Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
[linux-2.6/mini2440.git] / net / ipv6 / addrconf.c
blobe2d3b7580b76fa991140dd72ea130c28890a0f26
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),
1693 ipv6_addr_copy(&cfg.fc_dst, pfx);
1695 /* Prevent useless cloning on PtP SIT.
1696 This thing is done here expecting that the whole
1697 class of non-broadcast devices need not cloning.
1699 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1700 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1701 cfg.fc_flags |= RTF_NONEXTHOP;
1702 #endif
1704 ip6_route_add(&cfg);
1707 /* Create "default" multicast route to the interface */
1709 static void addrconf_add_mroute(struct net_device *dev)
1711 struct fib6_config cfg = {
1712 .fc_table = RT6_TABLE_LOCAL,
1713 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1714 .fc_ifindex = dev->ifindex,
1715 .fc_dst_len = 8,
1716 .fc_flags = RTF_UP,
1717 .fc_nlinfo.nl_net = dev_net(dev),
1720 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1722 ip6_route_add(&cfg);
1725 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1726 static void sit_route_add(struct net_device *dev)
1728 struct fib6_config cfg = {
1729 .fc_table = RT6_TABLE_MAIN,
1730 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1731 .fc_ifindex = dev->ifindex,
1732 .fc_dst_len = 96,
1733 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1734 .fc_nlinfo.nl_net = dev_net(dev),
1737 /* prefix length - 96 bits "::d.d.d.d" */
1738 ip6_route_add(&cfg);
1740 #endif
1742 static void addrconf_add_lroute(struct net_device *dev)
1744 struct in6_addr addr;
1746 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1747 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1750 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1752 struct inet6_dev *idev;
1754 ASSERT_RTNL();
1756 if ((idev = ipv6_find_idev(dev)) == NULL)
1757 return NULL;
1759 /* Add default multicast route */
1760 addrconf_add_mroute(dev);
1762 /* Add link local route */
1763 addrconf_add_lroute(dev);
1764 return idev;
1767 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1769 struct prefix_info *pinfo;
1770 __u32 valid_lft;
1771 __u32 prefered_lft;
1772 int addr_type;
1773 struct inet6_dev *in6_dev;
1775 pinfo = (struct prefix_info *) opt;
1777 if (len < sizeof(struct prefix_info)) {
1778 ADBG(("addrconf: prefix option too short\n"));
1779 return;
1783 * Validation checks ([ADDRCONF], page 19)
1786 addr_type = ipv6_addr_type(&pinfo->prefix);
1788 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1789 return;
1791 valid_lft = ntohl(pinfo->valid);
1792 prefered_lft = ntohl(pinfo->prefered);
1794 if (prefered_lft > valid_lft) {
1795 if (net_ratelimit())
1796 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1797 return;
1800 in6_dev = in6_dev_get(dev);
1802 if (in6_dev == NULL) {
1803 if (net_ratelimit())
1804 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1805 return;
1809 * Two things going on here:
1810 * 1) Add routes for on-link prefixes
1811 * 2) Configure prefixes with the auto flag set
1814 if (pinfo->onlink) {
1815 struct rt6_info *rt;
1816 unsigned long rt_expires;
1818 /* Avoid arithmetic overflow. Really, we could
1819 * save rt_expires in seconds, likely valid_lft,
1820 * but it would require division in fib gc, that it
1821 * not good.
1823 if (HZ > USER_HZ)
1824 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1825 else
1826 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1828 if (addrconf_finite_timeout(rt_expires))
1829 rt_expires *= HZ;
1831 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1832 dev->ifindex, 1);
1834 if (rt && addrconf_is_prefix_route(rt)) {
1835 /* Autoconf prefix route */
1836 if (valid_lft == 0) {
1837 ip6_del_rt(rt);
1838 rt = NULL;
1839 } else if (addrconf_finite_timeout(rt_expires)) {
1840 /* not infinity */
1841 rt->rt6i_expires = jiffies + rt_expires;
1842 rt->rt6i_flags |= RTF_EXPIRES;
1843 } else {
1844 rt->rt6i_flags &= ~RTF_EXPIRES;
1845 rt->rt6i_expires = 0;
1847 } else if (valid_lft) {
1848 clock_t expires = 0;
1849 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1850 if (addrconf_finite_timeout(rt_expires)) {
1851 /* not infinity */
1852 flags |= RTF_EXPIRES;
1853 expires = jiffies_to_clock_t(rt_expires);
1855 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1856 dev, expires, flags);
1858 if (rt)
1859 dst_release(&rt->u.dst);
1862 /* Try to figure out our local address for this prefix */
1864 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1865 struct inet6_ifaddr * ifp;
1866 struct in6_addr addr;
1867 int create = 0, update_lft = 0;
1868 struct net *net = dev_net(dev);
1870 if (pinfo->prefix_len == 64) {
1871 memcpy(&addr, &pinfo->prefix, 8);
1872 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1873 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1874 in6_dev_put(in6_dev);
1875 return;
1877 goto ok;
1879 if (net_ratelimit())
1880 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1881 pinfo->prefix_len);
1882 in6_dev_put(in6_dev);
1883 return;
1887 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1889 if (ifp == NULL && valid_lft) {
1890 int max_addresses = in6_dev->cnf.max_addresses;
1891 u32 addr_flags = 0;
1893 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1894 if (in6_dev->cnf.optimistic_dad &&
1895 !net->ipv6.devconf_all->forwarding)
1896 addr_flags = IFA_F_OPTIMISTIC;
1897 #endif
1899 /* Do not allow to create too much of autoconfigured
1900 * addresses; this would be too easy way to crash kernel.
1902 if (!max_addresses ||
1903 ipv6_count_addresses(in6_dev) < max_addresses)
1904 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1905 addr_type&IPV6_ADDR_SCOPE_MASK,
1906 addr_flags);
1908 if (!ifp || IS_ERR(ifp)) {
1909 in6_dev_put(in6_dev);
1910 return;
1913 update_lft = create = 1;
1914 ifp->cstamp = jiffies;
1915 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1918 if (ifp) {
1919 int flags;
1920 unsigned long now;
1921 #ifdef CONFIG_IPV6_PRIVACY
1922 struct inet6_ifaddr *ift;
1923 #endif
1924 u32 stored_lft;
1926 /* update lifetime (RFC2462 5.5.3 e) */
1927 spin_lock(&ifp->lock);
1928 now = jiffies;
1929 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1930 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1931 else
1932 stored_lft = 0;
1933 if (!update_lft && stored_lft) {
1934 if (valid_lft > MIN_VALID_LIFETIME ||
1935 valid_lft > stored_lft)
1936 update_lft = 1;
1937 else if (stored_lft <= MIN_VALID_LIFETIME) {
1938 /* valid_lft <= stored_lft is always true */
1939 /* XXX: IPsec */
1940 update_lft = 0;
1941 } else {
1942 valid_lft = MIN_VALID_LIFETIME;
1943 if (valid_lft < prefered_lft)
1944 prefered_lft = valid_lft;
1945 update_lft = 1;
1949 if (update_lft) {
1950 ifp->valid_lft = valid_lft;
1951 ifp->prefered_lft = prefered_lft;
1952 ifp->tstamp = now;
1953 flags = ifp->flags;
1954 ifp->flags &= ~IFA_F_DEPRECATED;
1955 spin_unlock(&ifp->lock);
1957 if (!(flags&IFA_F_TENTATIVE))
1958 ipv6_ifa_notify(0, ifp);
1959 } else
1960 spin_unlock(&ifp->lock);
1962 #ifdef CONFIG_IPV6_PRIVACY
1963 read_lock_bh(&in6_dev->lock);
1964 /* update all temporary addresses in the list */
1965 for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1967 * When adjusting the lifetimes of an existing
1968 * temporary address, only lower the lifetimes.
1969 * Implementations must not increase the
1970 * lifetimes of an existing temporary address
1971 * when processing a Prefix Information Option.
1973 if (ifp != ift->ifpub)
1974 continue;
1976 spin_lock(&ift->lock);
1977 flags = ift->flags;
1978 if (ift->valid_lft > valid_lft &&
1979 ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1980 ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1981 if (ift->prefered_lft > prefered_lft &&
1982 ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1983 ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1984 spin_unlock(&ift->lock);
1985 if (!(flags&IFA_F_TENTATIVE))
1986 ipv6_ifa_notify(0, ift);
1989 if (create && in6_dev->cnf.use_tempaddr > 0) {
1991 * When a new public address is created as described in [ADDRCONF],
1992 * also create a new temporary address.
1994 read_unlock_bh(&in6_dev->lock);
1995 ipv6_create_tempaddr(ifp, NULL);
1996 } else {
1997 read_unlock_bh(&in6_dev->lock);
1999 #endif
2000 in6_ifa_put(ifp);
2001 addrconf_verify(0);
2004 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2005 in6_dev_put(in6_dev);
2009 * Set destination address.
2010 * Special case for SIT interfaces where we create a new "virtual"
2011 * device.
