR(ifp)) {
2332 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2333 addrconf_dad_start(ifp, 0);
2334 in6_ifa_put(ifp);
2338 static void addrconf_dev_config(struct net_device *dev)
2340 struct in6_addr addr;
2341 struct inet6_dev * idev;
2343 ASSERT_RTNL();
2345 if ((dev->type != ARPHRD_ETHER) &&
2346 (dev->type != ARPHRD_FDDI) &&
2347 (dev->type != ARPHRD_IEEE802_TR) &&
2348 (dev->type != ARPHRD_ARCNET) &&
2349 (dev->type != ARPHRD_INFINIBAND)) {
2350 /* Alas, we support only Ethernet autoconfiguration. */
2351 return;
2354 idev = addrconf_add_dev(dev);
2355 if (idev == NULL)
2356 return;
2358 memset(&addr, 0, sizeof(struct in6_addr));
2359 addr.s6_addr32[0] = htonl(0xFE800000);
2361 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2362 addrconf_add_linklocal(idev, &addr);
2365 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2366 static void addrconf_sit_config(struct net_device *dev)
2368 struct inet6_dev *idev;
2370 ASSERT_RTNL();
2373 * Configure the tunnel with one of our IPv4
2374 * addresses... we should configure all of
2375 * our v4 addrs in the tunnel
2378 if ((idev = ipv6_find_idev(dev)) == NULL) {
2379 printk(KERN_DEBUG "init sit: add_dev failed\n");
2380 return;
2383 if (dev->priv_flags & IFF_ISATAP) {
2384 struct in6_addr addr;
2386 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2387 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2388 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2389 addrconf_add_linklocal(idev, &addr);
2390 return;
2393 sit_add_v4_addrs(idev);
2395 if (dev->flags&IFF_POINTOPOINT) {
2396 addrconf_add_mroute(dev);
2397 addrconf_add_lroute(dev);
2398 } else
2399 sit_route_add(dev);
2401 #endif
2403 static inline int
2404 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2406 struct in6_addr lladdr;
2408 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2409 addrconf_add_linklocal(idev, &lladdr);
2410 return 0;
2412 return -1;
2415 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2417 struct net_device *link_dev;
2418 struct net *net = dev_net(idev->dev);
2420 /* first try to inherit the link-local address from the link device */
2421 if (idev->dev->iflink &&
2422 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2423 if (!ipv6_inherit_linklocal(idev, link_dev))
2424 return;
2426 /* then try to inherit it from any device */
2427 for_each_netdev(net, link_dev) {
2428 if (!ipv6_inherit_linklocal(idev, link_dev))
2429 return;
2431 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2435 * Autoconfigure tunnel with a link-local address so routing protocols,
2436 * DHCPv6, MLD etc. can be run over the virtual link
2439 static void addrconf_ip6_tnl_config(struct net_device *dev)
2441 struct inet6_dev *idev;
2443 ASSERT_RTNL();
2445 if ((idev = addrconf_add_dev(dev)) == NULL) {
2446 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2447 return;
2449 ip6_tnl_add_linklocal(idev);
2452 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2453 void * data)
2455 struct net_device *dev = (struct net_device *) data;
2456 struct inet6_dev *idev = __in6_dev_get(dev);
2457 int run_pending = 0;
2458 int err;
2460 switch(event) {
2461 case NETDEV_REGISTER:
2462 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2463 idev = ipv6_add_dev(dev);
2464 if (!idev)
2465 return notifier_from_errno(-ENOMEM);
2467 break;
2468 case NETDEV_UP:
2469 case NETDEV_CHANGE:
2470 if (dev->flags & IFF_SLAVE)
2471 break;
2473 if (event == NETDEV_UP) {
2474 if (!addrconf_qdisc_ok(dev)) {
2475 /* device is not ready yet. */
2476 printk(KERN_INFO
2477 "ADDRCONF(NETDEV_UP): %s: "
2478 "link is not ready\n",
2479 dev->name);
2480 break;
2483 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2484 idev = ipv6_add_dev(dev);
2486 if (idev)
2487 idev->if_flags |= IF_READY;
2488 } else {
2489 if (!addrconf_qdisc_ok(dev)) {
2490 /* device is still not ready. */
2491 break;
2494 if (idev) {
2495 if (idev->if_flags & IF_READY) {
2496 /* device is already configured. */
2497 break;
2499 idev->if_flags |= IF_READY;
2502 printk(KERN_INFO
2503 "ADDRCONF(NETDEV_CHANGE): %s: "
2504 "link becomes ready\n",
2505 dev->name);
2507 run_pending = 1;
2510 switch(dev->type) {
2511 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2512 case ARPHRD_SIT:
2513 addrconf_sit_config(dev);
2514 break;
2515 #endif
2516 case ARPHRD_TUNNEL6:
2517 addrconf_ip6_tnl_config(dev);
2518 break;
2519 case ARPHRD_LOOPBACK:
2520 init_loopback(dev);
2521 break;
2523 default:
2524 addrconf_dev_config(dev);
2525 break;
2527 if (idev) {
2528 if (run_pending)
2529 addrconf_dad_run(idev);
2531 /* If the MTU changed during the interface down, when the
2532 interface up, the changed MTU must be reflected in the
2533 idev as well as routers.
2535 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2536 rt6_mtu_change(dev, dev->mtu);
2537 idev->cnf.mtu6 = dev->mtu;
2539 idev->tstamp = jiffies;
2540 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2541 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2542 stop IPv6 on this interface.
2544 if (dev->mtu < IPV6_MIN_MTU)
2545 addrconf_ifdown(dev, event != NETDEV_DOWN);
2547 break;
2549 case NETDEV_CHANGEMTU:
2550 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2551 rt6_mtu_change(dev, dev->mtu);
2552 idev->cnf.mtu6 = dev->mtu;
2553 break;
2556 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2557 idev = ipv6_add_dev(dev);
2558 if (idev)
2559 break;
2562 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2564 case NETDEV_DOWN:
2565 case NETDEV_UNREGISTER:
2567 * Remove all addresses from this interface.
2569 addrconf_ifdown(dev, event != NETDEV_DOWN);
2570 break;
2572 case NETDEV_CHANGENAME:
2573 if (idev) {
2574 snmp6_unregister_dev(idev);
2575 addrconf_sysctl_unregister(idev);
2576 addrconf_sysctl_register(idev);
2577 err = snmp6_register_dev(idev);
2578 if (err)
2579 return notifier_from_errno(err);
2581 break;
2584 return NOTIFY_OK;
2588 * addrconf module should be notified of a device going up
2590 static struct notifier_block ipv6_dev_notf = {
2591 .notifier_call = addrconf_notify,
2592 .priority = 0
2595 static int addrconf_ifdown(struct net_device *dev, int how)
2597 struct inet6_dev *idev;
2598 struct inet6_ifaddr *ifa, **bifa;
2599 struct net *net = dev_net(dev);
2600 int i;
2602 ASSERT_RTNL();
2604 if ((dev->flags & IFF_LOOPBACK) && how == 1)
2605 how = 0;
2607 rt6_ifdown(net, dev);
2608 neigh_ifdown(&nd_tbl, dev);
2610 idev = __in6_dev_get(dev);
2611 if (idev == NULL)
2612 return -ENODEV;
2614 /* Step 1: remove reference to ipv6 device from parent device.