2013 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2015 struct in6_ifreq ireq;
2016 struct net_device *dev;
2017 int err = -EINVAL;
2019 rtnl_lock();
2021 err = -EFAULT;
2022 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2023 goto err_exit;
2025 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2027 err = -ENODEV;
2028 if (dev == NULL)
2029 goto err_exit;
2031 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2032 if (dev->type == ARPHRD_SIT) {
2033 struct ifreq ifr;
2034 mm_segment_t oldfs;
2035 struct ip_tunnel_parm p;
2037 err = -EADDRNOTAVAIL;
2038 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2039 goto err_exit;
2041 memset(&p, 0, sizeof(p));
2042 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2043 p.iph.saddr = 0;
2044 p.iph.version = 4;
2045 p.iph.ihl = 5;
2046 p.iph.protocol = IPPROTO_IPV6;
2047 p.iph.ttl = 64;
2048 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2050 oldfs = get_fs(); set_fs(KERNEL_DS);
2051 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2052 set_fs(oldfs);
2054 if (err == 0) {
2055 err = -ENOBUFS;
2056 dev = __dev_get_by_name(net, p.name);
2057 if (!dev)
2058 goto err_exit;
2059 err = dev_open(dev);
2062 #endif
2064 err_exit:
2065 rtnl_unlock();
2066 return err;
2070 * Manual configuration of address on an interface
2072 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2073 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2074 __u32 valid_lft)
2076 struct inet6_ifaddr *ifp;
2077 struct inet6_dev *idev;
2078 struct net_device *dev;
2079 int scope;
2080 u32 flags;
2081 clock_t expires;
2082 unsigned long timeout;
2084 ASSERT_RTNL();
2086 if (plen > 128)
2087 return -EINVAL;
2089 /* check the lifetime */
2090 if (!valid_lft || prefered_lft > valid_lft)
2091 return -EINVAL;
2093 dev = __dev_get_by_index(net, ifindex);
2094 if (!dev)
2095 return -ENODEV;
2097 if ((idev = addrconf_add_dev(dev)) == NULL)
2098 return -ENOBUFS;
2100 scope = ipv6_addr_scope(pfx);
2102 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2103 if (addrconf_finite_timeout(timeout)) {
2104 expires = jiffies_to_clock_t(timeout * HZ);
2105 valid_lft = timeout;
2106 flags = RTF_EXPIRES;
2107 } else {
2108 expires = 0;
2109 flags = 0;
2110 ifa_flags |= IFA_F_PERMANENT;
2113 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2114 if (addrconf_finite_timeout(timeout)) {
2115 if (timeout == 0)
2116 ifa_flags |= IFA_F_DEPRECATED;
2117 prefered_lft = timeout;
2120 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2122 if (!IS_ERR(ifp)) {
2123 spin_lock_bh(&ifp->lock);
2124 ifp->valid_lft = valid_lft;
2125 ifp->prefered_lft = prefered_lft;
2126 ifp->tstamp = jiffies;
2127 spin_unlock_bh(&ifp->lock);
2129 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2130 expires, flags);
2132 * Note that section 3.1 of RFC 4429 indicates
2133 * that the Optimistic flag should not be set for
2134 * manually configured addresses
2136 addrconf_dad_start(ifp, 0);
2137 in6_ifa_put(ifp);
2138 addrconf_verify(0);
2139 return 0;
2142 return PTR_ERR(ifp);
2145 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2146 unsigned int plen)
2148 struct inet6_ifaddr *ifp;
2149 struct inet6_dev *idev;
2150 struct net_device *dev;
2152 if (plen > 128)
2153 return -EINVAL;
2155 dev = __dev_get_by_index(net, ifindex);
2156 if (!dev)
2157 return -ENODEV;
2159 if ((idev = __in6_dev_get(dev)) == NULL)
2160 return -ENXIO;
2162 read_lock_bh(&idev->lock);
2163 for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2164 if (ifp->prefix_len == plen &&
2165 ipv6_addr_equal(pfx, &ifp->addr)) {
2166 in6_ifa_hold(ifp);
2167 read_unlock_bh(&idev->lock);
2169 ipv6_del_addr(ifp);
2171 /* If the last address is deleted administratively,
2172 disable IPv6 on this interface.
2174 if (idev->addr_list == NULL)
2175 addrconf_ifdown(idev->dev, 1);
2176 return 0;
2179 read_unlock_bh(&idev->lock);
2180 return -EADDRNOTAVAIL;
2184 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2186 struct in6_ifreq ireq;
2187 int err;
2189 if (!capable(CAP_NET_ADMIN))
2190 return -EPERM;
2192 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2193 return -EFAULT;
2195 rtnl_lock();
2196 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2197 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2198 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2199 rtnl_unlock();
2200 return err;
2203 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2205 struct in6_ifreq ireq;
2206 int err;
2208 if (!capable(CAP_NET_ADMIN))
2209 return -EPERM;
2211 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2212 return -EFAULT;
2214 rtnl_lock();
2215 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2216 ireq.ifr6_prefixlen);
2217 rtnl_unlock();
2218 return err;
2221 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2222 static void sit_add_v4_addrs(struct inet6_dev *idev)
2224 struct inet6_ifaddr * ifp;
2225 struct in6_addr addr;
2226 struct net_device *dev;
2227 struct net *net = dev_net(idev->dev);
2228 int scope;
2230 ASSERT_RTNL();
2232 memset(&addr, 0, sizeof(struct in6_addr));
2233 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2235 if (idev->dev->flags&IFF_POINTOPOINT) {
2236 addr.s6_addr32[0] = htonl(0xfe800000);
2237 scope = IFA_LINK;
2238 } else {
2239 scope = IPV6_ADDR_COMPATv4;
2242 if (addr.s6_addr32[3]) {
2243 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2244 if (!IS_ERR(ifp)) {
2245 spin_lock_bh(&ifp->lock);
2246 ifp->flags &= ~IFA_F_TENTATIVE;
2247 spin_unlock_bh(&ifp->lock);
2248 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2249 in6_ifa_put(ifp);
2251 return;
2254 for_each_netdev(net, dev) {
2255 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2256 if (in_dev && (dev->flags & IFF_UP)) {
2257 struct in_ifaddr * ifa;
2259 int flag = scope;
2261 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2262 int plen;
2264 addr.s6_addr32[3] = ifa->ifa_local;
2266 if (ifa->ifa_scope == RT_SCOPE_LINK)
2267 continue;
2268 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2269 if (idev->dev->flags&IFF_POINTOPOINT)
2270 continue;
2271 flag |= IFA_HOST;
2273 if (idev->dev->flags&IFF_POINTOPOINT)
2274 plen = 64;
2275 else
2276 plen = 96;
2278 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2279 IFA_F_PERMANENT);
2280 if (!IS_ERR(ifp)) {
2281 spin_lock_bh(&ifp->lock);
2282 ifp->flags &= ~IFA_F_TENTATIVE;
2283 spin_unlock_bh(&ifp->lock);
2284 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2285 in6_ifa_put(ifp);
2291 #endif
2293 static void init_loopback(struct net_device *dev)
2295 struct inet6_dev *idev;
2296 struct inet6_ifaddr * ifp;
2298 /* ::1 */
2300 ASSERT_RTNL();
2302 if ((idev = ipv6_find_idev(dev)) == NULL) {
2303 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2304 return;
2307 ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2308 if (!IS_ERR(ifp)) {
2309 spin_lock_bh(&ifp->lock);
2310 ifp->flags &= ~IFA_F_TENTATIVE;
2311 spin_unlock_bh(&ifp->lock);
2312 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2313 in6_ifa_put(ifp);
2317 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2319 struct inet6_ifaddr * ifp;
2320 u32 addr_flags = IFA_F_PERMANENT;
2322 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2323 if (idev->cnf.optimistic_dad &&
2324 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2325 addr_flags |= IFA_F_OPTIMISTIC;
2326 #endif
2329 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2330 if (!IS_ERR(ifp)) {
2331 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2332 addrconf_dad_start(ifp, 0);
2333 in6_ifa_put(ifp);
2337 static void addrconf_dev_config(struct net_device *dev)
2339 struct in6_addr addr;
2340 struct inet6_dev * idev;
2342 ASSERT_RTNL();
2344 if ((dev->type != ARPHRD_ETHER) &&
2345 (dev->type != ARPHRD_FDDI) &&
2346 (dev->type != ARPHRD_IEEE802_TR) &&
2347 (dev->type != ARPHRD_ARCNET) &&
2348 (dev->type != ARPHRD_INFINIBAND)) {
2349 /* Alas, we support only Ethernet autoconfiguration. */
2350 return;
2353 idev = addrconf_add_dev(dev);
2354 if (idev == NULL)
2355 return;
2357 memset(&addr, 0, sizeof(struct in6_addr));
2358 addr.s6_addr32[0] = htonl(0xFE800000);
2360 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2361 addrconf_add_linklocal(idev, &addr);
2364 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2365 static void addrconf_sit_config(struct net_device *dev)
2367 struct inet6_dev *idev;
2369 ASSERT_RTNL();
2372 * Configure the tunnel with one of our IPv4
2373 * addresses... we should configure all of
2374 * our v4 addrs in the tunnel
2377 if ((idev = ipv6_find_idev(dev)) == NULL) {
2378 printk(KERN_DEBUG "init sit: add_dev failed\n");
2379 return;
2382 if (dev->priv_flags & IFF_ISATAP) {
2383 struct in6_addr addr;
2385 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2386 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2387 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2388 addrconf_add_linklocal(idev, &addr);
2389 return;
2392 sit_add_v4_addrs(idev);
2394 if (dev->flags&IFF_POINTOPOINT) {
2395 addrconf_add_mroute(dev);
2396 addrconf_add_lroute(dev);
2397 } else
2398 sit_route_add(dev);
2400 #endif
2402 static inline int
2403 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2405 struct in6_addr lladdr;
2407 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2408 addrconf_add_linklocal(idev, &lladdr);
2409 return 0;
2411 return -1;
2414 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2416 struct net_device *link_dev;
2417 struct net *net = dev_net(idev->dev);
2419 /* first try to inherit the link-local address from the link device */
2420 if (idev->dev->iflink &&
2421 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2422 if (!ipv6_inherit_linklocal(idev, link_dev))
2423 return;
2425 /* then try to inherit it from any device */
2426 for_each_netdev(net, link_dev) {
2427 if (!ipv6_inherit_linklocal(idev, link_dev))
2428 return;
2430 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2434 * Autoconfigure tunnel with a link-local address so routing protocols,
2435 * DHCPv6, MLD etc. can be run over the virtual link
2438 static void addrconf_ip6_tnl_config(struct net_device *dev)
2440 struct inet6_dev *idev;
2442 ASSERT_RTNL();
2444 if ((idev = addrconf_add_dev(dev)) == NULL) {
2445 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2446 return;
2448 ip6_tnl_add_linklocal(idev);
2451 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2452 void * data)
2454 struct net_device *dev = (struct net_device *) data;
2455 struct inet6_dev *idev = __in6_dev_get(dev);
2456 int run_pending = 0;
2457 int err;
2459 switch(event) {
2460 case NETDEV_REGISTER:
2461 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2462 idev = ipv6_add_dev(dev);
2463 if (!idev)
2464 return notifier_from_errno(-ENOMEM);
2466 break;
2467 case NETDEV_UP:
2468 case NETDEV_CHANGE:
2469 if (dev->flags & IFF_SLAVE)
2470 break;
2472 if (event == NETDEV_UP) {
2473 if (!addrconf_qdisc_ok(dev)) {
2474 /* device is not ready yet. */
2475 printk(KERN_INFO
2476 "ADDRCONF(NETDEV_UP): %s: "
2477 "link is not ready\n",
2478 dev->name);
2479 break;
2482 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2483 idev = ipv6_add_dev(dev);
2485 if (idev)
2486 idev->if_flags |= IF_READY;
2487 } else {
2488 if (!addrconf_qdisc_ok(dev)) {
2489 /* device is still not ready. */
2490 break;
2493 if (idev) {
2494 if (idev->if_flags & IF_READY) {
2495 /* device is already configured. */
2496 break;
2498 idev->if_flags |= IF_READY;
2501 printk(KERN_INFO
2502 "ADDRCONF(NETDEV_CHANGE): %s: "
2503 "link becomes ready\n",
2504 dev->name);
2506 run_pending = 1;
2509 switch(dev->type) {
2510 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2511 case ARPHRD_SIT:
2512 addrconf_sit_config(dev);
2513 break;
2514 #endif
2515 case ARPHRD_TUNNEL6:
2516 addrconf_ip6_tnl_config(dev);
2517 break;
2518 case ARPHRD_LOOPBACK:
2519 init_loopback(dev);
2520 break;
2522 default:
2523 addrconf_dev_config(dev);
2524 break;
2526 if (idev) {
2527 if (run_pending)
2528 addrconf_dad_run(idev);
2530 /* If the MTU changed during the interface down, when the
2531 interface up, the changed MTU must be reflected in the
2532 idev as well as routers.
2534 if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2535 rt6_mtu_change(dev, dev->mtu);
2536 idev->cnf.mtu6 = dev->mtu;
2538 idev->tstamp = jiffies;
2539 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2540 /* If the changed mtu during down is lower than IPV6_MIN_MTU
2541 stop IPv6 on this interface.
2543 if (dev->mtu < IPV6_MIN_MTU)
2544 addrconf_ifdown(dev, event != NETDEV_DOWN);
2546 break;
2548 case NETDEV_CHANGEMTU:
2549 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2550 rt6_mtu_change(dev, dev->mtu);
2551 idev->cnf.mtu6 = dev->mtu;
2552 break;
2555 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2556 idev = ipv6_add_dev(dev);
2557 if (idev)
2558 break;
2561 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2563 case NETDEV_DOWN:
2564 case NETDEV_UNREGISTER:
2566 * Remove all addresses from this interface.
2568 addrconf_ifdown(dev, event != NETDEV_DOWN);
2569 break;
2571 case NETDEV_CHANGENAME:
2572 if (idev) {
2573 snmp6_unregister_dev(idev);
2574 addrconf_sysctl_unregister(idev);
2575 addrconf_sysctl_register(idev);
2576 err = snmp6_register_dev(idev);
2577 if (err)
2578 return notifier_from_errno(err);
2580 break;
2583 return NOTIFY_OK;
2587 * addrconf module should be notified of a device going up
2589 static struct notifier_block ipv6_dev_notf = {
2590 .notifier_call = addrconf_notify,
2591 .priority = 0
2594 static int addrconf_ifdown(struct net_device *dev, int how)
2596 struct inet6_dev *idev;
2597 struct inet6_ifaddr *ifa, **bifa;
2598 struct net *net = dev_net(dev);
2599 int i;
2601 ASSERT_RTNL();
2603 if ((dev->flags & IFF_LOOPBACK) && how == 1)
2604 how = 0;
2606 rt6_ifdown(net, dev);
2607 neigh_ifdown(&nd_tbl, dev);
2609 idev = __in6_dev_get(dev);
2610 if (idev == NULL)
2611 return -ENODEV;
2613 /* Step 1: remove reference to ipv6 device from parent device.
2614 Do not dev_put!
2616 if (how) {
2617 idev->dead = 1;
2619 /* protected by rtnl_lock */
2620 rcu_assign_pointer(dev->ip6_ptr, NULL);
2622 /* Step 1.5: remove snmp6 entry */
2623 snmp6_unregister_dev(idev);
2627 /* Step 2: clear hash table */
2628 for (i=0; i<IN6_ADDR_HSIZE; i++) {
2629 bifa = &inet6_addr_lst[i];
2631 write_lock_bh(&addrconf_hash_lock);
2632 while ((ifa = *bifa) != NULL) {
2633 if (ifa->idev == idev) {
2634 *bifa = ifa->lst_next;
2635 ifa->lst_next = NULL;
2636 addrconf_del_timer(ifa);
2637 in6_ifa_put(ifa);
2638 continue;
2640 bifa = &ifa->lst_next;
2642 write_unlock_bh(&addrconf_hash_lock);
2645 write_lock_bh(&idev->lock);
2647 /* Step 3: clear flags for stateless addrconf */
2648 if (!how)
2649 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2651 /* Step 4: clear address list */
2652 #ifdef CONFIG_IPV6_PRIVACY
2653 if (how && del_timer(&idev->regen_timer))
2654 in6_dev_put(idev);
2656 /* clear tempaddr list */
2657 while ((ifa = idev->tempaddr_list) != NULL) {
2658 idev->tempaddr_list = ifa->tmp_next;
2659 ifa->tmp_next = NULL;
2660 ifa->dead = 1;
2661 write_unlock_bh(&idev->lock);
2662 spin_lock_bh(&ifa->lock);
2664 if (ifa->ifpub) {
2665 in6_ifa_put(ifa->ifpub);
2666 ifa->ifpub = NULL;
2668 spin_unlock_bh(&ifa->lock);
2669 in6_ifa_put(ifa);
2670 write_lock_bh(&idev->lock);
2672 #endif
2673 while ((ifa = idev->addr_list) != NULL) {
2674 idev->addr_list = ifa->if_next;
2675 ifa->if_next = NULL;
2676 ifa->dead = 1;
2677 addrconf_del_timer(ifa);
2678 write_unlock_bh(&idev->lock);
2680 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2681 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2682 in6_ifa_put(ifa);
2684 write_lock_bh(&idev->lock);
2686 write_unlock_bh(&idev->lock);
2688 /* Step 5: Discard multicast list */
2690 if (how)
2691 ipv6_mc_destroy_dev(idev);
2692 else
2693 ipv6_mc_down(idev);
2695 idev->tstamp = jiffies;
2697 /* Shot the device (if unregistered) */
2699 if (how) {
2700 addrconf_sysctl_unregister(idev);
2701 neigh_parms_release(&nd_tbl, idev->nd_parms);
2702 neigh_ifdown(&nd_tbl, dev);
2703 in6_dev_put(idev);
2705 return 0;
2708 static void addrconf_rs_timer(unsigned long data)
2710 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2712 if (ifp->idev->cnf.forwarding)
2713 goto out;
2715 if (ifp->idev->if_flags & IF_RA_RCVD) {
2717 * Announcement received after solicitation
2718 * was sent
2720 goto out;
2723 spin_lock(&ifp->lock);
2724 if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2725 /* The wait after the last probe can be shorter */
2726 addrconf_mod_timer(ifp, AC_RS,
2727 (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2728 ifp->idev->cnf.rtr_solicit_delay :
2729 ifp->idev->cnf.rtr_solicit_interval);
2730 spin_unlock(&ifp->lock);
2732 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2733 } else {
2734 spin_unlock(&ifp->lock);
2736 * Note: we do not support deprecated "all on-link"
2737 * assumption any longer.
2739 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2740 ifp->idev->dev->name);
2743 out:
2744 in6_ifa_put(ifp);
2748 * Duplicate Address Detection
2750 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2752 unsigned long rand_num;
2753 struct inet6_dev *idev = ifp->idev;
2755 if (ifp->flags & IFA_F_OPTIMISTIC)
2756 rand_num = 0;
2757 else
2758 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2760 ifp->probes = idev->cnf.dad_transmits;
2761 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2764 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2766 struct inet6_dev *idev = ifp->idev;
2767 struct net_device *dev = idev->dev;
2769 addrconf_join_solict(dev, &ifp->addr);
2771 net_srandom(ifp->addr.s6_addr32[3]);
2773 read_lock_bh(&idev->lock);
2774 if (ifp->dead)
2775 goto out;
2776 spin_lock_bh(&ifp->lock);
2778 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2779 idev->cnf.accept_dad < 1 ||
2780 !(ifp->flags&IFA_F_TENTATIVE) ||
2781 ifp->flags & IFA_F_NODAD) {
2782 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2783 spin_unlock_bh(&ifp->lock);
2784 read_unlock_bh(&idev->lock);
2786 addrconf_dad_completed(ifp);
2787 return;
2790 if (!(idev->if_flags & IF_READY)) {
2791 spin_unlock_bh(&ifp->lock);
2792 read_unlock_bh(&idev->lock);
2794 * If the defice is not ready:
2795 * - keep it tentative if it is a permanent address.