2615 Do not dev_put!
2617 if (how) {
2618 idev->dead = 1;
2620 /* protected by rtnl_lock */
2621 rcu_assign_pointer(dev->ip6_ptr, NULL);
2623 /* Step 1.5: remove snmp6 entry */
2624 snmp6_unregister_dev(idev);
2628 /* Step 2: clear hash table */
2629 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2630 bifa = &inet6_addr_lst[i];
2632 write_lock_bh(&addrconf_hash_lock);
2633 while ((ifa = *bifa) != NULL) {
2634 if (ifa->idev == idev) {
2635 *bifa = ifa->lst_next;
2636 ifa->lst_next = NULL;
2637 addrconf_del_timer(ifa);
2638 in6_ifa_put(ifa);
2639 continue;
2641 bifa = &ifa->lst_next;
2643 write_unlock_bh(&addrconf_hash_lock);
2646 write_lock_bh(&idev->lock);
2648 /* Step 3: clear flags for stateless addrconf */
2649 if (!how)
2650 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2652 /* Step 4: clear address list */
2653 #ifdef CONFIG_IPV6_PRIVACY
2654 if (how && del_timer(&idev->regen_timer))
2655 in6_dev_put(idev);
2657 /* clear tempaddr list */
2658 while ((ifa = idev->tempaddr_list) != NULL) {
2659 idev->tempaddr_list = ifa->tmp_next;
2660 ifa->tmp_next = NULL;
2661 ifa->dead = 1;
2662 write_unlock_bh(&idev->lock);
2663 spin_lock_bh(&ifa->lock);
2665 if (ifa->ifpub) {
2666 in6_ifa_put(ifa->ifpub);
2667 ifa->ifpub = NULL;
2669 spin_unlock_bh(&ifa->lock);
2670 in6_ifa_put(ifa);
2671 write_lock_bh(&idev->lock);
2673 #endif
2674 while ((ifa = idev->addr_list) != NULL) {
2675 idev->addr_list = ifa->if_next;
2676 ifa->if_next = NULL;
2677 ifa->dead = 1;
2678 addrconf_del_timer(ifa);
2679 write_unlock_bh(&idev->lock);
2681 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2682 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2683 in6_ifa_put(ifa);
2685 write_lock_bh(&idev->lock);
2687 write_unlock_bh(&idev->lock);
2689 /* Step 5: Discard multicast list */
2691 if (how)
2692 ipv6_mc_destroy_dev(idev);
2693 else
2694 ipv6_mc_down(idev);
2696 idev->tstamp = jiffies;
2698 /* Shot the device (if unregistered) */
2700 if (how) {
2701 addrconf_sysctl_unregister(idev);
2702 neigh_parms_release(&nd_tbl, idev->nd_parms);
2703 neigh_ifdown(&nd_tbl, dev);
2704 in6_dev_put(idev);
2706 return 0;
2709 static void addrconf_rs_timer(unsigned long data)
2711 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2713 if (ifp->idev->cnf.forwarding)
2714 goto out;
2716 if (ifp->idev->if_flags & IF_RA_RCVD) {
2718 * Announcement received after solicitation
2719 * was sent
2721 goto out;
2724 spin_lock(&ifp->lock);
2725 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2726 /* The wait after the last probe can be shorter */
2727 addrconf_mod_timer(ifp, AC_RS,
2728 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2729 ifp->idev->cnf.rtr_solicit_delay :
2730 ifp->idev->cnf.rtr_solicit_interval);
2731 spin_unlock(&ifp->lock);
2733 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2734 } else {
2735 spin_unlock(&ifp->lock);
2737 * Note: we do not support deprecated "all on-link"
2738 * assumption any longer.
2740 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2741 ifp->idev->dev->name);
2744 out:
2745 in6_ifa_put(ifp);
2749 * Duplicate Address Detection
2751 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2753 unsigned long rand_num;
2754 struct inet6_dev *idev = ifp->idev;
2756 if (ifp->flags & IFA_F_OPTIMISTIC)
2757 rand_num = 0;
2758 else
2759 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2761 ifp->probes = idev->cnf.dad_transmits;
2762 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2765 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2767 struct inet6_dev *idev = ifp->idev;
2768 struct net_device *dev = idev->dev;
2770 addrconf_join_solict(dev, &ifp->addr);
2772 net_srandom(ifp->addr.s6_addr32[3]);
2774 read_lock_bh(&idev->lock);
2775 if (ifp->dead)
2776 goto out;
2777 spin_lock_bh(&ifp->lock);
2779 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2780 idev->cnf.accept_dad < 1 ||
2781 !(ifp->flags&IFA_F_TENTATIVE) ||
2782 ifp->flags & IFA_F_NODAD) {
2783 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2784 spin_unlock_bh(&ifp->lock);
2785 read_unlock_bh(&idev->lock);
2787 addrconf_dad_completed(ifp);
2788 return;
2791 if (!(idev->if_flags & IF_READY)) {
2792 spin_unlock_bh(&ifp->lock);
2793 read_unlock_bh(&idev->lock);
2795 * If the defice is not ready:
2796 * - keep it tentative if it is a permanent address.
2797 * - otherwise, kill it.
2799 in6_ifa_hold(ifp);
2800 addrconf_dad_stop(ifp);
2801 return;
2805 * Optimistic nodes can start receiving
2806 * Frames right away
2808 if(ifp->flags & IFA_F_OPTIMISTIC)
2809 ip6_ins_rt(ifp->rt);
2811 addrconf_dad_kick(ifp);
2812 spin_unlock_bh(&ifp->lock);
2813 out:
2814 read_unlock_bh(&idev->lock);
2817 static void addrconf_dad_timer(unsigned long data)
2819 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2820 struct inet6_dev *idev = ifp->idev;
2821 struct in6_addr mcaddr;
2823 read_lock_bh(&idev->lock);
2824 if (idev->dead) {
2825 read_unlock_bh(&idev->lock);
2826 goto out;
2828 if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2829 read_unlock_bh(&idev->lock);
2830 addrconf_dad_failure(ifp);
2831 return;
2833 spin_lock_bh(&ifp->lock);
2834 if (ifp->probes == 0) {
2836 * DAD was successful
2839 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2840 spin_unlock_bh(&ifp->lock);
2841 read_unlock_bh(&idev->lock);
2843 addrconf_dad_completed(ifp);
2845 goto out;
2848 ifp->probes--;
2849 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2850 spin_unlock_bh(&ifp->lock);
2851 read_unlock_bh(&idev->lock);
2853 /* send a neighbour solicitation for our addr */
2854 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2855 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2856 out:
2857 in6_ifa_put(ifp);
2860 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2862 struct net_device * dev = ifp->idev->dev;
2865 * Configure the address for reception. Now it is valid.
2868 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2870 /* If added prefix is link local and forwarding is off,
2871 start sending router solicitations.