2796 * - otherwise, kill it.
2798 in6_ifa_hold(ifp);
2799 addrconf_dad_stop(ifp);
2800 return;
2804 * Optimistic nodes can start receiving
2805 * Frames right away
2807 if(ifp->flags & IFA_F_OPTIMISTIC)
2808 ip6_ins_rt(ifp->rt);
2810 addrconf_dad_kick(ifp);
2811 spin_unlock_bh(&ifp->lock);
2812 out:
2813 read_unlock_bh(&idev->lock);
2816 static void addrconf_dad_timer(unsigned long data)
2818 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2819 struct inet6_dev *idev = ifp->idev;
2820 struct in6_addr mcaddr;
2822 read_lock_bh(&idev->lock);
2823 if (idev->dead) {
2824 read_unlock_bh(&idev->lock);
2825 goto out;
2827 if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2828 read_unlock_bh(&idev->lock);
2829 addrconf_dad_failure(ifp);
2830 return;
2832 spin_lock_bh(&ifp->lock);
2833 if (ifp->probes == 0) {
2835 * DAD was successful
2838 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2839 spin_unlock_bh(&ifp->lock);
2840 read_unlock_bh(&idev->lock);
2842 addrconf_dad_completed(ifp);
2844 goto out;
2847 ifp->probes--;
2848 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2849 spin_unlock_bh(&ifp->lock);
2850 read_unlock_bh(&idev->lock);
2852 /* send a neighbour solicitation for our addr */
2853 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2854 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2855 out:
2856 in6_ifa_put(ifp);
2859 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2861 struct net_device * dev = ifp->idev->dev;
2864 * Configure the address for reception. Now it is valid.
2867 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2869 /* If added prefix is link local and forwarding is off,
2870 start sending router solicitations.
2873 if (ifp->idev->cnf.forwarding == 0 &&
2874 ifp->idev->cnf.rtr_solicits > 0 &&
2875 (dev->flags&IFF_LOOPBACK) == 0 &&
2876 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2878 * If a host as already performed a random delay
2879 * [...] as part of DAD [...] there is no need
2880 * to delay again before sending the first RS
2882 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2884 spin_lock_bh(&ifp->lock);
2885 ifp->probes = 1;
2886 ifp->idev->if_flags |= IF_RS_SENT;
2887 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2888 spin_unlock_bh(&ifp->lock);
2892 static void addrconf_dad_run(struct inet6_dev *idev) {
2893 struct inet6_ifaddr *ifp;
2895 read_lock_bh(&idev->lock);
2896 for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2897 spin_lock_bh(&ifp->lock);
2898 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2899 spin_unlock_bh(&ifp->lock);
2900 continue;
2902 spin_unlock_bh(&ifp->lock);
2903 addrconf_dad_kick(ifp);
2905 read_unlock_bh(&idev->lock);
2908 #ifdef CONFIG_PROC_FS
2909 struct if6_iter_state {
2910 struct seq_net_private p;
2911 int bucket;
2914 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2916 struct inet6_ifaddr *ifa = NULL;
2917 struct if6_iter_state *state = seq->private;
2918 struct net *net = seq_file_net(seq);
2920 for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2921 ifa = inet6_addr_lst[state->bucket];
2923 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2924 ifa = ifa->lst_next;
2925 if (ifa)
2926 break;
2928 return ifa;
2931 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2933 struct if6_iter_state *state = seq->private;
2934 struct net *net = seq_file_net(seq);
2936 ifa = ifa->lst_next;
2937 try_again:
2938 if (ifa) {
2939 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2940 ifa = ifa->lst_next;
2941 goto try_again;
2945 if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2946 ifa = inet6_addr_lst[state->bucket];
2947 goto try_again;
2950 return ifa;
2953 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2955 struct inet6_ifaddr *ifa = if6_get_first(seq);
2957 if (ifa)
2958 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2959 --pos;
2960 return pos ? NULL : ifa;
2963 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2964 __acquires(addrconf_hash_lock)
2966 read_lock_bh(&addrconf_hash_lock);
2967 return if6_get_idx(seq, *pos);
2970 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2972 struct inet6_ifaddr *ifa;
2974 ifa = if6_get_next(seq, v);
2975 ++*pos;
2976 return ifa;
2979 static void if6_seq_stop(struct seq_file *seq, void *v)
2980 __releases(addrconf_hash_lock)
2982 read_unlock_bh(&addrconf_hash_lock);
2985 static int if6_seq_show(struct seq_file *seq, void *v)
2987 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2988 seq_printf(seq,
2989 NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2990 NIP6(ifp->addr),
2991 ifp->idev->dev->ifindex,
2992 ifp->prefix_len,
2993 ifp->scope,
2994 ifp->flags,
2995 ifp->idev->dev->name);
2996 return 0;
2999 static const struct seq_operations if6_seq_ops = {
3000 .start = if6_seq_start,
3001 .next = if6_seq_next,
3002 .show = if6_seq_show,
3003 .stop = if6_seq_stop,
3006 static int if6_seq_open(struct inode *inode, struct file *file)
3008 return seq_open_net(inode, file, &if6_seq_ops,
3009 sizeof(struct if6_iter_state));
3012 static const struct file_operations if6_fops = {
3013 .owner = THIS_MODULE,
3014 .open = if6_seq_open,
3015 .read = seq_read,
3016 .llseek = seq_lseek,
3017 .release = seq_release_net,
3020 static int if6_proc_net_init(struct net *net)
3022 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3023 return -ENOMEM;
3024 return 0;
3027 static void if6_proc_net_exit(struct net *net)
3029 proc_net_remove(net, "if_inet6");
3032 static struct pernet_operations if6_proc_net_ops = {
3033 .init = if6_proc_net_init,
3034 .exit = if6_proc_net_exit,
3037 int __init if6_proc_init(void)
3039 return register_pernet_subsys(&if6_proc_net_ops);
3042 void if6_proc_exit(void)
3044 unregister_pernet_subsys(&if6_proc_net_ops);
3046 #endif /* CONFIG_PROC_FS */
3048 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3049 /* Check if address is a home address configured on any interface. */
3050 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3052 int ret = 0;
3053 struct inet6_ifaddr * ifp;
3054 u8 hash = ipv6_addr_hash(addr);
3055 read_lock_bh(&addrconf_hash_lock);
3056 for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3057 if (!net_eq(dev_net(ifp->idev->dev), net))
3058 continue;
3059 if (ipv6_addr_equal(&ifp->addr, addr) &&
3060 (ifp->flags & IFA_F_HOMEADDRESS)) {
3061 ret = 1;
3062 break;
3065 read_unlock_bh(&addrconf_hash_lock);
3066 return ret;
3068 #endif
3071 * Periodic address status verification
3074 static void addrconf_verify(unsigned long foo)
3076 struct inet6_ifaddr *ifp;
3077 unsigned long now, next;
3078 int i;
3080 spin_lock_bh(&addrconf_verify_lock);
3081 now = jiffies;
3082 next = now + ADDR_CHECK_FREQUENCY;
3084 del_timer(&addr_chk_timer);
3086 for (i=0; i < IN6_ADDR_HSIZE; i++) {
3088 restart:
3089 read_lock(&addrconf_hash_lock);
3090 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3091 unsigned long age;
3092 #ifdef CONFIG_IPV6_PRIVACY
3093 unsigned long regen_advance;
3094 #endif
3096 if (ifp->flags & IFA_F_PERMANENT)
3097 continue;
3099 spin_lock(&ifp->lock);
3100 age = (now - ifp->tstamp) / HZ;
3102 #ifdef CONFIG_IPV6_PRIVACY
3103 regen_advance = ifp->idev->cnf.regen_max_retry *
3104 ifp->idev->cnf.dad_transmits *
3105 ifp->idev->nd_parms->retrans_time / HZ;
3106 #endif
3108 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3109 age >= ifp->valid_lft) {
3110 spin_unlock(&ifp->lock);
3111 in6_ifa_hold(ifp);
3112 read_unlock(&addrconf_hash_lock);
3113 ipv6_del_addr(ifp);
3114 goto restart;
3115 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3116 spin_unlock(&ifp->lock);
3117 continue;
3118 } else if (age >= ifp->prefered_lft) {
3119 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3120 int deprecate = 0;
3122 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3123 deprecate = 1;
3124 ifp->flags |= IFA_F_DEPRECATED;
3127 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3128 next = ifp->tstamp + ifp->valid_lft * HZ;
3130 spin_unlock(&ifp->lock);
3132 if (deprecate) {
3133 in6_ifa_hold(ifp);
3134 read_unlock(&addrconf_hash_lock);
3136 ipv6_ifa_notify(0, ifp);
3137 in6_ifa_put(ifp);
3138 goto restart;
3140 #ifdef CONFIG_IPV6_PRIVACY
3141 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3142 !(ifp->flags&IFA_F_TENTATIVE)) {
3143 if (age >= ifp->prefered_lft - regen_advance) {
3144 struct inet6_ifaddr *ifpub = ifp->ifpub;
3145 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3146 next = ifp->tstamp + ifp->prefered_lft * HZ;
3147 if (!ifp->regen_count && ifpub) {
3148 ifp->regen_count++;
3149 in6_ifa_hold(ifp);
3150 in6_ifa_hold(ifpub);
3151 spin_unlock(&ifp->lock);
3152 read_unlock(&addrconf_hash_lock);
3153 spin_lock(&ifpub->lock);
3154 ifpub->regen_count = 0;
3155 spin_unlock(&ifpub->lock);
3156 ipv6_create_tempaddr(ifpub, ifp);
3157 in6_ifa_put(ifpub);
3158 in6_ifa_put(ifp);
3159 goto restart;
3161 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3162 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3163 spin_unlock(&ifp->lock);
3164 #endif
3165 } else {
3166 /* ifp->prefered_lft <= ifp->valid_lft */
3167 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3168 next = ifp->tstamp + ifp->prefered_lft * HZ;
3169 spin_unlock(&ifp->lock);
3172 read_unlock(&addrconf_hash_lock);
3175 addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3176 add_timer(&addr_chk_timer);
3177 spin_unlock_bh(&addrconf_verify_lock);
3180 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3182 struct in6_addr *pfx = NULL;
3184 if (addr)
3185 pfx = nla_data(addr);
3187 if (local) {
3188 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3189 pfx = NULL;
3190 else
3191 pfx = nla_data(local);
3194 return pfx;
3197 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3198 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3199 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3200 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3203 static int
3204 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3206 struct net *net = sock_net(skb->sk);
3207 struct ifaddrmsg *ifm;
3208 struct nlattr *tb[IFA_MAX+1];
3209 struct in6_addr *pfx;
3210 int err;
3212 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3213 if (err < 0)
3214 return err;
3216 ifm = nlmsg_data(nlh);
3217 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3218 if (pfx == NULL)
3219 return -EINVAL;
3221 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3224 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3225 u32 prefered_lft, u32 valid_lft)
3227 u32 flags;
3228 clock_t expires;
3229 unsigned long timeout;
3231 if (!valid_lft || (prefered_lft > valid_lft))
3232 return -EINVAL;
3234 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3235 if (addrconf_finite_timeout(timeout)) {
3236 expires = jiffies_to_clock_t(timeout * HZ);
3237 valid_lft = timeout;
3238 flags = RTF_EXPIRES;
3239 } else {
3240 expires = 0;
3241 flags = 0;
3242 ifa_flags |= IFA_F_PERMANENT;
3245 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3246 if (addrconf_finite_timeout(timeout)) {
3247 if (timeout == 0)
3248 ifa_flags |= IFA_F_DEPRECATED;
3249 prefered_lft = timeout;
3252 spin_lock_bh(&ifp->lock);
3253 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3254 ifp->tstamp = jiffies;
3255 ifp->valid_lft = valid_lft;
3256 ifp->prefered_lft = prefered_lft;
3258 spin_unlock_bh(&ifp->lock);
3259 if (!(ifp->flags&IFA_F_TENTATIVE))
3260 ipv6_ifa_notify(0, ifp);
3262 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3263 expires, flags);
3264 addrconf_verify(0);
3266 return 0;
3269 static int
3270 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3272 struct net *net = sock_net(skb->sk);
3273 struct ifaddrmsg *ifm;
3274 struct nlattr *tb[IFA_MAX+1];
3275 struct in6_addr *pfx;
3276 struct inet6_ifaddr *ifa;
3277 struct net_device *dev;
3278 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3279 u8 ifa_flags;
3280 int err;
3282 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3283 if (err < 0)
3284 return err;
3286 ifm = nlmsg_data(nlh);
3287 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3288 if (pfx == NULL)
3289 return -EINVAL;
3291 if (tb[IFA_CACHEINFO]) {
3292 struct ifa_cacheinfo *ci;
3294 ci = nla_data(tb[IFA_CACHEINFO]);
3295 valid_lft = ci->ifa_valid;
3296 preferred_lft = ci->ifa_prefered;
3297 } else {
3298 preferred_lft = INFINITY_LIFE_TIME;
3299 valid_lft = INFINITY_LIFE_TIME;
3302 dev = __dev_get_by_index(net, ifm->ifa_index);
3303 if (dev == NULL)
3304 return -ENODEV;
3306 /* We ignore other flags so far. */
3307 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3309 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3310 if (ifa == NULL) {
3312 * It would be best to check for !NLM_F_CREATE here but
3313 * userspace alreay relies on not having to provide this.
3315 return inet6_addr_add(net, ifm->ifa_index, pfx,
3316 ifm->ifa_prefixlen, ifa_flags,
3317 preferred_lft, valid_lft);
3320 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3321 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3322 err = -EEXIST;
3323 else
3324 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3326 in6_ifa_put(ifa);
3328 return err;
3331 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3332 u8 scope, int ifindex)
3334 struct ifaddrmsg *ifm;
3336 ifm = nlmsg_data(nlh);
3337 ifm->ifa_family = AF_INET6;
3338 ifm->ifa_prefixlen = prefixlen;
3339 ifm->ifa_flags = flags;
3340 ifm->ifa_scope = scope;
3341 ifm->ifa_index = ifindex;
3344 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3345 unsigned long tstamp, u32 preferred, u32 valid)
3347 struct ifa_cacheinfo ci;
3349 ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3350 + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3351 ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3352 + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3353 ci.ifa_prefered = preferred;
3354 ci.ifa_valid = valid;
3356 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3359 static inline int rt_scope(int ifa_scope)
3361 if (ifa_scope & IFA_HOST)
3362 return RT_SCOPE_HOST;
3363 else if (ifa_scope & IFA_LINK)
3364 return RT_SCOPE_LINK;
3365 else if (ifa_scope & IFA_SITE)
3366 return RT_SCOPE_SITE;
3367 else
3368 return RT_SCOPE_UNIVERSE;
3371 static inline int inet6_ifaddr_msgsize(void)
3373 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3374 + nla_total_size(16) /* IFA_ADDRESS */
3375 + nla_total_size(sizeof(struct ifa_cacheinfo));
3378 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3379 u32 pid, u32 seq, int event, unsigned int flags)
3381 struct nlmsghdr *nlh;
3382 u32 preferred, valid;
3384 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3385 if (nlh == NULL)
3386 return -EMSGSIZE;
3388 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3389 ifa->idev->dev->ifindex);
3391 if (!(ifa->flags&IFA_F_PERMANENT)) {
3392 preferred = ifa->prefered_lft;
3393 valid = ifa->valid_lft;
3394 if (preferred != INFINITY_LIFE_TIME) {
3395 long tval = (jiffies - ifa->tstamp)/HZ;
3396 preferred -= tval;
3397 if (valid != INFINITY_LIFE_TIME)
3398 valid -= tval;
3400 } else {
3401 preferred = INFINITY_LIFE_TIME;
3402 valid = INFINITY_LIFE_TIME;
3405 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3406 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3407 nlmsg_cancel(skb, nlh);
3408 return -EMSGSIZE;
3411 return nlmsg_end(skb, nlh);
3414 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3415 u32 pid, u32 seq, int event, u16 flags)
3417 struct nlmsghdr *nlh;
3418 u8 scope = RT_SCOPE_UNIVERSE;
3419 int ifindex = ifmca->idev->dev->ifindex;
3421 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3422 scope = RT_SCOPE_SITE;
3424 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3425 if (nlh == NULL)
3426 return -EMSGSIZE;
3428 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3429 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3430 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3431 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3432 nlmsg_cancel(skb, nlh);
3433 return -EMSGSIZE;
3436 return nlmsg_end(skb, nlh);
3439 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3440 u32 pid, u32 seq, int event, unsigned int flags)
3442 struct nlmsghdr *nlh;
3443 u8 scope = RT_SCOPE_UNIVERSE;
3444 int ifindex = ifaca->aca_idev->dev->ifindex;
3446 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3447 scope = RT_SCOPE_SITE;
3449 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3450 if (nlh == NULL)
3451 return -EMSGSIZE;
3453 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3454 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3455 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3456 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3457 nlmsg_cancel(skb, nlh);
3458 return -EMSGSIZE;
3461 return nlmsg_end(skb, nlh);
3464 enum addr_type_t
3466 UNICAST_ADDR,
3467 MULTICAST_ADDR,
3468 ANYCAST_ADDR,
3471 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3472 enum addr_type_t type)
3474 int idx, ip_idx;
3475 int s_idx, s_ip_idx;
3476 int err = 1;
3477 struct net_device *dev;
3478 struct inet6_dev *idev = NULL;
3479 struct inet6_ifaddr *ifa;
3480 struct ifmcaddr6 *ifmca;
3481 struct ifacaddr6 *ifaca;
3482 struct net *net = sock_net(skb->sk);
3484 s_idx = cb->args[0];
3485 s_ip_idx = ip_idx = cb->args[1];
3487 idx = 0;
3488 for_each_netdev(net, dev) {
3489 if (idx < s_idx)
3490 goto cont;
3491 if (idx > s_idx)
3492 s_ip_idx = 0;
3493 ip_idx = 0;
3494 if ((idev = in6_dev_get(dev)) == NULL)
3495 goto cont;
3496 read_lock_bh(&idev->lock);
3497 switch (type) {
3498 case UNICAST_ADDR:
3499 /* unicast address incl. temp addr */
3500 for (ifa = idev->addr_list; ifa;
3501 ifa = ifa->if_next, ip_idx++) {
3502 if (ip_idx < s_ip_idx)
3503 continue;
3504 err = inet6_fill_ifaddr(skb, ifa,
3505 NETLINK_CB(cb->skb).pid,
3506 cb->nlh->nlmsg_seq,
3507 RTM_NEWADDR,
3508 NLM_F_MULTI);
3510 break;
3511 case MULTICAST_ADDR:
3512 /* multicast address */
3513 for (ifmca = idev->mc_list; ifmca;
3514 ifmca = ifmca->next, ip_idx++) {
3515 if (ip_idx < s_ip_idx)
3516 continue;
3517 err = inet6_fill_ifmcaddr(skb, ifmca,
3518 NETLINK_CB(cb->skb).pid,
3519 cb->nlh->nlmsg_seq,
3520 RTM_GETMULTICAST,