2874 if (ifp->idev->cnf.forwarding == 0 &&
2875 ifp->idev->cnf.rtr_solicits > 0 &&
2876 (dev->flags&IFF_LOOPBACK) == 0 &&
2877 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2879 * If a host as already performed a random delay
2880 * [...] as part of DAD [...] there is no need
2881 * to delay again before sending the first RS
2883 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2885 spin_lock_bh(&ifp->lock);
2886 ifp->probes = 1;
2887 ifp->idev->if_flags |= IF_RS_SENT;
2888 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2889 spin_unlock_bh(&ifp->lock);
2893 static void addrconf_dad_run(struct inet6_dev *idev) {
2894 struct inet6_ifaddr *ifp;
2896 read_lock_bh(&idev->lock);
2897 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2898 spin_lock_bh(&ifp->lock);
2899 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2900 spin_unlock_bh(&ifp->lock);
2901 continue;
2903 spin_unlock_bh(&ifp->lock);
2904 addrconf_dad_kick(ifp);
2906 read_unlock_bh(&idev->lock);
2909 #ifdef CONFIG_PROC_FS
2910 struct if6_iter_state {
2911 struct seq_net_private p;
2912 int bucket;
2915 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2917 struct inet6_ifaddr *ifa = NULL;
2918 struct if6_iter_state *state = seq->private;
2919 struct net *net = seq_file_net(seq);
2921 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2922 ifa = inet6_addr_lst[state->bucket];
2924 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2925 ifa = ifa->lst_next;
2926 if (ifa)
2927 break;
2929 return ifa;
2932 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2934 struct if6_iter_state *state = seq->private;
2935 struct net *net = seq_file_net(seq);
2937 ifa = ifa->lst_next;
2938 try_again:
2939 if (ifa) {
2940 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2941 ifa = ifa->lst_next;
2942 goto try_again;
2946 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2947 ifa = inet6_addr_lst[state->bucket];
2948 goto try_again;
2951 return ifa;
2954 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2956 struct inet6_ifaddr *ifa = if6_get_first(seq);
2958 if (ifa)
2959 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2960 --pos;
2961 return pos ? NULL : ifa;
2964 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2965 __acquires(addrconf_hash_lock)
2967 read_lock_bh(&addrconf_hash_lock);
2968 return if6_get_idx(seq, *pos);
2971 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2973 struct inet6_ifaddr *ifa;
2975 ifa = if6_get_next(seq, v);
2976 ++*pos;
2977 return ifa;
2980 static void if6_seq_stop(struct seq_file *seq, void *v)
2981 __releases(addrconf_hash_lock)
2983 read_unlock_bh(&addrconf_hash_lock);
2986 static int if6_seq_show(struct seq_file *seq, void *v)
2988 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2989 seq_printf(seq,
2990 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2991 NIP6(ifp->addr),
2992 ifp->idev->dev->ifindex,
2993 ifp->prefix_len,
2994 ifp->scope,
2995 ifp->flags,
2996 ifp->idev->dev->name);
2997 return 0;
3000 static const struct seq_operations if6_seq_ops = {
3001 .start = if6_seq_start,
3002 .next = if6_seq_next,
3003 .show = if6_seq_show,
3004 .stop = if6_seq_stop,
3007 static int if6_seq_open(struct inode *inode, struct file *file)
3009 return seq_open_net(inode, file, &if6_seq_ops,
3010 sizeof(struct if6_iter_state));
3013 static const struct file_operations if6_fops = {
3014 .owner = THIS_MODULE,
3015 .open = if6_seq_open,
3016 .read = seq_read,
3017 .llseek = seq_lseek,
3018 .release = seq_release_net,
3021 static int if6_proc_net_init(struct net *net)
3023 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3024 return -ENOMEM;
3025 return 0;
3028 static void if6_proc_net_exit(struct net *net)
3030 proc_net_remove(net, "if_inet6");
3033 static struct pernet_operations if6_proc_net_ops = {
3034 .init = if6_proc_net_init,
3035 .exit = if6_proc_net_exit,
3038 int __init if6_proc_init(void)
3040 return register_pernet_subsys(&if6_proc_net_ops);
3043 void if6_proc_exit(void)
3045 unregister_pernet_subsys(&if6_proc_net_ops);
3047 #endif /* CONFIG_PROC_FS */
3049 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3050 /* Check if address is a home address configured on any interface. */
3051 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3053 int ret = 0;
3054 struct inet6_ifaddr * ifp;
3055 u8 hash = ipv6_addr_hash(addr);
3056 read_lock_bh(&addrconf_hash_lock);
3057 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3058 if (!net_eq(dev_net(ifp->idev->dev), net))
3059 continue;
3060 if (ipv6_addr_equal(&ifp->addr, addr) &&
3061 (ifp->flags & IFA_F_HOMEADDRESS)) {
3062 ret = 1;
3063 break;
3066 read_unlock_bh(&addrconf_hash_lock);
3067 return ret;
3069 #endif
3072 * Periodic address status verification
3075 static void addrconf_verify(unsigned long foo)
3077 struct inet6_ifaddr *ifp;
3078 unsigned long now, next;
3079 int i;
3081 spin_lock_bh(&addrconf_verify_lock);
3082 now = jiffies;
3083 next = now + ADDR_CHECK_FREQUENCY;
3085 del_timer(&addr_chk_timer);
3087 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3089 restart:
3090 read_lock(&addrconf_hash_lock);
3091 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3092 unsigned long age;
3093 #ifdef CONFIG_IPV6_PRIVACY
3094 unsigned long regen_advance;
3095 #endif
3097 if (ifp->flags & IFA_F_PERMANENT)
3098 continue;
3100 spin_lock(&ifp->lock);
3101 age = (now - ifp->tstamp) / HZ;
3103 #ifdef CONFIG_IPV6_PRIVACY
3104 regen_advance = ifp->idev->cnf.regen_max_retry *
3105 ifp->idev->cnf.dad_transmits *
3106 ifp->idev->nd_parms->retrans_time / HZ;
3107 #endif
3109 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3110 age >= ifp->valid_lft) {
3111 spin_unlock(&ifp->lock);
3112 in6_ifa_hold(ifp);
3113 read_unlock(&addrconf_hash_lock);
3114 ipv6_del_addr(ifp);
3115 goto restart;
3116 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3117 spin_unlock(&ifp->lock);
3118 continue;
3119 } else if (age >= ifp->prefered_lft) {
3120 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3121 int deprecate = 0;
3123 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3124 deprecate = 1;
3125 ifp->flags |= IFA_F_DEPRECATED;
3128 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3129 next = ifp->tstamp + ifp->valid_lft * HZ;
3131 spin_unlock(&ifp->lock);
3133 if (deprecate) {
3134 in6_ifa_hold(ifp);
3135 read_unlock(&addrconf_hash_lock);
3137 ipv6_ifa_notify(0, ifp);
3138 in6_ifa_put(ifp);
3139 goto restart;
3141 #ifdef CONFIG_IPV6_PRIVACY
3142 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3143 !(ifp->flags&IFA_F_TENTATIVE)) {
3144 if (age >= ifp->prefered_lft - regen_advance) {
3145 struct inet6_ifaddr *ifpub = ifp->ifpub;
3146 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3147 next = ifp->tstamp + ifp->prefered_lft * HZ;
3148 if (!ifp->regen_count && ifpub) {
3149 ifp->regen_count++;
3150 in6_ifa_hold(ifp);
3151 in6_ifa_hold(ifpub);
3152 spin_unlock(&ifp->lock);
3153 read_unlock(&addrconf_hash_lock);
3154 spin_lock(&ifpub->lock);
3155 ifpub->regen_count = 0;
3156 spin_unlock(&ifpub->lock);
3157 ipv6_create_tempaddr(ifpub, ifp);
3158 in6_ifa_put(ifpub);
3159 in6_ifa_put(ifp);
3160 goto restart;
3162 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3163 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3164 spin_unlock(&ifp->lock);
3165 #endif
3166 } else {
3167 /* ifp->prefered_lft <= ifp->valid_lft */
3168 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3169 next = ifp->tstamp + ifp->prefered_lft * HZ;
3170 spin_unlock(&ifp->lock);
3173 read_unlock(&addrconf_hash_lock);
3176 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3177 add_timer(&addr_chk_timer);
3178 spin_unlock_bh(&addrconf_verify_lock);
3181 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3183 struct in6_addr *pfx = NULL;
3185 if (addr)
3186 pfx = nla_data(addr);
3188 if (local) {
3189 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3190 pfx = NULL;
3191 else
3192 pfx = nla_data(local);
3195 return pfx;
3198 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3199 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3200 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3201 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3204 static int
3205 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3207 struct net *net = sock_net(skb->sk);
3208 struct ifaddrmsg *ifm;
3209 struct nlattr *tb[IFA_MAX+1];
3210 struct in6_addr *pfx;
3211 int err;
3213 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3214 if (err < 0)
3215 return err;
3217 ifm = nlmsg_data(nlh);
3218 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3219 if (pfx == NULL)
3220 return -EINVAL;
3222 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3225 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3226 u32 prefered_lft, u32 valid_lft)
3228 u32 flags;
3229 clock_t expires;
3230 unsigned long timeout;
3232 if (!valid_lft || (prefered_lft > valid_lft))
3233 return -EINVAL;
3235 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3236 if (addrconf_finite_timeout(timeout)) {
3237 expires = jiffies_to_clock_t(timeout * HZ);
3238 valid_lft = timeout;
3239 flags = RTF_EXPIRES;
3240 } else {
3241 expires = 0;
3242 flags = 0;
3243 ifa_flags |= IFA_F_PERMANENT;
3246 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3247 if (addrconf_finite_timeout(timeout)) {
3248 if (timeout == 0)
3249 ifa_flags |= IFA_F_DEPRECATED;
3250 prefered_lft = timeout;
3253 spin_lock_bh(&ifp->lock);
3254 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3255 ifp->tstamp = jiffies;
3256 ifp->valid_lft = valid_lft;
3257 ifp->prefered_lft = prefered_lft;
3259 spin_unlock_bh(&ifp->lock);
3260 if (!(ifp->flags&IFA_F_TENTATIVE))
3261 ipv6_ifa_notify(0, ifp);
3263 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3264 expires, flags);
3265 addrconf_verify(0);
3267 return 0;
3270 static int
3271 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3273 struct net *net = sock_net(skb->sk);
3274 struct ifaddrmsg *ifm;
3275 struct nlattr *tb[IFA_MAX+1];
3276 struct in6_addr *pfx;
3277 struct inet6_ifaddr *ifa;
3278 struct net_device *dev;
3279 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3280 u8 ifa_flags;
3281 int err;
3283 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3284 if (err < 0)
3285 return err;
3287 ifm = nlmsg_data(nlh);
3288 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3289 if (pfx == NULL)
3290 return -EINVAL;
3292 if (tb[IFA_CACHEINFO]) {
3293 struct ifa_cacheinfo *ci;
3295 ci = nla_data(tb[IFA_CACHEINFO]);
3296 valid_lft = ci->ifa_valid;
3297 preferred_lft = ci->ifa_prefered;
3298 } else {
3299 preferred_lft = INFINITY_LIFE_TIME;
3300 valid_lft = INFINITY_LIFE_TIME;
3303 dev = __dev_get_by_index(net, ifm->ifa_index);
3304 if (dev == NULL)
3305 return -ENODEV;
3307 /* We ignore other flags so far. */
3308 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3310 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3311 if (ifa == NULL) {
3313 * It would be best to check for !NLM_F_CREATE here but
3314 * userspace alreay relies on not having to provide this.
3316 return inet6_addr_add(net, ifm->ifa_index, pfx,
3317 ifm->ifa_prefixlen, ifa_flags,
3318 preferred_lft, valid_lft);
3321 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3322 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3323 err = -EEXIST;
3324 else
3325 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3327 in6_ifa_put(ifa);
3329 return err;
3332 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3333 u8 scope, int ifindex)
3335 struct ifaddrmsg *ifm;
3337 ifm = nlmsg_data(nlh);
3338 ifm->ifa_family = AF_INET6;
3339 ifm->ifa_prefixlen = prefixlen;
3340 ifm->ifa_flags = flags;
3341 ifm->ifa_scope = scope;
3342 ifm->ifa_index = ifindex;
3345 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3346 unsigned long tstamp, u32 preferred, u32 valid)
3348 struct ifa_cacheinfo ci;
3350 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3351 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3352 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3353 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3354 ci.ifa_prefered = preferred;
3355 ci.ifa_valid = valid;
3357 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3360 static inline int rt_scope(int ifa_scope)
3362 if (ifa_scope & IFA_HOST)
3363 return RT_SCOPE_HOST;
3364 else if (ifa_scope & IFA_LINK)
3365 return RT_SCOPE_LINK;
3366 else if (ifa_scope & IFA_SITE)
3367 return RT_SCOPE_SITE;
3368 else
3369 return RT_SCOPE_UNIVERSE;
3372 static inline int inet6_ifaddr_msgsize(void)
3374 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3375 + nla_total_size(16) /* IFA_ADDRESS */
3376 + nla_total_size(sizeof(struct ifa_cacheinfo));
3379 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3380 u32 pid, u32 seq, int event, unsigned int flags)
3382 struct nlmsghdr *nlh;
3383 u32 preferred, valid;
3385 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3386 if (nlh == NULL)
3387 return -EMSGSIZE;
3389 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3390 ifa->idev->dev->ifindex);
3392 if (!(ifa->flags&IFA_F_PERMANENT)) {
3393 preferred = ifa->prefered_lft;
3394 valid = ifa->valid_lft;
3395 if (preferred != INFINITY_LIFE_TIME) {
3396 long tval = (jiffies - ifa->tstamp)/HZ;
3397 preferred -= tval;
3398 if (valid != INFINITY_LIFE_TIME)
3399 valid -= tval;
3401 } else {
3402 preferred = INFINITY_LIFE_TIME;
3403 valid = INFINITY_LIFE_TIME;
3406 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3407 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3408 nlmsg_cancel(skb, nlh);
3409 return -EMSGSIZE;
3412 return nlmsg_end(skb, nlh);
3415 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3416 u32 pid, u32 seq, int event, u16 flags)
3418 struct nlmsghdr *nlh;
3419 u8 scope = RT_SCOPE_UNIVERSE;
3420 int ifindex = ifmca->idev->dev->ifindex;
3422 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3423 scope = RT_SCOPE_SITE;
3425 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3426 if (nlh == NULL)
3427 return -EMSGSIZE;
3429 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3430 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3431 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3432 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3433 nlmsg_cancel(skb, nlh);
3434 return -EMSGSIZE;
3437 return nlmsg_end(skb, nlh);
3440 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3441 u32 pid, u32 seq, int event, unsigned int flags)
3443 struct nlmsghdr *nlh;
3444 u8 scope = RT_SCOPE_UNIVERSE;
3445 int ifindex = ifaca->aca_idev->dev->ifindex;
3447 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3448 scope = RT_SCOPE_SITE;
3450 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3451 if (nlh == NULL)
3452 return -EMSGSIZE;
3454 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3455 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3456 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3457 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3458 nlmsg_cancel(skb, nlh);
3459 return -EMSGSIZE;