3521 NLM_F_MULTI);
3523 break;
3524 case ANYCAST_ADDR:
3525 /* anycast address */
3526 for (ifaca = idev->ac_list; ifaca;
3527 ifaca = ifaca->aca_next, ip_idx++) {
3528 if (ip_idx < s_ip_idx)
3529 continue;
3530 err = inet6_fill_ifacaddr(skb, ifaca,
3531 NETLINK_CB(cb->skb).pid,
3532 cb->nlh->nlmsg_seq,
3533 RTM_GETANYCAST,
3534 NLM_F_MULTI);
3536 break;
3537 default:
3538 break;
3540 read_unlock_bh(&idev->lock);
3541 in6_dev_put(idev);
3543 if (err <= 0)
3544 break;
3545 cont:
3546 idx++;
3548 cb->args[0] = idx;
3549 cb->args[1] = ip_idx;
3550 return skb->len;
3553 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3555 enum addr_type_t type = UNICAST_ADDR;
3557 return inet6_dump_addr(skb, cb, type);
3560 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3562 enum addr_type_t type = MULTICAST_ADDR;
3564 return inet6_dump_addr(skb, cb, type);
3568 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3570 enum addr_type_t type = ANYCAST_ADDR;
3572 return inet6_dump_addr(skb, cb, type);
3575 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3576 void *arg)
3578 struct net *net = sock_net(in_skb->sk);
3579 struct ifaddrmsg *ifm;
3580 struct nlattr *tb[IFA_MAX+1];
3581 struct in6_addr *addr = NULL;
3582 struct net_device *dev = NULL;
3583 struct inet6_ifaddr *ifa;
3584 struct sk_buff *skb;
3585 int err;
3587 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3588 if (err < 0)
3589 goto errout;
3591 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3592 if (addr == NULL) {
3593 err = -EINVAL;
3594 goto errout;
3597 ifm = nlmsg_data(nlh);
3598 if (ifm->ifa_index)
3599 dev = __dev_get_by_index(net, ifm->ifa_index);
3601 if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3602 err = -EADDRNOTAVAIL;
3603 goto errout;
3606 if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3607 err = -ENOBUFS;
3608 goto errout_ifa;
3611 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3612 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3613 if (err < 0) {
3614 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3615 WARN_ON(err == -EMSGSIZE);
3616 kfree_skb(skb);
3617 goto errout_ifa;
3619 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3620 errout_ifa:
3621 in6_ifa_put(ifa);
3622 errout:
3623 return err;
3626 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3628 struct sk_buff *skb;
3629 struct net *net = dev_net(ifa->idev->dev);
3630 int err = -ENOBUFS;
3632 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3633 if (skb == NULL)
3634 goto errout;
3636 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3637 if (err < 0) {
3638 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3639 WARN_ON(err == -EMSGSIZE);
3640 kfree_skb(skb);
3641 goto errout;
3643 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3644 errout:
3645 if (err < 0)
3646 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3649 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3650 __s32 *array, int bytes)
3652 BUG_ON(bytes < (DEVCONF_MAX * 4));
3654 memset(array, 0, bytes);
3655 array[DEVCONF_FORWARDING] = cnf->forwarding;
3656 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3657 array[DEVCONF_MTU6] = cnf->mtu6;
3658 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3659 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3660 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3661 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3662 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3663 array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3664 array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3665 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3666 #ifdef CONFIG_IPV6_PRIVACY
3667 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3668 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3669 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3670 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3671 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3672 #endif
3673 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3674 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3675 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3676 #ifdef CONFIG_IPV6_ROUTER_PREF
3677 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3678 array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3679 #ifdef CONFIG_IPV6_ROUTE_INFO
3680 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3681 #endif
3682 #endif
3683 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3684 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3685 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3686 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3687 #endif
3688 #ifdef CONFIG_IPV6_MROUTE
3689 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3690 #endif
3691 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3692 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3695 static inline size_t inet6_if_nlmsg_size(void)
3697 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3698 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3699 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3700 + nla_total_size(4) /* IFLA_MTU */
3701 + nla_total_size(4) /* IFLA_LINK */
3702 + nla_total_size( /* IFLA_PROTINFO */
3703 nla_total_size(4) /* IFLA_INET6_FLAGS */
3704 + nla_total_size(sizeof(struct ifla_cacheinfo))
3705 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3706 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3707 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3711 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3712 int bytes)
3714 int i;
3715 int pad = bytes - sizeof(u64) * items;
3716 BUG_ON(pad < 0);
3718 /* Use put_unaligned() because stats may not be aligned for u64. */
3719 put_unaligned(items, &stats[0]);
3720 for (i = 1; i < items; i++)
3721 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3723 memset(&stats[items], 0, pad);
3726 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3727 int bytes)
3729 switch(attrtype) {
3730 case IFLA_INET6_STATS:
3731 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3732 break;
3733 case IFLA_INET6_ICMP6STATS:
3734 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3735 break;
3739 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3740 u32 pid, u32 seq, int event, unsigned int flags)
3742 struct net_device *dev = idev->dev;
3743 struct nlattr *nla;
3744 struct ifinfomsg *hdr;
3745 struct nlmsghdr *nlh;
3746 void *protoinfo;
3747 struct ifla_cacheinfo ci;
3749 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3750 if (nlh == NULL)
3751 return -EMSGSIZE;
3753 hdr = nlmsg_data(nlh);
3754 hdr->ifi_family = AF_INET6;
3755 hdr->__ifi_pad = 0;
3756 hdr->ifi_type = dev->type;
3757 hdr->ifi_index = dev->ifindex;
3758 hdr->ifi_flags = dev_get_flags(dev);
3759 hdr->ifi_change = 0;
3761 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3763 if (dev->addr_len)
3764 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3766 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3767 if (dev->ifindex != dev->iflink)
3768 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3770 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3771 if (protoinfo == NULL)
3772 goto nla_put_failure;
3774 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3776 ci.max_reasm_len = IPV6_MAXPLEN;
3777 ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3778 + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3779 ci.reachable_time = idev->nd_parms->reachable_time;
3780 ci.retrans_time = idev->nd_parms->retrans_time;
3781 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3783 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3784 if (nla == NULL)
3785 goto nla_put_failure;
3786 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3788 /* XXX - MC not implemented */
3790 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3791 if (nla == NULL)
3792 goto nla_put_failure;
3793 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3795 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3796 if (nla == NULL)
3797 goto nla_put_failure;
3798 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3800 nla_nest_end(skb, protoinfo);
3801 return nlmsg_end(skb, nlh);
3803 nla_put_failure:
3804 nlmsg_cancel(skb, nlh);
3805 return -EMSGSIZE;
3808 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3810 struct net *net = sock_net(skb->sk);
3811 int idx, err;
3812 int s_idx = cb->args[0];
3813 struct net_device *dev;
3814 struct inet6_dev *idev;
3816 read_lock(&dev_base_lock);
3817 idx = 0;
3818 for_each_netdev(net, dev) {
3819 if (idx < s_idx)
3820 goto cont;
3821 if ((idev = in6_dev_get(dev)) == NULL)
3822 goto cont;
3823 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3824 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3825 in6_dev_put(idev);
3826 if (err <= 0)
3827 break;
3828 cont:
3829 idx++;
3831 read_unlock(&dev_base_lock);
3832 cb->args[0] = idx;
3834 return skb->len;
3837 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3839 struct sk_buff *skb;
3840 struct net *net = dev_net(idev->dev);
3841 int err = -ENOBUFS;