3462 return nlmsg_end(skb, nlh);
3465 enum addr_type_t
3467 UNICAST_ADDR,
3468 MULTICAST_ADDR,
3469 ANYCAST_ADDR,
3472 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3473 enum addr_type_t type)
3475 int idx, ip_idx;
3476 int s_idx, s_ip_idx;
3477 int err = 1;
3478 struct net_device *dev;
3479 struct inet6_dev *idev = NULL;
3480 struct inet6_ifaddr *ifa;
3481 struct ifmcaddr6 *ifmca;
3482 struct ifacaddr6 *ifaca;
3483 struct net *net = sock_net(skb->sk);
3485 s_idx = cb->args[0];
3486 s_ip_idx = ip_idx = cb->args[1];
3488 idx = 0;
3489 for_each_netdev(net, dev) {
3490 if (idx < s_idx)
3491 goto cont;
3492 if (idx > s_idx)
3493 s_ip_idx = 0;
3494 ip_idx = 0;
3495 if ((idev = in6_dev_get(dev)) == NULL)
3496 goto cont;
3497 read_lock_bh(&idev->lock);
3498 switch (type) {
3499 case UNICAST_ADDR:
3500 /* unicast address incl. temp addr */
3501 for (ifa = idev->addr_list; ifa;
3502 ifa = ifa->if_next, ip_idx++) {
3503 if (ip_idx < s_ip_idx)
3504 continue;
3505 err = inet6_fill_ifaddr(skb, ifa,
3506 NETLINK_CB(cb->skb).pid,
3507 cb->nlh->nlmsg_seq,
3508 RTM_NEWADDR,
3509 NLM_F_MULTI);
3511 break;
3512 case MULTICAST_ADDR:
3513 /* multicast address */
3514 for (ifmca = idev->mc_list; ifmca;
3515 ifmca = ifmca->next, ip_idx++) {
3516 if (ip_idx < s_ip_idx)
3517 continue;
3518 err = inet6_fill_ifmcaddr(skb, ifmca,
3519 NETLINK_CB(cb->skb).pid,
3520 cb->nlh->nlmsg_seq,
3521 RTM_GETMULTICAST,
3522 NLM_F_MULTI);
3524 break;
3525 case ANYCAST_ADDR:
3526 /* anycast address */
3527 for (ifaca = idev->ac_list; ifaca;
3528 ifaca = ifaca->aca_next, ip_idx++) {
3529 if (ip_idx < s_ip_idx)
3530 continue;
3531 err = inet6_fill_ifacaddr(skb, ifaca,
3532 NETLINK_CB(cb->skb).pid,
3533 cb->nlh->nlmsg_seq,
3534 RTM_GETANYCAST,
3535 NLM_F_MULTI);
3537 break;
3538 default:
3539 break;
3541 read_unlock_bh(&idev->lock);
3542 in6_dev_put(idev);
3544 if (err <= 0)
3545 break;
3546 cont:
3547 idx++;
3549 cb->args[0] = idx;
3550 cb->args[1] = ip_idx;
3551 return skb->len;
3554 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3556 enum addr_type_t type = UNICAST_ADDR;
3558 return inet6_dump_addr(skb, cb, type);
3561 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3563 enum addr_type_t type = MULTICAST_ADDR;
3565 return inet6_dump_addr(skb, cb, type);
3569 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3571 enum addr_type_t type = ANYCAST_ADDR;
3573 return inet6_dump_addr(skb, cb, type);
3576 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3577 void *arg)
3579 struct net *net = sock_net(in_skb->sk);
3580 struct ifaddrmsg *ifm;
3581 struct nlattr *tb[IFA_MAX+1];
3582 struct in6_addr *addr = NULL;
3583 struct net_device *dev = NULL;
3584 struct inet6_ifaddr *ifa;
3585 struct sk_buff *skb;
3586 int err;
3588 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3589 if (err < 0)
3590 goto errout;
3592 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3593 if (addr == NULL) {
3594 err = -EINVAL;
3595 goto errout;
3598 ifm = nlmsg_data(nlh);
3599 if (ifm->ifa_index)
3600 dev = __dev_get_by_index(net, ifm->ifa_index);
3602 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3603 err = -EADDRNOTAVAIL;
3604 goto errout;
3607 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3608 err = -ENOBUFS;
3609 goto errout_ifa;
3612 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3613 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3614 if (err < 0) {
3615 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3616 WARN_ON(err == -EMSGSIZE);
3617 kfree_skb(skb);
3618 goto errout_ifa;
3620 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3621 errout_ifa:
3622 in6_ifa_put(ifa);
3623 errout:
3624 return err;
3627 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3629 struct sk_buff *skb;
3630 struct net *net = dev_net(ifa->idev->dev);
3631 int err = -ENOBUFS;
3633 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3634 if (skb == NULL)
3635 goto errout;
3637 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3638 if (err < 0) {
3639 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3640 WARN_ON(err == -EMSGSIZE);
3641 kfree_skb(skb);
3642 goto errout;
3644 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3645 errout:
3646 if (err < 0)
3647 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3650 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3651 __s32 *array, int bytes)
3653 BUG_ON(bytes < (DEVCONF_MAX * 4));
3655 memset(array, 0, bytes);
3656 array[DEVCONF_FORWARDING] = cnf->forwarding;
3657 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3658 array[DEVCONF_MTU6] = cnf->mtu6;
3659 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3660 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3661 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3662 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3663 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3664 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3665 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3666 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3667 #ifdef CONFIG_IPV6_PRIVACY
3668 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3669 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3670 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3671 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3672 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3673 #endif
3674 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3675 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3676 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3677 #ifdef CONFIG_IPV6_ROUTER_PREF
3678 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3679 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3680 #ifdef CONFIG_IPV6_ROUTE_INFO
3681 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3682 #endif
3683 #endif
3684 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3685 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3686 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3687 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3688 #endif
3689 #ifdef CONFIG_IPV6_MROUTE
3690 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3691 #endif
3692 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3693 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3696 static inline size_t inet6_if_nlmsg_size(void)
3698 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3699 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3700 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3701 + nla_total_size(4) /* IFLA_MTU */
3702 + nla_total_size(4) /* IFLA_LINK */
3703 + nla_total_size( /* IFLA_PROTINFO */
3704 nla_total_size(4) /* IFLA_INET6_FLAGS */
3705 + nla_total_size(sizeof(struct ifla_cacheinfo))
3706 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3707 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3708 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3712 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3713 int bytes)
3715 int i;
3716 int pad = bytes - sizeof(u64) * items;
3717 BUG_ON(pad < 0);
3719 /* Use put_unaligned() because stats may not be aligned for u64. */
3720 put_unaligned(items, &stats[0]);
3721 for (i = 1; i < items; i++)
3722 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3724 memset(&stats[items], 0, pad);
3727 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3728 int bytes)
3730 switch(attrtype) {
3731 case IFLA_INET6_STATS:
3732 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3733 break;
3734 case IFLA_INET6_ICMP6STATS:
3735 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3736 break;
3740 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3741 u32 pid, u32 seq, int event, unsigned int flags)
3743 struct net_device *dev = idev->dev;
3744 struct nlattr *nla;
3745 struct ifinfomsg *hdr;
3746 struct nlmsghdr *nlh;
3747 void *protoinfo;
3748 struct ifla_cacheinfo ci;
3750 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3751 if (nlh == NULL)
3752 return -EMSGSIZE;
3754 hdr = nlmsg_data(nlh);
3755 hdr->ifi_family = AF_INET6;
3756 hdr->__ifi_pad = 0;
3757 hdr->ifi_type = dev->type;
3758 hdr->ifi_index = dev->ifindex;
3759 hdr->ifi_flags = dev_get_flags(dev);
3760 hdr->ifi_change = 0;
3762 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3764 if (dev->addr_len)
3765 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3767 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3768 if (dev->ifindex != dev->iflink)
3769 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3771 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3772 if (protoinfo == NULL)
3773 goto nla_put_failure;
3775 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3777 ci.max_reasm_len = IPV6_MAXPLEN;
3778 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3779 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3780 ci.reachable_time = idev->nd_parms->reachable_time;
3781 ci.retrans_time = idev->nd_parms->retrans_time;
3782 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3784 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3785 if (nla == NULL)
3786 goto nla_put_failure;
3787 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3789 /* XXX - MC not implemented */
3791 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3792 if (nla == NULL)
3793 goto nla_put_failure;
3794 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3796 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3797 if (nla == NULL)
3798 goto nla_put_failure;
3799 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3801 nla_nest_end(skb, protoinfo);
3802 return nlmsg_end(skb, nlh);
3804 nla_put_failure:
3805 nlmsg_cancel(skb, nlh);
3806 return -EMSGSIZE;
3809 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3811 struct net *net = sock_net(skb->sk);
3812 int idx, err;
3813 int s_idx = cb->args[0];
3814 struct net_device *dev;
3815 struct inet6_dev *idev;
3817 read_lock(&dev_base_lock);
3818 idx = 0;
3819 for_each_netdev(net, dev) {
3820 if (idx < s_idx)
3821 goto cont;
3822 if ((idev = in6_dev_get(dev)) == NULL)
3823 goto cont;
3824 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3825 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3826 in6_dev_put(idev);
3827 if (err <= 0)
3828 break;
3829 cont:
3830 idx++;
3832 read_unlock(&dev_base_lock);
3833 cb->args[0] = idx;
3835 return skb->len;
3838 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3840 struct sk_buff *skb;
3841 struct net *net = dev_net(idev->dev);
3842 int err = -ENOBUFS;
3844 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3845 if (skb == NULL)
3846 goto errout;
3848 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3849 if (err < 0) {
3850 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3851 WARN_ON(err == -EMSGSIZE);
3852 kfree_skb(skb);
3853 goto errout;
3855 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3856 errout:
3857 if (err < 0)
3858 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3861 static inline size_t inet6_prefix_nlmsg_size(void)
3863 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3864 + nla_total_size(sizeof(struct in6_addr))
3865 + nla_total_size(sizeof(struct prefix_cacheinfo));
3868 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3869 struct prefix_info *pinfo, u32 pid, u32 seq,
3870 int event, unsigned int flags)
3872 struct prefixmsg *pmsg;
3873 struct nlmsghdr *nlh;
3874 struct prefix_cacheinfo ci;
3876 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3877 if (nlh == NULL)
3878 return -EMSGSIZE;
3880 pmsg = nlmsg_data(nlh);
3881 pmsg->prefix_family = AF_INET6;
3882 pmsg->prefix_pad1 = 0;
3883 pmsg->prefix_pad2 = 0;
3884 pmsg->prefix_ifindex = idev->dev->ifindex;
3885 pmsg->prefix_len = pinfo->prefix_len;
3886 pmsg->prefix_type = pinfo->type;
3887 pmsg->prefix_pad3 = 0;
3888 pmsg->prefix_flags = 0;
3889 if (pinfo->onlink)
3890 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3891 if (pinfo->autoconf)
3892 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3894 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3896 ci.preferred_time = ntohl(pinfo->prefered);
3897 ci.valid_time = ntohl(pinfo->valid);
3898 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3900 return nlmsg_end(skb, nlh);
3902 nla_put_failure:
3903 nlmsg_cancel(skb, nlh);
3904 return -EMSGSIZE;
3907 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3908 struct prefix_info *pinfo)
3910 struct sk_buff *skb;
3911 struct net *net = dev_net(idev->dev);
3912 int err = -ENOBUFS;
3914 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3915 if (skb == NULL)
3916 goto errout;
3918 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3919 if (err < 0) {
3920 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3921 WARN_ON(err == -EMSGSIZE);
3922 kfree_skb(skb);
3923 goto errout;
3925 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3926 errout:
3927 if (err < 0)
3928 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3931 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3933 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3935 switch (event) {
3936 case RTM_NEWADDR:
3938 * If the address was optimistic
3939 * we inserted the route at the start of
3940 * our DAD process, so we don't need
3941 * to do it again
3943 if (!(ifp->rt->rt6i_node))
3944 ip6_ins_rt(ifp->rt);
3945 if (ifp->idev->cnf.forwarding)
3946 addrconf_join_anycast(ifp);
3947 break;
3948 case RTM_DELADDR:
3949 if (ifp->idev->cnf.forwarding)
3950 addrconf_leave_anycast(ifp);
3951 addrconf_leave_solict(ifp->idev, &ifp->addr);
3952 dst_hold(&ifp->rt->u.dst);
3953 if (ip6_del_rt(ifp->rt))
3954 dst_free(&ifp->rt->u.dst);
3955 break;
3959 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3961 rcu_read_lock_bh();
3962 if (likely(ifp->idev->dead == 0))
3963 __ipv6_ifa_notify(event, ifp);
3964 rcu_read_unlock_bh();
3967 #ifdef CONFIG_SYSCTL
3969 static
3970 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3971 void __user *buffer, size_t *lenp, loff_t *ppos)
3973 int *valp = ctl->data;
3974 int val = *valp;
3975 int ret;
3977 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3979 if (write)
3980 addrconf_fixup_forwarding(ctl, valp, val);
3981 return ret;
3984 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3985 int __user *name, int nlen,
3986 void __user *oldval,
3987 size_t __user *oldlenp,
3988 void __user *newval, size_t newlen)
3990 int *valp = table->data;
3991 int val = *valp;
3992 int new;
3994 if (!newval || !newlen)
3995 return 0;
3996 if (newlen != sizeof(int))
3997 return -EINVAL;
3998 if (get_user(new, (int __user *)newval))
3999 return -EFAULT;
4000 if (new == *valp)
4001 return 0;
4002 if (oldval && oldlenp) {
4003 size_t len;
4004 if (get_user(len, oldlenp))
4005 return -EFAULT;
4006 if (len) {
4007 if (len > table->maxlen)
4008 len = table->maxlen;
4009 if (copy_to_user(oldval, valp, len))
4010 return -EFAULT;
4011 if (put_user(len, oldlenp))
4012 return -EFAULT;
4016 *valp = new;
4017 addrconf_fixup_forwarding(table, valp, val);
4018 return 1;
4021 static struct addrconf_sysctl_table