3843 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3844 if (skb == NULL)
3845 goto errout;
3847 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3848 if (err < 0) {
3849 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3850 WARN_ON(err == -EMSGSIZE);
3851 kfree_skb(skb);
3852 goto errout;
3854 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3855 errout:
3856 if (err < 0)
3857 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3860 static inline size_t inet6_prefix_nlmsg_size(void)
3862 return NLMSG_ALIGN(sizeof(struct prefixmsg))
3863 + nla_total_size(sizeof(struct in6_addr))
3864 + nla_total_size(sizeof(struct prefix_cacheinfo));
3867 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3868 struct prefix_info *pinfo, u32 pid, u32 seq,
3869 int event, unsigned int flags)
3871 struct prefixmsg *pmsg;
3872 struct nlmsghdr *nlh;
3873 struct prefix_cacheinfo ci;
3875 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3876 if (nlh == NULL)
3877 return -EMSGSIZE;
3879 pmsg = nlmsg_data(nlh);
3880 pmsg->prefix_family = AF_INET6;
3881 pmsg->prefix_pad1 = 0;
3882 pmsg->prefix_pad2 = 0;
3883 pmsg->prefix_ifindex = idev->dev->ifindex;
3884 pmsg->prefix_len = pinfo->prefix_len;
3885 pmsg->prefix_type = pinfo->type;
3886 pmsg->prefix_pad3 = 0;
3887 pmsg->prefix_flags = 0;
3888 if (pinfo->onlink)
3889 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3890 if (pinfo->autoconf)
3891 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3893 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3895 ci.preferred_time = ntohl(pinfo->prefered);
3896 ci.valid_time = ntohl(pinfo->valid);
3897 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3899 return nlmsg_end(skb, nlh);
3901 nla_put_failure:
3902 nlmsg_cancel(skb, nlh);
3903 return -EMSGSIZE;
3906 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3907 struct prefix_info *pinfo)
3909 struct sk_buff *skb;
3910 struct net *net = dev_net(idev->dev);
3911 int err = -ENOBUFS;
3913 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3914 if (skb == NULL)
3915 goto errout;
3917 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3918 if (err < 0) {
3919 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3920 WARN_ON(err == -EMSGSIZE);
3921 kfree_skb(skb);
3922 goto errout;
3924 err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3925 errout:
3926 if (err < 0)
3927 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3930 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3932 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3934 switch (event) {
3935 case RTM_NEWADDR:
3937 * If the address was optimistic
3938 * we inserted the route at the start of
3939 * our DAD process, so we don't need
3940 * to do it again
3942 if (!(ifp->rt->rt6i_node))
3943 ip6_ins_rt(ifp->rt);
3944 if (ifp->idev->cnf.forwarding)
3945 addrconf_join_anycast(ifp);
3946 break;
3947 case RTM_DELADDR:
3948 if (ifp->idev->cnf.forwarding)
3949 addrconf_leave_anycast(ifp);
3950 addrconf_leave_solict(ifp->idev, &ifp->addr);
3951 dst_hold(&ifp->rt->u.dst);
3952 if (ip6_del_rt(ifp->rt))
3953 dst_free(&ifp->rt->u.dst);
3954 break;
3958 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3960 rcu_read_lock_bh();
3961 if (likely(ifp->idev->dead == 0))
3962 __ipv6_ifa_notify(event, ifp);
3963 rcu_read_unlock_bh();
3966 #ifdef CONFIG_SYSCTL
3968 static
3969 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3970 void __user *buffer, size_t *lenp, loff_t *ppos)
3972 int *valp = ctl->data;
3973 int val = *valp;
3974 int ret;
3976 ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3978 if (write)
3979 addrconf_fixup_forwarding(ctl, valp, val);
3980 return ret;
3983 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3984 int __user *name, int nlen,
3985 void __user *oldval,
3986 size_t __user *oldlenp,
3987 void __user *newval, size_t newlen)
3989 int *valp = table->data;
3990 int val = *valp;
3991 int new;
3993 if (!newval || !newlen)
3994 return 0;
3995 if (newlen != sizeof(int))
3996 return -EINVAL;
3997 if (get_user(new, (int __user *)newval))
3998 return -EFAULT;
3999 if (new == *valp)
4000 return 0;
4001 if (oldval && oldlenp) {
4002 size_t len;
4003 if (get_user(len, oldlenp))
4004 return -EFAULT;
4005 if (len) {
4006 if (len > table->maxlen)
4007 len = table->maxlen;
4008 if (copy_to_user(oldval, valp, len))
4009 return -EFAULT;
4010 if (put_user(len, oldlenp))
4011 return -EFAULT;
4015 *valp = new;
4016 addrconf_fixup_forwarding(table, valp, val);
4017 return 1;
4020 static struct addrconf_sysctl_table
4022 struct ctl_table_header *sysctl_header;
4023 ctl_table addrconf_vars[DEVCONF_MAX+1];
4024 char *dev_name;
4025 } addrconf_sysctl __read_mostly = {
4026 .sysctl_header = NULL,
4027 .addrconf_vars = {
4029 .ctl_name = NET_IPV6_FORWARDING,
4030 .procname = "forwarding",
4031 .data = &ipv6_devconf.forwarding,
4032 .maxlen = sizeof(int),
4033 .mode = 0644,
4034 .proc_handler = &addrconf_sysctl_forward,
4035 .strategy = &addrconf_sysctl_forward_strategy,
4038 .ctl_name = NET_IPV6_HOP_LIMIT,
4039 .procname = "hop_limit",
4040 .data = &ipv6_devconf.hop_limit,
4041 .maxlen = sizeof(int),
4042 .mode = 0644,
4043 .proc_handler = proc_dointvec,
4046 .ctl_name = NET_IPV6_MTU,
4047 .procname = "mtu",
4048 .data = &ipv6_devconf.mtu6,
4049 .maxlen = sizeof(int),
4050 .mode = 0644,
4051 .proc_handler = &proc_dointvec,
4054 .ctl_name = NET_IPV6_ACCEPT_RA,
4055 .procname = "accept_ra",
4056 .data = &ipv6_devconf.accept_ra,
4057 .maxlen = sizeof(int),
4058 .mode = 0644,
4059 .proc_handler = &proc_dointvec,
4062 .ctl_name = NET_IPV6_ACCEPT_REDIRECTS,
4063 .procname = "accept_redirects",
4064 .data = &ipv6_devconf.accept_redirects,
4065 .maxlen = sizeof(int),
4066 .mode = 0644,
4067 .proc_handler = &proc_dointvec,
4070 .ctl_name = NET_IPV6_AUTOCONF,
4071 .procname = "autoconf",
4072 .data = &ipv6_devconf.autoconf,
4073 .maxlen = sizeof(int),
4074 .mode = 0644,
4075 .proc_handler = &proc_dointvec,
4078 .ctl_name = NET_IPV6_DAD_TRANSMITS,
4079 .procname = "dad_transmits",
4080 .data = &ipv6_devconf.dad_transmits,
4081 .maxlen = sizeof(int),
4082 .mode = 0644,
4083 .proc_handler = &proc_dointvec,
4086 .ctl_name = NET_IPV6_RTR_SOLICITS,
4087 .procname = "router_solicitations",
4088 .data = &ipv6_devconf.rtr_solicits,
4089 .maxlen = sizeof(int),
4090 .mode = 0644,
4091 .proc_handler = &proc_dointvec,
4094 .ctl_name = NET_IPV6_RTR_SOLICIT_INTERVAL,
4095 .procname = "router_solicitation_interval",
4096 .data = &ipv6_devconf.rtr_solicit_interval,
4097 .maxlen = sizeof(int),
4098 .mode = 0644,
4099 .proc_handler = &proc_dointvec_jiffies,
4100 .strategy = &sysctl_jiffies,
4103 .ctl_name = NET_IPV6_RTR_SOLICIT_DELAY,
4104 .procname = "router_solicitation_delay",
4105 .data = &ipv6_devconf.rtr_solicit_delay,
4106 .maxlen = sizeof(int),
4107 .mode = 0644,
4108 .proc_handler = &proc_dointvec_jiffies,
4109 .strategy = &sysctl_jiffies,
4112 .ctl_name = NET_IPV6_FORCE_MLD_VERSION,
4113 .procname = "force_mld_version",
4114 .data = &ipv6_devconf.force_mld_version,
4115 .maxlen = sizeof(int),
4116 .mode = 0644,
4117 .proc_handler = &proc_dointvec,
4119 #ifdef CONFIG_IPV6_PRIVACY
4121 .ctl_name = NET_IPV6_USE_TEMPADDR,
4122 .procname = "use_tempaddr",
4123 .data = &ipv6_devconf.use_tempaddr,
4124 .maxlen = sizeof(int),
4125 .mode = 0644,
4126 .proc_handler = &proc_dointvec,
4129 .ctl_name = NET_IPV6_TEMP_VALID_LFT,
4130 .procname = "temp_valid_lft",
4131 .data = &ipv6_devconf.temp_valid_lft,
4132 .maxlen = sizeof(int),
4133 .mode = 0644,
4134 .proc_handler = &proc_dointvec,
4137 .ctl_name = NET_IPV6_TEMP_PREFERED_LFT,
4138 .procname = "temp_prefered_lft",
4139 .data = &ipv6_devconf.temp_prefered_lft,
4140 .maxlen = sizeof(int),
4141 .mode = 0644,
4142 .proc_handler = &proc_dointvec,
4145 .ctl_name = NET_IPV6_REGEN_MAX_RETRY,
4146 .procname = "regen_max_retry",
4147 .data = &ipv6_devconf.regen_max_retry,
4148 .maxlen = sizeof(int),
4149 .mode = 0644,
4150 .proc_handler = &proc_dointvec,
4153 .ctl_name = NET_IPV6_MAX_DESYNC_FACTOR,
4154 .procname = "max_desync_factor",
4155 .data = &ipv6_devconf.max_desync_factor,
4156 .maxlen = sizeof(int),
4157 .mode = 0644,
4158 .proc_handler = &proc_dointvec,
4160 #endif
4162 .ctl_name = NET_IPV6_MAX_ADDRESSES,
4163 .procname = "max_addresses",
4164 .data = &ipv6_devconf.max_addresses,
4165 .maxlen = sizeof(int),
4166 .mode = 0644,
4167 .proc_handler = &proc_dointvec,
4170 .ctl_name = NET_IPV6_ACCEPT_RA_DEFRTR,
4171 .procname = "accept_ra_defrtr",
4172 .data = &ipv6_devconf.accept_ra_defrtr,
4173 .maxlen = sizeof(int),
4174 .mode = 0644,
4175 .proc_handler = &proc_dointvec,
4178 .ctl_name = NET_IPV6_ACCEPT_RA_PINFO,
4179 .procname = "accept_ra_pinfo",
4180 .data = &ipv6_devconf.accept_ra_pinfo,
4181 .maxlen = sizeof(int),
4182 .mode = 0644,
4183 .proc_handler = &proc_dointvec,
4185 #ifdef CONFIG_IPV6_ROUTER_PREF
4187 .ctl_name = NET_IPV6_ACCEPT_RA_RTR_PREF,
4188 .procname = "accept_ra_rtr_pref",
4189 .data = &ipv6_devconf.accept_ra_rtr_pref,
4190 .maxlen = sizeof(int),
4191 .mode = 0644,
4192 .proc_handler = &proc_dointvec,
4195 .ctl_name = NET_IPV6_RTR_PROBE_INTERVAL,