4023 struct ctl_table_header *sysctl_header;
4024 ctl_table addrconf_vars[DEVCONF_MAX+1];
4025 char *dev_name;
4026 } addrconf_sysctl __read_mostly = {
4027 .sysctl_header = NULL,
4028 .addrconf_vars = {
4030 .ctl_name = NET_IPV6_FORWARDING,
4031 .procname = "forwarding",
4032 .data = &ipv6_devconf.forwarding,
4033 .maxlen = sizeof(int),
4034 .mode = 0644,
4035 .proc_handler = &addrconf_sysctl_forward,
4036 .strategy = &addrconf_sysctl_forward_strategy,
4039 .ctl_name = NET_IPV6_HOP_LIMIT,
4040 .procname = "hop_limit",
4041 .data = &ipv6_devconf.hop_limit,
4042 .maxlen = sizeof(int),
4043 .mode = 0644,
4044 .proc_handler = proc_dointvec,
4047 .ctl_name = NET_IPV6_MTU,
4048 .procname = "mtu",
4049 .data = &ipv6_devconf.mtu6,
4050 .maxlen = sizeof(int),
4051 .mode = 0644,
4052 .proc_handler = &proc_dointvec,
4055 .ctl_name = NET_IPV6_ACCEPT_RA,
4056 .procname = "accept_ra",
4057 .data = &ipv6_devconf.accept_ra,
4058 .maxlen = sizeof(int),
4059 .mode = 0644,
4060 .proc_handler = &proc_dointvec,
4063 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4064 .procname = "accept_redirects",
4065 .data = &ipv6_devconf.accept_redirects,
4066 .maxlen = sizeof(int),
4067 .mode = 0644,
4068 .proc_handler = &proc_dointvec,
4071 .ctl_name = NET_IPV6_AUTOCONF,
4072 .procname = "autoconf",
4073 .data = &ipv6_devconf.autoconf,
4074 .maxlen = sizeof(int),
4075 .mode = 0644,
4076 .proc_handler = &proc_dointvec,
4079 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4080 .procname = "dad_transmits",
4081 .data = &ipv6_devconf.dad_transmits,
4082 .maxlen = sizeof(int),
4083 .mode = 0644,
4084 .proc_handler = &proc_dointvec,
4087 .ctl_name = NET_IPV6_RTR_SOLICITS,
4088 .procname = "router_solicitations",
4089 .data = &ipv6_devconf.rtr_solicits,
4090 .maxlen = sizeof(int),
4091 .mode = 0644,
4092 .proc_handler = &proc_dointvec,
4095 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4096 .procname = "router_solicitation_interval",
4097 .data = &ipv6_devconf.rtr_solicit_interval,
4098 .maxlen = sizeof(int),
4099 .mode = 0644,
4100 .proc_handler = &proc_dointvec_jiffies,
4101 .strategy = &sysctl_jiffies,
4104 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4105 .procname = "router_solicitation_delay",
4106 .data = &ipv6_devconf.rtr_solicit_delay,
4107 .maxlen = sizeof(int),
4108 .mode = 0644,
4109 .proc_handler = &proc_dointvec_jiffies,
4110 .strategy = &sysctl_jiffies,
4113 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4114 .procname = "force_mld_version",
4115 .data = &ipv6_devconf.force_mld_version,
4116 .maxlen = sizeof(int),
4117 .mode = 0644,
4118 .proc_handler = &proc_dointvec,
4120 #ifdef CONFIG_IPV6_PRIVACY
4122 .ctl_name = NET_IPV6_USE_TEMPADDR,
4123 .procname = "use_tempaddr",
4124 .data = &ipv6_devconf.use_tempaddr,
4125 .maxlen = sizeof(int),
4126 .mode = 0644,
4127 .proc_handler = &proc_dointvec,
4130 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4131 .procname = "temp_valid_lft",
4132 .data = &ipv6_devconf.temp_valid_lft,
4133 .maxlen = sizeof(int),
4134 .mode = 0644,
4135 .proc_handler = &proc_dointvec,
4138 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4139 .procname = "temp_prefered_lft",
4140 .data = &ipv6_devconf.temp_prefered_lft,
4141 .maxlen = sizeof(int),
4142 .mode = 0644,
4143 .proc_handler = &proc_dointvec,
4146 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4147 .procname = "regen_max_retry",
4148 .data = &ipv6_devconf.regen_max_retry,
4149 .maxlen = sizeof(int),
4150 .mode = 0644,
4151 .proc_handler = &proc_dointvec,
4154 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4155 .procname = "max_desync_factor",
4156 .data = &ipv6_devconf.max_desync_factor,
4157 .maxlen = sizeof(int),
4158 .mode = 0644,
4159 .proc_handler = &proc_dointvec,
4161 #endif
4163 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4164 .procname = "max_addresses",
4165 .data = &ipv6_devconf.max_addresses,
4166 .maxlen = sizeof(int),
4167 .mode = 0644,
4168 .proc_handler = &proc_dointvec,
4171 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4172 .procname = "accept_ra_defrtr",
4173 .data = &ipv6_devconf.accept_ra_defrtr,
4174 .maxlen = sizeof(int),
4175 .mode = 0644,
4176 .proc_handler = &proc_dointvec,
4179 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4180 .procname = "accept_ra_pinfo",
4181 .data = &ipv6_devconf.accept_ra_pinfo,
4182 .maxlen = sizeof(int),
4183 .mode = 0644,
4184 .proc_handler = &proc_dointvec,
4186 #ifdef CONFIG_IPV6_ROUTER_PREF
4188 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4189 .procname = "accept_ra_rtr_pref",
4190 .data = &ipv6_devconf.accept_ra_rtr_pref,
4191 .maxlen = sizeof(int),
4192 .mode = 0644,
4193 .proc_handler = &proc_dointvec,
4196 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4197 .procname = "router_probe_interval",
4198 .data = &ipv6_devconf.rtr_probe_interval,
4199 .maxlen = sizeof(int),
4200 .mode = 0644,
4201 .proc_handler = &proc_dointvec_jiffies,
4202 .strategy = &sysctl_jiffies,
4204 #ifdef CONFIG_IPV6_ROUTE_INFO
4206 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4207 .procname = "accept_ra_rt_info_max_plen",
4208 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4209 .maxlen = sizeof(int),
4210 .mode = 0644,
4211 .proc_handler = &proc_dointvec,
4213 #endif
4214 #endif
4216 .ctl_name = NET_IPV6_PROXY_NDP,
4217 .procname = "proxy_ndp",
4218 .data = &ipv6_devconf.proxy_ndp,
4219 .maxlen = sizeof(int),
4220 .mode = 0644,
4221 .proc_handler = &proc_dointvec,
4224 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4225 .procname = "accept_source_route",
4226 .data = &ipv6_devconf.accept_source_route,
4227 .maxlen = sizeof(int),
4228 .mode = 0644,
4229 .proc_handler = &proc_dointvec,
4231 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4233 .ctl_name = CTL_UNNUMBERED,
4234 .procname = "optimistic_dad",
4235 .data = &ipv6_devconf.optimistic_dad,
4236 .maxlen = sizeof(int),
4237 .mode = 0644,
4238 .proc_handler = &proc_dointvec,
4241 #endif
4242 #ifdef CONFIG_IPV6_MROUTE
4244 .ctl_name = CTL_UNNUMBERED,
4245 .procname = "mc_forwarding",
4246 .data = &ipv6_devconf.mc_forwarding,
4247 .maxlen = sizeof(int),
4248 .mode = 0644,
4249 .proc_handler = &proc_dointvec,
4251 #endif
4253 .ctl_name = CTL_UNNUMBERED,
4254 .procname = "disable_ipv6",
4255 .data = &ipv6_devconf.disable_ipv6,
4256 .maxlen = sizeof(int),
4257 .mode = 0644,
4258 .proc_handler = &proc_dointvec,
4261 .ctl_name = CTL_UNNUMBERED,
4262 .procname = "accept_dad",
4263 .data = &ipv6_devconf.accept_dad,
4264 .maxlen = sizeof(int),
4265 .mode = 0644,
4266 .proc_handler = &proc_dointvec,
4269 .ctl_name = 0, /* sentinel */
4274 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4275 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4277 int i;
4278 struct addrconf_sysctl_table *t;
4280 #define ADDRCONF_CTL_PATH_DEV 3
4282 struct ctl_path addrconf_ctl_path[] = {
4283 { .procname = "net", .ctl_name = CTL_NET, },
4284 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4285 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4286 { /* to be set */ },
4287 { },
4291 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4292 if (t == NULL)
4293 goto out;
4295 for (i=0; t->addrconf_vars[i].data; i++) {
4296 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4297 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4298 t->addrconf_vars[i].extra2 = net;
4302 * Make a copy of dev_name, because '.procname' is regarded as const
4303 * by sysctl and we wouldn't want anyone to change it under our feet
4304 * (see SIOCSIFNAME).