4196 .procname = "router_probe_interval",
4197 .data = &ipv6_devconf.rtr_probe_interval,
4198 .maxlen = sizeof(int),
4199 .mode = 0644,
4200 .proc_handler = &proc_dointvec_jiffies,
4201 .strategy = &sysctl_jiffies,
4203 #ifdef CONFIG_IPV6_ROUTE_INFO
4205 .ctl_name = NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4206 .procname = "accept_ra_rt_info_max_plen",
4207 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4208 .maxlen = sizeof(int),
4209 .mode = 0644,
4210 .proc_handler = &proc_dointvec,
4212 #endif
4213 #endif
4215 .ctl_name = NET_IPV6_PROXY_NDP,
4216 .procname = "proxy_ndp",
4217 .data = &ipv6_devconf.proxy_ndp,
4218 .maxlen = sizeof(int),
4219 .mode = 0644,
4220 .proc_handler = &proc_dointvec,
4223 .ctl_name = NET_IPV6_ACCEPT_SOURCE_ROUTE,
4224 .procname = "accept_source_route",
4225 .data = &ipv6_devconf.accept_source_route,
4226 .maxlen = sizeof(int),
4227 .mode = 0644,
4228 .proc_handler = &proc_dointvec,
4230 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4232 .ctl_name = CTL_UNNUMBERED,
4233 .procname = "optimistic_dad",
4234 .data = &ipv6_devconf.optimistic_dad,
4235 .maxlen = sizeof(int),
4236 .mode = 0644,
4237 .proc_handler = &proc_dointvec,
4240 #endif
4241 #ifdef CONFIG_IPV6_MROUTE
4243 .ctl_name = CTL_UNNUMBERED,
4244 .procname = "mc_forwarding",
4245 .data = &ipv6_devconf.mc_forwarding,
4246 .maxlen = sizeof(int),
4247 .mode = 0644,
4248 .proc_handler = &proc_dointvec,
4250 #endif
4252 .ctl_name = CTL_UNNUMBERED,
4253 .procname = "disable_ipv6",
4254 .data = &ipv6_devconf.disable_ipv6,
4255 .maxlen = sizeof(int),
4256 .mode = 0644,
4257 .proc_handler = &proc_dointvec,
4260 .ctl_name = CTL_UNNUMBERED,
4261 .procname = "accept_dad",
4262 .data = &ipv6_devconf.accept_dad,
4263 .maxlen = sizeof(int),
4264 .mode = 0644,
4265 .proc_handler = &proc_dointvec,
4268 .ctl_name = 0, /* sentinel */
4273 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4274 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4276 int i;
4277 struct addrconf_sysctl_table *t;
4279 #define ADDRCONF_CTL_PATH_DEV 3
4281 struct ctl_path addrconf_ctl_path[] = {
4282 { .procname = "net", .ctl_name = CTL_NET, },
4283 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4284 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4285 { /* to be set */ },
4286 { },
4290 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4291 if (t == NULL)
4292 goto out;
4294 for (i=0; t->addrconf_vars[i].data; i++) {
4295 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4296 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4297 t->addrconf_vars[i].extra2 = net;
4301 * Make a copy of dev_name, because '.procname' is regarded as const
4302 * by sysctl and we wouldn't want anyone to change it under our feet
4303 * (see SIOCSIFNAME).
4305 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4306 if (!t->dev_name)
4307 goto free;
4309 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4310 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4312 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4313 t->addrconf_vars);
4314 if (t->sysctl_header == NULL)
4315 goto free_procname;
4317 p->sysctl = t;
4318 return 0;
4320 free_procname:
4321 kfree(t->dev_name);
4322 free:
4323 kfree(t);
4324 out:
4325 return -ENOBUFS;
4328 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4330 struct addrconf_sysctl_table *t;
4332 if (p->sysctl == NULL)
4333 return;
4335 t = p->sysctl;
4336 p->sysctl = NULL;
4337 unregister_sysctl_table(t->sysctl_header);
4338 kfree(t->dev_name);
4339 kfree(t);
4342 static void addrconf_sysctl_register(struct inet6_dev *idev)
4344 neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4345 NET_IPV6_NEIGH, "ipv6",
4346 &ndisc_ifinfo_sysctl_change,
4347 ndisc_ifinfo_sysctl_strategy);
4348 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4349 idev->dev->ifindex, idev, &idev->cnf);
4352 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4354 __addrconf_sysctl_unregister(&idev->cnf);
4355 neigh_sysctl_unregister(idev->nd_parms);
4359 #endif
4361 static int addrconf_init_net(struct net *net)
4363 int err;
4364 struct ipv6_devconf *all, *dflt;
4366 err = -ENOMEM;
4367 all = &ipv6_devconf;
4368 dflt = &ipv6_devconf_dflt;
4370 if (net != &init_net) {
4371 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4372 if (all == NULL)
4373 goto err_alloc_all;
4375 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4376 if (dflt == NULL)
4377 goto err_alloc_dflt;
4380 net->ipv6.devconf_all = all;
4381 net->ipv6.devconf_dflt = dflt;
4383 #ifdef CONFIG_SYSCTL
4384 err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4385 NULL, all);
4386 if (err < 0)
4387 goto err_reg_all;
4389 err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4390 NULL, dflt);
4391 if (err < 0)
4392 goto err_reg_dflt;
4393 #endif
4394 return 0;
4396 #ifdef CONFIG_SYSCTL
4397 err_reg_dflt:
4398 __addrconf_sysctl_unregister(all);
4399 err_reg_all:
4400 kfree(dflt);
4401 #endif
4402 err_alloc_dflt:
4403 kfree(all);
4404 err_alloc_all:
4405 return err;
4408 static void addrconf_exit_net(struct net *net)
4410 #ifdef CONFIG_SYSCTL
4411 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4412 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4413 #endif
4414 if (net != &init_net) {
4415 kfree(net->ipv6.devconf_dflt);
4416 kfree(net->ipv6.devconf_all);
4420 static struct pernet_operations addrconf_ops = {
4421 .init = addrconf_init_net,
4422 .exit = addrconf_exit_net,
4426 * Device notifier
4429 int register_inet6addr_notifier(struct notifier_block *nb)
4431 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4434 EXPORT_SYMBOL(register_inet6addr_notifier);
4436 int unregister_inet6addr_notifier(struct notifier_block *nb)
4438 return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4441 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4443 static void addrconf_net_exit(struct net *net)
4445 struct net_device *dev;
4447 rtnl_lock();
4448 /* clean dev list */
4449 for_each_netdev(net, dev) {
4450 if (__in6_dev_get(dev) == NULL)
4451 continue;
4452 addrconf_ifdown(dev, 1);
4454 addrconf_ifdown(net->loopback_dev, 2);
4455 rtnl_unlock();
4458 static struct pernet_operations addrconf_net_ops = {
4459 .exit = addrconf_net_exit,
4463 * Init / cleanup code
4466 int __init addrconf_init(void)
4468 int err;
4470 if ((err = ipv6_addr_label_init()) < 0) {
4471 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4472 err);
4473 return err;
4476 register_pernet_subsys(&addrconf_ops);
4478 /* The addrconf netdev notifier requires that loopback_dev
4479 * has it's ipv6 private information allocated and setup
4480 * before it can bring up and give link-local addresses
4481 * to other devices which are up.
4483 * Unfortunately, loopback_dev is not necessarily the first
4484 * entry in the global dev_base list of net devices. In fact,
4485 * it is likely to be the very last entry on that list.
4486 * So this causes the notifier registry below to try and
4487 * give link-local addresses to all devices besides loopback_dev
4488 * first, then loopback_dev, which cases all the non-loopback_dev
4489 * devices to fail to get a link-local address.
4491 * So, as a temporary fix, allocate the ipv6 structure for
4492 * loopback_dev first by hand.
4493 * Longer term, all of the dependencies ipv6 has upon the loopback
4494 * device and it being up should be removed.
4496 rtnl_lock();
4497 if (!ipv6_add_dev(init_net.loopback_dev))
4498 err = -ENOMEM;
4499 rtnl_unlock();
4500 if (err)
4501 goto errlo;
4503 err = register_pernet_device(&addrconf_net_ops);
4504 if (err)
4505 return err;
4507 register_netdevice_notifier(&ipv6_dev_notf);
4509 addrconf_verify(0);
4511 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4512 if (err < 0)
4513 goto errout;
4515 /* Only the first call to __rtnl_register can fail */
4516 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4517 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4518 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4519 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4520 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4522 ipv6_addr_label_rtnl_register();
4524 return 0;
4525 errout:
4526 unregister_netdevice_notifier(&ipv6_dev_notf);
4527 errlo:
4528 unregister_pernet_subsys(&addrconf_ops);
4530 return err;
4533 void addrconf_cleanup(void)
4535 struct inet6_ifaddr *ifa;
4536 int i;
4538 unregister_netdevice_notifier(&ipv6_dev_notf);
4539 unregister_pernet_device(&addrconf_net_ops);
4541 unregister_pernet_subsys(&addrconf_ops);
4543 rtnl_lock();
4546 * Check hash table.
4548 write_lock_bh(&addrconf_hash_lock);
4549 for (i=0; i < IN6_ADDR_HSIZE; i++) {
4550 for (ifa=inet6_addr_lst[i]; ifa; ) {
4551 struct inet6_ifaddr *bifa;
4553 bifa = ifa;
4554 ifa = ifa->lst_next;
4555 printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4556 /* Do not free it; something is wrong.
4557 Now we can investigate it with debugger.
4561 write_unlock_bh(&addrconf_hash_lock);
4563 del_timer(&addr_chk_timer);
4564 rtnl_unlock();
4566 unregister_pernet_subsys(&addrconf_net_ops);