4306 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4307 if (!t->dev_name)
4308 goto free;
4310 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4311 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4313 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4314 t->addrconf_vars);
4315 if (t->sysctl_header == NULL)
4316 goto free_procname;
4318 p->sysctl = t;
4319 return 0;
4321 free_procname:
4322 kfree(t->dev_name);
4323 free:
4324 kfree(t);
4325 out:
4326 return -ENOBUFS;
4329 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4331 struct addrconf_sysctl_table *t;
4333 if (p->sysctl == NULL)
4334 return;
4336 t = p->sysctl;
4337 p->sysctl = NULL;
4338 unregister_sysctl_table(t->sysctl_header);
4339 kfree(t->dev_name);
4340 kfree(t);
4343 static void addrconf_sysctl_register(struct inet6_dev *idev)
4345 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4346 NET_IPV6_NEIGH, "ipv6",
4347 &ndisc_ifinfo_sysctl_change,
4348 ndisc_ifinfo_sysctl_strategy);
4349 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4350 idev->dev->ifindex, idev, &idev->cnf);
4353 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4355 __addrconf_sysctl_unregister(&idev->cnf);
4356 neigh_sysctl_unregister(idev->nd_parms);
4360 #endif
4362 static int addrconf_init_net(struct net *net)
4364 int err;
4365 struct ipv6_devconf *all, *dflt;
4367 err = -ENOMEM;
4368 all = &ipv6_devconf;
4369 dflt = &ipv6_devconf_dflt;
4371 if (net != &init_net) {
4372 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4373 if (all == NULL)
4374 goto err_alloc_all;
4376 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4377 if (dflt == NULL)
4378 goto err_alloc_dflt;
4381 net->ipv6.devconf_all = all;
4382 net->ipv6.devconf_dflt = dflt;
4384 #ifdef CONFIG_SYSCTL
4385 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4386 NULL, all);
4387 if (err < 0)
4388 goto err_reg_all;
4390 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4391 NULL, dflt);
4392 if (err < 0)
4393 goto err_reg_dflt;
4394 #endif
4395 return 0;
4397 #ifdef CONFIG_SYSCTL
4398 err_reg_dflt:
4399 __addrconf_sysctl_unregister(all);
4400 err_reg_all:
4401 kfree(dflt);
4402 #endif
4403 err_alloc_dflt:
4404 kfree(all);
4405 err_alloc_all:
4406 return err;
4409 static void addrconf_exit_net(struct net *net)
4411 #ifdef CONFIG_SYSCTL
4412 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4413 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4414 #endif
4415 if (net != &init_net) {
4416 kfree(net->ipv6.devconf_dflt);
4417 kfree(net->ipv6.devconf_all);
4421 static struct pernet_operations addrconf_ops = {
4422 .init = addrconf_init_net,
4423 .exit = addrconf_exit_net,
4427 * Device notifier
4430 int register_inet6addr_notifier(struct notifier_block *nb)
4432 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4435 EXPORT_SYMBOL(register_inet6addr_notifier);
4437 int unregister_inet6addr_notifier(struct notifier_block *nb)
4439 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4442 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4444 static void addrconf_net_exit(struct net *net)
4446 struct net_device *dev;
4448 rtnl_lock();
4449 /* clean dev list */
4450 for_each_netdev(net, dev) {
4451 if (__in6_dev_get(dev) == NULL)
4452 continue;
4453 addrconf_ifdown(dev, 1);
4455 addrconf_ifdown(net->loopback_dev, 2);
4456 rtnl_unlock();
4459 static struct pernet_operations addrconf_net_ops = {
4460 .exit = addrconf_net_exit,
4464 * Init / cleanup code
4467 int __init addrconf_init(void)
4469 int err;
4471 if ((err = ipv6_addr_label_init()) < 0) {
4472 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4473 err);
4474 return err;
4477 register_pernet_subsys(&addrconf_ops);
4479 /* The addrconf netdev notifier requires that loopback_dev
4480 * has it's ipv6 private information allocated and setup
4481 * before it can bring up and give link-local addresses
4482 * to other devices which are up.
4484 * Unfortunately, loopback_dev is not necessarily the first
4485 * entry in the global dev_base list of net devices. In fact,
4486 * it is likely to be the very last entry on that list.
4487 * So this causes the notifier registry below to try and
4488 * give link-local addresses to all devices besides loopback_dev
4489 * first, then loopback_dev, which cases all the non-loopback_dev
4490 * devices to fail to get a link-local address.
4492 * So, as a temporary fix, allocate the ipv6 structure for
4493 * loopback_dev first by hand.
4494 * Longer term, all of the dependencies ipv6 has upon the loopback
4495 * device and it being up should be removed.
4497 rtnl_lock();
4498 if (!ipv6_add_dev(init_net.loopback_dev))
4499 err = -ENOMEM;
4500 rtnl_unlock();
4501 if (err)
4502 goto errlo;
4504 err = register_pernet_device(&addrconf_net_ops);
4505 if (err)
4506 return err;
4508 register_netdevice_notifier(&ipv6_dev_notf);
4510 addrconf_verify(0);
4512 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4513 if (err < 0)
4514 goto errout;
4516 /* Only the first call to __rtnl_register can fail */
4517 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4518 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4519 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4520 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4521 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4523 ipv6_addr_label_rtnl_register();
4525 return 0;
4526 errout:
4527 unregister_netdevice_notifier(&ipv6_dev_notf);
4528 errlo:
4529 unregister_pernet_subsys(&addrconf_ops);
4531 return err;
4534 void addrconf_cleanup(void)
4536 struct inet6_ifaddr *ifa;
4537 int i;
4539 unregister_netdevice_notifier(&ipv6_dev_notf);
4540 unregister_pernet_device(&addrconf_net_ops);
4542 unregister_pernet_subsys(&addrconf_ops);
4544 rtnl_lock();
4547 * Check hash table.
4549 write_lock_bh(&addrconf_hash_lock);
4550 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4551 for (ifa=inet6_addr_lst[i]; ifa; ) {
4552 struct inet6_ifaddr *bifa;
4554 bifa = ifa;
4555 ifa = ifa->lst_next;
4556 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4557 /* Do not free it; something is wrong.
4558 Now we can investigate it with debugger.
4562 write_unlock_bh(&addrconf_hash_lock);
4564 del_timer(&addr_chk_timer);
4565 rtnl_unlock();
4567 unregister_pernet_subsys(&addrconf_net_ops);